Use of a Novel High-Throughput Genetic Selection to Identify Regulators of Selected Vibrio Cholerae Late Genes; Characterization of Phob As a Regulator of Xds

Use of a novel high-throughput genetic selection to identify regulators of selected Vibrio cholerae late genes; characterization of PhoB as a regulator of xds

A thesis

submitted by

EMILYKATE MCDONOUGH

In partial fulfillment of the requirements for the degree of

Doctor of Philosophy

Molecular Microbiology

TUFTS UNIVERSITY

Sackler School of Graduate Biomedical Sciences

August, 2014

Advisor: Andrew Camilli

To everyone who inspired me to keep going.

ACKNOWLEDGEMENTS First and foremost, thank you Andy for mentoring me with such a kind and enthusiastic nature. Thank you for always being approachable and doing your best to make time for teaching. Thank you especially for the first few years in the lab, when I stumbled around my bench, hardly knowing anything about microbiology; you were always supportive and excited about even my smallest achievements. Thank you for always listening to my ideas and for letting me make my own decisions, even when you completely disagree with my choices. As I move through this world, whether in science or film, I will always be thankful for what I have learned and accomplished in your laboratory. I hope you continue to infect those around you with your passion, optimism and kindness.

To my committee, thank you for all of your input throughout the years. You have challenged me to think on my feet, a skill that is very useful in this field. Thank you for all of the guidance, encouragement, and kind criticism. Linc – thank you for your constant words of support. Even when I have felt discouraged by my work, whether you know my feelings or not, you have made a point to tell me otherwise. Also thank you for critiquing my writing so carefully. Michael – thank you for your mentorship. I love talking with you about both science and life – and I wish I’d done it more. You have always cared about me as a person, which has helped me remember that there is more to life than graduate school. Ralph – thank you for your ever-questioning ways; you are a true Obie, whatever that means. Having you on my committee has made me stronger and a more confident scientist.

Thank you to the Camilli lab members, both past and present members. The lab has been my home for seven years, and I will think fondly of many memories after I have left. Lobster bay – (Katie, Jason, Neil, Steph) thank you for all the conversations about science and life. Our little mascot is falling to pieces, but I hope his namesake will live on. Pick a good replacement for me! Free high fives. Ayman – thank you for opening my eyes to the world of Illustrator short cuts and teaching me the ins and outs of Word. Also thank you for your excitement about biochemistry and your undying willingness to help out. You have been a good source of moral support, and always provide a good laugh. Anne – as one of the strongest women I know, you have been a role model for me. I hope our paths will cross again soon. Rita – thank you for always listening to my ideas and making me feel like they mattered. Evan – I’m so glad we got to write the book chapter together, thanks for that experience. Also, thanks for making the lab a great place to be. Dave – I’m so grateful that you are in the lab. You have been an amazing resource for and critic of my work. Thanks for contributing so much to my thesis work and for making me a better scientist; you are a great one. Also thanks for all the camera talk. Revati and Heather – thanks for all the lunches and great scientific talks. You guys have been great friends. Ellie – I’m so lucky that you joined our lab. You have been a wonderful friend these past few years. Thanks for coffee, walks, cheering me up, morale support, etc. I’m excited to see where your life takes you – I know you’ll reach all your dreams.

Thank you to my friends for all of your undying support, and understanding of my lack of time to come play or visit you in NYC. There’s too many of you to thank individually, but know that you are all in my heart.

Kayle. Thank you. Thank you for putting up with all my stress and bursts of frustration. Thank you for providing me with so much love and support. Thank you for our scientific conversations. Thank you for reminding me to breathe and to live. I have grown so much from knowing you. I love you.

And finally, to my family, thank you. You have always believed in me and told me to follow my dreams. Popsy and momma, thank you for all the opportunities you have given me – I am so very lucky. I’m sorry I didn’t stay little, but I think this accomplishment is pretty awesome, eh? Hannie – you always say that I’m your role model sister, but you should know that I look up to you every day; you’re so smart, driven, and not afraid to stick your neck out. I admire the woman you’ve become.

ABSTRACT Vibrio cholerae, the causative agent of the severe diarrheal disease cholera, is a natural member of temperate aquatic environments around the world. V. cholerae undergoes adaptive shifts in gene expression throughout the various stages of its life cycle. Our laboratory identified 57 V. cholerae genes that are expressed specifically at or near the end of the infection cycle in the infant mouse model of infection. Many of these

‘late’ genes appear to be involved in preparing the bacteria for the shift from the host small intestine to the aquatic environment. In contrast, these same genes are not required for V. cholerae colonization of the mouse small intestine. These data led us to hypothesize that V. cholerae evolved to preinduce such dissemination genes while still in the host intestinal tract in order to optimize their chances of successfully transitioning to the aquatic environment.

To gain support for this hypothesis and to extend our understanding of gene regulation at this critical transition, the goal of my thesis project was to determine what signals V. cholerae senses in the small intestine that alert the organism to the upcoming change in environment. The first half of my thesis describes the implementation and results of a high throughput genetic selection we developed to identify transcriptional regulators of late genes. The selection was designed such that both activators and repressors could be identified, as it is not known if these genes are induced or de- repressed late in infection. We settled on three late genes as query genes for the selection: cdpA (encoding a cytoplasmic cyclic-diguanylate phosphodiesterase), emrD (encoding a efflux pump inner membrane subunit), and xds (encoding a secreted DNA exonuclease).

The selection results did not identify any regulators of cdpA. Conversely, pepA (encoding

5 a leucyl amino-peptidase) was identified as a repressor of emrD, and phoB (encoding the phosphate starvation response regulator) was identified as an activator of xds.

Additionally, I have provided evidence that while regulation of xds by PhoB is relevant to the host intestinal environment, expression of xds can occur early or late in infection depending on the physiological state of the inoculum used for infection.

In the second half of my thesis I have worked towards a deeper understanding of the biological significance of the regulation of xds by PhoB. As Xds is an exonuclease, we hypothesized that PhoB induces xds under phosphate limiting conditions as part of the upregulation of phosphate acquisition genes. Indeed, V. cholerae is known to be able to survive using DNA as a sole source of phosphate, and Xds has been described as important for this ability. In this work I describe the identification and preliminary characterization of two nucleotidases, UshA and CpdB, that contribute to the ability of V. cholerae to utilize nucleotides as a source of phosphate. I have shown that both of ushA and cpdB are induced by phosphate limiting conditions, further supporting the significance of nucleotides as sources of phosphate.

TABLE OF CONTENTS ACKNOWLEDGEMENTS...... 3 ABSTRACT...... 5 TABLE OF CONTENTS...... 7 LIST OF FIGURES ...... 11 LIST OF TABLES ...... 14 CHAPTER I: ...... 15 INTRODUCTION...... 15 I.1. V. CHOLERAE AS A MODEL WATERBORNE PATHOGEN...... 16 I.1.1. Importance of waterborne microbial pathogens...... 16 I.1.2. Role of the aquatic reservoir in transmission of waterborne pathogens...... 17 I.1.3. Vibrio cholerae as a model waterborne pathogen...... 19 I.2. OVERVIEW OF THE V. CHOLERAE VIRULENCE PROGRAM ...... 24 I.2.1. Primary virulence genes and their regulation...... 24 I.2.2. Quorum sensing and the regulation of virulence...... 26 I.2.3. Cyclic diguanylate and the regulation of virulence...... 29 I.2.4. Bistable expression of virulence factors late in infection ...... 31 I.3. LATE STAGE OF INFECTION: PREPARING FOR THE TRANSITION ...... 35 I.3.1. An RpoS and HapR mediated transition ...... 36 I.3.2. V. cholerae late genes...... 40 I.4. V. CHOLERAE IN THE AQUATIC ENVIRONMENT ...... 43 I.4.1. Dissemination of V. cholerae...... 44 I.4.2. Nutrient starvation and strategies for survival...... 46 I.4.2.1. Glycogen storage and utilization ...... 46 I.4.2.2. Chitin as an alternative source of carbon and nitrogen...... 48 I.4.2.3. Biofilm formation of V. cholerae and importance of the phenotype...... 49 I.4.3.4. Viable but non-culturable state...... 53 I.4.4. Transmission to a new host and hyperinfectivity...... 54 I.5. THE PHOSPHATE STARVATION RESPONSE IN V. CHOLERAE...... 58 I.5.1. Phosphate acquisition and storage systems in bacteria including V. cholerae ...... 58 I.5.1.1. Environmental sources of phosphate ...... 58 I.5.1.2. Transport of inorganic phosphate ...... 62 I.5.1.3. Polyphosphate storage and breakdown...... 65 I.5.2. The Pho regulon...... 66 I.5.2.1. The two-component system PhoB/R and control by PhoU ...... 66 I.5.2.2. Genes and programs induced by PhoB...... 68 I.5.2.3. Regulation of genes by PhoB ...... 69 I.5.3. Importance of the phosphate response of V. cholerae during infection ...... 69 I.6. DNA AS AN ALTERNATIVE SOURCE OF CARBON, NITROGEN, AND PHOSPHATE IN BACTERIA...... 70 I.6.1. General strategies for use of DNA as a source of nutrients ...... 70

1.6.2. Bacterial DNA nucleases ...... 71 I.6.3. Bacterial nucleotidases...... 73 1.6.4. V. cholerae interactions with DNA ...... 76 1.7. THESIS STATEMENT: AIMS AND RATIONALE ...... 77 1.7.1. Aims...... 77 I.7.2. Introduction to specific late genes studied in this thesis...... 78 I.3.2.1. cdpA ...... 78 I.3.2.2. emrD ...... 78 I.3.2.3. xds ...... 79 CHAPTER II:...... 80 GENETIC SELECTION TO IDENTIFY REGULATORS OF V. CHOLERAE LATE GENES ...... 80 II.1. INTRODUCTION ...... 81 II.2. RESULTS ...... 82 II.2.1. Design of a novel genetic selection ...... 82 II.2.1.1. Original selection design and problems therein ...... 85 II.2.1.2. Version two of the genetic selection ...... 92 II.2.2. Construction of saturating transposon libraries ...... 95 II.2.3. Antibiotic minimal inhibitory concentration (MIC) experiments with transposon libraries ...... 101 II.2.4. Design of data analysis...... 103 II.2.5. Identification and in vitro validation of putative late gene regulators...... 109 II.2.5.1. cdpA results ...... 112 II.2.5.2. emrD results ...... 121 II.2.5.3. xds results ...... 133 II.2.6. Investigation of the nature of PhoB regulation of xds...... 142 II.2.7. PhoB as a regulator of additional V. cholerae late genes...... 151 II.3. DISCUSSION ...... 156 II.3.1. Discussion of the novel genetic selection method ...... 156 II.3.2. cdpA selection discussion ...... 159 II.3.3. emrD selection discussion ...... 161 II.3.4. xds selection discussion...... 165 CHAPTER III: ...... 169 RE-EXAMINATION OF XDS EXPRESSION DURING INFECTION OF AN INFANT MOUSE...... 169 III.1. OVERVIEW ...... 170 III.2. RESULTS...... 170 III.2.1. PhoB regulates xds during infection of an infant mouse ...... 170 III.2.2. Induction of xds occurs early, rather than late in infection...... 174 III.2.3. xds is not required for early colonization of the infant mouse...... 177 III.3. DISCUSSION...... 180 CHAPTER IV: ...... 185

IDENTIFICATION OF FACTORS REQUIRED FOR THE USE OF NUCLEOTIDES AND DNA AS SOURCES OF PHOSPHATE BY V. CHOLERAE ...... 185 IV.1. OVERVIEW ...... 186 IV.2. RESULTS ...... 187 IV.2.1. PhoX is not required for utilization of eDNA as a phosphate source ...... 187 IV.2.2. Transposon mutant screen to identify additional V. cholerae phosphatases and preliminary validation of hits...... 190 IV.2.3. ushA is not required for growth on exogenous DNA...... 193 IV.2.4. UshA is a non-nucleotide-specific 5’ nucleotidase, putatively periplasmic 195 IV.2.5. Deletion of ushA does not affect infant mouse colonization of V. cholerae 201 IV.2.6. V. cholerae is able to utilize 5’ nucleotides as a source of nitrogen, but not carbon, and ushA is required for this phenotype ...... 203 IV.2.7. The competence pilus major subunit, PilA, is not required for growth on eDNA when supplied as the sole source of phosphate...... 205 IV.2.8. Identification of additional putative nucleotidases ...... 209 IV.2.9. Identification of CpdB as a 3’ nucleotidase ...... 210 IV.2.10. Deletion of cpdB and ushA does not affect the ability of V. cholerae to grow using eDNA as the sole source of phosphate ...... 216 IV.2.11. Genetic regulation of cpdB and ushA...... 220 IV.3. DISCUSSION ...... 222 IV.3.1. Discussion of the nucleotidases identified in this work...... 222 IV.3.2. Model for utilization of eDNA as a source of phosphate ...... 229 CHAPTER V:...... 233 SUMMARY AND PERSPECTIVES...... 233 CHAPTER VI: ...... 248 MATERIALS AND METHODS ...... 248 VI.1. CHAPER II ...... 249 VI.1.1. Media and bacterial strains...... 249 VI.1.2. Strain construction ...... 250 VI.1.3. MIC experiments on plasmid insertion deletions ...... 254 VI.1.4. Transposon mutagenesis ...... 254 VI.1.5. Determination of selection conditions...... 255 VI.1.6. High-throughput sequencing and genetic selection ...... 256 VI.1.7. Data analysis...... 257 VI.1.8. RNA purification and qRT-PCR ...... 259 VI.1.9. Purification of PhoBCA ...... 260 VI.1.10. Electromobility shift assays...... 262 VI.2 CHAPTER III...... 263 VI.2.1. Strains...... 263 VI.2.2. Recombination-based in vivo expression technology (RIVET) ...... 263 VI.2.3. Evaluation of xds in vivo fitness ...... 264 VI.3. CHAPTER IV...... 264 VI.3.1. Strain construction ...... 264

VI.3.2. Phosphatase screen ...... 265 VI.3.3. Phosphate growth curves ...... 266 VI.3.4. Nitrogen growth curves ...... 267 VI.3.5. Nucleotidase assays...... 267 VI.3.6. Infant mouse competition assay...... 269 VI.3.7. RNA purification and qRT-PCR ...... 269 VI. 4. STRAINS AND PRIMERS...... 271 VI.4.1. Chapter II ...... 271 VI.4.2. Chapter III...... 277 VI.4.3. Chapter IV ...... 278 REFERENCES...... 281

LIST OF FIGURES

Figure I-1. Simple life cycle of toxigenic V. cholerae...... 22

Figure I-2. Quorum sensing and starvation conditions coordinate expression of many V. cholerae genes...... 28

Figure I-3. cAMP-CRP repression and ToxT autoregulation create the bistable expression pattern of toxT...... 33

Figure I-4. Transcriptional changes of V. cholerae during infection...... 37

Figure I-5. Dissemination and transmission of V. cholerae in endemic areas...... 45

Figure I-6. Comparison of pst operons in E. coli and V. cholerae...... 64

Figure II-1. Possible transposon insertions leading to misregulation of the late genes. ... 84

Figure II-2. Venn diagram showing the high degree of overlap between selections from the original selection...... 86

Figure II-3. Effect of ompU deletion on transcript levels of cdpA and xds...... 90

Figure II-4. Effect of cysB or rseA deletions on expression of late genes...... 91

Figure II-5. Design of the genetic selection, version two...... 94

Figure II-6. pDL1098 delivery system for mTn10...... 96

Figure II-7. Flow chart of data analysis methods...... 105

Figure II-8. Fold enrichment of in cis Tn insertions...... 111

Figure II-9. In vitro validation test of putative cdpA repressors...... 116

Figure II-10. Ethanol stress activation of RpoE does not induce cdpA transcription. .... 119

Figure II-11. Impact of rpoE::mTn10 on expression of cdpA...... 120

Figure II-12. Induction of rpoE does not activate transcription of emrD...... 124

Figure II-13. VC0469 does not validate as a repressor of emrD...... 126

Figure II-14. Transposon insertions enriched in VC0465 – VC0469 in the emrD selections...... 127

Figure II-15. Transcription of emrD is induced in a gshB knockout strain...... 129

Figure II-16. Deletion of pepA increase transcription of emrD...... 130

Figure II-17. G482D point mutant of PepA behaves like a deletion of pepA...... 132

Figure II-18. Transposon insertions enriched in phoR...... 137

Figure II-19. PhoU is an indirect repressor of xds...... 139

Figure II-20. Induction of xds by low phosphate is dependent on PhoB activity...... 141

Figure 11-21. PhoBCA is a constitutively active protein in vivo...... 144

Figure II-22. Probes used for EMSA experiments...... 146

Figure II-23. PhoBCA binds to the xds promoter...... 147

Figure II-24. Rough estimation of the PhoB binding site in the xds promoter...... 149

Figure II-25. Competition binding assays with PhoBCA...... 150

Figure II-26. PhoB induces acgAB under low phosphate conditions...... 152

Figure II-27. PhoB binds directly to the acgAB promoter...... 154

Figure II-28. PhoB regulation of additional late genes...... 155

Figure II-29. γ-glutamyl cycle in Vibrio cholerae...... 163

Figure III-1. Recombination-based in vivo expression technology (RIVET) to measure xds transcription...... 172

Figure III-2. PhoB regulates xds during infection of an infant mouse...... 173

Figure III-3. Expression of xds occurs early, rather than late, during infection...... 176

Figure III-4. Early induction of xds is not required for infant mouse colonization...... 179

Figure IV-1. V. cholerae does not require phoX for the utilization of eDNA as a source of phosphate...... 188

Figure IV-2. Transposon insertions in nupC and ushA...... 192

Figure IV-3. ushA is not required for growth on eDNA as a source of phosphate...... 194

Figure IV-4. ushA, but not phoX, is required for growth on nucleotides as a source of phosphate...... 196

Figure IV-5. UshA is required for 5’ nucleotidase activity of V. cholerae...... 200

Figure IV-6. ushA is not required for infant mouse small intestinal colonization...... 202

Figure IV-7. ushA is required for utilization of 5’ nucleotides as sources of N...... 204

Figure IV-8. pilA does not contribute to growth on eDNA as a source of phosphate. ... 206

Figure IV-9. Affect of eDNA, nucleotides, and Pi limitation on expression of pilA...... 208

Figure IV-10. Growth of cpdB, yjjG, and VCA0545 mutants on 5’ nucleotides...... 211

Figure IV-11. Growth of wild type and ΔcpdBΔyjjGΔVCA0545 on eDNA...... 212

Figure IV-12. cpdB is required for growth on 3’CMP...... 214

Figure IV-13. cpdB, but not ushA, is required for 3’ nucleotidase activity...... 215

Figure IV-14. Deletion of cpdB, alone or together with ushA, does not affect growth on eDNA as a sole source of Pi...... 217

Figure IV-15. Triple mutant of xds, ushA, and cpdB inhibits growth on eDNA...... 219

Figure IV-16. Regulation of ushA and cpdB by phosphate and PhoB...... 221

Figure IV-17. Model for the utilization of eDNA as a source of phosphate in V. cholerae...... 231

Figure V-1. Alternative models of phosphate level and affect on gene expression during infection...... 239

LIST OF TABLES

Table I-1. Comparison of physiochemical parameters of typical environments for V. cholerae...... 18

Table I-2. A subset of V. cholerae genes induced late in the infant mouse model of colonization...... 41

Table II-1. MIC of putative regulators...... 88

Table II-2. Properties of transposon libraries in biological replicate...... 98

Table II-3. Properties of final mixed transposon libraries...... 100

Table II-4. Antibiotic concentrations used in the genetic selections...... 102

Table II-5. Putative repressors of cdpA identified in the minus IPTG selection...... 113

Table II-6. Putative activators of cdpA identified in the genetic selection...... 117

Table II-7. Putative repressors of emrD...... 122

Table II-8. Putative activators of emrD...... 123

Table II-9. Putative repressors of xds...... 134

Table II-10. Putative activators of xds...... 136

Table IV-1. Growth ability of various strains on 5’ nucleotides...... 197

Table VI-1. Strains used in Chapter II...... 271

Table VI-2. Primers used in Chapter II...... 273

Table VI-3. Strains used in Chapter III...... 277

Table VI-4. Strains used in Chapter IV...... 278

Table VI-5. Primers used in Chapter IV...... 279

CHAPTER I:

INTRODUCTION

Much of the text and figures/tables in this section are adapted from a book chapter I wrote with help from Evan Bradley. Permission for reprint was obtained from ASM Press:

McDonough, E., Bradley, E., Camilli, A. Regulating the Transition of Vibrio cholerae Out of the Host. In: Regulation of Bacterial Virulence. ML Vasil, AJ Darwin (ed) ASM Press, Washington, DC, 2013.

I.1. V. CHOLERAE AS A MODEL WATERBORNE PATHOGEN

I.1.1. Importance of waterborne microbial pathogens Waterborne microbial pathogens reside in virtually every aquatic environment on

Earth, including fresh, brackish and marine waters. These pathogens, which include various viruses, bacteria and protozoa, contribute to significant morbidity and mortality worldwide (Geldreich 1996). When ingested, most waterborne pathogens cause some type of gastrointestinal disease ranging from mild enteritis to explosive, secretory

(watery) diarrhea (Sharma, et al. 2003, Skovgaard 2007). The risk of disease from many of these pathogens is much higher in areas where the infrastructure does not allow for proper sanitation and wastewater treatment. Unfortunately, the impact of disease in the developing world is compounded by the high financial burden of prophylactic and curative treatment plans for many diseases (Geldreich 1996).

Waterborne pathogens also present a threat in the developed world. In rural areas, water from wells can be contaminated from sewage runoff, and in urban areas, wastewater treatment plants are known to be less than 100% effective in removing certain waterborne pathogens. For example, Cryptosporidium oocysts are able to survive the microbiocidal levels of chlorine used in water purification (Sharma, et al. 2003,

Skovgaard 2007). Enteric viruses, including rotavirus, are also problematic for treatment facilities, as the physical and chemical filtration systems only remove between 50% and

90% of the viruses present in sewage, which is not enough to fully prevent contamination of environmental water sources (Okoh, et al. 2010). Numerous bacterial pathogens, including Salmonella spp, Shigella spp and enterohemorrhagic Escherichia coli, can also be isolated in the treated effluent of wastewater facilities. Additionally, street level

16 sewage drainage pipes and wastewater treatment plants may be overwhelmed by an abundance of rain or flooding, resulting in contaminated drinking and recreational waters

(Geldreich 1996).

I.1.2. Role of the aquatic reservoir in transmission of waterborne pathogens Marine, brackish and fresh waters all present different challenges to the survival and persistence of waterborne pathogens, including, but not limited to, macronutrient and micronutrient deprivation, predation, ultraviolet light and fluctuations in pH, temperature, dissolved oxygen and osmolarity (Table I-1) (Freter, et al. 1961, Brookes, et al. 2004,

Alam, et al. 2007, Kamal, et al. 2007, Schild, et al. 2007, Nelson, et al. 2008, Lara, et al.

2009). In spite of these challenges, many pathogens have evolved to spend some portion of their life cycle in an aquatic environment. The benefits of being a waterborne pathogen are several. As water is essential to life, virtually all hosts of waterborne pathogens come in contact with contaminated water at some frequency. Additionally, water can provide a means to spread pathogens from one region to another, either with the current or with migrating birds, zooplankton, phytoplankton or insects that pathogens may associate with

(Geldreich 1996, Lipp, et al. 2002, Brookes, et al. 2004).

Although numerous organisms have been described as waterborne pathogens, the precise role of the aquatic environment in transmission of many of these organisms is unclear. What is known about transmission of waterborne pathogens suggests that a wide variety of strategies for spread and persistence are used by these pathogens. For some, such as the protozoan Cryptosporidium parvum and many enteric viruses, the aquatic environment may simply provide a place to quiescently persist until a new host is encountered (Sharma, et al. 2003, Skovgaard 2007, Okoh, et al. 2010). In contrast,

Table I-1. Comparison of physiochemical parameters of typical environments for V. cholerae.

Physiochemical Rice Water Stool Urban Pond parameter (RWS) (Dhaka) Estuarine Salinity (ppt) 3.4a 0.1a 0.1-19e pH 8.3-7b 6.6-6.9a 8.0f Temperature 37° Ch 26-28°Ca 27.2-31.4°Ce + a - - e Fixed nitrogen 938 mM NH4 59.8 mM NO2 8-40 mM NO3 + d 1.8 mM NH4 3- a d g Phosphate (PO4 ) 15.2 mM 0.0002 mM 0.05-0.13 mM Redox status Anaerobicb,c,i Relatively high Low oxygen: oxygen: 1-8 x 10-4%g 6% dissolved oxygena a(Nelson, et al. 2008) b(Freter, et al. 1961) c(Merrell, et al. 2002a) d(Schild, et al. 2007) e(Lara, et al. 2009) f(Alam, et al. 2007) g(Kamal, et al. 2007) hAs measured directly after exit from the human host iDetermined by direct measurement of redox potential and inference from gene expression data

Legionella spp and some members of the Vibrio genus are natural inhabitants of the aquatic environment, either in free-living or host-associated states. Some of these native inhabitants only cause incidental infection of humans (e.g. L. pneumophila, V. parahaemolyticus), while others are facultative human pathogens (V. cholerae) (Grimes

1991). Additionally, water contamination has been implicated in the transmission of numerous pathogens, although this may not be their dominant mode of transmission.

These include enterohemorrhagic E. coli, Campylobacter jejuni, Yersinia enterocolitica,

Helicobacter pylori, Pseudomonas aeruginosa, and Aeromonas spp (Grimes 1991,

Sharma, et al. 2003, Skovgaard 2007).

I.1.3. Vibrio cholerae as a model waterborne pathogen Many members of the halophilic Vibrio genus have been reported to cause disease in humans. These species are natural members of aquatic environments around the world, and thus pose a significant threat to many populations. One member of this genus, V. cholerae, is the causative agent of the severe diarrheal disease Asiatic cholera and is a model for waterborne facultative pathogens. Although Vibrio species are found within marine or brackish waters, V. cholerae is unique in that it can also persist in freshwater environments, a trait that is central to its epidemic potential (Faruque, et al. 1998a, Reidl and Klose 2002).

V. cholerae strains are divided into over 200 serogroups on the basis of their LPS

O-antigen. While numerous serogroups have contributed to localized outbreaks of cholera, only two serogroups have been reported to cause epidemics: O1 and O139.

Within these serogroups, only strains containing genes encoding the major virulence factors, cholera toxin (CT) and the toxin co-regulated pilus (TCP), are considered

19 toxigenic and of epidemic potential (Faruque, et al. 1998a, Reidl and Klose 2002, Safa, et al. 2010, Faruque 2011). These toxigenic strains are lysogens of the filamentous bacteriophage CTXΦ, which harbors at least one copy of the CT-encoding genes (ctxAB).

The CTXΦ has been demonstrated to be fully mobilizable and able to lysogenize new strains (Waldor and Mekalanos 1996). This ability of ctxAB to be transferred via CTXΦ infection may contribute to the emergence of new toxigenic V. cholerae strains.

The O1 serogroup of V. cholerae can be further divided into two biotypes: classical and El Tor. The world is currently experiencing the seventh cholera pandemic since 1817. Despite the limited data for the first five pandemics, V. cholerae O1 of the classical biotype is considered the major biotype responsible for the first six pandemics.

However the seventh pandemic, which began in 1961, has been dominated by V. cholerae

O1 of the El Tor biotype. It has been suggested, but not conclusively shown, that one evolutionary pressure contributing to the switch from classical to the El Tor biotype as the major pathogenic strain may be that El Tor is better able to survive within the aquatic environment (Faruque, et al. 1998a, Reidl and Klose 2002, Faruque 2011). In addition to harboring numerous genotypic differences, the classical and El Tor biotypes carry different CTXΦ genomes, annotated as CTXΦCla and CTXΦEl, respectively. Among other differences, these two bacteriophages encode different CT; CTXΦCla-encoded CT induces more severe diarrhea than CTXΦEl-encoded CT. Since 1992, the classical biotype has not been isolated from environmental samples and therefore, is considered extinct.

However, there have been recent reports of altered or atypical El Tor strains that harbor either CTXΦCla or a hybrid of CTXΦCla and CTXΦEl. It is unclear where the CTXΦCla was acquired by these atypical El Tor strains (Safa, et al. 2010). The combination of El

Tor environmental fitness and the high toxicity of the CTXΦCla-encoded CT may provide a significant fitness advantage for these atypical El Tor variants in the host and aquatic environment and potentially for transmission. It will be interesting to see if these strains become the dominant epidemic-causing V. cholerae.

While V. cholerae O1 remains the dominant serogroup causing disease, a new serogroup, V. cholerae O139, emerged in India and Bangladesh in 1992 and has been contributing to disease since then (Faruque, et al. 1998a, Faruque and Nair 2002, Reidl and Klose 2002, Faruque 2011). Interestingly, the O139 strain is closely related to O1 El

Tor, and may have emerged via replacement of the O1-antigen locus with newly transferred O139 O-antigen genes. As with V. cholerae O1, some O139 strains are toxigenic and others are not (Faruque and Nair 2002). Currently, there appears to be an evolutionary battle between toxigenic V. cholerae O1 and O139 strains. Since 1992 numerous O139 genetic variants have been isolated, suggesting this strain is evolving to compete with O1 strains for maintenance in the environmental niche (Faruque, et al.

1999, Faruque, et al. 2000, Faruque and Nair 2002, Faruque 2011). By following the changes in the O139 genome, a clearer picture may arise of why certain strains dominate during an epidemic – information that may prove useful for combating cholera.

A toxigenic V. cholerae encounters two disparate environments during its life cycle: the human intestinal tract and the aquatic environment (Figure I-1 and Table I-1).

During life in the environment, V. cholerae is often found in biofilms associated with phyto- and zooplankton but may also persist in a planktonic state (Lipp, et al. 2002). The bacteria enter the human host via ingestion of contaminated water or food and subsequently colonize the small intestine (Peterson 2002). Eventually, the bacteria are

Rice Water Stool -High phosphate concentration -High !xed nitrogen concentration -Salinity >3 ppt -Micro aerobic to anaerobic

Ingestion of esophagus contaminated water

stomach

large intestine

small intestine

Fresh water environment -Low phosphate concentration -Low !xed nitrogen concentration -Salinity <0.1 ppt -Relatively aerobic

Shedding in rice Vibrio cholerae water stool (RWS)

Figure I-1. Simple life cycle of toxigenic V. cholerae. V. cholerae present in contaminated fresh water sources are ingested and cause cholera in humans. The bacteria replicate in the small intestine and are shed back into the environment in rice water stool (RWS). The shed bacteria return to the aquatic environment to initiate another infection cycle. Key chemical differences between these environments are listed and measurements taken of the typical chemical composition of these environments are presented in Table I-1.

shed back into the environment in secretory diarrhea, which is termed rice water stool

(RWS) due to its appearance. Back in the environment, V. cholerae bacteria persist or, if conditions permit, multiply and spread via water currents until once again ingested by a host. During outbreaks, the mode of transmission may be dominated by rapid spread of the hyperinfectious form of V. cholerae that is shed in the RWS of cholera patients

(discussed in more detail below) (Merrell, et al. 2002, Hartley, et al. 2006). Although direct spread of cholera within households occurs with high frequency (Weil, et al. 2009,

Kendall, et al. 2010), larger outbreaks and epidemics occur via ingestion of contaminated fresh water used for drinking, food preparation, etc.

In order to mediate the transition from one environment to another, toxigenic V. cholerae undergoes adaptive shifts in gene expression throughout the various stages of its life cycle (e.g. DiRita 1992, Merrell, et al. 2002, Nielsen, et al. 2006, Childers and Klose

2007, Schild, et al. 2007, Nielsen, et al. 2010). Properly mediating the transition between host and environment is of particular importance to toxigenic V. cholerae, as it encounters numerous physical and biological stresses during these transitions.

Interestingly, there is evidence that V. cholerae has evolved to preemptively prepare itself for this transition by turning off virulence genes and turning on environmental survival genes while still in the human host in order to better survive the transition from host to environment (Schild, et al. 2007). This chapter will focus heavily on the genetic changes the bacterium undergoes as it transitions out of the host and into the aquatic environment, as well as nutrient acquisition strategies used by the organism after exiting the host.

I.2. OVERVIEW OF THE V. CHOLERAE VIRULENCE PROGRAM In order to better understand the changes in V. cholerae gene expression that occur during the transition out of the host, we must first discuss the state of gene expression prior to this transition. This prior, acute stage of infection within the small intestine is when V. cholerae multiplies on the epithelium and expresses virulence factors.

I.2.1. Primary virulence genes and their regulation For a toxigenic V. cholerae strain to colonize the small intestine and cause disease, it requires expression of TCP and CT (Faruque 2011). TCP is a type IV bundle- forming pilus initially expressed upon entry into the host (Lee, et al. 1999), and is absolutely essential for V. cholerae colonization of the small intestine (Herrington, et al.

1988). While the precise mechanism by which TCP contributes to colonization is unknown, the pilus is thought to allow for interactions between bacterial cells rather than bacterial-host interactions (Childers and Klose 2007). CT, an enterotoxin and a member of the A-B subunit type ADP-ribosylating toxins, is the virulence factor responsible for the profuse secretory diarrhea that is associated with cholera (Herrington, et al. 1988,

Peterson 2002, Childers and Klose 2007, Faruque 2011).

Some of the V. cholerae virulence genes are present in regions of the V. cholerae genome that are potentially horizontally transferable (Reidl and Klose 2002, Childers and

Klose 2007). The tcpA-F operon, which encodes TCP, is located within the Vibrio pathogenicity island 1 (VPI-1), which also contains genes for the accessory colonization factor (Acf) and critical virulence gene regulators (TcpPH and ToxT) (Childers and Klose

2007). Although the VPI-1 was initially shown to be horizontally transferrable (Karaolis, et al. 1999), this phenotype was not reproducible (Faruque, et al. 2003). Nevertheless, the 24

VPI-1 and other virulence gene loci should be subject to low frequency transfer by natural transformation, a process recently shown to occur in V. cholerae during biofilm growth on chitinous surfaces (Meibom, et al. 2005, Blokesch and Schoolnik 2008). As mentioned above, the genes encoding CT (ctxAB) are present in the CTXΦ prophage, which is associated with all known pandemic-causing strains (Reidl and Klose 2002).

Horizontal transfer of CTXΦ between strains during co-infection of the small intestine has been observed in experimentally infected animals (Waldor and Mekalanos 1996), and is likely to occur with some frequency within humans, who can be simultaneously infected with multiple toxigenic V. cholerae strains (Kendall, et al. 2010). TCP is the receptor for CTXΦ (Childers and Klose 2007), suggesting that the CTXΦ is most likely acquired by a strain only after acquisition of the VPI-1. VPI-1 transfer via natural transformation in the environment, followed by CTXΦ transfer via phage particles during human infection, may contribute to emergence of new, virulent, toxigenic strains. This hypothesis is supported by the frequent emergence of new toxigenic variants (Faruque, et al. 1998b, Faruque, et al. 1999, Faruque, et al. 2000, Faruque and Nair 2002, Stine, et al.

2008, Faruque 2011).

The regulation of ctxAB and tcpA-F expression has been studied extensively with respect to the factors that induce and repress transcription, though not exhaustively with respect to the activating signals. The major virulence regulator, ToxT, is responsible for inducing expression of ctxAB and tcpA-F. The gene encoding ToxT is located within the

VPI-1, directly downstream of the tcpA-F operon. Expression of toxT is under tight regulation of two different membrane-bound transcription factor complexes: TcpP/H

(within the VPI-1) and ToxR/S (encoded outside of the VPI-1 and CTXΦ). Although the

25 precise cues which activate signaling through TcpP/H or ToxR/S are yet to be elucidated, it is clear that toxT expression is induced through these regulators upon transition into a host, and may be affected by changes in bile concentration and/or temperature. Consistent with this, ToxR/S has been shown to also regulate the expression of V. cholerae outer membrane porins in response to bile salts (Childers and Klose 2007).

I.2.2. Quorum sensing and the regulation of virulence One pathway that helps orchestrate virulence gene expression is the quorum sensing system, which is the mechanism through which bacteria measure population density. Quorum sensing plays a large role in controlling gene expression, and thus the behavior of V. cholerae throughout its life cycle. An overview of this system and how it relates to virulence gene expression is shown in Figure I-2. The bacterium has at least two parallel quorum sensing systems, which converge at the transcription factor HapR, to detect fluctuations in the bacterial population density (Miller, et al. 2002). Expression of hapR (induced at high cell density) has been associated with repression of virulence

(Miller, et al. 2002, Zhu, et al. 2002, Kovacikova and Skorupski 2002b, Zhu and

Mekalanos 2003, Nielsen, et al. 2006). Specifically, HapR represses transcription of aphA, whose gene product is a transcription factor required for tcpPH expression (Figure

I-2) (Kovacikova and Skorupski 2002b).

In the context of a host, HapR appears to orchestrate changes in gene expression necessary for the various stages of the infection. Upon ingestion by a host, V. cholerae must pass through the gastric acid barrier in order to colonize the small intestine. It is unclear if hapR is expressed in these transitioning bacteria, however bacteria within a biofilm (where hapR is expressed) are more resistant to acid-shock and thus may be more

High Cell Density Low Cell Density

[AI-2] high [AI-1] high [AI-2] low [AI-1] low

LuxP LuxP Inner membrane Inner membrane

LuxQ CqsS LuxQ CqsS

P P LuxU + ATP LuxU LuxO LuxO ~ P

Nutrient limitation Reduced growth rate

qrr sRNAs

HapR RpoS cAMP-CRP hapR mRNA

AphA

Glycogen Ex vivo survival Chemotaxis and Bio!lm aphA tcpP biosynthesis genes Flagellar genes formation CT and TCP

Figure I-2. Quorum sensing and starvation conditions coordinate expression of many V. cholerae genes. V. cholerae are capable of monitoring their population density by production of small molecules known as auto inducers (AI-1 and AI-2) (Waters and Bassler 2005). At low cell densities, concentrations of these auto inducers are low. Under these conditions, the response regulator LuxO is phosphorylated by the Lux sensor kinases LuxQ and LuxU and the Cqs sensor kinase CqsS, leading to the production of the qrr sRNAs (Lenz, et al. 2004). The qrr sRNAs interfere with hapR mRNA translation, leading to reduced levels of this transcription factor. At high cell densities, auto inducers produced by the Lux and Cqs systems are at high concentrations. Under these conditions LuxU, LuxQ, and CqsS act as phosphatases for LuxO causing inactivation of the response regulator and lowered qrr sRNA levels which leads to increased levels of HapR. HapR has a number of regulator targets in V. cholerae, including activation of chemotaxis and flagellar genes (Nielsen, et al. 2006), as well as repression of aphA (Lin, et al. 2007). AphA is a transcription factor required for tcpPH transcription (Skorupski and Taylor 1999), which encodes a membrane-bound transcription factor that is required to activate ToxT expression (Goss, et al. 2010). ToxT is the major transcriptional activator of virulence genes, and it leads to production of cholera toxin (CT) and the toxin co-regulated pilus (TCP) (Childers and Klose 2007). Stationary phase signals at high cell density, including slowed growth rate and nutrient limitation, lead to increased levels of active CRP. In turn, CRP increases activation of HapR and induces activation of RpoS (Hengge-Aronis 2002, Silva and Benitez 2004). RpoS contributes to the activation of the chemotaxis and flagellar genes, which are hypothesized to be involved in detachment and escape from the intestinal epithelium at late time points of infection (Nielsen, et al. 2006). RpoS also regulates genes for glycogen biosynthesis (Bourassa and Camilli 2009) and genes important for survival in the environment (Schild, et al. 2007).

likely to reach the small intestine than planktonic bacteria (Zhu and Mekalanos 2003).

After passing the gastric acid barrier, such HapR-expressing bacteria entering the small intestine become more spread out, diluting the quorum sensing inducer signals

(autoinducers) and thus repressing hapR expression. Shutting off hapR allows V. cholerae to express virulence genes, which would otherwise be repressed (Miller, et al.

2002, Zhu, et al. 2002). Following colonization, V. cholerae multiplies, increasing in density until hapR is yet again turned on (Nielsen, et al. 2006), leading to repression of ctxAB and tcpA-F (Kovacikova and Skorupski 2002b). This represents our first clue as to how the transition out of the host may be regulated. The regulation of V. cholerae virulence genes by HapR appears to be evolutionarily selected for, as toxigenic strains have a higher percentage of intact quorum sensing systems than do non-toxigenic O1 or

O139 strains (Wang, et al. 2011b).

I.2.3. Cyclic diguanylate and the regulation of virulence Cyclic diguanylate (c-di-GMP), a nearly ubiquitous second messenger in bacteria, plays a large, yet not well-defined, role in transcriptional regulation of V. cholerae virulence genes. Two classes of proteins affect intracellular levels of c-di-GMP: diguanylate cyclases (DGC), which convert two GTP molecules into the second messenger, and specific phosphodiesterases (PDE), which break down the messenger into two GMP molecules. DCGs are characterized by their GGDEF domain, which is named for conserved amino acids. Likewise, PDEs harbor a conserved EAL or HD-GYP domain. As a second messenger, c-di-GMP effects expression of numerous phenotypes in bacteria including virulence gene expression, biofilm formation, and motility (Jenal and

Malone 2006, Tamayo, et al. 2007). Until recently c-di-GMP was considered to be

29 present only in bacteria as no diguanylate cyclases had been identified in eukaryotes or archea. However, identification of a DGC and characterization of the enzyme as a player in stalk cell differentiation in the eukaryotic organism Dictyostelium discoideum has changed this thinking (Chen and Schaap 2012).

In V. cholerae high levels of c-di-GMP are associated with an increase in biofilm formation and repression of virulence and motility genes (Tischler and Camilli 2004,

Tamayo, et al. 2005, Tischler and Camilli 2005, Jenal and Malone 2006, Srivastava, et al.

2013). The mechanism of c-di-GMP virulence gene regulation has not been clearly defined, but it appears to be different in classical and El Tor biotypes. In the classical biotype, deletion of the PDE VieA results in increased levels of c-di-GMP and repression of the virulence genes presumably through regulation of toxT (Tischler and Camilli

2005). VieA does not contribute to virulence regulation in El Tor (Lee, et al. 2001,

Beyhan, et al. 2006, Kamp, et al. 2013). However, high levels of c-di-GMP in the biotype still lead to repression of virulence genes through control of toxT (Tamayo, et al. 2008).

Suppression of c-di-GMP levels during infection has been proposed due to the second messengers’ effect on virulence gene expression. Late in infection and upon transition into the aquatic environment, the concentration of the second messenger is likely to again increase to promote biofilm formation and perhaps other processes. This simplistic model of c-di-GMP expression has been supported by work conducted in our laboratory (Tischler and Camilli 2005, Schild, et al. 2007, Tamayo, et al. 2008).

However, the regulation of c-di-GMP levels, and thus virulence genes, appears to be more complex. Recent work from the Waters lab has shown that c-di-GMP levels increase in response to bile, but not when bicarbonate is also present (Koestler and

Waters 2014). This work has resulted in a model in which c-di-GMP levels are high in the small intestinal lumen, where bile levels are also high, but low closer to the epithelial cells, where bicarbonate would be encountered. This model fits with the notion that virulence genes allow for colonization of the intestinal epithelial layer, but are not required for exit from the host.

I.2.4. Bistable expression of virulence factors late in infection It was previously shown, using the rabbit ligated ileal loop model of infection, that at 12 hours post-inoculation V. cholerae make a coordinated escape from the mucosal epithelial layer and swim into the intestinal lumen (Nielsen, et al. 2006).

Expanding on this, Nielsen and colleagues (2010) described a phenomenon in these detached bacteria where the expression of tcpA, which encodes the major pilin subunit of

TCP, is either low or high, i.e., the population bifurcates with some bacteria expressing virulence factors and others not. Many of the cells still expressing tcpA at the late stage of infection are localized in bacterial clumps, which fits with the idea that TCP may aid in bacteria-bacteria interactions (Childers and Klose 2007).

This bifurcation late in infection was shown to result from the bistable expression of ToxT. The bistable phenotype is dependent on a positive regulatory feedback loop between ToxT and tcpA-F (Figure I-3) (Nielsen, et al. 2010). Briefly, ToxT is a known regulator of tcpA-F and transcription from this operon can read through into the toxT gene, which lies directly downstream of tcpA-F (Yu and DiRita 1999). Protein levels within bacteria are known to vary between cells due to stochastic differences in regulation of transcription and translation, as well as degradation (Cai, et al. 2006). Thus, some V. cholerae cells will accumulate a large amount of ToxT, while others may contain

Carbon limitation Host infection signals

PTS glu TcpP/H ToxR/S Inner membrane

cya

ATP cAMP tcpPH tcpA-F toxT

toxT mRNA tcpA-F + toxT mRNA

CRP cAMP-CRP ToxT ctxAB

Figure I-3. cAMP-CRP repression and ToxT autoregulation create the bistable expression pattern of toxT. Upon transition into a host, in vivo signals, possibly including bile salts, temperature or pH, are sensed though TcpP/H and ToxR/S (Childers and Klose 2007). These transcription factors activate transcription of the toxT gene from its own promoter. Subsequently, ToxT, the major virulence regulator in V. cholerae, induces expression from the tcpA-F promoter. Read-through transcription from this operon promoter, which is directly upstream of toxT, results in an increase in toxT mRNA and ToxT protein. This generates a situation where toxT expression is amplified in a positive feedback loop, which is proposed to stochastically generate bacteria that express high levels of ToxT (Nielsen, et al. 2006). Stationary phase signals, presumably sensed late in infection, are thought to be responsible for extinguishing this regulatory cascade. Carbon limitation sensed though the phosphorylated glucose-specific PTS system (PTS-glu) stimulates adenylate cyclase (cya), which increases the pool of cAMP within the cell. Consequently, cAMP can bind and activate the catabolite repressor protein (CRP) (Skorupski and Taylor 1997). The cAMP-CRP complex directly inhibits expression of TcpP/H (Kovacikova and Skorupski 2001). As illustrated in Figure I-2, cAMP-CRP indirectly enhances expression of HapR (Liang, et al. 2007), which also acts to indirectly reduce TcpP/H expression. It is proposed that cAMP-CRP also inhibits expression from the tcpA-F promoter by an undetermined mechanism (Liang, et al. 2007, Nielsen, et al. 2010). This is hypothesized to dampen the ToxT autoregulatory circuit in a portion of the population (approximately 50%), leading to shut off of virulence gene expression (Nielsen, et al. 2010).

much less. Nielsen and colleagues (2010) propose that the positive feedback loop results in amplification of the differences in ToxT levels within the bacteria.

Activation of the catabolite repression protein (CRP) is also necessary for the bistability of ToxT (Nielsen, et al. 2010). CRP is turned on in response to limitation of favorable carbon sources, which is presumably encountered in the intestinal lumen late in infection when other high-cell density cues would be detected (Poncet, et al. 2009).

Active CRP represses tcpPH by competing with AphA at the tcpPH promoter

(Kovacikova and Skorupski 2001). This transcriptional repression at the tcpPH promoter extinguishes the ToxT autoregulatory loop. Therefore, in response to carbon starvation,

CRP appears to play a required role in producing the subpopulation of bacteria late in infection that have turned off expression of ToxT and downstream virulence factors.

The bistable phenotype can be modeled in vitro in a rich broth medium with the addition of bicarbonate, a known inducer of the virulence gene regulatory cascade

(Abuaita and Withey 2009, Nielsen, et al. 2010). Using this in vitro system, the bacterial population bifurcates, with respect to tcpA expression, upon entry into early stationary phase when carbon becomes limited. Approximately half of the cells repress expression of tcpA, whereas the other half continues to express the gene. The stable expression of tcpA is maintained for at least 3 hours in vitro after bacterial multiplication has ceased

(Nielsen, et al. 2010). Therefore, it is possible that a subpopulation of V. cholerae cells present in RWS may also be expressing virulence genes. However, contrary to this notion, a study that determined the V. cholerae transcriptome in human RWS found that the virulence genes were highly down-regulated (Merrell, et al. 2002). Thus, there is a

34 major shift in gene expression between V. cholerae acutely infecting the small intestine and those bacteria exiting the host.

The difference between these two studies could be explained by the fact that the bistable switch was observed in the rabbit ligated ileal loop, which is a closed system.

Although there is bacterial growth and fluid accumulation in the ligated ileal loop, there is no flow of intestinal contents, nor do the bacteria pass through the large intestine. Thus, it is not clear if the coordination and timing of gene expression in this model is comparable to what occurs in a human host, nor is it evident if the detached bacteria in the lumen of the ligated ileal loop are comparable to bacteria shed in human RWS.

Alternatively, the inconsistency between the two studies may be explained by their different approaches. Nielsen and colleagues (2010) looked at expression of virulence genes in individual cells, which allowed them to discover heterogeneity within the population. Conversely, Merrell and colleagues (2002) measured gene expression using microarray analysis. Although microarray data can be quite informative and powerful, these results are merely an average of all the bacteria in the population. This is a major downfall of microarrays and other population-level studies, and it certainly explains how researchers may have missed a small population of TCP expressors.

I.3. LATE STAGE OF INFECTION: PREPARING FOR THE TRANSITION Despite having turned off the expression of CT, TCP, and other ToxT-regulated virulence genes, V. cholerae cells shed in human RWS have a hyperinfectious phenotype hypothesized to aid in the rapid spread of cholera during outbreaks (Merrell, et al. 2002).

The discovery of this hyperinfectious form led to the prediction that V. cholerae undergoes additional transcriptional and/or phenotypic changes late in infection other

35 than simply turning off the known virulence genes. Indeed, this and other studies suggest that V. cholerae undergoes multiple transcriptional shifts subsequent to the acute stage of infection in the small intestine (Figure I-4) (Merrell, et al. 2002, Larocque, et al. 2005,

Nielsen, et al. 2006, Schild, et al. 2007, Nielsen, et al. 2010). For example, there is evidence that the bacteria make a coordinated exit from the intestinal surface and out of the host (Nielsen, et al. 2006). Additionally, the bacteria appear to turn on genes that are important for life in an aquatic environment prior to exiting the host (Schild, et al. 2007).

I.3.1. An RpoS and HapR mediated transition In order for V. cholerae cells to exit a host they must detach from the small intestinal epithelium and pass through the rest of the bowels. Using the rabbit ligated ileal loop model, Nielsen and colleagues (2006) described the V. cholerae ‘mucosal escape response’ during which the bacteria detach from the mucus-lined intestinal epithelium and move into the lumen. Activation of the stationary phase and stress responsive alternative sigma factor, RpoS, was shown to be necessary for this response. Specifically,

RpoS was proposed to induce motility and chemotaxis genes that would allow the bacterium to swim into the intestinal lumen (Nielsen, et al. 2006). RpoS was also shown to induce expression of the Hemagglutinin (HA) protease, a secreted enzyme that has been proposed to aid in V. cholerae detachment from the epithelial mucosal layer

(Finkelstein, et al. 1992, Yildiz and Schoolnik 1998, Nielsen, et al. 2006). Additionally,

RpoS was suggested to play a role in reducing the expression of CT and TCP, possibly through the activation of HapR and the subsequent repression of aphA by the quorum sensing regulator (Nielsen, et al. 2006).

Late: Shed in diarrhea Characterized by: -Repressed chemotaxis Middle: Exit to luminal !uid of the small -Hyperinfectivity intestine -Expression of ex vivo survival genes Early: Repilication at intestinal Characterized by: -Slowed growth (nutrient limitation?) Key regulators: epithelium -High cell density -? Luminal Characterized by: -Upregulated chemotaxis and !agellar genes !uid -Rapid growth -High expression of CT and TCP Key regulators: -RpoS Key regulators: -HapR -ToxT -cAMP-CRP Mucosal barrier

Villi (not to scale)

Figure I-4. Transcriptional changes of V. cholerae during infection. During the course of infection, V cholerae is hypothesized to undergo numerous transcriptional changes. Early: After initial attachment, bacteria express high levels of the major virulence regulator, ToxT (Withey and Dirita 2006, Childers and Klose 2007). ToxT activates virulence and colonization genes, which allow the bacterium to elaborate the toxin co-regulated pilus (TCP) and cholera toxin (CT) (Childers and Klose 2007). Additionally, these cells exhibit a fast growth rate (Nielsen, et al. 2010). Middle: As cell density increases and nutrients become limiting, HapR and cAMP-CRP are hypothesized to inhibit virulence gene expression and begin the process of exit from the small intestine (Kovacikova and Skorupski 2001, Kovacikova and Skorupski 2002, Nielsen, et al. 2010). While this occurs, RpoS and HapR enhance expression of factors for chemotaxis and motility, which allow the bacteria to escape the mucosal epithelial layer (Nielsen, et al. 2006, Schild, et al. 2007). Late: As V. cholerae is shed in RWS, chemotaxis becomes repressed and the bacteria enter a hyperinfectious state (Merrell, et al. 2002, Butler, et al. 2006, Liang, et al. 2007). During this transition from the intestinal mucosal layer to RWS, the bacteria regulates genes important for environmental survival (Schild, et al. 2007). The regulators important for this transition remain to be determined, although RpoS and PhoB may be key players in this transcriptional shift (Nielsen, et al. 2006, Schild, et al. 2007, Pratt, et al. 2009).

The role of RpoS during infection has been controversial. V. cholerae cells are likely exposed to stresses that are relevant to alternative sigma factor utilization

(Kovacikova and Skorupski 2002, Ding, et al. 2004), but the specific role of RpoS in bacterial colonization and replication is unclear. Two studies reported that RpoS is dispensable for intestinal colonization (Klose and Mekalanos 1998, Yildiz and Schoolnik

1998), while a third reported a subtle but statistically significant role (Merrell, et al.

2000). All three of these studies made use of the infant mouse model of cholera infection, and thus the variation in results may be due to differences in the V. cholerae strains used.

The strains used in the first two studies harbor natural frameshift mutations in hapR, the gene encoding the major quorum sensing regulator, while the strain used by Merrell and colleagues (2000) contains an intact hapR locus. Signaling of RpoS through HapR was reported to contribute to the V. cholerae mucosal escape response via HapR induction of several motility genes (Nielsen, et al. 2006). This may help explain why there appear to be differences in the importance of RpoS for virulence depending on the presence of a functional hapR locus. Perhaps in the absence of functional HapR signaling, the motility genes important for exit from the epithelium are induced through another pathway. The wild type strain used in this thesis harbors an intact hapR locus, and thus we expect that the mucosal escape response and the signaling interactions between RpoS and HapR, are intact.

Since RpoS is a stationary phase sigma factor, it has been proposed that, late in infection, V. cholerae may be phenotypically similar to stationary phase cells found in broth culture. In vitro, RpoS plays a role in survival of V. cholerae during carbon starvation, under hyperosmotic conditions and during oxidative stress (Yildiz and

Schoolnik 1998). It is likely that within a host, the bacterium encounters nutrient deprivation and oxidative stress, which may signal activation of rpoS.

In addition to the induction of rpoS during the mucosal escape response, there is further evidence that V. cholerae exiting a host has entered a stationary phase-like state.

Nielsen and colleagues (2006) reported that viable counts of V. cholerae isolated from a rabbit ligated ileal loop increase exponentially early in infection, but begin to level off around 6-8 hours post-inoculation, a time point that coincides with the early stages of the mucosal escape response (Nielsen, et al. 2006). This laboratory also reported that V. cholerae found in the lumen at the 8-hour time point has a severe reduction in expression of rrnB1, a gene that indicates whether a cell is actively growing (Nielsen, et al. 2010).

Though these expression data from the ligated ileal loop model may or may not correlate with expression during human infection, a separate study supports the idea of a stationary phase-like state for V. cholerae in its human host. By comparing the transcriptome of V. cholerae in vomitus (early infection) to that in RWS (late infection), LaRocque and colleagues (2005) found that genes involved in DNA repair, protein synthesis and energy production were expressed in bacteria present in vomitus, but not in the RWS bacteria.

The V. cholerae cells in vomitus are therefore believed to be derived from an actively growing population of bacteria within the upper small intestine, in contrast to RWS bacteria, which are non-growing. Additionally, it has long been known that RWS is a medium that does not support the growth of V. cholerae (Freter, et al. 1961). Taken together, these results suggest that the environment encountered by V. cholerae late in infection mimics the conditions of stationary phase culture growth, and that bacteria present in RWS may be phenotypically similar to those from an in vitro stationary phase

39 culture. This information may prove useful for designing in vitro models of the late stage of V. cholerae infection.

I.3.2. V. cholerae late genes A study by Schild and colleagues (2007) recently identified 57 V. cholerae genes that are expressed specifically at or near the end of the infection cycle in the infant mouse model of infection. Many of these ‘late’ genes appear to be involved in preparing the bacteria for the shift from the host small intestine into an aquatic environment. A number of these late genes with known or proposed functions are listed in Table I-2. During the transition to the aquatic environment V. cholerae typically undergoes numerous physiochemical shocks, including steep reductions in the following: temperature, osmolarity (when shed into fresh water), macronutrient concentration (carbon and energy sources) and micronutrient concentration (namely fixed nitrogen and phosphorus) (Table

I-1) (Schild, et al. 2007, Nelson, et al. 2008, Lara, et al. 2009). Some of the late genes aid in preparing V. cholerae for these shocks (a few examples are discussed below). The induction of these late genes makes evolutionary sense, as changing gene expression after transitioning to the environment may be too energetically costly or may simply be too late to prevent killing in the face of the numerous insults the bacterium will endure.

