Identification of an epithelial cell receptor responsible for Clostridium difficile TcdB-induced cytotoxicity

Significance

Clostridium difficile is a toxin-producing bacterium that is a frequent cause of hospital-acquired and antibiotic-associated diarrhea. The incidence, severity, and costs associated with C. difficile infection (CDI) are increasing, making C. difficile an important public health concern. As a toxin-mediated disease, there is significant interest in understanding the receptors that mediate the cellular entry and function of these toxins. The targeted disruption of toxin-receptor interactions could provide novel therapeutic strategies that can either augment or replace the need for antibiotic therapies in the treatment of CDI.

Abstract

Clostridium difficile is the leading cause of hospital-acquired diarrhea in the United States. The two main virulence factors of C. difficile are the large toxins, TcdA and TcdB, which enter colonic epithelial cells and cause fluid secretion, inflammation, and cell death. Using a gene-trap insertional mutagenesis screen, we identified poliovirus receptor-like 3 (PVRL3) as a cellular factor necessary for TcdB-mediated cytotoxicity. Disruption of PVRL3 expression by gene-trap mutagenesis, shRNA, or CRISPR/Cas9 mutagenesis resulted in resistance of cells to TcdB. Complementation of the gene-trap or CRISPR mutants with PVRL3 resulted in restoration of TcdB-mediated cell death. Purified PVRL3 ectodomain bound to TcdB by pull-down. Pretreatment of cells with a monoclonal antibody against PVRL3 or prebinding TcdB to PVRL3 ectodomain also inhibited cytotoxicity in cell culture. The receptor is highly expressed on the surface epithelium of the human colon and was observed to colocalize with TcdB in both an explant model and in tissue from a patient with pseudomembranous colitis. These data suggest PVRL3 is a physiologically relevant binding partner that can serve as a target for the prevention of TcdB-induced cytotoxicity in C. difficile infection.

Clostridium difficile infection (CDI) is the leading cause of antibiotic-associated diarrhea and pseudomembranous colitis in the United States (1, 2). Over the past decade, morbidity and lethality from CDI have increased (3, 4), and the need for new treatment options has become a priority.

The pathology associated with CDI is associated with the activities of two large, glucosylating toxins, TcdA and TcdB (5). Upon binding to the colonic epithelium, these toxins induce the fluid secretion, immune cell influx, and tissue damage associated with clinical manifestations of CDI (5). TcdA and TcdB have four functional domains: an N-terminal glucosyltransferase domain (GTD), an autoprotease domain, a pore-forming and delivery domain, and a combined repetitive oligopeptides (CROPS) domain, which extends from around residue 1830 to the C terminus and has been implicated in receptor binding. The toxins enter cells by receptor-mediated endocytosis (6). Acidification of the endosome is thought to trigger a structural change in the delivery domain, allowing for pore formation and translocation of the GTD into the cytosol (7, 8). Activation of the autoprocessing domain by eukaryotic inositol hexakisphosphate results in the release of the GTD into the cell, allowing access to substrates (8). The GTD transfers a glucose from UDP glucose onto the switch I region of Rho family GTPases such as Rho, Rac1, and Cdc42 (9, 10). These modifications cause a cytopathic effect resulting from rearrangement of the actin cytoskeleton and can lead to apoptosis (11). At higher concentrations, TcdB is also capable of inducing the production of reactive oxygen species, resulting in cell death by a necrotic mechanism (12, 13). 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 (14, 15). TcdA and TcdB are thought to have different receptors, based on sensitivity differences among cell types in vitro (16–19). Multiple receptors for TcdA have been proposed including Gal alpha 1–3Gal beta 1–4GlcNAc, blood antigens I, X, and Y, rabbit sucrase isomaltase, and gp96 (18, 20–22). The TcdA CROPS domain is thought to play a role in binding cell surface carbohydrates (18, 23, 24). Antibodies against the CROPS domains of TcdA and TcdB can block intoxication (25, 26), and excess TcdA CROPS domain can compete with TcdA holotoxin for cell binding (27). At the same time, truncations of TcdA and TcdB that lack the CROPS domains are still capable of intoxicating cells (7, 28, 29) and a homologous toxin from Clostridium perfringens, TpeL, lacks a CROPS domain entirely (29, 30). A receptor for TpeL has been identified (29), suggesting that a receptor-binding site for other large clostridial toxins could exist outside of the CROPS. A recent report indicates that chondroitin sulfate proteoglycan 4 (CSPG4) mediates TcdB-induced cytopathic and apoptotic events in HeLa and HT29 cells (31). 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 factor(s) responsible for TcdB binding and toxicity.

