A Clostridium difficile Cell Wall Glycopolymer Locus Influences Bacterial Shape, Polysaccharide Production and Virulence

Clostridium difficile is a diarrheagenic pathogen associated with significant mortality and morbidity. While its glucosylating toxins are primary virulence determinants, there is increasing appreciation of important roles for non-toxin factors in C. difficile pathogenesis. Cell wall glycopolymers (CWGs) influence the virulence of various pathogens. Five C. difficile CWGs, including PSII, have been structurally characterized, but their biosynthesis and significance in C. difficile infection is unknown. We explored the contribution of a conserved CWG locus to C. difficile cell-surface integrity and virulence. Attempts at disrupting multiple genes in the locus, including one encoding a predicted CWG exporter mviN, were unsuccessful, suggesting essentiality of the respective gene products. However, antisense RNA-mediated mviN downregulation resulted in slight morphology defects, retarded growth, and decreased surface PSII deposition. Two other genes, lcpA and lcpB, with putative roles in CWG anchoring, could be disrupted by insertional inactivation. lcpA - and lcpB - mutants had distinct phenotypes, implying non-redundant roles for the respective proteins. The lcpB - mutant was defective in surface PSII deposition and shedding, and exhibited a remodeled cell surface characterized by elongated and helical morphology, aberrantly-localized cell septae, and an altered surface-anchored protein profile. Both lcpA - and lcpB - strains also displayed heightened virulence in a hamster model of C. difficile disease. We propose that gene products of the C. difficile CWG locus are essential, that they direct the production/assembly of key antigenic surface polysaccharides, and thereby have complex roles in virulence.


Plasmids and mutagenesis
A list of the plasmids and primers used in this study is listed in Table 1. C. difficile gene-specific disruptions were designed using the intron-targeting tool ClosTron [26]. Seven open reading frames (ORFs) from the putative CWG locus, based on the C. difficile strain 630 genome sequence, were targeted for disruption: CD2783, CD2781 (mviN), CD2780 (pgm2), CD2779 (manC), CD2769, CD2765 (lcpA), and CD2766 (lcpB). The gene-specific ClosTron insertion site was chosen based on the Perutka algorithm [27]. Targeting regions were constructed and cloned into pMTL007C-E5 by DNA 2.0 (Menlo Park, CA; [26]). Knockdown of mviN expression was achieved via the inducible expression of a 170 base pair RNA fragment complementary to the ribosome-binding site and the first 30 codon-encoding nucleotides of the mviN mRNA. This DNA fragment was synthesized and cloned into pMTL82153 [28] by DNA2.0 (Menlo Park, CA) to generate pMC4. To facilitate inducible expression, the insert was amplified using primers MGM119 and MGM120 (Table 1), digested with SacI and BamHI restriction enzymes (New England Biolabs, Ipswich, MA), and cloned into similarly digested pRPF185 (kind gift from Drs. Neil Fairweather and Robert Fagan) to generate pMC6 [29]. The lcpA-and lcpB-complementing plasmids were derived via PCR amplification from JIR8094 chromosomal DNA using the corresponding primers (MC29/MC30, and MC25/MC26, respectively; Table 1). The In-Fusion HD Cloning Kit (Clontech Laboratories, Inc., Mountain View, CA) was used to clone the fragments into pRPF185 to generate pMC9 (for lcpA) and pMC10 (for lcpB). For a corresponding control, pRPF185 was digested with SacI and BamHI, Klenow Polymerase-treated, and re-ligated using the T4 DNA ligase kit (Thermo-Fisher Scientific Inc., Waltham, MA) to yield pMAC1.

C. difficile conjugation
Plasmids were transferred from E. coli to C. difficile via conjugation following the protocol of Heap et al. [30] with some modifications. ClosTron constructs for targeted lcpA and lcpB disruptions, respectively, were electroporated into E. coli CA434 [31]. E. coli CA434 containing one of the constructs of interest (donor strain) and C. difficile JIR8094 (recipient strain) were grown overnight in BHIS [24]. One milliliter (mL) of the overnight donor culture was centrifuged at 1,500 x g, washed in 1mL of BHIS, and pelleted again. The pellet was re-suspended in 200μL of the overnight recipient strain and plated on a BHIS plate incubated at 37°C anaerobically for 10 hours. The entire bacterial lawn was scraped off the plate, re-suspended in 1mL phosphate buffer saline (PBS), of which 200μL was plated on BHIS containing cefoxitin (8μg/ mL), cycloserine (250μg/mL), and thiamphenicol (15μg/mL), and incubated anaerobically for 24-48 hours. C. difficile transconjugants were passaged on BHIS containing erythromycin (10μL/ml) to select for potential gene-specific integrants. Potential integrants were screened by PCR using gene-specific primers, paired with a primer specific for the antibiotic resistance cassette (ErmRAM) sequence [26].

CWG visualization by transmission electron microscopy
C. difficile was visualized via thin-section electron microscopy using the pre-and post-fixation protocol of Davies and Boriello [13]. CWG visualization by scanning electron microscopy C. difficile capsular polysaccharides were visualized via scanning electron microscopy using the pre-and post-fixation protocols of Hammerschmidt et al. and Davies and Boriello [13,32]. Briefly, GV53 (C. difficile K14) was grown overnight in BHI broth, diluted 1:50, and grown to mid-log phase. The mid-log cultures were plated on BHI agar and grown for 24-36 hours. For pre-fixation, an agar plug was fixed in 0.1M sodium cacodylate buffer containing 0.1% ruthenium red, 3% glutaraldehyde, and 10mM L-lysine for 20 minutes on ice. The fixative solution was removed, and the plugs were washed with 0.1M sodium cacodylate buffer containing 0.1% ruthenium red. The plugs were fixed again in fixative solution without L-lysine for 3 hours at room temperature, washed in 0.1M sodium cacodylate buffer containing 0.1% ruthenium red, and then post-fixed in 0.1M sodium cacodylate buffer containing 0.1% ruthenium red and 1% osmium tetroxide for 1 hour at room temperature. Following post-fixation, the plugs were washed 3 times in 0.1M sodium cacodylate buffer containing 0.1% ruthenium red, subjected to an ethanol-water dehydration series consisting of 10%, 30%, 50%, 70%, 90%, and 100% at 10 minutes each on ice. At the final dehydration step, the plugs were warmed to room temperature and fresh 100% ethanol was added. Further processing and imaging of the plugs were done at the University of Arizona's Spectroscopy and Imaging Facilities, and visualization performed using a Hitachi S-4800 Type II Ultra-High Resolution Field Emission Scanning Electron Microscope. Alternatively, strains GV345 (lcpA -plcpA), GV346 (lcpB -plcpB), GV347 [lcpA -(Vector)], GV348 [lcpB -(Vector)] and GV349 [WT (Vector)] were grown overnight in BHI broth, harvested and fixed in 2% glutaraldehyde and submitted to the University Spectroscopy and Imaging Facilities at the University of Arizona for processing and imaging using the same microscope as above.

