Parabacteroides distasonis ameliorates hepatic fibrosis potentially via modulating intestinal bile acid metabolism and hepatocyte pyroptosis in male mice

1Hepatic fibrosis is a dynamic process characterized by excessive accumulation of extracellular matrix resulting from chronic liver injury. If left untreated, hepatic fibrosis ultimately leads to cirrhosis, liver failure, or hepatocellular carcinoma. The incidence of hepatic fibrosis is increasing worldwide, and there are no drugs available to protect against or decrease hepatic fibrosis. The gut microbiome, the secondary largest genome of the host, plays an important role in chronic hepatic injury. Various probiotics, such as Lactobacillus rhamnosus GG, Saccharomyces boulardii, and the selenium-enriched probiotics, Akkermansia muciniphila, Bifidobacterium and Bacteroides acidifaciens, can improve hepatic fibrosis, alcoholic and nonalcoholic fatty liver in mice and rats–. Notably, it was demonstrated that Parabacteroides distasonis (P. distasonis) improved obesity, inflammatory bowel disease, colorectal cancer, and testicular dysfunction, and prevented post-calorie restriction weight gain through producing or stimulating host production of various active metabolites such as 3-oxo lithocholic acid (3-oxoLCA), 3-oxo-Δ4-LCA, succinate, spermine, agmatine, indolelactic acid, melatonin, uracil and urocanic acid–. Thus, P. distasonis may play an important role in human health.

2Bacteroides was found to have bile salt hydrolase (BSH) activity that deconjugates primary bile acid derivatives produced in the liver. This transformation of conjugated bile acids to unconjugated derivatives that can then be subjected to further bacterial modifications, is essential for bile acid homeostasis. BSH is negatively correlated with nonalcoholic fatty liver disease (NAFLD) and liver cirrhosis clinically. BSH showed various activity in the host such as lipid-lowering, improving Clostridioides difficile infection, and enhancing adhesion ability in the gastrointestinal tract–. Uncovering the mechanism by which BSH influences hepatic fibrosis would be beneficial for the developing BSH-based therapeutics.

3Ileal farnesoid X receptor (FXR) controls bile acid transport across the intestinal epithelial cells into the blood where systemic bile acid metabolites influence inflammation, glucose homeostasis, obesity, insulin resistance, and NAFLD. In the gut, certain bile acid metabolites have antibacterial activity involved in the pathogenesis of inflammatory bowel disease and mucosal injury. Intestinal FXR antagonists such as ursodeoxycholic acid (UDCA), glycine-β-muricholic acid (GβMCA), and glycoursodeoxycholic acid (GUDCA) were suggested in mouse models to be of potential value for the treatment of primary biliary cirrhosis (PBC), NAFLD, obesity and atherosclerosis through regulating the FXR/SMPD3 axis and reducing the biosynthesis of intestine-derived ceramides–. GUDCA and a betulinic acid derivative, two intestinal FXR antagonists, improved methionine and choline-deficient (MCD) diet-induced nonalcoholic steatohepatitis and hepatic fibrosis. In addition, the probiotic Lactobacillus rhamnosus GG, reduced hepatic fibrosis by inhibiting intestinal FXR signaling and decreasing FGF15 expression, and promoting the hepatic synthesis and excretion of bile acids, thus indicating that inhibition of intestinal FXR signaling may be of value as a strategy to alleviate hepatic fibrosis.

4The purpose of the current study was to examine the protective role of P. distasonis in hepatic fibrosis. P. distasonis decreased taurochenodesoxycholic acid (TCDCA) levels by increasing BSH activity and inhibiting ileal FXR. The decrease of TCDCA improved activation of hepatic stellate cells (HSCs) through decreasing the mitochondrial permeability transition (MPT)-Caspase-11 pyroptosis pathway. Celastrol derived from the root of Tripterygium wilfordii plant promotes the growth of P. distasonis and prevents thioacetamide (TAA)- and MCD diet-induced hepatic fibrosis through increasing BSH activity and inhibiting ileal FXR signaling. Administration of P. distasonis leads to increased BSH activity, inhibition of intestinal farnesoid X receptor (FXR) signaling, and decreased TCDCA levels in liver. These findings provide a therapeutic strategy for hepatic fibrosis through supplementation with P. distasonis.

5Data mining from the BIG Data Center (CRA001920) found that clinical cholestasis was associated with decreased Bacteroidetes (phylum) and Parabacteroides (genus) levels (Fig. 1a, b). P. distasonis was decreased in clinical cholestasis patients (Fig. 1c, d), clinical hepatic fibrosis patients (Fig. 1e), and TAA-induced hepatic fibrosis in mice (Supplementary Fig. 1a). Receiver operating characteristic (ROC) analysis revealed that P. distasonis had high discriminatory power to clinically identify hepatic fibrosis (Fig. 1f). P. distasonis was negatively correlated with the hepatic fibrosis indexes including alkaline phosphatase (ALP), total bile acid (TBA), Col1a1, tissue inhibitor of metalloproteinase-1 (Timp1), smooth muscle actin 2 (Acta2), transforming growth factor β (Tgfb) mRNA expression and severity degree (Child and MELD scores) in clinical samples and in mice (Supplementary Fig. 1b, c, Supplementary Fig. 2d). These results showed that P. distasonis may play an important role in the development of hepatic fibrosis.

6Metabolomics analysis revealed that conjugated bile acids in serum and feces (e.g., TCDCA and glycochendeoxycholic acid (GCDCA)) were increased and unconjugated bile acids in feces (e.g., chenodeoxycholic acid (CDCA)) were decreased in hepatic fibrosis patients, thus indicating lower BSH activity (Fig. 1h, i, Supplementary Fig. 1d). Feces BSH levels were decreased, and unconjugated bile acid/conjugated bile acid ratios in serum and feces were decreased in clinical hepatic fibrosis patients (Fig. 1k–m). Lactobacillus (Genus), Bifidobacterium (Genus), Clostridium (Genus), Bacteroides (Genus), Parabacteroides (Genus) potential express BSH. Among all these bacteria, only P. distasonis levels were decreased in clinical cholestasis patients and in the TAA-induced hepatic fibrosis mouse model, implying that P. distasonis may have BSH activity (Supplementary Fig. 2e, f). P. distasonis levels were positively correlated with BSH activity in clinical feces samples (Fig. 1n).

