Anti-infective bile acids bind and inactivate a Salmonella virulence regulator

Abstract

Bile acids are prominent host and microbiota metabolites that modulate host immunity and microbial pathogenesis. However, the mechanisms by which bile acids suppress microbial virulence are not clear. To identify the direct protein targets of bile acids in bacterial pathogens, we performed activity-guided chemical proteomic studies. In Salmonella enterica serovar Typhimurium, chenodeoxycholic acid (CDCA) most effectively inhibited the expression of virulence genes, invasion of epithelial cells and interacted with many proteins. Notably, we discovered that CDCA can directly bind and inhibit the function of HilD, an important transcriptional regulator of S. Typhimurium virulence and pathogenesis. Our characterization of bile acid-resistant HilD mutants in vitro and in S. Typhimurium infection models, suggests HilD is one of the key protein targets of anti-infective bile acids. This study highlights the utility of chemical proteomics to identify the direct protein targets of microbiota metabolites for mechanistic studies in bacterial pathogens.

Graphical Abstract

Introduction

The discovery of new pathogens and the emergence of multi-drug resistant microbes demands a better understanding of host-microbe interactions to develop new therapeutic approaches for controlling infectious diseases. In this regard, the microbiota and diet are important factors that modulate host immunity and control susceptibility to enteric pathogens¹. Notably, the metabolites from microbiota and diet provide an important source of nutrients for host cells and microbes, but can also modulate immunity and directly inhibit pathogen infection. In particular, bile acids are a diverse family of microbiota metabolites that have emerged as key modulators of health and disease^{2, 3}.

Bile acids are initially synthesized in the liver via multiple-step oxidation of cholesterol and stored in the gallbladder. As amphipathic sterols, bile acids that are secreted into the small intestine can facilitate digestion of dietary fats and oils by serving as surfactants. Once secreted into the intestine, conjugated primary bile acids are deconjugated by gut microbiota that express bile salt hydrolase (BSH) enzymes to form free bile acids, which can then be further metabolized to generate secondary bile acids. These structurally diverse bile acids play important roles in host immunity and infection^{3, 4}. In the gut, the high levels (micromolar to millimolar) of bile acids can regulate intestinal barrier function and immunity by activation of the farnesoid X receptor (FXR)⁵ and Takeda G protein-coupled receptor 5 (TGR5)⁶, suppression of NLRP3 inflammasome activity⁷ or modulation of T_H17 and T_reg cell homeostasis^8-10.

In addition to regulating host immunity, bile acids exhibit antimicrobial properties and serve as signals to modulate pathogen virulence^11-14 as well as antibiotic resistance¹⁵. For example, secondary bile acids such as deoxycholic acid (DCA) and lithocholic acid (LCA) can inhibit Clostridium difficile growth and mediate colonization resistance in vivo¹⁶. Beyond inhibition of C. difficile growth, bile acids also can reversibly bind and neutralize the TcdB toxin¹³. Alternatively, LCA has been reported to induce chaining and biofilm formation of another Gram-positive pathogen, vancomycin-resistant Enterococcus¹⁷. Interestingly, isoallolithocholic acid levels are elevated in centenarians and exhibit antimicrobial effects against C. difficile and E. faecium¹⁸. In Gram-negative bacterial pathogens, such as Vibrio cholerae, the primary bile acid taurocholate (TCA) causes dimerization of transcription factor TcpP to active V. cholerae virulence cascade^{11, 12}. Moreover, bile acid stimulation of cholera toxin production in V. cholerae may also be mediated by targeting the transmembrane domain of ToxR and its downstream signaling¹⁴. In the context of this study, bile acids are also reported to inhibit the Salmonella virulence and attenuate infection^{19, 20} as well as induce antimicrobial resistance^{21, 22}. However, how specific bile acids inhibit Salmonella infection, their specific protein target(s) and mechanism(s) of action remain unclear.

Salmonella enterica are one of the leading causes of gastroenteritis (S. Typhimurium) as well as typhoid fever (S. Typhi), a systemic disease that kills an estimated ~200,000 people/year world-wide²³. Following oral ingestion, Salmonella can invade intestinal epithelial cells to compromise the gut epithelium and disseminate into host tissues to cause disease²⁴. Genetic and biochemical studies have shown that the Salmonella contain two major pathogenicity islands, Salmonella Pathogenicity Island 1 (SPI-1) and 2 (SPI-2), which are essential for bacterial invasion and dissemination, respectively²⁴. Notably, SPI-1 and 2 encode type III secretion systems (T3SSs) as well as secreted protein effectors that are injected into host cells for Salmonella invasion, replication and dissemination^{24, 25}. As T3SSs are also present in many Gram-negative bacteria and required for their pathogenesis²⁵, understanding how bile acids from microbiota inhibit these key virulence pathways in Salmonella may yield key protein targets and facilitate the design of new anti-infectives.

To investigate the bile acid mechanism(s) of action in Salmonella, we evaluated the activity of different bile acids on virulence gene expression and invasion of epithelial cells and then characterized the protein target(s) of the most anti-infective bile acid chenodeoxycholic acid (CDCA, 1) using chemical proteomics. Amongst the diverse proteins recovered by CDCA-specific photoaffinity labeling and chemical proteomics, we focused on HilD, an AraC/XylS-family transcriptional regulator of SPI-1 gene expression. Our subsequent biochemical studies and identification of bile/CDCA-resistant HilD mutant (Q39E, N44D, H95L), suggested microbiota-generated bile acids and other metabolites can directly interact and inhibit HilD dimerization and DNA-binding to suppress S. Typhimurium virulence. Indeed, the generation of this HilD mutant S. Typhimurium strain by CRISPR–Cas9 gene editing revealed that the hilD^{Q39E, N44D, H95L} strain is more virulent in conventional microbiota-sufficient but not in antibiotic-treated mouse models of infection, highlighting HilD as a key transcriptional regulator of Salmonella-microbiota interactions. Collectively, our results showcase the utility of chemical proteomics for characterizing microbiota-metabolite protein targets and revealed HilD as a key protein target of anti-infective microbiota metabolites in S. Typhimurium.

Results

Bile acids inhibit S. Typhimurium virulence.

To investigate the protein targets and mechanisms of bile acids in bacterial pathogens, we evaluated the activity of specific bile acids (Fig. 1a) on Salmonella virulence and infection. Previous studies demonstrated that 3% bile inhibited Salmonella invasion to HeLa cells¹⁹ and individual bile acids such as DCA can inhibit expression of sipC::lacZY of Salmonella with an IC₅₀ of 7.6 μM²⁰. However, the effect of individual bile acids on Salmonella invasion of epithelial cells has not been systematically evaluated. To investigate the activity of individual bile acids on Salmonella virulence, we evaluated different bile acids, including conjugated and microbiota-derived free bile acids (Fig. 1a, Supplementary Fig. 1) on S. Typhimurium (strain 14028s) infection of HT-29 cells, a human colorectal adenocarcinoma epithelial cell line. Amongst all the bile acids tested, we found CDCA at 500 μM exhibited significant inhibition of HT-29 cell line invasion (Fig. 1b). In addition, we monitored the activity of bile acids on HilA protein expression, as a readout of SPI-1 gene and protein expression, using a chromosomally HA-tagged HilA in S. Typhimurium strain (hilA-HA) that was previously generated²⁶. Consistent with the invasion assay, 3% bile acids and CDCA (500 μM) inhibited HilA-HA expression (Fig. 1c), in a dose-dependent manner (Fig. 1d, e). Gene expression analysis by qRT-PCR showed that hilA transcription was inhibited along with SPI-1 effector sipA, but expression of ftsZ, a key regulator of cell division, was not significantly affected (Fig. 1f). These results demonstrated that amongst the diverse bile acids we tested, CDCA most effectively inhibited Salmonella virulence gene expression and infection of epithelial cells.

Figure 1. Chenodeoxycholic acid most effectively inhibits S. Typhimurium virulence in vitro.

a, Chemical structure of free bile acids and conjugated bile acids. b, Gentamicin protection assay of S. Typhimurium grown with bile acids (concentration = 0.5 mM, 100 mM HEPES with pH 8.0 for free bile acids) infecting HT-29 cells at MOI = 10 (biologically independent samples, DMSO n = 10, DCA n = 7, others n = 6). c, d, e, SPI-1 transcriptional factor HilA-HA expression of S. Typhimurium (hilA-HA) grown with individual bile acids (c, concentration = 0.5 mM, 100 mM HEPES with pH 8.0 for free bile acids, biologically independent samples, DMSO n = 12, others n = 6), bile acid mixture (d, 0%-3%, biologically independent samples, no bile acids n = 6, others n = 3) and CDCA (e, 0.125 mM-0.5 mM, biologically independent samples, n = 6). Western blot was quantified by grayscale analysis. Statistical analysis for b, c, e, all bile acids were compared to DMSO with one-way ANOVA followed by Dunnett's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. f, Expression of SPI-1 and control genes of S. Typhimurium grown with DMSO, CDCA (concentrate on = 0.5 mM) measured by qRT-PCR (biologically independent samples, n = 6, results are pooled from two independent experiments). Unpaired t test (two-sided). Centerline, average. Error bar, SD. Comparisons with P > 0.05 were not considered significant.

Analysis of bile acid-interacting proteins in S. Typhimurium.

