Clostridium difficile Lipoprotein GerS Is Required for Cortex Modification and Thus Spore Germination

The Gram-positive, spore-forming bacterium Clostridium difficile is a leading cause of antibiotic-associated diarrhea. Because C. difficile is an obligate anaerobe, its aerotolerant spores are essential for transmitting disease, and their germination into toxin-producing cells is necessary for causing disease. Spore germination requires the removal of the cortex, a thick layer of modified peptidoglycan that maintains spore dormancy. Cortex degradation is mediated by the SleC hydrolase, which is thought to recognize cortex-specific modifications. Cortex degradation also requires the GerS lipoprotein for unknown reasons. In our study, we tested whether GerS is required to generate cortex-specific modifications by comparing the cortex composition of ΔgerS spores to the cortex composition of spores lacking two putative cortex-modifying enzymes, CwlD and PdaA. These analyses revealed that GerS, CwlD, and PdaA are all required to generate cortex-specific modifications. Since loss of these modifications in ΔgerS, ΔcwlD, and ΔpdaA mutants resulted in spore germination and heat resistance defects, the SleC cortex lytic enzyme depends on cortex-specific modifications to efficiently degrade this protective layer. Our results further indicate that GerS and CwlD are mutually required for removing peptide chains from spore peptidoglycan and revealed a novel interaction between these proteins. Thus, our findings provide new mechanistic insight into C. difficile spore germination.

the Peptostreptococcaceae family, whereas the members of the Clostridiaceae and Lachnospiraceae families encode catalytically competent CspC and CspA proteases (19). While it is unclear whether CspC pseudoproteases function as germinant receptors in other Peptostreptococcaceae family members, the catalytic mutations in CspC are strictly conserved (19). We recently identified an additional Peptostreptococcaceae family-specific protein required for C. difficile spore germination: the GerS lipoprotein (20). Using Targetron-based gene disruption of gerS, we showed that loss of C. difficile GerS leads to a Ͼ4-log reduction in spore germination in the JIR8094 strain background (27,28) because the cortex is not degraded (20). Interestingly, although C. difficile gerS mutant spores fail to hydrolyze their cortex, the pro-SleC zymogen is still proteolytically processed in response to germinants (20).
These observations suggest that C. difficile GerS is required for spore germination because it (i) modulates SleC cortex lytic enzyme activity downstream of CspB-mediated proteolytic cleavage or (ii) is necessary to generate the cortexspecific peptidoglycan modifications that are presumably needed by SleC to recognize and degrade the cortex layer. Previous studies in Bacillus subtilis have shown that muramic-␦-lactam (MAL) residues distinguish spore cortex peptidoglycan from the germ cell wall (29,30). Approximately half the N-acetylmuramic acid (NAM) residues in B. subtilis cortex peptidoglycan are converted to MAL residues, while~25% carry tetrapeptide side chains and the remainder have single D-Ala side chains due to the action of the LytH hydrolase (31).
The MAL modification is specifically recognized by cortex lytic enzymes, ensuring that these lytic enzymes do not degrade the vegetative germ cell wall of spores (29,32,33). In B. subtilis, two enzymes act sequentially on spore peptidoglycan to generate MAL residues: CwlD, an N-acetylmuramoyl-L-alanine amidase, and PdaA, an N-acetylmuramic acid deacetylase (34,35) (Fig. 1A). CwlD removes the tetrapeptide side chains attached to NAM residues, while PdaA deacetylates the NAM residues and performs lactam cyclization to produce MAL (34)(35)(36)(37). Loss of either CwlD or PdaA in B. subtilis leads to severe defects in cortex degradation and thus to spore germination on media containing germinants but not germination events upstream of cortex degradation (34,35,37).
In this study, we sought to test whether GerS is necessary for generating the muramic-␦-lactam (MAL) cortex-specific modifications that are presumably recognized by SleC. In order to test this hypothesis, we compared the muropeptide profiles of spores lacking either of the putative cortex-modifying enzymes CwlD and PdaA to spores lacking GerS. Since C. difficile CwlD and PdaA share 57% and 54% sequence similarity to their B. subtilis homologs (see Fig. S1 in the supplemental material), respectively, they likely have functions similar to those of their B. subtilis counterparts. The germination phenotypes and peptidoglycan compositions of gerS, cwlD, and pdaA mutant spores were compared to provide the first insights (to our knowledge) into the mechanisms controlling cortex modification in the Clostridia and their impact on spore physiology.

RESULTS
Characterization of ⌬gerS, ⌬cwlD, and ⌬pdaA mutants. To construct mutations in the gerS, cwlD, and pdaA genes, we used pyrE-based allele-coupled exchange (ACE) to make clean deletions (see Fig. S2 in the supplemental material) (38). Unlike the Targetronbased gene disruption method that we previously used to construct a gerS::ermB mutant in the JIR8094 strain background, the clean deletions prevented polar effects on downstream gene expression. Although mutation of the alanine racemase (alr2) gene downstream of gerS does not strongly alter germination (20,39), constructing a gerS in-frame deletion avoids effects on downstream gene expression. Another advantage of the pyrE-based ACE system is that it allows single-copy, chromosomal complementation of the gerS (C. difficile 630_34640 [CD630_34640]), cwlD (CD630_01060), and pdaA (CD630_14300) mutations and avoids introducing experimental artifacts associated with the overexpression of plasmid-based complementation constructs (19).
