USF Libraries
USF Digital Collections

Frequency distributions of Escherichia coli subtypes in various fecal sources over time and geographical space

MISSING IMAGE

Material Information

Title:
Frequency distributions of Escherichia coli subtypes in various fecal sources over time and geographical space application to bacterial source tracking methods
Physical Description:
Book
Language:
English
Creator:
Anderson, Matthew A. ( Matthew Alexander )
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
fecal coliforms
indicator organisms
ribotyping
antibiotic resistance analysis
population biology
Dissertations, Academic -- Microbiology -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Bacterial source tracking (BST) methods often involve the use of phenotypic or genotypic fingerprinting techniques to compare indicator bacteria such as Escherichia coli isolated from unknown sources against a library of fingerprints from indicator bacteria found in the feces of various known source animals. The predictive capability of a library is based in part on how well the library isolates reflect the true population diversity of indicator bacteria that can potentially impact a water body. The purpose of this study was to compare the behavior of E. coli population structures in the feces of humans, beef cattle and horses across different parameters. Ribotyping and antibiotic resistance analysis were used to "fingerprint", or subtype E. coli isolates. Significantly greater diversity was observed in the E. coli population of horses compared to the human or beef cattle sampled. Subtype sharing between individuals from all host categories was infrequent, therefore the majority of E. coli subtypes were sampled from a single individual. The dominant E. coli populations of nine individuals (three per host source category) were monitored over time, which demonstrated that E. coli subtypes within a host individual vary on a monthly time frame, and an increase in the frequency of subtype sharing was noted between individuals within the same source group over time. The E. coli population of a single human that had just finished antibiotic treatment was studied on a daily basis for one month. The loss of an E. coli subtype with high antibiotic resistance was observed over time, however there was a single dominant E. coli subtype that was present at every sampling event during the entire month. Geographic distinctiveness of E. coli populations was investigated by sampling four herds located in different geographical regions. We observed that E. coli populations are not geographically distinct, but are somewhat individual-specific, as most E. coli isolates had a subtype that was found in a single individual. This study defines factors that should be considered when constructing a successful BST library, and suggests that E. coli may not be the appropriate indicator organism for BST.
Thesis:
Thesis (M.S.)--University of South Florida, 2003.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Matthew A. Anderson.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 117 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001447466
oclc - 54067850
notis - AJN3910
usfldc doi - E14-SFE0000206
usfldc handle - e14.206
System ID:
SFS0024902:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001447466
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 040114s2003 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000206
035
(OCoLC)54067850
9
AJN3910
b SE
SFE0000206
040
FHM
c FHM
090
QR41.2
1 100
Anderson, Matthew A.,
q (Matthew Alexander)
0 245
Frequency distributions of Escherichia coli subtypes in various fecal sources over time and geographical space
h [electronic resource] :
application to bacterial source tracking methods /
by Matthew A. Anderson.
260
[Tampa, Fla.] :
University of South Florida,
2003.
502
Thesis (M.S.)--University of South Florida, 2003.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 117 pages.
520
ABSTRACT: Bacterial source tracking (BST) methods often involve the use of phenotypic or genotypic fingerprinting techniques to compare indicator bacteria such as Escherichia coli isolated from unknown sources against a library of fingerprints from indicator bacteria found in the feces of various known source animals. The predictive capability of a library is based in part on how well the library isolates reflect the true population diversity of indicator bacteria that can potentially impact a water body. The purpose of this study was to compare the behavior of E. coli population structures in the feces of humans, beef cattle and horses across different parameters. Ribotyping and antibiotic resistance analysis were used to "fingerprint", or subtype E. coli isolates. Significantly greater diversity was observed in the E. coli population of horses compared to the human or beef cattle sampled. Subtype sharing between individuals from all host categories was infrequent, therefore the majority of E. coli subtypes were sampled from a single individual. The dominant E. coli populations of nine individuals (three per host source category) were monitored over time, which demonstrated that E. coli subtypes within a host individual vary on a monthly time frame, and an increase in the frequency of subtype sharing was noted between individuals within the same source group over time. The E. coli population of a single human that had just finished antibiotic treatment was studied on a daily basis for one month. The loss of an E. coli subtype with high antibiotic resistance was observed over time, however there was a single dominant E. coli subtype that was present at every sampling event during the entire month. Geographic distinctiveness of E. coli populations was investigated by sampling four herds located in different geographical regions. We observed that E. coli populations are not geographically distinct, but are somewhat individual-specific, as most E. coli isolates had a subtype that was found in a single individual. This study defines factors that should be considered when constructing a successful BST library, and suggests that E. coli may not be the appropriate indicator organism for BST.
590
Adviser: Harwood, Valerie J.
653
fecal coliforms.
indicator organisms.
ribotyping.
antibiotic resistance analysis.
population biology.
690
Dissertations, Academic
z USF
x Microbiology
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.206



PAGE 1

Frequency Distributions of Escherichia coli Subtypes in Various Fecal Sources Over Time and Geographical Space: Applicati on to Bacterial Source Tracking Methods by Matthew A. Anderson A thesis submitted in partial fulfillment of the requirements for the degree of Masters of Science Department of Biology College of Arts and Sciences University of South Florida Major Professor: Valerie J. Harwood, Ph.D. Earl McCoy, Ph.D. John Lisle, Ph.D. Date of Approval: November 21, 2003 Keywords: Fecal coliforms, Indicator Or ganisms, Ribotyping, Antibiotic Resistance Analysis, Population Biology Copyright 2003, Matthew Anderson

PAGE 2

Table of Contents List of Tables iii List of Figures iv Abstract vi Chapter 1. Introduction and Objectives 1 E. coli as an Indicator Organism 1 Bacterial Source Tracking 3 Non Library-Based BST Methods 4 Library-Based BST Methods 6 Methods for Defining the Population Structure of E. coli in the Feces of Animals 10 Population Structure of E. coli 14 Variability in Pathogenic E. coli Strains 14 Variability of E. coli Within Host Species 15 Population Structure of E. coli Within Individual Hosts 16 Using Diversity Indices to Measure Variability in E. coli Populations 18 Applying Knowledge of the Population Genetics of E. coli to BST 19 Host Specificity 20 Temporal Stability 21 Primary versus Secondary Habitats 22 Geographic Variability in E. coli Populations 23 Objectives 24 Chapter 2. One-Time Intensive Sampling of Feces from Cattle, Horses and Humans 26 Materials and Methods 26 Sample Collection 26 Ribotyping 27 Antibiotic Resistance Analysis (ARA) 29 Statistical Analysis 31 Results 34 Population Distribution of E. coli Subtypes within Source Categories 34 Pattern Sharing within and Between Source Categories 43 Discussion 48 i

PAGE 3

Chapter 3. Sampling of Individual Host Animals From Three Source Categories Over Time 52 Materials and Methods 52 Sample Collection and Processing 52 Results 54 Population Structure of E. coli within Individuals Over Time 54 Pattern Overlap Between Individuals within a Source Category Over Time 67 Ribotyping 67 Antibiotic Resistance Analysis 69 Discussion 72 Chapter 4. Intensive Sampling of E. coli from One Human for One Month 76 Materials and Methods 76 Sample Collection and Processing 76 Results 77 Distribution of Ribotypes and Antibiotic Resistance Profiles Found in Human X Over Time 77 Discussion 84 Chapter 5. Sampling of Cattle Herds From Four Geographic Regions 87 Materials and Methods 87 Sample Collection and Processing 87 Results 88 Population Structure of E. coli in Various Herds of Beef Cattle 88 Discussion 95 Chapter 6. Discussion-Applications for Bacterial Source Tracking 98 References 104 ii

PAGE 4

List of Tables Table 1 PCR program used for the synthesis of the 16S rDNA probe. 29 Table 2 Antibiotics and their concentrations used for antibiotic resistance analysis of E. coli isolates. 30 Table 3 Diversity measurements of the population structure of E. coli within individuals from different source categories using ribotyping. 35 Table 4 Significance values determined by comparing the means of all diversity measurements obtained from each source category. 36 Table 5 Diversity measurements of the population structure of E. coli within individuals from different source categories using antibiotic resistance analysis. 38 Table 6 Significance values determined by comparing the means of all Diversity measurements obtained from each source category. 39 Table 7 The proportion of E. coli subtypes shared between multiple source categories. 47 Table 8 Ribotype sharing between individuals within the same source category over time. 68 Table 9 Antibiotic resistance pattern sharing between individuals over time for humans (A), cattle (B) and horses (C). 71 Table 10 Antibiotic resistance patterns (ARPs) of E. coli sampled from human X. 80 Table 11 Diversity measurements of E. coli populations within beef cattle individuals from different herds. 89 Table 12 Percentage E. coli ribotypes shared between each of the herds. 92 Table 13 Chi-square values for comparison of the frequency of ribotype sharing between herds. 93 iii

PAGE 5

List of Figures Figure 1 Example of a similarity dendrogram of ribotypes created using the Dice coefficient.. 33 Figure 2 Equations for the parameters used to define E. coli population structure. 33 Figure 3 Accumulation curves representing the number of unique patterns versus sampling effort for horses, humans, and cattle. 42 Figure 4 Proportion of E. coli isolates that demonstrate pattern sharing between different individuals from the same source category. 44 Figure 5 The E. coli population structure over time within Human A. 56 Figure 6 The E. coli population structure over time within Human B. 57 Figure 7 The E. coli population structure over time within Human C 58 Figure 8 The E. coli population structure over time within Cow A. 61 Figure 9 The E. coli population structure over time within Cow B. 62 Figure 10 The E. coli population structure over time within Cow C. 63 Figure 11 The E. coli population structure over time within Horse A. 64 Figure 12 The E. coli population structure over time within Horse. 65 Figure 13 The E. coli population structure over time within Horse C. 66 Figure 14 A ribotype gel using Hind III representing isolates from the first two days of sampling from human X. 77 Figure 15 A ribotype gel using Pvu II representing isolates from the first two days of sampling from human X. 77 iv

PAGE 6

Figure 16 Distribution of ARPs of E. coli isolates sampled from human X. 78 Figure 17 Similarity dendrogram for antibiotic resistance patterns of E. coli isolated from human X. 81 Figure 18 Distribution of antibiotic resistance patterns over time for E. coli isolated from human X. 82 Figure 19 Distribution of ribotypes over time for E. coli isolated from human X. 82 Figure 20 Frequency of observations of E. coli ribotype/ARP combinations in human X. 83 Figure 21 Sharing of E. coli ribotypes within and between herds, expressed as percentages of total E. coli isolates isolated from a given herd. 91 Figure 22 A similarity dendrogram of ribotypes generated from a subset of E. coli isolated from four cattle herds. 94 v

PAGE 7

Frequency Distributions of Escherichia coli Subtypes in Various Fecal Sources Observed Over Time and Geographical Space: Application to Bacterial Source Tracking Methods Matthew A. Anderson ABSTRACT Bacterial source tracking (BST) methods often involve the use of phenotypic or genotypic fingerprinting techniques to compare indicator bacteria such as Escherichia coli isolated from unknown sources against a library of fingerprints from indicator bacteria found in the feces of various known source animals. The predictive capability of a library is based in part on how well the library isolates reflect the true population diversity of indicator bacteria that can potentially impact a water body. The purpose of this study was to compare the behavior of E. coli population structures in the feces of humans, beef cattle and horses across different parameters. Ribotyping and antibiotic resistance analysis were used to fingerprint, or subtype E. coli isolates. Significantly greater diversity was observed in the E. coli population of horses compared to the human or beef cattle sampled. Subtype sharing between individuals from all host categories was infrequent, therefore the majority of E. coli subtypes were sampled from a single individual. The dominant E. coli populations of nine individuals (three per host source category) were monitored over time, which demonstrated that E. coli subtypes within a host individual vary on a monthly time frame, and an increase in the frequency of subtype sharing was noted between individuals within the same source group over time. vi

PAGE 8

The E. coli population of a single human that had just finished antibiotic treatment was studied on a daily basis for one month. The loss of an E. coli subtype with high antibiotic resistance was observed over time, however there was a single dominant E. coli subtype that was present at every sampling event during the entire month. Geographic distinctiveness of E. coli populations was investigated by sampling four herds located in different geographical regions. We observed that E. coli populations are not geographically distinct, but are somewhat individual-specific, as most E. coli isolates had a subtype that was found in a single individual. This study defines factors that should be considered when constructing a successful BST library, and suggests that E. coli may not be the appropriate indicator organism for BST. vii

PAGE 9

Chapter 1. Introduction and Objectives E. coli as an Indicator Organism When water has been polluted by fecal material it may pose a health risk to humans who become exposed to this water either through drinking or recreational use (i.e swimming). Waters contaminated by fecal material can also present a threat to the food industry, particularly in the case of shellfish. Fecal material, particularly that of human origin, may contain pathogenic bacteria, such as Salmonella spp, and Shigella spp, and pathogenic viruses, such as enteroviruses and rotaviruses, which cause human disease (6). Because of this danger, methods have been established to monitor the sanitary conditions of water by measuring the levels of indicator microorganisms in water samples (1). Water samples cannot feasibly be tested for every enteric pathogen that could be present because: 1) standardized laboratory tests are not available for many pathogens, and 2) previously undiscovered pathogens may be present that cannot be detected by any currently devised test (60). The indicator microorganism represents a relatively rapid, simple way to test water for the possibility of fecal pollution; however, not every organism that could be found in feces can act as a successful indicator microorganism. An effective indicator microorganism must have several qualities (44): 1) the organism should be non-pathogenic, 2) the organism should correlate well with the presence of enteric pathogens, 3) the organism needs to survive as long, if not longer, then the 1

PAGE 10

hardiest waterborne pathogen, 4) the organism should survive but not be able to reproduce in natural waters, and finally, 5) the organism should be easy to test for and the method for detection should be rapid. Early in the 20 th century it was suggested that members of the coliform group (total coliforms) be used as bacterial indicators for fecal pollution in waters (2). The coliform group is characterized as facultatively anaerobic, nonspore forming, gram negative rods that are found in the intestines of warm-blooded animals. Coliforms produce gas from the fermentation of lactose at 37C within 48 h. Members of this group include species of the genera Klebsiella, Citrobacter, Enterobacter, and Escherichia (6). A subgroup of the total coliforms is the fecal coliform group. Fecal coliforms differ from the total coliform group by their ability to grow at 44.5 C and includes the genus Escherichia and some members genus Klebsiella (6). The fecal coliform group, most notably Escherichia coli, is one of the primary indicator organisms used when monitoring for fecal pollution in fresh and potable waters (18). E. coli has many characteristics of a good indicator. It is generally non-pathogenic, there are standard methods of enumeration which are simple and rapid (18-24 hours), and it is found in large numbers in the feces of all warm blooded animals and some cold blooded animals (32). Fecal coliforms have been used as a standard of monitoring waters for fecal pollution; however recent studies have shown that fecal coliforms may have certain attributes that make them less than ideal indicators. Some studies have shown that fecal coliforms may be able to persist and even multiply in tropical and subtropical waters and sediment (9, 74) while other studies have demonstrated that fecal coliforms do not 2

PAGE 11

correlate well with the survival of pathogenic viruses (21, 45, 87) These results have led scientists to investigate the potential of other organisms as indicators of fecal pollution in waters. Bacterial Source Tracking The ability to identify the presence of fecal pollution in waters is only the first step in water quality assessment and control. One problem with using indicator microorganisms to detect fecal pollution is that generally, as in the case of E. coli, the organism is found in the feces of many different animals. More information about the source of the fecal pollution is needed to make reliable inferences about the probability of the presence of human pathogens. There are many benefits to knowing the source of fecal pollution. When assessing microbial risk, it is important to know if the fecal pollution is of human origin because human feces generally contains more pathogenic organisms than animal feces (41). Locating the source of fecal pollution can also assist in cleaning up and possibly preventing any further contamination of the impacted water body (28). Recent studies have attempted to develop a method capable of tracing indicator microorganisms found in polluted waters back to their source (i.e. animal species or point source). These methods are known as bacterial source tracking (BST) or microbial source tracking (MST) (72). To date, no standard or universally accepted BST method has been developed. The existing methods can be divided into two broad categories; library-based techniques and non library-based techniques. 3

PAGE 12

Non Library-Based BST Methods Non library-based BST methods are predicated on the hypothesis that the chosen indicator microorganism demonstrates some degree of host specificity. One of the first BST methods to be developed compared the concentration of fecal coliforms found in a water sample to the concentration of fecal streptococci present (20). The technique, known as the fecal coliform to fecal streptococci ratio (FC/FS), was used to determine animal vs. human origin of fecal pollution. The authors hypothesized that humans tend to have much larger concentrations of fecal coliforms than fecal streptococci in their feces, and that the ratio is reversed in all other animals. The authors proposed that a FC/FS ratio < 0.7 indicated pollution from a non-human host, and ratios 4 indicated human origin. The advantage to this technique is the simplicity and speed with which it is conducted, as results are obtained in 24-48 hrs. A problem with this technique is that studies have demonstrated fecal coliforms and fecal streptococci do not have equal survival rates in marine waters (30, 47). Meaning that the resulting ratio would not be applicable unless the waters were sampled at the moment of pollution. Furthermore, it has been observed (17, 73) that the proposed FC/FS ratios do not hold true for all animal species and that different individuals of the same species can have very different FC/FS ratios. Some humans have with large amounts of fecal streptococci in their feces, while animals have been observed with very high levels of fecal coliforms. For these reasons, the FC/FS ratio is not used today. It has been argued that the survival rate of current indicator bacteria does not correlate well with the survival of viruses in water (21, 45, 87). It is therefore important to develop BST methods using organisms that have similar, if not greater survival rates 4

