Unique gut microbiota and metabolomic profiling as biomarker of post-transplant recovery in acute-on-chronic liver failure after liver transplantation

Abstract

Abstract

Acute-on-chronic liver failure (ACLF) is a severe condition arising from chronic liver disease, characterized by acute decompensation, organ failure, and high short-term mortality. Poor outcomes have also been observed in patients with ACLF after liver transplantation (LT). Emerging evidence, including a study from our center, suggests that gut microbiota plays an important role in ACLF. Patients who underwent LT at our center between October 2022 and June 2024 were included. Fecal samples were collected within 1 month post-LT for 16S rRNA and untargeted metabolomic sequencing. In this study, 144 samples from 69 patients with ACLF, cirrhosis, or hepatocellular carcinoma (HCC) were analyzed. Distinct microbiota and metabolic profiles were observed among the groups. ACLF patients exhibited significantly altered beta diversity, with notable depletion of g__Anaerostipes. Metabolomic analysis revealed substantial differences, including enrichment of tangeritin and depletion of candesartan in the ACLF group. Network analysis identified g__Anaerostipes as a key node linking differential taxa and metabolites. A random forest model based on these features effectively distinguished patient groups, with the highest classification accuracy observed in HCC. Multi-omic signatures were also associated with early allograft dysfunction (EAD), particularly g__Lachnoclostridium. Several microbial and metabolic features, including g__Lachnoclostridium, showed significant correlations with clinical indicators. The gut microbiome after LT is closely associated with ACLF. This study offers valuable insights for further investigation into the pathogenesis and post-LT prognosis.

Key points

• ACLF patients have a unique gut microbiota and metabolic profile after LT

• g__Anaerostipes is the prominent biomarker of ACLF's multi-omics signature

• g__Lachnoclostridium is a promising indicator of recovery after LT

Supplementary Information

The online version contains supplementary material available at 10.1007/s00253-026-13774-5.

Introduction

Acute-on-chronic liver failure (ACLF) is a life-threatening syndrome characterized by acute hepatic decompensation in patients with chronic liver disease, frequently complicated by extrahepatic organ failure and high short-term mortality (Arroyo et al. 2015). Liver transplantation (LT) remains the only established effective treatment. Elucidation of the underlying mechanisms and identification of alternative therapeutic strategies are urgently needed.

Cirrhosis-associated immune dysfunction provides the immunological substrate for ACLF (Albillos et al. 2014), while high-grade systemic inflammation triggered by diverse insults drives disease progression (Clària et al. 2016). Gut-derived pathogen-associated molecular patterns (PAMPs), arising from impaired intestinal barrier integrity and microbiota dysbiosis, are key contributors to this inflammatory milieu (Clària et al. 2016; Trebicka et al. 2024).

Although LT improves survival in ACLF, patients with higher-grade disease exhibit lower 1-year survival and higher infection rates (Li et al. 2025), independent of preoperative stabilization (Sundaram et al. 2020). Compared to relatively stable advanced liver disease, ACLF is associated with poorer post-transplant outcomes (Cervantes-Alvarez et al. 2022; Karvellas et al. 2023; Kwong et al. 2024; Watt et al. 2010). The underlying mechanisms remain unclear, and the gut microbiota linked to ACLF may contribute to this discrepancy.

In this study, we identified a distinct gut microbial and metabolic profile in ACLF patients during the early post-transplant period and explored its association with postoperative recovery, providing new insights into ACLF pathophysiology and post-LT prognosis.

Methods

Study design

This retrospective observational study aimed to identify the distinct gut microbiome and metabolomic profiles of ACLF patients following liver transplantation and explore their potential associations with prognosis. Patients who underwent liver transplantation at our center between October 2022 and June 2024 were considered for inclusion. Exclusion criteria were as follows: age under 18 or over 70 years, received oral antibiotics within 1 month before surgery, received multi-organ transplantation, or those who did not receive tacrolimus post-transplantation. ACLF diagnosis was confirmed according to the European Association for the Study of the Liver-Chronic Liver Failure Consortium (EASL-CLIF) ACLF definition (Peng et al. 2025). Cirrhosis and hepatocellular carcinoma (HCC) diagnoses were confirmed through pathological examination of liver tissue.

