Contributions of Molecular Epidemiology to the Understanding of Infectious Disease Transmission, Pathogenesis, and Evolution

Contributions of Molecular Epidemiology to the Understanding of Infectious Disease Transmission, Pathogenesis, and Evolution BETSY FOXMAN, PHD

PURPOSE: Describe the contributions of molecular genetics to our understanding of the molecular epidemiology of infectious diseases caused by bacteria. METHODS: Synthesize the literature, highlighting work on Escherichia coli and Group B streptococcus. RESULTS: 1) Commensal bacteria are genetically and phenotypically diverse. 2) Disease-causing strains of commensal bacteria often have special characteristics than allow them to be distinguished from common inhabitants. 3) Colonization by commensal bacteria is dynamic. 4) Commensal bacteria are transmitted between individuals. CONCLUSIONS: Applications of epidemiologic principles to bacterial populations gives insight into the natural history of colonization and transmission in the human host. Ann Epidemiol 2007;17:148–156. Ó 2007 Elsevier Inc. All rights reserved. KEY WORDS:

Molecular epidemiology, Infectious disease, Bacterial pathogens.

INTRODUCTION Our bodies are home to trillions of bacteria. They live on our skin and in our nose, mouth, throat, bowel flora, and the vaginal cavity. The total number of bacterial cells outnumbers human cells by 10 to 1 (1). Bacteria present in our normal flora that do not cause disease are called commensals; however, commensal bacteria are not all the same, but rather a diverse group with different potentials to cause disease. A major research goal of my research group has been to distinguish between commensals with high and low potentials to cause disease. Until recently, I used to imagine that bacteria grew on surfaces in our body like colonies on an agar plate or floated aimlessly as separate individuals in urine and mucus, a life consisting of eating and procreating. However, they do not. Bacteria live in complex structures like biofilms, they communicate, and they exchange genetic material. They sacrifice for the greater good, and our health is dependent upon them as much as their health is dependent upon us. In addition to producing vitamins essential to our health, ‘‘good’’ bacteria help protect us from invasion

by other bacteria. For example, lactobacillus in the vaginal cavity protects against overgrowth of gram-negative organisms, and the presence of certain types of Streptococcal pneumoniae in the nasal pharynx protects against cocolonization with Staphylococcus aureus (2, 3). I discuss four aspects of commensal bacteria and how these aspects impact on the design, conduct, and interpretation of epidemiologic studies of bacterial species that can be either commensal or disease causing. (i) Commensal bacteria are genetically and phenotypically diverse. (ii) Diseasecausing strains of commensal bacteria often have special characteristics that allow them to be distinguished from common inhabitants. (iii) Colonization by commensal bacteria is dynamic. (iv) Commensal bacteria are transmitted between individuals. I discuss each of these points and illustrate with an example. I end the report with a discussion of the implications of these points with respect to studies of antibiotic resistance.

COMMENSAL BACTERIA ARE GENETICALLY AND PHENOTYPICALLY DIVERSE From the Center for Molecular and Clinical Epidemiology of Infectious Diseases, Department of Epidemiology, University of Michigan School of Public Health, Ann Arbor, MI. Address correspondence to: Betsy Foxman, Center for Molecular and Clinical Epidemiology of Infectious Diseases, Department of Epidemiology, University of Michigan School of Public Health, 109 Observatory St, Ann Arbor, MI 48109-2029. Tel.: (734) 764-5487; fax: (734) 764-3192. E-mail: [email protected]. Originally presented as the Wade Hampton Frost Lecture at the 2005 American Public Health Association Meetings, Philadelphia, PA, December 12, 2005. Received June 23, 2006; accepted September 18, 2006. Ó 2007 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010

The ability to rapidly sequence genetic material has led to nothing less than a revolution in our understanding of bacteria. One of the greatest revelations is that bacteria are more genetically diverse than most of us might have possibly imagined. This diversity flies in the face of many of the things that I believed we knew about bacteria. Most of us were trained to think of bacteria as clonal: dividing in two to create perfect copies of themselves. This is a misconception on a number of fronts. First, there are errors in replication; thus, each time a bacterium 1047-2797/07/$–see front matter doi:10.1016/j.annepidem.2006.09.004

