OBJECTIVES: Does exposure to subinhibitory concentrations of quinolones favour overexpression of efflux pumps or selection of target site mutations? METHODS: ATCC 49,619 (fully susceptible) and SP32 (clinical isolate with PmrA-mediated efflux and mutation in ParE) were exposed for 24 h in broth to ciprofloxacin, levofloxacin, moxifloxacin or garenoxacin at concentrations of 0.5x the MIC, with daily re-adjustments for up to 13 days. Efflux was detected phenotypically (decrease in MIC in the presence of reserpine), and expression of pmrA and patA/patB was measured by real-time PCR and comparative RT-PCR, respectively. Target site mutations were detected by sequencing of the quinolone resistance determining regions in parC, parE and gyrA. The clonal identity of isolates was checked by PFGE of genomic DNA. RESULTS: Ciprofloxacin selected for stable mutants with 2.5-5-fold MIC increases for ciprofloxacin, 2-3-fold for levofloxacin and 1.3-2-fold for garenoxacin and moxifloxacin [partial reversion with reserpine for ciprofloxacin, gemifloxacin and levofloxacin (SP32 strain only), but not for garenoxacin and moxifloxacin]. Increased MICs were associated with overexpression of patA/B but not pmrA. In contrast, exposure to levofloxacin, moxifloxacin or garenoxacin selected target site mutations (gyrA, parC, parE) in both strains. Increases in MIC caused by efflux were similar to those caused by target site mutations. CONCLUSIONS: Exposure of Streptococcus pneumoniae to subinhibitory MICs of ciprofloxacin, a substrate for efflux pumps, results in patA/B-mediated efflux whatever the initial level of expression of pmrA of the strain. Quinolones that are poorly (levofloxacin) or not affected (moxifloxacin, garenoxacin) in their activity by efflux transporters preferentially select for target site mutants.