To elucidate the protein targets of CDCA, we synthesized several bile acid photoaffinity reporters and employed chemical proteomics (Supplementary Fig. 2) to identify key interacting proteins in S. Typhimurium (14028s). Based on the anti-infective activity of CDCA (Fig. 1a, b), alk-X-CDCA (2) was synthesized, which contains a diazirine photocrosslinker at the 3 position and alkyne tag for bioorthogonal detection (Fig. 2a). Two additional bile acid photoaffinity reporters, alk-X-LCA (3) and alk-X-UDCA (4), were synthesized to serve as potential negative controls. (Fig. 2a, Supplementary Fig 3). These three photoaffinity reporters were then evaluated in S. Typhimurium infection and SPI-1 protein expression (HilA-HA) assays (Fig. 2b, c). Alk-X-CDCA exhibited similar activity to CDCA, while alk-X-LCA and alk-X-UDCA were relatively inactive in S. Typhimurium virulence assays at the same concentration. In cell UV-crosslinking, bioorthogonal labeling with azide-rhodamine dye and in-gel fluorescence profiling indicated that all the bile acid photoaffinity reporters interact with numerous proteins in Salmonella in a UV-dependent manner (Fig. 2d). Label-free quantitative (LFQ) proteomic analysis of bile acid reporter-crosslinked proteins after biotinylation and streptavidin affinity enrichment recovered 200-300 Salmonella protein hits for each individual bile acid reporter (Extended Data Fig. 1a-d). Further analysis revealed 129 proteins are shared by all three reporters (Fig. 2e), which include extracellular or secreted proteins (SipD, SipC, SopE2 and others), T3SS components (InvA, OrgA, SpaO, PrgH) and motility-related proteins (FilD, FlgK, FlgL, FliC), as predicted by Gene Ontology analysis. In addition, cytoplasmic and membrane proteins including fatty acid metabolic enzymes (FabD, FabF, FabI, AccC) (Fig. 2f) are also identified (Extended Data Fig. 1e). Interestingly, 97 protein hits were exclusively crosslinked by alk-X-CDCA (Fig. 2e, f), which more effectively inhibited Salmonella infection ex vivo (Fig. 2b, c), including several virulence-associated SPI-1 proteins such as OrgC, PrgK, HilD and InvC (Fig. 2f). As shown in the volcano plot (Fig. 2g), these proteins are enriched in a UV-dependent manner, which suggests the protein interactions between alk-X-CDCA are reversible and non-covalent. Collectively, these data suggest bile acids can interact with many proteins in Salmonella, including key proteins involved in bacterial virulence.

Figure 2. Chemoproteomic profiling of chenodeoxycholic acid-interacting proteins in S. Typhimurium.

a, Chemical structure of CDCA and bile acid reporters. b, Gentamicin protection assay of S. Typhimurium grown with bile acids reporters (concentration = 0.5 mM) infecting HT-29 cells at MOI = 10 (biologically independent samples, n = 6). One-way ANOVA followed by Tukey's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD; Comparisons with P > 0.05 were not considered significant. c, SPI-1 transcriptional factor HilA-HA expression of S. Typhimurium (hilA-HA) grown with bile acid reporters (concentration = 0.5 mM, biologically independent samples, DMSO n = 9, others n = 6). Western blot was quantified by grayscale analysis, normalized with GroEL expression. One-way ANOVA followed by Tukey's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. d, Bile acid reporters-treated S. Typhimurium cell lysates were reacted with azide-rhodamine and proteins were separated by SDS-PAGE and visualized by in-gel fluorescence. (Representative gel from 4 biological replicates) e, Venn diagram for the mass spectrometry datasets of three bile acid reporters. The significant protein targets compared to DMSO control under UV condition for each reporter were used for analysis. f, Log₂(LFQ intensity) of representative protein targets from group I and II of e (Log₂(LFQ intensity = 0) was replaced with 0). g, LFQ proteomic analysis of alk-X-CDCA-labeled proteins in S. Typhimurium (biologically independent samples, n = 4). Volcano plot (FDR = 0.05, s0 = 1). Pathogenesis proteins are colored by blue (Group I in e) and red (Group II in e).

Chenodeoxycholic acid binds virulence regulator HilD.

Amongst the bile acid (alk-X-CDCA)-interacting proteins we recovered (Fig. 2e, Group II), we focused on analysis of HilD, which was originally characterized in a group of genes involved in Salmonella hyper-invasion and a key transcriptional regulator of the SPI-1 gene expression^{27, 28} Based on the HilD protein sequence, it is predicted to be an AraC/XylS-like transcriptional regulator²⁹ that forms a homodimer as well as heterodimerizes with other AraC family proteins RtsA and HilC³⁰. The HilD C-terminus is predicted to bind DNA and regulate gene expression in Salmonella, whereas the N-terminus is predicted to contain a “jelly roll” motif (residue 29-104), which is found in other AraC/XylS-family transcriptional regulators such as ToxT from Vibrio cholerae³¹ (Supplementary Fig. 4). Long-chain fatty acids (LCFAs) have been reported to bind the “jelly roll” motif of ToxT and modulate its function³¹. Moreover, structural studies of another AraC/XylS-family protein Rns from E. coli demonstrated that Rns forms a homodimer and interacts with LCFAs at residues H20 and R75 in the regulatory domain³². Interestingly, dietary and microbiota metabolites such as LCFAs^33-35, SCFAs^{36, 37} as well as small molecules like myricanol³⁸ have also been reported to target HilD, which suggests AraC/XylS-family proteins may be broadly targeted by hydrophobic metabolites. To study the interaction of bile acids with HilD, we generated a chromosomal HA-epitope tagged HilD S. Typhimurium strain (hilD-HA) via CRISPR-Cas9 gene editing, which we previously established in Salmonella²⁶. Indeed, alk-X-CDCA efficiently photocrosslinked HilD-HA in S. Typhimurium in a UV-dependent manner by anti-HA antibody immunoprecipitation (Fig. 3a) or after biotinylation and streptavidin capture (Fig. 3b). Furthermore, alk-X-CDCA photocrosslinked HilD-HA in a dose-dependent manner (Fig. 3c). Consistent with proteomics data, HilD-HA was most efficiently photocrosslinked by alk-X-CDCA compared to the other two negative controls (Fig. 3d) and the labeling can be competed away by excess CDCA but not inactive bile acids such as cholic acid (CA) (Supplementary Fig. 5). These results suggest CDCA directly binds to HilD and may provide a biochemical mechanism for modulating the expression of SPI-1 virulence genes and S. Typhimurium infection.

Figure 3. HilD mutants confer resistance to bile and chenodeoxycholic acid inhibition of S. Typhimurium virulence.

a, c, d, S. Typhimurium (hilD-HA) were treated with bile acid reporters for 1 h and irradiated with 365 nm light. HilD-HA was immunoprecipitated from cell lysates and was reacted with az-rho, followed by SDS-PAGE in gel fluorescence scanning, and anti-HA immunoblotting. b, S. Typhimurium (hilD-HA) were treated with 0.5 mM of alk-X-CDCA for 1 h, followed by 365 nm light irradiation. The labeled proteins were reacted with az-biotin and enriched with streptavidin beads (“pull down”). Equal loading is validated by immunoblotting of input samples before enrichment (“Input”). a, b, d were repeated at least two times with similar results. c, dose dependent labeling was not repeated. e, Predicted structure of HilD dimer by Robetta. “Jelly roll” motif (putative binding pocket), DNA binding domain and N-terminus are highlighted by corresponding colors. f, Salmonella strains (hilA’-lacZ) expressing WT or single point mutated pBAD-hilD were treated with 2% bile acids for 4 h. Expression of hilA’-lacZ was measured by β-Galactosidase assay (biologically independent samples, n = 3). Expression percentage of bile acids treated bacteria was calculated against untreated bacteria. Centerline, average. Error bar, SD. g, HilA-HA expression of hilD^WT or hilD^mut S. Typhimurium (hilA-HA) strains grown with or without 2% bile acids. Western blot was quantified by grayscale analysis (biologically independent samples, WT strain n = 8, other n = 3). h, Gentamicin protection assay of S. Typhimurium grown with 2% bile acids infecting HT-29 cells at MOI = 10 (biologically independent samples, n = 6). Statistical analysis for g, h, two-way ANOVA followed by Sidak's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. Comparisons with P > 0.05 were not considered significant.

Identification of bile acid-resistant HilD mutants.

The alk-X-CDCA photocrosslinking suggests that HilD may be one of the key protein targets for bile acid (CDCA) inhibition of S. Typhimurium virulence gene expression (Fig. 2). To validate the significance of alk-X-CDCA-HilD photocrosslinking, we investigated potential bile acid-resistant HilD mutants. Based on previously published ToxT-LCFA binding models³¹, we generated several chromosomal hilD mutant strains by CRISPR-Cas9 gene editing at positions Asp 260, Leu 264 and Arg 267, which are adjacent to the “jelly roll” motif and have potential to bind with the carboxylic acid group of CDCA (Extended Data Fig. 2a). However, these HilD mutants are not resistant to CDCA in the SPI-1 protein expression assay (Extended Data Fig. 2b). Moreover, the R267A and the triple mutant (N260A, K264A and R267A) in the absence of bile acids exhibited significantly decreased SPI-1 expression, possibly due to disrupting the HilD structure and function (Extended Data Fig. 2b). We therefore employed error prone PCR to generate a HilD mutant library and screened their activity using the SPI-1 (hilA)-lacZ reporter strain in the presence and absence of 2.5 % bile acids (Extended Data Fig. 2c). Using this screening assay, we identified several bile acid-resistant HilD mutants after sequencing (Fig. 3e). We then generated individual HilD mutants that were further validated by the β-galactosidase assay (Fig. 3f). Several of these single HilD mutations are partially resistant to bile acids (Fig. 3f). These mutations are generally located in two regions of HilD, the N terminus and “jelly roll” motif, the putative small molecule binding pocket (Fig. 3e).

