Antagonism Between Gut Ruminococcus gnavus and Akkermansia muciniphila Modulates the Progression of Chronic Hepatitis B

Abstract

Background & aims: A long immune-tolerant (IT) phase lasting for decades and delayed HBeAg seroconversion (HBe-SC) in patients with chronic hepatitis B (CHB) increase the risk of liver diseases. Early entry into the immune-active (IA) phase and HBe-SC confers a favorable clinical outcome with an unknown mechanism. We aimed to identify factor(s) triggering IA entry and HBe-SC in the natural history of CHB.

Methods: To study the relevance of gut microbiota evolution in the risk of CHB activity, fecal samples were collected from CHB patients (n = 102) in different disease phases. A hepatitis B virus (HBV)-hydrodynamic injection (HDI) mouse model was therefore established in several mouse strains and germ-free mice, and multiplatform metabolomic and bacteriologic assays were performed.

Results: Ruminococcus gnavus was the most abundant species in CHB patients in the IT phase, whereas Akkermansia muciniphila was predominantly enriched in IA patients and associated with alanine aminotransferase flares, HBeAg loss, and early HBe-SC. HBV-HDI mouse models recapitulated this human finding. Increased cholesterol-to-bile acids (BAs) metabolism was found in IT patients because R gnavus encodes bile salt hydrolase to deconjugate primary BAs and augment BAs total pool for facilitating HBV persistence and prolonging the IT course. A muciniphila counteracted this activity through the direct removal of cholesterol. The secretome metabolites of A muciniphila, which contained small molecules structurally similar to apigenin, lovastatin, ribavirin, etc., inhibited the growth and the function of R gnavus to allow HBV elimination.

Conclusions: R gnavus and A muciniphila play opposite roles in HBV infection. A muciniphila metabolites, which benefit the elimination of HBV, may contribute to future anti-HBV strategies.

Keywords: Bile Salt Hydrolase; Cholestyramine; Cholic Acid; Immune Active.

Graphical abstract

Figure 1

Association between gut microbiota and CHB progression. (A) Histogram of LDA score of gut microbiota between IT and IA patients. Only features with LDA score >3.5 and P <.05 are shown. (B) Relative amounts (%) of indicated bacteria of each group are presented as aligned dot plots with bar graphs of means ± standard error of the mean (SEM). Mann-Whitney test was used to compare groups including IT (n = 40), IA (n = 30), spontaneous HBe-SC (n = 21), under therapy (n = 28), and HBe-SC with therapy (n = 18). (C) Fecal levels of R gnavus and A muciniphila and serum levels of HBeAg and ALT in patients with different ages and disease courses were compared. Results are given as dot plots with bar graphs of means ± SEM. Differences between groups were evaluated by Mann-Whitney test. (D) Heat map shows the correlation between bacterial levels and clinical parameters as indicated. Pearson R correlation analysis was applied to find the significant features. The bacterial levels of IT, IA, and spontaneous HBe-SC groups were compared for their correlation with age and serum levels of ALT and HBeAg. The association between bacterial levels and IT duration was calculated using data of 1–12-year age group. Bacterial data sets of IA group were further estimated for correlation with the occurrence of HBe-SC. ∗P < .05, ∗∗P < .005. A color-coded correlation scale is provided on the right of the plot. (E) Matrix of Spearman rank correlation coefficient from comparisons of indicated bacteria detected in all samples (n = 137) is shown. The correlogram was generated using corrplot package in R. The Spearman correlation was evaluated among the indicated bacterial and displayed significance (∗∗∗P < .0007) between R gnavus (RG) and A muciniphila (AM).