To characterize these mutant strains, we determined their apparent sporulation efficiency levels by measuring CFU levels on media containing germinants following heat treatment of sporulating cultures. Specifically, ΔgerS, ΔcwlD, and ΔpdaA strains were induced to sporulate, and the resulting samples were either heated to kill any vegetative cells or left untreated and then plated on media containing taurocholate (40). The ΔgerS, ΔcwlD and ΔpdaA mutants all exhibited an~5-log decrease in apparent sporulation efficiency relative to the wild type (P Ͻ 0.02, Fig. 1B). In fact, no CFU were detected in three of the four biological replicates performed, while CFU were detected in the three mutant strains at a frequency of~0.0005% in the fourth biological replicate, which is close to the limit of detection of this assay. The apparent sporulation efficiency of the 630Δerm ΔgerS strain was roughly equivalent to that previously reported for the JIR8094 gerS::ermB Targetron mutant (20), while the apparent sporulation efficiency of ΔcwlD was similar to that of the previously constructed 630Δerm ΔcwlD strain (41). ΔcwlD, and ΔpdaA strains and their complements measured for four biological replicates by heat-treating sporulation cultures and determining CFU on media containing germinant cultures relative to untreated cultures. The Δspo0A mutant served as a negative control because it cannot form spores (11). Averages of results from four biological replicates are shown along with the associated standard deviations. (C) Spore purification yields. Sporulating cells were harvested from 70:30 agar plates 48 h following plating. The optical density at 600 nm for all purified spore stocks resuspended in identical 600-l volumes was determined divided by the total number of 70:30 plates used per strain. Averages and standard deviations of results from three independent experiments are shown. (D) Germination efficiency of ΔgerS, ΔcwlD, and ΔpdaA spores and their complements. Averages of results from three biological replicates performed using two independent spore preparations are shown along with the associated standard deviations. The ΔsleC mutant served as a negative control because it is defective in cortex hydrolysis (21). Statistical analyses were performed using one-way ANOVA and Tukey's test. *, P Ͻ 0.05; **, P Ͻ 0.005; ***, P Ͻ 0.0005; ****, P Ͻ 0.0001; n.s., not statistically significant. Importantly, complementing the ΔgerS, ΔcwlD, and ΔpdaA strains with wild-type alleles of gerS, cwlD and pdaA, respectively, from the pyrE locus restored the apparent sporulation efficiency of the mutants to wild-type levels (Fig. 1B).
Since the apparent sporulation efficiency measured using the heat-treatment assay does not distinguish between defects in heat resistance, spore formation, germination, and/or outgrowth, we tested whether ΔgerS, ΔcwlD, and ΔpdaA spores would be purified as efficiently as wild-type spores. The spore purification yield for each strain was measured by inducing sporulation, purifying spores using density gradient centrifugation, resuspending spores in equivalent volumes, and determining their yields by measuring optical density at 600 nm (OD 600 ). No significant differences in purified spore yields were observed for the ΔgerS, ΔcwlD, and ΔpdaA strains relative to the wild-type strain or the germination-defective ΔsleC mutant (42) (Fig. 1C).
When ΔgerS, ΔcwlD, and ΔpdaA spores were plated on media containing germinant, an~2-to-3-log decrease in spore germination efficiency was observed in the mutants relative to the wild-type strain (P value Ͻ 0.0001, Fig. 1D). ΔpdaA spores exhibited a 5-fold-greater germination defect than the ΔgerS and ΔcwlD spores, but this difference was not statistically significant. The ΔsleC mutant, which lacks the primary cortex lytic SleC enzyme, exhibited a 3-log decrease in germination efficiency (P Ͻ 0.0001, Fig. 1D) as previously reported (25,42). Notably, ΔgerS, ΔcwlD, and ΔpdaA spores all exhibited delayed germination phenotypes, producing colonies from germinating spores more slowly than wild-type colonies and with kinetics similar to ΔsleC spore kinetics (25). In addition, as previously observed with C. difficile ΔsleC and ΔcspBAC mutants (25,42) and B. subtilis germination receptor mutants (43), the germination efficiencies for ΔgerS, ΔcwlD, and ΔpdaA spores differed between spore preparations for unknown reasons, so all analyses were performed on two independent biological replicate preparations of purified spores. While the 2-log germination defect of 630Δerm ΔgerS spores was considerably less severe than the 4-log germination defect reported for our JIR8094 gerS::ermB mutant (20), this discrepancy is consistent with the~2-to-3-log-less-severe germination defects observed for sleC and cspBAC Targetron mutant spores in the 630Δerm strain relative to the JIR8094 strain background (25). Although the reason for this difference is unclear, these results indicate that GerS, CwlD, and PdaA are critical for spore germination. Notably, the germination defects of ΔgerS, ΔcwlD, and ΔpdaA spores (~2 to 3 logs; Fig. 1D) were markedly less severe than their~5-log defect in apparent sporulation efficiency (Fig. 1B), suggesting that GerS, CwlD, and PdaA contribute to spore heat resistance.
To test this possibility, we exposed purified ΔgerS, ΔcwlD, and ΔpdaA spores to 60°C heat treatment, similarly to the sporulating cell analyses, and measured the capacity of heat-treated spores to germinate and form colonies on media containing germinant. While the germinated CFU levels for wild-type and ΔsleC spores were unchanged by the 60°C treatment, ΔgerS, ΔcwlD, and ΔpdaA germinated CFU levels decreased by~10-fold (P Յ 0.05, Fig. S3A), which was similar to the approximately Ϫ5-fold decrease previously reported for the gerS::ermB mutant in JIR8094 (20). The apparent increase in the heat sensitivity of ΔgerS, ΔcwlD, and ΔpdaA mutant spores relative to wild-type spores was not due to decreased dipicolinic acid (DPA) levels (44), since measurement of spore DPA levels revealed that ΔgerS and ΔcwlD spores had slightly higher levels of DPA than wild-type spores (P Ͻ 0.05, Fig. S3B).