7In order to determine the role of gut microbiota on hepatic fibrosis, bacteria in the intestinal tract were depleted by administration of an antibiotic cocktail (ampicillin, 0.25 mg/mL; neomycin, 0.25 mg/mL; metronidazole, 0.25 mg/mL; vancomycin, 0.125 mg/mL) and then the gut microbiota was reconstructed by fecal microbial transplantation (FMT) from healthy mice (Supplementary Fig. 3). Loss of gut microbiota increased TAA-induced hepatic fibrosis as revealed by histology, increased bile acids in serum and liver, decreased body weights, increased aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activities, and increased expression of proinflammatory factor and hepatic fibrosis gene mRNAs (interleukin-1 (Il1), Il6, tumor necrosis factor α (Tnfa) and Timp1) (Supplementary Fig. 3a–f). These pathological features including histology, AST activity, proinflammatory factor, and hepatic fibrosis gene expression (Il1, Il6, and Tgfb) could be improved following FMT from healthy mice to antibiotic cocktail-treated diseased mice (Supplementary Fig. 3g–k). These results show that gut microbiota play an important role in TAA-induced hepatic fibrosis.

8To evaluate the effect of P. distasonis in TAA-induced hepatic fibrosis, P. distasonis transplantation was carried out (Fig. 2a). P. distasonis (species) and Bacteroidetes (phylum) in cecum lumen were significantly increased after transplantation (Fig. 2b, Supplementary Fig. 4c). This resulted in improved histology and decreased ALT, proinflammatory factor, and hepatic fibrosis gene mRNAs (Il1, Il6, Timp1 and Tgfb) and proteins expression (IL1β, IL6, α-smooth muscle actin (αSMA), COL1A1, TGFβ and TIMP1) revealing that P. distasonis could improve TAA-induced hepatic fibrosis (Fig. 2c–g, Supplementary Fig. 4d). Non-targeted serum metabolomics further revealed that levels of 31 metabolites were altered by P. distasonis, including 3 bile acids, 7 acylcarnitines, and 5 lipids (Supplementary Fig. 4e, f). Targeted analysis of bile acids in serum, liver, duodenum, jejunum, ileum, and cecum content was then conducted. P. distasonis decreased bile acids in serum and liver and increased bile acids in ileum and cecum content, indicating that P. distasonis promoted the excretion of bile acids in enterohepatic circulation (Fig. 2h and Supplementary Fig. 4g). Both bile acids and acylcarnitines were positively correlated with liver injury indexes (e.g., ALT, Timp1 and Tgfb mRNAs) (Supplementary Fig. 5b). P. distasonis decreased bile acids in serum and liver in healthy mice (Supplementary Fig. 4g) but did not influence acylcarnitine levels (Supplementary Fig. 5c, d). Hepatic bile acid synthesis and transporter gene mRNAs (Fxr, organic anion transporting polypeptide 4 (Oatp4), and bile salt export pump (Bsep)) that were decreased by TAA, were reversed by P. distasonis treatment (Supplementary Fig. 4h). These results showed that P. distasonis promoted the excretion of bile acids and improved hepatic fibrosis.

9Since P. distasonis was positively correlated with BSH activity in hepatic fibrosis patients (Fig. 1n), we predicted that P. disasonis possessed BSH activity. As expected, P. distasonis decreased TCDCA levels and increased CDCA/TCDCA ratios by transforming TCDCA to CDCA in culture medium under anaerobic conditions (Fig. 3a). TCDCA was not decreased further with the highest P. distasonis employed in Fig. 3a, likely due to growth inhibition by high levels of P. distasonis. P. distasonis protein, from living P. distasonis in the culture medium, showed BSH activity and decreased TCDCA levels, as revealed by in vitro incubations (Fig. 3b). Protein extraction and incubations were carried out based on a previous study. Cecum content BSH activity and unconjugated/conjugated bile acid level were increased after P. distasonis treatment in mice, indicating that P. distasonis increased BSH activity (Fig. 3c).

10Ileal FXR participated in the reabsorption of bile acids. P. distasonis inhibited ileal FXR target gene expression (Fxr, ileal bile acid binding protein (Ibabp) and organic solute transporter α (Osta)), and proteins expression (FXR and OSTβ) (Fig. 3d, e, Supplementary Fig. 4k, l). Hepatic Fxr mRNA and FXR target gene mRNAs (Shp, and Bsep) were not decreased by P. distasonis (Supplementary Fig. 5a). Dual-luciferase reporter gene assays were performed with HEK293 and Caco-2 cells co-transfected with an FXR expression vector and tk-palinodromic ecdysteroid response element (EcRE)-luciferase expression plasmid, demonstrating that the FXR agonist CDCA significantly increased FXR signaling and P. distasonis and the FXR antagonist tauro-β-muricholic acid (TβMCA) significantly decreased FXR signaling (Fig. 3f). P. distasonis decreased Fxr mRNA and the FXR target gene mRNAs Fgf19 and Ostb, in Caco-2 cells after a 24 h exposure (Fig. 3g). These results showed that P. distasonis inhibited ileal FXR directly. GUDCA, a natural antagonist of ileal FXR, could decrease TAA-induced hepatic fibrosis and decrease the protective effects of P. distasonis as revealed by H&E and Masson trichrome staining, and AST and ALT activities (Fig. 3h, i). GUDCA treatment decreased serum and hepatic bile acids and promoted the excretion of bile acid in ileum and cecum contents, consistent with the effect of P. distasonis (Fig. 3j). These results showed that ileal FXR participated in the protective effects of P. distasonis, and inhibition of ileal FXR signaling may promote the excretion of bile acids and improve hepatic fibrosis. Other P. distasonis activities such as the production of active metabolites, including indolelactic acid, trimethylamine, cadaverine, lactic acid, and 2-hydeoxybutyric acid were also observed by culture medium metabolomics analysis (Supplementary Fig. 6).

11Serum and hepatic TCDCA is important in hepatic fibrosis. In serum of hepatic fibrosis patients, TCDCA and GCDCA were increased 494-fold and 24-fold, respectively (Fig. 1i, j). However, in the TAA-induced fibrosis mouse model, TCDCA was the only bile acid that after increasing by TAA, was decreased by treatment with P. distasonis and GUDCA and potentiated by antibiotics in serum and liver (Fig. 2h, 3j, Supplementary Fig. 3b). Serum TCDCA was positively correlated with hepatic fibrosis indexes in patients and in mice (Supplementary Fig.1b, c). Decreased TCDCA in serum and liver after P. distasonis treatment resulted from increased P. distasonis BSH activity that converts TCDCA to CDCA. Additionally, inhibition of ileal FXR signaling would decrease the absorption of TCDCA from the intestine (Fig. 3j).