Genes involved in chitin binding and catabolism were identified as late genes

(Schild, et al. 2007). As will be discussed later, V. cholerae can utilize chitin, which is the major component of the exoskeletons of numerous zooplankton, as a sole carbon and nitrogen source (Huq, et al. 1983, Watnick and Kolter 2000, Lipp, et al. 2002, Broza, et al. 2008). However, there is no known source of chitin within the human gastrointestinal tract, and therefore these genes are not required for growth during infection in an infant

Table I-2. A subset of V. cholerae genes induced late in the infant mouse model of colonization.

Gene / Operon Domains Associated process Phenotype VC0130 GGDEF family Cyclic-di-GMP - protein, EAL domain metabolism (cdpA) VC0200-VC0203 iron(III) ABC Iron uptake Attenuated early and late in transporter, permease infection protein VC0404 MSHA biogenesis Biofilm formation - protein (mshN) VC1593 GGDEF family Cyclic-di-GMP Decreased survival in pond protein (acgB) metabolism and RWS VC1962 C4-dicarboxylate Carbon metabolism Decreased growth on transport, succinate transcriptional regulatory protein (dctD-1) VC2369 sensor Histidine oxygen/redox sensor Decreased survival in pond kinase (fexB) and RWS VC2697 GGDEF family Cyclic-di-GMP Decreased survival in pond protein metabolism and RWS VCA0601 ABC transporter, Unknown Decreased survival in pond permease protein and RWS VCA0612 Cellulose degradation Chitin utilization, Decreased survival in pond, product, hypothesized decreased growth on chitin phosphorylase VCA0686 iron(III) ABC Iron uptake Decreased survival in pond transporter, permease protein VCA0774 Glycerol kinase (gplK) Carbon metabolism Decreased survival in pond, decreased growth on glycerol Adapted from (Schild, et al. 2007).

mouse (Schild, et al. 2007). Therefore, induction of chitin utilization genes prior to exit from the host makes sense as a transition strategy.

Three genes encoding DGCs were identified as late infection induced genes:

VC1593 (acgB), VC2370, and VC2697. As discussed above, DGC proteins are responsible for producing intracellular c-di-GMP, which is a second messenger that is thought to be present at low levels during the early stages of infection in order to repress virulence genes (Tischler and Camilli 2005, Tamayo, et al. 2008). The identification of

DGCs as induced late in the infant mouse infection supports a theory that c-di-GMP levels increase as the bacterium prepares for exit from the host. This concept complements the finding that bile, which is present primarily in the lumen of the gastrointestinal tract, increases c-di-GMP levels in V. cholerae (Koestler and Waters

2014). However, the presumed repression of motility by high c-di-GMP levels during this transition is counter-intuitive and indeed conflicts with the observation that RWS contains highly motile bacteria, which is the basis for the rapid clinical diagnosis of cholera using dark field microscopy (Benenson, et al. 1964). It is possible that motility is under differential control at this stage, or that the V. cholerae population in RWS is bifurcated with respect to motility, since the other half of the bacteria are in clumps, as mentioned above. The question of motility aside, the importance of c-di-GMP synthesis genes in the aquatic environment is well-supported due to their role in biofilm formation as discussed above, and it makes sense that V. cholerae would have evolved to turn these genes on just prior to exiting the host.

Eleven of the late genes/operons found by Schild and colleagues (2007) are under control of RpoS although, contradictorily, seven of these are repressed by this sigma

42 factor (Nielsen, et al. 2006). This suggests that the mucosal escape response may be distinct from the late gene induction program, and that there are two waves of transcriptional changes late in infection; the first corresponds to the mucosal escape response in which RpoS is a dominant inducer, and the second corresponds with the late gene expression program. Alternatively, the split regulation of late genes by RpoS may reflect the bifurcated population observed in the rabbit ileal loop model (Nielsen, et al.

2010). Differences between the two reports could also be due to the use of different infection models; Schild and colleagues (2007) used the infant mouse model in which there is no profuse fluid accumulation, whereas Nielsen and colleagues (2006) used the rabbit ligated ileal loop model, which exhibits fluid accumulation, but is a closed system so the timing and nature of gene expression could be effected. Regardless, from the late gene study and others it is clear that much is still unknown about the environment V. cholerae faces at the late stage of infection, as well as how the environment impacts gene regulation and phenotypic expression of the organism.

I.4. V. CHOLERAE IN THE AQUATIC ENVIRONMENT Although V. cholerae is widely accepted to be a natural member of aquatic environments in temperate parts of the world, the fate of the organism upon exiting its human host is a controversial topic within the field. Transmission of V. cholerae during outbreaks may occur human-to-human within households by contact with RWS, or the bacterium may spend a short period of time within an aquatic reservoir before infecting another human. Between outbreaks, V. cholerae must persist for long periods within the aquatic niche until a host ingests it, a period of time that can last weeks to years. The phenotypic state of the bacterium during short-term or long-term persistence in the

43 aquatic reservoir is unclear, and it seems that the bacterium may have numerous strategies for survival within this harsh environment.

I.4.1. Dissemination of V. cholerae The aquatic environment is hypothesized to play a key role in the dissemination of V. cholerae within and between regions. However, the nature of this role is contentious. During outbreaks the disease is generally transmitted through ingestion of contaminated freshwater, but it has proven difficult to isolate toxigenic strains from these water sources, and even more difficult during inter-epidemic periods (Islam 1992, Aulet, et al. 2007, Grim, et al. 2010). Furthermore, V. cholerae is known to survive optimally in brackish waters with salinities of 0.5%-3%. This has led to the hypothesis that during seasonal epidemics in areas around the Bay of Bengal (primarily India and Bangladesh), the organism spreads from coastal waters through estuaries and into the freshwater sources (Figure I-5) (Miller, et al. 1982). This spread may be due to seasonal weather patterns, such as the monsoon season. Alternatively, spread from coastal to fresh waters could be the result of diseased individuals traveling from one location to another, leading to contamination of new water sources (Mosley, et al. 1968, Sant, et al. 1975, Lara, et al.

2009). In the non-seasonal epidemics that occur elsewhere in the world, such as recent outbreaks in Zimbabwe and Haiti, the origin and pattern of V. cholerae spread may be quite different.

It has long been known that infected people, both symptomatic and asymptomatic, can transport toxigenic V. cholerae long distances, even across oceans (Mosley, et al.

1968, Sant, et al. 1975). The recent cholera outbreak in Haiti has reiterated the importance of humans as a dissemination vehicle for V. cholerae. The strain causing

contaminated water Contamination of Hyperinfectious Ingestion of spread

fresh water

Persistance in estuarine Survivial in river/fresh Cholera epidemics environments (salinity > 0.1%) water environment Improper sewage treatment

Ocean Estuary River/fresh water source

Population center

Figure I-5. Dissemination and transmission of V. cholerae in endemic areas. V. cholerae persist in the environment primarily in estuarine waters (salt concentrations greater then 0.1%) (Miller, et al. 1982, Miller, et al. 1985, Alam, et al. 2007, Lara, et al. 2009). V. cholerae is hypothesized to persist long-term in biofilms associated with chitinous surfaces (Meibom, et al. 2004, Meibom, et al. 2005), or in a viable but non- culturable state (VBNC) (Alam, et al. 2007). Weather events and disruption of fresh water ecosystems by human activity (Lara, et al. 2009) can lead to contamination of fresh water reservoirs where V. cholerae can survive for at least short periods of time (Nelson, et al. 2008, Bourassa and Camilli 2009). Cholera epidemics begin with ingestion of V. cholerae from contaminated water sources. Hyperinfectious spread of freshly shed V. cholerae is hypothesized to rapidly amplify the number of infected patients during an outbreak (Merrell, et al. 2002, Hartley, et al. 2006, Pascual, et al. 2006). Improper waste management can lead to continued contamination of drinking water sources, which prolongs the epidemic.

disease in Haiti is more genetically similar to toxigenic, clinical isolates in South Asia than to those causing sporadic disease in South America (Chin, et al. 2011). Although recently isolated environmental strains in South America were not compared with the

Haitian strain, it is unlikely that the two spatially distant populations of V. cholerae

(Haitian and South Asian) convergently evolved. Similarly, it seems unlikely that water currents or weather events are responsible for transporting strains that originated in South

Asia halfway around the globe to Haiti. It is most plausible, in this era of rapid global transportation that the Haitian strain was delivered to the Caribbean island by one or more asymptomatic people traveling from South Asia. Therefore, while acknowledging the importance of weather events and seasonal patterns in the dissemination and transmission of V. cholerae, it is also important to consider the impact of human activity.

I.4.2. Nutrient starvation and strategies for survival As discussed earlier in this chapter, after transmission to the aquatic environment

V. cholerae faces abrupt changes in availability of numerous nutrients including carbon sources, phosphate, and nitrogen (Kamal, et al. 2007, Schild, et al. 2007, Nelson, et al.

2008, Lara, et al. 2009). In this section I will discuss how V. cholerae faces these challenges through acquisition of carbon and nitrogen from alternative sources.

Additionally I will discuss how the organism uses biofilm formation and a viable but non-culturable (VBNC) state to survive in the harsh aquatic environment. A separate section of this introduction is dedicated the acquisition and importance of phosphate.

I.4.2.1. Glycogen storage and utilization Work from our laboratory has begun to explore the importance of glycogen storage and utilization in V. cholerae throughout the various stages of the life cycle.

Notably, Bourassa and Camilli (2009) showed that V. cholerae cells shed in human RWS contain glycogen granules, suggesting accumulation and storage of glucose during life in the host. Glycogen is a branched glucose polymer and is used by many bacteria as a carbon and energy reserve. Under nutrient limiting conditions, bacteria can break down the glycogen stores in order to provide an internal source of carbon (Wilson, et al. 2010).

Therefore, stored glycogen in V. cholerae may provide a source of carbon in the aquatic environment.

Glycogen storage in V. cholerae is induced under carbon sufficient, but nitrogen- and/or phosphate-limiting conditions in vitro, although the signals during infection are yet to be elucidated (Bourassa and Camilli 2009). Fixed nitrogen, phosphate and carbon levels in pond water are typically growth limiting (Table I-1) (Schild, et al. 2007, Nelson, et al. 2008). This, together with the fact that V. cholerae cells exiting the human host contain glycogen granules, suggests a potential role for glycogen stores in persistence of

V. cholerae in the aquatic environment. In support of this, glycogen storage was found to be advantageous to the bacterium for survival in pond water and also in a pond-to-host transmission assay (Bourassa and Camilli 2009).

As would be expected, V. cholerae genes involved in glycogen storage, but not utilization, are induced in human RWS (Merrell, et al. 2002b, Nelson, et al. 2008).

Additionally, glgB, a gene important for glycogen synthesis (Wilson, et al. 2010), is induced during experimental V. cholerae infection of human volunteers (Lombardo, et al.

2007). When human RWS is incubated in filter-sterilized pond water, glycogen degradation genes are induced, further indicating that glycogen stores provide an internal source of carbon for the bacterium until an external source is found. Also of note, RpoS

47 appears to be involved in regulating glycogen storage, indicating that part of the mucosal escape response may be to accumulate glycogen (Bourassa and Camilli 2009). Finally, our results presented in this thesis demonstrate that phosphate is limiting during infection, further supporting induction of glycogen storage within a host.

I.4.2.2. Chitin as an alternative source of carbon and nitrogen One source of carbon and nitrogen in the aquatic environment is chitin, a polymer of N-acetylglucosamine. Chitin is the major component of the exoskeletons of zooplankton and insects found within and around many water sources (Watnick and

Kolter 2000). Chitin is produced at a rate of 1011 metric tons per year, of which 109 metric tons are produced by copepods within aquatic reservoirs (Lipp, et al. 2002).

All members of the Vibrio species that have been tested are chitinolytic, meaning they have the ability to break down and metabolize chitin (Lipp, et al. 2002). In the laboratory, Vibrio species can survive in association with a wide variety of chitin sources as the sole supply of carbon and nitrogen, including the exoskeletons of zooplankton and insects found in fresh, brackish and marine waters (Huq, et al. 1983, Watnick and Kolter

2000, Lipp, et al. 2002, Broza, et al. 2008). Some Vibrio species have been reported to exhibit behavior of both chitin attachment and detachment, suggesting that they are able to remove themselves from a chitinous surface when all or most of the carbon and nitrogen has been consumed or when other signals, such as population density, trigger dispersal (Lipp, et al. 2002). In addition, V. cholerae exhibits the ability to chemotax towards chitin sources (Meibom, et al. 2004). As mentioned earlier, several chitin attachment and metabolism genes are V. cholerae late genes. Mutants of some of these genes showed a defect in their ability to compete with wild type V. cholerae in a host-to-

48 pond transition assay (Table I-2) (Schild, et al. 2007). Thus, chitin appears to be an important, perhaps the most important, source of carbon, and potentially nitrogen, for V. cholerae in the aquatic reservoir.

I.4.2.3. Biofilm formation of V. cholerae and importance of the phenotype Chitinous exoskeletons found in the aquatic environment not only provide a source of carbon and nitrogen for V. cholerae, but they also provide a surface on which the organism can form a biofilm (Watnick, et al. 1999, Watnick and Kolter 2000).

Biofilms are complex microbial communities that have a three-dimensional structure often consisting of pillars and fluid-filled channels, which are surrounded by bacterially secreted exopolysaccharides and released DNA. These microbial communities attach themselves to a variety of biotic and abiotic surfaces, and can be found in aquatic reservoirs, generally associated with zooplankton, insects or exoskeletons and other detritis. In the laboratory, biofilms are generally comprised of one species in order to simplify experiments, however in the natural environment they may contain a variety of different bacteria. Bacteria within a biofilm may be protected from numerous insults including toxic chemicals, antibiotics, thermal stress, oxidative stresses, UV radiation, bacteriophages and other predators. This makes the biofilm structure an ideal survival mechanism for many aquatic-dwelling bacteria that often face these insults (Watnick and

Kolter 2000, Huq, et al. 2008).

Formation of biofilms in aquatic environments is considered a developmental process since there are several distinct steps to formation of a mature biofilm during which the bacteria undergo numerous changes in gene expression (Watnick and Kolter

2000, Huq, et al. 2008). In many bacteria, including V. cholerae, quorum sensing plays a

49 major role in controlling biofilm formation (Huq, et al. 2008). At low cell density, motile bacteria attach to a surface to form a sparse monolayer. Multiplication on the surface fills in the monolayer. Additional multiplication, coupled with exopolysaccharide production, builds up the mature three-dimensional structure of the biofilm (Moorthy and Watnick

2004). As the biofilm matures, cell density provides a signal to stop building additional biofilm (Figure I-2). In V. cholerae, the quorum sensing regulator, HapR, represses biofilm formation through two distinct mechanisms. HapR decreases the level of the biofilm-promoting second messenger c-di-GMP (Waters, et al. 2008). HapR also represses expression of vpsT (Waters, et al. 2008), which encodes the positive transcriptional regulator of the Vibrio exopolysaccharide synthesis genes (Casper-Lindley and Yildiz 2004). VpsT is itself regulated post-translationally by c-di-GMP binding

(Krasteva, et al. 2010).

Other factors required for V. cholerae biofilm formation vary depending on the strain and conditions of the assay (Watnick, et al. 1999, Meibom, et al. 2004, Moorthy and Watnick 2004). The mannose-sensitive hemagglutinin (MSHA) pilus, which binds chitin and potentially other surfaces, is the key player in a model of biofilm formation proposed by Moorthy and Watnick (Moorthy and Watnick 2004). In this model, V. cholerae proceeds through three stages en route to a mature biofilm: 1) planktonic – free- swimming bacteria expressing flagella and surface attachment pili, 2) monolayer – a single layer of bacteria transiently attached to a surface via the MSHA pilus, and 3) the biofilm – a three-dimensional structure supported by an exopolysaccharide. The authors show that gene expression profiles of the three stages are different, supporting their idea of three distinct phases. They suggest that the MSHA pilus is required for monolayer

50 formation, both mediating attachment to a surface and suppressing expression of flagellar genes to prevent detachment (Moorthy and Watnick 2004). The fact that mshN, a gene in the MSHA pilus operon, is known to be expressed late in infection (Schild, et al. 2007), suggests that the bacteria may begin to elaborate the type IV pilus prior to exiting the host, possibly in preparation for future biofilm formation.

The requirement of the MSHA pilus and chitin in biofilm formation is not absolute, as V. cholerae can bypass the MSHA pilus/monolayer step under certain laboratory conditions (Watnick, et al. 1999) and biofilms can be formed on surfaces other than chitin (Watnick, et al. 1999, Moorthy and Watnick 2004). However in nature, monolayer formation on chitinous surfaces, aided by the MSHA pilus, may well be a necessary stage in maturation of a biofilm. One environmental trigger necessary for biofilm formation is the presence of monosaccharides (Kierek and Watnick 2003), which are normally in low concentrations in aquatic environments (Nelson, et al. 2008). The monolayer stage would give V. cholerae cells the opportunity to break down some underlying chitin to obtain the monosaccharides necessary for the next stage of biofilm formation.

In addition to providing shelter from a variety of environmental insults, biofilms may contribute to the evolution of bacteria. Rates of bacterial conjugation and exchange of extrachromosomal DNA are accelerated within biofilms (Watnick and Kolter 2000). In

V. cholerae, it was reported that HapR, in the presence of chitin, can mediate the expression of comEA, genes required for bacterial competence (Meibom, et al. 2005).

HapR is active when the population density is high, suggesting that V. cholerae competence is a phenotype of mature biofilms. Exchange of DNA within a biofilm may

51 promote the emergence of new toxigenic strains of V. cholerae via the horizontal transfer of the O-antigen genes, VPI-1, CTXΦ, or other loci. Indeed, natural transformation may be the only way for the VPI-1 to be transferred among strains, as self-transmissibility appears to be lacking (Faruque, et al. 2003).

Biofilm-associated V. cholerae may play a significant role in transmission of the bacterium to a human host. A potential role of biofilms in bacterial transmission is not without precedent. Drinking water contaminated with biofilms of E. coli, Aeromonas, and

Pseudomonas species are a potential threat to humans if ingested. Clumps of biofilm found in water have been reported to contain upward of 109 cells (Huq, et al. 2008). This is a significant number considering that the ID50 of in vitro grown V. cholerae is reported to be between 108-1011 bacterial cells (Cash, et al. 1974). In addition, a study conducted in Bangladesh found a 50% reduction in cholera cases when folded sari cloth was used to filter drinking water. The folded cloth was able to move debris larger than ~20 µm, which includes most biofilm-associated V. cholerae, but not planktonic cells or smaller clumps of bacteria (Huo, et al. 1996). Studies conducted in the lab have shown that V. cholerae cells associated with biofilms are more infectious than planktonic cells (Huq, et al. 2008,

Tamayo, et al. 2010). This phenotype is likely due to the physiological state of the bacteria (Tamayo, et al. 2010), but it remains a possibility that biofilms on chitinous surfaces may also serve to concentrate V. cholerae cells on suspended particles in water, resulting in a higher rate of infection when ingested. These two phenomena are not mutually exclusive, and either or both may be critically important to the ability of V. cholerae to initiate outbreaks from aquatic reservoirs.

I.4.3.4. Viable but non-culturable state One environmental survival behavior that several enteric bacteria, including V. cholerae, have been reported to exhibit is entrance into a viable but non-culturable

(VBNC) state (Trevors 2011) or, more conservatively, active but non-culturable (ABNC) state (Kell, et al. 1998). VBNC V. cholerae, which are generally coccoid, are defined as viable cells unable to be cultured on standard media. These cells can be enumerated in the laboratory using epifluorescent microscopy, and their viability (or activity) may be in some cases determined by the addition of yeast extract and nalidixic acid, which induces the metabolically active coccoid cells to elongate (Alam, et al. 2007, Aulet, et al. 2007,

Trevors 2011). The VBNC state is thought to function to protect bacteria from environmental insults and allow them to endure long periods of nutrient starvation. In the aquatic environment the VBNC state may protect bacterial cells from fluctuations in pH, salinity, temperature and nutrient availability. In the laboratory a VBNC state can be induced by altering many of these variables, although nutrient limitation is most commonly used (Trevors 2011).

The VBNC phenotype has been demonstrated in several pathogenic strains of V. cholerae, including V. cholerae O1 (classical and El Tor biotypes) and O139 (Colwell, et al. 1996, Baffone, et al. 2003, Alam, et al. 2007, Aulet, et al. 2007, Grim, et al. 2010,

Trevors 2011). It has been proposed that between cholera outbreaks, toxigenic V. cholerae may exist in a quiescent VBNC state in bodies of freshwater. In support of this,

VBNC V. cholerae have been isolated from freshwater sources around the world (Aulet, et al. 2007, Broza, et al. 2008, Grim, et al. 2010). Also, the V. cholerae VBNC phenotype can be induced in the laboratory with prolonged incubation in pond water at both 4°C and

53 room temperature (Colwell, et al. 1996, Alam, et al. 2007). However, the genetic components that contribute to the VBNC state have yet to be elucidated.

The epidemiological importance of VBNC V. cholerae is not yet apparent. The ability of VBNC cells to cause disease depends heavily on their ability to “resuscitate” into fully active, pathogenic bacteria. Although the triggers necessary for resuscitation are largely unknown, the ability of VBNC cells to revert to an infectious form upon passage through a host has been demonstrated in several pathogenic bacteria including toxigenic V. cholerae (Colwell, et al. 1996, Baffone, et al. 2003, Alam, et al. 2007). One of these studies reported that although V. cholerae O1 was resuscitated upon passage through the rabbit ligated ileal loop, there was no appreciable fluid accumulation in the ileal loops, suggesting that the VBNC cells are not as infectious as culturable cells (Alam, et al. 2007). Despite these reports, other researchers have argued that the appearance of resuscitated cells upon passage through an animal is merely the result of rare, culturable cells in the inoculum (Bogosian, et al. 1998, Kell, et al. 1998). Kell and colleagues

(1998) propose the use of strict statistical analysis of resuscitation data, something which is rarely done, taking into account the probability that any culturable cell existed in the inoculum. Until the resuscitation trigger is identified and the genetic mechanisms described, it will be hard to fully appreciate VBNC V. cholerae as epidemiologically important.

I.4.4. Transmission to a new host and hyperinfectivity Cholera has been endemic in Southeast Asia throughout recorded history (Barua

1992). Around the Bay of Bengal there are seasonal outbreaks (Lipp, et al. 2002) that appear to originate near coastal waters and travel inland over the course of the outbreak

(Broza, et al. 2008). Studies conducted in Bangladesh near the Karnaphuli estuary have shown that weather events and human disruption of the estuary environment can lead to contamination of river water further upstream from where Vibrio species are usually detected (Lara, et al. 2009). It is possible that these events, which can lead to contamination of fresh water sources for population centers, contribute to the beginning of seasonal outbreaks (Figure I-5). In Bangladesh, the region where most V. cholerae ecological research has been conducted, the outbreaks occur in a bimodal pattern: March-

June and September-December. This pattern correlates with blooms of zooplankton, insect and algae species that V. cholerae has been reported to associate with (Islam 1992,

Epstein 1993, Lipp, et al. 2002). Such blooms may contribute to transmission by both providing a carbon source in the environment for V. cholerae to grow on and by generating biofilms, thereby enhancing virulence and concentrating the infectious dose of bacteria.

During an outbreak of cholera, the number of cases in an area can increase explosively. This observation does not fit a model of cholera transmission that relies solely on ingestion of V. cholerae of low infectivity (Hartley, et al. 2006). Rather, the mode of V. cholerae transmission may be dominated by spread of the hyperinfectious form of the organism that is shed in the RWS of cholera patients (Merrell, et al. 2002,

Hartley, et al. 2006). The hyperinfectivity phenotype can be modeled in the laboratory by passaging V. cholerae cells through the infant mouse small intestine (Alam, et al. 2005).

Hyperinfectious V. cholerae cells isolated from either human RWS or infant mice are on average 10-100 fold more infectious than in vitro-grown bacteria (Merrell, et al. 2002,

Alam, et al. 2005). In a mathematical model that describes the number of cholera cases

55 that occur during an outbreak, the trend more closely fits the explosive spread of the disease that is described in epidemiological data when the hyperinfectious form of V. cholerae is incorporated (Hartley, et al. 2006).

Hyperinfectious bacteria are thought to contribute to human-to-human spread of

V. cholerae. This mode of transmission may be important for household spread of the disease (Harris, et al. 2008, Weil, et al. 2009, Kendall, et al. 2010). Supporting this, it has been shown that individuals who reside within the same household as an index case are at much greater risk of contracting cholera than individuals living in the same community.

Furthermore, the household members that contract cholera typically present with symptoms 2-3 days after the index case began shedding (Weil, et al. 2009), which is within the range of the incubation period of V. cholerae in humans. In addition to household spread via direct contact with RWS, the hyperinfectious bacteria may persist for a short period of time in water sources before being ingested by a new host.

Hyperinfectious V. cholerae have been shown to maintain the phenotype for at least 5 hours, but not more than 18 hours, in pond water (Merrell, et al. 2002). This suggests a transient nature of hyperinfectivity, and helps to explain why cholera outbreaks exhibit a seasonal pattern and are not without end.

The hyperinfectious phenotype goes a long way toward explaining the explosive nature of cholera outbreaks, however, much is still unknown about how the phenotype is induced or maintained in V. cholerae. As discussed below, several studies have reported means of artificially inducing hyperinfectivity or have found potential genes involved in the phenotype, but it is unclear if these are artifacts of laboratory manipulation or if they truly mimic the hyperinfectious form found in nature. Additionally, it is unknown if

56 hyperinfectivity is established via one pathway, or if there are multiple ways a bacterium can become hyperinfectious.

Acid-tolerized V. cholerae has been shown to exhibit a hyperinfectious phenotype in the infant mouse model of infection (Merrell and Camilli 1999). The increased infectivity of acid-tolerized bacteria was suggested to result from these bacteria being better able to survive passage through the gastric acid barrier, thus allowing more bacteria to reach and colonize the small intestine. However, a later study ruled out this mechanism by showing that non-acid-tolerized bacteria survive just as well, and that instead, the acid-tolerized bacteria multiply faster once they have colonized the small intestine

(Angelichio, et al. 2004). Therefore, this suggests that some aspect of the metabolic state of acid-tolerized bacteria may contribute to their hyperinfectivity.

Evidence that suppression of chemotaxis, but not motility, contributes to hyperinfectivity has been presented in several studies from our laboratory (Merrell, et al.

2002, Butler and Camilli 2004, Butler, et al. 2006). Chemotaxis genes are repressed in human RWS V. cholerae cells compared with in vitro grown bacteria (Merrell, et al.

2002), despite both populations of bacteria being motile. Correspondingly, bacteria isolated from human RWS were shown to be less able to chemotax to amino acid chemoattractants than in vitro-grown bacteria (Butler, et al. 2006). One possible explanation for the increased infectivity of motile, but non-chemotactic, bacteria is their ability to colonize the entire small intestine, whereas chemotactic strains colonize primarily the distal portion of the organ (Butler and Camilli 2004). Activation of chemotaxis genes late in infection via RpoS was shown to be important for the mucosal escape response (Nielsen, et al. 2006). This suggests that transcriptional changes occur

57 after the mucosal escape response in order to repress the chemotactic phenotype prior to exit from the host. Alternatively, the repression of chemotaxis may occur within a subset of cells exiting the host, similar to the bistable expression of tcpA described by Nielsen et al. (2010).

I.5. THE PHOSPHATE STARVATION RESPONSE IN V. CHOLERAE Phosphate is a fundamental nutrient for life being essential for the formation of various macromolecules, such as RNA and DNA, and for protein signaling cascades.

Indeed, phosphate is proposed to comprise 3% of the dry weight of all living organisms

(White and Metcalf 2007). Due to the importance of phosphate, bacteria have evolved several mechanisms to acquire it from the environment. Microbial organisms are known to use a wide variety of phosphate compounds including inorganic phosphate, organic phosphate compounds, and reduced phosphate compounds (White and Metcalf 2007).

The following section is a discussion of the many forms that phosphate takes in the environment as well as some of the ways bacteria have evolved to assimilate the nutrient.

I have focused primarily on the E. coli literature as the phosphate starvation response is very well described in the bacterium.

I.5.1. Phosphate acquisition and storage systems in bacteria including V. cholerae I.5.1.1. Environmental sources of phosphate Inorganic phosphate is the most readily available form of phosphate in the aquatic environment (White and Metcalf 2007). Reports of concentrations of inorganic phosphate in the aquatic environment range from micromolar to millimolar quantities (Kamal, et al.

2007, Schild, et al. 2007, Nelson, et al. 2008). These differences in concentration may be due to gross differences between the studies such as global location of the water sources

58 tested, seasonal differences in sample acquisition, or testing of fresh versus brackish waters. Alternatively, these differences may reflect slight differences in sampling methods; samples with higher apparent inorganic phosphate concentrations may contain more contamination from aquatic organisms. Three forms of inorganic phosphate exist in the environment: orthophosphate (Pi), polyphosphate (PolyPi), and pyrophosphate (PPi).

Most bacteria, including V. cholerae, have several dedicated Pi transporters. A detailed discussion of inorganic phosphate uptake systems follows this section.

Bacteria are able to utilize numerous organophosphates, which are characterized by a phosphorous-oxygen-carbon ester bond (e.g. sugar phosphates), as sources of phosphate. Although many organophosphates can cross the outer membrane of Gram- negative bacteria, they cannot be transported into the cytoplasm of cells with the exceptions of glycerol-3-phosphate and hexose-6-phosphates (van Veen 1997, Lamarche, et al. 2008). To utilize Pi from the non-transportable organophosphate compounds, the Pi group must be removed prior to transport by an inorganic Pi transporter. For example, V. cholerae alkaline phosphatase (PhoX) is able to remove the Pi group from several organic phosphate compounds including glucose-6-phosphate, glucose-1-phosphate, and β- glycerophosphate (Roy, et al. 1982). Additionally, some bacteria possess phosphatases specific to particular sugar phosphates, such as the glucose-1-phosphatase of E. coli, encoded at the agp locus (Pradel and Boquet 1988). Presumably once the Pi has been removed from the organophosphate both the sugar and Pi groups are transported into the cytoplasm for use as Pi and carbon sources.

Dedicated transporters of glycerol-3-phosphate and/or hexose-6-phosphates have been described in several organisms including E. coli, Chlamydia pneumoniae,

Lactococcus lactis (formerly Streptococcus lactis), Staphylococcus aureus, and

Salmonella enterica serovar Typhimurium (Kadner 1973a, Kadner and Winkler 1973b,

Rao and Torriani 1990, Island, et al. 1992, van Veen 1997, Schwoppe, et al. 2002).

Glycerol-3-phosphate transport in E. coli is mediated by two independent systems: GlpT and Ugp (Rao and Torriani 1990). However, it is unlikely that the GlpT system in E. coli contributes to acquisition of phosphate under Pi limiting conditions. Two lines of evidence support this hypothesis: 1) transport of glycerol-3-phosphate into the cell by

GlpT requires antiport of Pi (Elvin, et al. 1985) and, 2) the GlpT system is transcriptionally regulated by GlpR, not PhoB, and is activated by extracellular glycerol-

3-phosphate, not limiting Pi concentrations (Lin 1976, Elvin, et al. 1985). In contrast to this, the E. coli Ugp system, which relies on ATP-hydrolysis to drive transport of the sugar phosphate, is induced under Pi starvation conditions by direct control by PhoB

(Schweizer and Boos 1985, Overduin, et al. 1988, Hekstra and Tommassen 1993).

Additionally, increased internal, but not external, Pi inhibits Ugp activity (Brzoska, et al.

1994). Interestingly, the Ugp system is also controlled by carbon starvation via cAMP-

CRP regulation, suggesting that Ugp contributes to use of glycerol-3-phosphate as a carbon and Pi source (Kasahara, et al. 1991). Hexose-6-phosphates are transported in E. coli via UhpT, which much like the GlpT transporter utilizes antiport of Pi (Ambudkar, et al. 1986). Transcription of uhpT is controlled by a two-component-like system UhpABC, which senses periplasmic glucose-6-phosphate and induces uhpT via activation by UhpA

(Weston and Kadner 1988). In E. coli UhpT-mediated transport of glucose-6-phosphate does not allow for use of the sugar phosphate as a source of Pi for the bacterium, presumably due to the induction of alkaline phosphatase by the phosphate starvation

60 response, which can break down the sugar phosphate in the periplasm (Hoffer, et al.

2001).

Although V. cholerae harbors the genes for GlpT and Ugp, they have not been studied or confirmed for transport activity. However, a proteomic study of V. cholerae identified UgpB peptide in Pi starved wild type cells, suggesting that V. cholerae can use glycerol-3-phosphate as a source of Pi like E. coli (von Kruger, et al. 2006). To my knowledge, no one has studied this phenotype. Conversely, uptake of hexose-6- phosphates by a UhpT-like system in V. cholerae has been recently described. Unlike the antiport system of E. coli UhpT, the V. cholerae hexose-6-phosphate transporter is driven by ATP-hydrolysis such that no Pi is lost from the intracellular pool. Indeed, uptake of hexose-6-phosphate by V. cholerae was shown to promote survival under low Pi conditions (Moisi, et al. 2013). Interestingly, the V. cholerae hexose-6-phosphate transporter is not induced under low Pi conditions, but it does appear to be carbon- catabolite repressed through a cAMP-CRP independent manner (von Kruger, et al. 2006,

Moisi, et al. 2013).

Phosphonates are an alternative form of organophosphorous that some bacteria are able to utilize. These compounds are characterized by their chemically inert and hydrolytically stable carbon-phosphorous bond (White and Metcalf 2007, Kamat and

Raushel 2013). Many bacteria are able to utilize phosphonate compounds as sources of

Pi. To accomplish this, the compound is generally transported into the cytoplasm and broken down by one of many enzymes able to hydrolyze the C-P bond (Villarreal-Chiu, et al. 2012). V. cholerae transport and utilization of phosphonate has not been studied to

61 our knowledge, however, the organism does encode a putative phosphonoacetate hydrolase.

I.5.1.2. Transport of inorganic phosphate

Most bacteria encode two independent systems for the uptake of Pi from the environment: inorganic phosphate transport system (Pit) and phosphate-specific transport system (Pst/PhoU). Pit is a low affinity (Km = 25 to 38.3 µM in E. coli) transporter of Pi

(Rosenberg, et al. 1977, Willsky and Malamy 1980). Use of the ionophore, CCCP, revealed that Pit takes advantage of the proton motive force to drive transport across the inner membrane (Rosenberg, et al. 1977). E. coli encodes two Pit transporters, PitA and

PitB. PitB is induced by low phosphate conditions, whereas PitA is constitutive (Harris, et al. 2001). The substrate of Pit transport is metal phosphates (e.g. MgHPO4 and

CaHPO4) (van Veen, et al. 1993, van Veen, et al. 1994). Although a V. cholerae Pit system has not been identified, the organism does contain a gene that encodes a protein with homology (37% identity) to the E. coli PitA. Additionally, this putative PitA transporter was not identified as a phosphate starvation induced protein, however this work needs to be further confirmed (von Kruger, et al. 2006).

In contrast to Pit, the Pst/PhoU system represents a high affinity Pi transporter

(Km = 0.2 to 0.4 µM in E. coli) (Rosenberg, et al. 1977, Willsky and Malamy 1980). The

Pst/PhoU system functions as an ABC transporter comprised of multiple proteins: PstS,

PstC, PstA, PstB, and PhoU (Rao and Torriani 1990, van Veen 1997, Lamarche, et al.

2008). PstS is a periplasmic Pi binding protein that is responsible for shuttling Pi to the transport system (Medveczky and Rosenberg 1969, Willsky and Malamy 1976, Kubena, et al. 1986). PstC and PstA comprise an integral membrane complex and are both

62 considered important for Pi to transverse the inner membrane (Cox, et al. 1988, Cox, et al. 1989). PstB provides energy for transport of Pi by ATP-hydrolysis within the cytoplasm (Chan and Torriani 1996). Although PhoU is encoded in an operon with the pst genes, the protein does not contribute to Pi transport; the role of PhoU will be discussed below.

In E. coli, the genes encoding the Pst/PhoU system are present in a single operon

(pstSCAB-phoU). Conversely, V. cholerae encodes the Pst/PhoU system in two operons located on the large chromosome, which are separated by two genes (ppk and ppx) that are divergently transcribed from the pst genes. Additionally, V. cholerae encodes a second set of pst genes on the second, small chromosome (Figure I-6). In both organisms the pst genes are induced by low extracellular phosphate through the action of a two component system, PhoB/R, that will be discussed below.

Recently, a sodium-dependent phosphate cotransporter, NptA, was identified in V. cholerae (Lebens, et al. 2002). This transporter belongs to a class of eukaryotic Pi transporters, and no other example of this type has been described in bacteria (Werner and Kinne 2001, Virkki, et al. 2007). The V. cholerae NptA is most closely related to the sodium-dependent phosphate cotransporter from Caenorhabditis elegans. Since the gene is present in both toxigenic and nontoxigenic strains of V. cholerae, the bacterium is thought to have obtained the gene through horizontal gene transfer with a marine- dwelling eukaryotic organism. Alternatively, since the G/C content of the nptA gene is

A) V. cholerae Pst1

phoB phoR pstS1 ppx ppk pstC1 pstA1 pstB1 phoU

B) V. cholerae Pst2

VCA0068 pstS2 pstC2 pstA2 pstB2 VCA0074

C) E. coli Pst

bglG phoU pstB pstA pstC pstS glmS

Figure I-6. Comparison of pst operons in E. coli and V. cholerae. The pst genes are organized differently in V. cholerae and E. coli. In V. cholerae there are two regions of the genome that encode pst genes. A) Unlike the other pst regions, pst1 of V. cholerae is split in half by the genes encoding exopolyphosphatase (ppx) and polyphosphate kinase (ppk). pstS1 is transcribed in an operon with the two component system phoBR, while pstCAB-phoU is transcribed as a separate operon. B) Pst2 of V. cholerae is encoded on the second, smaller chromosome of the bacterium. The operon only contains pstSCAB with no phoU homolog. C) E. coli pst-phoU genes are present in a single, uninterrupted operon. phoBR genes are more than 300 open reading frames away from this operon.

similar to that of the V. cholerae genome, it is possible that the gene underwent convergent evolution in the Vibrio spp. (Lebens, et al. 2002). Although this scenario seems less likely than the former, due to the relatively high relatedness of the two proteins (30% identical). The biological importance of NptA in V. cholerae has not beendescribed, however it is not important for survival in the infant rabbit or pond water, nor was the gene identified as an in vivo induced gene (Lombardo, et al. 2007, Schild, et al. 2007, Kamp, et al. 2013).

I.5.1.3. Polyphosphate storage and breakdown Under phosphate starvation conditions numerous microbial species are able to acquire phosphate from an internal source: PolyP. As its name suggests, PolyP is a Pi polymer, which can range in size from 3 to over 300 residues. The Pi chain is synthesized from ATP through the action of polyphosphate kinase, encoded by ppk. Once synthesized, the PolyP chains are stored within the cell, generally in cytoplasmic granules. PolyP has been implicated in numerous phenotypes of bacteria including survival of various stresses, virulence, biofilm formation, and quorum-sensing. Of relevance here, PolyP is known to serve as a source of Pi under phosphate starvation conditions. Break down of PolyP into orthophosphate is facilitated by exopolyphosphatase, encoded by ppx (Kulaev and Kulakovskaya 2000).

V. cholerae is known to store large amounts of PolyP, which can be visualized by electron microscopy as electron dense bodies. Similar to other organisms, V. cholerae synthesizes PolyP using the kinase Ppk when grown under Pi replete conditions (Ogawa, et al. 2000). The role of PolyP storage in V. cholerae depends on the strain assayed. One ppk mutant El Tor strain exhibited a decrease in motility and attachment to abiotic

65 surfaces (i.e. decrease in the ability to form biofilms) (Ogawa, et al. 2000), while a ppk mutant in a different El Tor strain displayed no defect in these phenotypes (Jahid, et al.

2006). As the conditions used in the latter study were not described, it is unclear if the conflicting data are due to strain or experimental design differences. However, it is clear that PolyP contributes to the ability of V. cholerae to survive environmental stresses. The stores of PolyP were shown to be important for survival after a shift from a nutrient rich broth (2xYT) to a minimal medium (MOPS buffered + 0.4% glucose + 2mM Pi) (Ogawa, et al. 2000). Additionally, PolyP was shown to be important for survival of exposure to high salt, low pH, and hydrogen peroxide insults specifically under low Pi conditions

(Jahid, et al. 2006). Since the activity of Ppx was 100-fold less than that for E. coli, and nearly undetectable, it is unclear if V. cholerae is able to utilize PolyP as a source of Pi

(Ogawa, et al. 2000). However, since the Ppx protein was isolated from a high Pi grown bacterium, it is possible that higher activity would be seen for the protein after growth in

Pi -depleted conditions, when the enzyme is presumably required.

I.5.2. The Pho regulon

Upon transfer to a Pi -depleted environment (e.g. pond water), V. cholerae induces the phosphate starvation response. As part of the response, V. cholerae induces expression of Pi acquisition genes (e.g. pst and phoX), which comprise the Pho regulon.

Below is a discussion of the Pho regulon and what is known about the system in V. cholerae.

I.5.2.1. The two-component system PhoB/R and control by PhoU The Pho regulon is controlled by a two-component system, PhoB/R. PhoR is a membrane-bound autophosphorylating histidine kinase, which transfers its phosphate to

66 an aspartate residue on the response regulator, PhoB, under phosphate limiting conditions

(Makino, et al. 1985, Makino, et al. 1986, Makino, et al. 1989, Lamarche, et al. 2008).

This two-component signal transduction system is regulated by the Pst/PhoU system, whose role as a Pi transporter was discussed above. Under high phosphate conditions,

Pst/PhoU suppresses autophosphorylation of PhoR and thus represses the Pho regulon

(Lamarche, et al. 2008). PhoR activation is bimodal within a population, such that each cell appears to demonstrate an all-or-none response (Zhou, et al. 2005).

PhoU, which is not required for transport of Pi, is essential for the Pst/PhoU control over PhoR activity (Steed and Wanner 1993). However, the mechanism of this control is unclear. Recently it was demonstrated through both co-elution experiments and a bacterial-two-hybrid experiment that PhoU can bind both PstB (the ATPase component of the Pst/PhoU system) and PhoR (Gardner, et al. 2014). The roles of Pst/PhoU as a Pi transporter and a regulator of the Pho regulon can be separated, as a point mutation in

PstA abrogates Pi transport, but not activity of PhoB (Cox, et al. 1988). Regardless of its functionality, formation of the Pst/PhoU complex appears to be essential for control of

PhoB/R as null mutations in the Pst/PhoU system of V. cholerae and E. coli result in a constitutively active PhoB (Willsky, et al. 1973, Lamarche, et al. 2008, Pratt, et al.

2010).

The Pho regulon may respond to other cues in addition to phosphate levels.

Phosphorylation of PhoB by six different noncognate histidine kinases has been reported

(Wanner and Latterell 1980, Zhou, et al. 2005). The physiological importance of this surprising result is unknown and needs to be studied further. The Pho regulon has also been connected with stress response in many organisms including Citrobacter rodentium,

E. coli, and V. cholerae (Baek and Lee 2006, Von Krueger, et al. 2006, von Kruger, et al.

2006, Crepin, et al. 2008, Cheng, et al. 2012, Lery, et al. 2013). Therefore, there may be a connection between RpoS and PhoB regulated genes. This idea suggests that PhoB may be a major player in regulating V. cholerae late genes or mucosal escape genes.

I.5.2.2. Genes and programs induced by PhoB The Pho regulon has been heavily studied in E. coli. In this organism, PhoB induces over 200 genes/proteins in response to low phosphate (VanBogelen, et al. 1996).

This means that approximately 5% or more of the E. coli genome is altered during either phosphate limitation or activation of PhoB. The Pho regulon has not been fully described in V. cholerae, but a limited proteomic studied confirmed that there is much overlap between the E. coli and V. cholerae regulons (von Kruger, et al. 2006).

Most of the genes induced by phosphate limitation are involved in Pi assimilation from the environment. For example, in E. coli, PhoB induces transcription of the pst genes, pitB, ugbB, and phoA (von Kruger, et al. 1999, Harris, et al. 2001, Crepin, et al.

2008, Willsky and Malamy, 1976). With the exception of pitB the homologous protein products of these genes were also identified as induced in a low phosphate, PhoB- dependent manner in V. cholerae (von Kruger, et al. 2006). However, as mentioned above, a large number of genes identified as part of the Pho regulon are also characterized as stress response genes. Other than nutrient limitation, these genes have no obvious tie to phosphate homeostasis. As many of the genes identified as part of the Pho regulon have not been confirmed for direct regulation by PhoB, it is possible that much of the apparent overlap between the Pho regulon and stress response genes is due to induction of a general stress response to the nutrient starved condition.

I.5.2.3. Regulation of genes by PhoB Phosphorylation of PhoB by PhoR, and subsequent dimerization of the response regulator is required to activate the protein (McCleary 1996). PhoB binds to DNA at a particular sequence (termed Pho box) that has been described as two direct repeats separated by a 4-bp AT rich region: CTGTCAT-A(AT)A(TA)-CTGT(CA)A(CT)

(Makino, et al. 1993, Blanco, et al. 2011). Use of in silico promoter searches for this Pho box motif has led to successful identification of Pho regulon genes in E. coli (Yuan, et al.

2006, Chekabab, et al. 2014). Unlike E. coli, the interaction of PhoB with promoter sites in V. cholerae has not been well described. Three Pho boxes of high similarity to the E. coli Pho box were identified within the V. cholerae phoBR promoter and PhoB was demonstrated to bind these sequences (Diniz, et al. 2011). However, V. cholerae PhoB also interacts with a region of DNA within the tcpPH promoter that does not contain a putative Pho box (Pratt, et al. 2010). In this thesis I have described another PhoB- regulated gene, which does not harbor a cononcial Pho Box. Indeed my work suggests that PhoB directly binds to this regulator as well. Further research, perhaps using a combination of promoter-pull downs and RNAseq, will help elucidate how PhoB interact with the promoters of V. cholerae Pho regulon genes.

I.5.3. Importance of the phosphate response of V. cholerae during infection Dynamic activity of PhoB/R appears to be important during infection, as deletions in Pst/PhoU or PhoB/R highly attenuate virulence (von Kruger, et al. 1999, Merrell, et al.

2002, Lamarche, et al. 2008). The former is explained by the discovery that PhoB is a repressor of the virulence genes late in infection, mediated through repression of tcpPH

(Pratt, et al. 2010). Further evidence that PhoB/R is active late in infection is that genes in the V. cholerae Pho regulon are induced in human RWS (Nelson, et al. 2008) despite 69 the fact that RWS contains high levels of phosphate (Table I-1). The work completed in this thesis demonstrates that PhoB is also induced at an early time point in infection of an infant mouse (McDonough, et al. 2014). Thus the activity of PhoB may be high at both the early and late stages of infection. Further work is needed to determine whether the activity of PhoB is dynamic during infection or if it remains high throughout the infection.

I.6. DNA AS AN ALTERNATIVE SOURCE OF CARBON, NITROGEN, AND PHOSPHATE IN BACTERIA Being comprised of phosphate, sugars, and nucleic acids, DNA is a rich source of

Pi, carbon (C), and nitrogen (N). Extracellular DNA (eDNA) is present in picomolar to micromolar concentrations in aquatic environments, depending on the location tested

(Lorenz and Wackernagel 1994, Bjorkman and Karl 2005). The source of the DNA in aquatic environments is unclear, although much of it may be released from decomposing microbes, zooplankton, fish, or other aquatic-dwelling organisms. Additionally, several marine- and fresh water-dwelling bacterial species were found to secrete DNA under exponential growth conditions (Paul and David 1989, Nielsen, et al. 2007). The purpose of DNA secretion in microbes has been tied with biofilm formation, as DNA can help provide structure to a biofilm matrix (Whitchurch, et al. 2002, Qin, et al. 2007). However the DNA may also serve as source of nutrients. Below is a discussion of some strategies bacteria use to break down eDNA in order to aquire Pi, C, and N.

I.6.1. General strategies for use of DNA as a source of nutrients The interactions of microbial species found in estuarine environments and eDNA has long been studied (Paul, et al. 1988). Indeed, several organisms, including Ruegeria pomeroyi, Yersinia entericolitica, Pseudomonas aeruginosa, Shewanella spp. and 70

Corynebacterium glutamicum, have been shown to utilize DNA and/or nucleotides as sources of Pi, C, and N (Trulzsch, et al. 2001, Rittmann, et al. 2005, Pinchuk, et al. 2008,

Mulcahy, et al. 2010, Sebastian and Ammerman 2011). These species have developed methods to break down eDNA into nucleosides and Pi, which can be shuttled into the cytoplasm through dedicated nucleoside and Pi transporters (e.g. NupC and Pst).

Since bacteria can also take up intact DNA strands, i.e. natural transformation, this process has been suggested as a mechanism for nutrient acquisition (Sinha, et al.

2013). However, this hypothesis is highly debated in the natural competence field

(Johnston, et al. 2014). Considering that upon transport into the cytoplasm of single stranded DNA by competence machinery, the DNA is very quickly coated with single- stranded binding protein (SSB), it seems unlikely that this DNA could be accessed by degradation enzymes that would allow its use as a source of nutrients (Dubnau 1999).

The cognate strand of DNA not taken up by the cell is degraded into nucleotides, providing a source of energy for the transport of the intact strand across the inner membrane. Thus, it is possible that this broken-down strand is further degraded into nucleoside and Pi and used as nutrients. Until further work has been completed it will be unclear if natural competence is tied with nutrient acquisition. However, experiments conducted in this thesis have demonstrated that the competence machinery of V. cholerae is likely not involved in nutrient acquisition.

1.6.2. Bacterial DNA nucleases The first step of DNA degradation is completed by nucleases. These enzymes can be classified based on any number of traits, e.g. substrate specificity, direction of processivity, and possession of endo- or exonuclease activity. Most nucleases are

71 dependent on the presence of metal ions for catalysis. Within the bacterial cell nucleases contribute to DNA replication and repair. However, secreted nucleases are tied to defense and nutritional acquisition (Yang 2011).

Secreted nucleases have been isolated from a wide range of bacteria, including P. aeruginosa, Streptococcus pyogenes, Corynebacterium glutamicum, Synechocystis spp., and Aeromonas hydrophila (Chang, et al. 1992, Dodd and Pemberton 1996, Ishige, et al.

2003, Suzuki, et al. 2004, Buchanan, et al. 2006, Hasegawa, et al. 2010, Mulcahy, et al.

2010). The role of these nucleases is largely unknown. However, SpnA and three other cell wall-associated nucleases of Streptococcus pyogenes were shown to be required for full virulence of the pathogen through the dissolution of neutrophil extracellular traps

(NETs) (Buchanan, et al. 2006, Hasegawa, et al. 2010). NETs are a web-like structure comprised of neutrophil secreted-eDNA and antimicrobial proteins, such as neutrophil elastase and modified histones (Kawasaki and Iwamuro 2008, Branzk and

Papayannopoulos 2013, Zawrotniak and Rapala-Kozik 2013). NETs are important for killing by the host of a wide array of microbial pathogens, and thus the secreted nucleases of numerous pathogens may prove important for protection from NETs (Branzk and

Papayannopoulos 2013, Zawrotniak and Rapala-Kozik 2013). Alternatively, nucleases identified in water-dwelling organisms such as C. glutamicum and Synechocystis spp. are induced under low phosphate conditions, suggesting these nucleases play a role in acquisition of Pi from eDNA, although this has not been confirmed (Ishige, et al. 2003,

Suzuki, et al. 2004). Interestingly, P. aeruginosa, a microbe that can be found in both aquatic environments and within a human host, secretes a low phosphate-induced nuclease, EddB (Lewenza, et al. 2005, Mulcahy, et al. 2010). Although the interaction of

P. aeruginosa with NETs has not been defined, it will be interesting to see if this nuclease plays a role in both nutritional acquisition and defense against the host.

I.6.3. Bacterial nucleotidases The breakdown of DNA by secreted endo- and exo- nucleases leads to accumulation of extracellular nucleotides. Although nucleotides are easily transported across the outer membrane of Gram negative bacteria, they are considered incapable of permeating the inner membrane (Watanabe, et al. 2011). Using radiolabeled nucleotides

Mycoplasma mycoides and Vibrio parahaemolyticus were shown to transport nucleotides into the cytoplasm, however the authors did not preclude the possibility that the nucleoside and Pi were transported separately into the cell (Sakai, et al. 1987, Youil and

Finch 1988). Therefore, prior to uptake of the nucleotide components, the nucleotides must be broken down outside of the cytoplasm. The final product of nucleases is either 5’ or 3’ nucleotides, which are differentiated by attachment of the phosphate group to either the 5 or 3 position carbon in the (deoxy)ribose. Three main classes of phosphatases have been connected with bacterial break-down of extracellular 5’ or 3’ nucleotides: alkaline phosphatases, UDP-sugar hydrolases, and 2’3’ cyclic phosphodiesterases, and these will be discussed in turn.

Alkaline phosphatases are nonspecific phosphatases, which are able to hydrolyze a wide array of phospho-bonds and are induced under phosphate starvation conditions

(Rao and Torriani 1990). Three classes of alkaline phosphatase have been identified and studied: PhoA, PhoD, and PhoX (White 2009). PhoX is the most recently described class, and it is found primarily in marine bacteria. This class of alkaline phosphatases generally functions as monomers and requires Ca2+ binding for activation of the catalytic site

(Sebastian and Ammerman 2009). PhoX isolated from both R. pomeroyi and Shewanella spp. was found to be important for growth in nucleotides and/or DNA (Pinchuk, et al.

2008, Sebastian and Ammerman 2011). Interestingly, PhoX from R. pomeroyi is not required for removal of Pi from glucose-6-phosphate or β-glycerol-phosphate unlike the vast majority of alkaline phosphatases (Sebastian and Ammerman 2011). Conversely,

Campylobacter jejuni requires PhoX for growth on glucose-6-phosphate, but not ATP as the sole source of phosphate (van Mourik, et al. 2008). Although the role of PhoX in phosphorus assimilation from nucleotides has not been clearly defined, the alkaline phosphatase is unquestionably involved in this phenotype of some bacterial species. The experiments presented in this thesis demonstrate that PhoX is either not involved, or plays a minor role, in acquisition of Pi from eDNA and nucleotides.

UDP-sugar hydrolases are another broad class of enzymes with two catalytic activities: 1) UDP-sugar  UMP + sugar-Pi and 2) UMP  uridine + Pi (Neu 1967a).

The first step of this reaction was identified first, and is reflected in the name of the enzyme. The second process of this step represents 5’ nucleotidase activity, and is most relevant to this thesis. UDP-sugar hydrolases, which are extra-cytoplasmic, have broad substrate specificity for nucleotides, but they do not act on cyclic nucleotides (Neu

1967b, Rittmann, et al. 2005). These enzymes are able to act on the manufactured phosphatase substrates 5-bromo-5-chloro-3-indolyl phosphate (XP) and p-Nitrophenyl phosphate (pNPP), which are widely used in screens to identify phosphatases (Lee, et al.

2000). Two subclasses of UDP-sugar hydrolases have been described in S. enterica,

UshA and UshB (Edwards, et al. 1993). Since UshA is the more commonly researched

UDP-sugar hydrolase, I will focus on it in this section.

UshA is a highly conserved protein among bacterial species. Indeed UshA from

E. coli, S. Typhimurium, and E. aerogenes all share high amino acid sequence identity except for in the secretion signal sequence (Lee, et al. 2000). Despite sequence conservation, transcriptional regulation of ushA is not consistent between bacterial species. For example, phosphate starvation induces transcription of ushA in C. glutamicum, but not in E. coli (Burns and Beacham 1986, Ishige, et al. 2003, Kocan, et al. 2006). Although E. coli UshA is the most thoroughly studied UDP-sugar hydrolase in bacteria, its physiological function remains unclear. However, UshA of C. glutamicum and Shewanella spp. is essential for growth on eDNA and/or nucleotides (Rittmann, et al.

2005, Pinchuk, et al. 2008).

The final nucleotidase that has been tied with use of eDNA and/or nucleotides as a source of nutrients is CpdB, which is a 2’3’ cyclic phosphodiesterase. This enzyme has two independent active sites that catalyze the two-step reaction: 1) 2’3’ cyclic nucleotide

 3’ nucleotide and 2) 3’ nucleotide  nucleoside + Pi (Anraku 1964a, Anraku 1964b).

Although 2’3’ cyclic phosphodiesterase and 3’ nucleotidase activities have been described in several organisms including many of the Enterobacteriaceae, the physiological role of this enzyme is not well characterized (Neu 1968). CpdB of Yersinia pestis and Y. entericolicita is essential for growth on 2’3’ cAMP as the sole source of carbon. Additionally, the activity of CpdB in these species is carbon catabolite repressed, such that glucose inhibits expression of cpdB in a cAMP-CRP dependent manner

(Trulzsch, et al. 2001). The contribution of CpdB to phosphorus assimilation has not been studied to my knowledge; however, in this thesis I show that the protein is required for the ability of V. cholerae to utilize 3’ nucleotides as a source of Pi.

1.6.4. V. cholerae interactions with DNA

V. cholerae is known to grow on eDNA as a sole source of Pi (Seper, et al. 2011).

During its life cycle, the organism may encounter eDNA in both the aquatic environment or in the host. The aquatic environment provides ambient eDNA (as discussed above), but eDNA is also found within the matrix of V. cholerae biofilms (Seper, et al. 2011). V. cholerae can stimulate the release of NETs from neutrophils, and thus DNA from these

NETs could potentially provide a source of nutrients for the organism (Seper, et al.

