Variation in Resistance Traits, Phylogenetic Backgrounds, and Virulence Genotypes among Escherichia coli Clinical Isolates from Adjacent Hospital Campuses Serving Distinct Patient Populations

Abstract

Escherichia coli sequence type 13 (ST131), an emergent cause of multidrug-resistant extraintestinal infections, has important phylogenetic subsets, notably the H30 and H30Rx subclones, with distinctive resistance profiles and, possibly, clinical associations. To clarify the local prevalence of these ST131 subclones and their associations with antimicrobial resistance, ecological source, and virulence traits, we extensively characterized 233 consecutive E. coli clinical isolates (July and August 2013) from the University of Minnesota Medical Center-Fairview Infectious Diseases and Diagnostic Laboratory, Minneapolis, MN, which serves three adjacent facilities (a children's hospital and low- and high-acuity adult facilities). ST131 accounted for 26% of the study isolates (more than any other clonal group), was distributed similarly by facility, and was closely associated with ciprofloxacin resistance and extended-spectrum β-lactamase (ESBL) production. The H30 and H30Rx subclones accounted for most ST131 isolates and for the association of ST131 with fluoroquinolone resistance and ESBL production. Unlike ST131 per se, these subclones were distributed differentially by hospital, being most prevalent at the high-acuity adult facility and were absent from the children's hospital. The virulence gene profiles of ST131 and its subclones were distinctive and more extensive than those of other fluoroquinolone-resistant or ESBL-producing isolates. Within ST131, bla_CTX-M-15 was confined to H30Rx isolates and other bla_CTX-M variants to non-Rx H30 isolates. Pulsed-field gel electrophoresis documented a predominance of globally distributed pulsotypes and no local outbreak pattern. These findings help clarify the epidemiology, ecology, and bacterial correlates of the H30 and H30Rx ST131 subclones by documenting a high overall prevalence but significant segregation by facility, strong associations with fluoroquinolone resistance and specific ESBL variants, and distinctive virulence gene associations that may confer fitness advantages over other resistant E. coli.

INTRODUCTION

Escherichia coli sequence type 131 (ST131) has emerged globally over the past decade as the main cause of extraintestinal E. coli infections (1). Treatment is challenging due to characteristic resistance to fluoroquinolones and, increasingly, third-generation cephalosporins, which is mediated by extended-spectrum β-lactamases (ESBLs), primarily CTX-M-15 (2–4). ST131 belongs to E. coli phylogenetic group B2 and is most often closely associated with urinary tract infections but can infect any extraintestinal site (1, 2, 5). Recent reports of carbapenemase-producing ST131 isolates add urgency to this growing problem (6, 7).

The emergence of fluoroquinolone-resistant ST131 isolates has been due almost entirely to the ST131-H30 subclone, which carries the H30 variant of the type 1 fimbrial adhesin gene fimH (8). Extraintestinal E. coli infections in young children, the elderly, and long-term care facility residents are disproportionately due to ST131, with >85% of such isolates representing the ST131-H30 subclone (9). Recent whole-genome phylogenetic analyses confirmed that most CTX-M-producing ST131 isolates represent the H30Rx subset within the H30 subclone (10, 11), supporting that the ST131 epidemic has emerged via the expansion of nested subclones with successively broader resistance capabilities (12, 13). Although the H30Rx subclone appears to comprise a large fraction of CTX-M-producing E. coli isolates in North American populations (14, 15), the ecological and host associations of the H30Rx subclone have only begun to be explored.

The dramatic emergence of the H30 and H30Rx subclones remains unexplained. The signature resistance phenotypes of these subclones are well defined (8, 10, 14), but whether or how virulence also might have contributed to clonal expansion remains uncertain. The current prominence of these subclones as extraintestinal pathogens might suggest that they contain more virulence genes than non-ST131 E. coli or qualify more often as extraintestinal pathogenic E. coli (ExPEC). However, the available evidence supports these hypotheses only in comparison with other resistant E. coli (2, 16, 17).

