Identification of Host Factors Including an Epithelial Cell Receptor Responsible for Clostridium difficile TcdB-Mediated Cytotoxicity

By

Michelle Ellen LaFrance

Dissertation

Submitted to the Faculty of the

Graduate School of Vanderbilt University

in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

in

Biochemistry

May, 2016

Nashville, Tennessee

Approved:

Dana Borden Lacy, Ph.D.

Walter J. Chazin, Ph.D.

David Cortez, Ph.D.

Terence S. Dermody, M.D.

To my family

ii ACKNOWLEDGEMENTS

I am extremely thankful to Dr. Borden Lacy, who took a leap of faith and allowed me to join her lab to undertake a project that was outside the normal scope of the lab’s research at the time. The lab’s focus was structural biology, but she decided to take on a student who wished to pursue a molecular biology-based project. It was a risk, and it was a lot of hard work for both of us, but it paid off. Borden, you have helped make me a better scientist, a better writer, and a stronger person. Thank you. I would also like to acknowledge the members of the Lacy lab who have been there on this journey with me. It’s been a long, hard road, filled with ups and downs, but I’m grateful for knowing every one of you. Big, big thanks to Dr. Rory Pruitt and Dr. Kelly Gangwer for getting me started in the lab when I first joined. Thank you to Dr. Melissa Farrow for doing the microscopy and tissue sample work for the PVRL3 project. Thanks go to Dr. Heather Kroh, Ramya Chandrasekaran, and Stacey Seeback for reagents, helping hands, great lunches, and lasting friendships. Dr. Ryan Craven, benchmate extraordinaire, for never getting angry at my mess crowding his space. The rest of the Lacy lab, past and present, thank you. The Spiller lab: our next-door neighbors, friends, and extended lab family. Dr. Benjamin Spiller, always there with unique insight and questions to make us think harder about our science. Members of the Spiller lab past and present, you guys are awesome. Thank you for your friendship. This project would not have been possible without Dr. Donald H. Rubin and the post- doctoral fellow working in his lab, Dr. Jinsong Sheng, with whom we performed the -trap screen. Song Sheng constructed the gene trap library in the Caco-2 cell line and performed the genomic DNA sequencing. Dr. Rubin performed the identification of the disrupted and assisted in designing follow-up experiments. Dr. Rubin was always there with words of encouragement when this project was getting tough—which was often—and for that I am forever grateful. I would like to thank the members of the Dermody lab, especially Dr. Bernardo Mainou and Dr. Jennifer Anstadt. Their advice on follow-up experiments for receptor identification projects was invaluable. The Jam-A ectodomain construct was also a gift from the Dermody lab (Chapter III). Thank you to my thesis committee, Dr. Walter Chazin, Dr. David Cortez, and Dr. Terence Dermody, for their guidance and support.

iii The PVRL1, PVRL2, and PVRL3 ectodomain constructs used for pull down experiments (Chapter III), competition assays (Chapter III), and electron microscopy (Chapter IV) were a very generous gift from Dr. Lawrence Shapiro of Columbia University. Oliver Harrison, a former graduate student in his lab, patiently taught me how to express and purify these difficult via multiple email exchanges. This work was financially supported by the Burroughs Wellcome Foundation, the Cellular and Molecular Microbiology Training Program, the Vanderbilt Digestive Disease Research Center (NIH P30DK058404), and the National Institute of Allergy and Infectious Diseases (NIH R01AI095755). And last but certainly not least, my family. My brother Dan, sister Jackie, and Dan’s longtime girlfriend Michelle were there for me when I needed an ear to talk to, even though they were hundreds of miles away. The care packages of homemade food, pastries from Mike’s, and craft supplies kept me going through the hard times. I can’t thank you enough. You always seem to know exactly what I need when I need it. My parents set me on this journey when I was very young, and encouraged me to keep going throughout my entire school career. They have been my rock, my source of strength, telling me to keep going when I wasn’t sure I could. I owe them everything. Thank you, Mom and Dad. This is for you.

iv TABLE OF CONTENTS

DEDICATION ...... ii

ACKNOWLEDGEMENTS ...... iii

LIST OF TABLES ...... vii

LIST OF FIGURES ...... viii

LIST OF ABBREVIATIONS ...... ix

Chapter Page I. INTRODUCTION ...... 1

Historical Overview ...... 1 C. difficile Infection ...... 2 Symptoms and treatment of C. difficile infection ...... 2 Emergence of the hypervirulent NAP1 strain ...... 2 CDI as a toxin-mediated disease ...... 3 LCT Domain Structure and Function ...... 4 Combined Repetitive OligoPeptides (CROPS) and receptor-binding domain ...... 6 Delivery domain ...... 8 Autoprocessing domain ...... 9 Glucosyltransferase domain ...... 10 Effects of TcdB on cells ...... 11 The cytopathic pathway ...... 13 The cytotoxic pathway ...... 13 Known information about TcdA and TcdB receptor binding ...... 15 Poliovirus-Receptor Like/ Family ...... 18 Discovery of PVR as the poliovirus receptor ...... 18 Other PVR-like (PVRL) proteins as viral receptors ...... 19 PVR proteins as adherens junctions molecules ...... 21 Other roles for nectin proteins ...... 22 Research Objectives ...... 23

II. GENE-TRAP SCREEN TO IDENTIFY HOST FACTORS INVOLVED IN TCDB- MEDIATED CYTOTOXICITY ...... 25

Introduction ...... 25 Results ...... 25 Primary Screen: Gene-Trap ...... 25 Secondary Screen: RNAi viability assays ...... 28 AP2 is required for TcdB ...... 29 Galectin-3 is required for TcdB cytotoxicity ...... 31 Discussion ...... 35

v Methods ...... 37

III. IDENTIFICATION OF AN EPITHELIAL CELL RECEPTOR RESPONSIBLE CLOSTRIDIUM DIFFICILE TCDB-INDUCED CYTOTOXICITY ...... 42

Introduction ...... 42 Results ...... 43 PVRL3 is required for TcdB-mediated cytotoxicity ...... 43 PVRL3 is required for the cytopathic effect in Caco-2 cells but not HeLa cells .. 46 TcdB binds directly to the PVRL3 ectodomain ...... 51 Inhibition of the TcdB-PVRL3 interaction leads to cellular survival ...... 52 Discussion ...... 53 Methods ...... 56

IV. SUMMARY ...... 60

Conclusions ...... 60 Future directions ...... 61 Understanding the role of galectin-3 in TcdB-induced cytotoxicity ...... 61 Other gene-trap screen hits ...... 62 Interplay between PVRL3 and CSPG4, and possibly other as-yet unknown receptors ...... 63 Virus-PVR interactions as a guide for understanding the TcdB-PVRL3 structure ...... 64 Binding studies with PVRL3 ...... 66 Structure: crystallography and electron microscopy based studies of TcdB- PVRL3 complex ...... 68 Activation of NOX by PVRL3 ...... 69 Role of PVRL3 in a mouse model of infection ...... 70 Hypothetical applications ...... 71 Concluding Remarks...... 72 APPENDIX ...... 74 List of publications ...... 74

BIBLIOGRAPHY ...... 75

vi

LIST OF TABLES

Table Page 1. Large Clostridial Toxins ...... 5 2. Gene-Trapped Genes Identified as Involved in TcdB-Mediated Cytotoxicity ...... 26

vii LIST OF FIGURES

Figure Page 1.1 TcdA and TcdB structures ...... 5 1.2 Structure of the CROPS domain ...... …...... 6 1.3 Structures of the TcdB APD and GTD ...... 10 1.4 The GTPase cycle ...... 11 1.5 Mechanism of LCT function ...... 12 1.6 Crystal structure of the PVRL1 dimer ...... 22 2.1 AP2 is required for TcdB cytotoxicity and entry into cells, but not TcdA intoxication of cells ...... 30 2.2 Galectin-3 is required for TcdB-mediated cytotoxicity in Caco-2 and HeLa cells ...... 32 2.3 Effects of galectin-3 on TcdB ...... 34 3.1 Disruption of PVRL3 expression in Caco-2 cells results in resistance to TcdB ...... 44 3.2 Disruption of PVRL3 in HeLa cells results in cell survival ...... 45 3.3 PVRL3 mediates apoptosis in Caco-2 but not HeLa cells ...... 47 3.4 Knockdown of PVRL3 in HeLa does not inhibit TcdB-induced Rac1 glucosylation ...... 48 3.5 Cytopathic effects of TcdB on HeLa cells ...... 49 3.6 TcdB-induced cytopathic effects are impaired in Caco-2 E19 ...... 50 3.7 PVRL3 binds directly to TcdB ...... 51 3.8 TcdA binds to a variety of glycosylated proteins ...... 52 3.9 Extracellular inhibition of the TcdB-PVRL3 interaction confers protection against TcdB ...... 53 4.1 Gene-Trap Clone D05 is disrupted in EHD1 and E- expression ...... 63 4.2 RT-PCR profiles of common cell lines ...... 64 4.3 Proposed binding pocket for PVRL3 on TcdB ...... 66 4.4 PVRL3 D1 interacts directly with TcdB ...... 67 4.5 Microscale thermophoresis ...... 68 4.6 Residues critical for binding to ...... 69 4.7 Electron microscopy images of TcdB-PVRL3 complex ...... 70 4.8 Effects of PVRL3 in MEF cells ...... 72

viii LIST OF ABBREVIATIONS

AJ Adherens junction AP2S1 Adaptor Protein 2S1 APD Autoprocessing domain AREG Amphiregulin ARL6IP1 ADP-riboylation factor-like6 interacting protein 1 ATP Adenosine triphosphate BCA Bicinchoninic acid protein quantification assay bp CAPN12 12 CCNG1 Cyclin G1 CDH17 Cadherin 17 CDI C. difficile infection CHO Chinese hamster ovary cells CLTC heavy chain CMV Cytomegalovirus CPEB4 Cytoplasmic polyadenylation element binding protein 4 CRISPR Clustered regularly interspaced short palindromic repeats CRD Carbohydrate recognition domain CROPS Combined repetitive oligonucleotides CSPG4 Chondriotin sulfate proteoglycan 4 D1 Domain 1, N-terminal Ig domain of Ig superfamily protein D2 Domain 2 from N-terminus of Ig superfamily protein D3 Domain 3 from N-terminus of Ig superfamily protein DMEM Dulbecco’s modified Eagle’s medium DNA Deoxyribonucleic acid DRG1 Developmentally regulated GTP binding protein 1 eIF3B Eukaryotic initiation factor 3B EHD1 EPS15 homology domain containing-1 FAD Flavin adenine dinuleotide FAK Focal adhesion kinase

ix FBS Fetal bovine serum Fc Fragment, crystallizable FGL1 Fibrinogen-like 1 GAP GTPase activating protein GAPDH Glyceraldehyde-3-phosphate dehydrogenase GDI GDP dissociation inhibitor GDP Guanosine diphosphate GEF Guanine nucleotide exchange factor GFP Green fluorescent protein GST Glutathione S-transferase fusion protein tag GTP Guanosine triphosphate GTPase Guanosine triphosphatase GTD Glucosyltransferase domain ELISA Enzyme linked immunosorbent assay FNDC1 Fibronectin III domain containing 1 HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid HIV Human immunodeficiency virus HIS1H2BI Histone cluster 1, H2B

His6 Hexa-histidine tag HMGB1 High Mobility Group Box 1 hnRNPH1 Heterogeneous nuclear ribonucleoprotein H1 HSV Herpes simplex virus HVEM Herpesvirus entry mediator, HveA Ig Immunoglobulin IHC Immunohistochemistry

InsP6 Inositolhexakisphosphate IP Immunoprecipitation JAM Junctional adhesion molecule kDa Kilodaltons LCT Large Clostridial toxin LD50 Lethal dose, 50% LDH Lactate dehydrogenase LGALS3 Galectin-3

x LGALS3BP Galectin-3 binding protein LGALS7 Galectin-7 LRP1 Low-density lipoprotein receptor-related protein 1, receptor for TpeL MAPK1IP1L Mitogen activated 1 interacting protein 1-like MEF Mouse embryonic fibroblast, a common cell line MEM Minimal essential medium miR192 microRNA192 miR194-2 microRNA194-2 Mb Mega base pair MPC2 Mitochondrial pyruvate carrier 2 MST Microscale thermophoresis MV Measlesvirus N4PB2L2 NEDD4 binding protein 2 like 2 NADP+ Nicotinamide adenine dinucleotide phosphate (oxidized) NADPH Nicotinamide adenine dinucleotide phosphate (reduced) Ni-NTA Nickel nitrilotriacetic acid-agarose affinity purification resin NOX NADPH Oxidase complex NOXA1 NOX Activator-1 NOXO1 NOX Organizer-1 NRIP Nuclear receptor interaction protein PAK p21 activated kinase PARM1 Prostate androgen regulated mucin-like 1 PBS Phosphate buffered saline PCR Polymerase chain reaction PLIN2 Perilipin PMC Pseudomembranous colitis PMN Polymorphonuclear cell POU3F3 Pou class 3 homeobox 3 PRR15L Proline-rich region 15-like PV Poliovirus PVDF Polyvinylidene fluoride PVR Poliovirus receptor, CD155, Nectin-like 5 PVRL Poliovirus receptor-like, nectin

xi PVRL1 Poliovirus receptor-like 1, nectin-1, receptor for HSV PVRL2 Poliovirus receptor-like 2, nectin-2, receptor for HSV PVRL3 Poliovirus receptor-like 3, nectin-3, receptor for TcdB PVRL4 Poliovirus receptor-like 4, nectin-4, receptor for measlesvirus PZD PSD95-Dlg1-ZO1 domain RBM39 RNA binding motif 39 RIPA Radioimmunoprecipitation assay buffer, used for cell lysis ROS Reactive oxygen species RPS15A Ribosomal Protein small 15A RSPH3 Radial spoke 3 homolog RT-PCR Reverse-transcriptase polymerase chain reaction RNA Ribonucleic acid RNAi RNA interference RTX Repeats in Toxin SC5DL Sterol-C5-desaturase-like SDS-PAGE Sodium dodecylsulfate polyacrylamide gel electrophoresis SEM Standard error of the mean shRNA Short hairpin RNA siRNA Small interfering RNA SLAM Signaling lymphocyte activation molecule SLCO2B1 Solute carrier 2B1 SNX5 5 SPR Surface plasmon resonance Taf1d TATA box binding protein associated factor 1D TcdA Clostridium difficile toxin A TcdB Clostridium difficile toxin B Tcnα Clostridium novyi alpha toxin TcsH Clostridium sordelli hemorrhagic toxin TcsL Clostridium sordelli lethal toxin TM4SF1 Transmembrane 4 superfamily 1 TpeL Clostridium perfringens large cytotoxin TTR Transthyretin UDP Uridine diphosphate

xii CHAPTER I INTRODUCTION

Historical Overview Ivan Hall and Elizabeth O’Toole first identified the bacterium Clostridium difficile in 1935 during a study of the normal intestinal flora of infants in the first 10 days following birth (1). The difficulty of isolating and culturing this rod-shaped, spore-forming anaerobic bacterium led the researchers to name the new strain Bacillus difficilis. When injected into guinea pigs and rabbits, Hall and O’Toole found that B. difficilis was surprisingly pathogenic, causing subcutaneous edema, weight loss, intense pain, spasms, and death within hours in most animals. Post-mortem analysis showed that in all but one case, phagocytic leukocytes and edema were found near the site of inoculation. The researchers found no evidence of bacteria in other tissues and therefore did not believe that the infection had spread, yet the lungs and gallbladder had swollen in some animals, indicating a systemic reaction. Hall and O’Toole next tested for the presence of a soluble exotoxin produced by B. difficilis by injecting a culture filtrate into guinea pigs. Four of five animals injected with the culture filtrate died within 48 h; only the animal injected with the 1:1000 dilution of the filtrate survived the experiment. Hall and O’Toole also showed that the exotoxin was heat-labile, as boiling the sample before injecting it into animals ablated the toxin activity. This paper constituted both the first report of the isolation of the bacterium and of its associated toxigenic factors. Despite the highly pathogenic nature of the bacterium and the accompanying exotoxin seen in rodents, the infants from whom B. difficilis was initially isolated were asymptomatic. A follow up study was performed to determine whether the subcutaneous injection of the bacteria may have altered the results (2). Researchers found that feeding culture supernatant to animals did not result in illness, while injection did. Therefore, B. difficilis was assumed to be a normal part of the intestinal flora, only causing disease when outside of its normal environment. In the 1970s, both the bacteria and the toxins from B. difficilis, now renamed Clostridium difficile, were linked to antibiotic-associated pseudomembranous colitis (PMC) and found in stools of patients undergoing treatment with clindamycin (3-9). Very quickly, many other groups also correlated antibiotic use, PMC, presence of Clostridial toxins in stool samples, and

1 recurrent C. difficile infection, implicating C. difficile as the primary cause of PMC (reviewed in (10)).

C. difficile Infection

Symptoms and treatment of C. difficile infection C. difficile infection (CDI) most commonly presents as mild to moderate diarrhea; however, symptoms can range from asymptomatic carriage to life-threatening conditions such as PMC and toxic megacolon. Most commonly, CDI is contracted in a hospital setting and is normally referred to as a nosocomial infection (11), although the incidence of community- acquired CDI is increasing (12). Healthy patients rarely develop CDI. The major risk factor for CDI is treatment with antibiotics, which can disrupt the normal gut flora. If the patient then comes in contact with C. difficile spores, which are notoriously hardy and difficult to eradicate, C. difficile can colonize the colonic environment (11). PMC is characterized by of the colon and the formation of small yellow plaques, or lesions called pseudomembranes, on the colonic surface, which are filled with fibrin and neutrophils, as well as necrotized epithelial cells. Symptoms include severe abdominal pain and distension, fever, and severe diarrhea. In the most severe cases, the colonic muscle tone is damaged, such that the colon swells as is seen in ulcerative colitis. This can lead, paradoxically, to a decrease in diarrhea. The distension, referred to as toxic megacolon, can lead to rupture of the bowel if not treated promptly. The recommended course of treatment for all cases of CDI is cessation of current antibiotic treatment and administration of metronidizole for mild cases and vancomycin for severe cases (13). The capacity of C. difficile to form hardy spores makes it difficult to completely eliminate it from the environment, leading to a high recurrence rate, between 20- 30% (14, 15). In cases where perforation of the bowel has occurred, a partial colectomy to surgically remove damaged tissue may be required (11).

Emergence of the hypervirulent NAP1 strain A survey of eight hospitals in six states in 2001-2003 in the United States revealed a sharp increase in the number of cases of CDI and the severity of disease (16). The strains obtained from patients during the epidemic were compared with historical isolates and found to primarily represent a hypervirulent strain of C. difficile termed BI/NAP1/027. This strain has a

2 deletion in the tcdC , which is thought to be a negative regulator of toxin expression. In support of this idea, Warny and coworkers found that toxins A and B (see below) are indeed overproduced in the NAP1 strain compared with the historical strains (17), which may lead to the hypervirulent phenotype. The hypervirulent NAP1 strain is also more resistant to antibiotics, particularly fluoroquinolones, than historical strains (16). Increased use of fluoroquinolones in medical settings may have selected for the NAP1 strain over the historical strain and resulted in increased colonization with the NAP1 strain in patients. C. difficile is currently the most common cause of hospital-acquired diarrhea in the United States and has also become the most common cause of healthcare-associated infection in hospitals in the United States (12). The incidence of CDI has been rising for the last two decades (12, 18). In 2011, C. difficile was responsible for half a million infections and 29,000 deaths in the U.S. (12). The burden of cost for controlling C. difficile spread in critical care facilities alone is estimated to be in excess of $4.8 billion (19). Thus, there is a need for more study into the infection mechanism and development of new therapeutics.

CDI as a toxin-mediated disease Hall and O’Toole noted when they first discovered C. difficile that disease was caused by an exotoxin, since culture filtrate injected into animals resulted in edema and death (1). When research into the disease began again in the 1970s, it was noted that the disease was toxin- mediated and that the toxin was similar to that of C.C. sordellii, as antitoxin to C. sordellii lethal toxin was neutralizing (3, 4). In 1981, it was discovered that there were in fact two toxins produced by C. difficile (20) (21). Purified toxin A causes tissue damage and fluid accumulation in rabbit ileal loops, while purified toxin B does not. Toxin B, however, has a 20,000-fold higher LD50 in a tissue culture model (21). These papers began referring to toxin A (later referred to as TcdA) as an enterotoxin and toxin B (TcdB) as a cytotoxin. Other groups confirmed that TcdA was the primary cause of the tissue damage and fluid leakage in rodent models of intoxication, while TcdB did not cause tissue damage in these models (22, 23). The question then became, is TcdB necessary for C. difficile pathogenesis, or is TcdB more of an accessory factor? Strains of C. difficile pathogenic to humans were later isolated that were TcdA-/TcdB+, but no clinical isolate has yet been identified that is TcdA+/TcdB- (24, 25). These findings indicate that TcdB is required for pathogenesis in humans.

3 In 2009, Dena Lyras and coworkers successfully engineered isogenic toxin knockouts in C. difficile bacteria using homologous recombination to create truncated toxin genes and then used a hamster model of infection to show that TcdB is necessary for infection, while TcdA is dispensable (26). Although the tcdB knockout bacteria overproduce TcdA by western blot, these bacteria are still avirulent in hamsters. The authors found that the only animals that succumbed to infection were infected with wild-type revertants. In contrast, tcdA knockout bacteria do not differ from the wild-type controls. The same group went on to expand on the role of TcdA and TcdB in pathogenesis by creating isogenic knockouts in a hypervirulent NAP1 strain and examining intestinal tissue and distal organ damage in multiple animal models (27). In these experiments, the TcdA-/TcdB+ strain is significantly more virulent than the TcdA+/TcdB- strain in both mouse and hamster models, using metrics of weight loss, labored breathing, diarrhea, and survival. The double- knockout TcdA-/TcdB- strain is avirulent. Based on histopathology scores on gut tissue of infected mice, the TcdA-TcdB+ strain and wild-type bacterial infections cause severe damage to the surface epithelium, in contrast to what was previously seen in the rabbit ileal loop model of intoxication (21). There is polymorphonuclear (PMN) cell influx, severe submucosal edema, severe erosion of the epithelium, and ulceration of the mucosa (27). In contrast, the TcdA+TcdB- strain only caused mild edema and tissue damage. Finally, the Lyras group also examined distal tissue damage, as C. difficile causes multi- organ failure in rare cases (28). They found that only the TcdB-producing strains cause this phenomenon, again highlighting the role of TcdB in systemic disease.

LCT Domain Structure and Function TcdA and TcdB are members of a family of large Clostridial toxins (LCTs), which also includes the hemorrhagic and lethal toxins from C. sordellii (TcsH and TcsL, respectively), alpha toxin from C. novyi (Tcnα), and large cytotoxin from C. perfringens (TpeL). These toxins are very large single-chain polypeptides, on the order of 250-300 kDa, except for the slightly shorter TpeL, which lacks the C-terminal Combined Repetitve Oligo Peptides (CROPS) domain. The LCT family members are listed in Table 1. LCTs are multi-domain A/B toxins, with a catalytic N-terminal domain, or the A component, capable of causing toxicity to cells, and the remainder of the toxin, or the B component,

4 responsible for delivering the N-terminus to the cytosol (a review of A/B toxins can be found in (29)). LCTs cause cytotoxicity by modifying Rho- and Ras-family GTPases, inactivating them and disrupting downstream signaling events. Known Rho and Ras family targets of LCTs are listed in Table 1. The B component of the LCTs is further sub-divided into three structurally distinct domains: a combined repetitive oligopeptides (CROPS) domain at the C-terminus that is involved in carbohydrate binding, a delivery domain responsible for pore formation in the endosomal membrane, and an autoprocessing domain with protease activity capable of cleaving the N- terminal catalytic domain, releasing it into the cytosol (Figure 1.1A). Each of these domains and their functions is discussed in more detail below.

Table 1: Large Clostridial Toxins Toxin Species Mol. Weight Major Molecular Targets TcdA C. difficile 309 kDa RhoA/B/C, Rac1, Cdc42 TcdB C. difficile 270 kDa RhoA/B/C, Rac1, Cdc42 TcsL C. sordellii 270 kDa Rac1, H/K/N-Ras, Rap2, Ral TcsH C. sordellii 300 kDa Rac1 Tcnα C. novyi 250 kDa RhoA/B/C, Rac1 TpeL C. perfringens 191 kDa H/K/N-Ras

Figure 1.1: TcdA and TcdB structures. A. TcdA and TcdB are homologous A/B toxins consisting of four domains each. A component: Glucosyltransferase domain (GTD) in red, responsible for catalytic activity and cellular effects. B component consists of three domains: Autoprocessing domain (APD) in blue, responsible for cleaving and releasing GTD into the cytosol; Delivery domain in yellow, responsible for pore formation in the endosomal membrane following endocytosis; and Combined repetitive oligopeptides (CROPS) domain in green, involved in cell surface binding. B and C. Negative stain electron microscopy averages of TcdA (B) and TcdB (C) holotoxins. D. Crystal structure of TcdA residues 1-1832 fitted into the 20 Å EM map of TcdA. GTD in red, APD in blue, delivery domain in yellow, InsP6 in green. Last residue visible in the structure is 1802. Panels B, and C taken from (30); panel D from (31).

