Global Distribution and Epidemiologic Associations of Escherichia coli Clonal Group A,

Escherichia coli clonal group A (CGA) was first reported in 2001 as an emerging multidrug-resistant extraintestinal pathogen. Because CGA has considerable implications for public health, we examined the trends of its global distribution, clinical associations, and temporal prevalence for the years 1998-2007. We characterized 2,210 E. coli extraintestinal clinical isolates from 32 centers on 6 continents by CGA status for comparison with trimethoprim/sulfamethoxazole (TMP/SMZ) phenotype, specimen type, inpatient/outpatient source, and adult/child host; we adjusted for clustering by center. CGA prevalence varied greatly by center and continent, was strongly associated with TMP/SMZ resistance but not with other epidemiologic variables, and exhibited no temporal prevalence trend. Our findings indicate that CGA is a prominent, primarily TMP/SMZ-resistant extraintestinal pathogen concentrated within the Western world, with considerable pathogenic versatility. The stable prevalence of CGA over time suggests full emergence by the late 1990s, followed by variable endemicity worldwide as an antimicrobial drug-resistant public health threat.

Escherichia coli clonal group A (CGA) was fi rst reported in 2001 as an emerging multidrug-resistant extraintestinal pathogen.Because CGA has considerable implications for public health, we examined the trends of its global distribution, clinical associations, and temporal prevalence for the years 1998-2007.We characterized 2,210 E. coli extraintestinal clinical isolates from 32 centers on 6 continents by CGA status for comparison with trimethoprim/ sulfamethoxazole (TMP/SMZ) phenotype, specimen type, inpatient/outpatient source, and adult/child host; we adjusted for clustering by center.CGA prevalence varied greatly by center and continent, was strongly associated with TMP/SMZ resistance but not with other epidemiologic variables, and exhibited no temporal prevalence trend.Our fi ndings indicate that CGA is a prominent, primarily TMP/ SMZ-resistant extraintestinal pathogen concentrated within the Western world, with considerable pathogenic versatility.The stable prevalence of CGA over time suggests full emergence by the late 1990s, followed by variable endemicity worldwide as an antimicrobial drug-resistant public health threat.
E xtraintestinal infections caused by Escherichia coli are a substantial source of illness, death, and increased health care costs and have become increasingly challenging to manage because of the rising prevalence of resistance to fi rst-line antimicrobial drugs (1).The resistance problem is now recognized as having a prominent clonal component attributable in large part to the emergence and dissemination of specifi c antimicrobial drug-resistant clonal groups of extraintestinal pathogenic E. coli (2)(3)(4)(5)(6).
CGA fi rst came to attention during the late 1990s as a prominent cause of trimethoprim/sulfamethoxazole (TMP/ SMZ)-resistant urinary tract infections among otherwise healthy women across the United States (2,10).Isolates of CGA typically exhibit a fairly conserved virulence genotype that includes P fi mbriae (with the F16 structural subunit variant), group 2 capsule (with the K52 capsular antigen), and the aerobactin and yersiniabactin siderophore systems.They also commonly exhibit resistance to multiple antimicrobial agents other than TMP/SMZ, including tetracycline, chloramphenicol, streptomycin, and spectinomycin, with the corresponding resistance genes carried either on a large conjugative plasmid (2) or within a genomic resistance module (11).
CGA has been recognized primarily as a cause of community-acquired cystitis and pyelonephritis in adult women mainly in the United States (2,10,12,13).It is largely unknown to what extent CGA might have broader pathogenic capabilities with respect to anatomic site of infection (urine vs. nonurine), site of acquisition (hospital vs. community), and host age (adult vs. child).Likewise, although a global survey of E. coli clinical isolates from 2001 found CGA was signifi cantly associated with the United States (14), assessment of its distribution beyond the United States has been limited (5).Furthermore, no recent data are available regarding whether the overall prevalence of CGA is rising, stable, or waning, as can occur on a local level for CGA and other extraintestinal pathogenic E. coli clonal groups (3,13).Therefore, because of the major public health implications of CGA, we assessed the prevalence of this E. coli clonal group during 1998-2007 in multiple locales in the United States and internationally, paying specifi c attention to specimen type, inpatient versus outpatient status of host, and host age.

