Greater Ciprofloxacin Tolerance as a Possible Selectable Phenotype Underlying the Pandemic Spread of the H30 Subclone of Escherichia coli Sequence Type 131

Abstract

Minimum bactericidal concentrations (MBCs) for ciprofloxacin were significantly higher among 41 members of the H30 subclone within Escherichia coli sequence type 131 than among 48 other fluoroquinolone-resistant E. coli isolates. This MBC difference, which was not explained by ciprofloxacin MICs, gyrA, parC, and parE mutations, the presence of aac(6′)-Ib-cr, or organic solvent tolerance (a surrogate for efflux pump activity), conceivably could have promoted the pandemic emergence of the H30 sequence type 131 subclone.

TEXT

Escherichia coli is clonally diverse and increasingly antimicrobial resistant (1). The H30 subclone within E. coli sequence type 131 (ST131) has emerged inexplicably in many locales as the dominant E. coli lineage among human clinical isolates, including those resistant to fluoroquinolones or extended-spectrum cephalosporins (2–9).

We recently documented higher fluoroquinolone MICs for the ST131 H30 subclone (here, H30) isolates than for other fluoroquinolone-resistant E. coli isolates (10). Here, we assessed whether H30 isolates also have higher minimum bactericidal concentrations (MBCs) for ciprofloxacin and explored the associations between the ciprofloxacin MBCs, MICs, and fluoroquinolone resistance mechanisms.

Isolates.

The 89 fluoroquinolone-resistant (i.e., ciprofloxacin nonsusceptible) E. coli study isolates were from our study of fluoroquinolone MICs in relation to clonal background and resistance-associated traits (10). They included 41 H30 and 48 non-H30 E. coli isolates and were selected from larger source collections for diversity of gyrA and parC alleles, multilocus sequence types, and pulsed-field gel electrophoresis profiles (3, 7, 11, 12). The non-H30 group included 31 STs from 20 different ST complexes (as determined by eBURST software) and 4 non-H30 ST131 strains.

MICs and MBCs.

The ciprofloxacin MICs were determined by broth microdilution using standard methods (13), with doubling dilutions of ciprofloxacin (range, 0.5 to 2,048 mg/liter) in cation-adjusted Mueller-Hinton broth, plus turbidity-adjusted suspensions of each test isolate (0.5 McFarland standard; approximately 1 × 10⁸ CFU/ml). The MIC for a given titration series was the lowest ciprofloxacin concentration yielding no visible growth after overnight incubation.

For MBC determination, aliquots from each overnight MIC dilution series underwent quantitative plating to antibiotic-free agar. The MBC was the lowest ciprofloxacin concentration that yielded a ≥99.9% decrease in viable counts (14).

For each isolate, the MIC and MBC were determined initially in duplicate. If the duplicates disagreed by >2-fold for any result, up to 5 total determinations were done. The geometric mean value was used.

Resistance-associated traits.

The presence of aac(6′)-Ib-cr (encoding a fluoroquinolone- and aminoglycoside-modifying enzyme) (15), mutations in the quinolone resistance-determining region (QRDR) of gyrA, parC, and parE (16, 17) and the organic solvent tolerance (OST) (a proxy for efflux pump activity) (18) were determined previously (10).

Statistical analysis.

Comparisons involving continuous variables were tested using the Mann-Whitney U test. The correlation was analyzed by simple regression.

Distribution of ciprofloxacin MBCs versus resistance-associated traits in H30 and non-H30 isolates.

The 89 study isolates exhibited a broad range of ciprofloxacin MBCs (Fig. 1). Compared with the non-H30 isolates, the H30 isolates had significantly higher MBCs (median, 4-fold; P < 0.001). For both MICs and MBCs, the four non-H30 ST131 strains resembled more closely the non-ST131 strains than the H30 strains (not shown).

FIG 1.

Ciprofloxacin minimum bactericidal concentration (MBC) in relation to H30 subclone status and resistance mechanisms. (A) H30 versus non-H30 isolates. (B) Subgroups based on H30 status and the presence of aac(6′)-Ib-cr (aac) and/or nonsynonymous mutations at parE458. The circles indicate individual isolates. The horizontal bars show group medians (wide bars) and the 25th and 75th percentiles (narrow bars). P values (two-tailed) are by the Mann-Whitney U test. NS, not significant.

To identify the mechanistic correlates, the MBCs were compared statistically with the fluoroquinolone resistance mechanisms. All 41 H30 strains carried gyrA and parC QRDR mutations, and 10 of these additionally carried aac(6′)-Ib-cr. None carried a parE458 mutation. In contrast, of the 48 non-H30 strains, 27 carried gyrA and parC QRDR mutations only, 21 carried a parE458 mutation (in addition to gyrA and parC QRDR mutations, except in 1 strain), and 4 carried aac(6′)-Ib-cr (3 with gyrA and parC mutations only and 1 with gyrA, parC, and parE458 mutations).

aac(6′)-Ib-cr was associated with significantly higher MBCs among the H30 isolates (P = 0.04), and a similar trend was exhibited among the non-H30 isolates (Fig. 1B). Among the aac(6′)-Ib-cr-positive isolates, the MBCs did not differ significantly between the H30 and non-H30 isolates. Among the aac(6′)-Ib-cr-negative non-H30 isolates, the presence of a parE458 mutation was also associated with a higher MBC (P = 0.04). Nonetheless, among the aac(6′)-Ib-cr-negative isolates, the MBCs still were significantly higher among the H30 than the non-H30 isolates, whether the latter lacked or contained parE458 mutations (P = 0.004 and P < 0.001, respectively).

