To study the role of D-alanylation in C. difficile physiology and antibiotic resistance, we deleted the dltDABC operon by allelic exchange in the 630 strain. This deletion was confirmed by PCR (Supplementary Figure S1). The DNA of the mutant was sequenced to verify the genetic construct. The 630 strain and the ΔdltDABC mutant grew similarly in TY broth (Supplementary Figure S1) indicating that the inactivation of the dltDABC operon did not affect C. difficile growth. To further confirm the implication of the dltDABC operon in D-alanine incorporation at the surface, we performed a quantification assay of esterized D-alanine. As shown in Figure 1 and Supplementary Figure S2, a very low quantity of D-alanine in the cell-wall of the ΔdltDABC mutant induced or not with lysozyme (0.4 μg.mg⁻¹) was detected with a 10-fold decrease compared with the 630 strain not induced with lysozyme. We also observed a 4-fold increase of D-alanylation when 630 strain was grown in the presence of lysozyme, in agreement with the induction of expression of the dlt operon by lysozyme (Woods et al., 2016).

In this study, we intended to identify the specific site of D-alanylation of C. difficile glycopolymers. Therefore, we purified LTA and PSII from the ΔdltDABC mutant and the parental strain. LTA and PSII samples were characterized by NMR, using ¹H and ³¹P 1D spectra as well as 2D COSY, TOCSY and ¹³C-¹H HSQC experiments (Figure 2; Supplementary Figures S3–S5). The proton and carbon chemical shifts are in good agreement with the NMR data previously observed for PSII and o-deacylated LTA (Reid et al., 2012). Thus, LTA seems to be a repetition of a dimer consisting of two N-acetylglucosamine (GlcpNAc) residues (L and M) connected by an a (1-3) linkage. The repeating units seem connected by a 6–6 phosphodiester bridge (6-P-6) between C₆ of residues L and M, as observed on a ³¹P-¹H HMBC experiment (Supplementary Figure S4). Both residues are acetylated on C₂ and M seems to bear a glyceric acid on C₁. The terminal residues L_t and M_t are not observed, probably due to sensitivity issues. 15% of the L residues seem to be N-glucosamine (GlcpN), with the substitution of the N-acetyl group at C₂ for N. This corresponds to a set of signals labeled as L’, and M’ for the other residue in the modified units. Two signals at 1.59 and 1.47 ppm are observed for the 630 strain, which are assigned to CH₃ from alanine groups, as confirmed by COSY and ¹³C-¹H HSQC spectra (Supplementary Figure S3). In addition, a DOSY NMR experiment was acquired to determine if the signals from D-alanine are connected to the carbohydrate chain (Supplementary Figure S5). The signal at 1.47 ppm belongs to a small fast diffusing molecule (diffusion coefficient D = 700 μm².s⁻¹), probably free alanine. The other signal diffuses at the same slow rate as all other signals from the LTA, with a diffusion coefficient D′ = 22 μm².s⁻¹. This shows that the LTA from the 630 strain is substituted by an alanine, on almost 15% of the residues as suggested by the intensity of the CH₃ signal. Furthermore, the slow diffusion coefficient suggests a molecular mass of 171 ± 12 kDa for the LTA (Augé et al., 2009). ¹H NMR analysis of PSII from the 630 strain or the ∆dltDABC mutant did not show any signal that could be assigned to D-alanine, at 1.59 or 1.47 ppm (Figure 2), suggesting that PSII is not D-alanylated. Altogether, our results suggest that D-alanine esterification is exclusive to the LTA of C. difficile.