Abnormal forespore morphology in cwlD and pdaA mutants. Interestingly, while analyzing sporulation in these mutants by phase-contrast microscopy, we observed that the ΔcwlD and ΔpdaA mutants produced forespores with an abnormal, circular morphology (as opposed to an ovoid morphology) more frequently than the wild-type strain and the ΔgerS mutant (Fig. S4). To visualize these morphological defects with greater resolution, we analyzed sporulating cells of the mutants and their complements using transmission electron microscopy (TEM). A minimum of 50 sporulating cells of each mutant was examined 23 h after sporulation induction. Circular forespores were observed in 39% and 13% of ΔcwlD and ΔpdaA sporulating cells, respectively ( Fig. 2A), which represent 10-fold and 3-fold increases over wild-type cells consistent with our phase-contrast microscopy analyses (Fig. S4). The apparent increase in the frequency of circular forespores indicated by TEM relative to phase-contrast microscopy was likely a function of the higher degree of confidence in scoring cells as circular forespores. Importantly, complementation of the ΔcwlD and ΔpdaA mutants reduced the frequency of circular forespores by~3-fold. Taking the results together, the microscopy analyses indicate that the putative cortex-modifying enzymes, CwlD and PdaA, help generate and/or maintain wild-type spore morphology.
Interestingly,~50% of the circular forespores observed in the ΔcwlD mutant also exhibited coat tethering defects where the coat appeared to slough off the forespore (also known as "bearding" [45]). Coat detachment was observed in only 1 of 33 ΔcwlD forespores with ovoid morphology relative to 10 of 21 circular ΔcwlD forespores. This bearding phenomenon was elevated in ΔcwlD (20%) and ΔpdaA (10%) sporulating cells but not ΔgerS sporulating cells (5%) relative to the wild type (4%) (Fig. 2B). While it is unclear why coat localization defects were observed in the ΔcwlD and ΔpdaA strains, the observation that bearding correlates with circular forespore formation in the ΔcwlD mutant could imply that geometric cues play a role in tethering the coat to the forespore. In analyses of purified spores by TEM, the ΔgerS, ΔcwlD, and ΔpdaA spores appeared qualitatively similar to wild-type spores. The spore cortex of the mutant strains, however, appeared darker (more electron dense) than the wild type (Fig. S5). This darkened cortex phenotype for the ΔcwlD spores relative to the wild type has previously been observed by TEM in B. subtilis (33), so the more electron-dense cortex of the C. difficile mutant spores may indicate that GerS, CwlD, and PdaA are required to modify the cortex.
GerS, CwlD, and PdaA are necessary for cortex degradation. Given that B. subtilis cwlD and pdaA mutants are defective in spore germination because their cortex lytic enzymes, CwlJ and SleB, cannot recognize their substrate MAL (33), we tested whether cortex degradation was impaired in C. difficile mutants lacking gerS, cwlD, and pdaA using an optical-density-based assay. Cortex degradation can be indirectly assessed by measuring the change in optical density at 600 nm (OD 600 ) of spores as they germinate, hydrate, and thus decrease in optical density. Previous work has shown that cortex degradation constitutes the initial 20% drop in OD 600 observed in wild-type germinating spores and that the subsequent 20% drop can be attributed to core rehydration (44). The optical density of wild-type spores rapidly decreased in the initial 45 min after taurocholate addition and then remained largely stable after 3 h, while the optical density of ΔsleC spores remained unchanged over this time period because SleC is necessary for cortex degradation (21). The optical density of ΔcwlD spores did not significantly decrease over 3 h, while ΔgerS and ΔpdaA spores exhibited slight but significant drops at 2.5 and 1 h, respectively, following taurocholate treatment (Fig. 3A). These observations suggest that ΔgerS, ΔcwlD, and ΔpdaA mutant strains are defective in cortex degradation to differing degrees. While the inability to detect a delayed germination phenotype for ΔcwlD spores may reflect the low sensitivity of the OD 600 germination drop assay, the delayed germination phenotype of ΔgerS and ΔpdaA spores in the optical density assay is consistent with the qualitatively delayed germination observed in plate-based germination assays (data not shown).
To assess whether the cortex degradation defects in ΔgerS, ΔcwlD, and ΔpdaA spores were due to an inability to process the pro-SleC zymogen into its catalytically active form, we analyzed SleC cleavage following treatment with taurocholate. As expected, the predominant form of SleC observed in untreated samples was the pro-SleC zymogen for all strains. Exposure of wild-type, ΔgerS, ΔcwlD, and ΔpdaA spores to taurocholate resulted in similar levels of SleC processing (Fig. 3B). These results confirm our previous analyses of the gerS::ermB strain in the JIR8094 strain background (20), but they contrast with recent studies of a 630Δerm ⌬gerS mutant (46) where SleC cleavage was shown to decrease in the absence of GerS. However, the latter studies used different experimental conditions to assess SleC cleavage, and no complementation analyses of the ΔgerS strain were performed. Regardless, our observations strongly suggest that ΔgerS, ΔcwlD, and ΔpdaA spores are defective in spore germination because they inefficiently hydrolyze the cortex despite proteolytically activating SleC.