Accordingly, we addressed three primary study questions. First, we sought to determine the prevalence of ST131 and its H30 and H30Rx subclones in comparison to that of the major E. coli phylogroups and classic ExPEC lineages among consecutive E. coli isolates from three closely located hospital campuses in Minneapolis, MN, which serve highly distinct patient populations and share a clinical microbiology laboratory. Second, we compared the ST131 subclones with other E. coli according to virulence gene carriage and resistance phenotypes and genotypes. Third, we sought to determine to what extent the local ST131, H30, and H30Rx populations are distinct from, versus part of, the global ST131 expansion and whether they represent a local clonal outbreak versus diverse strains.

MATERIALS AND METHODS

Setting and isolates.

The University of Minnesota (UM) Medical Center-Fairview Infectious Diseases and Diagnostic Laboratory (UMMC IDDL) processes specimens from the medical center's East Bank and West Bank campuses and the UM Masonic Children's Hospital, which are closely colocated in central Minneapolis, MN. Each facility has an emergency department and intensive care units. The West Bank campus houses the orthopedic, behavioral health, long-term and acute rehabilitation, and obstetrical services. The East Bank campus serves primarily patients with more acute and complicated diagnoses and includes active stem cell transplant, solid organ transplant, and oncology services.

During the 7-week study period (July and August 2013), all clinical E. coli isolates from the IDDL were collected, deidentified, and submitted to the research laboratory. Collection date, specimen type, origin (facility, outpatient/inpatient status), ciprofloxacin phenotype, and ESBL status were recorded. The ciprofloxacin and ESBL phenotypes were determined by a Vitek 2 instrument (bioMérieux, Inc, Durham, NC), according to the manufacturer's instructions. Isolates with a ciprofloxacin MIC of ≥4 μg/ml were considered resistant.

Molecular methods.

Isolates were characterized molecularly using previously described methods (16). Major E. coli phylogenetic group, i.e., A, B1, B2, C (formerly typed as A), D, E (formerly typed as D), and F (also formerly typed as D) was determined by the revised multiplex PCR-based approach of Clermont et al. (18, 19). ST131 genotype was determined by PCR-based detection of ST131-specific single-nucleotide polymorphisms (SNPs) in gyrB and mdh (20). ST131 isolates were further tested for membership in the H30 and H30Rx ST131 subclones by subclone-specific SNP-based PCR (10). Additional lineage-specific PCR assays were used to detect other prominent ExPEC clonal lineages, including ST95 and ST73 (group B2) and ST69, clonal complex 31 (CC31), and ST405 (group D) (20–23). Additionally, a dual-locus sequence typing system based on fumC and fimH (CH typing) was used to presumptively define the ST complex of origin for 10 randomly selected group B2 isolates that were not from ST131, ST73, or ST95 (24).

The presence of 50 ExPEC-associated virulence genes (VGs) was assessed by multiplex PCR. A virulence score was the total number of virulence genes detected, adjusted for the multiple detection of pap (P fimbriae), sfa/foc (S and F1C fimbriae, respectively), and kps II (group 2 capsule) operons. Isolates were classified as ExPEC if positive for ≥2 of the following: papAH and/or papC (P fimbriae), sfa/focDE, afa/draBC (Dr family adhesins), iutA (aerobactin receptor), and kpsM II (25).

CTX-M-type β-lactamase genes were identified with two sets of previously described primers, one for the generic detection of all bla_CTX-M variants, and the second specific for bla_CTX-M-15 (26, 27).

Pulsed-field gel electrophoresis analysis.

