Groups of age matched C57BL/6 mice (6–12 weeks old) were used. Group sizes used in each experiments are indicated in the figure legends. Germ-free and gnotobiotic sDMDMm2 mice (Brugiroux et al., 2016) were established and maintained at the Clean Mouse Facility (CMF) of the Department of Clinical Research of the University of Bern. sDMDMm2 mice were colonized with a mouse-intestine derived 12-species mouse microbiota (Oligo-MM¹²) consisting of the following bacterial species: Acutalibacter muris sp. nov. KB18 (DSM 26090), Flavonifractor plautii YL31 (DSM26117), Clostridium clostridioforme YL32 (DSM 26114), Blautia coccoides YL58 (DSM 26115), Clostridium innocuum I46 (DSM26113), Lactobacillus reuteri I49 (DSM 32035), Enterococcus faecalis KB1 (DSM 32036), Bacteroides caecimuris sp. nov. I48 (DSM 26085), Muribaculum intestinale sp. nov. YL27 (DSM 28989), Bifidobacterium longum subsp. animalis YL2 (DSM 26074), Turicimonas muris sp. nov. YL45 (DSM 26109) and Akkermansia muciniphila YL44 (DSM 26127). All Oligo-MM¹² strains are available at http://www.dsmz.de/miBC. In-depth description of the Oligo-MM¹² consortium species and description of novel taxa are provided elsewhere (Brugiroux et al., 2016; Lagkouvardos et al., 2016). SPF mice were purchased from Envigo (formerly Harlan, The Netherlands). Germ-free and sDMDMm2 mice were exported from flexible film breeding isolators into sterile individually ventilated cages (IVC; model Sealsafe Plus, Tecniplast, Italy) for C. difficile infection experiments. All animal experiments were performed in accordance with the Swiss Federal and the Bernese Cantonal regulations and were approved by the Bernese Cantonal ethical committee for animal experiments under the license number BE 82/13.

C. scindens ATCC35704 (Morris et al., 1985) was grown in BHIS+ medium, which contains 37 g BHI, 5 g yeast extract, 40 ml salts solution (0.2 g CaCl₂, 0.2 g MgSO₄, 1 g K₂HPO₄, 1 g KH₂PO₄, 10 g NaHCO_3, and 2 g NaCl in 1 L _ddH₂O), 1 g L-cysteine and 2 g fructose per L _ddH₂O or on BHIS+ agar. Erythromycin (10 μg/mL) was added to BHIS+ medium to select for C. scindens ATCC35704 from fecal/cecal samples. C. difficile DH1916 (Burns et al., 2011) was grown in BHIS medium, BHIS agar or Clostridium difficile fructose agar (CDFA). Erythromycin (10 μg/mL), cycloserine (0.5 mg/mL), cefoxitin (16 μg/mL), and taurocholate (0.1%) were added to plates where necessary. BHIS+ plates for the enumeration of C. scindens were incubated in an anaerobic workstation (Don Whitley A45 HEPA, 10% CO₂, 10% H₂, 80% N₂) for 72 h at 37°C. CDFA or BHIS plates for the enumeration of C. difficile were incubated in anaerobic jars using an atmospheric generator (Genbox anaer, bioMérieux, France) for 24–48 h at 37°C or in the anaerobic workstation for 24–48 h at 37°C.

C. difficile DH1916 was grown overnight in BHIS medium. A 100 μL aliquot of the overnight culture was plated onto BHIS agar plates and incubated for 5 days in an anaerobic workstation at 37°C. Plates were removed from the workstation and stored at 4°C overnight. Using a loop, all cell material was scraped into 1 mL _ddH₂O (three plates per mL). The spore suspensions were re-incubated at 4°C overnight to free spores from mother cells and subsequently washed for at least 10 times in ice-cold _ddH₂O. Spore preps were controlled for their purity using phase contrast microscopy and spores were enumerated using a hemocytometer as well as by plating on BHIS plates containing 0.1% taurocholic acid. All spore preparations were heat-treated (65°C, 30 min) to ensure complete inactivation of any remaining vegetative cells. All Spore preps were stored at 4°C until use.