The primary outcome was early allograft dysfunction (EAD), which was defined as recipients presenting at least one of the criteria: (1) total bilirubin ≥ 10 mg/dL on day 7; (2) international normalized ratio > 1.6 on day 7; (3) alanine aminotransferases (ALT) or aspartate aminotransferases (AST) > 2000 international units (IU)/L within the first 7 days.

Post-LT treatment

All recipients received standard triple immunosuppressive therapy with tacrolimus, mycophenolic acid, and corticosteroids. The initial tacrolimus dose (0.10–0.15 mg/kg/day) was adjusted to maintain trough levels of 8–10 ng/mL. Mycophenolate mofetil was administered at 0.75 g every 12 h or enteric-coated mycophenolate sodium at 0.54 g every 12 h. Basiliximab was used for induction therapy, and intravenous antibiotics were provided as prophylaxis.

Sample collection

Fecal samples were collected 7 days, 14 days, and 21 days after LT in sterile tubes, immediately frozen, and stored at –80 °C until DNA and metabolite extraction. When multiple samples were available from the same patient, analyses were performed at the patient level, and repeated measures were not treated as independent observations.

16S sequencing and data analysis

Details of DNA extraction and library preparation are provided in the Supplemental Methods. Paired-end reads were merged, and amplicon sequence variants (ASVs) were inferred using DADA2 in QIIME2 (Bolyen et al. 2019), with taxonomic annotation based on the SILVA 138 database (Quast et al. 2013). Downstream analysis was performed in R (package MicrobiotaProcess, v1.13.2.994; Xu et al. 2023).

Alpha diversity was evaluated using six indices (Observed, Chao1, ACE, Shannon, Simpson, and Pielou). Beta diversity was assessed by principal coordinates analysis (PCoA) based on Bray–Curtis distances, and group differences were tested by permutational multivariate analysis of variance (ADONIS). Differential taxa were identified using linear discriminant analysis effect size (LEfSe) with linear discriminant analysis (LDA) > 2 as the threshold. Functional pathway prediction was conducted using PICRUSt2 (Douglas et al. 2020), with visualization via ggpicrust2 (v1.7.3; Yang et al. 2023). The sequences have been deposited under the accession numbers SRA PRJNA1418176.

Untargeted metabolomics analysis

Extraction and preparation of metabolites are detailed in the Supplemental Methods. Data preprocessing, normalization, and statistical analysis were performed using the R package metaX (Wen et al. 2017). Principal component analysis (PCA) and PCoA were applied to identify clustering trends and outliers, with group differences assessed by ADONIS (Bray–Curtis).

Differential metabolites were defined by orthogonal partial least squares–discriminant analysis (OPLS-DA) with variable importance in projection (VIP) > 1, fold change > 1.2, and p < 0.05 (Wilcoxon test).

Model building

The discriminatory performance of differential microbiota and metabolites was evaluated using random forest, support vector machine (SVM), and k-nearest neighbors (KNN) algorithms (caret v6.0–94, randomForest v4.7-1.1Max 2008; Wiener 2002). Features identified by LEfSe or VIP ranking were used for model input. Models were trained with fivefold cross-validation repeated three times. Classification performance was assessed by area under the curve (AUC), accuracy, and Cohen’s kappa. Feature importance was determined by mean decrease in accuracy and Gini index.

Statistical analysis

Statistical analyses were performed using R (v4.2.1, Posit Software, Boston, MA, USA), SPSS 26 (IBM, Chicago, IL, USA), and GraphPad Prism 8 (GraphPad Software, San Diego, CA, USA). Between-group comparisons were conducted using the Wilcoxon rank-sum test, unpaired t-test, ANOVA, chi-square, or Fisher’s exact test, as appropriate. Correlations among taxa, metabolites, and clinical indices were assessed by Spearman’s correlation and simple linear regression. Co-occurrence networks were constructed using nodes and edges with |r|> 0.2 and p < 0.05. Statistical significance was set at p < 0.05.

