Changes in qnr Prevalence and Fluoroquinolone Resistance in Clinical Isolates of Klebsiella pneumoniae and Enterobacter spp. Collected from 1990 to 2005

Abstract

Clinical isolates of Klebsiella pneumoniae and Enterobacter spp. collected from 1990 through 2005 at a tertiary care center were studied for qnr genes. Isolates bearing these genes emerged in the mid-1990s, coinciding with the time of a rapid increase in fluoroquinolone resistance. Sixty percent of these isolates were ciprofloxacin susceptible by CLSI breakpoints.

Plasmid-mediated qnr genes confer low-level quinolone resistance that is below the Clinical and Laboratory Standards Institute (CLSI) nonsusceptibility breakpoint but substantial enough to facilitate the selection of chromosomal mutations that confer higher-level resistance (4, 5). qnr has therefore been hypothesized to be a potential contributor to the increase in the prevalence of quinolone resistance among gram-negative bacteria. Epidemiological surveys have found qnrA, qnrB, and qnrS in various Enterobacteriaceae (9). These surveys generally have been performed with outbreak strains or isolates collected over a short period. Hence, little is known about the epidemiological patterns of these genes in a general clinical population over time. We therefore surveyed bloodstream isolates of Enterobacter spp. and Klebsiella pneumoniae collected over a 16-year period for qnr genes in order to more broadly characterize the epidemiology of these resistance elements in a clinical population.

(This work was presented in part at the 46th Interscience Conference on Antimicrobial Agents and Chemotherapy, San Francisco, CA, 2006.)

Methods.

All patient-unique bloodstream isolates had been prospectively collected at Hadassah Ein-Kerem Hospital, Jerusalem, since 1990 and kept at −70°C. For this study, we used Enterobacter spp., including Enterobacter cloacae and Enterobacter aerogenes, and K. pneumoniae and screened all available isolates from selected years from 1990 through 2005. Our selection strategy was to test all isolates from the first year for which all isolates were available (1990 for Enterobacter spp. and 1991 for K. pneumoniae) until the first year that qnr genes were detected. This allowed an accurate determination of the time at which qnr emerged in these populations. The qnr prevalence after emergence was determined by testing all isolates from selected subsequent years; these years were chosen based on convenience and complete isolate availability. All tested isolates were included in the final analysis. For E. cloacae, the following years were chosen: 1990 to 1995, 1997, 1999, 2001, 2003, 2004, and 2005 (for a total of 462 strains). We used all 117 available E. aerogenes isolates collected from 1990 to 2005. The 679 K. pneumoniae strains screened were isolated from 1991 to 1996, 1998, 2003, and 2005. Susceptibility for all patient-unique blood culture isolates recovered during the periods indicated for Enterobacter and Klebsiella were obtained from clinical laboratory records. Susceptibility testing throughout this time was performed according to CLSI (formerly NCCLS) standards. There was no change in the method of testing for resistance to ciprofloxacin or ceftazidime (CAZ), or in the breakpoints used, over the study period. Extended-spectrum β-lactamase (ESBL) production was determined according to the CLSI double-disk synergy test and the Etest ESBL (1). Screening was carried out by multiplex PCR amplification of qnrA, qnrB, and qnrS as previously described (11), except that the DNA polymerase was Taq DNA polymerase (New England Biolabs, Beverly, MA), and PCR conditions were 94°C for 30 s, 52°C for 30 s, and 72°C for 30 s, cycled 30 times.

Statistical methods.

Fisher's exact test was used to compare qnr prevalences before and after 1 January 1994 for Enterobacter spp. and before and after 1 January 1996 for K. pneumoniae. The Poisson regression analysis was used to compare the proportional ciprofloxacin resistances (defined by a MIC of ≥2) (1) prior to and after the first detection of qnr genes for each species. The logarithm of the number of isolates was used as the offset, and the Pearson chi-square statistic was used as the scaling factor to adjust for overdispersion. SAS statistical analysis software (SAS, Inc., Cary, NC) was used for calculations.

Enterobacter spp.