AEP Vol. 17, No. 2 February 2007: 148–156

Selected Abbreviations and Acronyms PCR Z polymerase chain reaction PFGE Z pulsed-field gel electrophoresis MLST Z multilocus sequence typing GBS Z group B Streptococcus UTI Z urinary tract infection ERIC Z enterobacterial repetitive intergenic consensus sequences ECO-SENS Z a cross-sectional study of acute, uncomplicated, community-acquired UTI in 16 western European countries and Canada

divides, there is some potential, albeit small, for change in genetic sequence. Second, bacteria can acquire new genetic information by a variety of mechanisms, including acquisition of plasmids, integration of bacterial viruses (phage), and, in some cases, direct uptake of free DNA. These factors all increase bacterial genetic diversity. Third, there is constant natural selection ongoing within the human host, and the selection is not applied equally to all bacterial species. For example, bacteria living in a biofilm, either single or multiple bacterial species that communicate and work together, can be protected from the selection process. Biofilms are found attached to such surfaces as your teeth, throat, and nasal pharynx and may be protected from certain selection factors. Because these bacteria live in communities, bacteria living inside the biofilm are better able to survive antibiotic therapy. Biofilms are one explanation why antibiotic-sensitive bacteria can be cultured after antibiotic therapy (4). The misconception that a bacterial species colonizing a single individual is clonal can lead to laboratory methods that unwittingly support the notion that bacteria are clonal. In clinical microbiology, if bacteria are cultured, we generally streak for isolation and randomly select one colony for study. Thus, we have only one colony for study. The colony will have developed from a single bacterium and most likely is a clonal group. However, if we believe there is a diverse population, we must select isolates in a way that will give a representative sample of the organisms present. The idea that commensal or pathogenic bacteria of a particular species are all the same, i.e., clonal, within an individual human host, rather than a population of individuals, may have been reinforced inadvertently by limitations of our typing techniques. Developments in molecular biology, in particular, of polymerase chain reaction (PCR), have greatly enhanced our ability to detect differences between bacterial isolates. However, even the goldstandard techniques, such as pulsed-field gel electrophoresis (PFGE), currently in use for epidemiologic studies are limited for detecting diversity. PFGE is an indirect measure of genetic content based on the number of restriction sites present that will be recognized and cut by the selected restriction enzyme. The cuts result in DNA fragments that, when separated by using electrophoresis, give a pattern.

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

149

For example, PFGE is used by PulseNet, part of the Centers for Disease Control and Prevention food surveillance system, to detect foodborne disease clusters (5). PFGE is extremely useful when we already suspect a relationship, similar to being able to correctly differentiate blood relatives from those who married into the family by looking at attendees at a family reunion. However, if we had a random group of individuals selected from one part of the country, we would do less well in relating individuals to each other, even if we included family name along with physical characteristics. The other gold-standard method is to genetically sequence a limited number of gene loci, called multilocus sequence typing (MLST). MLST was developed for determining evolutionary relationships, i.e., relationships on a very long time scale, and the gene loci sequenced are chosen because they are highly conserved. However, if we consider only very conserved genes, many strains of a bacterial species appear the same. For example, by using MLST, shigella are the same as Escherichia coli (6). Although MLST is a useful construct for studying the genetic origins of species, it is less useful for studying the dynamics of colonization within and transmission between human hosts. It also can be misleading if applied inappropriately in outbreak investigation because it generally is less discriminatory than PFGE and thus might inappropriately group nonoutbreak strains with outbreak strains. As an example of the diversity among commensal bacteria, consider group B Streptococcus (GBS). GBS is a common inhabitant of the bowel and vaginal flora and also is found in the throat and urethral opening. In our studies of healthy men and women at the University of Michigan, the prevalence of colonization was 34% for women and 20% for men (7). GBS causes a variety of infections, particularly during pregnancy and among neonates and the elderly or infirm. There are nine known serotypes, a phenotypic classification based on surface antigens. Serotype is associated with propensity to cause disease. We isolated GBSs from healthy students living in a single dormitory and typed the isolates by using PFGE. The dendrogram in Figure 1 shows the diversity among commensal isolates from healthy individuals (dendrogram created by Dr. Lixin Zhang by using PFGE results by Dr. Stephanie Borchardt). A dendrogram is a way of showing relationships between isolates; the scale shows how similar the isolates are by using a pairwise method of determining similarity; band sizes are compared between all possible pairs of isolates. Although the enzyme used, chosen because it is standard in the field for GBS, produces only eight bands, we see tremendous variety in the isolates, which all have the same serotype. The variety is even more evident when we compare the genetic sequence. An analysis of six sequenced strains and data in the literature found that only w80% of the genome

150

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

AEP Vol. 17, No. 2 February 2007: 148–156

FIGURE 1. Dendrogram of group B streptococcus serotype III isolates. Dotted lines mark isolates with 50%, 70%, and 90% similarity by using the nearest neighbor joining method.

of any one strain shared the core genome (8). GBS is clearly a diverse organism.