2013). However, since this work was completed ex vivo, it is unclear if V. cholerae encounters NETs during a natural infection.

Little is known about how V. cholerae interacts with eDNA. V. cholerae produces and secretes two nucleases into culture supernatant: Xds – with exonuclease activity and

Dns – with endonuclease activity (Focareta and Manning 1987, Focareta and Manning

1991a, Focareta and Manning 1991b, Seper, et al. 2011). Xds and Dns were demonstrated to break down eDNA present in biofilms, with the proposed purpose being escape from the matrix (Seper, et al. 2011). Although deletions of xds and dns have not affected virulence or colonization of the pathogen, Xds and Dns are also important for protection of V. cholerae from NETs (Focareta and Manning 1991a, Seper, et al. 2013).

Although the main role of Xds and Dns may be to break down structural eDNA, it is unsurprising that both are required for wild type growth on eDNA as a sole source of Pi

(Seper, et al. 2011). In Chapter IV of this thesis I will discuss the work I have done to build on the knowledge of how V. cholerae uses DNA as a source of Pi.

1.7. THESIS STATEMENT: AIMS AND RATIONALE

1.7.1. Aims When I began work on this thesis we were interested in elucidating what cues V. cholerae senses at the late stage of infection that lead to induction of the late genes. We hypothesized that late genes are induced by either global or specific regulators in response to nutritional or other environmental cues in the host small intestine. Therefore, we designed and implemented a high-throughput genetic selection to identify regulators of a specific set of late genes (VC0130/cdpA, VCA0083/emrD, and VC2621/xds). A discussion of each of these late genes and what was known about their regulation in 2007 follows this section. We chose these three late genes based on several criteria, however we were limited in our selection of late genes due to an agreement with Stefan Schild, the first author of the late gene paper. First and foremost, all three of the genes were identified as induced during human infection (Lombardo, et al. 2007). Additionally, we chose emrD and xds based on their induction in nearly 90% of the population by 21 hours post infection of an infant mouse. Finally all three chosen late genes showed a large difference in induction in vitro versus the 21 hour time-point of infection (Schild, et al.

2007).

Following identification of late gene regulators we aimed to characterize at least one of these regulators in terms of its role during infection. Chapter II of this thesis is a description of the high-throughput genetic selection and the results therein. Chapter III builds on one of these regulators, PhoB as a regulator of xds, and defines its role in vivo.

Chapter IV further expands on characterization of the physiological importance of PhoB regulation of xds by investigating the mechanism of Pi assimilation from DNA.

I.7.2. Introduction to specific late genes studied in this thesis I.3.2.1. cdpA Three of the late genes identified by our laboratory are in the DGC family of proteins (VC1593/acgB, VC2697, and VC2369) and are predicted to lead to an increase in c-di-GMP levels as the bacteria prepare to exit the host (Schild, et al. 2007). Exit of V. cholerae from a host into fresh water results in hypo-osmotic shock and nutritional limitation and therefore pre-induction of biofilm genes may be highly advantageous for the bacterium. Considering that formation of a biofilm appears to be disadvantageous to the bacterium while still in the host, it makes sense that V. cholerae also may temper the increase in c-di-GMP level by expressing cdpA, a known PDE, late in infection

(Lombardo, et al. 2007, Schild, et al. 2007, Tamayo, et al. 2008).

When I began this thesis, the mechanism of transcriptional activation of cdpA was unclear, but it was known to increase during biofilm formation possibly to limit the extent to which the biofilm forms (Tamayo, et al. 2008). Conversely, induction of acgB, one of the DGC late genes, is triggered by low phosphate growth both in vitro and in vivo

(Pratt, et al. 2009). Therefore, we hypothesized that PhoB may be the regulator of cdpA and/or additional late genes.

I.3.2.2. emrD EmrD is closely related to a group of uncharacterized Major Facilitator

Superfamily (MFS) cytoplasmic membrane transporters in the Moritella sp. (>60% identity). Transporters in the MFS are found throughout all kingdoms of life and the family comprises over 10,000 different proteins. Although some of these transporters have been thoroughly characterized, e.g. E. coli LacY and GlpT, the vast majority of them are of unknown function. MFS transporters have been shown to transport a wide

78 array of substrates including ions, carbohydrates, peptides, nucleotides, antibiotics, and other molecules. However, individual transporters generally display specificity for transport of structurally similar molecules (Yan 2013). The gene emrD is named as such based on homology to the E. coli protein EmrD (30% identity), which is itself in the MFS and is implicated in resistance to antibiotics and the ionophore, CCCP (Naroditskaya, et al. 1993, Yin, et al. 2006). V. cholerae EmrD-3, which has only 32% identity with EmrD-

1, was recently described as a multidrug efflux pump (Smith, et al. 2009). However, the substrate specificity of EmrD-1 is still unknown. Additionally, nothing is known about the regulation of this gene.

I.3.2.3. xds As discussed above Xds is one of two secreted nucleases of V. cholerae, the other being Dns (Focareta and Manning 1987, Focareta and Manning 1991a, Focareta and

Manning 1991b, Seper, et al. 2011). The nucleases are involved in escape from biofilms, protection from NETs, and utilization of DNA as a source of Pi (Seper, et al. 2011, Seper, et al. 2013). When I began work on this thesis it was known that dns is repressed by the quorum-sensing regulator, HapR (Blokesch and Schoolnik 2008). As mentioned previously, HapR levels are expected to increase as the infection progresses and the population density increases in the small intestine, and this would lead to repression of dns late in infection. This is surprising since it would imply that Xds and Dns are active at separate times during infection, despite the fact that they contribute to the same phenotypes.

CHAPTER II:

GENETIC SELECTION TO IDENTIFY REGULATORS OF V. CHOLERAE LATE GENES

Some of the text and figures/tables in this section are adapted from my manuscript with permission from the publisher:

McDonough, E., Lazinski, D.W., Camilli, A. Identification of in vivo regulators of the Vibrio cholerae xds gene using a high-throughput genetic selection. Mol. Microbiol. 2014 Feb. 20.

II.1. INTRODUCTION As described in Chapter I of this thesis, V. cholerae moves between two dissimilar environments: aquatic reservoirs and the intestinal tract of humans.

Accordingly, this pathogen undergoes adaptive shifts in gene expression throughout the various stages of its life cycle (e.g. DiRita 1992, Merrell, et al. 2002, Nielsen, et al. 2006,

Childers and Klose 2007, Schild, et al. 2007, Nielsen, et al. 2010). These dramatic shifts in gene expression are essential for survival and growth of the organism when it is exposed to challenging environments such as the acidic human stomach and the nutrient poor aquatic environment. The V. cholerae virulence gene regulatory cascade, which is induced within the first few hours of infection in the infant mouse model host (Lee, et al.

1999), has been well characterized (Miller, et al. 2002, Zhu, et al. 2002, Kovacikova and

Skorupski 2002b, Zhu and Mekalanos 2003, Nielsen, et al. 2006, Childers and Klose

2007). However, we are only beginning to understand changes in this pathogen at later stages of infection, when V. cholerae turns off virulence genes and induces expression of genes needed for the transition out of the host and into the aquatic environment (Merrell, et al. 2002a, Nielsen, et al. 2006, Schild, et al. 2007).

We were interested in learning more about how V. cholerae regulates transcription at the late stage of infection. As described in this chapter, we sought to identify regulators responsible for the infection (in vivo)-specific induction of three late genes: cdpA, emrD, and xds. To this end, we designed a genetic selection that couples a traditional antibiotic selection with high-throughput transposon sequencing (Tn-seq). We expected to identify both global and specific late gene regulators. We hoped that the

81 identification of specific late gene regulators would lead us to a better understanding of the environment V. cholerae encounters at the late stage of infection.

My first three years in the Camilli laboratory were spent designing, implementing, and validating results from a failed version of the genetic selection. I will briefly explain the design of this original selection, a small number of the validation results, and why we believe it was unsuccessful. After redesigning the genetic selection to circumvent the problems associated with the original version, we successfully identified regulators of two late genes, emrD and xds.

II.2. RESULTS

II.2.1. Design of a novel genetic selection In designing both the original and final genetic selections we took advantage of the fact that V. cholerae late genes are induced during infection, but not during growth in

LB broth (Schild, et al. 2007). In both cases we made transcriptional fusions of the late genes of interest to antibiotic resistance genes. In constructing the transcriptional fusions, we inserted the antibiotic-resistance gene with an optimal ribosomal binding site (RBS)

(Osorio, et al. 2005) immediately after the stop codon of the late gene. No genomic material was deleted in these strains. Due to the lack of appreciable late gene expression in LB, the fusion strains had very low minimal inhibitory concentrations (MICs) to the chosen antibiotics during in vitro growth. These fusion strains were subjected to transposon mutagenesis. The transposon libraries were plated on LB plates in the presence of antibiotic concentrations above the MIC to select for transposon mutants that misregulated the late genes, specifically that led to increased expression of the antibiotic resistance marker. The transposon mutants that led to survival on the antibiotic plates

82 were identified through the use of Tn-seq (van Opijnen, et al. 2009). We expected that we would find perhaps between one and five transcriptional regulators for each gene, considering that a single gene having more than five regulators would be highly unlikely.

However, we also considered the possibility that we could obtain many dozens, perhaps hundreds of unique transposon insertions that led to increased growth on the antibiotic.

Thus, we chose to use the high-throughput Tn-seq method to hopefully identify all of these sites and be able to rank order them in terms of the strength of their selection.

Although we made use of two different transposons for each version of the selection, both carried an isopropyl β-D-1-thiogalactopyranocide (IPTG) inducible and outward-reading promoter. The presence and absence of IPTG during selection allowed for the identification of both activators and repressors (Figure II-1). Transposon insertions that disrupted a late gene repressor would enhance expression of the late gene fusion. These transposon insertion mutants would be enriched in either the presence or absence of IPTG. Alternatively, elevated expression of the late gene fusion could occur in the presence of IPTG when the outward reading promoter of the transposon drove expression of an activator. A third, artifactual category of transposon insertions would be the result of a mutation in cis to a late gene that leads to constitutive expression of the antibiotic fusion if it is in the correct orientation for read-through. This situation would not be useful, except that it functions as an internal control for the activity of the IPTG- inducible promoter.

The use of antibiotic selection to identify regulators of bacterial genes is not novel. The novel aspect of our genetic selection is the use of Tn-seq to identify enriched transposon insertions. Since Tn-seq identifies all transposon insertions present in the pool

LB plates: No expression of late genes Antibiotic plates: Late genes expressed

Repressor Late gene AntibioticR Repres Tn sor Late gene AntibioticR

Activator Late gene AntibioticR Tn Activator Late gene AntibioticR

Late gene AntibioticR Tn Late gene AntibioticR

Figure II-1. Possible transposon insertions leading to misregulation of the late genes. There are three categories of transposon (Tn) insertions that could directly lead to increased expression of the late genes. Category A) could be selected for in both plus and minus IPTG conditions, while categories B) and C) would only be found in the plus IPTG condition. A) Expression of the late genes may be silent in LB growth due to the action of a repressor. Inactivating Tn insertions into the repressor gene would inhibit this LB-related repression and induce expression of the late gene. The orientation of the Tn is irrelevant. B) Late gene expression may require aid of a transcriptional activator that is not expressed during LB growth. Insertion of a Tn upstream of, or in the 5’-end of an activator could drive expression of the activator, and subsequently late gene, provided that the outward reading promoter is oriented to do so. C) Insertion of the Tn directly upstream of, or within the late gene could drive expression of the antibiotic resistance marker. These transposon insertions serve as a control for the activity of the IPTG- inducible promoter.

before and after selection, and the frequency at which they are present, we hypothesized that Tn-seq would allow us to identify both strong and weak transcriptional regulators.

Tn-seq also allowed for accurate determination of the complexity of our input transposon library, which gave us valuable information about the coverage of our selection.

II.2.1.1. Original selection design and problems therein In version one of the genetic selection, we made transcriptional fusions of late genes to bla, which encodes resistance to beta-lactam antibiotics such as carbenicillin

(Cb). We used a mini-Mariner transposon to mutagenize the fusion strains. The transposon libraries were selected on LB plates that were supplemented with 15 µg/mL

Cb (plus and minus IPTG), and aliquots of the library were also plated on LB or LB

IPTG plates as controls. Of note, the MIC for the unmutagenized fusion strains was 10

µg/mL. High-throughput sequencing of the transposon-genome junctions revealed that many more transposon mutants were able to grow than we had originally hypothesized.

Many of the genes identified by the selections could be categorized into functional groups: phosphate acquisition/sensing, iron uptake/metabolism, glutathione and sulfur metabolism, growth under anaerobic conditions, or general stress response genes (Figure

II-2). These data suggested to us that V. cholerae might induce late gene expression upon sensing changes in phosphate, iron, and/or oxygen availability in the host.

We found some overlap in the results between the genetic selections of all three late genes. This was especially apparent when the fold enrichment cut-off was dropped to

10. The overlap between the selections suggested to us that either the late genes we had chosen to study were coordinately regulated, or that many of the transposon insertions selected contribute directly to antibiotic/Cb resistance and not expression of the gene

emrD-bla VCA0686 narQ xds-bla VCA0731 VC0609 VC0817 VCA0254 pepA VC0069 VCA0177 gshB VC0469 gor pstS-1 VC2441 mnmA pstC-1 VC2443 dcuB pstA-1 VC2442 rstA pstB-1 ugd phoU ipk rseB ompU fexA rseA fexB

VC0124 ugpA VCA0640 cysB VC0104 glmU fruB VCA0216 ribE mrcA VC2288 nupC VC2500 VCA0796 zipA pta cdpA-bla dnaZX leuS

Key: Hypothetical protein Anaerobic respiration Phosphate metabolism/uptake Stress response Carbohydrate and lipid metabolism Nitrogen metabolism Iron metabolism/uptake Nucleotide metabolism Glutathione and sulfur metabolism Other

Figure II-2. Venn diagram showing the high degree of overlap between selections from the original selection. Genes with fold enrichment greater than 100 are represented in the Venn diagram. Genes shown as enriched in more than one selection are only shown if fold enrichment was greater than 100 in all selections. Functional categorization is according to Kegg orthologies, modules, and pathways (www.kegg.jp).

fusion per se. Indeed, two genes enriched in our selections, ompU and cysB, were previously implicated in resistance to another beta-lactam antibiotic, ampicillin (Ap)

(Oppezzo and Anton 1995, Nguyen, et al. 2009), and therefore are likely false-positives.

To determine whether genes enriched in multiple selections (as well as cysB) fell into the category of false positives, I determined the MIC of Cb for the knock-out mutants. The ΔompU mutant was made previously by Stefan Schild, a former post-doc in our laboratory. As I wanted a quick and easy method for deleting and validating genes, I made plasmid-insertion knock-outs of both rseA and cysB. I was unable to obtain a knockout of fexB at this time, and therefore I could not determine the MIC of this mutant.

It is important to note that these knockout strains were made using the wild type parental strain (i.e. they did not harbor the bla gene). Therefore, any increase in MIC when compared to the wild type strain was solely due to the deletion of the putative regulator.

However, we considered, but didn’t test the possibility, that the survival of transposon mutants on the Cb selection plates could be due to a combinatorial effect of a gene deletion and constitutive leaky expression of the bla gene from the fusion. The MIC of the wild type strain (no bla gene fusion) was determined to be 5 µg/mL. We found that disruption of ompU and rseA led to a modest increase in the MIC of less than 2-fold, whereas the MIC of the cysB mutant increased approximately 3-fold over that of the wild type (Table II-1). Thus, the MIC experiments supported the hypothesis that insertions in ompU, rseA, and cysB were selected due to their intrinsically increased resistance to Cb.

Despite the results from the MIC experiment, we still entertained the possibility that these genes also contributed to regulation of the late genes. To test this idea, we first used qRT-PCR to measure transcription of cpdA and xds in the wild type and a ΔompU

Table II-1. MIC of putative regulators.

Strain MIC of carbenicillin (µg/ml) Wild type 5 ΔompU >5 and <10 cysB::pGP704 >15 rseA::pGP704 >5 and <10

Three biological replicates were tested to determine the MIC of Cb in each strain. The MIC was defined as the concentration of antibiotic that knocked-down growth >100-fold when compared with the no drug control.

mutant (Figure II-3). Unfortunately, deletion of ompU did not lead to increased transcription of cdpA or xds. Likewise, transcription of cdpA, emrD, or xds was not increased in the rseA and cysB mutants as measured by qRT-PCR (Figure II-4).

Therefore, I concluded that rseA, cysB, and ompU were all false positive hits from the genetic selection.

Despite the high rate of false positives, the data from the original selection did lead us to identify two regulators: phoU as a repressor of xds, and gshB as a repressor of emrD. I will discuss both phoU and gshB and their role as regulators in section II.2.5. of this thesis, as they were also identified in the second version of the genetic selection.

However, since we were interested in publishing the high-throughput sequencing coupled antibiotic genetic selection as a novel method, we were interested in fine-tuning the selection in order to lower the false discovery rate.

#! $%&'()*+, '(#) ( : . + 7 % ' 1 # * 1 ( , 7 3 ) ( + ' 6 , , # ( 9 % & , 5 0 % ) / 0 & 3 , 7 . 4 8

!"# !"#$ %"& -,.,(/,0123,'

Figure II-3. Effect of ompU deletion on transcript levels of cdpA and xds. Expression of cdpA and xds was measured in the wild type and ΔompU strains using qRT-PCR. RNA was collected from three biological replicates grown to OD600 = 0.3 in LB. The expression levels are normalized to the housekeeping gene, rpoB, and set relative to wild type expression. The mean and SEM are shown.

#! $%&'()*+, +-./!011'*%$ ( >

2 +-./!011!+*, , ; % - 5 # ' 5 ( , ; 7 ) ( + ' : ! , , ( = % & , 9 4 % ) 3 4 & 7 , ; 2 8 <

!"# !"#$ %&'( )"*

-,2,(3,4567,'

Figure II-4. Effect of cysB or rseA deletions on expression of late genes. Expression of cdpA, emrD, and xds was measured in the wild type and in the rseA and cysB mutant strains using qRT-PCR. RNA was collected from two biological replicates grown to OD600 = 0.3 in LB. The expression levels are normalized to the housekeeping gene, rpoB, and set relative to wild type expression. The median and range are shown.

II.2.1.2. Version two of the genetic selection In designing the second version of the genetic selection we aimed to eliminate or greatly reduce the number of false positives that were selected in the original design.

Ultimately, we needed to increase the stringency of the genetic selection such that only increased transcription of the resistance genes would allow for survival on the antibiotics.

We considered increasing the concentration of Cb on the plates, as the concentration of

Cb we chose was only 5 µg/mL above the MIC of the fusion strains and we found that transposon insertions in some false positive genes were able to increase the resistance only slightly more than this amount. However, other false positive genes (e.g. cysB) increased resistance to much higher concentrations of Cb. Therefore, instead, we chose to add a second antibiotic resistance marker, neo – encoding resistance to neomycin and kanamycin (Kn), to the transcriptional fusion.

As antibiotics, Cb and Kn have distinct modes of action, and likewise bacteria have evolved different mechanisms of resistance against these drugs. Cb is a beta-lactam, which inhibits cell wall synthesis by binding to and inhibiting penicillin binding proteins

(PBPs) (Park and Strominger 1957, Zapun, et al. 2008, Sainsbury, et al. 2011). Cb and other beta-lactam antibiotics are inhibited via three potential mechanisms: 1) via the action of beta-lactamases, which break open the beta-lactam ring making it unrecognizable by the PBPs; 2) altered PBP structures; or 3) increased efflux of the antibiotics through outer membrane pores (Livermore 1984, Li, et al. 1994, Wilke, et al.

2005). Conversely, Kn is an aminoglycoside, which binds to the A-site of the ribosome and induces translational misreading (Davies and Davis 1968, Kotra, et al. 2000).

Resistance to Kn is primarily mediated through three mechanisms: 1) aminoglycoside-

92 modifying enzymes, whose actions result in altered and inactive forms of Kn; 2) altered influx or efflux of the antibiotic across the cytoplasmic membrane; and 3) alteration of the ribosome such that the aminoglycoside is no longer able to interact (Kotra, et al.

2000, Hollenbeck and Rice 2012). Considering the differences in mechanisms of action, modes of resistance, and location of activity, it is unlikely that a single transposon mutation would confer resistance to both Cb and Kn antibiotics independent of transcription of the reporters.

Figure II-5 outlines the basic protocol of the second version of our genetic selection. The strains containing the late gene-bla-neo fusions were subjected to mTn10 transposon mutagenesis. The specifics of this mTn10 transposon and an explanation of why we switched transposons is detailed in section II.2.2. The mTn10 carries lacI together with an isopropyl β-D-1-thiogalactopyranocide (IPTG) inducible outward- reading Ptac promoter. The transposon mutant library was plated on LB agar supplemented with Cb + Kn (+/- IPTG) to select for transposon insertions that caused enhanced expression of late gene-bla-neo. Three independent selections were performed for each fusion strain. Each of the independent selections used slightly different antibiotic concentrations, such that we could potentially identify different classes of regulators

(weak and strong).

In addition to including a second antibiotic marker in this new version of the genetic selection, we also used a wild type Tn library control, i.e. that lacked fusions to bla-neo, that we hoped would help us weed out any false positives. Any transposon hop enriched in the wild type selection would be considered background noise.

Ptac

Store aliquots late gene bla neo mTn10 in -80ºC

gDNA INPUT 1. Sequence mTn10 junctions with Illumina HiSeq

2. Determine ratio of OUTPUT OUTPUT/INPUT of each gDNA (Repressor mTn10 insertion mutants) Carbenicillin + 3. Identify enriched insertion Kanamycin mutants OUTPUT (Repressor or gDNA Activator mutants) Carbenicillin + Kanamycin + IPTG

Figure II-5. Design of the genetic selection, version two. Late genes were transcriptionally fused to the beta-lactamase (bla) and neomycin phosphotransferase (neo) genes, which encode resistance to carbenicillin and kanamycin, respectively. As in vivo induced genes, late genes are expressed at low levels during growth on LB agar, therefore, the fusion strains (cdpA-bla-neo, emrD-bla-neo, and xds- bla-neo) are unable to grow on plates containing concentrations of carbenicillin and kanamycin above the MIC. The fusion strains were mutagenized with a mini-Tn10 transposon, which carries an outward reading, IPTG-inducible promoter, Ptac. Mutants with increased expression of the bla-neo fusion were selected on either carbenicillin + kanamcyin or carbenicillin + kanamycin + 1 mM IPTG. Transposon insertions enriched on the plates without IPTG were those disrupting a repressor of the late genes. Conversely, transposon insertions enriched under IPTG positive conditions included both repressor knock-outs and insertions upstream of the query late gene activators. After selection, genomic DNA (gDNA) was isolated from pooled colonies and the samples were prepared for transposon junctional sequencing using the Illumina HiSeq. Input gDNA was prepped directly from the frozen transposon libraries.

II.2.2. Construction of saturating transposon libraries We considered three transposons for use in the selection: mMariner, mTn10, and mTn5, all of which were present on plasmids that allowed for in vivo transposition. These transposons had previously been used successfully in V. cholerae to create transposon libraries, although the complexities of these libraries are unknown (Chiang and

Mekalanos 2000, Judson and Mekalanos 2000, Merrell, et al. 2002b, Pratt, et al. 2009).

The mTn10 and mTn5 mutagenesis systems available at that time yielded, in my hands, low complexity libraries with plasmid integration rates upwards of 10%. Therefore, as mentioned above we made use of the mMariner system in the original design of the genetic selection. However, after high throughput sequencing the input transposon libraries (after completion of the selection version one) we found that complexities of our mMariner libraries were quite low (approximately 10,000 to 20,000 insertions in each library).

Just prior to the redesign of the genetic selection, David Lazinski, a research scientist in our laboratory, designed a transposon delivery plasmid (pDL1098) that was greatly improved from the systems we previously had in the laboratory (Figure II-6). The pDL1098 delivery plasmid, which harbors mTn10, is temperature-sensitive for replication, but requires high temperature for transposition. At the permissive temperature of 30°C the plasmid replicates and the transposase is repressed. At the non-permissive temperature of 40°C the transposase is induced and replication stops, resulting in loss of the vector in the multiplying population of bacteria. Prior to my use of the new delivery system, David demonstrated that he was able to obtain >100,000 unique transposon insertions in E. coli.

Figure II-6. pDL1098 delivery system for mTn10. The mTn10 transposon we used is carried on the pDL1098 delivery vector, which carries the gene encoding chloramphenicol resistance (cat). This plasmid is mobilizable, and can thusly be transferred into V. cholerae via conjugation with an appropriate E. coli donor strain. Additionally, the repA gene, which is required for replication of the plasmid, is a temperature sensitive variant of the pSC101 repA. The spectinomycin resistance- encoding mini-transposon carries an outward reading Ptac promoter along with the lac repressor near one inverted repeat (Tn10 IR). Near the other inverted repeat, the transposon harbors a double transcriptional stop signal (rrnB T1 and rrnB T2) to prevent outward reading transcription from that end of the mTn10. Transposition is mediated through the action of the Tn10 transposase, which is transcriptionally controlled by the lambda repressor. The plasmid encodes the cI 857 temperature sensitive variant of the lambda repressor. The map of pDL1098 was created using CLC Main Workbench version 6.5 (CLC Bio).

Since the transposon delivery vector had not yet been tested in V. cholerae prior to my work, we were concerned that there may be a propensity for “hot spot” insertions or that the transposition would not work as well as it had in E. coli. Therefore, I made two independent transposon libraries in each background that I planned to put through the genetic selection (four fusion strains and wild type). This allowed us to learn about the biology of the system, as well as determine if the libraries were of good quality for use in the selection.

Using Tn-seq we were able to accurately determine the complexity of our input transposon libraries. The specifics of the transposon library complexities we obtained are described in Table II-2. In general each library contained approximately 100,000 transposon insertions, which was similar to what was obtained in E. coli. The total number of transposon insertions represented by combined biological replicates for each strain was >130,000, with approximately 30 to 40 percent overlap between them. These data suggested to us that despite the high number of transposon insertions, each individual library was not yet at saturation, nor are the combined libraries. The most highly represented transposon insertions were very similar between all eight of the libraries. For example, two insertions in the gene encoding cholera toxin and at positions

1567747 and 1567756 in chromosome I were within the top 25 insertions in all eight libraries. These data suggest that there are hot spots in V. cholerae for mTn10 transposition, and in fact others observed this same phenomenon in the laboratory (Faith

Wallace-Gadsden and Heather Kamp, personal communication). However, none of the libraries exhibited insertions that represented more than 0.4 to 5% of the total transposon library.

Table II-2. Properties of transposon libraries in biological replicate.

Number of unique mTn10 insertion sites Number of Total Percent of total Library Library overlapping number of insertions represented Strain 1 2 insertions insertions in both libraries Wild type 99,654 88,655 50,107 138,202 36 cdpA-bla-neo 133,928 133,857 78,126 189,659 41 emrD-bla-neo 126,723 66,408 46,002 147,129 31 xds-bla-neo 140,262 150,430 85,557 205,135 42

In order to increase the complexity of the transposon libraries, aliquots of each independently derived library were mixed prior to subjection to the antibiotic selection.

The libraries were mixed roughly according to ratios of unique insertions in each library.

For example, the wild type transposon libraries contained 99,654 and 88,655 insertions each and the emrD-bla-neo transposon libraries contained 126,723 and 66,408 unique insertions each. These wild type libraries were mixed at a 50/50 ratio, whereas the emrD- bla-neo libraries were mixed at a 70/30 ratio. Each final mixed library contained greater than 140,000 unique transposon insertion sites and represented insertions in at least 3,400 or 3,885 (~90%) annotated open reading frames (ORFs) in the V. cholerae genome. Table

II-3 outlines the specifics of these mixed transposon libraries that functioned as my inputs for the genetic selection.

Table II-3. Properties of final mixed transposon libraries.

Number of Percentage Number of intergenic Percentage of unique Number of regions with of disrupted Fusion mTn10 genes mTn10 undisrupted intergenic Strain* insertion sites disrupted insertions genes regions cpdA-bla-neo 238,077 3,628 2,517 9.3 8.0 emrD-bla-neo 160,659 3,485 2,303 9.0 7.4 xds-bla-neo 169,198 3,526 2,352 9.1 7.5

*Because I was unable to determine antibiotic selection conditions for the wild type strain, I never sequenced a mixed library.

100

II.2.3. Antibiotic minimal inhibitory concentration (MIC) experiments with transposon libraries We performed a series of minimal inhibitory concentration (MIC) experiments to determine appropriate concentrations of Cb + Kn for our selection. In these experiments, we varied Cb and Kn concentrations, as well as the number of CFU plated. We examined both the mutagenized and non-mutagenized bla-neo fusion strains in these preliminary assays. We chose antibiotic and cell density combinations that fulfilled the following criteria: were able to completely inhibit growth of the unmutagenized strain, produced little to no background growth, allowed for development of discrete colonies, and exhibited reproducible degrees of killing. The chosen concentrations are described in

Table II-4. I found that in conditions where the concentration of at least one of the antibiotics was too close to the MIC, the selection was not consistent from plate to plate in terms of both background growth and percent kill. Additionally, the freshness of the antibiotics and dryness of the selection plates contributed greatly to experiment-to- experiment variability. We had a particularly tough time identifying antibiotic selection concentrations for the cdpA-bla-neo strain that yielded reproducible results even between technical replicates performed on the same day. The possible reasons for this are discussed in the discussion section of this chapter. We also had a hard time identifying antibiotic selection concentrations for the wild type strain. Like the cdpA-bla-neo strain, the wild type results were highly variable and inconsistent. Additionally, the wild type strain exhibited a filamentous phenotype under the microscope, which has been reported by others in the laboratory when V. cholerae is exposed to ampicillin. Due to the severe lack of reproducibility, we did not move forward with the wild type strain.

101

Table II-4. Antibiotic concentrations used in the genetic selections.

Minus IPTG concentrations Plus IPTG concentrations Average Average number of number of Fusion colonies per Fold colonies per Fold strain Cba Kna plate killingb Cb Kn plate killing cdpA-bla-neo 10 80 538 18401 10 50 1500 6600 12 65 867 11418 12 50 1500 6600 12 80 232 42672

emrD-bla-neo 16 40 756 10847 16 40 1005 8159 18 45 237 34599 18 45 881 9308 20 50 222 36937 20 50 620 13226

xds-bla-neo 16 40 338 4142 16 40 797 1757 18 45 604 2318 18 45 1017 1377 20 50 271 5166 20 50 1155 1212 a Antibiotic concentrations measured in µg/mL. b Fold killing was determined by dividing the CFU/mL of Tn library as determined by plating on LB plates, by the average number of colonies per plate

102

II.2.4. Design of data analysis Tn-seq experiments in our laboratory are generally analyzed using an equation that calculates the fitness contribution of each gene under the selection condition used. In order to calculate a true fitness value, this equation takes into account the net growth of the bulk population (van Opijnen, et al. 2009). In addition, we typically use a set of neutral genes to determine a normalized fitness value for each gene. The neutral genes are those that, when knocked out, do not contribute to the bacterium’s ability to survive and grow in the tested selection condition. The mean fitness of the set of neutral genes is set to one; the majority of the other genes should have fitness scores that also cluster around one. The experiments are designed such that the genes of interest are those that exhibit a fitness statistically significantly less than or greater than one (van Opijnen, et al.

2009, Klein, et al. 2012, Kamp, et al. 2013). However, the concept of fitness and the use of neutral genes does not apply to the experimental design described in this thesis because the majority of transposon mutants are killed in the presence of antibiotics.

Therefore, we developed methods of data analysis specific to this genetic selection technique, in order to determine transposon insertions that were significantly enriched

(outlined in Figure II-7). We analyzed the plus IPTG and minus IPTG selection data separately and differently because the type of transposon insertion mutation we were interested in was different for each condition (insertions upstream of activators vs. insertions within repressors). A more detailed description of the data analysis is described in Chapter VI.

In order to identify repressors of the three late genes, we utilized the minus IPTG selection data. From these data, we calculated a fold enrichment score for each gene

103

bla-neo # normalized output reads # normalized -Strand bias only in output reads -Strand =

insertion FE Insertions if: of interest in genes enriched in -IPTG -Not present of -Not directly upstream considered: regions Genes/intergenic ->2 insertions with FEinsertion >5 selections in all three -Represented # input reads ) genome = value selection value 3 p V. cholerae V. value value p ( = 3 value* selection 2 + p value p +IPTG # reads in output Tn X Tn in output # reads selection 1 + -IPTG value p Quality control and Mapping Quality control poly C-tail trimming -3’ >7 quality quality score) by (95% of bases must have Filter -FASTQ N16961 mapping to -Bowtie analysis Hopcount insertion Tn each at reads # sequence -Tallies of insertions location strand -Shows test signed-rank Wilcoxon >4 insertions -Required per gene output and input insertions -Compares in each gene correction -Bonferroni test signed-rank Wilzoxon Average Normalized output hopcount output hopcount Normalized in input # reads total in output # reads total selection 3 gene FE = 3 selection 2 + gene FE =

gene output standardized read frequency read output standardized ge FE =

selection 1 + gene gene FE Aggregate hopcount analysis hopcount Aggregate in each gene # sequencing reads -Tallies frequency* read Standardized in library number of reads in gene X / total # reads genome size of gene X / total size Avera input standardized read frequency read input standardized

104

Figure II-7. Flow chart of data analysis methods. The minus IPTG (represented by blue lines) and plus IPTG (represented by red lines) data sets were analyzed separately and differently. Each individual line represents a different antibiotic concentration. The solid thick lines indicate that the information from the separate selections was combined. After analysis, average FEgene scores and Wilcoxon signed-rank test p values were sorted in descending order and the lists were compared. The TUFC Galaxy Server (Giardine, et al. 2005, Blankenberg, et al. 2010, Goecks, et al. 2010) was used to perform the following analyses: quality control and mapping, hopcount analysis, aggregate hopcount analysis, and standardized read frequency. All other analyses were performed using the SAS-based statistics software package JMP pro (Version 10.0, SAS Institute Inc., Cary, NC).

105

(FEgene) in all three antibiotic selection outputs, for each late gene. As a first pass, genes that were not enriched to some extent in all three selections were discarded. An average

FEgene of the three independent selections was calculated as described in Figure II-7. The vast majority of genes with transposon insertions that were selected for exhibited average

FEgene scores <1; these were considered background, i.e. non-enriched genes. The first category of genes we considered as putative regulators were those that exhibited an average FEgene score >1,000.

In order to better determine the reproducibility of our FEgene results, we performed statistical analysis using the Wilcoxon signed-rank test on the minus IPTG data. The test compared read counts in the output to those, paired by insertion site, in the input, for each gene separately. The Wilcoxon signed-rank test makes three assumptions about the data:

1. The data are paired and from the same population. (In our case, each Tn

insertion has paired input and output data, and comes from the same

transposon library population.)

2. Each pair are random and independent. (Our Tn insertions were created

randomly and independently.)

3. The data are measured on an interval scale, meaning the difference

between 10 and 20 is the same as the difference between 80 and 90. (The

data we collected were measured in number of reads on the Illumina chip,

which roughly corresponds to the number of bacterial cells represented on

the selection plate. The difference between 10 and 20 bacterial cells is the

same as the difference between 80 and 90 cells.)

106

Essentially, the Wilcoxon signed-rank test allowed us to determine which genes exhibited a statistically significant difference in the number of reads in the output as compared with the input. A Bonferroni multiple comparisons analysis was performed on the Wilcoxon signed-rank test data in order to correct for the multiple comparisons to the single input data set and lower the false discovery rate. We calculated an average p value for each gene of the three minus IPTG outputs. Comparing the average p values with the average

FEgene scores, we were able to identify a list of putative repressors for each late gene of interest.

To identify putative activators of the late genes we analyzed the plus IPTG selection data. The genes of interest in the plus IPTG selection are downstream of the transposon-carried Ptac promoter. Instead of examining the plus IPTG data on the gene level, as we had done for the minus IPTG data, we analyzed the data at the insertion level. This analysis allowed us to identify enriched insertions, determine their directionality, and examine which downstream genes were potentially affected. Insertions were ranked by fold enrichment (FEinsertion) as outlined in Figure II-7 and described in more detail in Chapter VI. Although we were ultimately interested in transposon insertions within regions that could span both intergenic and genic regions, in order to easily analyze the data, we grouped transposon insertions by gene/ intergenic regions and first identified genes or intergenic regions of interest. Genes/ intergenic regions were considered of interest if they passed the following requirements: 1) insertions within the gene/ intergenic region showed strand bias in the output enriched transposons, but not in the input; 2) at least two insertions with an average FEinsertion >5 were enriched in the

107 output; and 3) the gene/ intergenic region was not directly upstream of the bla-neo fusion or identified in the minus IPTG selection.

Although we could have technically performed a statistical analysis of the plus

IPTG selection data at the insertion level, we did not due to lack of power in the analysis.

The region of DNA a transposon could insert and effectively induce an activator was potentially quite small when compared to the size of a gene. The additional requirement for insertions to be oriented in one direction cuts the number of potential observations for a statistical test in half. Thus, if we ran a statistical test, we would have risked losing too many potential hits as false negatives. For example, one of our true hits from the plus

IPTG selection only had two Tn insertions that were enriched and thus would have been discarded in any statistical test due to an inadequate number of observations.

Once a list of genes/ intergenic regions was identified using the above criteria, putative positive regulators of the late genes were determined. This was done by mapping the insertions to the V. cholerae genome and determining what coding regions were directly downstream of the transposon-encoded Ptac promoter. For several identified genes/ intergenic regions we were unable to identify putative activators. This was usually due to the Ptac promoter appearing to drive convergent transcription with the downstream genes. If only a few insertions were enriched within a particular gene, we considered the possibility that the transposon insertions could be driving transcription of a truncated protein with altered function. However, this possibility was ruled out if the two enriched insertions were located far apart from one another with multiple insertions in between that were not selected for.

108

II.2.5. Identification and in vitro validation of putative late gene regulators In this section I have outlined the genes identified as putative regulators of V. cholerae late genes, as well as the work conducted to validate these regulators in vitro. In summary, we identified regulators of both emrD and xds, but failed to identify regulators of cdpA.

The highest percentage of enriched transposon insertions in every plus IPTG selection were those directly upstream of the bla-neo fusions and oriented in the correct orientation to drive transcription of the antibiotic markers. While these transposon insertions confirmed the selective induction of the Ptac promoter in the presence of IPTG, they did not provide any insight into regulation of the late genes. In all three fusion backgrounds the insertions in cis to the bla-neo fusion were 1,000 to 10,000 fold more enriched in the plus IPTG selections than in the minus IPTG selections. This is visualized in Figure II-8 in which I have plotted the fold enrichment of each in cis Tn insertion (up to 10,000 bp upstream of the bla-neo fusion) as compared with the distance from the bla- neo fusion RBS. Also included is a LOWESS line, which is a general curve fitting algorithm for non-standard data sets. The LOWESS lines for the cdpA-bla-neo minus

IPTG selection are thrown off by the presence of highly selected insertions upstream of the fusion. However, the data points themselves clearly demonstrate the separation between the minus IPTG and plus IPTG selections.

As discussed, we hypothesized that we might see a difference between each of the antibiotic conditions. For example, we predicted that Tn insertions further away from the bla-neo fusion would be more enriched in the lower antibiotic concentrations than in the presence of strong antibiotic selection. However, by comparing the LOWESS lines, it

109

A) !"#$%!"#"$%&'( +))))) ./+)01(*)0203456 ./+,01(*)0203456 ./+)01(7) +)))) ./+,01(8* ./+,01(7)

+))) % ( "

= +)) A $ & ; ( "

0 +) @ # ' ? +

)-+

)-)+ ) *))) +)))) +*))) ,)))) /90:9!%;"<=0'>0&'()*+,

B) +-./%0+'+!12,* +))))) ./+801(C)0203456 ./+701(C*0203456 +)))) ./,)01(*)0203456

./+801(C) +))) ./+701(C*

% ./,)01(*) ( "

= +)) A $ & ; ( "

0 +) @ # ' ? +

)-+

)-)+ ) ,))) C))) 8))) 7))) +)))) /90:9!%;"<=0'>0/#

C) 3"0%0+'+!12,* +))))) ./+801(C)0203456 ./+701(C*0203456 ./,)01(*)0203456 +)))) ./+801(C) ./+701(C* ./,)01(*) +))) % ( "

= +)) A $ & ; ( "

0 +) @ # ' ? +

)-+

)-)+ ) *))) +)))) +*))) ,)))) /90:9!%;"<=0'>0/#

110

Figure II-8. Fold enrichment of in cis Tn insertions. The fold enrichment of each Tn insertion was plotted versus its distance from the A) cdpA-bla-neo, B) emrD-bla-neo, or C) xds-bla-neo fusion RBS. Closed symbols represent insertions enriched in the plus IPTG selections, whereas open symbols represent insertions enriched in the minus IPTG selections. Each color represents a different combination of antibiotic concentrations. LOWESS lines are drawn for each selection condition – solid lines for plus IPTG and dashed lines for minus IPTG selections. Note the differences in the x-axis scale. Analysis was completed using JMP pro (Version 11.0, SAS Institute Inc., Cary, NC).

111

does not appear that there is a clear relationship between antibiotic concentration and fold enrichment of distant insertions.

When looking specifically at the strength of the Ptac promoter, we were surprised by how far away individual transposon insertions were from the bla-neo fusion that were still able to support growth on the antibiotics. For example, in the xds-bla-neo plus IPTG selection, Tn insertions as far as ~8,500 bp (7 genes) away from the fusion were enriched greater than 1,000-fold (Figure II-8c). This implies that the Ptac promoter has the ability to promote the processivity of RNA polymerase for ~8,500 bases. Conversely, in the absence of IPTG, the (weakly) enriched in cis transposons (FEgene <100) were only present in the xds promoter, which is only ~2,800 bp away from the RBS of the fusion.

Of note, the in cis Tn insertions in the emrD-bla-neo plus IPTG selections were only enriched >1,000 fold as far as 4,000 bp away from the fusion. The reason for the difference between the apparent Ptac promoter strengths in the emrD-bla-neo and the other fusion strain selections is unclear. It is possible that chromosomally encoded terminators prevent read-through from Tn insertions further away.

II.2.5.1. cdpA results Insertions in 67 genes were selected for with greater than a 10-fold enrichment when compared to the input in the cdpA-bla-neo minus IPTG selections (Table II-5). Of these 67 genes, 17 exhibited FEgene scores >1,000. Eight genes with >1,000 FEgene scores were identified as statistically significant with average p values <0.05 (VC2368/fexB,

VC2369/fexA, VC2288, VC2415/pdhR, VC0095/ubiC, VC2289/apbE, VC0944/lipB, and

VC1907/cysB). However, only three of these hits passed the Bonferroni correction: fexB, ubiC, and cysB.

112

Table II-5. Putative repressors of cdpA identified in the minus IPTG selection.

Mean Mean fold Wilcoxon enrichment signed-rank Locus Protein ID Gene score test p value VC1842 hypothetical protein 10596 0.33 VC2650 hypothetical protein 9708 0.38 VC0891 exodeoxyribonuclease VII small subunit 8939 0.23 VC2267 transcription antitermination protein nusB 7395 NusB VC2369 aerobic respiration control sensor protein fexB 4248 1.3 x 10-19 ArcB VC2288 hypothetical protein 3323 0.0068 VC2368 two-component response regulator fexA 2967 0.0028 VC2720 putative DNA uptake protein 2648 0.10 VC0307 transcription termination factor Rho rho 2123 0.25 VC0593 2-amino-4-hydroxy-6- 1931 0.33 hydroxymethyldihydropteridine pyrophosphokinase VC2415 transcriptional regulator PdhR pdhR 1881 0.0062 VC0095 chorismate--pyruvate lyase ubiC 1625 1.6 x 10-6 VC0963 2-octaprenyl-3-methyl-6-methoxy-1,4- 1492 0.38 benzoquinol hydroxylase VC0967 hypothetical protein 1381 0.15 VC2289 thiamin biosynthesis lipoprotein ApbE apbE 1239 0.024 VC0944 lipoyltransferase lipB 1084 0.0014 VC1907 transcriptional regulator CysB cysB 1013 1.9 x 10-08 VC0120 porphobilinogen deaminase hemC 978 0.50 VC0096 hypothetical protein 888 0.055 VC1098 acetate kinase 875 0.0026 VC0943 lipoyl synthase 775 0.052 VC1890 NADH dehydrogenase 610 4.1 x 10-5 VC2529 RNA polymerase factor sigma-54 479 0.095 VCA0762 hypothetical protein yieM 469 2.8 x 10-10 VCA0763 hypothetical protein 442 5.3 x 10-16 VC2710 bifunctional (p)ppGpp synthetase II/ 434 0.50 guanosine-3',5'-bis pyrophosphate 3'- pyrophosphohydrolase VC2476 hypothetical protein 396 VC0814 transcriptional regulator, putative 383 0.42 VC1097 phosphate acetyltransferase 373 0.0041 VC0362 elongation factor Tu tuf 296 VC0321 elongation factor Tu tuf 244 VC1021 LuxO repressor protein 230 0.042

113

VC0966 phosphocarrier protein HPr hpr 214 0.17 VC1891 hypothetical protein 189 0.38 VC0130 GGDEF family protein cdpA 176 0.00051 VC0123 frataxin-like protein cyaY 168 0.013 VC0304 guanosine pentaphosphate 164 8.1 x 10-5 phosphohydrolase VC2420 flavodoxin FldB fldB 163 0.053 VC0610 iron(III) ABC transporter, ATP-binding 132 0.19 protein VC2081 zinc ABC transporter, periplasmic zinc- 109 3.1 x 10-5 binding protein VC0125 diaminopimelate decarboxylase 105 0.68 VC2649 serine acetyltransferase cysE 105 VC0305 ATP-dependent RNA helicase RhlB rhlB 91 0.0013 VC0609 iron(III) ABC transporter, permease 91 0.0085 protein VC2156 lipoprotein 91 0.19 VC0378 zinc uptake regulation protein, putative 85 0.13 VC2083 zinc ABC transporter, permease protein 82 0.042 VC2502 DNA polymerase III subunit chi 79 0.25 VC0851 small protein A 76 VC0608 iron(III) ABC transporter, periplasmic 76 0.38 iron-compound-binding protein VC2625 ribulose-phosphate 3-epimerase 67 VC2082 zinc ABC transporter, ATP-binding 60 0.022 protein VC0055 coproporphyrinogen III oxidase 60 1.0 x 10-5 VC0054 Sua5/YciO/YrdC family protein 53 0.22 VC1920 ATP-dependent protease LA 46 0.17 VC1148 DNA-binding transcriptional regulator hexR 43 0.29 HexR VC0002 flavodoxin 38 0.25 VC2735 hypothetical protein 34 0.27 VC0451 hypothetical protein 23 0.19 VC1816 hypothetical protein 19 0.50 VC1079 hypothetical protein 18 0.044 VC1049 LysR family transcriptional regulator 13 0.13 VC1162 hypothetical protein 12 0.38 VC0995 PTS system, N-acetylglucosamine- 12 0.019 specific IIABC component VC2030 ribonuclease E 11 0.28 VC0535 DNA mismatch repair protein 11 3.2 x 10-5 VCA0198 site-specific DNA-methyltransferase, 10 0.016 putative

Genes with no reported p value had fewer than four Tn insertions; the JMP pro 10 software package does not calculate Wilcoxon signed-rank scores when there are fewer than four observations.

114

Clean deletions of fexA, fexB, ubiC, pdhR, and VC2288 and a plasmid insertion knock-out of cysB were all tested for their effect on transcription of cdpA using qRT-PCR

(Figure II-9a, b, and c). Additionally, Heather Kamp, a post-doc in our laboratory previously made a clean deletion of VCA0762, which we also tested for its ability to regulate cdpA (Figure II-9d). Because I had previously seen differences in qRT-PCR results between broth grown and plate grown bacteria, I used plate grown bacteria to isolate RNA for these experiments in order to more closely mimic the selection conditions. The results from these experiments clearly show that these seven genes, which represent nearly half of the 17 genes identified by the minus IPTG genetic selection, do not affect regulation of cdpA.

Analysis of the plus IPTG data from this selection revealed insertions in four regions (genic or intergenic) that fulfilled all of our requirements: VC1841, VC0822,

VC0796/citC, and the intergenic region between VC2467/rpoE and VC2468 (Table II-6).

Since I was unable to determine which genes would be affected by insertions in VC1841,

VC0822, or citC, I only pursued the insertions in the region upstream of rpoE. We hypothesized that these insertions led to increased transcription of rpoE, which is encoded on the negative strand of the genomic DNA. RpoE is an alternative sigma factor that responds to envelope stress and is activated only after inactivation (via DegS and mis-folded OMPs) of its two anti-sigma factors, RseA and RseB (Collinet, et al. 2000,

Grigorova, et al. 2004, Chaba, et al. 2011). Interestingly, cdpA has an RpoE consensus- binding site with a 92% match to the S. Typhimurium consensus sequence (Skovierova, et al. 2006).

115

A) B) 10 10 ) ) rpoB rpoB expression expression expression expression 1 1 cdpA cdpA (normalized to to (normalized (normalized to to (normalized Relative Relative 0.1 0.1 Wild type ∆fexA ∆fexB ∆pdhR Wild type ∆VC2288 ∆ubiC Mutants tested Mutants tested

C) D) 10 10 ) ) rpoB rpoB expression expression expression expression 1 1 cdpA cdpA (normalized to to (normalized (normalized to to (normalized Relative Relative 0.1 0.1 Wild type pGP704::cysB Wild type ∆VCA0762 Mutants tested Mutants tested

Figure II-9. In vitro validation test of putative cdpA repressors. Wild type and mutant strains were grown on LB plates overnight. Bacterial growth was removed from the plates and RNA was isolated using the Qiagen RNEasy kit and Turbo DNA-free DNAse treatment. After cDNA synthesis using random-hexamer primed extension, gene specific primers were used to amplify rpoB and cdpA transcripts in each sample. All cdpA transcript levels were normalized to the housekeeping gene rpoB. Wild type expression levels were set to 1 and mutant expression levels were expressed as relative to wild type. Each panel represents qRT-PCR experiments that were performed on separate days, using different wild type controls. At least 2 biological replicates are represented for each strain. Mean and SEM are shown.

116

Table II-6. Putative activators of cdpA identified in the genetic selection.

Locus or Number of selected Mean fold Putative activator intergenic region Tn insertions enrichment identified ig-VC2467-VC2468a 10 206 rpoE VC1841 3 40 unclear VC0822 2 14 VC0823 - hypothetical VC0796 2 7 truncated VC0796 aIntergenic region between VC2467 and VC2468.

117

It has previously been shown that RpoE is required for V. cholerae growth in LB supplemented with 3% ethanol (Kovacikova and Skorupski 2002a), suggesting RpoE is active under these conditions. To test if induction of RpoE leads to increased transcription of cdpA, I grew the wild type strain in the presence of 3% ethanol, collected

RNA from these bacteria, and performed qRT-PCR. Under these conditions, RpoE transcript levels increased approximately 3-fold when compared with LB growth conditions. However, there was no change in cdpA expression (Figure II-10).

We hypothesized that the increased transcription from the IPTG induced Ptac promoter on the Tn would be much greater than the 3-fold increase obtained by ethanol stress, and might explain why there was induction of cdpA in the former but not the latter experiment. Therefore, I remade the most highly enriched Tn insertion upstream of rpoE in both the wild type and cdpA-bla-neo backgrounds, using only the half of the Tn that contains the Ptac promoter (rpoE::mTn10) (Figure II-11a). Growth of these strains in 1 mM IPTG induced expression of rpoE approximately 30-fold, but cdpA was still not expressed (Figure II-11b). Although I performed this qRT-PCR using plate grown bacteria, which mimics the selection condition, we considered that full activation of

RpoE would require envelope stress. In the genetic selection, the antibiotics may have provided this stress, particularly Cb, which alters cell wall integrity. In order to test this hypothesis, I assayed for expression of cdpA after growth on 1 mM IPTG plates with or without addition of antibiotics (Figure II-11c). Once again, no induction of cdpA occurred. Considering that the cdpA-bla-neo rpoE::mTn10 strain was able to grow on LB

+ 1 mM IPTG + 10 µg/mL Cb + 50 µg/mL Kn, we hypothesized that induction of rpoE

118

) LB + 3% EtOH rpoB

1 Relative expression (normalized to to (normalized

0.1 rpoE expression cdpA expression

Figure II-10. Ethanol stress activation of RpoE does not induce cdpA transcription. Four biological replicates of wild type were grown in either LB or LB supplemented with

3% ethanol at 37°C with aeration. When cells reached an OD600 0.4 – 0.5, RNA was isolated, cleaned of DNA contamination, and converted to cDNA using random-hexamer primers. qPCR using gene specific primers was used to determine the quantities of rpoB, rpoE, and cdpA transcripts in each sample. Transcript levels were normalized to the housekeeping gene rpoB and expression in 3% ethanol was set relative to LB alone. Mean and SEM are shown.

119

A) Ptac

VC2468 aad9 lacIIR rpoE

B) 100 Wild type background cdpA-bla-neo background ) LB rpoB 10 LB + IPTG

1 Relative expression (normalized to to (normalized

0.1 rpoE cdpA rpoE cdpA

C) 10 LB ) LB + IPTG

rpoB LB + IPTG + Cb+ Kn

1 Relative expression (normalized to to (normalized

0.1 cdpA expression Figure II-11. Impact of rpoE::mTn10 on expression of cdpA. A) Design of the rpoE::mTn10, which includes insertion of the inverted repeat (IR), Ptac promoter, lacI, and the spectinomicyin resistance marker (aad9), and was inserted in the promoter region of rpoE. B) Expression levels of rpoE and cdpA were measured using qRT-PCR in the wild type rpoE::mTn10 and cdpA-bla-neo rpoE::mTn10 strains from bacteria grown on either LB plates (white bars) or LB plates supplemented (black bars) with 1 mM IPTG. C) Expression of cdpA was measured using qRT-PCR in the cdpA-bla- neo rpoE::mTn10 strain. RNA was isolated from bacteria grown on LB plates that were supplemented with nothing (white bars), 1 mM IPTG (black bars), or 1 mM IPTG + 10 µg/mL Cb + 50 µg/mL Kn (checkered bars). For both B) and C), transcript levels were normalized to rpoB and set relative to expression in the LB control. At least three biological replicates are shown, with mean and SEM.

120 leads to increased survival on antibiotics. I did not perform a MIC experiment to confirm this. We concluded from these experiments that rpoE is not an activator of cdpA.

In summary, we were unable to identify any regulators of cdpA. I knocked out seven putative repressors of cdpA and tested their role in regulation of cdpA. Despite these genes being highly enriched in the selection output, they did not impact transcript levels of cdpA. Additionally, I tested rpoE as a putative activator of cdpA, but I was once again unable to alter transcription of the late gene.

II.2.5.2. emrD results Analysis of the minus IPTG selection results for the emrD-bla-neo fusion strain revealed four putative repressors with FEgene scores >10 (Table II-7). Two of these genes,

VC0469 and VC2501/pepA, were enriched >1,000 fold and exhibited statistical significance by the Wilcoxon signed-rank test. The plus IPTG selection identified rpoE as a putative activator. Additionally, insertions upstream of VC0988/tppB, VC0469, and

VC0556/gshA were enriched in the plus IPTG selection (Table II-8). In this section I will talk about results from validation with rpoE, VC0469, and pepA.

As described for cdpA, qRT-PCR was used to determine whether rpoE contributes to regulation of emrD (Figure II-12). First, expression of emrD was measured in the wild type strain after growth in LB supplemented with 3% ethanol and compared to expression of the late gene after LB growth (Figure II-12a). Second, expression of emrD was measured in wild type rpoE::mTn10 and emrD-bla-neo rpoE::mTn10 after growth of the bacteria on LB plates or LB plates supplemented with 1 mM IPTG (Figure II-12b).

Finally, emrD transcript was measured in the rpoE::mTn10 strains after subjecting the bacteria to both rpoE induction (1 mM IPTG) and/or antibiotic stress (1 mM IPTG, 10

121

Table II-7. Putative repressors of emrD.

Mean fold Mean Wilcoxon signed- Locus ProteinID Gene enrichment rank test p value VC0469 hypothetical protein 3694 6.4 x 10-6 VC2501 leucyl aminopeptidase pepA 1688 6.5 x 10-7 VC1304 fumarate hydratase, class I, fumA 226 0.27 putative VC0535 DNA mismatch repair protein mutS 13 0.11

122

Table II-8. Putative activators of emrD.