Modifications of the envelope charge have been reported to be involved in the perturbation of the physicochemical properties of the bacterial cell surface (Giaouris et al., 2009; Nguyen et al., 2011; Wu et al., 2021). Therefore, we investigated the role of D-alanylation of polysaccharides on C. difficile surface properties. Surface hydrophobicity was indirectly analyzed by measuring adhesion to the alkane hydrocarbon n-hexadecane. As shown in Figure 3A, the ΔdltDABC mutant exhibited a significantly lower affinity for n-hexadecane (12.14% ± 3.57%) than the parental strain (20.88 ± 3.07%). This result suggests that the surface charge modification led to a less hydrophobic surface for the ΔdltDABC mutant than the parental strain. No differences in PSII quantities or surface organization in Transmission Electronic Microscopy (TEM) were observed in the ΔdltDABC mutant compared to the parental strain (Supplementary Figure S6). Despite the fact that modification of surface charge impacts autolysis in other Firmicutes (Madela and Fixher, n.d.), the deletion of the dltDABC operon had no impact on the autolysis of C. difficile (Supplementary Figure S7). We also tested the impact of the deletion of the dltDABC operon on motility. Using a motility assay on 0.3% agar plates, we observed a significant increase in the motility of the ΔdltDABC mutant (19.3 mm ± 1.71) compared to the parental strain (11 mm ± 1.71; Figure 3B). This result was not caused by increased production of FliC in the mutant as observed on Western-Blot using an antibody raised against FliC (Supplementary Figure S8). In some Gram-positive bacteria, changes in the surface hydrophobicity can also impact adhesion (Giaouris et al., 2009; Pantaléon et al., 2018). We then investigated the impact of D-alanylation on adhesion on intestinal Caco-2/TC7 cells. As presented in Figure 3C, the adhesion of the ΔdltDABC mutant was significantly lower than that of the parental strain, indicating that the lack of D-alanylation decreases the ability of C. difficile to adhere to Caco-2/TC7 cells.

Since adhesion and motility are modified in the ΔdltDABC mutant, we hypothesized further consequences on other complex biological processes of C. difficile such as biofilm formation. Therefore, we first analyzed the early adhesion and biofilm formation steps on a PVC abiotic surface (Figure 4). Means of adhesive bacteria from the ΔdltDABC mutant were significantly lower than the parental strain at 2 h (Figure 4A). When bacteria multiply in the early biofilm on the PVC, at 4 h and 6 h, the ΔdltDABC mutant was significantly less able to form early biofilms than the 630 strain. By contrast, a difference was not observed after 8 h of early biofilm formation. These results suggest a delay in the phase of adhesion on the abiotic surface and the early steps of biofilm formation up to 6 h for the ΔdltDABC mutant. However, the resolution of this delay after 6 h suggests no further consequences on the biofilm formation properties of C. difficile. Therefore, we tested the ability of the two strains to form mature biofilms at 24 h and 48 h in BHIS, using a crystal-violet quantification assay. As presented in Figures 4B,C, we observed a significant increase in biofilm formation for the ΔdltDABC mutant at 24 h and 48 h compared to the parental strain. Overall, the impact on initial adhesion properties exhibited by the mutant had no consequences on the ability of the ΔdltDABC mutant to further form biofilm.

As observed using a dltD clostron mutant (McBride and Sonenshein; Woods et al.), the esterification of D-alanine at the LTA modulates the resistance to some CAMPs (Nisin, Polymyxin B and Gallidermin) and to lysozyme (McBride and Sonenshein, 2011; Woods et al., 2016). To confirm these results, we tested the susceptibility to lysozyme using an antimicrobial disk assay on Pep-M plates. As shown in Figure 5, we observed a greater size of the zone inhibition for the ∆dltDABC mutant (21 mm) compared to the parental strain (7 mm). Taken together, these results confirm the implication of the dltDABC operon in lysozyme sensitivity in C. difficile, in agreement with previous results (Woods et al., 2016). D-alanylation is a well-known resistance mechanism to antimicrobial compounds and this modification has been previously described in C. difficile as involved in the resistance to a few CAMPs (McBride and Sonenshein, 2011). Therefore, we tested the sensitivity of the ΔdltDABC mutant to molecules that have an activity close to the bacterial membrane, additional CAMPs (LL37 and nisin) and antibiotics (i) targeting specifically the penicillin binding proteins (amoxicillin, imipenem and cefotaxime), (ii) targeting the peptidoglycan synthesis (vancomycin, teicoplanin, bacitracin), and (iii) targeting indirectly the peptidoglycan synthesis by disturbing the membrane (daptomycin). As presented in Table 1, we observed increased sensitivity of the mutant to bacitracin (>7-fold) and nisin (2-fold), to antibiotics of the glycopeptide family, vancomycin and teicoplanin (2-fold) and to the lipopeptide daptomycin (2-fold). In contrast, we did not observe any modifications in susceptibility to β-lactams, cephalosporins or LL37.

