Background. Developing next-generation mechanism-based mathematical models (ngMBMs) that incorporate characteristics of bacterial strains as predictors of bacterial response to antibiotic treatment is essential to support model-informed precision dosing of antibiotics. This study aimed to develop a ngMBM that describes the time-courses of bacterial responses of 12 Pseudomonas aeruginosa (PA) strains with various resistance mechanisms treated with aztreonam (ATM), ciprofloxacin (CIP), and their combination.
Methods. The PAO1 wild-type reference strain and its 11 isogenic strains were investigated in 72-h static-concentration time-kill studies. Strains with single resistance mechanisms were: PAΔdacB, PAΔAD, PAΔADΔADh3 and PAΔADΔADh2ΔADh3 with AmpC β-lactamase hyperproduction caused by different mutations; PAΔmexR and PAΔmexZ with MexAB-OprM and MexXY-OprM efflux pump overexpression, respectively; and PAOD1 with reduced entry porins. The double mutants were: PAOD1ΔdacB, PAOD1ΔAD, PAΔADΔmexR and PAOD1ΔmexR. Clinically relevant CIP and ATM concentrations were studied. Total viable bacterial counts were determined. An ngMBM was developed to incorporate the resistance mechanisms and mechanisms of action of the antibiotics. Model evaluation included population-predicted fits, standard diagnostic plots, the Akaike information criterion, and biological plausibility of parameter estimates.
Results. All monotherapies failed (Figure 1). Synergy at 72h (≥2-log10 lower bacterial count than the most effective monotherapy) was observed with ATM 25 mg/L + CIP 0.5 mg/L for 9 of the 12 strains, with mutants containing the mexR deletion being the exception. PAΔmexR and PAOD1ΔmexR were synergistically killed by ATM 25 + CIP 1.0 mg/L, whilst PAΔADmexR displayed additive killing. The population fits of the final ngMBM well described the time-courses of bacterial response for all 168 experimental arms from 12 strains simultaneously, without estimating strain-specific drug effect parameters. The effect of ATM was best described by inhibition of both successful replication and growth, and that of CIP by direct bacterial killing, reflecting their mechanisms of action. The combination therapy effect was best modelled by subpopulation synergy. Incorporating the resistance mechanisms into the ngMBM was essential to differentiate the population fits. Resistance mechanisms were modelled via modified antibiotic degradation and influx into or efflux out of the cell compared to the wild-type, consistent with their biological effects. Combining the effects of the single mutations well described the double mutants in the model.
Conclusion. Combination therapy of ATM and CIP successfully killed bacteria and inhibited regrowth in PAO1 and 10 of the 11 mutant strains where monotherapies failed to do so. Integrating the resistance mechanisms into the ngMBM was crucial to accurately describe the responses of isogenic strains with a wide range of resistance mechanisms to antibiotic treatment.