Strains
Sets of unpublished human clinical extraintestinal E. coli isolates were obtained from 32 clinical microbiology laboratories and affi liated repositories worldwide (online Technical Appendix, wwwnc.cdc.gov/EID/pdfs/11-0488-Techapp.pdf).Submitters were asked to provide ≈25 consecutive TMP/SMZ-resistant extraintestinal E. coli isolates and 25 concurrent TMP/SMZ-susceptible extraintestinal E. coli controls (1 per patient), or more, as available.When possible, isolates were to be distributed evenly by inpatient versus outpatient source and, within each of these categories, by specimen type (urine vs. other).Information was requested about the adult (age >18 years) versus child (age <18 years) status of the hosts and the local overall prevalence of TMP/SMZ-resistant E. coli.
Isolates were submitted to the research laboratory (J.R.J.) in agar stabs and frozen at −80ºC in 15% glycerol pending further analysis.Selected isolates underwent additional screening, by breakpoint agar dilution and disk diffusion, for confi rmation of TMP/SMZ phenotype.

Molecular Analysis
Major E. coli phylogenetic group (A, B1, B2, D) was defi ned by triplex PCR (15).Group D isolates were assessed for CGA (i.e., clonal complex 69) status in a staged fashion.First, all were screened by PCR for CGAassociated single-nucleotide polymorphisms (SNPs) in fumC (16).All fumC SNP-positive isolates (which included all CGA isolates, plus any non-CGA isolates containing the same fumC SNPs) then underwent pulsed-fi eld gel electrophoresis (PFGE) analysis of XbaI-restricted total DNA (17).Those with >94% PFGE profi le similarity to a known CGA isolate (determined on the basis of previous fumC and gyrB SNP analysis or 7-locus MLST) (5) were defi ned as CGA because this degree of PFGE similarity reliably predicts identity by MLST (J.R. Johnson, unpub.data).The fumC SNP-positive isolates without a PFGE profi le match of >94% to a known CGA isolate were individually screened by PCR for CGA-associated SNPs in gyrB.If the isolates were positive for gyrB SNPs and for the CGA-associated fumC SNPs, which together provide highly accurate identifi cation of CGA isolates, they were defi ned as CGA isolates (5,6,16).

Statistical Analysis
Unpaired and paired comparisons of proportions were tested by using the Fisher exact and McNemar tests, respectively.Selected variables were assessed as predictors of CGA status by using generalized linear models based on the generalized estimating equation (GEE; logistic GEE regression) to account for clustering by locale, supplemented by univariable and multivariable logistic regression analysis, as needed.Statistical analyses were conducted by using SPSS version 19.0 (IBM, Somers, NY, USA).

Study Population
A total of 2,210 E. coli clinical isolates from 32 globally dispersed centers were studied (Table 1); the isolates differed from those used in a prior global survey (14).Each center provided a median of 54 isolates (range 24-320).All centers but 1 provided TMP/SMZ-susceptible and TMP/SMZ-resistant isolates in approximately equal numbers.The median year of isolation was 2002 (range 1998-2007).Seventeen centers provided isolates for the fi rst half of the study period only (1998-2002); 12 centers provided isolates for the last half of the study period only (2003)(2004)(2005)(2006)(2007); and 3 centers provided isolates for both halves of the study period.
The 32 centers were in 19 countries, each represented by a single center, except the United States, which was represented by 14 centers.The 6 inhabited continents were each represented by multiple centers, as follows: Africa, 2; Asia, 4; Australia/New Zealand, 2; Europe, 5; North America, 15; and South/Central America, 4. Of the 15 centers in North America, 1 was in Canada and 14 in the United States (5 in Minnesota and 9 in other states).
Of the 32 centers, 31 reported the age (i.e., adult vs. child) of the patients from whom the clinical specimens were obtained.All 31 centers provided isolates derived from specimens from adults (n = 1,909), and 19 centers also provided isolates derived from specimens from children (n = 250).Of the 32 centers, 31 provided urine-source isolates (n = 1,511); 28 centers also provided isolates from other (nonurine) sources (n = 653).Outpatient versus inpatient source for clinical samples was reported by 29 centers, all of which provided isolates from outpatient specimens (n = 1,135); 26 centers also provided isolates from inpatient specimens (n = 926).As reported by 31 centers, the prevalence of E. coli TMP/SMZ susceptibility, by center, ranged from 36% to 90% (median 78%).

