A gut microbiota-bile acid axis promotes intestinal homeostasis upon aspirin-mediated damage

SUMMARY

Aspirin-related gastrointestinal damage is of growing concern. Aspirin use modulates the gut microbiota and associated metabolites such as bile acids (BAs) but how this impacts intestinal homeostasis remains unclear. Herein, using clinical cohorts and aspirin-treated mice, we identified an intestinal microbe, Parabacteroides goldsteinii, whose growth is suppressed by aspirin. Mice supplemented with P. goldsteinii or its BA metabolite, 7-keto-lithocholic acid (7-keto-LCA), showed reduced aspirin-mediated damage of the intestinal niche and gut barrier, effects that were lost with a P. goldsteinii hdhA mutant unable to generate 7-keto-LCA. Specifically, 7-keto-LCA promotes repair of the intestinal epithelium by suppressing signaling by the intestinal BA receptor, farnesoid X receptor (FXR). 7-Keto-LCA was confirmed to be a FXR antagonist that facilitates Wnt signaling and thus self-renewal of intestinal stem cells. These results reveal the impact of oral aspirin on the gut microbiota and intestinal BA metabolism that in turn modulates gastrointestinal homeostasis.

Graphical Abstract

In brief

Li and Ding et al. identify an intestinal microbe, Parabacteroides goldsteinii, whose growth is suppressed by aspirin. They show that P. goldsteinii and its metabolite 7-keto-lithocholic acid promoted repair of aspirin-associated intestinal damage by suppressing the bile acid receptor FXR and maintaining Wnt signaling in intestinal stem cells.

INTRODUCTION

Aspirin has become the most commonly used drug because of its analgesic, antiplatelet and potential antineoplastic effects, and it is recommended to lower the risk of adverse cardiovascular outcomes and prevent cancer.^1–3 For years, the adverse effects of aspirin, such as gastrointestinal damage, have continued to cause concern,^4–6 notably since recent reports suggest that over 90% of patients receiving aspirin develop intestinal toxicity.^7,8 The mechanism by which aspirin induces intestinal injury has received considerable attention, yet controversies still exist. Aspirin inhibits prostaglandin-endoperoxide synthase 1 (PTGS1 or cyclooxygenase [COX] 1) and COX2, which are believed to mediate gastrointestinal damage.^6,9,10 However, further studies found that small bowel mucosal prostaglandins could be dramatically decreased without mucosal damage, which was also confirmed in COX1-knockout mice.^6,11,12 Thus, the inhibition of COX does not seem to be the only mechanism involved in aspirin-associated enteropathy. In addition, aspirin is an orally administered drug that is absorbed in the small intestine. There, it reaches its highest concentrations with more than 50% excreted in the feces.¹³ Therefore, the possibility cannot be excluded that the adverse effects of aspirin might be due in part to modifications of the gut microbiota.

Aspirin treatment was recently reported to be associated with altered composition of the gut microbiota.^14,15 However, which species of the gut microbiota are affected by aspirin and the underlying mechanism mediating the impact of aspirin on host intestinal homeostasis are still unclear. Recently, accumulating evidence have shown that the gut microbiota influences intestinal homeostasis and intestinal stem cell function via the modulation of metabolites, including bile acids.^16–22 Intestinal stem cells are essential in maintaining the gut barrier,^21,23 which is impaired during the development of inflammatory bowel disease^20,24 as well as the aspirin-related enteropathy.²⁵ Several studies have suggested pivotal roles for bile acids produced by commensal bacteria in regulating the expansion, differentiation and self-renewal of intestinal stem cells.^26,27 For instance, bile acid receptors s such as the farnesoid X receptor (FXR), are important players in coordinating the renewal and regeneration of intestinal stem cells in response to injury.²⁸ Several studies also indicated that the administration of aspirin analogs was associated with an altered gut microbiota.²⁹ However, it is still unclear whether and how aspirin-related gut dysbiosis and alteration of bile acids regulate homeostasis of the intestinal stem cell niche.

The current study reports that oral aspirin decreased levels of the bile acid 7-keto-lithocholic acid (7-keto-LCA) by depleting the Parabacteroides goldsteinii (P. goldsteinii) in the intestine, as revealed by integrative metagenomics and metabolomics analysis of humans and mice. Furthermore, we identified 7-keto-LCA as an endogenous antagonist of intestinal FXR, which is important in sustaining intestinal stem cell niche homeostasis and preserving the gut barrier. These results suggest that aspirin-induced intestinal damage is perturbed by a P. goldsteinii-bile acids-FXR axis. Supplementation with P. goldsteinii and 7-keto-LCA may be of potential translational value in the prevention of gastrointestinal damage.

RESULTS

Aspirin induces gut dysbiosis in humans and mice

To investigate how aspirin affects the gut microbiota and metabolites in humans, we collected stool samples from 23 healthy volunteers before (BA) and after (AA) being naively treated with 100 mg aspirin per day for 30 days (Figure 1A and S1A). Whole-genome shotgun sequencing of stool samples showed that although the Shannon indexes were not significantly changed, the Simpson indexes were decreased (Figure S1B). The gut microbiota composition was substantially reshaped after aspirin treatment (Figure 1B), which was supported by partial least squares discriminant analysis (PLS-DA). The variable importance in projection (VIP) score and linear discriminant analysis (LDA) score of the gut microbiota also indicated that Parabacteroides were the top gut microbiota that resulted in the group separation (Figure 1C and S1C–E). Subsequent analysis by co-occurrence network and comparison also confirmed the depletion of gut bacteria in the genus Parabacteroides (Figure 1D–F). We further determined the abundance of the main species and found that P. goldsteinii, Parabacteroides merdae (P. merdae) and Parabacteroides distasonis (P. distasonis) were all decreased (Figure 1G). Further canonical correspondence analysis and redundancy analysis both indicated that use of aspirin resulted in a shift of microbial composition (Figure S1F–G). We also re-analyzed the metagenomic data of gut microbiota in one previously published database.³⁰ Similarly, the abundance of P. goldsteinii instead of other species were found to be significantly decreased in the aspirin-user groups (Figure S1H).

Figure 1. Aspirin induces gut dysbiosis in humans and mice.

(A) Cohort of healthy volunteers treated with aspirin (BA-AA cohort). Stool samples were collected before the administration of aspirin (BA group) and 30 days after perioral aspirin administration (100 mg/d) (AA group), followed by metagenomics next-generation sequencing.

(B) Partial least squares discriminant analysis (PLS-DA) based on the relative genus abundances. N=23 individuals/group.

(C) Variable Importance in Projection (VIP) scores of PLS-DA showed the ability of different taxa to discriminate between groups. A taxon with VIP score > 1.5 was considered important in the discrimination.

(D) Gut microbial co-occurrence network analysis based on core genus (top 50 abundant genus) in the BA and AA groups. Connecting lines indicate the absolute values of Spearman’s rank correlation coefficient >0.30.

(E) Relative abundance of gut microbiota of the BA and AA groups. Red: Parabacteroides.

(F) Comparison of the relative abundances of the top 12 genera with the most significant differences between the BA and AA groups.

(G) Relative abundance of different species in genus of Parabacteroides based on metagenomics results. n=23 individuals/group.

(H) Top: C57BL/6J mice were treated with aspirin [2 mg/mL (11.1 mM) in drinking water] for 2 weeks and feces collected for 16S sequencing (n=6/group). Bottom: PLS-DA of 16S sequencing (n=6/group) (Asp 0d vs Asp 14d).

(I) The three indexes of α-diversity of mice fecal microbiota (Asp 0d vs Asp 14d).

(J) Abundance of the top 10 abundant species of gut microbiota from mice at different time points after aspirin treatment.

(H–J) Two times each experiment was repeated independently with similar results. (B and H) PERMNOVA Test. (F) False positive rate of two-tailed Wilcoxon rank-sum paired-test. (G and I) Two-tailed Wilcoxon rank-sum paired-test.

We further verified the effects of aspirin on gut microbiota in an animal model and the PLS-DA data also showed that the gut microbiota composition was altered by oral aspirin (Figure 1H). Similarly, the α-diversity (Figure 1I) and the abundance of P. goldsteinii was decreased in a time-dependent manner (Figure 1J, Figure S1I). Together, these data suggest that aspirin induced gut dysbiosis in both humans and mice, characterized by decreased abundance of P. goldsteinii.

Aspirin impairs gut barrier function and causes intestinal damage

To further determine the importance of P. goldsteinii in the disturbance of intestinal homeostasis by aspirin, the intestinal injury scores of mice treated with aspirin were evaluated (Figure 2A, Figure S2A).^31,32 A negative correlation between the abundance of P. goldsteinii and the severity of intestinal injury was determined (Figure S2B–D). Notably, the heatmap and principal component analysis (PCA) of the microbiota composition in intestinal contents indicated that the gut microbiota was different in mice treated with aspirin and neomycin, compared to the saline group (Figure 2B–C). Both aspirin and neomycin decreased α-diversity of the gut microbiota (Figure S2E), suppressed P. goldsteinii in SPF mice (Figure 2D). Furthermore, Periodic Acid-Schiff staining (PAS) showed that the intestinal mucosa became thinner after aspirin treatment, which was similar in mice upon treatment with neomycin (Figure 2F). This may be due to the suppressed differentiation of goblet cells (PAS+ cells, Figure S2F) and decreased proliferation of crypt cells (Ki67+ cells, Figure 2E). Moreover, intestinal permeability was increased by aspirin or neomycin (Figure 2G). Fecal occult blood test (FOBT) results suggested that intestinal bleeding rates were increased after aspirin and neomycin treatment (Figure 2H). TdT-mediated dUTP nick-end labeling (TUNEL) assays also revealed that apoptosis was driven by aspirin and antibiotic treatment (Figure 2I). We then measured plasma and intestinal levels of aspirin by ultra-performance liquid chromatography–coupled time-of-flight mass spectrometry (UPLC-ESI-QTOFMS), revealing no difference of bioavailability of aspirin (Figure S2G). Transmission electron microscopy (TEM) further showed an impaired gut barrier after aspirin treatment (Figure 2J), which was also supported by decreased levels of tight junction proteins ZO-1 and claudin-5 (Figure 2K). More interestingly, fecal microbiota transplantation (FMT) of commensal microbiota ameliorated aberrant phenotypes observed in aspirin-treated mice, while FMT from mice already being administered aspirin had no effect (Figure S2H–I).

Figure 2. Aspirin impairs gut barrier function and causes intestinal damage.

(A) C57BL/6J mice were subjected to oral aspirin treatment for 2 weeks by adding aspirin into the drinking water at a concentration of 2 mg/mL (11.1mM), with or without antibiotics (neomycin, Neo). Representative images of H&E staining of intestine and injury scores for each group are shown. Random 10 positions of distal jejunum were picked and the average scores calculated for all mice (n=10 replicates/group). Scale bar: 200 μm.

(B) Heatmap of different species in mice intestinal contents as treated in (A).

(C) PCA analysis of 16S sequencing results derived from intestinal contents of the indicated four groups.

(D) Relative abundance of P. goldsteinii based on 16S sequencing results. n=5/group.

(E–F) Ki67 (E) and Periodic Acid-Schiff (PAS, F) staining of distal jejunum of mice. The quantitative comparison of Ki67 staining of intestinal crypts (red dotted lines, n = 5/group) and PAS staining of villus (n=10/group) were shown. Twenty random points of one sample were picked and the average value for each mouse was calculated as one single point.