Thus, even in the absence of aac(6′)-Ib-cr, H30 strains have significantly higher MBCs than non-H30 strains, despite lacking parE458 mutations. This indicates a possible MBC-determining role for the unique combination of QRDR mutations and/or other factors in H30 strains.

Correlation between MBCs and MICs in H30 and non-H30 strains.

For more closely matched comparisons of MICs and MBCs in H30 versus non-H30 isolates, analyses were restricted to the 55 isolates (24 H30 and 31 non-H30) with QRDR replacement mutations involving only gyrA and parC (not parE458) and without aac(6′)-Ib-cr. Within this population, in each subset (H30 and non-H30), MBCs and MICs were correlated only moderately (Fig. 2). Interestingly, however, although the two regression curves were nearly parallel, the H30 curve was shifted significantly upward to higher MBCs. This shows that at a given MIC, the MBCs tended to be much higher among the H30 than the non-H30 isolates. An even weaker correlation between the MBCs and MICs was observed when outliers with very low MICs/MBCs were excluded (data not shown).

FIG 2.

Ciprofloxacin minimum bactericidal concentrations (MBCs) in relation to the MICs among 89 ciprofloxacin-resistant Escherichia coli isolates. H30 subclone members and non-H30 isolates are shown separately. Isolates containing aac(6′)-Ib-cr (aac) and/or nonsynonymous mutations at parE458 were excluded.

The OST score was associated positively with the MIC, both overall and among the H30 and non-H30 isolates separately (Table 1). However, it was associated with the MBC only among the non-H30 isolates.

TABLE 1.

P values for associations of MICs and MBCs with organic solvent tolerance among ciprofloxacin-resistant Escherichia coli isolates

Parameter	P valueb
	Total (n = 89)	Non-H30 (n = 48)	H30 (n = 41)
MIC	0.006	0.007	0.01
MBCa	NS	0.04	NS

^aMBC, minimum bactericidal concentration.

^bP values (by Pearson correlation) are shown where the P value is <0.05. NS, not significant (P ≥ 0.05).

Thus, although there were modest correlations between the MBCs and MICs, the sizeable MBC differences between the H30 and non-H30 isolates could not be explained by their MIC differences, indicating that the MBC is a separate phenotype from the MIC.

Comment.

We found that, compared with fluoroquinolone-resistant non-H30 isolates, the fluoroquinolone-resistant H30 isolates typically have approximately 4-fold higher MBCs that are not explained by their corresponding MICs. This identifies a novel fitness mechanism that may underlie the striking epidemiological success of the ST131 H30 lineage.

Our findings suggest that even when inhibited by ciprofloxacin, H30 strains are more likely than other fluoroquinolone-resistant E. coli isolates to survive and regrow once the ciprofloxacin concentrations drop below the MIC. This tolerance phenomenon may contribute to the H30 strain-associated clinical scenario of same-strain recurrent urinary tract infection, with the causative organism disappearing during ciprofloxacin therapy and then reemerging after treatment completion (19, 20; J. R. Johnson, unpublished data).

Regarding mechanisms, our data suggest that aac(6′)-Ib-cr may differentially raise the MBC, which could explain why aac(6′)-Ib-cr is being selected in strains with preexisting QRDR mutations (10). Additionally, parE458 replacement mutations were associated with increased MBCs. However, neither aac(6′)-Ib-cr carriage nor parE458 (absent in H30 strains) (10) could explain the H30 versus non-H30 MBC gap. Likewise, the OST, which appeared to have less impact on the MBC than on the MIC, was associated with MBCs only marginally (non-H30 isolates) or not at all (total population and H30 isolates). Thus, to the extent that OST reflects the net efflux pump activity, efflux pumps also are unlikely to underlie the higher MBCs of the H30 isolates. As such, the mechanism(s) for higher MBCs among H30 isolates remains undefined.

Antimicrobial tolerance, which was studied initially mostly in relation to beta-lactam agents and staphylococci, conceivably could result from technical artifacts (21). Our findings are unlikely to represent technical artifacts, since we noted strong associations of the MBCs with the H30 subclone, independently from the MICs, when the H30 and non-H30 isolates were tested in parallel. Studies of quinolone tolerance in E. coli (22–25) have identified SOS-regulated genes (25), hipA (24), and defects in DNA repair (22) as possible contributors; these warrant future assessment in H30 isolates.

The study limitations include the small numbers in certain subgroups, the reliance on the OST as a proxy for efflux pump activity (versus measuring expression of multiple specific pumps), and the absence of attention to porin expression or to MBCs and the MIC/MBC discrepancy for drugs other than ciprofloxacin. The strengths include the well-characterized study population and the analysis of resistance mechanisms in relation to MBCs.

In summary, we newly documented higher ciprofloxacin MBCs for members of the H30 subclone of ST131, compared with those for other fluoroquinolone-resistant E. coli isolates. These phenotypes could not be explained by differences in the MICs, gyrA, parC, and parE mutations, or OST or the presence of aac(6′)-Ib-cr. Our findings identify a selectable phenotype that may have contributed to the emergence and pandemic spread of H30.