Surface-Layer Protein A (SlpA) Is a Major Contributor to Host-Cell Adherence of Clostridium difficile

Clostridium difficile is a leading cause of antibiotic-associated diarrhea, and a significant etiologic agent of healthcare-associated infections. The mechanisms of attachment and host colonization of C. difficile are not well defined. We hypothesize that non-toxin bacterial factors, especially those facilitating the interaction of C. difficile with the host gut, contribute to the initiation of C. difficile infection. In this work, we optimized a completely anaerobic, quantitative, epithelial-cell adherence assay for vegetative C. difficile cells, determined adherence proficiency under multiple conditions, and investigated C. difficile surface protein variation via immunological and DNA sequencing approaches focused on Surface-Layer Protein A (SlpA). In total, thirty-six epidemic-associated and non-epidemic associated C. difficile clinical isolates were tested in this study, and displayed intra- and inter-clade differences in attachment that were unrelated to toxin production. SlpA was a major contributor to bacterial adherence, and individual subunits of the protein (varying in sequence between strains) mediated host-cell attachment to different extents. Pre-treatment of host cells with crude or purified SlpA subunits, or incubation of vegetative bacteria with anti-SlpA antisera significantly reduced C. difficile attachment. SlpA-mediated adherence-interference correlated with the attachment efficiency of the strain from which the protein was derived, with maximal blockage observed when SlpA was derived from highly adherent strains. In addition, SlpA-containing preparations from a non-toxigenic strain effectively blocked adherence of a phylogenetically distant, epidemic-associated strain, and vice-versa. Taken together, these results suggest that SlpA plays a major role in C. difficile infection, and that it may represent an attractive target for interventions aimed at abrogating gut colonization by this pathogen.


Introduction
Clostridium difficile is a gram-positive, anaerobic, spore-forming bacterium, and causes the antibiotic-associated diarrheal disease, C. difficile infection (CDI). It is also a leading cause of bacterial healthcare-associated infections in hospitals in the United States [1,2], having surpassed methicillin-resistant Staphylococcus aureus (MRSA) in some hospitals for this dubious distinction. Like many enteric pathogens, Clostridium difficile must associate with the intestinal mucosa to begin the process of host colonization [3,4]. While much has been reported on C. difficile pathogenesis in terms of the toxins it produces [5][6][7], relatively little is known about the interaction of the pathogen with the mammalian gut.
Multiple C. difficile adhesins have been described, including the flagellin FliC, the flagellar cap protein FliD [8], fibronectinbinding proteins [9] [10], a heat-shock protein, GroEL [11], the surface associated, heat-shock-induced adhesin, Cwp66 [12], and the surface layer protein, SlpA [13][14][15]. SlpA contains two biologically distinct entities, the high-molecular weight (HMW) and the low molecular weight (LMW) subunits, which are derived via Cwp84-mediated cleavage of a single precursor protein [16], and assemble on the bacterial surface into a paracrystalline lattice [17]. The crystal structure of the C. difficile strain 630 LMW subunit, and the low resolution small-angle X-ray scattering (SAXS) structure of a complex of the two subunits, was solved recently [18]. The two subunits associate with high affinity through the N-terminus of the HMW protein and the C-terminus of the LMW protein [18].
Cwp66 and SlpA are encoded by two genes in a 17-gene cluster that encodes many surface-associated proteins [19,20]. Such Slayer proteins (SLPs) provide structural integrity to the cells, act as molecular sieves, bind to host tissues and extracellular matrix proteins [13], and contribute to host cell adhesion and immune evasion [21][22][23]. Recently, Dingle and colleagues described 12 types of a 10 kb cassette containing cwp66, the secretory translocase secA2, and slpA that can be exchanged between unrelated genotypes [24].
We hypothesized that bacteria-host interactions play a prominent role in intestinal colonization by C. difficile, and were particularly interested in epidemic-associated (EA) strains [25][26][27]. CDI outbreaks involving these strains are characterized by increased morbidity/mortality, increased rates of disease recurrence, reduced cure rate with treatment and increased environmental predominance and spread [28]. In this study, we assessed the role of genotype, SlpA sequence and S-layer cassette type in bacterial adherence of historic/older C. difficile clinical isolates and more recent EA strains including those of the molecular types BI/ NAP1/027 and BK/NAP7,8/078 [restriction endonuclease analysis (REA) groups BI and BK, North American pulse field gel types NAP1 and NAP7/8, or PCR ribotypes 027 and 078].