From 1990 through 1993, 1 out of 191 Enterobacter sp. isolates (0.5%) was ciprofloxacin resistant. From 1994 through 2005, 61 out of the 786 Enterobacter blood isolates (7.8%) were ciprofloxacin resistant (P < 0.01) (Fig. 1). From 1990 through 1993, none of the 94 isolates had qnr, and from 1994 onwards, 33 out of 485 (6.8%) isolates had qnr (P < 0.01) (Fig. 1). The number of isolates found harboring qnrA, qnrB, and qnrS were 8, 19, and 6, respectively; 23 (70%) were fluoroquinolone susceptible by CLSI criteria (Table 1). qnrA was more prevalent until 2001; since then, qnrB predominated.

FIG. 1.

Percentages of Enterobacter sp. isolates with the qnr gene and ciprofloxacin resistance. No qnr genes were found between 1990 and 1993.

TABLE 1.

Ciprofloxacin susceptibilities of qnr-bearing isolates

Gene carried by indicated isolate(s)	No. of isolates with indicated ciprofloxacin susceptibility characteristica
	Susceptible	Intermediate or resistant
Enterobacter spp.
qnrA	3	5
qnrB	16	3
qnrS	4	2
K. pneumoniae
qnrA	1	2
qnrB	2	4
qnrS	0	1
Total	26 (60)	17 (40)

^aAccording to CLSI criteria (1). Values in parentheses are percentages.

Pulsed-field gel electrophoresis after restriction with XbaI was performed according to a standardized protocol (8) on 10 randomly selected qnrB⁺ samples. The band pattern was analyzed according to accepted criteria (13). Only two strains (Fig. 2, lanes 3 and 4), isolated 15 months apart, were related.

FIG. 2.

Pulsed-field gel electrophoresis of 10 selected Enterobacter isolates after restriction with XbaI on 10 qnrB⁺ samples (Salmonella enterica serotype Braenderup provided the size marker). The dates of collection (month/year) and lane numbers are shown on the right, and molecular sizes in kilobases are at the bottom.

K. pneumoniae.

The prevalence of ciprofloxacin resistance among K. pneumoniae isolates was higher than in Enterobacter spp., but the patterns of change in resistance prevalence were similar. Between 1991 and 1995, 68 out of 525 isolates (12.9%) were ciprofloxacin resistant. A steep increase in ciprofloxacin resistance occurred in 1996. From 1996 through 2005, 412 out of the 1,156 isolates (35.6%) were ciprofloxacin resistant (P < 0.01) (Fig. 3). Of the 679 K. pneumoniae bloodstream isolates screened, 10 carried qnr. The three qnrA genes were found in isolates from 1996 and 1998. Six qnrB genes and a single qnrS gene were found in 2003 and 2005. Three of the 10 isolates harboring these genes were ciprofloxacin susceptible by CLSI criteria (Table 1). There was a significant difference in the numbers of qnr-positive strains between the two time periods, 0 out of 391 strains from 1991 to 1995 and 10 (3.5%) out of 288 in 1996 to 2005 (P < 0.01) (Fig. 2). Therefore, the appearance of qnr in both Enterobacter spp. and K. pneumoniae coincided with the period of rapid increase in ciprofloxacin resistance among these organisms.

FIG. 3.

Percentages of K. pneumoniae isolates with the qnr gene and ciprofloxacin resistance. No qnr genes were found between 1991 and 1995.