Bioinformatic analyses of the C. difficile CWG locus
The RAST (Rapid Annotation using Subsystem Technology) sequence-based comparative tool [33] was used to compute predicted proteomes of 32 C. difficile strains whose genome sequences were obtained from National Center for Biotechnology Information GenBank (NCBI; GenBank; [34]), as well as specific isolates in the Vedantam Laboratory strain collection, and compared to the C. difficile strain 630 genome [35] (S1 Table). The conditional formatting function of the Microsoft Excel application (Microsoft, Redmond, WA) was used to generate the protein identity/C. difficile strain grid. Different colors were assigned to varying ranges of percent identities in each cell. PHYLIP [36] was used to compute and generate a cladogram where strains with similar protein identity profiles were clustered together.

Transcriptional analysis
For non-quantitative Reverse Transciptase (RT)-PCR analysis, JIR8094 was grown in BHI broth overnight then sub-cultured in BHI broth to a 600 nanometer (nm) optical density (OD) 600nm = 0.4. Cells were harvested by centrifugation, and RNA extracted using the GeneJET RNA Purification Kit (Thermo Fisher Scientific Inc., Waltham, MA). Extracted RNA was subsequently treated with DNAse (Thermo Fisher Scientific Inc., Waltham, MA) to remove any contaminating genomic DNA. Complementary DNA (cDNA) was synthesized from DNAsetreated RNA using the Superscript III First Strand kit (Thermo Fisher Scientific Inc., Waltham, MA). The intergenic regions of the following genes were amplified from the cDNA using conventional PCR, and the primers listed in Table 1: mviN and pgm2 (MC46 and MC47), pgm2 and manC (MC48 and MC49), manC and CD2778 (MC50 and MC51), and lcpA and lcpB (MC52 and MC53).
For cell wall fractionation and PSII shedding studies, strains were grown overnight, sub-cultured in 50mL BHI for 12 hours, harvested by centrifugation, and the supernatant (used for PSII harvesting) collected and concentrated through a 3kDa cutoff column to a final volume of 1.5mL. The remaining pellet (P1) was subjected to a total Surface-Layer Protein (SLP) shearing protocol as described by Fagan et al. [38]; this resulted in another supernatant (SLP fraction) and pellet (P2; finally used for cytosolic protein extraction). Briefly, for SLP shearing, the P1 pellet was washed in PBS, resuspended in 500μL of 0.2M glycine pH 2.2, and incubated at room temperature for 30 minutes with gentle agitation. Extracts were centrifuged at 16,000 x g for 15 minutes at 4°C and the supernatants were collected and neutralized with 2M tris-base. Finally, the remaining pellet (P2) was re-suspended in 50mM triethylammonium bicarbonate buffer, sonicated, and 6M urea was added to obtain a total cell lysate. For all subsequent immunoblot and SDS-PAGE analyses, all samples were normalized to total protein in the respective fraction (50μg for the supernatant fraction and 5μg for surface layer proteins; a representative Gel Code Blue-stained gel is presented in S6 Fig to show equivalent sample loading). Two independent biological replicates were performed for each strain. All SDS-PAGE analyses were performed using 4-20% TGX gels (Bio-Rad, Hercules, CA) followed by staining with Gel Code Blue (Thermo Fisher Scientific Inc., Waltham, MA).

Immunoblot analyses
PSII conjugated to the heat-labile enterotoxin B subunit (LTB) of enterotoxigenic Escherichia coli (ETEC) was a kind gift from Dr. Mario Monteiro [18]. We submitted this conjugate (PSII-LTB) to Alpha Diagnostics (San Antonio, Texas) for polyclonal antibody generation in rabbits (intramuscular immunization with 1mg of the conjugate per animal). The pre-immune serum and two bleeds were collected at Days 1, 53 and 67, respectively. For PSII immunoblot analysis, crude extracts normalized to cell number or total protein were diluted two-fold in PBS, and 5μL spotted onto an activated polyvinylidene fluoride (PVDF) membrane. The membrane was vigorously washed in deionized water for 20 minutes, blocked in 10% milk in TBST (Tris-buffered saline with 0.1% tween 20) for 15 minutes, and then washed in 5% milk in TBST for 2 minutes. After overnight incubation in primary antibody (pre-immune rabbit serum or PSII-LTB rabbit antiserum, 1:8,000), the membrane was washed once in 5% milk in TBST, twice in TBST for 5 minutes each, and once in 5% milk in TBST for 10 minutes. Following incubation with goat anti-rabbit horseradish-peroxidase-conjugated secondary antibody (Sigma-Aldrich, St. Louis, MO) at 1:10,000 dilution for 1 hour, the membrane was washed 5 times in TBST for 5 minutes each, and developed using SuperSignal West Femto chemiluminescent substrate (Thermo Fisher Scientific Inc., Waltham, MA).
Antisense RNA-mediated mviN down-regulation C. difficile JIR8094 containing pMC6 or pRPF185 (GV341 and GV342 respectively) were grown overnight in TY broth supplemented with thiamphenicol (15μg/mL). The strains were sub-cultured (1:50 or 1:100) in TY broth with or without anhydrotetracycline (500ng/mL) and grown for 12 hours. Following the induction, the cells were imaged using the Maneval's capsule stain as described by Corstvet et al. [39] and processed for PSII production using the EDTA--TEA extraction method and immunoblot analysis as described above.
C. difficile growth and toxin assessments C. difficile JIR8094 containing pMC6 or pRPF185 (GV341 and GV342 respectively), were grown overnight in TY broth supplemented with thiamphenicol as described above. The strains were sub-cultured (1:50), grown to an (OD) 600nm of approximately 0.6, and then diluted 1:100 in 30mL of TY broth with thiamphenicol and with or without 500ng/mL anhydrotetracycline. OD measurements were taken within a 24-hour time period. Two biological replicates were performed for each condition.
C. difficile JIR8094, and isogenic lcpAand lcpBmutants (GV343 and GV344 for the lcpAand lcpBmutants respectively) were grown overnight in BHI broth. The strains were standardized to (OD) 600nm = 0.1 and monitored over a 12-hour time period with (OD) 600nm measurements taken every hour. Three biological replicates were performed for each strain. Toxin data was measured at 72 hours from culture supernatants using the Wampole C. difficile Toxin A/B ELISA kit (Alere, Waltham MA, USA). Three biological replicates were used in the C. difficile toxin ELISA analysis.

