Identification, Immunogenicity and Crossreactivity of Type IV Pilin and Pilin-like Proteins from Clostridium difficile

Abstract

The Gram-positive anaerobe Clostridium difficile is the major cause of nosocomial diarrhea; manifestations of infection include diarrhea, pseudomembranous colitis, and death. Genes for type IV pili, a bacterial nanofiber often involved in colonization and until relatively recently described only in Gram-negatives, are present in all members of the Clostridiales. We hypothesized that any pilins encoded in the C. difficile genome would be immunogenic, as has been shown with pilins from Gram-negative organisms. We describe nine pilin or pilin-like protein genes, for which we introduce a coherent nomenclature, in the C. difficile R20291 genome. The nine predicted pilin or pilin-like proteins have relatively conserved N-terminal hydrophobic regions, but diverge at their C-termini. Analysis of synonymous and nonsynonymous substitutions revealed evidence of diversifying selective pressure in two pilin genes. Six of the nine identified proteins were purified and used to immunize mice. Immunization of mice with each individual protein generated antibody responses that varied in titer and crossreactivity, a notable result given the low amino acid sequence identity among the pilins. Further studies in other small mammals mirrored our results in mice. Our results illuminate components of the C. difficile type IV pilus, and help identify targets for an anti-C. difficile vaccine.

INTRODUCTION

Clostridium difficile is a Gram-positive, spore-forming, rod-shaped obligate anaerobe, now the leading cause of human health-care associated diarrhea. The bacterium was first isolated by Hall and O’Toole in 1935, and initially termed Bacillus difficilis owing to initial difficulties in culturing the organism (Hall and O’toole, 1935). Infection with the bacterium has a variety of manifestations, ranging from asymptomatic colonization of the colon to copious diarrhea, pseudomembranous colitis, and death (Kelly and LaMont, 2008). While antimicrobial therapy for infection is available, treatment often fails and relapse is common. Although the exact sequence of events in initial colonization with C. difficile is still under investigation, evidence from other intestinal pathogens suggests that attachment to epithelial cells, mediated by pili or fimbriae, non-fimbrial adhesins, or other surface molecules, is a requisite step in pathogenesis (Finlay and Falkow, 1997;Pizarro-Cerdá and Cossart, 2006).

Type IV pili (T4Ps) are bacterial surface appendages that mediate adherence, colonization, DNA transfer, and twitching motility, among other functions. The T4P structural subunits are called pilins, which derive from a precursor prepilin form after removal of a specific N-terminal peptide and modification of the nascent N-terminus by a prepilin peptidase (Strom et al., 1993). Other proteins with similar structures to those of pilins, termed pilin-like-proteins or minor pilins, are also cleaved by the prepilin peptidase and are involved in pilus biogenesis and dynamics (Winther-Larsen et al., 2005;Giltner et al., 2010). A complex nanomachine assembles pilin monomers into filaments that can extend several micrometers from the bacterial cell (Craig et al., 2004). In many bacterial species the pilus can also retract; the energy for extension and retraction are provided by distinct ATPases (Whitchurch et al., 1991). T4Ps have been long known and well studied in Gram-negative organisms, but only relatively recently have been better appreciated in Gram-positive bacteria and archaea. Genes for a type IV pilus were identified in the C. difficile genome (Varga et al., 2006) within the last decade, although fimbrial structures on C. difficile were observed by electron microscopy nearly two decades previously (Borriello et al., 1988).

Studies in Gram-negative bacteria indicate that pilin subunits are immunogenic. The major structural subunits from Pseudomonas aeruginosa, Neisseria gonorrhoeae, Burkholderia mallei and Vibrio cholerae T4Ps have been found to stimulate an immune response in mice and other small mammals (Koga et al., 1993;Forest et al., 1996;Fernandes et al., 2007b;Voss et al., 1996).. Antibodies to major pilus subunits were detected in patients with cholera (Attridge et al., 2004) and in children infected with enteropathogenic Escherichia coli (EPEC) (Martinez et al., 1999), and individuals experimentally infected with N. gonorrhoeae generated an antibody response to hypervariable regions of the major pilin (Forest et al., 1996). Volunteers infected with EPEC demonstrate specific anamnestic responses against homologous pilin upon rechallenge (Fernandes et al., 2007a).

T4Ps are immunogenic, extracellular, expressed at initial stages of infection, and composed of thousands of identical repeating subunits; thus, they present appealing vaccine targets. In passive transfer experiments, antibodies to the main V. cholerae pilin, TcpA, are protective against lethal cholera challenge in an infant mouse model (Sun et al., 1989). Moreover, trials of immunization with T4P subunits or whole pili have proven successful in conferring protection against V. cholerae and Dichelobacter nodosus (Voss et al., 1996;Stewart et al., 1985); further, a Moraxella bovis whole-pilin veterinary vaccine is commercially available (Piliguard^® Pinkeye TriView, Merck Animal Health). A vaccine directed against the C. difficile T4Ps may prove effective in preventing C. difficile colonization and disease.