C. difficile strains and media
All C. difficile human clinical isolates used in this study were originated from the Hines VA Hospital culture collection of Dr. Dale Gerding, or from the collection of Dr. Glenn Songer (Table 1). Four EA C. difficile strains (ribotype 027) and isolated from geographically distinct regions were chosen for in-depth analyses (BI-6, BI-8, BI-17, BI-23), as well as BI-1, a ''historic'' isolate from that pre-dates the epidemics. The non-EA but toxinproducing (toxigenic) strains included strain 630 (ribotype 012), and strain VPI10463 (ribotype 087), a known high-toxin producer [29], which are both rarely found in clinical settings. Strains J9 and K14 have caused hospital outbreaks, and are frequently isolated from hospital settings in the USA, but are not epidemic-associated [30][31][32][33]. Strains M3, M23 and T7 are non-toxigenic. Twenty-one strains, all belonging to the REA group BK and ribotype 078 were recovered from human, porcine or bovine CDI cases. This molecular type was historically associated with veterinary CDI, but is increasingly associated with fulminant human disease [34,35].
Strains were routinely cultured in Brain-Heart Infusion (BHI) broth or on BHI-agar plates (Difco, Buchs, Switzerland; 37 g/L) in an anaerobic chamber (Coy, Grasslake, MI) with 5% CO 2 , 5% H 2 and 90% N 2 . For adherence assays (described below), strains were grown to saturation in BHI broth overnight, clarified by centrifugation at 2000g, washed in phosphate-buffered saline (PBS), resuspended in fresh BHI broth at a ratio of 1:50, and grown anaerobically without agitation until exponential, or saturating growth phase was achieved as required.

Anaerobic bacterial adherence assays
To quantitate CD attachment to human host cells, we optimized an anaerobic bacterial adherence assay [6]. This assay uses a derivative of the Caco-2-derived human intestinal epithelial cell-line, Caco-2 BBE (C2 BBE ). C2 BBE host cells were cultured in high-glucose (25 mM) Dulbecco's Modified Eagle Medium (DMEM), 10% fetal bovine serum, 20 mM HEPES, 100 IU/ml penicillin, and 100 mg/ml streptomycin at 37uC in the presence of 5% CO 2 . Cells between passages 25 and 45 were grown as confluent monolayers (approximately 1.2610 6 cells) in 6-well plates, and transferred to antibiotic and serum-free DMEM 24 hours prior to adherence assays. All assay solutions were prereduced in the anaerobic chamber overnight. Since calcium ions are required for higher-order surface-layer protein assembly in C. difficile, DMEM was supplemented with 25 mM CaCl 2 (DMDM-Ca) prior to the adherence assays. DMEM-Ca was made by adding 1 mL of 1M anaerobic CaCl 2 to 40 mL anaerobic DMEM just prior to use. C2 BBE plates were introduced into the anaerobic chamber just before use, serum-free medium was removed, and exponential phase C. difficile applied at a multiplicity of infection of 20 in a total volume of 2 mL anaerobic DMEM-Ca. To exclude the effects of secreted proteins, including any toxins, all bacterial strains were washed and resuspended in anaerobic DMEM-Ca prior to incubation with host cells. Two mL of DMEM-Ca were applied to control wells.
Following a 40-minute incubation, host cells and adherent bacteria were washed twice with 1 mL of anaerobic phosphatebuffered saline (PBS), scraped, vortexed, serially diluted and plated to enumerate adherent C. difficile. Each experiment was performed in quadruplicate, and repeated at least three times in entirety. The percent adherence was calculated as the ratio of recovered C. difficile to input C. difficile, multiplied by 100. C2 BBE cells survive anaerobic conditions (,5% cell death), as confirmed by a Live-Dead staining assay (not shown; Life Technologies, Grand Island, NY). Also, immunofluorescence microscopy (not shown) indicated that C2 BBE host cells exhibited morphology consistent with viability (well rounded nuclei with uniform staining, normal actin stress fibers and uniformly distributed ZO-1 around host-cell periphery) despite exposure to anaerobic conditions.

Total surface protein interference assay
Confluent C2 BBE monolayers were incubated for 20 minutes prior to the addition of bacteria with anaerobic DMEM-Ca, and an equal volume of anaerobic PBS containing increasing amounts of purified, neutralized, dialyzed, anaerobic, total surface-layer protein (SLP) preparations from different C. difficile strains (described below). Exponential phase C. difficile were then added to the monolayers, adherence allowed to proceed for another 20 minutes, and attached bacteria enumerated as described above.

Antibody interference assay
Exponential phase C. difficile were resuspended in anaerobic DMEM-Ca and incubated with specific anti-LMW SlpA or anti-HMW SlpA (kind gift from Dr. Neil Fairweather), or both, antisera at a dilution of 1:1000 for one hour before addition to confluent C2 BBE monolayers. Control experiments confirmed that there was no significant growth/death of bacteria during this one hour incubation (not shown). As controls, an anti-TraG antiserum (an irrelevant E. coli conjugation protein, non-commercial; gift from Dr. David Hecht) and an anti-6-histidine antiserum (Qiagen, Valencia, CA, not shown) were used at dilutions of 1:1000 each.