Locus or intergenic Number of selected Mean fold Putative activator region Tn insertions enrichment identified 5 2887 unclear VC0465 7 2373 gshA VC0554 6 1222 tppB ig-VC0988-VC0989 2 895 rpoE ig-VC2467-VC2468

123

A) 10 LB

) LB + 3% EtOH rpoB

1 Relative expression (normalized to to (normalized

0.1 emrD expression

B) 10 Wild type emrD-bla-neo LB

) LB + IPTG rpoB

1 Relative expression (normalized to to (normalized

0.1 emrD expression emrD expression

C) 10 ) LB

rpoB 1 LB + IPTG LB + IPTG + Cb + Kn 0.1

0.01 Relative expression (normalized to to (normalized

0.001 emrD expression

Figure II-12. Induction of rpoE does not activate transcription of emrD. A) Transcript levels of emrD in LB or LB + 3% ethanol grown wild type bacteria were measured using qRT-PCR. B) emrD transcript was measured in wild type rpoE::mTn10 and emrD-bla-neo rpoE::mTn10 strains grown on LB plates or LB plates supplemented with 1 mM IPTG. C) emrD transcript was measured in emrD-bla-neo rpoE::mTn10 grown over night on LB plates, LB + 1 mM IPTG plates, or LB + 1 mM IPTG + 10 µg/mL Cb + 50 µg/mL Kn. In all cases (A, B, and C), transcript levels were normalized to rpoB and set relative to expression in LB. At least three biological replicates are shown, with mean and SEM.

124

µg/mL Cb, and 50 µg/mL Kn) (Figure II-12c). Since these experiments were conducted on the same days as the results shown in Figure II-10, the rpoE control PCR is not shown. As seen with cdpA, emrD transcript levels were not increased in any of the test conditions. On the contrary, induction of rpoE decreased emrD expression approximately

10-fold. This decrease was more severe when rpoE induction was combined with antibiotic treatment. Again, these results point to rpoE having a role in antibiotic-induced stress, rather than a regulator of late genes.

We next turned to VC0469 as a putative repressor of emrD. Despite the strong, statistically significant enrichment of VC0469 in the minus IPTG selection, deletion of the putative repressor had no effect on emrD transcript levels, when compared to transcription in the wild type strain (Figure II-13). VC0469 is encoded on the minus strand of the V. cholerae genome and is in a putative operon with gshB, which encodes glutathione synthetase. We hypothesized that Tn insertions in VC0469 would have disrupted transcription of gshB if the genes are truly co-transcribed, perhaps pointing to gshB as the late gene regulator. The clean deletion of VC0469 that I tested for effect on emrD transcription would leave the operon promoter intact and allow expression of gshB.

Additionally, insertions enriched in the plus IPTG selection were oriented such that the

Tn-encoded Ptac promoter could drive transcription of anti-sense RNA to gshB (Figure II-

14). The Tn insertions in VC0469 were enriched in the plus IPTG selection only if the

Ptac promoter was oriented such that it would divergently transcribe from gshB.

No Tn insertions in gshB were enriched in any of the selections because the input library contained virtually no insertions in the gene. This is likely not due to gshB being an essential gene, as we identified enriched Tn insertions in gshB in the original emrD

125

10 ) Wild type VC0469 rpoB

1 Relative expression (normalized to to (normalized 0.1 emrD expression

Figure II-13. VC0469 does not validate as a repressor of emrD. Wild type and ΔVC0469 were grown over night on LB plates. Bacterial growth was scraped up and RNA was isolated using the Qiagen RNEasy kit. Transcript levels of emrD were measured using qRT-PCR and normalized to rpoB levels. Expression of emrD in the mutant was set relative to the wild type. Two biological replicates of the wild type and four biological replicates of the mutant are represented with the mean and SEM.

126

-IPTG selection

Orientation of the VC0465 VC0466 VC0467 VC0468 VC0469 Tn encoded

Ptac

+IPTG selection

Orientation of the VC0465 VC0466 VC0467 VC0468 VC0469 Tn encoded

Ptac

Key for Tn insertions:

No enrichment in outputs 100 < FEgene < 1000 10 < FE < 100 1000 < FE < 10000 gene gene

Figure II-14. Transposon insertions enriched in VC0465 – VC0469 in the emrD selections. Each vertical line represents a single Tn insertion. Line above the horizontal arrows

(which represent the genes), represent Tn insertions oriented with the Ptac promoter driving transcription on the plus strand, where line below drive transcription on the minus strand. Black lines represent Tn insertions that were present in the input library but not enriched (FEgene < 10) in any of the outputs. Yellow, red, and orange lines represent Tn insertions that were enriched in the output libraries to various extents (see the key).

127

regulator selection. Although I have not made a clean deletion of gshB, I do have a plasmid-insertion knockout of the gene, which I tested for its affect on emrD transcription using qRT-PCR (Figure II-15). Indeed, emrD expression was induced in the gshB mutant.knockout, when compared to expression in the wild type strain. These data support the hypothesis that glutathione synthetase represses emrD.

Since glutathione synthetase does not harbor a DNA binding domain it is unclear how the protein regulates emrD. Interestingly, we identified gshA (which encodes glutamate-cysteine ligase) as a putative activator of emrD. Although we have not validated whether induction of gshA during LB growth can induce transcription of emrD, it is clear that regulation of the late gene is tied to the glutathione metabolic pathway.

Indeed, this idea is supported by results from the original selection, in which four genes with roles in glutathione/sulfur metabolism were identified as regulators of emrD.

The final gene that we validated for emrD was pepA. Again, we used qRT-PCR to compare transcript levels of emrD in the wild type and pepA deletion strains (Figure II-

16). As expected, emrD transcript levels increase approximately 20-fold when pepA is deleted, compared to the transcript levels in the wild type strain. These data support the hypothesis that pepA is a repressor of emrD. PepA is annotated in the V. cholerae

N16961 genome database as a hexameric leucyl aminopeptidase. However, the protein has also been shown to be a DNA-binding transcriptional regulator of the carAB operon in E. coli (Charlier, et al. 1995). Additionally, PepA is known to act as a Xer-site specific recombinase, which contributes to stable inheritance of some multi-copy plasmids, in E. coli (Stirling, et al. 1989). Although it hasn’t been determined if the V. cholerae PepA is

128

100 Wild type ) pGP704::gshB rpoB 10

1 Relative expression (normalized to to (normalized 0.1 emrD expression

Figure II-15. Transcription of emrD is induced in a gshB knockout strain. qRT-PCR was used to measure transcription of emrD in the wild type and gshB knockout strain (pGP704::gshB). RNA was isolated from plate grown bacteria and transcript levels of emrD were normalized to rpoB. Expression of emrD in the mutant was set relative to that seen in the wild type. Two biological replicates of the wild type and five biological replicates of the mutant are represented with the mean and SEM.

129

100 Wild type pepA )

rpoB 10

1 Relative expression (normalized to to (normalized

0.1 emrD expression

Figure II-16. Deletion of pepA increase transcription of emrD. Transcription of emrD was measured in the wild type and ΔpepA strains using qRT-PCR. RNA was isolated from plate grown bacteria. Transcript levels were normalized to rpoB and expression was set relative to the wild type levels. Four biological replicates are represented by the mean and SEM.

130

also capable of binding to DNA, the multi-functional protein has been implicated in repressing the virulence operon, tcpPH, in the classical biotype under non-inducing pH for virulence genes (Behari, et al. 2001).

A mutational analysis of the E. coli PepA has shown that the three functions of the protein – leucyl aminopeptidase, Xer-recombination, and DNA-binding transcription factor – can be studied independently by mutating specific residues (Charlier, et al.

2000). Alignment of the E. coli and V. cholerae PepA sequences reveals that the essential

DNA-binding residues and those involved in leucyl aminopeptidase activity are conserved in V. cholerae. We chose three point mutations to study in the V. cholerae

PepA: M106I, R165H, and G482D. In E. coli, M106I and R165H maintain aminopeptidase activity but are no longer able to repress the carAB operon, whereas

G482D maintains wild type expression of carAB with a 90% drop in aminopeptidase activity (Charlier, et al. 2000).

We successfully constructed the G482D mutant, but not the other two. We hypothesized that expression of emrD in this point mutant would be similar to that of the pepA deletion if PepA regulates emrD via its aminopeptidase activity. However, if PepA regulates emrD via its DNA-binding activity, this point mutant would behave like the wild type. We measured emrD expression in the G482D mutant using qRT-PCR and found that expression of emrD in the point mutant mimics that of the deletion strain, both being increased in transcription when compared with the wild type strain (Figure II-17).

The results strongly suggest that leucyl aminopeptidase activity of PepA, and not the putative DNA-binding activity, is responsible for regulating emrD transcription.

However, further analysis of the point mutant (i.e. to check that it is aminopeptidase

131

10 ) rpoB

1 Relative expression (normalized to to (normalized 0.1 Wild type pepA pepA G482D

Figure II-17. G482D point mutant of PepA behaves like a deletion of pepA. The transcript level of emrD was measured using qRT-PCR in the wild type, pepA deletion, and pepA G482D mutant. RNA was isolated from mid-exponential phase LB broth grown bacteria. Transcript levels are normalized to rpoB and set relative to the wild type sample. One biological replicate is shown for the wild type and pepA deletion strain. The mean and SEM of two independent isolates of the G482D point mutant are shown.

132 activity defective and that the protein is properly folded) should be conducted before further exploration into the mechanism of regulation.

In summary, we identified two repressors of emrD: gshB and pepA. Despite the fact that pepA has been implicated as a DNA-binding protein in other organisms, it appears that both gshB and pepA regulate emrD through indirect mechanisms. Further investigation is needed to determine the precise mechanism through which these regulators function, and whether or not they work in concert.

II.2.5.3. xds results For both fold enrichment scores and Wilcoxon signed-rank test results, the results unambiguously distinguished regulator genes from background genes in the xds-bla-neo selection (Table II-9). The five putative repressor genes identified with FEgene scores

>1000 (pstS, pstC, pstA, pstB, and phoU) were the only genes exhibiting statistical significance using the Wilcoxon signed-rank test. The proteins encoded by pstS, pstC, pstA, pstB, and phoU are part of the Pst/PhoU system, which is an ABC transporter of inorganic phosphate. Although the mechanism is unclear, Pst/PhoU also regulates the activity of the PhoB/R two-component system in response to the extracellular phosphate level. Under high phosphate conditions, Pst/PhoU inhibits autophosphorylation of the histidine kinase PhoR and thus represses activity of PhoB and expression of the Pho regulon. Under low phosphate conditions, PhoB/R is active and induces transcription of low phosphate response genes (Lamarche, et al. 2008). Null mutations in the Pst/PhoU system of V. cholerae, such as our enriched transposon insertions, mimic the low phosphate state resulting in constitutive activation of PhoB (Pratt, et al. 2009). Therefore,

133

Table II-9. Putative repressors of xds.

Mean Wilcoxon Mean fold signed-rank test p Locus Protein ID Gene enrichment value VC0724 phosphate ABC transporter, pstC-1 3559 1.2 x 10-32 permease protein VC0727 transcriptional regulator phoU 2738 3.3 x 10-7 PhoU VC0725 phosphate ABC transporter, pstA 2679 3.5 x 10-27 permease protein VC0726 phosphate ABC transporter, pstB 2648 1.8 x 10-14 ATP-binding protein VC0721 phosphate ABC transporter, pstS 1710 5.1 x 10-19 periplasmic phosphate- binding protein, putative VC0668 DNA mismatch repair mutH 16 0.38 protein

134

we hypothesized that PhoB is an activator of xds, while the pst/phoU genes are indirect repressors of the late gene.

We identified two insertions in the 5' end of phoR that were enriched in the plus

IPTG selection (Table II-10 and Figure II-18). The first insertion was located at 117 bp downstream of the initiator AUG and was enriched 174-fold. The other insertion, located at basepair 245, was enriched 1,722-fold. In both cases the transposon was oriented such that its Ptac promoter could drive expression of a 5’-truncated phoR gene in an IPTG- dependent manner. Although there were 57 insertions in both orientations throughout the phoR coding sequence in our input transposon library, we only identified these two insertions enriched in the output. Using the domain prediction algorithms built into

BLAST (NCBI), we determined that both transposon insertions could promote expression of truncated PhoR proteins that lack their periplasmic and transmembrane domains but still contain all domains thought to be important for phosphotransfer. In the more highly enriched mutant, an in-frame initiator codon was created at the transposon-chromosome junction. We hypothesize that expression of PhoR without its transmembrane domain leads to constitutive activation of the sensor kinase likely by removing the protein from the membrane and therefore preventing its interaction with the membrane-bound

Pst/PhoU inhibitor complex. Indeed, work done in E. coli demonstrated that PhoR truncation mutants lacking the transmembrane domain are able to activate PhoB in an unregulated manner (Yamada, et al. 1990).

135

Table II-10. Putative activators of xds.

Locus or Putative intergenic Number of selected Mean fold activator region Tn insertions enrichment identified VC0720 2 948 phoR

136

Ptac 2 ** Transposon 1

0 insertions in 1 13021299

phoR -1

-2

Ptac

PhoR 1 TM PAS HK 433

Figure II-18. Transposon insertions enriched in phoR. Each vertical line indicates the position of a single transposon insertion in phoR. Lines above phoR indicate that the orientation of the transposon carried Ptac promoter is the same as that of PphoR and the converse is true of the lines below phoR. The location of the known protein domains encoded by phoR are indicated: TM – predicted transmembrane domain, PAS – Per/Arnt/Sim, and HK – Histidine kinase domain. Two transposon insertions were selected specifically in the selection done in the presence of IPTG, these insertions are indicated by arrows. Transcription from the Ptac promoter carried on the transposon would lead to a truncated PhoR protein lacking a functional transmembrane domain.

137

To validate repressor activity of Pst/PhoU on xds I performed qRT-PCR on xds transcripts. I compared transcript level in wild type cells to that in a ΔphoU background

(Figure II-19a). I reasoned that deletion of phoU should have the same effect as deleting any of the other pst genes, resulting in constitutive activation of PhoB. In order to closely mimic the conditions of our genetic selection, I collected RNA for these experiments from LB agar plate grown bacteria. In support of our selection results, deletion of phoU resulted in an increase of xds expression approximately 90-fold, when compared to wild type. Furthermore, when we deleted phoB in the ΔphoU background, expression of xds was restored to the wild type level, confirming that the phoU and phoB are epistatic. As a control for these experiments, we measured expression of phoX, a V. cholerae pho regulon gene (von Kruger, et al. 2006, Pratt, et al. 2009), in wild type, ΔphoU, and

ΔphoUΔphoB strains (Figure II-19b). As expected, transcript levels of phoX followed the same pattern as xds. Together, these results support our conclusion from the selection results that PhoU is a repressor of xds, and that repression activity is dependent on the presence of the PhoB response regulator.

Since PhoB is active under low phosphate conditions, we hypothesized that xds would be induced in a phosphate-limiting environment when compared to growth in phosphate replete conditions. Recently, induction of xds under phosphate limiting conditions was reported in another V. cholerae El Tor strain, C6709 (Seper, et al. 2011).

These authors assayed expression of xds under high and low phosphate conditions using an xds-phoA transcriptional fusion reporter strain and measured alkaline phosphatase levels as a read-out for xds transcription. These results are confounded by the fact that V.

138

A) B)

xds ) ) 100 100 phoX rpoB rpoB 10 10

1 1 (normalizedto (normalizedto 0.1 0.1 Relative expression of of expression Relative Relative expression of of expression Relative

phoU phoB phoU phoB ∆ ∆ ∆ ∆ wild type wild type phoU phoU ∆ ∆

Figure II-19. PhoU is an indirect repressor of xds. Expression of A) xds and B) phoX were measured in wild type, ΔphoU, ΔphoUΔphoB strains using qRT-PCR. Transcript levels were normalized to the housekeeping gene rpoB, and set relative to the wild type strain. Bacteria were grown on LB agar plates at 37ºC for two hours to mimic selection conditions. Bacteria were subsequently removed from the plates and RNA was isolated. Four biological replicates are represented for each strain. Mean and SEM are shown.

139

cholerae has its own functional alkaline phosphatase that is regulated by PhoB

(Majumdar, et al. 2005, Pratt, et al. 2010). Therefore, under low phosphate conditions, the combined activity of the endogenous phosphatase and the reporter fusion was measured, and the relative contribution of each gene to the final readout was not assessed.

Here, I used qRT-PCR to directly measure transcript levels of xds in high and low phosphate conditions. I grew wild type V. cholerae in MOPS minimal media supplemented with 10 mM (high) or 0.1 mM (low) KH2PO4 as a phosphate source. As expected, I found that xds was induced under low phosphate conditions in the wild type background approximately 60-fold, when compared to growth in replete phosphate

(Figure II-20a). When I deleted phoB, expression of xds in low phosphate conditions dropped back to the level of xds expression seen in the wild type when it was grown in high phosphate. Expression of xds under 0.1 mM phosphate in the absence of phoB was restored when phoB was expressed from the Ptac promoter on the pMMB67EH vector. No complementation was observed when the empty vector was used. Expression of phoX was also measured in high or low phosphate MOPS media as a control for these experiments (Figure II-20b). Transcript levels of phoX followed the same pattern as xds expression, suggesting that PhoB is active in the wild type and complementing strains under low phosphate. However, the expression of phoX under low phosphate conditions in the phoB deletion strain was higher than expected (i.e. 31-fold increase in expression compared with the expression of phoX in the wild type strain grown under high phosphate conditions). This suggests that there may be other regulators active under low phosphate growth that contribute to induction of phoX. These results, presented in Figure

II-20, further support our hypothesis that expression of xds is induced in phosphate

140

#!+4:+ !"#+4:+ =7 -<0/-<'(% -<0/-<'(% + # " !

7 #!! + 2 ' 0 + & 1 % $ 0 + ) 0

/ #! ( + / 6 % . % - 5 ) , & % ' # + % 4 . * ) 0 ( 1 ' 3 &

% !"# $ 7 %(&' %(&' %(&' ¨ ¨ ¨ 8)&6+(9-% 8)&6+(9-% )%(&' 3-::;7

3-::;

;7 #!+4:+ !"#+4:+ -<0/-<'(% -<0/-<'(% * & (

7 #!!! % + ' 2 & 0 + % $ 1 #!! + 0 0 ) ( + / 6 / % % #! . 5 ) - & , ' % +

4 # . % 0 * ) ( 1 3 '

& !"# %

$ 7 %(&' %(&' %(&' ¨ ¨ ¨ 8)&6+(9-% 8)&6+(9-% )%(&' 3-::;7

3-::;

Figure II-20. Induction of xds by low phosphate is dependent on PhoB activity. The level of xds transcript was measured in various V. cholerae strains. Three biological replicates of each strain were grown overnight in MOPS media supplemented with either

10 mM or 0.1 mM KH2PO4 (phosphate). For plasmid containing strains, 1 mM IPTG was used to induce expression from the Ptac promoter. Transcript levels were normalized to rpoB and set relative to expression in the wild type when grown under high phosphate conditions. The mean and standard deviation are shown.

141 limiting conditions, and transcription of xds depends (perhaps solely) on the presence of activated PhoB.

II.2.6. Investigation of the nature of PhoB regulation of xds We were next interested in defining whether PhoB binds directly to the xds promoter in order to activate the gene. The binding of PhoB to promoter regions has been well studied in E. coli, and a consensus-binding site, or Pho box, has been described as two direct repeats separated by a 4-bp AT rich region: CTGTCAT-A(AT)A(TA)-

CTGT(CA)A(CT) (Makino, et al. 1993, Blanco, et al. 2011). The xds promoter region does not contain any sequence resembling an E. coli Pho box. However, the interaction of

PhoB with promoter sites in V. cholerae has not been well described. V. cholerae PhoB binds to three sites within the phoBR promoter, all of which are very similar to the E. coli

Pho box (Diniz, et al. 2011). However, V. cholerae PhoB also interacts with a region of

DNA within the tcpPH promoter that does not contain a putative Pho box (Pratt, et al.

2010). Therefore, despite the lack of an identifiable Pho box in the xds promoter, PhoB may still directly bind the xds promoter and regulate the gene.

Many methods are described for testing transcription factor – promoter interactions. We chose to study the interaction of PhoB with the xds promoter using electrophoretic-mobility shift assays (EMSAs). The DNA-binding activity of PhoB requires phosphorylation and subsequent dimerization (McCleary 1996). Rather than purifying the wild type protein, which needed to be phosphorylated before testing in the binding assays, we chose to use a constitutively active mutant of PhoB. This constitutively active mutant (PhoBCA; PhoB D10A/D53E) was previously determined to be in an active conformation when expressed in E. coli, where the allele was originally

142 described (Arribas-Bosacoma, et al. 2007). Additionally, the mutant was successfully used in V. cholerae to show binding of PhoB downstream of the tcpPH promoter (Pratt, et al. 2010). Despite the in vitro assays performed with the mutant PhoB protein, which support that PhoBCA maintains DNA-binding activity, the functionality of PhoBCA was not determined in either E. coli or V. cholerae. Therefore, we assessed the activity of this constitutively active mutant in V. cholerae by replacing the chromosomal phoB with phoBCA and measuring expression of phoX and xds in this strain as compared with the wildtype and a ΔphoU strain (Figure II-21). PhoBCA is constitutively active in V. cholerae, as seen by its ability to induce expression of phoX and xds, with similar levels of expression as seen in the ΔphoU mutant, when compared with the wild type expression. Thus, we felt that the PhoBCA protein could act as a biologically relevant protein for our EMSA studies1.

The EMSA probes were continuously labeled with Cy-5 dCTP in the presence of all four natural deoxynucleotides, such that 3-4 nucleotides/ DNA molecule would be labeled. The only exception was the tcpPH control probe, which was end-labeled with

Cy-3 as described in (Pratt, et al. 2010). The probes used in this study are described in

1 We made several attempts to use the wild type PhoB protein in the EMSA studies. After some initial trouble obtaining purified protein, we used either acetyl phosphate or the non-cognate Histidine kinase, VieS, as a phosphodonor to activate the protein. Both acetyl phosphate and non-cognate Histidine kinases have been used to phosphorylate bacterial proteins, including PhoB (McCleary and Stock 1994, McCleary 1996, Fisher et al 1995, Yamamoto et al 2005, Townsend et al 2013). In order to determine if the phosphorylation protocols were working, we utilized both the Phos-tag gel system, which preferentially retards the movement of phosphorylated proteins through an acrylamide gel, and MALDI-TOF mass spectrometry, which would reveal an increase in PhoB mass of approximately 80 Da upon phosphorylation. However, we were unable to detect any difference between the treated and untreated PhoB protein. 143

!" .,,,, *

) .,,, ( " 2 8 % 6 # 2 ! 7 + 2 6 % 6

5 .,, ( 2 5 ' + 4 + * ; % & 3 0 +

2 ., : + 4 1 % 6 ( 7 0 9 &

+ . /

,-. $%&'()*+ !"#$ !"#%&'

#" .,,,, 2 , #

" .,,, " ! 2 % 8 # 6 2 ! + 7 2 6 6 % .,, ( 2 5 5 ' + + 4 ; % * & 3 0 ., + 2 : 4 + 6 1 % ( 7 9 0 &

+ . /

,-. $%&'()*+ !"#$ !"#%&'

Figure 11-21. PhoBCA is a constitutively active protein in vivo. Transcript levels of A) xds and B) phoX were measured in wild type, ΔphoU, and phoBCA. For qRT-PCR, RNA was collected from 0.5 mL of bacteria grown in LB broth to

OD600 = 0.28-0.35. The expression levels were normalized to rpoB in each sample, and set relative to wild type expression. Four biological replicates were examined. The mean and SEM are shown.

144 detail in Figure II-22. A 440 bp product surrounding the xds promoter region was chosen as the query probe (probe 1). An internal region of the bla gene was used as a negative control. Two positive control probes were included: the V. cholerae phoBR promoter fragment used to describe the auto-regulatory effect of PhoB (Diniz, et al. 2011) and the tcpPH promoter (Pratt, et al. 2010). Diniz and colleagues (2011) described three PhoB binding sites in the phoBR promoter, which resulted in three distinctly shifted bands; the positive control probe used here contains all three sites. However, since the binding reaction conditions we used were different from those used by Diniz and colleagues (e.g. wild type PhoB vs. PhoBCA, radiolabeled vs. fluorescently labeled probes, and different buffer conditions), we did not necessarily expect to see the same binding profile.

However, we did use the same reaction conditions as those used to show binding of PhoB to the tcpPH promoter probe. Ayman Ismail, a post-doc in our laboratory who performed the tcpPH EMSA experiments, helped with purification of the PhoB protein and the initial gel shift experiments described here.

No binding of PhoBCA (up to 10 µM) was observed to the bla probe (Figure II-

23a). PhoBCA bound both the phoBR promoter and tcpPH promoter probes (Figure II-

23b,c). Three distinct shifts were observed for the phoBR promoter fragment (shift 1 –

2.5 µM; shift 2 – 5 µM; shift 3 – 10 µM), which was expected and consistent with the existence of three binding sites for PhoB. However, the pattern of binding was different than that originally described (Diniz, et al. 2011). Based on the previous report, we expected PhoBCA to bind to the tcpPH promoter fragment at a 1 µM concentration.

However, 10 µM of PhoBCA was required to induce shifting of this probe. It is likely that our protein preparation contained a mixture of both active and inactive protein. In support

145

A) -367 -274 -162 -94 -66 +1 +73 xds promoter Probe 1 Probe 2 Probe 3 Probe 4

B) -253 +1+10 phoBR promoter

C) -87 +1 +50 tcpPH promoter

D) +200 +502 +603 bla probe

Figure II-22. Probes used for EMSA experiments. The relevant region of the gene and/or promoter region is represented by a black bar. The tick marks indicate distance from the +1 site, which here indicates the translational start site (i.e. the A in the AUG start codon). The transcriptional start sites are not indicated in the figure, as they have not been defined for xds, phoBR, or tcpPH. Grey bars represent each probe, drawn to scale.

146

!" bla probe #" phoBR probe

#)*'+, -.)(/

&'()*'+, &'()*'+, -.)(/ -.)(/

01)#$!,234" 0 1 2.5 5 10 01)#$!,234" 0 1 2.5 5 10

$" tcpPH probe %" xds probe

#)*'+, #)*'+, -.)(/ -.)(/

&'()*'+, &'()*'+, -.)(/ -.)(/

01)#$!,234" 0 1 2.5 5 10 01)#$!,234" 0 1 2.5 5 10

Figure II-23. PhoBCA binds to the xds promoter. Increasing concentrations (0-10 µM) of PhoBCA were incubated with 5 nM of each labeled probe: A) bla negative control, B) phoBR positive control, C) tcpPH positive control, and D) xds promoter probe 1. The binding reaction was conducted in the presence of 0.1 mg/ml sheered calf thymus DNA in the EMSA buffer. Binding reactions were incubated in the dark at room temperature for 20 minutes and directly loaded on a 6% polyacrylamide native TBE gel and run at 150v for 50min in 1x TBE. Bands were visualized by scanning with a Fuji imager using the DGR1-Cy-5 filter.

147

of this, the concentration of PhoBCA required to shift phoBR was similar to that required of the unphosphorylated wild type protein to shift the same probe (Diniz, et al. 2011). We used the same purification protocol as Pratt and colleagues and much of the protocol was performed by Ayman Ismail, who originally purified the protein, so we did not expect huge differences in the protein activity. We did attempt to purify the protein several times, but each time the same level of activity was observed.

PhoBCA exhibited binding to the xds promoter at a concentration of 10 µM

(Figure II-23d), which was similar to the positive control. In order to roughly map the binding site of PhoB to the xds promoter, smaller PCR fragments were used as probes in the binding assay (Figure II-24). Complete binding of PhoBCA was observed to probe 3, which extended from the translational start site to 162 base pairs upstream. Probes 2 and

4 exhibited minimal binding. Therefore, we concluded that the binding site is located within probe 3. However, the binding pattern of Probe 3 and Probe 1 were slightly different, such that an additional binding site may exist between 274 and 367 base pairs upstream of the translational start site. Use of DNase I footprinting analysis could further define the binding site of PhoB within the xds promoter. Finally, we show that binding of

PhoBCA to the xds promoter can be competed out by unlabeled xds probe, but not an unlabeled bla probe (Figure II-25).

These EMSA results taken together with the fact that we did not identify any evidence of additional xds regulators in our genetic selection, lead us to conclude that

PhoB is indeed the sole direct regulator of xds. However, it is possible that another direct regulator of xds exists but is essential and cannot be disrupted by transposon mutagenesis.

148

-367 -274 -162 -94 -66 +1 +73 xds promoter

PhoB (µM) 0 1 2.5 5 10

Probe 1

PhoB (µM) 0 1 2.5 5 10

Probe 2

PhoB (µM) 0 1 2.5 5 10

Probe 3

PhoB (µM) 0 1 2.5 5 10

Probe 4

Figure II-24. Rough estimation of the PhoB binding site in the xds promoter. Overlapping fragments of the xds promoter were tested in the EMSA binding reactions. The number markings on the xds promoter represent distance from the translational, not transcriptional, start site (i.e. the A in the starting AUG start codon is +1). Probe 1 extends the entire length of the promoter (-367 to +73). Probe 2 extends from -274 to -94. Probe 3 extends from -162 to +1. Probe 4 extends from -66 to +73. For EMSAs, increasing concentrations (0-10 µM) of PhoBCA were incubated with 5 nM of each labeled xds promoter probe. The binding reaction was conducted in the presence of 0.1 mg/ml sheered calf thymus DNA in the EMSA buffer. Binding reactions were incubated in the dark at room temperature for 20 minutes and directly loaded on a 6% polyacrylamide native TBE gel and run at 150v for 50min in 1x TBE. Bands were visualized by scanning with a Fuji imager using the DGR1-Cy-5 filter. The dark red circle spanning -66 to -100 represents the most probably PhoB binding site. The light red circle spanning -300 to -330 represents a putative second PhoB binding site.

149

PhoB (µM) 0 2 4 6 8 10 12 8 8

xds Cy5 probe a b

Figure II-25. Competition binding assays with PhoBCA. Increasing concentrations (0-12 µM) of PhoBCA were incubated with the labeled probe 1 (5 nM) in the presence of 0.1 mg/ml sheered calf thymus DNA in the EMSA buffer. Specificity of the binding reaction was determined by competition with (a) 200-fold excess cold xds probe or (b) 200-fold excess cold non-specific probe (bla). Binding reactions were incubated in the dark at room temperature for 20 minutes and directly loaded on a 6% polyacrylamide native TBE gel and run at 150v for 50min in 1x TBE. Bands were visualized by scanning with a Fuji imager using the DGR1-Cy-5 filter.

150

Alternatively, PhoB-repressed redundant repressors may regulate xds, where each repressor on its own is sufficient to inhibit xds transcription. Further electro-mobility shift assays and promoter pull-downs could address the question of what directly regulates xds. Additionally, since we were unsure of the activity of our PhoBCA preparation, we are unable to speak to the precise affinity of PhoB for the xds promoter. Finally, footprinting experiments

II.2.7. PhoB as a regulator of additional V. cholerae late genes Having identified PhoB as a regulator of xds we hypothesized that PhoB would regulate additional late genes. Indeed, when we began work on this project two late genes, acgB and phoX, were described as part of the Pho regulon in the classical biotype

(von Kruger, et al. 2006, Schild, et al. 2007, Pratt, et al. 2009). Since the late genes were described in the El Tor biotype, we tested whether this regulation occurs in El Tor. Our qRT-PCR experiments earlier in this chapter verify the regulation of phoX by PhoB in our wild type strain. We measured transcript levels of acgA (encoding a c-di-GMP phosphodiesterase) and acgB (encoding a c-di-GMP diguanylate cyclase) in the wild type and ΔphoB strains after growth in minimal medium supplemented with varying phosphate concentrations (Figure II-26a,b). Both genes were induced in the wild type under phosphate limiting conditions (0.1 mM KH2PO4), but not in the phoB deletion strain. Transcript levels of phoB were measured in the wild type in all growth conditions to confirm the activation of the phosphate starvation response (Figure II-26c). As expected, phoB transcript levels were high only in the 0.1 mM KH2PO4 medium.

151

!" ./// '()*%%&'(%))*+, 0 " ,

$ @*9?0;A'% + * # ) ! .// & )

0 !"#$ % + ( ; 0 ' ? & %

% ./ 0 > * 9 % < : * ; 1 : 9 ( . % + , 8 =

/6. # 5 5 5 5 -

.012034 ./012034 /67012034 /6.012034 #" .//

0 '()$%%&'(%))*+,

" @*9?0;A'% , $ + *

# !"#$ ) ! & ) 0 %

+ ./ ( ; 0 ' ? & % % 0 > * % 9 < : *

; . 1 : ( 9 + % , 8 =

/6. # 5 5 5 -

.012034 ./0120345 /67012034 /6.012034

$" .// !"#$%%&'(%))*+, 0 " , $ + * # ) ! & ) 0

% ./ + ( ; 0 ' ? & % % 0 > * 9 % < : *

; . 1 : 9 ( % + , 8 =

/6. # 5 5 5 5 -

.012034 ./012034 /67012034 /6.012034

Figure II-26. PhoB induces acgAB under low phosphate conditions. Expression of A) acgA, B) acgB, and C) phoB were measured using qRT-PCR. RNA was collected from bacteria grown overnight in MOPS-glucose minimal medium supplemented with 10 mM KH2PO4, washed twice in no phosphate MOPS-glucose medium, and seeded into 2 mL MOPS-glucose cultures with varying amounts of

KH2PO4. Alternatively, cultures were grown overnight in LB and then back diluted in the same medium. All cultures were grown to an OD600 = 0.2 to 0.3. Transcript levels were normalized to rpoB and set relative to expression of the wild type in LB. The mean and standard error of two biological replicates is shown.

152

To determine if the regulation of acgAB by PhoB was mediated through direct binding of the response regulator, we performed EMSAs using a Cy-5 labeled acgAB promoter fragment and the PhoBCA protein (Figure II-27a). The gel shifts performed here were conducted on the same day as those presented in Figure II-22, and thus the control probes (bla, phoBR, and tcpPH) are not shown again. Since the regulation of acgAB by

PhoB was first described in the classical biotype, we assessed binding of PhoB to both the classical and El Tor promoter regions (Figure II-27b). Indeed, addition of PhoBCA to both acgAB promoter probes resulted in a shift in mobility of the probes. Addition of 5

µM PhoBCA resulted in an incomplete shift, whereas complete shifting of the probe was seen at 10 µM PhoBCA. These concentrations are similar to what was observed for the xds probe (Figure II-23d). These results, taken together with the qRT-PCR data demonstrate that PhoB activates a second late gene, acgB, in addition to xds.

In order to identify additional PhoB regulated late genes, we used the online promoter and transcription factor binding site prediction tool, Softberry BPROM

(Salamov 2011). Softberry BPROM searches entered genetic sequences for predicted -10 and -35 sites, as well as comparing the sequences to a databank of described E. coli transcription factor binding sites. We found six late genes with predicted Pho boxes in their promoter regions (Figure II-28a). None of these six genes were reported as induced by PhoB in a proteomic study, however, this study was not comprehensive, as it did not identify either Xds or AcgB as Pho regulated proteins (von Kruger, et al. 2006). Based on the fact that PhoB induces xds, we hypothesized that PhoB would also induce the other late genes. On the contrary, the location of the predicted Pho binding sites for VC0044,

VC0203, VC0280, and VC0353 suggested that the response regulator might repress these

153

acgAB probe 456778465 ;5'<$) #8$9:(* #8$9:(* +$%"&' ()$#*

!"#$%"&' ()$#*

,-$+./'0123 0 1 2.5 5 10 0 1 2.5 5 10

Figure II-27. PhoB binds directly to the acgAB promoter. Increasing concentrations (0-10 µM) of PhoBCA were incubated with 5 nM of each labeled probe. The binding reaction was conducted in the presence of 0.1 mg/ml sheered calf thymus DNA in the EMSA buffer. Binding reactions were incubated in the dark at room temperature for 20 minutes and directly loaded on a 6% polyacrylamide native TBE gel and run at 150v for 50min in 1x TBE. Bands were visualized by scanning with a Fuji imager using the DGR1-Cy-5 filter. Control reactions as shown in Figure II-23 were run on the same day as these shifts.

154

/012 )5 -&( -.# !"##$$ 314

/012 -&( -.# !"#%#& 314

/012 -&( -.# !"#%'# 314

/012 -&( -.# !"#&(& 314

/012 !"#&+, -&(314 -.#

/012 /012 -&( -.# !")#*(( 314 314

65 .# !"#&(& .# !")#*(( > > H H 8 8 ; ; C C 9 9 > > E $ E $ 2 2 # # F F ! ! 1

% . 1 8

% . 2 8 2 B B 1 1 B ; B ; > > E E = = 4 4 G G #?. #?. 789:2;<=> !"#$ 789:2;<=> !"#$ D;EC8F D;EC8F

.#2@A2=01B=0C;> #?.2@A2=01B=0C;>

Figure II-28. PhoB regulation of additional late genes. A) Using the Softberry BPROM online tool putative Pho boxes were identified in six late genes. Shown are: the predicted -10 and -35 promoter sites, the open reading frame of each gene (annotated with the black arrow), and putative Pho boxes (outlined by a PhoB box). VCA0955 harbors two putative Pho boxes. B) Transcript levels of VCA0353 and VCA0955 were measured using qRT-PCR in the wild type and ΔphoB strains. RNA was collected from bacteria grown overnight in MOPS minimal medium supplemented with either 0.1 mM or 10 mM KH2PO4. Transcript levels were normalized to the housekeeping gene rpoB. The mean and SEM of three biological replicates is shown for each condition.

155

genes. In order to determine what effect PhoB has on transcription of the six late genes, we used qRT-PCR to measure transcript levels of each gene in the wild type strain after growth in phosphate replete and phosphate limiting conditions. A phoB deletion strain was used as a control in the phosphate limiting conditions. Due to technical difficulties we were only able to measure transcription of two genes: VC0353 and VCA0955 (Figure

II-28b). PhoB represses both of these genes such that they are induced maximally in the phoB deletion strain. Surprisingly, VC0353 is induced under low phosphate conditions even in the presence of phoB. This suggests that either a high phosphate repressor or low phosphate activator besides PhoB additionally controls this gene. Conversely, VCA0955 is repressed in the wild type low phosphate condition, but fully induced in the phoB deletion strain regardless of the phosphate concentration in the medium. This may imply that VCA0955 is repressed after activation of PhoB via a non-cognate Histidine kinase.

II.3. DISCUSSION

II.3.1. Discussion of the novel genetic selection method When I began work on my thesis in 2008 we were excited about the idea of combining a traditional antibiotic genetic selection with high-throughput selection. At the time we considered that the addition of high-throughput sequencing to the antibiotic selection would allow for a clearer separation of true hits versus background as well as the knowledge of all genes and intergenic regions represented in the selection.

Additionally, the high-throughput aspect would allow for rapid, all-at-once sequencing of transposon insertions. This would be quite an improvement over the slow process of isolating limited numbers of individual colonies, isolating genomic DNA from the colonies, performing arbitrary-primed PCR of the Tn-genomic DNA junction, sequencing

156 each PCR product, and analyzing each sequencing result separately. We found that 50% to 99% of the Tn insertions able to grow on the minus IPTG selection plates represented insertions within true regulators of emrD and xds. Thus, sequencing a few dozen individually picked colonies, rather than high-throughput sequencing the entire population, may have been sufficient to confidently identify the most highly enriched insertions, however, we would not have obtained quantitative information nor been sure of identifying the moderately enriched insertions. On the other hand, the plus IPTG selection plates were overrun with bacteria harboring Tn insertions in cis to the bla-neo fusions. We only identified one true hit from the plus IPTG selections, phoR as a regulator of xds. Importantly, we would not have been able to identify this regulator without the high-throughput sequencing because phoR was but a small fraction of the enriched hits.

Since the high-throughput nature of the genetic selection allowed us to determine the frequency at which each Tn insertion was represented in the input and output pools, we hypothesized that we would be able to identify and differentiate both weak and strong regulators. We reasoned that Tn insertions in or near weak regulators would be represented at frequencies above the background, but lower than insertions in or near strong regulators. However, these classes of genes did not readily emerge from our selection data due to various sources of 'noise' such as variable selection on antibiotics, false positive hits, non-saturating Tn libraries, etc. As an alternative method to identify weak and strong regulators, we made use of three different antibiotic concentrations for each selection. The idea was that a strong regulator would be identified on the selection plates with even the highest amount of antibiotics, whereas a weak regulator would only

157 be identified on plates with a low level of antibiotics. In other words, a strong regulator would result in higher levels of bla-neo expression, which could potentially lead to a greater ability to survive on the antibiotic plates. Although we did identify some genes that exhibited up to 1000-fold differences in enrichment between the different antibiotic conditions, we were not necessarily able to identify a pattern between fold enrichment and antibiotic concentration. However, these results are not surprising considering that the products of bla and neo are enzymes, and therefore, an increase in the resistance enzymes would not necessarily translate to a linear increase in the MIC of the antibiotics unless the antibiotic substrate was in vast excess of the resistance enzyme. In the end, we ended up treating all antibiotic selection conditions as replicates, despite the fact that the conditions were not identical.

Therefore, taken together, in our particular case, the addition of high-throughput sequencing to the traditional antibiotic selection did not add major improvement to the method. However, due to high degree of separation between the fold enrichment of true hits versus the true enrichment of the background insertions, we were able to quickly and unambiguously differentiate the regulators from the background. Additionally, the deep sequencing of both our input and output libraries allowed us to clearly identify which Tn insertions were represented in our libraries. Since our original replicate Tn libraries exhibited a low level of overlap, we know that our selections were not saturating, and thus some potential regulators may have escaped detection.

In comparing the results from the first and second versions of the genetic selection, it became obvious that the most important aspect of the experiment was the stringency of the selection. In the first version of the selection we used only the bla

158 resistance marker as readout for induction of cdpA, emrD, and xds, and this resulted in a large number of false positives in the data. The addition of a second antibiotic resistance marker, neo, greatly increased the success of the selection by knocking down the background growth and reducing false positives that simply endowed V. cholerae with increased antibiotic resistance without increasing transcription of the reporter gene fusions. As I discussed above, the high-throughput aspect of the genetic selection did not make a difference in the minus IPTG selections; high-throughput sequencing didn’t help pull out true hits from false positives unless the antibiotic selection pressure was strong.

II.3.2. cdpA selection discussion The second-generation genetic selection worked very well for identifying regulators of both emrD and xds. However, we were unable to identify any regulators of cdpA, despite the fact that several genes were identified as >1000 fold enriched in the cdpA-bla-neo minus IPTG outputs with statistical significance according to the Wilcoxon signed-rank test. Since we did not attempt to validate all of the highly enriched genes, it is possible that we missed identifying the regulator(s) of cdpA. However, from our qRT-

PCR validation, we know that at least 7 highly enriched genes in the cdpA-bla-neo selections were false positive hits. I did not test whether these genes contribute to antibiotic resistance independent of the bla-neo fusion, and therefore I cannot concretely address whether they are random noise or biologically significant. However, as discussed above, we designed the genetic selection such that the concentration of antibiotics that we used would yield no growth when the non-Tn mutagenized bla-neo fusion control strain was plated, but 100-1,000 colonies when the Tn library was plated. Thus, we required that at least some colonies grew on the plate. Additionally, I had a particularly tough time

159 finding a concentration of antibiotics that yielded reproducible numbers of colonies with the cdpA-bla-neo strain, perhaps supporting the idea that anything that grew was merely random background.

The inability to identify a regulator of cdpA could be explained a number of ways.

The most simplistic explanation is that cdpA does not have a regulator, but that the gene is expressed at a constant level at all times. However, this seems unlikely since recombination-based in vivo expression technology (RIVET) analysis of cdpA expression showed no expression of the gene during growth in LB, and induction of the gene in 63% of the population by 24 hours post inoculation of an infant mouse (Schild, et al. 2007).

The second explanation stems from the idea that we designed an in vitro genetic selection in order to identify regulators of in vivo induced genes. Thus, if transcription of cdpA required the presence of a specific metabolite, perhaps to activate an inducer of the gene, and if that metabolite was absent from our in vitro selection conditions, then we would not have been able to identify the gene even in the plus IPTG selection. For example, we identified PhoB as a regulator of xds, however insertions upstream of phoB were not enriched in the plus IPTG selection since PhoB requires phosphorylation by

PhoR in order to be active as a transcriptional regulator. Therefore, increasing PhoB protein was ineffective without having also put the strain in a low phosphate condition. In the case of the xds selection we were lucky since disruption of the pst/phoU genes constitutively activates PhoB, and we also identified insertions that resulted in a constitutively active PhoR. However, genes analogous to pst/phoU and phoR may not exist for cpdA.

160

Finally, our genetic selection was designed to identify protein regulators only, and therefore, it is possible that cdpA is regulated via a different mechanism such as epigenetic control of transcription via Dam methylation or direct regulation by a riboswitch or small RNA (sRNA). For example, some bacteria, including V. cholerae are known to utilize riboswitches that bind c-di-GMP in their aptamer domains to regulate transcription or translation of various genes. Specifically, the Vc2 class I riboswitch of V. cholerae regulates the expression of the competence regulator, tfoX (Meibom, et al. 2004,

Sudarsan, et al. 2008). As a c-di-GMP phosphodiesterase-encoding gene, transcriptional regulation of cdpA via a c-di-GMP binding riboswitch would be plausible. We have not investigated this idea further, but a first step would be to look for a Vc2-like sequence or structure in the DNA upstream of cdpA.

II.3.3. emrD selection discussion Two putative regulators of emrD were identified: pepA and gshB. Although PepA has been shown to directly bind DNA in E. coli (Charlier, et al. 1995), neither V. cholerae PepA – encoding a leucyl aminopeptidase, nor GshB – encoding glutathione synthetase, contain canonical DNA-binding domains. We found that knock-out mutants of both pepA and gshB resulted in an increased expression of emrD. Additionally, our work with a PepA G482D point mutant, which was shown to have 90% reduced aminopeptidase activity in E. coli (Charlier, et al. 2000), suggests that PepA aminopeptidase activity must be intact in order for PepA to regulate emrD. However, we cannot rule out that this point mutant may behave differently in E. coli and V. cholerae, and thus further testing (e.g. aminopeptidase activity assays and use of alternative point mutants) would better support this idea.

161

A search of the KEGG Pathways Database (Kanehisa and Goto 2000, Kanehisa, et al. 2014) identifies PepA and GshB as players in glutathione metabolism. Specifically, these enzymes contribute to the γ-glutamyl cycle, which uses four main enzymes to make and break down glutathione (Figure II-29) (Orlowski and Meister 1970). Glutathione, which is a tripeptide (γ-glutamyl-cysteinyl-glycine), is produced through the action of

GshA (L-glutamate + L-cysteine + ATP  γ-glutamyl cysteine + ADP + Pi) and GshB

(γ-glutamyl cysteine + glycine + ATP  γ-glutamyl-cysteinyl-glycine +ADP + Pi)

(Masip, et al. 2006). Likewise, the breakdown of glutathione is a two-step process. First

γ-glutamyl transpeptidase mediates hydrolysis of the γ-glutamyl-cysteine bond. Although the protein was named a transpeptidase due to its ability to transfer the released γ- glutamyl to numerous amino acids, further investigation has demonstrated that this is a secondary function of the enzyme (Suzuki, et al. 1986). Regardless, the released glutamate from this hydrolysis can be shunted back in to glutathione synthesis by GshA.

V. cholerae encodes two putative γ-glutamyl transpeptidases: VC0194 and VCA0558.

After the glutamate residue has been removed from glutathione, aminopeptidases mediate the further break-down of cysteinyl-glycine to cysteine and glycine. E. coli encodes four aminopeptidases that contribute to this enzymatic activity: PepA, PepB, PepD, and PepN

(Suzuki, et al. 2001). Interestingly, although E. coli PepA was shown to be active against cysteinyl-linked peptides (Suzuki, et al. 2001), it is most commonly considered a leucyl aminopeptidase (Stirling, et al. 1989, Charlier, et al. 2000). Thus, based on homology

(81% identity), V. cholerae PepA was also annotated as a leucyl aminopeptidase, but the protein likely harbors cysteinylglycinase activity. Thus, PepA and GshB both contribute

162

Glutamate

Cysteine PepA GshA PepN Glycine VC0194 PepB VCA0558 Cysteinyl-glycine

Glutathione Υ-glutamyl-cysteine (Υ-glutamyl-cysteinyl-glycine) GshB

Figure II-29. γ-glutamyl cycle in Vibrio cholerae. The γ-glutamyl cycle results in the synthesis and break-down of glutathione. The enzymes involved in this pathway (in red) have not been studied in V. cholerae and are listed based on their homologies to E. coli proteins. The functions of each protein are as follows: GshA – γ-glutamyl cysteine synthase; GshB – glutathione synthetase; PepA, PepN, and PepB – aminopeptidase; and VC0194 and VCA0558 – γ-glutamyl transpeptidase. In our genetic selection, insertions in gshB and pepA were enriched in the plus and minus IPTG conditions, whereas insertions upstream of gshA were enriched only in the plus IPTG conditions.

163

to glutathione metabolism, and disruption of each of the genes would result in a break in the γ-glutamyl cycle, perturbing glutathione homeostasis.

In addition to PepA and GshB, we also identified several enriched insertions upstream of gshA that were selected in the plus IPTG conditions. These insertions likely lead to the over-expression of gshA, which encodes γ-glytamyl cysteine synthase, the first enzyme in glutathione synthesis. Over-expression of gshA would lead to an accumulation of glutathione and/or γ-glutamyl-cysteine in the cell, which could also be disruptive of the γ-glutamyl cycle. Additionally, glutathione is an inhibitor of GshA enzymatic activity such that over-expression of the protein may eventually lead to suppression of its activity

(Huang, et al. 1988).

It is not immediately clear how glutathione metabolism relates to regulation of emrD, which putatively encodes a cytoplasmic-membrane transporter. Both over- expression of gshA and deletion of gshB would lead to accumulation of γ-glutamyl- cysteine within the bacterium. Alternatively, deletion of pepA may result in accumulation of cysteinyl-glycine, although V. cholerae encodes 6 other putative aminopeptidases

(VC0067/ PepP, VC0755/ PepB, VC1494/ PepN, VC2261, VCA0812, and VCA0813) that could potentially compensate for the loss of pepA. If accumulation of these pathway intermediates is toxic to V. cholerae, it is possible that EmrD may function to transport these intermediates out of the cytoplasm to relieve toxicity. Thus, EmrD might be a small peptide transporter.

Glutathione has been tied to activity of the fumarate-nitrate reductase regulator

(FNR) in E. coli. FNR is a conserved global regulator in bacteria that regulates many genes, including those involved in anaerobic growth. FNR acts primarily as a dimer, and

164 it is held in this confirmation by iron-sulfur cluster interactions. Upon exposure to oxygen, the iron-sulfur cluster is oxidized and the FNR dimer dissociates (Green, et al.

2009). The reconstitution of the FNR dimer with an iron-sulfur cluster requires a two-step process, mediated by reducing agents. While dithiothreital (DTT) is used for in vitro reconstitution, glutathione functions as the in vivo reducing agent mediating dimer formation of FNR in E. coli (Tran, et al. 2000). Disruption of glutathione homeostasis

(e.g. gshB deletion) would lead to a inability to (re)activate FNR. Thus, FNR may act as a repressor of emrD in response to anoxia.

Through the identification of emrD regulators we hoped to identify the function of

EmrD and elucidate why the gene is induced at the late stage of infection. Neither gshB nor pepA have been identified as infection-induced genes (Merrell, et al. 2002a,

Lombardo, et al. 2007, Schild, et al. 2007). However, GshB has been implicated in the acid-tolerance response of V. cholerae, implying an early role for the protein during infection (Merrell, et al. 2002b). Additionally, gshB was shown to be repressed in rice water stool samples when compared with in vitro grown bacteria (Merrell, et al. 2002a).

Thus, expression of GshB during the early stages of infection, but not late, would allow for late induction of emrD via activation of the late gene in the absence of GshB. In summary, more work is required to determine the mechanism of PepA and GshB regulation of emrD, whether the regulation occurs during infection, and if this regulation is important in understanding of cholera as an infectious disease.

II.3.4. xds selection discussion We identified PhoB as an activator of xds that specifically induces expression of xds under phosphate starvation conditions. The qRT-PCR experiments we performed

165 suggest that PhoB is the sole regulator, or at least the major regulator, of xds from evidence that deletion of phoB resulted in complete repression of xds. It is conceivable that a weaker regulator, which does not induce expression enough to be detected by qRT-

PCR, contributes to xds activation. Alternatively, xds may be induced by an additional regulator that is not active under the conditions we have tested (e.g. LB and MOPS medias, and in the infant mouse small intestine). Additional evidence that PhoB is the sole regulator of xds was provided by our method wherein we used high-throughput sequencing in combination with saturating transposon mutagenesis to identify all regulators of xds, both weak and strong. This method identified the Pst/PhoU and PhoB/R systems as the sole regulatory system of xds. If a weak regulator does exist, it is possible that we were unable to select for it because the concentration of antibiotic that we chose was too high for the low level of xds-bla-neo expression driven by this hypothetical weak regulator. This possibility is unlikely to be true, because the concentrations of antibiotic that we used were only slightly above the level of resistance conferred by the basal level of expression of the xds promoter in its un-induced state.

In support of direct regulation of xds by PhoB, we found that PhoB can bind to the xds promoter in vitro. While performing these assays it became clear that EMSAs are not the best way to determine binding-affinity between a protein and a strand of DNA.

Indeed, the use of different buffers or minor buffer components can greatly change the apparent affinity of the regulator for the DNA. Additionally, determining the activity of a particular protein prep becomes essential for comparing one shift to another. There is no in vitro assay for PhoB activity, so we could not reliably compare shifts. Therefore,

EMSAs to determine the absolute affinity of a regulator for a promoter of interest is not

166 trivial. An alternative method is to determine the relative affinity of the regulator to the promoter when it is compared to the affinity of a negative control probe. Thus, the use of proper negative controls becomes greatly important in these assays. Here we used an internal fragment of bla as a negative control. This control is not as suitable as a promoter region would be, because PhoB could harbor low affinity binding to conserved -10 or -35 sites. Additionally, due to limitations with protein purification (e.g. a presumed low activity prep and inability to further concentrate the protein), we were unable to determine a differential affinity for the xds probe as compared with the bla control probe.

Although it is clear that more thorough binding assays are required to fully define the

PhoB-xds promoter interaction, the cold competition assays clearly show that the binding we observed in our EMSAs was specific.

Prior to our identification of PhoB as a regulator of xds, both the pst genes and phoR were described as induced during infection (Lombardo et al., 2007; Osorio et al.,

2005). Additionally, deletions in phoB or in pst/phoU, which result in constitutively high

PhoB activity, highly attenuate virulence (Merrell et al., 2002; Pratt et al., 2009; von

Kruger et al., 1999). This suggests that regulation of PhoB activity during infection is essential and dynamic. Combined with our results, this suggests that the regulatory connection we have defined is relevant to infection. This idea is further explored and developed in Chapter III.

The induction of xds by PhoB under phosphate limiting conditions, suggests that the exonuclease, Xds, may contribute to the phosphate starvation response. Recent work has provided evidence that V. cholerae can utilize eDNA as a phosphate source, and this ability is dependent on the presence of either xds or dns (encoding the other secreted

167 nuclease) (Seper et al., 2011). Degradation of DNA via Xds leads to accumulation of mainly monophosphorylated nucleotides. To date, no transporter of monophosphorylated nucleotides has been identified in V. cholerae. Thus, it is possible that a secreted or extracellular phosphatase would work together with Xds to provide free phosphate for uptake by either Pst/PhoU or another uptake system. This hypothesis is explored in

Chapter IV of this thesis. Additionally, the relevance of PhoB regulation of xds to infection is discussed in Chapter III.

168

CHAPTER III:

RE-EXAMINATION OF XDS EXPRESSION DURING INFECTION OF AN INFANT MOUSE

Some of the text and figures/tables in this section are adapted with permission from my manuscript:

McDonough, E., Lazinski, D.W., Camilli, A. Identification of in vivo regulators of the Vibrio cholerae xds gene using a high-throughput genetic selection. Mol. Microbiol. 2014 Feb. 20.

169

III.1. OVERVIEW The goal of our genetic selection, described in Chapter II, was to create an in vitro selection that would allow us to identify regulators of V. cholerae late genes. An essential aspect of our study was to validate that the regulators we identified were important for regulation during infection. In this chapter I describe the work we have done to validate

PhoB as an in vivo regulator of xds. As a follow-up to this work I show that xds is induced early in infection and that expression of xds is not required for the early colonization of an infant mouse small intestine.

III.2. RESULTS

III.2.1. PhoB regulates xds during infection of an infant mouse Since late genes are induced during infection, it is much easier to validate their activators than their repressors. A repressor would need to be artificially, constitutively expressed and kept active in order to prevent late gene expression in vivo, whereas deletion of an activator would suffice to prevent late induction. Constitutive activation of a V. cholerae gene during infection of a mouse is not trivial since we lack information on promoters that are active throughout infection. Inducible promoters, such as IPTG or arabinose inducible ones cannot be used since it would be impossible to maintain inducer level throughout infection. Furthermore, constitutive expression of a repressor is not guaranteed to prevent late gene expression, as the conditions may not be correct for activating the repressor. PhoU is a perfect example of this type of situation; repressor activity of PhoU relies heavily on an excess of phosphate in the growth condition, unlike the low phosphate conditions observed within the mouse small intestine. Thus, in order to validate the in vivo regulation of xds by the Pho genes, we chose to delete the xds

170 activator implicated by our results, phoB, and observe the effect on xds expression during infection.

We used RIVET to semi-quantitatively measure transcription of xds during infection of an infant mouse (Camilli and Mekalanos 1995)2. The gene tnpR, which encodes a res-site specific recombinase, was transcriptionally fused to xds via a plasmid insertion that disrupted the xds open reading frame (Schild, et al. 2007). This strain also contained a TnpR substrate, the resolvable kanamycin-resistance marked cassette (res- neo-sacB-res) (Osorio, et al. 2005)(Figure III-1). When xds is induced in the fusion strain, TnpR is co-expressed and mediates resolution of the cassette, resulting in the loss of Kn resistance. Therefore, the strain remains resistant to Kn so long as xds is not induced, but becomes sensitive to Kn upon induction of xds and resolution of the res cassette.