Proteomic analyses of mviN knock-down strains
Strains GV341 and GV342 were grown in TY broth with thiamphenicol (15 μg/mL) overnight, then diluted 1:50 in TY and thiamphenicol with or without anhydrotetracycline (500ng/mL). Samples were collected at mid-log and processed for iTRAQ proteomic analysis as described previously [40,41]. Briefly, a total of 100 μg of protein in 20 μl of 7M urea, 2M thiourea, and 4% CHAPS (Sigma-Aldrich, St. Louis, MO) was denatured in 2% SDS, reduced in 100mM tris-(2-carboxyethyl) phosphine, and alkylated in 84mM iodoacetamide. The urea concentration of the samples was diluted to 2M using 100mM TEAB prior to digestion with trypsin at a ratio of 1:10. iTRAQ reagent labeling was performed according to manufacturers' instructions (AB SCIEX, Framingham, MA). Strong cationic exchange (SCX) fractionation was performed using Waters 600E HPLC system, fractions collected, resuspended in acetonitrile/ trifluoroacetic acid, injected onto a Merck Chromolith CapRod column, and eluates automatically spotted on a stainless steel MALDI target plate for spot analysis (ABSciexTripleTOF 5600+ mass spectrophotometer; AB SCIEX, Framingham, MA). Protein identification and quantitation was performed using the Protein Pilot 3.0 software against the C. difficile strain 630 protein database, differentially abundant proteins were identified via multiple stringent statistical tests as described previously [40,41], and in the Statistics section below.

Immunofluorescence studies
To immobilize bacteria, 10μL of a 1:10 dilution of poly-L-lysine was spotted on a multi-well slide (MP Biomedicals, LLC., Santa Ana, CA) for 5 minutes, followed by 2 washes with PBS. Strains GV341 and GV342 were sub-cultured overnight in BHI broth with anhydrotetracycline (500ng/mL) then diluted to an (OD) 600nm = 0.1. Equal volumes were centrifuged at 3220 x g for 10 minutes, then re-suspended in 500μL of BHI. Ten microliters of the culture was spotted on poly-L-lysine coated slides, and fixed with 1.5% paraformaldehyde for 15 minutes at room temperature. The wells were washed 3 times with PBS for 5 minutes each, incubated in 50mM ammonium chloride for 5 minutes, washed 2 times with PBS for 5 minutes each, then blocked for 30 minutes with 2% IgG-free bovine serum albumin (BSA) in PBS. Cells were incubated in PSII-LTB rabbit antiserum (at a 1:100 dilution) for one hour, washed 5 times with 1% BSA for 5 minutes each, incubated with Alexafluor 488 at a 1:500 dilution (Life Technologies, Grand Island, NY) for one hour, followed by 6 washes in 1% BSA and 2 washes in PBS. Samples were mounted using ProLong Diamond reagent (Life Technologies, Grand Island, NY), allowed to cure for 24 hours, then imaged using the EVOSfl Cell Imaging System (Thermo Fisher Scientific Inc., Waltham, MA).
For FM4-64 and DAPI co-staining, samples were prepared as described above with the following modifications. Strains GV345 (lcpA -plcpA), GV346 (lcpB -plcpB), GV347 [lcpA -(Vector)], GV348 [lcpB -(Vector)] and GV349 [WT (Vector)] were sub-cultured overnight in BHI broth supplemented with thiamphenicol (15μg/mL), then immobilized on poly-L-lysine coated coverslips. Samples were processed for PSII staining as described above. Following PSII staining and incubation with Alexafluor 488, cells were washed 3 times with 1% BSA and 2 times in PBS for 5 minutes each. Cells were then stained with 10 μg/mL FM4-64FX (Life Technologies, Grand Island, NY) for 30 minutes. Cells were washed 4 times in PBS then mounted on glass slides using ProLong Gold with DAPI (Life Technologies, Grand Island, NY). Slides were imaged using a Deltavision RT deconvolution fluorescence microscope (MCB Imaging Facility, University of Arizona). The scale bar represents 3.26μm, and all images presented are representative of at least three different fields.

Biofilm studies
Biofilm assays were done as described by Pantaleon et al. [41]. Strains were grown overnight in BHI, then diluted 1:100 in BHIS supplemented with 1.8% glucose and grown for 72 hours in 24-well polystyrene plates (Costar, USA). The plate was washed two times with PBS then incubated at 37°C to dry. One milliliter of 0.2% crystal violet was added to each well, and the plate was incubated at 37°C for 30 minutes. The wells were washed twice with PBS, resuspended in 1mL of methanol/acetone (80%/20%), and the OD was measured at 570nm. Three biological replicates were performed in technical triplicate for each strain.

