>EPA
!& §
ill
(A C 25
S
United States
Environmental Protection
Agency
Environmental Research
Laboratory
Gulf Breeze, FL 32561
EPA/600/9-85/018
June 1985
Research and Development
Proceedings of the
Workshop:
Biodegradation
Kinetics, Navarre
Beach, Florida,
18-20 October 1983
o - o
« o B
-------
EPA/600/9-85/018
June 1985
J) PROCEEDINGS OF THE WORKSHOP:
(irj BIODEGRADATION KINETICS
L NAVARRE BEACH, FLORIDA
'^] 18-20 OCTOBER 1983
M
0
\o
VJ
^ Edited and Coordinated by
Al W. Bourquin and P. H. Pritchard
U.S. Environmental Protection Agency
Environmental Research Laboratory
Gulf Breeze, Florida 32561
William W. Walker
Gulf Coast Research Laboratory
East Beach Drive
Ocean Springs, Mississippi 39564
Rod Parrish
8330 Wild Lake Road
Pensacola, Florida 32506
U.S. Environmental Protection Agency
Region 5, Library (PL-12J)
77 West Jackson Bou!evar.d, 12th Floor
Chicago, IL 60604-3590
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
GULF BREEZE, FLORIDA 32561
-------
DISCLAIMER
The information in this document has been funded wholly or in part by
the U.S. Environmental Protection Agency. It has been subject to the
Agency's peer and administrative review and approved for publication as
an EPA document. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
11
-------
FOREWORD
The proliferation of organic chemicals, as a result of the need for new
and better agricultural and industrial chemicals, increases the potential for
environmental pollution, particularly in aquatic habitats. Inherent in any
fate study is the need for an accurate assessment of the potential for the
microbial degradation of the pollutant. However, there is little information
on the rates at which biodegradation will occur in the environment.
Furthermore, kinetic data from laboratory and field studies are often based on
assumptions that have no consensus or sound scientific basis. Thus, the
microbiologist currently has no consistent guidelines regarding appropriate
techniques for the determination of biodegradation rates in natural
environments, nor have microbiologists, in general, provided the chemical
industry and regulatory agencies with evaluations of the effect of
environmental parameters and chemical structures on biodegradation of
processes mediated by microbial populations.
This Workshop on Biodegradation Kinetics is a follow-up to a workshop
held in Gulf Breeze, Florida, and sponsored by the Environmental Research
Laboratory at Gulf Breeze. As was true for the first workshop, this workshop
on biodegradation kinetics was organized to offer representatives from
government, academia, and industry a forum for the examination of key issues
regarding the future direction and focus of scientific investigations in this
field. Participants addressed a number of questions of primary concern: What
are the differences and similarities between biodegradation rates in aquatic,
terrestrial, and laboratory environments? What methodological criteria must
be established to provide interchangeable degradation information? What is
the potential for a particular environment or its laboratory simulation to
dispose of a polluting chemical, and at what rate will it occur? We hope that
this publication of the workshop proceedings will provide an up-to-date
reference for professionals concerned with the fate, regulation, and
production of potential environmental pollutants.
_ ^nos
Director
Environmental Research Laboratory
Gulf Breeze, Florida
Til
-------
ABSTRACT
This workshop, held October 18-20, 1983, at Pensacola Beach, Florida,
focused on pertinent issues related to the scientific investigation of the
microbial degradation rates of organic chemicals in natural environments.
Participants discussed methodological criteria for these investigations and
the need for concentrating on the kinetic aspects of biodegradation. Position
papers dealing with the following topics were presented in open sessions: (1)
Statistical and Experimental Requirements for Modeling Decay Curves; (2) The
"Second Order" Approach Assumption, Limitations and Research Needs; (3)
Factors Controlling Biodegradation Rates in Microbial Communities; (4)
Application of Uptake and Mineralization Kinetics; (5) Relationships between
Chemical Structure and Biodegradation Rates; and (6) Extrapolation of
Laboratory Biodegradation Data to the Field. Discussions within each session
are summarized by the panel members in reports that include a consensus of the
direction and extent of research required for the description of
biodegradation rates of xenobiotic chemicals in natural environments. These
proceedings conclude with a summary report and suggestions for future research
in biodegradation kinetics. This report is submitted in fulfillment of
Contract No. CR 810789 by Gulf Coast Research Laboratory in conjunction with
Georgia State University Cooperative Agreement R809370 under the sponsorship
of the U.S. Environmental Protection Agency. This report covers a period from
April 4, 1984 to May 13, 1984, and work was completed as of June 30, 1984.
i v
-------
CONTENTS
Page
Foreword i i i
Abstract iv
Acknowledgment vi
Workshop Organization , 1
A. Objectives 1
B. Working Concept 1
C. Workshop Structure 1
D. EPA Perspective - By R. Brink, U.S. EPA 2
E. Ad-Hoc Advisory Committee 5
Statistical and Experimental Requirements for Modeling Decay Curves 6
The Second Order Rate Approach: Assumptions, Limitations, and
Research Needs 38
Factors Controlling Biodegradation Rates in Microbial Communities 72
Application and Uptake of Mineralization Kinetics 83
Relationships between Chemical Structure and Biodegradation Rates 117
Extrapolation of Laboratory Biodegradation Data to the Field 127
Summary and Research Plan for Biodegradation Studies 146
Appendix - Proposed Benchmark Chemical for Biodegradation Research....161
-------
ACKNOWLEDGMENT
The assistance of Ms. Gail Seidler, Gulf Breeze Environmental Research
Laboratory, and the Georgia State University Cooperative Agreement Office at
Gulf Breeze, in preparing material before the Workshop is gratefully
acknowledged.
VI
-------
I. Workshop Organization
A. Objectives
1. To evaluate the utility and limitations of approaches currently used
for the derivation of environmentally meaningful biodegradation rates
from laboratory studies.
2. To identify the future research required for the improvement of our
approaches in the development of a strategy for the extrapolation of
laboratory information on the kinetics of biodegradation to complex
environmental situations.
B. Working Concept
Biodegradation studies, in general, provide decay curves for a chemical
by using a particular method and inoculum. Decay curves vary under different
inocula or different experimental conditions. Application of this laboratory-
derived biodegradation information to the field requires that degradation
rates be calculated from the decay curves. This is usually accomplished by
developing mathematical rate equations or kinetic expressions that model the
decay curves. Because of uncertainties about the factors that control the
metabolic diversity and degradative potential of natural communities of
microorganisms, certain assumptions about the activity, concentration, and
substrate affinity of the biological catalysts within these communities and
about the effect of environmental parameters (sorption, temperature, multiple
substrates, etc.) on their activity must be made if useful kinetic expressions
are to be developed and the key, we believe, to measuring biodegradation rates
in the laboratory that can be used to predict the biological fate of a
chemical in the field. A critical evaluation of the logic associated with the
development of each assumption and the research needed to validate the
assumptions was the working theme for this workshop.
C. Workshop Structure
The workshop was divided into six sessions:
1. Statistical and Experimental Requirements for Modeling Decay Curves
2. The "Second Order" Approach: Assumptions, Limitations, and Research
Needs
3. Factors Controlling Biodegradation Rates in Microbial Communities
4. Application of Uptake and Mineralization Kinetics
5. Relationships Between Chemical Structure and Biodegradation Rates
6. Extrapolation of Laboratory Biodegradation Data to the Field
Each session was initiated with several 10-minute talks that briefly
summarized the status, limitations, and research needs of a particular
research approach used to study biodegradation kinetics. The talks, presented
below, were intended to provide representative data upon which subsequent
discussion was based. Position statements and suggestions were encouraged.
The objective of each session was to provide the session chairperson and
a panel of 3-5 individuals with interpretations, strategies, theories,
recommendations, and priorities that they could use to prepare a report to the
-------
workshop coordinator. Session reports were presented to the workshop
participants for further discussion and revision in a general plenary
session. The final panel reports are presented below.
D. An EPA Perspective by Robert Brink, U.S. EPA
My comments will be directed to the needs of the Office of Toxic
Substances (OTS) of EPA, although I am sure that any advances that can be made
in the ability to estimate environmentally realistic biodegradation rates will
be of value to many others in EPA and elsewhere.
The Toxic Substances Control Act (TSCA), under which OTS operates,
applies to all chemical substances not covered by prior acts and allows for
the regulation of those chemical substances that may or do present an
unreasonable risk of injury to human health or the environment.
An evaluation of the risk or likelihood that a chemical may cause an
adverse effect involves a determination of the hazard or adverse effects that
may occur if there is exposure and an assessment of the exposure
possibilities.
An exposure assessment considers the sources, pathways, concentrations
and potentially exposed populations. Chemical fate aspects of the exposure
assessment evaluate the transport and transformation possibilities following
release of the chemical to the environment and the fate conclusions are
important elements of the pathways and concentrations part of the exposure
assessment.
An assessment of the fate of a chemical released to the environment will
depend, in part, upon laboratory data that can be used to evaluate the
properties and processes that influence transport and transformations. In
actual practice, fate assessment includes a consideration of existing data
from reliable literature citations and estimation techniques that should be
considered and that may be used either to avoid unnecessary testing or to
guide the proper selection of laboratory testing. The overall approach is
summarized in Table 1 which lists, from top to bottom, the properties and
processes for which data may be needed and, from left to right, the various
levels of data input. In this scheme, laboratory data at one level will be
required only when data from a lower level are insufficient to answer relevant
exposure assessment questions. The testing may begin at any appropriate
spot. For example, if it is already known that the only significant
transformation processes will be biodegradation studies in soil and water, the
laboratory testing might begin with biodegradation studies that will provide
kinetic information and data on potential intermediates of concern. It is
also highly unlikely that any given chemical would require testing for all or
even a high percentage of the properties or processes. Existing knowledge
estimation techniques and the early base-set tests will provide information
that will preclude much of the other testing.
Environmentally realistic estimations of chemical concentrations of
nonpersistent substances require valid kinetic data with calculations of half-
lives under relevant environmental scenarios. For hydrolysis, the data from
the base-set will usually be sufficient to calculate environmental half-lives,
although there may be circumstances, such as environmental extremes of pH,
-------
that will dictate a need for upper tier simulations to obtain better rate
data. For selected environments, it may be necessary to conduct kinetic
studies at an upper tier, with simulations of particular environmental
variables, to obtain better rate values for leaching, evaporation, and
photodegradation. It is unlikely that the base-set biodegradation tests will
provide sufficient information for valid extrapolation to environmental rates
and it is anticipated that any determination of environmentally relevant
biodegradation rate data will require upper tier kinetic studies. It should
be emphasized that upper tier tests for rate determinations are justified only
for significant transfer transformation processes. For the organic compounds
evaluated by EPA, transformations in soil and water by micro-organisms are
often the most significant transformation processes.
The base set of fate tests will indicate the dominant transformation
processes that a chemical may undergo in natural environments. In some cases,
there will be indications of whether an organic substance is degraded
essentially to inorganic compounds (e.g. carbon dioxide and water) plus normal
metabolic intermediates. If a substance is extremely persistent, this will be
evident from the lack of reaction by any of the transformation processes. For
all of the situations where a chemical substance is observed neither to
persist nor to clearly degrade to innocuous inorganics and metabolites, the
investigator should be concerned with the question of persistent and
potentially toxic transformation products. In some cases, a good background
in chemistry and microbiology will enable the investigator to make informed
guesses about the potential products, but testing will be required to confirm
those guesses. The purposes of the degradation pathways testing are
twofold. One question to be answered is whether a toxic substance, when
degraded, shows concomitant loss of that toxicity. This is a combined
transformation and effects problem. The other question, which involves
analytical chemistry, deals with the identification of the products of
environmental transformation.
It is very important to us, in OTS, to be able to do better in assessing
the biodegradation rates of organic chemicals. This must be accomplished by
the use of a series of site-specific microcosms that have been shown to
replicate, to some necessary precision, the transformation rates in the
simulated environments. I hope that won't be necessary. It would be
preferable to be able to use some relatively simple, generic, bench-scale
laboratory procedures that would provide biodegradation rate constants that
could be translated into rates for specific environments by mathematical
manipulations that account for the relevant environmental factors that will
influence the rates. We hope that this workshop will consider the
possibilities and provide recommendations on the best research to pursue over
the next few years to achieve that goal.
-------
TABLF I
CHEMICAL FATE TESTING SCHEME
Data Source
Property or Process
to
.P
(0
D
Q) -p
,H W
ra QJ
-p -p
•H -H
U i4
Physical/Chemical Properties
Absorption, uv/visible spectra
Boiling Temperature
Density/Relative Density
Dissociation Constants in Water
Henry's Law Constant
Melting Temperature
Particle Size Distribution/Fiber Lngth.
Partition Coefficient (n-Octanol/Water)
pH of Water Soln or Suspension
Water Solubility
Vapor Pressure
Transport Processes
Evaporation from Water
Evaporation from Soil
Adsorption/Desorption
Uptake by Biota
Transformation
Biodegradation
Photodegdn.-Sunlight Photolysis
-Oxidn by OH radical or 0-,
-Indirect
Complex Formation
Hydrolysis
S-l
o
w
c
o
•H 03
-P Q)
(0 -P
^ 6
O -H
<-\ 4J
fO 03
U W
4J
0)
W
I cn
Q> -P
03 U3
(ti (U
CQ H
-y
«
o —
•r-t U3
•<-> Q)
Q> 4-1
C fi
•H «
<1) (0
Q (X
C o^
O £
•H \n
4J O
(0 U
rH O
3 >-i
e o
•H -rH
W 2
0) .C
-p -p
•H -H
Inappropriate and unnecessary data elements are indicated
by X marks within the boxes.
* Methods in the August, 1982 et seq. OTS Guidelines,
published by NTIS.
-------
E. AD-HOC ADVISORY COMMITTEE
Al W. Bourquin, Co-chairman
U.S. Environmental Protection Agency
Environmental Research Laboratory
Sabine Island
Gulf Breeze, FL 32561
(904) 932-5311
Hap Pritchard, Co-chairman
U.S. Environmental Protection Agency
Environmental Research Laboratory
Sabine Island
Gulf Breeze, FL 32561
(904) 932-5311
Donald Ahearn
Goergia State University
Research Office
131 Sparks Hall
Atlanta, GA 30303
(404) 658-4350
Martin Alexander
Department of Agronomy
Cornell University
Ithaca, NY 14853
(607) 256-3267
Robert Boethling
U.S. Environmental Protection Agency
Office of Pesticides and Toxic
Substances (TS-798)
Washington, DC 20460
(202) 382-3913
Peter Chapman
Department of Biochemistry
University of Minnesota
St. Paul, MN 55108
(612) 373-1303
Robert Hodson
Department of Microbiology
University of Georgia
Athens, GA 30602
(404) 542-1434
Harvey Holm
U.S. Environmental Protection Agency
Environmental Research Laboratory
College Station Road
Athens, GA 30601
(404) 250-3103
Carol Litchfield
Haskell Laboratories
DuPont Corporation
Elkton Road
Newark, DE 19711
(302) 366-5486
Jim Spain
Georgia State University
Cooperative Agreement Office
Environmental Research Laboratory
Sabine Island
Gulf Breeze, FL 32561
(904) 932-5311
William Walker
Gulf Coast Research Laboratory
Coast Beach Road
Ocean Springs, MS 39564
(601) 875-2244
-------
STATISTICAL AND EXPERIMENTAL REQUIREMENTS FOR MODELING DECAY CURVES
Objective: To discuss the limitations and possibilities for
statistically verifying that a kinetic expression
appropriately describes a decay curve or set of decay curves.
Biodegradation rates are commonly determined from decay curves assessed
by using certain statistical treatments of the data. Improper or insufficient
use of statistical methods and incomplete data sets, however, frequently
results in equivocal assessments of degradation rates. What statistical
analysis is necessary to verify that a kinetic expression appropriately
describes a decay curve? What experimental criteria (replication, sampling,
data points, precision, time course, etc.) should be established to assure the
completeness of data sets used for statistical analysis? When should
nonlinear or linear regressions be applied, and what advantages do they
offer? At what point does the number of outlying data points mean that a
model does not fit the data? What logic and discriminative framework should
be used to select models that fit the experimental data?
Panel Members: Joe Robinson (chairperson), Montana State University
David Currie, McGill University
Walter Maier, University of Minnesota
-------
PROBLEMS IN NON-LINEAR PARAMETER ESTIMATION
David Currie
Introduction
In recent years the use of nonlinear models has become increasingly
common in biological studies. The statistical treatment of such models
presents several difficulties which are not encountered in ordinary linear
regression analysis. As a result, uncautious treatment of nonlinear models
can often produce very poor parameter estimates and model predictions. It
is the purpose of this paper to describe three pitfalls in fitting nonlinear
models which are often observed in biological studies. I shall concentrate
specifically on the problem of obtaining the best estimates of the model
parameters (i.e., constants). A good, readable introduction to the basics of
regression analysis and some of its other potential problems is given by
Gujarati (1978) and more advanced treatments will be found in Draper and
Smith (1981) and Beck and Arnold (1977).
In seeking the "best" estimates of model parameters, I specifically mean
estimates which are both accurate (i.e., unbiased) and precise (i.e., having
low variance). The precision of estimates almost entirely determines one's
ability to test hypothesis. Imprecision undermines the ability to reject the
null hypothesis and increase the probability of retaining the null hypothesis
when it is in fact false (Type II error). More insidious yet is the effect
of inaccuracy (i.e., bias) estimation. Bias is the systematic tendency to
overestimate or underestimate a population mean value, and can cause differences
among hypothesis to appear to be statistically significant, when they are in
fact entirely spurious. In the discussion which follows, I shall consider
three common practices which tend to yield inaccurate and imprecise parameter
estimates: poor experimental design, inappropriate fitting techniques, and
inappropriate model choice.
POOR EXPERIMENTAL DESIGN
The design matrix, that collection of values of the independent variable
(x) at which the experimenter decides to take observations, is rarely given
much explicit consideration before an experiment is carried out. It can be
shown (Draper and Smith 1981) that, for any model, the variance of parameter
estimates will depend upon two things: the magnitude of the experimental
error (which the experimenter often cannot control) and the design matrix
(which usually can be controlled).
The effect of the design matrix is most easily illustrated with a linear
relationship (Figure 1). If one wishes to estimate the slope of a straight
line (the heavy line in the center of the figure) with a given amount of
experimental error around the line (depicted by the vertical bars), then
taking observations over a relatively narrow range of x values (interval A)
-------
will lead to a much broader confidence interval around the estimate of the
slope than will a wider spread of observations (interval B).
The situation is consideraly less obvious when a model is nonlinear.
This is due to the fact that estimates of nonlinear parameters depend upon the
placement of observations with respect to specific points on the curve.
Consider, for example, a simple exponential decay model:
where y is usually the measured amount of a decaying substance, and x is an
independent variable such as time or distance. Suppose that a, the initial
quantity of substance y, is known or fixed by the experimenter, and one wishes
to estimate the decay constant b. To what extent does the choice of values of
x influence the estimate of b?
Figure 2 depicts this decay curve and offers several conceivable
arrangements of values of x at which to measure y. The variance of the
estimate of b [i.e., var(b)] which results from each of these designs is shown
in Figure 2 as a multiple of the experimental error o . Var(b) is seen to
change by a factor of three over the possible designs shown here. Intuitively
appealing designs that spread the observation over a wide range of x values
yield estimates of b which are 50-100% more variable than the estimate found
by taking all of the observations at x = 1/b. It can be shown that this i.s,
in fact, the design which minimizes the variance* of b.
Each observation that one takes contributes a specifiable amount of
information toward estimating b. The amount of information conveyed by a
given observation y, depends entirely upon the value of x^ at which y^ is
measured. This relationship is shown in Figure 3. In brief, for the
exponential function, the most information is contained in an observation
taken at x = 1/b. The farther from x = 1/b an observation is taken, the less
information it contributes toward estimating b.
In general, minimum variance estimates of the parameters of any fairly
simple model will be obtained by concentrating all of the observations at a
number of specific points equal to the number of parameters in the model. In
many cases (such as a straight line), one or more of the optimum observation
points will be at extremes of the range of x. The exact points may be
determined by optimization of the variance-covan'ance matrix (Draper and Smith
1981 discuss the variance-covariance matrix, and Box and Lucas 1959 deal with
the optimization thereof), but it is not easy going. Alternatively, one may
use an approximate method to find these points as outlined in Appendix 1 and
discussed by Beck and Arnold (1977).
The obvious drawback of optimal designs is that they provide no means of
checking the fit of one's model to the data. In practice, it is often
desirable to sacrifice a certain amount of precision for the sake of
ascertaining that one's model is adequate. It is still possible in such cases
to evaluate the effect of the design matrix on parameter estimates. In
general, if observations are spread out in the neighborhood of the optimal
design points, a minimum of precision is sacrificed. For any given model, one
may generate a sensitivity relationship such as Figure 3 and use this
information to design one's experiments (or better still, use the covariance
matrix approach).
-------
Consider a second example, the familiar Michaelis-Menten model (or
alternatively, the formally equivalent Monod function or Langmuir isotherm):
Vs
v =
K + s
where v is the rate of reaction, s is the substrate concentration, and V and K
are fitted constants (representing a maximum reaction velocity and a half-
saturation constant, respectively). The most precise parameter estimates are
obtained by taking half of the observations at s = K, and the other half at
s »K (i.e., as high as possible) (Currie 1982).
Figure 4 (taken from Currie 1982) appraises the effect of using geometric
sequences (4A and 4B) and linear sequences (4C and 40) of seven observations
to estimate K and V. The linear sequences of observations s^ are defined by:
s1 = i_a_ for i = 1, 2,..., 7
(for example, when a = 0.5, the observations were taken at x values of 0.5,
1.0, 1.5, 2.0, 2.5, ~3~.0, 3.5) and the geometric sequences by:
s = a_ 21"1 for i = 1, 2,..., 7
(for a_ = 0.5, the observations were taken at 0.5, 1.0, 2.0, 4.0, 8.0, 16.0,
32.0). In essence, _a_ is a factor which serves to telescope the observations
such that when a_ is small, the observations are concentrated nearer the
origin, and when _a_ is large, the observations are more spread out.
The lower panels of Figure 4 (4B and 4D) show the variance of estimates
of K (i.e., K) as a function of the design matrix as characterized by a_. It
is immediately evident that the linear sequences of observations in FigUre 40
yield much higher variance estimates of K than do the geometric sequences in
Figure 4B. Moreover, among each of these classes there are obvious minima in
the variance of K which are obtained. The variance of parameter estimates
using the best geometric sequence of observations is approximately twice as
great as that obtained with the optimal design. However, the geometric spread
of observations does allow one to verify the fit of the model while still
producing reasonably precise parameter estimates. The catch, of course, is
that one must have some preliminary estimate of K in order to design the
experiment (similarly, for the exponential decay model, it is necessary to
have some preliminary estimate of b). Given this information (usually from an
exploratory experiment), one may then optimize the experimental design.
Not only is the precision of parameter estimates affected by the
experimental design, the accuracy is affected as well. In Figure 4 panels A
and C, the ratio of the estimated half-saturation constant (K) to its true
value (K) is shown. This is a measure of bias: a ratio of 1.0 indicates an
unbiased estimate of K, whereas deviations from 1.0 indicate less and less
accurate estimates. It is evident that some sequences of observations yield
badly biased estimates of K. The geometric sequences of observations (4A) are
seen to yield estimates of K which are also more accurate than those from the
linear sequences (4C).
The effect of experimental design on bias in parameter estimates is more
difficult to determine than its effect on variance; in the present case, it
-------
was determined by Monte Carlo simulations, which are fairly involved. As a
rule of thumb (which may not invariably be correct), it appears that the more
precise parameter estimates also tend to be reasonably accurate.
INAPPROPRIATE FITTING TECHNIQUE
The most common means of fitting nonlinear models is to apply a
transformation which linearizes the relationship and then to fit the data by
the ordinary least-squares method. Generally speaking, this is
inappropriate. Least-squares regression involves the assumption that the
error in the dependent variable is normally distributed and of equal magnitude
at all x. When a dependent variable is transformed, the error around that
variable is also transformed and, depending upon the original distribution,
the final error distribution will often be something very far from normal
which varies with x.
Figure 5 shows the bias and variance of estimates of K in the Michealis-
Menten model using the same data as in Figure 4. However, in the case of
Figure 4, the untransformed data were fitted with a direct, nonlinear least-
squares method (Bliss and James 1966; Hanson e^aj_. 1967), whereas in Figure 5
the data were subjected to a Woolf linearizing transformation first, and were
then fitted by ordinary linear least-squares. A comparison of the two figures
reveals that the Woolf transformation produces parameter estimates which are
strongly biased and which have very much higher variance than those found with
the untransformed procedure (Figure 4). The Woolf transformation has been
compared to other linearizing transformations of the Michaelis-Menten
function, and it has been shown to be the best of the possibilities (Dowd and
Riggs 1965; Atkins and Nimmo 1975). As a rule, linearizing transformations
will yield high biased, high variance parameter estimates, whereas direct
untransformed fits of nonlinear functions will be much more reliable.
Nonlinear fitting routines are now available in virtually all of the major
main-frame statistical packages (e.g., NLIN in SAS, NONLINEAR in SPSS).
Linearizing transformations may be appropriate in cases where a transformation
is applied in order to normalize the error distribution. Obviously, the
rationale here is that attention must be paid to the assumptions of one's
fitting technique.
INAPPROPRIATE MODEL
A third common practice which leads to poor parameter estimates is that
of fitting an inappropriate function for the sake of ease of computation. A
frequent example is the case in which a straight line is fitted to something
called "the initial linear portion" or the "first-order portion" of a
nonlinear function. In most cases, theoretical considerations or examination
of the residuals around the fitted function (e.g., Draper and Smith 1981)
should be sufficient to show that there is no initial linear portion.
Application of a linear model to an "approximately linear" curve will
invariably lead to biased parameter estimates.
As an example, consider^the data shown in Figure 6, which show the time
course of incorporation of P orthophosphate in a bacterial culture. The
goal of this experiment was to estimate the rate of orthophosphate
10
-------
incorporation, which is known to decline as orthophosphate is depleted from
solution. A very common procedure would be to fit a straight line to an
initial "linear" portion of the curve and to take its slope as the rate of
phosphorus incorporation. On cursory examination of the data, the assumption
of initial linearity does not seem unreasonable. However, it is unclear
exactly what portion of the curve should be considered as linear. The numbers
in the upper left portion of the figure represent estimates of the slopes of
straight lines fitted to progressively shorter initial segments of the
curve. It can be seen that as the latest observations are sequentially
eliminated from consideration, the estimated slope of the "initial linear
portion" steadily increases. This is due to the fact that indeed there is no
linear portion of the curve. As a result, the estimated rate of
orthophosphate uptake is systematically underestimated, to an extent which
depends upon the length of the period of observation.
It is evident that the initial rate of uptake is the biologically
interesting parameter in this example. The solution to this fitting
difficulty is to choose a function which eliminates the pattern in the
residuals, and which is therefore insensitive to the duration of the
experiment. A second degree polynomial was found to fulfill these criteria,
and yielded the rate estimate in the lower right portion of the figure. This
rate represents an unbiased estimate of the initial rate of orthophosphate
uptake.
It is worth mentioning in passing that the straight line fits these data
very well, as judged by the correlation coefficient or an F-test. Moreover,
the second degree term of the polynomial fit was never significant according
to a partial-F test. The second degree term was nonetheless retained in order
to satisfy the assumption (of any regression method) that the error around the
function is unbiased (i.e., that the expected value of the residuals is
zero). Failure to satisfy this assumption is evidence of an inappropriate
model, which causes parameter estimates to be biased.
SUMMARY
Fitting nonlinear models to experimental data poses a number of unique
statistical problems in addition to those encountered in ordinary linear
regression. A large number of biological studies involving nonlinear models
make fundamental statistical errors which cause the parameter estimates and
model predictions to be biased and highly variable. Many of these errors
result from neglect of the basic assumptions of the fitting techniques used.
Three especially common inappropriate procedures are: (1) neglect of the
experimental design matrix (that set of values of the independent variable at
which the experimenter determines to measure the dependent variable);
(2) fitting data which have been transformed to remove nonlinearity without
consideration of the effect on the error distribution; and (3) fitting the
wrong function for the sake of ease of computation (e.g., fitting a straight
line to the "initial linear portion" of a nonlinear function). Statistically
appropriate alternatives to these practices are readily available.
11
-------
LITERATURE CITED
1. Atkins, G. I. and I. A. Nimmo. 1975. A comparison of seven methods for
fitting the Michaelis-Menten equation. Biochemical Journal 149:775-777.
2. Beck, J. V. and K. J. Arnold. 1977. Parameter Estimation in Engineering
and Science. New York: Wiley. 501 pp.
3. Bliss, C. I. and A. T. James. 1966. Fitting the rectangular hyperbola.
Biometrics 22:573-602.
4. Box, G. E. P. and H. L. Lucas. 1959. Design of experiments in nonlinear
situations. Biometrika 46:77-90.
5. Currie, D. J. 1982. Estimating Michaelis-Menten parameters: Bias,
variance and experimental design. Biometrics 38:907-919.
6. Dowd, J. E. and D. S. Riggs. 1965. A comparison of estimates of
Michaelis-Menten kinetic constants from various linear
transformations. Journal of Biological Chemistry 240:863-869.
7. Draper, N. and H. Smith. 1981. Applied Regression Analysis (Second
edition). New York: Wiley. 709 pp.
8. Gujarati, D. 1978. Basic Econometrics. New York: McGraw Hill. 462 pp.
9. Hanson, K. R., R. Lung, and E. Haver 1967. A computer program for
fitting data to the Michaelis-Menten function. Biochemical and
Biophysical Research Communications 29:194-197.
APPENDIX
Consider a model in which y is a function of x and a parameter 8:
y = f(x, G )
The variance of the estimated parameter 0 is approximated by:
7 n 7
var(9) = a / \ b
1=1
where
A 5f
A -69
2 2
and a is the error variance. Therefore A evaluated at a given value of x is
a measure of the amount of information that that observation contributes
toward estimating the model parameter (Beck and Arnold 1977).
12
-------
When a model involves two or more parameters, the situation becomes more
complex since each observation contributes information to the estimation of
every parameter. An exact solution to the problem of designing the experiment
is given by Box and Lucas (1959; or Appendix I in Currie 1982). However, in
most cases the method given above, carried out for each parameter in the
model, gives an adequate indication of how to design an experiment.