DISEASE-CAUSING STRAINS OF COMMENSAL BACTERIA OFTEN HAVE SPECIAL CHARACTERISTICS THAT ALLOW THEM TO BE DISTINGUISHED FROM COMMON INHABITANTS This point is made most easily by using a specific organism and infection. Urinary tract infection (UTI) is an infection anywhere in the urinary tract usually caused by inhabitants of the bowel flora. Estimated annual incidences of UTI are 12% for women and 3% for men (9); annual total direct

and indirect costs were $1.6 billion in 1995 (10). The costs are greater now, not just because of inflation, but also because bacteria that cause UTI are becoming increasingly resistant to antibiotics. Therefore, more expensive alternative antibiotics are needed. When disease is caused by a commensal bacterial species, it commonly was held that the disease can be attributed to host factors. E coli, a common bowel inhabitant, causes the vast majority of UTIs. Urine is an excellent medium for growing bacteria; thus, we might posit that introduction of E coli from bowel flora into the bladder from either the bladder or vaginal cavity should be sufficient to initiate the disease process. Thus, UTI should occur when the host engages in such activity as sexual intercourse that moves E coli into the

AEP Vol. 17, No. 2 February 2007: 148–156

bladder; disease risk is enhanced if the host urinates infrequently, has a shorter urethra, or has a shorter distance between the rectal opening and urethral opening and is unable to fully empty the bladder. This models fits with several lines of evidence: UTI is more common in women then men, consistent with the hypothesis regarding closer distance between urethral and rectal opening and that of shorter urethral length (9). UTI is associated with sexual activity, which moves bacteria from the rectum into the vaginal cavity into the urethra, and postcoital prophylactic antibiotic treatment is an effective therapy in some women with chronic recurring UTIs. However, other evidence fits less well. On average, women presenting with a first UTI have had more than one sex partner. Therefore, movement of bacteria into the bladder is insufficient (11). High bacterial counts are found in women’s bladders after sexual activity, but bacteriuria usually spontaneously clear without symptoms ever developing (12). Recent antibiotic therapy for any reason is associated with increased risk for a symptomatic episode. Two inferences can be made from this counter evidence. First, whereas there are normally sterile protected sites in the body, in reality, protected sites are only sterile because the body works continuously to keep them sterile. Bacteria frequently colonize all body sites with an opening to environment, such as the bladder or the nipple, albeit for short periods. Even blood often is transiently colonized; e.g., if you brush your teeth vigorously, bacteria from your mouth enter the bloodstream. Second, the presence of high bacterial counts is not sufficient to cause disease: not all E coli are equally likely to cause UTI. E coli are diverse: the species designation E coli includes a wide range of bacteria that vary in size from 4.5 to 5.5 megabases. An estimated 3 megabases are held in common, or 65% of the genetic material (13). Conversely, humans and chimpanzees have sequence homology of 95% to 97%. Whereas a diverse group of E coli cause UTI, they can be separated from other E coli when typed by using conserved genes. There is even a designation: uropathogenic E coli (14). Although this implies a distinct group of uropathogenic E coli, the strains classified as uropathogenic E coli are a highly diverse group. No one factor or set of factors has yet to be identified that allows us to uniquely classify uropathogens. For example, we screened several collections of E coli for the presence of several genes that may have some role in UTI pathogenesis. Table 1 lists results of screening collections of isolates causing pyelonephritis, isolates causing cystitis, and commensal fecal isolates from healthy individuals (15). Because the E coli genome is heterogeneous in gene content, genes can be present or absent, as well as occur in different forms or alleles. Here, we examine only the presence and absence of genes associated with

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

151

TABLE 1. Presence and absence of uropathogenic virulence factors among Escherichia coli causing pyelonephritis and cystitis, compared commensal E coli isolates. Adapted from Marrs et al., 2002 Virulence signature 0000000010000 0000100010000 0001100010000 0011100010000 0011100110000 1001100010100 1011100010100 1101111010001 1111110010100

Pyelonephritis (n Z 153)

Cystitis (n Z 237)

Commensal (n Z 269)

0.6 1.3 0.6 1.9 3.2 8.4 29.4 8.4 4.5

5.5 4.2 8.4 7.2 8.4 4.6 10.6 13.9 1.3

16.0 3.0 10.8 11.2 2.6 8.9 5.2 4.5 1.5

UTI virulence, creating a binary score that shows the presence or absence of all 13 genes screened, called the virulence signature. Note that although 29% of pyelonephritis strains have the same set of genes, these genes also are found in the other groups. No more than 14% of isolates from women with a first UTI share the same virulence signature, and the most common virulence signature differs among collections. Note also that the most common virulence signature among fecal isolates has only one of the genes in the screen set, but this virulence signature is found in all collections. To find bacterial characteristics that distinguish between disease-causing and commensal strains, we essentially need to conduct a case–control study. This requires: (i) collections of representative invasive and commensal strains, and (ii) the capability to screen the collections for putative characteristics, ideally in a high-throughput format. If disease-causing and commensal isolates are populations of bacteria, not single clones, we must sample from those populations. This means using a different sampling strategy than is standard clinically because our goal is to select a representative sample from the diverse array of isolates present within a species, rather than identify what species is present. To determine the appropriate sampling fraction, we need to know the range of diversity present: in a properly collected midstream urine sample from an outpatient, only one uropathogen is present 95% of the time; i.e., 5% of the time, there is a second strain present of the same species. Thus, if we only pick one isolate, we will always miss the second strain. Multiple strains must be picked to increase the probability of sampling the second strain, if present. However, we do not know how often multiple strains are present for other protected sites, such as breast milk. In addition, commensals often grow in sites that are not protected, such as the nose, throat, and bowel flora, where there is an ecology of microorganisms. When we culture bacteria from these sites, we generally use culture media that suppress the growth of organisms other than those of interest. There