Location of essential C. difficile T4P components and verification of their functions is still underway. We hypothesized that there would be multiple genes for pilins, minor pilins, or pilin-like proteins within the C. difficile genome, and furthermore that any pilins would be immunogenic as has been demonstrated with pilins of various Gram-negative organisms. Here, we demonstrate the presence of numerous T4P pilin genes in multiple strains of C. difficile, as well as their immunogenicity in mice and other small mammals.

MATERIALS AND METHODS

Pilin gene identification

Three C. difficile pilin genes (pilA1, pilA2, and pilU) were previously identified in strain 630 (Varga et al., 2006). The presence of those pilins in R20291, a BI/NAP1/027 strain isolated in 2006 from the Stoke Mandeville hospital outbreak in Buckinghamshire, United Kingdom, was confirmed by BLAST search. The additional pilins were identified by BLAST search of the R20291 genome using amino-terminal sequences from the previously identified pilins, and by use of PilFind (Imam et al., 2011).

Analysis of pilin nucleotide sequences

Nucleotide sequences for each pilin were obtained from GenBank/NCBI or PATRIC. Sequences were aligned in Clustal Omega and manually edited. Rates of synonymous and nonsynonymous substitutions were calculated with SNAP (http://hcv.lanl.gov/content/sequence/SNAP/SNAP.html). Calculation of ω, tests for positive selection, and identification of positively-selected sites was completed using CodeML from PAML version 4.7 (http://abacus.gene.ucl.ac.uk/software/paml.html). Three models from CodeML were used: M0, which estimates an overall ω for the entire sequence; M7, which uses the beta function to model the distribution of ω over 0 ≤ ω ≤ 1 and accounts for sites with low purifying selection (Yang 2005); and M8, which adds another component to M7 to better model ω in situations where some (but not all) codons are under positive selection. For each pilin, M7 and M8 were compared using a likelihood-ratio test (LRT) tests for the presence of positively selected sites(Yang and Bielawski, 2000); a Bayes empirical Bayes analysis identified specific sites under selection in pilins for which the LRT was positive (Yang et al., 2005). Maximum-likelihood phylogenetic trees for use in CodeML were assembled in MEGA 5.2.

Protein expression and purification

DNA sequences of seven pilin genes lacking the codons for the signal peptide and the N-terminal hydrophobic domain were codon-optimized for expression in E. coli, commercially synthesized (Genscript), cloned into the pET30b vector downstream of the hexa-histidine tag sequence, and transformed into Escherichia coli BL21(DE3) cells (Invitrogen). The precise codon-optimized sequences for each pilin are listed in Supplemental Table 1. After inoculation of 1 L Luria broth + kanamycin with 20 mL turbid overnight culture, cells were grown to OD600 = 0.5 at 30 °C and induced with 1.0 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). In pilot studies to determine optimal pilin expression conditions, 100 mL flasks of Luria broth with kanamycin were inoculated with 2 mL of turbid overnight culture and grown at 30 °C or 37 °C to OD600 = 0.5, at which point cultures were induced with 1.0 mM IPTG. One milliliter samples were taken from each flask hourly for 5 hours after induction, after induction overnight and 24 hours. Samples were centrifuged, resuspended in 100 μL Laemmli buffer, boiled for 10 minutes, separated by SDS-PAGE, and Coomassie stained. The stained gels were scanned with an Odyssey imaging system, and the intensity of the pilin bands and a control band were quantified. The combination of temperature and induction time with the highest ratio of pilin band intensity to control band intensity was selected as the optimal pilin expression condition. After optimized expression for each pilin, cultures were pelleted by centrifugation at 5000 × g for 10 minutes at 4 °C (Beckman Coulter); pellets were stored at −20 °C. Cell pellets were resuspended in 50 mM NaH₂PO₄, 300 mM NaCl, 20 mM imidazole, pH 8.0 with protease inhibitors (Roche) and lysed in a French pressure cell at 1200 psi (SLM Aminco); lysates were centrifuged at 35000 × g for 30 min. Supernatants containing each fusion protein were applied to nickel-nitrilotriacetic acid (Ni-NTA) agarose (Qiagen) and incubated, with rotation, at 4 °C for 1 hr. After washing, protein was eluted from the resin with increasing concentrations of imidazole in 50 mM NaH₂PO₄, 300 mM NaCl, pH 8.0. Column fractions of PilU, PilV, and PilW were further purified by size-exclusion chromatography using a Sephacryl S-100 column. For ELISA assays, immunoblotting and immunoadsorption, the N-terminal purification tag was cleaved from each purified pilin protein with recombinant enterokinase (Novagen) and removed by incubation with Ni-NTA resin.

