Mice were fed with a Normal Chow diet (NC), High Fat (HF, 25% fat) or High-Fat High-Sucrose diet (HFHS, 25% fat, 30% sucrose) for 8 weeks (Figure 1A). No differences were observed in food and water consumption, but weight gain were increased in HFHS-fed mice and blood glucose were increased in both HFHS and HF-fed compared to NC-fed mice (Figures 1B–D). Dietary intake was quite stable (2–4 g/day) under ad libitum conditions, whereas energy intake was higher after HFHS (4,775.7 kcal/kg) and HF (4,649 kcal/kg) than NC (3,339 kcal/kg), as expected. Consequently, body weight over the 8-week dietary intervention was significantly higher in HFHS-fed mice than NC-fed mice or the HF (Figure 1B). Blood glucose levels were measured after eight weeks of diets and 12-h fasting (Figure 1C), and metabolic tests (oGTT and ITT) confirmed that only HFHS diet induced prediabetes characterized by reduced impaired glucose tolerance by insulin resistance (Figures 1D,E).

Unexpectedly, endoscopic lesions were observed in the colonic mucosa of HFHS-fed and HF-fed mice not treated with Dextran-sulfate sodium (DSS) (Figure 1F), and the endoscopic score (Figure 1G, Supplementary Table 1) was significantly higher (1.4-fold increase; p = 0.0089) than in NC-fed mice not treated with DSS. These endoscopic lesions were associated with a thinning and shortening of the colon length (-10.9 mm; p = 0.012) (Figures 1L,M). At the histological level (Figures 1H,I, Supplementary Table 2), some HFHS-fed animals showed a chronic inflammatory cell infiltrate (lymphoplasmacytic, Supplementary Table 3), with no significant increase in the global score (Figure 1I).

Then we used DSS colitis as a control and we confirmed that clinical manifestations of colitis appeared earlier in HFHS-fed mice (day 3) and that the colitis was more severe, as reflected by decreased body weight (Supplementary Figure 1A) and the higher Disease Activity Index (DAI) in the HFHS DSS⁺ group compared with NC DSS⁺ animals (3.8 vs. 2.7 at day 10, respectively; p < 0.0001) (Supplementary Figure 1B). Likewise, the endoscopic score indicated a severe colonic inflammation (Supplementary Figures C,D) that was also confirmed histologically (Supplementary Figures 1E,F, Supplementary Table 1). This was further supported by the thinning and shortening of the colon length (Supplementary Figures 1G,H). In agreement with these findings, an enlarged spleen was observed in either HFHS DSS⁺ mice (about larger 3-fold maximum size compared to NC DSS⁺; p = 0.0001) (Supplementary Figures 1K,L) or HFHS DSS⁻ mice (1.5-fold on average compared to NC DSS⁻; p = 0.11; Figures 1J,K). In contrast, no significant decrease of colon length was observed in HF-fed mice, even with colitis (Supplementary Figures 1I,J). On day 5 of DSS-treatment, intestinal permeability was measured by FITC-Dextran, showing that only HFHS-fed mice treated with DSS had increased permeability compared to both NC or HF DSS⁺ group (Supplementary Figure 1D). Furthermore, HFHS increased early mortality during DSS-colitis whereas HF-fed mice have similar mortality to NC-fed (30% of HFHS-fed mice from the 3rd day; p = 0.0066 vs. 20% of NC or HF-fed mice died after 7 days; Supplementary Figure 1C).

Altogether, these data showed that sucrose added to fat (rather than fat alone) aggravates the severity of DSS-induced colitis clinically, endoscopically, and histologically. Under physiological conditions, such diet induces a pre-IBD state characterized by endoscopic lesions, reduced colon length, and inflammatory cell infiltration in the colonic mucosa.