Figure 2

A muciniphila and R gnavus oppositely influence progression of HBV infection in BALB/c HBV-HDI mouse model. (A) Schematic time schedule of the experiment. HBV-DNA1.2 plasmid was injected into the tail veins of mice. A muciniphila and R gnavus were anaerobically grown in modified GAM broth. These bacteria were intragastrically gavaged to mice 3 times/wk signified by black arrowheads, in which each of them represents 1 time of gavaging feed. Serum and fecal samples were collected at times indicated. Fresh fecal samples of day 0 were picked before HBV injection. Ctrl mice were gavaged with modified GAM broth only. (B) Fecal counts of A muciniphila and R gnavus (mean percentage of relative abundance ± SEM) were assessed by 16S-rRNA NGS. (C) Serial serum titers of HBsAg, HBeAg, anti-HBs, and anti-HBe antibodies (mean percentage of persistent rate ± SEM) were examined. (D) Histogram of percent of cases with ALT flares (≥50 U/L, upper), clearance of HBsAg and HBeAg, HBe-SC and HBs-SC at 10 wpi (lower). Student t test was applied to identify statistical significance. ∗P < .05, ∗∗P < .005, ∗∗∗P < .0005.

Figure 3

R gnavus promotes HBV persistence in C57BL/6J mice that predisposed to HBV infection. (A) Schematic time schedule for generation of HBV-HDI mouse model. Mice (n = 8/group) were tail-vein injected with plasmid encoding HBV-DNA1.2. Live R gnavus was intragastrically gavaged to mice 3 times/wk as arrowheads indicate. Ctrl mice were gavaged with broth. Serum and fecal samples were collected at days 2 and 2–10 wpi, and fecal samples of day 0 were picked before HBV injection. (B) Fecal counts of A muciniphila and R gnavus were assessed by 16S rRNA NGS and reported as percentages of relative abundance ± SEM. (C) Serum levels of HBsAg, HBeAg, anti-HBs, and anti-HBe were examined at indicated time points and plotted as mean percentages of persistent rate ± SEM. Difference between groups was estimated with Student t test. ∗P < .05, ∗∗∗P < .0005. (D) Histogram of percentage of cases achieving clearance of HBsAg and HBeAg, as well as HBe-SC at 10 wpi. Different color codes symbolize groups of mice as described in B.

Figure 4

A muciniphila but not R gnavus changes the tendency of HBV chronicity toward rapid clearance. (A) Schematic time schedule of establishment of C3H/HeN HBV-HDI mouse model (6-wk-old, n = 8/group). HBV-carried mice were developed by hydrodynamic injection of HBV-DNA1.2 plasmids into tail veins. Live A muciniphila was intragastrically gavaged to mice 3 times/wk during 1–24 wpi. Black arrowhead denotes each gavage time point. Ctrl mice were gavaged with broth. Serum and fecal samples were serially collected at times indicated. Fecal samples of day 0 were picked before HBV injection. (B) 16S rRNA NGS profiled and quantified fecal counts of A muciniphila and R gnavus (mean percentages of relative abundance ± SEM). (C) Serum levels of HBsAg and HBeAg (mean percentages of persistent rate ± SEM) were consecutively monitored. (D) Serum HBV viral load (mean copy number ± SEM) was measured by quantitative PCR. Color codes symbolize mice groups described in B. (E) Hepatic expression of HBcAg (arrows) was detected by immunohistochemistry. (F) Percentage of cases with ALT flares (detected at 6 wpi) and HBsAg/HBeAg clearance and HBe-SC/HBs-SC (detected at 24 wpi) was calculated. Color codes represent different mice groups as described in B. (G) Schematic time schedule of experiment established in young C3H/HeN HBV-HDI mouse model (3-wk-old, n = 6/group). Experiment protocol was scheduled as described above with only a slight modification wherein the duration was shortened to 10 wpi. (H) Fecal counts of A muciniphila and R gnavus were measured by 16S rRNA NGS and reported as percentages of relative abundance ± SEM. (I) Serum levels of HBsAg and HBeAg were examined at indicated time points, compared by Student t test, and plotted as mean percentages of persistent rate ± SEM. (J) Histogram of percentage of cases succeeded in clearance of HBsAg and HBeAg, as well as HBe-SC and HBs-SC, at 10 wpi. Data were evaluated by Student t test. ∗P < .05, ∗∗P < .005, ∗∗∗P < .0005.