Results

Study population

A total of 69 patients with 144 samples were included in the study: 24 with ACLF, 32 with cirrhosis, and 13 with HCC. Table 1 summarized the characteristics of the three patient groups. Significant differences in hepatic and renal function indexes were observed, likely reflecting their diagnostic relevance in distinguishing ACLF from cirrhosis. The distribution of etiologies also varied markedly, possibly due to differences in disease prevalence in China and progression patterns within the liver disease population. Sample distribution was consistent among the three groups (Supplemental Table. S1).

Table 1.

The characters of inclusive patients

Parameter	ACLF	Cirrhosis	HCC	p
n	24	32	13
Baseline characters
Age, years (mean (SD))	45.83 (9.11)	47.31 (9.71)	52.15 (9.19)	0.150
Height, m (mean (SD))	1.68 (0.06)	1.66 (0.07)	1.69 (0.06)	0.285
Weight, kg (mean (SD))	65.50 (13.95)	63.59 (12.52)	69.04 (12.00)	0.443
BMI, kg*m−2 (mean (SD))	23.24 (4.32)	23.07 (4.02)	24.19 (3.79)	0.700
Primary disease (n (%))				0.046
HBV	17 (70.8)	15 (46.9)	12 (92.3)
Alcoholic liver disease	0 (0.0)	4 (12.5)	0 (0.0)
Autoimmune liver disease	1 (4.2)	4 (12.5)	0 (0.0)
Mixed etiologies	2 (8.3)	7 (21.9)	1 (7.7)
Others	4 (16.7)	2 (6.2)	0 (0.0)
ALT, U/L (mean (SD))	63.42 (43.41)	64.56 (69.20)	64.31 (131.60)	0.998
AST, U/L (mean (SD))	117.58 (90.15)	113.09 (134.21)	75.38 (88.06)	0.518
TBil, µmol/L (mean (SD))	404.44 (206.58)	197.51 (208.65)	77.05 (131.64)	< 0.001
DBil, µmol/L (mean (SD))	256.37 (135.17)	132.59 (151.77)	46.11 (92.97)	< 0.001
BA, µmol/L (mean (SD))	329.32 (196.53)	319.47 (715.26)	121.52 (138.94)	0.436
INR (mean (SD))	2.53 (1.12)	1.59 (0.38)	1.41 (0.49)	< 0.001
Scr, µmol/L (mean (SD))	150.50 (162.52)	66.88 (24.27)	79.00 (37.67)	0.008
BUN, mmol/L (mean (SD))	10.96 (8.42)	5.24 (2.21)	7.01 (5.70)	0.002
UA, µmol/L (mean (SD))	236.58 (177.91)	243.31 (124.33)	281.08 (116.82)	0.649
Alb, g/L (mean (SD))	35.13 (5.32)	32.07 (6.22)	33.10 (4.51)	0.138
Hb, g/L (mean (SD))	97.50 (17.68)	101.38 (23.19)	112.15 (18.62)	0.123
Plt, *109/L (mean (SD))	67.83 (35.75)	69.91 (48.50)	65.23 (36.02)	0.943
WBC, *109/L (mean (SD))	6.25 (2.19)	5.74 (4.61)	4.02 (2.29)	0.188
Lym, *109/L (mean (SD))	0.96 (0.49)	0.79 (0.46)	0.90 (0.45)	0.392
Neu, *109/L (mean (SD))	4.56 (1.91)	4.13 (4.24)	2.67 (2.28)	0.241
Post-LT characters
Antibiotic type (n (%))				0.788
Beta-lactam	24 (100.0)	32 (100.0)	13 (100.0)
Engpolymyxin	9 (37.5)	5 (15.6)	2 (15.4)
Glycopeptides	21 (87.5)	28 (87.5)	11 (84.6)
Anti-fungal drugs	19 (79.2)	28 (87.5)	10 (76.9)
Antibiotic duration (mean (SD))	18.71 (6.18)	25.00 (15.92)	26.31 (17.11)	0.163
C0 of tacrolimus after LT, µmol/L (mean (SD))
7 days	6.51 (2.65)	7.30 (3.53)	5.85 (2.23)	0.317
14 days	7.78 (2.72)	8.32 (2.88)	7.48 (4.00)	0.667
21 days	8.16 (3.62)	7.90 (3.05)	6.85 (2.13)	0.493