To determine the respective impact of fat vs. sugar on the immune system, we first performed cytometry analysis on the spleen and MLNs of mice. Interestingly, we observed that all studied immune cells population were decreased in MLNs of HFHS-fed mice, whereas they were increased in HF-fed mice compared to NC-fed mice (Figures 2A–F). Specifically, numbers of leukocytes, myeloid cells, and lymphocytes, including B and T cells, were decreased in HFHS-fed mice, whereas they were increased in HF-fed mice compared to NC-fed mice (Figures 2A–F). However, these cells were not significantly disturbed in the spleen (Figures 2G–K) except myeloid cells representing 8.5% of leukocytes in HFHS-fed mice, while they only represented 2.1% in NC-fed mice (Figure 2H, Supplementary Figure 2A). Likewise, T-cells represented 25.9% of lymphocytes in HFHS-fed mice vs. 31.9% in control animals (Figure 2J, Supplementary Figure 2A). Among the lymphocytes in the MLNs, while NK cells amounted to an average of 18.7% CD45+CD3- cells in NC-fed mice, they expanded to 24.6% in HFHS-fed mice and 45.8% in HFD-fed mice (Supplementary Figure 2A). Likewise, T-cell subpopulations, including CD4+ and CD8+ cells, were strongly decreased in HFHS-fed mice MLNs but not significantly in the spleen, compared to control animals (Figures 2L–O). Significantly, Th17 cells were strongly decreased in MLNs of HFHS-mice and slightly increased in the spleen than NC-fed animals (Figures 2P,U).

We next focused on HFHS-fed mice, and to further dissect the local and systemic immune response, we measured cytokine levels in plasma and colonic tissues. As expected, cytokines levels in plasma (Supplementary Figures 2B–D) were significantly increased in DSS-treated mice fed with NC compared to untreated animals (IL-1β: +25.39 pg/mL, p = 0.0212; IL-6: +77.34 pg/mL, p = 0.0005; TNF-α: +108.8 pg/mL, p = 0.0351). Unexpectedly, these cytokines were not increased in plasma of colitic HFHS-fed mice compared to not-treated HFHS-fed animals excepted for IL-6 (+75.56 pg/mL, p = 0.0002; Supplementary Figure 2C). Then we measured levels of several chemokines in the colonic tissues of HFHS-fed mice. In contrast to plasma, chemokine levels were higher in colonic tissues of HFHS-fed colitic mice compared to DSS-untreated animals (RANTES: +295.0 pg/mL, p = 0.015; Eotaxins: +760.7 pg/mL, p = 0.037 and KC: +55.67 pg/mL, p = 0.026; Supplementary Figures 2E–G).

In summary, these findings indicate that fat and sugar induced similar intestinal lesions but have a differential effect on the immune system. In contrast to HF that increased immune cells, sucrose enrichment decreases the number of immune cells in both MLNs and the spleen. While HFHS causes impaired systemic cytokine response, chemokines recruitment is maintained in the colon of these animals.

To better understand how HFHS may predispose to disease risk in later life, transcriptomic analyses were performed on colonic tissues of DSS-untreated mice. Interestingly, in healthy mice, HFHS caused a strong gene down-regulation compared to NC-fed mice (Figure 3A). Statistical analyses identified 154 differentially expressed genes, 86% (132) being under-expressed in HFHS-fed mice (FDR <0.05; fold change > 30%) (Supplementary Dataset 1). These genes were predominantly involved in response to diverse cellular stresses such as HSF1-mediated heat-shock, external stimuli, unfolded protein stress via HSP90 chaperone complex, and to a lesser extent in the regulation of transcription and translation, as well as in Class I MHC mediated antigen processing and presentation (Figure 3B). Functional annotations with OpenTargets confirmed an association between the HFHS gene signature and nutritional and metabolic diseases such as diabetes (62 targets), abnormalities of the immune system (75 targets), and infectious diseases (104 targets). Interestingly, they also unraveled an association between HFHS and “colitis” (37 targets) or IBD (47 targets) despite the absence of DSS treatment, which is consistent with the pre-IBD state observed at the endoscopic and histologic levels (all p-values < 0.001) (Supplementary Table 4). Among the upregulated genes in HFHS-fed mice, we observed Shisa5, which encodes an endoplasmic reticulum protein involved in the apoptotic process (+1.5-fold; FDR = 0.029) and Slamf7, involved in cell-cell interactions by modulating the activation and differentiation of a wide variety of immune cells and in response to anti-TNF therapy in IBD (24, 25) (+1.6-fold; FDR = 0.046). Conversely, Rgs1, which encodes for a protein inhibiting B cells chemotaxis, was strongly under-expressed (−2.22-fold; FDR = 0.033) in HFHS-fed mice (Figure 3C), while being overexpressed in active IBD (26). Stress-related genes were under-expressed in HFHS-fed mice: Hsph1, which encodes a member of the heat shock protein 70 family (−3-fold; FDR = 0.036), Hspa8 (−2.5-fold; FDR = 0.024), Hspa5 (−2.1-fold; FDR = 0.013), and Rb1cc, indirectly implicated in autophagy via Atg16L1 interaction (−1.9-fold; FDR = 0.025), all coding for heat shock proteins (HSP) involved in autophagy and endoplasmic reticulum stress. Moreover, chaperones associated with these HSP were also downregulated in HFHS-fed mice: Dnajb1 (−1.8-fold; FDR = 0.0002) and Dnaja1 (−2.1-fold; FDR = 0.0005), both implicated in the regulation of heat shock proteins, or Chordc1, encoding a protein proposed to act as co-chaperone for HSP90 (−2.2-fold; FDR = 0.003) and Ahsa2, a pseudogene proposed as an HSP90 stimulating co-chaperone (−2.2-fold; FDR = 0.003) (Figure 3D). Overall, HFHS modulated several molecular pathways involved in cellular stress-related pathways in the healthy intestine of mice.