12TCDCA induced toxicity in mouse primary hepatic cells while the deconjugated metabolite CDCA had no effect on these cells (Fig. 4a, b). TCDCA potentiated TAA-induced liver injury in mice as shown by histological analysis, AST levels, proinflammatory factors and hepatic fibrosis mRNA (Il1, Il6, and Timp1) levels, and serum metabolomics (Fig. 4c, d, Supplementary Fig. 7b–d). Furthermore, TCDCA potentiated the toxicity of TAA and deoxycholic acid (DCA) in mouse primary hepatic cells (Supplementary Fig. 8). These results showed that TCDCA contributed to hepatic fibrosis.

13Bile acids could induce MPT-Caspase-11 pyroptosis under disease conditions, and therefore we predicted TCDCA-induced hepatic fibrosis through MPT-Caspase-11 pyroptosis. The loss of mitochondrial cristae was revealed by transmission electron microscopy (TEM), altered mitochondrial respiratory chain gene mRNA expression, and the dissipation of mitochondrial transmembrane potential (Δψm) revealing that TCDCA induced MPT in mouse primary hepatic cells (Fig. 4e–g). TCDCA induced the Caspase-11 pyroptosis pathway (apoptosis protease activating factor 1 (Apaf-1)-Caspase-11-Caspase-3-gasdermin E (GSDME)) in mouse primary hepatic cells and in TAA-induced liver injury (Fig. 4h, i, Supplementary Fig. 8k, l). However, the Caspase-1 pyroptosis pathway remained unchanged (NOD-like receptor family pyrin domain containing 3 (NLRP3)-Caspase-1-gasdermin D (GSDMD)/IL1β) (Fig. 4h, i). P. distasonis and FMT improved Caspase-11 pyroptosis in TAA-induced hepatic fibrosis mice (Fig. 2g, Supplementary Fig. 3l). These results showed TCDCA-induced MPT and Caspase-11 pyroptosis. GCDCA, another bile acid increased in clinical serum and feces samples, could also induce MPT and Caspase-11 pyroptosis (Fig. 1i, Supplementary Fig. 9a–e). Apoptosis and ferroptosis played unimportant roles in the toxicity of TCDCA and GCDCA (Supplementary Figs. 9f, 10).

14Hepatocytes, HSCs, and Kupffer cells were treated with TCDCA (6.25–100 μM) to evaluate the role of Caspase-11 pyroptosis in different cells. TCDCA-induced Caspase-11 pyroptosis in mouse primary hepatocytes after 6 h and 24 h TCDCA treatment, as shown by increased Caspase-11 and Caspase-3 mRNA expression, induced balloon-like bubbles as indicated by microscopic examination and multiple pores formed in the membranes as revealed by TEM (Figs. 4e, 5b–d),. Furthermore, TCDCA could induce Caspase-11 pyroptosis in mouse primary HSC and human HSC (LX2 cell) 48 h after TCDCA treatment as revealed by increased Caspase-11 (Caspase-4 in humans) and Caspase-3 mRNA expression and balloon-like bubbles (Fig. 5b, e, f). Caspase-11 pyroptosis was not increased by 6.25–100 μM TCDCA in mice and human primary Kupffer cells 24 h after TCDCA treatment (Supplementary Fig. 11a). These results showed that Caspase-11 pyroptosis was induced by TCDCA in hepatocytes and HSCs rather than in Kupffer cells.

15To evaluate the effect of MPT and Caspase-11 pyroptosis in TCDCA-induced liver injury, the MPT agonist lonidamine and antagonist bongkrekic acid (BKA), and Caspase-11 inhibitor wedelolactone were tested in mouse primary hepatocytes (Fig. 5g–m). Lonidamine potentiated TCDCA-induced Caspase-11 pyroptosis, while BKA improved TCDCA-induced Caspase-11 pyroptosis in mouse primary hepatocytes (Fig. 5g–j). The increased cell viability, the decreased malondialdehyde (MDA), and Caspase-11 and Caspase-3 mRNA expression showed that the Caspase-11 inhibitor wedelolactone improved TCDCA-induced Caspase-11 pyroptosis in mouse primary hepatocytes (Fig. 5k–m). These results showed that MPT and Caspase-11 pyroptosis participated in TCDCA-induced liver injury.

16To evaluate Caspase-11 pyroptosis-activated HSC, LX2 cells were treated with Caspase-4 protein. Caspase-4 protein was internalized by LX2 cells, and increased hepatic fibrosis and pyroptosis gene mRNA and protein expression (Fig. 5n–p). A previous study found that Caspase-1 protein is internalized by LX2 cells. Therefore, TCDCA-induced MPT-Caspase-11 pyroptosis in hepatocytes and HSC, and Caspase-11 pyroptosis activated HSCs (Fig. 5a). On the other hand, TCDCA activated HSC by stimulating the secretion of IL1β, TNFα and TGFβ protein by Kupffer cells, as shown by increased Timp1 and Acta2 mRNA expression in HSC after the Kupffer cell supernatant treatment (Supplementary Fig. 11b–e).

17A previous study observed that celastrol from the roots of Tripterygium Wilfordii plant increased P. distasonis in high-fat diet-induced obese mice, suggesting that celastrol may promote the growth of P. distasonis. In vivo analysis found that celastrol increased Bacteroidetes (phylum), Bacteroidaceae (family), and P. distasonis (species) in the TAA-induced hepatic fibrosis model (Fig. 6a–c, Supplementary Fig. 12a–e). Celastrol could also increase the level of P. distasonis in healthy mice (Fig. 6d). In vitro analysis found that celastrol from 48.8 nM to 12.5 μM effectively promoted the growth of P. distasonis (Fig. 6f). Quantitative and qualitative experiments revealed that celastrol promoted biofilm formation (Fig. 6g, h). Culture medium metabolomics profiling found that various gut microbiota metabolites were changed in P. distasonis following celastrol treatment (Supplementary Fig. 12f, g), and the altered pathways included phenylalanine, tyrosine, and tryptophan biosynthesis, phenylalanine metabolism, and tryptophan metabolism (Supplementary Fig. 12h). These results showed that celastrol could promote the growth of P. distasonis.