SPF mice were antibiotic-pretreated with Clindamycin (100 μL by intraperitoneal injection, 2 mg/mL in PBS) or Streptomycin (100 μL by gavage, 200 mg/mL in PBS) 24 h before infection. sDMDMm2 mice were inoculated with C. scindens ATCC35704 by gavage of 10⁹–10¹⁰ CFU C. scindens and colonized for 4 or 7 days before C. difficile infection or sacrifice, as specified. C. scindens pre-colonization was performed in flexible film isolators, and the mice were subsequently exported into sterile IVCs for C. difficile infection. C. difficile DH1916 was administered by gavage of 10³ CFU of spores suspended in 100 μL sterile PBS.

Lipocalin-2 was measured from terminally collected cecal content using a commercial ELISA kit (R&D Systems, USA). The wells of a 96-well plate were coated with lipocalin-2 capture antibody (1:200 in PBS, 50 μL, incubated at 4°C overnight), washed with wash buffer (0.05% Tween-20 in PBS) and blocked with 150 μL blocking buffer (2% BSA in PBS, 3 h at RT). The diluted samples (10-fold dilutions over four wells) and the standards (duplicate, 3-fold dilutions over eight wells) were applied and the plates were incubated at 4°C overnight. After washing with wash solution, 100 μL of biotinylated anti-lipocalin-2 antibody was applied (1:200 in blocking buffer, 50 μL) and incubated for 1 h at room temperature. The wells were washed with wash buffer and treated with horseradish-peroxidase-streptavidin (Biolegend, USA, 1:1000 in PBS, 100 μL) for 1 h at room temperature. Development was performed using 100 μL of substrate buffer (0.1 M NaH₂PO₄, pH 4.0) with 0.01% ABTS and 0.1% H₂O₂. The absorbance at 415 nm was measured with a plate reader. Standards and samples were fitted using a non-linear regression and lipocalin-2 values were calculated back to ng of lipocalin-2 per g of sample.

Cecal tissue was fixed in 4% PFA/PBS immediately after the sacrifice of the mice and kept at 4°C for 24 h. The cecum tissue was rehydrated using a 20% Sucrose/PBS solution at 4°C for at least 24 h. The ceca were embedded in Optimal Cutting Temperature (OCT) medium (Tissue-Tek, Sakura, Japan) and stored at −80°C. Sections of 7 μm thickness were cut, air dried and stained with hematoxylin, and eosin. Cecal histopathology was evaluated in a blinded manner by a pathologist using an established scoring system for acute intestinal inflammation as described (Barthel et al., 2003). Scoring criteria were: (1) Submucosal edema; 0 = no pathological changes, 1 = mild edema; the submucosa is <0.20 mm wide, 2 = moderate edema; the submucosa is 0.21 to 0.45 mm wide, 3 = profound edema; the submucosa is >0.46 mm wide. (2) Polymorphonuclear cell (PMN) infiltration into the lamina propria; PMN were enumerated in 10 high-power fields (400x magnification in a field diameter of 420 μm) and the average number of PMN/high-power field was calculated, 0 ≤ 5 PMN/high-power field, 1 = 5–20 PMN/high-power field, 2 = 21–60 PMN/high-power field, 4 ≥ 100 PMN/high-power field. (3) Loss of goblet cells; Goblet cells were enumerated in 10 high-power fields (400x magnification in a field diameter of 420 μm) of cecum epithelium and the average of Goblet cells per high-power field was calculated, 1 = 11–28 goblet cells/high-power field, 2 = 1–10 goblet cells/ high-power field, 3 ≤ 1 goblet cell/ high-power field. (4) Epithelial integrity; 0 = no pathological changes, 1 = epithelial desquamation, 2 = erosion of the epithelial surface (gaps of 1–10 epithelial cells per lesion), 3 = epithelial ulceration (gaps of >10 epithelial cells per lesion). The combined histopathological score was determined as the sum of these individual scores.