ACLF acute-on-chronic liver failure, HCC hepatocellular carcinoma, BMI body mass index, HBV hepatitis B virus, ALT alanine transaminase, AST aspartate transaminase, TBil total bilirubin, DBil direct bilirubin, BA bile acid, INR international normalized ratio, Scr serum creatinine, BUN blood urea nitrogen, UA uric acid, Alb albumin, Hb hemoglobin, Plt platelet, WBC white blood cell, Lym lymphocyte, Neu neutrophil, C₀ trough blood concentrations, LT liver transplantation

Unique gut microbiota and metabolomic profiling in patients with ACLF

There was no significant difference in bacterial composition between time points (Supplemental Fig S1). Microbial analysis showed that the accumulation curve indicated that most samples reached a plateau for the observed index (Supplemental Fig. S2A). No significant differences in read counts were observed among the three groups (Supplemental Fig. S2B); however, the ACLF group showed a marked reduction in ASV count (Fig. 1A). Alpha diversity did not differ significantly (Supplemental Fig. S2C), whereas beta diversity differed significantly among all groups and in each pairwise comparison (Fig. 1B). At the phylum level, microbial composition varied significantly (Fig. 1C), with notable differences in taxa such as p__Fusobacteriota (p = 0.0378) and p__Synergistota (p = 0.0374). At the genus level, LEfSe analysis identified g__UBA1819 as enriched in ACLF, g__Anaerostipes in cirrhosis, and g__Intestinibacter in HCC (Fig. 1D). Functional predictions by PICRUSt2 revealed group-specific pathways, with alpha-linolenic acid metabolism enriched in the ACLF group (Fig. 1E).

Fig. 1.

Unique gut microbiota profiling in patients with ACLF. A Boxplot illustrating the sample ASV numbers between groups. B PCoA plot illustrating the distribution of groups, including the degree of explanation of PCs, distribution over the PCs, Observed, and Shannon. PCoA coordinates were calculated once using the full dataset and the same distance matrix. Colors indicate different grouping variables. C Bar chart showing the top 10 species composition at the phylum level across the groups. D Differential taxon identified by the LEfSe between groups. E The differences in predictive functional pathways between the two groups. ACLF, acute-on-chronic liver failure; HCC, hepatocellular carcinoma; ASV, amplicon sequence variants; PCoA, principal coordinates analysis; LEfSe, linear discriminant analysis effect size

In metabolic analysis, the base peak chromatograms in both positive and negative ion modes exhibited high overlap among all quality control (QC) (Supplemental Fig. S3A and S3B). PCA of QC distributions showed consistent clustering along the first two principal components (Supplemental Fig. S3C). Over 90% of metabolites of QC displayed a coefficient of variation below 0.3 across samples, indicating robust instrumental stability and high data quality (Supplemental Fig. S3D). A broad spectrum of metabolites, primarily lipids, benzenoids, and amino acids (Fig. 2A) was identified and annotated as involved in amino acid metabolism and other pathways (Fig. 2B). Numerous differential metabolites were detected between each pair of groups (Fig. 2C), with no overlap observed among the three pairwise comparisons (Fig. 2D). In total, 352 features were significantly enriched or depleted between the ACLF and cirrhosis groups. Among these, tangeritin showed the greatest enrichment in the ACLF group, while candesartan exhibited the most pronounced depletion (Fig. 2E).

Fig. 2.

Unique gut metabolic profiling in patients with ACLF. A Bar chart of metabolite classification in all samples. B Bar chart of pathway classification in all samples. C Differential metabolites identified by the Wilcoxon rank-sum test and OPLS-DA. D Venn diagram showing the differential features shared among the group pairs. E Volcano plot illustrating the differential metabolites. ACLF, acute-on-chronic liver failure; HCC, hepatocellular carcinoma; OPLS-DA, orthogonal partial least squares-discriminant analysis

Network analysis highlighted g__Anaerostipes as a central integrative node

Differential metabolite enrichment analysis (Fig. 3A-C) revealed taurine and hypotaurine metabolism as a hallmark pathway in ACLF, while amino acid biosynthesis characterized cirrhosis, and drug metabolism by other enzymes was a feature of HCC. To explore the potential link between metabolites and taxon, network analysis was performed between above pathway-related metabolites and differential microbiota. The results identified g__Anaerostipes as a key node linking multi-omics features (Fig. 3D).