13
-------
B
Figure 1. h schematic representation of the confidence interval
around the estimate of the siooe of a straight line, when
observations are taken over a narrow rannp (,\) versus a
broader ranqe (B).
14
-------
c
d
e
—» 2.86 a'
-*- 2.28 a'
2.50 a:
1.86 a
1.48 a2
4.10 a'
.-bx
0
4/b
Figure 2. The effect of the experimental design on
the precision of estimation of the decay constant b.
Each array of diamonds (a-f) represents a series of five
observations spread in various ways across the range
of x. The resulting variance of the estimate of b
is shown as a multiple of the error variance of a2.
15
-------
1.0 i
0.75-
O)
CD
_CC
CD
oc
0.50
0.25
y. = ae
3 1
-bX:
0
1/b
2/b
3/b
X-
Figure 3. Tne relative amount of information which
an observation y. taken at a given x. contribute
to the estimate of b.
16
-------
14
o
>
1-0
0-75
050
025 -
B
0
0-5
1-0 1-5
a
0 0-5 1-0 1-5
a
Figure 1. The bias (K/K) and variance (var(K)) of estimates
of the Michaelis-Menten K, as a function of the design matrix.
Panels A and B represent the effect of geometric sequences of
observations; panels C and D show the effect of linear
sequences. In both cases, a_ serves to telescope out the observations
such that then a is small, observations are concentrated near
the origin, and when a_ is large, observations are more spread out.
A nonlinear fitting technique was used. See text.
17
-------
20-
1-5
1-0
2-0
o
V.
0
0-5 10
a
-i r
\
0 0-5
1-0
a
—I—
1-5
Figure 5. Same as Figure 4, using a linear fit to Woolf transformed
data.
18
-------
% PARTICULATE
•a c r+ -s
O "O O fD
3
O
3
7T -5
fD O
CO
O -5
tt>
to
QJ C
—' -5
< co fD
cr ro -"•
-J. . < fD
c+ fD CO
a>
cr
ii -i-
O 3
b H-'
CO -"•
CO D»
o
T3
ro
a>
CO fl)
3- <-t-
O fD
-J Q-
rt-
fD to
O
-J.T3
3 fD
Q.
fD
fD
-S
CO
CO (-f-
fD T
LQ Ql
CO
3 3"
r+ rt-
oo
O -••
-h 3
fD
CU CO
Ol
0 O
1 1 1
o "
CO
m
§ "-
0 o
^L.
O
CO
CO
0 '
^ifc.
•"? M> j o- i cr
J II II
o Jo
o b
CO CO
> > ro ro .
00 o>
, »
cr
II
p
b
CO
CO
(O
» >
o
o
cr
II
o
b
ro
O)
-
cr
II
o
b
ro
a>
»• .
cr
II
0
b
_k.
00
cu "0
fD
Q.
-------
EXPERIMENTAL REQUIREMENTS FOR MODELING DECAY CURVES
W. J. Maier
ABSTRACT
The objective in analyzing and correlating data on the decay
of organic substrates is to develop mathematical models that can
be used to characterize rates of decay. Mathematical models are
needed to extrapolate the available information to different
environments or to predict the time required to reach some
predetermined concentration.
Mathematical curve fitting per se is not adequate for this
purpose if the physical-chemical-biological significance of the
coefficients has not been defined. A prerequisite for a useful
model is that the equations' parameters correspond to the rate
controlling reactions. The parameters can be adjusted to reflect
different environmental conditions.
UNSTRUCTURED MODELS
Biochemically mediated decay of organics stems from the metabolic
processes of the microflora. Rates are closely related to the concentration
of organisms as an important description even if the microorganism
concentrations remain constant with time. The framework for modeling is given
by:
ds _ px
dt y
(1)
dt
- k .x
(2)
where s = substrate concentration
x = concentration of microorganisms
u = growth rate coefficient
y = yield coefficient
K. = endogenous decay coefficient
For those situations where the decay rate is mediated by extracellular enzymes
(Eg), it may be necessary to relate its concentration to x. The parameters u
ana y must be defined in terms of the important process variables.
A number of alternate expressions have been proposed for y and y as shown
in Table 1. The most appropriate expression to use for a specific application
depends on the range of substrate concentration of the decay curve and the
presence of other competing organic substances.
20
-------
TABLE 1
Alternate Expressions Proposed for
the Growth Rate (y) and yield (y) Coefficients
y = y
m
Exponential growth, unlimited
y =
VHT
Monod growth equation
y =
y =
ym
K
s
S
+ S + S2
Ki
y S
m
Ki
f(Sj)
Monod equation with substrate
inhibition coefficient Ki
Monod equation with terms to
account for substrate inhibition
and alternate substrate effects
due to Sj substrates
u =
K
Mm, S
In n
s,n
V + f(SJ)
Summation of growth rates
represents concurrent utilization
of alternate substrates i
Y = constant
r- Y
1 + kd/y
Y = f(T)
Yield coefficient is dependent on
rates of endogenous decay and
growth
Yield coefficient is temperature
dependent
* y is a function of temperature, pH, and ionic strength
The first two relationships in Table 1 for p are widely used. The advantage
of the simpler models is that the resulting equation can be linearized or
solved explicitly. Cell mass concentration can be expressed in terms of
equivalent substrate concentration, thereby eliminating it as a dependent
variable. This allows calculating the numerical values of y and K from
measurements of substrate decay curves providing an independent measure of
initial cell mass concentration is available.
Solution of the third equation in Table 1 which incorporates the effect
of substrate inhibition requires making some simplifying assumptions. K can
be evaluated from data at low concentration by neglecting the Ki term and
conversely, Ki can be evaluated from high substrate concentration data. It
is, however, necessary to factor in the time rate of change of cell mass
concentration. Application of the last two relationships for growth rate
21
-------
expressions in Table 1 requires the concurrent analysis of several sets of
decay curves to define the interaction effects, namely, inhibition by
competing substrate and growth enhancement by concurrent substrate
utilization. The equations cannot be solved explicitly. One approach is to
correlate the data using the differential forms and cross correlate to
describe the effects of alternate substrates. Cell mass concentration is a
variable and must be factored into the analysis.
It is important to note that cell mass concentration (x) is a time
dependent variable regardless of the relationship used to define y. In some
situations (when the initial cell mass concentration is large compared to the
rate of additional cell production), it is acceptable to treat x as a
constant, thereby simplifying the data analysis procedures. It is, however,
necessary to obtain an independent measurement of the initial active cell mass
concentration in order to characterize the substrate decay curve or to
calculate the numerical value of y from a decay curve.
In the more general situation where cell mass concentration changes with
time, numerical integration of equations 1 and 2 can be used to simulate
substrate decay curves.
Computerized subroutines for solving coupled differential equations are
readily available. They can be used to predict decay curves from known values
of y , K , k ,, Y and they can be used to estimate the values of these
parameters from measured substrate decay curves. Parameter estimation
subroutines are available to calculate the best fit values for simple decay
curves as well as for a sequence of decay curves starting with different
initial concentration of cell mass or substrate.
Examples of the use of numerical analysis of laboratory test data are
illustrated in the following figures.
Figure 1 shows decay curves of 2,4-D using acclimated inoculum at two
different initial concentrations. The data (points) from both test runs were
fitted by nesting the data. The kinetic parameters were estimated using a
fitting subroutine.
Figure 2 shows 2,4-D decay curves and attendent increase in cell mass.
Initial cell mass concentration was calculated using fitting subroutine.
Figure 3 shows an application of parameter estimation to fit 3,5-DCB
decay curve. Lines illustrate the effects of changing the numerical values of
the parameters y , K$, XQ. Best fit calculations become progressively more
time consuming arVti less reliable for estimating multiple parameters.
Figure 4 shows measured decay curves of 2,4-D with and without addition
of alternate substrate (nutrient broth). 2,4-D acclimated inoculum was added
at time zero. Filtered inoculum had lower active cell mass and much lower
protozoa concentrations.
Figure 5 shows measured decay of 2,4-D (same as Figure 4a-4b) and model
calculations of substrate decay curves. Model calculations incorporate
effects of substrate inhibition, substrate interaction, and concurrent
substrate utilization.
22
-------
SEMI-STRUCTURED MODELS
A major drawback of unstructured models is that they are limited to
describing the kinetics of balanced growth, sometimes referred to as steady
state growth kinetics. Time dependent growth phenomena have been described in
a variety of situations. The most obvious cases occur when a new substrate
that requires acclimation of enzyme systems is suddenly made available.
Recent advances in modeling time dependent growth phenomena are largely
based on advances in understanding of the regulatory processes that control
enzyme production and activity. Recognition that some catabolic enzymes are
constitutive, e.g., formed at constant rates and amounts, while others are
inducible, e.g., affected by concentration of specific substrates or
metabolites, has been particularly helpful in explaining and modeling the
growth behavior of unacclimated cultures. Changes in growth rate are seen to
be the result of changes in enzyme activity levels; different enzyme
controlled reactions can become the rate limiting step in the overall sequence
of growth related metabolism; hence, the need for a structured model.
However, it has also become evident that complete stepwise modeling of growth
processes is extremely complicated and probably not feasible for routine
use. As a result, model development has focused on semi-structured models as
a compromise between rigor and practicality. Semi-structured models have
found favor because they can be used to describe time-dependent growth
phenomena without exceeding practical limits on the number of rate
coefficients, equilibrium constants, and rate equations needed to describe the
system.
One version of a two-enzyme model has been described to illustrate its
application for analyzing and correlating lag phase behavior. The model
considers two rate controlling steps, namely, substrate transport by permeases
followed by intracellular metabolism. The data represent the observed lag
phase when a 2,4-D acclimated culture is challenged with mixtures of 2,4-D and
glucose. The model assumes that transport and conversion of 2,4-D and glucose
are mediated by their respective permease enzyme systems E-^ and E^; P} and
?2 are internally available intermediates and are used for growth and energy
metabolism by their respective enzyme systems, E2} and Ep^. The presence of
P-^ and P~ can affect permease enzyme production or activity; repression of
permease fay the other intermediate was assumed. The model is schematized in
Table 2.
The corresponding mathematical formulations for calculating the time-
concentration changes of the two substrates (Sj and S-), their corresponding
intracellular products (P^ and P-), enzymes (E,,, E,p, iLi, E-p) and overall
cell mass formation (X) subject to material balance considerations are
summarized in Table 3a; the symbols are defined in Table 3b.
Published information on kinetic coefficients for transport enzymes and
overall catabolic enzyme systems was used. Test data using single substrates
were used to calculate coefficients for the acclimated culture. The system of
differential equations was solved by numerical integration using a Runge-Kutta
subroutine available through the University of Minnesota Computer Center. The
initial values of substrate and cell mass are measured.
23
-------
TABLE 2
2-enzyme model, concurrent substrate utilization
Version II - Different intracellular intermediates
Case I - Glucose permease repressed by 2,4-D intermediate
2,4-D permease repressed by glucose intermediate
Uptake:
1?
?2
- -
Growth
The model:
Uptake:
!2
1?
Lf-
c
12 2
c + P
C12 2
Growth:
F + p
L21 P
j. p
V
'
1?
ie-
3?
•
21
24
-------
TABLE 3a
Rate Equations
[Ell.TOT][SlJ
dt ~ "31 S7 k + S
X o -i
dt
[EH.TOT][S1] . el
" K
dt ' 31 bl k$ + Sx " 31 f1 kp vX
>
" K " K
"dT" " K32 b2 k$ + S2 " K32 f2 kp vX
dX el [E?1 Tni][Pl] e? ^E?? TOT^PO]
5£ = k ' v — dL> IUI i + \e ' v -A ^2, TUT 2 ,
dt K31yP1 fj kp vX + P! + K32yP2 f2 kp vX -HT^ "
dEHJOT _ Cl dX
dt . -. Dm , cTf
dl + ]1P2 X
dE12,TQT = C2 dX
dt . . , m0 "dT
d2 + I
dE21,TOT dX
dt ' Bl dt
dE22,TOT _ ^X
dt ^2 dt
25
-------
TABLE 3b
Explanations and Symbols
Sj: glucose concentration, mg/1
S2' 2,4-D concentration, mg/1
Pj_: glucose Intracellular intermediate, mg/1
P2: 2,4-D intracellular intermediate, mg/1
X: biomass, mg/1
Eij JQ-J-: total glucose transport enzyme, mg/1
Ej2 JOT: total 2,4-D transport enzyme, mg/1
Eoj YQT: total glucose Intermediate metabolic enzyme, mg/1
E22 TOT: total 2,4-D Intermediate metabolic enzyme, mg/1
ko : Michaelis constant for S^, 1=1,2; mg/1
31 '
kp : Michaelis constant for P.., 1=1,2; mg/1
K3.j, k' .: enzyme rate constants, 1=1,2; hrs~*
kd: biomass death rate coefficient, hrs"1
yp : yield coefficient, mg biomass/mg P^; 1=1,2
a,, bi, e^, f ^: molecular weights for S^, E^., P^, E2l-
v: specific volume of biomaterial, 1/mg
26
-------
MULTIPLE COMPONENT MODELS AND STEPWISE DEGRADATION OF COMPLEX SUBSTRATES
Stepwlse degradation of complex substrates can Involve one or more
species of microorganisms acting concurrently. Oxidation of ammonia to
nitrite and nitrate by nitrosommas is a classical example of a two species
process. Measurements of monochlorobiphenyl degradation show evidence of
sequential blodegratlon.
Models that explicitly simulate substrate utilization with formation and
utilization of intermediates by independently acting microorganism populations
have been defined. These models have the same form as the single substrate
models described in previous sections. The only difference is that the
Inventory of each microorganism population is calculated using a series of
coupled differential equations. The availability of the second step substrate
(product from first step) is dependent on the Sites of utilization of both
species.
The principal criterion for determining whether it is necessary to use a
muIt1 component model 1s the formation and accumulation of product
Intermediates. If the measured concentrations of intermediates are negligibly
small, 1t 1s reasonable to use a single culture model. If intermediate
products tend to accumulate at a slow rate, the use of a multicomponent model
is recommended.
PROTOZOA PREDATION EFFECTS
The role of protozoa 1n controlling bacterial populations has been
recognized and documented by a number of investigators in wastewater
treatment. For example, the rate of 2,4-D utilization with 3 micron filtered
Inoculum resulted in higher maximum rates of substrate removal. This has been
ascribed to the effect of removing protozoa thereby reducing predation
effects. Figure 7 illustrates the phenomenology for 2,4-D utilization and
attendent population dynamics of bacteria and protozoa.
Protozoa predation has usually been modeled indirectly by introducing the
endogenous decay term kd shown in equation 2. This approach appears to be
satisfactory for analyzing continuous flow reactor systems. However, it is
not adequate for simulating the phenomenology described in Figure 7 which
represents batch test data and is typical of shock loading situations.
Explicit modeling of protozoa predation has been defined. A prey
predator model has been used to simulate the data in Figure 7. Kinetic
coefficients that describe the growth of protozoa and bacterial decay were
calculated and give a reasonable fit of the data.
The methods for measuring microbial concentrations (bacteria and
protozoa) are time consuming and relatively few studies have been published.
As a result, 1t has not been possible to validate prey predator models
adequately. Additional test data that are representative of conditions in
natural bodies of water are needed to define the potential importance of
protozoa and other bacterial dieoff mechanisms that could affect population
dynamics.
27
-------
SUMMARY
It 1s recommended that the measurement and modeling of substrate decay
phenomena be broadened to include characterization of the accompanying changes
in active cell mass or enzyme concentrations. Quantitative descriptions of
cell mass are essential in order to obtain precise information on rates and
allow extrapolation of data to other environments.
The form of the rate equation that is best suited for analyzing test data
depends on the complexity of the environmental system to be characterized.
Substrate inhibition effects must be incorporated at elevated concentrations;
the critical concentrations vary considerably from 100+ milligrams per liter
to 300+ micrograms per liter for 2,4-D and PCP? respectively. The effects of
alternate substrates can be an important factor in regard to the utilization
of the target substrate and in enhancing the proliferation of active cells.
The effects of alternate substrates should be considered at all concentration
levels. This also implies that the decay curves of alternate substrate should
be measured concurrently with target substrate decay rates.
It is further recommended that semi-structured models be used for
analyzing the correlating data that represent transients or acclimation
periods. The greater complexity of these models is compensated for by the
advantages that accrue from more detailed mechanistic descriptions of the rate
controlling steps. For example, the semi-structured modeling approach can be
used to simulate the effects of specific enzyme systems thereby giving insight
into mechanisms. Simulation studies can also be used to specify enzyme
activity measurements that are needed to describe the transients or
acclimation process.
28
-------
CONCENTRATION, mg/l
CQ
C
-5
n
rr
o
o
T3
n>
3
O
X
<<
Cu
O
n>
rf
o
14
Q.
Q.
O
X ?^
o ex
7; -< r:
~s. • 3
35
H 3—33--
co \ n i -^ -^J"
n
c
-5
<
CD
O O
O O O i— K^
O 00 ho
M OC — I
-------
CO
o
C
-5
(D
3 n
Q. -i-
rt> O
3 3"
0
0 -5
ro o
— ' 3"
n>
3 3
CL> O
I/) X
co <<
OJ
o o
o o>
3 c~*"
O -J.
(D O
3
<-+ CD
-5 O
H
2
m
^
5
c
^•D
CO
QJ
rt Q-
O Q-
3 fD
(/i O
• CD
O
c
-5
O
H-
• * SI
S,-2,4-D CONCENTRATION, mg/l
X, = ACTIVE CELL MASS CONCENTRATION, _mg/I
00
*
o
—t—
0= X
o
-
O
O
-------
0.0
25
3,5-DCB TEST DATA
MODEL CALCULATIONS
\
\C
\
0.412 0.257 0.257 0.405
\ Ks 24'6 12.7 24.6 24.6
0.000218
50
75 100 125
TIME, HOURS
150 175
Figure 3. 3,5 Dichlorobenzoate decay curves.
-------
CONCENTRATION, mg/I
CONCENTRATION, mg/
CO
Q> IQ
13 C
O- -S
ro
—' c:
o -s
3 fD
n>
CU Q.
3 •
Q.
s; ro
3- I
o
3
c: cr
c4- -J.
-S O
_.. CL
3
H
m
i
CO
S
o
8-
§•
M -
o
r/
f
il
%
£J
£-
» pt
*Ul
5"n c ~n
,S?q
5sS»
s«s°
SiEU
„ 3 5
IzSg
(D /
01 00
ro +» O) uu (-, rvj -»»
OoOOOOOO
Of)
CU U3
3 C
Q. -S
ro
<-+ Cu
c-l- "
-S -f*
— '• I
n> o
3
(-1- cr
cr o
-s Q-
O CD
-s
EU
O-
Cu
rt
Cu
O
3
ro
—i C
o -s
Co O
3 •
Q.
3
c cr
<-+ -••
-s o
_•. o.
n> rt>
3 U3
rt- -s
o>
cr O.
-S Cu
O r+
<-t- -".
rr O
• 3
-------
2,4-D CONCENTRATION, mg/l
2,4-D CONCENTRATION, mg/l
CO
oo
c
-s
fD
O
O
O
3
O
c
-s
ro
3
Q)
O
3
O
-h
no
o>
3
Q.
fD
13
CO
O
t-h
3"
O N
b oi
b en
1 i —
v_>
6 r
NUTRENT BROTH, mg/l_
ro 01 ->i O
Ot O 01 O
ro -
O
S
o
t-o
Ui
K>
m
o
o
o
-p-
r-o
c
o
o
hO
o >—
^O t-o
0 0
1 — ' K>
O ^-n
J> ^-J
O
C 0
1 — ' 1 — '
oo ro
CO — 1
0 0
.£> UJ
l-o
UD
c
ro
tn
cr
o
o
3
O
c
-s
CO
N
EU
rl-
o'
3
O
-h
ro
DJ
3
CL
-S
fD
CO
O
ISJ
O
CD
O
8
NUTRIENT BROTH, mg/l_
t— NJ
O l_n
J>- --J C
^ 00
-------
Figure 6 shows a comparison of the test data and model calculations for a
batch test with 2,4-D acclimated inoculum using a mixture of glucose and 2,4-
D. The latter is utilized immediately; glucose utilization starts after
approximately 20 hours. The 20-hour lag phase in glucose utilization,
followed by concurrent utilization, and enhanced overall rates of 2,4-D
utilization as cell mass concentrations increase, was correctly simulated.
The model predicts a gradual increase in glucose transport and catabolism
until a balanced enzyme activity distribution that describes concurrent
substrate utilization is achieved. The latter condition represents pseudo
balanced growth with concurrent utilization of both substrates. The excellent
agreement between data and model simulation is noteworthy because the findings
of these studies are in accord with basic research using defined systems.
34
-------
Tfgure 6
Model Calculations, Version II, Case I
A : Experimental data on glucose
Q : Experimental data on 2,4-0
80.
35
-------
CONCENTRATION — mg/l
GO
CTi
in
c
en
CT •
EU
n
r+ ro
fD >»
-5 -P»
-<• I
o> n
Q. !-•
O
•a o
rt-ia
o •>-,
03
o
a.
n>
IQ
cu
a.
cu
rt-
o'
=5
Cu
3
O-
c
cr
0)
•s
NUMBERS OF BACTERIA 8 PROTOZOA
-------
PANEL REPORTS
Our recommendations on statistical techniques to be applied to future
studies in biodegradation must be divided into two sets. One set of
recommendations applies to those individuals (Group I) who feel that
estimating biodegradation rate coefficients to within an order of magnitude of
their correct value is acceptable. This acceptance largely results from the
level of noise (error) in the data from which these parameters are
estimated. A different set of recommendations applies to the second group
(Group II), which is interested in more precise parameter estimates and
mechanistic model-bullding. Presumably, individuals in the latter group work
with data several-fold more precise than the data to which the first group has
access.
RECOMMENDATIONS FROM GROUP I
1. Never propose a model that is more complex than your data will
support. This becomes especially important when the level of error is
high (e.g., when the standard deviation of replicate measurements of
the dependent variable exceeds 10% of the average value of the
dependent variable).
2. If more complex models are to be used, then the addition of higher-
order terms (e.g., nonlinear terms) should be tested for by using an
F-test.
3. Use nonlinear parameter estimation (NPE) techniques to estimate
microbial rate coefficients, when possible, but recognize that
estimating these coefficients by using linearized forms of nonlinear
models probably will not produce disastrous results. Fitting the data
directly to a nonlinear model, rather than to a linearized form, can
buy the investigator protection even when the level of noise is high.
RECOMMENDATIONS FROM GROUP II
1. Given a data set, propose several competing models to represent the
underlying biological process. Differences among these models may be
initially assessed via simulation. The simulated decay curves may
subsequently be used to aid the design of an experimental program.
2. Estimate all parameters for the proposed models via NPE. The standard
errors of the parameter should also be approximated. Explicitly state
the updating method used to fit the data directly to the nonlinear
models. No single method for estimating the parameters of nonlinear
models is best and there are several algorithms available. The most
commonly used NPE method -- typically available on main-frame computer
systems — is the Levenberg Marquardt algorithm. The Gaussian method
is another popular NPE technique.
3. Select the "best" of the proposed models by using an F-test or similar
model-building test.
37
-------
4. Once the appropriate model has been selected, additional experiments
may be run to obtain even more precise estimates of the microbial rate
coefficients. In these cases, optimal design criteria should be used
to select the range of measurement of the dependent variable. This
will ensure that the variance of the estimated parameters is
minimized.
As a simple example, assume that the first-order model has been
found to best represent a given data set by using an F-test. Now
additional experiments are to be run with the goal of obtaining more
precise estimates of K, the first-order decay coefficient. The
optimal design for these experiments is to center the measurements of
the dependent variable (presumably, substrate concentration) on the
point at which the dependent variable is 37% of its starting value.
Of course, data obtained for the optimally designed experiments should
be analyzed by using an NPE method, again stating the technique of
choice.
5. Robust regression (e.g., bi-weight regression) may be used to minimize
the influence of outliers when data are fitted to a particular
nonlinear model, but only when it is known a priori that the data are
best represented by the chosen nonlinear mode~"L
GENERAL REFERENCES
The following general references may be used for details on fitting
nonlinear models to data. Further, some of these references provide
information on optimally designing experiments for parameter estimation. The
textbook by Beck and Arnold pays a great deal of attention to the statistical
assumptions necessary for the correct application of least-squares analysis.
1. Beck and Arnold (1977). Parameter Estimation in Engineering and
Science. John Wiley & Sons, Inc.
2. Draper and Smith (1981). Applied Regression Analysis, 2nd ed. John Wiley
& Sons, Inc.
3. Wonnacott and Wonnacott (1971). Econometrics. John Wiley & Sons, Inc.
THE "SECOND ORDER" APPROACH:
ASSUMPTIONS, LIMITATIONS, AND RESEARCH NEEDS
Objective: To review the conceptual framework and assumptions
associated with deriving biodegradation rates from decay curves
which appear first order with respect to chemical concentration and
whose slopes are linearly related to catalyst concentration. To
develop a strategy for future research.
Many chemicals appear to degrade exponentially when exposed to inocula
taken directly from the field. Plotting the logarithm of the chemical
concentration against time produces a straight line. The slope of this line
38
-------
is generally considered to be a rate constant which is first order with
respect to chemical concentration if the catalyst concentration remains
constant during the experiment and the chemical concentration is below the
half saturation constant (K$). Moreover, it is assumed that differences in
the first order rates from one experiment to another are due to differences in
the catalyst concentration. How valid are these assumptions? Do they need
further experimental verification and how? Are there special cases where they
apply? What range of chemical concentrations needs to be tested? Are
catalyst concentrations being determined properly and effectively? Are
biodegradation rate constants for any particular chemical the same, regardless
of the source or composition of the biological catalysts? What experimental
evidence is needed to substantiate or refute this concept? What alternative
mathematical expressions can be used or developed to describe exponential
decay?
Panel Members: Bob Boethling (Chairperson), U.S. EPA
John Rogers, Battelle Northwest Laboratory
David Lewis, U.S. EPA
Joe Suflita, University of Oklahoma
John Wilson, U.S. EPA
John Rodgers, North Texas State University
Rick Cripe, U.S. EPA
39
-------
THE " SECOND - ORDER " MODEL - AN OVERVIEW
Robert Boethling
THEORY
Kinetic models to predict rates of biodegradation of chemical
substances have generally relied upon the empirically derived Monod
equation (Monod 1949). According to Monod:
d[B]
dt
u
max
CS]
(1)
Kc + [S]
where [B] is the microbial "concentration," V m^x the maximum rate of
growth of the population, [S] is the concentration of the growth - limiting
substrate, and Ks is [S] when the growth rate is 1/2 V niax. It is evident
that the Monod equation is an expression of the rate of growth of the
population, d[B]/dt, in terms of population size and substrate concentration,
To derive an expression for the rate of biodegradation of a substrate S,
equation (1) may be modified by addition of a yield coefficient,
Y = d[B]/d[S]
(2)
which describes the efficiency of conversion of substrate into microbial
biomass. By multiplying equation (1) by 1/Y adding a negative sign to
convert equation (1) to an expression for loss of S, one obtains equation
(3).
d[B]
dt
d[S]
don
max
KS
[B] [S]
+ [S]
-dLS]
dt
[B]LS]
;+ CS.1
(3)
The expression -d[SJ/dt is equivalent to the rate of disappearance of the
growth - limiting substrate.
To make use of equation (3), one must assume that the rate of
population growth is limited by the substrate S over the expected range
of [S]. This is not likely in environmental situations, however, because
many other compounds that can serve as sources of carbon and energy are
commonly available.
40
-------
The difficulty of studying bacterial growth kinetics at environmentally
relevant concentrations of S, where S is truly growth - limiting, is
amply illustrated by the work of Shehata and Marr (1971) with pure cultures
of Escherichia coli. To remove traces of utilizable carbon that would
have precluded growth limitation by the added substrate glucose at parts -
per - billion (ppb) levels of that sugar, procedures such as distillation
of water from permanganate solution and exhaustive washing of glassware
with chromi acid were required.
A closely related problem arises from consideration of the yield
coeeficient Y. When substrate is growth limiting, Y approaches zero as
substrate is utilized for growth and energy, and [S] approaches zero.
Although not obvious from equation (2), this phenomenon is well documented
and is a reflection of the diversion of substrate metabolism from growth
to the need for "energy of maintenance" under conditions of carbon and
energy needs may be provided by metabolism of those substrates. The
practical outcome is that the "correct" value of Y to use in equation (3)
cannot easily be determined.
If one is willing to disregard these theoretical difficulties,
equation (3) can still be used to derive first-and second-order rate
expressions for biodegradation. If the concentration of substrate, [S],
is much greater than Ks, Ks + [S] approximates [S] and equation (3)
reduces to:
-d[S] = Umax [B] (4)
dt Y
In this case the rate of substrate disappearance is proportional to
the microbial concentration, [R], but not to substrate concenration.
This is the usual situation in laboratory studies of microbial growth in
pure culture, where substrate is plentiful. Environmentally, we are
usually more concerned with situations in which [S] is lower. If [S] is
much lower than Ks, Ks + [Sj approximates Ks and equation (3) reduces to:
-dLS] = Umax[B] [S] (5)
dt YKS
Equation (4) is a first-order kinetic expression because only one
term is present as an independent variable. Equation (5) is a second-
order expression because substrate disappearance is dependent upon two
variables, [B] and [S].
The kinetic constants U max, and Ks and yield coefficient Y (assumed
to be constant) may be expressed as a single constant K:
= U,nax (6)
YKC
41
-------
Equation (5) then becomes:
-d[S] = K [B] [S] (7)
dt
Equation (7) is the second-order rate expression used EPA's Athens
laboratory, and K is the second-order rate constant. Because of experimental
difficulties in measuring U max, Y and Ks for many organic chemicals, K
is usually determined from equation (7) rather than equation (6).
Rearrangement and integration (assuming [B] is constant) of equation (7)
yields equation (8):
In LS] = -K[B]t + In LS0] (8)
where [S0] is the concentration of substrate at the start of the experiment.