Antimicrobial susceptibility testing
Susceptibility to the cathelicidin antimicrobial peptide (CAMP) LL-37 was determined for the isogenic parent (JIR8094) and the lcpAmutant (GV343) by measuring the minimum inhibitory concentration (MIC) as described previously [40]. Susceptibility to the antibiotics vancomycin and metronidazole was determined by agar disk diffusion, and measuring the zone of inhibition (ZOI). Briefly, bacteria were sub-cultured to mid-exponential growth phase then normalized to an (OD) 600nm = 0.1. One-hundred microliters of the normalized suspension was plated as a lawn on Brucella agar (BD Biosciences, San Jose, CA), and an antibiotic disc [vancomycin (5μg/mL) or metronidazole (50μg/mL)] was placed in the center. The ZOI was measured after 24 hours of growth.

Adhesion assays
Caco-2BBe cells (ATCC CRL-2102) were propagated in 6-well plates using Dulbecco's Modification of Eagle's Medium (DMEM with 4.5g/L glucose, L-glutamine and sodium pyruvate) (Corning, Corning, NY) supplemented with 10% FBS and 0.5% HEPES. One day before the assay, cells were incubated in serum-free medium (0.5% DMEM, 0.53% F12, 0.12% sodium bicarbonate, 1.78% HEPES and 0.5% mannose). Adhesion assays were performed as described by Merrigan et al. with minor modifications [42]. Briefly, strains JIR8094, GV343 (lcpA -) and GV344 (lcpB -) were grown overnight in BHI broth then sub-cultured in 50mL BHI, grown to an (OD) 600nm = 0.4, pelleted and resuspended in serum free media supplemented with 0.025M calcium chloride. Caco-2BBe cells were moved into the anaerobic chamber, serum free medium was removed from each well of the 6-well plate and 2mL of the bacterial suspension was added to each well at a multiplicity of infection (MOI) of 50. Bacteria were allowed to adhere for 40 minutes anaerobically at 37°C (the bacterial inoculum was plated on BHI plates at this point). Following incubation, bacteria were removed, Caco-2BBe cells were washed 3 times with 1mL PBS, scraped off each well, and plated on BHIS plates to enumerate percent adherence. Three biological replicates were performed in technical quadruplicate for each strain.

Golden Syrian hamster infections
The Golden Syrian hamster model of infection was used to study the impact of lcpA-and lcpBmutants in C. difficile virulence. Eight-week old male hamsters were orally administered clindamycin (30mg/kg; "Cleocin", by prescription, University of Arizona Health Center Pharmacy) three days prior to infection. Five hundred spores were then also administered orally, and hamsters monitored for disease symptoms (wet-tail, ruffled coat, lethargy) throughout the course of the experiment. Moribund animals or those meeting the criteria for euthanasia were administered 270mg/kg commercial euthanizing solution (Euthanasia III, MedPharma Inc, Pomona, CA). Euthanized animals were dissected for visualization of gross pathology, and cecal contents harvested and either visualized via phase-contrast microscopy, or plated on selective medium for recovery and molecular typing of C. difficile (using 16s-23s rDNA intergenic fragment profiling and comparison with the organisms used for infection). All animal studies were approved by the Institutional Animal Care and Use Committee of the University of Arizona.

Statistical analysis
Multiple statistical tests were employed to determine significance for experiments involving quantitation. For biofilm studies, growth and toxin assessments and adhesion assays, Student's t tests were performed to compute differences between WT (Vector) and lcpB -(Vector), and errors bars calculated from standard deviation(s). For proteomics studies, technical replicates and standard sensitivity curve analyses were used [43], and fold change cutoffs were calculated assuming a false discovery rate (FDR) of 10%. Further, for each protein we identified, a hypergeometric test was performed to calculate a p-value for its abundance in the test condition compared to its abundance in the control dataset (only values 0.05 were accepted). A protein was classified as differentially abundant only when both FDR and p-value criteria were satisfied.

C. difficile exhibits a capsule-like cell surface glycopolymer layer
To investigate whether the capsular polysaccharide layer varied in clinically-relevant isolates, CWG layers were visualized in phylogenetically-diverse strains of C. difficile, including those belonging to outbreak-associated molecular types. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) of ruthenium red-strained C. difficile strain GV60 (ribotype 053; outbreak-associated) revealed the presence of an electron-dense surface CWG layer (Fig 1A). GV60 staining was markedly different from that exhibited by the C. difficile strain 630 (ribotype 012; Fig 1B), suggesting strain-to-strain variations in CWG production.

PSII is a conserved C. difficile-specific antigen
Multiple surface-associated polysaccharides have been identified in C. difficile, and the structures of a subset of the molecules solved, including PSII (Fig 2A) [15,16]. Using a semi-quantitative PSII-specific immunoblot assay (Fig 2B), we detected varying levels of PSII in surface extracts from strains belonging to ten different C. difficile molecular types ("ribotypes"; 078, 003, 027, 017, 001, 053, 106, 010, 015, and 020; Fig 2B), but not from Clostridium perfringens (GV339) or Clostridium botulinum (S1B Fig). Purified PSII was used as a positive control, and pre-immune serum as a negative control to confirm anti-PSII antibody specificity (S1A and S1B Fig). Further, surface-localized PSII was confirmed in multiple isolates of a single C. difficile ribotype (RT027; S1B Fig). Highest PSII levels were noted for a Ribotype 003 isolate; however, this strain also demonstrated some cross-reactivity in a pre-immune serum-alone control immunoblot (S1A Fig). Thus, PSII is a common C. difficile-specific polysaccharide whose abundance varies between different clinical isolates.