PAGE 13

than pathogenic viruses in aquatic environments. One such BST method uses male specific (F+) RNA coliphages as the indicator organism of fecal pollution. F+ coliphages are viruses specific to coliforms and attach to the sex pili of these bacteria (77). The F+ RNA coliphages can be divided into four major groups (I, II, III, IV) by serotyping (58). Testing fecal samples from many different animals, including humans, established a link between the source of the coliphage and its serotype. Human feces tend to be populated with coliphages of serotypes II or III. Animal feces are usually dominated by coliphages of serotype I and sometimes IV. An exception is pig feces, which also contain serotype II coliphages. Using this information it may be possible to identify the source of fecal pollution by serotyping the F+ RNA coliphage found in polluted waters (58). A study by Hsu et al. attempted to improve upon this technique by using oligonucleotide probes to identify these four groups rather than serotyping (37). The study found that genotypic probing was as accurate as serotyping, however genotypic probing was more rapid and, more importantly, less expensive than serotyping. An advantage to using this method is that the process only involves one major step (serotyping or probing) in order to determine the source of pollution. The disadvantage to using this method is that discrimination can only be achieved on the human vs. nonhuman level; currently, identification of a specific animal source is not possible. Another disadvantage to this method is that studies have shown that not all individuals harbor F+ RNA coliphage in their feces (34). Meaning that although fecal pollution may be identified using standard fecal coliform counts, if there are no coliphage present, the source cannot be determined. 5

PAGE 14

Library-Based BST Methods Library-based methods work by subtyping (fingerprinting) the test organism, and creating a known source library of these fingerprints which acts as a predictive tool to determine the source of fecal pollution. The first step in making the known source library is to collect fecal samples from a host species, then isolate bacteria from a number of different individuals within the specific host species. These isolates are then subtyped with the chosen technique, and their patterns are entered in the library. The library should consist of isolates collected from source animals that could potentially pollute the watershed being studied. Enough isolates should be collected from each source category in order to adequately represent the dominant fingerprints in the indicator organism population. Once the library is complete it acts as a predictive tool to determine the origin of fecal pollution found in the watershed. To accomplish this task, environmental isolates are collected from the polluted waters and typed using the chosen technique. These patterns are compared to those in the library and the most probable source of the isolate is determined. Although the basic process is similar in all library based methods, many different subtyping techniques have been developed. These techniques can be separated into two types: phenotypic techniques and genotypic techniques. One of the most frequently utilized phenotypic BST techniques is known as antibiotic resistance analysis (ARA) (84). ARA exposes each isolate to a number of different antibiotics at varying concentrations. By scoring growth on the various antibiotic concentrations, a fingerprint for each isolate is created. The theory behind ARA is that different host sources are exposed to different types of antibiotics over time, 6

PAGE 15

therefore the E. coli population within these hosts should display a level of host specificity based on their susceptibility to antibiotics. There have been a number of studies using ARA that have successfully discriminated between different host sources (28, 33, 80, 86). These studies show that ARA is highly successful at distinguishing human vs. nonhuman fecal pollution, and can distinguish between specific animal sources at a rate that is significantly higher than random (80). Another phenotypic BST technique is carbon utilization profiles (CUP). This technique, which has been accomplished using the Biolog system in the U.S. (7) and the PhenePlate system in Europe (78), assesses growth on a number of different carbon sources (as many as 96). Hagedorn et al (27) conducted a study using CUPs to determine sources of fecal pollution. The results of this study were very similar to those found for ARA, i.e. the technique was successful at classifying human vs. nonhuman pollution (~88%) but this classification rate dropped when trying to distinguish between specific source animals. Phenotypic BST techniques have the advantage of being relatively inexpensive and rapid compared to genotypic techniques (80), which allows the development of a large, representative, library in a relatively short period of time. Genotypic typing techniques make direct use of an organisms genetic material (i.e. DNA) in order to generate subtypes. One such method is known as ribotyping (59). Ribotyping is a molecular method that targets the ribosomal RNA genes, i.e. the 16S rRNA gene, in order to find genetic variation within the gene and in surrounding DNA. The 16S rRNA gene is so highly conserved, that organisms from a single species have nearly identical 16S rRNA genes and the genetic variation within a species is located in the surrounding DNA due to mutations, duplications, deletions, or horizontal genetic 7

PAGE 16

transfer between organisms. Ribotyping relies on the genetic variation in these regions to discriminate between members of the indicator organism population. Restriction enzymes digest genomic DNA, resulting in many different sized DNA fragments, which are separated by electrophoresis and blotted onto a membrane. The membrane is probed using labeled 16S rDNA probes. Any fragment containing a portion of the 16S rRNA gene will hybridize with the probe and be detected. The result is a banding pattern determined by the genetic structure of the test organism. Studies have been done using ribotyping as a method of discriminating between indicator organisms from different sources (4, 10, 59). These studies have shown that ribotyping can potentially be useful for BST. Carson et al. (10) used a library of 287 isolates from human and various animal sources to determine the capability of their library to discriminate between sources. Their study found that greater than 90% of the isolates were correctly classified into the proper source categories, however, the library used in this study was very small (287 isolates). It has been suggested that small libraries do not adequately represent the true diversity of the indicator organism population, resulting in misleadingly high correct classification rates for the libraries (80). One major advantage to using ribotyping is its ability to be automated (8), which is faster, up to five times as fast, and more reproducible due reduced variability between gels caused by human error. However, with automation comes an increase in cost and the inability to adjust the method at the different steps in the process. Other genetic typing techniques have also been applied to BST. Two of these methods are repetitive extragenic palindromic PCR (REP-PCR) (11, 16) and amplified fragment length polymorphism (AFLP) (25, 29). Like ribotyping, these techniques 8

PAGE 17

utilize organisms DNA to create a fingerprint. REP-PCR primers target the short repetitive sequences found in the genome of the organism. The multiple amplicons generated are separated by electrophoresis, forming a pattern comprised of 20-30 bands in E. coli (11). Two restriction enzymes are used in AFLP to digest genomic DNA. The digested DNA is then ligated to adaptors specific for the restriction enzymes. The different sized fragments of ligated DNA are amplified by two rounds of PCR. The first round of PCR uses primers specific for the adaptor regions. The second round of PCR is performed with fluorescently labeled primers, which allow visualization of the banding pattern. Amplicons are separated by gel electrophoresis, creating a banding pattern representative of the genetic variability of an organism (25, 29). All of the library-based methods discussed thus far have demonstrated some ability to distinguish between bacteria isolated from different fecal sources, however many of the studies that tested the effectiveness of these methods were flawed. A shared problem in many of these studies is the size of the library constructed. Generally, a large library (1000s of isolates) is needed in order to adequately represent the diversity of subtypes that could be in the test organism (80). Phenotypic methods have the advantage of being able to produce these large libraries at a fraction of the cost necessary for genotypic methods. Because of this difference, BST studies that use phenotypic typing often have libraries with thousands of isolates (33, 80, 86), while studies using genotypic typing usually have libraries of only a few hundred isolates (4, 10, 16, 59), and these isolates are often unevenly distributed between sources. 9

PAGE 18

Another criticism of many BST studies is the manner in which the librarys efficacy as an effective predictive tool is tested. Most studies have utilized a self-cross of the library, in which library isolates are the test data set and the calibration data set, to assess the accuracy of the library (10, 11, 27, 33, 84). It has been argued that this is not an effective test of the librarys ability to identify the source of an environmental isolate because this method assumes that the library is an accurate representation of the total diversity found in the test organism populaiton, which generally is not true (80). Whitlock et al. suggested that isolates from a known source that are not present in the library should be used as test organisms to determine the accuracy of a BST library (80). The final flaw common to most BST methods is the lack of studies that provide information on the population structure of the test organism within different host species. Such information will be useful for improving current BST methods. Methods for Defining the Population Structure of E. coli in the Feces of Animals Many of the library-based BST methods use E. coli as the test organism to create a library of fingerprints from known sources (10, 16, 33, 80). The choice of E. coli is prompted in part by the fact that it is one of the standard indicator organisms used to detect fecal pollution in recreational and drinking waters in the United States and other countries (1) and by the fact that E. coli is found in the feces of all warm blooded animals. It is important to observe the population structure of E. coli within hosts and to investigate the behavior of E. coli populations across different variables (i.e. between host sharing, temporal stability, and geographic variability) in order to maximize the success rate of BST methods. 10

PAGE 19

Many studies have investigated the population genetics of E. coli (14, 22, 26, 48, 57, 69, 81). One of the first methods used to broadly define the population structure of E. coli was serotyping, which assesses differences in specific surface antigens of the cell. In the case of E. coli the standard antigens used are the lipopolysaccharide (O), the capsular (K) and the flagellar (H) antigen (40). In a series of studies Orskov et al. (53) and Orskov and Orskov (54-56) found as many as 170 different O antigens, 71 different K antigens, and 56 different H antigens, along with many combinations of the three. This was the first data to demonstrate the extreme antigenic variability demonstrated by members of this species. Because of this large diversity in antigen combinations, serotypes became one of the first operational taxonomic units (OTU) to be used when studying E. coli population structure (65). As more studies were conducted it was found that unique serotypes do not represent genetically distinct E. coli. This was discovered through a technique known as multilocus enzyme electrophoresis (MLEE). MLEE was first used to study population genetics of eukaryotic organisms (31, 42). In this process a fingerprint is developed by measuring the electrophoretic mobility of 12-20 cellular enzymes (66). The mobility of the enzymes through the gel is influenced by differences in their molecular weights. These different allelic variants, known as electromorphs, of the same enzyme will demonstrate electrophoretic bands at different positions on a gel. The presence or absence, as well as the allelic variation, of the enzymes being used for the study will create a unique fingerprint for the test organism known as its elecrophoretic type (ET). Based on the MLEE results, calculations can be made to determine the genetic distance between different ETs, which helps to establish the genetic lineage of the test organism. 11

PAGE 20

In 1973 Milkman first used MLEE to demonstrate genetic variation in E. coli (48). The study utilized five different enzymes and found that MLEE could be used as a method to show variability within E. coli. Subsequent studies compared the discriminatory capability of MLEE and serotyping. Ochman and Selander (52) used MLEE of 12 enzymes on 142 E. coli isolates characterized as having the K1 serotype. These isolates were obtained from human hosts located in both Europe and the United States. Their results revealed that fourteen distinct electrophoretic types were found within the K1 serotype. A similar study was done by Caugant et al. (13) that directly compared serotyping and MLEE using 261 E. coli isolates of various known serotypes. The results of this study demonstrated that the amount of genetic diversity found within a single antigen serotype using MLEE approaches the amount of diversity one would find if they randomly chose strains, demonstrating that serotyping lacks significant discrimination. In one particular O serotype (O8), six different electrophoretic types were identified out of six isolates. Not all serotypes presented this extreme variability, but all single antigen serotypes had at least two different electrophoretic types. Another important conclusion of this study was that isolates were not serotyped could be typed by MLEE. Other studies were conducted which confirmed these results, and established MLEE as the better tool for distinguishing genetic variability in E. coli (46, 50, 62, 67). Multilocus enzyme electrophoresis remains today a useful method for assessing variability in E. coli. Recently, additional genetic typing methods, such as ribotyping and restriction fragment length polymorphism (RFLP) of specific genes, have been applied to studies of the population structure of E. coli. In 1990 Arthur et al. (3) compared the 12

PAGE 21

ability of RFLP and MLEE to distinguish genetic variants of E. coli. RFLP analysis of the rrn (rRNA) operons and MLEE of thirteen enzymes were conducted on E. coli isolated from 20 patients with urinary or biliary tract infections. RFLPs of the rrn operon correlated very well with the electrophoretic types of MLEE; isolates with identical RFLPs had identical or closely related electrophoretic types. The methods were judged equally discriminatory. In 1995, Maslow et al. (46) performed a similar study using 187 E. coli isolates from the bloodstream of human patients to compare serotyping, MLEE, and restriction-fragment length polymorphism (RFLP) of the ribosomal RNA operon. MLEE provided the most discrimination of all three techniques, followed by RFLP and then serotyping; however, all identical RFLP patterns were shared by closely related electrophoretic types, leading to the conclusion that RFLP of the ribosomal RNA genes is just as effective as MLEE at discriminating between closely related variants of E. coli. These data were confirmed in a study by Silveira et al. using E. coli isolated from diseased birds. The study analyzed 69 E. coli isolates by MLEE (five enzymes) and RFLP of the ribosomal RNA operon. MLEE distinguished 33 different electrophoretic types, while twenty-four different RFLP patterns were observed. Although MLEE was more discriminatory than RFLP, all RFLP patterns were shared by closely related electrophoretic types (70). These studies indicate that subtyping E. coli populations by typing techniques such as MLEE and RFLP, provides a more accurate representation of the variability in E. coli populations than serotyping. 13

PAGE 22

Population Structure of E. coli Once it was determined that there were methods to effectively discriminate between genetically distinct E. coli subtypes, many studies were performed to determine the variability of E. coli in the feces of animals. These studies fall under three categories: 1.) investigations of the genetic variability in pathogenic E. coli, 2.) investigations of the genetic variability found in the natural E. coli populations of different host species, and 3.) studies that take a closer look at the population structure of E. coli found in the feces of individual hosts and the interactions of these E. coli populations between different hosts. In the third category, population structure is defined as the number of different subtypes, as well as their distributions, in the E. coli population of single individuals. Variability in Pathogenic E. coli Strains E. coli is a member of the natural intestinal flora of all healthy warm blooded animals. However, some E. coli strains have the potential to cause disease in and death of their hosts. Selander et al. (67) typed 63 E. coli isolates from infants with neonatal septicemia or meningitis by MLEE (21 enzymes). They found 39 unique electrophoretic types, demonstrating high genetic diversity. Isolates causing the same disease tended to be closely related. Other studies were conducted on different human pathogenic strains of E. coli that confirmed these results (50, 62). Not all studies have been limited to human pathogenic E. coli. Silveira et al. (70) studied the genetic diversity found in avian pathogenic E. coli strains by MLEE (five enzymes) and RFLP and demonstrated similar relationships between disease symptoms and genetic subtypes. All of these studies 14

PAGE 23

demonstrated that subtype diversity in pathogenic E. coli is high, and that strains of E. coli causing the same disease cluster when measured using genetic distance, suggesting that they share a lineage. Variablity of E. coli Within Host Species Pathogenic E. coli strains represent a small fraction of the total E. coli species. Study of the diversity and distribution of commensal E. coli strains provides insight into the evolution of different genetic lineages, how their distribution changes over geographic distance, and will help refine methods of water quality assessment like bacterial source tracking. Studies of the variability in natural E. coli populations usually fall within two categories: 1) studies that measure variation within a species and 2) studies that measure variation within individual hosts. Studies that investigate the population structure of E. coli present a broad perspective of the genetic variability that can be found within different hosts. Whittam et al. (83) performed a study using MLEE (12 enzymes) to assess the genetic variability found in 1,705 E. coli isolates of human origin. These isolates were collected from the feces or urine of multiple human hosts from different regions of North America and pooled together for analysis. Of the 302 different electrophoretic types identified, three major groups were found using principal components analysis. Because the isolates from individual hosts were pooled, within-individual variability could not be determined. It is therefore, not clear whether the observed variability in E. coli was widely distributed among individuals, or was associated with a small subset of the hosts. 15

PAGE 24

Similar results were obtained by Gordon in his study of the E. coli population of feral mice in two regions of Australia (22). This is a notable study because it dealt with the E. coli of a host species other than human. One isolate was obtained from the small intestine of each mouse. Over a period of seven months, 447 E. coli isolates were typed by MLEE (13 enzymes) and randomly amplified polymorphic DNA (RAPD) analysis. Fifty different RAPD genotypes and 60 unique ETs were observed, demonstrating a level of variability similar to that found for human E. coli populations in other studies (15, 83). It is important to note that the data in these previous studies represent an overview of the E. coli variability in a host species without looking specifically at how this variability is distributed among individual hosts. Population Structure of E. coli Within Individual Hosts Many studies of the population genetics of E. coli pool isolates collected from individuals in order to observe the genetic variability of the E. coli populations from within the entire host species (12, 22, 39, 68, 83). Far fewer studies have investigated the structure of the E. coli population found in individual animals and how these populations interact with other individuals of the same host species. Little is known about the short-term (i.e. month-to-month) stability of E. coli populations in individual hosts. Understanding the population structure of E. coli at the individual host level can lead to better models of the evolutionary genetics of natural E. coli populations. One of the first studies to investigate the population genetics of E. coli from an individual host was performed by Caugant et al. in 1981 (15), in which a single human was sampled over an 11-month period. Escherichia coli was isolated from fecal swabs at 16

PAGE 25

varying intervals and was typed by MLEE (15 enzymes). This study demonstrated a high level of subtype variability in the E. coli population of this single human through time. The E. coli population turned over every two to four weeks, and new ETs were observed at each sampling session. A total of 550 isolates was collected during the study, and 53 different ETs were observed. Two ETs persisted over a number of sampling dates, and these were designated resident strains. The authors hypothesized that non-resident E. coli types must be immigrants that are entering the population through food or contamination and named these transients (14). One disadvantage to the Caugant et al. study is that only one human was sampled. A follow up study (14) assessed the E. coli population variability of individual members of five different families, in two different states (New York and Massachusetts), by subtyping E. coli isolated from the feces by MLEE (15 enzymes). Members of one family in New York were sampled twice at an interval of three months, and all other families were sampled once. The E. coli population of one of the human individuals from New York demonstrated a very high level of genetic diversity in their E. coli population. The remaining family members demonstrated very low genetic diversity in their E. coli populations and many of these remaining members demonstrated only one ET making up their entire E. coli population. These results demonstrated that humans generally have an E. coli population of low genetic diversity, however a small proportion of human individuals harbor highly variable E. coli populations (13). Whittam performed a study of the E. coli populations found in avian hosts (81). Escherichia coli isolates (n = 280) from five different domestic birds collected at two sampling events per bird were typed by MLEE (13 enzymes). On average, birds 17

