EPA/600/R-05/064
June 2005
Microbial Source Tracking Guide Document
National Risk Management Research Laboratory
Office of Research and Development
U. S. Environmental Protection Agency
Cincinnati, OH 45268
-------
Notice
The intent of this guide is to provide the reader with insight into various tools and approaches used to track
sources of fecal contamination impacting water quality in streams, rivers, lakes, and marine beaches.
Descriptions of research and several case studies gathered through workshops, literature searches, and phone
interviews are also provided. An effort was made to showcase programs, activities, and analyses that
incorporated diverse Microbial Source Tracking (MST) approaches and tools. EPA does not support or
condone any of the uses of the MST data presented here; nor does it endorse any of the organizations
discussed in the case studies. An extensive interpretive review of the scientific literature is included for those
interested in learning more about the field.
This document does not impose legally binding requirements on states, authorized tribes, or the regulated
community and does not substitute for Clean Water Act (CWA) or Safe Drinking Water Act (SDWA)
requirements, EPA's regulations, or the obligations imposed by consent decrees or enforcement orders. EPA,
through its Office of Research and Development, participated in the research described here, and this
published guide has been subjected to the Agency's review and has been approved for publication as an EPA
document.
-------
Foreword
The U. S. Environmental Protection Agency (EPA) is charged by Congress with protecting the Nation's land,
air, and water resources. Under a mandate of national environmental laws, the Agency strives to formulate
and implement actions leading to a compatible balance between human activities and the ability of natural
systems to support and nurture life. To meet this mandate, EPA's research program is providing data and
technical support for solving environmental problems today and building a science knowledge base
necessary to manage our ecological resources wisely, undrestand how pollutants affect our health, and
prevent or reduce environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for investigation of
technological and management approaches for preventing and reducing risks from pollution that threaten
human health and the environment. The focus of the laboratory's research program is on methods and their
cost-effectiveness for prevention and control of pollution to air, land, water and subsurface resources;
protection of water quality in public water systems; remediation of contaminated sites, sediments and ground
water; prevention and control of indoor air pollution; and restoration of ecosystems. NRMRL collaborates
with both the public and private sector partners to foster technologies that reduce the cost of compliance
and to anticipate emerging problems. NRMRL's research provides solutions to environmental problems by:
developing and promoting technologies that protect and improve the environment; advancing scientific and
engineering information to support regulatory and policy decisions; and providing the technical support and
information transfer to ensure implementation of environmental regulations and strategies at national, state,
and community levels.
This publication has been produced as part of the Laboratory's strategic long-term research plan. It is
published and made available by EPA's Office of Research and Development to assist the user community
and to link researchers with their clients.
Sally Gutierrez,
Director
National Risk Management Research Laboratory
-------
Acknowledgements
This document was prepared by the U.S. Environmental Protection Agency's Office of Research and Develop-
ment in cooperation with the following contributing authors. Thanks to Thomas Edge, Environment Canada; John
Griffith, Southern California Coastal Waters Research Program; Joel Hansel, USEPA Region 4; Valerie J. (Jody)
Harwood, University of South Florida; Michael Jenkins, U.S. Department of Agriculture; Alice Layton, University
of Tennessee; Marirosa Molina, USEPA Ecosystems Research Division; Cindy Nakatsu, Purdue University; Robin
Oshiro, USEPA Office of Water; Michael Sadowsky, University of Minnesota; Jorge Sanlo Domingo, USEPA Of-
fice of Research and Development; Orin Shanks, USEPA Office of Research and Development; Gerald Stelma,
USEPA Office of Research and Development; Jill Stewart, National Oceanic and Atmospheric Administration;
Donald Stoeckel, U.S. Geological Survey; Bruce Wiggins, James Madison University; and Jayson Wilbur, Worces-
ter Polytechnic Institute.
Working meeting participants included: Steve Harmon, USEPA Office of Research and Development; Bonita
Johnson, USEPA Region 4; Stephanie Harris, USEPA Region 10; Samuel Myoda, Delaware Department of Natu-
ral Resources and Environmental Control; Donald Reasoner, USEPA Office of Research and Development; Jutta
Schnieder, Virginia Department of Environmental Quality; and Mano Sivagenesan, USEPA Office of Research and
Development.
Other who contributed: Charles Hagedorn, Virginia Polytechnic Institute and State University and Peter Hartel,
University of Georgia.
Other people consulted: Sally Gutierrez, USEPA Office of Research and Development and Ronald Landy, USEPA
Region 3.
The Lead Editor was Jorge Santo Domingo, USEPA Office of Research and Development. The Technical Editor
was Jean Dye, USEPA Office of Research and Development.
The lead editor gratefully acknowledges the help of Sue Schock (USEPA) for setting up the contract for the work-
ing meetings, Science Applications International Corporation (SAIC) for organizing the meetings, and Jill Neal for
helping formatting the manuscript and for administrative support. We also want to recognize the leadership of Ron-
ald Landy (USEPA Region 3), Bobbey Smith (USEPA Region 9), Donald Reasoner (USEPA-ORD/NRMRL), and
Sally Gutierrez (USEPA-ORD/NRMRL), who in many ways were responsible for several of the USEPA-ORD and
regional offices sponsored activities associated with this document. The working meetings were supported by funds
from the Water Supply and Water Resources Division and by Regional Applied Research Effort (RARE) Program
of the Office Science Policy.
The cover illustration and graphic support was provided by Teresa Ruby (CSC Inc.). Thanks to Steve Wilson for
excellent desktop publishing support. We also recognize the help provided by present and former members of the
Microbial Contaminant Control Branch (Catherine Kelty, Randy Revetta, Jingrand Lu, Donald Reasoner, Margaret
Williams, and Joyce Simpson) and the participants of the Microbial Source Tracking workshop held in August 2003
in Cincinnati, OH. Special thanks to the reviewers of this document: Sheridan Haack (USGS), Bane Schill (USGS),
Donald Stoeckel (USGS), Mike Jenkins (USDA), Brian Robinson (NOAA), Laura Webster (NOAA), Jan Gooch
(NOAA), Sally Gutierrez (EPA), Tom Edge (Environment Canada), James Goodrich (EPA), Gene Rice (EPA), Rich
Haugland (EPA), and Robin Oshiro (EPA).
-------
Table of Contents
Foreword iii
Acknowledgements iv
List of Figures vii
List of Tables viii
Executive Summary 1
Chapter 1. Introduction to Fecal Source Identification 3
Chapter 2. Decision Criteria 7
2.1 Introduction 7
2.2 Choice of method 7
2.2.1 Step 1: Is the problem adequately defined? 7
2.2.2 Step 2: Has an adequate sanitary survey been conducted? 9
2.2.3 Step 3: How many sources were identified in the sanitary survey? 9
2.2.4 Step 4: Is the watershed/study area of manageable size? 10
2.2.5 Step 5: What is the desired level of discrimination? 10
2.3 Explanation of Resolution/Outcome Endpoints 10
Chapter 3. Microbial Source Tracking Approaches 15
3.1 Introduction 15
3.2 Cultivation versus cultivation-independent microbial targets 15
3.3 Cultivation-dependent/library-dependent methods 17
3.3.1 Phenotypic methods 18
3.3.2 Genotypic methods 20
3.4 Cultivation-dependent/library-independent methods 27
3.5 Cultivation-independent/library-independent methods 29
3.6 Identification and quantification of specific bacteria 32
Chapter 4. Data Collection and Analysis in Library-dependent Approaches 41
4.1 Introduction 41
4.2 Data collection 41
4.3 Numerical representation of isolate profiles 42
4.4 Library construction and validation 44
4.5 Measuring spatial and temporal variability 45
4.6 Techniques for classification and discriminant analysis 50
4.7 Practical issues and Chapter summary 52
-------
Table of Contents (Continued)
Chapters. Methods Performance 55
5.1 Introduction 55
5.2 Universal Quality Measures 55
5.3 Method-Specific Performance Criteria 57
5.3.1 Library-Dependent Methods 57
5.3.2 Library-Independent Methods 59
5.4 Conclusions 61
Chapter 6. Assumptions and Limitations of MST Methods 63
6.1 Introduction 63
6.2 Host specificity of specific strain/pattern/marker (SPM) 63
6.3 Widespread distribution of SI and SPM in host populations 65
6.4 Stability of the signal 66
6.5 Transferable methodology 67
6.6 Temporal stability within the host 67
6.7 Geographic stability 68
6.8 Representative sampling 69
6.9 Persistence of SPMs in environmental waters 70
6.10 Persistence of SPM in primary vs. secondary habitats 74
6.11 Relevance of SI to regulatory tools 74
6.12 Relevance of SI to human health 75
6.13 Summary 75
Chapter 7. Applications of MST Approaches 77
Case 1. St. Andrews Park (Georgia) 78
Case 2. Tampa Bay (Florida) 80
Case 3. Vermillion River (Minnesota) 83
Case 4. Anacostia River (Maryland/District of Columbia) 85
Case 5. Accotink Creek, Blacks Run, and Christians Creek (Virginia) 88
Case 6. Avalon Bay (California) 92
Case 7. Holmans Creek (Virginia) 94
Case 8. Homosassa Springs (Florida) 96
Literature Cited 98
Glossary and Important Terms 120
Glossary of Acronyms 123
-------
List of figures
Figure 2.1 General criteria used to decide which method or approach should be applied in a
particular MST study 8
Figure 2.2 Criteria used to discriminate between human and all other sources of fecal
contamination 11
Figure 2.3 Criteria used to discriminate between different host groups 12
Figure 2.4 Criteria used to discriminate between different individual farms of the same
type of host group 13
Figure 3.1 Schematic representation of currently available MST methods 16
Figure 4.1 Curve representation of two genotypic fingerprints. Data is represented in terms
of band intensity (gray scale; Y axis) and migration distance (X-axis) 43
Figure 4.2 Multidimensional scaling map of the ten USEPA regional offices constructed
from the table of geographic distances in Table 4.1 46
Figure 4.3 PCA and MDS plots of a small library of isolates where different colors indicate
different source categories. Each color represents a bacterial strain isolated from
a different source 49
Figure 4.4 Relationships between similarity measures for binary profiles 52
Figure 5.1 Library-dependent specificity calculation for human detection 56
Figure 5.2 Sensitivity or RCC calculation for human detection in a 100 ml control sample 59
Figure 5.3 Measuring limit of detection for human host-specific PCR assay 60
Figure 5.4 Potential internal control for an environmental sample that will be analyzed
with a PCR based method 61
Figure 6.1 Illustration of some strains/patterns/markers (SPMs) currently utilized in MST methods 65
-------
List of tables
Table 3.1 Summary of logistics of methods tested for MST 37
Table 3.2 Comparison of advantages and disadvantages of source tracking methods 38
Table 4.1 Geographic distances between each pair of USEPA regional offices 46
Table 4.2 Common similarity measures for binary profiles 48
Table 4.3 Techniques for identifying patterns of spatio-temporal variability in the isolate profiles 49
Table 4.4 Summary of common classification rules 52
Table 4.5 Comparison of software packages commonly used for analyses 53
Table 5.1 Summary of Quality Measure Controls 57
Table 6.1 Characteristics of an ideal source identifier (SI) and those of a useful source identifier 64
-------
Executive Summary
Approximately 13% of surface waters in the United States do not meet designated use criteria as determined
by high densities of fecal indicator bacteria. Although some of the contamination is attributed to point
sources such as confined animal feeding operation (CAFO) and wastewater treatment plant effluents, non-
point sources are believed to contribute substantially to water pollution. Microbial source tracking (MST)
methods have recently been used to help identify nonpoint sources responsible for the fecal pollution of wa-
ter systems. Moreover, MST tools are now being applied in the development of Total Maximum Daily Loads
(TMDL) as part of Clean Water Act requirements and in the evaluation of the effectiveness of best manage-
ment practices. It is evident that MST is transitioning from the realm of research to that of application.
This is not a regulatory document; rather, this document was designed to be used as a reference guide by
those considering MST tools for water quality evaluations and TMDL-related activities. This document is
also relevant to people interested in source water protection where accurate identification of fecal pollution is
required to implement reliable management practices. However, in a broader sense, water quality managers
addressing public health issues, beach/shellfish closures, microbial risk management, and ecosystem restora-
tion should also benefit from the extensive materials contained in this document. Since some of the tools
discussed are used in other areas of microbial water quality, environmental scientists and engineers in general
would benefit from several of the Chapters of this document.
The guide is divided into seven Chapters. None of the Chapters is intended to stand alone; thus, the reader
is encouraged to consult as many Chapters as possible to put into context the comments and suggestions
made in various sections. A brief introduction to MST and the goals of this guide document are provided in
Chapter 1. Many of the criteria used to decide which method to use in source identification are discussed in
Chapter 2. Details relevant to each of the most current approaches used in MST and ways data are collected
and analyzed are explained in the next two Chapters. Performance standards for MST studies are discussed
in Chapter 5, followed by a critical evaluation of the general assumptions behind and limitations to applica-
tion inherent in the various approaches. Examples of MST application are presented in Chapter 7.
Most MST studies have relied on matching "fingerprints" from bacterial strains isolated from a water system
to those isolated from various hosts (e.g., humans, cows, pigs, raccoons, deer, geese, chickens, etc.) or known
environmental sources (e.g., municipal wastewater). In essence, fingerprints are based on phenotypic traits
(e.g., antibiotic resistance analysis) or genotypic profiles (e.g., rep-PCR, ribotyping) of individual micro-
bial strains. Typically, hundreds of fingerprints of pure culture bacteria isolated from different sources (or
known-source library) are generated in MST studies (Chapter 3). Although results from several studies sup-
port the use of the library-dependent approaches for MST, accuracy of these approaches in field application
has been questioned because of various problems associated with the target organisms. Some of these prob-
lems relate to the level of complexity introduced by spatial and temporal vectors, the stability of the markers
used, and issues of sampling design (Chapter 6). More recently, library-independent approaches have been
proposed based on the amplification of host-specific markers. Reports are beginning to surface summarizing
results of studies that evaluated library-independent methods against real-world samples (Chapter 7). Much
-------
less is known about the library-independent approaches than the library-dependent approaches; therefore, it
is not possible at this time to recommend one approach over the other except in specific circumstances (out-
lined in Chapter 2). It should be noted that in some cases, more than one approach could be utilized for the
purpose of identifying fecal pollution sources. Furthermore, in some circumstances it might be necessary to
use more than one approach to validate preliminary results obtained with a particular approach.
The complexity of environmental samples and the different variables affecting microbial survival and host
specificity have an indirect impact of the efficacy of all MST tools. Moreover, selection of MST tools and
approaches are dependent in large part on the goals of an individual effort. In all cases, accomplishing proj-
ect goals will be impacted by the availability of technical and financial support. As a consequence, various
MST methods might be deemed appropriate at sites with similar characteristics (Chapter 2). Regardless of
project-specific criteria, the ultimate MST goal can generally be summarized as identification of the major
sources of fecal contamination impacting the water system in question. In some cases, this goal has been
achieved, while in others, the lack of strong experimental design and poor understanding of the limitations
of MST have resulted in insufficient data analysis and poor decision making. Hence, environmental manag-
ers must consult the scientific literature and, whenever possible, consult experienced practitioners prior to
embarking on source identification studies.
This document builds upon a history of cooperative work among federal, state, and academic partners. To aid
our understanding of the reliability that can be expected from various MST tools and approaches, the EPA
Office of Research and Development organized several multiagency meetings with the purpose of receiv-
ing input from scientists in specialized areas such as population genetics, population biology, host-microbe
interactions, microbial physiology, and microbial ecology. MST researchers from academia, USEPA regions,
states, federal government (USDA, USGS, NOAA, and USEPA), and Environment Canada participated in
these meetings. Most of contributing authors of this guide document participated in the aforementioned
meetings and in similar meetings in the U. S., Canada, and Europe; in many ways, this document captures
the most relevant elements and suggestions that were discussed in prior meetings.
This document should be cited as follows:
USEPA (U.S. Environmental Protection Agency). 2005. Microbial Source Tracking Guide Document. Office
of Research and Development, Washington, DC EPA-600/R-05/064. 131 pp.
-------
Chapter 1. Introduction to Fecal Source Identification
The Clean Water Act establishes that the states must adopt water quality standards that are compatible with
pollution control programs to reduce pollutant discharges into waterways. In many cases the standards have
been met by the significant reduction of loads from point sources under the National Pollutant Discharge
Elimination System (NPDES). Point sources are defined as "any discernable, confined and discrete convey-
ance, including but not limited to any pipe, ditch or concentrated animal feeding operation from which pol-
lutants are or may be discharged". However, more than 30 years after the Clean Water Act was implemented,
a significant fraction of the U.S. rivers, lakes, and estuaries continue to be classified as failing to meet their
designated uses due to the high levels of fecal bacteria (USEPA, 2000b). As a consequence, protection from
fecal microbial contamination is one of the most important and difficult challenges facing environmental sci-
entists trying to safeguard waters used for recreation (primary and secondary contact), public water supplies,
and propagation offish and shellfish.
Microbiological impairment of water is assessed by monitoring concentrations of fecal-indicator bacteria
such as fecal coliforms and enterococci (USEPA, 2000a). These microorganisms are associated with fe-
cal material from humans and other warm-blooded animals and their presence in water is used to indicate
potential presence of enteric pathogens that could cause illness in exposed persons (Dufour, 1984). Fecally
contaminated waters not only harbor pathogens and pose potential high risks to human health, but they also
result in significant economic loss due to closure of shellfish harvesting areas and recreational beaches (Rabi-
novici et al., 2004). For effective management of fecal contamination to water systems, the sources must be
identified prior to implementing remediation practices. Millions of dollars are spent each year on monitoring
fecal-indicator bacteria in water and attempting to develop reliable methods for fecal source tracking. Reli-
able and accurate fecal source identification methods are imperative for developing best management prac-
tices (BMPs) to control fecal contamination from relevant animal sources, to protect recreational-water users
from water-borne pathogens, and to preserve the integrity of drinking source water supplies.
The immediate demand for methods in MST has been stimulated by the current total maximum daily load
(TMDL) requirements that states, territories, and tribes must comply with in the next five to ten years. A
TMDL specifies the theoretical amount of a pollutant that a waterbody can receive and still meet water qual-
ity standards. Strict waste load allocations from point sources like sewage treatment plants or industrial dis-
charge pipe have already been established with the purpose of meeting regulatory standards. For this reason
it is believed that nonpoint-pollutant sources are mostly responsible for many water system impairments,
especially after storm events. Most nonpoint sources are associated with agricultural operations, although
urban associated pollution is also an important contributor due to the increase in residential, commercial, and
industrial development, use of manure as fertilizers, persistence of combined sewer overflows, and mal-
functioning septic systems. Wildlife is often assumed to be a relevant source of pollution in cases where no
obvious contribution could be assigned to human activity and livestock farming. Due to the variety of poten-
tial fecal sources impacting watersheds, fecal source identification is a challenging task that often requires
multidisciplinary teams to effectively implement.
-------
Various approaches have been used to identify fecal sources in water samples (Sinton, 1996; Jagals, 1995;
Simpson et al., 2002). For example, chemical analyses have been used to detect human-associated markers
like caffeine, fragrances, and detergents (Glassmeyer et al., 2005). Fecal constituents (e.g., fecal sterols, fecal
stanols, and secretory immunoglobulins) have also been considered as source identifiers, since different con-
geners are preferentially present in different animal species. Some of the chemical-based approaches for fecal
source identification are gaining acceptance within the environmental community; however, issues that relate
to specificity, sensitivity, microbial biodegradation, and adsorption must be further investigated in order to
validate their use as reliable source identification tools. While this document will focus on source tracking
tools and approaches that use microorganisms as the source identifiers, it should be noted that chemical ap-
proaches could also be used in fecal source identification studies.
Early attempts to classify fecal sources based on microbial source identifiers focused on discriminating
contamination sources in a broad fashion (i.e., human vs. nonhuman categories) based on the fecal coliforms
to fecal streptococci (FC-FS) ratio. It is now widely accepted that this approach cannot accurately differenti-
ate between human and animal sources because differences in die-off rates between fecal coliforms to fecal
streptococci could affect FC-FS ratios in aged the ratios used to classify the sources are not consistently valid
for different animals. Variability in bacterial survival rates between fecal coliforms and fecal streptococci
affects this ratio, particularly when a temporal component is added to the equation. While the FC-FS ratio is
seldom used in contemporary source identification studies, it should be recognized that the work of Geldreich
and colleagues (Geldreich and Kenner, 1969; Geldreich et al., 1968; Geldreich and Clarke, 1966; Geldreich
et al., 1964) is in large part responsible for encouraging other scientists to develop and evaluate new tools to
discriminate between the different sources of fecal pollution.
More recently, a number of microbial source tracking (MST) approaches have been developed to associ-
ate various animals with fecal pollution of natural waters. MST is based on the assumption that, given the
appropriate method and source identifier, the source of pollution can be detected. In general terms, MST
methods could be grouped as library dependent methods (LDMs) and library independent methods (LIMs).
LDMs require the development databases of genotypic or phenotypic fingerprints for bacterial strains iso-
lated from suspected fecal sources. Fingerprints of isolates from contaminated water are compared with
these libraries for classification. Bacterial indicators of fecal contamination (e.g., E. coli and enterococci) are
commonly used for LDM development. LIMs do not depend on the isolation of targeted source identifier
as detection is performed via the amplification of a genetic marker by a Polymerase Chain Reaction (PCR)
step, although some methods often require a pre-enrichment to increase the sensitivity of the approach. Some
LIMs target the!6S rDNA (which is vital for protein synthesis and therefore present in all bacteria), while
others target function-specific genes (which are present in a particular bacterial group) for PCR primer devel-
opment. The advantages, limitations, and applications of a majority of these methods will be discussed in the
following Chapters.
Several MST tools are now being applied in the development of TMDL plans and in the evaluation of best
management practices. However, due the relatively recent development of MST, most environmental man-
agers and scientists have little training and experience in the application of MST methods to TMDL plans. To
date there is no single method that could be applied to all types of fecally contaminated water systems. This
is due, in part, to the fact that several factors can control the level of complexity of a particular water system,
which has a direct impact in choosing the best method for the identification of primary sources of pollution.
Moreover, there is a lack of consistency among the various laboratories performing some of the MST tech-
niques that keeps them from sharing data. The Office of Research and Development recognizes the impor-
tance of effective pollution management measures and the need to develop, evaluate, validate, and standard-
ize methods that could help stakeholders address current fecal pollution issues.
-------
The purpose of this guide is to provide scientists, engineers, and environmental managers with a comprehen-
sive, interpretive analysis of the current and relevant information (based on both lab and field data) related to
MST. Descriptions of the various MST approaches, data collection tools, data analysis procedures, method
application, performance standards, and assumptions and limitations associated with the field of MST will
be provided in different chapters. The chapters were written by a diverse group of professionals from aca-
demia and government agencies. Regional and state environmental professionals were also consulted during
different stages of this particular effort. Many of the contributing authors are recognized leaders in MST and
applied environmental microbiology. While the information herein presented is contemporary, it should be
noted that MST is a very intense and dynamic field, and therefore, the reader is encouraged to consult the
scientific literature frequently.
-------
Chapter 2. Decision Criteria
2.1 Introduction
A number of methods, both genomic and phenotypic, have been developed for use in microbial source track-
ing (MST). Some of these methods are library-dependent (i.e., rely on fingerprint databases of culturing
microorganisms) and some are library-independent (i.e., normally performed by nucleic acid amplification
techniques that do not require cultivation of microorganisms). Comparison studies have shown that no single
method is clearly superior to the others (Griffith et al., 2003; Stewart et al., 2003; Stoeckel et al., 2004).
Therefore, no single method has emerged as the method of choice for determining sources of fecal contami-
nation in all fecally impaired wate bodies. The decision on which method to use depends on the unique set
of circumstances associated with the specific study area in question, the results of sanitary surveys, as well
as budgetary and time constraints. In some situations, a rather coarse method will suffice, particularly if it is
only necessary to distinguish between human and animal fecal sources or between domestic animal and wild-
life sources. In other situations, it may be necessary to identify the species of domesticated animal or even
the specific herd or flock that is the major contributor of fecal pollution, both of which require more precise
methods.
2.2 Choice of method
The general MST decision tree that appears in this Chapter (Figure 2.1) was created to assist state and local
authorities in deciding whether or not MST methods are necessary to determine the sources of fecal pollution
in their particular watershed or bathing beach and, if so, which group of methods might be most appropriate
for their needs. Identification of an appropriate group of MST methods is the outcome of a series of decision
points. A menu of methods appears at each decision point, allowing the potential users to make informed
decisions. The reader is directed to Chapter 3 and 6 to evaluate the advantages and disadvantages of specific
MST methods.
The following steps outline the process as shown in the decision tree. These steps are meant to serve as a
guide to the reader as to the decision points in the tree.
2.2.1 Step 1: Is the problem adequately defined?
MST can be used in a number of circumstances. First, the problem to be addressed must be adequately de-
fined and the desired outcomes considered. For example, if the problem is bacterial exceedences that result
in beach advisories/closures, there are many variables to be determined. These include: the conditions under
which exceedences are likely to occur, the bacterial indicator species of concern, and the desired outcome
(removal of future advisories, determination if human pollution is a source, etc.)
Problem definition can vary for the same situation. In the case of TMDLs, the problem and desired outcome
may initially be defined to determine if human feces are contributing to the exceedences so that a prioritiza-
tion scheme can be fulfilled, i.e., if human feces are present, the area becomes a high-priority target for man-
-------
Has a fecal pollution
problem been
sufficiently defined?
Define
" problem
Has a sanitary
survey been
conducted9
Conduct a sanitary
survey of the stud\ area
How main
possible fecal
sources were
determined1'
Library
independent
method
confirms?
Is the study area
of manageable
sue?
Dissect study area to
smaller size
\
What le\ el of
discrimination
' i
Human vs
All Others
(-1)
r
Species
(Cattle vs.
Horses vs.
Human etc)
(P2)
ecal source
is desired'1
l
1
Groups
(Human vs
Wildlife vs.
Livestock;
(*3)
Individuals
(Specific
Cows, etc.)
(«4)
Figure 2.1 General criteria used to decide which method or approach should be applied in a particular MST study.
-------
agement action due to the known risks associated with this type of contamination. Once a TMDL is sched-
uled, the definition may change to a desire to know every source that may be contributing and, if possible, to
what degree.
Failure to adequately define the problem and desired outcomes prior to initiating the decision tree make it
unlikely that MST will serve a useful function and achieve results that can be acted upon. Given this, readers
are strongly cautioned about proceeding without this information.
2.2.2 Step 2: Has an adequate sanitary survey been conducted?
A sanitary survey can be used to evaluate and document sources of contaminants that might adversely affect
public health. Although sanitary surveys are frequently associated with drinking water supply systems, they
can be used to identify sources of pollution and to provide information on source controls and identification,
persistent problems such as exceedance of water quality standards, magnitude of pollution from sources,
and management actions and links to controls. A Registered Sanitarian or professional with experience in
these areas should perform the survey. A sanitary survey can be an effective tool for protecting human health
and can provide information that helps in designing monitoring programs and selecting sampling locations,
times, and frequencies.
In this instance, the sanitary survey should be of sufficient rigor to identify all of the potential sources within
the study area, as well as the conditions under which unacceptable contamination should occur. The spatial
and temporal extent of the contamination is typically based on local conditions including tidal cycle, near-
shore currents, dam releases, and rainfall. Lack of a sufficient survey will hinder the overall approach to
identifying the source of pollution in the study area.
For information on how to conduct sanitary surveys, the reader is referred to EPA's National Beach Guidance
and Required Performance Criteria for Grants -June 2002 (EPA 823-B-02-004), Appendix G. This dcument
is publically available at the following electronic site; http://www.epa.gov/waterscience/beaches/grants/guid-
ance/factsheet.pdf
2.2.3 Step 3: How many sources were identified in the sanitary survey?
Single source: It is quite possible that the sanitary survey will identify a single, dominant source of con-
tamination within the watershed. In this case, MST is likely unnecessary and remediation of the source is
warranted. However, some resource managers may desire a confirmatory test to back up the result of the
sanitary survey. In this case, one option would be to use a library independent method, assuming there is
an available technique that targets the source identified by the sanitary survey. Use of a library independent
method in this scenario is advantageous because these methods can confirm the findings of the sanitary sur-
vey without investing the time and money necessary to build a library. However, it may also be cost effec-
tive to employ a library dependent method if an appropriate local database already exists. If MST results
confirm the findings of the survey, then remediation is again warranted. If the confirmatory test fails to sub-
stantiate the findings of the survey or remediation fails to fix the problem, this would indicate a failure in the
sanitary survey. In such a case the sanitary survey should be repeated. In some cases a new survey strategy
should be considered.
Multiple sources: Proceed to the next step.
-------
2.2.4 Step 4: Is the watershed/study area of manageable size?
This is a rather subjective step, but experience in the field has shown that the smaller the watershed/study
area under examination, the greater the chance of success in determining the cause of the exceedence and the
likelihood of success at correcting the problem. In general terms, watersheds or study areas with drainage
areas greater than 14 digit USGS hydrologic unit code in size are not amenable to using MST. An excep-
tion to this general statement is that non-library based methods may prove useful in larger area evaluations if
the desired outcome is to know whether human fecal contamination is present. If previous steps have been
performed on areas greater than the 14-digit zone, it is strongly recommended that the size of the affected
watershed or drainage area be whittled down by use of extensive targeted sampling as previously document-
ed by Kuntz et al. (2003). In addition, a new sanitary survey may be necessary as the original one applies
specifically to the larger area.
2.2.5 Step 5: What is the desired level of discrimination?
As previously noted in Step 1, positive identification of a human source may be sufficient for some purposes.
However, more detailed information about all fecal sources may be necessary to address a different set of
objectives. Step 5 is meant to lead the reader to the set of methods which will provide the level of resolution
necessary to fulfill the objectives of the study. Possible discriminations are: 1) humans vs. all other sources,
2) species specific results (humans vs. cows vs. horses vs. deer etc.), 3) group comparisons (humans vs.
livestock vs. wildlife), and 4) specific individual hosts (cows from a certain farm vs. other farms vs. other
livestock on farms vs. human etc).
2.3 Explanation of Resolution/Outcome Endpoints
The following section will describe the different options available to discriminate between the different case
scenarios related to source tracking.
#1 Humans vs. All Other Sources and #2 Species Specific Results (Figure 2.2)
Both library independent and library dependent methods are amenable to the resolution of single species.
Library independent methods may be appropriate if techniques have been developed that target the desired
species. For example, methods that have been proposed to identify human fecal contamination include PCR
for host-specific Bacteroides species, E. coli toxin genes, or human-associated viruses (see Chapter 3). Other
species can likewise be targeted, although a limited number of methods currently exist for all species that
may be desired (Dick et al., 2005a; Dick et al., 2005b). As a result of sequencing efforts of fecal bacteria
as well as fecal microbial communities (Xu et al., 2003; Backhead et al., 2005; Eckburg et al., 2005), the
number of host specific assays is likely to increase significantly in the near future. If there is not a library-in-
dependent method available to target the desired species, a new method may need to be developed or library-
dependent methods should be considered. Likewise, if presence/absence results will not suffice to meet the
study objectives, and the available library independent methods are not capable of providing quantitative
results, then library-dependent methods should again be considered.
An adequate library must be available or developed in order to effectively utilize library dependent methods.
At present, it is not possible to provide generic guidance for what would constitute an adequate library for
any MST study. Readers should examine Chapter 3 to determine the requirements for library based methods.
Assuming that a library is available or developed, the level of discrimination should be determined to lead to
the appropriate suite of methods.
-------
As a caution, the use of 'weighed estimates' or 'quantitation' in these flow charts does not imply that an ex-
act, quantitative assessment is provided by these methods. With changing conditions in a watershed, robust-
ness of the base library, and other methodological considerations, the best that current technology can do is
to give a general idea as to the level of contribution from sources at the time the assessment is done. Results
from these types of analyses should be regarded as an estimate of contribution, rather than a well-defined
Is the level ol
quantitation of the
a\ailable librai\
independent method
amenable to \our
obiectives'1
Has a library
independent
method targeting
desired species
been developed'.'
Do you have a library or
\vill a libran of
sufficient si/e be
created9 (See Methods
Performance for details)
Reassess
desired
resolution or
quantitation
requirements
Level of
discrimination'
Libran based
Phenotypic
Methods
Libran based
Genotypic
Methods
Figure 2.2 Criteria used to discriminate between human and all other sources of fecal contamination.
-------
fraction associated with each source. With continuing evolution of the technology and methods for source
tracking, it is possible that precise quantitative results will be possible in the future.
#3 Host Group Comparison (Humans vs. Livestock vs. Wildlife) (Figure 2.3)
This track is very similar to #1 and #2; however, non-library based methods are not considered here because
the resolution of these methods is insufficient to discriminate to the level required. Library-based methods
only are applicable and come with the same caveat concerning a sufficient library as expressed in #1 and #2.
While there are non-library based methods that show promise for presence/absence analyses (e.g., the meth-
od that employs ruminant specific primers for Bacteroides), they do not currently offer the resolution neces-
sary to make a group comparison. In the case of the ruminant primers, the method will detect cows equally
as well as deer. More sensitive non-library-based methods continue to be developed and may become an
option for this type of group comparison in the future.
Do \ ou ha\ c a libran or
\\ill a libran,' of
sufficient si/e be
created'? (See Methods
Performance for details)
Reassess
desired
resolution
Level of
discrimination0
Libran- based
Phenotypic
Methods
Libran based
Genotypic
Methods
Figure 2.3 Criteria used to discriminate between different host groups.
-------
#4 Individual Hosts (Figure 2.4)
The only methods now available that produce this type of result are library based genotypic methods. Again,
these methods come with the caveat that a sufficient library must be available in order to get substantive re-
sults. Ideally, the library should be developed at the time of the study to counteract temporal variations that
have been observed in genomic libraries (Jenkins et al., 2003).
Do you ha\e a library or
will a library of
sufficient si/_c be
created'.' (See Methods
Performance for details)
Reassess
desired
resolution
Library based
Genotypic
Methods
Figure 2.4 Criteria used to discriminate between different individual farms of the same type of host group.
-------
Chapter 3. Microbial Source Tracking Approaches
3.1 Introduction
Numerous approaches have been used to determine potential sources of fecal contamination in the envi-
ronment. These methods are at various stages of development and validation. Accordingly, this Chapter
serves only as a resource for users to make an informed decision on the approach that best suits their needs
and financial resources. Currently one method cannot answer all questions and it is likely that this will not
change in the near future. This Chapter focuses on methods based on phenotypic and genotypic analysis of
microorganisms that have been used for source tracking. A number of the methods described in this Chapter
can be, or have already been, adapted for different target organisms. Chapter 6 reviews the target organisms
and factors that must be considered when appropriate methods are being chosen. The Chapter on case stud-
ies provides more detail on the successful application of some methods. Methods for MST are dynamic with
a number of new approaches being developed, such as gene chips with toxin genes and/or fecal indicator
sequences, and biosensors for the detection of target organisms.
Methods currently used for microbial source tracking fall into a few broad categories, genotypic versus phe-
notypic analysis of either cultivated target organisms, or cultivation-independent approaches by direct analy-
sis of samples from the environment (Figure 3.1). Genotypic analyses are based on some aspect of an organ-
ism DNA sequence, whereas phenotypic analysis measures a trait that is expressed. Genotypic methods differ
by targeting specific genes or by measuring genetic polymorphism (differences) in the genome. Genotypic
methods that have been used for microbial source tracking are: strain specific PCR (e.g., 16S rRNA gene,
host-specific toxin genes, or phage specific sequences), ribotyping, whole genome restriction fragment length
polymorphism (RFLP) analysis using pulse field gel electrophoresis (PFGE), repetitive element sequence
PCR (rep-PCR) fingerprint profiles, random amplification of polymorphic DNA (RAPD), and amplified frag-
ment length polymorphism (AFLP). Most of these methods require selective cultivation of indicator bacteria
from water samples as well as from fecal sources that are used to construct a host reference library, with the
exception of methods that detect bacterial host-specific genes (e.g., Bacteroides sp. 16S rDNA sequences)
using PCR. The two most often used phenotypic methods for MST, antibiotic resistance and carbon source
utilization, also require cultivation of the indicator bacteria. Each of these methods will be described in
detail in the following sections.
3.2 Cultivation versus cultivation-independent microbial targets
Many of the methods first tested for microbial source tracking used a cultivation approach for E. coli, entero-
cocci, and coliphage, as these organisms are used as indicator of fecal pollution in waters. Standard methods
for the cultivation of E. coli and enterococci and coliphage have been previously described (USEPA, 2000;
USEPA, 200la, b). Although EPA has standard cultivation methods, caution must be taken when comparing
studies in the literature. Often different methods have been used to cultivate and confirm the target organ-
ism (Harwood et al., 2003; Myoda et al., 2003). For example, E. coli may be isolated on mTEC, MI, mFc, a
combination of mENDO and NA-MUG, or by using commercial systems such as Colilert™ and Colitag™.
Some E. coli confirmatory tests used either singly or in combination are: IMViC - indole production, methyl
-------
Cultivation - Independent
Library-Independent
I
Sample
Concentration for processing
(can be stored at -20"C)
Direct sample anah sis
(possible but not common)
I
'f
Extract nucleic
acids
I
PCR
Cultivation
- Dependent
Library - Independent
Library'
- Dependent
Isolate or Enrich Target Organism(s)
I
I
Target Verfi cation
Specific Bacteria
(c g .HtHlcnnih's
Hlfldohaclt'riiini
SlIVpHXfXClH
Rhodocnct.it1>}
Viruses
(c g . Emero-. Adeno-.
\mises. or coliphagesi
Confirmation
using host
infection
xtract nuclici
1
Toxin
Gene PCR
I
Phenotypic
Analyses
i
Antibiotic
Resistance
4
Carbo
Utilizat
Genotypic
Analyses
n
on
i '
r
Blot and
hybridize \vith
Direct cell PCR
I 4
Extract nucleic
acids
gene specific I Rep-PCR | j RADP
Pick plaques probe
do RNase test
I
Serotype Genotype
Blot and hybridize
with rRNA gene
probe
Figure 3.1 Schematic representation of currently available MST methods.
red reaction, Voges-Proskauer test, failure to grow on citrate-minimal media, MUG (4-methylumbelliferyl-B-
D-glucuronide) hydrolysis (test for B-glucuronidase), indole production, gas formation on lactose, failure to
express urease, failure to express oxidase, and Analytical Profile Index (API) biotyping system. There is still
a need for researchers to standardize detection and confirmation methods for all indicators to ensure the same
organism is isolated and study results are comparable. The discriminatory power of each method may vary
when different target organisms are used and therefore each target organism must be tested independently to
assess the value of a method.
An alternative approach for studying microbial ecology has been prompted by research that estimated that
only a small fraction (0.1 to 10%) of bacterial species have been cultivated from most environments (Ran-
jard et al., 2000; Staley and Konopka, 1985; Torsvik et al., 2002). Most relevant to MST is the analysis of
gastrointestinal microbes, which indicates that some 400 different species of bacteria may be found in ani-
mal intestines and populations are in the order of 10" g"1 of contents (reviewed by Zoetendal, 2004). In-
testinal microflora have been well characterized in a number of animal hosts including humans (Suau et al.,
1999), swine (Leser et al., 2002; Pryde et al., 1999) and cattle (Ramsak et al., 2000). Collectively, and when
compared to cultivation-dependent methods, cultivation-independent methods suggest that the numerically
dominant bacteria in animal colons are anaerobic and belong to the low G+C Gram-positive and Cytophaga-
Flavobacter-Bacteroides bacterial phyla. Common genera in animal intestines are Bacteroides, Eubacte-
rium, Clostridium, Ruminococcus, Peptococcus, Peptostreptococcus, Bifidobacterium and Fusobacterium
(Matsuki et al., 2002). However, these bacteria are not readily cultivated in the laboratory, which has limited
-------
their use as fecal indicators in the past. In contrast, the more easily cultivated fecal indicator bacteria, E. coli
and Enterococcus sp., are present in lower concentrations. A number of molecular genetic methods and kits
have been developed to isolate nucleic acids from organisms or environmental samples without need for
cultivation, making it possible to use alternative targets. After extraction, a number of methods can be used
to examine DNA directly or indirectly after amplification by the polymerase chain reaction (PCR). The PCR
technique is an extremely useful, sensitive and rapid method that can be applied to both laboratory-cultivated
organisms and nucleic acids directly obtained from environmental samples. Nucleic acid replication via
PCR is automated in the laboratory resulting in an approximately 106-fold amplification of a target nucleotide
sequence. This approach provides a means to examine targets that are not readily cultivated and may not be
in high numbers in the environment, but nevertheless serve as better indicators of fecal sources.
Wantmore details on PCR?
sfpteto Reaction fPCB) is a method in which a targetpHAse^uejce is ]
i-farget sequences. Mmethbds
quirements: (i) target primer(s); (ii) each of the four nucleotides (adenine, cytosine,;
(iii) ttienttal totepnt DNA polymerase s, PCR tubes, etdl)-ajfe'i
3.3 Cultivation-dependent/library-dependent methods
Many methods that rely on the cultivation and isolation of the target microorganisms also require the creation
of a reference library. Reference libraries are built using isolates taken from known hosts or environmental
sources. In most cases isolates are taken from fecal samples, if possible, collected directly from the animal
or directly after excretion to ensure there is limited contamination from other sources. However, some in-
vestigators believe sewage lagoons and animal waste holding ponds provide isolates more representative of
survivors that would most likely be found in the environment. Most libraries have been built using isolates
taken from potential sources in the region being studied. Currently, there are conflicting opinions on the
geographic and temporal stability of source libraries, likely arising from a number of factors including: dif-
ferences in library sizes, sampling method, and data analysis method. Isolates with identical patterns from
the same fecal sample, are presumed to be clones and should be discarded from the library otherwise an inac-
curate statistical bias will occur. A number of approaches that have been used to determine the accuracy of
-------
libraries are discussed in the method performance Chapter. However, the most important unsolved factor is
the size of library necessary to successfully identify host sources.
