Intestinal gluconeogenesis shapes gut microbiota, fecal and urine metabolome in mice with gastric bypass surgery

Intestinal gluconeogenesis (IGN), gastric bypass (GBP) and gut microbiota positively regulate glucose homeostasis and diet-induced dysmetabolism. GBP modulates gut microbiota, whether IGN could shape it has not been investigated. We studied gut microbiota and microbiome in wild type and IGN-deficient mice, undergoing GBP or not, and fed on either a normal chow (NC) or a high-fat/high-sucrose (HFHS) diet. We also studied fecal and urine metabolome in NC-fed mice. IGN and GBP had a different effect on the gut microbiota of mice fed with NC and HFHS diet. IGN inactivation increased abundance of Deltaproteobacteria on NC and of Proteobacteria such as Helicobacter on HFHS diet. GBP increased abundance of Firmicutes and Proteobacteria on NC-fed WT mice and of Firmicutes, Bacteroidetes and Proteobacteria on HFHS-fed WT mice. The combined effect of IGN inactivation and GBP increased abundance of Actinobacteria on NC and the abundance of Enterococcaceae and Enterobacteriaceae on HFHS diet. A reduction was observed in the amounf of short-chain fatty acids in fecal (by GBP) and in both fecal and urine (by IGN inactivation) metabolome. IGN and GBP, separately or combined, shape gut microbiota and microbiome on NC- and HFHS-fed mice, and modify fecal and urine metabolome.


Abbreviations
The worldwide rise in the prevalence of obesity is associated with increased incidence of various metabolic disorders. In this context, bariatric surgeries have emerged as the most effective therapies to treat obesity and one of its comorbidities such as type 2 diabetes mellitus (T2DM) 1,2 . The Roux-en-Y gastric bypass (GBP) procedure is one of the most performed and efficient bariatric surgeries. After GBP, patients exhibit reduced food intake and a considerable and long-term weight loss of up to 30% 2,3 . Improvement or even remission of T2DM and its associated complications as hepatic steatosis and fibrosis is observed in most patients 3,4 . However, the improvements in glycaemic control cannot be exclusively correlated to weight loss. Indeed, many type 2 diabetic patients

Effects of intestinal gluconeogenesis inactivation on gut microbiota and microbiome in normal chow and high-fat high-sucrose fed mice.
Please note first that throughout the manuscript the term microbiota refers to ecological structure (relative abundance in %) of gut microbes, whereas the term microbiome refers to microbial inferred functions.
We assessed the effect of IGN inactivation on both gut microbiota and microbiome in NC-fed mice (these mice underwent a laparotomy (identified as "Lap" group) to serve as control for mice that underwent GBP). On a NC, iG6PC-KO mice had a statistically significant higher relative abundance of bacteria from phylum Proteobacteria, such as Desulfovibrio, and from phylum Bacteroidetes, such as Odoribacter, Alistipes, Rikenella as well as bacteria from phylum Firmicutes such as Candidatus Arthromitus, compared to WT mice (Fig. 1A). The overall diversity of gut microbiota of NC-fed iG6PC-KO was significantly different from that of WT mice, mostly based on Chao-1 index (Fig. 1B) as confirmed by comparison of overall microbial profiles (Fig. 1C). In addition, the microbiome of NC-fed iG6PC-KO exhibited an imputed functional statistically significant increase of nitrogen metabolism (Fig. 1D).
Then, we studied the impact of IGN inactivation on gut microbiota and microbiome of both WT and iG6PC-KO mice fed a high-fat/high-sucrose (HFHS) diet. The major change was the statistically significant higher abundance of genus Helicobacter (Proteobacteria phylum) in the gut microbiota of HFHS-fed iG6PC-KO mice compared to WT mice (Fig. 1E). The overall diversity of gut microbiota of HFHS-fed iG6PC-KO was significantly different from that of WT mice, mostly based on Chao-1 index (Fig. 1F) as confirmed by comparison of overall microbial profiles (Fig. 1G). In addition, the microbiome of HFHS-fed iG6PC-KO showed two statistically significant higher bacterial motility-related inferred functions (Fig. 1H). Altogether, these data indicate that IGN inactivation affects gut microbiota composition and functions regardless of diet.