GerS, CwlD, and PdaA are required to generate the cortex-specific modification muramic-␦-lactam. The results thus far were consistent with the hypothesis that ΔgerS, ΔcwlD, and ΔpdaA mutants cannot degrade the cortex because they fail to generate the cortex-specific MAL modifications required for SleC to recognize its cortex substrate. This hypothesis is based on the observation that B. subtilis cortex lytic enzymes require MAL residues for cortex degradation to proceed (33) and that B. subtilis CwlD and PdaA are sequentially required to generate MAL residues (34,35). To directly test this hypothesis, we analyzed the muropeptide profiles of our mutant strains. Cortex peptidoglycan was isolated from wild-type, ΔgerS, ΔcwlD, and ΔpdaA spores and digested with muramidase, which cleaves between NAM and N-acetylglucosamine (NAG) residues but not between MAL and NAG residues (30,47). The digestion products were separated using reverse-phase high-performance liquid chromatography (HPLC), and the identities of the muropeptide compounds within individual peaks were determined using mass spectrometry (MS) ( Fig. 4; see also Fig. S6 and Table S1 in the supplemental material for details regarding peak assignments).
Several peaks were identified in the peptidoglycan isolated from all four mutant spores: these peaks corresponded to the disaccharide (DS, NAG-NAM) carrying the L-Ala-D-Glu-m-Dpm-D-Ala tetrapeptide (TP, with Љm-DpmЉ referring to meso-2,6-diaminopimelic acid) (peak 2) and to two DS-TP (disaccharide tetrapeptides) cross-linked together (peak 9). A large fraction of the DS-TP was deacetylated on the NAG residue (GlucN, peak 3), consistent with the observation that large portions of NAG (N-acetylglucosamine) residues are deacetylated in vegetative C. difficile peptidoglycan (48).
Notably, a single peak (peak 12) was unique to the peptidoglycan isolated from wild-type spores. Peak 12 corresponds to a tetrasaccharide containing MAL (TS MAL ), the distinguishing feature of cortex peptidoglycan (33). Approximately one-third of peptidoglycan fragments from the wild type contained this cortex-specific modification (Fig. 4). Since this peak was absent from ΔgerS, ΔcwlD, and ΔpdaA spores, these results indicate that GerS, CwlD, and PdaA are all required for MAL synthesis and thus presumably for SleC to recognize its cortex substrate.
Remarkably, the muropeptide profile of ΔcwlD spores was identical to that of ΔgerS spores in peak presence and intensity. Peaks 2 and 3 increased markedly in ΔgerS and ΔcwlD spores relative to wild-type and ΔpdaA spores. These peaks contain disaccharide tetrapeptides (DS-TP) with and without acetylation, respectively, the major muropeptide expected if the putative amidase activity of C. difficile CwlD prevents removal of peptide side chains as previously observed for B. subtilis ΔcwlD spores (33,49). The relative increases in peaks 2 and 3 resulted from the lack of MAL in ΔgerS and ΔcwlD peptidoglycan, which allows mutanolysin to produce more DS species. Increases in peaks 4 and 5 may result from the dramatically increased abundance of peaks 2 and 3, providing more substrate for an unidentified endopeptidase that cleaves the peptide side chains. Regardless, since the muropeptide profiles of ΔgerS and ΔcwlD spores were essentially indistinguishable, these results strongly suggest that GerS is required for CwlD amidase activity to generate the substrate required for SleC recognition. Alternatively, GerS may directly modify the cortex in a CwlD-dependent manner.
Two muropeptides were unique to the ΔpdaA strain. Peak 1 consists of a tetrasaccharide (TS) that lacks any peptide side chains and contains no MAL. This species corresponds to the predicted product of CwlD-mediated removal of the tetrapeptide and is consistent with analyses of B. subtilis CwlD produced in Escherichia coli (35). Peak 8 contained the second muropeptide exclusively produced in the ΔpdaA mutant, a tetrasaccharide tetrapeptide variant that differs from peaks 6 and 7 by the position and extent of NAG deacetylation. Notably, ΔpdaA spores did not accumulate the large amounts of the DS-TP variants (peaks 2 and 3) observed in ΔgerS and ΔcwlD spores, presumably because CwlD (and/or GerS) removes some of the peptide side chains. Taking the data together, the muropeptide profile analyses indicate that GerS and CwlD activities are coupled and mediate tetrapeptide removal, whereas PdaA likely acts downstream of GerS and CwlD in modifying the peptide side chain to generate muramic-␦-lactam residues. C. difficile spore cortex peptidoglycan was purified and digested using mutanolysin. The resulting muropeptides were reduced, separated by reversed-phase HPLC, and identified by mass spectrometry. Peaks are numbered as indicated in Table S1; the peaks labeled ЉBЉ represent buffer components. ЉNAGЉ represents N-acetylglucosamine, ЉNAMЉ represents N-acetylmuramic acid, ЉMALЉ represents muramic-␦-lactam, and ЉNAMohЉ represents reduced N-acetylmuramic acid.