3.3.1 Phenotypic methods
Antibiotic resistance
Antibiotic resistance was developed as a method for source tracking based on the demonstrated phenomenon
that bacteria from hosts exposed to antibiotics will develop resistance to those antibiotics, and on the hypoth-
esis that this selective pressure would be a mechanism for discriminating among fecal bacteria from various
hosts. Antibiotics are used to prevent and treat infections in humans and domestic animals and to increase
growth rates in animal production. Bacteria resistant to antibiotics used in animal feed (Bryan et al., 2004)
have been found in poultry litter (Kelley et al., 1998), cattle feces (Dargatz et al., 2003), and in swine ma-
nure (Smalla et al., 2000). Throughout the literature, different permutations of antibiotics and concentrations
(range in ug/ml) have been used for antibiotic resistance tests including: amoxicillin (4-128), ampicillin (10),
bacitracin (10-100), cephalothin (sodium salt) (10-50), chloramphenicol hydrochloride (4), chlortetracy-
cline hydrochloride (20-80), chlortetracycline (20-80), doxycycline hydrochloride (4), erythromycin (5-50),
gentamicin (1-20), kanamycin monosulfate (3-50), monensin, (5-250), moxalactam-sodium salt (0.2-1),
nalidixic acid-sodium salt (3-25), neomycin sulfate (3-50), norfloxacin (0.1), oxytetracycline hydrochloride
(20-100), penicillin G-potassium salt (20-200), polymixin B (1-10), rifampicin (2-16), streptomycin sulfate
(20-800), sulfathiazole (500), tetracycline hydrochloride (4-64), trimethoprinrsulfamethoxazole (1:19 ratio)
(0.2- 5), and vancomycin (2.5-30). There is currently no standard suite of antibiotics and concentrations used
for antibiotic resistance testing. Antibiotics are best chosen after determining potential animal fecal sources
and antibiotics used in their treatment. Furthermore, the antibiotics chosen must be appropriate to the source
identifier utilized, i.e., E. coli and other fecal coliforms are intrinsically resistant to vancomycin; therefore, its
use with this class of source identifier is not informative.
This method has been used extensively because it is rapid, relatively simple, and relatively inexpensive.
Furthermore, it requires less technical expertise than molecular methods and no specialized equipment. There
are three approaches that have been used in MST studies antibiotic resistance analysis (ARA), multiple an-
tibiotic resistance (MAR) and Kirby-Bauer antibiotic susceptibility. In MAR studies, bacteria are tested for
resistance to different antibiotics (Parveen et al., 1997). ARA differs slightly by including different concen-
trations of each antibiotic being tested (Wiggins, 1996; Wiggins et al., 1999). The Kirby-Bauer antibiotic
susceptibility test has been a standard method for use in clinical studies and uses small filter disks that have
been impregnated with antibiotics. The zone of growth inhibition around the disks is used to quantify resis-
tance. Some MST researchers believe that ARA provides the most information of the three antibiotic-based
approaches. A potential problem when using antibiotic resistance as a phenotypic source tracking method
is the transfer of resistance genes between bacteria. Genes conferring antibiotic resistance have been found
on a variety of mobile genetic elements including plasmids, transposons, and conjugative transposons that
provide a means for lateral transfer of the genes (Bass et al., 1999; Kruse et al., 1994; Ohlsen et al., 2003;
Salyers et al., 1995; Smalla et al., 2000). Although indigenous bacteria have the potential to transfer antibiot-
ic resistance genes to fecal bacteria after bacteria from fecal sources enter the environment, this would have
to occur at very high frequency to affect the overall proportion of resistant cells in the fecal host population.
Even if gene transfer frequencies were as high as 1%, which is much higher than has been reported (Smalla
et al., 2000), their detection will be unlikely with current antibiotic resistance protocols unless there is exten-
sive regrowth of the recipients in the environment.
-------
Application of antibiotic resistance to MST
Among the different antibiotic resistance approaches available, ARA is the most common method in MST
studies (Booth et al., 2003; Choi et al., 2003; Graves et al., 2002; Hagedorn et al., 1999; Harwood et al.,
2000; Harwood et al., 2003; Whitlock et al., 2002; Wiggins, 1996; Wiggins et al., 1999, Wiggins et al.,
2003), and has been utilized in many TMDL studies. Regardless of the specific method, they all first require
cultivation of the target organism, most times E. coll or enterococci (Harwood et al., 2003; Parveen et al.,
1997; Wiggins, 1996).
Basic antibiotic resistance methodology
For ARA and MAR antibiotic resistance analysis is carried out by first developing a database of antibiotic
resistance patterns (ARPs) of indicator bacteria isolated from the feces or sewage of known animal sources.
Colonies are isolated by membrane filtration or by streaking onto the appropriate selective-differential media.
These isolates are transferred to a 96-well microplate filled with growth medium, incubated, and then rep-
lica-plated on a battery of antibiotic-containing media. Multiple concentrations of each antibiotic are used
for ARA, while a single antibiotic concentration is used for MAR. The isolates are then scored positive or
negative for growth on each plate. Plates with no antibiotic addition are used as positive controls. Typically,
the ARP of each isolate consists of approximately 30 data points. The procedure for determining the ARPs of
isolates requires four to five days.
Bacterial ARPs from known sources are then analyzed using discriminant analysis, which is a form of multi-
ple analysis of variance. Discriminant analysis uses the ARPs from known sources to generate the predictive
equations (the "classification rule") that will be used to classify unknown isolates by source. The accuracy
of the database is assessed by using ARPs of the isolates from known sources as test data. This procedure
generates a source-by-source matrix that provides the rate of correct classification for each source. Overall
performance is measured by averaging the rates of correct classification (ARCC) for each source. Fecal bac-
teria isolated from polluted water are then processed in the same manner as the known isolates, and identi-
fied using discriminant analysis. More rigorous tests can be utilized to validate the predictive accuracy of the
database, i.e., ARPs of isolates from samples that are not included in the library can be used to challenge the
database's predictive capability.
The Kirby-Bauer disk diffusion antibiotic susceptibility test is performed following the NCCLS protocol
(National Committee for Clinical Laboratory Standards 1999). In this method filter paper disks with known
concentrations of antibiotics are placed into a Petri plate that has been heavily inoculated with the bacterium
of interest. The antibiotics diffuse from the disks into the agar making a gradient of antibiotic concentra-
tions. The plate is incubated usually for about 24 h, then the zone of inhibition surrounding the disk is mea-
sured, which indicates the antibiotic sensitivity of that isolate. This diameter is called the minimum inhibi-
tory concentration (MIC) for that antibiotic. The size of the growth inhibition zones can vary due to: (1) the
culture medium used; (2) incubation conditions; (3) the rate of antibiotic diffusion; and (4) the concentrations
of the antibiotics used. All these factors must be kept constant to make between experiment comparisons.
An antibiotic sensitive control must be used for comparisons (e.g., E. co/i ATCC 25923). Tests of each
isolate must be replicated to ensure reproducibility. Isolates are scored as sensitive, intermediate, or resistant
compared to the control for each antibiotic used. For all three approaches isolates are classified based on a
combination of the antibiotics (and concentrations if known) to which they are sensitive and resistant.
Carbon utilization
This method compares differences in the utilization of several carbon and nitrogen substrates by different
bacterial isolates. Substrate utilization can be rapidly scored by the formation of a purple color due to the
-------
reduction of a tetrazolium dye included with the substrates and automatically detected using a microplate
reader. Isolates are typically classified using only a subset of indicative substrate, for example, Hagedorn et
al. (2003) used only 30 of the 95 wells for their analysis.
This method was first investigated for potential use in MST because it is rapid, simple and requires little
technical expertise. It has been most successfully used for identification of isolated clinical Gram-negative
bacteria (Holmes et al., 1994). Its use in analysis of environmental samples has been questioned due to vari-
ability and poor reproducibility (Konopka et al., 1998; Tenover et al., 1995). It is possible to test substrate
utilization of each isolate using an array of substrates in the laboratory but the method has been simplified
by the availability of commercial microwell plates containing substrates. Most commonly used are Biolog
microplates (Hayward, CA), and more recently PhenePlate (PhP plates; Stockholm, Sweden).
Application of carbon utilization to MST
This method has been tested for use in MST only at a small scale (Hagedorn et al., 2003; Harwood et al.,
2003; Wallis and Taylor, 2003). In one study, 30 enterococci strains were isolated from stream sites where an
obvious source of pollution was apparent and analyzed using the Biolog system. Using a 365 isolate source
library, classification of sample isolates correctly matching the presumptive sources ranged from 86.6-93.3%.
However, in another study using the PhenePlate system, a larger number of enterococci isolates (1,766) from
six sources were compared, diversity was very high in wastewater samples (Simpson's Diversity Index=
0.95) and seabird feces (DI = 0.72) but much lower in animal feces such as cows (DI=0.32) (Wallis and
Taylor, 2003). High diversity increases the size of library needed to differentiate hosts. In a controlled study,
results of carbon utilization were compared to antibiotic resistance and found to be comparable (Harwood
et al., 2003). Although positive identification was high (93%), there were also a number of false positives
(51.5%). The authors speculated that the library was too small, resulting in the false positives.
Carbon utilization methodology
Very little preparation by the user is necessary since microwell plates with 95 different substrates may be
purchased from Biolog, (Hayward, CA) or those with 24 substrates from PhenePlate (Stockholm, Sweden).
Since the PhenePlate has only 24 substrates, one plate can be used for replication or different isolates. Dif-
ferent microwell plates are used for the analysis of Gram-positive bacteria (e.g., GP2 MicroPlate™, Biolog)
and Gram-negative bacteria (e.g., GN2 MicroPlates™, Biolog). Isolates are first grown and a liquid suspen-
sion of cells at a standardized turbidity is used to inoculate the microplates. After incubation at 37°C for 24 h,
presence or absence of growth is indicated by purple dye formation and is assessed manually or automatical-
ly using a plate reader (MicroLog™ System, Biolog). Discriminant analysis of the binary data from known
sources is then typically used to determine the substrate combination that best distinguishes the host.
3.3.2 Genotypic methods
Molecular (DNA) typing or fingerprinting tools are used to differentiate specific microorganisms. Bacteria, in
particular E. coli strains, have been analyzed by a variety of genotyping methods that vary in their sensitiv-
ity and technical complexity. Genotypic methods requiring a reference library fall into two categories, direct
analysis of the genome or indirect analysis after PCR. The sensitive and rapid nature of the PCR method
and its ability to amplify target sequences approximately 106-fold has made it an attractive method, and is
commonly used in many of the newer source tracking approaches. This section discusses the cultivation-
dependent library methods, but PCR is used both in cultivation dependent and independent approaches. The
latter are discussed toward the end of this Chapter. In general, methods that employ PCR are usually more
-------
rapid than those that directly examine the genome. The advantages of using PCR based method are: only a
small amount of starting DNA material is needed, often bacterial cells can be used without performing DNA
extraction, some analyses can be automated reducing labor costs, and most produce highly reproducible and
accurate fingerprint profiles. All the methods listed in this section require a laboratory and personnel with at
least basic equipment and expertise in molecular genetics.
rep-PC'R DNA fingerprinting
The repetitive element sequence-based PCR (rep-PCR) DNA fingerprinting technique (de Bruijn, 1992;
Versalovic et al., 1991; Versalovic et al., 1994) uses the polymerase chain reaction and primers to amplify
specific portions of the microbial genome, which are subsequently visualized following electrophoresis.
The primers used for rep-PCR DNA fingerprinting are complementary to naturally occurring, multi-copied,
conserved, repetitive, DNA sequences present in the genomes of most Gram-negative and Gram-positive
bacteria (Lupski and Weinstock, 1992). The repetitive elements are usually comprised of duplicated genes,
interspersed repetitive extragenic palindromes (REP) and other palindromic unit sequences, intergenic repeat
units (IRU), enterobacterial repetitive intergenic consensus (ERIC) sequences, bacterial interspersed mosaic
elements (BIME), short tandemly repeated repetitive (STRR) sequences, and Box elements (Sadowsky and
Hur, 1998). Three major families of repetitive sequences have been generally used for rep-PCR DNA finger-
printing: repetitive extragenic palindromic (REP) sequences (35-40 bp), enterobacterial repetitive intergenic
consensus (ERIC) sequence (124-127 bp) and the 154 bp Box element (Versalovic et al., 1994). The use of
these primer(s) coupled with PCR leads to amplification of the specific genomic regions located between
adjacent REP, ERIC or Box elements. While the methods done using these sequences should be referred to
as REP-PCR, ERIC-PCR and Box-PCR genomic fingerprinting, respectively, collectively the technique is
referred to as rep-PCR genomic fingerprinting (Versalovic et al., 1991; Versalovic et al., 1994). The resulting
mixture of amplified DNA fragments is resolved in agarose gels, producing a banding profile referred to as
a rep-PCR genomic DNA fingerprint (Versalovic et al., 1994). Thus, the banding pattern serves as a "finger-
print" for strain identification or analysis of microbial populations. Bacteria having identical fingerprints are
regarded as being the same strain, and those having nearly identical or similar banding patterns are regarded
as being genetically related.
This method has been used extensively because it is rapid and relatively simple. Among the molecular ge-
notyping methods, it is the least expensive and requires less technical expertise. Only the basic equipment
present in most laboratories performing molecular genetic analyses is necessary unless higher throughput
and greater accuracy is desired and one chooses to use an automatic sequencer or fluorescence scanner (see
HEFERP below).
Application of rep-PCR to MST
The rep-PCR DNA fingerprinting technique is relatively quick, easy, and inexpensive to perform, and lends
itself to high throughput applications, making it an ideal method for microbial source-tracking studies (Car-
son et al., 2003; Dombek et al., 2000; Johnson et al., 2004; Lipman et al., 1995; McLellan et al., 2003). In
studies where rep-PCR have been compared to other methods, it has been shown to give better predictions
than ribotyping (Carson et al., 2003), and rDNA intergenic spacer region (ISR)-PCR (Seurinck et al., 2003).
Overview of rep-PCR methodology.
The rep-PCR DNA fingerprinting technique is amenable for use with DNA templates produced using a vari-
ety of methods. These include liquid cultures, colonies, and purified DNA. Using colonies directly instead
-------
of performing DNA extraction reduces the time and cost of using this method, particularly in comparison to
other genetic fingerprinting methods. Among the primers used for rep-PCR, the Box primer AIR has proven
to be the most useful in distinguishing environmental isolates of E. coli (Dombek et al., 2000). There are
two general methods to perform rep-PCR DNA fingerprinting, differing in the way in which the DNA frag-
ments are visualized. In the first more conventional method, the resulting DNA fragments in agarose gels
are visualized following staining in ethidium bromide (de Bruijn, 1992). Despite careful attention to detail,
it is often difficult to get rep-PCR gels to run consistently straight and avoid lane distortions, which makes
alignment and comparisons within and between gels difficult. To overcome these major limitations, a second
method has been developed, a horizontal, fluorophore-enhanced, rep-PCR (HFERP) technique (Johnson et
al., 2004). The technique is similar to that previously described for use with a DNA sequencer (Rademaker
and deBruijn, 1997; Versalovic et al., 1995). In HEFERP, however, a standard horizontal agarose gel elec-
trophoresis system and a dual-wavelength scanner are used. HFERP is ideal for high throughput analyses of
bacteria and the protocol can be geared for 96 well microplates using colonies (details at http://www.ecoH-
rep.umn.edu/a_hferpoverview.shtml).
Randomly Amplified Polymorphic DNA (RAPD) analysis
The Random Amplified Polymorphic DNA Analysis (RAPD), and Arbitrary Primed Polymerase Chain Reac-
tion (AP-PCR) techniques (Welsch and McClelland, 1990; Williams et al., 1990) represent two independent-
ly developed, but conceptually-related methods that have found extensive use in studies of microbial epi-
demiology, diversity, population genetics, taxonomy, evolution, and ecology (Mathieu-Daude et al., 1998).
Both methods rely on the fact that PCR conditions done using arbitrary primers at low stringency (AP-PCR)
or with non-selective primers at high stringency (RAPD) produce a series of strain specific PCR products
that depend on both the primer and template used. When separated on agarose gels and stained with ethidium
bromide, these PCR products produce a series of species- or strain-specific bands that act as a fingerprint of
the bacterial genome. A subsequently developed method, DNA Amplification Fingeiprinting (DAF) (Caeta-
no-Anolles et al., 1992) differs from AP-PCR and RAPD in that a polyacrylamide gel and silver staining is
frequently used to visualize the PCR products.
RAPD analyses are relatively inexpensive when compared to other molecular methods like ribotyping and
pulse field gel electrophoresis (PFGE), require no previous knowledge of the genome examined, are ame-
nable to using colonies, boiled preps, or purified DNA, and can be scaled-up for high throughput analyses.
However, it has been shown that RAPD analyses are susceptible to the buffers used, cycle number, primer
choice, and method of DNA preparation (Hopkins and Hilton, 2000; Mathieu-Daude et al., 1998, Wang et al.,
1993). Consequently, it has been reported that RAPD analyses may not be reproducible and suffer from lab-
to-lab variation (Hilton et al., 1997; Hopkins and Hilton 2000; Penner et al., 1993). Nevertheless, Wang and
co-workers (1993) reported that RAPD analyses were more sensitive than multilocus enzyme electrophoresis
in differentiating among E. coli strains. It has been suggested that some variation may be eliminated by the
use of standardized reagents and kits (Hopkins and Hilton, 2000).
Application of RAPD to MST
RAPD analyses have been used to examine genetic diversity of E. coli obtained from animals (Aslam et al.,
2003), feedlots (Galland et al., 2001), humans (Pacheco et al., 1997; Vogel et al., 2000), and in culture collec-
tions (Wang et al., 1993). There has been considerable interest in using RAPD analyses to detect and analyze
E. coli O157:H7 (Galland et al., 2001; Hopkins and Hilton, 2000; Radu et al., 2001) and enterotoxigenic E.
coli (Pacheco et al., 1996; Pacheco et al., 1997). Ting et al. (2003) reported that RAPD fingerprints might be
-------
useful for differentiating among human and non-human sources of E. coli contamination. However, in gen-
eral terms RAPD analyses have only been preliminarily tested for use in MST studies (2003).
Overview of RAPD methodology.
While RAPD and AP-PCR DNA fingerprinting have sometimes been used synonymously, in AP-PCR a
single or sometimes two arbitrary primers are used in PCR under low stringency conditions and priming is
done with sequences having the best match, with some mismatches. In contrast, RAPD DNA fingerprinting
is often done at high stringency conditions using primers with low selectivity that anneal at the Tm of the
primer. This is thought to result in priming of genomic DNA with less mismatches than is seen with AP-PCR
(Mathieu-Daude et al., 1998). RAPD DNA fingerprinting is typically carried-out using 10-mer random prim-
ers. These primer sets are commercially available (e.g. Genosys Biotechnologies or Amersham Biosciences
Ready-to-Go RAPD Analysis) and can be initially screened for discrimination ability using the organism of
interest. Several RAPD primers have been found to be useful to differentiate among E. coli strains (Madico
et al., 1995; Pacheco et al., 1997; Wang et al., 1993).
Amplified Fragment Length Polymorphism (AFLP) analysis
Amplified fragment length polymorphism (AFLP) is a powerful and sensitive DNA fingerprinting technique,
which was originally developed to map plant genomes (Blears et al., 1998; Lin and Kuo, 1995). It uses a
combination of genomic DNA digestion with restriction enzymes and PCR. In this method short adaptors are
ligated (attached) to the digested fragment ends to provide sufficient length of known sequence for primers
to be used for PCR. To amplify all of the digested fragments by PCR would result in a multitude of prod-
ucts that would be too difficult to resolve. To overcome this problem, additional PCR primers are used for a
second round of PCR. These primers differ from the initial primers by the addition of 1-3 nucleotide bases
resulting in the amplification of just a subset of the initial fragments. The addition of more nucleotides to the
end of the primers increases the specificity and decreases the number of resultant PCR products. Separate
reactions using different primers sets are often used and the data combined providing a substantial number of
data points to be used to discriminate isolates. If sufficient number of primers are used the entire genome can
be accurately sampled using this approach (Arnold et al., 1999). However, in most cases only about three
primer sets are needed to obtain sufficient resolution between isolates. Currently, there is no standard set of
primers designated for MST or for any bacterial species.
The need to conduct genomic DNA digestion and PCR makes this method more time consuming and more
expensive than other methods that use only PCR. Of all the PCR based methods this one can produce the
most bands, which provides a better chance for distinguishing isolates but also increases the need to precisely
discriminate bands. Using an automatic sequencer improves band discrimination, decreases time and labor
but adds to the costs both in purchasing the equipment and supplies.
Application of AFLP to MST
The AFLP method has been used to fingerprint different bacterial species and is reported to be more sensitive
in the detection of DNA polymorphism in them (Clerc et al., 1998; Lin and Kuo, 1995; Restrepo et al., 1999;
Valsangiacomo et al., 1995). The majority of studies have been focused towards epidemiology and not MST.
The number of MST studies to date is limited but suggests that its resolution is as good, or better, than most
other genetic fingerprinting MST methods (Guan et al., 2002; Hahm et al., 2003a, b; Leung et al., 2004).
-------
AFLP was compared to MAR and 16S ribosomal (rRNA) gene sequences in E. coli collected from livestock,
wildlife, or human feces (Guan et al., 2002). Discriminant analysis indicated AFLP was better than MAR
and rRNA gene sequence analysis at assigning isolates correctly to each source. Another study comparing
E. coli isolates obtained from cattle, humans and pigs using AFLP and ERIC-PCR revealed similar results
(Leung et al., 2004). There was greater than 90.6-97.7% correct classification using AFLP and 0-75% for
ERIC-PCR. A third study compared a number of different methods but mainly examined E. coli serotype
O157:H7 isolates and only a small number of environmental isolates (Hahm et al., 2003a). However, that
study and some follow-up work (Hahm et al., 2003b) suggested that AFLP resolved strain differences in E.
coli at the same level as PFGE. More fundamental research is still needed to determine the best primer sets
to use for different levels of discrimination between isolates. At the same time, when considering this ap-
proach the expertise, time and cost factor should be compared to other genotyping methods for the accuracy
achieved.
Overview of AFLP methodology
Isolates are analyzed using an AFLP fingerprinting kit following the instruction of the manufacturer (Gibco
BRL). Briefly, DNA is extracted from cultures using any standard total genomic DNA isolation method.
Purified DNA is then digested with a frequently cutting and a less frequently cutting restriction enzyme Msel
and EcoRI, respectively, and the fragments are ligated to EcoRI and Msel adapters to generate template DNA
for PCR amplification. This restriction-ligation mixture is diluted and amplified with EcoRI and Msel core
sequence primers for pre-selective amplification. Selective amplification is then performed using primer sets
with additions of 1-3 arbitrary nucleotide sequences on the 3' end of each. Eight primers of each EcoRI and
Msel adapters are provided with the AFLP kit. A total of 64 combinations of primer pairs can be used for
PCR amplification. Three commonly used selective primer sets are: EcoRI-A (FAM™) plus Msel-C, EcoRI-
0 (FAM™) plus Msel-CG and EcoRI-C (NED™) plus Msel-C. The primers used for PCR amplification are
fluorescently labeled (e.g., FAM™ and NED™) for automatic detection of the different size products using
an automatic sequencer. This also allows high throughput analysis of AFLP patterns. Labeled size markers
(DNA size markers) are included in each lane to ensure accuracy of band detection and differentiation. The
typical size range of amplification products is between 50 and 4000 bp. The number of isolates that can be
analyzed in a single run depends on the automated sequencer that is being used. All information collected
from the sequencer is then transferred to a fingerprint analysis program (e.g., Bionumerics, Applied Maths).
The data is binary, based on the presence and absence of bands in each profile (see data analysis Chapter).
Pulse Field Gel Electrophoresis (PFGE)
The most common genotyping method used in epidemiological investigations is pulse field gel electropho-
resis (PFGE) of total genomic DNA after restriction enzyme digestion using an infrequently cutting enzyme
(Tenover et al., 1995). It involves direct analysis of the microbial genome and PCR is not performed. Diges-
tion of total genomic DNA by an infrequently cutting restriction enzyme, results in the production of 10 to 30
large fragments. These fragments are too large to be separated in a standard agarose gel electrophoresis unit
because the gel pore size limits their migration. To overcome this limitation PFGE was developed in which
the orientation of the electric field is changed at different intervals allowing the large DNA molecules to re-
orient themselves at regular intervals and "snake" through the pores. The most commonly used instruments
apply a contour-clamped homogeneous electric field (CHEF) (Chu et al., 1986). To optimize separation it is
often necessary to vary the angle, pulse time and voltage. The top of the line CHEF electrophoresis unit is
computerized; the desired fragment size range is entered and optimal separation conditions are automatically
obtained. Fragment sizes are determined by comparison to molecules of known size. This also provides a
means to perform between gel fingerprint comparisons.
-------
The PFGE technique is time consuming and very tedious, thus may not be suitable for rapid identification of
large number of strains (Willshaw et al., 1997) often necessary for MST. While PFGE requires a specialized
gel rig with multiple electrodes configured in a hexagonal design, a chiller and pump, and programmable
power supply, the operator does not require special molecular skills. However, the PFGE aparatus is more
expensive that conventional gel electrophoresis. Since only a limited number samples can be processed per
gel, the number of available apparatuses is the limiting factor for high throughput analysis using this method.
Sample preparation does require some training but with experience many samples can be prepared daily.
Application of PFGE to MST
This method has been described as "superior to most other methods for biochemical and molecular typing"
(Olive and Bean, 1999). The Centers for Disease Control and Prevention (CDC) has adopted this method
for their "National Molecular Subtyping Network for Foodborne Disease Surveillance" mainly to discrimi-
nate E. coli O157:H7 and other foodborne pathogens. They have developed a network for health agencies to
quickly compare molecular PFGE genotype data at a centralized website called PulseNet (http://www.cdc.
gov/pulsenet/). It has been used successfully to rapidly compare PFGE profiles of suspect cultures with those
in the national database at CDC. In the future, this could serve as a model if EPA adopts any of the geno-
typic fingerprinting approaches for MST. However, publications of MST studies are much more limited. In
a beach study, PFGE of E. coli was better for discriminating host sources compared to the other fecal coli-
forms, Klebsiella, Citrobacter, and Enterobacter sp. (McLellan et al., 2001). As mentioned in the previous
section, Hahm et al. (2003a) found levels of discrimination using PFGE similar to AFLP. The same study
and another by McLellan et al. (2001) found that methods such as rep-PCR were less discriminatory than
PFGE. However, high resolution between fingerprint patterns is not always ideal when genetic diversity is
high between isolates taken from the same host animal. Greater genetic diversity translates into an increase
in reference library size needed to differentiate isolates from different hosts. Also, care must be taken in the
restriction enzymes chosen for this analysis because no relationship between fragment pattern and source
was seen when the restriction enzyme Sfil was used (Parveen et al., 2001).
Overview of PFGE methodology.
There are no standardized methods for PFGE for MST, but protocols set for CDC studies can be used for E.
coli isolates. This method can be used on any bacteria, but conditions for optimal DNA extraction must first
be determined. Isolates are first grown using standard conditions then DNA is extracted using an agarose
plug total genomic DNA isolation method, which minimizes undesired breakage of the DNA. The cells are
pelleted by centrifugation then suspended in unmolten low melt agarose or equivalent agarose specialized for
PFGE. Sufficient microbial biomass must be used to have at least 1 (ig of DNA in the plugs used for diges-
tion. While still liquid, the agarose/cell solution is transferred to plug mold where it is left to solidify. Once
solid, the plugs are removed from the mold and put through a series of steps to lyse the cells, remove proteins
and degrade RNA. Depending on the protocol used, this process can take from a few hours to two days.
Purified DNA still embedded in the agarose plugs is then digested with a rare cutting restriction enzyme.
The most commonly used restriction enzyme is Xba\ but others have also been tested and show variable
results. Electrophoresis performed at 14°C with 6 V/cm, angle 120°, linear ramping factor and 30 hr run-
ning time will separate digested DNA fragments ranging between 100 kb and 500 kb in size. Gels are stained
with ethidium bromide after fragments have been separated. It is often necessary to destain for several hours
to optimize band contrast. Gel images can be digitized and then entered into a fingerprint analysis program
(e.g., Bionumerics, Applied Maths). The data is binary, based on the presence and absence of bands in each
profile (Chapter 4).
-------
Ribotyping
Ribotyping is a version of restriction fragment polymorphism (RFLP)-Southern hybridization analysis
(Demezas, 1998; Sadowsky, 1994) that has found wide application in the subtyping of a variety of Gram-
negative and Gram-positive bacteria (Olive and Bean, 1999). It is another method that does not include PCR,
except in the making of the labeled rDNA probe. The technique has been broadly used in molecular epide-
miology (Bingen et al., 1992; Bingen et al., 1996; Picard et al., 1991), and taxonomic identification (Brisse
et al., 2000) studies, including those with E. coli (LiPuma et al., 1989; Stull et al., 1988; Tarkka et al., 1994).
RFLP patterns of bacterial genomic DNA made with moderate cutting enzymes contain too many fragments
for easy analysis, but ribotyping takes advantage of selective hybridization of a limited number of frag-
ments for strain differentiation. Ribotyping is based on the detection of genetic differences in the genomic
sequences within or flanking the 16S and 23S rRNA genes. Since rRNA genes exist as several copies (2-11)
in the bacterial genome and are highly conserved among bacteria, (Grimont and Grimont, 1986), hybridiza-
tion of restriction enzyme-digested genomic DNA with labeled rDNA probes produces a ladder of labeled
fragments that resemble a bar code. In addition, it has been recognized that since ribotyping produces rela-
tively few bands for each strain (approx. 5-15 for E. coli, depending upon the enzyme used and the strain),
the technique is amenable to computerized analyses (Lefresne et al., 2004, Machado et al., 1998). If greater
discrimination between strains is desired, more than one restriction enzyme can be used to digest DNA, and
the banding patterns produced by each enzyme are combined to form a composite pattern (Harwood et al.
2003; Jenkins et al., 2003).
Ribotyping is a relatively demanding procedure requiring multiple steps and some specialized equipment.
The need for specialized training, high supply costs and the time required to complete the procedure are dis-
advantages of using this method. However, the recent development of an automated ribotyping instrument,
the Riboprinter (DuPont-Qualicon, Wilmington, Delaware) has promoted renewed interest in using ribotyp-
ing as a molecular tool for epidemiological, MST, and clinical studies (Ito et al., 2003). However, the instru-
ment has limited throughput, analyses are relatively expensive, and there have been reports that automated
riboprinting may not be as reliable as manual methods (Grif et al., 1998). Despite these shortcomings, several
MST studies have used automated riboprinters to examine genetic diversity and groupings of fecal bacteria
from known animal sources and the environment.
Application of ribotyping to MST
Ribotyping has been widely used in microbial source tracking studies (Farag et al., 2001; Carson et al.,
2001; Carson et al., 2003; Hartel et al., 1999; Hartel et al., 2002; Harwood et al., 2003; Jenkins et al., 2003;
Parveen et al., 1999; Scott et al., 2003). While the authors of these studies used the same basic technique,
different laboratories have used different restriction enzymes in their analyses, and some have used a two-
enzyme scheme, making comparisons difficult. As with any genotypic method, lab-to-lab variation, issues of
repeatability, within and between gel variability and methods of analysis often make comparison of results
done in different laboratories difficult (Lefrense et al., 2004). Moreover, several different studies done using
slightly different procedures have reported variable results with respect to the ability of ribotyping to differ-
entiate among bacteria isolated from different animal hosts (Carson et al., 2003; Hartel et al., 2002; Parveen
et al., 1999; Scott et al., 2003). Furthermore, database size, geographic distribution of the isolated bacteria,
and the presence of replicate isolates in the bacterial source library impact the ability of ribotyping to differ-
entiate among bacteria at the host species level (Scott et al., 2002; Scott et al., 2003).
Overview of ribotyping methodology
The ribotyping method is carried out in multiple steps. The technique involves restriction enzyme digestion
of genomic DNA, separation of fragments by gel electrophoresis, immobilization of DNA fragments to a
-------
solid matrix (e.g., nylon membrane) by Southern transfer and subsequent hybridization using a labeled probe
of the E. coll rRNA genes or the entire operon (Grimont and Grimont, 1986). Several different procedures
can be used to isolate bacterial DNA (see Sadowsky, 1994) for ribotyping and several different restriction
enzymes may need to be tried to show differences at the strain level (Lefresne et al., 2004; Martin et al.,
1996; Parveen et al., 1999; Scott et al., 2002). However, while EcoRl, Pvull and Hindlll have frequently
been used for source tracking studies (Carson et al., 2001; Hartel et al., 2003; Scott et al., 2003; Vogel et al.,
2000) it has been suggested that two enzyme systems should be routinely used to increase the technique's
discrimination ability (Scott et al., 2003). The probes used for subsequent hybridization analysis can vary in
the different regions of the E. coli rRNA operon used, but most investigators use the entire E. coli rrnB rRNA
operon (Atwegg et al., 1989), only the 16S and 23S rRNA genes from E. coli, or mixtures of oligonucle-
otides complementary to specific regions in the operon (Gustaferro and Persing, 1992; Lafresne et al., 2004).
The probe is usually generated by PCR, but can also be generated by nick translation or random primer label-
ing (Ausubel et al., 2004) and labeled with 32P-, DIG-, or chemiluminescent-labels (Gustaferro and Persing,
1992; Regnault et al., 1997). Next hybridized fragments that constitute the ribotype banding patterns are
detected using autoradiography or color formation. When the Riboprinter (DuPont-Qualicon, Wilmington,
Delaware) is used, the sample (typically one bacterial colony) is added into the first tube and the instrument
automatically carries out subsequent steps.
3.4 Cultivation-dependent/library-independent methods
When the target for MST is typically found in low numbers, it is first necessary to enrich the sample or
obtain isolates. Enrichments are typically performed under conditions that favor the target organism. These
methods are based on presence or absence of the target organism or gene therefore a source library is unnec-
essary.
F+RNA coliphage typing
F+RNA coliphages can help distinguish human and animal waste contamination by typing isolates into one
of four subgroups (Alderisio et al., 1996; Brion et al., 2002; Cole et al., 2003; Griffin et al., 2000). Ecology
studies have demonstrated that groups I and IV are generally associated with animal feces, whereas groups
II and III are more sewage-specific (Furuse, 1987). Schaper et al. (2002a) found these associations to be
statistically significant but also noted that exceptions occur. Serotyping or genotyping can be used for typing
of F+RNA coliphages. In serotyping, group-specific antisera are used whereas in genotyping, hybridization
with group specific oligonucleotides is used (Beekwilder et al., 1996; Hsu et al., 1995).
Coliphage cultivation techniques are simple with low supply costs (only plates and media), but require an
overnight incubation step. Molecular methods have also been developed that allow for more rapid char-
acterization of coliphages. For example, Vinje et al. (2004) have developed an RT-PCR and reverse line
blot hybridization technique capable of rapid detection and genotyping of coliphages. Additionally, phage
characterization studies are underway which may allow for identification of more refined and host-specific
subgroups. These advances could lead to an improved and more specific phage genotyping system.
Application of F+RNA coliphage typing to MST
The use of coliphage typing for MST is library independent, but can only currently be used to broadly dis-
tinguish human and animal fecal contamination. Coliphages have been detected in domestic, hospital, and
slaughterhouse wastewaters (Funderburg and Sorber, 1985) and from treated wastewaters (Gantzer et al.,
1998) but there appears to be some limitation when individual samples are used (Noble et al., 2003). Quan-
-------
titative source tracking using F+RNA coliphage typing may be problematic owing to differential survival
characteristics of the subgroups (Brion et al., 2002; Schaper et al., 2002b).
Overview of coliphage typing methodology
Methods for isolation of coliphages include two standard USEPA procedures. One is Method 1601, a two-
step enrichment procedure (USEPA, 200la). The second is Method 1602, the single agar layer procedure
(USEPA, 2001b). Method 1601 is more sensitive than 1602, but may not be the best choice for isolation of
F+RNA coliphages meant to be subsequently typed for MST. The enrichment step likely excludes or masks
other strains that may have been present in the original sample, typically resulting in only one strain of phage
isolated from any given sample. The single agar layer procedure, on the other hand, is a pour plate technique
from which viruses can be easily isolated for subsequent typing.
Isolated viruses are grown in the presence of RNase A to distinguish F+RNA coliphages from F+DNA
coliphages. F+RNA coliphages cannot form plaques when RNase A is present. Then either a serotyping or
genotyping method is used to identify the F+RNA coliphages. For serotyping, virus infectivity is tested in the
presence of group-specific antisera. Inhibition of infectivity in the presence of a particular antiserum identi-
fies the group to which an isolate belongs. Coliphages are genotyped by using hybridization of group specific
labeled probes. Nucleic acid isolation is not necessary and plaques can be used directly for hybridizations.
Group I, II, III, or IV specific probe sequences are used for hybridization (Hsu et al., 1995; Beekwilder et al.,
1996). Identification of human source contamination is indicated by the hybridization of group II or III and
animal sources by group I or IV.
Gene specific PCR
Gene specific PCR methods have been developed for E. coli carried by humans (Oshiro et al., 1997), cattle
and swine (Khatib et al., 2002; Khatib et al., 2003,). It is anticipated that methods will soon be available for
E. coli carried by several other species of mammals and by birds. These methods are based on the discovery
that certain enterotoxin genes are carried almost exclusively by E. coli that infect individual species of warm-
blooded mammals. The STIb gene, the LTIIa gene and the STII gene are carried only by E. coli of human,
bovine and swine origin, respectively. Similarly, enterococci virulence genes have been used as targets for
host specific markers (Scott et al., 2005).
This two-step approach is relatively simple and can be performed within two working days. The biggest
advantages of these gene specific methods are that they are highly specific and they are library independent.
The biggest disadvantage is that the toxin genes are carried only by a small number of isolates, which makes
it necessary to perform a cultural enrichment step prior to testing by PCR.
Application of gene specific PCR to MST
This method is still in the developmental stages and there are no publications with its application for MST.
However, there is some indication that the prevalence of these genes in animal waste systems is greater than
previously expected (Chern et al., 2004) suggesting that it has potential in the future.
Overview of gene specific PCR methodology
Samples (1 L) are collected in sterile containers and shipped on ice to the laboratory and are processed within
24 hours. Samples are processed using membrane filtration, with filters being placed on mTEC agar and
-------
mTEC agar plus Congo Red. The mTEC plates are incubated for 1.5 hours at 35°C then at 44°C overnight.
The 10° and 10"' dilutions are harvested after 24 hours and the DNA is extracted. When samples contain suf-
ficient particulate matter to clog the filters, several filters of the 10"' dilution could be used. DNA extracts are
pooled and stored at - 80°C until nested PCR amplification. Two sets of primers are used for each toxin trait,
an outer primer set and a second set. All PCR amplicons are visualized through gel electrophoresis. Confir-
mation may be done by restriction fragment analysis or Southern blot hybridization using probes previously
designed specifically for each toxin.
Recently, magnetic beads were used to increase the sensitivity of the LTIla biomarker for cattle. In this
method, total DNA was extracted either from the mTEC medium colonies or directly from the environmental
samples. Next, the LTIla gene was removed from the DNA mixture by hybridization with magnetic beads
containing the LTIla probe. Finally PCR was used to amplify the LTIla gene for detection by gel electropho-
resis and staining. The combination of magnetic beads followed by PCR resulted in an increase in sensitivity
over the nested PCR technique by as much as 10,000 fold, even in the presence of PCR inhibitors such as
humic acids (Tsai et al., 2003).
3.5 Cultivation-independent/library-independent methods
Cultivation-independent methods for MST are primarily based on nucleic acid techniques arising from the
field of molecular microbial ecology. Molecular microbial ecology began in the 1980's with the develop-
ment of a phylogenetic framework for the placement of all organisms into one of three domains (Bacteria,
Archaea, or Eukarya) based solely its rRNA gene sequences (Head et al., 1998; Olsen et al., 1986). As rRNA
gene sequences accumulated into publicly assessable databases (Ribosomal Database Project (RDP), http://
rdp.cme.msu.edu, GenBank at the National Center for Biotechnological Information (NCBI), http://www.
ncbi.nlm.nih.gov), the level of classification based on rRNA gene sequences increased. Today, most organ-
isms can be classified from Kingdom to the genus-species level based on their rRNA gene sequences. Phylo-
genetic analysis of microbial communities based on rRNA gene sequences has been applied to many envi-
ronments including soil, water, extreme environments and animal gastrointestinal tracts (Zoetendal, 2004).
In molecular microbial ecology, methods can be broadly grouped into three categories: 1) those designed to
characterize or identify the members of a bacterial community; 2) those designed to measure large changes
in community structure; and 3) those designed to identify or quantify specific members of a community (for
reviews see Head et al., 1998; Zoetendal, 2004).