5 Combined Repetitive OligoPeptides (CROPS) and receptor-binding domain The gene for TcdA was sequenced in 1990 (32, 33). When the genes for TcdA and TcdB were first cloned, it was noted that the 3’ ends were repetitive (34). Seven long repeats of 50 residues each and 23 short repeats of 21 residues were identified for TcdA. The number of short repeats was later updated to 31, based on more accurate analysis (35). The TcdB C- terminus contains four long repeats and 19 short repeats. Taken together, von Eichel-Streiber and Sauerborn termed the repeating C-terminus the Combined Repetitive Oligo-Peptides, or CROPS (34). The CROPS domain of TcdA covers residues 1809-2710, and for TcdB, the domain consists of residues 1810-2366. There is homology to the repetitive carbohydrate- binding sequences at the C-terminus of a Streptococcal glucosyltransferase, GtfB, which is involved in biofilm production (34). This led researchers to propose that TcdA and TcdB bind carbohydrates at their C-termini, consistent with earlier work showing that TcdA could bind to glycosylated moieties on rabbit erythrocytes (36).

Figure 1.2: Structure of the CROPS domain. A. Crystal structure of the last two repeats of the TcdA CROPS domain, including a Galα1-3Galβ1- 4GlcNAc derivative shown in red. Short repeats in dark blue, long repeats in green, short repeats following long repeats with loops coordinating saccharides in cyan. PDB ID: 2G7C (37) B and C. Models of the CROPS domains of TcdA (B) and TcdB (C) based on the solved crystal structure fragments. Short repeats in blue, long repeats in purple. Figures taken from (35).

A portion of the TcdA CROPS domain containing the C-terminal 127 residues has been crystallized (35). The fragment consists of one long repeat and three short repeats; each repeat adopts a β-hairpin fold followed by a loop. The same group later crystallized a larger fragment of

6 the TcdA CROPS (37) (Figure 1.2) encompassing the C-terminal 255 residues, spanning nine short repeats and two long repeats. This structure was also crystallized with a Galα1-3Galβ1- 4GlcNAc derivative, two of which can be seen in the structure and in Figure 1.2A in red. The crystal structure shows several short repeats, shown in Figure 1.2A in dark blue, followed by a long β-hairpin repeat in green. The loop after the hairpin on the long repeat (green) forms a pocket in which the primary galactose binds. After the next short repeat (cyan), another loop forms a pocket to accommodate the third sugar (37). This pocket is shallower, which allows for more space to accommodate the attached glycolipid or glycoprotein chain. The packing of the β- hairpins results in a 120° rotation of each long repeat relative to the previous, giving the overall shape of the CROPS domain a solenoid structure (30, 35, 37) (Figures 1.1B, C, and D, Figures 1.2 B and C). Although the TcdA CROPS binds glycans, the TcdB CROPS has not been demonstrated to. In 2008, Dingle and coworkers showed that fragments of the TcdA CROPS domain competes for binding with TcdA to cells and agglutinate rabbit erythrocytes (38). Full-length TcdB CROPS domain, however, does not compete for TcdB binding or neutralize the effects of TcdB on cells, suggesting that the receptor-binding site for TcdB is outside of the CROPS domain. In addition, a glycan microarray using both the TcdA and TcdB CROPS fragments identified several more potential interacting ligands for TcdA,. Their data is uploaded to the Consortium for Functional Glycomics database. Unfortunately, the array yielded no significant hits for TcdB. Although it is possible that the chip used did not have the correct linkages necessary for TcdB recognition, it is also possible that TcdB CROPS does not recognize glycans. The CROPS domain is not required for LCT function in vitro. Several labs have reported that truncations of TcdA and TcdB lacking the CROPS domain are capable of intoxicating cells in culture, albeit at higher concentrations (39-41). It seems likely that the CROPS domain is involved binding to the cell surface, but specific, high-affinity receptor interactions occur outside of the CROPS domain. Two groups have identified a small region between the CROPS domain and the delivery domain that may be involved in receptor binding. In 2014, Schorch and colleagues identified a receptor for TpeL, the large cytotoxin from C. perfringens (41). TpeL is the only LCT identified to date that does not have a CROPS domain, and the authors mapped the receptor-binding site to the C-terminal 444 residues of TpeL. Yuan et al. identified chondroitin sulfate proteoglycan 4

7 (CSPG4) as a receptor involved in TcdB-mediated cytopathic effects (see below) in HeLa cells (42). Again, the authors mapped the binding site of CSPG4 to the 350 residues N-terminal to the CROPS domain of TcdB. Receptor binding is followed by endocytosis of the toxins. The LCTs use a clathrin- mediated mechanism of endocytosis to enter cells (43). By knocking down multiple genes in the clathrin pathway, as well as using pharmacological inhibitors, TcdB was demonstrated to gain access to the cytosol of multiple cell types by a clathrin-, and not -, mediated mechanism. They then extrapolated their data to conclude that other LCTs, including TcdA, Tcnα, and TcsL, also use the clathrin pathway. Their data for this were not as strong, and mostly showed that these toxins use a -dependent pathway. This observation will be discussed further in Chapter II.

Delivery domain Following receptor ligation and endocytosis, the N-terminal catalytic subunit must pass through the endosomal membrane to reach the cytosol. For many A/B toxins, this process involves structural rearrangement and insertion of hydrophobic segments into the endosomal membrane to induce pore formation (29). This has not been explicitly shown for the LCTs. Qa’Dan et al. gathered evidence that TcdB undergoes conformational changes upon treatment with a low pH buffer that leads to exposure of regions of hydrophobicity (44). Using a fluorescent probe that binds hydrophobic regions of a protein, they noted that following a drop to pH 4.0, fluorescence signal increased significantly, indicating a conformational change of the protein. TcdB is capable of forming pores using a 86Rb+ release assay (45). By pre-loading chilled CHO cells with 86Rb+, binding TcdB to the cell surface, and acidifying the extracellular milieu, the toxin inserted into the plasma membrane and created an ion channel through which the efflux of 86Rb+ was measured by scintillation counting. Channels were also formed by inserting TcdB into artificial lipid bilayers upon lowering of pH. Acidification is a necessary step. Bafilomycin, an inhibitor of the endosomal proton pump, can inhibit TcdB cytopathicity in cells, presumably by blocking pore formation (44, 45), and showing that the toxin requires an acidified to function properly. The overall topology of TcdA at neutral pH is that of a head domain and two tails: a shorter, inner tail and the longer, solenoid-shaped outer tail that is the CROPS domain (Figure

8 1.1B and D). The head domain is composed of the autoprocessing and glucosyltransferase domains, while the inner tail is comprised of the delivery domain (Figure 1.1D). (31). The 3.3 Å structure of residues 1-1806 of TcdA reveals a complex delivery domain unlike anything in the PDB to date. The delivery domain of TcdA encompasses residues 842- 1806 and map to the inner tail of the electron microscopy structure. Four highly hydrophobic alpha helices (residues 1026-1135) wrap around the outside of the delivery domain from head region to the end of the tail, and are thought to contain the pore- forming residues. Mutations in this hydrophobic region disrupt access of the toxin to the cell cytoplasm, inhibit 86Rb+ release, and decrease toxicity of TcdA (31). The delivery domain secondary structure then adopts a series of β-sheets as it travels back to the head region of the toxin.

Autoprocessing domain After the pore is formed in the endosomal membrane, the N-terminal two domains of the toxin pass into the cytosol. Cleavage of the N-terminal glucosyltransferase domain (GTD) from the rest of the toxin and the endosome allows it to access GTPase substrates. The GTD alone is released into the cytosol, while the other domains of the toxin remain in the endosome (46). TcdB cleavage depends, not on host proteins, but on the small molecule inositolhexakisphosphate (InsP6), and was proposed to represent an autoproteolytic aspartate protease. (47). Egerer et al. noted the homology to a domain within the Vibrio cholerae RTX toxin, which is an autoproteolytic cysteine protease (48, 49). A homology search identified a catalytic triad of D587, H653, and C698 (residue numbers in TcdB) and a leucine at the cleavage site (L544 in TcdA or L543 in TcdB). Mutation of the catalytic triad or the leucine cleavage site abolished cleavage of the GTD from the rest of the toxin. The crystal structures of the autoprocessing domains (APD) for both TcdA and TcdB have been determined (50, 51). For TcdB, the structure encompasses residues 544-797 (51) (Figure 1.3A). The core is a β-sheet, with a smaller three-strand β-sheet called β-flap (Figure

1.3A, blue) coordinating the InsP6 cofactor. Several basic residues form a positively charged binding pocket which binds InsP6. Binding of InsP6 transmits an allosteric signal across the APD to the catalytic protease site.

9

Figure 1.3: Structures of the TcdB APD and GTD. A. Crystal structure of the autoprocessing domain, with InsP6 and catalytic triad indicated. PDB ID 3PEE (51). B. Crystal structure of TcdB GTD with UDP-glucose in blue and Mn2+ as a green sphere. C. Zoom of B with catalytic residues indicated. B&C: PDB ID 2BVM (52).

Glucosyltransferase domain The TcdA and TcdB glucosyltransferase domains (GTDs) use UDP-glucose as the hexose donor substrate (53, 54). In vitro GTPase substrates for TcdA and TcdB include RhoA, Rac1, and Cdc42, but not H-Ras, Rab5, or Arf1, indicating specificity of the enzymes. Modification on RhoA was mapped to threonine 37 by mass spectrometry of a pepsin digest of an in-vitro modified RhoA peptide. This threonine residue is located in the critical Switch-1 region of the GTPase, essential for downstream effector binding (see below). Additional targets for the LCTs have been identified by mass spectrometry and are listed in Table 1 (55). The TcdB GTD consists of 543 amino acids of mixed α/β structure (52). The catalytic core consists of 7 mostly parallel β strands in a Rossman-like fold. The glucosyltransferase reaction includes the GTPase and UDP-glucose as substrates and relies on a manganese ion cofactor. A DXD motif, highly conserved across glucosyltransferases, is involved in Mn2+ and UDP-glucose binding. The two aspartate residues in TcdB are D286 and D288. In the crystal structure, UDP-glucose is cleaved, with the glucose hydrogen bonded to D270, R273, and D286, and the (cleaved) glycosidic bond is solvent-exposed, allowing for transfer to the GTPase substrate. The intracellular molecular targets of the LCTs are Rho- and Ras- family GTPases. These proteins are molecular switches that control many cellular functions, including cell cycle control, actin cytoskeleton dynamics, vesicular trafficking, and apoptosis. The GTPase cycles between “on” and “off” by regulating its nucleotide binding state: its active form is bound to GTP, and after hydrolysis, the GTPase changes conformation to adopt its inactive, GDP-bound form

10 (56). Cycling is aided by several accessory factors. The intrinsic GTPase activity, the catalysis of GTP to GDP, is facilitated by GTPase activating proteins (GAPs). The inactive GDP-bound form is activated again by guanine nucleotide exchange factors (GEFs) which exchange GDP for GTP. In its GTP-bound form, the GTPase can bind effector proteins and initiate downstream signaling events. The inactive GDP-bound GTPases are bound by GDP dissociation inhibitors (GDIs), inhibiting the exchange of nucleotides, keeping the protein in an inactive conformation.

Figure 1.4: The GTPase cycle. GTPases (Rac1 here as an example) are low molecular weight GTP binding proteins that are in an active conformation when bound to GTP (right) and inactive when bound to GDP (left). When in the active conformation, GTPases interact with effector proteins and influence cell activities. GTPase activating proteins (GAPs) increase the rate of hydrolysis of GTP to GDP; guanine nucleotide exchange factors (GEFs) exchange GDP for GTP and activate the GTPase. Far left: GDP- dissociation inhibitors (GDIs) sequester the GTPase in the cytosol, away from the rest of the cycle.

The major targets of the TcdB GTD include RhoA/B/C, Rac1, and Cdc42 (55). Specifically with respect to C. difficile infection, Rac1 inactivation leads to loss of cell polarity, focal adhesions, and actin filaments in the colonic tissue, ultimately causing pathology. The GTD inactivates its target by monoglucosylation of Thr37 on RhoA or Thr35 on Rac1 and Cdc42, located on the Switch-1 while the GTPase is in the GDP-bound conformational state. After monoglucosylation, GTPases cannot interact with downstream effectors and cannot be activated by GEFs (57, 58). This effectively inhibits Rho-dependent GTPase signaling in the cell.

Effects of TcdB on cells TcdB activates two cellular pathways, designated as cytopathic and cytotoxic (Figure 1.5). The cytopathic pathway is GTD-dependent, characterized by the inhibition of Rho-family GTPases, breakdown of the actin cytoskeleton, and eventual apoptosis. The cytotoxic pathway is GTD-independent, and is characterized by assembly and activation of the NADPH Oxidase complex and eventual necrosis. These pathways are described in greater detail below.

11

Figure 1.5: Mechanism of LCT function. A. Cytopathic effect common to all LCTs. Toxins ligate glycoproteins through interactions with their CROPS domains (green) as well as high-affinity receptors. Following endocytosis, a pH-dependent structural change of the delivery domain (yellow) allows for translocation of the autoprocessing (APD, blue) and glucosyltransferase (GTD, red) domains into the host cytosol. InsP6 allosterically activates the APD domain, which cleaves the GTD, which inactivates host GTPases, leading to downstream effects. B. Cytotoxic effects specific to TcdB. TcdB ligates receptors on the cell surface and endocytoses in a clathrin-dependent mechanism. Ligation of the high-affinity receptor also triggers assembly and activation of the NADPH Oxidase (NOX) complex, including membrane subunits NOX1 and p22, cytoplasmic subunits NOXA1 and NOXO1, and activated Rac1. Production of reactive oxygen species (ROS) results in lipid peroxidation, DNA damage, and ultimately cellular necrosis.

12 The cytopathic pathway Rho and Rac inactivation and subsequent disregulation of the actin cytoskeleton leads to loss of focal adhesion contacts at cell junctions (59-61). In the context of disease, this is probably the source of fluid efflux and diarrhea. When it was reported as the first substrate for TcdA and TcdB, RhoA was assumed to be the major target for the toxins and responsible for the phenotypes (53, 54). However, cells expressing a constitutively active Rac1, a mutant that cannot be glucosylated by TcdB due to the conformation of the target Thr35 in Switch-1, do not show the cytopathic rounding effects (62), while cells expressing RhoA with the corresponding mutation show the rounding phenotype, suggesting that Rac1 is the more relevant target. Furthermore, Rac1 modification by the LCTs leads to dephosphorylation of downstream targets such as PAK, leading to loss of focal adhesions (61, 63). In intestinal epithelial cells, TcdA-induced apoptosis involves cleavage of caspases 3, 6, 8, and 9 (64). Caspase-8 and -9 are fully activated by 18 h post intoxication, and caspase-6 is activated at this timepoint, with full activation by 24 h post intoxication(65). Caspase-3 activation was detectable by 24 h post intoxication, with full activation by 48 h. Specific inhibition of these caspases reduced DNA fragmentation (64), and activation of the caspase cascade is dependent on glucosyltransferase activity (66). Activated, cleaved Bid, a pro-apoptotic Bcl2 family member protein, is detectable at 6 h post intoxication with TcdA; however, this activation is independent of caspase activity (65). TcdA also induces cytochrome c release from the mitochondria detectable at 6, 18, and 24 h, indicating damage to the mitochondria (64), presumably due to Bid activation.

The cytotoxic pathway TcdB also activates a second pathway of cell death. In 2012, a former graduate student in our lab, Nicole Chumbler, published that TcdB is also capable of causing GTD-independent cell death in HeLa cells (67). She noted that the morphology of the cells at low concentrations of wild-type TcdB (10 pM) is similar to observations made by other researchers, that is, that the cells were rounded due to loss of actin cytoskeleton organization. However, at high concentrations (10 nM), the cells had lost membrane integrity, and all that remained visible was cellular debris. An LDH release assay was performed to test for cytoplasmic enzymes in the media supernatant, and an HMGB1 release assay to test for nuclear proteins in the cytosol.

13 Both assays were positive, showing breakdown of plasma and nuclear membranes, respectively, and she published a model that TcdB induces a necrotic cell death. Most researchers study TcdB at picomolar and lower concentrations, and the new study at higher concentrations generated controversy (68). However, a previous study done at Vanderbilt in 2010 quantified the C. difficile toxins in the stools of infected patients by ELISA and correlated the concentration of toxin with the severity of disease (69). For patients with moderate to severe CDI, the median concentration of toxin measured in stool was approximately 200 pM, while stools of 9 patients exhibiting no symptoms of C. difficile disease had TcdB concentrations of 0.1-3 pM. This finding was confirmed in 2015 by Song and coworkers, who were developing a digital ELISA for quantifying TcdA and TcdB in stools for diagnostic use (70). Their cutoff for “C. difficile negative” vs. “C. difficile positive” was determined to be approximately 0.1 pM for their assay, but the vast majority of their C. difficile positive patient samples were several thousand fold above this cutoff. We begin to see the effects of the cytotoxic pathway on tissue culture cells at 100 pM (Figure 2.2D&E), and with maximum effects detectable at about 10-fold higher concentrations. Considering the patients in the both the 2010 and 2015 studies were suffering from severe diarrhea, we can assume that the samples were diluted and the concentration of toxin at the site of infection was likely significantly higher. We therefore think the 0.1-1 nM concentrations of TcdB where we see necrosis occurring in cell cultures and tissue are relevant in the context of a physiological infection scenario. A post-doctoral fellow in our lab, Melissa Farrow, began to elucidate the mechanism of cell death by using siRNA to knock down various components of pathways that could lead to necrotic cell death (71). Knockdown of Rac1 but not RhoA or Cdc42 in both HeLa and Caco-2 cells led to cell survival. Rac1 is involved in the activation of the NADPH oxidase (NOX) complex (72), leading her to hypothesize that TcdB may function in generating a ROS response, and that this may be the mechanism by which the cells undergo necrosis. The NADPH oxidase (NOX) complex is a multi-subunit, membrane-bound complex that catalyzes the reduction of oxygen to superoxide using NADPH as a cofactor for catalysis (73). Originally identified in neutrophils as part of the oxidative burst of the phagosome, it is now known that many cell types express a NOX complex (74). The complex is formed by a membrane subunit NOX protein, another p22PHOX, two cytosolic components: the NOX activator 1 (NOXA1) and the NOX organizer 1 (NOXO1), and Rac1 (74). NOX1 is the major membrane NOX isoform expressed in the colonic epithelium (75). Activation

14 of the complex involves phosphorylation of NOXO1, which then stimulates the translocation of the cytosolic components to associate with the membrane-bound components NOX1 and p22PHOX (74). Rac1-GTP then binds to NOXA1 (76), which activates NOX1. NOXO1 and NOXA1 are each involved in an electron transfer chain that ultimately + results in the oxidation of NADPH to NADP and the reduction of O2 to superoxide. NOXA1 transfers electrons from NADPH to an FAD cofactor, and NOXO1 further moves those electrons from the flavin to NOX1, which catalyzes the reduction of O2 to superoxide. Farrow et al. showed that TcdB generates a NOX response in tissue culture cells using a fluorescent ROS indicator and H2O2 as a positive control. The ROS signal was specific to TcdB, as TcdA was not capable of generating such a response, and was independent of GTD activity, as a GTD mutant also generated a ROS response (71). In collaboration with the Goldenring lab at Vanderbilt, we developed a colonic explant model to study the effects of C. difficile toxins on mammalian tissue (67). Histological staining of tissue samples that have been treated with 10 nM TcdB shows that the surface layer of the epithelium is heavily damaged, while in the control sample the surface remains intact. Pre- treatment with an inhibitor of the NOX complex prevents TcdB-induced damage of the epithelium, supporting the mechanistic data and suggesting a new avenue toward treatment options (71). While an active Rac1 is necessary for activation of the NOX complex, it is also one of the primary targets of TcdB glucosylation and inactivation. The question arose of the timing of the activation cascade: whether Rac1 activated the NOX complex before TcdB was endocytosed and gluocosylated it. Farrow et al. performed a pull-down of active GTPases and using their affinity for effector proteins and quantified the amount of GTPase recovered using an enzyme-linked immunosorbent assay (ELISA). They observed a transient burst of GTPase activity before the inactivation that signaled the glucosylation event. More work is necessary to answer the new question of what is activating Rac1 in this early step.

Known information about TcdA and TcdB receptor binding The earliest report of a host factor that could mediate binding of a C. difficile toxin to the surface of a cell was by Krivan et al. in 1986 (36). These researchers showed that TcdA is capable of agglutinating rabbit erythrocytes in culture and that this agglutination is inhibited by excess glycans containing a sequence of Galα1-3Galβ1-4GlcNAc. Interestingly, in this study,

15 TcdA did not to agglutinate erythrocytes from several other species, and TcdB was incapable of agglutinating any erythrocytes tested. More work was done to identify the specific rabbit glycoproteins bound by TcdA. A glycoprotein expressed on rabbit ileal brush border was purified by fractionation of the tissue in 1996 and identified as sucrase isomaltase (77). TcdA binds this glycoprotein specifically on the carbohydrate side chains, and not via a protein-protein interaction. Furthermore, pre-treating tissue in an ileal loop model with a blocking antibody against the receptor protected the tissue from fluid secretion and inflammation induced by TcdA. In 1990, Tucker and Wilkins used an ELISA to screen for binding of TcdA to human glycoproteins (78). They identified several blood antigens, specifically antigens I, X, and Y, with affinity for TcdA. These glycans have similar compositions to those previously identified, and included Galβ1-4GlcNAc at their core. These antigens are present on the surface of the human intestinal epithelium and therefore would be available for binding during the course of CDI. Antigen X is also present on the surface of granulocytes, which are present in large numbers in pseudomembranes in patients with PMC. Finally in 2008, Na et al. reported a human receptor for TcdA on the surface of cultured human cells (79). They chemically crosslinked biotinylated TcdA to the surface of HT29 cells and isolated a complex of TcdA and a receptor on streptavidin beads. They identified the receptor as gp96 by mass spectrometry and were able to protect HT29 cells from TcdA- mediated cytopathic effects by using siRNA or a blocking antibody against gp96. The authors did not study the role of gp96 glycosylation, so the premise that TcdA could bind to a gp96- linked glycan rather than by a protein-protein interaction cannot be ruled out. The evidence that the TcdA CROPS domain is involved in cell surface binding is strong. In addition to the binding data and crystallographic evidence of glycan interactions (35, 37, 38), antibodies against the TcdA CROPS can neutralize toxins in tissue culture and in animal models (80), and excess TcdA CROPS can compete for TcdA binding to cultured cells (38). TcdA CROPS cannot compete for TcdB binding, again supporting the model that TcdA and TcdB bind to different receptors. However, TcdB CROPS does not compete for TcdB holotoxin binding, suggesting that the majority of TcdB binding occurs outside of the CROPS domain. At the time I began this work, no receptors for TcdB had been identified. Addressing this critical gap in knowledge was a primary motivation for the work presented in this dissertation. The general approach for identifying host factors important for TcdB intoxication is described in

16 Chapter II, and my identification of poliovirus receptor-like protein 3 (PVRL3) as a TcdB receptor is described in Chapter III. A few months prior to my publication of the PVRL3 result, Yuan et al. reported results from an independent study showing that chondroitin sulfate proteoglycan 4 (CSPG4) could serve as a TcdB receptor (42). Using an shRNA screen in HeLa cells, the authors assayed for cytopathic “rounding effects” at low concentrations of TcdB. Cells that remained flat, and therefore insensitive to toxin, were subjected to sequencing; two different shRNAs targeting chondroitin sulfate proteoglycan 4 (CSPG4) were identified as top candidates this screen. The authors used CRISPR/Cas9 to knock out CSPG4 in HeLa cells and showed that these cells were deficient in Rac1 glucosylation, cell rounding, and cell death at low concentrations of TcdB, all consistent with the TcdB-mediated cytopathic pathway (Figure 1.5A) and that this phenotype could be rescued when CSPG4 was expressed in the knockout background. At higher concentrations of TcdB, cell death was still observed in the CRISPR knockout cells, suggesting that at least one other receptor remained to be identified. The authors used recombinant protein fragments to narrow down the binding sites on both TcdB and CSPG4 by pull down. CSPG4 contains two laminin-G domains at the N- terminus, followed by several glycosaminoglycan repeats, a transmembrane helix, and a short cytoplasmic domain. An Fc-tagged construct of the two laminin-G domains is sufficient to pull down TcdB. Furthermore, rescue of the CRISPR-mutated CSPG4-/- HeLa cells with a CSPG4 construct that had the two laminin-G domains deleted was not possible. The binding site on TcdB was refined to amino acids 1500-1852. A GST-fusion protein of amino acids 1500-2366, including this domain and the CROPS, bound CSPG4, but a GST-fusion of the CROPS alone, residues 1852-2366, did not. Finally, Yuan et al. tested their receptor in a mouse intoxication model using NG2 (the rodent homolog of CSPG4) knockout mice. There was a dramatic increase in serum IL-8 5 h post intoxication in wild-type mice, which was reduced in NG2 knockout mice. However, the survival curves between wild-type and knockout mice were nearly indistinguishable. These data suggest that NG2/CSPG4 plays a role in the inflammatory response in response to systemic TcdB intoxication, but additional receptors are responsible for TcdB function in the context of disease. Another LCT from Clostridium perfringens, TpeL, lacks a CROPS domain (81). The low- density lipoprotein receptor-related protein 1 (LRP1) was identified as the primary receptor for TpeL by using a genetic screen in a human haploid cell line and assaying for cytopathic effects

17 (41). LRP1 knockout mouse embryonic fibroblasts (MEFs) showed impaired Ras glucosylation and toxin binding to the cell surface as compared to wild-type MEFs. Finally, the binding sites on both the receptor and the toxin were established to be Cluster IV on the heavy chain of LRP1 and the very C-terminus of TpeL. At the end of their paper from 2014, Schorch et al. formally put forth a model where LCTs can utilize multiple receptors and cell surface binding interactions. First, the toxin ligates the surface of the cell using the CROPS domain. The multiple repeats each bind a glycosylated moiety on the cell surface with a low affinity for any particular binding event, allowing for the toxin to associate and dissociate quickly, moving along the surface of the cell. For TcdB, this glycan binding would occur through an as-yet unidentified mechanism. Next, a high-affinity receptor interaction occurs outside of the CROPS region, triggering endocytosis of the toxin into the cell. As TpeL lacks a CROPS domain, this toxin would skip this first step and ligate its high- affinity receptor as a primary step. Schorch et al. identified the receptor-binding region of TpeL to be within the C-terminal 444 residues of the toxin, C-terminal of the delivery domain. Consistent with this model, CSPG4 binds TcdB in the C-terminal 350 residues of the delivery domain. This finding suggests that there may be a conserved receptor-binding domain only recently discovered for the LCTs. TcdB differs from the other LCTs in that it activates two pathways: the GTD-dependent cytopathic effect as well as the GTD-independent, NOX-activated cytotoxic pathway that is distinct from other LCTs. It is possible that the toxin uses different receptors to activate these different pathways. If so, the NOX-activated cytotoxic pathway receptor may not bind the same site in TcdB as that identified for the cytopathic pathway receptor CSPG4 (42). These are questions that will be addressed in the final chapter of Conclusions and Future Directions.