Fig. 3.

Network highlighted g__Anaerostipes as a central integrative node. A Differential metabolic pathways between group ACLF and cirrhosis. B Differential metabolic pathways between group ACLF and HCC. C Differential metabolic pathways between group cirrhosis and HCC. D Network illustrating the correlations between taxon and metabolites. ACLF, acute-on-chronic liver failure; HCC, hepatocellular carcinoma

Differential characteristics effectively distinguish patients in random forest models

To further investigate the multi-omics feature differences between groups, fivefold three times repeated cross-validation of machine learning classification model was established using differential taxon and metabolites. The results indicated that the random forest model outperformed KNN and SVM in AUC, accuracy, and kappa (Fig. 4A). Notably, the median AUC for all subgroups exceeded 75%, with the highest classification ability observed in the HCC group (Fig. 4B). In the feature importance evaluation, indigo carmine was found to contribute most significantly to the model (Fig. 4C).

Fig. 4.

Differential characteristics effectively distinguished patients in machine learning models. A Performance of a threefold cross-validated machine learning model in categorizing patients. B Performance of random forest model in categorizing subgroup. C Importance of variants in random forest model. ACLF, acute-on-chronic liver failure; HCC, hepatocellular carcinoma; AUC, area under curve; RF, random forest; SVM, support vector machine; KNN, k-nearest neighbor

Multi-omics was associated with early allograft dysfunction

EAD developed in 24 patients. Except for blood urea nitrogen, no significant differences in characteristics were observed between the EAD and Control groups (Table 2). There was no difference in sample distribution between groups EAD and control (Supplemental Table 2). Overall gut microbiota composition differed significantly between groups (Fig. 5A). LEfSe analysis at the genus level identified g__Lachnoclostridium as enriched in the EAD group (Fig. 5B). Several differential metabolites were detected, with diflufenican enriched and N,N,N-trimethyl-L-alanyl-L-proline betaine depleted in EAD (Fig. 5C). These metabolites were primarily enriched in pathway epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor resistance (Fig. 5D).

Table 2.

The baseline characters of group EAD and control

Parameter	EAD	Control	p
n	24	45
Baseline characters
Diagnosis (n (%))			0.191
ACLF	8 (33.3)	16 (35.6)
Cirrhosis	14 (58.3)	18 (40.0)
HCC	2 (8.3)	11 (24.4)
Age, years (mean (SD))	47.12 (11.24)	48.02 (8.62)	0.713
Height, m (mean (SD))	1.67 (0.07)	1.67 (0.06)	0.564
Weight, kg (mean (SD))	64.46 (14.36)	65.70 (12.23)	0.710
BMI, kg*m−2 (mean (SD))	23.26 (4.33)	23.38 (3.94)	0.906
Primary disease (n (%))			0.829
HBV	16 (66.7)	28 (62.2)
Alcoholic liver disease	2 (8.3)	2 (4.4)
Autoimmune liver disease	2 (8.3)	3 (6.7)
Mixed etiologies	3 (12.5)	7 (15.6)
Others	1 (4.2)	5 (11.1)
ALT, U/L (mean (SD))	57.50 (37.51)	67.64 (91.16)	0.604
AST, U/L (mean (SD))	93.75 (65.05)	114.91 (130.64)	0.459
TBil, µmol/L (mean (SD))	244.66 (226.95)	247.92 (233.70)	0.956
DBil, µmol/L (mean (SD))	153.58 (149.15)	162.43 (161.01)	0.824
BA, µmol/L (mean (SD))	228.66 (201.49)	315.97 (609.75)	0.499
INR (mean (SD))	2.00 (1.05)	1.83 (0.77)	0.452
Scr, µmol/L (mean (SD))	66.79 (26.87)	115.02 (125.37)	0.068
BUN, mmol/L (mean (SD))	5.22 (2.58)	8.82 (7.18)	0.021
UA, µmol/L (mean (SD))	214.46 (147.45)	266.02 (138.77)	0.155
Alb, g/L (mean (SD))	31.66 (5.69)	34.22 (5.60)	0.077
Hb, g/L (mean (SD))	98.46 (23.21)	103.98 (19.65)	0.301
Plt, *109/L (mean (SD))	56.42 (30.01)	74.64 (45.77)	0.083
WBC, *109/L (mean (SD))	4.57 (2.37)	6.13 (4.01)	0.085
Lym, *109/L (mean (SD))	0.82 (0.57)	0.89 (0.41)	0.537
Neu, *109/L (mean (SD))	3.03 (1.67)	4.53 (3.80)	0.071
Post-LT characters
Antibiotic type (n (%))			0.847
Beta-lactam	45 (100.0)	24 (100.0)
Engpolymyxin	12 (26.7)	4 (16.7)
Glycopeptides	40 (88.9)	20 (83.3)
Anti-fungal drugs	36 (80.0)	21 (87.5)
Antibiotic duration (mean (SD))	26.17 (19.90)	21.40 (8.91)	0.1737
C0 of tacrolimus after LT, µmol/L (mean (SD))
7 days	6.62 (3.19)	7.00 (2.80)	0.629
14 days	7.91 (3.27)	8.11 (2.60)	0.794
21 days	7.64 (2.97)	8.07 (3.39)	0.604