Total community analysis
Identification using 16S rRNA gene clone libraries
Microbial communities from environmental samples are frequently analyzed by the construction of 16S
rRNA gene clone libraries. Clone libraries can also be made from other genes but currently the gene with the
most available information is the 16S rRNA gene. Clone library construction and analysis is one of the more
expensive and time-consuming cultivation-independent methods. The generation of clone libraries requires
the combination of several molecular biological techniques including, nucleic acid extractions, PCR, DNA
ligation, bacterial transformation, and plasmid isolation, which may take up to a week to perform. In recent
years, these methods have been simplified by the use of commercial kits. Therefore, laboratory technicians
with minimal training can successfully generate clone libraries. DNA sequencing involves the use of costly
equipment and many laboratories send their DNA to specialized facilities for sequencing at a cost rang-
ing from around $4.00 to $20.00 a sequence. Thus, a large portion of the total cost is based on the number
of clones sequenced. DNA sequence analysis of clone libraries generates a large amount of electronically
-------
archival data, which may be time consuming to process. The analysis of this type of data requires, at a mini-
mum, an understanding of the publicly available sequence matching databases and programs. Realistically,
the time and cost to perform this method, one-month and $5-1 OK for 100 clones, does not make it an appro-
priate choice for MST. Its value lies in research and development of new approaches for MST.
Application of!6S rRNA gene clone libraries to MST
Construction of clone libraries from water samples for MST is not widely used because hundreds of se-
quences are needed to accurately profile an entire community. However, with regards to MST, the cloning
and sequencing of microbial communities from contaminated sites is useful for research purposes. At least
two studies (Cho and Kim, 2000; Simpson et al., 2004) demonstrated that the native microbial communities
in water are changed by the addition of fecal contamination. In both these studies fecal bacteria indicative
of the host source either bovine (Cho and Kim, 2000) or equine (Simpson et al., 2004) were detected. Also,
the construction and analysis of smaller clone libraries (< 50 sequences) from environmental samples can be
used to verify the specificity of specific primers (such as Bacteroides specific primers) used in PCR assays or
verify the presence of host-specific bacteria in the environmental sample.
Overview of!6S rRNA gene clone library methodology
In this method, nucleic acids are extracted and then amplified using primers designed to match the 16S rRNA
genes from as many bacterial species as possible (for a review of available general primers see Baker et al.,
2003). The 16S rRNA genes from the microbial community are cloned into plasmids and transformed into
E. coli to construct a library containing many individual E. coli colonies, each containing a different 16S
rRNA gene. Individual E. coli colonies are propagated, and the 16S rRNA genes carried in the plasmid are
isolated and sequenced. The 16S rRNA sequences representing the microbes from the environmental sample
are analyzed by comparison with other sequences in publicly available databases using the BLAST program
at NCBI (www.ncbi.nlm.nih.gov/BLAST) or the Similarity program at RDP (rdp.crne.msu.edu/html). Addi-
tionally, taxonomic or similarity relationships can be determined using cluster analysis and tree construction
programs based on the number of matching base pairs between the sequences (Olsen, et al., 1986). When
phylogenetic trees are constructed, the relationships between microbial sequences are generally presented as
OTUs (operational taxonomic units), clusters, or clades, because phenotypic information is needed to de-
scribe or confirm bacterial species.
Community structure by fingerprinting
Fingerprinting methods are often used to monitor changes in a community or to compare communities be-
cause the expense and labor involved in the construction and analysis of clone libraries limits the number of
samples that can be analyzed (Table 1). Essentially, all of the cultivation-independent fingerprinting methods
examine DNA size or conformation profiles generated from a microbial community after PCR amplifica-
tion of rRNA genes, or randomly amplified DNA fragments. The amplicons may be separated based on
sequence-specific melting behavior of amplicons by denaturing gradient gel electrophoresis (DGGE) or
temperature gradient gel electrophoresis (TGGE) (Muyzer and Smalla, 1998). In addition, one of the primers
used for PCR amplification may be labeled fluorescently, and amplicons can be separated by size before re-
striction enzyme digestion (Length Heterogeneity Restriction Fragment Length Polymorphisms, LH-RFLP)
or after restriction enzyme digestion (Terminal Restriction Fragment Length Polymorphisms, T-RFLP) (Liu
et al., 1997). The underlying principle for all of these fingerprinting methods is that differences in band-
ing patterns result from differences in microbial species comprising the community. Amplification with
generalized PCR primers from environmental samples usually results in a large number of bands, which are
-------
analyzed by band matching computer programs and statistically using cluster analysis. DNA bands can be
extracted from the gels and sequenced to identify the key members of the microbial community.
PCR methods using standard thermocyclers are relatively inexpensive and easy to perform with minimal
training. Most fingerprinting methods can be performed in about one day with the electrophoresis separa-
tion run overnight, thus allowing data analysis the next morning. Differences in cost between fingerprinting
methods will occur depending on the type of post-analysis performed. In general, electrophoresis methods
with better resolution require more costly equipment. For instance, gel electrophoresis equipment designed
to separate PCR products by temperature gradient gel electrophoresis is more expensive than standard gel
electrophoresis equipment. LH-RFLP and T-RFLP are one of the most expensive fingerprinting methods
because separation of DNA fragments differing by only a single base pair requires acrylamide gels and DNA
sequencing equipment. As with DNA sequencing, the separation of the DNA fragments on an automated
DNA sequencer may be subcontracted to a specialized facility.
Application of community structure to MST
Although, fingerprinting analyses of fecal samples have been used to demonstrate host-specificity of the mi-
crobial community with the animal host (Zoetendal et al., 2004), cultivation-independent community analy-
sis by fingerprinting has not been widely applied to MST studies. This is probably in part because in water
samples the portion of the community that can be linked to host specificity may be very small compared
to indigenous microbial community. Fingerprinting methods can be linked with more specific primers to
produce fewer DNA bands. In one study relevant to MST, LH-PCR methodology was used with Bacteroides
primers to identify a band size distinctive of bovine specific Bacteroides (276 bp) (Field et al., 2003). Addi-
tional digestion of the PCR amplified sequences with restriction enzymes (T-RFLP) resulted in the detection
of two additional markers for bovine-specific Bacteroides and one marker for human-specific Bacteroides.
These researchers previously demonstrated that LH-PCR could be used with Bifidobacterium specific prim-
ers to detect a bovine-specific amplicon of 453 bp. Digestion of the Bifidobacterium amplicons with restric-
tion enzymes resulted in human and bovine specific fragments (Bernhard and Field, 2000a).
Overview of community structure methodology
Detailed explanations of community structure analysis are available from other sources (Liu et al., 1997;
Muyzer et al., 1996; Nakatsu and Marsh, 2005). Briefly, both methods use PCR to amplify the rRNA gene.
Typically, universal primers targeting the small subunit, 16S rRNA gene in bacteria are used to amplify se-
quences directly from DNA or RNA extracted from environmental samples. However, primers that amplify
specific groups such as Bacteroides is often more useful for MST. In general, primers selected for T-RFLP
amplify almost the entire 16S rRNA gene whereas in DGGE primers generating PCR products less than 500
bp are selected to reduce the occurrence of artifacts. In DGGE, the PCR products are directly analyzed by
gel electrophoresis whereas in T-RFLP the PCR products are first digested with frequently cutting restriction
enzymes before electrophoresis. In T-RFLP either one or both primers are labeled with different fluorescent
tags to allow visualization and distinction of the end fragments using an automatic sequencing system. In
DGGE, the PCR products are separated in gels composed of a gradient of chemical denaturants that causes
differences in DNA migration based on their sequence. In both methods differences in migration of PCR
amplicons either because of fragment sizes or sequence composition, generate a fingerprint of the community
and a view of its complexity.
Alternate targets
While most of the culture independent/library independent methods have targeted fecal bacteria and viruses,
eukaryotic cells have also been suggested as useful markers for fecal source identification. For example,
-------
species of the genus Cryptosporidium has been shown to exhibit some degree of host specificity based on
sequence differences in the rRNA gene (Xiao et al., 2004). These differences have been used to characterize
the primary fecal sources of surface water and wastewater (Xiao et al., 2001). However, because C. parvum
is found in relatively low numbers in environmental waters with moderate level of fecal contamination, their
use in MST will have the same problems as with enteric viruses, this is, the need of concentration steps from
large volumes of water.
Recently, PCR-based assays targeting host mitochondrial genes were used to discriminate between human,
bovine, porcine, and ovine fecal samples (Martellini et al., 2005). The assays were developed to produce
PCR products of different length facilitating their use in a multiplex PCR approach. The use of host mito-
chondrial PCR approaches is based on the fact that as gut epithelial cells become senescent they are shed into
the gut lumen, after which they become part of the animal feces. The presence of relatively large numbers
mitochondrial genes per eukaryotic cell increases significantly the detection sensitivity of this method. This
is a significant advantage over other gene specific PCR methods which normally target markers with less
than five copies per cell. The expected limited survival of gut epithelial cells might limit the use of this ap-
proach to recent fecal contamination events in areas nearby fecal inputs.
3.6 Identification and quantification of specific bacteria
Identification and quantification of microbes in environmental samples by cultivation independent methods is
dependent on sequence information derived from clone libraries (see above section) or sequencing of genes
from cultivated organisms. Identification and quantification methods can be divided into direct probing
methods not requiring PCR or PCR-based methods.
Direct probing of specific genes
Originally, direct probing methods were used to quantify microbes in cultivation-independent studies
(Giovannoni et al., 1998; Stahl et al., 1998). Hybridization methods usually use small oligonucleotide
sequences (less than 25 base pairs) called probes designed to hybridize with target DNA sequences.
Direct probing methods are moderately time-consuming and may require specialized training depending on
the method used to label and detect the probe. In recent years, the use of radioactive probes, which require
licensees and training to use, have been replaced by non-radioactive labels. Filter membrane hybridiza-
tion methods, such as dot blot hybridization or Southern blot hybridization, require multiple handling steps
including DNA extraction, blocking of nontargeted sequences, hybridization of targeted gene, and washing
unincorporated probe. The total process may take one to three days depending on the method used to mea-
sure the amount of probe bound to the filter. The cost for reagents is relatively inexpensive. Fluorescent in
situ hybridization (FISH) using fluorescently labeled probes can also be performed directly on bacterial cells
on a microscope slide. The total process of fixing the cells to a slide followed by hybridization and wash-
ing takes one to two days and the cost of the reagents is also relatively inexpensive. However, visualization
of the fluorescent signal in bacterial cells requires the use of a high quality epifluorescence or confocal laser
scanning microscope and specialized imaging software. This equipment is expensive and requires special-
ized training.
Application of direct probing to MST
This method has not been used directly in any MST studies. Although numerous probes for quantifying fecal
bacteria have been designed for dot blot hybridization (Matsuki et al., 2002; Wang et al., 2002), the method
is used infrequently because quantitative PCR (QPCR) methods have a detection limit 0.01% compared to
10% for dot blot hybridization (Malinen et al., 2003). FISH is an effective method for monitoring population
-------
changes in fecal samples (Franks et al., 1998) but has not been widely applied to MST because the concen-
trations of bacteria in water samples are generally too low to measure by FISH and fluorescent microscopy.
In addition, it is possible for the the intracellular rRNA concentration of fecal bacteria in aquatic environ-
ments below the detection limits of this technique, primarily due to nutritional stress. However, the coupling
of flow cytometry with FISH may improve the sensitivity of detection and the number of samples that can be
processed (Rigottier-Gois et al., 2003) allowing future MST applications.
Overview of direct probing methodology
In dot blot hybridizations, DNA extracts are bound to nylon membranes and probes are labeled with radioac-
tive 32P or non-radioactive labels. After hybridization and washing, the amount of radioactivity remaining
on the filter corresponds to the amount of target signal present in the sample. In FISH, the probe is labeled
with a fluorescent compound. The probe is hybridized with whole cells that are treated to make them more
permeable. Cells that hybridize to the probe fluoresce when viewed under a fluorescent microscope (DeLong
et al., 1989). Results for FISH are generally reported as the percent of the population that is positive to each
of the group-specific probes (Santo Domingo et al., 1998).
Target specific PCR-based methods
In the 1980's, several bacteria including Bacteroides (Fiksdal et al., 1985), Bifidobacterium (Resnick and
Levin, 1981) and Rhodococcus coprophilus (Mara and Oragui, 1981) were suggested as alternative host-spe-
cific fecal indicators to E. coli and coliforms. Although several of these bacteria showed promise, most of
them were difficult to cultivate and required lengthy incubations periods (up to 3 weeks for R. coprophilus}
before colonies could be enumerated, thus making them impractical for MST. With the advent of cultiva-
tion-independent methods, several of these bacteria have been and are being reevaluated for use with MST.
Enterococcus has also been suggested as an alternative host-specific indicator and has been well studied by
cultivation-dependent methods. Therefore, it is logical that cultivation-independent assays have also been
developed for Enterococcus.
In addition to the basic PCR method described earlier, variations have been developed that include the
simultaneous detection of several DNA targets (multiplex PCR), increasing the sensitivity of detection by
using two amplification steps (nested PCR) (Yang and Rothman, 2004) and quantifying the initial template
by quantitative PCR (QPCR) also known as real time PCR (RT-PCR). PCR assays and real-time PCR assays
have also been designed to detect and quantify common fecal bacteria in both humans (Bartosch et al., 2004;
Liu et al., 2003; Malinen et al., 2003; Wang et al., 1996; Wang et al., 1997) and cattle (Tajima, 2001). Ulti-
mately, some of these assays may prove useful for MST, but they need to be tested for host-specificity before
they can be applied for MST because not all fecal bacteria reflect host-specificity. Because each assay is
specific to one species or subset of microbes, multiple assays will be needed for each environmental sample
to potentially identify several sources of contamination (e.g., both human and cattle). In addition, the com-
bination of assays may strengthen the argument for the source of contamination. For instance, samples with
positive results for ruminant-specific Bacteroides, R. coprophilus and a Streptococcus bovis would indicate
cattle as a source of fecal contamination. Similarly samples with positive results for human-specific Bacte-
roides (or B. fragilis), Bif. adolescentis or B. dentium and Enterococcus would indicate human as a source of
fecal contamination.
Target specific PCR-based methods are probably the least expensive of the cultivation-independent methods.
PCR-based methods require minimal personnel training and can be performed within one day. Although
minimal training is needed to perform PCR, laboratories routinely performing target specific PCR must in-
corporate quality control measures to prevent cross-contamination of samples and false positives. Presence-
-------
absence PCR assays are less expensive than QPCR assays because they can be performed using standard
thermocyclers and inexpensive gel electrophoresis equipment. QPCR requires a thermocycler with a fluo-
rescent detector that costs at least $20,000 more than the standard thermocycler. However, presence-absence
PCR assays are more time consuming than QPCR assays requiring 2-3 hours for the PCR step and 1-3 hours
for the gel electrophoresis step. In QPCR, the complete PCR assay and analysis can be performed in less
than three hours. Some QPCR thermocyclers are designed to be used in the field and can provide data within
30 minutes. Individual QPCR assays are also slightly more expensive than presence-absence assays because
an additional fluorescently labeled probe must be added to the reaction.
i..,.,,.V ,w. ''. ••!,*<*.« ,.ri"'Vv v^1*'***.-;:-,^:, -,-'« . ,.,\f .''., JJi'-lA, •••.•ff..J>jLil.,Tv V
Application of target specific PCR to MST
Bacteroides
Currently, Bacteroides assays are the most widely used cultivation independent host-specific microbial
assays for MST. The use of Bacteroides as a potential indicator was proposed in the 1980's because the
-------
amount of Bacteroides that could be cultivated from human fecal samples was around 1,000 fold greater
than the levels of E. coli in human feces (Fiksdal, 1985). Additional research using cultivation independent
methods indicated that the Bacteorides-Porphyromonas-Prevotella group comprised 10-60% of the intestinal
population from many animals including humans (Franks et al., 1998; Harmsen et al., 2002), cattle (Wood et
al., 1998) and horse (Daly and Shirazi-Beechey, 2003). Kreader (1995) developed PCR primers and specific
hybridization probes to distinguish three Bacteroides species and demonstrated that the B.fragilis group (B.
distasonis and B. thetaiotaomicron) and B, vulgatus were at higher concentrations in human feces than in
farm animal species (cattle, swine, horses, goats and sheep, and poultry). Bernhard and Field (2000b) dem-
onstrated that Bacteroides isolated from ruminant and humans were host-specific and designed PCR primers
to distinguish human-specific and ruminant-specific Bacteroides. The human-specific Bacteroides presence/
absence PCR assay was used as part of a tiered approach to identify fecal contamination as human or non-
human (Boehm et al., 2002). Recently, QPCR assays have been developed for the detection of all Bacteroi-
des species (Dick and Field, 2004), human-specific Bacteroides (Seurinck et al., 2005) and bovine-specific
Bacteroides (Layton, unpublished). A QPCR assay for the detection of B.fragilis from human fecal samples
has been developed, but this assay has not been tested for host-specificity against fecal samples from non-hu-
man sources (Malinen et al., 2003).
Bifidobacterium
Bifidobacterium are a well-studied group of beneficial intestinal bacteria that have also been proposed as
fecal indicator species. Several Bifidobacterium species have been proposed as being human host-specific
including Bif. adolescentis (Matsuki et al., 2004; Bonjoch et al., 2004), Bif. dentium (Nebra et al., 2003;
Bonjoch et al., 2004) and Bif. longum (Matsuki et al., 2004). General PCR primers have been developed to
detect all Bifidobacterium (Kaufmann et al., 1997), and several PCR platforms have been designed to detect
individual species. These include PCR amplification with genus-specific Bifidobacterium primers followed
by hybridization with a species-specific probe for Bif. dentium (Nebra et al., 2003) and multiplex PCR for the
detection of Bif. adolescentis and Bif. dentium (Bonjoch et al., 2004). QPCR assays have been designed to
quantify Bif. longum (Malinen et al 2003; Matsuki et al 2004), Bif. adolescentis and Bif. dentium (Matsuki et
al., 2004). Two concerns with the use of Bifidobacterium as an indicator may be their short survivability in
water (50% reduction in 10 hours; Resnik and Levin, 1981), and its lower concentration in human feces than
Bacteroides (Sghir et al., 2000). The combination of these two factors may make it more difficult to detect in
the environment than Bacteroides. However, both Bif. dentium and Bif. adolescentis have been found in hu-
man sewage but not animal wastewaters (Bonjoch et al., 2004). In addition, the detection of human associ-
ated Bifidobacterium in water samples may indicate recent contamination events.
Streptococcus Lancefield Group D
The taxonomic group, Streptococcus Lancefield Group D contains both Streptococcus and Enterococ-
cus. These bacteria are routinely isolated from fecal samples and were originally named according to the
host from which they were isolated implying host specificity. It was generally believed that E. faecalis and
E.faecium (formerly S. faecalis and S.faeciurn) were associated with humans (Vancanneyt et al., 2002),
whereas S. bovis were specific to ruminants (Whitehead and Cotta, 2000). However, more recent literature
indicates that S. bovis isolates may not be completely host-specific, as S. bovis isolated from clinical samples
may cause approximately 24% of the streptococcal infections resulting in endocarditis, meningitis and septi-
cemia (Whitehead and Cotta, 2000). Although not applied to MST, primers have been designed to differenti-
ate S. bovis strains isolated from rumen and humans sources (Whitehead and Cotta, 2000). Several QPCR
assays also have been developed to detect Enterococcus species for application to drinking water and rec-
reational water regulations. Frahm and Obst (2003) published primers and a probe sequence that matches a
range of Enterococcus species, whereas Santo Domingo et al. (2003) published primers and a probe sequence
-------
specific for E.faecalis. For MST applications, additional research is needed to confirm host-specificity of
the S. bovis and Enterococcus groups (Vancanney et al., 2002).
Rhodococcus coprophilus
This target has not been used in any MST studies and is still being tested for its distribution among hosts.
Rhodococcus coprophilus was proposed as an indicator of fecal contamination from farm animals (Mara and
Oragui, 1981). This bacterium inhabits the digestive system of almost all grazing animals and is passed
to other animals grazing on the contaminated grass via the fecal-oral route. The design of a TaqMan-based
QPCR assay by Savill et al. (2001) allows continued testing of this bacterium as an indicator. Additional
information is needed on the prevalence of this bacterium in the U. S. and the amount of bacteria contained
in feces. It is likely that this bacterium persists longer in the environment than either Bacteroides or Bifido-
bacterium as it is aerobic and is passed between grazing animals.
Overview of target specific PCR methodology
Application of target specific PCR assays to water samples generally requires concentration of water samples
for two reasons. First, in a PCR reaction the amount of target-containing sample added is only a few micro-
liters (uL). Given the dispersed and dilute nature of bacteria in water, larger samples are needed for repre-
sentative sample. Second, assuming a worst-case scenario where the detection of one copy of DNA in a PCR
reaction is equal to one culturable bacterium, very high concentrations of bacteria (e.g., approx. 106) would
be needed in the environmental sample. This is a worst-case scenario because even for easily cultivated
bacteria such as E. coli only about 1 % of the population can be re-grown from an environmental sample,
thus the actual number of target bacteria in the sample is higher. For most situations a 100 ml water sample
is suitable for analysis. Water samples are often concentrated by filtering a 100 ml aliquot through a 0.45-
\im membrane filter. After filtration the DNA can be extracted from the filter (Boehm et al., 2003; Frahm
and Obst, 2003), the bacteria enriched in nonselective broth (Frahm and Obst, 2003) or selective agar (Santo
Domingo et al., 2003), or the bacteria can be eluted or washed off the filter and PCR performed without DNA
extraction (Fode-Vaughan et al., 2001).
Identification and quantification of specific viruses
Identification of enteric viruses with limited host ranges can help distinguish sources of fecal pollution in wa-
ter (Noble et al., 2003). Human-specific adenoviruses (Jiang et al., 2001; Pina et al.., 1998) and enteroviruses
(Griffin et al., 1999; Noble and Fuhrman, 2001) are candidate indicators for human fecal contamination. Bo-
vine enteroviruses (Ley et al., 2002) and bovine and porcine adenoviruses (de Motes et al., 2004) have been
proposed for detection of animal-source fecal contamination. Similarly, teschoviruses have been used as an
indicator of porcine fecal contamination (Jiminex-Clavero et al., 2003). Additional viral targets could also be
appropriate for MST depending on host specificity and pending development of molecular assays.
Application of host-specific viruses to MST
This method is in developmental stages and the number of studies applying this approach is still limited,
although recently assays targeting enteric viruses were used to detect human and bovine fecal contamination
in coastal waters (Fong et al., 2005). In addition, an MST methods comparison study found that detection of
human viruses has among the lowest false positive rates for tested methods (Griffith et al., 2003; Noble et
al., 2003). That is, human viruses were not identified in samples that lacked human-source contamination.
However, the study also demonstrated that this approach fails to always detect contamination from individual
humans; human viruses were detected in samples seeded with sewage but not in samples seeded with fecal
-------
Table 3.1 Summary of logistics of methods tested for MST
METHOD
Antibiotic
Resistance
Carbon
Utilization
Profiles
rep-PCR
RAPD
AFLP
PFGE
Ribotyping
Phage Sero-
or Geno-
typing
Gene Spe-
cific PCR
Host-specific
PCR
Host-specific
QPCR
Targets tested
• E. coli
• Fecal streptococci
• Enterococcus spp.
• E. coli
• Fecal streptococci
• Enterococcus spp.
• E. coli
• E. coli
• E. call
• E. coli
• Enterococcus
spp.
• E. coli
• Fecal strepto
cocci
• Enterococcus
spp.
• F+ coliphage
• E. coli toxin
genes
• Bacteroides
• Bifidobacteria
• Enterococcus
• Rhodococcus
• F+ coliphage
• Enterovirus
• Adenovirus
• Bacteroides
• Rhodococcus
• Bifidobacteria
Cultivation
Individual
Isolates
Individual
Isolates
Individual
Isolates
Individual
Isolates
Individual
Isolates
Individual
Isolates
Individual
Isolates
Individual
Isolates
Sample
Enrichment
None
None
Library
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
Major Equipment Needs}
None
None
Plate reader (optional)
Thermal cycler
Agarose gel electrophoresis
units
Gel documentation system
Fluorescence scanner for
HEFERP
Thermal cycler
Agarose gel electrophoresis
units
Gel documentation system
Thermal Cycler
Automated sequencer
Thermal cycler
Pulse Field Gel Electropho-
resis
Gel documentation system
Agarose gel electrophoresis
units
Gel blotting/Hybridization
oven
Gel documentation system
Hybridization oven
None if serotyping
Thermal cycler
Agarose gel electrophoresis
units
Thermal cycler
Agarose gel electrophoresis
units
Fluorescent Thermal Cycler
Major Costs
Antibiotics
96-well microplates
Microplates with sub-
strates (e.g., Biolog, Phene
Plate)
PCR reagents
PCR disposable
Gel electrophoresis
PCR reagents
PCR disposable
Gel electrophoresis
reagents
DNA extraction kit
AFLP kit ($5 per reaction)
Plug prep, reagents
Restriction enzymes
Gel electrophoresis
reagents
DNA purification reagents
Gel electrophoresis
reagents
Restriction enzymes
Hybridization/
detection solutions
Labeled gene probe
Hybridization/
detection solutions
Labeled gene probe or
Phage specific antigen
PCR reagents
PCR disposables
Filtration units
PCR reagents
PCR disposable
Filtration units
PCR reagents/label
PCR disposable
Time Required*
4-5 days
2-5 days
1 day
1 day
5 days
2-4 days
1-3 days
1 -3 days
2 days
6-8 hours
1 -3 hours
J All methods require standard microbiological equipment, such as, micropipettors ($200-300 each), microcentrifuge ($1-2K), in methods requir-
ing cultivation growth chambers (incubators) are needed
Major equipment costs are in the range of: Microcentrifuge ($ 1-2K), thermal cycler ($5K), thermal cycler with fluorescence detector for quan-
titative PCR ($25,000-$90,000), automated sequencer ($55K), submarine agarose gel unit with power supply ($ 1 -2 K), PFGE unit ($ 11-25K),
riboprinter ($175K), gel documentation system ($2-15K), statistical analysis software ($8-15K) needed for all library-dependent methods
Reagent costs: PCR ($2-$10/reaction including primers), filters to concentrate water samples ($4/ sample), all molecular method using gel elec-
trophoresis require agarose and buffer solutions
Time after enrichments or isolation performed, time for isolation dependent on target and method used for isolation and confirmation can vary
considerably. Also, time required for data analysis for library dependent methods are not included because it is highly variable and dependent on
available gel and data analysis software.
-------
Table 3.2 Comparison of advantages and disadvantages of source tracking methods*
METHOD
ADVANTAGES
DISADVANTAGES
Antibiotic Resistance
Rapid; easy to perform
Requires limited training
May be useful to differentiate host source
Require reference library
Requires cultivation of target organism
Libraries geographically specific
Libraries temporally specific
Variations in methods in different studies
Carbon Utilization Profiles
Rapid; easy to perform
Requires limited training
Require reference library
Requires cultivation of target organism
Libraries geographically specific
Libraries temporally specific
Variations in methods in different studies
Results often inconsistent
rep-PCR
Highly reproducible
Rapid; easy to perform
Requires limited training
May be useful to differentiate host source
Requires reference library
Requires cultivation of target organism
Libraries may be geographically specific
Libraries may be temporally specific
RAPD
Rapid; easy to perform
May be useful to differentiate host source
Requires reference library
Requires cultivation of target organism
Libraries may be geographically specific
Libraries may be temporally specific
Has not been used extensively for source
tracking
AFLP
Highly reproducible
May be useful to differentiate host source
Labor-intensive
Requires cultivation of target organism
Requires reference library
Requires specialized training of personnel
Libraries may be geographically specific
Libraries may be temporally specific
Variations in methods used in different studies
PFGE
Highly reproducible
May be useful to differentiate host source
Labor-intensive
Requires cultivation of target organism
Requires specialized training of personnel
Requires reference library
Libraries may be geographically specific
Libraries may be temporally specific
Ribotyping
Highly reproducible
Can be automated
May be useful to differentiate host source
Labor-intensive (unless automated system
used)
Requires cultivation of target organism
Requires reference library
Requires specialized training of personnel
Libraries may be geographically specific
Libraries may be temporally specific
F+ RNA coliphage
Distinguishes human from animals
Subtypes are stable characteristics
Easy to perform
Does not require a reference library
Requires cultivation of coliphages
Sub-types do not exhibit absolute host
specificity
Low in numbers in some environments
Gene specific PCR
Can be adapted to quantify gene copy
number
Virulence genes may be targeted;
providing direct evidence that potentially
harmful organisms present
Does not require reference library
Require enrichment of target organism
Sufficient quantity of target genes may not
be available requiring enrichment or large
quantity of sample
Requires training of personnel
Primers currently not available for all
relevant hosts
-------
Table 3.2 Comparison of advantages and disadvantages of source tracking methods (Cont.)
Host-specific PCR
Virus specific PCR
• Does not require cultivation of target
organism
• Rapid; easy to perform
• Does not require a reference library
• Host specific
• Easy to perform
• Does not require reference library
• Little is known about survival and distribution
in water systems
• Primers currently not available for all relevant
hosts
• Low in numbers, requires large sample size
• Not always present even when humans present
All methods require validation, personnel trained in basic microbiology, and in some cases basic molecular biology skills (e.g.,
PCR and agarose gel electrophoresis), and only those requiring specialized training are labeled.
material from individual humans. These results are consistent with the low carriage rate of viruses in the hu-
man population (Payment and Hunter, 2001).
Overview of host-specific viruses methodology
Molecular methods such as PCR allow rapid detection of viruses. These assays also tend to be more sensi-
tive than traditional cell culture, which can be technically difficult, time consuming, and inefficient (Schwab
et al., 1995). Concentration and purification of viral nucleic acids from environmental samples can be chal-
lenging, but advances are being made within research laboratories to address these issues. Quantitative PCR
assays have been developed for some viruses, which allow levels of viral contamination from various sources
to be quantified.
-------
Chapter 4. Data Collection and Analysis in Library-dependent Approaches
4.1 Introduction
Data collection and analysis are two critical components of microbial source tracking (MST) that require
careful attention in the planning stages of any study. Different approaches to MST produce different types
of numerical data and consequently require different considerations and strategies in sampling and analysis.
This Chapter highlights key issues in sampling design and data representation and discusses several statisti-
cal methods commonly used in various stages of MST. The discussion will be limited to library-dependent
methods as they pose the most technical challenge from a statistical point of view. Library-independent MST
approaches use host-specific markers to identify contaminant sources (See Sections 4.3-4.4.). The existence
of host-specific markers reduces the dependence on libraries, which are subject to geographic and temporal
variability. However, while the presence of a host-specific marker enables source identification with near
certainty, the absence of the same marker does not necessarily exclude any host from consideration. In addi-
tion, there are currently only a limited number of hosts for which such markers have been found. Therefore,
at present, library-independent approaches may need to be used in conjunction with a library-dependent
approach and the associated statistical analysis would require similar adaptation. For example, a simple
two-stage procedure could be used which first screens for host-specific markers and then resorts library-de-
pendent methods and statistical analyses if none are found.
4.2 Data collection
Effective MST requires that appropriate data are collected to meet the objectives of the study. For example,
an analysis that indicates cattle as the major source of fecal contamination to a stream on 70% of dates
sampled may not be particularly meaningful if the stream did not exceed regulatory criteria on those days.
Despite dominance by cattle contamination on most dates, humans could very well be the major source on
exceedance dates and, therefore, the logical target of remediation efforts. The sampling plan must be de-
signed around the objectives of the study.
Applications of source tracking could use various sampling schedules to accomplish their objectives. For ex-
ample, in applications to total maximum daily load (TMDL) water quality assessments, it might be essential
to evaluate contributions at all concentration levels across all seasons, while for application to beach closures
it might be more important to evaluate contributions when concentrations exceed regulatory limits during the
recreational season.
Most water bodies, whether streams, lakes, or aquifers, are not well mixed so a single sample does not rep-
resent the entire water body. In moving water, in particular, short-term variability must be considered be-
cause a single enriched particle can greatly skew the results from that sample. Furthermore, transient animal
populations mean that potential contributors change with season and hydrology creates different flowpaths
from those contributors with weather and season. A single sample should rarely, if ever, be interpreted as a
comprehensive indicator of pollution status across the entire water body, the entire year, or all flows.
-------
Some general principles to follow in sampling watersheds of various kinds include:
1. Composite samples are preferred to single dip samples in order to include more of the entire cross-
sectional area or volume of the sampled water body.
2. Taking several replicate samples or compositing samples over time helps to even out short-term vari
ability. (Hyer and Moyer, 2003)
3. Existence of transient animal populations implies that the known-source library may not be useful
in all seasons (Haack et al., 2003). This stresses the need that the known-source library should be
collected concurrently with water samples.
4. Different sources of fecal contamination could be expected in storm flow from base flow and this
should be taken into account in the sampling plan (Hartel, 2004). For instance, fecal pollution in base
flow is generally considered to be from ground water seep (including leaky sewer lines and leach
fields), direct deposition by wildlife, and various NPDES-permitted effluents. Fecal pollution in storm
flow, on the other hand, is transported with overland flow (including field-spread manure), stormwa-
ter discharges (including combined-sewer overflows), and other flooded areas (Tian, 2004).
4.3 Numerical representation of isolate profiles
As has been stated in previous Chapters, the majority of currently applied approaches for MST are library-
dependent. That is to say, they rely on a collection of isolate profiles (fingerprints, banding patterns or dis-
crete data) from each source category and the information contained in this library of isolate profiles forms
the basis for classifying indicator organisms of unknown origin by source category. Both genotypic and
phenotypic library-dependent approaches are currently employed for MST. Genotypic approaches character-
ize isolates based on DNA-based characteristics, often visualized as banding patterns of DNA fragments on
agarose or polyacrylamide gels, whereas phenotypic approaches characterize isolates based on their observ-
able physiology or growth characteristics on specific laboratory media, or via quantitative measurements of
traits like cell surface antigens or resistance to antibiotics. Compilations of genotypic and/or phenotypic
characteristics can be measured and used to define a reproducible profile or fingerprint for each isolate.
However, there are usually several ways to represent an isolate's characteristic profile as numerical data, and
decisions about data representation can have a significant impact on both sampling and analysis strategies
and outcomes.
Genotypic data, such as an isolates DNA fingerprint, can generally be represented numerically as (1) a "con-
tinuous" intensity curve where peaks represent the location (fragment size) of bands and the heights (and/or
areas) of the peaks are a quantitative measure of a bands intensity (See Figure 4.1), (2) a discrete listing of
band locations and intensities, defining presence and magnitude of a finite set of bands from a list of pos-
sible band fragment sizes or (3) a discrete profile listing of band locations, as binary (presence/absence) data,
defining only the presence of a finite set of bands from a list of possible band fragment sizes (See Figure
4.1). While the latter method has frequently been used for genotypic profiles having simple banding patterns
(those having a limited number of bands), more complex band patterns are often analyzed using data derived
from fragment location and band intensity. In cases where numerical differences in band intensity are theo-
retically meaningful, either of the quantitative representations is preferable to the binary representation. For
example, in the analysis of PFGE profiles, high band intensity may indicate the presence of multiple frag-
ments of similar length. Similarly, in PCR-generated DNA fingerprints, enhanced band intensity may be due
to PCR bias, or target copy number. In extreme cases, alternate numerical representation that takes these
factors into account might yield useful information.
-------
o
o
CN
tfi
C
T3
C
CO
CO
o
o
o
kO
o -
I
0
200 400 600 800
Migration Distance
1000 1200 1400
Figure 4.1 Curve representation of two genotypic fingerprints. Data is represented in terms of band intensity (gray
scale) and migration distance (in pixels).
-------
Thus, a key factor to consider when deciding whether to use quantitative values or presence/absence char-
acter tables is the interpretability of the quantities measured by each variable in the numerical profile. A
confounding influence, however, is that laboratory and image processing protocols can also affect numeri-
cal representation and, consequently, the analysis of genotypic profiles. In particular, it is essential that data
profiles for each isolate be carefully aligned such that all common bands are, in fact, positioned at the same
location. This task is more difficult than it may at first seem. The default settings of software packages com-
monly used for genotypic fingerprint analysis are designed to aid, but cannot by themselves ensure, accurate
alignment. Therefore, the incorporation of subjective judgments by an experienced analyst is required.
The numerical representation of phenotypic isolate profiles is more straightforward. Phenotypic profiles
sometimes consist of a series of quantitative measurements of phenotypic traits, such as in antibiotic resis-
tance analysis (ARA) in which growth in the presence of serial concentrations of antibiotics is tested. Phe-
notypic profiles can also consist of binary character tables, such as in multiple antibiotic resistance (MAR),
where resistance to only one concentration of each of several antibiotics is measured and carbon utilization
patterns (CUP), where a substrate may or may not support growth of a strain (Chapters 3 and 6). In this
case, the only numerical representation is a profile of binary variables which indicates growth or absence of
growth under the test conditions. Although discrete data such as that collected in ARA could also be depict-
ed as binary character tables, it is preferred to record quantitative values (maximum concentration at which
growth was not inhibited for each antibiotic) because data for each concentration of antibiotic tested are not
independent.
4.4 Library construction and validation
Sample size and library representativeness
Library-dependent MST studies require the creation of a known source database to which unknown field
isolates are compared. Library size and the representativeness of strains in a known-source the library are
two major considerations that need to be carefully assessed before embarking on any MST study. The same
considerations must be given to MST studies done using phenotypic or genotypic data, although the final
number of strains in a known source database may vary depending on the methods chosen. Moreover, based
on usually empirical information, one must carefully weigh decisions on whether to take a large number of
samples from a few animals or a lesser number of samples from a large number of animals. Generally speak-
ing, a library needs to be large enough to (1) capture the total genetic diversity present within the population
of indicator bacteria in a given host animal and (2) be of sufficient size so that environmental isolates can be
reliably typed to host origin. The ultimate size of the known source database library is also linked to the size
of the watershed under consideration and the number of potential sources in the watershed. For example, a
smaller library will be needed if a watershed is primarily inhabited by a limited number of potential animal
sources that occupy a limited geographic location.
The genetic diversity of indicator bacteria (most people use databases consisting of E. coli or enterococci) in
a given animal host is related to feeding habit, food sources, diet variation in a host animal group (Hartel et
al., 2003), fecal contamination from other animals, temporal and geographic variation of bacterial genotypes
within and between animal species (Gordon, 2001; Hartel et al., 2002; Scott et al., 2002; Jenkins et al., 2003)
and the number of strains in a single animal (McLellan et al., 2003). Accordingly, estimates of library sizes
are often difficult to make without empirical data. Generally speaking, most genotypic-based MST studies
that have been conducted to date have used relatively small host origin databases, containing between 35 and
approximately 500 isolates (Johnson et al., 2004). A small library size makes comparisons to populations of
E. coli and Enterococcus in the environment difficult, mostly due to the large number of unidentified strains
that result from such analyses. Recently, Johnson and coworkers (Johnson et al., 2004) reported that library
-------
size and representativeness have a major influence on the accuracy of MST studies. In contrast, many phe-
notypic-based MST studies, mostly done using antibiotic resistance patterns, have used larger known-source
libraries consisting of about 1,000 - 6,000 isolates (Johnson et al., 2004). In many cases, however, the strains
examined have been isolated from the same source animal or sample, introducing biases due to the presence
of multiple replications of the same bacterial genotype.
There are several methods available to measure the representativeness of known-source libraries. Many of
these methods, however, are empirical in nature. Rarefaction analysis has been considered a useful tool for
comparing species richness and diversity. This type of analysis has been used in MST studies and provides a
statistical method for estimating the number of genotypes that are expected to be present in a random sample
of individuals. The data requirements for the rarefaction analyses are not exacting and do not require abun-
dance information (Koellner et al., 2004). Rarefaction analysis estimates the rarity of a given genotype in a
population by calculating a series that approximates the number of genotypes present in randomly and suc-
cessively drawn subsets of the original database. This method allows for the generation of a rarefaction curve
that allows comparison of the observed richness (diversity) among randomized library entries by averag-
ing randomizations of the observed accumulation curve (Heck et al., 1975). If a library is "saturated" with
genotypes, the rarefaction curve will appear to have a horizontal asymptote, indicating that additional library
entries do not appreciably increase the number of new genotypes uncovered. In contrast, rarefaction curves
that appear linear indicate that the library is not saturated with respect to diversity of genotypes. As such, ad-
ditional library entries are needed to be useful to type unknown environmental isolates. As a consequence, it
has been suggested that a library size of tens of thousands ofE. coli isolates may be needed to capture all the
genetic diversity present in natural populations (Mansour Samadpour, personal communication).
The representativeness and fidelity of known-source libraries can also be ascertained by applying jackknife
analysis and reporting the average rate of correct classification (ARCC). This method of analysis is fre-
quently reported in MST studies (e.g., Harwood et al., 2000.). The ARCC simply calculates the number of
library isolates assigned to the correct source group when the library is queried using "hold-out" or Jackknife
analyses. To do this, each isolate is individually removed from the database. The degree of similarity of the
removed isolate to those remaining in each source group is determined, and then the average rate of correct
classification is determined. Library entries that are incorrect or small libraries containing insufficient entries
to capture all the genetic diversity will have lower ARCC values.
