Clostridioides difficile toxins alter host metabolic pathway and bile acid homeostasis gene expression in colonic epithelium

TL;DR

This study shows that Clostridioides difficile and its toxins disrupt host metabolism and bile acid regulation in the gut, leading to inflammation and disease.

Contribution

The study reveals novel insights into how C. difficile toxins alter host metabolic and bile acid pathways in colonic epithelial cells.

Findings

C. difficile infection reduces PPAR signaling and cholesterol metabolism gene expression in mice.

Toxins TcdA and TcdB alter PPAR-regulated and cholesterol metabolic genes in human cell cultures.

Farnesoid X receptor pathways are significantly affected by toxin exposure in cell models.

Abstract

A major risk factor for acquiring Clostridioides difficile is antibiotic usage that disrupts a healthy microbial gut community, facilitating the establishment of infection. Once established, C. difficile secretes exotoxins (TcdA and TcdB) that are internalized into host colonic epithelial cells where they disrupt gut barrier function and induce hyperinflammation resulting in severe diarrhea and possibly leading to death. We employed three different platforms to explore gene expression of cells in the gut when exposed to C. difficile or its toxins, TcdA and TcdB. An antibiotic-treated mouse model of Clostridioides difficile infection (CDI) was used to identify differential gene expression with a NanoString Technologies mouse inflammatory gene panel consisting of 770 genes, including a subset of bile acid (BA) homeostasis and nuclear receptor genes. In the cecal tissue of mice with CDI, reduced expression was observed for genes involved in peroxisome proliferator-activated receptor (PPAR) signaling and cholesterol and glucose metabolism, while a significant increase in expression was observed for IL-17 related inflammatory genes. Similarly, Caco-2 cell culture and primary human colonic epithelial cells (hCE) exposed to toxins for 24 h showed altered expression in several PPAR-regulated and cholesterol metabolic genes similar to those found in mice. These cell culture experiments also revealed significant alterations in gene expression of the Farnesoid X receptor BA regulatory pathway. Together, these data suggest that exposure to C. difficile and its toxins may alter host cholesterol metabolic processes, including BA transport and synthesis.

Figures

Differential gene expression in mouse cecal tissue during C. difficile (CDI) infection compared to antibiotic-treated mice. (A) Mouse model of CDI experimental details. Mice (n = 4 per treatment group) were exposed to one of three treatments: no treatment or wild type (WT), cefoperazone treatment only (Cef), and cefoperazone-treated mice followed by challenge with 105 spores of CDI. (B) Percent baseline weight loss of mice from days 0–2 post challenge. (C) Toxin activity assay from the cecal content of CDI mice and Cef mice. An unpaired t-test was used for statistical analysis with P < 0.01 (P = 0.0032). (D) Significantly differentially expressed genes clustering to one of the four Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways identified in the gene set enrichment analysis are represented in the heatmap as gene counts from individual mice treated with Cef (light green) or CDI (dark green). (E) Volcano plots of genes differentially expressed in Cef mice compared to WT mice or (F) CDI mice compared to Cef mice. Colored circles correspond to the KEGG pathways represented in D. Differential expression analysis was performed using nSolver Advanced analysis software v.2.0.134 using Benjamini-Hochberg P value adjustment with P < 0.05.

Toxin exposure alters expression of PPAR and FXR signaling pathways in Caco-2 monolayers and primary human colonocytes. (A) Caco-2 gene expression. Cells were treated with 100 pM of either TcdA or TcdB, or 100 pM of both TcdA and TcdB and incubated for 24 h. (B) hCE gene expression. Cells were treated with 100 pM of either TcdA or TcdB. OCA (200 nM) was also added as an FXR agonist and positive control. No treatment control consisted only of media with no other additions. Tris (20 mM), pH 7.5, and 150 mM NaCl diluted to 2% in RMM acted as the no treatment vehicle control. For A, asterisks denote statistical significance determined by Mann-Whitney t-test compared to the no treatment control. *P < 0.05, P < 0.01. For B, asterisks denote statistical significance determined by unpaired t-test compared to no treatment control. *P < 0.05, P < 0.01.