Intestinal gluconeogenesis inactivation reduces acetate in fecal and urine metabolome. Since IGN inactivation affects gut microbiota composition and functions on both NC-and HFHS-feeding, as reported above, and to avoid masking effects of HFHS diet on iG6PC-KO genotype 19 , we evaluated the impact of IGN inactivation on both fecal and urine metabolome in NC-fed mice only. The overall fecal metabolome profile of iG6PC-KO mice was statistically significantly different from that of WT mice ( Fig. 2A), due to a significant reduction in the levels of acetate (Fig. 2B), and trimethylamine (TMA) and its precursor choline (Fig. 2C). Other changes were observed in the amount of fecal sugars, amino acids and of esters and other metabolites, without a significant impact on the overall profile ( Supplementary Fig. 1). The overall urine metabolome profile was not significantly affected by IGN inactivation though the amounf of SCFAs, esters and other metabolites showed statistically significant different profiles from those of WT mice (Fig. 3). The amount of TMA and its precursors together with urine volume out of 48 h did not significantly vary (Supplementary Fig. 2A,B). These data were associated with an increased food intake over 48 h ( Supplementary Fig. 2C) with no significant effect on body weight ( Supplementary Fig. 2D). Altogether, these data indicate that IGN inactivation affects both fecal and urine metabolome. www.nature.com/scientificreports/ Effects of gastric bypass on gut microbiota, microbiome, fecal and urine metabolome in normal chow fed mice. GBP is one of the most promising and effective treatments for severe obesity, also inducing a huge amelioration of T2DM. To understand the impact of GBP on gut microbiota, we analysed both gut microbiota and microbiome in NC-and HFHS-fed WT mice after GBP surgery. GBP was previously reported not to change food intake on NC 20,21 , thus, there was no rationale to study a pair-fed (PF) group. On NC, GBP induced a statistically significant shift of gut microbiota towards Firmicutes and Deltaproteobacteria (Fig. 4A) compared to control mice that underwent laparotomy. This shift was associated with a microbiome significantly different between groups (Fig. 4D). The overall diversity (Fig. 4B) and gut microbiota profile (Fig. 4C) were not statistically different. GBP induced a significant change in overall fecal metabolome (Fig. 5A), due to increased fecal glucose (Fig. 5B), reduced fecal acetate content (Fig. 5C), a general significant reduction in fecal amino acid content (Fig. 5D) and a significant change in the amount of several esters (  for D, **P < 0.01, for H with alpha value for the factorial Kruskal-Wallis test among classes and alpha value for the pairwise Wilcoxon test between subclasses set both at 0.01 and threshold on the logarithmic LDA score for discriminative features set at 3. ("Lap" stands for laporotomized). Please note that in Fig. 1H  www.nature.com/scientificreports/ Gastric bypass modifies gut microbiota and microbiome in high-fat high-sucrose fed mice. GBP was reported to decrease food intake during HFHS feeding 22,23 . Thus, to avoid the impact of decreased food intake on gut microbiota, we set up a group of pair-fed (PF) WT mice (HFHS_WT_Lap_PF mice) that underwent laparotomy and received the same amount of food as GBP mice. During HFHS feeding, we observed that GBP in WT mice induced a statistically significant shift towards Proteobacteria, compared to both WT-Lap and WT-Lap-PF mice (Fig. 6A). The overall diversity was significantly highly different, based on Chao-1 and Fisher-alpha indices (Fig. 6B) as well as the general gut microbial profile (Fig. 6C). GBP also induced a statistically significant increase in many inferred microbial functions related to chemicals metabolism ( Fig. 6D) with an overall microbiome profile significantly different from that of control mice not PF (Fig. 6E).
