Fecal carriage of multidrug-resistant organisms increases the risk of hepatic encephalopathy in patients with cirrhosis: insights from gut microbiota and metabolite features

This prospective study was conducted at Taipei Veterans General Hospital between October 2018 and April 2022. Cirrhosis was diagnosed based on histological, clinical, biochemical, endoscopic, and imaging findings suggestive of cirrhosis in patients with chronic liver disease [27]. Patients with previous episodes of hepatic encephalopathy (HE) episode or active hepatocellular carcinoma status; with malignancies other than hepatocellular carcinoma; with human immunodeficiency virus infection or severe comorbidities, such as chronic renal failure, heart failure, or chronic obstructive pulmonary disease; and who received proton pump inhibitor, non-steroidal anti-inflammatory drugs, antibiotics, or probiotics within 1 month were excluded. Twenty-two healthy adults without underlying systemic disease were enrolled as healthy controls. This study was approved by the Institutional Review Board of Taipei Veterans General Hospital (IRB No., 2017-09-013 C and 2019-08-013 A). Written informed consent was obtained from each participant.

Demographic characteristics, laboratory data, and medical history were collected. Blood and stool samples were collected on the day of enrollment. Fresh stool samples were collected using the Commode Specimen Collection System (Thermo Fisher, MA USA). Stool samples from 18 hospitalized patients with cirrhosis were collected directly in the ward, whereas samples from the 70 outpatients with cirrhosis and 22 healthy participants were collected at home and transported to the laboratory within 2 h of collection. All stool samples were collected following standardized protocols as previously described to minimize individual variability across sample handling [28]. On arrival at the laboratory, all stool samples were immediately processed and stored at − 80 °C until analysis. Patients with cirrhosis were followed for at least 1 year, or until death or liver transplant. During the follow-up period, clinical events were recorded, which included (1) complications of cirrhosis, such as SBP, overt HE [29], newly developed or worsening ascites, acute kidney injury, first or recurrent variceal bleeding; (2) newly diagnosed hepatocellular carcinoma; (3) bacterial infections; (4) death; and (5) liver transplant.

Stool samples were inoculated onto selective agar plates at 37 °C for 24 h to detect MDRO. The identification of MDRO was based on bacterial cultures and antimicrobial susceptibility testing. MDRO was defined as bacteria resistant to at least one agent in three or more antimicrobial categories, as previously described [30]. These included methicillin-resistant Staphylococcus aureus, vancomycin-resistant enterococci, and multidrug-resistant Gram-negative bacteria, such as Enterobacterales, Pseudomonas aeruginosa, and Acinetobacter spp. Enterobacterales were classified as third-generation cephalosporin-resistant pathogens if they were not susceptible to one or more third-generation cephalosporins and multidrug-resistant Gram-negative bacteria were considered carbapenem-resistant pathogens if they exhibited resistance to carbapenem antibiotics. Bacterial isolates were identified using matrix-assisted laser desorption-ionization time-of-flight mass spectrometry (bioMérieux SA, Marcy l’Etoile, France). The antibiotic susceptibility test was performed using the VITEK2 system (bioMérieux), with results interpreted according to the 2021 Clinical and Laboratory Standards Institute guidelines [31]. Healthy adults or patients with cirrhosis in which MDRO were detected in fecal specimens were considered carriers of MDRO. Patients with cirrhosis were then divided into the MDRO carrier and non-carrier groups for further analysis. The residual stool samples were placed in 10% glycerol vials and stored at − 80 °C to further assess the microbiome.

For each stool specimen, total genomic DNA was extracted using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocols. The extracted total DNA was stored at − 80 °C for further PCR amplification. To analyze the bacterial composition, the hypervariable V3–V4 region of 16S ribosomal RNA (rRNA) genes was amplified by PCR using the 341 F (V3) and 805R (V4) primers designed for Illumina sequencing [32]. For the analysis of fungal composition, fungal-specific internal transcribed spacer (ITS) genes of distinct regions of ITS1 were amplified using the primers of ITS5-1737 F and ITS2-2043R.