Total soluble protein isolation
To obtain total cellular protein, exponential phase C. difficile (O.D. 600 nm = 0.5) were harvested, and lysed by sonication (55% power; 12 pulses of 15 seconds each). Cellular debris was removed by centrifugation at 6500g for 30 minutes at 4uC. A protease inhibitor solution (EDTA-free Complete Cocktail, 16 final concentration, Roche, Indianapolis, IN) was added to the resultant supernate. This supernate was then centrifuged at 265,000g for 2.5 hr at 4uC to fractionate the sample into soluble proteins in the supernate and insoluble proteins in the pellet. The pellet was washed with PBS, and proteins dispersed with gentle sonication (45% power, 3 pulses of 15 seconds each). Equal amounts (30 mg) of total soluble proteins were subjected to SDS-PAGE on 15% Tris-HCl gels (Biorad, Hercules, CA), and stained with Gel-Code Blue (Pierce, Rockford, IL) to visualize protein bands.

S-layer protein extraction
SlpA and other surface-layer proteins (SLPs) were extracted from multiple C. difficile strains using 0.2M glycine pH 2.2, as described by Calabi et al [15]. Briefly, 50 mL of exponential phase C. difficile culture were harvested at by centrifugation (3000g for 20 minutes), washed in PBS, and resuspended in 200 ml of 0.2M glycine pH 2.2 and incubated at room temperature for 30 minutes. After centrifugation to remove the cell pellet (16,000g for 15 minutes at 4uC), the resultant supernate containing surface proteins (SLPs) was dialyzed into 10 volumes of PBS using 10 kDa molecular weight-cutoff centrifugation-based filters (Millipore, Billerica, MA), and pre-reduced before use.

Protein quantitation
Total CD soluble protein, surface layer extracts, and recombinant protein were measured using the bicinchoninic acid (BCA) protein assay kit (Pierce, Rockford, IL), according to the manufacturer's instructions. A standard curve constructed using a gradient of known concentrations of bovine serum albumin (BSA) was prepared and tested for each assessment.

DNA sequencing of slpA
For sequencing, genomic DNA was isolated from exponential phase cultures using Qiagen DNeasy columns (Germantown, Maryland), according to the manufacturer's instructions. DNA concentration was determined by a Beckman spectrophotometer, and DNA was aliquoted and frozen at 220uC.
Since SlpA is highly variable, primers to amplify and sequence the slpA gene were designed inward from neighboring conserved regions. To sequence the gene in entirety, the upstream primer (slpAF: 59-ATGTTGGGAGGAATTTAAGAAATG-39) was designed to include part of the conserved signal sequence of SlpA. The downstream primer (slpAR: 59-ACCTCCACCAGTTTT-CATCTCTGC-39) was designed within the adjacent secA2 gene [36]. To assess slpA conservation between the ribotype 078 strains, we sequenced a gene fragment predicted to encode part of the HMW subunit from all 21 strains used in this study (similar to that performed by Eidhin et al [36]). Primers for this sequencing approach covered nucleotides 1614-2021 of slpA. 100 nanograms of genomic DNA was used as template for PCR reactions with primers at 40 pmol concentration. The Failsafe PCR system (Epicentre Biotechnologies, Madison, WI) consisting of the Failsafe polymerase and a buffer mix (Buffer E) was used to amplify C. difficile DNA. Reactions of 50 ml volume were amplified for 32 cycles including a 94uC denaturation step (30 seconds), annealing from 45-53uC (2 min), and a 72uC extension step for 3 min. Annealing temperature were optimized for each strain using the temperature gradient feature of the BioRad i-cycler (Hercules, CA). PCR products were visualized using agarose gel electrophoresis, purified by QiaQuick PCR purification columns (Qiagen, Germantown, MD), and 20 ng of purified DNA used for sequencing. Sequence data were assembled and analyzed using Vector NTI software (Invitrogen, Carlsbad, CA).

Immunodetection
For Western blotting experiments, 30 mg of total soluble protein and 5 mg of SLP extracts were electrophoresed on denaturing 4-20% gradient Tris-HCl acrylamide gels (Biorad, Hercules, CA) and transferred overnight to 0.45 mM nitrocellulose membrane at 50 volts using neutral 16 Tris-glycine (TG) buffer (2.5 mM Tris-Cl, 19.2% Glycine) at 4uC in a Trans-Blot cell (BioRad, Hercules, CA). Membranes were blocked for 1 hour at room temperature or overnight at 4uC, using 1% blocker from the Roche Western Blotting Kit. Primary antisera to the HMW and LMW SlpA subunits were used at 1:100,000 dilution. Primary antibodies were incubated for 60 mins in 0.5% blocker in Tris-buffered saline (TBS), at room temperature with shaking. Membranes were washed three times with TBS-Tween (TBST; 50 mM Tris-Cl, 150 mM NaCl, pH 7.5 and 0.1% Tween 20) for 10 minutes. Membranes were incubated with secondary antibodies (goat antirabbit IgG-POD conjugate; Roche, Indianapolis, IN) for 30 minutes, and subjected to four 15-minute washes with TBST. Proteins were visualized using the POD chemiluminescent detection system in the Roche Western Blotting kit according to manufacturer's instructions.
Proteomic identification of C. difficile surface proteins Surface protein extracts were electrophoresed using denaturing 4-20% Tris-HCl PAGE (Biorad, Hercules, CA) and stained with Coommassie Brilliant Blue (Sigma, St. Louis, MO). The seven most prominent bands were excised, and proteins identified using liquid chromatography/mass spectrometry (LC/MS) analyses. All mass spectrometry analyses were performed at the University of Minnesota Mass Spectrometry Consortium.