(G) Intestinal permeability analysis using diamine oxidase (DAO) (left), FITC-labeled dextran (middle) and lipopolysaccharide (LPS) (Right). N=5 replicates/group.

(H) Fecal occult blood test (FOBT) using feces samples of mice from four groups. n=15 replicates/group.

(I) Representative images of TUNEL staining of intestine. Statistical analysis of average TUNEL-positive area is shown. N=5 replicates/group. Twenty random points of one sample were picked and the average value for each mouse was calculated as a single point.

(J) Transmission Electron Microscopy (TEM) of intestinal tissues from mice as treated in (A). Representative images from 1 of 5 mice were shown. Green arrows point to intact cell-cell junction and red arrows point to a disrupted cell junction. Scale bar: 5 μm.

(K) The mRNA levels of ZO-1 and claudin5 in intestinal tissues from mice as treated in (A). n=5 replicates/group. Expression of target gene mRNAs were normalized as to Actb.

(A–D) Two times each experiment was repeated independently with similar results. (E–J) Three times each experiment was repeated independently with similar results. (A, B, E, F, G, I and K) One-way ANOVA followed by Fisher’s LSD post hoc test. (C) PERMANOVA test. (H) Fisher’s precision probability test.

We also analyzed whether the colon epithelium was damaged by aspirin. However, no severe damage of colon epithelium was observed (Figure S3A). Nevertheless, the PCA analysis based on 16S sequencing of fecal samples also showed a distinct composition of the gut microbiota (Figure S3B). Similar to above results, the relative abundance of P. goldsteinii was also decreased in neomycin-treated and aspirin-treated mice (Figure S3C, D). Pearson’s r analysis further showed a positive correlation between fecal and intestinal composition of P. goldsteinii (Figure S3E). Together, these data suggest that aspirin impairs gut barrier function, potentially through inducing gut dysbiosis and suppressing P. goldsteinii.

Aspirin damage intestine through regulation of gut microbiota

Two other classic antibiotics, clindamycin and streptomycin, were also used to eradicate bacteria (CS group, Figure S3H). Interestingly, the relative abundance of P. goldsteinii were not changed in CS group (Figure S3F, G). Moreover, CS group with preserved P. goldsteinii seemed to be more resistant to aspirin’s effect on intestinal permeability (Figure S3H–I), mucin layer thickness (Figure S3J), cryptal cell proliferation and apoptosis (Figure S3K–L), and incidence of FOBT (Figure S3M). These results suggest that P. goldsteinii eradicated by neomycin, instead of other antibiotics, may play roles in intestinal injury.

We further tested these findings in germ-free (GF) mice. One group of GF mice were directly treated with aspirin for 2 weeks, while a cohort of GF mice (rGF group) were first naturally reconstituted with bacteria by co-housing with SPF mice, followed by aspirin treatment (Figure S4A). The 16S sequencing of the intestinal contents revealed successful colonization of P. goldsteinii in the aspirin-treated rGF mice (Figure S4B). We found that aspirin had no significant destructive effects in GF mice. However, rGF mice treated with aspirin had notable higher injury scores and deteriorated intestinal phenotype (Figure S4A, C–G). These data demonstrated that reconstitution of gut microbiota may mediate aspirin’s destructive effects on intestine.

P. goldsteinii is depleted by aspirin and restores bile acids spectrum.

We next investigated the mechanism underlying the suppressive effects of aspirin on P. goldsteinii. The intestinal concentration of aspirin in this model were quantified by LC/MS, ranging from 1 to 4 mM (Figure S5A). We showed a dose-dependent inhibition of growth in P. goldsteinii treated with aspirin (Figure 3A and S5B). Untargeted metabolomics analysis revealed that aspirin-related metabolites were enriched in compounds associated with amino acid metabolism (Figure S5C–D). Among them, L-glutamine, citruline and L-proline were found to be central components of interaction network (Figure 3B); these were verified in subsequent assays (Figure S5E). Supplementation with L-glutamine and citruline rescued the suppression of P. goldsteinii growth by aspirin (Figure 3C and S5F), indicating that aspirin may affect P. goldsteinii by interfering with the biosynthesis of both metabolites.

Figure 3. P. goldsteinii depleted by aspirin restores the gut microbiota and bile acid metabolite spectrum.

(A) P. goldsteinii growth curves with medium aspirin concentration ranging from 0 to 4 mmol/L (0.18 mg/mL-0.72 mg/mL). n=6 replicates/group. All statistical analysis were conducted between the respective groups and Asp set at 0 mM (Black).

(B) Metabolite enriched pathways and network based on metabolomics data derived from medium of P. goldsteinii treated with or without aspirin. Blue frame: different metabolites. Red line: positive-correlated in metabolism pathways; Blue line: negative-correlated in metabolism pathways.

(C) P. goldsteinii growth curves in response to aspirin and supplement of amino acids. Aspirin concentration: 4 mmol/L. n=6 replicates/group. All statistical analysis were conducted between respective groups and Asp group (Red).

(D) Left: Supplementation of P. goldsteinii (1×10⁸ CFU/0.2 mL, 3 times/week) to mice for 2 weeks, along with or without oral aspirin [(2 mg/mL (11.1mM)] for another 2 weeks. Intestinal contents were subjected to metabolomic analysis. n=8 replicates/group. Right: Bile acids concentrations.

(E) Mean percentages of different bile acids in the respective groups are shown. The area of the whole ring represents the concentration of the total bile acids.

(F) Concentrations of Tα/βMCA, 7-keto-LCA and UDCA in intestinal content of different groups of mice.

(G) Heatmap of the correlation is shown between the abundance of species of intestinal microbiota and the bile acid levels in the intestine of mice from Asp-treated mice and control mice.

(H) The concentrations of different bile acids in feces of BA-AA cohort in Figure 1A.

(I) Correlation between 7-keto-LCA and UDCA concentrations with P. goldsteinii abundance in stool samples of the BA-AA cohort. The samples with extremely low abundance (<10⁻⁴) were excluded.

(J) The abundance of the 7-α-hydroxysteroid dehydrogenase (hdhA) gene of the microbiota in individuals in the BA-AA cohort in Figure 1A. n=23 individuals/group.

(K) Synthetic pathways of BA in human and mice. In the gut of humans and mice, UDCA and 7-keto-LCA were generated by intestinal microbiota from CDCA, using hdhA. In mice, muricholic acid (MCA) was synthesized from CDCA by hepatic cells-generated 6β-hydroxylase and conjugated with taurine.

(L) The genome sequences of several species in Parabacteroides were determined using PacBio Sequel and Illumina MiSeq sequencers. Gene clusters containing hdhA identified in these species are shown.

(M) In vitro culture of P. goldsteinii (OD=0.1) were supplemented with CA, CDCA, LCA and UDCA. The medium was analyzed for bile acid concentrations. Concentrations of 7-keto-LCA and UDCA are shown. n=3 replicates/group.

(A–C) Three times each experiment was repeated independently with similar results. (D-G, M) Two times each experiment was repeated independently with similar results. (A, C, F and M) One-way ANOVA followed by Fisher’s LSD post hoc test. (G and I) Pearson correlation coefficient. (H and J) Two-tailed Wilcoxon matched-pairs signed rank test.

We then determined the regulatory effects of P. goldsteinii on the host metabolites. Untargeted metabolomic detection showed that intestinal bile acid pool was reduced after aspirin treatment (Figure 3D); this was restored by P. goldsteinii transplantation (Figure 3E). In particular, the intestinal concentrations of tauro-α/β-muricholic acid (Tα/βMCA), ursodeoxycholic acid (UDCA) and 7-keto-LCA were dramatically decreased by aspirin, while P. goldsteinii transplantation preserved them (Figure 3F). Moreover, a positive correlation was found between the abundance of P. goldsteinii and bile acids (Figure 3G). In addition, the metabolite profiling in the stool in the BA-AA cohort (see Figure 1A) showed similar changes and correlation (Figure 3H–I). These were also verified by the Mantel test (Figure S5G). The abundance of P. goldsteinii in stool samples of another cohort were also detected (asp-user-nonuser cohort), involving 70 aspirin nonusers and 105 aspirin users. Concordantly, the abundance of P. goldsteinii were decreased in the fecal samples of aspirin users (Figure S5H), and positively correlated with the fecal concentrations of UDCA and 7-keto-LCA (Figure S5I).

The above results suggest a bioactive role for P. goldsteinii in modulating bile acid patterns. Previously, 7-α/β-hydroxysteroid dehydrogenase (7-α/β-HSDH), that transforms CDCA into 7-keto-LCA and UDCA, was purified and characterized in Parabacteroides.³³ We performed further analysis of the metagenomic data of BA-AA cohort. The hdhA gene (encoding 7-α-HSDH) copy number was dramatically reduced after aspirin treatment (Figure 3J). Since previous studies suggest that hdhA encoded 7-α-HSDH are capable of transforming CDCA into 7-keto-LCA and UDCA in vivo and ex vivo^33,34 and β-MCA³⁵ in mice, we predicted the synthetic pathways of bile acid metabolism (Figure 3K). Further sequences orthologous to hdhA were identified in Parabacteroides, including P. goldsteinii, P. merdae and P. distasonis (Figure 3L). We also found that 7-keto-LCA but not UDCA was produced by P. goldsteinii in vitro (Figure 3M). A good correlation between bile acid concentrations in the intestinal contents and fecal samples was verified (Figure S5J). It was reported that Parabacteroides decreased taurine-conjugated bile acids by bile salt hydrolase (BSH),³⁶ our data indicated increasing concentrations of TβMCA by P. goldsteinii. This may be due to the expanding pool of total MCA in the bile acid pool (Figure S5K–M), while BSH activity did not change (Figure S5M). Together, these results suggest that the P. goldsteinii depleted by aspirin restores homeostasis of the intestinal bile acid spectrum.

P. goldsteinii ameliorates aspirin-induced intestinal damage through hdhA

We next investigated whether P. goldsteinii harboring hdhA is beneficial in preventing aspirin-induced intestinal damage. We constructed a hdhA-deficient P. goldsteinii mutant strain (PGΔhdhA) by knocking out the hdhA gene (Figure S6A). LC–MS analysis confirmed that the mutant strain was unable to convert CDCA into 7-keto-LCA (Figure S6B). In vitro culturing of both strains indicated no effects of hdhA on the growth of P. goldsteinii (Figure S6C). The qPCR results of fecal samples revealed that knocking out of hdhA in P. goldsteinii did not affect its colonization to the host (Figure S6D). Then SPF mice were treated with aspirin, together with transplantation of P. goldsteinii. Live P. goldsteinii (LPG) and wild-type P. goldsteinii (PGwt) but not heat-killed P. goldsteinii (HPG) or PGΔhdhA, ameliorated the injury scores (Figure 4A). Moreover, LPG and PGwt instead of HPG or PGΔhdhA restored the intestinal permeability (Figure 4B). PAS staining further showed that LPG and PGwt but not HPG or PGΔhdhA reversed the damage to the intestinal mucus barrier induced by aspirin (Figure 4C), and had positive effects in preserving goblet cells (Figure 4D). In addition, other aberrant phenotypes revealed by the TUNEL and FOBT assays demonstrated that LPG and PGwt protected intestinal epithelial cells (Figure 4E–F). Furthermore, TEM of the gut barrier showed robust tight junctions in the LPG group. These tight junctions were found to be damaged in the HPG group, similar to those of the aspirin group (Figure 4G). The PCoA analysis suggest that LPG transplantation may help recover the aspirin-induced gut dysbiosis (Figure S7A). Colonization of P. goldsteinii was verified by 16S sequencing (Figure S7B–C) and qRT-PCR (Figure S7D) after LPG transplantation in the aspirin-treated mice.