Poliovirus-Receptor Like/Nectin Protein Family

Discovery of PVR as the poliovirus receptor The cellular receptor for poliovirus (PV) was isolated and characterized using a genetic screen in which L cells expressing phage libraries were tested for susceptibility to poliovirus infection (82). Termed PVR or CD155, this 100 kDa, heavily glycosylated protein is a member of the immunoglobulin superfamily of proteins, as it has three extracelluar Ig domains. It also has a single transmembrane helix and a small cytoplasmic domain that differs among splice forms.

18 Expressing the CD155 gene in mouse L cells, which do not normally express CD155, renders these cells susceptible to poliovirus infection. Mutational analysis identified the N-terminal Ig domain (D1), encoded by residues 40- 136 of CD155, as the necessary domain for virus binding using L cells and assaying for virus infectivity (83, 84). Other mutations, including a deletion of the C-terminal cytoplasmic region, a deletion of the third Ig domain, or swapping this domain for an IgG domain, had no effect on viral entry (84). Deletion of the second Ig domain, amino acids 137-256, yielded a receptor that was deficient in allowing viral infectivity. This finding was due to conformational defects or blocking the availability of the D1 rather than a direct effect of deleting the D2; a longer linker between the D1 and the rest of the receptor restored activity, consistent with a model where the virus binds the D1, and that deletion of the D2 was inhibiting binding by causing spatial changes, moving the D1 closer to the cell surface and away from potential binding partners in the extracellular space. No cell surface binding of poliovirus is detected in cells expressing PVR constructs where the D1 or the transmembrane helix was deleted (83). Interestingly, deletion of the C-terminal cytoplasmic domain of PVR does not affect infectivity. The method of poliovirus uptake is still not fully understood, whether the virus enters through direct fusion at the plasma membrane or receptor-mediated endocytosis (85). Evidence in support of receptor-mediated clathrin-independent endocytosis is accumulating (86). How this occurs without signaling through the PVR cytoplasmic tail is unclear.

Other PVR-like (PVRL) proteins as viral receptors Alphaherpesviruses bind to the surface of a broad range of host cell types using five envelope glycoproteins: gB, gC, gD, gH, and gL (87). The initial event is mediated by gC binding to glycosaminoglycans, particularly heparan sulfate. This event is followed by gB, gH, and gL binding to the cell surface in quick succession, followed by binding of gD (88). Herpesvirus entry mediator (HVEM or HveA) was the first co-receptor identified for HSV (89, 90). A member of the tumor necrosis factor receptor family, HveA is only expressed on lymphoid cells, and therefore other receptors remained to be discovered. An expression screen using CHO cells was used to identify poliovirus receptor like-2 (PVRL2) as a co-receptor for herpes simplex virus-2 (HSV-2) and certain mutant strains of HSV- 1 (88). These cells express heparan sulfate to allow initial gC binding to the cell surface; the secondary event of gD binding to PVRL2 allows for internalization of the virus. Based on

19 sequence conservation, poliovirus receptor like-1 (PVRL1) and PVR were tested for their capacity to act as co-receptors for HSV (91). PVRL1 functions as a receptor for all viruses tested, including wild-type HSV-1. PVR failed to function as an HSV receptor but mediates entry for the related alphaherpesviruses porcine pseudorabies virus and bovine herpesvirus-1. Although PVRL1 and PVRL2 bind various alphaherpesviruses, the most well studied interaction is wild-type HSV-1 binding to PVRL1. As with the poliovirus-PVR interaction, the HSV-PVRL1 interaction is mediated primarily through the D1 of the receptor. An Fc-fusion of the D1 can compete for binding to the cell surface (92). Nonpermissive cells transformed with a construct expressing the D1 domain fused to the transmembrane helix, without the D2 or D3 domains, become permissive to herpesvirus infection, indicating that D1 alone is sufficient to allow infection (92). In fact, in contrast to the situation with poliovirus, the transmembrane helix is also dispensable (93). When soluble PVRL1 D1 was added to the medium of receptor- deficient CHO cells during an infection assay, a drastic increase in the number of plaques was noted when compared to the control plates. Two groups in parallel identified PVRL4 as the epithelial receptor for virus (MV) (94, 95). MV infection begins when virus particles enter the host and the hemagglutinin (H) protein interacts with receptors on cells in the airway epithelium. This binding event then triggers the viral F protein to induce fusion of the viral membrane with the host cell membrane (96). Laboratory-adapted and vaccine strains of MV use the complement factor CD46 as a receptor on epithelial cells (97, 98) and signaling lymphocyte activation molecule (SLAM, or CD150) as a receptor on lymphoid cells (99, 100). Wild-type and clinical isolates of MV also use SLAM as a receptor, but do not use CD46 to enter airway epithelial cells. Both groups screened multiple laboratory cell lines for susceptibility to MV infection, and by microarray, established the upregulated gene products in common among the susceptible cell lines. Of gene products chosen for further study, only transfection of PVRL4 into nonpermissive cells made these cells permissive to MV infection; transfection of other PVR family proteins did not act as functional receptors for MV (94). Knockdown of PVRL4 by siRNA (94, 95) or blocking the interaction of the virus with the receptor with antibodies against D1, but not D2 or D3 (95), inhibited infection. A D1-Fc fusion protein also blocked infection (95). High-resolution structures have been solved for several viral glycoprotein-receptor complexes: poliovirus bound to PVR, by cryo electron microscopy (101, 102), and HSV gD bound to PVRL1 (103) and MV hemagglutinin bound to PVRL4 (104), both by x-ray crystallography. Comparisons of the structures reveal that all three viruses bind to the

20 membrane-distal Ig domain (D1) of their respective receptors and make contact with the C’ strand, C’-C’’ loop, and also the F-G loop. Mutational analysis of the F-G loop shows that a conserved phenylalanine residue is required for binding of MV and herpesvirus to PVRL4 and PVRL1, respectively, but not required for binding of PV to PVR. This same residue is required for dimerization of the PVR proteins and formation of cell junctions (see below) (105). HSV and PV also interact with their receptors using the C, C’, and C’’ strands, effectively making contact with one entire face of the Ig β-sandwich. Mutations in these strands, particularly in the C’ strand, reduce binding of the virus to the receptor and infectivity of the virus in cultured cells (102, 103). The overall structure of the MV receptor-binding pocket is shaped to accommodate several receptors, including CD46, SLAM, and PVRL4 (104). This may explain why there are fewer specific contacts made between MV and PVRL4 compared with the other virus-receptor interactions, and why most are made on the tips of the loops. Nonetheless, several specific interactions were identified in the crystal structure and confirmed by mutational analysis, including the entire F-G loop and C’ strand, as well as residues Q30-Q33 of the B-C loop.

PVR proteins as adherens junctions molecules The PVR family of proteins, collectively also referred to as nectins, are normally localized to the plasma membrane at adherens junctions (AJs), as their normal role is in junction formation (106). They form Ca2+-independent trans-hetero-dimers in various combinations as the first step in cell-cell contact. The formation of an AJ between two moving cells begins when the cells first collide. PVR and PVRL3 at the leading edges of each cell make contact, forming trans-dimers (107). These dimers are short-lived, and as they dissociate, PVR is endocytosed from the surface in a clathrin-dependent manner (108), leaving PVRL3 free to associate with PVRL1, forming a nascent AJ (109). The nectin proteins associate with the F-actin binding protein afadin at their C-terminus. Afadin then recruits α- and β- catenins to the growing junction, bringing with them E-cadherin, establishing the finished AJ. The PVRL1-PVRL3 heterodimer is the strongest interaction of any combination, measured at 2.3 nM (110). PVRL1 also interacts with PVRL4; PVRL3 interacts with PVR and PVRL2 (105, 110). These interactions, as with the interactions with viruses, are mediated through the N-terminal Ig domains (105). A conserved phenylalanine on the F-G loop on one monomer intercalates into a hydrophobic pocket formed by a conserved glycine on the C’-C’’

21 loop on the opposing monomer. The backbone α-carbon of the glycine (Figure 1.3, cyan) forms the base of the pocket, and a hydrophobic residue on either side form the rest of the pocket. Mutating the F-G phenylalanine prevents dimer formation both in cell culture and in crystal contacts.

Figure 1.6: Crystal structure of the PVRL1 dimer. Individual monomers are in green and blue. Contacts are made through the N-terminal Ig domains (D1, indicated). Right, zoom of the critical D1 dimer interactions. Phe129 from one monomer fits into a hydrophobic pocket formed by the alpha carbon of Gly86 (cyan) and the side chain of Met85 from the second monomer. PDB ID 4FMF (105)

Other roles for nectin proteins PVRL1 and PVRL3 are involved in several signaling pathways targeting cell growth and differentiation. PVRL3 interacts with αvβ3 , both the active and inactive conformation. Ligation of PVRL3 activates integrin signaling, phosphorylating FAK and activating Srk, leading to activation of Rac1 (111) (112). PVRL3 is also involved in PDGF-dependent anti-apoptotic signaling regulated through the PI3K-Akt pathway (Kanzaki 2008). Interactions with PDGFR and

22 VEGFR activate Rho and Ras- family GTPases (112-114), activating several signaling cascades. PVRL1 and PVRL3 knockout mice are viable, although they are blind due to a malformation of the contact sites between the pigmented and non-pigmented epithelial cell layers in the eye (115). Additionally, PVRL3-null males are infertile; proper spermatid formation relies on PVRL3-PVRL2 contacts between spermatids and Sertoli cells in the testis (116). One known human mutation has been reported as a translocation at 46,XY,t(1;3)(q31.3;q13.13), or a breakpoint roughly 500kb upstream of the pvrl3 locus, and gave rise to a phenotype of greatly reduced PVRL3 expression (117). The patient presented with severe congenital cataracts and developmental delay. A karyotype revealed the mutation to be a translocation of one arm of 1 and one arm of chromosome 3. When testing expression levels of genes upstream and downstream of both breakpoints on 1 and 3 by RT-PCR, only pvrl3 transcript levels were affected by the translocation. Scans of the patient’s eyes were very similar to the studies conducted with pvrl3-null mice, strengthening the link between PVRL3 mutations and ocular defects.

Research Objectives When I began this work, little was known about the host factors required for TcdB function in cells. The toxin required an acidified endosome and InsP6 for the GTD to access its substrates, Rho family GTPases. TcdB activated two independent cell death pathways: one caspase-dependent, the other caspase-independent (118). There were many gaps in our knowledge of how TcdB intoxication affected cells, including the receptors ligated, the endocytic pathway used to gain access to the interior of the cell, and the mechanism of caspase- independent cell death. We used a genomic screen in a human epithelial cell line to identify genes required for TcdB cytotoxicity in an effort to understand the function of TcdB in cells. In Chapter II, I describe the gene-trap method of mutagenesis and whole-genome screening in the Caco-2 intestinal epithelial cell line. The gene-trap is a method of insertional mutagenesis in which a viral vector inserts randomly into the mammalian genome. The vector used here has a promoterless neomycin resistance marker used to select for vectors that have inserted into actively transcribed genes. The stop codon of the neomycin resistance marker then disrupts expression of the mammalian gene.

23 I used two assays to measure cell viability: Cell-TiterGLO and Cell-ToxGLO. Cell- TiterGLO assays for ATP levels in the cell, as when a cell dies, the ATP is quickly hydrolyzed. Normalizing the ATP levels of intoxicated cells to unintoxicated cells gives a percentage of viable cells in the plate well. Cell-ToxGLO assays for the release of cytosolic enzymes into the media supernatant, signifying the loss of plasma membrane integrity, which is a hallmark of necrosis. Follow up experiments on candidate genes from the gene-trap screen include other methods of genetic disruption such as knockdown of mRNA by RNAi and knockout at the DNA level with CRISPR/Cas9, pull-downs with purified proteins to show direct interactions, and Rac1 glucosylation assays to follow entry of the toxins into cells. We have access to an antibody that recognizes the Switch-1 region of Rac1 that we can use for immunoblots. When TcdA or TcdB enters the cell and modifies Rac1, the antibody can no longer recognize the epitope and the band is lost on the immunoblot. We also have an antibody that recognizes a different epitope that we can use as a total Rac1 control. Of the 58 sequenced clones from the gene-trap screen, we identified 45 gene products that, when disrupted, decrease TcdB-mediated cytotoxicity. We show that Clone D02, disrupted in the AP2, is resistant to TcdB, but sensitive to TcdA similar to wild-type Caco-2. This is contrary to published reports, and is the subject of ongoing work in the lab. We also show that galectin-3 is required for full potency of TcdB and binds directly to TcdB. In Chapter III, we show that PVRL3 functions as a receptor for TcdB in epithelial cells. Disruption of PVRL3 expression in Caco-2 or HeLa cells by RNAi or CRISPR/Cas9 resulted in cell survival. TcdB interacted with PVRL3 directly by pull-down, and inhibition of the TcdB- PVRL3 interaction extracellularly with either a blocking antibody or purified PVRL3 ectodomain also resulted in cell survival. In line with published reports (42), knockdown of PVRL3 in HeLa cells, which express CSPG4, did not inhibit the cytopathic “rounding” effects; however, in Caco- 2 cells, which do not express CSPG4, disruption of PVRL3 resulted in impaired Rac1 modification and cell rounding. Results from the whole genome screen give insight into the mechanism of TcdB intoxication of cells and the host factors necessary for toxin action. Future work on these factors will delineate the pathways used by TcdB and add to the understanding of C. difficile pathogenesis.

24 CHAPTER II GENE-TRAP SCREEN TO IDENTIFY HOST FACTORS INVOLVED IN TCDB-MEDIATED CYTOTOXICITY

Introduction TcdB intoxication of cells is a multi-step process involving ligation of host cell receptors, activation of the endocytic pathway, pore formation in the endosomal membrane, autoproteolysis triggered by inositol hexakisphosphate (InsP6), and glucosylation of host cell GTPases. With the exception of the endosomal proton pump, InsP6, and Rho family GTPases, no other host factors had been identified as being critical to TcdB function at the beginning of this work. The Cover and Rubin labs at Vanderbilt published results from a collaboration in which they used gene-trap mutagenesis to identify host factors involved in C. perfringens ε-toxin pathogenesis (119). We used a similar approach to identify host proteins and pathways that function in TcdB pathogenesis, and therefore established a collaboration with Dr. Rubin to perform a gene-trap screen in a colonic epithelial cell line. This chapter describes the results from that screen. We performed detailed follow-up on one gene product, PVRL3, which will be described in Chapter III of this dissertation. We also performed follow-up work on several other candidates from the screen, including the clathrin adapter protein AP2, and the family of galectin proteins including galectin-3 and galectin-3 binding protein. Findings show that AP2 is required for TcdB cytotoxicity and entry into cells, consistent with published data, but not for TcdA cytotoxicity, which is inconsistent with published data. Galectin-3 is required for TcdB cytotoxicity as shown by siRNA knockdown; knockdown of other galectin genes did not show this phenotype. Purified galectin-3 binds to TcdB by immunoprecipitation, suggesting that TcdB interactions with galectin-3 are required for full cytotoxic potency of TcdB. Future work on other candidates from the gene-trap screen will provide further insight into mechanism of action of TcdB.

Results

Primary Screen: Gene-Trap In collaboration with Jinsong Sheng and Donald Rubin, we performed a gene-trap screen using the human colonic epithelial cell line Caco-2 to identify cellular factors required for TcdB-

25 mediated cytotoxicity. Cells were mutagenized with a retroviral gene-trap vector that inserts a promoterless neomycin gene at random locations throughout the genome (120). Insertions that occur in actively transcribed regions are expected to confer resistance to neomycin (G418) and prevent expression of the disrupted gene. A library of G418-resistant Caco-2 cells was challenged with three successive treatments of 15 nM TcdB. Surviving cells were propagated from single clones, and the genomic DNA was sequenced with vector-specific primers to identify the location of the genetic disruption. We sequenced 58 clones and identified 45 loci where insertions occurred. Results are presented in Table 2.

Table 2: Gene-Trapped Genes Identified as Involved in TcdB-Mediated Cytotoxicity Clone ID Integration Site Gene Gene Gene Description RNAi ID Name Fold- Survival Plasma Membrane and Extracellular Matrix A04 NM_032532 84624 FNDC1 Fibronectin III Domain -1.1 Containing 1 B02 NM_015393 25849 PARM1 Prostate Androgen -1.2 Regulated Mucin-like 1 B09 NM_001042507 653499 LGALS7A/B Galectin-7 1.1 B11 NM_005567 3959 LGALS3BP Galectin-3 Binding Protein 1.7 C02.1 NM_004467 2267 FGL1 Fibrinogen-like 1 1.3 C02.2 NM_001144663 1015 CDH17 Cadherin 17 1.3 D05.1 NM_014220 4071 TM4SF1 Transmembrane 4 1.2 superfamily 1 D05.2 NR_029829 406970 miR194-2 microRNA involved in E- Not tested cadherin expression E05 NM_014887 10443 N4BP2L2 NEDD4 BP 2 like 2 1.7 E08, E19 NM_015480 25945 PVRL3/ Poliovirus receptor-like 3, 1.7 Nectin3 Nectin3 E10 NM_002306 3958 LGALS3 Galectin-3 1.6 E11 NM_001145211 11309 SLCO2B1 Solute carrier 2B1 1.0 Endocytosis and Vesicle Trafficking D02, E04 NM_004069 1175 AP2S1 AP2 Adaptor molecule 1.8 D05.2 NM_006795 10938 EHD1 EH domain containing 1 1.4 D11 NM_001077206 22872 Sec31A Sec31 homolog, Secretion 1.1 D13 NM_004859 1213 CLTC Clathrin Heavy Chain 1.3 F03.2 NM_002867 5865 Rab3B Rab3B, RAS oncogene 1.0 family member F10 NM_014426 27131 SNX5 Sorting nexin 5 1.4 Cell Cycle and Apoptosis B02, C03, NM_001657 374 AREG Amphiregulin -1.1 D14 B04.1 NM_015161 23204 ARL6IP1 ADP-ribosylation factor- 1.3 like 6 Interacting Protein 1

26 B09 NM_144691 147968 CAPN12 Calpain12 1.1 C04.1 NM_004060 900 CCNG1 Cyclin G1 1.2

Gene Expression A04 NM_031924 83861 RSPH3 Radial spoke 3 homolog -1.1 A07, B10 NM_006236 5455 POU3F3 Pou class 3 homeobox 3 1.6 B01 NM_005257 2627 GATA6 GATA-6 1.0 B04.1 NM_001019 6210 RPS15A Ribosomal Protein Small -1.1 15A B04.2 NM_024116 79101 Taf1d TATA box Binding Protein -1.1 Assoc. Factor 1D B07 NM_018442 55827 NRIP Nuclear receptor interaction -1.1 protein C04.2 NM_004902 9584 RBM39 RNA binding motif 39 1.1 C06 NM_030627 80315 CPEB4 Cytoplasmic polyadenylation 1.0 element binding protein 4 (D12 D06) NM_024320 79170 PRR15L/ Proline-rich region 15-like 1.0 E15 ATAD4 E10 NM_144578 93487 MAPK1IP Mitogen activated protein 1.2 1L kinase 1 interacting protein 1-like F03.1 NM_004147 4733 DRG1/ Developmentally 1.4 NEDD3 regulated GTP BP1 F09 NM_001037283 8662 eIF3B Eukaryotic initiation factor 1.1 3B B08, E17 NM_003525.2 8346 HIS1H2BI Histone cluster 1, H2B -1.5 E12 NM_005520 3187 hnRNPH1 Heterogeneous nuclear -1.3 ribonucleoprotein H1 Other / Unknown B07 NM_001143674 25874 MPC2 Mitochondrial Pyruvate -1.4 Carrier 2 C05 NM_000371 7276 TTR Transthyretin -1.2 C09 NM_001122 123 PLIN2 Perilipin 1.1 D01 NM_001024956 6309 SC5DL Sterol-C5-desaturase-like 1.1 D05.2 NR_029578 406967 miR192 microRNA Not tested D05.2 XM_002342818 645960 Hypothetical Hypothetical similar to RL16 Not tested E03 NW_001838769 Unknown F12 hmm2826564 unknown Conserved unknown gene Not tested F13 NM_004134 3313 HSPA9B HSPA9 heat shock 70kDa Not tested protein 9 Identical insertions indicated enclosed in parentheses. Independent insertions indicated with a comma. Bold typeface indicates survival in RNAi secondary screen in Column 6 to be 1.2-fold or higher above controls.

We identified 12 gene products that localize to the cell surface or extracellular matrix including three in the galectin family of proteins, which are involved in glycan binding, six involved in clathrin-mediated endocytosis and vesicle trafficking, four involved in cell fate

27 determination, 14 in chromatin maintenance and gene expression, and nine with other or unknown function. Five additional clones (C09.1, B01.1, E09, C05.1, and D16) proved to be repetitive DNA sequences and could not be mapped to discrete locations in the genome. The insertion in clone E10 fell in the non-coding region between the genes encoding galectin-3 and MAPK1IP1L. The insertion in clone A04 fell in the non-coding region between the genes encoding FNDC1 and RSPH3. The insertion in clone B09 falls in the intergenic region between LGALS7 and CAPN12. The insertion in clone B02 fell in the non-coding region between the genes encoding amphiregulin and PARM1, in the promoter region of PARM1 and downstream of the coding region of amphiregulin. Although we kept PARM1 on the list as potentially disrupted by the gene-trap insertion, two other clones were also disrupted in different locations within the amphiregulin gene, suggesting this is a relevant target. MPC2 and IQWD1 are transcribed in opposite directions but the genes overlap. The gene-trap insertion lies within this overlapping region, disrupting both. Caco-2 cells express many adhesion proteins and therefore were difficult to separate during the cloning period. B04, C02, C04, D05, and F03 proved to be mixtures of two clones each; these clones were re-designated as B04.1, B04.2, C02.1, C02.2, C04.1, C04.2, D05.1, D05.2, F03.1, and F03.2. The insertion in clone B04.1 disrupted a non-coding region between the genes encoding RPS15A and ARL6IP1. The new clone designated as D05.1 harbors an insertion in the TM4SF1 gene, while D05.2 is mutated in EHD1. The EHD1 gene locus also encodes two microRNAs, miR192 and miR194-2, both of which would also be disrupted by the gene-trap insertion.

Secondary Screen: RNAi viability assays To narrow this list and choose candidates for future study, Stacey Seeback, Dr. Melissa Farrow, and I used RNAi to knock down each of the genes on the list in HeLa cells. After 48 h of treatment with the interfering RNAs, we challenged the cells with TcdB, normalized the viability signal to that of an unintoxicated control, and calculated fold change compared to a control RNAi. All small RNAs were siRNAs except when testing PVRL3, which was found to have a relatively long protein half-life and required the use of constitutively expressed short hairpin RNAs. Results are reported in the final column of Table 1. Genes for which knockdown resulted in 1.2 fold or greater survival as compared to the control RNAi are notated in bold.

28 Fourteen gene products had a 1.2-fold or greater survival as compared to controls. This shorter list includes clathrin heavy chain and its adaptor protein AP2 as well as many of the vesicle trafficking proteins: sorting nexin 5, NEDD3, and EHD1. Galectin-3 and its associated binding protein are on this list as well, although galectin-7 is not. Also, MAPK1IP1L, the other gene possibly disrupted in the E10 clone along with galectin-3, did not make the cutoff. Finally, several membrane and extracellular proteins, including PVRL3, cadherin-17, fibrinogen-like1, and transmembrane 4 superfamily 1 make up the rest of the list. I decided to pursue PVRL3 in more detail, the data for which is in Chapter III of this thesis, as well as clathrin/AP2 and the galectin3/galectin3BP genes.