To precisely determine the increased sensibility of the ΔdltDABC mutant to bacitracin and vancomycin, we analyzed the survival of both strains in the presence of those antibiotics at various concentrations including the MICs of the two strains. After 24 h of bacitracin exposure to four different concentrations (MIC, MIC/2, MIC/4 and MIC/8 for 630 strain), we observed a significant decrease in survival to bacitracin of the ΔdltDABC mutant in comparison with the wild-type strain at all the concentrations tested (Figure 6A). After 24 h of vancomycin exposure to four different concentrations (2XMIC, MIC, MIC/2 and MIC/4 for 630 strain), we observed a significant decrease in the survival from the ΔdltDABC mutant in comparison with the wild-type strain at 2 μg.mL⁻¹ and 1 μg.mL⁻¹ of vancomycin (Figure 6B). Overall, these results highlight the role of D-alanylation in antibiotics susceptibility in C. difficile.

To further evaluate the potential of D-alanylation as a druggable inactivation target in C. difficile, we investigated the effect of DLT-1, a DltA specific inhibitor first described and tested in B. subtilis (May et al., 2005; Coupri et al., 2019). First of all, we assessed the impact of the inhibitor on the quantities of C. difficile cell-wall esterified D-alanine quantities. As presented in Figure 7, we tested DLT-1 on 3 different C. difficile strains: 630, UK1 (a ribotype 027) and E1 (a 078 ribotype) and quantified the cell-wall esterified D-alanine with and without the inhibitor. The 630 strain possesses a significantly higher quantity of cell-surface D-alanine than the two other tested strains. DLT-1 at 1 mM exhibited a significant decrease in the esterified D-alanine amount detected in cell-wall for all the tested strains, approximatively four-fold (Figure 7). In addition, the effect of the inhibitor was not improved at 2 mM. However, the quantity of D-alanine recovered after DLT-1 inhibition for the 630 strain (2.5 fold reduction in comparison to the 630 strain without treatment) was still higher than the amount detected in the ΔdltDABC mutant (10 fold reduction in comparison to the 630 strain without treatment, Figure 1), suggesting only a partial inhibition of D-alanylation of the LTA. We therefore evaluated the impact of DLT-1 on bacitracin and vancomycin survival of the parental strain. The addition of DLT-1 did not modify the profile of survival of the wild-type strain in the presence of an increasing concentration of bacitracin (from 75 μg.mL⁻¹ to 600 μg.mL⁻¹; Figure 8A) or vancomycin (4 μg.mL⁻¹, 2 μg.mL⁻¹, 1 μg.mL⁻¹ and 0.5 μg.mL⁻¹; Figure 8B). Even if DLT-1 significantly reduced the level of D-alanylation of C. difficile strain 630, the drastic survival changes observed for the ΔdltDABC mutant were not detected in the presence of the DLT-1 inhibitor.

C. difficile strains and plasmids used in this study are presented in Supplementary Table S1. C. difficile strains were grown anaerobically (5% H₂, 5% CO₂, 90% N₂) in Brain Heart Infusion broth (BHI), or in a peptone-containing medium (Pep-M; Ng et al., 2013). For solid media, agar was added to a final concentration of 17 g.L⁻¹. Cefoxitin (Cfx, 25 μg.mL-1) or thiamphenicol (Tm, 15 μg.mL-1) were added to C. difficile cultures when necessary. Escherichia coli strains were grown in LB broth. When indicated, ampicillin (100 μg.mL⁻¹) or chloramphenicol (15 μg.mL⁻¹) was added to the culture medium.