Full text

INTRODUCTION

Clostridioides difficile infection (CDI) remains an urgent public health threat by the CDC with an incidence of approximately 500,000 cases per year in the USA and 29,000 deaths (1–3). A major risk factor for acquiring C. difficile is antibiotic usage that disrupts a healthy microbial gut community facilitating CDI establishment (4, 5). Once colonization is established, C. difficile secretes toxins (TcdA and TcdB) that are internalized in colonic epithelial cells where they inactivate Rho and Rac family GTPases by glucosylation (6–8). Inactivation of these small GTPases disrupts the actin cytoskeleton and cellular tight junctions, ultimately causing apoptosis or necrosis (6). Immune cell infiltration is important in defense against CDI. However, the prolonged influx of pro-inflammatory mediators leads to increased tissue damage that can result in severe diarrhea, toxic megacolon, pseudomembranous colitis, and even death (6, 9). Current treatments include subsequent antibiotic treatments such as vancomycin and fidaxomicin, leaving limited treatment options for the 30% of patients that develop recurrent CDI. For some patients, introducing intact, healthy microbiota from a fecal microbiota transplantation (FMT) and using newer Food and Drug Administration-approved microbiota-focused therapeutics like Vowst and Rebyota are their last resort.

C. difficile is exquisitely sensitive to gut-specific bile acids (BAs). Spore germination occurs in the presence of taurocholate (TCA), a primary BA produced by the liver (10). Alternatively, C. difficile growth is limited by secondary BAs (primary BAs modified by gut microbes), namely, deoxycholate (DCA) and lithocholate (LCA) (11, 12). The loss of gut bacteria due to antibiotic usage depletes secondary BAs and is associated with colonization of C. difficile. Recovery of secondary BAs in the intestinal BA pool is observed with successful FMT treatment (13, 14). The primary regulator of BA synthesis is the nuclear receptor Farnesoid X receptor (FXR). FXR is activated by high levels of BAs in the intestinal lumen. Activation of FXR leads to increased expression of key players in BA homeostasis, including BA transporters (FABP6, apical sodium dependent bile acid transporter [ASBT], and OSTα/β) and FGF19, the secreted signaling molecule that triggers downregulation of BA production in the liver, creating a negative feedback loop (15). FXR activation also protects against cell damage by stabilizing barrier integrity and reducing inflammation by inhibiting transcription of nuclear factor kappa-light-chain-enhancer of activated B cells regulated cytokines such as IL-1β (16–19). Several BAs serve as FXR agonists, including the primary BA chenodeoxycholate and secondary BAs, DCA and LCA, and their conjugated derivatives (20). Enriching the intestinal BA pool with BAs that function as FXR agonists may enhance FXR activity, reduce toxin-mediated inflammation in CDI, and serve as a promising therapeutic strategy during infection.

It is well established that C. difficile toxins induce changes to the host cellular structure and inflammatory response (6–8, 21), but little is known about BA modification during CDI. Previous work has shown that C. difficile triggers an influx of TCA that accumulates in the ileal lumen, altering the BA pool in a toxin-dependent manner (22). The underlying mechanisms of this finding remain unclear. A deeper understanding of the interaction between C. difficile and BA regulation is needed before exploring FXR agonists as a potential therapeutic. In this study, we leveraged two in vitro and one in vivo platform to assess gene expression alterations in host BA homeostasis genes when exposed to C. difficile or purified toxins. Platforms included a primary CDI mouse model, toxin exposure to Caco-2 cells, and primary colonocytes derived from the intestinal crypts of a healthy human donor (hCE). Differential gene expression in genes of interest, including FXR and its target genes involved in BA homeostasis and cholesterol transporters, was evaluated. Our findings show mice challenged with C. difficile had reduced gene expression in metabolic pathways, including glycolysis, cholesterol metabolism, and peroxisome proliferator-activated receptor (PPAR) signaling, and increases in several inflammatory genes in the gut. Gene expression of select genes identified in mice was also decreased in the hCE cells but increased in the Caco-2 immortal cell line when exposed to toxins alone. C. difficile toxin exposure can alter the expression of BA homeostasis genes without FXR activation and suggests that toxins could be modifying BA and cholesterol transport in colonocytes.

MATERIALS AND METHODS

Animals and housing

C57BL/6J mice (males) were purchased from Jackson Laboratories (Bar Harbor, ME) and quarantined for 1 week prior to starting the antibiotic water administration to adapt to the new facilities and avoid stress-associated responses. Following quarantine, the mice were housed with autoclaved food, bedding, and water. Cage changes were performed weekly in a laminar flow hood by laboratory staff. Mice had a 12 h light and dark cycle and were housed at an average temperature of 70°F and 35% humidity.