Next-generation sequencing was performed using the Illumina MiSeq Desktop Sequencer following the standard protocol. Raw reads from the sequencing of 16S rRNA and fungal ITS genes were processed using QIIME 2 version 2024.5 and DADA2 plugin [33‐35]. The 16S rRNA (V3–V4) and ITS1 sequencing were conducted with an average sequencing depth of 35,000 and 100,000 reads per sample, respectively. Rarefaction was performed at 20,000 reads per sample to normalize sequencing depth. To normalize the variations in sequence depth across samples, amplicon sequence variant abundance information was rarefied to the minimum sequence depth using the QIIME script (single_rarefaction.py) [36]. Further analyses of alpha and beta diversities were both performed using the normalized data. Relative abundance normalization was then used to analyze sequencing data. The taxonomic compositions of all samples were identified.

Bacterial and fungal DNA quantification was performed using real-time PCR. The total bacterial DNA was quantified by amplifying 16S rRNA genes using specific primers as previously described [37, 38], and the fungal DNA was quantified using the Femto Fungal DNA Quantification Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s protocol [39].

Plasma lipopolysaccharides (LPS) were measured using enzyme-linked immunosorbent assay kits (Cloud-Clone Corp, Katy, TX, USA) according to the manufacturer’s instructions.

Plasma samples from all participants were immediately stored at −80 °C until analysis. An untargeted metabolomics approach was used. Each plasma sample (60 μL) was spiked with two internal standards (6 ppm lysine-¹³C6, 1 ppm stearic acid-¹³C18), selected to represent two distinct ends of the metabolite spectrum in the 9-min liquid chromatography gradient based on prior liquid chromatography mass spectrometry (LC–MS) data from our laboratory: lysine-¹³C6 (hydrophilic) elutes at 0.87 ± 0.07 min, whereas stearic acid-¹³C18 (hydrophobic) elutes at 3.85 ± 0.25 min. The samples were then deproteinized by adding 240 μL 100% methanol [40], followed by centrifugation at 4 °C and 13,000 g for 10 min. The supernatant was vacuum-dried and reconstituted in 60 µL ultrapure water for subsequent LC–MS analysis.

LC–MS analysis was performed on a Waters Xevo G2-S Q-Tof tandem mass spectrometer coupled with an Acquity UPLC system (Waters). A BEH C18 column (2.1 × 100 mm 1.7 μm, Waters) was used, and maintained at 40 °C. The total run time was 9 min, at a flow rate of 0.3 mL/min. The gradient started at 1% mobile phase B (acetonitrile with 0.1% ammonium hydroxide) from 0 to 0.5 min, and then increased to 100% B from 0.5 to 4 min. The composition was held at 100% B for 1 min, followed by a return to the initial composition over 1 min. The final conditions were maintained for an additional 3 min. Mobile phase A was an aqueous solution containing 0.1% ammonium hydroxide. Leucine enkephalin ([M − H] −  = 554.2615 m/z) was used for continuous mass calibration. MS^E data were acquired under both low and high collision energy conditions with a mass scan range of 50–1200 m/z. A pooled quality control sample was prepared by mixing plasma samples, aliquoting, and analyzing during each batch of LC–MS analysis to provide a basis for normalization. Following quality control-based normalization using the statTarget R package [41], the normalized metabolic feature levels from three measurements of each individual sample were averaged before statistical analysis.

The ionic features were tentatively identified by matching monoisotopic mass (m/z), fragmentation pattern, and retention time (if available) with the in-house library, Human Metabolome Database (http://www.hmdb.ca/), and the Human Microbial Metabolome Database (https://mimedb.org) [42]. Only [M − H] − adduction form was considered for compound identification. The MS1 and MS/MS mass tolerance were set to 20 ppm and 50 ppm, respectively.

Clinical data were expressed as the median (25 th to 75 th percentiles) or as counts, as appropriate. The chi-square or Fisher’s exact test was used to analyze categorical variables. The normality of continuous variables was assessed using the Kolmogorov–Smirnov test. Because the continuous variables were not normally distributed, Mann–Whitney U test was applied to compare continuous variables between MDRO carriers and non-carriers in the cirrhotic population. Cox regression analysis was performed to identify predictors of clinical outcomes. A logistic regression model was used to identify risk factors for the fecal carriage of MDROs. The cutoff values of the Model for End-Stage Liver Disease (MELD) score and LPS levels were determined using Youden's index. Variables with p < 0.1 in the univariate analysis were included in the multivariable analysis. Statistical significance was defined as p < 0.05. The Kruskal–Wallis test was used to assess the differences in metabolites between healthy controls and patients with and without MDRO, and between healthy controls and patients with and without HE. Data were considered significant when p < 0.05. All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 24.0 (IBM Corp. Armonk, NY, USA).