The ∆dltDABC knock-out mutant was obtained using an allele exchange method using the inducible toxicity of the CD2517.1 type I toxin (Peltier et al., 2020). Briefly, 1 kb fragments located upstream and downstream of the four genes to be deleted (from dltD to dltC) were PCR amplified from 630 genomic DNA using primers IMV1286/IMV1287 and IMV1288/IMV1289 (Supplementary Table S2). Purified PCR fragments were then introduced into the pMSR plasmid using the Gibson Assembly^® Master Mix (Biolabs). The sequences of the resulting cloning inserts were verified by sequencing. The obtained plasmids, introduced in HB101 (RP4) E. coli strain, were transferred by conjugation into the C. difficile 630 strain. Transconjugants were selected on BHI plates supplemented with Tm and C. difficile selective supplement (SR0096, Oxoid). Isolation of faster growing single-crossover integrants was performed by serial restreaking on BHI plates supplemented with Cfx and Tm. Single-crossover integrants were then restreaked on BHI plates containing 200 ng.mL⁻¹ of anhydrotetracycline (ATc) allowing the isolation of double crossover events. After confirmation of plasmid loss (Tm sensitive clones), the presence of the expected deletion was checked by PCR using oligonucleotides IMV1273 and IMV1284. The dltDABC mutant and its reference strain 630 were sequenced using Illumina with paired-end 300 bp reads by the diaeresis around Plateforme Microbiologie Mutualisée (P2M – Institut Pasteur). The platform provided filtered pair-end reads, de novo assembly and annotation. Deletion of the dltDABC operon was confirmed using breseq (v0.35.7; Deatherage and Barrick, 2014) with default parameters using filtered reads. This software allows to detect mutation relying on read mapping onto the assembled and annotated reference genomes. The GenBank accession number for the ΔdltDABC mutant is JAUPES000000000.

Ester-linked D-alanine quantification assay was performed as previously described (McBride and Sonenshein, 2011) with modifications (Catherwood et al., 2020). Overnight cultures of strains 630 and 630 ΔdltDABC in BHI broth were harvested by centrifugation at 5000 g for 15 min. Bacterial pellets were stored at −80°C before analysis. After three washing with 0.1 M MES (pH 6), pellets were resuspended in 1 mL of 0.1 M MES (pH 6) containing 0.2% SDS. Samples were then boiled for 15 min to extract the cell wall. After centrifugation, samples were washed five times in 0.1 M MES (pH 6) and lyophilized for 1 h. Pellets were weighed as total cell wall extract for future comparison. After resuspension of the pellets in 0.5 mL of sodium pyrophosphate (pH 8.3), samples were incubated for 3 h at 60°C to release D-alanine residues. After centrifugation, D-alanine was quantified in the supernatant. The assay reagent contained 2.5 μM of Rhodotorula gracilis D-amino acid oxidase, 15 U.mL⁻¹ horseradish peroxidase and 0.2 mM of Amplex Red in 0.1 M sodium pyrophosphate (pH 8.3). 500 μL of the sample were mixed with 500 μL of the assay reagent and incubated at 37°C for 15 min. The reaction was stopped by the addition of 0.1% of SDS. After centrifugation, OD_555nm was measured for each supernatant. To estimate D-alanine concentration, a standard D-alanine curve was performed. D-alanine relative quantities were calculated by dividing D-alanine concentration by previously weighed total cell-wall extracts.

LTA and PSII were purified as previously described (Cox et al., 2013). C. difficile strain 630 was grown in BHI broth supplemented with 0.5 g.L⁻¹ cysteine-HCl, 5 g.L⁻¹ yeast extract, 1 g.L⁻¹ glucose (BHISG) and harvested at an OD_600nm of 1. The bacterial cells were centrifugated (8,000 rpm, 4°C, 20 min), killed by adding phenol to 4% and washed with 10 mM phosphate buffered saline, pH 7.4. To isolate the LTA and PS-II, the cells were first extracted in boiling water for 30 min and the resulting solution was separated by low-speed centrifugation. The supernatant was dialyzed against water and lyophilized. Contaminating proteins and nucleic acids were removed from a 5 mg.mL⁻¹ aqueous solution of the lyophilized material by precipitation with 15% trichloroacetic acid 16 h at 4°C, low-speed centrifugation followed by dialysis of the supernatant against water. The water-soluble material was separated by anion exchange chromatography on a HiTrap Q column using an H₂O 1 M NaCl gradient to give PSII. The remaining cells were subjected to extraction with 45% phenol (68°C, 30 min). The water phase was separated from the phenol phase and cell debris by centrifugation. The phenol phase and cell debris were then re-extracted with more water and treated as above. The two water phases were combined and dialyzed against water to eliminate traces of phenol, and then lyophilized. The dried sample was dissolved in water to give a 1–2% solution (w/v) and treated with deoxyribonuclease I (DNase; 0.01 mg.mL⁻¹) and ribonuclease (RNase; 0.01 mg.mL⁻¹) for 3 h at 37°C, then treated with proteinase K (0.01 mg.mL⁻¹) for 3 h. The sample was then dialyzed against water overnight and lyophilized. The resulting LTA containing sample was purified by anion exchange chromatography as above.