PAGE 26

harbored diverse E. coli populations comprised of 4-5 ETs per individual out of twenty-five isolates. This study is one of the few to investigate the variability of E. coli population within individuals from a host species other than human. Using Diversity Indices to Measure Variability in E. coli Populations Various techniques have been discussed that subtype E. coli isolates in order to determine the structure of E. coli populations. This type of information can be analyzed by diversity indices, which are based on formulas that take into account both the number of different subtypes found within a population as well as the relative abundance of these subtypes. No one diversity index is universally accepted; however, Hills diversity numbers are a well known, and well studied, family of diversity indices (35). Three different indices that are a part of Hills diversity measurements are the richness estimator (S), the Shannon index (H), and the Simpson index (). The richness estimator represents the all subtypes within a population, including the rarest subtypes but does not take into account the abundance of these subtypes. Shannon's index is a measure of the difficulty in predicting the identity of subsequent subtypes. In general terms, Shannons index can be used as a rough measure of the abundant subtypes in a population. This diversity measurement takes into account both the total number of subtypes and their frequency within the population, and is one of the most widely used diversity indices. Simpsons index is a measure of the most abundant subtypes in a population. It represents the probability that two subtypes chosen at random will be the same. Because it is a probability estimate, the Simpsons value is always between one and zero. Often, the reciprocal of the Simpson value (1/is presented to give a better 18

PAGE 27

idea of the change in the diversity within a population. When using the reciprocal, increasing values signify increasing diversity within the population (43). Diversity indices have been used as tools in ecological studies to measure the diversity of animal species within a given area. More recently, microbiological studies have used these indices to measure the diversity of bacterial populations (4, 36, 71). A study performed by Avery et al. (4), used Simpsons index to measure the diversity of E. coli O157 subtypes determined by pulsed field gel electrophoresis (PFGE) and ribotyping. The diversity measurement was used to compare the two methods and determine which method had a greater discriminatory ability. Applying Knowledge of the Population Genetics of E. coli to BST The typing techniques outlined above were originally applied to E. coli as an attempt to understand variability in the population dynamics of this organism, generally in large groups of animals. This information should also be useful in refining the use of E. coli as an indicator organism for fecal pollution, and to improve the predictive capabilities of BST methods. Although there are many different BST methods, most of them are based on the same assumptions. These assumptions are especially true for library-based methods. Gordon lists four assumptions that must be met for a successful library-based BST method regardless of the indicator organism being used for construction of the library (23): 1) the population structure of the organism must show some degree of host specificity, 2) the population structure of the organism should be stable through time, 3) the genetic composition of the population should be the same in both the fecal 19

PAGE 28

environment and the external environment (i.e. soil, water), 4) the indicator organism population should demonstrate some geographic distinctiveness, with populations being distinct to specific regions. Using the information that has been gathered by previous studies on the genetic structure of E. coli populations as a base, several studies have been performed to determine the validity of these assumptions for E. coli. Host Specificity In order for BST to be effective, the microorganism being tested must show some host specificity, ideally demonstrating some overlap between individuals from the same source category. In a previously mentioned study, the E. coli population of feral mice was subtyped by MLEE (12 enzymes) and 60 ETs were observed out of 447 E. coli isolates (22). Although a large number of subtypes were observed, 48% of all E. coli isolates shared only three ETs. Another study (19) demonstrated the sharing of E. coli subtypes within host species using the restriction enzyme Xba I to generate restriction endonuclease digestion profiles (REDP) of the genomic DNA of E. coli O157:H7, which were isolated from 29 cattle located on different farms in Wisconsin. They found 20 different REDPs among the E. coli O157:H7 isolates, and observed that some REDPs were shared between multiple cows on the same farm. Collectively, these studies suggest that there is some degree of subtype sharing between individuals from a single host species. Few studies have rigorously investigated the extent to which E. coli subtypes are shared between host species. If E. coli subtypes were shared equally among many 20

PAGE 29

different species, BST methods would not be useful for detecting the source of fecal contamination in water. One study demonstrated overlap of E. coli subtypes between families and their pets (14). The E. coli isolates of five human families and their pets, were typed by MLEE (15 enzymes). E. coli subtypes were more frequently shared within families (including pets) than between families small percentage of the shared isolates within families were found in both human members and their pets. Specifically one ET was shared between humans and their dog, one ET was shared between humans and their cat, and three ETs were shared between cats and dogs. Temporal Stability Gordon (23) states that for an organism to be used successfully in a BST method the microorganisms population structure must remain stable through time. It has been demonstrated in a number of studies that E. coli populations from a host species do not remain stable over time (15, 19, 22, 39). One study investigated the E. coli population within a human host by MLEE (13 enzymes) for a period of eleven months and found that complete turnover of the subtypes within the E. coli population occurred every two to four weeks (15). Studies have also noted a lack of temporal stability in hosts other than human, including yearling steers (39). Two yearling steer herds (n = 6 per herd) were sampled over a period of nine months, and 451 E. coli isolates were ribotyped using two restriction enzymes (EcoRI and PvuII. The two herds were located on the same study site (Goergia). This study demonstrated high genetic diversity in the E. coli populations of steers over time, as 240 ribotypes were identified and only twenty of these appeared in 21

PAGE 30

more than one sampling session. This study did not follow specific individual steers over time, therefore variability on the individual host level could not be determined. The important conclusion from this study was that within geographically distinct host groups there is a high diversity in E. coli isolates collected over extended sampling periods. All of these studies demonstrate temporal instability in E. coli populations in various hosts, which may compromise the usefulness of E. coli for BST methods. However, if there is a large frequency of subtype sharing within a host species, then temporal instability of the E. coli population within an individual host may not be detrimental to the success of BST methods. For example, if a host loses a particular genotype at one time, that genotype may be picked up and shared by another individual of the same host group. The genotype will continue to be specific for that host group and hopefully will continue to discriminate source. Primary versus Secondary Habitats If an organism is to be used successfully in BST it must be present in the fecal material of the host one wishes to trace. Another assumption that must be met for BST, which is often overlooked, is that the dominant subtypes of the E. coli population found in the feces of a given host should be the same as the dominant subtypes found in the environmental waters and soil impacted by that host. Savageau (63) first stated the demand theory that identified two separate habitats for E. coli. According to this theory, the intestinal tract of an animal is the primary habitat of E. coli and the outside environment (i.e. water and sediments) is the secondary habitat. Because these two habitats represent such different conditions based on, for example temperature, nutrient 22

PAGE 31

availability, and water availability, it is possible that certain E. coli subtypes would be better adapted to survive in each habitat. Dominant subtypes from feces may not represent the dominant subtypes that would survive in the external environment and subsequently be sampled in the water. The demand theory has been tested in several studies. MLEE was used to compare the subtypes of the E. coli population isolated from the primary (gastrointestinal tract feces) and the secondary habitats (litter, water) of birds (81). Escherichia coli isolates from the two separate habitats represented genetically distinct subpopulations, supporting the demand theory. A similar study compared E. coli populations in human fecal samples and septic tanks of households by MLEE (thirteen enzymes) (24). In one household there was a large difference between the E. coli populations subtypes found in the feces versus what was found in the septic tank, while in a second household there was no significant difference between the E. coli populations. Geographic Variability in E. coli Populations Whittam et al. (82) assessed the genetic variation of E. coli populations found in humans from the countries of Sweden, Tonga, and the United States using MLEE (twelve enzymes) and found no distinction between regions. The amount of genetic variability sampled within a single region was just as great as the amount of variation observed between regions. However, it is important to note that the sample size for this study was very small, as only 178 E. coli isolates were collected from humans in three different countries. Such under-sampling may prevent geographic variability from being observed. The previously cited feral mouse study (22) reported analogous results. Variability in the 23

PAGE 32

E. coli population was assigned to within-region variability rather than between-region variability, implying that the amount of variability found within a single location is so great that it does not allow differentiation of E. coli populations based on geography. These studies suggest that E. coli populations are not geographically distinct. However, it is important to understand the limitations of these studies. As mentioned earlier, the total E. coli populations were probably under-sampled, and as more E. coli isolates are collected from these different regions, a larger degree of geographic separation may become evident. Another difficulty with studying geographic variation is determining how large or how small a region needs to be in order observe distinct populations. In all of the previous studies, the different sampling regions were vast distances apart. While this would seem to promote greater geographic variation, it may be possible that E. coli populations are, in fact, geographically distinct but on a much smaller scale. Beyond this scale, the large genetic variability inherent in the species will mask any differentiation present. Also, a lack of geographic differentiation may be beneficial to BST methods as it allows single libraries to be used for many watersheds. Objectives The current study is an investigation of the population genetics of E. coli in three types of hosts: humans, beef cattle, and horses. Two bacterial source tracking (BST) methods, ribotyping and antibiotic resistance analysis, were used to investigate the structure of E. coli populations within and between the three host populations (cattle, horses and humans). This study was conducted in four phases, and the results are discussed with respect to their application to BST. 24

PAGE 33

1) The first phase focused on intensive sampling of beef cattle, horse, and human individuals in order to characterize and compare the E. coli population structures in their feces. The population structure was defined as the specific subtypes observed, the number of different subtypes and their relative abundance in the E. coli populations. The overarching hypothesis was that the E. coli population structure would differ based on source. 2) In the second phase, individuals from three types of host were sampled over time in order to determine and compare the temporal stability of the E. coli populations within the feces individuals. The hypothesis was that members of the E. coli populations within individuals would not be stable over time, and that new E. coli types would be observed at each sampling event; however, the diversity of the E. coli populations within an individual would remain stable over time. 3) The third phase of the study investigated the E. coli population of a single human for one month that had just finished antibiotic treatment with Trimethoprim/Sulfamethaxazole. The hypothesis was that an outside influence such as antibiotic treatment would influence the diversity of the E. coli population within this individual. 4) The fourth phase of the study assessed the geographic distinctiveness of E. coli populations within individuals from four beef cattle herds located in different geographic regions. The hypothesis was that E. coli populations would be geographically distinct on the herd level, with a greater frequency of subtype sharing occurring between individuals from a single herd than between individuals from different herds. 25

PAGE 34

Chapter 2. One-Time Intensive Sampling of Feces from Beef Cattle, Horses and Humans Materials and Methods Sample Collection Feces from three different host source categories (humans, horses and beef cattle) were analyzed. Fecal samples were collected from five individuals per source category. All cattle were from the same herd, all horses were stabled at the same farm, and all humans worked in the same laboratory. All individuals remained healthy over the course of the study, and none received antibiotics. Fecal samples were collected from cattle and horse feces by stabbing one sterile swab multiple times into a fresh fecal mass, while humans were sampled via a direct anal swab. A total of three swab samples per individual was collected in a single day at approximately 3 h intervals. The swabs were then streaked onto mFC agar plates (100mm diameter), which were incubated overnight in a water bath at 44.5C (1). Twenty-five blue colonies per swab (75 colonies per individual) were transferred with sterile toothpicks into the wells of microtitre plates that contained EC broth amended with 4-methylumbelliferyl--D-glucuronide (MUG) (50g/ml). glucuronidase activity, which is characteristic of E. coli, was assessed by MUG cleavage, which was determined by fluorescence upon excitation with ultraviolet light (5). Only MUG-positive colonies were further analyzed. Each MUG-positive 26

PAGE 35

isolate was characterized by a unique code based on how many isolates had been collected, the source of the isolate and the location of the isolate on its specific microtitre plate. As an extra verification step, ten percent of the MUG-positive fecal coliform isolates were confirmed as E. coli by an oxidase test and API 20E biochemical test strips (Biomerieux, France), following the manufacturers instructions. Based on these tests 98% of the MUG-positive isolates were confirmed as E. coli. All E. coli isolates were stored at -80C for further analysis. Ribotyping Fifteen E. coli isolates per individual animal (five isolates per swab) were subtyped by ribotyping. Based on preliminary data (data not shown), and the cost of ribotyping, it was decided that fifteen isolates should provide an adequate representation of the E. coli population within individuals based on financial constraints. E.coli isolates were ribotyped by modifying a previously published protocol (59). Isolates were grown in 2ml of brain heart infusion (BHI) broth (Becton Dickinson, Sparks, MD) overnight at 37C while shaking for aeration. Genomic DNA was purified using the DNeasy tissue kit (Qiagen, Valencia, CA) following the manufacturers instructions. DNA concentrations were determined spectrophotometrically by UV absorbance in a Beckman spectrophotometer (Beckman Coulter, Fullerton, CA) after diluting each sample (20 L of purified DNA into 50 L of 10mM Tris-HCL). Purified DNA (1.5 g mL -1 ) was digested using 3 l buffer E (Promega, Madison, WI) and 0.7 l of Hind III (10U L -1 ) (Promega, Madison, WI) in a 27

PAGE 36

1.5ml centrifuge tube, for a final volume of 30 l. The mixture was incubated at 37C for 2.5 hrs. After digestion with Hind III, DNA fragments were separated on a 1% agarose gel (15 cm 15 cm) for 17 hrs at 30 V. An E. coli positive control (ATCC 9637) was run on every gel, and two digoxigenin (DIG)-labeled DNA molecular weight markers (Hind III digested) with a size range of 831 bp 21,226 bp (Roche, Indianapolis, IN) were always run on the first and last lane of each gel. The gels were blotted onto positively charged nylon membranes (Roche Diagnostics, Indianapolis, IN) for 2 hrs using a vacuum blotter. DNA was cross-linked to the membrane using a UV crosslinker (Spectrolinker XL-1000, Spectronics Corp, Westbury NY) and the membrane was hybridized with a hybridization buffer (blocking solution, 10% N-lauroylsarcosine, 10% SDS, sodium citrate, and NaCl) amended with probe overnight at 68C. An appropriate volume of probe was added to the hybridization buffer to create a 15% DIG-labeled 16S rDNA probe solution. The DIG-labeled probe was created using a PCR probe synthesis kit (Roche Diagnostics, Indianapolis, IN) in a thermocycler (Tpersonal, Whatman Biometra, Gottingen, Germany,) (Table 1). DNA isolated from E. coli ATCC 9637 with the DNeasy tissue kit (Quiagen, Valencia, CA) was used as the template for each reaction. The primers used were previously published universal primers targeting the 16S rRNA gene: Eco8F (5-AGAGTTTGATCTGGCTCAG-3) and Eco1492RC (5-GGTTACCTTGTTACGACTT-3) (76). Each reaction followed the PCR program shown in Table 1. The resulting amplicon is 1484 base pairs in length. 28

PAGE 37

The DNA fragments that hybridized with the probe were immunologically detected using the protocol described in the DIG nucleic acid detection kit (Roche Diagnostics, Indianapolis, IN). The developed membranes were scanned as a .tif file into a computer and imported to the BioNumerics program (Applied Maths, Belgium) for analysis. The membrane was dried and laminated for storage. Antibiotic Resistance Analysis (ARA) All of the MUG-positive E. coli isolates (~70 per individual) were analyzed using antibiotic resistance analysis. Antibiotic plates were prepared prior to analysis using 150 mm diameter petri plates with Mueller Hinton agar (Difco, Becton Dickinson, Sparks MD) amended with antibiotics (Table 2). Escherichia coli isolates were grown in 96-well microtitre plates filled with 180 L/well of EC broth and were incubated overnight at 37C. These cultures were diluted by pipeting 20 L of inoculum from each well into separate microdilution tubes filled Initial Denaturation Amplification Final Extension 35 cycles : 94C for 2 minutes 94C for 1 min 72C for 7 minutes 50C for 1 min 72C for 2 min Table 1. PCR program used for the synthesis of the 16S rDNA probe 29

PAGE 38

with 580 L of sterile nanopure water. This dilution method was based on the NCCLS method, which uses sterile nanopure water as the diluent when determining the minimum inhibitory concentration of an anti-microbial agent (51). The diluted culture (100l) was pipeted into a separate well in a new microtitre plate. Up to 96 E. coli isolates were prepared as inoculum cultures for each microtitre plate. The diluted E. coli inoculum was stamped onto each of the antibioic plates listed in Table 2 with a sterile, 96-prong replicator. The plates were incubated overnight at 37C. Growth on the antibiotic plates was scored and recorded. Any discernible growth determined visually at each position was considered a positive score. Antibiotic Concentrations (g ml -1 ) Amoxicillin (AMX) 4.0 128.0 Cephalothin Sodium Salt (CEP) 8.0 32.0 Chloramphhenicol (CLP) 4.0 NA a Chlortetracycline Hydrochloride (CHT) 20.0 80.0 Doxycycline Hydrochloride (DOX) 4.0 NA Gentamycin Sulfate (GEN) 1.0 NA Kanamycin Monosulfate (KAN) 3.0 NA N alidixic Acid Sodium Salt (NA) 3.0 NA N eomycin (NEO) 3.0 NA N orfloxin (NOR) 0.1 NA Oxytetracycline Hydrochloride (OXY) 20.0 NA Penicillin G Potassium (PEN) 20.0 200.0 Rifampicin (RIF) 2.0 16.0 Streptomycin Sulfate (STR) 20.0 80.0 Tetracycline Hydrochloride (TET) 4.0 64.0 Trimethoprim (TRI) 0.25 1.0 Trimethoprim/Sulfamethoxazole (TS) 5.0 NA Table 2. Antibiotics and their concentrations used for antibiotic resistance analysis of E. coli isolates a NA (not applicable) signifies that only a single concentration was used for that p articular antibiotic 30

PAGE 39

Statistical Analysis Banding patterns created by ribotyping (ribotypes) were analyzed using BioNumerics software (Applied Maths, Belgium). Ribotypes were compared by constructing a similarity dendrogram using the Dice coefficient algorithim with maximum similarity. Dice coefficient was chosen because the algorithm relies on the presence and absence of bands when determining similarity. The software optimization setting was 0.2 and the position tolerance setting was 0.7. It was observed that repeated runs (n = 38) of the control strain were 90% similar (data not shown), therefore patterns showing at least 90% similarity with the Dice coefficient were considered identical. The accuracy of this similarity value was confirmed by eye. An example of a similarity dendrogram is presented in Figure 1. Character patterns representing antibiotic resistance profiles (ARPs) created by ARA were also analyzed using BioNumerics software (Applied Maths, Belgium). ARPs were compared by constructing a similarity dendrogram using the binary coefficient known as simple matching (>50% mean). Simple matching was chosen because it is bias towards ARPs that demonstrate high levels of resistance and preliminary studies have shown that there is greater variability in ARPs of high resistance than ARPs of lower resistance. Repeated runs (n = 40) of the control were 94% similar, therefore ARPs were considered the same if they were at least 94% similar. Diversity indices were used to assess the structure of E. coli populations. The observed frequency and distribution of E. coli subtypes, determined by ribotype or ARP, were compared within and between three host categories. Three different diversity 31