As a test of our RIVET strains, and as further confirmation of our in vitro qRT-

PCR results, we first measured resolution mediated by the xds-tnpR fusion in MOPS media supplemented with high or low phosphate (Figure III-2a). Resolution of the res cassette in the wild type strain was low in high phosphate (mean = 5.0%; SD = 5.6) and high in low phosphate (mean = 97.7%; SD = 0.6). Conversely, when we deleted phoB, the res cassette was no longer induced in low phosphate (mean = 4.0%; SD = 0), when compared with high phosphate (mean = 2.3%; SD = 2.5). These results confirmed our

2 The use of qRT-PCR to quantitatively measure xds induction at early and late time points during infection would be ideal. However, we made numerous attempts to perform qRT-PCR from in vivo samples and were unsuccessful. Likely the high ratio of mouse RNA to V. cholerae RNA at the early time points made it difficult for us to perform this experiment. In lieu of qRT-PCR quantitative data, RIVET at least provides evidence that the gene is induced throughout the entire population of bacteria infecting the mouse, and that this induction is dependent on the presence of phoB.

171

A) xds OFF KanR SucroseS

Pxds

xds’ tnpR mob ori R6K bla ‘xds res neo sacB res

B) xds ON KanS SucroseR

Pxds

xds’ tnpR mob ori R6K bla ‘xds res neo sacB res

TnpR 1 TnpR

2 3 TnpR sacB

neo res

TnpR TnpR

Figure III-1. Recombination-based in vivo expression technology (RIVET) to measure xds transcription. For RIVET the xds open reading frame was disrupted by a single cross-over event with the plasmid pIVET-VC2621, resulting in a fusion of tnpR, which encodes the res-site specific resolvase, to the xds promoter. The RIVET strain also harbored the res cassette, which was inserted in to the endogenous lacZ gene. The res cassette used in this study carries a neo gene, which encodes resistance to Kn, and the sacB gene, which when expressed during growth on sucrose is toxic to the bacterium. A) When the xds promoter is off, no tnpR is made and the res cassette remains in the chromosome. Thus, the cells are resistant to Kn, but sensitive to growth on sucrose. B) When xds is induced, tnpR is also transcribed (1). The resolvase recognizes and binds to the res sites in 1:1 stoichiometry (2). Irreversible loss of the res cassette by resolution is mediated TnpR (resolvase) (3). These bacteria, which have lost the res cassette, are sensitive to Kn, but resistant to growth on sucrose.

172

23 43 &!(56( !:&(56( 78,+789/* 78,+789/* !"#$%&'( !"#$%$(

&!! &!!

%! %! 1 1 , , 0 0 / / . . - $! - $! , , + + * *

) #! ) #! ( ( ' ' "! "!

! !

)*(+# )*(+# )*(+# )*(+ ¨ ¨ ¨ ¨ ;0-<(/=7* ;0-<(/=7* ;0-<(/=7* ;0-<(/=7*

Figure III-2. PhoB regulates xds during infection of an infant mouse. Expression of xds was measured in the wild type (filled in circles) and ΔphoB (open circles) strains using RIVET. A) Resolution of the res cassette was determined after 4 hours of growth in MOPS media supplemented with 10 mM or 0.1 mM KH2PO4 (phosphate). B) Resolution of the res cassette was measured after 4 hours of growth in LB (in vitro) or 24 hours post infection of an infant mouse (in vivo). Each data point represents the resolution frequency from either an individual culture tube or mouse. The black bars indicate the mean resolution.

173

qRT-PCR results and demonstrated that regulation of xds by PhoB can be measured using

RIVET.

Next, we tested resolution of the res cassette in the wild type and ΔphoB RIVET strains during infection of infant mice (Figure III-2b). To ensure that resolution was not occurring during preparation of the inocula, we first measured resolution in both of these strains grown in LB broth (the medium used to grow the inocula). Neither the wild type

RIVET (mean = 3.7%; SD = 2.5) nor the ΔphoB RIVET (mean = 1.7%; SD = 1.5) strains exhibited appreciable resolution in LB. When the wild type RIVET strain was passaged through an infant mouse, nearly all colony-forming units (CFU) recovered from the small intestine at 24 hours had resolved the cassette (mean = 93%; SD = 8.0), consistent with xds being induced in vivo. Resolution did not occur in the ΔphoB RIVET strain after passage in vivo (mean = 1.1%; SD = 1.2). Therefore, PhoB appears to be an essential regulator of xds during infection.

III.2.2. Induction of xds occurs early, rather than late in infection Since xds is both a PhoB regulated gene and in vivo-induced, we hypothesized that V. cholerae encounters low phosphate conditions in the infant mouse small intestine.

Although others have provided some evidence that PhoB and the phosphate starvation response are important during infection (von Kruger, et al. 1999) we were interested in specifying at what time during infection phosphate becomes limiting. Our RIVET data suggested that PhoB is the dominant regulator of xds during infection of an infant mouse, as demonstrated by the evidence that deletion of phoB nearly completely prevented resolution of the res cassette. This conclusion was also supported by the results obtained from our Tn-seq selection where only insertions that resulted in PhoB activation were

174 found to significantly enhance expression from the xds promoter. Therefore, we reasoned that PhoB activation, and thus low phosphate levels, could be measured during infection of an infant mouse by using the xds RIVET strain as a biosensor for phosphate.

The RIVET expression profile of xds described by Schild and colleagues (2007) suggested that xds is induced late, implying, based on our work presented here, that phosphate becomes limiting late in infection. However, we had observed that xds expression in a phosphate depleted medium, as measured by in vitro RIVET, is delayed when using plate grown bacteria, rather than mid-exponential phase broth grown bacteria, as the inoculum. Schild and colleagues (2007) did not use mid-exponential phase bacteria, nor was care taken to exclude exogenous phosphate from the inoculum.

Additionally, the original xds RIVET expression profile was described using a modified version of RIVET that utilizes both a weaker ribosome binding site for the tnpR transcript and a mutated res substrate that has a lower affinity for TnpR (Osorio, et al. 2005, Schild, et al. 2007). Therefore, it remains a possibility that xds is induced earlier during infection in response to phosphate limitation in the small intestine.

We designed an in vivo RIVET experiment tailored to specifically identify when phosphate becomes limiting in the infant mouse during infection. Importantly, we used mid-exponential phase bacteria, washed twice, and resuspended in MOPS medium (no phosphate) as the inoculum to ensure that the strain could quickly respond to limiting phosphate by resolving. To test this prior to performing the in vivo experiment, we first measured expression of the RIVET inoculum over time, after transfer to MOPS medium lacking phosphate (Figure III-3a). We observed a lag of resolution of the res cassette of 2 to 2.5 hours after transfer to the phosphate-depleted condition. These results suggested

175

78 !"#$%'(&!*/012%&31# 98 !"#$%$&!*/012%&31#

,++ ,++

.+ .+ # # 1 1 3 3 & & % % 2 2 -+ -+ 1 1 0 0 / / ; ; (+ (+ ! ! : : '+ '+

+ +

'!)* 4!)* (!)* !,!)* (!)* !.!)* ,-!)* !"#$%& +56!)* ,56!)* !'56!)* 456!)* !"#$%& ,'!)* !'+!)* '(!)*

Figure III-3. Expression of xds occurs early, rather than late, during infection. Expression of xds was measured using RIVET after transfer of bacteria from a high phosphate medium (LB) to either A) MOPS minimal medium lacking phosphate or B) an infant mouse gastrointestinal tract. Each dot represents a single mouse or culture. Black bars represent the mean resolution for each time point.

176

that the soonest we would be able to detect resolution of the cassette in vivo would be 2 hours post-inoculation.

We measured resolution of the res cassette at 4, 8, 12, 16, 20, and 24 hours post inoculation of infant mice with the RIVET strain (xds-tnpR lacZ::res-sacB-neo-res)

(Figure III-3b). We found that expression of the gene within the population began by 4 hours, and by 12 hours nearly the entire population had resolved. Interestingly, the 4 hour time-point exhibited a bifurcation in the population in terms of xds expression; two of the mice exhibited ~38% resolution at 4 hours post-inoculation, whereas the other two mice exhibited ~88% resolution. The reasons for this bifurcation phenotype are unknown, and further replicates are required to determine if the phenotype is real or merely a sample size error. From this RIVET time-course experiment, we conclude that the mouse small intestine is a low phosphate environment, at least as early as 4 hours post-inoculation.

This low phosphate environment triggers PhoR/B activation resulting in induction of xds and likely other genes in the Pho regulon.

III.2.3. xds is not required for early colonization of the infant mouse Previous studies with V. cholerae strains harboring deletions in either xds alone or double deletions of xds and dns (encoding the other secreted DNase) have concluded that xds is not important for colonization or virulence (Focareta and Manning 1991a, Seper, et al. 2011, Seper, et al. 2013). Although each study was performed using a slightly different model of infection, in each case the fitness of the mutant strains was determined at the late stage of infection. According to our results xds is an early-induced gene and, therefore, we reasoned that the exonuclease may play a role in survival or growth during the early stage of colonization (e.g. nutrient acquisition). If the role of xds at an early time

177 point is small, the fitness disadvantage may be recovered at later stages of infection, thus providing a possible explanation for why earlier studies were unable to find a fitness defect for an xds deletion mutant. Furthermore, in our in vivo RIVET experiments we found that the manner in which the inoculum is prepared, specifically with regard to phosphate content, greatly impacts the timing of xds induction. At least two of the previous studies (Seper, et al. 2011, Seper, et al. 2013), used a high phosphate containing inoculum, which we found delays the induction of xds. Details of the inoculum used by the third study were not provided (Focareta and Manning 1991a). Therefore, we hypothesized that an xds deletion may elicit a fitness disadvantage during an early stage of infection when the inoculum is prepared under low phosphate conditions.

To test this possibility, we compared wild type and Δxds small intestinal colonization after 8 hours of infection, which represents the time when most of the population has induced xds in our timecourse experiment. The bacteria were prepared for infection as was described for the RIVET timecourse experiment, except that in addition to a MOPS inoculum lacking phosphate we also included a control infection using bacteria resuspended in MOPS with 10 mM KH2PO4 (high phosphate). We performed single strain mouse infections in order to control for complementation in trans of Xds, which is a secreted factor. We were unable to detect a statistically significant difference in the number of recoverable CFUs in the mouse small intestine at 8 hours post infection when comparing wild type and Δxds strains, regardless of the phosphate concentration in the inoculum (Figure III-4). Therefore, we conclude that although xds is induced during infection, the role that the gene plays is either dispensable or redundant in our infection model.

178

!"+-8+ ;: 79:,79.23 79:,79.23

!"& 3 1 0 2 , 3

2 %

1 !" 0 + / / . - ,

+ $

* !" ) ( ' !"#

!"# !"# ¨ ¨ 40/5+2673 40/5+2673

Figure III-4. Early induction of xds is not required for infant mouse colonization. Fitness of wild type and Δxds were measured after 8 hours of infection in infant mice. Each dot represents the recovered number of V. cholerae from a single mouse, with at least 8 mice in each group. We used the Mann-Whitney U Test to compare the number of CFU/small intestine of mice colonized with either wild type or Δxds strains when the inoculum was prepared in high phosphate (p = 1.0) or no phosphate (p = 0.94). Additionally, we combined all of the high phosphate and no phosphate measurements and compared them using the Student’s T-test. We were unable to find a statistically significant difference between colonization levels and high or no phosphate inoculum (p = 0.13). The black bars represents the median values.

179

III.3. DISCUSSION In Section II of this thesis we identified PhoB as a regulator of the infection- induced gene, xds, and we provided in vitro validation of this. In Section III we used a

RIVET-based assay to demonstrate that PhoB is the major regulator responsible for induction of xds during infection of an infant mouse. Our identification of PhoB as a regulator of the infection-induced gene, xds, corroborates prior work that identified pst genes and phoR as transcriptionally induced during infection (Osorio, et al. 2005,

Lombardo, et al. 2007). Additionally, dynamic activity of PhoB appears to be important during infection since deletions in phoB or in pst/phoU, which result in constitutively high PhoB activity, highly attenuate virulence (von Kruger, et al. 1999, Merrell, et al.

2002, Pratt, et al. 2009).

Using our RIVET-based assay, we measured the kinetics of xds induction both during incubation in MOPS medium supplemented with no source of phosphate and during infection of an infant mouse. Since PhoB is the major regulator of xds, we reasoned that we could use the resolution of the res cassette in the xds RIVET strain as a readout for PhoB activation. Interestingly, we observed a 2-2.5 hour lag in resolution of the res cassette after transition from replete to depleted phosphate medium. Considering that resolution and loss of the res cassette can occur within 20 min of tnpR induction

(Camilli, et al. 1994), and that the processes of transcription and translation take on the order of minutes (Young and Bremer 1976, Vogel and Jensen 1994), this result suggests that activation of PhoB takes approximately 1.5-2 hours after transition to the phosphate limited environment. This result is surprising considering that the activation of PhoB is thought to occur directly after sensing of limiting extracellular phosphate by the Pst/PhoU

180 complex. This seeming contradiction could be explained if PhoB is a weak activator of xds. Indeed, our EMSA experiments presented in Chapter II demonstrate that PhoB binds to the xds promoter with an affinity approximately 4-fold lower than its affinity to the phoBR promoter. That is, if a high level of phosphorylated PhoB protein is required to induce transcription of xds, this could explain the delay in resolution of the res cassette.

Bearing in mind this hypothesis, by using the xds RIVET strain, we may be overestimating the lag of PhoB activation after transition to liming phosphate conditions.

Further experiments using fusions of other PhoB-regulated genes to tnpR, such as phoX or even phoBR, could address this question of timing and affinity.

PhoB is induced as soon as 4 hours post-infection of an infant mouse, according to results from our in vivo RIVET assay. After subtracting the 2-2.5 hour lag observed in

PhoB activation in vitro, the time post-infection when a low phosphate condition was first encountered was 1.5-2 hours (and possibly less if xds transcription truly requires high levels of phosphorylated PhoB). Given that some time is needed post-inoculation to traverse the stomach, it is likely that as soon as bacteria reached the small intestine, they sensed a low concentration of phosphate. In other words, a low phosphate condition very likely pre-exists in the mouse small intestine prior to colonization, and is likely not due to depletion of phosphate by the bacterium or to sequestration of extracellular phosphate by the host innate immune response.

As described in the results, we observed a bifurcation phenotype at the four hour post-inoculation time-point, in which half of the mice harbored a population of bacteria with ~88% resolution, and the other mice harbored a population with ~38% resolution.

Systems that exhibit all-or-none phenotypes, such as activation of the Pho regulon, often

181 result in highly variable expression near the induction threshold. Therefore, this bifurcation phenotype may reflect that phosphate levels experienced by bacteria at the 4 hour time-point post-inoculation, are near the concentration that leads to induction of the

Pho regulon. Alternatively, the levels of phosphate V. cholerae senses within the mouse at the four hour time-point may be variable, and may depend on the precise infection conditions, or the physiological state of the individual mouse. It is also possible that the time it takes to traverse from the stomach to the small intestine (where we hypothesize the low phosphate environment is sensed), may be highly variable within and between bacterial populations. Again, this could be due to differences between infected mice.

Since we have not observed a bifurcation in any of the in vitro xds RIVET experiments, we hypothesize that this phenomenon is specific to the mouse infection.

Our RIVET data showing early induction of xds conflicts with the late induction of xds that was previously described (Schild et al., 2007). Differences between these two studies in the physiological states of the bacterial inoculum (plate grown versus mid- exponential broth grown) likely contribute to the differences in the rates of resolution that were observed. We hypothesize that internal phosphate stores, such as poly-phosphate, may differ between plate grown and mid-exponential phase bacteria, and that the presence of an internal phosphate store may delay activation of PhoB in a low phosphate environment. Furthermore, in contrast to this study, the inoculum used by Schild and colleagues (2007) was diluted in LB, a high phosphate-containing medium, and this likely also contributed to the delay in resolution. Finally, the ribosome binding site used by Schild and colleagues (2007) to drive TnpR translation was weaker than the one used in this study. Due to less efficient translation, more transcripts were needed to achieve the

182 critical level of resolvase required for resolution and this undoubtedly exacerbated the observed delay. Importantly, in this study, we identified low phosphate as the major inducer of xds expression and this knowledge enabled us to promote xds induction in vitro. We therefore could compare the rate of resolution of our inoculum in vitro with that in vivo, and this enabled the calibration of our RIVET-based clock. At the time of the

Schild et al. (2007) study, there was no knowledge of xds regulation and xds induction in vitro could not be achieved, and therefore calibration of their RIVET-based clock was not possible.

Despite our success with the RIVET-based assay, it has several limitations. The assay is only semi-quantitative, in that once a bacterium reaches the critical level of xds- tnpR transcription, resolution can occur within that cell. That is, there is not a one to one ratio of transcript to read out as for qRT-PCR assays. Additionally, loss of the res cassette is an irreversible event and therefore, we are unable to address whether transcription of phoB is off later in infection. Use of qRT-PCR on infection samples would allow us to quantitatively measure xds transcription as well as to describe an xds transcription profile over the course of an infection. Unfortunately, despite numerous attempts we were unable to get qRT-PCR to work from mouse infection samples.

We hypothesize that differences in the physiological state of V. cholerae and/or the phosphate concentration in contaminated water being ingested may have an impact on the timing of xds and Pho regulon induction during human infection, although the effect of this timing, if any, on the course of the illness is unknown. Despite the fact that fresh water is generally considered phosphate limiting (Nelson, et al. 2008), micro-pockets of high phosphate may exist. Additionally, human treatment or contamination of water may

183 introduce phosphate into the otherwise phosphate-free environment. It is unlikely that V. cholerae populations ingested by susceptible humans are always coming from depleted phosphate conditions. With respect to Xds, based on our and others’ experiments it seems that the induction of xds is not essential for mouse infection (Focareta and Manning

1991a, Seper, et al. 2011, Seper, et al. 2013), and therefore, the differences in timing may be irrelevant to the bacterial survival. On the other hand, phoB is critical for virulence in animal models of colonization, as either loss of phoB or unregulated PhoB activation (in pst/phoU mutant backgrounds) both highly attenuate virulence (von Kruger, et al. 1999,

Merrell, et al. 2002b, Pratt, et al. 2009, Kamp, et al. 2013). Thus, although Xds does not play an essential role in infection, other members of the Pho regulon do, presumably because they are needed to scavenge phosphate from alternative sources present in the small intestine.

In summary the results described in this chapter have expanded upon our knowledge of xds regulation and induction during infection. We have shown that PhoB is the major activator of xds in the infant mouse, and we provide evidence that this induction occurs very early in the course of an infection.

184

CHAPTER IV:

IDENTIFICATION OF FACTORS REQUIRED FOR THE USE OF NUCLEOTIDES AND DNA AS SOURCES OF PHOSPHATE BY V. CHOLERAE

185

IV.1. OVERVIEW In Chapters II and III of this thesis I have provided evidence that the phosphate starvation response regulator, PhoB, induces expression of xds, which encodes a secreted exonuclease. From these results, we hypothesized that V. cholerae expresses Xds during phosphate starvation in order to utilize eDNA as a source of phosphate. Indeed, while we performed this work, Seper and colleagues (2011) demonstrated that V. cholerae is able to use eDNA as a source of Pi and that the two secreted nuclease, Xds and Dns, are required for this phenotype.

Degradation of DNA results in accumulation of nucleotides. In order to access the

Pi from these nucleotides, bacteria must separate the phosphate group from the nucleoside group. Although nucleosides, which lack a phosphate group, are readily transported into the cytoplasm of many Gram negative bacteria via Nup transporters, it remains unclear whether the phosphorylated nucleotides can cross the inner membrane (Watanabe, et al.

2011). Therefore we hypothesized that expression of a periplasmic or extracellular phosphatase would allow for growth on eDNA by releasing phosphate from the mononucleotides liberated by Xds and Dns. Presumably, once the phosphate is removed, it can be taken up into the cell by any of the phosphate transporters, i.e. Pst/PhoU, PiT, or

Na-Pi cotransporter. In this final results chapter, I describe work in which nucleotide phosphatases are identified and minimally characterized.

186

IV.2. RESULTS

IV.2.1. PhoX is not required for utilization of eDNA as a phosphate source Many organisms have been shown to use eDNA and exogenous nucleotides as sources of Pi. In several of these organisms including R. pomeroyi and Shewanella spp. alkaline phosphatase has been described as the major phosphatase responsible for removing the Pi from nucleotides (Pinchuk, et al. 2008, Sebastian and Ammerman 2009).

The V. cholerae alkaline phosphatase, PhoX, is expressed in the periplasm and the gene is regulated by PhoB (von Kruger, et al. 2006, Pratt, et al. 2009). Therefore, we hypothezied that PhoX would provide the required phosphatase activity for growth on eDNA as a source of phosphate.

3 To determine if phoX is required for growth on eDNA as a sole source of Pi , I performed growth curves of wild type, ∆xds∆dns, and ∆phoX (Figure IV-1). I used sheared salmon sperm DNA for these experiments, as this was what others in the V. cholerae field successfully used (Seper, et al. 2011). The ∆xds∆dns strain was used as a negative control to ensure that there was no Pi contamination of the DNA. Additionally, we tested all of the strains for growth in MOPS-glucose medium supplemented with 10 mM KH2PO4, to ensure that the strains were able to grow equally well in Pi replete

3 It is important to note that the concentration of 0.5 mM used in these experiments refers to the total amount of Pi in the DNA, and not the molarity of the DNA. The stock of sheared salmon sperm DNA that I used is at 10 mg/mL. Pi has a mass of 80 Da, while dNMPs have an average mass of 330 Da.

Therefore, to determine the Pi concentration in my stock of DNA: 10 mg/mL x (80 Da/330 Da) = 2.42 mg/mL Pi.

The molarity of Pi in the solution, where 94.9g/mole is the molar mass of Pi = -3 -3 (2.42 x 10 g / 10 L) x (1 mole/94.9g) = 26 mM Pi

For DNA growth curve experiments, the DNA was added to the growth medium to a final concentration of 0.5 mM Pi. 187

A) 0.5 mM phosphate from DNA B) No phosphate & & '()*+,-./ !"#+ "$# ! ! %&'() ! ! $ !%& $ !%& 6 6 5 5

!%!& !%!&

! "! #! $! ! "! #! $! 0(1/+234 0(1/+234

Figure IV-1. V. cholerae does not require phoX for the utilization of eDNA as a source of phosphate. After overnight growth in a MOPS-glucose medium (10 mM KH2PO4), strains were back diluted to an OD600 ~0.05 and grown to mid-exponential phase in the same medium. The bacteria were washed two times in MOPS medium containing no phosphate and inoculated into 2 mL MOPS-glucose media cultures with either 0.5 mM phosphate from sheared salmon sperm DNA A) or no phosphate B). Strains were grown at 37°C with aeration. Shown is a representative graph of the mean of two biological replicates assayed on the same day. The growth assay was performed twice, with the same results each time.

All strains grew equally well in high phosphate MOPS medium (10 mM KH2PO4).

188

conditions. Indeed, none of the strains we tested (in any of the DNA growth curves discussed in this chapter), exhibited a growth defect when compared with wild type. As had been reported earlier, we found that wild type was able to use eDNA as a source of

Pi, however the rate of growth was severely decreased when compared to wild type growth in either a phosphate replete or limiting conditions. Growth of the phoX deletion strain closely matched the wild type. Thus, we concluded that either phoX does not contribute to utilization of eDNA as a source of phosphate, or its role is redundant with other phosphatases/nucleotidases.

In both this experiment and others, there was a high level of variation in terms of length of lag time before exponential growth, as well as growth rate. Although this variation occurred even between biological replicates assayed on the same day, the day- to-day variation was much more severe. Therefore, in most of the DNA growth curves I have shown either a single curve or the mean of two curves performed on a single day.

Although plotting error bars is more ideal, this makes the data nearly uninterpretable.

Additionally, I am interested in defining the proteins that are required for growth on eDNA as a source of Pi. As such, I am looking for a severe growth defect (i.e. categorical information; growth versus no growth). This is also why I am not reporting doubling times, as they are complicated to calculate for these particular curves, and they aren’t necessary for the question I am asking.

It is unclear to me what the source of the growth variation is. One possibility is that slight differences in starting optical densities from day-to-day or replicate-to- replicate could result in better or worse growth of the strain. For example, a replicate that starts at a higher density than another would have more secreted nucleases working to

189 break up the supplied eDNA, which may give this replicate a slight advantage. An alternative explanation is that the bacteria used to seed different replicates may have stored differing amounts of PolyP during the initial growth in high phosphate MOPS- glucose medium. A replicate that on average stored a greater amount of PolyP would be able to use the alternative source of phosphate to start the initial growth. Deleting ppx and/or ppk in all of the strain backgrounds tested, and repeating the eDNA growth curves could test this hypothesis. However, since, as I stated above, I am interested in growth versus no growth states, I did not pursue the source of this variation further.

IV.2.2. Transposon mutant screen to identify additional V. cholerae phosphatases and preliminary validation of hits To identify additional periplasmic or surface-associated V. cholerae phosphatases

I performed a transposon screen in a phoX deletion background using the colorimetric phosphatase substrate, 5-bromo-4-chloro-3-indolyl phosphate (XP), which turns blue upon removal of the phosphate group and subsequent oxidation of the molecule. While the wild type is dark blue on LB XP plates due to product accumulation in the periplasm, a phoX deletion strain is light blue on LB XP plates, suggesting the existence of another phosphatase other than phoX. The phoX deletion strain was mutagenized with a mTn10 transposon (as described in Chapter II). Although we did not determine the precise number of unique insertions in this library, based on our previous results with eight independently made libraries, we estimated a complexity of 50,000 to 100,000 insertions.

The library was plated on 20 LB XP plates (100 mM diameter) and approximately 40,000 mTn10 mutants were screened for white color on LB XP plates. I identified 7 white colonies. The location of the transposon insertions in these 7 colonies was determined by

190 using arbitrary-primed PCR and DNA-sequencing (Figure IV-2) (Hava and Camilli

2002).

Of these 7 colonies, 5 represented unique insertion sites present throughout the ushA coding region (encoding a bifunctional UDP-sugar hydrolase/ 5’ nucleotidase) and

2 unique insertions were in the 5’ end of nupC (encoding a nucleoside permease family protein). Clean deletion (in-frame deletion of coding sequence leaving start-stop codons and a FRT scar) of ushA in the wild type background results in white colonies on LB XP plates. In contrast a clean deletion of nupC was still blue on LB XP plates, even in a phoX deletion strain, which suggests that expression of a truncated NupC somehow inhibits cleavage of XP. Since Stefan Schild is working on characterizing the Nup transporters in

V. cholerae (personal communication), and we can see no obvious tie to phosphatase activity, we have not continued working on this protein.

191

A) Transposon insertions in ushA:

Orientation of the Tn encoded

Ptac UshA SS MPP C-term

B) Transposon insertions in nupC:

Orientation of the Tn encoded

Ptac NupC N-term Gate C-term

Figure IV-2. Transposon insertions in nupC and ushA. Seven total Tn insertions were identified in A) ushA and B) nupC. A single line represents each gene, and the Tn insertions are represented by triangles. Triangles on top of the line indicate that the Tn-encoded Ptac promoter is oriented in the same direction as the gene, those below the line are oriented opposite the gene. The predicted protein domains are shown beneath the gene. SS = secretion signal sequence, MPP = metallophosphatase domain, Gate = involved in determining nucleoside specificity.

192

IV.2.3. ushA is not required for growth on exogenous DNA Since we were interested in identifying phosphatases that specifically played a role in utilization of eDNA as a source of Pi, I next tested if ushA is required for this phenotype (Figure IV-3). Again, the ∆xds∆dns strain was used as a control to test for Pi contamination of the salmon sperm DNA. The ΔushA strain exhibited a very slight kinetic defect in growth compared with the wild type. However, by 25 hours of growth, both strains reached the same optical density. From these curves we concluded that ushA is either redundant or does not play a major role in DNA utilization.

Of note, wildtype, ∆ushA, and ∆xds∆dns all doubled a few times in medium lacking any Pi source. This suggested that these strains accumulate internal phosphate stores (e.g. PolyP) during the pre-growth in MOPS glucose potassium phosphate medium, and are able to utilize this phosphate store. In order to test this hypothesis that the consistent, but minimal growth seen in the no phosphate condition is due to break down of internal phosphate stores, I included the phoB deletion strain in these experiments. PhoB regulates the gene required for break down of PolyP stores, ppx

(Kulaev and Kulakovskaya 2000). Indeed, deletion of phoB abrogates the initial growth seen in the no phosphate condition, supporting my hypothesis. Alternatively, this initial small growth in the no phosphate medium may be due to contaminating phosphate carried over in the bacteria following the wash steps. The inability of the phoB deletion strain to grow in the no phosphate condition also supports this second hypothesis, as the

PhoB regulator is needed for expression of the high affinity phosphate uptake machinery,

Pst/PhoU, which is likely responsible for the uptake any low level of contaminating Pi.

Therefore, to more precisely determine whether ppx (vs. contaminating phosphate) is

193

A) 0.5 mM phosphate from DNA B) No phosphate 2(34+567* " " !"#! "$# %#&' ! ! ! ! (&)* 1 1 0 0 !&" !&" / /

!&!" !&!" ! "! #! $! %! ! "! #! $! %! '()*+,-. '()*+,-.

Figure IV-3. ushA is not required for growth on eDNA as a source of phosphate. This growth assay was performed as described in Figure IV-1, using mid-exponential phase bacteria, which were washed twice before putting in to the test conditions. The growth medium used was MOPS-glucose supplemented with either 0.5 mM phosphate from sheared salmon sperm DNA A) or no phosphate B). Shown is a representative graph of the mean of two biological replicates assayed on the same day.

194

responsible for the initial growth even in the absence of Pi, ppx itself should be deleted and the resulting strain tested for growth on eDNA. This experiment is not essential to my story, so I did not pursue it further.

IV.2.4. UshA is a non-nucleotide-specific 5’ nucleotidase, putatively periplasmic The main product of DNA degradation by Xds and Dns is thought to be monophosphorylated nucleotides. Additionally, UshA and PhoX are both phosphatases, which would act directly on the nucleotides rather than DNA. Therefore, I simplified the growth assays and tested whether the ushA and phoX deletion strains could grow using

5’dGMP and 5’dCMP as sources of Pi. Figure IV-4 shows that deletion of ushA completely inhibits the ability of the bacterium to grow on a mixture of these nucleotides, but deletion of phoX has no effect. These results suggest that UshA is the only 5’ nucleotidase active in the growth conditions used with the ability to act on either 5’dGMP or 5’dCMP.

The nucleotides 5’dGMP and 5’dCMP represent one purine and one pyrimidine, respectively. Therefore, to investigate whether UshA preferentially cleaves Pi from one molecule over the other, I performed these growth curves with each nucleotide separately. The ushA deletion strain was unable to grow when either a purine or a pyrimidine nucleotide was supplemented as the sole source of Pi (Table IV-1).

Additionally, the deletion of ushA could be complemented in this growth assay with a

FLAG-tagged ushA strain (discussed below). Further investigation with other 5’ monophosphorylated nucleotides supports the conclusion that UshA is the major 5’ nucleotidase in V. cholerae that is active in the conditions I’ve tested (Table IV-1).

195

A) & 0.1 mM 5’ dGMP/ 5’ dCMP B) & No phosphate !"#$ %&'( !"#$ %#&'( ! ! ! ! $ $ !'& !'& 0 0 / /

!'!& !'!& ! "!! #!! $!! %!! &!!! ! "!! #!! $!! %!! &!!! ()*+,-*. ()*+,-*.

Figure IV-4. ushA, but not phoX, is required for growth on nucleotides as a source of phosphate. Overnight MOPS-glucose (10 mM KH2PO4) cultures were back diluted to an OD600 ~ 0.05 and grown to mid-exponential phase. Cultures were washed two times in MOPS medium containing no phosphate and inoculated into 200 µL MOPS-glucose medium with either A) 0.1 mM dCMP/dGMP or B) no phosphate. Strains were grown at 37°C with aeration in a 96-well plate. The mean of four biological replicates, assayed on two separate days, are shown.

196

Table IV-1. Growth ability of various strains on 5’ nucleotidesa.

5'dGMP/ Strain 5'dGMP 5'dCMP 5'AMP 5'CMP 5'dCMPb 5'NTPc 5'dNTPc Wildtype +++ +++ +++ +++ +++ +++ ++* ΔxdsΔdns +++ +++ +++ +++ ND ND ND ΔphoB + + ND ND + ND ND ΔphoX +++ +++ ND ND +++ ND ND ΔushA - - - - - + + ushA-flag ++ ++ ND ND ND ND ND ΔphoXΔushA - - ND ND - ND ND a All nucleotides were added to a final concentration of 0.1 mM available Pi. For example, three times as much 5’AMP was added as compared with 5’NTP, which contains three phosphate groups for every one phosphate group of 5’AMP. bEquimolar mixture of 5’dGMP/ 5’dCMP. cEquimolar mixture of 5’NTP or 5’dNTP. *Grew to the same final OD600, but at a much slower growth rate. ND = not determined +++ = Wild type level of growth ++ = Intermediate growth + = Low level of growth - = No growth

197

The ΔxdsΔdns mutant displayed wild type growth in all tested conditions. This was as expected, as the nucleases should not be required for acquiring Pi from nucleotides. The ΔphoB strain initially grew at the same rate as the wild type, but it only doubled once before ceasing to grow further. Since the ΔushA strain does not grow at all, this small amount of growth cannot be from contaminating Pi in the nucleotide stocks.

Since strains were switched from a high phosphate medium into the test condition with no phosphate source other than 5’ nucleotides, it is possible that this initial growth of

ΔphoB was due to break down of PolyP stores by Ppx. However, since the phoB deletion strain failed to double even once in control medium lacking any source of Pi, this is unlikely to be the case.

As I will demonstrate below, ushA is not regulated by PhoB. Therefore, in the phoB deletion strain UshA is expressed and can lead to accumulation of extracellular Pi.

While it is assumed that a small number of Pst/PhoU complexes exist in the membrane at all times, PhoB is required for upregulating expression of the pst genes. Thus, the initial growth and then rapid cessation of growth of ΔphoB could be due to an inability to transport enough Pi across the cytoplasmic membrane.

In order to more directly test the 5’ nucleotidase activity of UshA, I performed nucleotidase assays, which use a mixture of ascorbic acid and molybdate to detect Pi in solution upon release by nucleotidase activity (Edwards, et al. 1993). Since I was interested in learning whether UshA is required for 5’ nucleotidase activity, and not interested in defining the biochemistry of V. cholerae UshA, I used whole cell lysates for these assays. I expected that the wild type bacterium would harbor 5’ nucleotidase

198 activity, while the ushA mutant should have undetectable levels of activity. Indeed, V. cholerae cell lysate harbors 5’ nucleotidase activity against 5’ dGMP (Activity = 4.97 pmole/min; SE = 0.53), and the activity is decreased 14-fold in the ushA deletion strain to what I believe is the background level (Activity = 0.35 pmole/min; SE = 0.12) (Figure

IV-5a). Both the no substrate control and a substrate only control had no detectable accumulation of Pi (data not shown).

The online program, PSORT (http://www.psort.org/psortb/index.html) categorizes UshA as a periplasmically localized protein, consistent with it having a predicted secretion signal sequence according to SignalP (http://www.cbs.dtu.dk/services/SignalP-4.1/)

(Nakai and Kanehisa 1991, Petersen, et al. 2011). Since nucleotides are not thought to be able to traverse the inner membrane, I hypothesized that comparison of UshA nucleotidase activity between cell lysates and unlysed cells would allow me to address the localization of the protein. Indeed, the nucleotidase activity obtained from unlysed, whole cells, is nearly identical to that of cell lysates for 5’ dGMP (Activity = 6.06 pmole/min; SE = 0.75). Additionally, the V. cholerae nucleotidase activity against 5’ dCMP is nearly twice that of 5’dGMP (Rate = 11.41 pmole/min; SE = 1.16) (Figure IV-

5b). These data suggest that UshA is secreted from the cytoplasm and remains either in the periplasm or cell-associated. Since I did not assay the culture supernatants, I was unable to address whether UshA is present in extracellular space. The increased activity with the 5’dGMP substrate may suggest that UshA preferentially cleaves pyrimidine nucleotides. Further work with a purified protein could address this question of substrate affinity.

199

A) -** Cell lysates B) -** Whole cells

,+* ,+* dCMP dGMP ,** ,** !"#$%&'() !"#$ +* +* (/2#)%3)#)45)$ (/2#)%3)#)45)$ * * * + ,* ,+ * + ,* ,+ ."/)%0/1 ."/)%0/1

Figure IV-5. UshA is required for 5’ nucleotidase activity of V. cholerae. Wild type and ΔushA strains were grown to an OD600 of ~0.5 in 10 mL LB cultures. Cultures were washed once in 10 mM Tris pH 8 and either A) lysed by sonication or B) assayed directly as unlysed cells. At time 0 lysates/ whole cells were mixed with either 10 mM 5’ dGMP (closed symbols) or 10 mM 5’ dCMP (open symbols). At 0, 5, 10, and 15 min after addition of the substrate, aliquots of the reaction were removed and mixed with 0.1 N HCl to prevent further enzymatic activity. After all samples were collected, cellular debris was removed by centrifugation and the supernatants were incubated with the ammonium molybdate solution (1% ascorbic acid and 1 N H2SO4) at 45°C for 20 min. Picomoles of phosphate released by enzymatic activity was determined by measuring the OD at 820nm and converting to pmole through use of a standard curve. The mean and standard error of at least two replicates are shown for each assay.

200

IV.2.5. Deletion of ushA does not affect infant mouse colonization of V. cholerae We were interested if 5’ nucleotidase activity (encoded by ushA) is important for colonization. Previous reports did not report ushA as preferentially induced during infection of infant mice (versus in vitro), or as contributing to colonization of the infant rabbit small intestine (Lombardo, et al. 2007, Schild, et al. 2007, Kamp, et al. 2013).

However, none of these studies were completely saturating, and thus could have missed ushA. Additionally, the infant rabbit and infant mouse models of infection are quite different (Ritchie, et al. 2010, Kamp, et al. 2013). Therefore, we tested the fitness of an ushA deletion mutant when competed with the wild type in the infant mouse model of infection (Figure IV-6). The median competitive index for the experiment was 1.7, and this was not statistically different from 1 as measured by a Wilcoxon signed rank test

(GraphPad Prism version 5.0; GraphPad Software; San Diego, CA). Interestingly the

ΔushA strain was 100-fold outcompeted by the wild type in one of seven mice. This could either be a chance error in the experiment (e.g. only part of the intended inoculum was administered, or poor colonization by the inoculum resulting in a bottleneck) or due to real biological differences in the mice or bacteria. For example, if the outlier mouse had a more phosphate limiting environment than the other six mice, it is possible that ushA was more essential in that mouse. Regardless, overall it is clear that ushA is not essential for infant mouse colonization. This suggests that 5’ nucleotides may not be a major alternative source of phosphate in the infant mouse small intestine.

201

(' 4 - 3 %

2 ( $ 1 # # - " $ ! 0 . /

6 '&( . # - " , ! 5 + '&'( * ) '&''( !"#$%&# !"#$

Figure IV-6. ushA is not required for infant mouse small intestinal colonization. Wild type (ΔlacZ) and ΔushA strains were grown to mid-exponential phase (OD600 ≈ 0.5). Excess phosphate was removed from the media via two rounds of washing in MOPS glucose media (no phosphate source). Bacteria were resuspended to an OD600 = 0.025, mixed 1:1, and back diluted 1:10 in the same medium. Five-day-old infant mice were inoculated with 50 µL of bacteria. After 24 hours mice were euthanized and the small intestines were homogenized and plated on LB Sm plates supplemented with X-gal. Each dot represents an individual mouse, and the line represents the median competitive index. Competitive indices were calculated by the following equation: (ΔushA out/ wild type out) / (ΔushA in/ wild type in).

202

IV.2.6. V. cholerae is able to utilize 5’ nucleotides as a source of nitrogen, but not carbon, and ushA is required for this phenotype

In addition to Pi, nucleotides provide a source of N and C. Like Pi, fixed-nitrogen is also typically at growth-limiting concentrations in aquatic environments, while its availability in the human small intestine is unknown. Therefore we hypothesized that V. cholerae could use nucleotides as a source of N and C and that ushA would be important for this phenotype. I tested whether V. cholerae could utilize nucleotides as a source of nitrogen by growing the bacteria in a modified M9-glucose medium that lacks a fixed- nitrogen source. The cultures were supplemented with 0.5 mM ammonium chloride

(readily usable nitrogen source), 2.5 mM 5’dGMP/ 5’dCMP, or no source of fixed- nitrogen (Figure IV-7a,b,c) and assayed for growth. As predicted, I found that V. cholerae is able to utilize 5’ nucleotides as a N source, and this ability is dependent on ushA.

Adenosine and cytidine deaminases, which are found in the cytoplasm and catalyze the removal of the amine group from adenosine and cytidine, have been well characterized in

E. coli (Carter 1995). Therefore, I hypothesized that V. cholerae acquires N primarily from the amino group of nucleosides, rather than the N within the pyrimidine or purine ring. I tested this hypothesis by complementing the ushA deletion with addition of thymidine to the growth medium. Thymidine is the only nucleoside in DNA that does not contain an amino group, and therefore, if my hypothesis is correct, the nucleoside should not be able to complement the ΔushA strain. Indeed, the ΔushA strain was unable to grow when thymidine was supplied as the only source of N (Figure IV-7d). Complementation with cytosine, which bypasses the need for UshA, could further address this question.

203

D9 !"#$%&$'(4$$)* E9 '?$'13@?A6B F?G@H6 / / 01*2$3456 01*2$3456 !"#$ !"#$ ! ! ! ! - !"/ - !"/ ; ; : :

!"!/ !"!/ ! +!! ,!! -!! .!! /!!! ! +!! ,!! -!! .!! /!!! 71%6$8%9 71%6$8%9

)9 +"#$%&$2)&<=2>&< ;9 +"#$%&$7C4%121B6 / / 01*2$3456 01*2$3456 !"#$ !"#$ ! ! ! ! - !"/ - !"/ ; ; : :

!"!/ !"!/ ! +!! ,!! -!! .!! /!!! ! +!! ,!! -!! .!! /!!! 71%6$8%9 71%6$8%9

Figure IV-7. ushA is required for utilization of 5’ nucleotides as sources of N. Bacterial cultures were grown to mid-exponential phase in M9-glucose supplemented with 0.5 mM ammonium chloride, washed twice in medium lacking N, and inoculated in 200 µL of M9-glucose supplemented with A) 0.5 mM Ammonium chloride, B) no fixed- nitrogen source, C) 2.5 mM 5’dCMP/ 5’dGMP, or D) 2.5 mM thymidine at an OD600 ~ 0.05. Growth was monitored over night in a 96-well plate using the plate reader. The mean of three biological replicates is plotted.

204

IV.2.7. The competence pilus major subunit, PilA, is not required for growth on eDNA when supplied as the sole source of phosphate Despite the fact that UshA contributes to nearly 100% of the 5’nucleotidase activity in V. cholerae under the conditions we have tested, deletion of ushA does not prohibit use of eDNA as the sole source of Pi (Figure IV-3). Therefore, we hypothesized that eDNA was somehow being transported into the cytoplasm. One such possible mechanism is the natural competence machinery, whereby the competence pilus could be taking up strands of DNA that are then degraded inside of the bacterium. If the competence pilus is involved in utilization of eDNA as a source of Pi, this would be a newly described function for this protein complex in V. cholerae. The competence pilus in Haemophilus influenzae was recently shown to be induced under nucleotide limiting conditions, suggesting that using competence for nutrient acquisition would not be unique to V. cholerae (Sinha, et al. 2013).

In order to test this hypothesis, I made a clean deletion of pilA, which encodes the major competence pilin in V. cholerae, in both the wild type and ΔushA backgrounds (Metzger and Blokesch 2014). I tested these mutants in the eDNA growth assay, and found that deletion of pilA does not impact the ability of V. cholerae to grow under the assay conditions (Figure IV-8). Additionally, a ΔushAΔpilA mutant behaves exactly like the single mutant, ΔushA, which displays a slight kinetic growth defect at first, but eventually reaches the same optical density as the wild type. Thus, I concluded from this experiment that the competence pilus of V. cholerae does not significantly contribute to the ability of the organism to use eDNA as a source of Pi.

205

A) !"#$%&$'()*'(+,-$./)%$012 B) 1)$'()*'(+,- 3! 3! >?@A$,B'- !"#$ ! ! 3 3 %&'$ ! ! = = %&'$( !"#$ 0 0 )*"$ *+"$$ < < !"3 !"3

!"!3 !"!3 ! 3! 4! 5! 6! #! ! 3! 4! 5! 6! #!

78&9$:(; 7?%-$:(;

Figure IV-8. pilA does not contribute to growth on eDNA as a source of phosphate. This growth assay was performed as described in Figure IV-1, using mid-exponential phase bacteria, which were washed twice before putting into the test conditions. The growth medium used was MOPS-glucose supplemented with either 0.5 mM phosphate from sheared salmon sperm DNA A) or no phosphate B). Shown is a representative graph of the mean of two biological replicates assayed on the same day.

206

Induction of pilA is known to occur in the presence of chitin or GlcNac (Meibom, et al. 2004), however the role of nucleotides/DNA or phosphate limitation in pilA induction has not been tested to my knowledge. I reasoned that if pilA is transcribed independently of GlcNac/chitin in response to nucleotides/DNA or phosphate limitation, this could suggest that even though the pilus isn’t required for growth on eDNA as a source of Pi, it may nevertheless contribute to the phenotype. However, in order for this hypothesis to fit with the existing data, the competence pilus would have to be inactive in the ΔxdsΔdns mutant, which is unable to grow using eDNA as a source of Pi. Therefore, I hypothesized that nucleotides (which are produced by Xds and Dns), rather than eDNA would induce expression of pilA.

To test this, I measured expression of pilA in wild type and ΔxdsΔdns strains that were grown in high phosphate MOPS-glucose, washed twice and resuspended in one of three conditions: no phosphate source, 0.5 mM phosphate from DNA, or 0.1 mM phosphate from 5’dCMP/5’dGMP (Figure IV-9a). Contrary to my hypothesis, pilA was induced by eDNA (approximately 10-fold) but not in the presence of nucleotides.

Additionally, induction of pilA was independent of xds and dns. Next, I measured expression of pilA after transition from high phosphate to no phosphate medium to determine if phosphate starvation induces transcription of the pilus gene (Figure IV-9b). I measured expression of phoB as a control for the Pho regulon. Expression of phoB was as expected and exhibited 10-fold induction in the phosphate depleted condition when compared with the high phosphate condition. However, pilA expression levels were the same in both conditions. Taken together, these results suggest that pilA is induced by the presence of eDNA, but not nucleotides or phosphate limiting conditions. However, it is

207

A) # 8%69(-:). . 7 %

- !"#(;<(9=<>?91<> , 6

. !"@(;<(ABC 5 $ ( # 2 ! *

( !"# % ( + * + - . 5 ) 4 3 !"!# 8%69(-:). )*+, *-+,, D-5,%2

B) #! . 7 % - $%&'()'*+)',-. , 6

. # 5 $ /*0()'*+)',-. ( # 2 ! * ( % ( + * + -

. !"# 5 ) 4 3 !"!# !"#$ !%&' 1.2. ;.,+E5.9

Figure IV-9. Affect of eDNA, nucleotides, and Pi limitation on expression of pilA. Expression of pilA was measured using qRT-PCR. Bacterial cultures were grown to mid- exponential phase in MOPS-glucose medium supplemented with 10 mM KH2PO4, washed twice in no phosphate MOPS, and resuspended in MOPS-glucose medium supplemented with: A) No Pi MOPS-glucose medium, MOPS-glucose medium + 0.1 mM

5’dCMP/5’dGMP, or MOPS-glucose medium + 0.5 mM Pi from DNA, or B) 10 mM

KH2PO4 or no source of Pi. RNA was isolated from cultures after A) four hours or B) two hours of incubation at 37°C with aeration. In both A) and B) expression levels are shown as relative to the housekeeping gene rpoB. The mean and standard error of at least two biological replicates are shown.

208

unlikely that the competence pilus contributes substantially to growth on eDNA as the sole source of phosphate as pilA was induced in the ΔxdsΔdns mutant, but this strain is unable to use eDNA as a source of Pi.

IV.2.8. Identification of additional putative nucleotidases Since the competence machinery does not appear to contribute to growth on eDNA as a source of Pi, we hypothesized that V. cholerae produces an additional nucleotidase that compensates for the loss of ushA during growth on eDNA. We believed that this nucleotidase would demonstrate activity against 3’nucleotides, as UshA accounts for the 5’ nucleotidase activity in the cell. In order for this hypothesis to have merit, V. cholerae would have to be able to produce both 5’ and 3’ nucleotides from DNA. The precise activities of the V. cholerae Xds and Dns extracellular nucleases have not been demonstrated as yet, and thus it is unclear what nucleotide-related substrates are produced by their activity on eDNA. For example, it is unknown if they exhibit 3’ – 5’ processivity and/or 5’ – 3’ processivity, whether they release mono-, di- or poly- nucleotides, and whether 5’ or 3’ phosphates are revealed after cleavage. BLAST analysis of the two nucleases suggested that while Dns has strong identity to EndA-type endonucleases that release 5’-phosphorylated nucleotides, Xds harbors a YhcR domain.

While Xds is a exonuclease, YhcR is an endonuclease from Bacillus subtilis that releases

3’ monophosphorylated nucleotides (Oussenko, et al. 2004, Seper, et al. 2011).

Therefore, it seems likely that V. cholerae has the capacity to release both 5’ and 3’ phosphorylated nucleotides.

In order to identify additional V. cholerae nucleotidases I performed two searches.

First, I used the UshA sequence as a query for a BLAST analysis of the V. cholerae

209 genome. I identified two putative nucleotidases: VC2562/CpdB (25% identical) and

VCA0545 (23% identical). Using these genes as queries for additional BLAST analysis, I found that CpdB clearly resides in the class of 2’3’ cyclic phosphodiesterases.

Alternatively, VCA0545 has 63% identity to NadN, a 5’ nucleotidase from Moritella dasanensis. PSORT analysis (Nakai and Kanehisa 1991) of VCA0545 and CpdB suggested that both proteins are localized to the periplasmic space. As a secondary search for nucleotidases, I searched the V. cholerae genome annotation for “nucleotidase”.

Using this search I again found CpdB, as well as VCA0608/YjjG. PSORT predicts YjjG as a cytoplasmic protein, however a different algorithm (TMHMM) predicts the protein is secreted from the cell (Nakai and Kanehisa 1991, Krogh, et al. 2001).

IV.2.9. Identification of CpdB as a 3’ nucleotidase I used natural co-transformation to make most of the possible single, double, and triple deletion mutants of cpdB, VCA0545, and yjjG by giving naturally competent V. cholerae PCR constructs for all three deletions along with a selectable marker DNA fragment (Dalia, et al. 2014). The mutants I obtained are: ΔcpdB, ΔyjjG,

ΔyiiGΔVCA0545, and ΔcpdBΔyjjGΔVCA0545. None of the strains exhibited a growth defect on 5’dGMP or 5’dCMP when either was supplied as a Pi source (Figure IV-10).

Additionally, the triple mutant grew as well as wild type in the eDNA growth condition

(Figure IV-11).

Since I could not identify any activity of the three putative nucleotidases against

5’ nucleotides, I next tested their action against 3’ nucleotides. Unfortunately, 3’ deoxynucleotides are either not commercially available, or are prohibitively expensive; therefore, I have examined 3’ nucleotidase activity using RNA nucleotides. First, I

210

!"#7%& B4 !"#$%& C4 , , ' ' ' ' *

* '-,

'-, 6 6 5 5

'-', '-', ' ('' )'' *'' +'' ,''' ' ('' )'' *'' +'' ,'''

./012304 ./012304

892:;9<:;=>1 $4 ,

?/@#2>A:1

' !"#$ ' * '-, %&&' 6

5 %&&' !"#$%&% !"#$ %&&' !"#$%&%

'-', ' ('' )'' *'' +'' ,'''

./012304

Figure IV-10. Growth of cpdB, yjjG, and VCA0545 mutants on 5’ nucleotides. Overnight MOPS-glucose (10 mM KH2PO4) cultures were back diluted to an OD600 ~ 0.05 and grown to mid-exponential phase. Cultures were washed two times in MOPS medium containing no phosphate and inoculated into 200 µL MOPS-glucose medium with A) 0.1 mM 5’dCMP, B) 5’dGMP, or C) no phosphate source. Strains were grown at 37°C with aeration in a 96-well plate. One biological replicate is shown.

211

!"#$%&'() !"#$ %&&' *+,-./. ' ! ! $ 1

0 !&'

!&!' ! "! #! $! %!

()*+,-./

Figure IV-11. Growth of wild type and ΔcpdBΔyjjGΔVCA0545 on eDNA. This growth assay was performed as described in Figure IV-1, using mid-exponential phase bacteria, which were washed twice before putting into the test conditions. The growth medium used was MOPS-glucose supplemented with 0.5 mM phosphate from sheared salmon sperm DNA. One biological replicate is shown.

212

determined the ability of V. cholerae to grow using 3’AMP or 3’CMP as a sole source of phosphate and found that the wild type strain is able to utilize 3’AMP, but not 3’CMP, as a source of phosphate (Figure IV-12). As expected, deletion of ushA did not impact the ability of V. cholerae to grow on 3’AMP. However, the ΔcpdB strain is unable to grow on 3’AMP, suggesting that this gene encodes a 3’ nucleotidase. Deletion of the other two putative nucleotidases did not affect the ability of V. cholerae to grow in 3’AMP (data not shown).

In order to further support our finding that cpdB is a 3’ nucleotidase, we performed nucleotidase assays using 3’AMP and 3’CMP as substrates for the assay. As suggested by the growth assays, we found that V. cholerae is unable to hydrolyze 3’CMP, but it is able to remove the Pi from 3’AMP. Furthermore, while ushA is not required for

3’AMP nucleotidase activity, cpdB is required (Figure IV-13). I have not performed this nucleotidase assay with the single deletion of cpdB, however, while I was working on this document Heather Kamp, a post-doc in our laboratory, performed this experiment.

Heather’s data support my findings with the triple mutant. Since other 3’ nucleotide substrates are not available, we were unable to determine if CpdB is required for growth and/or 3’ nucleotidase activity on other substrates.

213

A) 0.1 mM 3’ AMP B) 0.1 mM 3’ CMP

& & ! ! ! ! $ $ 0 0 / /

!'& !'&

! "!! #!! $!! %!! &!!! ! "!! #!! $!! %!! &!!! ()*+,-*. ()*+,-*.

C) No phosphate 1)23,456+ & !"#$ ! !

$ %&'( 0 /

!'&

! "!! #!! $!! %!! &!!! ()*+,-*.

Figure IV-12. cpdB is required for growth on 3’CMP. Overnight MOPS-glucose (10 mM KH2PO4) cultures were back diluted to an OD600 ~ 0.05 and grown to mid-exponential phase. Cultures were washed two times in MOPS medium containing no phosphate and inoculated into 200 µL MOPS-glucose medium with either A) 0.1 mM 3’AMP, B) 3’CMP, or C) no phosphate source. Strains were grown at 37°C with aeration in a 96-well plate. The mean of four biological replicates, assayed on two separate days, are shown.

214

A) 3’ AMP B) 3’ CMP 1$,0'23*& ǻ()*+ ǻ!"#$ǻ%&&'ǻ9&$

*%+,&'-&,&./&0 *%+,&'-&,&./&0

! ! " ! ! " #$%&'(%) #$%&'(%)

Figure IV-13. cpdB, but not ushA, is required for 3’ nucleotidase activity. Wild type and mutant strains were grown to an OD600 of ~0.5 in 10 mL LB cultures. Cultures were washed once in 10 mM Tris pH 8 and assayed directly as unlysed cells. At time 0 whole cells were mixed with either A) 10 mM 3’AMP or B) 10 mM 3’CMP. At 0, 5, 10, and 15 min after addition of the substrate, aliquots of the reaction were removed and mixed with 0.1 N HCl to prevent further enzymatic activity. After all samples were collected, cellular debris was removed by centrifugation and the supernatants were incubated with the ammonium molybdate solution (1% ascorbic acid and 1 N H2SO4) at

45°C for 20 min. Picomoles of Pi released by enzymatic activity was determined by measuring the OD at 820nm and converting to pmole through use of a standard curve. The mean and standard error of at least two replicates are shown for each assay.

215

IV.2.10. Deletion of cpdB and ushA does not affect the ability of V. cholerae to grow using eDNA as the sole source of phosphate Thus far, I have identified UshA as the major 5’ nucleotidase able to support growth on 5’ nucleotides, and CpdB as the major 3’ nucleotidase able to support growth on 3’ nucleotides. Thus, I hypothesized that deletion of both ushA and cpdB would prevent growth on eDNA when it was supplied as the only phosphate source. To test this,

I performed a growth assay with wild type, ΔushA, ΔcpdB, and ΔushAΔcpdB. Contrary to my hypothesis, the double deletion strain is still able to grow as well as wild type with phosphate supplied from eDNA (Figure IV-14).

