Microbiome as a Potential Diagnostic and Predictive Biomarker in Severe Alcoholic Hepatitis

Soon Sun Kim; Jung Woo Eun; Hyo Jung Cho; Do Seon Song; Chang Wook Kim; Young Seok Kim; Sung Won Lee; Yoon-Keun Kim; Jinho Yang; Jinhee Choi; Hyung Joon Yim; Jae Youn Cheong

Disclosures

Aliment Pharmacol Ther. 2021;53(4):540-551. 

In This Article

Results

Study Population and Clinical Course

Demographic and clinical parameters of the 24 patients with SAH and 24 healthy controls included in the study are summarised in Table 1. The median age was 50 years, and 62.5% of the subjects were men in both groups. The median MDF score for patients with SAH was 60.5 (interquartile range [IQR], 41.3–68.7) and the median Model for End-stage Liver Disease (MELD) score was 24.1 (IQR, 21.5–27.4). Sixteen patients (66.7%) received corticosteroid treatment and eight patients (33.3%) who had a contraindication for corticosteroid were treated with pentoxifylline. As described above, 14 patients were randomly assigned to receive rifaximin treatment, and eight patients completed 4 weeks of rifaximin treatment. As shown in Table 2, liver function markers, including MDF, MELD score, bilirubin, prothrombin time (international normalised ratio) and albumin, significantly improved after 4 weeks of rifaximin treatment.

Microbiome Analysis of Bacteria and Bacteria-derived EVs in Patients With SAH and Healthy Subjects

At the community level, SAH samples (gut bacteria and EVs) had generally lower alpha diversity than those of the healthy control group (Figure 2A,B). PCoA at the genus level demonstrated a clear separation between the groups (Figure 2C), and the UniFrac scores were significantly increased in the SAH groups compared with those of the healthy control groups (Figure 2D).

Figure 2.

Alpha- and beta-diversity comparison of microbiomes in severe alcohol hepatitis (SAH) patients and healthy controls. Box-and-whiskers plots based on (A) ACE and (B) Shannon indices for alpha diversity. (C) Principal coordinate analysis (PCoA) of bacterial beta diversity based on the UniFrac distance metric at the genus level. Green circles represent bacteria of healthy controls (BT_HC), blue circles represent bacteria of SAH patients (BT_SAH), yellow circles represent extracellular vesicles of healthy controls (EV_HC) and red circles represent extracellular vesicles of SAH patients (EV_SAH). (D) The beta diversity among patients with SAH was significantly higher than that among healthy controls. All P-values were determined by the two-sided Kruskal-Wallis H test. *P < 0.05; **P < 0.01; ***P < 0.001

Relative changes in microbial abundance at the phylum, class, family and genus levels are presented in Figure 3 and Table S1. At the phylum level, Verrucomicrobia was only detected in the healthy control groups (3.2% in BT_HC and 1.6% in EV_HC), but not in the SAH groups. Bacteroidetes was also less abundant in BT_SAH (15.2%) and EV_SAH (15.2%) samples than in the BT_HC (28.9%) and EV_HC (24.0%) samples. By contrast, Fusobacteria were only detected in the SAH groups (1.6% in BT_SAH and 1.4% in EV_SAH). The Firmicutes/Bacteroidetes ratios were significantly higher in SAH groups (BT_SAH and EV_SAH) than in healthy controls (Figure S1A).

Figure 3.

Relative abundance of the microbiome identified in the bacteria of healthy controls (BT_HC), bacteria of severe alcoholic hepatitis patients (BT_SAH), extracellular vesicles of healthy controls (EV_HC) and extracellular vesicles of severe alcoholic hepatitis patients (EV_SAH) at the (A) phylum, (B) class, (C) family and (D) genus levels