PAGE 40

indices and one eveness index from the Hills family of indices were used: 1.) Richness estimator (S), 2.) Shannons index (H), 3.) Simpsons index (), and 4.) Pielous eveness (J) (Figure 2) (43). These measurements do not demonstrate the total diversity of the E. coli population in any given individual because the number of E. coli isolates analyzed was only a very small proportion of the total number of E. coli isolates present in a fecal sample. However, the sample size utilized in this study was as large as was practical. Furthermore, because the sample sizes were similar per individual, the diversity measurements can be used to present a comparison between source categories. One-way ANOVA with Dunnett's post test was performed using GraphPad InStat version 3.00 (GraphPad Software, San Diego California) on all diversity measurements to compare E. coli population structure within the different source categories. Accumulation curves were used to compare E. coli populations in different source categories by estimating the number of new patterns observed in a given individual as a function of sampling effort for each different source. Accumulation curves can be used to obtain an estimate of the sampling effort needed in order to most accurately represent a population (38). Chi-square analysis was performed (GraphPad Software, San Diego California) to compare the amount of subtype sharing between source categories observed with ribotyping and antibiotic resistance analysis results. For all statistical tests, significance was determined at P < 0.05. 32

PAGE 41

Figure 1. Example of a similarity dendrogram of ribotypes created using the Dice coefficient. The vertical line represents a 91% similarity cutoff, which was used as the criterion for calling ribotypes identical. 100 95 90 85 80 75 70 65 60 Cow10Cow11Horse45Horse46Horse44Horse47Cow6Cow7Cow5Horse144Horse146Hum190Hum191Hum192Hum193Hum194Hum205Hum206Hum5Hum6Cow49Horse143Cow111Cow112Cow110Horse145 Figure 2. Equations for the parameters used to define E. coli population structure. Pattern Richness (S) = # of patterns Shannon index (H) = p i ln(p i ) pi = # isolates with pattern (i) / total isolates Pielous eveness index (J) = H/ln S Simpson index () = (ni(ni-1))/(n(n-1)) ni = # isolates with pattern i n = total # of isolates 33

PAGE 42

Results Population Distribution of E. coli Subtypes Within Source Categories Population structure is defined as the number of different subtypes and their relative abundance in a given population sample. The population structure of E. coli, as determined by ribotyping and ARA, was compared both within individuals from the same source categories and between individuals from different source categories (humans, horses and beef cattle). The hypothesis was that the E. coli population structure would differ based on source, i.e. humans, horses, or cattle. Ribotyping results suggest that the population structure of E. coli, as determined by the diversity indices, differs depending upon host source (Table 3). One-time samples of the feces of individual horses (n = 5) contained a significantly more diverse E. coli population than either cow or human individuals based on all diversity measurements, although eveness was not significantly different among different hosts (Table 4). The E. coli populations of individual horses demonstrated a richness of S = 9.2, had a mean Shannon value of H=2.0 and a mean Simpson value of 1/. There was no significant difference in any diversity measurements of the E. coli populations of beef cattle and humans (Table 4). The E. coli populations of individual humans had a mean richness of S =2.2, had a mean Shannon value of H=0.43, and a mean Simpson value of 1/. While the E. coli populations of individual cows had a mean richness S =3.4, had a mean Shannon value of H=0.67 and a mean Simpson value of 1/. 34

PAGE 43

Source Richness Estimator (S) Shannon Index (H') Pielous Eveness (J) Simpson Index (1/) Cow A (n = 15) 6 1.52 0.85 4.4 Cow B (n = 15) 3 0.88 0.80 2.4 Cow C (n = 15) 5 0.70 0.43 1.9 Cow D (n = 15) 1 0 1.00 1.0 Cow E (n = 15) 2 0.25 0.36 1.1 Mean 3.4a 0.67a 0.69a 2.2ab Human A (n = 14) 6 1.59 0.89 5.8 Human B (n = 15) 1 0 1.00 1.0 Human C (n = 15) 2 0.58 0.84 1.7 Human D (n = 15) 1 0 1.00 1.0 Human E (n = 15) 1 0 1.00 1.0 Mean 2.2a 0.43a 0.95a 2.1a Horse A (n = 14) 9 2.01 0.91 10.2 Horse B (n = 14) 11 2.26 0.94 30.4 Horse C (n = 15) 9 1.95 0.78 8.1 Horse D (n = 15) 13 2.52 0.98 52.6 Horse E (n = 15) 4 1.27 0.79 4.0 Mean 9.2b 2.0b 0.90a 21.1b Table 3. Diversity measurements of the population structure of E. coli within individuals from different source categories using ribotyping. Values that share the same letter within columns are not significantly different. 35

PAGE 44

Table 4. Significance values determined by comparing the means of all diversity measurements obtained from each source category. Means were compared using ANOVA. If significantly different, subsequent pairwise comparisons were performed for each source category to determine the difference. If the ANOVA was not significant, no additional tests were performed. Significant values are highlighted. Ribotyping Results ANOVA for Richness P = 0.0039 ANOVA for Simpson index P = 0.0153 Horse Cow Human Horse Cow Human Horse 0.0050.05 Cow Cow Human 0.0010.05 Human 0.010.05 ANOVA for Shannon index P = 0.0011 ANOVA for eveness index P = 0.189 Horse Cow Human Horse Cow Human Horse 0.010.05 Human 36

PAGE 45

Diversity measurements obtained when antibiotic resistance analysis (ARA) was used to subtype E. coli isolates also suggested that the population structure of E. coli found in the feces differs depending on host source (Table 5). The E. coli populations of human individuals were significantly less diverse than the E. coli populations of both cattle and horse individuals based on all diversity measurements. There was no significant difference noted between the eveness values (J) of any source category. There was no significant difference in the diversity measurement obtained for the E. coli populations of either cattle individuals or horse individuals according to the richness estimator and the Shannon index. However, the E. coli populations of horse individuals demonstrated a significantly higher Simpson value (1/=5.5) than the E. coli populations of cattle (1/=3.0) (Table 6). Similar trends were observed when ribotyping or antibiotic resistance analysis was used to subtype E. coli isolates. In general, the E. coli populations of horses appear to be the most diverse, followed by the E. coli populations of cattle and finally the E. coli populations of humans being the least diverse. For all source categories the diversity measurements were higher using antibiotic resistance analysis than was observed with ribotyping. One reason may be because four times the number of isolates was processed by ARA compared to ribotyping. This difference was due to the ease of processing ARA isolates as well as the relative cost difference of processing ARA isolates compared to ribotyping isolates. 37

PAGE 46

Source Richness Estimator (S) Shannon Index (H') Pielous Eveness (J) Simpson Index (1/) Cow A (n = 71) 14 1.35 0.51 2.2 Cow B (n = 68) 11 1.69 0.71 3.7 Cow C (n = 43) 11 1.77 0.74 4.5 Cow D (n = 52) 6 0.80 0.44 1.6 Cow E (n = 68) 9 1.43 0.65 3.0 Mean 10.2a 1.4a 0.61a 3.0a Human A (n = 68) 6 0.80 0.44 1.9 Human B (n = 70) 4 0.47 0.34 1.3 Human C (n = 71) 1 0 1.00 1.0 Human D (n = 67) 4 0.48 0.34 1.3 Human E (n = 65) 7 1.21 0.64 2.5 Mean 4.4b 0.59b 0.55a 1.6b Horse A (n = 66) 11 1.84 0.75 3.1 Horse B (n = 55) 12 2.12 0.85 7.4 Horse C (n = 48) 8 1.54 0.74 3.1 Horse E (n = 60) 11 2.0 0.83 8.2 Mean 10.5a 1.86a 0.79a 5.5c Table 5. Diversity measurements of the population structure of E. coli within individuals from different source categories using antibiotic resistance analysis. Values within the same column that share a letter are not significantly different. 38

PAGE 47

Table 6. Significance values determined by comparing the means of all diversity measurements obtained from each source category. Means were compared using ANOVA. If significantly different, subsequent pairwise comparisons were performed for each source category to determine the difference. If the ANOVA was not significant, no additional tests were performed. Significant values are highlighted. Antibiotic Resistance Analysis Results ANOVA for Richness P = 0.0024 ANOVA for Simpson index P = 0.0317 Horse Cow Human Horse Cow Human Horse P >0.05 0.0010.05 P <0.001 Horse Cow 0.01
PAGE 48

Accumulation curves demonstrate the number of new subtypes that are observed within an individual as E. coli isolates are collected. Ribotype accumulation curves demonstrated that, on average, the dominant E. coli populations in humans were made up of the fewest number of ribotyptes and apparently required the fewest number of isolates in order to represent the dominant ribotypes present in the E. coli population (Figure 3A). Both human and cattle accumulation curves demonstrated that collecting ten to twelve isolates was enough to adequately represent the diversity within the dominant E. coli populations of these two source categories using the ribotyping technique. The E. coli populations of horses were the most diverse and were represented by the greatest number of different ribotypes. Because the slope of the accumulation curve for E. coli ribotypes from horse feces did not approach 0 in this graph, the number of isolates needed to adequately represent the dominant E. coli population is unknown, and more than 15 isolates should have been analyzed (Figure 3A). ARA results also demonstrated that humans had the least diverse E. coli population, requiring the fewest isolates of the three source categories to represent the dominant ARPs in their E. coli population (Figure 3B). A much greater difference between E. coli populations of human and cattle were observed in the accumulation curves based on ARA compared to those based on ribotyping. Using ARA, the cattle accumulation curve is more similar to the horse accumulation curve, with both curves demonstrating a diverse E. coli population with many different ARPs observed. The accumulation curves for horses and cattle, based on ARA, demonstrated that collecting over fifty isolates for an individual adequately represents the dominant ARPs in their E. coli population. Around twenty-five isolates are required to adequately represent the 40

PAGE 49

dominant ARPs in the E. coli populations of humans. For all source categories, the slope of the ARP accumulation curves was steeper then the slope of the ribotype accumulation curves. These graphs demonstrate that the rate of accumulation of new subtypes is different based on both the source of the E. coli population (i.e. cattle, humans or horses) and the typing technique used. Therefore, different sampling strategies must be used, depending on the individual being sampled and the typing technique being used, in order to obtain an accurate representation of their dominant E. coli populations. 41

PAGE 50

Figure 3. Accumulation curves representing the number of unique patterns versus sampling effort for horses, humans, and cattle. 42 ARP Accumulation02468101216111621263136414651Number of isolatesNumber of new ARP s Mean Horse Mean Cattle Mean Human Ribotype Accumulation 012345678910123456789101112131415Number of isolatesNumber of new ribotype s Mean Horse Mean Human Mean Cattle A B

PAGE 51

Pattern Sharing Within and Between Source Categories The number of sampled E. coli isolates that share a specific ribotype or antibiotic resistance pattern (ARP) was observed during a one time sampling event in order to observe pattern sharing between individuals within the same source category (Figure 4). The hypothesis was that there would be greater pattern sharing between E. coli isolates from individuals within a source category than between individuals from different source categories. E. coli isolated from different human individuals did not share any ribotypes (Figure 4A), although ribotype sharing by E. coli within individuals was very high (Table 3). For example, all E. coli isolates from human B shared the same ribotype, but, this ribotype was not shared by any other human individuals. Three percent of horse E. coli isolates shared a ribotype that was found in two different horse individuals, while the remaining ninety seven percent of E. coli isolates had a ribotype observed in a single horse individual. No horse E. coli isolates had a ribotype found in three or more horse individuals (Figure 4C). The E. coli populations of cattle demonstrated the most ribotype sharing of the three source categories (Figure 4B). Forty three percent of cattle E. coli isolates shared a ribotype found in two different cattle individuals; however, there were no specific ribotypes shared between more than two cattle individuals. Overall, for all three source categories, humans, horses and cattle, greater ribotype sharing was observed between E. coli isolates within single individuals than between E. coli isolates from different individuals of the same source category (Chi-sqare test, P <0.001 for all sources). 43

PAGE 52

A B C D E F Cow Ribotypesn = 74 Horse Ribotypesn = 74 Human Ribotypesn = 73 Figure 4. Proportion of E. coli isolates that demonstrate pattern sharing between different individuals from the same source category. A, B, and C show ribotyping results; D, E, and F show ARA results. Human ARPsn = 341 Cow ARAn = 302 Horse ARPs*n = 214 1 individual 2 individuals 3 individuals 4 individuals The horse source category for ARA has a small number of isolates compared to the other source categories because patterns of four, rather than five individuals are represented 44

PAGE 53

Greater sharing of patterns between different individuals within a source category was observed using antibiotic resistance analysis (Figures 4D-4F); however, four times as many E. coli isolates were analyzed using ARA. For all source categories, the majority of E. coli isolates had an ARP found in two or more individuals. In fact, compared to ribotyping data, a large number of E. coli isolates from each source category shared an ARP that was found in four different individuals within their source category. Using antibiotic resistance analysis, the horse source category demonstrated the largest amount of ARP sharing (Figure 4F). ARP sharing within individuals is presented in Table 5. Although greater pattern sharing was observed between different individuals using ARA than was seen with ribotyping, there is still greater sharing of ARPs between E. coli isolates from a single individual, than between E. coli isolates from different individuals within the same source category for all source categories (Chi-sqare test, P < 0.001 for all source categories). E. coli subtypes isolated from all individuals in each source category were compared in order to observe ribotype and ARP sharing between different source categories (Table 7). Most of the sampled E. coli isolates (71%) had a ribotype that was only found in a single source category, i.e. human only, horse only, or cow only (Table 7A). Over 22% of the total E. coli isolates shared a ribotype that was observed in both the horse and cow source categories and 6.7% of the total E. coli isolates had a ribotype that was shared by members of all three source categories. No ribotypes were shared by humans and cattle only, or by humans and horses only. Although 31.1% of the total E. coli isolates had a ribotype found only in horses, 63.1% of the total ribotypes were found 45

PAGE 54

in horses only. This difference is because the horse source category has a diverse E. coli population with many ribotypes being represented by single isolates. In contrast, only 18.5% of the total ribotypes are found in humans only, however 28.5% of the total E. coli isolates have a ribotype found only in humans. The diversity of the E. coli populations of humans are low, therefore many isolates share a single ribotype. Comparitively greater sharing of antibiotic resistance patterns (ARPs) between source categories was observed (Table 6B). The majority of E. coli isolates (77%) had an antibiotic resistance pattern that was observed in two or more source categories. In fact 59% of the E. coli isolates analyzed by ARA had an antibiotic resistance pattern that was found in all three source categories. However, only 14.3% of the total number of ARPs were shared between three source categories. These data suggest that a small number of dominant ARPs are shared between source categories and are represented by a large number of E. coli isolates. There is a single ARP that is represented by almost 400 E. coli isolates and is shared by individuals from all three source categories (data not shown). A very small proportion of E. coli isolates had an antibiotic resistance profile found in either horses and humans only (three percent), or cattle and humans only (less than one percent), a trend shared with ribotyping. Also, as seen with ribotyping, horses have the largest number of ribotypes (44.8%) but these ribotypes are represented by a small number of E. coli isolates (11.6%). 46

PAGE 55

Source Distribution of all ribotypes in source categories (n = 222) Distribution of unique ribotypes (n = 65) Human only 28.8% 18.5% Cow only 10.9% 12.3% Horse only 31.1% 63.1% Human/Horse 0% 0% Human/Cow 0% 0% Horse/Cow 22.5% 4.6% Human/Horse/Cow 6.7% 1.5% A. Ribotyping Table 7. The proportion of E. coli subtypes shared between multiple source categories B. Antibiotic Resistance Analysis Source Distribution of all ribotypes in source categories (n = 857) Distribution of unique A RPs (n = 56) Human only 2.4% 5.3% Cow only 9.0% 16.1% Horse only 11.6% 44.8% Human/Horse 3.4% 1.7% Human/Cow 0.6% 1.7% Horse/Cow 14.8% 16.1% Human/Horse/Cow 58.2% 14.3% 47

PAGE 56

Discussion The current study systematically compared the E. coli populations of individuals from different animal sources using two subtyping methods and demonstrated two major findings: 1) the frequency of subytpe sharing between individuals is determined by which subtyping technique is being used. Ribotypes were often limited to single individuals with very little ribotype sharing between individuals, while antibiotic resistance patterns were frequently shared between individuals both within and between source categories. And 2) the diversity of an E. coli population within an individual is determined by source and the subtyping technique utilized. Very few studies have directly compared E. coli populations between different species. Caugant et al. performed a study that subtyped 655 E. coli isolates by multilocus enzyme electrophoresis (MLEE) from thirty-four individuals (twenty three humans, six dogs, and five cats) (14). Eighty-five percent of E. coli subtypes were limited to a single individual, and 15% of the E. coli subtypes were shared between individuals. Four percent of the subtypes were shared between individuals from different species. These results are similar to the ribotyping data presented in the current study; however, the MLEE study did not observe a difference in the diversity of E. coli populations from different species. This difference may be explained by differences in the physical proximity of host species. The host species of the MLEE study included humans and their pets (dogs and cats); which were living in the same home. The current study sampled humans, horses and beef cattle; which are all located in different areas. The close proximity of the humans, dogs and cats sampled in the MLEE study, coupled with 48