Antibody generation

Purified tagged pilin protein samples were sent to Rockland Immunochemicals, where they were used to immunize mice. Pre-immunization bleeds were taken 1 day prior to intradermal immunization with 50 μg purified pilin plus Freund’s complete adjuvant. Mice were boosted with 50 μg antigen and incomplete Freund’s adjuvant intradermally at 7 days post immunization, and subcutaneously with 50 μg antigen and incomplete Freund’s adjuvant at 14 and 33 days post-immunization. Terminal bleeds were taken 59 days after immunization.

Polyclonal antibodies to untagged PilA1 were raised in guinea pigs, antibodies to untagged PilA2 were raised in rats, and antibodies to untagged PilJ were raised in rabbits (Rockland Immunochemicals). For all animals, pre-immunization bleeds were taken prior to initial intradermal immunization. Guinea pigs were immunized intradermally with 100 μg antigen plus complete Freund’s adjuvant, and boosted on day 7 intradermally, and 14 and 28 subcutaneously, with 50 ug protein and incomplete Freund’s adjuvant. Test bleeds were taken on day 38; guinea pigs were boosted again, and terminal bleeds taken on day 80. Rats were immunized intradermally with 100 μg antigen plus complete Freund’s adjuvant, and boosted on day 7 intradermally, and 14 and 28 subcutaneously, with 50 ug protein and incomplete Freund’s adjuvant. Test bleeds were taken on day 38. The terminal bleed for one rat was taken on day 59; the two remaining rats were boosted again and terminal bleeds were taken on day 80. Rabbits were immunized intradermally with 400 μg antigen plus complete Freund’s adjuvant, boosted on day 7 intradermally, and 14 and 28 subcutaneously, with 200 ug protein and incomplete Freund’s adjuvant. Test bleeds were taken on day 38, and terminal bleeds taken on day 59.

ELISA

Unless otherwise noted, all solutions were used at 50 μl/well. Nunc Maxisorp 96-well plates were coated overnight with purified cleaved pilin or pilin-like protein, brought to 10 μg/mL in phosphate-buffered saline with 0.05% Tween-20 (PBST) 50 μl/well. Blank wells were coated with plain PBST. After coating, plates were blocked with 5% bovine serum albumin (Sigma) in PBST for 1 hr at 37 °C, 100 μL/well. Serum samples diluted 1:500 in PBST were added and serially diluted with one volume PBST in plate to a final volume of 50 μL/well. All sera were run in triplicate. Normal mouse serum (KPL) was loaded at 1:500 in PBST. Blank wells were loaded with PBST. Samples were incubated on plate for 2 hours at room temperature. Peroxidase-tagged goat anti-mouse-IgG (H + L) (KPL) was added at a 1:1,000 dilution and incubated for 30 minutes at 37 °C. Plates were developed with Sureblue Safestain (KPL) for 30 minutes at room temperature. Optical density at 655 nm (OD655) was read with a microplate reader (BioRad model 680). Blanks were averaged and subtracted from the sample and standard wells. Normal mouse serum (KPL) was used to provide a standard against which the experimental serum could be judged. The average plus two standard deviations of the OD655 with normal mouse serum was taken as the nonspecific normal mouse background OD. For experimental samples, triplicate wells were averaged; the highest dilution with an OD655 greater than normal mouse background was taken as the antibody titer.

Immunoadsorbance and immunoblotting

Upon terminal bleed, guinea pig antibodies to PilA1were found to cross-react with PilA2. This crossreactivity to PilA2 was eliminated by adsorption to PilA2 immobilized on Aminolink columns (Thermo Scientific). Columns were prepared according to kit protocols. Briefly, tagged PilA2 in 0.1 M NaH₂PO₄ was incubated with resin and 50 mM NaCNBH₃ at 4 °C overnight with end-over-end rotation. Unbound PilA2 was washed from resin with additional 0.1 M NaH₂PO₄ buffer. For immunoadsorption, guinea pig anti-PilA1 was diluted 1:1 in 0.1 M NaH₂PO₄ and incubated with resin at 4 °C for 1 hour with rotation. Unbound antibody was washed from column and concentrated back to original volume.

