The spread of drug resistance in Gram Negative Bacteria. By Hillary Fanya Berman

Transcription

1 The spread of drug resistance in Gram Negative Bacteria By Hillary Fanya Berman A Dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Infectious Diseases and Immunity in the Graduate Division of the University of California, Berkeley Committee in charge: Professor Lee W. Riley, Chair Professor Kimmen Sjölander Professor Sangwei Lu Fall 2013

2

3 Abstract The Spread of Drug Resistance in Gram Negative Bacteria by Hillary Fanya Berman Doctor of Philosophy in Infectious Diseases and Immunity University of California, Berkeley Professor Lee W. Riley, Chair Drug resistant infections with Gram negative bacteria have become increasingly common. Some of these Gram negative bacteria are pan resistant. To better understand the spread of drug resistant infections we used E. coli as a model system. By using this system we investigated both the spread of drug resistance genes and bacteria that are drug resistant. Drug resistance genes and their mobile genetic elements are frequently identified from environmental saprophytic organisms. It is widely accepted that the use of antibiotics in agriculture selects for drug resistant microorganisms, which are then spread from the farm environment to humans through the consumption of contaminated food products. We wished to identify novel drug resistance genes from microbial communities on retail food products. To do this, we created metagenomic plasmid libraries from microbiota isolated from retail spinach samples. From these libraries we identified five unique plasmids that increased resistance to antimicrobial drugs. These plasmids were identified in E. coli that grew on plates that contained ampicillin (pamp), aztreonam (pazt), ciprofloxacin (pcip), trimethoprim (ptrm), and trimethoprim-sulfamethoxazole (psxt). We identified open reading frames with similarity to known classes of drug resistance genes in the DNA inserts of all five plasmids. These drug resistance genes conferred resistance to fluoroquinolones, cephalosporins, and trimethoprim, which are classes of antimicrobial drugs frequently used to treat human Gram negative bacterial infections. These results show that novel drug resistance genes are found in microbiota on retail produce items. Food saprophytes may serve as an important reservoir for new drug-resistance determinants in human pathogens. The clinical management of infections caused by E. coli, including meningitis, is greatly complicated when the organism becomes resistant to broad-spectrum antibiotics. We sought to characterize the antimicrobial susceptibility, multilocus sequence type (MLST), and presence of known drug resistance genes of E. coli that caused meningitis between 1996 and 2011 in Salvador, Brazil. We then compared these findings to E. coli isolates from community acquired urinary tract infections (UTI) that occurred during the same time period and in the same city. We found 19% of E. coli from cases of meningitis and less than 1% of isolates from UTI to be resistant to third- 1

4 generation cephalosporins. The sequence types of E. coli from cases of meningitis included ST 131, ST 69, ST 405, and ST 62, which were also found among isolates from UTI. These sequence types of E. coli have previously been isolated from produce items and food animals. Additionally, among the E. coli isolates that were resistant to thirdgeneration cephalosporins, we found genes that encode the extended-spectrum betalactamases CTX-M-2, CTX-M-14, and CTX-M-15. These observations demonstrate that, compared to E. coli isolated from cases of community acquired UTI, those isolated from cases of meningitis are more resistant to third-generation cephalosporins, even though the same sequence types are shared between the two forms of extraintestinal infections. The results of our investigation of retail food products indicate that drug resistant genes are frequently found in these products and that these food items may be a reservoir for drug resistance genes found in human pathogens. We also found that many of the sequence types of E. coli that cause both UTI and meningitis have previously been isolated from retail food products. These findings suggest that bacteria that are considered to be normal flora of food products may play an important role of the spread of drug resistant genes and drug resistant bacteria. 2

5 Table of Contents CHAPTER ONE... 1 Introduction... 1 Overview of Drug Resistance Strategies of Gram Negative Bacteria... 4 Antibiotic Degrading or Altering Enzyme... 4 Modification, Replacement, or Protection of the Drug Target... 4 Exclusion or Active Efflux of the Antibiotic... 5 Dissertation Chapters and Overview... 6 Chapter One... 6 Chapter Two... 6 Chapter Three... 6 Chapter Four... 6 Chapter Five... 7 Human Subjects Statement... 7 CHAPTER TWO... 8 Identification of Novel Antimicrobial Resistance Genes from Microbiota on Retail Spinach.. 8 Abstract... 9 Background... 9 Results and Discussion Isolation of antibiotic resistant clones Identification of antibiotic resistance genes and phylogenetic analysis Conclusions Experimental Procedures Metagenome plasmid library construction and screening Competing interests Authors contributions Acknowledgements CHAPTER THREE Bioinformatics Analysis of Drug Resistance Conferring Amino Acid Sequences Introduction Methods Bioinformatics pipeline Bioinformatics methods used pcip Results and Discussion pamp Results and Discussion Conclusions and Future Directions CHAPTER FOUR Distribution of strain type and antimicrobial susceptibility of Escherichia coli causing meningitis in a large urban setting in Brazil Abstract Introduction Methods Isolate collection Multi- locus sequence typing (MLST) and sequence analysis Data analysis Results Characterization of E. coli causing meningitis Comparison of E. coli that caused UTI and meningitis Discussion Acknowledgements CHAPTER FIVE i

6 Conclusions and Future Directions Chapter Two Chapter Three Chapter Four Future Directions APPENDIX Tables and Figures Table 1: Minimum inhibitory concentrations (MIC) for E. coli containing the indicated plasmids Table 2: Open reading frames identified in each of the 5 drug resistance- conferring plasmids Table 3: Results of antimicrobial susceptibility testing of 36 E. coli isolates from cases of meningitis Table 4: Drug resistant phenotypes, beta- lactamase (blatem, blashv, or blaoxa), and sequence types found in collections of E. coli from meningitis and urinary tract infection (UTI) Figure 1: Diagram of open reading frames and DNA insert size Figure 2:... Error! Bookmark not defined. Alignment of Qnr proteins, Qnr- like proteins, and the novel Cip- Qnr protein... Error! Bookmark not defined. Figure 3: Qnr protein family Tree Figure 4: Predicted Cip- Qnr Structure Figure 5: Model of Amp- beta- lactamase Figure 6:... Error! Bookmark not defined. pamp- beta- lactamase alignment... Error! Bookmark not defined. Figure 7: Sequence types vs. year of isolate collection REFERENCES AND BIBLIOGRAPHY ii

7 Chapter One Introduction 1

8 Infections caused by drug resistant bacteria are a major cause of morbidity and mortality. Each year more than two million people in the United States acquire a drug resistant bacterial infection [1]. Additionally, at least 23,000 people die each year as the direct result of a drug resistant infection [1]. Infections with drug resistant Gram negative bacteria are particularly concerning [2]. These Gram negative multidrug- resistant infections often occur while a person is hospitalized. These infections are often caused by species belonging to commensal flora of the human host and it has become clear that the drug resistance genes are spreading in the community [1]. The spread of antimicrobial resistance genes has made previously manageable bacterial infections increasingly more difficult to treat [2]. Additionally, there has been a gradual decline in the development of new antimicrobial drugs, especially those that target Gram negative bacterial pathogens [2]. Drug resistant Gram negative bacteria pose a particular challenge because of the diverse strategies they employ to become drug resistant and the ease with which the drug resistance genes are spread [1-3]. The aim of this dissertation is to better understand the reservoirs and spread of drug resistance genes and the drug resistant bacteria that carry these genes. We wish to better understand their spread so that the emergence of novel drug resistance genes in human pathogens can be better predicted, with the ultimate aim of preventing these infections entirely. Development of new antimicrobial drugs to treat Gram negative bacteria has proven to be particularly challenging. There are two major types of challenges to the development of new antimicrobial drugs to treat Gram negative infections. The first challenge is biological, and the second challenge is conducing the clinical trials necessary to show that novel drugs are effective [2]. There are two biological reasons why developing drugs for Gram negative bacteria is difficult. The first is that Gram negative bacteria have an outer membrane that is located outside of the peptidoglycan layer. The permeability of the outer membrane is tightly controlled and many potential drugs are unable to cross it. The second reason is there are many different mechanisms of resistance that are already present in multidrug resistant Gram negative bacteria. It has proven to be extremely challenging to design a drug to overcome current mechanisms of drug resistance [3]. The second set of challenges, concerning the clinical trials, are the result of the current regulatory process of approval for antimicrobial drugs and how this impacts the expected return on investment for companies developing these drugs. Designing clinical trials that fairly and ethically test the effectiveness of antimicrobial drugs for the treatment of serious Gram negative bacterial infections are uniquely challenging. This is because, due to ethical concerns, it not possible to test a new antimicrobial drug in a placebo controlled trial. This means that much larger and prohibitively expensive non- 2