For a more discriminating analysis of the genetic similarities of the ST131 isolates, pulsed-field gel electrophoresis (PFGE) analysis was done as described previously (28), based on a standardized protocol (29) (Fig. 1). BioNumerics software version 6.6 (Applied Maths) was used for digital capture and analysis of profiles. Based on pairwise Dice similarity coefficients, isolates exhibiting >94% profile similarity (≈3-band difference) to the index isolate for an established pulsotype, implying close genetic similarity, were assigned to that pulsotype. A PFGE profile dendrogram was constructed according to the unweighted pair group method.

FIG 1.

Pulsed-field gel electrophoresis profile dendrogram of ST131 study isolates. The scale is the percent profile similarity according to Dice similarity coefficients. Columns to the right of the dendrogram give associated data on the isolates. cath, catheter; ESBL, extended-spectrum β-lactamase; FQ-R, fluoroquinolone resistant; midstr, midstream; neph, nephrostomy; PFGE, pulsed-field gel electrophoresis pulsotype (as defined within a large private PFGE library); und, undefined; Ur, urethral. Hyphens indicate absence of a particular characteristic.

Statistical methods.

Comparisons of proportions and continuous variables were tested using Fisher's exact test and the Mann-Whitney U test, respectively (all 2-tailed). The significance criterion was a P value of <0.05.

RESULTS

Study population.

During the 7-week study period (July and August 2013), the UMMC IDDL recovered 233 E. coli isolates. Of these, 211 (91%) were from unique patients, and 22 (9%) were from 11 patients (two per patient). The East Bank facility contributed the most isolates (150 [64%]), the West Bank facility fewer (65 [29%]), and the children's hospital the fewest (18 [7%]). The commonest specimen type was urine (80%). Most isolates (69%) were collected from outpatients in one of the three emergency departments.

Overall, 65 (28%) of the 233 isolates were ciprofloxacin resistant, and 16 (7%) had an ESBL-positive phenotype. Fifteen (94%) of the 16 ESBL-positive isolates contained bla_CTX-M, and 10 (63%) contained bla_CTX-M-15. Ciprofloxacin resistance and ESBL production were closely correlated, with all 16 (100%) ESBL-positive isolates being ciprofloxacin resistant versus only 49/217 (23%) ESBL-negative isolates (P < 0.001).

Phylogenetic group, clonal group, and pulsotype distribution.

The 233 study isolates were predominantly (70%) from phylogenetic group B2, and the remainder were from groups A, B1, and D (8 to 9% each) and C, E, and F (1 to 2% each) (Table 1). According to clonal group-specific PCR, ST131 accounted for 61 (26%) of the 233 study isolates, ST131-H30 for 42 (18%), and ST131-H30Rx for 11 (5%). The H type of the non-H30 ST131 isolates was not determined, although previous epidemiologic studies suggest that O16 ST131 isolates are likely H41, and non-O16 ST131 isolates are likely H22 (30).

TABLE 1.

Characteristics of 233 E. coli clinical isolates according to source hospital

Traita	Prevalence of trait (no. of isolates [column %]) at:				P valueb
	Total (n = 233)	Children's (n = 18)	West Bank (n = 65)	East Bank (n = 150)
Cipr	65 (28)	1 (6)	14 (22)	50 (33)	0.02
ESBL	16 (7)	0 (0)	0 (0)	16 (11)	<0.001
Group A	20 (9)	1 (6)	2 (3)	17 (11)
Group B1	22 (9)	1 (6)	8 (12)	13 (9)
Group B2	162 (70)	13 (72)	48 (74)	101 (67)
ST131	61 (26)	4 (22)	16 (25)	41 (27)
H30	42 (18)	0 (0)	11 (17)	31 (21)	0.049
H30Rx	11 (5)	0 (0)	1 (2)	10 (7)
ST73	20 (9)	1 (6)	7 (11)	12 (8)
ST95	24 (10)	2 (11)	10 (15)	12 (8)
Group C	4 (2)	0 (0)	0 (0)	4 (3)
Group D	19 (8)	3 (17)	5 (8)	11 (7)
ST69	4 (2)	2 (11)c	1 (2)	1 (1)	0.03c
ST405	3 (1)	0 (0)	0 (0)	3 (2)
Group E	2 (1)	0 (0)	1 (2)	1 (1)
Group F	4 (2)	0 (0)	1 (2)	3 (2)

^aCip^r, ciprofloxacin resistant; ESBL, extended-spectrum β-lactamase production.