A plot of In [S] vs. t should yield a straight line having a slope of -K[B]
and a y intercept of In [S0]. The second-order rate constant, K, is
obtained from the positive slope of equation (8) by simply dividing by
the microbial concentration, [B]. The positive slope of equation (8),
K[B], is also referred to as the pseudo first-order rate constant. It is
clear from equation (8) that only disappearance of parent compound and
microbial concentraion must be determined experimentally.
Unfortunately, the assumption that [S] is typically much lower than
Ks, necessary to derive equation (7), may be a major theoretical stumbling
block. Values of Ks for naturally occurring organic compounds such as
sugars and amino acids are usually in the ppb range or lower (Hodson
1977, Wright and Hobbie 1965). Fewer Ks values have been obtained for
xenobiotic chemicals, but existing data do suggest that ppb values of Ks
are not uncommon. For example, Button et al. '(1981) recently obtained Ks
values for acetylene, benzene, and ethylacetate degradation in seawater
samples of 0.25 to 2.8 ppb. In the same study, Vmax for degradation of
toluene in seawater was reached at a substrate concentration less than or
equal to 52.7 ppb, suggesting that Ks was significantly lower. Moreover,
using estuarine water samples, Bartholomew and Pfaender (1983) have
observed Ks values well below 10 ppb for several chemicals, including m-
cresol and chlorobenzene. Baughman et al. (1980) state that Ks values
typically range from 0.1 to 10 parts per million (ppm) in natural waters,
but provide no references in support of that conclusion. Since concentrations
of xenobiotics in natural waters are often in the low ppb range (Sheldon
and Hites 1979), it is clear from the foregoing that [S] cannot be assumed
to be much lower than Ks. Thus, it seems likely that equation (3) will
in many cases not be approximated by equation (5). In those cases the
simple second-order rate model represented by equations (5) through (8)
is invalid.
The conclusion is not that this approach should be abandoned, but
rather that more research is needed. Particularly needed is better
knowledge of Ks values for biodegradation of organic chemicals by natural
microbial populations. But more fundamentally, we need better knowledge
of how microbial populations degrade organic chemicals at low, environmentally
relevant concentrations. For example, evidence (Rubin et al. 1982)
suggests that the bacteria responsible for biodegradation of organic
chemicals at very low substrate concentrations (oligotrophic bacteria)
42
-------
may be different from those carrying out similar transformations at higher
substrate levels, so that rate constants determined in studies using
substrate concentrations of 0.1-10 ppm may not accurately relect degradation
kinetics at lower concentrations.
43
-------
PRACTICE
Regardless of theoretical difficulties, the simple second-order rate
model may be useful as long as it "works." But inadequate validation
remains a serious deficiency, and careless manipulation of environmental
samples used in degradation studies can invalidate conclusions. With
respect to validation, the model predicts that the rate of biodegradation
of a substrate should be proportional to the microbial concentration, and
that second-order rate constants should therefore be reproducible from
site to site and sample to sample. Although a general relationship
between population size and biodegradation rate is intuitively reasonable,
few quantitative studies are available to support it. Paris et al (1981)
established such a relationship for three chemicals, the butoxyethyl
ester of 2,4-dichlorophenoxyacetic acid (2,4-DBE), malathion, and
chlorpropham (CIPC). Site-to-site reproducibility of the second-order
rate constants was excellent, with coefficients of variation for each
chemical of less than 65% over all sites. Larson and Payne (1981) recently
provided similar evidence for linear alkylbenzene sulfonates (LAS)
degradation in samples of Ohio River water.
On the other hand, several studies have failed to demonstrate a
significant relationship between biodegradation rate and microbial
concentration, or have shown that other factors may be more important in
determining biodegradation rates. In a study of biodegradation of several
organic chemicals in fresh, marine, and estuarine waters, Bartholomew and
Pfaender (1983) found that measurements of microbial numbers did not
correlate very well with measurements of microbial activity, and in
particular, with the maximum rate of utilization of the added substrate m-
cresol. Nesbitt and Watson (1980a, b) studied the biodegradation of 2,4-
dichlorophenoxyacetic acid in river water, and showed that degradation
was primarily dependent upon rnicrobial activity rather than the total
number of microorganisms. They also demonstrated that microbial activity
was related to the concentration of inorganic nutrients, the suspended
sediment load, and the dissolved organic carbon content of the water.
Yordy and Alexander (1980) demonstrated that the rate of disappearance
of N-nitrosodiethoanolaniine in samples were collected. These authors
suggested that the presence of competent microorganisms in water samples
was more important than total population size. Boethling and Alexander
(1979) observed apparent thresholds at ppb concentrations for degradation
of certain organic chemicals in samples of stream water. These observations
do not support a simple second-order rate model, since at low substrate
concentrations biodegradadation rates should be directly proportional to
the initial substrate concentration.
Spain et al. (1980) showed that adaptation of microbial populations
by prior exposure to an organic chemical could have a marked effect on
biodegradation rates observed in subsequent exposures to the chemical.
Measurements of total microbial population size did not reflect the
altered biodegradation rates. These and other studies show that environemtal
factors may overshadow the effects of population size. This statement has
important implications for prediction of biodegradation rates in natural
waters. It says that a smiple second-order rate constant determined
44
-------
using a given water sample cannot be used to predict degradation rates at
other sites, without consideration of factors other than total population
size that may affect the rates.
It is significant that (1) the compounds studies by Paris et al.
(1981) are known to be biodegraded initially by hydrolysis, catalyzed by
enzymes that are probably constitutive, and (2) biodegradation was followed
by measuring disappearance of parent compound. The choices of test
compound and analytical method thus acted to maximize the probability
that biodegradation rates would be proportional to total population size.
It is reasonable to suppose that different results might be obtained if
test compounds and conditions required degradation by less obvious routes.
Would degradation rates always be proportional to total population size
if the transformation of interest were carried out by few bacteria, via
complicated, inducible pathways? Would strict proportionality be obtained
if the concerted action of several different bacterial strains were
required? These phenomena are very common. Couldn't prior exposure to
the chemical of interest, or to chemicals having related structures, have
greater impact on degradation rates than total proulation size?
Many of these problems could be avoided if there existed good methods
for measuring not the total microbial population, but only that segment
of the population active in carrying out the desired chemical transformation,
Unfortunately, such methods are not available for most chemicals. This
is an area in which innovative research is badly needed. In the absence
of methods for measurement of transformer populations or activity,
calculation of second-order rate constants must depend upon measurements
of total population size. Examples of methods commonly used by
microbiologists include plate counts, determination of total adenosine
triphosphate (ATP), and direct counting by epifluorescence microscopy.
Which method is used really doesn't matter, since they all suffer from
the same drawbacks relative to their application to the simple second-
order rate model.
The preceding discussion points to the need for much greater effort
in the validation of the model. The most critical need is for testing of
more compounds, representing a wide variety of chemical structures of
varying complexity. The biodegradation of these chemicals should be
studied using samples of natural waters that would be expected to present
a variety of adaptation possibilities: water from streams receiving
industrial effluents, pristine lakes, saltwater environments, etc.
Furthermore, biodegradation should be determined by measuring a parameter
such as C02 evolution in addition to disappearance of parent compound.
The approach of Larson and Payne (1981) is encouraging in this regard, as
these investigators used ring-labelled LAS and followed production of
C02 from cleavage of the benzene ring. Ring cleavage requires the
action of inducible enzymes. The choice of LAS may be criticized from
the standpoint of model validation, however, since their use as detergents
has no doubt led to the widespread development in nature of microbial
populations competent to degrade them.
Another issue worthy of mention is related to the changes in microbial
numbers and activity, population diversity, and nutrient concentrations,
that usually occur during long-term incubation of environmental samples.
45
-------
It has long been known, for example, that changes in population density
of 100 fold or more can occur in samples stored for long or even short
(less than 24 hours) periods of time, even when the samples are refrigerated
(ZoBell 1946). Accordingly, certain precautions must be observed in any
biodegradation study in which degradation in environmental samples is
determined. Such changes are difficult or impossible to avoid in long-
term biodegradation tests, but they can be minimized during the period
from sample collection to the start of the test, and changes in microbial
numbers over time, at least, can be easily followed.
It is therefore unfortunate that Paris et al. (1981) do not indicate
how long or under what conditions wAter samples were held before they were
received in the laboratory. Moreover, significant changes may have
occurred during the 48-hour period between receipt of the samples and the
start of biodegradation tests. Paris et al.(1981) state that bacterial
numbers did not vary significantly over the course of their experiments.
Such behavior could be expected if one or more days had elapsed between
sample collection and the start of biodegradation experiments, because
bacterial numbers in stored water samples typically increase rapidly, and
then decrease very slowly over several days or longer (ZoBell 1946).
Considering the fact that water samples were collected from 40 sites in
18 states, delays of one to several days would seem inevitable. If the
microbial numbers thus determined do not reflect in situ values, second-
order rate constants calculated from those numbers cannot have environmental
significance for the sites from which the samples were collected.
SUMMARY AND CONCLUSIONS
The second-order model for prediction of biodegradation of organic
chemicals in natural waters offers beguiling simplicity. One needs only
to follow disappearance of parent compound and microbial concentration
over time. The pseudo first-order rate constant is determined from a plot
of In chemical concentration vs. time, and the second-order rate constant
is obtained from the pseudo first-order rate constant by dividing the
latter by the microbial concentration. The second-order constant is
presumably site and sample independent and can be used, given a measurement
of mocrobial concentration for any site, to predict a half-life for the
chemical of interest at that site. Structure-activity relationships can
be established by correlation of the second-order biodegradation rate
constants with other known physical/chemical or fate parameters, such as
alkaline hydrolysis rate constants (Wolfe et al, 1980).
Numerous theoretical and practical problems intervene in the
implementation of this model, however. Theoretical problems include the
false assumption that the test chemical is growth limiting, the uncertain
meaning of the yield coefficient Y in such cases, and probably most
important, the potentially false assumption that natural chemical
concentrations are much lower than Ks. Foremost among practical problems
is that the model is as yet largely unvalidated. If one is willing to
suspend reservations about their handling of water samples, Paris et al.
(1981) have indeed demonstrated the site independence of second-order
rate constants. But only three chemicals have been extensively tested,
46
-------
and the chemicals had structures for which some degree of site independence
of the second-order rate constants should not be terribly surprising.
Moreover, there exists a substantial body of evidence to suggest that,
for many chemicals, various environmental factors can overshadow the
effects of microbial numbers, thus precluding the establishment of
significant correlations between total numbers and degradation rates.
The most urgent need is for more research to demonstrate the practical
usefulness of the model. Many chemicals having a variety of structures
should be studied, and proper attention should be paid to storage and
handling of water samples. Concomitantly, microcosm studies should be
initiated with the same chemicals, and the results from the two techniques
compared. Until these experiments have been performed, it cannot be
assumed that second-order rate constants are site independent, or even
independent of sample for samples collected over time at a given site.
If such independence cannot be shown, second-order rate constants offer
no advantage over first-order constants.
LITERATURE CITED
Bartholomew, G.W. and F.K. Pfaender. 1983. Influence of spatial and
temporal variations on organic pollutant biodegradation rates in
an estuarine environment. Appl. Environ. Microbiol. 45:103-109.
Baughman, G. L., D. F. Paris, and W. C. Steen. 1980. Chapter 6.
Quantitative expression of biotransformation. Quantitative
expression of biotransformation rate. In A. W. Maki, K. L.
Dickson, and J. Cairns, Jr. (eds), Biotransformation and fate of
chemicals in the aquatic environment. Washington, DC: American
Society of Microbiology, pp. 105-111.
Boethling, R. S. and M. Alexander. 1979. Effect of concentration
of organic chemicals on their biodegradation by natural microbial
communities. Appl. Environ. Microbiol. 37:1211-1216
Button, D. K, D. M. Schel1, and B.R. Robertson. 1981. Sensitive
and accurate methodology for measuring the kinetics of
concentration-dependent hydrocarbon metabolism rates in seawater
by microbial communities. Appl. Environ. Microbiol. 41:936-941.
Hodson, R. E. 1977. On the role of bacteria in the cycling of
dissolved organic matter in the sea. Ph.D. dissertation.
University of California, San Diego, Scripps Institution of
Oceanography.
Larson, R. J. and A. G. Payne. 1981. Fate of the benzene ring of
linear alkylbenzene sulfonate in natural waters. App. Environ.
Microbiol. 4.h 621-627.
Monod, J. 1949. The growth of bacterial cultures. Ann. Rev.
Microbiol . _3: 37-394.
Nesbitt, H. J. and J. R. Watson. 1980. Degradation of the herbicide
47
-------
2,4-D in river water - I. Description of study area and survey of
rate determining factors. Water Res. 14:1683-1688.
Nesbitt, H. J. and J. R. Watson. 1980. Degradation of the herbicide
2,4-D in river water - II. The role of suspended sediment,
nutrients and water temperature. Water Res. 14:1689-1694.
Paris, D. F., W. C. Steen, G. L. Baughman, and J. T. Barnett. 1981.
Second-order model to predict microbial degradation of organic
compounds in natural waters. Appl. Environ. Microbiol. 41:603-609.
Rubin, H. E., R. V. Subba-Rao, and M. Alexander. 1982.
Rates of mineralization of trace concentrations of aromatic
compounds in lake water and sewage samples. Appl. Environ.
Microbiol. 4_3: 1133-1138.
Shehata, T. E. and A. G. Marr. 1971. Effect of nutrient concentration
on the growth of Escherichia coli. J. Bacteriol. 107: 210-216.
Sheldon, L. S. and R. A. Hites. 1979. Sources and movement of
organic chemicals in the Delaware River. Environ. Sci. Technol.
U: 571-579.
Spain, J. C., P. H. Pritchard, and A. W. Bourquin. 1980. Effects
of adaptation on biodegradation rates in sediment water cores
from estuarine and freshwater environments. Appl. Environ. Microbiol
4^:762-734.
Wolfe, N. L., D. F. Paris, W. C. Steen, and G. L. Baughman.
Correlation of microbial degradation rates with chemical structure.
Environ. Sci. Technol. _14_: 1143-1144.
Wright, R. T. and J. E. Hobbie. 1965. The uptake of organic solutes
in lake water. Limnol. Oceanogr. 10:22-28.
Yordy, J. R. and M. Alexander. 1980. Microbial metabolism of N-
nitrosodiethanolamine in lake water and sewage. Appl. Environ.
Microbiol. J39:559-565.
ZoBell, C. E. 1946. Marine microbiology. Waltham, MA: Chronica
Botanica.
48
-------
THE SECOND-ORDER APPROACH TO PREDICTIVE MODELING
John E. Rogers
ABSTRACT
To estimate the rate of degradation and thus the persistence
of organic compounds in a wide range of environments will require
an in-depth understanding of the microbiological transformation
pathways (reaction sequences) of major compound classes and the
effects of physical-chemical (environmental) parameters as well as
microbiological population dynamics on the rates of these
reactions. Although not complete, a wealth of information is
available in the open literature describing the metabolism of
major compound classes, especially in the area of aerobic
degradation. Anaerobic metabolism has not been characterized to
the same degree. The microbiological degradation and
transformation rates of numerous organics have been determined in
environmental samples. Unfortunately, much of this data base is
sample specific and not transferable to other sites. Only in
recent years have attempts been made to quantitate these rates so
they can be used to estimate the persistence of specific organics
in a range of environments. These studies have led to the
development of a number of mathematical models to estimate
persistence of organics.
DEVELOPMENT OF PREDICTIVE MODELS
The development of viable research and predictive models for the
degradation of organic compounds in the environment requires the interactions
of the following two research areas.
- The development of environmentally applicable mathematical
representations of the microbiological processes involved in the
transformation and ultimate degradation of organic compounds.
- The design and development of procedures to produce laboratory data
compatible with the mathematical representations.
Previously these areas have, in many respects, developed independently.
In the future, as microbiological degradation models become more
sophisticated, it is essential that they be investigated in an interactive
manner, each process in turn being directed by the other as new information
and methodologies are developed. This will ensure the most rapid and complete
development of useful microbiological models. This does not preclude
theoretical studies. However, the development of models is generally of
little practical value when no consideration is given to the availability of
laboratory data to exercise the model or worse if no consideration is given to
the development of laboratory techniques to provide the data. This is also
true of degradation studies, where the usefulness of the data, as input for
available or developing computer models, has not been considered. Continued
efforts to integrate these two areas will lead to better microbiological
models and a better understanding of microbiological degradation in the
environment.
49
-------
An important first step in this direction has been to use a pseudo
second-order rate equation to describe the microbiological transformation of
organics in fresh and marine waters. There are two constraints, however, to
the use of this equation: the substrate concentration must be less than
saturating (assuming that microorganisms can be considered catalysts) and it
must not support growth of the microorganisms. The utility of this equation
has been demonstrated over a large number of waters for several compounds
which include a number of phenol derivatives (Paris et al. 1982), malathion,
and chloropropham (Paris et al. 1981) as well as butoxyethyl ester of 2,4-
dichlorophenoxyacetic acid (Paris et al. 1981; Rogers et al. 1983). The
pseudo second-order rate constants were reproducible within ~65% for the
different waters examined within a given laboratory. This reproducibility
offered significant encouragement for the prediction of the environmental fate
in water of organics that met the constraints listed above.
Early in these studies it was observed that certain compounds were
transformed only after a variable and concentration dependent lag or
adaptation period. In the laboratory pseudo second-order rate constants are
determined by dividing the measured pseudo first-order rate constant by the
bacterial concentration. The pseudo first-order rate constant is obtained
from the slope of a log (residual organic) vs. time plot. When adaptation to
a compound occurs, the pseudo first-order rate constant is determined from the
linear portion of the log plot. It was immediately apparent from these
studies that the use of second-order rates to model the degradation of
compounds which first require adaptation is only a partial model. A complete
model would include a mathematical expression for adaptation and the
transformation rate. For situations where growth is occurring during
adaptation, the curve can be represented by the Monod growth equation for
example. In situations where growth does not occur but the organic substrate
concentration is saturating, the resulting data can be represented by the
Michaelis-Menten equation for enzyme kinetics. We have observed that for some
compounds the adaptation period is far greater than the time during which
degradation occurs. In this situation one can simply state that the compound
resides in the environment at a specific concentration for a specific period
of time.
UTILITY OF SECOND-ORDER MODELS
The potential utility of the second-order models rests on the ability of
laboratory studies to determine an average pseudo second-order rate constant,
using a number of different waters with different microbial populations. The
average rate constant can then be used to predict degradation rates at other
sites if the microbial population and organic substrate concentrations are
known. This combined methodology can also be used when the degradation rate
is described by alternative mathematical expressions. For example, if the
Michaelis-Menten equation (1) were used, one would simply determine the
average values for Km, Vmax and the bacterial population of a number of waters
and then use these average values to estimate degradation rates in other
waters where the substrate concentration and bacterial populations are
known. The bacterial population in this case is used to set the value of Vma
from an average specific activity term (sp. act. = Vmax/bacterial population)
determined in the initial evaluation studies. Similar examples can be derived
from other mathematical expressions for the degradation rate.
50
-------
A continued research effort in this area, which I will term the "second-
order approach to predictive modeling," should address the areas of adaptation
kinetics, reaction kinetics, and quantitation of biomass. Biomass has been
included because it becomes important when extrapolating from basic research
models to widely applicable predictive models which are normally constructed
to use readily available or easily attainable data. For example, the
potential for the degradation of organic compounds cannot be quantitated by
using a most probable number technique (MPN) which utilizes 14-C-labeled
substrates. Using this technique, one obtains a compound specific biomass
measure. 14-C-MPN techniques can be expensive. However, this technique can
be shown to be proportional to the total bacterial colony-forming units
obtained by standard water quality methods; for example, the initially costly
14-C-MPN data base can be expanded cost effectively to a wider range of
environments. This would make the best use of 14-C-MPN which is a good
research tool and standard plate count methods which is a widely used
nonspecific enumeration technique.
SUMMARY
A research effort should include at least the following set of tasks and
subtasks.
Task 1. Identify adaptation phenomena for different compound classes at
different concentrations.
Subtask 1. Classify compound classes as to 1, 2, and 3 below.
1. No adaptation.
2. Adaptation is approximately equal to half the time
required to degrade the compound.
3. Adaptations are markedly longer than time required
for degradation.
Subtask 2. Define the kinetics of the adaptation process
(growth, adjustment to compound, induction, etc.)
Subtask 3. Define computer inputs and mathematical expression
for different adaptation processes (e.g., Monod kinetics which
require input of binding constants and specific growth rates).
Task 2. Identification of specific transformation kinetic processes.
Subtask 1. Characterize compound classes as to pseudo first-
order, pseudo second-order, and Michaelis-Menten kinetics.
Subtask 2. Define computer inputs and mathematical expressions
for different kinetic processes.
Task 3. Develop site independent biomass measures.
Subtask 1. Develop compound specific biomass measures (e.g.,
MPN).
51
-------
Subtask 2. Develop general biomass measures which relate to
specific measures (e.g., cfu).
LITERATURE CITED
1. Paris, D. F., W. C. Steen, G. L. Baughman, and J. T. Barnett, Jr. 1981.
Second-order model to predict microbial degradation of organic compounds
in natural waters. Appl. Environ. Microbiol. 41:603-609.
2. Paris, D. F., N. L. Wolfe, and W. C. Steen. 1982. Structure-activity
relationships in microbial transformation of phenols. Appl. Environ.
Microbiol. 44:153-158.
3. Rogers, J. E., S. W. Li, and L. J. Felice. 1984. Microbiological
transformation kinetics of xenobiotics in the aquatic environment.
KPA-Grant No. CR810436-01-0.
52
-------
USE OF FIRST- AND SECOND-ORDER MICROBIAL
TRANSFORMATION RATE COEFFICIENTS FOR PREDICTIVE MODELS
David L. Lewis
ABSTRACT
In our research at the Athens-ERL, we have assumed that
microbial transformations of xenobiotic chemicals are a result of
enzymatic reactions, thus following Michaelis-Menten kinetics.
Based on Michaelis-Menten kinetics, we have used first- and
second-order transformation rate coefficients for predicting
transformation rates by suspended, attached, and sediment-
associated microbiota. Our second-order transformation rate
coefficients have been determined using either total plate counts
or transformer enumerations for suspended organisms (1), including
blended aufwuchs (2) and ratios of surface area to container
volume for surfaces, including sediments, covered with attached
organisms (3,4,5).
APPLICATION OF TRANSFORMATION RATE COEFFICIENTS
The application of transformation rate coefficients to predicting
transformation rates in field situations utilizes numerous assumptions
relating to (i) the applicability of the Michaelis-Menten equation and its
derivative equations, (ii) the use of correlative parameters such as plate
counts, direct counts, and colonized surface areas for determining second-
order transformation rate coefficients, and (iii) the effects of various
environmental parameters on transformation rates. The strengths and
weaknesses of these assumptions have been discussed in a recent manuscript on
the application of single- and multiphasic-kinetics to predictive modeling for
aquatic ecosystems (6). An analysis of these assumptions has led us to
conclude that in addition to further research on kinetics, other areas need to
be investigated to improve our predictive capabilities. These areas include
(i) suppression of microbial transformations of xenobiotic chemicals by
diauxie, biologically produced inhibitors, and other xenobiotic chemicals;
(ii) adaptation; (iii) mass-transport effects; and (iv) multiphasic kinetics.
FUTURE RESEARCH NEEDS
In general we feel that transformation rate coefficients based on
Michaelis-Menten kinetics are most applicable for predicting transformation
rates of xenobJotic chemicals in low environmental concentrations. A K value
as low as 10~ M^ was observed for transformation of diethyl phthalate figure
1). Therefore, the maximum concentration of xenobiotic chemicals for which
microbial transformation rates can be predicted without the confounding
effects of saturating some or all of the enzyme systems may be as low as
10 _M_. Transformation rates of higher xenobiotic chemical concentrations may
also be confounded by multiphasic kinetics, toxic effects of the xenobiotic
chemical, and adaptation. Figure 2 shows the general divisions of confounding
effects and applicable transformation rate coefficients for various xenobiotic
chemical concentration ranges.
53
-------
We believe that transformation rate coefficients may be satisfactorily
estimated using SAR (7,8,9), or by using relative transformation rate
coefficients compared to coefficients determined for microbial transformations
of reference chemicals (6).
There have been many frustrations associated with taking simple
approaches to predicting microbial transformation rates. These frustrations
have tended to compel us toward increasing the sophistication of our
approaches. However, we believe that our methodology must incorporate
analytical methods commonly available to industries and governmental agencies,
thereby preserving practicality. We hope to demonstrate that even diverse and
complex environments operate within sufficiently narrow bounds and without an
inordinate number of essential principles so that good estimates of
transformation rates can be made using relatively simple models.
LITERATURE CITED
1. Paris, D. F., W. C. Steen, G. L. Baughman, and J. T. Barnett, Jr. 1981.
Second-order model to predict microbial degradation of organic compounds
in natural waters. Appl. Environ. Microbiol. 41:603-609.
2. Lewis, D. L. and H. W. Holm. 1981. Rates of transformation of methyl
parathion and diethyl phthalate by aufwuchs microorganisms. Appl.
Environ. Microbiol. 42:698-703.
3. Lewis, D. L., H. P. Kollig, and T. L. Hall. 1983. Predicting 2,4-
dichlorophenoxyacetic acid ester transformation rates in periphyton-
dominated ecosystems. Appl. Environ. Microbiol. 46:146-151.
4. Lewis, D. L., R. B. Kellogg, and H. W. Holm. 1984. Comparison of
microbial transformation rate coefficients of xenobiotic chemicals
between field-collected and laboratory microcosm microbiota. ASTM
(STP). In press.
5. Lewis, D. L., H. W. Holm, H. P. Kollig, and T. L. Hall. 1984. Transport
and fate of diethyl phthalate in aquatic ecosystems. Environ. Tox.
Chem. In press.
6. Lewis, D. L. 1984. Application of single- and multiphasic Michaelis-
Menton kinetics to predictive modeling for aquatic ecosystems.
Environ. Toxicol. Chem. j3(3) .
7. Paris, D. F., N. L. Wolfe, and W. C. Steen. 1982. Structure-activity
relationships in microbial transformation of phenols. App. Environ.
Microbiol. 44:153-158.
8. Wolfe, N. L., D. F. Paris, W. C. Steen, and G. L. Baughman. 1980.
Correlation of microbial degradation rates with chemical structure.
Environ. Sci. Technol. 14:1143-1144.
54
-------
9. Paris, D. F., N. L. Wolfe, and W. C. Steen. 1984. Microbial
transformation of esters of chlorinated carboxylic acids. Personal
Communication.
55
-------
CD
en
o
Time (h)
Figure 1. Saturation of Brevibacterium sp. with diethyl phthalate (DEP).
Kffl was approximately 80 ug liter"1 (4 X 1C)-7M).
56
-------
1CT2-
£ 10~-
o
(D
O
£
O
O
cti
-i-j
£
'o
OH
l(T6-d
10"
Spill or
Discharge
Concentrations
Optimum
Analytical
Concentrations
Environmental
Concentrations
Microbial transformation rates are:
• First—order to enzyme concentration
• Zero—order to pollutant concentration
• Frequently subject to toxic effect of pollutant
• Frequently subject to population adaptation
Applicable rate coefficients are:a
Microbial transformation rates are:
• First-order to enzyme concentration
• Uultiphasic (series of first-,mixed—,and
zero-order) to pollutant concentration
• Sometimes subject to toxic effect of pollutant
• Sometimes subject to population adaptation
Applicable rate coefficients are:
Microbial transformation rates are:
• First-order to enzyme concentration
• First-order to pollutant concentration
• Rarely subject to toxic effect of pollutant
• Rarely subject to population adaptation
Applicable rate coefficients are:
Transformation rate coefficients are based on rates of loss of parent
compound and include kg. nag liter" * h~^ per biomass concentration
or per ratio of periphyton-colonized surface area to volume; kj, h"1;
kfo. liters cell"* h~*; and kA. liters m~^ n-l
Figure 2. Summary of microbial transformation rate information
based on Michaelis-Menten kinetics, experimental data, and
various assumptions used in the Exposure Analysis Modeling
System (EXAMS).
57
-------
THE USE OF PROGRESS CURVES TO OBTAIN
SECOND-ORDER RATE KINETIC ESTIMATES
Joseph M. Sufiita
ABSTRACT
The second-order rate model provides a firm theoretical
foundation upon which predictions can be made regarding the rate
of xenobiotic biodegradation. The advantages and assumptions of
this approach are well known and already explained in the
preceeding papers. Most importantly the kinetic expression is
formulated in terms of substrate decay which is (and in my opinion
should be) the parameter routinely measured. Secondly, the theory
is valid at low substrate concentrations which is the likely
scenario in the natural environment. Lastly, the expression is
site independent so different habitats can be compared.
The problems of this approach are equally well known. Most
obvious are the practical constraints associated with (a) the
assay of actively metabolizing biomass, and (b) the measurement of
trace concentrations of substrate. In this paper I will
demonstrate an under-utilized method of analyzing kinetic data
from biodegradation experiments and show its application to
second-order rate expression.
INTRODUCTION TO PROGRESS CURVE THEORY
Progress curves are experiments designed to follow the depletion of an
initially saturating amount of substrate through the zero-, mixed- and first-
order regions. An idealized progress curve is shown in Figure 1. By
appropriate analysis of progress curves, one can obtain the various Michaelis-
Menten parameters (K and V ) as well as the first-order decay constants for
experiments where microbial growth over the course of the experiment is an
insignificant factor.
Usually the calculation of Km and V are performed by using one of
several linearizing transformations of the Michaelis-Menten equation such as
the Lineweaver-Burk double reciprocal plot. This approach requires several
substrate-velocity data pairs and conclusions are usually drawn from 5 or 6
data points at best. It is becoming well known that the parameter estimates
using these techniques can be notoriously inaccurate.