9 inferiority trials must be done instead. Due to these challenges associated with testing the drug for FDA approval, and because antimicrobial drugs are only used for a short time, they are perceived to have a small return on investment for the pharmaceutical companies that have traditionally taken these drugs though the clinical testing phase [2]. Because of the challenges associated with developing new antimicrobial drugs, it is unlikely there will be new drugs to treat these infections in the near future and these infections will remain untreatable. Therefore, a better understanding of how drug resistance genes and drug resistant Gram negative bacteria are spread is imperative to prevent these infections. Drug resistant infections often occur while a patient is in a hospital, but these infections are caused by species of bacteria that are normal constituents of the human microbiota [2, 4]. The majority of drug resistance in these bacteria, in particular Gram negative bacteria, is caused by genes that originated as part of the genome of another species of bacteria. These genes are thought to originate from environmental microbes, are carried on mobile genetic elements, and are spread by horizontal gene transfer [5]. This is thought to occur because most antibiotic drugs are natural products, or derivatives of natural products, that are produced by microbes [6, 7]. Consequently, many naturally occurring genes that confer resistance to antibiotic drugs already exist, even before the drugs are developed for medical use [6-8]. These resistance genes become problematic once they are spread to clinically important bacteria on mobile elements. Mobile elements include plasmids, integrons, and transposons [9]. It is widely accepted that the use of antimicrobial drugs in clinical and agricultural settings has selected for antimicrobial resistance [10]. Over half of all antimicrobial drugs used in the United States are used in agriculture [1, 11]. Antimicrobial drugs are used in agriculture for therapeutic, prophylactic, and growth promotion purposes [12]. Consequently, drug resistant bacteria are frequently found in retail meat and produce items [13, 14]. Contaminated food products have been implicated in numerous outbreaks of drug resistant enteric pathogens, including Salmonella, E. coli, and Campylobacter [1, 12, 13, 15, 16]. These drug resistant bacteria may directly cause drug resistant infections or be a reservoir of drug resistance genes that can be spread to other bacteria that reside in the human microbiome. Numerous national and international programs screen food products for the presence of antibiotic resistant bacteria. Surveillance programs to screen retail food products for contamination with antimicrobial resistant bacteria are part of national and international strategies to contain the spread of antimicrobial resistance [12, 13]. These screening programs look for food borne pathogens and organisms known to colonize humans, such as E. coli and Salmonella [13, 15]. 3

10 The microbiota of food products includes many commensal and saprophytic organisms that are subjected to the same selective pressure from antimicrobial drugs as food borne pathogens. However, since the vast majority of commensal and saprophytic bacteria do not cause human infections, they are generally not included in the screening programs that address issues of food safety. Consequently, these bacteria are rarely screened for antimicrobial resistance. When the microbiota has been screened, saprophytic organisms have been found to harbor the same resistance genes as human pathogens. The saprophytic organisms that make up the majority of microbiota on food may be a previously unrecognized reservoir for the antimicrobial resistance genes that are eventually found in human pathogens [17]. We studied E. coli as a model for Gram negative bacteria that are drug resistant to better understand the spread of both drug resistant bacteria and genes. We focused on E. coli for two reasons. First, E. coli has been extensively used as a model organism so there are many tools available for its study; importantly, these tools include functional metagenomic libraries. Second, E. coli is often the cause of drug resistant infections. An overview of the study methodology we employed is given below and is fully described in chapters two through four of this dissertation. Overview of Drug Resistance Strategies of Gram Negative Bacteria There are three broad mechanisms of bacterial antimicrobial drug resistance. They include: (1) alteration or degradation of the antimicrobial drug; (2) modification, protection, or replacement of the drug target; (3) exclusion or efflux of the drug. In Gram negative bacteria the first and third mechanisms are particularly problematic. These mechanisms are described in greater detail below. Antibiotic Degrading or Altering Enzyme The alteration or degradation of antimicrobial drugs is a common mechanism of drug resistance. In this strategy of resistance an enzyme changes the structure or composition of the drug to render it inactive. Two examples of these enzymes are beta- lactamases and aminoglycoside acetyltransferases [6]. Beta- lactamases break the beta- lactam ring found in beta- lactam, cephalosporin, and carbapenems drugs. Once the ring is hydrolyzed these drugs no longer function. Aminoglycoside acetyltransferase add an acetyl group to an aminoglycoside. Once these acetyl groups are added at the appropriate position, the drug is no longer functional [6]. More than 1000 beta- lactamases have been described in Gram negative bacteria, which greatly challenges the development of new antimicrobial agents. Modification, Replacement, or Protection of the Drug Target The modification, replacement, or protection of the microbial drug target frequently involves the acquisition of a drug resistance gene from another organism. In 4

11 this strategy, the bacterial component that is targeted by the drug is replaced, altered, or protected, so that the drug is no longer able to effectively inhibit its action. Examples of these mechanisms include the recently discovered cfr gene, the acquisition of various penicillin binding proteins, Qnr proteins, and mutations in the genes that code for DNA gyrase. The cfr gene codes for a methyltransferase that methylates residues on the ribosome. The methylation of these residues prevents the binding of many different classes of antimicrobial drugs that interact with the ribosome [18, 19]. The expression of penicillin binding proteins that have been acquired through horizontal gene transfer from another species is a common and problematic mechanism of resistance [6, 20]. A penicillin binding protein is expressed that has been acquired from another species of bacteria and has a drastically lower affinity for the relevant beta- lactam drugs. The presence of a penicillin binding protein that is not inhibited by the antimicrobial drug is able to complement the host penicillin binding proteins that are inhibited by the drug. This renders the bacteria resistant to a beta- lactam or cephalosporin drug to which it was formally susceptible [20]. The hypothesized mechanism of action of the Qnr proteins is an example of protection of the bacterial components from the effects of the antibiotic drugs. Although a not fully understood mechanism, it is thought that Qnr proteins bind to DNA- gyrase, a target of quinolone drugs. This binding prevents the quinolones from intercalating into the DNA of the bacteria causing cell death and damage. It is not thought that Qnr proteins alter either the drug or the DNA gyrase to which it binds. It is also not believed that Qnr proteins are enzymes [21, 22]. It is common to find mutations in the genes that code for DNA gyrase in bacteria that are resistant to fluoroquinolones. These mutations prevent the drug from binding to its target, DNA gyrase, and render the bacteria resistant [6]. Exclusion or Active Efflux of the Antibiotic A common mechanism of resistance in Gram negative bacteria is the exclusion of drugs though alterations in the permeability of the outer membrane and the active efflux of antimicrobial drugs. The loss of porins, which decreases the permeability of the cells outer membrane, in combination with carbapenase enzymes, is thought to cause resistance to carbapenem drugs [23]. 5

12 The active efflux of drugs by the acquisition of efflux pump from another species of bacteria is a common mechanism of resistance. One example of this is tetracycline efflux pumps. In this case an efflux pump that has affinity for tetracyclines is acquired though horizontal gene transfer so the drug is actively removed from the cell. This makes the concentrations of the drug in the cell so low that the drug becomes ineffective [6]. Dissertation Chapters and Overview Chapter One Chapter One describes the rationale and provides an overview of the dissertation. It also contains a brief background on the mechanisms of drug resistance found in Gram negative bacteria. Chapter Two In chapter two we identified novel drug resistance genes from the microbiota of a retail produce item. To do this we used two functional metagenomic libraries that were constructed with DNA isolated from the microbiota of retail spinach. The libraries were grown in E. coli and screened for the ability to confer drug resistance to E. coli. From these libraries we isolated and characterized five novel drug resistance genes. Chapter Three In chapter three we further characterized the novel drug resistance sequences we identified in chapter two. To do this we preformed an extensive bioinformatics analysis. Chapter Four In chapter four we investigated the spread of drug resistance bacteria. To do this we studied clinical isolates of E. coli that caused extraintestinal infections in a large urban community in Brazil. We compared the strain type, drug resistance genes, and drug resistance phenotype of E. coli that caused two different kinds of extraintestinal infection in the same city between 1996 and The two types of infection we compared were meningitis and urinary tract infection. This study and its results are described in chapter four. 6

13 Chapter Five In chapter five we discussed the conclusions and future directions of the research presented in chapters two through four. Human Subjects Statement My dissertation research does not involve human subjects research. The research done as part of this dissertation meets the specific requirements for research with coded information and biological specimens that are not human subjects research. This is because the following two conditions have both been met. (1) The private information or specimens were not collected specifically for the currently proposed project through an interaction or intervention with living individuals; and (2) the investigator(s) cannot readily ascertain the identity of the individual(s) to whom the coded private information or specimens pertain as a result of the following circumstance; the investigators and the holder of the key have entered into an agreement prohibiting the release of the key to the investigators under any circumstances, until the individuals are deceased (NB: DHHS regulations for humans subjects research do not require the IRB to review and approve this agreement); there is IRB- approved written policies and operating procedures for a repository or data management center that prohibit the release of the key to the investigator under any circumstances, until the individuals are deceased; or there are other legal requirements prohibiting the release of the key to the investigators, until the individuals are deceased. Therefore, by definition the work does not include human subjects data. 7

14 Chapter Two Identification of Novel Antimicrobial Resistance Genes from Microbiota on Retail Spinach Hillary F. Berman and Lee W. Riley BMC Microbiology, Accepted November