Combined effect of GBP and IGN inactivation on gut microbiota, microbiome, fecal and urine metabolome in normal chow fed mice. Next, we evaluated the combined effect of GBP and IGN inactivation on gut microbiota, microbiome, fecal and urine metabolome during NC. iG6PC-KO mice that underwent GBP showed a statistically significant higher abundance of Proteobacteria and Actinobacteria, compared to Lap iG6PC-KO mice (Fig. 7A). None of the diversity indices was significantly changed (Fig. 7B), though the overall gut microbiota profiles were significantly dissimilar (Fig. 7C) and the microbiome of iG6PC-KO mice that underwent GBP showed biggest significant changes in pathways related to amino acids and cyano amino acids metabolism and biosynthesis of secondary metabolites (Fig. 7D). Note that iG6PC-KO mice did affect neither food intake nor body weight ( Supplementary Fig. 4). These data indicate that GBP may counteract IGN inactivation-induced change in gut microbiota. It is noteworthy that in iG6PC-KO mice GBP did not significantly change fecal metabolome ( Supplementary  Fig. 5) and slightly significantly changed urine metabolome (Fig. 8A,B and Supplementary Fig. 6).
We did not identify any cluster of parameters from gut microbiota, microbiome and fecal metabolites (Supplementary Fig. 7A). By contrast, the genus Lactobacillus was significantly and positively correlated with microbial pathway related to transcription proteins and urine succinate ( Supplementary Fig. 7B-D).

Combined effect of GBP and IGN inactivation on gut microbiota and microbiome in HFHS-fed mice.
On HFHS, iG6PC-KO mice that underwent GBP showed a statistically significant higher relative abundance of bacteria from family Enterococcaceae and from order Enterobacteriales, compared to both laparotomized iG6PC-KO mice and PF laparotomized iG6PC-KO mice (Fig. 9A). The overall diversity, mostly based on Chao-1 and Fisher-alpha indices, as well as the general gut microbiota profile of iG6PC-KO mice that underwent GBP were significantly different from the other groups of mice ( Fig. 9B-C). Microbial inferred functions related to amino acid metabolism were also significantly enriched (Fig. 9D), though the general microbiome profile was not significantly divergent ( Fig. 9E), compared to the other groups of mice. All these data showed that GBP in iG6PC-KO mice is able to modulate gut microbiota and microbiome during both NC and HFHS diets.

GBP induces specific changes in gut microbiota and microbiome of NC-and HFHS-fed WT mice.
To specifically determine the role of diet in relation to WT genotype, we compared gut microbiota and microbiome of all of the WT mice fed either a NC or a HFHS diet in this work. All groups of mice displayed at least a specific microbial taxon ( Supplementary Fig. 8A). In detail, on NC, the phyla Bacteroidetes and Actinobacteria had a significantly higher abundance in laparotomized mice, while bacteria from order Lactobacilalles showed a significantly higher abundance in GBP mice. On HFHS diet, the phylum Proteobacteria together with bacteria from genus Bacteroides had a significantly higher abundance in GBP mice, whereas laparotomized mice had a significantly higher abundance of Firmicutes and PF laparotomized HFHS-fed mice had a significantly higher abundance of family Rikenellaceae and genus Bilophila. The overall diversity of gut microbiota of all groups of mice was highly and significantly divergent, mostly based on Chao-1 and, to a lesser extent, to Fisher-alpha indices ( Supplementary Fig. 8B). The overall profile of gut microbiota of all groups of mice was also significantly diverse, with NC-fed laparotomized mice being the most divergent group compared to HFHSfed mice (Supplementary Fig. 8C). Consistently, all groups of WT mice displayed microbial inferred functions   Fig. 8D). The profile of gut microbiome of all groups of mice was also significantly diverse, with HFHS-fed GBP mice being the most diverse group when compared to HFHS-fed mice ( Supplementary Fig. 8E).

GBP induces specific changes in gut microbiota and microbiome of NC-and HFHS-fed iG6PC-KO mice.
