HexSDF Is Required for Synthesis of a Novel Glycolipid That Mediates Daptomycin and Bacitracin Resistance in C. difficile

ABSTRACT Clostridioides difficile is a Gram-positive opportunistic pathogen responsible for 250,000 hospital-associated infections, 12,000 hospital-associated deaths, and $1 billion in medical costs in the United States each year. There has been recent interest in using a daptomycin analog, surotomycin, to treat C. difficile infections. Daptomycin interacts with phosphatidylglycerol and lipid II to disrupt the membrane and halt peptidoglycan synthesis. C. difficile has an unusual lipid membrane composition, as it has no phosphatidylserine or phosphatidylethanolamine, and ~50% of its membrane is composed of glycolipids, including the unique C. difficile lipid aminohexosyl-hexosyldiradylglycerol (HNHDRG). We identified a two-component system (TCS), HexRK, that is required for C. difficile resistance to daptomycin. Using transcriptome sequencing (RNA-seq), we found that HexRK regulates expression of hexSDF, a three-gene operon of unknown function. Based on bioinformatic predictions, hexS encodes a monogalactosyldiacylglycerol synthase, hexD encodes a polysaccharide deacetylase, and hexF encodes an MprF-like flippase. Deletion of hexRK leads to a 4-fold decrease in daptomycin MIC, and that deletion of hexSDF leads to an 8- to 16-fold decrease in daptomycin MIC. The ΔhexSDF mutant is also 4-fold less resistant to bacitracin but no other cell wall-active antibiotics. Our data indicate that in the absence of HexSDF, the phospholipid membrane composition is altered. In wild-type (WT) C. difficile, the unique glycolipid HNHDRG makes up ~17% of the lipids in the membrane. However, in a ΔhexSDF mutant, HNHDRG is completely absent. While it is unclear how HNHDRG contributes to daptomycin resistance, the requirement for bacitracin resistance suggests it has a general role in cell membrane biogenesis.

. C. difficile is predicted to have approximately 50 TCSs, with slight variations being observed between strains (46). However, the vast majority of these TCSs remain unstudied, with only a few TCSs having any identified function or signal. The cprABCK locus has been found to respond to a subset of lantibiotic compounds and regulate the expression of antibiotic resistance mechanisms, mainly in the form of an ABC transporter (47,48). This system shares a similar mechanism to the bacitracin resistance TCS in B. subtilis, bceRSAB (49,50). The response regulator Spo0A is a major regulator of sporulation; however, its cognate histidine kinase has not been identified, and the manner in which Spo0A becomes activated is still unclear (51)(52)(53). The WalRK TCS is essential and required for proper cell wall morphology in C. difficile (54). The genes encoding the CmrRST TCS are phase variable and consist of two response regulators (RRs) and a single histidine kinase (HK). CmrRST regulates a variety of cell functions, including colony morphology and motility (55,56). The agrACDB locus, homologous to the agr locus in S. aureus, modulates virulence by increasing production of flagella and TcdA (57)(58)(59). The RR CD1688 has been shown to be a repressor of sporulation that functions by inhibiting expression of spoIIR, an important early-stage regulator of sporulation (60). Loss of CD1688 leads to increased sporulation (60). Understanding the roles of these unstudied TCSs can give a better understanding of how C. difficile responds to its environment, particularly the complex host environment.
Here, we describe the identification of the TCS, hexRK, and its regulon, hexSDF. We show that HexRK and HexSDF mediate daptomycin and bacitracin resistance in C. difficile. We also show that HexSDF is required for synthesis of the novel glycolipid HNHDRG and that loss of HexSDF leads to large changes in lipid content of the cell membrane.

RESULTS
Identification of hexRK and hexSDF. We performed a transposon insertion sequencing (Tn-seq) experiment in which we sought to identify genes that are required for growth in the presence of daptomycin. We generated a transposon insertion library containing ;80,000 colonies using pRPF215 as previously described (61). The transposon library was then exposed to 0 and 0.5 mg/mL daptomycin. Upon sequencing, we found there were ;45,000 unique transposon insertions, which was lower than expected. However, we identified one gene of interest, cdr20291_2610 (here referred to as hexK for histidine kinase regulating synthesis of a hexose-based glycolipid), which had fewer insertions recovered when cells were grown with daptomycin (see Table S1 in the supplemental material). hexK is predicted to encode a TCS sensor histidine kinase (HK) with 2 transmembrane domains and an approximately 120-amino-acid extracytoplasmic domain as predicted by DeepTMHMM (62). hexK is immediately downstream of cdr20291_2611 (here referred to as hexR for response regulator controlling synthesis of a hexose-based glycolipid), a predicted TCS RR. Through bioinformatic analysis, we found that HexK is homologous to the relatively uncharacterized HK from B. subtilis, YrkQ (Fig. S1). Due to our interest in signal transduction, we focused on HexRK for survival in the presence of daptomycin.