Prior to colitis, we investigated the impact of diet on the fecal microbiota by comparing its diversity and composition before starting the different diets (baseline or B) and after eight weeks of dietary intervention (day zero or D0). We observed that HFHS-fed mice had a lower α-diversity at baseline for the Chao (Figure 4A) and Shannon (Supplementary Figure 3A) indices because of the acclimatization period (diet was not the same throughout the acclimatization period and the experiment). To evaluate the diet's impact beyond the initial founder effect, we calculated a fold change of the α-diversity indices at D0 relative to baseline. We observed no significant difference between each diet group's fold changes, both for Chao (Figure 4B) and Shannon indices (Supplementary Figure 3B). These results indicated that the dietary intervention did not strongly influence the α-diversity of the intestinal microbiota. Considering β-diversity, we observed a strong association with the diet for each time point (Supplementary Table 5), and for both Bray-Curtis and Jaccard indices (Figure 4C, Supplementary Figure 3C), with explained variance >42%. However, within diet groups, a significant difference between time points was observed only for HFHS-fed mice, with explained variance >25% for Jaccard index and >71% for Bray-Curtis index (Supplementary Table 5). These results indicate that despite the initial difference between animal groups, the HFHS-fed mice had an essential change in the composition of their fecal microbiota, strongly affecting core abundant bacteria, and to a lesser extent, the prevalence of less abundant species. Consistently, we identified five Operational Taxonomic Units (OTUs) indicators for diet, with OTUs defined as follows: i) abundance/prevalence is significantly associated with diet at D0 but not at baseline, and ii) fold change between D0 and B is significantly associated with diet (Supplementary Figure 7). Precisely, two OTUs belonging to the order Clostridiales (Otu036, Otu044) showed an apparent decrease in their relative abundance in HFHS-fed mice while remaining at a similar abundance in NC-fed mice (Figures 4D,E, Supplementary Figure 3). Two OTUs belonging to the family Lachnospiraceae (Otu052, Otu146) showed a strong increase of their relative abundance and prevalence in HFHS-fed mice (Figures 4F,G, Supplementary Figure 3, Supplementary Table 5). One OTU belonging to the genus Barnesiella (OTU194) was not found at baseline but was ubiquitous at D0 in NC-fed mice, while present in only one HFHS-fed animal, in lower abundance (Figure 4H, Supplementary Figures 3P–U). These results suggest that HFHS can significantly impact individual bacteria, most of them belonging to Firmicutes, known to produce short-chain fatty acids (SCFAs) (27) (Supplementary Figure 7). These biomarkers represent together more than 90% of the OTUs, while Firmicutes represent ~50% of the entire community, indicating that bacteria belonging to this phylum might be significantly impacted by the dietary intervention.

We next measured SCFAs concentration in caecal contents and feces of mice, and showed that microbial changes induced by HFHS were followed by a significant decrease in caecal and fecal SCFAs concentrations (Figures 4I,J). Indeed, caecal propionate and butyrate concentrations were decreased, whereas isovalerate concentration increased in HFHS-fed mice compared to control animals (p < 0,05) (Figure 4I). Likewise, fecal acetate, propionate, and especially butyrate concentrations were profoundly reduced in HFHS-fed mice compared to control animals (p < 0,05) (Figure 4J).