Purified untagged pilins were boiled for 10 minutes in Laemmli buffer and applied to precast 4–15% gradient Mini-PROTEAN TGX polyacrylamide gels (Bio-Rad). Proteins were transferred to polyvinylidene difluoride (PVDF) membranes, blocked for 1 hour with 5% non-fat dry milk, then incubated at 4 °C overnight with polyclonal rabbit anti-PilJ at a 1:10,000 dilution, polyclonal rat anti-PilA2 at 1:1000 dilution, or polyclonal guinea pig anti-PilA1 at 1:1,000 dilution. Membranes were washed in PBST, incubated with IRDye 800CW donkey anti-rabbit IgG H+L, donkey anti-guinea pig IgG H+L, or goat anti-rat IgG H+L (Li-Cor Biosciences) as appropriate for 1 hour, and infrared signals were detected and quantified using the Odyssey imaging system (Li-Cor Biosciences).

RESULTS

T4P pilin gene sequences, analysis, and nomenclature

The C. difficile R20291 genome encodes at least nine putative pilin or pilin-like proteins (Table 1). Pilins were identified by BLAST searches based on the pilin genes previously identified in strain 630 (Varga et al., 2006) and by using PilFind (Imam et al., 2011). Four of these genes are located in the main T4P gene cluster, one is located in a smaller T4P cluster, two are located near one another, and the remaining three are scattered throughout the genome (Figure 1). We have named each pilin in accordance with prior T4P nomenclature as well as its predicted function.

Table 1.

Nomenclature and size of nine pilin genes in C. difficile R20291

NCBI gene identifier	NCBI protein accession number	Mature Protein Designation	Mature pilin predicted length (aa)	Mature pilin predicted size (kDa)
CDR20291_3350	YP_003219825	PilA1	164	17.5
CDR20291_3155	YP_003219630	PilA2	108	11.7
CDR20291_1084	YP_003217582	PilA3	168	19.4
CDR20291_0683	YP_003217184	PilJ	262	29.2
CDR20291_3343	YP_003219818	PilK	500	57.7
CDR20291_3344	YP_003219819	PilU	166	18.8
CDR20291_3345	YP_003219820	PilV	178	20.6
CDR20291_2191	YP_003218677	PilW	158	17.3
CDR20291_1081	YP_003217579	PilX	113	13.7

Figure 1.

Although each predicted pilin protein is relatively well conserved across strains, the nine pilins diverge from each other. All the predicted prepilin proteins share the N-terminal prepilin peptidase cleavage site and a hydrophobic N-terminal domain, but their amino acid sequences differ significantly after this region (Tables 1 and 2, Figure 2). Mature pilins are predicted to range from 108 to 500 amino acids in length, and 11.2 to 57.7 kDa in mass. Of the nine pilins, eight are present in all 18 C. difficile genomes annotated with protein sequences available in the NCBI genome database; the only one not present in all genomes is found in thirteen of the eighteen genomes. All strains analyzed are capable of producing both toxin A and toxin B. Amino acid sequences of each of the nine pilins are well conserved across strains; minimum and average percent identity for each pilin range from 76% to 96% and 95% to 99% (relative to the gene in R20291) respectively, across the eighteen strains (Supplemental Table 2).

Table 2.

Pilin amino acid sequence percent identity and maximum identity within C. difficile R20291

Subject sequence	Query Sequence
	PilA1		PilA2		PilA3		PilJ		PilK		PilU		PilV		PilW		PilX
	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity	Percent coverage	Maximum identity
PilA1	--	--	79	29	82	28	18	63	6	21	32	38	22	30	37	74	26	30
PilA2	56	29	--	--	31	24	41	29	2	33	10	37	9	100	45	35	9	33
PilA3	76	28	39	24	--	--	No significant similarity		10	33	12	32	34	35	57	50	19	83
PilJ	28	63	94	29	No significant similarity		--	--	6	44	49	50	26	30	45	52	56	30
PilK	19	21	15	33	40	24	18	44	--	--	72	35	31	24	17	45	16	50
PilU	21	48	15	37	No significant similarity		45	48	38	29	--	--	74	24	93	29	91	22
PilV	24	30	15	100	24	35	18	30	13	24	52	32	--	--	27	32	10	46
PilW	43	74	63	37	8	50	21	58	5	33	10	50	15	50	--	--	9	33
PilX	19	30	10	33	13	83	23	30	4	50	90	22	6	46	17	28	--	--

Figure 2.

The gene CDR20291_3350 is the first open reading frame in the main T4P gene cluster, and shows the greatest polymorphism across all C. difficile strains. For these and other reasons (see discussion), it is our hypothesis that this protein is the major pilin of the C. difficile T4P, and thus propose pilA1 as the gene designation and PilA1 the name of the mature protein. In several other T4P systems, the ‘A’ designation indicates the major pilin, as in TcpA of V. cholerae and PilA of P. aeruginosa. This designation is not universal, however; the major pilins in Neisseria species are designated PilE.