SlpA subunit purification
To produce recombinant SlpA subunits, the portions of slpA corresponding to the LMW subunit and the HMW subunit were each cloned individually from CD strains 630 and K14. Using the Gateway System (Invitrogen), amplified products were ligated into the entry vector pENTR/SD/D-TOPO. The entry vector was recombined with the destination vector pET-DEST-42, which contains an IPTG-inducible promoter, and a C-terminal V5 and 6x-histidine tag. Following sub-cloning into E. coli DH5a, pET-DEST-42 expression vectors with slpA constructs were transformed into Rosetta E. coli to correct the codon bias of E. coli for expression. For selection purposes during cloning and expression, the following antibiotics were used in liquid and solid media: ampicillin (200 mg/ml), carbenicillin (50 mg/ml), kanamycin (25 mg/ml), chloramphenicol (25 mg/ml).
The conditions for recombinant protein synthesis and purification differed for different constructs. Briefly, Rosetta E. coli containing the constructs were grown and induced using Novagen Overnight Express Autoinduction Terrific Broth (TB) medium (Life Technologies, Grand Island, NY). Following induction, bacteria were pelleted by centrifugation, and frozen at 280uC. Pellets were lysed using Bugbuster (EMD Chemicals, Gibbstown, N.J) in the presence of protease inhibitors (EDTA-free Complete Cocktail, 16final concentration, Roche, Indianapolis, IN). For the LMW subunit of 630, recombinant protein was purified from the soluble fraction of cell lysates using cobalt affinity chromatography (Talon-spin columns, Clontech, Mountain View, CA). The HMW subunit of 630 and the LMW subunit of K14 both degraded in the soluble fraction, so these products were isolated from the insoluble fraction, based on the protocol of Fagan et al [18]. Briefly, cell pellets were lysed using BugBuster lysis buffer (EMD Chemicals, Gibbstown, N.J), and insoluble protein from inclusion bodies was purified according to manufacturer's directions. These inclusion bodies were dissolved in 8M urea, 150 mM NaCl 10 mM HEPES, and subjected to cobalt affinity chromatography. The purified recombinant proteins were re-folded by step-wise dialysis to native buffer conditions (150 mM NaCl 10 mM HEPES) to yield soluble refolded proteins.
Because strain K14 HMW construct degraded extensively in E. coli, it was purified by electroelution from CD surface protein extracts. K14 SLP extracts were mixed with 8M urea and Laemmli loading buffer, heated at 70uC for 15 minutes and subsequently electrophoresed in an SDS-PAGE gel. Portions of the unstained gel corresponding to the HMW subunit were excised, minced, and placed in the vertical tubes of a BioRad electroelution apparatus, and allowed to migrate out of the gel slices over three hours. The electroeluted protein solution was dialyzed against native buffer to remove SDS, denatured in 8M urea and refolded as described above.

Recombinant SlpA protein interference assay
Confluent parental Caco-2 cells, (which exhibit similar adherence to the Caco-2 BBE line, data not shown) in 24 well plates were incubated for 20 minutes with 250 uL DMEM-Ca and 250 uL of 150 mM NaCl 10 mM HEPES buffer containing increasing amounts of anaerobic recombinant protein. Exponential-phase C. difficile (MOI 20, 250 mL volume of inoculum prepared as described above) were added to the monolayers and incubated for another 20 minutes. Availability of purified recombinant subunits precluded multiple replicates and data shown are representative of one experiment performed in triplicate, with up to nine concentrations of protein tested in each set.

Toxin production
Total C. difficile toxin (TcdA+TcdB) amounts were determined from equivalent volumes of cell culture supernates of bacterial cultures grown anaerobically for 72 hours. Supernates were clarified by centrifugation, and total toxins determined following manufacturer instructions in the C. DIFFICILE TOX A/B II kit (TechLabs, Blacksburg, VA). All determinations were made from two biological replicates of samples, each tested in triplicate.