To measure the expression of calprotectin, cecal tissue was taken at the time of necropsy. The cecal tissue was emptied from all fecal matter, washed in sterile PBS, and immediately stored in RNAlater (Qiagen, Germany). RNA was extracted from the tissue using the RNEasy kit (Qiagen, Germany) according to the manufacturer's instructions. RNA quality was analyzed with the Agilent Bioanalyzer system using RNA microchips (Agilent Technologies, USA). Amplification of calprotectin and the housekeeping gene β-actin was performed using the following primers and probes: calprotectin probe (CTC TGC TAC TCC TTG TGG CTG TCT), calprotectin fw (AGT GTC CTC AGT TTG TGC AGA), calprotectin rev (CAG GGA CCC AGC CCT AG), β-actin probe (CCA GTT GGT AAC AAT GCC ATG TTC), β-actin fw (TAG GCA CCA GGG TGT GAT G), β-actin rev (TGC CAG ATC TTC TCC ATG TCG). Both probes were labeled with 6-carboxyfluorescein (FAM) on the 5′ and the Black hole quencher (BHQ) on the 3′ end. All probes were synthesized and tested by Microsynth (Switzerland) according to MIQE guidelines (Bustin et al., 2009). RT-PCR was performed in duplicates in a volume of 20 μL using the TaqMan RNA to C_t 1-Step kit (Applied Biosystems, USA) on a QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher, USA) with the following program: 15 min 48°C, 10 min 95°C followed by 40 cycles of 15 s 95°C, and 1 min 60°C. Fluorescence was measured for each cycle after then step of 60°C. C_t-values for both genes were averaged over both measurements and the mean C_t of calprotectin was subtracted from the mean C_t-value of β-actin to normalize.

Deuterated CDCA and deoxycholic acid were used as recovery standards. Approximately 10 milligrams of freeze-dried cecal content was homogenized with 150 μL of H₂O and spiked with 20 μL of recovery standards (100 μM), using an automated Precellys 24 bead-based homogenizer (5000 rpm; 1 × 20 s). Mixed samples were equilibrated on ice for 1 h. An amount of 500 μL of ice-cold alkaline (5% ammonia in ACN) was added to the homogenate, which was then vigorously vortexed, and continuously shaken for 1 h at room temperature. The mixture was centrifuged at 16,000 g for 10 min and the supernatant collected. The pellet was extracted with another 500 μL of ice-cold alkaline ACN. Supernatants from the two extractions steps were pooled and dried in a rotational vacuum concentrator (Christ, Germany) before reconstitution in 100 μL of 50:50 ammonium acetate (5 mM): methanol (MeOH) pH 6. The supernatant was kept at −20°C until LC-MS injection.

Both qualitative and quantitative analyses were conducted on a Agilent 6530 Accurate-Mass Q-TOF LC/MS mass spectrometer coupled to an Agilent 1290 series UHPLC system (Agilent Technologies, USA). The separation was achieved using a Zorbax Eclipse-Plus C18 column (2.1 × 100 mm, 1.8 μm; Agilent Technologies, USA) heated at 50°C. A binary gradient system consisted of 5 mM ammonium acetate pH 6 in water as eluent A and acetonitrile as eluent B. The sample separation was carried out at 0.4 mL/min over a 19 min total run time using the following program: 0–7 min, isocratic 21% B; 7–7.5 min, 21–25% B; 7.5–11.5 min, 24–30% B; 11.5–14.5 min, 30–32% B; 14.5–17.5 min, 32–63% B; 17.5–19 min, 63–100% B. The system was reequilibrated in initial conditions for 2 min. The sample manager system temperature was maintained at 4°C and the injection volume was 0.5 μL. Mass spectrometer detection was operated in negative ionization mode using the Dual AJS Jet stream ESI Assembly. The QTOF instrument was operated in the 4 GHz high-resolution mode (typical resolution 17,000 (FWHM) at m/z 1000) in profile mode and calibrated in negative full scan mode using ESI-L solution (Agilent Technologies, USA). Internal calibration was performed during acquisition via continuous infusion of a reference mass solution [5 mM purine, 1 mM HP-921 (Agilent reference mass kit, Agilent Technologies USA) in 95% MeOH acidified with 0.1% formic acid] and allowed to permanently achieve a mass accuracy better than 5 ppm. HR mass spectra were acquired over the range of m/z 300–700 at an acquisition rate of three spectra/s. AJS settings were as follows: drying gas flow, 8 L/min; drying gas temperature, 300°C; nebulizer pressure, 35 psi; capillary voltage, 3500 V; nozzle voltage, 1000 V; fragmentor voltage, 175 V; skimmer voltage, 65 V; octopole 1 RF voltage, 750 V. Data was processed using the MassHunter Workstation (Agilent Technologies, USA). According to this method, 36 bile acids (Table S1) were quantified by external calibration curves. Standard stock solutions were prepared at a concentration of 10 mM in methanol. Bile acid solutions were mixed together and diluted with 50:50 ammonium acetate (5 mM):MeOH to construct standard curves between 5 and 5000 nM. Extracted ions chromatograms (XIC) were based on a retention time (RT) window of ±0.5 min with a mass-extraction-windows (MEW) of ±30 ppm centered on m/z_theor of each bile acid.