C. difficile encodes a putative CWG locus
To identify candidate genes involved in CWG biosynthesis in general, and PSII production in particular, we performed an in silico scan of the C. difficile strain 630 genome for ORFs with predicted glycotransferase, glycopolymer export, and glycopolymer modulatory functions. Similar to Willing et al [21], we identified an approximately 80 kilobase locus; we further observed that ORFs within this region exhibited striking similarity to the CWG biosynthetic genes of S. pneumoniae [21], including those encoding a predicted initiating transferase (CD2783), glycosyltransferases, a flippase-type exporter (MviN), a polymerase (CD2777), and proteins that anchor CWGs to the cell surface (LcpA and LcpB; Fig 3A). RT-PCR analysis of the intergenic regions between mviN and pgm2, pgm2 and manC, manC and CD2778, and lcpA and lcpB suggests that mviN-CD2778 and lcpA-lcpB are co-transcribed (S2A Fig). The sequence-based RAST Prokaryotic Genome Annotation Server [33] revealed that this locus is highly conserved amongst phylogenetically diverse C. difficile strains (Fig 3B and S1 Table). In addition, and notably, this CWG cluster is absent in C. perfringens and C. botulinum, consistent with a putative role in the production of the C. difficile-specific antigen, PSII (Fig 2B).

MviN depletion is associated with decreased PSII deposition
In general, perturbation of the later steps of CWG biosynthesis, including the polysaccharide exporter, result in intracellular accumulation of toxic intermediates, or sequestration of the common lipid carrier from other essential pathways [11,44,45]. As such, attempts to disrupt the corresponding genes in other organisms have been unsuccessful [45][46][47][48]. Consistent with these observations, we were unable to disrupt five genes in the CWG biosynthetic locus (CD2783, mviN, CD2780, CD2779, and CD2769) via ClosTron mutagenesis. However, mviN expression was successfully diminished using an antisense RNA approach, and resulted in multiple phenotypes. First, mviN-specific antisense RNA-expressing bacteria exhibited a growth defect (S3A Fig). Second, though MviN downregulation did not affect total CWG production, there were slight morphology defects as visualized by Maneval's capsule staining (S4 Fig). Third, specific detection and visualization of PSII revealed lower abundance of the polysaccharide in cells expressing mviN antisense RNA (Fig 4A). Consistent with this observation, MviN downregulation also resulted in decreased cell surface PSII staining, as visualized by immunofluorescence (Fig 4B). mviN is the third gene in a cluster of 19 genes, and we confirmed that is co-transcribed with its immediate downstream member pgm2, and that the next two genes were also each co-transcribed with their immediate neighbors (pgm2 and manC and manC and CD2778; S2A Fig). Fourth, fully quantitative proteomic analyses also revealed that in the mviN knockdown strain, expression of multiple CWG gene products was decreased [including mviN itself (1.5 fold) and its immediate downstream genes CD2780 (1.9 fold) and CD2779 (1.6 fold); Table 2]. Expression of other, less proximal genes such as those whose products are predicted to synthesize, anchor and polymerize CWGs, was also significantly perturbed. Therefore, it is likely that multiple gene products in the CWG cluster, including MviN play a role in PSII and CWG deposition.

The lcpB mutant displays altered PSII shedding and deposition
Two genes in the CWG locus, lcpA and lcpB, were amenable to insertional inactivation, (lcpAand lcpB -, respectively) [26]. In Gram-positive organisms, LCP proteins play several roles in cell wall biosynthesis, including the anchoring of teichoic acid and/or capsular polysaccharide The orientation of the genes in the putative CWG locus is shown along with the generalized predicted functions of each gene product [21]. The color of each gene corresponds to its predicted function shown in the legend below the locus. (B) The locus is highly conserved at the amino acid level (95%-100%) across diverse C. difficile strains. A total of 32 clinical isolates were analyzed and compared to the reference strain CD630 (boxed in green). A list of all the strains used is provided in S1 Table. doi:10.1371/journal.ppat.1005946.g003  to the cell wall [49][50][51][52][53][54]. The function and contribution of LCP proteins to CWG biosynthesis in C. difficile, however, is unknown. In Staphylococcus aureus, LCP mutants shed more capsular polysaccharide from their cell surface as a result of improper anchoring of the polymers to the cell wall [50,51]. Thus, we investigated PSII shedding to explore a possible CWG anchoring function of LcpA and LcpB in C. difficile. Relative to the parent strain, more PSII was shed from the lcpBmutant; this was significantly reversed in plasmid-complemented strains (Fig 5A). Only modest changes were observed for the lcpAmutant. Overall, the variation in shed PSII levels described above was not due to PSII production differences, since cytosolic PSII abundance was unchanged in parent and all mutant strains (S7 Fig). Notably, in the parent strain, co-localization with FM464, a plasma membrane-specific dye, confirmed functional surface presentation of PSII (Fig 5B). In contrast, PSII anchoring to the cell surface was defective in both lcpAand lcpBmutants as evidenced by loss of co-localization with FM4-64; this was most severe in the lcpBstrain. Additionally, the lcpBmutant also displayed altered cell morphology with curved, elongated cells (S5 Fig). Plasmid complementation of both lcpAand lcpBmutants restored PSII co-localization with FM4-64 similar to the parent strain, and also rescued the lcpBstrain morphology defect.