We next introduced these mutations into chromosomal hilD in S. Typhimurium by CRISPR-Cas9 gene editing and reevaluated their activity against bile acids (Fig. 3g, Extended Data Fig. 3). Indeed, these mutations exhibited significant resistance to 2% bile acids (Fig. 3g). The N-terminal double mutant (R13H, A16E) strain was partially resistant to bile acids, while the triple and quadruple mutated strains exhibited nearly full resistance to bile acids (Fig. 3g). For the “jelly roll” motif binding pocket mutations, single and double (Q39E, N44D) mutated strains were only partially resistant to bile acids, while the triple mutant (Q39E, N44D, H95L) exhibited the most significant bile acid resistance (Fig. 3g). These mutant strains also share similar resistance towards CDCA (Extended Data Fig. 3a). We then chose three representative S. Typhimurium HilD mutant strains HilD^{N3K,V7A,R13H,A16E} (N-terminal mutation), HilD^N44D and HilD^{Q39E,N44D,H95L} (putative binding pocket mutation) and assayed invasion of HT-29 epithelial cells in the presence and absence of 2% bile acids (Fig. 3h). Among these three strains, the S. Typhimurium HilD triple mutant strain (Q39E, N44D, H95L) was most resistant to bile acids in HT-29 infection assay (Fig. 3h).

The HilD-CDCA docking study suggests these three residues are around the CDCA binding pocket (Extended Data Fig. 3b, c). Of note, Gln39 and Asn44, which bear polar uncharged side chains are mutated to Glu39 and Asp44 bearing negative charged side chains, which might interrupt the interaction of HilD with CDCA, which has a carboxylic functional group. The HilD^H95L mutation within the “jelly roll” motif might also act in a similar manner. These results suggest HilD is an important protein target for inhibition of S. Typhimurium virulence by bile acids, which may bind in the “jelly roll” motif of HilD (Extended Data Fig. 3b, c). Intriguingly, it was reported that the hilD^N44A Salmonella strain, which express mutant HilD protein on a low copy-number hilD^N44A plasmid, showed significant resistance to long chain fatty acid, such as cis-2-hexadecenoic acid (HDA)³⁴.

To investigate the function of residue Asn44 more systematically, we knocked in other hilD mutations by CRISPR-Cas9 system, including N44A, N44K and N44F. We found that the N44A mutant strain showed similar resistance to 2% bile acids with N44D in both SPI-1 expression assay and Salmonella invasion assay (Extended Data Fig. 4a, b). In addition, we tested the HilD N44D mutant strain against HDA and found it showed partial resistance similar to N44A, but the triple mutant strain (Q39E, N44D, H95L) was the most resistant (Extended Data Fig. 4c, d).

CDCA interferes with HilD dimerization and DNA binding.

To characterize direct bile acid-HilD interactions, we expressed the recombinant protein using an intein-based protein expression strategy, akin to ToxT³¹, and purified the protein by size exclusion column (SEC) for in vitro assays (Supplementary Fig. 6). Analysis of the recombinant HilD protein constructs showed that alk-X-CDCA photocrosslinked HilD^WT more effectively than the HilD^{Q39E,N44D,H95L} mutant (Fig. 4a, b). It has been suggested that small molecules are able to affect AraC/XylS-like proteins by inhibiting protein dimerization³⁹, protein-DNA binding^{31, 40} and/or effecting protein stability^{20, 34, 36}. We evaluated HilD stability and found that CDCA had no major effect on protein turnover, with or without the expression of Lon protease (Extended Data Fig. 5a-d). However, CDCA administration did decrease the protein expression levels of HilD-HA (Extended Data Fig. 5e-h) as well as hilD transcript levels (Extended Data Fig. 5i), suggesting anti-infective bile acids may directly affect HilD-regulated gene expression in S. Typhimurium.

Figure 4. Chenodeoxycholic acid interferes with HilD dimerization and DNA binding.

a, Purified HilD^WT/mut proteins (10 μM) were treated with alk-X-CDCA (125 μM) for 30 min and irradiated with 365 nm light for 5 min on ice. Reaction mixture was reacted with az-rho, and precipitated by MeOH, followed by SDS-PAGE in gel fluorescence scanning, and coomassie blue staining. b, Quantification of labeling efficiency in a by grayscale analysis. The ratios were normalized to HilD^WT. (biologically independent samples, n = 6) Mann-Whitney test (two-sided). Centerline, average. Error bar, SD. c, EMSA of hilC promoter 5’-biotin-labelled PCR product (0.1 nM) in presence of 200 nM HilD^WT with increasing concentration of CDCA and CA. e. EMSA of hilC promoter 5’-biotin-labelled PCR product (0.1 nM) in presence of 200 nM HilD (WT or mut) with increasing concentration of CDCA. d, f, Quantification of inhibition of HilD-DNA interaction in c, e by grayscale analysis. Inhibition of HilD-DNA interaction was analyzed by free-DNA / total-DNA. (biologically independent samples, n = 4) Unpaired t test (two-sided). Centerline, average. Error bar, SD. Comparisons with P > 0.05 were not considered significant. g. BS³-mediated HilD (HilD^WT and HilD^mut) crosslinking in presence of DMSO or CDCA (400 μM). i-l, Mass distributions of HilD monomer and dimer for HilD^WT and HilD^mut with or without 100 μM of CDCA by mass photometry (HilD_{total monomer} = 10 nM). h, m, Quantification of dimerization ratio by grayscale analysis (h, n = 8) or mass photometry (m, HilD^WT and HilD^mut n = 8, HilD^WT and HilD^mut with CDCA n = 6). unpaired t test (two-sided). Centerline, average. Error bar, SD. ** P < 0.01, *** P < 0.001, **** P < 0.0001. For h, HilD^WT (DMSO vs CDCA), P < 0.0001; HilD^{Q39E+N44D+H95L} (DMSO vs CDCA), P = 0.0052; HilD^WT vs HilD^{Q39E+N44D+H95L} (DMSO), P = 0.0002.

We then investigated whether CDCA directly affected HilD binding to DNA by electrophoresis mobility shift assay (EMSA). HilD specifically bound to a 5’-biotin-labeled hilC promoter fragment with a dose dependent manner³³, but not the open reading frame (ORF) fragment of hilC, a reported HilD target gene (Supplementary Fig. 7a, b). Notably, CDCA inhibited HilD and DNA binding in a dose-dependent manner (Fig. 4c, d). However, biotin-EBNA DNA and Epstein-Barr Nuclear Antigen extract binding was not affected with the same concentration of CDCA (Supplementary Fig. 7c). CA, which shows weaker inhibitory activity on SPI-1 protein expression (Fig. 1c) and HilD binding ability (Supplementary Fig 5), exhibited less inhibition of HilD and DNA binding (Fig. 4c, d). Moreover, the analysis of the HilD^{Q39E,N44D,H95L} mutant demonstrated that this bile acid-resistant mutant is less sensitive to CDCA (Fig. 4e, f). A previous study suggested small molecules inhibited ToxT and DNA binding either by changing the conformation of protein DNA binding domain³¹ or inhibiting protein dimerization. Indeed, CDCA decreased HilD dimer ratio via the bis(sulfosuccinimidyl)suberate (BS³) crosslinking assay in vitro (Fig. 4g, h). Moreover, the HilD^{Q39E,N44D,H95L} mutant exhibited higher levels of dimerization compared to HilD^WT and was also more resistant to CDCA (Fig. 4g, h).

We also quantified HilD dimerization levels by mass photometry. At the concentration of 10 nM, we observed two peaks for both HilD^WT and HilD^{Q39E,N44D,H95L} mutant in mass photometry, which represent HilD monomer and dimer respectively (Supplementary Fig. 8). Consistent with the BS³-mediated crosslinking assay, we found the HilD^{Q39E,N44D,H95L} mutant is more dimeric (Fig 4i, j, m). When treated with CDCA, the dimer ratio decreased for WT HilD, but the HilD^{Q39E,N44D,H95L} mutant was more resistant to CDCA (Fig 4k, l, m, Supplementary Fig. 9). AraC/XylS-family proteins can interact with diverse metabolites. In fact, HilD has recently been reported to sense or interact with other dietary or microbiota metabolites including long-(oleate) and short-chain fatty acids^{33, 34, 36}. Indeed, the HilD^{Q39E,N44D,H95L} mutant is also resistant to suppression of SPI-1 genes by these other dietary and microbiota metabolites (Extended Data Fig. 6), suggesting the small molecule binding pocket in the predicted “jelly roll” motif may be able to accommodate other ligands, which is consistent with the results from other AraC/XylS-family proteins^{31, 39, 40}. Our results suggest anti-infective bile acids (i.e. CDCA) and other anti-infective metabolites may bind to HilD and affect its structure as proposed in the ToxT-LCFA interaction⁴⁰, which interferes with HilD dimerization³⁹, DNA binding and regulation of SPI-1 gene expression (Supplementary Fig. 10).