152

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

also is the potential to suppress the growth of an unusual strain of the species of interest. If we wish to determine the etiologically relevant period or ask such questions as whether individuals are colonized with new strains after antibiotic therapy or there is reemergence of previous strains, we need to conduct studies to describe the dynamics of colonization. Describing the dynamics requires representative samples from the bacterial populations and frequent sampling; thus, large numbers of samples. The implication of all this sampling is that we need to have large numbers of isolates from each individual for testing. For example, recent studies of the dynamics of commensal E coli colonization picked 28 isolates per specimen (16). Thus, a high-throughput method is essential. We developed a novel microarray platform that makes it feasible to rapidly screen 1500 bacterial isolates in duplicate on a single slide for the presence or absence of genes. We call this a library on a slide (17). After the slides are prepared, each screening experiment takes 3 days. We also are developing a molecular typing system based on the presence or absence of genetic material that can be conducted by using this high-throughput format; thus, all 1500 isolates can be typed at once. Figure 2 shows an example of a library on a slide that was probed for a gene. Each dot represents the entire genome of a single bacterium that is positive for the gene sequence. Because we can determine the molecular type of large numbers of isolates, we can ask such questions as ‘‘What are the dynamics of colonization?’’ Because we can screen large numbers of isolates, we also can rapidly test the predictions made from bioinformatic analyses of the genetic sequence of human pathogens about the potential importance of putative virulence genes. The human pathogen sequencing project has made available the genetic sequence of several pathogens. In the past, we learned much from studies of a handful of laboratory strains; however, bacteria are much more diverse than we previously imagined. Whereas a virulence factor might be very important in disease in a single individual, that does not necessarily imply that it is always important or is even found in similar clinical cases of disease. We can examine genetic sequences from the human pathogen project and make guesses about function, but we do not know that a putative gene coding for a virulence factor is actually associated with disease. Library on a slide facilitates the identification of these associations. We conduct our case–control studies on isolates that caused disease by using commensal isolates as controls. These hypothesis-generating studies enable us to rapidly determine whether a putative virulence gene occurs with any frequency among population-based isolates and whether it is differentially present among disease-causing versus commensal isolates. Factors that appear more frequently among invasive isolates can be pursued by using standard molecular biologic techniques to determine the mechanism and function.

AEP Vol. 17, No. 2 February 2007: 148–156

My colleagues and I at the Center for Molecular and Clinical Epidemiology of Infectious Diseases at the University of Michigan currently are conducting such case–control studies for a number of different bacterial species: E coli, GBS, Haemophilus influenzae, and Mycobacterium tuberculosis. These studies allow us to evaluate quickly whether genes or gene regions for which the putative function suggests they are associated with disease actually are associated with disease. Thus, we are applying epidemiologic methods to make sense of the genetic sequences generated by the human pathogen sequencing project.

COLONIZATION BY COMMENSAL BACTERIA IS DYNAMIC If commensal bacteria are diverse and have their own transmission system, how often do they turn over within the individual, i.e., what are the dynamics of colonization? The incidence and duration of colonization are key parameters to understanding the transmission system; they also are essential for designing studies because we need to know how frequently samples should be collected and what the etiologically relevant period is for the disease of interest. However, we know surprising little about the dynamics of colonization with commensal bacteria. We have the general sense that colonization is dynamic, but the timing is unclear. Thus, for most organisms, we do not know how frequently we should be collecting specimens, and we have yet to discover the etiologically relevant period. Our group recently conducted a study of the incidence of GBS colonization. We collected specimens every 3 weeks for 12 weeks. By using the observed incidence and prevalence, we estimated the average duration of colonization to be 13.7 weeks for women and 8.5 weeks for men. However, our limit of detection was 3 weeks, and some individuals carried GBS for only one interval. In addition, some individuals carried GBS for the entire study period; therefore, longer follow-up would have been desirable (18). Because GBS is a major cause of neonatal disease, but is prevented by prophylactic treatment with intrapartum antibiotics, our results emphasize the need to determine colonization at time of labor or to vaccinate at a time that will prevent colonization during the etiologic risk period. Duration of colonization also varied by serotype. For GBS, different serotypes have a different propensity to cause disease. Because there are different dynamics of colonization by each serotype, we also might ask if the transmission system is the same for each serotype. For example, we observed that sex partners of persons colonized with GBS have a very high rate of GBS colonization (19). However, GBS also is isolated from individuals who have never engaged in sexual activity. One possible explanation is transmission by another route.