It has been suggested that plasmid-mediated quinolone resistance genes could, in part, account for the strong linkage between plasmid-mediated ESBL resistance and quinolone resistance (7, 12). To test for coresistance, we examined the prevalence of CAZ resistance among qnr⁺ and qnr-deficient Klebsiella and Enterobacter isolates. The relative risk for CAZ resistance in qnr⁺ K. pneumoniae isolates was 1.8 (95% confidence interval [CI], 1.3 to 2.5); in Enterobacter isolates it was 3.5 (95% CI, 2.7 to 4.5). It is important to determine whether these differences in CAZ resistance arise from a truly independent association between qnr genes and CAZ resistance genes or are simply a by-product of the association of qnr with fluoroquinolone resistance, which itself may be associated with CAZ resistance for reasons unrelated to qnr genes. To address this concern, we stratified the qnr-tested isolates by quinolone susceptibility (Table 2). The association between CAZ resistance and qnr was no longer significant in K. pneumoniae but was still strong in fluoroquinolone-susceptible Enterobacter spp. Fluoroquinolone-susceptible Enterobacter spp. carrying a qnr gene were 3.4 times more likely to be CAZ resistant than those not carrying qnr, suggesting the possibility of cotransmission of these resistance elements in this genus. Because CAZ resistance in Enterobacter spp. is often due to the derepressed expression of chromosomal AmpC β-lactamase (6), we examined the qnr-positive CAZ-resistant Enterobacter strains for the ESBL phenotype. Of the 24 isolates, 16 (66.7%) were ESBL positive. Because AmpC β-lactamases can mimic the phenotype of ESBL (3), we matched each strain with a qnr-negative CAZ-resistant isolate from the same year. Of these control strains, only five (22.8%) were ESBL positive (P < 0.01). This suggests that CAZ resistance in qnr-positive Enterobacter strains was associated with a true ESBL-mediated mechanism. Thus, the epidemiologic association between qnr and CAZ resistance in Enterobacter could plausibly be a reflection of a genetic linkage between these resistance elements on plasmids. Further work is required to examine this hypothesis.

TABLE 2.

Distribution of ciprofloxacin and CAZ susceptibilities among qnr⁺ and qnr-deficient isolates^a

Susceptibility characteristic of indicated isolate(s)	No. of isolates with indicated characteristic				Relative risk	95% CI
	qnr+		qnr-deficient
	CAZ-R	CAZ-S	CAZ-R	CAZ-S
K. pneumoniae
FQ-S	1	2	134	309	1.1	0.2-5.5
FQ-I/R	7	0	131	18	1.1	0.9-1.3
Total	8	2	265	327	1.8	1.3-2.5
Enterobacter spp.
FQ-S	15	8	95	397	3.4	2.4-4.8
FQ-I/R	9	1	16	3	1.1	0.8-1.4
Total	25	8	111	400	3.5	2.7-4.5

^aFQ, fluoroquinolone; S, susceptible; I, intermediate; R, resistant.

Conclusion.

Ciprofloxacin was introduced into the Hadassah hospitals in September 1989. The prevalence of quinolone resistance among the two taxa studied increased little until the mid-1990s, at which point it rapidly increased. Notably, this increase coincided with the entry of qnr genes into the bacterial population. This penetration appears to have occurred in multiple clones. A causal link between qnr genes and increased fluoroquinolone resistance cannot be proven by such a temporal relationship. It is possible, for example, that the increase in fluoroquinolone resistance drove the increase in qnr prevalence by providing a selective advantage to qnr plasmids. It is also possible that the emergence of qnr in this time frame was coincidental. However, the demonstration in vitro (5, 10) that qnrA strongly facilitates the selection of high-level chromosomal fluoroquinolone resistance in wild-type Enterobacteriaceae supports the suggestion that qnr genes have contributed to the emergence of fluoroquinolone resistance in this population. While the prevalence of these genes is still low, qnr plasmids may be unstable in Enterobacteriaceae (14); the proportion of isolates in which these genes are found may thus underrepresent their true contribution to the emergence of resistance. Still, the emergence of qnr cannot account for the entire increase in resistance in this population.

As has been demonstrated previously (2), a substantial fraction of the qnr-carrying isolates were susceptible to fluoroquinolones according to CLSI criteria. Indeed, 18 (72%) of 25 qnrB-bearing isolates were classified as ciprofloxacin susceptible. Given the potential of these strains for developing resistance, it is not clear that the current breakpoints adequately reflect the expected patient outcomes. Also sobering is the epidemiologic association demonstrated here between qnr genes and apparent ESBL resistance in Enterobacter spp., suggesting that qnr genes and certain ESBLs are frequently cotransmitted and thus coselected.