^bP values (for 3-group comparison across the 3 facilities, unless specified otherwise) are shown for comparisons that yielded a P value of ≤0.05. For all other comparisons, P > 0.05.

^cFor children's hospital versus other sites, P = 0.03.

ST95 accounted for only 24 (10%), ST73 for 20 (9%), ST69 for 4 (2%), and ST405 for 3 (1%) isolates, whereas CC31 was not detected. Supplemental CH typing of 10 randomly selected group B2 isolates from the 57 isolates that did not type as ST131, ST73, or ST95 identified six minor STs, including ST12 (n = 3) ST14 (n = 2), ST127 (n = 2), ST141 (n = 1), and 2 undefined STs (n = 1 each). Based on this distribution and the overall prevalence of non-ST131, -ST73, and -ST95 group B2 isolates (24%), and each of these minor STs was estimated to account for 3 to 7% of the total population. Thus, ST131 and its H30 subset, with 26% and 18% total prevalence rates, respectively, were by far the most prevalent clonal subsets within the population, outstripping all other group B2 STs and even the non-B2 major phylogenetic groups.

XbaI PFGE analysis resolved 38 different pulsotypes, which accounted for one to 12 isolates each (Fig. 1). The three most prevalent pulsotypes, i.e., types 968 (20%), 800 (7%), and 812 (10%), represented the top pulsotypes in a previous global PFGE survey of ST131 isolates (28). There were only two instances of paired isolates with indistinguishable profiles, in each of which pairs of matching isolates were collected from different sites from the same patient at the same facility.

Resistance versus phylogeny.

In contrast to the expectation that group B2 isolates should be more susceptible than others (31), ciprofloxacin resistance showed no significant phylogenetic group associations, and ESBL production was significantly associated with group C only (Table 2). In contrast, at the clonal group level, both ciprofloxacin resistance and ESBL production were strongly associated with ST131, H30, and H30Rx (Table 2). Specifically, all H30 and H30Rx isolates, and 44 (72%) of ST131 isolates overall, were ciprofloxacin resistant. Similarly, ST405 (group D) was significantly associated with ESBL production. In contrast, ST73 and ST95 (group B2) were significantly associated with ciprofloxacin susceptibility.

TABLE 2.

Ciprofloxacin and extended-spectrum β-lactamase phenotypes by phylogenetic and clonal group

Phylogenetic/clonal group	Prevalence of phylogenetic/clonal group (no. of isolates [column %]) for:a			P valueb	Prevalence of phylogenetic/clonal group (no. of isolates [column %]) for:		P valueb
	Total (n = 233)	Cips (n = 168)	Cipr (n = 65)		ESBL negative (n = 217)	ESBL positive (n = 16)
Group A	20 (9)	17 (10)	3 (5)		18 (8)	2 (13)
Group B1	22 (9)	18 (11)	4 (6)		22 (10)	0 (0)
Group B2	162 (70)	112 (67)	50 (77)		153 (71)	9 (56)
ST131	61 (26)	17 (10)	44 (68)	<0.001	52 (24)	9 (56)	0.008
H30	42 (18)	0 (0)	42 (65)	<0.001	34 (16)	8 (50)	0.003
H30Rx	11 (5)	0 (0)	11 (17)	<0.001	5 (2)	6 (38)	<0.001
ST73	20 (9)	20 (12)	0 (0)	0.001	20 (9)	0 (0)
ST95	24 (10)	24 (14)	0 (0)	<0.001	24 (11)	0 (0)
Group C	4 (2)	1 (1)	3 (5)		2 (1)	2 (13)	0.02
Group D	19 (8)	15 (9)	4 (6)		17 (8)	2 (13)
ST69	4 (2)	4 (2)	0 (0)		4 (2)	0 (0)
ST405	3 (1)	1 (1)	2 (3)		1 (0)	2 (13)	0.01
Group E	2 (1)	2 (1)	0 (0)		2 (1)	0 (0)
Group F	4 (2)	3 (2)	1 (2)		3 (1)	1 (6)

^aCip^s, ciprofloxacin susceptible; Cip^r, ciprofloxacin resistant.