AEP Vol. 17, No. 2 February 2007: 148–156

GBS serotype Ib also is a major fish pathogen (20), and GBS is a major cause of lactation mastitis in cows (21), suggesting the potential for foodborne transmission. Our study dealt with colonization only, not disease. GBS-related disease almost always occurs in a host that is compromised in some way, although we are seeing increasing numbers of GBS UTIs in otherwise healthy young women. What is uncertain for both GBS-related diseases and other diseases caused by commensal bacteria is how often commensal bacteria become pathogens, rather than disease following the acquisition of a more pathogenic strain of the commensal species. The observation that invasive disease from a commensal organism frequently follows antibiotic use is one argument in support of the hypothesis, suggesting that existing flora may protect against colonization with a more virulent strain. As many meeting abstracts used to end: further study is needed.

COMMENSAL BACTERIA ARE TRANSMITTED BETWEEN INDIVIDUALS If commensal bacteria are diverse and colonization is dynamic, bacteria must be transmitted from other individuals. That is, commensal bacteria are not just there. In addition, if we believe that disease-causing strains have special characteristics that allow them to inhabit protected body sites, these characteristics also may be associated with transmission. To return to the UTI example, the reason that uropathogenic E coli have special characteristics is because there is a benefit to them to be able to live in the urinary tract. Examined from the perspective of the bacterium, living in the

FIGURE 2. Library on a slide microarray platform compared with a United States quarter. Each spot on the slide contains the total genomic DNA of a strain of Escherichia coli. Photograph and slide by Dr. Lixin Zhang.

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

153

urinary tract has a number of benefits: there is lots of food and little competition. This means you can transmit your progeny to additional hosts in large enough doses that there is a good probability of colonization, especially from people who do not believe they need to wash their hands after urinating. The doubling time of E coli can be as little as 30 minutes; thus, several generations can grow between urinary voidings. Thus, if uropathogenic E coli have a propensity to cause UTI, they also must have a transmission system. We have some evidence that uropathogenic E coli are transmitted between sex partners at a greater rate than E coli that live in the rectal flora. We conducted a study among heterosexual couples in which we compared cocolonization rates among couples in which the woman had a UTI with those for couples in which the woman did not have a UTI. UTI couples were twice as likely to share E coli strains as non-UTI couples (22). We also analyzed the data from the perspective of the E coli. We had, on average, three isolates per couple: from UTI couples, the uropathogen and additional E coli strains from the rectal flora of the participating man and woman. We also had E coli strains from rectal flora of non-UTI couples. All isolates were typed by using PFGE to distinguish between strain types. In this analysis, uropathogens were six times more likely to be found in both members of a couple compared with nonuropathogens. If we believe uropathogens are transmitted, but that there are a large number of diverse strains of uropathogens, there must be a series of propagated epidemics that together lead to the overall high UTI endemic rate of 12% among women. Because we do not have a way to definitively identify uropathogens, outbreaks have been difficult to detect. There have been two reports of clusters of multidrug-resistant clones; both detected because of a unique antibiotic profile. There also is some evidence of seasonality, which potentially suggests transmission of an infectious agent. Explanations of seasonality can be grouped into three classes: pathogen appearance and disappearance, environmental changes, and host-behavior changes. We detected a seasonal pattern in a study of pyelonephritis, or kidney infection, which is the more severe form of UTI, by using insurance claims in Korea (23). Although there were geographic differences in incidence ranging from 26 to 45 cases/10,000, we observed summer peaks in each area. Southern areas showed greater incidence and earlier peak times, but there was no clear pattern by latitude and differences were not significant. Although I give only one or two examples of each statement, I believe they generally hold. Commensal bacteria are genetically and phenotypically diverse. Disease-causing strains of commensal bacteria often have special characteristics that allow them to be distinguished from common inhabitants. Colonization by commensal bacteria is dynamic. Commensal bacteria are transmitted between individuals.