15 Abstract Background: Drug resistance genes and their mobile genetic elements are frequently identified from environmental saprophytic organisms. It is widely accepted that the use of antibiotics in animal husbandry selects for drug resistant microorganisms, which are then spread from the farm environment to humans through the consumption of contaminated food products. We wished to identify novel drug resistance genes from microbial communities on retail food products. Here, we chose to study the microbial communities on retail spinach because it is commonly eaten raw and has previously been associated with outbreaks of bacterial infections. Results: We created metagenomic plasmid libraries from microbiota isolated from retail spinach samples. We identified five unique plasmids that increased resistance to antimicrobial drugs in the E. coli host. These plasmids were identified in E. coli that grew on plates that contained ampicillin (pamp), aztreonam (pazt), ciprofloxacin (pcip), trimethoprim (ptrm), and trimethoprim- sulfamethoxazole (psxt). We identified open reading frames with similarity to known classes of drug resistance genes in the DNA inserts of all 5 plasmids. These drug resistance genes conferred resistance to fluoroquinolones, cephalosporins, and trimethoprim, which are classes of antimicrobial drugs frequently used to treat human Gram negative bacterial infections. These results show that novel drug resistance genes are found in microbiota on retail produce items. Conclusions: Here we show that microbiota of retail spinach contains DNA sequences previously unidentified as conferring antibiotic resistance. Many of these novel sequences show similarity to genes found in species of bacteria, which have previously been identified as commensal or saprophytic bacteria found on plants. We showed that these resistance genes are capable of conferring clinically relevant levels of resistance to antimicrobial agents. Food saprophytes may serve as an important reservoir for new drug- resistance determinants in human pathogens. Key Words: Antibiotic Resistance, Gram Negative Bacteria, Metagenomic Library Background The spread of antimicrobial resistance genes has made previously manageable bacterial infections increasingly more difficult to treat. In addition, there has been a gradual decline in the development of new antimicrobial drugs, especially against Gram negative bacterial pathogens. The identification of genes in Gram negative bacteria that confer resistance to cephalosporins, carbapenems, and fluoroquinolones has created fears that we are returning to the pre- antibiotic era [2]. These multidrug- resistant infections often occur in hospitals and are frequently caused by species belonging to the normal microbiota of the human host [2, 24]. This suggests that the microbiota of the patients themselves is the reservoir for many of the organisms that cause hospital 9

16 acquired infections. Furthermore, recent work has demonstrated that the intestinal microbiota of humans and food animals are a reservoir of drug resistance genes [25]. Consequently, a better understanding of how drug resistance genes enter the human microbiota is imperative to better prevent drug resistant infections. Drug resistance genes and their mobile genetic elements are frequently identified from environmental saprophytic organisms. These include samples taken from soil, water, and wild animals [9, 26, 27]. Additionally, these genes have been identified in environmental samples from ancient and pristine environments samples that have never been exposed to human activity [26-28]. Due to the great diversity of antibiotic resistance genes found in environment, it has been hypothesized that environmental microbes serve as a reservoir of drug resistance genes and that a few then enter human pathogens [26, 27]. These drug resistance genes are spread between bacteria via mobile genetic elements, such as plasmids, transposons, and integrons [5, 9, 26, 27]. The detection of mobile genetic elements and drug resistance genes in the environment has led to numerous studies and policies to address the effects of environmental exposure to antimicrobial agents on human pathogens [29, 30]. It is widely accepted that the use of antibiotics in animal husbandry selects for drug resistant microorganisms, which are then spread from the farm environment to humans through the consumption of contaminated food products [31]. Numerous studies of bacterial pathogens in food products, such as Campylobacter and Salmonella, have demonstrated that the use of antimicrobial drugs in agriculture can result in drug resistant infections in humans [16, 29-32]. However, the majority of studies have been limited to species of zoonotic pathogens that cause foodborne disease and these studies are frequently done as part of national surveillance programs for food safety. Species of bacteria that are not usually considered foodborne pathogens, but nonetheless are found in both the human and food product microbiota, are usually not included in studies of drug resistant bacteria in retail food products. Studies of microbiota of animals demonstrated that commensal organisms are a reservoir of antimicrobial drug resistance genes [25]. These studies include the identification of antimicrobial resistance genes from animal feces including chickens and cows [25, 33-35]. Animal manure is frequently used as fertilizer in agriculture and may contribute to the spread of drug resistance genes. The spread of drug resistance genes by commensal bacteria on food products is an area that requires further study [31]. Produce items, which are frequently eaten raw, are one way consumers are exposed to microbiota on retail food products [36]. Previous work has shown that the normal microbiota of retail produce items harbors clinically relevant drug resistance genes [17]. However, previous studies have relied on PCR based methods to identify known drug resistance genes, which limits the number and types of drug resistance that could potentially be identified. Other studies used functional metagenomic libraries to identify novel antimicrobial resistance genes from environmental samples in a sequence independent manner [37, 38]. We wanted to apply this sequence independent 10

17 approach to investigate the presence of antimicrobial resistance genes on retail spinach. We chose to study the microbial communities on retail spinach because it is commonly eaten raw and has previously been associated with outbreaks of bacterial infections [39]. To do this, we made two metagenomic plasmid libraries with DNA isolated from the microbiota of retail spinach. One library was made from a cultured sample of spinach microbiota while the other was made in a culture independent manner. We then screened these libraries for their ability to confer resistance to antibiotics to an E. coli host. Results and Discussion Isolation of antibiotic resistant clones The first plasmid library, which was constructed from a cultured sample, contained 160 Mb of inserted DNA. The second library, which was constructed from an uncultured sample, contained 140 Mb of inserted DNA. We first constructed a cultured library because we wished to enrich for microbial DNA to increase the chances of cloning DNA sequences that contained drug- resistance genes. We constructed a library from an uncultured sample because we wanted to identify potential drug- resistance genes from bacterial organisms that cannot be cultivated in artificial medium. From the cultivated library, we isolated four different antimicrobial resistance- conferring clones. From the uncultivated library, we isolated one additional antimicrobial resistance- conferring clone. The mean size of the plasmid DNA inserts in both libraries was two Kb. We identified five unique plasmids that conferred increased drug resistance (minimum inhibitory concentration or MIC) to the host E. coli. Each plasmid was named after the antimicrobial agent to which it conferred resistance (Table 1). These plasmids were identified in E. coli that grew on plates that contained ampicillin (pamp), aztreonam (pazt), ciprofloxacin (pcip), trimethoprim (ptrm), and trimethoprim- sulfamethoxazole (psxt). The plasmids pamp, pazt, pcip, and ptrm were isolated from the library made from a cultured sample. The plasmid psxt was isolated from the library made from an uncultured sample. pamp increased the MIC of ampicillin 4 fold (4ug/ml to 16 ug/ml). pazt increased the MIC of the host strain 96 fold to aztreonam (.125 ug/ml to 12 ug/ml), 10 fold to cefepime (.096 ug/ml to 1ug/ml), and 6 fold to piperacillin ( 2ug/ml to 12 ug/ml). Additionally, pazt encoded an ESBL phenotype as measured by the ceftazidime, ceftazidime/clavulanic acid ESBL Etest, (TZ 16 ug/ml and TZL 1 ug/ml). pcip caused a 62- fold increase in resistance to ciprofloxacin (<.002 ug/ml to.125ug/ml) as well as a 31- fold increase in resistance to levofloxacin (.012 ug/ml to.38 ug/ml). The MIC of trimethoprim for E. coli carrying ptrm increased >258 fold (.124 ug/ml to >32ug/ml) 11

18 and 7 fold to trimethoprim- sulfamethoxazole (.064 ug/ml to.5 ug/ml). psxt caused a 15- fold increase in resistance to trimethoprim- sulfamethoxazole (.064 ug/ml to 1 ug/ml) as well as an >256- fold increase in resistance to trimethoprim alone (.125ug/ml to >32 ug/ml). Identification of antibiotic resistance genes and phylogenetic analysis We identified open reading frames with similarity to known classes of drug resistance genes in DNA inserts of all 5 plasmids. They are summarized in Figure 1 and Table 2. In pazt we identified a sequence with 94% identity at the nucleotide level to a gene that encodes penicillin- binding protein 1A identified in Bacillus subtilis subsp. Spizizenii [GenBank, gb CP ]. The expression of altered penicillin binding proteins are known to confer resistance to beta- lactam and cephalosporin antibiotics in various clinically important pathogens [40]. However, the ability of this sequence to confer clinically relevant levels of cephalosporin resistance or an ESBL phenotype has not been previously reported. In pamp we identified a beta- lactamase gene with 71 % identity to the ERP- 1 gene that encodes a class A extended spectrum beta- lactamase found in Erwinia persicin [GenBank, gb AY ] [41]. When transformed into an E. coli host, ERP- 1 was reported to increase resistance to penicillins and cephalosporins, including piperacillin, cefotaxime, and ceftazidime [41]. Surprisingly, we found that pamp did not increase the MIC of the host E. coli to piperacillin, ceftazidime, or cefotaxime. Also, pamp did not increase the MIC of this host strain to any tested cephalosporin, monobactam, or carbapenem (cefepime, aztreonam, or imipenem). However, this isolate tested positive for the presence of a beta- lactamase by the nitrocefin assay. These results suggest that the novel sequence we identified in pamp is distinct from ERP- 1 in terms of the spectrum of drug- resistance phenotype it encodes. However, it is also possible that these observations are artifacts due to poor expression of the gene in a heterologous host. Surprisingly, when the pcip DNA sequence was submitted to BlastN, only two other sequences in the NCBI non- redundant nucleotide database were identified. The sequences were part of whole genome sequences of Exiguobacterium antarcticum and Exiguobacterium sibiricum. These species were identified in a frozen Antarctic lake and a core sample of the Siberian tundra [42, 43]. When submitted to BlastP, a fluoroquinolone resistance protein from Oceanobacillus sp. Ndiop was identified. This quinolone resistance protein was a predicted pentapeptide repeat protein (PRP) [GenBank, ref ZP_ ]. One known class of plasmid- mediated quinolone resistance conferring sequences is called QNR[44]. QNRs are pentapeptide repeat proteins and have been associated with extended spectrum beta- lactamases [44]. In addition to the PRP, the pcip DNA insert contained a second open reading frame that 12