Figure 5

HBV clearance, induced by A muciniphila but inhibited by R gnavus, in germ-free mice. (A) Schematic time schedule for establishment of germ-free C3H/HeN HBV-HDI mice model. Each black arrowhead represents 1 time point of A muciniphila/R gnavus administration, and frequency of gavaging was 1 time/wk. (B) PCR targeting V3 and V4 16S rRNA hypervariable regions was conducted on stool samples harvested at 10 wpi. (C) 16S rRNA NGS quantified the fecal counts of A muciniphila and R gnavus (mean percentages of relative abundance ± SEM). Student t test was used to test for differences between Ctrl and A muciniphila/R gnavus-treated mice (∗∗∗P < .0005, left). Percentage of cases achieving clearance of HBsAg and HBeAg, HBe-SC and HBs-SC was assessed (right). (D) Serum HBsAg and HBeAg (mean percentages of persistent rate) of germ-free C3H/HeN HBV-HDI mice were time-serially measured. Data were evaluated by Student t test. ∗P < .05, ∗∗P < .005, ∗∗∗P < .0005.

Figure 6

Metabolomic changes during disease progression of CHB. (A and B) Stools (A) and sera (B) samples collected from IT and IA patients were subjected to UHPLC-Orbitrap-MS–based untargeted metabolite analysis. Weighted heat map images were constructed according to the log₁₀-transformed abundance of metabolites. The top 20 metabolites identified from stool samples (A) and top 18 from sera (B) are shown. (C) Cholesterol biosynthesis and metabolism pathway are schematically shown. Results of UHPLC-Orbitrap-MS–quantified metabolites are depicted as box plots and analyzed by Student t test. (D) Spearman correlation coefficients assessing relationships between relative amounts of gut microbiota and the metabolites identified. All the indicated metabolites were isolated from fecal samples of IT and IA patients except GCA, which was extracted from their serum samples.

Figure 7

Reverse regulation between A muciniphila and R gnavus in CA biosynthesis. (A) UHPLC-Orbitrap-MS–based untargeted metabolite analysis was achieved on liver tissues and sera samples of C3H/HeN HBV-HDI mice fed with modified GAM broth (Ctrl) or A muciniphila (n = 8/group). Box plots showed the differences of CA scores. (B) Quantitative PCR analysis of mRNA levels of HBV core (upper) and FXRα (lower) in liver tissues. Changes in expression levels were determined by 2^−ΔΔCt. (C) Linear regression analysis revealed the correlation between fecal A muciniphila and hepatic CA detected at 24 wpi. (D) Schematic time schedule of the study protocol. Mice were fed daily with either 0.5% CA or 2% cholestyramine diets at the same administration time as indicated. A muciniphila was intragastrically gavaged to CA-feeding mice 3 times/wk as specified by black arrowheads. Serum HBeAg was plotted as mean percentage of persistent rate ± SEM. (E) Fecal counts of R gnavus and A muciniphila were determined by 16S rRNA NGS. Results of 2–10 wpi were compared with the respective values of 0 wpi. Significant differences between groups depicted in D and E were estimated by Student t test. ∗P < .05, ∗∗P < .005, ∗∗∗P < .0005. (F) Heat map depicting the top 8 bacteria identified by NGS analysis using fecal samples of indicated mice collected after sacrifice. Color code scale shows relative abundance (%) of bacteria indicated. (G) Histologic features detected by H&E staining (upper) on liver tissues of mice fed with CA, CA+A muciniphila, and cholestyramine. Black arrows denote the necrotic hepatocytes surrounding the portal area. ZO-1 (green)-specific immunofluorescence assay (lower) was performed on ileum tissue sections of mice indicated. Hoechst fluorescence (blue) displays distribution of nuclei.