154

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

Finally, I discuss how these statements are especially salient to studies of antibiotic resistance. Antibiotic resistance has been identified as a major public health problem in several Institute of Medicine and other government reports (24–28) and is first on the list of targeted areas in the Centers for Disease Control and Prevention strategy for preventing emerging infectious diseases (29). Antibiotic resistance poses a direct threat to human health by decreasing the effectiveness of treatment; the need to use alternative, frequently more toxic treatments; and prolonging illness, all of which increase the risk for more serious morbidity and the potential for mortality. There also are economic costs, including the need for alternative more expensive therapies and longer duration of therapy (28). The American Society of Microbiology estimated in 1995 that treating resistant infections in the United States cost $4 billion annually, or approximately 0.5% of total health care costs. A research goal identified in the 2003 Institute of Medicine report (28) is to ‘‘develop a fuller understanding of how microbes evolve when faced with drugs that threaten their survival’’ as a means to identify more effect control methods. Bacteria can acquire resistance to antibiotics in two ways. The first way is spontaneous mutation. Although point mutations occur at a low rate, approximately 2  10 3 per genome, because the number of bacteria in our bodies is so high, in the trillions, the probably of a resistance mutation occurring that can be selected for with antibiotic treatment is not as small as we might hope. Second, many bacteria readily take up DNA from the environment and exchange genetic material with each other that includes genes coding for resistance to antibacterial agents. Bacterial sex occurs both within and between species. Thus, commensal bacteria are likely to have resistant genes and may be a reservoir for the emergence and proliferation of antibiotic resistance genes in disease-associated bacteria. Resistance to antibiotics among E coli isolated from the bowel of healthy individuals has increased during the past 20 years. Therefore, understanding both the transmission of resistance among commensal bacteria and transmission of bacteria among people are essential to identify effective strategies to limit the spread of antibiotic resistance. Although there are many different important questions regarding antibiotic resistance, one question is the relative importance of horizontal spread, i.e., transmission of antibiotic resistance genes between bacteria, versus vertical spread, which is through division and transmission of a single resistant bacterial strain, or clone, between people, also known as clonal expansion. Once resistance is acquired through bacterial sex, the now resistant bacteria also will undergo clonal expansion. However, the potential public health interventions are different if the primary explanation

AEP Vol. 17, No. 2 February 2007: 148–156

for resistance is transmission of a single resistant clone between individuals versus transmission of resistance between bacteria. We addressed this question among uropathogenic E coli. For many years, the first-line treatment has been a combination of two drugs, trimethoprim and sulfamethoxazole, and resistance rates were low, presumably because E coli had to become resistant to both trimethoprim and sulfamethoxazole for the treatment to be ineffective. However, between the late 1980s and late 1990s, the resistance of UPEC to trimethoprim-sulfamethoxazole doubled in most places (30, 31). A clinical rule of thumb for a maximal acceptable level of treatment failure caused by resistance is 20%, which trimethoprim-sulfamethoxazole is rapidly approaching. Because trimethoprim-sulfamethoxazole resistance can be transmitted through a mobile genetic element called an integron, which may include genes for resistance to other antibiotics, treatment with an antibiotic other than trimethoprim-sulfamethoxazole in the previous 2 weeks increases bacterial resistance to trimethoprim-sulfamethoxazole two-fold; treatment with trimethoprim-sulfamethoxazole increases resistance 17-fold (30). In 2001, a report suggested that 9% to 11% of trimethoprim-sulfamethoxazole resistance among uropathogenic Escherichia coli (UPEC) might be attributed to a single widespread clone (32). The clone was defined by use of a PCR-based method of molecular typing based on the presence of repetitive elements called ERIC. PCR products then are run on a gel giving a band pattern. The investigators observed a distinct pattern, which they called pattern A. Preliminary results suggested that clonal expansion had a major role in the spread of antibiotic resistance. My group had participated in the original report, lending a subset of our isolates. However, we had some concerns with regard to the validity of using the particularly typing method, ERIC typing, to infer clonality. Thus, we explored further. We tested 90 isolates from women with UTIs and 79 rectal isolates collected from healthy women during the same period for the presence of ERIC pattern A and found 23. If ERIC pattern A represents a clone, we would anticipate that all pattern A isolates would share the same mechanism of antibiotic resistance. We measured this indirectly by screening for the presence of integrons, a type of transposon capable of integrating new gene cassettes, that carries trimethoprim-sulfamethoxazole resistance. Integrons can be different sizes, here, from 1.2 to 2.1 kilobases, and there are two types, class I and class II. There are 17 known genes coding for trimethoprim, called dihydrofolate reductase DFR genes; we observed three in this set. Among the 23 isolates showing pattern A, we observed six combinations of the genes. However, only 61% of the ERIC-type pattern A had the same combination of genes coding for trimethoprim-sulfamethoxazole resistance, inconsistent with the hypothesis of

AEP Vol. 17, No. 2 February 2007: 148–156

clonal expansion, but consistent with the hypothesis of horizontal spread of resistance (33). To further explore the potential for horizontal versus vertical spread of trimethoprim-sulfamethoxazole resistance, we screened 350 trimethoprim-sulfamethoxazole–resistant isolates from the ECO-SENS collection for the presence of genes coding for trimethoprim and sulfamethoxazole resistance, as well as the presence of integrons that carry these resistances (34). ECO-SENS is a cross-sectional study of acute, uncomplicated, community-acquired UTIs in 16 western European countries and Canada (35). The multiple mechanisms of trimethoprim and sulfamethoxazole resistance included in our screen were found in all geographic regions. We then classified each isolate by the presence or absence of each sul allele, integrated dfr alleles, and integron type. We observed a wide variety of gene combinations. There was no significant difference in the distribution of individual gene alleles, integrons, or gene combinations by country or region. Clustering of genetic mechanisms by country would suggest clonal expansion. The lack of clustering is consistent with the hypothesis of horizontal gene transfer. Therefore, we concluded that in both Michigan and the ECO-SENS collection that horizontal gene transfer has a greater role in the spread of trimethoprim-sulfamethoxazole resistance than clonal expansion. Thus, for trimethoprim-sulfamethoxazole, intervention strategies should not be directed toward limiting dissemination of a few clones, but should seek ways to minimize the spontaneous emergence of resistance in multiple locations.