19 showed similarity to a beta- lactamase domain containing protein (Figure 2). However, pcip did not increase the MIC of the host E. coli to any of the tested beta- lactam antibiotics (ampicillin and piperacillin), cephalosporin antibiotics (ceftazidime, cefotaxime, cefepime), or carbapenem antibiotics (imipenem). Additionally, pcip did not test positive for beta- lactamase production by the nitrocefin assay. This indicated that the predicted beta- lactamase is either not expressed or does not function as a beta- lactamase. We subcloned the PRP sequence in pcip (pprp:1b) in order to confirm that the predicted PRP was able to confer resistance to ciprofloxacin. As compared to the empty vector, the pprp:1b increased the MIC of ciprofloxacin of the host E. coli 7 fold (.016 to.125) and the MIC of levofloxacin 5 fold (.047 to.25 ug/ml). The MIC conferred by pprp:1b to the host E. coli is consistent with previously reported MICs from other QNR sequences found in human pathogens [44]. We found that the ptrm DNA insert has 84% identity to a region of the Erwinia pyrifoliae DSM complete genome [GenBank, emb FN ]. This region contains the fola gene, which encodes a dihydrofolate reductase (DHFR). DHFR is the target of trimethoprim [6]. The expression of a DHFR that is not susceptible to trimethoprim is a well- known mechanism of resistance [6]. However, the acquisition of the fola gene from Erwinia pyrifoliae DSM has not previously been shown to confer resistance to trimethoprim. Similarly, the DNA insert from psxt has 88% identity to a region of the Pseudomonas fluorescens SBW25 complete genome [45]. This region also encodes a predicted dihydrofolate reductase. This sequence from Pseudomonas fluorescens SBW25 has not previously been shown to confer resistance to trimethoprim or trimethoprim sulfamethoxazole. Conclusions Here we show that microbiota of retail spinach contains previously unidentified antibiotic resistance- conferring genes and that functional metagenomic libraries can be used to screen retail food products for drug resistance genes in a sequence independent manner. Furthermore, due to the limited amount of DNA that can be cloned into a plasmid library and the requirement that the drug resistance gene be expressed in a heterologous host, it is likely we only identified a fraction of drug resistance conferring genes present in our spinach samples. Although none was identical in DNA sequence, many of these novel sequences show sequence similarity to genes found in species of bacteria that have previously been identified as commensal or saprophytic bacteria found on plants [41, 45]. This suggests the sequences we identified are not the result of contamination from animals or humans. We showed that these resistance genes are capable of conferring clinically 13

20 relevant levels of resistance to commonly used classes of antimicrobial agents, including cephalosporins and fluoroquinolones. The novel antimicrobial resistance genes we identified include beta- lactamases, a pentapeptide repeat protein, a penicillin binding protein, and putative dihydrofolate reductase genes. These types of resistance mechanisms are some of the most common and clinically problematic mechanisms of drug resistance found in pathogens [6]. We do not know at this time if these genes will become clinically important, and one limitation of this study is that we did not analyze these genes for their potential for horizontal transfer to human pathogens. Recent functional genomics analysis of environmental soil samples has not only identified drug- resistance genes with identical nucleotide sequences from human pathogens, but also mobile gene sequences providing evidence for possible horizontal gene transfers [46]. Further studies using sequence independent methods to identify antimicrobial resistance genes from retail food products should be done to better understand the role of saprophytes as a reservoir for new drug- resistance genes. Experimental Procedures Metagenome plasmid library construction and screening Two metagenomic plasmid libraries of spinach microbiota were constructed. One was based on DNA extracted from cultured bacteria and the other was based on DNA extracted from uncultured spinach wash. The metagenomic DNA used to create the first library was obtained by washing twenty five grams of bagged baby spinach in PBS. A description of the spinach used to create the library has been previously published [17]. Briefly, the spinach was purchased from a supermarket located in Berkeley, California in They included organic as well as non- organic spinach. One milliliter of the PBS wash was then used to inoculate 50 ml TSB. This culture was grown at 37 C with shaking overnight. The culture was then centrifuged at 10,000 x g for 10 minutes. DNA was extracted from the resulting pellet by the phenol chloroform method. The metagenomic DNA used to create the second library was obtained by washing six bags of baby spinach in two liters of PBS. Six different brands of spinach were purchased from three retailers located in Berkeley, California in The spinach was incubated in PBS at room temperature for two hours. The resulting wash was then filtered through sterilized cheesecloth and a sterilized coffee filter to remove spinach debris. The filtered wash was then centrifuged at 10,000 x g for 20 minutes. DNA was extracted directly from the resulting pellet with the Gnome DNA isolation Kit, MP Biomedical. Two plasmid libraries with metagenomic DNA inserts were constructed in psmart- LC kan vector in the E. coli host, E. Cloni (Lucigen corp., Middleton WI). The psmart vector confers resistance to kanamycin and has transcriptional terminators 14

21 flanking the cloning sites. Consequently, transcription of the cloned sequences requires a native promoter. E. coli clones containing the two plasmid libraries were then screened for resistance to antimicrobial agents on Mueller Hinton agar plates containing one of the following 16 antimicrobial agents: ampicillin, carbenicillin, ticarcillin, amoxicillin/clavulanic acid, ticarcillin/clavulanic acid, cefotaxime, ceftazidime, aztreonam, meropenem, gentamicin, nalidixic acid, ciprofloxacin, trimethoprim, trimethoprim- sulfamethoxazole, chloramphenicol, or tetracycline. The phenotype of resistance to antimicrobial agents was confirmed by retransforming the recombinant plasmid into E. Cloni. Growth on Muller Hinton agar containing kanamycin was used as a positive control for transformation. The acquisition of drug resistance from the transformation of the plasmid was demonstrated by growth on Mueller Hinton agar containing the corresponding antimicrobial agent. We used the empty vector, psmart, in the E. coli host as a negative control for antibiotic stability. Antimicrobial susceptibility testing The MIC of each E. coli clone was determined by Etest (Biomerieux, France) according to manufactures recommendations. All Etests were repeated in at least two independent experiments. ATCC 29522, ATCC , and ATCC were used for control as recommended by the manufacturer. In accordance with manufactures recommendations, less than a fourfold difference in MIC was considered to be with in the expected margin of error for this test. Nitrocefin test A colony was spotted onto a sterile Petri dish and then covered with nitrocefin, as previously described [47]. E. Cloni containing the empty vector (psmart) was used as a negative control. Sequencing and data analysis The sequence of the DNA insert in the resistance conferring plasmid was determined by primer walking at the University of California, Berkeley sequencing facility. The sequences were assembled with Geneious Version 5.6, (Biomatters, New Zealand). The insert sequences were then submitted to ORFinder and the BLAST suit of programs at NCBI [48]. The nucleotide sequences of the insert from each plasmid have been deposited in Genbank with the following accession numbers: pamp: KF791056, pazt: KF791057, pcip: KF791058, ptrm: KF791059, psxt: KF Cloning: Standard protocols for ligation independent cloning in to vector 1B, QB3 Macrolab, University of California, Berkeley were used. 15

22 Competing interests The authors declare no competing interests. Authors contributions HB carried out the laboratory and bioinformatics studies and drafted the manuscript. LWR conceived of the study, and participated in its design and coordination and helped to draft the manuscript. Both authors read and approved the final manuscript. Acknowledgements We would like to acknowledge the invaluable contributions of Eva Raphael and Olivera Marjanovic to the coordination of this study and the laboratory work required to create the first metagenomic library. This project was supported in part by NIAID/NIH grant (R01AI059523) and T32AI

23 Chapter Three Bioinformatics Analysis of Drug Resistance Conferring Amino Acid Sequences 17

24 Introduction As described in chapter two, we used a functional metagenomic library to isolate unknown DNA sequences that confer resistance to antibiotics from retail spinach. We preformed a bioinformatics analysis to identify these sequences and generate a hypothesis about their mechanism of action. A brief overview of our analysis pipeline is given here, followed by a brief description of each method. Finally, we discuss the results, conclusions, and future directions of the analysis of pcip- PRP and Amp- beta- lactamase. Here we will focus on the results of the protein sequence analysis since the results of the analysis of the nucleic acid sequences were presented in chapter two. Methods Bioinformatics pipeline We started by submitting the sequence to BlastN and BlastX against the NCBI databases [49]. From this first search we were able to determine the type of organism the DNA had come from and if the sequence was closely related to know drug resistance genes. We then submitted the sequence to ORFinder at NCBI. ORFinder predicts start and stop codons of open reading frames (ORFs) in all six frames [50]. ORFinder also predicts the protein sequence from each identified ORF. With this information, we submitted each possible nucleic acid and corresponding protein sequence to the appropriate Blast flavor (BlastN or BlastP). From this analysis we determined what the gene sequence was likely to code for. We then preformed a literature review to determine if the predicted protein could plausibly confer resistance to antimicrobial drugs. Once a plausible ORF was identified, we generated multiple sequence alignments with relevant known drug resistance genes or protein sequences and predicted the protein structure to generate a hypothesis about its mechanism of action. We also generated phylogenetic trees to determine the relationship of our novel gene or protein sequence to known drug resistance sequences. Bioinformatics methods used BLAST search of databases We used the BlastN, BlastP, and BlastX programs. We used these programs to search protein or nucleic acid sequence databases at NCBI. These programs use a matrix to generate a score of a comparison between two sequences. This creates a 18