Figure 4. P. goldsteinii ameliorates aspirin-induced intestinal damage through hdhA.

(A) Supplementation of live P. goldsteinii (LPG), heat-killed P. goldsteinii (HPG), wild-type P. goldsteinii (PG-WT) or hdhA-deficient P. goldsteinii (PGΔhdhA) (1×10⁸ CFU/0.2 mL, 3 times/week) by gavaging mice for 2 weeks, with or without oral aspirin treatment (2 mg/mL [11.1 mM]) for another 2 weeks. Representative images of H&E-stained intestine and injury scores for each group are shown. Random 10 positions of distal jejunum were picked and average scores calculated for all the mice (n=10 replicates/group). Scale bar: 200 μm.

(B) Intestinal permeability analysis using DAO (left), FITC-labeled dextran (middle) and LPS (Right). n=5 replicates /group.

(C–D) PAS staining showing representative images of each group (left) and statistical analysis (right) of mucosal thickness (D) and PAS-positive Goblet cells (C). n=10 individuals /group.

(E) TUNEL staining of intestine were shown and random 10 position of distal jejunum were picked from all the mice. n=5 individuals/group.

(F) FOBT using feces samples of mice as treated in (a). n=15 replicates/group.

(G) The representative images of the structure of the intestinal gut barrier of aspirin-treated mice with LPG or HPG. Green arrows point to intact cell-cell junction under electron microscope and red arrows point to a disrupted cell junction. Yellow dotted line marked an apoptotic cell.

(H) Supplementation of UDCA and 7-keto-LCA to mice for 2 weeks, with or without oral aspirin (2 mg/mL [11.1 mM]) for another 2 weeks. Representative images of H&E-stained intestine and injury scores for each group are shown. Random 10 positions of distal jejunum were picked and the average scores calculated for all mice (n=10 replicates/group). Scale bar: 200 μm.

(I) Intestinal permeability analysis using DAO (left), FITC-labeled dextran (middle) and LPS (Right). n=5 replicates /group.

(J–K) PAS staining (n=10 replicates/group) and TUNEL+ area (n=5 replicates/group) of intestinal tissues from mice treated in (h).

(L) FOBT using feces samples of mice as treated in (h). n=15 replicates/group.

(M) Relative mRNA levels of markers of intestinal tight junction in mice intestinal tissues, including ZO-1 (left) and claudin-5 (right). n=5 replicates /group.

(A–G) Two times each experiment was repeated independently with similar results. (H–M) Three times experiment was repeated independently with similar results. (A, B, C, D, E, H, I, J, K and M) One-way ANOVA followed by Fisher’s LSD post hoc test. (F and L) Fisher’s precision probability test.

We further established a gain-of-function model of hdhA by E. coli (Figure S6E). Restoration of hdhA in E.coli DH5α (EcD:hdhA) was able to colonize the mouse gut as well as the wild-type strain (EcD:wt) (Figure S6F). EcD:hdhA instead of EcD:wt decreased the intestinal injury scores, preserved the mucinous thickness and gut barrier, increased the numbers of proliferating cells in crypts, and decreased the apoptotic area and FOBT incidence (Figure S6G–L). In addition, the culture medium was analyzed for bile acids indicating that hdhA in E. coli functioned in the production of 7-keto-LCA (Figure S6M).

The role of UDCA and 7-keto-LCA in the phenotype was next determined in vivo. Both UDCA and 7-keto-LCA protected against aspirin-induced intestinal epithelial damage (Figure 4H). Intestinal bile acid concentrations showed an increase of UDCA and 7-keto-LCA (Figure S7E). Moreover, analysis of the intestinal permeability showed opposing effects of UDCA and 7-keto-LCA against aspirin (Figure 4I). In addition, PAS staining further supported the protective effects of these bile acids in maintaining the intestinal mucus barrier (Figure 4J and S7F) and preserving goblet cells (Figure S7G). Both bile acids showed protective effects against aspirin-induced damage (Figure 4K–L and S7H), and cryptal cell proliferation (Figure S7J–K). TEM revealed that UDCA and 7-keto-LCA helped maintain the tight junctions (Figure S7I). The mRNA levels of tight junction markers were increased by LPG transplantation, and gavage of UDCA and 7-keto-LCA (Figure 4M). These data demonstrated that P. goldsteinii, as well as the bile acids it produces through hdhA, protects against aspirin-induced intestinal damage.

Bile acids receptor FXR mediates aspirin’s effects on the intestine

FXR, a bile acid receptor, has a key role in intestinal stem cell expansion and gut barrier^28,37. Previously, we found that modulating FXR activity promoted the transdifferentiation of gut cells toward an intestine-like phenotype³⁸. Thus, we generated a mouse model with intestine-specific knockout of FXR. Fxr-floxed mice (Fxr^fl/fl) and intestine-specific Fxr-null mice (Fxr^ΔIE, Figure S8A–B) were treated with aspirin and showed attenuated intestinal damage (Figure 5A), gut barrier destruction (Figure 5B), growth inhibition (Figure 5C), apoptosis (Figure 5D), mucosa thickness (Figure 5E), goblet cell decrease (Figure S8C) and FOBT rates (Figure S8D) caused by aspirin. More interestingly, the number of crypt cells with nuclear distribution of β-catenin and Olfm4 were decreased after aspirin treatment, which were attenuated by knockout of intestinal FXR (Figure 5F). Furthermore, the numbers of goblet cells and Paneth cells were both increased after knockout of intestinal FXR in response to aspirin (Figure 5G). To determine if inhibition of P. goldsteinii was driven by aspirin or the subsequent intestinal injury, the intestine contents were analyzed by 16S sequencing. Aspirin changed the main composition of the microbiota in both Fxr^fl/fl and Fxr^ΔIE mice (Figure S8E–F). The relative abundance of P. goldsteinii were both decreased in these mice (Figure S8G), indicating that inhibition of P. goldstenii was directly induced by aspirin, but not the result of intestine injury. Together, these data suggest that FXR partially mediate aspirin’s effects on the intestine.

Figure 5. Intestinal-specific knockout of FXR attenuates aspirin-induced intestinal damage.

(A) Intestinal-specific knockout of Fxr-mice (Fxr^ΔIE) and Control genotyped mice (Fxr^fl/fl) were treated with aspirin (2 mg/mL [11.1 mM] in drinking water) for 2 weeks and intestinal injury scores for each group shown. n=10 replicates /group. Scale bar: 200 μm.

(B) Intestinal permeability analysis. n=5 replicates /group.

(C–E) Ki 67, TUNEL and PAS staining of intestine from mice as treated in (A). n=5 replicates/group in (C) and (D), n=10 replicates/group in (E). Each point represents the average value of 10 random visions or villus and crypts for every mouse.

(F) IHC of β-catenin and Olfm4 in intestinal crypts from mice as treated in (A). n=10 replicates/group. White arrow heads mark the cells in crypt with nuclear β-catenin expression.

(G) Expression of mucin2 and lysozyme in intestinal villi and crypts from mice as treated in (A), marks Goblet cells and Paneth cells respectively. n=10 replicates/group. White arrow heads mark the cells in crypt with Lysozyme expression. Each point represents the average value of 20 random villus or crypts for every mouse.

(A–G) Two times each experiment was repeated independently with similar results. (A, B, C, E, F and G) One-way ANOVA followed by Fisher’s LSD post hoc test. (D) Two-tailed Wilcoxon rank-sum test.

P. goldsteinii promotes intestinal self-renewal by regulating bile acid metabolism

We and others previously reported that bile acids play pivotal roles in intestinal stemness,^20,27,28 and are indispensable for tissue repair.^37,39 We then asked whether P. goldsteinii and bile acids regulate intestinal stem cell function during self-repair. Aspirin was found to suppress cell proliferation in the intestinal niche, which was promoted by P. goldsteinii transplantation (Figure 6A). Moreover, the intestinal organoids cultured from primary isolated crypts also showed decreased expansion, budding numbers and surface areas in aspirin-treated mice, which were recovered by P. goldsteinii transplantation (Figure 6B). To observe the effects of P. goldsteinii on intestinal stem cells more closely, a coculture device was developed and oxygen concentrations was measured (Figure 6C). The viability of P. goldsteinii were evaluated (Figure S9A) and when cocultures with organoids, increased organoid growth was noted (Figure 6D). More specifically, an increased proliferative ability and a decreased apoptotic percentage were found in LPG-cocultured organoids compared with organoids that were cocultured with HPG (Figure 6E). The numbers of the goblet cells and Paneth cells were also increased in organoids of LPG group (Figure 6F). The co-culture system was also tested without CDCA in the upper medium, revealing that no significant differences were observed in organoid growth (Figure S9B), apoptosis (Figure S9C) and differentiation (Figure S9D). However, stem cell marker mRNAs such as Lgr5, Olfm4 and Ascl2, and intestinal permeability marker mRNAs, were increased after coculture with LPG (Figure 6G), potentially providing mechanistic insights.

Figure 6. P. goldsteinii promotes intestinal self-renewal by modulating bile acid metabolism.

(A) Ki67 and EdU staining of distal jejunum of mice treated as in Figure 4A. P. g: Live P. goldsteinii. Red/white dotted line marks single intestinal crypt. Ten random points of one sample were picked and the average value for each mouse was calculated as one single point. Red: EdU; Blue: DAPI. n=10 replicates/group.

(B) Intestinal organoids were cultured using crypts from mice treated as in (A). Forming rates were calculated by organoids/crypts ratio. Buddings were marked by red triangles. Ten random visions of each mice-generated sets of organoids were picked and data from five biological repeats are shown.

(C) Description of the co-culture system enabling mutual interaction between intestinal organoids and anaerobic bacteria. O₂ concentrations were measured 24 h after assembling. Upper media were supplemented with CDCA at a concentration of 10 μM.

(D) Live P. goldsteinii (LPG) or heat-killed P. goldsteinii (HPG) were co-cultured with secondary intestinal organoids from C57BL/6J mice for 4 days. Forming rates, budding numbers and area of organoids were compared. n=10 biological replicates/group.

(E) EdU (upper set) and TUNEL (bottom set) staining of organoids after being cocultured with HPG or LPG. White dotted line marks a single organoid. Scale bars: 20 μm. Red: EdU; Green: TUNEL; Blue: DAPI. n=20 biological replicates/group.

(F) Immunofluorescence of Mucin2 and lysozyme. Red: Mucin2 (Muc2); Yellow: lysozyme (Lyso); Blue: DAPI. n=20 biological replicates/group. White arrows marked the surface cells of organoids with positive staining.

(G) After co-culturing, organoid levels of mRNAs encoding stem cell markers and permeability markers were measured by qRT-PCR.

(H) Supplement of UDCA and 7-keto-LCA were administered to mice together with aspirin as in Figure 4H. Intestinal crypts were extracted and primary organoids were cultured, and representative images of organoids shown (left). Forming rates, budding numbers and area of organoids were analyzed (right). n=10 replicates/group.