CONCLUSIONS In the studies presented here, I show the effective application of epidemiologic principles to studies of bacterial populations, which gives us insight into the natural history of bacterial colonization and transmission to the human host. By examining representative isolates from normal bacterial flora, we discovered that bacteria are diverse, have different potentials to cause disease, and may have different transmission systems. However, much remains to be done. Measuring the dynamics of strain-specific colonization is a key parameter for determining the etiologic relevant period and modeling the overall transmission system; the dynamics are unknown for most bacterial species. In addition, we have no idea of the extent that host behavior, such as hand washing and diet, modify the dynamics. In addition, disease is an interaction between the host and agent; the ability to examine this interface by using proteomics and genomics is almost upon us. Ultimately, insights gained from studies of bacteria that are both disease causing and commensals will help us more effectively prevent colonization with diseasecausing bacteria and maintain protective bacteria, which,

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

155

in turn, will help preserve antibiotic therapies and identify alternative methods to both prevent and treat bacterial infections. The author thanks her colleagues Drs Carl Marrs and Lixin Zhang for their comments on this presentation and their contributions to the work presented; Dr Lixin Zhang for creating the figures; and staff members Ms. Patricia Tallman and Dr. Usha Srinivasan and the many students who contributed to the work presented here.

REFERENCES 1. Sears C. A dynamic partnership: Celebrating our gut flora. Anaerobe. 2005;11:247–251. 2. Bogaert D, van Belkum A, Sluijter M, Luijendijk A, de Groot R. Colonisation by Streptococcus pneumoniae and Staphylococcus aureus in healthy children. Lancet. 2004;363:1871–1872. 3. Regev-Yochay G, Dagan R, Raz M, Carmeli Y, Shainberg B, Derazne E. Association between carriage of Streptococcus pneumoniae and Staphylococcus aureus in children. JAMA. 2004;292:716–720. 4. Chambless JD, Hunt SM, Stewart PS. A three-dimensional computer model of four hypothetical mechanisms protecting biofilms from antimicrobials. Appl Environ Microbiol. 2006;72:2005–2013. 5. Gerner-Smidt P, Hise K, Kincaid J, Hunter S, Rolando S, Guttierez EP. PulseNet USA: A five-year update. Foodborne Pathog Dis. 2006;3:9–19. 6. Lan R, Alles M, Donohoe K, Martinez M, Reeves P. Molecular evolutionary relationships of enteroinvasive Escherichia coli and Shigella spp. Infect Immun. 2004;72:5080–5088. 7. Manning SD, Neighbors K, Tallman PA, Gillespie B, Marrs CF, Foxman B. Prevalence of group B streptococcus colonization and potential for transmission by casual contact in healthy young men and women. Clin Infect Dis. 2004;39:380–388. 8. Tettilin H, Masignani V, Cieslewicz JM, Donati C, Medini D, Ward NL, et al. Genome analysis of multiple pathogenic isolates of Streptococcus agalactiae: implications for the microbial ‘‘pan-genome.’’ PNAS. 2005;102:13950–13955. 9. Foxman B, Brown P. Epidemiology of urinary tract infections: Transmission and risk factors, incidence, and costs. Infect Dis Clin North Am. 2003;17:227–241. 10. Foxman B, Barlow R, d’Arcy H, Gillespie B, Sobel J. Urinary tract infection: Estimated incidence and associated costs. Ann Epidemiol. 2000;10:509–515. 11. Foxman B, Geiger AM, Palin K, Gillespie B, Koopman JS. First-time urinary tract infection and sexual behavior. Epidemiology. 1995;6:162–168. 12. Nicolle L, Bradley S, Colgan R, Rice J, Schaeffer A, Hooton TM. Infectious Diseases Society of America guidelines for the diagnosis and treatment of asymptomatic bacteriuria in adults. Clin Infect Dis. 2005;40:643–654. 13. Yang F, Zhang X, Chen L, Jiang Y, Yan Y, Tang X, et al. Genome dynamics and diversity of Shigella species, the etiologic agents of bacillary dysentery. Nucl Acids Res. 2005;33:6445–6448. 14. Donnenberg M, Whittam T. Pathogenesis and evolution of virulence in enteropathogenic and enterohemorrhagic Escherichia coli. J Clin Invest. 2001;107:539–548. 15. Marrs CF, Zhang L, Tallman P, Manning SD, Somsel P, Raz R. Variations in 10 putative uropathogen virulence genes among urinary, faecal and periurethral Escherichia coli. J Med Microbiol. 2002;51:138–142. 16. Schlager TA, Hendley JO, Bhang JL, Wobbe CL, Stapleton A. Variation in frequency of the virulence-factor gene in Escherichia coli clones colonizing the stools and urinary tracts of healthy prepubertal girls. J Infect Dis. 2003;188:1059–1064.