HilD mutant is resistant to microbiota inhibition in vivo.

Our results suggest that S. Typhimurium HilD^{Q39E,N44D,H95L} strain may be more resistant to microbiota suppression of infection and more virulent in vivo than the WT S. Typhimurium. To investigate the significance of microbiota metabolite inhibition of HilD function, we evaluated the activity of WT and the HilD^{Q39E,N44D,H95L} mutant S. Typhimurium strains in mice. In alignment with previous studies⁴¹, the ΔhilD strain demonstrated less pathogenesis compared to S. Typhimurium (14028s) in conventional specific pathogen-free (SPF) C57BL/6J mice (Supplementary Fig. 11). Oral infection of microbiota sufficient SPF-C57BL/6J mice showed that the HilD^{Q39E,N44D,H95L} mutant strain induced greater weight loss, mortality and inflammation compared to WT S. Typhimurium (Fig. 5a, b, Supplementary Fig. 12), suggesting that this HilD mutant strain is resistant to gut microbiota-mediated suppression of S. Typhimurium virulence. Notably, the S. Typhimurium levels in the feces of these mice were similar between HilD^WT and HilD^{Q39E,N44D,H95L} mutant strains (Fig. 5c), indicating the differences in pathogenesis are unlikely due to S. Typhimurium colonization levels in vivo. Previous studies have shown that antibiotic treatment of mice with streptomycin depletes over 90% of their gut microbiota, increases the levels of inactive conjugated bile acids and significantly decreases amounts of the anti-infective metabolites (such as SCFAs, LCFAs and free bile acids), which renders mice more susceptible to S. Typhimurium infection^{34, 42, 43}. Indeed, our analysis of streptomycin-treated mice orally-infected with WT S. Typhimurium and the HilD^{Q39E,N44D,H95L} mutant strain exhibited similar levels of pathogenesis and pathogen load in the feces (Fig. 5d-f), suggesting HilD may be a key target of anti-infective microbiota species and metabolites. Indeed, we found that fecal metabolites from gut-microbiota sufficient mice inhibit the virulence of WT Salmonella but not the HilD^{Q39E,N44D,H95L} mutant strain ex vivo (Extended Data Fig. 7). Taken together, these results suggest that microbiota-derived metabolites such as the free bile acids and others (such as LCFAs) may contribute to Salmonella virulence inhibition via HilD in vivo.

Figure 5. S. Typhimurium HilD^{Q39E,N44D,H95L} mutant is resistant to microbiota-suppression of virulence in vivo.

a-c, C57BL/6 mice were orally infected with 10⁹ CFU of S. Typhimurium (n = 12). The survival (a, d), mouse weight (b, e), and Salmonella CFU in feces (c, f) were monitored daily. For d-f, the mice were orally treated with 20 mg of streptomycin 24 hour before infection (n = 8). For statistical analysis, a, d, Gehan-Breslow-Wilcoxon test; a, P = 0.0014. b, e, c, f, Two-way ANOVA followed by Sidak's multiple comparisons test, adjusted P value. For b, e, Comparisons were conducted on day 4. b, P < 0.0001. Centerline, average. Error bar, SD. Comparisons with P > 0.05 were not considered significant.

HilD mutant is resistant to CDCA inhibition in vivo.

To further directly investigate the bile acid-HilD interaction in vivo, we performed CDCA dietary supplementation studies. For these studies, streptomycin-treated C57BL/6J mice were supplemented with or without 0.2% CDCA in their diet and infected with HilD^WT or HilD^{Q39E,N44D,H95L} mutant S. Typhimurium strains. We found that the HilD^{Q39E,N44D,H95L} mutant strain is more virulent compared to the WT strain in CDCA dietary supplementation model by percent survival and weight loss analysis (Fig. 6a-c). In contrast, no significant difference was observed in streptomycin-pretreated mice without CDCA supplementation (Fig. 6d-f). Previous work demonstrates that supplementary CDCA can elicit an intestinal antimicrobial program and attenuates Salmonella pathogenesis⁴⁴. Our results demonstrates that beyond modulating host immunity, CDCA may also directly attenuates Salmonella infection by inhibiting HilD in vivo. These results show that in addition to other microbiota metabolites^33-36, anti-infective bile acids such as CDCA may target HilD to attenuate S. Typhimurium virulence in vivo.

Figure 6. S. Typhimurium HilD^{Q39E,N44D,H95L} mutant is more virulent in CDCA-supplemented mice.

C57BL/6 mice were fed with or without 0.2% CDCA supplements (a-c, CDCA is supplemented (n = 15); d-f, CDCA is not supplemented (n = 12)). The mice were orally treated with 20 mg of streptomycin. 24-hour post streptomycin treatment, mice were orally infected with 10⁴ CFU of S. Typhimurium (14028s) (hilA-HA, hilD^WT or hilD^{Q39E+N44D+H95}). The survival (a, d), mouse weight (b, e), and Salmonella CFU in feces (c, f) were monitored. Average mouse weight curves were censored after the first mouse in each group reached 80% of baseline weight. Results are pooled from three independent experiments. For analysis, a, d, Log-rank (Mantel-Cox) test, a, P = 0.0204. b, Two-way ANOVA followed by Sidak's multiple comparisons test, adjusted P value. Comparisons were conducted on day 7, P = 0.029, Centerline, average. Error bar, SD. e, Two-way ANOVA followed by Sidak's multiple comparisons test, adjusted P value. Comparisons were conducted on day 5, Centerline, average. Error bar, SD. Comparisons with P > 0.05 were not considered significant. Dotted lines indicated the limit of detection.

Discussion

Intestinal microbiota provides an important barrier to enteric pathogen infection in humans and animals, but the precise mechanisms of specific bacterial species and their metabolites have been unclear and beginning to emerge¹. For example, our previous studies have shown that Enterococcus peptidoglycan remodeling and muropeptides enhance intestinal barrier function via nucleotide-binding oligomerization domain-containing protein 2 (NOD2) activation and limit S. Typhimurium and Clostridium difficile pathogenesis⁴⁵. Alternatively, commensal Enterobacteriaceae can use aerobic metabolism to compete with S. Typhimurium growth⁴⁶ or sequester essential metals⁴⁷ in vivo. Moreover, SCFAs from other microbiota species can suppress S. Typhimurium growth and pathogenesis⁴⁸. Using chemical proteomics, we previously demonstrated that site-specific acylation of another key transcriptional regulator HilA provides a mechanism by which SCFAs can attenuate S. Typhimurium virulence²⁶. In addition, microbiota species can produce significant amounts (μM – mM) of bile acids that are reported to regulate both innate and adaptive immunity in mammals as well as attenuate the growth and/or virulence mechanisms of enteric pathogens such as S. Typhimurium^{19, 20}. CDCA for example has been reported to enhance host intestinal barrier function to limit infection⁴⁴. However, the molecular mechanism(s) by which bile acids directly inhibit the virulence of S. Typhimurium was unclear.

To investigate the anti-infective mechanism(s) of bile acids, we evaluated a panel of different bile acids, including both conjugated bile acids and microbiota-derived free bile acids, on S. Typhimurium virulence assays ex vivo. Amongst all the bile acids we tested, CDCA displayed the most anti-infective activity. The activity variation of bile acids might arise from the different solubility and permeability as well as protein target engagement. Indeed, our proteomics analysis of both active and inactive bile acid photoaffinity reporters in S. Typhimurium enabled us to identify HilD and other virulence associated proteins, such as OrgC, PrgK and InvC. In this study, we found that inhibition of Salmonella invasion to HT-29 cells is correlated with the inhibition of SPI-1 virulence gene expression level by anti-infective bile acids (Fig. 1b, c). We therefore focused on the anti-infective activity and interaction of CDCA with HilD, a key transcriptional regulator of SPI-1 genes in S. Typhimurium.

To dissect the CDCA-HilD interaction, we identified and characterized bile acid-resistant HilD mutants and identified key three mutations, Q39E, N44D and H95L. These three residues are located around the HilD “jelly roll” motif and potential ligand binding pocket. Our subsequent analysis of recombinant HilD protein constructs in vitro revealed that the Q39E, N44D, H95L triple mutant was more resistant to CDCA inhibition of HilD protein homodimerization and DNA binding as well as suppression SPI-1 gene expression and invasion of epithelial cells ex vivo. It is possible that CDCA inhibition of HilD homodimerization and DNA binding may also perturb its protein-protein interactions with other factors (such as HilE, HilC and RtsA)^{30, 49}. Additional biochemical and high-resolution structural studies will be needed explore these possibilities.

