Gut bacteria convert glucocorticoids into progestins in the presence of hydrogen gas

Abstract

Recent studies suggest that human-associated bacteria interact with host-produced steroids, but the mechanisms and physiological impact of such interactions remain unclear. Here, we show that the human gut bacteria Gordonibacter pamelaeae and Eggerthella lenta convert abundant biliary corticoids into progestins through 21-dehydroxylation, thereby transforming a class of immuno- and metabo-regulatory steroids into a class of sex hormones and neurosteroids. Using comparative genomics, homologous expression, and heterologous expression, we identify a bacterial gene cluster that performs 21-dehydroxylation. We also uncover an unexpected role for hydrogen gas production by gut commensals in promoting 21-dehydroxylation, suggesting that hydrogen modulates secondary metabolism in the gut. Levels of certain bacterial progestins, including allopregnanolone, better known as brexanolone, an FDA-approved drug for postpartum depression, are substantially increased in feces from pregnant humans. Thus, bacterial conversion of corticoids into progestins may affect host physiology, particularly in the context of pregnancy and women's health.

Keywords: 21-dehydroxylation; Eggerthella lenta; hydrogen gas; neurosteroids.

Figure 1.. Targeted steroid profiling reveals that the gut microbiome and pregnancy state influence progestin levels in the GI tract.

A) Concentrations of THPs were higher in SPF compared to GF female C57BL/6 mouse feces (n=5 and n=6 GF and SPF female mice per group, respectively, Welch’s t test). B) Concentrations of the corticoids 3α5αTHDOC, 3α5βTHDOC, 3β5αTHDOC, 3β5βTHDOC, 11-dehydrocorticosterone, and corticosterone in the bile of humans with primary sclerosing cholangitis (PSC) patients or biliary stricture (n=24 samples, one sample per patient). C) Proposed pathway for the formation of 3α5αTHP (allopregnanolone) and 3β5αTHP (isoallopregnanolone) from the abundant biliary steroid 3α5αTHDOC. HSDH, hydroxysteroid dehydrogenase. D) SPF pregnant female mouse fecal slurries 21-dehydroxylated 3α5αTHDOC while GF female mouse fecal slurries did not display this activity. Fecal slurries were incubated with 3α5αTHDOC for 48 h. THP formation was quantified by UHPLC-MS (n=3 biological replicates per fecal pellet. Two fecal pellets from SPF mouse 3 were tested). E-F) Culture pools from healthy female (E) and male (F) human feces 21-dehydroxylated 3α5αTHDOC. Feces were incubated with 3α5αTHDOC for 7 d and THP formation was quantified by UHPLC-MS (n=3 biological replicates per donor). G) Feces from pregnant human donors contained substantially higher levels of THPs than non-pregnant human feces (n=24 male samples, 17 female non-pregnant samples, and 7 pregnant samples, Kruskal-Wallis with Dunn’s multiple comparisons test). All data are presented as mean ± SEM.

Figure 2.. Human-associated Gordonibacter and Eggerthella strains 21-dehydroxylate 3α5αTHDOC.

A-D) A culture pool isolated from female donor F3 was grown in media containing 0.5% Arg and 21-dehydroxylation of 3α5αTHDOC (100 μM) was assayed after 96 h (A), (n=3 biological replicates per group). Four culture pools were created by dividing a single serial dilution plate into four quadrants (B). Quadrant pool 1 performed 21-dehydroxylation while pools 2–4 lacked this activity (C), (n=3 biological replicates per group). Comparative metagenomics analysis on each quadrant revealed G. pamelaeae as a potential 21-dehydroxylator, as it was the only species present in the active quadrant but absent in the inactive quadrants (D). E) G. pamelaeae DSM 19738 weakly 21-dehydroxylated 3α5αTHDOC (100 μM). This activity was significantly increased in co-culture with E. coli Nissle (EcN). EcN monoculture did not perform this transformation (n=3 biological replicates per condition, Browne-Forsythe ANOVA with Dunnett’s multiple comparisons test and Welch’s correction). F) Screening of an Actinobacterial strain library demonstrated strain-level variation in 21-dehydroxylation activity. Gordonibacter and closely related E. lenta strains 21-dehydroxylated 3α5αTHDOC in co-culture with EcN (n=3 biological replicates per condition). N.D. not detected. All data are presented as mean ± SEM.

Figure 3.. H₂ gas is necessary and sufficient for induction of robust 21-dehydroxylation in E. lenta.

A-C) E. lenta 14A monoculture (A), EcN monoculture (B) and E. lenta 14A and EcN co-culture (C) were imaged through Scanning Electron microscopy (SEM). Representative images are shown. White arrows indicate lysed (“ghost”) cells. D) Syringe-filtered spent supernatant from EcN induced 21-dehydroxylation in E. lenta 14A in 7 d, while vacuum-filtered supernatant lacks this ability, suggesting that a gas may be the inducing factor (n=3 biological replicates per condition). E) E. lenta 14A monoculture under a hydrogen atmosphere 21-dehydroxylated 3α5αTHDOC to a similar extent as E. lenta 14A and EcN co-culture under anaerobic chamber conditions in 72 h. Nitrogen gas was used as a non-reductive atmospheric control. No 21-dehydroxylation activity was observed from E. lenta 14A or EcN monocultures under a nitrogen atmosphere, EcN monoculture under a hydrogen atmosphere, or media alone under a hydrogen or nitrogen atmosphere (n=3 biological replicates per culture condition, n=1 replicate for media controls). F-G) E. coli BW25113 mutants containing deletions in hydrogen production pathways produced less hydrogen (F) and promoted less 21-dehydroxylation (G) than the E. coli parent strain in co-culture with E. lenta 14A. Dissolved H₂ levels and THP production from 3α5αTHDOC were quantified after 8 h and 48 h, respectively (n=3 biological replicates per culture condition, one-way ANOVA with Dunnett’s multiple comparisons test). H) E. lenta 14A and EcN synergistically produced H₂ gas in co-culture. Dissolved hydrogen gas concentrations in E. lenta 14A and EcN mono- and co-cultures were quantified after 48 h (n=3 biological replicates per condition). I-J) E. lenta 14A and G. pamelaeae DSM 19743 co-cultures with EcN synergistically produced H2 (I) and performed 21-dehydroxylation (J) (n=3 biological replicates per condition, one-way ANOVA followed by Tukey’s multiple comparisons test). K-L) Fecal culture from pregnant human donor F2 produced H₂ in 48 h (K) and THPs in 48 h (L) (n=3 biological replicates per condition, Welch’s t test). M) Both Gram-positive (pink) and Gram-negative (blue) gut microbes promoted 21-dehydroxylation in E. lenta 14A (n=3 biological replicates per condition). N.D., not detected. All data are presented as mean ± SEM.