156

Foxman MOLECULAR EPIDEMIOLOGY AND INFECTIOUS DISEASES

AEP Vol. 17, No. 2 February 2007: 148–156

17. Zhang L, Foxman B, Gilsdorf JR, Marrs CF. Bacterial genomic DNA isolation using sonication for microarray analysis. Biotechniques. 2005;39:640–644.

27. Davis SA, Gordon DM. The influence of host dynamics on the clonal composition of Escherichia coli populations. Environ Microbiol. 2002;4:306–313.

18. Foxman B, Gillespie B, Manning SD, Howard LJ, Tallman P, et al. Incidence and duration of group B Streptococcus by serotype among male and female college students living in a single dormitory. Am J Epidemiol. 2006;163:544–551.

28. Knobler SL, Lemon SM, Narafi M, Borrough T. The Resistance Phenomenon in Microbes and Infectious Disease Vectors: Implications for Human Health and Strategies for Containment, Workshop Report, Forum on Emerging Infections, Institute of Medicine. Washington, DC: National Academy; 2003.

19. Manning SD, Tallman P, Baker CJ, Gillespie B, Marrs CF, Toxman B. Determinants of co-colonization with group B Streptococcus among heterosexual college couples. Epidemiology. 2002;13:533–539. 20. Evans JJ, Klesius PH, Gilbert PM, Shoemaker CA, Al Sarawi MA, et al. Characterization of beta-haemolytic group B Streptococcus agalactiae in cultured seabream, Sparus auratus L, and wild mullet, Liza klunzingeri (Day), in Kuwait. J Fish Dis. 2002;25:505–513.

29. Prevention C. Preventing emerging infectious diseases: A strategy for the 21st century. Overview of the updated CDC plan. Morbid Mortal Wkly Rep Recommend Rep. 1998;47(RR-15):1–14. 30. Brown PD, Freeman A, Foxman B. Prevalence and predictors of trimethoprim-sulfamethoxazole resistance among uropathogenic Escherichia coli isolates in Michigan. Clin Infect Dis. 2002;34:1061–1066.

21. Dogan B, Schukken YH, Santisteban C, Boor KJ. Distribution of serotypes and antimicrobial resistance genes among Streptococcus agalactiae isolates from bovine and human hosts. J Clin Microbiol. 2005;43:5899–5906.

31. Hooton TM, Foxman B, Fritsche TR, Nicolle LE. Acute uncomplicated cystitis in an era of increasing antibiotic resistance: A proposed approach to empiric therapy. Clin Infect Dis. 2004;39:75–80.

22. Foxman B. Epidemiology of urinary tract infections: Incidence, morbidity, and economic costs. Am J Med. 2002;113(Suppl 1A):S5–13. 23. Ki M, Park T, Choi B, Foxman B. The epidemiology of acute pyelonephritis in South Korea, 1997-1999. Am J Epidemiol. 2004;160:985–993.

32. Manges AR, Johnson JR, Foxman B, O’Bryan TT, Fullerton KE, et al. Widespread distribution of urinary tract infections caused by a multidrug-resistant Escherichia coli clonal group. N Engl J Med. 2001;345:1007–1013.

24. Harrison P, Lederberg J. Antimicrobial Resistance: Issues and Options, Workshop Report, Forum on Emerging Infections, Institute of Medicine. Washington, DC: National Academy; 1998.

33. France AM, Kugeler KM, Freeman A, Zalewski CA, Blahna M, Zhang L. Clonal groups and the spread of resistance to trimethoprim-sulfamethoxazole in uropathogenic Escherichia coli. Clin Infect Dis. 2005;40:1101–1107.

25. Council NR. The Use of Drugs in Food Animals: Benefits and Risks. Committee on Drug Use in Food Animals, Panel on Animal Health, Food Safety, and Public Health, Board on Agriculture, National Research Council. Washington, DC: National Academy; 1999.

34. Blahna MT, Zalewski CA, Reuer J, Kahlmeter G, Foxman B, Marrs CF. The role of horizontal gene transfer in the spread of trimethoprim-sulfamethoxazole resistance among uropathogenic Escherichia coli in Europe and Canada. J Antimicrob Chemother. 2006;57:666–672.

26. Resistance ITF. Annual Report on a Public Health Action Plan to Combat Antimicrobial Resistance, CDC Publication. Available at: http:// www.cdc.gov/drugresitance. Accessed March 1, 2006.

35. Kahlmeter G, for ECO.SENS. An international survey of the antimicrobial susceptibility of pathogens from uncomplicated urinary tract infections: The ECO.SENS Project. J Antimicrob Chemother. 2003;51: 69–76.