The gene CDR20291_3155 is predicted to encode a small pilin located in the small T4P gene cluster. We speculate that the presence of two or more accessory genes in that cluster hints at the possibility of distinct pili, with this particular pilin as the major structural subunit. Therefore, we propose pilA2 as the gene designation and PilA2 as the name for the mature protein encoded by CDR20291_3155.

Gene CDR20291_1084 is the fourth gene in a possible cluster of T4P-associated genes, and is predicted to encode a protein of almost the same size as PilA1. Given that it is one of two pilins with codons subject to positive selection (see below), we believe that it may also make distinct pili, and thus propose pilA3 and PilA3 as the gene designation and mature protein name, respectively. Gene CDR20291_1081 is located upstream of PilA3; it has less DNA variability than PilA3, and lacks sites under positive selection. Although it is the first of that particular gene set, we consider it to be a minor pilin. Following the Pseudomonas nomenclature (Alm et al., 1996), we have adopted pilX and PilX as its gene and protein designations. Separating PilA3 and PilX are two genes with single transmembrane domains; though they have low amino acid percent identity with PilMN and PilO in the largest C. difficile T4P gene cluster, it is tempting to speculate that these two genes encode proteins involved in T4P biogenesis.

Gene CDR20291_0683 is predicted to encode a large pilin. It is present in every strain examined, although it is not part of either the large or small T4P gene clusters. It also is one of the less diverse pilins across the 18 genomes (Supplemental Table 2). Structural analysis (Piepenbrink et al., submitted) indicates the protein encoded by CDR20291_0683 contains two globular C-terminal pilin domains, rather than the single domain typical of pilins in Gram-negative bacteria (Craig et al., 2004). Molecular modeling indicates that pili composed primarily of the product of this gene would be wider than any observed, indicating that pili composed entirely of this protein are unlikely. Given its size, low sequence diversity, and location outside of the gene cluster, it likely encodes a minor pilin. We propose pilJ as the gene designation and PilJ as the mature protein identifiers for gene CDR20291_0683, with ‘J’ indicating ‘Janus’ for the two C-terminal pilin domains. Precedent for this designation exists in the EPEC bundle-forming pilus, where one of the minor pilins is termed BfpJ, and in the T2S minor pseudopilin GspJ (Stone et al., 1996).

Genes CDR20291_3344 and CDR20291_3345 are located adjacent to one another in the T4P gene cluster, and their predicted products share a relatively high degree of amino acid sequence similarity with each other (Figure 1, Table 2). Their amino acid sequences have low variability across strains (Supplemental Table 2), leading us to believe they are minor pilins. Following the model of minor pilins in Pseudomonas (FimU) and Neisseria (PilV), (Alm and Mattick, 1996;Alm and Mattick, 1995), we propose pilU and pilV, and PilU and PilV as the gene and mature protein identifiers for CDR20291_3344 and CDR20291_3345, respectively.

Immediately downstream of pilV in the main T4P cluster lies the CDR20291_3343 open reading frame, predicted to encode a protein of 57.7 kDa with similarities to pilin-like proteins including a possible prepilin peptidase cleavage site followed by a stretch of hydrophobic residues. However, this unusually large potential prepilin-like protein lacks the conserved glutamic acid at position +5. We propose pilK for this gene and PilK for the mature protein as the GspK proteins of T2S systems are also large and lack the conserved glutamic acid (Korotkov and Hol, 2008).

Gene CDR20291_2191 is not located near any other T4P genes, and is present in only thirteen of eighteen strains studied; it is the only pilin not present in all eighteen strains. We consider it to be a minor pilin. Following the Pseudomonas nomenclature (Alm et al., 1996), we have termed the gene and mature protein pilW and PilW, respectively.

A recent review (Melville and Craig, 2013) identified these same nine pilins in C. difficile strain 630 that we list in R20291, through a combination of BLAST searches and PilFind results. Of these nine, the authors list PilA3 and PilX as “putative PilD dependent proteins,” identified by PilFind rather than by BLAST, because they are either not near other pilus-associated genes, or they bear little resemblance to other pilins. The authors do not discuss these two pilins further. Analysis of the proteins separating PilA3 and PilX indicates that they are possible T4P biogenesis proteins (Figure 1). Furthermore, the lack of amino acid similarity between the pilins is not necessarily a valid reason for discounting PilA3 and PilX as true pilins, given the large differences in amino acid sequences even among the pilins in the largest gene cluster.