Statistical analyses
The SPSS (SPSS, Chicago, IL) and StatView (SAS, San Francisco, CA) software packages were used for statistical analyses. Significance was determined using analysis of variance (ANOVA) to enable comparison between multiple groups of continuous numerical data. The Protected Least Significant Difference test was used for posthoc analyses.

C. difficile strains exhibit varied adherence to host epithelial cells in culture
Clinically diverse C. difficile strains displayed varying abilities to adhere to intestinal epithelial cells (Figure 1, Figure 2), even if they were phylogenetically closely related. Further, for each strain tested, a range of adherence was observed (reflecting interexperimental variability); however the range was consistent. Thus, the average adherence range was approximately 8%-11% for strain 630, 2-4% for strain K14, 5%-9% for strain BI-17, 8-14% for strain M3, and 9%-12% for strain T7. Across all C. difficile strains tested, the mean adherence value was 7.1%. As a group, EA BI/NAP1/027 strains exhibited a mean adherence value of 4.6%.
We also tested a group of ribotype 078 strains (REA group BK and PFGE type NAP 7,8,9) derived from different mammalian sources with symptomatic CDI, to examine adherence to host cells of the same or different species; these strains exhibited a range of adherence corresponding to their provenance. As a group, ribotype 078 strains derived from human CDI cases showed a trend to adhere at higher levels to human intestinal epithelial monolayers, compared with those derived from porcine cases (lower adherence) and bovine cases (least adherence) (p = 0.11; Figure 2). Finally, similar adherence profiles to those seen in Figure 1 and Figure 2 were obtained when a subset of the C. difficile strains tested above (630, BI-17) were used in assays with different intestinal-origin host cells (HT-29, T-84, CaCo-2; not shown) and different multiplicities of infection (50,100; not shown).
Growth phase influences C. difficile adherence to epithelial cells Entry into stationary phase, and consequent nutrient limitation, alters C. difficile physiology in multiple ways, including the increased production of toxins. To determine if nutrient limitation influenced C. difficile adherence, the attachment proficiency of several strains during the stationary phase of growth was evaluated. Figure 3A shows the mean adherence of C. difficile to epithelial cells in multiple independent assays. Epidemic-associated strain BI-17 showed a significant decrease in adherence in stationary phase (p#0.05). Strain 630 showed a trend to decreased adherence (p#0.07), while strain K14 did not display appreciable differences between the two conditions. Several other unrelated strains tested [J32, and M3 (non-toxigenic)] also exhibited lower adherence in stationary phase (Fig. 3B).
Toxin production and C. difficile adherence to epithelial cells are not correlated To determine if there was any correlation between toxin production and bacterial adherence, we evaluated the 21 ribotype 078 strains shown in Figure 2, since these represented the largest number of phylogenetically related isolates in this study. No correlation was evident between the ability to produce toxins and host-cell attachment. However, as a group, the seven ribotype 078 strains of bovine origin produced more toxins A and B than strains of human or porcine origin (Figure 4). These bovine-origin strains also had the lowest adherence when compared with human or porcine 078 strains (Figure 2), but were comparable to all other non-078 strains tested including ribotype 027 strains (Figure 1). Direct toxin effects on adherence were ruled out since all assays were performed with washed, exponentially-growing bacteria, both conditions at which no TcdA or TcdB is present [37].

C. difficile strains exhibit distinct surface protein profiles; identification of SlpA
We hypothesized that the differences in attachment to host epithelial cells could result from changes in protein expression, particularly surface-anchored molecules. Indeed, the various epidemic-associated C. difficile strains studied here (BI-6, BI-8, BI-17), as well as those that were toxigenic (but not epidemicassociated), displayed gross differences in soluble protein abundances (not shown). To specifically explore differences in expression and abundance of surface molecules likely involved in attachment to host cells, sheared surface-layer proteins (SLPs) of several C. difficile strains were separated via SDS-PAGE, and visualized ( Figure 5A). The various strains exhibited distinct profiles of surface proteins, some correlating to the known size of C. difficile cell wall proteins [38].
Two approaches, immunoblot analysis and mass spectrometry, were taken to identify the surface-associated molecules observed in Fig. 5A. Western blotting experiments using polyclonal anti-HMW SlpA and anti-LMW SlpA antisera revealed that the bands corresponding to the highly abundant proteins in Fig. 5A were indeed SlpA subunits ( Figure 5B). The EA strains (BI-6, BI-8, BI-17) produced surface layer protein (SLP) profiles that were distinct from their non-epidemic counterparts. The HMW S-layer protein migration corresponded to approximately 48 kDa, while the LMW S-layer protein corresponded to approximately 30 kDa.
The LMW subunit has been previously noted to vary more in size and antigenicity [14,15,36].
Several SLP bands (corresponding to those numbered in Fig. 5A) were excised and subjected to MALDI mass spectrometry analyses, which revealed that all were homologs of C. difficile cell wall proteins (Cwp20, Cwp84, a ,66 kDa Cwp and SlpA; not shown), sharing similarity to those of the sequenced epidemic-  . Adherence of ribotype 078 strains derived from different mammalian sources with symptomatic CDI. Panel A. Twenty-one ribotype 078 strains were tested for adherence to human-derived enterocytes, using the protocol described in the Methods section. Isolates were divided into those of human origin (dark gray bars; 7 total); porcine origin (medium gray bars; 7 total) and bovine origin (light gray bars; 7 total). All isolates were tested in quadruplicate, and assays were repeated in entirety at least once. The epidemic-associated BI-17 strain (red bar) was used as an internal control. Panel B. Average adherence of each group of 078 strains in comparison with a human EA strain, BI17. doi:10.1371/journal.pone.0078404.g002 associated C. difficile strain from Quebec QCD-32g58 (Genbank accession #AAML00000000.4; GI:145694830).