Importantly, the HilD^{Q39E,N44D,H95L} S. Typhimurium mutant was also resistant to other anti-infective microbiota metabolites ex vivo and more virulent in microbiota-sufficient and CDCA-supplemented mice, which indicates HilD may be one of the key targets of anti-infective microbiota and dietary metabolites in S. Typhimurium in vivo. Previous studies found that LCFAs from colon and caecum of mice inhibit Salmonella virulence ex vivo³⁴ as well as inhibit HilD function in Salmonella in vitro^33-35. Our CDCA supplement experiments suggest that in addition to LCFAs and other microbiota metabolites^33-36, anti-infective bile acids such as CDCA may also target HilD to regulate Salmonella virulence in vivo. It will be interesting to characterize specific microbiota species and pathways that can generate CDCA and other anti-infective bile acids^{12, 50}. In addition, our proteomic studies suggest anti-infective bile acids may also target other virulence associated proteins (OrgC, PrgK, InvC), which should be investigated in the future. Collectively, our studies demonstrate the utility of chemical proteomics for investigating the mechanisms of microbiota metabolites and demonstrate that bile acids may directly target key virulence factors in microbial pathogens, which highlights potential small molecule protein targets for anti-infective development.

Methods

Microbial Strains and Growth Conditions

All strains used are listed in Supplementary Table 1. All Salmonella Typhimurium strains used were derivatives of S. Typhimurium 14028S³⁶. Salmonella strains were cultured at 37 °C in liquid Miller Luria-Bertani (LB) medium [10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl] (Becton Dickinson, Difco™), SPI-1 inducing LB medium [10 g/L tryptone, 5 g/L yeast extract, 300 mM NaCl], or on Salmonella Shigella agar (Becton Dickinson). Cultures were grown at 37 °C in Multitron shaking incubator (INFORS HT) at 220 rpm.

Gentamicin protection assay of S. Typhimurium grown with bile acids

1:50 dilutions of overnight Miller LB cultures of Salmonella Typhimurium strain 14028s (WT or mutant) were grown in 4 mL SPI-1 inducing LB (with or without bile acids) for 4 h at 37 °C with 220 rpm shaking. Cultures were washed twice with cold PBS to remove bile acids and then re-suspended to cold PBS to measure OD 600. HT-29 cells (human colon epithelial cell line, ATCC, HTB-38™) were cultured in 12-well tissue culture plates about 90% of confluency. Wells were added with Salmonella cells at an MOI = 10:1 and centrifuged at 1000 g for 5 min at room temperature. Cells were incubated at 37 °C with 5% CO₂ for 2 h to allow invasion. The media was removed, and cells are washed with PBS (100 μg/mL gentamicin) three times. Then new DMEM media containing 100 μg/mL gentamicin are added and cells are incubated for an additional hour to kill extracellular Salmonella. Wells were then washed 3 times with PBS, and cells were lysed with 500 μl 1% Triton X-100 PBS. Lysates were serially diluted and drip-dropped on Salmonella Shigella agar plates (BD 211597) to determine the number of invaded bacteria.

Western blot analysis of SPI-1 expression

1:50 dilutions of overnight Miller LB cultures of Salmonella Typhimurium strain 14028s (hilA-HA) were grown in 4 mL SPI-1 inducing LB (with or without bile acids) for 4 h at 37 °C with 220 rpm shaking. Cultures were washed twice with cold PBS to remove bile acids and then re-suspended to cold PBS to measure OD 600. 600 μL of bacteria culture was spined down and the supernatant was removed. Bacteria pellet was resuspended in 100 μL of 1x Laemmli buffer and boiled for 10 min. The resulted samples were subject to western blot analysis. Western blot was quantified by grayscale analysis with ImageJ 1.53e., normalized with GroEL expression (1:3000, ab90522, abcam) or OD 600 value.

Salmonella Quantitative Reverse-Transcription PCR

1:50 dilutions of overnight Miller LB cultures of Salmonella Typhimurium strain 14028s were grown in 4 mL SPI-1 inducing LB (with or without bile acids) for 4 h at 37 °C with 220 rpm shaking. 500 μl of Salmonella cultures were processed with RNeasy Mini Kit (Qiagen) per manufacturer’s manual. Concentrations of purified RNA were normalized to 100 ng/μl with RNase-free water. Quantitative Reverse-Transcription PCR (qRT-PCR) were performed with Power SYBR Green RNA-to-C_T 1-Step Kit (Applied Biosystems) per manufacturer’s manual and primers listed in Supplementary Table 2. rpoS and rpoD genes are used as ΔCT calculating controls.

Preparation of Salmonella bacterial total cell lysates

1:50 dilutions of overnight Miller LB cultures of Salmonella Typhimurium strain 14028s were grown in 4 mL SPI-1 inducing LB for 4 h at 37 °C with 220 rpm shaking. For bile acid reporters labeling experiments, bacteria were firstly cultured for 3 h in LB and another 1 h with bile acid reporters in DMSO or DMSO as the negative control. S. Typhimurium cells were pelleted at 16000 g for 1 min, and pellets were washed with cold PBS once and re-suspended in 3 mL cold PBS. For in-cell photo-crosslinking experiments, bacteria in 3 mL cold PBS was transferred to petri dish (35 mm X 10 mm) and subjected to UV irradiation at 365 nm on ice for 5 min using a Spectrolinker XL-1000 UV crosslinker (Spectronics) at a distance of 3-5 cm. Bacteria were then harvested and centrifuged 5000 g at 4 °C for 10 min. Bacteria were lysed with 200 μl lysis buffer (phosphate-buffered saline (PBS) containing 0.5% Nonidet P-40, 1X EDTA-free protease inhibitor cocktail (Roche), 0.5 mg/mL lysozyme (in dH₂O) (Sigma), and 1:1,000 dilution of Benzonase (Millipore)). After re-suspension, pellets were sonicated for 30 s (10 s X 3), then were incubated on ice for 30 min. The bacteria were further sonicated for another 30 s (10 s X 3). Cell lysates were centrifuged at 8200 g for 5 min to remove cell debris and supernatants were collected. Protein concentration was estimated by BCA assay with BCA Protein Assay Kit (Thermo).

In-gel fluorescence analysis of bile acid reporter labeling

For in-gel fluorescence analysis of bile acid reporters-labeled Salmonella proteome, from the bile acid reporters-treated or control total cell lysates prepared as described above, 45 μl of each total cell lysates (~50 μg) was added with 5 μl of click chemistry reagents as a 10X master mix (az-Rho⁵³: 0.1 mM, 10 mM stock solution in DMSO; tris(2-carboxyethyl)phosphine hydrochloride (TCEP): 1 mM, 50 mM freshly prepared stock solution in dH₂O; tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA): (0.1 mM, 2 mM stock in 4:1 t-butanol: DMSO); CuSO₄ (1 mM, 50 mM freshly prepared stock in dH₂O)). Samples were mixed well and incubated at room temperature for 1 h. After incubation, samples were mixed with 200 μl cold methanol and incubate at −20 °C overnight. Sample proteins were precipitated at 18000 g for 15 min at 4 °C. After gently removing the aqueous layer, protein pellets were washed with 200 μl cold methanol, spinning down at 18000 g for 15 min at 4 °C, and liquid was gently decanted. After washing twice, pellets were allowed air-dried before boiling with 1X Laemmli buffer. Samples were boiled with 1X Laemmli buffer 95 °C for 5 min before being loaded onto a 4-20% Tris-HCl gel (Bio-Rad) for SDS-PAGE. In-gel fluorescence scanning was performed using a Typhoon 9400 imager (Amersham Biosciences) or ChemiDoc Imaging Systems and analyzed by Image Lab 6.1.

For in-gel fluorescence analysis of bile acid reporter-labeled HilD-HA, from the bile acid reporters-treated or control total cell lysates prepared as described above, 250 μg of each total cell lysates were immunoprecipitated with 20 μl PBS-T-washed Pierce™ Anti-HA Magnetic Beads (Thermo Scientific). Samples were mixed well and incubated at 4 °C for 2 h with end-to-end rotation. After samples were washed with 500 μl PBS-T 3 times, 45 μl of PBS-T (PBS with 0.1% Tween-20) was added to each sample. 5 μl of click chemistry reagents as a 10X master mix mentioned above were added to each sample. After incubation at room temperature for 1 h, samples were washed with 500 μl PBS-T 3 times. Samples were boiled with 2X Laemmli buffer at 95 °C for 5 min before being loaded onto a 4-20% Tris-HCl gel (Bio-Rad) for SDS-PAGE. In-gel fluorescence scanning was performed using a Typhoon 9400 imager (Amersham Biosciences) or ChemiDoc Imaging Systems. Proteins were then transferred onto 0.45 μm nitrocellulose membrane (Bio-Rad) with Trans-Blot Turbo Transfer System (Bio-Rad) at 25 V for 30 min. The membrane was blocked with 5% non-fat milk in PBS-T for 60 min, and anti-HA rabbit antibody ab9110 (1:5,000, abcam) or H6908 (1:2000, Sigma) was added to solution before incubating membrane at 4°C overnight. The membrane was washed with PBS-T 3 times, and incubated with 1:10,000 goat polyclonal anti-rabbit HRP (Abcam, ab97051) in PBS-T with 5% non-fat milk at room temperature for 1 h. The membrane was washed with PBS-T 3 times, and imaged with Clarity Western ECL substrate (Bio-Rad) and ChemiDoc XRS+ System (Bio-Rad).