Results

To identify cellular factors involved in TcdB-mediated cytotoxicity, we used a genetic selection to isolate toxin-resistant Caco-2 cells. First, Caco-2 cells were mutagenized with a retroviral gene-trap vector that inserts a promoterless neomycin gene at random locations throughout the chromosome (32). 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 could be involved in 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 when measuring ATP as a viability indicator, and sensitivity could be restored through complementation with a lentiviral PVRL3 overexpression plasmid (Fig. 1A). The knockdown of PVRL3 expression was incomplete, however (Fig. 1B). The gene-trap insertion occurred within the promoter of the pvrl3 locus, leaving the coding region of both alleles intact. To confirm the results obtained with 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 Western blot (Fig. 1C) and a significant increase in cell survival at both an 18-h (Fig. 1D) and 48-h (see Fig. S2A) time point. These data suggest that PVRL3 expression is necessary 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 (Fig. S1), the PVRL3 knockdown cells were still sensitive to the TcdA-induced apoptotic cell death that occurs after 48 h (Fig. 1E).

Fig. 1.

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) Western blot of whole-cell lysates from cells used in D. (D) Caco-2 cells were stably transduced with four different shRNAs targeting pvrl3 transcript or a nontargeting shRNA (control). Cells were seeded and intoxicated, and viability was measured and plotted as in A. Results represent the mean ± SEM of six experiments. (E) Caco-2 cells transduced with the nontargeting shRNA (control) and two PVRL3 shRNAs from D were plated and treated with serial dilutions of TcdA. ATP levels were measured at 48 h 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 D and E.

We next wanted to evaluate 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 (Fig. 2B, Inset) resulted in increased cell survival at a 4.5-h time point, as indicated by ATP concentration (Fig. 2A) and lactate dehydrogenase (LDH) release, an alternative indicator of TcdB-mediated necrosis (Fig. 2B). We also disrupted the expression of PVRL3 in HeLa cells using the CRISPR/Cas9 system (Fig. 2C, Inset). The CRISPR-mutated HeLa cells were significantly less sensitive to TcdB, whereas the complemented mutants were as sensitive as the wild-type cells (Fig. 2C).

Fig. 2.

Disruption of PVRL3 in HeLa results in cell survival. (A) HeLa cells were transduced with two different shRNAs targeting pvrl3 or a nontargeting shRNA (control). Cells were seeded and intoxicated for 4.5 h. Viability was assayed as in Fig. 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. (Inset) Western blot of whole-cell lysates of cells used in A and B probing for PVRL3 and GAPDH (loading control). (C) 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. (Inset) Western blot of whole cell lysates of cells used in C. 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.

Unlike in Caco-2 cells (Fig. S2A), knockdown of PVRL3 expression in HeLas did not confer resistance to the apoptotic events that occur when cells are exposed to toxin for 48 h (Fig. S2B). Furthermore, the knockdown did not confer resistance to TcdB-induced cytopathic “rounding” events (Fig. S3). This finding is consistent with observations from Yuan et al. that CSPG4 plays a role in TcdB-mediated cytopathic and apoptosis effects in HeLa cells (31). It is also consistent with our observation that, whereas PVRL3 is expressed in both cell types, CSPG4 is only expressed in HeLa cells (Fig. S2C).

As PVRL3 is a membrane-bound protein with three extracellular Ig domains, we hypothesized that TcdB could use PVRL3 as a binding partner on the surface of cells. To test whether TcdB can directly bind to PVRL3, pull-down assays with a His-tagged extracellular domain of PVRL3 were performed. TcdB, TcdB_1–1834, TcdB CROPS, and TcdA_1–1832 were biotinylated, conjugated to streptavidin–agarose beads, and incubated with purified PVRL3_1–359–His₆ ectodomain. PVRL3_1–359–His₆ coeluted with full-length TcdB and TcdB_1–1834, but not with TcdB CROPS, TcdA_1–1832, or empty beads (Fig. 3A). The TcdA CROPS is known to bind a broad array of carbohydrates (17). Full-length TcdA binds to PVRL3 but also a number of unrelated glycosylated proteins (Fig. S4). The unrelated junctional adhesion molecule (JAM-A) ectodomain (residues 1–235) did not bind any of the TcdB constructs (Fig. 3B), indicating a specific interaction between TcdB and PVRL3.

Fig. 3.

PVRL3 binds directly to TcdB. Purified recombinant TcdB holotoxin, TcdB_1–1834, TcdB CROPS, and TcdA_1–1832 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.

After demonstrating that TcdB could bind directly to PVRL3, we wanted to determine if 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 plated and incubated with the anti-PVRL3 antibody or an unrelated isotype control antibody, then challenged with toxin for 18 h (Fig. 4A). 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.

Fig. 4.