For the PSII dot blot, exponential phase cultures were harvested by centrifugation. The supernatant fraction was recovered and precipitated with 10% TCA for 30 min. The supernatant and the total crude cell fractions were treated with 100 μg.mL⁻¹ of proteinase K (Sigma) for 1 h at 37°C. Samples were then serially diluted and 5 μL of each dilution were spotted onto an activated polyvinylidene difluoride membrane (PVDF). The membrane was washed in H₂O, blocked for 15 min in TBST (20 mM Tris–HCl, 150 mM NaCl, 0.05% Tween20, pH 7.5) containing 10% milk, and then washed in 5% milk in TBST for 2 min. After overnight incubation in PSII-LTB rabbit antiserum (1,10,000), the membrane was washed once in TBST with 5% milk, twice in TBST for 5 min, and once in TBST with 5% milk for 10 min. Following incubation with goat anti-rabbit horseradish-peroxidase-conjugated secondary antibody at 1:10,000 dilution for 1 h, the membrane was washed 5 times in TBST for 5 min and revealed using the SuperSignalWest Femto chemiluminescent substrate.

Lyophilized LTA samples were solubilized in deuterated water. 1,4-dioxane was used as a reference, with δ¹H = 3.75 ppm, and IUPAC recommendations were followed to reference ¹³C and ³¹P spectra (ref: Pure Appl. Chem., Vol. 80, No. 1, pp. 59–84, 2008). All NMR experiments were performed on a Bruker Avance III 400 MHz spectrometer equipped with a Prodigy Cryoprobe. The temperature was regulated at 298 K and standard Bruker parameter sets were chosen. DOSY experiments used a standard BPPLED pulse sequence with a diffusion time d20 = 200 ms and gradient pulses p30 = 2.5 ms. Diffusion data were analyzed using the DynamicsCenter software implemented in Topspin. The molecular mass of LTA was estimated from the diffusion coefficient using the calculator available.¹

Overnight cultures were diluted in fresh BHI medium to a final OD_600nm of 0.05 and incubated at 37°C in anaerobic conditions for 5 h for all assays. Surface hydrophobicity was assessed as the affinity for the apolar solvent n-hexadecane, as previously described (Bellon-Fontaine et al., 1996). The OD_600nm of the suspensions (A0) was measured using a spectrophotometer. The bacterial suspension (2.4 mL) was mixed and vortexed for 2 min with 0.4 mL n-hexadecane (Sigma). The mixture was decanted for 15 min to ensure complete separation of the two phases. The absorbance of the water phase (A) was then measured. The percentage of hydrophobic properties was subsequently calculated by the following equation: % hydrophobicity = ((A0-A)/A0)*100.

For motility assays, 5 μL of exponential-growth-phase cultures were cultivated on BHI plates containing 0.3% agar. Plates were incubated for 24 h at 37°C and the zone of motility was then measured. To test biofilm formation, C. difficile strains were grown in BHISG broth overnight at 37°C in anaerobic conditions. For biomass quantification and bacterial cell counts, 1 mL of an overnight suspension in BHIS was resuspended at OD_600nm 0.05 and cultivated in polystyrene 24-well plates (Costar^®) for 24 h or 48 h in anaerobic conditions at 37°C. Biofilm biomass was quantified by the classical crystal violet (CV; Acros Organics, United States) staining method. A two-step washing procedure with 1 mL of PBS was applied. Biofilms were air-dried for 10 min at 37°C. 1 mL of 0.2% CV (W/V) was applied to biofilms for 30 min. After the removal of CV, wells were washed twice with 1 mL of PBS. Biofilm biomass was solubilized with an 80:20 ethanol/acetone (V/V) solution, scrapped from the plate and quantified by the measure of OD_570nm (V-1200 Spectrophotometer, VWR).