To specifically determine the role of diet in relation to iG6PC-KO genotype, we compared the gut microbiota and microbiome of all of the iG6PC-KO mice fed either a NC or a HFHS diet in this work. All groups of mice displayed specific taxa of gut microbiota ( Supplementary Fig. 9A). In detail, on a NC, the class of Bacilli had a significantly higher abundance in laparotomized iG6PC-KO mice whereas the gut microbiota of GBP iG6PC-KO mice showed significantly higher abundance of phyla Proteobacteria, Bacteroidetes and Actinobacteria. On a HFHS feeding, bacteria from order Enterobacteriales and genus Roseburia showed a significantly higher abundance in GBP iG6PC-KO mice; genus Anaerovorax had a significantly higher abundance in laparotomized iG6PC-KO mice whereas the gut microbiota of PF iG6PC-KO mice was characterized by a significantly higher abundance of bacteria from Christensenellaceae family and genus Alistipes. The overall diversity of gut microbiota of all of the groups of mice was highly and significantly different, based on both Chao-1 and Fisher-alpha indices ( Supplementary Fig. 9B). The overall profile of gut microbiota of all groups of mice was significantly diverse, with a more pronounced diet effect ( Supplementary Fig. 9C), compared to WT mice ( Supplementary Fig. 9C). These data suggest a synergic effect by GBP and diet on gut microbiota in iG6PC-KO mice. Regarding gut microbiome, all groups of iG6PC-KO mice displayed specific microbial inferred functions enriched by GBP and diet ( Supplementary Fig. 9D) and a significantly diverse overall profile ( Supplementary  Fig. 9E). Finally, it is noteworthy that the comparative analysis of both gut microbiota ( Supplementary Fig. 10) and microbiome ( Supplementary Fig. 11) could identify at least a specific microbial taxon and microbial inferred functions enriched in each of the ten groups of mice in this study. Only the gut microbiota of HFHS-fed iG6PC-KO GBP mice did not significantly vary compared to all the other groups of mice (since it does not appear in Supplementary Fig. 10) whereas its microbiome showed specific enriched inferred functions Supplementary  Fig. 11. Overall, these data show that GBP induces highly specific changes in both gut microbiota and microbiome in both WT and iG6PC-KO mice, regardless of diet.
The major changes induced by both gastric bypass and intestinal glucose inactivation on gut microbiota, microbiome as well as fecal and urine metabolome are summarized in Fig. 10.

Discussion
In this study, we evaluated separately and combinedly the effects of both IGN inactivation and GBP on gut microbiota and microbiome in NC-and HFHS-fed mice. Our data highlight that both IGN and GBP have the   Fig. 9D to generate LDA score. ("Lap" stands for laporotomized; "PF" stands for pair-feeding). "n" for: HFHS_iG6PC_KO_GBP = 5, HFHS_iG6PC_KO_Lap = 6, HFHS_iG6PC_KO_Lap_PF = 6. www.nature.com/scientificreports/ capacity to modulate gut microbiota and microbiome regardless of the two diets. We hypothesized that IGN inactivation might change the content of SCFAs in the lumen, leading to a modification in bacterial cross-feeding and a final shaping of gut microbiota. Consistently with our hypothesis, we found a reduction in the amount of fecal acetate, which is the major SCFA produced. In accordance with our data, IGN inactivation in HFHS-fed mice supplemented with fructooligosaccharides also decreased cecal acetate content without measurable changes in propionate and butyrate 12 . We also showed that IGN inactivation on NC induced a dysbiosis with significantly higher relative abundance of bacteria belonging to major phyla such as Firmicutes (significantly higher relative abundance of genus CandidatusArthomitus and family LachnospiraceaeNK4A136), Proteobacteria (significantly higher relative abundance of family Desulfovibrionaceae and genus Desulfovibrio) and Bacteroidetes (significantly higher relative abundance of genera Odoribacter, Alistipes and Rikenella).
In line with our previous report, during HFHS-feeding, mice with IGN inactivation displayed higher abundance of Firmicutes 11 . HFHS-fed mice with IGN inactivation also had a significant higher abundance of Proteobacteria such as Helicobacter, which was not observed during HFHS feeding or IGN inactivation alone. This evidence may suggest a detrimental effect of the suppression of IGN on gut microbial community. In fact, during NC feeding, WT mice compared to iG6PC-KO mice had a significant higher abundance of genus Parabacteroides, shown to alleviate obesity and metabolic alterations in both ob/ob and HFD-fed mice, by activating IGN via the production of succinate 24 . Additionally, once combined with a nutritional challenge such as a HFHS diet, IGN inactivation may be deleterious for the host via the subsequent increase of pathogens such as Helicobacter 25,26 . This interpretation is corroborated by the significant higher bacterial motility-related inferred functions observed in the gut microbiome of HFHS-fed iG6PC-KO mice, the efficacy of bacterial motility being a distinguishing trait of pathogens 27 . Further analyses including multiple time-points of fecal sampling during HFHS diet should improve our knowledge on the causality relationship of the diet for the changes in gut microbiota composition.