I have proposed two possible explanations for the ability of the ΔushAΔcpdB V. cholerae to utilize eDNA as a source of phosphate. First, there may exist additional 3’ nucleotidases specific for thymidine or guanosine nucleotides that are responsible for the continued growth in eDNA. However, due to availability and/or cost limitations I was unable to obtain 3’TMP or 3’GMP to test this. Thus, I cannot rule out this possibility.

Additionally, I was unable to address whether cpdB is able to act against deoxy nucleotides, which are more relevant to eDNA.

The second possible explanation is that either Xds or Dns harbors an uncharacterized phosphatase activity in addition to the nuclease activity. Combined nuclease/nucleotidase activity within a single protein is not a novel concept. One such protein has been described in Leishmania spp. and it was shown to be important for uptake of purines as well as protection from NETs (Hammond and Gutteridge 1984,

Guimaraes-Costa, et al. 2014). Although BLAST analysis of Dns revealed that the EndA

216

A) 0.5 mM phosphate from DNA B) No phosphate 2)34,567+ ' ' !"#$ %&'( ! ! ! ! !"#$) %&'( $ $ 1 1 !&' !&' 0 0

!&!' !&!' ! "! #! $! %! ! "! #! $! %! ()*+,-./ ()*+,-./

Figure IV-14. Deletion of cpdB, alone or together with ushA, does not affect growth on eDNA as a sole source of Pi. This growth assay was performed as described in Figure IV-1, using mid-exponential phase bacteria, which were washed twice before putting into the test conditions. The growth medium used was MOPS-glucose supplemented with either A) 0.5 mM phosphate from sheared salmon sperm DNA, or B) no phosphate source. One biological replicate is shown.

217

domain accounts for the entire amino acid sequence, the Xds protein is largely comprised of sections of unknown function. Therefore, I hypothesized that Xds, rather than Dns, is likely to harbor non-nuclease functions (e.g. phosphatase activity).

To test whether Xds or Dns may have phosphatase activity, I performed growth analysis using eDNA as a Pi source with wild type, ΔxdsΔdns (a negative control for growth), ΔxdsΔushAΔcpdB, and ΔdnsΔushAΔcpdB. A previous report suggests that single deletions of xds or dns have only minor growth defects when eDNA is supplied as the sole phosphate source (Seper, et al. 2011). Considering this, together with the fact that the double deletion ΔushAΔcpdB is able to grow as well as wild type in this condition, I reasoned that I would see growth in the triple mutants unless I managed to delete all phosphatase activity in the bacterium when adding the deletion of either xds or dns to the ΔushAΔcpdB background. I have only tested these strains once and this assay needs to be repeated before publishing or drawing definitive conclusions. However, while the ΔdnsΔushAΔcpdB strain grew as well as wild type, the ΔxdsΔushAΔcpdB exhibited a severely diminished growth rate (Figure IV-15). Thus, these data support the hypothesis that Xds may indeed harbor phosphatase activity.

218

'()*+,-./ & ! !"#$ "%# ! $

6 !"#$ &#'($ )*"+

5 "%#+ &#'($ )*"+ !%&

!%!& ! "! #! $! 0(1/+234

Figure IV-15. Triple mutant of xds, ushA, and cpdB inhibits growth on eDNA. This growth assay was performed as described in Figure IV-1, using mid-exponential phase bacteria, which were washed twice before putting into the test conditions. The growth medium used was MOPS-glucose supplemented with 0.5 mM phosphate from sheared salmon sperm DNA. One biological replicate is shown.

219

IV.2.11. Genetic regulation of cpdB and ushA Since both ushA and cpdB are important for utilization of phosphate derived from nucleotides, I hypothesized that these genes would be induced under low phosphate conditions when their action would be advantageous. However, considering that I identified the ushA phenotype under high phosphate conditions when PhoB is inactive, I hypothesized that PhoB does not contribute to regulation of ushA. To test this, I used qRT-PCR to measure the expression of cpdB and ushA following two hours of incubation in either high- or no-phosphate MOPS-glucose medium. Prior to the two-hour incubation, the cells were grown to mid-exponential phase in high phosphate MOPS-glucose medium. As a control, I measured expression of phoB, which is induced in low phosphate as part of the phosphate starvation response (Diniz, et al. 2011). Both ushA and cpdB are induced by phosphate limiting conditions (Figure IV-16a); cpdB exhibited a 10-fold increase in transcription when switched from high- to no-phosphate (comparable to induction of phoB), whereas ushA exhibited a 2-fold transcriptional increase. To test whether the phosphate response regulator, PhoB, was responsible for the increased expression of cpdB and ushA in low phosphate, I measured expression of these genes in the wild type, ΔphoU (in which phoB is constitutively active), and ΔphoUΔphoB (Figure

IV-16b). Despite the robust induction of the control gene, phoX, I was unable to demonstrate a dependence of cpdB or ushA expression on PhoB activation. Therefore, cpdB and ushA are induced in low phosphate by a PhoB-independent mechanism.

220

A) #! $%&'()'*+,'-./ / 0*(,'*+,'-./ : % . - 9

/ # 5 $ ( # 2 ! * * % ( + * + .

/ !"# 5 , 8 7 !"!# !"#$ %!&$ '(") 1/2/(3/-+45/6

#! B) ;<( / : %

. # !"#, - 9 !"#,( !"#$ / 5 $

( !"# # 2 ! * * % ( + * !"!# + . / 5 ,

8 !"!!# 7 !"!!!# !"#+ %!&$ '(") 1/2/(3/-+45/6

Figure IV-16. Regulation of ushA and cpdB by phosphate and PhoB. A) Bacterial cultures were grown to mid-exponential phase in MOPS-glucose medium supplemented with 10 mM KH2PO4, washed twice in no phosphate MOPS, and resuspended in MOPS-glucose medium supplemented with either 10 mM KH2PO4 or no phosphate. RNA was isolated from cultures after two hours of incubation at 37°C with aeration. B) RNA was collected from mid-exponential phase LB grown bacteria. In both A) and B) expression levels are shown as relative to the housekeeping gene rpoB. The mean and standard error of at least three biological replicates are shown.

221

IV.3. DISCUSSION In the first two sections of this thesis I described our finding that PhoB is an activator of the gene encoding a secreted exonuclease, xds. This regulation occurs under phosphate limiting conditions. This led us to the hypothesis that Xds is involved in Pi assimilation from eDNA. Indeed, as we completed the work discussed in Chapters II and

III, Seper and colleagues (2011) reported that V. cholerae can use eDNA as a source of Pi and that both of the secreted nucleases, Xds and Dns, are required for this ability. The missing link between nuclease-mediated break down of DNA into nucleotides and uptake of Pi by Pst/PhoU (or another system) is the removal of the Pi from the nucleotide. Thus, as described in this chapter, we focused our work on identifying the phosphatases/ nucleotidases that allow growth on DNA as a source of Pi.

IV.3.1. Discussion of the nucleotidases identified in this work We first hypothesized that PhoX, the V. cholerae alkaline phosphatase, would provide this phosphatase activity. PhoX has been tied to growth on DNA and/or nucleotidase activity in other organisms (Pinchuk, et al. 2008, Sebastian and Ammerman

2011). Here, we found that a deletion of phoX grew as well as the wild type when using eDNA or 5’ nucleotides as a source of Pi. However, we have not ruled out whether PhoX can act on other nucleotides, as our eDNA growth assay only shows whether a gene is required for growth. These results are not surprising considering that V. cholerae PhoX was previously demonstrated to be inactive against 5’AMP or 5’ATP (Roy, et al. 1982).

Although these authors did not test PhoX for phosphatase activity against other nucleotides, it is unlikely that the generally non-specific phosphatase would act on some nucleotides but not others. Additionally, we have not demonstrated that phoX is

222 expressed under the conditions under which bacterial growth has been assayed. We assume that the concentration of eDNA added to the culture medium results in an apparent low phosphate concentration for the bacteria, especially considering the extremely slow growth rate exhibited. Under such low phosphate conditions, phoX would be induced. Thus, we conclude that PhoX is unlikely involved in Pi acquisition from

DNA and nucleotides.

Using both a blue/white (XP) screen and BLAST searches, we identified several putative phosphatases. Of these, two were found to be required for growth on nucleotides: UshA – UDP-sugar hydrolase/ 5’ nucleotidase, and CpdB – 2’3’ cyclic phosphodiesterase/ 3’ nucleotidase. These proteins have been described and implicated in

Pi acquisition in several organisms, and thus it is not surprising that we have identified them as such in V. cholerae (Trulzsch, et al. 2001, Rittmann, et al. 2005, Pinchuk, et al.

2008). Indeed, we have shown that V. cholerae UshA is required for growth in 5’ nucleotides (dGMP, dCMP, AMP, and CMP) and that it accounts for nearly all of the 5’ nucleotidase activity against dGMP and dCMP. We also demonstrated that CpdB is an active 3’ nucleotidase that is required for growth of V. cholerae on 3’AMP as a source of

Pi. Although we have focused on describing V. cholerae UshA and CpdB as 5’ and 3’ nucleotidases, respectively, we have not determined if these proteins carry the bifunctional activities (UDP-sugar hydrolase and 2’3’ phosphodiesterase, respectively) as described in other organisms. By using purified protein and enzymatic assays and/or point mutants one could address this question. I hypothesize that the bifunctional aspects of the proteins are not essential for acquisition of Pi.

223

We have characterized UshA as putatively periplasmic in location. This localization fits with the function of UshA to remove the Pi group from nucleotides prior to uptake of each part into the cytoplasm. The periplasmic localization was based both on

PSORT analysis and nucleotidase assays using whole cells as the source of enzyme in the reaction. Although cells were not intentionally lysed for the whole cell nucleotidase assays, it is possible that a small amount of lysis occurred in the sample preparation or during incubation in the assay buffer. One way to control for this would be to plate for viable counts prior to sample preparation, as well as after completion of the assay4. An alternative method for localization of UshA is to tag the protein with an immuno-reactive peptide and perform western blots on fractionated bacterial samples. I have performed preliminary studies using a FLAG-tagged (C-terminal) UshA protein. I showed that this protein was able to complement the ushA deletion strain for growth in 5’ nucleotides and

I’m currently working on optimizing the fractionation protocol such that I can localize the protein. I have not performed any localization studies with CpdB, but I hypothesize that the protein is also located either in the periplasm or in the extracellular space.

Studies with E. coli UshA have shown that the protein is localized both cytoplasmically and periplasmically, however nucleotidase activity is only present in the periplasm. This seeming contradiction has been explained through the discovery of a cytoplasmic inhibitor of UshA (Glaser, et al. 1967, Neu 1967b, Innes, et al. 2001). The

4 Indeed, Heather Kamp conducted several nucleotidase experiments for me while I was writing this document. Heather performed this plating control that I have described and she found that there is 99.9 – 99.99% drop in viability of the cells upon transitioning to the assay buffer. Therefore, while I cannot use the “whole cell” data to determine localization, I can still draw the conclusion that 5’dGMP is a better substrate for UshA than 5’dCMP. 224 gene encoding this inhibitor has not been identified, and thus I cannot search for a homolog in V. cholerae. However, this suggests that in localization studies we may find the protein localized to both the cytoplasm and the periplasm. If this were the case, a more accurate way to localize the nucleotidase activity would be to perform nucleotidase activity assays on fractionated cells.

We have preliminarily investigated genetic regulation of ushA and cpdB. We hypothesized that phosphate limitation would induce both nucleotidases. Indeed, both of these genes were expressed upon transition to a phosphate limiting environment, although the induction of ushA was slight. Using a phoU deletion strain, which constitutively activates PhoB, we were unable to induce expression of either ushA or cpdB. This suggests that expression of ushA and cpdB under phosphate limitation is not dependent on PhoB. Our identification of ushA as a PhoB-independent, low phosphate induced gene is in direct contrast to that published by von Kruger and colleagues (2006). This group identified UshA as absent from the periplasmic fraction of phoB mutant of V. cholerae, but not the wild type, when the strains were grown under low phosphate conditions and using 2D-gel electrophoresis (von Kruger, et al. 2006). The difference between my work and that of von Kruger and colleagues (2006) may be explained by the use of different biotype strains of V. cholerae (El Tor vs. classical). However, it is possible that PhoB in addition to a phosphate limitation signal is required for expression and this could also explain the difference between the two studies. This could be tested by measuring expression of the genes under high and low phosphate conditions in the wild type and

ΔphoB strain as was done for xds in Chapter II.

225

Expression of ushA in relation to phosphate concentrations is not consistent between bacteria. For example, ushA is expressed under low phosphate conditions in C. glutamicum and this expression is regulated through the PhoB homolog, PhoS (Rittmann, et al. 2005, Kocan, et al. 2006). Alternatively, E. coli ushA is not regulated by phosphate conditions or PhoB (Burns and Beacham 1986). The low level of induction of V. cholerae ushA led us to hypothesize that an additional regulator is required for full induction of the gene. An alternative explanation for the low level of ushA expression is that the gene is not regulated and the 2 to 3-fold change observed in our qRT-PCR samples was noise in the method. Adding additional replicates to the experiment will address this question.

Examination of the promoter region of ushA using the online promoter prediction software, Softberry BPROM (Solovyev and Salamov 2011), revealed putative binding sites for NarL (overlapping with the predicted -35 site) and FNR (near the predicted +1 transcriptional start site). NarL is the response regulator of the NarXL two-component system, which activates in response to extracellular nitrate (Stewart 1993). As discussed in Chapter II, FNR is a conserved global regulator in bacteria that regulates many genes including those involved in anaerobic growth and whose activity is controlled via the oxygen levels within the cell (Green, et al. 2009). Co-regulation of promoters by FNR and NarL is widely accepted, and generally results in activation of target genes involved in anaerobic respiration (Stewart 1993, Myers, et al. 2013). For example, FNR and NarL work together in E. coli to activate transcription of napF, which encodes nitrate reductase, in the presence of nitrate and absence of oxygen (Myers, et al. 2013).

Reduction of nitrate to nitrite under anaerobic conditions is coupled to the electron

226 transport chain and helps form the proton gradient that allows for the generation of energy within the cell (Stewart 1993).

If V. cholerae ushA is induced by FNR and NarL binding, this would presumably result in the accumulation of ammonia and ribonucleosides inside the bacterium under anaerobic conditions, in the presence of nitrate. It is unclear why the organism would specifically induce ushA under these conditions. V. cholerae encodes a nitrate reductase, and thus the organism can likely respire using nitrate. One potential argument for the uptake of ammonia during anaerobiosis is to allow regeneration of the nitrate pool and further anaerobic respiration. However, ammonia conversion to nitrate is a process that requires oxygen, and thus acquisition of additional ammonia via the break down of nucleosides would not provide an additional source of nitrate for the organism when FNR is active. Thus the connection between FNR, NarL, and UshA is unclear when considering the traditional model of FNR/NarL regulation. It is possible that FNR/NarL act to repress UshA, although to my knowledge this would be the first description of these regulators coordinately repressing transcription. Alternatively, induction of UshA by FNR/NarL could be unrelated to anaerobic respiration, but tied to nucleotide and/or amino acid metabolism. For example, in E. coli, FNR and NarL coordinately induce expression of upp, which encodes uracil phosphoribosyltransferase and is required for generation of UMP from uracil (Myers, et al. 2013). Also, nucleosides taken up through

Nup transporters and subsequently deaminated are important for purine salvaging in the pathogen, H. pylori (Miller, et al. 2012). The ammonia removed from nucleosides could either be recycled to build different nucleoside/nucleotide bases, or used for amino acid metabolism. Indeed, ammonia is required for asparagine synthesis in V. cholerae

227

(Fresquet, et al. 2004). Despite several hypotheses, it remains unclear why FNR and

NarL would induce ushA transcription in V. cholerae.

Interestingly, we found that ushA is required for the use of 5’ nucleotides as a source of N. Although we didn’t directly test whether a nucleoside deaminase is required for this phenotype, we propose that the source of the N is from the amine group based on the fact that thymidine is unable to complement the deletion of ushA. Additionally, preliminary experiments suggest that limitation of a N source in the growth medium does not induce expression of ushA, but it may in fact repress the gene (data not shown). This might suggest that ushA is under control of the NtrC regulator, which controls activity of a N limitation response (Zimmer, et al. 2000). Taken together, it is clear that a more thorough analysis of ushA regulation either through a targeted or open approach will yield potentially interesting results about the biology of the bacterium.

Unike ushA, we found that cpdB is greatly induced (10-fold) by transition to phosphate depleted conditions. Regulation of cpdB in other organisms has not been deeply studied, however the gene is not induced by phosphate limitation in S.

Typhimurium (Kier, et al. 1977). Alternatively, cpdB of both S. Typhimurium and

Yersinia spp. are under control of carbon catabolite repression (Kier, et al. 1977,

Trulzsch, et al. 2001). A cAMP-CRP binding site has been identified in the cpdB promoter of Yersinia entericolitica, and CpdB allows growth of this species on 2’3’- cAMP as the sole source of carbon (Trulzsch, et al. 2001). Using the online promoter prediction tool, Softberry BRPOM (Solovyev and Salamov 2011), I identified two putative CRP binding sites in the V. cholerae cpdB promoter. Thus, it will be interesting

228 to see if cpdB from this organism is carbon catabolite repressed, as well as induced by phosphate starvation.

IV.3.2. Model for utilization of eDNA as a source of phosphate

The utilization of eDNA as a source of Pi requires break down of the DNA strands into nucleotides, removal of the Pi from the nucleotide, and uptake of the Pi into the cell

(Figure IV-17). The two known secreted nucleases of V. cholerae, Xds and Dns, are required for growth on eDNA such that a double deletion is unable to grow (Seper, et al.

2011). This suggests that if there is an additional secreted nuclease produced by V. cholerae, it does not significantly contribute to this phenotype under the conditions tested. Prior to the work described in the chapter no nucleotidases involved in utilization of DNA/nucleotides as a source of Pi (or N or C) had been described in V. cholerae.

Despite our identification that ushA and cpdB are required for growth on extracellular 5’ and 3’ nucleotides as the sole source of Pi, the double deletion strain,

ΔcpdBΔushA is able to grow as well as the wild type strain when using eDNA instead of nucleotides. This suggests to us that an additional phosphatase/nucleotidase is compensating for the loss of CpdB and UshA. Since we were unable to test some nucleotides as substrates for CpdB and UshA, it is possible that this additional phosphatase is substrate-specific for 5’TMP, 3’TMP, 3’GMP, or selected deoxynucleotides. We did identify other nucleotidases (yjjG and VCA0545) that may provide this activity. Alternatively PhoX, which does not act against any of the tested nucleotides may be the missing phosphatase. Using a quintuple knock-out

(ΔushAΔcpdBΔyjjGΔphoXΔVCA0545) could address this question.

229

3’

O O

3’

5’ O

3’ O

O O

3’

O P

5’ 5’

5’ O O 3’

O O

P O

5’

3’

5’

O O

3’

O P O 5’

3’

O 3’ O 5’

5’

O P P

O O O

3’

O 5’

3’

3’ O 5’

O O

O P

5’ P

O O O

3’ O O 3’

O 5’ O O O

O 5’

5’ 3’

O O

O P O

O 3’ 5’ O O P

O O O 3’

5’

P Xds O O P

5’ O 3’ O 5’

O 3’

O O O

3’ O 5’ P P

O O P O

O O O

5’ O O P 5’ O 3’

O 5’ O O

O 3’ 5’

O P 3’ O O O P

O O O

O P O 3’ 5’

O O

5’

3’ O P O 5’ O O

O 3’ P O

O O 3’

5’ O O O

O 3’ P

O 5’ O O 5’

O 3’ O

O P P 5’ O

O 3’ O O O

3’ O 5’ P O O

O 5’

O 3’ O 5’ P O 3’ P

O O

O 5’ O O 3’ O P

O O O 5’

P O 3’ 3’ 5’ O

O O P

5’ O O

3’ O P

O O

O O 3’ O

5’ O O 3’

3’ O 5’ P

O O

O P

5’ 3’

O O

O 3’ O 5’ P

O 5’ O 3’ O O 5’ P O O 3’ O O O O 5’ P O O 3’ O P P O 5’ O O 3’ O O 5’ O 5’ P Dns O 5’ O O 3’ O O 5’ P O O 3’ O 5’ P O 3’ O 3’ 3’ O O Extracellular P

Xds O O O O P space O O 5’

3’

O O 5’ P O O 3’ UshA 5’ 5’ O O P 3’ 3’ O O 5’ O O P 3’ O O Periplasm CpdB Factor X? O O P O O O O P O O O O Low [P ] P i O O O O P O O PstS PhoR Nup PstC PstA PhoU P PstB PhoB PhoB P Cytoplasm O O P xds phoX O O

O O phosphate starvation genes P O O

230

Figure IV-17. Model for the utilization of eDNA as a source of phosphate in V. cholerae. Extracellular DNA is broken down by Xds and Dns in the extracellular space. Dns is an endonuclease in the EndA family of nuclease, and presumably cleaves at the 3’ carbon, leaving a 5’ phosphate attached to the DNA strand. Alternatively, Xds is an exonuclease and may cleave at the 5’ carbon, leaving a 3’ phosphate. Additionally, preliminary evidence suggests that Xds has phosphatase activity against DNA strands. Once produced, nucleotides can pass across the outer membrane into the periplasm through porins. Within the periplasm, UshA and CpdB remove phosphate groups from 5’ and 3’ nucleotides, respectively. An additional nucleotidase active against 3’TMP, 3’GMP, or 5’TMP may also act in the periplasm. Released phosphate can traverse the inner membrane via the Pst/PhoU system, while nucleosides can pass through the nucleoside transporters (e.g. NupC). Low phosphate conditions induce transcription of phosphate starvation genes such as xds and phoX, in a PhoB-dependent manner. Additionally, ushA and cpdB are transcribed under phosphate limiting conditions in a PhoB-independent manner.

231

The alternative explanation, which we performed preliminary experiments to test, is that either Dns or Xds harbors nucleotidase activity in addition to nuclease activity.

Indeed, when we tested a triple deletion of xds, ushA, and cpdB for growth in eDNA as the phosphate source, the strain was unable to grow. However, the ΔdnsΔushAΔcpdB triple mutant behaved similarly to the wild type strain. Thus, we preliminarily concluded that UshA, CpdB, and Xds provide the nucleotidase activity required for growth on eDNA. The fact that the double deletion ΔushAΔcpdB is able to grow as well as wild type suggests that Xds is able to provide a significant level of nucleotidase activity. The source of eDNA in these experiments was sheared into 100-500 bp lengths, which would provide potentially many exposed Pi groups for Xds to remove. I hypothesize that the use of unsheared eDNA in these experiments would make it much more difficult for V. cholerae to acquire phosphate from eDNA using only Xds.

Taken together, the results presented in this chapter have expanded our knowledge of V. cholerae Pi assimilation from eDNA. I have described the nucleotidases required for the phenotype, as well as investigated their regulation. Additionally, I am the first to show that V. cholerae can utilize both 5’ and 3’ nucleotides as a source of Pi, but that the organism is unable to use 3’CMP.

232

CHAPTER V:

SUMMARY AND PERSPECTIVES

233

As a waterborne pathogen V. cholerae must adapt to rapid changes in the surrounding environment. Understanding the adaptation strategies utilized by V. cholerae may be an important step towards controlling the pathogen. The transition of entry into the host, and subsequent virulence gene expression has traditionally received more attention by researchers than have the later stages of infection or dissemination from the host. However, broad changes in gene expression patterns during late stages of infection and dissemination have recently been described using RIVET and microarray analysis

(Larocque et al., 2005; Nelson et al., 2008; Nielsen et al., 2006; Nielsen et al., 2010;

Schild et al., 2007). Additionally, a recent Tn-Seq study from our lab defined genes that are important for host-to-pond survival, using an infant rabbit model of infection (Kamp et al., 2013). Together, these studies have provided insight into the genetic changes of V. cholerae during the late stage of infection and transition to the environment.

V. cholerae cells exiting a host may either end up encountering a new host or persisting for an extended period of time in aquatic reservoirs. It is highly unlikely that the organism can predict which of these scenarios it will encounter. In fact, there is evidence that the bacterium may prepare for both situations by expressing genes necessary for environmental survival as well as those which contribute to a hyperinfectious phenotype. There are two explanations as to how the bacterium can prepare for both distinct situations. One possibility is that each bacterium shed in RWS expresses genes necessary to survive in both situations. This may be energetically taxing to the bacterium, and thus seems unlikely. Alternatively, there could be two subpopulations of V. cholerae exiting a host: hyperinfectious (e.g. motile but non-

234 chemotactic) and ready for the aquatic environment (e.g. chemotactic and with late genes for chitin utilization expressed).

Despite the accumulating data on the transcriptional changes that V. cholerae undergoes during the infection, it has been difficult to determine the precise signals that are inducing these changes. This problem is particularly acute for signals in the host that trigger transcriptional changes in V. cholerae during the late stages of infection. The late stage of infection is thought to confront the bacteria with a high population density, nutrient starvation, and intense host immune pressure including attack by neutrophils, which is consistent with the hypothesis that the bacteria enter a non-growing stationary phase-like phenotype as they prepare to exit the host (Nielsen, et al. 2006, Nielsen, et al.

2010). Both nutrient starvation and quorum sensing likely contribute to regulating the induction of both the mucosal escape response and/or the late gene program. In support of nutrient starvation acting as an important inducing signal, representative genes in all five V. cholerae iron acquisition systems are in fact late genes, suggesting that iron becomes growth limiting at the late stage of infection (Schild, et al. 2007). In an attempt to determine possible inducing signals within a host, one group used mass spectrometry to measure the levels of various metabolites present in cecal fluid collected from infected infant rabbits, as compared with metabolites present in in vitro culture supernatants

(Mandlik, et al. 2011). Although these metabolites do not provide a definitive list of inducing signals, they have provided information about what V. cholerae may be sensing

(or not sensing) in the host. Pi levels were not reported in that study. Here, we took the alternative approach of identifying transcriptional regulators that are required for induction of specific genes that are induced during infection. By identifying such

235 regulators, it was our goal to uncover potential signals involved in inducing transcriptional changes within a host.

In Chapter II we utilized a novel in vitro high-throughput genetic selection to identify regulators of three described late genes: cdpA, emrD, and xds. We were unable to identify any regulators of cdpA. We briefly described the regulation of emrD by PepA and GshB. The connection between these two regulators and emrD is unclear and requires further investigation. However, since both GshB and PepA are connected to glutathione metabolism, there may be a connection between the redox state of the bacterium and induction if emrD. Alternatively, induction of emrD may occur in gshB and pepA deletion strains in order to relieve toxic stress by pumping glutathione intermediates out of the cell. Finally, we described the Pi-responsive PhoB as the major regulator of xds. As expected based on this result, phosphate limiting concentrations induces expression of xds. Thus we hypothesized that phosphate limitation late in infection may be a key cue that V. cholerae senses prior to its transition to the aquatic environment.

Contrary to this hypothesis, we have described follow-up experiments in Chapter

III that demonstrate an induction of xds very early in infection of the infant mouse. As discussed in Chapter III, these results are in conflict to the identification of xds as a late gene (Schild, et al. 2007). However, numerous differences between our study and the late gene study can easily explain the conflicting data. The most important of these variables is the condition under which the inoculum is prepared. If the inoculum is prepared under conditions of excess Pi, then excess phosphate in the inoculum fluid will extend the time until the Pho regulon gets induced during infection. In contrast, if the inoculum is

236 prepared under conditions where phosphate is absent from the inoculum fluid, then the bacteria can respond more rapidly to phosphate limiting conditions in the host intestinal tract. Using the latter inoculum preparation condition, we conclude that a low concentration of phosphate exists in the infant mouse small intestine and this serves as a signal for the induction of in vivo-specific genes. Although we did not test this, the storage of PolyP stores in a mouse inoculum could also alter the timing of Pho regulon induction.

Our identification of xds as an early induced gene, rather than a late induced gene, does not contradict all of the results presented by Schild and colleagues (2007). For example, we have no evidence that cdpA and emrD are not late induced. We also show that PhoB regulates two otherwise uncharacterized late genes: VC0353 and VCA0955.

Unlike xds, PhoB represses these genes. Thus, induction of VC0353 and VCA0955 late in infection fits with our finding that the small intestine is phosphate limiting and reports that RWS contains a high level of phosphate, which would repress PhoB activity and allow for induction of the two late genes (Nelson, et al. 2008). Regardless, our work does demonstrate that large unbiased screens, such as the late gene screen, can have unintended and hidden biases.

In light of our findings that PhoB regulates numerous late genes, a new contradiction arises; if VC0353 and VCA0955 are repressed by PhoB and xds is induced by PhoB, how were Schild and colleagues (2007) able to identify all three of these genes as late induced? One potential explanation is that PhoB activity (and thus phosphate levels) change throughout infection such that the timing of xds induction is separate from that of VC0353 and VCA0955 (Figure V-1). Consistent with this hypothesis, infection

237

A) High 100 Percent resolution

50 concentration Apparent phosphate phosphate Apparent Low 0

RWS

Inoculum Stomach

Small intestine Large intestine

Phosphate levels VC0353 expression xds expression VCA0955 expression B) High 100 Percent resolution

50 concentration Apparent phosphate phosphate Apparent Low 0

RWS

Inoculum Stomach

Small intestine Large intestine

238

Figure V-1. Alternative models of phosphate level and affect on gene expression during infection A) A model for the fluctuation in phosphate level during infection based on Schild and colleagues (2007) where phosphate levels are high in the early stage of infection (<7 hours post inoculation). Once the contaminating phosphate in the inoculum is depleted, phosphate levels drop, which allows activation of PhoB and subsequently induction of xds. Towards the end of infection phosphate levels once again increase, which allows induction of the PhoB repressed late genes, VC0353 and VCA0955. Schild and colleagues (2007) reported that xds exhibits 90% resolution by 21 hours post inoculation. This resolution is likely reached prior to the loss of PhoB activity, which would halt transcription of xds. Since RIVET cannot be used to determine when a gene turns off, it is unknown whether xds remains on, or turns off. B) Based on the data presented in this thesis we propose a model of infection in which phosphate levels are low throughout much of infection. We hypothesize that phosphate level would increase late in infection, as presented in the alternative model, allowing for induction of VC0353 and VCA0955 and repression of xds. Again, although we predict that xds is repressed at the tail end of infection, we have shown its continuous expression in this model based on the output of our RIVET data.

239

studies with both phoB deletion mutants and strains with constitutive PhoB activity suggest that the activity of PhoB is dynamic (von Kruger, et al. 1999, Merrell, et al.

2002b, Pratt, et al. 2009). Additionally, Schild and colleagues (2007) reported induction of xds beginning at 7 hours post infection, whereas VC0353 and VCA0955 were not induced until 21 hours post infection. Thus, the original late gene screen supports a model of differential timing of xds versus VC0353/VCA0955 transcriptional induction.

Furthermore, we have reported that while xds is solely regulated by PhoB under phosphate limiting conditions, VC0353 requires an additional low phosphate specific cue for full activation. In contrast, PhoB regulates VCA0955 independent of the phosphate concentration, potentially via activation of PhoB by a non-cognate Histidine kinase.

These results taken together further demonstrate that our work does not significantly conflict with the findings of Schild and colleagues (2007).

The relevance of small intestinal low phosphate to human V. cholerae infections is unclear. As mentioned above, one main difference between our study and that of Schild and colleagues (2007) is the presence of contaminating Pi in the inoculum used by the latter study. Thus, we hypothesize that the state of V. cholerae with regards to phosphate concentrations or PolyP stores prior to ingestion by a human host may greatly impact the timing of induction of the Pho regulon. RWS is reportedly a high phosphate environment, whereas the aquatic environment may be either phosphate replete (e.g. polluted) or deplete (the usual state of natural bodies of water) depending on the body of water examined (Kamal, et al. 2007, Schild, et al. 2007, Nelson, et al. 2008). Thus human-to- human transmission of V. cholerae may result in delayed induction of the Pho regulon when compared with pond-to-human infections.

240

Human passage of V. cholerae induces the hyperinfectious state of the organism

(Merrell, et al. 2002a). Thus it would be interesting to see if PhoB and phosphate contribute to this phenotype. For example, hyperinfectivity can be modeled using infant mouse passaged bacteria (Alam, et al. 2005). Additionally, the hyperinfectivity phenotype is maintained outside of the host for at least 5 hours (Merrell, et al. 2002a).

Therefore, testing the involvement of PhoB in hyperinfectivity could be accomplished by taking mouse passaged bacteria, washing them twice in a no phosphate MOPS medium and incubating in the same medium for at least 2 hours (when our work suggests the organism senses the switch in phosphate concentration), but less than 5 hours. The control samples would be incubated in a high phosphate MOPS medium. After incubation, the bacteria would then be inoculated into new infant mice in a 1:1 competition. If repression of PhoB by high phosphate levels in RWS (or at a late stage of infection) is required for hyperinfectivity, the bacteria incubated in high phosphate

MOPS will outcompete those incubated in low phosphate MOPS. If this hypothesis is supported by the proposed experiments, it would suggest that removal of phosphate from

RWS or contaminated water could prevent hyperinfectious spread of the bacterium.

Unfortunately, removal of phoshate is much harder than the reverse, and I do not know of any agents that would render the phosphate completely unavailable.

It is clear that much is unknown about fluctuations in phosphate concentrations within a host and how these changes impact the Pho regulon. In an attempt to address this question, we used a RIVET-based assay to describe the timing of xds induction during infection. However, using this assay, we were only able to address the initial induction of the Pho regulon, as excision of the res cassette is an irreversible process. Thus, we were

241 unable to address what occurs with regard to PhoB activity late in infection. Additionally, we have demonstrated that PhoB regulation of in vivo-induced genes is complicated.

While some genes are solely regulated by the response regulator (e.g. xds), other genes appear to integrate additional signals that affect the timing of their expression (e.g.

VC0353 and VCA0955). Thus describing the timing of PhoB activity during infection will not necessarily correlate with the expression of PhoB-regulated genes or with the Pi availability in the host. Since we have shown that xds acts as a traditional PhoB regulated gene (i.e. induced by low phosphate conditions by PhoB), use of a more quantitative assay such as qRT-PCR to measure transcript levels of xds throughout the course of an infection could specifically address the time of PhoB activation and potentially Pi concentrations within the host. Although we attempted to do these experiments, we were unsuccessful likely due to a high level of mouse RNA contamination (even after a eukaryotic RNA removal kit was employed). Use of Taq-man probes, which are highly specific for the transcript but also expensive, rather than SYBR green may address this issue.

Modeling induction of PhoB based on population averages (RIVET and qRT-

PCR) and time-course data does not give a clear picture of where in the host PhoB is induced. These types of studies (like our own) assume a model of infection where V. cholerae bacteria move from one site of the body to the next in a coordinated manner, and that the time post inoculation specifically correlates with a location in the host. This idea is reflected in the models presented in Figure V-1. In reality, we do not fully understand what occurs during infection. Work with the rabbit ligated ileal loop model and immunofluorescence microscopy has allowed researchers to examine bacteria at the

242 single cell level during infection. This work revealed that as the infection progresses some bacteria remain attached to the intestinal epithelial cells, whereas others move into the lumen and prepare for ‘escape’ (Nielsen, et al. 2006, Nielsen, et al. 2010). These results support the idea that the movement of V. cholerae through the intestinal tract is not coordinated. However, the researchers were using a closed model of infection, and it is unclear how related this model is to human infection. Use of the infant rabbit model of

V. cholerae infection, which allows colonization of the small intestine and production of secretory diarrhea, in combination with immunofluorescence microscopy may help address this question (Ritchie, et al. 2010).

After describing xds as regulated by PhoB, we sought to understand this connection further. Since Xds is a secreted exonuclease, we hypothesized that low phosphate induction of xds by PhoB would allow V. cholerae to consume eDNA as an alternative source of Pi. Indeed, while we completed experiments in Chapter II and III,

Seper and colleagues (2011) demonstrated that V. cholerae could use eDNA as the sole source of Pi and that Xds, along with the other secreted nuclease, Dns, are required for the phenotype. Thus, we sought to further characterize the mechanism through which V. cholerae uses eDNA as a Pi source.

In this thesis I have described results that demonstrate the requirement of a 5’ and

3’ nucleotidase (UshA and CpdB, respectively), for the ability of V. cholerae to grow using nucleotides as a Pi source. Additionally, I have shown that these nucleotidases are induced by low phosphate conditions in PhoB independent manner. Since the double deletion of ushA and cpdB is able to grow using eDNA as the sole source of Pi, I was unable to show their necessity for the phenotype. This suggests that additional

243 nucleotidases/phosphatases contribute to growth on eDNA. Indeed, I have shown preliminary evidence that Xds may provide this additional activity. Prior to my work,

UshA and CpdB had not been characterized in V. cholerae or any of the Vibrionaceae.

The physiological relevance of Pi acquisition from eDNA by V. cholerae is unclear. DNA is a known structural component of V. cholerae biofilms. In this context,

Xds and Dns were demonstrated as important for allowing individual bacteria to degrade the DNA structure and escape from the biofilm (Seper, et al. 2011). Since DNA is a rich source of Pi, it seems likely that the organism would take advantage of the released nucleotides and consume them as nutrients. V. cholerae forms biofilms in the aquatic environment where nutrients can be quite limiting. Therefore, I hypothesize that the genes involved in acquiring nutrients from eDNA would be important for survival of V. cholerae in a biofilm. Since E7946, the strain of V. cholerae that I have performed all of my experiments with, does not readily form biofilms, I have been unable to test this hypothesis.

The source of DNA within the V. cholerae biofilm is unknown. Neisseria spp. are known to secrete chromosomal DNA, a phenotype that has been hypothesized as important in both natural competence as well as biofilm formation (Hamilton, et al. 2005,

Lappann, et al. 2010). Like Neisseria spp., V. cholerae might actively secrete DNA for use in the biofilm matrix. Alternatively, the eDNA isolated from the biofilm matrix of V. cholerae may be released upon cell lysis. Aside from the DNA present in V. cholerae biofilms, the aquatic environment contains a great deal of extracellular DNA. Reported concentrations of DNA in the aquatic environment range from picomolar to micromolar amounts (Lorenz and Wackernagel 1994, Bjorkman and Karl 2005). Although we have

244 used a concentration of eDNA approximately 10-fold less than Seper and colleagues

(2007), we did not attempt to titrate the concentration of eDNA to determine whether picomolar or micromolar amounts of eDNA are enough to support survival of V. cholerae. However, the DNA found in biofilms is likely more concentrated, and may be similar to the concentration of DNA that we have used.

Formation of a biofilm on chitin, which is abundant in the aquatic environment, induces natural transformation in V. cholerae (Meibom, et al. 2005). Since the studies that identified eDNA as a structural component of V. cholerae biofilms did not use chitin as the surface to which the bacterium attached, it is unknown if V. cholerae uses eDNA in chitin-attached biofilms as a structural component. Thus, it is unclear if this biofilm associated eDNA is also present for assisting in DNA exchange as is suggested for

Neisseria spp. Additionally, in our studies we explored whether the competence pilus is involved in uptake of eDNA for nutrient acquisition. We were unable to find a connection between competence and use of eDNA as nutrients, either by phenotype testing or transcriptional evaluation. This is unsurprising, as the secreted nucleases are known to diminish natural transformation of V. cholerae. Also, expression of the nucleases does not correlate with expression of competence genes as demonstrated for dns by Blokesch and colleagues (2008), and our work demonstrates this for xds. Thus, while it is probable that V. cholerae can utilize biofilm eDNA as both a source of nutrients and a substrate for the competence machinery, these traits are likely separated.

Separation of these phenotypes may be due to induction of the phenotypes at different times during biofilm formation, or due to expression of the phenotypes in spatially separated locations in the mature biofilm. For example, expression of xds may occur on

245 the surface of the biofilm, where the exonuclease is required for escape of planktonic bacteria. Alternatively, expression of the competence phenotype may occur within the center of the biofilm, where the majority of the eDNA is found and could be used for natural transformation.

Within a human host V. cholerae may encounter eDNA in the form of NETs secreted by attacking neutrophils. Indeed, the secreted nucleases, Xds and Dns, were shown to be important in defending V. cholerae from NET attack by breaking down the structural DNA of the NET (Seper, et al. 2013). Since we have shown that the small intestine of an infant mouse is a phosphate limiting environment, we hypothesize that

DNA from NETs can also serve as a source of phosphate for the pathogen. The ΔxdsΔdns mutant, which is unable to grow when eDNA is supplied as the sole Pi source, displays no colonization defect during infection of infant mice (Focareta and Manning 1991a).

However, the infant mouse model is known to have an impaired innate immunity

(Bogaert, et al. 2009). Indeed infection of adult mice with ΔxdsΔdns revealed a slight defect in colonization when compared with the wild type strain. Additionally, infection of neutropenic adult mice eliminated this defect, which demonstrated that neutrophils were involved in the decreased survival of the mutant (Seper, et al. 2013). Since the nucleases contribute to break-down of NETs, it is hard to know if they also contribute to Pi assimilation from this DNA during infection. Alternatively, a nucleotidase mutant, which is unable to grow using eDNA as a Pi source, is unlikely to contribute to NET dissolution.

Thus, once we have identified a nucleotidase mutant that is unable to grow in the presence of eDNA as the sole source of Pi, we will be able to test whether V. cholerae

246 uses NET DNA as a source of Pi. This mutant may be ΔxdsΔushAΔcpdB, although further testing is required.

Taken together, the work described in this thesis has expanded the field substantially. We have described a novel high throughput genetic selection that we successfully used to identify regulators of V. cholerae late genes. The method has broad application for identifying regulators of conditionally expressed genes, and could be used with any genetically tractable organism. In follow-up experiments from this genetic selection, we demonstrated that phosphate is limiting in the small intestine during V. cholerae infection, as early as 2-4 hours post inoculation of an infant mouse. This knowledge may have impact on the methods used to study infection-induced, PhoB

regulated genes. Additionally, we explored the use of eDNA as a source of Pi and we have identified two key players in this phenotype, which may provide an important source of nutrients both in the host and in the aquatic environment.

247

CHAPTER VI:

MATERIALS AND METHODS

248

All animal experiments were done in accordance with NIH guidelines, the Animal

Welfare Act, and US federal law. Tufts University School of Medicine’s Institutional

Animal Care and Use Committee approved the experimental protocol "B2013-44" that was used for this study. All animals were housed in a centralized and AAALAC- accredited research animal facility that is fully staffed with trained husbandry, technical, and veterinary personnel.

PCR primers used are listed in IV.1.4. All V. cholerae strains were constructed using standard molecular techniques in an Sm resistant derivative of the clinical isolate

E7946 (Mekalanos 1983). Unless stated otherwise, all mutations generated in this study were confirmed by Sanger sequencing by the Tufts University Core Facility (TUCF).

Plasmids were maintained in Escherichia coli DH5αλpir (oriR6K containing plasmids) or DH5α. Unless otherwise noted, the donor strain, E. coli SM10λpir, was used for conjugative transfer of plasmids.

VI.1. CHAPER II

VI.1.1. Media and bacterial strains Bacterial strains were propagated in LB broth with aeration or on LB agar at

37°C, unless otherwise noted. When indicated bacteria were grown in MOPS-glucose minimal media [1x MOPS salts (40 mM MOPS pH 7.4 (3 –(N- morpholino)propanesulfonic acid) (Sigma Aldrich), 4 mM tricine, 0.1 mM FeSO4 

7H2O, 9.5 mM NH4Cl, 0.28 mM KCl, 0.53 mM MgCl2  6H2O, and 50 mM NaCl); 1x

NRES (25 mM of each of the amino acids N, R, E, and S); 1x trace metals (0.005%

MgSO4, 0.0005% MnCl2  4H2O, 0.0005% FeCl3, and 0.0004% nitrilotriacetic acid); and

0.5% glucose] , supplemented with either 0.1 mM (low) or 10 mM (high) KH2PO4.

249

Unless otherwise noted, antibiotics were used at the following concentrations: 100 µg ml-

1 streptomycin (Sm), 100 µg ml-1 ampicillin (Ap), 2 µg ml-1 chloramphenicol (Cm), 50 µg ml-1 spectinomycin (Sp) and 50 µg ml-1 kanamycin (Kn). Addition of 1 mM IPTG to either plates or broth was used to induce transcription from the Ptac promoter.

VI.1.2. Strain construction The cdpA, emrD, and xds genes were transcriptionally fused to two antibiotic- resistance genes, bla and neo, at the native chromosomal locus to generate synthetic operons (late gene-bla and late gene-bla-neo). The fusions were constructed by using splicing by overlap extension (SOE) PCR with the primers listed below, followed by allelic exchange using pCVD442catlac, which is a derivative of the suicide plasmid pCVD442 (Donnenberg and Kaper, 1991). Two successive SOE and allelic exchanges were done as described (Lee et al., 1998; Pratt et al., 2010), except that 2 µg ml-1 Cm was used in place of Ap and bacterial conjugation was performed on 0.8 µm AAWP filters

(Millipore) placed on an LB agar plate for 2 hours at 37°C. Firstly, three primers pairs for each gene fusion were used to generate bla fusions to cdpA, emrD, and xds. These strains were used for the original selection. Secondly, primers listed in the table below were used to generate the bla-neo fusions that were crossed into the chromosome of the bla fusion strains to generate the final cdpA-bla-neo, emrD-bla-neo, and xds-bla-neo fusions. The bla-neo genes were inserted immediately after the late genes, each with their own ribosomal binding site as designed by Osorio et al., 2005 resulting in the sequence

TTTAGGATACATTTTT laying between the late gene stop codon and bla start codon, and between the bla stop codon and neo start codon. No genomic sequence was removed during this construction.

250

Plasmid insertion knock-outs of rseA and cysB were created using the suicide vector, pGP704 (Miller and Mekalanos 1988), with a cat gene in place of bla (Laboratory plasmid). Internal fragements of rseA and cysB were amplified using the PCR primers: rseAins F/ rseaAins R and cysBins F/ cysBins R, respectively. PCR products and the pGP704 plasmid were digested with KpnI and SacI, and subsequently ligated together.

The ligation product was transformed into DH5αλpir for amplification of the vector, and subsequently transferred in to SM10λpir for conjugation with the wild type. Filter matings were performed and colonies were screen for the insertion mutation prior to confirmation via sequencing.

Deletions of phoU, pepA, fexA, fexB, pdhR, ubiC, VC0469, and VC2288 were completed using the Trans-FLP method, which utilizes the natural transformability of chitin-grown V. cholerae and Flp-recombination, as described previously. Briefly, a gene of interest is replaced with a selectable FRT cassette, which is flipped out of the genome by the Flp recombinase carried on pBR-flp (Blokesch 2012; De Souza Silva and

Blokesch, 2010). However, we constructed a new SpR-marked FRT cassette and a new

CmR-marked Flp-delivery vector, pDL1098-flp. The new delivery vector has two major differences from the pBR-flp plasmid. Firstly, pDL1098-flp carries an RP4 oriT, and is therefore transferable via conjugation with a suitable E. coli donor (we used MFDpir

(Ferrieres et al., 2010)). Secondly, similar to pBR-flp, transcription of flp in pDL1098-flp is induced at 40°C. However, unlike pBR-flp, 40°C is a non-permissive temperature for pDL1098-flp replication, and therefore loss of the FRT cassette via recombination at the

FRT sites by Flp, can be completed in the same step as curing of the plasmid from the strain. Taking into account the differences in pBR-flp and pDL1098-flp, the following

251 modifications were made to the Trans-FLP protocol. After growth at 40°C for 6 hours the cells were plated on LB plates with no antibiotic and left at 37°C overnight. The following morning, colonies were replicated onto LB agar, and LB agar supplemented with either Sp or Cm. Colonies that failed to grow in the presence of Sp and Cm were checked by PCR for the deletion. We used the primer pairs F1/R1 and F2/R2 primer pairs listed in the primer table to create the FRT cassette SOE for replacement of each of the loci. Except for ΔphoU, these mutants were only confirmed by colony PCR using primers outside of the SOE PCR product to ensure that the insertion occurred in the correct location of the genome.

The rpoE::mTn10 strains were made via natural transformation of V. cholerae strains (wild type, cdpA-bla-neo, and emrD-bla-neo) as previously described, using a

SOE product of three PCRs (rpoEtn1_F1/ rpoEtn1_R1, rpoEtn1_F2/ rpoEtn1_R2, and rpoEtn1_F3/ rpoEtn1_R3). The SOE product was comprised up the rpoE gene and the region of gDNA upstream of the rpoE promoter, separated by part of mTn10 (inverted repeat, Ptac, lacR, and aad9. The mTn10 section was inserted between bases 2649857 and

2649858 on chromosome I. Transformants were selected on Sp100, as the SpR gene from the mTn10 was maintained in the SOE product.

The PepA G482D point mutant was created using natural co-transformation of V. cholerae (Dalia, et al. 2014). The SOE product was the product of two PCRS (pepA pt mut F1/pepA 482D R1 and pepA 482D F2/pepA pt mut R2). The 482D mutation was created by swapping the GGT codon for GAT at G482, and this change was centered in the pepA 482D R1 and pepA 482D F2 primers. Wild type cells were transformed with the SOE product (4µg) and pBAD33kan which was propagated in and isolated from E.

252 coli TG1 cells (900 ng). Transformants were isolated on Kn75 plates and screened for the point mutation using discriminatory PCR with isolated gDNA and the primers: pepA

482D check F and pepA 482D check R. The ultimate base (3’ end) of the pepA 482D check F primer matched the point mutation (A), and the penultimate base matched neither the wild type nor the point mutation (C). For this discriminating PCR, the following program was used with Taq polymerase (Phoenix Laboratory, Tufts University): 95°C –

5 min; and 18 cycles of 95°C – 30 sec, 55°C – 30 sec, and 72°C – 2 min. Out of 12 colonies screen, 2 were correct.

Deletion of phoB, in either the wildtype or ΔphoU background, was completed using SOE products ligated into the pCVD442lac allelic exchange vector. The map and sequence of pCVD442lac, which was derived from pCVD442, is available in

Supplementary Figure 3. The ΔphoB SOE product was constructed with the following primer pairs: phoBF1/ phoBR1 and phoBF2/ phoBR2. Allelic exchange was performed as described above.

The complementation plasmid pMMB-phoB was constructed in the pMMB67EH vector. V. cholerae phoB was amplified from E7946 genomic DNA using the primers phoB-pMMB-EcoRI-F and phoB-pMMB-BamHI-R. The PCR fragment was digested with EcoRI and BamHI restriction enzymes and then ligated into pMMB67EH that had been similarly digested. pMMB-phoB and pMMB67EH were cloned in E. coli DH5α, and transferred into the conjugation donor strain E. coli MFDpir (Ferrieres et al., 2010).

The plasmids were moved into V. cholerae using filter mating as described above.

253

The mTn10 delivery plasmid pDL1098, was constructed for this study using standard molecular biology methods. A map for pDL1098 is shown in Figure II-6. Use of this plasmid to create the transposon library is described in below.

VI.1.3. MIC experiments on plasmid insertion deletions The MIC of Cb for wild type, ΔompU, cysB::pGP704, and rseA::pGP704 was calculated (Table II-1). Individual colonies (3 biological replicates) were used to inoculate LB cultures supplemented with either 100 μg/mL (wild type and ΔompU) or

100 μg/mL and 2 μg/mL Cm. Cultures were grown at 37°C with aeration until the optical density of each culture reached 0.27 – 0.35. When the proper OD600 was reached, cultures were plated on LB plates (100 μL of 10-5 and 10-6 dilutions) or LB plates supplemented with either 5, 10, or 15 μg/mL Cm (100 μL of 10-3 and 10-5 dilutions). After overnight growth at 37°C, MIC plates were scored. The MIC was defined as the concentration of

Cb that resulted in a 100-fold knock down of growth when compared with the no drug control.

VI.1.4. Transposon mutagenesis The mTn10 delivery vector pDL1098 is temperature-sensitive for replication, but requires high temperature for transposition. At the permissive temperature of 30°C the plasmid replicates and the transposase is repressed. At the non-permissive temperature of

40°C the transposase is induced and replication stops, resulting in loss of the vector in the multiplying population of bacteria. For mTn10 mutagenesis of the V. cholerae xds-bla- neo strain, pDL1098 was transferred into the strain via filter mating with an E. coli

MFDpir mating strain (Ferrieres, et al. 2010). Exconjugates were selected on LB agar supplemented with 2.5 µg ml-1 Cm and 50 µg ml-1 Sp and grown at 30°C. Two

254 independent transposon libraries were made from single colonies by inoculating each into a 2 ml LB culture supplemented with 2.5 µg ml-1 Cm and 50 µg ml-1 Sp. Cultures were grown overnight at 30°C with aeration. In the morning, 0.5 ml of the turbid culture was passaged into 50 ml of LB media supplemented with 100 µg ml-1 Sp and pre-warmed to

40°C. After overnight, aerated growth at 40°C, 100 µl of the culture was further passaged into another 100 ml of pre-warmed LB supplemented with 100 µg ml-1 Sp. A second passage at 40°C in the presence of Sp was found necessary to completely eliminate bacteria harboring the transposon delivery plasmid. The final libraries were frozen in aliquots at -80°C in 20% glycerol.

VI.1.5. Determination of selection conditions We performed a series of minimal inhibitory concentration (MIC) experiments to determine concentrations of Cb + Kn for our selection. In these experiments, we varied

Cb and Kn concentrations, as well as the number of CFU plated. We chose antibiotic and cell density combinations as close to the MIC as possible and that fulfilled the following criteria: produced little to no background growth, allowed for development of discrete colonies, and exhibited reproducible degrees of killing. In conditions where the concentration of at least one of the antibiotics was too close to the MIC, the selection was not consistent from plate to plate in terms of both background growth and percent kill.

Additionally, the freshness of the antibiotics and dryness of the selection plates contributed greatly to experiment-to-experiment variability.

A total of 6 selection conditions for emrD and xds were chosen in order to increase the robustness of our selections: 16 µg/mL Cb + 40 µg/mL Kn, 18 µg/mL Cb +

45 µg/mL Kn, and 20 µg/mL Cb + 50 µg/mL Kn, all +/- 1 mM IPTG. For cdpA we used 5

255 selection conditions: 10 µg/mL Cb + 80 µg/mL Kn, 12 µg/mL Cb + 65 µg/mL Kn, 12

µg/mL Cb + 80 µg/mL Kn, 10 µg/mL Cb + 50 µg/mL Kn + IPTG, and 12 µg/mL Cb + 50

µg/mL Kn + IPTG. When 106 of the un-transposed bla-neo fusion strains were plated on these antibiotic conditions, no colonies grew. However, when 106 transposon mutagenized bla-neo funsions were plated on these plates, an average of 650 colonies grew (Table II-4). The inability of the un-transposed bla-neo fusion strains to grow on our selection plates suggested that spontaneous mutations, which could contribute to background noise in our selection, would not be a significant problem. The fold-killing was approximately the same for each antibiotic concentration. Additionally, we found all three conditions selected for the same transposon insertions, and have thus treated the conditions as biological replicates in our analysis.

VI.1.6. High-throughput sequencing and genetic selection The complexity of each transposon library was determined by Tn-seq using the

HTM-PCR method of preparing the samples for sequencing as described (Klein, et al.

2012, Lazinski and Camilli 2013), except that nested mTn10-specific primers, OLJ363 and OLJ385, were used. Final PCR products were cleaned using QIAquick PCR purification kit (Qiagen). The sample was sequenced on the Illumina HiSeq 2500 using the mTn10-specific sequencing primer OLJ386 and read from a single-end for 50 nucleotides. Primers used for barcoding samples are listed in primer table.

For the genetic selection, one aliquot of each transposon library was thawed and mixed in order to increase the overall complexity of the final library. The pooled library was then plated on LB agar plates supplemented with Cb and Kn at predetermined concentrations. All plates used in the genetic selection were made two days prior to usage

256 with freshly made aliquots of antibiotics and IPTG, when added. The plates were dried in a hood for 20 min and left until usage at room temperature. Approximately 1.5 x 106 CFU were plated on an average of 10 plates for each selection condition. After overnight growth at 37°C, the colonies from each selection condition were pooled and genomic

DNA was isolated from approximately 5 x 109 CFU using a DNeasy Blood and Tissue

Kit (Qiagen). Genomic DNA for the input sample was isolated directly from freezer stocks of the transposon libraries. Genomic DNA samples were subject to Tn-seq as above, to determine the relative frequencies of each transposon insertion in the input and post-selection (output) populations.

VI.1.7. Data analysis Sequence reads of transposon junctions were analyzed using the TUCF Galaxy

Server (Blankenberg, et al. 2010) (Giardine, et al. 2005) (Goecks, et al. 2010). Sequence reads were trimmed of 3’ poly-C-tails as described (Klein, et al. 2012). Low quality reads were discarded using the ‘Filter by quality’ script, which assigns a FASTQ quality score to each base in a single read. We retained any reads in which ≥95% of the bases had a quality score of >7. Reads were mapped to the V. cholerae N16961 genome (GenBank

Accession No. NC_002505 and NC_002506) using Bowtie. The default Bowtie parameters were used with two modifications: we required Bowtie to guarantee that each read was the best in terms of quality and stratum and we allowed Bowtie to backtrack 800 times when attempting to align a read. The TUCF custom scripts, ‘hopcount’ and

‘aggregate hop table’, were run on the mapped reads. Hopcount tallies the number of sequencing reads at every insertion site, which was used to determine the complexity of the input library, as well as the severity of the genetic selection (i.e. we learned how

257 many transposon insertions are represented in the output as compared with the input).