However, such difficulty is partially circumvented through the use and
analysis of progress curves. Because the Michaelis-Menten model is a
differential velocity equation, its integrated form is valid over the entire
course of the reaction. The apparent K and V parameters can tse determined
by measuring the loss of substrate (or production of product) several times
during the course of the experiment incorporating as many data points as
desired. The information is then plotted using the integrated Michaelis-
Menten equation (Figure 2). From such a plot, or one of several other
linearizations of the Michaelis-Menten equation, one can easily derive the
Michael is parameters (Km and Vmax) as well as the first-order decay rate
max nr
58
-------
APPLICATION OF PROGRESS CURVES TO EXPERIMENTAL DATA
Figure 3 shows a progress curve for the anaerobic degradation of 4-amino-
3,5-dichlorobenzote by sediment microorganisms. Using the same data set, the
information was plotted using several linearizations of the integrated
Michaelis-Menten equation. The various equations and the associated
regression coefficients are shown in Figure 3. I purposely chose an equation
which did not give the highest R , estimated the kinetic parameters K» and
V , and then used a computer to simulate a progress curve based on those
kinetic estimates. The solid line is computer simulated, and the data are
superimposed on it. As can be seen, the fit is fajrly good, giving a Km
estimate of 30ym and a Vmax estimate of 1.5 ymoles 1~ h" .
Let's examine another progress curve, Figure 4a, this time following the
anaerobic dehalogenation of 3-chlorobenzote by an enriched methanogenic
consortium. On this arithmetric plot we can get a fairly good straight line
through most of the data, suggesting that it is zero-order. If we are truly
in the proper substrate range, we should get the curve shown in Figure 4b, a
semi-log plot. Note that if a regression line was drawn through the initial
data points, we would still get a fairly good straight line with an R of
about 0.98. In this particular case it covers at least one order of magnitude
before the curve bends off going to the true first-order region. The same is
true for the degradation of 3,5-dichl orobenzote (Figure 5). The same basic
pattern is seen: again quite a long region of linearity before the curve
gradually enters the first-order region.
We can analyze this progress curve data and generate the Michaelis-Menten
parameters and simulate progress curves based on our kinetic estimates. As
can be seen in Figure 6, we can estimate K 's of about 67 and 47 yM for the
mono and dichlorosubstrates respectively and V values of about 24 and 8.
The data in this figure are superimposed on the computer simulated solid line
based on the kinetic estimates. As can be seen, the fit is fairly good in
both cases. The important point is that the kinetic parameters can be
estimated from a single experiment.
Several questions arise. Can this approach be extrapolated to more
complex systems other than enrichments? How does it all relate to second-
order rate theory? Knowing Km and Vm we can easily calculate a first-order
decay constant which is simply V /K • According to second-order rate
theory, we need to normalize tne first-order rate by the biomass
concentration. The open question is what do we use as a measure of the
biomass which is actively involved in substrate metabolism?
In the saturating or zero-order region of substrate decay, the rate of
degradation is independent of substrate concentration and dependent on the
amount of catalyst. Therefore, the zero-order rate of substrate decay should
be an accurate reflection of the quantity of active catalyst in the system.
The zero-order rate is simply the Vmax term. We then need experimental
verification of this model. Ideally, we should be able to experimentally vary
biomass, run progress curves at initially saturating substrate concentrations,
and normalize the observed first-order decay rates by the zero-order decay
rates.
59
-------
This can be seen in an experiment done with anoxic sediment slurries.
The amount of biomass was varied by dilution with filter sterilized anaerobic
lake water. Progress curves were performed on these samples with 4-amino-3,5-
dichlorobenzoate as substrate and kinetic estimates were obtained using the
integrated Michaelis-Menten equation. As can be seen in Figure 7, dilution
had very little effect on the biochemistry of the reaction; the apparent K
was estimated to be about 30 yM in each case. However, dilution did have air?
affect on the amount of active biomass and therefore the V of the
reaction. It was approximately halved in the diluted sediment slurry*. If the
first-order decay rate is then normalized by the maximum velocity, then
essentially the same second-order proportionality constant is obtained. This
evidence supports the contention that the rate of biodegradation is
proportional to both substrate concentration and biomass and that V
estimates are a reasonable measure of the concentration of active catalytic
units. It therefore does not matter if the active catalyst is an exoenzyme,
individual cells, or a consortium of organisms.
Table 1 illustrates that there is fairly good agreement between the
expected and measured rates of 3-chlorobenzoate metabolism when the biomass
content of sediment slurries was varied by dilution. Thus, a decrease in
biomass resulted in a corresponding drop in the zero-order rate of decay.
FUTURE RESEARCH NEEDS
In conclusion, it can be seen that progress curves can be used to obtain
parameter estimation of the salient kinetic constants including the second-
order rate constant. In addition, it is believed that the V a of a reaction
provides a reasonable estimate of the concentration of active catalyst.
The advantages of using progress curves are:
a. Kinetic estimations are derived from a single experiment.
b. With accurate Kp values established with a prior research effort, any
lab can measure V and make predictions at any substrate
concentration; no need to go to low substrate concentrations.
c. Expressions can be amended mathematically to account for other
factors like partitioning, heterogeniety, etc. (see below).
d. Conservative rate estimations can be made.
The disadvantages of progress curve analysis are:
a. It may be impossible to get saturation for some substrates.
b. There is no accounting for lag or acclimation.
c. Math is sensitive to "renegade" points.
It is also assumed in the use of progress curves that a decrease in
reaction velocity is due only to decreasing saturation of the catalyst. In
addition, Michaelis-Menten kinetics only describe substrate consumption when
this process is either (a) unlinked to growth, or (b) when the amount of
growth occurring is less than that which gives sigmoidal substrate depletion
curves.
60
-------
It is very important in performing progress analysis to:
a. Choose saturating initial substrate concentrations (i.e., 2-4 K ) and
follow degradation to first-order region (O.IK ).
b. Take at least 10 data points covering the range in #1.
c. Use the linearized form of the integrated Michaelis-Menten equation
or nonlinear least-squares analysis of data to get parameter
estimations.
d. Simulate a progress curve based on the derived estimates of Km and
Vm,^ to test reliability of parameter estimates.
max
LITERATURE CITED
Suflita, J. M., J. A. Robinson, and J. M. Tiedje. 1983. Kinetics of
microbial dehalogenation of haloaromatic substrates in methanogenic
environments. Appl. Environ. Microbiol. 45:1466-1473.
61
-------
IDEAL PROGRESS CURVE
CTi
ro
c
0>
D.
0>
N
ro
o.
-a
3
IQ
0)
(/)
to
O
C
-i
CD
18
Zer°-°rder decay
nits
~ 12-
(D
*-•
CO
u.
+*
(0
X)
3
CO
6-
Kinetic Parameters
Vmax= ko
Mixed-order decay
First-order decay (k^
30
Time (Units)
-------
IDEALIZED PLOT OF INTEGRATED
MICHAELIS-MENTEN EQUATION
«n S0/S)/t = (-1/Km)(S0-S/t) + Vmax/Km
CO
•»•».
o
CO
Slope = -1/K
m
max
(Sft-S)/t
Figure 2. Idealized plot of integrated Michael is-: lenton equation.
63
-------
140-1
CTi
CO fD
v
Ul
I CO
Q. •
o
-s •
o n
CO O
&> fD
3 IQ
OO 01
O
3
O
i
CU
3
O
Un S0/S)/t
t/P
P/t
a.
>^
W
<
cc
CO
m
D
CO
(-1/KmXSo-S/t)+Vmax/Km .805
Km/Vmax [
-------
1200i
COOH
COOH
1000
800
5
>«»
UJ
CO
m
D
CO
600-
400
200
CH,
Figure 4a. Anaerobic dehalogenation of 3-chlorobenzoate by an enriched
methanogenic consortium.
65
-------
0
10
20
3'0 40
TIME (MRS)
50
60
70
Figure 4b. The dehalogenation of 3-chlorobensoic acid by an enriched
methanogenic community.
66
-------
0 30 60 90 120 150 180 210 240 270
TIME (MRS)
Figure 5. The pattern of 3,5-dichlorobenzoic acid dehalogenation to 3-
chlorobenzoate by an enriched methanogenic community.
67
-------
x
F
O •
CO =1
HI
<
cc
CO
m
CO
h* N.
• •
CO N
CM
N. N.
CO *fr
CO N
CO CO
*- o
N
o
o
CNJ
N
m «
CO 10
CO*
o
(O
IO
CM
o
o
CO
cc
10
N.
o
to
IO
CM
3-LVdlSanS
Figure 6. Simulated progress curves on mono and dichlorosubstrates determined
by the kinetic parameters.
68
-------
TABLE 1.
CTl
Zero-order rates of 3-Chlorobenzoate
degradation in acclimated sediment
diluted to various extents
Sediment Degradation rate (p moles i h )
(%)a
100
75
50
25
Actual
2.71 ±0.08
2.32 ±0.11
1.45 ±0.13
0.64 ±0.1 7
Expected
2.03
1.35
0.68
a) 100% = 50mg (dry weight) ml
-------
"vj
o
-5
fD
3 •
10
r^ -o
3" 1
fD O
-•• -5
3 fD
<•+ 10
fD (/>
-5 O
QJ c:
n- -s
fD <
Q. fD
o o
3 3"
O
-S
O
cr
fD
3
Nl
O
CXI
r+
fD
140-n
120
100
Sediment
99
50
Km
CpM)
29.9
29.7
(AT moles •!
V1)
1.49
0.65
k1
0.05
0.022
(AI moles I )
0.0335
0.0338
3- -S
fD r+
w' ""
1 -£=•
31
*
fD DJ
3 3
H- ->•
fD 3
3 O
1
fD CO
-Q •
c (_n
O> 1
c-h Q.
HI
H
DC
H
eo
CO
CO
80-
60-
4O-
50% SEDIMENT
A
99% SEDIMENT
20-
200 300
TIME (MRS)
-------
PANEL REPORT
The term "second order," previously used to describe a particular
approach to site-independent biodegradation kinetics, was felt to be
misleading and should be changed to "site-independent biodegration
kinetics." The term "second order," developed from initial studies by
utilizing a second-order rate equation derived from Monad kinetics, is now
considered to be limited to the number of compounds and sites for which it is
applicable. A site-independent mathematical approach should examine a number
of potential kinetics expressions, letting the data dictate which expression
is to be used by testing the fit of these equations by nonlinear regression.
In a practical sense, the approach would use a series of equations for which
the appropriate constants (Km, V , K2) have been developed in preliminary
laboratory studies, and which given some simple measure (active biomass,
substrate concentration, soil moisture, etc.) for the specific site, could
then be applied to a wide variety of environments.
Research needs may be divided into two major areas:
1. The range of chemical structures and habitats studied to date has not
been adequate to define the usefulness of this approach. This is
particularly true with respect to structure, since most of the
compounds for which we have data are degraded initially by hydrolytic
enzymes. We should now focus on compounds such as chlorinated
hydrocarbons and heterocyclics that are degraded by other routes. Of
general importance is that the compounds selected do not closely
resemble common naturally occurring compounds or other compounds
expected to be ubiquitous in aquatic environments. It is also
important that second-order rate constants be obtained for a wide
variety of aquatic, terrestrial, and subsurface environments. The
rate constants should then be compared not only within but also across
environmental types to determine the degree of site independence.
2. An important gap in our knowledge concerns the length of the lag or
acclimination period before degradation is observed. To our
knowledge, there is no firm theoretical or mathematical foundation
that can be used to predict when a particular compound may start to
degrade. We suggest that research be directed to understand the
reasons for acclimation periods and develop the appropriate
descriptive mathematical equations. It would be extremely useful if
the site independent microbial kinetic approach could be used to
predict not only biotransformation rates, but also acclimation
periods. As an initial approach, it might be worthwhile to try to
correlate acclimation periods with some measure(s) of biomass.
As general guidance we also offer the following suggestions:
1. The methods used for measuring catalyst concentrations represent an
area of special concern in any study designed to evaluate the
usefulness of the second-order approach. We suggest that simple
measures of total biomass such as plate counts or ATP levels be
tried first. It should be recognized, however, that measures of
total biomass may not be appropriate for many chemicals. Thus,
71
-------
measures of specific degraders or, even better, of degrader
activity are preferred. Possible parameters include V values,
specific degrader MPN's, and microbial numbers as determined by
microautoradiography. There may be other parameters and methods
that give even better results, however, and these should also be
explored.
2. Test chemical concentrations should range from very low to much higher
levels. Whereas very low concentrations simulate conditions in
many aquatic environments, high concentrations may be more relevant
near point source discharges, in spill situations, or in areas
receiving leachate from hazardous waste dumpsites.
3. The analytical regimen used to measure biodegradation should be
versatile, so that (a) the complete mineralization of the substrate
can be detected and quantified should that eventuality occur, and
(b) parent compound depletion and product formation can be measured
if complete mineralilzation is not the ultimate fate of the parent
compound. A mass balance should be established in all studies.
4. To estimate variability, one needs an adequate number of replicates
for each environment. Usually three to six replicates are
considered adequate.
Panel Members: Martin Alexander, Chairman
Bob Hodson, University of Georgia
Dennis Focht, University of California
Dennis Laskowski, Dow Chemical
Bill Walker, Gulf Coast Research Laboratory
Herb Fredrickson, Georgia State University
Tom Federle, University of Alabama in Birmingham
FACTORS CONTROLLING BIODEGRADATION RATES IN MICROBIAL COMMUNITIES
Objective: To examine the physical and biological factors which control the
rates of biodegradation and discuss how these rate-controlling
factors can be incorporated into kinetic expressions of
biodegradation. To develop a strategy for future research.
Biodegradation studies frequently produce chemical disappearance curves
which reflect rate limitations due to such factors as changes in population
size or composition, adaptation, competition for substrates, cometabolism,
sorption to surfaces, and inorganic nutrient concentrations. What evidence
suggests that these factors will be important in predicting biodegradation
rates in the field? Can the effect of these factors be deciphered from the
shape of the decay curve? What experimental techniques will provide more
quantitative information on these rate-controlling factors? Can we assume
that the size and activity of a degrader population within a microbial
community is constant from site to site? Is it important to develop kinetic
expressions which will accommodate different rate-controlling factors? What
research is required to correctly model these special aspects of
biodegradation?
72
-------
UPTAKE OF DISSOLVED ORGANIC COMPOUNDS
BY AQUATIC MICROHETEROTROPHIC POPULATIONS:
MULTIPHASIC AND SIMPLE MICHAELIS-MENTEN KINETIC PATTERNS
Robert E. Hudson
ABSTRACT
Concentrations of individual dissolved organic compounds in
natural waters are very low, in the range of picomolar to, at
most, micromolar. The low ambient concentrations are due to
efficient uptake of the organic substrates by microorganisms
possessing specific, high affinity transport systems. When these
systems are examined individually, most are found to possess
finite V and K, values and, thus, to exhibit simple Michaelis-
Menten uptake kinetics. Since the mid-1960's, when radiotracer
studies of the uptake of dissolved organic compounds from natural
waters were first carried out, and until recently, it has been
assumed that natural mixed-species assemblages of microorganisms
also exhibited simple Michaelis-Menten uptake kinetics. This
assumption formed the basis for models used to predict the fate of
both natural and pollutionally-derived organic compounds in
freshwater and marine environments. Recent evidence, however,
indicates that natural assemblages of aquatic microheterotrophs
possess a high degree of "kinetic diversity" when presented with
dissolved organic compounds. Simple hyperbolic kinetic patterns,
indicative of the entire population's having a single Kt and V ,
are observed less frequently than are complex, multiphasic kinetic
patterns indicative of the presence of a wide variety of
individual transport systems within the population.
INTERPRETATION OF KINETIC PATTERNS
Wright and Hobbie (1966) were among the first to study intensively the
relationships between the concentrations of dissolved organic substrates in
natural waters and the rates at which these substrates were taken up by the
natural aquatic microbial populations. They employed a linearization plot
derived from studies of enzyme-substrate interactions to determine graphically
kinetic parameters for the uptake of organic compounds. In this procedure the
value t/f, the incubation time (t) divided by the fraction (f) of labeled
substrate taken up during the incubation, is plotted versus (A), the
concentration of added substrate. The experimental procedures used initially
involved determining t/f over a relatively narrow range (e.g., between one and
ten micromolar) of substrate concentrations. The resulting data often fit a
straight line fairly well, a fact that was interpreted as indicating that
uptake was via a single transport system. With the Wright-Hobbie
representation of the uptake data, the slope of the line of best fit is taken
to be the inverse of the V ; the Y-intercept is the in situ turnover time
for the substrate, and the X-intercept is equivalent to the sum (Kt + Sn)
where Sn is the in situ substrate concentration.
After this approach became widely used, the data were nearly always
treated as though the relationship between t/f and (A) was theoretically
73
-------
linear no matter how well the experimental data fit a straight line. It
became generally assumed that the microbial population in a particular body of
water possessed a single Kt and Vmax for uptake of a given compound.
Concentrations of microbially utilizable compounds are exceedingly low.
Therefore, it was also assumed that the lack of diversity with regard to the
kinetic parameters was the result of long-term selective pressure toward
acquisition of lower and lower hL values. Gradually, it was assumed, all the
organisms would evolve the same low K£ (that is, the same high affinity for
uptake). These assumptions are themselves based on the tacit assumption that
an entire aquatic habitat is uniform with respect to concentration of any
particular dissolved organic compound. They take no account of the potential
heterogeneity that could exist within a water body on either a spatial or
temporal basis. For selective pressure to result in only a single low Kt, for
example, the organisms must never experience transiently high concentrations
of substrate. If they did, one might assume that some of the organisms would
have opted for exploitation of the "rich microzone" niche, and that this
adaptation would be reflected in higher Kt and V values than those in
organisms adapted to the ultra-low substrate concentrations of bulk water.
Microzones of organic enrichment do occur in natural waters. Such zones
include the immediate vicinity of photosynthesizing phytoplankton (even those
cells that are healthy), surfaces of suspended detrital particles, and the
guts of aquatic animals. These regions might produce sustained high
concentrations of dissolved organic substrates. In addition, transiently high
concentrations might result from the fragmentation of organisms such as algae
and protozoans when they are fed upon by zooplankton, and from the excretion
of feces by aquatic animals. These considerations lead one logically to
assume that the local concentrations of dissolved organic compounds in natural
waters may at times greatly exceed the overall bulk concentration that would
be measured after filtration (homogenization) of a water sample for organic
analysis. Instead of monotonously low, nanomolar levels, the
microheterotrophs might encounter, at various times, a wide range of substrate
concentrations from nanomolar to perhaps many micromolar or even millimolar in
enriched microzones.
RECENT FINDINGS
We have examined the uptake kinetics over a wide concentration range for
a number of naturally occurring organic compounds by natural
microheterotrophic populations in freshwater and marine environments (Azam and
Hodson 1981; Kirchman and Hodson, unpublished). When the substrate
concentration range used is increased to include concentrations that can be
expected to occur in organically rich microzones and the data are treated in a
manner similar to that of Wright and Hobbie (1966), straight line plots are
rarely obtained. Instead, the data plotted on the "linearization" graph
produce a hyperbola-like curve. No single 1C or V can be derived from the
curves, but rather the curves suggest that a wide range of Kt and Vmax values
are represented in the population. Some components of the population appear
to be adapted to using the ultra-low levels of organic substrate typical of
bulk water, whereas other components are adapted to take advantage of very
high concentrations where and when they might occur. The overall effect is
the failure of the substrate uptake rate to saturate. The population as a
whole can rapidly respond to a range of substrate concentrations that
stretches over many orders of magnitude.
74
-------
FUTURE RESEARCH NEEDED
Many pollutants are taken up (and subsequently degraded) by
microorganisms because of the similarity of their molecular structure to
naturally occurring substrates that the microbes are adapted to utilize.
Models to describe the rate of removal of such compounds from natural waters
have often assumed that the kinetics of uptake of the pollutant will reflect
those of the natural analog. And the assumption for the natural compounds has
previously been that they are removed via simple Michaelis-Menten kinetics by
the natural mixed populations. These models predicted that uptake (removal)
rates would saturate at some finite concentration of the compound in the
water. What we now know about the multiphasic nature of uptake kinetics
exhibited by many natural populations suggests that pollutant uptake rates
might not saturate but continue to increase as higher and higher V ax uptake
mechanisms come into play. At this time the presence or absence of
multiphasic kinetics in the uptake of xenobiotic compounds has not been
examined. However, the implications of the possible presence of this kinetic
pattern with regard to modeling the fate of pollutants in aquatic environments
warrant concentrating some research in this area.
LITERATURE CITED
1. Azam, F. and R. E. Hodson. 1981. Multiphasic kinetics for D-glucose
uptake by assemblages of natural marine bacteria. Mar. Ecol. Prog.
Series. 6:213-222.
Wright, R. T. and J. E. Hobbie. 1966. Use of glucose
bacteria and algae in aquatic ecosystems. Ecology 47:447-
2. Wright, R. T. and J. E. Hobbie. 1966. Use of glucose and acetate by
' - - ' - - -464.
75
-------
APPLICATION OF 3/2 ORDER KINETICS FOR SUBSTRATE
BIODEGRADATION IN SOIL
Dennis D. Focht
ABSTRACT
The kinetics of mineralization of carbonaceous substrates is
explained by a deterministic model which is applicable for either
growth or non-growth conditions in soil. The mixed order nature
(referred to as "3/2 order") does not require a priori decisions
about reaction order, discontinuity periods of lag or stationary
phases, or corrections for endogenous mineralization rates.
MODEL FORMULATION
The differential equation is
HS I
= K,S + K SE (1)
When K E is expressed as a function of time we have
- $ = K1S + K2St
for linear growth (i.e., K'E = K2t) and
- f = Kis + EoeVlts
for exponential growth, (i.e., K'E = EQeut). The rate of C02 product formation
dp - ds + K
dt-~dt + Ko
where K0is the zero-order rate constant for mineralization of indigenous soil
organic matter or, in the case of C0? evolution, the rate of mineralization
from residual product that has become part of the new humus. Upon integration of
equations 21, 2b, and 3 and by substitution we nave
-K,t - (K?t2)/2
P = SoCl-e ] + K0t (4a)
-K,t - ~(eut - 1)
] + K0t (4b)
for linear and exponential growth, respectively.
APPLICATION OF THE MODEL
C02 evolution data from soil were fit to the above equation by a
nonlinear regression analysis with the "NLIN" program by SAS on an IBM 750
computer using the Marquardt algorith for stepwise iteration. This program
gives a rapid convergence even when the initial estimates are considerably
76
-------
wrong. Linear growth (equation 2a) is applicable when initial cell densities
are relatively high and do not increase more than 100 fold, while exponential
growth (equation 2b) is applicable to low cell densities which increase by
several orders of magnitude in response to substrate additions. Because the
computer program might seem formidable and not accessible to non-mathematics,
a linear transformation of equation 4a can be used after the values for S0
and K0 are determined by a simple linear regression analysis of the zero-
order portion of the reaction (see Figure 1). The transformation to the linear
form:
Y = -iq - K2t/2 (5a)
is then possible in which
ln[(So - P + K t)/S ] (5b)
Y =
t
This equation was found to be an adequate approximation for equation 4a if
there are sufficient data points in the linear portion of the curve to
determine SQ and K0 accurately. (It should be noted here that S0 is not
strictly defined as the initial substrate concentration, but rather the
fraction of substrate that is readily converted to C02« This redefinition obviates
the problem of accounting for substrate which goes into humus or biomass.
FUTURE RESEARCH NEEDS
The advantages of the 3/2 order kinetics to the Monod equation are as
follows: (1) only 2 (K]_, K^) rather than 4 (Vmax, Ks, Y, X0) interdependent
constants have to be determined by nonlinear regression analysis (in lieu of
equations 5a,b); (2) the required initial estimate can be easily oFFained
from a linearized form (equations 5a, b) rather than from an interval estimate
of a differential equation; (3) substrate or product formation can be expressed
as an explicit function of time; (4) biomass concentration does not have to
be known.
The immediate question regarding the model is: what intrinsic
characteristics do the rate constants possess? The zero-order constant K0
represents the indigenous gross metabolic activity of the soil. Certainly
this value will differ among soils, so it represents a base value: i.e., a
soil with a higher K0 would probably also effect more rapid degradation of
added substrates than one with a lower K0. Thus, S0 the fraction of readily
mineralized carbon would be dependent upon K0. Compartive degradation rates
of any given substrate might be made on this basis. In the absence of growth,
the meaning of Kj, the "first-order" rate constant becomes very clear, and
one would thus be justified in using a "half-life" expression for the degradation
rate. The intrinsic meaning of l<2, the "second-order" rate constant, is not
clear at this time. However, it is probaly more yermaine to address the
question of how one expresses the nonlinear rate process of a sigmoidal curve.
Since we are concerned with expressing some quantitative meaning to the upper
and lower rates of degradation, it is justifiable then to define the kinetic
parameters Vm, the maximum rates of mineralization, and tm, the time at which
this occurs, this can be determined by taking the second derivative of equation
2a to express the change in velocity.
77
-------
d
dt
2 =
_
dt
K2t dS + K2S = d2P
'
and when d2S/dt2 =
dt
6)
dt
0, dS/dt = Vm, so that
K2t
m
( 7)
where Sm is the substrate concentration (i.e., the saturation constant)
8)
This clearly states that K2
when there is no inflection
Thus the tm value is useful
length of the achieved.
K^; otherwise, t[n 0, a situation that develops
point (i.e., the rate is essentially first order).
quantitiatively in telling us something about the
When exponential growth gives a better fit than linear growth, the KI term
has always been found to be negligible. Thus equations 2b and 4b are reduced
to
dt
= Eo e
1 - e
+ K
(10)
If we take the second derivative of equation 9 to find the inflection point
as we did in the development of equations 6-8, we obtain for exponential
growth
-Eo/
(11)
Sm - V
(12)
In ( /E0)
The application of the linear and exponential growth models to the mineralization
of biphenyl in soil can be seen in Fig. 1 Linear growth gives a better fit
with inoculation, while exponential growth gives a better fit in lieu of
inoculation: presumably this is due to the low indigenous population density
of biphenyl degraders.
78
-------
200-
0
0
50
Fig. 1. Fitting of the 3/2 order kinetic model to CO? evolution
from biphenyl additions (0.33%) to soil. Addition of
1Q5 cells/g of the biphenyl-degrader Acinetobacter P6
(Furukawa) (Q) and no inoculation (•) are respectively
best fit for linear (S0 = 17, KI = 0.0050, K? = 0.045)
and exponential (S0 = 15, y = 0.18, E0 = 0.0059) growth
The linear portion of both curves (K0 = 0.090) is the
same as the control (Q) without biphenyl. Complete
mineralization of biphenyl would bring about 240 pmoles
C02/g.
79
-------
FACTORS CONTROLLING DEGRADATION RATES IN SOILS
D. A. Laskowski
ABSTRACT
In keeping with the workshop theme, this paper attempts to
summarize the needs for characterizing degradation kinetics of
chemicals in soils. Kinetics is becoming increasingly important
because of its key role in any evaluation of environmental impact
of chemicals. Modeling is definitely emerging as the premier tool
for fate analysis and degradation kinetics are mandatory in such
exercises.
KINETIC ANALYSIS OF SOIL DEGRADATION
Unfortunately there is not much kinetic data available on soil
degradation, even though much effort has been spent in studying chemical fate
in soil. Few experiments have been carried out that quantify the statistical
variability in a rate constant when switching from one soil to another.
Effects of climate on the rate constant have seldom been addressed. It is not
known if kinetic data from soil can be used in aquatic environments and vice
versa. Researchers believe the differences are so great between aquatic and
terrestrial systems that kinetic data from one are not transportable to the
other. But where is the evidence? Do rate constants from an aquatic system
truly belong to a different population when temperature, moisture, and
inherent variability within water and soil populations are factored out?
FUTURE RESEARCH NEEDS
Our assessment of the progress in learning about degradation kinetics of
chemicals is that there has not been much learned in the last ten years.
There are many questions but few answers. Questions that we ourselves have
arrived at over the years are summarized below:
1. What effect does climate (soil temperature and moisture) have on
degradation rate and does this effect vary from one soil to another?
2. What is the relationship between initial concentration and degradation
rate? Soils seem to have a finite capacity to degrade chemicals. How
much does this vary from one soil to another?
3. What influence does time of contact have on degradation kinetics?
There is suggestion that chemicals become less available for
degradation as residence time in soil increases.
4. Finally, how much variation is encountered in changing from one soil
to another? How is this variation distributed statistically? Can it
be reduced by factoring out correlations between rate and soil
property data that can be obtained readily?
80
-------
We embarked on a program to address some of these questions since it is
not possible to model fate effectively without this information. Soils are
being collected from across the country and will be used to examine
relationships between degradation rate and the variables' initial
concentration, soil temperature, soil moisture, and soil. We hope to have at
least some answers from our experiments over the next couple of years.
PANEL REPORT
Environmental factors govern the rate and extent of biodegradation and,
indeed, may determine whether or not biodegradation occurs. In addition,
environmental factors determine the differences observed between
biodegradation rates in laboratory tests and in the field, between different
sites in natural ecosystems, and also at the same site but at different
times. Nevertheless, inadequate information exists on which environmental
factors in nature govern the rate and on the quantitative relationship between
the ecologically significant factors and changes in rates. This information
is essential for meaningful predictions of rates in natural ecosystems based
on laboratory tests and of rates in different environments.
In some instances there is general agreement on the importance of
individual environmental factors, yet, even in these instances, the
quantitative impact of increasing intensity of that factor on biodegradation
rate is uncertain. In other instances only a general perception exists that a
particular factor may be significant and, in these cases, the role of that
factor must be established and then its quantitative impact must be
established. Not all physical, chemical, and biological properties or
characteristics of natural environments will have a significant effect, or any
effect, on the rates of biodegradation. Yet widespread agreement does not
exist on which factors should be ruled out as being unimportant.
Thus, we recommend that the environmental factors that are of primary
importance in governing biodegradation rates be characterized. The factors of
primary importance are: (a) temperature, (b) aeration, (c) concentration of
xenobiotic, (d) history of prior exposure to xenobiotic or related chemicals,
(e) salinity, (f) surfaces (primarily in flowing streams), (g) inorganic
nutrients (primarily in aquatic environments), (h) moisture level of soils,
and (i) soil and sediment type (including organic matter content, caution
exchange capacity, texture, and clay type). Other factors that probably are
of lesser importance or that are important under specific circumstances or for
particular chemicals are (a) sorption, (b) suspended solids (aquatic systems),
and (c) pH.
The significance of grazing on the biodegrading species by protozoa and
other organisms, the importance of the sediment-water and air-water
interfaces, and the use of rates of community metabolism as a predictor of
biodegradation rates should also be considered. In each instance the
quantitative relationship between the intensity of the factor and
biodegradation should be established. Prime attention should be given to
assessing which of these are the major factors affectiny biodegradation rates,
which are minor, and which have no detectable influence.