The lcpBmutant exhibits morphological defects
Many bacteria have common lipid precursors for CWG and peptidoglycan biosynthesis, and cell shape is determined by appropriate deposition of each component in the cell wall [11,55]. The lcpBmutant, but not the lcpAmutant, displayed a modest growth defect (S3B Fig) with bacterial titers approximately one order of magnitude less than those of the parent strain at both exponential and stationary growth phases (S2 Table). Curiously, the lcpBmutant also had remarkably altered morphology (S5 Fig), with cells appearing longer and thicker, harboring multiple septa, and with marked curvature; this was vividly apparent via scanning electron microscopy (Fig 6). Transmission electron microscopy of the lcpBstrain also revealed improper cell septum localization and structure, as well as a more diffuse cell wall. These morphological defects were significantly rescued via complementation using plasmid-encoded lcpB.
Cell wall remodeling and alterations were also explored in both Lcp mutants. First, total Surface-Layer Proteins (SLPs), integral to the C. difficile para-crystalline surface layer, were visualized. The SLP profile of lcpAmutant and its complement were comparable to that of the parent strain. However, multiple low molecular weight proteins were recovered from the lcpBmutant SLP extracts; these were mutant-specific since complementation with plasmid-encoded LcpB rescued the phenotype (Fig 7A). Second, biofilm formation was assessed for all strains; the lcpBmutant, but not lcpAmutant, produced robust biofilms compared to the parent strain, a phenotype that was partially complemented (Fig 7B). SDS-PAGE analyses are representative of a minimum of three biological replicates.
Of note, neither mutant produced altered levels of the large clostridial toxins TcdA and/or TcdB (S8 Fig). Finally, the lcpAstrain also exhibited no alterations in susceptibility to the antimicrobials vancomycin and metronidazole, or the cationic peptide LL-37 [measured as zone of inhibition (ZOI) and minimum inhibitory concentration (MIC) S3 Table]. The lcpBstrain could not be tested for these phenotypes due to a severe growth defect in the recommended MIC medium [40,56].
The lcp mutants are hypervirulent in the hamster model of acute CDI Given the pleiotropic alterations in the lcpBmutant, including changes to surface glycopolymer profiles, and considering that there were no differences in toxin production in the mutant Minimal increase in shed PSII from the lcpAmutant [lcpA -(Vector); Row 2] and corresponding plasmid complementation (lcpA -plcpA; Row 3) is also shown. UD, "undiluted". All immunoblots are representative of a minimum of two biological replicates. (B) Immunofluorescence demonstrating PSII co-staining with the cellsurface-specific dye FM4-64 on the isogenic parent (Row 1 images). Alteration and re-localization of PSII in the lcpAstrain (Row 2 images) and lcpBstrain (Row 4 images). Altered morphology, with a curved and elongated phenotype and multiple septae are also visible in the lcpBstrain. Complementation-mediated restoration of PSII co-localization with FM4-64 in both mutants is shown in Rows 3 and 5 images respectively, as well as the morphology defect rescue in lcpBstrain (Row 5 images). compared to the parent strain, we sought to determine if cell-surface-associated phenotypes influenced C. difficile virulence in the acute Golden Syrian hamster model of infection. In a pilot study, hamsters infected with the mutants (four hamsters infected with the lcpAmutant and five hamsters infected with the lcpBmutant) succumbed to the infection much earlier than the isogenic parent strain-infected control (Fig 8A). This hypervirulent phenotype was confirmed for the lcpBmutant in a powered study, where the average time to death for the mutant-infected hamsters (n = 10) was approximately 50% less than animals infected with the isogenic parent strain (n = 5) (Fig 8B; average time to death was 65 hours for mutant-infected hamsters, and 152 hours for wild-type infected hamsters). Additionally, cecal contents from the lcpBmutant-infected hamsters yielded bacteria with helical morphology (Fig 8B, inset).