16S rRNA specific primers and hydrolysis probes were designed using Primer Express 3 (Applied Biosystems, Life Technologies, USA). For duplex qPCR assays, hydrolysis probes were 5′-labeled with either 6-carboxyfluorescein (FAM) or 6-carboxyhexafluorescein (HEX). Each probe was conjugated with the black hole quencher 1 (BHQ1) at the 3′ end. Primers and probes were synthesized by Metabion (Planegg, Germany). qPCR conditions were established according to the MIQE guidelines (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) (Bustin et al., 2009). All primers were designed for an optimal annealing temperature of 60°C. qPCR standard curves were determined using linearized plasmid as DNA template. Plasmid DNA was diluted in H₂O containing 100 ng/μl yeast tRNA (Roche, Switzerland). Standard curves were run on a Roche Lightcycler96 instrument in triplicates. Efficiency of each qPCR reaction was calculated based on the slope of standard curves (qPCR efficiency: (10^{(−1/slope of standard curve)}−1) × 100) using 10-fold dilutions of template. Efficiencies for all qPCR reactions were within the range of 90–110%. For all experiments the software LightCylcer96 version 1.1 reproduced standard curves based on single DNA template with known DNA quantity as well as the efficiency derived from the standard curve of each qPCR assay of the initial run of the standard curves. Specificity was confirmed by performing an assay for each primer/probe pair using an equimolar mixture of all linearized plasmids except for the one to be tested as template. One PCR reaction (total volume: 20 μl) contained 300 nM of each primer, 250 nM of the corresponding hydrolysis probe (Table S2), FastStart Essential DNA Probes Master (Roche, Switzerland), and 5 ng template gDNA. PCR reactions with DNA templates extracted from feces or cecal content were run in duplicates. PCR conditions were: 95°C for 10 min, followed by 45 cycles of 95°C for 15 s, and 60°C for 1 min. Fluorescence for each cycle was recorded after the step at 60°C. Quantification cycle (Cq) as well as the baseline were automatically determined by the software LightCycler96 version 1.1 (Roche, Switzerland). The detection limit ranged between 1 and 248 16S rRNA gene copies/5ng template gDNA (Table S2).

Statistical analyses were performed as described in Figure Legends using GraphPad Prism 6 (Version 6.07) or R (Version 3.1.2). Hierarchical clustering in Figures 2, 4 used the dist function (Canberra distance) as well as the hclust function with the “complete” method. The NMDS analysis in Figure 5 used the function metaMDS of the R package vegan as well as the adonis function for statistical analysis of the differences between groups.