C. difficile surface protein preparations promote bacterial adherence to host cells
To explore a possible link between SlpA variations and differential C. difficile adherence to host cells, the ability of fresh, pre-reduced surface layer proteins (crude or purified preparations) as well as SlpA antibodies, respectively, to competitively block bacterial attachment was assessed. When confluent C2 BBE monolayers were pre-incubated with increasing amounts of total SLP preparations, C. difficile adherence to epithelial cells was reduced in a dose-dependent manner (up to 80% adherence inhibition with the highest concentration of SLP used; p#0.0001; Figure 6). This dose-dependent adherence reduction was evident irrespective of the toxigenic status of the C. difficile strain tested [ Figure 6A, strain BI-17 (toxigenic); Figure 6B, strain M3 (nontoxigenic)]. In control experiments, PBS alone (Fig. 6) or PBS with 50 mg bovine serum albumin (not shown) did not significantly interfere with C. difficile binding to C2 BBE cells.  . Toxin production of ribotype 078 isolates. Toxin testing was performed as described in the Methods section. As above, isolates were divided into those of human origin (dark gray bars; 7 total); porcine origin (medium gray bars; 7 total) and bovine origin (light gray bars; 7 total). Two biological replicates of each strain were assayed, and each toxin measurement was obtained in triplicate. The clinically uncommon low toxin producing strain 630 and high toxin producing strain VPI 10463 (clear bars) were used as internal controls. doi:10.1371/journal.pone.0078404.g004 Surface protein-based inhibition of C. difficile adherence is not strain specific In the hamster model of CDI, colonization with a non-toxigenic strain of C. difficile efficiently prevents colonization by a toxigenic C. difficile strain [39], although the mechanism for this effect is not known. To test if the adherence interference observed above was strain-specific, we used total SLP preparations from two C. difficile isolates of different phylogenetic clades, and tested their abilities to inhibit adherence of the parent strain from which they were derived, as well as the non-cognate strain. SLPs prepared from the non-toxigenic C. difficile strain M3 significantly inhibited adherence of M3 vegetative cells, as well as those of the epidemic-associated strain BI-17, and BI-17 SLPs similarly inhibited adherence of both BI-17 and M3 vegetative cells (Figure 7). The degree of inhibition was almost identical in both sets of assays ($85%; p#0.0001).