AP2 is required for TcdB endocytosis Soon after we completed the initial screening in 2010, Papatheodorou et al. reported that TcdB enters cells using a clathrin-mediated mechanism of endocytosis (43). Two of our clones had mutations in clathrin heavy chain and the clathrin adaptor protein AP2 (clones D13 and D02, respectively), consistent with this report. Clone D13 was non-viable from the liquid nitrogen stocks, and we were therefore unable to perform subsequent experiments using these cells; however, D02 was viable and subjected to further study. We were unable to detect noticeable differences in expression of AP2 by conventional RT-PCR compared with GAPDH loading control (data not shown). This could indicate only a slight defect in gene expression, which was observed in other clones (see clone E19, Chapter III of this thesis), but which nonetheless could have a significant phenotypic effect. Wild-type and D02 Caco-2 cells were plated and challenged with serial dilutions of TcdA and TcdB (Figure 2.1A). Consistent with published data, D02 cells were significantly resistant to TcdB across all toxin concentrations tested. Interestingly, however, D02 was as sensitive to TcdA as wild-type Caco-2 cells. This is in contrast to published data by Papatheodorou et al. and is the focus of ongoing thesis work of another graduate student in the lab.

29

Figure 2-1: AP2 is required for TcdB cytotoxicity and entry into cells, but not TcdA intoxication of cells. A. Wild-type Caco-2 and Caco-2 D02 cells with a disruption in the AP2 locus were plated and challenged with TcdB and TcdA. After 18h, viability was read for the cells challenged with TcdB, and after 48h, viability was read for those challenged with TcdA. Signal was normalized to a no-toxin control for each cell type and toxin. Results are mean ± SEM of six experiments. **P < 0.01, ***P < 0.005; P values relative to wild-type cells. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test. B. Wild-type and D02 cells were plated and shifted to 4°C to inhibit endocytosis. Cells were intoxicated with TcdB for 1h at 4°C. One replicate plate of each cell type was lysed (0 min), and a second was shifted to 37°C to allow toxin entry into cells for 45 min before lysate was harvested. Immunoblots of whole-cell lysates were probed for TcdB, unglucosylated Rac1, and GAPDH as a loading control.

Next, I assayed whether D02 cells had a defect in TcdB entry into cells. I plated wild- type and D02 Caco-2 cells, and 24 h later chilled the plates to 4°C to inhibit endocytosis. I intoxicated the cells with TcdB, lysed one plate of each cell type, shifted a second replicate plate to 37°C to allow endocytosis to occur, lysed the cells after 45 minutes, and immunoblotted whole-cell lysates (Figure 2.1B) for Rac1 and TcdB. TcdB undergoes autoproteolysis once it enters the cell, and the cleaved N-terminus is free to modify host GTPases including Rac1. We have reagents in the lab to probe immunoblots for multiple toxin domains, which we can then use to determine the cleavage state of the toxin, and an antibody specific to the Switch-1 region of Rac1, which is the target of TcdB glucosylation. When TcdB modifies the GTPase, the antibody can no longer recognize its epitope, and the band is absent on an immunoblot. I observed that at 4°C, there is a comparable amount of TcdB bound to the surface of both wild-type and D02 Caco-2 cells, indicating that the difference in viability seen in Figure 2A is not due to a binding defect (Figure 2.1B, lanes 1 and 2). However, at 37°C, TcdB holotoxin signal is lost in the wild-type Caco-2 lane, indicating that the toxin has entered and is cleaved. It

30 is still present in the D02 lane, suggesting that the toxin is still on the surface of the cell and has not been endocytosed. In wild-type cells at 37°C, there is no signal for Rac1, indicating that the TcdB glucosyltransferase domain has entered the cytosol and has modified its substrate Rac1, preventing the antibody from binding the epitope. However, Rac1 modification is inhibited in D02 cells, consistent with the model that TcdB requires clathrin-mediated endocytosis to enter cells.

Galectin-3 is required for TcdB cytotoxicity We also chose to follow up on the cluster of three galectin genes from Table 2. Galectins are two-domain proteins, with a carbohydrate recognition domain at the C-terminus and a collagen-like multimerization domain at the N-terminus (121, 122). Galectin-3 (clone E10), galectin-3-binding protein (clone B11), and galectin-7 (clone B09) were all candidates in our screen. Galectin-7 is not normally expressed in the human colon, and in fact is only normally expressed in keratinocytes (123-125), so although it was interesting in that it may have been expressed in tissue culture, we chose not to study it further. Galectin-3, on the other hand, is ubiquitously expressed, including in the human colon (126). Therefore, we chose to follow this potentially interesting target gene. Disruption of galectin-3 expression either by gene-trap mutagenesis or siRNA in Caco-2 cells resulted in significant cell survival as compared to control cells (Figure 2.2A and B). Disruption of the galectin-3-binding protein locus by gene-trap mutagenesis (clone B11) had a similar effect on viability (Figure 2.2A), although expression levels of galectin-3 or galectin-3- binding protein were not assessed by immunoblot or RT-PCR in this clone. By immunoblot, we saw a faint band in the E10 and lgals3 siRNA-treated cell lanes (Figure 2.2B), but RT-PCR confirmed complete absence of lgals3 transcript in the E10 clone (Figure 2.2C). This finding suggests that the galectin-3 antibody displays cross-reactivity with another galectin protein expressed in Caco-2 cells.

31

Figure 2.2: Galectin-3 is required for TcdB-mediated cytotoxicity in Caco-2 and HeLa cells. A. Caco-2, Caco-2 E10, Caco-2 B11, and Caco-2 transfected with either siRNA against the lgals3 transcript or a control siRNA were seeded and treated for 18 h with serial dilutions of TcdB. For each cell line, ATP was measured as a readout of viability and normalized to signal from untreated cells. Results represent the mean ± SEM from five experiments. B. An Immunoblot of whole-cell lysates of cells used in A, except B11, was probed for galectin-3 or GAPDH (loading control). C. RT-PCR on wild-type Caco-2 or Caco-2 E10 total RNA for lgals3 transcript or gapdh transcript (loading control). D. HeLa cells were transfected with siGENOMETM siRNAs against the indicated genes and 48 h later challenged with serial dilutions of TcdB. Viability was read and plotted as in A. Results represent the mean ±SEM of six experiments. E. HeLa cells were transfected with pools of ON-TARGETplusTM siRNAs against the lgals3 transcript or a control siRNA and challenged with serial dilutions of TcdB. Viability was read and plotted as in A. Results represent the mean ± SEM of six experiments. F. Deconvolution of the pool of four siRNAs used in E, transfected individually into HeLa cells. Viability was read and plotted as in A. G. Immunoblot of whole-cell lysates from cells used in E. Viability assays show *P < 0.05, **P < 0.01, ****P < 0.001; P values relative to wild-type cells are shown in A and control siRNA in D, E, F. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test.

32

Galectin-3 is a member of a family of 14 soluble, galactoside-binding lectins, and since galectin-7 was also identified in our original screen, we determined whether any other galectin protein family members had effects similar to those of galectin-3. I tested nine other galectins using siRNA knockdown in HeLa cells. Only knockdown of galectin-3 showed robust protection against intoxication with TcdB (Figure 2.2D); knockdown of other galectin family members showed no protection, or, in some cases, made the cells more sensitive to TcdB-mediated cell death (galectin-8 and galectin-13). Galectins 5, 6, and 11 do not exist in the , and galectin-10 and galectin-14 were not tested in this experiment. The siRNAs used in panels 3A, B, and D were from Dharmacon’s siGENOME generation of products, which has off-target effects. Using the more specific ON-TARGETplus siRNA library, we found that galectin-3 siRNAs completely knocked down protein expression by immunoblot (Figure 2.2G) and provided protection in HeLa cells from intoxication by TcdB (Figure 2.2E). The galectin-3 siRNA used in this experiment was made up of a pool of four sequences purchased from Dharmacon. I deconvoluted the pool and tested each sequence individually; all four siRNAs tested positive for protection against TcdB intoxication (Figure 2.2F). I then tested another cell line to determine if I could rescue a negative phenotype. BT549 cells are a breast cancer epithelial cell line that lacks expression of galectin-3 and is naturally resistant to TcdB (Figure 2.3A). I transduced a lentivirus containing the gene for galectin-3 into these cells and isolated three clonal cell populations that overexpressed galectin-3 (Figure 2.3B); each of these clones was sensitive to TcdB compared to the wild-type parental cells.

33

Figure 2.3: Effects of galectin-3 on TcdB A. Wild-type BT549 cells and three clones harboring an overexpression vector were plated and challenged with serial dilutions of TcdB. Viability was measured and plotted as in Figure 2.3. Viability assays show *P < 0.05, **P < 0.01, ****P < 0.001; P values relative to wild-type cells. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test. B. Immunoblot of whole-cell lysates of cells used in A probed for galectin-3 and GAPDH (loading control). C. Purified recombinant galectin-3 and TcdB were mixed at an equimolar ratio. The reaction was split in two: one anti-TcdB IP was performed and one anti-galectin-3 IP was performed. D. HeLa cells (galectin-3 positive) and BT549 cells (galectin-3 negative) were plated side-by-side and allowed to equilibrate to 4°C to inhibit endocytosis. Cells were intoxicated with TcdB for 1 h at 4°C. One replicate well was lysed, and a second replicate well was shifted to 37°C for 45 min to allow endocytosis and processing of the toxin. Immunoblots of whole cell lysates were probed for the TcdB GTD and GAPDH (loading control).

Because galectin-3 has affinity for the same types of glycans that TcdA is known to bind (N-acetyllactosamine derivatives, such as Galα1–3Galβ1–4GlcNAc) (36, 37, 127), and because

34 TcdB had long been hypothesized to bind glycosylated proteins or lipids but an interaction had not been demonstrated, I hypothesized that galectin-3 functions as a scaffold between TcdB and its glycan binding partner. To test if TcdB can bind directly to galectin-3, I cloned, expressed, and purified recombinant galectin-3 and performed a co-immunoprecipitation with purified recombinant TcdB. In Figure 2.3C, I precipitated a complex containing both galectin-3 and TcdB, showing that these proteins do indeed interact. After showing that TcdB and galectin-3 formed a complex, I attempted to determine whether there was a binding defect on cells using wild-type BT549 cells. Because these cells have a defect in galectin-3 expression, I reasoned that the difference in viability might be due to a defect in TcdB binding to cells. To test this hypothesis, I plated BT549 cells in parallel with HeLa cells, which bind TcdB very well. I shifted the cells to 4°C to inhibit endocytosis and intoxicated the cells with 10 nM TcdB for 1 h. I lysed one well, shifted a second to 37°C for 45 min to allow endocytosis of the toxin, and then lysed the cells. Immunoblotting for the toxin revealed that there was no defect in the capacity of BT549 cells to bind toxin, as I had hypothesized (Figure 2.3D, lanes 1&2). However, there may have been a defect in the capacity of the cells to endocytose the toxin, as there was no processed toxin in the BT549 lane compared with the HeLa lane at 37°C (Figure 2.3D, lanes 3&4). This experiment was performed before the BT549 cells harboring the galectin-3 overexpression vector were made; repeating this experiment with a timecourse assaying for toxin entry could shed some light on the role of galectin-3 in TcdB-mediated cytotoxicity, since as of now its role is still unknown.

Discussion Here we describe a method for identifying host factors involved in toxin pathogenesis in mammalian cells. Using a viral insertion mechanism of mutagenesis, we generated a library of stable genetic mutants that can be screened phenotypically for mammalian genes and pathways involved in any number of different areas of study. Some of the candidates identified in the gene-trap screen, when knocked down by siRNA in HeLa cells during our secondary screen, had little effect, or in some cases, a negative effect on cell viability when challenged with TcdB. There could be several reasons for this observation. The first is that in the beginning, we left genes on the list when we were unsure which gene was affected by the gene-trap insertion. For example, clone E10, harbored an

35 insertion in the intergenic region between two genes: galectin-3 and MAPK1IP1L. Follow up analysis with RNAi confirmed that galectin-3 was the relevant target in this clone. Another reason is that a different subset of genes could have been expressed in HeLa cells, the cell type in which we performed the siRNA screening, as opposed to the gene-trapped line Caco-2. If a gene were not expressed in HeLa cells, we would not expect phenotype, which would explain a relative value of 1 in our secondary screen. Due to the relatively large number of genes on our list, we did not perform RT-PCR analysis on every gene to confirm expression in HeLa cells before beginning our secondary screen. The major reason why the siRNA data differed from the gene-trap data is probably that the gene-trap disruptions were rarely complete knockouts. As mentioned above when describing the D02 clone, and as I will discuss again in Chapter III when describing the E19 clone, many of the mutants displayed only a partial loss of expression. The gene-trap library was made using a diploid cell line, and several of the clones were heterozygous. Many had mutations in the promoter region, including the galectin-3 and PVRL3 mutants, leaving their coding regions intact. Partial losses of function in genes like histones and transcription factors may have allowed these mutants to be viable long enough for transcriptional profiles to adapt and upregulate other chromatin remodeling and transcription factors to take their places during the several weeks while the gene trap screen library was propagating. Perhaps those that were upregulated happened to be beneficial to the cell under conditions of oxidative stress, such as the aberrant NADPH oxidase signaling that TcdB activates. Therefore, these mutants survived when the gene-trap libraries were tested. However, siRNA experiments are done on a much faster time scale—only 72 h—and at complete loss of expression. There is no time for the cells to adapt under these conditions. Two of the siRNAs that had the most negative effect on viability were a histone and an hnRNP siRNA. Completely removing a histone or an hnRNP from the cell, without allowing time for adaptation, would be disruptive to homeostasis. To then challenge the cell with a toxin known to activate oxidative stress pathways could lead to even more DNA damage and cell death at a faster rate. One interesting result that came from our work was the fact that clone D02, mutated in the clathrin adaptor protein AP2, is not resistant to TcdA. Papatheodorou et al. found that both TcdA and TcdB enter cells using a clathrin-mediated mechanism of endocytosis (43). Our data

36 in Figure 2 support their model for TcdB, but ongoing studies in the lab are investigating the mechanism of why our result for TcdA is not consistent with their published data. Entamoeba histolytica infection of the liver is dependent on expression of galectin-1 and galectin-3, the mechanism of which begins with galectin-1 mediated-E. histolytica adhesion to liver endothelial cells, followed by full invasion of the tissue mediated by galectin-1 and galectin- 3 interactions on endothelial and epithelial cells(128). Although we show that the mechanism of galectin-3 action is not solely by mediating binding of TcdB to the cell surface (Figure 2.3), it may be acting by linking co-receptors to signal for endocytosis to occur. Some viruses are known to bind proteinaceous and glycan receptors simultaneously to signal for endocytosis. For example, HSV-1 binding to host cells first involves viral glycoproteins gB and gC to heparan sulfate, followed by gD binding to PVRL1 or PVRL2 (88, 91). It is possible that the LCTs follow a similar two-receptor mode of entry. While the TcdA CROPS domain could be responsible for glycan binding, the TcdB CROPS has not yet been shown to have the capacity for carbohydrate binding, despite its homology to TcdA. We show that TcdB can bind directly to galectin-3, and galectin-3 binds the same carbohydrates as TcdA. An open question is where the galectin-3 binding protein fits, as the gene-trap clone B11 exhibits the same resistance to TcdB as the E10 galectin-3 mutant.

Methods Cell culture Caco-2 cells were maintained in MEM supplemented with 10% FBS + 1% non-essential amino acids. HeLa cells were maintained in DMEM + 10% FBS. BT549 cells were maintained in DMEM/F-12 supplemented with 10% FBS. BT549 cells were transduced with lentivirus containing the gene for galectin-3 and selected using 0.5 µg/ml puromycin for 96 h.

Intoxication assays HeLa cells and BT549 cells were seeded in 6 well plates at a density of 1E6 cells per well and 24 h later shifted to 4°C to inhibit endocytosis. Cells were then intoxicated with 10 nM TcdB for 1 h at 4°C. One well for each cell type was lysed with RIPA buffer, while a second well was shifted to 37°C for 45 min to allow endocytosis and processing of the toxin. Lysates were normalized by BCA assay (Pierce) and run on SDS-PAGE. Immunoblots were probed for the TcdB GTD (20B3) and GAPDH loading control.

37 Construction of gene trap library The U3neoSV1 retroviral vector (129) was obtained from Zirus, Inc. Caco-2 cells were plated in 75 cm2 flasks and incubated with U3neoSV1 (MOI=0.1) at 37°C for 1 h in the presence of 4 µg/ml Polybrene. Transduced cells were selected using G418 at a concentration of 750 µg/ml and grown to confluence.

Gene trap screen Gene-trapped Caco-2 cells were plated in 10 cm dishes and challenged with 15 nM native TcdB for 4 h at 37°C, after which media was exchanged, and cells were left to recover for 96 h. After two subsequent rounds of toxin selection, individual surviving cells were grown into clonal colonies, isolated using trypsin-soaked whatman filters, and allowed to propagate. Clones were verified as being toxin-resistant using the viability assay described below. Genomic DNA was isolated from each clone using a QIAmp DNA Blood Maxi (QIAGEN). The gene-trap vector and surrounding genomic DNA sequence was isolated by digestion of genomic DNA with EcoRI or BamHI, self-ligation, transformation of the resulting plasmids into E. coli, and growth on selective media.

Viability assays Cells were seeded at a density of 2500 cells/well in a 384 well plate for 16 h and incubated with serial dilutions of toxin. Viability in Caco-2 cells was measured 18 h or 48 h post intoxication with Cell-TiterGLO (Promega) as per the manufacturer’s instructions. For HeLa cells, viability was measured 4.5 h post intoxication with Cell-TiterGLO (Promega) as per the manufacturer’s instructions. For BT549 cells, viability was measured 18 h post intoxication with Cell-TiterGLO. For Cell-TiterGLO assays, luminescence was read on a BioTek® Synergy plate reader. For each replicate plate, no-toxin control for each cell line was scaled to 100% for Cell-TiterGLO assays. Data are presented as the mean and SEM of multiple replicate experiments.

Protein expression and purification Native TcdB was expressed and purified from C. difficile supernatant as described (22) and used for the initial gene trap screening. Recombinant toxins or toxin fragments were used in subsequent experiments and purified as in previous publications (130). In brief, proteins are expressed in Bacillus megaterium, lysates are clarified, and toxins are purified in three chromatography steps: Ni-NTA, anion exchange, and size exclusion chromatography.

38 Galectin-3 coding sequence (see below) was cloned into pET28 and expressed in E. coli BL21 as an N-terminal His6-tagged fusion protein. Lysate was clarified and protein was purified over Ni-NTA and size exclusion chromatography.

TcdB-galectin-3 co-immunoprecipitation Purified recombinant TcdB and purified recombinant galectin-3 were mixed at approximately equimolar concentrations in PBS and allowed to bind for 1 h at 4°C. A 20 µl aliquot was kept as “input,” and the rest of the binding reaction was split into two tubes. To one tube, 2 µg of anti- galectin-3 antibody (Santa Cruz sc-53127) was added. To the other, 1µg of anti-TcdB antibody (Abcam ab775832) was added. Tubes were allowed to incubate end-over-end with antibody overnight at 4°C. Protein-G bead slurry (GE) was added to bind the immunoglobulin, and immunocomplexes were pulled down, washed 5 times with PBS, and eluted from the beads by boiling in Laemmli buffer. Products were run on western blot and probed for both TcdB and galectin-3 using antibodies mentioned above.

Knockdown by siRNA Caco-2 or HeLa cells were plated and the next day transfected with a pool of four Dharmacon siGENOME siRNAs against lgals3 transcript (M-010606) or a control siRNA (D-001206-13) using Lipofectamine® RNAiMAX (Invitrogen) as per the manufacturer’s instructions. After 48 h of treatment with siRNA, cells were assayed for viability as described above or lysed with RIPA buffer. Expression of galectin-3 was determined by immunoblotting whole-cell lysates for galectin-3 (Santa Cruz sc-56108) and GAPDH as a loading control (Santa Cruz sc-25778). HeLa cells were also transfected with Dharmacon SMARTpool siRNAs against other galectin genes: lgals1 (M-011718), lgals2 (M-020134), lgals4 (M-012232), lgals7 (M-011719), lgals8 (M- 010607), lgals9 (M-011319), lgals12 (M-010683), lgals13 (M-017203) and tested for viability, and further tested with a pool of four Dharmacon ON-TARGETplus lgals3 siRNAs (SMARTpool L-010606) and the associated non-targeting control siRNA (D-001810-10). The lgals3 ON- TARGETplus SMARTpool was deconvoluted into four lgals3 siRNAs, transfected individually into HeLa cells (J-010606-06, J-010606-07, J-010606-08, J-010606-09), and tested for effects on viability as described above. For follow up on other gene-trap hits, HeLa cells were plated and transfected with Dharmacon siGENOME SMARTpool siRNAs against the following transcripts: his1h2bi (M-011445), brp44

39 (M-020233), hnrnph1 (M-012107), parm1 (M-020227), ttr (M-012554-02), rps15a (M-013542), areg (M-017435), taf1d (M-014314), rsph3 (M-018730), fndc1 (M-024901), iqwd1 (M-020878), cpeb4 (M-014636), gata6 (M-008351), slco2b1 (M-007442), rab3b (M-008825), prr15l (M- 014338), col2a1 (M-013073), sec31a (M-014166), eif3b (M-019196), capn12 (M-005802), rbm39 (M-011965), sc5dl (M-009745), plin2 (M-019204), mapk1ip1l (M-016369), ccng1 (M- 003216), tm4sf1 (M-010610), arl6ip1 (M-004321), fgl1 (M-007904), (M-004001), cdh17 (M- 011829), snx5 (M-012524), nedd3 (M-019818), (M-019022), pou3f3 (M-020135), lgals3bp (M-008016), n4bp2l2 (M-014177), ap2s1 (M-011833), as well as the control, lgals3, and lgals7 siRNAs mentioned above, and the control and pvrl3 shRNAs described in Chapter III. Triplicate wells were either left untreated or intoxicated with 10 nM TcdB for 4 h at 37°C, at which time viability was read as described above. Replicate wells were averaged and signal for each RNAi was normalized to the no-toxin control. Finally, data were presented as a fold-change of viability versus the control RNAi.

Lentiviral overexpression of galectin-3 The vector pLJM1 containing the Rap2A coding sequence was obtained from Addgene (plasmid ID: 19311), and was digested with SalI and EcoRI to remove the Rap2A coding sequence. Digestion products were separated on a 0.7% agarose gel, and the vector band was gel purified using a Wizard SV Gel and PCR Clean-up Kit (Promega A9281) using the manufacturer’s instructions. The coding sequence of galectin-3 cloned into pSPORT6 was acquired from the Vanderbilt Microarray Shared Resource Core. An EcoRI site was engineered at the 3’ end of the coding sequence in pCMV-SPORT6-lgals3 by quickchange PCR. The resulting construct was treated with SalI and EcoRI, and the band containing the lgals3 coding sequence was gel purified and ligated into the linearized pLJM1 vector. To produce lentivirus for galectin-3 overexpression, HEK293T cells were plated in a 10 cm dish and transfected with 8 µg total DNA and 12 µL Fugene-6 (Promega). DNA was a mixture of 3 µg ∆R8.91 + 1 µg pCMVG + 4 µg pLJM1-lgals3. The packaging plasmids ∆R8.91 and pCMVG were kindly donated by the Aiken lab (Vanderbilt University). After 48 hours of transfection, the total 8 ml of media was harvested, filtered through a 0.45 um filter, and frozen in 250 µL aliquots at -80°C. BT549 cells were plated in 75 cm2 flasks and incubated with 500 µL overexpression lentivirus-containing supernatant diluted in 3 mL serum-free media for 4 h, then complete media overnight. Transduced cells were selected for by culturing cells in 0.5 µg/ml puromycin for 96 h. Puromycin-resistant cells were cloned by limited dilution in later experiments and checked for

40 galectin-3 expression by immunoblot. Three galectin-3 positive clones were used for subsequent experiments.

Rac glucosylation assay Wild-type Caco-2 and Caco-2 D02 gene-trap cells were lifted with trypsin, counted, and seeded in 6-well dishes at 1E6 cells per well and allowed to incubate overnight at 37°C. Dishes were allowed to equilibrate to room temperature for 20 minutes and then incubated at 4°C for 1 h to inhibit endocytosis. Cells were intoxicated with 10 nM TcdB for 1 h at 4°C. One replicate well for each cell type was harvested: media was aspirated, cells were washed once with PBS, and cells were lysed with RIPA buffer. A second replicate well for each cell type was shifted to 37°C. After 45 minutes, lysate was harvested as described. Protein lysates were normalized by BCA assay (Pierce), and run on SDS-PAGE. Immunoblots were probed for TcdB glucosyltransferase domain (20B3), GAPDH (sc-25778), and Rac1 (BD 610650).