4.5 Measuring spatial and temporal variability
There are several statistical techniques available to measure and compare patterns of spatial and temporal
variability. Among these are exploratory graphical techniques such as multi-dimensional scaling (MDS) or
principal components analysis (PCA) and confirmatory analyses performed using statistical techniques such
as multivariate analysis of variance (MANOVA). The goal of an exploratory analysis is to identify patterns
of variation in the data relevant to assumptions and hypotheses (Chapter 3). The goal of a confirmatory
analysis is to test the validity of specific assumptions and hypotheses which may have been formulated based
on observations made during an exploratory analysis.
Multidimensional Scaling
MDS (Torgerson, 1958) is a technique for representing a dissimilarity (or distance) matrix in relatively few
dimensions. To illustrate the usefulness of MDS, consider a data set like the distance tables at the back of a
road atlas, giving driving distances between major cities. The MDS algorithm could accurately reconstruct a
map of the United States from this matrix of distances. The distance between pairs of cities in a MDS map
of the United States would be roughly proportional to the corresponding geographic distances. (However,
-------
the map itself might be rotated or inverted.) This is illustrated below in the Figure 2 which is an MDS map
of the ten USEPA Regional Offices, which was constructed by applying MDS to the table of geographic
distances in Table 4.1. If the user did not know the geographical relation between the cities listed in the road
atlas, the MDS map would be helpful for identifying geographic relationships.
1000--
cvi
0--
-tXX)--
*SEA
SF
*DEN *KC
* DAL
NY **BOS
Of! * PHI
*ATL
-2000
H 1
•1000 0
Dimension 1
1000
2000
Figure 4.2 Multidimensional scaling map of the ten USEPA regional offices constructed from the table of geographic
distances in Table 4.1
Table 4.1 Geographic distance between each pair of USEPA regional offices.
Boston
New York
Philadelphia
Atlanta
Chicago
Dallas
Kansas City
Denver
San Francisco
Seattle
Boston
0
200
300
1100
1000
1750
1440
2000
3130
3020
New York
0
110
850
810
1560
1230
1790
2930
2840
Philadelphia
0
750
790
1440
1170
1740
2900
2820
Atlanta
0
710
820
820
1430
2480
2630
Chicago
0
920
540
1020
2170
2050
Dallas
0
510
780
1750
2130
Kansas City
0
610
1860
1860
Denver
0
1260
1340
San Francisco
0
810
Seattle
0
-------
In the context of MST, MDS plots are based on a matrix of numerical inter-isolate dissimilarity measures
(instead of driving distances). Patterns of inter-isolate variation can be represented in a two or three-dimen-
sional plot in which distances between points are roughly proportional the dissimilarity between the isolates
they represent. As with PCA, this technique allows the multi-dimensional isolate profile data to be plotted in
two or three dimensions and aids in the identification of major sources of variation.
among binary ^di|^l|i^ii^6.^4^^^
,_„ y»/v*«\ s\.ji.iJ$if4fuf, »S^( .. ^i> U,* .4 V
-------
Table 4.2 Common similarity measures for binary profiles
Coefficient
Jaccard(1901)
Sokal and Mi-
chener(1958)
Dice (1945)
Ochiai (1957)
Kulczynski
(1928)
Russell and Rao
(1940)
Mathematical
expression
a/(a+b+c)
(a+d)/p
2a/(2a+b+c)
a/
sqrt((a+b)(a+c))
a(2a+b+c)/
[2(a+b)(a+c)]
a/p
Value when
a=0 (no
matches)
0
d/p
0
0
0
0
Value when
b=c=0 (no
mismatches)
1
1
1
1
1
a/(a+d)
1 - value is a
distance metric
Yes
Yes
No
No
No
Yes
Suitability for
MDS
High
Moderate
Moderate
Moderate
Low
Low
Alternate
names
Coefficient of
community
Simple
matching
Sorensen (1948)
Coefficient of
closeness
Jeffrey's x
Principal components analysis
PC A (Hotelling, 1933) is a statistical technique for dimension reduction and identification of dependence
patterns among variables. PCAuses the interdependence between the original set of variables, as measured
by correlation or covariance, to reduce the data set to a smaller set of variables called principal components.
The principal components reproduce patterns present in the full set of variables and are easier to visualize.
For example, two or three principal components can sometimes be used to summarize data for 50 or more of
descriptive variables such as bands in a fingerprint. The major assumption of PCA is that the dependence
between variables is fully described in terms of pairwise covariances and that this covariance structure is
similar for the entire population.
While PCA and MDS are both useful techniques for representing multivariate data in relatively few dimen-
sions, it should be emphasized that both the underlying assumptions and objectives of these two methods
are quite different. In particular, PCA is based on measurements and assumptions about variable interde-
pendence and MDS is based on measurements and assumptions about inter-isolate similarity (See Table 3).
As a matter of practice, however, one often observes strong similarities between MDS and PCA plots (See
Figure 4.3). If there is a natural similarity among profiles, whether based on host of origin, time frame, or
geographic location, it may be detectable by either approach. Further discussion of both MDS and its con-
nections to PCA can be found in Cox and Cox (2001).
More about Principal Components Analysis
-------
; exercised when drawmg conclusions
application of PC A and its underlying theory can be found in Jolliffe (2002),
Table 4.3 Techniques for identifying patterns of spatio-temporal variability in the isolate profiles
Method
Principal components
analysis (PCA)
Multidimensional scal-
ing (MDS)
Multivariate analysis of
variance (MANOVA)
Non-parametric
MANOVA
Reference
Retelling (1933)
Turgorsen(1952)
Wilks (1932)
Anderson (2002)
Exploratory or
Confirmatory
Exploratory
Exploratory
Confirmatory
Confirmatory
Objective
Represent the variation in a
large number of variables by
a small number of principal
components.
Represent an interobject dis-
tance matrix in relatively few
dimensions.
Test for statistically signifi-
cant differences between the
means of specified groups of
isolate profiles.
Test for statistically signifi-
cant differences between the
means of specified groups of
isolate profiles.
Relevant
Assumptions
Covariance is an appropriate
measure of variable interde-
pendence.
Dependence between
variables is similar for all
isolates.
Selected inter-isolate distance
metric is appropriate.
Multivariate normality
Dependence between
variables is similar for all
observations
Selected inter-isolate distance
metric is appropriate.
£ o
•..
**
-500
500
PCA axis 1
£
Q
.** .
• •» •
« * •».
«
» •
V •
I 1 I I ! I I
-03 -02 -01 00 01 02 03 04
MDS axis 1
Figure 4.3 PCAand MDS plots of a small library of isolates where different colors indicate different source categories.
-------
Multivariate analysis of variance
If patterns of spatial or temporal dependence are suspected or observed in an exploratory analysis such as
PCA or MDS, a multivariate analysis of variance (MANOVA) (Wilks, 1932) can be used to test for signifi-
cant spatial or temporal effects. However, in order for valid conclusions to be drawn from a typical MANO-
VA, data must satisfy the assumption of multivariate normality. This assumption is rarely met by MST data
and therefore resampling-based methods for MANOVA (Anderson, 2003), which do not assume, multivariate
normality are preferable. Unfortunately, this methodology is not currently available in standard software
packages. Therefore, it should be emphasized that results of standard MANOVA must be interpreted with
caution.
Isolate identification
Once a library of isolate profiles from each potential source has been collected, a rule for identifying the
most likely source of isolates of unknown origin must be constructed. Statistical methods to accomplish this
are referred to as discriminants or classification rules. This section describes the general process of con-
structing and evaluating classification rules in the context of MST and discusses the assumptions of several
classification rules commonly used for MST. Certain types of classification rules and their corresponding
assumptions are more appropriate for different types of MST data, but the same general process should be
followed in the construction and assessment of classification rules regardless of the type of data (Hastie et al.,
2002).
1. Declone isolates from each feces sample by deleting identical patterns within each single feces
sample in the database. These are essentially duplicate observations and it would be inappropriate to
have replicate observations in the training, validation, and test sets (below).
2. Randomly divide data into training isolates (-50%), validation isolates (-25%) and test isolates
(-25%).
3. Use the training isolates to construct various classification rules.
4. Estimate the accuracy of each rule by attempting to classify the validation isolates.
5. Select the most accurate rules and refine them. Refinement techniques include variable selection and
the adjustment of tuning parameters.
6. Once a (single) best rule has been selected, use the test isolates to estimate generalizability of the
rule. This step is important to predict how well the classification rule will work in real-world
application.
4.6 Techniques for classification and discriminant analysis
There are several different techniques for classification of isolates to source categories based on the known-
source library. This section attempts to give some details of a few commonly used procedures for discrimi-
nant analysis and identify situations where each is appropriate.
Linear and Quadratic Discriminant Analysis
As with most statistical techniques, at the foundation of the most commonly-used and well-known classifica-
tion rules is the assumption that the training data (known-source library) is derived from random samples of
a population for which the variation between samples is well-described by a normal distribution. In particu-
lar, linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA) assume that the data from
each population follow a multivariate normal distribution. Implicit in this assumption is the notion that each
-------
individual variable follows a (univariate) normal distribution and that the dependence structure among the
variables is fully characterized by a matrix of pairwise covariances.
In the case of binary profiles, the concept of a mean, or "consensus," fingerprint surrounded by a cloud of
variants, distributed at similarity distances that follow a normal distribution, is not meaningful. Therefore,
the assumption of multivariate normality makes it difficult to justify the use of LDA or QDA for classifica-
tion of binary profiles resulting from genotypic fingerprints because these presence/absence data do not fol-
low a normal distribution.
LDA makes the additional assumption that the matrix of pairwise covariances is the same for samples from
each population. However, in the context of MST, this may not always be true (e.g., two bands might be
positively correlated for one source category and negatively correlated for another source category). Finally,
classification rules based on the assumption of multivariate normality use estimates of covariance matrices
that tend to be poor unless sample sizes are very large. Thus classification rules based on LDA and QDA can
often perform poorly when sample sizes are not very large. In conclusion, LDA should be used with caution
for MST, but QDA seems to be a somewhat reasonable approach for data resulting from phenotypic profiles
when sample sizes are large.
Nearest-neighbor rules
In addition to LDA and QDA there are several other types of classification rules. When the assumptions
required by LDA and QDA are inappropriate, nearest-neighbor rules are the most common alternative. There
are several varieties of nearest neighbor rules, but they share the general characteristic of classifying objects
based on the group membership of the most similar objects of known origin. These rules do not assume any
explicit form for the data distribution, such as multivariate normality, but in order to provide reasonable clas-
sifications, similar objects, as measured by some distance or similarity coefficient, must come from the same
population.
The three most common types of nearest neighbor rules in the MST literature are the (1) maximum similar-
ity, (2) average similarity and (3) k-nearest-neighbor rules. Maximum similarity classification simply assigns
isolates of unknown origin to the source of the most similar isolate in the library. Average similarity mea-
sures the similarity between the isolate of unknown origin and all isolates of known origin and then assigns
the unknown isolate to the source with which the unknown isolate profile has the highest average similarity.
The ^-nearest neighbor rules (Fix and Hodges, 1951) are somewhat of a compromise between these two
methods. For some specified value of &, the A; most similar objects are identified and the isolate of unknown
origin is assigned to the source with the largest representation among the k nearest neighbors. Surprisingly,
little research has been conducted regarding the choice of the value of k. However, for the simple case of
two multivariate normal populations of comparable group sizes, Enas and Choi (1986) recommend select-
ing k to be approximately between n2/8 and n3/8 depending on whether there are small or large differences
between the group covariance matrices. So, even for sample sizes of n= 1000 the recommended value of k is
somewhere between 5 and 13. Thus a large number of neighbors are not advisable. Further information on
the theory and implementation of nearest neighbor rules can be found in Dasarathy (1991).
Epidemiological matching
So called "epidemiological matching" is another approach that has been used for isolate identification.
(Note: Statistically, this can be viewed as a generalization of a maximum-similarity classification rule.) This
-------
practice involves clustering isolate profiles into subtypes and assigning an isolate of unknown origin to a
source category only if it is similar enough to all the isolates of a particular subtype, which themselves are all
associated with the same source. Definition of subtypes is accomplished via complete linkage hierarchical
cluster analysis, which establishes a minimum similarity for all isolate profiles within a subtype. Determin-
ing the value of this minimum similarity value depends on both the quality of the data and on the similarity
being used. For example, Figure 4.4 illustrates the fact that the simple matching coefficient is always larger
than the Russell-Rao coefficient, the Jaccard coefficient is always larger than the Russell-Rao coefficient and
the Dice coefficient is always larger than the Jaccard coefficient. Therefore, it is difficult to establish guide-
lines for establishing subtypes beyond stating that the relative magnitudes of similarity measures should be
considered.
Russell - Rao = — < = Simple matching
P P
a
Jaccard =
a
a + b + c
s.ao - — ^ jaccara
p (a + b + c) + d a + b + c
a
a + b + c_
a(b + c)
(2a + b + c\ a + b + c)
2a
2a + b + c
= Dice
Figure 4.4 Relationships between similarity measures for binary profiles.
Table 4.4 Summary of common classification rules
Rule
Linear Discriminant Analysis
(LDA)
Quadratic Discriminant
analysis (QDA)
Average similarity
Maximum similarity
(1 -nearest neighbor)
k-nearest neighbor
Epidemiological matching
Assumptions
Multivariate normality
Common covariance structure
for each group
Multivariate normality
Appropriate similarity
Appropriate similarity
Appropriate similarity
Appropriate similarity
Suitability classification of
genotypic profiles
Low
Low
Low
High
High
High
Suitability classification of
phenotypic profiles
Low
Moderate
Low
Moderate
High
Moderate
4.7 Practical issues and Chapter summary
Total cost of misclassification
In a formal decision-theoretic framework the cost of a classification error is factored into the evaluation of a
classification rule. For example, in the context of MST, it might be more costly to identify a poultry farm as
the source of contamination, when in fact wild geese are the true source of contamination, than it would be
to make an error in the reverse direction. Additional examples include unnecessary human sewer upgrades
-------
using public money, BMP for livestock waste management at a portion of the farmer's personal income, or
wildlife management plans at a small amount of public money. Incorrect identification of contamination
sources is also likely to have political costs in addition to monetary costs. However, most software packages
do not let users specify the costs of each type of misclassification error. An alternative protection against
costly errors is requiring a threshold of evidence before any classification can be made, since it is often
preferable to make no attempt at classification rather than classify incorrectly. An example of an analysis of
MST data which includes the use of threshholding is found in Ritter et al. (2003).
Software
The techniques and tools for data management and analysis discussed in this Chapter require software for
implementation. There are several software packages available for image processing, library management
and data analysis. These packages vary in cost, capabilities and ease of user interface. Table 4.5 attempts to
give some indication of the relative strengths and weaknesses of some commonly used software packages.
Table 4.5 Comparison of software packages commonly used for analyses associated with microbial
source tracking
Software
Bionumerics
SAS
R
ImageQuant
Company
Applied Maths, Belgium
SAS Institute, Gary NC
CRAM, www.r-project.org
Capabilities
Image analysis, data management
and statistical analysis
Statistical analysis
Statistical analysis
Image analysis
Ease of use
High
Moderate
Low
Flexiility
Low
Moderate
High
Moderate
Cost
High
High
Free for academic use
Low
Summary
In conclusion, we reemphasize that there are several critical decisions to be made regarding data collection
and analysis in any MST study and reiterate the most important ideas below.
1. The sampling plan must be designed around the objectives of the study.
2. There are usually several ways to represent an isolate's characteristic profile as numerical data, and
decisions about data representation can have a significant impact on both sampling and analysis
strategies and outcomes.
3. The genetic diversity of indicator bacteria in a given animal host is influenced by several factors.
Accordingly, estimates of library sizes are often difficult to make without empirical data.
a. Generally speaking, most genotypic-based MST studies that have been done to date have used
relatively small host origin databases, containing between 35 and about 500
b. In contrast, many phenotypic-based MST studies, mostly done using antibiotic resistance
patterns, have used known-source libraries consisting of about 1,000 - 6,000 isolates
c. In some of the more extreme cases a significantly large library (i.e., fingerprints for 20,000 to
40,000 E. coli isolates) may be needed to capture all the genetic diversity present in natural
populations
-------
4. Certain types of classification rules and their corresponding assumptions are more appropriate for
different types of MST data, but the same general process should be followed in the construction and
assessment of classification rules regardless of the type of data
a. LDA should be used with caution for MST, but QDA seems to be a somewhat reasonable
approach for data resulting from phenotypic profiles when sample sizes are large.
b. When QDA is inappropriate, nearest-neighbor rules are the most common alternative.
-------
Chapter 5. Methods Performance
5.1 Introduction
The goal of Microbial Source Tracking (MST) is to associate a microorganism from a polluted site with an
human or animal source to infer the origin of fecal pollution. This information is vital to managers, stake-
holders, and other interest groups that play a role in contracting MST studies, performing water quality moni-
toring, risk assessment, and protection and restoration of U.S. surface waters. Decision makers require high
quality data. Quality control strategies measure confidence in data and help ensure proper use of methods.
As a result, researchers have developed quality measures to assess the performance of each MST method. A
comparison of quality measures revealed a core group of performance criteria that all MST methods share in
common. This Chapter will organize and define MST universal quality measures and provide an overview of
method-specific performance criteria that can be used to evaluate the quality of data and overall performance
of each MST approach.
5.2 Universal Quality Measures
Although MST researchers use a wide array of techniques to identify fecal pollution in surface waters, all
methodologies should adhere to a strict set of quality measures. These measures are organized into five
quality control issues including specificity, precision, control samples, quality assurance documentation, and
minimum number of controls. Recommendations for each quality control issue are discussed below.
Specificity. Specificity refers to the ability of a particular MST method to discriminate between different
animal fecal sources. The specificity of a method can be described as the proportion of samples that are
negative [test negatives (TN) + false positives (FP)] that test negative [test negatives (TN)]. Specificity is
mathematically expressed as:
TN x 100%
TN + FP
A specificity percentage should be reported for each animal fecal source included in a MST study. Although
there is currently no consensus, specificity values below 80% percent reflect questionable discriminatory
power. Managers should use data with caution and may need to consider data from an alternative MST ap-
proach. Specificity control standards should be prepared at concentrations easily detected by the respective
MST method and should consist of a pool of fecal samples acquired from animal sources in the same geo-
graphic context as water samples. The minimum number of individual animal fecal samples will be depen-
dent on the complexity of the watershed system (see Chapter 4). Currently, there is no agreement on how
to calculate this number. A conservative estimate might be a minimum often individuals per animal source.
Because specificity control standards are generated for each watershed, specificity must be established for
each geographic location tested. It is also ideal to perform specificity controls before applying a particular
MST method to test samples. Many researchers will collect test samples during specificity testing and ar-
chive samples until specificity is confirmed in the watershed of interest.
-------
Precision. Precision or reproducibility is important for all MST applications and is measured through the use
of replicates. Replicates are repetitions of an assay or part of an assay and fall into two categories: identical
replicates and experimental replicates. Identical replicates are assays performed simultaneously using the
same method preparations and same reagents (i.e. antibiotics, media, PCR reagents, etc.). Identical replicates
serve two functions. They can preserve data. If one replicate fails, the other can potentially still provide
data. They can also be used to monitor variability or low precision in a test sample batch. A sample batch is
a set of test samples prepared and processed together through all steps of the MST method. Approximately
10% of all samples tested should be replicated. Replicate sample results should be in agreement. Experi-
mental replicates are assays that share the same reagents, while the sample preparations come from similar,
but not identical samples. They provide crucial information about the overall precision of the method. For
example, if a researcher wishes to test the reliability of identifying human fecal pollution in a watershed, it is
inappropriate to assay just one water sample. A number of samples must be analyzed to determine whether
there is any variation in method response. If variability is prevalent, researchers can evaluate analyst perfor-
mance, quality of reagents, proper equipment function, or sample matrix characteristics to increase precision.
Control Samples. Control samples are quality measures that monitor the proper performance of MST meth-
ods and screen for the presence or absence of extraneous microorganisms or nucleic acids introduced into
a MST experiment. All MST methods should incorporate method positive controls and negative controls.
Method positive controls verify whether a MST process is performing adequately (Figure 5.1). These con-
trols should be obtained from a known source and should always yield a predefined result when the MST
method is conducted correctly. For example, ARA laboratories commonly use enterococci and E. coll strains
with known multiple antibiotic-resistant patterns as method positive controls. If the expected antibiotic-re-
sistance pattern is not observed, the researchers should reject all data with the same sample materials and
request immediate resampling. For culture-independent methods, control template should be tested at a
concentration ten times above the limit of detection. Method positive controls should be performed for each
batch of test samples.
Negative controls are used to monitor for the introduction of extraneous materials into an experiment. They
are divided into two categories including field blanks and method blanks. Field blanks monitor for the intro-
Data:
Sou
Cc
P
Equation:
rce Test Negatives False P
>w 900 1(
g 850 1!
Test Negatives
Test Negatives + False Positive
900 + 850
„ Ann
900 + 850 + 100 + 150
Conclusion: Human specific pattern detectable percentage is 8
ositives
)0
50
_ y 100
js
= 87.5%
7.5%
Figure 5.1 Library-dependent specificity calculation for human detection.
-------
duction of extraneous material into MST experiments during field sample handling, transport, and storage.
In the field, sterile water should be transferred to a sample collection tube and processed as a test sample. A
positive result indicates the presence of contamination most likely due to poor aseptic technique in the field,
contact with other samples, or damaged storage containers. The method blank is designed to screen for con-
tamination throughout the entire MST process. This control determines whether glassware, filters, handling
procedures, media, reagents, or lab environment introduce extraneous material into samples. In the labora-
tory, the control is processed in the same manner as a test sample except that sterile water is substituted for
an environmental sample. At least one method blank should be performed for each sample batch.
Quality Assurance Documentation. The results of all quality measuring data and method validation should
be thoroughly documented, published in future studies, and easily accessible to management personnel. In
addition to laboratory standard record keeping procedures including equipment calibration and maintenance
schedules, reagent catalogs, quality measure data, sample processing notes, and routine documentation back-
up, MST researchers should pay careful attention to sample acquisition documentation. Information describ-
ing animal fecal sampling location and date should be consistently documented for each MST experiment.
Access to this information will be imperative for future research concerning library and genotypic target geo-
graphical and temporal stability. It may also be useful to record the diet of individual animals used for fecal
sampling. All documentation should be reviewed by a laboratory supervisor for accuracy and completeness.
Quality assurance documentation reviews ensure that all method quality requirements were met, and that any
deficiencies are properly noted in the final report.
Minimum Number of Controls. The disadvantages of performing quality measure controls are that they
result in additional cost, can occupy space in the laboratory, and can consume more sample. However, these
controls are the only way to validate a MST method and ensure that data from test samples are genuine.
Each researcher should weigh these disadvantages against the need for precise and accurate information
when deciding how many controls to run in each experiment. Table 5.1 lists each quality measure control
type, summarizes their importance, and lists recommended frequencies for a typical MST study.
5.3 Method-Specific Performance Criteria
5.5.7 Library-Dependent Methods.
Library-dependent methods compare traits from cultivated fecal isolates collected from water samples with
a library of cultivated isolates from known fecal sources. The known source library acts as a predictive tool
Table 5.1 Summary of Quality Measure Controls
Description
Specificity
Identical Replicates
Experimental Replicates
Method Positive Control
Field Blank
Method Blank
Purpose
Verify ability to discriminate between animal
sources
Monitor variability between test
replicates in sample batch
Monitor method variability between sample
batches
Verify method process performing
correctly
Verify that not contamination introduced during
sample acquisition
Verify that no contamination introduced during
entire method process
Frequency
Establish for each MST geographic
location tested
10% of the number of field samples tested per
batch
At least 10% of field samples tested per batch
One control per sample batch
5% of the number of field samples
collected
At on control per sample batch
-------
to determine the source of fecal pollution. Library-based methods include carbon utilization profiles, antibi-
otic resistance assays (ARA), ribotyping, pulsed-field gel electrophoresis (PFGE), amplified fragment length
polymorphism (AFLP), and repetitive PCR (rep-PCR). The utility of a particular method is directly related
to the ability of a library to accurately represent and characterize fecal sources present in a watershed. Un-
suitable libraries lead to inaccurate information and poor management decisions. Researchers evaluate the
quality of libraries based on composition, size, continuity, sensitivity, and minimal detectable percentage.
Library Composition. The first step in library construction is to collect fecal samples from host species, then
isolate bacteria from a number of different individuals. Most source tracking libraries are composed of either
enterococci or E. coli isolates. Culture methods designed to isolate these microorganisms can sometimes
allow the growth of other microorganism species. As a result researchers should perform additional tests to
confirm the identity of the target organisms (enterococci or E. coli} in a library. Ideally, the library should
consist of 100% of the target indicator organism. The library should be comprised of isolates collected from
source animals impacting the local watershed. Potential fecal pollution sources can be identified by perform-
ing a sanitary survey of the watershed.
Library Size. The ideal library should contain enough isolates from each host species to characterize the
dominant traits of an indicator organism population. Some researchers suggest that small libraries misrep-
resent population diversity of indicator organisms in surface waters. However, it remains undefined what
constitutes the optimal library size partly because few studies to date have rigorously evaluated this problem.
Wiggins and colleagues (2003) conclude that a library should be as large as it needs to be representative.
Library representativeness is a measure of how well a library classifies the patterns found in a target micro-
organism from each of the host species found in a watershed. Representativeness is estimated by comparing
the ARCC from a resubstitution analysis with the ARCC from a cross-validation analysis (see Chapter 4,
Data Collection and Analysis for review). If the difference in ARCC values is less than 5%, then the library
is representative. Researchers currently construct libraries based on sample accessibility, cost, and practical
experience. As a general guideline, libraries should contain at least 1,000 isolates per host species of interest.
Library Continuity. The ideal library should be able to classify fecal isolates from numerous geographical
areas and should be representative over time. However, factors such as season, diet, and horizontal gene
transfer (movement of DNA from one bacterial cell to another) can create library discontinuity (Bryant,
1959; Hungate, 1966; Ogimoto and Imai, 1981; Stewart and Bryant, 1988; Harmseri et al., 2000). Initial
studies indicate that geographic variability can be high and that libraries should be constructed from local
samples only (Hartel et al., 2002; Wiggins et al., 2003). The longest a library has been shown to be stable is
12 months (Wiggens et al., 2003). Thus, library continuity should be re-evaluated at least once a year until
additional studies indicate otherwise.
Library Sensitivity. Library sensitivity measures the detectable percentage of isolated target microorganisms
exhibiting a host-specific pattern (Figure 5.2). The sensitivity of a method is described as the proportion of
samples that are positive [test positives (TP) + false negatives (FN)] that test positive [test positives (TP)].
Sensitivity is mathematically expressed as:
TP x 100
TP + FN
A sensitivity value is also referred to as the rate of correct classification (RCC) and should be reported for
each animal fecal source included in a MST study. In addition, researchers commonly report an average rate
of correct classification value (ARCC) or mean of all RCC values. Sensitivity values should be determined
from a set of characterized standards (from known fecal sources).
-------
Sensitivity
Data:
Test Positives
Test Positives + False Negatives
x100
Test positives = 850
False negatives = 150
850
850 + 150
x100 =85%
Conclusion: Human rate of correct classification is 85%
Figure 5.2 Sensitivity or RCC calculation for human detection in a 100 ml control sample.
Minimal Detectable Percentage (MDP). The Minimal Detectable Percentage is a measure of the lower limit
for considering that a source is present in a sample (Whitlock et al., 2002; Harwood et al., 2003; Wiggins
et al., 2003). Its value is based on the average frequency of misclassification of the known sources in the
library. The MDP can be used to estimate the likelihood that an isolate that is not from a given source will
be classified into that source, and therefore provide the basis for a significance cut-off when predicting the
sources of isolates in water samples (Harwood et al., 2003). Several methods of determining the MDP have
been proposed, and although there is not yet consensus on the best method, all MST studies should present a
value of the MDP and the method that was used to determine it.
5.3.2 Library-Independent Methods
Library-independent methods rely on genotypic traits to identify sources of fecal pollution. These methods
do not require isolate cultivation. Library-independent methods include T-RFLP community analysis and the
detection of host-specific microbial DNA sequences. Host-specific strategies target 16S rDNA from Bacte-
roides, toxin and adhesion DNA sequences, and numerous phage loci. These methods rely on PCR technol-
ogy and can detect small quantities of nucleic acids in a few hours. However, an increased limit of detection
elevates the risk of amplifying extraneous nucleic acid templates. Inhibitory substances can co-extract with
nucleic acids during sample purification and concentration (Wilson, 1997). In some cases, PCR inhibition
may be the cause of false-negative reactions and can dramatically decrease the limit of detection.
Limit of Detection. The limit of detection is the minimum concentration or copy number of a control DNA
target that routinely yields a PCR product (Figure 5.3). Detection limits are measured by adding a range of
control DNA template concentrations (i.e. 1, 10, 102, 103, and 104 copies) to PCR test reactions. PCR control
DNA templates can be any of the following: 1) purified total nucleic acid extract from a microorganism con-
taining the sequence of interest, 2) the whole microorganism, which can be used when the DNA template is
released by heating before or during amplification, 3) a specific DNA template containing the entire sequence
to be amplified, including primer binding sites, or 4) a cloned DNA fragment containing a modified form of
the DNA target (see Inhibition of Nucleic Acid Amplification section). After limit of detection is established
for a MST method, researchers should include a control containing the minimum detectable quantity for each
sample batch tested. This control will ensure that each PCR assay is performing at an optimal level.
-------
Experi
Data:
Concli
ment: Test for detection of 1, 10, 102, 103, and 104 target copies.
Each test reaction tested in triplicate.
Copy#
1
10
102
103
104
Trial 1
-
+
+
+
+
Trial 2
-
-
+
+
+
Trial 3
-
-
+
+
+
ision: Human limit of detection is 1 02 copies.
Assay will occasionally detect 10 copies.
Figure 5.3 Measuring limit of detection for human host-specific PCR assay.
Confirmation of PCR data. Most host-specific PCR methods measure the presence or absence of a target
DNA sequence in an environmental sample. For example, a water test sample is collected and concentrated
on a filter. DNA from microorganisms adhering to the filter surface are extracted, purified, and amplified us-
ing primers that target a specific sequence or group of sequences. If the target DNA is present, the researcher
will observe a PCR product on an agarose gel. Two strategies can be used to validate the authenticity of the
resulting PCR product. First, the researcher should report the PCR product size (base pairs). For example,
the human host-specific 16S rDNA PCR primer set HF134 and 708R (Bernard and Field, 2000) should
yield a PCR product of approximately 574 base pairs. Second, the resultant PCR product can be cloned and
sequenced. Sequencing is more time consuming and expensive, but it is the only way to definitively prove
detection of target DNA. Sequencing will also help build a database of sequences that can be used to evalu-
ate genetic variation of target DNA over time and in different geographic locales.
Extraneous Nucleic Acids. PCR methods that exhibit a low specificity may be contaminated with extraneous
nucleic acids found in the laboratory environment or reagents. Nucleic acids from equipment, other samples,
and previously synthesized amplicons can contaminate PCR reactions. Extraneous nucleic acids from these
sources can be eliminated with physical barriers. Sample preparation, nucleic acid extractions, PCR cocktail
assembly and amplifications, and post-PCR manipulations should occur in separate work areas. If laboratory
space is limited, separation of pre-PCR (sample filtration, nucleic acid extraction, and PCR cocktail as-
sembly) from post-PCR (i.e. gel visualization, molecular cloning, etc.) manipulations is most critical. Each
area should contain dedicated equipment and be cleaned with 0.6% sodium hypochlorite (NaOCl) after each
use. In addition to physical barriers, a unidirectional workflow between areas (i.e. sample preparation -^
extractions -> PCR cocktail assembly and amplification -> post-PCR analyses) should be used to reduce the
potential for contamination.
PCR reactions may also amplify nucleic acids present in extraction and PCR reagents, which cannot be
eliminated with physical barriers. For example, several studies have documented the presence of eubacte-
rial DNA in Taq DNA polymerase preparations (Hughes et al., 1994; Schmidt et al.. 1991; Rand and Houck,
1990) and others suspect the presence of cow, pig, and chicken DNA in commercially prepared deoxynucleo-
-------
side triphosphates (Shanks et al., in press). Reagents should be opened only in dedicated work areas and
used exclusively for MST analyses. To screen for extraneous nucleic acids in PCR reagents, researchers
should perform at least 20 no template PCR reactions with the reagents prior to the initiation of a study. Re-
searchers should also do at least one method blank before environmental water samples are processed in the
laboratory to monitor for extraneous nucleic acids in extraction reagents.
Inhibition of Nucleic Acid Amplification. PCR methods that exhibit a reduced limit of detection may be
inhibited by substances that co-extracted with nucleic acids from water samples. Inhibition may be total
or partial and can manifest as complete reaction failure or as a reduced limit of detection. Some inhibitory
substances observed in environmental samples include detergents, humic acids, polysaccharides, fats, and
other cellular debris (Wilson, 1997). To monitor the impact of inhibition, researchers can perform a matrix
spike control for each suspected environmental sample. A matrix spike contains the minimum quantity of
detectable control DNA template and is added directly into a PCR reaction containing sample extract. These
controls are critical for quantitative PCR applications. The matrix control DNA template should be eas-
ily distinguished from wild-type sequences present in the sample extract. Matrix control DNA templates
should be prepared from a cloned DNA fragment containing a modified form of the target sequence by size,
by restriction mapping, and/or by an alternative probe recognition sequence. Modified control DNA can be
prepared by in vitro generation of deletions, insertions, or other sequence changes. For example, a modified
control DNA template engineered with a 20 bp insertion allows for gel visualization of wild-type and modi-
fied control DNA sequences simultaneously (Figure 5.4).
Overlap Extension PCR Mutaaenesis:
1
I
5'—X
•X
•x1
2% Aaarose Gel:
Wild-type (574 bp)
Modified Template (594 bp)
Figure 5.4 Potential internal control for an environmental sample that will be analyzed with a PCR based method.
The top panel illustrates the construction of modified control template using overlap extension PCR
(Higuchi et al., 1988). The bottom panel shows discrimination of wild-type template (574 bp) from
modified DNA template (594 bp) on a 2% agarose gel.
5.4 Conclusions
A comparison of quality measures for each available MST method uncovered a shared set of method per-
formance criteria. These criteria are organized into five key quality measure issues. Specificity verifies the
-------
ability of an MST method to discriminate between different animal sources present iin a watershed. Preci-
sion quality measures variability between test sample replicates and independent test sample batches. Con-
trol samples screen for the presence of extraneous microorganisms or nucleic acids introduced during the
MST process and ensure that experimental technique, consumables, and equipment are functioning properly.
Thorough quality assurance documentation of all parts of the MST process, especially method validation and
sample acquisition encourage the accurate transfer of information from laboratory scientists to decision-mak-
ing management. Finally, the incorporation of quality measures at recommended frequencies ensures the
validation of high quality data and responsible data interpretation.
In addition to universal performance criteria, some MST methods require additional quality measures.
Library-dependent methods must pay careful attention to library construction. Factors including library
sensitivity, composition, size, and continuity directly impact the quality of MST data. Library-independent
methods that utilize PCR strategies require rigorous adherence to quality standards that measure the limit of
detection and that reduce contamination of MST experiments with extraneous nucleic acids originating from
the laboratory environment, equipment, consumables, and reagents. Additional controls must also be includ-
ed that monitor for the presence of inhibiting substances that often co-extract with nucleic acids recovered
from environmental samples.
Accurate characterization of the source of fecal pollution in a watershed allows managers to identify the
most appropriate management action to restore or protect an impaired waterway. Although it may not be
feasible to include all of the recommended controls, the more controls used the more confidence a decision
maker will have when evaluating MST data. In addition, quality measure recommendations will help bring
more uniformity to MST research, will lead to more effective method evaluations, and the practice of sound
science.
-------
Chapter 6. Assumptions and Limitations of MST Methods
6.1 Introduction
Just as no "ideal" indicator organism for the assessment of water quality has been identified, an active body
of research continues to seek the ideal source identifier (SI) for fecal contamination in environmental waters.
This section will define the characteristics that MST practitioners seek in an ideal SI, which could theoreti-
cally be a chemical, a virus, a bacterium or other microorganism, or a gene(s). In many MST applications,
the SI is subtyped ("fingerprinted") in order to discriminate between particular subtypes that are associated
with various host sources. Many discriminatory characteristics of Sis are used in MST methods, including SI
strain/species, fingerprint pattern, or genetic marker; therefore these will be grouped under the acronym SPM
(species/pattern/marker). The ideal characteristics of Sis and SPMs will be compared with the more realistic
expectations for good or useful SIs/SPMs.
Every new field of scientific inquiry must make some practical assumptions about the effect of variables on
the application of the method. Part of the process of maturation of that field is framing the assumptions as
scientific hypotheses, followed by rigorous hypothesis testing. MST investigators are actively involved in
this process, as highlighted in several recent reviews (Scott et al., 2002; Simpson et al., 2002; Stewart et al.,
2003). In this Chapter, the assumptions made about various organisms and methods currently used for MST
will be discussed in conjunction with the hypotheses that have been tested. Assumption/hypotheses that
remain to be tested will be outlined, and the known limitations of and concerns about the methods will be
presented.
Table 6.1 outlines the characteristics of a hypothetical, ideal source identifier, and contrasts them with the
characteristics of a useful SI. MST investigators have identified many SI candidates, and MST approaches
have focused on various SPMs, some of which are illustrated in Figure 6.1. None of the Sis currently in use
have been demonstrated to have all the characteristics listed. Many methods are in an early stage of develop-
ment, and further research may demonstrate that some possess all or most of these attributes.
6.2 Host specificity of specific strain/pattern/marker (SPM)
The ideal SI would be unique to a host species, and have no alternative sources. Furthermore, the SI would
be represented by variants, each of which would be unique to a host species that contributes contamination
to water bodies. MST would be a much simpler field if all of the fecal microorganisms we use as indicator
organisms were strongly and specifically associated with the gastrointestinal tract of their respective hosts;
however, many fecal bacterial strains appear not to be host-specific. Strains that inhabit multiple host types
have been termed "transient" (Harwood et al., 2003; Myoda et al., 2003), a term borrowed from earlier work
on E. coli population dynamics. In the population dynamics literature the term "transient" had a different
meaning, as it described subtypes that were not observed consistently in host individuals (reviewed in Hartl
and Dykhuizen, 1984); i.e., they were sampled only once or infrequently from an individual. Other MST
practicioners have utilized the term "cosmopolitan" to describe the multiple-host phenomenon (Field et al.,
2003; Whitlock et al., 2002), which refers to the organism's ability to inhabit various host species, and im-
-------
Table 6.1 Characteristics of an ideal source identifier (SI) and those of a useful source identifier3.
Characteristic
Host specificity
Distribution in host
Stability of pattern/ marker
Temporal stability in host
Geographic range/stability
Representative sampling
Ideal SI
Specific strain/pattern/marker (SPM)
found only in one host species.
Found in all members of all populations
of a host species.
Not subject to mutation or methodologi-
cal variability
No temporal variability within host indi-
viduals or host populations
SPM associated with a particular host are
constant across broad geographic ranges
The diversity of the SI in host popula-
tions and in water is represented by a
small sample size
Useful SI
Specific SPM is differentially distributed
among host species of interest.
Found in the waste streams from host
species that could impact the study area
Rarely subject to mutation; methodology
has defined reproducibilityb
Temporal variability in individuals is
balanced by temporal stability in host
populations
SPM associated with a particular host
can be consistently identified across the
geographic area to be studied
The diversity of the SI in host popula-
tions and in water can be represented by
a reasonable sample size
Survival in water
A. Rate of decay
B. Abundance in 1° vs. 2° habitat
Quantitative assessment
Relevance to regulatory tools
Relevance to health risk
Consistent decay rate in various types of
waters and habitats; no growth under any
conditions
The distribution of SPMs in source
material, i.e. feces, does not change after
delivery to the water
The relative and absolute contribution
of each host to SI concentration can be
assessed
The SI itself is also used to regulate
water quality, i.e. coliform, enterococci
The SI itself constitutes a health risk
Predictable decay rate in various types of
waters and habitats; no growth under the
conditions of the study area; all SPMs
decay at the same rate after leaving host
The distribution of SPMs in water bears
a significant resemblance to that found in
contaminating fecal material
SI may not be quantitative, but accurately
indicates presence/absence of source, e.g.
conventional PCR markers
The SI is correlated with a regulatory
water quality parameter
The SI is correlated with health risk
a Strain/ pattern/marker is abbreviated SPM.
b Methodological reproducibility refers to the ability to generate the same pattern (i.e. a DNA or phenotypic profile) or result (i.e.
PCR +/-) from independent assays.
plies nothing about the length of the habitation, which might be long-term or short-term, nor the geographic
distribution. It should be noted that apparent lack of host specificity (observation of a SPM in more than one
host) could be due to insufficient discrimination in the typing method; however, even very highly discrimina-
tory methods such as PFGE identify cosmopolitan isolates.
The cosmopolitan distribution of some SPMs undoubtedly has a negative influence on MST applications,
but efforts to understand the impact of this phenomenon are complicated by the fact that discrimination
between microbial subtypes (strains with different fingerprints) depends upon the method utilized for subtyp-
ing (Guan et al., 2002; Johnson et al., 2004). The difference in discriminatory capability of the various MST
methods has made comparison of studies that rely on different analytical methods extremely difficult; how-
ever, cosmopolitan host distribution is well-documented for E. coli. Multilocus enzyme electrophoresis of
E. coli revealed that 24 of 270 electrophoretic types were found in more than one (up to seven) distinct hosts
(Ochman et al., 1983). Genotyping by REP PCR revealed some identical E. coli subtypes in gull feces and
-------
B
S* «.. «M ! - t -
zsss s;*U .