25 pairwise alignment that is scored to generate the E- value [49, 51, 52]. The E- value represents the number of different alignments with the same or greater value than a set score by chance and is used to rank the hits. The final result of this is a list of ranked pairwise alignments. Multiple sequence alignment Multiple sequence alignments are made of three or more protein or nucleic acid sequences that are partially or completely aligned. Residues that are aligned are presumed to be homologous and to have a common ancestor. Multiple sequence alignments are used to build phylogenetic trees, build profiles to search databases, as well as to assign predicted functions and protein structures. It is assumed that homologous residues are likely to retain similar functions and structures. Profiles are based on the idea that the probability that an amino acid will be in a position can be estimated based on the other amino acid residues that are already present in the other sequences at that position in your alignment [51]. Here we used MUSCLE to make the protein alignments discussed below [53]. The alignments were preformed in Jalview or Geneious as indicated below [54, 55]. Phylogenetic trees Phylogenetic trees are a graph that represents the inferred evolutionary relationship of homologues sequences of nucleic acids, proteins, or whole organisms. There are numerous methods that can be used to make phylogenetic trees, but here we used the neighbor joining method in the program Jalview [51, 54, 56]. Structure prediction and critical residue identification Structure prediction and the identification of critical residues can be used to test and further develop hypotheses about the mechanism of action of a protein. Structures are generally predicted by searching a database to find a homologous protein that has been experimentally studied to elucidate its structure. Phyre2, the program used here, predicts the secondary structure of an amino acid sequence and then uses the secondary structure profile to search a database of known structures and their secondary structure profiles [57]. Here we used structure prediction and structural alignments to develop hypotheses about the mechanism of action and to determine if features experimentally shown to be important for the function of a protein are present in the novel variants we identified. Additionally, we used a predicted structure to map critical residues to further develop hypotheses about substrate specificity [51]. The structures were visualized and manipulated in Pymol and Geneious as indicated below [55, 58]. 19

26 pcip Results and Discussion Here we further characterized the novel pentapeptide repeat protein (PRP) identified from the pcip plasmid described in chapter two. In chapter two we focused our analysis on the pcip- PRP gene sequence. Here we focus on the protein sequence. To avoid confusion we will refer to the protein encoded by the pcip- PRP as Cip- Qnr. As described in chapter two, Qnr proteins are pentapeptide repeat proteins that are known to cause resistance to fluoroquinolone drugs [59]. To further investigate Cip- Qnr we compared our novel sequence to known sequences of Qnr proteins. We used MUSCLE to make a multiple sequence alignment with the amino acid sequence of Cip- Qnr and a representative sequence of each Qnr family (Figure 2) [54, 60]. There are six known families of Qnr proteins found in clinical isolates, which are QnrA, B, C, D, S, and V. A family is defined as having at least 70% identity between either nucleotides or derived amino acids [61]. Additionally, Qnr- like proteins that have not been found in clinical isolates, but have been studied, were also included in the alignment (EsfQnr, AhQnr, Mcbg, and MfpA). Based on this alignment, we constructed a phylogenetic tree in Jalview using a neighbor joining method (Figure 3) [54]. Cip- Qnr clustered with EsfQnr (amino acid sequence identity of 35%), but was not closely related to any of the known classes of Qnrs commonly found in clinical isolates. This is not surprising given the low sequence identity between these two proteins, as compared to the rest of the Qnr proteins. Therefore, this clustering my be an example of long branch attraction, when the sequences that are the most different cluster together, not because they are similar to each other, but because they are not related to anything else in the tree. We then used Phyre2 to generate a predicted structure of Cip- Qnr (Figure 4) [57]. The mechanism by which Qnrs and other Qnr- like proteins confer resistance to quinolones is not well understood. Therefore we were unable to identify critical residues in these Cip- Qnr. However, based on the Phryre2 model, Cip- Qnr contains a loop region which MfpA, the protein structure that the model is based on, does not contain [62, 63]. Studies of AhQnr have shown that deleting a similar loop region renders it inactive so it is no longer able increase resistance to quinolones [64]. Conclusions and Future Directions From the above analysis I have concluded that the Cip- Qnr we identified is truly novel and not closely related to known classes of Qnr proteins. One future direction is to further analyze a pairwise alignment and structural alignment between Cip- Qnr and AhQnr to determine if the predicted loop regions are either homologous or could be 20

27 functionally equivalent. We could then test the importance of the loop region to the Cip- Qnr function experimentally by site directed mutagenesis. If I were to do these analyses again, I would focus on identifying conserved residues in the Cip- Qnr. The mechanism of action of the Qnr proteins is unknown, but it is not thought to be an enzyme. Identifying conserved residues with a function that could be tested experimentally may lead to new hypothesis about the mechanism of action of the Qnr proteins. As it is thought that Qnrs directly interact with the DNA gyrase, I would also like to directly model the interaction between the DNA gyrase and Qnr proteins. pamp Results and Discussion As described in chapter two, in pamp we identified a novel beta- lactamase ( Amp- beta- lactamase) with 71% identity at the nucleotide level to ERP- 1, a class A extended spectrum beta- lactamase found in Erwinia persicin [65]. Using the Ambler classification system, beta- lactamases are classified into classes A- D based on amino acid sequence similarity [66]. To further investigate Amp- beta- lactamase, we generated an amino acid multiple sequence alignment with an extensively studied class A beta- lactamase, TEM- 1. We also predicted the structure of Amp- beta- lactamase (Figure 5 and 6). We made a multiple sequence alignment with ERP- 1, Amp- beta- lactamase, d1hoza (the protein used to generate the structural of model), and the TEM- 1 protein sequence (Figure 5). We chose to compare Amp- beta- lactamase to TEM- 1 rather than a more closely related beta- lactamase because TEM- 1 has been extensively studied as a model for group A beta- lactamases [67, 68]. Based on this previous work on TEM- 1 we were able to map residues that had been experimentally shown to alter the substrate specificity of the enzyme. We first mapped the residues identified as critical in the catalytic site atlas (CAS) [69]. The catalytic site atlas has a very restrictive definition of critical residues which does not include residues that are involved in ligand binding [69]. From this analysis we determined that the four CAS critical residues, Ser- 70, Lys- 73, Ser- 130, and Gly- 166, were conserved between all sequences we investigated. All reported residue numbering is based on the consensus Ambler alignment scheme [68]. We next identified residues that have been experimentally shown to alter the substrate specificity of the enzyme but did not meet the strict definition of a CAS critical residue. We did not identify any differences between ERP- 1 and the novel pamp beta- lactamase at the residues that we examined. However, we identified numerous 21

28 differences at residues known to alter the substrate specificity of the TEM- 1 enzyme in both ERP- 1 and the AMP- beta- lactamase. The substitutions we identified can be categorized into two broad groups. First, there were substitutions of the exact same amino acids at the residue that had previously been experimentally shown to alter the substrate specificity of the enzyme. Second, there were substitutions at positions that had previously been associated with changes in substrate specificity; however, the specific amino acid substitutions observed here had not been previously studied. First, we identified substitutions in ERP- 1 and the novel Amp- beta- lactamase at position 104, 182, and 237 that have previously been shown to alter the substrate specify of beta- lactamase enzymes. There was a Glu- 104 to Ser- 104 conversion. Previous work has found that this substitution changed the substrate specificity, probably because this residue is located at the entrance to the binding pocket [67]. There was an Met- 182 to Thr- 182 conversion that has previously been found in an enzyme that was resistant to clavulanic acid type inhibitors [67]. However, that enzyme also had a change at residue 69 that was later shown to mediate the clavulanic acid resistance [67]. ERP- 1 and the Amp- beta- lactamase also have a conversion at position 69, which is described below. Therefore, when combined, these two substitutions may contribute to the extended spectrum of ERP- 1. However, Amp- beta- lactamase does not show resistance to beta- lactamase inhibitor combinations like ERP- 1 does, even though it has the same substitutions. We observed an Ala- 237 to Ser- 237 substitution. This substitution has previously been reported as a natural variant found in a beta- lactamase from Proteus vulgaris. When Ser- 237 was replaced by Ala- 237 in this enzyme, the efficiency against oxyimino- cephalosporins was reduced [67]. Second, we identified substitutions of residues that are known to be important for substrate specificity, but the effects of the exact substitutions described here are unknown. We identified this type of substitution at 69, 240, and 244. There was a Met- 69 to Cys- 69 conversion. There are other examples of enzymes with Met- 69 to Cys- 69 conversion, but the phenotypic result of this conversion is unknown [67, 68]. Previous work has found that an Met- 69 to either a Leu- 69 or Iso- 69 conversion at this position changes the substrate specificity of the enzyme [67]. This position is particularly important because it is located next to the highly conserved and critical Ser- 70 that makes up part of the back wall of the binding pocket [67]. There was a Glu- 240 to Gly- 240 conversion at position 240 but the importance of this conversion is unknown. There was an Arg- 244 to Thr- 244 conversion. Mutagenesis studies in the presence of clavulanic acid and naturally resistant variants have previously identified Ser- 244 or Cys- 244 conversions [67]. 22

29 Conclusions and Future Directions We identified many substitutions in Amp- beta- lactamase as compared to TEM- 1. But we did not identify any substitutions between ERP- 1 and Amp- beta- lactamase at the residues we examined. This suggests that the differences in resistance phenotype conferred by ERP- 1 and the Amp- beta- lactamase reported in chapter 2 may be due to the poor expression of the Amp- beta- lactamase. Alternatively, it could be due to substitutions at residues that were not examined as part of this analysis. To further test these hypotheses one future direction would be to experimentally characterize the enzyme kinetics of the novel AMP- beta- lactamase. If I were to do the analysis of the AMP- beta- lactamase again, I would focus on identifying the differences between CTX- M enzymes, ERP- 1, and Amp- beta- lactamase. There are known differences between the CTX- M type enzymes and TEM- 1 that we did not address here [66, 67, 70, 71]. These differences may be important because our preliminary analysis of Amp- beta- lactamase, which is not presented here, and previous work with ERP- 1, suggests that they are more closely related to CTX- M type enzymes than TEM- 1[65]. 23