To highlight the role of the gut microbiota in this experimental context, we performed fecal microbiota transplantation (FMT) in germ-free (GF) mice before inducing DSS-colitis (Supplementary Figure 4A). Despite the initial difference in GF-mice bodyweight before FMT, those who received microbiota from HFHS-fed mice recovered their initial weight after transplantation, contrary to NC-fed mice transplanted (Supplementary Figure 4B). Then, GF-mice that received microbiota from HFHS or HF-fed mice developed higher colitis than mice transplanted with fecal microbiota from control mice, as illustrated by higher bodyweight loss (Supplementary Figure 4D), survival rate (Supplementary Figure 4C). Importantly, at day 8, GF-mice that received HFHS-fed microbiota had an increased DAI (Supplementary Figure 4E).

To assess whether the changes induced by HFHS could be restored, we analyzed the gut microbiome and transcriptome of mice fed eight weeks with NC after eight weeks of HFHS (Figure 5A). Colonic transcriptome reprogramming detected in HFHS-fed mice was only partially reversible in HFHS+NC-fed mice (Figure 5B). Detailed results are described in Supplementary Material. Briefly, further hierarchical clustering identified six dysregulated signatures in HFHS+NC-fed mice, predominantly involved in unfolded protein response (UPR), autophagy, heat shock protein response, and metabolic pathways that play a crucial role in intestinal homeostasis (Figure 5B, Supplementary Table 6). Among upregulated genes in HFHS, many genes involved in p53/TP53-dependent apoptosis returned to regular expression levels (as observed in NC-fed mice) after switching diet while many HSP-related genes were down-regulated in HFHS-fed mice failed to recover their baseline expression in HFHS+NC-fed mice (Figures 5C–E). Regarding disease-associated targets, genes from clusters 2 and 5 were still associated with immune system diseases (75 targets) and gastrointestinal diseases (60 targets) including IBD (27 targets), ulcerative colitis (20 targets), Crohn's disease (15), “colitis” (22 targets), and digestive system infectious diseases (44 targets) (all p-values < 0.09; Supplementary Table 7).

Similar observations were made on the composition of the gut microbiota. Detailed results are described in Supplementary Material. Briefly, NC-fed mice did not vary in α-diversity, while HFHS-fed mice had a lower diversity at D0 compared to baseline, and the HFHS+NC-fed group had a higher diversity at D0 compared to baseline only for the Chao index (Figures 6A,B). Regarding β-diversity, all groups were significantly different from each other at D0, but the HFHS-fed mice showed the highest distance to the other groups (Figure 6C). Among the indicators identified in the eight-week protocol, Otu036, Otu052, and Otu194 showed an interesting pattern, where the HFHS+NC-fed mice were in line with NC-fed mice and opposed to HFHS-fed mice, despite an initial difference according to diet group (Figures 6D–I, Supplementary Figure 7).

These results indicate that (i) HFHS has a consistent effect on microbial composition, given that four out of five biomarkers showed a similar pattern in both protocols, and (ii) a return to NC diet after eight weeks of HFHS can restore microbial abundances for four out of five diet-specific species identified in the eight-week protocol. Consistently, colitis severity was improved in HFHS-fed mice (Supplementary Figure 6).

Taken together, these results suggest that the long-term harmful effects of HFHS on gut microbiota and transcriptome can be partially corrected even after the return to a normal chow: some transcriptomic imprints lasted, and some effects persisted after exposure to HFHS on the gut microbiota.

Male C57BL/6J mice were purchased from Janvier Labs (Le Genest-Saint-Isle, France). All procedures were performed following guidelines established by the European Convention for the protection of Laboratory Animals (54) and with a project approval (authorization N°APAFIS#16630) delivered by the French ministry of research. The animals were maintained on a strict 12-h light-dark cycle and were housed at 22–23°C, in cages containing a maximum of 5 animals, with ad libitum access to food and water. All animal experiments were repeated at least two times on two separate occasions. Before dietary interventions, mice were randomized to ensure that no incidental pre-diet differences in body weight existed between the different groups. Mice were fed with a High-fat high-sucrose Diet (HFHS; U8960 v5, Safe Diets, Augy, France), High Fat diet (HF U8960 v143, Safe Diets, Augy, France), or Normal Chow diet (NC; A04, Safe Diets, Augy, France) for 8 or 16 weeks (Figure 1A).

For each dietary regimen (NC, HFHS, or HF), half of the animals were treated with DSS to induce colitis after eight weeks of diet (Figure 1A). Mice were continued on the same diet during colitis, induced by administration of 3% dextran sulfate sodium (DSS, molecular weight 36 000–50 000, MP Biomedicals, Strasbourg, France) dissolved in drinking water for five days. DSS solution was replaced thereafter by normal drinking water for another five days as previously described (55). Mice were sacrificed on the 10th day or upon reached the endpoint (i. e. > 20% weight loss).