(I) Statistical analysis of EdU (Left) and TUNEL (right) staining of organoids generated from mice treated as in (H). n=20 biological replicates/group.

(J) Staining of mucin2 (Green) and lysozyme (Red) in intestinal villi and crypts from mice as treated in (H). Each point represents the mean values of 30 random villus or crypts for every mouse. White arrow heads mark Paneth cells. n=10 replicates/group. (A-J) Three times each experiment was repeated independently with similar results. (A, B, H, I and J) One-way ANOVA followed by Fisher’s LSD post hoc test. (C, D and G) Fisher’s LSD post hoc test. (E and F) Two-tailed Wilcoxon rank-sum test.

More interestingly, both UDCA and 7-keto-LCA abrogated the suppression of primary organoids cell proliferation induced by aspirin (Figure 6H). Moreover, it showed recovered growth and stem cell expansion in organoids derived from bile acid-treated mice (Figure 6I and Figure S9E). Similar results were found in secondary organoids treated with aspirin in combination with UDCA and 7-keto-LCA in vitro (Figure S9F). The numbers of goblet cells and Paneth cells were both increased by UDCA and 7-keto-LCA supplementation in vivo (Figure 6J). These effects were also consistent with those in the intestinal cell line HT-29, showing the antiapoptotic and proliferative functions of UDCA and 7-keto-LCA against aspirin-induced intestinal damage (Figure S9G–H). Together, these data suggest that P. goldsteinii and its bile acid metabolites help maintain intestinal stem cell function during intestinal self-repair.

P. goldsteinii-producing 7-keto-LCA is an FXR antagonist that facilitates Wnt signaling

Considering that microbial UDCA and TβMCA are both proven intestinal FXR antagonists,^37,40,41 we tested the function of 7-keto-LCA on FXR. 7-Keto-LCA was docked into the human FXR ligand binding domain (PDB 3DCT, R4), establishing bonds with surrounding residues (Figure 7A and S10A–I). We then applied the TR-FRET FXR coactivator assay to determine the direct binding. Neither 7-keto-LCA and UDCA showed agonistic activity as compared to CDCA (Figure S10J). Moreover, we found that 7-keto-LCA (IC50 = 32.8 μM) and UDCA (IC50 = 59.7 μM) were both FXR antagonists (Figure 7B). The expression of two FXR downstream markers, fibroblast growth factor 19 (FGF19) and small heterodimer partner (SHP), were increased by CDCA. But this was abrogated by TβMCA, UDCA, and 7-keto-LCA in both intestinal cell lines (Figure 7C and S10K–L). Luciferase reporter gene assays further revealed that both UDCA and 7-keto-LCA markedly inhibited FXR transactivation activity (Figure 7D, Figure S10M). The IC50s, calculated through this method (Figure S10N), were similar to TR-FRET data. These findings were also verified in organoid co-culture models (Figure S10O). Since LCA was also reported as a pregnane X receptor (PXR) agonist⁴², we measured mRNA levels of PXR downstream targets genes (Figure S10P). These results showed that PXR was indeed activated by LCA, but not by 7-keto-LCA, suggesting that 7-keto-LCA may not activate PXR as suspected.

Figure 7. 7-keto-LCA is an FXR antagonist that facilitates Wnt signaling.

(A) 7-Keto-LCA is identified as a possible FXR antagonist as revealed by docking of 7-keto-LCA (carbon atoms in red) into the human FXR binding pocket (PDB 3DCT).

(B) TR-FRET FXR coactivator recruitment assay to assess whether 7-keto-LCA and UDCA are FXR antagonists in the presence of the agonist CDCA (20 μM). n=4 replicates/treatment. LCA was used as a negative control.

(C) HT-29 cells were treated with different bile acids and the synthetic FXR agonists Fexaramine D (FexD) and GW4064. The mRNA levels of FXR target genes were quantified by qRT-PCR. n=5 replicates/group.

(D) pGL3 luciferase reporter gene vector was containing the FXR-binding sequence was generated (left). Luciferase activity were detected after treatment of control (DMSO), CDCA, 7-keto-LCA and UDCA. n=5 replicates/treatment.

(E) Secondary organoids were treated with CDCA (20 μM), 7-keto-LCA (20 μM) or UDCA (20 μM) for 5 days and next-generation sequencing conducted. Heatmap of changes in stem cell gene signature (Lgr5-Ascl2) is shown.

(F) PCA analysis of indicated gene signature in (E).

(G) Gene sets enrichment analysis (GSEA) of sequencing results between 7-keto-LCA group and DMSO group, targeting proliferation and apoptosis pathways.

(H) qRT-PCR results of mRNAs encoded by the FXR targets Fgf15 and Shp (left) and intestinal stem cell markers including Lgr5, Ascl2 and Olfm4 (right) in organoids treated as indicated in (E). n=5 replicates/treatment. mRNA levels were normalized as to β-actin.

(I) Expression of Catb and Olfm4 mRNAs in intestinal crypts from mice as treated in Figure 4A and H. White arrowheads mark cells in the crypt with nuclear β-catenin expression. n=10 biological replicates/group. Each point represents the mean values of 20 random crypts for every mouse.

(B–D, H–I) Three times each experiment was repeated independently with similar results. (E–G) Two times each experiment was repeated independently with similar results. (B) Non-liner regression curve fit. (C, D, H and I) One-way ANOVA followed by Fisher’s LSD post hoc test.

To assess whether UDCA and 7-keto-LCA regulate intestinal stemness, RNA-sequencing of intestinal organoids treated with bile acids was performed. Interestingly, both UDCA and 7-keto-LCA promoted the expression of the stem cell signature (Lgr5-Ascl2) (Figure 7E). Corresponding PCA revealed separated clusters between the control and CDCA groups (Figure 7F). Furthermore, treatment with 7-keto-LCA counteracted the observed suppression of many genes, most notably those involved in the Wnt signaling and proliferation-related pathways but not in the apoptosis pathway (Figure 7G, and S10Q). These findings provide possible mechanistic insights into how FXR antagonists preserve stem cell expansion. This result was consistent with the PCR results showing decreased expression of FXR target gene mRNAs and increased expression of stem cell marker mRNAs in organoids (Figure 7H). Moreover, nuclear translocation of β-catenin was increased in 7-keto-LCA-treated cells lines and organoids (Figure S10R–S). The in vivo results also showed that 7-keto-LCA and P. goldsteinii transplantation both abrogated the suppression of β-catenin activity by aspirin in the intestinal crypts (Figure 7I). A possible explanation might be that the physical binding of FXR with β-catenin was modulated by bile acids, as coimmunoprecipitation (co-IP) showed increased levels of the FXR-β-catenin interaction after UDCA and 7-keto-LCA treatment (Figure S10T). These data suggest that Fxr antagonist 7-keto-LCA facilitates Wnt signaling to preserve the stemness of intestinal crypt cells.

DISCUSSION

It was previously established that the microbiota are key regulators of the gut bile acid pool composition^33,37,43 and intestinal stem cell function.^27,28 Herein, we identified P. goldsteinii as a critical regulator of bile acid metabolite patterns during aspirin-induced intestinal damage and the subsequent self-repair driven by intestinal stem cells. This was due to its ability to ameliorate aspirin-induced insufficiency of specific bile acids and sustaining the function of intestinal stem cells (Graphic abstract).

Although recent studies are providing increasing evidence that medications can contribute substantially to microbiota variation,^44,45 no studies have explored the interaction between the gut commensal microbiota and aspirin. Here, we established a single-drug-intervention cohort involving healthy volunteers. The results revealed that the consistent decrease in Parabacteroides abundance was due to the direct effect of aspirin treatment. Another report also demonstrated that there was a remarkable decrease in Parabacteroides levels in healthy individuals taking aspirin, based on 16S sequencing data.⁴⁶ Interestingly, Parabacteroides was shown to correlate with protection against intestinal inflammation.^47–49 However, few of these reports focused on the effects of Parabacteroides on host bile acid metabolism. Among the Parabacteroides, we found that only the species P. goldsteinii was strongly correlated with UDCA and 7-keto-LCA levels, which were both decreased after aspirin treatment. The gut microbiota is known to influence host intestinal homeostasis, mostly via metabolic pathways, particularly bile acid metabolism³⁷. We found that the levels of UDCA and its precursor 7-keto-LCA, identified in this study as endogenous FXR antagonists, were decreased after aspirin treatment. Intestinal CDCA is converted to 7-keto-LCA and UDCA by gut microbial 7-α/β-HSDH.³⁷ Herein, we identified the hdhA gene in P. goldsteinii and activity toward 7-keto-LCA in vivo and in vitro. These findings were consistent with a previous report in Parabacteroides.³³ Interestingly, our results also suggest that the reduction of the hdhA is less striking than the reduction of P. goldsteinii abundance, indicating that other gut commensals carry this gene but are not affected by aspirin. One possible reason for their failure to protect the intestine may be that allelic enzymes encoded by the hdhA gene in different bacteria have low enzyme activity. Future studies are needed to explore the function of hdhA genes encoded in the genomes of other microorganisms and their activity in producing 7-keo-LCA.

It should be noted that humans lack key enzymes to generate TβMCA and βMCA in liver and intestine as these are the main bile acid components in mice.⁵⁰ However, the FXR antagonist 7-keto-LCA, uncovered in the present study, was found to be relatively abundant in the human gut, suggesting that it could be important in mediating FXR signaling in human intestine. From this point of view, 7-keto-LCA and UDCA may be playing similar roles in the human gut as TβMCA does in mice⁵¹. The other issue is that recent reports suggest a potential role for Parabacteroides in deconjugation of the taurine group from Tα/βMCA catalyzed by BSH.³⁶ However, the current study indicates increasing concentrations of TβMCA in intestine after P. goldsteinii transplantation. One possibility may be that the anabolism of α/βMCA was promoted by P. goldsteinii garaging.

The present study is in agreement with previous work showing the effects of P. goldsteinii on intestinal permeability. Notably, two earlier studies revealed a protective role for LPS derived from P. goldsteinii, which acts as a toll-like receptor 4 antagonist in the gut barrier.^47,48 However, to our knowledge, few studies have linked gut injury to the effects of aspirin on gut dysbiosis and the subsequent regulation of bile acids. Here, the current findings suggest that aspirin damages the intestinal barrier at least partly through its influences on the gut microbiota-bile acid axis. These effects increase intestinal permeability and result in apoptosis of epithelial cells and decreased capacity for intestinal repair. It was reported that the aspirin eugenol ester induced metabolomic changes that influence FXR signaling in liver²⁹. These findings seemed to be exclusively limited to high-fat diet-induced animal models. By contrast, our study suggests that aspirin itself regulates the microbiota-bile acid axis in the gut. We found that P. goldsteinii depleted by aspirin plays vital roles in sustaining intestinal stem cell proliferation by producing 7-keto-LCA. Bile acids, such as UDCA have been clinically used for years. Thus, this study might shed light on developing practical strategies to prevent aspirin-associated enteropathy. Although a previous study indicated that some bacteria might have profound effects on the absorption of aspirin,¹⁴ our data showed different results. This may be due to the concentrations of aspirin used in the drinking water of mice, the animal model employed, and the use of different antibiotics.