^bP values are shown where P < 0.05; for all other comparisons, P ≥ 0.05.

These findings suggested that the paucity of overall phylogenetic group associations with ciprofloxacin resistance and ESBL production might be due to a “diluting” effect from the presence of ST131 within group B2. Supporting this hypothesis, repetition of the analysis without ST131 yielded a quite different picture, with group B2 being significantly associated with ciprofloxacin susceptibility and ESBL negativity (P = 0.004 and 0.002, respectively) (Table 3). Additionally, phylogenetic group C now exhibited a significant association with ciprofloxacin resistance (P = 0.006) and an even stronger association with ESBL production (P = 0.008) (Table 3); the same was true for ST405 (for ciprofloxacin-resistant isolates, 67% [ST405] versus 11% [others], P = 0.04; for ESBL production, 67% [ST405] versus 3% [others], P = 0.004). In contrast, ST73 and ST95 lost their significant associations with ciprofloxacin susceptibility (not shown).

TABLE 3.

Ciprofloxacin and extended-spectrum β-lactamase phenotypes by phylogenetic group after excluding the 61 ST131 isolates

Phylogenetic group	Total no. of isolates (% of 172)	Prevalence of phylogenetic/clonal group (no. of isolates [column %]) for:a		P valuea	Prevalence of phylogenetic/clonal group (no. of isolates [column %]) for:a		P valuea
		Cips (n = 151)	Cipr (n = 21)		ESBL negative (n = 165)	ESBL positive (n = 7)
A	20 (12)	17 (11)	3 (14)		18 (11)	2 (29)
B1	22 (13)	18 (12)	4 (19)		22 (13)	0 (0)
B2 (non-ST131)	101 (59)	95 (63)	6 (29)	0.004	101 (61)	0 (0)	0.002
C	4 (2)	1 (1)	3 (14)	0.006	2 (1)	2 (29)	0.008
D	19 (11)	15 (10)	4 (19)		17 (10)	2 (29)
E	2 (1)	2 (1)	0 (0)		2 (1)	0 (0)
F	4 (2)	3 (2)	1 (5)		3 (2)	1 (14)

^aP values are shown where P < 0.05; for all other comparisons, P ≥ 0.05.

Hospital distribution.

Hospital campus of origin was significantly associated with ciprofloxacin resistance, ESBL phenotype, and ST131 subclone. Specifically, 64 (98%) of 65 ciprofloxacin-resistant isolates were from the East and West Bank adult facilities (versus 0% for the children's facility, P = 0.02; Table 1). Similarly, all 16 ESBL-expressing isolates (100%) originated from the East Bank facility (versus none from the West Bank and children's facilities; P < 0.001) (Table 1).

The major phylogenetic groups and ST131 were distributed similarly across the 3 hospital campuses (Table 1). In contrast, the ST131-H30 subclone exhibited a marked gradient across facilities, being most abundant at the East Bank facility (21% of all isolates), intermediate at the West Bank facility (11%), and absent from the children's facility (P = 0.049) (Table 1). The H30Rx group followed this trend, accounting for 7% of the East Bank isolates, 2% of the West Bank isolates, and no children's hospital isolates.

