1                                             NCEA-C-1765
 2                                            November 2006
 3
 4
 5

 6  Toxicological Reviews of Cyanobacterial
 7                       Toxins:
        Microcystins LR, RR, YR and LA
10

11

12

13

14

15

16

17

18

19

20

21               National Center for Environmental Assessment
22                  Office of Research and Development
23                 U.S. Environmental Protection Agency
24                     Cincinnati, OH 45268
25

-------
1                                      DISCLAIMER
2
o
J
4          This report is an external draft for review purposes only and does not constitute Agency
5   policy. Mention of trade names or commercial products does not constitute endorsement or
6   recommendation for use.
                                                      DRAFT - DO NOT CITE OR QUOTE

-------
 1                                 TABLE OF CONTENTS
 2
 o
 3
 4    LIST OF TABLES	vi
 5    LIST OF FIGURES	vii
 6    LIST OF ACRONYMS	viii
 7    FOREWORD	x
 8    AUTHORS, CONTRIBUTORS, AND REVIEWERS	xi
 9
10    1.  INTRODUCTION	1
11
12    2.  CHEMICAL AND PHYSICAL INFORMATION	3
13
14    3.  TOXICOKINETICS	5
15       3.1.  ABSORPTION	5
16       3.2.  DISTRIBUTION	6
17            3.2.1. Organ Distribution	6
18            3.2.2. Cellular Uptake	8
19            3.2.3. Subcellular Localization of Cystolic Protein Binding	9
20       3.3.  METABOLISM	10
21       3.4.  ELIMINATION	13
22       3.5.  PHYSIOLOGICALLY-BASED TOXICOKINETIC MODELS	14
23
24    4.  HAZARD IDENTIFICATION	15
25       4.1.  STUDIES IN HUMANS - EPIDEMIOLOGY, CASE REPORTS, CLINICAL
26            CONTROLS	15
27            4.1.1. Oral Exposure	15
28                 4.1.1.1. Short-Term Studies and Case Reports	15
29                 4.1.1.2. Long-Term Studies and Epidemiological Studies	19
30            4.1.2.   Inhalation Exposure	24
31       4.2.  ACUTE, SHORT-TERM, LONG-TERM AND CHRONIC STUDIES AND
32            CANCER BIOASSAYS IN ANIMALS - ORAL AND INHALATION	24
33            4.2.1. Oral Exposure	25
34                 4.2.1.1. Acute Studies	25
35                 4.2.1.2. Short-Term Studies	28
36                 4.2.1.3. Subchronic Studies	31
37                 4.2.1.4. Chronic Studies	34
38                 4.2.1.5. Initiation/Promotion Studies - Cyanobacterial Extracts	36
39            4.2.2. Inhalation Exposure	36
40                 4.2.2.1. Acute Studies	36
41                 4.2.2.2. Short-Term Studies	37
42                 4.2.2.3. Subchronic and Chronic Studies	37
                                            iii      DRAFT - DO NOT CITE OR QUOTE

-------
 1                               TABLE OF CONTENTS cont.
 2
 o
 3
 4      4.3.  REPRODUCTIVE/DEVELOPMENTAL STUDIES - ORAL AND INHALATION...39
 5           4.3.1.  Oral Exposure	39
 6                  4.3.1.1. Purified Microcystins	39
 7                  4.3.1.2. Cyanobacterial Extracts	39
 8           4.3.2.  Inhalation Exposure	39
 9      4.4.  OTHER STUDIES	40
10           4.4.1.  Neurological Effects	40
11           4.4.2.  Immunological Effects	40
12           4.4.3.  Hematological Effects	41
13           4.4.4.  Effects by Dermal Exposure	41
14           4.4.5.  Effects by Parenteral Exposure	42
15                  4.4.5.1.  Effects in Humans after Parenteral Exposure	42
16                  4.4.5.2.  Effects in Animals after Parenteral Exposure	43
17           4.4.6.  Effects by Intratracheal or Intranasal Instillation	54
18           4.4.7.  Mechanistic Studies	54
19                  4.4.7.1.  Target Organ/Cell Type Specificity	55
20                  4.4.7.2.  Characterization of Subcellular Effects in the Liver	56
21                  4.4.7.3.  Molecular Target: Inhibition of Type 1 and 2A Protein
22                          Phosphatases	58
23                  4.4.7.4.  Cytoskeletal  Effects	62
24                  4.4.7.5.  Apoptosis	63
25                  4.4.7.6.  Lipid Peroxidation	64
26                  4.4.7.7.  Prevention of Liver Toxicity and Lethality	65
27                  4.4.7.8.  Extra-Hepatic Effects of Mi crocystins	66
28           4.4.8.  Genotoxicity and Cell Proliferation	67
29           4.4.9.  Structure-Activity Relationships	69
30      4.5.  MODE OF ACTION - NONCANCER AND  CANCER	70
31           4.5.1.  Target Organ Specificity	70
32           4.5.2.  Key Events in the Mode of Action  for Liver Toxicity and Hemorrhage	71
33                  4.5.2.1.  Molecular Events	71
34                  4.5.2.2.  Cellular Effects	73
35           4.5.3.  Conclusion	73
36      4.6.  SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS	74
37           4.6.1.  Oral	75
38           4.6.2.  Inhalation	78
39      4.7.  WEIGHT-OF-EVIDENCE EVALUATION AND CANCER
40           CHARACTERIZATION	79
41           4.7.1.  Summary of Overall Weight of Evidence	79
42           4.7.2.  Synthesis of Human, Animal and Other Supporting Evidence	79
43      4.8.  SUSCEPTIBLE POPULATIONS AND LIFE STAGES	81
44
                                                iv       DRAFT - DO NOT CITE OR QUOTE

-------
 1                             TABLE OF CONTENTS cont.
 2
 o
 3
 4    5.  DOSE-RESPONSE ASSESSMENTS	82
 5       5.1.  NARRATIVE DESCRIPTION OF THE EXTENT OF THE DATABASE	82
 6       5.2.  ORAL REFERENCE DOSE	83
 7           5.2.1.  Acute Oral RfD	84
 8           5.2.2.  Short-Term Oral RfD	85
 9                 5.2.2.1. Choice of Principal Study and Critical Effect	85
10                 5.2.2.2. Methods of Analysis	85
11                 5.2.2.3. RfD Derivation	89
12           5.2.3.  Subchronic Oral RfD	90
13                 5.2.3.1. Choice of Principal Study and Critical Effect	90
14                 5.2.3.2. Methods of Analysis	91
15                 5.2.3.3. RfD Derivation	92
16           5.2.4.  Chronic Oral RfD	92
17                 5.2.4.1. Choice of Principal Study and Critical Effect	92
18                 5.2.4.2. RfD Derivation	93
19       5.3.  INHALATION REFERENCE CONCENTRATION	93
20       5.4.  CANCER ASSESSMENT	95
21
22    6.  MAJOR CONCLUSIONS IN THE CHARACTERIZATIONS OF HAZARD AND
23       DOSE RESPONSE	96
24       6.1.  HUMAN HAZARD POTENTIAL	96
25       6.2.  DOSE RESPONSE	97
26
27    7.  REFERENCES	98
28
29    APPENDIX A	A-l
                                           v       DRAFT - DO NOT CITE OR QUOTE

-------
 1                                     LIST OF TABLES
 2
 3
 4   No.                                      Title
 5
 6   4-1    Relative Risk of Colorectal Cancer and Microcystin Concentration by Drinking
 7          Water Source	20
 8
 9   4-2    Incidence of Liver Lesions in Mice and Rats Treated with a Single Dose of
10          MCLR	26
11
12   4-3    Serum Enzyme Levels and Relative Liver Weights (Mean + Standard Deviation
13          in Rats Ingesting MCLR in Drinking Water	29
14
15   4-4    Incidence of Liver Lesions in Rats Ingesting MCLR in Drinking Water for
16          28 Days	30
17
18   4-5    Blood Chemistry Results (Mean + Standard Deviation) for Mice Treated
19          with MCLR for 13 Weeks	32
20
21   4-6    Incidence of Liver Histopathology in Mice Treated with MCLR for
22          13 Weeks	33
23
24   4-7    Incidence and Severity of Nasal Cavity Lesions in Mice Inhaling Microcystin
25          Aerosol for 7 Days	38
26
27   4-8    LDso Values of Purified Microcystin Congeners by Intraperitoneal Administration	44
28
29   4-9    Intraperitoneal LD50 Values for Bloom Extracts	49
30
31   4-10   Studies Comparing Protein Phosphatase Inhibition Activity of Microcystin
32          Congeners	59
o o
JJ
34   4-11   Summary Noncancer Results in All Animal Studies of Oral Exposure to Purified
35          Microcystin-LR	76
36
37   5-1    Available Dose-Response Information for Oral Exposure to Purified MCLR	84
38
39   5-2    Incidence of Liver Lesions Used for BMD Modeling	86
40
41   5-3    BMD Modeling Results for Heinze (1999) Liver Lesion Data	87
42
43   5-4    BMD Modeling Results for Fawell et al. (1999) Chronic Liver Inflammation Data	91
44
45   5-5    BMD Modeling Results for Fawell etal. (1999) ALT Data in Male Mice	92
46
47   5-6    Summary of Reference Dose Values	94
                                                vi       DRAFT - DO NOT CITE OR QUOTE

-------
 1                                     LIST OF FIGURES
 2
 o
 J
 4    2-1    General Structure of Microcystins	3
 5
 6    3-1    Structures of GSH and Cys Conjugates of Microcystins LR, YR, andRR	12
 7
 8    4-1    Relationship between Colorectal Cancer and Microcystin Concentration in
 9           River and Pond Water in Haining City, China (Zhou et al., 2002)	22
10
11    4-2    Schematic Representation of Interactions between Microcystin-LR and the
12           Catalytic Site of Protein Phosphatase 1	61
13
14    5-1    Exposure Response Array for Oral Exposure: All Studies of Purified
15           Microcystin-LR	83
16
17    5-2    Probit Model Fit to Liver Lesion Incidence Data from Heinze (1999)	88
                                                vii      DRAFT - DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31



AFB1
ALP
ALT
AST
BMD
BMDL
BMDS
BMR
BUN
CI
DEN
ELISA
EPA
ESR
FETAX
GD
GFR
GGT
GST-P
HPLC
i.p.
i.v.
LCso
LD50
LDH
LOAEL
MAPK
MCAR
                   LIST OF ACRONYMS

AflatoxinBl
Alkaline phosphatase
Alanine aminotransferase
Aspartate aminotransferase
Benchmark dose
Benchmark dose, lower confidence limit
Benchmark dose software
Benchmark response
Blood urea nitrogen
Confidence interval
Diethylnitrosamine
Enzyme-linked immunosorbent assay
Environmental Protection Agency
Electron spin resonance
Frog Embryo Teratogenicity Assay-Xenopus
Gestation day
Glomerular filtration rate
y-Glutamyltransferase
Glutathione S-transferase (placental form)
High pressure liquid chromatography
Intraperitoneal
Intravenous
Concentration lethal to 50% of population
Dose lethal to 50% of population
Lactate dehydrogenase
Lowest-observed-adverse-effect level
Mitogen-activated protein kinase
Microcystin-AR
                              Vlll
DRAFT - DO NOT CITE OR QUOTE

-------
1
2
o
6
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
MCLA
MCLR
MCRR
MCYM
MCYR
MPT
MW
NMR
NOAEL
OATP
PAS
POD
PP1
PP2A
RfC
RfD
ROS
RVR
SDH
SRR
TDI
TEF
TUNEL
UF
VAF
WHO
Microcystin-LA
Microcystin-LR
Microcystin-RR
Mi crocy stin-YM
Microcystin-YR
Mitochondrial permeability transition
Molecular weight
Nuclear magnetic resonance
No-observed-adverse-effect level
Organic acid transport protein
Periodic acid-Schiff
Point of departure
Protein phosphatase 1
Protein phosphatase 2A
Reference concentration
Reference dose
Reactive oxygen species
Renal vascular rate
Sorbitol dehydrogenase
Standardized rate ratio
Tolerable Daily Intake
Toxicity equivalency factor
Terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick end-labeling
Uncertainty factor
Virus antibody free
World Health Organization
                               IX
DRAFT - DO NOT CITE OR QUOTE

-------
 1                                          PREFACE
 2
 3
 4          The Safe Drinking Water Act (SDWA), as amended in 1996, requires the Environmental
 5   Protection Agency (EPA) to publish a list of contaminants which, at the time of publication, are
 6   not subject to any proposed or promulgated national primary drinking water regulations, that are
 7   known or anticipated to occur in public water systems, and which may require regulations under
 8   SDWA.  This list, known as the Contaminant Candidate List (CCL), was first published in 1998
 9   and then again in 2005. The 1998 and 2005 CCLs include "cyanobacteria (blue-green algae),
10   other freshwater algae, and their toxins" as microbial contaminants.
11
12          In 2001, a meeting was held among EPA, researchers from the drinking water industry,
13   academia and government agencies with expertise in the area of fresh water algae and their
14   toxins. The goal of this meeting was to convene a panel of scientists to assist in identifying a
15   target list of algal toxins that are likely to pose a health risk in source and finished waters of the
16   drinking water utilities in the U.S. Toxin  selection was based on four criteria: health effects,
17   occurrence in the United States, susceptibility to drinking water treatment and toxin stability.
18   Microcystins were identified at this meeting as being toxins of high priority based on those
19   criteria.
20
21          The National Center for Environmental Assessment has prepared this Toxicological
22   Review of Cyanobacterial Toxins: Microcystins (LR, RR, YR and LA) as one in a series of dose-
23   response assessments to support the health assessment of unregulated contaminants on the CCL.
24    The purpose of this document is to compile and evaluate the available data regarding
25   microcystin toxicity to aid the Office of Water in regulatory decision making. It is not intended
26   to be a comprehensive treatise on the chemical or toxicological nature of microcystins.
27
28          In Section 6, Major Conclusions in the Characterization of Hazard and Dose Response,
29   EPA has characterized its overall confidence in the quantitative and qualitative aspects of the
30   hazard and dose response by addressing knowledge gaps, uncertainties, quality of data and
31   scientific controversies. The discussion is intended to convey the limitations of the assessment
32   and to aid and guide the Office of Water in the ensuing steps of the human health risk assessment
33   of microcystins.
                                                         DRAFT - DO NOT CITE OR QUOTE

-------
 1                     AUTHORS, CONTRIBUTORS AND REVIEWERS
 2
 o
 3
 4   AUTHORS
 5
 6   Belinda Hawkins, Ph.D. (Chemical Manager)
 7   National Center for Environmental Assessment
 8   Office of Research and Development
 9   U.S. Environmental Protection Agency
10   Cincinnati, OH
11
12   Heather Carlson-Lynch
13   Syracuse Research Corporation
14   Syracuse, NY
15
16   Marc Odin
17   Syracuse Research Corporation
18   Syracuse, NY
19
20   Julie Stickney
21   Syracuse Research Corporation
22   Syracuse, NY
23
24
25   REVIEWERS
26
27   INTERNAL EPA REVIEWERS
28
29   Joyce Donohue, Ph.D.
30   Office of Water
31   Washington, DC
32
33   Elizabeth Hilborn, D. V.M., M.P.H.
34   National Health and Environmental Effects Research Laboratory
35   Office of Research and Development
36   Research Triangle Park, NC
37
38   James Sinclair, Ph.D.
39   Office of Water
40   Cincinnati, OH
41
42   Jeff Swartout
43   National Center for Environmental Assessment
44   Office of Research and Development
45   Cincinnati, OH
                                               xi       DRAFT - DO NOT CITE OR QUOTE

-------
 1                                     1. INTRODUCTION
 2
 o
 J
 4          This toxicological review presents background and justification for hazard and dose-
 5   response assessments of microcystins LR, RR, YR and LA.  U.S. Environmental Protection
 6   Agency (EPA) toxicological reviews may include oral reference dose (RfD) and inhalation
 7   reference concentration (RfC) values for chronic and less-than-lifetime exposure durations and a
 8   carcinogenicity assessment.
 9
10          The RfD and RfC provide quantitative information for use in risk assessments for health
11   effects known or assumed to be produced through a nonlinear (possibly threshold) mode of
12   action. These reference values are defined as an estimate of an exposure, designated by duration
13   and route, to the human population (including susceptible subgroups), that is likely to be without
14   an appreciable risk of adverse effects. Reference values may be derived for acute (<24 hours),
15   short-term (up to 30 days), subchronic (up to  10% of average lifespan) and chronic (up to
16   lifetime) exposures, all considered to be continuous exposures throughout the duration  specified.
17    A reference value is derived from a BMDL (a statistical  lower confidence limit on the
18   benchmark dose), a no-observed-adverse effect level (NOAEL), lowest-observed-adverse-effect
19   level (LOAEL) or other suitable point of departure with uncertainty/variability  factors applied to
20   reflect limitations of the data used. The RfD is expressed in units of mg/kg-day, and the RfC in
21   units of mg/m3.
22
23          The carcinogenicity assessment provides information on the carcinogenic hazard
24   potential of the  substance in question and quantitative estimates of risk from oral exposure and
25   inhalation exposure. The information includes a weight-of-evidence judgment  of the likelihood
26   that the agent is a human carcinogen and the conditions under which the carcinogenic effects
27   may be expressed. Quantitative risk estimates are presented in three ways.  The slope factor is
28   the result of application of a low-dose extrapolation procedure and is presented as the risk per
29   mg/kg-day. The unit risk is the quantitative estimate in terms of either risk per |J,g/L drinking
30   water or risk per |J,g/m3 air breathed.  Another form in which risk is presented is a drinking water
31   or air concentration providing cancer risks of 1 in 10,000; 1 in 100,000; or 1 in  1,000,000.
32
33          Development of these hazard identification and dose-response  assessments for
34   microcystins has followed the general guidelines for risk assessment as set forth by the National
35   Research Council (1983). EPA guidelines that were used in the development of this assessment
36   include the following: Guidelines for the Health Risk Assessment of Chemical Mixtures (U.S.
37   EPA, 1986a), Guidelines for Mutagenicity Risk Assessment (U.S. EPA, 1986b), Guidelines for
38   Developmental Toxicity Risk Assessment (U.S. EPA, 1991), Guidelines for Reproductive Toxicity
39   Risk Assessment (U. S. EPA, 1996), Guidelines for Neurotoxicity Risk Assessment (U. S. EPA,
40   1998a), Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), Supplemental Guidance
41   for Assessing Susceptibility from Early-Life Exposure to Carcinogens  (U. S. EPA, 2005b),
42   Recommendations for and Documentation of Biological Values for Use in Risk Assessment (U.S.
43   EPA, 1988), (proposed) Interim Policy for Particle Size and Limit Concentration Issues in
44   Inhalation Toxicity (U.S. EPA, 1994a), Methods for Derivation of Inhalation Reference
45   Concentrations and Application of Inhalation Dosimetry (U.S. EPA, 1994b), Use of the
46   Benchmark Dose Approach in Health Risk Assessment (U.S. EPA, 1995), Science Policy Council
                                                         DRAFT - DO NOT CITE OR QUOTE

-------
 1    Handbook: Peer Review (U.S. EPA, 1998b, 2000a, 2005c), Science Policy Council Handbook:
 2    Risk Characterization (U.S. EPA, 2000b), Benchmark Dose Technical Guidance Document
 3    (U.S. EPA, 2000c), Supplemental Guidance for Conducting Health Risk Assessment of Chemical
 4    Mixtures (U.S. EPA, 2000d) and A Review of the Reference Dose and Reference Concentration
 5    Processes (U. S. EPA, 2002).
 6
 7          Microcystins are a group of at least 80 naturally occurring hepatotoxins produced by
 8    freshwater cyanobacteria (blue-green algae) including Microcystis, Anabaena, Nodularia,
 9    Nostoc and Oscillatoria (Duy et al., 2000). Microcystins were first isolated from cyanobacterial
10    extracts in the 1980s (WHO, 1999).
11
12          Much of the toxicological research on microcystins has focused on a single congener,
13    microcystin-LR (MCLR). In addition to MCLR, this report focuses on three other major
14    microcystin congeners:  microcystin-YR, microcystin-RR and microcystin-LA (abbreviated as
15    MCYR, MCRR and MCLA throughout this document). Literature searches were conducted for
16    studies relevant to the derivation of toxicity and carcinogenicity values for these four
17    microcystin congeners.  The following databases were searched: MEDLINE (PubMed),
18    TOXLINE, BIOSIS, CANCERLIT, TSCATS, CCRIS, DART/ETIC, EMIC, GENETOX, HSDB
19    andRTECS. The relevant literature was reviewed through May 2006.
                                                       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2
 3
 4
 5
 6
 7
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
                   2. CHEMICAL AND PHYSICAL INFORMATION
       Microcystins are monocyclic heptapeptide toxins produced by a number of
cyanobacterial species, including members ofMicrocystis, Anabaena, Nodularia, Nostoc and
Oscillatoria (Duy et al., 2000). At least 80 microcystin congeners have been identified. A
general structure for microcystins is shown in Figure 2-1.
                                                                     Me
                                                                        Me  OMe
                                                            H        I         II
                                                            C=C—C=C—C—C—C—Ph
                                                                 H
                                                                    H   H   H
                                                                        8
                                                                                      10
                       Figure 2-1. General Structure of Microcystins
       Microcystins are monocyclic heptapeptides consisting of D-alanine (Ala); two variable
amino acids (at positions X and Z in Figure 2-1); D-p-methylaspartic acid (MeAsp);
(2S,3S,8S,9S)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid (Adda); iso-
linked D-glutamic acid (Glu) and TV-methyl dehydroalanine (MDha). Structural variations occur
in all seven of the amino acid peptides, but most commonly in the L-amino acids at positions X
and Z in Figure 2-1 (shown as "variable" amino acids in the figure). The most common L-amino
acids at position X are leucine (L), arginine (R) and tyrosine (Y), while those at position Z are
arginine (R) and alanine (A). The congeners take their names from the L-amino acids in these
positions.  For example, the microcystin congener with leucine in the X position and arginine in
the Z position is microcystin-LR.

       Little information on the chemical and physical properties of microcystins was located.
Duy et al.  (2000) provided the most thorough (albeit general) review of the properties of
microcystins. The microcystins identified to date have molecular weights (MWs) ranging from
900 to 1200. Microcystins are nonvolatile and generally quite hydrophilic, although a few have
lipophilic properties.  Microcystins are soluble in water, ethanol and methanol, and insoluble in
                                                         DRAFT - DO NOT CITE OR QUOTE

-------
1   acetone, ether, chloroform and benzene.  Laboratory studies show microcystins to be stable
2   under changes in temperature and pH.  Microcystins are stable in sunlight; however, in the
3   presence of pigments (type unspecified) and sunlight, microcystins can be decomposed or
4   isomerized (Duy et al., 2000).  Limited information suggests that microcystins can
5   bioaccumulate in aquatic organisms; these toxins have been measured in a number offish and
6   aquatic invertebrates.
                                                       DRAFT - DO NOT CITE OR QUOTE

-------
 1                                    3. TOXICOKINETICS
 2
 3
 4          The available information on toxicokinetics of microcystins is primarily focused on
 5   exposure via injection routes, either intravenous or intraperitoneal. Few data are available on the
 6   oral and inhalation exposure routes. No data on the absorption, metabolism or elimination of
 7   microcystins after in vivo oral or inhalation exposure in humans or animals were located in the
 8   literature. Toxicokinetic data available from parenteral routes of exposure are of uncertain
 9   relevance to oral and inhalation exposure routes.  The database on microcystins does not contain
10   any toxicokinetic models for microcystins.
11
12          A number of studies on the toxicokinetics of microcystins have used 3H-dihydro-MCLR
13   (usually produced by reduction of the MDha moiety in MCLR with tritiated sodium
14   borohydride) as a test material.  While there are similarities between the organ distribution,
15   hepatocellular uptake and clinical syndrome after exposure to dihydroMCLR and MCLR
16   (Meriluoto et al.,  1990), there are differences in the binding of these compounds to molecular
17   targets. Craig et al. (1996) showed that, while dihydroMCLA was capable of inactivating
18   protein phosphatase 2Ac through a rapid binding mechanism, it did not subsequently form a
19   covalent bond with PP2Ac, while MCLA did. Further information on the structural requirements
20   for microcystin toxicity is provided in Section 4.4.9. Potential differences in the binding of
21   dihydro-microcystin analogs dictates that caution be exercised in generalizing toxicokinetic
22   information derived using these compounds to microcystins as a group.  In particular,
23   information on the subcellular localization of dihydromicrocystins may not be applicable to
24   microcystins containing an intact MDha residue.
25
26   3.1.   ABSORPTION
27
28          Pulmonary absorption of MCLR (purified from a bloom sample) was rapid following
29   intratracheal instillation in mice (Ito et al.,  2001).  Immunostaining of the lung occurred within 5
30   minutes, followed by a lag period of 60 minutes before staining was observed in the liver. Based
31   on the positive immunostaining of alveoli,  it was concluded that absorption occurred at the
32   alveoli. The lethal dose level for intratracheal injection was similar to a lethal dose for
33   intraperitoneal (i.p.) injection (i.e., both about 100 |j,g/kg). The authors reported that the lungs
34   were not affected by microcystin administration, but it is unclear whether a detailed
35   histopathological evaluation of the lungs was conducted in addition to the immunostaining.
36
37          The occurrence of hepatotoxicity and lethality following oral exposure to microcystins
38   (see Section 4.2) is evidence of oral absorption of the  toxin. However, quantitative assessments
39   of oral absorption were not located. Ito et  al. (1997a)  qualitatively studied the oral absorption
40   and distribution of MCLR (purified from a bloom  sample) following gavage dosing in mice (500
41   |j,g/kg). Immunostaining techniques indicated that MCLR was absorbed primarily in the small
42   intestine, although some absorption did occur in the stomach. Erosion was observed in the
43   surface epithelial cells and in the lamina propria of the small intestine villi. Erosion may
44   facilitate uptake of the toxin into the bloodstream.
45
                                                          DRAFT - DO NOT CITE OR QUOTE

-------
 1          The oral bioavailability of MCLR was indirectly studied in in situ experiments using
 2   isolated intestinal loops of rats (Dahlem et al., 1989).  Rats given an infusion of MCLR (>95%
 3   pure by high pressure liquid chromatography (HPLC) with UV detection) into the ileum showed
 4   clinical signs (i.e., labored breathing and circulatory shock) and evidence of liver toxicity within
 5   6 hours of a single 5 mg/kg dose. Liver toxicity was assessed as an increase in the liver-to-body
 6   weight ratio and the presence of gross and histopathological liver lesions characteristic of
 7   microcystin toxicity (i.e., enlarged livers, hepatocyte rounding and disassociation, hemorrhage).
 8   Infusion of a similar dose into a jejunal loop produced a lower degree of liver toxicity, as
 9   compared to the ileal loop infusion.  These results suggest that there could be site-specificity in
10   intestinal absorption of MCLR; however, differences in absorptive surface area were not taken
11   into account in the experiment.  In vitro experiments reported in this publication indicated that
12   cholestyramine, a bile acid sequestrant, bound MCLR, and the presence of cholestyramine in the
13   ileal loop infusion significantly reduced MCLR liver toxicity (Dahlem et al., 1989).
14
15          Oral absorption of 3H-dihydromicrocystin was also demonstrated using ileal loop
16   exposure in swine (Stotts et al., 1997a,b). In the exposed swine, the maximum blood
17   concentration of the toxin occurred 90 minutes after dosing.
18
19   3.2.   DISTRIBUTION
20
21          The distribution of microcystins is limited due to the poor ability of these toxins to cross
22   cell membranes. Microcystins are primarily taken up into the liver by the multispecific active
23   transport system for bile acids.  Once inside the cell, these toxins bind covalently to cytosolic
24   proteins, resulting in retention in the liver.  The cytosolic proteins bound by microcystins have
25   been identified as the protein  phosphatase enzymes  (PP1 and PP2A). It should be noted that
26   dihydromicrocystin analogs do not appear to form covalent bonds with  PP1 and PP2A, although
27   they are able to rapidly bind and inactivate the enzymes (Craig et al., 1996). Binding to and
28   inhibition of these phosphatase enzymes are directly related to the mechanism of toxicity for
29   microcystins and are further discussed in Section 4.4.7.3.  This section will describe the overall
30   organ distribution, cellular uptake, subcellular localization and protein binding of microcystins.
31
32   3.2.1.  Organ Distribution
33
34          The organ distribution of a 125I-labelled heptapeptide toxin (MW 1019) isolated from
35   Microcystis aeruginosa (while not identified by the study authors as such, probably because the
36   toxin had not yet been  named, this is assumed to be a microcystin) was  investigated in female
37   rats following intravenous (i.v.) administration (Falconer et al., 1986; Runnegar et al., 1986).
38   The heptapeptide  toxin was purified by HPLC prior to reaction with 125I in the presence of Nal
39   and lactoperoxidase. Labelling of the toxin was confirmed by HPLC and mouse bioassay.  The
40   highest tissue concentrations of microcystins were detected in the liver  and kidney. After 30
41   minutes, 21.7% of the administered dose was present in the liver, 5.6% was present in the
42   kidneys, 7% remained  in the gut contents, and 0.9% was cleared in the urine (Falconer et al.,
43   1986). The balance of the administered dose was not reported; however, the authors reported
44   that no significant accumulation was observed in other organs or tissues.
45
                                                          DRAFT - DO NOT CITE OR QUOTE

-------
 1          Brooks and Codd (1987) reported extensive liver uptake following i.p. injection of 125
 2   Hg/kg of a 14C-labelled toxin extracted from M. aeruginosa strain 7820 (assumed to be a
 3   microcystin) in mice.  Seventy percent of the radiolabel was found in the liver after 1 minute,
 4   increasing to almost 90% after 3 hours. Radiolabel was also found in the lungs, kidneys, heart,
 5   large intestine, ileum and spleen.
 6
 7          The distribution of 3H-MCLR (>95% pure) was evaluated following i.p. injection of a
 8   sublethal (45 ng/kg) or lethal (101 ng/kg) dose in mice (Robinson et al., 1989). The tissue
 9   distribution of radiolabel was similar after injection of either a lethal or a sublethal dose. Liver
10   accumulation reached a maximal value of 60% by 60 minutes. For the 101 |J,g/kg dose, the liver,
11   intestine and kidney contained 56, 7 and 0.9% of the radiolabel, respectively. Heart, spleen,  lung
12   and skeletal muscle each contained less than  1% of the radiolabel.
13
14          Immunostaining methods were used to evaluate the organ distribution following
15   intratracheal instillation of MCLR purified from a bloom sample (Ito et al., 2001).  Following
16   instillation of a lethal dose (100 ng/kg), the lung, liver, small intestine and kidney were
17   positively stained for MCLR. Intense staining was observed in the lung by 5 minutes post-
18   instillation, followed by the kidney (10 minutes), the small intestine (45 minutes) and the liver
19   (60 minutes). After approximately 90 minutes, bleeding began around the hepatic central vein.
20   The authors described the pathological changes in the  liver as essentially the same as those seen
21   following oral or i.p. injection exposure routes. Intratracheal instillation of a sublethal dose (50
22   Hg/kg) resulted in immunostaining of the lung,  liver, kidney, cecum and large intestine (Ito et al.,
23   2001). No discernable pathological changes were observed at this dose level. Ito et al. (2002)
24   synthesized glutathione and cysteine conjugates of microsystin-LR and administered them by
25   intratracheal instillation in mice.  These conjugates are, according to the authors, known
26   metabolites of microcystins.  The metabolites were demonstrated to be less toxic than the parent
27   compound (lethal doses were about 12-fold higher than the MCLR lethal dose) and were
28   distributed primarily to the kidney and intestine, as opposed to the liver.
29
30          The distribution of MCLR (purified from a bloom sample) following oral gavage
31   administration to mice (500 ng/kg) was investigated using immunostaining methods (Ito et al.,
32   1997a). MCLR was detected in large amounts in the villi of the small intestine.  Erosion of the
33   villi was observed, which may have enhanced absorption of the toxin into the bloodstream.
34   MCLR was also present in the blood plasma, liver, lungs, kidneys and heart.
35
36          The distribution of 3H-dihydroMCLR in mice was shown to differ for the oral and i.p.
37   injection routes of exposure (Nishiwaki et al., 1994).  Intraperitoneal injection of 3H-
38   dihydroMCLR resulted in rapid and continuous uptake in the liver, with approximately 72%  of
39   the administered dose present in the liver after 1 hour. The  3H-dihydroMCLR was synthesized
40   by reduction of N-methyldehydroalanine from microcystin-LR.  Small amounts of radiolabel
41   were found in the small intestine (1.4%), kidney and gallbladder (0.5%), lungs (0.4%) and
42   stomach (0.3%) following i.p. injection. Oral administration of 3H-dihydroMCLR resulted in
43   much lower concentrations in the liver, with less than  1% of the administered dose found in the
44   liver at either 6 hours or 6 days post administration. 3H-DihydroMCLR is rapidly distributed to
45   the liver of swine following i.v. injection or ileal loop  infusion (Stotts et al., 1997a,b).  Smaller
46   amounts were distributed to the kidneys, lungs, heart, ileum and spleen.


                                                 7        DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2          MCLR was not found in the milk of dairy cattle that were exposed toM aeruginosa cells
 3   via drinking water (Orr et al., 2001) or ingestion of gelatin capsule containing the cells (Feitz et
 4   al., 2002).
 5
 6   3.2.2.  Cellular Uptake
 7
 8          The cellular uptake of 3H-dihydroMCLR was evaluated using primary rat hepatocytes in
 9   suspension and in isolated perfused rat liver (Eriksson et al., 1990a; Hooser et al., 1991a). The
10   uptake (as measured by scintillation counting of washed cells) of a mixture of unlabelled MCLR
11   and 3H-dihydroMCLR was shown to be specific for freshly isolated rat hepatocytes (Eriksson et
12   al., 1990a).  Uptake was negligible in human hepatocarcinoma cells (Hep G2), mouse fibroblast
13   (NIH-3T3) and human neuroblastoma cells (SH-SY5Y). The uptake of 3H-dihydroMCLR was
14   also shown to be inhibited by bile acid transport inhibitors such as antamanide,
15   sulfobromophthalein and rifampicin, and by the bile salts cholate and taurocholate (by
16   competing for the bile acid transporter).
17
18          The uptake of 3H-dihydroMCLR was demonstrated to be rapid for the first 5-10 minutes,
19   followed by a plateau, in both rat hepatocyte suspensions and the isolated perfused rat liver
20   (Hooser et al., 1991a). Uptake was measured as radioactivity in fractionated cells versus
21   radioactivity in medium.  The uptake of 3H-dihydroMCLR was inhibited by incubation of
22   suspended rat hepatocytes at 0°C, suggesting the involvement of an energy-dependent process
23   (Hooser et al., 1991a). Uptake was also inhibited by preincubation of hepatocytes with
24   rifampicin, presumably via competitive inhibition of the bile acid transporter (Hooser et al.,
25   199 la).
26
27          Many studies have demonstrated that inhibition of microcystin uptake at the bile acid
28   transporter reduces or eliminates the liver toxicity observed following in vitro or in vivo
29   exposures (Runnegar et al.,  1981, 1995a; Runnegar and Falconer, 1982; Hermansky et al.,
30   1990a,b; Thompson and Pace, 1992).  The human organic acid transport protein (OATP) was
31   shown to mediate the transport of 3H-microcystin (type not specified) in Xenopus laevis oocytes,
32   and this uptake was inhibited by sulfobromophthalein and taurocholate. This transport protein is
33   found in the human brain and may be related to the acute neurotoxicity seen in hemodialysis
34   patients exposed to microcystins (see  Section 4.4.5.1).
35
36          Runnegar et al. (1991) studied the influence of dose level and exposure time on the
37   uptake of 125I-microcystin-YM in isolated rat hepatocytes (measured as radioactivity in
38   centrifuged cell pellet). Hepatocyte uptake was initially rapid with a plateau in the uptake rate
39   observed after 10 minutes. The initial uptake rate (in the first minute of exposure) increased
40   with increasing concentration, but cumulative uptake ceased at a dose that resulted in plasma
41   membrane blebbing.
42
43          Microcystin-YM uptake by isolated rat hepatocytes was temperature-dependent and was
44   inhibited 70-80% by the addition of sodium deoxycholate or sulfobromophthalein (Runnegar et
45   al., 1995b).  This provides evidence to indicate that microcystin uptake occurs by carrier
46   mediated transport, most likely via the bile acid transporter. Pretreatment of mice with bile acid
                                                         DRAFT - DO NOT CITE OR QUOTE

-------
 1   transporter inhibitors (cyclosporine A, rifamycin, trypan blue and trypan red) abolished
 2   microcystin toxicity, suggesting limited or no uptake of microcystins (Runnegar et al., 1995b).
 3   Further, in vitro preincubation of hepatocytes with bile acids or bile acid transport inhibitors
 4   (taurocholate, trypan blue, cholate, sulfobromophthalein, cyclosporine A, trypan red and
 5   rifamycin) each decreased the uptake of microcystin-YM, as measured by assays for protein
 6   phosphatase inhibition in cell lysates (Runnegar et al., 1995b).  Pretreatment with protein
 7   phosphatase inhibitors (i.e., microcystin-YM and calyculin A) also resulted in the inhibition of
 8   both microcystin-YM uptake and protein phosphatase inhibition, suggesting that the bile acid
 9   transporter is itself regulated by serine/threonine phosphorylation.
10
11          Many cell types and established cell lines, including both rodent and some human cells,
12   have been evaluated for potential susceptibility to microcystin uptake and toxicity.  Primary
13   isolated hepatocytes have been shown to be the most sensitive to cytotoxicity, due to the
14   presence of the organic ion/bile acid transport system (Eriksson et al., 1990b).  In addition,
15   primary cultures of liver cells cease to express these bile acid transport proteins after 2-3 days of
16   being maintained in culture.  Therefore, established liver cell lines are generally not useful for
17   evaluating microcystin toxicity (Eriksson and Golman, 1993; Heinze et al., 2001). Chong et al.
18   (2000) evaluated microcystin toxicity in eight permanent cell lines (including rodent, primate
19   and human cell lines), only two of which (human oral epidermoid carcinoma KB cells and rat
20   Reuber H35 hepatoma H-4-II-E cells) showed cytotoxicity following MCLR exposure. The
21   toxic response in these cells  was  most evident if MCLR was added when the cells were seeded.
22   Established monolayers were more resistant to microcystin toxicity. Mechanistic studies that
23   evaluate organ and cell type  specificity for microcystins are further discussed in Section 4.4.7.1.
24
25   3.2.3.  Subcellular Localization and Cytosolic Protein Binding
26
27          Tissue distribution was evaluated in mice given i.v. injection of a sublethal dose of
28   3H-MCLR (Robinson et al.,  1991a). The liver contained approximately 67% of the radiolabel by
29   60 minutes, and the amount of hepatic radioactivity did not change throughout the 6-day study
30   period, despite urinary and fecal  elimination of 24% of the administered dose.  The subcellular
31   distribution of radioactivity in the liver demonstrated that approximately 70% of the hepatic
32   radiolabel was present in the cytosol. In vitro experiments showed that radiolabeled microcystin
33   in the liver was bound to high molecular weight cytosolic proteins (Robinson et al., 1991b).  The
34   nature of the binding was demonstrated to be covalent, saturable and specific for a protein with a
35   molecular weight of approximately 40,000. Binding was inhibited by okadaic acid (a potent
36   inhibitor of serine/threonine  phosphatases [1 and 2A]), suggesting that the target protein is
37   protein phosphatase 1 or 2A. A discussion of protein phosphatase binding and inhibition by
38   microcystins is provided under mechanistic studies in Section 4.4.7.3, below. Binding proteins
39   for MCLR were found in cytosol derived from several different organs, suggesting that liver
40   specificity is not due to limited distribution of target proteins.  Covalent binding to hepatic
41   proteins may be responsible  for the long retention of microcystins in the liver.  Lin and Chu
42   (1994) evaluated the kinetics of MCLR distribution in serum and liver cytosol derived from
43   mice. Uptake of pure MCLR, as analyzed by direct  competitive enzyme-linked immunosorbent
44   assay (ELISA), into the serum was shown to be rapid following an i.p. injection of 35 ng/kg
45   (sublethal dose).  The toxin reached a maximum concentration in the serum by 2 hours and in
46   liver cytosol by 12 hours post-injection. MCLR was shown to be bound to liver cytosolic
                                                          DRAFT - DO NOT CITE OR QUOTE

-------
 1   proteins and the kinetics of binding were correlated with inhibition of protein phosphatase 2 A
 2   activity.  The maximum decrease in enzyme activity was observed 6-12 hours following
 3   injection.
 4
 5          Pace et al. (1991) demonstrated significant accumulation of 3H-MCLR in isolated
 6   perfused liver despite a low overall extraction ratio (16% in liver, 79% in perfusate). In the
 7   liver, radiolabel corresponding to MCLR (15%) and a more polar metabolite (85%) was
 8   primarily found in the cytosolic fraction.
 9
10          The subcellular distribution of 3H-dihydroMCLR was evaluated using primary rat
11   hepatocytes in suspension and the isolated perfused rat liver (Hooser et al., 1991a).
12   3H-dihydroMCLR was  primarily localized in the cytosolic fraction in both the hepatocytes and
13   liver. In the hepatocytes, precipitation with trichloroacetic acid indicated that approximately
14   50% of the 3H-dihydroMCLR was found as free toxin, while the remaining 50% was bound to
15   cytosolic proteins. Since little of the radiolabel was in the insoluble pellet containing insoluble
16   actin and other elements, the authors suggested that 3H-dihydroMCLR did not bind significantly
17   to actin or other cytoskeletal proteins (Hooser et al., 1991a).
18
19          The subcellular protein binding of 3H-dihydroMCLR was evaluated in rat liver
20   homogenates (Toivola et al., 1994). Most  of the radiolabeled toxin (80%) was bound to
21   cytosolic proteins. 3H-dihydroMCLR was shown to bind both protein phosphatase  1 (PP1) and
22   protein phosphatase 2A (PP2A); however, PP2A was detected primarily  in the cytosol, while
23   PP1 was found in the mitochondrial and post-mitochondrial particulate fraction (membrane
24   proteins).  The binding  of microcystins to PP1  and PP2A and the inhibition of protein
25   phosphatase activity is further discussed in Section 4.4.7.3 (Mechanistic Studies).
26
27          Limited information in humans exposed to microcystins intravenously via dialysate
28   indicates that a large proportion of microcystins in the blood are bound to proteins.  Hilborn et
29   al. (2005) compared two techniques for measuring microcystin in the serum of six patients. Use
30   of ELISA, which detects free microcystins, resulted in serum microcystin concentrations ranging
31   from 8 to 51% of the concentrations obtained using gas  chromatography/mass spectrometry
32   (GC/MS) detection of 2-methyl-3-methoxy-4-phenylbutyric acid (MMPB, which is derived from
33   both free and protein-bound microcystins by chemical oxidation). These results indicate that
34   microcystins are bound to proteins in human blood, and that analysis for microcystins using
35   ELISA techniques may underestimate total blood concentrations.
36
37   3.3.   METABOLISM
38
39          Urinary and  fecal metabolites of MCLR were analyzed in samples collected 6 and 12
40   hours following i.v.  injection of a sublethal dose of 3H-MCLR in mice (Robinson et al., 1991a).
41   Approximately 60% of the radiolabel in both the urine and the feces was associated with the
42   parent compound. MCLR was metabolized in  liver cytosol preparations to a product that binds
43   to a high molecular weight cytosolic protein (Robinson  et al., 1991b). The parent compound
44   also binds to this protein, which has been suggested to be the  catalytic subunit of protein
45   phosphatase 2A. In isolated perfused rat liver, binding of both the parent toxin (3H-MCLR) and
46   a more polar metabolite to cytosolic proteins was also demonstrated (Pace et al., 1991). Polar
                                                10       DRAFT - DO NOT CITE OR QUOTE

-------
 1   metabolites accounted for 65-85% of the hepatic cytosol radiolabel. Metabolites of MCLR were
 2   not further characterized in these studies.
 3
 4          3H-Dihydromicrocystin is not extensively metabolized in swine liver after i.v. injection
 5   or ileal loop exposure, and is primarily present in hepatic tissues as the parent compound (Stotts
 6   etal., 1997a,b).
 7
 8          Administration of 125 ng/kg of Microcystis toxin 7820 to mice resulted in decreased
 9   levels of cytochrome b5 and cytochrome P450 (Brooks and Codd,  1987). Pretreatment of mice
10   with microsomal enzyme (mixed function oxidase) inducers (p-naphthoflavone,
11   3-methylcholanthrene and phenobarbital) was shown to eliminate this effect on hepatic enzymes
12   and to extend survival and reduce liver toxicity (i.e., changes in liver weight). In an in vitro
13   study using mouse liver microsomes, cytochrome P450 associated  enzyme activity (i.e.,
14   metabolism of aminopyrene and p-nitrophenol) was not altered by  an unidentified toxin isolated
15   fromM aeruginosa (assumed to be a microcystin; Cote et al., 1986).
16
17          The hepatic metabolism of MCRR and MCLR (purified from blooms) was studied
18   following i.p. injection in mice and rats (Kondo et al., 1996). Glutathione and cysteine
19   conjugates were identified at 3 and 24 hours in both mouse and rat livers. Structural
20   modification of the 3-amino-9-methoxy-2,6,8,-trimethyl-10-phenyldeca-4,6-dienoic acid (Adda)
21   and methyldehydroalanine (MDha) moieties of the toxins was indicated.  Figure 3-1 shows the
22   glutathione and cysteine conjugates of microcystins.
23
24          Kondo et al. (1992) demonstrated that glutathione and cysteine conjugates of MCLR and
25   MCYR were less toxic than the parent compounds based on LD50 estimates, but were still
26   significantly toxic (LD50 values ranged from 217 to 630 ng/kg in mice). Glutathione and
27   cysteine conjugates of MCLR were shown to inhibit protein phosphatases 1 and 2A in  vitro to
28   the same degree as MCLR; however, these metabolites were primarily distributed to the kidney
29   and intestine following intratracheal instillation in mice (Ito et al., 2002). This result suggests
30   that the lower toxicity of glutathione and cysteine conjugates may be related to distribution to
31   excretory  organs and elimination of metabolites in vivo.  Metcalf et al. (2000) also demonstrated
32   that microcystin conjugates with glutathione, cysteine-glycine and  cysteine were less toxic  in the
33   mouse bioassay; however, these conjugates were also shown to be  weaker inhibitors of protein
34   phosphatases 1 and 2A in vitro.  Takenaka (2001) illustrated that glutathione conjugates of
35   MCLR are formed by glutathione S-transferase enzymes found in both rat liver cytosol and
36   microsomes.
37
                                                11       DRAFT - DO NOT CITE OR QUOTE

-------
                                            H,,,  >C°2H
 2
 o
 J
 4
 5
 6
 7
                    X    Y
      Microcystin LR  Leu  Arg
      Microcystin YR  Tyr  Arg
      Microcystin RR  Arg  Arg
                                 GSH conjugate Z =
                                                  H2N
                                                  HOOC  H
                                 Cys conjugate Z =
                                                                9   H
                                                 hLN     COOH
                                                                    ,COOH
 9
10
11
Figure 3-1.  Structures of GSH and Cys Conjugates of Microcystins LR, YR and RR
                            (Kondoetal., 1992)
                                              12       DRAFT - DO NOT CITE OR QUOTE

-------
 1          Several studies have demonstrated an increase in hepatic glutathione levels following
 2   exposure to microcystins (Ding et al., 2000a; Boua'icha and Maatouk, 2004; Gehringer et al.,
 3   2004). MCLR was shown to induce the de novo synthesis of glutathione in mice exposed to a
 4   toxic sublethal dose (75% of the LD50) (Gehringer et al., 2004).  Increased transcription of
 5   glutathione-S-transferase was also demonstrated in this study.
 6
 7   3.4.    ELIMINATION
 8
 9          Limited information on the elimination of microcystins from the human body is available
10   from follow-up of dialysis patients exposed to microcystins intravenously (see Section 4.4.5.1
11   for further detail).  In two separate incidents in Brazil (one in Caruaru, one  in Rio de Janeiro),
12   microcystins were detected in patients' serum more than 50 days after documented exposure
13   (Hilborn et al., 2005; Scares et al., 2006).
14
15          The excretion of microcystins was investigated in female rats (Falconer et al., 1986).
16   The  blood half-life was measured following i.v. administration of a 125I-labelled heptapeptide
17   toxin extracted firomM aeruginosa (MW 1019, assumed to be a microcystin). A biphasic blood
18   elimination curve was demonstrated, with the first component having a half-life of 2.1 minutes
19   and the second component having a half-life of 42 minutes. After  120 minutes, 9.4% of the
20   administered dose was present in the intestinal contents and 2.9% was present in the urine,
21   suggesting that biliary excretion plays a significant role in elimination of microcystins.  Biliary
22   excretion was  also demonstrated in isolated perfused rat liver (Pace et al., 1991).  In the bile
23   collected 10-20 minutes after toxin exposure, 78% of the radiolabel was associated with the
24   parent toxin, while the remaining radiolabel was associated with more polar metabolites.
25
26          MCLR excretion was also evaluated in mice (Robinson et al.,  1991a). A biexponential
27   plasma elimination curve was observed following i.v. injection of a sublethal dose of 3H-MCLR.
28   Plasma half-lives of 0.8 and 6.9 minutes were reported for the first and second phase of
29   elimination, respectively. Approximately 24% of the administered dose was eliminated in the
30   urine (9%) and feces (15%) throughout the 6-day study period. Approximately 60% of the
31   excreted microcystin, measured at 6 and  12 hours following injection, was present as the parent
32   compound.
o o
34          Ito et al. (1997a) demonstrated that MCLR is secreted in the mucous of goblet cells from
35   both the small and large intestine of mice following administration by oral gavage (500 |j,g/kg).
36   MCLR was not detected in urine in this study.
37
38          Stotts et al. (1997a,b) evaluated the toxicokinetics of 3H-dihydroMCLR in swine
39   following i.v. injection and ileal loop exposure. Elimination of 3H-dihydromicrocystin was rapid
40   and followed a biphasic pattern, suggesting that the liver rapidly removes the toxin from the
41   blood. Clearance from the blood is slower at higher dose levels, presumably due to the liver
42   toxicity and circulatory shock observed at high doses. 3H-Dihydromicrocystin was detected in
43   the bile as early as 30 minutes after i.v. injection. Following ileal loop exposure, the
44   concentration of toxin was consistently higher in the portal venous blood as compared to
45   peripheral blood. This suggests that first pass metabolism may play a role in the clearance of
46   dihydroMCLR.
                                                13        DRAFT - DO NOT CITE OR QUOTE

-------
1
2   3.5.   PHYSIOLOGICALLY-BASED TOXICOKINETIC MODELS
3
4         No physiologically based toxicokinetic models have been developed for microcystins.
                                           14       DRAFT - DO NOT CITE OR QUOTE

-------
 1                               4. HAZARD IDENTIFICATION
 2
 3
 4   4.1.    STUDIES IN HUMANS - EPIDEMIOLOGY, CASE REPORTS, CLINICAL
 5          CONTROLS
 6
 7          A number of case reports and epidemiological investigations have examined the
 8   relationship between human exposure to cyanobacteria and various health endpoints. In all of
 9   these studies, humans were exposed to blooms of cyanobacteria in environmental settings.  As a
10   result, the potential for co-exposure to multiple cyanobacterial toxins and/or other
11   microorganisms or compounds to contribute to observed toxicity cannot be ruled out.
12
13   4.1.1.  Oral Exposure
14
15          4.1.1.1.  Short-Term Studies and Case Reports
16
17          Dillenberg and Dehnel (1960) reported on a series of animal poisonings and human
18   exposures to cyanobacterial blooms in various lakes of Saskatchewan, Canada during the
19   previous year. Several cases of individual or group human exposures during recreational
20   activities were reported.  In general, the symptoms were gastrointestinal in nature, including
21   nausea, stomach pain and diarrhea; headache and muscle weakness were also reported.  Stool
22   samples from three of the victims showed evidence of cyanobacteria (Microcystis and
23   Anabaend).  In addition,  water samples from the lakes in which the victims had been swimming
24   showed cyanobacteria. At the time of this report, cyanobacterial toxins had not been fully
25   characterized. Thus, no data on the nature or quantity of toxins in the affected waters were
26   provided.
27
28          Billings (1981) reported a series of outbreaks of human illness potentially associated with
29   exposure to cyanobacteria in two Pennsylvania lakes. Swimmers in both lakes reported
30   symptoms,  including headache, abdominal cramping, nausea, vomiting, diarrhea, hay fever-like
31   symptoms,  ear aches, eye irritation, sore throat, sneezing, runny  nose and swollen lips within a
32   few hours of swimming in the waters.  Investigation by the state Departments of Environmental
33   Resources and Health served to rule out common bacterial, protozoal and viral agents in the
34   outbreaks.  In the second lake, a bloom ofAnabaena was identified. Indirect evidence (rapid
35   onset of symptoms, absence of other potential causative agents and consistency with previous
36   reports of health effects after exposure to cyanobacteria) led the investigators to postulate a role
37   for exposure to Anabaena in the health outcomes.
38
39          Turner et al. (1990) reported a similar type of outbreak among army recruits who had
40   consumed reservoir water during canoe exercises.  Detailed case reports were presented for two
41   recruits. Both 16-year-old recruits presented with several days'  history of malaise, sore throat,
42   blistering around the mouth, dry cough, pleuritic pain and abdominal pain.  One also had
43   experienced vomiting and diarrhea.  Physical examination revealed fever, left basal pulmonary
44   consolidation (pneumonia) and abdominal tenderness in both patients.  Blood tests revealed low
45   platelet counts in both patients. Both were tested for a variety of pathogens, including
46   Leptospim, Legionella, Chlamydia, Coxiella, Mycoplasma and influenza and adenovirus, all
                                                15        DRAFT - DO NOT CITE OR QUOTE

-------
 1   with negative result.  The authors reported similar symptoms (sore throat, headache, abdominal
 2   pain, dry cough, diarrhea, vomiting and blistered mouth) in 16 additional soldiers who had taken
 3   part in the canoe exercises.  The reservoir contained a bloom of cyanobacteria, primarily M
 4   aeruginosa. Further, a sample of the bloom taken the day after the patients were admitted into
 5   the hospital showed microcystins to be present, including MCLR. High levels of Escherichia
 6   coli were also found in reservoir water sampled 2 weeks later. The authors suggested that
 7   microcystin exposure may have had a role in the pulmonary consolidation and low platelet count
 8   of the two patients, citing evidence from studies in mice (the authors cited Falconer et al., 1981
 9   and Slatkin et al., 1983).
10
11          Teixeira et al. (1993) characterized an epidemic of gastroenteritis, primarily among
12   children, in the vicinity of the Itaparica Dam in Bahia, Brazil. The authors collected data on the
13   incidence of treatments for diarrhea between February and May of 1988. Timing of 1118 cases
14   of diarrhea in local health units was recorded, showing a spike in the incidence of gastroenteritis
15   coinciding  with the flooding of the Itaparica Dam reservoir. Most (about 70%) of the cases
16   involved children under the age of 5 years. Additional data were collected on the age, residence,
17   symptoms, foods consumed, source of drinking water and travel for 76 patients given outpatient
18   treatment for diarrhea.  Fecal, blood and urine samples were collected from these patients and
19   analyzed for chemical and biological contaminants (i.e., bactedologic, virologic, cholinesterase,
20   heavy metals).  In addition,  water samples were analyzed for chemical and biological
21   contaminants (i.e., organophosphates, carbamates, heavy metals, fecal coliform). Analysis of
22   biological samples showed no contaminants.  Untreated water samples showed high levels of
23   fecal coliform, but samples  of treated water did not contain significant levels.  Untreated water
24   samples also revealed high counts (1104-9755 units per mL) ofAnabaena andMicrocystis cells,
25   4-32 times  the World Health Organization (WHO)  maximum acceptable cell count for untreated
26   water (300 units/mL at the time). No data were provided on levels of cyanobacteria in the
27   treated water.  It is not clear from the publication whether affected persons were  exposed to
28   treated or untreated water. This study does not provide information on health effects of
29   microcystin exposure, but provides some circumstantial evidence for gastrointestinal  effects
30   from exposure to cyanobacteria.
31
32          A case control study investigated the incidence of gastrointestinal and dermatological
33   symptoms among persons exposed to Murray River water (Australia) (el Saadi and Cameron,
34   1993; el Saadi et al., 1995).  Physicians in 8 of 11 towns along the Murray River participated in
35   the study, recruiting 102 gastrointestinal and 86 dermatological cases between January and
36   March, 1992. Gastrointestinal cases were patients with abdominal pain, vomiting or diarrhea;
37   dermatological  cases had rash, itching or blistering of the mouth. Control patients (132) were
38   selected as the next patient entering the office after each case, when possible. For each study
39   participant, age, sex, primary source of drinking and domestic water (rain/spring, untreated river
40   water or chlorinated river water from a town supply) and recreational water contact (none, river
41   or lake contact, or other, such as pool contact) during the previous week were recorded.  River
42   water samples were collected and cyanobacteria identified and quantified. Anabaena,
43   Aphanizomenon and Planktothrix were the most common cyanobacteria identified in the
44   samples; small numbers of M aeruginosa were infrequently identified.  Both univariate and
45   multivariate analysis of the  data showed the odds of having gastrointestinal symptoms to be
46   raised in persons drinking chlorinated river water or using untreated river water for domestic
                                                16        DRAFT - DO NOT CITE OR QUOTE

-------
 1   purposes. Likewise, both types of statistical analyses showed increased odds of having
 2   dermatological symptoms for persons less than 20 years of age and for persons using untreated
 3   river water for domestic purposes.  The proportion of patients with gastrointestinal symptoms
 4   and the proportion with dermatological symptoms both correlated with mean weekly log
 5   cyanobacterial cell count in the river, although the correlation was statistically significant only
 6   for gastrointestinal symptoms. However, when upper and lower reaches of the river were
 7   analyzed separately, nonsignificant correlations were observed.
 8
 9          No data on cyanobacterial toxins in the river water were provided.  The symptoms
10   reported in this study cannot be readily attributed to any particular toxin (if they are indeed
11   associated with toxin exposure rather than exposure to the living cyanobacterial cells) due to the
12   absence of toxin analyses, as well as the identification of genera with potential to produce
13   multiple toxins.  In addition, the potential for other microbial or chemical contaminants in the
14   untreated river water was not evaluated in this study.
15
16          Pilotto et al. (1997) conducted a prospective study of gastrointestinal and dermatological
17   symptoms among people exposed to cyanobacteria at water recreation sites in Australia. Study
18   participants were individuals 6 years of age and older who were present at one of several water
19   bodies that were both used for recreational purposes and expected to have algal blooms during
20   the summer.  Interviewers visited these sites on several Sundays and invited  all individuals to
21   participate. Participants completed a questionnaire to evaluate health status and the nature and
22   duration of water-contact activities.  In addition, subjects were asked whether they had
23   symptoms or recreational water contact in the 5 days prior to study initiation, in order to control
24   for the effects of prior health conditions and prior exposure on study findings. Five hundred and
25   fourteen persons had either pre-existing symptoms or water contact in the days prior to initial
26   interview.  Participants were contacted by telephone 2 and 7 days later, at which time the
27   occurrence of diarrhea, flu-like  symptoms, skin rashes, mouth ulcers, fevers or eye or ear
28   infections in the intervening time was recorded. Water samples for cyanobacterial cell count and
29   toxin analysis were collected at 10 a.m. and 2 p.m. on the day of initial interview.  Toxin
30   presence was assessed by mouse bioassay (i.p. injection of 500 mg freeze-dried cells/kg body
31   weight).
32
33          Of 1029 persons invited to participate, 921 persons participated in the study (Pilotto et
34   al., 1997). Interviewers were able to contact 845 of these persons by telephone 2 days after
35   initial interview, and 852 persons 7 days later. No differences in the reporting of gastrointestinal
36   and dermatological symptoms were found between those who had water contact and those
37   without water contact (on the day of the initial interview) when contacted 2 days later (Pilotto et
38   al., 1997). However, when subjects with water contact or symptoms prior to initial interview
39   were excluded, a significant trend to increasing symptom occurrence with duration of exposure
40   was  observed among persons contacted 7 days after initial interview. Cyanobacterial cell count
41   showed some correlation with symptom occurrence, but presence/absence of hepatotoxins did
42   not.  The authors postulated that any association between symptoms and exposure resulted from
43   the allergenic nature of the cells rather than exposure to toxins.

