Clinical severity score-guided metagenomic analysis identifying gut microbial taxonomic and functional markers for severe hepatitis E progression

ABSTRACT Although gut microbiome alterations have been reported in hepatitis E (HE), the taxonomic and functional determinants of disease severity remain poorly defined. Here, we performed shotgun metagenomic sequencing of fecal samples from 125 individuals spanning acute non-icteric hepatitis (ANIH), acute icteric hepatitis (AIH), acute liver failure (ALF), and healthy controls. We constructed a s urrogate clinical severity index from routine blood parameters and developed an integrative analytical framework that combined XGBoost-based taxonomic modeling with LASSO-driven functional feature selection to explore microbiome-severity associations. The taxonomic model discriminated patients from controls (area under the curve [AUC] = 0.944) and identified nine bacterial species significantly associated with severity, independent of age, sex, and body mass index (BMI) (permutation test, P < 0.05), with Veillonella atypica emerging as the most robust candidate biomarker. The signature featured enrichment of lactate-utilizing Veillonella spp. and Ligilactobacillus salivarius , alongside depletion of beneficial commensals ( Dorea longicatena , Ruminococcus timonensis , Eubacterium ramulus ), collectively suggesting a pathogenic “lactate axis.” The HE-enriched lactate utilizers were positively associated with 13 core severity-increasing KOs involved in oxidative stress adaptation ( npr , hemQ , NUDT1 , nfr1 ) and secretion/biofilm formation ( gspD , fhaC , vpr , sinR ) and negatively with three severity-decreasing KOs, including the butyrate fermentation gene (K14534). Strikingly, the stringent 16-KO core (intersection of four methods) explained more variance in disease severity than the broader set identified by at least three methods ( R ² = 0.61 vs. 0.47). Collectively, these findings reveal a strong link between microbiome-derived lactate metabolism and HE severity, highlighting its potential as a basis for microbiome-based severity stratification and motivating further mechanistic exploration. IMPORTANCE Hepatitis E virus (HEV) infection ranges from ANIH to ALF, yet the role of the gut microbiome remains poorly understood across the severity spectrum. By performing shotgun metagenomic sequencing on fecal samples from 125 individuals and applying an integrative framework combining XGBoost-based taxonomic modeling with LASSO-driven functional selection, we identified robust microbiome–severity associations as quantified by a surrogate clinical severity index derived from routine blood parameters. These associations were characterized by enrichment of lactate-utilizing Veillonella species (notably Veillonella atypica ), depletion of beneficial butyrate-producing commensals, and severity-linked shifts in microbial functional gene profiles. Such patterns are consistent with a perturbed microbial “lactate axis” in gut–liver crosstalk, although all associations remain correlational. Our findings nominate candidate microbial markers for severity stratification in hepatitis E and provide a hypothesis-generating framework to guide future mechanistic studies and microbiome-based therapeutic strategies.