Pull-down of endogenous HilD-HA by bile acid reporters

From the bile acid reporter-treated or control total cell lysates prepared as described above, 90 μl of each total cell lysates (~100 μg) was added with 10 μl of click chemistry reagents as a 10X master mix (Az-biotin: 0.1 mM, 10 mM stock solution in DMSO; tris(2-carboxyethyl)phosphine hydrochloride (TCEP): 1 mM, 50 mM freshly prepared stock solution in dH₂O; tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA): 0.1 mM, 2 mM stock in 4:1 t-butanol: DMSO; CuSO₄: 1 mM, 50 mM freshly prepared stock in dH₂O). Samples were mixed well and incubated at room temperature for 1 h. After incubation, samples were mixed with 400 μl cold methanol and incubate at −20 °C overnight. Sample proteins were precipitated at 18000 g for 15 min at 4 °C. After gently removing the aqueous layer, protein pellets were washed with 400 μl cold methanol, spinning down at 18000 g for 15 min at 4 °C, and liquid was gently decanted. After last wash, pellets were let air dried before being re-solubilized in 100 μl 4% SDS PBS with bath sonication (5 μl solution was kept for “input” before enrichment). Solutions were then diluted with 300 μl PBS, and incubated with 20 μl PBS-T-washed High Capacity NeutrAvidin agarose (Pierce) (500 μl PBS-T-washed twice, 2500gX60s) at room temperature for 1 h with end-to-end rotation. The agarose was washed with 500 μl 1% SDS PBS 3 times, 500 μl 1M Urea PBS 3 times, and 500 μl PBS 3 times. Samples were boiled with 2X Laemmli buffer 95 °C for 5 min before being loaded onto a 4-20% Tris-HCl gel (Bio-Rad) for SDS-PAGE. Equal loading is validated by immunoblotting of input samples before enrichment (“Input”).

Label-Free Quantitative proteomics of bile acid reporters in Salmonella

1:50 dilutions of overnight Miller LB cultures of Salmonella Typhimurium strain 14028s WT were grown in 20 mL SPI-1 inducing LB (300 mM NaCl) for 3 h at 37 °C with 220 rpm shaking. Cultures were incubated with 0.5 mM bile acid reporters in DMSO or DMSO as the negative control for another 1 h at 37 °C with 220 rpm shaking. The culture medium was centrifuged at 5000 g for 10 min. Bacteria pellets was re-suspended in 10 mL cold PBS. For in-cell photo-crosslinking experiments, bacteria in 10 mL cold PBS was transferred to petri dish (100 mm X 15 mm) and subjected to UV irradiation at 365 nm on ice for 5 min using a Spectrolinker XL-1000 UV crosslinker (Spectronics) at a distance of 3-5 cm. Bacteria were then harvested and centrifuged 5000 g at 4 °C for 10 min. After re-suspension in 1 mL lysis buffer, bacteria were sonicated for 15 sec with a Sonic Dismembrator Model 500 (Fisher Scientific) with 5 sec on and 10 sec off per cycle. Cell lysates were centrifuged at 16000 g for 20 min to remove cell debris and supernatants were collected. Each total cell lysates was added with 100 μl of click chemistry reagents as a 10X master mix (az-Biotin: 0.1 mM, 10 mM stock solution in DMSO; tris(2-carboxyethyl)phosphine hydrochloride (TCEP): 1 mM, 50 mM freshly prepared stock solution in dH₂O; tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA): 0.1 mM, 2 mM stock in 4:1 t-butanol: DMSO; CuSO₄: 1 mM, 50 mM freshly prepared stock in dH₂O). Samples were mixed well and incubated at room temperature for 1 h. After incubation, samples were mixed with 4 mL cold methanol and incubated at −20 °C overnight. Protein pellets were centrifuged at 5000 g for 30 min at 4°C, pellets were transferred to 2.0 mL centrifuge tube and were washed with 1 mL cold methanol 3 times. After last wash, pellets were let air dried before being re-solubilized in 250 μl 4% SDS PBS with bath sonication. Solutions were diluted with 750 μl PBS, and incubated with 60 μl PBS-T-washed High Capacity NeutrAvidin agarose (Pierce) (500 μl PBS-T-washed twice, 2500 g X 60 s) at room temperature for 1 h with end-to-end rotation. The agarose was washed with 500 μl 1% SDS PBS 3 times, 500 μl 4M Urea PBS 3 times, and 500 μl PBS 3 times and then reduced with 500 μl 10 mM DTT (Sigma) in PBS for 30 min at 37 °C, and alkylated with 500 μl 50 mM iodoacetamide (Sigma) in PBS for 30 min in dark. 50 μl NH₄HCO₃ (10 mM) was added to the tube. Neutravidin-bound proteins were digested on bead with 400 ng Trypsin/Lys-C mix (Promega) at 37 °C overnight with shaking. Digested peptides were collected (2500 g for 60 s) and lyophilized before being desalted with custom-made stage-tip containing Empore SPE Extraction Disk (3M). Peptides were eluted with 2% acetonitrile, 2% formic acid in dH₂O.

Peptide LC-MS analysis was performed with a reversed phase C18 column coupled to an Orbitrap mass spectrometer. MS spectra was measured at a resolution of 60,000. The 28 samples data were pooled and processed with MaxQuant version 2.1.0.0. The peptides were identified from the MS/MS spectra searched against the database (Salmonella Typhimurium 14028s, Proteome ID: UP000002695). Label free quantification experiments in MaxQuant were performed using the built-in label free quantification algorithm enabling the “Match between runs” option. Other parameters were used as preset in the software⁵⁴. The search results from MaxQuant were analyzed by Perseus 2.0.5.0. Briefly, the DMSO and bile acid reporters-labeled samples were grouped correspondingly. The results were cleaned to filter off reverse hits and contaminants. Only proteins that were identified in 3 out of 4 sample replicates and with more than two unique peptides were subjected to subsequent statistical analysis. LFQ intensities were used for measuring protein abundance and logarithmized. Signals that were originally zero were imputed with random numbers from a normal distribution, whose mean and standard deviation were chosen to best simulate low abundance values below the noise level (Replace missing values by normal distribution – Width = 0.3; Shift = 1.8). The summary of proteomics data is shown in the Supplementary Dataset.

Salmonella CRISPR–Cas9 genome editing

The Salmonella CRISPR-Cas9 genome editing method was modified from reference²⁶. Overnight culture of S. Typhimurium strains were diluted 1:50 to 20 ml fresh LB medium, and were grown at 37 °C in a shaking incubator at 220 rpm until the OD600 reached 0.5–0.7. Cells were pelleted at 5,000 g for 10 min at 4 °C, and washed twice with 10 ml ice-cold 10% glycerol. Cell pellets were resuspended in 100 μl 10% glycerol, and aliquoted as 50 μl per tube, then stored in −80 °C. Electrocompetent parent S. Typhimurium strains were transformed with pKD46 via electroporation with Gene Pulser II (Bio-Rad) at 2.5 kV and 25 μF in a 2-mm cuvette, and selected on carbenicillin agar plates at 30 °C overnight. The resulting S. Typhimurium pKD46 strains were made into electrocompetent cells (ice-cold 10% glycerol) after being grown at 30 °C with 0.2% arabinose and carbenicillin until the OD600 reached 0.5–0.7. S. Typhimurium pKD46 electrocompetent cells were transformed with 2 μl pCas9⁵⁵ (~100ng, see Supplementary Table 3 for spacer sequence) and 2 μg ssDNA editing template (Supplementary Table 4) and recovered at 37 °C for 2 h. Then resulting bacteria were selected on chloramphenicol agar plates at 37 °C overnight. Colonies on the plate were streaked onto chloramphenicol agar plates, and colonies from the plates were randomly picked for colony PCR to confirm editing (50%~100% efficiency), successfully edited colonies were further confirmed by sanger sequence. The pCas9 was cured by growing bacteria in plain LB agar.

Construction and screening of mutant hilD library

Mutant pTXB1-HilD library was generated with GeneMorph II EZClone Domain Mutagenesis Kit (Agilent Technologies) per manufacturer’s manual. Specifically, WT pTXB1-HilD (0.5/1 μg) was mutagenized for 30 cycles of PCR amplification using the primers pTXB1-M-F and pTXB1-M-R (Supplementary Table 2). The resulting PCR product (mutant hilD fragment) was purified by agarose gel. Then WT pTXB1 (template plasmid) and megaprimer (mutant hilD fragment) were subjected to EZClone reaction, followed by Dpn I-digestion of template plasmid. Dpn I-digested DNA was transferred to XL10-Gold ultracompetent cells to enlarge the mutant pTXB1-HilD library.

The plasmid library extracted from XL10-Gold cell is transformed to electrocompetent reporter strain (Salmonella Typhimurium 14028s, tetRA-hilD-3XFLAG, attλ::pDX1::hilA’-lacZ) via electroporation with Gene Pulser II (Bio-Rad) at 2.5 kV and 25 μF in a 2-mm cuvette, and the resulting bacteria were selected on screening agar plates (200 mg/L IPTG, 100 mg/L Carbenicillin, 100 mg/L 100 mg/L X-gal, 2.5 % bile acids mixture). The blue colonies were picked and steaked on the screening agars to confirm the resistance (blue color generation). The hilD fragment of plasmids from double-confirmed blue strains were sent for sequencing.