To confirm that loss of hexRK led to a decrease in daptomycin resistance, as the Tnseq data suggested, we utilized CRISPR interference (CRISPRi) to knock down expression of the hexRK operon. We found knockdown of hexR leads to an ;4-fold decrease in daptomycin resistance (Fig. 1B). We then constructed a DhexRK deletion using CRISPR mutagenesis, and we found that loss of HexRK decreased daptomycin resistance ;4-fold (Fig. 1C). We then asked if we could restore daptomycin resistance by expression of HexR D56E (a mutant of the putative phosphorylation site), which mimics phosphorylation of HexR and likely increases HexR activity. When hexR D56E is expressed from a xylose-inducible promoter in a DhexRK mutant, we observed an increase in daptomycin resistance to near-wild-type levels (Fig. 1D). Taken together, our data suggest that HexRK is required for daptomycin resistance.
Identification of the HexR regulon. We hypothesized that HexRK likely regulates expression of genes involved in daptomycin resistance. Thus, we performed transcriptome sequencing (RNA-seq), comparing WT with DhexRK to identify any genes that were differentially regulated in the absence of hexRK. Previous RNA-seq experiments showed hexRK was expressed at relatively high levels in log-phase growth (54); thus, we performed RNA-seq experiments on cultures grown to mid-log phase (optical density at 600 nm [OD 600 ], 0.7 to 0.8) (Table S2).
We found that 32 genes, not including hexRK, were differentially regulated between wild type (WT) and DhexRK with a P value of ,0.05 and a log 2 fold change of greater than 2 or less than 22. Of these 32 genes in WT versus DhexRK, 24 showed an increase in expression, and 8 showed a decrease in expression (Table 1; Fig. 2A). Since our data suggest that HexRK is required for daptomycin resistance, we were most interested in those genes whose expression was dependent upon HexRK. We were particularly interested in a three-gene operon, cdr20291_2614-2613-2612 (hexSDF), located immediately upstream of hexRK (Fig. 1A). We have named the genes hexSDF due to their involvement in the synthesis of a hexose-based glycolipid, and based on predictions, HexS appears as a synthase, HexD a deacetylase, and HexF a flippase. The genes in the hexSDF operon were expressed ;12-fold lower in the DhexRK mutant (Table 1; Fig. 2A). The RNA-seq revealed other differentially regulated genes; however, several of these genes are reported to be phase variable, including cdr0440, cdr0962, cdr0963, and cdr1514 (63)(64)(65)(66). As such, we chose to focus our efforts on better understanding the function of the hexSDF operon. To confirm that HexRK is required for hexSDF expression, we constructed a P hexS -sLuc opt reporter. We found that expression of the P hexS -sLuc opt reporter was reduced by 26-fold in DhexRK compared to WT, indicating that expression of hexSDF is dependent upon HexRK (Fig. 2B). Since we found that daptomycin resistance decreases in the absence of HexRK, we were interested if daptomycin was an activator of the HexRK system. We tested if the P hexS -sLuc opt was induced after a 2-h treatment with various concentrations of daptomycin. We found, under these conditions, that daptomycin did not lead to increased expression of the hexSDF locus (Fig. 2C). We also tested growth with antibiotics for up to 6 h but did not detect any induction. hexK (cdr2610) is annotated as a TCS histidine kinase, and hexR (cdr2611) is annotated as a TCS response regulator. hexF (cdr2612) is annotated as a MprF-like flippase, hexD (cdr2613) is annotated as a polysaccharide deacetylase, and hexS (cdr2614) is annotated as a monogalactosyldiacylglycerol synthase. The neighboring gene adjacent to hexK is cdr2609 and is annotated as a putative membrane protein. The gene neighboring hexS is cdr2615 (glyA) and is annotated as a putative serine hydroxymethyltransferase. (B) MICs of CRISPRi knockdown with 2 guides, each targeting a different location within hexR. These knockdowns result in ;4-fold decreases in daptomycin resistance. Data were analyzed using a one-way analysis of variance (ANOVA) with Dunnett's multiple-comparison test. ****, P , 0.0001; ***, P , 0.001. (C) DhexRK has ;4-fold-lower daptomycin MIC than WT. Data were analyzed with an unpaired t test; ****, P , 0.0001. (D) Overproduction of HexR D56E in DhexRK restores daptomycin resistance to WT levels compared to DhexRK with empty pAP114 (EV). Data were analyzed using one-way ANOVA using Sidak's multiple-comparison test. *, P , 0.05; ns, P . 0.05.