Given our observations regarding the pilin amino acid sequence diversity across strains, we decided to investigate how the diversity in pilin amino acid sequences across different strains was reflected in their nucleotide sequences, and thus undertook an analysis of rates of synonymous and nonsynonymous mutations in the coding sequences for each pilin. For the majority of the pilins, rates of both synonymous (dS) and nonsynonymous (dN) substitutions were low, (Figure 3, dotted and dashed lines, and data not shown). Those nonsynonymous substitutions that did occur tended to cluster neat the C-termini of the proteins (Figure 3 and data not shown). As the N-termini are buried inside the pilus and play a critical role in pilin-pilin interactions (Craig et al., 2004), the relative paucity of substitutions there is consistent with the results of similar analyses conducted in other T4P pilin genes (Parge et al., 1995;Blank et al., 2000). PilA1 and, surprisingly, PilW have the greatest number of sites with synonymous or nonsynonymous substitutions, while other pilins, particularly PilU, PilV, and PilX, have relatively few (Figure 3, data not shown).

Figure 3.

The ratio of nonsynonymous substitutions per site to synonymous substitutions per site, called dN/dS or ω, is a standard measure of selective pressure. Values of ω for the full-length nucleotide sequences were generally close to 0 (Table 3), congruent with the theory that most sites in a protein are subject to purifying selection (with a resulting low ω) due to functional constraints (Yang and Bielawski, 2000).

Table 3.

Codons under positive selection in pilin genes

Pilin	ω	Positive Likelihood Ratio Test at p < 0.05	Residues under positive selection (R20291 sequence)
PilA1	0.328	Yes	146 N, 147P, 150 S
PilA2	0.060	No	--
PilA3	0.235	Yes	96 S
PilJ	0.120	No	--
PilK	0.171	No	--
PilU	0.179	No	--
PilV	0.137	No	--
PilW	0.148	No	--
PilX	0.564	No	--

^aBold text indicates p > 99%, plain text indicates p > 95%

It is unlikely that entire genes will be subject to positive selection; rather, individual codons undergo selection (Yang et al., 2005). These sites of positive selection can be identified by statistical methods; the likelihood-ratio test (LRT) tests for the presence of positively selected codons in a protein. If the LRT is positive, a Bayes empirical Bayes (BEB) approach can determine which sites are under positive selection (Yang et al., 2005). Of the nine pilins, only pilA1 and pilA3 have a positive LRT, indicating each had codons under positive selection. The BEB analysis identified three pilA1 codons under positive selection, all of which were clustered in the C-terminal region of the pilin (listed in Table 3 and indicated by arrows in Figure 3A); this region is likely to be exposed to the environment rather than buried in the body of the pilus (Craig et al., 2004). pilA3 has one codon predicted to be under positive selection, also in the C-terminal region of the protein (listed in Table 3 and indicated by an arrow in Figure 3B).

Pilin purification, immunogenicity and cross-reactivity in animals

Artificial genes with codons optimized for E. coli expression and lacking the signal sequence and conserved N-terminal hydrophobic domain were synthesized for seven pilin and pilin-like proteins. Six of seven proteins were soluble and successfully purified (Figure 4); only PilX proved insoluble and difficult to purify.

Figure 4.

To test the immunogenicity of the six purified pilin and pilin-like proteins, mice were immunized intradermally with purified protein and adjuvant and boosted subcutaneously at 1 and 2 weeks after the initial injection. Terminal bleeds were collected 59 days after the initial injection. Preimmune and terminal sera from each mouse were tested by ELISA for antibody response against the homologous protein. Although all pilin and pilin-like proteins induced a detectable antibody response, those responses varied by pilin. Immunization with PilW, PilU, or PilV led to a higher titer homologous antibody response than immunization with any of the other three proteins, while PilA1 generated the weakest responses of any pilin, and was the only pilin to which an immunized mouse produced no detectable antibody (Figure 5).

Figure 5.

After determining the titer of antibodies against the immunizing pilin or pilin-like protein, we next measured the antibody titer against the five heterologous proteins (Figure 5). Notably, the highly conserved N-terminus of each protein was excluded from the purified soluble protein antigens, and thus all cross-reactive responses are due to epitopes present in less conserved regions of the proteins. Antibodies raised against PilJ were almost completely specific to their immunizing antigen. In contrast, antibodies raised against PilW reacted to all six purified pilins; antibody titers against each individual pilin varied among the five mice. Antibodies raised against PilU reacted strongly to both PilU and PilV, and vice versa, suggesting strong epitope conservation between these proteins. Interestingly, antibodies raised against PilA2 proved more reactive to PilJ than to PilA2.