Hopcount also provides directionality of the transposon and its outward reading IPTG- inducible Ptac promoter. Aggregate hop table totals the number of sequencing reads in each gene. Included in the aggregate hop table output is a standardized frequency of transposon insertions for every gene (standardized read frequency), corrected for the gene size and calculated by the equation: standardized read frequencey = ((number of reads in gene X/ total number of reads in Illumina library)/ (size of gene X/ total genome size)).

Enriched genes in the IPTG negative selections, which represent putative repressors of cdpA, emrD, and xds, were identified by two methods. We calculated fold enrichment for every gene (FEgene), where FEgene = standardized read frequency in output/ standardized read frequency in input. Genes were ordered genes by their FEgene score and the average FEgene of the three antibiotic conditions was used to calculate the reported

FEgene scores. We also used the Wilcoxon signed-rank test as calculated by the JMP statistics software package (Version 10, SAS Institute Inc., Cary, NC) to determine whether a gene was statistically significantly enriched. We used this test to compare paired (by insertion site) Hopcount values of input and output for every gene. For this test, output Hopcount values were normalized to the input by multiplying the number of reads at each insertion site in the output by the ratio: total reads in the input/ total reads in the output. For genes with fewer than 4 insertion sites, a p value was not determined.

Statistically significant p values were considered less than 1.18 x 10-5 according to the

Bonferonni correction. The p values reported in Tables II-5, II-7, and II-9 are averages of the three antibiotic selection conditions.

258

The +IPTG selection data was analyzed using the same methods as for the -IPTG data in order to confirm putative repressors. However, we performed additional analyses at the transposon insertion level, as opposed to the gene level, in an attempt to identify putative activators of cdpA, emrD, and xds. Fold enrichment of individual transposon insertions (FEinsertion) was calculated for both genic and intergenic insertions and insertions were sorted in rank order of FEinsertion. To calculate FEinsertion, we first normalized the output Hopcount values to the input as described above. FEinsertion was calculated as: normalized output reads/ input reads at every insertion site. Any insertions present in genes identified in the selection without IPTG were discarded from the analysis. Additionally, insertions directly upstream of the bla-neo resistance markers were discarded. Enriched insertions were considered of interest if they fulfilled the following requirements: 1) exhibiting greater than 5-fold enrichment, 2) at least two insertions in the same intergenic/ genic region were enriched, and 3) enriched insertions in the same region exhibit strand bias (i.e., orientation of the Ptac promoter) in the output, but not the input.

VI.1.8. RNA purification and qRT-PCR All RNA was isolated from bacteria grown in broth unless otherwise stated otherwise. LB broth grown bacteria were grown to OD600 ~ 0.3 and aliquots were mixed

1:2 with RNAprotect Bacteria Reagent (Qiagen). For RNA isolation from bacteria grown in MOPS media, bacteria were grown overnight and 0.25 to 1 ml of culture was mixed with two volumes of RNAprotect. For RNA isolation from plate grown bacteria, fresh colonies from semi-confluent plates were scraped up in LB and adjusted to an OD600= 10, and frozen in 25% glycerol at -80°C in 100 µl aliquots of ~109 CFU. Frozen aliquots

259 were thawed and plated on LB plates at a density of ~109 CFU per plate and incubated at

37°C for two hours. Bacteria were removed from the plates by resuspending in LB and diluted in two volumes of RNAprotect. If RNA was not isolated the same day, samples were centrifuged at room temperature and the pellets were stored at -80°C. RNA was isolated from the cells using the RNeasy Mini Kit (Qiagen).

Any contaminating DNA in the RNA prep was digested using the TURBO DNA- free kit (Ambion, Life Technologies). cDNA was synthesized from 0.25 to 1 µg of DNA- free RNA using an iScript cDNA Synthesis Kit (Bio-Rad). DNA contamination was assessed through the use of control samples lacking reverse transcriptase. cDNA was diluted 5-fold in DEPC treated water and 5 µl was used as a template for qRT-PCR.

SYBR green qRT-PCR was conducted as described previously (Pratt, et al. 2009). Refer to Table S3 for the primers used in these experiments. All Ct values were corrected for primer efficiency as determined by a standard curve using genomic DNA template.

Transcript levels were normalized to the housekeeping gene, rpoB. Transcript levels were reported either as relative to rpoB or as fold change over wild type, this is specified in the figure legends.

VI.1.9. Purification of PhoBCA PhoBCA was purified as previously described (Pratt, et al. 2010). E. coli BL21

DE3 pET-GST-PhoB was grown overnight in LB + Ap at 37°C. In the morning, the culture was back-diluted to an OD600 = 0.01 in 1 L of LB + Ap and grown at 37°C with aeration until the culture reached OD600 ~0.7, at which point 1 mL of 1M IPTG was added to the culture to induce expression of GST-PhoB. Flasks were incubated at 20°C for 16 hr with aeration. After protein production, the 1 L of culture was pelleted at 4,500

260 rpm for 20 min at 4°C. Pellets were resuspended in lysis buffer (20 mM Tris HCl pH 8.0,

150 mM NaCl, ½ tablet of Roche protease inhibitor cocktail). After resuspension, β- mercaptoethanol was added to the cells at 5 mM and the preparation was frozen at -80°C until they were used for purification of GST-PhoB.

Throughout all of the purification steps, the sample was kept on ice or work was done in the 4°C room, unless otherwise stated. For purification, cells were thawed and lysed via sonication (50% amplitude, ½ sec on and ½ sec off, 15 sec, 4-5 cycles). Lysed cells were clarified by centrifugation in an SS34 rotor at 18,000 rpm for 45 min at 4°C.

The supernatant was incubated with Glutathione Sepharose 4B beads (GE Healthcare

Life Sciences) in a gravity flow column for 20 min with gentle rolling. The bound protein was washed five times with GST-wash buffer (20 mM Tris HCl pH 8.0, 150 mM NaCl, 1 mM DTT) to remove any unbound contaminating proteins. For elution, buffer was added to the beads (100 mM Tris HCl pH 8.0, 20 mM reduced glutathione, 150 mM NaCl, 1 mM DTT), gently mixed, and then the beads were allowed to site for 15 min before the liquid was removed from the column. Three elution samples were collected; elution 1 and

2 were combined.

The protein was further purified with anion exchange using the AKTA S15Q 2 mL column. The sample eluted from the glutathione beads was diluted 1 in 10 in a no salt buffer, QB1A (20 mM Tris HCl pH 8.0, 1 mM DTT). After equilibration of the S15Q column with QB1A, the sample was applied to the column. The protein was eluted from the column using a 0-30% gradient of QB1B buffer (20 mM Tris HCl pH 8.0, 1 M NaCl,

1 mM DTT), which was allowed to develop over 30 column volumes. The peak fraction was incubated overnight at 4°C with TEV protease (purified from E. coli by Ayman

261

Ismail). The next day the cleaved protein was incubated with Glutathione Sepharose 4B beads as before and the flow-through, which contained the untagged PhoBCA protein, was collected. Buffer exchange was performed using the AKTA and a Superose12 gel filtration column in order to transfer the purified protein into EMSA buffer (10 mM Tris

HCl pH 8.0, 100 mM KCl, 5% glycerol, 1 mM DTT).

All steps of the purification, as well as the final protein purity, were monitored using 15% SDS-PAGE and Coomassie staining.

VI.1.10. Electromobility shift assays EMSA probes were either end-labeled with Cy-3 (tcpPH probe) or continuously labeled (the rest) by incorporating dCTP-Cy-5 into the PCR reaction. PCR primers are listed in the primer table. The labeled nucleotide was used at 1/20 the concentration of the unlabeled dCTP such that 1/20 cytidine residues would be labeled within each probe. The high fidelity DNA polymerase, Phusion (Thermo Scientific/ Finnzymes), was used for the PCR reactions to avoid sequence errors in the EMSA probes. The PCR reactions were cleaned by extraction from a 2% agarose gel and the Qiagen Gel Extraction Kit. Finally, probes were run through a Performa Spin Column (EdgeBio) and diluted in dH20 to 50 nM.

For EMSA, the protein was diluted to the desired concentrations in a total volume of 15 µL of EMSA buffer. The binding reaction was started by adding the reaction mix: 2

µL probe PCR (5 nM final), 1.5 µL sheared calf thymus DNA (50 µg/mL final), 1 µL 5x

EMSA buffer, and 0.5 µL bovine serum albumin (BSA) (150 µg/mL final). Reactions were incubated for 30 min in the dark and subsequently run on 6% native polyacrylamide gel and run at 150 min for 35-45 min depending on the length of the probe. The gel box

262 was kept in the dark. The fluorescent bands were visualized using the Fujifilm FLA-9000 imagine scanner.

VI.2 CHAPTER III

VI.2.1. Strains For RIVET assays, the resolvable kanamycin-resistance marked cassette (res-neo- sacB-res) (Osorio et al., 2005) was inserted into the lacZ gene of the wild type and

ΔphoB strains via natural transformation as described (Blokesch 2012; De Souza Silva and Blockesch, 2010). The transcriptional fusion xds-tnpR, carried on the pIVET-

VC2621 plasmid (Schild et al., 2007), was moved into both res cassette-containing strains via conjugation with the E. coli SM10λpir donor. Exconjugates were selected on

LB agar supplemented with Sm, Kn, and Ap at 30°C.

VI.2.2. Recombination-based in vivo expression technology (RIVET) Resolution assays were performed as described except that Kn was used in place of tetracycline (Pratt, et al. 2009). Additionally, for in vivo resolution analysis, 5-day-old

CD-1 mice were intragastrically inoculated with plate grown bacteria, which were resuspended in LB. A single mouse received either 2 x 104 xds-tnpR lacZ::res-neo-sacB-

600 7 res CFU (OD = 0.001) or 2 x 10 ΔphoB xds-tnpR lacZ::res-neo-sacB-res CFU (OD600

= 0.5). The higher dose for the latter strain was necessary due to its attenuated colonization phenotype. For the timecourse RIVET experiment, xds-tnpR lacZ::res-neo- sacB-res was grown to mid-exponential phase in LB broth (OD600 = 0.3), washed twice in

MOPS minimal medium (no phosphate added), and resuspended to an OD600 = 0.001.

Mice were intragastically inoculated with 50 µl of the prepared inoculum. Mice were euthanized in groups of 4 or 5 at hours post inoculation: 4, 8, 12, 16, 20 and 24.

263

Resolution was assessed as described above. For in vitro control, four independent inoculums were prepared the same way as described above. Resolution was measured in half hour increments for four hours.

VI.2.3. Evaluation of xds in vivo fitness Bacterial inocula were prepared as described for the RIVET timecourse experiment except that wild type and Δxds strains were resuspened in MOPS containing either 0 mM KH2PO4 or 10 mM KH2PO4 to an OD600 = 0.005. Five-day-old CD-1 mice were intragastrically inoculated with 50 µL of inocula. Eight hours post inoculation, the bacterial load in the infant mouse small intestine was assessed by homogenizing the small intestine in 1 mL of LB and plating dilutions on LB + Sm plates. Each inoculum was tested in at least 8 mice using single strain infections.

VI.3. CHAPTER IV

VI.3.1. Strain construction The single mutants – ΔushA, ΔnupC, and ΔpilA – were constructed using the

Flp/FRT method as described in VI.1.2. The mutants – ΔushAΔcpdB and ΔushAΔpilA were constructed using Flp/FRT and ΔushA as the parental strain. The triple mutants –

ΔxdsΔushAΔdns and ΔdnsΔushAΔcpdB were constructed using two rounds of Flp/FRT and either Δxds or Δdns as the parental strain.

Natural co-transformation was used to make the mutants: ΔcpdB, ΔyjjG,

ΔyjjGΔVCA0545, and ΔcpdBΔyjjGΔVCA0545. The selectable marker used in this construction was pBAD33kan isolated from E. coli TG1 cells. The PCR constructs used for transformation resulted in the exchange of the desired open reading frame for a FRT scar (GAAGCAGCTCCAGCCTACA), leaving only the start and stop codon of the

264 deleted gene. Tranformants were selected on LB plates with 75 µg/mL Kn and subsequently screen by MASC PCR in order to the genotype of each transformant at all loci of interest (Wang and Church 2011a).

The ushA-FLAG strain was made using natural co-transformation with the pBAD33kan selectable marker. The FLAG tag

(GACTACAAAGACGATGACGACAAG) was incorporated by PCR directly before the stop codon of ushA. Correct insertion of the FLAG-tag was confirmed by PCR and sequencing.

VI.3.2. Phosphatase screen The mTn10 library was constructed in ΔphoX using pDL1098 as described in

VI.1.4. Aliquots of the library were thawed, diluted to a 10-5 dilution in LB, and 225 µL were plated on each 150 mm LB plate (supplemented with 100 µg/mL Sp and 40 µg/mL

XP). Plates were incubated overnight at 37°C and white colonies were identified (7/

~40,000) and colony purified on the same medium. Genomic DNA was prepped of each white colony and arbitrary PCR followed by sequencing of the PCR product was used to determine the location of each Tn insertion (Hava and Camilli 2002). Briefly, two rounds of PCR were performed using the primers 1) Arb1/olj363 and 2) Arb2/olj386. For PCR 1 the following program was used: 95°C – 5 min, followed by 6 rounds of 95°C – 30 sec,

30°C – 30 sec, and 72°C – 1min; followed by 30 rounds of 95°C – 30 sec, 45°C – 30 sec, and 72°C – 1 min. For PCR 2 the following program was used: 95°C – 5 min, followed by 35 rounds of 95°C – 30 sec, 55°C – 30 sec, and 72°C – 1min. PCR products were cleaned and sent for sequencing using the primer olj386.

265

VI.3.3. Phosphate growth curves Strains were struck on LB Sm plates and grown overnight at 37°C. Single colonies were used to inoculate 2 mL cultures of MOPS-glucose medium supplemented with 10 mM KH2PO4 (described in VI.1.1), and grown overnight at 37°C with aeration.

In the morning, strains were back diluted to an OD600 ~0.05 and grown to mid- exponential phase in the same medium. Bacteria were washed twice in MOPS-glucose medium (no phosphate) and inoculated in to the growth curve test cultures.

For eDNA curves, strains were inoculated into 2 mL cultures of MOPS-glucose medium supplemented with the desired source of Pi (either KH2PO4, DNA, or nothing).

The source of DNA used was sheared salmon sperm DNA (Life Technologies). Culture tubes were grown at 37°C with aeration and OD600 readings were taken through the glass tube after blanking the spectrophotometer with the appropriate medium. New glass tubes were always used to avoid misreading of the optical density due to scratches in the glass, as well as to avoid Pi contamination from the phosphoric acid used to wash the dishes.

For nucleotide curves, strains were inoculated into 200 µL MOPS-glucose medium supplemented with the desired source of Pi (either KH2PO4, nucleotides, or nothing). Nucleotides were added at a concentration of 0.1 mM total Pi, that is, if two nucleotides were added, each was used at 0.05 mM. Additionally, when NTPs were used, the nucleotide was added to a final concentration of 0.033 such that the final Pi concentration was 0.1 mM (three Pi per nucleotide). The cultures were grown in 96-well plates at 37°C with aeration using the BioTek Synergy Plate Reader. Optical density readings were measured and recorded every 15 min using the Gen5 Data Analysis

266 software. All nucleotides, except dNTP and NTP, were obtained in power or salt form and resuspended in dH2O to a concentration of 0.1 M.

VI.3.4. Nitrogen growth curves For these growth curves a modified M9-glucose medium was used. First 5x salts were made in 1 L dH2O (33.9g Na2HPO4, 15g KH2PO4, 2.5g NaCl). Next 2x medium was made by diluting 5x salts with dH20 and 1M MgCl2 (final concentration of 2 mM) and 1 M CaCl2 (final concentration of 0.2 mM).

Single colonies were used to inoculate 2 mL cultures of modified M9-glucose medium supplemented with 0.5 mM NH4Cl and grown overnight at 37°C with aeration.

Wild type and mutants grew to an OD600 of 0.3 in this medium. In the morning cultures were back-diluted to 0.02 in 5 mL cultures of the same medium and grown at 37°C with aeration until the cells reached mid-exponential phase (OD600 ~0.15 in this medium).

Bacteria were washed twice in modified M9-glucose medium (no nitrogen source) and inoculated into 200 µL total volume modified M9-glucose cultures that were supplemented with the desired nitrogen source. The cultures were grown in 96-well plates at 37°C with aeration using the BioTek Synergy Plate Reader. Optical density readings were measured and recorded every 15 min using the Gen5 Data Analysis software.

VI.3.5. Nucleotidase assays Single colonies were used to inoculate 10 mL LB cultures were grown at 37°C with aeration to an optical density of ~0.5. For each replicate, the equivalent of 10 mL of

OD600 =0.5 culture was centrifuged in a 15 mL conical tube at 4,500 x g for 20 min at room temperature. The supernatants were removed and the cultures were resuspended in

267

1 mL 10 mM Tris HCl pH 7.5. The cultures were spun again and the pellets were resuspended in 0.5 mL 10 mM Tris HCl pH 7.5. At this point cells were either directly assayed or lysed. Cell lysis was achieved by transferring the cells to 2 mL eppendorf tubes and sonicating for 1 min with 50% amplitude, ½ sec on and ½ sec off. Cell lysates were clarified by spinning the tubes at 8,000 x g for 10 min at 4°C.

Nucleotidase assays were performed as described previously with a few modifications (Edwards, et al. 1993). Accordingly, 100 µL of the cells (either lysed or unlysed) were mixed with 890 µL of assay buffer (150 µL 0.5 M Sodium Acetate pH 6.0,

30 µL 150 mM CoCl2, 30 µL 480 mM CaCl2, and water to a final volume of 890 µL) and equilibrated to 37°C for 5 min. The nucleotidase reaction was started by the addition of

10 µL 100 mM nucleotide substrate and samples were immediately mixed by vortexing and placed at 37°C. At times 0, 5, 10, and 15 min after addition of the substrate, 150 µL samples of each reaction were removed and transferred to eppendorf tubes with 100 µL

0.1 N HCl and placed on ice to stop the enzymatic reaction. Once all samples were acquired, the tubes were centrifuged at 16,000 x g for 5 min at 4°C to spin out the cell debris. Subsequently, 60 µL of the supernatant was mixed with 140 µL of the development reagent (1 part 10% ascorbic acid, 6 parts 0.42% Ammonium molybdate in

1 N H2SO4). Samples were incubated at 45°C for 20 min, after which 150 µL of each sample was transferred to a 96-well plate and the absorbance at 820 nm was measured using a BioTek Synergy Plate Reader.

To convert the absorbance readings to pmoles of Pi released, a standard curve was performed. Five-fold dilutions of KH2PO4 corresponding to 1000, 200, 40, 8, 1.6, or 0 pmoles of Pi were mixed with the assay buffer and then mixed with the development

268 reagent and incubated as described above. After the absorbance was measured, the readings were plotted against starting concentration and the slope corresponded to the conversion factor (i.e. absorbance readings in subsequent assays were converted to pmoles released by the slope). The standard curve was performed twice, in duplicate each time.

VI.3.6. Infant mouse competition assay

Bacterial inocula were grown to mid-exponential phase in LB broth (OD600 = 0.3), washed twice in MOPS minimal medium (no phosphate added), and resuspended to an

OD600 = 0.025. Wild type (ΔlacZ ) and ΔushA were mixed 1:1 and diluted 1:10, such that the final OD600=0.005. Seven, 5-day-old CD-1 mice were intragastically inoculated with

50 µl of the prepared inoculum and the remaining input was plated on LB Sm plates supplemented with X-gal in order to determine the input ratio of ΔlacZ:ΔushA. After 24 hours, mice were euthanized and the output ratio of ΔlacZ:ΔushA in the infant mouse small intestine was assessed by homogenizing the small intestine in 1 mL of LB and plating dilutions on LB Sm X-gal plates.

VI.3.7. RNA purification and qRT-PCR For MOPS grown cultures, bacteria were grown to mid-exponential phase in

MOPS-glucose medium supplemented with 10 mM KH2PO4, washed twice in no phosphate MOPS, and resuspended in MOPS-glucose medium supplemented with the desired source of Pi, if any. Bacteria were incubated in the test medium for the amount of time described in each figure legend after which aliquots were removed and mixed 1:2 with RNAprotect. For LB grown cultures, bacteria were grown to mid-exponential phase in LB (OD600 ~0.3) and the mixed 1:2 with RNAprotect.

269

RNA was isolated, converted to cDNA, and analyzed by qRT-PCR using the methods described in VI.1.8. Expression levels are shown relative to rpoB, but not normalized to wild type expression.

270

VI. 4. STRAINS AND PRIMERS Primers and strains that were used in multiple chapters are only listed in Chapter

II tables.

VI.4.1. Chapter II

Table VI-1. Strains used in Chapter II. Strain Genotype or Phenotype Reference E. coli DH5α F- Δ(lacZYA-argF) U169 recA1 endA1 hsdR17 supE44 thi-1 Laboratory strain gyrA96 relA1 DH5αλpir F- Δ(lacZYA-argF)U169 recA1 endA1 hsdR17 supE44 thi-1 Laboratory strain gyrA96 relA1 λ::pir SM10λpir thi recA thr leu tonA lacY supE RP4-2-Tc::Mu λ::pir, KnR Laboratory strain MFDpir MG1655 RP4-2-Tc::[ΔMu1::aac(3)IV-ΔaphA-Δnic35- (Ferrieres et al., 2010) ΔMu2::zeo] ΔdapA::(erm-pir) ΔrecA TG1 pBAD33kan F' [traD36 proAB+ lacIq lacZΔM15] supE thi-1 Δ(lac-proAB) Laboratory strain Δ(mcrB-hsdSM) 5, (rk-, mk-), carrying pBAD33kan BL21 DE3 pET-GST-PhoB F- ompT hsdSB(rB-, mB-) gal dcm (DE3), expressing GST- (Pratt et al., 2010) PhoB from an IPTG inducible promoter. V. cholerae Wild type E7946 El tor Ogawa, HapR+, ApR Laboratory strain cdpA-bla Transcriptional fusion of cdpA to bla This study emrD-bla Transcriptional fusion of emrD to bla This study xds-bla Transcriptional fusion of xds to bla This study ΔompU In frame deletion of ompU Laboratory strain cysB::pGP704 Plasmid insertion knock-out of cysB, ApR This study rseA::pGP704 Plasmid insertion knock-out of rseA, ApR This study cdpA-bla-neo Transcriptional fusion of cdpA to bla and neo This study emrD-bla-neo Transcriptional fusion of emrD to bla and neo This study xds-bla-neo Transcriptional fusion of xds to bla and neo This study; (McDonough et al., 2014) ΔfexA In frame deletion of fexA This study ΔfexB In frame deletion of fexB This study ΔubiC In frame deletion of ubiC This study ΔpdhR In frame deletion of pdhR This study ΔVC2288 In frame deletion of VC2288 This study ΔVCA0762 In frame deletion of VCA0762 Laboratory strain rpoE::mTn10 Remake of mTn10 insertion at bp 2649857 on chromosome I. This study Tn inverted repeat, Ptac, lacI, and aad9 were inserted. cdpA-bla-neo rpoE::mTn10 Remake of mTn10 insertion at bp 2649857 on chromosome I in This study the cdpA-bla-neo fusion emrD-bla-neo rpoE::mTn10 Remake of mTn10 insertion at bp 2649857 on chromosome I in This study the emrD-bla-neo fusion ΔVC0469 In frame deletion of VC0469 This study gshB::pGP704 Plasmid insertion knock-out of gshB, ApR This study ΔpepA In frame deletion of pepA This study PepA G482D PepA point mutation, confers loss of aminopeptidase activity in This study; (Charlier et E. coli al., 2000)

271

ΔphoU In frame deletion of phoU This study; (McDonough et al., 2014) ΔphoUΔphoB In frame deletion of phoU and phoB This study; (McDonough et al., 2014) ΔphoB In frame deletion of phoB This study; (McDonough et al., 2014) ΔphoB (pMMB) In frame deletion of phoB carrying pMMB, ApR This study; (McDonough et al., 2014) ΔphoB (pMMB-phoB) In frame deletion of phoB carrying pMMB-phoB, ApR This study; (McDonough et al., 2014) PhoBCA PhoB D10A/D53E constitutively active mutant in place of the (Pratt et al., 2010) wild type locus Plasmids q R pMMB pMMB67EH IncQ lacI bla (Ap ) Ptac rrnB (Furste et al., 1986) pMMB-phoB pMMB67EH with the phoB ORF cloned into EcoRI and BamHI This study; (McDonough restriction sites et al., 2014) pDL1098 Temperature-sensitive mTn10 delivery vector, Fig. S1, CmR, This study; (McDonough SpR et al., 2014) pDL1098-flp Temperature-sensitive Flp recombinase delivery vector, Fig. S2, This study; (McDonough CmR et al., 2014) pCVD442lac Allelic exchange vector, Fig. S3, ApR Laboratory plasmid pCVD442catlac Allelic exchange vector, Fig. S4, CmR Laboratory plasmid pBAD33kan Laboratory plasmid pET-GST-PhoB (Pratt et al., 2010) pGP704-cat Cat marker replacing the bla marker from pGP704 Laboratory plasmid; (Miller and Mekalanos, 1988)

272

Table VI-2. Primers used in Chapter II. Primer use Primer name Sequence (5' to 3' orientation) Strain construction cdpA-bla cdpA-bla-F1 CGCTACGCGTTTAACAACAA cdpA-bla-R1 AAAAATGTATCCTAAATTAACCCAAGCGTGAAGG cdpA-bla-F2 GCATTGGTAACGCAAGATAGACGCGTCA cdpA-bla-R2 AAAACCCGATCCGTTTAGGT universal-bla-F3 TTTAGGATACATTTTTATGAGTATTCAACATTTCCGT cdpA-bla-R3 CTATCTTGCGTTACCAATGCTTAATCAGTGA emrD-bla emrD-bla-F1 TTGCGAGACAGCTATGATG emrD-bla-R1 AAAAATGTATCCTAAATTACATCGGTACTTGCGC emrD-bla-F2 GCATTGGTAAGTAAGCTGAGTTTACCGTC emrD-bla-R2 AGCCACTATCGCCACAAAAG universal-bla-F3 TTTAGGATACATTTTTATGAGTATTCAACATTTCCGT cdpA-bla-R3 CTCAGCTTACTTACCAATGCTTAATCAGTGA xds-bla xds-bla-F1 AGTCAGCAAGCACCAGAGGT xds-bla-R1 AAAAATGTATCCTAAACTAGCGACGGCGACGCCA xds-bla-F2 GCATTGGTAAAAAGTGGCATTTTTGATAAACC xds-bla-R2 CGGTGTGGATACCAAACACA universal-bla-F3 TTTAGGATACATTTTTATGAGTATTCAACATTTCCGT xds-bla-R3 ATGCCACTTTTTACCAATGCTTAATCAGTGA cysB::pGP704 cysBins F CATGGGTACCGGTGTCATGCCAACTC cysBins R CATGGAGCTCAAGCGGCAAGCATTTAACG rseA::pGP704 rseAins F CATGGGTACCGTAGTCCTGCAGCAAGG rseAins R CATGGAGCTCGACAATGAGCTGATTTTAGG cdpA-bla-neo bla-F1 ATGCGCATGCATGAGTATTCAACATTTCCGT bla-R1 AAAAATGTATCCTAAATTACCAATGCTTAATCAGTGA neo-VC0130-F2 GTTTTTCTAACGCAAGATAGACGCGTCA cdpA-bla-R2 AAAACCCGATCCGTTTAGGT neo-F3 TTTAGGATACATTTTTATGAGCCATATTCAACGG neo-VC0130-R3 CTATCTTGCGTTAGAAAAACTCATCGAGCATC emrD-bla-neo bla-F1 ATGCGCATGCATGAGTATTCAACATTTCCGT bla-R1 AAAAATGTATCCTAAATTACCAATGCTTAATCAGTGA neo-VCA0083-F2 GTTTTTCTAAGTAAGCTGAGTTTACCGTC emrD-bla-R2 AGCCACTATCGCCACAAAAG neo-F3 TTTAGGATACATTTTTATGAGCCATATTCAACGG neo-VCA0083-R3 CTCAGCTTACTTAGAAAAACTCATCGAGCATC xds-bla-neo bla-F1 ATGCGCATGCATGAGTATTCAACATTTCCGT bla-R1 AAAAATGTATCCTAAATTACCAATGCTTAATCAGTGA neo-VC2621-F2 GTTTTTCTAAAAAGTGGCATTTTTGATAAACC xds-bla-R2 CGGTGTGGATACCAAACACA neo-F3 TTTAGGATACATTTTTATGAGCCATATTCAACGG neo-VC2621-R3 ATGCCACTTTTTAGAAAAACTCATCGAGCATC FRT-cassette FRT-kan-F ATTCCGGGGATCCGTCGAC FRT-kan-R TGAACAACATGGACTGGCA ΔfexA fexA_FRT_F1 TTTAATACTCGCATCCCTTCCTC fexA_FRT_R1 GAAGCAGCTCCAGCCTACAGAAGATTAACCTCTTCTTTTATATCTAATTAGA fexA_FRT_F2 GTCGACGGATCCCCGGAATTTGCATTAGCGTTACCTAAACTTG fexA_FRT_R2 TTTGGCTGTGGTGGTACAAG fexA_FRT_F0 TCACTGCCACCAATTTATTGG fexA_FRT_R0 TTGATTTGTTAGTCAAATTGGGC

273

ΔfexB fexB_FRT_F1 GAAAACCTCACCACGTTACAGC fexB_FRT_R1 GTCGACGGATCCCCGGAATCATTCACTGCTCCGTCAGATAAA fexB_FRT_F2 GAAGCAGCTCCAGCCTACATGATAAAAAAAGACCCTAGCCA fexB_FRT_R2 CAAGGTGGAAAGTAAAAACC fexB_FRT_F0 CTCACTTGATTGTTGGGTTACCC fexB_FRT_R0 GGCTGAGTGGCATTGGCTAT ΔubiC ubiC_FRT_F1 AGGTTAAGCCAGACCCGCT ubiC_FRT_R1 GTCGACGGATCCCCGGAATCAATTTGACGCTTTACCCGA ubiC_FRT_F2 GAAGCAGCTCCAGCCTACATAAATGACCGCTGTAAAAGCGC ubiC_FRT_R2 CGAAAGGCTATTCGGTGGAA ubiC_FRT_F0 GCGTGACCGCTTCAAGTAGAA ubiC_FRT_R0 GGTGCGCCACTTTATTCAACG ΔpdhR pdhR_FRT_F1 ATCAGTTCACGTTTTAGCGTG pdhR_FRT_R1 GTCGACGGATCCCCGGAATAGCCATAATTTATTGTTCTTCTACTCT pdhR_FRT_F2 GAAGCAGCTCCAGCCTACACAATAGCGGCCATTGCGATT pdhR_FRT_R2 AAGCGAAGGTTGGATCGTAAG pdhR_FRT_F0 GCAACTCATGATTGATCATCTGG pdhR_FRT_R0 TTGCTGGGTGAGCGTAGCTT ΔVC2288 VC2288_FRT_F1 AGGTTAAGCCAGACCCGCT VC2288_FRT_R1 GTCGACGGATCCCCGGAATCAATTTGACGCTTTACCCGA VC2288_FRT_F2 GAAGCAGCTCCAGCCTACATAAATGACCGCTGTAAAAGCGC VC2288_FRT_R2 CGAAAGGCTATTCGGTGGAA VC2288_FRT_F0 GCGTGACCGCTTCAAGTAGAA VC2288_FRT_R0 GGTGCGCCACTTTATTCAACG rpoE::mTn10 rpoEtn1_F1 CCCAAGTGACTTGCCAGAAG rpoEtn1_R1 CAAAATCATTAGGGGATTCATCAGTGCCGAGTCTCTAAAGGTGC rpoEtn1_F2 ATTATAAGGGGCCTGCCACCTTATCGAGAAAGTTCATTATTGGGTG rpoEtn1_R2 GCAATTGTTAGTGCGATTAAAACG rpoEtn1_F3 GCACCTTTAGAGACTCGGCACTGATGAATCCCCTAATGATTTTG rpoEtn1_R3 CACCCAATAATGAACTTTCTCGATAAGGTGGCAGGCCCCTTATAAT rpoEtn1_F0 TGAACAACATGGACTGGCA rpoEtn1_R0 CACTGGCTCATCTTGGCTC ΔVC0469 VC0469_FRT_F1 CATTGAAGCCGTTCGCCAG VC0469_FRT_R1 GTCGACGGATCCCCGGAATCATGCGGATAATTCCTAATC VC0469_FRT_F2 GAAGCAGCTCCAGCCTACATAAATGGAGATAAACCATGATTAA VC0469_FRT_R2 AATATCCTTCGAGGTGGTGAC gshB::pGP704 gshBins F CATGGGTACCACGCGCCAAGCAGTAAG gshBins R CATGGAGCTCCAATTCAAATCTGAGCAAAC ΔpepA VC0469_FRT_F1 CCTGCATCACCACCAATTC VC0469_FRT_R1 GTCGACGGATCCCCGGAATCATGCGTACTCCTACATCC VC0469_FRT_F2 GAAGCAGCTCCAGCCTACATAATCGCAAAAGGGCCTTC VC0469_FRT_R2 GCTCAATTACCACGCCAC PepA G482D pepA pt mut F1 TTCCGGCTTAACCGTCACTACG pepA 482D R1 GGACTAGTAGCGAGACTGGACGATCGGTTGAGCCTTTCGCGGCACC pepA 482D F2 GGTGCCGCGAAAGGCTCAACCGATCGTCCAGTCTCGCTACTAGTCC pepA pt mut R2 GACCATTGAACTCTTTCGCAGCC pepA 482D check F AATGCCTCTGAGCAATCGAG pepA 482D check R TAGTAGCGAGACTGGACGGT ΔphoU phoU-FRT-F1 TCGGAATCACCACGGGAG phoU-FRT-R1 GTCGACGGATCCCCGGAATCATGGCATATCCTTACTAAAC phoU-FRT-F2 GAAGCAGCTCCAGCCTACATAAGCGGCGCTGTCCGTC phoU-FRT-R2 GCAAACTTACAAGGCGTGG ΔphoB phoBF1 GGTCTAGAGTCGTTTCGACCGCTATTTC

274

phoBR1 CCTCTTAGGCCCTTCTAGACATAATTGATTAACCTTTG phoBF2 GTCTAGAAGGGCCTAAGAGGGTGTAGAACAAATC phoBR2 GGTCTAGATCATCACACCATGGGCTTTA ΔphoB phoB-pMMB- GATCGAATTCATGTCTAGAAGGATTCTGGTTGTTG (pMMB- EcoRI-F phoB) phoB-pMMB- GAGCGAGTCGGATCCTTAGGCTTTGGTTGAAAAACGATAC BamHI-R PhoBCA PhoBCA F1 CAGCGTAAAGGATTAGCTCTG PhoBCA R1 CAACAACCAGAATCCTTCTAGACATAATTGATTAACCTTTGTTTTTCACC PhoBCA F2 GGTGAAAAACAAAGGTTAATCAATTATGTCTAGAAGGATTCTGGTTGTTG PhoBCA R2 ACCGTCTGGATCAGCTTATCATGTCCAGCGTCTTCTAAGG PhoBCA F3 CCTTAGAAGACGCTGGACATGATAAGCTGATCCAGACGGT PhoBCA R3 GACCCTTTACCACCATTGAC PhoBCA F01 GAAGGATTCTGGTTGTTGAAAC PhoBCA F02 CTGATCTCGTCCTGCTTGGA PhoBCA R0 GACGAATGTGGACGTCTACC EMSA probes bla bla F TGAAGCCATACCAAACGACG bla R GGAAGCTAGAGTAAGTAGTTCGCC phoBR phoBR probe F CAGTTTTTAAACCACATTGTTG phoBR probe R CCTTTGTTTTTCACCAACG tcpPH tcpPH probe F CAA CTG CAA AAT TAG ATT GCA AAT AA tcpPH probe R CAT TCG TTC CAC CAA AGG TTA TC, 5' Cy3 labeled xds probe 1 xds 1 F TCTTAAACGGCCTTAATGAGTAC xds 1 R ATAGTCAGAGCGGCGAATG xds probe 2 xds 2 F AATAAATCGACGCCGTGG xds 2 R TGTTACTGTTTTTCATGCAACAC xds probe 3 xds 3 F TCTATTGATGAGATGAACATCTCG xds 3 R GATGTACCTTCTCCTCCCTATTT xds probe 4 xds 4 F ACACAATACTGGTTCAATACACC xds 4 R GCGCCACTAAGCCAAAAC qPCR primers rpoB rpoBF CTGTCTCAAGCCGGTTACAA rpoBR TTTCTACCAGTGCAGAGATGC cdpA cdpA-qRT F CTTTATGTCGAGACGGGTGGTTAC cdpA-qRT R CGGCTCATCATCCAAACGTAAG emrd emrD-qRT F GATTAGCCGTGATTTACAGG emrD-qRT R GCCATAAAATAGCTGTGAGC xds xds-qRT F GTATTACAACCTGACGCAGCTC xds-qRT R GGAAGGTAAGATCTCGATGGCTTG rpoE rpoE qRT F ACACTTGGTTGTACCGTATTGCT rpoE qRT R AATTCAGCTTCTTCAGCATCAAC phoB PhoBqF AGGGCTATCAGGCGGTTGAG PhoBqR TACCACCAGGCAACATCCAG phoX VCA0033 qPCR F CGGTGTCACCATTGTTGAAG VCA0033 qPCR R TGATCCGACGATTACGTTCA acgA acgAqF CTCGGTTCAACATCTACAAGCAC acgAqR ATCAGGCTGTAATTCAGACCAC acgB acgBqF GCAAACAGATCCGAGTGCTATG acgBqR CTGAATGTGATTACCGTGAACC VC0353 vc0353qF CGCAAGATCCACAAGCTAACC vc0353qR TCCAAGCGAGATCCCAAACTG VCA0955 vca0955qF GTCATGGAATGGGCAAATGAGC

275

vca0955qR GCTTGTTGATAGAAGGCCTCC Illumina sequencing primers HTM-PCR OLJ363 GTGTGGGCACTCGACATATGACAAG OLJ376 GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCTGGGGGGGGGGGGGGGG OLJ385 AATGATACGGCGACCACCGAGATCTACACTCTTTGGGGGCCAAAATCATTAG GGGATTCATCAG Input BC37 CAAGCAGAAGACGGCATACGAGATCACTGTGTGACTGGAGTTCAGACGTGT Barcode* GCTCTTCCGATCT Cm16 Kn40 OLJ533 CAAGCAGAAGACGGCATACGAGATCGCGCGGTGACTGGAGTTCAGACGTGT Barcode* GCTCTTCCGATCT Cm16 Kn40 OLJ536 CAAGCAGAAGACGGCATACGAGATGCGCGCGTGACTGGAGTTCAGACGTGT +IPTG GCTCTTCCGATCT Barcode* Cm18 Kn45 OLJ534 CAAGCAGAAGACGGCATACGAGATCTCTCTGTGACTGGAGTTCAGACGTGTG Barcode* CTCTTCCGATCT Cm18 OLJ537 CAAGCAGAAGACGGCATACGAGATGGGGGGGTGACTGGAGTTCAGACGTGT Kn45+IPTG GCTCTTCCGATCT Barcode* Cm20 Kn50 OLJ535 CAAGCAGAAGACGGCATACGAGATGAGAGAGTGACTGGAGTTCAGACGTGT Barcode* GCTCTTCCGATCT Cm20 Kn50 OLJ538 CAAGCAGAAGACGGCATACGAGATGTGTGTGTGACTGGAGTTCAGACGTGT +IPTG GCTCTTCCGATCT Barcode* Illumina OLJ386 ACACTCTTTGGGGGCCAAAATCATTAGGGGATTCATCAG sequencing primer

*Barcode sequence is underlined.

276

VI.4.2. Chapter III Table VI-3. Strains used in Chapter III. Strain Genotype or Phenotype Reference V. cholerae xds RIVET xds-tnpR lacZ::res-neo-sacB-res, ApR, KnR (Schild et al., 2007) ΔphoB xds RIVET ΔphoB xds-tnpR lacZ::res-neo-sacB-res, ApR, KnR This study; (McDonough et al., 2014) Δxds In frame deletion of xds Laboratory strain Plasmids pIVET-VC2621 tnpR delivery vector containing VC2621 (xds) fragment, (Schild et al., 2007) OriR6K, ApR

277

VI.4.3. Chapter IV Table VI-4. Strains used in Chapter IV. Strain Genotype or Phenotype Reference V. cholerae ΔxdsΔdns In frame deletion of xds and dns Laboratory strain ΔphoX In frame deletion of phoX Laboratory strain ΔushA In frame deletion of ushA This study ΔnupC In frame deletion of nupC This study ushA-FLAG C-terminal FLAG tagged UshA This study ΔlacZ In frame deletion of lacZ Laboratory strain ΔpilA In frame deletion of pilA This study ΔushAΔpilA In frame deletion of ushA and pilA This study ΔcpdB In frame deletion of cpdB This study ΔyjjG In frame deletion of yjjG This study ΔyjjGΔVCA0545 In frame deletion of yjjG and VCA0545 This study ΔcpdBΔyjjGΔVCA0545 In frame deletion of cpdB, yjjG, and VCA0545 This study ΔushAΔcpdB In frame deletion of ushA and cpdB This study ΔxdsΔushAΔcpdB In frame deletion of xds, ushA, and cpdB This study ΔdnsΔushAΔcpdB In frame deletion of dns, ushA, and cpdB This study Plasmids pDL1098-flp Temperature-sensitive Flp recombinase delivery vector, Fig. This study; (McDonough S2, CmR et al., 2014) pBAD33kan Laboratory plasmid pDL1098 Temperature-sensitive mTn10 delivery vector, Fig. S1, CmR, This study; (McDonough SpR et al., 2014)

278

Table VI-5. Primers used in Chapter IV. Primer use Primer name Sequence (5' to 3' orientation) Arbitrary primed PCR Arb1 (Hava and Camilli 2002) olj363 GTGTGGGCACTCGACATATGACAAG Arb2 (Hava and Camilli 2002) olj386 ACACTCTTTGGGGGCCAAAATCATTAGGGGATTCATCAG qPCR primers pilA From Ankur Dalia From Ankur Dalia cpdB cpdB qRT F AGATAAAGCCTCCGATCAAAT cpdB qRT R GATCAAATCACCGTTATCGAC ushA ushA qPCR F GTACCAGAATCAGACCTACAAGA ushA qPCR R GGATTATCAAATTCGTGGTTAC Strain construction ΔushA ushA FRT F1 TCACATCGAGTTAGCACGTCTG ushA FRT R1 GTCGACGGATCCCCGGAATCATTGTCATACCTTTGAACTGATG ushA FRT F2 GAAGCAGCTCCAGCCTACATAATAAGGTTTGACTCGCAAAGTTG ushA FRT R2 AGAGGTTACAGGAGTGCGTCAG ushA R0 CTTTGCGCACTTTGATGAAT ΔnupC nupC FRT F1 TACACTGAGCTGCAACGCATTG nupC FRT R1 GTCGACGGATCCCCGGAATCAAATTGTGAGTAGAACAGGAAAGG nupC FRT F2 GAAGCAGCTCCAGCCTACATGATCACAGATTGATGGATTGAG nupC FRT R2 GTTAAGGGTAATAGTGCCTTCAGC nupC R0 CCGACTAAAAACTCCACCTGA ushA-FLAG ushA FRT F1 see above ushA C Flag R1 CTTGTCGTCATCGTCTTTGTAGTCACGATAAACAATCTCGCCCGCTG ushA C Flag F2 ACTACAAAGACGATGACGACAAGTAATAAGGTTTGACTCGCAAAGT TGAATTGCG ushA FRT R2 see above ushA R0 see above ΔpilA pilA FRT F1 CAGATATTGAAACAGGCGACGA pilA FRT R1 GTCGACGGATCCCCGGAATCATATGCCTTGCTACACAAGGG pilA FRT F2 GAAGCAGCTCCAGCCTACATAATGCTCACCAACCTTGTTGC pilA FRT R2 GCCATACTAACCCAATACACTCAT ΔcpdBΔyjjGΔVCA0545 VCA0545 FRT F1 GGTGTGAAAAGTACCAAGGGA VCA0545 FRT R1 TGTAGGCTGGAGCTGCTTCGGCATACGTCTTCTCTCTTTTC VCA0545 FRT F2 GAAGCAGCTCCAGCCTACATAAAACGGATATCTCTTTGCCTT VCA0545 FRT R2 CAGTTCCAAAGCTCACTCC VCA0545 FRT R0 CGTTCGGCTTACCATTTTTCT VCA0608 FRT F1 TTTCGTTGGATGTTGACACTG VCA0608 FRT R1 TGTAGGCTGGAGCTGCTTCTTCATGATGATCTCCTTAAAATCAG VCA0508 FRT F2 GAAGCAGCTCCAGCCTACATAATGCCATGAATAAGCGAGG VCA0608 FRT R2 TCATCACCTCTTTCTATTCACC VCA0608 FRT R0 AGAGCTAGAGAAACTGGAAGAA cpdB FRT F1 TCTCGGTCTCTCCCTGTAAATG cpdB FRT R1 TGTAGGCTGGAGCTGCTTCTTCACTCATAACCAAATTGTGATGTG cpdB FRT F2 GAAGCAGCTCCAGCCTACATAAGCACCGATAATGCCCCTATTG cpdB FRT R2 GTAAGACTTTGGGCTGTTTTCG

279 cpdB FRT R0 CCTCATAGAAAAGAAAACAGCC ABD725 GAAGCAGCTCCAGCCTACA

280

REFERENCES

281

Abuaita, B. H., and J. H. Withey. 2009. Bicarbonate induces Vibrio cholerae virulence gene expression by enhancing ToxT activity. Infect Immun 77:4111-4120. Alam, A., R. C. Larocque, J. B. Harris, C. Vanderspurt, E. T. Ryan, F. Qadri, and S. B. Calderwood. 2005. Hyperinfectivity of human-passaged Vibrio cholerae can be modeled by growth in the infant mouse. Infect Immun 73:6674-6679. Alam, M., M. Sultana, G. B. Nair, A. K. Siddique, N. A. Hasan, R. B. Sack, D. A. Sack, K. U. Ahmed, A. Sadique, H. Watanabe, C. J. Grim, A. Huq, and R. R. Colwell. 2007. Viable but nonculturable Vibrio cholerae O1 in biofilms in the aquatic environment and their role in cholera transmission. Proc Natl Acad Sci U S A 104:17801-17806. Ambudkar, S. V., T. J. Larson, and P. C. Maloney. 1986. Reconstitution of sugar phosphate transport systems of Escherichia coli. J Biol Chem 261:9083-9086. Angelichio, M. J., D. S. Merrell, and A. Camilli. 2004. Spatiotemporal analysis of acid adaptation-mediated Vibrio cholerae hyperinfectivity. Infect Immun 72:2405- 2407. Anraku, Y. 1964a. A new cyclic phosphodiesterase having a 3'-nucleotidase activity from Escherichia Coli B. I. Purification and some properties of the enzyme. J Biol Chem 239:3412-3419. Anraku, Y. 1964b. A new cyclic phosphodiesterase having a 3'-nucleotidase activity from Escherichia Coli B. II. Further studies on substrate specificity and mode of action of the enzyme. J Biol Chem 239:3420-3424. Arribas-Bosacoma, R., S. K. Kim, C. Ferrer-Orta, A. G. Blanco, P. J. Pereira, F. X. Gomis-Ruth, B. L. Wanner, M. Coll, and M. Sola. 2007. The X-ray crystal structures of two constitutively active mutants of the Escherichia coli PhoB receiver domain give insights into activation. J Mol Biol 366:626-641. Aulet, O., C. Silva, S. G. Fraga, M. Pichel, R. Cangemi, C. Gaudioso, N. Porcel, M. A. Jure, M. C. de Castillo, and N. Binsztein. 2007. Detection of viable and viable nonculturable Vibrio cholerae O1 through cultures and immunofluorescence in the Tucuman rivers, Argentina. Rev Soc Bras Med Trop 40:385-390. Baek, J. H., and S. Y. Lee. 2006. Novel gene members in the Pho regulon of Escherichia coli. FEMS Microbiol Lett 264:104-109. Baffone, W., B. Citterio, E. Vittoria, A. Casaroli, R. Campana, L. Falzano, and G. Donelli. 2003. Retention of virulence in viable but non-culturable halophilic Vibrio spp. Int J Food Microbiol 89:31-39. Barua, D. 1992. History of Cholera, p. 1-36. In D. Barua and W. B. Greenough, III (ed.), Cholera. Plenum Publishing Corporation, New York, NY. Behari, J., L. Stagon, and S. B. Calderwood. 2001. pepA, a gene mediating pH regulation of virulence genes in Vibrio cholerae. J Bacteriol 183:178-188. Benenson, A. S., M. R. Islam, and W. B. Greenough, 3rd. 1964. Rapid Identification of Vibrio Cholerae by Darkfield Microscopy. Bull World Health Organ 30:827- 831.

282

Beyhan, S., A. D. Tischler, A. Camilli, and F. H. Yildiz. 2006. Differences in gene expression between the classical and El Tor biotypes of Vibrio cholerae O1. Infect Immun 74:3633-3642. Bjorkman, K. M., and D. M. Karl. 2005. Presence of dissolved nucleotides in the North Pacific Subtropical Gyre and their role in cycling of dissolved organic phosphorus. Aquat. Microb. Ecol. 39:193-203. Blanco, A. G., A. Canals, J. Bernues, M. Sola, and M. Coll. 2011. The structure of a transcription activation subcomplex reveals how sigma(70) is recruited to PhoB promoters. EMBO J 30:3776-3785. Blankenberg, D., G. Von Kuster, N. Coraor, G. Ananda, R. Lazarus, M. Mangan, A. Nekrutenko, and J. Taylor. 2010. Galaxy: a web-based genome analysis tool for experimentalists. Curr Protoc Mol Biol Chapter 19:Unit 19 10 11-21. Blokesch, M., and G. K. Schoolnik. 2008. The extracellular nuclease Dns and its role in natural transformation of Vibrio cholerae. J Bacteriol 190:7232-7240. Bogaert, D., D. Weinberger, C. Thompson, M. Lipsitch, and R. Malley. 2009. Impaired innate and adaptive immunity to Streptococcus pneumoniae and its effect on colonization in an infant mouse model. Infect Immun 77:1613-1622. Bogosian, G., P. J. Morris, and J. P. O'Neil. 1998. A mixed culture recovery method indicates that enteric bacteria do not enter the viable but nonculturable state. Appl Environ Microbiol 64:1736-1742. Bourassa, L., and A. Camilli. 2009. Glycogen contributes to the environmental persistence and transmission of Vibrio cholerae. Mol Microbiol 72:124-138. Branzk, N., and V. Papayannopoulos. 2013. Molecular mechanisms regulating NETosis in infection and disease. Semin Immunopathol 35:513-530. Brookes, J. D., J. Antenucci, M. Hipsey, M. D. Burch, N. J. Ashbolt, and C. Ferguson. 2004. Fate and transport of pathogens in lakes and reservoirs. Environ Int 30:741-759. Broza, M., H. Gancz, and Y. Kashi. 2008. The association between non-biting midges and Vibrio cholerae. Environ Microbiol 10:3193-3200. Brzoska, P., M. Rimmele, K. Brzostek, and W. Boos. 1994. The Pho regulon- dependent Ugp uptake system for glycerol-3-phosphate in Escherichia coli is trans inhibited by Pi. J Bacteriol 176:15-20. Buchanan, J. T., A. J. Simpson, R. K. Aziz, G. Y. Liu, S. A. Kristian, M. Kotb, J. Feramisco, and V. Nizet. 2006. DNase expression allows the pathogen group A Streptococcus to escape killing in neutrophil extracellular traps. Curr Biol 16:396- 400. Burns, D. M., and I. R. Beacham. 1986. Nucleotide sequence and transcriptional analysis of the E. coli ushA gene, encoding periplasmic UDP-sugar hydrolase (5'- nucleotidase): regulation of the ushA gene, and the signal sequence of its encoded protein product. Nucleic Acids Res 14:4325-4342. Butler, S. M., and A. Camilli. 2004. Both chemotaxis and net motility greatly influence the infectivity of Vibrio cholerae. Proc Natl Acad Sci U S A 101:5018-5023. Butler, S. M., E. J. Nelson, N. Chowdhury, S. M. Faruque, S. B. Calderwood, and A. Camilli. 2006. Cholera stool bacteria repress chemotaxis to increase infectivity. Mol Microbiol 60:417-426.

283

Cai, I., N. Friedman, and X. Xie. 2006. Stochastic protein expression in individual cells at the single molecule level. Nature 440:358-362. Camilli, A., D. T. Beattie, and J. J. Mekalanos. 1994. Use of genetic recombination as a reporter of gene expression. Proc Natl Acad Sci U S A 91:2634-2638. Camilli, A., and J. J. Mekalanos. 1995. Use of recombinase gene fusions to identify Vibrio cholerae genes induced during infection. Mol Microbiol 18:671-683. Carter, C. W., Jr. 1995. The nucleoside deaminases for cytidine and adenosine: structure, transition state stabilization, mechanism, and evolution. Biochimie 77:92-98. Cash, R. A., S. I. Music, J. P. Libonati, M. J. Snyder, R. P. Wenzel, and R. B. Hornick. 1974. Response of man to infection with Vibrio cholerae. I. Clinical, serologic, and bacteriologic responses to a known inoculum. J Infect Dis 129:45- 52. Casper-Lindley, C., and F. H. Yildiz. 2004. VpsT is a transcriptional regulator required for expression of vps biosynthesis genes and the development of rugose colonial morphology in Vibrio cholerae O1 El Tor. J Bacteriol 186:1574-1578. Chaba, R., B. M. Alba, M. S. Guo, J. Sohn, N. Ahuja, R. T. Sauer, and C. A. Gross. 2011. Signal integration by DegS and RseB governs the σE-mediated envelope stress response in Escherichia coli. Proc Natl Acad Sci U S A 108:2106-2111. Chan, F. Y., and A. Torriani. 1996. PstB protein of the phosphate-specific transport system of Escherichia coli is an ATPase. J Bacteriol 178:3974-3977. Chang, M. C., S. Y. Chang, S. L. Chen, and S. M. Chuang. 1992. Cloning and expression in Escherichia coli of the gene encoding an extracellular deoxyribonuclease (DNase) from Aeromonas hydrophila. Gene 122:175-180. Charlier, D., G. Hassanzadeh, A. Kholti, D. Gigot, A. Pierard, and N. Glansdorff. 1995. carP, involved in pyrimidine regulation of the Escherichia coli carbamoylphosphate synthetase operon encodes a sequence-specific DNA- binding protein identical to XerB and PepA, also required for resolution of ColEI multimers. J Mol Biol 250:392-406. Charlier, D., A. Kholti, N. Huysveld, D. Gigot, D. Maes, T. L. Thia-Toong, and N. Glansdorff. 2000. Mutational analysis of Escherichia coli PepA, a multifunctional DNA-binding aminopeptidase. J Mol Biol 302:411-426. Chekabab, S. M., G. Jubelin, C. M. Dozois, and J. Harel. 2014. PhoB activates Escherichia coli O157:H7 virulence factors in response to inorganic phosphate limitation. PLoS One 9:e94285. Chen, Z. H., and P. Schaap. 2012. The prokaryote messenger c-di-GMP triggers stalk cell differentiation in Dictyostelium. Nature 488:680-683. Cheng, C., M. J. Wakefield, J. Yang, M. Tauschek, and R. M. Robins-Browne. 2012. Genome-wide analysis of the Pho regulon in a pstCA mutant of Citrobacter rodentium. PLoS One 7:e50682. Chiang, S. L., and J. J. Mekalanos. 2000. Construction of a Vibrio cholerae vaccine candidate using transposon delivery and FLP recombinase-mediated excision. Infect Immun 68:6391-6397. Childers, B. M., and K. E. Klose. 2007. Regulation of virulence in Vibrio cholerae: the ToxR regulon. Future Microbiol 2:335-344.