Mouse model of C. difficile infection and sample collection

Methodologies of this primary CDI model have been previously published (23, 24). Groups of 5-week-old C57BL/6J mice (male, four mice per treatment group) were given cefoperazone (0.5 mg/mL) in their drinking water ad libitum for 5  days, followed by a 2 day washout with regular drinking water ad libitum (Fig. 1A). Mice were then challenged with 10^5^ spores of C. difficile R20291 pyrE via oral gavage at day 0. Mice were monitored for weight loss and clinical signs of CDI (lethargy, inappetence, diarrhea, and hunched posture) for 2 days post-challenge. Fecal pellets were collected daily for C. difficile enumeration. Uninfected controls included cefoperazone-treated mice and conventional mice. Animals were euthanized 2 days post-challenge. The contents and tissue from the cecum and colon were collected immediately at necropsy, flash frozen in liquid nitrogen, and stored at −80°C until further analysis. To prevent RNA degradation, snips of cecal and colonic tissue were also stored in RNA-Later at −80°C until further analysis.

*Differential gene expression in mouse cecal tissue during C. difficile (CDI) infection compared to antibiotic-treated mice. (A) Mouse model of CDI experimental details. Mice (n = 4 per treatment group) were exposed to one of three treatments: no treatment or wild type (WT), cefoperazone treatment only (Cef), and cefoperazone-treated mice followed by challenge with 105 spores of CDI. (B) Percent baseline weight loss of mice from days 0–2 post challenge. (C) Toxin activity assay from the cecal content of CDI mice and Cef mice. An unpaired t-test was used for statistical analysis with *P < 0.01 (P = 0.0032). (D) Significantly differentially expressed genes clustering to one of the four Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways identified in the gene set enrichment analysis are represented in the heatmap as gene counts from individual mice treated with Cef (light green) or CDI (dark green). (E) Volcano plots of genes differentially expressed in Cef mice compared to WT mice or (F) CDI mice compared to Cef mice. Colored circles correspond to the KEGG pathways represented in D. Differential expression analysis was performed using nSolver Advanced analysis software v.2.0.134 using Benjamini-Hochberg P value adjustment with P < 0.05.

Vero cell cytotoxicity assay

Toxin activity was tested as described previously (23). Cecal contents were resuspended in 1:10 dilution wt:vol in phosphate-buffered saline and filtered with a 0.22 µm filter. Filtrates were added to wells seeded with 10^4^ cells per well and incubated overnight. Toxin A from VPI 10463 was used as a positive control, and antitoxin (List Labs 152C, and TechLab, T5003) was used to quench activity.

RNA isolation and gene expression analysis

RNA extraction was performed on cecal tissue using PureLink RNA Mini kit (Thermo Fisher, 12183025) following the manufacturer’s protocol. The RNA was treated with Turbo DNase (Thermo Fisher, AM2239) to remove genomic DNA contamination. Samples were quantified using Qubit (Thermo Fisher Scientific) and run on an Agilent Bioanalyzer NanoChip (Santa Clara, CA) to assess the quality of RNA. The nCounter Mouse Fibrosis v.2 gene expression panel and custom code set was purchased from NanoString Technologies (Seattle, WA) and consists of 770 mouse genes, including 6 internal reference genes and 19 custom genes of interest (Table S1). The nCounter assay was performed using 100 ng of total RNA. Hybridization reactions were performed according to the manufacturer’s instructions with 5 µL diluted sample preparation reaction, and samples were hybridized overnight. Hybridized reactions were purified using the nCounter Prep Station (NanoString Technologies), and data collection was performed on the nCounter Digital Analyzer (NanoString Technologies) following the manufacturer’s instructions to count targets. Nanostring nCounter Max platform was utilized at the Microbiome Shared Resource Lab at Duke University to perform these assays.