44          Pilotto et al. (1999) evaluated the relationship between cyanobacterial exposure  and
45   perinatal outcomes in an ecological study conducted in Australia.  Cyanobacterial monitoring
                                                 17       DRAFT - DO NOT CITE OR QUOTE

-------
 1   data (cell counts) were collected from raw drinking water supplies in 156 towns.  Perinatal
 2   outcome data were obtained from several registries (for calendar years 1992-1994) and the
 3   following variables assessed: premature birth (<36 weeks), low birth weight (<2.5 kg), very low
 4   birth weight (<1.5 kg) and congenital defects (at least one). Maternal residence at birth was used
 5   to assess exposure based on cyanobacterial cell counts. Exposure was assessed at various
 6   gestational periods, either as the proportion of time with cyanobacterial exposure (proportion of
 7   weeks with non-zero levels) or average alert level (alert level 1 = <2000 cells/mL; alert level 2 =
 8   2000-15,000 cells/mL; alert level 3 = <15,000 cells/mL). Data on 32,700 births were collected,
 9   although the numbers with exposure data in different gestational  periods varied. A significant
10   difference in the incidence of low birth weight and very low birth weight babies was observed
11   between mothers with and without cyanobacterial exposure during the first trimester. Very low
12   birth weight incidences increased with increasing exposure to cyanobacteria (as measured by the
13   proportion of first trimester with non-zero cyanobacterial cell counts). At the highest exposure
14   category (exposure to cyanobacteria during 100% of first trimester), the odds ratio (OR) was
15   1.42 (95%  confidence interval [CI] = 1.00-2.02).  When exposure was assessed as the average
16   alert level (cell concentration interval), there was a significant increase in congenital defects at
17   the highest average alert level of 2.5-3.0  (OR = 2.03, 95% CI = 1.37-3.01), but the trend was not
18   significant.  This study had a number of limitations, including a lack of individual exposure data
19   and lack of data on cyanobacteria or toxins in the finished water (after various treatment
20   processes). Further, because the measure of exposure was cyanobacteria rather than toxin, it is
21   difficult to interpret this study in the context of microcystin effects.
22
23          Falconer et al. (1983) compared the hepatic enzyme levels in patients served  by a public
24   water supply contaminated with a bloom of M. aeruginosa with levels in patients living in areas
25   served by other water supplies.  Enzymes assessed in the study were y-glutamyltransferase
26   (GGT),  aspartate aminotransferase (AST), alanine aminotransferase (ALT) and alkaline
27   phosphatase (ALP). The study population consisted of all patients served by a single hospital
28   laboratory and referred for liver function tests before,  during and after a bloom of M aeruginosa
29   in the Malpas Dam reservoir of Australia. Patients were classified either as residents of the city
30   of Armidale, which uses the reservoir for drinking water supply,  or residents of neighboring
31   towns with independent water supplies. Liver function test results within each comparison group
32   were further sorted by date into three categories: testing during the 5 weeks before the first signs
33   of the bloom appeared, testing during the 3-week bloom or the 2  weeks following copper sulfate
34   treatment of the bloom (identified as the  high-risk time interval due to the cell lysis and
35   subsequent toxin release) or testing during the 5 weeks that followed.
36
37          Results of plasma enzyme analyses were compared before, during and after the bloom
38   among residents of Armidale and surrounding areas (Falconer et  al., 1983).  Analysis of variance
39   was used to assess differences in enzyme levels between comparison groups and between times
40   within comparison groups.  Results of the statistical analysis indicated a significant rise in GGT
41   levels in residents of Armidale during the bloom period.  ALT levels in Armidale residents
42   increased during the bloom period, but the change was not statistically significant.
43
44          The authors noted substantial variability in enzyme levels, attributing this finding to the
45   imprecise method of selecting study participants (Falconer et al., 1983).  It should be noted that
46   several of the enzyme measurements for  the referent population were associated with one
                                                 18       DRAFT - DO NOT CITE OR QUOTE

-------
 1   individual requiring repeat analysis for chronic kidney disease. Alcoholism, which can increase
 2   GGT levels, was reported to occur at about the same proportion (7-10%) in the groups assessed
 3   before and during the bloom, although it was substantially lower in the post-bloom group of
 4   Armidale residents. The authors concluded that the change in GGT among Armidale residents
 5   before and during the bloom period might potentially be associated with exposure to M
 6   aeruginosa.
 1
 8          4.1.1.2. Long-Term Studies and Epidemiological Studies
 9
10          Zhou et al. (2002) conducted a retrospective cohort analysis of colorectal cancer and
11   exposure to microcystins in drinking water in a Chinese province in which an association had
12   been reported previously (Jiao et al., 1985; Chen et al., 1994). Cases of primary colorectal
13   adenocarcinoma between 1977 and 1996 from eight randomly selected towns within Haining
14   City of Zhejiang Province were used as the study population.  Cases were identified using the
15   local cancer registry and independently verified by two pathologists. Drinking water source used
16   longest during the lifetime was used as a surrogate measure of exposure to microcystins.
17   Information on drinking water source  was obtained by interview of patients or family members
18   of deceased cases.  In each of the eight towns, 10 water sources (3  rivers, 3 ponds, 2 wells and 2
19   taps) were randomly selected and sampled for microcystins twice in each of the months of June
20   through September (total of eight samples from each source).  Water samples were analyzed for
21   microcystins by ELISA; the authors did not specify the targeted congeners. The authors do not
22   specify the nature of the "tap" water sources, but the text implies that the tap water derives from
23   one or more treatment plants.
24
25          The average incidence rate of colorectal cancer across all of the study areas was
26   8.37/100,000 per year.  The incidence rate was compared among the four different water sources,
27   with well water users serving as the referent population.  Compared with the incidence among
28   well water users, the colorectal cancer incidence rates among users of the other water sources
29   were significantly increased. Tap water use was associated with a  relative risk of 1.88, while
30   river and pond water use were both associated with a relative risk over 7.0.  There was no
31   difference in colorectal cancer incidence between river and pond water users. The authors
32   suggested that exposure to trihalomethane compounds might account for the increase in
33   incidence among tap water users. Table 4-1  shows the incidence rate, relative risk and 95% CIs
34   for these exposure comparisons.
35
36          Microcystins were detected at  concentrations exceeding 50 pg/mL (considered by the
37   authors to be the limit for positive detection) only in river and pond water,  and the average
38   concentrations in these sources were substantially higher (30- to 50-fold) than well or tap water.
39   A similar proportion (about 25%) of the residents in each of the eight towns used river and pond
40   water for drinking water, allowing an  analysis comparing the average microcystin concentration
41   in river and pond water in each town with the incidence rate by town. This analysis showed a
42   strong correlation between incidence rate and concentration of microcystin (Spearman
43   correlation
                                                19       DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-1 . Relative Risk of Colorectal Cancer and Microcystin Concentration by Drinking
Water Source (Zhou et al., 2002)
Water
Source
Well
water
Tap
water
River
water
Pond
water
Colorectal
Cancer
Incidence
Rate per
100,000
3.61

6.77

28.5

27.76

Relative
Risk of
Colorectal
Cancer
_

1.88

7.94

7.7

95% CI
_

1.39-2.54

6.11-10.31

5.75-10.30

Number of
Microcystin
Samples >50
pg/mL
0/12

0/17

25/69

6/35

Mean
Microcystin
Concentration
(pg/mL)
0.73

4.85

141.08

106.19

Maximum
Microcystin
Concentration
(pg/mL)
9.13

11.34

1083.43

1937.94

20     DRAFT - DO NOT CITE OR QUOTE

-------
 1   coefficient = 0.88, p<0.01). Figure 4-1 shows the relationship between colorectal cancer
 2   incidence and average microcystin concentration.
 3
 4          This study provides suggestive evidence for an association between microcystin exposure
 5   and colorectal cancer. It is also consistent with earlier reports of an association between
 6   drinking river or pond water and incidence of colorectal cancer in the Zhejiang Province of
 7   China (Jiao et al., 1985; Chen et al., 1994; studies published in Chinese and not translated for
 8   this review). However, because demographic information was not provided in the report, it is
 9   not clear whether dietary, genetic and lifestyle factors associated with colorectal cancer have
10   been adequately controlled in the analysis.  Further, other potential biological and chemical
11   contaminants in the river and pond water were not considered.
12
13          Several  epidemiological studies have examined the relationship between drinking water
14   source (well, river, pond or ditch)  and liver cancer in Haimen City, Jian-Su Province, China, an
15   area with an elevated hepatocarcinoma incidence (Yu, 1989; Yu et al., 1989).  These studies,
16   published in Chinese and not translated for this review, showed an increased risk of primary liver
17   cancer associated with consumption of pond or ditch  water (Ueno et al., 1996). According to
18   Health Canada (2002), Yu (1989)  showed that consumption of pond or ditch water was
19   associated with an 8-fold increase in liver cancer incidence when compared with well water
20   consumption. Health Canada (2002) reported that a larger study of 65 counties in China, also
21   published in Chinese (Chen et al.,  1991), showed the opposite; that consumption of deep well
22   water was associated with  an increased risk of liver cancer.
23
24          Ueno et al. (1996) conducted a survey of microcystin content in drinking water supplies
25   in Haimen City to test the hypothesis that microcystins in surface drinking water supplies could
26   contribute to the higher incidence  of liver cancer. Microcystins were measured by ELISA in
27   shallow and deep wells, as well as in ponds/ditches and river waters.  The authors did not
28   indicate which congeners were targeted by the ELISA.  Occurrence of microcystins  was higher
29   in pond/ditch water (17% reported as positive with concentration >50 pg/mL) and river water
30   (32% positive) samples than in shallow wells (4% positive) or deep wells (no detections >50
31   pg/mL).  Further, microcystin concentrations averaged  across the drinking water types were
32   different, averaging 101, 160 and 68 pg/mL in pond/ditch, river and  shallow well samples,
33   respectively.  These data, while suggestive,  do not directly associate exposure to microcystins
34   and liver cancer, since individual exposures were not measured or estimated, and other
35   biological or chemical contaminants in the surface waters have not been ruled out.
36
37          In a case-control study of liver cancer in Haimen City, conducted by Yu et al. (2002), a
38   variety of liver cancer risk factors  were evaluated, including hepatitis B and C virus infection,
39   aflatoxin Bl or microcystin exposure, smoking, drinking, diet and genetic polymorphisms. From
40   a pool of 248 patients with hepatocellular carcinoma  and 248 age-, sex- and residence-matched
41   controls, 134 paired cases and controls assented to blood samples for virus infection and ALDH2
42   and CYP2E1 gene polymorphism  analyses.  Data from  these analyses were combined with
43   questionnaire information on possible lifestyle and dietary risk factors for liver cancer.
44   Microcystin exposure was assessed categorically based on drinking water supply (tap, deep or
                                                21        DRAFT - DO NOT CITE OR QUOTE

-------
 2
 4
 6
 8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
46
47
48
49
50
51
52
53
54
            Incidence of colorectal cancer vs. microcystin concentration (Zhou et al., 2002)
     14
  o
  o
  o  10
  o  12
     10 -
  o
  o
      6 -
      2 -
                                                            y = 0.0267x + 3.0079
                                                                R2 = 0.8277
                   50         100        150        200        250
                       Microcystin concentration (pg/mL) in river and pond
300
350
Figure 4-1.  Relationship between Colorectal Cancer and Microcystin Concentration in River
                 and Pond Water in Haining City, China (Zhou et al., 2002)
                                                   22
                                                     DRAFT - DO NOT CITE OR QUOTE

-------
 1   shallow well, river, ditch or pond), as in earlier studies (Yu et al., 1989). Neither univariate nor
 2   multivariate analysis of the data indicated an association between consumption of river, pond or
 3   ditch water and hepatocellular carcinoma. Hepatitis B virus infection was strongly associated
 4   with primary liver cancer, and history of i.v. injection was also identified as a risk factor (Yu et
 5   al., 2002).
 6
 7          Fleming et al. (2002) conducted an ecological epidemiological investigation of the
 8   relationship between drinking water source and incidence of primary liver cancer in Florida.
 9   The study was prompted by data showing cyanobacteria and toxins, especially microcystins, in
10   surface drinking water sources in Florida. The study population consisted of all cases of primary
11   hepatocellular carcinoma reported to the Florida state cancer registry between 1981 and 1988.
12   The study population was divided into comparison groups consisting of those served by surface
13   drinking water supply and those using other sources.  Residence at the time of diagnosis was
14   used to place cases into the various comparison groups. Surface water treatment plants  and their
15   service areas were geocoded, as were deep groundwater treatment plants. Several comparisons
16   were made.  First, incident cases residing in the service area of a surface water treatment plant
17   were compared with those residing in the service area of a deep groundwater treatment plant.
18   Within this comparison, there were several referent groups; one randomly sampled from the
19   available groundwater service areas, one matched on median income and rent, one matched on
20   ethnic makeup and one matched on income, rent and ethnicity.  Second, incident cases in the
21   surface water service area were compared with equally-sized buffer areas surrounding the
22   surface water service area, but not served by the treatment plant. Finally, incident cases were
23   compared with the incidence in the general Florida population.
24
25          Evaluation of the individual incidence rates in the 18 surface water service areas with the
26   groundwater service areas did not reveal any statistically significant differences among the
27   individual incidence rates. When the service areas were pooled, residence in a  surface water
28   service area was associated with  a statistically significant reduced risk of hepatocellular
29   carcinoma compared with either groundwater service areas (standardized rate ratios [SRR]
30   ranged from 0.8 to 0.98 for the four groundwater comparison groups) or the general Florida
31   population (SRR  of 0.8). It should be noted that the measure of exposure, residence within a
32   surface water service area, was estimated as the average size plus two  standard deviations of the
33   service area for this comparison.
34
35          When comparisons were made between residence in the actual (i.e., not estimated as
36   above) surface water service areas and residence in the buffer areas surrounding the service
37   areas, a statistically significant increase in the incidence of hepatocellular carcinoma was
38   observed for those residing within the surface water service area (SRR = 1.39, CI = 1.38-1.4).
39   Analyses of 1990 census data suggested that the ethnic and socioeconomic backgrounds of the
40   service areas and  buffer areas were similar, although the authors did not report these data.
41   Interestingly, the  incidence of hepatocellular carcinoma in the buffer areas was significantly
42   lower than that in the general Florida population (SRR = 0.59).
43
44          An ecological  study such as this is useful for generating hypotheses, but not for
45   establishing an exposure-response relationship due to the lack of exposure data on individuals.
46   In this case in particular, there is strong potential for misclassification  of exposure.  Residence in
                                                23        DRAFT - DO NOT CITE OR QUOTE

-------
 1   a surface water service area at the time of diagnosis of hepatocellular carcinoma is a poor
 2   measure of potential exposure to cyanobacterial toxins, especially given residential mobility and
 3   likely latency time for cancer development. Further, the initial comparisons with groundwater
 4   service areas used GIS-generated estimates of surface water service areas rather than actual
 5   service areas, leading to greater potential for misclassification.
 6
 7          Fleming et al. (2004) also conducted an ecological study assessing the relationship
 8   between incident colorectal cancer and proximity to a surface drinking water treatment plant,
 9   with the latter representing a surrogate for exposure to cyanobacteria.  Methods for this study
10   were identical to those described above for Fleming et al. (2002) except that colorectal cancer
11   data were abstracted from the Florida Cancer Data System from 1981-1999. As with Fleming et
12   al. (2002), comparisons were made between the colorectal cancer incidence rates in the 18
13   surface water treatment service areas with several referent groups (a random group of
14   groundwater treatment service areas, a group  of groundwater treatment service areas matched on
15   median income and rent, a group of groundwater treatment service areas matched on ethnic
16   makeup, a group of groundwater treatment service areas matched on both median income and
17   ethnicity,  groups residing in an equally-sized  buffer areas surrounding the surface water service
18   area and, finally, the general Florida population).  Mann Whitney rank sum tests of all
19   comparisons did not suggest an association between colorectal cancer and residence at time of
20   diagnosis  in a surface water treatment area (details not provided). This ecological study is
21   subject to the same limitations as described above for Fleming et al. (2002).
22
23   4.1.2.  Inhalation Exposure
24
25          No studies of human exposure to microcystins via inhalation were identified in the
26   materials reviewed for this document.
27
28   4.2.    ACUTE, SHORT-TERM, SUBCHRONIC AND CHRONIC STUDIES AND
29          CANCER BIOASSAYS IN ANIMALS - ORAL AND INHALATION
30
31          Early research on cyanobacterial toxins examined the effects of exposure to
32   cyanobacterial organisms rather than on the toxins now known to produce many of the
33   toxicological effects. In the case of microcystins, the isolation and characterization of important
34   toxin(s) did not occur until the 1980s (WHO,  1999). As a result, many studies have used various
35   extracts of cyanobacterial blooms as test substances in toxicological experiments. These studies
36   contribute to the hazard identification for cyanobacterial toxins, but, as discussed below, are not
37   useful  for dose-response assessment.
38
39          The quantity of an individual cyanobacterial toxin in different bloom samples and
40   extracts varies widely, being influenced by a number of different factors. Some toxins are
41   produced by more than one genus of cyanobacteria. For example, microcystins have been shown
42   to be produced by Microcystis, Anabaena, Planktothrix, Nostoc and others (WHO,  1999). Some
43   species (e.g., Anabaend) can produce more than one cyanotoxin (WHO, 1999). Even within a
44   species, different strains produce varying levels of toxin; some produce little or no toxin at all.
45   Growth conditions can also contribute to the level of toxin produced by a given species and
46   strain (WHO, 1999).  Finally, any given sample may contain multiple genera, species or strains
                                                24        DRAFT - DO NOT CITE OR QUOTE

-------
 1   of cyanobacteria, as well as other contaminating organisms.  Some of these variables will also
 2   apply to cyanobacterial cells cultured in a laboratory, although clonal cultures may be
 3   characterized as to toxin content. In general, there is no clear means of predicting the toxin
 4   content in a given bloom sample or cell extract.
 5
 6          In addition to the variations in toxin production within bloom and/or culture samples,
 7   there are variations in toxin concentration depending on the method used for producing a
 8   material for toxicological administration.  Among the studies reviewed, the administered
 9   material included lyophilized bloom samples in solution, cell-free extracts, cell-free lysates,
10   partially purified toxins, purified toxins and others.  As endotoxins, microcystins exist primarily
11   within the cyanobacterial cell, and are released when cells are lysed.  As a result, studies of
12   extracts obtained by removing intact cells may not contain much, if any microcystin. Purified or
13   partially-purified toxins were used in a number of studies; however, the specific microcystin
14   congener or congeners may not have been identified. As a result, data using bloom samples, cell
15   extracts or partially purified toxins cannot be considered reliable information in relating
16   exposure to a given toxin with toxicological effect.
17
18          Giving due consideration to the limitations of algal extract studies, a distinction is made
19   between toxicological data obtained using purified microcystins and toxicological data obtained
20   using a bloom sample or extract. These data are discussed separately in this report, with the
21   latter data being considered supplemental due to the confounding factors outlined above. As a
22   result,  studies of cyanobacterial extracts are not reviewed in depth.
23
24   4.2.1.  Oral Exposure
25
26          4.2.1.1. Acute Studies
27
28          4.2.1.1.1. Purified Microcystins
29
30          Fawell et al. (1999) conducted acute, subchronic and developmental toxicity studies of
31   MCLR in mice and/or rats. In the acute portion of the study, single oral gavage doses of MCLR
32   (purity not specified) in aqueous solution were given to male and female CR1 :CD-
33   1(ICR)BR(VAF plus) mice and CR1 :CD(SD)BR(VAF plus) rats (five per sex per species).
34   Doses  of 500, 1580 and 5000 ng/kg body weight were administered.  Untreated control groups
35   were not included. The animals were observed for up to 14 days prior to sacrifice and necropsy.
36    Microscopic examinations of the lung and liver were conducted. LD50 values were calculated.
37
38          Oral LD50 values were estimated to be about 5000 ng/kg for mice  and over 5000 ng/kg
39   for rats. Animals that died showed clinical signs, including hypoactivity and piloerection;
40   however, clinical signs were absent in survivors.  Body weights among surviving animals were
41   not affected during the 14-day follow-up.  Necropsy of the animals that died showed darkly
42   discolored and distended livers, as well as pallid kidneys, spleen and adrenals.  Livers of all
43   animals that died had moderate  or marked centrilobular hemorrhage.  The incidence and severity
44   of liver lesions increased in a dose-dependent fashion, as shown in Table 4-2.
45
                                                25        DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-2. Incidence of Liver Lesions in Mice and Rats Treated with a Single Dose of MCLR
(Fawell et al., 1999)
Dose
Og/kg)
Number
Animals
Treated
Mortality
Diffuse
Hemorrhage
Moderate
Centrilobular
Hemorrhage
Marked
Centrilobular
Hemorrhage
Centrilobular
Necrosis
Cytoplasmic
Vacuolation
Mice
500
1580
5000
10
10
10
0
1
5
2
2
1
0
2
7
0
1
0
0
0
2
0
0
0
Rats
500
1580
5000
10
10
10
0
0
1
8
7
8
0
0
1
0
0
1
0
0
1
0
0
1
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

21
22
       Diffuse hemorrhage in the liver was seen in rats and mice of all dose groups, but the
incidence was not clearly related to dose. Because an untreated control group was not included
in this study, it is not possible to say whether the liver effects were treatment-related; thus,
neither a NOAEL nor a LOAEL could be determined.  The single mouse death at 1580 ng/kg
indicates that this is a frank effect level (PEL) in this species; the PEL in rats was 5000 ng/kg
with a single rat death at this dose.

       Yoshida et al. (1997) assessed the acute oral toxicity  of purified MCLR (>95% pure by
HPLC) in female BALB/c mice. Preliminary experiments using doses of 16.8 and 20 mg/kg
resulted in death within 160 minutes in two mice; therefore, doses of 8.0, 10.0 and 12.5 were
chosen for LDso determination. MCLR in saline solution was administered via gavage to a total
of five 6-week-old mice. Two control mice received saline via gavage. Mortality was observed
over a 24-hour period, and dead animals, including those in the preliminary study, were
immediately necropsied. One surviving mouse was sacrificed  and necropsied 24 hours after
treatment; the remainder were sacrificed and necropsied after a week. The liver, kidneys and
lung were sectioned and examined by light microscopy. Electron microscopy was used to
identify apoptotic cells in the livers of treated mice. The remaining tissues were subjected to
histopathological analysis.
       Mortality within 24 hours was 0/1 at 8 mg/kg, 0/2 at  10 mg/kg and 2/2 at 12.5 mg/kg.
The oral LD50 was calculated to be 10.9 mg/kg.  No effects on  the stomach, intestine, skin or
                                                26
                                                    DRAFT - DO NOT CITE OR QUOTE

-------
 1   organs other than the liver and kidneys were observed. Liver effects were pronounced in
 2   animals that died, including centrilobular hemorrhage and hepatocyte degeneration, as well as
 3   free hepatocytes in the veins of mice receiving doses in excess of 12.5 mg/kg (in the preliminary
 4   experiments).  Effects on the kidneys included proteinaceous eosinophilic materials in the
 5   Bowman's spaces of mice receiving high doses (>12.5 mg/kg).  In a single mouse treated with
 6   10.0 mg/kg and sacrificed after 24 hours, evidence of hepatocellular necrosis was observed in the
 7   centrilobular and midzonal regions, and single cell death (possibly apoptotic) was reported in the
 8   centrilobular region, as well as surrounding necrotic areas.  In the other mouse treated with 10
 9   mg/kg and the two mice treated with 8.0 mg/kg (all sacrificed 1 week after treatment), the livers
10   contained hypertrophic hepatocytes in the centrilobular region and fibrosis in the centrilobular
11   and midzonal regions. A few apoptotic cells were observed in these animals. No kidney effects
12   were reported in animals that survived treatment for at least 24 hours.  No control group was
13   included, so neither a NOAEL nor a LOAEL could be determined from this study. The deaths of
14   both mice given 12.5 mg/kg MCLR indicate that this is an FEL.
15
16          Ito et al. (1997a) compared the acute effects of MCLR on the livers and gastrointestinal
17   tracts of young and aged mice. Single doses of 500 ng/kg MCLR (purity not specified)
18   dissolved in ethanol and diluted in saline were administered via oral gavage to aged (29 mice
19   aged 32 weeks) and young (12 mice aged 5 weeks) male ICR mice.  Three aged and three young
20   untreated mice served as controls. Twenty-two aged mice were sacrificed at 2 hours, five mice
21   at 5 hours, and two mice at 19 hours after treatment; four young mice were sacrificed at each
22   time point. Liver damage and gastrointestinal erosion were evaluated.
23
24          The results showed marked differences between young and aged mice in both liver
25   damage and gastrointestinal effects.  In young mice, no liver pathology or gastrointestinal
26   changes were reported.  In contrast, 18 of 29 aged mice treated with the same dose developed
27   pathological changes of the liver. Among the aged mice, 8  of 29 had liver injury of the highest
28   severity, characterized as bleeding, disappearance of many hepatocytes in the whole liver and
29   friable tissue (severity rating of+4). Five of 29 mice had liver changes characterized by
30   bleeding and disappearance of hepatocytes in centrilobular region (severity rating of+3).
31   Necrosis in the centrilobular region was observed in 4 of 29 mice (severity rating of+2) and
32   eosinophilic changes in the centrilobular region were indicated for one mouse (severity rating of
33   +1).
34
35          Gastrointestinal effects observed in the treated aged mice included necrosis to  one-third
36   depth of the mucosa and severe duodenal damage (including decreased villi density, separation
37   of epithelial cells from lamina propria and edema of both the submucosa and villi).  Details of
38   the incidence of these effects were not reported; however, the authors indicated that the degree of
39   liver injury was related to the severity of gastrointestinal effects. Regeneration of intestinal
40   tissues was evident in some of the mice sacrificed at later time points (5 and 19 hours  after
41   treatment). Among untreated aged mice, serum enzyme levels (AST and ALT) were not
42   different, but gastrointestinal condition was somewhat compromised in the aged mice. Aged
43   mice had thinning of surface epithelial cells with consequent exposure of lamina propria and
44   glands in some areas.  The authors postulated that the oral uptake of MCLR was dependent on
45   gastrointestinal tract erosion and the loss of permeability in capillaries of the villi.  This study
                                                27        DRAFT - DO NOT CITE OR QUOTE

-------
 1   identified a freestanding LOAEL of 500 ng/kg (in aged mice only) for liver and gastrointestinal
 2   effects.
 3
 4          In a poorly described study, Fitzgeorge et al. (1994) administered MCLR via gavage to
 5   newly weaned CB A/BALBc mice. The commercially-obtained compound was described only as
 6   "suitably purified."  The LDso was estimated to be 3000 ng/kg, and increases in liver (43%) and
 7   kidney (5.9%) weights were reported. The authors reported that there was no change in lung or
 8   spleen weight; other endpoints were either not examined or not reported.
 9
10          Rogers et al. (2005) evaluated the potential synergism between MCLR and anatoxin-a
11   administered by gavage to CD-I mice (sex not specified).  A total of 60 fasted mice were given
12   gavage doses of 0, 500 or 1000 ng/kg MCLR (purity >98%) followed 50 minutes later with
13   gavage doses of 0, 500, 1000 or 2500 |J,g/kg anatoxin-a (purity >95%).  Controls were given
14   distilled water by  gavage. While not specified, group sizes are presumed to have been five
15   animals per treatment. The animals were observed for clinical signs of toxicity, loss of appetite
16   and mortality; body weight was measured before treatment and 3 hours later.  The duration of
17   observation was not reported.  No deaths, clinical signs of toxicity or differences in body weight
18   were observed. Effect levels cannot be identified from this study due to inadequate reporting
19   and because few toxicological endpoints were evaluated.
20
21          4.2.1.1.2.  Cyanobacterial Extracts
22
23          Rao et al. (2005) compared the acute oral effects of microcystin extract in aged (36
24   weeks old) and young (6 weeks old) Swiss albino mice. A single LD50 dose of extract was
25   administered to male mice; mortality occurred after 4-5 hours. Both groups of mice had
26   increased relative liver weight and DNA fragmentation compared to control, but there was no
27   difference between the age groups. In contrast, glutathione depletion and lipid peroxidation
28   were significantly greater in the aged mice when compared with young mice.  Further, while
29   most serum enzymes were increased over controls in both groups, GGT was increased to a
30   greater extent in aged mice than in young mice.
31
32          4.2.1.2. Short-Term Studies
33
34          4.2.1.2.1.  Purified Microcystins
35
36          Heinze (1999) evaluated the effects of MCLR in drinking water on 11-week-old male
37   hybrid rats (Fl generation of female WELS/Fohm x male BDIX).  Groups of 10 rats were given
38   doses of 0, 50 or 150 ng/kg body weight for 28 days in drinking water.  Water consumption was
39   measured daily and rats were weighed at weekly intervals.  Dose estimates provided by the
40   authors were not adjusted to account for incomplete drinking water consumption (3-7% of
41   supplied water was not consumed over the 28-day period).  The test material was obtained
42   commercially, but the authors did not report a measure of purity. After 28 days of exposure, rats
43   were sacrificed by exsanguination under ether anesthesia.  Organ weights (liver, kidneys,
44   adrenals, thymus and spleen) were recorded and hematology,  serum biochemistry and
45   histopathology of liver and kidneys were evaluated.
46
                                                28       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2
 3
 4
 5
 6
 7
       Hematological evaluation demonstrated an increase in the number of leukocytes in rats in
the highest dose group (38% increase).  Serum biochemistry showed significantly increased
mean levels of ALP and lactate dehydrogenase (LDH) in both treatment groups (84 and 100%
increase in LDH, 34 and 33% increase in ALP in low and high doses, respectively) and no
changes in mean levels of ALT or AST.  A dose-dependent increase in relative liver weights was
observed (17 and 26% at the low and high doses, respectively).  Table 4-3 shows the mean
enzyme levels and relative liver weights.
Table 4-3. Serum Enzyme Levels and Relative Liver Weights (Mean + Standard Deviation) in
Rats Ingesting MCLR in Drinking Water (Heinze, 1999)
Parameter
Relative liver weight (g/100 g body weight)
Lactate dehydrogenase (microkatals/L)
Alkaline phosphatase (microkatals/L)
Control
n=10
2.75 + 0.29
16.64 + 4.48
9.67 + 2.20
50 ng/kg
n=10
3.22 + 0.34*
30.64 + 5.05*
13.00 + 3.81*
150ng/kg
n=10
3.47 + 0.49*
33.58+1.16*
12.86+1.85*
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
* p<0.05 when compared with control
       The authors also reported a dose-dependent increase in absolute liver weights, although
the data were not provided. No statistically significant changes in other organ weights or body
weights were observed. In treated animals, histopathological alterations in the liver were
classified as toxic hepatosis. The incidence of liver lesions is summarized in Table 4-4.  Lesions
were spread diffusely throughout the parenchyma and included increased cell volume, increased
mitochondria, cell necrosis, activation of Kupffer cells and increased amounts of periodic acid-
Schiff (PAS)-positive substances, indicating cell damage.  Liver lesions were observed in both
treatment groups, but the severity of the damage was increased in the 150 ng/kg dose group. No
effects on the kidneys were observed.  A NOAEL could not be determined from this study. The
lowest dose tested (50 ng/kg) represents a LOAEL based on liver lesions, increased relative liver
weights and changes in liver enzymes  (ALP and LDH).

       4.2.1.2.2.  Cyanobacterial Extracts

       Davidson (1959) treated groups of three mice with drinking water or feed mixed with
different extracts (crude, aqueous and  filtered) from aNostoc rivulare water bloom for 21 days.
The only effects reported were ruffled hair and nervousness in the mice treated with crude
extract. Kalbe (1984) observed no change  in body weight among juvenile mice and rats given
filtrates from two different water blooms of M. aeruginosa for 2-8 weeks.
                                               29
                                                    DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-4. Incidence of Liver Lesions in Rats Ingesting MCLR in Drinking Water for 28
Days (Heinze, 1999)
Liver Histopathology
Control
n=10
50 |ig/kg
n=10
150 ng/kg
n=10
Degenerative and Necrotic Hepatocytes with Hemorrhage
Slight
Moderate
Intensive damage
0
0
0
4
6
0
0
6
3
Degenerative and Necrotic Hepatocytes without Hemorrhage
Slight
Moderate
Intensive damage
0
0
0
0
0
0
0
1
0
PAS-positive Material
Slight
Moderate
Intensive damage
1
0
0
5
5
0
0
8
2
Activation of Kupffer Cells
Slight
Moderate
Intensive damage
0
0
0
0
10
0
0
10
0
Lipid Granules and Droplets
Slight
Moderate
Intensive damage
0
1
0
4
2
0
0
1
0
30
DRAFT - DO NOT CITE OR QUOTE

-------
 1          Orr et al. (2003) exposed yearling beef cattle to live cultures of M aeruginosa in
 2   drinking water in an effort to evaluate whether microcystins accumulated in the liver or blood of
 3   the animals. Four steers were treated for 28 days, and four untreated steers served as controls.
 4   No effects on body weight, weight gain, food or water consumption or plasma enzymes (GOT,
 5   glyceraldehydes dehydrogenase, AST or bilirubin) were observed. The authors reported no
 6   detectable microcystins (by HPLC and gas chromatography-mass spectrometry) in either plasma
 7   samples collected throughout treatment or in samples of the liver collected upon sacrifice at the
 8   termination of exposure. Analysis of the liver samples by ELISA showed measurable
 9   microcystins; however, the authors indicated that these results likely represented cross-reaction
10   with something besides microcystins, given the failure  of the more sensitive HPLC analysis to
11   detect microcystins.
12
13          Schaeffer et al. (1999) reported the results of an unpublished 1984 study in which
14   Aphanizomenon flos-aquae, a cyanobacterium consumed as a food supplement, was fed to mice
15   in the diet.  The authors used recent analysis of the A. flos-aquae, which often coexists with
16   Microcystis species, to estimate the microcystin content in the material consumed by the mice.
17   Analysis of i]\Q A. flos-aquae samples used in the feeding study showed an average concentration
18   of 20+5 |j,g MCLR per gram of A. flos-aquae.  The authors estimated the daily exposure of
19   MCLR in the exposed mice to range from 43.3 ng/kg body weight per day to  333.3 |j,g/kg-day.
20   No clinical signs of toxicity were reported, and no effects on mortality, body weight, organ
21   weights or histology were observed  in the treated mice. In addition, no effects on reproductive
22   parameters were reported in five treated mice (highest dose group) allowed to breed.
23
24          4.2.1.3.  Subchronic Studies
25
26          4.2.1.3.1. Purified Microcystins
27
28          Fawell et al. (1999) conducted acute, subchronic and developmental toxicity studies of
29   MCLR given via gavage to Crl :CD-1(ICR)BR (VAF plus) mice (age not specified). MCLR was
30   obtained commercially and administered in distilled water. The concentration in the dosing
31   solution was verified by HPLC with UV detection.  Daily oral gavage doses of 0, 40, 200 or
32   1000 |J,g/kg body weight were given to groups of 15 male and 15 female mice for 13 weeks.
33   Daily clinical observations were made, body weight and food consumption were recorded
34   weekly, and eye examinations were  conducted prior to  and at the conclusion of treatment.
35   Hematology and serum biochemistry were evaluated for seven mice of each treatment group
36   during the final week of treatment. Upon sacrifice after 13 weeks, gross examination of organs
37   and microscopic evaluation of tissues were performed.  All tissues were examined in the control
38   and high dose animals, while only lungs, liver and kidney were examined in the other treated
39   animals.
40
41          Mean body weight gain was  decreased approximately 15% in all treated male groups.
42   Mean terminal body weights differed from controls by  about 7% in these groups. No dose-
43   related trends were evident for body weight gain or body weight in males. The only body weight
44   change observed in females was an increase in body weight gain in the 200 |j,g/kg-day group.
45   Hematological evaluation showed slight (10-12%) decreases in mean hemoglobin concentration,
46   red blood cell count and packed cell volume among females receiving 1000 ng/kg body weight.


                                                31        DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2
 3
 4
 5
 6
ALP, ALT and AST levels were significantly elevated (2- to 6-fold higher) in the high-dose
males, and ALP and ALT were likewise elevated (2- and 6-fold higher, respectively) in high
dose females.  ALT and AST were also elevated (2-fold) in the mid-dose males. GGT was
slightly decreased in some treatment groups.  Serum albumin and protein were reduced (13%) in
males of the mid- and high-dose groups. Table 4-5  shows the blood chemistry results.
Table 4-5. Blood Chemistry Results (Mean + Standard Deviation) for Mice Treated with MCLR for 13 Weeks
(Fawelletal., 1999)
MCLR Dose
(|ag/kg-day)
Alkaline
Phosphatase
(ALP) (U/l)
Alanine
Aminotransferase
(ALT) (U/l)
Aspartate
Aminotransferase
(AST) (U/l)
Gamma
Glutamyl
Transaminase
(GGT) (U/l)
Total
Protein
(g%)
Albumin
(g%)
Male
Control
40
200
1000
91+22.2
95 + 29.2
94 + 32.3
232b+ 103.2
27 + 8.0
37+17.2
59a+28.0
159c+75
68 + 27.7
64+12.2
121b+43.7
121b+26.3
6+1.0
4 + 0.7
3C + 0.4
4 + 0.4
5.5 + 0.32
5.1+0.26
4.8b + 0.29
4.8C + 0.21
3.2 + 0.19
3.0 + 0.13
2.8C + 0.13
2.8C + 0.11
Female
Control
40
200
1000
167 + 24.6
187 + 76.2
156 + 33.4
339b+ 123.7
32+11.3
25 + 7.8
27 + 9.4
220b+ 149.1
101+38.3
74+13.2
74 + 22.1
144 + 71.7
4+1.0
3+0.5
3+0.0
3+0.4
5.1+0.30
5.2 + 0.28
5.3+0.31
5.1+0.22
3.1+0.14
3.2 + 0.16
3.4a+0.14
3.1+0.18
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
a Significantly different from controls at p<0.05
b Significantly different from controls at pO.Ol
0 Significantly different from controls at p<0.001
       Histopathological changes in the liver were reported in the males and females of the mid-
and high-dose groups, with a dose-related increase in incidence and intensity. The liver lesions
were multifocal and observed throughout the liver lobule.  Table 4-6 summarizes the incidence
of liver histopathological changes.  Sex-related differences in liver pathology were not apparent.
No lesions were found in other tissues.

       The authors characterized the 40 ng/kg body weight dose as a clear NOAEL and
indicated that histopathological changes observed in the 200 ng/kg dose group were not severe.
The mid dose (200 |j,g/kg-day) represents a LOAEL based on the liver histopathological changes
and statistically significant blood chemistry changes. The WHO (1999) used the NOAEL value
of 40 |j,g/kg-day from this study as the basis for its provisional Tolerable Daily Intake for
MCLR.
                                                 32
                                                     DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-6. Incidence of Liver Histopathology in Mice Treated with MCLR for 13 Weeks
(Fawell et al., 1999)
Liver Histopathology
Male
Acute inflammation
Chronic inflammation
Congestion
Hepatocyte vacuolation
Hemosiderin deposits
Hepatocyte degeneration
Female
Autolysis
Chronic inflammation
Congestion
Hepatocyte vacuolation
Hemosiderin deposits
Hepatocyte degeneration
Control
n=15
0
1
3
5
0
0
n=15
0
5
0
5
0
0
40 |j,g/kg-day
n=15
1
2
0
5
0
0
n=15
0
8
0
5
0
0
200 ng/kg-day
n=15
0
4
0
6
0
1
n=15
0
8
0
11
1
1
1000 ng/kg-day
n=15
0
15
1
O
15
14
n=15
1
14
1
8
14
9
33
DRAFT - DO NOT CITE OR QUOTE

-------
 1          4.2.1.3.2.  Cyanobacterial Extracts
 2
 3          Falconer et al. (1994) administered dried bloom materials in the drinking water of pigs
 4   for 44 days. Plasma samples collected over 56 days showed dose- and time-dependent increases
 5   in GOT, ALP and total bilirubin, as well as a decrease in plasma albumin. Dose-related changes
 6   in the incidence and severity of histopathological changes of the liver were also observed,
 7   including cytoplasmic degeneration, hepatic cord disruption, single cell necrosis, periacinar
 8   degeneration, congestion and Kupffer cell proliferation.
 9
10          4.2.1.4.  Chronic Studies
11
12          4.2.1.4.1. Purified Microcystins
13
14          Ueno et al. (1999) evaluated the toxicity of MCLR in mice chronically exposed via
15   drinking water.  Two hundred 6-week-old female BALB/c mice were randomly assigned to
16   receive either no treatment or drinking water (ad libitum) containing 20 [ig/L MCLR for 7
17   days/week.  The MCLR had been isolated from lyophilized algal bloom materials from Lake
18   Suwa in Nagano, Japan and had been characterized as 95% pure by HPLC. Twenty animals
19   from each group were sacrificed at 3, 6 and 12 months, while the remaining 40 animals were
20   retained for chronic toxicity evaluation and sacrificed at 18 months.
21
22          Weekly  estimates of food and water consumption and daily observations for clinical
23   signs of toxicity, morbidity and mortality were recorded. Body weights were recorded weekly
24   for the first 2 months, biweekly up until the first year  and monthly until sacrifice. At 3, 6, 12 and
25   18 months, blood was obtained from 20 animals from each group. Samples from 10 animals per
26   group were used for hematological evaluation, and samples from 10 additional animals were
27   used for serum biochemistry evaluation. At each scheduled sacrifice time, complete necropsy of
28   10 animals per group was conducted.  Animals from the chronic toxicity group were necropsied
29   when moribund or dead (prior to scheduled sacrifice)  or upon sacrifice at 18 months. Relative
30   and absolute organ weights  (liver, kidneys, spleen, thymus, adrenal, ovaries, brain, heart and
31   uterus) were recorded for 9-10 animals per group at each scheduled sacrifice, and
32   histopathological evaluation of these and numerous other organs was conducted. Finally, three
33   to five animals per group were subjected to immunohistochemistry of the liver upon sacrifice to
34   determine the distribution of MCLR in the liver.
35
36          Based on weekly estimates of water consumption, the authors calculated the average total
37   intake of MCLR over 18 months to be 35.5 |j,g/mouse. No clinical signs of toxicity were
38   observed in either of the groups, and survival in the control and chronic treatment groups was
39   similar. No statistically significant differences in body weight, food consumption, water
40   consumption or hematology were observed; however, hematology data from the 3-month
41   sacrifice were lost due to sampling errors.  Treated mice were reported to have a statistically
42   significant decrease in ALP at month 12 (13%) and a  significant increase in cholesterol at month
43   18 (22%). Neither effect was considered by the authors to be lexicologically significant in the
44   absence of other treatment-related effects; however, the increase in cholesterol could be related
45   to the interaction of MCLR with bile acid transport in the liver.
                                                34       DRAFT - DO NOT CITE OR QUOTE

-------
 1          A decrease in heart weight among treated mice sacrificed at month 12 was not considered
 2   treatment-related in the absence of histopathological changes.  Sporadic changes in absolute and
 3   relative thymus weight in treated mice were observed, but histological and morphometric
 4   evaluation of the thymus revealed no abnormalities attributable to exposure. In contrast to other
 5   studies, the authors observed no difference in the incidence of liver histopathology between
 6   treated and control mice.  Immunohistochemistry of the liver revealed no accumulation of
 7   MCLR. This study identifies a free-standing NOAEL of 2.7 |j,g/kg-day in female mice
 8   (calculated assuming a 24.5 g body weight and an exposure duration of 548 days).
 9
10          Ito et al. (1997b) evaluated the carcinogenicity and liver toxicity of chronic gavage doses
11   of MCLR.  A water bloom from Lake Suwa, Japan served as the source of the MCLR, which
12   was isolated and dissolved in ethanol and saline for dosing. The purity of the isolated MCLR
13   was not specified. Twenty-two ICR mice (13 weeks old) were given either 80 or 100 gavage
14   doses of 80 ng/kg MCLR over the course of 28 weeks.  Ten mice were sacrificed after 80
15   treatments, five were sacrificed after 100 treatments and seven were withdrawn from treatment
16   after 100 doses and sacrificed 2 months later. Three mice served as untreated control. Although
17   the authors did not specify the nature of the postmortem examinations, it appears that the liver
18   was the only organ examined. No change in mean liver weight was observed in the MCLR-
19   treated animals compared with controls. The authors reported "light" injuries to hepatocytes in
20   the vicinity of the central vein in 8 of 15 mice sacrificed immediately after treatment, and in 5 of
21   7 mice that were withdrawn from treatment for 2 months after exposure.  No fibrous changes or
22   neoplastic nodules were observed.  Analysis for MCLR and its metabolites by
23   immunohistochemistry failed to detect either the parent compound or any metabolites in the
24   livers of mice  sacrificed immediately after treatment.
25
26          Thiel (1994) briefly reported the results of a chronic toxicity study of MCLA in vervet
27   monkeys.  The report is a brief summary published in the proceedings of an international
28   workshop; a published version of this study was not located. According to the summary, three
29   monkeys were given increasing intragastric doses of MCLA for 47 weeks, while three other
30   monkeys served as controls.  Doses increased from 20 ng/kg at the commencement of the study
31   to 80 ng/kg at study termination.  The rate of dosage increase was not reported. Monthly
32   measures of body weight and clinical signs (respiration, pulse, temperature) showed no effect of
33   treatment. Blood was withdrawn monthly; hematological  parameters examined were hematocrit,
34   bilirubin, hemoglobin, erythrocyte and leukocyte count and platelet count. No statistically
35   significant changes in hematological parameters were observed. No changes were observed in
36   serum biochemistry  analyses (albumin, globulins and electrolytes, as well as AST,  LDH, ALP,
37   ALT and GGT). Histopathological examination of the liver and other organs (not specified) did
38   not show any differences in treated monkeys when compared with controls.
39
40          4.2.1.4.2.  Cyanobacterial Extracts
41
42          Falconer et al.  (1988) conducted a chronic exposure experiment using an extract of aM
43   aeruginosa water bloom in Swiss Albino mice.  A concentration-dependent increase in mortality,
44   reduced body weight and  a concentration-dependent increase in ALT levels were observed
45   among groups of mice receiving serial dilutions of the extract as their drinking water for a year.
46   There was some evidence that bronchopneumonia incidence was related to concentration of


                                                3 5       DRAFT - DO NOT CITE OR QUOTE

-------
 1   extract. No significant differences in liver histopathology were observed, although the observed
 2   liver changes were slightly more prevalent in treated animals.  The data showed some indication
 3   of sex differences in susceptibility; male mice showed effects (including mortality and enzyme
 4   level increases) at lower concentrations than females.
 5
 6          4.2.1.5. Initiation/Promotion  Studies - Cyanobacterial Extracts
 7
 8          Falconer (1991) and Falconer and Buckley (1989) reported evidence of skin tumor
 9   promotion by extracts of Microcystis. Microcystis extract was administered via drinking water
10   to mice pretreated topically with an initiating dose of dimethylbenzanthracene (DMB A). The
11   total skin tumor weight in mice drinking Microcystis extract was significantly higher than that of
12   mice receiving only water after initiation.  The number of tumors per mouse was only slightly
13   increased in mice receiving extract; the weight difference was largely due to the weight of
14   individual tumors (Falconer and Buckley, 1989). The total weight of tumors in this group also
15   exceeded that of mice pretreated with DMBA and subsequently treated with topical croton oil,
16   with or without concurrent consumption of Microcystis extract. Details of the tumor incidence
17   in the mice were not provided by the authors. When Microcystis extract was provided in the
18   drinking water of mice pretreated with two oral doses of N-methyl-N-nitroso-urea, no evidence
19   of promotion of lymphoid or duodenal  adenomas and adenocarcinomas was observed.  No
20   primary liver tumors were observed (Falconer and Humpage, 1996).
21
22          Humpage et al. (2000) administered M. aeruginosa extract in drinking water to mice
23   pretreated with azoxymethane. Mice were sacrificed at intervals up to 31 weeks after
24   commencement of extract exposure.  Enzyme analysis showed a concentration-dependent
25   increase in ALP and decrease in albumin in mice treated with extract. The authors observed a
26   concentration-dependent increase in the mean area of aberrant crypt foci of the colon, although
27   the number of foci per colon and the number of crypts per focus were not different among the
28   groups. The authors proposed that increased cell proliferation caused the increase in size of foci.
29   Histological examination of the livers of mice treated with extract showed more leukocyte
30   infiltration in animals treated with the highest concentration of extract compared to those
31   receiving a low concentration.
32
33   4.2.2.  Inhalation Exposure
34
35          All available studies of inhalation exposure used MCLR, and there were no inhalation
36   studies using cyanobacterial  extracts.
37
38          4.2.2.1. Acute Studies
39
40          Fitzgeorge et al. (1994) conducted experiments in CBA/BALBc mice with MCLR
41   administered via intranasal instillation  and inhalation.  This study is poorly  described, giving few
42   details of study design and findings.  A single experiment with mice (number unspecified)
43   inhaling a fine aerosol (particle size 3-5 |j,m) with 50 |j,g/L MCLR for an unspecified duration of
44   time did not result in any deaths, clinical signs of toxicity or histopathological changes. The
45   nature of the examinations was not reported. The authors estimated the delivered dose of MCLR
46   to be very small (about 0.0005 ng/kg).