CwlD is critical for GerS signal peptide processing and incorporation into mature spores. Since the gerS and cwlD mutants produced identical spore muropeptide profiles and shared similar levels of spore heat sensitivity and germination defects, the putative amidase activity of CwlD appears to depend on GerS. To gain insight into the relationship between GerS and CwlD, we tested whether GerS might affect CwlD levels in sporulating cells and/or purified spores and vice versa. Our previous plasmidbased complementation analyses of a gerS Targetron mutant revealed that GerS follows a typical Gram-positive lipoprotein biogenesis pathway (50) in which GerS is translocated across the outer forespore membrane by its N-terminal signal peptide, becomes lipidated on a conserved cysteine (Cys22), and then is processed to remove its signal peptide, leaving the protein tethered to the membrane by virtue of its N-terminal lipidation (20,51). Accordingly, both full-length GerS and cleaved GerS were detectable in wild-type sporulating cells, while only cleaved GerS was detected in mature spores (20). Both full-length GerS and cleaved GerS were detected in sporulating ΔcwlD and ΔpdaA cells, but the proportion of full-length GerS was dramatically higher in ΔcwlD sporulating cells than in ΔpdaA and wild-type cells (Fig. 5A). Furthermore, when ΔcwlD spores were purified, the levels of GerS were markedly reduced, whereas wild-type levels of GerS were present in ΔpdaA spores (Fig. 5B). Notably, complementation of cwlD restored signal peptide processing to GerS in sporulating cells and increased GerS levels in mature spores (Fig. 5B). All together, these results indicate that CwlD is critical for normal processing of GerS, leading to defects in GerS incorporation into or stability in mature spores or both. In contrast with GerS, CwlD levels in sporulating cells were unchanged among the different mutant strains (Fig. 5B). However, Western blot analyses revealed that CwlD levels in mature spores appeared to be slightly (~2-fold) reduced in ΔgerS spores relative to wild-type spores, although this difference was not statistically significant (data not shown). CwlD also underwent cleavage in mature spores, with the cleaved variant of CwlD being more abundant than full-length CwlD in wild-type, ΔgerS, and ΔpdaA strains.
PdaA levels were unchanged in both sporulating cells and mature spores of wildtype and the mutant strains; both full-length and cleaved forms of PdaA were detected at equivalent levels in sporulating cells and mature spores in all the strains examined. Interestingly, although the B. subtilis homologs of PdaA and CwlD have predicted signal peptides, only PdaA has a putative signal peptide cleavage site, A 19 X 20 A 21 (34,37). In contrast, both C. difficile CwlD and PdaA lack predicted signal peptide sequences, although they both are predicted to contain transmembrane domains based on signal peptide 4.1 (52) and TMHMM v. 2.0 analyses. Since both full-length and cleaved CwlD and PdaA were detected by Western blotting, both proteins undergo some form of posttranslational processing, with processed CwlD being the predominant form detected in mature spores.
GerS and CwlD physically interact. Since GerS processing in sporulating cells and GerS levels in spores were reduced in the absence of CwlD, we wondered whether CwlD might affect GerS processing through a direct interaction. To this end, we tested GerS binding to CwlD using co-affinity purification in an E. coli-based expression system. Purification of soluble CwlD and GerS from E. coli required deleting their transmembrane domain, so we deleted the region encoding CwlD's putative transmembrane sequence and GerS's signal peptide. C-terminally His-tagged CwlD lacking its N-terminal 25 residues (CwlD Δ25 -His 6 ) was coproduced with untagged GerS lacking its N-terminal 26 residues (GerS Δ22 ). As a specificity control, we also coproduced CwlD Δ25 -His 6 with untagged CspBA, which is also made in the mother cell during sporulation but was not expected to interact with CwlD. Notably, when CwlD Δ25 -His 6 was coproduced with untagged GerS in E. coli, untagged GerS was copurified with CwlD Δ25 -His 6 on Ni 2ϩ -affinity resin at stoichiometric levels (Fig. 6). In contrast, untagged CspBA did not copurify with CwlD Δ25 -His 6 . Importantly, untagged GerS Δ22 exhibited little nonspecific binding to the Ni 2ϩ -affinity resin (Fig. 6). These results indicate that the soluble domains of recombinant GerS and CwlD directly interact with high affinity. They further suggest that GerS and CwlD directly interact in C. difficile sporulating cells, which may be related to GerS's dependence on CwlD to undergo signal peptide processing.

DISCUSSION
Although we previously showed that C. difficile spore germination depends on the GerS lipoprotein (20), the mechanism underlying this requirement remained unknown. In this study, we determined that GerS is necessary for CwlD-mediated cortex modification in C. difficile (Fig. 4) and that these cortex modifications are required for the SleC hydrolase to efficiently recognize its cortex substrate during germination ( Fig. 1 and 3). Our results further revealed that GerS and CwlD directly interact (Fig. 6) and that GerS incorporation into spores depends on CwlD (Fig. 5). Since GerS homologs are found only in the members of the Peptostreptococcaceae family (20), our results suggest that C. difficile employs a unique mechanism for cortex modification relative to the pathway defined in B. subtilis (29) and possibly most other spore formers based on gene conservation.
Although the muropeptide profiles of C. difficile ΔcwlD and ΔpdaA suggest that C. difficile CwlD and PdaA have activities similar to those of their B. subtilis homologs, our studies raise the issue of why C. difficile CwlD's presumed amidase activity depends on GerS, since CwlD in C. difficile and CwlD in B. subtilis share high amino acid similarity (58%). One possibility is that C. difficile CwlD amidase activity requires heterodimer formation with GerS, in contrast with B. subtilis CwlD, which functions independently (35). In vitro analyses of the peptidoglycan-modifying activities of recombinant CwlD and GerS, singly and in combination, would shed light on the ability of CwlD and/or the GerS-CwlD complex to modify peptidoglycan to generate NAM substrates for PdaA.