Various studies have proven that aspirin inhibits intestinal cell proliferation in multiple cancer models, most of which consider the inhibition of COX1/2 to be central to its preventive mechanisms perpetuating proinflammatory signals that promote proliferation.^52,14 Here, we explored the influences of aspirin on intestinal stemness, and found that these effects are mediated by P. goldsteinii. It’s accepted that interactions between gut and its sustainable proliferation and differentiation are largely determined by its microbial composition, predominantly through secreted metabolites. Several studies have suggested the pivotal roles of microbial bile acids in regulating the function and self-renewal of intestinal stem cells.^26,53 In particular, we recently reported that cholic acid inhibits peroxisome proliferator-activated receptor α (PPARα), resulting in impaired intestinal stem cell renewal.²⁰ Another study suggested that aspirin reduced the stem-like state in the colon stem cell niche that is mediated by its direct inhibitory effects on the Wnt pathway.²⁵ The interplay between FXR activity and Wnt signaling was verified in earlier studies.²⁸ Here, we provide evidence for a 7-keto-LCA-FXR-Wnt signaling cascade that is important in maintaining the self-renewal of intestinal stem cells, and its inhibition due to aspirin treatment and the subsequent depletion of P. goldsteinii.

The lack of tools for the coculture of intestinal organoids with anaerobic bacteria has been a challenge for exploring the microbiota-host interactions, due to the contradictory oxygen conditions required to culture them. Several methods were reported, including the injection of bacteria into organoids⁵⁴ and the generation of monolayer cells from organoids.^55,56 These methods are promising but technically difficult to apply due to the complexity of the system structure. Here, we developed a simple and practical coculture method based on the modification of easily assembled transwell inserts. This system partially separates the medium into two physical spaces with different oxygen conditions while enabling metabolite interaction between anaerobic bacteria and organoids. Although absence of the direct attachment of bacteria to epithelial cells may limit the application of this method, it is still a useful transwell-based barrier system for the assessment of the effects of bacterial metabolites on the intestinal microenvironment.

The limitations of this study need to be discussed. Other aspirin-suppressed species that have potential functions should not be ignored. For example, P. distasonis was reported to produce bile acid metabolites that can serve as TGR5 and FXR agonists and potentially can represented a probiotic in the treatment of inflammatory arthritis⁵⁷ and hepatic fibrosis.³⁶ Thus, further studies should focus on the function of these gut commensals in regulating gut dysbiosis and microenvironment. In addition, direct evidence in vivo is needed to support the potential for FXR antagonism by P. goldstenii or 7-keto-LCA. Another issue is that the Cre+ mouse could be included as a control group to exclude the possibility that Cre expression might be toxic to intestinal cells in which it is expressed, although our data do not support such an interaction.

In conclusion, the current study established that the interaction between the intestinal microbiota and bile acids is critical for maintaining intestinal stem cell function and the gut barrier through their enzyme-mediated modification of unconjugated bile acids. These findings suggest that microbiota-based therapies may potentially alter bile acid patterns due to aspirin-induced decrease of gut commensal microbiota that metabolize bile acids. The subsequent phenotypes of impaired gut barrier function and stemness in gastrointestinal disorders may be reversible by supplementation of bile acids as well as commensal gut bacteria.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for reagents should be directed to and will be fulfilled by the corresponding author Yue Wu (yue.wu@xjtu.edu.cn).

Materials availability

This study did not generate new unique reagents

Data and code availability

• Data of metagenomic sequencing, 16S rRNA sequencing, and metabolomics have been deposited at NGDC database: PRJCA018155) or MetaboLights database: MTBLS8146). Data of RNA sequencing of organoids has been deposited at NGDC database: OMIX005261. These data are publicly available as of the date of publication. Their accession numbers can be found in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available upon reasonable request from the corresponding author, Yue Wu (yue.wu@xjtu.edu.cn).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-NR1H4	Abcam	Cat# ab235094;
Rabbit anti-NR1H4	Proteintech	Cat#25055-1-AP; RRID: AB_2879874
Rabbit anti-β-Catenin	Cell Signaling Technology	Cat# 8480: RRID: AB_11127855
Ki67 antibody	Abcam	Cat#ab15580
Rabbit anti-Muc2	Abcam	Cat# ab272692; RRID: AB_2888616
Rabbit anti-Lysozyme C-1/2	Cell Signaling Technology	Cat# 60487
Rabbit anti-Olfm4	Cell Signaling Technology	Cat# 39141; RRID: AB_2650511
Anti-rabbit IgG, HRP-linked Antibody	Cell Signaling Technology	Cat#7074; RRID: AB_2099233
Alexa Fluor 488-conjugated goat anti-rabbit IgG	Abcam	Cat# ab150077; RRID: AB_2630356
Alexa Fluor 594-conjugated goat anti-rabbit IgG	Abcam	Cat# ab150080; RRID: AB_2650602
Bacterial and virus strains
Parabacteroides goldsteinii	Japan Collection of Microorganisms	Cat# JCM13446
Parabacteroides Merdae	Japan Collection of Microorganisms	Cat# JCM 13405
Parabacteroides distasonis	Japan Collection of Microorganisms	Cat# JCM13400
Escherichia coli DH5α	Thermo Scientific	Cat#EC0112
Biological Samples
Human feces	-	-
Chemicals, peptides, and recombinant proteins
Aspirin	MedChemExpress	Cat# HY-14654
NaCl	Sigma-Aldrich	Cat# 7647-14-5
FITC-dextran (4kDa)	Sigma-Aldrich	Cat# 46944
5-ethynyl-20-deoxyuridine (EdU)	MedChemExpress	Cat # HY-118411
L-Glutamine	Sigma-Aldrich	Cat# 49419
L-Proline	Sigma-Aldrich	Cat# P0380
L-Citruline	Sigma-Aldrich	Cat# C7629
T-beta-muricholic acid (TβMCA)	MedChemExpress	Cat# HY-135103
Ursodeoxycholic acid (UDCA)	MedChemExpress	Cat# HY-13771
7-Ketolithocholic acid (7-keto-LCA)	MedChemExpress	Cat# HY-W018512
Fexaramine D	MedChemExpress	Cat# HY-10912
GW4064	MedChemExpress	Cat# HY-50108
Chenodeoxycholic acid (CDCA)	MedChemExpress	Cat# HY-76847
Cholic acid (CA)	MedChemExpress	Cat# HY-N0324
Lithocholic acid (LCA)	MedChemExpress	Cat# HY-B0172
DMEM medium	ThermoFisherScience	Cat# 11965126
IntestiCult Organoid Growth Medium	Stemcell Technologies	Cat# 06005
Gentle Cell Dissociation Reagent	Stemcell Technologies	Cat# 07174
D-PBS	Stemcell Technologies	Cat# 37350
DMEM/F12	Stemcell Technologies	Cat# 36254
GFR Matrigel	Corning	Cat# 356231
24 wells, polystyrene plate	Corning	Cat# CLS3526
DAPI	Sigma-Aldrich	Cat# D9542
Difco Fluid Thioglycollate Medium	BD Biosciences	Cat# 0048064
Opti-MEM medium	Gibco	Cat# 31985-070
Lipofectamine™ 2000	Invitrogen	Cat# 11668-019
Critical commercial assays
ToxinSensor Chromogenic LAL Endotoxin Assay Kit	Make Research Easy	Cat# L00350
Mouse Diamine Oxidase (DAO) ELISA Kit	Bioswamp	Cat# MU30134
Fecal Occult Blood Test (FOBT) kit	ABON	Cat# V277200
TdT-mediated dUTP nick-end labeling (TUNEL) kit	BIOSCIENCE	Cat# T6013L
RNeasy Micro Kit	Qiagen	Cat# 74004
Evo M-MLV RT Kit with gDNA Clean for qPCR	Accurate Biology	Cat# AG11705
SYBR Green Premix Pro Taq HS qPCR kit	Accurate Biology	Cat# AG11701
Periodic Acid Schiff (PAS) Stain Kit	Solarbio Life Sciences	Cat# G1280
Modified Hematoxylin-Eosin (HE) Stain Kit	Solarbio Life Sciences	Cat# G1121
Annexin V-FITC Apoptosis Detection Kit	BD Biosciences	Cat# 556547
CCK8 kit	MedChemExpress	Cat# HY-K0301
LanthaScreen® TR-FRET FXR Coactivator Assay Kit	ThermoFisherScience	Cat# PV4833
Co-immunoprecipitation (Co-IP) Kit	ThermoFisherScience	Cat# 26149
Experimental models: Cell lines
HT29	ATCC	Cat#HTB-38
IEC-6	ATCC	CRL-1592
Deposited Data
Raw data files for 16S / metagenomic sequencing	This work	PRJCA018155
Raw data files for metabolomic analysis	This work	MTBLS8146
Raw data files for genome sequencing of P. goldsteinii	This work	PRJCA018155
Raw data files for RNA sequencing of organoids	This work	OMIX005261
Experimental models: Organisms/strains
Mouse C57BL/6J	Beijing Vital River Laboratory Animal Technology	N/A
Nr1h4 flox [Fxrfl/fl] mice	Cyagen Biosciences	Cat# S-CKO-04890
PVillin-Cre mice	Cyagen Biosciences	Cat# T000142
Oligonucleotides
Primers see Table S4
Recombinant DNA
pGL3-basic	Shanghai Sangon Biotech	N/A
pGL3-2000bp~+500bp DNA sequence of SHP	Shanghai Sangon Biotech	N/A
pGL3--2000bp~+500bp NA sequence of FXR	Shanghai Sangon Biotech	N/A
Software and algorithms
Graphpad Prism 9	GraphPad Software	N/A
FlowJo	Tree Star,Inc.	https://www.flowjo.com/
R v4.3.0	R Development Core Team	https://www.r-project.org/
MetaboAnalyst 5.0	Xia J et al, Nucl. Acids Res (2009)	https://www.metaboanalyst.ca/
Fast QC v0.11.8	FastQC	https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
MetaPhlAn2 v2.7.7	Truong DT et al, Nature Methods (2015)	https://huttenhower.sph.harvard.edu/metaphlan2/
ImageJ	National Institutes of Health (NIH)	https://imagej.nih.gov/ij/
Other
4-well Chamber Slide w/removable wells	ThermoFisherScience	Cat# 154917
Transwell 12mm (0.4um)	CORNING	Cat# 3460
Falcon® 70 μm cell Strainer	CORNING	Cat# 352350

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Ethics statement

This study was approved by the Ethics Committee of Xi’an Jiaotong University (Approval No. 2020-299) and comply with the Declaration of Helsinki principles and local laws. All animal procedures adhered to the US National Institutes of Health’s Guide for the Care and Use of Laboratory Animals. Strain culture experiments were performed following a protocol approved by the Xi’an Jiaotong University Administrative Panel on Biosafety. Metabolomics and metagenomic studies on human stool samples were performed with approval by the Hospital’s Protection of Human Subjects Committee.