Germ-Free (GF) C57BL/6J mice were housed at gnotobiotic facilities at the Hannover Medical School (Hannover, Germany), and Fecal Microbiota Transplantation (FMT) experiments were performed at Technical University Dresden (Dresden, Germany). For microbial reconstitution of GF mice, 8-week-old GF C57BL/6J mice received fecal material obtained from individually WT C57BL/6J HFHS-fed or HF-fed or NC-fed mice. Stool pellets were freshly dissolved in PBS with 0.05% L-cysteine hydrochloride (Sigma-Aldrich; 4 pellets per 5 ml of PBS), and mice received 100 μl solution by gavage. After one week, colitis was induced by administration of 2% DSS dissolved in drinking water for five days. DSS solution was replaced thereafter by normal drinking water for another three days.

For gut microbiota analysis, fresh stool samples from isolated mice were collected in tubes, immediately frozen in liquid nitrogen upon collection, and stored at −80°C until DNA isolation. Blood was collected by submandibular bleeding (56) in heparinized tubes (Microtainer® BD medical, Franklin Lakes, NJ), and the plasma was obtained by centrifugation at 2000 xg > 3min at room temperature. The entire colon was removed from the caecum to the anus. The colon length was measured, then washed with phosphate-buffered saline PBS 1X to remove the remaining content. The spleen was removed and weighed as previously described (57). Samples from the colon were taken, divided into 0.5 cm long pieces, snap-frozen in liquid nitrogen, and stored at −80°C until use.

Metabolic tests were performed with 10 mice per group using standard procedures (58). Oral glucose tolerance tests (OGTT) were performed with mice fasted overnight for 6 h. Blood glucose concentrations were determined before and after administering glucose solution orally via gavage (2g glucose/kg). Blood glucose levels were determined at defined post-gavage time points by analyzing blood obtained from the tail vein with a portable glucometer (Glucometer OneTouch® Verio Reflect, LifeScan Europe GmbH, Zug, Switzerland). For insulin tolerance tests (ITT), mice were fasted for 3 h and then injected with human insulin (0.75 U/kg i.p.; NovoRapid®, NovoNordisk A/S, Denmark). Blood glucose levels were monitored as described for the OGTT assay.

On the day of the assay (day 5 of DSS-treatment), 4 kDa fluorescein isothiocyanate (FITC)-dextran (Sigma-Aldrich, St. Louis, MO) was dissolved in PBS 1X to obtain a stock solution of 80 mg/mL. Mice fasted for 4 h before gavage with 60 mg/100g. Blood was collected 3 h following gavage on heparinized tubes (Microtainer® BD medical, Franklin Lakes, NJ). FITC-Dextran was quantified in plasma by fluorescence (γ Ex/Em 485 nm/535 nm, ID3, Molecular devices) as previously described (11).

Mice were daily weighed and evaluated for clinical symptoms. The Disease Activity Index (DAI) was determined on a scale from 0 to 4 and calculated as the mean of three individual subscores (body weight loss, stool consistency, and blood in the stool) as previously described (59).

Coloscopy was performed on the last day of the study (day 10), just before the mice were sacrificed. Mice were anesthetized by isoflurane inhalation. Distal colon and rectum were examined using a rigid Storz Hopkins II mini-endoscope (Storz, Tuttlingen, Germany) coupled to a basic Coloview system (with a xenon 175 light source and an Endovision SLB Telecam; Storz). All images were displayed on a computer monitor and recorded with video capture software (Studio Movie Board Plus from Pinnacle, Menlo Park, CA). Three independent trained-readers performed a blind determination of endoscopic scores. Ulcerative colitis endoscopic index of severity (UCEIS) was calculated as the sum of three subscores (vascular pattern, bleeding, and erosions/ulcers), as previously described (60). The final grade was defined as the mean of the three independent assessments.

Colon samples were extensively washed and fixed with 4% paraformaldehyde for 24 h under agitation. Samples were washed, paraffin-embedded, and sectioned. Colitis was histologically assessed on 5 μm sections stained with hematoxylin-eosin-saffron (HES) stain. The histological colitis score was calculated blindly by expert pathologists, as previously described (61). Briefly, disease scoring based on six histological features: acute inflammatory cell infiltrate (polymorphonuclear cells in the lamina propria), crypt abscesses, mucin depletion, surface epithelial integrity, chronic inflammatory cell infiltrate (round cells in the lamina propria), and crypt architectural irregularities. Each feature was graded on a 4-point scale corresponding to none, mild, moderate, or severe. The final grade was defined as the mean of the two independent assessments.