EAD early allograft dysfunction, BMI body mass index, HBV hepatitis B virus, ALT alanine transaminase, AST aspartate transaminase, TBil total bilirubin, DBil direct bilirubin, BA bile acid, INR international normalized ratio, Scr serum creatinine, BUN blood urea nitrogen, UA uric acid, Alb albumin, Hb hemoglobin, Plt platelet, WBC white blood cell, Lym lymphocyte, Neu neutrophil, C₀ trough blood concentrations, LT liver transplantation

Fig. 5.

Gut microbiome was also associated with EAD. A PCoA plot illustrating the distribution of groups, including the degree of explanation of PCs, distribution over the PCs, Observed, and Shannon. PCoA coordinates were calculated once using the full dataset and the same distance matrix. Colors indicate different grouping variables. B Differential taxon identified by the LEfSe between groups. C Volcano plot illustrating the differential metabolites. D Differential metabolic pathways between groups. EAD, early allograft dysfunction; ACLF, acute-on-chronic liver failure; HCC, hepatocellular carcinoma; PCoA, principal coordinates analysis; LEfSe, linear discriminant analysis effect size

Certain features were found to correlate with clinical indicators

To further examine the relationship between multi-omic features and clinical outcomes, simple linear regression was conducted between the abundance of specific taxon or metabolites and clinical parameters across patient groups. Numerous features showed significant associations (Fig. 6A), with several consistent correlations observed across subgroups—for example, between 1,4-dihydroxyheptadec-16-en-2-yl acetate and AST levels (Fig. 6B). Additionally, microbial and metabolic signatures associated with EAD, g__Lachnoclostridium, were closely linked to postoperative recovery, ALT levels, after LT (Fig. 6C).

Fig. 6.

Lachnoclostridium was found to correlate with clinical indicators. A Bar plot illustrating the number of related features in every subgroup. B Venn diagram showing the related features shared among the subgroups. C Differential features significantly correlated with the clinical indicators. ACLF, acute-on-chronic liver failure; HCC, hepatocellular carcinoma; ALT, alanine transaminase; AST, aspartate transaminase; TBil, total bilirubin; INR, international normalized ratio; Scr, serum creatinine

Discussion

In this study, we profiled gut microbial and metabolomic features across patients with ACLF, cirrhosis, and HCC during the early post-transplant period, revealing distinct multi-omic signatures associated with disease state and postoperative recovery. ACLF patients exhibited a unique beta diversity pattern, characterized by depletion of specific taxa, most notably g__Anaerostipes. Metabolomic analyses further identified significant alterations, including enrichment of tangeritin and depletion of candesartan. Integrative network analysis highlighted g__Anaerostipes as a central node linking differential microbial taxa and metabolites. In addition, multi-omic features, including g__Lachnoclostridium, were associated with EAD and correlated with key clinical indicators, suggesting a potential role for gut microbial–metabolic interactions in post-transplant recovery.