                                                36       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2          A brief abstract describes a study of acute MCLR exposure via inhalation (Creasia,
 3   1990). Details of study design and results were not reported. The LCso for mice exposed to a
 4   MCLR aerosol (nose only) for 10 minutes was reported to be 18 |J,g/L (mg/m3) air with a 95%
 5   confidence interval of 15.0-22.0 |j,g/L (mg/m3). Based on studies of lung deposition after
 6   exposure of mice to the LCso concentration, an LD50 of 43 ng/kg body weight was estimated.
 7   The authors reported that histological lesions in mice killed by aerosol exposure were similar to
 8   those in mice dosed intravenously with MCLR.
 9
10          4.2.2.2. Short-Term Studies
11
12          Benson et al. (2005) exposed groups of six male BALB/c mice to monodisperse
13   submicron aerosols of MCLR via nose-only inhalation for 30, 60 or 120 minutes each day for 7
14   consecutive days.  The concentration of MCLR was 260-265 ng/m3 and  doses deposited in the
15   respiratory tract were estimated to be 3, 6 and 12.5 |J,g/kg body weight. Control mice were
16   exposed  to the aerosolized vehicle (20% ethanol in water). Clinical signs were recorded daily.
17   Sacrifice by injection of Euthasol occurred the day after the last exposure.  Blood was collected
18   by cardiac puncture and serum was subjected to clinical chemistry analysis (blood urea nitrogen
19   [BUN], creatinine, total bilirubin, ALP, AST,  ALT, total protein, albumin and globulin).  Organ
20   weights (adrenals,  lung, liver, kidney, spleen and thymus) were recorded and histopathological
21   examinations of the liver, respiratory tract tissues, adrenals, kidney, spleen, thymus,
22   gastrointestinal tract and testes were conducted.
23
24          No clinical signs or effects on body weight or organ weights were observed following
25   exposure to MCLR aerosol. Histopathological examination revealed treatment-related lesions in
26   the nasal cavity only.  Lesions were not observed in the liver, other organs or in other parts of the
27   respiratory tract. As indicated in Table 4-7, the incidence and severity of nasal lesions increased
28   with length of the daily exposure period.  The lesions consisted primarily of necrosis or
29   inflammation of respiratory epithelial cells and degeneration, necrosis and atrophy of olfactory
30   epithelial cells. Necrotic lesions of olfactory epithelial cells were generally larger patches, while
31   few cells were involved in respiratory epithelial cell necrosis.
32
33          4.2.2.3. Subchronic and Chronic Studies
34
35          No subchronic or chronic animal studies evaluating the inhalation route of exposure were
36   identified in the materials reviewed for this document.
                                                 37       DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-7. Incidence and Severity of Nasal Cavity Lesions in Mice Inhaling Microcystin
Aerosol for 7 Days (Benson et al., 2005)
Lesion
Severity
Control
Daily Exposure Period (minutes)
30
60
120
Respiratory Epithelial Necrosis
Turbinate 1
Turbinate 2
Minimal
Mild
Moderate
Mild
Moderate
0/6
0/6
0/6
0/6
0/6
1/6
0/6
0/6
0/6
0/6
0/6
6/6
0/6
6/6
0/6
0/6
0/6
2/6
3/6
3/6
Respiratory Epithelial Inflammation
Turbinate 1
Turbinate 2
Mild
Mild
0/6
0/6
1/6
1/6
0/6
0/6
1/6
0/6
Olfactory Epithelial Degeneration, Necrosis and Atrophy
Turbinate 1
Turbinate 2
Turbinate 3
Mild
Moderate
Mild
Moderate
Mild
Moderate
Marked
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
6/6
0/6
6/6
0/6
0/6
4/6
1/6
0/6
6/6
0/6
4/6
2/6
38
DRAFT - DO NOT CITE OR QUOTE

-------
 1   4.3.    REPRODUCTIVE/DEVELOPMENTAL STUDIES - ORAL AND INHALATION
 2
 3   4.3.1.  Oral Exposure
 4
 5          4.3.1.1.  Purified Microcystins
 6
 7          Fawell et al. (1999) conducted acute, subchronic and developmental toxicity studies of
 8   MCLR given via gavage to Crl :CD-1(ICR)BR (VAF plus) mice (age not specified).  MCLR (0,
 9   200, 600 or 2000 ng/kg) was administered to groups of 26 mice on days 6-15 of pregnancy. The
10   mice were sacrificed on day 18 and necropsied.  Weight and sex of the fetuses were recorded,
11   and external, visceral and skeletal examinations performed.  Seven of 26 dams receiving 2000
12   Hg/kg died and 2 others were sacrificed prematurely due to morbidity.  Altered liver appearance
13   was noted during gross examination of these animals.  Surviving dams in this group did not
14   display any clinical signs of toxicity or differences in body weight or food consumption. The
15   authors reported that fetal body weight was significantly lower than controls and there was
16   delayed skeletal ossification at the highest dose; however, the data were not presented in the
17   publication.  These effects may be associated with maternal toxicity. According to the authors,
18   no effects on resorption or litter size were observed, nor were there increases in external, visceral
19   or skeletal abnormalities in fetuses of any treatment group. Data on reproductive  and
20   developmental parameters were not provided. Based on the authors' description of the findings,
21   it seems evident that the 600 |j,g/kg-day dose represents a NOAEL for both maternal and
22   developmental toxicity in mice. Further, the deaths of seven dams receiving 2000 |j,g/kg-day
23   clearly identify this dose as a maternal PEL. However, in the absence of the data  showing
24   incidences of reproductive and developmental parameters, it is difficult to determine
25   conclusively whether the high dose also represents a LOAEL for developmental toxicity.
26
27          4.3.1.2.  Cyanobacterial Extracts
28
29          Falconer et al. (1988) conducted a limited study of reproductive effects using an extract
30   from anM aeruginosa bloom sample. Eight female mice that had been given l/4th dilution of
31   the extract as drinking water (estimated to contain 14 |j,g/mL of unspecified microcystin toxin)
32   since weaning were mated with similarly treated males. No difference in number of litters, pups
33   per litter, sex ratio or litter weight were observed.  Reduced brain size was reported to occur in 7
34   of 73 pups from treated parents and in none of 67 pups from controls. The litter distribution of
35   the affected pups was not reported by the authors.  One of the small brains was examined
36   histologically, revealing extensive damage to the hippocampus.
37
38   4.3.2.  Inhalation Exposure
39
40          No reports of developmental or reproductive toxicity by the inhalation route of exposure
41   were identified in the materials reviewed for this document.
42
                                                39       DRAFT - DO NOT CITE OR QUOTE

-------
 1   4.4.    OTHER STUDIES
 2
 3   4.4.1.  Neurological Effects
 4
 5          Neurological effects have been reported to occur with acute lethal doses of MCLR.
 6   Clinical signs such as hypoactivity and piloerection have been observed in mice and rats exposed
 7   to lethal doses of MCLR by oral or i.p. administration (Hooser et al., 1989a; Fawell et al., 1999).
 8   These signs were observed in animals that subsequently succumbed to the lethal exposures. No
 9   reports of neurotoxicity by the inhalation route of exposure were identified in the materials
10   reviewed for this document.
11
12          Maidana et al. (2006) reported that long-term memory retrieval (as assessed by step-
13   down inhibitory avoidance task) was impaired in rats receiving intrahippocampal injection of
14   0.01 or 20 |J,g/L of a microcystin extract from Microcystis strain RST 9501.  Exposure at 0.01
15   |j,g/L also impaired spatial learning (as assessed by performance on the radial arm maze), but
16   exposure at the higher concentration did not.  The authors indicated that the primary microcystin
17   produced by this strain is [D-Leu1] MCLR, a variant of MCLR.  Oxidative damage, as measured
18   by lipid peroxides and DNA damage, was increased in tissue homogenates of the hippocampus
19   from treated animals.
20
21          Foxall and Sasner (1981) conducted limited in vitro studies on the neurological effects of
22   a crude extract from a bloom of M. aeruginosa. Little detail  on experiment design was reported;
23   frog and mouse heart, frog sartorius muscle, frog sciatic nerve and mouse ileum were used in the
24   experiments. The authors reported that the extract had no effect on electrical or mechanical
25   events.
26
27   4.4.2.  Immunological Effects
28
29          Evaluation  of the immunotoxicity of microcystins in vivo was reported in only two
30   studies, both of which used a cyanobacterial cell extract rather than purified microcystins.  Shirai
31   et al. (1986) reported that mice, immunized i.p.  with either live or sonicated cells from a
32   Microcystis water bloom, developed delayed-type hypersensitivity when challenged 2 weeks
33   later with a subcutaneous injection of sonicated cells.  Delayed hypersensitivity was assessed by
34   footpad swelling, which was increased approximately 2-fold over controls at the highest doses of
35   cells. It is not clear whether an endotoxin in the bloom sample was responsible for the
36   development of hypersensitivity, or whether the antigenic epitope existed on other components
37   of the sample.
38
39          Shen et  al. (2003) assessed the effect of cyanobacterial cell extract on immune function.
40   Mice received 14 daily i.p. injections containing a cell-free extract from a water bloom
41   dominated by M. aeruginosa. Doses were reported as 16, 32 and 64 mg lyophilized cells/kg
42   body weight or as 4.97, 9.94 and 19.88 ng/kg microcystin equivalents. HPLC analysis indicated
43   that  the microcystin content of the extract was 79.53%, although specific congeners in the extract
44   were not reported.  The following immunotoxicity  endpoints were examined: phagocytosis,
45   lymphocyte proliferation and antibody production in response to sheep red blood cells.
46   Phagocytic capacity was reduced at the two highest doses,  but percentage phagocytosis was not


                                                40       DRAFT - DO NOT CITE OR QUOTE

-------
 1   affected. B-lymphocyte proliferation was significantly reduced (33% compared to controls at 32
 2   mg/kg), while changes in T-lymphocyte proliferation were mild, and deemed biologically
 3   insignificant. Finally, humoral immune response, as measured by antibody-forming plaques,
 4   was reduced in a dose-dependent manner in treated mice.
 5
 6          Chen et al. (2004, 2005) evaluated the role of nitric oxide generation and macrophage
 7   related cytokines on the reduced phagocytic capacity induced by pure MCLR. A dose-dependent
 8   inhibition of nitric oxide production was observed in activated macrophages, and a repressive
 9   effect was seen in cytokine formation at the mRNA level (e.g., IL-lp, TNF-a, GM-CSF, IFN-y)
10   after either 24 hour (Chen et al., 2004) or 6 hour treatment (Chen et al., 2005). Hernandez et al.
11   (2000) indicated that MCLR enhanced the early spontaneous polymorphonuclear leukocyte
12   (PMN) adherence (not late or PMN stimulated early or late) at low concentrations, suggesting
13   that microcystins may affect the immune system.
14
15   4.4.3.  Hematological Effects
16
17          Several studies have noted thrombocytopenia in laboratory animals treated with
18   microcystins or bloom extracts purportedly containing microcystins (Slatkin et al., 1983; Adams
19   et al., 1985, 1988; Takahashi et al., 1995). Early investigations explored whether microcystins
20   had a direct effect on platelets, and whether platelets might be responsible for pulmonary
21   thrombi (Slatkin et al., 1983; Jones, 1984). However, in vitro studies have shown that MCLR
22   neither induces nor impedes the aggregation of platelets (Adams et al., 1985). Pulmonary
23   thrombi apparently consist of necrotic hepatocytes circulating in the blood (see Section
24   4.4.5.2.1).  More recent information supports the hypothesis that hematological effects observed
25   in animals acutely exposed to microcystins are secondary  effects of liver hemorrhage (Takahashi
26   etal., 1995).
27
28          Takahashi et al. (1995) reported  dose-dependent reductions in erythrocyte count,
29   leukocyte count, hemoglobin concentration, hematocrit and coagulation parameters 1 hour after
30   rats were exposed to MCLR (100 and 200 ng/kg i.p).  None  of these parameters changed until
31   after massive liver hemorrhage commenced.  Further, hematological changes such as increased
32   prothrombin time and fibrin deposition in the renal glomeruli were not observed. The authors
33   concluded that the depletion of blood components occurred as a result of liver hemorrhage.
34
35          Interestingly,  mild thrombocytopenia was reported in 1-week-old mice treated with a
36   large i.p. dose of MCLR, even though none of these mice  died (Adams et al., 1985).
37
38   4.4.4.  Effects by Dermal Exposure
39
40          No  animal studies evaluating  the effects in animals of dermal exposure to purified
41   microcystins were identified in the materials reviewed for this document;  only one study using a
42   cyanobacterial extract was located. Davidson (1959) applied a crude extract from a bloom of TV.
43   rivulare to  the shaved backs of three  mice every 2 hours for  a total of 12 hours. Heavy scales
44   were observed on the treated areas; the scales were gone within 4 days and hair regrowth
45   occurred in the following weeks.  Dermal application of an aqueous extract or aqueous filtrate
46   did not result in any effects.
                                                41       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2   4.4.5.  Effects by Parenteral Exposure
 3
 4          4.4.5.1.  Effects in Humans after Parenteral Exposure
 5
 6          In February of 1996, unfinished water from a reservoir with a cyanob acted al bloom was
 7   used at a hemodialysis center in Caruaru Brazil, leading to numerous deaths among the patients
 8   treated with the water (Jochimsen et al., 1998; Pouria et al., 1998; Carmichael et al., 2001;
 9   Azevedo et al., 2002). The first report of this incident was published by Jochimsen et al. (1998),
10   and Pouria et al. (1998) published follow-up information on the status of the patients. Azevedo
11   et al. (2002) provided the most up-to-date information on patient status after the incident. Using
12   water samples from the reservoir, serum and tissue samples from patients, and a variety of
13   methods, each publication identified microcystins as the primary causative factor in the deaths
14   and disease among patients. Carmichael et al. (2001) used analytical results from tissue samples
15   with dialysate volumes to estimate the concentration of microcystins in the water to which
16   patients were exposed, as a direct measure of exposure concentration was not available.
17
18          Of the 131 patients undergoing dialysis during the 4 days when unfinished water was
19   supplied to the center, 116 experienced symptoms, including visual disturbances, nausea,
20   vomiting and muscle weakness; 100 developed acute liver failure; and 52 had died as of
21   December, 1996 (Azevedo et al., 2002). The acute presentation of the disease included malaise,
22   weakness, dizziness,  vertigo, visual disturbances and blindness, nausea, vomiting and abdominal
23   pain.  Clinical signs included hepatomegaly and jaundice. Biochemistry showed high
24   concentrations of bilirubin and ALP, moderate increases in AST and ALT, hypoglycemia,
25   hypoalbuminemia and severe hypertrigliceridemia. Major hematology findings were slightly
26   low platelet count (within normal range) and reduced platelet aggregation, as well as red blood
27   cell abnormalities (anisocytosis, acanthocytosis  and schistocytosis) in some patients. Liver
28   biopsy and autopsy samples showed severe, diffuse individual hepatocyte necrosis throughout
29   the liver lobule, with cell-plate disruption and apoptosis; however, no intrahepatic hemorrhage
30   was observed. Leukocyte infiltration and canalicular cholestasis were also observed.
31
32          Microcystins  were implicated  as the major contributing factor to patient death and
33   morbidity (Jochimsen et al., 1998; Pouria et al.,  1998; Carmichael et al., 2001; Azevedo et al.,
34   2002). Evidence for  the role of microcystins was derived from a variety of sources discussed by
35   Azevedo et al. (2002) and Carmichael et al. (2001). Quantitative analysis of the phytoplankton
36   in the reservoir from  which the water  was supplied to the dialysis center showed that
37   cyanobacteria represented about 99% of the phytoplankton in the reservoir,  although the species
38   present during the outbreak were not identified.  Analyses of filter systems in the dialysis center
39   showed microcystins; the carbon filter also had cylindrospermopsin.  Analyses of patient sera
40   and liver samples provided additional evidence. Patient sera were analyzed for other potential
41   toxins (chlorines, chloramines, trace elements, heavy metals, agricultural compounds and
42   pesticides), but none  were found (Pouria et al., 1998). Both serum and liver analyses for
43   microcystins revealed MCYR, MCLR and MCAR. None of the biological samples contained
44   cylindrospermopsin.  Finally, physiological effects observed in the patients closely  mirrored
45   effects observed in laboratory animals exposed to microcystins. Specifically, the liver damage
46   observed in patients was similar to that observed in mice (Jochimsen et al., 1998). Using
                                                42       DRAFT - DO NOT CITE OR QUOTE

-------
 1   microcystin concentrations measured in patients' livers and typical dialysis volume, Carmichael
 2   et al. (2001) estimated the average concentration of microcystins in the dialysate to be 19.5 |j,g
 3   microcystins/L.
 4
 5          Scares et al. (2006) reported another incident involving human exposure to microcystins
 6   via dialysate. In November, 2001, 44 patients of a renal dialysis center in Rio de Janeiro, Brazil,
 7   were assumed to be exposed to microcystins after a bloom of Microcystis andAnabaena
 8   occurred in the reservoir supplying water to the center. The concentration of microcystins in the
 9   drinking water was 0.4 [ig/L by ELISA.  In the water used at the center (after treatment by
10   activated carbon column), the concentration was 0.32 [ig/L. Of the 44 patients exposed, 90%
11   had serum microcystin concentrations above the limit of detection (0.16 ng/mL). Twelve of the
12   patients were selected for 2-month follow-up monitoring of serum levels. Over the follow-up
13   period, serum concentrations ranged from <0.16 to 0.96 ng/mL.  The highest serum
14   concentrations occurred 1  month after initial exposure. The authors did not provide any
15   information on health effects from the exposure.
16
17          Pilotto et al. (2004) reported that about 20% of 114 volunteers subjected to skin-patch
18   testing using cell suspensions and extracts of various cyanobacterial cultures (including two
19   strains of M aeruginosd) experienced a mild dermal skin reaction (erythema). The reaction did
20   not vary with cyanobacterial species.
21
22          4.4.5.2. Effects in Animals after Parenteral Exposure
23
24          4.4.5.2.1. Acute Studies with Parenteral Exposure
25
26          The acute toxicity of microcystins administered i.p. has been extensively studied.  The
27   primary target organ for acute microcystin toxicity is the liver; effects have also been observed
28   in the kidney, lungs and gastrointestinal tract.  Some of the effects observed in organs other than
29   the liver are believed to be secondary to liver effects.
30
31          A number of references report LDso estimates for injected MCLR (Slatkin et al., 1983;
32   LeClaire et al., 1988; Lovell et al., 1989a; Hermansky et al., 1990c; Miura et al., 1991; Stotts et
33   al., 1993; Gupta et al., 2003); two report LD50 values for MCRR and MCYR (Stotts et al.,  1993;
34   Gupta et al., 2003). Table 4-8 summarizes the available estimates of microcystin LD50 values
35   after injection exposure to purified microcystins.
36
37          As the table shows, the LDso for MCLR in mice ranges between 30 and 60  ng/kg.  The
38   acute lethality of MCYR is slightly lower than MCLR; LD50 estimates for MCYR were 111 and
39   171 ng/kg (Gupta et al., 2003 and Stotts  et al., 1993, respectively). The LD50 for MCRR is
40   higher still, with LDso values estimated as 235 ng/kg (Gupta et al., 2003) and 650 |J,g/kg (Stotts
41   et al., 1993).  In rats, the LDso for MCLR was similar to that in mice. There is some evidence
42   that
                                                43        DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-8. LD50 Values of Purified Microcystin Congeners by Intraperitoneal Administration
Sex/Strain
Purity
Vehicle
LD50 (95% CI)
Duration
Comments
Reference
MCLR-Mice
Male/Balb/C
Female/Swiss
albino Hale-
Stoner
Female/NIH
non- Swiss
Male/Swiss
Webster
Not specified/
CBA/Balbc
Female/Swiss
albino
Male/Swiss
albino
75%
NS
>95%
NA
NS
NS
NS
NS*
NS
Distilled
water
0.09% saline
NS
NS
Methanol and
PBS
32.6(±1.2)ng/kg
60 ng/kg
61 ng/kg
<100 ng/kg
250 ng/kg
43. 0|ig/kg (37.5-49.4)
43 ng/kg (37.5-49.4)
24 hours
NS
NS
NS
NS
24 hours
24 hours
Impurities tested for
toxicity at high doses
with negative results


LD50 by up and down
method

LD50 by up and down
method
LDso by up and down
method
Lovell et al.,
1989a
Slatkin et al.,
1983
Hermansky et
al., 1990c
Stotts et al.,
1993
Fitzgeorge et
al., 1994
Gupta et al.,
2003
Rao et al.,
2005
44
DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-8 cont.
Sex/Strain
Purity
Vehicle
LD50 (95% CI)
Duration
Comments
Reference
MCLR-Rats
Male/Fischer
Male/Fischer
344
NS
>95%
NS
saline
50 ng/kg (36-68)
Fed rats: 122 ng/kg (106-141)
Fasted rats: 72 ng/kg (60-83)
72 hours
25 hours
Intraarterial injection.
Abstract only

LeClaire et al.,
1988
Miura et al.,
1991
MCRR-Mice
Male/Swiss
Webster
Female/Swiss
albino
NS
NS
0.09% saline
NS
-650 ng/kg
235.40 ng/kg (202.3-272.8)
NS
24 hours
LD50 by up and down
method
LDso by up and down
method
Stotts et al.,
1993
Gupta et al.,
2003
MCYR-Mice
Male/Swiss
Webster
Female/Swiss
albino
NS
NS
0.09% saline
NS
-171 ng/kg
1 10.6 ng/kg (81.7-149.6)
NS
24 hours
LDso by up and down
method
LD50 by up and down
method
Stotts et al.,
1993
Gupta et al.,
2003
2   * Not specified.
                                                      45
DRAFT - DO NOT CITE OR QUOTE

-------
 1   the LD50 for MCLR was higher in fed rats (122 ng/kg) than in fasted rats (72 ng/kg) (Miura et
 2   al., 1991).
 3
 4          In general, death occurs quickly in animals receiving a lethal injected dose of MCLR.
 5   Mice typically die within 1-2 hours of a lethal i.p. dose of MCLR (Adams et al., 1988; Gupta et
 6   al., 2003). Mean time to death for mice exposed to a lethal dose of MCRR or MCYR is also
 7   within 2 hours (Gupta et al., 2003).  In mice, an i.p. dose of 100 ng/kg MCLR typically results in
 8   100% mortality (Adams et al., 1988; Hooser et al., 1989a; Hermansky et al., 1990c). Hooser et
 9   al. (1989a) compared the effects of MCLR in mice and rats and observed significant differences
10   in survival time; both male and female rats given less than 240 ng/kg survived between 20 and
11   32 hours; rats receiving higher doses died within 8 hours. In contrast, female mice receiving  100
12   Hg/kg died within 1.5 hours (Hooser et al., 1989a).  Miura et al. (1991) administered MCLR i.p.
13   to fed and fasted rats and reported a protective effect of feeding. Median time to death for rats
14   given 100 ng/kg MCLR was 32 hours in fed rats and less than 2 hours in fasted rats.
15
16          The sequence of events leading to death in laboratory rodents has been extensively
17   studied (Slatkin et al.,  1983; Adams et al., 1988; Hooser  et al., 1989a,b, 1990; Takahashi et al.,
18   1995). In general, similar effects have been reported in both rats and mice, but effects occur
19   later in rats than in mice (Hooser et al., 1989a). Gross and microscopic changes in swine treated
20   intravenously are similar to those observed in rodents (Lovell et al., 1989b).
21
22          Ten minutes after mice received a lethal i.p. dose, clinical signs, enzyme changes and
23   liver weight changes were generally absent (Slatkin et al., 1983; Adams et al., 1988). Beginning
24   approximately 20 minutes after dosing in mice, liver weights increased as the livers became
25   suffused with blood (Slatkin et al., 1983; Adams et al., 1988; Hermansky et al., 1990c). Adams
26   et al. (1988) estimated that as much as 44% of the total blood volume was located  in the liver 30
27   minutes after a lethal dose of MCLR in mice. Similarly,  in swine treated intravenously, blood
28   volume lost to the liver was estimated to be about 40% (Beasley et al., 2000). At 30 minutes,
29   there were isolated areas of hepatic necrosis, and at 45 minutes, there was marked  liver
30   congestion and widespread hepatic necrosis (Hermansky etal., 1990c).  Pulmonary thrombi
31   observed at the time of death were generally believed to result from necrotic hepatocytes (Adams
32   et al., 1988). Other pulmonary effects observed at 30-60 minutes after exposure included
33   congestion, bronchial epithelial hyperplasia and necrosis, edema and hemorrhage (Gupta et al.,
34   2003). In general, hepatic enzyme levels show little or no change until 20-30 minutes after
35   dosing, when hepatic hemorrhage is beginning (Adams et al., 1988; Hooser et al.,  1989a;
36   Hermansky et  al., 1990c; Takahashi et al., 1995).
37
38          At 60 minutes, there was severe disassociation of hepatocytes, hepatocyte loss and
39   hemorrhage, with disintegration of liver architecture (Hooser et al., 1991b; Guzman et al., 2003).
40   Thrombocytopenia coincided with hepatic hemorrhage as blood accumulated in the liver (Slatkin
41   et al., 1983; Adams et al., 1988; Takahashi et al., 1995).
42
43          The available studies demonstrate a very steep dose-response curve for MCLR acute
44   toxicity. In female NIH non-Swiss outbred mice, the only change observed after i.p.
45   administration of 50 |j,g MCLR/kg was Kupffer-cell hyperplasia, while all mice receiving 100
                                                46       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Hg/kg died (Hermansky et al., 1990c).  Hooser et al. (1989a) reported that male rats dosed i.p.
 2   with 20, 40 or 80 ng/kg and females dosed with 40 |J,g/kg MCLR showed no clinical signs of
 3   toxicity nor gross or microscopic lesions in the liver or other organs examined, while 120 ng/kg
 4   was lethal in some rats and 160 |J,g/kg was consistently lethal. Lovell et al. (1989a) administered
 5   a sublethal i.p. dose (about 25 ng/kg) of MCLR to male mice and reported a significant increase
 6   in liver weight (8.7%), but no clinical signs or hepatic lesions.
 7
 8          Induction of apoptosis is believed to be one mechanism for liver damage in acutely
 9   exposed animals (Hooser, 2000). Yoshida et al.  (1998) treated mice with single sublethal doses
10   of MCLR (20, 30 and 45  ng/kg i.p.) and observed them for 7 days.  Multiple apoptotic cells were
11   noted in the centrilobular regions of the livers of these mice. Hooser (2000) reported widespread
12   apoptosis in most hepatocytes after rats were treated with a single i.p. dose of 500 ng/kg MCLR.
13   Yoshida et al. (1998) reported the occurrence of two types of apoptotic hepatocytes in mice
14   given sublethal doses of MCLR; one that revealed MCLR by immunohistochemistry and one
15   that did not. The authors suggested that the latter type may contain MCLR that had lost the
16   antigenic epitope, or may have become apoptotic via other means, including ischemia or
17   hypoxia. Apoptosis induction is not restricted to the liver; Botha et al. (2004) reported
18   significantly increased apoptotic indices in the gastrointestinal tract of BALB/c mice as early as
19   8 hours after a single 75% LD50 dose (specific dose not reported) of MCLR i.p. The authors
20   observed immunohistochemical evidence of MCLR in the lamina propria and postulated that
21   MCLR was involved in the induction of apoptosis. The role of apoptosis in microcystin-induced
22   liver toxicity is further discussed in Section 4.4.7.5 (Mechanistic Studies).
23
24          Effects outside the liver have been reported after acute injection exposure to MCLR.  It
25   remains unclear whether such effects may be indirectly related to hepatotoxicity.  Some studies
26   have shown increases in kidney weight (Hooser et al., 1989a; Lovell et al., 1989a) or other signs
27   of kidney damage (LeClaire et al., 1988; Zhang et al., 2002) in rats and mice after injection of
28   sublethal doses of MCLR. Lovell et al. (1989a)  observed dilation of cortical tubules and
29   eosinophilic granular or fibrillar material in the cortical tubules after MCLR administration.  In
30   addition to reports  of kidney effects, there are scattered reports of cardiac effects, including
31   degeneration and necrosis of myocardial cells after i.p. or i.v. exposure to MCLR (LeClaire et
32   al., 1988; Zhang et al., 2002).  LeClaire et al. (1995) reported that mechanisms such as reflex
33   tachycardia and increased cardiac output, which  typically would allow the heart to compensate
34   for the acute hypotension caused by blood pooling in the liver, were impaired in rats given a
35   lethal dose of MCLR. The authors  suggested a cardiogenic component to the toxicity of MCLR.
36   Oishi and Watanabe (1986) observed tachycardia in mice 20 minutes after i.p. injection of
37   lyophilized cells fromM aeruginosa.  Finally, a few studies suggest that acute i.p. exposure to
38   MCLR can affect lipid peroxidation levels in both the intestinal mucosa (Moreno et al., 2003)
39   and liver (Towner et al., 2002) of rats.
40
41          Two injection studies support the finding by Ito et al. (1997a) by  oral exposure that
42   young animals are  not as  susceptible to the acute toxic effects of microcystins as older animals.
43   Adams et al. (1985) administered lethal doses of MCLR via i.p. injection to 1-, 2- and 3-week-
44   old mice.  None of the  1- or 2-week-old mice died, whereas 23 of 31 mice aged 3 weeks died
45   within 2 hours. The 3-week-old mice that survived were rechallenged with MCLR a week later,
                                                47        DRAFT - DO NOT CITE OR QUOTE

-------
 1   and all died.  Rao et al. (2005) reported that time to death decreased with age in mice treated i.p.
 2   with MCLR.
 3
 4          Guzman and Solter (2002) evaluated the acute effects of repeated injection of MCLR.
 5   Male BALB/c mice were injected with 45 ng/kg MCLR daily for 2, 4 or 7 days. Livers of mice
 6   receiving four or seven daily doses were pale and moderately enlarged, with an accentuated
 7   reticular pattern; absolute and relative liver weights were statistically increased over controls.
 8   Histopathology revealed apoptotic hepatocytes in the centrilobular region of mice receiving two
 9   doses, and marked hepatocytomegaly, disorganized hepatocytes, multinucleated hepatocytes and
10   cytoplasmic vacuolation in mice receiving four or seven doses.  Guzman et al. (2003) reported
11   immunostaining of some  centrilobular hepatocytes following two doses of 45 ng/kg; however,
12   protein phosphatase activity was not affected.
13
14          Acute toxicity of bloom extracts is highly variable, likely reflecting the variable toxin
15   content of algal blooms.  LDso values estimated for various bloom extracts range from  14 mg dry
16   weight cells/kg body weight to 1924 mg dry weight cells/kg (see Table 4-9). These studies used
17   a variety of test materials, including lyophilized cyanobacterial  cells, cell-free lysates, etc.
18
19          Early studies of purified but unidentified toxins firomM aeruginosa show essentially the
20   same pattern of acute hepatotoxicity and time to death after i.p.  injection of lethal doses in rats
21   and mice that is seen with purified MCLR (Elleman et al., 1978; Foxall and Sasner, 1981;
22   Falconer et al., 1981; Theiss et al., 1984, 1985, 1988; Jones and Carmichael, 1984; Siegelman et
23   al., 1984; Dabholkar and  Carmichael,  1987). In addition, many studies of cyanobacterial bloom
24   extracts (primarily M. aeruginosa) administered via i.p. injection to laboratory rodents show
25   similar effects (Ashworth and Mason, 1945; Ohtake et al., 1989; Rao et al.,  1994; Porfmo et al.,
26   1999; Sabour et al., 2002).
27
28          Jackson et al. (1984) also observed liver pathology in sheep exposed intraruminally to a
29   bloom sample identified asM aeruginosa.  Time to death ranged from 18 to 48 hours post
30   exposure. In animals that succumbed, the livers were hemorrhagic and necrotic.
31
32          Effects on organs other than the liver have been reported in some studies of bloom
33   extracts. Bhattacharya et al. (1997) observed changes suggesting distal tubular dysfunction
34   (proteinurea, as well as decreases in kidney  LDH and AST levels) in rats injected with LD50
35   doses of M aeruginosa extract, but no histopathological changes. Dose-dependent increases in
36   urea and creatinine and decreases in total protein and albumin were also observed. Picanco et al.
37   (2004) reported that i.p. injection of an extract from a culture of M aeruginosa (strain NPJB-1)
38   into either young or mature mice resulted in increased alveolar collapse and increased number of
39   polymorphonuclear and mononuclear cell infiltrations when compared with saline-treated
40   controls. There was a very low concentration of contaminating bacteria in the culture (i.e., the
41   culture was not axenic) used in this study, and the authors acknowledged that materials in the
42   extract other than microcystins may have contributed to the pulmonary effects.
43
                                                48       DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-9. Intraperitoneal LD50 Values for Bloom Extracts
Microcystin Source
Solution of lyophilized
cells of M. aemginosa
Solution of lyophilized
cells of M aemginosa
Solution of lyophilized
cells of M ichthyoblabe
Cell-free lysate of M.
aemginosa
Solution of lyophilized
bloom sample
Solution of lyophilized
bloom sample
Cell-free lysate of M.
aemginosa
Purified toxin fromM.
aemginosa
purified toxin of M.
aemginosa
Species
Rats
Mice
Mice
Mice
Mice
Mice
Mice
Mice
Mice
Sex/Strain
Male/Jcl:
Wistar
Male/Jcl:ICR
Male/Swiss
Male/Swiss
Not specified/
Balb/C
Male/Swiss
albino
Male/Swiss
albino
Not specified/
white, strain
unspecified
Male/white,
strain
unspecified
Vehicle
Saline
Saline
0.9% saline
0.9% saline
Saline
0.9% saline
Not
specified
Ethanol and
water
Not
specified
LD50 (95% CI)
67.4 mg dry weight
cells/kg, 2 hours
14.4 mg dry weight
cells/kg, 1 hour
502-1 924 mg cell dry
weight/ kg body weight
43 1 mg/kg dry weight, 24
hours
25-250 mg dry weight
phytoplankton sample/kg
154.28 mg algae/kg, 48
hours
3.5 g extract/kg, 24 hours
466+13 |ag/kg
56 |ag/kg (43-60)
Duration
2 hours
1 hour
Not
specified
24 hours
Not
specified
48 hours
24 hours
Not
specified
Not
specified
Comments

None of mice surviving
past 1 hour died within 1
week
Microcystin content
ranged from 0.73-0.78
Hg/g
LD50 by up and down
method
Bloom dominated by M.
aemginosa. MCLR
content of samples ranged
from 53-952 |ag/gDW
biomass
M. aemginosa was 95%
of bloom biomass
MCLR dose
approximately 9.625
mg/kg
Congener not
identified/specified
Congener not
identified/specified
Reference
Oishi and
Watanabe, 1986
Oishi and
Watanabe, 1986
Sabouretal.,
2002
Raoetal., 1994
Tarczynska et
al., 2000
Porfino et al.,
1999
Raoetal., 2005
Bishop etal.,
1959
Ellemanetal.,
1978
49
DRAFT - DO NOT CITE OR QUOTE

-------
 1          In further support for the findings of age-dependent liver effects from purified
 2   microcystins, Foxall and Sasner (1981) reported that neonatal and juvenile (age not specified)
 3   mice had no clinical symptoms or liver pathology after i.p. injections of a crude lysate from a
 4   bloom of M. aeruginosa. Details of the experiment and results were not provided. The authors
 5   reported that the mice did not die until they were 20 days of age, but it is not clear from the
 6   publication whether the mice treated as neonates died later or only mice that were at least 20
 7   days old when treated died. Mature mice were also treated for comparison, but the results were
 8   not reported. The authors concluded that "young animals were not sensitive to the toxin but
 9   developed sensitivity as they matured".
10
11          4.4.5.2.2. Short-Term Studies with Parenteral Exposure
12
13          Guzman and Solter (1999) and Solter et al. (1998, 2000) evaluated the  effects on rats of
14   short-term administration of MCLR via continuous i.p. infusion.  After 28 days of exposure at
15   16, 32 or 48 |j,g/kg-day, there were dose-dependent increases in serum  levels of sorbitol
16   dehydrogenase (SDH), AST, GOT, ALP and bile acids, while a dose-dependent decrease in
17   serum albumin and a decrease in ALT synthesis were also observed (Solter et al., 1998, 2000).
18
19          Immunohistochemistry on the liver showed evidence of bioaccumulation of MCLR in
20   liver cytosol, with measured liver concentrations increasing at a greater rate than the
21   administered dose (Solter et al.,  1998). Apoptotic cells and cytoplasmic vacuolation were
22   observed in the livers of rats receiving 32 and 48 |j,g/kg-day.  A later study exposing rats at the
23   same doses showed evidence for oxidative damage in the liver, as measured by dose-dependent
24   increases in  malondialdehyde, a lipid peroxidation byproduct (Guzman and Solter, 1999). This
25   observation is consistent with evidence for oxidative stress after short-term exposure to MCLR.

26          4.4.5.2.3. Subchronic Studies with Parenteral Exposure
27
28          Shi et al. (2002; Chinese publication, only abstract reviewed) reported  oxidative stress in
29   rats injected with i.p. doses of 4, 8 or 12 |j,g/kg-day MCLR for 35 days. Serum GGT and whole
30   blood glutathione were decreased, while LDH and AST increased after exposure, with no change
31   in ALT levels.  Hepatocyte proliferation and apoptosis were also observed. Oxidative stress and
32   apoptosis are discussed further in Section 4.4.7 (Mechanistic Studies).
33
34          Elleman et al. (1978) administered daily i.p. injections of a purified toxin from a bloom
35   of M aeruginosa to white male rats (strain unspecified) for 6 weeks. Doses were reported as
36   fractions of the LDioo (0.75, 0.5, 0.25 and 0). These doses correspond  to 52.5, 35 and 17.5 ng/kg
37   based on the reported LDioo (70 ng/kg). Two mice from each group were sacrificed weekly for
38   necropsy and histopathological examination of the liver, kidney, heart, lung, spleen and brain.
39   Eleven of 16 high-dose and 5 of 14 mid-dose mice died prior to scheduled sacrifice; none of the
40   mice receiving the low dose died early. Of the 11 high dose animals that died  prematurely, nine
41   died during the first week with symptoms of acute toxicity and liver hemorrhage. In the
42   remaining groups, progressive liver changes were seen with each week, and dose-dependent
43   pathology was observed. The authors noted numerous mitotic figures in hepatocytes of the low-
44   dose mice early on.  Other histopathological findings in the liver were  hepatocyte degeneration,
                                                50        DRAFT - DO NOT CITE OR QUOTE

-------
 1   scattered necrosis, fibrosis and mononuclear cellular infiltration; details of the severity of these
 2   findings in each group were not reported.
 o
 4          4.4.5.2.4. Chronic Studies with Parenteral Exposure
 5
 6          Milutinovic et al. (2002, 2003) evaluated the kidney effects of chronic i.p. administration
 7   of MCLR and MCYR in rats.  Doses of 10 ng/kg were administered to groups of five male
 8   Wistar rats every other day for 8 months. After sacrifice under CO2 anesthesia, the kidneys were
 9   removed, fixed, sectioned and stained. During exposure, the treated rats exhibited clinical signs
10   of toxicity and reduced body weight.  Microscopic examination of the kidneys of treated animals
11   showed collapsed glomeruli and dilated tubules with eosinophilic casts and some cytoplasmic
12   vacuolation.  The interstitial space was infiltrated with lymphocytes.  More renal corpuscles
13   were significantly damaged in the MCLR-treated group than the MCYR-treated group.
14   Cytoskeletal abnormalities and DNA damage typical of apoptosis or necrosis were also observed
15   in tubular epithelial cells (Milutinovic et al., 2002, 2003). Although they did not report details of
16   the liver pathology in this study, the authors noted that the kidneys were more damaged than the
17   livers, suggesting that adaptation to exposure may have occurred in the livers.
18
19          4.4.5.2.5. Initiation/Promotion Studies with Parenteral Exposure
20
21          Nishiwaki-Matsushima et al. (1992) demonstrated that i.p administration of MCLR could
22   enhance the number and area of glutathione S-transferase (placental form; GST-P) positive foci
23   in a medium-term rat liver bioassay.  In male F344 rats pretreated with 200 mg/kg
24   diethylnitrosamine (DEN)  and partially hepatectomized, the number of GST-P positive foci was
25   significantly increased when the rats were subsequently treated with 10 ng/kg MCLR i.p. twice a
26   week.  In a follow-up experiment, rats were pretreated with DEN and then given twice  weekly
27   i.p. doses of 10, 25 or 50 ng/kg MCLR. A dose-dependent increase in the number and  area of
28   GST-P positive foci was observed in the animals treated with MCLR (Nishiwaki-Matsushima et
29   al., 1992).
30
31          Ohta et al. (1994) also used the two-stage rat liver bioassay model to evaluate the
32   promotion capability of MCLR in rats pretreated with DEN, but not subjected to partial
33   hepatectomy. After treatment with 200 mg/kg DEN, rats given twice weekly i.p. injections of 25
34   Hg/kg MCLR had significantly more GST-P positive foci and significant increases in the area of
35   such foci when compared with DEN pretreatment alone. MCLR alone had negligible initiating
36   capability.
37
38          Hu et al. (2002) reported significant enhancement of gamma-GT foci in a two-stage
39   medium-term rat bioassay. Microcystin treatment (congener not specified) in DEN-pretreated
40   rats resulted in 100% incidence of gamma-GT foci, while DEN treatment alone resulted in foci
41   in only 22% of rats.  Immunohistochemistry showed that microcystin exposure reduced
42   expression of the bax gene and increased expression of the bcl-2 gene. The Bax protein induces
43   apoptosis, while the Bcl-2  protein inhibits apoptosis (Klassen, 2001). This finding suggests that
44   apoptosis is inhibited by low doses of microcystin; in contrast, higher doses appear to induce
45   apoptosis (see Section 4.4.7.5 below).
                                                51       DRAFT - DO NOT CITE OR QUOTE

-------
 1          Sekijima et al. (1999) conducted a similar experiment using either DEN or aflatoxin Bl
 2   (AFB1) as an initiator (with partial hepatectomy) and MCLR or a combination of AFB1 and
 3   MCLR for promotion. In rats pretreated with 200 mg/kg DEN and subsequently given 10 ng/kg
 4   MCLR i.p., there were increases in both the number and area of GST-P positive foci, but this
 5   difference did not reach statistical significance.  In rats pretreated with DEN and subsequently
 6   given either AFB1 alone or a combination of MCLR and AFB1, a statistically significant
 7   increase in number and area of foci was observed. The effect on GST-P positive foci of
 8   combined treatment with MCLR and AFB1 was not synergistic, however; the number and area
 9   of foci in animals treated with both was not larger than the sum of the foci induced by each
10   compound individually. When altered hepatic foci of all types (including basophilic and
11   eosinophilic/clear, rather than only GST-P positive foci) were analyzed, the number of foci was
12   significantly greater in rats treated with both MCLR and AFB1 than in those treated with either
13   toxin alone.  To assess the effect of MCLR on initiation by AFB1, Sekijima et al. (1999)
14   pretreated rats with AFB1 followed by twice weekly i.p. injections of 1 or 10 |J,g/kg MCLR for 6
15   weeks.  The number and area of GST-P positive foci were significantly increased in animals
16   given MCLR compared with controls. There was no difference in number or area of foci
17   between the two doses of MCLR. The authors suggested that the higher dose may have had
18   cytotoxic effects on hepatocytes.
19
20          Ito et al. (1997b) treated 13 ICR mice with 100 i.p. injections (5 times/week) of 20 |ig/kg
21   MCLR over 28 weeks. Five mice were sacrificed immediately after the last injection, while
22   eight mice were withdrawn from treatment and sacrificed 2 months later. Three untreated mice
23   served as controls. Relative liver weights were 4.7% of body weight in the control mice, 9% in
24   mice sacrificed immediately after the last treatment, and 6.8% in those withdrawn from treatment
25   and sacrificed later. Statistical comparisons among the relative liver weights were not provided,
26   nor were the data with which to perform these comparisons.  Neoplastic nodules were observed
27   in the livers of all mice of both treatment groups. The nodules ranged in size up to 5 mm in
28   diameter. The mean numbers of nodules in the treated animals (7.7 and 9.9 nodules per cm2 area
29   in the groups  sacrificed immediately and 2 months later, respectively) were not significantly
30   different from the controls (control data not provided).  The incidence of nodule development in
31   the few control mice was not reported.  The types of nodules were characterized as A) weakly
32   staining with  hematoxylin and eosin and with small nuclei; B) intensely staining with eosin but
33   not with PAS or C) mainly occupied with fat droplets.  The small number of animals in the
34   treatment and control groups limits the usefulness of these data.
35
36          4.4.5.2.6.  Developmental/Reproductive Studies with Parenteral Exposure
37
38          Chernoff et al. (2002) investigated the developmental toxicity of MCLR in CD-I mice.
39   Pregnant mice were treated with i.p. or subcutaneous doses of MCLR (95% pure) on gestation
40   days (GDs) 7-8, 9-10 or 11-12.  Doses of 0, 32, 64 and 128  ng/kg were administered i.p., while
41   only the 128 |J,g/kg dose level was administered subcutaneously.  Mice were sacrificed by CO2
42   inhalation on GD 17.  After litter and gravid uterus weights were recorded, fetuses were
43   examined for gross malformations and preserved for skeletal examination. Livers of the dams
44   were examined grossly and subjected to histopathology. No effects on maternal weight gain,
45   litter size, average fetal weight or incidence of gross or skeletal abnormalities were observed.
46   Histological examination of the maternal livers showed no effects of treatment.


                                               52       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2          In a separate experiment by the same researchers (Chernoff et al., 2002), pregnant mice
 3   treated with 32-128 ng/kg MCLR (via i.p. injection on GD 7-8, 9-10 or 11-12) were allowed to
 4   give birth, and the growth and viability of the offspring were followed for 5 days. A different,
 5   apparently more potent, lot of MCLR was used in this experiment. Mortality exceeded 50% (19
 6   of 35 dams) in the 64 |J,g/kg dose group and only 1 of 34 animals in each of the 96 and 128 |J,g/kg
 7   dose groups survived treatment. Among the surviving animals, there were no effects on
 8   viability, birth weight or growth of litters during the brief follow-up period.
 9
10          Experiments with rabbit whole embryo cultures in vitro suggest that low concentrations
11   of MCLR (10-20 |jM) can alter the organization of actin filaments and microtubules, although
12   cell morphology is not significantly affected (Frangez et al., 2003; Zuzek et al., 2003). At high
13   concentrations (100 |jM), MCLR causes cell rounding and loss of adhesion properties, with
14   consequent cell detachment and dispersion. Frangez et al. (2003) showed that the zona pellucida
15   (a glycoprotein envelope surrounding the ovum) forms an effective barrier against the effects of
16   MCLR, as rabbit whole embryos embedded in zona pellucida were not affected even at high
17   concentrations of MCLR.
18
19          Development of mouse embryos in culture was inhibited by a purified toxin from a
20   bloom dominated by Microcystis (Sepulveda et al.,  1992). At a concentration of 120 |j,g/mL of
21   toxin in the culture medium, development of two-cell embryos was halted and cytolysis occurred
22   in some embryos. Disruption of the actin cortex was also observed in these embryos.  At 60
23   Hg/mL, two-thirds of the embryos divided once more, and the remainder did not develop further.
24    In eight-cell embryos treated with 120 |j,g/mL toxin, compaction was prevented or reversed.
25   Embryos at this stage treated with 240 |j,g/mL did not develop further;  cells were rounded and
26   lysed.  The specific toxin used in this experiment was not identified.
27
28          Conflicting results have been observed in teratogenicity testing of purified microcystins
29   in the Frog Embryo Teratogenicity Assay-Xenopus (FETAX) assay with X. laevis embryos.  In
30   one study, MCLR at concentrations of 25-250 |J,g/L was shown to induce both skeletal and soft
31   tissue malformations mX. laevis embryos (Dvorakova et al., 2002).  In contrast, Fischer and
32   Dietrich (2000) reported no effects on mortality, malformation or growth in these embryos after
33   exposure to either MCLR or MCRR at concentrations up to 2000  |J,g/L. Dvorakova et al. (2002)
34   attributed the differing results to interlaboratory variability or variability in sensitivity of the
35   embryos. O'Brien et al.  (2003) reported no effects of MCYR on X. laevis embryo mortality,
36   malformation or growth.
37
38          Interestingly, Dvorakova et al. (2002) demonstrated that biomass from & Microcystis
39   species  (wesenbergii) that does not produce microcystins could induce malformations in
40   Xenopus embryos.  O'Brien et al. (2003) also tested two extracts each from Plantothrix
41   rubescens andM aeriigmosa, reporting that all four extracts resulted in facial narrowing and
42   growth retardation in 96-hour Xenopus embryos, while purified MCYR had no effects.
43
                                                53       DRAFT - DO NOT CITE OR QUOTE

-------
 1   4.4.6.  Effects by Intratracheal or Intranasal Instillation
 2
 3          Ito et al. (2001) evaluated the distribution of MCLR after intratracheal instillation of
 4   lethal doses in male ICR mice and included a limited description of toxic effects.  MCLR in
 5   saline solution was instilled at various doses (50, 75, 100, 150 and 200 ng/kg) into 34 mice; 3
 6   mice were sham-exposed as controls.  Mortality was 100% in 12 mice receiving doses of 100
 7   Hg/kg and greater. At 75 |J,g/kg, two of four mice died, while no deaths occurred in  18 mice
 8   given 50 ng/kg intratracheally.  The time course of hepatotoxicity was further evaluated in eight
 9   mice given intratracheal doses of 100 ng/kg. One mouse was sacrificed at each of 5, 10, 20, 30,
10   45, 60, 90 and 120 minutes. Immunostaining for MCLR showed the toxin in the lungs within 5
11   minutes and in the liver after 60 minutes. Hemorrhage in the liver was observed after 90 minutes
12   and became severe by 120 minutes.
13
14          Fitzgeorge et al. (1994)  conducted experiments in CBA/BALBc mice with MCLR
15   administered via intranasal instillation and inhalation.  This study is poorly described, giving few
16   details of study design and findings.  The LD50 for intranasal instillation of MCLR was equal to
17   the i.p. LD50 (250 ng/kg).  Liver and kidney weights were increased in the animals receiving
18   MCLR intranasally (41.6 and 7.5% respectively). The authors further evaluated the relationship
19   between dose and liver weight increase after intranasal instillation of MCLR. At single
20   intranasal doses of 31.3, 62.5, 125, 250 and 500 ng/kg, liver weight increased proportionally (0,
21   1.5, 24.4,  37.4 and 87%).  Seven daily intranasal doses of 31.3 ng/kg, a dose that had produced
22   no liver weight change after a single dose, resulted  in a liver weight increase of 75%. Fitzgeorge
23   et al. (1994) reported histopathological findings, but failed to specify which findings resulted
24   from single doses and which resulted from the multiple-dose experiment reported in the same
25   publication.  Findings included  necrosis of respiratory and  olfactory epithelium in the nasal
26   mucosa and centrilobular necrosis with hemorrhage in the liver.  Early changes in the liver
27   included vacuolar degeneration and  necrosis of hepatocytes near the central vein.  The adrenal
28   glands showed effects as well, with vacuolation and necrosis of the inner cortex, as well as
29   congestion of medullary blood vessels. No histopathological changes were observed in the
30   trachea, lungs, esophagus, pancreas, spleen, lymph  nodes, kidneys or brain.
31
32   4.4.7.  Mechanistic Studies
33
34          Many mechanistic studies have been conducted to characterize the toxicology of
35   microcystins. These studies include in vivo investigations in laboratory animals, in situ studies
36   in isolated perfused organ systems and in vitro assays in  isolated cell preparations. Mechanistic
37   studies have  evaluated many aspects of microcystin toxicity,  including: 1) the reason for target
38   organ and cell type specificity of microcystins, 2) description of the subcellular effects that occur
39   in susceptible cells, 3) interaction with serine and threonine protein phosphatases  (i.e., PP1 and
40   PP2A) as  the molecular target for microcystins, 4) the role  of cytoskeletal effects, 5) the
41   importance of oxidative stress and apoptosis as a mode of toxic action and 6) the use of
42   chemoprotectants to reduce toxicity.  Each of these topics is discussed in further detail below.
43   Mechanistic  data related to the genotoxicity of microcystins is presented below in Section 4.4.8
44   (Genotoxicity and Cell Proliferation).
45
                                                 54       DRAFT - DO NOT CITE OR QUOTE

-------
 1          4.4.7.1. Target Organ/Cell Type Specificity
 2
 3          Oral and injection studies in laboratory animals have demonstrated that the liver is the
 4   primary target organ for microcystin toxicity (see Section 4.2). Mechanistic studies suggest that
 5   the target organ specificity is directly related to the limited ability of microcystins to cross cell
 6   membranes in the absence of an active transport system, such as the bile acid transporter in
 7   hepatocytes.  Evidence of the importance of the bile acid transporter to liver toxicity is provided
 8   by studies that used bile acids and bile acid transport inhibitors (Runnegar et al., 1981, 1993,
 9   1995b; Runnegar and Falconer, 1982; Eriksson et al., 1990a; Hermansky et al., 1990a,b;
10   Hermansky et al., 1991). These studies demonstrated that the liver toxicity produced by in vitro
11   or in vivo exposures to microcystins was reduced or eliminated by inhibition of hepatocellular
12   uptake using bile acid transport inhibitors (e.g., antamanide, sulfobromophthalein and
13   rifampicin) and bile salts (i.e., cholate and taurocholate).  Additional discussion of the cellular
14   uptake of microcystins is provided in Section 3.2 (Distribution).
15
16          Runnegar et al. (1993) demonstrated that i.p. injection of mice with MCYM or MCLR
17   caused inhibition of liver protein phosphatase activity followed by evidence of liver toxicity (i.e.,
18   increased liver weight). Kidney protein phosphatase activity was unchanged following the in
19   vivo exposure even at lethal doses. In vitro exposure of kidney extracts to microcystins did
20   result in a decrease in kidney phosphatase activity, and no difference in sensitivity was observed
21   between liver and kidney phosphatase inhibition.  This result suggests that target organ
22   specificity is most likely due to slower intercellular uptake of microcystins in the  kidney.
23
24          The cell type specificity of microcystins was investigated using isolated rat hepatocytes,
25   rat renal epithelial cells (ATCC 1571) and rat skin fibroblasts (ATCC 1213) (Khan et al., 1995;
26   Wickstrom et al., 1995). The time course of light microscopic and ultrastructural effects was
27   examined following in vitro exposure to MCLR (Khan et al., 1995). Effects were noted after 4
28   minutes in hepatocytes, 1 hour in renal cells and 8 hours in fibroblasts.  Similar lesions observed
29   in all cell  types included blebbing, loss of cell-cell contact, clumping and rounding, cytoplasmic
30   vacuolization and redistribution of cellular organelles.  Effects that were seen only in
31   hepatocytes include loss of microvilli, whirling of rough ER, dense staining and dilated cristae of
32   mitochondria and pinching off of membrane blebs. The nuclear changes typical of apoptosis
33   were seen in renal cells and fibroblasts.  Cell type differences may be related to the specific
34   proteins that were overphosphorylated within each cell type. The authors postulated that the lack
35   of apoptotic changes in hepatocytes in this study might be related to the short exposure duration
36   or the failure of their transmission electron microscopic method to examine severely damaged
37   cells that had detached from the coverslips. Wickstrom et al. (1995) evaluated the changes in
38   cytoskeletal morphology after MCLR exposure in these cell types.  High concentrations and long
39   incubation times were required  for cytoskeletal changes in kidney and skin cells; however, the
40   nature of the changes was similar in all cell types (e.g., actin aggregation). Wickstrom et al.
41   (1995) suggested that microcystins may enter kidney cells and fibroblasts via pinocytosis.
42
43          McDermott et al. (1998) treated several cell types with MCLR (primary rat hepatocytes,
44   human fibroblasts, human endothelial cells, human epithelial cells and rat promyelocytes).
45   Hepatocytes underwent membrane blebbing, cell shrinkage, organelle redistribution, chromatin
                                                 55        DRAFT - DO NOT CITE OR QUOTE

-------
 1   condensation and, in some cells, DNA fragmentation. Similar changes were observed in the
 2   other cell types, but a longer duration of exposure was required.
 3
 4          Matsushima et al. (1990) demonstrated that injection of MCYR into mouse skin
 5   epithelial cells and human fibroblasts resulted in morphological changes in cell shape (i.e.,
 6   spindle shape to round forms). These cells were thought to be resistant to microcystins;
 7   however, microcystins YR, LR and RR produced a dose-dependent inhibition of protein
 8   phosphatase activity using the partially purified enzyme derived from mouse skin cell cytosol.
 9   The authors suggest that absence of a direct effect in these cells is due to lack of penetration
10   through the cell membrane.
11
12          Many cell types and established cell lines have been evaluated for potential susceptibility
13   to microcystin uptake and toxicity. Primary isolated hepatocytes have been shown to be the
14   most  sensitive to cytoxicity, due to the presence of the organic ion/bile acid transport system
15   (Eriksson et al., 1987, 1990b). Uptake was negligible in human hepatocarcinoma cells (Hep
16   G2), mouse fibroblasts (NIH-3T3), erythrocytes and human neuroblastoma cells (SH-SY5Y).
17   Hepatic endothelial cells have also been shown to be resistant to microcystin toxicity (Solow et
18   al., 1989; Runnegar et al., 1994). Primary cultures of liver cells cease to express these bile acid
19   transport proteins after 2-3 days of being maintained in culture. Therefore, established liver cell
20   lines  are not generally useful for evaluating microcystin toxicity (Eriksson et al., 1994; Battle et
21   al., 1997; Heinze et al., 2001).
22
23          4.4.7.2. Characterization of Subcellular Effects in the Liver
24
25          The liver effects that occur following in vivo exposures to microcystins are generally
26   discussed in Section 4.2 (Animal Studies). Many additional mechanistic studies describe liver
27   histopathology, ultrastructural changes and biomarkers of cytotoxicity in either isolated perfused
28   rat liver (Berg et al., 1988; Pace et al., 1991; Runnegar et al.,  1995b) or primary isolated
29   hepatocytes (Runnegar et al.,  1981; Runnegar and Falconer, 1982; Aune and Berg, 1986;
30   Runnegar and Falconer, 1986; Berg and Aune, 1987; Runnegar et al., 1987; Falconer and
31   Runnegar,  1987a,b; Thompson et al., 1988; Solow et al., 1989; Mereish et al., 1989;  Mereish and
32   Solow, 1990; Eriksson et al., 1990a; Boe et al., 1991; Khan et al.,  1995; Runnegar et al., 1995b;
33   Yea et al., 2001; Batista et al., 2003).
34
35          Toxicological effects of microcystins in the isolated perfused rat liver were similar to
36   those demonstrated following in vivo exposure (Pace et al., 1991).  During a 60-minute exposure,
37   MCLR caused liver engorgement and cessation of bile flow.  Electron microscopy revealed loss
38   of sinusoidal architecture, dilation of bile canaliculi and the space of Disse and decreased
39   intracellular contact. Mitochondrial swelling, disruption of endoplasmic reticulum and
40   formation of whorls and loss of desmosomal intermediate filaments were also observed.
41   Mitochondrial function was impaired, with inhibition of state 3 respiration and a decrease in the
42   respiratory control index.
43
44          Runnegar et al. (1995b) demonstrated a decrease in protein phosphatase activity in
45   perfused rat liver exposed to MCYM.  Cessation of bile flow, increased perfusion pressure,
46   decreased protein secretion and decreased glucose secretion were also observed. Histological
                                                 56       DRAFT - DO NOT CITE OR QUOTE

-------
 1   changes included hepatocytes swelling, loss of sinusoidal architecture, pyknotic nuclei and
 2   extensive necrosis.  Exposure to high concentrations of toxin extracts in the isolated perfused
 3   liver produced loss of cord architecture due to hepatocyte disassociation, membrane damage and
 4   cytolysis and nuclear effects (pyknosis, karyokinesis, karyolysis) (Berg et al., 1988).
 5   Ultrastructural effects included swollen mitochondria, vacuoles, necrosis, abnormal nuclei, bile
 6   canaliculi lacking microvilli and whorls of rough endoplasmic reticulum.
 7
 8          Studies in primary isolated hepatocytes have described the morphological and
 9   histopathological changes induced by microcystins that relate to loss of sinusoidal architecture
10   and cytotoxicity (Runnegar et al., 1981; Runnegar and Falconer,  1982; Aune and Berg, 1986;
11   Runnegar and Falconer, 1986; Berg and Aune, 1987; Runnegar et al., 1987; Falconer and
12   Runnegar,  1987a,b; Thompson et al., 1988; Solow et al., 1989; Mereish et al., 1989; Mereish and
13   Solow, 1990; Eriksson et al., 1990a; Boe et al., 1991; Runnegar et al., 1995b; Khan et al., 1995;
14   Yea et al., 2001; Batista et al., 2003). Microcystin exposure to hepatocytes in suspension or
15   cultured in a monolayer results in membrane blebbing that becomes more pronounced and
16   localized in one region of the cell surface.  Cells are observed to be rounded in appearance and
17   become dissociated from one another.  Microfilaments are reorganized as a compact spherical
18   body in the vicinity of the blebbing, while the rest of cell is depleted of filamentous actin.
19   MCLR disrupts hepatocellular morphology within minutes, leading to loss of sinusoidal
20   architecture and hemorrhage. Morphological changes in hepatocytes (i.e., blebbing, rounding)
21   have been shown to occur prior to any effect on cell  membrane integrity (measured as LDH
22   leakage or  release of radiolabeled adenine nucleotides) or cell viability (generally measured as
23   decreased trypan blue exclusion) (Runnegar et al., 1981; Runnegar and Falconer, 1982; Aune
24   and Berg, 1986; Ding et al., 2000a).
25
26          Thompson et al. (1988) described the time course of cellular effects of microcystins (type
27   not specified) on cultured rat hepatocytes.  Disintegration of attachment matrix occurred by 15
28   minutes, followed by cells clustered in groups with no extracellular material at 1 hour,  and
29   release of cells from plates between 2 and 4 hours. LDH release did not occur until after these
30   visual effects, but was dose-related.
31
32          Similar toxicological effects were observed in isolated human hepatocytes (Yea et al.,
33   2001; Batista et al.,  2003). MCLR produced blebbing, fragmentation and hepatocyte
34   disassociation.  Cytotoxicity, as measured by LDH leakage, occurred after morphological
35   changes were evident.  Yea et al. (2001) indicated that cytotoxicity in human hepatocytes was
36   observed at a concentration (1  |jM) that did not affect rat hepatocytes.  Batista et al. (2003) also
37   reported a slightly higher susceptibility to microcystin-induced morphological change in human
38   hepatocytes as compared to rat hepatocytes.
39
40          The ultrastructural effects observed following microcystin exposure in isolated rat
41   hepatocytes (i.e.,  condensation of chromatin, segregation of organelles, separated by apoptotic
42   microbodies, decrease in cell volume and increase in cell density) suggest that hepatocyte cell
43   death is apoptosis and not necrosis (Boe et al., 1991). Microcystin exposure in hepatocytes
44   causes cell shrinkage, externalization of membrane phosphatidylserine, DNA fragmentation and
45   chromatin condensation, indicating a rapid apoptosis (Ding et al., 2000b). Apoptosis is
46   discussed in further detail below as a possible mode of action for microcystin liver toxicity.
                                                 57       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Several studies have suggested that microcystins may increase the release and decrease the
 2   reincorporation of arachadonic acid into cellular membranes (Adams et al., 1985; Naseem et al.,
 3   1990, 1991; Nobre et al., 2001).
 4
 5          4.4.7.3. Molecular Target: Inhibition of Type 1 and 2A Protein Phosphatases
 6
 7          The primary molecular target of microcystins has been identified as serine and threonine
 8   protein phosphatases PP1/PP2A.  Protein phosphatases dephosphorylate proteins while protein
 9   kinases phosphorylate them.  Together, protein kinases and phosphatases maintain the balance of
10   phosphorylation and dephosphorylation of key proteins involved in cell cycle regulation.
11   Because more than 97% of protein phosphates occur at serine and threonine residues (Gehringer,
12   2004), the PP1 and PP2A protein phosphatases are particularly important. Inhibition of these
13   enzymes results in the increased phosphorylation of a number of regulatory proteins.
14   Importantly, PP1 is believed to be the major phosphorylase a phosphatase in the liver (Runnegar
15   et al., 1993).  PP2A, the major soluble serine/threonine phosphatase, regulates several mitogen-
16   activated protein kinases (Gehringer, 2004).
17
18          Microcystins bind to these enzymes under both in vivo and in vitro study conditions,
19   resulting in an inhibition of enzyme activity leading to an increase in protein phosphorylation.
20   Microcystins have been shown to directly inhibit the activity of PP1 and PP2A derived from
21   several different species (i.e., fish, mammals, plants) and cell types  (Honkanen et al., 1990;
22   MacKintosh et al., 1990; Matshushima et al., 1990; Yoshizawa et al., 1990; Sim and Mudge,
23   1993; Xu et al., 2000; Leiers et al., 2000; Becchetti et al., 2002). Microcystins do not alter
24   protein kinase activity, suggesting the balance of phosphorylation and dephosphorylation is
25   related to protein phosphatase inhibition alone. Microcystins have been used as a tool to
26   investigate the importance of serine and threonine phosphorylation to specific cellular functions.
27   The regulatory effects of phosphorylation on sodium channel opening in renal cells (Becchetti et
28   al., 2002), smooth and skeletal muscle contraction (Hayakawa and Kohama, 1995; Knapp et al.,
29   2002) and insulin secretion (Leiers et al., 2000) have been  studied.
30
31          Runnegar et al.  (1993) demonstrated the inhibition  of PP1 and PP2A activity in the liver
32   following i.p. injection of MCYM and MCLR.  Increased protein phosphorylation preceded the
33   observed increase in liver weight and was correlated with hepatotoxicity. Decreased
34   phosphatase activity was also  demonstrated in the isolated  perfused liver  (Runnegar et al.,
35   1995b).

