Anti-Listeria mechanisms, safety, and predictive modeling of Lacticaseibacillus casei UTMB9: Probiotic profiling targeting virulence gene expression and biofilm formation

Abstract

There is growing interest in probiotics due to their potential to confer health benefits. This study aimed to explore the potential probiotic characteristics, gene expression linked to biofilm formation, and anti-biofilm properties of Lacticaseibacillus casei UTMB9. L. casei UTMB9 tolerates acidic environments, with survival rates of 6.90, 7.59, and 7.96 log CFU/mL at pH 2.5, 3.5, and 4.5, respectively, and exhibits slight growth inhibition under bile concentrations up to 0.7 %. Viability declined modestly from 8.51 to 6.61 log colony forming unit (CFU)/mL under simulated gastrointestinal conditions. Surface hydrophobicity (40.89 %), auto-aggregation (30.9 %), co-aggregation (40.81 %), and adhesion (11.89 %) support its adherence potential. Antimicrobial assays showed most potent inhibition against Listeria monocytogenes (9.70 mm) versus E. coli (4.74 mm). Scanning Electron Microscopy (SEM) imaging revealed pronounced damage to L. monocytogenes cells after exposure to cell-free supernatant (CFS), accompanied by significant downregulation of prfA and flaA, and marked antibiofilm suppression. It also demonstrated antioxidant activity (2,2-diphenyl-1-picrylhydrazyl (DPPH) 43.58 %, 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) 47.47 %, linoleic acid oxidation inhibition 30.28 %) and moderate anticancer activity (IC₅₀ ≈ 55–62 mg/mL). The strain reduced cholesterol uptake by 41.9 %, was broadly antibiotic-sensitive (except for partial resistance to ampicillin), and lacked biogenic amine production, DNase, or hemolytic activity. In the second part of this study, Gaussian Process Regression (GPR) was used to predict acidity and bile salt. GPR accurately predicted acidity and bile tolerance (MAPE = 0.22 % and 0.18 %; R² ≥ 0.99). These findings position L. casei as a promising probiotic agent with robust antimicrobial, antibiofilm, antioxidant, and predictive model-supported features.