Clinical isolates of E. coli, Klebsiella pneumoniae, and K. oxytoca were obtained from 19 US hospitals during phase IV (2002–2004) of Project ICARE, which focused on nonoutbreak isolates of Enterobacteriaceae with decreased susceptibility to extended-spectrum cephalosporins, fluoroquinolones, or carbapenems. The 19 hospitals were located in 13 states throughout the country and had an approximately equal mix of teaching and nonteaching medical centers. Duplicate isolates from the same patient were excluded.

Isolates were tested for susceptibility to the antimicrobial drugs listed in Table 1 by using the broth microdilution reference method (BMD) described by the Clinical and Laboratory Standards Institute (7). Isolates were stored at –70°C and were subcultured to trypticase soy agar plates containing 5% defibrinated sheep blood (BD BioSciences, Sparks, MD, USA) >2 times before testing. For each organism, BMD tests were inoculated by using a cell suspension equivalent to a 0.5 McFarland standard. Enterobacteriaceae isolate identifications provided by participating laboratories were confirmed by colony shape, spot tests (8), and Vitek GNI+ cards (bioMérieux, Durham, NC, USA). Differences were resolved by using reference biochemical tests (9).

Isolates were analyzed by polymerase chain reaction (PCR) amplification techniques to determine whether a class 1 integron was present. Integrons were detected by PCR amplification of a class 1 integrase–specific fragment of the IntI1 gene as previously described (1). The primer sequences used were IntI1-F: 5´-TCT CGG GTA ACA TCA AGG-3´ and IntI1-R: 5´-AGG AGA TCC GAA GAC CTC-3´.

Amplifications were performed in 10 μL of Taq PCR Master Mix (Qiagen, Valencia, CA, USA), 1.5 mmol/L MgCl₂, 5 pmol/L each primer, and 1 μg template DNA (1). DNA was extracted with a Genelite Bacterial Genomic DNA Kit (Sigma, Saint Louis, MO, USA). Amplification specifications were as follows: 5 min at 94°C followed by 35 cycles of 1 min at 94°C, 1 min at 55°C, and 30 s at 72°C. PCR products were analyzed by gel electrophoresis with 1.5% agarose gels. All PCRs included positive and negative controls.

A statistical comparison of the frequencies of integron presence in E. coli and Klebsiella spp. was conducted by using odds ratios and 95% confidence intervals. Intermediate and resistant isolates were pooled as nonsusceptible for analysis. Multivariate logistic regression analysis was conducted with 1 representative from each major class of drugs tested. Data validation and analysis were performed by SAS statistical software version 9.1 (SAS Institute, Cary, NC, USA). Fit of the data to this model was evaluated by the Hosmer-Lemeshow χ² test.

A positive test result for class 1 integrons was found for 181 (57%) of the 320 bacterial isolates screened, including 103 (49%) of 209 E. coli isolates and 78 (70%) of 111 Klebsiella spp. isolates. A positive test result in E. coli isolates was significantly associated with nonsusceptibity to gentamicin, tobramycin, ciprofloxacin, chloramphenicol, and trimethoprim-sulfamethoxazole (Table 1). A positive test result in Klebsiella isolates was significantly associated with nonsusceptibility to gentamicin, tobramycin, cefpodoxime, ceftazidime, and trimethoprim-sulfamethoxazole.

Multivariate analysis showed that the only drug variables associated with a positive test result for class 1 integrons in E. coli were nonsusceptibility to gentamicin and trimethoprim-sulfamethoxazole (with a strong interaction between the 2 variables), while an inverse association was present between a positive integron test result and nonsusceptibility to cefepime (Table 2). For Klebsiella spp., a positive association existed between a positive integron test result and nonsusceptibility to gentamicin and trimethoprim-sulfamethoxazole, as well as an inverse association between a positive integron test result and nonsusceptibility to aztreonam.

More than half of the selected isolates from US hospitals that we tested were positive for class 1 integrons. The prevalence of class 1 integrons in E. coli was 49% in selected non-outbreak isolates from hospitalized patients from 2002 to 2004 in our study, 52% in consecutive urinary tract isolates in 2001 in Sweden (10), and 15% in isolates from blood in hospitals in Sweden in 1998 and 1999 (11). The prevalence of integrons was 70% in Klebsiella isolates from hospitalized patients in our study and 73% in extended-spectrum β-lactamase (ESBL)–producing Klebsiella spp. collected from 3 hospitals in Australia from 1989 to 1999 (12). These data cannot be directly compared because of differences in selection criteria and testing procedures, but as a whole they suggest that prevalence of integrons in these bacteria in the United States is high, as it is in other areas of the world. Carbapenem resistance was infrequent in the isolates we tested, so we could not address the association of integron carriage and metallo-β-lactamases that has been observed in other parts of the world (2).

Our study found an inverse association between integron presence and nonsusceptibility to cefepime in E. coli and to aztreonam in Klebsiella spp., which indicates that resistance determinants for these drugs are not frequently carried by integrons. One possible reason for the lack of an association between nonsusceptibility to cefepime and integron carriage is the relatively recent approval of cefepime for use in the United States. CTX-M-type β-lactamases are the most common resistance determinants in cefepime-resistant isolates worldwide, but these enzymes are just beginning to appear in the United States (13).

In phase IV of Project ICARE, participating sites were asked to submit isolates that showed decreased susceptibility to fluoroquinolones, extended-spectrum cephalosporins, or carbapenems. We did not track whether isolates were from patients in intensive care units or not. Thus, our isolates made up a convenience sample strongly associated with drug resistance mechanisms carried as integron gene cassettes and otherwise, and they are not necessarily representative of isolates from the United States in general. We did not examine carriage of >1 integron, as documented in earlier studies (10).

In these selected isolates, the crude association between a positive test result for class 1 integrons and nonsusceptibility to a fluoroquinolone (ciprofloxacin) in E. coli, and nonsusceptibility to a β-lactam (ceftazidime) in Klebsiella spp. failed to remain as independent variables in the multivariate model. This finding suggests that the mechanisms of resistance to fluoroquinolones and β-lactams are associated with mechanisms of resistance to the drugs that remained independently associated with class 1 integrons (aminoglycosides and trimethoprim-sulfamethoxazole), but for reasons unrelated to integron carriage. A previous study also found that ciprofloxacin resistance no longer predicted integron presence in Enterobacteriaceae when multivariate analysis was conducted (1). A Spanish study (14) found no association between integron carriage and β-lactam resistance in ESBL-producing E. coli strains unless strains contained metallo-β-lactamases, which were infrequently encountered in our study population. A biologic mechanism for association of resistance to fluoroquinolones and aminoglycosides (i.e., the inactivation of ciprofloxacin by the aac (6´)-Ib-cr aminoglycoside acetylating gene) has been recently reported (15).

In our study, both crude and multivariate associations between integron positivity and nonsusceptibility varied for the E. coli and Klebsiella groups. This finding emphasizes that studies of this type must assess whether analysis should combine results for organisms of different genus and species into larger groups (e.g., Enterobacteriaceae).

When determinants encoding resistance to a variety of antimicrobial classes are contained within an integron, use of any of these agents may select for and enhance expression of the other determinants (2). Thus, in the example of our studied isolates, use of aminoglycosides may lead to spread of a multidrug-resistant bacterial strain such as K. pneumoniae. The differential association of class 1 integrons with resistance in the isolates we studied suggests that integrons facilitate the spread of antimicrobial drug resistance in the United States.