36          The relationship between phosphatase inhibition by microcystins and changes in
37   cytoskeletal structure and cell morphology has been reviewed (Eriksson and Golman, 1993).
38   Inhibition of protein phosphatase activity by MCLH and 7-dmMCRR was associated with an
39   increased phosphorylation of cytoskeletal  and cytosolic proteins (Eriksson et al., 1990b).
40   Concentrations that produce a marked increase in protein phosphorylation were accompanied by
41   a complete reorganization of microfilament network. The  cytoskeletal  effects of microcystins
42   are discussed in further detail below. Microcystin LH and  7-dmMCRR were equipotent
43   inhibitors of purified PP1 and PP2A; however, higher concentrations of 7-dmMCRR were
44   required to increase protein phosphorylation. Table 4-10 shows studies with comparative data
45   on inhibition  of protein phosphatases (IC50s) by MCLR, MCYR, MCRR and MCLA.
                                                58        DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-10. Studies Comparing Protein Phosphatase Inhibition Activity of Microcystin
Congeners
Reference
IC50 (nM)
MCLR
MCLA
MCYR
MCRR
PP2A Inhibition
Craig etal., 1996
Nishiwaki-Matsushima et al., 1991
Matsushima et al., 1990
0.15
0.28
7.6
0.16




4.5

0.78
5.8
PP1 Inhibition
MacKintosh et al., 1995
0.2

0.2

Mixture of PPs
Yoshizawa et al., 1990
1.6

1.4
3.4
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
       The molecular interaction between microcystins and protein phosphatases has been
evaluated using immunoprecipitation, autoradiography, reverse phase liquid chromatography,
X-ray crystallography, nuclear magnetic resonance (NMR) solution structures, and molecular
dynamics simulation (Runnegar et al., 1995b; MacKintosh et al., 1995; Goldberg et al., 1995;
Craig et al., 1996; Bagu et al., 1997; Mattila et al., 2000; Mikhailov et al., 2003; Maynes et al.,
2004, 2006).  Molecular modeling and molecular dynamics simulations have indicated that
microcystins bind in a Y-shaped groove containing the catalytic site on the surface of PP1
(Mattila et al., 2000).  Studies with PP1  suggest that the C-terminal P12-P13 loop of PP1
(containing residues 268-281) is important for microcystin-protein phosphatase interactions as
well as for substrate recognition (Maynes et al., 2004, 2006).  Information available to date
indicates that the binding process primarily involves the amino acids Glu, Adda, Leu and MDha
of microcystins.  Figure 4-2 shows a schematic representation of the interactions between
microcystin-LR and protein phosphatase 1; these interactions are discussed further below.

       Microcystins LR,  LA and LL interact with the catalytic subunits of PP1 and PP2A in two
phases.  The first phase occurs within minutes and consists of rapid inactivation of the
phosphatase.  The second, slower phase of interaction represents a covalent interaction that takes
place within several hours (Craig et al.,  1996). The initial binding and inactivation of protein
phosphatases appears to result from several non-covalent interactions that are still being
elucidated.  Mattila et al.  (2000) demonstrated an interaction of the Glu  amino acid (reported as
                                                59
                                                     DRAFT - DO NOT CITE OR QUOTE

-------
 1   IGlu in the publication) carboxyl group of MCLR with a metal ion (Fe, Mn) in the PP1 catalytic
 2   site. Glu appears to be an important component because esterification eliminates toxicity
 3   (Namikoshi et al., 1993; Rinehart et al., 1994). Herfmdal and Selheim (2006), in a review of the
 4   mechanisms of microcystin toxicity, indicated that the Adda side chain is involved in a
 5   hydrophobic interaction between the Trp 206 and IlelSO residues in the hydrophobic groove of
 6   PPL Mattila et al. (2000) suggested that the long side chain of the Adda residue may contribute
 7   to orienting the toxin into the catalytic site.  The Adda amino acid residue of microcystins plays
 8   an important role in the inhibition of protein phosphatase activity (Nishiwaki-Matsushima et al.,
 9   1991; Gulledge et al., 2002, 2003a,b). Isomerization of the diene from 4E,6E to 4E,6Z on the
10   Adda chain (see Figure 2-1) eliminates the toxic activity of microcystins (Harada et al., 1990;
11   Nishiwaki-Matsushima et al., 1991; Stotts et al., 1993).  Microcystin analogues containing only
12   Adda and one additional amino acid are capable of substantial inhibition of PP1 and PP2A, while
13   modifications to the Adda structure abolished the inhibition (Gulledge et al., 2003b). Finally,
14   Herfmdal and Selheim (2006) indicated that the L-Leu of MCLR participates in a hydrophobic
15   interaction with Tyr 272 of PP1  (on the 012-013 loop).
16
17          The second phase of interaction between microcystins and protein phosphatase consists
18   of covalent bonding (Craig et al., 1996). Immunoprecipitation and autoradiography methods
19   indicate that covalent bonds result from the interaction between the methylene of the MDha
20   residue of microcystins and the thiol of Cys273 located at the C-terminal of PPL  NMR solution
21   structures and X-ray crystallography data on the MCLR/PP1 complex illustrate the covalent
22   linkage at Cys-273 (Goldberg et al., 1995; Bagu et al., 1997). Site-directed mutagenesis
23   replacing  Cys273 in PP1 results in a loss of microcystin binding (MacKintosh et al., 1995;
24   Maynes et al., 2004). Based on  sequence similarity between PP1 and PP2A, it has been
25   suggested that Cys-266 is the site of equivalent covalent linkage between PP2A and microcystins
26   (Craig etal., 1996).
27
28          Microcystin analogues containing a reduced MDha residue are not capable of covalent
29   binding to protein phosphatases. MacKintosh et al. (1995) reported that a reduction of the MDha
30   residue of MCYR by ethanethiol abolished  covalent binding to PPL Likewise, Craig et al.
31   (1996) showed that reduction of the MDha residue of MCLA abolished the covalent binding
32   phase with PP2A. Maynes et al. (2006) confirmed the lack of covalent interaction by
33   determining the crystal structure of dihydroMCLA bound to PP1.  Their work showed that the
34   012-013 loop of PP1 takes on a different conformation when the covalent bond is absent, and
35   that other interactions (including hydrogen bonding) are  responsible for the bond between
36   dihydroMCLA and PP1.
37
38          The importance of covalent bonding between microcystins and protein phosphatases to
39   toxicity resulting from the enzyme inhibition is uncertain, as other interactions are apparently
40   responsible for the rapid inactivation of the enzymes (Herfmdal and Selheim, 2006).
41   Modifications to either molecule (microcystin or protein phosphatase) to prevent covalent
42   bonding generally decrease, but do not eliminate, the toxic action (Meriluoto et al., 1990;
43   MacKintosh et al., 1995; Hastie et al., 2005).
                                                60       DRAFT - DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
               P12-P13loop

                 I
Figure 4-2.  Schematic Representation of Interactions between Microcystin-LR and the Catalytic Site of Protein Phosphatase 1
                                                             61
                                                                               DRAFT - DO NOT CITE OR QUOTE

-------
 1          Microcystins may also bind to other molecular targets in addition to protein
 2   phosphatases. Chen et al. (2006) used bioinformatic approaches to identify human liver aldehyde
 3   dehydrogenase 2 (ALDH2) as a potential molecular target of MCLR.  After screening a phage
 4   display library to identify potential ligands specific for MCLR, Chen et al. (2006) used
 5   molecular docking studies to show that MCLR could bind to ALDH2. The authors postulated
 6   that this interaction could lead to aldehyde-induced reactive oxygen species (ROS) and
 7   apoptosis.
 8
 9          4.4.7.4. Cytoskeletal Effects
10
11          The cytoskeletal effects of microcystins in the liver have been visually demonstrated in
12   several studies using light, electron and fluorescent microscopy (Runnegar and Falconer, 1986;
13   Eriksson et al.,  1989; Hooser et al., 1989a,b, 1991b; Falconer and Yeung, 1992). Ultrastructural
14   changes in rats  given a lethal dose of microcystin A1 include the following: a widening of
15   intracellular spaces; progressive disassociation followed by rounding, blebbing and invagination
16   of hepatocytes; loss of microvilli in the space of Disse; breakdown of the endothelium;
17   hemorrhage;  and loss of lobular architecture (Hooser et al., 1989b).  No effects were noted in
18   endothelial cells or Kupffer cells.  In isolated hepatocytes, actin aggregates were seen at the base
19   of the membrane blebs.  As membrane blebs grew larger and were drawn toward one pole of the
20   cell, the microfilaments were organized toward the  same pole, resulting in rosette formation with
21   a condensed band of microfilaments at the center (Runnegar and Falconer, 1986; Eriksson et al.,
22   1989; Hooser et al.,1991b; Falconer and Yeung, 1992; Wickstrom et al., 1995; Ding et al.,
23   2000a). Frangez et al. (2003) also demonstrated cytoskeletal changes in rabbit primary  whole
24   embryo cultured cells.  Actin and microtubule disorganization was demonstrated to lead to
25   detachment and cellular toxicity.
26
27          The observed reorganization of microfilaments that leads to alteration of hepatocyte
28   morphology was not shown to be due to effects on actin polymerization (Runnegar and Falconer,
29   1986; Eriksson et al., 1989; Falconer and Yeung, 1992). Instead,  microcystins cause an increase
30   in the phosphorylation of cytokeratin intermediate filament proteins (Falconer and Yeung,  1992;
31   Ohta et al., 1992; Wickstrom et al., 1995; Blankson et al., 2000).  Toivola et al. (1997) evaluated
32   the effects  of MCLR on hepatic keratin intermediate filaments in primary hepatocytes cultures.
33   A disruption  of the desmoplakin2 organization  at the cell surface (disorganization of
34   desmosomes) is followed by a dramatic reorganization of the intermediate filament and
35   microfilament networks, resulting in intermediate filaments being organized around a condensed
36   actin core.  The major target proteins for microcystin-induced hyperphosphorylation include
37   keratins 8 and 18 and desmoplakin (DP) I/II. Keratins 8 and 18 are the major proteins of
38   intermediate  filaments in hepatocytes; DP I and II attach keratin filaments in epithelial cells to
39   desmosomes. Hyperphosphorylation of DPI/II leads to loosening of cell junction and loss of
40   interactions with cytoplasmic intermediate filaments.  The hyperphosphorylation of keratin
41   proteins leads to increased solubility (caused by disassembly or prevention of subunit
42   polymerization), leading to the observed morphological effects. Phosphopeptide mapping shows
     1 The authors refer to the test compound as microcystin-A, which may reflect an old nomenclature no longer in use.
     Available information is insufficient to identify the congener with current nomenclature.
     2 Desmoplakin is the principal plaque protein in a desmosome, which is a localized thickening of membrane that
     serves as an adhesion junction connecting contiguous cells.


                                                 62       DRAFT - DO NOT CITE OR QUOTE

-------
 1   four specific tryptic peptides in soluble keratin 18 that are highly phosphorylated; however, no
 2   specific phosphorylation sites have been identified for keratin 8.  A Ca2+/calmodulin-dependent
 3   kinase may be involved in regulating the serine-specific phosphorylation of keratin proteins 8
 4   and 18.
 5
 6          Some investigators have suggested that generation of reactive oxygen species may play a
 7   role in the cytoskeletal changes induced by microcystins. Ding et al. (2001) illustrated
 8   generation of superoxide and hydrogen peroxide radicals preceding microfilament
 9   disorganization and cytotoxicity. Hepatocellular glutathione levels were affected by
10   microcystins, and administration of N-acetylcysteine was shown to protect against cytoskeletal
11   alterations (Ding et al., 2000a).
12
13          4.4.7.5. Apoptosis
14
15          The ultrastructural changes observed in hepatocytes after microcystin exposure suggest
16   that cell death is related to apoptosis and not necrosis. These changes include cell shrinkage
17   (decreased volume and increased density), condensation of chromatin and segregation of
18   organelles separated by apoptotic microbodies (Boe et al., 1991; Fladmark et al., 1998;
19   McDermott et al., 1998; Ding et al., 2000b; Mankiewicz et al., 2001). The effects of
20   microcystins on the signaling pathways involved in rapid apoptosis have been investigated in
21   several studies (Ding et al., 1998a,b, 2000b, 2001, 2002; Ding  and Ong, 2003).  Mitochondrial
22   permeability transition (MPT)  is considered to be a critical rate-limiting event in apoptosis.
23   Oxidative stress may play a role in the induction of MPT and the onset of apoptosis.  In cultured
24   hepatocytes exposed to microcystins, an increase in the generation of ROS preceded the onset of
25   MPT, mitochondrial depolarization and apoptosis.  A dose- and time-dependent increase in ROS
26   and lipid peroxidation, measured as malondialdehyde formation, was shown to  precede
27   morphological changes in hepatocytes and release of LDH.  The addition of deferoxamine or
28   cyclosporine A inhibited the formation of ROS and delayed the onset of MPT and cell death.
29   The addition of superoxide dismutase prevented collapse of cytoskeleton and release of LDH
30   from isolated hepatocytes. An early surge of mitochondrial Ca2+ was shown to  occur prior to
31   MPT and cell death.  Prevention of this Ca2+ surge by one of several methods (i.e., chelation of
32   intracellular Ca2+, blockage of the mitochondrial Ca2+ uniporter or use of mitochondrial
33   uncoupler) prevented MPT and cell death.  Electron transport chain inhibitors (e.g., rotenone,
34   actinomycin A, oligomycin or  carbonyl cyanide m-chlorophenylhydrazone) also inhibited the
35   onset of MPT.  MCLR caused  the release of cytochrome c through MPT, which is considered
36   universal in mitochondrial apoptosis; however, caspases -9 and -3 were not activated.  The
37   increase in intracellular Ca2+ may instead facilitate the activation of calpain, which occurred
38   following exposure to microcystins (Ding and Ong, 2003). Botha et al. (2004)  demonstrated that
39   apoptosis and  oxidative stress can be induced in nonhepatic cells by microcystins.  LDH leakage
40   and increased apoptotic indices were observed in the human colon carcinoma cell line (CaCo2)
41   and MCF-7  cells (deficient in pro-caspase-3).  These changes were accompanied by increased
42   H2O2 formation and increased  calpain  activity.
43
44          Western blot analysis has been used to show that MCLR increases the expression of p53
45   and the pro-apoptotic Bax protein in both cultured rat hepatocytes treated with MCLR in vitro
46   and rat liver after in vivo exposure (Fu et al., 2005). Expression of the anti-apoptotic protein
                                                63        DRAFT - DO NOT CITE OR QUOTE

-------
 1   Bcl-2 was decreased in vitro, but in vivo MCLR treatment did not lead to a difference in the
 2   expression of this protein. This finding suggests that MCLR may induce apoptosis through other
 3   mechanisms in addition to the inhibition of protein phosphatases.
 4
 5          4.4.7.6. Lipid Peroxidation
 6
 7          Several studies have investigated the role of glutathione homeostasis and lipid
 8   peroxidation in microcystin-induced liver toxicity  (Runnegar et al., 1987; Eriksson et al., 1989;
 9   Bhattacharya et al., 1996; Ding et al., 2000a; Towner et al., 2002; Gehringer et al., 2003a,b,
10   2004; Boua'icha and Maatouk, 2004). Ding et al. (2000a) indicated that microcystin exposure in
11   isolated hepatocytes resulted in an initial increase in glutathione synthesis followed by a later
12   depletion of glutathione.  MCLR was shown to induce the de novo synthesis of glutathione in
13   mice exposed to a toxic sublethal dose (75% of the LD50) (Gehringer et al., 2004). Increased
14   transcription of glutathione-S-transferase was also demonstrated in his study. Gehringer et al.
15   (2004) suggest that increased lipid peroxidation induced by microcystins is accompanied by an
16   increase in glutathione peroxidase, transcriptional  regulation of glutathione-S-transferase and
17   glutathione peroxidase and de novo synthesis of glutathione.  Boua'icha and Maatouk (2004) also
18   reported that a low noncytotoxic concentration (2 ng/mL) of MCLR in primary rat hepatocytes
19   caused an initial increase in ROS formation and an increase in glutathione; however, a decrease
20   in lipid peroxidation was observed in this study. Electron spin resonance (ESR) spin trapping
21   techniques have demonstrated the formation of two possible lipid-derived free radical
22   metabolites in rat liver following in vivo exposure  to MCLR (Towner et al., 2002). Vitamin E
23   and selenium supplementation in mice provided some protection against liver toxicity and
24   lethality by MCLR (Gehringer et al., 2003a,b). Measures of liver toxicity included serum
25   enzyme determination, lipid peroxidation, glutathione levels and histopathology. Hermansky et
26   al. (1991) reported that membrane active antioxidants (i.e., vitamin E, silymarin and glutathione)
27   provided some protection from microcystin toxicity (i.e., LDH leakage) and lethality; however,
28   free radical scavengers and water  soluble antioxidants were ineffective (see below).
29
30          Several studies have reported MCLR-induced increases in lipid peroxidation as well as
31   decreases in antioxidant enzymes  (Moreno et al., 2005; Jayaraj et al., 2006).  Jayaraj et al. (2006)
32   measured oxidative stress in mice treated i.p. with an LDso dose of MCLR. Significant increases
33   in heat shock protein-70 and hepatic lipid peroxidation were observed. Further, GSH was
34   depleted, and there were decreases in the activity of glutathione peroxidase, superoxide
35   dismutase, catalase, glutathione reductase and glutathione-S-transferase in the animals treated at
36   the LD50. Similarly, Moreno et al. (2005) reported significant reductions in glutathione
37   peroxidase, glutathione reductase, superoxide dismutase and catalase, along with increases in
38   lipid peroxidation, in both the liver and kidney of rats treated intraperitoneally with single doses
39   of MCLR.
40
41          Some studies report the absence of lipid peroxidation during microcystin-induced
42   hepatotoxicity. A time-dependent leakage of LDH, ALT and AST was observed in liver slices
43   with no change observed for glutathione content or lipid peroxidation (Bhattacharya et al.,  1996).
44   In addition, Runnegar et al. (1987) indicated that glutathione depletion did not occur until after
45   morphological changes (i.e., blebbing) were observed.  Eriksson et al. (1989) indicated that rapid
                                                 64       DRAFT - DO NOT CITE OR QUOTE

-------
 1   deformation of isolated rat hepatocytes by MCLR was not associated with alterations in
 2   glutathione homeostasis.
 o
 4          4.4.7.7. Prevention of Liver Toxicity and Lethality
 5
 6          Several types of agents have been evaluated as potential chemoprotectants against
 7   microcystin toxicity, including inhibitors of bile acid transport, microsomal enzyme inducers,
 8   calcium channel blockers, free radical scavengers, water-soluble antioxidants and membrane
 9   active antioxidants. It was initially reported that preincubation of hepatocytes with chemicals
10   that interfere with uptake of bile acids (sulfobromophthalein, rifampicin, sodium cholate and
11   sodium deoxycholate) also prevents hepatocyte deformation (Runnegar et al., 1981; Runnegar
12   and Falconer,  1982).
13
14          Hermansky et al. (1991) evaluated several possible chemoprotectants by measuring LDH
15   leakage and lethality following i.p. injection of MCLR.  No protective effect was observed using
16   calcium channel blockers, free-radical scavengers or water-soluble antioxidants administered
17   prior to MCLR.  Membrane active antioxidants,  such as vitamin E, silymarin and glutathione,
18   provided some protection from microcystin toxicity and lethality. Phenobarbital provided partial
19   protection; however, tetrachlorodibenzo-^-dioxin did not afford protection,  suggesting that the
20   protective effect may not be related to microsomal enzyme induction.  Rifampicin and
21   cyclosporine A provided complete protection by blocking uptake of microcystins at the bile acid
22   transporter.  Hermansky et al. (1990a,b)  reported that rifampicin can be given 15 minutes after
23   MCLR injection and still prevent lethality, while cyclosporine A prevents lethality only if given
24   0.5-3 hours before MCLR injection.
25
26          Silymarin and dithioerythritol, both antioxidants, were shown to reduce MCLR toxicity,
27   as measured by LDH and adenine nucleotide release and light microscopy in primary cultures of
28   adult rat hepatocytes (Mereish and Solow, 1990). Dithioerythritol and silymarin have both been
29   shown to increase the content of reduced thiols (i.e., glutathione). Silymarin has additionally
30   been shown to stabilize membranes, inhibit lipoxygenase, reduce leukotrienes, scavenge free
31   radicals and increase protein synthesis.
32
33          Mereish et al. (1991) indicated that silymarin pretreatment inhibited microcystin liver
34   toxicity in mice (evidenced by histopathology and serum enzyme levels) following  i.p. injection,
35   but not oral administration.  As discussed above, vitamin E and selenium supplementation in
36   mice provided some protection against liver toxicity and lethality by MCLR (Gehringer et al.,
37   2003a,b).
38
39          Thompson and Pace (1992) evaluated several types of agents for their ability to protect
40   against MCLR toxicity in cultured hepatocytes.  Toxicity was measured  as morphology under
41   light microscopy, LDH release and protein synthesis inhibition.  The uptake of MCLR into
42   hepatocytes was also measured. Cytochalasins D&E,  fungal metabolites that interfere with actin
43   polymerization into microfilaments, were shown to protect against LDH release and provided
44   moderate protection from rounding and clustering of cells; however, these compounds produced
45   cytotoxicity themselves at concentrations that were required for protection against microcystin
46   toxicity.  Cholate and deoxycholate are competitive inhibitors for the bile acid transporter.
                                                 65       DRAFT - DO NOT CITE OR QUOTE

-------
 1   These compounds provided some protection against LDH leakage, but were also cytotoxic.
 2   Trypan blue and trypan red also provided some protection related to blocking microcystin uptake
 3   into hepatocytes. The antibiotic rifampicin was shown to prevent microcystin uptake and
 4   toxicity at low non-cytotoxic concentrations, suggesting a possible therapeutic use in
 5   microcystin poisoning.
 6
 7          Rao et al. (2004) confirmed that pretreatment with cyclosporine-A, rifampin and
 8   silymarin each provided 100% protection against a lethal dose of MCLR (although the route was
 9   not specified, it is assumed to be i.p., based on the lethal dose). Protected animals had
10   significantly reduced glutathione and increased hepatic lipid peroxidation up to 7 days after
11   treatment, but levels were returned to normal by 14 days.
12
13          Adams et al. (1985) demonstrated that MCLR lethality following i.p. injection in mice
14   was reduced by pretreatment with a single subcutaneous injection of carbon tetrachloride.
15   Lethality was also reduced in young mice in this study (no deaths at 1 and 2 weeks of age, 23/31
16   deaths at 3 weeks of age), suggesting that normal hepatic function is necessary for the uptake
17   and hepatotoxicity of MCLR. The microsomal enzyme inhibitors SKF525A and cobalt chloride
18   produced no effect on MCLR lethality, indicating that microsomal metabolism is not critical for
19   MCLR toxicity.  The administration of hydrocortisone was also shown to protect against MCLR
20   lethality in mice, possibly due to a decrease in the release of arachadonic acid from membrane
21   phospholipids. Naseem et al. (1990) demonstrated that pretreatment of cultured rat hepatocytes
22   with glucocorticoids (flucinolone, dexamethasone and hydrocortisone) reduced the release of
23   arachadonic acid and metabolites caused by MCLR.
24
25          p-Carotene and lutein inhibited the effect of MCLR on hepatocyte morphology in mouse
26   primary hepatocyte cultures (Matsushima-Nishiwaki et al., 1995).  P-Carotene protected the
27   cytokeratin network from disassembly and suppressed the hyperphosphorylation of cytokeratins
28   8 and 18.  Several carotenoid analogs were evaluated, and the protective effect appeared to be
29   related to the number of trans configured double-bonds in the carotenoid.
30
31          4.4.7.8.  Extra-Hepatic Effects of Microcystins
32
33          An isolated perfused kidney model was used to evaluate the kidney toxicity of MCLR
34   (Nobre et al., 1999, 2001). MCLR produced vascular, glomerular and tubular effects in the
35   exposed kidney.  An increase in perfusion pressure was followed by an increase in the
36   glomerular filtration rate (GFR), increased urinary flow rate and a reduction in tubular transport
37   at the proximal tubules.  Histopathological changes included protein in the urinary spaces, but
38   were not further described. Dexamethazone and indomethacin were shown to antagonize the
39   effects of MCLR on perfusion pressure, renal vascular resistance (RVR), GFR and urinary flow.
40    These results suggest a role for phospholipase A2 and cyclooxygenase in the kidney toxicity of
41   microcystins. Nobre et al. (2003) utilized rat peritoneal macrophages exposed to MCLR to
42   further investigate the role of inflammatory mediators in the isolated perfused kidney model.
43   Macrophage supernatants from exposed rats caused an increase in RVR, GFR and urinary flow
44   and reduced Na+ transport. These effects were reduced by cyclohexamide, dexamethasone and
45   quinacrine, further suggesting the involvement of PLA2 and other inflammatory mediators in
46   microcystin-induced kidney toxicity.
                                                66       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2          Moreno et al. (2003) investigated the effects of MCLR on intestinal physiology following
 3   an i.p. injection of 100 ng/kg in rats. Lipid peroxidation was increased in both the serum and the
 4   intestinal mucosa of treated rats. With the exception of sucrase, intestinal brush border enzymes
 5   were unaffected by MCLR exposure.  An increase in the specific activity of acid phosphatase
 6   and succinate dehydrogenase in intestinal homogenates suggests an effect of MCLR on
 7   lysosomal and mitochondrial membranes, respectively.  Nobre et al. (2004) used perfused rat
 8   ileal segments and ligated intestinal loops to evaluate the effect of MCLR on electrolyte and
 9   water secretion.  MCLR caused significant secretion of water, sodium, potassium and chloride.
10   Aziz (1974) observed that a dialyzable component of whole cell lysate fromM aeruginosa
11   caused fluid accumulation in the ligated small intestine of guinea pigs.
12
13          Sicihska et al. (2006) evaluated the effects of MCLR on human erythrocytes in vitro.
14   MCLR exposure resulted in the formation of echinocytes, hemolysis, conversion of
15   oxyhemoglobin to methemoglobin, and a decrease in membrane fluidity. In addition, measures
16   of oxidative stress were affected in treated erythrocytes; glutathione reductase and superoxide
17   dismutase activity were decreased, while ROS and lipid peroxidation were increased.
18
19          Several studies have evaluated the effects of MCLR on immune system components in
20   vitro (Lankoff et al., 2004; Teneva et al., 2005; Chen et al., 2005; Kujbida et al., 2006).  Lankoff
21   et al. (2004) reported that MCLR inhibited B-cell proliferation in human and chicken  peripheral
22   blood lymphocytes at all concentrations tested, and decreased T-cell proliferation only at the
23   highest concentration. Apoptosis was enhanced in both human and chicken lymphocytes
24   (Lankoff et al., 2004). Similarly, MCLR was cytotoxic to mouse splenocytes, and caused
25   apoptosis in B-cells but not in T-cells (Teneva et al., 2005).
26
27          Kujbida et al. (2006) assessed the effects of MCLR and [Asp3]-MCLR on human
28   polymorphonuclear lymphocytes (PMNs) in vitro. Both compounds caused migration of
29   neutrophils in a chemotaxis chamber, suggesting that PMNs may migrate from the blood stream
30   to organs that concentrate microcystins,  such as the liver. In addition, both caused a dose-related
31   increase in ROS production as measured by chemiluminescence of PMN degranulation products
32   that accompany ROS production. The phagocytosis of Candida albicans by PMNs was
33   increased after exposure to either compound, but only MCLR increased the intracellular killing
34   of C. albicans.  These findings suggest the possibility that PMNs may mediate some of the toxic
35   effects of microcystins.
36
37   4.4.8.  Genotoxicity and Cell Proliferation
38
39          Available data give conflicting results when purified MCLR has been tested for
40   mutagenicity.  Pure MCLR did not induce mutations in the Ames assay either with or without
41   metabolic activation, although microcystin-containing extracts did induce mutations (Ding et al.,
42   1999). A crude toxin extracted from M.  aeruginosa did not induce mutations in the Ames assay
43   (Grabow et al., 1982). In contrast, Suzuki et al. (1998) observed increased ouabain resistance
44   mutation frequency in human embryo fibroblast cells treated with MCLR (purity not specified).
45   Similarly, Zhan et al. (2004) reported a 5-fold increase over control in the frequency of
46   thymidine kinase mutations when human lymphoblastoid TK6 cells were treated with
                                                67       DRAFT - DO NOT CITE OR QUOTE

-------
 1   commercially-obtained MCLR. More slow-growing mutants were observed than fast-growing
 2   mutants, suggesting that the mutation damage was larger than the TK locus, and that MCLR
 3   induced large deletions, recombinations or rearrangements. Repavich et al. (1990) reported that
 4   Ames assays (using strains TA98, TA100 and TA102) of a purified hepatotoxin (supplied by
 5   Wright State University and presumed to be microcystin) were negative, as were Bacillus subtilis
 6   multigene sporulation assays.
 7
 8          The conflicting information on mutagenicity may be related to differences in the cell
 9   uptake of MCLR. For example, the failure of MCLR to induce mutations in bacterial cells  may
10   be related  to poor uptake.  Zhan et al. (2004) reported that MCLR is not taken up by many cell
11   types, including bacteria; however, the authors did not provide references to support this
12   assertion.  While hepatocytes take up MCLR at a significant rate, other cell types show limited
13   or no uptake unless measures are taken to enhance the penetration of the cells by MCLR. The
14   cellular uptake of microcystins is discussed in detail in Sections 3.2 (Distribution) and 4.4.7
15   (Mechanistic Studies).
16
17          A number of studies have reported DNA damage after MCLR treatment in vivo (Rao and
18   Bhattacharya, 1996),  and in primary rat hepatocytes (Ding et al., 1999) and human hepatoma
19   cells (Zegura et al., 2003, 2004). Recent studies suggest that apoptosis may be intimately linked
20   to observations of DNA damage in cells treated with MCLR. Lankoff et al. (2004) showed a
21   strong correlation between DNA damage, as measured by the comet assay, and the induction of
22   apoptosis, as measured by the terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick
23   end-labeling (TUNEL) assay, in human lymphocytes. Other evidence has suggested that the
24   comet assay can give a false positive measure of DNA damage when apoptosis is induced, as
25   DNA fragmentation is one consequence of apoptosis induction (Lankoff et al., 2004).  The
26   authors postulated that earlier reports of DNA damage measured by the comet assay may have
27   been related to early stages of apoptosis due to cytotoxicity rather than a direct effect on DNA.
28   The induction of apoptosis appears to be dose-related. Humpage and Falconer (1999)  showed
29   that low (picomolar) concentrations of commercially-obtained MCLR induced cytokinesis and
30   inhibited apoptosis in primary mouse hepatocytes, while higher (nanomolar) concentrations
31   resulted in the inverse effects.
32
33          Boua'icha et al. (2005) reported that noncytotoxic concentrations of MCLR did not cause
34   the formation of hydrophobic DNA adducts in primary cultured rat hepatocytes treated in vitro,
35   but did decrease the amount of endogenous hydrophobic adducts. MCLR was also shown to
36   cause a dose-  and time-dependent increase in the formation of 8-oxo-7,8-dihydro-2'-
37   deoxyguanosine (a measure of oxidative DNA damage) both in cultured hepatocytes and in rat
38   liver cells  after in vivo treatment via i.p. injection (Maatouk et al., 2004; Boua'icha et al., 2005).
39
40          Conflicting results have been reported in studies of MCLR-induced clastogenicity.
41   MCLR (commercially-obtained) has induced micronuclei in human lymphoblastoid cells and
42   mouse bone marrow erythrocytes (Ding et al., 1999; Zhan  et al., 2004). Lankoff et al.  (2004)
43   observed no effect of MCLR on the incidence of chromosomal aberrations in human peripheral
44   blood lymphocytes. Observations of polyploidy in MCLR-treated cells (Humpage and Falconer,
45   1999; Lankoff et al., 2003) may be related to its  effects on cytokinesis. Lankoff et al. (2003)
46   showed that MCLR, through its effect on microtubules, damages the mitotic spindle, leading to
                                               68        DRAFT - DO NOT CITE OR QUOTE

-------
 1   the formation of polyploid cells. Repavich et al. (1990) reported a dose-related increase in
 2   chromosome breakage in human lymphocytes exposed to a purified hepatotoxin (presumed to be
 3   a microcystin).
 4
 5          Mechanistic evidence provides support for the hypothesis that MCLR can act as a
 6   promoter at low doses.  Zhu et al. (2005) reported that MCLR can transform immortalized
 7   colorectal crypt cells, resulting in anchorage-independent growth and enhanced proliferation.
 8   MCLR has been shown to increase the expression of the bcl-2 protein (that inhibits apoptosis)
 9   and decrease the expression of the bax protein (that induces apoptosis) (Hu et al., 2002).
10   Further, MCLR upregulates the transcription factors c-fos and c-jun, leading to abnormal
11   proliferation (Zhao and Zhu, 2003).  Gehringer (2004) reviewed the molecular mechanisms
12   leading to promotion by MCLR and the related tumor promoter, okadaic acid.  Gehringer (2004)
13   reported that MCLR inhibits protein phosphatase PP2A, which regulates several mitogen-
14   activated protein kinases (MAPK). The MAPK cascade regulates transcription of genes required
15   for cell proliferation, including c-jun and c-fos.  In  addition, activation of the MAPK cascade has
16   been postulated to inhibit apoptosis and thus increase cell proliferation. Finally, Gehringer
17   (2004) noted that MCLR has been reported to increase phosphorylation of p53, which is
18   involved in the regulation of the cell cycle and apoptosis as well as the carcinogenic process.
19
20   4.4.9.  Structure-Activity Relationships
21
22          With a few exceptions, microcystin congeners exhibit i.p. LDso values between 50 and
23   300 ng/kg in mice (Rinehart et al., 1994; WHO,  1999).  MCLR is one of the most potent
24   congeners (i.p. LD50 approximately 50 ng/kg). Limited comparative testing of in vitro protein
25   phosphatase inhibition (IC5o) of MCLR, -RR and -YR resulted in IC5o values of 1.6, 3.4 and 1.4
26   nM, respectively (Yoshizawa et al., 1990), indicating that microcystin congeners may be
27   relatively similar in protein inhibition potency.  Pharmacokinetic differences among the various
28   microcystin congeners may be at least partially responsible for observed variations in lethal
29   potency (Ito et al., 2002). Microcystin congeners of varying hydrophobicity were shown to
30   interact differently with  lipid monolayers (Vesterkvist and Meriluoto, 2003).  Effects on
31   membrane fluidity could alter the cellular uptake of these toxins.
32
33          Wolf and Frank (2002) proposed toxicity equivalency factors (TEFs) for the four major
34   microcystin congeners based on LD50 values obtained after i.p. administration. The proposed
35   TEFs, using MCLR as the index compound (TEF=1.0) were 1.0 for MCLA and MCYR and 0.1
36   for MCRR.  The application of TEFs based on i.p. LDso values to assessment of risk from oral or
37   dermal exposure is questionable given that differences in liphophilicity and polarity of the
38   congeners may lead to variable absorption by non-injection routes  of exposure.
39
40          The molecular interaction between microcystins and the catalytic subunits of protein
41   phosphatases has been extensively studied (see Section 4.4.7.3 for more detail). The interaction
42   was shown to occur in two phases. The first phase occurs within minutes and results in rapid
43   inactivation of the phosphatase (Craig et al., 1996). The amino acids Glu and Adda appear to be
44   important for the rapid inactivation of the protein phosphatases and for subsequent toxicity
45   (Harada et al., 1990; Nishiwaki-Matsushima et al.,  1991; Namikoshi et al., 1993; Rinehart et al.,
46   1994; Gulledge et al., 2002, 2003a,b).  The carboxyl group of the Glu residue in MCLR
                                                69        DRAFT - DO NOT CITE OR QUOTE

-------
 1   apparently interacts with a metal (Fe, Mn) ion in the PP1 catalytic site (Mattila et al., 2000). The
 2   Adda side chain is involved in a hydrophobic interaction between the Trp 206 and IlelSO
 3   residues in the hydrophobic groove of PP1 (Herfindal and Selheim, 2006). The few apparently
 4   non-toxic (i.p. LDso values >1000 ng/kg) microcystin congeners exhibit structural alterations in
 5   the Adda or Glu regions (Harada et al.,  1990; Stotts et al., 1993; Rinehart et al., 1994).
 6
 7          The second, slower phase of interaction represents a covalent interaction between the
 8   methylene of the MDha residue of microcystins and the thiol of Cys273 located at the C-terminal
 9   of PP1 that takes place over several hours (Craig et al., 1996).  Microcystin analogues containing
10   a reduced MDha residue are not capable of covalent binding to protein phosphatases
11   (MacKintosh et al., 1995; Craig et al., 1996; Maynes et al., 2006). The importance of covalent
12   bonding to the toxic effect of microcystins is uncertain, as other interactions are apparently
13   responsible for the rapid inactivation of the enzymes (Herfindal and Selheim, 2006).
14   Modifications to either molecule (microcystin or protein phosphatase) to prevent covalent
15   bonding generally decrease, but do not  eliminate the toxic action (Meriluoto et al., 1990;
16   MacKintosh et al., 1995; Hastie et al., 2005; Herfindal and Selheim, 2006).
17
18   4.5.   MODE OF ACTION - NONCANCER AND CANCER
19
20          Microcystins appear to result in different cellular effects depending on dose. In a review
21   of the mechanistic data on microcystins, Gehringer (2004) postulated a dualistic response for
22   microcystins and okadaic acid (another potent inhibitor of PP1  and PP2A). Gehringer (2004)
23   outlined evidence suggesting that at high doses, microcystins cause alterations in cellular
24   structure and function that may lead to cell death via apoptosis  or necrosis, while at low doses,
25   microcystins inhibit apoptosis and cause cell proliferation.  The high-dose effects are likely to be
26   responsible for the acute toxicity and lethality of microcystins and are discussed below.

27   4.5.1.  Target Organ Specificity
28
29          The liver is the primary site of toxicological action for microcystins after oral, i.v., i.p.
30   and intranasal instillation exposure.  Acute and short-term exposures to microcystins have
31   resulted in intrahepatic hemorrhage in both rats and mice (Ito et al.,  1997a; Fawell et al., 1999;
32   Heinze, 1999). One subchronic study in mice also showed liver effects (Fawell et al., 1999),
33   including hepatocyte degeneration, chronic inflammation and hemosiderin deposits, but no
34   hemorrhage.  Liver hemorrhage, resulting from apoptosis and necrosis of hepatocytes leading to
35   disintegration of hepatic architecture, appears to be the most prominent effect observed in
36   available toxicological studies,. The main reason for this target organ specificity is the greater
37   cellular uptake of microcystins by hepatocytes compared with other cells.  Microcystins are
38   actively transported into hepatocytes by the bile acid transporter system, while uptake by other
39   cell types is limited by the lack of an active transport system. In vitro studies demonstrate that
40   preincubation of hepatocytes with compounds that block the uptake of bile acids prevent damage
41   to the hepatocyte from microcystin exposure (Runnegar et al.,  1981, 1993, 1995a;  Runnegar and
42   Falconer,  1982; Eriksson et al., 1990a; Hermansky et al., 1990a,b, 1991).  Treatment of various
43   cell types with microcystins results in rapid damage to hepatocytes, while  changes in other cell
44   types occur after much longer exposure (Khan et al., 1995; Wickstrom  et al., 1995; McDermott
45   etal., 1998).
                                                 70       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2   4.5.2.  Key Events in the Mode of Action for Liver Toxicity and Hemorrhage
 3
 4          Available mechanistic data indicate that the hepatotoxic effects of microcystins begin
 5   with one or two molecular effects:  inhibition of protein phosphatases and induction of oxidative
 6   stress. At present, it is not clear whether the two effects are linked, whether they lead
 7   independently to similar cellular effects (e.g., cytoskeletal damage and apoptosis), or whether
 8   one effect is the predominant cause of hepatocellular damage. In both cases, the initial cellular
 9   effects are associated with cytoskeletal changes and the induction of apoptosis.  These alterations
10   in the structural integrity and function of hepatocytes lead to profound hepatotoxicity and
11   hemorrhage. In summary, the key events in the hepatotoxicity of microcystins appear to be:
12
13       1.  Molecular events (inhibition of protein phosphatase and/or induction of oxidative stress);
14      2.  Cellular effects (cytoskeletal damage and/or apoptosis); and
15      3.  Tissue damage (altered liver structure and function, and intrahepatic hemorrhage).
16
17          The molecular events and cellular effects leading to tissue damage are discussed further
18   below.
19
20          4.5.2.1. Molecular Events
21
22          Protein Phosphatase Inhibition. As discussed above in Section 4.4.7.3, microcystins are
23   potent inhibitors of serine and threonine PP1 and PP2A.  Inhibition of these protein phosphatases
24   results in aberrant phosphorylation of a number of cellular proteins, with the potential  for
25   multiple effects on the cell.  Current data suggest that the inhibition of PP1/PP2A by
26   microcystins can trigger cytoskeletal damage and apoptosis.

27          One outcome of microcystin-induced inhibition of PP1 and PP2A is the
28   hyperphosphorylation of cytokeratin intermediate filament proteins (Falconer and Yeung, 1992;
29   Ohta et al., 1992; Wickstrom et al., 1995; Blankson et al., 2000).  Specifically, microcystin
30   exposure results in hyperphosphorylation of keratins 8 and 18 and desmoplakin I/II (Toivola et
31   al., 1997).  It has been suggested that protein kinases PKC, PKA or the calcium/calmodulin-
32   dependent kinase may play a role in the hyperphosphorylation of these proteins (Gehringer,
33   2004). The hyperphosphorylation of desmoplakin I/II results in the loosening of cell junctions
34   and loss of interactions with cytoplasmic intermediate filaments, while the hyperphosphorylation
35   of keratins 8 and  18 leads to increased solubility. Some of the morphological changes observed
36   in hepatocytes (e.g., blebbing, rounding) may result from the  hyperphosphorylation of
37   cytokeratin intermediate filament proteins.
38
39          Guzman et al. (2003) reported an increase in the phosphorylation of p53 in rat livers after
40   i.p. exposure to MCLR.  PP1 and PP2A help to regulate the activity of p53 through
41   dephosphorylation. Thus, inhibition of these enzymes can result in hyperphosphorylation of
42   p53. Increases in the phosphorylation of p53 can cause an increase in the transcription of
43   p21WAF1, which in turn inhibits cyclin D-, E- and A-dependent kinases.  The result of the latter
44   inhibition is to stall the cell cycle in Gl phase (Gehringer, 2004). This disruption of cell cycling
45   can allow for repair of DNA damage or for apoptosis to occur.  For MCLR, there is evidence of
                                                 71       DRAFT - DO NOT CITE OR QUOTE

-------
 1   hyperphosphorylation of p53, but other steps in this cascade of events have not yet been
 2   investigated.
 3
 4          Oxidative Stress.  A number of studies indicate that oxidative stress may play a role in
 5   microcystin-induced hepatotoxicity. As noted above in Section 4.4.7.5, studies of
 6   chemoprotectants show that several antioxidants can provide protection against the toxicity of
 7   microcystins.  Among the antioxidants shown to protect against the effects of microcystin, either
 8   in vitro or in vivo, are vitamin E, silymarin, dithioerythritol, desferoxamine, N-acetylcysteine,
 9   superoxide dismutase and glutathione. Further, dose and time-dependent increases in reactive
10   oxygen species have been shown to precede morphological changes in hepatocytes, and the
11   addition of superoxide  dismutase prevents the cytoskeletal collapse caused by microcystins.
12   Ding and Ong (2003) have proposed two primary pathways by which microcystins increase
13   oxidative stress leading to cell death.  First, microcystins may deplete glutathione, leading to
14   oxidative damage and cell death. Second, microcystins may increase the production of ROS by
15   disrupting the mitochondrial electron transport chain, leading to mitochondrial permeability
16   transition and apoptosis.
17
18          Microcystins may enhance oxidative stress by altering glutathione homeostasis; however,
19   the importance of glutathione homeostasis in MCLR-induced hepatotoxicity is not clear.
20   Glutathione serves as an intracellular antioxidant, by scavenging free radicals, by serving as a
21   substrate for the reduction of hydrogen peroxide by glutathione peroxidase and by detoxifying
22   xenobiotics. In addition, depletion of glutathione can disrupt microfilament structures in some
23   cell types (Ding and Ong, 2003). MCLR lethality has been prevented in mice by pretreatment
24   with glutathione (Hermansky et al., 1991), and hepatocytes pretreated with a glutathione
25   precursor were likewise protected from MCLR toxicity.   Some studies have shown glutathione
26   depletion after microcystin exposure (Runnegar et al., 1987); however, depletion did not occur
27   until after membrane blebbing had been observed. Other studies have reported an increase in
28   glutathione after MCLR exposure (Ding et al., 2000a; Bou'iacha and Maatouk, 2004).  Finally,
29   Eriksson et al. (1989) indicated that the rapid deformation of rat hepatocytes after MCLR
30   exposure was not associated with changes in glutathione levels. Thus, the role of glutathione
31   homeostasis in MCLR-induced hepatotoxicity has not yet been determined.
32
33          A variety of studies  have demonstrated the importance of mitochondrial permeability
34   transition in the apoptotic cascade induced by MCLR (see Ding and Ong, 2003; Gehringer,
35   2004). In particular, studies have shown that pretreating hepatocytes with cyclosporin A, a
36   specific inhibitor of MPT, prevented cell death from microcystin exposure (Ding and Ong, 2003;
37   Gehringer, 2004). Ding and Ong (2003) outlined the following pathways for MCLR-induced
38   apoptosis via MPT.  First, microcystin disrupts the mitochondrial electron transport chain,
39   leading to the release of reactive oxygen species from mitochondria and mitochondrial
40   permeability transition. MPT triggers a release of cytochrome c and mitochondrial calcium.
41   Cytochrome c may activate one or more caspases that trigger apoptosis, although neither
42   caspase-9 nor caspase-3 appear to be involved based on current information (Ding and Ong,
43   2003; Gehringer, 2004).  The release of mitochondrial calcium activates calpain and
44   calcium/calmodulin-dependent protein kinase II, both of which also lead to apoptosis.
45
                                                72       DRAFT - DO NOT CITE OR QUOTE

-------
 1          4.5.2.2. Cellular Effects
 2
 3          Cytoskeletal Changes.  Morphological changes observed in hepatocytes treated in vitro
 4   with MCLR include membrane blebbing, cell rounding and dissociation.  Membrane blebs
 5   become localized in one region of the cell, and microfilaments are reorganized as a compact
 6   spherical body near the blebbing (Runnegar and Falconer, 1986; Eriksson et al., 1989; Hooser et
 7   al., 1991b; Falconer and Yeung, 1992; Wickstrom et al., 1995; Ding et al., 2000a).  These
 8   morphological changes occur before cell viability or cell membrane integrity is affected.
 9   Electron microscopy of isolated perfused rat liver showed that these cellular effects led to loss of
10   sinusoidal architecture, dilation of bile canaliculi and the space of Disse and decreased
11   intercellular contact (Pace et al., 1991). Intrahepatic hemorrhage results from the breakdown of
12   liver structure, and the liver is rapidly engorged with blood.
13
14          Apoptosis.  Membrane blebbing is also a characteristic of the apoptotic process
15   (Gehringer, 2004). A growing body of evidence indicates that microcystin exposure can trigger
16   apoptosis. Hooser (2000) used several visualization (light and electron microscopy) and
17   analytical techniques (TUNEL and electrophoresis to evaluate DNA laddering) to demonstrate
18   widespread apoptosis in the livers of rats  3 hours after an i.p. dose of MCLR.  Characteristic
19   apoptotic changes including cell rounding, shrinkage, disassociation, loss of microvilli and
20   chromatin margination and condensation  were observed in a majority of hepatocytes. The author
21   postulated that the rapidity with which the apoptotic process occurred  overwhelmed the
22   phagocytic capacity of the liver, such that apoptotic hepatocytes depleted their energy stores and
23   later underwent necrosis. In mice, apoptotic hepatocytes have also been observed, but not to the
24   degree reported in rats. Hooser (2000) postulated that intrahepatic hemorrhage and death
25   occurred so quickly in mice that cellular changes characteristic of apoptosis did not have time to
26   develop.

27   4.5.3.  Conclusion
28
29          The mechanisms by which microcystins induce hepatic damage have not been fully
30   elucidated. Available evidence suggests roles for both protein phosphatase inhibition and
31   oxidative stress as important molecular events, since chemoprotectant studies show that
32   pretreatment with compounds that inhibit either of these effects can protect against
33   hepatotoxicity in MCLR-treated animals. It is possible, even likely, that microcystin exposure
34   triggers a series of independent or linked  events that cause cytoskeletal damage and/or apoptosis,
35   given the numerous cellular functions controlled by PP1 and PP2A, as well as the number of
36   effects triggered by oxidative stress. These cytoskeletal and apoptotic changes apparently lead
37   to the altered hepatic structure/function and intrahepatic hemorrhage observed in animal studies.
                                                 73       DRAFT - DO NOT CITE OR QUOTE

-------
 1   4.6.   SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS
 2
 3          The preponderance of toxicological data on the effects of microcystins is restricted to the
 4   MCLR congener. A single, poorly-described study, reported only in a secondary source is
 5   available for the LA congener. Data on the YR and RR congeners are limited to i.p. LD50 values
 6   and measures of relative inhibition of protein phosphatases. As a result, this section largely
 7   describes the available information on the toxic effects of MCLR, with limited reference to other
 8   congeners.
 9
10          Anecdotal reports indicate that, in humans, exposure to cyanobacterial blooms (including
11   microcystin-producing genera) can result in neurological, gastrointestinal and dermatological
12   symptoms, such as headache; muscle weakness; eye, ear and throat irritation; nausea; stomach
13   pain; diarrhea; blistering around the mouth; and hay-fever like symptoms (Dillenberg and
14   Dehnel, 1960; Billings, 1981; Turner et al., 1990; Teixeira et al., 1993; el Saadi and Cameron,
15   1993). Effects were reported in persons exposed via recreational contact (swimming, boating)
16   and drinking water. Turner et al. (1990) also reported pneumonia in army recruits exposed to a
17   cyanobacterial bloom.  Symptoms occurring after exposure to cyanobacteria cannot be directly
18   attributed to microcystin toxins (or other endotoxins); some effects may result from exposure to
19   the cyanobacterial cells themselves, or from exposure to multiple toxins in the bloom.
20
21          The primary noncancer health effect of exposure to MCLR is liver damage. The liver is
22   targeted largely because hepatocytes are among only a few cell types that actively take up
23   microcystins, which do not readily cross the cell membrane.  Severe liver damage (diffuse
24   individual hepatocyte necrosis, cell-plate disruption and apoptosis) occurred in dialysis patients
25   exposed to microcystins3 in dialysate (Jochimsen et al., 1998; Pouria et al., 1998; Carmichael  et
26   al., 2001; Azevedo et al., 2002).  At high  acute doses in laboratory animals, MCLR caused
27   potentially fatal hemorrhaging within the liver. While the liver is the usual target of microcystin
28   toxicity, there have been some reports of effects in other systems, including hematological,
29   kidney, cardiac, neurological and gastrointestinal effects. It has been suggested that some effects
30   in other organs observed after high doses of MCLR may result from ischemia or hypoxia caused
31   by hepatic hemorrhage. However, some effects outside the liver have been observed in the
32   absence of hemorrhage.
33
34          Much of the toxicological data on microcystins are limited to reports of liver effects after
35   single lethal or sublethal doses administered via i.p. injection. These studies indicate that
36   injected doses of 50-200 |J,g/kg MCLR or MCYR are usually lethal in mice and rats within a few
37   hours (Adams et al., 1988; Hooser et al., 1989a; Hermansky et al., 1990c;  Stotts et  al., 1993;
38   Gupta et al., 2003; Rao et al., 2005).  Miura et al. (1991) showed that the median time to death is
39   greatly increased in fed rats (32 hours) when compared with fasted rats (less than 2 hours).  The
40   authors suggest that fasting may increase the sensitivity of animals to the mitochondrial toxicity
41   of microcystins, although this could not be conclusively demonstrated. In the liver, MCLR
42   destroys the cytoskeleton of hepatocytes,  leading to hepatocyte disassociation, degeneration,
     3 Exposure was to untreated water containing cyanobacteria. The presence of microcystins was confirmed in patient
     biopsy samples; however, it is possible that the patients may also have been exposed to other microbial or chemical
     contaminants.


