Nutritional matrix shifts the temperature-salinity response of biofilm formation by Salmonella Heidelberg and Listeria monocytogenes on stainless steel.

Food residues on processing surfaces may modify microbial responses relative to standardized laboratory media, potentially limiting model transferability. This study evaluated viable biofilm-associated populations of Salmonella Heidelberg and Listeria monocytogenes on AISI 304 stainless steel under combined NaCl (1.09-3.91%) and temperature (6.88-42.12 °C) conditions. A Central Composite Rotatable Design was conducted in Brain Heart Infusion broth (BHI) and chicken juice, with 22 independent observations for each pathogen-matrix combination. Complete second-order polynomial models were fitted and evaluated using analysis of variance, lack-of-fit tests, residual diagnostics, and heteroscedasticity-consistent inference. Model performance varied across pathogen-matrix combinations, with R2 values of 0.767 and 0.828 for S. Heidelberg in chicken juice and BHI, respectively, and 0.507 and 0.718 for L. monocytogenes. Significant lack of fit was detected for S. Heidelberg in chicken juice and L. monocytogenes in BHI. Cross-matrix transfer of BHI-derived models to chicken juice showed no significant directional bias, but prediction errors remained substantial, with RMSE values of 0.94 and 1.10 log CFU/cm2 for S. Heidelberg and L. monocytogenes, respectively. Response-surface structure differed significantly between matrices for S. Heidelberg (p = 0.0016), whereas evidence was weaker for L. monocytogenes (p = 0.071). These findings demonstrate matrix-dependent responses and indicate that relationships obtained in standardized media may not be directly transferable to food-derived matrices.