41 CHAPTER III IDENTIFICATION OF AN EPITHELIAL CELL RECEPTOR RESPONSIBLE CLOSTRIDIUM DIFFICILE TCDB-INDUCED CYTOTOXICITY

Introduction TcdA and TcdB induce a cytopathic ‘rounding’ in epithelial cells, which is dependent on the N-terminal glucosyltransferase domain inactivating Rho-family GTPases, including Rac1. TcdB also activates the cytotoxic pathway at higher concentrations of toxin, which is GTD independent and leads to large amounts of reactive oxygen species and ultimately necrosis. We speculate that both mechanisms are important in the context of disease; the cytopathic effects promote inflammation and disruption of the tight junctions, whereas the TcdB-induced necrosis contributes to the colonic tissue damage observed in severe cases of CDI. Although TcdA and TcdB are homologs, they appear to perform separate, nonredundant functions (26, 131). TcdA and TcdB are thought to use different receptors, based on sensitivity differences among cell types in vitro (36, 38, 132, 133). Multiple receptors for TcdA have been proposed including Galα1–3Galβ1–4GlcNAc, blood antigens I, X, and Y, rabbit sucrase isomaltase, and gp96 (36, 77-79). The TcdA CROPS domain is thought to play a role in binding cell surface carbohydrates (34, 36, 37). Antibodies against the CROPS domains of TcdA and TcdB can block intoxication (134, 135), and excess TcdA CROPS domain can compete with TcdA holotoxin for cell binding (136). At the same time, truncations of TcdA and TcdB that lack the CROPS domains are still capable of intoxicating cells (39-41) and a homologous toxin from Clostridium perfringens, TpeL, lacks a CROPS domain entirely (41, 81). A receptor for TpeL has been identified (41), suggesting that a receptor-binding site for other large clostridial toxins could exist outside of the CROPS. Another report indicates that chondroitin sulfate proteoglycan 4 (CSPG4) mediates TcdB-induced cytopathic and apoptotic events in HeLa cells and HT29 cells (42). CSPG4 does not mediate the necrotic effects that occur at higher TcdB concentrations and CSPG4 binds TcdB outside the CROPS; these observations are consistent with a dual receptor hypothesis. This study represents an independent effort to define the cellular factors responsible for TcdB binding and toxicity.

42 Results

PVRL3 is required for TcdB-mediated cytotoxicity To identify cellular factors involved in TcdB-mediated cytotoxicity, we used a genetic selection to isolate toxin-resistant Caco-2 cells (see Chapter 2). First, Caco-2 cells were mutagenized with a retroviral genetrap vector that inserts a promoterless neomycin gene at random locations throughout the genome (120). Insertions that occur in actively transcribed regions are expected to confer resistance to neomycin (G418) and prevent expression of the disrupted gene. A library of G418-resistant Caco-2 cells was challenged with three successive treatments of 15 nM TcdB. Surviving cells were propagated from single clones, and the genomic DNA was sequenced with vector-specific primers to identify the location of the genetic disruption. Of the 61 sequenced clones, we identified two mutant clones harboring mutations in the poliovirus receptor-like 3 (PVRL3 or nectin-3) locus. As this gene encodes a membrane protein with three extracellular Ig domains and is related to several known viral receptors, we hypothesized that PVRL3 is required for TcdB binding to the cell surface. We used one of the gene-trapped cell lines, designated E19, for further study. Wild-type and E19 Caco-2 cells were challenged with serial dilutions of TcdB. The E19 cells were resistant to cell death relative to wild-type Caco-2 cells when measuring ATP as a viability indicator, and sensitivity could be restored through complementation with a lentiviral PVRL3 overexpression plasmid (Figure 3.1A). Knockdown of PVRL3 expression was incomplete, however (Figure 3.1B). The gene-trap insertion occurred within the promoter of the pvrl3 locus, leaving the coding region of both alleles intact. To confirm results obtained using the gene-trap mutant, to achieve more efficient expression knockdown, and to further rule out the possibility of the genetic insertion affecting multiple loci, we transduced Caco-2 cells with four different shRNAs targeting the pvrl3 transcript and compared these cells with cells transduced with a nontargeting shRNA. Knockdown of PVRL3 by shRNA resulted in protein levels that were nearly undetectable by immunoblot (Figure 3.1D) and a significant increase in cell survival at both 18 (Figure 3.1C) and 48 h (Figure 3.2A). These data suggest that PVRL3 expression is required for TcdB-mediated cell death in Caco-2 cells. The consequence of disrupting PVRL3 expression appears to be specific to TcdB. Whereas Caco-2 cells are insensitive to TcdA at 18 h (Figure 3.1E), the PVRL3 knockdown cells were still sensitive to the TcdA-induced apoptotic cell death that occurs after 48 h (Figure 3.1F).

43

Figure 3.1: Disruption of PVRL3 expression in Caco-2 cells results in resistance to TcdB. A. Disruption of PVRL3 expression in Caco-2 cells results in resistance to TcdB. A. Caco-2, Caco-2 E19, and E19 stably overexpressing PVRL3 were seeded and treated for 18 h with serial dilutions of TcdB. For each cell line, ATP was measured as a readout of viability and normalized to signal from untreated cells. Results represent the mean ± SEM from eight experiments. B. Western blot of whole-cell lysates of cells used in A were probed for PVRL3 or GAPDH (loading control). C. Caco-2 cells were stably transduced with four different shRNAs targeting pvrl3 transcript or a non- targeting shRNA (control). Cells were seeded and intoxicated, and viability was measured and plotted as in A. Results represent the mean ±SEM of six experiments. D. Western blot of whole-cell lysates from cells used in C. E&F. Caco-2 cells transduced with the non-targeting shRNA (control) and two pvrl3 shRNAs from D were plated and treated with serial dilutions of TcdA. ATP levels were measured at 18 h (E) or 48 h (F) and viability was plotted as in A and D. Results are mean ± SEM from four experiments. Viability assays show *P < 0.05, **P < 0.005, ***P < 0.001; P values relative to wild-type cells are shown in A and control shRNA in C. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test.

We next evaluated whether the role of PVRL3 in TcdB-induced killing was restricted to Caco-2 cells by testing an unrelated cell type. Knockdown of PVRL3 in HeLa cells using two different shRNAs (Figure 3.2C) resulted in increased cell survival at 4.5-h, as indicated by ATP concentration (Figure 3.2A) and lactate dehydrogenase (LDH) release, an alternative indicator of TcdB-mediated necrosis (Figure 3.2B). We also disrupted the expression of PVRL3 in HeLa cells using the CRISPR/Cas9 system (Figure 3.2E). The CRISPR-mutated HeLa cells were

44 significantly less sensitive to TcdB, whereas the complemented mutants were as sensitive as the wild-type cells (Figure 3.2D).

Figure 3.2: Disruption of PVRL3 in HeLa cells results in cell survival. A. HeLa cells were transduced with two different shRNAs targeting pvrl3 or a non-targeting shRNA (control). Cells were seeded and intoxicated for 4.5 h. Viability was assayed as in Figure 1. Results represent the mean ± SEM of five experiments. B. Cells from A were seeded and challenged with TcdB or digitonin for 4.5 h, and necrosis was measured using an LDH release assay. Signal was normalized to unintoxicated cells. Digitonin control represents maximum signal to determine the dynamic range of the assay. Results represent the mean ± SEM of three experiments. C. Western blot of whole-cell lysates of cells used in A and B probing for PVRL3 and GAPDH (loading control). D. Two CRISPR guide RNAs were designed against pvrl3 and transfected into HeLa cells. These cells, wild-type HeLa cells, and the CRISPR mutants complemented with a PVRL3 overexpression vector were challenged with TcdB or digitonin for 4.5 h. Cell viability was normalized as in B. Results represent the mean ± SEM from four experiments. E. Western blot of whole cell lysates of cells used in D. Viability assays show *P < 0.05, **P < 0.005; P values relative to control shRNA cells are shown in A and B and wild type in C. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test.

45 PVRL3 is required for the cytopathic effect in Caco-2 cells but not HeLa cells Unlike in Caco-2 cells (Figure 3.1A), knockdown of PVRL3 expression in HeLa cells did not confer resistance to the apoptotic events that occur when cells are exposed to toxin for 48 h (Figure 3.3B). Furthermore, the knockdown did not confer resistance to TcdB-induced Rac1 modification or cytopathic “rounding” events (Figs. 3.4 and 3.5). These findings are consistent with observations from Yuan et al. that CSPG4 functions in TcdB-mediated cytopathic and apoptotic effects in HeLa cells (42). They are also consistent with our observation that, whereas PVRL3 is expressed in both cell types, CSPG4 is only expressed in HeLa cells (Figure 3.3C).

46

Figure 3.3: PVRL3 mediates apoptosis in Caco-2 but not HeLa cells. A. Caco-2 cells transduced with shRNAs used in Figure 1E were seeded and challenged with TcdB for 48 h. Viability was read and normalized as in Figure 1. Results represent mean ± SEM of four experiments. *P < 0.05; **P < 0.005, ***P < 0.001. P values are relative to control shRNA. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test. B. HeLa cells transduced with shRNAs used in Figure 2 were seeded and challenged with TcdB for 48 h. Viability was measured and normalized as in Figure 2A. Results represent mean ± SEM of three experiments. C. Western blots of whole cell lysates from wild-type Caco-2 and wild-type HeLa cells were probed for CSPG4, PVRL3, and GAPDH (loading control).

47

Figure 3.4: Knockdown of PVRL3 in HeLa does not inhibit TcdB-induced Rac1 glucosylation. HeLa cells were transduced with an shRNA targeting the pvrl3 transcript or a control shRNA. Cells were plated and intoxicated with 10 pM TcdB for indicated amounts of time, then lysed with RIPA buffer and run on SDS-PAGE. Immunoblots were probed for PVRL3, CSPG4, unglucosylated Rac1, and loading controls.

48

Figure 3.5. Cytopathic effects of TcdB on HeLa cells. HeLa cells transduced with shRNAs used in Figure 2 were seeded, challenged with 1 pM TcdB, and imaged at 0-, 30-, 60-, and 90-min time points. The transition to round cells is indicative of the cytopathic effect. Photographs taken by Melissa Farrow.

In the E19 gene-trapped Caco-2 clone, which has reduced PVRL3 expression, Rac1 modification and cell rounding are impaired upon intoxication with TcdB (Figure 3.6). These phenotypes are rescued when PVRL3 is expressed in E19 cells. We also show that apoptosis is PVRL3-dependent in Caco-2 cells (Figure 3.3A). These data support a model where in the absence of CSPG4, PVRL3 functions as a secondary receptor for the cytopathic pathway.

49

Figure 3.6: TcdB-induced cytopathic effects are impaired in Caco-2 E19. A. Caco-2 wild-type, E19, and E19 harboring an overexpression vector for PVRL3 were plated and intoxicated with 10 pM TcdB for 1h, then lysed with RIPA buffer and run on SDS-PAGE. Immunoblots were probed for PVRL3, unglucosylated Rac1, and loading controls. B. Cells from (A) were plated and intoxicated with 1 nM TcdB for 1 h, and imaged. The transition to round cells is indicative of the cytopathic effect.

50 TcdB binds directly to the PVRL3 ectodomain As PVRL3 is a membrane-bound protein with three extracellular Ig domains, we hypothesized that TcdB directly interacts with PVRL3 on the surface of cells. To test whether TcdB can directly bind to PVRL3, pull-down assays were conducted with a His-tagged extracellular domain of PVRL3. TcdB, TcdB1–1834, TcdB CROPS, and TcdA1–1832 were biotinylated, conjugated to streptavidin–agarose beads, and incubated with purified PVRL31–359–

His6 ectodomain. PVRL31–359–His6 coeluted with full-length TcdB and TcdB1–1834, but not with

TcdB CROPS, TcdA1–1832, or empty beads (Figure 3.7A). The TcdA CROPS binds a broad array of carbohydrates (38). Full-length TcdA binds to PVRL3 but also a number of unrelated glycosylated proteins (Figure 3.8). The unrelated JAM-A ectodomain (residues 1–235) did not bind any of the TcdB constructs (Figure 3.7B), indicating a specific interaction between TcdB and PVRL3.

Figure 3.7 PVRL3 binds directly to TcdB. Purified recombinant TcdB holotoxin, TcdB1–1834, TcdB CROPS, and TcdA1–1832 were biotinylated and bound to streptavidin–agarose beads. Toxin-bound beads were then incubated with purified recombinant ectodomains of PVRL3 (A) or JAM-A (B). Complexes were eluted by boiling in Laemmli buffer, separated by SDS/PAGE, and either stained with colloidal blue or transferred to PVDF membranes. Western blots were probed for the His tag on the ectodomain constructs. The recombinant TcdB proteins also have His tags, and the CROPS domain is visible because it is similar in size to the PVRL3 ectodomain.

51

Figure 3.8. TcdA binds to a variety of glycosylated proteins. Biotinylated recombinant TcdA holotoxin was bound to streptavidin–agarose beads. Toxin-bound beads were then incubated with purified recombinant PVRL3 ectodomain (PVRL3 1–359), 1% FBS, anti-GAPDH antibody (rabbit IgG), or anti-PVRL1 antibody (mouse IgG). Complexes were eluted in Laemmli buffer, and immunoblots were performed for the His6 tag on the PVRL3 ectodomain. Secondary antibodies against rabbit IgG and mouse IgG were pooled and used to detect heavy chains of the antibodies.

Inhibition of the TcdB-PVRL3 interaction leads to cellular survival After demonstrating that TcdB could bind directly to PVRL3, we determined whether this interaction could be inhibited on the cell surface and lead to cell survival. We tested whether a monoclonal antibody recognizing the N-terminal, D1 Ig domain of PVRL3 could block intoxication of Caco-2 cells. Cells were incubated with the anti-PVRL3 antibody or an unrelated isotype control antibody and challenged with toxin for 18 h (Figure 3.9A). The anti-PVRL3 antibody protected Caco-2 cells from TcdB cytotoxicity at concentrations of 2 ng/μL (13 nM) and 4 ng/μL (27 nM) but not at 1 ng/μL (7 nM). This dose-dependent inhibition supports the hypothesis that PVRL3 acts as a binding partner for TcdB on the surface of cells.

52

Figure 3.9 Extracellular inhibition of the TcdB–PVRL3 interaction confers protection against TcdB. A. Caco-2 cells were pretreated with a monoclonal antibody recognizing the D1 domain of PVRL3 or an isotype control antibody and challenged for 18 h with TcdB. Viability was measured and plotted as in Figure 1. B. TcdB was pre-bound to purified recombinant PVRL1, PVRL2, or PVRL3ectodomain for 16 h. HeLa cells were challenged with toxin–ectodomain complexes for 4.5 h and cell viability was measured using an LDH release assay. Results are plotted as fold change of signal relative to no-toxin control cells. A and B. Results are means ± SEM from three experiments. *P < 0.05, **P <0.005 are relative to buffer controls. Statistics calculated using two-way ANOVA and Dunnett’s multiple comparison test.

We also tested whether preincubation of TcdB with purified PVRL3 ectodomain could inhibit cytotoxicity. At the same time, we tested the role of two related proteins, PVRL1 and PVRL2, for the capacity to inhibit cytotoxicity. The glycosylated ectodomains of human PVRL1, PVRL2, and PVRL3 were expressed and purified from the supernatant of transfected HEK293 cells, incubated with TcdB, and applied to HeLa cells for 4.5 h (Figure 3.9B). At 15 nM TcdB, prebinding with PVRL3 resulted in baseline levels of signal in an LDH release assay, comparable to untreated cells. At 75 nM TcdB, where TcdB was present in molar excess, this protection began to diminish but was still statistically significant. Prebinding with PVRL1 or PVRL2 had no effect on cell viability, suggesting that the protective effect is specific for PVRL3.

Discussion The goal for this study was to identify a host cell receptor for TcdB. While numerous studies have emphasized the use of haploid cell lines for identifying pathogen host factors (137), we wanted to start with Caco-2 cells since numerous studies have documented the utility of these cells as a model for TcdB intoxication (138) and the intestinal epithelium in general (139). Since colleagues had previously succeeded in identifying host factors important for Clostridium perfringens ε-toxin intoxication by constructing a gene-trap library using Madin

53 Darby canine kidney (MDCK) cells (119), we adopted a similar strategy in Caco-2 cells. We found two TcdB-resistant clones with disruptions in the PVRL3 locus and were able to confirm the importance of PVRL3 in TcdB-induced cytotoxicity using multiple cell lines and genetic approaches. Genetic disruption of PVRL3 expression through gene-trap insertion (Figure 3.1A & B), shRNA knockdown (Figure 3.1C, 3.2A-C), or CRISPR/Cas9 mutagenesis (Figure 3.2D&E) confers resistance to TcdB-induced cell death. Furthermore, ectopic expression of PVRL3 restores the sensitivity of the E19 Caco-2 (Figure 3.1A) or CRISPR-mutated HeLa cells (Figure 3.2D&E) to TcdB-mediated cell death. Our data suggest that PVRL3 is a receptor for TcdB. First, we observe a direct binding interaction between PVRL3 and TcdB using purified proteins (Figure 3.7). This binding interaction can be interrupted on cells with either an anti-PVRL3 antibody (Figure 3.9A) or purified PVRL3 ectodomain (Figure 3.9B) to provide protection against TcdB-induced cytotoxicity. These data indicate that the TcdB-PVRL3 interaction is relevant in a physiological context. PVRL3 belongs to a family of four proteins: PVRL1 to PVRL4 (also called Nectin-1 to Nectin-4, respectively) (116). These proteins consist of three extracellular immunoglobulin-like domains, a single transmembrane helix, and a short cytoplasmic domain that binds afadin (106). They are related to the poliovirus receptor, PVR, (also called Necl-5/Tage4/CD155) which has a similar structure but does not bind afadin. Many of the PVR/PVRL family members are involved in virus entry. PVR was first identified as the poliovirus receptor (82), followed by PVRL1 and PVRL2 as herpesvirus receptors (91, 140), and most recently PVRL4 as an epithelial receptor for measles virus (94, 95) and other morbilliviruses (141, 142). This study represents the first report of a pathogen using PVRL3 as a receptor and also the first instance of a bacterial toxin using a member of this family of proteins as a receptor. The nectins form homotypic and heterotypic trans dimers, which contribute to adherens junction formation in many vertebrate tissues (143). Individual subtypes are typically expressed in distinct but overlapping patterns that, in the case of the auditory epithelium, lead to a checkerboard cellular pattern (144). PVRL3 can form heterotypic adhesions with PVRL1, PVRL2, and PVR. As with viruses and other toxins that exploit proteins localized in junctions (145-148), one question is how the pathogen gains access to these sites. It is possible that the function of the CROPS is to mediate interaction with other receptors on the apical surface until PVRL3 becomes accessible. Accessibility could occur through the natural turnover of cells in the colon (149) or though toxin-mediated disruption of the tight junctions (150-152). On the other

54 hand, the accessibility of PVRL3 may not be as limited as initially assumed. Note that in IHC staining, PVRL3 is highly expressed and not limited to junctions (125). A related question is how PVRL3 binding contributes to cytotoxicity. TcdB cytotoxicity depends on access to an acidified endosome (45, 153), a process mediated by clathrin- mediated endocytosis (43). The interaction of PVR with PVRL3 causes down-regulation of PVR from the cell surface through a clathrin-mediated endocytic mechanism (108). Microbeads coated with a PVRL3-IgG Fc fusion placed on the surface of NIH3T3 cells induce the assembly of clathrin heavy chain or adaptin B, components needed for the formation of endocytotic clathrin-coated vesicles in response to PVR binding. Thus, PVRL3 is capable of mediating endocytic entry in response to ligand binding. We propose that a similar entry mechanism takes place for TcdB. Chrondroitin sulfate proteoglycan 4 (CSPG4) mediates TcdB-induced rounding of HeLa cells. We observed that HeLa cells with depleted levels of PVRL3 remain susceptible to TcdB- induced cytopathic Rac1 glucosylation (Figure 3.4), ‘rounding effects’ (Figure 3.5), and apoptosis (Figure 3.3B). However, E19 gene-trapped Caco-2 cells were deficient in these cytopathic effects (Figure 3.6), and Caco-2 cells knocked down for PVRL3 were significantly more viable at 48 h (Figure 3.3A). As Caco-2 cells do not express detectable levels of CSPG4 (Figure 3.3C), we propose that in the absence of CSPG4, PVRL3 can function as a receptor for the cytopathic pathway as well as the receptor for the cytotoxic pathway. Schorch et al. recently proposed a dual-receptor mechanism for large clostridial toxins by which the CROPS domain allows the toxin to dock onto the cell surface by interacting with oligosaccharides, followed by toxin binding to a high-affinity receptor (41). Our data are compatible with this model, as binding to PVRL3 is independent of the CROPS domain (Figure 3.7). However, we note that both PVRL3 and CSPG4 bind TcdB outside of the CROPS domain (42). More work is needed to address whether there is a specific cellular host factor that mediates interaction with the CROPS and how this interaction contributes to the interaction we observe here with PVRL3. Another important point is that CSPG4 is not expressed on the surface of the human colon (154) and therefore cannot be a receptor for the initial phases of C. difficile infection, but PVRL3 is highly expressed on the surface of the colonic epithelium (125). The challenges of C. difficile infection have led to an interest in developing therapeutic approaches that target the activity of the toxins. Future work designed to define the PVRL3- TcdB interface should be able to guide the design of small molecules or biologics that block this interaction.

55

Methods Cell culture Caco-2 cells were maintained in MEM +10% FBS +1% HEPES +1% Non-essential amino acids +1% NaPyruvate. G418 at a concentration of 750 µg/ml was added to the media for gene-trap clone selection, and shRNA clones were selected for with 5 µg/ml puromycin. HeLa cells were maintained in DMEM +10% FBS +1% NaPyruvate. Lentiviral overexpression, shRNA transduction, and CRISPR transfection were selected for using 0.5 µg/ml puromycin.

Transduction of cultured cells Lentiviral particles encoding shRNAs against PVRL3 and a scrambled control were obtained from Open Biosystems (catalog numbers RHS4348, VGH5518-200277059, VGH5518- 200285736, VGH5518-200298879, VGH5518-200306796). To produce lentivirus for PVRL3 overexpression, HEK293T cells were plated in a 10 cm dish and transfected with 8 µg total DNA and 12 µL Fugene-6 (Promega). DNA was a mixture of 3 µg ∆R8.91 + 1 µg pCMVG + 4 µg pLJM1-PVRL3. The plasmids ∆R8.91 and pCMVG were kindly donated by the Aiken lab (Vanderbilt University). After 48 hours of transfection, the total 8 ml of media was harvested, filtered through a 0.45 µm filter, and frozen in 250 µL aliquots at -80°C. Caco-2 or HeLa cells were plated in 75 cm2 flasks and incubated with 2 µL shRNA or 500 µL overexpression lentivirus-containing supernatant diluted in 3 mL serum-free media for 4 h, then complete media overnight. Transduced cells were selected for by culturing Caco-2 cells in 5 µg/ml puromycin for 96 h; HeLa cells were selected for using 0.5 µg/ml puromycin for 96 h. Puromycin-resistant cells were pooled for later experiments. Expression of PVRL3 was determined by immunoblotting whole-cell lysates for PVRL3 (Santa Cruz sc-28637) and GAPDH as a loading control (Santa Cruz sc-25778). Lysates were also probed for CSPG4 (Santa Cruz sc-20162).

Cytopathicity Assay Wild-type, E19, and E19-PVRL3 overexpressing Caco2 cells were plated in 6-well dishes, intoxicated with 1 nM TcdB, and imaged 1 h post intoxication. HeLa cells transduced with non- targeting or two independent pvrl3-targeting shRNAs were intoxicated with 1 pM TcdB and imaged every 30 min over a 90-min period. Images were acquired at 20°— on an Olympus IX73 microscope with a Q color 3 camera.

56

CRISPR/Cas9 mutagenesis Two CRISPR guide RNAs were designed to target the second exon of the human pvrl3 genomic coding sequence. Targeting sequence 1 was created using the oligos 5’- CACCGTAGCTGGACCAATTATTG-3’ and 5’-AAACCCACAATAATTGGTCCAGC-3’. Targeting sequence 2 was created using the oligos 5’-CACCGCTCCATAGGTGCCTTAGC-3’ and 5’- AAACCCAGCTAAGGCACCTATGG-3’. Pairs of oligos encoding the guide RNA and the complimentary strand were cloned into the BbsI sites of pX459 (Addgene plasmid ID: 48139) (155) and transfected into HeLa cells using Lipofectamine LTX. Transfected cells were selected with puromycin for 48 h. Surviving cells were pooled and used for further study.

Viability Assays Cells were seeded at a density of 2500 cells/well in a 384 well plate for 16 h and incubated with serial dilutions of toxin. Viability in Caco-2 cells was measured 18 h or 48 h post intoxication with Cell-TiterGLO (Promega) as per the manufacturer’s instructions. For HeLa cells, viability was measured 4.5 h or 48 h post intoxication with either Cell-TiterGLO or Cell-ToxGLO (Promega) as per the manufacturer’s instructions. For both Cell-TiterGLO and Cell-ToxGLO assays, luminescence was read on a BioTek® Synergy plate reader. For each replicate plate, no-toxin control for each cell line was scaled to 100% for Cell-TiterGLO assays, or 1 for Cell- ToxGLO. Data are presented as the mean and SEM of multiple replicate experiments.