We first addressed the question of whether the sDMDMm2 mice colonized with the 12-species Oligo-MM¹² consortium (Brugiroux et al., 2016), devoid of any known 7-dehydroxylating microbial activity, at steady-state (without prior antibiotic treatment) are susceptible or resistant to CDI. We compared this new model to previously established CDI models including completely axenic (germ-free) mice (Pawlowski et al., 2010), antibiotic- (streptomycin- and clindamycin-) pretreated SPF mice (Buffie et al., 2011; Ng et al., 2013) and unmanipulated (and highly CDI resistant) SPF mice. All animals were infected with 10³ CFU of C. difficile spores of strain DH1916 (Ribotype 027; Burns et al., 2011) and sacrificed 72 h post infection. Quantification of C. difficile in feces at 6 and 24 h post infection, and in cecal content at 72 h post infection revealed high C. difficile colonization levels in germ-free, antibiotic-treated SPF and gnotobiotic sDMDMm2 mice (Figure 1A). In contrast, unmanipulated SPF animals, apart from the gastrointestinal transit of the inoculum detected around 6 h post-infection, showed no evidence for productive intestinal C. difficile colonization. In all CDI-susceptible mouse models that we tested, the levels of the intestinal inflammation markers lipocalin-2 (Figure 1B) and calprotectin (Figure 1C) were strongly elevated over the background levels measured in non-infected control mice and the CDI-resistant SPF mice. Large intestinal histopathology revealed marked mucosal inflammation in C. difficile-infected germ-free, antibiotic-treated SPF, and gnotobiotic sDMDMm2 animals, but not in the non-antibiotic-treated SPF mice (Figures 1D,E). Clindamycin-treated uninfected controls showed an elevated baseline histopathology of unknown etiology (see Figure 1D) compared to non-antibiotic-treated SPF mice. In summary, the sDMDMm2 CDI model is non-colonization-resistant against C. difficile and closely reproduces the colonization dynamics and pathogenesis of the widely used antibiotic-pretreatment mouse models.

The available metagenomic information of the Oligo-MM¹² consortium (Lagkouvardos et al., 2016) predicts that the Oligo-MM¹² microbiota is able to deconjugate bile acids in vivo, but that none of the 12 strains harbors the metabolic pathway for bile acid 7α-dehydroxylation to generate secondary bile acids. To test this hypothesis and to identify discrete bile acid metabolome profiles associated with CDI susceptibility in the different models we carried out targeted UHPLC-HRMS based metabolomics. Instead of C. difficile inoculation, we sacrificed cohorts of germ-free, streptomycin-treated SPF, clindamycin-treated SPF, unmanipulated SPF, and gnotobiotic sDMDMm2 mice and quantified 36 bile acid species (nine of which were below the minimal detectable concentrations in all samples and omitted from the analysis; see Figure 2A and Table S1) in cecum content extracts. An additional group included in this analysis were sDMDMm2 animals that had been colonized for 7 days with C. scindens, which we hypothesized would increase the large intestinal concentrations of secondary bile acids in vivo.

Heat map analysis with unsupervised hierarchical clustering (Canberra distance) of individual mice based on the quantities of each bile acid derivative showed distinct bile acid profiles that grouped the animals according to their experimental treatment (Figure 2B). Germ-free mice were distinguished from all microbially colonized groups by the near-complete absence of deconjugated primary as well as secondary bile acids. Bile acid deconjugation by bile acid hydrolases is a common intestinal microbial metabolic activity, and correspondingly absent from germ-free animals, but evident in all microbially colonized animal groups tested in this experiment. Both antibiotic-treated CDI infection models cluster with the untreated SPF animals but were characterized by an overall reduction of all bile acids, specifically a marked decrease in secondary bile acids and nearly undetectable amounts of DCA and LCA in particular. Germ-free animals, as well as sDMDMm2 + C. scindens and to some extent the sDMDMm2 gnotobiotes accumulated higher cecal concentrations of the conjugated primary bile acid TCA than SPF mice. Similar to the other CDI-susceptible groups, Oligo-MM¹²-colonized sDMDMm2 mice lacked most secondary bile acids. The exception was 7-oxodeoxycholic acid (7-oxoDCA), which remained at levels equal or higher than found in SPF control mice. Hence, CDI-susceptible animals were distinguished from CDI-resistant animals by reduced levels of the six secondary bile acid species DCA, 12-oxolithocholic acid (12-oxoLCA), LCA, ursodeoxycholic acid (UDCA), hyodeoxycholic acid (HDCA), and ωMCA (Figure 2C). Finally, of these six secondary bile acid species the levels of four, DCA, LCA, UDCA, and 12-oxoLCA were fully or partially restored to the concentrations found in SPF control mice by the additional colonization of sDMDMm2 mice with C. scindens.