Anti-SlpA antibodies block C. difficile adherence to host cells
Since surface extracts contain proteins other than SlpA, the impaired adherence observed in Figures 6 and 7 could have resulted from interference by any of these molecules. To define the specific involvement of SlpA in adherence, C. difficile strain 630 was pre-incubated with specific anti-LMW SlpA or anti-HMW SlpA antisera for one hour before addition of the bacteria to confluent C2 BBE monolayers. Pre-incubation of C. difficile 630 bacteria with antibodies against either SlpA subunit, but not against an irrelevant protein (E. coli protein TraG), significantly reduced adherence of that strain by approximately 50%, indicating that both subunits of SlpA contribute to C. difficile adherence (Figure 8; p#0.0001). The anti-SlpA antibodies also significantly reduced adherence of the epidemic-associated strain BI-17 (not shown).  C. difficile adherence is inhibited by pre-coating host cells with surface layer protein preparations. Total SLP protein from strain BI-17 (toxigenic) or strain M3 (non-toxigenic) were overlaid on C2 BBE monolayers, and followed by adherence assays for the cognate C. difficile vegetative bacteria. For comparative purposes, data were converted to percent adjusted adherence, with adherence of the control (no added SLP) set to 100%; therefore, no error bars are shown. Asterisks indicate significant differences (p#0.01). Epidemic-associated, ribotype 027 strains are shown as red bars, and non-toxigenic strains are shown as green bars. doi:10.1371/journal.pone.0078404.g006 SlpA sequence varies between C. difficile strains, and impacts bacterial adherence to host cells To correlate SlpA variations with adherence profiles of the different strains, we determined the sequences of the entire coding and upstream regions of the slpA genes of multiple toxigenic, toxigenic/epidemic-associated, and non-toxigenic C. difficile clinical isolates. slpA from four strains from the epidemic-associated clade (BI-1, BI-6, BI-8 and BI-17; BI-1 predates current outbreaks), two toxigenic strains J9 and K14, and two nontoxigenic strains, M3 and T7 was sequenced. Tables 2 and 3 show the percentage amino acid identity between strains, and by subunit. SlpA sequence was identical (99-100%; entire sequence) at the amino acid level among the four epidemic-associated, ribotype 027 strains. Strain J9, despite being in a different clade, showed the highest degree of sequence similarity to the epidemicassociated strains. The non-toxigenic strain M3 displayed the least sequence conservation with the epidemic-associated strains, consistent with the phylogenetic divergence of non-toxigenic strains.
At the subunit level, the HMW subunit was more conserved in predicted amino acid sequence, while the LMW subunit was more divergent, consistent with previous surveys of slpA sequences [36]. The only highly conserved sequences in the LMW subunit were the N-terminal signal sequence, and the C-terminal portion predicted to be involved in interaction with the HMW subunit. The HMW subunit was conserved over the whole sequence assessed, consistent with its role as the peptidoglycan anchor.    For ribotype 078 strains, we observed high level slpA sequence identity among all 21 isolates used in this study. Sequence identity was .95% when either a HMW-encoding fragment was sequenced (our strains; Table 4), or the whole gene investigated (in silico analyses; all publicly-available, full-length 078-specific slpA; (NCBI, ADVM01000008.1, ADNX01000091.1, NC_017174.1, ADDE01000040.1, ABKL02000030.1; not shown), and nine as yet unpublished ribotype 078 genomes [Glenn Songer, personal communication; (not shown)]. However, SlpA sequence conservation was ,60% between ribotype 078 and non-078 strains ( Table 4).
Given the relative conservation of sequences for the HMW subunit, we hypothesized that the LMW subunit sequence divergences contributed to host-cell adherence differences. To test this, a pilot study evaluated the ability of purified recombinant HMW and LMW SlpA subunits (of both strains 630 and K14), to interfere with adherence of strain 630. Data were calculated as IC 50 values (subunit concentration required for 50% inhibition of strain 630 adherence) based on a non-linear least-squares curve fit. Consistent with our hypothesis, the IC 50 for the LMW subunit was lower (1.1 mM and 2.2 mM for LMW from 630 and K14, respectively) than that observed for the HMW subunit (3.7 mM and 4.4 mM HMW from 630 and K14, respectively; Table 5). However, due to the limited number of data points, the confidence intervals of the curves for each subunit overlapped, indicating that they were not significantly different. Interestingly, the IC 50 values were consistent with the adherence proficiency of the corresponding parent strains (strain 630 showed higher adherence than strain K14), with lower concentrations of the 630 SlpA subunits being required for 50% attachment inhibition.