18Celastrol could protect against TAA-induced hepatic fibrosis, including histology, decrease AST and ALT activities, proinflammatory factor expression and hepatic fibrosis gene and protein expression, and serum and liver metabolites (Supplementary Figs. 13a–f, 14e). Celastrol decreased bile acid levels in serum and liver, especially TCDCA, and increased bile acid levels in duodenum, jejunum, ileum, cecum content, urine, and feces, indicating that celastrol promoted the excretion of bile acids (Fig. 6i, Supplementary Figs. 13j, 14e, 15). Consistent with the effect of P. distasonis, celastrol increased BSH activity, inhibited ileal FXR, and attenuated Caspase-11 pyroptosis in mice (Fig. 6j–l, Supplementary Figs. 14g, 15i). The data also demonstrated that ileal FXR plays an important role in the protective effects of celastrol on hepatic fibrosis as revealed using Fxr-null mice and the intestinal FXR inhibitor GUDCA (Supplementary Fig. 16). After pretreatment with antibiotics, the protective effects of celastrol were decreased (Supplementary Fig. 3m–o), and FMT from 14-day celastrol treated mice could improve TAA-induced hepatic fibrosis (Supplementary Fig. 3g–l) which showed that the gut microbiota has an important role in the protective effects of celastrol. These results revealed that celastrol regulates intestinal bile acid metabolism and protects against hepatic fibrosis.

19P. distasonis treatment increased P. distasonis levels in MCD diet-induced steatohepatitis and hepatic fibrosis (Supplementary Fig. 17a). H&E, Oil red O, Masson trichrome, immunohistochemistry and immunofluorescence stainings showed that P. distasonis improved MCD diet-induced hepatic fibrosis, and decreased lipid accumulation (Fig. 7b). Furthermore, P. distasonis improved AST and ALT levels, and hepatic fibrosis gene mRNA and protein expression (Fig. 7c–e, g). Targeted metabolomics analysis revealed that bile acids were decreased in serum and increased in ileum and cecum (Fig. 7h, Supplementary Fig. 17b). P. distasonis increased BSH activity, inhibited ileal FXR signaling, and improved Caspase-11 pyroptosis in the MCD diet-induced hepatic fibrosis model (Fig. 7f, g, i–l). Furthermore, celastrol improved MCD diet-induced steatohepatitis and hepatic fibrosis, which increased BSH activity decreased ileal FXR pathway, and alleviated bile acids level in the liver (Supplementary Figs. 18, 19). The protective effects of celastrol were decreased in Fxr-null mice, indicating that ileal FXR may play an important role in the development of hepatic fibrosis (Supplementary Fig. 19i, j).

20The present study provides evidence that P. distasonis contributes to improving hepatic fibrosis through regulating bile acid homeostasis as graphically shown in Fig. 8. The biological function of P. distasonis in regulating the generation of bile acids and the targets of the host were demonstrated. The current study determined the role of gut microbiota, bile acid metabolism, and related host targets in hepatic fibrosis. The most common mouse models to study hepatic fibrosis include the chemical-based models (TAA, CCl4, ethanol), diet-based models (MCD diet, HFD), a surgery-based model, and a genetically modified model. Since a previous study found that celastrol could protect against CCl4- and HFD-induced chronic liver injury, the hepatic fibrosis models induced by TAA and MCD were used in the current study.

21Bile acids, diets, intestinal pH, and gastrointestinal motility likely influence the decrease of P. distasonis in hepatic fibrosis patients. Bile acids can exert both direct effects and indirect effects on the gut microbiota, including bacterial membrane damage, changes in membrane lipid compositions, and targeting nuclear receptors. Furthermore, diets such as a high-fat western diet and alcohol consumption may play a key role in modulation of the gut microbiota composition through influencing intestinal pH that can promote the development of proinflammatory gut microbiota and increase lipopolysaccharides.

22P. distasonis is the anaerobic strain belonging to the genus Parabacteroides, which commonly colonize the gastrointestinal tract of numerous species. It was found that P. distasonis has beneficial and pathogenic effects on human health and was associated with decreased colorectal cancer incidence, obesity, and inflammatory bowel disease and prevent post-calorie restriction weight gain, although it may induce depressive-like behavior in a mouse model of Crohn’s disease,,. A previous study found that P. distasonis improved triptolide-induced testicular dysfunction through increasing microbial-derived polyamines which could upregulate heat shock protein 70. In a previous study, P. distasonis increased unconjugated bile acids, such as UDCA and LCA indicating increased BSH activity in cecum content. Clinical studies reported that NAFLD and neonatal cholestasis disease decreased P. distasonis levels,. In the present study, P. distasonis levels were decreased in hepatic fibrosis patients and supplementation of P. distasonis could improve TAA- and MCD diet-induced liver fibrosis in mice through regulating bile acid metabolism, indicating that P. distasonis plays an important role in hepatic fibrosis. Since research on the protective effects of P. distasonis on liver fibrosis is limited, the present study provides prospects to employ P. distasonis to improve hepatic fibrosis.

23Intestinal FXR and BSH play an important role in human health. VSH#3 probiotics (L. paracasei, L. plantarum, L. acidophilus, L. delbrueckii subspecies L. bulgaricus, B. longum, B. breve, B. infantis, Streptococcus thermophilus), colesevelam and resveratrol promoted the excretion of bile acids and had a beneficial role in metabolic diseases through increasing BSH and inhibiting ileal FXR–. Another study found decreased BSH activity inhibited ileal FXR through increasing intestinal TβMCA levels. In the present study, P. distasonis and celastrol improved hepatic fibrosis by increasing BSH activity and inhibiting ileal FXR. Apart from direct inhibition of P. distasonis on ileal FXR, several bile acids such as UDCA, ωMCA, and βMCA were increased in cecum after P. distasonis and celastrol treatments and were reported to inhibit intestinal FXR signaling in mice,, which may contribute to the inhibition of intestinal FXR signaling.