81
-------
Representative chemicals should be chosen as benchmark compounds to be
used in biodegradation studies. They should represent xenobiotics that
intrinsically have different rates of biodegradation, and whose metabolism is
initiated by markedly different types of enzyme reactions. These compounds
should be chosen with care.
To facilitate the comparison of biodegradation rates under a variety of
environmental conditions, we feel that a set of standard reference compounds
(representative of various class of xenobiotics) be examined in a manner that
produces easily comparable data. Using these various test compounds, we
should compare controlled laboratory experiments to environmental studies.
Individual factors should be manipulated in the laboratory and their effect on
biodegradation assessed. The natural spatial and temporal variability in a
geographically and trophically diverse set of systems should be exploited by
using regression studies to correlate biodegradation with these factors.
Environmental factors may have different effects with different
compounds. For example, the Q-.Q for the biodegradation of one compound may be
2, while the biodegradation of another may be 20, and thus they have different
temperature coefficients. Or, with respect to aeration, the rate of
hydrocarbon degradation would be drastically affected (from relative rates of
100% to 0% while the rate of glucose degradation would be relatively
unchanged, (e.g., 100-75%) by varying aeration status. Hence, we recommend
that research on the role of environmental factors be designed to evaluate
their impacts on compounds whose biodegradation will probably be affected
differently or to different extents by changes in the same factor.
The confidence intervals for the quantitative effects of each of these
environmental factors must be defined. Then, the significance of each of
these factors and the interactions among them must be validated by field
trials. If these field trials suggest that factors other than those proposed
here are important, these factors should be defined and their quantitative
impact evaluated.
The apparent lag phase or period for acclimation is important because it
may be a significant part of the time that the compound persists. Yet, almost
nothing is known of factors affecting the length of this phase or its cause
(i.e., time for population growth, induction or enzymes, or genetic
changes). Therefore, research is needed to establish these factors and to
determine the reason for this apparent lag phase.
82
-------
APPLICATION OF UPTAKE AND MINERALIZATION KINETICS
Objective: To evaluate the potential application of uptake and
mineralization kinetics to the biodegradation of xenobiotic
chemicals by natural microbial communities.
The biological transformation of naturally occurring organic materials at
concentrations typically found in receiving waters or soil has been
extensively studied by kinetic analysis of the rates at which radiolabeled
substrates are taken up and/or mineralized to carbon dioxide by heterogeneous
microbial communities. These studies are based on the demonstration of
saturation-type kinetics from which estimations of specific and maximum
turnover rates, half-saturation constants, and specific activity indices can
be derived. Only recently has this kinetic approach been applied to the
transformation of xenobiotic chemicals. Is this a reasonable application, and
what new information or insights can we expect to gain from this approach
which cannot be obtained by other approaches? Can saturation kinetics be
demonstrated for most xenobiotic chemicals? Can specific activity indices be
used as a method for estimating the fraction of the microbial population
involved in biodegradation? Must xenobiotic chemicals be used as carbon
and/or energy sources by microorganisms before these kinetic techniques can be
applied?
Panel Members: Carol Litchfield, Dupont
Fred Pfaender, University of North Carolina
Bob Larson, Proctor and Gamble
Don Button, University of Alaska
Roy Ventullo, University of Dayton
83
-------
HISTORICAL BACKGROUND TO THE MEASUREMENT OF
HETEROTROPHIC POTENTIAL IN THE AQUATIC ENVIRONMENT
C. D. Litchfield
ABSTRACT
My purpose here is to present an historical background to the
measurement of heterotrophic activity in the aquatic
environment. This will enable all of us to have the same frame of
reference for evaluating the application of this technique to
biodegradation studies. These applications will be described by
the other members of this panel.
HISTORICAL BACKGROUND
In 1951 Steeman-Nielsen published the first paper on the use of C
labeled carbon to estimate the production of organic matter in the ocean by
phytoplankton (5). The next year he applied this method to field studies and
described the technique which is still used, essentially unchanged since that
paper was published (6). In 1962 Parsons and Strickland applied the
radiocarbon labeled concept to the measurement of heterotrophic processes
using C-labeled organic substrates. They noted that a Hanes transformation
plot (1) of the data derived using Michaelis-Menten kinetics usually resulted
in a straight line. From the slope and intercepts of that line they then
calculated their estimations of heterotrophic potential (4). In that paper,
however, they cautioned, "the method is rapid and convenient and can give
values of relative heterotrophic potential. As with the radiochemical method
for measuring marine photosynthesis, the exact interpretation of results
presents many problems...." (4) (the emphases are mine).
As useful as the procedure was, they did not recognize, however, that the
majority of the carbon was not converted to biomass. In fact, it is not
unusual for the percent converted to biomass to account for only 10-30% of the
total carbon used. By 1966 it had become apparent that a major correction was
needed to account for the respired CO-, and such corrections would obviously
be important to any interpretation of neterotrophic potential. Two different
approaches to solving this problem were presented by Kadoata and his co-
workers (3) and by Williams and Askew (7). These papers provided the basis
for the publication in 1969 by Hobbie and Crawford of their paper,
"Respiration corrections for bacterial uptake of dissolved organic compounds
in natural waters" (2).
With this newer information, then, the procedure began to be widely
applied in microbial ecological studies, and numerous papers have been
published discussing the assumptions, applications of the technique to
different ecosystems, and the problems with the interpretation of
heterotrophic uptake/mineralization studies.
84
-------
REVIEW OF THE BASIC TECHNIQUE
I would now like to review very briefly the basic technique used in
measuring heterotrophic activity, the calculations which are used to derive
"heterotrophic potential," and the assumptions inherent or implied in adapting
Michaelis-Menten kinetics to ecological studies. Typically the tests are
performed in an enclosed sample bottle where the respired CC^ is trapped on a
suspended piece of filter paper moistened with phenethylamine or KOH.
Determination of the amount of C02 respired is performed using a liquid
scintillation counter and the resulting data are used to derive the estimates
of heterotrophic mineralization. To determine the heterotrophic uptake, the
incubation mixture is filtered, washed, and the filter also counted. The
amount of radiolabeled carbon in or on the cells is the basis for the
calculations of heterotrophic uptake. Several concentrations of the substrate
and several time periods (usually five minutes to three hours or less) are
necessary unless one already knows the in situ concentrations of the substrate
and the biomass of organisms able to utilize fnat particular substrate.
There are many assumptions inherent in applying Michaelis-Menten kinetics
to such studies:
o The substrate is present at saturating concentrations.
o There is no significant change in the substrate concentration during
the incubation period.
o There is no change in the concentration of the biomass of the actively
metabolizing population.
o There are no activators required nor inhibitors present.
o In the equation below, K3 is the rate limiting step:
Bacteria (B) + Substrate (S) *nr~" BS - B + Product(s)(P)
Although not part of the kinetic assumptions, other requirements in the
mechanics of these studies are that: samples must be incubated at in situ
temperatures and pressures, if applicable; added substrate concentrations must
be less than the background or natural substrate level; and the oxygen, redox,
and pH conditions should approximate in situ conditions.
If all of these assumptions and conditions have been met, the data can
then be plotted using the equation:
K, + S
t/f = 1(s) + -t LL
WT v V
max max
The resulting linear plot is shown in Figure 1 where t/f is plotted
against the added substrate concentration, and the slope of the line is
l/Vmax, the y-intercept is a function of the in situ substrate concentration
plus an indication of the affinity of that population for the substrate (i.e.,
a high K^. will require a high substrate concentration to activate the
microbial population).
85
-------
If
invalid
data.
all of the conditions and
to apply Michaelis-Menten
assumptions have not
kinetic equations to
been met, then it is
the analysis of the
FUTURE RESEARCH NEEDS
These types of studies can be performed using substrates not commonly
found in the environment, xenobiotics, or even naturally occurring organic
compounds. We have been tracking changes in the heterotrophic mineralization
rates for several naturally occurring materials. The results with urea are
shown in Figure 2 where the amount of urea C mineralized per 24 hours is
plotted for different zones in the New York Bight apex. Using the
heterotrophic mineralization/potential approach, then, we are able to estimate
or compare relative microbial activities in different marine environments over
a prolonged time.
If the investigator remembers the limitations of this method and applies
the resulting data appropriately, heterotrophic potential measurements can
provide a relatively rapid indication of the ability of the existing microbial
population to transform and metabolize organic compounds entering the
envi ronment.
1.
2.
3.
4.
5.
6.
7.
Dixon, M. and E.
Academic Press.
LITERATURE CITED
C. Webb. 1964. The nzymes, 2nd edition. New York
Hobbie, J. E. and
bacterial uptake
Limnol. Oceanogr.
C. C. Crawford. 1969. Respiration corrections for
of dissolved organic compounds in natural waters.
14:528-532.
Kadota, H., Y. Hatta, and H. Miyoshi. 1966. A new method for estimating
the mineralization activity of lake water and sediments. Memoir Res.
Inst. Food Sci. Kyoto Univ. 27:28-30.
Parsons, T. R. and J. D. H. Strickland. 1962. On the production of
particulate organic carbon by heterotrophic processes in sea water.
Deep-Sea Res. Oceanog Abst. 8^:211-222.
Steeman-Nielsen, E. 1951.
matter in the sea by means
Measurement
of carbon 14.
of the production
Nature 167:684.
of organic
Steeman-Nielsen, E. 1952.
organic production in the sea.
140.
The use of radioactive carbon for measuring
J. Cons. Perm. Int. Explor. Mer 18:117-
Williams, P. J.
mineralization
Deep-Sea Res.
LeB. and C. Askew. 1968. A method of measuring the
by microorganisms of organic compounds in sea water.
15:365-375.
86
-------
STATION 2 0-IOcm
15 MR 1974
J
I I
J I
5 10
nm [S]
14
Figure 1. Typical Hanes plot of C-mineralization data. The plot shown is
for one location in the 0-10 cm portion of a core from the sediments
at Sandy Hook Bay, Xew York.
87
-------
Seasonal Influence on the Amount of Urea-C
Utilized Per 24 hr Day In Different
Zones of the New York Bight Apex
'Surface
Layer
•fay July
1§T7
CrulM
Figure 2. Seasonal influence on the amount of Urea-C utilized per 24 hr
day in different zones of the New York Bight Apex.
88
-------
APPLICATION OF UPTAKE AND MINERALIZATION KINETICS
Frederic K. Pfaender
ABSTRACT
Over the last several years we have been investigating the
application of heterotrophic uptake kinetics to the measurement of
pollutant biodegradation. Various versions of the technique have
been used, with varying degrees of success, for many years to
assess the heterotrophic potential of natural aquatic bacteria,
using substrates like glucose, acetate, and amino acids. The
approach is based on Michaelis-Menten kinetics, and was developed
to describe the kinetics of enzyme-substrate reactions. The
application of this approach to mixed populations of intact
microorganisms has often been questioned. The saturation response
has been shown empirically in many, but not all, environments and
for 3 variety Of Substrates. The fact that saturation is obtained
for many pollutants is taken as an indication that this kindjsf kinetic
approach can be applied to describing the activities of microorganisms
in environmental samples.
APPLICATION OF MICHAELIS-MENTEN KINETICS TO POLLUTANT DEGRADATION
The use of Michaelis-Menten kinetics to describe pollutant biodegradation
rates requires that certain assumptions be made:
1. For the measured kinetic parameters to be accurate, the amount of
enzyme present during the incubation period must remain constant. The
extension of this assumption to environmental samples means that the
numbers of organisms and/or the amount of biomass should remain
constant during the incubation. This, of necessity, will require a
relatively short incubation period, since it is reasonably well known
that the numbers of organisms do change while confined in a container.
2. It is also required that the concentration of substrate not change
significantly during the course of the measurement. These two
assumptions can only be satisfied if the measurement is made over a
relatively short period of time. Long-term incubations will result in
both an increase in the number of organisms present and a decrease in
the concentration of the substrate, due to its breakdown by the
microorganisms.
3. The assumption is made that the transport systems present are
responding only to the compound added as substrate, and not to other
compounds present in the environment.
4. In our particular case, we make the assumption that concentration of
the specific pollutant substrate present in the environmental sample
is negligible compared to the amount we add as substrate. This
assumption allows us to disregard the natural substrate concentration
in the calculation of the kinetic parameters.
89
-------
We feel that the use of this kinetic approach to measuring pollutant
blodegradation has several very positive features. From past experience in
our laboratory, we know that confining samples within a bottle results in
significant changes in both the size and activity of the microbial
community. We also know that there is a period of approximately 12 to 16
hours before these changes become significant. Therefore, if biodegradation
rates can be measured during this time period, a rate can be generated that
might be quite close to that which would occur under natural environmental
conditions. Since the assumptions required for use of this technique
necessitate short-term incubations, this methodology appears quite appropriate
for answering questions about real world rates. In addition, the kinetic
approach offers the advantage of being able to calculate several useful
biodegradation' parameters. These include V , the maximum potential
velocity that the indigenous community canrafl:tain, Km or the half-saturation
constant which gives an indication of the concentration range to which the
community is adapted. In addition, a rate constant, which we have called K,
and has the same units as a first-order rate constant, can be calculated from
the kinetic data in several different ways. These include the dividing of
V by Km, or taking the slope of the linear portion of the velocity vs.
concentration plot. Two additional non-Michaelis-Menton parameters can be calculated
from the data generated in these experiments. The velocity of metabolism witK
any particular added concentration can be obtained from the data generated by
multiplying the concentration added by the f/t factor. In addition, the
turnover time for the added xenobiotic at any particular concentration can
be measured directly from the linearization of the kinetic data.
INTERPRETATION OF EXPERIMENTAL RESULTS
We can calculate several parameters from the uptake and mineralization
data, but have questions about which one is most useful. V may represent
the maximum potential velocity, and when we look at the saturation plots we
see that for some compounds saturation occurs at what may be environmentally
realistic concentrations (cresol Figure 1, and chlorobenzene Figure 2). For
other compounds, however, (NTA) many hundreds of mg/1 are needed, which are
concentrations far in excess in what might actually occur. We began
calculating K, because it is derived from the slope of the linear portion of
the curve, generally in the lower part of the concentration range. We felt
that it gave a realistic additional measure of biodegradation. When we
do correlations of biodegradation rate measures and environmental
Characteristics (Table 1), V gives relationships that make sense. In
coastal environments where nutrient concentrations are low, we get a strong
relationship with nitrogen and phosphorous species and chlorophyll, which is
the principal source of nutrients. In the lake and river environments where
concentrations of nutrients are many times higher than the coastal
environments, nitrogen and phosphorous do not appear to be strongly related to
biodegradation rates. Although not shown in the table, when K, and Vmax are
correlated with one another and with environmental parameters, they generally
Correlate well with one another and with the same characteristics of the
environment. While it is reassuring that these two measures of biodegradation
appear to be describing similar processes, it does not help a great deal in
determining which one gives the most realistic or useful measure of
biodegradation. Since both are obtainable from the same set of laboratory
measurements, it is always possible to calculate and report both.
90
-------
POTENTIAL FOR OBSERVING SATURATION KINETICS
We have always been able to obtain saturation if appropriate
concentrations are used. As mentioned previously, for some compounds this
requires higher concentrations than are likely to occur in the environment.
We are also aware of several reports in the literature of saturation not being
obtained with many natural substrates. Since we have looked at only four or
five pollutant compounds, in only a few environments, it is difficult to make
generalizations about many different classes of materials. This raises
questions about the conditions under which saturation may be obtained. In
attempting to shed light on this question, we have used several techniques to
estimate the number of degraders of specific compounds. Table 2 shows the
results of this evaluation in some fresh water samples using two methods,
plate counting and microautoradiography. The important point of this table is
that the specific degraders constitute a relatively small part of the total
community. I believe the reason we are able to obtain saturation is because
this small community behaves in a manner more similar to a pure enzyme system
than would the larger, more diverse community that might be metabolizers of
naturally occurring substrates. Since the technique we use was developed tc
measure the activity of pure enzyme systems, I do not consider it surprising
that we are able to obtain saturation when working with a very small part of
the total assemblage of microorganisms present in an environmental sample.
These questions, however, are still far from being answered.
EFFECTS OF INCUBATION PERIODS
The uptake and mineralization technique that we have used to assess
pollutant biodegradation requires that the measurements be made over a
relatively short period of time. During this short incubation period only a
small proportion of the label added is metabolized, usually from one to three
or four percent. This obviously raises the question of whether this short-
term rate can be extrapolated to the longer periods of time necessary to
achieve complete breakdown of the pollutant. In an attempt to gain some
understanding of this we have conducted time course experiments in which
extended incubation periods were used and the amount metabolized over time
measured. Table 3 shows the results of this study for two compounds,
chlorobenzene and meta-cresol. We have used KI to predict the percent of the
compound that would be degraded at different time periods and compared this to
what was observed in the time course study. We have also used V to predict
how many micrograms of the material would have been metabolizeci at the same
time periods and compared this with the micrograms actually converted to coz
and cells. As you can see, for both compounds and both methods of estimating
biodegradation, there is good agreement between the two methods for an initial
period of several days. However, after five to seven days confinement in the
time course studies, the amount of degradation slows down, yielding an
apparent underestimation of the amount of pollutant that would be biodegraded,
as predicted by the kinetic approach. It is interesting to note the good
agreement between the V prediction and the micrograms actually
metabolized. This would tend to indicate that in the containers used for the
time course study the community is functioning at close to V . Since we do
not have information on the long-term degradation of these compounds in the
91
-------
environment tested, we have no way to determine which method is providing the
most accurate estimate of persistence. However, good agreement between the
two techniques during the early stages of Incubation and the fact that changes
are known to occur in the container after some period of confinement would
tend to point to the kinetic approach, giving a more realistic estimate of the
amount of time necessary for biodegradation.
FUTURE RESEARCH NEEDS
We are relatively confident that the uptake and mineralization approach
has potential as a rapid and useful measure of pollutant biodegradation.
However, additional research needs to be addressed to several pertinent
questions:
1. To how broad a range of compound and environment can this approach be
successfully applied? At this point we have examined four or five
compounds in three coastal and four freshwater environments. These
sample sites are all in one geographic part of the United States, and
certainly do not represent the range of environmental characteristics
that might be found throughout even the United States. f
2. Under what environmental conditions does saturation occur? While we
know that it occurs for the compounds and environments we have tested,
neither represent the range of materials or sites that are
available. The methodology offers the opportunity to calculate
several degradation parameters (K,, specific velocities and turnover
times) even if saturation is not obtained.
3. How valid are extrapolations based on the uptake of only a small part
of the added compound? In the comparisons we have done with time-
course studies, the extrapolations looked to be reasonably good, at
least for a period of several days. One of the problems we face in
testing these extrapolations is in determining what to compare the
extrapolations to. While time-course studies are extremely useful in
determining whether a compound is degradable and required for
identifying degradation products and mechanisms, their use for
estimating environmental rates of biodegradation appear questionable,
but there are few other ways to do comparisons.
4. Does adaptation occur at the low environmental concentrations of most
pollutants? This is a question that relates not only to this
technique, but to all of the methods that we are discussing. What we
really need to know is whether the low concentrations of those
pollutants that we find in most environments constitute a selective
pressure for the community of microorganisms that are present. This
raises additional questions of how do we measure adaptation and what
factors in the environment might be influencing whether or not
organisms can adapt.
5. Why do metabolic rates change when natural communities are confined?
Are the observed changes in rate due to alterations in the biomass
present, alterations in the species composition or organisms present,
or the result of changes in substrate concentration that result from
microbial activity.
92
-------
Is the assumption that the microbes are responding to only the added
substrate valid? This question is important for an evaluation of the
uptake and mineralization technique, since one of the assumptions we
make is that the organisms respond only to the compound added. I
think, however, this question has broader interest in terms of
evaluating the specificity of the enzymatic mechanisms available for
the degradation of pollutants. It is quite well known from laboratory
studies that many organisms have transport systems that function with
several different substrates. In environments where nutrients are
scarce, it would be to the organism's advantage to be able to utilize
a range of compounds with as few enzymes as possible. In our
particular case, this is significant because the kinetics may be
influenced by not only the substrate we add, but by other compounds
that may be present in the environmental sample.
93
-------
0.40--
. . 400
S-
o>
01
3.
QJ
0.30--
0.20--
300
200
T/F (
ra
4->
O
0.10..'
100
1 I I h-
30 60 90 120
Concentration (yg-liter~ )
Figure 1. M-cresol uptake by Jordan Lake microbial community on 6/22/82,
(A—A) - velocity, (•— •) - linear transformation.
-------
80 -
60 -
40 •
s-
CD
20
15
30
-1
800
. 600
. 400
T/F
(h)
200
Concentration (yg-liter )
Figure 2- Total uptake of ch'borobenzene by Lake Michie microbial community on 11/2/82,
(A—A) - velocity, (D-D) - linear transformation.
-------
Table 1. Correlation of nutrient concentrations and metabolic parameters
Coastal Systems
Lakes
Rivers
m-Cresol NTA ? Max Ami no
N03 + NO -N
NH4
P04
Tot Part. Wt.
Chlorophyl 1 a
+ = Correlated
* = Correlated
** = Correlated
*** = Correlated
Vm= vm,~ Uptake Acid Met
max max
.82*** .81** .76*** -.08
.69** .44 .59* -.22
.74** .80** .67** -.16
.08 -.11 .01 -.19
.85*** .06 .62** -.20
at 90? level
at 95? level
at 99? level
at 99.9? level
? Max Ami no ? Max Ami no
Uptake Acid Met Uptake Acid Het
.23 -.36 -.18 .07
.49 -.36 .11 .06
.49 -.39 .16 .01
- _
_ _
96
-------
Table 2. Enumeration of specific degraders for m-Cresol
in North Carolina rivers and lakes
Sample Source
Characteristic Haw EnoMichie Jordan
Total heterotrophs
(CPU x 104/ml) 23.0 14.5 13.8 9.5
m-Cresol degraders
(CPU x lOVml) 3.30 1.36 .30 4.0
% m-Cresol degraders .14 .09 .02 .42
Acridine Orange fi
Direct Count (x 10b/ml) 4.3 5.7 8.2 16.0
m-Cresol degraders microauto-
radiography (x KT/ml) 18.5 25.0 30.1 14.3
% m-Cresol degraders 4.3 4.3 3.7 0.89
97
-------
Table 3. Comparison of persistence predicted by heterotrophic uptake and time
course studies with Jordan Lake
% Metabolized ug Metabolized
Time ~T7v"
Compound (days) Prediction Observed Prediction Observed
Chlorobenzene 1 5.5 3.5-6 .033 .03- .07
4 22.1 18-39 .134 .06-.186
7 38.6 41-52 .235 .09-.237
14 77.2 32-57 .470 .12-.175
m-Cresol .5 8.4 .5-8 .09 .037-.087
1 16.8 2-16 .18 .082-.21
4 67.2 50-62 .72 .34-5.6
7 117.2 45-62 1.26 .34-3.7
14 235 45-62 2.52 .34-3.87
98
-------
HYDROCARBON BIODEGRADATION: KINETIC CONSTANTS AND THEIR APPLICATION
O.K. Button, K.S. Craig and B.R. Robertson
We have examined the persistence of hydrocarbons in aquatic environments
with focus on those hydrocarbons in the dissolved phase as they are
transformed and biodegraded by bacteria. Our main experimental approach
has been to follow radiolabeled hydrocarbons through both laboratory and
environmental systems with data interpreted from biochemical and kinetic
points of view. Our main interest is in the biological principles of
hydrocarbon transport mechanisms and kinetics.
Early work led to the development of a sensitive technique for
following the rate of C0£ liberation from hydrocarbons (Button et al.,
1981a). We were then able to locate sites of high degradative activity.
Rates were fast in Port Aransas, Texas, near industrialized Corpus Christi
(manuscript in preparation). In Port Valdez, Alaska, rates were even
faster in a thin layer at 50 m depth due to ballast water discharge
(Button, et al., 1981b).
In Resurrection Bay, Alaska, rates of toluene metabolism were
initially slow, but could be induced to higher values, perhaps due to
a microflora subsisting on conifer-derived terpenes (Button, Appl. Environ.
Microbiol., in press). By exposing seawater samples to a range of toluene
concentrations, we found that induction proceeded at its maximal rate
after 48 h and at half its maximal rate (Kinc|) at a substrate concentration
of 1.9 ug per liter (manuscript in preparation).
Microautoradiographic techniques, together with epifluorescence
microscopy, gave the original population of toluene oxidizers in this
pristine fjord. About 8% of the bacteria (27 yg/liter) had the capacity
to metabolize toluene. The process of induction raised the specific
affinity for toluene (aA) of this natural population from a base of 4.2
liters/g cells x hours to 11.6.
Induction raised the specific affinity of the marine isolate
Pseudomonas T2 from a base of 0.03 to 500. The saturation constants kj
for natural assemblages of bacteria in seawater and for Ps. T2 were
found to have values of 2 and 44 yg/liter, respectively, about the same
as the concentrations that gave half-maximal rates of induction, K-jnc|.
This similarity between Ky and K-jnc| is explained if toluene metabolism
proceeds according to constitutive systems with the above values of Kj
and then induces additional enzyme (such as toluene dioxygenase) at the
rate-limiting step. Then induction and transport are controlled by the
same process. A peculiarity of this system is that while the values for
specific affinity are quite usual, the values of Kj are very small
(manuscripts in preparation).
The formation of large quantities of metabolic products (type
compounds shown in Fig. 1) appears to be a phenomenon only of very dilute
microbial populations (Fig. 2), such as those in seawater. The recycling
step of product reaccumulation by the bacteria is subject to different
types of transport systems, perhaps of different classes (see below) and
99
-------
certainly with different kinetics. This phenomenon of dilute solutions
necessitates a correction in the basic assumption that the kinetic
constants for biodegradation are biomass-independent. The biomass dependency
can be demonstrated by numerically simulating steady state systems using
a continuous culture model and solving for biodegradation rates using the
kinetic constants for toluene utilization, together with the product
formation and re-utilization kinetics determined from 3-methylcatechol
metabolism experiments.
Basic formulations for biodegradation kinetics and their relation to
Monod kinetics are shown in Fig. 3 (Button, 1983). Also shown is the
relationship between the kinetics and enzyme content of the organism at
the rate limiting step.
Consideration of these kinetics has led to a model for the active
transport of hydrocarbons called vectorial partitioning. It is only the
fourth transport mechanism known and is shown in Fig. 4. Recent inhibitor
data support this mechanism.
The understanding of biodegradation in aquatic systems has increased
dramatically over the last few years. Four areas in need of additional
work are:
1) Concentration-dependent kinetics and specificity of induction,
particularly the extent to which hydrocarbons induce low-level metabolic
activity in the bulk of the natural aquatic microflora when the inducers
are sustained at very low concentrations over long periods. These data
are useful because they reflect the concentrations of a pollutant which
aquatic organisms are experiencing but which are rather difficult to
measure. Also data define the concentrations of one hydrocarbon that
would be expected to assist in priming a system for the biodegradation of
another.
2) The nature and kinetics of ampholite degradation. While apparent
Michael is constants for the transport of hydrocarbons seem to differ
radically from those for oxygenated substrates due to different mechanisms
of accumulation, many anthropogenic hydrocarbons as well as their products
of biodegradation are only mildly polar. Little is known of the decomposition
kinetics of these compounds as they move from recognition by non-polar
transport mechanisms to being accumulated by transport systems for polar
substrates. For example, if an ampholite is accumulated by vectorial
partitioning, it may be of little use to study its degradation at
concentrations above nanomolar levels.
3) The chemical nature of organic products formed from hydrocarbon
biodegradation. Apparently hydrocarbon biodegradation, as studied in the
laboratory, contains metabolic loops which proceed outside the cell. In
natural aquatic environments the biomass is too low for product
reaccumulation to occur, so these products are lost, generating a range
of unusual organic chemicals which can only be biodegraded with what may
be rather low-affinity transport mechanisms.
100
-------
4) The mechanism for active transport of non-polar substrates. If
vectorial partitioning is the main mechanism for hydrocarbon transport,
knowledge of this mechanism is useful for accurate anticipation of how
the process will proceed for new compounds without having to test each alone,
101
-------
REFERENCES
Button, O.K., Schell, D.M., and Robertson, B.R., Appl. Environ.
Microbiol. 42, 936-941 (1981a).
Button O.K., Roberts, B.R., and Craig, K.S., Appl. Environ.
Microbiol. _42,_ 709-719, (1981b).
Button, O.K., Trends Biochem. Sciences 8, 121-124 (1983).
102
-------
3 -Methyl
catechol
<*-Hydroxy-
keto acid
o-Cresol
1. Chemical nature of products formed from toluene by Pseudomonas T2.
0)
o
o:
c
o
•^
0
E
L_
.c
k_
Q>
V.
*Q>
C
Q>
_3
O
4
3
2
100
200
300
Toluene (A), ^g/liter
2.
Formation of organic products (P) and carbon dioxide (Q) as a function
of initial toluene concentration by Pseudomonas T2.
103
-------
Polar Substrate
Non Polar
Michaelian
C = k —
X
V
a =
X
max
KT
V
max
- v
K
T
J_
X
Monod // = V
m 3 x m 3 x
i/
M ~ KT
Y = constant
3. Formulation of biodegradation kinetics based on the specific affinity and
second order kinetics.
Chemiosmotic
A,
Group
Translocotion
Ligand Taxi
Cout
/
.....
'in
Vectprial Dout-
Partitioning \
-OH-^DOHin
4. Model for transport of hydrocarbons into bacteria by vectorial partitioning
together with other models for active transport.
104
-------
A HETEROTROPHIC BIODEGRADATION POTENTIAL (HBP)
APPROACH FOR PREDICTING BIODEGRADATION RATES
IN NATURAL ECOSYSTEMS
Robert J. Larson
ABSTRACT
Numberous investigtors (1, 4, 5, 10, 12-14) have shown that
the rate of uptake and/or mineralization of radiolabeled natural
substrates in aquatic systems can be described by a hyberbolic
enzyme-saturation model of the Michaelis-Menten type (9).