In addition to the interdependence between CwlD and GerS in modifying the cortex, CwlD regulates GerS signal peptide processing and incorporation into spores (Fig. 5). It is unclear whether the GerS processing defect in ΔcwlD spores alters GerS function, but mutation of GerS's cysteine 22 lipidation site (20) in the presence of CwlD prevents signal peptide processing without altering GerS function or incorporation into spores (20). The GerS signal peptide processing defect of ΔcwlD spores may reflect the need for GerS to bind CwlD in order to be recognized by the C. difficile lipoprotein signal peptidase (Lsp [51]). Alternatively, CwlD-mediated stabilization of GerS after its transport across the outer forespore membrane or CwlD-dependent transport of GerS across this outer membrane could account for reduced GerS levels in ΔcwlD spores. Distinguishing between these possibilities requires that methods be developed for localizing proteins to the intermembrane space between the outer and inner forespore membranes. Unfortunately, current biochemical methods for removing coat layers disrupt the outer forespore membranes and strip cortex proteins from the spore (20).
Interestingly, both C. difficile CwlD and PdaA were detected in mature spores in both cleaved and uncleaved states, even though both proteins lacked any putative signal peptide cleavage site (Fig. 5), unlike their B. subtilis counterparts (34,37,52). While it is unclear whether cleaved C. difficile CwlD and PdaA are functional, their release from the outer and inner forespore membranes, respectively, could allow them to more efficiently modify cortex peptidoglycan, since their transcriptional regulation indicates they are produced in separate compartments. C. difficile cwlD, like gerS, is controlled by the mother cell-specific SigE sigma factor, while pdaA is controlled by the foresporespecific SigG sigma factor (53)(54)(55). The regulation of C. difficile cwlD contrasts slightly with that of B. subtilis, where cwlD is transcribed in both the mother cell and forespore (35,37). Given that C. difficile CwlD appears to be released from the outer forespore membrane into the intermembrane space, whereas lipidation of GerS presumably retains GerS on the membrane, it is unclear whether cleaved CwlD remains associated with GerS on the outer forespore membrane. Regardless, the observation that GerS processing and incorporation into spores depend on CwlD (Fig. 5), CwlD's presumed amidase activity depends on GerS (Fig. 4), and GerS and CwlD bind stoichiometrically in E. coli (Fig. 6) strongly suggests that GerS function and CwlD function are intimately related.
Indeed, the phenotypes of gerS and cwlD mutants, from their germination defects to their muropeptide profiles, were almost entirely overlapping (Fig. 1, 3, and 4). The pdaA mutant was also defective in generating cortex-specific muramic-␦-lactam (MAL) modifications and exhibited spore germination defects and heat sensitivity comparable to those seen with ΔgerS and ⌬cwlD spores, even though the ΔpdaA spores produced a muropeptide profile that was slightly different from the profiles of the ΔgerS and ΔcwlD spores (Fig. 4). Taken together, our results indicate that C. difficile cortex modification is critical for spore germination and affects the apparent heat sensitivity of purified spores.
Notably, although the germination phenotypes of ΔgerS and ΔcwlD spores were indistinguishable from each other, C. difficile ΔcwlD exhibited circular rather than ovoid spore morphology three times more frequently than ΔgerS or wild-type cells (TEM, Fig. 2). It is unclear why loss of CwlD but not GerS altered spore morphology given that these strains give rise to spores with identical muropeptide profiles; indeed, alterations to spore morphology in B. subtilis cwlD mutant spores have not been mentioned in the literature (33,56). Accordingly, it would be interesting to test whether CwlD's ability to modify the cortex and regulate spore morphology depends on its enzymatic activity.
Taken together, our results confirm that the highly conserved CwlD and PdaA enzymes play roles in modifying the cortex similar to those played by their B. subtilis counterparts while also revealing a novel role for the GerS lipoprotein in regulating this process. The tight interaction between CwlD and GerS, their largely overlapping functions, and the conservation of GerS exclusively in the Peptostreptococcaceae family suggest that the mechanism underlying CwlD-mediated cortex modification may exhibit significant differences in C. difficile relative to B. subtilis and other spore-forming bacteria. Since we previously showed that GerS is required for virulence in a hamster model of C. difficile infection (20), the germination defects of cwlD and pdaA cortex modification mutants will likely reduce lethality in animal models of infection and reduce C. difficile disease recurrence. These analyses suggest that inhibiting GerS function during spore formation could reduce infectious C. difficile spore formation and thus disease transmission.
The Escherichia coli strains used for HB101/pRK24-based conjugations and BL21(DE3)-based protein production are listed in Table S2. E. coli strains were grown at 37°C with shaking at 225 rpm in Luria-Bertani broth (LB). The medium was supplemented with chloramphenicol (20 g/ml), ampicillin (50 g/ml), or kanamycin (30 g/ml) as indicated.
E. coli strain construction. All primers are listed in Table S3. All plasmid constructs were cloned into DH5␣, and all sequences were confirmed using Genewiz. To create the gerS* complementation construct, primer pair 2181 and 2182 was used to amplify gerS, including 387 bp of its upstream region. To construct the gerS-alr2 complementation construct, primer pair 2181 and 2408 was used to amplify 387 bp upstream and downstream of the alr2 gene. While both the gerS and gerS-alr2 complementation constructs restored wild-type levels of germination to ⌬gerS (see Fig. S7 in the supplemental material), GerS levels were consistently higher in the gerS-alr2 complemen-tation construct than in the gerS construct. To create the cwlD complementation construct, primer pair 2362 and 2450 was used to amplify cwlD and 274 bp of its upstream region. To create the pdaA complementation construct, primer pair 2451 and 2457 was used to amplify pdaA and 198 bp upstream and 138 bp downstream of the pdaA gene. The resulting PCR products were recombined into pMTL-YN1C by Gibson assembly. The plasmids were transformed into E. coli DH5␣, and the resulting plasmids were confirmed by sequencing and then transformed into HB101/pRK24.