GBP induced significantly higher abundance of Firmicutes and Proteobacteria on NC (limited sample size due to complexity of GBP surgery may have underpowered the experimental comparison between NC_WT_Lap and NC_WT_GBP groups) and significantly higher abundance of Firmicutes, Bacteroidetes and Proteobacteria on HFHS diet. Firmicutes are major producers of SCFAs such as butyrate. Therefore, a shift in gut microbiota towards SCFAs-producing bacteria could account, at least in part, for the metabolic benefits of GBP, since SCFAs induce multiple metabolic benefits for the host 28 including the activation of IGN [11][12][13] . We did not observe changes in fecal butyrate content regardless of genotype, GBP or diet in our study. Indeed, changes in the amount of propionate and butyrate are more difficult to highlight in feces since they are mainly absorbed by colonocytes 29,30 . A similarity between the shift in gut microbiota induced by IGN inactivation and GBP was observed at the level of phyla (Firmicutes, Bacteroidetes and Proteobacteria). However, at sublevels such as family, genus and species the two shifts showed some differences, underlining the impact of both IGN inactivation and GBP on NC and HFHS.
More importantly, IGN but not GBP induced changes in urine acetate levels. Butyrate is the preferred energy substrate for colonocytes, while acetate crosses gut barrier to reach the portal vein. Propionate is utilised by the intestine and the liver. This is not the case of acetate. As a result, acetate is the most abundant SCFA in the systemic circulation whereas only small amounts of butyrate and propionate could be found in periphery 31 . Microbiota shaping by IGN inactivation might thus control systemic acetate plasma levels and consequently negatively impact on host metabolism. Overall, the detrimental effects deriving from IGN inactivation agree with the previous observation that SCFAs activate IGN 28 thus leading to huge metabolic benefits.
It is noteworthy that GBP counteracted changes induced by IGN inactivation on gut microbiota and microbiome during both NC and HFHS diet. On one hand, we observed that NC-fed iG6PC-KO mice that underwent GBP had higher bifidobacteria at the taxonomic levels of order, family and genus. Bifidobacteria are known probiotics 32 , therefore the combination of GBP and IGN inactivation during NC feeding appears to benefit the host by shifting gut microbiota towards an enrichment in beneficial microbes. By contrast, on a HFHS feeding, www.nature.com/scientificreports/ the combination of GBP and IGN inactivation appears to disadvantage the host, by shifting gut microbiota towards an enrichment in detrimental microbes, such as those from genus Escherichia_Shigella, which are known enterobacteria pathogens 33 . During HFHS feeding, GBP induced a significant decrease in Fisher and Chao-1 diversity indices in both WT and iG6PC-KO mice. These indices are related to rare microbial taxa, suggesting that this reduction is not affecting the overall gut microbiota diversity. The family Desulfovibrionaceae was significantly and positively correlated with the indole alkaloid biosynthesis microbial pathway, when combining microbiota and microbiome in both all WT and iG6PC-KO mice in this study. A vast body of publications points out the role of indole as a microbial metabolite responsible for many regulatory effects on host functions, such as inflammation and gut barrier 28 . The Desulfovibrionaceae family showed a significant higher relative abundance in the gut microbiota of both NC-and HFHS-fed WT mice that underwent GBP. This evidence appears in keeping with the benefits of GBP on the detrimental changes induced on gut microbiota either by IGN inactivation on NC or by a HFHS diet. However, the relative abundance of Desulfovibrionaceae family was also significantly higher in gut microbiota of NC-fed iG6PC-KO mice. This evidence suggests that the presence of Desulfovibrionaceae family may be necessary but not sufficient to explain alone the beneficial effects of GBP to the host. Rather, the overall GBP-modified gut microbiota would account for these positive effects.