Recently, saturation kinetic models have also been applied to
biodegradation of various xenobiotic compounds in ground and
surface water systems (6, 11). The application of saturation
models to xenobiotic biodegradation has important implications for
environmental fate research. These models directly relate the
rate of biodegradation of a specified substrate to the activity of
the degrading microbial population. As such, they may be useful
in normalizing microbial numbers for microbial activity and
extrapolating biodegradation rate data to different environmental
systems.
THEORY
Tl]fi general equation relating the rate of uptake (and/or mineralization)
of a C-labeled substrate to the concentration of added substrate (when the
background substrate concentration is unknown) can be written as:
K + (Sb + A) [1]
where v is the degradation rate (ng/liter/h), V is the maximum degradation
rate (ng/liter/h), Su is the background substrate concentration (yg/liter), A
is the added substrate concentration (pg/liter) and K is the half-saturation
constant where v = 0.5 V (ng/liter). When the initial degradation rate is
constant, v can also be defined as:
v = f_ (Sb + A)
t - [2]
105
-------
where f is the fraction taken up and/or mineralized at time t (h).
Traditionally, equations [1] and [2] have been combined, inverted, and
rewritten as follows to estimate kinetic constants (8):
1 (Sb + A) . Vmax (Sb + A)
t K + (Sb + A) [3]
' = "may
t (K + Sb) + A [4]
1 = l/Vmax (A) + (K + Sb)
[5]
A plot of t/f vs. A (equation [5]) yields a straight line with y-
intercept T , x-intercept (K + Sb), and a slope of 1/V (Figure 1). Tp is
defined as the time required for complete degradation or a chemical at its in
situ concentration, V is the extrapolated maximum degradation rate ~aT
infinite substrate concentration, and (K + S, ) represents the sum of the half
saturation constant and the background substrate concentration. Although
unweighted linear regression of t/f vs. A (equation [5]) is commonly used to
estimate kinetic constants, direct nonlinear regression analysis of the
uninverted equation (equation [4]) can also be used to estimate these
constants. Nonlinear techniques offer special advantages over standard linear
techniques, in that they avoid many of the statistical biases inherent in
linear transformations of nonlinear data (8). They also give better estimates
of kinetic constants than linear techniques and are readily available to
anyone with access to a microcomputer.
Equations [4] and [5] have traditionally been used to determine kinetic
constants when A > S. . A special situation occurs, however, when S^ is either
much greater than A or so low as to be considered negligible. When Sb » A,
turnover times can be calculated directly from the ratio t/f using a single
concentration of A. The concentration of added substrate is so low that it
does not influence the in situ degradation rate (3), and observed Tn values
represent the actual turnover time for the background level of substrate
present. Since only one substrate concentration is tested by this technique,
however, V , and K + $b values cannot be calculated. When Sb « A, as would
be the case when testing xenobiotics in previously unexposed systems, $b can
be considered zero for all practical purposes and ignored when combining [1]
and [2]. Under these conditions, the following equation can be written:
v = 1. A= Vmax • A
t K1 = A [6]
106
-------
where the half saturation constant is written as K' to differentiate it from
the half saturation constant (K) of equation [5] which includes an S^ term.
Strictly speaking, turnover times have no meaning when S^j-O. Therefore, in
systems where the in situ concentration of xenobiotic is negligible, T values
cannot be calculated, although a ratio t/f may be obtained. However, equation
[6] can be used to determine the empirical kinetic constants, V and K'.
These constants indicate the "potential" for biodegradation of specific
xenobiotics in natural ecosystems and thus have use in defining the overall
heterotrophic biodegradation potential (HBP) of the natural microbial
population present. When some background level of the substrate is present,
T can be calculated from equation [4] by the relationship T = K"/V , where
K" = K + Sb (8) and Vmax has units of ug/liter/h.
GENERATION OF KINETIC CONSTANTS
The ability to generate HBP kinetic constants has important implications
for biodegradation testing of xenobiotics. Use of these constants with other
measurements of microbial biomass such as colony forming units, CPU (2),
acridine-orange-direct-counts, AODC (15), and most-probable-number (MPN)
determinations (7) should allow microbial numbers and degradation activity to
be correlated directly via a specific activity index (SAI). The concept of
SAI was initially proposed to characterize the activity of individual cells on
natural substrates (13). Turnover ratee (T = i/Tr.) or vmax values were
divided by bacterial numbers (CPU, AODC) to yield Tp or V SAI's. In
principle, a similar approach should be applicable to degradation of
xenobiotic compounds, if these compounds are mineralized as carbon and/or
energy sources.
Figures 2, 3, and 4 illustrate the results of HBP studies on three
detergent chemicals, linear alkylbenzene sulfonate (LAS), sodium
nitrilotriacetate (NTA) and dodecylnony1ethoxylate (CjoEg) in ground and
surface waters. All three materials were testea over a range of
concentrations and rate data (v = f/t . A) and were fit by nonlinear
regression techniques to equation [6] to estimate the kinetic parameters K'
and vmax* Vjnax was then dlvided b^ the number of viable bacteria (CPU/liter)
to generate Vmax SAI's in the different environmental samples tested.
As the data in Figures 2-4 indicate, biodegradation rate data for all
three materials are accurately described by a saturation kinetic model
(equation [6]). Precise estimates of the kinetic constants K1 and V can be
obtained (Table 1), even though maximum degradation rates vary for the
detergent chemicals and natural water samples tested. Moreover, normalization
of Vj7iax for CPU resulted in SAI values that agreed relatively well, both
within and between compound classes. This good agreement suggests that
consistent kinetic results can be obtained if degradation data are normalized
for both microbial numbers and microbial activity using an HBP/SAI approach.
FUTURE RESEARCH NEEDS
HBP kinetic approaches which incorporate both microbial numbers and
microbial activity measurements (SAI) appear to show promise for predicting
107
-------
biodegradation rates of radiolabeled xenobiotics in natural ecosystems. In
theory, these approaches should be applicable to die-away studies in which
disappearance of parent compound is measured by specific analytical methods.
More research on different xenobiotic compounds is needed, however, to
establish the applicability of HBR measurements to analytical studies. In
addition, several important research areas need to be addressed to determine
the predictive value of an HBP/SAI approach in estimating biodegradation rates
in the environment. These areas include (but are not limited to) the:
- effect of incubation period on kinetic measurements,
- significance of T value when S^ ^0
- appropriate biomass measurements for generating SAI values,
- and comparability of rate data derived from short-term HBP assays vs.
long-term biodegradability die-away assays.
Information is needed in these areas not only to formulate a conceptual
framework for using microbial activity measurements in environmental fate
research but also to establish a practical basis for predicting biodegradation
rates of xenobiotics in natural ecosystems.
LITERATURE CITED
1. Azam, F. and 0. Holm-Hansen. 1973. Use of tritiated substances in the
study of heterotrophy in seawater. Limnol. Oceanogr. 23:191-196.
2. Buck, J. F. 1979. The plate count in aquatic microbiology. Native
Aquatic Bacteria: Enumeration, Activity and Ecology, ASTM STP 695,
J. W. Costerton and R. R. Colwell, (Eds.), American Society for Testing
and Materials, pp. 19-28.
3. Gocke, K. 1977. Comparison of methods for determining the turnover times
of dissolved organic compounds. Mar. Biol. 42:131-141.
4. Hobbie, J. E. and C. C. Crawford. 1969. Respiration corrections for
bacterial uptake of dissolved organic compounds in natural waters.
Limnol. Oceanogr. _l£:528-532.
5. Ladd, T. I., R. M. Ventullo, P. M. Wallis, and J. W. Costerton. 1982.
Heterotrophic activity and biodegradation of labile and refractory
compounds by groundwater and stream microbial populations. Appl.
Environ. Microbiol. 44:321-329.
6. Larson, R. J. 1983. Kinetic and ecological approaches for predicting
biodegradation rates of xenobiotic organic chemicals in natural
ecosystems. In M. J. Kluq and C. A. Reddy (Eds.), Current Perspectives
in Microbial Ecology. American Society for Microbiology, pp. 677-686.
108
-------
7. Lehmicke, L. G. , R. T. Williams, and R. L. Crawford. 1979. 14C-most-
probable-number method for enumeration of active heterotrophic
microorganisms in natural waters. Appl . Environ. Microbiol. 38:644-
649.
8. Li, W. K. W. 1983. Consideration of errors in estimating kinetic
parameters based on Michaelis-Menten formalism in microbial ecology.
Limnol. Oceanogr. j?8_:185-190.
9. Michaelis, M. and M. L. Menten. 1913. Kinetics of invertase action.
Biochem. Z. 49:333-369.
10. Parsons, T. R. and J. D. Strickland. 1962. On the production of
particulate organic carbon by heterotrophic processes in sea water.
Deep-Sea Res. 8:211-222.
11. Pfaender, F. K. and G. W. Bartholomew. 1982. Measurement of aquatic
biodegradation rates by determining heterotrophic uptake of radiolabeled
pollutants. Appl. Environ. Microbiol. 44:159-164.
12. Wright, R. T. and J. E. Hobbie. 1966. Use of glucose and acetate by
bacteria and algae in aquatic ecosystems. Ecology 47:447-464.
13. Wright, R. T. 1978. Measurement and significance of specific activity in
the heterotrophic bacteria of natural waters. Appl. Environ. Microbiol.
14. Wright, R. T. and B. K. Burnison. 1979. Hetertrophic activity measured
with radiolabeled organic substrates, pp. 140-155. _Ir^ J. W. Costerton
and R. R. Colwell (Eds.), Native Aquatic Bacteria: Enumeration,
Activity, and Ecology, ASTM Technical Publication No. 695, Amer. Soc.
Testing Material, Philadelphia, PA.
15. Zimmermann, R. and L. Meyer-Reil. 1974. A new method of fluorescence
staining of bacterial populations on membrane filters. Kiel.
Meeresforsch 30:24-27.
109
-------
Table 1. HBP kinetic constants for biodegradation of detergent chemicals in
ground and surface waters.3
Compound
LAS
NTA
C12E9
v
Water (ng/liter/h)
surface 539
(429-649)
ground 138
(111-165)
ground 121
(105-136)
surface 8081
(4575-11587)
K1
(ng/liter)
902
(546-1259)
63
(32-94)
388
(256-520)
504
(98-1008)
SAI
r2 (ng/cell/h)
0.99 1.5 x 10"5
0.99 1.7 x 10"5
0.99 3.3 x 10"5
0.99 3.0 x 10"5
a Values in parentheses are 95% confidence intervals of true means.
110
-------
1/Vmax (A)
y-intercept = T
max
max
-10
x-jntercept
Figure 1. Linear transformation of hyperbolic saturation curve
(equation (4)) as derived by Wright and Hobbie (12).
Ill
-------
LAS RAPID CREEK RIVER WATER
Dotted Contours Are 95.0% Conf. Limits Of True Mean
480-
400-
320-
O
o
LU
SUBSTRATE CONCENTRATION (yg/l) x 101
Figure 2. Rate or degradation of LAS in river water as a function
of LAS concentration. Data have been analyzed by equation
(6), and parameter estimates are summarized in Table 1.
112
-------
NTA CANADIAN GROUND WATER
Dotted Contours Are 95.0% Conf. Limits Of True Mean
O)
c
O
g
LU
120-
100-
40
80
240
SUBSTRATE CONCENTRATION
Figure 3. Rate of degradation of NTA in groundwater as a function
of INTA concentration. Data have been analyzed by equation
(6), and parameter estimates are summarized in Table 1.
113
-------
C12E9 OHIO RIVER WATER
Dotted Contours Are 95.0% Conf. Limits Of True Mean
560
C E OHIO GROUND WATER
Dotted Contours Are 95.0% Conf. Limits Of True Mean
0 20 40 60 80 100 120
SUBSTRATE CONCENTRATION Cug//) x 101
0
20 40 60 80 100 120
SUBSTRATE CONCENTRATION (fjg/l) x 101
Figure 4. Rate of degradation of Ci2E9 concentration. Data
have been analyzed by eqdation (6), and parameter
estimates are summarized in Table 1.
114
-------
PANEL REPORT
The purpose of this panel was to evaluate the applicability of the short-
term (approximately 24 hours) heterotrophic potential method to the
biodegradation of xenobiotic compounds. Before discussing the applicability
of heterotrophic potential measurements, we believe it is crucial that certain
terms be defined to avoid confusion and problems with semantics.
The technique commonly referred to as heterotrophic potential is an
activity assessment based on saturation kinetics in which metabolic velocity
is measured over a range of substrate concentrations. In addition to yielding
a velocity at each concentration added, the maximum potential velocity (Vma )
that the natural community may attain can be established. The technique is
based on the assumptions specified by Michaelis-Menten which require short-
term incubations to avoid significant changes in substrate concentration,
active biomass, and the enzymatic rate 1 imitating step during the tes.ting
period. Other useful biodegradation parameters can also be obtained from this
data set. These parameters are defined below.
1. Specific Activity Index is a measure of the amount of microbial
activity per unit biomass in which the biodegradation rate is divided
by the number of organisms active on that compound. The units are
rate/unit biomass.
2. V is the rate of biodegradation at saturating substrate
concentrations; the units are g substrate/liter'hr.
3. Turnover time is the reciprocal of turnover rate, which is the
fraction of material biodegraded at subaaturating substrate concentrations
divided by time, in units of time.
4. Kt is the concentration of a compound at half V in units of grams
\f - . nid A
per liter.
There are several advantages to the heterotrophic potential testing
method to survey the biodegradabi1ity of xenobiotics.
1. The test is relatively simple to perform.
2. A wide range of concentrations of the compound can be tested in a
relatively short period.
3. These substrate concentrations approximate realistic environmental
concentrations.
4. The test is short-term (approximately 24 hours); therefore changes in
certain environmental parameters (biomass, etc.) do not interfere with
subsequent data analyses.
5. There are statistical techniques (nonlinear regression) available for
data analyses.
115
-------
6. Several useful biodegradation rate constants and kinetic constants can
be calculated
7. The procedure is cost effective compared to long-term testing now
performed.
As with any test method, there are also disadvantages which can restrict
the uses of this method:
1. If "adaptation" by the microbial community is required, it may not be
detected.
2. A radiolabeled compound is usually required to perform the tests at
environmentally realistic concentrations.
3. For the test to be valid, it must be run at a range of concentrations,
including levels well in excess of K..
4. Kinetically diverse microbial assembleges are not strictly Michaelian.
Further research is needed to assess the applicability of heterotrophic
biodegradation rates in complex environmental systems. These research areas
are:
1. Expansion of the data base by application of the method to other
ecosystems, in particular soil and subsurface environments, by using a
range of compounds (benchmark compounds) with different
physical/chemical properties.
2. Definition of the range of compounds for which the method is
applicable by systematic testing of different classes of chemicals.
3. Determination of the most "appropriate" measure of microbial biomass
for estimating numbers of substrate utilizers and Specific Activity
Indices.
Also, the panel recommends the following:
1. Data analysis should be performed by nonlinear regression techniques
to estimate kinetic constants and associated statistical parameters.
2. In all testing the importance of the dilutent must be emphasized.
3. Rate constants and kinetic constants obtained from different testing
protocols should be compared.
4. Future improvements in methods and theory should be incorporated into
the assessment of biodegradability as technology permits.
116
-------
Panel Members: Peter Chapman, University of Minnesota
Doris Paris, U.S. EPA
Gerald Neimi, University of Minnesota, Duluth
Dave Gibson, University of Texas
Bill Gledhill, Monsanto
Phil Howard, Syracuse Research Corporation
RELATIONSHIPS BETWEEN CHEMICAL STRUCTURE AND BIODEGRADATION RATES
Objective: To discuss concepts for relating the biodegradation rate of a
chemical to its structure.
Because of the extremely large number of chemicals which the Office of
Toxic Substances will be required to regulate, it is important to evaluate
structure-activity relationships (SAR) as a potential tool for applying
biodegradation information for representative chemicals to other similar
chemicals for which relatively little fate information is available. Current
SAR studies use BOD's and persistence testing data to compare chemical
structure and chemical properties to biodegradability. These data are
frequently generated using sewage sludge or other environmentally unrealistic
inocula. Can biodegradation rate information from studies using natural
inocula and intact microbial communities be used successfully in SAR
studies? What limitations and disadvantages are inherent in this approach?
What methods and quantitative analysis of the data should be used for
obtaining biodegradation rate information? Is the information too site-
specific? What chemical classes should be tested? What aspects of chemical
structure or chemical properties should be considered?
117
-------
STRUCTURE-ACTIVITY RELATIONSHIPS (SARs)
P. J. Chapman
ABSTRACT
In attempting to relate activity to structure in
biodegradation studies, the object is to determine whether one can
perceive in the structure of a given chemical (or in certain of
its properties) features which reveal its biodegradability or its
persistence. Put another way, we might ask are there ways of
looking at the structures or derived parameters of organic
compounds which can indicate whether their biodegradation is
permitted or constrained in the environment? This approach, if
successful, would of course be invaluable to the EPA Office of
Toxic Substances in its evaluation of the numerous chemicals for
which regulatory decisions are required.
EVALUATION OF ASSUMPTIONS
Is it possible that for kinetic purposes we can treat the complex
microflora of different environments as if they are pure cultures? Or, to
take the analogy one step further, as if we were dealing with a purified
enzyme system challenged by a range of potential substrates? This latter
analogy at least enables us to see that we can apply this information to a
new, potential substrate in an informed way once we find the answers to the
following questions:
(i) What is the mechanism of the reaction catalyzed?
(ii) What is the range of compounds serving as substrates and at what
relative rates for given concentrations?
(iii) What groups in the substrate are recognized by the active site and
what steric factors limit the range of substrates?
But can we now predict the rate of its attack? Or are we merely limited to
statements such as "it looks as if it ought to fit at the active site?"
We should realize that in other areas some degree of success has already
been achieved by this approach. Perhaps we can view the many different
environments of our planet in the same way that a pharmacologist views tissues
or target organs and asks: What is the fate of this drug in these tissues
when it is administered? Pharmacologists already have at hand the knowledge
about what methods of administration to use to best reach a given tissue and
what types of compounds will enter these tissues and undergo metabolic
transformation. In other words, the design of tissue-targeted drugs has
considerable rationale. As long ago as 1690 the British philosopher, John
Locke, wrote in his "Essay Concerning Human Understanding": "Did we know the
mechanical affections of the particles of rhubarb, hemlock, opium and a man as
a watchmaker does a watch we should be able to tell beforehand that rhubarb
will purge, hemlock kill and opium make a man sleep." Note that Locke
recognized the need to know both the "mechanical affectations" of man as well
as those of different particles. The pharmacologists' success in relating
specific structural features of drugs to precise biological effects is based
118
-------
on exactly those principles. Their application 1s relevant to an assessment
of the toxlcological effects of chemicals in our environment, but we should,
however, realize that our level of understanding of the fate of chemicals in
our environment will be limited by what we know of the potential of the
environment to deal with chemicals of different types. Whether that can be
treated in such a systematic fashion remains to be seen. No matter whether
the assessment is toxicological or biodegradative, one is obliged to make
reference to a data base -- an organized body of information to which
reference can be made for precedents and examples whenever a new chemical
comes under the gaze of a regulatory agency. What would our assessment now be
of the early alkylbenzene sulfonate detergents in use in the late 1950's?
Hopefully we would now be able to recognize that the extensive branching of
the alkyl sidechains — a feature introduced by the synthetic methods employed
-- is an environmentally undesirable feature in such compounds because highly
branched alkanes show limited rates of attack and the detergent sidechains are
the usual sites of attack when biodegradation is initiated. Unfortunately we
had to experience the effects of these slowly degradable detergents before
research revealed that detergents with linear sidechains, the presently-used
LAS detergents, are much more readily biodegradable.
FEASIBILITY OF SAR APPROACHES FOR BIOOEGRADATION
t
The above example raises a question about the general feasibility of the SAR
approach -- our knowledge of the biodegradation of different classes of
organics has not begun to keep in step with the range of different synthetic
compounds in current use, not to mention those presently under development.
It would appear that at this time we do not possess a sufficiently adequate
data base from which to assess different chemicals. Thus we really have at
least two important questions to deal with. Is the SAR approach a feasible
one? If so, then what specific information is needed to make the approach
generally useful? If the answer to the first is in the affirmative, then it
is necessary to evaluate different approaches perhaps by determining the
predictive value of each in relation to subsequently performed laboratory and
field studies with a representative number of test chemicals.
FUTURE RESEARCH NEEDS
Whatever approach is adopted, a body of information on biodegradation
must be available and organized in such a way as to be useful and accessible
to all interested parties. Furthermore it should be based on' standardized,
widely acceptable procedures. Whether we rely on a large body of existing
biodegradation data such as the extensive BOD values in the literature, on
pure culture studies of pathways and biotransformations or, more
realistically, make use of some measure of the rate of disappearance of parent
compound, are matters for discussion. Personally I favor the latter approach
coupled wherever possible to some measure of the completeness of the
biodegradation process using, for example, percent conversion to carbon
dioxide or, in anaerobic situations, to carbon dioxide and methane. I think
it should be stressed that if one adopts disappearance of parent compound as a
measure of biodegradation, that an apparently rapid reaction such as,
hydrolysis of the methyl ester of 24D or of methyl parathion, will yield
readily degradable methanol plus the bulk of the molecule which may be
considerably more resistant to attack. In other words, we have to have an
approach that will also provide information about products so that these too
can be evaluated in turn.
119
-------
Additional questions about concentrations of test compounds, type of
inoculum, whether sediment, should be present or if nutrients should be added,
are all reflections of the type of environment we are trying to simulate in
the laboratory. Do we want to try for the best possible circumstances for
biodegradation or the worst possible scenario? To me there seems to be a very
real need to assess the potential for aerobic biodegradation in both types of
environments and to follow up with anaerobic studies for compounds which
emerge as persistent or slowly biodegradable. We could ask this question
another way by saying how far can we extrapolate findings from one set of
environmental conditions, inocula, concentrations, etc.
120
-------
A PRELIMINARY MODEL TO PREDICT BIODEGRADABILITY
FROM CHEMICAL STRUCTURE
G. J. Niemi
Ronald R. Regal
Dinah D. Vaischnav
Oilman D. Veith
ABSTRACT
The long-term success in understanding the impacts of
potentially hazardous chemicals in order to ensure environmental
protection is dependent, in part, on the development of predictive
models that reduce the need for cost-prohibitive tests on
individual chemicals. Under the assumption that chemicals of
similar structure have similar behavior, knowledge on the
structure of a chemical allows the estimation of its activity by
means of comparisons with chemicals of known activity. We present
the results of a preliminary analysis on the prediction of whether
a chemical is persistent or not persistent on the basis of
structure-activity relationships. We use a data base derived from
a literature search of the chemicals that have been tested for
biochemical oxygen demand (BOD) and apply multivariate analysis to
develop a prediction model.
APPROACH FOR PREDICTION OF BOD
We calculated three sets of molecular connectivity indices for those
chemicals with BOD data [framework, simple or bond-corrected, and valence-
corrected indices ( Kier and Hall, Molecular Connectivity in Chemistry and
Drug Research, Acadtmic Press, NY, 1976)],
Our approach for the prediction of BOD was a two-step process:
(1) identification of chemicals using similar structure with K-means
clustering (' Dixon and Brown, Biomedical Computer Programs, P Series,
Univ. of California Press, 1979) and (2) separate discrimination of
chemicals within a cluster of high BOD (defined as those with BOD values > 13)
and those with low BOD (those with BOD values <_ 13). We performed the cluster
analysis with principal components that were calculated for each chemical from
a separate principal component analysis of 45 variables (a combination of
variables calculated from the 3 molecular connectivity indices) for 16,121
chemicals. We assumed that this large data base was representative of the
multivariate space occupied by a substantial number of industrial chemicals.
For example, there are currently about 45,000 chemicals registered in the
Toxic Substance Control Act (TSCA) inventory.
MODEL DEVELOPMENT AND APPLICATION
A total of 340 chemicals (120 with low BOD and 220 with high BOD) was
identified from the data base assembled by Vaischnav. Because the K-means
clustering algorithm uses Euclidean distances to define clusters, chemicals
that are outliers in the data base have a major influence on the definition of
clusters. Therefore, we identified two major parts of the principal component
space, "outer" and "inner" space, and performed separate clustering analysis
on each. The outer space chemicals were defined approximately as those
chemicals > 2 standard deviations from the mean for any of the first 8
121
-------
principal components. This division resulted in the identification of 62
chemicals (18%) in the outer space and 278 (72%) chemicals in the inner
space. Subsequently, we identified 4 clusters (clusters 1-4, Table 1) in the
outer space group of chemicals and 6 clusters (clusters 5-10, Table 1) in the
inner space group of chemicals.
Three of the clusters (clusters 1, 2, and 5) required no discrimination
between high and low BOD groups because two of the outer space clusters
included only low BOD chemicals, while one of the inner space clusters
included 95% (18/19) of chemicals with high BOD. Discrimination between high
and low BOD chemicals in the remaining 7 clusters was performed with stepwise
discriminant function analysis (DFA) (Dixon and Brown, op. cit.). We included
variables up to 6th order from each connectivity index separately in the DFA
to identify those variables or combinations of variables that best separated
the groups. In the summary (Table 1) we selected those variables that
provided the best discrimination in terms of the number of chemicals correctly
classified into the low BOD group.
The clustering/discrimination model resulted in the correct
classification of 78% (92/120) of the chemicals identified as having low BOD
and 77% (170/220) of the chemicals identified as having high BOD. However,
there is a differential probability of correct classification based on which
part of the subspace or cluster the chemical is assigned. For example, if a
new chemical is classified into the outer space group of chemicals, then there
is a high probability (about 94%) that the chemical can be correctly
classified as having a high or low BOD. Similarly, if a chemical is
classified into cluster 3 of the inner space, then there is about an 82%
chance of correct classification. In contrast, the discrimination
probabilities in cluster 4 or 5 of the inner space for the low BOD groups are
only slightly above 50%.
FUTURE RESEARCH NEEDS
It is apparent that chemicals are persistent because of a variety of
microbial mechanisms that relate to degradability. Therefore, a global
approach that puts all persistent chemicals in one group and all non-
persistent chemicals in another, such as a discriminant analysis of all
chemicals considered simultaneously, is unsatisfactory. Our second approach
was to classify chemicals into similar, relatively homogeneous groups, and
then search for the common structural elements that separated persistent from
non-persistent chemicals. In general, this preliminary model is relatively
good for the prediction of persistence, but we believe that the model can be
improved considerably. The following should be considered in subsequent
applications.
1. We suspect errors in the BOD values due to a lack of acclimation of
chemicals or due to test measurements. Re-evaluation and re-testing
of chemicals will be necessary to clear potential problems in the data
base.
2. The number of possible combinations of clusters and variables that can
be used for 340 chemicals is enormous. If chemicals within a cluster
have low BOD but different structural reasons for being persistent,
122
-------
then there is no guarantee that discrimination analysis can recognize
these structural differences. A variety of clustering models using
combinations of different numbers of clusters and variables needs to
be evaluated.
3. We used the multivariate techniques described here as exploratory
tools to elucidate structural configurations that may be associated
with biodegradation. Therefore, we included variables with
relatively low F-ratios to enter into the discrimination equation.
Subsequent investigations, preferably with a larger sample size,
need to be attentive to potentially spurious correlations in the
process of screening a large number of variable for inclusion in
a model. For this reason we have emphasized the potential
usefulness of this overall strategy and have not focused on
specific structural attributes that were associated with
discrimination.
123
-------
Table 1. Summary of cluster analysis and discriminant function analysis for
340 industrial chemicals.
SPACE CLUSTER ANALYSIS
N-Chemi
High
BOD
Outer 1 0
2 0
3 5
4 13
Inner 5 18
6 35
7 10
8 60
9 29
10 50
cals
Low
BOD
6
16
13
9
1
18
12
15
14
16
DISCRIMINATION ANALYSIS
% Correctly
Classified
High Low Variables Included
BOD BOD -F Ratio
-
_
60 85 SCL4
SCL6
85 78 FCL5
FP6
FP4
FCL3
_
60 78 SPC6
SPC5
80 83 VPO
VP1
VCL4
VCH6
83 53 VPO
VPS
VP4
VCL3
72 64 SCL5
SCH5
SPC4
SPC5
76 75 SP1
SP3
SP5
SCH6
- 4.95
- 2.03
- 4.92
- 5.40
- 2.84
- 1.28
- 1.18
- 2.67
- 1.16
- 2.23
- 3.83
- 2.27
- 4.78
- 7.60
- 3.04
- 1.37
- 2.63
- 3.79
- 2.55
- 1.65
- 2.63
- 4.12
- 1.42
- 5.91
Total
220
120
77
78
124
-------
PANEL REPORT
To obtain information on the biodegradation of representative compounds
for a data base by which to assess the biodegradation of many other similar
chemicals of interest to OTS, we employed two general approaches. The first
of these makes use of pure cultures, an approach that reveals only the genetic
potential of microorganisms to bring about biodegradation. It does not supply
information about the role of these organisms under environmentally realistic
conditions. The second approach employs mixed cultures under conditions that
simulate environmental circumstances and provides biodegradation information
such as the kinetics of compound disappearance and the need for acclimation
periods. It should be stressed that the only test systems from which reliable
data extrapolation can occur are those that employ mixed cultures and
environmentally realistic conditions.
In the view of this panel, it does not appear to be possible to predict
biodegradation rates from structure-activity relationships (SAR). For a small
category of molecules, however, it does appear possible to make qualitative or
even semi-quantitative judgments. The limitation is due to lack of a
sufficient data base. Compounds which are identifiable as the simpler members
of discrete chemical classes, such as aromatic hydrocarbons, esters, phenols,
and amides, generally fall into this category.
If we are to explore whether predictions of a qualitative or perhaps at
best a semi-quantitative nature are possible, it is necessary to enlarge and
extend the base of information from which extrapolations might be made. In
this connection it is important that an acceptable procedure is adopted by a
majority of workers in this area. Use of BOD data has suggested that SAR
correlations can be profitably pursued but it appears that a more acceptable
methodology should provide a reliable measure of primary biodegradation under
conditions simulating different environments and with regard to ranges of
substrate concentration variations such as temperature, aerobic/anaerobic
conditions, presence of sediment, eutrophic versus oligotrophic water, etc.
This panel recommends that such a procedure be adopted for the purposes of
generating information useful for SAR considerations but wherever practicable
should be accompanied by parallel studies of rate measurements of ultimate
biodegradation.