Raw data from NanoStrings nCounter was analyzed using the nSolver Analysis Software v4.0. Gene counts for each sample were background subtracted based on the geometric mean of negative probes. Gene counts for each sample were normalized based on the geometric mean of positive probes and to a panel of housekeeping genes: Acad9, Armh3, Cnot10, Gusb, Mtmr14, Nol7, Nubp1, Pgk1, Ppia, and Rplp0. Differential expression analysis was performed using nSolver Advanced analysis software v.2.0.134 using Benjamini-Hochberg P value adjustment. Probes used custom annotations as detailed in Table S2. Further analysis was performed in R v.4.4.1. Ensembl and National Center for Biotechnology Information gene IDs were annotated using biomaRt v.2.60.1 with the GRCm39 gene ensemble. Gene set enrichment analysis (GSEA) was performed on differentially expressed genes using a minimum gene set size of 3 using clusterProfiler v.2.1.6 to perform GSEA with Benjamini-Hochberg correction. Graphing was performed using tidyverse v.2.0.0 and ComplexHeatmap v.2.20.0. Normalized gene counts in heatmaps were standardized by subtracting the mean and dividing by the standard deviation for each probe. All computational analysis is available at doi 10.5281/zenodo.13350814.

RNA extraction from Caco-2 cells

Twenty-four-well plates were seeded with 5 × 10^4^ Caco-2 (American Type Culture Collection, HTB-37) cells per well and grown in Dulbecco’s Modified Eagle Medium supplemented with 2 mm L-glutamine, 10% fetal bovine serum, and 5,000 U/mL penicillin plus 5,000 µg/mL streptomycin cocktail. Plates were incubated in 5% CO_2_ at 37°C with media changes every 2–3 days. After the cells reached confluence (3–4 days), they were incubated for an additional 7 days. After 7 days post-confluency, cells were treated with 100 pmol of recombinantly purified TcdA, TcdB, or TcdB lacking the glycosyl-transferase domain activity (TcdB-GTD^mut^) for 24 h before RNA extraction. TcdB-GTD^mut^ contains mutations in the GTD domain at the following residues: W102, D286, and D288. This mutant also lacks residues 2101–2366 in the C-terminus of the CROP domain, allowing complete exposure of the receptor-binding domain (DRBD). After incubation, cells were lysed with Trizol reagent. Chloroform (200 µL) was added to the lysate, mixed, and incubated for 15 min. The lysate was centrifuged at 4°C for 15 min, and the aqueous layer was removed and added to equal volumes of 100% ethanol. The RNA from the ethanol mixture was purified using the Zymo Research RNA Clean and Concentrator-25 (Zymo Research, cat #R1017) per manufacturer’s instructions. DNA was degraded on the column with Zymo Research DNAse 1 (Zymo Research, cat #E1010) per manufacturer’s protocol. RNA concentration was measured with Qubit.

RNA extraction from human colonocytes

Primary human transverse colon epithelial cells were acquired from Altis Biosystems and cultured as previously described (25, 26). Cells were plated onto a 96-well Transwell plate (Corning, 3392) and grown in Altis RepliGut Growth Media. Once cells reached confluence, Altis RepliGut Maturation Medium (RMM) was applied to promote cellular differentiation and polarization. hCE barrier integrity was monitored daily via TEER using the EVOM Auto (EVA-MT-03-01, WPI). Treatments used for this study were (i) no treatment (NT): cells were grown in RMM only; (ii) vehicle: RMM media with 2% vehicle buffer (20 mM Tris, pH 7.5, 150 mM NaCl), (iii) obeticholic acid (OCA) added to RMM at a concentration of 200 nM; (iv and v) TcdA or TcdB: toxin diluted in vehicle and added to RMM to a final concentration of 100 pM toxin and 2% vehicle buffer. At 24 h post-treatment, RNA lysates were collected in Buffer RLT (Qiagen, cat #79216). RNA extraction followed the Qiagen Rneasy Plus extraction kit (Qiagen, cat #74134).

Quantitative reverse transcription PCR

RNA from Caco-2 cells and human colonocytes was normalized to 500 and 100 ng, respectively, and used as a template for reverse transcription reactions using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher cat #4368814). The resulting cDNA was used for quantitative PCR with the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad cat #1725270). For relative quantification, the ΔΔCt method was used to normalize the gene of interest to the internal expression control, GAPDH, followed by normalization of toxin-treated cells to no treatment controls. Data were reflected as a fold change comparison between toxin-treated cells to no treatment controls.