Mesenteric lymph nodes and spleen samples were washed and conserved in complete RPMI (10 % FCS). Samples were first grinded on a nylon mesh (40 μm cell strainer, Greiner), and large debris was removed. Cell suspensions were filtered through a 70 μm mesh (Miltenyi Biotec) and centrifuged at 4 °C, 600 xg, for 5 min. Cell suspensions were then washed and resuspended in phosphate-buffered saline (PBS) supplemented with 2% bovine serum albumin (BSA) for counting on an automated cell before direct cell surface staining. For intracellular staining, cells were fixed and permeabilized with a commercially available fixation/permeabilization kit (eBioscience). Single-cell suspensions were stained with antibodies to the following markers: CD45-BV421, CD3-APC/Fire750, CD4-FITC, CD25-BV510, CD8a-BV711, CD335-PECy7, CD19-BV605, CD11b-PERcpCY5.5, RORγt-APC, and Foxp3-PE in presence of FCBlock CD16/32; all antibodies were obtained from Biolegend. The gating strategy is detailed in Supplementary Figure 8. Gallios cytometer (Beckman) was used for cell acquisition, and the flow cytometry data were analyzed with Kaluza software.

Total protein was extracted from colonic tissue by lysing homogenized tissue with Bio-plex Cell lysis kit (Bio-rad #171304011, Bio-Rad Laboratories Inc, USA) and quantified by using the bicinchoninic acid assay method. All samples were normalized to 900 μg/mL for the assay. Cytokine levels in plasma and colonic lysates were measured using Bio-Plex Pro mouse cytokines 23-plex assay group I assay (Bio-Rad #M60009RDPD, Bio-Rad Laboratories Inc, USA), a BioPlex 200 instrument, and Milliplex Analyst software. Each biological replicate was assayed in technical duplicate. According to the manufacturer's instructions, protein concentrations were determined based on a standard curve described by the manufacturer-provided protocol and values for Lot #64212941.

Total RNA was extracted from 20 mg of colon sample using Trizol (Invitrogen, Carlsbad, California, USA) according to the manufacturer's recommendations. RNA was further digested with TurboTM DNase (Thermo Fisher, Waltham, Massachusetts, USA) and phenol-chloroform extracted. The quality of total RNAs was attested by O.D. 260 nm/O.D. 280 nm and O.D. 260 nm/O.D. 230 nm determination by spectrophotometry using a Nanodrop ND-1000 (NanoDrop Technologies; Wilmington, DE, USA) and using a 2,100 Bioanalyzer (Agilent Technologies; Massy, France). Samples with RNA integrity number > 8 were selected. RNA samples were aliquoted and stored at −80 °C, and later amplified and biotinylated using the Affymetrix GeneChip® WT PLUS reagent kit, then hybridized to Affymetrix Mouse Clariom S cartridges following the manufacturer's instructions. The arrays were washed and scanned according to the protocol GeneChip® Expression Wash, Stain and Scan for Cartridge Arrays.

Fluorescence values corresponding to raw expression data for every 37 samples were extracted from CEL files using the R oligo package (https://bioconductor.org/) with the corresponding platform definitions (pd.clariom.s.mouse). Positive and negative control probes were removed, which left 21.881 probes, each corresponding to a unique and well-annotated gene. Quality control steps, data normalization, and unsupervised explorations were conducted as described previously (62). Corrections for batch effects between microarrays were performed with Combat (package sva from Bioconductor). Statistical analyses were achieved using linear modeling with empirical Bayes, p.values were computed by applying a moderated two-way t-test and adjusted for false discovery rate (FDR) following the Benjamini–Hochberg procedure. Hierarchical clustering heat maps were obtained on gene-median-centered data with uncentered correlation as a similarity metric. Volcano plots were rendered using EnhancedVolcano (Bioconductor). Functional annotations were performed with the OpenTargets platform (63) for disease associations and gene ontology enrichments and ReactomePA (64) for pathway analyses in mice. For all experiments, p.values < 0.001 or FDR < 0.05 indicated statistical significance, depending on the test availability and/or relevance.