β-Galactosidase activity assay

The single amino acid mutated pBAD-hilD plasmids were generated via QuikChange XL site-directed mutagenesis Kit (Agilent Technologies) according to sequence of pBAD-hilD from blue colonies. Electrocompetent reporter strain (tetRA-hilD-3XFLAG attλ::pDX1::hilA’-lacZ) was transformed with mutant pBAD-hilD plasmids via electroporation with Gene Pulser II (Bio-Rad) at 2.5 kV and 25 μF in a 2-mm cuvette, and selected on carbenicillin agar plates at 37 °C overnight. Overnight culture of reporter strains with WT or mutant pBAD-hilD plasmids were diluted 1:50 to 4 mL LB medium (carbenicillin and arabinose), and were grown at 37 °C in a shaking incubator at 220 rpm for 4 h. β-Galactosidase activity assay was based on published methods with some modification. Briefly, cultures were incubated on ice for 10 min and bacteria was washed with Z buffer (60 mM Na₂HPO₄.7H₂O, 40 mM NaH₂PO₄.H₂O, 10 mM KCl, 1 mM MgSO₄, 50 mM β-mercaptoethanol, pH to 7.0) three times. OD 600 of cultures were measured, and cultures were diluted 1:5 with cold Z buffer and permeabilized by adding 100 μl chloroform and 50 μl 0.1% SDS, following by vortex and equilibrating in 28 °C for 5 min. Reaction was started by adding 0.2 mL substrate, o-nitrophenyl-β-D-galactoside (4 mg/mL) and incubated at 28 °C. The reaction was stopped after sufficient yellow color has developed by adding 0.5 mL 1M Na₂CO₃. 1 mL reaction mixture was transferred to a new tube and centrifuged at 18000 g for 5 min. OD 420 and OD 550 of supernatant was measured. Miller Units = 1000 x [(OD₄₂₀ - 1.75 x OD₅₅₀)] / (Reaction time x V x OD₆₀₀).

HilD expression and purification

HilD was cloned from Salmonella Typhimurium strain 14028s and ligated into pTXB1 vector (New England Biolabs). HilD-CBD (intein-chitin-binding domain) fusion protein was expressed in LB media using BL21-CodonPlus (DE3)-RIL (Agilent Technologies) E. coli. Briefly, overnight culture of strains with pTXB1-HilD-CBD was inoculated to LB media (25 μg/mL chloramphenicol and 200 μg/mL Ampicillin) at 37 °C. IPTG (100 mg/L) was added when OD 600 reached 0.6-0.8 to induce the protein expression. Protein was expressed at 25 °C in Multitron shaking incubator (INFORS HT) at 220 rpm for 16 h. Cells were harvested by centrifugation (5000 g for 15 min), washed with cold PBS. After re-suspension in column buffer (20 mM Tris pH 8.0, 1 mM EDTA, 500 mM NaCl, 1mM EDTA, 0.1% Triton X-100 and 40 μM PMSF), bacteria were sonicated for 5 min with a Sonic Dismembrator Model 500 (Fisher Scientific) with 5 sec on and 10 sec off per cycle. Cell lysates were centrifuged at 30000 g for 45 min to remove cell debris and supernatants were collected. The chitin-affinity column (New England Biolabs) was used per manufacturer’s manual to purify HilD-CBD fusion protein. HilD was cleaved from CBD on column by the addition of 100 mM DTT with cleavage buffer (20 mM Tris pH 8.0, 1 mM EDTA, 150 mM NaCl) and incubation for 16 h at 4 °C. The cleaved HilD protein was eluted with cleavage buffer. The eluted HilD was further purified by HiTrap SP Sepharose Fast Flow cation-exchange column (GE Healthcare) using a gradient from 100% buffer A (20 mM Tris pH 6.8, 50 mM NaCl, 10 mM DTT) to 100% buffer B (20 mM Tris pH 6.8, 1 M NaCl, 10 mM DTT), with the protein peak corresponding to HilD eluting at 45% buffer B. HilD fractions were collected and then concentrated using an Amicon ultracentrifugal filter unit (Millipore) with a molecular-weight cutoff of 10 kDa. The concentrated protein was subjected to a Superdex 200 Increase 10/300 GL column (GE Healthcare) pre-equilibrated with buffer (20 mM HEPES, 500 mM NaCl, 2 mM TCEP, pH 6.8). Fractions containing the target protein were combined and concentrated. Protein concentration was estimated by The Bradford Assay (Bio-Rad).

Electrophoretic mobility shift assays (EMSA)

A double stranded 5’-biotinylated DNA fragment of the hilC promoter region (corresponding to −162 to +46) or a nonspecific DNA fragment (corresponding to +60 to +269 of hilC) was amplified by 5’-biotinylated primers (Supplementary Table 2) from S. Typhimurium genomic DNA. EMSA were carried out using the LightShift Chemiluminescent EMSA Kit (Thermo Scientific™, 20148) following the manufacturer’s instructions. Briefly, purified HilD proteins or control protein were incubated with indicated bile acids or DMSO control at 4 °C. Then 5’-biotinylated DNA fragment was added and incubated for 30 min at room temperature. The mixture was then loaded on 5% Criteriontm TBE Polyacrylamide Gel (Bio-Rad) then transferred onto a positively charged nylon membrane (Thermo Scientific) and detected by chemiluminescence.

Chemical cross-linking of HilD in vitro

2.5 μM of Purified HilD proteins (WT or Q39E+N44D+H95L, 20 mM HEPES, 500 mM NaCl, 2 mM TCEP, pH 6.8) were incubated with indicated bile acids or DMSO control at 4 °C. The proteins were crosslinked by incubation with 0.6 mM BS³ (bis(sulfosuccinimidyl)suberate, Thermo Scientific Pierce, A39266) at room temperature for 1 h. Crosslinking was terminated by incubation with 50 mM Tris, pH 7.5 for 15 min. 1/3 volume of 4 X Laemmli buffer was added to reaction mixture and resulted mixture was boiling at 95 °C for 5 min, and analyzed by SDS-PAGE. The proteins were visualized by silver staining. dimer ratio = dimer / (monomer + dimer).

HilD analysis by mass photometry⁵⁶

HilD proteins (WT or Q39E+N44D+H95L) were purified by size exclusion column and diluted to 40 nM in DPBS (corning, 21-031-CV, with or without 100 μM of CDCA). Proteins treated with CDCA were incubated on ice for 4 hours before measurement. High precision microscope coverslips (No. 1.5H, 24 x 50 mm, 170±5 μM) were cleaned with repeated alternating washes of Milli-Q water and isopropanol, and dried using a clean stream of air. Clean coverslips were assembled with silicone gaskets (CultureWell™ reusable gasket, 3mm diameter x 1 mm depth, Grace Bio-Labs), and samples imaged using a Refeyn Two MP Mass Photometer (Refeyn.com). The instrument was focused against cold PBS buffer (15 μl). Protein samples (5 μl) were then added to the cold PBS (15 μl), mixed thoroughly, and imaged for 60 seconds at 300 frames per second(fps) using the Refeyn AcquireMP (v. 2.4.1) software suite. Reported concentrations reflect sample concentrations in the final 20 μl solution. Mass Photometry data was analyzed using Refeyn DiscoverMP software suite (v.2.4.3). Raw MP contrast values were converted to mass using a standard calibration of BSA (ThermoFisher, 23209) and Thyroglobulin (Millipore, 609310). Data displayed above gaussian fits represent apex molecular weight, standard deviation (σ), and the number of events within the gaussian fit.

Animal Experiments

Mus musculus C57BL/6J female mice of 7 weeks old were purchased from the Jackson Laboratory (000664) and maintained under SPF conditions. Housing conditions: mice were housed at 68-79 °F with 30%-70% humidity. The lights are on from 6 am to 6 pm. Mice 8 to 10 weeks of age were used for the study. Animal care and experimentation were consistent with the National Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee of Scripps Research.

S. Typhimurium infection of mice

Specific-pathogen-free C57BL/6J mice were gavaged with/without a single dose of 20 mg of streptomycin 24 h before infection. Overnight bacterial cultures of S. Typhimurium (hilA-HA, hilD^WT or hilD^mutant) strains were diluted 1:50, and were grown at 37 °C in a shaking incubator at 220 rpm for 4 h. S. Typhimurium were then washed and re-suspended in cold LB medium at 10¹⁰ CFU/mL. Mice were gavaged with 100 μl of the bacterial suspension (10⁹ CFU). Leftover inoculate were serially tenfold diluted and plated to confirm the number of CFU administered. For CDCA administration study, LabDiet® 5K52 and LabDiet® 5K52 supplemented with 0.2% CDCA were used throughout the experiment. Both diets were formulated and manufactured by Research Diets, Inc., NJ. Mice fed with or without CDCA were orally treated with 20 mg of streptomycin. 24-hour post streptomycin treatment, mice were orally infected with 10⁴ CFU of S. Typhimurium (14028s) (hilA-HA, hilD^WT or hilD^mutant).

For the S. Typhimurium infection survival assay, weight loss of mice was monitored starting just before infection, and mice were euthanized when they reached 80 % baseline weight, appeared hunched or moribund or exhibited a visibly distended abdomen (indicative of peritoneal effusion), whichever occurred first. Death was not used as an end-point. Colony-forming units (CFU) in the feces were determined by plating 50 μl or five serial dilutions of feces suspended in sterile PBS on Salmonella Shigella agar plates (BD 211597).