Phylogenetic Group and CGA Status versus TMP/SMZ Phenotype
Within the total group of isolates, phylogenetic group distribution was signifi cantly associated with TMP/SMZ phenotype (Table 2).Although the susceptible (n = 1,083) and resistant (n = 1,127) populations exhibited the same rank order for phylogenetic group prevalence (i.e., B2 >D >A >B1), absolute prevalences differed greatly by TMP/ SMZ phenotype.That is, among susceptible isolates, group B2 predominated overwhelmingly, being nearly 3× as prevalent as group D. In contrast, among resistant isolates, phylogenetic groups B2 and D were closely matched (approximately one third of isolates each).Accordingly, group D was strongly associated with TMP/SMZ resistance.

CGA Status in Relation to Geography
Of the 32 centers, 26 provided at least 1 CGA isolate (Table 1).The prevalence of CGA by center varied greatly, ranging from 0% to 34% (median 5%) for TMP/SMZresistant isolates and from 0% to 9.4% (median 1.4%) for TMP/SMZ-susceptible isolates.In all but 2 centers, CGA was at least as prevalent among TMP/SMZ-resistant as among TMP-SMZ-susceptible isolates; in 5 of the centers, the difference in prevalence was statistically signifi cant.The 5 centers with the highest prevalence of CGA isolates (2 in the United States, 3 in other countries) had >20% CGA prevalence among resistant isolates, and another 5 (1 United States, 4 in other countries) had 10%-19% prevalence.At the other extreme, 6 centers (3 in the United States, 3 in other countries) had no CGA isolates.
Accordingly, data for Africa and Asia were combined for comparison with data from other regions.Among TMP/ SMZ-susceptible isolates, CGA was similarly prevalent among the isolates from Africa and Asia combined and the isolates from other areas (3.3% vs. 2.8%, p>0.10).In contrast, among TMP/SMZ-resistant isolates, CGA was signifi cantly more prevalent among isolates from areas other than Africa and Asia than it was among isolates from Africa and Asia (12.5% vs. 4.0%, p<0.001) (Table 3).Likewise, CGA was signifi cantly associated with TMP/ SMZ resistance among the isolates from areas other than Africa and Asia (p<0.001) but not among the isolates from Africa and Asia (p>0.10)(Table 3).

CGA Status versus Other Variables
We also examined the prevalence of CGA in relation to other variables, after stratifi cation for TMP/SMZ phenotype (Table 4).For each variable (i.e., specimen type, host age group, host inpatient/outpatient status, and year isolate was obtained from patient specimen), CGA was signifi cantly more prevalent among TMP/SMZ-resistant than TMP/ SMZ-susceptible isolates.In contrast, for a given TMP/ SMZ phenotype, the prevalence of CGA varied minimally in relation to the other variables.Specifi cally, CGA was similarly (and, in some instances, slightly more) prevalent among isolates from nonurine versus urine specimens, children versus adults, inpatients versus outpatients, the fi rst half versus the second half of the study period (Table 4), and centers with isolates with a below-median versus above-median prevalence of TMP/SMZ resistance (not shown).This fi nding suggested that continent and TMP/ SMZ status were closely associated with CGA status, whereas other study variables were not.