30 Chapter Four Distribution of strain type and antimicrobial susceptibility of Escherichia coli causing meningitis in a large urban setting in Brazil Hillary Berman 1, Maria Goreth Barberino 2, Edson Duarte Moreira Jr. 2, Lee Riley 1, Joice N. Reis 2,3 1. Division of Infectious Disease and Vaccinology, School of Public Health, University of California, Berkeley, CA Gonçalo Moniz Research Center, Oswaldo Cruz Foundation, Salvador, Brazil 3. Federal University of Bahia, Salvador, Brazil 24

31 Abstract The clinical management of meningitis caused by Escherichia coli is greatly complicated when the organism becomes resistant to broad- spectrum antibiotics. We sought to characterize the antimicrobial susceptibility, multilocus sequence type (MLST), and presence of known drug resistance genes of E. coli that caused meningitis between 1996 and 2011 in Salvador, Brazil. We then compared these findings to E. coli isolates from community acquired urinary tract infections (UTI) that occurred during the same time period and in the same city. We found 19% of E. coli from cases of meningitis and less than 1% of isolates from UTI to be resistant to third- generation cephalosporins. The sequence types of E. coli from cases of meningitis included ST 131, ST 69, ST 405, and ST 62, which were also found among isolates from UTI. Additionally, among the E. coli isolates that were resistant to third- generation cephalosporins, we found genes that encode the extended- spectrum beta- lactamases CTX- M- 2, CTX- M- 14, and CTX- M- 15. These observations demonstrate that compared to E. coli isolated from cases of community acquired UTI, those isolated from cases of meningitis are more resistant to third- generation cephalosporins even though the same sequence types are shared between the two forms of extraintestinal infections. Introduction The prevalence of antimicrobial drug resistance in Gram negative bacteria continues to increase worldwide, which complicates the clinical management of infections caused by these organisms [2]. These infections include enteric as well as extraintestinal infections such as urinary tract infections (UTI), sepsis, and meningitis [72, 73]. Escherichia coli is a Gram negative bacillus that is capable of causing urinary tract infections (UTI) and meningitis, but it is also a normal constituent of the gut microbiota of mammals [72, 74]. It is thought that the E. coli that cause extraintestinal infections first colonize the gut [72]. From the gut E. coli are then able to spread to the urinary tract and blood stream [72]. Animal models of E. coli meningitis have shown that once E. coli reaches a high enough concentration in the blood stream it is able to cross the blood brain barrier and cause meningitis [75, 76]. The severity of meningitis has been correlated with the concentration of bacteria in the bloodstream [72]. E. coli is the leading cause of UTI, and second leading cause of neonatal meningitis in developed countries [72, 77]. Although it is not as common, E. coli also causes meningitis in adults [77, 78]. Isolates of E. coli that have caused neonatal meningitis in developed countries have been extensively studied [75, 79-85]. These studies have found that some serotypes and sequence types predominate among E. coli that cause meningitis. However, these same serotypes and sequence types are found in E. coli isolated from other sources such as urinary tract infections, poultry, and from healthy human 25

32 digestive tracts [80, 81, 86]. Serotypes of E. coli that have been isolated from cases of meningitis include O18:K1, O45:K1, O1:K1, and O83:K1 [82, 85, 87, 88]. The multilocus sequence types (MLST) isolated from cases of meningitis include ST 95, ST 62, ST 131, and ST 69 [82, 85, 86, 88-90]. When combinations of typing methods and more discriminant typing methods are used, isolates that have caused meningitis can be separated from isolates from other sources [80, 81]. However, most studies examined a limited number of collections of E. coli available for study obtained from different geographic sites and time periods and hence, it is unclear how significant these differences are. There are recent reports of meningitis caused by E. coli that are resistant to third- generation cephalosporins and produce extended spectrum beta- lactamases (ESBL) [89, 91]. The spread of E. coli that are resistant to broader- spectrum cephalosporin antibiotics is problematic because these antibiotics are recommended for the treatment of meningitis caused by E. coli [76, 77]. We sought to characterize E. coli isolated from UTI and meningitis patients by both MLST and the presence of ESBL genes in order to better understand the spread and distribution of drug resistance in E. coli that cause extraintestinal infections. Methods Isolate collection The isolates characterized in this study were originally collected as part of three separate studies of meningitis and community acquired UTI that occurred in Salvador, Brazil. Due to sample loss, only a subset of the UTI isolates were characterized. All E. coli isolates from the meningitis collection were included in the current study. As part of active hospital based surveillance, E. coli isolates from cases of meningitis that occurred between 1996 and 2011 in Salvador were consecutively collected. The methodology and results of this study were previously published [92]. E. coli from community acquired UTI were collected as part of two separate studies, one that occurred between 2001 and 2002, and another that occurred between 2008 and The results and methodology of the study of E. coli from UTI that occurred between 2001 and 2002 has been published elsewhere [93]. The study of E. coli collected from UTI from 2008 to 2009 is unpublished (Personal communication, Barberino et al.). The antimicrobial susceptibility of the UTI E. coli isolates, reported here, was determined as part of this previous study. The 2008 to 2009 UTI study by Barberino et al. occurred at a private outpatient emergency room in Salvador, Brazil, and was conducted in accordance with all relevant human subjects protections (IRB 180/2008). 26

33 As previously published, the isolates from the and the UTI collections were classified as susceptible or non- susceptible to each antimicrobial agent tested. Non- susceptible isolates include isolates that showed either intermediate or resistant phenotypes. Antimicrobial susceptibility testing The susceptibility of the E. coli isolates from cases of meningitis was determined with the MicroScan BP38 panels manufactured by Siemens. ATCC and ATCC strains were used as reference for all tests. According to manufactures published break points, isolates were classified as susceptible, intermediate, and resistant to each drug. Where indicated, isolates were classified as either susceptible, or non- susceptible. The non- susceptible isolates contained all those with either the intermediate or resistant phenotype. DNA extraction, beta- lactamase detection and sequencing DNA was extracted with the Promega Wizard Genomic DNA kit according manufactures recommendations. Meningitis isolates were screened for the presence of bla TEM, bla SHV, bla OXA, and bla CTX- M gene sequences with multiplex PCR protocols previously published by Dallenne et al. [94]. UTI isolates were only screened for bla TEM, bla SHV, and bla OXA. The primer sets used were: Multiplex I TEM, SHV and OXA- 1- like, and Multiplex II CTX- M group 1, group 2 and group 9 [94]. All reactions and gel electrophoresis were performed according to previously published protocols [94]. For each reaction, DNA known to contain the relevant gene sequence was used as a positive control, and water was used as negative control. The PCR products that tested positive for a beta- lactamase were submitted for sequencing at the University of California, Berkeley, DNA Sequencing Facility. The resulting sequences were then submitted to BlastN for identification [49]. Multi- locus sequence typing (MLST) and sequence analysis MLST was preformed according to previously published protocols [86]. The sequences were submitted to the online MLST database at for sequence type identification. All sequencing was done at the University of California, Berkeley, DNA sequencing Facility. The resulting sequences were cleaned, assembled, and analyzed with Geneious

34 Data analysis All data analysis and graphs were generated with STATA 10. P- values were calculated with Fisher s exact test. Results Characterization of E. coli causing meningitis Thirty- six E. coli isolates from cases of meningitis were collected from patients in Salvador, Brazil, from 1996 to Figure 7 displays the year of isolation and sequence type of each of these isolates. The age of the patient who had meningitis was known for 21 isolates. Thirteen (62%) of these isolates were from children less than two years of age. Three novel sequences types 3701, 3702, and 3703 were identified among these isolates. All of the isolates with novel sequence types were isolated from cases of meningitis in children who were less than two years old. Among the meningitis isolates, 19% were resistant to ceftazidime and cefotaxime, 17% were resistant to aztreonam, and 14% were resistant to ciprofloxacin (Table 3). All isolates that were resistant to third- generation cephalosporins were found to have caused meningitis after 2002 (Figure 7). We did not find resistance to carbapenems, cefotetan, or tigecycline. Eighty- six percent of isolates carried bla TEM, 13% carried bla SHV, and none carried bla OXA. Thirty- six percent of isolates tested positive for bla CTX- M. All isolates that were resistant to third- generation cephalosporins carried a bla CTX- M type gene. Five percent of isolates tested positive for bla CTX- M group one, 3% for bla CTX- M group nine, and 28% for bla CTX- M group two. Of the 13 isolates that tested positive for CTX- M, 10 yielded interpretable sequences. bla CTX- M - 2 was identified in isolates ST652, ST62, ST405, ST127, ST167, ST359, and ST362. bla CTX- M- 15 was found in one ST998 isolate. bla CTX- M- 14 was found in one ST127 isolate. Comparison of E. coli that caused UTI and meningitis The E. coli isolated from cases of meningitis had significantly (P<0.001) higher frequency of resistance to third- generation cephalosporins than E. coli that were isolated from UTI. The study of 544 cases of UTI in Salvador that took place in found less than 1% to be non- susceptible (intermediate or resistant phenotypes) to third- generation cephalosporins [93]. The study of 411 cases of community acquired UTI that took place between found all isolates to be susceptible to third- generation cephalosporins. The current study of 36 E. coli from cases of meningitis found that 22% were non- susceptible to third- generation cephalosporins and that 19% were resistant to third- generation cephalosporins. 28