Bacterial profiling using 16S amplicon sequencing was performed as described previously (65). Briefly, DNA from fecal pellets was isolated using the PowerSoil DNA Isolation Kit (MoBio) according to the manufacturer's directions. Individual amplicons were tagged with specific multiplex identifier (MID) barcodes and pooled for library construction before sequencing. The 16S rRNA gene variable region V3-V4 was amplified and sequenced on an Illumina MiSeq with 2 × 250 bp. Raw sequences were first trimmed to remove bad quality tails and filtered for size using CUTADAPT (66) v1.1.4. Then USEARCH (67, 68) v10.0.240 was used to merge forward and reverse reads. Merged reads were filtered for size and expected error rate before dereplication and denoising, simultaneously removing chimeras. Reads present in less than two copies were filtered out as they likely represent PCR or sequencing errors. A custom R (69) script was used to transfer sequences into MOTHUR (70) v1.39.5, with which further processing was done according to the MiSeq SOP (71). Briefly, sequences were aligned to the SILVA (72) v132 reference database, classified against the RDP release 11 (trainset 16) (73) database, binned into operational taxonomic units (OTUs) at the species-level 97% identity threshold, and rarefied to 12,000 sequences per sample. A detailed account of the procedure is available in Supplementary Dataset 2. This procedure filtered 4 763 346 sequences for 147 samples, and lead 3 152 719 sequenced to be retained (66%) before subsampling (Supplementary Dataset 3). Additional formatting scripts are available in Dataset S4. The OTU Table obtained from MOTHUR was imported in R (69) for statistical analysis. Alpha diversity indices (Chao and Shannon) were calculated with vegan (74) package version 2.5–6. Fold changes between time points were calculated as: log2 (target value/reference value). Correlations between α-diversity indices or fold change and experimental parameters were tested using the Wilcoxon rank sum test. Beta diversity was evaluated with the vegan (74) package version 2.5–6, using the Bray-Curtis and Jaccard distances to build principal coordinate analysis (PCoA). Fold changes of OTU abundances between time points were calculated as: log2 [(target value+0.1)/ (reference value+0.1)]. The addition of 0.1 to both values avoids infinite values due to zero abundances, while keeping the ratios in the same trend. Correlations with experimental parameters were assessed with multivariate analysis of variance using package pairwiseAdonis (75). Biomarkers were found by correlating individual OTUs with experimental parameters with four complementary tests: Kruskal-Wallis rank sum test was done on OTU abundances and fold changes (i) or non-zero OTU abundances (ii); Pearson's chi-square test was performed on OTU prevalence (iii); indicator species analysis (iv) was conducted with package indicspecies (76). To evaluate the strength of association, Eta was calculated for Kruskal-Wallis tests and Cramer's V for chi-square tests. Where applicable, p-values were adjusted for multiple testing using the false discovery rate (“FDR”) correction. All necessary data and scripts for analysis are available in Supplementary Dataset 5.

The concentrations of short-chain fattyacids (SCFAs; i.e. acetate, propionate, butyrate, valerate, caproate, isobutyrate, isovalerate, and isocaproate) were measured in the cecal contents and stool samples of mice. Analyses were performed as previously described (77). Samples were water-extracted, and proteins were precipitated with phosphotungstic acid. Analyzes were performed by gas chromatography using a system (Autosystem; Perkin-Elmer, St. Quentin en Yvelines, France) equipped with a split/splitless injector, a flame ionization detector, and a capillary column (length, 15 m; inside diameter, 0.53 mm; film thickness, 0.5m) impregnated with SP100 (Nukol; Supelco, Saint-Quentin-Fallavier, France). The internal standard used was 2-ethylbutyrate. All samples were analyzed in duplicate. Data were collected, and peaks were integrated using Turbochrom software (Perkin-Elmer, Courtaboeuf, France).

All data are expressed as means ± SEM. Two-group comparisons were performed using Mann-Whitney U-test for non-parametric data or t-test for parametric data. ANOVA with Tukey's post-hoc test for parametric data or Kruskal Wallis with Dunn's correction for non-parametric data were used to compare more than two groups. p values < 0.05 were considered significant. Statistical details and the exact value of “N” can be found in the Figure legends. Statistical analyses were performed with GraphPad Prism 6 software (GraphPad Software, Inc, La Jolla, CA). All authors had access to the all data and have reviewed and approved the final manuscript.