(A)Courtesy of Michael Sadowsky (B) Courtesy of Valerie J. Harwood
(C) Courtesy of Charles Hagedorn
(D) Courtesy of Troy M. Scott
Figure 6.1 Illustration of some strains/patterns/markers (SPMs) currently utilized in MST methods. (A) rep-PCR
patterns of E. coli isolates - each vertical lane represents one pattern. (B) ribotype pattern of one £. coli
isolate (C) carbon source utilization pattern of one Enterococcus isolate (D) specific genetic marker (esp
of E. faecium) amplified by PCR of Enterococcus DNA
sewage (McLellan et al., 2003). A total of 22% of all distinct E. coli ribotypes (two-enzyme) isolated from
cattle, chickens, horses and swine were shared by some combination of host species (Hartel et al., 2002),
which represented 66% of all isolates tested. Absolute specificity was also lacking in F-specific coliphages;
three serotypes (Type 1,11 and III) were found in municipal wastewater, and each of these was also found in
animal feces (Cole et al., 2003). Only Type IV coliphages were specific to animal feces. No coliphage type
was specific to human-derived wastewater, although Type II coliphages were the dominant serotype isolated
from wastewater. Because E. coli, Enterococcus spp. and coliphages are commensal fecal indicators that are
broadly distributed in feces and are widely used by the regulatory and MST community, we suggest that a
better understanding of cosmopolitan distribution, and how profoundly it affects MST methods, is particu-
larly important in these organisms. Furthermore, as new methods are developed their host specificity or host
range should be fully explored.
It has been suggested (Simpson et al., 2002) that host specificity would be augmented if the MST target con-
tributed to the specific interaction between host and fecal microbe. Candidates include the genes that code for
microbial appendages such as pili and adhesins, which mediate attachment to cells of the host gastrointestinal
tract. One method capitalizing on this approach is PCR amplification of the gene for the enterococcal surface
protein (esp) of E. faecium (Scott et al., 2004), which, though promising, requires further validation. Enteric
viruses, which rely on specific cell surface receptors to bind to host cells, are inherently species-specific and
have been used to assess the presence of human fecal contamination in environmental waters (Griffin et al.,
1999; Jiang et al., 2001).
6.3 Widespread distribution of SI and SPM in host populations
An MST marker that is adopted for water quality and total maximum daily load (TMDL) assessment and
restoration throughout the U.S. will of necessity be widely distributed in host populations across the country.
Thus, relatively rare markers such as some genes associated with pathogens are likely to be less useful than
more common markers, even though they may be highly host-specific. For example, in Europe a bacterio-
-------
phage (bacterial virus) that infects Bacteroides fragilis HSP40 was found only in human sewage and in sew-
age-contaminated waters (Tartera et al., 1989). This bacteriophage was considered a promising candidate for
a human-specific fecal marker; however, its limited distribution in sewage (Scott et al., 2002) and the relative
difficulty of the method (Leclerc et al., 2000) have probably contributed to its rare use status in the U.S. F-
specific coliphages are common in sewage, but it has been estimated that only -3% of humans carry this type
of phage (reviewed in Leclerc et al., 2000).
The hypothesis that other proposed SPMs have widespread distribution in the gastrointestinal tracts of their
respective hosts must be tested. Included among these are the species-specific genetic markers amplified
from Bacteroides (Bernhard and Field 2000a, b), the toxin genes of E. coli found in pigs and cattle (Chern et
al., 2004; Khatib et al., 2002; Tsai et al., 2003), and the esp gene of Ent.faecium (Scott et al., 2005). Some
information is available for E. coli toxin genes LTIIA and STII, as the prevalence of species-specific forms
of these markers was measured in animal waste from farms in several states (Khatib et al., 2002; Khatib et
al., 2003). More than 93% of samples from cattle waste lagoons were positive for the cattle-specific LTIIA
marker when >103 E. coli were screened, and the frequency of positive results rose to 100% when >105 E.
coli were screened (Khatib et al., 2002). The swine-specific STII marker was found in 100% of samples
when 35 E. coli were screened (Khatib et al., 2003).
Ideally, host-specific SPMs should be present at about the same density in separate populations of a given
host species, which would provide greater confidence that sampling effort was adequate when using stan-
dardized protocols. Furthermore, it would be advantageous if host-specific SPMs were found at about the
same density in various individual animals within a host population, which would facilitate accurate quanti-
fication. Very little is known about these concerns for any of the methods, except that the majority of animals
in a herd do carry E. coli, but generally do not carry enterotoxigenic E. coli (Chern et al., 2004).
6.4 Stability of the signal
A required characteristic for a useful SPM is stability of the "signal", whether that signal is a phenotypic
pattern, a genetic pattern, or a PCR product. The assumption that genetic patterns/markers are a more stable
type of signal than phenotypic patterns has appeared frequently in MST literature (Parveen et al., 1999; Scott
et al., 2002; Simpson et al., 2002), due in part to the fact that bacterial phenotypes (traits such as antibiotic
resistance or the ability to use a particular carbon source) are influenced by environmental conditions as well
as the genetic makeup of the organisms. This assumption should, however, be tested in the context of an
MST study, in which the testing occurs in a controlled laboratory environment under near-optimal growth
conditions. While it is known that some bacteria lose resistance to antibiotics when selective pressure (anti-
biotic presence) is removed, as occurs when bacteria are cultured from feces or water samples, it is unknown
whether this phenomenon occurs often enough to significantly impact the accuracy of MST studies based on
antibiotic resistance patterns. Similarly, the frequency and consequences of transfer of antibiotic resistance
genes from one SI to another are not established for MST.
The gene(s) for ribosomal RNA (rRNA) are frequently targeted for MST studies (Carson et al., 2003;
Parveen et al., 1999) because these genes mutate relatively rarely. Ribosmal RNAs are an integral component
of the ribosome, the protein-synthesizing "machinery" of the cell, and certain regions of rRNA are very high-
ly conserved (change very little, if at all, over thousands of generations). The low mutation rates of the rRNA
genes do contribute to the stability of many types of fingerprints; in fact, sequencing of rRNA genes within
a species such as E. coli generally results in very little strain discrimination (Guan et al., 2002). Ribotyping
as it is used for MST should, perhaps, be clarified as "genomic ribotyping", since the chromosomal DNA
is isolated, cut with restriction enzymes, and chromosomal fragments are separated by electrophoresis (see
-------
Chapter 3-Methods). Labeled fragments of the rRNA gene(s) are then used as probes to identify the gene loci
on the chromosome. This method can be quite discriminatory, even within a species, because much of the
variability in patterns is due to variation outside the conserved rRNA operons. Although it has been assumed
that ribotypes represent a very stable form of signal, no comparisons with phenotypic or other genotypic
methods have been published.
A linked assumption of many MST methods is that mutations in host individuals that could change the speci-
ficity of the SPM are very rare. An individual could lose the ability to support the SPM if, e.g., a receptor in
the gastrointestinal tract experienced decreased affinity for the SPM. Conversely, an individual from a dif-
ferent species might acquire the ability to support the SPM by mutation or horizontal gene transfer. While a
recent mutation in a host population would not be a major concern, because few individuals would carry the
mutation, over generations it could pose a problem, particularly in isolated host populations.
6.5 Transferable methodology
It is assumed that MST methods will be transferable across laboratories. The ability to successfully perform
many of these methods will be dependent upon the relative expertise of laboratory technicians, the equipment
and facilities available, and the extent to which protocols are standardized and made "user friendly." As pro-
tocols are being developed, every effort should be made to include rigorous controls and streamlined tech-
niques into MST methods. The error associated with the method, whether described in terms of false-posi-
tives and false-negatives, or Type I and Type II error, should be thoroughly explored. An important aspect of
the analytic parameters used for matching patterns is that as the similarity index required to call two patterns
the same becomes more stringent, the number of distinct patterns (ribotypes, for example) identified increas-
es (Hartel et al., 2002). The similarity values imposed for pattern matching must not be chosen arbitrarily,
but should rely on measurements of the inherent variability of the method. For example, if E. coli isolate X
is ribotyped ten times on ten separate occasions, what is the similarity of those patterns? The discriminatory
power of the method cannot be greater than its inherent variability, i.e. if ten replicate measurements of the
ribotype of E. coli X are 92% similar, only ribotypes that are less than 92% similar can legitimately be called
different ribotypes. Ideally, a confidence interval should also be calculated to better define differences that
should be considered significant, although this has not been practiced in the literature. It is important to keep
in mind that development of any MST method that analyzes patterns based on band-matching algorithims
requires confirmation of pattern matches and nonmatches by eye before one can rely on the matches called
by the software.
6.6 Temporal stability within the host
The ideal SI should exhibit stability within individual host animals and within host populations over time.
Although a good deal of information is available on the temporal stability of E. coli populations in host ani-
mals, very few studies have addressed the temporal stability of other Sis.
Previous studies on the temporal variability of E. coli established the concept of transient vs. resident popu-
lations of E. coli in the gastrointestinal tract. Caugant et al. (1981) defined a "transient" population as one
observed at only one sampling point, while a "resident" population was one observed at more than one
sampling point. Transient vs. resident populations are a particularly relevant MST concern if the range of
subtypes estimated in natural populations of E. coli (100-1000 per host species by multilocus enzyme elec-
trophoresis) (Selander et al., 1987) are found to be comparable in other fecal indicator bacteria. Over an
11-month period, only 5.6% of the E. coli isolated from the feces of a single human host were considered
"resident"(Caugant et al., 1981), and a total of 53 electrophoretic types were identified using multilocus
-------
enzyme electrophoresis. In another study, resident E. coli populations from multiple hosts accounted only
for 8% of all the electrophoretic types identified (Ochman et al., 1983). A study on temporal stability of E.
coli in humans, cattle and horses defined a "persistent" ribotype as one that was sampled from an individual
in two consecutive sample events (Anderson 2003; Anderson et al., 2003). At least one persistent ribotype
was observed per human, although only four of 36 (11%) of the ribotypes observed in the three humans were
persistent. E. coli populations of horses and cattle tended to display higher diversity (more subyptes per host)
than those of humans; however, they followed a similar trend in that most of the E. coli subtypes observed
were not persistent (Anderson 2003; Anderson et al., 2003). These studies indicate a high probability that the
E. coli subtype(s) obtained from a single host at a given time are not representative of the E. coli population
in the animal's feces over time. Such a limitation has major repercussions in the establishment of host origin
libraries, which may require continuous updating in order for a particular MST methodology to be able to
track the host species (Jenkins et al., 2003) over an extended period of time.
While temporal stability of the SPM in individual host animals is an ideal characteristic for MST, temporal
stability at the larger host population level is a characteristic of a useful SI. In a recent study on the temporal
stability of E. coli ribotypes in cattle herds, individual cattle in the herds were sampled at random during four
sample events (Jenkins et al., 2003). The E. coli ribotypes that were observed in more than one sample event
("residents") represented only 8.3% of 240 ribotypes. Among the 20 resident ribotypes, no ribotype was
found at all four sampling times or in all of the steers sampled. Although many E. coli isolates were analyzed
per cow (~11 to -25), individual cattle were not resampled throughout the study. Thus, it could be argued
that the observed variability was as likely due to undersampling of individuals in the herd as temporal vari-
ability. However, in support of the above results are data from an eight-month study of three beef cattle from
one herd (Anderson 2003; Anderson et al., 2003) that were repeatedly sampled. E. coli ribotype variability
in the feces of these animals was high, sharing between herd members was low, and temporal variability in
the dominant ribotypes within each animal was consistently noted. Evidence of the temporal variability of E.
coli populations in other species was observed in humans and horses (Anderson 2003; Anderson et al., 2003).
Two humans that lived together tended to share E. coli ribotypes with each other, but not with a human work-
ing in the same room, while horses in the same herd shared very few subtypes (Anderson, 2003; Anderson et
al., 2003). However, investigations of temporal stability carried out on a larger scale (and with a different SI)
were more encouraging, as the temporal stability of a large library of Enterococcus spp. subtyped by antibi-
otic resistance analysis was demonstrated for up to a year (Wiggins et al., 2003).
6.7 Geographic stability
Several assumptions based on the geographic distribution of an ideal SI can be identified: (a) SPMs sampled
from one population of a host species will be similar to SPMs sampled from another population of the same
host species, and a predictive relationship can be established between the two; and (b) SPMs sampled from
host populations separated by broad geographic ranges will exhibit a high similarity index and accurately
track the host species. A hypothesis that could be contradictory to (a) and (b) has also been proposed: (c)
SPMs exhibit geographic structure, that is, the similarity of SPMs in various populations of one host species
is directly proportional to their geographic distance from one another (Gordon, 2001).
Studies indicate that hypothesis (c) regarding geographic structure for populations of the same host species
is not met for E. coli populations; however, this assumption is probably the least important one for most
MST applications. Very little of the variability in E. coli populations of humans seems to be attributable to
geographic separation (Caugant et al., 1984; Whittam et al., 1983); which may be partly due to the mobil-
ity of human populations (Gordon, 2001). Caugant et al. (1984) reported that little geographic structure
was observed in E. coli populations of families living within the same city, where only 6% of the variability
-------
was explained by geographic distance. Only 1% of the variability was explained by geographic distance for
families living in different cities. Geographic structure accounted for only a small percentage (i.e., 2%) of the
variability in E. coli subtypes in mice (Gordon, 1997). While studies that have compared E. coli population
structure in various animals have found significant contributions to diversity from both geographic loca-
tion and host source (Gordon and Lee, 1999; Souza et al., 1999), only a small percentage of the variability
(<20%) was accounted for by these factors. One study on livestock did find geographic structure in E. coli
populations in cattle and horses, i.e. more ribotypes were shared in host populations in closer geographic
proximity; however, no geographic structure was observed for E. coli from chickens and swine (Hartel et al.,
2002).
Ideally, host populations in all parts of the U.S. would share similar Sis and SPMs so that nationwide (or
more inclusive) databases could be constructed. Studies completed to date suggest that this ideal will not be
met, at least for library-based methods. In a study performed across a relatively broad geographic area in
Florida, E. coli from beef, dairy, poultry, swine and human hosts were ribotyped by a one-enzyme procedure
(Scott et al., 2003). Although the method accurately differentiated E. coli originating from human vs. non-
human hosts, it failed to distinguish among the different non-human host species across the broad geographic
region. The diversity and distribution ofE. coli ribotypes differed in captive vs. wild deer (Hartel et al.,
2003), which was attributed to diet. The diets of host animals may differ significantly by geographic region,
providing one of the drivers for geographic variability of commensal bacterial populations in one host spe-
cies. E. coli and Enterococcus libraries from three geographic regions were assessed for broad geographic
applicability (Dontchev et al., 2003). Subtyping methods used were antibiotic resistance analysis (ARA),
ribotyping (one-enzyme) and pulsed field gel electrophoresis (PFGE). The regional sublibraries (Florida,
Shenendoah Valley VA and southwest VA) identified isolates collected from within the region significantly
more accurately than they identified isolates from outside the region. A three-region merged library identified
the source of isolates much less accurately than each of the regional libraries, and this generalization held
true for each of the methods and Sis.
The geographic applicability of an Enterococcus ARA library was broadened by increasing library size and
representation of isolates from a number of watersheds in the Shenendoah Valley region of Virginia (Wiggins
et al., 2003). Six watershed-specific libraries were merged to produce a library of 6,587 isolates, which iden-
tified the source of enterococci fairly accurately across the combined geographic area. The geographic range
of the merged library was limited, as it identified isolates from southwest Virginia and Florida significantly
less accurately than isolates from the six-watershed region.
6.8 Representative sampling
One of the most important assumptions of any MST method is that the SI population can be adequately
sampled so that all (or most) SPMs are represented. The assumption of representative sampling is extremely
important with respect to sampling of both host fecal material and SPMs in water samples. Many factors im-
pose limits on the amount of material or isolates that can be analyzed, including cost and time. Under-sam-
pling of SI populations in fecal sources leads to nonrepresentative libraries, which may have high correct
classification rates (internal accuracy) but low predictive accuracy for isolates that are not included in the
library (Whitlock et al., 2002; Wiggins et al., 2003). Furthermore, nonreprepresentative libraries will display
neither temporal nor geographic stability. Various estimates of E. coli subtype diversity within host popula-
tions have been advanced, e.g. between 100 and 1000 (Milkman, 1973; Selander et al., 1987). Rarefaction
analysis of an E. coli rep-PCR library determined that a library size of 1535 isolates from humans and twelve
animal species was not close to saturation (Johnson et al., 2004), which demonstrates the great diversity in E.
coli genotypes. A 2:1 ratio of total isolates analyzed to estimated subtype richness has been suggested as a
-------
minimal requirement for capturing diversity (Jenkins et al., 2003; Parveen et al., 1999), further increasing the
sampling effort needed. Complicating the issue is the fact that different host species and sample types (for
example, human feces vs. sewage) contain E. coli populations of differing richness (Stoeckel et al., 2004),
indicating that sampling effort should be adjusted based on host species and sample type. The apparently
low frequency of "resident" E. coli subtypes compared to "transients" may be more a reflection of sampling
limitations than it is a true characteristic of E. coli populations (Jenkins et al., 2003). Achieving representa-
tive sampling of E. coli populations in environmental waters will be affected by similar concerns; i.e. high-
diversity E. coli populations were found in both pristine and anthropogenically impacted waters (Chivukula
and Harwood, 2004).
6.9 Persistence of SPMs in environmental waters
Microbial source tracking studies contain many implicit assumptions about the survival of a chosen source
identifier. Many of these assumptions are based upon simplified or idealized views of microorganism sur-
vival characteristics. For example, it may be assumed that the decay rate of E. coli SPMs entering a stream
directly from the feces of a herd of cattle will be exactly the same as the E. coli SPMs entering the stream
from a failing septic system. However, a significant difference in decay rate might influence the relative
numbers of E. coli SPMs recovered downstream from these fecal sources, which would in turn lead to inac-
curate assessment of initial fecal loads. It is important that such assumptions be recognized and understood
when choosing a SI and designing or interpreting MST studies.
The failure of the fecal coliform/fecal streptococcus (FC/FS) ratio for fecal pollution source tracking is a
lesson to heed in the current pursuit of MST methods. The lesson is particularly relevant when considering
attempts to quantify source contributions for total maximum daily load (TMDL) assessments. The FC/FS ra-
tio has been criticized for aspects such as differential decay rates for fecal coliform and fecal streptococci in
aquatic environments (American Public Health Association, 1995; Simpson et al., 2002). Initial assumptions
about the comparable survival of coliforms and streptococci proved invalid after further study, and use of the
FC/FS ratio as a MST method has decreased in recent years. The lesson identifies the importance of testing
survival assumptions for MST Sis before methods are widely applied to source tracking problems.
In order for a microorganism to be considered an ideal source indicator, it must meet a number of criteria
pertaining to its survival in aquatic environments. An ideal SI would not exhibit any population growth upon
entering aquatic environments. It would also have SPM decay rates that are constant over space and time.
For example, SPM decay rates would not vary between water types (e.g. temperate freshwater lake or tropi-
cal saltwater beach) or across aquatic habitats within a watershed (e.g. lake water column or river sediment).
In addition, an ideal SI would have SPM decay rates that would be constant between its primary fecal habitat
and secondary aquatic habitats. Any variance from these ideal survival characteristics could have important
implications for interpreting results from MST studies.
None of the currently used source identifiers are known to meet all of these ideal survival criteria. Therefore,
it is important to understand their survival characteristics to determine where they can still be useful under
the conditions of a specific MST study. The survival of some source identifiers has been better studied than
others, and in these cases, their survival characteristics may be sufficiently predictable to make the micro-
organism a useful source identifier under the conditions of a specific MST study area and time. In other
cases, important survival hypotheses remain untested and survival characteristics poorly known. This lack
of information can compromise the value of the source identifier. The following section explores several as-
sumptions about survival for three of the more commonly used source identifiers: E. coli, Enterococcus sp.,
and Bacteroides spp.
-------
Escherichia coli: (i) The SPM decay rates are always negative after it enters water.
Escherichia coli has been regarded as a good practical indicator of fecal pollution that generally survives in
aquatic environments between 4 and 12 weeks (Edberg et al., 2000). There are many studies indicating its
decay rate is negative after entering water environments such as: river water (Grabow et al., 1975), ground-
water (Filip et al., 1987), and seawater (Rozen and Belkin, 2001). However, there are a growing number of
reports suggesting that some E. coli SPM decay rates may not be negative under certain conditions in aquatic
environments.
A number of studies have provided evidence that suggests that E. coli can multiply in certain tropical and
subtropical environments (Byappanahalli et al., 2003a; Carrillo et al., 1985; Desmarais et al., 2002; Hardina
and Fujioka, 2001; Rivera et al., 1988; Solo-Gabriele et al., 2000; Byappanahalli and Fujioka, 2004). For
example, one study found high levels of E. coli in Florida riverbank soils, and suggested that E. coli could
be washed into the water during high tides (Solo-Gabriele et al., 2000). Associated laboratory experiments
found that E. coli was capable of increasing by several orders of magnitude in these soils, and suggested
the importance of soil properties and periodic wetting and drying as influential for "multiplication". Micro-
cosm experiments by Byappanahalli and Fujioka (2004) indicated that E. coli has the capacity to multiply in
tropical soils, but the bacteria require suitable nutrient and moisture conditions availability. A Tropical Water
Quality Indicator Workshop in 2001 agreed upon a consensus statement that fecal indicator bacteria like E.
coli can multiply and persist in soil, sediment, and water in some tropical/subtropical environments (e.g.
Hawaii, Guam, Puerto Rico, south Florida) (Fujioka and Byappanahalli, 2003).
The question of E. coli multiplication in certain temperate environments is also under investigation. For
example, E, coli counts in sand and water gradually increased over the bathing season at a Lake Michigan
beach (Whitman and Nevers, 2003), which was attributed to higher survival rates (lower decay rates), and
perhaps growth, in warmer temperatures. Growth of E. coli associated with the macro-alga Cladophora
mats in the Great Lakes has also been investigated (Byappanahalli et al., 2003b; Whitman and Nevers, 2003;
Whitman et al., 2003). E. coli survived over 6 months in Lake Michigan Cladophora algal mats (sun-dried
and stored at 4°C) and then quickly multiplied when moisture was returned (Whitman et al., 2003). The
authors suggested that Cladophora could be a secondary habitat and source for E. coli in certain beach areas,
although the case for natural multiplication needed further validation.
(ii) The SPM decay rates are constant across aquatic habitats.
There have been numerous studies to investigate E. coli survival in aquatic environments (Santo Doming et
al., 1989) and in laboratory microcosms simulating aquatic habitats (Santo Domingo et al., 2000). However,
it can be difficult to compare survival studies across different microcosm designs and experimental condi-
tions (e.g. many microcosm studies have been conducted under filtered water or sterile conditions). For this
reason, the studies reviewed below are from field studies or laboratory experiments conducted under non-
sterile conditions, and survival results are identified based upon whether they were obtained from field stud-
ies or from laboratory microcosms.
There are numerous studies to indicate that E. coli decay rates are not constant across aquatic habitats. Mi-
croorganisms entering aquatic habitats might generally be expected to survive longer under colder tempera-
tures, or if they are attached to particles. For example, a number of studies have found significantly lower
decay rates for E. coli in sediments than in the associated water column (Burton et al., 1987; Craig et al.,
2004; Gerba and McLeod, 1976; LaLiberte and Grimes, 1982). Craig et al. (2004) found E. coli at > 5 x 103
CFU/100 g after 28 days in the sediments of saltwater microcosms, while they were undetectable after 7 days
in microcosms containing only water. In situ measurements of fecal coliforms in both water and sediment of
-------
their river and beach study showed that a rain event caused an intial peak of similar levels in both river water
and sediment, which was followed by a more rapid decline of fecal coliform numbers in water than in sedi-
ments. Two days after the peak, levels of fecal coliforms were 100 times greater in river sediment compared
to water. The authors concluded that there was extended persistence of fecal coliforms in the coastal sedi-
ments compared to water (Craig et al., 2004).
There is also evidence of variable E. coli decay rates across different types of sediment. E. coll decay rates
varied according to sediment type, with the greatest rates of decay occurring in beach sediment microcosms
consisting of large particle size and high organic carbon (Craig et al., 2004). Burton et al. (1987) found
enhanced survival ofE. coli in sediments with high proportions of clay and nutrients compared with sandy
low-nutrient sediments.
These laboratory and field experiments are consistent with observations about E. coli persistence from field
surveillance studies. Higher E. coli counts were consistently found in stream and bank sediments than in
the stream water of a small Indiana watershed (Byappanahalli et al., 2003a). The authors suggested that the
widespread and consistent occurrence of E. coli in the watershed could be attributable to long term per-
sistence (and/or multiplication) of E. coli in soil and sediment, and the subsequent erosion and washing of
sediment-borne E. coli into the water. Considerable progress is being made toward understanding the per-
sistence of E. coli in beach habitats that may prove informative for MST studies, i.e. E. coli counts per unit
weight were 3-17 times higher in sand than in the water column at 6 freshwater bathing beaches in the Great
Lakes (Wheeler-Aim et al., 2003). Similarly, E. coli counts in foreshore sand were typically several orders
of magnitude higher than in the water at a Lake Michigan beach (Whitman and Nevers, 2003). These results
strongly suggest that E. coli decay rates are lower in beach sand than in the water column, and that beach
sand could be a significant reservoir for longer term persistence, and subsequent resuspension of E. coli into
beach waters.
Elevated salinity has a detrimental effect on fecal coliform and E. coli survival, particularly in the water col-
umn. Numerous studies have shown that the decay rates of these organisms are much greater at marine/estua-
rine salinities compared to freshwater (Hood et al., 2002; Sinton et al., 2002). Solar radiation also increases
E. coli decay rates (Sinton et al., 2002; Whitman et al., 2004).
Enterococcus: (i) The SPM decay rates are always negative after it enters water.
Many studies indicate that culturable enterococci decline after entering aquatic environments (Sinton et al.,
1993; Sinton et al., 2002). However, there are also a growing number of reports suggesting that Enterococ-
cus SPM decay rates may not be negative under certain conditions in aquatic environments.
Several studies have provided evidence indicating that Enterococcus spp. may be able to multiply in certain
tropical and subtropical environments (Desmarais et al., 2002; Fujioka et al., 1999). A Tropical Water Qual-
ity Indicator Workshop in 2001 reached a consensus statement that fecal indicator bacteria like Enterococ-
cus can multiply and persist in soil, sediment, and water in some tropical/subtropical environments (Hawaii,
Guam, Puerto Rico, south Florida) (Fujioka and Byappanahalli, 2003). However, microcosm experiments by
Byappanahalli and Fujioka (2004) suggested that enterococci might require more complex nutrients than E.
coli and, thus are less likely to multiply in tropical soils.
Enterococci may also be able to multiply under certain conditions in temperate aquatic environments. En-
terococcus spp. on drift seaweed at recreational beaches in New Zealand exceeded seawater levels by 2-4
orders of magnitude (Anderson et al., 1997). The presence of genetically identical (clonal) enterococci
dominating seaweed populations was suggested as evidence that active growth or selection was occurring,
-------
and that enterococci could be washed off into surrounding water. Similarly, a study in southern California
suggested that a tidal saltwater marsh was serving as a source rather than a sink for Enterococcus contami-
nation of nearby coastal beaches (Grant et al., 2001). The possible growth of Enterococcus spp. associated
with Cladophora mats in the Great Lakes has also been investigated (Byappanahalli et al., 2003b; Whitman
et al., 2003).
Enterococcus'. (ii) The SPM decay rates are constant across aquatic habitats.
There have been fewer studies of Enterococcus spp. decay rates in aquatic ecosystems than E. coli; however,
available information suggests that enterococci decay rates are not constant across aquatic habitats. Decay
rates of enterococci from municipal waste stabilization pond effluents differed in river water depending upon
salinity, season and sunlight exposure (Sinton et al., 2002), i.e. decay rates were higher in more saline wa-
ters, in the summer, and when exposed to increased sunlight. Some Enterococcus species have been associ-
ated with occurrence on plants (Mundt, 1961; Geldrich and Kenner, 1969) and in insects (Martin and Mundt,
1972) which may suggest the possibility of more diverse SPM survival strategies which should be tested.
Bacteroides: (i) The SPM decay rates are always negative after it enters water.
Enteric anaerobes like Bacteroides spp. and Bifidobacterium spp. have been suggested as indicators of recent
fecal pollution because they are believed to have predictably negative decay rates and survival times of
hours in oxygenated waters (Carrillo et al., 1985; Fiksdal et al., 1985; Resnick and Levin, 1981). Bernhard
and Field (2000a) suggested that ease of detection and longer survival in water made Bacteroides-Prevotella
genetic markers superior to those of Bifidobacterium.
Bacteroides fragilis did not maintain culturability as well as E. coli or Enterococcus faecalis in dialysis bags
suspended in aerobic freshwaters (Fiksdal et al., 1985), but immunofluorescence assays demonstrated 18%
persistence after 192 hours. Another study found that Bacteroides cells could survive for up to 6 days in
drinking water under oxygen-stressed conditions (Avelar et al., 1998).
Like all MST methods, the usefulness of nonlibrary-based methods such as PCR detection of Bacteroides is
based upon the assumption that these anaerobic bacteria do not multiply upon entering aquatic ecosystems.
While the few studies conducted to date suggest this is the case, the fate and ecology of anaerobes like Bac-
teroides spp. in aquatic ecosystems remains poorly understood. It is possible that certain aquatic habitats,
such as sediments, may provide suitable environments for anaerobic Bacteroides spp. to exhibit population
growth. It is noteworthy that it has taken many years of studying the survival/growth of E. coli in diverse
aquatic ecosystems to better understand some of its potential limitations as a source identifier. Hypotheses
related to the possibility of Bacteroides sp. multiplication in certain unique aquatic habitats (e.g. anoxic sedi-
ments) need to be tested.
(iii) The SPM decay rates are constant across aquatic habitats.
There have been few studies of Bacteroides spp. survival in aquatic ecosystems, and so there is insufficient
information to evaluate whether SPM decay rates are constant across aquatic habitats. One study (Kreader,
1998) found that persistence of PCR-detectable DNA from the fecal anaerobe B. distasonis was dependent
upon temperature and predation. Laboratory and in situ studies in river water found that B. distasoni was
detectable by PCR for at least two weeks at 4° C, but for only 4-5 days at 14°C, 1-2 days at 24°C, and 1 day
at 30°C. Although the PCR method detected both dead and living bacteria, predators were considered impor-
tant factors in the decline of both dead and living cells. The author stressed that seasonal variation in the B.
distasoni decay rate would need to be considered for any water monitoring applications.
-------
6.10 Persistence of SPM in primary vs. secondary habitats
Savageau (1983) advanced the concept that the gastrointestinal tract is a primary habitat for.E. coli, while
external environments such as soil and water are secondary habitats. In a recent review related to bacterial
source tracking (BST), Gordon (2001) asserted that for any bacterial species that is used to identify human
and animal sources of fecal pollution in surface water, several assumptions must be validated. One of these
assumptions is that "the clonal composition of the species isolated from soil and water [secondary habitat]
represents the clonal composition of the species in the host populations responsible for the fecal inputs [pri-
mary habitat] to the environment". The rationale for this statement is clear: if the fecal SPM(s) that are used
as the source identifier persists poorly in the water relative to other SPMs, the source-specific fecal signal
will rapidly disappear.
Several studies on the distribution of Escherichia coli subtypes in the primary habitat vs. secondary habitat
showed distinct differences in subtype distribution between the two. One hundred thirteen distinct E. coli
electrophoretic types (determined by multi locus enzyme electrophoresis or MLEE) were isolated from bird
feces and the litter on which they had defecated. Only 10% of the clones were found in both the primary
and secondary habitat (Whittam, 1989). Another study (Gordon et al., 2002) compared electrophoretic
types (ETs) of E. coli from feces of two human couples, each representing a household, and E. coli from
each household's septic tank. This study indicated that E. coli clones from a secondary habitat such as a
septic tank can differ significantly from the primary habitat such as the couples' feces. Ribotyping of E. coli
from dog feces, untreated wastewater and contaminated soil inoculated into water showed that the dominant
subtypes in the primary habitat were distinct from those in the secondary habitat, and that certain "survivor"
strains could be identified (Hood et al., 2003). E. coli clones isolated from swine manure slurry (a primary
source, but secondary habitat) were compared to those isolated from soil inoculated with the same slurry by
the genotypic method ERIC-PCR (Topp et al., 2003). Although a major shift in community structure was
evident upon comparison of isolates from the secondary (manure slurry) vs. tertiary (soil) habitats, many
subtypes were shared between the two habitats. However, one SPM that was prominent in manure was not
recovered from soil, indicating differential survival of SPMs.
All types of F-specific RNA coliphages apparently do not have the same decay rate in water, as Type IV
strains were less persistent than Types I, II and III in one study (Brion et al., 2002). F-specific RNA coli-
phages also may experience higher inactivation rates in warm waters compared to cooler waters (Cole et al.,
2003).
Based on the above mentioned phenotypic and genotypic studies, E. coli appears to be a questionable candi-
date as a source tracking organism, although genetic fingerprinting of E. coli is the basis for some commer-
cial source tracking enterprises. For library/culture independent methods such as the PCR for Bacteroides
(Bernhard and Field, 2000b; Bernhard et al., 2003) and E. coli toxin genes (Khatib et al., 2002; Khatib et
al., 2003) the primary habitat versus secondary habitat criterion for validity may be less stringent, but is still
applicable. The library independent method is binary; a genetic signal specific for an animal host is either
detected in an environmental sample or it is not. However, if the signal (DNA in this case) is very short-lived
in the water compared to indicator organisms and pathogens, it will not serve its purpose. Furthermore, ef-
forts are underway to develop quantitative PCR protocols for some markers, and the efficacy of these meth-
ods will rely to a certain extent on the primary-vs.-secondary habitat hypothesis.
6.11 Relevance of SI to regulatory tools
Indicator bacteria such as coliforms have been used for over a century as indicators of fecal contamination
in water. In the U.S., indicator bacteria (fecal coliforms, E. coli and enterococci) are the standard by which
-------
microbial water quality in environmental waters is measured. Currently, almost all MST studies, whether
carried out on bathing beaches, in reservoirs, or for total maximum daily load (TMDL) assessments, are
responses to exceedances of indicator bacteria standards. Understandably, water quality managers prefer a
SI that is directly connected to the regulatory parameter (indicator bacteria) for assessment of fecal sources;
however, as MST methods are tested and validated in the field, a method that utilizes one or more alternative
Sis may show greater utility than methods that use conventional indicator organisms.
An assumption of MST methods that employ Sis other than conventional indicator bacteria is that the results
generated by the SI will have some discernible relationship with indicator bacteria levels. The failure of indi-
cator organism and SI to correlate is not a priori a reason to discard the SI, particularly if it is associated with
human health risk (see below) (Field et al., 2003). The interpretation of the results may, however, prove more
complex when the SI is not an indicator organism, particularly in the case of TMDL assessment. Very little is
known about these relationships in environmental waters, making them an essential area for further study.
6.12 Relevance of SI to human health
The ultimate goal of MST is to determine the host species responsible for fecal pollution from among many
possible candidates; however, simply discriminating human fecal material from nonhuman is of practical
use for water quality managers (Harwood et al., 2003; Myoda et al., 2003; Stewart et al., 2003). The useful-
ness of human vs. nonhuman source discrimination is due in part to the assumption that human fecal material
poses a greater human health risk than other types of fecal material (Scott et al., 2002). Although some of the
rationale for this assumption is based on indirect evidence (i.e. the majority of gastroenteritis associated with
recreational water use is caused by viruses, and human enteric viruses are highly host-specific), direct evi-
dence also exists. Detection of enteric viruses, which are exclusively of human source, was correlated with
gastroenteritis in swimmers in marine waters (Haile et al., 1999), and a meta-analysis of epidemiological
studies showed that enteric viruses were strongly associated with gastroenteritis (Wade et al., 2003). Thus,
Sis that can discriminate human vs. nonhuman fecal pollution should be useful, provided they have some as-
sociation with human health outcomes and/or pathogens.
The indicator organism paradigm is based on the assumption that indicator organisms are predictive of hu-
man health risk. Much debate and many epidemiological studies have explored this assumption (reviewed in
Wade et al., 2003). The 1986 standards for recreational water quality specify the use of E. coli (and not fecal
coliforms) for freshwater bodies, and enterococci for freshwater and marine water bodies (U.S. Environmen-
tal Protection Agency, 1986). A meta-analysis of the epidemiological literature on gastroenteritis resulting
from recreational water use found that E. coli was significantly associated with gastroenteritis in fresh water,
and that enterococci were significantly associated with gastroenteritis in marine water (Wade et al., 2003),
supporting the use of these organisms as indicators of human health risk. Coliphages were also predictive of
gastroenteritis, although fewer studies were available for analysis. Much work remains to be conducted on
the correlation of alternative Sis with human health risk in environmental waters.
6.13 Summary
• None of the source identifiers currently used meet the criteria for an ideal SI, including those that are
indicator organisms recognized for regulatory uses.
• The ecology and population biology of some source identifiers, particularly fecal coliforms/£. coli,
are much better understood than that of others, such as the enterococci and Bacteroides spp. While
-------
the high genetic diversity of E. coli allows great discrimination between subtypes, it also complicates
development of known source libraries.
The correlation of novel Sis such as Bacteroides with levels of conventional indicator organisms
and/or with human health outcomes has not been determined, but should be if public health effects
are under consideration.
-------
Chapter 7. Application of MST Approaches
This Chapter presents a series of case studies involving application of several MST methods. The intent of
this Chapter is to provide some real-world examples of how various MST methods have been applied. There
have been far more studies than can be covered in this Chapter, so several have been chosen as examples.
Many of these examples were compiled based on communications with the authors of the studies. How-
ever, others were been written based only on published reports and/or journal articles, and thus are not as
complete. We have tried to include examples of studies using MST methods in current use, as well as some
projects involving multiple techniques.
Each of the following case studies follows the same general outline. First is a general description of the wa-
tershed, with a statement of the problem and the goals and objectives of the project. Next is a brief descrip-
tion of the methodology used, including the classification method. (For more detailed information on the
methods, please refer to Chapter 3 in this document.) For the studies that used a library-based MST method,
a description of the library is included, with information on known source samples and evaluation methods.
Following that is a section on sampling considerations, describing how and when the water samples were
collected. Finally, a section on the outcomes of the study follows, with a summary of the major results and
conclusions, and information on follow-up studies and implementation efforts.
This Chapter includes 8 case studies (presented in no particular order) which illustrate the use of many, but
not all, currently applied MST methods:
Case 1. Saint Andrews Park (Georgia). Targeted sampling and Enterococcus speciation.
Case 2. Tampa Bay (Florida). ARA with fecal coliforms, ribotyping with E. coli, and human-
pathogenic enterovirus detection.
Case 3. Vermillion River (Minnesota). rep-PCR with E. coli.
Case 4. Anacostia River (Maryland/District of Columbia). ARA and PFGE with enterococci.
Case 5. Accotink Creek, Blacks Run, and Christians Creek (Virginia). Two-enzyme ribotyping with
E. coli.
Case 6. Avalon Bay (California). Host-specific Bacteroides/Prevotella markers and human- pathogenic
enterovirus detection.
Case 7. Holmans Creek (Virginia). ARA with E. coli.
Case 8. Homosassa Springs (Florida). F+ RNA coliphage genotyping.
Several validation steps have been identified as being essential as part of the design of any new MST study
(refer to Chapter 5 for details). These include precision measurements, positive and negative controls, ex-
ternal validation standards (including known source field samples to test library classification accuracy and
primer specificity), spiked samples (including a matrix spike for PCR on community DNA extracts), and
consideration of independent ancillary data (land use data, sanitary surveys, results by multiple methods,
etc.). When reading these case studies, keep in mind that many of them were initiated several years ago, and
the designers did not have the benefit of what has been learned in subsequent years. With 20/20 hindsight,
it is easy to point out limitations of even the best contemporary studies. Every study could be improved on,
-------
given more time, more money, and better understanding of the approaches. The purpose of this Chapter is
not to criticize, but to learn from the past and to use these practical examples as guides when designing new
source tracking projects.
Case 1. St. Andrews Park (Georgia)
Source of information: Hartel, P., K. Gates, and K. Payne. 2004. Targeted sampling of St. Andrews Park
on Jekyll Island to determine sources of fecal contamination
A. General description
1. Watershed description. Saint Andrews Park is located on the southern tip of Jekyll Island facing
St. Andrews Sound. The park beach is approximately 1.3 km long and is bounded by Beach Creek
at the northern end and the tip of Jekyll Island at the southern end. Previous fecal coliform sampling
of the park suggested that fecal contamination might have originated from a number of locations
north of the park. A sampling of those creeks and pipes emptying into the Jekyll River, which flows
north of the park into St. Andrews Sound, and of the sound itself, was conducted. Several creeks
showed high counts of fecal enterococci. One broken sewer pipe, servicing a local restaurant, was
observed and subsequently repaired.
2. Problem definition. Recently, high numbers of fecal coliforms were observed during beach
monitoring of the park, and these numbers resulted in a beach advisory.
3. Statement of objectives. To use targeted sampling and enterococcal speciation to identify sources of
fecal contamination to St. Andrews Park during calm weather conditions, and, if weather conditions
in the one-month sampling period permit, during stormy weather conditions.