35 We compared E. coli that caused both UTI and meningitis by comparing the MLST of isolates with identical drug resistant phenotypes. The isolates were classified as susceptible or non- susceptible to eight classes of antimicrobial drugs because different antimicrobial susceptibility testing methods, drugs, and breaks points were used for each collection. Only classes of drugs that all isolates were tested with were included in this analysis. These classes were tetracyclines (TET), aminoglycosides (AMI), trimethoprim- sulfamethoxazole (TSX), penicillins (PEN), first generation cephalosporins (CEPH1), third generation cephalosporins (CEPH3), fluoroquinolones (FLR), and nitrofurantoin (NIT). Each drug resistance phenotype was named after the classes of drugs for which the isolates were non- susceptible. The drug resistance phenotypes found in both the UTI and meningitis collections are: TET+TSX+PEN+CEPH1, TSX+PEN+CEPH1, TET+TSX+PEN, PEN+CEPH1, TET+TSX+CEPH1, TET+TSX, TET+TSX+FLR, and TET+TSX+FLR+CEPH1 (Table 4). Available UTI isolates with these phenotypes were submitted for MLST and screened for the bla TEM, bla SHV, and bla OXA gene sequences. From these isolates ten sequence types were identified in both the meningitis and UTI collections. These sequence types were ST10, ST62, ST69, ST93, ST127, ST131, ST167, ST405, ST998, and ST2609. Of these common sequence types, three types had the same drug resistance phenotypes in both the meningitis and UTI collections. They included ST10, ST62, and ST69. Of these three sequence types, only ST69 and ST62 showed the same beta- lactamase genotype in the UTI and meningitis collection. Isolates with the phenotype TET+SXT+PEN+CEPH1, that were ST69, and that carried the bla TEM gene, were isolated from a case of meningitis in 2008 and from six cases of UTI from 2008 to Isolates with the phenotype TET+ SXT+ PEN, that are ST69, and that carry bla TEM, were isolated from a case of meningitis in 2001, six cases of UTI that occurred between 2001 and 2002, and one case of UTI that occurred between 2008 and Isolates with the phenotype SXT+ PEN+ CEPH1, that were ST62, and that carried bla TEM gene, were isolated from two cases of meningitis in 2009 and ST62 with bla TEM and the same drug resistance phenotype was also isolated from one UTI that occurred between 2001 and 2002, and a UTI that occurred between 2008 and Discussion To better understand the spread of drug resistant E. coli, we compared MLST and presence of bla TEM and bla SHV beta- lactamase genes from E. coli isolated from UTI and meningitis cases. We strain typed the E. coli by an MLST protocol which is commonly used to characterize drug resistant E. coli isolates [86, 95]. We identified a number of internationally distributed sequence types that are frequently drug resistant. These sequence types have also previously been isolated from food products. These sequence types include the ST10 complex, ST69, ST117, ST131, and ST405 [95]. ST69, previously 29

36 labeled CgA, was first reported from UTI patients in Berkeley, USA, and has been reported from UTI patients in Rio de Janeiro also Brazil [96, 97]. ST62 has previously been associated with K1 E. coli, a serotype that frequently causes neonatal meningitis [86]. Our findings further highlight the worldwide distribution of these lineages, and that they are capable of causing meningitis. The results of antimicrobial susceptibility testing of the E. coli that caused meningitis show that 19% were resistant to third generation cephalosporins. These drugs are currently recommended for the treatment of meningitis suspected to be caused by Gram negative organisms [77]. In this study we also identified extended spectrum beta- lactamase genes in all meningitis isolates that were resistant to third generation cephalosporins. In addition to the presence of bla TEM and bla SHV genes, this study identified bla CTX- M- 2, bla CTX- M- 15, and bla CTX- M- 14 genes among E. coli strains that caused meningitis. Cases of meningitis caused by ESBL producing E. coli have been also been reported from Turkey, France, Algeria, Thailand, Germany, and Brazil [89, 98]. The E. coli that caused meningitis were more resistant to third generation cephalosporins than those that caused community acquired UTI. While the numbers of E. coli that caused meningitis are small (N=36), this difference was still statistically significant. Unfortunately, the UTI isolates that were non- susceptible to third- generation cephalosporins were lost, and we were unable to investigate these isolates further. The results of this study support the findings of previous studies comparing E. coli from cases of meningitis and E. coli isolated from other sources [79, 80, 87]. These studies showed that the same sequence types are frequently found in E. coli that are isolated from meningitis and other sources and once further typing schemes are applied differences can be found. In our study, these overlapping sequence types included the most common E. coli sequence types in the UTI collection. When the drug resistance profiles and beta- lactamase genotype were compared, many of the E. coli strains from the meningitis cases could be further differentiated from those that caused UTI. The meningitis- causing E. coli strains may express additional factors that render them more invasive. Acknowledgements The authors would like to acknowledge the team at the Gonçalo Moniz Research Center, who coordinated the meningitis study and assisted with the laboratory work. This team included Jailton Azevedo Silva, Maira Santos, and Lorena Freire. This project was supported in part by the Fogarty International Center and the National Institute of Allergy and Infectious Diseases under grant number R25TW

37 Chapter Five Conclusions and Future Directions 31

38 Conclusion In this dissertation we investigated the distribution of drug resistance in Gram negative bacteria using E. coli as a model system. Because the bacteria that are drug resistant can be spread separately from the drug resistance genes, we investigated each independently. We described the results of these investigations in chapters two, three and four of this dissertation. The conclusions of the studies detailed in each chapter are further described below. From this work we concluded that retail produce items contain saprophytic organisms that carry novel drug resistance genes. Furthermore, a review of an MLST database showed that the sequence types of E. coli that cause meningitis and UTI have been found in retail food product items. Both of these findings highlight the need to better understand the microbiota of food products in order to prevent the spread of drug resistant genes and bacteria. The human microbiota is exposed to environmental microorganisms found on animals, plants, and soils through food products, especially those eaten raw. Previous work has shown that environmental samples, such as soils, harbor novel drug resistance genes. Retail produce and meat products have also been shown to harbor drug resistant pathogens such as E. coli. The work presented here shows that retail food products also harbor novel drug resistance genes. We believe the drug resistance genes and microbes on retail produce originate from the environment in which the produce item is grown. Consequently, this work demonstrates another link between environmental microorganisms and drug resistant human pathogens. Chapter Two We screened the microbiota of retail baby spinach for genes that could confer antimicrobial resistance to E. coli. From this screen we identified five novel antimicrobial resistance genes. These genes conferred resistance to penicillins, cephalosporins, monobactam, fluoroquinolones, and folate pathway inhibitors. Furthermore, because these sequences did confer resistance, we know each of the sequences must contain the regulatory elements necessary for expression in E. coli. From these findings we concluded that the microbiota found on retail baby spinach is a source of novel drug resistance genes. We expect that microbiota on other retail produce items will also harbor drug resistance genes. Furthermore, these resistance genes can readily be expressed in E. coli and confer high enough levels of drug resistance that treatment of an infection caused by E. coli with the corresponding drug would be expected to fail. This is particularly concerning because the drugs that these sequences confer resistance to are used to empirically treat serious Gram negative bacterial infections. Chapter Three 32

39 We further analyzed the predicted amino acid sequences of the CIP- Qnr and AMP- beta- lactamase, originally described in chapter two. From this analysis we concluded that these were novel sequences and generated further hypotheses pertaining to the mechanism of action of these two proteins. This finding also suggests the constantly evolving nature of drug resistance genes in saprophytes in the environment as well as in food consumed by humans. Chapter Four To further investigate the distribution of drug resistant bacteria and genes, we studied E. coli that caused extraintestinal infections in Salvador, Brazil. We determined the sequence type of the E. coil that caused meningitis and community- acquired urinary tract infection (UTI). We then compared these sequence types to each other and to a database of isolates that had been collected from around the world [86]. When we compared the sequence types of the UTI isolates to the meningitis isolates, we found that the most common sequence types were found in isolates that caused both types of infection. We also identified E. coli isolates from cases of both UTI and meningitis that are ST 131 and ST 69. These sequence types have previously been isolated from healthy humans, human infections, and from retail food products from around the world. We also compared the drug resistance phenotype and beta- lactamase genes found in E. coli that cause meningitis and UTI. In some cases the E. coli that caused meningitis had the same sequence type, drug resistance phenotype, and beta- lactamase genes as the E. coli isolates that caused UTI. These isolates include ST 69. Because these isolates were collected from the same city, and in some cases during the same year, it is reasonable to infer that these E. coli isolates are closely related. These findings further highlight the importance of screening the microbiota on food products for drug resistance and potentially pathogenic bacteria because previous studies suggest that the E. coli that cause UTI, especially ST 69, are spread by contaminated food products [95]. Future Directions To better understand and predict the emergence of novel drug resistance genes and drug resistant infections, it is important to gain a better understanding of how drug resistant bacteria and drug resistance genes are spread. Future studies to expand this understanding should focus on the spread of drug resistance genes from the microbiota of food products to human pathogens at a population level. To do this it is important to study the microbiota of retail food products and then the pathogens that cause infection in the people who have consumed these retail food products. These sorts of studies would require sampling food products and clinical isolates bacterial isolates from the same community over time. Examples of the sorts of 33

40 studies that would be critical to address this question are surveys of the known drug resistance genes and their sequences found in various types of retail food products such as produce and meat, as well as, surveys of the drug resistance genes found in human pathogens isolated from both hospital and community acquired infections. Appendix Tables and Figures 34