                                                 74        DRAFT - DO NOT  CITE OR QUOTE

-------
 1   apoptosis and necrosis (Hermansky et al., 1990c; Hooser et al., 1991b). Hepatic hemorrhage and
 2   disintegration of the liver architecture follow quickly (Hooser et al., 1991b). Effects reported to
 3   occur outside the liver include pulmonary thrombi derived from necrotic hepatocytes, kidney
 4   effects such as dilation of cortical tubules and eosinophilic material in the cortical tubules, and
 5   degeneration and necrosis of myocardial cells (Adams et al., 1988; LeClaire et al.,  1988; Zhang
 6   et al., 2002). As previously stated, some of these effects may occur secondary to hepatic
 7   hemorrhage.
 8
 9          Injection studies suggest a very steep dose-response curve for acute liver effects from
10   microcystin exposure. In several studies, mice and rats receiving single i.p. doses of 20-40
11   Hg/kg MCLR showed no clinical toxicity and few or no gross or microscopic effects in the liver
12   or other organs (Hooser et al., 1989a; Lovell et al., 1989a; Hermansky et al., 1990c), while i.p.
13   doses of 50-200 ng/kg are usually lethal within a few hours (Hooser et al., 1989a; Hermansky et
14   al.,  1990c;  Stotts et al., 1993; Gupta et al., 2003).
15
16   4.6.1.  Oral
17
18          Table 4-11  provides a summary of the noncancer effects from repeated-dose oral studies
19   of MCLR toxicity in laboratory animals.  The table includes all of the studies that used purified
20   microcystins as the test substance. As the table indicates, the toxicological database for effects
21   of microcystins after oral exposure is limited.
22
23          Liver Effects.  One study of human exposure to drinking water before, during and after a
24   bloom of M. aeruginosa reported a significant increase in GGT levels during the bloom
25   compared with levels before the bloom (Falconer et al.,  1983).  The study population consisted
26   of all persons subjected to liver function tests in the area served by the affected drinking water
27   supply; as such, it is not representative of the general population.  The liver is the primary  target
28   organ when laboratory animals are exposed to high doses of MCLR. Oral exposure to single 500
29   Hg/kg doses of MCLR caused diffuse hemorrhage in the liver of mice and rats; more pronounced
30   liver damage occurred at higher doses (Ito et al.,  1997a; Fawell et al., 1999). Young mice  (5
31   weeks old) did not develop signs of hepatotoxicity at 500 ng/kg MCLR, while aged mice (32
32   weeks old) developed clear signs (Ito et al.,  1997a). This difference may result in part from
33   differences in gastrointestinal absorption of microcystins, but cannot be entirely explained by
34   absorption  differences, since similar age-dependent effects were reported after i.p.  exposure
35   (Adams et al., 1985; Rao et al., 2005).
36
37          A single 28-day study of oral exposure to 50 or 150 ng/kg MCLR in drinking water
38   showed increased liver weight, slight to moderate liver lesions with hemorrhages and increased
39   ALP and LDH in rats exposed at 50  |j,g/kg-day (Heinze, 1999). A subchronic study in mice
40   using a similar dose range identified a LOAEL of 200 ng/kg (Fawell et al., 1999).  At this  dose,
41   mild liver lesions including chronic inflammation, hemosiderin deposits and single hepatocyte
42   degeneration were observed, as well as increased ALT and AST in male animals. The 40 ng/kg
43   dose was identified as a NOAEL.  Mild hepatocyte injury was reported in mice given 80 or 100
44   gavage doses of 80 ng/kg each over  28 weeks, corresponding to time-weighted average doses of
45   33-41 |j,g/kg-day (Ito et al., 1997b).  Based on the report, it appears that a limited postmortem
                                                75        DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-11. Summary Noncancer Results in All Animal Studies of Oral Exposure to Purified Microcystin-LR
Species
Sex
Average
Daily Dose
(|ag/kg-day)
Exposure
NOAEL
Og/kg-
day)
LOAEL
(Hg/kg-
day)
Responses
Comments
Reference
Acute Exposure
Rat
Mouse
Mouse
Mouse
M/F
M/F
F
M
500, 1580,
5000
500, 1580,
5000
8000,
10000,
12500
0,500
Single
gavage
Single
gavage
Single
gavage
Single
gavage
ND
ND
ND
ND
5000*
1580*
12500*
500 (aged
mice
only)
Mortality; diffuse hepatic
hemorrhage at lower doses
Mortality; diffuse hepatic
hemorrhage at lower dose
Mortality (2/2); hypertrophic
hepatocytes, fibrosis in
centrilobular and midzonal
regions at lower doses
Centrilobular hepatic
hemorrhage and necrosis;
necrosis of intestinal mucosa
and duodenal damage
No untreated controls. Dose-
dependent increase in
hepatotoxicity
No untreated controls. Dose-
dependent increase in
hepatotoxicity
No untreated controls. 1-2
animals/dose group.
Effects observed in aged (32
week-old) mice; no effects on
liver or gastrointestinal tract in
young (5 week-old) mice
Fawell et al.,
1999
Fawell et al.,
1999
Yoshidaetal.,
1997
Ito et al.,
1997a
Short-Term Exposure
Rat
M
0, 50, 150
Drinking
water, 28
day
ND
50
Slight to moderate
degenerative and necrotic
hepatocytes with
hemorrhages; increased serum
enzymes (ALP and LDH)

Heinze, 1999
76
DRAFT - DO NOT CITE OR QUOTE

-------
Table 4-11. cont.

Species


Sex

Average
Daily Dose
(|ag/kg-day)

Exposure

NOAEL
(|4,g/kg-
day)
LOAEL
(Hg/kg-
day)

Responses


Comments


Reference

Subchronic Exposure
Mouse





M/F





0, 40, 200,
1000




Daily
gavage,
13 weeks



40





200





Minimal/slight chronic
inflammation with
hemosiderin deposits and
single hepatocyte
degeneration; increased serum
enzymes (ALT and AST)






Fawell el al.,
1999




Chronic Exposure
Mouse



Mouse




F



Not
given



0,3



Not
available



Drinking
water, 18
months

Gavage,
80 (o,g/kg,
80-100
times over
28 weeks
3



ND




ND



ND




No effects on survival, body
weight, hematology, serum
biochemistry, organs or
histopathology
Light injuries to hepatocytes
in the vicinity of the central
vein


Minor changes in ALP and
cholesterol not considered
lexicologically significant by
researchers
Only liver examined; only Ihree
conlrol animals; dosing
frequency unclear


Ueno el al.,
1999


Ilo el al.,
1997b



Developmental Toxicity
Mouse




F




0, 200, 600,
2000



Gavage,
GD 6-15



600




2000*




Maternal mortality (7/26) and
morbidity (2/26 humanely
sacrificed); reduced fetal body
weight, delayed skeletal
ossification
Aulhors defined 600 |ag/kg-day
as NOAEL bul did nol presenl
dala on reproductive or
developmental parameters lo
support identification of LOAEL
Fawell el al.,
1999



: Frank Effect Level (PEL)
                                                     77
DRAFT - DO NOT CITE OR QUOTE

-------
 1   examination was conducted in this study, which was primarily aimed at evaluating
 2   carcinogenicity. No liver or other toxicity was reported in female mice given approximately 3
 3   |j,g/kg-day MCLR in drinking water for 18 months (Ueno et al., 1999).
 4
 5          Neurological Effects.  The database contains scattered reports of neurological symptoms
 6   after exposure to high doses of MCLR.  Dialysis patients exposed to microcystins in dialysate
 7   reported symptoms such as visual disturbance, blindness, vertigo, headache and muscle
 8   weakness (Jochimsen  et al., 1998).  Clinical signs in mice and rats orally exposed to lethal doses
 9   (about 5000 |J,g/kg) include hypoactivity and piloerection (Fawell et al., 1999).
10
11          Other Organs. Gastrointestinal effects (necrosis, duodenal damage) were observed in
12   aged mice exposed orally to single 500 ng/kg doses of MCLR (Ito et al., 1997a).  Kidney effects
13   including eosinophilic materials in the Bowman's spaces were observed in two mice exposed to
14   a lethal dose of 12.5 mg/kg (Yoshida et al., 1997).  Female mice exposed subchronically to 1000
15   Hg/kg had slight increases in hemoglobin concentration, erythrocyte count and packed cell
16   volume (Fawell et al., 1999).  Milutinovic et al. (2002, 2003) briefly reported that kidney effects
17   are more pronounced than liver effects in rats chronically exposed to i.p. doses of MCLR and
18   MCYR (time weighted average dose,  5 ng/kg for 8 months). Details of the liver examinations
19   were not reported in this study, limiting the usefulness of these data.
20
21          Developmental Effects. A single oral study of developmental toxicity in mice reported
22   maternal toxicity, liver effects and deaths in some dams treated at the highest dose of MCLR
23   (2000 |J,g/kg during GD 6-15), along with reduced fetal body weight and delayed skeletal
24   ossification. No effects on reproductive or developmental parameters were observed in other
25   treatment groups, and 600 ng/kg was identified as a NOAEL for developmental toxicity (Fawell
26   et al., 1999).  One study of developmental toxicity after i.p. injection of 32-238 ng/kg MCLR in
27   mice confirmed the lack of developmental or reproductive effects in the absence of maternal
28   toxicity (Chernoff et al., 2002).  A study in which an extract ofM aeruginosa (estimated to
29   contain about 14 |J,g/L unspecified toxin) was administered in the drinking water to mice before
30   and during pregnancy revealed small brains in 7 of 73 pups from  treated parents  and none in
31   untreated controls (Falconer et al., 1988).  The litter distribution of the affected pups was not
32   reported by the authors.  It is not possible to attribute this effect to microcystin exposure, as the
33   extract may have contained other compounds.

34          In vitro studies suggest that MCLR can disrupt the cytoskeleton of embryonic cells,
35   causing cell detachment, retarding division or causing cytolysis (Sepulveda et al., 1992; Frangez
36   et al., 2003; Zuzek et al., 2003). MCLR effects on these and other cell types may be limited by
37   the degree of uptake.  Frangez et al. (2003) showed that an intact  zona pellucida prevented
38   effects in rabbit whole embryo cultures.
39
40   4.6.2.  Inhalation
41
42          Very limited information is available on the toxicity of MCLR via inhalation exposure.
43   The available data indicate that short-term inhalation of a low concentration of MCLR can cause
44   local damage to the epithelial  cells of the nasal cavity.  A single study of inhalation exposure in
                                                78        DRAFT - DO NOT CITE OR QUOTE

-------
 1   mice revealed dose-dependent damage to the respiratory and olfactory epithelial cells of the
 2   nasal cavity (Benson et al., 2005). Exposure occurred over 7 days at 260 ng/m3 for 30, 60 and
 3   120 minutes per day (authors estimated the deposited doses as 3, 6 and 12 ng/kg).  No effects on
 4   the liver or other organs were observed.
 5
 6          Several limited lines of evidence suggest that high doses of MCLR via respiratory
 7   exposure routes can lead to systemic uptake with subsequent liver effects. Systemic uptake of
 8   MCLR by respiratory routes of exposure has been demonstrated in studies of acute, high-dose
 9   exposure (Creasia, 1990; Fitzgeorge et al., 1994; Ito et al., 2001). Importantly, the LD50 for
10   MCLR given via either intranasal or intratracheal instillation is similar to that of MCLR given
11   via i.p. injection (Fitzgeorge et al., 1994; Ito et al., 2001).  As with i.p. and oral exposure, liver
12   hemorrhage is the proximate cause of death in animals lethally dosed via intranasal or
13   intratracheal instillation (Fitzgeorge et al., 1994; Ito et al., 2001). Further evidence of systemic
14   effects comes from a brief abstract describing lethality in mice exposed via inhalation (nose
15   only) to MCLR aerosols. Creasia (1990) reported an LC50 for MCLR of 18 mg/m3 air for 10
16   minutes (authors estimated the deposited dose as 45 ng/kg), and indicated that histopathological
17   findings in deceased mice were similar to those reported after i.v. dosing.  Ito et al. (2001)
18   suggested that MCLR could enter the bloodstream either via local damage to the nasal mucosa
19   leading to exposure of the nasal blood vessels, or through transport to the lung and absorption
20   into  alveolar capillaries.
21
22   4.7.    WEIGHT-OF-EVIDENCE EVALUATION AND CANCER
23          CHARACTERIZATION
24
25   4.7.1.  Summary of Overall Weight of Evidence
26
27          Applying the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), there is
28   inadequate information to assess carcinogenic potential of microcystins by the oral, dermal or
29   inhalation routes of exposure. One poorly-described long-term carcinogenicity bioassay found
30   no increase in neoplastic liver nodules after gavage administration in mice (Ito et al., 1997b).
31   The few available epidemiological studies that suggest a positive association between liver or
32   colorectal cancers and microcystins are limited by ecological study design, poor measures of
33   exposure, potential coexposure to other microbial or chemical contaminants and, in some cases,
34   failure to control for known liver and colorectal risk factors.  MCLR has been shown to have a
35   promotional effect in two-stage rat liver bioassays using i.p. administration; however, the
36   relevance of this effect to environmental exposures is uncertain. Mechanistic information
37   provides some support for a possible promotional effect of MCLR.
38
39   4.7.2.  Synthesis of Human, Animal and Other Supporting Evidence
40
41          Several human epidemiological studies have reported an association between
42   consumption of drinking water containing cyanobacteria and microcystins and liver or colon
43   cancer in certain areas of China (Yu et al., 1989 and Yu, 1989 as cited in Ueno et al., 1996; Zhou
44   et al., 2002).  In all of these studies, the use of a surface drinking water supply was used as a
45   surrogate for exposure to microcystins. Individual exposure to microcystins was not estimated.
                                                79        DRAFT - DO NOT CITE OR QUOTE

-------
 1   Further, it is not clear whether these studies adequately controlled for confounding factors, such
 2   as hepatitis infection or aflatoxin exposure.
 3
 4          Ito et al. (1997b) conducted the only study of oral carcinogenicity of a purified
 5   microcystin. In this study, chronic gavage doses of MCLR over 28 weeks failed to induce
 6   neoplastic nodules of the liver in mice. Limited information from two-stage, medium-term rat
 7   liver bioassays where MCLR was administered i.p. suggest that MCLR can act as a promoter,
 8   increasing the  number and/or size of GST (placental form) positive foci in livers of rats
 9   pretreated with an initiating agent (Nishiwaki-Matsushima et al.,  1992; Ohta et al., 1994;
10   Sekijima et al., 1999; Hu et al., 2002). In one such study, MCLR alone showed no initiating
11   activity (Ohta  et al., 1994).  Ito et al. (1997b) observed an increase in the size of neoplastic liver
12   nodules in mice given 100 i.p. injections of MCLR without an initiating agent; however, the
13   numbers of treated and control animals were small.
14
15          Studies of cyanobacterial extract  also suggest a possible promotional effect.  In mice
16   given an extract of M. aeruginosa in drinking water, the mean area of aberrant crypt foci of the
17   colon was significantly increased, although the number of foci was not affected (Humpage et al.,
18   2000). Similarly, the total weight of skin tumors was increased in mice given an extract of
19   Microcystis in drinking water after topical DMBA pretreatment (Falconer and Buckley, 1989;
20   Falconer, 1991).  It is not possible to determine whether the observed effects resulted from
21   exposure to microcystins or to other contaminants in the extracts.
22
23          Mechanistic data indicate that at low doses, MCLR may increase cell proliferation.
24   MCLR has been shown to increase the expression of the bcl-2 protein (that inhibits apoptosis)
25   and decrease the expression of the bax protein (that induces apoptosis) (Hu et al., 2002).
26   Further, MCLR upregulates the transcription factors c-fos and c-jun, leading to abnormal
27   proliferation (Zhao and Zhu, 2003).  Gehringer (2004) reviewed the molecular mechanisms
28   leading to promotion by  MCLR and the related tumor promoter, okadaic acid. Gehringer (2004)
29   reported that MCLR inhibits protein phosphatase PP2A, which regulates several MAPKs. The
30   MAPK cascade regulates transcription of genes required for  cell proliferation, including c-jun
31   and c-fos. In addition, activation of the MAPK cascade has been postulated to inhibit apoptosis
32   and thus increase  cell proliferation. Finally, Gehringer (2004) noted that MCLR has been
33   reported to increase phosphorylation of p53, which is involved in the regulation of the cell cycle
34   and apoptosis.

35          Genotoxicity studies of MCLR have given conflicting results, with negative findings in
36   Ames assays (Grabow et al., 1982; Ding et al.,  1999) while positive results were observed with
37   human cell lines (Suzuki et al., 1998; Zhan et al., 2004). Evidence for MCLR-induced DNA
38   damage as measured by the comet assay  has been called into question by the finding that
39   apoptosis can lead to false positive findings in this assay (Lankoff et al., 2004). There is some
40   evidence for a clastogenic effect of MCLR (Ding et al., 1999; Zhan et al., 2004).
41
                                                80        DRAFT - DO NOT CITE OR QUOTE

-------
 1    4.8.    SUSCEPTIBLE POPULATIONS AND LIFE STAGES
 2
 3           Little information is available on potentially susceptible populations.  Studies in
 4    laboratory rodents suggest that the acute effects of MCLR may be more pronounced in adult or
 5    aged animals than in juvenile animals (Adams et al., 1985; Ito et al., 1997a; Rao et al., 2005). In
 6    these studies, young animals showed little or no effect at MCLR doses lethal to adult animals.
 7    Age-dependent differences in toxicity were observed after both oral and i.p. exposure,
 8    suggesting that differences in gastrointestinal uptake were not entirely responsible for the effect
 9    of age. The relevance of these age-related differences to acute toxicity in humans is unclear.
10
11           Available information does not suggest any pronounced gender differences in response to
12    microcystins.  Studies with algal extracts suggest the possibility that male mice may be more
13    sensitive than female mice to oral exposure to algal extracts (Falconer et al., 1988).  However,
14    the relevance of this finding to human microcystin exposure is uncertain given the potential for
15    coexposure to  other contaminants in algal extracts.
16
17           Because microcystins inhibit the action of protein phosphatases (PP1  and PP2A),
18    coexposure to  other compounds that inhibit these enzymes (for example, okadaic acid) may
19    enhance the toxicological effects of microcystins.
                                                81       DRAFT - DO NOT CITE OR QUOTE

-------
 1                            5.  DOSE RESPONSE ASSESSMENTS
 2
 3
 4   5.1.    NARRATIVE DESCRIPTION OF THE EXTENT OF THE DATABASE
 5
 6          The available information on the toxicokinetic behavior of microcystins in humans or
 7   animals after oral or inhalation exposure is limited to a single study of the organ distribution of
 8   dihydro-MCLR. No other data are available on the absorption, distribution, metabolism or
 9   elimination of microcystins via environmentally relevant exposure routes.  Acute lethality data
10   show a significant difference in lethal doses via injected and oral routes of exposure, suggesting
11   that the toxicokinetic behavior of microcystins is an important determinant of health effects after
12   oral exposure; thus, the deficiencies in this category of data are significant.
13
14          The mode by which microcystins affects its primary target organ, the liver, is remarkably
15   well-studied. There are abundant mechanistic data ranging from target organ specificity down to
16   molecular targets. In vitro studies using human hepatocytes show effects similar to those in
17   animal hepatocytes, indicating that the mode of toxicological action is similar. Human
18   hepatocytes appear to be more susceptible to the action of MCLR than rat hepatocytes (Yea et
19   al., 2001; Batista et al.,  2003). Further evidence for the relevance of this mode of toxicological
20   action to humans comes from reports of human exposure. Liver histopathology on humans
21   exposed to MCLR via dialysate showed effects similar to those seen in animals,  although
22   intrahepatic hemorrhage was not observed (Azevedo et al., 2002).
23
24          The toxicological database for microcystins is almost exclusively limited to  data on a
25   single congener, MCLR. Data on the other congeners is restricted to  in vitro studies of protein
26   phosphatase inhibition,  i.p. LD50 measures and a single, poorly-described toxicological
27   evaluation of MCLA in primates. The database on the oral toxicity of MCLR is adequate to
28   support the derivation of RfD values. Human data on the oral toxicity of MCLR are limited by
29   potential co-exposure to other cyanobacterial toxins and microorganisms. There are three
30   studies of acute oral exposure to MCLR in two laboratory animal species; however, none of
31   these identified a NOAEL, and the minimum dose tested was the same for all three. Further,
32   neither of the two experiments published in Fawell et al. (1999) nor the experiment  by Yoshida
33   et al. (1997) used an untreated control group. One animal study evaluated the oral toxicity of
34   MCLR after short-term (<30 days) exposure and and one after subchronic (30-90 days)
35   exposure.  The short-term study (Heinze, 1999) used a small number  of animals  (10/dose) and
36   did not identify a NOAEL, but was otherwise of good study quality.  The subchronic study
37   (Fawell et al., 1999) used an adequate number of animals (30/dose) and identified both a
38   NOAEL and LOAEL. Two chronic exposure studies are available, but one of these (Ito et al.,
39   1997b) apparently conducted only a limited examination of the liver for toxicity, and the other
40   (Ueno et al., 1999) used a single dose and did not identify a LOAEL.  A single, well-conducted
41   developmental toxicity  study in the mouse is available; however, the  results are presented  only
42   briefly and without any supporting data in the publication (Fawell et al., 1999).  The noncancer
43   database is missing a chronic toxicity study in a second species, as well as a multigeneration
44   reproductive toxicity study and neurotoxicity study.
45
                                                82       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
       The database on the inhalation toxicity of MCLR is inadequate for the derivation of any
RfC.  There are no human data on the inhalation of MCLR.  There is a single well-reported
animal study addressing inhalation exposure to MCLR for 7 days (Benson et al., 2005).  This
study used only one exposure concentration with daily exposure for 30, 60 or 120 minutes and,
as such, is not adequate for short-term RfC derivation.

       The available data on carcinogenicity are inadequate for carcinogenicity assessment.
There is no well-conducted long-term carcinogenicity bioassay for microcystin.  Several studies
using an initiation-promotion protocol are available, as are limited mechanistic data suggesting a
potential promoting effect of microcystins.

5.2.    ORAL REFERENCE DOSE

       Data considered in deriving oral reference dose for each exposure duration are
summarized in the following exposure-response array (Figure 5-1) as well as in Table 5-1 below.
Due to the limited toxicological database for microcystins, both the table and the figure include
all studies in the published literature that examined the oral toxicity of purified MCLR in
laboratory animals, with one exception. The publication by Ito et al. (1997b) did not provide the
dosing frequency  or information to estimate an average daily dose; thus, this study is not
included in the table or figure.
        100000
              10000
               100
                10

1
Ra


1
1
ts



Mi

1
ce
• FEL
• LOAEL
ONOAEL

Fawell etal.
(1999)


Fawell etal.
(1999)
Ac

The vertical lines
represent the range
^ of dos^s t^st^d in a
given study.
Mice

* °
Aged ^
i'licc Mice
Rats Mice
o
Mice
Yoshidaet Ito etal. Heinze Fawell etal. Ueno et al. Fawell etal.
al. (1997) (1997a) (1999) (1999) (1999) (1999)
Lite Short-term Longer-term Chronic Develop-
mental
22   Figure 5-1. Exposure Response Array for Oral Exposure: All Studies of Purified Microcystin-
23   LR
                                                83
                                                     DRAFT - DO NOT CITE OR QUOTE

-------
Table 5-1. Available Dose-Response Information for Oral Exposure to Purified MCLR
Species
Sex
Average Daily
Dose
Og/kg-day)
Exposure
NOAEL
(Hg/kg-
day)
LOAELorFEL
Og/kg-day)
Reference
Acute Exposure
Rat
Mouse
Mouse
Mouse
M/F
M/F
F
M
500, 1580, 5000
500, 1580, 5000
8000, 10000,
12500
0,500
Single gavage
Single gavage
Single gavage
Single gavage
ND
ND
ND
ND
5000*
1580*
12500*
500 (aged mice
only)
Fawelletal., 1999
Fawelletal., 1999
Yoshidaetal.,
1997
Itoetal., 1997a
Short-Term Exposure
Rat
M
0, 50, 150
Drinking
water, 28 d
ND
50
Heinze, 1999
Subchronic Exposure
Mouse
M/F
0, 40, 200, 1000
Daily gavage,
13 weeks
40
200
Fawelletal., 1999
Chronic Exposure
Mouse
F
0,3
Drinking
water, 18
months
o
J
ND
Uenoetal., 1999
Developmental Toxicity
Mouse
F
0, 200, 600, 2000
Gavage,
GD 6-15
600
2000*
Fawell et al., 1999
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
 : Frank Effect Level (PEL)
5.2.1.  Acute Oral RfD

       The acute oral data for MCLR are inadequate for the derivation of an acute RfD. There
are four studies of acute exposure to MCLR (Yoshida et al., 1997; Ito et al., 1997a; Fawell et al.,
1999); however, none of the available studies identified a NOAEL.  Yoshida et al. (1997)
conducted a LDso determination using a small number of mice (5) treated with a single gavage
dose, and no untreated controls. Histopathology of surviving mice (one at 8 mg/kg and two at 10
mg/kg) showed liver lesions (hypertrophic hepatocytes with centrilobular and midzonal fibrosis).
The high dose in this study was an PEL based on deaths of the 2 treated animals.  Fawell et al.
                                               84
                                                   DRAFT - DO NOT CITE OR QUOTE

-------
 1   (1999) administered single gavage doses of 500, 1580 or 5000 ng/kg MCLR to groups of rats
 2   and mice (10/dose/species). There were no untreated control groups in this study. In both
 3   species, diffuse hepatic hemorrhage was observed at the low dose (500 ng/kg); however, in the
 4   absence of untreated controls for comparison, this dose cannot be identified as a LOAEL. Ito et
 5   al. (1997a) administered single gavage doses of 500 ng/kg to young (5 weeks old) and aged (32
 6   weeks old) mice. Centrilobular hemorrhage and hepatocyte necrosis, as well as gastrointestinal
 7   lesions, were observed in the aged mice, while no effect was observed in the young mice. This
 8   study identifies a freestanding LOAEL. The absence of an acute study of adequate quality
 9   precludes derivation of an acute oral RfD.
10
11   5.2.2.  Short-Term Oral RfD
12
13          5.2.2.1. Choice of Principal Study and Critical Effect
14
15          A single short-term study and a single developmental toxicity study of orally-
16   administered MCLR are available. Heinze (1999) evaluated the effects of MCLR in drinking
17   water in 11-week-old male hybrid rats. Groups of 10 rats (5 of each sex) were given
18   approximate doses of 0, 50 or 150 ng/kg body weight for 28 days.  Serum biochemistry showed
19   significantly increased mean levels of ALP and LDH in both treatment groups (84 and 100%
20   increase in LDH, 34 and 33% increase in ALP in low and high doses, respectively).  A dose-
21   dependent increase in relative liver weights was observed at both dose levels (17 and 26% at the
22   low and high doses, respectively).  Liver lesions were observed in both treatment groups, but the
23   severity of the damage was increased in the 150 ng/kg dose group.  Moderate to severe
24   degenerative and necrotic hepatocytes with hemorrhage was observed in 0 of 10 controls, 6 of 10
25   low-dose and 9 of 10 high-dose rats.
26
27          Fawell et al. (1999) evaluated the developmental toxicity of MCLR administered via
28   gavage to mice at doses of 0, 200, 600 and 2000 ng/kg  on GDs 6-15.  Seven of 26 dams
29   receiving 2000 ng/kg died and two others were sacrificed prematurely due to morbidity.  At this
30   dose, fetal body weight was significantly lower than controls and there was delayed skeletal
31   ossification; these effects may have been associated with maternal toxicity. Data on
32   reproductive and developmental parameters were not provided in the reference; thus,  a LOAEL
33   for developmental toxicity could not be determined.  This study identified a NOAEL  of 600
34   |j,g/kg-day for both developmental and maternal effects. The high dose of 2000 ng/kg was an
35   FEL based on maternal mortality. The study by Heinze (1999) identified a lower LOAEL and
36   more sensitive effect (hepatotoxicity) than the developmental  toxicity study did; thus, this study
37   was chosen as the basis for the short-term RfD.
38
39          5.2.2.2. Methods of Analysis
40
41          Liver toxicity observed by Heinze (1999) included liver lesions, serum enzyme changes,
42   and changes in relative liver weight. All three of these endpoints were considered for
43   determining the point of departure for RfD derivation.
44
                                                85       DRAFT - DO NOT CITE OR QUOTE

-------
 1          All quantal models in U.S. EPA's Benchmark Dose (BMD) software (version 1.3.2) were
 2   fit to the incidence data for liver lesions in rats (males and females combined) exposed to MCLR
 3   in the drinking water for 28 days (Heinze, 1999).  The incidence data for liver lesions are
 4   reported in Table 4-4 (Section 4.2.1.2.1). Of the liver lesions observed, the category of
 5   degenerative and necrotic hepatocytes with hemorrhage showed a strong dose-related trend with
 6   greater incidence and severity with higher dose, and no control animals were affected. As the
 7   table shows, at 50 ng/kg, 4/10 rats had slight lesions and 6/10 had lesions of moderate severity.
 8   At 150 ng/kg, 6/10 rats had moderate lesions and 3/10 had intensive damage. For BMD
 9   modeling, the moderate and severe lesion categories were collapsed into one. The data modeled
10   are shown in Table 5-2.
11
Table 5-2. Incidence of Liver Lesions Used for BMD Modeling (Heinze, 1999)

Lesion incidence
0 ng/kg-day
0/10
50 |j,g/kg-day
6/10
150 |j,g/kg-day
9/10
12
13
14          In accordance with the U.S. EPA (2000c) BMD methodology, the default benchmark
15   response (BMR) of 10% increase in extra risk was used. The high response rate (60%) at the
16   lowest dose with a positive response precludes the use of a lower BMR for this analysis.  Models
17   were run using the default restrictions on parameters built into the BMDS. The modeling results
18   are shown in Table 5-3.  Adequate fits were achieved with all models, except the quantal
19   quadratic. While the gamma, multistage, quantal linear and Weibull all converged on the same
20   model, the log probit model gave the best fit,  as assessed by AIC. Figure 5-2 shows the fit of the
21   log probit model to the data.  Appendix A contains the full model outputs. The BMD and
22   BMDL estimated by the log probit model for  the liver lesion data are 11.0 and 6.4 |j,g/kg-day,
23   respectively.
24
25          The linear model for continuous data was fit to the increased LDH and ALP levels
26   reported in Heinze (1999). These data are shown in Table 4-3 (Section 4.2.1.2.1). With only
27   three observations  in each of these datasets, there were not enough data points to use the
28   remaining models (polynomial, power, or Hill), which each have more than three parameters.4
29
30          The linear model did not provide adequate fit to either dataset as measured by goodness-
31   of-fit criteria (see Appendix A for model outputs). The linear model was also fit to the relative
32   liver weight changes (also three observations) reported in Heinze (1999). These data are shown
33   in Table 4-3 (Section 4.2.1.2.1). In accordance with the U.S. EPA methodology, the default
34   BMR of one standard deviation change from the control mean was used, and the polynomial
35   coefficients were restricted to be positive.  The linear model provided an adequate fit to the data
36   (see Appendix A for model output).  The BMD and BMDL estimated by the linear model for the
37   relative liver weight data are 85  and 58 |j,g/kg-day, respectively.
     4 The number of parameters describing the shape of the dose-response curve cannot exceed the number of dose
     groups (U.S. EPA, 2000c).
                                                86       DRAFT - DO NOT CITE OR QUOTE

-------
Table 5-3. BMD Modeling Results for Heinze (1999) Liver Lesion Data
Model
Log-probit (slope >1)
Gamma (power >1)
Multistage (degree=l)*
Quantal Linear
Weibull (power >1)
Log-logistic (slope >1)
Quantal Quadratic
Logistic
Probit
Degrees
of
Freedom
2
2
2
2
2
1
2
1
1
x2
0.01
0.09
0.09
0.09
0.09
0.00
5.77
3.43
3.50
% Goodness
of Fit
p-Va\ue
0.99
0.96
0.96
0.96
0.96
1
0.06
0.06
0.06
AIC
21.97
22.05
22.05
22.05
22.05
23.96
26.17
28.31
28.43
BMD
Og/kg-
day)
11.04
6.31
6.31
6.31
6.31
10.14
24.81
19.43
19.69
BMDL
(Hg/kg-day)
6.38
3.92
3.92
3.92
3.92
1.24
19.01
11.40
12.31
* Degree of polynomial initially set to (n-1) where n= number of dose groups including control;
model selected is lowest degree model providing adequate fit. Betas restricted to >0.
                                         87
DRAFT - DO NOT CITE OR QUOTE

-------
 1
 3
 5
 7
 9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
44
45
46
47
48
49
50
    0.
                            Probit Model with 0.95 Confidence Level
           Probit
            BMD
                   BM
                            4       6       8       10      12      14      16

                                         dose
   15:0607/142005
Figure 5-2.  Probit Model Fit to Liver Lesion Incidence Data from Heinze (1999)
                                                   88
                                                DRAFT - DO NOT CITE OR QUOTE

-------
 1          5.2.2.3. RfD Derivation
 2
 3          The BMDL of 6 |ig/kg-day from the Heinze (1999) data is the lowest BMDL among the
 4   modeled datasets and was used as the point of departure (POD) for the short-term RfD. Dividing
 5   the BMDL of 6 |j,g/kg-day by a composite uncertainty factor (UF) of 1000 results in a short-term
 6   RfD for MCLR of 6xlO'6 mg/kg-day.
 7
 8          Short-term RfD            =      BMDL-UF
 9                                     =6 ng/kg-day -  1000
10                                     =      0.000006 mg/kg-day or 6x106 mg/kg-day
11
12          The composite UF of 1000 includes a factor of 10 for interspecies extrapolation, a factor
13   of 10 to account for interindividual variability in the human population and a factor of 10 for
14   database limitations, as follows.
15
16      •  A default 10-fold UF for intraspecies differences was used to account for potentially
17          susceptible individuals in the human population. There is insufficient information on the
18          toxicity of microcystins in exposed humans. Cases of human poisoning have been
19          attributed to ingestion of water containing microcystin-producing cyanobacteria, but no
20          dose-response information is available.  There is no information on the degree to which
21          humans of varying gender, age, health status or genetic makeup might vary in the
22          disposition of,  or response to, ingested microcystins.  There are some data to suggest that
23          adult or aged rodents may be more susceptible than young rodents to the acute toxicity of
24          MCLR (Ito et al., 1997a). Further, studies with algal extracts suggest the possibility that
25          male mice may be more sensitive than female mice to oral exposure to algal extracts
26          (Falconer et al., 1988); however, the relevance of this finding to human microcystin
27          exposure is unclear.

28      •  An interspecies UF of 10 was used to account for differences in response between
29          laboratory rodents and humans. No information is available on the toxicity of purified
30          microcystins in humans, and data on toxicokinetic differences between animals and
31          humans in the disposition of ingested microcystins are not available. Limited data from
32          in vitro studies suggests that human hepatocytes may be more susceptible to the effects of
33          MCLR than rat hepatocytes (Yea et al., 2001; Batista et al., 2003),  supporting the use of a
34          full 10-fold UF.

35      •  A 10-fold UF is used to account for deficiencies in the database. Database deficiencies
36          include the lack of a detailed developmental toxicity study, a neurotoxicity study, a
37          multi-generation reproductive toxicity study and supporting information on systemic
38          toxicity in a second species.
                                                89        DRAFT - DO NOT CITE OR QUOTE

-------
 1   5.2.3.  Subchronic Oral RfD
 2
 3          5.2.3.1. Choice of Principal Study and Critical Effect
 4
 5          A single subchronic oral toxicity study is available for MCLR. Fawell et al. (1999)
 6   identified both a NOAEL (40 |j,g/kg-day) and a LOAEL (200 |j,g/kg-day for slight liver injury)
 7   after subchronic exposure of male and female mice.  Fawell et al. (1999) administered daily oral
 8   gavage doses of 0, 40, 200 or 1000 |j,g MCLR per kg body weight to groups of 15 male and 15
 9   female mice for 13 weeks. Histopathological changes in the liver and serum enzyme changes
10   were reported in the mid- and high-dose groups. Both the histopathology and the serum
11   enzymes showed dose-dependent changes. The authors considered the liver changes in the 200
12   |j,g/kg-day dose group to represent a minimal effect.
13
14          The NOAEL from this 90-day study (40 |j,g/kg-day) is only slightly lower than the
15   LOAEL identified in the 28-day study above (50 |j,g/kg-day). Further, Heinze (1999) observed
16   more severe effects in rats exposed to 50 |j,g/kg-day for 28 days in drinking water than Fawell et
17   al. (1999) observed in mice exposed via gavage doses of 200 |j,g/kg-day for 90 days. The reason
18   for this difference in response is not clear. Both studies used a commercially-produced test
19   material from the same manufacturer. In the drinking water study, MCLR was dissolved in
20   ethanol and diluted to a stock solution that was subsequently used to prepare drinking water. It
21   is possible that the small intake of ethanol may have potentiated the hepatic effects of MCLR;
22   however, there are no data to determine whether this is likely or not. In the gavage  study, test
23   solutions were prepared with distilled water and the concentration was confirmed by HPLC with
24   UV detection (Fawell et al., 1999). The accuracy of dosing in the gavage study was likely to be
25   greater than in the drinking water study. The authors of the drinking water study indicated that
26   the MCLR solution was prepared daily, and water consumption was measured daily. Between 3
27   and 7% of the water solution administered over the 28 days was not consumed, and the dose
28   estimates were not corrected for this loss (Heinze, 1999). This loss of administered dose would
29   lead to a small overestimate of the LOAEL in the drinking water study, leading to a further
30   discrepancy in the results of the two studies.
31
32          The drinking water study used smaller group sizes (10 males/dose) than the gavage study
33   (15/sex/dose or 30/dose).  However, the incidence of liver lesions (with necrosis and
34   hemorrhage) in the drinking water study increased from 0% to 100% (including slight, moderate,
35   and intensive lesions) between the control and low-dose group, and there was a dose-related
36   change in the severity of the lesions, leaving little question that the effect was treatment-related.
37
38          These studies appear to contradict evidence from acute parenteral studies indicating that
39   mice are more sensitive to the acute effects of MCLR. Typically, mice die within a few hours of
40   a lethal injected dose, while rats may survive 24-48 hours.  Species-specific differences in oral
41   absorption of MCLR do not appear to account for the discrepancy between these studies; in an
42   acute study of orally-administered MCLR using both mice  and rats, mortality occurred at a lower
43   dose in mice (1580 ng/kg) than in rats (5000 ng/kg; Fawell et al., 1999). It is possible that the
44   more mild effects in the mice in the subchronic study resulted from an adaptive response to
45   MCLR exposure.  The longer exposure duration may have allowed for liver regeneration and
                                                90       DRAFT - DO NOT CITE OR QUOTE

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
repair that was not possible in the shorter-term study; however, there is no information to support
this hypothesis. In fact, Ito et al. (1997) reported that light injuries to hepatocytes were still
evident in five of seven mice 2 months after treatment with MCLR had ceased.

       5.2.3.2 Methods of Analysis

       The data from both Heinze (1999) and Fawell et al. (1999) were considered for
identifying the point of departure for the subchronic RfD derivation. Results of the modeling for
Heinze (1999) are reported in Section  5.2.2.2. Among the liver lesions reported by Fawell  et al.
(1999), only chronic inflammation showed evidence of a dose-response relationship.
Consequently, all quantal models in U.S. EPA's Benchmark Dose Software (BMDS) were  fit to
the incidence data for chronic liver inflammation in male and female mice reported by Fawell et
al. (1999). The incidence data are reported in Table 4-6. In accordance with the U.S. EPA
methodology, the default BMR of 10% increase in extra risk was used.  Models were run using
the default restrictions on parameters built into the BMDS.  Adequate fits were achieved with all
models. For both male and female mice, the probit model provided the best fit, as assessed by
AIC.  Table 5-4 gives the results for the best fit models; Appendix A contains the full outputs for
all models.
Table 5-4. BMD Modeling Results for Fawell et al. (1999) Chronic Liver Inflammation Data

Male:
Probit Model
Female:
Probit Model
Degrees
of
Freedom
2
2
X2
0.21
0.94
Goodne
ss of Fit
p-Va\ue
0.90
0.63
AIC
40.75
72.84
BMD
Og/kg-
day)
107.59
86.34
BMDL
Og/kg-
day)
66.45
56.92
20
21
22
23
24
25
26
27
28
29
30
       Fawell et al. (1999) also reported significant increases in ALP, AST, and ALT in high-
dose animals. These data are shown in Table 4-5 (Section 4.2.1.3.1).  Of these, only the increase
in ALT in male mice showed a dose-response trend amenable to modeling. All continuous
models but the Hill model5 were fit to the ALT data for male mice reported in Fawell et al.
(1999). In accordance with the U.S. EPA methodology, the default BMR of one standard
deviation change from the control mean was used, and the polynomial coefficients were
restricted to be positive.  Only the linear model with a nonhomogenous variance provided an
adequate fit to the data. Table 5-5 gives the results from the linear model (see Appendix A for
model  output). The BMD and BMDL estimated by the linear model (nonhomogenous variance)
for the ALT increases in male mice are 82 and 58 |j,g/kg-day, respectively.
     5 There were too few dose groups to apply the Hill model.
                                                91
                                                    DRAFT - DO NOT CITE OR QUOTE

-------
Table 5-5. BMD Modeling Results for Fawell et al. (1999) ALT Data in Male Mice

Linear
model, non-
homogenous
variance
Degrees
of
Freedom
2
Goodness
of Fit
/>-Value
0.10
AIC
447.68
BMD
Og/kg-
day)
81.84
BMDL
(Hg/kg-day)
58.37
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
       The BMDL from the 28-day drinking water study (6 |j,g/kg-day) is approximately an
order of magnitude lower than any of the BMDL values from the 90-day gavage study (57-66
Hg/kg-day).  Details of the BMD modeling and analysis of the data from Heinze (1999) are
provided above in Section 5.2.2.2, Method of Analysis, under Short-Term Oral RfD.

       5.2.3.3. RfD Derivation

       The BMDL of 6 ng/kg-day from the Heinze (1999) data is used as the POD for the
subchronic RfD.  A composite UF of 1000 is used to derive the subchronic RfD, including a
factor of 10 for interspecies extrapolation, a factor of 10 for interindividual variability, and a
factor of 10 for database limitations (see Section 5.2.2.3 above for details). Although the BMDL
comes from a 28-day study, a UF for exposure duration is not proposed, based on the lower
toxicity observed in the 90-day gavage study conducted by Fawell et al. (1999). The subchronic
RfD is, therefore, set equal to the short-term RfD  of 0.006 |j,g/kg-day or 6xlO"6 mg/kg-day.
       Subchronic RfD
5.2.4.  Chronic Oral RfD
BMDL -H UF
6 ng/kg-day - 1000
0.006 ng/kg-day or 6x10 ° mg/kg-day
                                                               -6
       5.2.4.1.  Choice of Principal Study and Critical Effect

       Two chronic studies of the oral toxicity of MCLR were identified.  Ito et al. (1997b)
conducted a chronic gavage study in mice with 80 to 200 doses (80 |j,g/kg-day) given over 28
weeks; however, the control group was very small (3 animals) and the postmortem examination
was apparently limited to the liver.  This study was not given further consideration for RfD
development given these study quality concerns.

       Ueno et  al. (1999) evaluated the toxicity of MCLR in female mice chronically exposed
via drinking water 7 days/week.  The authors conducted a comprehensive postmortem
examination.  No treatment-related effects were identified, and the authors observed no
                                               92
                                                    DRAFT - DO NOT CITE OR QUOTE

-------
 1   difference in the incidence of liver histopathology between treated and control mice. It is
 2   important to note that immunohistochemistry of the liver revealed no accumulation of MCLR.
 3   This study identified a free-standing NOAEL of approximately 3 |j,g/kg-day in female mice.
 4
 5          Although Ueno et al. (1999) used only a single dose level and identified a freestanding
 6   NOAEL, it was chosen for RfD derivation because it was a well-conducted chronic  study using a
 7   relevant exposure route (drinking water). The BMDL of 6 |j,g/kg-day from modeling (see
 8   Section 5.2.2.2) of Heinze (1999) compares favorably with the free-standing NOAEL of 3
 9   |j,g/kg-day reported by Ueno et al. (1996), providing support for the use of the NOAEL from
10   Ueno etal. (1999).
11
12          5.2.4.2.  RfD Derivation
13
14          As noted above, the NOAEL of 3 |j,g/kg-day from the study by Ueno et al. (1999) is used
15   as the POD for the chronic RfD. Because this study used a single dose, it was not possible to use
16   BMD modeling to identify the POD. A composite UF of 1000 is used to derive the chronic RfD,
17   including a factor of 10 for interspecies extrapolation, a factor of 10  for interindividual
18   variability and a factor of 10 for database limitations (see Section 5.2.2.3 above for details).
19   Dividing the NOAEL of 3 |j,g/kg-day by a composite UF of 1000 results in a chronic RfD for
20   MCLR of 3xlO"6 mg/kg-day.
21
22          Chronic RfD               =      NOAEL-UF
23                                     =3 ng/kg-day - 1000
24                                     =      0.003 ng/kg-day or 3x106 mg/kg-day
25
26          In 1999, the WHO published a provisional Tolerable Daily Intake (TDI) for MCLR based
27   on the subchronic gavage study later published by Fawell et  al. (1999).  The WHO used the
28   NOAEL  of 40 ng/kg-day with a composite UF of 1000 to derive a TDI  of 0.04 ng/kg-day or
29   4xlO"5 mg/kg-day. The composite uncertainty factor included UFs of 10-fold each for
30   interindividual variability, interspecies extrapolation, and database deficiencies (WHO
31   specifically cited the lack of chronic toxicity and  carcinogenicity studies). WHO (1999) did  not
32   evaluate either Ueno et al. (1999) or Heinze (1999), which may not have been published at the
33   time.
34
35          Table 5-6 provides a summary of the RfD values derived for MCLR in this report.
36
37   5.3.    INHALATION REFERENCE CONCENTRATION
38
39          The available data do not provide adequate information for the derivation of inhalation
40   RfCs for MCLR.  Two acute  inhalation studies were identified in the literature. In a poorly
41   described study, Fitzgeorge et al.  (1994) conducted a single experiment with mice (number
42   unspecified) inhaling a fine aerosol (particle size  3-5 |jm) with 50  |jg/L MCLR for an unspecified
43   duration of time. There were apparently no deaths,  clinical signs of toxicity or histopathological
44   changes;  however, the authors gave few details of study design and findings. A brief abstract
45   describes a study of acute microcystin exposure via inhalation (Creasia, 1990).  The LCso for
                                                93       DRAFT - DO NOT CITE OR QUOTE

-------
Table 5-6. Summary of Reference Dose Values

Acute
Short-term
Sub chronic
Chronic
RfD (mg/kg-day)
NA
6xlO'6
6xlO'6
3 x ID'6
Critical Effect

Hepatotoxicity
Hepatotoxicity
No effects observed
Principal Study

Heinze, 1999
Heinze, 1999
Uenoetal., 1999
94
DRAFT - DO NOT CITE OR QUOTE

-------
 1   mice exposed to a MCLR aerosol (nose only) for 10 minutes was reported to be 18 (ig/L (mg/m3)
 2   air with a 95% confidence interval of 15.0-22.0 |j,g/L (mg/m3). The authors reported that
 3   histological lesions in mice killed by aerosol exposure were similar to those in mice dosed
 4   intravenously with MCLR. Neither of these studies provides adequate basis for an acute RfC.
 5
 6          Only  one well-conducted study of inhalation exposure to MCLR was identified. Benson
 7   et al. (2005) exposed groups of six male BALB/c mice to monodisperse submicron aerosols of
 8   MCLR via nose-only inhalation for 30, 60 or 120 minutes each day for 7 consecutive days. The
 9   concentration of MCLR was 260-265 |J,g/m3. Histopathological examination revealed treatment-
10   related lesions in the nasal cavity only.  The incidence and severity of nasal lesions increased
11   with daily exposure duration. This study used only one  exposure concentration, and as such,  the
12   data are of limited utility for RfC derivation. Further, extrapolation of the effects from this study
13   for the purpose of deriving a short-term RfC would be associated with substantial uncertainty
14   given the brief exposure time (30-120 minutes/day) and duration (7 days). There are no
15   subchronic or chronic animal studies evaluating the inhalation route of exposure.
16
17          Route-to-route extrapolation is not considered appropriate for microcystins based on
18   current data.  Limited available information indicates that inhalation exposure to microcystins
19   may cause point-of-entry effects (Benson et al., 2005), while oral exposure leads to
20   hepatotoxicity.  Data from intratracheal and intranasal instillation studies  show hepatic effects
21   after exposure via these routes; however, the relevance of this information to inhalation
22   exposures is uncertain.
23
24   5.4.    CANCER ASSESSMENT
25
26          No dose-response or other information is available regarding the carcinogenicity of pure
27   microcystins.
                                                 95       DRAFT - DO NOT CITE OR QUOTE

-------
 1         6. MAJOR CONCLUSIONS IN THE CHARACTERIZATIONS OF HAZARD
 2                                   AND DOSE-RESPONSE
 3
 4
 5   6.1.   HUMAN HAZARD POTENTIAL
 6
 7          Microcystins are a group of naturally occurring hepatotoxins produced by freshwater
 8   cyanobacteria.  No studies of the absorption, distribution, metabolism or elimination of
 9   microcystins LR, YR, RR or LA in vivo have been conducted. Acute lethality data suggest a
10   significant difference in the lethal dose after oral exposure to MCLR when compared with
11   injection routes of exposure, suggesting low absorption of orally-administered MCLR. The
12   cellular uptake and distribution of MCLR has been extensively studied, showing preferential
13   uptake of MCLR by hepatocytes due to the presence of a bile acid transporter.
14
15          Data on human exposures to microcystin-producing cyanobacteria have shown
16   gastrointestinal and dermal effects; however, it is not clear whether these are effects  of
17   microcystins, other endotoxins, or the microorganisms themselves. The preponderance of the
18   available toxicological studies in animals employed the MCLR congener. Both oral  and
19   parenteral exposure studies in laboratory animals point to the liver as the primary target organ of
20   MCLR, and parenteral exposure studies suggest a steep dose-response curve for the hepatotoxic
21   effects. The toxicological database is limited to only a few studies using oral exposure to
22   purified microcystins; all of these used the MCLR congener, and most were conducted in mice.
23   The database includes four studies of acute exposure, one 28-day and one 90-day study, and two
24   chronic studies.  In all of the studies where a toxicological effect was observed, the primary
25   target organ was the liver (one study also identified gastrointestinal lesions). The liver effects
26   included hepatocyte degeneration and necrosis, inflammation, fibrosis, hypertrophy and
27   hemorrhage.  Humans exposed to microcystins via dialysate suffered acute liver failure and, in
28   many cases, death. Liver biopsies conducted on the decedents showed hepatocyte necrosis and
29   apoptosis, but no intrahepatic hemorrhage.
30
31          A single well-described study of short-term (7 days) inhalation exposure identified the
32   upper respiratory tract as a target organ.  Damage to the respiratory and olfactory epithelial cells
33   of the nasal cavity was observed in mice in this study.  Studies using intratracheal  or intranasal
34   instillation have shown systemic effects, including liver toxicity, after these exposures.
35
36          A single oral  developmental toxicity study with inadequate data reporting indicated that
37   developmental effects in the  absence of maternal toxicity are not likely, and that developmental
38   effects, if any, would occur at much higher doses than liver effects.
39
40          Mechanistic studies indicate that the inhibition of protein phosphatases 1 and 2A  (a well-
41   established molecular effect of microcystin exposure) and/or oxidative stress play  a role in the
42   hepatotoxicity of MCLR. Cellular effects in hepatocytes include membrane blebbing, cell
43   rounding and dissociation, and apoptosis. These cellular effects lead to alterations in the liver
44   structure and function as well as intrahepatic hemorrhage.
45
                                                 96       DRAFT - DO NOT CITE OR QUOTE

-------
 1    6.2.     DOSE RESPONSE
 2
 3           The available oral data were sufficient for derivation of short-term, subchronic and
 4    chronic oral RfDs for MCLR.  Based on a BMDL of 6 |j,g/kg-day for hepatotoxicity in rats
 5    exposed to MCLR in drinking water for 28 days (Heinze, 1999), an RfD of 0.006 |j,g/kg-day
 6    (6xlO"6 mg/kg-day) was derived for short-term and subchronic exposure durations. A UF of
 7    1000 was used to derive the RfD. The UF comprises component factors of 10 for interspecies
 8    extrapolation, 10 for interindividual variability and 10  for database deficiencies.  The subchronic
 9    RfD did not include an additional UF for extrapolating from a 28-day study because a subchronic
10    (90-day) gavage study identified a higher NOAEL. A  chronic RfD of 0.003 |ig/kg-day (3xlO"6
11    mg/kg-day) was derived from a free-standing NOAEL of 3 |j,g/kg-day in female mice chronically
12    exposed via drinking water. A composite UF of 1000 was used, with factors of 10 each for
13    interindividual variability, interspecies extrapolation and database deficiencies. Inhalation RfCs
14    were not derived since there were no studies of adequate quality for this purpose. There is
15    inadequate evidence to evaluate the carcinogenicity of microcystins LR, RR, YR and LA.
                                                97       DRAFT - DO NOT CITE OR QUOTE

-------
 1                                      7.  REFERENCES
 2
 3
 4   Adams, W.H., R.D. Stoner, D.G. Adams et al.  1985.  Pathophysiologic effects of a toxic peptide
 5   fromMicrocystis aeruginosa. Toxicon. 23(3):441-447.

 6   Adams, W.H., J.P. Stone, B.  Sylvester et al.  1988.  Pathophysiology of cyanoginosin-LR, in
 7   vivo and in vitro studies.  Toxicol. Appl. Pharmacol. 96(2):248-257.

 8   Ashworth, C.T. and M.F. Mason. 1945. Observations on the pathological changes produced by
 9   a toxic substance present in blue-green algae (Microcystis aeruginosa). Am. J. Pathol.
10   22:369-383.

11   Aune, T. and K. Berg. 1986. Use of freshly prepared rat hepatocytes to study toxicity of blooms
12   of the blue-green algae Microcystis aeruginosa and cyanotoxin Oscillatoria agardhii. J.
13   Toxicol. Environ. Health.  19(3):325-336.

14   Azevedo, S.M.F.O., W.W. Carmichael, E.M. Jochimsen et al.  2002. Human intoxication by
15   microcystins during renal dialysis treatment in Caruaru, Brazil. Toxicology. 181-182:441-446.

16   Aziz, K.M.S.  1974. Diarrhea toxin obtained from a waterbloom-producing species, Microcystis
17   aeruginosaKutzing. Science.  183:1206-1207.

18   Bagu, J.R., B.D. Sykes, M.M. Craig and C.F.B. Holmes.  1997. A molecular basis for different
19   interactions of marine toxins with protein phosphatase-1.  Molecular models for bound
20   motuporin, microcystins, okadaic acid, and calyculin A. J. Biol. Chem. 272(8):5087-5097.

21   Batista, T., G. de Sousa, J.S.  Suput et al. 2003. Microcystin-LR causes the collapse of actin
22   filaments in primary human hepatocytes.  Aquat. Toxicol. 65(1):85-91.

23   Battle, T., C. Touchard, HJ. Moulsdale et al.  1997. New cell substrates for in vitro evaluation
24   of microcystin hepatocytoxicity. Toxicol. In Vitro. ll(5):557-567.

25   Beasley, V.R., R.A. Lovell, K.R. Holmes et al. 2000.  Microcystin-LR decreases hepatic and
26   renal perfusion and causes circulatory shock, severe hyperglycemia, and terminal hyperkalemia
27   in intravascularly dosed  swine. J. Toxicol. Environ. Health-Part A.  61(4):281-303.

28   Becchetti, A., B. Malik,  G. Yue et al.  2002.  Phosphatase inhibitors increase the open
29   probability of ENaC in A6 cells. Am. J. Physiol. Renal Physiol.  283(5):F1030-F1045.

30   Benson, J.M., J. A. Hutt, K. Rein et al.  2005.  The toxicity of microcystin LR in mice following
31   7 days of inhalation exposure.  Toxicon. 45(6):691-698.

32   Berg, K. and T. Aune. 1987. Freshly prepared rat hepatocytes used in screening the toxicity of
33   blue-green algal blooms. J. Toxicol. Environ. Health.  20(1/2): 187-197.
                                                98       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Berg, K., J. Wyman, W.W. Carmichael and A.S. Dabholkar.  1988. Isolated rat liver perfusion
 2   studies with cyclic heptapeptide toxins of Microcystis and Oscillatoria (freshwater
 3   cyanobacteria). Toxicon. 26(9):827-837.

 4   Bhattacharya, R., P.V.L. Rao, A.S.B. Bhaskar et al.  1996. Liver slice culture for assessing
 5   hepatoxicity of freshwater cyanobacteria.  Human Exp. Toxicol.  15(2): 105-110.

 6   Bhattacharya, R., K. Sugendran, R.S. Dangi and P.V.L. Rao. 1997. Toxicity evaluation of
 7   freshwater cyanobactenum Microcystis aeruginosaPCC-7806. 2. Nephrotoxicity in rats.
 8   Biomed. Environ. Sci.  10(1):93-101.