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 Fig. 1. (B) TcdB was prebound to purified recombinant PVRL1, PVRL2, or PVRL3 ectodomain 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.

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 their ability to inhibit cytotoxicity. The glycosylated ectodomains of human PVRL1, PVRL2, and PVRL3 were expressed and purified from the media supernatant of transfected HEK293 cells, incubated with TcdB, and applied to HeLa cells for 4.5 h (Fig. 4B). 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.

Finally, we wanted to evaluate whether PVRL3 and/or CSPG4 would be present and accessible for TcdB binding in a physiologically relevant context. A healthy section of colonic tissue, freshly excised from a human donor, was stained with anti-PVRL3 and anti-CSPG4 antibodies. Whereas PVRL3 was clearly visible on the surface epithelium, CSPG4 was not detected in these samples (Fig. S5), confirming previous observations made that CSPG4 is not expressed on the surface epithelium of the colon (33). Similar tissue samples were treated with a Flag-tagged version of TcdB, fixed, and then stained with anti-FLAG and anti-PVRL3 antibodies. We observed significant overlap between PVRL3 and the signal from anti-FLAG, suggesting that TcdB and PVRL3 colocalize in tissue (Fig. 5A). Tissues that were not exposed to toxin or primary antibodies did not stain similarly (Fig. S6). Likewise, in a sample from a clinical patient presenting with C. difficile-associated pseudomembranous colitis, an antibody specific to TcdB revealed that the TcdB present from the course of infection colocalized with PVRL3 (Fig. 5B). These data, showing that the PVRL3–TcdB interaction can be observed in tissue obtained from a C. difficile-infected individual, demonstrate the physiological relevance of this interaction.

Fig. 5.

TcdB and PVRL3 colocalize in human colonic explants. (A) Human colonic explants were treated with 10 nM FLAG–TcdB for 3 h, formalin fixed, and unstained sections were mounted. Sections were stained with FLAG or PVRL3 and imaged at 63×. The overlay image represents FLAG (green) and PVRL3 (red). The 63× images were further magnified and are shown as the zoom panels. (B) A human colonic sample from a patient infected with C. difficile was fixed and unstained sections were mounted. Sections were stained with antibodies against TcdB or PVRL3. Overlay represents TcdB (green) and PVRL3 (red). Image magnification and zoom are as in A.

Discussion

The goal for this study was to identify a host cell receptor for TcdB. Whereas numerous studies have emphasized the use of haploid cell lines for identifying pathogen host factors (34), we wanted to start with Caco-2 cells because numerous studies have documented their utility as a cellular model for TcdB intoxication (35) and the intestinal epithelium in general (36). Because we had previously succeeded in identifying host factors important for C. perfringens ε-toxin intoxication by constructing a gene-trap library in Madin Darby canine kidney (MDCK) cells (37), we adopted a similar strategy in Caco-2 cells. We used a high concentration of toxin (15 nM) and selected for clones that were resistant to cell death. In contrast, the study that resulted in the identification of CSPG4 as a TcdB receptor was conducted with low concentrations of toxin and relied on the selection of shRNA-disrupted HeLa cells that were resistant to cytopathic effects (31).

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 (Fig. 1A), shRNA knockdown (Figs. 1D and 2 A and B), or CRISPR/Cas9 mutagenesis (Fig. 2C) confers resistance to TcdB-induced cell death. Furthermore, ectopic expression of PVRL3 restores the sensitivity of the E19 Caco-2 (Fig. 1A) or CRISPR-mutated HeLa cells (Fig. 2C) 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 (Fig. 3). This binding interaction can be interrupted on cells with either an anti-PVRL3 antibody (Fig. 4A) or purified PVRL3 ectodomain (Fig. 4B) to provide protection against TcdB-induced cytotoxicity. Furthermore, we show that PVRL3 is highly expressed on the epithelial surface of the human colon and that PVRL3 and TcdB colocalize in toxin-treated healthy tissue and in tissue derived from a C. difficile-infected individual (Fig. 5). 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) (38). These proteins consist of three extracellular Ig-like domains, a single transmembrane helix, and a short cytoplasmic domain that binds afadin. They are related to the poliovirus receptor, PVR, (also called Necl-5/Tage4/CD155) which has a similar sequence structure but does not bind afadin. Many of the PVR/PVRL family members have been shown to be involved in virus entry. PVR was first identified as the poliovirus receptor (39), followed by PVRL1 and PVRL2 as herpesvirus receptors (40, 41), and most recently PVRL4 has been identified as an epithelial receptor for measles virus (42, 43) and other morbilliviruses (44, 45). This study represents the first report to our knowledge of a pathogen using PVRL3 as a receptor, and also the first instance to our knowledge of a bacterial toxin using a member of this family of proteins as a receptor.