41 Table 1: Minimum inhibitory concentrations (MIC) for E. coli containing the indicated plasmids. MICs in the presence of the plasmid with the metagenomic DNA insert compared to the empty vector (psmart or 1B) are shown. In accordance with manufactures recommendations, less than a 4- fold increase was considered to be within the margin of error. Drug Resistant Plasmid Name Antimicrobial drug Drug Resistance Plasmid MIC (ug/ml) Empty Vector MIC (ug/ml) Fold Increase pamp ampicillin pamp piperacillin pamp cefotaxime <.25 <.25 0 pamp ceftazidime <.5 <.5 0 pamp cefepime pamp Aztreonam pamp Imipenem pcip ciprofloxacin <0.002 >62.5 pcip levofloxacin pcip Ampicillin pcip piperacillin pcip ceftazidime <.5 <.5 0 pcip cefotaxime <.25 <.25 0 pcip cefepime pcip Imipenem ptmp trimethoprim > >258 ptmp trimethoprim- sulfamethoxazole pazt Aztreonam pazt cefepime pazt ceftazidime pazt piperacillin psxt trimethoprim- sulfamethoxazole psxt trimethoprim > >256 pprp:1b ciprofloxacin pprp:1b levofloxacin

42 Table 2: Open reading frames identified in each of the 5 drug resistance- conferring plasmids. OrFinder, BlastX and BlastP were used to identify each open reading frame. Plasmid ORF Predicted Protein Length Most Similar Protein, Genbank Accession Number pamp 1 95 Putative Iron- Chelator Esterase EL pamp Putative Sodium Dependent Transporter ZP_ pamp Sugar Efflux Transporter YP_ pamp Extended Sepctrum Beta- lactamase ERP- 1 AAAL pamp Peptidyl- Dipeptidase YP_ pcip Pentapeptide- Repeat Containing Protein YP_ pcip Beta- lactamase Domain Containing Protein YP_ ptmp Diemethyladenoanine Transferase YP_ ptmp Protein ApaG ( Protein Cor D) YP_ ptmp Bis ( 5'- nucleoayl) Tetraphosphotase YP_ Organism E- value AA Identity Bacillus Subtilis subsp inaquoaorm 7 e % Bacillus Subtilis 2 e % subsp inaquoaorm Erwinia % tasmaniensis Et1/99 Erwinina perscina 2 e % Erwinia tasmaniensis Et1/99 Exiguobacterium antarticum Exiguobacterium antarticum ptmp Protein FolA YP_ Erwinia pyrifoliae DSM ptmp 5 91 Rhtb Family Transporter YP_ pazt Hypothetical Protein BsI_26310 ZP_ pazt Penicillin Binding Protein 2c ZP_ psxt Dihydrofolate Reductase YP_ psxt Sodium: Dicarboxylate Symporter ZP_ e % 9 e % 1 e % Erwinia tasmanienaia 1 e % Erwinia amylovora 2 e % CFBP1430 Erwinia billingiae 4 e % 3 e % Erwinina amylovora CFB e % Bacillus Subtilis 2 e % subsp inaquoaorm Bacillus Subtilis % Pseudomonas fluorsencens Pseudomonas fluorsencens 4 e % 3 e % 36

43 Table 3: Results of antimicrobial susceptibility testing of 36 E. coli isolates from cases of meningitis. The categories marked with N/A were not tested. Sensitive Intermediate Resistant N % N % N % Amikacin Ampicillin/Sulbactam Aztreonam N/A N/A 6 17 Cefazolin Cefepime N/A N/A 5 14 Cefotaxime Cefoxitin Ceftazidime N/A N/A 7 19 Ceftriaxone Cefuroxime N/A N/A 8 22 Cephalothin Chloramphenicol Ciprofloxacin N/A N/A 5 14 Gentamicin N/A N/A 5 14 Levofloxacin N/A N/A 4 11 Nitrofurantoin N/A N/A 1 3 Piperacillin/ Tazobactam Piperacillin Tetracycline N/A N/A Tobramycin N/A N/A 5 14 Trimethoprim/ Sulfamethoxazole N/A N/A

44 Table 4: Drug resistant phenotypes, beta- lactamase (blatem, blashv, or blaoxa), and sequence types found in collections of E. coli from meningitis and urinary tract infection (UTI). Meningitis UTI UTI Phenotype N Sequence Types bla N Sequence Types bla N Sequence Types bla TET+TSX +PEN+CEP H , 678, 69 bla TEM 6 394, 127, 998, 38, 73, bla TEM 18 38, 69, 93, 104, 31, 998, 14, 131, 73, 405, 62, 372, 3928 bla TEM, bla SHV TSX+PEN +CEPH , 2609, 62 bla TEM 4 62, 38, 73, 127 bla TEM 15 69, 62, 10, 12,14, 23, 73, 38, 12,2609 bla TEM TET+TSX +PEN 4 69, 746, 3702 bla TEM , 167, 976, 3704, 73, 69, 131,3910, 1312, 3704, 205, 38, 10 bla TEM 7 69, 95, 14, 38, 74, 827 bla TEM, bla SHV PEN+CEPH , 62 bla TEM 3 127, 10, 73 bla TEM , 73, 3706, 38, 127 bla TEM TET+TSX +CEPH bla TEM 1 46 bla TEM Non e None None TET+TSX 1 10 bla TEM, bla SHV 10 10, 3903, 167, 73, 127, 46, 3904, 12, 998 bla TEM None TET+TSX +FLR TET+TSX +FLR+CEP H bla TEM, bla SHV None 0 None None None 0 None None bla SHV 38

45 Figure 1: Diagram of open reading frames and DNA insert size. Open reading frame labels correspond to those in table 2. Broken arrows correspond to truncated open reading frames. Open reading frames marked with a * show similarity to known drug resistance genes. Figure not drawn to scale. pamp% 1" 3" 4*" 5" 4"Kb" pcip% 2" 1*" 2" 1.5"Kb" ptmp% 1" 2" 3" 4*" 5" 2.5"Kb" psxt% 1*" 2" 1.5"Kb" pazt% 1" 2*" 2.5"Kb" 39

46 ! pcip$qnr( EsfQnr( Mcbg( Mfpa( QnrA( QnrB( QnrC( QnrD( QnrS( QnrV( AhQnr( pcip$qnr( EsfQnr( Mcbg( Mfpa( QnrA( QnrB( QnrC( QnrD( QnrS( QnrV( AhQnr( Figure'2:' Alignment!of!Qnr!proteins,!Qnr1like!proteins,!and!the!novel!Cip1Qnr!protein.!! The!Cip1Qnr!is!highlighted!in!blue.!!AhQnr,!which!had!a!loop!region!that!was!critical!to!its!activity,!is!highlighted!in!red.!!The! alignment!was!made!with!muscle!and!visualized!in!jalview. 40

47 35.21%' 32.31%' 24.20c%' 24.08%' 23.74%' 20.91%' 20.45%' 19.68%' 19.44%' 19.21%' EsfQnr' AhQnr' QnrS' MfpA' QnrV' QnrA' QnrC' mcbg' QnrD' QnrB' Figure 3: Qnr protein family Tree Neighbor joining tree made from the protein sequences using blossom 62 scores in Jalview. The alignment which the tree was based on was made using MUSCLE. As well as a list of the amino acid percent identity each Qnr and Qnr- like protein as compared to the novel Cip- Qnr, calculated from the alignment in figure 1. Cip- Qnr is highlighted in red in the tree. 41

48 Figure 4: Predicted Cip- Qnr Structure We used Phyre to predict the structure of Cip- Qnr. (Top) predicted structure of Cip- Qnr displayed with Pymol. (Bottom) Predicted structure of Cip- Qnr ( green) aligned with 2bm5 (blue), the crystal structure from the SCOP family pentapeptide repeat, which the structure prediction is based on. 42

49 Figure 5: Model of Amp- beta- lactamase. This model was made using the Phyre2 server and visualized in Geneious. Residues highlighted in pink, are critical residues identified by CAS and are located with in the binding pocket (Ser- 70, Lys- 73, Ser- 130, and Gly- 166). 43

50 !!!! Figure'6:' pamp%beta%lactamase!alignment!!!pamp%beta%lactamase!alignment!made!with!muscle!and!visualized!in!geneious.!!the!residue!numbering!reported!here!and!in! the!text!is!based!on!the!consensus!ambler!numbering!system,!and!therefore!does!not!correspond!to!the!exact!residue! numbers!in!the!alignment.!residues!highlighted!in!red!are!the!critical!residues!identified!by!cas!and!described!in!the!text! (Ser%70,!Lys%73,!Ser%130,!and!Gly%166).!!The!residues!highlighted!in!Yellow,!are!the!residues!that!have!known!substitutions! that!alter!the!substrate!specificity!of!the!enzymes!(glu%104%ser,!met%182%thr!and!ala%237%ser).!!the!residues!highlighted!in! blue!are!the!residues!that!have!previously!been!found!to!be!important!for!substrate!specificity,!but!the!effects!of!the! substitutions!observed!here!and!unknown!(met%69%cys,!glu%240%gly,!arg%244%thr).!p62593,!is!tem%1,!and!d1hzoa,!is!the! sequence!of!the!protein!that!the!structure!prediction!in!figure!2!is!based!on.!! 44

51 Figure 7: Sequence types vs. year of isolate collection. Isolates that were resistant to third- generation cephalosporins are labeled with the bla CTX- M type gene they carry. Isolates marked with a * did not yield an interpretable DNA sequences and are labeled with the name of the bla CTX- M group they tested positive for. 45