PAGE 57

the fact that all had an omnivorous/carnivorous diet, may explain why no difference in the E. coli population structure of these individuals was detected. Although few studies directly compare the E. coli populations of different host species, studies have investigated the population structure of E. coli within individuals from a single species. Another MLEE study subtyped the E. coli population of a single human for eleven months (15). At any given time the majority of E. coli isolates within this human displayed a single dominant subtype, and the diversity of the population was very low. These results are comparable to observations made in the current study, i.e. diversity of human E. coli populations were low at any given time point, and were usually dominated by a single subtype. A study performed by Whittam used MLEE to subtype the E. coli populations of five individual birds (81). An average of 4.2 subtypes per bird was observed when 20-30 isolates were analyzed. Although MLEE was used as the subtyping method in each of these studies, different enzymes were chosen for each method, which complicates comparison these studies. However, in a broad sense, Whittams and Caugants studies demonstrate that the structure of E. coli populations, as determined by MLEE, differed based on the source animal, which agrees with our findings. A recent study performed by Simpson et al used denaturing gradient gel electrophoresis (DGGE), with primers for the V3 variable region of 16S rDNA gene, to fingerprint the domimant bacterial population in the feces of dogs (71). Their study was designed to assess differences in the enteric bacterial populations based on breed, age, and diet of the dogs. Each dog appeared to have a distinct enteric bacterial population based on DGGE banding patterns, however banding patterns did tend to cluster on a 49

PAGE 58

similarity dendrogram according to the age of the dog. The DGGE study did not specifically investigate the E. coli population of dogs, however in a general sense their observations support those of the current study, i.e. there is high variability between the fecal bacterial populations of host individuals. A goal of the current study was to determine the number of isolates needed to adequately represent the diversity of E. coli subtypes within a population. A gram of feces may contain billions of E. coli cells (44), therefore it is impossible to sample the entire E. coli population of a given individual. However, the assumption made for this study is that the chosen sample size, although only a proportion of the total E. coli population, should act as a representation of the diversity of the numerically dominant subtypes that make up the E. coli population. Accumulation curves were used as a tool to determine the adequacy of the chosen sample size in the intensive sampling phase of this study. According to the accumulation curves, ten to twelve E. coli isolates adequately represented the dominant ribotypes in the feces of humans and beef cow individuals, while more isolates were needed to represent the dominant antibiotic resistance patterns (20-25 for humans, and 45-50 for beef cattle and horses). These accumulation curves only represent the number of dominant subtypes that are present in a fecal swab sample processed as described in Chapter 2 (Materials and Methods) and do not represent the total diversity of the entire E. coli population within the feces, which is probably much higher. This limitation is acknowledged and accepted because this particular sampling procedure is an accepted practice in microbiological studies in general and in BST studies in particular. Therefore, even though the total diversity is not represented, understanding the behavior of this limited E. 50

PAGE 59

coli population has practical implications when attempting to optimize bacterial source tracking methodology. Another limitation of the current study is the small number of individuals (n=5) sampled from each host species. The current study demonstrated a very low frequency of ribotype sharing between individuals within a source category, and in the case of humans there was no ribotype sharing at all. In another study in this laboratory (17), with a large sample size of individuals (sixty individuals per source category), ribotype sharing was demonstrated between individuals within a source category and between source categories. It is understood that the current study may misrepresent the true frequency of ribotype sharing that occurs within a source category due to the small sample size. A greater frequency of subtype sharing both within and between source categories was observed for ARA compared to ribotyping, but many more isolates were analyzed for ARA. Thus, direct comparisons of the frequency of sharing by the two subtyping methods must be cautiously interpreted. In summary, the current study has demonstrated that the diversity of the dominant E. coli populations of feces differ based on source and the chosen subtyping technique. These results have strong implications for understanding and improving the efficiency and accuracy of bacterial source tracking. 51

PAGE 60

Chapter 3. Sampling of Individual Host Animals from Three Source Categories Over Time Materials and Methods Sample Collection and Processing Three individuals per source category from three different source categories (human, horse, and cattle) were sampled over several months. These individuals were among the individuals sampled in the one time intensive sampling event (Chapter 2 Materials and Methods). Human individuals were sampled for twelve months, horse individuals were sampled for eight months, and individual beef cows were sampled for seven months. During the sampling period, a one-time intensive sampling event was conducted for all individuals (Chapter 2 Materials and Methods). Once a month, a single fecal swab sample was obtained from each individual. The fecal swabs were processed for E. coli isolation as described previously (see Chapter 2, Materials and Methods). Twenty-five blue colonies per swab were transferred from mFC to a microtitre plate containing EC broth amended with MUG. Five MUG-positive E. coli isolates per individual were analyzed by ribotyping each month, and all MUG-positive E. coli isolates were analyzed by ARA. Ribotypes and antibiotic resistance profiles (ARPs) were analyzed using BioNumerics as described previously (Chapter 2, Materials and Methods). One-way 52

PAGE 61

ANOVA with Dunnett's post test was performed (GraphPad Software, San Diego California) on the mean richness estimator (S) values for the entire sampling session in order to compare the accumulated richness within individuals over time. 53

PAGE 62

Results Population Structure of E. coli Within Individuals Over Time The dominant E. coli population in the feces of three human individuals was assessed on a monthly basis for a twelve-month period using both ribotyping and antibiotic resistance analysis to subtype E. coli (Figures 5-7). The hypothesis was that the specific subtypes within an E. coli population for a single individual would not be stable over time. Figure 5A represents the E. coli subtypes sampled from human A over time, as determined by ribotyping. Each color/pattern in the figures represents a unique ribotype, which was assigned a number. The number of dominant E. coli ribotypes isolated per sample event from human A varied little over time. For each month (except August 2002) the E. coli population was represented by one to two dominant ribotypes, suggesting that the richness of the population remained relatively consistent over time. However, there was a large amount of variation in the specific ribotypes observed over time. The ribotypes of E. coli isolated from human A observed in June 2002 and July 2002 were not the same ribotypes observed in later months, i.e. December 2002 and January 2003 (Figure 5A). Many ribotypes were only present for one month and then not seen again, although ribotype 1 was a persistent ribotype and was observed in five different months. In this case, a persistent ribotype is defined as any ribotype observed at more than one consecutive sampling event. Similar results were observed for the dominant E. coli populations of the other two humans (Figures 7A and 8A). Humans B and C demonstrated a consistent richness value for their E. coli populations over time, containing one to two different ribotypes for 54

PAGE 63

any given month. A number of different ribotypes were observed throughout the sampling period and many ribotypes were only observed at one sampling event. The E. coli populations of human B demonstrated one very stable ribotype (ribotype 2) that was present in 8 different months and persisted from August 2002 until May 2003 (Figure 7A). There was no significant difference (ANOVA, P =0.142) among the accumulated richness of E. coli ribotypes within human individuals, although human B appeared to have the lowest accumulated richness as only four ribotypes were observed throughout the entire sampling period. The dominant E. coli population structure within humans was also monitored using ARA (Figures 5B-7B). More patterns were identified by subtyping with ARA than with ribotyping. There was no significant difference (P =0.415, ANOVA) in the richness of E. coli populations within human individuals over time, i.e. similar richness in antibiotic resistance patterns (ARPs) was observed at any given month for all human individuals. The E. coli populations demonstrated temporal instability by ARA as different ARPs were observed throughout the sampling period. More ARPs demonstrated persistence in human individuals than was observed with ribotyping. For example, the E. coli population of human A contained four different ARPs that demonstrated persistence (ARPs 7,10,16,19) while with ribotyping there was only one persistent subtype (ribotype 1). However, this result was not unexpected due to the greater amount of subtype variability in ARA that was demonstrated previously. 55

PAGE 64

Figure 5. The E. coli population structure over time within Human A. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for all months. Human A Ribotype02468101214Jun-02Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03May-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 A Human A ARA01020304050607080Jun-02Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03May-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 B 56

PAGE 65

Figure 6. The E. coli population structure over time within human B. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained f or all m onths. Human B Ribotyping0246810121416Jun-02Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03May-03Number of isolates with pattern 1 2 3 4 A Human B ARA01020304050607080Jun-02Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03May-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 B 57

PAGE 66

Figure 7. The E. coli population structure over time within human C. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for a ll m onths. Human C Ribotyping 0246810121416Jun-02Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03May-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 A Human C ARA01020304050607080Jun-02Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03May-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 B 58

PAGE 67

The dominant E. coli populations of three individuals from the beef cattle source category was monitored on a monthly basis for eight months by ribotyping and ARA (Figures 8-10). The E. coli populations of all three cow individuals displayed similar results with ribotyping. The dominant E. coli populations of beef cattle are unstable through time as new ribotypes were observed at each sampling event throughout the eight-month sampling period. There was no significant difference (ANOVA, P =0.8294) in the accumulated richness of ribotypes between different individuals from the beef cattle source, with a range of thirteen to nineteen different ribotypes observed within these cattle over eight months. The E. coli populations of beef cattle individuals displayed very few persistent ribotypes. Most ribotypes were observed only once and then not found in any other months. The dominant E. coli population of cow A demonstrated two persistent ribotypes (ribotype 10 and 12) and the E. coli populations of cow B and C each demonstrated only one ribotype that persisted for more than one month (ribotype 10 and ribotype 3, respectively) (Figures 8A-10A). Generally, these persistent ribotypes were observed at two sampling events, however, ribotype 10 in cow B was observed in four sampling months (Figure 9A). Overall, E. coli populations in cattle were highly variable. Antibiotic resistance analysis revealed results similar to what was observed with ribotyping the E. coli populations of cattle over time (Figures 8B-10B). There was no significant difference (ANOVA, P =0.272) in the accumulated richness of ARPs within beef cattle individuals over time. Also, there was no temporal stability in the dominant E. coli populations of beef cattle determined by ARA, with new ARPs observed each month. Persistent ARPs were more frequently observed than persistent ribotypes in the 59

PAGE 68

E. coli populations of cattle. For example, five persistent ARPs were isolated from beef cow B, but only one ribotype demonstrated persistence. Again, this is not unexpected due to the greater amount of subtype variability observed with ARA compared to ribotyping and also the fact the more isolates were processed by ARA. The dominant E. coli populations of three horses were monitored on a monthly basis for eight months using both ribotyping and ARA (Figures 11-13). Both methods revealed no significant difference in the richness of the dominant E. coli populations of all horses at any given time (ANOVA, P =0.161 and 0.670, for ribotyping and ARA respectively). The dominant E. coli populations of all horses were temporally unstable, as determined by ribotyping and ARA. Many ribotypes and ARPs in the horse E. coli populations were only observed at one sampling event. Only 4% of the many ribotypes observed in horse individuals demonstrated any persistence through time. In the E. coli populations of horses A and C there was one (ribotype 23) and three (ribotypes 15, 23,25) ribotypes that persisted for more than one month. Horse B was the only individual, out of all three source categories, with an E. coli population that did not have a single persistent ribotype during the entire sampling period. 60

PAGE 69

Figure 8. The E. coli population structure over time within cow A. Different colors represent different ribotypes (A) or ARPs (B). Samples were not ob tained f o r a ll m onths. Cow A Ribotyping0246810121416Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 A Cow A ARA01020304050607080Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 B 61

PAGE 70

Figure 9. The E. coli population structure over time within cow B. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for all months. Cow A Ribotyping0246810121416Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 A Cow B ARA010203040506070Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 B 62

PAGE 71

Figure 10. The E. coli population structure over time within cow C. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for all months. Cow C Ribotyping0246810121416Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 A Cow C ARA05101520253035404550Jul-02Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 B 63

PAGE 72

64 Horse A ARA010203040506070Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 Horse A Ribotyping0246810121416Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 B A Figure 11. The E. coli population structure over time within horse A. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for all months.

PAGE 73

Figure 12. The E. coli population structure over time within horse B. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for all monthsa. Horse B Ribotyping0246810121416Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 A Horse B ARA0102030405060Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 B 65

PAGE 74

Figure 13. The E. coli population structure over time within horse C. Different colors represent different ribotypes (A) or ARPs (B). Samples were not obtained for all months. Horse C Ribotyping0246810121416Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Number of isolates with pattern 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 A Horse C ARA05101520253035404550Aug-02Sep-02Oct-02Nov-02Dec-02Jan-03Feb-03Mar-03Apr-03Number of isolates with ARP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 B 66

PAGE 75

Pattern Overlap Between Individuals Within a Source Category Over Time Ribotyping Results from the one-time sampling event of five human, beef cattle, and horse individuals (Chapter 2, Results) demonstrated that in one month there was little overlap in E. coli ribotypes between individuals in the same source category (Figure 4). The overlap between the three individuals of each source category was compiled for the entire experiment (over seven, eight, or twelve months) (Table 8). A shared ribotype is any ribotype that was observed within two different individuals regardless of which sampling date it was observed. For example, in Table 8, because ribotype 6c was observed in cow A in July and then in cow C in August it would be considered a shared ribotype. Only one ribotype was shared by more than one human over time. Ribotype 1 was observed in both human A and human B, and was shared in four different months (Table 8). The fact that human A and human B live in the same household may have contributed to the sharing of ribotype 1. E. coli ribotypes appear to be shared between different cattle within a herd (Table 8). Eight different ribotypes were shared between individual cattle through time but were not shared in consecutive months. For example, ribotype 7 appears in cow A in August 2002 and then is not found again until March 2003 in cow D. This type of distribution was observed with a number of the beef cow ribotypes. Horse individuals shared the largest number of ribotypes compared to the other two source categories. Twelve different ribotypes were shared between individual horses over eight months (Table 8). The E. coli ribotypes isolated from horses were not shared 67

PAGE 76

between consecutive months but were sampled sporadically. For example, ribotype 5 was first observed in horse B in October 2002, then in horse A in December 2002 and again in horse E in March 2003. Human A Human B Human C Cow A Cow B Cow C Horse A Horse B Horse C Jun-02 X X X X X X Jul-02 6c 8c X X X Aug-02 1m 1m 7c 8c 2c,6c 3h Sep-02 3h 2h,3h,8h,9h, 11h Oct-02 4c 8c 2h,11h 5h,7h,9h 10h Nov-02 1m 11h Dec-02 1m 1m X X X 5h,11h 10h,12h 6h,12h Jan-03 1m 1m 1c 1c,2c,3c 2c 1h Feb-03 1m 3c 4c 5c 1h,8h 4h,12h 3h,12h Mar-03 1m 5c,8c 8c 7c 2h,7h 4h,5h Apr-03 X X X 6h,7h,8h May-03 1m 1m X X X X X X Table 8. Ribotype sharing by individuals within the same source category over time. Each number/letter represents a different ribotype. The letters correspond to a source category (m=human, c=cow, h=horse). signifies that there were no shared patterns in that month x signifies that the individual was not sampled in that month 68

PAGE 77

Antibiotic Resistance Analysis (ARA) A far greater amount of antibiotic resistance pattern (ARP) sharing over time was observed in all three source categories than was observed by ribotyping. Due to the difference in sample sizes for each typing method, four times as many isolates were processed by ARA at each sampling event compared to ribotyping, resulting in a comparison of hundreds more ARA isolates. Twenty different ARPs were shared by the E. coli populations of human individuals over time (Table 9A). Half of these ARPs were observed in the E. coli population of all three human individuals at some point during the sampling. Generally, the shared ARPs were present for no more than two months for any individual, however three ARPs (6,8,13) were observed at more than 6 sampling events each. ARP 13 seems to follow the same trend observed for ribotype 1m in the previous section. This result was not unexpected because all E. coli isolates from humans A and B that had ribotype 1 also had ARP 13. This ARP is found in human B for eight months and is shared by human A in three of those months and human C in one month. As mentioned earlier, the fact that human A and B live in the same household may have influenced the sharing of this E. coli subtype. Eighteen different ARPs were shared by E. coli isolates from different individual beef cattle during the sampling period (Table 9B). Over half of these ARPs (60%) were shared by all three cattle at some point during the sampling period, which was very similar to the results observed for the human source category. The E. coli populations of horses demonstrated the greatest amount of ARP sharing through time (Table 9C). Thirty-four different ARPs were shared by more than one horse individual at some point 69

PAGE 78

during the sampling period. These results demonstrate that the source category with the least diverse E. coli populations (humans) tends to have the lowest frequency of subtype sharing over time, while the source category with the most diverse E. coli populations (horses) tends to have the highest frequency of subtype sharing over time. 70

PAGE 79

Date Human A Human B Human C Jun-02 15,17,18 16,17 6,20 Jul-02 5,11 5,13,14 3,15,18 Aug-02 2,6,12,13,14 7,10,12,13 Sep-02 13 6,7,8 Oct-02 1,2,3,6,8,19 2,3,13 1,8,19 Nov-02 X X X Dec-02 6,8,9,12,13 12,13 19,20 Jan-03 16,17,20 16 Feb-03 6 12.13 10 Mar-03 1,6,7,8,11 4 2,10 Apr-03 4,6,7,8,9 13 1,4,11,12,13,14 May-03 1,13,14 13,15 1,5,8 Date Cow A Cow B Cow C Jul-02 5 3,7,10,14,17 3,5 Aug-02 1,3,4,6,10,15,16,17 7,8,9,10,11,12,13,16 3,4,6,7,8,10,11,15 Sep-02 X 4,8,10,13,16 X Oct-02 3,4,6,7 3,8,10,12,14,16 3,7,10,13,18 Nov-02 3 1,3,4,9,12,14,16,17 3,7,9,10,14 Dec-02 X X X Jan-03 X X X Feb-03 3,8,11,12,14,15,16 7,8,10,12,14,16 3,4,5,10,13,17 Mar-03 7,8,9,12,14,15.16,18 2,8,9,12,13,14,16 2,4,13 Date Horse A Horse B Horse Aug-02 7,8,11,13,25,26,33 6,7,9,12,13,22,26 X Sep-02 1,6,10,26,31,34 1,2,12,16,20,21,22,23,25 ,26,32 2,3,14,16,20,23,27,33 Oct-02 1,6,7,11,12,13,14,23 ,25,26,32 3,12,13,15,21,23,32,34 5,13,15,16,20,21,22,24 Nov-02 X X X Dec-02 7,9,10,15,26,33,34 6,7,8,15,23,25,32 2,6,14,20,23,24,26,28 ,29 Jan-03 17,18,19 X X Feb-03 1,6,10,12,23,26,27,30 5,10,15,17,23,24,26 8,18,23,28,34 Mar-03 4,5,14,23,26,28,29,30 ,32,33 7,9,19,20,25,26,30,32 5,7,9,11,13,14,20,23 ,26,30,32 Apr-03 1,6,14,24,25,28 6,14,15,24,27,28,30,33 2,4,23,27,28,31,33 C B A Table 9. Antibiotic resistance pattern sharing between individuals over time for humans (A), cattle (B) and horses (C). Within a given table, each number represents a unique ARP. Highlighted numbers are noteworthy ARPs. 71