24It should be noted that the causal role of bacterial BSH, intestinal FXR signaling, and altered bile acid levels in liver fibrosis after P. distasonis administration was not validated in vivo in the current study. However, the role of BSH, conjugated bile acids, and ileal FXR signaling in bile acid homeostasis have been firmly document in earlier studies. Bacterial BSH deconjugates conjugated bile acids secreted from the liver and the BSH inhibitor caffeic acid phenethyl ester increases conjugated bile acids and decreases deconjugated bile acid in the intestine,. Deconjugated bile acids that result from BSH activity are less soluble and their reabsorption by intestinal epithelial cells is decreased thus they are preferentially eliminated in the feces. Previous studies found that inhibition of ileal FXR or ileal FXR knockout mice increase bile acids in the ileum and decrease bile acids in liver and serum, indicating that inhibition of ileal FXR promotes the fecal elimination of bile acids,,. Notably, by inhibition of ileal FXR signaling, the conjugated bile acids TUDCA and GUDCA are markedly increased in healthy mice and the ileal conjugated bile acids TαMCA and TβMCA are increased and hepatic TCA and TDCA levels and serum TDCA and THDCA levels are decreased in HFD-induced mice. In ileal FXR knockout mice, serum TβMCA, TωMCA, TCA, CDCA and hepatic TαMCA, TβMCA, TωMCA, TCDCA, and TUDCA are decreased. After GUDCA treatment in the present study, bile acid levels in ileum and cecum were increased, and bile acid levels in serum and liver were decreased, thus indicating that inhibition of ileal FXR led to increased bile acid excretion. Finally, administration of P. distasonis leads to increased BSH activity, inhibition of intestinal FXR signaling, and decreased TCDCA levels in liver.

25Inhibiting the reabsorption and promoting the excretion of bile acid have become a hotspot in drug discovery. Some drugs were developed for treating liver disease such as UDCA, TUDCA, odevixibat, maralixibat, colestyramine, and colesevelam. Notably, UDCA and TUDCA were used for the treatment of PBC through inhibiting intestinal FXR signaling and promoting the excretion of bile acids. In 2021, odevixibat and maralixibat were approved by the FDA for the treatment of pruritus induced by cholestasis. Odevixibat and maralixibat decrease the reabsorption of bile acid through inhibiting ileal apical sodium-dependent bile acid transporter (ASBT), encoded by a FXR target gene, and promote the excretion of bile acids. The bile acid sequestrants colestyramine and colesevelam promote bile acid excretion through complexing with bile acids in the intestinal lumen, and improving pruritus induced by cholestasis clinically and hepatic fibrosis in mice. Therefore, inhibition of FXR and other bile acid transporters in ileum may promote the elimination of bile acids and improve hepatic fibrosis.

26The mitochondria is a central hub sensing various stimuli and thereby controlling cell fates, including cell proliferation, differentiation, death, and immunological response. MPT is induced by the persistent opening of the MPT pore, which is usually activated by Ca2+ and reactive oxygen species (ROS). Pyroptosis is mediated by inflammatory caspases which is featured with pore formation on plasma membrane, cell lysis, and swelling. Both Caspase-1-induced inflammasome and Caspase-11-induced pyroptosome were found to play key roles in pyroptosis. The Caspase-11 pryoptosis pathway, including Apaf-1-Caspase-11-Caspase-3-GSDME, is a general sensor of MPT. A previous study found DCA and CDCA activated MPT-Caspase-11 pyroptosis. The present study found TCDCA and GCDCA could activate MPT-Caspase-11 pyroptosis. Hepatic fibrosis is accompanied by activation of Caspase-1 and Caspase-11 pyroptosis. Earlier work found that hepatocytes induce Caspase-1 pyroptosis and release NLRP3 inflammasome and IL1β to the supernatant,. Hepatocytes activate HSC through internalizing the NLRP3 inflammasome and IL1β,. The present study found that Caspase-11 pyroptosis occurs in primary hepatocytes and HSC, which activate HSC through releasing the Caspase-11 pyroptosome. These studies suggested that TCDCA may activate HSC through inducing MPT-Caspase-11 pyroptosis.

27In the present study, TCDCA levels were increased in serum and feces of hepatic fibrosis patients. It was also reported that TCDCA and GCDCA were increased in NAFLD liver tissue. TCDCA activates HSC through releasing TNFα and TGFβ by Kupffer cells. Therefore, TCDCA may play an important role in the development of hepatic fibrosis. TCDCA induces liver injury in rats and cell toxicity in vitro. Previous studies found that TCDCA initially binds to the mitochondrial outer membrane and subsequently the structure of the inner boundary membrane disintegrates and MPT is induced. Furthermore, TCDCA contributes to apoptosis through the activation of the caspase cascade, PKC/JNK and nuclear factor-κB (NF-κB) pathways,. The present study demonstrated that TCDCA induces MPT-Caspase-11 pyroptosis, and subsequently Caspase-11 pyroptosis activates HSC. Clinical serum TCDCA levels in chronic liver diseases were 0.7-15.1 μM. The present study found that TCDCA levels in serum were 0.0004-146 μM in the patients with hepatic fibrosis and 246-559 μM in mice with hepatic fibrosis. Notably these levels are higher than the concentration (6.25–100 μM) of TCDCA-induced pyroptosis in hepatic cells in vitro, suggesting that the increased TCDCA could induce pyroptosis during hepatic fibrosis.

28Celastrol could protect acute liver injury or cholestasis induced by acetaminophen, α-naphthyl isothiocyanate, CCl4 and TAA,. Celastrol protects against CCl4-induced hepatic fibrosis through activating AMPK/SIRT3 signaling. Celastrol also protects against renal fibrosis through upregulating the cannabinoid receptor 2, improving cardiac fibrosis through inhibiting microRNA-21/ERK1/2, and improving pulmonary fibrosis through activating autophagy,. Here, celastrol was found to protect against TAA- and MCD diet-induced hepatic fibrosis through increasing P. distasonis in mice. An earlier study found that celastrol reduced intestinal lipid absorption to antagonize obesity by resetting the gut microbiota under HFD feeding. Antibiotics treatment blocked the celastrol-induced lipid-lowing effect and FMT from celastrol-treated obese mice reduced the body weight gain of HFD-fed mice suggesting a critical role for the gut microbiota in mediating the anti-obesity effects of celastrol under HFD. Increased P. distasonis was observed after celastrol and HFD co-treatments,. The present work demonstrated that celastrol increased P. distasonis levels directly in vivo and in vitro and loss of gut microbiota decreased the protective effects of celastrol, indicating that the protective effect of celastrol on liver fibrosis is closely associated with its promotion of P. distsonis growth. Similarly, celastrol could inhibit intestinal FXR and increase the excretion of bile acids, which decreases the levels of hepatic TCDCA and prevents TAA- and MCD diet-induced hepatic fibrosis. The function of P. distasonis may also be influenced by celastrol. Celastrol promots the production of indoleacrylic acid, phenylacetylglutamine, and tryptophan in P. distasonis in vitro, which can be explored by further studies on the prokaryotic transcriptome (Supplementary Fig. 6b). A previous study found that polysaccharide increased the levels of short-chain fatty acids of Lactobacillus faecis by regulating the expression of idh, metE, adh2, mvK in L. faecis as revealed by study of the Lactobacillus faecis transcriptome and proteome. Thus celastrol may also influence related genes in P. distasonis.