Human subjects

Human feces samples (BA-AA cohort in Figure 1A, Supplementary Table S1) were collected from 23 healthy volunteers who received oral aspirin treatment for 30 d. It is recommended by European Society of Cardiology that the oral dose of aspirin should preferably be 75–150 mg for all patients after percutaneous coronary intervention (PCI)³. Thus, a low dose of aspirin (100 mg q.d.) was used for all volunteers to determine its effects on the gut microbiota. All of the subjects enrolled satisfied the following criteria: Healthy volunteers aged 40- to 60-years-old and had not been treated with aspirin with the past 1 year before this study. Human feces samples of aspirin user/non-user cohort (Figure S5H–I) were collected from the Health Examination Center of First Affiliated Hospital of Xi’an Jiaotong University. This cohort include 70 aspirin nonusers and 105 aspirin users (100 mg q.d. for 1–3 months, Supplementary Table S2). All human subjects are yellow race, East Asian ancestry and the ethnicity is Han Chinese. The information of human subjects related to sex is shown in Table S1 and S2. The socioeconomic status of the study participants was not recorded. After collection, feces were snap frozen in dry ice, and stored at −80 °C until analysis. This study protocol was approved by the Health Research Ethics Board of First Affiliated Hospital of Xi’an Jiaotong University, and the study was conducted in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent. The metagenomic data in a reported study involving healthy controls and patients with cardiometabolic disease³⁰ were retrieved and divided into two groups (Aspirin non-user N=321, Aspirin User N=573) according to the history of aspirin usage (Figure S1H). The retrieved microbial data of these patients and basic information including sex were shown in Supplementary Table S3. The information of gender, race, ancestry, ethnicity and socioeconomic status of the participants was unable to retrieve from the primary database.³⁰

Mice

Female C57BL/6J mice 8- to 10-week-old (body weights around 20 g) were obtained from Beijing Vital River Laboratory Animal Technology Co. Ltd. Preliminary results revealed that the intestinal injury induced by aspirin was not related to mouse gender. Because female mice are generally more docile than male mice, female mice were used to establishment the animal models. Nr1h4=flox [Fxr^fl/fl] and PVillin-Cre mice were purchased from Cyagen Biosciences Inc (Beijing, China). These two mouse lines were crossed to generate the Fxr^fl/fl Villin-Cre (+) genotype and the Fxr^fl/fl controls and designated as gut-specific knockout mice (Fxr^ΔIE) as previously described^41,58. All animal protocols were approved by the Animal Care and Use Committee of Xi’an Jiaotong University. All the mice were housed in a standard specific-pathogen-free (SPF) environment. Eight-week-old female C57BL/6J mice were fed with normal diet and drinking water with vehicle or 2 mg/ml (11.1mM) aspirin for 0, 1, 3, 7 and 14 days. Non-fasting animals were anaesthetized and sacrificed at each time point respectively. Intestinal contents and small intestinal tissue were collected. In Figure 2, two groups of mice were pretreated with 3-day neomycin (1mg/ml in drinking water), along with vehicle or 2 mg/ml (11.1mM) aspirin for 2 more weeks. The other two groups of mice were administered drinking water with vehicle or 2 mg/ml (11.1 mM) aspirin. To determine the aspirin doses for mice, adjustment of human doses (100 mg/day/individual) were calculated, which is transformed based by body surface area. The body surface area was calculated through the Meeh-Rubner formula:

$$\text{Body}\text{Surface}\text{Area}\left({\text{m}}^2\right)=9.1\times \text{body}\text{weight}{\left(\text{g}\right)}^{2/3}/10000.$$

Thus, human doses (100 mg/60 kg/day) will be equivalent to mouse doses of 11.1 mg/20 g/day. It was reported that the volume of daily water consumption per mouse is 4 to 6 mL in average. After calculation, the concentration of aspirin in drinking water was to 2 mg/mL (equals 11.1 mM).

For the germ-free mouse experiment, the mice were administered sterilized saline or aspirin as indicated. Female GF mice were naturally reconstituted with gut bacteria by moving their cages to an SPF environment and co-housing with SPF mice (2 GF mice housed with three SPF mice in one cage) for 2 weeks to generate the rGF mice. Intestinal contents from aspirin-fed GF or rGF mice were collected and subjected to 16S sequencing. For gavage of P. goldsteinii, the mice were orally administered saline or bacteria by gavage every other day.

For bile acid intervention (Figure 4H), two groups of mice (10-week-old female C57BL/6J) were treated with UDCA (50 mg/kg/d) or 7-ketoLCA (50 mg/kg/d) through daily oral gavage, followed by 2 mg/ml (11.1mM) aspirin in combination with UDCA (50 mg/kg/d) or 7-keto-LCA (50 mg/kg/d) by oral gavage daily. When preparing the bile acids before gavage, the bile acids were dissolved in dimethyl sulfoxide (DMSO) first, and then the DMSO containing bile acids were dissolved in saline. For the other two groups, 10-week-old female C57BL/6J mice were administered drinking water with vehicle or 2 mg/ml (11.1mM) aspirin for 2 weeks. For the Fxr gene function experiments in Figure 5 and Figure S8, 8-week-old female Fxr^fl/fl and Fxr^ΔIE mice were treated with aspirin (2 mg/ml, 11.1mM) or vehicle-containing drinking water for 2 weeks and anaesthetized and sacrificed at the end of experiment.

Bacteria

P. goldsteinii, P. Merdae, P. distasonis and E.coli was obtained from the Japan Collection of Microorganisms (JCM, see the Key Resources table), which were previously isolated from healthy individual^59,60. It was routinely cultured in medium at 37 °C in a COY type B anaerobic chamber using gas mix consisting of 5% hydrogen, 10% carbon dioxide and 85% nitrogen, and hydrogen was maintained at ~3.3% using an anaerobic gas infuser.

Organoids and cell culture

Crypts were isolated from mice as described with modifications^20,54. Small intestines were washed in ice-cold PBS containing 2% BSA. The intestines were opened longitudinally and gently washed with cold PBS to remove the luminal contents. The tissue was cut into 2 mm² fragments and further washed 20 times with cold PBS until the supernatant was clear. Intestinal samples were then incubated with 20 ml Gentle Cell Dissociation Reagent (#07174, NovoBiotechnology, Beijing China) at 20 rpm for 15 min at room temperature and then the lysis buffer was replaced with cold PBS containing 0.1% BSA. The fragments were vigorously shaken and the suspension filtered through a 70-mm filter. Enriched crypts were washed once with cold PBS and resuspended with DMEM/F12 medium (Gibco). The crypts were counted, plated with Matrigel (Corning #356231) and IntestiCult Organoid Growth Medium with Supplement 1 and 2 (STEMCELL Technologies #06005). Organoids were exposed to DMSO, aspirin (2 mM), UDCA (20 μM, MCE#HY-13771), 7-ketoLCA (20 uM, MCE #HY-W018512), TβMCA (20 μM, MCE#HY-135103) or CDCA (10 μM MCE#HY-76847) for 5 days or as indicated. The human intestinal cancer cell lines HT29 and rat small intestinal cell lines IEC-6 were acquired from ATCC and cultured according to supplier’s instructions. In Figure 7B–D, the cells were synchronized for 12 h in serum-free medium and exposed to TβMCA (20 μM, MCE#HY-135103), UDCA (20 μM, MCE #HY-13771) and 7-keto-LCA in the presence of fexaramine D (FexD, 10 μM, MCE #HY-10912) or GW4064 (1 μM MCE #HY-50108) for 24h as described³⁸. RNA was extracted by TRizol after exposure.

METHOD DETAILS

DNA extraction and preparation

Genomic DNA from human stool samples clinically collected was extracted by Stool Genomic DNA Kit (#CW20925, CWBIO, China). Degradation and contamination of DNA were monitored on 1% agarose gels. DNA purity was checked by use of a Nano-Photometer spectrophotometer (IMPLEN, CA, USA). The Qubit DNA Assay Kit in Qubit 2.0 Fluorometer (Life Technologies, CA, USA) measured the concentration of DNA.

Metagenomic sequencing

Metagenomics sequencing were conducted as described⁵⁸. DNA, 700 ng per sample, was used for sample preparation. Sequencing libraries were established by NEB Next Ultra DNA Library Prep Kit for Illumina (# E7370L, NEB, USA); the manufacturer’s recommendations and index codes were adopted to attribute sequences to each sample. The fragmented DNA ends were repaired, polyA-tailed, and ligated with a sequencing adaptor for Illumina sequencing. PCR amplification and purification (AMPure XP system) were performed. The concentration of DNA was measured by the Qubit DNA Assay Kit in Qubit 2.0 Fluorometer (Life Technologies, CA, USA) and diluted to 2 ng/μL. The insert size of library was assessed by using the Agilent Bioanalyzer 2100 system. qPCR ensured accurate concentration (> 3 nM) of the library. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using HiSeq 4000 PE Cluster Kit (Illumina) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina HiSeq 4000 platform and 150-bp paired-end reads were generated.

16S rRNA gene amplicon sequencing

In Figure 1H–J and Figure S1I, fecal samples were collected for the analysis and comparison of longitudinal changes in the microbiota. For Figure 2 and 3, intestinal contents including cecum contents were collected for 16S sequencing and metabolomic analysis. Briefly, the mice were killed and the whole length of the small intestine and the cecum collected. The gut was flushed three times with cold PBS in order to remove all of the intestinal contents, including the cecum contents from the gut. The intestinal contents, including feces left in the intestine, were collected for the following experiments. The amplicon libraries were created by PCR amplification of the V4 variable region using primers with common adaptor sequences: 515F and 806R⁶¹. Barcoded reverse and non-barcoded forward primers were used with Taq DNA polymerase Master Mix (TONBO Biosciences) according to the manufacturer’s directions. Samples were amplified in triplicate with the thermocycler protocol as described⁶². 16S rRNA gene amplicons were cleaned using Mag-Bind RxnPure Plus beads (Omega Bio-tek). Samples were mixed in equimolar amounts before sequencing on Illumina’s MiSeq platform using 2×250 bp paired-end runs. The resulting ASVs were assigned taxonomy by mapping with the SILVA 132 database⁶³.

The α-diversity, β-diversity and network visualization

The α-diversity analysis was performed using usearch software as described⁶⁴, while β-diversity analysis was performed using the ‘vegan’ package in R v 3.6.1 to assess the feature richness in BA-AA cohorts individuals or mice feces. The α-diversity indexes (Shannon, Simpson and Chao1 indexes) was calculated by usearch based on the relative abundance of each taxon. Chao 1 index represents the number of detected species, which is equivalent to the richness of the species, and disregards the abundance of each species. In contrast, Shannon and Simpson indexes are related to the richness of species and the relative abundance of each species. The Shannon index is weighted in favor of the richness of species, while the Simpson index places more emphasis on the relative abundance of each species. β-Diversity analysis was conducted using partial least squares discriminant analysis (PLS-DA) or principal component analysis (PCA) based on Bray-Curtis dissimilarity index matrices. The networks of species-species correlations were visualized by Cytoscape v3.9.1 and ChiPlot (https://www.chiplot.online/). The Mantel test were conducted and visualized in R v3.6.1 (see key resources table) using the “ggcor” package.

Bile acid analysis

Bile acids were analyzed as described.⁵⁸ In this study, we utilized ultra-high performance liquid chromatography (UHPLC) coupled with a quadrupole time-of-flight (QTOF) to identify bile acids in human and mice feces. A triple quad with a 100X dynamic range were utilized for precise quantification of bile acids using authentic standard (standard curves). Stool samples or intestinal content samples were dissolved and bacteria medium samples were prepared by precipitation. The bile acid concentrations in the supernatants were measured by a UPLC/Synapt G2-Si QTOF MS system (Waters Corp., Milford, MA) with an ESI source. Chromatographic separation was operated on an Acquity BEHC18 column (100 mm × 2.1 mm i.d., 1.7 μ m, Waters Corp.). Column temperature was 45 °C, and the flow rate was 0.4 ml/min. The mobile phase included a mixture of 0.1% formic acid in water and 0.1% formic acid in acetonitrile. The gradient elution was applied and MS detection proceeded in negative mode. A mass range of m/z 50 to 850 was acquired. Standards for all bile acids were used to identify the different bile acid metabolites detected by LC-MS.

