Toxins produced by community-associated methicillin-resistant Staphylococcus aureus (CA-MRSA) contribute to virulence. We developed a statistical approach to determine an optimum sequence of antimicrobials to treat CA-MRSA infections based on an antimicrobial's ability to reduce virulence. In an in vitro pharmacodynamic hollow fiber model, expression of six virulence genes (lukSF-PV, sek, seq, ssl8, ear, and lpl10) in CA-MRSA USA300 was measured by RT-PCR at six time points with or without human-simulated, pharmacokinetic dosing of five antimicrobials (clindamycin, minocycline, vancomycin, linezolid, and trimethoprim/sulfamethoxazole (SXT)). Statistical modeling identified the antimicrobial causing the greatest decrease in virulence gene expression at each time-point. The optimum sequence was SXT at T0 and T4, linezolid at T8, and clindamycin at T24-T72 when lukSF-PV was weighted as the most important gene or when all six genes were weighted equally. This changed to SXT at T0-T24, linezolid at T48, and clindamycin at T72 when lukSF-PV was weighted as unimportant. The empirical p-value for each optimum sequence according to the different weights was 0.001, 0.0009, and 0.0018 with 10,000 permutations, respectively, indicating statistical significance. A statistical method integrating data on change in gene expression upon multiple antimicrobial exposures is a promising tool for identifying a sequence of antimicrobials that is effective in sustaining reduced CA-MRSA virulence.
Keywords: Staphylococcus aureus; antimicrobials; hollow fiber model; mathematical modeling; virulence.