Synthesis of a Novel Glycolipid Requires HexSDF mBio
Loss of hexRK and hexSDF leads to daptomycin and bacitracin susceptibility. HexS is annotated as a putative monogalactosyldiacylglycerol synthase, HexD is a predicted polysaccharide deacetylase, and HexF is annotated as MprF-like. In other organisms, MprF is known to contribute to daptomycin resistance ( Fig. 1A) (67)(68)(69). MprF from S. aureus and B. subtilis is a bifunctional protein that adds lysine onto phosphatidylglycerol, generating Lys-PG, and then flips Lys-PG from the inner leaflet of the membrane to the outer leaflet (70)(71)(72). Bioinformatic analysis reveals that HexF contains homology to the flippase domain of MprF, but HexF does not appear to contain the domain required for lipid lysinylation (Fig. S2). Based on the bioinformatic predictions of the hexSDF operon, we hypothesized that the operon likely encodes proteins that alter biosynthesis of membrane lipids.
To determine if HexSDF is required for daptomycin resistance, we utilized CRISPRi to knock down expression of the hexSDF operon. We found that knockdown of hexSDF led to a ;4-fold decrease in daptomycin MIC (Fig. 3A). Having demonstrated that CRISPRi knockdowns of hexSDF decreased daptomycin resistance, we constructed a deletion of the hexSDF operon (DhexSDF) and found that this resulted in an ;8-fold decrease in daptomycin resistance compared to WT (Fig. 3B). We then complemented these mutants by expressing hexSDF from a xylose-inducible promoter. We found that overproduction of HexSDF in WT C. difficile R20291 did not lead to significant increases in daptomycin resistance; however, when HexSDF is overproduced in either DhexRK or

Synthesis of a Novel Glycolipid Requires HexSDF mBio
DhexSDF, daptomycin resistance is restored to WT levels ( Fig. 3C). Taken together, these data suggest that HexSDF is required for HexRK-mediated daptomycin resistance. We sought to determine if loss of hexSDF affected resistance to other antimicrobials. We screened several other cell wall-targeting antimicrobials, including vancomycin, bacitracin, nisin, polymyxin, ampicillin, and lysozyme. We found that loss of hexSDF decreased bacitracin resistance by ;4-fold ( Fig. 3D) but did not result in altered resistance to other antimicrobials (Fig. S3). Expression of hexSDF in DhexSDF resulted in bacitracin MIC comparable to WT; however, overproduction of HexSDF in WT did not affect bacitracin MIC (Fig. 3D).
We wanted to better understand the contribution of each gene in the hexSDF operon to daptomycin and bacitracin resistance. Thus, we constructed deletions of each gene individually. We found that loss of hexS had the greatest effect on daptomycin resistance, decreasing the daptomycin MIC by ;4-fold (Fig. 3E). The hexD and hexF mutations lead to an ;2-fold decrease in the daptomycin resistance (Fig. 3E). We also investigated how the individual deletions affected bacitracin resistance and found trends similar to what was found with daptomycin ( Fig. S4A). We next overproduced pairs of genes from the hex operon and found that P xyl -hexSF, P xyl -hexSD, and P xyl -hexSDF all led to increases in daptomycin resistance (Fig. 3F), while P xyl -hexDF did not lead to an increase in resistance (Fig. 3F). We found that the bacitracin MICs showed a remarkably similar trend where HexS expression is essential for increased bacitracin resistance (Fig. S4B). Combined, these data indicate that HexS is essential for increased daptomycin and bacitracin resistance.