Having generated a body of primary biodegradation data for various
classes of chemicals, it should be evident what chemical or physical
parameters might logically be explored for correlations that might provide
indicators of biodegradation. Chemicals should be selected so that they can
be grouped into different clusters based on chemical classes by functional
groups. within these clusters applications of statistical techniques can
establish whether significant correlation exists between structure and
biodegradation rates.
We suggest that the following are workable approaches which should be
implemented into research plans:
1. Choose optimal test systems acceptable for wide-spread use.
2. Choose chemicals or classes of chemicals from a broad cross-section of
those representative of different classes of environmental interest.
125
-------
3. Determine whether biodegradation data for simple compounds can be
related to more complex structures possessing several of the same
structural entities.
126
-------
Panel Members: Gary Klecka, Dow Chemical
Dick Raymond, Groundwater Environmental Consultants
Jim Spain, Georgia State University
Harvey Holm, U.S. EPA
John Walker, U.S. EPA
Robert Brink, U.S EPA
Art Kaplan, U.S. Army-Natick
EXTRAPOLATION OF LABORATORY ftlODEGRADATION DATA TO THE FIELD
Objective: To develop a conceptual and experimental framework for testing the
usefulness of kinetic expressions of biodegradation processes in
predicting that biotic fate of chemicals in complex natural
situations.
Decay curves are commonly obtained from laboratory experiments in which
little attempt is made to maintain conditions similar to those in the field.
Resulting biodegradation rates may, therefore, be potentially inaccurate
estimations of rates in the field. What should be done to test the
environmental significance of these rate determinations? Are field studies
required? How should field studies be designed? What is the minimum size and
scope for a field test? What additional environmental factors should be
examined for their effect on biodegradation? How will laboratory-derived
rates and their associated kinetic expressions be used to predict
biodegradation in the field where the effect of variations in environmental
factors may be additive or integrative? What is the role of mathematical
models in relating laboratory data to the field?
127
-------
EXTRAPOLATION OF LABORATORY BIODEGRADATION DATA TO THE FIELD
Gary M. Klecka
ABSTRACT
Studies on the biodegradation of chemicals in the environment
are extremely complex. In addition to microbial transformation,
other environmental processes such as volatilization,
photodecomposition, hydrolysis, and adsorption are likely to
influence the concentration of a particular chemical. As a
result, model systems are usually established in the laboratory to
study the decomposition of a compound under controlled conditions,
and the rates observed are then extrapolated to describe the
process in the field. Since the key to successful extrapolation
ultimately depends on the laboratory data, special attention must
be paid when designing and conducting experiments such that the
information obtained will enable one to make a reliable estimate
of the rates at which a process will occur in the environment.
FACTORS CONTROLLING THE FATE OF CHEMICALS IN THE ENVIRONMENT
r
Fundamental to predicting the biodegradation of a chemical in ttfe
environment is the identification of the environmental compartments in which
the chemical is likely to reside. To address this question, a number of
simple equilibrium models have been developed, such as the one recently
described by Neely and Mackay (6). Once the partitioning of the compound is
known, the importance of biodegradation in controlling the fate of the
chemical in the environment will require an understanding of: (a) the
mechanism(s) involved in decomposition, (b) the rates at which these processes
occur, and (c) factors which are likely to influence the rates.
Laboratory studies with pure cultures of microorganisms and individual
compounds have revealed considerable information regarding the mechanisms
involved in the biodegradation of a wide variety of organic chemicals. Many
of these reactions can be classified as oxidation, reduction, hydrolysis, or
conjugation. One of the most important features of microbial degradation is
the ability of aerobic microorganisms to catalyze the incorporation of
molecular oxygen into the substrate (4). Thus, the bacterial degradation of
many aromatic molecules has been shown to require molecular oxygen for both
hydroxylation and ring-fission reactions. Alternatively, a number of
compounds, including several of the organochlorine insecticides, have been
shown to be degraded more rapidly under anaerobic conditions (7). Thus, a
knowledge of the mechanisms involved in microbial metabolism provides an
indispensable guide for understanding the fate of organic chemicals in the
envi ronment.
In order to define the role of biodegradation in determining the
environmental concentration of a particular chemical, a quantitative
expression describing the rate of the reaction is required. The microbial
degradation of a number of chemicals in natural waters and soils has been
shown to be described by first-order kinetics, and thus the rates of these
biological reactions can be expressed in terms of rate constants. Since
128
-------
first-order rate constants are often subject to considerable variability, the
extrapolation of rate constants to different systems will require an
understanding of the importance of microbial numbers or activities in
determining the reaction rate. One approach to this problem involves
normalizing rate constants by the size of the total microbial population as
enumerated by conventional techniques (2). However, a considerable number of
studies have shown that biodegradation rates do not consistently correlate
with any measurement of the total microbial population (1, 3, 10). Thus,
further studies are necessary to determine the relationship between reaction
rate constants and the concentration or activity of the specific
microorganisms responsible for degradfng a compound.
The extrapolation of laboratory-derived rate constants to the environment
will also require an understanding of the various biological, chemical, and
physical factors that influence biodegradation rates, and the integration of
these effects into mathematical models. Biological variables include
concentration, spatial distribution, species composition, activity, and
previous history of the microbial population. In addition, biodegradation may
be affected by interactions among members of the community such as mutualism,
commensalism, competition and predation, although little is known about the
influence of microbial interactions on degradation rates. Substrate
concentration may also be a significant factor affecting the susceptibility of
an organic compound to microbial attack. There is some evidence suggesting
the existence of threshold concentrations below which biodegradation does not
occur; however, the data are inconclusive. The biodegradation of organic
chemicals is known to be affected by a number of environmental variables,
including sorption, mixing, temperature, oxygen concentration, redox
potential, pH, ionic strength, and the presence of organic and inorganic
nutrients. Unfortunately, the significance of many of these variables is
poorly understood, and few quantitative relationships are available. These
effects will need to be evaluated if laboratory data are to 'be successfully
extrapolated to predict the fate of chemicals in the environment.
USE OF LABORATORY DATA TO PREDICT FATE IN THE FIELD
In spite of the many uncertainties, several investigators have had
considerable success in using laboratory biodegradation data to predict the
fate of a chemical in the environment. The key to these initial attempts
involved the application of site-specific rate constants. This approach was
used by Games (5) for modeling the fate of linear alkylbenzene sulfonate (LAS)
in Rapid Creek, South Dakota. LAS is introduced into the stream from a single
point-source input of domestic sewage. The choice of LAS as the test
substrate simplified the modeling process, since biodegradation has been
identified as the principal transformation mechanism. Rate constants were
measured using water and water-sediment mixtures obtained from sites
downstream from the Rapid City wastewater treatment plant. The U.S.
Environmental Protection Agency Exposure Analysis Modeling System (EXAMS) was
then used for predicting the fate of LAS in Rapid Creek, and the results were
compared to actual LAS concentrations in the stream below the treatment
plant. Steady-state concentrations of LAS in the water and sediment predicted
using the EXAMS model agreed fairly well with the measured concentrations.
However, a sensitivity analysis of the model to errors in flow rate, aqueous
and sediment biodegradation rate constants, dispersion coefficient and
129
-------
adsorption coefficient revealed that the model calculations were highly
dependent on the least understood parameters, namely the dispersion
coefficient and the sediment biodegradation rate constant. Variation in
either of these parameters resulted in a significant deviation of the
predicted concentrations from the measured values. It is clear that a better
understanding of sediment biodegradation and sediment-water interface dynamics
is required to improve predictions of the fate of chemicals in the aquatic
envi ronment.
Walker (8) demonstrated the use of site-specific rate constants for
predicting the fate of the herbicide propyzamide in experimental field
plots. Laboratory studies conducted using soil samples obtained from the test
site indicated that biodegradation was first-order with respect to propyzamide
concentration, and the effects of temperature and soil moisture on the
reaction rate could be described using simple equations. Field studies were
conducted on two separate occasions, and test sites were treated to simulate
both surface applied and incorporated herbicide formulations. Measurements of
propyzamide concentration, soil temperature, and moisture content were
performed throughout the field experiment. For the computer simulations,
biodegradation rate constants were corrected using data obtained for
variations in soil temperature and moisture, and changes in herbicide
concentrations were predicted by integration of the decomposition rates
calculated for the entire experimental period. The decomposition of
propyzamide predicted using the computer model was found to correlate well
with the measured concentrations of both surface applied and incorporated
forrnul ations.
Walker (8) subsequently modified the basic simulation model to include
methods for estimating soil temperatures and moisture contents from standard
meteorological data. When used in conjunction with site-specific degradation
rate constants, the model has been shown to predict the environmental fate of
the pesticides simazine, atrazine, propyzamide, linuron, metamitron,
trifluralin, metribuzin, chlorthal-dimethyl, prometryne, asulam, and
napropamide (9). With certain compounds, such as simazine, atrazine,
propyzamide, and metribuzin, there was a tendency to underestimate the rates
of disappearance. With these pesticides, it appears that mobility of the
compound in the soil was responsible for the inconsistencies between predicted
and measured concentrations. Deviations were also noted in an attempt to
model the fate of surface applied formulations of napropamide, where loss of
the compound was affected by photodecomposition. These observations emphasize
the importance of accounting for all fate processes when attempting to model
the fate of chemicals in the environment.
130
-------
FUTURE RESEARCH NEEDS
In summary, predicting the role of biodegradation in determining the fate
of chemicals in the environment in many cases will involve the extrapolation
of laboratory data to the field. It is clear that the key to successful
predictions will require a fundamental understanding of:
1. mechanisms involved in decomposition
2. kinetics of biological reactions
3. relationship of microbial activity to reaction rates
4. environmental variables likely to influence the rates.
Although a considerable amount of research remains before reliable
extrapolations of laboratory data can be made on a routine basis, several
investigators have had considerable success by applying simple models in
conjunction with site-specific rate constants. It is likely that further
studies of this type will be important for improving our ability to predict
the fate of chemicals in the environment.
LITERATURE CITED
Bartholomew, G. W. and F. K. Pfaender. 1983.
temporal variations of organic pollutant
estuarine environment. Appl. Environ.
Influence of spatial and
biodegradation rates in an
Microbiol. 45:103.
Baughman, G. L.
expression of
Dickson, and
Chemicals in
, D. F. Paris, and W. C. Steen. 1980. Quantitative
biotransformation rate, p. 105. In A. W. Maki, K. L.
J. Cairns, Jr., Eds., Biotransformation and Fate of
the Aquatic Environment. Washington, D.C.: American
Society for Microbiology.
Bourquin, A. W., J. C. Spain, and P. H. Prichard. 1982. Biodegradation
activity correlations with biological and environmental variables.
Abstr. Annu. Meet. Am. Soc. Microbiol. N91.
Da g 1 ey, S.
biosphere.
1977. Microbial degradation of organic compounds in the
Survey Prog. Chem. 8:121.
Games, L. M. 1982. Field validation of exposure analysis modeling system
(EXAMS) in a flowing stream, p. 325. In K. L. Dickson, A. W. Maki, and
J. Cairns, Jr., Eds., Modeling the Fate of Chemicals in the Aquatic
Environment. Ann Arbor, MI: Ann Arbor Science Publishers.
Neel y , W. B. and D. Mackay. 1982. Evaluative model for estimating
environmental fate, p. 127. In K. L. Dickson, A. W. Maki, and J.
Cairns, Jr., Eds., Modeling the Fate of Chemicals in the Aquatic
Environment. Ann Arbor, MI: Ann Arbor Science Publishers.
Sethunathan, N. 1973. Microbial degradation of insecticides in flooded
soil and in anaerobic cultures. Residue Rev. 47:143.
131
-------
8. Walker, A. 1974. A simulation model for prediction of herbicide
persistence. J. Environ. Qual. 2/396.
9. Walker, A. 1978. Simulation of the persistence of eight soil-applied
herbicides. Weed Res. 18:305.
10. Wright, R. T. 1978. Measurement and significance of specific activity in
the heterotrophic bacteria of natural waters. Appl. Environ.
Microbiol. 36:297.
132
-------
EXTRAPOLATION OF
LABORATORY BIODEGRADATION DATA TO THE FIELD
Richard L. Raymond
ABSTRACT
Field experience utilizing biodegradation as a cleanup tool
for hydrocarbon contaminated soils and groundwaters has shown that
many factors, presently beyond the control of the operator, affect
the kinetics of degradation. It seems highly unlikely that any
laboratory test can be devised that will come close to describing
the kinetics to be expected in the field. A few examples will
illustrate the problem.
USE OF LABORATORY DATA FOR FATE PREDICTIONS IN THE FIELD
Numerous laboratory and pilot plant studies have shown that many pure
hydrocarbons and mixtures thereof are degradable at rates above
0.5 Ibs/gal/day. The conditions for this high rate of turnover optimize
oxygen demand, substrate and inorganic nutrient concentrations, temperature,
and physical mixing. Optimization is rarely achievable in the field. In
Table 1, Table 2, and Figure 1, the effect of substrate type and
concentration, soil composition, and probably adsorption on the rate of
biodegradation is illustrated for two biofarming projects.
Equally dramatic effects are shown in field situations for groundwater
cleanup of gasoline. Two examples are shown in Figures 2 and 3. Rates in
gallons of aquifer storage capacity are 0.003 and 0.005 Ib/gal/month,
respectively. In groundwater cleanup, oxygen transfer is the most limiting
factor in the rate of degradation.
LITERATURE CITED
1. Raymond, R. L., J. 0. Hudson, and V. W. Jamison. Land application of
oil. AIChE Symposium Series. Water, 1978. pp. 340-356.
2. Raymond, R. L., V. W. Jamison, and J. 0. Hudson. Beneficial stimulation
of bacterial activity in groundwaters containing petroleum products.
AIChE Symposium Series. Water, 1976. pp. 390-404.
133
-------
Table 1. Biodegradation rates observed in landfarming.
Substrate Loading Rate
Ibs/cu. ft. Ibs/cu. ft./month
Decane 0.012 0.0061
Octadecane 0.01 0.0070
Tetramethylpentadecane 0.01 0.0020
Biphenyl 0.01 0.0070
Phenanthrene 0.02 0.0100
Phenylundecane 0.01 0.0200
#2 Fuel Oil 1.6 0.1100
#6 Fuel Oil 1.6 0.0750
Crude Oil 1.6 0.0810
These rates are orders of magnitude less than predicted from laboratory data.
134
-------
Table 2. Percentage reduction in oil concentration
Location of Field Sites
Type of Oil
After one year:
Used crankcase oils
Crude oils
Home heating oils (#2)
Residual oil (#6)
After three years:
Used crankcase oils
Crude oils
Home heating oil (#2)
Residual oil (#6)
Marcus Hook,
Pennsylvania
69.2
54.2
86.0
48.5
90.6
87.3
98.3
89.2
Tulsa,
Oklahoma
73.8
77.5
90.0
65.5
99+
99+
99+
94.0
Corpus Christi,
Texas
60.8
54.2
86.0
59.4
79.8
78.0
94.0
85.5
Avera
67.9
61.9
87.3
57.8
89.8
88.1
97.1
89.5
135
-------
5
*
I2O
100
80
60
40
20
o>
E
120
100
80
60
40
20
6/1 7/1 8/1 9/1 IO/I
CH -(CH ) -CH
3283
6/1 7/1 8/1 9/1 IO/I
6/1 7/1 8/1 9/1 IO/I
120
IOO
80
60
40
20
6/1 7/1 8/1 9/1 IO/I
Figure 1. Examples of biodegradation under field conditions.
136
-------
120
100
z
- 80
1 60
40
20
160
140
_, 120
z loo
\ 80
E
60
40
20
204
T
6/1 7/1 8/1 9/1 (0/1
6/1 7/1 8/1 9/1 IO/I
Figure 1 (continued). Examples of biodegradation under field
conditions.
137
-------
Figure 2. Millville, NJ, biostimulation
Aeration capacity
Nutrient distribution
Nutrient composition and quantity
Ammonium sulfate
Sodium phosphates
Trace inorganics - Mg, Mn, Fe, etc.
Water control
Duration
Gasoline degraded
20 SCFM
Batch and continuous
1200 Ibs.
600 Ibs.
25 Ibs.
Injection and discharge to waste
6 months
1000 gals.
Figure 3. Whitemarsh Township PA, biostimulation
Aeration capacity
Nutrient distribution
Nutrient compostion and quantity
Ammonium sulfate
Sodium phosphates
Water control
Duration
Gasoline degraded
25 SCFM
Batch
58 tons
29 tons
Discharged to waste
One year
45,360 gals.
138
-------
COMPARISON OF £-NITROPHENOL BIODEGRADATION
IN FIELD AND LABORATORY TEST SYSTEMS
J. C. Spain
INTRODUCTION
Laboratory test systems provide the most practical means to
obtain data that can be used to predict the biodegradation and
fate of organic pollutants. More accurate predictions could
probably be made if fate tests could be conducted in the field for
each chemical under consideration. Such studies have often been
conducted for the application of pesticides to soil, but
constraints of time and expense do not permit such studies to be
carried out routinely in aquatic habitats. The second best
approach is to conduct a few field tests with selected chemicals
and to compare the results with those from laboratory tests to
assess the relevance and utility of the laboratory data. When the
strengths and weaknesses of the test systems are understood, the
degree of confidence with which laboratory data can be
extrapolated can be evaluated.
We have used several types of biodegradation test systems, including
shake flasks, eco-cores, and microcosms, to study the biodegradation of _p_-
nitrophenol (PNP) in the laboratory. In most instances, microbial communities
degraded nitrophenol after a lag period of several days. The length of the
lag period was variable, however; and in samples from estuarine or marine
sites and some freshwater sites, there was no biodegradation for weeks. The
inclusion of sediment also seemed to affect the biodegradation, but it was not
clear whether the effect made the results more or less realistic.
The purpose of this study was to compare biodegradation of j}_-nitrophenol
and concomitant responses of microbial communities in laboratory~test systems
with those in the field. We prepared laboratory test systems with samples
from a freshwater pond and then treated all of the laboratory systems and the
pond simultaneously with the test compound so that direct comparisons could be
made. Two questions were of primary importance: (1) Is the adaptation
process in the field similar to that in the laboratory? (2) Which type of
laboratory test system (what level of complexity) best predicts the field
results?
RESULTS FROM FIELD STUDY
In the pond, PNP disappeared slowly for the first 120-150 hours
(Figure 1), then much more rapidly. Because of fluctuations in the pond water
level due to rain and runoff, it was not clear from the shape of the curve
between 100 and 200 hours whether the biodegradation rate had increased or
whether the test compound was being washed out of the pond. Therefore, when
the PNP concentration fell below 50 ug/1 in the pond, an additional aliquot of
139
-------
PNP was added to bring the concentration to 134 yg/1. After 23 hours with no
measurable rain or runoff, the nitro compound was no longer detectable, which
Indicated that the biodegradation rate had increased dramatically.
The effects of prior exposure on the microbial community were determined
by measuring the biodegradation rate of PNP in the pond exposed to the nitro
compound and in the control which had no prior exposure. The test compound
disappeared from the exposed pond within one day (Figure 2). During the same
period, there was no detectable degradation in the control pond. The rate of
biodegradation in the first 24 hours was two orders of magnitude faster in the
exposed pond than in the control (-9.84 compared to 0.0998 ug/1). After an
acclimation period of 3-4 days, the test compound was also degraded in the
control pond (data not shown).
Bacteria able to mineralize PNP were enumerated in samples taken from the
pond during the initial exposure to determine whether changes that would
account for the Increased degradation rates could be detected in the microbial
community. There was an increase of three orders of magnitude in the number
of PNP degraders in the exposed pond between 100 and 240 hours (Figure 3),
which corresponded to the time period during which PNP degradation increased
dramatically. There was no commensurate increase in total heterotrophs, and
there was no proliferation of PNP degraders in the control pond.
RESULTS FROM LABORATORY TESTS
Results in laboratory test systems were similar to those in the pond.
For example, data from the small microcosms and from shake flasks with sediment
are shown in Figures 4 and 5, respectively. The shake flasks without sediment
(Figure 6) were the only exception to the close correlation between laboratory
and field data. The lag period extended for almost 500 hours, and the
subsequent degradation rate was slower than in the other systems.
Increases in biodegradation rates correlated well with increases in
populations of PNP degrading bacteria. We are currently investigating the
relationships among microbial growth, acclimation, and degradation of
xenobiotic compounds.
The results of this study show that there can be dramatic increases in
microbial degradation rates in the field after exposure of the resident
microbial community to the test compound. Such increases must be considered
in predicting the fate of pollutants that could serve as growth substrates for
bacteria. Our earlier work has shown that there can be wide variations in the
abilities of communities to adapt to degrading a test compound. We now have a
better understanding of how laboratory test systems can be used to assess the
biodegradation responses. For example, the role of the sediment in the
acclimation and the degradation is readily apparent. We are currently
conducting experiments to determine the factors that control such responses so
that they can be predicted.
140
-------
CONCLUSIONS
Several conclusions can be drawn regarding extrapolation of laboratory
data to predict results in the field:
1. Laboratory systems must include the components that affect the
biodegradation rates and the complexity necessary to simulate the
environment under consideration. In this case, inclusion of sediment
from the test site provided an acceptable level of agreement.
2. Measurement of biomass or any other surrogate for predicting
biodegradation probably will not provide predictions that are as
accurate as measuring the biodegradation rate in the environment.
3. If a simple biodegradation test system can be shown to predict the
biodegradation process in the environment, the most reasonable
approach would be to use the test system with an inoculum from the
site to predict biodegradation at that site. That is to say, if some
parameter must be measured to predict biodegradation rates at a
particular site, then measuring the biodegradation rate itself with a
simple laboratory system would be the most effective means currently
available.
LITERATURE CITED
1. Spain, J.C., P.H. Pritchard, and A.W. Bourquin. 1980. Effects of
adaptation on hiodegradation rates in sediment/water cores from
estuarine and freshwater environments. Appl. Environ. Microbiol.
40:726-734.
2. Spain, J.C., and P.A. Van Veld. 1983. Adaptation of natural
microbial communities to degradation of xenobiotic compounds:
effects of concentration, exposure time, inoculum, and
chemical structure. Appl. Environ. Microbiol. 45:428-435.
3- Spain, J.C., P.A. Van Veld, C.A. Monti, P.H. Pritchard, and
C.R. Cripe. 1984. Comparison of j>-nitrophenol degradation in
field and laboratory test systems. Appl. Environ. Microbiol.
(Accepted).
4. Van Veld, P.A., and J.C. Spain. 1983. Degradation of selected
xenobiotic compounds in three types of aquatic test systems.
Chemosphere. 12:1291-1305.
141
-------
350
250
OL
Z
Q.
150
50
•
50 100 150
HOURS
200
250
Figure 1. Concentration of PNP in the test pond, ihe test
pond received an additional treatment of PNP after
220 hours.
250
Q_
Q_
50
CONTROL
TREATED
20
60
HOURS
100
Figure 2. Concentration of PNP in treated and control ponds
The treated pond was exposed to PNP for 10 days;
the control pond received no pretreatment.
142
-------
I06
HETEROTROPHS
£L
TREATED
CONTROL
PNP DEGRADERS
48
144
HOURS
240
Figure 3.
Total heterotrophs and PNP degraders enumerated
by MPN technique in previously treated and control
ponds. Samples were diluted appropriately and
incubated in PNP nutrient media for 2 weeks.
200
o»
100
50
. A! .
100
200
300
HOURS
Figure 4. Degradation of PNP in small (3 1) microcosms.
The microcosms contained sediment and water
(removed prior to treatment from the pond).
The microcosms received an additional treatment
of PNP after 180 hours.
143
-------
200
en
I 100
0
100 200
HOURS
Figure 5. Degradation of PNP in flasks containing water
and 500 mg sediment/1. The sediment was removed
from the pond prior to treatment. The flasks
received an additional treatment of PNP after
170 hours.
200
- 150
CL
Z
CL
100
50
Figure 6.
200 400 600
HOURS
Degradation of PNP 1n flasks containing
water from the field site. The flasks received
an additional treatment of PNP after 600 hours,
144
-------
PANEL REPORT
The importance of field studies has become increasingly evident to
participants at this workshop. At this point, there is considerable
uncertainty among workshop participants regarding the validity for
extrapolating laboratory data to predict the biodegradation of a chemical in
the environment. Consequently, the validity of laboratory studies must be
tested in the field.
Fundamental to this process is the selection of laboratory test systems
that are similar to the various environmental compartments (soil, water,
groundwater, etc.). Each of these laboratory systems should incorporate the
important components of the compartment under consideration, such as a field
source of inoculum, the appropriate solid surfaces (e.g., sediment), and
environmental conditions. Once a laboratory system is developed, it will then
be important to verify the ability of the system to predict the biodegradation
of a chemical in the corresponding field site. The design of these
experiments should incorporate a statistical analysis, in order to minimize
the variability likely to be encountered in the environment. Should the
results of initial trials indicate that the laboratory system does not predict
the field observations, the system should be modified as appropriate.
Upon establishment of a reliable laboratory model, the system should be
validated by using several chemicals of different classes and enough field
sites to establish confidence in the system. Following validation, the
laboratory system can then be used as a tool to predict the biodegradation of
other chemicals at a given site. Furthermore, the laboratory system can also
be used with confidence to address a variety of additional questions, such as
the relationship of biodegradation rates among different sites and the effects
of a controlling environmental factor on the biodegradation rate.
145
-------
SUMMARY AND RESEARCH PLAN FOR BIODEGRADATION STUDIES
BASED ON OUTPUTS AND DISCUSSIONS
AT THE BIODEGRADATION KINETICS WORKSHOP
J. M. Suflita
P. H. Prichard
INTRODUCTION
The microbial ecology effort of the EPA has evolved to the point where it
is no longer sufficient for researchers to enumerate the numbers and types of
microorganisms in a particular environment. Past experience has taught us
that this approach provides little insight into the cycling of materials and
energy in the ecosystem to which the organisms belong. More importantly we
have become interested in and involved in the activity of microorganisms in
nature and wish to describe that activity in a quantitative manner.
Therefore, we feel that future research should describe kinetics of microbial
activity, predominant variables controlling the reaction rates, and potential
techniques to extrapolate laboratory biodegradation information to the field.
The greater degree of complexity and sophistication inherent in this
research endeavor presents numerous problems in approach and methodology. The
biodegradation kinetics work has pointed up a number of equivocal approaches
and interpretational difficulties in biodegradation studies. However, during
the course of this meeting, certain underlying generalizations were
established; these can serve as a basis for the formulating of future research
directives.
We have attempted to organize these directives into a specific research
plan. Clearly, the various avenues of kinetic research are not mutually
exclusive. Progress in one area will assist the development of another.
However, we feel strongly that certain research areas deserve an immediate
effort, whereas others will evolve in the future.
Kinetic models of biodegradation processes were generally considered by
the participants at the workshop as necessary elements of a research
program. Models are based on assumptions that can form the basis for specific
testing of hypothesis and related experimentation. They force us, for
pragmatic reasons, to describe the complexity of biodegradation with a minimum
number of measurable parameters. The parameters idealy represent an
integration of the most important biotic and abiotic factors affecting the
biodegradation of xenobiotic chemicals. Although the time and research needed
to search for and test those parameters will be extensive, we feel that the
potential outcome will justify the effort. Kinetic models will also provide
predictive capabilities that will be invaluable for our making regulatory
decisions. Such models will require us to examine the factors controlling
adaptation events in natural ecosystems, since these phenomena must be
included in our predictions. We believe that the testing of kinetic models
represents a major research directive.
Equally important are research endeavors focusing on the problem of
extrapolating laboratory biodegradation data to the field and extending
146
-------
biodegradation kinetic predictions from one ecosystem to another. The
development of a conceptual framework for extrapolation will be the impetus
and direction for gathering data from both field and laboratory experiments.
It will also force us to make decisions about how to determine efficiently the
spatial and temporal variability in biodegradation rates. In addition, it
will facilitate the regulatory decision-making process and base it on observed
variability.
These two areas, model testing and extrapolation, will therefore form the
basis for our research plan. Discussions and panel reports from the workshop
helped define the goals of this plan.
MODEL TESTING
KINETIC MODELS FOR BIODEGRADATION
Several types of kinetic models of varying degrees of complexity were
discussed at the workshop. However, two simple models, a first-order decay
model and the so-called "second-order approach" or Athens model, as we will
subsequently refer to it, seem to be the most reasonable to incorporate into a
research plan.
A first-order decay model contends that the rate of substrate (xenobiotic
chemical) disappearance over time in any environmental sample will depend on
the substrate concentration by a proportionality constant (k-^) if other
conditions in the test system (cell numbers, cell activity, nutrients,
competing substrates, etc.) remain the same over the course of an
experiment. The proportionality constant is a reflection of the actively
degradating microbial biomass in the sample being tested and the inherent rate
at which the enzymatic machinery in the cells will attack the xenobiotic
chemical in question. It obviously will vary, depending on when and where the
environmental sample is taken. The model's dependency on substrate
concentration will only be true when substrate concentrations are relatively
low.
The Athens model reasonably contends that the rate of biodegradation is
proportional to both substrate concentration and actively degrading biomass.
Typically, a first-order rate constant (kj) is normalized by a biomass
measurement to produce a proportionality constant (kp). In this case k-, is
actually a reflection of the Michaelis kinetic parameters Vmax and Km (R-, =
V /Km). The proportionality constant, k~, is specific for each xenobiotic
cnemical, and its value is assumed to be the same for all microbial
communities, regardless of their size, activity, or source (site-
independent). This suggests that the biochemical kinetic constants for
degrading any one xenobiotic chemical are essentially "universal."
The Athens model, however, suffers from several practical constraints and
inadequately tested assumptions. Most obvious in this respect is the term in
the expression which requires information on the actively degrading biomass.
Several biomass measurements have been used in attempts to test this model.
Measurements of CPU, AODC counts, LPS measurement, ATP, etc. have found
limited utility with relatively few substrate classes in only a few
habitats. To be sure, there is general disagreement on the "best" biomass
measurement currently in use.