Cloning PVRL3: Total RNA from Caco-2 cells was isolated using a Qiagen RNeasy Mini kit. Complementary DNA (cDNA) was prepared in a 2-step RT-PCR reaction: in the first step, 10 ng of RNA was converted to cDNA with Superscript II RT (Invitrogen) using the manufacturer’s instructions and random hexamers as primers. In the second step, a conventional PCR was performed in a 50 µL reaction volume with 2.5 µL of cDNA template and 2.5 µL each primer. To clone full-length PVRL3 from Caco-2 cDNA, the PVRL3 coding sequence was amplified using forward primer containing a SalI site, and a reverse primer containing a NotI site (5’- GTTAGGTCGACGATGGCGCGGACCCTGCG-3’ and 5’- CTTCTGCGGCCGCCTAAACATACCACTCCCTCCTGGA-3’). A band at the expected size of approximately 1700 bp was excised, gel purified, digested with NotI and SalI, and ligated into the linearized pCMV-SPORT6 vector, transformed and amplified in E. coli strain XL1-Blue, and

57 sequenced. The pCMV-SPORT6-PVRL3 construct (pBL490) was then used as a template to make the lentiviral overexpression plasmid pLJM1-PVRL3 (pBL548) as described below. The vector pLJM1 containing the Rap2A coding sequence was obtained from Addgene (plasmid ID: 19311), and was digested with SalI and EcoRI to remove the Rap2A coding sequence. Digestion products were separated on a 0.7% agarose gel, and the vector band was gel purified using a Wizard SV Gel and PCR Clean-up Kit (Promega A9281) using the manufacturer’s instructions. An EcoRI site was engineered at the 3’ end of the coding sequence in pCMV-SPORT6-PVRL3 by quickchange PCR. The resulting construct was treated with SalI and EcoRI, and the band containing the PVRL3 coding sequence was gel purified and ligated into the linearized pLJM1 vector. Plasmids for the recombinant expression of TcdB (pBL377) (30), TcdA (pBL282) (50), and TcdA1-1832 (pBL515) (50) were previously described. A plasmid for the recombinant expression of TcdB1-1834 (pBL575) was generated by loop-out mutagenesis with pBL377 to remove the region encoding residue 1835 to the end of the gene. A plasmid for the recombinant expression of the TcdB CROPS domain (pBL281) was generated by amplifying DNA encoding TcdB residues 1827-2366 with primers that encoded the BamH1 (5’) and XhoI (3’) restriction sites. The fragment was digested with BamH1 and XhoI and ligated into a modified pET27b vector. The vector encodes an N-terminal His tag and confers kanamycin resistance to E. coli. A plasmid for the recombinant expression of TcdB with an N-terminal 3xFLAG tag (pBL705) was generated by six consecutive rounds of quickchange mutagenesis using pBL377 as a template.

Protein expression and purification Constructs containing the coding sequence for the PVRL1, PVRL2, and PVRL3 ectodomains cloned into pCEP4 were gifts from Lawrence Shapiro (Columbia University) (105). The coding sequence for the JAM-A ectodomain cloned into pcDNA3.1 (145) was a gift from Terrence Dermody (Vanderbilt University). The JAM-A ectodomain coding sequence was PCR amplified with primers encoding a HindIII site at the 5’ end and a NotI site at the 3’ end and cloned into pCEP4 (pBL687). All ectodomains were expressed in HEK293T cells and purified as described (105).

Pull-downs Purified TcdA, TcdB, and TcdB fragments were chemically biotinylated using the EZ-Link Sulfo- NHS-Biotin kit (Thermo 21217). Excess unbound biotin was removed from the protein

58 preparations using a Sephadex G-25 desalting column (GE 52-1308-00 BB). 150 pmol of each biotinylated protein was bound to 20 µL streptavidin agarose beads (EMD-Millipore 16-126) for 1 h at 4°C in 20 mM HEPES pH 8.0 + 50 mM NaCl. Beads were spun down and washed once to remove excess unbound toxin and then incubated with 1.5 nmol ectodomain in 20 mM Tris pH 8.0 + 150 mM NaCl at 4°C for 18 h. Beads were washed three times with 20 mM Tris pH 8.0 + 150 mM NaCl + 0.5 % Triton X-100 and complexes were eluted by boiling in Laemmli buffer. Eluted proteins were run on SDS-PAGE and stained with Colloidal Blue (Invitrogen LC6025).

Immunoblots were also performed, probing for the His6 tag on the ectodomain (Abcam ab18184).

Antibody blocking Caco-2 cells were plated at a density of 2500 cells per well of a 384-well plate and incubated overnight at 37°C. The plate was incubated at 4°C for 1 h to inhibit endocytosis, and then incubated with a monoclonal antibody against PVRL3 (Sigma SAB1402559) or a monoclonal antibody against CagA (Santa Cruz sc-28368) at a final concentration of 4 ng/µL, 2 ng/µL, or 1 ng/µL. Cells were incubated at 4°C for 1 h to allow antibody binding, and then serial dilutions of TcdB were added to quadruplicate wells and allowed to incubate for 18 h at 37°C. Cell viability was read using Cell-TiterGLO as described above.

Ectodomain Competition HeLa cells were seeded at 2500 cells/well in a 384 well plate and allowed to incubate at 37°C for 16 h. Serial dilutions of PVRL1, PVRL2, and PVRL3 ectodomain were prepared and added to serial dilutions of purified TcdB. Complexes were allowed to incubate for 16 h at 4°C, at which time they were added to cells for 4.5 h at 37°C. Viability was assayed using Cell-ToxGLO as described above.

Statistics For experiments using cultured cell lines, statistical analyses were performed using a two-way ANOVA, and p-values were generated using a Dunnett’s multiple comparisons test in GraphPad Prism. Data are presented as means ± SEM.

This chapter is published as LaFrance, et al. (2015), Proc Natl Acad Sci USA, pp. 7073-8.

59 CHAPTER IV SUMMARY

Conclusions Clostridium difficile is currently responsible for a major health crisis in the United States. Understanding the mechanism of action of its virulence factors, TcdA and TcdB, will go a long way in directing the design of future therapeutics. At the start of this project, host factors used by TcdB to adhere to the surface of colonic epithelial cells and gain access to the cell cytosol were unknown. The primary motivation for this work was to address this knowledge gap. In Chapter II, I describe genetic screen using mammalian cells to identify host factors required for TcdB cytotoxicity. We sequenced 58 clones resistant to nanomolar concentrations of TcdB, used RNAi to confirm that disruption of these genes led to resistance to toxin, and followed up on two candidates: the clathrin adaptor protein AP2 and the galactose-specific binding protein galectin-3. We showed that in accordance with the Aktories group’s published data, AP2 is required for TcdB cytotoxicity and entry into cells (Figure 2.1A and B), but contrary to their published work, AP2 is not required for TcdA-mediated cell death (Figure 2.1A). We also showed that galectin-3 is required for TcdB-mediated cytotoxicity in both HeLa and Caco-2 cells (Figure 2.2) and that this role for galectin-3 appears to be unique among members of the galectin family (Figure 2.2D). A galectin-3 negative cell line, BT549, is resistant to TcdB (Figure 2.3A and B). These cells become sensitive to TcdB when galectin-3 is expressed using a lentiviral vector. Furthermore, galectin-3 can bind directly to TcdB (Figure 2.3C). These data show that AP2 and galectin-3 are required for TcdB-mediated cytotoxicity. AP2 is necessary for endocytosis of TcdB; the role of galectin-3 is currently unknown. In Chapter III, I show that PVRL3 is an epithelial cell receptor for TcdB (156). The gene- trapped cell line E19 from the screen described in Chapter II was resistant to intoxication with TcdB, but sensitivity was rescued when PVRL3 was expressed using a lentiviral plasmid (Figure 3.1). Knockdown of PVRL3 by shRNA in Caco-2 (Figure 3.1) or HeLa cells (Figure 3.2A and B) or CRISPR/Cas9 knockout (Figure 3.2C) of PVRL3 in HeLa cells protected against TcdB intoxication. As with the gene-trap mutant, TcdB sensitivity in the CRISPR/Cas9 knockout cells could be rescued by overexpression of PVRL3 using a lentiviral vector. TcdB binds directly to purified PVRL3 ectodomain (Figure 3.7), and the interaction occurs outside of the TcdB CROPS domain. Finally, inhibition of the PVRL3-TcdB interaction either by preincubation with excess

60 PVRL3 ectodomain or with an antibody specific for the receptor blocked intoxication with TcdB on cells (Figure 3.9). After determining that PVRL3 is a receptor for the cytotoxic pathway ending in necrosis, we also examined the role of PVRL3 in the cytopathic pathway. We showed that in HeLa cells, the cytopathic response is not PVRL3-specific (Figure 3.3, 3.4, 3.5), which is consistent with published data by Yuan et al. showing that in HeLa cells, this pathway is driven by CSPG4 (42). However, in Caco-2 cells, which do not express detectable levels of CSPG4, the cytopathic pathway is PVRL3-dependent (Figure 3.3 and 3.6), suggesting that in the absence of CPSG4, PVRL3 can function as a receptor for this pathway and cytotoxic pathway. The sequenced clones from the gene-trap screen returned 45 discrete genome loci. In addition to the work mentioned above on PVRL3, CLTC/AP2, an offshoot of which is now the focus of Ramya Chandrasekaran’s doctoral work, and the galectin gene cluster, there are many genes from this list for which no follow up work has been done. The candidates in Table 2 could lead to many more projects in the lab.

Future directions

Understanding the role of galectin-3 in TcdB-induced cytotoxicity The galectin-3 story remains unfinished. Thus far, I have shown that galectin-3 plays a role in TcdB-mediated cytotoxicity and that galectin-3 can bind directly to TcdB. However, what role galectin-3 plays in TcdB-mediated pathogenesis remains unclear. Since I have stopped working on this project, we have developed many tools in the lab that would be helpful in moving this work forward. The first logical step would be to engineer a CRISPR/Cas9 knockout in HeLa cells and rescue the knockout using the lentiviral vector I have already made. The standard assay for inhibition of galectins is to competitively block the carbohydrate recognition domain (CRD) using excess lactose in the media; sucrose is used as a negative control. I attempted this assay, but my results were inconsistent. I had been using our standard ATP readout as a viability indicator, but the cells consumed the excess sugar in the medium as a carbon source, shifting the baseline as they metabolized the carbon source as they were challenged with TcdB at the same time. Now that we have other viability assays for use in the lab, this lactose inhibition assay is once again an option. We also have other constructs of TcdB that we did not have at the time that I began this project, namely, TcdB1-1834. Although I did observe direct binding between the TcdB holotoxin

61 and galectin-3, I did not ascribe the binding to any specific domain. My hypothesis was is TcdB has lost its capacity to bind carbohydrates at its C-terminal CROPS domain and that instead, galectin-3 performs this function. It would be interesting to determine where on the toxin galectin-3 interacts.

Other gene-trap screen hits There are several candidates that fall into the category of vesicle trafficking and recycling. Not much is known about how TcdB traffics through the cell, except that it enters through clathrin-mediated entocytosis and requires an acidified endosome to properly function. EPS15 homology domain-containing protein-1 (EHD1) is thought to be involved in endosomal recycling, particularly with containing growth factor receptors (157) and (158). This would suggest that clone D05 is defective in recycling endosomes that contain PVRL3 back to the plasma membrane after the first round of toxin endocytosis. On the other hand, clone D05 also disrupts two microRNAs. One of these, miR192, leads to the downregulation of a transcriptional repressor for E-cadherin. Disruption of miR192 should lead to the downregulation of E-cadherin, which is what we observe by RT-PCR (Figure 4.1). The other microRNA, miR194-2, is not well studied, but also could have important regulatory effects.

Figure 4.1: Gene-Trap Clone D05 is disrupted in EHD1 and E-Cadherin expression. Total RNA was extracted from wild-type and D05 Caco-2 cells. RT-PCR was performed with gene- specific primers as indicated, and no-enzyme (-RT) negative controls. Actin-β serves as a loading control. Data collected together with Ramya Chandrasekaran.

Jovic et al. studied the effects of EHD1 on the recycling of β1 integrin by using a pulse- chase experiment, where they bound an antibody to surface-bound integrin (pulse), washed away unbound antibody, allowed endocytosis and after 2 h, added more antibody to track recycling over time (chase) (158). This assay was conducted using HeLa cells treated with a

62 control siRNA and an siRNA against EHD1 and found that there was a significant delay in the kinetics of β1 integrin recycling in EHD1-depleted cells. We have established a similar entry assay for the endocytosis of TcdA and TcdB, which we can use for both immunoblotting and confocal microscopy. Using HeLa cells expressing an siRNA against EHD1, and tracking the entry of labeled TcdB, we can determine whether EHD1 is involved in receptor recycling of PVRL3. We also have a GFP-labeled construct of PVRL3 that can be used with HeLa cells if we want to track the endocytosis and recycling of PVRL3.

Interplay between PVRL3 and CSPG4, and possibly other as-yet unknown receptors We now know that there are at least two receptors for TcdB: CSPG4 (42) and PVRL3 (156). In Caco-2 cells expressing PVRL3 shRNAs, at higher concentrations of TcdB and 48 h of incubation time, I still observed about 25% cell death (Figure 3.3, shRNA 1). This result could be explained by incomplete knockdown, or it could mean that there is yet another receptor that has yet to be identified. Due to the homology of the other nectin family members to PVRL3, I wanted to know whether another nectin family member could act as a lower affinity receptor for TcdB in the absence of PVRL3. By RT-PCR, Caco-2, HeLa, and 293T cells, which are sensitive to the cytotoxic effects of TcdB, also express PVRL1, PVRL2, and PVR, but not PVRL4 (Figure 4.2). HT29 cells express PVRL4 but not PVRL3 and are not sensitive to TcdB cytotoxicity. I did not test PVRL4 for this reason and did not test PVR because of lack of time and reagents.

Figure 4.2: RT-PCR profiles of common cell lines. Total RNA was extracted from Caco-2, HeLa, 293T, and HT29 cells. RT-PCR was performed with gene- specific primers as indicated, and no-template (NTC) and no-enzyme (-RT) negative controls. Actin-β served as a loading control.

PVRL1 is most similar to PVRL3 by primary sequence and could be the best candidate for a low-affinity binding partner in the absence of PVRL3. However, the ectodomain of PVRL1

63 failed to compete for binding of TcdB on cells (Fig. 3.9), suggesting that if PVRL1 can function as an alternate receptor, the affinity is significantly less than that of the TcdB-PVRL3 interaction. Follow-up on PVRL1 and other nectins in a PVRL3-deficient cell line, such as the CRISPR- mutated HeLa cells (Figure 3.2), still would be warranted. It is also possible that another, completely different receptor has yet to be found. Doing another screen in a PVRL3-/-CSPG4-/- background would be one way to identify it. I have created stable, PVRL3-/- HeLa cells by CRISPR/Cas9, and the CRISPR oligos used by Yuan et al. are published and could be used to make the double knockout. A CSPG4/PVRL3 double knockout cell line could be a useful reagent to have in the lab when studying the interplay between the two pathways and the specific downstream effects of the roles of PVRL3 or CSPG4.

Virus-PVR interactions as a guide for understanding the TcdB-PVRL3 structure Poliovirus may interact with its receptor using PVR dimerization-mimickry, creating a high-affinity interaction with PVR by taking advantage of the free energy reduction of burying a solvent-exposed phenyl ring in a hydrophobic pocket (101). The PVR proteins each have a phenylalanine on the F-G loop forced to stay in a solvent-exposed conformation by a nearby proline in a highly conserved TFPxG motif and a pocket formed by a glycine and a large hydrophobic residue approximately 20 Å away on the C’-C’’ loop. The model suggests that the viral F184 on VP3 interacts with the hydrophobic pocket on PVR, mimicking the dimerization motif used by the nectin family members to create adherens junctions. Likewise, the crystal structure of the HSV gD/PVRL1 complex reveals that the conserved PVRL1 F129 binds in a hydrophobic pocket between a side chain F223 and an amphipathic helix on gD (103). Mutating F129 on PVRL1 to serine abolishes binding of HSV to the receptor (103). Looking more closely at the primary sequence of TcdB, and in particular where it differs from TcdA, and mapping regions of differences onto the surface of a homology model of TcdB based on the published crystal structure of TcdA (31) (Figure 4.3A), I noticed that there were two regions on the surface of TcdB that are similar to the conserved dimerization motifs of the PVR proteins on either side of a cleft in the toxin (Figure 4.3B). TcdB also has a solvent- exposed phenylalanine in a constrained position due to a nearby proline, F780 on TcdB, located in the autoprocessing domain, in a similar SFNP sequence. About 20 Å across the cleft, located on the delivery domain, there is a patch of hydrophobicity similar to those in the nectin

64 sequences, containing a glycine and a phenylalanine residue (Figure 4.3B and C, highlighted in white).

Figure 4.3: Proposed binding pocket for PVRL3 on TcdB. A. Homology model of TcdB1-1834 based on the published crystal structure of TcdA1-1832 (31). GTD in red, APD in blue, Delivery domain in gold, CSPG4 binding domain in cyan. B. Zoom of proposed binding cleft. F780, G895, and F896 indicated with arrows, and side chains in sticks. C, same as (B) but with space fill.

The spacing across the cleft is an appropriate size to fit the receptor. I hypothesize that PVRL3 binds in the cleft formed by the three domains of TcdB, that the F153 of PVRL3 crucial for dimerization (105) binds into a hydrophobic pocket on the delivery domain of the toxin formed by G895 and F896 (Figure 4.3B and C), and that TcdB F780 binds to the hydrophobic pocket on the receptor, mimicking PVRL3 dimerization. My hypothesized hydrophobic glycine pocket on the delivery domain is missing on TcdA, sterically blocked and filled in by a threonine, which could explain why TcdA does not bind to the receptor (Figures 3.7, 4.4). We are currently testing this theory by making mutations in this region of TcdB to make it more TcdA-like, blocking this pocket with larger residues, and testing the effects on PVRL3 binding and toxicity in cells. We are also making other mutations on the autoprocessing domain, including F780A, to test its effects on PVRL3 binding and toxicity in

65 cells. On the receptor side, we are making an F153A mutant. In addition to making this mutant in the recombinant D1 and testing binding, we are planning to express this mutant receptor in the CRISPR-edited PVRL3-null HeLa cells (Figure 3.2) and test the capacity of TcdB to intoxicate these cells.

Binding studies with PVRL3 We have begun to define the binding site between PVRL3 and TcdB by making shorter constructs of both the toxin and receptor. The other members of the PVR family are viral receptors, and PV, HSV, and MV bind to the N-terminal Ig domain of their cognate receptors. Based on this knowledge, I hypothesized that TcdB would bind the D1 of PVRL3. I cloned the D1 into a pET vector as a StrepII-tagged recombinant protein, expressed it in E. coli to inclusion bodies, refolded it, and used this protein for preliminary binding studies (Figure 4.4).

Figure 4.4: PVRL3 D1 interacts directly with TcdB. PVRL3 D1 (residues 1-165) was recombinantly expressed and refolded as a StrepII-tagged protein and bound to streptactin beads. Receptor-bound beads were incubated with purified recombinant TcdB holotoxin, TcdB1–1834, TcdB CROPS, TcdB glucosyltransferase domain (GTD), TcdB autoprocessing domain (APD) or TcdA1–1832. Complexes were eluted by boiling in Laemmli buffer, separated by SDS- PAGE, and stained with Coomassie blue.

As with the full-length, fully glycosylated extracellular region of PVRL3, the shorter, non- glycosylated D1 bound to TcdB holotoxin and TcdB1-1834 but not the C-terminal CROPS domain. I also tested binding to the N-terminal GTD and APD and found that PVRL3 D1 bound to both

66 (Figure 4.4). The D1 bound to the APD in every replicate of the pull down I performed, but the GTD bound in only roughly half of the experiments (data not shown), leading me to think that the Kd of the interaction to the APD was quite strong but that the interactions with the GTD may not be as specific. This pull down does show, however, that TcdB interacts with PVRL3 at the D1 and that glycosylation is not required for the interaction. We also had the opportunity to determine a preliminary Kd for the interaction between the full-length ectodomain and the TcdB holotoxin by microscale thermophoresis (MST) in collaboration with Dr. Benjamin Spiller and the Department of Pharmacology during a product demonstration (Figure 4.5). The binding affinity was measured at 240 ± 20 nM.

Figure 4.5: Microscale thermophoresis. 100 nM TcdB was complexed with serial dilutions of PVRL3 ectodomain. Samples were loaded into MST capillaries and analysis was performed using Monolith NT.115. Data generated by NanoTemper Technologies, Inc.

There is still much to be done in this area. Ultimately we would like to generate a high- resolution structure of the TcdB-PVRL3 complex (see below), but this may be slow to achieve. Therefore, we will work on mutational studies using these reagents. Based on mutational studies with virus-receptor interactions, point mutants on the C’ strand and F-G loop of PVRL3 D1 would be a good place to start (Figure 4.6).

67

Figure 4.6: Residues critical for binding to nectins. Alignment of D1 residues of PVRL1, PVRL3, PVRL4, and PVR. Residues highlighted in pink have been shown by mutagenesis to be critical for HSV binding to PVRL1 (103), MV binding to PVRL4 (104), and PV binding to PVR (102). Boxes demark the residues on the C’-C’’ and F-G loops critical for nectin dimerization.

The proposed binding surface on TcdB overlaps with the previously identified receptor- binding site for CSPG4 (42) (Figure 4.3, indicated in cyan). During my ectodomain competition assays, I noticed that pre-binding PVRL3 ectodomain to TcdB not only blocked the cytotoxic effects of TcdB but also prevented rounding of HeLa cells (data not shown). This finding indicates that pre-binding of PVRL3 to TcdB also blocks binding of TcdB to CSPG4, the primary receptor for the cytopathic pathway in HeLa cells.

Structure: crystallography and electron microscopy based studies of TcdB-PVRL3 complex Knowledge of how the toxin and receptor interact would go a long way to dictate anti- toxin therapies that target the first stage of cellular intoxication. We have begun by using negative-stain electron microscopy to study the structure of the TcdB-PVRL3 complex. TcdB1-

1810 and PVRL31-359-His6 were mixed at a 1:2 molar ratio and used for conventional negative stain electron microscopy with uranyl formate (0.7% mass/volume) as described (30). Single particle images were collected and a sampling is shown in Figure 4.7. Single particles have given us a low-resolution look at the complex, which is consistent with the model that the receptor binds to the cleft formed between the delivery, autoprocessing, and glucosyltransferase domains of the toxin (Figure 4.3).

68

Figure 4.7: Electron microscopy images of TcdB-PVRL3 complex. TcdB1-1810 was complexed with TcdB1-359 and imaged by negative stain electron microscopy. Single particles are shown in the bottom of each panel. Top of each panel: TcdB outlined in white/gray, PVRL3 Ig domains outlined in pink. Images taken by Michelle LaFrance, Stacey Seeback, and Heather Kroh, in collaboration with the laboratory of Dr. Melanie Ohi.

Future work will include multiple particle averages to give us more information on the structure of the complex. Mutagenesis of both the toxin and the receptor and subsequent imaging by EM will give us information on the residues required for the interaction. Finally, since we have successfully demonstrated binding of the smaller D1 domain to TcdB1-1834 (Figure 4.4), we are working toward an x-ray crystal structure of TcdB in complex with the PVRL3 D1.

Activation of NOX by PVRL3 A major question left unanswered is how, if the TcdB GTD is inactivating Rac1 by glucosylating the Switch-1 region, does the NOX complex, which requires an activated Rac1, stay activated for the length of time we observe NOX-mediated cell death. The answer may lie in the normal role of PVRL3, and the signaling pathways it is involved in during the formation of cell adhesions. PVRL3 activates multiple signaling pathways once it has formed trans- heterodimers on the extracellular space. Interactions in cis with αvβ3 integrin activate focal adhesion kinase (FAK), which then phosphorylates Src and ultimately leads to activation of Rac1 (110).

69 Alternatively, interactions with afadin lead to recruitment of E-cadherin, which leads to a transient activation of Rac1 at nascent focal adhesions (159). Teasing apart the mechanism by which ligation of PVRL3 leads to the activation of Rac1 can be done using tools we already have in the lab. We have on hand the whole-genome siRNA library, so taking a targeted approach and knocking down multiple genes induced by PVRL3 signaling, then challenging these cells with TcdB and assaying for activation of Rac1 using a pull-down for active GTPases (71) (130), assaying for the activation of the NOX complex using a fluorescent indicator dye (71), or assaying for loss of membrane integrity with our LDH release assay, would be feasible. Once we have narrowed the list of candidate genes, we can study the relevant pathways using CRISPR/Cas9 stable knockout cell lines and complementation experiments.

Role of PVRL3 in a mouse model of infection Although we have defined a role for PVRL3 in tissue culture models of TcdB intoxication, we have yet to establish a role for this receptor in C. difficile infection. The mouse model of C. difficile infection is widely used (27), and PVRL3 knockout mice have been successfully bred (115, 117). Unfortunately, when we reached out to the Takai group about collaborating on this project using their PVRL3 knockout mice, they declined. Therefore, we have developed reagents to make our own knockout mice using CRISPR/Cas9. I have designed guide RNAs that target the mouse pvrl3 gene sequence at two locations: one guide RNA targets the start codon, while the other targets the splice site between Intron 1 and Exon 2 (Figure 4.8). I cloned these guide sequences into our CRISPR/Cas9 vectors and transfected them into MEF cells, showing that each sequence disrupts gene expression and that PVRL3 is necessary for TcdB- mediated cytotoxicity in mouse cells.