4. Date of study. Completed June 3, 2004
B. Analytical approach
1. Method description. The method chosen was targeted sampling followed by Enterococcus
speciation. Targeted sampling has four steps. The first step is to divide the sampling into two
conditions: base and storm. The second step is to conduct intensive samplirig(s) of the contaminated
waterway, collecting as many samples as possible in one day. Collecting the samples in this manner
reduces temporal variability. The third step is to combine the fecal bacterial numbers with GPS data.
The fourth step is to conduct MST at "hot" areas (i.e., those sites containing relatively high fecal
bacterial numbers). The process is then repeated for storm conditions.
Given the circumstances of St. Andrews Park, with its limited number of potential fecal-source
categories (i.e., humans, pets, and wildlife) and the limited one-month sampling time, the simplest,
quickest, and least expensive MST method was considered to be the one based on Enterococcus
faecalis. In this phenotypic method, enterococci are speciated biochemically and the percentage of
enterococci represented by Ent. faecalis determined. High percentages of Ent. faecalis are associated
with humans and some wild birds (Wheeler et al., 2002).
All confirmed enterococci from Quanti-tray wells were speciated according to a modification of the
Manero and Blanch (1999) protocol. The protocol was modified to identify only three fecal
enterococcal species, Ent. faecalis, Ent. faecium, and Ent. gallinarum. In a further test for the
presence of human-associated Ent. faecalis, approximately 100 isolates were spotted on each of two
0.45-micron membranes on 5-cm Petri plates containing mEI agar (Becton-Dickinson). The plates
were incubated at 41±0.5 °C for 24 hours and were sent by overnight mail to Biological Consulting
Service of North Florida (Gainesville, FL). Their proprietary method (Scott et al., 2005) tests for
the presence or absence of a human-specific factor in enterococci isolates.
-------
2. Target organisms. Enterococci, recovered using Enterolert™ as primary cultivation (enrichment)
medium with recovery and colony isolation using Enterococcosel agar. Confirmed enterococci were
speciated (targeting Ent. faecalis) and tested for a human-specific marker as described above.
3. Statistical approach/classification method. A high proportion of enterococci as Ent. faecalis was
taken to indicate presence of human or avian fecal sources. Presence of human-specific marker used
to differentiate human from avian fecal contamination sources.
C. Sampling considerations
1. Number and frequency of samples. Targeted sampling, twice (21-22 April and 04 May, 2004).
Number of samples not reported but indicated to be about 60 on April 21-22. Fifteen samples
collected 04 May.
2. Type of sample (depth-width integrated or a simple grab). Grab.
3. When collected (season, flow conditions). Spring; one set under calm (low suspended sediments)
and the other under windy (high suspended sediments) conditions
4. Volume of sample and concentration factor. 100 mL analyzed for each sample - Enterococci
colonies evaluated for host-specific factor
5. Evaluation and validation
a. Spiked samples. None reported
b. Blind samples. None reported
c. Negative controls. None reported
d. Comparisons to independent ancillary data. Turbidity, land use pattern for one area (marsh)
D. Outcomes
1. Summary of results and conclusions. During calm weather, highest concentrations of enterococci
were detected in the upper reaches of Beach Creek, the sediments of the creek, and the bathing area.
Species composition in creek sediments and bathing area sediments were different, which was
taken to indicate effects by different enterococci sources. The large proportion of Ent. faecalis in the
upper reaches of Beach Creek was interpreted to implicate wild birds or humans as a source. The
conclusion that wild birds, not humans, were a major source in the upper reaches of Beach Creek
was supported by the marshy character of the area, which makes a human source unlikely at that
location. Though there was no statistical correlation between turbidity and enterococci
concentration, coincidence of high enterococci concentrations and high turbidity in windy weather
was taken as evidence that sediments were a source of elevated water-column numbers during windy
weather.
Human-specific adhesin factor was not detected in any of 200 isolates tested. This was interpreted
as evidence that human sources were not major contributors of enterococci to the test area.
However, the incidence rate of the human-specific marker in enterococci colonizing the human
population is unknown, and there was no mention of a positive control in marker detection by the
research method used, which might limit the interpretability of this result. Human population size,
local approaches to control human waste, or proximity of human residences to the affected area,
factors which were certainly considered in the study, were not mentioned in the report as further
corroborating data.
2. Implementation efforts based on the study. None reported
3. Follow-up monitoring. None reported
-------
Case 2. Tampa Bay (Florida)
Source of information: J.B. Rose, J.H. Paul, M.R. McLaughlin, V. J. Harwood, S. Farrah, M. Tamplin,
J. Lukasik, M. Flanery, P. Stanek and H. Greening. 2000. Healthy Beaches Tampa Bay:
Microbiological monitoring of water quality conditions and public health impacts. Final Project
Report.
A. General description
1. Watershed description. Tampa Bay is located on the west central coast of Florida, opening to the
Gulf of Mexico. This is a shallow subtropical estuary, one of the largest in the southeastern U.S. It
is valued for its ecosystem, fisheries, recreational uses and as a port. The drainage basin is
approximately 2300 square miles and includes 9 major and 76 minor sub-basins. The major
tributaries in the Bay are the Hillsborough, Alafia, Little Manatee and Manatee Rivers, while minor
systems include Alligator Creek, Joe's Creek (Pinellas County), Rocky Creek, Double Branch Creek,
Sweetwater Creek (northwest Hillsborough County), Tampa Bypass Canal, Delaney Creek, Bullfrog
Creek (central and south Hillsborough County), and Frog Creek (Manatee County). Freshwater
inputs are very important to the Bay and are associated with rainfall, with about 60% of the annual
precipitation occurring from June to September. Along with this freshwater input come contaminants
originating from point and non-point sources.
2. Problem definition. Risk to swimmers using polluted beaches has been a major issue associated
with the setting of ambient water quality standards and discharge limits to recreational sites.
Prevention of disease depends on the use of appropriate fecal indicators. However, the finding that
the most widely used fecal contamination indicator, fecal coliforms and more specifically E. coli,
grow naturally on vegetation in warm climates clearly brings into question whether these or other
indicators developed for temperate climates are applicable in Florida and other southeastern areas.
In recent years, total and fecal coliform bacterial indicators have not been able to consistently
indicate the persistence of pathogens, especially viruses in surface waters. F-specific RNA coliphage,
enterococci and Clostridium perfringens have been suggested as better indicators of fecal
contamination and public health risks in tropical and sub-tropical regions.
3. Statement of objectives. This study examined traditional and alternative pollution indicators, as
well as the presence of pathogenic viruses, and their association with environmental variables
(salinity, rainfall, stream flow) in fresh and marine water systems of the Tampa Bay area. The goals
of this study were: 1) to determine appropriate indicators for microbiological water quality for
recreational sites in area beaches and for Tampa Bay; and 2) to determine the occurrence of
pathogens along with indicators in Tampa Bay watersheds and area beaches, their associated sources
(animal vs human), public health risks and potential for management. The final goal of this project
was to form the baseline for other studies and help to develop a long-term strategy for addressing or
enhancing Florida water quality.
4. Date of study. Sampling began in June 1999 and ended in August 2000.
B. Analytical approach
1. Method description. ARA (using a combination of 32 antibiotics and antibiotic concentrations).
Ribotyping was performed by the method of Parveen et al. (1999), using Hindlll. Enterovirus counts
were carried out on human cells lines.
-------
2. Target organisms. Fecal coliforms for ARA, E. coll for ribotyping, enterovirus.
3. Statistical approach/classification method. Library-dependent methods used linear discriminant
analysis. ARA: Classification was performed 6-way (chicken vs. cow vs. dog vs. human vs. pig vs
wild). Ribotyping: Classification was performed 2-way (human vs. animal). Library-independent
method used detection of human-pathogenic enterovirus or Bacteroides fragiUs phage to indicate
presence of human fecal contamination.
C. Library considerations
1. When collected. ARA: Not reported. Ribotyping: A previous isolate collection was used (Parveen
et al., 1997), plus 59 newly-collected isolates.
2. Sources included
a. numbers of samples (reference feces) of each source. ARA: Not reported. Ribotyping: Not
reported.
b. numbers of isolates from each sample (average). ARA: Not reported. Ribotyping: Not
reported.
c. library size. ARA: 3,309 fecal coliform isolates, of which 1,154 are from humans and the
remainder are from chickens, cattle, dogs, pigs and wild animals (mostly wild birds and raccoons).
Ribotyping: 238 isolates (114 human, 124 animal).
3. Evaluation and validation
a. testing for representativeness (cross-validation, holdouts, blind samples). ARA: Not reported.
The ARCC of the ARA library was not reported. Ribotyping: Not reported. The ARCC of the ribo
typing library was 82%
b. testing for random classification. ARA: Not reported. Ribotyping: Not reported.
c. comparisons to independent ancillary data. Compared to other fecal indicators including fecal
coliforms, coliphage, B. fragilis phage, Clostridium and enterococci.
D. Sampling considerations
1. Number and frequency of samples. Twenty-two sites were chosen in Tampa Bay for this study.
The final choices were based on watershed representation, areas of concern in regard to pollution,
accessibility and previously sampled sites. Eleven sites of primarily rural or suburban land use were
chosen in Hillsborough and Manatee counties. Six additional sites were located in highly urban areas,
and 4 beach sites were chosen to represent various types, including urban, heavy boat use,
recreational site in rural area, and pristine unpopulated beach. A control site was located in the middle
of the bay. Each site was sampled monthly for a period of approximately one year for traditional
and alternative fecal indicators, which included fecal coliforms, E. coli, enterococci, C. perfringens
and coliphage. Ten of the sites were chosen for in-depth testing (including antibiotic resistance
analysis of fecal coliform isolates, ribotyping of E. coli isolates, and enterovirus detection). These
sites were monitored 6 times throughout the study.
2. Type of sample (depth-width integrated or a simple grab). Grab samples.
3. When collected (season, flow conditions). Sampling began in June 1999 and ended in August 2000.
4. Number of isolates per sample. ARA: 48. Ribotyping: 1-5.
-------
E. Outcomes
1. Summary of results and conclusions. Perhaps one of the most striking findings of this study is the
extent to which wild animals dominate as a source of fecal coliform and E. coli isolates. Over
the course of the study, wild animal isolates dominated each site according to ARA. Ribotyping
results were consistent; in 74% of all samples (n=53) the majority of isolates were identified as non-
human. However, all sites displayed some level of human fecal pollution according to the three
methods used (ribotyping, ARA and enterovirus counts). The three different methods did not always
coincide on their detection of the presence or absence of human contamination, however the data
collected over the course of the study unambiguously documents the presence of human fecal sources.
Level of agreement among the two library-dependent methods (antibiotic resistance analysis and
ribotyping) and enterovirus counts was assessed for each sampling event. Sites were scored positive
for human impact when >20% of isolates were identified as human by ribotyping and by ARA, and
when any enterovirus counts were detected. Sites were scored negative for human impact when <20%
of isolates were identified as human by ribotyping and by ARA, and when no enterovirus counts
were detected (<1/100 ml). Ribotyping and ARA results agreed for 31 of 53 samples (58%).
Ribotyping and enterovirus results agreed for 29 of 52 (56%) samples. ARA and enterovirus results
agreed most frequently, as positive results at the same sites were noted for 38 of 55 sampling events
(69%). All three methods agreed for 21 of 51 samples (41%). There was no correlation between the
percent of isolates identified as human by ribotyping and enterovirus counts. The Spearman rank cor
relation test (used for non-normally distributed data) showed a significant correlation between the
percent of isolates identified as human by ARA and enterovirus counts (p < 0.05; r = 0.324).
The percentage of isolates identified as human by ARA was significantly con-elated with enterovirus
counts, but the percentage of isolates identified as human by ribotyping was not significantly
correlated with enterovirus counts. This discrepancy points to the need for including the fingerprints
of more isolates from known, local sources in the respective databases. In the case of ARA, we have
seen dramatic improvements in correct classification rates by adding fingerprints from local sources.
The genetic and phenotypic variability of indicator bacteria such as E. coli is quite great, therefore
any information that can be obtained on the fingerprints of actual contamination sources to a
watershed is extremely valuable. Encouragingly, ribotyping, ARA and enterovirus counts agreed on
the presence/absence of human sources in 41% of samples. The probability of the three methods
agreeing by chance alone is 0.125 (0.5 X 0.5 X 0.5), therefore the three methods agree on the
presence of contamination far more frequently than would be predicted by a purely stochastic process.
2. Implementation efforts based on the study. None reported.
3. Follow-up monitoring. Improvements to the databases (ribotyping and ARA) are underway to
increase accuracy.
-------
Case 3. Vermillion River (Minnesota)
Source of information: Sadowsky, M. 2004. "Determination of Fecal Pollution Sources in Minnesota
Watersheds". Technical Report prepared for the Legislative Commission on Minnesota Resources.
A. General description
1. Watershed description. The Vermillion River Watershed encompasses 372 square miles, mostly
located through central Dakota County south of the Twin Cities metropolitan area. The main stem
originates in Scott County to the west and flows generally northeast to the City of Hastings. Current
land use in the watershed is still dominated by agriculture with suburban areas and smaller urban
growth centers interspersed throughout the watershed.
2. Problem definition. In 1998, the Vermillion River main stem, from Empire Township to the dam
in Hastings, was listed on the Federal Clean Water Act's 303(d) list of impaired waters for fecal
coliform bacteria. The river was not meeting its designated use (primary contact - swimming)
standard due to high bacteria levels. Also in 1998, the Vermillion River was placed on the Minnesota
Pollution Control Agency's (MPCA) list of waters in need of a total maximum daily load (TMDL)
assessment for fecal coliform. In 1999 the MPCA, with the help of local agencies and citizens,
collected fecal coliform samples throughout the Vermillion River watershed to begin determining the
extent of the bacterial problem. These data indicate that the river and its tributaries have bacteria
levels in excess of the MPCA's state standard of 200 organisms/100 ml of sample.
3. Statement of objectives. The study was conducted to determine the major sources of fecal pollution
in the watershed.
4. Date of study. April 2001 through December 2003.
B. Analytical approach
1. Method description. HFERP (Horizontal, Fluorophore-Enhanced Rep-PCR.)
2. Target organisms.
3. Statistical approach-classification method. A 4-way analysis was performed (domesticated vs
human vs wildlife vs pets). Each test isolate was assigned to the group of the known-source isolate
with which it had maximum similarity with 1 % optimization using a curve-based (Pearson correlation
coefficient) calculation as applied by BioNumerics software. Robustness of this classification
was evaluated using the custom script ID Bootstrap within BioNumerics, and classifications were
rejected for probabilities less than 90%.
C. Library considerations
1. When collected. July 1999 through November 2002, from known sources from central Minnesota,
Duluth, and the far-western edge of Wisconsin.
2. Sources included
a. numbers of individuals of each source. Cat (37), Chicken (86), Cow (115), Deer (64), Dog (71),
Duck (42), Goat (36), Goose (73), Horse (44), Human (197), Pig (111), Sheep (37), Turkey (69).
-------
b. numbers of unique isolates from each source. Cat (48), Chicken (144), Cow (189), Deer
(96), Dog (106), Duck (81), Goat (42), Goose (135), Horse (78), Human (210), Pig (215),
Sheep (61), Turkey (126).
3. Evaluation and validation
a. testing for representativeness (cross-validation, holdouts, blind samples). Using jackknife
analysis with 1% optimization and maximum similarities using a curve-based (Pearson
correlation coefficient) calculation. The ARCC using this approach was 74%.
b. testing for random classification. None
c. comparisons to independent ancillary data. None.
D. Sampling considerations
1. Number and frequency of samples. Ten sites were sampled along the Vermillion River during
each sampling event. Stream samples were collected on 07/11/01, 08/08/01, 09/05/01, 10/03/01,
03/27/02, 05/01/02, 06/05/02, and 07/02/02.
2. Type of sample (depth-width integrated or a simple grab). Grab samples.
3. When collected (season, flow conditions). Collected from 07/01-07/02. Samples were collected
during periods of high and low flow.
4. Number of isolates per sample. The average number of isolates for each site on each sampling date
was 25.
E. Outcomes
1. Summary of results and conclusions. Identifications indicated that 14% of unknowns matched with
geese, 12% with pigs, 12% with cats, 10% with cows, 9% with human, 9% with deer, 9% with
sheep, and 9% with turkey. The remaining percentages (30%) then fall off to match with the other
groups or remained unclassified. The conclusion was that geese, pigs, cats, cows, humans, deer,
sheep, and turkeys were the dominant sources of fecal pollution in the watershed.
2. Implementation efforts based on the study. None.
3. Follow-up monitoring. None.
-------
Case 4. Anacostia River (Maryland/District of Columbia)
Source of information: Hagedorn, C., K. Porter, and A. H. Chapman. 2003. Bacterial Source Tracking to
Identify Sources of Fecal Pollution in the Potomac and Anacostia Rivers and Rock Creek,
Washington, D.C. Final Project Report.
A. General description
1. Watershed description. The Anacostia River watershed is located in the Maryland counties of
Montgomery (34%) and Prince George (49%), and in the District of Columbia (17%). It is a 456 km2
drainage area and contains 15 km of river (plus an additional 25 km represented by two major
tributaries), with 2% of the land in agricultural use, 28% in forest and park, and 70% in residential
and industrial (urban). The possible/suspected sources of fecal contamination in the Anacostia River
watershed are humans, waterfowl, seagulls and other shore birds, pigeons, starlings, dogs, and cats,
deer, raccoons, muskrats, cattle and horses. The river is a tidal embayment with minimal recharge at
its lower end where it empties into the Potomac River.
2. Problem definition. The Anacostia River does not meet the Clean Water Act national goal of
"fishable or swimmable" standards. It is on the Priority List of impaired waters due to elevated fecal
coliform levels and adversely affected benthic aquatic organisms.
3. Statement of objectives. The study was conducted to determine the major sources of fecal pollution
in the stream, and especially to determine if human fecal pollution was present.
4. Date of study. July 2002 through May 2003.
B. Analytical approach
1. Method description. ARA, using 30 combinations of antibiotic x concentration, and PFGE using the
restriction enzyme Not\.
2, Target organisms. Enterococcus spp.
3. Statistical approach-classification method. Linear discriminant analysis. Classification was per
formed 5-way (bird vs. human vs. livestock vs. pets vs. wildlife).
C. Library considerations
1. When collected. May 2002 through May 2003, from Four-Mile Run (Arlington County, Va), the
Lower Potomac area (Coan and Little Wicomico Rivers), the area around Colonial Beach (Va), and
from the Upper Potomac area (Harper's Ferry to Great Falls), from the Blue Plains Wastewater
Facility, and the greater D.C. area.
2. Sources included
a. numbers of samples of each source. Bird: 40; Human: 31; Livestock: 23; Pets: 52; Wildlife: 22.
b. numbers of isolates from each sample (average). Bird: 6; Human: 12; Livestock: 12; Pets: 6;
Wildlife: 12.
-------
c. library size. ARA= 1,806 isolates (248 bird, 430 human, 699 livestock, 168 pets, and 261
wildlife); PFGE = 750 isolates (150 per source for each of the five sources), all drawn from the
samples in a (above), no more that 8 isolates per sample for the PFGE library (6 or less for
most).
3. Evaluation and validation
a. testing for representativeness (cross-validation, holdouts, blind samples). The ARCC of the
ARA library was 89% and the ARCC of the PFGE library was 93%. The pulled-sample ARCCs
were 74% for ARA and 81% for PFGE. Blind samples were all human isolates, as this was the
most important source in the project. For ARA, the RCC for new sets of human isolates were 70%
at the start of the study (when the library was roughly two-thirds completed), and 79% at the end
of the study with the complete library. The RCC for blind samples with PFGE was 2% to 5%
above the ARA values.
b. testing for random classification. Random rate of classification for ARA was 26%, or about 6%
above the random expectation of 20% for a 5-way classification. Random rate of classification for
PFGE was 24%, or about 4% above the random expectation of 20%.
c. comparisons to independent ancillary data. Seventeen combined sewer outflows (CSOs) are
located on the Washington, D.C. portion of the river (10 km of the 15-km Anacostia River
mainstem are located within the District). The city's NPDES permit allows 2.1 billion gallons of
treated sewage per year to be discharged into the river. This limit is exceeded in most years, but
information regarding the actual dates that the overflows occur or the amounts discharged are not
readily available. What is known is that the discharges are almost always the result of storms and
overflows.
D. Sampling considerations
1. Number and frequency of samples. Six sites were sampled along the Anacostia River. Samples
were collected monthly between July 2002 and May 2003 (10 months). For quality control purposes,
10 duplicate samples were collected, one each month. An additional two sets of samples were taken
immediately after heavy storms, one in the fall and one after a snowmelt in the winter for a total of 82
samples collected on 12 dates.
2. Type of sample (depth-width integrated or a simple grab). Grab samples.
3. When collected (season, flow conditions). Samples were collected during periods of high and low
flow.
4. Number of isolates per sample. 24 for ARA, 8 for PFGE.
£. Outcomes
1. Summary of results and conclusions. The dominant sources over all 10 months of sampling were
(using ARA) birds (31%), wildlife (25%), and humans (24%), followed by pets (20%). Livestock
detections were essentially non-existent. There was a seasonality trend, as bird and wildlife sources
dominated during the low-flow warm weather months (July, August, September, and October),
-------
whereas human and bird sources dominated during the high flow-cold weather months (January,
February, March, and April). Storm events (both in October) elevated the human signature to levels
found during high flow, even though the two storms occurred at the end of the low flow months. A
March snowmelt event elevated the human signature even higher (42.4%), indicating that high flow
events were related to an increased human signature (any major high-flow event in the city results in
sewer overflows). The PFGE water sample results mirrored those from ARA, (except that wildlife
and bird sources were each reduced by an average of 3% to 4% and human was higher by the same
amount), and the two datasets had an r2 value of 82.6%.
2. Implementation efforts based on the study. None.
3. Follow-up monitoring. None.
-------
Case 5. Accotink Creek, Blacks Run, and Christians Creek (Virginia)
Source of information: Hyer, K. E. and D. L. Moyer. 2003. Patterns and Sources of Fecal Coliform
Bacteria in Three Streams in Virginia, 1999-2000. USGS Water-Resources Investigations Report 03-
4115
A. General description
1. Watershed description. Areas of three Virginia streams were chosen for evaluation in the reported
study: Accotink Creek, drainage area 25 mi2, human population greater than 110,000, was primarily
urban; Blacks Run, drainage area 20 mi2, human population about 34,700, was mixed urban and
agricultural; Christians Creek, drainage area 107 mi2, human population about 12,000, was primarily
agricultural. Extensive base-flow, event-flow, and continuum sampling was done in each watershed
over a period of 20 months. Microbial source tracking by use of ribotyping was performed on E. coli
isolates collected at a single, state-determined water-quality compliance point for each watershed.
2. Problem definition. Surface-water impairment by fecal coliform bacteria is a water-quality issue
of national scope and importance. In Virginia, more than 175 stream segments are on the
Commonwealth's 1998 303(d) list of impaired waters because of elevated concentrations of fecal
coliform bacteria. In Virginia, total maximum daily load (TMDL) assessments will need to be
developed over the next 10 years for all impaired water bodies identified on the State's 1998 303(d)
list. Establishment of TMDLs in waters contaminated by fecal coliform bacteria is difficult because
the potential sources of the bacteria are numerous and the magnitude of their contributions is
commonly unknown. Potential sources of fecal coliform bacteria include all warm-blooded animals
(humans, pets, domesticated livestock, birds, and wildlife). The lack of information on bacteria
sources makes it difficult to develop accurate load allocations, technically defensible TMDLs, and
appropriate source-load reduction measures. Information about the major fecal coliform sources that
impair surface-water quality would represent an improvement in the development of technically
defensible TMDLs.
3. Statement of objectives. This study was performed to demonstrate the field application of a BST
method and to identify the sources of fecal coliform bacteria in three streams on Virginia's 1998
303(d) list of impaired waters. The three streams sampled during this study were selected because
they represent a range of land uses (urban, agricultural, and mixed urban/agricultural) and most of the
potential fecal coliform sources that are likely to be encountered throughout the Commonwealth.
4. Date of study. 1999-2000.
B. Analytical approach
1. Method description. The known-source E. coli reference collection of Dr. Mansour Samadpour
(Institute for Environmental Health, Seattle Washington; more than 50,000 isolates at the time) was
used and supplemented by known-source samples in the studied watersheds. Isolates were
characterized by ribotyping using restriction enzymes EcoRl and Pbull.
2. Target organisms. E. coli
3. Statistical approach/classification method. 1:1 matching. The approach used in this study was that
ribotypes (strains) of E. coli are specific to host species; therefore, any stream-isolated E. coli with
-------
a ribotype that matched a known-source isolate could be assigned to that host species as the source.
Where there was a match to one source, isolates were classified to that source. Where there was a
match to more than one source, isolates were classified as transient. Where there was no match in the
library, isolates were classified as unknown.
C. Library considerations
1. When collected. Three sets of isolates were used as the known-source library: 1) 50,000 isolates
from the IEH collection, collected over approximately 5-10 years prior to the current study, national
coverage; 2) 450 isolates previously collected in Virginia, many by George Simmons, in the
approximately 5-10 years prior to the current study; 3) 723 isolates collected in the three watersheds
concurrently with water sample collection.
2. Sources included
a. numbers of samples of each source. Though the distribution of samples among hosts was not
noted in the manuscript, 723 source samples were collected during the study from a humans, pets,
domestic animals, and wildlife.
b. numbers of isolates from each sample (average). One.
c. library size. The overall known-source library comprised more than 50,000 isolates.
3. Evaluation and validation
a. testing for representativeness (cross-validation, holdouts, blind samples). 23 isolates from the
known-source library were re-submitted as 66 blinds (some were submitted as duplicates or
triplicates). The lab had prior knowledge of which 23 isolates were being used. Blind isolates
were re-analyzed and matched in all cases to the correct identity among the 23 isolates used.
b. testing for random classification. None.
c. comparisons to independent ancillary data. Multiple lines of evidence were used to evaluate
whether MST results were reasonable in these study streams. The authors began by evaluating
populations and distributions of known fecal sources, and land-use patterns in each watershed.
They conducted continuum sampling to evaluate longitudinal trends in fecal-indicator
concentrations in the main stem, in tributaries, and in effluents discharged to the main stem. They
also evaluated seasonal and flow-related trends in fecal-indicator concentrations. These data were
interpreted in terms of transport pathways and animal distributions in the watersheds to indicate
expected sources of fecal-indicator bacteria.
Several quality control elements were considered to evaluate the interpretation of MST data in this
study, and provided further information about some unexpected results. The unexpectedly high
contribution by waterfowl in the urban Accotink Creek watershed was consistent with the results
of a prior study in a neighboring urban watershed, Four Mile Run (Simmons et al., 1999).
Contributions of bacteria from human sources were independently evaluated by sampling for
wastewater organics compounds. In all three streams, detectable concentrations of caffeine and
cotinine were present, consistent with MST-indicated contributions of human wastewater to the
-------
streams. The interpretation that poultry waste was in Christians Creek was supported by total
arsenic data collected by Hancock et al. (2000). The poultry feed amendment Roxarsone contains
arsenic, which is generally excreted by the birds. Arsenic-bearing poultry litter is ultimately land-
applied on the surrounding agricultural fields. Total arsenic concentrations increased during a
storm event, supporting the hypothesis that field-applied poultry litter was flushed into streams.
D. Sampling considerations
1. Number and frequency of samples. Between 400 and 450 water-isolated E. coli were evaluated
for each of three watersheds. Samples were taken on two schedules - routine monitoring samples
(2/3 of samples) were collected approximately every 6 weeks and event-oriented samples, targeted at
storm flow, were collected as available (5 events, 1/3 of samples). For routine monitoring, 4-8 sam
pies were collected 5 minute intervals to represent small-scale variability in concentration and sourc
es. For event-oriented samples, 10 samples were collected across the hydrograph to represent small-
scale variability in concentration and sources during rain events.
2. Type of sample (depth-width integrated or a simple grab). Depth-width integrated samples using
three depth-integrated transits (routine monitoring) and grab samples from the centroid of flow (storm
flow samples)
3. When collected (season, flow conditions). Samples were collected for 20 months over all seasons.
Of the samples, 61% were taken during low-flow condition, 39% during storm-flow condition. Storm
samples were collected across the hydrograph (10 samples).
4. Number of isolates per sample. 3-5 per water sample. Multiple samples from the same sites on the
same dates were not composited.
E. Outcomes
1. Summary of results and conclusions. Overall, about 65% of isolates could be assigned to a source
in this study. Of the remaining 35%, some had no match in the library (unknown) and others matched
to multiple sources (transient). Classification was made to the species level with some exceptions (for
example, some bird-origin feces could be classified to species, but others could only be classified to
"avian" or "poultry"). The MST results were a combination of the expected and the unanticipated.
Fecal-indicator sources in Accotink Creek, the urban setting, were affected by human and pet feces, as
expected, but were also strongly influenced by waterfowl. Blacks Run fecal-indicator bacteria were
a mixture of human, pet, and livestock sources, as expected. Fecal-indicator concentrations in
Christians Creek had a larger human and pet component than expected (about 25% of isolates),
compared with livestock and poultry (about 50%). A further unexpected finding in all three water
sheds was that relative contributions from each major source were about the same during both
base-flow and storm-flow periods, despite the expectation that different transport pathways would
dramatically change relative contributions from different sources. Lastly, the study detected seasonal
patterns in the contributions of bacteria from cattle and poultry sources in Blacks Run and Christians
Creek; this seasonal pattern was consistent with the land management strategies used in each
watershed.
2. Implementation efforts based on the study. Volunteer implementation along with cost share
implementation in support of the TMDL document. Exclusion fencing of cattle was one of the major
implementation efforts.
-------
3. Follow-up monitoring. Based on the results of this initial study, DEQ developed and submitted a
TMDL to the USEPA in 2002 that included a goal to reduce the human sources of fecal coliform
bacteria by 99%. The TMDL for Accotink Creek was approved by USEPA in July 2002. As a follow-
up step to the TMDL, USGS initiated another study in cooperation with Fairfax County Stormwater
Planning Division (SWPD), City of Fairfax, and DCR to help identify the distribution of fecal
coliform and locate the precise sources of human fecal coliform inputs to Accotink Creek. This
second study began in mid-to-late 2001 and will continue for 3 years. The field-work portion of the
study is anticipated to be completed in late 2004. Staff from SWPD is currently assisting USGS field
sampling efforts and laboratory analysis for some parameters.
-------
Case 6. Avalon Bay (California)
Source of information: Boehm, A. B., Fuhrman, J. A., Mrse, R. D. and Grant, S. B. 2003. Tiered approach
for identification of a human fecal pollution source at a recreational beach: Case study at Avalon Bay,
Catalina Island, California. Environ. Sci. Tech. 37(4), 673-680.
A. General description
1. Watershed description. The impacted coastline is a 500-m stretch of sandy beach located in Avalon
Bay, on the southeast side of Catalina Island, California (area 200 km2). Avalon (area 6.9 km2)
is the largest town on the island with 3500 year-round residents. The city's primary source of revenue
is tourism; on a typical summer day 17,500 tourists arrive via ferry, cruise ship, or personal vessel,
and up to 400 vessels are anchored in the bay. Rainfall in this region occurs primarily from
November through March, and consequently, during the summer-time study, there is no rainfall. As
is the case for virtually any coastal community, there are many potential sources of fecal
contamination in Avalon Bay. Sewer trunk lines run parallel to the beach, approximately 20 m from
the shoreline. Nuisance runoff is directed into the sewer system by low-flow diverters; however, some
of the runoff enters the ocean untreated through small drains that discharge to the sand, particularly
during periods when streets are being washed down by City staff. Secondary treated sewage is
released at a rate of approximately 2158 m3 d"1 southeast of the bay through an outfall that terminates
100 m from the coast, at a depth of 65 m. A pier with restrooms, restaurants, and recreational
establishments extends from the shoreline near the southeast end of the beach. In addition, pigeons
and sea gulls congregate to feed and nest near the shoreline.
2. Problem definition. During the summers of 2000 and 2001, water samples from Avalon Beach
frequently exceeded the single sample standard for enterococci; thus, signs were posted at the beach
warning swimmers not to enter the water. Based on historical data, this was not necessarily a new
problem, but was magnified with the new, more stringent state water quality regulations that were
instated in the summer of 1999.
3. Statement of objectives. City officials were not able to readily identify and remedy the pollution
source, and thus the study was commissioned. At the outset of the study, it was not clear to what
extent the following potential sources impacted water quality in Avalon Bay: effluent from the sewage
treatment plant, nuisance runoff, feces of birds and other wild animals, contaminated subsurface
water, and boat sewage collection tanks. The latter was not expected to contribute much to the
pollution because the city has an aggressive dye program to reduce illicit discharges into the bay.
B. Analytical approach
1. Method description. A three-tiered approach for determining sources of human and nonhuman fecal
indicator bacteria (FIB) at a recreation beach that utilizes both standard assays for FIB and novel
detection techniques for human-specific Bacteroides/Prevotella and enterovirus. The first tier
documents the spatiotemporal variability of the pollution signal and takes into account the possible
influence of sunlight and tides on FIB concentrations in coastal waters. The second tier consists of
source studies. Studies in the first two tiers identify pollution sources and "hot spots" using only
standard FIB tests. The third and final tier consists of selectively sampling FIB sources and hot spots
for the enteric bacteria Bacteroides/Prevotella and enterovirus using nucleic acid detection techniques
to determine if fecal contamination, indicated by FIB, is of human origin. This study illustrates how
-------
measurements made with traditional indicators, in conjunction with more novel indicators, can lead to
source identification and mitigation.
2. Target organisms. Bacteroides/Prevotella and enterovirus.
3. Statistical approach-classification method. Based on presence/absence of host-specific PCR
product. Sensitivity of the Bacteroides/Prevotella method was estimated at 1 ug/5-50 mL of seawater.
Detection limit of the enterovirus method was approximately 1 PFU per 2-20 L of seawater.
C. Sampling considerations
1. Number and frequency of samples. 33 samples, collected between 9/19/2001 and 10/29/2001.
2. Type of sample (depth-width integrated or a simple grab). Grab samples.
3. When collected (season, flow conditions). Summer, no rainfall events included.
4. Volume of sample and concentration factor. For Bacteroides/Prevotella, bacteria from water
samples were collected by filtration of 1-4 L. Most amplifications were perform using 1 and 10 ng
of extracted DNA, equivalent to about 5-50 mL of seawater, chosen to provide a compromise between
sensitivity and inhibition of the assay. For enterovirus, 2-20 L of water was filtered.
5. Evaluation and validation
a. Spiked samples. All sets of assays included positive controls in which a small amount (1-100 pg)
of human fecal DNA extract or cultured poliovirus was added to replicates of the field samples to
see if reactions were inhibited by the matrix.
b. Blind samples. Not done.
c. Negative controls. All sets of assays included negative controls (no DNA or poliovirus added).
d. Comparisons to independent ancillary data. Source tracking was performed on samples from
locations which were identified using the first two tiers of the procedure.
D. Outcomes
1. Summary of results and conclusions. FIB in Avalon Bay appear to be from multiple, primarily land-
based, sources including bird droppings, contaminated subsurface water, leaking drains, and runoff
from street wash-down activities. Multiple shoreline samples and two subsurface water samples
tested positive for human-specific bacteria and enterovirus, suggesting that at least a portion of the
FIB contamination is from human sewage.
2. Implementation efforts based on the study. Based on the results of the study, the city of Avalon
slip-lined their sewer lines that run along the beach.
3. Follow-up monitoring. Not mentioned in report.
-------
Case 7. Holmans Creek (Virginia)
Source of information: Noto, M., K. Hoover, E. Johnson, J. McDonough, E. Stevens, and B. A.
Wiggins. 2000. "Use of Antibiotic Resistance Analysis (ARA) to Identify Nonpoint Sources of Fecal
Contamination in the Holmans Creek Watershed". Technical Report prepared for the Lord Fairfax
Soil and Water Conservation District.
A. General description
1. Watershed description. Holmans Creek is located in Shenandoah County, Virginia. It is a 11,988
acre drainage area and contains 12 miles of stream, with 72% of the land in agricultural use, 26%
forested, and 2% mixed urban land use. All of the homes use septic systems and wells or cisterns.
Holmans Creek feeds the North Fork of the Shenandoah River and flows eventually into the
Chesapeake Bay. The possible/suspected sources of fecal contamination in the Holmans Creek
watershed are beef and dairy cattle (cattle), chickens and turkeys (poultry), failing septic systems
(human), and geese.
2. Problem definition. Holmans Creek does not meet the Clean Water Act national goal of "fishable or
swimmable" standards. It is on the Priority List of impaired waters due to elevated fecal coliform
levels and adversely affected benthic aquatic organisms.
3. Statement of objectives. The study was conducted to determine the major sources of fecal pollution
in the stream.
4. Date of study. July 1999 through January 2001.
B. Analytical approach
1. Method description. ARA, using 16 antibiotics (51 concentrations total).
2. Target organisms. Enterococci.
3. Statistical approach-classification method. Linear discriminant analysis. Classification was
performed 4-way (cattle vs poultry vs human vs geese).
C. Library considerations
1. When collected. July 1999 through January 2001, from known sources located within the watershed.
2. Sources included
a. numbers of samples of each source. Cattle (3 animals/sample): 26; Poultry litter (multiple
animals/sample): 11; Septic tanks (1 household/sample): 42; Geese (3 animals/sample): 7.
b. numbers of isolates from each sample (average). Cattle: 18; Poultry: 23; Septic tanks 19;
Geese 14.
3. Evaluation and validation
a. testing for representativeness (cross-validation, holdouts, blind samples). The ARCC of the
library was 73%. The Minimum Detectable Percentage (MDP) for each source type was
-------
determined to be 18% by averaging the percentages of other source types that were misclassified as
that type. Further representativeness sampling was not done at the time, but subsequent cross-
validation and holdout analysis showed that the library was reasonably representative for human
and livestock sources, but was not representative for the wild (goose) samples.
b. testing for random classification. Not done.
c. comparisons to independent ancillary data. See section G.
D. Sampling considerations
1. Number and frequency of samples. Nine sites were sampled along Holmans Creek during each
sampling event. Stream samples were collected on 7/23/99, 9/29/99, 11/18/99, 2/15/00, 2/19/00 (after
a heavy storm), 7/20/00, 9/20/00, and 1/25/01.
2. Type of sample (depth-width integrated or a simple grab). Grab samples.
3. When collected (season, flow conditions). Collected over a year and a half. Samples were collected
during periods of high and low flow. One set of samples were taken immediately after a heavy storm.
4. Number of isolates per sample. The goal for each sample was 46 isolates, but some samples had
fewer. The average number of isolates per sample was 41.
E. Outcomes
1. Summary of results and conclusions. Human sources were dominant in five of eight sampling
events, and at four of nine locations. In 53 of the 64 samples, the proportion of human was above the
MDP, and human was the dominant source in 29 of the 64 samples. Cattle was the dominant source on
three of eight sampling days, and at five of nine locations. The proportion of cattle was above the
MDP in 52 of 64 samples, and cattle was the dominant source in 26 of them. Poultry and geese fecal
contributions were low throughout the sampling period. The conclusions were that humans and cattle
are the dominant sources of fecal pollution in the watershed.
2. Implementation efforts based on the study. Based on the results of this study, a septic system
maintenance project was undertaken in the watershed. This project identified numerous straight pipes
discharging sewage directly into the stream, and found that approximately 25% of the septic systems
in the watershed were failing. Through the use of cost-share funds, many of these systems have been
repaired or replaced. The Implementation Plan for this watershed calls for removal of all straight
pipes, all failing septic systems must be identified and corrected, and all livestock must be excluded
from the stream.
3. Follow-up monitoring. Stream monitoring in this watershed has been continuing. Samples from
the same sites have been collected quarterly from 2002 - 2004. The results from the newer sample
indicate that the percentage of human pollution has decreased from the 2001 levels. Subsequent
classification of the samples was performed using a larger, regional library that was determined to be
representative for all sources (using cross-validation, holdouts, and random classification).
-------
Case 8. Homosassa Springs (Florida)
Source of information: Griffin, D. W., R. Stokes, J. B. Rose, and J. H. Paul III. 2000. Bacterial indicator
occurrence and the use of F+ specific RNA coliphage assay to identify fecal sources in Homosassa
Springs, Florida. Microbial Ecology 39:56-64.
A. General description
1. Watershed description. The Homosassa Springs State Wildlife Park (HSSWP) is a 180-acre
complex that surrounds Homosassa River's main spring (Homosassa Main). HSSWP is the home of
numerous animals including birds, deer, bobcats, alligators, a hippopotamus, a permanent group of
manatees, and fish. The Homosassa Main consists of three separate vents, each with its own distinct
chemical signature, which have a combined average discharge of approximately 2,944 liters s'1. To the
southeast of the park is the Southeast Fork of the Homosassa River. The Southeast Fork is fed by a
closely associated group of springs, which have a combined average discharge of approximately 1,953
liters s"1. The waters of these two sources and the immediate region in the river receiving these waters
appear clear.
2. Problem definition. Water quality issues in the Homosassa River system have received the attention
of local citizen groups and the media. Of particular concern were the elevated levels of coliforms
and fecal coliforms found in Homosassa River downstream of HSSWP, which have been attributed
to Park animals. The Florida Department of Health (DOH), which has been monitoring water quality
at a site just downstream of the park (an area which was to have been designated as a swimming site),
found that fecal indicator concentrations consistently exceed recreational use standards (> 200 fecal
coliform colony forming units (CPU) 100 ml"1).