Fecal microbiota transplantation (FMT)

For FMT experiments in Supplementary Figure S2H–I, donor mice (8-week-old, female, n=7) were untreated (NC group), treated with saline (Saline group) or aspirin (Asp group) for more than 2 weeks. Fecal samples (above 150 mg) of each mouse were collected from day 14 to day 20 and stored in sterile tubes and homogenized in 1 ml of PBS. After centrifugation (2,000g for 10 min), bacteria-enriched supernatants were collected and centrifuged (5 min at 15,000g). Bacterial pellets were washed twice with PBS, resuspended in 700 μL of saline with 20% (v/v) glycerol, and stored at −80°C. For recipient mice, control mice and aspirin-treated mice (2 weeks, saline or aspirin in drinking water [2 mg/mL]) were treated daily with fecal microbiota transplants from each donor group daily via oral gavage (200 ul once daily) for 2 weeks as described^47,65. The method of euthanasia and collection were as described above.

For P. goldsteinii and E. coli transplantation, bacteria were collected by centrifugation and resuspended in sterilized saline. Mice were treated with 2×10⁸ colony-forming units of live P. goldsteinii (200 μL, LPG), wild-type P. goldsteinii (PGwt), hdhA-deficient P. goldsteinii (PGΔhdhA), wild-type E. coli DH5α(EcD:wt) and E.coli transfected with hdhA (EcD;hdhA), by oral gavage every other day (3 times a week) as described previously.⁴⁷ Heat-killed P. goldsteinii (200 μL, HPG) was prepared by heating bacteria at 100 °C for 15 min in a water bath. There were six groups with four groups treated with LPG/HPG/PGwt/ PGΔhdhA, followed by 2 mg/ml (11.1 mM) aspirin in drinking water along with FMT for 2 weeks (3 times/week). The other two groups after 2 weeks of saline treatment, were administered drinking water containing vehicle or aspirin (2 mg/ml, 11.1mM) for 2 weeks.

Co-culture of P. goldsteinii and organoids

An anaerobic environment chamber was designed using a 0.4 um transwell chamber in which P. goldsteinii was cultured. Medium was replaced every other day. The cell inserts together with upper anaerobic medium were changed every 48 h, which help keep the medium anaerobic as longer as we can. CDCA (10μM) were added to the upper medium in Figure 6D–G. The images of organoids were captured and the number of buds were counted under an Olympus microscope. Clonogenicity (colony-forming efficiency) was calculated by plating 150 crypts and assessing organoid formation 3–7 days or as specified after initiation of cultures. For secondary organoid assays, primary organoids were dissociated by Gentle Cell Dissociation Reagent (GCDR, STEMCELL Technologies), centrifuged and resuspended with cold DMEM/F12. Cells were seeded onto Matrigel as previously described. Secondary organoids were enumerated on day 3–5, unless otherwise specified.²⁰

Bacterial strains and culture conditions

Bacteria are routinely cultured in medium at 37 °C in a COY type B anaerobic chamber using gas mix consisting of 5% hydrogen, 10% carbon dioxide and 85% nitrogen, and hydrogen was maintained at ~3.3% using an anaerobic gas infuser. All media and plastic ware were pre-reduced in the anaerobic chamber for at least 24 h before use. Where appropriate, OD measurements were performed in Balch-type anaerobic tubes using a GENESYS 30 spectrophotometer (Thermo Fisher Scientific). For assays in metabolites analysis of cell suspensions in Figure 3, P. goldsteinii was cultured in defined medium with supplemented aspirin (0–4 mmol/L) with or without amino acids (L-glutamine #49419, L-citruline #C7629, L-proline #P0380, all purchased from Sigma-Aldrich, China) for the indicated times. Then the cells were harvested by centrifugation and resuspended to an OD of ~1.0 before the addition of substrates.

Construction of hdhA-deficient P. goldsteinii

The hdhA internal fragment (789 bp) gene was transformed into a pGERM suicide vector that contained P. goldsteinii (ermG) and Escherichia coli (bla)-selective markers as described⁶⁶. After conjugation, the E. coli donor strain was aerobically incubated in Luria broth supplemented with ampicillin and daptomycin at 37 °C until reaching an OD600 of 0.3. The P. goldsteinii recipient was anaerobically cultured in YCFA medium at 37 °C until an OD600 of 1.0 was reached. The donor culture (1.0 ml) and the recipient culture (1.0 ml) were mixed and further centrifuged to provide the mating mix that was next placed on YCFA-medium agar plates. The plate was transferred into an anaerobic station after incubation under aerobic conditions and then the obtained bacteria were suspended in 5 ml Gifu anaerobic medium. After incubation, cell suspension was added to the YCFA-medium agar plates with erythromycin for mutant selection. The erythromycin-resistant colonies were acquired after 4 d of anaerobic incubation. Plasmid insertion into the target gene was then verified by PCR in the genome DNA.

Construction of E.coli:hdhA

To construct the hdhA-expressing E. coli strain for colonization in mice, the gene encoding hdhA was cloned from related bacterial isolates of P. goldsteinii, whose signal peptide was replaced with universe signal peptide pelB and was integrated into the genome of E.coli DH5α at the attB site using the CRISPR-Cas9 system. Then the pTargetF plasmid was eliminated by culturing strains with 0.5 mM IPTG and 50 μg/mL kanamycin alone at 30°C for over 14 h, and then the pCas plasmid was finally removed by culturing the strains at 42°C overnight in LB medium.

Aspirin detection

The levels of aspirin from plasma and intestinal contents were determined by ultra-high-performance liquid chromatography-quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS) as described¹⁴. The detection sample was mixed in 5 mL 4-Cl-phenylalanine (0.39 mg/mL, internal standard). Aspirin was extracted twice by 400 mL cold methanol. Supernatants were dried under nitrogen stream and detected by UPLC (1290 Infinity LC; Agilent, Santa Clara, CA) and MS (6550 QTOF-MS; Agilent).

Chiu’s score

Scores were independently assessed by two independent pathologists and the standards were described in Figure S2A. The mean scores were calculated based on average values of 10 randomly picked visions from one slice of intestinal tissues and all the individuals in each group were analyzed statistically.

EdU staining

To measure the proliferation of intestinal epithelial cells, 5-ethynyl-20-deoxyuridine (EdU, MCE # HY-118411) at 100 mg/kg was intraperitoneally injected into mice 12 h prior to sacrifice. Intestines were collected and EdU positive cells were detected by EdU staining. EDU at 20 uM incubated for 2 h with organoids and cell lines before harvesting. Organoids and cell lines were collected and EdU positive cells were detected by Edu staining.

Immunohistochemistry (IHC)

For IHC, the target molecules were performed on tissue using Ki67 antibody (#ab15580, Abcam, Shanghai, China). The slides were incubated with HRP-conjugated secondary antibodies (Dako). The proteins were visualized in situ with DAB chromogenic substrate (Dako), followed by counterstaining with hematoxylin as described.^38,67 The results of IHC were independently scored by two independent observers. The Ki-67 positive area was evaluated by ImageJ software in 10 random visons of one slice and mean area were calculated and statistically analyzed.

Immunofluorescence

Cells were plated onto glass coverslips and fixed with 4% paraformaldehyde for 20 min and permeabilized with 0.1% Triton X-100 in PBS for 15 min. Organoids were coated by optimal cutting temperature compound and proceeded to frozen section as well as Formaldehyde Fixed Paraffin Embedded section of fresh intestinal tissues. The following antibodies are used: Blocking solution was applied for 1 h at room temperature. Primary antibodies were applied at 4 °C overnight. FITC or Cy5-conjugated secondary antibodies were loaded and incubated for 2 h at room temperature. Immunostaining signals and DAPI-stained nuclei were visualized at room temperature using a confocal microscope (FV10i; Olympus) equipped with a 10×/0.30 NA objective lens (Olympus) and Fluoview software (version 4.3; Olympus). Quantitation of beta-catenin distribution was evaluated by blinded observers on a per-slide basis, in 100 cells of each condition. For better visualization, the images were adjusted using the levels and brightness/contrast tools in Photoshop according to the guidelines for the presentation of digital data as described.³⁸

Measurement of oxygen concentration

The oxygen concentration was measured with a potable fiber optic oxygen meter (Microx 4 trace; PreSens) that was calibrated according to the manufacturer’s instruction. Measurements were performed at a consistent position within the cell inserts and a bottom well of plates. When measuring the oxygen concentration in the co-culture system, the culture plate was transferred into the anaerobic chamber, the plastic sheets detached, and the oxygen meter applied as described previously.

Co-Immunoprecipitation (Co-IP)

Immunoprecipitation was done with the Pierce Classic IP Kit (#26146, Pierce Biotechnology, US) according to the manufacturer’s protocol as described.⁶⁷ Cells were lysed with IP Lysis/Wash Buffer for 5 min, the lysate was centrifuged at 13000 g for 10 min, and then the supernatant fraction was precleared using Control Agarose Resin for 1 h, adding 10 μg primary antibodies and 40 μl Protein A/G Agarose Resin to 600 μl precleared lysate, and incubated overnight at 4°C. Normal IgG (10 μg) was used as a control antibody. The resin was washed twice with 200 μl Lysis/Wash Buffer and once with 100 μl 1x conditioning buffer. A volume of 50 μl 2x sample buffer with 20 mM DDT was added and incubated at 100 °C for 10 min. It was then centrifuged at 1000 × g for 3 min to collect the eluate. For Co-IP, the IP eluate was examined by western blot with another primary antibody including anti-FXR (#25055-1-AP, Proteintech, Wuhan Sanying, China) and anti-β-catenin (#8480, Cell Signaling Technology, Shanghai, China).

Periodic acid schiff stain (PAS)

Paraffin-embedded intestinal sections were dewaxed in xylene and rehydrated through gradient alcohols. After acidification with periodate, the sections were washed with pure water with chaffer to avoid light staining, then the sections were dehydrated and sealed. Finally, the digital sections were prepared using a slice scanner (Pannoramic MIDI).

Intestinal permeability detection with fluorescein-isothiocyanate (FITC)-dextran

Intestinal permeability was assessed in vivo following oral administration of fluorescein-isothiocyanate (FITC)-dextran (4kDa; #46944Merck, Darmstadt, Germany), a high molecular weight glucose polymer that is neither digested nor absorbed by healthy mice. The animals were given FITC-dextran (0.4 mg/g) orally 4 h before blood collection. Whole blood was obtained through the capillary vein of the eyeball. Serum was diluted with phosphate buffered saline (PBS) and fluorescence was monitored using a multipurpose microplate reader. Excited at 485 nm, emitted at 528 nm. Standard curves were prepared by dilution of FITC-dextran in PBS. Serum from mice that did not receive FITC-dextran was used as background control.

