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


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

In This Article

Materials and Methods

Subjects and Grouping

This case-control study enrolled the 24 patients with SAH and 24 healthy controls. Faecal samples were collected from patients with SAH who visited the six university hospitals in South Korea between December 2016 and January 2019. SAH was defined as modified MDF >32.[4] Patients were treated with oral corticosteroids (40 mg/day) or with pentoxifylline (1200 mg/day) in cases of a contraindication for corticosteroids, such as infection, pancreatitis, acute kidney injury and gastrointestinal bleeding. Faecal samples were collected before corticosteroid or pentoxifylline administration. As part of the clinical trial conducted to investigate the effects of rifaximin on SAH (NCT02485106), 14 patients were randomly selected to be treated with rifaximin (1200 mg/day), and faecal samples were collected after the 4-week follow-up from 8 patients who completed the 4-week rifaximin therapy. Faecal samples were not obtained from 3 patients who underwent liver transplantation within 4 weeks, 2 patients who were lost to follow-up and 1 patient who withdrew the consent. The study protocol was approved by the Institutional Review Board of Ajou University Hospital, Suwon, South Korea (AJRIB-BMR-KSP-18-397 and AJIRB-BMR-KSP-18-299), and informed consent was obtained from all participants.

Healthy volunteers were enrolled from Haewoondae Baek Hospital in Busan, South Korea, among individuals visiting the hospital for regular health screenings during the study period. After completion of the check-up, we selected healthy controls who were confirmed to have no known diseases and normal laboratory test results. This study was approved by the Institutional Review Board of Haewoondae Baek Hospital (IRB No. 129792-2015-064).

The gut microbiome from faecal samples was compared among the following four groups: (a) bacteria of healthy controls (BT_HC), (b) bacteria of patients with SAH (BT_SAH), (c) bacteria-derived EVs of healthy controls (EV_HC) and (d) bacteria-derived EVs of patients with SAH (EV_SAH). The effect of rifaximin treatment on the microbiome profile was also examined based on comparison before (preRXM) and after rifaximin treatment (postRXM) in bacteria and bacteria-derived EVs respectively (Figure 1). Furthermore, MDF and Lille scores were used to determine whether selected SAH-specific taxa correlated with disease severity or prognosis. For the disease severity at presentation, patients were assigned to a high MDF score (H) or a low MDF score (L) group based on the median (60.5) MDF score. In terms of prognosis and treatment response, patients were divided into the response (R, Lille score <0.45) and nonresponse (NR, Lille score ≥0.45) groups according to Lille score after 7 days of treatment.[16]

Figure 1.

Flowchart of the microbiome study design. EV, extracellular vesicles; HC, healthy controls; postRXM, SAH patients 4 weeks after rifaximin treatment; preRXM, SAH patients before rifaximin treatment; SAH, severe alcoholic hepatitis

EV Isolation and DNA Extraction From Human Faecal Samples

The human faecal samples were filtered through a cell strainer after being diluted in 10 mL of phosphate-buffered saline for 24 hours. The faecal samples were subjected to differential centrifugation at 10 000 g for 10 minutes at 4°C, and the pellet comprising bacteria were separated from the supernatant comprising EVs. The EVs were purified by sterilising the supernatant through a 0.22 μm filter to completely remove bacteria and foreign particles.

The bacteria and EV samples were boiled for 40 minutes at 100°C. The supernatant was collected after 30 minutes, and centrifuged at 18 214 g and 4°C to eliminate remaining floating particles and waste. DNA was then extracted from the bacteria and bacterial EV membrane using the DNeasy Power Soil kit (QIAGEN) according to the manufacturer guidelines, and quantified using the QIAxpert system (QIAGEN).

Bacterial Metagenomic Analysis

Bacterial genomic DNA was amplified with primers targeting the V3-V4 hypervariable regions of 16S rDNA (16S_V3_F 5'-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3' and 16S_V4_R 5'-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3'). Cycling conditions were 95°C for 3 minutes; 35 cycles of 95°C for 30 seconds, 58°C for 30 seconds and 72°C for 30 seconds; and 72°C for 5 minutes. The libraries were prepared using the polymerase chain reaction (PCR) products according to the MiSeq System guidelines (Illumina), and quantified using QIAxpert (QIAGEN). Each amplicon was then quantified, set to an equimolar ratio, pooled and sequenced on a MiSeq platform (Illumina) according to the manufacturer's recommendations.

Analysis of Bacterial Composition in the Microbiota

Raw reads were initially processed with a quality check (QC) and filtering of low-quality (<Q25) reads by trimmomatic 0.32.[17] For reads passing the QC, paired-end sequence data were merged using PandaSeq.[18] Primers were then trimmed with ChunLab Inc's in-house program at a similarity cut-off of 0.8. Sequences were cleaned using the Mothur pre-clustering program, which merges sequences and extracts unique sequences allowing up to two differences between the sequences.[19] The EzTaxon database (http://www.eztaxon-e.org/) was used for taxonomic assignment with BLAST 2.2.22, and pairwise alignment was used to calculate similarity.[20,21] Microbiome taxonomic profiles were analysed with the BIOiPLUG program (ChunLab Inc). The uchime and nonchimeric 16S rRNA database from EzTaxon were used to detect chimeras for reads with a best-hit similarity rate below 97%.[22] Sequence data were then clustered using -Hit and UCLUST, and diversity analysis was carried out.[23,24]

Statistical Analysis

Linear discriminant analysis (LDA) effect size (LEfSe) based on Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was used to predict how taxonomic differences between faecal microbiota of two groups impact their microbial metabolic potential in the main functional classes (Kyoto Encyclopedia of Genes and Genomes categories).[25] LEfSe uses the Kruskal–Wallis rank-sum test to identify features with significantly different abundances of assigned taxa between groups, and uses LDA to estimate the size effect of each feature based on a cut-off value of P < 0.05 and LDA score >2.0. The PICRUSt and LEfSe analyses were performed using the EzBioCloud database (ChunLab Inc).[26] Alpha diversity was calculated according to the abundance-based coverage estimator (ACE) and Shannon score. Beta diversity at the genus level was determined through principal coordinate analysis (PCoA) and the UniFrac distance metric.[27] Each candidate taxa accuracy for SAH was assessed by the area under the curve (AUC) based on receiver operating characteristic (ROC) curves analysis. The Firmicutes/Bacteroidetes ratios were calculated by dividing the abundance of Firmicutes by that of Bacteroidetes for each patient.

All statistical parameters were calculated using GraphPad Prism 8.0 software (GraphPad Software Inc). Values are expressed as the median and quartiles of the data. Differences with a P-value less than 0.05 were deemed to be statistically significant.