To produce greater volumes of sera for future experiments, we immunized additional species of small mammals with selected pilins. Our results with these species mirror our experiences in mice. As in mice, immunization with PilJ produced a specific and robust response in rabbits (Figure 6A). In contrast, guinea pigs immunized with PilA1 required several boosts to generate significant responses; these antibodies also cross-reacted with PilA2, and immunoadsorption was necessary to select for antibodies specific for PilA1 (Figure 6B). This experience is reminiscent of the poor immunogenicity of PilA1 in mice. Of three rats immunized with PilA2, two responded only to PilJ on Western blot; the third responded specifically to PilA2 (Figure 6C). The ability to generate antibodies crossreactive with PilJ was also a feature of PilA2 immunization in mice.

Figure 6.

DISCUSSION

Insights into C. difficile T4P composition

Our findings confirm that the C. difficile genome encodes pilins and pilin-like proteins and that these proteins are immunogenic in mice. Previous research had identified three genes encoding putative pilin or pilin-like proteins in the C. difficile 630 genome (Varga et al., 2006). This list was expanded to seven pilins and two putative pilins in a recently published review (Melville and Craig, 2013). We confirmed the presence of all nine genes in the R20291 genome, and present evidence that the two putative pilins are T4P pilin genes. Of these nine, eight are present in all 18 publically available C. difficile genomes; pilW is found in the genomes of all available ribotype 027 and PFGE NAP1 strains, but in only three of six strains that do not belong to these groups (Supplemental Table 2, p=.044, Fisher’s exact test). Although we have identified nine pilins in the R20291 genome, we remain open to the identification of more pilins and pilin-like proteins; indeed, it was only during the preparation of this manuscript that we located PilK and PilA3.

Recent immunogold experiments done by our collaborators, using the guinea pig anti-PilA1 antibody and the rabbit anti-PilJ antibody described herein, demonstrated PilA1 and PilJ incorporated into pili on the bacterial surface (Piepenbrink et al., under revision). Preliminary results of flow cytometry and fluorescent microscopy from our lab also suggest the presence of PilJ on the cell surface. The crossreactivity of the mouse anti-PilU, anti-PilV, and anti-PilW antibodies, and the low titer of the rat anti-PilA2 antibodies, precluded their use in immunogold or other applications; we have not yet verified the presence of PilA2, PilU, PilV, or PilW on the C. difficile surface.

Previous structural studies of pili indicate that the C-terminal globular domains of the major pilin are exposed in the pilus structure, while the N-terminal hydrophobic alpha-helices are buried in the main body of the pilus (Craig et al., 2003). For ease of purification, we removed the sequences encoding N-terminal hydrophobic regions from the synthetic genes, and expressed the proteins without those regions. We believe that we are justified in examining immunogenicity of such truncated pilins and pilin-like proteins, as the excluded regions are not accessible to the host immune system in intact pili and would not be expected to contribute to protective immunity. Thus, any observed crossreactivity between pilins is due to conserved epitopes in the divergent C-terminal regions, rather than to the more highly conserved N-terminus.

Pilus-expressing Gram-negative bacteria contain multiple genes for minor pilins and pilin-like proteins that play different roles in biogenesis, in addition to one or more genes for the pilus major structural subunit. Given the example of the Gram-negative T4Ps, it is likely that several of the C. difficile pilin genes encode pilin-like proteins or minor pilins.

Based on our results, we speculate that pilA1 encodes the major pilin of a T4P. Not only is it located at the beginning of the main T4P gene cluster (the only cluster encoding prepilin peptidases) and present in all analyzed genomes, it is the most genetically diverse of the pilin genes across the 18 genomes analyzed, and one of only two pilin genes with codons subject to positive selection. The diversity of the genes and the evidence for selective pressure exerted on codons suggest diversification of exposed epitopes in response to host immune pressure or other factors. The most likely target of the host immune response is the major pilin, as it is the most common pilin present and therefore subject to a selective pressure not experienced by any other pilin. That PilA1 is a relatively weak immunogen, may present a selective advantage for the bacteria during infection. Our results do not exclude the possibility of multiple major pilins, either producing unique pili or together forming heterogeneous pili.

Given its location in an apparent operon with several additional T4P biogenesis genes, pilA2 may also encode a major pilin protein, albeit one that is unusually small. Its high degree of sequence conservation suggests that PilA2 may not be expressed to the same degree as or may play a different role than PilA1. Likewise, pilA3 may encode a major pilin protein, given its location downstream of three putative T4P genes and the presence of an amino acid under positive selection in its C-terminal region. Of note, since the predicted size of PilA1 is consistent with other major pilins, and given its amino acid sequence diversity, evidence for diversifying selection, and localization in a complete operon, we have elected to switch the nomenclature for pilA1 and pilA2 relative to the original designations of homologous genes in C. perfringens (Varga et al., 2006).

