Fig. 1: Experimental design and functional screening of infant-derived Bifidobacterium strains.

a The experimental design comprises three successive phases: In vitro Screening: Twelve infant-derived strains were initially evaluated based on four criteria: dissociation ability, enzyme activity, tolerance, and growth kinetics. This process identified Bifidobacterium animalis subsp. lactis H22B656 as the original candidate strain. Host-mediated In vivo Adaptation: GF mice served as an evolution bioreactor. Following colonization with H22B656, mice were subjected to increasing dietary fat content (30% to 60%) over 7 weeks. Sixty colonies were isolated from feces at different time points. Based on the specific activity ratio, the adapted strain B. animalis subsp. lactis W5S9 was selected. NAFLD Efficacy Validation: The therapeutic potential was evaluated in SPF mice under a high-fat diet (HFAD). Mice were divided into four groups (Control, HFAD, HFAD + H22B656, and HFAD + W5S9; n = 6 mice per group) for an 8-week intervention. Efficacy was assessed via biochemistry, histology, targeted bile acid (BA) metabolomics, and metagenomics to compare the metabolic regulation capabilities of the original and adapted strains. b Calcium precipitation assay. Twelve candidate strains were cultured anaerobically on MRS-cys agar containing 0.3% (w/v) conjugated bile acids for 48 h. The bar plot shows the mean diameter of the calcium-precipitation halo for each strain. c Specific BSH activity measured by a ninhydrin-based colorimetric assay using glycine-conjugated bile acids as substrate. d Heatmaps of survival rates under acid (pH 2.0, 2 h) and bile acid (0.3% (w/v), 4 h) stress, calculated as OD₆₀₀₍treatment₎/OD₆₀₀₍control₎. Data are from n = 3 biologically independent experiments. e Growth kinetics. Strains were inoculated at 2% (v/v) into MRS-cys broth and incubated anaerobically at 37 °C for 36 h, with OD₆₀₀ recorded every 4 h. For all in vitro experiments (b–e), data are derived from n = 3 biologically independent bacterial cultures. Data in (b), (c), and (e) are presented as mean ± SD. Panels (a–e) were created with BioRender (Zhang, J. (2026) https://BioRender.com/3qyx3m3). Source data are provided as a Source Data file.