147
-------
In addition, the derivation of the model assumes that substrate
concentration is much less than K^. Since many laboratory experiments use
starting substrate concentrations in the range of 20-200 yg/1, it is assumed
that Km must therefore be relatively high. To ensure that degradation rates
are measured in the first-order region, substrate concentrations should be at
least one-tenth the Kffl value. Practically, this means Km values should be in
the low yg/1 range. Thus, it is quite possible that a biodegradation rate
determination might be unknowingly performed in the zero-order region rather
than the first-order region. Before substrate concentrations can be assumed
to be well below K , an indication of the magnitude in K for different
xenobiotic chemicals must be obtained. The implicit assumption that K values
are constant relative to other parameters such as Vm;)V also needs to be
tested. max
ACTIVITY MEASUREMENTS
It is apparent in the development and use of biodegradation kinetic
models that (a) the activity of a microbial community (which includes
different bacterial populations), (b) the relativeconcentrations of the
active bacteria, and (c) the environmental factors controlling their activity
need to be assessed using a minimum number of parameter estimates. The more
encompassing and integrative the parameter estimates, the less are required.
We believe that EPA should support research to develop activity measurements
that are consistent with known biological theory, easily performed by a
variety of laboratories, and applicable to a wide variety of xenobiotic
chemicals and ecosystems. The initial use of total viable cell numbers in a
microbial community was, in essence, an attempt to provide an "activity"
measurement. It was assumed that all bacteria, or some constant proportion
thereof, were active in degrading a chemical and that they were all equally
active. Evidence presented at this workshop and from other sources indicates
that this "activity" measurement is apparently applicable to a limited number
of chemicals and ecosystems. This is due in large part to problems in the
original assumptions. Similarly, measurements of the numbers of specific
chemical degraders within a population are limited because of methodological
problems and the realization that the mere presence of a degrader does not
necessarily mean it is active. Thus, new approaches for relating the activity
of a microbial community to its degradative potential are required.
Two promising approaches are the measurement of the maximum velocity
(Vm,J of degradation and the first-order decay rates (k,).
ma x i
In the saturating or zero-order region of substrate decay, the
biodegradation rate (Vmax) is limited by the active catalyst concentration and
not by the substrate concentration. Therefore, the rate observed at
saturating substrate levels is a direct reflection of the actively degrading
biomass. This is exactly the quantity required in the Athens model (that is
k2 = k,/V )> but it uses a rate of biodegradation instead of cell number.
At first glance, it would appear that the., units for this proportionality
constant are unusual (i.e., liters.umoles or reciprocal of K units).
However, when this kinetic constant is properly employed, -dS/dt will have
recognizable units (i.e., iimoles.l~.h~ ).
The first-order region of substrate decay is also a reflection of
activity degrading biomass when the substrate is present at concentrations
148
-------
well below the K of the reaction. In this region, a constant fraction of
substrate undergoes conversion to product at any given moment, and the
absolute velocity decreases as the substrate is consumed. Thus, the rate for
a given reaction is substrate dependent.
It is important to note that first-order kinetics will often be displayed
if some other factor mitigates the consumption of a substrate by the requisite
microorganisms. For instance, this behavior will be observed if
biodegradation is limited by diffusion of the molecule to the site of
microbial activity. Such phase transfer resistance is common for the
consumption of gaseous substrates or possible for chemicals that tend to
partition strongly to solids.
Both approaches (V , k^ have minimal interpretation difficulties, and
each is a direct measure of activity. Secondly, they are not limited to a
particular type of activity (e.g., hydrolysis, hydroxylation, deamination,
etc.), and they are not restricted to a given chemical class or series of
structurally related compounds. Both approaches are also independent of the
nature of the catalytic unit. It matters little if the biotransformation is
actually catalyzed by an extracellular enzyme, a bacterium, a fungus, or an
anaerobic consortium of organisms. The approach can be used in a variety of
environments, including single phase systems (i.e., water alone) or multiphase
systems (i.e., sediment/water slurries). In addition, the activity
measurement will already reflect an integration of the environmental factors
influencing biodegradation. In general, V determinations do not require
the analytical sophistication associated witn the measurement of low substrate
concentrations; as a corollary then, the requirements for radiolabeled
material may largely be circumvented. Biodegradation rates obtained at
saturating concentrations of substrate (V a ) permit subsequent predictions of
rates at any concentration (assuming the Athens model is essentially
correct). Biodegradation rates obtained in the first-order decay region are
only predictive at lower substrate concentrations. Finally, the suggested
activity measurements can be made by all laboratories using relatively simple
incubation systems.
METHODOLOGY
Ideally we would like to see research designed to give all the salient
kinetic parameters (K[r), V , kp and kp) in a single experiment or set of
experiments. It is desirable to get the maximum amount of pertinent
information from a limited amount of experimentation, particularly when the
inoculum tends to be difficult to obtain. Based on the information presented
at this workshop, it would appear that there are two promising methods to
accomplish this goal: (a) progress curve experiments and (b) experiments
generating typical substrate concentration-velocity data pairs. Both
approaches have advantages and disadvantages, and neither is applicable to all
chemicals in all environments. The techniques should be chosen with special
regard to practical constraints such as the ease of obtaining the measurements
and of gaining the maximum amount of information with a minimum of inoculum,
time, effort, and cost.
1. Progress curves. Progress curves are experiments designed to follow
substrate depletion or product formation through the zero-, mixed-, and first-
order regions. By analyzing these experiments using either nonlinear
149
-------
parameter estimation techniques (as recommended by the statisticians) or
linearizing transformations of the data, the various Michaelis constants V
and K can be obtained, and thus k, and k? can also be calculated. The rear x
advantage to this technique over others js_ that the kinetic parameter
estimates can be derived from a single experiment. Progress curve experiments
also lend themselves to manipulating environmental factors (i.e.,
partitioning, temperature, etc.) which will help describe quantitatively the
influence of these factors on biodegradation processes.
A limitation of this technique may be the difficulty of obtaining
saturation concentrations for some substrates. However, this remains to be
seen. If the degradation of xenobiotic compounds is enzymatically catalyzed,
saturation kinetics should be observed.
Obviously, we realize that progress curve experiments are not designed as
simulations of complex environments. Rather, they are tools to help us
understand the intricacies of nature with an eye toward predicting
biodegradation kinetics.
Unlike other types of experiments designed to estimate biodegradation
rate parameters, the assumptions in this experimental approach are
straightforward. Simply stated, decreases in reaction velocity with time are
due only to decreasing catalyst saturation, and the biodegradative process is
unlinked to significant changes in catalyst concentration over the course of
the experiment.
The latter assumption presumes that the particular biodegradation process
is unrelated to significant microbial growth. Of course, this assumption will
not always be the case, and microorganisms will proliferate as a result of the
metabolism of a xenobiotic carbon and energy source. However, we feel that
the recommended approach and ensuing rate estimations are basically
conservative. There will be a certain amount of error associated with any
rate estimation, and from an environmental policy viewpoint, we feel that it
is better to err on the conservative or low side rather than to overestimate
the natural biodegradation rate. If microbial growth does occur as a result
of xenobiotic compound utilization, this ultimate expression of metabolism
will be reflected in an adaptation phenomenon (see below), and rate
estimations can be periodically updated.
2. Substrate concent rat i on -velpc i ty experiments. The same kinetic
parameters can also be obtained by performing typical sTfbstrate concent ration -
velocity experiments and analyzing the results by nonlinear regression or
linearizing techniques. These experiments involve determining the initial
rate of substrate disappearance at several different substrate
concentrations. Either labeled or unlabeled substrates can be used, depending
on analytical sensitivity required. Typically, such experiments assume that
the reaction proceeds to a negligible extent during the course of the assay
(i.e., less than 5% substrate depletion). Under some and perhaps most
experimental conditions, it may not be possible to restrict the reaction to
this extent. At best, these experiments usually employ 5 or 6 data points.
This limitation is not inherent in progress curve analysis, since the
Michael is-Menten expression is a differential velocity equation that is valid
over the entire concentration range and as many data points as desired can be
incorporated into the analysis.
150
-------
Several important factors may limit the widespread applicability of the
substrate concentration-velocity method. First, radiolabeled substrates are
often required to detect very low concentrations or small changes in substrate
concentrations against a large background. Considering the myriad of
chemicals and their potential availability and cost, it is unlikely that
techniques relying on radiolabeled materials will gain widespread acceptance.
Secondly, parameter estimations from substrate concentration-velocity
data pairs often rely on linearized plotting techniques. Good examples
include heterotrophic activity experiments, which usually use plots analogous
to the double reciprocal plot of Lineweaver-Burk. However, to obtain accurate
parameter estimates using this approach, some a priori knowledge is required,
since the chosen substrate concentration range is critical and must be in the
neighborhood of K . If the chosen substrate range is too high (relative to
Km), the observed velocity is insensitive to changes in substrate. Thus V
can be accurately estimated, but K will be inordinately low. If the chosen
range is much lower than Km, the curve will intercept the graph axis too
closely to the origin to allow accurate Km or Vmax estimates (Km and V will
be very large). The substrate increments chosen can also critically influence
parameter estimation. Other techniques of plotting substrate-velocity data
have similar ideal concentration ranges. As a corollary to the above
cautions, saturation is almost never checked in these experiments to verify
that one is truly in the appropriate concentration range. Nevertheless, if
appropriate concentration ranges are chosen, accurate parameter estimation can
be made using this technique.
Lastly, from a practical standpoint, many substrate concentration-
velocity experiments rely on CO,, formation and detection. The assumption is
that primary degradation is the rate limiting process and that subsequent
conversion to CO.? is relatively rapid. However, even a cursory review of the
literature reveals that COo may not represent the ultimate fate of the test
substrate. Primary degradation reactions may yield intermediates that can
bind, sorb, or are generally more persistent than, starting chemicals. Thus,
before experiments are designed, C02 formation must first be correlated with
substrate disappearance.
RESEARCH PLAN
MODEL VALIDATION
Research effort should continue to test further the usefulness of the
Athens model. We need to determine if it is desirable and feasible to obtain
a k^ proportionality constant for every xenobiotic chemical (or class of
chemical if the benchmark approach is utilized). Research to date has shown
that the use of total numbers of bacteria as the normalizer for different
degradative activities among microbial populations has been partially
successful. Some compounds that are apparently degraded by hydrolytic
mechanisms give consistent k2 values when the first-order rate is divided by
total cell numbers. However, further work along these lines is not
recommended for two reasons. First, there does not seem to be a consistency
trend indicating which chemicals are amendable to the total cell number
approach. For example, some chemicals that are hydrolytically attacked do not
give consistent ko values while others do. Secondly, since the activity of
microbial communities is unpredictably related to cell numbers, the conceptual
basis behind using total cell numbers is obscure.
151
-------
We believe, as we have indicated above, that a more direct measure of
microbial activity is needed to obtain and compare k^ values. To determine
which activity measure to use, we recommended that a decision be made relative
to the magnitude and variability in values of the Michaelis half saturation
constant, K , for a variety of xenobiotic chemicals. In other words,
assumptions ^kl>out Km used in the formulation of the Athens model must be
carefully and thoroughly tested. There is currently a paucity of kinetic
information upon which to make such decisions. We feel strongly that this is
where the research effort in EPA should be focused.
The conceptual framework for a research effort involving the Michaelis
half-saturation constant is shown in Figure 1. The Km values for xenobiotic
chemicals can be relatively high or low. Enzymes capable of metabolizing
pollutant chemicals that closely resemble native substrate(s) would be
expected to exhibit high affinities for the xenobiotic compound and thus low
K 's. Chemicals totally foreign to the environment would probably have enzyme
systems with relatively high K values. If K is consistently high for most
xenobiotic chemicals, then an adequate measure of microbial degradative
activity for any environmental sample would be the first-order decay constant
(ki). This would be an integrative measure representing the inherent rate at
which the catalytic unit(s) would attack the chemical, the number of catalytic
units involved, and the factors that control catalysis. Knowledge that
-------
nr
HIGH
v
RELATIVELY
CONSTANT
LOW
RELATIVELY
CONSTANT
4T—
YES
"T"
NO
MEASURE
k1 OR Vmax
(I^POSSlM)
YES
MEASURE
max
INTERMEDIATE CASE
(CAN USUALLY MAKE
PREDICTIONS BASED
ON KINETIC MODEL)
v
WORSE CASE
(NO GENERALIZING
KINETIC MODEL
APPLICABLE)
BEST CASE
(CAN USE KINETIC
MODELS TO
PREDICT RATES)
Figure 1. Flow diagragm for research strategy in testing kinetic models of
biodegradation.
153
-------
If Km values for xenobiotic chemicals are reasonably constant relative to
V values, then three hypotheses follow. First, the average affinity of the
catalytic units for the substrate is the same or similar from one environment
to another. Similarity may only be apparent after certain categorizations,
i.e., oligotrophic versus eutrophic water bodies, polluted versus unpolluted
areas, naturally occurring versus exotic chemicals, etc. It may be that the
biochemical mechanisms associated with substrate transport into the cells and
initial catabolic attack are quite similar. Biodegradation processes are
therefore controlled more by environmental factors and substrate chemistry
rather than by biochemical factors, i.e., a large array of metabolic machinery
has apparently not evolved to transform xenobiotic chemicals. Second,
assuming biodegradation can occur, the rate of transformation will be
controlled largely by the number of actively degrading microorganisms or
catalytic units. This activity will encompass and reflect the effect of the
immediate environment. This activity, as we have suggested above, can be
equated to the V value. Knowledge about the relative constancy in Km would
be a major step in determining if the complexity of biodegradation rates in
natural ecosystems can be generalized. If K 's are also relatively constant
for a general chemical class, then perhaps all chemicals in a given class can
be expected to have invariant K 's in widely different environments. Third,
since ki = V /K , a relatively constant K will mean that k, varies as does
V . Knowledge of either kj or Vmax can therefore be used in biodegradation
rate predictions.
Information on K will facilitate derivation of k,, values (k]/V ) and
the testing of the Athens model. Experiments to determine k~ neecT to be
conducted with a variety of chemicals and environmental samples from diverse
areas. If k- can be shown to be the same or similar for chemicals by
normalizing differences in degradation rates with V , then V would appear
to be an excellent parameter for assessing actively degrading biomass. Thus,
V or k, could be used as the basic parameter estimate for determining
cWeria for extrapolation.
ESTABLISHING KINETIC PARAMETERS
To be sure, there is a limited amount of information on the basic kinetic
parameters of Km, V , k^, and k£ governing pollutant biodegradation. Our
current thoughts on biodegradation kinetics are an integration and a direct
outgrowth of the information presented at the workshop. We feel that in order
to test models for biodegradation kinetics, efforts using judiciously selected
model substrates in a number of habitats should be expanded toward
establishing the parameters mentioned above.
The criteria for selection of the model substrates include: (a) their
potential for analysis, (b) their high water solubility, (c) their lack of
significant sorption, (d) the availability of radiolabeled compounds, (e) the
amount of background information on their biodegradation, (f) their ability to
be metabolized aerobically and anaerobically, (g) representativeness of
different chemical classes, and (h) their potential for exhibiting a
relatively wide range of biodegradation rates. A potential list of chemical
types that could be tested is presented in the workshop report. In addition,
we suggest that the initial efforts start with three substrates:
4-nitrobenzoate, 4-cresol, and 2,4-dichlorophenoxyacetic acid. These
compounds should be tested by establishing their kinetic parameters (Km, Vmax»
154
-------
ki, and k~) in habitats ranging from oligotrophic environments to eutrophic
ones. For example, biodegradation in a polluted portion of an estuary may be
compared with a pristine area. Another example would include freshwater lakes
of differing (but well defined) trophic status. These habitats as well as
others (sewage, sludge, streams, rivers, open ocean, groundwater, etc.) should
be used to help determine if predictive kinetic models of biodegradation are
truly feasible.
The techniques used in this effort should include the shake flask, since
it is this method which is the most likely to be used on a widespread basis.
The design and analysis of these experiments will be those described earlier
in this document. Degradation studies with suspended sediments should also be
included.
ENVIRONMENTAL FACTORS AFFECTING BIOOEGRADATION
As pointed out by the workshop panel report, there are many factors that
can serve to accelerate or retard the biodegradation of xenobiotic
susbstances. Considering the multiplicity of these factors, one questions
whether or not it is expeditious for EPA to support research to quantitate the
influence of these factors.
Our approach to assessing metabolic activity by measuring laboratory
biodegradation rates seems to be a way of accounting for these environmental
factors without actually knowing what they are. For instance, such results
will tell us if a biodegradation rate is relatively slow or fast. Even though
such experiments will not allow us to conclude why this is so. We feel that
by measuring rates we can successfully integrate the many environmental
factors influencing biodegradation and can limit the number of false positive
indications of biodegradation.
It is fully recognized that it is desirable to learn as much as possible
about the factors controlling biodegradation rates in the environment.
However, we believe that pragmatically a detailed knowledge of these factors
is not necessary for rate predictions. As more and more factors become
elucidated, quantitative relationships describing their influence on
biodegradation will be established. The basic biodegradation kinetic model
will thus grow in degree of sophistication to accommodate these
relationships. However, for each new factor described, field information on
when and where that factor will have its effect must be obtained to make more
sophisticated predictions. The subsequent monitoring effort will increase
exponentially with each new factor or concern. Our philosophy suggests that
research efforts must initially be oriented in the direction of finding
parameter estimates that integrate the effect of these environmental factors
into a small number of measurements. We feel that either V or k,
measurements lend themselves to this goal and therefore should be strongly
emphasized. If this concept does not eventually prove useful, we will perhaps
have to investigate more intensively the factors that control biodegradation
rates in order to find a better quantitative basis for environmental
predictions of biodegradation rates.
Work should also continue on the influence of environmental factors,
because the resulting information will eventually help extrapolation research
(see next section). It may also provide the basis for experiments that
155
-------
attempt to manipulate the environment to improve biodegradation capacity or
rates for the purposes of eliminating certain in-place environmental
pollutants. Dr. Raymond's paper describing the use of nutrient additions to
accelerate hydrocarbon degradation in ground water is an excellent example.
ADAPTATION
Numerous published investigations and extensive discussions at this
workshop have indicated that many xenobiotic chemicals do not degrade
immediately when added to an environmental sample. A lag or adaptation period
of varying lengths is observed before a chemical is rapidly degraded. The
extent of the adaptation period directly affects the exposure of the biota to
the pollutant. The incidence of adaptation and the mechanism(s) controlling
adaptation should therefore be as extensively studied as biodegradation
rates. Therefore, lag phases must be considered in possible regulatory
decisions.
At this point, the length of time before the onset of rapid degradation
is impossible to predict since the phenomena is poorly understood. Therefore,
it is difficult to model this portion of a decay curve. Probable reasons for
adaptation include the requirement for bacterial growth before significant
parent chemical is transformed, specific substrate concentrations before the
requisite enzymes are inducted, the need to deplete competing substrates
before the compound of interest degrades, or the need to exchange the
essential genetic material. Other mechanisms of adaptation can certainly be
postulated.
At present, a generalizing descriptive model for adaptation has not been
proposed. This is largely because of an absence of information on the
mechanisms controlling adaptation. However, the adaptation event itself can
be readily observed experimentally. Adaptation, or the length of time before
accelerated biodegradation occurs, is an integrating parameter which, like the
V value, does not require a knowledge of the exact adaptive mechanisms or
the factors controlling it. We therefore believe that research efforts should
investigate the length of the lag or adaptation period. Three major
approaches should be considered. First, the effect of substrate
concentrations on the length of the lag period should be cleary defined for a
variety of chemical types and environmental samples. Arguments that
adaptation will not occur at environmentally realistic
concentrations (ug/1) must be experimentally substantiated. Second, the
relationship of adaptation to the growth of organisms capable of degrading the
substrate must be investigated. This effort is needed to establish whether
lag periods are simply a function of the initial number of degraders in the
environmental sample and their eventual growth rate in that sample. This
assumes that if the degraders are present they will commence growth
immediately following exposure to the substrate. Examples where degraders are
present but never grow are assumed for now to be infrequent and may
consequently form the basis for research in the future. If growth can be
quantitatively linked to the lag period prior to adaptation, the development
of descriptive models becomes feasible.
Third, the natural variability in adaptation time should be examined to
determine if regulatory decisions can possibly be based on certain
consistencies in the adaptation response. For example, will the adaptation
156
-------
time for a specific xenobiotic chemical always be the same if environmental
samples are always taken from the same general area or does the adaptation
time always vary widely in time and space? If consistencies can be observed,
it will be important subsequently to relate them to adaptation mechanisms,
chemical structure trophic conditions, topographic and hydrogeologic features,
and previous exposure to pollutants.
EXTRAPOLATION
Extrapolation of laboratory data to the field is a simple concept that is
poorly defined. Extrapolation means different things to different
investigators, and a clear plan that outlines the research questions and
suggests practical, experimental approaches does not exist. In an effort to
correct this situation, we propose a research strategy that would involve:
(a) establishing the "environmental relevance" of laboratory-derived data, and
(b) categorizing the variation in biodegradation rates in natural environments
in such a manner to permit predictions of exposure concentrations.
ENVIRONMENTAL RELEVANCE OF LABORATORY DATA
It is frequently argued that xenobiotic chemical biodegradation rates
obtained from simple laboratory tests (e.g., river die-away, shake flasks,
etc.) are environmentally insignificant since, by design, these tests replace
the complexity associated with natural environments with expediency,
standardization, and reproducibility. Laboratory tests are important for
regulation and chemical registration, but it is not clear how accurately this
information reflects what actually happens in the field. To be sure, industry
cannot examine every chemical in a laboratory system that simulates real world
conditions, but information can be supplied that will delineate the
qualitative and quantitative degree to which a simple laboratory method
represents field events. Simple awareness of the deficiencies of a laboratory
test may be very useful from a regulatory or extrapolation viewpoint.
Modifications of existing tests or the addition of supplemental tests will
likely be the most productive outcome of this type of relevancy study. Such
modifications will greatly improve our confidence in making fate predictions
from laboratory information.
We propose that the environmental relevance of laboratory test data can
be most efficiently determined using microcosm studies and field studies.
Based on recent information, microcosms appear to provide good simulations of
the field. They can be dosed indiscriminately, and they are isolated from
uncontrollable weather conditions. They certainly provide a means of studying
biodegradation processes under complex conditions. Microcosm studies
therefore, in our opinion, are the methods of choice for such relevancy
studies. Whenever possible, the fate of a xenobiotic chemical should also be
examined under field conditions. Only experience in both types of studies
will determine our confidence in and limits to the the use of microcosms.
A research plan involving microcosms and field studies would consist of
using biodegradation information from simple laboratory tests to see if it can
adequately explain the biotic fate of a chemical in more complex ecosystems
such as microcosms or field studies. If this hypothesis proves correct, then
we believe that the environmental relevance of the laboratory test data will
have been established.
157
-------
If it cannot, then manipulation of the complex system or the isolation of
component parts thereof may indicate how to modify appropriately the simple
laboratory test so that it will accurately describe natural events. In some
cases, complexity may control biodegradation rates to such an extent as to
prohibit reduction to simpler terms. If this is the case, a routine and
standardizable complex laboratory test system must eventually be developed and
tested.
The use of laboratory data to explain events in complex systems is
complicated by the fact that biodegradation will not be the only fate process
affecting a chemical. Dilution, sorption to sediments, diffusion,
bioturbation, bioaccumulation, photolysis, hydrolysis, volatility, etc., may
also influence the fate of the chemical. Recent investigations indicate that
the use of appropriate controls, physical and chemical tracers, unique
analytical techniques, and simple mathematical models will permit a
determination of the biotic fate of a chemical in a complex system and will
allow a comparison with information generated in simple laboratory tests. We
recommend that these research efforts be continued and expanded. Test
chemicals have to be selected carefully, and ancillary experiments will have
to be carried out for additional information. Microcosms that simulate
specific portions of ecosystems need to be developed and tested. Appropriate
field sites will have to be located. Dosing, analytical, and data analysis
procedures need to be developed.
We feel strongly that microcosms and field studies have relatively little
predictive value. They are generally too small to be truly representative of
any large ecosystems. They are instead excellent validation tools. Kinetic
models for biodegradation, sites of active biodegradation, and factors that
control biodegradation rates can be tested in these systems and their
environmental significance determined.
VARIATIONS IN BIODEGRADATION RATES
Estimating biodegradation rates using laboratory studies requires, as we
have proposed above, the determination of V value (assuming a need for a
site-independent kinetic constant k2) or tne first-order rate constant.
Environmental samples are usually collected from a field site for this purpose
and then used in a laboratory biodegradation test system. The resulting
estimation of biodegradation, k-,, or the activity of the microbial community
(V ) is representative of that site only. Its representativeness of the
rest of the ecosystem is unknown. The values are likely to vary widely,
depending on when and where the sample is taken. Little appreciation for this
natural variation currently exists. Is it so extensive as to force us to
analyze biodegradation rates over very small spans of space and time? Perhaps
biodegradation rates or the activities of microbial communities responsible
for the biodegradation are not as widely varying as we suspect. Or perhaps
variability can be categorized relative to certain general charactristics of
the ecosystem. Will eutrophication tend to decrease or increase
variability? Will different degrees of pollution or prior exposure to
xenobiotic chemicals affect variability, or will the absolute rates just
increase or decrease? Thus, we believe that an extensive research effort
should be centered on measuring the variability in biodegradation rates. This
appears to be the only way that we will be able to truly extrapolate
laboratory data to the field, to extrapolate biodegradation information from
158
-------
one environmental site to another, and to extrapolate biodegradation
information about one chemical to another in the same chemical class.
We suggest that the temporal and spatial variability of k1 or V in one
typical ecosystem (estuary, river, lake, etc.) be established. Statistical
designs, which are readily available and adequately tested, will have to be
employed for the most efficient estimation of this variability. In that
ecosystem, the largest area that will consistently give biodegradation rates
within a pre-specified range must eventually be determined. In an estuary,
for example, the area considered might include the water (at an average depth)
and sediment for the main embayment. Determination of the range will be based
on the actual observed variability, on the toxicological properties of the
chemical, and on certain ecosystem characteristics. The objective will be to
determine the frequency and size of "hot spots" or "low spots," if any. The
range in biodegradation rates tolerated will essentially control the
significance of hot spots. For example, one portion of the main embayment in
an estuary may consistently give degradation rates of V values that fall
outside the observed variability in the rest of the embayment. This hot spot
may not, however, be significant if one is willing to accept the additional
variability. Acceptance or nonacceptance will again depend on the observed
data, the chemical being tested, and certain regulatory considerations.
Efforts must then continue to identify hot spots in other corners of the
ecosystems and during other seasons of the year. The maximal use of
statistical design will be crucial for optimized determination of variability.
With an estimate of the variability for the biodegradation of one
chemical in one ecosystem at hand, cluster analysis of the data should be
performed to determine the minimum number of measurements necessary to detect
similar variability in any new ecosystem. If the number is too large then the
variability in degradation rates in aquatic environments is too extreme for us
to make generalizations over large areas within a ecosystem. Degradation
rates become very site specific, and extrapolation of laboratory data to the
field is unfeasible. However, if the number of required measurements is
reasonable, then examination of the temporal and spatial variability in other
ecosystems should be carried out.
We propose that two or three geographically different field sites be
chosen for each major ecosystem type (e.g., rivers, estuaries, streams, lakes,
etc.) to measure the variability in k^, Vmax> or Km« For example, two or
three different lakes could be selected in different geographic areas to
represent this ecosystem type. The selection would be based, in part, on
accessibility for experimental purposes and, in part, on environmental
characteristics of the field site. We believe that initially field sites
representing extremes (i.e., polluted versus pristine, eutrophic versus
oligotrophic) should be examined for their variability. If variability is
"similar" for each of these extremes, then it can be argued that site-to-site
extrapolation is feasible within certain limits of error. A chemical, for
example, may show an average half-life of X days plus or minus some error term
for one particular site of bounded environment. If variability can be assumed
to be the same from one site to another, the average half life for a chemical
determined at a new site can be assigned the same variability as the original
site.
159
-------
The "large number of chemicals examined by OTS precludes an assessment of
variability in biodegradation rates for each new chemical. Some form of
structure activity relationship is obviously necessary. We propose that the
biodegradation rates of structural analogues within a given chemical class can
be ranked from the fastest to the slowest. The consistency of the ranking(s)
should then be examined using samples taken from several field sites at
different times. If one were to examine the chemical class of chlorinated
biphenyls, for example, would monochlorinated biphenyls always degrade faster
than dichlorinated biphenyls, regardless of the source of the inoculum used in
the biodegradation experiments? Would 4,4-dichlorobiphenyl always degrade
faster than 3,4-dichlorbiphenyl, etc.? A consistent ranking would prove
invaluable. It would mean that the variability in biodegradation rates in
time and space for one representative of a chemical class (i.e., a "benchmark"
chemical) could be linked to other chemicals of that class through the ranking
analysis. Consistent rankings may only be possible if different subsets are
established. Ranking using samples from polluted areas may be different from
those using samples from nonpolluted areas. Extensive eutrophication may also
result in different rankings. Likewise, rankings using samples from river
water should not necessarily be expected to agreee with those using estuarine
samples.
We suggest that the major chemical classes of concern to OTS and EPA be
identified and a representative member of each class be selected. The partial
list of chemicals discussed at this workshop is an excellent start in this
direction. Work can then begin to establish consistencies in the rankings of
their biodegradation rates with other chemicals within those classes.
160
-------
Appendix: Proposed Benchmark Chemicals for Biodegradation Research
The following chemicals were recommended by an ad hoc committee who met
briefly on the last day of the workshop:
1. pentachlorphenol
2. p-nitrobenzoate
3. 2-mercaptobenzothizole (MBT)
4. pthalate esters
5. phenylcarbamate or thiocarbonate
6. decalin or phenanthrene
7. phenoxybenzoate
8. ABS
9. carboxyazobenzene (or similar)
10. polyethylene glycol
11. Dalapon
12. tertiary or quadamine (EDTA or NT)
13. cyclohexane carboxylic ac.
14. substituted pyridine
15. parachlorobiphenyl
The criteria for choosing a benchmark chemical were:
1. Range of physical properties (e.g., anionic, cationic, hydphobicity)
2. Range of chemical structure (e.g., heterocyclic - aromatic and
aliphatic branched, polymeric, straight chain)
3. First-order fate process -- biodegradation *
4. Variety of initial reaction mechanisms
5. Safety in handling
6. Availability and cost
7. Environmental relevance
The purpose of using benchmark chemicals was to compare laboratory test
systems and to test the integrity of those systems.
161
------- |