Logistic GEE Models and Multivariable Analysis
To account for possible confounding of these associations because of clustering by center, we used logistic GEE regression models to assess associations of CGA with TMP/SMZ phenotype, continent (as Africa/ Asia vs. other), and the other nongeographic variables.Univariable analyses identifi ed the same signifi cant associations (or lack thereof) with CGA as noted initially; only TMP/SMZ phenotype and continent were confi rmed as signifi cant correlates of CGA status (Table 5).
Accordingly, we constructed a multivariable logistic GEE regression model based on TMP/SMZ phenotype and continent to assess the independent association of these 2 variables with CGA status.However, the model did not run to completion, possibly because of small numbers in certain cells (not shown).Univariable logistic regression analysis yielded results for these 2 variables separately that were similar to those obtained with the (univariable) generalized linear models (Table 5), which provided evidence that clustering by center had little effect on the associations.Therefore, a multivariable logistic regression model was constructed with TMP/SMZ phenotype and continent as the candidate predictor variables.This model yielded results similar to those of the univariable models, which provided evidence that the associations of CGA with TMP/ SMZ phenotype and continent are largely independent of each other (Table 5).

Discussion
In this global survey for the recently recognized E. coli lineage CGA among extraintestinal clinical isolates from humans during 1998-2007, we identifi ed strong associations of CGA with TMP/SMZ resistance and with regions other than Africa and Asia; this evidence indicates that CGA is primarily a TMP/SMZ-resistant pathogen concentrated within the Western world.In contrast, we found no association of CGA with other epidemiologic variables, which suggests that CGA is a similarly prominent pathogen among children and adults, among inpatients and outpatients, and within and outside the urinary tract.Finally, the fairly stable prevalence of CGA throughout the study period suggests that CGA had fully emerged by the late 1990s and now is an endemic public health threat in many centers worldwide.
The observed overall association of CGA with TMP/ SMZ resistance is consistent with the fi ndings of multiple studies (2,10,(12)(13)(14)(18)(19)(20).However, we did not fi nd this association in Africa and Asia.Overall prevalence of CGA was also lowest in these regions.Taken together, these fi ndings suggest that the TMP/SMZ-resistant variants of CGA had a selective advantage in the Western world but not in Asia and Africa, which led to the lineage's expansion in Europe, the Americas, and Australia but not in Asia and Africa.Why such an expansion seemingly has not occurred in Africa and Asia is unclear.One possibility is that TMP/ SMZ-resistant CGA isolates emerged fi rst in the Western world and have had insuffi cient time to diffuse to and expand within Africa and Asia.Alternatively, Africa and Asia may already have had an abundance of successful endemic TMP/SMZ-resistant clones competing with CGA for the same niche, effectively excluding it, or conditions in Africa and Asia may be somehow less permissive to the dispersal and expansion of this clonal group.Further comparisons of the TMP/SMZ-resistant populations from Africa and Asia versus other locales could be informative in this regard.
Clear-cut variation in the prevalence of CGA was evident at the continent level.However, marked differences also were apparent even among closely located centers, as noted in our smaller global survey (14).For example, whereas 1 Minneapolis center had no CGA isolates, another center had a high prevalence of CGA.To what extent these differences are real, rather than a refl ection of the inherent imprecision of small samples, is unclear.However, because different hospitals in the same locale often serve different patient populations and may draw from different catchment areas, the possibility of true variation by hospital is plausible.The determinants of this local variation, if real, would be potentially useful to discover as a step toward developing preventive measures.The center with the highest prevalence was in Denver, Colorado, USA, which also was the site of a previous survey with a high CGA prevalence; that survey involved different isolates than those included here (12).This consistency across studies suggests that Denver may be a focus of high-level endemicity for CGA.