29In summary, P. distasonis decreased hepatic TCDCA levels, increased BSH activity, inhibited ileal FXR and improved hepatic fibrosis. The decrease of hepatic TCDCA improved the activation of HSC by decreasing MPT-Caspase-11 pyroptosis in hepatocytes. Taken together, these findings suggest that supplementation with P. distasonis may be a promising means to ameliorate hepatic fibrosis, however, whether patients with hepatic fibrosis might benefit from supplementation with P. distasonis needs to be investigated in further studies.

30Sample 1: Sixty-two patients with hepatic fibrosis and 25 healthy volunteers were recruited to collect serum samples. Sample 2: Seventeen patients with hepatic fibrosis and 10 healthy volunteers were recruited to collect fecal samples. Sample 3: Ten patients with hepatic fibrosis and 10 healthy volunteers were recruited to collect the serum and fecal samples. The patient information is summarized in Supplementary Tables 1–3. All volunteers were evaluated by clinicians during the study ensuring that they remained in good health. Written informed consents were obtained from each participant. The patients and healthy volunteers had no participant compensation. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki. The study protocol was approved by the Conjoint Health Research Ethics Board of the Secondary Affiliated Hospital of Kunming Medical University (Registration number: PJ-2017-25) and the Conjoint Health Research Ethics Board of West China Hospital, Sichuan University (Registration number: ChiCTR2200067222).

31All animal experiment protocols were approved by the Animal Care and Use Committee of West China Hospital, Sichuan University (20220217001). Male 6-week-old Fxr-null mice (C57BL/6 J background) were previously described. Male 6-week-old C57BL/6 J mice were purchased from GemPharmatech Co., Ltd. (Jiangsu, China) and maintained under a standard 12-h light/12-h dark cycle environment with free access to water and rodent chow (Double lion experimental animal feed technology Co., LTD.). Mice were maintained in a temperature 22 ± 1 °C and humidity-controlled room (40–65%).

32To investigate the relationship between P. distasonis and the severity of hepatic fibrosis, mice were randomly assigned into six groups (n = 6): (1) Con-6 week; (2) TAA-6 week; (3) Con-8 week; (4) TAA-8 week; (5) Con-10 week; (6) TAA-10 week. TAA groups were treated with TAA for 6, 8, and 10 weeks (i.p., 200 mg/kg dissolved in physiologic saline solution, three times per week). To determine the role of gut microbiota in TAA-induced hepatic fibrosis, mice were randomly assigned to 3 groups (n = 6): (1) Control; (2) TAA; (3) TAA + Antibiotic. TAA and TAA + Antibiotic groups were treated with TAA for 6 weeks (200 mg/kg). The mixture of antibiotics including ampicillin (0.25 mg/mL), neomycin (0.25 mg/mL), metronidazole (0.25 mg/mL), and vancomycin (0.125 mg/mL) were dissolved in water and supplied ad libitum to eliminate the gut microbiota. After TAA treatment for 1 week, the TAA + Antibiotic group was treated with antibiotics for 5 weeks.

33P. distasonis, obtained from American Type Culture Collection (ATCC8503, USA), was cultured in brain-heart infusion fluid medium (Huankai, China) and tryptic soy agar/broth solid medium (with 5% sheep blood, Solarbio, China). To determine the protective effect of P. distasonis on TAA-induced hepatic fibrosis, the mice were randomly assigned to 5 groups (n = 6): (1) Control; (2) TAA; (3) TAA + P. distasonis (TAA + P.d.); (4) TAA + heat-killed P. distasonis (TAA + P.d.-H); (5) P. distasonis (P.d.). TAA, TAA + P.d., and TAA + P.d.-H groups were treated with TAA for 6 weeks (200 mg/kg). After TAA treatment for 1 week, all groups were treated with antibiotics for 1 week. After antibiotics treatment, the TAA + P.d. and P.d. groups were given P. distasonis by oral transplantation for 4 weeks (i.g., 2×108 CFU dissolved in PBS, once a day); heat-killed P. distasonis was parallelly treated.

34To investigate the protective effects of celastrol on TAA-induced hepatic fibrosis, wild-type, and Fxr-null mice were randomly assigned to 3 groups (n = 6): (1) Control; (2) TAA; (3) TAA + Celastrol. TAA and TAA + Celastrol groups were treated with TAA for 6 weeks (200 mg/kg). After TAA treatment for 1 week, TAA + Celastrol group was orally treated with celastrol for 5 weeks (10 mg/kg dissolved in 1% DMSO + 2% Tween 80 + 97% water, three times per week). Furthermore, to investigate the effect of celastrol on bile acid metabolism in healthy mice, mice were orally treated with celastrol for 1 day (n = 5) and 2 weeks (n = 6), respectively (10 mg/kg, three times per week).

35To validate the function of gut microbiota on the protective effect of celastrol, the mice were randomly assigned to 5 groups (n = 6): (1) Control; (2) TAA; (3) TAA + Celastrol; (4) TAA + Antibiotic; (5) TAA + Antibiotic+Celastrol. TAA, TAA + Celastrol, TAA + Antibiotic and TAA + Antibiotic+ Celastrol groups were treated with TAA for 6 weeks (200 mg/kg). After TAA treatment for 1 week, the TAA + Antibiotic+Celastrol group was co-treated with celastrol (10 mg/kg) and antibiotics for 5 weeks.