3. Statement of objectives. This study was designed to assess microbial water quality and to
differentiate fecal sources contributing to the contamination previously observed in HSSWP and its
adjacent waters.
4. Date of study. November of 1997 and November of 1998.
B. Analytical approach
1. Method description. F+ specific RNA coliphage genotyping. Types 11 and 111 coliphage are
associated with human sources of fecal contamination and Types I and IV are associated with non-
human sources.
2. Target organisms. F+ specific RNA coliphage.
3. Statistical approach/classification method. Direct match of specific oligoriucleotide probes.
C. Sampling considerations
1. Number and frequency of samples. Seven sites in November 1998 and nine sites in November
1998.
2. Type of sample (depth-width integrated or a simple grab). Grab samples.
-------
3. When collected (season, flow conditions). November of two consecutive years.
4. Volume of sample and concentration factor 20-L samples concentrated by vortex flow filtration
(>70% coliphage recovery) to 40-60 ml, of which 1 ml aliquots were used for coliphage analysis.
5. Evaluation and validation
a. Spiked samples. None reported. No reference feces from local animals were positive for F+
RNA coliphage. A reference human-waste stream was positive for human-associated types II and
III coliphage.
b. Blind samples. None reported.
c. Negative controls. None reported.
d. comparisons to independent ancillary data. Several factions have attributed the fecal indicator
prevalence to HSSWP animals. The watershed also contains many residences with older septic
tanks.
D. Outcomes
1. Summary of results and conclusions. F+ specific RNA coliphage analysis indicated that fecal
contamination at all sites that had F+ RNA coliphage was from animal sources (mammals and birds).
These results suggest that animal (either indigenous or residents of HSSWP) and not human sources
influenced microbial water quality in the area of Homosassa River covered by this study.
2. Implementation efforts based on the study. None reported.
3. Follow-up monitoring. None reported.
-------
Literature Cited
Alderisio, K. A., D. A. Wait, and M. D. Sobsey. 1996. Detection and characterization of male-specific RNA
coliphages in a New York City reservoir, p. 133-142, In Watershed Restoration Management, J. J. McDon-
nell, D. L. Leopold, J. B. Stribling, and L. R. Neville (ed.), New York City Water Supply Studies. American
Water Resources Association, Herndon, VA.
Altwegg, M., F. W. Hickman-Brenner, and J. J. Farmer 111. 1989. Ribosomal RNA gene restriction patterns
provide increased sensitivity for typing Salmonella typhi strains. J. Infect. Dis. 160:145-149.
American Public Health Association. 1995. In Standard Methods for the Examination of Water and Wastewa-
ter. Washington DC: American Public Health Association, Inc.
Anderson, M. A. 2003. Frequency distributions of Escherichia coli subtypes in various fecal sources over
time and geographical space: Application to bacterial source tracking methods, pp. 117. Tampa, FL: Univer-
sity of South Florida.
Anderson, M. A., J. E. Whitlock, and V. J. Harwood. 2003. Frequency distributions of Escherichia coli sub-
types in various fecal sources: Application to bacterial source tracking methods. American Society for Micro-
biology General Meeting. Washington, DC.
Anderson, S. A., S. J. Turner, and G. D. Lewis. 1997. Enterococci in the New Zealand environment: implica-
tions for water quality monitoring. Water Sci. Techno/. 35:325-331.
Arnold, C., L. Metherell, J. P. Clewley, and J. Stanley. 1999. Predictive modelling of fluorescent AFLP: a
new approach to the molecular epidemiology of E. coli. Res. Microbiol. 150:33-44.
Aslam, M., F. Nattress, G. Greer, C. Yost, C. Gill, and L. McMullen. 2003. Origin of contamination and ge-
netic diversity of Escherichia coli in beef cattle. Appl. Environ. Microbiol. 69:2794-2799.
Ausubel, F. M., R. Brent, R. E. Kingston, D. D. Moore, J.G. Seidman, J. A. Smith, and K. Struhl. 2004. Cur-
rent Protocols in Molecular Biology. John Wiley and Sons, Hoboken, NJ.
Avelar, K. E., S. R. Moraes, L. J. Pinto, W. d. G. Silva e Souza, R. M. Domingues, and M. C. Ferreira.
1998. Influence of stress conditions on Bacteroides fragilis survival and protein profiles. Zentralbl Bakteriol
287:399-409.
Backhed F, R. E. Ley, J. L. Sonnenburg, D. A. Peterson, and J. I. Gordon. 2005. Host-bacterial mutualism in
the human intestine. Science 307:1915-1920.
Baker, G. C., J. J. Smith, and D. A. Cowan. 2003. Review and re-analysis of domain-specific 16S primers.
J. Microbiol. Meth. 55:541-555.
Bartosch, S., A. Fite, G. T. Macfarlane, and M. E. T. McMurdo. 2004. Characterization of bacterial commu-
nities in feces from healthy elderly volunteers and hospitalized elderly patients by using real-time PCR and
effects of antibiotic treatment on the fecal microbiota. Appl. Environ. Microbiol. 70:3575-3581.
-------
Bass, L., C. A. Liebert, M. D. Lee, A. O. Summers, D. G. White, S. G. Thayer and J. J. Maurer. 1999. Inci-
dence and characterization of integrons, genetic elements mediating multiple-drug resistance, in avian Esch-
erichia coli, Antimicrob, Agents Chemother. 43: 2925-2929.
Beekwilder, J., R. Nieuwenhuizen, A. H. Havelaar, and J. van Duin. 1996. An oligonucleotide hybridization
assay for the identification and enumeration of F-specific RNA phages in surface water. /. Appl. Bacterial.
80:179-186.
Bernhard, A. E., T. Goyard, M. T. Simonich, and K. G. Field. 2003. Application of a rapid method for identi-
fying fecal pollution sources in a multi-use estuary. Water Res. 37:909-913.
Bernhard, A.E., and K.G. Field. 2000a. Identification of nonpoint sources of fecal pollution in coastal wa-
ters by using host-specific 16S ribosomal DNA genetic markers from fecal anaerobes. Appl. Environ. Micro-
biol. 66:1587-1594.
Bernhard, A.E., and K.G. Field. 2000b. A PCR assay to discriminate human and ruminant feces on the basis
of host differences \nBacteroides-Prevotella genes encoding 16S rRNA. Appl. Environ. Microbiol. 66:4571-
4574.
Bingen, E., E. Denamur, N. Brahimi, and J. Elion. 1996. Genotyping may provide rapid identification of
Escherichia coli Kl organisms that cause neonatal meningitis. Clin. Infect. Dis. 22:152-156.
Bingen, E. H., E. Denamur, B. Picard, P. Goullet, N.Y. Lambert-Zechovsky, N. Brahimi, J.-C. Mercier, F.
Beaufils and J. Elion. 1992. Molecular epidemiology unravels the complexity of neonatal Escherichia coli
acquisition in twins. J. Clin. Microbiol. 30:1896-1898.
Blears, M. J., S. A. De Grandis, H. Lee, and J. T. Trevors. 1998. Amplified fragment length polymorphism
(AFLP): a review of the procedure and its applications. J. Indust. Microbiol. Biotech. 21:99-114.
Boehm, A. B., J. A. Fuhrman, R. D. Mrse, and S. B. Grant. 2002. Tiered approach for identification of a
human fecal pollution source at a recreational beach: Case study at Avalon Bay, Catalina Island, California.
Environ. Sci. Technol. 37:673-680.
Bonjoch, X., E. Balleste and A. R. Blanch. 2004. Multiplex PCR with 16S rRNA gene-targeted primers of
Bifidobacterium spp. to identify sources of fecal pollution. Appl. Environ. Microbiol. 70: 3171-3175.
Booth, A. M., C. Hagedorn, A. K. Graves, S. C. Hagedorn and K. H. Mentz. 2003. Sources of fecal pollution
in Virginia's Blackwater River. J. Environ. Eng. 129: 547-552.
Brion, G. M., J. S. Meschke, and M. D. Sobsey. 2002. F-specific RNA coliphages: occurrence, types, and
survival in natural waters. Water Res. 36:2419-2425.
Brisse, S., C. M. Verduin, D. Milatovic, A. Fluit, J. Verhoef, S. Laevens, P. Vandamme, B. Tummler, H. A.
Verbrugh, and A. van Belkum. 2000. Distinguishing species of the Burkholderia cepacia complex and Burk-
holderia gladioli by automated ribotyping. J. Clin. Microbiol 38:1876-1884.
Bryan, A., N. Shapir, and M. J. Sadowsky. 2004. Frequency and distribution of tetracycline resistance genes
in genetically diverse, nonselected, and nonclinical Escherichia coli strains isolated from diverse human and
animal sources. Appl Environ. Microbiol 70:2503-2507.
-------
Bryant, M. P. 1959. Bacterial species of the rumen. Bacterial. Rev. 23:125-153.
Burton, G. A., Jr., Gunnison, D. and Lanza, G. R. 1987 Survival of pathogenic bacteria in various freshwater
sediments. Appl. Environ. Microbiol. 53:633-638.
Byappanahalli, M., D. A. Shively, M. B. Nevers, M. J. Sadowsky, and R. L. Whitman. 2003. Growth and sur-
vival of Escherichia coli and enterococci populations in the macro-alga Cladophora (Chlorophyta). FEMS
Microbiol. Ecol. 46:203-211.
Byappanahalli, M. and R. Fujioka. 2004 Indigenous soil bacteria and low moisture may limit but allow fae-
cal bacteria to multiply and become a minor population in tropical soils. Water Sci. Techno!. 50:27-32.
Caetano-Anolles, G., B. J. Bassam, and P. M. Gresshoff. 1992. Primer-template interactions during DNA
amplification fingerprinting with single arbitrary oligonucleotides. Mol. Gen. Genet. 235:157-165.
Carrillo, M., E. Estrada, and T. C. Hazen. 1985. Survival and enumeration of the fecal indicators Bifidobacte-
rium adolescentis and Escherichia coli in a tropical rain forest watershed. Appl Environ. Microbiol. 50:468-
476.
Carson, A. C., B. L. Shear, M. R. Ellersieck, and A. Asfaw. 2001. Identification of fecal Escherichia coli
from humans and animals by ribotyping. Appl. Environ. Microbiol. 67:1503-1507.
Carson, C. A., B. L. Shear, M. R. Ellersieck, and J. D. Schnell. 2003. Comparison of ribotyping and repeti-
tive extragenic palindromic-PCR for identification of fecal Escherichia coli from humans and animals. Appl.
Environ. Microbiol. 69:1836-1839.
Caugant, D. A., Levin, B. R. and Selander, R. K. 1984. Distribution of multilocus genotypes of Escherichia
coli within and between host families. J. Hyg. (Land) 92:377-384.
Chern, E. C., Y. L. Tsai, and B. H. Olson. 2004. Occurrence of genes associated with enterotoxigenic and
enterohemorrhagic Escherichia coli in agricultural waste lagoons. Appl. Environ. Microbiol. 70: 356-362.
Chivukula, V. and V. J. Harwood. 2004. Impact of fecal pollution on the microbial diversity in natural waters.
American Society for Microbiology General Meeting. New Orleans, LA.
Cho, J.C., and S.J. Kim 2000. Increase in bacterial community diversity in subsurface aquifers receiving
livestock wastewater input. Appl. Environ. Microbiol 66: 956-965.
Choi, S., W. Chu, J. Brown, S. J. Becker, V. J. Harwood and S. C. Jiang. 2003. Application of enterococci
antibiotic resistance patterns for contamination source identification at Huntington Beach, California. Mar.
Poll. Bull. 46:748-755.
Chu, G., D. Vollrath and R. W. Davis. 1986. Separation of large DNA molecules by contour-clamped homo-
geneous electric fields. Science 234:1582-1585.
Clerc, A., C. Manceau, and X. Nesme. 1998. Comparison of randomly amplified polymorphic DNA with am-
plified fragment length polymorphism to assess genetic diversity and genetic relatedness within genospecies
III of Pseudomonas syringae. Appl. Environ. Microbiol. 64:1180-1187.
-------
Clesceri, L. S., A. E. Greenberg, and A. D. Eaton. 1998. Standard methods for the examination of water and
waste water. APHA AWWA WEF, Washington, DC.
Cole, D., S. C. Long, and M. D. Sobsey. 2003. Evaluation of F+ RNA and DNA coliphages as source-spe-
cific indicators of fecal contamination in surface waters. Appl. Environ. Microbiol. 69:6507-6514.
Cox, T. F., and M. A. A. Cox. 2001. Multidimensional Scaling, 2nd ed., Chapman and Hall.
Craig, D. L., H. J. Fallowfield, and N. J. Cromar. 2004. Use of microcosms to determine persistence of
Escherichia coli in recreational coastal water and sediment and validation with in situ measurements. J. Appl.
Microbiol. 96:922-930.
Daley, K., and S.P. Shirazi-Beechey. 2003. Design and evaluation of group-specific oligonucleotide probes
for quantitative analysis of intestinal ecosystems: their application to assessment of equine colonic micro-
flora. FEMS Microbiol. Ecol. 44:243-252.
Dargatz, D. A., P. J. Fedorka-Cray, S. R. Ladely, C. A. Kopral, K. E. Ferris, and M. L. Headrick. 2003.
Prevalence and antimicrobial susceptibility of Salmonella spp. isolates from US cattle in feedlots in 1999 and
2000. J. Appl. Microbiol. 95:753-761.
Dasarathy, B. V. 1991. Nearest Neighbor Pattern Classification Techniques. IEEE Computer Society Press.
Los Alamitos, CA
de Bruijn, F. J. 1992. Use of repetitive (repetitive extragenic palindromic and enterobacterial repetitive in-
tergeneric consensus) sequences and the polymerase chain reaction to fingerprint the genomes ofRhizobium
meliloti isolates and other soil bacteria. Appl. Environ. Microbiol. 58:2180-2187.
DeLong, E.F., G.S. Wickhan, and N.R. Pace. 1989. Phylogenetic strains: Ribosomal RNA-based probes for
the identification of single cells. Science. 243:13600-1363.
Demezas, D. 1998. Fingerprinting bacterial genomes using restriction fragment length polymorphisms, pp.
383-398. In Bacterial Genomes: Structure and Analysis., F. J. de Bruijn, J.R. Lupski, and G. Weinstock (ed.).
Chapman and Hall, New York.
de Motes, C. M., P. Clemente-Casares, A. Hundesa, M. Martin, and R. Girones. 2004. Detection of bovine
and porcine adenoviruses for tracing the source of fecal contamination. Appl. Environ. Microbiol. 70:1448-
1454.
Desmarais, T. R., Solo-Gabriele, H. M. and Palmer, C. J. 2002 Influence of soil on fecal indicator organisms
in a tidally influenced subtropical environment. Appl. Environ. Microbiol. 68:1165-1172.
Dice, L. R. 1945. Measures of the amount of ecological association between species. Ecology 26:297-302
Dick, L. K., A. E. Bernhard, T. J. Brodeur, J. W. Santo Domingo, J. M. Simpson, S. P. Walters, and K. G.
Field. 2005a. Host distributions of uncultivated fecal Bacteroidales bacteria reveal genetic markers for fecal
source identification. Appl. Environ. Microbiol. 71:3184-3191.
-------
Dick, L. K., M. T. Simonich, K. G. Field. 2005b. Microplate subtractive hybridization to enrich for Bacteroi-
dales genetic markers for fecal source identification. Appl. Environ. Microbiol 71:3179-3183.
Dick, L. K., and K. G. Field. 2004. Rapid estimation of numbers of fecal Bacteroidetes by use of a quantita-
tive PCR Assay for 16S rRNA genes. Appl. Envir. Microbiol. 70: 5695-5697.
Dombek, P. E., L. K. Johnson, S. T. Zimmerley, and M. J. Sadowsky. 2000. Use of repetitive DNA sequences
and the PCR to differentiate Escherichia coli isolates from human and animal sources. Appl. Environ. Micro-
biol. 66:2572-2577.
Dontchev, M., J. E. Whitlock, and V. J. Harwood. 2003 Ribotyping of Escherichia coli and Enterococcus
spp. to determine the source of fecal pollution in natural waters. American Society for Microbiology General
Meeting. Washington, DC.
Dufour, A. P. 1984. EPA health effects criteria for fresh recreational waters. Office of Research and Develop-
ment, United States Environmental Protection Agency, Research Triangle Park, North Carolina.
Eckburg, P. B., E. M. Bik, C. N. Bernstein, E. Purdom, L. Dethlefsen, M. Sargent, S. R. Gill, K. E. Nelson,
and D. A. Relman. 2005. Diversity of the human intestinal microbial flora. Science. 308:1635-1638.
Edberg, S. C., E. W. Rice, R. J. Karlin, and M. J. Allen. 2000. Escherichia coli: the best biological drinking
water indicator for public health protection. Symp. Ser. Soc. Appl. Microbiol. 106S-116S.
Enas, G. C., and S. C. Choi. 1986. Choice of the smoothing parameter and efficiency of k-nearest neighbor
classification. Comput. Math. Applic. 12A:235-244.
Farag, A. M., J. N. Goldstein, D. F. Woodward, and M. Samadpour. 2001. Water quality in three creeks in the
backcountry of Grand Teton National Park, USA. J. Fresh Water Ecol. 16:135-143.
Field, K.G., A.E. Bernhard, and T. J. Brodeur. 2003. Molecular approaches to microbiological monitoring:
Fecal source detection. Environ. Moni. Assess. 81: 313-326.
Field, K. G., E. C. Chern, L. K. Dick, J. Fuhrmann, J. Griffith, P. Holden, M. G. LaMontagne, J.Le, B. Olson,
and M. T. Simonich. 2003. A comparative study of culture-independent, library-independent genotypic meth-
ods of fecal source tracking. J. Water Health 1:181 -194.
Fiksdal, L., J. S. Maki, S. J. Lacroix, and J. T. Staley. 1985. Survival and detection of Bacteroides spp, pro-
spective indicator bacteria. Appl. Environ. Microbiol. 49:148-150.
Filip, Z., D. Kaddu Mulindwa, and G. Milde. 1987. Survival and adhesion of some pathogenic and faculta-
tive pathogenic microorganisms in groundwater. Water Sci. Technol. 19:1189-1190.
Fode-Vaughan, K. A., C. F. Wimpee, C. C. Remsen, and M. L. Perille Collins. 2001. Detection of bacteria in
environmental samples by direct PCR without DNA extraction. BioTechniques 31:598-607.
Fong, T. T., D. W. Griffin, and E. K. Lipp. 2005. Molecular assays for targeting human and bovine enteric
viruses in coastal waters and their application for library-independent source tracking. Appl Environ Micro-
biol. 71:2070-2078.
-------
Frahm, E., and U. Obst. 2003. Application of fluorogenic probe technique (TaqMan PCR) to the detection
of Enterococcus spp. and Escherichia coli in water samples. J. Microbiol. Meth. 52:123-131.
Franks, A. H., H. J. M. Harmsen, G. C. Raangs, G. J. Jansen, F. Schut, and G. W. Welling. 1998. Variations
of bacterial populations in human feces measured by fluorescent in situ hybridization with group-specific
16S rRNA-targeted oligonucleotide probes. Appl. Environ. Microbiol. 64:3336-3345.
Fujioka, R. S., and M. N. Byappanahalli. 2003. Proceedings and Report: Tropical Water Quality Indicator
Workshop, pp. 90. Honolulu, HI, Water Resource Center, University of Hawaii at Manoa.
Fujioka, R. S., C. Sian-Denton, M.Borja, J. Castro, and K. Morphew. 1999. Soil: the environmental source of
Escherichia coli and enterococci in Guam's streams. J. Appl. Microbiol. Symposium Supplement 85:83S-89S.
Funderburg, S. W, and C. A. Sorber. 1985 Coliphages as indicators of enteric viruses in activated sludge.
Water Res. 19:547-555.
Furuse, K. 1987. Distribution of coliphages in the environment: general considerations. In Phage Ecology, S.
M. Goyal, C. P. Gerba, and G. Bitton (ed.), p. 87-124, John Wiley and Sons, New York
Galland, J. C., D. R. Hyatt, S. S. Crupper, and D. W. Acheson. 2001. Prevalence, antibiotic susceptibility, and
diversity of Escherichia coli O157:H7 isolates from a longitudinal study of beef cattle feedlots. Appl.
Environ. Microbiol. 67:1619-1627.
Gantzer, C., A. Maul, J. M. Audic, and L. Schwartzbrod. 1998. Detection of infectious enteroviruses, entero-
virus genomes, somatic coliphages, and Bacteroides fragilis phages in treated wastewater. Appl. Environ.
Microbiol. 64:4307-4312.
Geldreich, E. E., and B. A. Kenner. 1969. Concepts of fecal streptococci in stream pollution. J. Water Pollut.
Control Fed. 41:R336-352.
Geldreich E. E., L. C. Best, B. A. Kenner, and D. J. Van Donsel. 1968. The bacteriological aspects of storm-
water pollution. J. Water Pollut. Control Fed. 40:1861-1872.
Geldreich, E. E., and N. A. Clarke. 1966. Bacterial pollution indicators in the intestinal tract of freshwater
fish. Appl. Microbiol. 14:429-437.
Geldreich, E. E., Kenner B. A., and P. W. Kabler. 1964. Occurrence of coliforms, fecal coliforms, and fecal
streptococci on vegetation and insects. Appl. Microbiol. 12:63-69.
Gerba, C. P., and J. S. McLeod. 1976 Effect of sediments on the survival of Escherichia coli in marine wa-
ters. Appl. Environ. Microbiol. 32:114-120.
Ginzinger, D.G. 2002. Gene quantification using real-time quantitative PCR: an emerging technology hits
the mainstream. Exp. Hematol. 30:503-512.
Giovannoni, S. J., E. F. DeLong, G. J. Olsen, and N. R. Pace. 1998. Phylogenetic group-specific oligode-
oxynucleotide probes for identification of single microbial cells. J. Bacterial. 170:720-726.
-------
Gordon, D. M. 1997. The genetic structure of Escherichia coli populations in feral house mice. Microbiology
143:2039-2046.
Gordon, D. M. 2001. Geographical structure and host specificity in bacteria and the implications for tracing
the source of coliform contamination. Microbiology. 147:1079-1085.
Glassmeyer, S. T., E. T. Furlong, D. W. Kolpin, J. D. Cahill, S. D. Zaugg, S. L. Werner, M. T. Meyer, D. D.
Kryak. 2005. Transport of chemical and microbial compounds from known wastewater discharges: potential
for use as indicators of human fecal contamination. Environ. Sci. Technol. 39:5157-5169.
Gordon, D. M., S. Bauer, and J. R. Johnson. 2002. The genetic structure of Escherichia coli populations in
primary and secondary habitats. Microbiology. 148:1513-1522.
Gordon, D. M., and J. Lee. 1999. The genetic structure of enteric bacteria from Australian mammals.
Microbiology. 145:2673-2682.
Grabow, W. O. K., O. W. Prozesky, and J. S. Burger. 1975. Behavior in a river and dam of coliform bacteria
with transferable or non-transferable drug resistance. Water Res. 9:777-782.
Grant, S. B., B. F. Sanders, A. B. Boehm, J. A. Redman, J. H. Kim, R. D. Mrse, A. K. Chu, M. Gouldin, C.
D. McGee, N. A. Gardiner, B. H. Jones, J. Svejkovsky, G. V. Leipzig, and A. Brown. 2001. Generation of en-
terococci bacteria in a coastal saltwater marsh and its impact on surf zone water quality. Environ Sci Technol
35:2407-2416.
Graves, A. K., C. Hagedorn, A. Teetor, M. Mahal, A. M. Booth, and R. B. Reneau, Jr. 2002. Antibiotic resis-
tance profiles to determine sources of fecal contamination in a rural Virginia watershed. J. Environ. Qual. 31:
1300-1308.
Grif, K., H. Karch, C. Schneider, F. D. Daschner, L. Beutin, T. Cheasty, H. Smith, B. Rowe, M. P. Dierich,
and F. Allerberger. 1998. Comparative study of five different techniques for epidemiological typing of Esch-
erichia coli O157. Diag. Microbiol. Infect. Dis. 32:165-176.
Griffin, D. W., C. J. Gibson, E. K. Lipp, K. Riley, J. H. Paul, and J. B. Rose. 1999. Detection of viral patho-
gens by reverse transcriptase PCR and of microbial indicators by standard methods in the canals of the
Florida Keys. Appl. Environ. Microbiol. 65:4118-4125.
Griffin, D. W., R. Stokes, J. B. Rose, and J. H. Paul. 2000. Bacterial Indicator occurrence and the use of an
F(+) specific RNA coliphage assay to identify fecal sources in Homosassa Springs, Florida. Microb. Ecol.
39:56-64.
Griffith, J. F., S. B. Weisberg, and C. D. McGee. 2003. Evaluation of microbial source tracking methods us-
ing mixed fecal sources in aqueous test samples. J. Water Health. 1:141-51.
Grimont, F., and P. A. D. Grimont. 1986. Ribosomal ribonucleic acid gene restriction patterns as potential
taxonomic tools. Ann. Inst. Pasteur Microbiol. 137B: 165-175.
Guan, S., R. Xu, S. Chen, J. Odumeru, and C. Gyles. 2002. Development of a procedure for discriminating
among Escherichia coli isolates from animal and human sources. Appl. Environ. Microbiol. 68:2690-2698.
-------
Gustaferro, C. A., and D.H. Parsing. 1992. Chemiluminescent universal probe for bacterial ribotyping. J.
Clin. Microbiol. 30:1039-1041.
Guttman, L. 1950. The basis for scalogram analysis. In Measurement and Prediction. S. A.Stauffer, L. Gutt-
man, E. A. Suchman, P. L. Lazarsfeld, S. A. Star, & J. A. Clausen (ed.), Vol.4, Princeton University Press,
Princeton, New Jersey.
Haack, S. K., L. R. Fogarty, and C. Wright. 2003. Escherichia coli enterococci at beaches in the Grand
Traverse Bay, Lake Michigan: sources, characteristics, and environmental pathways. Environ. Sci. Techno!.
37:3275-3282.
Hagedorn, C., S. L. Robinson, J. R. Filtz, S. M. Grubbs, T. A. Angier and R. B. Reneau. 1999. Determining
sources of fecal pollution in a rural Virginia watershed with antibiotic resistance patterns in fecal
streptococci. Appl. Environ. Microbiol. 65:5522-5531.
Hagedorn, C, J. B. Crozier, K. A. Mentz, A. M. Booth, A. K.Graves, N. J. Nelson, and R. B. Reneau. 2003.
Carbon source utilization profiles as a method to identify sources of faecal pollution in water. J. Appl.
Microbiol. 94:792-799.
Hahm, B. K., Y. Maldonado, E. Schreiber, A. K. Bhunia, and C. H. Nakatsu. 2003. Subtyping of foodborne
and environmental isolates of Escherichia coli by multiplex-PCR, rep-PCR, PFGE, ribotyping and AFLP. J.
Microbiol. Meth. 53:387-399.
Hahm, B. K., A. K. Bhunia, and C. H. Nakatsu. 2003. Application of AFLP for discriminating Escherichia
coli isolated from livestock, wildlife and humans. American Society for Microbiology Annual Meeting.
Washington D.C.
Haile, R. W, J. S. Witte, M. Gold, R. Cressey, C. McGee, R. C. Millikan, A. Glasser, N. Harawa, C. Ervin, P.
Harmon, J. Harper, J. Dermand, J. Alamillo, K. Barrett, M. Nides, and G. Wang. 1999. The health effects of
swimming in ocean water contaminated by storm drain runoff. Epidemiology 10:355-363.
Hand, D. J., 1997. Construction and assessment of classification rules. John Wiley and Sons. Chichester, UK
Hardina, C. M., and R. S. Fujioka. 2001. Soil: the environmental source of Escherichia coli and enterococci
in Hawaii's streams. Environ Tox Water Qual 6:185-195.
Harmsen, H. J. M., G. C. Raangs, T. He, J. E. Degener, and G. W. Welling. 2002. Extensive set of 16S
rRNA-based probes for detection of bacteria in human feces. Appl. Environ. Microbiol. 68:2982-2990.
Harmsen, H. J., A. C. Wildeboer-Veloo, G. C. Raangs, A. A. Wagendorp, N. Klijn. 2000. Analysis of intesti-
nal flora development in breast-fed and formula-fed infants by using molecular identification and detection
methods. J. Pediatr. Gastroenterol. Nutr. 30:61-70.
Hartel, P. G., E. A. Frick, A. L. Funk, J. L. Hill, J. D. Summer, and M. B. Gregory. 2004. Sharing of ribotype
patterns of Escherichia coli isolates during baseflow and stormflow conditions. USGS Scientific Investiga-
tions Report 2004.
-------
Hartel, P. G., W. I. Segars, N. J. Stern, J. Steiner, and A. Buchan. 1999. Ribotyping to determine the host
origin of Escherichia coll isolates in different water samples, p. 377-382. In Wildland hydrology, D.S. Olsen
and J.P. Potyondy (ed.). Am. Water Resour. Assoc., Herndon, VA.
Hartel, P. G., J. D. Summner, J. L. Hill, J. V. Collins, J. A. Entry, and W. I. Segars. 2002. Geographic vari-
ability of Escherichia coli ribotypes from animals in Idaho and Georgia. J. Environ. Qual. 31:1273-1278.
Hartel, P. G., J. D. Summer, and W. I. Segars. 2003. Deer diet affects ribotype diversity of Escherichia coli
for bacterial source tracking. Water Res. 37:3263-3268.
Hartigan, J. 1975. Clustering Algorithms. Wiley, New York, NY.
Hartl, D. L., and D. E. Dykhuizen. 1984 The population genetics of Escherichia coli. Annu. Rev. Genet.
18:31-68.
Harwood, V. J., J. Whitlock, and V. Withington. 2000. Classification of antibiotic resistance patterns of in-
dicator bacteria by discriminant analysis: use in predicting the source of fecal contamination in subtropical
waters. Appl Environ. Microbiol. 66:3698-704.
Harwood, V. J., B. Wiggins, C. Hagedorn, R. D. Ellender, J. Gooch, J. Kern, M. Sarnadpour, A. C. H. Chap-
man and B. J. Robinson. 2003. Phenotypic library-based microbial source tracking methods: efficacy in the
California collaborative study. J. Water Health 1:153-166.
Hastie, T., Tibshirani, R. and Friedman, J. 2002. The Elements of Statistical Learning. Springer Series in
Satistiscs, Springer-Verlag, New York, NY, USA.
Head, I. M., J. R.Saunders, and R.W. Pickup. 1998. Microbial evolution, diversity, and ecology: A decade of
ribosomal RNA analysis of uncultivated microorganisms. Microbial Ecol. 35:1-21.
Hilton, A.C., J.G. Banks, and C.W. Penn. 1997. Optimization of RAPD for fingerprinting Salmonella. Lett.
Appl. Microbiol. 24:243-248.
Holmes, B. L., M. Costas, M. Ganner, S. L. W. On, and M. Stevens. 1994. Evaluation of Biolog system for
identification of some Gram-negative bacteria of clinical importance. J. Clin. Microbiol. 32:1970-1975.
Hopkins, K. L., and A. C. Hilton. 2000. Methods available for the sub-typing of Escherichia coli O157.
World J. Microbiol. Biotechnol. 16:741-748.
Hood, K. L., J. E. Whitlock, and V. J. Harwood. 2003. Factors that influence the persistence of bacterial indi-
cator organisms in fresh and saline subtropical waters. American Society for Microbiology General Meeting,
Washington, DC.
Hood, K. L., J. E. Whitlock, M. R. McLaughlin, J. B. Rose, and V. J. Harwood. 2002. Survival and finger-
print stability of indicator organisms in subtropical waters. American Society for Microbiology General
Meeting, Salt Lake City, UT.
Hotelling, H. 1933. Analysis of a complex of statistical variables into principal components. J. Educational
Psychol. 24:417-441, 498-520.
-------
Hsu, F. C., Y. S. Shieh, J. van Duin, M. J. Beekwilder, and M. D. Sobsey. 1995. Genotyping male-specific
RNA coliphages by hybridization with oligonucleotide probes. Appl, Environ. Microbiol. 61:3960-3966.
Hughes, M. S., L. A. Beck, R. A. Skuce. 1994. Identification and Elimination of DNA Sequences in Taq
DNA Polymerase. J. Clin. Microbiol. 32:2007-2008.
Hungate, R. E. 1966. The rumen and its microbes. Academic Press, New York, NY.
Hyer, K. E., and D. L. Moyer. 2003. Patterns and sources of fecal coliform bacteria in three streams in Vir-
ginia, 1999-2000., USGS Water Resources Investigations Report 03-4115.
Ito,Y, Y. linuma, H. Baba, Y. Sugino, Y. Hasegawa, K. Shimokata, S. Ichiyama, T. Hasegawa, and M. Ohta.
2003. Evaluation of automated ribotyping system for characterization and identification of verocytotoxin-
producing Escherichia coli isolated in Japan. Jpn. J. Infect. Dis. 56:200-204.
Jaccard, P. 1901. Distribution de la flore alpine dans le Bassin des Dranes et dans quelques regions voisines.
BullSoc. Vaud. Sci. Nat., 37:241-272.
Jagals, P., W. O. K. Grabow, and J. C. De Villiers. 1995. Evaluation of indicators for assessment of human
and animal faecal pollution of surface run-off. Water Sci Technol. 31:235-241.
Jenkins, M.B., P. G. Hartel, T. J. Olexa, and J. A. Stuedemann. 2003. Putative temporal variability of Esch-
erichia coli ribotypes from yearling steers. J. Environ. Qual. 32:305-309.
Jiang, S., R. Noble, and W. Chu. 2001. Human adenoviruses and coliphages in urban runoff-impacted coastal
waters of Southern California. Appl. Environ. Microbiol. 67:179-184.
Jimenez-Clavero, M. A., C. Fernandez, J. A. Ortiz, J. Pro, G. Carbonell, J. V. Tarazona, N. Roblas, and V.
Ley. 2003. Teschoviruses as indicators of porcine fecal contamination of surface water. Appl Environ
Microbiol. 69:6311-6315.
Johnson, L. K., M. B. Brown, E. A. Carruthers, J. A. Ferguson, P. E. Dombek, and M. J. Sadowsky. 2004.
Sample size, library composition, and genotypic diversity influence accuracy of determining sources of fecal
pollution among natural populations of Escherichia coli from different animals. Appl. Environ. Microbiol.
70:4478-4485.
Johnson, R. A. and Wichern, D. W. 2002. Applied Multivariate Statistical Analysis, Prentice-Hall.
Jolliffe, I. T. 2002. Principal Component Analysis, Springer-Verlag, New York.
Kaufmann, P., A. Pfefferkorn, M. Teuber, and L. Meile. 1997. Identification and quantification of Bifidobac-
terium species isolated from food with genus-specific 16S rRNA-target probes by colony hybridization and
PCR.Appl. Environ. Microbiol. 63:1268-1273.
Kelley, T. R., O. C. Pancorba, W. C. Merka and H. M. Barnharts. 1998. Antibiotic resistance of bacterial litter
isolates. Poultry Sci. 77:243-247.
-------
Khatib, L. A., Y. L. Tsai, and B. H. Olson. 2002. A biomarker for the identification of cattle fecal pollution in
water using the LTIIa toxin gene from enterotoxigenic Escherichia coli. Appl. Microbiol. Biotech. 59:97-104.
Khatib, L. A., Y. L. Tsai, and B. H. Olson. 2003. A biomarker for the identification of swine fecal pollution
in water, using the STI1 toxin gene from enterotoxigenic Escherichia coli. Appl. Microbiol. Biotech. 63:231-
238.
Klein, D. 2002. Quantification using real-time PCR technology: applications and limitations. Trends Mol.
Med 8:257-260.
Koellner, T., Hersperger, A. M. and Wohlgemuth, T. 2004. Rarefaction method for assessing plant species
diversity on a regional scale. Ecography 27:532-544.
Konopka, A., L Oliver, and R. F. Turco. 1998. The use of carbon substrate utilization patterns in environmen-
tal and ecological microbiology. Micro. Ecol. 35:103-115.
Kreader, C. A. 1998. Persistence of PCR-detectable Bacteroides distasonis from human feces in river water.
Appl. Environ. Microbiol. 64:4103-4105.
Kreader, C. A. 1995. Design and evaluation of Bacteroides DNA probes for the specific detection of human
fecal pollution. Appl. Environ. Microbiol. 61:1171-1179.
Kruse, H., and H. S0rum. 1994. Transfer of multiple drug resistance plasmids between bacteria of diverse
origins in natural microenvironments. Appl. Environ. Microbiol. 60: 4015-4021.
Kulczynski, S. 1928. Zespoly roslin w Pieninach Bull. Int. Acad. Po. Sci. Lettres Ser. B Suppl. 2:57-203
Kuntz, R.L., P.O. Hartel, D.G. Godfrey, J.L. McDonald, K.W. Gates, and W.I. Segars. 2003. Targeted sam-
pling protocol as prelude to bacterial source tracking with Enterococcusfaecalis. J. Environ. Qual. 32:2311-
2318.
LaLiberte, P., and D. J. Grimes, 1982. Survival of Escherichia coli in lake bottom sediment. Appl. Environ.
Microbiol. 43:623-628.
Leclerc, H., S. Edberg, V. Pierzo, and J. M. Delattre. 2000. Bacteriophages as indicators of enteric viruses
and public health risk in groundwaters. J. Appl. Microbiol. 88:5-21.
Lefresne, G., E. Latrille, F. Winger, and P. A. D. Grimont. 2004. Repeatability and reproducibility of ribotyp-
ing and its computer interpretation. Res. Microbiol. 155:154-161.
Leser, T.D., J.Z. Amenuvor, T.K. Jensen, R.H. Lindecrona, M. Boye, and K. M011er. 2002. Culture-indepen-
dent analysis of gut bacteria: the pig gastrointestinal tract microbiota revisisted. Appl. Environ. Microbiol.
68:673-690.
Leung, K. T., R. Mackereth, Y.-C. Tien, E. Topp. 2004. A comparison of AFLP and ERIC-PCR analyses for
discriminating Escherichia coli from cattle, pig and human sources. FEMS Microbiol. Ecol. 47:111-119.
-------
Ley, V., J. Higgins, and R. Payer 2002. Bovine enteroviruses as indicators of fecal contamination. Appl
Environ Microbiol. 68:3455-3461.
Lin, J.-J., and J. Kuo. 1995. AFLP: a novel PCR-based assay for plant and bacterial DNA fingerprinting.
Focus 17:66-70.
Lipman, J. A., A. de Nijs, T. J. G. M. Lam, and W. Gaastra. 1995. Identification of Escherichia coli strain
from cows with clinical mastitis by serotyping and DNA polymorphism patterns with REP and ERIC prim-
ers. Vet. Microbiol. 43:13-19.
LiPuma, J. J., T. L. Stull, S. E. Dasen, K. A. Pidcock, D. Kaye and O. M. Korzeniowski. 1989. DNA poly-
morphisms among Escherichia coli isolated from bacteriuric women. J. Infect. Dis. 159:526-532.
Liu, C. X., Y. L. Song, M. McTeague, A.W. Vu, H. Wexler, and S. M. Finegold. 2003. Rapid identification
of the species of the Bacteroides fragilis group by multiplex PCR assays using group- and species-specific
primers. FEMS Microbiol. Lett. 222:9-16.
Liu, W. T., T. L. Marsh, H. Cheng, and L. J. Forney. 1997. Characterization of microbial diversity by deter-
mining terminal restriction fragment length polymorphisms of genes encoding 16S rRNA. Appl. Environ.
Microbiol. 63:4516-4522.
Livak, K. J., S. J. A. Flood, J. Marmaro, W. Giusti, and K. Deetz. 1995. Oligonucleotides with fluorescent
dyes at opposite ends provide a quenched probe system useful for detecting PCR product and nucleic acid
hybridization. PCR Meth. Appl. 4:357-362.
Lupski, J. R., and G. M. Weinstock. 1992. Short, interspersed repetitive DNA sequences in prokaryotic ge-
nomes. J. Bacterial. 174:4525-4529.
Machado, J., F. Grimont and P. A. D. Grimont. 1998. Computer identification of Escherichia coli rRNA gene
restriction patterns. Res. Microbiol. 149:119-135.
Madico, G., N. S. Akopyants, and D. E. Berg. 1995. Arbitrarily primed PCR DNA fingerprinting of Esch-
erichia coli O157:H7 strains by using templates from boiled cultures. J. Clin. Microbiol. 33:1534-1536.
Malinen, E., A. Kassinen, T. Rinttila, and A. Palva. 2003. Comparison of real-time PCR with SYBR Green I
or 5'-nuclease assays and dot-blot hybridization with rDNA-targeted oligonucleotide probes in quantification
of selected faecal bacteria. Microbiology. 149:269-277.
Mara, D. D., and J. Oragui. 1981. Occurrence of Rhodococcus coprophilus and associated Actinomycetes in
feces, sewage, and freshwater. Appl. Environ. Microbiol. 42: 1037-1042
Martellini, A., P. Payment, and R. Villemur. 2005. Use of eukaryotic mitochondrial DNA to differentiate hu-
man, bovine, porcine and ovine sources in fecally contaminated surface water. Water Res. 39:541-548.
Martin, I. E., S.D. Tyler, K. D. Tyler, R. Khakhria, and W. M. Johnson. 1996. Evaluation of ribotyping as
epidemiologic tool for typing Escherichia coli serogroup O157 isolates. J. Clin. Microbiol. 34:720-723.