70

Figure 4.8: Effects of PVRL3 in MEF cells. Two CRISPR guide RNAs were designed against pvrl3 and transfected into MEF cells. These cells and wild-type MEF cells were challenged with TcdB for 4.5 h. Cell viability was normalized to no-toxin control for each cell line. Results represent the mean ± SEM from four experiments. Inset: immunoblot of whole cell lysates probed for PVRL3 or GAPDH as a loading control.

Studying C. difficile pathogenesis using a pvrl3-/- animal would give new insight into the role of TcdB in infection and help to tease apart the ongoing question of the individual roles of TcdA and TcdB in disease.

Hypothetical applications A small molecule antitoxin that can block the PVRL3-TcdB interaction on the surface of the cell, before the toxin has a chance to enter, may be an attractive candidate for C. difficile treatment for several reasons: it does not have to cross the plasma membrane, therefore bypassing the potential problem of uptake into the cell, and it inhibits the first stage of intoxication, so all downstream signaling is blocked. There is also the possibility that a drug that binds the TcdB receptor-binding pocket in such a way that it blocks nectin binding could be modified to also bind either the HSV, MV, or PV receptor binding pocket such that it also could be an antiviral therapy. PVRL4 has recently been identified as a tumor marker, and MV is a candidate therapy for various cancers that overexpress PVRL4 (160, 161). Since PVRL4 is normally highly expressed only in placenta and stomach cells (125), the Edmonson vaccine strain of MV, which uses PVRL4 as a receptor, is relatively safe to use as an anti-tumor agent. Upregulation of PVRL3 has not been associated with cancer or other diseases as of yet, but if it is in the future, attenuated mutants of TcdB may be of use as targeted therapeutics.

71

Concluding Remarks When I began this work, there were no known cell surface binding partners for TcdB. There are now two identified receptors: PVRL3 (156) and CSPG4 (42). CSPG4 is expressed on HeLa and smooth muscle cells but not on colonic epithelial cells, while PVRL3 is expressed on many cultured cell types and the surface of the colonic epithelium. There are many exciting open questions left to explore. For example, why does TcdB only activate the NOX complex at higher (≥100 pM) concentrations of toxin? In Caco-2 cells at low concentrations of TcdB, where PVRL3 is the only receptor, we observe apoptosis at low concentrations of TcdB and necrosis at higher concentrations. This observation suggests that assembly or activation of the NOX complex on the endosome only happens at higher concentrations of TcdB. What drives this concentration threshold-dependent switch to the redox-active state? Another question left unanswered is the role of the TcdB CROPS domain. At the beginning of this work, the CROPS domain was assumed to be the receptor binding domain; however, both PVRL3 and CSPG4 bind TcdB outside this domain. So, then, what is the function of this domain? Does it make other interactions with the cell surface? Does it have another, unknown function? A third question is that of accessing the receptor. PVRL3 is primarily located at adherens junctions and is tightly complexed with other junctional proteins. How, then, does the toxin ligate it? We and others have observed by histologial staining that some PVRL3 is localized at the apical surface (125, 156). Alternatively, as cell turnover occurs in the intestine, the toxin may have temporary access to open junctions. PVRL3 plays a major role in TcdB intoxication of epithelial cells in tissue. It is our model that TcdB interacts with PVRL3 first in the very beginning stages of tissue intoxication, activating both the NOX-mediated necrotic cell death at high concentrations of TcdB and the GTD- dependent Rac1 inactivation at lower concentrations that leads to loss of focal adhesions, fluid secretion, and diarrhea. Once TcdB has intoxicated the surface layer of cells in the tissue, those cells have lost their focal adhesions, and the toxins are able to move further into the tissue, then TcdB may access CSPG4-expressing cells and disseminate further into other organs. By this time, however, the patient is already suffering from a severe case of CDI and significant damage has

72 already been done to the colon. Therefore, focusing on PVRL3 as a relevant drug target and moving forward with therapeutics targeting the TcdB-PVRL3 interaction would be warranted.

73 APPENDIX

LIST OF PUBLICATIONS

LaFrance, M.E., Farrow, M.A., Chandrasekaran, R., Sheng, J., Rubin, D.H., and Lacy, D.B. Identification of an epithelial cell receptor responsible for Clostridium difficile TcdB-induced cytotoxicity. Proc Natl Acad Sci USA, 2015. 112(22): p. 7073-8.

74 BIBLIOGRAPHY

1. Hall, I.C. and O'Toole, E. Intestinal flora in newborn infants with a description of a new pathogenic anaerobe, Bacillus difficilis. Amer. J. Dis. Child., 1935. 49: p. 390-402.

2. Snyder, M.L. Further Studies on Bacillus difficilis (Hall and O'Toole). 1937. 60(2): p. 223- 231.

3. Rifkin, G.D., Fekety, F.R., and Silva, J. Antibiotic-induced colitis implication of a toxin neutralised by Clostridium sordellii antitoxin. Lancet, 1977. 2(8048): p. 1103-6.

4. Larson, H.E. and Price, A.B. Pseudomembranous colitis: Presence of clostridial toxin. Lancet, 1977. 2(8052-8053): p. 1312-4.

5. Bartlett, J.G., Chang, T.W., Gurwith, M., Gorbach, S.L., and Onderdonk, A.B. Antibiotic- associated pseudomembranous colitis due to toxin-producing clostridia. N Engl J Med, 1978. 298(10): p. 531-4.

6. Bartlett, J.G., Moon, N., Chang, T.W., Taylor, N., and Onderdonk, A.B. Role of Clostridium difficile in antibiotic-associated pseudomembranous colitis. Gastroenterology, 1978. 75(5): p. 778-82.

7. Rifkin, G.D., Silva, J., and Fekety, R. Gastrointestinal and systemic toxicity of fecal extracts from hamsters with clindamycin-induced colitis. Gastroenterology, 1978. 74(1): p. 52-7.

8. Katz, L., LaMont, J.T., Trier, J.S., Sonnenblick, E.B., Rothman, S.W., Broitman, S.A., and Rieth, S. Experimental clindamycin-associated colitis in rabbits. Evidence of toxin- mediated mucosal damage. Gastroenterology, 1978. 74(2 Pt 1): p. 246-52.

9. Onderdonk, A.B., Hermos, J.A., Dzink, J.L., and Bartlett, J.G. Protective effect of metronidazole in experimental ulcerative colitis. Gastroenterology, 1978. 74(3): p. 521-6.

10. Voth, D.E. and Ballard, J.D. Clostridium difficile toxins: mechanism of action and role in disease. Clin Microbiol Rev, 2005. 18(2): p. 247-63.

11. Kelly, C.P., Pothoulakis, C., and LaMont, J.T. Clostridium difficile colitis. N Engl J Med, 1994. 330(4): p. 257-62.

12. Lessa, F.C., Mu, Y., Bamberg, W.M., Beldavs, Z.G., Dumyati, G.K., Dunn, J.R., Farley, M.M., Holzbauer, S.M., Meek, J.I., Phipps, E.C., Wilson, L.E., Winston, L.G., Cohen, J.A., Limbago, B.M., Fridkin, S.K., Gerding, D.N., and McDonald, L.C. Burden of Clostridium difficile infection in the United States. N Engl J Med, 2015. 372(9): p. 825-34.

13. Kelly, C.P. and LaMont, J.T. Clostridium difficile--more difficult than ever. N Engl J Med, 2008. 359(18): p. 1932-40.

14. Eyre, D.W., Walker, A.S., Wyllie, D., Dingle, K.E., Griffiths, D., Finney, J., O'connor, L., Vaughan, A., Crook, D.W., Wilcox, M.H., and Peto, T.E.A. Predictors of First Recurrence

75 of Clostridium difficile Infection: Implications for Initial Management. Clinical Infectious Diseases, 2012. 55(suppl 2): p. S77-S87.

15. Lowy, I., Molrine, D.C., Leav, B.A., Blair, B.M., Baxter, R., Gerding, D.N., Nichol, G., Thomas, W.D., Leney, M., Sloan, S., Hay, C.A., and Ambrosino, D.M. Treatment with monoclonal antibodies against Clostridium difficile toxins. N Engl J Med, 2010. 362(3): p. 197-205.

16. McDonald, L.C., Killgore, G.E., Thompson, A., Owens, R.C., Kazakova, S.V., Sambol, S.P., Johnson, S., and Gerding, D.N. An epidemic, toxin gene-variant strain of Clostridium difficile. N Engl J Med, 2005. 353(23): p. 2433-41.

17. Warny, M., Pepin, J., Fang, A., Killgore, G., Thompson, A., Brazier, J., Frost, E., and McDonald, L.C. Toxin production by an emerging strain of Clostridium difficile associated with outbreaks of severe disease in North America and Europe. Lancet, 2005. 366(9491): p. 1079-84.

18. Hall, A.J., Curns, A.T., McDonald, L.C., Parashar, U.D., and Lopman, B.A. The roles of Clostridium difficile and norovirus among gastroenteritis-associated deaths in the United States, 1999-2007. Clinical Infectious Diseases, 2012. 55(2): p. 216-23.

19. Dubberke, E.R. and Olsen, M.A. Burden of Clostridium difficile on the Healthcare System. Clinical Infectious Diseases, 2012. 55(suppl 2): p. S88-S92.

20. Banno, Y., Kobayashi, T., Watanabe, K., K., U., and Nozawa, Y. Two toxins (D-1 and D- 2) of Clostridium difficile causing antibiotic-associated colitis; purification and some characterization. Biochem. Int., 1981. 2: p. 629-635.

21. Taylor, N.S., Thorne, G.M., and Bartlett, J.G. Comparison of two toxins produced by Clostridium difficile. Infection and immunity, 1981. 34(3): p. 1036-43.

22. Lyerly, D.M., Saum, K.E., MacDonald, D.K., and Wilkins, T.D. Effects of Clostridium difficile toxins given intragastrically to animals. Infection and immunity, 1985. 47(2): p. 349-52.

23. Triadafilopoulos, G., Pothoulakis, C., O'Brien, M.J., and LaMont, J.T. Differential effects of Clostridium difficile toxins A and B on rabbit ileum. Gastroenterology, 1987. 93(2): p. 273-9.

24. Sambol, S.P., Merrigan, M.M., Lyerly, D., Gerding, D.N., and Johnson, S. Toxin gene analysis of a variant strain of Clostridium difficile that causes human clinical disease. Infection and immunity, 2000. 68(10): p. 5480-7.

25. Rupnik, M., Kato, N., Grabnar, M., and Kato, H. New types of toxin A-negative, toxin B- positive strains among Clostridium difficile isolates from Asia. J Clin Microbiol, 2003. 41(3): p. 1118-25.

26. Lyras, D., O'Connor, J.R., Howarth, P.M., Sambol, S.P., Carter, G.P., Phumoonna, T., Poon, R., Adams, V., Vedantam, G., Johnson, S., Gerding, D.N., and Rood, J.I. Toxin B is essential for virulence of Clostridium difficile. Nature, 2009. 458(7242): p. 1176-9.

76 27. Carter, G.P., Chakravorty, A., Pham Nguyen, T.A., Mileto, S., Schreiber, F., Li, L., Howarth, P., Clare, S., Cunningham, B., Sambol, S.P., Cheknis, A., Figueroa, I., Johnson, S., Gerding, D., Rood, J.I., Dougan, G., Lawley, T.D., and Lyras, D. Defining the Roles of TcdA and TcdB in Localized Gastrointestinal Disease, Systemic Organ Damage, and the Host Response during Clostridium difficileInfections. mBio, 2015. 6(3): p. e00551-15.

28. Dobson, G., Hickey, C., and Trinder, J. Clostridium difficile colitis causing toxic megacolon, severe sepsis and multiple organ dysfunction syndrome. Intensive care medicine, 2003. 29(6): p. 1030.

29. Henkel, J.S., Baldwin, M.R., and Barbieri, J.T. Toxins from bacteria. EXS, 2010. 100: p. 1-29.

30. Pruitt, R.N., Chambers, M.G., Ng, K.K.-S., Ohi, M.D., and Lacy, D.B. Structural organization of the functional domains of Clostridium difficile toxins A and B. Proc Natl Acad Sci USA, 2010. 107(30): p. 13467-72.

31. Chumbler, N.M., Rutherford, S.A., Zhang, Z., Farrow, M.A., Lisher, J.P., Farquhar, E., Giedroc, D.P., Spiller, B.W., Melnyk, R.A., and Lacy, B.D. Crystal Strucure of Clostridium difficile Toxin A. Nature Microbiology, 2016. 1(1).

32. Dove, C.H., Wang, S.Z., Price, S.B., Phelps, C.J., Lyerly, D.M., Wilkins, T.D., and Johnson, J.L. Molecular characterization of the Clostridium difficile toxin A gene. Infection and immunity, 1990. 58(2): p. 480-8.

33. Sauerborn, M. and von Eichel-Streiber, C. Nucleotide sequence of Clostridium difficile toxin A. Nucleic acids research, 1990. 18(6): p. 1629-30.

34. von Eichel-Streiber, C. and Sauerborn, M. Clostridium difficile toxin A carries a C- terminal repetitive structure homologous to the carbohydrate binding region of streptococcal glycosyltransferases. Gene, 1990. 96(1): p. 107-13.

35. Ho, J.G.S., Greco, A., Rupnik, M., and Ng, K.K.-S. Crystal structure of receptor-binding C-terminal repeats from Clostridium difficile toxin A. Proc Natl Acad Sci USA, 2005. 102(51): p. 18373-8.

36. Krivan, H.C., Clark, G.F., Smith, D.F., and Wilkins, T.D. Cell surface binding site for Clostridium difficile enterotoxin: evidence for a glycoconjugate containing the sequence Gal alpha 1-3Gal beta 1-4GlcNAc. Infection and immunity, 1986. 53(3): p. 573-81.

37. Greco, A., Ho, J.G.S., Lin, S.-J., Palcic, M.M., Rupnik, M., and Ng, K.K.-S. Carbohydrate recognition by Clostridium difficile toxin A. Nat Struct Mol Biol, 2006. 13(5): p. 460-1.

38. Dingle, T., Wee, S., Mulvey, G.L., Greco, A., Kitova, E.N., Sun, J., Lin, S., Klassen, J.S., Palcic, M.M., Ng, K.K.S., and Armstrong, G.D. Functional properties of the carboxy- terminal host cell-binding domains of the two toxins, TcdA and TcdB, expressed by Clostridium difficile. Glycobiology, 2008. 18(9): p. 698-706.

77 39. Genisyuerek, S., Papatheodorou, P., Guttenberg, G., Schubert, R., Benz, R., and Aktories, K. Structural determinants for membrane insertion, pore formation and translocation of Clostridium difficile toxin B. Mol Microbiol, 2011. 79(6): p. 1643-54.

40. Olling, A., Goy, S., Hoffmann, F., Tatge, H., Just, I., and Gerhard, R. The Repetitive Oligopeptide Sequences Modulate Cytopathic Potency but Are Not Crucial for Cellular Uptake of Clostridium difficile Toxin A. PLoS ONE, 2011. 6(3): p. e17623.

41. Schorch, B., Song, S., van Diemen, F.R., Bock, H.H., May, P., Herz, J., Brummelkamp, T.R., Papatheodorou, P., and Aktories, K. LRP1 is a receptor for Clostridium perfringens TpeL toxin indicating a two-receptor model of clostridial glycosylating toxins. Proc Natl Acad Sci USA, 2014. 111(17): p. 6431-6.

42. Yuan, P., Zhang, H., Cai, C., Zhu, S., Zhou, Y., Yang, X., He, R., Li, C., Guo, S., Li, S., Huang, T., Perez-Cordon, G., Feng, H., and Wei, W. Chondroitin sulfate proteoglycan 4 functions as the cellular receptor for Clostridium difficile toxin B. Cell research, 2015. 25: p. 157-168.

43. Papatheodorou, P., Zamboglou, C., Genisyuerek, S., Guttenberg, G., and Aktories, K. Clostridial glucosylating toxins enter cells via clathrin-mediated endocytosis. PLoS ONE, 2010. 5(5): p. e10673.

44. Qa'Dan, M., Spyres, L.M., and Ballard, J.D. pH-induced conformational changes in Clostridium difficile toxin B. Infection and immunity, 2000. 68(5): p. 2470-4.

45. Barth, H., Pfeifer, G., Hofmann, F., Maier, E., Benz, R., and Aktories, K. Low pH-induced formation of ion channels by clostridium difficile toxin B in target cells. J Biol Chem, 2001. 276(14): p. 10670-6.

46. Pfeifer, G., Schirmer, J., Leemhuis, J., Busch, C., Meyer, D.K., Aktories, K., and Barth, H. Cellular uptake of Clostridium difficile toxin B. Translocation of the N-terminal catalytic domain into the cytosol of eukaryotic cells. J Biol Chem, 2003. 278(45): p. 44535-41.

47. Reineke, J., Tenzer, S., Rupnik, M., Koschinski, A., Hasselmayer, O., Schrattenholz, A., Schild, H., and von Eichel-Streiber, C. Autocatalytic cleavage of Clostridium difficile toxin B. Nature, 2007. 446(7134): p. 415-9.

48. Sheahan, K.-L., Cordero, C.L., and Satchell, K.J.F. Autoprocessing of the Vibrio cholerae RTX toxin by the cysteine protease domain. EMBO J, 2007. 26(10): p. 2552- 61.

49. Egerer, M., Giesemann, T., Jank, T., Satchell, K.J.F., and Aktories, K. Auto-catalytic cleavage of Clostridium difficile toxins A and B depends on cysteine protease activity. J Biol Chem, 2007. 282(35): p. 25314-21.

50. Pruitt, R.N., Chagot, B., Cover, M., Chazin, W.J., Spiller, B., and Lacy, D.B. Structure- function analysis of inositol hexakisphosphate-induced autoprocessing in Clostridium difficile toxin A. J Biol Chem, 2009. 284(33): p. 21934-40.

78 51. Shen, A., Lupardus, P.J., Gersch, M.M., Puri, A.W., Albrow, V.E., Garcia, K.C., and Bogyo, M. Defining an allosteric circuit in the cysteine protease domain of Clostridium difficile toxins. Nat Struct Mol Biol, 2011. 18(3): p. 364-371.

52. Reinert, D.J., Jank, T., Aktories, K., and Schulz, G.E. Structural basis for the function of Clostridium difficile toxin B. J Mol Biol, 2005. 351(5): p. 973-81.

53. Just, I., Wilm, M., Selzer, J., Rex, G., von Eichel-Streiber, C., Mann, M., and Aktories, K. The enterotoxin from Clostridium difficile (ToxA) monoglucosylates the Rho proteins. J Biol Chem, 1995. 270(23): p. 13932-6.

54. Just, I., Selzer, J., Wilm, M., von Eichel-Streiber, C., Mann, M., and Aktories, K. Glucosylation of Rho proteins by Clostridium difficile toxin B. Nature, 1995. 375(6531): p. 500-3.

55. Genth, H., Pauillac, S., Schelle, I., Bouvet, P., Bouchier, C., Varela-Chavez, C., Just, I., and Popoff, M.R. Haemorrhagic toxin and lethal toxin from Clostridium sordellii strain vpi9048: molecular characterization and comparative analysis of substrate specificity of the large clostridial glucosylating toxins. Cell Microbiol, 2014. 16(11): p. 1706-21.

56. Cherfils, J. and Zeghouf, M. Regulation of Small GTPases by GEFs, GAPs, and GDIs. Physiological Reviews, 2013. 93(1): p. 269-309.

57. Herrmann, C., Ahmadian, M.R., Hofmann, F., and Just, I. Functional consequences of monoglucosylation of Ha-Ras at effector domain amino acid threonine 35. J Biol Chem, 1998. 273(26): p. 16134-9.

58. Sehr, P., Joseph, G., Genth, H., Just, I., Pick, E., and Aktories, K. Glucosylation and ADP ribosylation of rho proteins: effects on nucleotide binding, GTPase activity, and effector coupling. Biochemistry, 1998. 37(15): p. 5296-304.

59. Boehm, C., Gibert, M., Geny, B., Popoff, M.R., and Rodriguez, P. Modification of epithelial cell barrier permeability and intercellular junctions by Clostridium sordellii lethal toxins. Cell Microbiol, 2006. 8(7): p. 1070-85.

60. Waschke, J., Baumgartner, W., Adamson, R.H., Zeng, M., Aktories, K., Barth, H., Wilde, C., Curry, F.E., and Drenckhahn, D. Requirement of Rac activity for maintenance of capillary endothelial barrier properties. American journal of physiology Heart and circulatory physiology, 2004. 286(1): p. H394-401.

61. Geny, B., Grassart, A., Manich, M., Chicanne, G., Payrastre, B., Sauvonnet, N., and Popoff, M.R. Rac1 inactivation by lethal toxin from Clostridium sordellii modifies focal adhesions upstream of actin depolymerization. Cell Microbiol, 2010. 12(2): p. 217-32.

62. Halabi-Cabezon, I., Huelsenbeck, J., May, M., Ladwein, M., Rottner, K., Just, I., and Genth, H. Prevention of the cytopathic effect induced by Clostridium difficile Toxin B by active Rac1. FEBS Lett, 2008. 582(27): p. 3751-6.

79 63. May, M., Wang, T., Müller, M., and Genth, H. Difference in F-Actin Depolymerization Induced by Toxin B from the Clostridium difficile Strain VPI 10463 and Toxin B from the Variant Clostridium difficile Serotype F Strain 1470. Toxins, 2013. 5(1): p. 106-119.

64. Brito, G.A.C., Fujji, J., Carneiro-Filho, B.A., Lima, A.A.M., Obrig, T., and Guerrant, R.L. Mechanism of Clostridium difficile toxin A-induced apoptosis in T84 cells. The Journal of infectious diseases, 2002. 186(10): p. 1438-47.

65. Carneiro, B.A., Fujii, J., Brito, G.A.C., Alcantara, C., Oriá, R.B., Lima, A.A.M., Obrig, T., and Guerrant, R.L. Caspase and bid involvement in Clostridium difficile toxin A-induced apoptosis and modulation of toxin A effects by glutamine and alanyl-glutamine in vivo and in vitro. Infection and immunity, 2006. 74(1): p. 81-7.

66. Gerhard, R., Nottrott, S., Schoentaube, J., Tatge, H., Olling, A., and Just, I. Glucosylation of Rho GTPases by Clostridium difficile toxin A triggers apoptosis in intestinal epithelial cells. Journal of medical microbiology, 2008. 57(Pt 6): p. 765-70.

67. Chumbler, N.M., Farrow, M.A., Lapierre, L.A., Franklin, J.L., Haslam, D.B., Haslam, D., Goldenring, J.R., and Lacy, D.B. Clostridium difficile Toxin B causes epithelial cell necrosis through an autoprocessing-independent mechanism. PLoS Pathog, 2012. 8(12): p. e1003072.

68. Li, S., Shi, L., Yang, Z., Zhang, Y., Perez-Cordon, G., Huang, T., Ramsey, J., Oezguen, N., Savidge, T.C., and Feng, H. Critical roles of Clostridium difficile toxin B enzymatic activities in pathogenesis. Infection and immunity, 2015. 83(2): p. 502-13.

69. Ryder, A.B., Huang, Y., Li, H., Zheng, M., Wang, X., Stratton, C.W., Xu, X., and Tang, Y.-W. Assessment of Clostridium difficile Infections by Quantitative Detection of tcdB Toxin by Use of a Real-Time Cell Analysis System. J Clin Microbiol, 2010. 48(11): p. 4129-4134.

70. Song, L., Zhao, M., Duffy, D.C., Hansen, J., Shields, K., Wungjiranirun, M., Chen, X., Xu, H., Leffler, D.A., Sambol, S.P., Gerding, D.N., Kelly, C.P., and Pollock, N.R. Development and Validation of Digital Enzyme-Linked Immunosorbent Assays for Ultrasensitive Detection and Quantification of Clostridium difficile Toxins in Stool. J Clin Microbiol, 2015. 53(10): p. 3204-3212.

71. Farrow, M.A., Chumbler, N.M., Lapierre, L.A., Franklin, J.L., Rutherford, S.A., Goldenring, J.R., and Lacy, D.B. Clostridium difficile toxin B-induced necrosis is mediated by the host epithelial cell NADPH oxidase complex. Proc Natl Acad Sci USA, 2013.

72. Hordijk, P.L. Regulation of NADPH oxidases: the role of Rac proteins. Circulation research, 2006. 98(4): p. 453-62.

73. Brown, D.I. and Griendling, K.K. Nox proteins in signal transduction. Free radical biology & medicine, 2009. 47(9): p. 1239-53.

80 74. Panday, A., Sahoo, M.K., Osorio, D., and Batra, S. NADPH oxidases: an overview from structure to innate immunity-associated pathologies. Cellular and Molecular Immunology, 2015. 12(1): p. 5-23.

75. Szanto, I., Rubbia-Brandt, L., Kiss, P., Steger, K., Banfi, B., Kovari, E., Herrmann, F., Hadengue, A., and Krause, K.-H. Expression of NOX1, a superoxide-generating NADPH oxidase, in colon cancer and inflammatory bowel disease. The Journal of pathology, 2005. 207(2): p. 164-76.

76. Takeya, R., Ueno, N., Kami, K., Taura, M., Kohjima, M., Izaki, T., Nunoi, H., and Sumimoto, H. Novel human homologues of p47phox and p67phox participate in activation of superoxide-producing NADPH oxidases. J Biol Chem, 2003. 278(27): p. 25234-46.