Figure 8

Cross-correlation time-series analysis of fecal CA and serum HBeAg in accordance with abundance of R gnavus and A muciniphila. (A and B) Serial sampling was conducted to obtain stool and serum samples from CHB patients who received ETV or TAF therapy, as well as patients who experienced HBe-SC. M, months. Fecal counts of bacteria were analyzed by 16S rRNA NGS, and fecal CA level was measured by targeted BAs liquid chromatography coupled to tandem mass spectrometry and serum titer of HBeAg by immunoassay. Patients who showed positive correlation between R gnavus and fecal CA are presented in A, and those who demonstrated negative correlation between A muciniphila and fecal CA are displayed in B. (C) Spearman correlation was estimated using serial sampling data of A and B for revealing relationships among A muciniphila, R gnavus, fecal CA, and serum HBeAg. ∗P < .03, ∗∗P < .002.

Figure 9

A muciniphila obstructs the BAs bioconversion activity of R gnavus. (A) Schematic diagram representing the in vitro cholesterol metabolism assay. Each reaction contained 65 mg liver slices in 3 mL modified GAM medium with or without R gnavus (2 × 10⁷), A muciniphila (2 × 10¹⁰), and 1 mmol/L cholesterol. Reactions (No. 9–12) contained A muciniphila-CM, which was prepared by centrifuging A muciniphila culture at 15,000g for 20 minutes and passed through a 0.22-μm pore size filter to fully remove bacterial cells. (B) UHPLC-Orbitrap-MS–based untargeted metabolite analysis was performed on supernatant obtained from A. Result was shown as fold relative to respective reactions without cholesterol supplement. Only a subset of metabolites that have significant fold changes are selected for visualization in a heat map. A, A muciniphila; Chol, cholesterol; CM, A muciniphila-CM; R, R gnavus. (C) Heat map clustering analysis constructed on basis of fold change in targeted BAs liquid chromatography coupled to tandem mass spectrometry profiling. Culture supernatants depicted in A were subjected to this assay. Primary and secondary BAs are highlighted in red and green fonts, respectively. (D) In vitro BSH enzymatic activity was measured using TCA, GCA, TCDCA, T-α-MCA, and T-β-MCA as substrates in the presence or absence of BSH of R gnavus. Protein expression of BSH was confirmed by Coomassie blue-stained sodium dodecyl sulfate–polyacrylamide gel electrophoresis analysis. (E) UHPLC-Orbitrap-MS characterization of A muciniphila-CM. Volcano diagrams displayed up-regulated (red) and down-regulated (green) metabolites that identified by negative and positive ion modes after normalized to the modified GAM control broth. Metabolites showed greatest value on the x-axis, which were found only in the A muciniphila-CM but completely absent from the GAM broth.

Figure 10

A muciniphila-CM eliminates HBV infection through its antibacterial and antiviral activities. (A) Flow cytometry assessing survival of R gnavus on co-cultivation with A muciniphila live cells or A muciniphila-CM for 24 hours. After performing propidium iodide stain, live (green) and dead (purple) FITC-R gnavus were analyzed. (B) Anti-HBV activity of A muciniphila-CM was measured by co-culturing with HBV-transfected Huh-7 cells in comparison with ribavirin (800 μmol/L) treatment. Culture media were harvested for detecting the secreted HBsAg, HBeAg, and extracellular and intracellular HBV-DNA 24 hours after treatment. Culture media were collected for measurement of extracellular HBV-DNA. Cultured cells were lysed for assessing intracellular HBV-DNA. (C) Schematic time schedule for establishing HBV-HDI mice administered with A muciniphila and A muciniphila-CM. Gavaging was performed 3 times/week as specified by black arrowheads with red rim. Day 0 (time of HBV injection) and day 2 are denoted by d0 and d2, respectively. Serum levels of HBsAg and HBeAg of each group were measured (mean percentages of persistent rate ± SEM). Student t test. ∗P < .05, ∗∗∗P < .0005.