Discussion
In this study, we confirmed the presence of a surface CWG layer in C. difficile strains belonging to diverse phylogenic groups (ribotypes). Further, our data suggest that several genes within a CWG biosynthetic locus are essential for C. difficile viability. Specifically, our studies are consistent with a key role for the putative flippase ortholog MviN, and the LytR-CpsA-Psr (LCP) ortholog LcpB, in surface deposition of the C. difficile-specific polysaccharide PSII, and in the regulation of bacterial cell shape and virulence. diffuse cell wall structure (black arrow) is shown. Scale bar for TEM is 100nm. All images are representative of a minimum of 10 fields visualized, and at least two biological replicate preparations.
doi:10.1371/journal.ppat.1005946.g006  (Bars 1, 2 and 3), but the lcpBmutant [lcpB -(vector)] produces a more robust biofilm, that is only partially restored to wild-type levels via plasmid complementation (Bars 4 and 5). Three biological replicates (each in technical triplicate) were performed for each strain. The error bars are represented by standard deviation estimates as well as Student's t tests to compute differences between WT (Vector) and lcpB -(Vector). Significance is p < 0.05. doi:10.1371/journal.ppat.1005946.g007 C. difficile MviN displays 30% identity to its Escherichia coli ortholog. MviN is predicted to be a flippase, with a role in exporting lipid II for peptidoglycan biosynthesis, and is essential for viability in E. coli [57]. Consistent with a role as a flippase, mviN downregulation in C. difficile resulted in reduced surface PSII display. Since mviN appears to be co-transcribed with other member(s) of the CWG locus, and since down-regulation of mviN expression results in concomitantly altered abundance of other CWG locus products including LcpB, its precise contribution to the phenotypes described above remains to be established.
Of the three known CWG biosynthetic pathways, the C. difficile CWG cluster is similar to that of the streptococcal Wzx/Wzy pathway [22,45,48], which assembles complex polysaccharides, each consisting of multiple monosaccharides in a single repeating unit with a branched structure [48]. Correspondingly, PSII is comprised of repeating hexasaccharide units, and also has a branched structure [15]. The main components of the Wzx/Wzy pathway include an initiating transferase, glycosyltransferases specific for the monosaccharide units in the CWG repeating unit, a flippase-type exporter, a polymerase, and enzymes usually from the LCP protein family, that anchor the CWG to the cell surface [22,23,48]. Similar gene products, notably the putative MviN flippase, and LcpA and LcpB are encoded within the C. difficile CWG biosynthetic cluster. Further, and reminiscent of the S. pneumoniae Wzx/Wzy system, several genes in the C. difficile cluster could not be disrupted. As in S. pneumoniae, it is likely that inactivation of genes involved in the later steps of C. difficile CWG biosynthesis leads to accumulation of toxic intermediates, or sequesters common, limiting, lipid precursors in dead-end products [11,44,45]. Consistent with the apparent specificity of PSII to C. difficile, genes in the CWG locus display synteny and are highly conserved within this species, but absent in other related clostridia. Since both the mviN-knockdown strain and the lcpBmutant display alterations in PSII deposition, the genes in this cluster likely direct multiple aspects of C. difficile CWG biology, including export. Collectively, our data, as well as the published literature, particularly in the S. pneumoniae system, support the accompanying model for PSII biosynthesis in C. difficile (Fig 9; [22]).
Recently, there has been an increasing appreciation for a direct role for LCP proteins in anchoring CWGs to the bacterial cell wall [49]. LCP proteins are widespread amongst Gram-positive bacteria, and most organisms harbor more than one homolog within their genome [49][50][51][52]. The solved crystal structure of the S. pneumoniae LCP homolog Cps2A contained tightly-bound pyrophosphoryl-lipid precursors of CWGs, suggesting a role for LCP proteins in anchoring teichoic acids (TA) and/or capsular polysaccharide (CPS) to peptidoglycan, most likely at the C6-OH linkage of N-acetylmuramic acid residues; this function has also been explored in Staphylococcus aureus and Bacillus subtilis [49][50][51][52]. All three organisms encode multiple LCP homologs in their genomes, with partial functional redundancy. While single LCP mutants displayed modest defects, double-and triple-LCP mutants shed CPS and/or TAs into the medium, and had severe growth and morphology defects [49][50][51][52]. In contrast, our data suggest non-redundant roles for C. difficile LcpA and LcpB, despite 64% amino acid identity. Relative to the lcpBstrain, the lcpAstrain displayed modest alterations, particularly in the in vitro phenotypes. Given the gene organization and RT-PCR analysis (S2B Fig), it is likely that the lcpAdefects are due to polar effects on LcpB expression. In contrast, the lcpBstrain displayed a range of phenotypes, which were at least partly, if not fully, complemented by plasmid-encoded lcpB.
The lcpBmutant sheds more PSII into the supernatant compared to its isogenic parent, and this results in little to no PSII surface staining; therefore, we conclude that LcpB is required for maintenance of deposited PSII on the bacterial surface. It is as yet unclear where PSII is anchored; a recent study reveals roles for LCP proteins beyond peptidoglycan-mediated anchoring [53].
In addition to a defect in PSII biosynthesis and deposition, the lcpBmutant also displays significant morphological alteration(s). In Bacillus anthracis, one of six LCP homologs (lcpD) is required for proper attachment of CWGs to peptidoglycan. However, LcpD, in concert with another homolog LcpB3, is also involved in maintaining cell size and proper chain length formation [54]. The authors speculated that this extra role could be due to differences in cell morphology between cocci versus bacilli, as bacilli require extra machinery to maintain their rod shape, whereas cocci do not. Interestingly, mviN-knockdown also impacted cell morphology, resulting in curved cells (S4 Fig). It is possible that CWG biosynthetic genes, including LCP proteins are associated with enzymes involved in maintaining proper rod-shape, and that disruption of the corresponding genes may lead not only to improper anchoring of CWGs, but also a defect in peptidoglycan deposition and/or rod shape formation [49]. We therefore propose that lcpB disruption perturbs PSII deposition as well as other processes in cell wall biosynthesis.
Given the pleotropic CWG-and morphology-altering impacts of lcpB disruption, the marked hypervirulence of the lcpBmutant (as well as lcpAstrain) in hamsters was unexpected. The increased virulence could not be attributed to alterations in toxin levels since the amount of toxin produced by the mutants was not significantly different from the isogenic parent strain. Further, the presence of the ermB cassette cannot be considered to be a contributing factor toward virulence since published mutants in the JIR8094 background have displayed both hypervirulent [58] and avirulent [59][60][61] phenotypes in the same hamster model of infection. There are several other possible explanations for the heightened virulence of our mutants, all implicating a role for non-toxin virulence factors. First, reduced surface capsular polysaccharides (CPS) may influence bacterial adherence to epithelial cells [62,63]. Streptococcus pneumoniae colonizes more efficiently when it produces less CPS, possibly due to enhanced attachment to epithelial cells via exposed adhesins [32]. We performed adhesion assays to Caco-2BBe human intestinal epithelial cells [42] but observed either no difference (lcpAmutant) in adhesion compared to the parent strain, or decreased adhesion (lcpBmutant; S9A  Fig). Thus, it is unlikely that epithelial cell adhesion contributes to the increase in virulence in either mutant. Second, the altered surface morphology may facilitate biofilm formation [64]. Indeed, our studies revealed robust biofilm formation by the lcpBstrain (Fig 7B), but it is unlikely that this phenotype profoundly influences virulence since hamster lethality was observed in less than 48 hours. Third, and most plausible for this study, depletion of CWGs may uncover pro-inflammatory bacterial ligands, as has been noted for the unencapsulated S. suis mutants [64]. Thus, the lack of cell surface PSII deposition in the lcpBstrain could expose other, pro-inflammatory, surface factors. The lcpBmutant displays an altered cell wall protein profile (Fig 7A), specifically the presence of lower molecular weight products that are absent in the parent strain extracts. Consistent with the display of altered ligands in the mutants, human intestinal epithelial cells treated with surface layer extracts from the lcpBstrain secreted increased amounts of the pro-inflammatory cytokine Interleukin-8 (IL-8; S9B Fig; two independent methodologies). Fourth, a remodeled cell surface may alter susceptibility to killing by host antimicrobial peptides, a first line of defense during infection. The lcpAstrain showed similar susceptibility as the parent strain to LL-37, an intestine-specific cathelicidin antimicrobial peptide ( [40]; S3 Table; the lcpBstrain could not be tested because it failed to grow robustly on the approved testing medium). Finally, increased shedding of PSII may be directly hyper-immunostimulatory and thus detrimental. However this is relevant only to the lcpB -Fig 9. C. difficile PSII biosynthesis model [22,23,48]. One repeating unit of PSII is assembled on a lipid carrier (undecaprenyl phosphate) in the bacterial cytoplasm, exported to the cell surface, polymerized and anchored to peptidoglycan, the cell membrane, or cell wall proteins. Sequentially: (1) a predicted initiating transferase (CD2783) transfers the first sugar of the PSII repeating unit to the lipid carrier; (2) a second glycosyltransferase catalyzes the committed step of the pathway by transferring the second sugar to the repeating unit, and cytoplasmic glycosyltransferases, including ManC and Pgm2, synthesize the rest of the PSII unit; (3) a polysaccharide flippase (MviN) transports the PSII repeating unit from the cytoplasm to the cell surface; (4) a polymerase (CD2777) polymerizes the PSII chain extracellularly and the lipid carrier is recycled by an unknown mechanism (indicated by the asterisk); (5) finally, surface-anchoring factors (LcpA and/or LcpB) catalyze the transfer and anchoring of fully polymerized PSII to peptidoglycan or the cell membrane. Consistent with the NMR determinations derived in Ganeshapillai et al. [15], we have included the phosphate unit only on the mannose residue on polymerized PSII. However, the mechanistic basis of this phospho-sugar linkage has not been explored.
doi:10.1371/journal.ppat.1005946.g009 mutant, and therefore cannot be invoked as a common mechanism for increased virulence. Future studies will aim to parse between the above possibilities.
For all the results described in this work, we employed the C. difficile strain JIR8094, an erythromycin-sensitive derivative of the clinical isolate CD630. The primary rationale for this was two-fold. First, strain 630 and its derivatives have a more prolonged lethality profile in the hamster model of CDI compared with other isolates (range 4-7 days; [7,59,60]), thus enabling robust time-to-death estimations, and second, multiple molecular assessments have been performed in this strain background allowing for comparison between published studies. However, as proof-of-principle, lcpB was also disrupted in a ribotype 027 (outbreak-associated) isolate, as well as a non-toxigenic strain of C. difficile (S10 Fig). In both cases, preliminary bright-field microscopy visualizations revealed altered bacterial morphology similar to that observed with the JIR8094 lcpBstrain (Fig 6).
In conclusion, we have shown that proteins encoded by the C. difficile CWG locus, especially LcpB, influence PSII shedding, cell shape, and virulence. Future studies will focus on elucidating the steps involved in PSII attachment to the cell wall, as well as an investigation of the impact of cell morphology and cell surface composition on C. difficile-induced host immune responses. PSII has been proposed as a vaccine candidate for CDI prevention, and several studies have assessed its immunogenic and protective potential [18,19,65,66]. Since PSII, and possibly other CWGs, fundamentally affect key aspects of C. difficile pathogenesis, it is imperative to understand CWG biosynthetic mechanisms to optimally engineer next-generation anti-CDI biotherapeutics and other potential vaccine candidates.  Table. LL-37 susceptibility for the isogenic parent (JIR8094) and lcpAmutant strain. A scrambled LL-37 control is also shown. The numbers are representative of 1 biological replicate for the isogenic parent and 2 biological replicates for the lcpAmutant. Susceptibility to vancomycin and metronidazole (ZOI measurements). The numbers are representative of 3 biological replicates for the isogenic parent, and 2 biological replicates for the lcpAmutant. (DOCX) S1 Fig. Anti-PSII antiserum specificity. (A) Purified PSII was used as a positive control. Following CWG extraction using the EDTA-TEA method, EDTA-TEA buffer was probed with PSII-LTB rabbit antiserum (top) and extracts from various strains were probed with preimmune serum to verify lack of cross-reactivity (bottom). Ribotype 003 exhibited slight crossreactivity with the pre-immune serum, but the amount is negligible compared to the total amount of PSII detected in Fig 2B. (B) PSII was also detected from other 027 ribotypes as well as Clostridium perfringens and Clostridium botulinum. PSII-LTB rabbit antiserum reacts robustly with all 027 strains tested but not C. perfringens or C. botulinum (top). These strains were also probed with pre-immune serum to verify lack of cross-reactivity (bottom). The REA types of the 027 strains tested are as follows: 027-1 = BI-1 (GV44), 027-2 = BI-6 (GV45), 027-3 = BI-8 (GV46), 027-4 = BI-17 (GV48), and 027-5 = BI-23 (GV51) (UD, "undiluted"; the ramps indicate increasing to decreasing concentration from left to right). (TIF) Maneval's capsule staining was performed for the lcpAand lcpBmutants (A). Brightfield images were taken using the EVOSfl microscope for WT (Vector), lcpB -(Vector) and lcpB -plcpB (B). There is noticeable alteration in cell morphology of the lcpBbut not lcpAmutant compared to the wild-type strain (A), and this difference is partially complemented with a plasmid-encoded copy of lcpB (B). (TIF)