When considering the impact on gut microbiota of both IGN inactivation and GBP during NC and HFHS feeding, the most striking result is the fact that every group of mice, regardless of genotype and diet, had at least a specific taxon and/or a microbial function enriched. This evidence underlines deep modifications induced by IGN inactivation and GBP on gut microbiota. It must be mentioned that the changes in microbiota induced by the diet were studied through the comparison of NC to HFHS fed mice. However, the specific effects of the HFHS diet on microbiota and microbiome might be improved by further analyses including multiple time-points of fecal sampling to decipher to which extent any diet account for the observed changes.
By showing that GBP may also counteract, at least in part, IGN inactivation-induced gut microbiota dysbiosis, our study also provides a new rationale susceptible to account for the metabolic benefits of GBP. Finally, the inactivation of IGN reduces the intestinal capacity to metabolize specific bacterial metabolites such as SCFAs, changing both ecological structure and functions of the gut microbiota.

Materials and methods
Animal models. All experiments were performed according to the principles and guidelines established by the European Convention for the Protection of Laboratory Animals. Protocols were approved by our regional animal care committee (C2EA-55, Université Lyon 1, Lyon) and by the Ministry of Higher Education and Research (Agreement project number: Apafis#11929-2017102421331413 v1). Male C57Bl/6 J wild-type mice (WT) were purchased from Charles River Laboratories at 4 weeks of age. Male iG6PC-KO mice, with an intestine-specific disruption of the catalytic subunit (G6pc1) of glucose-6 phosphatase, the key enzyme in endogenous glucose production, were generated as described previously 34 . Briefly, 7-8-wk-old iG6PC-KO mice received a daily 100 µL injection of Tamoxifen (10 mg/mL, Sigma) on 5 consecutive days. WT mice did not receive tamoxifen as they will undergo major surgery and a HFHS diet for 20 weeks (see below). In fact, the putative metabolic effect of tamoxifen in adult mice decreases sharply after 5 weeks, thus its effect will likely be dominated by nutritional and/or surgical cues 35,36 .
All mice were housed in the animal facility of Lyon 1 University under controlled temperature (22 ± 2 °C) and lighting (12 h light/dark cycle with light at 7 a.m.) with free access to food and water.
To induce obesity, 4-weeks old WT and iG6PC-KO mice were placed on high-fat/high-sucrose (HFHS) diet for 20 weeks prior to surgery. HFHS diet, consisting of 36.1% fat, 35% carbohydrates (50% maltodextrine + 50% sucrose) and 19.8% proteins, was produced by the Unité de Préparation des Aliments Expérimentaux (UE SAAJ INRAE, Jouy-en-Josas, France). For experiments in lean animals, WT and iG6PC-KO mice were maintained on standard diet (SAFE A04, Augis, France) and surgery was performed at 24-weeks old. All animals were maintained on their respective diet after surgery.
The duodenal-jejunal bypass surgery (GBP) was performed as previously described in 24-weeks old mice 37 . For each genotype (WT and iG6PC-KO), GBP (duodenal-jejunal)-operated mice and sham-operated mice groups were constituted at the day of surgery; sham operation consisting in a laparotomy 37 . Since GBP mice fed a HFHS diet ate less during the first ten days after the operation, experiments on HFHS fed mice involved a third group of sham-operated PF mice (sham-PF) mice 37 . Twenty-two days after surgery, mice were placed fro 48 h in a metabolic cage for collecting urine and feces.
Twenty-five days after surgery, mice were fasted for 6 h and next killed by cervical dislocation. The caecum content was sampled and immediately frozen at − 80 °C for further analyses. All groups co-housed almost 6 months (20 weeks plus 25 days) since the beginning of the diet, which equilibrates gut microbiota of mice before any intervention 38 . Mice were randomly associated to control or any other treatment (HF/HS and/or surgery) group. Given the murine models used in this work, researchers were aware of the group allocation during all the study. ARRIVE guidelines were followed. "n" = 35 for WT mice (NC_WT_Lap = 4; NC_WT_GBP = 3; HFHS_ WT_Lap = 9; HFHS_WT_Lap_PF = 10; HFHS_WT_GBP = 9); "n" = 26 for KO mice (NC_iG6PC_KO_Lap = 5; NC_iG6PC_KO_GBP = 4; HFHS_iG6PC_KO_Lap = 6; HFHS_iG6PC_KO_Lap_PF = 6; HFHS_iG6PC_KO_ GBP = 5). Exclusion criteria of mice were only based on the success of surgery and on the quality and quantity of the collected urine and feces samples.