 9   Billings, W.H.  1981.  Water-associated human illness in northeast Pennsylvania and its
10   suspected association with blue-green algae blooms.  In: The Water Environment: Algal  Toxins
11   and Health, W.W. Carmichael, Ed.  Plenum Press, New York, NY. p. 243-255.

12   Bishop, C.T., E.F.L.J. Anet and P.R. Gorham. 1959. Isolation and identification of the fast-
13   death factor in Microcystis aeruginosa NRC-l.  Can. J. Biochem. Physiol. 37(3):453-471.

14   Blankson, H., E.M. Grotterod and P.O. Seglen.  2000. Prevention of toxin-induced cytoskeletal
15   disruption and apoptotic liver cell death by the grapefruit flavonoid, narigin. Cell Death Diff.
16   7(8):739-746.

17   Boe, R., B.T. Gjersten, O.K.  Vintermyr et al. 1991. The protein phosphatase inhibitor okadaic
18   acid induces morphological changes typical of apoptosis in mammalian cells. Exp. Cell Res.
19   195(l):237-246.

20   Botha, N, M. van de Venter,  T.G. Downing et al. 2004.  The effect of intraperitoneally
21   administered microcystin-LR on the gastrointestinal tract of Balb/c mice. Toxicon.
22   43(3):251-254.

23   Boua'icha, N. and I. Maatouk. 2004. Microcystin-LR and nodularin induce intracellular
24   glutathione alteration, reactive oxygen species production and lipid peroxidation in primary
25   cultured rat hepatocytes.  Toxicol. Lett.  148(l-2):53-63.

26   Boua'icha, N., I. Maatouk, MJ. Plessis and F. Perin. 2005. Genotoxic potential of microcystin-
27   LR and nodularin in vitro in primary cultured rat hepatocytes and in vivo in rat liver. Environ.
28   Toxicol. 20(3):341-347.

29   Brooks, W.P. and G.A. Codd.  1987. Immunological and toxicological studies on Microcystis
30   aeruginosa peptide toxin. Br. Phycol. J.  22(3):301.

31   Carmichael, W.W., S.M.F.O. Azevedo, IS. An et al. 2001. Human fatalities from
32   cyanobacteria:  Chemical and biological evidence for cyanotoxins.  Environ. Health Perspect.
33   109(7):663-668.

34   Chen, J., T.C. Campbell, J.Li and R.Peto. 1991.  Diet, Life-style and Mortality in China.  A
35   Study of the Characteristics of 65 Chinese Counties, Cornell University Press, Ithaca, NY. (As
36   cited in Health Canada, 2002).


                                                99        DRAFT - DO NOT CITE OR QUOTE

-------
 1   Chen, K., Y.Z. Shen and G.F. Shen.  1994.  Study on incidence rate of some cancer in areas with
 2   difference in drinking water sources. Chin. J. Public Health.  12(3): 146-148 (As cited in Zhou et
 3   al., 2002). (Chinese)

 4   Chen, T., X. Zhao, Y Liu et al.  2004. Analysis of immunomodulating nitric oxide, iNOS and
 5   cytokines MRNA in mouse macrophages induced by microcystin-LR.  Toxicology.
 6   197(l):67-77.

 7   Chen, T., P. Shen, J. Zhang et al.  2005.  Effects of microcystin-LR on patterns of iNOS and
 8   cytokine mRNA expression in macrophages in vitro. Environ. Toxicol. 20(1):85-91.

 9   Chen, T., J. Cui, Y. Liang et al. 2006. Identification of human liver mitochondrial aldehyde
10   dehydrogenase as a potential target for microcystin-LR. Toxicology. 220(1):71-80.

11   Chernoff, N, E.S. Hunter III, L.L. Hall et al. 2002.  Lack of teratogenicity of microcystin-LR in
12   the mouse and toad. J. Appl. Toxicol. 22(1):13-17.

13   Chong, M.W.K., K.D. Gu, P.K.S. Lam et al. 2000. Study of the cytotoxicity of microcystin-LR
14   on cultured cells.  Chemosphere.  41(1-2):143-147.

15   Cote, L-M., R. A. Lovell, E.H. Jeffrey et al. 1986. Failure of blue-green algae (Microcystis
16   aeruginosd) hepatotoxin to alter in vitro mouse liver enzymatic activity. J. Toxicol.- Toxin Rev.
17   52(2):256.

18   Craig, M., H.A. Luu, T.L. McCready et al.  1996. Molecular mechanisms underlying the
19   interaction of motuporin and microcystins with type-1 and type-2A protein phosphatases.
20   Biochem. Cell Biol. 74(4):569-578.

21   Creasia, D.A. 1990.  Acute inhalation toxicity of microcystin-LR with mice.  Toxicon.
22   28(6):605.

23   Dabholkar, A.S. and W.W. Carmichael.  1987. Ultrastructural changes in the mouse liver
24   induced by hepatoxin from the freshwater cyanobacterium Microcystis aeruginosa strain 7820.
25   Toxicon. 25(3):285-292.

26   Dahlem, A.M., A.S. Hassan, S.P.  Swanson et al.  1989. A model system for studying the
27   bioavailability of intestinally administered microcystin-LR, a hepatotoxic peptide from the
28   cyanobacterium Microcystis aeruginosa.  Pharmacol. Toxicol. 64(2): 177-181.

29   Davidson, F.F.  1959.  Poisoning of wild and domestic animals by a toxic waterbloom of Nostoc
30   rivulare Kuetz.  J. Am. Water Works Assoc. 51:1277-1287.

31   Dillenberg, H.O. and M.K. Dehnel. 1960. Toxic water bloom in Saskatchewan, 1959. Can.
32   Med. Assoc. J.  83:1151-1154.

33   Ding, W.X. and C.N. Ong. 2003. Role of oxidative stress and mitochondrial changes in
34   cyanobacteria-induced apoptosis and hepatotoxicity. FEMS Microbiol. Lett.  220(1): 1-7.
                                               100       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Ding, W.X., H.M. Shen, Y. Shen et al. 1998a. Microcystic cyanobacteria causes mitochondrial-
 2   membrane potential alteration and reactive oxygen species formation in primary cultured rat
 3   hepatocytes. Environ. Health Perspect.  106(7):409-413.

 4   Ding, W.X., H.M. Shen, H.G. Zhu and C.N. Ong.  1998b. Studies on oxidative damage induced
 5   by cyanobacteria extract in primary cultured rat hepatocytes.  Environ. Res. 78(1): 12-18.

 6   Ding, W.X., H.M. Shen, H.G. Zhu, B.L. Lee and C.N. Ong.  1999. Genotoxicity of microcystic
 7   cyanobacterial extract of a water source in China.  Mutat Res.  442(2):69-77.

 8   Ding, W.X., H.M. Shen and C.N. Ong. 2000a. Microcystic cyanobacteria extract induces
 9   cytoskeletal disruption and intracellular glutathione alteration in hepatocytes.  Environ. Health
10   Perspect. 108(7):605-609.

11   Ding, W.X., H.M. Shen and C.N. Ong. 2000b. Critical role of reactive oxygen species and
12   mitochondrial permeability transition in microcystin-induced rapid apoptosis in rat hepatocytes.
13   Hepatology. 32(3):547-555.

14   Ding, W.X., H.M. Shen and C.N. Ong. 2001. Critical role of reactive oxygen species formation
15   in microcystin-induced cytoskeleton disruption in  primary cultured hepatocytes. J. Toxicol.
16   Environ. Health Part A.  64(6):507-519.

17   Ding, W.X., H.M. Shen and C.N. Ong. 2002. Calpain activation  after mitochondrial
18   permeability transition in microcystin-induced cell death in rat hepatocytes. Biochem. Biophys.
19   Res. Commun. 291(2):321-331.

20   Duy, T.N., P.K.S. Lam, G.R. Shaw et al.  2000.  Toxicology and risk assessment of freshwater
21   cyanobacterial (blue-green algal) toxins in water.  Rev. Environ. Contam. Toxicol. 163:113-185.

22   Dvorakova, D., K. Dvorakova, L. Blaha et al.  2002. Effects of cyanobacterial biomass and
23   purified microcystins on malformations in Xenopus laevis:  Teratogenesis assay  (FETAX).
24   Environ. Toxicol. 17(6):547-555.

25   Elleman, T.C., I.R. Falconer, J.R. Jackson andM.T.C. Runnegar.  1978.  Isolation,
26   characterization and pathology of the toxin from aMicrocystis aeruginosa (=Anacystis cyanea)
27   bloom.  Aust. J. Biol. Sci.  31(3):209-218.

28   el Saadi, O. and A.S. Cameron.  1993. Illness associated with blue-green algae.  Med. J. Aust.
29   158(11):792-793.

30   el Saadi, O., AJ. Esterman, S. Cameron and D.M. Roder. 1995. Murray River water raised
31   cyanobacterial cell counts, and gastrointestinal and dermatological symptoms. Med. J. Aust.
32   162(3): 122-125.

33   Eriksson, I.E. and R.D. Golman.  1993.  Protein phosphatase inhibitors alter cytoskeletal
34   structure and cellular morphology. Adv. Prot. Phosphatases.  7:335-357.
                                               101       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Eriksson, I.E., J.A.O. Meriluoto and T. Lindholm. 1987. Cyanobacterial toxins - physiological
 2   and ecological effects. In: Rapport till Finlans Akademi. p. 1-12.

 3   Eriksson, I.E., G.I.L. Paatero, J.A.O. Meriluoto et al.  1989. Rapid microfilament reorganization
 4   induced in isolated rat hepatocytes by microcystin-LR, a cyclic peptide toxin.  Exp. Cell Res.
 5   185(1):86-100.

 6   Eriksson, I.E., L. Gronberg, S. Nygard et al. 1990a.  Hepatocellular uptake of
 7   3H-dihydromicrocystin-LR, a cyclic peptide toxin. Biochim. Biophys. Acta. 1025(l):60-66.

 8   Eriksson, I.E., D. Toivola, J.A.O. Meriluoto et al.  1990b.  Hepatocyte deformation induced by
 9   cyanobacterial toxins reflects inhibition of protein phosphatases. Biochem. Biophys. Res.
10   Commun. 173(3): 1347-1353.

11   Eriksson, I.E., D.M. Toivola, M. Reinikainen et al.  1994.  Testing of toxicity in cyanobacteria
12   by cellular assays.  In: Detection Methods for Cyanobacterial Toxins, G.A. Codd, T.M. Jefferies,
13   C.W. Keevil and E. Potter, Eds.  Royal Society of Chemistry, Cambridge,  p. 75-84.

14   Falconer, I.R.  1991.  Tumor promotion and liver injury caused by oral consumption of
15   cyanobacteria. Environ. Toxicol. Water Qual. 6(2):177-184.

16   Falconer, I.R. and T.H. Buckley. 1989. Tumour promotion by Microcystis sp., a blue-green alga
17   occurring in water  supplies.   Med. J. Aust.  150(6):351.

18   Falconer, I.R. and A.R. Humpage. 1996. Tumour promotion by cyanobacterial toxins.
19   Phycologia.  35:74-79.

20   Falconer, L.R. and M.T.C. Runnegar.  1987a. Effects of the peptide toxin from Microcystis
21   aeruginosa on intracellular calcium, pH and membrane integrity in mammalian cells.  Chem-
22   Biol. Interact.  63(3):215-225.

23   Falconer, L.R. and M.T.C. Runnegar.  1987b. Toxic peptide from the blue-green alga
24   Microcystis aeruginosa- effects  on hepatocytes and thymocytes.  Biochem. Soc.  Trans.
25   15(3):468-469.

26   Falconer, I.R. and S.K. Yeung.   1992. Cytoskeletal changes in hepatocytes induced by
27   Microcystis toxins  and their relation to hyperphosphorylation of cell proteins.  Chem-Biol.
28   Interact.  81(1-2): 181-196.

29   Falconer, I.R., A.R.B. Jackson, J. Langley and M.T.C.  Runnegar. 1981.  Liver pathology in
30   mice in poisoning by the blue-green alga Microcystis aeruginosa. Aust.  J.  Biol.  Sci. 34(2): 174-
31   187.

32   Falconer, I.R., A.M. Beresford and M.T.C. Runnegar.  1983.  Evidence of liver damage by toxin
33   from a bloom of the blue-green alga, Microcystis aeruginosa.  Med. J. Aust. 1(11):511-514.
                                               102       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Falconer, I.R., T. Buckley and M.T.C. Runnegar. 1986. Biological half-life organ distribution
                    • 125
 2   and excretion of  I-labeled toxic peptide from the blue-green alga Microcystis aeruginosa.
 3   Aust.J. Biol. Sci. 39(1): 17-21.

 4   Falconer, L.R., J.V. Smith, A.R.B. Jackson et al.  1988.  Oral toxicity of a bloom of the
 5   cyanobacterium Microcystis aeruginosa administered to mice over periods of up to one year. J.
 6   Toxicol. Environ. Health. 24(3):291-305.

 7   Falconer, I.R., M.D. Burch, D.A. Steffensen, M. Choice and O.R. Coverdale. 1994. Toxicity of
 8   the blue-green alga (Cyanobacterium) Microcystis aeruginosa in drinking water to growing pigs,
 9   as an animal model for human injury and risk assessment.  Environ. Toxicol. Water Qual.
10   9:131-139.

11   Fawell, J.K., R.E. Mitchell, D.J. Everett and R.E. Hill.  1999.  The toxicity of cyanobacterial
12   toxins in the mouse. 1.  Microcystin-LR.  Human Exp. Toxicol. 18(3):162-167.

13   Feitz, A.J., T. Lukondeh, M.C. Moffitt et al.  2002.  Absence of detectable levels of the
14   cyanobacterial toxin (microcystin-LR) carry-over into milk. Toxicon.  40(8): 1173-1180.

15   Fischer, WJ. and D.R. Dietrich. 2000.  Toxicity  of the cyanobacterial cyclic heptapeptide toxins
16   microcystin-LR and -RR in early lifestages of the African clawed frog. Aquat. Toxicol.
17   49(3):189-198.

18   Fitzgeorge, N.L.M., S. A. Clark and C.W. Kelvin. 1994. Routes of intoxication.  In: Detection
19   Methods for Cyanobacterial (Blue-Green Algal) Toxins. G.A. Codd, T.M. Jeffreies, C.W.
20   Kelvin and E. Potter, Eds.  Royal Society of Chemistry, Cambridge, U.K. p. 69-74.

21   Fladmark, K.E., M.H. Serres, N.L. Larsen et  al.  1998.  Sensitive detection of apoptogenic toxins
22   in suspension cultures of rat and salmon hepatocytes.  Toxicon.  36(8):1101-1114.

23   Fleming, L.E., C. Rivero, J. Burns et al. 2002.  Blue green algal (cyanobacterial) toxins, surface
24   drinking water and liver cancer in Florida.  Harmful Algae. 1(2): 157-168.

25   Fleming, L.E., C. Rivero, J. Burns, C. Williams, J.A. Bean and W.B. Stephan.  2004.
26   Cyanobacteria exposure, drinking water and colorectal cancer.  In: Harmful Algae 2002.
27   Proceedings of the Xth International Conference on Harmful Algae.  K.A. Steidinger, J.H.
28   Landsberg, C.R.  Tomas and G.A. Vargo, Eds. Florida Fish and  Wildlife Conservation
29   Commission and Intergovernmental Oceanographic Commission of UNESCO, Tallahassee, FL.
30   p. 470-472.

31   Foxall, T.L. and JJ. Sasner, Jr. 1981. Effects of a hepatic toxin from the cyanophyte
32   Microcystis aeruginosa.  In: The Water Environment:  Algal Toxins and Health,  W.W.
33   Carmichael, Ed.  Plenum Press, New York, NY. p. 365-387.

34   Frangez, R.,  M.C. Zuzek, J. Mrkun et al. 2003. Microcystin-LR affects cytoskeleton and
35   morphology  of rabbit primary whole embryo cultured cells in vitro.  Toxicon.  41(8):999-1005.
                                               103       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Fu, W.Y., J.P. Chen, X.M. Wang and L.H. Xu. 2005. Altered expression of p53, Bcl-2 and Bax
 2   induced by microcystin-LR in vivo and in vitro.  Toxicon. 46(2): 171-177.

 3   Gehringer, M.M.  2004.  Microcystin-LR and okadaic acid-induced cellular effects:  A dualistic
 4   response.  FEES Lett.  557(l-3):l-8.

 5   Gehringer, M.M., S. Govender, M. Shaw and T.G. Downing.  2003 a. An investigation of the
 6   role of vitamin E in the protection of mice against microcystin toxicity.  Environ. Toxicol.
 7   18(2): 142-148.

 8   Gehringer, M.M., K.S. Downs, T.G. Downing et al. 2003b. An investigation into the effects of
 9   selenium supplementation on microcystin hepatotoxicity. Toxicon.  41(4):451-458.

10   Gehringer, M.M., E.G. Shephard, T.G. Downing et al.  2004.  An investigation into the
11   detoxification of microcystin-LR by the glutathione pathway in Balb/c mice. Int. J. Biochem.
12   CellBiol. 36(5):931-941.

13   Goldberg, J., H.B. Huang, Y.G. Kwon, et al.  1995. Three-dimensional  structure of the catalytic
14   subunit of protein serine/threonine phosphatase-1. Nature 376(6543)745-753.

15   Grabow, W.O.K., W.C. Du Randt, O.W. Prozesky and W.E. Scott.  1982. Microcystis
16   aeruginosa toxin: Cell culture toxicity, hemolysis, and mutagenicity assays. Appl. Environ.
17   Microbiol. 43(6): 1425-1433.

18   Gulledge, B.M., J.B. Aggen, H.B. Huang et al. 2002. The microcystins and nodularins: cyclic
19   polypeptide inhibitors of PP1 and PP2A.  Curr. Med. Chem. 9(22): 1991-2003.

20   Gulledge, B.M., J.B. Aggen and A.R. Chamberlin. 2003a.  Linearized and truncated microcystin
21   analogues as inhibitors of protein phosphatases 1 and 2A. Bioorg. Medicinal Chem.. Lett.
22   13(17):2903-2906.

23   Gulledge, B.M., J.B. Aggen, H. Engetal.  2003b. Microcystin analogues comprised only of
24   Adda and a single additional amino acid retain moderate activity as PP1/PP2A inhibitors.
25   Bioorg. Medicinal Chem. Lett. 13(17):2907-2911.

26   Gupta, N., S.C. Pant, R. Vijayaraghavan and P.V. Rao. 2003. Comparative toxicity evaluation
27   of cyanobacterial cyclic peptide toxin microcystin variants (LR, RR, YR) in mice. Toxicology.
28   188(2-3):285-296.

29   Guzman, R.E. and P.F. Solter.  1999. Hepatic oxidative stress following prolonged sublethal
30   microcystin-LR exposure. Toxicologic Pathol. 27(5):582-588.

31   Guzman, R.E. and P.F. Solter. 2002. Characterization of sublethal microcystin-LR exposure in
32   mice. Vet. Pathol. 39(l):17-26.

33   Guzman, R.E., P.F. Solter and M.T. Runnegar. 2003. Inhibition of nuclear protein phosphatase
34   activity in mouse hepatocytes by the cyanobacterial toxin microcystin-LR.  Toxicon.
35   41(7):773-781.


                                               104       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Harada, K., K. Ogawa, K. Matsuura et al. 1990. Structural determination of geometrical
 2   isomers of microcystins LR and RR from cyanobacteria by two-dimensional NMR spectroscopic
 3   techniques. Chem. Res. Toxicol.  3(5):473-481.

 4   Hastie, C.J., E.B. Borthwick, L.F. Morrison et al. 2005. Inhibition of several protein
 5   phosphatases by a non-covalently interacting microcystin and a novel cyanobacterial peptide,
 6   nostocyclin. Biochim. Biophys. Acta. 1726:187-193.

 7   Hayakawa, K.  and K. Kohama. 1995. Reversible effects of okadaic acid and microcystin-LR on
 8   the ATP-dependent interaction between actin and myosin. J. Biochem.  117(3):509-514.

 9   Health Canada. 2002. Guidelines for Canadian Drinking Water Quality: Supporting
10   Documentation — Cyanobacterial Toxins-Microcystin-LR. Water Quality and Health Bureau,
11   Healthy Environments and Consumer Safety Branch, Health Canada, Ottawa, Ontario.
12   Available at http://www.hc-sc.gc.ca/ewh-semt/pubs/water-eau/doc_sup-appui/index_e.html.

13   Heinze, R. 1999.  Toxicity of the cyanobacterial toxin microcystin-LR to rats after 28 days
14   intake with the drinking water. Environ. Toxicol.  14(1):57-60.

15   Heinze, R., J. Fastner, U. Neumann and I. Chorus.  2001. Testing cyanobacterial toxicity with
16   primary rat hepatocyte and cell-line  assays.  In: Cyanotoxins:  Occurrence, Causes,
17   Consequences, I. Chorus, Ed.  Springer-Verlag, New York, NY.  p. 317-324.

18   Herfindal, L. and F. Selheim.  2006.  Microcystin produces disparate effects on liver cells in a
19   dose dependent manner. Mini. Rev. Med. Chem. 6(3):279-285.

20   Hermansky, S.J., PJ. Casey and SJ.  Stohs.  1990a.  Cyclosporin  A—a chemoprotectant against
21   microcystin-LR toxicity. Toxicol. Lett.  54(2-3):279-285.

22   Hermansky S.J., S.N. Wolff and SJ.  Stohs.  1990b.  Use of rifampin as an effective
23   chemoprotectant and  antidote against microcystin-LR toxicity. Pharmacology. 41(4):231-236.

24   Hermansky, S.J., SJ. Stohs, R.S. Markin and WJ. Murray. 1990c. Hepatic lipid peroxidation,
25   sulfhydryl status, and toxicity  of the blue-green algal toxin microcystin-LR in mice.  J. Toxicol.
26   Environ. Health. 31(1):71-91.

27   Hermansky, S J., SJ. Stohs, Z.M. Eldeen et al.  1991.  Evaluation of potential chemoprotectants
28   against microcystin-LR hepatotoxicity in mice.  J. Appl. Toxicol.  ll(l):65-73.

29   Hernandez, M., M. Macia, C. Padilla and F.F. Del Campo. 2000. Modulation of human
30   polymorphonuclear leukocyte adherence by cyanopeptide toxins. Environ. Res. 84(l):64-68.

31   Hilborn, E.D.,  W.W. Carmichael, M. Yuan et al. 2005. Serologic evaluation of human
32   microcystin exposure. From:  Interagency International Symposium on Cyanobacterial Harmful
33   Algal Blooms, Research Triangle Park, NC.
                                               105       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Honkanen, R.E., J. Zwiller, R.E. Moore et al. 1990. Characterization of microcystin-LR, a
 2   potent inhibitor of type 1 and type 2A protein phosphatases. J. Biol. Chem.
 3   265(32): 19401-19404.

 4   Hooser, S.B.  2000. Fulminant hepatocyte apoptosis in vivo following microcystin-LR
 5   administration to rats. Toxicol. Pathol. 28(5):726-733.

 6   Hooser, S.B., V.R. Beasley, R.A. Lovell et al.  1989a.  Toxicity of microcystin LR, a cyclic
 7   heptapeptide hepatoxin from Microcystis aeruginosa to rats and mice. Vet. Pathol.
 8   26(3):246-252.

 9   Hooser, S.B., L.L. Waite, V.R. Beasley et al.  1989b.  Microcystin-A induces morphologic and
10   cytoskeletal hepatocyte changes in vitro. Toxicon 27(1):50-51.

11   Hooser, S.B., V.R. Beasley, EJ. Basgall, W.W. Carmichael, and W.M. Haschek.  1990.
12   Microcystin-LR-induced ultrastructural changes in rats.  Vet. Pathol. 27(1):9-15.

13   Hooser, S.B., M.S. Kuhlenschmidt, A.M. Dahlem et al. 1991a. Uptake and subcellular
14   localization of tritiated dihydro-microcystin-LR in rat liver.  Toxicon. 29(6):589-601.

15   Hooser, S.B., V.R. Beasley, L.L. Waite et al.  1991b.  Actin filament alterations in rat
16   hepatocytes induced in vivo and in vitro by microcystin-LR, a hepatoxin from the blue-green
17   alga, Microcystis aeruginosa.  Vet. Pathol. 28(4):259-266.

18   Hu, Z., H. Chen, Y. Li et al. 2002.  [The expression of bcl-2 and bax genes during microcystin
19   induced liver tumorigenesis.]  Zhonghua Yu  Fang Yi Xue ZaZhi. 36(4):239-242. (Chinese)

20   Humpage, A.R. and LR. Falconer. 1999. Microcystin-LR and liver tumor promotion: Effects on
21   cytokinesis, ploidy, and apoptosis in cultured hepatocytes.  Environ. Toxicol.  14(l):61-75.

22   Humpage, A.R., SJ. Hardy, EJ. Moore et al. 2000. Microcystins (cyanobacterial toxins) in
23   drinking water enhance the growth of aberrant crypt foci in the mouse colon.  J. Toxicol.
24   Environ. Health Part A. 61(3):155-165.

25   Ito, E., F. Kondo andK.-I. Harada.  1997a. Hepatic necrosis in aged mice by oral administration
26   of microcystin-LR. Toxicon.  35(2):231-239.

27   Ito, E., F. Kondo, K. Terao and K.-I. Harada.  1997b.  Neoplastic nodular formation in mouse
28   liver induced by repeated intraperitoneal injections of microcystin-LR.  Toxicon.
29   35(9):1453-1457.

30   Ito, E., F. Kondo and K. Harada.  2001.  Intratracheal administration of microcystin-LR, and its
31   distribution.  Toxicon. 39(2-3):265-271.

32   Ito, E., A. Takai, F. Kondo et al.  2002. Comparison of protein phosphatase inhibitory activity
33   and apparent toxicity of microcystins and related compounds.  Toxicon.  40(7): 1017-1025.
                                               106       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Jackson, A.R.B., A. Mclnnes, I.R. Falconer and M.T.C. Runnegar.  1984.  Clinical and
 2   pathological changes in sheep experimentally poisoned by the blue-green alga Microcysstis
 3   aeruginosa.  Vet. Pathol. 21(1): 102-113.

 4   Jayaraj, R., T. AnandandP.V. Rao. 2006. Activity and gene expression profile of certain
 5   antioxidant enzymes to microcystin-LR induced oxidative stress in mice. Toxicology.
 6   220(2-3): 136-146.

 7   Jiao, D.A., G.F. Shen, Y.Z. Shen and G.M. Zheng. 1985. The case-control study of colorectal
 8   cancer. Chin. J. Epidemio. 6:285-288 (As cited in Zhou etal., 2002). (Chinese)

 9   Jochimsen, E.M., W.W. Carmichael, J.S. An et al.  1998.  Liver failure and death after exposure
10   to microcystins at a hemodialysis center in Brazil. New Engl. J. Med. 338(13):873-878.

11   Jones, C.L.A.  1984. Biochemical, hematological, and hepatotoxicological studies of
12   hepatotoxins from Microcystis aeruginosa strain 7820, andAnabaenaflos-aquae strain S-23-G.
13   M.S. Thesis, Wright State Univ., Dayton, OH.

14   Jones, C.L.A. and W.W. Carmichael. 1984. Comparison of hepatotoxins from the cyanobacteria
15   Anabaenaflos-aquae and Microcystis aeruginosa.  Fed. Proc. 43(3):A1716.

16   Kalbe, L.  1984.  Animal experiments on the oral toxicity of blue-green algae waterblooms.
17   Limnol. (Berlin) 15(2):559-562.

18   Khan, S.A., S. Ghosh, M.L. Wickstrom et al.  1995. Comparative pathology of microcystin-LR
19   in cultured hepatocytes, fibroblasts and renal epithelial cells. Natural Toxins. 3(3): 119-128.

20   Klassen, C.D., Ed.  2001.  Casarett and Doull's Toxicology: The Basic Science of Poisons, 6th
21   ed. McGraw-Hill, New York, NY.

22   Knapp J.,  S. Aleth,  F. Balzer et al.  2002.  Calcium-independent activation of the contractile
23   apparatus  in smooth muscle of mouse aorta by protein phosphatase inhibition. Naunyn
24   Schmiedebergs Arch. Pharmacol.  366(6):562-569.

25   Kondo, F., Y. Ikai,  H. Oka et al. 1992.  Formation, characterization, and toxicity of the
26   glutathione and cysteine conjugates of toxic heptapeptide microcystins.  Chem. Res. Toxicol.
27   5(5):591-596.

28   Kondo, F., H. Matsumoto, S. Yamada et al. 1996.  Detection and identification of metabolites of
29   microcystins formed in vivo in mouse and rat livers. Chem. Res. Toxicol. 9(8): 1355-1359.

30   Kujbida, P., E. Hatanaka, A. Campa, P. Colepicolo and E. Pinto.  2006.  Effects of microcystins
31   on human polymorphonuclear leukocytes. Biochem. Biophys. Res. Commun.  341(l):273-277.

32   Lankoff, A., A. Banasik, G. Obe et al. 2003.  Effect of microcystin-LR and cyanobacterial
33   extract from Polish reservoir drinking water on cell cycle progression, mitotic spindle, and
34   apoptosis in CHO-K1 cells. Toxicol. Appl. Pharmacol.  189(3):204-213.
                                               107       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Lankoff, A., L. Krzowski, J. Glab et al. 2004. DNA damage and repair in human peripheral
 2   blood lymphocytes following treatment with microcystin-LR.  Mutat. Res.  559(1-2): 131-142.

 3   LeClaire, R.D., W.B. Lawrence, K. A. Bostian and K. A. Mereish. 1988. Acute toxicity of
 4   microcystin-LR in the rat: A comparative dose-response study using serum chemistries and
 5   mortality as indices. Toxicologist. 8(1):221.

 6   LeClaire, R.D., G.W. Parker and D.R. Franz.  1995. Hemodynamic and calorimetric changes
 7   induced by microcystin-LR in the rat.  J. Appl. Toxicol.  15(4):303-311.

 8   Leiers, T., A. Bihlmayer, H.P.T. Ammon and M.A. Wahl.  2000.  [Ca2+](i)- and insulin-
 9   stimulating effect of the non-membranepermeable phosphatase-inhibitor microcystin-LR in
10   intact insulin-secreting cells (RINmSF).  Br. J. Pharmacol. 130(6):1406-1410.

11   Lin, J.R. and F.S. Chu.  1994. Kinetics of distribution of microcystin-LR in serum and liver
12   cytosol of mice:  an immunochemical analysis. J. Agric. Food Chem. 42(4): 1035-1040.

13   Lovell, R.A., DJ. Schaeffer, S.B. Hooser et al. 1989a. Toxicity of intraperitoneal doses of
14   microcystin-LR in two strains of male mice. J. Environ. Pathol. Toxicol. Oncol.  9(3):221-237.

15   Lovell, R.A., K.R. Holmes, W.W. Carmichael and V.R. Beasley.  1989b. Hemodynamic and
16   pathologic effects of microcystin-LR administered intravenously to swine.  Toxicon. 27(1):60.
17   (Abstract)

18   Maatouk, L, N. Bouaicha, MJ. Plessis and F. Perin. 2004. Detection by 32P-postlabelling of
19   8-oxo-7,8-dihydro-2'-deoxyguanosine in DNA as biomarker of microcystin-LR- and nodularin-
20   induced DNA damage in vitro in primary cultured rat hepatocytes and in vivo in rat liver. Mutat.
21   Res. 564(1):9-20.

22   MacKintosh, C., K.A. Beattie, S. Klumpp et al. 1990. Cyanobacterial microcystin-LR is a
23   potent and specific inhibitor of protein phosphatases  1 and 2A from both mammals and higher
24   plants. FEES Lett. 264(2): 187-192.

25   MacKintosh, R.W., K.N. Dalby, D.G. Campbell, P.T. Cohen, P. Cohen and C. MacKintosh.
26   1995. The Cyanobacterial toxin microcystin binds covalently to cysteine-273  on protein
27   phosphatase 1. FEES Lett. 371(3):236-240.

28   Maidana, M., V. Carlis, F.G. Galhardi  et al. 2006. Effects of microcystins over short- and long-
29   term memory and oxidative stress generation in hippocampus of rats. Chem. Biol. Interact.
30   159(3):223-234.

31   Mankiewicz, J., M. Tarczynska, K.E. Fladmark et al. 2001. Apoptotic effect of Cyanobacterial
32   extract on rat hepatocytes and human lymphocytes.  Environ. Toxicol.  16(3):225-233.

33   Matsushima, R.,  S. Yoshizawa,  M.F. Watanabe et al.  1990. In vitro and in vivo effects of
34   protein phosphatase inhibitors, microcystins and nodularin on mouse skin and fibroblasts.
35   Biochem. Biophys. Res. Commun. 171(2):867-874.
                                               108       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Matsushima-Nishiwaki, R., Y. Shidoji, S. Nishiwaki et al. 1995. Suppression by carotenoids of
 2   microcystin-induced morphological changes in mouse hepatocytes. Lipids.  30(11): 1029-1034.

 3   MattilaK., A. Annila and T.T. Rantala.  2000.  Metal ions mediate the binding of cyanobacterial
 4   toxins to human protein phosphatase I: A computational study. Oulu University Library, Oulun
 5   Yliopisto, Oulu.

 6   Maynes JT, K.R. Perreault, M.M. Cherney, H.A. Luu, M.N. James and C.F. Holmes. 2004.
 7   Crystal structure and mutagenesis of a protein phosphatase-1 :calcineurin hybrid elucidate the
 8   role of the P12-P13 loop in inhibitor binding. J. Biol. Chem.  279(41):43198-43206.

 9   Maynes, J.T., H.A. Luu, M.M. Cherney et al. 2006.  Crystal structures of protein phosphatase-1
10   bound to motuporin and dihydromicrocystin-LA: Elucidation of the mechanism of enzyme
11   inhibition by cyanobacterial toxins. J. Mol. Biol. 356(1):111-120.

12   McDermott, C.M., C.W. Nho, W. Howard and B. Holton. 1998. The cyanobacterial toxin,
13   microcystin-LR can induce apoptosis in a variety of cell types.  Toxicon. 36(12):1981-1996.

14   Mereish, K. A. and R. Solow. 1990. Effect of antihepatotoxic agents against microcystin-LR
15   toxicity in cultured rat hepatocytes. Pharmacol. Res.  7(3):256-259.

16   Mereish, K.A., R. Solow,  Y. Singh et al.  1989.  Comparative toxicity of cyclic polypeptides and
17   depsipeptides on cultured rat hepatocytes.  Toxicologist.  9(1):68.

18   Mereish, K.A., D.L. Bunner, D.R. Ragland and D.A. Creasia.  1991. Protection against
19   microcystin-LR induced hepatoxicity by Silymarin: Biochemistry, histopathology and lethality.
20   Pharm. Res. 8(2):273-277.

21   Meriluoto, J.A., S.E. Nygard, A.M. Dahlem and I.E. Eriksson.  1990.  Synthesis, organotropism
22   and hepatocellular uptake of two tritium-labeled epimers of dihydromicrocystin-LR, a
23   cyanobacterial peptide toxin analog. Toxicon.  28(12): 1439-1446.

24   Metcalf J.S., K.A. Beattie, S. Pflugmacher and G.A. Codd. 2000. Immuno-crossreactivity and
25   toxicity assessment of conjugation products of the cyanobacterial toxin, microcystin-LR.  FEMS
26   Microbiol. Lett.  189(2):155-158.

27   Mikhailov, A., A.S. Harmala-Brasken, J. Hellman et al. 2003. Identification of ATP-synthase as
28   a novel intracellular target for microcystin-LR. Chem-Biol. Inter. 142(3):223-237.

29   Milutinovic, A., B. Sedmak, I. Horvat-Znidarsic and D. Suput.  2002.  Renal injuries induced by
30   chronic intoxication with microcystins. Cell Mol. Biol. Lett.  7(1):139-141.

31   Milutinovic A, M. Zivin, R. Zore-Pleskovic, B. Sedmak and D. Suput. 2003. Nephrotoxic
32   effects of chronic administration of microcystins-LR and-YR.  Toxicon. 42(3):281-288.

33   Miura, G.A., N.A. Robinson, W.B. Lawrence and J.G. Page.  1991. Hepatoxicity of
34   microcystin-LR in fed and fasted rats.  Toxicon. 29(3):337-346.
                                               109       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Moreno, I.M., A. Mate, G. Repetto et al.  2003.  Influence of microcystin-LR on the activity of
 2   membrane enzymes in rat intestinal mucosa. J. Physiol. Biochem.  59(4):293-299.

 3   Moreno, I, S. Pichardo, A. Jos et al.  2005. Antioxidant enzyme activity and lipid peroxidation
 4   in liver and kidney of rats exposed to microcystin-LR administered intraperitoneally.  Toxicon.
 5   45(4):395-402.

 6   Namikoshi M., BW Choi, F Sun, KL Rinehart, WR Evans and W.W. Carmichael.  1993.
 7   Chemical  characterization and toxicity of dihydro derivatives of nodularin and microcystin-LR,
 8   potent cyanobacterial cyclic peptide hepatotoxins. Chem. Res. Toxicol. 6(2):151-158.

 9   Naseem, S.M., Hines, H.B. andD.A. Creasia. 1990. Inhibition of microcystin-induced release
10   of cyclooxygenase products from rat hepatocytes by anti-inflammatory steroids. Proc. Soc. Exp.
11   Biol. Med. 195(3):346-349.

12   Naseem, S.M., K.A. Mereish, R. Solow and H.B. Hines.  1991. Microcystin-induced  activation
13   of prostaglandin synthesis and phospholipids metabolism in rat hepatocytes. Toxicol. In Vitro.
14   5(4):341-345.

15   National Research Council. 1983. Risk Assessment in the Federal Government: Managing the
16   Process. National Academy Press, Washington, DC.

17   Nishiwaki R., T. Ohta, E. Sueoka et al. 1994. Two significant aspects of microcystin-LR:
18   specific binding and liver specificity.  Cancer Lett. 83(l-2):283-289.

19   Nishiwaki-Matsushima, R., S. Nishiwaki, T. Ohta et al. 1991. Structure-function relationships
20   of microcystins, liver tumor promoters, in interaction with protein phosphatase. Jpn. J. Cancer
21   Res. 82(9):993-996.

22   Nishiwaki-Matsushima, R., T. Ohta,  S. Nishiwaki et al. 1992. Liver tumor promotion by the
23   cyanobacterial cyclic peptide toxin microcystin-LR.  J. Cancer Res.  Clin. Oncol.
24   118(6):420-424.

25   Nobre, A.C.L., M.C.M. Jorge, D.B. Menezes et al. 1999.  Effects of microcystin-LR in isolated
26   perfused rat kidney. Brazilian J. Med. Biol. Res.  32(8):985-988.

27   Nobre, A.C.L., G.R. Coelho, M.C.M. Coutinho et al. 2001. The role of phospholipase A(2) and
28   cyclooxygenase in renal toxicity induced by microcystin-LR.  Toxicon.  39(5):721-724.

29   Nobre, A.C.L., A.M.C. Martins, A. Havt et al. 2003. Renal effects  of supernatant from rat
30   peritoneal macrophages activated by microcystin-LR:  Role protein  mediators. Toxicon.
31   41(3):377-381.

32   Nobre, A.C.L., S.M. Nunes-Monteiro, M.C.S.A. Monteiro et al. 2004. Microcystin-LR promote
33   intestinal secretion of water and electrolytes in rats.  Toxicon.  44:555-559.

34   O'Brien, E., S. Altheimer, N. Kreke and D.R. Dietrich.  2003.  Teratogenicity  of cyanobacterial
35   extracts to FETAX embryos. Toxicol.  Sci.  72(Suppl. 1):246.


                                               110      DRAFT - DO NOT CITE OR QUOTE

-------
 1   Ohta, T., R. Nishiwaki, J. Yatsunami et al.  1992.  Hypersphosphorylation of cytokeratins 8 and
 2   18 by microcystin-LR, a new liver tumor promoter, in primary cultured rat hepatocytes.
 3   Carcinogenesis.  13(12): 2443-2447.

 4   Ohta, T., E. Sueoka, N. lida et al.  1994. Nodularin, a potent inhibitor of protein phosphatases 1
 5   and 2A, is a new environmental carcinogen in male F344 rat liver. Cancer Res.
 6   54(24):6402-6406.

 7   Ohtake, A., M. Shirai, T. Aida et al.  1989.  Toxicity of Microcystis species isolated from natural
 8   blooms and purification of the toxin. Appl. Environ. Microbiol. 55(12):3202-3207.

 9   Oishi, S. andM.F. Watanabe.  1986. Acute toxicity of Microcystis aeruginosa and its
10   cardiovascular effects.  Environ. Res. 40(2):518-524.

11   Orr, P.T., GJ. Jones, R.A. Hunter et al. 2001.  Ingestion of toxic Microcystis aeumginosa by
12   dairy cattle and implications for microcystin contamination of milk. Toxicon.
13   39(12):1847-1854.

14   Orr, P.T., GJ. Jones, R.A. Hunter and K. Berger.  2003.  Exposure of beef cattle to sub-clinical
15   doses of Microcystis aeruginosa: Toxin bioaccumulation, physiological effects and human
16   health risk assessment.  Toxicon. 41(5):613-620.

17   Pace, J.G., N.A. Robinson, G.A. Miura et al. 1991.  Toxicity and kinetics of [3H]microcystin-LR
18   in isolated perfused rat livers.  Toxicol. Appl. Pharmacol. 107(3):391-401.

19   Pican90 M.R., R.M. Scares, V.R. Cagido, S.M.F.O. Azevedo, P.R.M. Rocco and W.A. Zin.
20   2004. Toxicity of cyanobacterial extract containing microcystins to mouse lungs.  Braz. J. Med.
21   Biol. Res. 37(8): 1225-1229.

22   Pilotto, L.S., R.M. Douglas, M.D. Burch et al.  1997. Health effects of exposure to
23   cyanobacteria (blue-green algae) during recreational water-related activities. Aust. NZ J. Pub.
24   Health.  21(6):562-566.

25   Pilotto, L.S.,E.V. Klewer, R.D. Davies, M.D. Burch and R.G. Attewell.  1999.  Cyanobacterial
26   (blue-green algae) contamination in drinking water and perinatal outcomes. Aust. NZ J. Pub.
27   Health.  23(2): 154-158.

28   Pilotto L., P. Hobson, M.D. Burch, G. Ranmuthugala, R. Attewell and W. Weightman. 2004.
29   Acute skin  irritant effects of cyanobacteria (blue-green algae) in healthy volunteers. Aust. NZ J.
30   Pub. Health.  28(3):220-224.

31   Porfmo, Z., M.P. Ribeiro, C.S. Esstevam et al.  1999. Hepatospenomegaly caused by an extract
32   of cyanobacterium Microcystis aeurignosa bloom collected in the Manguaba lagoon,  Alagoas-
33   Brazil. Rev. Mi crobiol. 30(3):278-285.

34   Pouria, S, A.  de Andrade, J. Barbosa et al. 1998.  Fatal microcystin intoxication in
35   haemodialysis unit in Caruaro, Brazil.  Lancet. 352:21-26.
                                                111       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Rao, P.V.L. and R. Bhattacharya.  1996.  The cyanobacterial toxin microcystin-LR induced
 2   DNA damage in mouse liver in vivo. Toxicology.  114(l):29-36.

 3   Rao, P.V.L., R. Bhattacharya and S.D.  Gupta.  1994.  Isolation, culture, and toxicity of the
 4   cyanobacterium (blue-green alga) Microcystis aeruginosa from a freshwater source in India.
 5   Bull. Environ. Contam. Toxicol. 52(6):878-885.

 6   Rao, P.V., R. Jayaraj and A.S. Bhaskar. 2004. Protective efficacy and the recovery profile of
 7   certain chemoprotectants against lethal poisoning by microcystin-LR in mice. Toxicon.
 8   44(7):723-730.

 9   Rao P.V.L., N. Gupta, R. Jayaraj et al.  2005. Age-dependent effects on biochemical variables
10   and toxicity induced by cyclic peptide toxin microcystin-LR in mice.  Comp. Biochem. Physiol.
11   Part C: Toxicol. Pharmacol.  140(1):11-19.

12   Repavich, W.M., W.C. Sonzogni, J.H.  Standridge, R.E. Wedepohl and L.F. Meisner. 1990.
13   Cyanobacteria (blue-green algae) in Wisconsin waters:  Acute and chronic toxicity. Water Res.
14   24(2):225-231.

15   Rinehart K.L., M. Namikoshi and B.W. Choi. 1994.  Structure and biosynthesis of toxins from
16   blue-green algae (cyanobacteria).  J. Appl. Phycol.  6:159-176.

17   Robinson, N.A., G. A. Miura, C.F.  Matson et al.  1989.  Characterization of chemically tritiated
18   microcystin-LR and its distribution in mice. Toxicon. 27(9): 1035-1042.

19   Robinson, N.A., J.G. Pace, C.F. Matson et al. 1991a. Tissue distribution, excretion and hepatic
20   biotransformation of microcystin-LR in mice. J. Pharmacol. Exp. Therapeut. 256(1): 176-182.

21   Robinson, N.A., C.F. Matson and J.G. Pace. 1991b.  Association of microcystin-LR and its
22   biotransformation product with a hepatic-cystolic protein. J. Biochem. Toxicol. 6(3):171-180.

23   Rogers, E.H., E.S. Hunter, V.C. Moseretal. 2005. Potential developmental toxicity of
24   anatoxin-a, a cyanobacterial toxin.  J. Appl. Toxicol.  25(6):527-534.

25   Runnegar, M.T.C. and I.R. Falconer. 1982.  The in vivo and in vitro biological effects of the
26   peptide hepatotoxin from the blue-green alga Microcystis aeruginosa.  S. Afr. J. Sci.
27   78:363-366.

28   Runnegar, M.T.C. and I.R. Falconer. 1986.  Effect of toxin from the cyanobacterium
29   Microcystis aeruginosa on ultrastructural morphology and actin  polymerization in isolated
30   hepatocytes. Toxicon.  24(2): 109-115.

31   Runnegar, M.T.C., I.R. Falconer and J. Silver. 1981. Deformastion of isolated rat  hepatocytes
32   by a peptide hepatoxin from the blue-green alga Microcystis aeruginosa. Nayn-Schmied Arch.
33   Pharmacol.  317(3):268-272.
                                                112       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Runnegar, M.T.C., I.R. Falconer, T. Buckley and A.R.B. Jackson.  1986.  Lethal potency and
 2   tissue distribution of 125I-labelled toxic peptides from the blue-green alga Microcystis
 3   aeruginosa. Toxicon. 24(5):506-509.

 4   Runnegar, M.T.C., J. Andrews, R.G. Gerdes and I.R. Falconer.  1987. Injury to hepatocytes
 5   induced by a peptide toxin from the cyanobacterium Microcystis aeruginosa. Toxicon.
 6   25(11):1235-1239.

 7   Runnegar, M.T.C., N. Berndt and N. Kaplowitz.  1991.  Identification of hepatic protein
 8   phosphatases as novel critical targets for the hepatotoxicity of microcystin in vivo. Hepatology.
 9   14(4):A159.

10   Runnegar, M.T.C., S. Kong and N. Berndt.  1993. Protein phosphatase inhibition and in vivo
11   hepatotoxicity of microcystins. Am. J. Physiol. 265(2):G224-G230.

12   Runnegar, M.T.C., L.D. Deleve and N. Berndt.  1994.  The effects of the protein phosphatase
13   inhibitors and microcystin and calyculin A differ in hepatocytes and hepatic endothelial cells.
14   FASEBJ.  8(7): 1231.

15   Runnegar, M., N. Berndt, S.M. Kong, E.Y. Lee and L. Zhang. 1995a. In vivo and in vitro
16   binding of microcystin to protein phosphatases 1 and 2A. Biochem. Biophys. Res. Commun.
17   216(1):162-169.

18   Runnegar, M.T.C., N. Berndt and N. Kaplowitz.  1995b. Microcystin uptake and inhibition of
19   protein phosphatases:  effects of chemoprotectants and self-inhibition in relation to known
20   hepatic transporters.  Toxicol.  Appl. Pharmacol.  134(2):264-272.

21   Sabour, B., M. Loudiki, B. Oudra et al. 2002.  Toxicology  of & Microcystis ichthyoblabe
22   waterbloom from Lake Oued Mellah (Morocco). Environ.  Toxicol. 17(1):24-31.

23   Schaeffer, D.J., P.B. Malpas and L.L. Barton.  1999. Risk assessment of microcystin in dietary
24   Aphanizomenon flos-aquae. Ecotoxicol. Environ. Safety. 44(1):73-80.

25   Sekijima, M., T.  Tsutsumi, T. Yoshida et al.  1999.  Enhancement of glutathione S-transferase
26   placental-form positive liver cell foci development by microcystin-LR in aflatoxin Bi-initiated
27   rats. Carcinogenesis.  20(1): 161-165.

28   Sepulveda, M.S., M. Rojas and F. Zambrano.  1992. Inhibitory effect of ^Microcystis sp.
29   (cyanobacteria) toxin on development of preimplantation mouse  embryos. Comp. Biochem.
30   Physiol.-C Pharmacol. Toxicol. Endocrinol.  102(3):549-553.

31   Shen, X., G.R. Shaw, G.A. Codd et al.  2003.  DNA microarray analysis of gene expression in
32   mice treated with the cyanobacterial toxin, cylindrospermopsin.  Can. Tech. Rep. Fish Aquat.
33   Sci. 2498:49-51.

34   Shi, W., H. Zhu, X. Yan and Z. Zhou.  2002. [Sub-acute hepatotoxicity of low doses of
35   microcystins.] Huan Jing Ke Xue. 23(5):47-51.  (Chinese)
                                               113       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Shirai, M., Y. Takamura, H. Sakuma et al. 1986. Toxicity and delayed type hypersensitivity
 2   caused by Microcystis blooms from Lake Kasumigaura. Microbiol. Immunol. 30(7):731 -73 5.

 3   Sicihska, P., B. Bukowska, J. Michalowicz and W. Duda.  2006.  Damage of cell membrane and
 4   antioxidative system in human erythrocytes incubated with microcystin-LR in vitro. Toxicon.
 5   47(4):387-397.

 6   Siegelman, H.W., W.H. Adams, R.D. Stoner and D.N.  Slatkin. 1984.  Toxins  of Microcystis
 7   aeruginosa and their hematological and histopathological effects. In: Seafood Toxins, E.P.
 8   Ragelis, Ed. American Chemical Society, Washington, DC. p. 407-413.

 9   Sim, A.T.R. and L.M.  Mudge.  1993. Protein phosphatase activity in cyanobacteria:
10   consequences for microcystin toxicity analysis.  Toxicon.  31(9): 1179-1186.

11   Slatkin, D.N., R.D. Stoner, W.H. Adams et al.  1983. Atypical pulmonary thrombosis caused by
12   a toxic cyanobacterial  peptide.  Science. 220:1383-1385.

13   Scares, R.M., M. Yuan, J.C. Servaites et al. 2006.  Sublethal exposure from microcystins to
14   renal insufficiency patients in Rio de Janeiro, Brazil. Environ. Toxicol. 21(2):95-103.

15   Solow, R., K. Mereish, G.W. Anderson, Jr. et al. 1989. Effect of microcystin-LR on cultured rat
16   endothelial cells.  Toxicologist. 9(1): 160.

17   Solter, P.P., G.K. Wollenberg, X. Huang et al.  1998. Prolonged sublethal exposure to the
18   protein phosphatase inhibitor microcystin-LR results in multiple  dose-dependent hepatotoxic
19   effects. Toxicol. Sci.  44(l):87-96.

20   Solter, P., Z.L. Lui and R. Guzman. 2000. Decreased  hepatic ALT synthesis is an outcome of
21   subchronic microcystin-LR toxicity. Toxicol. Appl. Pharmacol.  164(2):216-220.

22   Stotts, R.R., M. Namikoshi, W.M. Haschek et al. 1993. Structural modifications imparting
23   reduced toxicity in microcystins from Microcystis spp.  Toxicon.  31(6):783-789.

24   Stotts, R.R., A.R.  Twardock, W.M.  Haschek et al. 1997a. Distribution of tritiated
25   dihydromicrocystin in swine.  Toxicon. 35(6):937-953.

26   Stotts, R.R., A.R.  Twardock, G.D. Koritz et al.  1997b. Toxicokinetics of tritiated
27   dihydromicrocystin-LR in swine.  Toxicon. 35(3):455-465.

28   Suzuki, H., M.F. Watanabe, Y.P. Yu et al. 1998. Mutagenicity of microcystin-LR in human
29   RSA cells. Int. J.  Mol. Med. 2(1): 109-112.

30   Takahashi, O., S. Oishi and M.F. Watanabe.  1995.  Defective blood coagulation is not causative
31   of hepatic haemorrhage induced by microcystin-LR. Pharmacol. Toxicol. 76(4):250-254.

32   Takenaka, S. 2001.  Covalent glutathione conjugation  to cyanobacterial hepatotoxin
33   microcystin-LR by F344 rat cytosolic and microsomal  glutathione S-transferases. Environ.
34   Toxicol. Pharmacol. 9(4):135-139.
                                               114       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Tarczynska M., G. Nalecz-Jawecki, M. Brzychcy et al. 2000.  The toxicity of cyanobacterial
 2   blooms as determined by microbiotests and mouse assays. In: New Microbiotests for Routine
 3   Toxicity Screening and Biomonitoring, Persoone et al., Eds. Kluwer Academic/Plenum
 4   Publishers, New York, NY. pp. 527-532.

 5   Teixeira, M.G.L.C., M.C.N. Costa, V.L.P. Carvalho et al.  1993. Gastroenteritis epidemic in the
 6   area of the Itaparica Dam, Bahia, Brazil.  Bulletin of PAHO. 27(3):244-253.

 7   Teneva, I, R. Mladenov, N. Popov and B. Dzhambazov.  2005. Cytotoxicity and apoptotic
 8   effects of microcystin-LR and anatoxin-a in mouse lymphocytes. Folia. Biol. (Praha).
 9   51(3):62-67.

10   Theiss, W.C., W.W. Carmichael, M.M. Mullins and R.H. Bruner.  1984.  Hepatotoxicity of the
11   blue-green algae (cyanobacteria)Microcystis aeruginosa.  Toxicologist.  4(1): 12.

12   Theiss, W.C., W.W. Carmichael, W.W. Mullins et al.  1985. Direct rapid hepatotoxicity of the
13   cyanobacteria Microcystis aeruginosa. Toxicon.  23(1):40-41.

14   Theiss, W.C., W.W. Carmichael, J. Wyman and R. Bruner. 1988. Blood pressure and
15   hepatocellular effects of the cyclic heptapeptide toxin produced by Microsystis aeruginosa strain
16   PCC-7820. Toxicon. 26(7):603-613.

17   Thiel, P.  1994. The South African contribution to studies on the toxic cyanobacteria and their
18   toxins. In: Toxic Cyanobacteria: Current Status of Research and Management.  Proceedings of
19   an International Workshop.  Adelaide, Australia, March 22-26.  D.A. Steffensen and B.C.
20   Nicholson, Ed.  Australian Centre for Water Quality Research, Salisbury, Australia,  pp. 23-27.

21   Thomspon, W.L. and J.G. Pace. 1992. Substances that protect cultured hepatocytes from the
22   toxic effects of microcystin-LR. Toxicol. In Vitro.  6(6):579-587.

23   Thompson, W.L.,  M.B. Allen and K.A. Bostian.  1988. The effects of microcystin on
24   monolayers of primary rat hepatocytes. Toxicon. 26(1):44.

25   Toivola, D., I.E. Eriksson and D.L. Brautigan.  1994.  Identification of protein phosphatase 2A
26   as the primary target for microcystin-LR in rat liver homogenates.  FEES Lett.  344(2-3): 175-
27   180.

28   Toivola, D.M., R.D. Goldman, D.R. Garrod and I.E. Eriksson.  1997. Protein phosphatases
29   maintain the organization and structural interactions of hepatic keratin intermediate filaments.  J.
30   CellSci.  110(Pt. l):23-33.

31   Towner, R. A., S. A. Sturgeon and K.E. Hore. 2002. Assessment of in vivo oxidative lipid
32   metabolism following acute microcystin-LR-induced hepatotoxicity in rats.  Free Radical Res.
33   36(1):63-71.

34   Turner, P.C., AJ.  Gammie, K. Hollinrake and G.A. Codd. 1990. Pneumonia associated with
35   contact with cyanobacteria.  Br. Med. J. 300(6737):1440-1441.
                                               115       DRAFT - DO NOT CITE OR QUOTE

-------
 1   Ueno, Y., S. Nagata, T. Tsutsumi et al. 1996. Detection of microcystins, a blue-green algal
 2   hepatotoxin, in drinking water sampled in Haimen and Fusui, endemic areas of primary liver
 3   cancer in China, by highly sensitive immunoassay.  Carcinogenesis.  17(6):1317-1321.