36To determine the function of gut microbiota on the protective effect of celastrol, mice were randomly assigned to 4 groups: (1) Control; (2) TAA; (3) TAA + microbial transplantation from control group (TAA + Con-FMT); (4) TAA + microbial transplantation from celastrol 10 mg/kg group (TAA + Cela-FMT). Fresh gut microbiota from the donor mice was prepared by suspending 50 mg cecum content in 1 mL PBS followed by centrifugation at 500 g. The volume of microbial suspension was 200 μL/mouse/day. The donor mice in the TAA + Con-FMT group were healthy mice. The donor mice in the TAA + Cela-FMT group were treated with celastrol mice for 2 weeks. The TAA, TAA + Con-FMT, and TAA + Cela-FMT groups were treated with TAA for 6 weeks (200 mg/kg). After TAA treatment for 1 week, all groups were treated with antibiotics for 1 week. After antibiotics treatment, FMT was conducted for 4 weeks starting from the 15th day in the TAA + Con-FMT and TAA + Cela-FMT groups.

37To investigate the toxic effect of TCDCA on liver injury, the mice were randomly assigned into four groups (n = 6): (1) Control; (2) TAA; (3) TAA + TCDCA; (4) TCDCA. TAA + TCDCA and TCDCA groups were orally treated with TCDCA (200 mg/kg) for 5 consecutive days,. After TCDCA treatment for 4 days, TAA and TAA + TCDCA groups were given a single intraperitoneal dose of TAA (300 mg/kg dissolved in physiologic saline solution).

38To investigate the effect of intestinal FXR on hepatic fibrosis, mice were randomly assigned into 4 groups (n = 5): (1) TAA; (2) TAA + GUDCA; (3) TAA + GUDCA + Cela; (4)TAA + GUDCA + P.d. All groups were treated with TAA for 6 weeks (200 mg/kg). After TAA treatment for 1 week, TAA + GUDCA group was orally treated with GUDCA for 5 weeks (50 mg/kg dissolved in 2% DMSO + 2% Tween 80 + 96% water, once a day), TAA + GUDCA + Cela group was co-treated with GUDCA (50 mg/kg) and celastrol (10 mg/kg) for 5 weeks, and TAA + GUDCA + P.d. group was co-treated with GUDCA (50 mg/kg) and P. distasonis (2 × 108 CFU) for 5 weeks.

39To determine the protective effect of P. distasonis on MCD diet-induced hepatic fibrosis, the mice were randomly assigned to 5 groups (n = 6): (1) MCS; (2) MCD; (3) MCD + P. distasonis (MCD + P.d.); (4) MCD + heat-killed P. distasonis (MCD + P.d.-H); (5) P. distasonis (P.d.). MCD, MCD + P.d., and MCD + P.d.-H groups were fed a MCD diet for 6 weeks (TP-3001, Trophic Animal Feed High-Tech Co., Ltd, China). The control MCS group was treated with MCS diet for 6 weeks. After MCD diet treatment for 1 week, all groups were administered antibiotics for 1 week. After antibiotics treatment, the MCD + P.d. and P.d. groups were given P. distasonis by oral transplantation for 4 weeks (i.g., 2 × 108 CFU dissolved in PBS, once a day); heat-killed P. distasonis was parallelly treated. To investigate the protective effect of celastrol on MCD diet-induced hepatic fibrosis, the wild-type mice (n = 8) and Fxr-null mice (n = 6) were randomly assigned to three groups: (1) MCS; (2) MCD; (3) MCD + Celastrol. MCD and MCD + Celastrol groups were treated with MCD diet for 6 weeks. After MCD diet treatment for 1 week, the MCD + Celastrol group was orally treated with celastrol for 5 weeks (10 mg/kg, three times per week). MCD diet (TP3005G) and MCS diet (TP3005GS) were purchased from Trophic Animal Feed High-Tech Co., Ltd, China.

40Microbial genomic DNA in cecum content was extracted using the Stool Genomic DNA kit (CWBIO, China). High-throughput sequencing and analysis were performed by the Beijing Genomics Institute (Shenzhen, China). All the qualified DNA was used to construct a library. Paired-end reads were generated with the Illumine HiSeq/MiSeq platform, and the reads with sequencing adapters, N base, poly base, and low quality were filtered. Species composition and abundance analysis was then carried out.

41P. distasonis was cultured in brain-heart infusion fluid medium under anaerobic conditions at 37 °C, and the value of OD570 reached 0.6. To determine the effects of celastrol on the growth of P. distasonis, 2% (v/v) fluid medium containing P. distasonis was then added to fluid medium containing different concentrations of celastrol (48.8 nM–12.5 μM based on the maximum solubility) for 24 h. Absorbance values were measured at 570 nm. Crystal violet was used to quantify biofilm formation. Culture medium samples for metabolomics was prepared as follows: 500 μL culture medium were minced with 500 μL 50% aqueous acetonitrile. After centrifugation at 16,000 × g for 20 min, 100 μL supernatant were minced with 300 μL acetonitrile. A 5 μL aliquot was injected into the LC-MS system after centrifugation at 16,000 × g for 20 min. 5 μM chlorpropamide was used as the internal standard.

42The preparation of plasma, liver, duodenum, jejunum, ileum, cecum content, feces, and urine samples were carried out using the methods described previously,. The liquid chromatography system obtained a Vanquish autosampler, Vanquish detector, Vanquish pump (Thermo Fisher, USA) equipped with a ACQUITY UPLC HSS T3 column (2.1 × 100 mm, 1.8 μM, Waters, USA). The Vanquish UHPLC-Q Exactive plus MS (Thermo Fisher, USA) was collected in both positive and negative modes, which was operated in full-scan mode at m/z 60 to 900. The mobile phase comprised 0.1 % aqueous formic acid and acetonitrile containing 0.1% formic acid. The flow rate was 0.3 mL/min with a gradient ranging from 2% to 98% acetonitrile in a 17 min run. Principal component analysis (PCA) and orthogonal projection to latent structures-discriminant analysis (OPLS-DA) were carried out using SIMCAP 13.0 (Umetrics, Kinnelon, NJ). Targeted analysis was carried with a Thermo Xcalibur Qual Browser. The chemical structures of the altered metabolites were identified by authentic standard and MS/MS fragmentation patterns in Supplementary Tables 4–9.