PilK was named for its similarities to GspK proteins, namely its size and, more importantly, a hydrophobic residue in place of a Glu+5. Craig et al. hypothesize that, in an assembled pilus, the Glu+5 of one pilin forms a salt bridge with positively charged N-terminal Phe1 of the pilin previously extracted from the membrane (Craig et al., 2006). If this salt bridge is essential, the lack of Glu+5 on a minor pilin would preclude the addition of that pilin into the main body of a pilus. It could, however, be the first pilin added to a nascent pilus and thus form the pilus tip. Indeed, GspK of the ETEC T2SS is predicted to form the tip of its pseudopilus (Korotkov and Hol, 2008).

The strong cross-reactivity between antibodies raised against PilU and PilV, combined with the (relative) sequence conservation and location of pilU and pilV next to each other in the gene cluster suggest that they are the result of a gene duplication and thus may encode proteins with similar structures and roles in pilus biogenesis. Their localization downstream of pilA1 further suggests that they encode minor pilin or pilin-like proteins. The PilU protein was previously assumed to encode the major pilin subunit and antibodies raised against this protein (referred to only as PilA) were found to decorate pili connecting bacteria to cells in hamsters infected with C. difficile (Goulding et al., 2009). Details regarding the sequence of the antigen used to raise the serum were not provided. Given our results demonstrating extensive cross-reactivity among pilins and pilin-like proteins lacking their conserved N-terminal domains, it is not clear which protein(s) are recognized by these antibodies. Considering its relative conservation and position far from the start of the gene cluster, we suggest that pilU more likely encodes a minor pilin protein rather than the major pilin.

Of the seven pilins expressed in E. coli, the only one that we could not easily purify for study was PilX. Although we could express the protein in E. coli, it proved insoluble even when lacking the N-terminal hydrophobic region. That PilX was the only insoluble protein of the seven leads us to believe that it may have a fundamentally different role than the other six soluble pilins.

Vaccine potential of C. difficile pilins

Determining which pilin and pilin-like proteins are immunogenic, and to what degree, is a critical step in identifying optimal components of an experimental vaccine directed against C. difficile T4Ps. T4Ps are composed of thousands of repeating monomers, are extracellular and easily accessible to the host immune system; moreover, they are often important for initial colonization and biofilm formation (Tacket et al., 1998;O’Toole and Kolter, 1998) Thus, they present excellent targets for vaccines to prevent pathogen colonization. Neutralizing antibodies directed against pili could lead to immune clearance of C. difficile before they establish themselves in the colon. Prior pilin-based vaccines have met with some success: a multivalent M. bovis whole-pilin vaccine was protective against heterologous strain challenge (Lepper et al., 1995) and is now commercially available (Piliguard^® Pinkeye TriView, Merck Animal Health), while immunization with V. cholerae TcpA generates some protective immunity in a V. cholerae mouse model (Sun et al., 1989;Voss et al., 1996). A vaccine directed against a single N. gonorrhoeae pilus subtype and tested in humans was successful in generating antibodies against that specific subtype; however, the vaccine granted no protection against N. gonorrhoeae infection (Boslego et al., 1991). The sequence diversity of the gonococcal pilin, granted by its ability to recombine pilE with pilS cassettes, seems to preclude the generation of pan-reactive anti-pilus antibodies and the development of an effective anti-gonococcal T4P vaccine.

The most obvious component of a C. difficile pilus vaccine would be the major pilin, which we currently believe to be PilA1. The direct anti-PilA1 antibody response is weak, which may invalidate it as the sole component of a vaccine. Instead, PilW could be one possible component of a multivalent T4P vaccine, as immunization with that protein generates strong responses to many of the other pilin and pilin-like proteins. This cross-reactivity suggests that it could confer protection against pilins from non-R20291 C. difficile strains. Furthermore, immunization with PilW generated stronger antibody responses to PilA1 than immunization with PilA1 itself; if PilA1 is indeed the main structural subunit, such crossreacting antibodies could circumvent the poor PilA1 immunogenicity. PilJ, which is expressed extracellularly (Piepenbrink et al., submitted), present in every strain of C. difficile examined, and induces a strong antibody response after immunization, is another strong candidate for inclusion in a vaccine.

Given that C. difficile is a colonic pathogen, mucosal immunity may provide optimal protection. Both the Sabin polio and rotavirus vaccines take advantage of oral administration, and while C. difficile is a bacterium rather than a virus, the same principle applies. Ideally, C. difficile vaccine would combine a toxoid, to prevent toxin-induced pathology (Kotloff et al., 2001;Wang et al., 2012), with pilin or pilin-like proteins to prevent colonization, pathology, and transmission.

Future work will seek to confirm the major pilus subunit, the roles of the other pilins and pilin-like proteins in pilus function, and to test the protection afforded by a pilin-based vaccine against C. difficile infection in an animal model. This approach, if successful, could lead to clinical studies.