 4   Ueno, Y., Y. Makita, S. Nagata et al.  1999.  No chronic oral toxicity of a low-dose of
 5   microcystin-LR, a cyanobacterial hepatoxin, in female Balb/C mice.  Environ. Toxicol.
 6   14(l):45-55.

 7   U.S. EPA.  1986a. Guidelines for the Health Risk Assessment of Chemical Mixtures. Fed. Reg.
 8   51(185):34014-34025.

 9   U.S. EPA.  1986b. Guidelines for Mutagenicity Risk Assessment. Fed. Reg.
10   51(185):34006-34012.

11   U.S. EPA.  1988. Recommendations for and Documentation of Biological Values for Use in
12   Risk Assessment.  U.S. Environmental Protection Agency, Office of Research and Development,
13   Washington, DC. EPA 600/6-87/008. NTIS PB88-179874/AS.

14   U.S. EPA.  1991. Guidelines for Developmental Toxicity RiskAssessment. Fed. Reg.
15   56(234):63798-63826.

16   U.S. EPA.  1994a. Interim Policy for Particle Size and Limit Concentration Issues in Inhalation
17   Toxicity Studies. Fed. Reg.  59(206):53799.

18   U.S. EPA.  1994b. Methods for Derivation of Inhalation Reference Concentrations and
19   Application of Inhalation Dosimetry.  U.S. Environmental Protection Agency, Office of
20   Research and Development, Washington, DC. EPA/600/8-90/066F.  Available at
21   http://www.epa.gov/iris/backgr-d.htm.

22   U.S. EPA.  1995. Use of the Benchmark Dose Approach in Health Risk Assessment. U.S.
23   Environmental Protection Agency. U.S. Environmental Protection Agency, Office of Research
24   and Development, Washington, DC.  EPA/630/R-94/007. Available at
25   http://www.epa.gov/iris/backgr-d.htm.

26   U.S. EPA.  1996. Guidelines for Reproductive Toxicity Risk Assessment. Fed. Reg.
27   61(212):56274-56322.

28   U.S. EPA.  1998a. Guidelines for Neurotoxicity Risk Assessment. Fed. Reg.
29   63(93):26926-26954.

30   U.S. EPA.  1998b. Science Policy Council Handbook: Peer Review. Prepared by the Office of
31   Science Policy, Office of Research and Development, Washington, DC. EPA/100/B-98/001.
32   Available at http://www.epa.gov/iris/backgr-d.htm.

33   U.S. EPA.  2000a. Science Policy Council Handbook: Peer Review, 2nd ed. U.S. Environmental
34   Protection Agency, Office of Research and Development, Office of Science Policy, Washington,
35   DC. EPA/100/B-00/001. Available at http://www.epa.gov/iris/backgr-d.htm.
                                              116       DRAFT - DO NOT CITE OR QUOTE

-------
 1   U.S. EPA. 2000b.  Science Policy Council Handbook: Risk Characterization. U.S.
 2   Environmental Protection Agency, Office of Research and Development, Office of Science
 3   Policy, Washington, DC. EPA/100/B-00/002. Available at
 4   http://www.epa.gov/iris/backgr-d.htm.

 5   U.S. EPA. 2000c.  Benchmark Dose Technical Guidance Document: External Review Draft.
 6   U.S. Environmental Protection Agency, Office of Research and Development, Office of Science
 7   Policy, Washington, DC. EPA/630/R-00/001. Available at
 8   http://www.epa.gov/iris/backgr-d.htm.

 9   U.S. EPA. 2000d.  Supplemental Guidance for Conducting for Health Risk Assessment of
10   Chemical Mixtures. U.S. Environmental Protection Agency, Office of Research and
11   Development, Office of Science Policy, Washington, DC. EPA/630/R-00/002. Available at
12   http://www.epa.gov/iris/backgr-d.htm.

13   U.S. EPA. 2002. A Review of the Reference Dose and Reference Concentration Processes. U.S.
14   Environmental Protection Agency, Risk Assessment Forum, Washington, DC.
15   EPA/630/P-02/0002F.  Available at http://www.epa.gov/iris/backgr-d.htm.

16   U.S. EPA. 2005a.  Guidelines for Carcinogen Risk Assessment. U.S. Environmental Protection
17   Agency, Risk Assessment Forum, Washington, DC. EPA/630/P-03/001B. Available at
18   http://www.epa.gov/iris/backgr-d.htm.

19   U.S. EPA. 2005b.  Supplemental Guidance for Assessing Susceptibility from Early-Life
20   Exposure to Carcinogens. U.S. Environmental Protection Agency, Risk Assessment Forum,
21   Washington, DC. EPA/630/R-03/003F. Available at http://www.epa.gov/iris/backgr-d.htm.

22   U.S. EPA. 2005c.  Peer Review Handbook. 3rd edition. Review draft.  Science Policy Council,
23   Washington, DC. Available at: http://intranet.epa.gov/ospintra/scipol/prhndbk05.doc.

24   Vesterkvist, P.S. and J.A. Meriluoto.  2003. Interaction between microcystins of different
25   hydrophobicities and lipid monolayers. Toxicon. 41(3): 349-355.

26   WHO (World Health Organization). 1999.  Toxic Cyanobacteria in Water: A Guide to their
27   Public Health Consequences, Monitoring, and Management, I. Chorus and J. Bartram, Eds.
28   E&FN Spon, London, UK

29   Wickstrom, M.L., S.A. Khan, W.M. Haschek et al. 1995. Alterations in microtubules,
30   intermediate filaments and microfilaments induced by microcystin-LR in cultured cells.
31   Toxicologic. Pathol. 23(3):326-337.

32   Wolf, H.-U. and C. Frank. 2002.  Toxicity assessment of cyanobacerial toxin mixtures.
33   Environ. Toxicol.  17(4): 395-399.

34   Xu, L., P.K.S. Lam, J. Chen et al. 2000.  Comparative study on in vitro inhibition of grass carp
35   (Ctenopharyngodon idellus) and  mouse protein phosphatases by microcystins. Environ. Toxicol.
36   15(2):71-75.
                                               117      DRAFT - DO NOT CITE OR QUOTE

-------
 1   Yea, S.S., Y.I. Yang, W.H. Jang and K.H. Paik.  2001.  Microcystin-induced proinflammatory
 2   cytokines expression and cell death in human hepatocytes. Hepatology. 34(4 Pt. 2 Suppl.):516A

 3   Yoshida, T., Y. Makita, S. Nagata et al.  1997. Acute oral toxicity of microcystin-LR, a
 4   cyanobacterial hepatotoxin in mice.  Nat. Toxins. 5:91-95.

 5   Yoshida, T., Y. Makita, T. Tsutsumi et al.  1998.  Immunohistochemical localization of
 6   microcystin-LR in the liver of mice: A study on the pathogenesis of microcystin-LR-induced
 7   hepatotoxicity.  Toxicologic. Pathol.  26(3):411-418.

 8   Yoshizawa, S., R. Matsushima, M.F. Watanabe et al. 1990. Inhibition of protein phosphatases
 9   by Microcystis and nodularin associated  with hepatotoxicity. J. Cancer Res. Clin. Oncol.
10   116(6):609-614.

11   Yu, S.-Z. 1989.  Drinking water and primary liver cancer. In: Primary Liver Cancer, Z.Y. Tang,
12   M.C. Wu and S.S. Xia, Ed.  China Academic Publishers, New York, NY. p. 30-37 (as cited in
13   Ueno et al., 1996 and Health Canada, 2002).

14   Yu, S.-Z., Z.-Q. Chen, Y.-K. Liu, Z.-Y. Huang and Y.-F. Zhao.  1989.  The aflatoxins and
15   contaminated water in the etiological study of primary liver cancer. In: Mycotoxins and
16   Phycotoxins'88, S. Natori, K. Hashimoto and Y. Ueno, Eds. Elsevier, Amsterdam, p. 37-44 (as
17   cited in Ueno et al., 1996 and Health Canada, 2002).

18   Yu, S.Z., X.E. Huang, T. Koide et al. 2002. Hepatitis B and C viruses infection, lifestyle and
19   genetic polymorphisms as risk factors for hepatocellular carcinoma in Haimen, China.  Jpn. J.
20   Cancer Res.  93(12): 1287-1292.

21   Zegura, B., B. Sedmak and M. Filipic. 2003. Microcystin-LR induces oxidative DNA damage
22   in human hepatoma cell line HepG2. Toxicon.  41(l):41-48.

23   Zegura, B., T.T. Lah and M. Filipic. 2004. The role of reactive oxygen species in microcystin-
24   LR-induced DNA damage. Toxicology.  200(l):59-68.

25   Zhan, L., M. Sakamoto, M. Sakuraba et al. 2004. Genotoxicity of microcystin-LR in human
26   lymphoblastoid TK6 cells.  Mutat. Res.  557(l):l-6.

27   Zhang, Z., S. Kang, C. Chen et al. 2002. [The acute toxic effects of microcystin-LR in SD rats.]
28   Zhonghua Yu Fang YiXueZaZhi.  36(5):295-297. (Chinese)

29   Zhao, J.M. and H.G. Zhu. 2003. [Effects of microcystins on cell cycle and expressions of c-fos
30   andc-jun.] Zhonghua Yu Fang Yi Xue Za Zhi.  37(l):23-25.  (Chinese)

31   Zhou, L., H.  Yu and K. Chen.  2002. Relationship between microcystin in drinking water and
32   colorectal cancer. Biomed. Environ. Sci. 15(2):166-171.

33   Zhu, Y., X. Zhong, S. Zheng, Z. Ge, Q. Du and S. Zhang. 2005. Transformation of
34   immortalized colorectal crypt cells by microcystin involving constitutive activation of Akt and
35   MAPK cascade.  Carcinogenesis.  26(7): 1207-1214.


                                               118       DRAFT - DO NOT CITE OR QUOTE

-------
1   Zuzek, M.C., M. Kosec, J. Mrkum et al.  2003.  Microcystin-LR causes reorganization of actin
2   filaments and microtubules in rabbit whole embryo cultures. Toxicol. Lett. 144(Suppl 1):568.
                                             119      DRAFT - DO NOT CITE OR QUOTE

-------
 1                                       APPENDIX A
 2
 3
 4    Contents: Benchmark Dose Modeling Output Files for:
 5
 6    1) Heinze, 1999 Liver Lesions
 7    2) Fawell et al., 1999 Male Chronic Liver Inflammation
 8    3) Fawell et al., 1999 Female Chronic Liver Inflammation
 9    4) Fawell et al., 1999 Male and Female (combined) Chronic Liver Inflammation
10    5) Heinze, 1999 Relative Liver Weight Changes
11    6) Heinze, 1999 Lactate Dehydrogenase Changes
12    7) Heinze, 1999 Alkaline Phosphatase Changes
13    8) Fawell et al., 1999 Male Alanine Aminotransferase Changes
                                             A-l        DRAFT: DO NOT CITE OR QUOTE

-------
 1    Heinze,1999. Liver Lesions
 2    ====================================================================
 3             $Revision: 2.2 $ $Date: 2001/03/14 01:17:00 $
 4            Input Data File: C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
 5            Gnuplot Plotting File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
 6                                              Thu Jul 14 15:03:08 2005
 7     ====================================================================
 8
 9     BMDS MODEL RUN

11
12       The form of the probability function is:
13
14       P[response]= backgrounds- (1-background)*CumGamma[slope*dose,power] ,
15       where CumGamma(.)  is the cummulative Gamma distribution function
16
17
18       Dependent variable = COLUMN2
19       Independent variable = COLUMN1
20       Power parameter is restricted as power >=1
21
22       Total number of observations = 3
23       Total number of records with missing values = 0
24       Maximum number of iterations = 250
25       Relative Function Convergence has been set to: le-008
26       Parameter Convergence has been set to:  le-008
27
28
29
30                      Default Initial (and Specified) Parameter Values
31                         Background =    0.0454545
32                              Slope =    0.0153804
33                              Power =      1.02976
34
35
36               Asymptotic Correlation Matrix of Parameter Estimates
37
38               ( *** The model parameter(s)  -Background    -Power
39                     have been estimated at a boundary point, or have been
40    specified by the user,
41                     and do not appear in the correlation matrix )
42
43                      Slope
44
45         Slope            1
46
47
48
49                              Parameter Estimates
50
51           Variable           Estimate             Std. Err.
52         Background                   0               NA
53              Slope           0.0166997          0.00500499
54              Power                   1               NA
55
56    NA - Indicates that this parameter has hit a bound
57         implied by some inequality constraint and thus
58         has no standard error.
59
60
61
62                            Analysis of Deviance Table
63
64           Model      Log(likelihood)  Deviance  Test DF     P-value
                                           A-2       DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30












Full model -9.98095
Fitted model -10.0255 0.089063 2
Reduced model -20.7944 21.6269 2

AIC: 22.051


Goodness of Fit



0.9564
<.0001






Scaled
Dose Est._Prob. Expected Observed Size Residual


0.0000 0.0000 0.000 0
50.0000 0.5661 5.661 6
150.0000 0.9183 9.183 9



10 0
10 0.2162
10 -0.2115

Chi-square = 0.09 DF = 2 P-value = 0.9553


Benchmark Dose Computation

Specified effect = 0.1

Risk Type = Extra risk

Confidence level = 0.95

BMD = 6.30914

BMDL = 3.92229
Gamma Multi-Hit Model with 0.95 Confidence Level
1
0.8
1 0.6
£=
g
| 0.4
0.2
0


Gamma Multi-Hit
^--^^^
<,,//
; ./
/
~ /
/














:
:
:


~
:
BMDL BMD :
0 20 40 60 80 100 120 140
dose









160

31
        15:0307/142005
                                           A-3
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2    ====================================================================
 3            Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
 4            Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS\_CYANO
 5    TOX REV\HEINZE_LIVER_LESIONS.(d)
 6            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 7    DOCUMENTS\_CYANO TOX REV\HEINZE_LIVER_LESIONS.pit
 8                                              Thu Mar 09 11:41:20 2006
 9     ====================================================================
10
11     BMDS MODEL RUN
12    	
13
14       The form of the probability function is:
15
16       P[response]  = I/[1+EXP(-intercept-slope*dose)]
17
18
19       Dependent variable = COLUMN2
20       Independent variable = COLUMN1
21       Slope parameter is not restricted
22
23       Total number of observations = 3
24       Total number of records with missing values = 0
25       Maximum number of iterations = 250
26       Relative Function Convergence has been set to:  le-008
27       Parameter Convergence has been set to: le-008
28
29
30
31                      Default Initial Parameter Values
32                         background =            0   Specified
33                          intercept =     -2.28075
34                              slope =    0.0300564
35
36
37               Asymptotic Correlation Matrix of Parameter Estimates
38
39               ( *** The model parameter(s)   -background
40                     have been estimated at a boundary point, or have been
41    specified by the user,
42                     and do not appear in the correlation matrix )
43
44                  intercept        slope
45
46     intercept            1        -0.75
47
48         slope        -0.75            1
49
50
51
52                              Parameter Estimates
53
54           Variable           Estimate             Std. Err.
55          intercept            -2.02314            0.772069
56              slope           0.0344016           0.0123743
57
58
59
60                            Analysis of Deviance Table
61
62           Model      Log(likelihood)  Deviance  Test DF     P-value
63         Full model        -9.98095
64       Fitted model        -12.1529       4.34392      1         0.03714
                                          A-4       DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
_L\7
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Reduced model

AIC:





Dose Est

0.0000 0
50.0000 0
150.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

-20.7944 21.6269 2 <.

28.3058


Goodness of Fit


._Prob. Expected Observed Size

.1168 1.168 0 10
.4248 4.248 6 10
.9584 9.584 9 10

3.43 DF = 1 P-value = 0.0639


Computation

0.1

Extra risk

0.95

19.4327

11.4032

.0001






Scaled
Residual

-1.15
1.121
-0.9248
















30
31
32
                                   Logistic Model with 0.95 Confidence Level
            0.6
            0.4
                    Logistic
                     BMDL
                            BMD
                                   40
                                           60
                                                  80


                                                dose
                                                          100
                                                                 120
                                                                         140
                                                                                160
          11:41 03/092006
                                                        A-5
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Logistic Model $Revision:  2.1 $ $Date: 2000/02/26 03:38:20 $
 3            Input Data File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
 4            Gnuplot Plotting File:   C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
 5                                              Thu Jul 14 15:04:46 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]
14
15
16       Dependent variable = COLUMN2
17       Independent variable = COLUMN1
18       Slope parameter is restricted as slope >= 1
19
20       Total number of observations =  3
21       Total number of records with missing values = 0
22       Maximum number of iterations =  250
23       Relative Function Convergence has been set to: le-008
24       Parameter Convergence has been  set to: le-008
25
26
27
28       User has chosen the log transformed model
29
30
31                      Default Initial  Parameter Values
32                         background =             0
33                          intercept =      -5.97477
34                              slope =       1.63093
35
36
37               Asymptotic Correlation  Matrix of Parameter Estimates
38
39               ( *** The model parameter(s)   -background
40                     have been estimated at a boundary point, or have been
41    specified by the user,
42                     and do not  appear in the correlation matrix )
43
44                  intercept        slope
45
46     intercept            1        -0.99
47
48         slope        -0.99            1
49
50
51
52                              Parameter Estimates
53
54           Variable           Estimate             Std.  Err.
55         background                   0               NA
56          intercept            -5.97477             4.77026
57              slope             1.63093             1.12507
58
59    NA - Indicates that this parameter has hit a bound
60         implied by some inequality constraint and thus
61         has no standard error.
62
63     Warning: Likelihood for the fitted model larger than the Likelihood for the
64    full model.
                                          A-6       DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
_L /
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37


Model
Full model
Fitted model
Reduced model

AIC:





Dose Est

0.0000 0
50.0000 0
150.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


Analysis of Deviance Table

Log (likelihood) Deviance Test DF P-value
-9.98095
-9.98095 -3.55271e-015 1 -1
-20.7944 21.6269 2 <.0001

23.9619


Goodness of Fit

Scaled
._Prob. Expected Observed Size Residual

.0000 0.000 0 10 0
.6000 6.000 6 10 -3.44e-015
.9000 9.000 9 10 -1.123e-014

0.00 DF = 1 P-value = 1.0000


Computation

0.1

Extra risk

0.95

10.137

1.23783


A-7
DRAFT: DO NOT CITE OR QUOTE

-------
3   0.6
    0.4
                         Log-Logistic Model with 0.95 Confidence Level
             Log-Logistic
         BMDL
                 BMD
                    20       40
                                   60
                                           80




                                        dose
                                                  100     120      140     160
  15:0407/142005
                                                 A-8         DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Multistage Model.  $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
 3            Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS\_CYANO
 4    TOX REV\HEINZE_LIVER_LESIONS.(d)
 5            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 6    DOCUMENTS\_CYANO TOX REV\HEINZE_LIVER_LESIONS.pit
 7                                              Thu Mar 09 11:37:17 2006
 8     ====================================================================
 9
10     BMDS MODEL RUN

12
13       The form of the probability function is:
14
15       P[response]  = background +  (1-background)*[1-EXP(
16    -betal*dose^l-beta2*dose^2) ]
17
18       The parameter betas are restricted to be positive
19
20
21       Dependent variable = COLUMN2
22       Independent variable =  COLUMN1
23
24     Total number of observations  =  3
25     Total number of records with  missing values = 0
26     Total number of parameters in model = 3
27     Total number of specified parameters = 0
28     Degree of polynomial = 2
29
30
31     Maximum number of iterations  =  250
32     Relative Function Convergence has been set to: le-008
33     Parameter Convergence has been  set to: le-008
34
35
36
37                      Default  Initial Parameter Values
38                         Background  =    0.0617654
39                            Beta(l)  =     0.015138
40                            Beta (2)  =            0
41
42
43               Asymptotic Correlation Matrix of Parameter Estimates
44
45     (  *** The model parameter(s)   -Background    -Beta(2)
46     have been estimated at a  boundary point, or have been specified by the
47    user,  and do not appear in the correlation matrix )
48
49                    Beta(l)
50
51       Beta(l)             1
52
53
54
55                              Parameter Estimates
56
57           Variable           Estimate             Std.  Err.
58         Background                   0               NA
59            Beta(l)            0.0166997          0.00582148
60            Beta(2)                    0               NA
61
62    NA  - Indicates that this parameter has hit a bound
63         implied by some inequality  constraint and thus
64         has no standard error.
                                          A-9       DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
_L /
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39


Model
Full model
Fitted model
Reduced model

AIC:




Dose Est

i: 1
0.0000 0.
i: 2
50.0000 0.
i: 3
150.0000 0.

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

Analysis of Deviance Table

Log (likelihood) Deviance Test DF
-9.98095
-10.0255 0.089063 2
-20.7944 21.6269 2

22.051


Goodness of Fit

._Prob. Expected Observed


0000 0.000 0

5661 5.661 6

9183 9.183 9

0.09 DF = 1 P-value =


Computation

0.1

Extra risk

0.95

6.30914

3.92229



P-value

0.9564
<.0001






Size Chi^2 Res.


10 -1.000

10 0.138

10 -0.244

0.7623














A-10
DRAFT: DO NOT CITE OR QUOTE

-------
                          Multistage Model with 0.95 Confidence Level
    0.6
o   0.4
    0.2
            Multistage
          BiylDL BMD




            0
                    20      40      60      80      100     120      140




                                        dose
                                                                         160
  11:37 03/092006
                                                A-l 1         DRAFT: DO NOT CITE OR QUOTE

-------
 1     ====================================================================
 2            Multistage Model.  $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
 3            Input Data File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
 4            Gnuplot Plotting File:   C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
 5                                              Thu Jul 14 15:05:31 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(
14    -betal*dose^l) ]
15
16       The parameter betas are restricted to be positive
17
18
19       Dependent variable = COLUMN2
20       Independent variable =  COLUMN1
21
22     Total number of observations =  3
23     Total number of records with missing values = 0
24     Total number of parameters in model = 2
25     Total number of specified parameters = 0
26     Degree of polynomial = 1
27
28
29     Maximum number of iterations =  250
30     Relative Function Convergence has been set to: le-008
31     Parameter Convergence has been  set to: le-008
32
33
34
35                      Default  Initial Parameter Values
36                         Background  =    0.0617654
37                            Beta(l)  =     0.015138
38
39
40               Asymptotic Correlation Matrix of Parameter Estimates
41
42               ( *** The model parameter(s)  -Background
43                     have been estimated at a boundary point, or have been
44    specified by the user,
45                     and do not appear in the correlation matrix )
46
47                    Beta(l)
48
49       Beta(l)             1
50
51
52
53                              Parameter Estimates
54
55           Variable           Estimate             Std. Err.
56         Background                   0               NA
57            Beta(l)            0.0166997          0.00582148
58
59    NA - Indicates that this parameter has hit a bound
60         implied by some inequality  constraint and thus
61         has no standard error.
62
63
64
                                          A-12      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
IT-
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36


Model
Full model
Fitted model
Reduced model

AIC:




Dose Est


i : 1
0.0000 0.
i: 2
50.0000 0.
i: 3
150.0000 0.

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

Analysis of Deviance Table

Log (likelihood) Deviance Test DF
-9.98095
-10.0255 0.089063 2
-20.7944 21.6269 2

22.051


Goodness of Fit

. Prob . Expected Observed



0000 0.000 0

5661 5.661 6

9183 9.183 9

0.09 DF = 2 P-value =


Computation

0.1

Extra risk

0.95

6.30914

3.92229



P-value

0.9564
<.0001






Size Chi^2 Res.



10 0.000

10 0.138

10 -0.244

0.9553














                                 Multistage Model with 0.95 Confidence Level
            0.8
                    Multistage
                                   40
                                          60
                                                  80




                                               dose
                                                         100
                                                                120
                                                                        140
                                                                               160
37
          15:0507/142005
                                                       A-13
DRAFT: DO NOT CITE OR QUOTE

-------
A-14      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2    ====================================================================
 3            Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
 4            Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS\_CYANO
 5    TOX REV\HEINZE_LIVER_LESIONS.(d)
 6            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 7    DOCUMENTS\_CYANO TOX REV\HEINZE_LIVER_LESIONS.pit
 8                                              Thu Mar 09 11:49:47 2006
 9     ====================================================================
10
11     BMDS MODEL RUN
12    	
13
14       The form of the probability function is:
15
16       P[response]  = CumNorm(Intercept+Slope*Dose) ,
17
18       where CumNorm(.)  is the cumulative normal distribution function
19
20
21       Dependent variable = COLUMN2
22       Independent variable = COLUMN1
23       Slope parameter is not restricted
24
25       Total number of observations = 3
26       Total number of records with missing values = 0
27       Maximum number  of iterations = 250
28       Relative Function Convergence has been set to: le-008
29       Parameter Convergence has been set to: le-008
30
31
32
33                      Default Initial (and Specified) Parameter Values
34                         background =            0   Specified
35                          intercept =     -1.57069
36                              slope =    0.0201403
37
38
39               Asymptotic Correlation Matrix of Parameter Estimates
40
41               ( *** The model parameter(s)   -background
42                     have been estimated at a boundary point, or have been
43    specified by the user,
44                     and do not appear in the correlation matrix )
45
46                  intercept        slope
47
48     intercept            1        -0.75
49
50         slope        -0.75            1
51
52
53
54                              Parameter Estimates
55
56           Variable           Estimate             Std. Err.
57          intercept              -1.209            0.433188
58              slope           0.0190194          0.00553748
59
60
61
62                            Analysis of Deviance Table
63
64           Model      Log(likelihood)  Deviance  Test DF     P-value
                                          A-15      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
     Full  model
   Fitted  model
  Reduced  model

            AIC:
    -9.98095
    -12.2154        4.46887
    -20.7944        21.6269

    28.4308
                      Goodness  of   Fit
                                  0.03452
                                  <.0001
Dose
0.0000
50.0000
150.0000
Est ._Prob.
0.1133
0.3982
0.9499
Expected
1.133
3.982
9.499
Observed
0
6
9
Size
10
10
10
Scaled
Residual
-1.131
1.304
-0.7234
 Chi-square =
3.50
DF = 1
P-value  =  0.0613
   Benchmark Dose Computation

Specified  effect =             0.1

Risk Type         =       Extra risk

Confidence level =            0.95

              BMD =         19.6901

             BMDL =        12.3138
                            Probit Model with 0.95 Confidence Level
      S   0.6
          0.4
               Probit
                 BMDL
                       BMD
                            40     60     80    100    120    140

                                      dose
                                                                160
33
        11:4903/092006
                                             A-16
                                                   DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
 3            Input Data File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
 4            Gnuplot Plotting File:   C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
 5                                              Thu Jul 14 15:06:15 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = Background
14                   + (1-Background) * CumNorm(Intercept+Slope*Log(Dose)),
15
16       where CumNorm(.)  is the cumulative normal distribution function
17
18
19       Dependent variable = COLUMN2
20       Independent variable = COLUMN1
21       Slope parameter is restricted as slope >= 1
22
23       Total number of observations = 3
24       Total number of records with missing values = 0
25       Maximum number of iterations = 250
26       Relative Function Convergence has been set to: le-008
27       Parameter Convergence has been set to: le-008
28
29
30
31       User has chosen the log transformed model
32
33
34                      Default Initial (and Specified) Parameter Values
35                         background =            0
36                          intercept =     -3.68466
37                              slope =            1
38
39
40               Asymptotic Correlation Matrix of Parameter Estimates
41
42               ( *** The model parameter(s)   -background    -slope
43                     have been estimated at a boundary point, or have been
44    specified by the user,
45                     and do not appear in the correlation matrix )
46
47                  intercept
48
49     intercept            1
50
51
52
53                              Parameter Estimates
54
55           Variable           Estimate             Std. Err.
56         background                   0               NA
57          intercept            -3.68338            0.323658
58              slope                   1               NA
59
60    NA - Indicates that this parameter has hit a bound
61         implied by some inequality constraint and thus has no standard  error.
62
63
64
                                          A-17      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35


Model
Full model
Fitted model
Reduced model

AIC:





Dose Est
0.0000 0
50.0000 0
150.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


Analysis of Deviance Table

Log (likelihood) Deviance Test DF P-value
-9.98095
-9.98638 0.0108688 2 0.9946
-20.7944 21.6269 2 <.0001

21.9728


Goodness of Fit

Scaled
. Prob . Expected Observed Size Residual
.0000 0.000 0 10 0
.5904 5.904 6 10 0.06155
.9078 9.078 9 10 -0.08513

0.01 DF = 2 P-value = 0.9945


Computation

0.1

Extra risk

0.95

11.0433

6.37572


                                   Probit Model with 0.95 Confidence Level
            0.6
            0.4
            0.2
                  Probit
                   BMDL  BMD
                    0      20      40      60     80      100     120     140     160


                                               dose
36
37
          15:0607/142005
                                                       A-18
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Quantal Linear Model $Revision: 2.2 $ $Date: 2000/03/17  22:27:16  $
        Input Data File: C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
        Gnuplot Plotting File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
                                          Thu Jul  14 15:07:01  2005
 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(-slope*dose)]
   Dependent variable = COLUMN2
   Independent variable = COLUMN1

   Total number of observations = 3
   Total number of records with missing values = 0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =    0.0454545
                          Slope =    0.0129727
                          Power =            1   Specified
           Asymptotic Correlation Matrix of Parameter Estimates

           ( *** The model parameter(s)  -Background     -Power
                 have been estimated at a boundary point, or have been
specified by the user,
                 and do not appear in the correlation matrix )
     Slope
             Slope

                 1
       Variable
     Background
          Slope
                     Parameter Estimates

                     Estimate
                             0
                     0.0166997
                             Std. Err.
                                NA
                           0.00500498
NA -
Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
                   Analysis of Deviance Table

                              Deviance  Test DF
Log(likelihood)
     -9.98095
     -10.0255
     -20.7944

       22.051
P-value
                                0.089063
                                 21.6269
     0.9564
    <.0001
                                          A-19
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
/
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25





Dose Est


0.0000 0
50.0000 0
150.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


Goodness of Fit


. Prob . Expected Observed Size


.0000 0.000 0 10
.5661 5.661 6 10
.9183 9.183 9 10

0.09 DF = 2 P-value = 0.9553


Computation

0.1

Extra risk

0.95

6.30914

3.92229




Scaled
Residual


0
0.2162
-0.2115















26
27
                                Quantal Linear Model with 0.95 Confidence Level
            0.6
            0.4
            0.2
                     Quantal Linear
                                   40
                                           60
                                                  80

                                                dose
                                                          100
                                                                 120
                                                                         140
                                                                                 160
          15:07 07/142005
                                                        A-20
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Quantal Quadratic Model $Revision: 2.2 $ $Date: 2000/03/17  22:27:16  $
        Input Data File: C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
        Gnuplot Plotting File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
                                          Thu Jul  14  15:07:40  2005
 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(-slope*dose^2)]
   Dependent variable = COLUMN2
   Independent variable = COLUMN1

   Total number of observations = 3
   Total number of records with missing values =  0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =    0.0454545
                          Slope =  8.64849e-005
                          Power =            2   Specified
           Asymptotic Correlation Matrix of Parameter Estimates

           ( *** The model parameter(s)  -Background     -Power
                 have been estimated at a boundary point, or have  been
specified by the user,
                 and do not appear in the correlation matrix )
     Slope
             Slope

                 1
       Variable
     Background
          Slope
                     Parameter Estimates

                     Estimate
                             0
                   0.000171114
                             Std. Err.
                                NA
                         5.74454e-005
NA -
Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
                   Analysis of Deviance Table

                              Deviance  Test DF
Log(likelihood)
     -9.98095
     -12.0873
     -20.7944

      26 . 1745
P-value
                                 4.21262
                                 21.6269
     0.1217
    <.0001
                                          A-21
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
/
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25





Dose Est


0.0000 0
50.0000 0
150.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


Goodness of Fit


. Prob . Expected Observed Size


.0000 0.000 0 10
.3480 3.480 6 10
.9787 9.787 9 10

5.77 DF = 2 P-value = 0.0558


Computation

0.1

Extra risk

0.95

24.814

19.01




Scaled
Residual


0
1.673
-1 .725















26
27
                                Quantal Quadratic Model with 0.95 Confidence Level
            0.8





        1   0.6
        •S

        c
        .2

        1   0.4
        L_



            0.2
                      Quantal Quadratic
                        BMDL
                               BMD
                            20
                                    40
                                            60
                                                    80

                                                 dose
                                                           100
                                                                   120
                                                                           140
                                                                                  160
          15:07 07/142005
                                                        A-22
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Weibull Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16  $
        Input Data File: C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.(d)
        Gnuplot Plotting File:  C:\BMDS\HEINZE_MOD_AND_INT_WITH_HEMORRHAGE.plt
                                          Thu Jul  14  15:08:20  2005
 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(-slope*dose^power)]
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Power parameter is restricted as power >=1

   Total number of observations = 3
   Total number of records with missing values =  0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =    0.0454545
                          Slope =    0.0129727
                          Power =            1
           Asymptotic Correlation Matrix of Parameter Estimates

           ( *** The model parameter(s)  -Background     -Power
                 have been estimated at a boundary point, or have  been
specified by the user,
                 and do not appear in the correlation matrix )
     Slope
Slope

    1
       Variable
     Background
          Slope
          Power
        Parameter Estimates

        Estimate
                0
        0.0166997
                1
  Std.  Err.
     NA
0.00500498
     NA
NA - Indicates that this parameter has hit a bound
     implied by some inequality constraint and thus
     has no standard error.
       Model
     Full model
   Fitted model
      Analysis of Deviance Table

Log(likelihood)   Deviance  Test DF
     -9.98095
     -10.0255      0.089063      2
            P-value

                 0.9564
                                          A-23
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
_L\7
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Reduced model

AIC:





Dose Est

0.0000 0
50.0000 0
150.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


-20.7944 21.6269 2 <.

22.051


Goodness of Fit


._Prob. Expected Observed Size

.0000 0.000 0 10
.5661 5.661 6 10
.9183 9.183 9 10

0.09 DF = 2 P-value = 0.9553


Computation

0.1

Extra risk

0.95

6.30914

3.92229


.0001






Scaled
Residual

0
0.2162
-0.2115

















31
32
                                  Weibull Model with 0.95 Confidence Level
        3   0.6
            0.4
                  Weibull
                                   40      60      80      100     120     140


                                               dose
                                                                               160
          15:0807/142005
                                                       A-24
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
Fawell et al.,1999   Male Chronic Inflammation
         $Revision: 2.2 $ $Date: 2001/03/14 01:17:00  $
        Input Data File: C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:45:24  2005
 BMDS MODEL RUN
   The form of the probability function is:

   P[response]= background+(1-background)*CumGamma[slope*dose,power],
   where CumGamma(.)  is the cummulative Gamma distribution  function
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Power parameter is restricted as power >=1

   Total number of observations = 4
   Total number of records with missing values =  0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =      0.09375
                          Slope =   0.00193589
                          Power =          1.3
           Asymptotic Correlation Matrix of Parameter Estimates

             Background        Slope        Power

Background            1        0.037        0.041

     Slope        0.037            1             1

     Power        0.041            1             1
       Variable
     Background
          Slope
          Power
       Model
     Full model
   Fitted model
  Reduced model
        Parameter Estimates

        Estimate             Std. Err.
        0.0999826            0.054818
        0.0276638            0.488219
          8.09035             116.456
                        Analysis of Deviance Table
Log(likelihood)
     -18.2628
     -18.4512
     -39.4295
Deviance  Test DF
P-value
  0.376765
   42.3333
     0.5393
    <.0001
                                          A-25
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
«
o
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
AIC:





Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 1

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


42.9024


Goodness of Fit


. Prob . Expected Observed Size
.1000 1.500 1 15
.1000 1.500 2 15
.2667 4.001 4 15
.0000 15.000 15 15

0.37 DF = 1 P-value = 0.5429


Computation

0.1

Extra risk

0.95

170.825

53.1183







Scaled
Residual
-0.4301
0.4304
-0.0002949
0.00692

















30
31
                               Gamma Multi-Hit Model with 0.95 Confidence Level
        £   0.6
            0.4
            0.2
                    Gamma Multi-Hit
                   BMDL
                              BMD
                              200
                                          400         600


                                               dose
                                                                800
                                                                           1000
          15:45 12/282005
                                                       A-26
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
 3            Input Data File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:   C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
 5                                              Thu Mar 09 11:56:58 2006
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = I/[1+EXP(-intercept-slope*dose)]
14
15
16       Dependent variable = COLUMN2
17       Independent variable = COLUMN1
18       Slope parameter is not restricted
19
20       Total number of observations = 4
21       Total number of records with missing values = 0
22       Maximum number of iterations = 250
23       Relative Function Convergence has been set to:  le-008
24       Parameter Convergence has been set to: le-008
25
26
27
28                      Default Initial Parameter Values
29                         background =            0   Specified
30                          intercept =     -2.07771
31                              slope =   0.00552538
32
33
34               Asymptotic Correlation Matrix of Parameter Estimates
35
36               ( *** The model parameter(s)   -background
37                     have been estimated at a boundary point, or have been
38    specified by the user,
39                     and do not  appear in the correlation matrix )
40
41                  intercept        slope
42
43     intercept            1        -0.72
44
45         slope        -0.72            1
46
47
48
49                              Parameter Estimates
50
51           Variable           Estimate             Std. Err.
52          intercept            -2.49527            0.612166
53              slope          0.00805129          0.00337645
54
55
56
57                            Analysis of Deviance Table
58
59           Model      Log(likelihood)  Deviance  Test DF     P-value
60         Full model        -18.2628
61       Fitted model        -18.4278      0.329996      2          0.8479
62      Reduced model        -39.4295       42.3333      3         <.0001
63
64               AIC:         40.8556
                                          A-27      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
/
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Goodness of Fit
c
Dose Est . Prob . Expected Observed Size I

0.0000 0.0762 1.143 1 15
40.0000 0.1022 1.533 2 15
200.0000 0.2921 4.382 4 15
1000.0000 0.9962 14.942 15 15

Chi-square = 0.28 DF = 2 P-value = 0.8680


Benchmark Dose Computation

Specified effect = 0.1

Risk Type = Extra risk

Confidence level = 0.95

BMD = 111.719

BMDL = 70.686

Scaled
Residual

-0.139
0.3983
-0.217
0.2408
















28
29
                                  Logistic Model with 0.95 Confidence Level
            0.6
            0.4
                   Logistic
                    BMDL
                           BMD
                               200
                                          400         600


                                               dose
                                                                800
                                                                           1000
          11:5603/092006
                                                       A-28
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Logistic Model $Revision: 2.1 $ $Date: 2000/02/26  03:38:20  $
        Input Data File: C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:47:11  2005


 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]


   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Slope parameter is restricted as slope >= 1

   Total number of observations = 4
   Total number of records with missing values = 0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
   User has chosen the log transformed model
                  Default Initial Parameter Values
                     background =    0.0666667
                      intercept =     -10.0178
                          slope =      1.86367
           Asymptotic Correlation Matrix of Parameter Estimates

             background    intercept        slope

background            1       0.0022      -0.0023

 intercept       0.0022            1            -1

     slope      -0.0023           -1            1
       Variable
     background
      intercept
          slope
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
  Parameter Estimates

  Estimate
  0.0999997
   -65.9875
    12.1748
                             Std. Err.
                            0.0547725
                              5466.64
                              1031.77
Analysis of Deviance Table

           Deviance  Test DF
Log(likelihood)
     -18.2628
     -18.4512
     -39.4295

      42.9024
P-value
             0.376844
              42.3333
     0.5393
    <.0001
                                          A-29
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
/
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26





Dose Est


0.0000 0
40.0000 0
200.0000 0
1000.0000 1

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL


Goodness of Fit


. Prob . Expected Observed Size


.1000 1.500 1 15
.1000 1.500 2 15
.2667 4.000 4 15
.0000 15.000 15 15

0.37 DF = 1 P-value = 0.5428


Computation

0.1

Extra risk

0.95

188.582

94.487




Scaled
Residual


-0.4303
0.4303
-4.206e-005
0.0004285















                                 Log-Logistic Model with 0.95 Confidence Level
            0.6
            0.4
                     Log-Logistic
                      BMDL
                                BMD
                               200
                                           400         600

                                                dose
                                                                  800
                                                                             1000
27        15:4712/282005
28
                                                        A-30
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21  03:38:21  $
        Input Data File: C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
                                          Thu Mar 09  11:59:44  2006
 BMDS MODEL RUN
   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(
-betal*dose^l-beta2*dose^2-beta3*dose^3)]

   The parameter betas are restricted to be positive
   Dependent variable = COLUMN2
   Independent variable = COLUMN1

 Total number of observations = 4
 Total number of records with missing values = 0
 Total number of parameters in model = 4
 Total number of specified parameters = 0
 Degree of polynomial = 3
 Maximum number of iterations = 250
 Relative Function Convergence has been set to:  le-008
 Parameter Convergence has been set to: le-008
                  Default Initial Parameter Values
                     Background =            0
                        Beta(l) =            0
                        Beta(2) =            0
                        Beta(3) = 1.00264e+011
           Asymptotic Correlation Matrix of Parameter Estimates

           ( *** The model parameter(s)  -Beta(2)
                 have been estimated at a boundary point, or have  been
specified by the user,
                 and do not appear in the correlation matrix )

             Background

Background            1

   Beta(l)        -0.65

   Beta(3)         0.14
       Variable
     Background
        Beta(l)
        Beta(2)
        Beta(3)
Beta(l)
-0.65
1
-0.42
Parameter
Estimate
0.0782274
0.000933123
0
7.10432e-009
Beta(3)
0.14
-0.42
1
Estimates
Std. Err
0.200883
0.00219636
NA
1.57018e-008
                                          A-31
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.


Model
Full model
Fitted model
Reduced model

AIC:




Dose Est
i: 1
0.0000 0.
i: 2
40.0000 0.
i: 3
200.0000 0.
i: 4
1000.0000 0.

Chi -square =

Benchmark Dose
Specified effect
Risk Type
Confidence level

BMD

BMDL

Analysis of Deviance Table
Log (likelihood) Deviance Test DF
-18.2628
-18.3175 0.10941 1
-39.4295 42.3333 3

42.635


Goodness of Fit

._Prob. Expected Observed

0782 1.173 1

1124 1.686 2

2774 4.161 4

9997 14.996 15

0.11 DF = 1 P-value =

Computation
0.1
Extra risk
0.95

104.279

37.4815


P-value
0.7408
<.0001






Size Chi^2 Res.

15 -0.160

15 0.210

15 -0.054

15 1.000

0.7438









A-32
DRAFT: DO NOT CITE OR QUOTE

-------
                         Multistage Model with 0.95 Confidence Level
o   0.6
    0.4
            Multistage
          BMDL    3MD
                       200
                                  400         600




                                       dose
                                                         800
                                                                    1000
  11:5903/092006
                                               A-33         DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Multistage Model.  $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
 3            Input Data File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
 5                                              Thu Mar 09 12:00:48 2006
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(
14    -betal*dose^l-beta2*dose^2) ]
15
16       The parameter betas are restricted to be positive
17
18
19       Dependent variable = COLUMN2
20       Independent variable =  COLUMN1
21
22     Total number of observations =  4
23     Total number of records with missing values = 0
24     Total number of parameters in model = 3
25     Total number of specified parameters = 0
26     Degree of polynomial = 2
27
28
29     Maximum number of iterations =  250
30     Relative Function Convergence has been set to: le-008
31     Parameter Convergence has been  set to: le-008
32
33
34
35                      Default  Initial Parameter Values
36                         Background  =            0
37                            Beta(l)  =            0
38                            Beta(2)  = 1.01264e+014
39
40
41               Asymptotic Correlation Matrix of Parameter Estimates
42
43                 Background      Beta(l)      Beta (2)
44
45    Background            1        -0.65         0.33
46
47       Beta(l)         -0.65            1        -0.75
48
49       Beta(2)          0.33        -0.75            1
50
51
52
53                              Parameter Estimates
54
55           Variable           Estimate             Std. Err.
56         Background           0.0889744            0.205309
57            Beta(l)         8.80277e-005          0.00292558
58            Beta(2)         5.76916e-006        6.80122e-006
59
60
61
62                            Analysis of Deviance Table
63
64           Model      Log(likelihood)   Deviance  Test DF     P-value
                                          A-34      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
1 1
_L J.
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Full model
Fitted model
Reduced model

AIC:




Dose Est

i: 1
0.0000 0.
i: 2
40.0000 0.
i: 3
200.0000 0.
i: 4
1000.0000 0.

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

-18.2628
-18.4529 0.380206 1
-39.4295 42.3333 3

42.9058


Goodness of Fit

._Prob. Expected Observed Size


0890 1.335 1 15

1005 1.508 2 15

2893 4.340 4 15

9974 14.961 15 15

0.35 DF = 1 P-value = 0.5556


Computation

0.1

Extra risk

0.95

127.726

39.8667


0.5375
<.0001






Chi^2 Res.


-0.275

0.363

-0.110

1.003
















                                  Multistage Model with 0.95 Confidence Level
        o   0.6
            0.4
                    Multistage
                   BMDL
                            BMD
                               200
                                           400
                                                      600
                                                                 800
                                                                            1000
36
                                                dose
          12:0003/092006
                                                        A-35
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21  03:38:21  $
        Input Data File: C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28 15:47:43 2005
 BMDS MODEL RUN
   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(
-betal*dose^l) ]

   The parameter betas are restricted to be positive
   Dependent variable = COLUMN2
   Independent variable = COLUMN1

 Total number of observations = 4
 Total number of records with missing values = 0
 Total number of parameters in model = 2
 Total number of specified parameters = 0
 Degree of polynomial = 1
 Maximum number of iterations = 250
 Relative Function Convergence has been set to: le-008
 Parameter Convergence has been set to: le-008
                  Default Initial Parameter Values
                     Background =            0
                        Beta(l) = 1.04991e+017
           Asymptotic Correlation Matrix of Parameter Estimates

             Background      Beta(l)

Background            1        -0.37

   Beta(l)        -0.37             1
                          Parameter Estimates
       Variable
     Background
        Beta(l)
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
  Estimate
   0.048937
  0.0025684
                             Std. Err.
                              0.16314
                          0.000904753
Analysis of Deviance Table

           Deviance  Test DF
Log(likelihood)
     -18.2628
     -20.3189
     -39.4295

      44.6378
P-value
              4.11219
              42.3333
      0.128
    <.0001
                                          A-36
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Dose Est
i: 1
0.0000 0.
i: 2
40.0000 0.
i: 3
200.0000 0.
i: 4
1000.0000 0.

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL
Goodness of Fit
._Prob. Expected

0489 0.734

1418 2.127

4310 6.465

9271 13.906

2 . 94 DF = 2


Computation

0.1

Extra risk

0.95

41.0219

25.943
Observed Size Chi^2 Res.

1 15 0.381

2 15 -0.070

4 15 -0.670

15 15 1.079

P-value = 0.2298













                                 Multistage Model with 0.95 Confidence Level
        o   0.6
            0.4
                    Multistage
                               200
                                          400
                                                     600
                                                                800
                                                                           1000
30
31
                                               dose
          15:47 12/282005
                                                       A-37
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ===================================================================
 2            Probit Model $Revision:  2.1 $ $Date: 2000/02/26 03:38:53 $
 3            Input Data File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:   C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
 5                                              Thu Mar 09 12:02:26 2006
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P [response]  = CumNorm(Intercept+Slope*Dose) ,
14
15       where CumNorm(.)  is the cumulative normal distribution function
16
17
18       Dependent variable = COLUMN2
19       Independent variable = COLUMN1
20       Slope parameter is not restricted
21
22       Total number of observations  = 4
23       Total number of records with  missing values = 0
24       Maximum number  of iterations  = 250
25       Relative Function Convergence has been set to: le-008
26       Parameter Convergence has been set to: le-008
27
28
29
30                      Default Initial (and Specified) Parameter Values
31                         background  =            0   Specified
32                          intercept  =     -1.32259
33                              slope  =   0.00345221
34
35
36               Asymptotic Correlation Matrix of Parameter Estimates
37
38               ( *** The model parameter(s)   -background
39                     have been estimated at a boundary point, or have been
40    specified by the user,
41                     and do not  appear in the correlation matrix )
42
43                  intercept        slope
44
45     intercept            1        -0.65
46
47         slope        -0.65            1
48
49
50
51                              Parameter Estimates
52
53           Variable           Estimate             Std. Err.
54          intercept            -1.42508            0.306506
55              slope          0.00435347          0.00171942
56
57
58
59                            Analysis of Deviance Table
60
61           Model      Log(likelihood)  Deviance  Test DF     P-value
62         Full model        -18.2628
63       Fitted model        -18.3773       0.229031      2          0.8918
64      Reduced model        -39.4295        42.3333      3         <.0001
                                          A-38      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

AIC:





Dose Est

0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL



40.7546


Goodness of Fit


. Prob . Expected Observed Size

.0771 1.156 1 15
.1055 1.582 2 15
.2897 4.345 4 15
.9983 14.974 15 15

0.21 DF = 2 P-value = 0.9002


Computation

0.1

Extra risk

0.95

107.589

66.4468








Scaled
Residual

-0.151
0.3512
-0.1963
0.16

















                                   Probit Model with 0.95 Confidence Level
        £   0.6
            0.4
            0.2
                  Probit
                    BMDL  BMD
                              200
                                         400
                                                     600
                                                                800
                                                                           1000
31
32
                                               dose
          12:02 03/092006
                                                       A-39
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
 3            Input Data File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
 5                                              Wed Dec 28 15:48:13 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = Background
14                   + (1-Background) * CumNorm(Intercept+Slope*Log(Dose)),
15
16       where CumNorm(.)  is the cumulative normal distribution function
17
18
19       Dependent variable = COLUMN2
20       Independent variable = COLUMN1
21       Slope parameter is restricted as slope >= 1
22
23       Total number of observations = 4
24       Total number of records with missing values = 0
25       Maximum number of iterations = 250
26       Relative Function Convergence has been set to: le-008
27       Parameter Convergence has been set to: le-008
28
29
30
31       User has chosen the log transformed model
32
33
34                      Default Initial (and Specified) Parameter Values
35                         background =    0.0666667
36                          intercept =     -5.60926
37                              slope =      1.03389
38
39
40               Asymptotic Correlation Matrix of Parameter Estimates
41
42                 background    intercept        slope
43
44    background            1       0.0018       -0.002
45
46     intercept       0.0018            1           -1
47
48         slope       -0.002           -1            1
49
50
51
52                              Parameter Estimates
53
54           Variable           Estimate             Std. Err.
55         background                 0.1           0.0547723
56          intercept            -22.1534             1087.37
57              slope             4.01215             205.229
58
59
60
61                            Analysis of Deviance Table
62
63           Model      Log(likelihood)  Deviance  Test DF     P-value
64         Full model        -18.2628
                                          A-40      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
9
10
1 1
_L J.
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Fitted model
Reduced model

AIC:




Dose Est

0.0000 0
40.0000 0
200.0000 0
1000.0000 1

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL
-18.4512 0.376844 1
-39.4295 42.3333 3

42.9024


Goodness of Fit

._Prob. Expected Observed

.1000 1.500 1
.1000 1.500 2
.2667 4.000 4
.0000 15.000 15

0.37 DF = 1 P-value =


Computation

0.1

Extra risk

0.95

181.665

95.2612
0.5393
<.0001





Scaled
Size Residual

15 -0.4303
15 0.4303
15 -1.174e-005
15 0.0004249

0.5428













                                   Probit Model with 0.95 Confidence Level
31
32
            0.8
        B   0.6


        I
        c
        .2


        S   0.4
            0.2
                   Probit
                      BMDL
          15:48 12/282005
                               BMD
                               200
                                           400         600



                                                dose
                                                                 800
                                                                             1000
                                                       A-41
DRAFT: DO NOT CITE OR QUOTE

-------
 1     ====================================================================
 2            Quantal Linear Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
 3            Input Data File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:   C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
 5                                              Wed Dec 28 15:48:41 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(-slope*dose)]
14
15
16       Dependent variable = COLUMN2
17       Independent variable = COLUMN1
18
19       Total number of observations  = 4
20       Total number of records with  missing values = 0
21       Maximum number of iterations  = 250
22       Relative Function Convergence has been set to: le-008
23       Parameter Convergence has been set to: le-008
24
25
26
27                      Default Initial (and Specified) Parameter Values
28                         Background  =      0.09375
29                              Slope  =    0.0033673
30                              Power  =            1   Specified
31
32
33               Asymptotic Correlation Matrix of Parameter Estimates
34
35               ( *** The model parameter(s)  -Power
36                     have been estimated at a boundary point, or have been
37    specified by the user,
38                     and do not appear in the correlation matrix )
39
40                 Background        Slope
41
42    Background            1        -0.23
43
44         Slope        -0.23            1
45
46
47
48                              Parameter Estimates
49
50           Variable           Estimate             Std. Err.
51         Background            0.048937           0.0439067
52              Slope           0.0025684         0.000746108
53
54
55
56                            Analysis of Deviance Table
57
58           Model      Log(likelihood)   Deviance  Test DF     P-value
59         Full model        -18.2628
60       Fitted model        -20.3189        4.11219      2           0.128
61      Reduced model        -39.4295        42.3333      3         <.0001
62
63               AIC:         44.6378
64
                                          A-42      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

Goodness of Fit
._Prob. Expected
.0489 0.734
.1418 2.127
.4310 6.465
.9271 13.906

2.94 DF = 2


Computation

0.1

Extra risk

0.95

41.0219

25.943

Scaled
Observed Size Residual
1 15 0.3183
2 15 -0.09393
4 15 -1.285
15 15 1.086

P-value = 0.2298














                                Quantal Linear Model with 0.95 Confidence Level
            0.6
            0.4
                     Quantal Linear
                  BMDL BMD
                               200
                                          400
                                                      600
                                                                 800
                                                                            1000
27
28
                                                dose
          15:48 12/282005
                                                        A-43
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Quantal Quadratic Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
 3            Input Data File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
 5                                              Wed Dec 28 15:49:14 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(-slope*dose^2)]
14
15
16       Dependent variable = COLUMN2
17       Independent variable = COLUMN1
18
19       Total number of observations = 4
20       Total number of records with missing values = 0
21       Maximum number of iterations = 250
22       Relative Function Convergence has been set to: le-008
23       Parameter Convergence has been set to: le-008
24
25
26
27                      Default Initial (and Specified) Parameter Values
28                         Background =      0.09375
29                              Slope =  3.3673e-006
30                              Power =            2   Specified
31
32
33               Asymptotic Correlation Matrix of Parameter Estimates
34
35               ( *** The model parameter(s)   -Power
36                     have been estimated at a boundary point, or have  been
37    specified by the user,
38                     and do not appear in the correlation matrix )
39
40                 Background        Slope
41
42    Background            1        -0.32
43
44         Slope        -0.32            1
45
46
47
48                              Parameter Estimates
49
50           Variable           Estimate             Std. Err.
51         Background            0.091177           0.0532315
52              Slope        6.03482e-006        3.53595e-006
53
54
55
56                            Analysis of Deviance Table
57
58           Model      Log(likelihood)   Deviance  Test DF     P-value
59         Full model        -18.2628
60       Fitted model        -18.4542      0.382862      2          0.8258
61      Reduced model        -39.4295       42.3333      3         <.0001
62
63               AIC:         40.9085
64
                                          A-44      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Goodness of Fit
Dose Est._Prob. Expected
0.0000 0.0912 1.368
40.0000 0.0999 1.499
200.0000 0.2861 4.291
1000.0000 0.9978 14.967

Chi-square = 0.36 DF = 2


Benchmark Dose Computation

Specified effect = 0.1

Risk Type = Extra risk

Confidence level = 0.95

BMD = 132.132

BMDL = 85.4708

Scaled
Observed Size Residual
1 15 -0.3298
2 15 0.4317
4 15 -0.1665
15 15 0.1808

P-value = 0.8372














                               Quantal Quadratic Model with 0.95 Confidence Level
            0.6
            0.4
                      Quantal Quadratic
                      BMDL   BMD
                               200
                                           400
                                                      600
                                                                  800
                                                                             1000
27
28
                                                dose
          15:49 12/282005
                                                        A-45
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Weibull Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16  $
        Input Data File: C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_MALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:49:46  2005
 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(-slope*dose^power)]
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Power parameter is restricted as power >=1

   Total number of observations = 4
   Total number of records with missing values = 0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =      0.09375
                          Slope = 4.78181e-007
                          Power =      2.28256
           Asymptotic Correlation Matrix of Parameter Estimates

             Background        Slope        Power

Background            1        -0.33         0.31

     Slope        -0.33            1            -1

     Power         0.31           -1            1
                          Parameter Estimates
       Variable
     Background
          Slope
          Power
  Estimate
  0.0959027
.27341e-006
    2.27222
                             Std. Err.
                            0.0577365
                         1.17833e-005
                              1.67024
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
Analysis of Deviance Table

           Deviance  Test DF
Log(likelihood)
     -18.2628
     -18.4284
     -39.4295

      42.8569
P-value
             0.331283
              42.3333
     0.5649
    <.0001
                                          A-46
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
,J
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

Goodness of Fit
._Prob. Expected
.0959 1.439
.1009 1.514
.2712 4.067
.9998 14.997

0.33 DF = 1


Computation

0.1

Extra risk

0.95

145.973

48.796

Scaled
Observed Size Residual
1 15 -0.3845
2 15 0.4168
4 15 -0.03917
15 15 0.05665

P-value = 0.5678














                                  Weibull Model with 0.95 Confidence Level
        £   0.6
            0.4
            0.2
                  Weibull
                   BMDL
                             BMD
                              200
                                          400
                                                     600
                                                                800
                                                                           1000
26
27
                                               dose
          15:49 12/282005
                                                       A-47
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
Fawell et al., 1999 Female Chronic Inflammation
         $Revision: 2.2 $ $Date: 2001/03/14 01:17:00  $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:50:58  2005
 BMDS MODEL RUN
   The form of the probability function is:

   P[response]= background+(1-background)*CumGamma[slope*dose,power],
   where CumGamma(.)  is the cummulative Gamma distribution  function
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Power parameter is restricted as power >=1

   Total number of observations = 4
   Total number of records with missing values =  0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =      0.34375
                          Slope =   0.00408479
                          Power =      2.14573
           Asymptotic Correlation Matrix of Parameter Estimates

           ( *** The model parameter(s)  -Power
                 have been estimated at a boundary point, or have  been
specified by the user,
                 and do not appear in the correlation matrix )
             Background

Background            1

     Slope        -0.42
                          Slope

                          -0.42

                              1



                     Parameter Estimates
       Variable
     Background
          Slope
          Power
                     Estimate
                      0.385764
                     0.0020737
                             1
   Std.  Err.
  0.0913888
0.000852726
      NA
NA -
Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
                                          A-48
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35


Model
Full model
Fitted model
Reduced model

AIC:





Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

Analysis of Deviance Table

Log (likelihood) Deviance Test DF
-33.9494
-34.4584 1.01801 2
-40.7516 13.6045 3

72.9167


Goodness of Fit


. Prob . Expected Observed
.3858 5.786 5
.4347 6.520 8
.5943 8.914 8
.9228 13.842 14

1.02 DF = 2 P-value =


Computation

0.1

Extra risk

0.95

50.8081

27.7368



P-value

0.6011
0.003496






Scaled
Size Residual
15 -0.4172
15 0.7709
15 -0.4808
15 0.1531

0.5996














                               Gamma Multi-Hit Model with 0.95 Confidence Level
            0.6
            0.4
            0.2
                    Gamma Multi-Hit
                              200
                                          400
                                                     600
                                                                800
                                                                           1000
36
                                               dose
          15:50 12/282005
                                                        A-49
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Thu Mar 09 12:16:37 2006
 BMDS MODEL RUN


   The form of the probability function is:

   P[response]  = I/[1+EXP(-intercept-slope*dose)]
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Slope parameter is not restricted

   Total number of observations = 4
   Total number of records with missing values = 0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial Parameter Values
                     background =            0   Specified
                      intercept =    -0.342416
                          slope =   0.00261455
           Asymptotic Correlation Matrix of Parameter Estimates

           (  *** The model parameter(s)  -background
                 have been estimated at a boundary point, or have been
specified by the user,
                 and do not appear in the correlation matrix )
              intercept

 intercept            1

     slope        -0.51
       Variable
      intercept
          slope
       slope

       -0.51

           1



  Parameter Estimates

  Estimate
  -0.374017
 0.00300128
                             Std. Err.
                             0.335179
                           0.00110665
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
Analysis of Deviance Table

           Deviance  Test DF
Log(likelihood)
     -33.9494
     -34.4216
     -40.7516

      72.8432
P-value
             0.944503
              13.6045
     0.6236
   0.003496
                                          A-50
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
/
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Goodness of Fit
Dose Est . Prob . Expected

0.0000 0.4076 6.114
40.0000 0.4368 6.553
200.0000 0.5563 8.345
1000.0000 0.9326 13.989

Chi-square = 0.94 DF = 2


Benchmark Dose Computation

Specified effect = 0.1

Risk Type = Extra risk

Confidence level = 0.95

BMD = 80.3245

BMDL = 48.7137


Scaled
Observed Size Residual

5 15 -0.5851
8 15 0.7534
8 15 -0.1792
14 15 0.0114

P-value = 0.6243















                                  Logistic Model with 0.95 Confidence Level
29
30
31
            0.6
            0.4
                   Logistic
                   BMDL  BMD
                               200
                                          400
                                                     600
                                                                800
                                                                           1000
                                               dose
          12:1603/092006
                                                       A-51
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Logistic Model $Revision: 2.1 $ $Date: 2000/02/26  03:38:20  $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:51:45  2005


 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]


   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Slope parameter is restricted as slope >= 1

   Total number of observations = 4
   Total number of records with missing values = 0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
   User has chosen the log transformed model
                  Default Initial Parameter Values
                     background =     0.333333
                      intercept =     -5.18041
                          slope =            1
           Asymptotic Correlation Matrix of Parameter Estimates

             background    intercept        slope

background            1        -0.51         0.47

 intercept        -0.51            1        -0.99

     slope         0.47        -0.99            1
       Variable
     background
      intercept
          slope
       Model
     Full model
   Fitted model
  Reduced model
        Parameter Estimates

        Estimate
         0.426119
         -12.3243
          2.07233
            Std. Err.
           0.0985863
             8.90473
             1.36766
                        Analysis of Deviance Table
Log(likelihood)
     -33.9494
     -34.5387
     -40.7516
Deviance  Test DF
P-value
   1.17872
   13.6045
     0.2776
   0.003496
                                          A-52
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
AIC:





Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL
75.0774


Goodness of Fit


. Prob . Expected Observed Size
.4261 6.392 5 15
.4314 6.471 8 15
.5448 8.172 8 15
.9310 13.966 14 15

1.17 DF = 1 P-value = 0.2789


Computation

0.1

Extra risk

0.95

132.532

11.3311





Scaled
Residual
-0.7267
0 .7971
-0.08904
0.03513















                                 Log-Logistic Model with 0.95 Confidence Level
            0.6
            0.4
                     Log-Logistic
                 BMDL
                             BMD
                               200
                                           400
                                                      600
                                                                 800
                                                                             1000
28
29
          15:51 12/282005
                                                dose
                                                        A-53
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21  03:38:21  $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Thu Mar 09  12:17:59  2006
 BMDS MODEL RUN
   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(
-betal*dose^l-beta2*dose^2-beta3*dose^3)]

   The parameter betas are restricted to be positive
   Dependent variable = COLUMN2
   Independent variable = COLUMN1

 Total number of observations = 4
 Total number of records with missing values =  0
 Total number of parameters in model = 4
 Total number of specified parameters = 0
 Degree of polynomial = 3
 Maximum number of iterations = 250
 Relative Function Convergence has been set to:  le-008
 Parameter Convergence has been set to: le-008
                  Default Initial Parameter Values
                     Background =     0.417955
                        Beta(l) =   0.00124572
                        Beta(2) =            0
                        Beta(3) =  9.2085e-010
           Asymptotic Correlation Matrix of Parameter Estimates

           ( *** The model parameter(s)  -Beta(2)
                 have been estimated at a boundary point, or  have  been
specified by the user,
                 and do not appear in the correlation matrix  )

             Background

Background            1

   Beta(l)        -0.66

   Beta(3)         0.55
       Variable
     Background
        Beta(l)
        Beta(2)
        Beta(3)
Beta(l)
-0.66
1
-0.93
Parameter
Estimate
0.403311
0.00148059
0
7 . 06214e-010
Beta(3)
0.55
-0.93
1
Estimates
Std. Err
0.162784
0.00267062
NA
2.70652e-009
                                          A-54
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.