For gut bacterial and fungal analyses, the alpha diversity was evaluated using Faith’s phylogenetic diversity (Faith’s PD) and Shannon indices, with Kruskal–Wallis and Wilcoxon tests applied for statistical analysis. Beta diversity was evaluated using Bray–Curtis and Unweighted UniFrac distances and were visualized through principal coordinates analysis and nonmetric multidimensional scaling (NMDS). Results were tested for significance using the permutational multivariate analysis of variance test. Differences in diversity were further analyzed using the R vegan package. Quality filtering was performed using DADA2, and rarefaction curves confirmed sequencing saturation across all samples. To identify candidate taxa most likely to explain the differences between groups, linear discriminant analysis (LDA) effect size (LEfSe) analysis was performed with an α value = 0.05 (using Kruskal–Wallis and Wilcoxon tests) and an effect size threshold of 3 for bacteria and 2 for fungus on LDA [43, 44].

Correlations between bacteria/fungi and metabolic profiles were evaluated using Spearman’s rank test. Statistical computations were considered significant when Spearman's rank correlation coefficient > 0.2 and p < 0.05. The values of the correlation tests were adjusted using the Benjamini–Hochberg method, and a false discovery rate-adjusted p < 0.05 was considered statistically significant for multiple comparisons.

Eighty-eight patients with cirrhosis and 22 healthy volunteers were enrolled. The median age and sex distributions were comparable across groups (Supplementary Table 1). The fecal MDRO colonization rate was higher in patients with cirrhosis than in the healthy controls (33% vs. 9.1%, p = 0.026) (Supplementary Table 2). Of the 29 patients with fecal MDRO carriage, six patients (30.7%) were colonized with ≥ 2 MDROs. The MDRO most isolated was third-generation cephalosporin-resistant Escherichia coli (58.6%), followed by vancomycin-resistant Enterococcus spp. (48.3%).

Demographic characteristics were similar between MDRO carriers and non-carriers, including age, sex, cirrhosis etiology, comorbidities, and severity of liver disease (Table 1). However, MDRO carriers had higher plasma LPS levels (15.1 vs 10.4 ng/L, p = 0.006) and a higher admission rate within 30 days (34.5% vs 13.6%, p = 0.002). During a median follow-up of 16.4 months (range, 0.7–40.4 months), 36 patients (40.1%) developed complications associated with cirrhosis within 1 year (Table 2). HE occurred more frequently in MDRO carriers than in non-carriers (20.7 vs 3.4%, p = 0.008). However, other complications, including infectious events, did not differ significantly. Among infectious events, five patients developed SBP, whereas ten had non-SBP infections, including bacteremia (n = 5), intraabdominal infections (n = 2), aspiration pneumonia (n = 2), and urinary tract infection (n = 1). The positive culture rate for these infectious events was 53%, but none of the pathogens were MDROs.

Univariate Cox regression analysis showed that plasma LPS levels ≥ 14.9 ng/mL and fecal MDRO carriage were independent predictors for HE occurrence, which maintained their statistical significance on multivariable analysis, further supporting their association with HE occurrence (Table 3). Table 4 shows the risk factors associated with fecal MDRO colonization in patients with cirrhosis. In the univariate analysis, higher LPS levels (≥ 11.9 ng/mL) and prior admission in the last 30 days were significant risk factors for fecal MDRO colonization. However, in multivariable analysis, prior admission within 30 days did not predict MDRO colonization. Only LPS ≥ 11.9 ng/mL (odds ratio = 3.84; p = 0.009) remained an independent risk factor for MDRO colonization. Consistently, higher fecal bacterial DNA loads were observed in MDRO carriers compared with non-carriers and healthy adults (Supplementary Fig. 1 A).

Compared to healthy adults, Proteobacteria was predominant in fecal samples of patients with cirrhosis (Fig. 1A). At the family level, Bacteroidaceae, Enterobacteriaceae, Lactobacillaceae and Streptococcaceae increased, whereas Lachnospiraceae and Ruminococcaceae decreased. The richness and evenness of fecal bacteria measured by the Faith’s PD index and Shannon index were significantly reduced in patients with cirrhosis (Fig. 1B). In addition, the principal component analyses of the unweighted UniFrac distance and the Bray–Curtis distance showed a significant bacterial dissimilarity between these two groups (both p = 0.001; Fig. 1C).