Finally, we wanted to determine if HexSDF is important for daptomycin resistance in other strains of C. difficile. C. difficile 630 (PCR ribotype 012) is a widely used and well-studied strain of C. difficile and thus the most appropriate strain to use for comparison to our findings in R20291. Strain 630, like nearly all strains of C. difficile, encodes the hex locus. We utilized CRISPRi to knockdown hexSDF in 630Derm and found, similar to R20291, an ;2to 4-fold decrease in daptomycin MIC compared to the Neg sgRNA (Fig. S5). This suggests that HexSDF is involved in daptomycin resistance in multiple strains of C. difficile.
HexSDF is required for production of HNHDRG. Based on bioinformatic predictions of the hexSDF operon, we hypothesized that the alterations in daptomycin and bacitracin resistance were due to changes in membrane lipid composition. Several lines of evidence supported this hypothesis. First, HexS is annotated as a monogalactosyldiacylglycerol synthase,  Table S2 in the supplemental material). (B) The luciferase reporter P hexS -sLuc opt shows significantly decreased activity in DhexRK compared to WT, confirming the differential expression identified in RNA-seq. Data were analyzed with an unpaired t test; ****, P . 0.0001. (C) The P hexS -sLuc opt reporter does not show altered luminescence in the presence of multiple concentrations of daptomycin after incubation for 2 h.

Synthesis of a Novel Glycolipid Requires HexSDF mBio
indicating that it likely is adding sugar groups to a lipid (46,73). Second, HexF is annotated as an MprF-like protein.
In other organisms such as S. aureus, MprF is known to contribute to daptomycin resistance by converting phosphatidylglycerol into lysophosphatidylglycerol (67)(68)(69). Importantly, HexF only contains homology to the flippase domain of MprF (Fig. S2), meaning that HexF could play a role in flipping lipids across the membrane. Finally, HexD is annotated as a putative polysaccharide deacetylase, potentially removing an N-acetyl group, resulting in the production of an amino sugar. Previous work defining the polar lipids in C. difficile revealed the presence of a glycolipid with an amino sugar HNHDRG (34). Based on these bioinformatic predictions, we hypothesized that HexSDF is required for production of the novel glycolipid HNHDRG.

Synthesis of a Novel Glycolipid Requires HexSDF mBio
When HexSDF is overproduced in a DhexSDF mutant, HNHDRG is detected in similar levels to WT (Fig. 4A). Interestingly, we found that the three other major glycolipids, MHDRG (Fig. 4B), DHDRG (Fig. 4C), and THDRG (Fig. 4D), increased in DhexSDF compared to WT. The levels of all three glycolipids were also decreased to levels lower than WT when HexSDF was overproduced (Fig. 4B to D). In contrast, phosphatidylglycerol (Fig. 4E) and cardiolipin (Fig. 4F) levels were not significantly different in the DhexSDF mutant or strains overproducing HexSDF; however, the levels of both did trend upward in the DhexSDF mutant. We next sought to determine the contribution of individual genes in the hex operon. We found overproduction of HexDF did not lead to accumulation of either HNHDRG (Fig. 4A). This suggests that HexS is absolutely required for HNHDRG production. Interestingly, we found that overproduction of HexSF leads to accumulation of HNacHDRG, a previously undetected glycolipid which we predict is the acetylated intermediate of HNHDRG, but there was no detectable Synthesis of a Novel Glycolipid Requires HexSDF mBio HNHDRG ( Fig. 4A and G). This suggests that the putative deacetylase HexD is required for deacetylation of HNacHDRG to produce HNHDRG. In contrast, overproduction of HexSD leads to accumulation of near-wild-type levels of HNHDRG (Fig. 4G), suggesting HexF is not essential for HNHDRG production. In toto, we hypothesize that HexS is required for incorporation of N-acetyl-hexosyl onto the MHDRG precursor, resulting in an HNacHDRG intermediate, and that HexD is required for deacetylation, which leads to production of HNHDRG.