43Primer sets for RNA analysis are shown in Supplementary Table 10. The following antibodies were used in western blot, immunohistochemistry and immunofluorescence: Apaf-1 (sc-65891, Santa Cruz Biotechnology, USA; WB 1:200 dilution), Caspase-11 (ab180673, Abcam, UK; WB 1:1000 dilution), Caspase-3 (9662 s, Cell Signaling Technology, USA; WB 1:1000 dilution), GSDME (ab215191, Abcam, UK; WB 1:1000 dilution), αSMA (D4K9N, 19245 s, Cell Signaling Technology, USA; WB 1:1000 dilution; IF 1:100 dilution; IHC 1:800 dilution), COL1A1 (E8F4L, 72026 s, Cell Signaling Technology, USA; WB 1:1000 dilution; IHC 1:100 dilution), TIMP1 (ab179580, Abcam, UK; WB 1:1000 dilution), TGFβ (ab215715, Abcam, UK; WB 1:1000 dilution; IHC 1:200 dilution), IL6 (D5W4V, 12912 s, Cell Signaling Technology, USA; WB 1:1000 dilution; IF 1:200 dilution), IL1β (D3H1Z, 12507 s, Cell Signaling Technology, USA; WB 1:1000 dilution), FXR (sc-25309, Santa Cruz Biotechnology, USA; WB 1:100 dilution), SHP (ab232841, Abcam, UK; WB 1:500 dilution), OSTβ (bs-2128R, Bioss, China; WB 1:1000 dilution), NLRP3 (D4D8T, 15101 s, Cell Signaling Technology, USA; WB 1:1000 dilution), Caspase-1 (E2G2I, 89332 s, Cell Signaling Technology, USA; WB 1:1000 dilution), GSDMD (sc-393656, Santa Cruz Biotechnology, USA; WB 1:200 dilution), CK19 (ab254186, Abcam, UK; IF 1:100 dilution), CD31 (ab7388, Abcam, UK; IF 1:100 dilution), GPX4 (sc-166570, Santa Cruz Biotechnology, USA; WB 1:200 dilution), COX2 (sc-376861, Santa Cruz Biotechnology, USA; WB 1:200 dilution) and GAPDH (14C10, 2119 s, Cell Signaling Technology, USA; WB 1:1000 dilution).

44To evaluate the BSH activity of P. distasonis in vivo, BSH activity in the feces and cecum content was measured by LC-MS based on the generation of CDCA from TCDCA. The unconjugated/conjugated bile acid ratios in serum, cecum content, and feces were used to show BSH activity in vivo. To evaluate the BSH activity of P. distasonis in vitro, P. distasonis was co-cultured with 25 μM TCDCA in brain-heart infusion fluid medium under anaerobic conditions at 37 °C for 24 h, and the generation of CDCA from TCDCA in the medium was measured by LC-MS.

45AST, ALT, ALP, catalase (CAT), MDA, total cholesterol (TC), total triglycerides (TG), Fe2+/Fe3+ (Nanjing Jiancheng Bioengineering Institute, China), lactate dehydrogenase (LDH, Cayman, USA), mitochondrial membrane potential (Beytime, Shanghai, China) were measured following the manufacturer’s instructions. H&E, Oil red O, Sirius red, and Masson trichrome staining were carried out.

46Primary hepatic cells, including hepatocytes, HSCs, and Kupffer cells were obtained from mouse liver by collagenase perfusion and percoll centrifugation–. Mouse primary HSC and Kupffer cells were identified by Oil red O staining and F4/80 antibody, respectively. Human LX2 cell and human primary Kupffer cells were obtained from Procell Life Science&Technology Co., Ltd. (CL-0560) and Qingqi biotech company (BFN608007039, China), respectively.

47Mouse primary hepatic cells were cultured in William E medium containing 10% fetal bovine serum. Cells were harvested 24 h after incubation with TCDCA (6.25, 25, and 100 μM), GCDCA (6.25–600 μM), CDCA (6.25, 25, and 100 μM), TAA (model drug, 16 mM) and DCA (model drug, 200 μM). Mouse primary hepatocytes were cultured in William E medium containing 10% fetal bovine serum, and the drug concentration was selected as follows: 6.25, 25, and 100 μM TCDCA, 1, 10, and 100 μM Caspase-11 inhibitor wedelolactone, 10 μM, 100 μM and 1 mM MPT agonist lonidamine, 200 nM MPT antagonist BKA. Mouse primary HSC and human LX2 cell were cultured in DMEM medium containing 10% fetal bovine serum which were harvested after incubation with TCDCA (6.25, 25, and 100 μM), Caspase-4 protein (0.5, 1, and 2 U/mL, ab51994, Abcam, UK), Kupffer cell supernatant treated with 100 μM TCDCA (half of the medium was exchanged with the supernatant in 12-well plate). Mouse and human primary Kupffer cells were cultured in DMEM medium containing 10% fetal bovine serum which were harvested after incubation with TCDCA (6.25, 25, and 100 μM).

48Primary hepatic cells after 100 μM TCDCA treatment were collected and fixed at 4 °C in 0.1 M phosphate buffered 3% glutaraldehyde and postfixed in 1% osmium tetroxide in the same buffer. After rinsing in distilled water, the cells were dehydrated in graded acetone series and embedded in SPI-Pon812. Ultrathin sections were cut and attached to copper grids with Formvar film, then stained with 2% uranyl acetate and Reynolds lead citrate, and examined under a Hitachi HT7800 electron microscope.

49For luciferase reporter gene assays, HEK293 and Caco-2 cells were transfected with FXR, tk-EcRE-luciferase and renilla-luciferase. Lipofectamine 2000 was used in the transection procedure (Invitrogen, Grand Island, NY). Transfected cells were treated with FXR agonist CDCA (50 μM), P. distasonis (2.5 × 104 and 5 × 104 CFU/mL) and FXR inhibitor TβMCA (60 μM) for 24 h2. Furthermore, Caco-2 cells were treated with FXR agonist CDCA (150 μM), P. distasonis (2.5 × 104 and 5 × 104 CFU/mL) and FXR inhibitor TβMCA (60 μM) for 24 h2, and the gene expression of Fxr, Fgf19, and Ostb were evaluated. HEK293 (CRL-1573) and Caco-2 (HTB-37) were obtained from ATCC.

50Data analyses such as ROC curves, correlation analysis, and Sankey plots were performed by GraphPad Prism 6.01 (GraphPad, USA) and OriginPro 9.0 (OriginLab, USA). Pathway enrichment was conducted with MetaboAnalyst 4.0. Differences between two groups were tested using the Student’s t test. Differences among multiple groups were tested using one-way ANOVA followed by Dunnett’s post hos comparisons. P value <0.05 was considered significant. Two times each experiment was repeated independently with similar results.

51Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.