Protein expression and purification of TcdA and TcdB

Expression and isolation of recombinant TcdA, TcdB, and TcdB-GTD^mut^ were performed as described by Yang et al. (27, 28). Purified proteins were a gift from Roman Melnyk and John Tam. Briefly, transformed Bacillus megaterium was inoculated into Luria–Bertani medium (LB) containing tetracycline and grown to an A600 of 1.6, followed by overnight xylose induction at 30°C. Bacterial pellets were collected, resuspended with 20 mM Tris, pH 8/0.1 M NaCl, and passed twice through an EmulsiFlex C3 microfluidizer (Avestin, Ottawa, ON) at 15,000 psi. The resulting lysate was clarified by centrifuging at 18,000 × g for 20 min. TcdB and TcdA were purified by nickel affinity chromatography followed by anion exchange chromatography using HisTrap FF Crude and HiTrap Q columns (Cytiva), respectively. TcdB-GTD^mut^ was purified similarly with the exceptions of initial purification used a cobalt affinity column, and the eluent was biotinylated at 4°C with 15× molar excess maleimide biotin (Thermo 21901BID) and then followed by a HiTrap Q column. Fractions containing purified protein were verified by SDS-PAGE, then pooled and diafiltered with a 100,000 MWCO ultrafiltration device (Corning) into 20 mM Tris, pH 7.5/150 mM NaCl. Finally, glycerol was added to 5% vol/vol; the protein concentration was estimated by A280, divided into single-use aliquots, and stored at −80°C.

Statistical analysis

An unpaired t-test was used for the toxin assay (Fig. 1C) statistical analysis (P < 0.05). For the toxin assay (Fig. 1C) and quantitative reverse transcription CR (qRT-PCR) data (Fig. 2A and B), statistical tests were performed in GraphPad Prism v.8. For qRT-PCR data, normalization assays were conducted using the Shapiro-Wilk test. Differential expression analysis for NanoStrings data was performed using nSolver Advanced analysis software v.2.0.134 using the Benjamini-Hochberg P value adjustment. Genes with a log2 fold change of >0.5 or <−0.5 and P < 0.05 were called as differentially expressed genes (Fig. 1E and F). Caco-2 expression data were not normally distributed (Shapiro-Wilk test), and treatments were compared to NT control using the Mann-Whitney non-parametric test (Fig. 2A; Fig. S2). hCE data showed normal distribution, and an unpaired t-test was used to compare treatments to NT (Fig. 2B). Significance for qRT-PCR analysis is P < 0.05.

*Toxin exposure alters expression of PPAR and FXR signaling pathways in Caco-2 monolayers and primary human colonocytes. (A) Caco-2 gene expression. Cells were treated with 100 pM of either TcdA or TcdB, or 100 pM of both TcdA and TcdB and incubated for 24 h. (B) hCE gene expression. Cells were treated with 100 pM of either TcdA or TcdB. OCA (200 nM) was also added as an FXR agonist and positive control. No treatment control consisted only of media with no other additions. Tris (20 mM), pH 7.5, and 150 mM NaCl diluted to 2% in RMM acted as the no treatment vehicle control. For A, asterisks denote statistical significance determined by Mann-Whitney t-test compared to the no treatment control. *P < 0.05, **P < 0.01. For B, asterisks denote statistical significance determined by unpaired t-test compared to no treatment control. *P < 0.05, *P < 0.01.

RESULTS

CDI decreases gene expression of host cholesterol metabolism in mice

An antibiotic-treated mouse model of CDI was used to investigate changes in host gene expression during acute infection (Fig. 1A). CDI mice were symptomatic 2 days post-challenge with C. difficile spores, evidenced by weight loss compared to either wild type (WT) or antibiotic-treated mice (Cef) (Fig. 1B). Significantly greater toxin activity was observed in the cecal content of CDI mice compared to the Cef mice as expected (23), confirming active CDI (Fig. 1C; P = 0.0032, unpaired t-test). CDI mice had a C. difficile cecal load of approximately 1.15 × 10^8^ CFU/g. At the time of necropsy, cecal tissue was collected from each treatment group for later RNA extraction and differential gene expression leveraging NanoString Technologies. We used the mouse inflammatory gene panel consisting of 770 genes and a subset of BA homeostasis and nuclear receptor genes. GSEA was used to identify Kyoto Encyclopedia of Genes and Genomes pathways with significant differentially expressed genes (DEs) (Fig. S1). Four pathways were identified: the IL-17 signaling pathway consisted of four genes with increased expression (Tnf, Ccl2, Mmp13, and Mmp3) during CDI compared to Cef, while 25 genes with reduced expression were spread between glycolysis/gluconeogenesis metabolism, cholesterol metabolism, and PPAR signaling pathway (Fig. 1D; Fig. S1A through D). Most genes with reduced expression were associated with cholesterol metabolism, where several genes are also regulated through PPAR signaling pathways (Fig. 1D; Fig. S1C and D). Volcano plots comparing host gene expression of Cef mice compared to WT mice (Fig. 1E) show a small number of genes with altered expression (12 increased and 45 decreased), indicating that exposure to the antibiotic alone had a subtle effect on gene expression. However, volcano plots comparing host gene expression in CDI compared to Cef mice (Fig. 1F) indicate that exposure to CDI does significantly impact gene expression, with 17 genes significantly increased in expression and 135 significantly decreased in expression (Fig. 1F). Reduction in cholesterol transport genes was prominent, namely, the apolipoproteins, Apoa1, Apoa4, Apoc3, and Apob (Fig. 1D and F). BA-related genes, Cyp27a1 and Scl51a (OSTα), had significantly reduced expression in mice with CDI (Fig. 1D and F; Table S1). Expression of several nuclear receptors including VDR (Nr1i1), CAR (Nr1i3), LXRb (Nr1h2), LXRα (Nr1h3), and RXRA (Table S1) was also significantly decreased in mice with CDI.