DISCUSSION
Identification of an operon required for production of a novel glycolipid. The C. difficile membrane consists of ;50% glycolipids, including MHDRG, DHDRG, THDRG, and HNHDRG. We do not currently know the nature of the hexosyl groups on these lipids, and in fact, we do not know if MHDRG is a single type of glycolipid or a mixture of two or more glycolipids with different sugars. HNHDRG is a novel glycolipid that makes up ;16% of the membrane WT C. difficile and, to date, has not been described in other bacteria (34). The principal contribution of this investigation is the identification of HexRK and HexSDF as essential for the production of the novel glycolipid HNHDRG. We identified these genes in a Tn-seq screen for genes involved in daptomycin resistance. We found that knockdowns or knockouts of either hexRK or hexSDF lead to decreases in daptomycin and bacitracin resistance. We find that in the absence of HexSDF, there is almost no detectable HNHDRG. We could complement these defects with a plasmid expressing hexSDF from a xylose-inducible promoter. This argues that HexSDF is required for resistance to daptomycin and bacitracin as well as production of the novel glycolipid HNHDRG.
Model for HNHDRG synthesis. Based on the lipidomics data, we hypothesize HNHDRG is likely produced by HexS catalyzing the addition of an N-acetyl-hexose to an MHDRG (Fig. 5). This lipid is then flipped to the outer leaflet of the membrane by HexF, and the N-acetyl group is deacetylated by HexD, producing HNHDRG on the outer leaflet of the membrane (Fig. 5). This model suggests deacetylation is required for production of HNHDRG and is supported by lipidomics data from strains lacking HexD, which show the accumulation of HNacHDRG (Fig. 4G). While we have identified the genes required for synthesis of HNHDRG, the steps of HNHDRG synthesis are still not entirely defined. Currently, little is known about membrane lipid synthesis in C. difficile and if it differs from better-studied organisms.
In B. subtilis, all known glycolipid synthesis is dependent upon a single enzyme, UgtP, which catalyzes the addition of glucose to diacylglycerol to create monoglucosyl diacylglycerol (MGDG) (74). In B. subtilis, UgtP works processively to add an additional glucose group to MGDG to synthesize diglucosyl diacylglycerol (DGDG) and a third glucose to DGDG to create triglucosyl diacylglycerol TGDG (74). We think it is unlikely that HexS is responsible for catalyzing the addition of two different sugars, one without an N-acetyl group and one with an N-acetyl group, onto diradylglycerol. Instead, we hypothesize that another enzyme, potentially one of the three C. difficile genes predicted to encode UgtP-like homologs (cdr20291_0008, cdr20291_1186, and cdr20291_2958) synthesizes MHDRG and HexS and then adds N-acetyl-hexosyl to MHDRG to generate HNacHDRG. This is supported by our data showing that HexS is absolutely required for HNHDRG synthesis. Consistent with this model, we find increased levels of MHDRG, DHDRG, and THDRG in all strains lacking HexS. We propose that in the absence of HexS, MHDRG is no longer being consumed to generate HNHDRG; thus, more MHDRG is available to be used to synthesize DHDRG and THDRG. We also find that overproduction of HexS leads to significant decreases of MHDRG, DHDRG, and THDRG, which is consistent with this model. However, at this time, we cannot rule out the possibility that HexS adds hexosyl-N-acetyl-hexosyl directly to diradylglycerol (DRG).
As noted earlier, we find that overproduction of HexSF without HexD results in the production of HNacHDRG but not HNHDRG. This is consistent with the role of HexD in deacetylation of HNacHDRG to produce HNHDRG. However, we were surprised that expression of HexSD without HexF produced HNHDRG at levels equivalent to all three genes alone. We hypothesize three possible explanations for this phenomenon. One possibility is that Synthesis of a Novel Glycolipid Requires HexSDF mBio HexD functions intracellularly prior to flipping of HNHDRG. A second possibility is that HexF is not required for flipping. The third possibility is that in the absence of HexF, other flippases are capable of mediating transport/flipping of HNacHDRG, where it is acted on extracellularly by HexD. It is interesting to note that C. difficile encodes two additional genes (cdr20291_2010 and cdr20291_3180) that are predicted to encode proteins with homology to HexF/MprF and may also function as flippases (Fig. S6). Taken together, these data support a model in which HexS adds N-acetyl-hexosyl to MHDRG, leading to production of HNacHDRG. HNacHDRG is then deacetylated by HexD to produce HNHDRG. Identification of the HexRK regulon. HexRK represents a family of TCSs that have undefined functions. We found that HexK shares homology with YrkQ in B. subtilis. YrkQ is a relatively unstudied TCS, and B. subtilis is not known to produce HNHDRG, nor is it known what signals activate YrkPQ. To date, we have been unable to identify a signal to which HexRK responds; neither daptomycin ( Fig. 2C) nor bacitracin (see Fig. S4C in the supplemental material) leads to activation of the P hexS -sLuc opt reporter, and our RNA-seq data suggest hexRK has a high basal expression (54). While the signals are not known, both the HexRK and YrkPQ regulons have been identified (75). Similar to the C. difficile HexR regulon, the YrkP regulon is small, ;8 genes, and consists mostly of proteins with predicted function involved in lipid or membrane biogenesis. This includes YkcB, a predicted C 55 -P-Glc glucosyltransferase; YkcC, a predicted lipoteichoic acid glycosyltransferase; YrkO, a membrane protein with 10 predicted transmembrane domains and DUF418 domain, which may be involved in transport; YrkN, a putative GNAT N-acetyltransferase; and YrkR, a 4-transmembrane domain protein with homology to phosphate starvation-inducible protein PsiE. This raises the possibility that HexRK and YrkPQ are responsible for regulating genes involved in membrane or cell envelope biogenesis in multiple organisms. Consistent with this, the YrkP regulon is activated in daptomycin-resistant mutants of B. subtilis (76). However, it is not clear how much YrkPQ contributes to daptomycin resistance in B. subtilis. It is also important to note that B. subtilis does not encode an operon homologous to hexSDF.