PAGE 80

Discussion Two findings emerged from the analysis of E. coli subtypes in individuals over time. The most obvious result, which was observed in every individual sampled, was that the dominant E. coli populations within individuals are temporally variable, i.e, new subtypes were frequently observed within the populations over time. The second finding of this study was that subtype sharing (repeatedly observed subtypes) between individuals within a source category increased over time. For this phase, subtype sharing is defined as a subtype being observed in two different individuals regardless of the time the subtype is observed. Few studies have followed E. coli populations over time, and ours is the first to monitor the subtypes of E. coli populations, as determined by ribotyping and antibiotic resistance analysis, in specific individuals from different animal sources over time. Caugant subtyped the E. coli population of a single human by MLEE twenty-two times over an eleven-month period (15). New E. coli subtypes appeared every month, and a total of 53 different subtypes were observed throughout the sampling period. However, three subtypes persisted throughout the sampling period. Furthermore, the diversity of E. coli subtypes remained relatively consistent over time, as a single dominant subtype comprised the majority of the sampled isolates at any given time. It is important to note that the sampling events in Caugants study were not evenly spaced. In some months the subject was sampled daily, while other sample events were separated by as many as five months. Therefore, although the dominant E. coli subtypes sampled tended to vary over time, it is difficult to interpret the frequency of change in this population. In spite of the 72

PAGE 81

methodological discrepancies, the findings from the MLEE study (15) generally support the findings of this study: when individuals are sampled on a month-by-month regime, variability over time is observed in their respective E. coli populations. On the other hand, some relatively persistent E. coli subtypes were observed in most individuals. A major consideration in interpretation of the results of the current study, which has been mentioned previously, is the small sample size used in the current study. Only five isolates were collected per individual each month. This number was chosen because other BST studies have collected five isolates or less per individual when creating a library (10, 33, 80), and for the sake of practicality. It is possible that a temporally stable E. coli population within an individual has so many different subtypes that each month a different combination of subtypes would be collected, giving the false impression of temporal instability. It would be important to test for this scenario, if the true diversity of a population was being studied. However, the current study investigated the temporal stability of E. coli populations as it applies to BST. Therefore, although the results of the current study do not represent the dynamics of the total E. coli population, the results do represent the dominant E. coli population that would be used when creating a BST library. The second finding of the current study was that subtype sharing between individuals within a host species increases over time. One explanation for this observation may be that E. coli subtypes are cycled between individuals within a source category. This idea would be most applicable to herd animals, such as beef cattle and horses, because of the close proximity of individuals within a population. Horses, with the most diverse E. coli populations, demonstrated the largest frequency of subtype 73

PAGE 82

sharing over time, while humans, with the least diverse E. coli populations, demonstrated the lowest frequency of subtype sharing over time. These results support the above hypothesis. However, another cause of the observed increase in subtype sharing over time may be a function of sampling effort coupled with fluctuations in the dominant E. coli subtype(s) in the feces at any given time. As more isolates are collected over time, the possibility of observing shared E. coli subtypes increases. Such E. coli subtypes may have been present in multiple individuals at all times but were missed because of the small number of isolates collected from each individual at any given time. Other studies have also observed high levels of subtype sharing within individuals from the same host species over time. The MLEE subtyping study of Australian feral mice (22) determined that 48% of mice shared one of three common subtypes. This frequency of subtype sharing is greater than that observed in the current study, probably because the MLEE study sampled 447 individuals while the current study only sampled three. Ribotyping of E. coli isolates from yearling steers, collected at four sampling events over a fifteen month period, (39) found that 8.3% of observed ribotypes were shared by multiple steers from different sampling events. A higher frequency of ribotype sharing over time was observed in the E. coli populations of horses and beef cattle in the current study. Sixteen percent of the observed beef cattle ribotypes over time were shared at different sampling events, and 13% of the observed horse ribotypes were shared at different sampling events. The frequency of sampling events over time is one possible reason for higher frequency of subtype sharing observed in the current study. The yearling steers were sampled four times over fifteen months, while the cattle and horses were sampled eight and nine times, respectively, over nine months. 74

PAGE 83

In summary, the dominant E. coli populations of all individuals were variable over time when using the described sampling procedure for collecting E. coli isolates, and the frequency of subtype sharing increased over time, suggesting a possible cycling of E. coli subtypes between individuals within a source category. These results demonstrate that E. coli may not be a good candidate for BST libraries if the library is meant for any long-term use. 75

PAGE 84

Chapter 4. Intensive Sampling of E. coli from One Human for One Month Materials and Methods Sample Collection and Processing One human individual was sampled every day for two weeks, followed by weekly sampling for the remainder of one month. Single samples were obtained at each sampling session using rectal swabs. E. coli isolates were processed as described previously (Chapter 2, Materials and Methods). Four MUG-positive E. coli isolates per swab were analyzed by ribotyping and all MUG-positive E. coli isolates (20-24 isolates) were analyzed by ARA. Isolates chosen for ribotyping were a subset of the isolates subtyped by ARA. Two rounds of ribotyping were performed on these isolates. The first round was the procedure described previously (Chapter 2 Materials and Methods), using Hind III as the restriction enzyme in the DNA digestion. The protocol was slightly modified in the second round by substituting Pvu II as the restriction enzyme used in the DNA digestion as well as adding bovine serum albumin (BSA) (Promega, Madison WI) to the digestion mix. BSA was added to increase enzyme (Pvu II) performance according to the manufacturers instructions. Ribotypes were analyzed and compared using BioNumerics software as described previously 76

PAGE 85

Results Distribution of Ribotypes and Antibiotic Resistance Profiles Found in Human X Over Time The fecal bacteria of human X were sampled on a daily basis for two weeks, and once a week for the subsequent two weeks in order to investigate the E. coli population structure within its feces. Human X was chosen for sampling because it had been taking trimethoprim/sulfamethoxazole twice a day for seven days for the treatment of an urinary tract infection. Treatment was completed one week prior to the first sampling event. The hypothesis was that antibiotic use would alter the population structure of E. coli in this human, resulting in greater diversity. All of the E. coli isolates sampled over the course Figure 14. A ribotype gel using Hind III representing isolates from the first two days of sampling from human X. 1 2 3 4 5 6 7 Lane 1; E. coli + control Lanes 2,4 & 7; ribotype 1 Lanes 5 & 6; ribotype 2 Lane 3; empty Figure 15. A ribotype gel using Pvu II representing isolates from the first two days of sampling from human X. 1 2 3 4 5 6 7 Lane 1; E. coli + control Lanes 2,3,4 & 7; ribotype 1 Lanes 5 & 6; ribotype 2 77

PAGE 86

of the experiment shared a single Hind III ribotype when the criterion of 90% similarity was applied. However, upon closer inspection of the membranes, two separate ribotypes were observed (Figure 14). The two Hind III ribotypes differed by a single band in the high molecular weight region of the gel. In order to obtain greater discrimination, ribotyping was performed on these isolates a second time with the restriction enzyme Pvu II (Figure 15), which resolved two distinct subtypes in the E. coli population of human X. Ribotypes 1 and 2, as they were designated, consistently linked with the two ribotypes observed by eye using Hind III. Ribotype 1 made up 91% of E. coli isolates from human X. Antibiotic resistance analysis of the isolates from human X revealed ten different antibiotic resistance patterns (ARPs) (Table 9). The vast majority of E. coli isolates were represented by one of two ARPs. Seventy percent of the E. coli isolates from human X shared ARP 3, while 21% shared ARP 9 (Figure 16). Inspection of the two dominant ARPs revealed that ARP 3 represented an antibiotic resistance pattern showing little Figure 16. Distribution of ARPs of E. coli isolates sampled from human X. Each number corresponds to a unique ARP. Distribution of E. coli isolates among ARPsn = 383 1 2 3 4 5 6 7 8 9 10 78

PAGE 87

resistance to any antibiotics, while ARP 9 represented an antibiotic resistance pattern showing heavy resistance to many antibiotics. Figure 17 is a similarity dendrogram demonstrating isolates that share ARP 3 and ARP 9. The colored squares represent the level of resistance and isolate has to the given antibiotics. The darker the square, the higher the level of resistance. For example, isolate 266 had ARP 9 and displayed resistance to the highest levels of amoxicillin, chloremphenicol, chlortetracycline, doxycycline, oxytetracycline, penicillin, tetracycline, trimethoprim, to the second highest level of cephalothin, and rifampicin, and was susceptible to gentamycin, kanamycin, neomycin, naladixic acid, norfloxacin and streptomycin (Figure 17, Table 10). The E. coli population of human X was also monitored over time by subtyping with ARA (Figure 18). ARP 3, representing the weakly resistant isolates, was observed at every sampling event and was often the dominant pattern at that time. ARP 9, representing the heavily resistant isolates, was consistently present in the first two weeks of sampling and was not observed again for the remainder of the sampling period. Similar results were observed with the E. coli isolates subtyped by ribotyping (Figure 19). Ribotype 1 was found at every sampling date and was the dominant ribotype at each sampling event except one (3/26). Ribotype 2 was only observed at four sampling dates, all within the first two weeks of the sampling period. 79

PAGE 88

Antibiotic Resistance Pattern 1 2 3 4 5 6 7 8 9 10 AMX 0 0 0 0 0 4 32 32 32 0 CEP 8 8 0 0 0 0 8 0 8 8 CLP 4 4 4 4 4 4 4 4 4 20 CHT 20 20 20 20 20 20 20 20 80 4 DOX 0 0 0 0 0 0 4 0 4 0 GEN 0 0 0 0 0 0 0 0 0 0 KAN 0 3 0 3 0 0 0 0 0 0 NA 0 0 0 0 3 0 3 0 0 0 NEO 0 0 0 0 0 0 0 0 0 0 NOR 0 0 0 0 0 0 0 0 0 0 OXY 0 0 0 0 0 0 20 0 20 20 PEN 20 20 20 20 20 20 200 20 200 200 RIF 2 2 2 2 2 2 2 2 2 2 STR 0 0 0 0 0 0 20 0 20 0 TET 0 0 0 0 0 0 0 0 64 64 TRI 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 1 1 TS 0 0 0 0 0 0 5 0 5 5 Table 10. Antibiotic resistance patterns (ARPs) of E. coli sampled from human X. Each number represents a unique ARP. 80

PAGE 89

Figure 17. Similarity dendrogram for antibiotic resistance patterns of E. coli isolated from human X. The numbers correspond to isolate identifiers. The colored squares demonstrate levels of resistance to the various antibiotics. The darker the square, the higher the resistance. 1 0 0 9 5 9 0 8 5 8 0 7 5 7 0 A M X C E P C L P C H T D O X G E N K A N N A N E O N O R O X Y P E N R I F S T R T E T T R I T S 265361731782743708217927537183276372841703627417136375268364269365270366175271367792723688017727336981180266267169172761737717478176 ARP 9 ARP 3 Isolates 81

PAGE 90

Figure 18. Distribution of antibiotic resistance patterns over time for E. coli isolated from human X. Each color and number represents a unique pattern. Human X ARA0510152025303/24/20033/26/20033/28/20033/30/20034/1/20034/3/20034/5/20034/7/20034/9/20034/11/20034/13/20034/15/20034/17/20034/19/20034/21/2003# of isolates with ARP 1 2 3 4 5 6 7 8 9 10 Figure 19. Distribution of ribotypes over time for E. coli isolated from human X. Each color and number represents a unique ribotype. Human X -Ribotyping00.511.522.533.544.53/24/20033/26/20033/28/20033/30/20034/1/20034/3/20034/5/20034/7/20034/9/20034/11/20034/13/20034/15/20034/17/20034/19/20034/21/2003# of isolates with ribotype 2 1 82

PAGE 91

The presence of two dominant ribotypes and two dominant ARPs in the E. coli population in human X prompted the decision to link the ribotypes with their respective ARPs. We hypothesized that ribotype 1 would consistently link with ARP 3 (weakly resistant) and that ribotype 2 would consistently link with ARP 9 (heavily resistant). Ribotype 2 isolates did consistently link with ARP 9, however, E. coli isolates with ribotype 1 were found to have either ARP 3 or 9 during the sampling period (Table 10). E. coli isolates exhibiting ribotype 1 and ARP 9 were only observed in the second week of sampling and were not subsequently observed. The proportion of E. coli isolates with each combination of ribotypes and ARPs is presented in Figure 20. Only 6% of all ribotyped E. coli isolates have the combination of ribotype 1 with ARP 9. Number of E. coliisolates with Ribotype/ARP combination Ribotype 1/ARP 3 Ribotype 1/ARP 9 Ribotype 2/ARP 9 Other Figure 20. Frequency of observations of E. coli ribotype/ARP combinations in human X. The Other group represents those ribotypes that did not display ARP 3 or 9. 83

PAGE 92

Discussion Based on the data obtained from monthly sampling of different individuals (Chapter 3, Results) it was determined that for some individuals, the entire E. coli population can turn over in one month. One concern was that a single month was too long an interval to observe the actual dynamics of an E. coli population, i.e a population may turn over multiple times in a single month. In order to observe more detailed population dynamics over time, the E. coli population of a single individual (human X) was sampled more frequently (on a daily basis). It was not known until after the sampling period was complete that human X had finished antibiotic treatment with trimethoprim/sulfamethoxazole five days before sampling began. Therefore, it is not valid to use the E. coli population of human X during this sampling period to judge the daily population dynamics of an unperturbed E. coli population. However, because of human Xs antibiotic treatment prior to sampling, the current study was able to observe the effect of outside influences, like antibiotic treatment, on the E. coli population of a human individual. The results of the current study demonstrated that there were two dominant subtypes in the E. coli population of human X during the first two weeks of sampling, as determined by ribotyping and antibiotic resistance analysis. One of these subtypes was present throughout the sampling period while the second subtype was only observed in the first two weeks. The antibiotic resistance pattern that was lost represents a heavily resistant pattern, leaving a single dominant pattern of weak resistance. We hypothesized that the antibiotic treatment selected for the subtype carrying the heavily resistant pattern. 84

PAGE 93

However, once treatment ended, this E. coli subtype would no longer have a selective advantage, and it may not have been competitive with the subtype that became dominant in the absence of selective pressures. These antibiotic resistance data correspond to the shift in subtype dominance with ribotyping. The two dominant ribotypes and the two dominant ARPs isolated from human X exhibited similar behavior over time. Ribotypes were linked to their respective ARPs with the expectation that ribotype 1 would consistently link with ARP 3, the weakly resistant pattern, and ribotype 2 would consistently link with ARP 9, the heavily resistant pattern. The majority of ribotypes were associated with their respective ARPs in this fashion, however 6% of the E. coli ribotypes from human X were designated ribotype 1 with ARP 9. One reason for this discrepancy may be that the heavy resistance found in ARP 9 is plasmid driven. Therefore, any ribotype with this plasmid would demonstrate the appropriate resistance pattern (ARP 9), including both ribotype 1 and 2. Further research needs to be conducted on the E. coli isolated from human X to determine if this hypothesis is correct. Ribotyping using Hind III alone was not able to resolve differences in certain E. coli subtypes. By using a second restriction enzyme (Pvu II) on the same isolates in a separate digest, greater discrimination was achieved. The diversity observed in E. coli populations presented in previous sections of this study based on ribotyping (Chapter 2 and 3, Results) is therefore a conservative estimate of the true diversity present. Most ribotyping studies utilize a single restriction enzyme (Hind III) (10, 49, 59, 75). Therefore the results of the current study, although a conservative estimate, have practical implications when addressing issues involving other ribotyping BST studies. 85

PAGE 94

Throughout this study, the term subtype is used as an overarching definition for any unique fingerprint of an E. coli isolate determined by the chosen typing method. Based on this definition, the variability of E. coli subtypes within an individual is not only determined by individual host dynamics but, in part, determined by the subtyping technique. In the current study, more antibiotic resistance patterns were observed in E. coli populations than ribotypes, suggesting greater discriminatory capability for ARA compared to ribotyping. These results demonstrate the need to obtain preliminary data on indicator organism population structure using the subtyping technique that will be used in the actual study. For example, it would not be useful to use the results from phenotypic subtyping as a guide to the sampling strategies for a genotypic subtyping study. As a scientist, one chooses a subtyping method and acknowledges the limitations inherent in the method. Therefore, it must be understood that the conclusions of this study are an estimation of the population dynamics of E. coli as determined by the chosen typing method and may not represent the dynamics of E. coli populations that would be observed using other typing techniques. 86

PAGE 95

Chapter 5. Sampling of Cattle Herds From Four Geographic Regions Materials and Methods Sample Collection and Processing Samples were obtained from five beef cattle individuals within a herd. Four herds were each sampled one time from four different geographic regions of varying distance from one another. Herd TFL was located in Tampa, Florida, herd PCFL was located in Plant City, Florida, herd GFL was located in Gainesville, Fl, and herd HMS was located in Hattiesburg, Mississippi. Samples were obtained using fecal swabs and E. coli isolates were processed as described previously (Chapter 2 Materials and Methods). Fifteen MUG-postive E. coli isolates were processed using ribotyping and the ribotypes were compared using BioNumerics software as described previously (Chapter 2 Materials and Methods). One-way ANOVA with Dunnett's post test was performed using GraphPad InStat version 3.00 (GraphPad Software, San Diego California) to determine differences in the E. coli population structures of different herds. Chi-square analysis (GraphPad Software, San Diego California) was performed to compare the frequency of subtype sharing observed between individuals within different herds and also to compare the frequency of ribotype sharing between individuals from different herds. Significance was determined at P <0.05. 87