Toxin exposure in Caco-2 cells and human primary colonocytes disrupts cholesterol and bile acid transport gene expression

To determine if toxins alone can modify transcription of the PPAR signaling pathway and cholesterol metabolism genes in human colon cells in a similar manner as mouse cecal tissue, we exposed differentiated Caco-2 cells grown in a flat-bottom well system to toxins TcdA and TcdB or both at 100 pM for 24 h. Caco-2 cells exposed to TcdA had significantly altered expression of the gene encoding CAR (NR1I3) (Fig. 2A). Caco-2 cells exposed to TcdB had significantly increased expression of PPAR signaling genes PPARG and CPT-1, as well as the apolipoprotein APOA4 and the nuclear receptor CAR (NR1I3) (Fig. 2A). When both toxins (TcdA + TcdB) were combined at 100 pM each, similar increases in expression to TcdB only exposure were observed, indicating that TcdB is responsible for the changes in expression. This was not true for PPARG, where expression decreased upon exposure to both toxins (Fig. 2A).

Cyp27a1 and Slc51a were the only BA homeostasis genes with significant changes in gene expression levels in the mouse study. The lack of other FXR-regulated BA homeostasis genes may be attributed to high levels of the FXR antagonist TβMCA (29), which mice typically produce, potentially inhibiting FXR activation and, in turn, inhibiting expression of the FXR-regulated BA genes. We wanted to explore FXR-regulated BA homeostasis genes further in an antagonist-free environment and used qRT-PCR to determine if toxins altered gene expression in Caco-2 cells. Significant increases in gene expression were observed in cells exposed to all three treatments, TcdA, TcdB, and TcdA + TcdB (Fig. 2A). TcdB and TcdA + TcdB exposed cells had significantly increased expression in NR1H4 (FXR), FGF19, NR0B2 (SHP), and Slc10a2 (ASBT) genes, but not FABP6. TcdA-exposed cells had significantly increased expression in FGF19 and NR0B2 only. NR0B2 had the highest increase in expression when exposed to TcdB and TcdA + TcdB, with a 65- and 50-fold increase, respectively. All other genes tested had less than 10-fold increases in expression.

To determine if disruption of Rho-GTPase signaling is responsible for the observed changes in expression seen in only the FXR-regulated genes, we exposed Caco-2 cells to 100 pM of the TcdB mutant lacking glycosyl-transferase activity (TcdB-GTP^mut^), which retains all other functional characteristics of the protein. The DE seen with fully active TcdB is lost with the mutant protein (Fig. S2), indicating that TcdB is altering gene expression via disruption of GTPase signaling.

Caco-2 cells are derived from an immortal adenocarcinoma in which expression of NR1H4 and FABP6 differs from primary colonic epithelial cells (30, 31). To avoid confounding variables using cancer cell lines, we implemented the RepliGut platform (Altis Biosystems, Durham, NC) comprising differentiated primary colonic epithelial (hCE) cells derived from intestinal crypts of human transplant-grade donors (25, 26). hCEs were exposed to differing treatments for 24 h and expression levels for the same PPAR pathway and cholesterol genes assayed in the Caco-2 cells were measured. SCD1 was the only gene to result in significant differences; however, instead of increasing in expression like the Caco-2 cells, gene expression decreased in hCE cells (Fig. 2B). In fact, expression of other PPAR pathway genes resulted in little change or decreased expression in the hCE cells (Fig. 2B).