Significant differences in bacterial composition between SAH patients and healthy controls are summarised in Table S2. A total of 267 and 244 taxa were significantly altered between the groups for the bacteria and EV samples, respectively. The top 20 increased or decreased bacterial and EV taxa in SAH patients, and their relative abundances in the SAH and healthy control groups are visualised in Figure 4A,B respectively. Among the taxa with increased abundance in BT_SAH, Bacilli (class), Lactobacillales (order) and Veillonella (genus) showed the highest AUC values (0.896, 0.896 and 0.864, respectively) for predicting SAH (Figure 4C). Among taxa with decreased abundance in BT_SAH, Eubacterium_g23, Oscillibacter and Clostridiales showed the highest AUC values (0.949, 0.924 and 0.904, respectively) for predicting SAH (Figure 4D). In analysis of the EVs of microbes, increased Veillonella, Veillonella parvula group and Lactobacillales, and decreased Eubacterium_g23, Oscillibacter and Christensenellaceae abundances showed the highest AUC values for predicting SAH (Figure 4E,F). Detailed AUC values and 95% confidence intervals of each taxon are summarised in Table S3.

Figure 4.

Relative abundance of predicative microbial features and diagnostic power for the prediction of severe alcoholic hepatitis (SAH). Top 20 taxa with increased (upper) and decreased (lower) abundance in the (A) bacteria of SAH patients (BT_SAH) and (B) extracellular vesicles (EVs) of SAH patients (EV_SAH). Box-and-whiskers plots showing the relative abundance for each group in the cohort. Receiver operating characteristic curves of genera with (C) increased and (D) decreased abundance in the gut microbiome of SAH patients. Receiver operating characteristic curve of genera with (E) increased and (F) decreased abundance in the microbe-derived EVs of SAH patients

Changes in the Microbiome After Rifaximin Treatment

The alpha diversity (ACE score) of bacteria was lower in patients with SAH before rifaximin treatment than that of healthy controls, and was not affected by rifaximin treatment (Figure 5A). By contrast, beta diversity (UniFrac score) was higher in patients with SAH than that in the healthy control group and decreased significantly after rifaximin treatment (Figure 5B). However, there was no difference in the alpha diversity of EVs between the two groups, and the beta diversity was significantly higher in patients with SAH before rifaximin treatment, which did not change after 4 weeks of rifaximin treatment (Figure 5C).

Figure 5.

Effect of rifaximin on the gut microbiota composition. (A) Schematic of the experimental design; arrows indicate the days of rifaximin treatment. (B) Box-and-whiskers plots of alpha diversity (left) and beta diversity (right) based on UniFrac distances in the gut microbiome. (C) Box-and-whiskers plots of alpha diversity (left) and beta diversity (right) based on UniFrac distances in the microbe-derived extracellular vesicles. All P-values were determined by the two-sided Kruskal-Wallis H test. **P < 0.01; ***P < 0.001. BT_HC, bacteria of healthy controls; BT_postRXM, bacteria of post-rifaximin treatment group; BT_preRXM, bacteria of pre-rifaximin treatment group; EV_HC, extracellular vesicles of healthy controls; EV_postRXM, extracellular vesicles of post-rifaximin treatment group; EV_preRXM, extracellular vesicles of pre-rifaximin treatment group

Relative changes in microbial abundance between the three groups were evaluated using LefSe analysis. At the phylum level, the Firmicutes/Bacteroidetes ratio of the BT_postRXM group was higher than that of the BT_preRXM group; however, the ratio was not significantly different between the EV_preRXM and EV_postRXM groups (Figure S1B,C).

A total of 104 taxa showed significantly different abundances in the BT_HC, BT_preRXM and BT_postRXM groups, and 94 taxa exhibited significantly different abundances among the EV_HC, EV_preRXM and EV_postRXM groups (Table S4). Among them, there were 29 common taxa that were altered in the BT and EV groups, which are listed in Table S5.