Adhesion assays on PVC were performed on 24-well plates (Costar^™). Overnight cultures were diluted into fresh BHISG broth to a final OD_600nm of 0.05. Plates were inoculated with 1 mL of cultures for 2 h, 4 h, 6 h and 8 h. After two wash steps with 1 mL of PBS to remove non-adherent bacteria, adhesive cells were resuspended in 1 mL of PBS. Viable bacteria were then plated on BHISG and cultivated at 37°C for 24 h.

Cell adhesion assays were carried out on Caco-2/TC7 cells. Cells were cultivated to confluence from frozen stock in 24-well plates (Costar^®) in an MDEM medium. On infection day, Caco-2/TC7 cells were washed once with 1 mL of MDEM medium and incubated in 500 μL of MDEM before infection. Bacteria were grown in BHI broth overnight and diluted at an OD_600nm of 0.05 in fresh BHI. After 3 h at 37°C, OD_600nm measurements were reported on the 630 strain growth curve to calculate CFU/mL. Strains were then diluted to a final concentration of 2×10⁵ CFU/mL and centrifugated at 7000 rpm for 10 min. Pellets were resuspended in an MDEM medium and 500 μL of bacterial suspension (10⁵ CFU/mL) were added to the Caco-2/TC7 cells. After 1 h of incubation at 37°C in anaerobic conditions, cells were washed twice with 1 mL of PBS. Cells lysis was carried out with 1 mL of Saponin and adherent bacteria were plated on BHI medium, incubated for 48 h at 37°C in anaerobic conditions.

For Lysozyme sensitivity assays, exponential-phase cultures of C. difficile strains in Pep-M medium were plated on Pep-M agar plates. Lysozyme (800 μg) was placed on a 6-mm paper disk. The growth inhibition diameter was measured after 24 h of incubation at 37°C.

Strains cultivated overnight were diluted to an OD_600nm of 0.05 in fresh BHI medium and incubated at 37°C for 5 h to obtain bacteria in exponential phase before conducting MICs. MICs were determined on BHI plates by E-test (bioMérieux) after 24 h incubation at 37°C. MICs were also determined in liquid culture as follows: 100 μL of BHI were distributed in a 96-well microplate (Bio-Rad, CA, United States). 100 μL of a 4X antibiotic solution were added into the first well and serially diluted to half. After a 100-time dilution of these cultures, 100 μL were distributed in each well of the plates, that were incubated for 24 h at 37°C. The MIC was visually determined as the lowest antibiotic concentration that inhibited bacterial growth.

To quantify survival to antibiotic treatment, strains cultivated overnight were diluted to an OD_600nm of 0.05 in fresh BHI medium and incubated at 37°C for 5 h to obtain bacteria in exponential phase. Bacteria were diluted 100-times to obtain approximately 5 10⁵ CFU.mL⁻¹ and then distributed on a 96-well microplate (Bio-Rad). Antibiotic solutions were prepared at concentrations 4 times higher than the MIC and serially diluted in half. A DltA inhibitor DLT-1 (May et al., 2005; Coupri et al., 2019) was added when indicated. Samples were collected immediately (T0) to confirm the inoculum size and after 24 h of incubation at 37°C to determine the survival of the different strains by plate counting on BHI agar.

All tests were performed using Graphpad prism software. For analyses of more than three samples, the normality, using a Shapiro wilk test, and the variance, using a Bartlett test were checked. In case of these hypotheses were confirmed, the parametric test of ANOVA was performed followed by a Tukey test. In case at least one of these hypotheses were not confirmed, the nonparametric test of Kruskal Wallis was performed. For the comparison of two samples, the normality, using a Shapiro wilk test, and the variance, using a Fisher test were checked. In case of these hypotheses were confirmed, the parametric test of t-test was performed. In case at least one of these hypotheses were not confirmed, the nonparametric test of Mann Witney was performed.