Significant differences in gene expression were observed in the BA regulatory genes in hCE (Fig. 2B). NR1H4 expression was significantly reduced when exposed to either TcdA or TcdB compared to the no treatment control. These data are opposite of NR1H4 expression in Caco-2 cells. However, like the Caco-2 cells, FGF19 expression was significantly increased in the hCE cells when exposed to either TcdA or TcdB alone, but only TcdB significantly increased NR0B2 expression. Significant decreases in the expression of FABP6 were observed for both toxins, but only TcdB exposure resulted in significantly reduced SLC10A2 expression. The addition of OCA, an agonist of FXR, at 200 nM did not increase NR1H4 expression, but OCA exposure increased expression of FGF19 and FABP6, both of which are induced by activated FXR, suggesting OCA exposure did activate FXR, leading to increased expression of these genes. Interestingly, activation of FXR via OCA is expected to increase NR0B2 expression; however, OCA exposure resulted in no change. The DE observed for NR0B2 and FABP6 when exposed to toxin is not through FXR and suggests that TcdA and TcdB can alter expression of BA homeostasis genes without activation of FXR by an agonist.

DISCUSSION

It is well known that C. difficile is exquisitely sensitive to BAs, namely, TCA which acts as a spore germinant (32). Wexler et al. have shown that TCA concentration increases in the intestinal lumen in the early stages of CDI and is dependent on toxin (22). However, the mechanism by which C. difficile is able to modify the host response with regard to BA synthesis is not known. This study used three different systems to interrogate the host BA response to C. difficile toxin exposure. In this study, we provide evidence that C. difficile is able to modify cholesterol transport and metabolism, including BA transport, in a toxin-dependent manner.

Differential gene expression in the mouse cecum did not result in significant changes in BA regulatory genes. However, several genes that decreased in expression are important in cholesterol and lipid transport. These include apolipoprotein genes Apoa1, Apoa4, Apoc3, and Apob (Fig. 1D and F). These apolipoproteins are important components of chylomicron and high-density lipoprotein structure, which pack and transport cholesterol and lipids to the lymph system from intestinal epithelial cells (33–35). The reduction in gene expression of intestinal apolipoproteins hints at reduced lipid and cholesterol transport into the lymph and bloodstream. Ppara was also significantly reduced in cecal tissue and correlates with the reduced apolipoproteins as PPARs are transcriptional regulators of Apoa1, Apoc3, and Apoa2 (36, 37). Further support for the disruption of lipid transport was seen with the reduced expression of intracellular lipid and cholesterol transport proteins, Plin4 and Ncp1. Like the CDI model, expression of APOA4 was decreased in the hCE when exposed to toxins, indicating that inhibition of cholesterol transport by C. difficile is dependent on toxins. Modification of cholesterol transport is seen in various infections to enhance pathogenesis. Cryptosporidium parvum, an intestinal parasite infecting ileal cells, reduces apolipoprotein expression and protein content in these cells to reduce cholesterol transport during infection (38). Human cytomegalovirus (HCMV) modifies cholesterol transport by enhancing cholesterol efflux rather than inhibition. HCMV mobilizes lipid rafts via manipulation of actin filaments to facilitate cholesterol binding to APOA1 and increasing cholesterol efflux (39).

Despite no significant changes in expression in FXR BA regulatory genes in the CDI mouse model, we did observe significant changes when exposed to toxins TcdA and TcdB in the two different colonic epithelial cell types. Caco-2 and hCE expression data show increased expression of BA regulatory genes NR0B2 and FGF19. These proteins are important not only in BA regulation but also have a role in lipid and cholesterol metabolism. SHP (NR0B2) is an orphan response regulator that binds to transcriptional regulatory proteins, altering their ability to initiate transcription. FGF19 enhances SHP activity through phosphorylation of SHP at T58. This partnership between SHP and FGF19 is able to modify the expression of key cholesterol and BA transport genes (40, 41). The ASBT gene SLC10A2 is downregulated through SHP repression of the LRH-1 transcriptional regulator (42, 43). The increase in NR0B2 and FGF19 expression and the significant reduction in SLC10A2 expression in hCE cells suggest that toxins initiate the SHP/FGF19 partnership to reduce influx of BAs into the cells by repressing SLC10A2 expression and may explain increased concentrations of TCA in the intestinal lumen in early infection. The SHP/FGF19 partnership is also important in downregulating cholesterol transport into cells by repressing SREBF2, a major cholesterol metabolic transcriptional regulator, and therefore inhibiting transcription of Ncp1l1, a major intestinal cholesterol transporter (40). SHP is also a repressor of MTP, an APOB chaperone, and APOA4 integral to chylomicron assembly (44).