Mechanism of HNHDRG daptomycin resistance. Since phosphatidylglycerol is required for daptomycin to kill target cells, we initially expected that loss of HexSDF may lead to an increase in phosphatidylglycerol content and thus a decrease in daptomycin resistance. However, we found there was not a significant change in phosphatidylglycerol levels in DhexSDF (Fig. 4F). The subtle changes we observed in phosphatidylglycerol are still present in strains lacking either HexD or HexF. Consistent with this, we find that loss of HexD or HexF also leads to a decrease in daptomycin resistance (Fig. 3E). Taken together, it suggests that HNHDRG is playing a more direct role in daptomycin resistance.

Synthesis of a Novel Glycolipid Requires HexSDF mBio
We are unsure how HNHDRG increases daptomycin resistance. However, one potential clue comes from the screening of the DhexSDF mutant for sensitivity to additional antibiotics, where we find HexSDF is required for resistance to bacitracin and daptomycin but not other cell wall-targeting antibiotics. It has been hypothesized that since C. difficile does not produce PE or PS, HNHDRG serves to balance the charge of the membrane, as it can form a positively charged zwitterion (34). This could potentially explain the resistance to daptomycin and bacitracin. Both daptomycin and bacitracin require cationic metal ions to function. Daptomycin requires Ca 21 , while bacitracin prefers Zn 21 , though it can utilize other divalent metal ions (21,26,77). We hypothesize that HNHDRG decreases the ability of positively charged metal ions to access the membrane, limiting the activity of both daptomycin and bacitracin. However, there is the caveat that loss of HNHDRG does not affect the activity of cationic antimicrobial peptides, such as lysozyme (Fig. S3), which one may expect if HNHDRG functioned by alteration of membrane charge. While the exact mechanism of resistance is unclear, we hypothesize that HNHDRG is affecting resistance directly, as opposed to alterations in other membrane lipid species being responsible for resistance.

MATERIALS AND METHODS
Bacterial strains, media, and growth conditions. Bacterial strains are listed in Table 2. The C. difficile strains used in this study are derivatives of R20291 (78) and 630Derm (79). C. difficile strains were grown on tryptone-yeast (TY) medium that consisted of 3% tryptone, 2% yeast extract, and 2% agar (for solid medium). TY was supplemented as needed with thiamphenicol at 10 mg/mL (Thi 10 ). Conjugations were performed on solid brain heart infusion (BHI) media supplemented with yeast extract (3.65% BHI, 0.5% yeast extract, 2% agar) and plated on TY with thiamphenicol at 10 mg/mL, kanamycin at 50 mg/mL, and cefoxitin at 50 mg/mL. C. difficile strains were maintained at 37°C in an anaerobic chamber (Coy Laboratory Products) in an atmosphere of ;2 to 2.5% H 2 , 5% CO 2 , and 85% N 2 .