To create the recombinant protein expression constructs for co-affinity purifications and antibody production, primer pair 2545 and 2546 was used to amplify cwlD lacking its first 25 codons, which encode the transmembrane domain, and the stop codon, using C. difficile genomic DNA as the template. The resulting PCR product was cloned into pET22b to result in a CwlD Δ25 -His 6 fusion construct using Gibson assembly. The resulting plasmids were confirmed by sequencing and then transformed into BL21(DE3) cells.
To create the pdaA expression construct for antibody production, primer pair 1458 and 1459 was used to amplify full-length pdaA without its stop codon. The resulting PCR product was digested with NdeI and XhoI and cloned into pET22b digested with the same enzymes to generate a construct encoding C-terminally His 6 -tagged PdaA. The cspBA expression construct was cloned using a series of primers and plasmids pKS01 and pKS02 as the template, which harbor codon-optimized cspA and cspB, respectively, and were a kind gift of Joseph Sorg. Primer pair 1505 and 1529 was first used for PCR analysis of codon-optimized cspB. Primer pair 1507 and 1508 was used for PCR analysis of codonoptimized cspA lacking its stop codon. The resulting PCR template was used in a second PCR with primer pair 1530 and 1508 to extend the 5= end of the PCR product to overlap the codon-optimized cspB PCR. The resulting PCR product was used in a SOE PCR reaction with the cspB template to generate codon-optimized cspBA lacking its stop codon. This product was cloned into pET28a by digesting both with NcoI and XhoI and ligating the products. The resulting plasmid, pET28a-cspBA, was then used as the template with primer pair 1505 and 1625. The resulting PCR product harbors codon-optimized cspBA with a stop codon; this product was gel purified and then digested with NcoI and XhoI and ligated into pET28a cut with the same enzymes.
Bioinformatic analyses. Homologs of C. difficile 630 CwlD (CD630_01060) and PdaA (CD630_14300) were identified using NCBI PSI-BLAST. Homologs identified in Peptostreptococcaceae family members gave an E value of Ͻe Ϫ124 , in Clostridium spp. an E value of Ͻe Ϫ48 , in Bacillus spp. an E value of Ͻe Ϫ49 , and in Paenibacillus spp. an E value of Ͻe Ϫ62 . Both signal peptide cleavage sites (SignalP 4.1 [52]) and transmembrane helices (TMHMM v. 2.0 [60]) were predicted for C. difficile GerS, CwlD, and PdaA.
C. difficile strain construction. Allele-coupled exchange (ACE [38]) was used to construct clean deletions of gerS, cwlD, and pdaA in the 630Δerm ΔpyrE mutant using uracil and 5-fluoroorotic acid to select for plasmid excision as previously described (44). Flanking primers used to screen for deletions of gerS, cwlD, and pdaA are shown in Fig. S2 and provided in Table S3. Colonies that appeared to harbor gene deletions were confirmed using a primer that binds within the region with a primer that binds to the flanking region as a negative control. At least two independent clones from each allelic exchange were phenotypically characterized.
Complementation strains were constructed as previously described using CDDM to select for restoration of the pyrE locus via recombination of the complementation construct into that locus (44). Two independent clones from each complementation strain were phenotypically characterized.
Sporulation. C. difficile strains were inoculated from glycerol stocks overnight onto BHIS plates containing taurocholate (TA) (0.1% [wt/vol], 1.9 mM). Colonies arising from these plates were used to inoculate liquid BHIS cultures, which were grown to early stationary phase, back-diluted 1:50 into BHIS, and grown until the cultures reached an OD 600 between 0.35 and 0.75. Sporulation was induced on 70:30 agar plates as previously described (45) for 24 h. Sporulating cells were harvested into phosphatebuffered saline (PBS), and sporulation levels were visualized by phase-contrast microscopy.
Phase-contrast microscopy. Sporulating cultures were enumerated for visible signs of sporulation using phase-contrast microscopy. A minimum of 450 cells from three independent replicates were counted, and the percentage of circular forespore and free spore formation was determined relative to the total number of cells with visible signs of sporulation combined with the number of free spores. The cell counts were averaged, and the standard deviation was calculated. Statistical significance was determined using a one-way analysis of variance (ANOVA) and Tukey's test.
Sporulation efficiency assay. Heat-resistant, functional spore formation was measured in sporulating C. difficile cultures after 20 to 24 h as previously described (40). The apparent sporulation efficiency represents the average ratio of heat-resistant CFU to total CFU for a given strain relative to the average ratio determined for the wild type based on a minimum of three biological replicates. Statistical significance was determined using a one-way ANOVA and Tukey's test.
Spore purification. Sporulation was induced on 70:30 agar plates for 2 to 3 days as previously described (45). Spores were washed 6 times in ice-cold water, incubated overnight in water on ice, treated with DNase I (New England Biolabs) at 37°C for 60 min, and purified on a HistoDenz (Sigma-Aldrich) gradient. Phase-contrast microscopy was used to assess spore purity (Ͼ95% pure); the optical density of the spore stock was measured at OD 600 ; and spores were stored in water at 4°C. Despite the circular spore morphology, spore purification yields for the cwlD and pdaA mutants were comparable to those of the gerS mutant and wild-type spores (Fig. 1C). Spore purification yields were determined from four biological replicates performed using two independent spore preparations. Four 70:30 plates were used to induce sporulation for each strain; purified spores were resuspended in 600 l, and the total optical density of this mixture was determined for each strain.