Interestingly, FABP6 gene expression was also significantly reduced in hCE when exposed to toxins. FABP6 is an intracellular BA transporter and is highly expressed during FXR activation. The reduced expression supports the reduced BA transport during toxin exposure. However, this reduction appears to be independent of FXR activation, as OCA exposure to hCE cells enhances FABP6 expression. The mechanism for the reduction in expression is unknown. Similarly, FXR activation is known to increase NR0B2 expression (45). However, increased expression of NR0B2 in the hCE cells is not observed with OCA exposure, suggesting that the increased expression observed with toxin exposure in this study is independent of FXR activation. Further studies are needed to determine if C. difficile is able to alter regulation of BA homeostasis genes and BA transport in a toxin-dependent manner.

Repression of BA transport may be advantageous for C. difficile by effectively increasing the concentrations of BA in the lumen to enhance spore germination or growth. BA transport has been observed in the presence of other toxins and pathogens. The fungal mycotoxin deoxynivalenol alters BA transport by inhibiting SLC10A2 FABP6 and SLC51A (OSTα) (46). Similarly, infection with Salmonella Typhimurium in the porcine ileum reduces expression of SLC10A2 and FABP6 (47). The increase in FGF19 also suggests that C. difficile toxins not only can inhibit BA transport but may also have the potential to disrupt BA synthesis in the liver since FGF19 is a signaling protein that is transported to the liver to initiate reduction of BA synthesis.

C. difficile toxins are able to promote epithelial disruption by inhibiting the Rho and Rac family GTPase signaling pathways that maintain cytoskeletal structure through glucosylating the GTPase via the GTD domain of the toxins (6–8). This disruption also promotes altered expression of FXR-regulated genes (Fig. 2A; Fig. S2), suggesting that toxins are facilitating a complex response in epithelial cells via GTPase disruption to promote disease and the success of C. difficile. Mouse studies with mutants of the C. difficile strain R20291 in the glucosyltransferase domain of both TcdA and TcdB are required for severe symptoms in the CDI mouse model, demonstrating severe weight loss, diarrhea, and edema but are not required for increased epithelial damage, underscoring the importance of this domain in the development of severe disease (21). Utilizing such mutants for in vivo studies would clarify the importance of GTPase signaling in regulating the BA homeostasis genes tested here.

Three different platforms were utilized to interrogate the effects C. difficile toxins have on BA regulatory pathways. Gene expression in the CDI mouse model did not result in changes in FXR regulatory genes as was observed in the tissue culture experiments. This may be due to the high levels of the BA TbMCA found in the mouse inhibiting FXR activation. The BA pool was not measured in this study and may have an unknown impact. The presence of other microbes as well as the host immune response complicates our understanding. Tissue culture studies are advantageous as they reduce complicated interactions of an in vivo system. However, Caco-2 cell lines are derived from colonic cancer cells which inherently have altered gene expression and metabolism, where results need to be interpreted cautiously. Interestingly, the hCE expression data obtained are more like the expression observed in the CDI mouse model than the expression observed in Caco-2 cells, supporting previous work identifying hCE as a better mimic of cells in vivo (48). Again, caution is needed as stem cells obtained from another individual may respond differently.

There remain some limitations in our animal study design that includes the use of male mice only. We have tested sex as a biological variable many times in our mouse model of CDI and do not see any changes in clinical signs of disease between male and female. Indeed, the host transcriptional responses observed cannot be generalized to both sexes, and further exploration is needed to understand if our observations span both sexes. Also, to reduce the number of mice, we used a single time point to assess host gene expression levels, which limits our understanding of how bacterial kinetics affects this. Looking at 2 days post-C. difficile challenge is where mice lose the most weight and experience severe clinical signs of disease. Measuring host gene expression levels at the point of the most severe toxin-mediated inflammation may miss important changes in gene/protein levels that lead to the progression of severe disease. Finally, the use of C. difficile tcdA and tcdB mutant strains would solidify the role toxins have in altering bile acid and cholesterol transport as observed in this study.