PAGE 96

Results Population Structure of E. coli in Various Herds of Beef Cattle The dominant members of the E. coli populations of five individual cows from four different beef cattle herds from various geographical regions were subtyped by ribotyping. Ribotypes were analyzed and diversity indices were calculated as a method of comparison (Table 11). The hypothesis was that E. coli population structures are determined by host and not by location, therefore there would be no difference in the population structure of E. coli in the feces of individuals from different herds. One way ANOVAs determined no significant difference between the E. coli populations of individuals from different beef cattle herds according to the richness estimator (P=0.388), Shannon index (P= 0.120), Simpson index (P=0.054) and Pielous eveness (P=0.388). Ribotype sharing was assessed to determine if E. coli ribotypes within beef cattle demonstrate any geographic distinctiveness. A geographically distinct E. coli population would demonstrate a greater frequency of subtype sharing between individuals from a specific region then between individuals from different regions. Figure 21 represents the distribution of E. coli ribotypes among cattle individuals from all four herds in order to demonstrate sharing within herds versus sharing between herds. More than half of the E. coli isolates from each of the four herds had ribotypes that were found only in one specific herd (their herd of origin). The ribotypes from herds PCFL and GFL were the least broadly distributed, as the majority of E. coli isolates (55% and 62% respectively) 88

PAGE 97

Richness Estimator (S) Shannon Index (H') Pielou's Eveness Index (J') Simpson's Index (1/) Herd TFL (Tampa Fl) Cow A (n = 15) 6 1.52 0.85 4.4 Cow B (n = 15) 3 0.88 0.8 2.4 Cow C (n = 15) 1 0 1 1 Cow D (n = 15) 5 0.7 0.43 1.9 Cow E (n = 15) 2 0.25 0.36 1.1 Mean 3.4 0.67 0.69 2.2 Herd PCFL (Plant City Fl) Cow AA (n=13) 4 1.27 0.91 4.5 Cow BB (n = 15) 6 1.62 0.9 6.3 Cow CC (n = 15) 5 1.4 0.88 3.7 Cow DD (n = 15) 5 1.23 0.77 3.1 Cow EE (n = 15) 6 1.68 0.93 6.7 Mean 5.2 1.44 0.88 4.9 Herd GFL (Gainseville FL) Cow AAAA (n = 15) 8 1.79 0.86 6.3 Cow BBBB (n = 15) 6 1.46 0.82 4.2 Cow CCCC (n = 11) 4 1.16 0.84 3.2 Cow DDDD (n = 15) 5 1.52 0.94 5.5 Cow EEEE (n = 15) 3 0.48 0.44 1.3 Mean 5.2 1.28 0.78 4.1 Herd HMS (Hattiesburg MS) Cow AAA (n = 15) 1 0 1 1 Cow BBB (n = 15) 6 1.4 0.77 3.5 Cow CCC (n = 15) 7 1.62 0.83 4.5 Cow DDD (n = 15) 1 0 1 1 Cow EEE (n = 15) 4 0.95 0.69 2.2 Mean 3.8 0.79 .86 2.4 Table 11. Diversity measurements of E. coli populations within beef cattle individuals from different herds. 89

PAGE 98

from these herds had a ribotype observed in only a single individual beef cow within a herd. About thirty percent of E. coli isolates from each of the three Florida herds had a ribotype that was shared between more than one herd. The largest number of E. coli isolates demonstrating ribotype sharing between multiple herds was observed in herd HMS (47%). These data suggest that E. coli populations are not geographically distinct, as the frequency of between herd ribotype sharing is greater than the frequency of within herd ribotype sharing, however most ribotypes are limited to single individuals. The frequency of ribotype sharing between individuals within a herd was compared for all herds. There was significantly greater sharing of ribotypes between individuals found in each of the Florida herds than was observed between individuals found in the Mississippi herd (Chi-square, P < 0.01 for all tests). There was no significant difference between the amount of ribotype sharing between herds observed in either Florida herds or the Mississippi herd (Chi-square, P =0.833). The E. coli isolates that had a ribotype found in more thane herdt were further analyzed to determine if there was preferential ribotyping sharing, with a single herd sharing its riboytpes with only one of the others, or if all herds were sharing their ribotypes equally amongst themselves (Table 12). A single herd was held constant and the amount of sharing between this herd and the remaining three was compared using multiple chi-square tests. Significance values are presented in Table 13. Significantly greater sharing of ribotypes was observed between herd PCFL and HMS than was observed between herd PCFL and the two remaining Florida herds (TFL and GFL). There was significantly greater sharing of ribotypes between herds TFL and GFL than between GFL and the remaining two herds. These results suggest preferential ribotype 90

PAGE 99

sharing between certain herds, however it does not appear that the preference is based on region, as one of the Florida herds has a greater frequency of ribotype sharing with the Mississippi herd instead of the other Florida herds. Figure 21. Sharing of E. coli ribotypes within and between herds, expressed as percentages of total E. coli isolates isolated from a given herd. Ribotype Sharing Within and Between Herds0%10%20%30%40%50%60%70%TFL (n = 75)PCFL (n = 73)GFL (n = 66)HMS (n = 71)% of E. coli isolate s SingleIndividual >1 individual >1 herd 91

PAGE 100

Herd Ribotypes shared with herd TFL Ribotypes shared with herd PCFL Ribotypes shared with herd GFL Ribotypes shared with herd HMS Herd TFL (Tampa Fl) n = 75 100% 24% 23% 11% Herd PCFL (Plant City Fl) n = 73 1% 100% 6% 22% Herd GFL (Gainesville FL) n = 66 1% 22% 100% 6% Herd HMS (Hattiesburg MS) n = 71 1% 44% 4% 100% Table 12. Percentage E. coli ribotypes shared between each of the herds. 92

PAGE 101

Herd Comparison of the amount of sharing between a given herd and the remaining three herds. TFL vs HMS p = 0.0193 PCFL GFL vs HMs p = 0.009 TFL vs GFL p = 0.84 PCFL vs HMs p = 0.76 TFL PCFL vs GFL p = 0.32 HMS vs GFL p = 0.301 PCFL vs HMS p = 0.74 GFL PCFL vs TFL p = 0.0058 HMS vs TFL p = 0.027 PCFL vs TFL p = 0.10 HMS PCFL vs GFL p = 0.015 GFL vs TFL p = 0.64 Table 13. Chi-square values for comparison of the frequency of ribotype sharing between herds. A similarity dendrogram was produced from the E. coli isolates of all four herds in order to determine which, if any, of the E. coli isolates cluster together (Figure 22). We hypothesized that isolates from the same herd would be found within the same clusters. There was no single cluster that contained all of the E. coli isolates from a single herd. However, E. coli isolates from the same herd were found within the same cluster (cluster A and B in Figure 22). There are clusters made up of E. coli isolates from different herds as well (cluster C in Figure 22). In the dendrogram there is no separation of E. coli isolates based on region, i.e. Florida E. coli isolates and Mississippi E. coli isolates are scattered throughout the dendrogram. This confirms the observations presented in Table 12, demonstrating that E. coli populations are not geographically distinct. 93

PAGE 102

Figure 22. A similarity dendrogram of ribotypes generated from a subset of E. coli isolated from four cattle herds. An equal number of isolates from each herd were selected at random for inclusion in the dendrogram. Dice (Opt:0.20%) (Tol 0.7%-0.7%) (H>0.0% S>0.0%) [0.0%-100.0%]ribotyping 1 0 0 9 0 8 0 7 0 6 0 5 0 4 0 3 0 2 0 1 0 Cow416Cow417Cow430Cow432Cow415Cow35Cow386Cow240Cow242Cow241Cow445Cow447Cow255Cow256Cow257Cow370Cow372Cow270Cow271Cow50Cow52Cow431Cow225Cow227Cow355Cow356Cow357Cow340Cow342Cow371Cow310Cow311Cow312Cow325Cow327Cow286Cow287Cow341Cow446Cow401Cow402Cow326Cow95Cow96Cow97Cow80Cow81Cow82Cow65Cow66Cow67Cow400Cow36Cow37Cow285Cow272Cow51Cow385Cow387Cow226GainesvilleGainesvilleGainesvilleGainesvilleGainesvilleTampaGainesvillePlant CityPlant CityPlant CityGainesvilleGainesvillePlant CityPlant CityPlant CityHattiesburgHattiesburgPlant CityPlant CityTampaTampaGainesvillePlant CityPlant CityHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgHattiesburgPlant CityPlant CityHattiesburgGainesvilleGainesvilleGainesvilleHattiesburgTampaTampaTampaTampaTampaTampaTampaTampaTampaGainesvilleTampaTampaPlant CityPlant CityTampaGainesvilleGainesvillePlant City C B A 94

PAGE 103

Discussion Two major findings, based on ribotyping results, were observed in this phase of the current study. The first finding was that no geographically-driven differences in the E. coli population structures could be discerned in beef cattle. The second major finding was that E. coli populations of beef cattle did not form distinctive groups on the herd level or the state level. The frequency of subtype sharing between individuals within beef cattle herds was no greater than the frequency of subtype sharing between individuals from different beef cattle herds. No greater frequency of ribotype sharing was observed between beef cattle herds located in Florida than between beef cattle herds from different states (i.e. Florida and Mississippi). In general, 62% of the E. coli ribotypes sampled from beef cattle were confined to a single individual. Similar results have been observed in studies of E. coli populations of individual humans from different regions. One such study subtyped the E. coli populations of five families (34 individuals) by MLEE (14). These families were located in two separate regions, New York and Massachusetts. The results of this study were mentioned previously (Chapter 2, Discussion) when discussing the E. coli population structure of individuals from different animal categories. Eighty-five percent of E. coli subtypes were limited to a single individual, and only a small proportion (7%) of E. coli subtypes were shared between individuals from different families. Five percent of the subtypes were shared between individuals from the same state and only two percent of the E. coli subtypes were shared between individuals from different states. These results suggest that geographic distinctiveness on the state level may exist in E. coli populations from humans, however more individuals from different states should be sampled before this 95

PAGE 104

hypothesis is accepted. Another study investigated geographic distinctiveness of E. coli populations in humans from different countries (82), subtyping 178 E. coli isolates by MLEE collected from humans located in the United States, Sweden, and Tonga. Each human sampled provided a single E. coli isolate. The study demonstrated that there is little genetic differentiation between E. coli subtypes from each of the three regions and that the majority of humans sampled had a unique E. coli subtype. Both studies (14, 82) observed very limited, if any, geographic distinctiveness in the E. coli populations of humans, and most subtypes were confined to single individuals. Results from these studies support the observations of the current study, i.e. most E. coli subtypes were limited to a single beef cow. However, the current study demonstrated a much greater frequency of subtype sharing between individuals. The hypothesis was that greater sharing of subtypes would be observed between the Florida herds than between the Florida and Mississippi herds, and that E. coli populations would be geographically distinct on the state level. This was not observed in this study. Sharing between herds appeared random with respect to geographic distance. One Florida herd (PCFL) shared more subtypes with the Mississippi herd (HMS) than with any other herd, while the other two Florida herds (TFL and GFL) demonstrated greater sharing of subtypes between each other. It should be noted that the sample size for the current study was relatively small, as only three herds from Florida were sampled and only one herd from another state (Mississippi) was sampled, and only five cattle were sampled from each herd. Beef cattle herds can be made up of hundreds of cows. It is possible that if more individuals from a single herd were sampled, a greater frequency of subtype sharing within a herd would be observed. More herds from within Florida as 96

PAGE 105

well as from different states would need to be subtyped and compared in order to determine the true extent, if any, of geographic distinctiveness. In summary, the E. coli populations of beef cattle displayed no measurable geographical distinctiveness as the frequency of subtype sharing within herds was no greater than the frequency of subtype sharing between herds. 97

PAGE 106

Chapter 6. Discussion Implications for Bacterial Source Tracking (BST) Each part of the larger study addresses some question relevant to bacterial source tracking (BST) methods that rely on E. coli as the indicator organism: Is the population structure of E. coli dependent on host species? Is there greater subtype sharing between individuals within a hosts species than between individuals from different host species? Are the dominant E. coli populations of individuals invariant or variable over time? What effect does a major physiological perturbance, like antibiotic therapy, have on the temporal variability of an E. coli population? Does geographic distance correlate with the observed variability in E. coli subtypes in cattle? The perfect indicator organism for a BST library would have the following characteristics: 1) it would demonstrate a large amount of subtype sharing within a host species but no subtype sharing between host species, 2) it would have a stable population within individuals over time, and 3) it would exhibit no geographically-associated variability, i.e. individuals within host species would share the same subtypes regardless of location. This final characteristic of the perfect indicator differs from the previously mentioned criteria set out by in Gordon (23). Gordon argues that having a geographically 98

PAGE 107

distinct indicator population is necessary because this characteristic allows a well-sampled library to be highly effective when used on a specific watershed. However, limiting the use of a library to single watersheds requires the construction of a large number of libraries. It has been argued by others that the perfect indicator organism should have no geographic distinction, allowing for the construction of a single library that is applicable over a large geographic area (72). A better understanding of the population dynamics of E. coli within various hosts can aid the resolution of the questions above and help determine whether E. coli is the appropriate indicator microorganism for BST. Optimizing sampling strategy to achieve adequate representation of the E. coli subtypes within host animals, without repeatedly sampling sister clones is necessary for efficient construction of BST libraries. A well-sampled BST library is one that fully represents the diversity of the indicator organism in all sources (80, 85). This study has demonstrated that the diversity of E. coli populations differ based on host species using either ribotyping or antibiotic resistance analysis, therefore sampling strategies need to be adjusted when collecting isolates from various host species. Previous BST studies have utilized arbitrarily chosen sample sizes for library construction (10, 28, 33, 80, 84). Some studies collected the same number of isolates for each source category, while others utilized the available isolates without a defined sampling plan. The results of this study support the argument that the E. coli population structure in various hosts should provide the basis for designing a sampling plan. For example, due to the significantly greater diversity observed in horse E. coli populations compared to those in humans and cattle, more isolates should be collected from a single horse than from a single human. 99

PAGE 108

A consistent observation was that many E. coli ribotypes were limited to single host animals. These data suggest high subytpe variability in E. coli populations determined by ribotyping. Other studies using genetic subtyping techniques on E. coli isolates support this high variability (39, 46, 50). The high genetic variability in E. coli suggests that BST libraries would require very large numbers of isolates from many different individuals to be effective (72, 80). The frequency of ribotype sharing between individuals from the same host species was relatively low, which was unexpected, particularly since the five horses and five cattle were from the same herd. There was no significant difference in the frequency of ribotype sharing within a host species vs. between host species. It is important to note that the sample size in this study was small, as only five individuals for each source category were sampled, and only fifteen isolates from each individual were ribotyped. Greater ribotype sharing might be observed with larger sample sizes (17). In contrast to ribotype sharing, a much greater frequency of antibiotic resistance subtype sharing was observed between individuals. Within each host species, the majority of E. coli isolates had an antibiotic resistance pattern (ARP) found in more than one individual. There was also a large number of E. coli isolates (58%) represented by an ARP found in all three source categories. In fact, a single ARP made up of almost 400 E. coli isolates was observed in all three source categories. Such a large frequency of ARP sharing between sources can severely impair the accuracy of a BST library, as unknown E. coli isolates may have an ARP belonging to a number of different host species. Another observation that has implications for the structure of BST libraries is the fact that over time, the subtypes of E. coli populations within individuals vary. One 100

PAGE 109

sample event per individual will therefore not capture the diversity that is present over time, so more work is needed to the make a representative library, including sampling individuals more than once over time. Having a temporally variable E. coli population within host species may also limit the long term effectiveness of a BST library (23). The current study did not detect geographically distinct E. coli populations in beef cattle. The majority of E. coli isolates sampled had ribotypes that were observed in single beef cows. Furthermore, the frequency of ribotype sharing between beef cattle from the same region was no greater than the frequency of ribotype sharing between beef cattle from different regions. These results suggest that a BST library would only be applicable to the region where the host animals sampled are located. These results are supported by another study of geographic variability that subtyped 1,800 isolates collected from humans, swine, poultry, beef, and dairy cattle located in three regions of Florida by ribotyping (64). The study found that a library constructed of isolates from three regions of Florida was consistently effective for each region. This implies that a BST library could be used over large geographic areas if isolates are collected from all areas. Region-specific E. coli ribotypes in the beef cattle populations sampled may not have been observed because of the small sample size used in this study, as mentioned previously. The results of the current study suggest that E. coli doe not have the appropriate characteristics of the perfect BST organism: there is high subtype variability in E. coli populations, which dictates that very large numbers of isolates are required for constructing a BST library. Subtype sharing between source categories makes source identification problematic. E. coli populations are temporally variable, limiting the long101

PAGE 110

term effectiveness of BST libraries; and, E. coli subtypes do not appear to be extensively shared by host species from different geographic regions, limiting the effective area of a BST library. Therefore, other indicator organisms must be considered for BST, and many studies have begun to use Enterococcus spp. (17, 27, 78, 79, 86). A study using ARA on enterococci isolated from known sources in Virginia (85), addressed many of the same issues discussed in the current study. The study observed that an enterococci BST library was applicable over large geographic areas, however the library was more effective for the areas where samples were collected. The enterococci study (85) demonstrated that their BST library remained effective over one year, demonstrating the potential for long term use. These results suggest that enterococci are better BST organisms than E. coli, however further research is needed on Enterococcus spp. before it can become the standard BST organism. Two subtyping techniques were used in this study, ribotyping and antibiotic resistance analysis (ARA). Ribotyping is a much more expensive and time consuming method than antibiotic resistance analysis. For this reason, up to five times as many isolates were subtyped using ARA compared to ribotyping. Although results from the subtyping methods generally agreed, two major differences were observed. First, larger diversity values were consistently associated with E. coli populations subtyped by ARA. One reason for the observed difference may be the difference in sample size; more isolates were subtyped with ARA so there was a greater chance of observing relatively rare subtypes. The second major difference was observed in the amount of subtype sharing between individuals from different source categories. E. coli isolates subtyped by ARA demonstrated significantly greater amounts of subtype sharing between host species 102