Gut-derived PAMPs and microbial dysbiosis in ACLF

In patients with ACLF, the gut serves as a primary source of PAMPs, which play a central role in triggering systemic inflammation and driving disease progression (Clària et al. 2016). Microbial products—such as lipopolysaccharide, peptidoglycan, and bacterial DNA—enter the portal circulation as a result of increased intestinal permeability (Kim et al. 2021), which is influenced by hypertension (Simbrunner et al. 2019), mucosal injury (Llorente and Schnabl 2015), systemic inflammation (Kronsten and Shawcross 2025), and alterations in the gut microbiome (Schroeder 2019). Therefore, characterization of the intestinal microbial spectrum in ACLF is critical for elucidating disease mechanisms and informing therapeutic interventions.

Role of g__Anaerostipes depletion in ACLF

In this study, g__Anaerostipes was identified as a characteristic genus in ACLF and functioned as a key node linking gut microbial and metabolomic features. Anaerostipes has been increasingly recognized as a beneficial genus in liver-related disorders. Mendelian randomization analysis suggests that Anaerostipes serve a potential protective role against intrahepatic cholangiocarcinoma (Chen et al. 2023), and its abundance is reduced in advanced hepatocellular carcinoma and aging populations (Huo et al. 2023; Zhang et al. 2023). As a core taxon in cirrhosis, Anaerostipes-mediated fatty acid synthesis has been associated with portal vein free fatty acid levels (Ali et al. 2023), intestinal barrier injury biomarkers (Efremova et al. 2024), and portal vein thrombosis (Huang et al. 2023). In rat models, treatments such as Yinchen Wuling powder have been shown to alleviate liver fibrosis while promoting Anaerostipes colonization (Zhang et al. 2021). Similarly, dietary fiber (Kovynev et al. 2025), exercise (Kovynev et al. 2025), conjugated linoleic acid (Gao et al. 2022), and Shenling Baizhu powder (Zhang et al. 2018) have been shown to increase the relative abundance of Anaerostipes in models of metabolic dysfunction-associated steatotic liver disease (MASLD), accompanied by reductions in body weight, adiposity, and plasma glucose levels.

Functionally, Anaerostipes is increasingly recognized as a probiotic due to its ability to produce butyrate, a short-chain fatty acid with multiple protective roles in intestinal and systemic homeostasis. First, butyrate serves as a major energy source for colonic epithelial cells, increasing oxygen consumption (Litvak et al. 2018) and maintaining an anaerobic environment that suppresses the growth of aerobic pathogens such as Salmonella and Escherichiacoli (Manson et al. 2008; Parada Venegas et al. 2019), reducing the risk of opportunistic infections. Moreover, microbial butyrate enhances macrophage-mediated pathogen clearance (Flemming 2019). Butyrate stabilizes hypoxia-inducible factor (HIF) in epithelial cells by promoting mitochondrial oxygen consumption or inhibiting HIF-prolyl hydroxylase (Wang et al. 2021), thereby strengthening the intestinal barrier via upregulation of claudin and MUC2 (Wang et al. 2021; Zheng et al. 2017). On the other hand, butyrate acts as a ligand for aryl hydrocarbon receptors and G protein–coupled receptors, mediating anti-inflammatory responses and preserving intestinal homeostasis (Marinelli et al. 2019; Yip et al. 2021), which limits PAMPs translocation during ACLF progression.

Consistent with recent integrative microbiome analyses in liver disease, depletion of butyrate-producing anaerobes has been linked to immune dysregulation and adverse clinical outcomes (Abenavoli et al. 2023). Collectively, loss of Anaerostipes in ACLF may exacerbate barrier dysfunction, amplify PAMP translocation, and impair post-transplant recovery.

Relevance of g__Lachnoclostridium to postoperative recovery

A strong association was identified between the gut microbiome and postoperative recovery, with EAD-associated g__Lachnoclostridium showing significant correlations with clinical indicators. Over the past decade, Lachnoclostridium has been extensively studied and linked to MASLD. Mendelian randomization analyses suggest that increasing the relative abundance of Lachnoclostridium may be protective against MASLD (Dai et al. 2023). In mouse or rat models of MASLD, substances such as bacterial cellulose (Han et al. 2024), N-acetylcysteine (Ding et al. 2022b), dihydroquercetin (Wang et al. 2022), Da Chaihu decoction (Cui et al. 2020), Dendrobium officinale (Tian et al. 2023), Qushi Huayu (QSHY) decoction (Ni et al. 2023), and sesamolin (Yu et al. 2022) have been shown to alleviate hepatic injury, accompanied by significant alterations in the abundance of Lachnoclostridium. These findings suggest that Lachnoclostridium may modulate lipid and carbohydrate metabolism, intervening in liver damage. Moreover, Lachnoclostridium has been shown to attenuate MASLD progression through a microbiota oscillation–dependent mechanism (Xia et al. 2023).