Taxonomic and predicted functional analysis of gut microbiota. Total  Metabolomic analysis. Sample preparation. Fecal samples (100 mg) were homogenized using the Fast-Prep-24 (MP Biomedicals, Irvine, CA, USA) homogenizer in 0.5 ml of phosphate buffer (0.2 M, pH = 7) prepared in deuterium oxide (D 2 O) and containing 1 mM trimethylsilylpropionic acid (TSP). Samples were left on ice for 1 min and homogenized again. Samples were then centrifuged (10000g, 10 min, 4 °C) and supernatants were collected. The remaining pellet was further extracted as described above. Supernatants obtained from two runs of extraction were combined and centrifuged at 10,000g for 10 min at 4 °C. A total of 600 µL of supernatant was transferred into 5 mm NMR tubes. As for urines, they were collected over 48 h on metabolic cages. Prior to analysis, urine samples were thawed at room temperature. Then, 200 µL of phosphate buffer (0.2 M, pH = 7) prepared in D 2 O and containing 1 mM TSP was added to 500 µL of urine sample. The mixture was vortexed, centrifuged (5500g, 15 min, 4 °C), and 600 µL of supernatant were transferred to a 5 mm NMR tube.
1 H NMR acquisition. 1 H NMR spectra were obtained at 300 K, on a Bruker Avance III HD 600 MHz NMR spectrometer (Bruker Biospin, Rheinstetten, Germany) operating at 600.13 MHz for proton frequency using an inverse detection 5 mm 1 H-13 C-15 N-31 P cryoprobe. 1 H NMR spectra of urine samples were acquired using the « noesypr1d » (Bruker Library) pulse sequence with water suppression during the relaxation delay (5 s) and mixing time (100 ms). A total of 256 transients were collected into 65,536 data points using a spectral width of 20 ppm and an acquisition time of 2.7 s. 1 H NMR spectra of fecal samples were acquired using the Carr-Purcell-Meiboom-Gill (CPMG) spin echo pulse sequence with presaturation, with a total spin-echo delay of 320 ms to attenuate broad signals from macromolecules. A total of 256 transients were collected into 65,536 data points using a spectral width of 20 ppm, a relaxation delay of 5 s and an acquisition time of 2.7 s. Prior to Fourier transformation, an exponential line broadening function of 0.3 Hz was applied to the Free Induction Decays (FID). All NMR spectra were phase-and baseline-corrected and referenced to the chemical shift of TSP (0 ppm) using Topspin (V3.2, Bruker Biospin, Germany). Original sources of method description as already published 41-43 . Statistical analysis. Data for diversity indices and metabolome analyses are presented as mean ± SD. Statistical analyses were performed by 2-way ANOVA followed by a 2-stage linear step-up procedure of Benjamini, Krieger and Yekutieli to correct for multiple comparisons by controlling the False Discovery Rate (< 0.05), by using GraphPad Prism version 9.3.0 for Windows Vista (www. graph pad. com, GraphPad Software, San Diego, CA). Values were considered significant starting at P < 0.05 or as reported. For the taxonomical and predictive functional analysis of gut microbiota significant values were considered starting at P < 0.05 or as reported. Principal Component Analysis graphs were drawn and related statistical analyses were performed using PAST 4 0.08 (https:// www. nhm. uio. no/ engli sh/ resea rch/ infra struc ture/ past/) and calculating Bray-Curtis distance with 1-way PERMANOVA analysis with Bonferroni's correction. Heat-maps based on Pearson distance and a complete linkage were drawn by using PermutMatrix 1.9.4 (http:// www. atgc-montp ellier. fr/ permu tmatr ix/) 44 . Original sources of method description as already published [41][42][43] .