PAGE 111

compared to ribotyping results. The major conclusion from these results was that the population structure of E. coli was determined not only by source, but also by the typing technique and sample size chosen. Many studies support this observation and have demonstrated different E. coli population dynamics within the same isolate set based on subtyping techniques (4, 11, 46, 61, 65). In summary, preliminary studies of E. coli populations using two different subtyping techniques (ribotyping and ARA) suggest that E. coli may not be the appropriate microorganism for library-based bacterial source tracking for the following reasons: 1) Different sampling strategies are needed when collecting isolates from each source category to construct a representative BST library. An E. coli BST library may not be effective for long periods of time due to the instability of E. coli populations within individuals. BST libraries may only be effective in the limited region where isolates were collected because E. coli populations tend to be individual-specific. Preliminary studies must be performed using the chosen subtyping technique because E. coli population structures differ based on technique as well as source. 103

PAGE 112

References 1. American Public Health Association. 1998. Standard Methods for the Examination of Water and Wastewater, 20th edition ed. American Public Health Association, Washington DC. 2. American Public Health Association. 1905. Standard Methods of Water Analysis,, 1 ed. American Public Health Association, Washington, DC. 3. Arthur, M., R. D. Arbeit, C. Kim, P. Beltran, H. Crowe, S. Steinbach, C. Campanelli, R. A. Wilson, R. K. Selander, and R. Goldstein. 1990. Restriction fragment length polymorphisms among uropathogenic Escherichia coli isolates: pap-related sequences compared with rrn operons. Infect Immun. 58(2):471-9. 4. Avery, S. M., E. Liebana, C. A. Reid, M. J. Woodward, and S. Buncic. 2002. Combined use of two genetic fingerprinting methods, pulsed-field gel electrophoresis and ribotyping, for characterization of Escherichia coli O157 isolates from food animals, retail meats, and cases of human disease. J Clin Microbiol. 40(8):2806-12. 5. Bitton, G., Koopman, B., and Jung, K. 1995. An assay for the enumeration of total coliforms and Escherichia coli in water and wastewater. Water Environment Research. 67(6):906-909. 6. Bitton, G. 1994. Wastewater Microbiology. Wiley-Liss, Inc., New York. 7. Bochner, B. R. 1996. The Biolog MicroStation System and General Procedures for Identifying Environmental Bacteria and Yeast. In W. P. Olson (ed.), Automated Microbial Identification and Quantitation Technologies for the 2000's. Interpharm Press, NY, New york. 8. Bruce, J. 1996. Automated system rapidly identifies and characterizes microorganisms in food. Food Technology. January:77-81. 9. Byappanahalli, M. N., and R. S. Fujioka. 1998. Evidence that tropical soil can support the growth of Escherichia coli. Wat Sci Tech. 38(12):171-174. 10. Carson, C. A., 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(4):1503-7. 11. Carson, C. A., B. L. Shear, M. R. Ellersieck, and J. D. Schnell. 2003. Comparison of Ribotyping and Repetitive Extragenic Palindromic-PCR for Identification of Fecal Escherichia coli from Humans and Animals. Appl Environ Microbiol. 69(3):1836-9. 12. Caugant, D. A., B. R. Levin, G. Lidin-Janson, T. S. Whittam, C. Svanborg Eden, and R. K. Selander. 1983. Genetic diversity and relationships among strains of Escherichia coli in the intestine and those causing urinary tract infections. Prog Allergy. 33:203-27. 104

PAGE 113

13. Caugant, D. A., B. R. Levin, I. Orskov, F. Orskov, C. Svanborg Eden, and R. K. Selander. 1985. Genetic diversity in relation to serotype in Escherichia coli. Infect Immun. 49(2):407-13. 14. Caugant, D. A., B. R. Levin, and R. K. Selander. 1984. Distribution of multilocus genotypes of Escherichia coli within and between host families. J. Hyg. Camb. 92:377-384. 15. Caugant, D. A., B. R. Levin, and R. K. Selander. 1981. Genetic diversity and temporal variation in the E. coli population of a human host. Genetics. 98(3):467-90. 16. 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 Microbiol. 66(6):2572-7. 17. Dontchev, M., Whitlock, J.E., and Harwood, V. J. 2003. Presented at the 103rd General Meeting, American Society for Microbiology, Washington, D.C. 18. Dufour, A. P. 1984. Health Effects Criteria for Fresh Recreational Waters EPA-600/1-84-004. U. S. Environmental Protection Agency. 19. Faith, N. G., J. A. Shere, R. Brosch, K. W. Arnold, S. E. Ansay, M. S. Lee, J. B. Luchansky, and C. W. Kaspar. 1996. Prevalence and clonal nature of Escherichia coli O157:H7 on dairy farms in Wisconsin. Appl Environ Microbiol. 62(5):1519-25. 20. Geldrich, E. E., and B. A. Kenner. 1969. Concepts of fecal streptococci in stream pollution. J Water Pollut Control Fed. 41:R336-R352. 21. Gerba, C. P., S. M. Goyal, R. L. Labelle, I. Cech, and G. F. Bodgan. 1979. Failure of indicator bacteria to reflect the occurrence of enteroviruses in marine waters. Am J Public Health. 69(11):1116-1119. 22. Gordon, D. M. 1997. The genetic structure of Escherichia coli populations in feral house mice. Microbiology. 143(Pt 6):2039-46. 23. Gordon, D. M. 2001. Geographical structure and host specificity in bacteria and the implications for tracing the source of coliform contamination. Microbiology. 147(Pt 5):1079-85. 24. Gordon, D. M., S. Bauer, and J. R. Johnson. 2002. The genetic structure of Escherichia coli populations in primary and secondary habitats. Microbiology. 148(Pt 5):1513-22. 25. 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(6):2690-8. 26. Guttman, D. S., and D. E. Dykhuizen. 1994. Clonal divergence in Escherichia coli as a result of recombination, not mutation. Science. 266(5189):1380-3. 27. 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(5):792-9. 28. Hagedorn, C., S. L. Robinson, J. R. Filtz, S. M. Grubbs, T. A. Angier, and R. B. Reneau, Jr. 1999. Determining sources of fecal pollution in a rural Virginia watershed with antibiotic resistance patterns in fecal streptococci. Appl Environ Microbiol. 65(12):5522-31. 105

PAGE 114

29. 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 Methods. 53(3):387-99. 30. Hanes, N. B., and R. Fragala. 1967. Effect of seawater concentration on survival of indicator bacteria. Water Pollut Control Fed. 39:97. 31. Harris, H. 1966. Enzyme polymorphism in man. Proc. R. Soc. Lond. Ser. B. 164:298-310. 32. Harwood, V. J., J. Butler, D. Parrish, and V. Wagner. 1999. Isolation of fecal coliform bacteria from the diamondback terrapin (Malaclemys terrapin centrata). Appl Environ Microbiol. 65(2):865-7. 33. Harwood, V. J., J. Whitlock, and V. Withington. 2000. Classification of antibiotic resistance patterns of indicator bacteria by discriminant analysis: use in predicting the source of fecal contamination in subtropical waters. Appl Environ Microbiol. 66(9):3698-704. 34. Havelaar, A. H., Pot-Hogeboom, W.M., Furuse K., Pot, R.,and Hormann, M.P. 1990. F-Specific RNA bacteriophages and sensitive host strains in faeces and wastewater of human and animal origin. J Appl Bacteriol. 69(1):30-37. 35. Hill, M. O. 1973. Diversity and eveness: A unifying notation and its consequences. Ecology. 54:427-432. 36. Hill, T. C. J., Walsh, K.A, Harris, J.A., and Moffett, B.F. 2003. Using ecological diversity measures with bacterial communities. FEMS Microbiology Ecology. 43:1-11. 37. 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(11):3960-6. 38. Hughes, J. B., Hellmann, J.J., Ricketts, T.H., and Bohannan, J.M. 2001. Counting the Uncountable: Statistical Approaches to Estimatiing Microbial Diversity. Appl Environ Microbiol. 67(10):4399-4406. 39. Jenkins, M. B., P. G. Hartel, T. J. Olexa, and J. A. Stuedemann. 2003. Putative temporal variability of Escherichia coli ribotypes from yearling steers. J Environ Qual. 32(1):305-9. 40. Kauffmann, F. 1944. Zur serolgie der coli-gruppe. Acta Pathol. Microbiol. Scand. 21:20-45. 41. Krumperman, P. H. 1983. Multiple antibiotic resistance indexing of Escherichia coli to identify high-risk sources of fecal contamination of foods. Appl Environ Microbiol. 46(1):165-170. 42. Lewontin, R. C., and J. L. Hubby. 1966. A molecular approach to the study of genic heterozygosity in natural populations. II. Amount of variation and degree of heterozygosity in natural populations of Drosophila pseudoobscura. Genetics. 54:595-609. 43. Ludwig, J. A., and Reynolds, J. F. 1988. Statistical Ecology. A primer on methods and computing. John Wiley and Sons, New York. 44. Maier, R. M., Pepper, I.L. and Gerba, C.P. 2000. Environmental Microbiology. Academic Press, San Diego. 106

PAGE 115

45. Marzouk, Y., S. M. Goyal, and C. P. Gerba. 1980. Relationship of viruses and indicator bacteria in water and wastewater of Israel. Water Res. 14(11):1585-1590. 46. Maslow, J. N., T. S. Whittam, C. F. Gilks, R. A. Wilson, M. E. Mulligan, K. S. Adams, and R. D. Arbeit. 1995. Clonal relationships among bloodstream isolates of Escherichia coli. Infect Immun. 63(7):2409-17. 47. Miescier, J. J., and V. J. Cabelli. 1982. Enterocooci and other microbial indicators in municipal wastewater effluents. J Water Pollut Control Fed. 54(12):1599-1606. 48. Milkman, R. 1973. Electrophoretic variation in Escherichia coli from natural sources. Science. 182(116):1024-6. 49. Morrison, D., Woodford, N., Barrett, S.P., Sisson, P., and Cookson, B.D. 1999. DNA banding pattern polymorphism in vancomycin-resistant Enterococcus feacium and criteria for defining strains. J Clin Microbiol. 37(4):1084-1091. 50. Nagy, B., R. A. Wilson, and T. S. Whittam. 1999. Genetic diversity among Escherichia coli isolates carrying f18 genes from pigs with porcine postweaning diarrhea and edema disease. J Clin Microbiol. 37(5):1642-5. 51. National Committee for Clinical Laboratory Standards. 2001. Performance standards for antimicrobial susceptibility testing. National Committee for Clinical Laboratory Standards, Wayne, PA. 52. Ochman, H., T. S. Whittam, D. A. Caugant, and R. K. Selander. 1983. Enzyme polymorphism and genetic population structure in Escherichia coli and Shigella. J Gen Microbiol. 129(Pt 9):2715-26. 53. Orskov, F., I. Orskov, and A. J. Furowicz. 1972. Four new Escherichia coli O antigens, O148, O151, O152, O153, and one new H antigen, H50, found in strains isolated from enteric diseases in man with a discussion on the future numbering of K antigens. Acta Pathol Microbiol Scand [B] Microbiol Immunol. 80(3):435-40. 54. Orskov, I., and F. Orskov. 1976. Five new Escherichia coli K antigens, K95, K96, K97, K98 and K100. Acta Pathol Microbiol Scand [B]. 84B(6):321-5. 55. Orskov, I., and F. Orskov. 1970. The K antigens of Escherichia coli. Re-examination and re-evaluation of the nature of L antigens. Acta Pathol Microbiol Scand [B] Microbiol Immunol. 78(5):593-604. 56. Orskov, I., and F. Orskov. 1977. Special O:K:H serotypes among enterotoxigenic E. coli strains from diarrhea in adults and children. Occurrence of the CF (colonization factor) antigen and of hemagglutinating abilities. Med Microbiol Immunol (Berl). 163(2):99-110. 57. Orskov, I., F. Orskov, A. Birch-Andersen, M. Kanamori, and C. Svanborg-Eden. 1982. O, K, H and fimbrial antigens in Escherichia coli serotypes associated with pyelonephritis and cystitis. Scand J Infect Dis Suppl. 33:18-25. 58. Osawa, S., K. Furuse, and I. Watanabe. 1981. Distribution of ribonucleic acid coliphages in animals. Appl Environ Microbiol. 41(1):164-8. 59. Parveen, S., K. M. Portier, K. Robinson, L. Edmiston, and M. L. Tamplin. 1999. Discriminant analysis of ribotype profiles of Escherichia coli for differentiating human and nonhuman sources of fecal pollution. Appl Environ Microbiol. 65(7):3142-7. 107

PAGE 116

60. Pipes, W. O. 1982. Bacterial Indicators of Pollution. CRC Press, Inc, Boca Raton, FL. 61. Pupo, G. M., R. Lan, P. R. Reeves, and P. R. Baverstock. 2000. Population genetics of Escherichia coli in a natural population of native Australian rats. Environ Microbiol. 2(6):594-610. 62. Rodrigues, J., I. C. Scaletsky, L. C. Campos, T. A. Gomes, T. S. Whittam, and L. R. Trabulsi. 1996. Clonal structure and virulence factors in strains of Escherichia coli of the classic serogroup O55. Infect Immun. 64(7):2680-6. 63. Savageau, M. A. 1983. Escherichia coli habitats, cell types, and melecular mechanisms of gene control. Amer. Nat. 122:732-744. 64. Scott, T. M., Parveen, S., Portier, K.M, Rose, J.B, Tamplin, M.L., Farrah, S.R., Koo, A., and Lukasik, J. 2003. Geographical variation in ribotype profiles of Escherichia coli isolates from humans, swine, poultry, beef, and dairy cattle in Florida. Appl Environ Microbiol. 69(2):1089-1092. 65. Selander, R. K., and Levin, B.R. 1980. Genetic structure and variation in natural populations of Escherichia coli, p. 1625-1648. In F. C. Neidhardt (ed.), Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology. Merican Society for Microbiology, Washington DC. 66. Selander, R. K., D. A. Caugant, H. Ochman, J. M. Musser, M. N. Gilmour, and T. S. Whittam. 1986. Methods of multilocus enzyme electrophoresis for bacterial population genetics and systematics. Appl Environ Microbiol. 51(5):873-84. 67. Selander, R. K., T. K. Korhonen, V. Vaisanen-Rhen, P. H. Williams, P. E. Pattison, and D. A. Caugant. 1986. Genetic relationships and clonal structure of strains of Escherichia coli causing neonatal septicemia and meningitis. Infect Immun. 52(1):213-22. 68. Selander, R. K., and B. R. Levin. 1980. Genetic diversity and structure in E. coli populations. Science. 210:545-547. 69. Selander, R. K., and B. R. Levin. 1980. Genetic diversity and structure in Escherichia coli populations. Science. 210(4469):545-7. 70. Silveira, W. D., M. Lancellotti, A. Ferreira, V. N. Solferini, A. F. P. De Castro, E. G. Sehling, and M. Brocchi. 2003. Determination of the clonal structure of avian Escherichia coli strains by isoenzyme and ribotyping analysis. J. Vet. Med. B. 50:63-69. 71. Simpson, J. M., B. Martineau, W. E. Jones, J. M. Ballam, and R. I. Mackie. 2002. Characterization of fecal bacterial populations in canines: effects of age, breed and dietary fiber. Microb Ecol. 44(2):186-97. 72. Simpson, J. M., J. W. Santo Domingo, and D. J. Reasoner. 2002. Microbial source tracking: state of the science. Environ Sci Technol. 36(24):5279-88. 73. Sinton, L. W., Finlay, R.K., and Hannah, D. J. 1998. Distinguishing human from animal faecal contamination in water: a review. New Zealand journal of marine and freshwater research. 32:323-348. 74. 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(1):230-237. 108

PAGE 117

75. Soto, S. M., Martinex, N., Guerra, B., Gonzalez-Hevia, M.A., Mendoza, M.C. 2000. Usefulness of genetic typing methods to trace epidemiologically Salmonella serotype Ohio. Epidemiol Infect. 125(3):481-490. 76. Urakawa, H., K. Kita-Tsukamoto, and K. Ohwada. 1999. 16S rDNA restriction fragment length polymorphism analysis of psychrotrophic vibrios from Japanese coastal water. Can J Microbiol. 45:1001-1007. 77. Wagner, E. K., and Martinez, J.H. 2003. Basic Virology, 2nd ed. Blackwell Science, Malden, MA. 78. Wallis, J. L., and Taylor, H.D. 2003. Phenotypic population characteristics of the enterococci in wastwater and animal faeces: implications for the new European directive on the quality of bathing waters. Water Sci Technol. 47(3):27-32. 79. Wheeler, A. L., P. G. Hartel, D. G. Godfrey, J. L. Hill, and W. I. Segars. 2002. Potential of Enterococcus faecalis as a human fecal indicator for microbial source tracking. J Environ Qual. 31(4):1286-93. 80. 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(17):4273-82. 81. Whittam, T. S. 1989. Clonal dynamics of Escherichia coli in its natural habitat. Antonie Van Leeuwenhoek. 55(1):23-32. 82. 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(1):67-83. 83. Whittam, T. S., H. Ochman, and R. K. Selander. 1983. Multilocus genetic structure in natural populations of Escherichia coli. Proc Natl Acad Sci U S A. 80(6):1751-5. 84. 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(11):3997-4002. 85. Wiggins, B. A., Cash, P.W., Creamer, W.S., Dart, S.E., Garcia, P.P., Gerecke, T.M., Han, J., Henry, B.L., Hoover, K.B., Johnson, E.L., Jones, K.C., McCarthy, J.G., McDonough, J.A., Mercer, S.A., Noto, M.J., Park, H., Phillips, M.S., Purner, S.M., Smith, B.M, Stevens, E.N. 2003. Use of antibiotic resistance analysis for representativeness testing of multiwatershed libraries. Appl Environ Microbiol. 69(6):3399-3405. 86. 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(8):3483-6. 87. Wyer, M. D., J. M. Fleisher, J. Gough, D. Kay, and H. Merrett. 1995. An investigation into parametric relationships between Enterovirus and faecal indicator organisms in the coastal waters of England and Wales. Water Res. 29(8):1863-1868. 109