CGA has been reported primarily as a urine pathogen among ambulatory women (2,10,13,19,20), which might be interpreted as indicating that urine is the favored niche or context of the clonal group.However, CGA has been reported in other clinical contexts, including, for example, as a cause of community-acquired pneumonia in a male renal transplant recipient (21).We found no association of CGA with urine versus nonurine (extraintestinal) source, inpatient versus outpatient host status, and host age (child vs. adult).This absence of discernible niche specialization suggests that CGA is a generalist, able to cause different types of infection in diverse host populations, in the hospital  and community alike.In terms of niche specialization, CGA is analogous to E. coli O18:K1:H7, which, although best known as an agent of neonatal meningitis, is also a prominent cause of acute cystitis in women (22,23).The pathogenic versatility of CGA has no doubt contributed to its epidemiologic success.We found no evidence of a time trend for the prevalence of CGA, even with locale taken into account, which suggests that CGA had already emerged and established widespread endemicity by 1998, the start of the study period.Overall, CGA was not as prevalent in this study as it was in the initial reports from the mid-to late 1990s (2,10).This discrepancy could refl ect selection bias rather than a true prevalence decrease by the time of the present study (i.e., our study included all patient specimens sent to clinical microbiology laboratories; the early studies included specimens from women with acute cystitis and uncomplicated pyelonephritis).
Even in regions in which prevalence was highest, CGA accounted for only a minority of TMP/SMZresistant isolates.This fi nding suggests that other resistant clonal groups are likely present, some of which could be similarly or more prominent compared with CGA (3,5,6,13,20,24,25).Identifi cation of such clonal groups and investigation of their epidemiology could help clarify the basis for the non-CGA component of TMP/SMZ resistance in E. coli, which represents a major ongoing public health threat.
Study limitations must be acknowledged.First, lack of information regarding the infected host (e.g., clinical symptoms, underlying health status, and sex) limits our understanding of the study population and precludes assessment of these variables in relation to CGA status.Second, variability by center in the completeness of epidemiologic data reporting reduced power for analyses involving those variables and may have introduced unrecognized bias.Third, limited sampling of certain geographic regions (especially Africa and Australia/ New Zealand) and host groups (children) constrained the inferences that could be drawn about those variables.Fourth, uncertainty regarding how closely each site followed the requested selection criteria allowed for possibly biased sample distribution by site.
Study strengths also must be acknowledged.First, the large sample size enhanced statistical power and permitted subgroup analyses that were not possible in previous studies.Second, the broad geographic distribution and multicenter design improved generalizability and allowed analyses by region.Third, the availability of basic epidemiologic data for most isolates enabled statistical analysis of these variables.Fourth, the predominantly prospective, consecutive sampling would be expected to provide a more broadly representative sample than a sample limited to a specifi c syndrome or host group.Fifth, the combined use of univariable and multivariable modeling, including adjustment for clustering, enabled optimal assessment of associations between individual variables and CGA.
In summary, our global survey for CGA during 1998-2007 identifi ed strong associations of CGA with TMP/ SMZ resistance and non-African/Asian origin but not with other epidemiologic variables-evidence that suggests CGA is a similarly prominent extraintestinal pathogen among children and adults, for inpatients and outpatients, and within and outside the urinary tract.The fairly stable prevalence of CGA through the 10-year study period suggests that CGA had fully emerged by the late 1990s and now is endemic worldwide as an antimicrobial drugresistant public health threat.

Table 4 .
Prevalence of clonal group A, by clinical/host variables, among 2,210 extraintestinal Escherichia coli isolates from 32 globally distributed centers, 1998-2007* clinical variable include only isolates for which status with respect to the particular variable was known.‡p values, by Fisher exact test, for comparisons of TMP/SMZ-susceptible vs. -resistant isolates within each subgroup.For all comparisons between subgroups within a given category (whether overall or by TMP/SMZ phenotype), p>0.10.