Figure 4.. The Elen_2451–2454 gene cluster encodes molybdenum-dependent 21-dehydroxylase activity.

A) A putative 21-dehydroxylase gene cluster encoding a predicted molybdopterin oxidoreductase (Elen_2453) was identified through comparative genomics of an Actinobacteria strain library. B) The gene cluster is conserved amongst 21-dehydroxylating Coriobacteriaceae. Genetic alignment of putative 21-dehydroxylation gene cluster in Eggerthella and Gordonibacter strains was performed using the clinker pipeline. Genes in non-producer strains are shown in shades of grey. C-E) E. lenta non-producer strains containing the putative gene cluster produced less hydrogen in co-culture with EcN than E. lenta producer controls. H₂ production (C), redox potential (D), and THP production (E) from 3α5αTHDOC by E. lenta strains in co-culture with EcN were quantified at 48 h (n=3 biological replicates per culture condition, one-way ANOVA with Dunnett’s multiple comparisons test). F) 21-dehydroxylation activity was inhibited by tungstate. Either tungstate or molybdate was added to E. lenta 14A + E. coli Nissle co-cultures incubated with 3α5αTHDOC and 21-dehydroxylation activity was measured after 7 d (n=3 biological replicates per culture condition, one-way ANOVA with Dunnett’s multiple comparisons test). G) Homologous expression of Elen_2451–2454 resulted in 21-dehydroxylation of 3α5αTHDOC. Selected molybdopterin oxidoreductase-containing gene clusters were transformed into E. lenta DSM 2243 and expression was induced with 50 μM cumate. THP production was quantified after 8 h in cumate-induced and uninduced (DMF) controls (n=3 biological replicates per condition). H) G. urolithinfaciens (Gu) cannot 21-dehydroxylate 3α5αTHDOC in 7 d and does not contain Elen_2451–2454 homologs in its genome. I) Heterologous expression of a stable construct containing Elen_2451–2453 in the non-producer Gu DSM 27213 resulted in 21-dehydroxylation of 3α5αTHDOC. THP production was quantified after 8 h in cumate-induced and uninduced (DMF) controls (n=3 biological replicates per condition). J) Gordonibacter was the only genus that was significantly enriched in abundance in pregnant donor feces samples in comparison to non-pregnant donor feces, while Bernardetia was significantly enriched in non-pregnant donors (compositional differential abundance analysis was conducted using the ALDEx2, taxa with a Benjamini Hochberg False Discovery Rate-corrected Welch’s t-test of <0.1 were described as statistically significant). K-L) Feces from pregnant donors contained a higher percent abundance of Gordonibacter and Eggerthella genera (K) and higher estimated counts per million (CPM of summed genera, normalized by gene length, determined by gene-based taxonomic profiling) of Elen_2451–2454 homologs (L) than feces from healthy non-pregnant donors (n=7 subjects per group, Mann-Whitney test). M) CPM of Elen_2451–2454 correlated with the concentration of total THPs in feces in pregnant and non-pregnant human feces (n=7 subjects per group, Spearman correlation). All data are presented as mean ± SEM.

Figure 5.. 21-dehydroxylation occurs in vivo and is dependent on the Elen_2451–2454 cluster.

A) C57BL/6 female GF mice were gavaged with fecal slurry from pregnant donor P7 or PBS. THPs were quantified in feces at days 10, 16, and 28. B) Higher concentrations of total THPs were observed in feces at days 10, 16, and 28 in mice that received the FMT of donor P7 compared to PBS-gavaged mice (n=6 and 7 PBS- and P7 slurry-gavaged mice, respectively, one-tailed Welch’s t-test). C) C57BL/6 female GF mice were colonized with either E. lenta 14A (cluster-containing) + EcN or left in the GF state. D) Higher concentrations of total THPs were observed in feces at 7 d in co-colonized mice compared to GF female mice (n=5 and 6 GF and E. lenta 14A + EcN co-colonized mice, respectively, Welch’s t-test). E) C57BL/6 female GF mice were colonized with either E. lenta 14A (cluster-containing, producer) + EcN or E. lenta A2 (cluster-deficient, non-producer) + EcN. F) Higher concentrations of total THPs were observed in feces at 7 d in mice co-colonized with E. lenta 14A + EcN compared to mice co-colonized with E. lenta A2 + EcN (n=15 mice per group, Welch’s t-test). G) Concentrations of individual THPs in co-colonized mouse feces at 7 d (n=15 mice per group, Welch’s t-test). N.D., not detected. All data are presented as mean ± SEM.