Plasmid and bacterial strain construction. All plasmids are listed in Table 3. Plasmids were constructed using Gibson Assembly (New England Biolabs, Ipswich, MA). Regions of the plasmids constructed using PCR were verified by DNA sequencing. Oligonucleotide primers used in this work were synthesized by Integrated DNA Technologies (Coralville, IA) and are listed in Table S3 in the supplemental material. All plasmids were propagated using OmniMax-2 T1R as a cloning host. CRIPSR-Cas9 deletion plasmids were passaged through E. coli strain MG1655 before transformation into B. subtilis strain BS49 (80). CRISPR-Cas9 plasmids were built on the backbone of pJK02 (81) with some modification (82). CRISPR mutagenesis was performed as previously described (82). Briefly, CRISPR-Cas9, single guide RNA (sgRNA), and 500-bp upstream and downstream homology are provided on a plasmid delivered through conjugation. Genes targeted with sgRNA are subject to double-stranded break via Cas9; homologous recombination is possible through the supplied homology. This recombination even results in a markerless deletion.
For xylose-inducible overexpression constructs, genes of interest were amplified using PCR, and the oligonucleotides are listed in Table S3. PCR amplicons were then inserted into the plasmid pAP114 at the SacI and BamHI sites as described previously (83). For CRISPRi constructs, two guides were created for each targeted gene of interest, and guides were amplified using PCR. The oligonucleotide sequences for the guides are listed in Table S3. PCR amplicons were then inserted into the pIA33 backbone at the MscI and NotI sites as previously described (83). For nanoluciferase reporter constructs, promoters of interest were amplified using PCR; the oligonucleotides are listed in Table S3. PCR amplicons were then inserted into pAP24 (84) at the KpnI and SacI sites using Gibson assembly. These plasmids were then passaged through E. coli HB101/pRK24 for conjugation into C. difficile (85).
Antibiotic MIC determination. Overnight cultures of C. difficile were subcultured, grown to late log phase (OD 600 of 1.0), and then diluted into TY to 10 6 CFU/mL. For strains containing plasmids with xylose-inducible elements, 1% xylose was added to the overnight cultures and subcultures. A series of antibiotic concentrations were prepared in a 96-well plate in 50 mL TY broth. Wells were inoculated with 50 mL of the diluted late-log-phase culture (0.5 Â 10 5 CFU/well) and grown at 37°C for 16 h. For strains containing vectors with xylose-inducible elements, 1% xylose was included in the media. After the incubation period, the 96-well plates were spun down for 5 min at 5,000 rpm, and the MIC was determined based on the presence of cell pellets. For lysozyme MICs, the steps remained the same, with the exception that after the 16 h incubation, 5 mL from each well was plated on TY for viability. The MIC from this viability plate was read the following day, with the MIC being defined as the lowest concentration of lysozyme at which 5 or fewer colonies were found per spot.
Nanoluciferase assays. Overnight cultures of C. difficile were subcultured 1:100 and grown until mid-log phase (OD 600 , ;0.5). For antibiotic induction experiments, a series of antibiotic concentrations were prepared in a 96-well plate in 50 mL TY broth. Wells were inoculated with 50 mL of mid-log-phase culture. Inoculated plates were allowed to incubate for 2 h at 37°C in the anaerobic chamber. Fifty microliters from each well was then transferred to solid white 96-well plates. A 1:100 dilution was made of the Synthesis of a Novel Glycolipid Requires HexSDF mBio Nano-Glo luciferase assay (Promega) working solution. Fifty microliters of the diluted Nano-Glo solution was added to each well in the white plate. Plates were allowed to rest at room temperature for 5 min prior to luminescence measurement. Luminescence was measured using an Infinite M200 Pro plate reader (Tecan). Luminescence values were normalized to OD 600 . RNA extraction. Overnight cultures were grown in TY and then subcultured 1:100 in 100 mL of TY. Subcultures were grown to an OD 600 of 0.7 to 0.8, mixed 1:1 with a solution of 50% ethanol and 50% acetone, and immediately frozen at 280°C. Samples were stored at 280°C until they were ready for harvest. Extraction of RNA was performed using the RNeasy extraction kit (Qiagen) as previously described (54). The RNA integrity number (RIN) was determined using the Agilent Bioanalyzer 2000 at the University of Iowa.
RNA-seq and analysis. Sequencing was performed by (SeqCenter, Pittsburgh, PA) using 51-bp paired-end (PE) reads. We achieved a read depth of ;80 to 120Â per sample. Data analysis was done using the ProkSeq (86) pipeline, which performs trimming and quality control (QC), and sequence alignment using Bowtie2, quantification of RNA expression levels using FeatureCounts, and differential expression analysis using DEseq2 (87). The full RNA-seq data set can be found in Table S2.