Discussion
While the function of the C. difficile toxins in disease has been widely evaluated and appreciated, the role of non-toxin virulence factors is less well-defined. It is increasingly evident, however, that such virulence factors contribute significantly to disease. For example, the ability of non-toxigenic strains to prevent colonization and/or disease induction by toxigenic strains in the hamster model suggests the importance of colonization during C. difficile infection [40]. Additionally, up to 80% of strains isolated from patients with recurrent CDI are genotypically identical to the initial infecting strain, indicating the existence of robust persistence mechanisms [41]. We, and others, have shown that EA strains have increased spore production that may contribute to their ability to thrive in the patient and/or the environment [37,42]. However, a recent assessment of a large number of EA and non-EA strains indicate that this trend may not generalize to the 027 clade as a whole [43]. The studies described here address one nontoxin mechanism vital to establishing colonization: the interaction of C. difficile with host intestinal epithelial cells.
Using a quantitative, anaerobic assay to measure C. difficile vegetative cell attachment to host intestinal epithelial cells, we demonstrated that C. difficile strains adhere to host cells, and that SlpA contributes to this adherence. Importantly, adherence values of C. difficile were consistently similar to those seen for other enteric pathogens [44]. The various strains of C. difficile displayed diverse adherence capabilities, both within and between clades. Since bacteria typically harbor a repertoire of adhesins whose expression can be altered via a variety of regulatory mechanisms, it is not surprising that even closely related strains of C. difficile vary so widely in their overall adherence [45]. Thus, while the nontoxigenic strains such as M3, M23, and T7 were highly adherent, so were several toxigenic strains. These observations underscore the independent importance of non-toxin virulence factors in the establishment of C. difficile colonization. Interestingly, we did not  observe any correlation (positive or negative) between C. difficile adherence and the ability to produce the large clostridial toxins TcdA and TcdB. This finding is consistent with our previous test of C. difficile tcdA/tcdB isogenic mutants [6]. Consistent with previous studies [14,36], our sequencing results indicated that for all strains tested, the SlpA LMW subunit was more variable, while the HMW subunit was more conserved. The four newer, epidemic-associated ribotype 027 isolates harbored identical slpA coding regions [30]. SlpA, especially the LMW subunit, is also known to be antigenic, leading to the suggestion that it could be included a candidate in a multi-component vaccine for the prevention of CDI [46,47].
Ribotype 078 strains, once thought to be restricted to veterinary populations, exhibited robust adherence to the human intestinal monolayers used in this study. Interestingly, 078 strains recovered from human infections adhered most strongly to C2 BBE cells compared with those recovered from porcine or bovine CDI, suggesting possible adaptation of human-derived 078 strains to the human GI tract. However, this putative adaptation could not be solely attributed to SlpA, since we found slpA sequences to be highly conserved in this ribotype.
A recent study has provided a useful framework for understanding the high variability in slpA and adjacent genes. Using whole genome sequencing, Dingle and colleagues [24] found that a 10KB cassette (''S-layer cassette'') comprising slpA, cwp66, and secA2 exhibited a high degree of variability as compared to the surrounding genome, and suggested that horizontal gene transfer was responsible for this variation. Twelve S-layer cassettes were described (Types 1-12), as well as one additional type that was a hybrid of the Type 2 and Type 6 cassettes, and found only in two closely related strains of PCR ribotype 078 and 193. To assess if the adherence variations we observed in our twenty-one PCR ribotype 078 strains ( Figure 2) could be correlated to potential Slayer cassette types, we performed additional DNA sequence analyses on all twenty-one of our ribotype 078 strains, as well as fourteen published and unpublished C. difficile 078 strain genomes, and found almost invariant slpA sequence and highly conserved (.91%) S-layer cassette sequences. Taken together these data suggest that ribotype 078 strains may represent a sub-clade of C. difficile that are monophyletic at slpA as well as the S-layer cassette locus. Therefore, the strain-to-strain variation in host-cell adherence we observed in the PCR ribotype 078 strains is likely due to factors other than SlpA, or beyond the S-layer cassette.
The inhibition of bacterial adherence to host cells by SLP preparations and purified SlpA proteins, as well as by antibodies specific to SlpA, strongly supports a role for this molecule in C. difficile vegetative cell attachment to enterocytes. This is consistent with the known interactions of SlpA with extracellular matrix proteins of epithelial cells [13], although the interaction(s) mechanisms are undefined. SlpA is, however, not a typical adhesin, in that the S-layer lattice covers the whole bacterium, and as the bacterial cell grows, new surface-layers are deposited. Therefore, increased expression of SlpA would likely not provide additional binding epitopes on the surface. However, the stability and abundant distribution of SlpA may be advantageous under certain conditions. For example, if other adhesins are downregulated during stationary phase, SlpA is always present to serve as an alternate, high avidity attachment molecule.
Our cross-interference assays confirmed that adherence inhibition occurs when using SLP preparations from non-cognate C. difficile strains. In the animal model of CDI, infection with one C. difficile strain prevents the subsequent colonization of another [39]. The mechanism for this exclusion is not known, and may involve competition for mucosal niche occupancy or direct gut epithelial cell binding. Secreted surface molecules such as SlpA could therefore present an innovative approach to prevent C. difficile colonization by competitively excluding incoming bacteria from host-cell attachment sites in high-risk veterinary or human populations. This idea is supported by studies of other mucosally-associated bacteria. For example, surface layer extracts from Lactobacillus helveticus inhibit the adherence of enterohemorrhagic E. coli O157:H7 to epithelial cells [48]. Conversely, targeting/ sequestering SlpA on bacteria may be another approach to prevent C. difficile colonization, and recent efforts in support of this approach have focused on rational design of surface-protein blocking agents [49].
Adherence interference assays with recombinant subunit proteins implicated a greater role for the LMW subunit compared to the HMW subunit. This is consistent with crystal and solution structure models of SlpA, which suggest greater surface exposure for the LMW subunit, with the HMW subunit predicted to be more closely associate with the peptidoglycan layer [18]. Studies by Calabi et al, however, found that HMW SlpA bound to gastrointestinal tissues, while the LMW SlpA failed to do so [13]. Clearly, further studies exploring the topology of SlpA-host cell surface interactions are warranted.
Taken together, the results from the studies presented here indicate that that some C. difficile strains, including those associated with recent CDI outbreaks, have robust adherence to human intestinal epithelial cells that is mediated, in large part, by Surface-Layer Protein A. The LMW subunit, despite high variability, may be more involved than the HMW subunit in mediating adherence of SlpA, and therefore of C. difficile, to host cells. However, sequence variability may confound immune responses to CDI that are based on SLP recognition, and merits attention if whole SlpAbased anti-CDI interventions are explored.

Supporting Information
Table S1 Plasmids constructed and used in this study. (DOCX)