Martin, J. D. and J. O. Mundt. 1972. Enterococci in insects. Appl. Microbiol. 24:575-580.
-------
Mathieu-Daude, R, K. Evans, F. Kullmann, R. Honeycutt, T. Vogt, J. Welsch, and M. McClelland, 1998. Ap-
plications of DNA and RNA Fingerprinting by the arbitrary primed polymerase chain reaction, pp. 414-436.
In Bacterial Genomes, F. de Bruijn, J. R. Lupski and G. M. Weinstock (ed.), Chapman and Hall, New York,
NY.
Matsuki, T., K. Watanabe, J. Fujimoto, Y. Miyamtoto, T. Takada, K. Matsumoto, H. Oyaizu, and R. Tanaka.
2002. Development of 16S rRNA-gene-targeted group-specific primers for the detection and identification
of predominant bacteria in human feces. Appl. Environ. Microbiol 68:5445-5451.
Matsuki, T., K. Watanabe, J. Fujimoto, Y. Kado, T. Takada, K. Matsumoto, and R. Tanaka. 2004. Quantita-
tive PCR with 16S rRNA-gene-targeted species-specific primers for analysis of human intestinal Bifidobacte-
ria. Appl. Environ. Microbiol. 70:167-173.
McLachlan, G. J. 1992. Discriminant Analysis and Statistical Pattern Recognition. John Wiley and Sons,
New York, NY.
McLellan, S. L., A. D. Daniels and A. K. Salmore. 2001. Clonal populations of thermotolerant Enterobacteri-
aceae in recreational water and their potential interference with fecal Escherichia coli counts. Appl. Environ.
Microbiol. 67:4934-4938.
McLellan, S. L., A. D. Daniels and A. K. Salmore. 2003. Genetic characterization of Escherichia coli popu-
lations from host sources of fecal pollution using DNA fingerprinting. Appl. Environ. Microbiol. 69:2587-
2594.
Milkman, R. 1973. Electrophoretic variation in Escherichia coli from natural sources. Science 182:1024-
1026.
Muyzer, G., S. Hottentrager, A. Teske, and C. Wawer. 1996. Denaturing gradient gel electrophoresis of PCR-
amplified 16S rDNA-A new molecular approach to analyse the genetic diversity of mixed microbial com-
munities, p. 3.4.4: 1-23. In Molecular microbial ecology manual, A. Akkermans, van Elsas, J.D. and F. J. de
Bruijn (ed.), Kluwer Academic Publishers, Nowell, MA
Muyzer, G., and K. Smalla. 1998. Application of denaturing gel electrophoresis (DGGE) and temperature
gradient gel electrophoresis (TGGE) in microbial ecology. Antonie Van Leeuwenhoek 73:127-141.
Myoda, S. P., C. A. Carson, J. J. Fuhrmann, B.-K. Hahm, P. G. Hartel, R. L. Kuntz, C. H. Nakatsu, M. J. Sa-
dowsky, M. Samadpour and H. Yampara-lquise. 2003. Comparing genotypic bacterial source tracking meth-
ods that require a host origin database. J. Water Health 1:167-180.
Nakatsu, C. H., and T. L. Marsh. 2005. Analysis of microbial communities with denaturing gradient gel elec-
trophoresis and terminal restriction fragment length polymorphism. In Methods for general and molecular
bacteriology. C. A. Reddy, T. M. Schmidt (eds.) ASM Press, Washington D. C. (in press)
National Committee for Clinical Laboratory Standards. 1999. Methods for dilution antimicrobial susceptibil-
ity tests for bacteria that grow aerobically. Villanova, PA: NCCLS.
Nebra, Y, X. Bonjoch, and A. R. Blanch. 2003. Use of Bifidobacterium dentium as an indicator of the origin
of fecal water pollution. Appl. Environ. Microbiol. 69:2651-2656.
-------
Noble, R. T., and J. A. Fuhrman 2001. Enteroviruses detected by reverse transcriptase polymerase chain reac-
tion from the coastal waters of Santa Monica Bay, California: low correlation with bacterial indicator levels.
Hydrobiology 460:175-184.
Noble, R.T., S.A. Allen, A. D. Blackwood, W. Chu, S. C. Jiang, G. L. Lovelace, M. D. Sobsey, J. R. Stewart,
and D. A. Wait. 2003. Use of viral pathogens and indicators to differentiate between human and non-human
fecal contamination in a microbial source tracking comparison study. J. Wat. Health. 1:195-204.
Ochiai, A. 1957. Zoogeographical studies on the soleoid fishes found in Japan and its neighboring regions 11,
Bull. Jap. Soc. Sci. Fish. 22:526-530.
Ogimoto, K., and S. Imai. 1981. Atlas of rumen microbiology. Japan Scientific Society Press, Tokyo, Japan.
Ohlsen, K., T. Ternes, G. Werner, U. Wallner, D. Loffler, W. Ziebuhr, W. Witte and J. Hacker. 2003. Impact of
antibiotics on conjugational resistance gene transfer in Staphylococcus aureus in sewage. Environ. Microbiol.
5:711-716.
Olive, D. M., and P. Bean. 1999. Principles and applications of methods for DNA-based typing of microbial
organisms. J. Clin. Microbiol. 37:1661-1669.
Olsen, G. J., D. J. Lane, S. J. Giovannoni, and N. R. Pace. 1986. Microbial ecology and evolution: A ribo-
somal RNA approach. Ann. Rev. Microbiol. 40:337-365.
Oshiro, R. K. and B. H. Olson. 1997. Occurrence of STh toxin gene in wastewater. In Coliforms and E.
coli: Problem or Solution?, D. Kay and C. Fricker (ed.), pp. 255-259. Royal Society of Chemistry, Cam-
bridge, UK.
Pacheco, A. B., B. E. Guth, D. F. de-Almeida, and L. C. S. Ferreira. 1996. Characterization of enterotoxigen-
ic Escherichia coli by random amplification of polymorphic DNA. Res. Microbiol. 147:175-182.
Pacheco, A. B., B. E Guth, K. C. Scares, L. Nishimura, D. F. de Almeida, and L. C. S. Ferreira. 1997. Ran-
dom amplification of polymorphic DNA reveals serotype-specific clonal clusters among enterotoxigenic
Escherichia coli strains isolated from humans. J. Clin. Microbiol. 35:1521-1525.
Parveen, S., R. L. Murphree, L. Edmiston, C. W. Kaspar, K. M. Portier, and M. L. Tamplin. 1997. Associa-
tion of multiple-antibiotic-resistance profiles with point and nonpoint sources of Escherichia coli in Apala-
chicola Bay. Appl. Environ. Microbiol. 63:2607-2612.
Parveen, S., K. M. Portier, K. Robinson, L. Edmiston, and M. L. Tamplin. 1999. Discriminant analysis of ri-
botype profiles of Escherichia coli for differentiating human and nonhuman sources of fecal pollution. Appl.
Environ. Microbiol. 65:3142-3147.
Parveen, S., N. C. Hodge, R. E. Stall, S. R. Farrah, and M. L. Tamplin. 2001. Phenotypic and genotypic char-
acterization of human and nonhuman Escherichia coli. Water Res. 35:379-386
Payment, P, and P. R. Hunter. 2001. Endemic and epidemic infectious intestinal disease and its relationship
to drinking water. In L. Fewtrell and J. Bartram (ed.), Water Quality: Guidelines, Standards and Health. As-
sessment of risk and risk management for water-related infectious disease. 1WA Publishing, on behalf of the
World Health Organization, London.
-------
Penner, G. A., A. Bush, R Wise, W. Kim, L. Domier, K. Kasha, A. Laroche, G. Scoles, S. J. Molnar, and G.
Fedak. 1993. Reproducibility of random amplified polymorphic DNA (RAPD) analysis among laboratories.
PCR Meth. Applicat. 2:341-345.
Picard, B., N. Picard-Pasquier, R. Krishnamoorthy and P. Goullet. 1991. Characterization of highly virulent
Escherichia coli strains by ribosomal DNA restriction fragment length polymorphism. FEMS Microbiol. Lett.
82:183-188.
Pina, S., M. Puig, F. Lucena, J. Jofre, and R. Girones. 1998. Viral pollution in the environment and in shell-
fish: human adenovirus detection by PCR as an index of human viruses. Appl Environ. Microbiol 64:3376-
3382.
Pryde, S. E., A. J. Richardson, C. S. Stewart, and H. J. Flint. 1999. Molecular analysis of the microbial
diversity present in the colonic wall, colonic lumen and cecal lumen of a pig. Appl. Environ. Microbiol.
65:5372-5377.
Rabinovici, S. J. M., R. L. Berhknopf, A. M. Wein, D. L. Coursey, and R. L. Whitman. 2004. Economic and
health risk trade-offs of swim closures at a Lake Michigan Beach. Environ. Sci. Technol. 38:2737-2745.
Rademaker, J. L. W., and F. J. de Bruijn. 1997. Characterization and classification of microbes by rep-PCR
genomic fingerprinting and computer-assisted pattern analysis, p. 151-171. In DNA markers: protocols, ap-
plications, and overviews, G. Caetano-Anolles, and P. M. Gresshoff (ed.), John Wiley and Sons, New York,
NY.
Radu, S., O. W. Ling, G. Rusul, M. I. Karim, and M. Nishibuchi. 2001. Detection of Escherichia coli O157:
H7 by multiplex PCR and their characterization by plasmid profiling, antimicrobial resistance, RAPD and
PFGE analyses. J. Microbiol. Meth. 46:131-139.
Ramsak, A., M. Peterka, K. Tajima, J.C. Martin, J. Wood, M.E.A. Johnston, R.I. Aminov, H.J. Flint, and G.
Avgustin. 2000. Unraveling the genetic diversity of ruminal bacteria belonging to the CFB phylum. FEMS
Microbiol. Ecol. 33:69-79.
Rand, K. H., and H. Houck. 1990. Taq polymerase contains bacterial DNA of unknown origin. Mol. Cell.
Probes 4:445-450.
Ranjard, L., F. Poly, and S. Nazaret. 2000. Monitoring complex bacterial communities using culture-inde-
pendent molecular techniques: application to soil environment. Res.Microbiol. 151:167-177.
Regnault, B., F. Grimont, and P. A. Grimont. 1997. Universal ribotyping method using a chemically labelled
oligonucleotide probe mixture. Res. Microbiol 148:649-659.
Resnick, I. G., and M. A. Levin. 1981. Assessment of Bifido bacteria as indicators of human fecal pollution.
Appl. Environ. Microbiol. 42:433-438.
Restrepo, S., M. Duque, J. Tohme, and V. Verdier. 1999. AFLP fingerprinting: an efficient technique for de-
tecting genetic variation of Xanthomonas axonopodis pv. manihotis. Microbiology 145:107-114.
-------
Rigottier-Gois, L., A. -G. Le Bourhis, G. Gramet, V. Rochet, and J. Dore. 2003. Fluorescent hybridisation
combined with flow cytometry and hybridisation of total RNA to analyse the composition of microbial com-
munities in human faeces using 16S rRNA probes. FEMS Microbiol. Ecol. 43:237-245.
Ritter K. J., E . Carruthers, C. A. Carson, R. D. Ellender, V. J. Harwood, K. Kingsley, C. Nakatsu, M. Sad-
owsky, B. Shear, B. West, J. E. Whitlock, B. A. Wiggins, J. D. Wilbur. 2003. Assessment of statistical meth-
ods used in library-based approaches to microbial source tracking. J. Water Health. 1:209-223.
Rivera, S. C., T. C. Hazen,. and G. A. Toranzos. 1988. Isolation of fecal coliforms from pristine sites in a
tropical rain forest. Appl. Environ. Microbiol. 54:513-517.
Rozen, Y., and S. Belkin. 2001. Survival of enteric bacteria in seawater. FEMS Microbiol. Rev. 25:513-529.
Russell, P. F., and T. R. Rao. 1940. On habitat an association of species of anopheline larvae in southeastern
Madras. J. Malaria lust. India. 3:153-178.
Sadowsky, M. J. 1994. Microbial DNA fingerprinting and restriction fragment length polymorphism analysis.
pp. 647-664. In Methods of Soil Analysis, Chemical and Microbiological Properties of Soils. R. W. Weaver,
J. S. Angle, and P. Bottomley (ed.), ASA-SSSA, Madison, WI.
Sadowsky, M. J., and H.-G. Hur. 1998. Use of endogenous repeated sequences to fingerprint bacterial ge-
nomic DNA, 399-413. In Bacterial genomes: structure and analysis, J.R. Lupski, G. Weinstock, and F. J. de
Bruijn (ed.), Chapman and Hall, New York, NY.
Salyers, A. A., N. B. Shoemaker, A. M. Stevens and L.-Y. Li. 1995. Conjugative transposons: An unusual and
diverse set of integrated gene transfer elements. Microbiol. Rev. 59:579-590.
Santo Domingo, J. W, S. C. Siefring, and R. A. Haugland. 2003. Real-time PCR method to detect Entero-
coccus faecalis in water. Biotech. Lett. 25:261-263.
Santo Domingo, J. W, S. Harmon, and J. Bennett. 2000. Survival of Salmonella Species in river water. Curr.
Microbiol. 40:409-417
Santo Domingo, J. W. M. G. Kaufman, M. J. Klug, and J. M. Tiedje. 1998. Characterization of the cricket
hindgut microbiota with fluorescently labeled rRNA-targeted oligonucleotide probes. Appl. Environ. Micro-
biol. 64:752-755.
Santo Domingo, J. W, F. A. Fuentes, and T. C. Hazen. 1989. Survival and activity of Streptococcus faecalis
and Escherichia coli in petroleum-contaminated tropical marine waters. Environ. Pollut. 56:263-281.
Savageau, M. A. 1983. Escherichia coli habitats, cell types, and molecular mechanisms of gene control.
American Naturalist 122:732-744.
Savill, M. G., S. R. Murray, P. Scholes, E. W. Maas, R. E. McCormick, E. B. Moore, and B. J. Gilpin. 2001.
Application of polymerase chain reaction (PCR) and TaqMan™ PCR techniques to the detection and identifi-
cation of Rhodococcus coprophilus in faecal samples. J. Microbiol. Meth. 47:355-368.
-------
Schaper, M., J. Jofre, M. Uys, and W. O. Grabow. 2002a. Distribution of genotypes of F-specific RNAbac-
teriophages in human and non-human sources of faecal pollution in South Africa and Spain. J. Appl. Micro-
biol. 92:657-667.
Schaper, M., A. E. Duran, and J. Jofre. 2002b. Comparative resistance of phage isolates of four genotypes of
F-specific RNA bacteriophages to various inactivation processes. Appl. Environ. Microbiol. 68:3702-3707.
Schmidt, T. M., B. Pace, and N. R. Pace. 1991. Detection of DNA contamination in Taq polymerase.
Biotechniques 11:176-177.
Schwab, K. J., R. De Leon, and M. D. Sobsey 1995. Concentration and purification of beef extract mock
eluates from water samples for the detection of enteroviruses, hepatitis A virus, and Norwalk virus by reverse
transcription-PCR. Appl. Environ. Microbiol. 61:531-537.
Scott, T. M., T. M. Jenkins, J. Lukasik, and J. B. Rose. 2005. Potential use of a host associated molecular
marker in Enterococcm faecium as an index of human fecal pollution. Environ. Sci. Technol. 39:283-287.
Scott, T. M., S. Parveen, K. M. Portier, J. B. Rose, M. L. Tamplin, S. R. Farrah, A. Koo, and J. Lukasik.
2003. Geographical variation in ribotype profiles of Escherichia coll isolates from humans, swine, poultry,
beef, and dairy cattle in Florida. Appl. Environ. Microbiol. 69:1089-1092.
Scott, T. M., J. B. Rose, T. M. Jenkins, S. R. Farrah, and J. Lukasik. 2002. Microbial source tracking: current
methodology and future directions. Appl. Environ. Microbiol. 68:5796-5803.
Selander, R. K., J. M. Musser, D. A. Caugant, M. N. Gilmour, and T. S. Whittam. 1987. Population genetics
of pathogenic bacteria. Microb. Pathog. 3:1-7.
Seurinck, S., T. Defoirdt, W. Verstraete, S. D. Siciliano. 2005. Detection and quantification of the human-
specific HF183 Bacteroides 16S rRNA genetic marker with real-time PCR for assessment of human faecal
pollution in freshwater. Environ. Microbiol. 7:249-259.
Seurinck, S., W. Verstraete, and S. D. Siciliano. 2003. Use of 16S-23S rRNA Intergenic spacer region PCR
and repetitive extragenic palindromic PCR analyses of Escherichia coli isolates to identify nonpoint fecal
sources. Appl. Environ. Microbiol. 69:4942-4950.
Sghir, A., G. Gramet, A. Suau, V. Rochet, P. Pochart, and J. Dore. 2000. Quantification of bacterial groups
within human fecal flora by oligonucleotide probe hybridization. Appl. Environ. Microbiol. 66:2263-2266.
Shi, G. R. 1993. Multivariate data analysis in palaeoecolgy and palaeobiology - a review. Palaeogeography,
Palaeoclimatology, Palaeocology. 105:1999-234.
Simpson, J. M., J. W. Santo Domingo, and D. J. Reasoner. 2004. Assessment of equine fecal contamination:
the search for alternative bacterial source-tracking targets. FEMS Microbiol. Ecol. 47:65-75.
Simpson, J. M., J. W. Santo Domingo, and D. J. Reasoner. 2002. Microbial Source Tracking: state of the
science. Environ. Sci. Technol 36:5279-5288.
-------
Sinton, L. W., Hall, C. H., Lynch, P. A. and Davies-Colley, R. J. 2002. Sunlight inactivation of fecal indica-
tor bacteria and bacteriophages from waste stabilization pond effluent in fresh and saline waters. Appl.
Environ. Microbiol. 68:1122-1131.
Sinton, L. W., R. K. Finlay, and D. J. Hannah. 1998. Distinguishing human from animal faecal contamination
in water: a review. N. Z. J. Mar. Freshwat. Res. 32:323-348.
Smalla, K., H. Heuer, A. Gotz, D. Niermeyer, E. Krogerrecklenfort, and E. Tietze. 2000. Exogenous isola-
tion of antibiotic resistance plasmids from piggery manure slurries reveals a high prevalence and diversity in
IncQ-like plasmids. Appl. Environ. Microbiol. 66:4854-4862.
Sneath, P. H. A., and R. R. Sokal. 1973. Numerical Taxonomy. Freeman, San Francisco, CA.
Sokal, R. R., and C. D. Michener. 1958. A statistical method for evaluating systematic relationships. Uni-
versity of Kansas Science Bulletin 38:1409-1438.
Solo-Gabriele, H. M., M. A. Wolfert, T. R. Desmarais, and C. J. Palmer. 2000. Sources of Escherichia coli in
a coastal subtropical environment. Appl. Environ. Microbiol. 66:230-237.
Sorensen, T. 1948. A method of establishing groups of equal amplitude in plant sociology based on similarity
of species content and its application to analyses of the vegetation on Danish commons. Biol. Skr. 5:1-34.
Souza, V., M. Rocha, A. Valera, and L. E. Eguiarte. 1999. Genetic structure of natural populations of Esch-
erichia coli in wild hosts on different continents. Appl. Environ. Microbiol. 65:3373-85.
Stackelberg, P. E., E. T. Furlong, M. T. Meyer, S. D. Zaugg, A. K. Henderson, D. B. Reissman. 2004. Persis-
tence of pharmaceutical compounds and other organic wastewater contaminants in a conventional drinking-
water-treatment plant. Sci. Total Environ. 329:99-113.
Stahl, D. A., B. Flesher, H. R. Mansfield, and L. Montgomery. 1988. Use of phylogenetically based hybrid-
ization probes for studies of ruminal microbial ecology. Appl. Environ. Microbiol. 54:1079-1084.
Staley, J. T. and A. Konopka. 1985. Measurement of in situ activities of nonphotosynthetic microorganisms
in aquatic and terrestrial habitats. Ann. Rev. Microbiol. 39:321-346.
Statsoft, Inc. 2004. Electronic Statistics Textbook. Tulsa, OK: Statsoft WEB: http://www.statsoft.com/text-
book/stathome. html.
Stewart, C. S. and M. P. Bryant. 1988. In The rumen microbial ecosystem, P. N, Hobson, (ed.) Elsevier Ap-
plied Science, London, United Kingdom.
Stewart, J. R., R. D. Ellender, J. A. Gooch, S. Jiang, S. P. Myoda, and S. B. Weisberg. 2003. Recommenda-
tions for microbial source tracking: lessons from a methods comparison study. J. Water Health 1:225-231.
Stoeckel, D. M., C. M. Kephart, V. J. Harwood, M. A. Anderson, and M. Dontchev. 2004 Diversity of fecal
indicator bacteria subtypes: implications for construction of microbial source tracking libraries. American
Society for Microbiology General Meeting. New Orleans, LA.
-------
Stoeckel, D.M., M. V. Mathes, K. E. Hyer, C. Hagedorn, H. Kator, J. Lukasik, L. O'Brien, T. W. Fenger, M.
Samadpour, K. M. Strickler, and B. A. Wiggins. 2004. Comparison of seven protocols to identify fecal con-
tamination sources using Escherichia coli. Environ. Sci. Tech. 38:6109-6117.
Stull, T. L., J. J. LiPuma, and T. D. Edlind. 1988. A broad-spectrum probe for molecular epidemiology of
bacteria: ribosomal RNA. J. Infect. Dis. 157:280-286.
Suau, A., R. Bonnet, M. Sutren, J. -J. Godon, G. R. Gibson, M. D. Collins, and J. Dore. 1999. Direct analy-
sis of genes encoding 16S rRNA from complex communities reveals many novel molecular species within
the human gut. Appl. Environ. Microbiol. 65: 4799-4807.
Tajima, K., R. I. Aminov, T. Nagamine, H. Matsui, M. Nakamura, and Y. Benno. 2001. Diet-dependent
shifts in the bacterial population of the rumen revealed with real-time PCR. Appl Environ. Microbiol.
67:2766-2774.
Tarkka, E., H. Ahman and A. Siitonen. 1994. Ribotyping as an epidemiologic tool for Escherichia coli.
Epidemiol Infect. 112: 263-274.
Tartera, C., F. Lucena, and J. Jofre. 1989 Human origin of Bacteroides fragilis bacteriophages present in the
environment. Appl. Environ. Microbiol. 55:2696-701.
Tenover, F. C., R. D. Arbeit, R. V. Goering, P. A. Mickelsen, B. E. Murray, D. H. Persing, and B. Swamina-
than. 1995. Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis:
Criteria for bacterial strain typing. J. Clin. Microbiol. 33:2233-2239.
Tian, Y.Q., P. Gong, J. D. Radke, and J. Scarborough, 2002. Spatial and Temporal Modeling of Microbial
Containments on Grazing Farmlands. J. Environ. Quality 31:860-869.
Ting, W. T., D. Johnson, A. Holler, K. Tran, and C. Tseng. 2003. A study of the sources of E. coli contamina-
tion at Marquette Park Beach by random amplified polymorphic DNA typing. Annual Meeting of the Ameri-
can Society for Microbiology, Washington, DC.
Topp, E., M. Welsh, Y.-C. Tien, A. Dang, G. Lazarovits, K. Conn, and H. Zhu. 2003 Strain-dependent vari-
ability in growth and survival of Escherichia coli in agricultural soil. FEMS Microbiol. Ecol. 44:303-308.
Torgerson, W. S. 1952. Multidimensional scaling: 1. Theory and method. Psychometrika 17:401-419.
Torsvik, V., L. Ovreas and J. F. Thingstad. 2002. Prokaryotic diversity—magnitude, dynamics, and control-
ling factors. Science 296: 1064-1066.
Tsai, Y-L, J.Y. Le, and B.H. Olson. 2003. Magnetic bead hybridization to detect eriterotoxigenic Escherichia
coli strains associated with cattle in environmental water sources. Can. J. Microbiol. 49:391-398.
U.S. Environmental Protection Agency. 1986 Bacteriological ambient water quality criteria
for marine and fresh recreational waters. Federal Register 51(45), p. 8012-8016.
U.S. Environmental Protection Agency. 1984. Health Effects Criteria for Fresh Recreational Waters. Office
of Research and Development, Washington, DC. EPA-600/1-84-004. 44 pp.
-------
U.S. Environmental Protection Agency. 2000a. Improved enumeration methods for the recreational water
quality indicators: Enterococci and Escherichia coli. Office of Science and Technology, Washington, DC.
EPA/82 l/R-97/004. 55pp.
U.S. Environmental Protection Agency. 2000b. Atlas of America's polluted waters. Office of Water (4503F),
EPA 840-BOO-002. United States Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency. 200la. Method 1601. Male-specific (F+) and somatic coliphage in
water by two-step enrichment procedure. EPA Office of Water, Washington, DC. EPA 821-R-01-030. 40 pp.
U.S. Environmental Protection Agency. 2001b, Method 1602. Male-specific (F+) and somatic coliphage in
water by single agar layer (SAL) procedure. EPA Office of Water, Washington, DC. EPA 821-R-01-029. 38
pp.
Valsangiacomo, C., F. Baggi, V. Gaia, T. Balmelli, R. Peduzzi, and J. -C. Piffaretti. 1995. Use of amplified
fragment length polymorphism in molecular typing of Legionella pneumophilia and application to epidemio-
logical studies./ Clin. Microbiol. 33: 1716-1719.
Vancanneyt, T. M., A. Lombardi, C. Andrighetto, E. Knijff, S. Torriani, K. J. Bjorkroth, C. M. A. P. Franz,
M. R. F. Moreno, H. Rvets, L. De Vuyst, J. Swings, K. Kersters, F. Dellaglio, W.H. Holzapfel. 2002. Intra-
species genomic groups in Enterococcus faecium and their correlation with origin and pathogenicity. Appl.
Environ. Microbiol. 68:1381-1391.
Versalovic, J., V. Kapur, T. Koeuth, G. H. Mazurek, T. S. Whittam, J. M. Musser, and J. R. Lupski. 1995.
DNA fingerprinting of pathogenic bacteria by fluorophore-enhanced repetitive sequence-based polymerase
chain reaction. Arch. Pathol. Lab. Med. 119:23-29.
Versalovic, J., T. Koeuth, and J. R. Lupski. 1991. Distribution of repetitive DNA sequences in eubacteria and
application to fingerprinting of bacterial genomes. Nucl. Acids Res. 19:6823-6831.
Vesalovic, J., M. Schneider, F. J. de Bruijn, and J. R. Lupski. 1994. Genomic fingerprinting of bacteria using
repetitive sequence-based polymerase chain reaction. Methods MoL Cell. Biol. 5:25-40.
Vinje, J., S. J. G. Oudejans, J. R. Stewart, M. D. Sobsey, and S. C. Long. 2004. Molecular detection and ge-
notyping of male-specific coliphages by RT-PCR and reverse line blot hybridization. Appl Environ
Microbiol. 70:5996-6004.
Vogel, L., E. van Oorschot, H.M. Maas, B. Minderhoud, and L. Dijkshoorn. 2000. Epidemiologic typing of
Escherichia coli using RAPD analysis, ribotyping and serotyping. Clin. Microbiol. Infect. 6:82-87.
Wade, T. J., N. Pai, J. N. Eisenberg, and J. M. Colford, Jr. 2003. Do U.S. Environmental Protection Agency
water quality guidelines for recreational waters prevent gastrointestinal illness? A systematic review and
meta-analysis. Environ. Health Perspect. 111:1102-1109.
Wallis, J. L., and H. D. Taylor. 2003. Phenotypic population characteristics of the enterococci in wastewater
and animal faeces: implications for the new European directive on the quality of bathing waters. Water Sci.
Technol. 47:27-32.
-------
Wang, G., T. S. Whittam, C. M. Berg, and D. E. Berg. 1993. RAPD (arbitrary primer) PCR is more sensitive
than multilocus enzyme electrophoresis for distinguishing related bacterial strains. Nucl. Acids Res. 21:5930-
5933.
Wang, R. F., W. W. Cao, and C. E. Cerniglia. 1996. PCR detection and quantification of predominant anaero-
bic bacteria in human and animal fecal samples. Appl. Environ. Microbiol. 62: 1242-1247.
Wang, R. F., W. W. Cao, and C. E. Cerniglia. 1997. PCR detection of Ruminiococcus spp. in human and
animal faecal samples. Mol. Cell. Probes 11:259-265.
Wang, R.F., S. J. Kim, L. H. Robertson, and C.E. Cerniglia. 2002. Development of a membrane-array
method for the detection of human intestinal bacteria in fecal samples. Mol. Cell Probes 16: 341-350.
Welsh, J., and M. McClelland. 1990. Fingerprinting genomes using PCR with arbitrary primers. Nucl. Acids
Res. 18:7213-7218.
Wheeler-Aim, E., J. Burke, and A. Spain. 2003. Fecal indicator bacteria are abundant in wet sand at freshwa-
ter beaches. Water Res. 37:3978-3982.
Whitehead, T.R., and M.A. Cotta. 2000. Development of molecular methods for identification of Streptococ-
cus bovis from human and ruminal origins. FEMS Microbiol. Lett. 182:237-240.
Whitlock, J. E., D. T. Jones and V. J. Harwood. 2002. Identification of the sources of fecal coliforms in an
urban watershed using antibiotic resistance analysis. Water Res. 36:4273-4282.
Whitman, R. L., and M. B. Nevers. 2003. Foreshore sand as a source of Escherichia coli in nearshore water
of a Lake Michigan beach. Appl. Environ. Microbiol. 69:5555-5562.
Whitman, R. L., M. B. Nevers, G. C. Korinek, and M. N. Byappanahalli. 2004. Solar and temporal effects
on Escherichia coli concentration at a Lake Michigan swimming beach. Appl. Environ. Microbiol. 70:4276-
4285.
Whitman, R. L., D. A. Shively, H. Pawlik, M. B. Nevers, and M. N. Byappanahalli. 2003. Occurrence of
Escherichia coli and enterococci in Cladophora (Chlorophyta) in nearshore water and beach sand of Lake
Michigan. Appl. Environ. Microbiol. 69:4714-4729.
Whittam, T. S. 1989. Clonal dynamics of Escherichia coli in its natural habitat. Antonie Van Leeuwenhoek
55:23-32.
Whittam, T. S., H. Ochman, and R. K. Selander. 1983 Geographic components of linkage disequilibrium in
natural populations of Escherichia coli. Mol. Biol. Evol. 1:67-83.
Wiggins, B. A. 1996. Discriminant analysis of antibiotic resistance patterns in fecal streptococci, a method
to differentiate human and animal sources of fecal pollution in natural waters. Appl. Environ. Microbiol.
62:3997-4002.
Wiggins, B. A., R. W. Andrews, R. A. Conway, C. L. Corr, E. J. Dobratz, D. P. Dougherty, J. R. Eppard, S. R.
Knupp, M. C. Limjoco, J. M. Mettenburg, J. M. Rinehardt, J. Sonsino, R. L. Torrijos and M. E. Zimmerman.
-------
1999. Use of antibiotic resistance analysis to identify nonpoint sources of fecal pollution. Appl. Environ.
Microbiol 65:3483-3486.
Wiggins, B. A., P. W. Cash, W. S. Creamer, S. E. Dart, P. P. Garcia, T. M. Gerecke, J. Han, B. L. Henry, K.
B. Hoover, E. L. Johnson, K. C. Jones, J. G. McCarthy, J. A. McDonough, S. A. Mercer, M. J. Noto, H. Park,
M. S. Phillips, S. M. Purner, B. M. Smith, E. N. Stevens, and A. K. Varner 2003. Use of antibiotic resistance
analysis for representativeness testing of multiwatershed libraries. Appl. Environ. Microbiol. 69:3399-3405.
Wilbur, J. D., J. K. Ghosh, C. H. Nakatsu, S. M. Brouder, and R.W. Doerge. 2002. Variable selection for
high-dimensional multivariate binary data with application to microbial community DNA fingerprint analy-
sis. Biometric 58:378-386
Wilks, S. S. 1932. Certain Generalization in the Analysis of Variance. Biometrika 24:471-494.
Williams, J. G. K., A. R. Kubelik, K. J. Livak, J. A. Rafalski and S. V Tingey. 1990 DNA polymorphisms
amplified by arbitrary primers are useful as genetic markers. Nucl. Acids Res. 18:6531-6535.
Willshaw, G.A., H.R. Smith, T. Cheasty, P.G. Wall, and B. Rowe.1997. Vero cytotoxin-producing Escherich-
ia co// 0157 outbreaks in England and Wales, 1995: Phenotypic methods and genotypic subtyping. Emerg.
Infect. Dis. 3:561-565.
Wilson, 1. G. 1997. Inhibition and facilitation of nucleic acid amplification. Appl. Environ. Microbiol.
63:3741-3751.
Wood, J., K .P. Scott, G. Avgustin, C. J. Newbold, and H. J. Flint. 1998. Estimation of the relative abun-
dance of different Bacteroides and Prevotella ribotypes in gut samples by restriction enzyme profiling of
PCR-amplified 16S rRNAgene sequences. Appl. Environ. Microbiol. 64: 3683-3689.
Xiao L., R. Payer, U. Ryan, S. J. Upton. 2004. Cryptosporidium taxonomy: recent advances and implications
for public health, din. Microbiol. Rev. 17:72-97.
Xiao L., A. Singh, J. Limor J, T. K. Graczyk, S. Gradus, A Lai. 2001. Molecular characterization of Crypto-
sporidium oocysts in samples of raw surface water and wastewater. Appl. Environ. Microbiol. 67:1097-1101.
Xu, J., M. K. Bjursell, J. Himrod, S. Deng, L. K. Carmichael, H. C. Chiang, L. V. Hooper, J. I. Gordon..
2003. A genomic view of the human-Bacteroides thetaiotaomicron symbiosis. Science 299:2074-2076.
Yang, S., and R. E. Rothman. 2004. PCR-based diagnostics for infectious diseases: uses, limitations, and
future applications in acute-care settings. Lancet Infect. Dis. 4:337-348.
Zoetendal, E. G., C. T. Collier, S. Koke, R. I. Mackie, and H. R. Gaskins. 2004. Molecular ecological analy-
sis of the gastrointestinal microbiota: a Review. J. Nutr. 134:465-472.
-------
Glossary of Relevant Terms
16S and 23S rRNA genes- These are part of the ribosomal RNA genes that microbiologists use for the phy-
logenetic identification of bacteria. Due to the different levels of sequence conservancy they are also used
in the development of methods to detect bacteria in complex samples. The terms 16S rDNA and 23S rDNA
have recently been used to discriminate between the genes and rRNA transcripts.
Antibiotic Resistance Analysis (ARA) - Method that uses resistance to antibiotics to generate phenotypic
profiles of bacterial indentifier.
Clean Water Act (CWA) - An act passed by the U.S. Congress to control water pollution (formerly referred
to as the Federal Water Pollution Control Act of 1972). Public Law 92-500, as amended. 33 U.S.C. 1251 et
seq.
Clean Water Act Section 303(d) - annual report to Congress from EPA that identifies those waters for which
existing controls are not sufficiently stringent to achieve applicable water quality standards.
Clean Water Act Section 305(b) - biennial reporting requires description of the quality of the Nation's
surface waters, evaluation of progress made in maintaining and restoring water quality, and description of the
extent of remaining problems by using biological data to make aquatic life use support decisions.
Clone - A population of identical microorganisms derived from the same genetic lineage. All of the bacteria
in one culture, or one colony identical clones (unless a mutation occurs).
Coliphage - A bacterial virus (i.e., bacteriophage) that infects E. coli. Coliphages have been proposed as
potential indicators for the presence of enteric viruses in fecally impacted waters.
Confined Animal Feeding Operation (CAFO) - A lot or facility where animals have been, are, or will be
stabled or confined and fed or maintained for a total of 45 days or more in any 12 month period; and where
crops, vegetation, forage growth, or post-harvest residues are not sustained over any portion of the lot facil-
ity in the normal growing season and more than 1,000 animal units are confined at the facility or from 301
to 1,000 animal units are confined at the facility and it also meets one of the specific criteria addressing the
method of discharge.
Cosmopolitan - Describes strains that are found in more than one host species. "Transient" is sometimes
used synonymously.
DNA - Deoxyribonucleic acid. Encodes for the genetic material of living organisms with the exception of
some classes of viruses.
F+RNA - RNA male-specific coliphages.
False-negative - A source is not identified when it is actually present.
False-positive - A source is identified when it is not actually present.
Genotype - The analysis is based directly on the DNA of the organism. Ribotyping and PCR are both geno-
typic analyses.
-------
Library - In MST is normally refered to the group of fingerprints generated from microbial isolates col-
lected from the potential sources (i.e., animal feces) impacting a watershed. MST libraries should not be
confused with gene cloning libraries. Fingerprints are based on phenotypic traits (e.g., antibiotic resistance
analysis) or genotypic profiles (e.g., rep-PCR, ribotyping) of individual microbial strains
Library dependent methods (LDMs) - MST methods that require the development of a source library.
Library independent methods (LIMs) - MST methods that do not require the development of a source
library.
Microbial source tracking - Approach or approaches intended to identify the fecal sources impacting a wa-
ter system. Other terms that relate to MST are bacterial source tracking (when bacteria is the target), micro-
bial source identification, and fecal source identification.
Non-Point Source Pollution - pollution that occurs when rainfall, snowmelt, or irrigation runs over land or
through the ground, picks up pollutants, and deposits them into rivers, lakes, and coastal waters or introduces
them into ground water.
Point Source Pollution - Identifiable inputs of waste that are discharged via pipes or drains primarily (but
not exclusively) from industrial facilities and municipal treatments plants into rivers, lakes, and ocean.
Phenotype - Characteristics of an organism that rely on translation of genetic information into proteins. An-
tibiotic resistance patterns and carbon source utilization patterns represent phenotypes, as they are mediated
by enzymes and other proteins.
Quantitative PCR - Also known as real time PCR. The principles of QPCR are similar to those of conven-
tional PCR techniques with the exception that in each round of amplification the accumulation of PCR prod-
ucts is quantified using a fluorescence detector. Using host specific methods it is possible to quantify levels
of pollution of PCR from different animal types.
Restriction fragment length polymorphism (RFLP) - A type of polymorphism detectable in a genome by
the size differences in DNA fragments generated by restriction enzyme analysis.
Source identifier (SI) -A general category for the analytes used for MST. E. coli, enterococci, PCR bands
and caffeine are all examples of Sis.
Species/pattern/marker (SPM) - A specific species, pattern or marker that is indicative of a particular host
species. ARA patterns of enterococci, ribotypes of E. coli and the human-specific DNA band of Bacteroides
are examples of SPMs.
RNA - Ribonucleic acid. This polymer is primarily involved in protein synthesis.
Subtype - A microbial strain possessing a distinctive pattern or marker. Electrophoretic types, ribotypes, rep-
PCR patterns and antibiotic resistance patterns all define bacterial subtypes. Coliphage types I-IV are also
subtypes.
-------
Total Maximum Daily Load (TMDL) - TMDL is a calculation of the maximum amount of a pollutant that
a waterbody can receive and still meet water quality standards, and an allocation of that amount to the pol-
lutant's sources. Water quality standards are set by States, Territories, and Tribes. They identify the uses for
each waterbody, for example, drinking water supply, contact recreation (swimming), and aquatic life support
(fishing), and the scientific criteria to support that use. A TMDL is the sum of the allowable loads of a single
pollutant from all contributing point and nonpoint sources. The calculation must include a margin of safety
to ensure that the waterbody can be used for the purposes the State has designated. The calculation must also
account for seasonal variation in water quality. The Clean Water Act, section 303, establishes the water qual-
ity standards and TMDL programs.
Type I error - Occurs when a difference is identified that does not really exist (analogous to false-positive).
Type II error - Occurs when a difference that does exist is not identified (analogous to false-negative).
-------
Glossary of Acronyms
AFLP
ARA
ARCC
ARP
BMP
BOX-PCR
BST
CUP
DA
DFA
DGGE
rDNA
EcoRI
ER1C-PCR
FISH
Hindlll
ISR-PCR
MLEE
MRA
MST
NOAA
PCR
PFGE
PvuII
QPCR
rep-PCR
REP-PCR
RFLP
rRNA
TMDL
TRFLP
USDA
USEPA
USGS
Amplified fragment length polymorphism
Antibiotic resistance analysis
Average rate of correct classification
Antibiotic resistance profiling
Best management practices
Repetitive polymerase chain reaction using BOX primers
Bacterial source tracking
Carbohydrate utilization profiling
Discriminant analysis
Discriminant function analysis
Denaturing gradient gel electrophoresis
Ribosomal ribonucleic acid gene
Restriction endonuclease derived from Escherichia coli
Enterobacterial repetitive intergenic consensus sequences polymerase
chain reaction
Fluorescent in situ hybridization
Restriction endonuclease derived from Haemophilus influenzae
Intergenic spacer region polymerase chain reaction
Multilocus enzyme electrophoresis
Multiple resistance analysis
Microbial source tracking
National Oceanic and Atmospheric Administration
Polymerase chain reaction
Pulse filed gel electrophoresis
Restriction endonuclease derived from Proteus vulgaris
Quantitative PCR
Repetitive polymerase chain reaction
Repetitive extragenic palindromic sequence polymerase chain reaction
Restriction fragment length polymorphism
Ribosomal ribonucleic acid
Total maximum daily load
Terminal restriction fragment length polymorphism
United States Department of Agriculture
United States Environmental Protection Agency
United States Geological Survey
------- |