Serum lipopolysaccharide (LPS) detection

Serum lipopolysaccharide was quantified using a ToxinSensor Chromogenic LAL Endotoxin Assay kit (#L00350 GenScript, Nanjing, China) according to the manufacturer’s instructions. Briefly, serum samples were diluted 1:10 to in LAL Reagent Water, four standards (0.01 EU/ml-0.1 EU/ml) and blank controls without samples were prepared, then processed according to the manufacturer’s protocol. Finally mixed well with the liquid in the test tube and the absorbance of 545nm was determined.

Serum diamine oxidase (DAO) detection

Serum DAO was quantified using a Mouse Diamine Oxidase ELISA Kit (#MU30134, Bio-Swamp, Wuhan, China) according to the manufacturer’s instructions. Serum samples were added to wells and combined with DAO antibody. After washing, TMB substrate solution were added. The reaction was terminated by the addition of a sulphuric acid solution and the color change measured spectrophotometrically at a wavelength of 450 nm. The concentration of DAO in the samples was then determined by comparing the OD of the samples to the standard curve.

Fecal occult blood test (FOBT)

Fecal occult blood test was used to check for hidden blood in the feces. Fresh stool (20 mg) was mixed with normal saline and tested with the inside dipstick of the kit-One step Fecal Occult Blood Test (Suzhou Abogen Biosciences, China) following the instruction of manufacturer.

TdT-mediated dUTP nick-end labeling (TUNEL) assay

Tissue and Organoids cell apoptosis was detected with an in-situ cell death detection kit (#11684817910, Roche, Swiss) following the manufacture’s instruction. Briefly, paraffin-embedded intestinal sections were dewaxed in xylene and rehydrated through gradient alcohols. The slides were then incubated with 0.1% Triton X-100 in 0.1% sodium citrate buffer for 15 min followed by PBS washing. Next, the samples were incubated with the TUNEL reaction mixture in a humidified chamber at room temperature for 60 min avoid light. This was followed by PBS washes and incubation with 1 mg/mL DAPI (#28718-90-3, Merck, Darmstadt, Germany) for 10 min. After staining, the sections were scanned by a fluorescent section scanner.

Transmission electron microscopy (TEM)

Fresh intestinal tissues of distal jejunum derived from mice of indicated groups were collected and cut into 3×3 mm² pieces and fixed in fixative reagent (#G1102, Servicebio, Wuhan, China) at 4 °C. After being embedded in resin, ultrathin sections (70 nm) were cut and stained. The structures of intestinal sections were analyzed using a Hitachi HT-HT7700 TEM.

RNA extraction and PCR

Total RNA for cultured cells or organoids was extracted by TRizol reagent (Invitrogen, CA, USA). The total RNA of fresh frozen tissues was extracted using a RNeasy Mini Kit (#74104, Qiagen, Hilden, Germany). The PCR primers are listed in Supplementary Table S4. Reverse transcription (RT) PCR was performed using the PrimeScript RT reagent kit (TaKaRa, Dalian, China) following the manufacturer’s instruction. Quantitative RT PCR was performed using SYBR Premix Ex Taq II (TaKaRa) and measured in a LightCycler 480 system (Roche, Basel, Switzerland) as described.³⁸ Gapdh mRNA or 16S RNA were used as the internal controls. 2^−△△CT referred to the fold-change of the RNA expression of one sample when compared to the calibration sample.

TR-FRET FXR coactivator recruitment assay

CDCA, 7-keto-LCA, and UDCA were tested for direct FXR activation or repression using the commercial TR-FRET FXR coactivator recruitment assay kit (Thermo Fisher, LanthaScreen^™, Cat# PV4833).

Luciferase reporter gene assay

Luciferase reporter plasmids for Shp promoter were constructed by inserting the promoter sequences of Shp gene (−2000 bp) into the pGL3-promoter vectors. HT-29 cells were co-transfected with appropriate plasmids in 24-well plates. After 24 h post-transfection, the cells were exposed to UDCA (50–200 μM, MCE #HY-13771) or 7-keto-LCA (50–200 μM, MCE #HY-W018512) in the presence of CDCA (20 μM, 10 μM MCE #HY-76847) for 24 h. Then the cells were harvested and lysed for luciferase assays. Luciferase activity was measured using a Dual-Luciferase Reporter Assay System (Promega, WI, USA) according to the manufacturer’s protocol. Firefly luciferase activity was normalized to Renilla luciferase used as an internal control. The transfection experiments were performed in triplicate for each plasmid construct as described.³⁸

RNA sequencing and GSEA

Total RNA of organoids was extracted as described and RNA quality was confirmed using the Agilent 2100 Bioanalyzer. RNA-seq libraries were prepared from three biological replicates for each experimental condition and sequenced on an Illumina HiSeq 2500, 4000, or NextSeq500 using barcoded multiplexing and a 100-bp read length. Image analysis and base calling were done with Illumina CASAVA-1.8.2. The quality of the reads was assessed with fastqc. Reads were mapped against the reference genome and transcript annotation (GRCm38.p6) using STAR as described.²⁸ For Figure 7E, the mRNAs were filtered with minimum 10 reads and fold changes were calculated from treatments (n = 3). Top 30 downregulated genes from the LGR5-ASCL2 signature in CDCA group were shown in the heatmap. For Figure S10, differentially expressed genes of Wnt/β-catenin signature (fold change>1.5) in CDCA group compared with DMSO group were shown. Row z-scores (n = 3) were calculated from the matrix of normalized expression using R. For GSEA, normalized expression of gene matrix from RSEM results was used with downloaded gene signatures (https://www.gsea-msigdb.org/gsea/index.jsp). GSEA was performed with the default setting (GSEA 4.2.3 software).⁶⁸

Flow apoptosis assay

Cells were seeded in 6-well plates at 2 × 10⁵ per well and harvested using trypsin 24h after exposure to indicated bile acids. The Annexin V-FITC Apoptosis Detection Kit (BD Biosciences) was used for apoptosis assays as described before.⁶⁷ Cells (1×10⁴) were starved in serum-free medium for 24 h, stained according to the manufacturer’s protocol and sorted using a fluorescence-activated cell sorting sorter (BD), and the data were analyzed using the Modfit software (BD).

CCK8 assays

For HT-29 cell viability assays, gradient concentrations of UDCA or 7-keto-LCA (5 uM,10 uM,20 uM) were administered along with CDCA (10 uM) at 72 h. Cells were seeded in 96-well plates (2×10³/well) and treated as indicated conditions. CCK8 was then added to serum-free medium, and the absorbance was detected by Varioskan Flash Multimode Reader (Thermo-Fisher, Waltham, MA USA) at 450nm with a reference wavelength at 650 nm.

Untargeted quantification of metabolites

The concentration of aspirin in P. goldsteinii inhibitory assays were determined as follows: The intestinal concentration of aspirin in this model is ranging from 100–400 mg/L as Figure S2G showed. If converted to molar concentrations, the concentration of aspirin in the gut will be range from 0.56 to 2.22 mM. Thus, to test the inhibitory effects of aspirin on P. goldsteinii, we used different concentrations of aspirin in the in vitro model, including 0.1, 0.5, 1, 2, 4, 8 and 16 mM. We only show the results of P. goldsteinii growth using aspirin concentration of 1, 2 and 4 mM, which are similar to the actual gut concentration of aspirin in mice. For untargeted metabolites detection in Figure 3A–C and S3, high-resolution LC-MS analysis was performed with a Thermo Fisher Scientific Vanquish Horizon UHPLC System coupled with a Thermo Q Exactive HF hybrid quadrupole-orbitrap high-resolution mass spectrometer equipped with a heated electrospray ionization (HESI) ion source as described.⁶² The medium of each group was extracted and injected (n=3). Each sample was analyzed in positive and negative modes with an m/z range of 150–800. Heatmap of different metabolites were generated by ClustVis online tool (https://biit.cs.ut.ee/clustvis/). Metabolite sets enrichment and network were established by MetaboAnalyt 5.0 (https://www.metaboanalyst.ca/MetaboAnalyst/ModuleView.xhtml).

Whole genome sequencing of bacteria

Whole genome sequencing of bacteria was described as before.³³ The extracted genomic DNA of Parabacteroides was sheared to yield DNA fragments. The genome sequences were determined by the whole-genome shotgun strategy using PacBio Sequel and Illumina MiSeq sequencers. The library of the PacBio Sequel sequencing was prepared using the SMRTbell template prep kit 2.0 (target length = 10–15 kb) without DNA shearing. After the removal of the internal control and adaptor trimming by Sequel, error correction of the trimmed reads was performed using Canu v.1.8. De novo hybrid assembly of the filter-passed MiSeq reads and the corrected Sequel reads were performed using Unicycler v.0.4.8, which contained checks for overlapping and circularization to generate circular contigs. The gene prediction and annotation of the generated contigs were performed using the Rapid Annotations based on Subsystem Technology (RAST) server and Prokka software tool.

QUANTIFICATION AND STATISTICAL ANALYSIS

To test significant differences in metagenomic features (such as species and genus), the two-tailed Wilcoxon rank-sum test adjusted by the Benjamini–Hochberg procedure to control the false discovery rate (FDR) < 5% was used. For metabolomics data, the two-tailed Wilcoxon rank-sum test was performed. For categorical variables (such as smoking and gender), the Chi-square test or Fisher’s exact test was performed. In animal and cell experiments (group number > 2), the data distribution was determined using the Shapiro–Wilk normality test and variance homogeneity was evaluated using Levene’s equal variance test. The sample distribution was determined using a Kolmogorov–Smirnov normality test. For data that passed the Kolmogorov–Smirnov normality test and Levene’s equal variance test, significance was calculated by one-way ANOVA followed by Fisher’s least significant difference (LSD) post-hoc test. For data that passed the Shapiro–Wilk normality test but did not show heterogeneity of variance, significance was calculated by Welch’s ANOVA followed by Games-Howell post-hoc test. For non-normally distributed data, significance was calculated by the nonparametric Kruskal–Wallis test followed by the pairwise Wilcoxon rank sum post-hoc test. In animal and cell experiments (group number = 2), data that passed the Shapiro–Wilk test and Levene’s test were compared for differences using the two-tailed Student’s t-test; data that passed the Shapiro–Wilk test but did not show heterogeneity of variance were compared for differences using Welch’s t-test; data that failed to pass the Kolmogorov–Smirnov test were compared for differences using the two-tailed Wilcoxon rank-sum test. GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA) was used for statistical treatment. The sample sizes were determined by power analysis using StatMate version 2.0 (GraphPad Software, San Diego, CA) and based on animal litter numbers and controlling for sex and age within experiments. Experimental data were shown as the Mean ± SD or otherwise indicated. No data were excluded from data analysis. Two-tailed unpaired Student’s t-test and one-way ANOVA with Tukey’s correction were used for all comparisons of mice-related experiments. Mann Whitney’s u-test were used for comparisons of values that didn’t coincide with normal distribution. P values < 0.05 were considered significant. Correlation analysis of gut microbiome and host metabolome were investigated using nonparametric Spearman’s test. No data were excluded from the analyses. For all animal experiments, littermates were grouped by a researcher unaware of the experimental design. Grouped littermates were assigned to treatment arms randomly before starting experiments.

Highlights.

• Aspirin users have marked gut microbiota dysbiosis, contributing to intestinal injury

• P. goldsteinii, depleted by aspirin, alleviates aspirin-related gut barrier disruption

• P. goldsteinii and its metabolite 7-keto-LCA maintain intestinal homeostasis

• 7-keto-LCA, an FXR antagonist, facilitates Wnt signaling and stem cell proliferation