Mapping Antimicrobial Synergism in Sorbate-Based Ternary Preservative Systems Against Listeria innocua and Salmonella Typhimurium Using Multivariate Analysis

Designing multi-ingredient preservation systems is crucial for reducing the use of synthetic additives while maintaining food safety. This study evaluated the antimicrobial efficacy of ternary mixtures combining two natural antimicrobials (thymol, carvacrol, eugenol, citral, vanillin) and potassium sorbate (PS) against Listeria innocua and Salmonella Typhimurium at pH 4.5 and 5.5. The fractional inhibitory concentration index (FICI) and the total minimum inhibitory concentration (MIC) were determined to identify optimal synergistic combinations. Multivariate analysis, including Principal Component Analysis and Partial Least Squares (PLS) regression with standardized coefficients, was applied to decipher the relative impact and hierarchy of the predictor variables. At pH 4.5, concentrations of PS ≤ 64 ppm combined with any of the natural components resulted in higher synergistic mixtures (low FICI 0.318–0.378) for L. innocua. For Salmonella, only combinations of thymol, carvacrol, eugenol, and PS ≤ 32 ppm obtained the lowest FICI (0.252–0.344) at pH 4.5, while at pH 5.5, 128 ppm PS, thymol, and vanillin were required for a similar FICI (0.363). The PLS models revealed a distinct shift in variable importance between the synergistic index and the MIC. For the FICI model, mixture components exerted the primary influence; potassium sorbate displayed the highest predictive weight (standardized coefficient: +0.8736), followed by eugenol (+0.7757), carvacrol (+0.6206), and citral (+0.5119). This indicates that maximum synergism (lowest FICI) is constrained to lower fractional concentrations of natural compounds, thereby avoiding saturation of the cellular target site. Bacterial species (+0.2418) and pH (+0.3146) contributed less, indicating a homogeneous effect across the tested bacteria. For the total MIC model, potassium sorbate (+0.4701) and vanillin (+0.4492) regulated the antimicrobial quantity requirements. For the strains evaluated, the coefficients of the bacterial type in both models indicate that the ternary mixtures performed similarly. These findings demonstrate that integrating multivariate PLS modeling provides a robust framework for optimizing natural-synthetic antimicrobial blends, significantly reducing dependence on potassium sorbate through tailored synergism.