Spore heat sensitivity. To assess spore heat sensitivity,~2 ϫ 10 7 spores were resuspended in 510 l of water (OD 600 of 1.6). The solution was aliquoted as 100-l volumes into two tubes. A 10-l volume was serially diluted for the untreated condition. The remaining four tubes were incubated for 15 min at room temperature or 60°C for 30 min. Samples were serially diluted in PBS following heat treatment and plated onto BHIS-TA. After~23 h, colonies arising from germinated spores were counted. The remaining 90 l of the spores was pelleted and resuspended in EBB buffer ( Germination assay. Germination assays were performed as previously described (20). Approximately 1 ϫ 10 7 spores (OD 600 of 0.35) were resuspended in 100 l of water, and 10 l of this mixture was removed for 10-fold serial dilutions in PBS. The dilutions were plated on BHIS-TA, and colonies arising from germinated spores were enumerated at 23 h. Germination efficiencies were calculated by averaging the CFU produced by spores for a given strain relative to the number produced by wild-type spores for at least three biological replicates from two independent spore preparations. Statistical significance was determined by performing a one-way analysis of variance (ANOVA) on natural log-transformed data using Tukey's test. The data were transformed because the use of two independent spore preparations resulted in a nonnormal distribution.
OD 600 kinetics assay. Germination was induced as previously described (44). About 1.4 ϫ 10 7 spores (OD 600 of 0.48) were suspended in BHIS and exposed to 1% taurocholate (19 mM), and the OD 600 was measured every 3 min for 45 min, followed by every 15 min for a total of 3 h. The change in OD 600 over time was determined by calculating the ratio of the OD 600 measured for TA-treated samples to that measured for untreated samples. Statistical significance was measured by performing a repeatedmeasures ANOVA and Tukey's test.
Total DPA quantification using fluorescence. To evaluate the total amount of DPA contained within spores using terbium fluorescence, approximately 2 ϫ 10 7 spores from each strain were resuspended in 1 ml of buffer 1 [0.3 mM (NH 4 ) 2 SO 4 , 6.6 mM KH 2 PO 4 , 15 mM NaCl, 59.5 mM NaHCO 3 , and 35.2 mM Na 2 HPO 4 ] and incubated at 37°C (background) or 95°C (total DPA) for 1 h. Following heat treatment, 10 l of supernatant was added to 115 l of buffer 2 (1 mM Tris, 150 mM NaCl) with 200 M terbium chloride (Alfa Aesar). Samples were prepared in an opaque 96-well plate (PerkinElmer) and evaluated after 15 min of incubation with terbium chloride using a Synergy H1 microplate reader (BioTek; 270-nm excitation, 420 nm cutoff, 545-nm reading, gain of 100). Reported relative fluorescent unit (RFU) values represent the background fluorescence (wild-type, ΔgerS, ΔcwlD, ΔpdaA, and ΔsleC spores, incubated at 37°C, with terbium) and were subtracted from the fluorescence detected for each strain after 95°C treatment. The data represent results from three biological replicates.
Transmission electron microscopy. Sporulating cultures (23 h) were fixed and processed for electron microscopy at the University of Vermont Microscopy Imaging Center as previously described (61). A minimum of 50 full-length sporulating cells were used for phenotype counting, with the exception that 25 full-length sporulating cells were counted to analyze the phenotype of the cwlD complementation strain.
Muropeptide analysis. Purified spores were used for cortex peptidoglycan purification, muramidase digestion, and HPLC separation of muropeptides essentially as described previously (30). Peaks were collected, further purified by HPLC using a volatile buffer system, and subjected to amino acid analysis as described previously (62) using matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) to analyze muropeptides as described previously (63). Briefly, muropeptides were dissolved in HPLC-grade water and 1 l of each sample was spotted and dried on a MALDI plate. Muropeptide spots were covered with 1 l of 4 mg/ml ␣-cyano-4-hydroxycinnamic acid-50% (vol/vol) acetonitrile-0.2% (vol/vol) trifluoroacetic acid-20-mM ammonium chloride and air dried. Mass spectra were acquired using a 4800 MALDI-TOF/TOF mass spectrometer (Applied Biosystems) in positive ion mode with a mass-to-charge range of 500 to 2,000 using the average from 1,500 individual laser shots. Parent ions were further fragmented utilizing tandem mass spectrometry (MS/MS) in 1-kV positive operating mode, and tandem mass spectra were generated from the sum of 2,500 individual laser shots.
SleC cleavage assay. Germinant-induced processing of the pro-SleC zymogen was assessed using 2 ϫ 10 7 spores per strain suspended in samples containing 100 l of water and 100 l of BHIS. The spore solution was aliquoted in 90-l volumes into two tubes containing either 10 l of water or 10 l of a 10% taurocholate (19 mM) solution. After incubation for 20 min at 37°C, the samples were pelleted for 5 min at 15,000 rpm, 50 l of EBB buffer was added, and samples were prepared for Western blot analysis as described above.
Protein purification for antibody production and His 6 tag pulldown assays. E. coli BL21(DE3) strains (see Table S2 in the supplemental material) were grown for protein purification. The anti-CwlD and anti-PdaA antibodies used in this study were raised against CwlD Δ25 -His 6 , which lacks its first 25 codons, and against full-length PdaA-His 6 in rabbits by Cocalico Biologicals (Reamstown, PA). Δ25-CwlD-His 6 was purified from E. coli strains 2045 using Ni 2ϩ -affinity resin as described in more detail below. PdaA-His 6 was purified similarly as well except that an additional gel filtration chromatography step was used.
For the co-affinity purification assays, E. coli BL21(DE3) strains were transformed with two plasmids to express combinations of Δ25-CwlD-His 6 or untagged Δ22-GerS or untagged CspBA or empty vector and purified using Ni 2ϩ -affinity resin as described above.