We selected the taxa that increased or decreased in the SAH patients relative to those of controls but were then restored to control levels after rifaximin treatment for further analysis. Eight taxa had reduced abundance in the BT_preRXM group compared to that in the BT_HC group and increased following treatment in the BT_postRXM group, whereas seven taxa showed the opposite pattern with higher microbial abundance in the BT_preRXM group than in the controls that decreased after rifaximin treatment (BT_postRXM group) (Figure 6A). In EVs, 10 taxa increased before rifaximin treatment and then decreased after treatment, and 13 taxa showed the opposite direction of change (Figure 6B). Figure 6C,D present representative taxa exhibiting restoration to control levels after rifaximin treatment. In both the gut bacteria and EVs, the abundances of Clostridium ramosum, Clostridium_g6, Fusobacteriaceae, Fusobacterium, Veillonella and Veillonella parvula group were higher before rifaximin treatment and then decreased after treatment. By contrast, the abundances of Lachnospira, PAC001129_s, PAC001162_s, Prevotella, Prevotellaceae, Roseburiacecicola group, Ruminococcus, Sporobacter and Subdoligranulum were lower in patients with SAH and increased after rifaximin treatment (Table S5).

Figure 6.

Comparisons of the effects of rifaximin treatment. Bar graphs represent the relative abundance of predictive microbial taxa in the (A) gut microbiome and (B) microbe-derived extracellular vesicles (EVs). Representative bacterial genera that showed changes in abundance following rifaximin treatment in the (C) gut microbiome and (D) microbe-derived EVs. Bar charts show changes in relative abundance before and after rifaximin treatment in each patient with severe alcohol hepatitis (SAH). BT_postRXM, bacteria of post-rifaximin treatment group; BT_preRXM, bacteria of pre-rifaximin treatment group; EV_postRXM, extracellular vesicles of post-rifaximin treatment group; EV_preRXM, extracellular vesicles of pre-rifaximin treatment group

Role of Selected Taxa in Predicting Disease Severity and Treatment Response

Next, we analysed the association between the 15 selected taxa and disease severity or treatment response. Table S6 summarises the mean abundance of each taxon in MDF(H), MDF(L), Lille(NR) and Lille(R) groups. Only the abundance of Prevotella was different significantly according to Lille score. Responders to SAH treatment (Lille score <0.45) showed higher Prevotella abundance than did nonresponders (LDA effect size 4.32, P-value = 0.046) only in BT (Figure 7A). No taxon demonstrated significant difference according to MDF score.

Figure 7.

Gut microbiota composition and treatment response in patients with severe alcoholic hepatitis. (A) Violin curve shows the relative abundance of Prevotella in bacteria of treatment responders (Lille[R]) and treatment nonresponders (Lille[NR]). (B) Box-and-whiskers plots of alpha diversity in the gut microbiome (BT) and microbe-derived extracellular vesicles (EVs) based on treatment response. (C) Beta diversity based on UniFrac distances in BT and EV based on treatment response. Treatment response was defined based on Lille score after 7 days of treatment (responder <0.45 vs nonresponder ≥0.45) [Colour figure can be viewed at wileyonlinelibrary.com]

At the community level, the alpha diversity of the treatment response group (BT_Lille[R]) was higher than that of the BT_Lille(NR), but the beta diversity did not differ between the two groups (Figure 7B,C). In terms of MDF score, the alpha and beta diversities of the MDF(H) groups were not significantly different from those of the MDF(L) groups.

Predicted Functional Composition of Metagenomes Based on Rifaximin Treatment

LEfSe analysis based on the PICRUSt dataset showed that metabolic pathway, ABC transporters and biosynthesis of secondary metabolites pathways were commonly enriched in both bacteria and microbe-derived EVs of patients with SAH (Figure 8A,B). However, there were no significant changes in enriched gene pathways of microbe-derived EVs before and after treatment with rifaximin. The porphyrin and chlorophyll metabolic pathway was the only enriched gene pathway in the BT_postRXM group compared to that in the BT_preRXM group (Figure 8C).

Figure 8.

Predicted functional composition of metagenomes in patients with severe alcohol hepatitis (SAH). Linear discriminant analysis effect size (LEfSe) analysis was conducted on metabolic functions using Kyoto Encyclopedia of Genes and Genomes categories in the gut microbiome (A) and microbe-derived extracellular vesicles (EVs) (B) of patients with SAH. (C) Only one pathway was enriched in the gut microbiome after rifaximin treatment, and no pathway was enriched in EVs. LEfSe and false discovery rate (FDR) q-values were used to rank the pathways enriched in each phenotype

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....