In addition to MASLD, Lachnoclostridium has been implicated in other liver diseases. In liver cancer patients, the colonization of Lachnoclostridium has been linked to responses to immunotherapy (Lee et al. 2022), likely through its effects on tertiary lymphoid structures (Zhao et al. 2023) and its activation of CD8 + T cells (Yu et al. 2024). Lactobacillus plantarum ZY08 alleviated alcohol-related hepatic steatosis, liver injury, and intestinal barrier dysfunction by restoring Lachnoclostridium abundance and reducing plasma endotoxin levels (Ding et al. 2022a). GalN/LPS-induced acute liver injury promoted Lachnoclostridium colonization (Liu et al. 2022). Moreover, Si-Wu-Tang (Xue et al. 2021) and Shaoyao Gancao decoction (Li et al. 2022) mitigated CCl₄-induced liver fibrosis, accompanied by changes in Lachnoclostridium. Alhagi-honey has been shown to alleviate heat stress–induced hepatic injury, potentially by modulating Lachnoclostridium levels in association with liver enzyme activity (Xu et al. 2024).

Recent evidence further highlights that gut microbiota composition influences metabolic inflammation and immune reprogramming after liver transplantation, underscoring the dynamic interplay between microbial ecology and graft outcomes (Jing and Jiang 2025). The bidirectional behavior of Lachnoclostridium across disease states and interventions suggests context-dependent functions, reflecting complex host–microbe interactions during liver injury and recovery.

Clinical implications and limitations

Our findings suggest that early post-transplant gut microbial and metabolic signatures may provide insights into ACLF pathophysiology and postoperative prognosis. However, several limitations of our study should be acknowledged. The small sample size, single-center design, and lack of external validation reduce the generalizability of our findings. Further validation of the association between unique gut microbiota and metabolomic profiles with ACLF and EAD is needed through basic experimental studies. Potential preoperative factors influencing the gut microbiota, such as aspects of clinical management, were not incorporated into the experimental design due to limited accessibility. Although patients received relatively uniform intravenous antibiotic regimens, these treatments may have substantially altered gut microbial composition. The potential interplay between immunosuppressive therapy and the gut microbial ecosystem warrants further investigation. Additionally, post-transplant microbiome characteristics do not fully replicate the pre-transplant state. The statistical power of subgroup analyses and the adjustment for confounding factors require further strengthening. Due to sample size limitations and the follow-up duration, the predictive value of early post-transplant microbiome features for long-term outcomes could not be assessed. Nevertheless, this study provides a multi-omic framework for understanding gut–liver crosstalk in ACLF and highlights potential microbial targets for improving transplant outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Y.Z, K.C and Y.M participated in research design, X.X, J.Z, Y.Z, K.C and Y.M participated in the writing of the paper, X.X, J.Z, J.J and P.D participated in the performance of the research, X.X, J.Z, J.J and P.D participated in data analysis.

Funding

This study was supported by the National Natural Science Foundation of China (grant number: 81771722 and 82570782).

Data availability

The sequences have been deposited under the accession numbers SRA PRJNA1418176.

Declarations

Ethics approval and consent to participate

All the LT recipients received the allografts from donation after citizen’s death (DCD). All the transplantations performed in our center were approved by the DCD Ethics Committee of the Third Xiangya Hospital, Central South University. The allograft was attributed by the China Organ Transplant Response System. The study protocol was approved by the Ethics Committee of the Third Xiangya Hospital of Central South University, Changsha, China (No. 22207). Written informed consent was obtained from all study participants. Experiments were carried out in accordance with the ethical guidelines set by the Declaration of Helsinki 1964 and its later amendments.

Competing interests

The authors declare no competing interests.