Lipidomics. Overnight cultures were grown in TY supplemented with 1% xylose and 10 mg/mL thiamphenicol in biological triplicate. Overnight cultures were then subcultured to an OD 600 of 0.05 in TY supplemented with 1% xylose and 10 mg/mL thiamphenicol. Subcultures were allowed to grow until an OD 600 of 0.6 to 0.7 was reached, at which point the cells were harvested and pelleted at 8,000 rpm for 20 min. Biological replicates (3) were grown on different days.
Lipid extraction and LC-MS/MS were performed by Cayman Chemical Company. After thawing, cells were mixed with 5 mL methanol, transferred to 7-mL Precellys tubes containing 0.1-mm ceramic beads (Bertin Technologies; CK01 lysing kit), and homogenized with 3 cycles at 8,800 rpm for 30 s, with 60-s pauses between cycles. Then, 800 mL of the homogenized mixtures was transferred to 8-mL screw-cap glass tubes. A methyl tert-butyl ether (MTBE)-based liquid-liquid extraction protocol was used by first adding 1.2 mL methanol containing a mixture of deuterated internal standards covering several major lipid categories (fatty acids, glycerolipids, glycerophospholipids, sphingolipids, and sterols) and then 4 mL MTBE. The mixture was incubated on a tabletop shaker at 500 rpm at room temperature for 1 h and then stored at 4°C for 60 h to maximize lipid extraction. After bringing the samples to room temperature, phase separation was induced by adding 1 mL water to each sample. The samples were vortexed and then centrifuged at 2,000 Â g for 15 min. The upper organic phase of each sample was carefully removed using a Pasteur pipette and transferred into a clean glass tube. The remaining aqueous phase was reextracted with 2 mL of the upper phase of MTBE/methanol/water at 10:3:2.5 (vol/vol/vol). After vortexing and centrifuging, the organic phase was collected and combined with the initial organic phase. The extracted lipids were dried overnight in a SpeedVac vacuum concentrator. The dried lipid extracts were reconstituted in 200 mL n-butanol-methanol at 1:1 (vol/vol) and transferred into autosampler vials for analysis by LC-MS/MS. Aliquots of 5 mL were injected into an Ultimate 3000 ultraperformance liquid chromatography (UPLC) system connected to a Q Exactive Plus Orbitrap mass spectrometer (Thermo Scientific). An Accucore C 30 2.6 mm, 150-by 2.1-mm HPLC column (Thermo Scientific) was used, using mobile phases A (acetonitrile/water/formic acid 60:40:0.1 [vol/vol/vol], containing 10 mM ammonium formate) and B (acetonitrile/isopropanol/formic acid 10:90:0.1 [vol/vol/vol], containing 10 mM ammonium formate). Lipids were eluted at a constant flow rate of 300 mL/min using a gradient from 30% to 99% mobile phase B over 30 min. The column temperature was kept at a constant 40°C. Polarity switching was used throughout the gradient to acquire high-resolution MS data (resolution, 75,000) and data-dependent MS/MS data.
Data analysis was performed using Lipostar software (version 2; Molecular Discovery) for detection of features (peaks with unique m/z and retention time), noise and artifact reduction, alignment, normalization, and lipid identification. Automated lipid identification was performed by querying the Lipid Maps Structural Database (LMSD), modified by Cayman to include many additional lipids not present in the LMSD.
Statistical analysis of the data presented in Fig. 4 was performed on the summed peak areas of lipids with the same head groups and normalized to the WT values to allow for comparison between strains. Data were analyzed with one-way analysis of variance using Dunnett's multiple-comparison test.
The resulting read data were analyzed using the Transit package for Tn-seq analysis (89). Briefly, the reads were demultiplexed using Sabre (90). The data were then trimmed and aligned to eliminate poorquality reads using the TRANSIT preprocessor (89). TRANSIT aligned the trimmed reads to the C. difficile R20291 genome using the Burrows-Wheeler Aligner (BWA), tabulated Tn insertion counts, and calculated conditional essentially utilizing the resampling comparison method (89).
Data availability. RNA-seq data were submitted to the NCBI GEO repository and assigned accession number GSE220119.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.