Next, we analyzed how colonization with C. scindens and the concomitant changes in secondary bile acid concentrations correlated with the course of CDI. For this purpose a cohort of sDMDMm2 mice was infected with C. difficile 4 days after association with C. scindens. One sub-cohort was sacrificed 24 h after infection (day 5 after C. scindens colonization), the other 72 h after infection (day 7 after C. scindens colonization). At the time of sacrifice, we quantified C. scindens (Figure 3A) and C. difficile densities (Figure 3B), cecal mucosal pathology (by histopathological scoring, luminal lipocalin-2 protein, and mucosal calprotectin mRNA quantification; Figures 3C–F), and cecal luminal bile acid composition (by UHPLC-HRMS in sDMDM + C. scindens mice; Figure 4). At the time of infection, C. scindens had reached levels of 10⁸ −10⁹ CFU/g feces (Figure 3A). The C. scindens-associated mice at 24 h post infection showed partial resistance against C. difficile colonization compared to controls (Figures 3B–F), with significantly reduced fecal and cecal C. difficile loads and absence of marked cecal pathology. This partially protected phenotype was associated with the presence of secondary bile acid DCA in the cecal contents (Figures 4A,B, light blue symbols; reference dataset of uninfected sDMDMm2 mice colonized with C. scindens for 7 days from Figure 2 is shown for comparison; orange symbols, indicated by plus sign). However, we observed a complete loss of protection by 72 h post infection, when large intestinal colonization levels of C. difficile and cecal pathology were no longer significantly different from controls (Figures 3B–F). Median C. scindens densities were decreased at this time point compared to before infection with C. difficile. The loss of protection was further associated with a near-complete absence of microbial bile acid transformation products in cecum contents in these animals (Figures 4A,B; dark blue symbols). The concentrations of secondary and primary deconjugated bile acids were near or below detection limits, very similar to the bile acid profiles observed in germ-free animals (see Figure 2B). These data show that colonization with C. scindens impacts the early colonization dynamics but could not fully prevent C. difficile colonization and pathogenesis. This is consistent with the previously reported effects of DCA and LCA in inhibition of vegetative growth (Sorg and Sonenshein, 2008; Buffie et al., 2014) described in vitro and ex vivo. The association between CDI-induced intestinal pathology and the abolition of intestinal microbial bile acid transformations (deconjugation and 7-dehydroxylation) to germ-free mouse-like levels prompted us to analyze the effect of CDI on the large intestinal Oligo-MM¹² microbiota composition.

To analyze the impact of CDI on sDMDMm2 microbiota composition, sDMDMm2 mice were pre-colonized for 7 days with C. scindens followed by infection with C. difficile for 72 h (or uninfected controls). For comparison, we analyzed non-pretreated sDMDMm2 mice infected with C. difficile for 72 h (and uninfected controls). We analyzed the bacterial consortium composition over time in the feces and at the endpoint in cecal contents using an established 16S rRNA-based TaqMan qPCR method (see materials and Methods; Brugiroux et al., 2016) including probes for all Oligo-MM¹² species as well as C. scindens and C. difficile (Figure 5A, Figure S1). We quantified the fecal (−7 days, −4 days, 0 h, 24 h post infection) as well as cecal (72 h post infection) Oligo-MM¹² composition C. scindens-associated and non-associated mice to determine changes in response to C. scindens colonization alone, as well as in response to C. difficile infection (Figure 5A, Figure S1). Non-parametric multidimensional scaling (using the metaMDS function of the vegan package in R) with the qPCR data of the 12 bacterial strains of the Oligo-MM¹² microbiota (excluding the quantities of C. difficile and C. scindens) of all fecal samples showed that the colonization of sDMDMm2 animals with C. scindens led to minor changes in Oligo-MM¹² composition (compare yellow and green symbols in Figure 5B). These changes were mainly driven by variability in the densities of C. clostridioforme and A. muris sp. nov. (Figure S1) over the time of the colonization with C. scindens and were not statistically significant.

As in non-pretreated sDMDMm2 mice, sDMDMm2 + C. scindens associated mice after 72 h of infection with C. difficile showed a severely perturbed cecal microbial consortium, whose species composition differed significantly from uninfected control animals (Figures 5A,C, p = 0.008 for sDMDMm2; p = 0.001 for sDMDMm2 + C. scindens). Pre-colonization with C. scindens had no statistically significant influence on this outcome (Figure 5C, p = 0.145, compare blue and red circles). The observed shift in microbiota composition was associated with a decreased density of C. scindens, as well as most of the Oligo-MM¹² bacterial species, reflected in a decreased total bacterial density (Figure S1).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.