Bild vergrößern

The phylogenetic diversity of the fecal bacteria measured by Faith’s PD index and Shannon index decreased significantly in patients with and without MDRO carriage when compared with healthy controls (Fig. 2A). No significant difference was observed between patients with and without MDROs; however, a trend toward a decrease in alpha diversity was observed in patients with cirrhosis carrying MDROs. According to unweighted UniFrac metrics, a significant dissimilarity in the fecal bacteria was observed both between healthy controls and patients with cirrhosis regardless of MDRO carriage (p = 0.001) and between patients with and without MDRO colonization (p = 0.033). However, despite overall differences across groups, MDRO-associated bacterial dissimilarity was not significant between MDRO carriers and non-carriers when measured using the Bray–Curtis distance (p = 0.134) (Fig. 2B). Furthermore, LEfSe analysis showed a prominent abundance of Streptococcus salivarius in MDRO carriers, whereas Megamonas genus was abundant in non-carriers (Fig. 2C).

Bild vergrößern

Total 4869 metabolic features were detected in plasma from patients with cirrhosis using untargeted metabolomic analysis, of which 1618 metabolites were putatively identified. Thirty-one metabolites differed significantly among healthy controls, and patients with and without MDRO carriage (Supplementary Table 3). The correlation analysis showed that abundance of six metabolites were significantly associated with dominant bacterial taxa in patients carrying MDROs (Fig. 3). The abundance of Clostridioides difficile was positively correlated with the abundance of isoaustin and 2,3-butanediol glucoside, both of which were significantly higher in MDRO carriers. Conversely, C.difficile was negatively correlated with DG (14:0/18:0/0:0), thelephoric acid, and 5-(3',4',5'-trihydroxyphenyl)-gamma-valerolactone, all of which were significantly reduced in MDRO carriers compared with non-carriers. Additionally, Streptococcus salivarius exhibited a negative correlation with PE-NMe2(24:0/20:3), which was negatively regulated in MDRO carriers compared with non-carriers. However, no definitive positive correlation was observed between Streptococcus salivarius and other metabolites. Among these six metabolites, only isoaustin was significantly elevated in  MDRO carriers who developed HE, compared with MDRO carriers without HE and healthy controls (Fig. 4).

Bild vergrößern

Bild vergrößern

Isoaustin is a fungal derived metabolite [45] and we observed that fecal samples from patients with cirrhosis carrying MDROs showed a notable increase in fungal DNA expression compared with healthy controls and patients without MDRO carriage (Supplementary Fig. 1B). To further investigate the influence of MDROs on fungal composition, we analyzed gut fungal profiles in healthy adults and patients with cirrhosis (Supplementary Fig. 2). Patients with cirrhosis, compared with healthy controls, showed a distinct change in fungal composition with an increased abundance of Candida at the genus level. While fungal alpha diversity measured by the Shannon index showed no significant difference between groups, Faith’s PD index revealed significantly greater phylogenetic richness in patients with cirrhosis. In the beta diversity analysis, unweighted UniFrac distances indicated substantial dissimilarity in fungal communities between patients with cirrhosis and healthy controls (p = 0.022), reflecting fungal dysbiosis in cirrhosis. The LEfSe analysis corroborated these findings, highlighting an elevated abundance of Candida in patients with cirrhosis. Further analyses of the fungal composition within the cirrhotic group revealed distinct patterns between MDRO carriers and non-carriers (Fig. 5). The taxa bar plot also showed an increased abundance of Candida, Pichia, and Saccharomyces in MDRO carriers compared with non-carriers at the genus level (Fig. 5A). Alpha diversity comparisons showed no significant differences between MDRO carriers and non-carriers on Faith’s PD and Shannon index (Fig. 5B). However, the beta diversity analysis using Bray–Curtis distances demonstrated significant differences in fungal communities between the two groups (p = 0.017; Fig. 5C). The LEfSe analysis further identified a predominant abundance of Saccharomycetes and Candida albicans specifically in MDRO carriers (Fig. 5D). Furthermore, we assessed the correlation between isoaustin levels and the dominant fungal profile in MDRO carriers. A positive correlation was observed between isoaustin levels and specific fungi, notably the genera Stemphylium and Stemphylium vesicarium, in MDRO carriers (Fig. 5E). This association suggests potential metabolic interactions unique to the MDRO-associated fungal profiles.

Bild vergrößern