Model
Full model
Fitted model
Reduced model

AIC:




Dose Est
i: 1
0.0000 0.
i: 2
40.0000 0.
i: 3
200.0000 0.
i: 4
1000.0000 0.

Chi -square =

Benchmark Dose
Specified effect
Risk Type
Confidence level

BMD

BMDL


Analysis of Deviance Table
Log (likelihood) Deviance Test DF
-33.9494
-34.4013 0.903921 1
-40.7516 13.6045 3

74.8026


Goodness of Fit

. Prob . Expected Observed

4033 6.050 5

4376 6.565 8

5587 8.381 8

9330 13.995 14

0.90 DF = 1 P-value =

Computation
0.1
Extra risk
0.95

70.9904

28.0638



P-value
0.3417
0.003496






Size Chi^2 Res.

15 -0.291

15 0.389

15 -0.103

15 0.005

0.3421









A-55
DRAFT: DO NOT CITE OR QUOTE

-------
                      Multistage Model with 0.95 Confidence Level
1
0.9
0.8
T3 0.7
"o
< 0.6
£=
g
"o
<5 0.5
0.4
0.3
0.2

MultistE

Q6 —
^-^^^ :
,^~" \












BMDL








/^

^^ J
/^" \
^ :
j
.
H
BMD :
0 200 400 600 800 1000
dose
12:17 03/092006
                                          A-56
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21  03:38:21  $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:52:16  2005
 BMDS MODEL RUN
   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(
-betal*dose^l-beta2*dose^2) ]

   The parameter betas are restricted to be positive
   Dependent variable = COLUMN2
   Independent variable = COLUMN1

 Total number of observations = 4
 Total number of records with missing values = 0
 Total number of parameters in model = 3
 Total number of specified parameters = 0
 Degree of polynomial = 2
 Maximum number of iterations = 250
 Relative Function Convergence has been set to:  le-008
 Parameter Convergence has been set to: le-008
                  Default Initial Parameter Values
                     Background =     0.417495
                        Beta(l) =   0.00110779
                        Beta(2) = 1.05844e-006
           Asymptotic Correlation Matrix of Parameter Estimates

             Background      Beta(l)      Beta(2)

Background            1        -0.67         0.58

   Beta(l)        -0.67             1        -0.95

   Beta(2)         0.58        -0.95             1
       Variable
     Background
        Beta(l)
        Beta(2)
        Parameter Estimates

        Estimate
         0.402286
       0.00142259
    Std. Err.
    0.165802
  0.00323206
     7.46169e-007
3.28803e-006
       Model
      Analysis of Deviance Table

Log(likelihood)  Deviance  Test DF
              P-value
                                          A-57
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
8
9
10
1 1
_L J.
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Full model
Fitted model
Reduced model

AIC:




Dose Est

i: 1
0.0000 0.
i: 2
40.0000 0.
i: 3
200.0000 0.
i: 4
1000.0000 0.

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

-33.9494
-34.4149 0.93098
-40.7516 13.6045

74 .8297


Goodness of Fit

._Prob. Expected Observed


4023 6.034 5

4360 6.540 8

5635 8.453 8

9317 13.975 14

0.93 DF = 1 P-value


Computation

0.1

Extra risk

0.95

71.3892

27.9852


1 0.3346
3 0.003496






Size Chi^2 Res.


15 -0.287

15 0.396

15 -0.123

15 0.026

= 0.3347














                               Multistage Model with 0.95 Confidence Level
1
0.9
0.8
T3 0.7
S
0
< 0.6
£=
g
~o
<5 0.5
0.4
0.3
0.2

Multista

9e
H
^— ""^
^^--"~"~^ -









(





/


BMDL





/








^^-^ :
//^ :
,/ :
// J



-
H
H
H
BMD :
0 200 400 600 800 1000
dose
36
37
         15:52 12/282005
                                                   A-58
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Multistage Model.  $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
 3            Input Data File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:   C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
 5                                              Thu Mar 09 12:20:10 2006
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(
14    -betal*dose^l) ]
15
16       The parameter betas are restricted to be positive
17
18
19       Dependent variable = COLUMN2
20       Independent variable =  COLUMN1
21
22     Total number of observations =  4
23     Total number of records with missing values = 0
24     Total number of parameters in model = 2
25     Total number of specified parameters = 0
26     Degree of polynomial = 1
27
28
29     Maximum number of iterations =  250
30     Relative Function Convergence has been set to: le-008
31     Parameter Convergence has been  set to: le-008
32
33
34
35                      Default  Initial Parameter Values
36                         Background  =     0.377465
37                            Beta(l)  =   0.00221127
38
39
40               Asymptotic Correlation Matrix of Parameter Estimates
41
42                 Background      Beta(l)
43
44    Background            1        -0.45
45
46       Beta(l)         -0.45            1
47
48
49
50                              Parameter Estimates
51
52           Variable           Estimate             Std. Err.
53         Background            0.385756              0.1375
54            Beta(l)           0.00207378         0.000944904
55
56
57
58                            Analysis of Deviance Table
59
60           Model      Log(likelihood)   Deviance  Test DF     P-value
61         Full model        -33.9494
62       Fitted model        -34.4584        1.01801      2          0.6011
63      Reduced model        -40.7516        13.6045      3        0.003496
64
                                          A-59      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
J
4
5
6
7
/
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
AIC:
Dose Est

i: 1
0.0000 0.
i: 2
40.0000 0.
i: 3
200.0000 0.
i: 4
1000.0000 0.

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

72.9167
Goodness of Fit
. Prob . Expected Observed Size


3858 5.786 5 15

4347 6.520 8 15

5943 8.914 8 15

9228 13.842 14 15

1.02 DF = 2 P-value = 0.5996


Computation

0.1

Extra risk

0.95

50.8061

27.7368

Chi^2 Res.


-0.221

0.402

-0.253

0.148
















                                Multistage Model with 0.95 Confidence Level


-a
~o
£
<
£=
g
"o
TO



1
0.9
0.8
0.7
0.6

0.5
0.4
0.3
0.2

Mult

SI8Q6 —
^ — — j
^^~ \
/^ -








/'


BMDL
0


/

x-/ ;

j
^
'-_
BMD :
200 400 600 800 1000
                                            dose
32
         12:2003/092006
                                                    A-60
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Probit Model $Revision:  2.1 $ $Date: 2000/02/26 03:38:53 $
 3            Input Data File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:   C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
 5                                              Thu Mar 09 12:21:25 2006
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P [response]  = CumNorm(Intercept+Slope*Dose) ,
14
15       where CumNorm(.)  is the cumulative normal distribution function
16
17
18       Dependent variable = COLUMN2
19       Independent variable = COLUMN1
20       Slope parameter is not restricted
21
22       Total number of observations  = 4
23       Total number of records with  missing values = 0
24       Maximum number  of iterations  = 250
25       Relative Function Convergence has been set to: le-008
26       Parameter Convergence has been set to: le-008
27
28
29
30                      Default Initial (and Specified) Parameter Values
31                         background  =            0   Specified
32                          intercept  =    -0.216871
33                              slope  =   0.00162326
34
35
36               Asymptotic Correlation Matrix of Parameter Estimates
37
38               ( *** The model parameter(s)   -background
39                     have been estimated at a boundary point, or have been
40    specified by the user,
41                     and do not  appear in the correlation matrix )
42
43                  intercept        slope
44
45     intercept            1        -0.53
46
47         slope        -0.53            1
48
49
50
51                              Parameter Estimates
52
53           Variable           Estimate             Std. Err.
54          intercept           -0.223386             0.20693
55              slope          0.00172875         0.000560405
56
57
58
59                            Analysis of Deviance Table
60
61           Model      Log(likelihood)  Deviance  Test DF     P-value
62         Full model        -33.9494
63       Fitted model         -34.421       0.943277      2           0.624
64      Reduced model        -40.7516        13.6045      3        0.003496
                                          A-61      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
           AIC:          72.842
                     Goodness  of  Fit
Dose
0.0000
40.0000
200.0000
1000.0000
Est. Prob.
0.4116
0.4387
0.5487
0.9339
Expected
6.174
6.581
8.230
14.008
Observed
5
8
8
14
Size
15
15
15
15
Scaled
Residual
-0.6161
0.7385
-0.1196
-0.00859
 Chi-square =
0.94
DF = 2
P-value = 0.6252
   Benchmark Dose Computation

Specified effect =            0.1

Risk Type        =      Extra risk

Confidence level =           0.95

             BMD =         86.341

            BMDL =       56.9234
                                          A-62
                                                DRAFT: DO NOT CITE OR QUOTE

-------
e   0.6
    0.4
    0.2
           Probit
            BMDL  3MD
                           Probit Model with 0.95 Confidence Level
                       200
                                  400
                                             600
                                                        800
                                                                   1000
                                       dose
  12:21 03/092006
                                               A-63         DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Probit Model $Revision: 2.1 $ $Date: 2000/02/26  03:38:53  $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:52:51  2005
 BMDS MODEL RUN
   The form of the probability function is:

   P[response] = Background
               + (1-Background) * CumNorm(Intercept+Slope*Log(Dose)

   where CumNorm(.) is the cumulative normal distribution  function
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Slope parameter is restricted as slope >= 1

   Total number of observations = 4
   Total number of records with missing values =  0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
   User has chosen the log transformed model
                  Default Initial  (and Specified) Parameter Values
                     background =     0.333333
                      intercept =     -5.62294
                          slope =            1
           Asymptotic Correlation Matrix of Parameter Estimates

             background    intercept        slope

background            1        -0.46         0.43

 intercept        -0.46            1        -0.99

     slope         0.43        -0.99             1
       Variable
     background
      intercept
          slope
        Parameter Estimates

        Estimate
         0.431473
           -7.732
           1.2906
 Std.  Err.
0.0937984
  4.97551
  0.75516
       Model
     Full model
      Analysis of Deviance Table

Log(likelihood)  Deviance  Test DF
     -33.9494
           P-value
                                          A-64
                                                DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
   Fitted model
  Reduced model

            AIC:
   -34.5602        1.22166
   -40.7516        13.6045

    75.1204
                       Goodness  of   Fit
                                     0.269
                                 0.003496
Dose
0.0000
40.0000
200.0000
1000.0000
Est ._Prob.
0.4315
0.4323
0.5370
0.9327
Expected 01
6.472
6.485
8.055
13.991
^served £
5
8
8
14
3ize
15
15
15
15
Scaled
Residual
-0.7674
0.7897
-0.02871
0.009746
 Chi-square =
1.21
DF =  1
P-value  =  0.2706
   Benchmark Dose Computation

Specified effect =             0.1

Risk Type        =       Extra risk

Confidence level =            0.95

              BMD =         148.118

             BMDL =        48.5214
                            Probit Model with 0.95 Confidence Level
      t   0.6
          0.4
               Probit
                BMDL
                        BMD
                         200
                                  400       600

                                      dose
                                                    800
                                                             1000
        15:52 12/282005
                                             A-65
                                                   DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Quantal Linear Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
 3            Input Data File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:   C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
 5                                              Wed Dec 28 15:54:53 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(-slope*dose)]
14
15
16       Dependent variable = COLUMN2
17       Independent variable = COLUMN1
18
19       Total number of observations  = 4
20       Total number of records with  missing values = 0
21       Maximum number of iterations  = 250
22       Relative Function Convergence has been set to: le-008
23       Parameter Convergence has been set to: le-008
24
25
26
27                      Default Initial (and Specified) Parameter Values
28                         Background  =      0.34375
29                              Slope  =   0.00194591
30                              Power  =            1   Specified
31
32
33               Asymptotic Correlation Matrix of Parameter Estimates
34
35               ( *** The model parameter(s)  -Power
36                     have been estimated at a boundary point, or have been
37    specified by the user,
38                     and do not appear in the correlation matrix )
39
40                 Background        Slope
41
42    Background            1        -0.42
43
44         Slope        -0.42            1
45
46
47
48                              Parameter Estimates
49
50           Variable           Estimate             Std. Err.
51         Background            0.385756           0.0913888
52              Slope          0.00207382         0.000852767
53
54
55
56                            Analysis of Deviance Table
57
58           Model      Log(likelihood)   Deviance  Test DF     P-value
59         Full model        -33.9494
60       Fitted model        -34.4584        1.01801      2          0.6011
61      Reduced model        -40.7516        13.6045      3        0.003496
62
63               AIC:         72.9167
64
                                          A-66      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL
Goodness of Fit
._Prob. Expected
.3858 5.786
.4347 6.520
.5943 8.914
.9228 13.842

1.02 DF = 2


Computation

0.1

Extra risk

0.95

50.8051

27.7368
Scaled
Observed Size Residual
5 15 -0.4171
8 15 0.771
8 15 -0.4808
14 15 0.153

P-value = 0.5996













                                Quantal Linear Model with 0.95 Confidence Level
            0.6
            0.4
            0.2
                     Quantal Linear
                               200
                                           400
                                                      600
                                                                 800
                                                                            1000
26
27
                                                dose
          15:54 12/282005
                                                        A-67
DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Quantal Quadratic Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
 3            Input Data File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
 4            Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
 5                                              Wed Dec 28 15:55:22 2005
 6     ====================================================================
 7
 8     BMDS MODEL RUN

10
11       The form of the probability function is:
12
13       P[response]  = background + (1-background)*[1-EXP(-slope*dose^2)]
14
15
16       Dependent variable = COLUMN2
17       Independent variable = COLUMN1
18
19       Total number of observations = 4
20       Total number of records with missing values = 0
21       Maximum number of iterations = 250
22       Relative Function Convergence has been set to: le-008
23       Parameter Convergence has been set to: le-008
24
25
26
27                      Default Initial (and Specified) Parameter Values
28                         Background =      0.34375
29                              Slope = 1.94591e-006
30                              Power =            2   Specified
31
32
33               Asymptotic Correlation Matrix of Parameter Estimates
34
35               ( *** The model parameter(s)   -Power
36                     have been estimated at a boundary point, or have been
37    specified by the user,
38                     and do not appear in the correlation matrix )
39
40                 Background        Slope
41
42    Background            1        -0.23
43
44         Slope        -0.23            1
45
46
47
48                              Parameter Estimates
49
50           Variable           Estimate             Std. Err.
51         Background            0.449003           0.0779228
52              Slope        2.16129e-006        1.00801e-006
53
54
55
56                            Analysis of Deviance Table
57
58           Model      Log(likelihood)   Deviance  Test DF     P-value
59         Full model        -33.9494
60       Fitted model        -34.6163       1.33385      2          0.5133
61      Reduced model        -40.7516       13.6045      3        0.003496
62
63               AIC:         73.2326
64
                                          A-68      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL
Goodness of Fit
._Prob. Expected
.4490 6.735
.4509 6.764
.4946 7.420
.9365 14.048

1.32 DF = 2


Computation

0.1

Extra risk

0.95

220.792

153.94
Scaled
Observed Size Residual
5 15 -0.9007
8 15 0.6416
8 15 0.2998
14 15 -0.05091

P-value = 0.5181













                               Quantal Quadratic Model with 0.95 Confidence Level
            0.6
            0.4
            0.2
                      Quantal Quadratic
                          BMDL
                                  BMD
                               200
                                           400
                                                      600
                                                                  800
                                                                             1000
26
                                                dose
          15:55 12/282005
                                                        A-69
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
        Weibull Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16  $
        Input Data File: C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.(d)
        Gnuplot Plotting File:  C:\BMDS\FAWELL_FEMALE_CHRONIC_INFLAMMATION.plt
                                          Wed Dec 28  15:55:56  2005
 BMDS MODEL RUN


   The form of the probability function is:

   P[response] = background +  (1-background)*[1-EXP(-slope*dose^power)]
   Dependent variable = COLUMN2
   Independent variable = COLUMN1
   Power parameter is restricted as power >=1

   Total number of observations = 4
   Total number of records with missing values =  0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial  (and Specified) Parameter Values
                     Background =      0.34375
                          Slope =  0.000217656
                          Power =      1.31712
           Asymptotic Correlation Matrix of Parameter Estimates

             Background        Slope        Power

Background            1        -0.77         0.76

     Slope        -0.77            1            -1

     Power         0.76           -1            1
       Variable
     Background
          Slope
          Power
        Parameter Estimates

        Estimate
         0.396812
       0.00106435
          1.09925
            Std. Err.
            0.136488
          0.00719029
            0.999439
                        Analysis of Deviance Table
       Model
     Full model
   Fitted model
  Reduced model

           AIC:
Log(likelihood)
     -33.9494
     -34.4529
     -40.7516

      74.9057
Deviance  Test DF
P-value
   1.00699
   13.6045
     0.3156
   0.003496
                                          A-70
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Dose Est
0.0000 0
40.0000 0
200.0000 0
1000.0000 0

Chi -square =


Benchmark Dose

Specified effect

Risk Type

Confidence level

BMD

BMDL

Goodness of Fit
._Prob. Expected
.3968 5.952
.4327 6.491
.5792 8.689
.9271 13.906

1.01 DF = 1


Computation

0.1

Extra risk

0.95

65.3741

27.7679

Scaled
Observed Size Residual
5 15 -0.5025
8 15 0.7864
8 15 -0.3601
14 15 0.09327

P-value = 0.3151














                                  Weibull Model with 0.95 Confidence Level
27
28
29
30
            0.6
            0.4
            0.2
                  Weibull
                  BMDL  BMD
                              200
                                          400
                                                     600
                                                                800
                                                                           1000
                                               dose
          15:55 12/282005
                                                       A-71
DRAFT: DO NOT CITE OR QUOTE

-------
 1    Heinze,  1999 Relative Liver Weight Changes
 2
 4           Polynomial Model.  Revision: 2.2  Date: 9/12/2002
 5           Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS \_CYANO
 6   TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT. (d)
 7           Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 8   DOCUMENTS \_CYANO TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT .pit
 9                                             Thu May 18 09:37:38 2006
10    ====================================================================
11
12    BMDS MODEL RUN
13   _____________________________________________________________________
14
15      The form of the response function is:
16
17      Y[dose]  = beta_0 + beta_l*dose + beta_2*dose^2 + ...
18
19
20      Dependent variable = MEAN
21      Independent variable =  COLUMN1
22      rho is set to 0
23      The polynomial coefficients  are restricted to be positive
24      A constant variance model is fit
25
26      Total number of dose groups  = 3
27      Total number of records with missing values = 0
28      Maximum number of iterations = 250
29      Relative Function Convergence has been set to: le-008
30      Parameter Convergence has been set to: le-008
31
32
33
34                     Default  Initial Parameter Values
35                             alpha =       0.1466
36                               rho =            0   Specified
37                            beta_0 =      2.84857
38                            beta_l =   0.00447143
39
40
41
42                                    Parameter Estimates
43
44                                                            95.0% Wald Confidence
45   Interval
46          Variable         Estimate        Std. Err.     Lower Conf. Limit
47   Upper Conf. Limit
48             alpha         0.143276        0.0369936           0.0707695
49   0.215782
50            beta_0          2.84857         0.101163              2.6503
51   3.04685
52            beta_l       0.00447143       0.00110819          0.00229942
53   0.00664343
54
55
56              Asymptotic Correlation Matrix of Parameter Estimates
57
58                     alpha       beta_0       beta_l
59        alpha            1     5.6e-009     1.9e-009
60       beta_0     5.6e-009            1        -0.73
61       beta_l     1.9e-009        -0.73            1
62
63
64        Table of Data and Estimated Values of Interest
                                          A-72      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
 Dose
Res.
    0
   50
  150
        Obs Mean
           Obs Std Dev   Est Mean
     Est Std Dev
10
10
10
2.75
3.22
3.47
0.29
0.34
0.49
2.85
3.07
3.52
0.379
0.379
0.379
-0.824
1.24
-0.412
  Model Descriptions for likelihoods  calculated
 Model Al:         Yij
           Var{e(ij) }

 Model A2:         Yij
           Var{e(ij)
             = Mu (i) + e(ij)
               Mu(i) + e(ij )
               Sigma(i)^2
 Model  R:         Yi = Mu + e(i)
            Var{e(i) } =
              Likelihoods of Interest

   Model      Log(likelihood)   DF
    Al           15.381120       4
    A2           16.880747       6
  fitted         14.144766       2
     R            7.133405       2
                                                    AIC
                                                 -22.762240
                                                 -21.761494
                                                 -24.289533
                                                 -10.266809
 Test 1:
levels

 Test 2:
 Test 3 :
 Does response and/or variances differ among dose

 (A2 vs. R)
 Are Variances Homogeneous  (Al vs A2)
 Does the Model for the Mean Fit  (Al vs. fitted)
   Test

   Test 1
   Test 2
   Test 3
            Tests of Interest

  -2*log(Likelihood Ratio)  Test df
              19.4947
              2.99925
              2.47271
                      4
                      2
                      1
p-value

  <.0001
  0.2232
  0.1158
The p-value for Test 1 is less than  .05.   There  appears
to be a
difference between response and/or variances  among the
dose levels.
It seems appropriate to model the data

The p-value for Test 2 is greater than  .05.   A
homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than  .05.
chosen appears
to adequately describe the data
 Benchmark Dose Computation
                                    The model
                                          A-73
                                                DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
Specified effect  =

Risk Type


Confidence level  =

              BMD  =


             BMDL  =
           3.8
           3.6
           3.4
       c
       o
       Q_
       0   3.2
       ce
       c
       ro
       0
           2.8
           2.6
                  Linear
         09:3705/182006
Estimated standard deviations  from the  control mean


      0.95

  84.6525


  57.9321




       Linear Model with 0.95 Confidence Level
                                        BMD
                                                 BMD
                            20
                               40
              60
   80

dose
100
120
140
160
                                              A-74
                                                    DRAFT: DO NOT CITE OR QUOTE

-------
 1    Heinze,  1999  Lactate Dehydrogenase Changes
 2
 4           Polynomial Model.  Revision: 2.2  Date: 9/12/2002
 5           Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS \_CYANO
 6   TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT. (d)
 7           Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 8   DOCUMENTS \_CYANO TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT .pit
 9                                             Thu May 18 09:40:36 2006
10    ====================================================================
11
12    BMDS MODEL RUN
13   _____________________________________________________________________
14
15      The form of the response function is:
16
17      Y[dose]  = beta_0 + beta_l*dose + beta_2*dose^2 + ...
18
19
20      Dependent variable = MEAN
21      Independent variable =  COLUMN1
22      rho is set to 0
23      The polynomial coefficients  are restricted to be positive
24      A constant variance model is fit
25
26      Total number of dose groups  = 3
27      Total number of records with missing values = 0
28      Maximum number of iterations = 250
29      Relative Function Convergence has been set to: le-008
30      Parameter Convergence has been set to: le-008
31
32
33
34                     Default  Initial Parameter Values
35                             alpha =      15.6395
36                               rho =            0   Specified
37                            beta_0 =        20.22
38                            beta_l =        0.101
39
40
41
42                                    Parameter Estimates
43
44                                                            95.0% Wald Confidence
45   Interval
46          Variable         Estimate        Std. Err.     Lower Conf. Limit
47   Upper Conf. Limit
48             alpha           29.028          7.49498             14.3381
49   43.7179
50            beta_0            20.22          1.43994             17.3978
51   23.0422
52            beta_l            0.101        0.0157738            0.070084
53   0.131916
54
55
56              Asymptotic Correlation Matrix of Parameter Estimates
57
58                     alpha       beta_0       beta_l
59        alpha            1     6.2e-015     1.5e-015
60       beta_0     6.2e-015            1        -0.73
61       beta_l     1.5e-015        -0.73            1
62
63
64        Table of Data and Estimated Values of Interest
                                          A-75      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
 Dose
Res.
    0
   50
  150
        Obs Mean
           Obs Std Dev   Est Mean
      Est Std Dev
10
10
10
16.6
30.6
33.6
4.48
5.05
1.16
20.2
25.3
35.4
5.39
5.39
5.39
-2.1
3.15
-1.05
  Model Descriptions for likelihoods calculated
 Model Al:         Yij
           Var{e(ij) }

 Model A2:         Yij
           Var{e(ij)
             = Mu (i) + e(ij)
               Mu(i) + e(ij )
               Sigma(i)^2
 Model  R:         Yi = Mu + e(i)
            Var{e(i) } =
              Likelihoods of Interest

   Model      Log(likelihood)   DF
    Al          -54.666589       4
    A2          -46.093905       6
  fitted        -65.523922       2
     R          -78.954450       2
                                                    AIC
                                                 117 .333177
                                                 104.187810
                                                 135.047844
                                                 161.908899
 Test 1:
levels

 Test 2:
 Test 3 :
 Does response and/or variances differ among dose

 (A2 vs. R)
 Are Variances Homogeneous  (Al vs A2)
 Does the Model for the Mean Fit  (Al vs. fitted)
   Test

   Test 1
   Test 2
   Test 3
            Tests of Interest

  -2*log(Likelihood Ratio)  Test df
              65.7211
              17.1454
              21 .7147
                      4
                      2
                      1
 p-value

   <.0001
0.0001892
   <.0001
The p-value for Test 1 is less than  .05.   There  appears
to be a
difference between response and/or variances  among  the
dose levels.
It seems appropriate to model the data

The p-value for Test 2 is less than  .05.   Consider
running a
non-homogeneous variance model

The p-value for Test 3 is less than  .05.   You may want
to try a
different model
 Benchmark Dose Computation
Specified effect =
                                          A-76
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1   Risk Type        =     Estimated standard deviations  from the control mean
2
o
J
4   Confidence  level =          0.95
5
6               BMD =       53.3442
7
8
9              BMDL =       39.9199
                                        A-77      DRAFT: DO NOT CITE OR QUOTE

-------
 1    ====================================================================
 2            Polynomial Model.  Revision:  2.2  Date: 9/12/2002
 3            Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS\_CYANO
 4    TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT.(d)
 5            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 6    DOCUMENTS\_CYANO TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT.pit
 7                                              Thu May 18 09:42:43 2006
 8     ====================================================================
 9
10     BMDS MODEL RUN

12
13       The form of the response function is:
14
15       Y[dose]  = beta_0 + beta_l*dose +  beta_2*dose^2 + ...
16
17
18       Dependent variable = MEAN
19       Independent variable =  COLUMN1
20       The polynomial coefficients  are restricted to be positive
21       The variance is to be modeled as  Var(i)  = alpha*mean(i)^rho
22
23       Total number of dose groups  = 3
24       Total number of records with missing values = 0
25       Maximum number of iterations = 250
26       Relative Function Convergence has been set to: le-008
27       Parameter Convergence has been set to: le-008
28
29
30
31                      Default  Initial Parameter Values
32                              alpha =      15.6395
33                                rho =            0
34                             beta_0 =        20.22
35                             beta_l =        0.101
36
37
38
39                                     Parameter Estimates
40
41                                                             95.0% Wald Confidence
42    Interval
43           Variable         Estimate        Std. Err.     Lower Conf. Limit
44    Upper Conf. Limit
45              alpha     2.44639e+006     4.30425e+006       -5.98979e+006
46    1.08826e+007
47                rho         -3.55294         0.556161              -4.643
48    -2.46289
49             beta_0          21.3722          1.64609              18.146
50    24.5985
51             beta_l        0.0863674        0.0134905           0.0599266
52    0.112808
53
54
55               Asymptotic Correlation Matrix of Parameter Estimates
56
57                      alpha          rho       beta_0       beta_l
58         alpha            1        -0.99         0.19        -0.23
59           rho        -0.99            1        -0.21         0.25
60        beta_0         0.19        -0.21            1        -0.89
61        beta_l        -0.23         0.25        -0.89            1
62
63
64         Table of Data and Estimated Values of Interest
                                          A-78      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
 Dose
Res.
    0
   50
  150
        Obs Mean
           Obs Std Dev   Est Mean
      Est Std Dev
10
10
10
16.6
30.6
33.6
4.48
5.05
1.16
21.4
25.7
34.3
6.79
4.9
2.93
-2.2
3.2
-0.808
 Model Descriptions for likelihoods  calculated
 Model Al:         Yij
           Var{e(ij) }

 Model A2:         Yij
           Var{e(ij) }

 Model A3:         Yij
           Var{e(ij) }
 Model  R:         Yi
            Var{e(i) }
            Model
             Al
             A2
             A3
           fitted
              R
             = Mu (i) + e(ij)
               Mu(i) + e(ij )
               Sigma(i)^2

               Mu(i) + e(ij )
               alpha* (Mud) )

               Mu + e(i)
               Sigma^2
              Likelihoods of Interest

              Log(likelihood)   DF
                -54.666589       4
                -46.093905       6
                -54.022155       5
                -61.092641       4
                -78.954450       2
                                 AIC
                              117.333177
                              104.187810
                              118 . 044309
                              130.185281
                              161.908899
 Test 1:
levels?

 Test 2 :
 Test 3:
 Test 4:
          Explanation of Tests

 Does response and/or variances differ among Dose

 (A2 vs. R)
 Are Variances Homogeneous?  (Al vs A2)
 Are variances adequately modeled?  (A2 vs. A3)
 Does the Model for the Mean Fit?  (A3 vs.  fitted)
   Test

   Test 1
   Test 2
   Test 3
   Test 4
            Tests of Interest

  -2*log(Likelihood Ratio)  Test df
              65.7211
              17 . 1454
              15.8565
               14.141
                      4
                      2
                      1
                      1
    p-value

   <.0001
0.0001892
   <.0001
0.0001696
The p-value for Test 1 is less than  .05.   There  appears
to be a
difference between response and/or variances  among the
dose levels
It seems appropriate to model the data

The p-value for Test 2 is less than  .05.   A
non-homogeneous variance
model appears to be appropriate
                                          A-79
                                                DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2    The p-value for Test 3 is less than .05.   You may want
 3    to consider a
 4    different variance model
 5
 6    The p-value for Test 4 is less than .05.   You may want
 7    to try a different
 8    model
 9
10
11     Benchmark Dose Computation
12    Specified effect =             1
13
14    Risk Type        =     Estimated standard deviations from the control mean
15
16
17    Confidence level =          0.95
18
19                 BMD =       78.6091
20
21
22                BMDL =       63.5158
23
24
25
26
27
                                          A-80      DRAFT: DO NOT CITE OR QUOTE

-------
 1    Heinze,  1999 Alkaline Phosphatase Changes
 2
 o
 J
 4    ====================================================================
 5            Polynomial Model.  Revision: 2.2  Date: 9/12/2002
 6            Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY DOCUMENTS\_CYANO
 7    TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT.(d)
 8            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 9    DOCUMENTS\_CYANO TOX REV\MODELING\HEINZE_ENZYMES_AND_LIVER_WT.pit
10                                              Thu May 18 09:46:03 2006
11     ====================================================================
12
13     BMDS MODEL RUN
14    	
15
16       The form of the response function is:
17
18       Y[dose]  = beta_0 + beta_l*dose + beta_2*dose^2 + ...
19
20
21       Dependent variable = MEAN
22       Independent variable =  COLUMN1
23       rho is set to 0
24       The polynomial coefficients  are restricted to be positive
25       A constant variance model is fit
26
27       Total number of dose groups  = 3
28       Total number of records with missing values = 0
29       Maximum number of iterations = 250
30       Relative Function Convergence has been set to: le-008
31       Parameter Convergence has been set to: le-008
32
33
34
35                      Default  Initial Parameter Values
36                              alpha =      7.59287
37                                rho =            0   Specified
38                             beta_0 =      10.6414
39                             beta_l =            0
40
41
42
43                                     Parameter Estimates
44
45                                                             95.0% Wald Confidence
46    Interval
47           Variable         Estimate        Std. Err.     Lower Conf. Limit
48    Upper Conf. Limit
49              alpha          7.93453          2.04869             3.91918
50    11.9499
51             beta_0          10.6414          0.75283             9.16591
52    12.117
53             beta_l        0.0180286       0.00824683          0.00186506
54    0.0341921
55
56
57               Asymptotic Correlation Matrix of Parameter Estimates
58
59                      alpha       beta_0       beta_l
60         alpha            1     l.le-006    -4.7e-007
61        beta_0     l.le-006            1        -0.73
62        beta_l    -4.7e-007        -0.73            1
63
64
                                          A-81      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 J
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
 Dose
Res.
Table of Data and Estimated Values of  Interest

       N    Obs Mean    Obs Std Dev    Est Mean
                       Est Std Dev
    0    10
   50    10
  150    10
             9.67
               13
             12.9
 2.2
3.81
1.85
10.6
11.5
13.3
2.82
2.82
2.82
  Model Descriptions for likelihoods calculated
 Model Al:         Yij
           Var{e(ij) }

 Model A2:         Yij
           Var{e(ij) }

 Model  R:          Yi
            Var{e(i) }
            Model
             Al
             A2
           fitted
              R
                 = Mu (i) + e (ij)
                 = Mu (i) + e (ij)
                 = Sigma(i)*2

                 = Mu + e(i)
                  Likelihoods of Interest

                  Log(likelihood)   DF
                    -43.827730       4
                    -40.832314       6
                    -46.068366       2
                    -48.794188       2
                    AIC
                  95.655461
                  93.664628
                  96.136733
                 101.588375
 Test 1:
levels

 Test 2:
 Test 3 :
     Does response and/or variances differ  among  dose

     (A2 vs. R)
     Are Variances Homogeneous  (Al vs A2)
     Does the Model for the Mean Fit  (Al vs.  fitted)
   Test

   Test 1
   Test 2
   Test 3
                Tests of Interest

      -2*log(Likelihood Ratio)  Test df
                  15.9237
                  5.99083
                  4 .48127
         4
         2
         1
     p-value

    0.0003485
      0.05002
      0.03427
The p-value for Test 1 is less than  .05.  There  appears
to be a
difference between response and/or variances  among  the
dose levels.
It seems appropriate to model the data

The p-value for Test 2 is greater than  .05.   A
homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is less than  .05.  You  may  want
to try a
different model
 Benchmark Dose Computation
 -1.09
  1.64
-0.545
                                          A-82
                                                DRAFT: DO NOT CITE OR QUOTE

-------
 1    Specified effect =             1
 2
 3    Risk Type        =     Estimated standard deviations from the control mean
 4
 5
 6    Confidence  level =          0.95
 7
 8                BMD =       156.243
 9
10
11               BMDL =       87.6609
12
13
14
15
16
                                         A-83      DRAFT: DO NOT CITE OR QUOTE

-------
 1    Fawell et al.,  1999 Male Alanine Aminotransferase Changes
 2
 3    ====================================================================
 4            Polynomial Model.  Revision:  2.2  Date: 9/12/2002
 5            Input Data File: C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 6    DOCUMENTS\_CYANO TOX REV\MODELING\FAWELL_MALE_ALT.(d)
 7            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 8    DOCUMENTS\_CYANO TOX REV\MODELING\FAWELL_MALE_ALT.pit
 9                                              Tue May 09 12:59:55 2006
10     ====================================================================
11
12     HMDS MODEL RUN
13    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
14
15       The form of the response function is:
16
17       Y[dose]  = beta_0 + beta_l*dose + beta_2*doseA2 + ...
18
19
20       Dependent variable = MEAN
21       Independent variable = COLUMN1
22       rho is set to 0
23       The polynomial coefficients  are restricted to be positive
24       A constant variance model is fit
25
26       Total number of dose groups  = 4
27       Total number of records with missing values = 0
28       Maximum number of iterations = 250
29       Relative Function Convergence has been set to: le-008
30       Parameter Convergence has been set to: le-008
31
32
33
34                      Default Initial Parameter Values
35                              alpha =            1
36                                rho =            0   Specified
37                             beta_0 =      30.4717
38                             beta_l =     0.129124
39
40
41
42                                     Parameter Estimates
43
44                                                             95.0% Wald
45    Confidence Interval
46           Variable         Estimate        Std. Err.     Lower Conf. Limit
47    Upper Conf. Limit
48              alpha          1584.79           289.34             1017.69
49    2151.88
50             beta_0          30.4717          6.47011             17.7905
51    43.1529
52             beta_l         0.129124        0.0126792            0.104273
53    0.153974
54
55
56               Asymptotic Correlation Matrix of Parameter Estimates
57
                                          A-84      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
o
6
4
5
6
7
9
10


12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
alpha beta 0 beta 1
alpha 1 -6.1e-OC>8 9.2e-OC>7
beta 0 -6.1e-008 1 -0.61
beta 1 9.2e-007 -0.61 1


Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .


-

0 15 27 8 30.5 39.8
40 15 37 17.2 35.6 39.8
200 15 59 28 56.3 39.8
1000 15 159 75 160 39.8



Model Descriptions for likelihoods calculated


Model Al: Yij = Mu(i) + e(ij)
Var{e (ij ) } = SigmaA2

Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma (i)A2

Model R: Yi = Mu + e(i)
Var{e(i) } = SigmaA2


Likelihoods of Interest

Model Log (likelihood) DF AIC
Al -250.943933 5 511.887866
A2 -216.540867 8 449.081734
fitted -251.046212 2 506.092423
R -281.663312 2 567.326624

Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2 )
Test 3: Does the Model for the Mean Fit (Al vs. fitted)

Tests of Interest

Test -2*log (Likelihood Ratio) Test df p-value

Test 1 130.245 6 <.0001
Test 2 68.8061 3 <.0001
Test 3 0.204557 2 0.9028

The p-value for Test 1 is less than .05. There appears
to be a







ChiA2





-0.338
0.133
0.263
-0.0579








































A-85
DRAFT: DO NOT CITE OR QUOTE

-------
 1    difference between response and/or variances among the
 2    dose levels.
 3    It seems  appropriate to model the data
 4
 5    The p-value for Test 2  is less than .05.   Consider
 6    running a
 7    non-homogeneous variance model
 8
 9    The p-value for Test 3  is greater than .05.   The model
10    chosen appears
11    to adequately describe  the data
12
13
14
15     Benchmark Dose Computation
16    Specified effect =             1
17
18    Risk Type        =     Estimated standard deviations from the control mean
19
20
21    Confidence level =          0.95
22
23                 BMD =       308.304
24
25
26                BMDL =       252.245
27
                                          A-86      DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 6
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
        Polynomial Model. Revision: 2.2  Date:  9/12/2002
        Input Data File: C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
DOCUMENTS\_CYANO TOX REV\MODELING\FAWELL_MALE_ALT.(d)
        Gnuplot Plotting File:  C:\DOCUMENTS AND  SETTINGS\HCLYNCH\MY
DOCUMENTS\_CYANO TOX REV\MODELING\FAWELL_MALE_ALT.pit
                                          Tue May 09 13:00:39  2006
 HMDS MODEL RUN


   The form of the response function is:

   Y[dose] = beta 0 + beta l*dose + beta 2*doseA2
   Dependent variable = MEAN
   Independent variable = COLUMN1
   The polynomial coefficients are restricted to be positive
   The variance is to be modeled as Var(i) = alpha*mean(i)Arho

   Total number of dose groups = 4
   Total number of records with missing values = 0
   Maximum number of iterations = 250
   Relative Function Convergence has been set to: le-008
   Parameter Convergence has been set to: le-008
                  Default Initial Parameter Values
                          alpha =             1
                            rho =             0
                         beta_0 =      30.4717
                         beta 1 =     0.129124
Confidence Interval
       Variable         Estimate
Upper Conf. Limit
          alpha         0.098324
0.32663
            rho          2.15231
2.72391
         beta_0          28.6439
33.2276
         beta_l         0.141692
0.177912
Parameter Estimates



       Std.  Err.

        0.116485

        0.291636

         2.33865

       0.0184799
   95.0% Wald

Lower Conf. Limit

      -0.129982

        1.58072

        24.0602

       0.105472
           Asymptotic Correlation Matrix of Parameter  Estimates
                                          A-87
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
alpha rho beta 0 beta 1
alpha 1 -0.99 O."06 -0.11
rho -0.99 1 -0.065 0.18
beta 0 0.06 -0.065 1 -0.39
beta 1 -0.17 0.18 -0.39 1


Table of Data and Estimated Values of Interest

Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .
_

0 15 27 8 28.6 11.6
40 15 37 17.2 34.3 14.1
200 15 59 28 57 24.3
1000 15 159 75 170 79



Model Descriptions for likelihoods calculated


Model Al: Yij = Mu(i) + e(ij)
Var{e(ij) } = SigmaA2

Model A2 : Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma (i)A2

Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = alpha* (Mu (i )) Arho

Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2


Likelihoods of Interest

Model Log (likelihood) DF AIC
Al -250.943933 5 511.887866
A2 -216.540867 8 449.081734
A3 -217.537818 6 447.075636
fitted -219.838125 4 447.676251
R -281.663312 2 567.326624


Explanation of Tests

Test 1: Does response and/or variances differ among Dose
levels?
(A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2 )
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)

Tests of Interest









ChiA2



-0.549
0.739
0.321
-0.556







































A-88
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2       Test    -2*log(Likelihood Ratio)   Test df        p-value
 o
 J
 4       Test 1              130.245          6          <.0001
 5       Test 2              68.8061          3          <.0001
 6       Test 3               1.9939          2           0.369
 7       Test 4              4.60062          2          0.1002
 8
 9    The p-value for Test 1 is less than .05.   There appears
10    to be a
11    difference between response and/or variances among the
12    dose levels
13    It seems appropriate to model the data
14
15    The p-value for Test 2 is less than .05.   A
16    non-homogeneous variance
17    model appears to be appropriate
18
19    The p-value for Test 3 is greater than .05.   The
20    modeled variance appears
21     to be appropriate here
22
23    The p-value for Test 4 is greater than .05.   The model
24    chosen seems
25    to adequately describe the data
26
27
28     Benchmark Dose Computation
29    Specified effect =             1
30
31    Risk Type        =     Estimated standard deviations from the control mean
32
33
34    Confidence level =          0.95
35
36                 BMD =       81.8426
37
38
39                BMDL =       58.3727
                                          A-89      DRAFT: DO NOT CITE OR QUOTE

-------
                                 Linear Model with 0.95 Confidence Level
 1
 2
 o
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
       I
       o
       8-
       
-------
 1
 2
 o
 6
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
                  Default Initial Parameter Values
                          alpha =
                            rho =
                         beta_0 =
                         beta_l =
                         beta 2 =
                                    1
                                    0
                              28.5738
                             0.160635
                                    0
Confidence Interval
       Variable
Upper Conf. Limit
          alpha
0.32663
            rho
2.72391
         beta_0
33.2276
         beta_l
0.177912
         beta 2
               Estimate

              0.0983241

                2.15231

                28.6439

               0.141692

                      0
             Parameter Estimates



                    Std. Err.

                     0.116485

                     0.291636

                      2.33865

                    0.0184799

                           NA
                          95.0% Wald

                       Lower Conf. Limit

                             -0.129982

                               1.58072

                               24.0602

                              0.105472
NA - Indicates that this parameter has hit a bound
     implied by some inequality constraint and thus
     has no standard error.
           Asymptotic Correlation Matrix of Parameter Estimates
     alpha
       rho
    beta_0
    beta 1
         alpha
             1
         -0.99
          0.06
         -0.17
             rho
           -0.99
               1
          -0.065
            0.18
            beta_0
              0.06
            -0.065
                 1
             -0.39
            beta_l
             -0.17
              0.18
             -0.39
                 1
The following parameter(s) have been estimated at a
boundary
point or have been specified.  Correlations are not
computed:

beta 2
 Dose
Res .
     Table of Data and Estimated Values of Interest

            N    Obs Mean    Obs Std Dev   Est Mean   Est Std Dev   ChiA2
    0
   40
15
15
27
37
17.2
28.6
34.3
11.6
14.1
-0.549
 0.739
                                          A-91
                                                DRAFT: DO NOT CITE OR QUOTE

-------
1 200 15 59 28 57
2 1000 15 159 75 170
3
4
5
6 Model Descriptions for likelihoods calculated
7
9 Model Al: Yij = Mu(i) + e(ij)
10 Var{e(ij)} = SigmaA2
11
12 Model A2: Yij = Mu(i) + e(ij)
13 Var{e(ij)} = Sigma (i)A2
14
15 Model A3: Yij = Mu(i) + e(ij)
16 Var{e(ij)} = alpha* (Mu (i )) Arho
17
18 Model R: Yi = Mu + e(i)
19 Var{e(i) } = SigmaA2
20
21
22 Likelihoods of Interest
23
24 Model Log (likelihood) DF
25 Al -250.943933 5
26 A2 -216.540867 8
27 A3 -217.537818 6
28 fitted -219.838125 4
29 R -281.663312 2
30
31
32 Explanation of Tests
33
34 Test 1: Does response and/or variances differ
35 levels?
36 (A2 vs. R)
37 Test 2: Are Variances Homogeneous? (Al vs A2 )
38 Test 3: Are variances adequately modeled? (A2
24.3 0.321
79 -0.556




















AIC
511.887866
449.081734
447.075636
447.676251
567.326624




among Dose



vs . A3 )
39 Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
40
41 Tests of Interest
42
43 Test -2*log (Likelihood Ratio) Test df
44
45 Test 1 130.245 6
46 Test 2 68.8061 3
47 Test 3 1.9939 2
48 Test 4 4.60062 1
49
50 The p-value for Test 1 is less than .05. There
51 to be a



p-value

<.0001
<.0001
0.369
0.03196

appears

52 difference between response and/or variances among the
53 dose levels
54 It seems appropriate to model the data
55
56 The p-value for Test 2 is less than .05. A
57 non-homogeneous variance





A-92
DRAFT: DO NOT CITE OR QUOTE

-------
 1
 2
 o
 6
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
model appears to be appropriate

The p-value for Test 3 is  greater than  .05.
modeled  variance appears
 to be appropriate here
                       The
The p-value for Test 4 is  less than .05.   You may want
to try  a  different
model
 Benchmark  Dose Computation
Specified effect =              1
Risk Type


Confidence  level =

              BMD =


             BMDL =
Estimated  standard deviations  from the control  mean


     0.95

  81.8426


  58.3727

   Polynomial Model with 0.95 Confidence Level
       I
       I
       in
       
-------
 1
 2     ====================================================================
 3            Power Model.  $Revision:  2.1 $ $Date: 2000/10/11 20:57:36 $
 4            Input Data File:  C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 5    DOCUMENTS\_CYANO TOX REV\MODELING\FAWELL_MALE_ALT.(d)
 6            Gnuplot Plotting File:   C:\DOCUMENTS AND SETTINGS\HCLYNCH\MY
 7    DOCUMENTS\_CYANO TOX REV\MODELING\FAWELL_MALE_ALT.pit
 8                                              Tue May 09 13:03:23 2006
 9     ====================================================================
10
11     HMDS MODEL RUN
17    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
13
14       The form of the response function is:
15
16       Y[dose]  = control + slope *  doseApower
17
18
19       Dependent variable = MEAN
20       Independent variable = COLUMN1
21       The power is restricted to be greater than or equal to 1
22       The variance is to be modeled as Var(i)  = alpha*mean(i)Arho
23
24       Total number of dose groups  = 4
25       Total number of records with missing values = 0
26       Maximum number of iterations = 250
27       Relative Function Convergence has been set to: le-008
28       Parameter Convergence has been set to: le-008
29
30
31
32                      Default Initial Parameter Values
33                              alpha =      1692.21
34                                rho =            0
35                            control =           27
36                              slope =     0.519763
37                              power =     0.801589
38
39
40               Asymptotic Correlation Matrix of Parameter Estimates
41
42                      alpha          rho      control        slope        power
43
44         alpha            1        -0.99         0.13        -0.46         0.48
45
46           rho        -0.99            1        -0.12         0.41        -0.43
47
48       control         0.13        -0.12            1        -0.76         0.74
49
50         slope        -0.46         0.41        -0.76            1           -1
51
52         power         0.48        -0.43         0.74           -1            1
53
54
55
56                              Parameter Estimates
57
                                          A-94      DRAFT: DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
9
10
11
12
13
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Variable Estimate Std. Err.
alpha 0.098324 0.141722
rho 2.15231 0.343016
control 28.6439 3.51452
slope 0.141692 0.218496
power 1 0.231677


Table of Data and Estimated Values of Interest

Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .
_

0 15 27 8 28.6 11.6
40 15 37 17.2 34.3 14.1
200 15 59 28 57 24.3
1000 15 159 75 170 79



Model Descriptions for likelihoods calculated


Model Al: Yij = Mu(i) + e(ij)
Va r { e ( i j ) } = Si gma A 2

Model A2 : Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma (i)A2

Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = alpha* (Mu (i )) Arho

Model R: Yi = Mu + e(i)
Var{e(i) } = SigmaA2


Likelihoods of Interest

Model Log (likelihood) DF AIC
Al -250.943933 5 511.887866
A2 -216.540867 8 449.081734
A3 -217.537818 6 447.075636
fitted -219.838125 5 449.676251
R -281.663312 2 567.326624


Explanation of Tests

Test 1: Does response and/or variances differ among Dose levels?
(A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2 )
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)











ChiA2



-0.142
0.191
0.083
-0.144





































A-95
DRAFT: DO NOT CITE OR QUOTE

-------
 1                         Tests of Interest
 2
 3       Test    -2*log(Likelihood Ratio)     d.f        p-value
 4
 5       Test 1              130.245          6         <.00001
 6       Test 2              68.8061          3         <.00001
 7       Test 3               1.9939          2           0.369
 8       Test 4              4.60062          1         0.03196
 9
10    The p-value for Test 1 is less than .05.   There appears to be a
11    difference between response and/or variances among the dose levels
12    It seems appropriate to model the data
13
14    The p-value for Test 2 is less than .05.   A non-homogeneous variance
15    model appears to be appropriate
16
17    The p-value for Test 3 is greater than .05.   The modeled variance appears
18     to be appropriate here
19
20    The p-value for Test 4 is less than .05.   You may want to try a different
21    model
22
23
24     Benchmark Dose Computation
25    Specified effect =             1
26
27    Risk Type        =     Estimated standard deviations  from the control mean
28
29    Confidence level =          0.95
30
31                 BMD =       81.8426
32
33
34                BMDL =       58.3727
35
36
37
                                          A-96      DRAFT: DO NOT CITE OR QUOTE

-------