284

Chin, C. S., J. Sorenson, J. B. Harris, W. P. Robins, R. C. Charles, R. R. Jean- Charles, J. Bullard, D. R. Webster, A. Kasarskis, P. Peluso, E. E. Paxinos, Y. Yamaichi, S. B. Calderwood, J. J. Mekalanos, E. E. Schadt, and M. K. Waldor. 2011. The origin of the Haitian cholera outbreak strain. N Engl J Med 364:33-42. Collinet, B., H. Yuzawa, T. Chen, C. Herrera, and D. Missiakas. 2000. RseB binding to the periplasmic domain of RseA modulates the RseA:sigmaE interaction in the cytoplasm and the availability of sigmaE.RNA polymerase. J Biol Chem 275:33898-33904. Colwell, R. R., P. Brayton, D. Herrington, B. Tall, A. Huq, and M. M. Levine. 1996. Viable but non-culturable Vibrio cholerae 01 revert to a cultivable state in the human intestine. World Journal of Microbiology and Biotechnology 12:28-31. Cox, G. B., D. Webb, J. Godovac-Zimmermann, and H. Rosenberg. 1988. Arg-220 of the PstA protein is required for phosphate transport through the phosphate- specific transport system in Escherichia coli but not for alkaline phosphatase repression. J Bacteriol 170:2283-2286. Cox, G. B., D. Webb, and H. Rosenberg. 1989. Specific amino acid residues in both the PstB and PstC proteins are required for phosphate transport by the Escherichia coli Pst system. J Bacteriol 171:1531-1534. Crepin, S., M. G. Lamarche, P. Garneau, J. Seguin, J. Proulx, C. M. Dozois, and J. Harel. 2008. Genome-wide transcriptional response of an avian pathogenic Escherichia coli (APEC) pst mutant. BMC Genomics 9:568. Dalia, A. B., E. McDonough, and A. Camilli. 2014. Multiplex genome editing by natural transformation. Proc Natl Acad Sci U S A 111:8937-8942. Davies, J., and B. D. Davis. 1968. Misreading of ribonucleic acid code words induced by aminoglycoside antibiotics. The effect of drug concentration. J Biol Chem 243:3312-3316. Ding, Y., B. M. Davis, and M. K. Waldor. 2004. Hfq is essential for Vibrio cholerae virulence and downregulates sigma expression. Mol Microbiol 53:345-354. Diniz, M. M., C. L. Goulart, L. C. Barbosa, J. Farache, L. M. Lery, A. B. Pacheco, P. M. Bisch, and W. M. von Kruger. 2011. Fine-tuning control of phoBR expression in Vibrio cholerae by binding of PhoB to multiple Pho boxes. J Bacteriol 193:6929-6938. DiRita, V. J. 1992. Co-ordinate expression of virulence genes by ToxR in Vibrio cholerae. Mol Microbiol 6:451-458. Dodd, H. N., and J. M. Pemberton. 1996. Cloning, sequencing, and characterization of the nucH gene encoding an extracellular nuclease from Aeromonas hydrophila JMP636. J Bacteriol 178:3926-3933. Dubnau, D. 1999. DNA uptake in bacteria. Annu Rev Microbiol 53:217-244. Edwards, C. J., D. J. Innes, D. M. Burns, and I. R. Beacham. 1993. UDP-sugar hydrolase isozymes in Salmonella enterica and Escherichia coli: silent alleles of ushA in related strains of group I Salmonella isolates, and of ushB in wild-type and K12 strains of E. coli, indicate recent and early silencing events, respectively. FEMS Microbiol Lett 114:293-298.

285

Elvin, C. M., C. M. Hardy, and H. Rosenberg. 1985. Pi exchange mediated by the GlpT-dependent sn-glycerol-3-phosphate transport system in Escherichia coli. J Bacteriol 161:1054-1058. Epstein, P. R. 1993. Algal blooms in the spread and persistence of cholera. Biosystems 31:209-221. Faruque, S. M., M. J. Albert, and J. J. Mekalanos. 1998a. Epidemiology, genetics, and ecology of toxigenic Vibrio cholerae. Microbiol Mol Biol Rev 62:1301-1314. Faruque, S. M., Asadulghani, M. N. Saha, A. R. Alim, M. J. Albert, K. M. Islam, and J. J. Mekalanos. 1998b. Analysis of clinical and environmental strains of nontoxigenic Vibrio cholerae for susceptibility to CTXPhi: molecular basis for origination of new strains with epidemic potential. Infect Immun 66:5819-5825. Faruque, S. M., and G. B. Nair. 2002. Molecular ecology of toxigenic Vibrio cholerae. Microbiol Immunol 46:59-66. Faruque, S. M., Nair, G. B., and Takeda, Y. 2011. Molecular Epidemiology of Toxigenic Vibrio cholerae. In T. R. a. S. K. Bhattacharya (ed.), Epidemiological and Molecular Aspects on Cholera. Springer Science+Business Media, LLC. Faruque, S. M., M. N. Saha, Asadulghani, D. A. Sack, R. B. Sack, Y. Takeda, and G. B. Nair. 2000. The O139 serogroup of Vibrio cholerae comprises diverse clones of epidemic and nonepidemic strains derived from multiple V. cholerae O1 or non-O1 progenitors. J Infect Dis 182:1161-1168. Faruque, S. M., A. K. Siddique, M. N. Saha, Asadulghani, M. M. Rahman, K. Zaman, M. J. Albert, D. A. Sack, and R. B. Sack. 1999. Molecular characterization of a new ribotype of Vibrio cholerae O139 Bengal associated with an outbreak of cholera in Bangladesh. J Clin Microbiol 37:1313-1318. Faruque, S. M., J. Zhu, Asadulghani, M. Kamruzzaman, and J. J. Mekalanos. 2003. Examination of diverse toxin-coregulated pilus-positive Vibrio cholerae strains fails to demonstrate evidence for Vibrio pathogenicity island phage. Infect Immun 71:2993-2999. Ferrieres, L., G. Hemery, T. Nham, A. M. Guerout, D. Mazel, C. Beloin, and J. M. Ghigo. 2010. Silent mischief: bacteriophage Mu insertions contaminate products of Escherichia coli random mutagenesis performed using suicidal transposon delivery plasmids mobilized by broad-host-range RP4 conjugative machinery. J Bacteriol 192:6418-6427. Finkelstein, R. A., M. Boesman-Finkelstein, Y. Chang, and C. C. Hase. 1992. Vibrio cholerae hemagglutinin/protease, colonial variation, virulence, and detachment. Infect Immun 60:472-478. Fisher, S. L., W. Jiang, B. L. Wanner, and C. T. Walsh. 1995. Cross-talk between the histidine protein kinase VanS and the response regulator PhoB. Characterization and identification of a VanS domain that inhibits activation of PhoB. J Biol Chem 270:23143-23149. Focareta, T., and P. A. Manning. 1991a. Distinguishing between the extracellular DNases of Vibrio cholerae and development of a transformation system. Molecular Microbiology 5:2547-2555. Focareta, T., and P. A. Manning. 1987. Extracellular proteins of Vibrio cholerae: molecular cloning, nucleotide sequence and characterization of the

286

deoxyribonuclease (DNase) together with its periplasmic localization in Escherichia coli K-12. Gene 53:31-40. Focareta, T., and P. A. Manning. 1991b. Genetic analysis of the export of an extracellular DNase of Vibrio cholerae using DNase-beta-lactamase fusions. Gene 108:31-37. Fresquet, V., J. B. Thoden, H. M. Holden, and F. M. Raushel. 2004. Kinetic mechanism of asparagine synthetase from Vibrio cholerae. Bioorg Chem 32:63- 75. Freter, R., H. L. Smith, Jr., and F. J. Sweeney, Jr. 1961. An evaluation of intestinal fluids in the pathogenesis of cholera. J Infect Dis 109:35-42. Gardner, S. G., K. D. Johns, R. Tanner, and W. R. McCleary. 2014. The PhoU protein from Escherichia coli interacts with PhoR, PstB, and metals to form a phosphate-signaling complex at the membrane. J Bacteriol 196:1741-1752. Geldreich, E. E. 1996. Pathogenic agents in freshwater resources. Hydrological Processes 10:315-333. Giardine, B., C. Riemer, R. C. Hardison, R. Burhans, L. Elnitski, P. Shah, Y. Zhang, D. Blankenberg, I. Albert, J. Taylor, W. Miller, W. J. Kent, and A. Nekrutenko. 2005. Galaxy: a platform for interactive large-scale genome analysis. Genome Res 15:1451-1455. Glaser, L., A. Melo, and R. Paul. 1967. Uridine diphosphate sugar hydrolase. Purification of enzyme and protein inhibitor. J Biol Chem 242:1944-1954. Goecks, J., A. Nekrutenko, and J. Taylor. 2010. Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol 11:R86. Goss, T. J., C. P. Seaborn, M. D. Gray, and E. S. Krukonis. 2010. Identification of the TcpP-binding site in the toxT promoter of Vibrio cholerae and the role of ToxR in TcpP-mediated activation. Infection and immunity 78:4122-4133. Green, J., J. C. Crack, A. J. Thomson, and N. E. LeBrun. 2009. Bacterial sensors of oxygen. Curr Opin Microbiol 12:145-151. Grigorova, I. L., R. Chaba, H. J. Zhong, B. M. Alba, V. Rhodius, C. Herman, and C. A. Gross. 2004. Fine-tuning of the Escherichia coli sigmaE envelope stress response relies on multiple mechanisms to inhibit signal-independent proteolysis of the transmembrane anti-sigma factor, RseA. Genes Dev 18:2686-2697. Grim, C. J., E. Jaiani, C. A. Whitehouse, N. Janelidze, T. Kokashvili, M. Tediashvili, R. R. Colwell, and A. Huq. 2010. Detection of toxogenic Vibrio cholerae 01 in freshwater lakes of the former Soviet Republic of Georgia. Environmental Microbiology Reports 2:2-6. Grimes, D. 1991. Ecology of estuarine bacteria capable of causing human disease: A review. Estuaries 14:345-360. Guimaraes-Costa, A. B., T. S. DeSouza-Vieira, R. Paletta-Silva, A. L. Freitas- Mesquita, J. R. Meyer-Fernandes, and E. M. Saraiva. 2014. 3'- nucleotidase/nuclease activity allows Leishmania parasites to escape killing by neutrophil extracellular traps. Infect Immun 82:1732-1740. Hamilton, H. L., N. M. Dominguez, K. J. Schwartz, K. T. Hackett, and J. P. Dillard. 2005. Neisseria gonorrhoeae secretes chromosomal DNA via a novel type IV secretion system. Mol Microbiol 55:1704-1721.

287

Hammond, D. J., and W. E. Gutteridge. 1984. Purine and pyrimidine metabolism in the Trypanosomatidae. Mol Biochem Parasitol 13:243-261. Harris, J. B., R. C. LaRocque, F. Chowdhury, A. I. Khan, T. Logvinenko, A. S. Faruque, E. T. Ryan, F. Qadri, and S. B. Calderwood. 2008. Susceptibility to Vibrio cholerae infection in a cohort of household contacts of patients with cholera in Bangladesh. PLoS Negl Trop Dis 2:e221. Harris, R. M., D. C. Webb, S. M. Howitt, and G. B. Cox. 2001. Characterization of PitA and PitB from Escherichia coli. J Bacteriol 183:5008-5014. Hartley, D. M., J. G. Morris, Jr., and D. L. Smith. 2006. Hyperinfectivity: a critical element in the ability of V. cholerae to cause epidemics? PLoS medicine 3:e7. Hasegawa, T., M. Minami, A. Okamoto, I. Tatsuno, M. Isaka, and M. Ohta. 2010. Characterization of a virulence-associated and cell-wall-located DNase of Streptococcus pyogenes. Microbiology 156:184-190. Hava, D. L., and A. Camilli. 2002. Large-scale identification of serotype 4 Streptococcus pneumoniae virulence factors. Mol Microbiol 45:1389-1406. Hekstra, D., and J. Tommassen. 1993. Functional exchangeability of the ABC proteins of the periplasmic binding protein-dependent transport systems Ugp and Mal of Escherichia coli. J Bacteriol 175:6546-6552. Hengge-Aronis, R. 2002. Signal transduction and regulatory mechanisms involved in control of the sigma(S) (RpoS) subunit of RNA polymerase. Microbiol Mol Biol Rev 66:373-395, table of contents. Herrington, D. A., R. H. Hall, G. Losonsky, J. J. Mekalanos, R. K. Taylor, and M. M. Levine. 1988. Toxin, toxin-coregulated pili, and the toxR regulon are essential for Vibrio cholerae pathogenesis in humans. J Exp Med 168:1487-1492. Hoffer, S. M., N. van Uden, and J. Tommassen. 2001. Expression of the Pho regulon interferes with induction of the uhpT gene in Escherichia coli K-12. Arch Microbiol 176:370-376. Hollenbeck, B. L., and L. B. Rice. 2012. Intrinsic and acquired resistance mechanisms in enterococcus. Virulence 3:421-433. Huang, C. S., W. R. Moore, and A. Meister. 1988. On the active site thiol of gamma- glutamylcysteine synthetase: relationships to catalysis, inhibition, and regulation. Proc Natl Acad Sci U S A 85:2464-2468. Huo, A., B. Xu, M. A. Chowdhury, M. S. Islam, R. Montilla, and R. R. Colwell. 1996. A simple filtration method to remove plankton-associated Vibrio cholerae in raw water supplies in developing countries. Appl Environ Microbiol 62:2508- 2512. Huq, A., E. B. Small, P. A. West, M. I. Huq, R. Rahman, and R. R. Colwell. 1983. Ecological relationships between Vibrio cholerae and planktonic crustacean copepods. Appl Environ Microbiol 45:275-283. Huq, A., C. A. Whitehouse, C. J. Grim, M. Alam, and R. R. Colwell. 2008. Biofilms in water, its role and impact in human disease transmission. Curr Opin Biotechnol 19:244-247. Innes, D., I. R. Beacham, and D. M. Burns. 2001. The role of the intracellular inhibitor of periplasmic UDP-sugar hydrolase (5'-nucleotidase) in Escherichia coli: cytoplasmic localisation of 5'-nucleotidase is conditionally lethal. J Basic Microbiol 41:329-337.

288

Ishige, T., M. Krause, M. Bott, V. F. Wendisch, and H. Sahm. 2003. The phosphate starvation stimulon of Corynebacterium glutamicum determined by DNA microarray analyses. J Bacteriol 185:4519-4529. Islam, M. S., Alam, M.J., and P.K.B Neogi. 1992. Seasonality and toxigenicity of Vibrio cholerae non-01 isolated from different components of pond ecosystems of Dhaka City, Bangladesh. World Journal of Microbiology and Biotechnology 8:160-163. Island, M. D., B. Y. Wei, and R. J. Kadner. 1992. Structure and function of the uhp genes for the sugar phosphate transport system in Escherichia coli and Salmonella Typhimurium. J Bacteriol 174:2754-2762. Jahid, I. K., A. J. Silva, and J. A. Benitez. 2006. Polyphosphate stores enhance the ability of Vibrio cholerae to overcome environmental stresses in a low-phosphate environment. Appl Environ Microbiol 72:7043-7049. Jenal, U., and J. Malone. 2006. Mechanisms of Cyclic-di-GMP Signaling in Bacteria. ANNUAL REVIEW OF GENETICS 40:385. Johnston, C., B. Martin, G. Fichant, P. Polard, and J. P. Claverys. 2014. Bacterial transformation: distribution, shared mechanisms and divergent control. Nat Rev Microbiol 12:181-196. Judson, N., and J. J. Mekalanos. 2000. TnAraOut, a transposon-based approach to identify and characterize essential bacterial genes. Nat Biotechnol 18:740-745. Kadner, R. J. 1973a. Genetic Control of the Transport of Hexose Phosphates in Escherichia coli: Mapping of the uhp Locus. J Bacteriol 116:764-770. Kadner, R. J., and H. H. Winkler. 1973b. Isolation and characterization of mutations affecting the transport of hexose phosphates in Escherichia coli. J Bacteriol 113:895-900. Kamal, D., A. N. Khan, M. A. Rahman, and F. Ahamed. 2007. Study on the physico chemical properties of water of Mouri River, Khulna, Bangladesh. Pak J Biol Sci 10:710-717. Kamat, S. S., and F. M. Raushel. 2013. The enzymatic conversion of phosphonates to phosphate by bacteria. Curr Opin Chem Biol 17:589-596. Kamp, H. D., B. Patimalla-Dipali, D. W. Lazinski, F. Wallace-Gadsden, and A. Camilli. 2013. Gene fitness landscapes of Vibrio cholerae at important stages of its life cycle. PLoS Pathog 9:e1003800. Kanehisa, M., and S. Goto. 2000. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28:27-30. Kanehisa, M., S. Goto, Y. Sato, M. Kawashima, M. Furumichi, and M. Tanabe. 2014. Data, information, knowledge and principle: back to metabolism in KEGG. Nucleic Acids Res 42:D199-205. Karaolis, D. K., S. Somara, D. R. Maneval, Jr., J. A. Johnson, and J. B. Kaper. 1999. A bacteriophage encoding a pathogenicity island, a type-IV pilus and a phage receptor in cholera bacteria. Nature 399:375-379. Kasahara, M., K. Makino, M. Amemura, A. Nakata, and H. Shinagawa. 1991. Dual regulation of the ugp operon by phosphate and carbon starvation at two interspaced promoters. J Bacteriol 173:549-558. Kawasaki, H., and S. Iwamuro. 2008. Potential roles of histones in host defense as antimicrobial agents. Infect Disord Drug Targets 8:195-205.

289

Kell, D. B., A. S. Kaprelyants, D. H. Weichart, C. R. Harwood, and M. R. Barer. 1998. Viability and activity in readily culturable bacteria: a review and discussion of the practical issues. Antonie Van Leeuwenhoek 73:169-187. Kendall, E. A., F. Chowdhury, Y. Begum, A. I. Khan, S. Li, J. H. Thierer, J. Bailey, K. Kreisel, C. O. Tacket, R. C. LaRocque, J. B. Harris, E. T. Ryan, F. Qadri, S. B. Calderwood, and O. C. Stine. 2010. Relatedness of Vibrio cholerae O1/O139 isolates from patients and their household contacts, determined by multilocus variable-number tandem-repeat analysis. J Bacteriol 192:4367-4376. Kier, L. D., R. Weppelman, and B. N. Ames. 1977. Regulation of two phosphatases and a cyclic phosphodiesterase of Salmonella Typhimurium. J Bacteriol 130:420- 428. Kierek, K., and P. I. Watnick. 2003. Environmental determinants of Vibrio cholerae biofilm development. Appl Environ Microbiol 69:5079-5088. Klein, B. A., E. L. Tenorio, D. W. Lazinski, A. Camilli, M. J. Duncan, and L. T. Hu. 2012. Identification of essential genes of the periodontal pathogen Porphyromonas gingivalis. BMC Genomics 13:578. Klose, K. E., and J. J. Mekalanos. 1998. Distinct roles of an alternative sigma factor during both free-swimming and colonizing phases of the Vibrio cholerae pathogenic cycle. Mol Microbiol 28:501-520. Kocan, M., S. Schaffer, T. Ishige, U. Sorger-Herrmann, V. F. Wendisch, and M. Bott. 2006. Two-component systems of Corynebacterium glutamicum: deletion analysis and involvement of the PhoS-PhoR system in the phosphate starvation response. J Bacteriol 188:724-732. Koestler, B. J., and C. M. Waters. 2014. Bile acids and bicarbonate inversely regulate intracellular cyclic di-GMP in Vibrio cholerae. Infect Immun 82:3002-3014. Kotra, L. P., J. Haddad, and S. Mobashery. 2000. Aminoglycosides: perspectives on mechanisms of action and resistance and strategies to counter resistance. Antimicrob Agents Chemother 44:3249-3256. Kovacikova, G., and K. Skorupski. 2002a. The alternative sigma factor sigma(E) plays an important role in intestinal survival and virulence in Vibrio cholerae. Infect Immun 70:5355-5362. Kovacikova, G., and K. Skorupski. 2001. Overlapping binding sites for the virulence gene regulators AphA, AphB and cAMP-CRP at the Vibrio cholerae tcpPH promoter. Mol Microbiol 41:393-407. Kovacikova, G., and K. Skorupski. 2002b. Regulation of virulence gene expression in Vibrio cholerae by quorum sensing: HapR functions at the aphA promoter. Mol Microbiol 46:1135-1147. Krasteva, P. V., J. C. Fong, N. J. Shikuma, S. Beyhan, M. V. Navarro, F. H. Yildiz, and H. Sondermann. 2010. Vibrio cholerae VpsT regulates matrix production and motility by directly sensing cyclic di-GMP. Science 327:866-868. Krogh, A., B. Larsson, G. von Heijne, and E. L. Sonnhammer. 2001. Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes. J Mol Biol 305:567-580. Kubena, B. D., H. Luecke, H. Rosenberg, and F. A. Quiocho. 1986. Crystallization and x-ray diffraction studies of a phosphate-binding protein involved in active transport in Escherichia coli. J Biol Chem 261:7995-7996.

290

Kulaev, I., and T. Kulakovskaya. 2000. Polyphosphate and phosphate pump. Annu Rev Microbiol 54:709-734. Lamarche, M. G., B. L. Wanner, S. Crepin, and J. Harel. 2008. The phosphate regulon and bacterial virulence: a regulatory network connecting phosphate homeostasis and pathogenesis. FEMS Microbiology Reviews 32:461-473. Lappann, M., H. Claus, T. van Alen, M. Harmsen, J. Elias, S. Molin, and U. Vogel. 2010. A dual role of extracellular DNA during biofilm formation of Neisseria meningitidis. Mol Microbiol 75:1355-1371. Lara, R. J., S. B. Neogi, M. S. Islam, Z. H. Mahmud, S. Yamasaki, and G. B. Nair. 2009. Influence of catastrophic climatic events and human waste on Vibrio distribution in the Karnaphuli estuary, Bangladesh. Ecohealth 6:279-286. Larocque, R. C., J. B. Harris, M. Dziejman, X. Li, A. I. Khan, A. S. Faruque, S. M. Faruque, G. B. Nair, E. T. Ryan, F. Qadri, J. J. Mekalanos, and S. B. Calderwood. 2005. Transcriptional profiling of Vibrio cholerae recovered directly from patient specimens during early and late stages of human infection. Infect Immun 73:4488-4493. Lazinski, D. W., and A. Camilli. 2013. Homopolymer tail-mediated ligation PCR: a streamlined and highly efficient method for DNA cloning and library construction. Biotechniques 54:25-34. Lebens, M., P. Lundquist, L. Soderlund, M. Todorovic, and N. I. Carlin. 2002. The nptA gene of Vibrio cholerae encodes a functional sodium-dependent phosphate cotransporter homologous to the type II cotransporters of eukaryotes. J Bacteriol 184:4466-4474. Lee, K. S., S. B. Song, K. E. Kim, Y. H. Kim, S. K. Kim, B. H. Kho, D. K. Ko, Y. K. Choi, Y. K. Lee, C. K. Kim, Y. C. Kim, J. Y. Lim, Y. Kim, K. H. Min, and B. L. Wanner. 2000. Cloning and characterization of the UDP-sugar hydrolase gene (ushA) of Enterobacter aerogenes IFO 12010. Biochem Biophys Res Commun 269:526-531. Lee, S. H., S. M. Butler, and A. Camilli. 2001. Selection for in vivo regulators of bacterial virulence. Proc Natl Acad Sci U S A 98:6889-6894. Lee, S. H., D. L. Hava, M. K. Waldor, and A. Camilli. 1999. Regulation and temporal expression patterns of Vibrio cholerae virulence genes during infection. Cell 99:625-634. Lenz, D. H., K. C. Mok, B. N. Lilley, R. V. Kulkarni, N. S. Wingreen, and B. L. Bassler. 2004. The small RNA chaperone Hfq and multiple small RNAs control quorum sensing in Vibrio harveyi and Vibrio cholerae. Cell 118:69-82. Lery, L. M., C. L. Goulart, F. R. Figueiredo, K. S. Verdoorn, M. Einicker-Lamas, F. M. Gomes, E. A. Machado, P. M. Bisch, and W. M. von Kruger. 2013. A comparative proteomic analysis of Vibrio cholerae O1 wild-type cells versus a phoB mutant showed that the PhoB/PhoR system is required for full growth and rpoS expression under inorganic phosphate abundance. J Proteomics 86:1-15. Lewenza, S., R. K. Falsafi, G. Winsor, W. J. Gooderham, J. B. McPhee, F. S. Brinkman, and R. E. Hancock. 2005. Construction of a mini-Tn5-luxCDABE mutant library in Pseudomonas aeruginosa PAO1: a tool for identifying differentially regulated genes. Genome Res 15:583-589.

291

Li, X. Z., D. Ma, D. M. Livermore, and H. Nikaido. 1994. Role of efflux pump(s) in intrinsic resistance of Pseudomonas aeruginosa: active efflux as a contributing factor to beta-lactam resistance. Antimicrob Agents Chemother 38:1742-1752. Liang, W., A. Pascual-Montano, A. J. Silva, and J. A. Benitez. 2007. The cyclic AMP receptor protein modulates quorum sensing, motility and multiple genes that affect intestinal colonization in Vibrio cholerae. Microbiology 153:2964-2975. Lin, E. C. 1976. Glycerol dissimilation and its regulation in bacteria. Annu Rev Microbiol 30:535-578. Lin, W., G. Kovacikova, and K. Skorupski. 2007. The quorum sensing regulator HapR downregulates the expression of the virulence gene transcription factor AphA in Vibrio cholerae by antagonizing Lrp- and VpsR-mediated activation. Molecular microbiology 64:953-967. Lipp, E. K., A. Huq, and R. R. Colwell. 2002. Effects of global climate on infectious disease: the cholera model. Clin Microbiol Rev 15:757-770. Livermore, D. M. 1984. Penicillin-binding proteins, porins and outer-membrane permeability of carbenicillin-resistant and -susceptible strains of Pseudomonas aeruginosa. J Med Microbiol 18:261-270. Lombardo, M. J., J. Michalski, H. Martinez-Wilson, C. Morin, T. Hilton, C. G. Osorio, J. P. Nataro, C. O. Tacket, A. Camilli, and J. B. Kaper. 2007. An in vivo expression technology screen for Vibrio cholerae genes expressed in human volunteers. Proc Natl Acad Sci U S A 104:18229-18234. Lorenz, M. G., and W. Wackernagel. 1994. Bacterial gene transfer by natural genetic transformation in the environment. Microbiol Rev 58:563-602. Majumdar, A., A. Ghatak, and R. K. Ghosh. 2005. Identification of the gene for the monomeric alkaline phosphatase of Vibrio cholerae serogroup O1 strain. Gene 344:251-258. Makino, K., M. Amemura, S. K. Kim, A. Nakata, and H. Shinagawa. 1993. Role of the sigma 70 subunit of RNA polymerase in transcriptional activation by activator protein PhoB in Escherichia coli. Genes Dev 7:149-160. Makino, K., H. Shinagawa, M. Amemura, T. Kawamoto, M. Yamada, and A. Nakata. 1989. Signal transduction in the phosphate regulon of Escherichia coli involves phosphotransfer between PhoR and PhoB proteins. J Mol Biol 210:551- 559. Makino, K., H. Shinagawa, M. Amemura, and A. Nakata. 1986. Nucleotide sequence of the phoR gene, a regulatory gene for the phosphate regulon of Escherichia coli. J Mol Biol 192:549-556. Makino, K., H. Shinagawa, and A. Nakata. 1985. Regulation of the phosphate regulon of Escherichia coli K-12: regulation and role of the regulatory gene phoR. J Mol Biol 184:231-240. Mandlik, A., J. Livny, W. P. Robins, J. M. Ritchie, J. J. Mekalanos, and M. K. Waldor. 2011. RNA-Seq-based monitoring of infection-linked changes in Vibrio cholerae gene expression. Cell Host Microbe 10:165-174. Masip, L., K. Veeravalli, and G. Georgiou. 2006. The many faces of glutathione in bacteria. Antioxid Redox Signal 8:753-762. McCleary, W. R. 1996. The activation of PhoB by acetylphosphate. Mol Microbiol 20:1155-1163.

292

McCleary, W. R., and J. B. Stock. 1994. Acetyl phosphate and the activation of two- component response regulators. J Biol Chem 269:31567-31572. McDonough, E., D. W. Lazinski, and A. Camilli. 2014. Identification of in vivo regulators of the Vibrio cholerae xds gene using a high-throughput genetic selection. Mol Microbiol 92:302-315. Medveczky, N., and H. Rosenberg. 1969. The binding and release of phosphate by a protein isolated from Escherichia coli. Biochim Biophys Acta 192:369-371. Meibom, K. L., M. Blokesch, N. A. Dolganov, C. Y. Wu, and G. K. Schoolnik. 2005. Chitin induces natural competence in Vibrio cholerae. Science 310:1824-1827. Meibom, K. L., X. B. Li, A. T. Nielsen, C. Y. Wu, S. Roseman, and G. K. Schoolnik. 2004. The Vibrio cholerae chitin utilization program. Proceedings of the National Academy of Sciences 101:2524-2529. Mekalanos, J. J. 1983. Duplication and amplification of toxin genes in Vibrio cholerae. Cell 35:253-263. Merrell, D. S., S. M. Butler, F. Qadri, N. A. Dolganov, A. Alam, M. B. Cohen, S. B. Calderwood, G. K. Schoolnik, and A. Camilli. 2002a. Host-induced epidemic spread of the cholera bacterium. Nature 417:642-645. Merrell, D. S., and A. Camilli. 1999. The cadA gene of Vibrio cholerae is induced during infection and plays a role in acid tolerance. Mol Microbiol 34:836-849. Merrell, D. S., D. L. Hava, and A. Camilli. 2002b. Identification of novel factors involved in colonization and acid tolerance of Vibrio cholerae. Mol Microbiol 43:1471-1491. Merrell, D. S., A. D. Tischler, S. H. Lee, and A. Camilli. 2000. Vibrio cholerae requires rpoS for efficient intestinal colonization. Infect Immun 68:6691-6696. Metzger, L. C., and M. Blokesch. 2014. Composition of the DNA-uptake complex of. Mob Genet Elements 4:e28142. Miller, C. J., B. S. Drasar, and R. G. Feachem. 1982. Cholera and estuarine salinity in Calcutta and London. Lancet 1:1216-1218. Miller, C. J., R. G. Feachem, and B. S. Drasar. 1985. Cholera epidemiology in developed and developing countries: new thoughts on transmission, seasonality, and control. Lancet 1:261-262. Miller, E. F., S. Vaish, and R. J. Maier. 2012. Efficiency of purine utilization by Helicobacter pylori: roles for adenosine deaminase and a NupC homolog. PLoS One 7:e38727. Miller, M. B., K. Skorupski, D. H. Lenz, R. K. Taylor, and B. L. Bassler. 2002. Parallel quorum sensing systems converge to regulate virulence in Vibrio cholerae. Cell 110:303-314. Miller, V. L., and J. J. Mekalanos. 1988. A novel suicide vector and its use in construction of insertion mutations: osmoregulation of outer membrane proteins and virulence determinants in Vibrio cholerae requires toxR. J Bacteriol 170:2575-2583. Moisi, M., S. Lichtenegger, S. Tutz, A. Seper, S. Schild, and J. Reidl. 2013. Characterizing the hexose-6-phosphate transport system of Vibrio cholerae, a utilization system for carbon and phosphate sources. J Bacteriol 195:1800-1808.

293

Moorthy, S., and P. I. Watnick. 2004. Genetic evidence that the Vibrio cholerae monolayer is a distinct stage in biofilm development. Molecular Microbiology 52:573-587. Mosley, W. H., S. Ahmad, A. S. Benenson, and A. Ahmed. 1968. The relationship of vibriocidal antibody titre to susceptibility to cholera in family contacts of cholera patients. Bull World Health Organ 38:777-785. Mulcahy, H., L. Charron-Mazenod, and S. Lewenza. 2010. Pseudomonas aeruginosa produces an extracellular deoxyribonuclease that is required for utilization of DNA as a nutrient source. Environ Microbiol 12:1621-1629. Myers, K. S., H. Yan, I. M. Ong, D. Chung, K. Liang, F. Tran, S. Keles, R. Landick, and P. J. Kiley. 2013. Genome-scale analysis of Escherichia coli FNR reveals complex features of transcription factor binding. PLoS Genet 9:e1003565. Nakai, K., and M. Kanehisa. 1991. Expert system for predicting protein localization sites in Gram-negative bacteria. Proteins 11:95-110. Naroditskaya, V., M. J. Schlosser, N. Y. Fang, and K. Lewis. 1993. An E. coli gene emrD is involved in adaptation to low energy shock. Biochem Biophys Res Commun 196:803-809. Nelson, E. J., A. Chowdhury, J. Flynn, S. Schild, L. Bourassa, Y. Shao, R. C. LaRocque, S. B. Calderwood, F. Qadri, and A. Camilli. 2008. Transmission of Vibrio cholerae is antagonized by lytic phage and entry into the aquatic environment. PLoS pathogens 4:e1000187. Neu, H. C. 1967a. The 5'-nucleotidase of Escherichia coli. I. Purification and properties. J Biol Chem 242:3896-3904. Neu, H. C. 1967b. The 5'-nucleotidase of Escherichia coli. II. Surface localization and purification of the Escherichia coli 5'-nucleotidase inhibitor. J Biol Chem 242:3905-3911. Neu, H. C. 1968. The 5'-nucleotidases and cyclic phosphodiesterases (3'-nucleotidases) of the Enterobacteriaceae. J Bacteriol 95:1732-1737. Nguyen, D. T., T. C. Ngo, H. H. Tran, T. P. Nguyen, B. M. Nguyen, K. Morita, and M. Ehara. 2009. Two different mechanisms of ampicillin resistance operating in strains of Vibrio cholerae O1 independent of resistance genes. FEMS Microbiol Lett 298:37-43. Nielsen, A. T., N. A. Dolganov, G. Otto, M. C. Miller, C. Y. Wu, and G. K. Schoolnik. 2006. RpoS controls the Vibrio cholerae mucosal escape response. PLoS pathogens 2:e109. Nielsen, A. T., N. A. Dolganov, T. Rasmussen, G. Otto, M. C. Miller, S. A. Felt, S. Torreilles, and G. K. Schoolnik. 2010. A bistable switch and anatomical site control Vibrio cholerae virulence gene expression in the intestine. PLoS Pathog 6. Nielsen, K. M., P. J. Johnsen, D. Bensasson, and D. Daffonchio. 2007. Release and persistence of extracellular DNA in the environment. Environ Biosafety Res 6:37- 53. Ogawa, N., C. M. Tzeng, C. D. Fraley, and A. Kornberg. 2000. Inorganic polyphosphate in Vibrio cholerae: genetic, biochemical, and physiologic features. J Bacteriol 182:6687-6693.

294

Okoh, A. I., T. Sibanda, and S. S. Gusha. 2010. Inadequately treated wastewater as a source of human enteric viruses in the environment. Int J Environ Res Public Health 7:2620-2637. Oppezzo, O. J., and D. N. Anton. 1995. Involvement of cysB and cysE genes in the sensitivity of Salmonella Typhimurium to mecillinam. J Bacteriol 177:4524-4527. Orlowski, M., and A. Meister. 1970. The gamma-glutamyl cycle: a possible transport system for amino acids. Proc Natl Acad Sci U S A 67:1248-1255. Osorio, C. G., J. A. Crawford, J. Michalski, H. Martinez-Wilson, J. B. Kaper, and A. Camilli. 2005. Second-generation recombination-based in vivo expression technology for large-scale screening for Vibrio cholerae genes induced during infection of the mouse small intestine. Infect Immun 73:972-980. Oussenko, I. A., R. Sanchez, and D. H. Bechhofer. 2004. Bacillus subtilis YhcR, a high-molecular-weight, nonspecific endonuclease with a unique domain structure. J Bacteriol 186:5376-5383. Overduin, P., W. Boos, and J. Tommassen. 1988. Nucleotide sequence of the ugp genes of Escherichia coli K-12: homology to the maltose system. Mol Microbiol 2:767-775. Park, J. T., and J. L. Strominger. 1957. Mode of action of penicillin. Science 125:99- 101. Pascual, M., K. Koelle, and A. P. Dobson. 2006. Hyperinfectivity in cholera: a new mechanism for an old epidemiological model? PLoS medicine 3:e280. Paul, J. H., and A. W. David. 1989. Production of extracellular nucleic acids by genetically altered bacteria in aquatic-environment microcosms. Appl Environ Microbiol 55:1865-1869. Paul, J. H., M. F. Deflaun, and W. H. Jeffrey. 1988. Mechanisms of DNA utilization by estuarine microbial populations. Appl Environ Microbiol 54:1682-1688. Petersen, T. N., S. Brunak, G. von Heijne, and H. Nielsen. 2011. SignalP 4.0: discriminating signal peptides from transmembrane regions. Nat Methods 8:785- 786. Peterson, K. M. 2002. Expression of Vibrio cholerae virulence genes in response to environmental signals. Curr Issues Intest Microbiol 3:29-38. Pinchuk, G. E., C. Ammons, D. E. Culley, S. M. Li, J. S. McLean, M. F. Romine, K. H. Nealson, J. K. Fredrickson, and A. S. Beliaev. 2008. Utilization of DNA as a sole source of phosphorus, carbon, and energy by Shewanella spp.: ecological and physiological implications for dissimilatory metal reduction. Appl Environ Microbiol 74:1198-1208. Poncet, S., E. Milohanic, A. Maze, J. Nait Abdallah, F. Ake, M. Larribe, A. E. Deghmane, M. K. Taha, M. Dozot, X. De Bolle, J. J. Letesson, and J. Deutscher. 2009. Correlations between carbon metabolism and virulence in bacteria. Contrib Microbiol 16:88-102. Pradel, E., and P. L. Boquet. 1988. Acid phosphatases of Escherichia coli: molecular cloning and analysis of agp, the structural gene for a periplasmic acid glucose phosphatase. J Bacteriol 170:4916-4923. Pratt, J. T., A. M. Ismail, and A. Camilli. 2010. PhoB regulates both environmental and virulence gene expression in Vibrio cholerae. Mol Microbiol.

295

Pratt, J. T., E. McDonough, and A. Camilli. 2009. PhoB regulates motility, biofilms, and cyclic di-GMP in Vibrio cholerae. J Bacteriol 191:6632-6642. Qin, Z., Y. Ou, L. Yang, Y. Zhu, T. Tolker-Nielsen, S. Molin, and D. Qu. 2007. Role of autolysin-mediated DNA release in biofilm formation of Staphylococcus epidermidis. Microbiology 153:2083-2092. Rao, N. N., and A. Torriani. 1990. Molecular aspects of phosphate transport in Escherichia coli. Mol Microbiol 4:1083-1090. Reidl, J., and K. E. Klose. 2002. Vibrio cholerae and cholera: out of the water and into the host. FEMS Microbiol Rev 26:125-139. Ritchie, J. M., H. Rui, R. T. Bronson, and M. K. Waldor. 2010. Back to the future: studying cholera pathogenesis using infant rabbits. MBio 1. Rittmann, D., U. Sorger-Herrmann, and V. F. Wendisch. 2005. Phosphate starvation- inducible gene ushA encodes a 5' nucleotidase required for growth of Corynebacterium glutamicum on media with nucleotides as the phosphorus source. Appl Environ Microbiol 71:4339-4344. Rosenberg, H., R. G. Gerdes, and K. Chegwidden. 1977. Two systems for the uptake of phosphate in Escherichia coli. J Bacteriol 131:505-511. Roy, N. K., R. K. Ghosh, and J. Das. 1982. Monomeric alkaline phosphatase of Vibrio cholerae. J Bacteriol 150:1033-1039. Safa, A., G. B. Nair, and R. Y. Kong. 2010. Evolution of new variants of Vibrio cholerae O1. Trends Microbiol 18:46-54. Sainsbury, S., L. Bird, V. Rao, S. M. Shepherd, D. I. Stuart, W. N. Hunter, R. J. Owens, and J. Ren. 2011. Crystal structures of penicillin-binding protein 3 from Pseudomonas aeruginosa: comparison of native and antibiotic-bound forms. J Mol Biol 405:173-184. Sakai, Y., K. Toda, Y. Mitani, M. Tsuda, S. Shinoda, and T. Tsuchiya. 1987. Properties of the membrane-bound 5'-nucleotidase and utilization of extracellular ATP in Vibrio parahaemolyticus. J Gen Microbiol 133:2751-2757. Salamov, V. S. a. A. 2011. Automatic Annotation of Microbial Genomes and Metagenomic Sequences., p. 61-78. In R. W. Li (ed.), Megagenomics and its Applications in Agriculture, Biomedicine, and Environmental Studies. Nova Science Publishers. Sant, M. V., W. N. Gatlewar, and S. K. Bhindey. 1975. Epidemiological studies on cholera in non-endemic regions with special reference to the problem of carrier state during epidemic and non-epidemic period. Prog Drug Res 19:594-601. Schild, S., R. Tamayo, E. J. Nelson, F. Qadri, S. B. Calderwood, and A. Camilli. 2007. Genes induced late in infection increase fitness of Vibrio cholerae after release into the environment. Cell Host & Microbe 2:264-277. Schweizer, H., and W. Boos. 1985. Regulation of ugp, the sn-glycerol-3-phosphate transport system of Escherichia coli K-12 that is part of the pho regulon. J Bacteriol 163:392-394. Schwoppe, C., H. H. Winkler, and H. E. Neuhaus. 2002. Properties of the glucose-6- phosphate transporter from Chlamydia pneumoniae (HPTcp) and the glucose-6- phosphate sensor from Escherichia coli (UhpC). J Bacteriol 184:2108-2115. Sebastian, M., and J. W. Ammerman. 2009. The alkaline phosphatase PhoX is more widely distributed in marine bacteria than the classical PhoA. ISME J 3:563-572.

296

Sebastian, M., and J. W. Ammerman. 2011. Role of the phosphatase PhoX in the phosphorus metabolism of the marine bacterium Ruegeria pomeroyi DSS-3. Environ Microbiol Rep 3:535-542. Seper, A., V. H. Fengler, S. Roier, H. Wolinski, S. D. Kohlwein, A. L. Bishop, A. Camilli, J. Reidl, and S. Schild. 2011. Extracellular nucleases and extracellular DNA play important roles in Vibrio cholerae biofilm formation. Mol Microbiol 82:1015-1037. Seper, A., A. Hosseinzadeh, G. Gorkiewicz, S. Lichtenegger, S. Roier, D. R. Leitner, M. Rohm, A. Grutsch, J. Reidl, C. F. Urban, and S. Schild. 2013. Vibrio cholerae evades neutrophil extracellular traps by the activity of two extracellular nucleases. PLoS Pathog 9:e1003614. Sharma, S., P. Sachdeva, and J. S. Virdi. 2003. Emerging water-borne pathogens. Appl Microbiol Biotechnol 61:424-428. Silva, A. J., and J. A. Benitez. 2004. Transcriptional regulation of Vibrio cholerae hemagglutinin/protease by the cyclic AMP receptor protein and RpoS. Journal of bacteriology 186:6374-6382. Sinha, S., J. Mell, and R. Redfield. 2013. The availability of purine nucleotides regulates natural competence by controlling translation of the competence activator Sxy. Mol Microbiol 88:1106-1119. Skorupski, K., and R. K. Taylor. 1997. Control of the ToxR virulence regulon in Vibrio cholerae by environmental stimuli. Molecular microbiology 25:1003-1009. Skorupski, K., and R. K. Taylor. 1999. A new level in the Vibrio cholerae ToxR virulence cascade: AphA is required for transcriptional activation of the tcpPH operon. Mol Microbiol 31:763-771. Skovgaard, N. 2007. New trends in emerging pathogens. Int J Food Microbiol 120:217- 224. Skovierova, H., G. Rowley, B. Rezuchova, D. Homerova, C. Lewis, M. Roberts, and J. Kormanec. 2006. Identification of the sigmaE regulon of Salmonella enterica serovar Typhimurium. Microbiology 152:1347-1359. Smith, K. P., S. Kumar, and M. F. Varela. 2009. Identification, cloning, and functional characterization of EmrD-3, a putative multidrug efflux pump of the major facilitator superfamily from Vibrio cholerae O395. Arch Microbiol 191:903-911. Solovyev, V., and A. Salamov. 2011. Automatic Annotation of Microbial Genomes and Metagenomic Sequences., p. 61-78. In R.W.Li (ed.), Metagenomics and its Applications in Agriculture, Biomedicine and Environmental Studies. Nova Science Publishers. Srivastava, D., M. L. Hsieh, A. Khataokar, M. B. Neiditch, and C. M. Waters. 2013. Cyclic di-GMP inhibits Vibrio cholerae motility by repressing induction of transcription and inducing extracellular polysaccharide production. Mol Microbiol 90:1262-1276. Steed, P. M., and B. L. Wanner. 1993. Use of the rep technique for allele replacement to construct mutants with deletions of the pstSCAB-phoU operon: evidence of a new role for the PhoU protein in the phosphate regulon. J Bacteriol 175:6797- 6809. Stewart, V. 1993. Nitrate regulation of anaerobic respiratory gene expression in Escherichia coli. Mol Microbiol 9:425-434.

297

Stine, O. C., M. Alam, L. Tang, G. B. Nair, A. K. Siddique, S. M. Faruque, A. Huq, R. Colwell, R. B. Sack, and J. G. Morris, Jr. 2008. Seasonal cholera from multiple small outbreaks, rural Bangladesh. Emerg Infect Dis 14:831-833. Stirling, C. J., S. D. Colloms, J. F. Collins, G. Szatmari, and D. J. Sherratt. 1989. xerB, an Escherichia coli gene required for plasmid ColE1 site-specific recombination, is identical to pepA, encoding aminopeptidase A, a protein with substantial similarity to bovine lens leucine aminopeptidase. EMBO J 8:1623- 1627. Sudarsan, N., E. R. Lee, Z. Weinberg, R. H. Moy, J. N. Kim, K. H. Link, and R. R. Breaker. 2008. Riboswitches in eubacteria sense the second messenger cyclic di- GMP. Science 321:411-413. Suzuki, H., S. Kamatani, E. S. Kim, and H. Kumagai. 2001. Aminopeptidases A, B, and N and dipeptidase D are the four cysteinylglycinases of Escherichia coli K- 12. J Bacteriol 183:1489-1490. Suzuki, H., H. Kumagai, and T. Tochikura. 1986. gamma-Glutamyltranspeptidase from Escherichia coli K-12: purification and properties. J Bacteriol 168:1325- 1331. Suzuki, S., A. Ferjani, I. Suzuki, and N. Murata. 2004. The SphS-SphR two component system is the exclusive sensor for the induction of gene expression in response to phosphate limitation in Synechocystis. J Biol Chem 279:13234-13240. Tamayo, R., B. Patimalla, and A. Camilli. 2010. Growth in a biofilm induces a hyperinfectious phenotype in Vibrio cholerae. Infect Immun 78:3560-3569. Tamayo, R., J. T. Pratt, and A. Camilli. 2007. Roles of cyclic diguanylate in the regulation of bacterial pathogenesis. ANNUAL REVIEW OF MICROBIOLOGY 61:131. Tamayo, R., S. Schild, J. T. Pratt, and A. Camilli. 2008. Role of cyclic Di-GMP during El Tor biotype Vibrio cholerae infection: characterization of the in vivo- induced cyclic Di-GMP phosphodiesterase CdpA. Infect Immun 76:1617-1627. Tamayo, R., A. D. Tischler, and A. Camilli. 2005. The EAL domain protein VieA is a cyclic diguanylate phosphodiesterase. J Biol Chem 280:33324-33330. Tischler, A. D., and A. Camilli. 2004. Cyclic diguanylate (c-di-GMP) regulates Vibrio cholerae biofilm formation. Mol Microbiol 53:857-869. Tischler, A. D., and A. Camilli. 2005. Cyclic diguanylate regulates Vibrio cholerae virulence gene expression. Infect Immun 73:5873-5882. Townsend, G. E., 2nd, V. Raghavan, I. Zwir, and E. A. Groisman. 2013. Intramolecular arrangement of sensor and regulator overcomes relaxed specificity in hybrid two-component systems. Proc Natl Acad Sci U S A 110:E161-169. Tran, Q. H., T. Arras, S. Becker, G. Holighaus, G. Ohlberger, and G. Unden. 2000. Role of glutathione in the formation of the active form of the oxygen sensor FNR ([4Fe-4S].FNR) and in the control of FNR function. Eur J Biochem 267:4817- 4824. Trevors, J. T. 2011. Viable but non-culturable (VBNC) bacteria: Gene expression in planktonic and biofilm cells. J Microbiol Methods 86:266-273. Trulzsch, K., A. Roggenkamp, C. Pelludat, A. Rakin, C. Jacobi, and J. Heesemann. 2001. Cloning and characterization of the gene encoding periplasmic 2',3'-cyclic phosphodiesterase of Yersinia enterocolitica O:8. Microbiology 147:203-213.

298 van Mourik, A., N. M. Bleumink-Pluym, L. van Dijk, J. P. van Putten, and M. M. Wosten. 2008. Functional analysis of a Campylobacter jejuni alkaline phosphatase secreted via the Tat export machinery. Microbiology 154:584-592. van Opijnen, T., K. L. Bodi, and A. Camilli. 2009. Tn-seq: high-throughput parallel sequencing for fitness and genetic interaction studies in microorganisms. Nat Methods 6:767-772. van Veen, H. W. 1997. Phosphate transport in prokaryotes: molecules, mediators and mechanisms. Antonie Van Leeuwenhoek 72:299-315. van Veen, H. W., T. Abee, G. J. Kortstee, W. N. Konings, and A. J. Zehnder. 1993. Mechanism and energetics of the secondary phosphate transport system of Acinetobacter johnsonii 210A. J Biol Chem 268:19377-19383. van Veen, H. W., T. Abee, G. J. Kortstee, W. N. Konings, and A. J. Zehnder. 1994. Translocation of metal phosphate via the phosphate inorganic transport system of Escherichia coli. Biochemistry 33:1766-1770. VanBogelen, R. A., E. R. Olson, B. L. Wanner, and F. C. Neidhardt. 1996. Global analysis of proteins synthesized during phosphorus restriction in Escherichia coli. J Bacteriol 178:4344-4366. Villarreal-Chiu, J. F., J. P. Quinn, and J. W. McGrath. 2012. The genes and enzymes of phosphonate metabolism by bacteria, and their distribution in the marine environment. Front Microbiol 3:19. Virkki, L. V., J. Biber, H. Murer, and I. C. Forster. 2007. Phosphate transporters: a tale of two solute carrier families. Am J Physiol Renal Physiol 293:F643-654. Vogel, U., and K. F. Jensen. 1994. The RNA chain elongation rate in Escherichia coli depends on the growth rate. J Bacteriol 176:2807-2813. von Kruger, W. M., S. Humphreys, and J. M. Ketley. 1999. A role for the PhoBR regulatory system homologue in the Vibrio cholerae phosphate-limitation response and intestinal colonization. Microbiology 145:2463-2475. von Kruger, W. M., L. M. Lery, M. R. Soares, F. S. de Neves-Manta, C. M. Batista e Silva, A. G. Neves-Ferreira, J. Perales, and P. M. Bisch. 2006. The phosphate- starvation response in Vibrio cholerae O1 and phoB mutant under proteomic analysis: disclosing functions involved in adaptation, survival and virulence. Proteomics 6:1495-1511. Waldor, M. K., and J. J. Mekalanos. 1996. Lysogenic conversion by a filamentous phage encoding cholera toxin. Science 272:1910-1914. Wang, H. H., and G. M. Church. 2011a. Multiplexed genome engineering and genotyping methods applications for synthetic biology and metabolic engineering. Methods Enzymol 498:409-426. Wang, Y., H. Wang, Z. Cui, H. Chen, Z. Zhong, B. Kan, and J. Zhu. 2011b. The prevalence of functional quorum-sensing systems in recently emerged Vibrio cholerae toxigenic strains. Environ Microbiol Rep 3:218-222. Wanner, B. L., and P. Latterell. 1980. Mutants affected in alkaline phosphatase, expression: evidence for multiple positive regulators of the phosphate regulon in Escherichia coli. Genetics 96:353-366. Watanabe, K., S. Tomioka, K. Tanimura, H. Oku, and K. Isoi. 2011. Uptake of AMP, ADP, and ATP in Escherichia coli W. Biosci Biotechnol Biochem 75:7-12.

299

Waters, C. M., and B. L. Bassler. 2005. Quorum sensing: cell-to-cell communication in bacteria. Annu Rev Cell Dev Biol 21:319-346. Waters, C. M., W. Lu, J. D. Rabinowitz, and B. L. Bassler. 2008. Quorum sensing controls biofilm formation in Vibrio cholerae through modulation of cyclic di- GMP levels and repression of vpsT. J Bacteriol 190:2527-2536. Watnick, P., and R. Kolter. 2000. Biofilm, city of microbes. J Bacteriol 182:2675-2679. Watnick, P. I., K. J. Fullner, and R. Kolter. 1999. A Role for the mannose-sensitive hemagglutinin in biofilm formation by Vibrio cholerae El Tor, p. 3606-3609, vol. 181. Am Soc Microbiol. Weil, A. A., A. I. Khan, F. Chowdhury, R. C. Larocque, A. S. Faruque, E. T. Ryan, S. B. Calderwood, F. Qadri, and J. B. Harris. 2009. Clinical outcomes in household contacts of patients with cholera in Bangladesh. Clin Infect Dis 49:1473-1479. Werner, A., and R. K. Kinne. 2001. Evolution of the Na-P(i) cotransport systems. Am J Physiol Regul Integr Comp Physiol 280:R301-312. Weston, L. A., and R. J. Kadner. 1988. Role of uhp genes in expression of the Escherichia coli sugar-phosphate transport system. J Bacteriol 170:3375-3383. Whitchurch, C. B., T. Tolker-Nielsen, P. C. Ragas, and J. S. Mattick. 2002. Extracellular DNA required for bacterial biofilm formation. Science 295:1487. White, A. E. 2009. New insights into bacterial acquisition of phosphorus in the surface ocean. Proc Natl Acad Sci U S A 106:21013-21014. White, A. K., and W. W. Metcalf. 2007. Microbial metabolism of reduced phosphorus compounds. Annu Rev Microbiol 61:379-400. Wilke, M. S., A. L. Lovering, and N. C. Strynadka. 2005. Beta-lactam antibiotic resistance: a current structural perspective. Curr Opin Microbiol 8:525-533. Willsky, G.R., R.L. Bennett, and M.H. Malamy. 1973. Inorganic phosphate transport in Escherichia coli: involvement of two genes which play a role in alkaline phosphatase regulation. J. Bacteriol 113: 529-539. Willsky, G.R., and M.H. Malamy. 1976. Control of the synthesis of alkaline phosphatase and the phosphate-binding protein in Escherichia coli. J. Bacteriol 127: 595-609. Willsky, G.R., and M.H. Malamy. 1980. Characterization of two genetically sebarable inorganic phosphate transport systems in Escherichia coli. J. Bacteriol 144:356- 365. Wilson, W. A., P. J. Roach, M. Montero, E. Baroja-Fernandez, F. J. Munoz, G. Eydallin, A. M. Viale, and J. Pozueta-Romero. 2010. Regulation of glycogen metabolism in yeast and bacteria. FEMS Microbiol Rev 34:952-985. Withey, J. H., and V. J. Dirita. 2006. The toxbox: specific DNA sequence requirements for activation of Vibrio cholerae virulence genes by ToxT. Mol Microbiol 59:1779-1789. Yamada, M., K. Makino, H. Shinagawa, and A. Nakata. 1990. Regulation of the phosphate regulon of Escherichia coli: properties of phoR deletion mutants and subcellular localization of PhoR protein. Mol Gen Genet 220:366-372. Yamamoto, K., K. Hirao, T. Oshima, H. Aiba, R. Utsumi, and A. Ishihama. 2005. Functional characterization in vitro of all two-component signal transduction systems from Escherichia coli. J Biol Chem 280:1448-1456.

300

Yan, N. 2013. Structural advances for the major facilitator superfamily (MFS) transporters. Trends Biochem Sci 38:151-159. Yang, W. 2011. Nucleases: diversity of structure, function and mechanism. Q Rev Biophys 44:1-93. Yildiz, F. H., and G. K. Schoolnik. 1998. Role of rpoS in stress survival and virulence of Vibrio cholerae. J Bacteriol 180:773-784. Yin, Y., X. He, P. Szewczyk, T. Nguyen, and G. Chang. 2006. Structure of the multidrug transporter EmrD from Escherichia coli. Science 312:741-744. Youil, R., and L. R. Finch. 1988. Isolation and characterization of Mycoplasma mycoides subsp. mycoides mutants deficient in nucleoside monophosphate transport. J Bacteriol 170:5922-5924. Young, R., and H. Bremer. 1976. Polypeptide-chain-elongation rate in Escherichia coli B/r as a function of growth rate. Biochem J 160:185-194. Yu, R. R., and V. J. DiRita. 1999. Analysis of an autoregulatory loop controlling ToxT, cholera toxin, and toxin-coregulated pilus production in Vibrio cholerae. J Bacteriol 181:2584-2592. Yuan, Z. C., R. Zaheer, R. Morton, and T. M. Finan. 2006. Genome prediction of PhoB regulated promoters in Sinorhizobium meliloti and twelve proteobacteria. Nucleic Acids Res 34:2686-2697. Zapun, A., C. Contreras-Martel, and T. Vernet. 2008. Penicillin-binding proteins and beta-lactam resistance. FEMS Microbiol Rev 32:361-385. Zawrotniak, M., and M. Rapala-Kozik. 2013. Neutrophil extracellular traps (NETs) - formation and implications. Acta Biochim Pol 60:277-284. Zhou, L., G. Gregori, J. M. Blackman, J. P. Robinson, and B. L. Wanner. 2005. Stochastic activation of the response regulator PhoB by noncognate histidine kinases. Journal of Integrative Bioinformatics. Zhu, J., and J. J. Mekalanos. 2003. Quorum sensing-dependent biofilms enhance colonization in Vibrio cholerae. Developmental cell 5:647-656. Zhu, J., M. B. Miller, R. E. Vance, M. Dziejman, B. L. Bassler, and J. J. Mekalanos. 2002. Quorum-sensing regulators control virulence gene expression in Vibrio cholerae. Proc Natl Acad Sci U S A 99:3129-3134. Zimmer, D. P., E. Soupene, H. L. Lee, V. F. Wendisch, A. B. Khodursky, B. J. Peter, R. A. Bender, and S. Kustu. 2000. Nitrogen regulatory protein C-controlled genes of Escherichia coli: scavenging as a defense against nitrogen limitation. Proc Natl Acad Sci U S A 97:14674-14679.

301