Fecal lipocalin-2 quantification by ELISA

Fecal samples were homogenized using the FastPrep-24 5G (4m/s, Adapter: QuickPrep, Time: 5 sec, Cycle:1, Lysing Matrix: D) (MP Biomedicals) in PBS containing 0.1% Tween 20 (10mg/ml) and stored at −20°C until analysis. Once thawed, samples were centrifuged at 20,000 rpm for 5 minutes to clarify and then supernatants were serially diluted (1:2, 1:20, 1:200) for analysis using the Mouse Lipocalin-2/NGAL DuoSet ELISA kit (R&D SYSTEMs) according to manufacturer’s instructions. A standard curve was generated and concentrations of Lcn-2 per gram of feces was calculated by linear regression analysis (GraphPad Software).

Statistics

Statistical analyses were conducted by GraphPad Prism 8.0. Briefly, pairwise comparisons were performed with unpaired two-sided t-test or two-sided Man-Whitney test. For comparisons of multiple groups with one variable, one-way ANOVA followed by multiple comparisons was used. For comparisons of multiple groups with two variables, two-way ANOVA followed by multiple comparisons is used. Adjusted P value was provided for multiple comparisons. Specific statistical tests are identified in corresponding figure legends. Data are presented as mean ± SD. Comparisons with P > 0.05 were not considered significant.

Extended Data

Extended Data Fig. 1. Chemoproteomic analysis of bile acid reporter-interacting proteins in S. Typhimurium.

a, b, c, LFQ proteomic analysis of bile acid reporters-labeled proteins in S. Typhimurium under UV irradiation (biologically independent samples, n = 4). S. Typhimurium cell lysates were reacted with az-biotin for the enrichment of bile acid reporters-labeled proteins with streptavidin beads and identification by LC-MS/MS. Volcano plots were presented for alk-X-CDCA (a), alk-X-LCA (b) and alk-X-UDCA (c). d, Gene ontology analysis of the shared 129 protein targets identified by LFQ proteomic reveals their association with pathogenesis (analyzed by The Database for Annotation, Visualization and Integrated Discovery (DAVID)). The names of the 16 proteins specifically assigned to the functional cluster of “virulence” are shown. e, Cellular component analysis of protein hits in a by gene ontology (GO).

Extended Data Fig. 2. Identification of bile acid resistant HilD mutations.

a. Three amino acids (N260, K264, R267) which are adjacent to potential binding pocket are highlighted in Robetta predicted HilD dimer. Green: DNA binding domain, orange: “jelly roll” motif, the putative binding domain. b. An overnight culture of S. Typhimurium (hilD-HA^WT or hilD-HA^mutants generated by CRISPR-Cas9) were diluted 1/50 into 4 mL SPI-1 inducing LB aliquots containing DMSO or CDCA (0.5 mM) and incubated for 4 h at 37 °C with 220 rpm shaking. SPI-1 effectors and flagella components levels in S. Typhimurium growth media were monitored by SDS-PAGE followed by Coomassie blue staining⁵¹. Experiments were repeated at least two times with similar results. c. Schematic for screening of mutant hilD library. Generally, hilD mutant library was generated from pTXB1-HilD by error prone PCR. The mutant library was transformed to reporter strain (S. Typhimurium tetRA-hilD-3XFLAG attλ::pDX1::hilA’-lacZ) and screened by plates (IPTG, carbenicillin, X-gal and 2.5 % bile acids)⁵². Plasmids from blue colonies were sequenced.

Extended Data Fig. 3. Investigation of HilD and CDCA interaction.

a. HilA-HA expression of HilD-mutant strains in presence of CDCA (500 μM). SPI-1 transcriptional factor HilA-HA expression of hilD^WT or hilD^mutant S. Typhimurium (hilA-HA) strains grown with or without 0.5 mM of CDCA. Western blot was quantified by grayscale analysis. Two-way ANOVA followed by Sidak's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. (biologically independent samples, n = 3). Comparisons with P > 0.05 were not considered significant. b, c, Docking of CDCA into HilD model. HilD protein model was generated based on AraC family protein Rns (PDB 6XIV³²) by SWISS-MODEL. The structure model of HilD was generated with the protein preparation wizard using Maestro software (12.4, Schrödinger, LLC), and Q39, N44, H95 were chosen as centroid of selected residues and the Ligand Docking protocol was used to model potential binding modes. Green: DNA binding domain, orange: “jelly roll” motif, the putative binding domain.

Extended Data Fig. 4. HilD Asparagine 44 mutations affect Salmonella response to bile acids and long chain fatty acid.

a, c, HilA-HA expression of hilD^WT or hilD^mutant S. Typhimurium (HilA-HA) strains grown with or without 2% bile acids (a) and 25 μM HDA (cis-2-hexadecenoic acid) (c). Western blot was quantified by grayscale analysis. b, Gentamicin protection assay of S. Typhimurium grown with 2% bile acids infecting HT-29 cells at MOI = 10. Statistical analysis for a, b, c, (a, biologically independent samples, n = 3; b, biologically independent samples, n = 6; c, biologically independent samples, n = 3) Two-way ANOVA followed by Sidak's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. Comparisons with P > 0.05 were not considered significant. d, Western blot for c.

Extended Data Fig. 5. CDCA inhibits hilD transcription and has little effect on HilD stability.

a, c, e, Overnight culture of Salmonella Typhimurium (hilD-HA for a, e; hilD-HA, lon :: cat for c) was subcultured 1:50 to 20 mL LB (300 mM NaCl). Bacteria was cultured for 3.5 h and another half an hour with treatment of CDCA or DMSO control (a, c) or cultured for 4 h with treatment of CDCA or DMSO (e). Then these cultures were treated with rifampin (100 ug/ml), streptomycin (200 ug/ml), and spectinomycin (50 ug/ml) to halt gene transcription and protein translation and kept at 37 °C with 220 rpm. Samples are picked at indicated time point. HilD-HA protein level was analyzed by Western blot. b, d, f, western blot was quantified by grayscale analysis for a, c, e correspondingly. For b, d Data was analyzed by two-way ANOVA and Sidak's multiple comparisons test. Comparisons with P > 0.05 were not considered significant. Centerline, average. Error bar, SD (biologically independent samples, n = 3). e, f was repeated twice with similar results. g, h, Overnight culture of Salmonella (HilD-HA) was subcultured 1:50 to 20 mL LB (300 mM NaCl). Bacteria was cultured for 4 h with treatment of CDCA or DMSO. HilD-HA and GroEL protein level was analyzed by Western blot (biologically independent samples, n = 6). Unpaired t test (two-sided). Centerline, average. Error bar, SD. i. Expression of SPI-1 and control genes of S. Typhimurium grown with DMSO, CDCA (concentration = 0.5 mM) measured by qRT-PCR (biologically independent samples, n = 6, results are pooled from two independent experiments, one independent experiment was carried out together with Fig. 1f, three replicates of control gene ftsZ were shared with 1f). Unpaired t test (two-sided). Centerline, average. Error bar, SD. Comparisons had P > 0.05 were not considered significant.

Extended Data Fig. 6. HilD^{Q39E+N44D+H95L} Salmonella is resistant to other gut microbiota metabolized small molecules.

a, SPI-1 transcriptional factor HilA-HA expression of S. Typhimurium (hilA-HA, hilD^{WT or Q39E+N44D+H95}) grown with propionic acid (10 mM), palmitic acid (0.1 mM) and cis-2-hexadecenoic acid (0.025 mM). b, Western blot was quantified by grayscale analysis, normalized with OD 600 value. All metabolites were compared to DMSO with one-way ANOVA and Dunnett's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. (biologically independent samples, n = 3). Comparisons with P > 0.05 were not considered significant. c. chemical structures of propionic acid, palmitic acid and cis-2-hexadecenoic acid.

Extended Data Fig. 7. Mice fecal metabolites inhibit Salmonella virulence.

a, feces were collected from C57BL/6 mice one day post with or without 20 mg streptomycin treatment. Feces from 4 mice were pooled together and added to lysing matrix D (Methanol : H₂O = 9 : 1, 100 mg/mL). Feces were then homogenized by Fastprep-24 5G (MP Biomedicals, 4 m/s, 3 cycles, 5s for each cycle). Supernatant after centrifugation (18 000 g, 4 °C, 30 min) were lyophilized and the corresponding residues were redissolved in LB for bacterial culture (metabolites from 50 mg of feces were dissolved in 1 mL of LB). Overnight culture of Salmonella (hilA-HA, hilD^{WT or Q39E+N44D+H95}) was subcultured 1:50 to LB with or without metabolites. Bacteria were cultured for 4 h at 37 °C. Expression of HilA-HA and GroEL was analyzed by western blot analysis. b, c, Western blot was quantified by grayscale analysis, normalized with GroEL expression. Bacteria treated with metabolites were compared to bacteria treated with LB by one-way ANOVA and Dunnett's multiple comparisons test, adjusted P value. Centerline, average. Error bar, SD. (biologically independent samples, n = 3). Comparisons with P > 0.05 were not considered significant.

Data availability

Proteome database (UP000002695) used in this study is available from Uniprot (https://www.uniprot.org/proteomes/UP000002695). The mass spectrometry proteomics RAW data have been deposited to the ProteomeXchange Consortium, via the PRIDE partner repository with the dataset identifier PXD034373. Rns crystal structure (PDB ID 6XIV) and ToxT crystal structure (PDB ID 3GBG) are available on the RCSB Protein Data Bank. The source data underlying Figs. 1-6 and Extended Data Figs. 1-7 are provided as source data files. All the other data supporting the findings of this study are available within the article and its supplementary information files, and from the corresponding author on reasonable request.