The nectins can form homotypic and heterotypic trans dimers, which contribute to adherens junction formation in many vertebrate tissues (46). Individual subtypes are typically expressed in distinct but overlapping patterns that, in the case of the auditory epithelium, lead to a checkerboard cellular pattern (47). PVRL3 can form heterotypic adhesions with PVRL1, PVRL2, and PVR. As with viruses and other toxins that exploit proteins localized in junctions (48–51), 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 (52) or through toxin-mediated disruption of the tight junctions (48, 53, 54). On the other hand, the accessibility of PVRL3 may not be as limited as we would have initially assumed. We note that in our immunohistochemistry (IHC) staining, PVRL3 is highly expressed and not limited just to junctions (Fig. 5).

A related question is how PVRL3 binding contributes to cytotoxicity. TcdB cytotoxicity depends on access to an acidified endosome (55, 56), a process which is mediated by clathrin-mediated endocytosis (6). The interaction of PVR with PVRL3 has been shown to cause down-regulation of PVR from the cell surface through a clathrin-mediated endocytic mechanism (57). Thus, PVRL3 is capable of mediating endocytic entry in response to ligand binding. We propose that a similar entry mechanism takes place for TcdB.

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 (29). Our data are compatible with this model, as binding to PVRL3 is independent of the CROPS (Fig. 4). Also, we observed that HeLa cells with depleted levels of PVRL3 were still susceptible to TcdB-induced cytopathic rounding effects (Fig. S3) and the apoptosis events that can be detected at 48 h (Fig. S2B). The report that CSPG4 mediates TcdB-induced rounding of HeLa cells is consistent with the idea of a second receptor, although we note that both PVRL3 and CSPG4 bind TcdB outside of the CROPS domain (31). Furthermore, we were not able to detect CSPG4 expression in either Caco-2 cells (Fig. S2C) or human colonic tissue (Fig. S5). Unlike HeLa cells, Caco-2 cells that were disrupted in PVRL3 expression were resistant to the TcdB-induced cell death events that occur when cells are exposed to lower concentrations of toxin for 48 h (Fig. S2A). More work is needed to address whether there is a specific cellular host factor that binds the TcdB CROPS, and the role of these receptors in animal infection models.

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 and/or biologics that block this interaction.

Methods

Reagents and Cell Culture.

Detailed descriptions of the cloning, protein purification, and cell culture methods are included in SI Methods.

Genetrap.

The U3neoSV1 retroviral vector (58) was obtained from Zirus. Caco-2 cells were plated in 75-cm² flasks and incubated with U3neoSV1 (multiplicity of infection, MOI = 0.1) at 37 °C for 1 h in the presence of 4 µg/mL Polybrene. Transduced cells were selected for using G418 at a concentration of 750 µg/mL and grown to confluence. 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 kit (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 Escherichia coli, and growth on selective media.

Viability Assays.

Cells were seeded at a density of 2,500 cells per 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 postintoxication with Cell-Titer-Glo (Promega). For HeLa cells, viability was measured 4.5 h or 48 h postintoxication with either CellTiter-Glo or CellTox-Glo (Promega) using a BioTek Synergy plate reader. No-toxin control was scaled to 100% for CellTiter-Glo assays or onefold change for CellTox-Glo.

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 preparations using a Sephadex G-25 desalting column (GE Healthcare, 52-1308-00 BB). A total of 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 6.9 + 50 mM NaCl. Beads were spun down and washed once to remove excess unbound toxin and then incubated with 1.5 nmol ectodomain, 1% FBS, 4 µg anti-GAPDH (sc25778), or 4 µg anti-PVRL1 (Santa Cruz, sc-21722) antibody 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). Western blots were also performed, probing for the His₆ tag on the ectodomain (Abcam, ab18184). Secondary antibodies against mouse IgG (Cell Signaling, 7076S) were used to detect anti-His₆ and anti-PVRL1 primary antibodies and rabbit IgG (Cell Signaling, 7074S) was used to detect anti-GAPDH primary antibody.

Antibody Blocking.

Caco-2 cells were plated at a density of 2,500 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. Viability was read using CellTiter-Glo as described above.

Ectodomain Competition.

HeLa cells were seeded at 2,500 cells per well in a 384-well plate and allowed to incubate at 37 °C for 16 h. Serial dilutions of PVRL1, PVRL2, and PVRL3 ectodomains 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 CellTox-Glo as described above.

Colonic Explants.

Human colonic tissue samples were obtained by the Cooperative Human Tissue Network from consented, deidentified donors under Vanderbilt Institutional Review Board-approved protocol 031078. Experiments were conducted according to published protocols (13) and are further described in SI Methods.

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.