S6 Fig. Normalization for PSII immunoblots.
Fifty micrograms of total protein from the "shed" (supernatant) fraction of WT (Vector), lcpA -(Vector), lcpA -plcpA, lcpB -(Vector) and lcpB -plcpB were run on a 4-20% TGX gel and stained with Gel Code Blue. The staining was evenly distributed between all strains indicating normalized protein in the "shed proteins" fraction. A representative gel is shown for a total of two biological replicates for each strain. mutants produce similar levels of toxin compared to the isogenic parent strain control. Data are normalized to total protein in the supernatant (represented by arbitrary units, AU). Three biological replicates were performed for each strain in each set. A Student's t test was performed and indicated no significance between any of the strains (p > 0.05). The error bars denote standard deviation. (TIF) S9 Fig. Epithelial cell adhesion assays and IL-8 analysis. (A) Adhesion to Caco-2BBe epithelial cells was determined in the lcpAand lcpBmutants compared to the wild-type strain. The target MOI was 50 for the wild-type and lcpAmutant, but 100 was used for the lcpBmutant due to variability in percent adhesion with lower MOIs. The percent adhesion of both mutants trended less than the wild-type strain, but a Student's t test indicate that the differences are not significant in either mutant (p > 0.05). A total of three biological replicates were done four times for each strain and error bars represent standard deviation. (B) IL-8 secretion from HT-29 epithelial cells was determined by stimulation of HT-29's with 250μg of total protein from surface layer extracts from the lcpBmutant and wild-type strains. Data from two different methodologies are shown (ELISA, bar graph, 1 biological replicate; and Cytokine Profiler Array, immunoblot, 1 biological replicate); purified flagellin from Salmonella typhimurium is the positive control (blue bar), and all samples were normalized using no-protein control.