77. Pothoulakis, C., Gilbert, R.J., Cladaras, C., Castagliuolo, I., Semenza, G., Hitti, Y., Montcrief, J.S., Linevsky, J., Kelly, C.P., Nikulasson, S., Desai, H.P., Wilkins, T.D., and LaMont, J.T. Rabbit sucrase-isomaltase contains a functional intestinal receptor for Clostridium difficile toxin A. Journal of Clinical Investigation, 1996. 98(3): p. 641-9.

78. Tucker, K.D. and Wilkins, T.D. Toxin A of Clostridium difficile binds to the human carbohydrate antigens I, X, and Y. Infection and immunity, 1991. 59(1): p. 73-8.

79. Na, X., Kim, H., Moyer, M.P., Pothoulakis, C., and LaMont, J.T. gp96 is a human colonocyte plasma membrane binding protein for Clostridium difficile toxin A. Infection and immunity, 2008. 76(7): p. 2862-71.

80. Lyerly, D.M., Phelps, C.J., Toth, J., and Wilkins, T.D. Characterization of toxins A and B of Clostridium difficile with monoclonal antibodies. Infection and immunity, 1986. 54(1): p. 70-6.

81. Amimoto, K., Noro, T., Oishi, E., and Shimizu, M. A novel toxin homologous to large clostridial cytotoxins found in culture supernatant of Clostridium perfringens type C. Microbiology, 2007. 153(Pt 4): p. 1198-206.

82. Mendelsohn, C.L., Wimmer, E., and Racaniello, V.R. Cellular receptor for poliovirus: molecular cloning, nucleotide sequence, and expression of a new member of the immunoglobulin superfamily. Cell, 1989. 56(5): p. 855-65.

83. Koike, S., Ise, I., and Nomoto, A. Functional domains of the poliovirus receptor. Proc Natl Acad Sci USA, 1991. 88(10): p. 4104-8.

84. Freistadt, M.S. and Racaniello, V.R. Mutational analysis of the cellular receptor for poliovirus. Journal of virology, 1991. 65(7): p. 3873-6.

85. Brandenburg, B. and Zhuang, X. Virus trafficking - learning from single-virus tracking. Nat Rev Microbiol, 2007. 5(3): p. 197-208.

86. Mueller, S., Wimmer, E., and Cello, J. Poliovirus and poliomyelitis: a tale of guts, brains, and an accidental event. Virus Research, 2005. 111(2): p. 175-93.

81 87. Spear, P.G. Entry of alphaherpesviruses into cells. Seminars in Virology, 1993(4): p. 167-180.

88. Warner, M.S., Geraghty, R.J., Martinez, W.M., Montgomery, R.I., Whitbeck, J.C., Xu, R., Eisenberg, R.J., Cohen, G.H., and Spear, P.G. A cell surface protein with herpesvirus entry activity (HveB) confers susceptibility to infection by mutants of herpes simplex virus type 1, herpes simplex virus type 2, and pseudorabies virus. Virology, 1998. 246(1): p. 179-89.

89. Nicola, A.V., Ponce de Leon, M., Xu, R., Hou, W., Whitbeck, J.C., Krummenacher, C., Montgomery, R.I., Spear, P.G., Eisenberg, R.J., and Cohen, G.H. Monoclonal antibodies to distinct sites on herpes simplex virus (HSV) glycoprotein D block HSV binding to HVEM. Journal of virology, 1998. 72(5): p. 3595-601.

90. Whitbeck, J.C., Peng, C., Lou, H., Xu, R., Willis, S.H., Ponce de Leon, M., Peng, T., Nicola, A.V., Montgomery, R.I., Warner, M.S., Soulika, A.M., Spruce, L.A., Moore, W.T., Lambris, J.D., Spear, P.G., Cohen, G.H., and Eisenberg, R.J. Glycoprotein D of herpes simplex virus (HSV) binds directly to HVEM, a member of the tumor necrosis factor receptor superfamily and a mediator of HSV entry. Journal of virology, 1997. 71(8): p. 6083-93.

91. Geraghty, R.J., Krummenacher, C., Cohen, G.H., Eisenberg, R.J., and Spear, P.G. Entry of alphaherpesviruses mediated by poliovirus receptor-related protein 1 and poliovirus receptor. Science, 1998. 280(5369): p. 1618-20.

92. Cocchi, F., Lopez, M., Menotti, L., Aoubala, M., Dubreuil, P., and Campadelli-Fiume, G. The V domain of herpesvirus Ig-like receptor (HIgR) contains a major functional region in herpes simplex virus-1 entry into cells and interacts physically with the viral glycoprotein D. Proc Natl Acad Sci USA, 1998. 95(26): p. 15700-5.

93. Kwon, H., Bai, Q., Baek, H.-J., Felmet, K., Burton, E.A., Goins, W.F., Cohen, J.B., and Glorioso, J.C. Soluble V domain of Nectin-1/HveC enables entry of herpes simplex virus type 1 (HSV-1) into HSV-resistant cells by binding to viral glycoprotein D. Journal of virology, 2006. 80(1): p. 138-48.

94. Noyce, R.S., Bondre, D.G., Ha, M.N., Lin, L.-T., Sisson, G., Tsao, M.-S., and Richardson, C.D. Tumor cell marker PVRL4 (nectin 4) is an epithelial cell receptor for measles virus. PLoS Pathog, 2011. 7(8): p. e1002240.

95. Mühlebach, M.D., Mateo, M., Sinn, P.L., Prüfer, S., Uhlig, K.M., Leonard, V.H.J., Navaratnarajah, C.K., Frenzke, M., Wong, X.X., Sawatsky, B., Ramachandran, S., Mccray, P.B., Cichutek, K., von Messling, V., Lopez, M., and Cattaneo, R. Adherens junction protein nectin-4 is the epithelial receptor for measles virus. Nature, 2011. 480(7378): p. 530-3.

96. Navaratnarajah, C.K., Leonard, V.H.J., and Cattaneo, R. Measles virus glycoprotein complex assembly, receptor attachment, and cell entry. Current topics in microbiology and immunology, 2009. 329: p. 59-76.

82 97. Dörig, R.E., Marcil, A., Chopra, A., and Richardson, C.D. The human CD46 molecule is a receptor for measles virus (Edmonston strain). Cell, 1993. 75(2): p. 295-305.

98. Naniche, D., Varior-Krishnan, G., Cervoni, F., Wild, T.F., Rossi, B., Rabourdin-Combe, C., and Gerlier, D. Human membrane cofactor protein (CD46) acts as a cellular receptor for measles virus. Journal of virology, 1993. 67(10): p. 6025-32.

99. Tatsuo, H., Ono, N., Tanaka, K., and Yanagi, Y. SLAM (CDw150) is a cellular receptor for measles virus. Nature, 2000. 406(6798): p. 893-7.

100. Hsu, E.C., Iorio, C., Sarangi, F., Khine, A.A., and Richardson, C.D. CDw150(SLAM) is a receptor for a lymphotropic strain of measles virus and may account for the immunosuppressive properties of this virus. Virology, 2001. 279(1): p. 9-21.

101. Strauss, M., Filman, D.J., Belnap, D.M., Cheng, N., Noel, R.T., and Hogle, J.M. Nectin- Like Interactions between Poliovirus and Its Receptor Trigger Conformational Changes Associated with Cell Entry. Journal of virology, 2015. 89(8): p. 4143-4157.

102. Zhang, P., Mueller, S., Morais, M.C., Bator, C.M., Bowman, V.D., Hafenstein, S., Wimmer, E., and Rossmann, M.G. Crystal structure of CD155 and electron microscopic studies of its complexes with polioviruses. Proc Natl Acad Sci USA, 2008. 105(47): p. 18284-9.

103. Di Giovine, P., Settembre, E.C., Bhargava, A.K., Luftig, M.A., Lou, H., Cohen, G.H., Eisenberg, R.J., Krummenacher, C., and Carfi, A. Structure of herpes simplex virus glycoprotein D bound to the human receptor nectin-1. PLoS Pathog, 2011. 7(9): p. e1002277.

104. Zhang, X., Lu, G., Qi, J., Li, Y., He, Y., Xu, X., Shi, J., Zhang, C.W.-H., Yan, J., and Gao, G.F. Structure of measles virus hemagglutinin bound to its epithelial receptor nectin-4. Nat Struct Mol Biol, 2013. 20(1): p. 67-72.

105. Harrison, O.J., Vendome, J., Brasch, J., Jin, X., Hong, S., Katsamba, P.S., Ahlsen, G., Troyanovsky, R.B., Troyanovsky, S.M., Honig, B., and Shapiro, L. Nectin ectodomain structures reveal a canonical adhesive interface. Nat Struct Mol Biol, 2012. 19(9): p. 906-15.

106. Takai, Y., Irie, K., Shimizu, K., Sakisaka, T., and Ikeda, W. Nectins and nectin-like molecules: roles in , migration, and polarization. Cancer science, 2003. 94(8): p. 655-67.

107. Ikeda, W., Kakunaga, S., Itoh, S., Shingai, T., Takekuni, K., Satoh, K., Inoue, Y., Hamaguchi, A., Morimoto, K., Takeuchi, M., Imai, T., and Takai, Y. Tage4/Nectin-like molecule-5 heterophilically trans-interacts with Nectin-3 and enhances cell migration. J Biol Chem, 2003. 278(30): p. 28167-72.

108. Fujito, T., Ikeda, W., Kakunaga, S., Minami, Y., Kajita, M., Sakamoto, Y., Monden, M., and Takai, Y. Inhibition of cell movement and proliferation by cell-cell contact-induced interaction of Necl-5 with nectin-3. The Journal of Cell Biology, 2005. 171(1): p. 165-73.

83 109. Takai, Y., Ikeda, W., Ogita, H., and Rikitake, Y. The immunoglobulin-like cell adhesion molecule nectin and its associated protein afadin. Annu Rev Cell Dev Biol, 2008. 24: p. 309-42.

110. Mandai, K., Rikitake, Y., Mori, M., and Takai, Y. Nectins and Nectin-Like Molecules in Development and Disease. Current Topics in Developmental Biology, 2015. 112: p. 197- 231.

111. Miyata, M., Rikitake, Y., Takahashi, M., Nagamatsu, Y., Yamauchi, Y., Ogita, H., Hirata, K.-I., and Takai, Y. Regulation by Afadin of Cyclical Activation and Inactivation of Rap1, Rac1, and RhoA Small G Proteins at Leading Edges of Moving NIH3T3 Cells. J Biol Chem, 2009. 284(36): p. 24595-24609.

112. Amano, H., Ikeda, W., Kawano, S., Kajita, M., Tamaru, Y., Inoue, N., Minami, Y., Yamada, A., and Takai, Y. Interaction and localization of Necl-5 and PDGF receptor beta at the leading edges of moving NIH3T3 cells: Implications for directional cell movement. Genes to cells : devoted to molecular & cellular mechanisms, 2008. 13(3): p. 269-84.

113. Ikeda, W., Kakunaga, S., Takekuni, K., Shingai, T., Satoh, K., Morimoto, K., Takeuchi, M., Imai, T., and Takai, Y. Nectin-like molecule-5/Tage4 enhances cell migration in an integrin-dependent, Nectin-3-independent manner. J Biol Chem, 2004. 279(17): p. 18015-25.

114. Kinugasa, M., Amano, H., Satomi-Kobayashi, S., Nakayama, K., Miyata, M., Kubo, Y., Nagamatsu, Y., Kurogane, Y., Kureha, F., Yamana, S., Hirata, K.-i., Miyoshi, J., Takai, Y., and Rikitake, Y. Necl-5/poliovirus receptor interacts with VEGFR2 and regulates VEGF-induced angiogenesis. Circulation research, 2012. 110(5): p. 716-26.

115. Inagaki, M., Irie, K., Ishizaki, H., Tanaka-Okamoto, M., Morimoto, K., Inoue, E., Ohtsuka, T., Miyoshi, J., and Takai, Y. Roles of cell-adhesion molecules nectin 1 and nectin 3 in ciliary body development. Development, 2005. 132(7): p. 1525-37.

116. Miyoshi, J. and Takai, Y. Nectin and nectin-like molecules: biology and pathology. Am J Nephrol, 2007. 27(6): p. 590-604.

117. Lachke, S.A., Higgins, A.W., Inagaki, M., Saadi, I., Xi, Q., Long, M., Quade, B.J., Talkowski, M.E., Gusella, J.F., Fujimoto, A., Robinson, M.L., Yang, Y., Duong, Q.T., Shapira, I., Motro, B., Miyoshi, J., Takai, Y., Morton, C.C., and Maas, R.L. The cell adhesion gene PVRL3 is associated with congenital ocular defects. Human genetics, 2012. 131(2): p. 235-50.

118. Qa'Dan, M., Ramsey, M., Daniel, J., Spyres, L.M., Safiejko-Mroczka, B., Ortiz-Leduc, W., and Ballard, J.D. Clostridium difficile toxin B activates dual caspase-dependent and caspase-independent apoptosis in intoxicated cells. Cell Microbiol, 2002. 4(7): p. 425-34.

119. Ivie, S.E., Fennessey, C.M., Sheng, J., Rubin, D.H., and McClain, M.S. Gene-Trap Mutagenesis Identifies Mammalian Genes Contributing to Intoxication by Clostridium perfringens ε-Toxin. PLoS ONE, 2011. 6(3): p. e17787.

84 120. Osipovich, A.B., White-Grindley, E.K., Hicks, G.G., Roshon, M.J., Shaffer, C., Moore, J.H., and Ruley, H.E. Activation of cryptic 3' splice sites within introns of cellular genes following gene entrapment. Nucleic acids research, 2004. 32(9): p. 2912-24.

121. Leffler, H., Carlsson, S., Hedlund, M., Qian, Y., and Poirier, F. Introduction to galectins. Glycoconj J, 2004. 19(7-9): p. 433-40.

122. Ochieng, J., Furtak, V., and Lukyanov, P. Extracellular functions of galectin-3. Glycoconj J, 2004. 19(7-9): p. 527-35.

123. Magnaldo, T., Bernerd, F., and Darmon, M. Galectin-7, a human 14-kDa S-lectin, specifically expressed in keratinocytes and sensitive to retinoic acid. Dev Biol, 1995. 168(2): p. 259-71.

124. Magnaldo, T., Fowlis, D., and Darmon, M. Galectin-7, a marker of all types of stratified epithelia. Differentiation, 1998. 63(3): p. 159-68.

125. Uhlen, M., Fagerberg, L., Hallstrom, B.M., Lindskog, C., Oksvold, P., Mardinoglu, A., Sivertsson, A., Kampf, C., Sjostedt, E., Asplund, A., Olsson, I., Edlund, K., Lundberg, E., Navani, S., Szigyarto, C.A.-K., Odeberg, J., Djureinovic, D., Takanen, J.O., Hober, S., Alm, T., Edqvist, P.-H., Berling, H., Tegel, H., Mulder, J., Rockberg, J., Nilsson, P., Schwenk, J.M., Hamsten, M., Von Feilitzen, K., Forsberg, M., Persson, L., Johansson, F., Zwahlen, M., Von Heijne, G., Nielsen, J., and Ponten, F. Tissue-based map of the human proteome. Science, 2015. 347(6220): p. 1260419-1260419.

126. Lotz, M.M., Andrews, C.W., Korzelius, C.A., Lee, E.C., Steele, G.D., Clarke, A., and Mercurio, A.M. Decreased expression of Mac-2 (carbohydrate binding protein 35) and loss of its nuclear localization are associated with the neoplastic progression of colon carcinoma. Proc Natl Acad Sci USA, 1993. 90(8): p. 3466-70.

127. Bachhawat-Sikder, K., Thomas, C.J., and Surolia, A. Thermodynamic analysis of the binding of galactose and poly-N-acetyllactosamine derivatives to human galectin-3. FEBS Letters, 2001. 500(1-2): p. 75-9.

128. Petropolis, D.B., Faust, D.M., Deep Jhingan, G., and Guillen, N. A new human 3D-liver model unravels the role of galectins in liver infection by the parasite Entamoeba histolytica. PLoS Pathog, 2014. 10(9): p. e1004381.

129. Murray, J.L., Mavrakis, M., McDonald, N.J., Yilla, M., Sheng, J., Bellini, W.J., Zhao, L., Le Doux, J.M., Shaw, M.W., Luo, C.-C., Lippincott-Schwartz, J., Sanchez, A., Rubin, D.H., and Hodge, T.W. Rab9 GTPase is required for replication of human immunodeficiency virus type 1, filoviruses, and measles virus. Journal of virology, 2005. 79(18): p. 11742-51.

130. Pruitt, R.N., Chumbler, N.M., Rutherford, S.A., Farrow, M.A., Friedman, D.B., Spiller, B., and Lacy, D.B. Structural determinants of Clostridium difficile toxin A glucosyltransferase activity. J Biol Chem, 2012. 287(11): p. 8013-20.

85 131. Kuehne, S.A., Cartman, S.T., Heap, J.T., Kelly, M.L., Cockayne, A., and Minton, N.P. The role of toxin A and toxin B in Clostridium difficile infection. Nature, 2010. 467(7316): p. 711-713.

132. Chaves-Olarte, E., Weidmann, M., Eichel-Streiber, C., and Thelestam, M. Toxins A and B from Clostridium difficile differ with respect to enzymatic potencies, cellular substrate specificities, and surface binding to cultured cells. Journal of Clinical Investigation, 1997. 100(7): p. 1734-41.

133. Frisch, C., Gerhard, R., Aktories, K., Hofmann, F., and Just, I. The complete receptor- binding domain of Clostridium difficile toxin A is required for endocytosis. Biochem Biophys Res Commun, 2003. 300(3): p. 706-11.

134. Frey, S.M. and Wilkins, T.D. Localization of two epitopes recognized by monoclonal antibody PCG-4 on Clostridium difficile toxin A. Infection and immunity, 1992. 60(6): p. 2488-92.

135. Orth, P., Xiao, L., Hernandez, L.D., Reichert, P., Sheth, P.R., Beaumont, M., Yang, X., Murgolo, N., Ermakov, G., DiNunzio, E., Racine, F., Karczewski, J., Secore, S., Ingram, R.N., Mayhood, T., Strickland, C., and Therien, A.G. Mechanism of action and epitopes of Clostridium difficile toxin B-neutralizing antibody bezlotoxumab revealed by X-ray crystallography. J Biol Chem, 2014. 289(26): p. 18008-21.

136. Sauerborn, M., Leukel, P., and von Eichel-Streiber, C. The C-terminal ligand-binding domain of Clostridium difficile toxin A (TcdA) abrogates TcdA-specific binding to cells and prevents mouse lethality. FEMS Microbiol Lett, 1997. 155(1): p. 45-54.

137. Carette, J.E., Guimaraes, C.P., Varadarajan, M., Park, A.S., Wuethrich, I., Godarova, A., Kotecki, M., Cochran, B.H., Spooner, E., Ploegh, H.L., and Brummelkamp, T.R. Haploid genetic screens in human cells identify host factors used by pathogens. Science, 2009. 326(5957): p. 1231-5.

138. Du, T. and Alfa, M.J. Translocation of Clostridium difficile toxin B across polarized Caco- 2 cell monolayers is enhanced by toxin A. The Canadian journal of infectious diseases = Journal canadien des maladies infectieuses, 2004. 15(2): p. 83-8.

139. Meunier, V., Bourrié, M., Berger, Y., and Fabre, G. The human intestinal epithelial cell line Caco-2; pharmacological and pharmacokinetic applications. Cell Biol Toxicol, 1995. 11(3-4): p. 187-94.

140. Cocchi, F., Menotti, L., Mirandola, P., Lopez, M., and Campadelli-Fiume, G. The ectodomain of a novel member of the immunoglobulin subfamily related to the poliovirus receptor has the attributes of a bona fide receptor for herpes simplex virus types 1 and 2 in human cells. Journal of virology, 1998. 72(12): p. 9992-10002.

141. Delpeut, S., Noyce, R.S., and Richardson, C.D. The tumor-associated marker, PVRL4 (nectin-4), is the epithelial receptor for morbilliviruses. Viruses, 2014. 6(6): p. 2268-86.

142. Pratakpiriya, W., Seki, F., Otsuki, N., Sakai, K., Fukuhara, H., Katamoto, H., Hirai, T., Maenaka, K., Techangamsuwan, S., Lan, N.T., Takeda, M., and Yamaguchi, R. Nectin4

86 is an epithelial cell receptor for canine distemper virus and involved in neurovirulence. Journal of virology, 2012. 86(18): p. 10207-10.

143. Satoh-Horikawa, K., Nakanishi, H., Takahashi, K., Miyahara, M., Nishimura, M., Tachibana, K., Mizoguchi, A., and Takai, Y. Nectin-3, a new member of immunoglobulin- like cell adhesion molecules that shows homophilic and heterophilic cell-cell adhesion activities. J Biol Chem, 2000. 275(14): p. 10291-9.

144. Togashi, H., Kominami, K., Waseda, M., Komura, H., Miyoshi, J., Takeichi, M., and Takai, Y. Nectins establish a checkerboard-like cellular pattern in the auditory epithelium. Science, 2011. 333(6046): p. 1144-7.

145. Barton, E.S., Forrest, J.C., Connolly, J.L., Chappell, J.D., Liu, Y., Schnell, F.J., Nusrat, A., Parkos, C.A., and Dermody, T.S. Junction adhesion molecule is a receptor for reovirus. Cell, 2001. 104(3): p. 441-51.

146. Bergelson, J.M. Intercellular junctional proteins as receptors and barriers to virus infection and spread. Cell host & microbe, 2009. 5(6): p. 517-21.

147. Guttman, J.A. and Finlay, B.B. Tight junctions as targets of infectious agents. Biochimica et biophysica acta, 2009. 1788(4): p. 832-41.

148. Katahira, J., Sugiyama, H., Inoue, N., Horiguchi, Y., Matsuda, M., and Sugimoto, N. Clostridium perfringens enterotoxin utilizes two structurally related membrane proteins as functional receptors in vivo. J Biol Chem, 1997. 272(42): p. 26652-8.

149. Pentecost, M., Otto, G., Theriot, J.A., and Amieva, M.R. Listeria monocytogenes invades the epithelial junctions at sites of cell extrusion. PLoS Pathog, 2006. 2(1): p. e3.

150. Hecht, G., Pothoulakis, C., LaMont, J.T., and Madara, J.L. Clostridium difficile toxin A perturbs cytoskeletal structure and tight junction permeability of cultured human intestinal epithelial monolayers. Journal of Clinical Investigation, 1988. 82(5): p. 1516-24.

151. Hecht, G., Koutsouris, A., Pothoulakis, C., LaMont, J.T., and Madara, J.L. Clostridium difficile toxin B disrupts the barrier function of T84 monolayers. Gastroenterology, 1992. 102(2): p. 416-23.

152. Coyne, C.B. and Bergelson, J.M. Virus-induced Abl and Fyn kinase signals permit coxsackievirus entry through epithelial tight junctions. Cell, 2006. 124(1): p. 119-31.

153. Henriques, B., Florin, I., and Thelestam, M. Cellular internalisation of Clostridium difficile toxin A. Microb Pathog, 1987. 2(6): p. 455-63.

154. Terada, N., Ohno, N., Murata, S., Katoh, R., Stallcup, W.B., and Ohno, S. Immunohistochemical study of NG2 chondroitin sulfate proteoglycan expression in the small and large intestines. Histochemistry and cell biology, 2006. 126(4): p. 483-90.

155. Ran, F.A., Hsu, P.D., Wright, J., Agarwala, V., Scott, D.A., and Zhang, F. Genome engineering using the CRISPR-Cas9 system. Nat Protoc, 2013. 8(11): p. 2281-308.

87 156. LaFrance, M.E., Farrow, M.A., Chandrasekaran, R., Sheng, J., Rubin, D.H., and Lacy, D.B. Identification of an epithelial cell receptor responsible for Clostridium difficile TcdB- induced cytotoxicity. Proc Natl Acad Sci USA, 2015. 112(22): p. 7073-8.

157. Rotem-Yehudar, R., Galperin, E., and Horowitz, M. Association of insulin-like growth factor 1 receptor with EHD1 and SNAP29. J Biol Chem, 2001. 276(35): p. 33054-60.

158. Jović, M., Naslavsky, N., Rapaport, D., Horowitz, M., and Caplan, S. EHD1 regulates beta1 integrin endosomal transport: effects on focal adhesions, cell spreading and migration. Journal of cell science, 2007. 120(Pt 5): p. 802-14.

159. Kitt, K.N. and Nelson, W.J. Rapid Suppression of Activated Rac1 by and Nectins during De Novo Cell-Cell Adhesion. PloS ONE, 2011. 6(3): p. e17841.

160. Takano, A., Ishikawa, N., Nishino, R., Masuda, K., Yasui, W., Inai, K., Nishimura, H., Ito, H., Nakayama, H., Miyagi, Y., Tsuchiya, E., Kohno, N., Nakamura, Y., and Daigo, Y. Identification of Nectin-4 Oncoprotein as a Diagnostic and Therapeutic Target for Lung Cancer. Cancer research, 2009. 69(16): p. 6694-6703.

161. Zhang, S.-C., Wang, W.-L., Cai, W.-S., Jiang, K.-L., and Yuan, Z.-W. Engineered measles virus Edmonston strain used as a novel oncolytic viral system against human hepatoblastoma. BMC cancer, 2012. 12: p. 427.

88