Clonal associations of bla_CTX-M and bla_CTX-M-15.

bla_CTX-M was significantly more prevalent among ST131 (13% versus 4%), ST131-H30 (19% versus 4%), and ST131-H30Rx (55% versus 4%) isolates than others (P < 0.05 for each comparison) (Table 4). Likewise, bla_CTX-M-15 was significantly more frequent among ST131 (10% versus 2%), ST131-H30 (14% versus 2%), and ST131-H30Rx (55% versus 2%) isolates than others (P < 0.05 for each comparison) (Table 4). Within ST131, bla_CTX-M-15 (n = 6) was confined to the H30Rx subclone, whereas other bla_CTX-M variants (n = 2) were confined to the non-Rx subset within H30 (Table 4).

TABLE 4.

Associations of bla_CTX-M and bla_CTX-M-15 with ST131 and its clonal subsets

Gene	Total no. of isolates (n = 233)	No. with trait (column %)		P valuea	No. with trait (column %)		P valuea	No. with trait (column %)		P valuea
		ST131 (n = 61)	Others (n = 172)		H30 (n = 42)	Others (n = 191)		H30Rx (n = 11)	Others (n = 222)
blaCTX-M	15	8 (13)	7 (4)	0.03	8 (19)	7 (4)	0.001	6 (55)	9 (4)	<0.001
blaCTX-M-15	10	6 (10)	4 (2)	0.02	6 (14)	4 (2)	0.003	6 (55)	4 (2)	<0.001

^aP value by Fisher's exact test.

Virulence genes.

Of the 50 tested VGs, 31 (62%) were significantly associated (negatively or positively) with ST131 and/or its clonal subsets (Table 5). Although most associations were negative, nine VGs were significantly overrepresented in ST131, including iha (adhesin siderophore receptor), sat and usp (toxins), fyuA and iutA (siderophore receptors), traT and ompT (outer membrane proteins), kpsM II (group 2 capsule), and malX (pathogenicity island marker). These same nine genes were also significantly overrepresented among H30 isolates and seven also among H30Rx isolates, despite the smaller group size. Molecular ExPEC status was significantly associated with both ST131 and its H30Rx subclone, with the H30 subclone (100% ExPEC) exhibiting a similar trend.

TABLE 5.

Virulence gene distribution by ST131 and its H30 and H30Rx subclones

Category	Genea	Prevalence of trait (no. of isolates [column %])				P value vs all others forb:
		Total (n = 233)	ST131 (n = 61)	H30 (n = 42)	H30Rx (n = 11)	ST131	H30	H30Rx
Adhesins	papAH	75 (32)	6 (10)	2 (5)	1 (9)	<0.001	<0.001
	papC	79 (34)	6 (10)	2 (5)	1 (9)	<0.001	<0.001
	papEF	75 (32)	6 (10)	2 (5)	1 (9)	<0.001	<0.001
	papG	62 (27)	6 (10)	1 (2)	1 (9)	<0.001	<0.001
	Allele GII	46 (20)	5 (8)	2 (5)	1 (9)	0.008	0.005
	Allele GIII	29 (12)	1(2)	0 (0)	0 (0)	0.001	0.003
	sfa/focDE	40 (17)	0 (0)	0 (0)	0 (0)	<0.001	<0.001
	sfaS	18 (8)	0 (0)	0 (0)	0 (0)	0.005	0.05
	focG	13 (6)	0 (0)	0 (0)	0 (0)	0.02
	afa/draBC	21 (9)	11 (18)	4 (10)	4 (36)	0.008		0.01
	iha	84 (36)	54 (89)	39 (93)	11 (100)	<0.001	<0.001	<0.001
	hra	65 (28)	2 (3)	1 (2)	1 (9)	<0.001	<0.001
Toxins	hlyD	56 (24)	4 (7)	1 (2)	1 (9)	<0.001	<0.001
	cnf1	42 (18)	1 (20)	1 (2)	1 (9)	<0.001	0.002
	sat	86 (37)	55 (90)	40 (95)	11 (100)	<0.001	<0.001	<0.001
	vat	97 (42)	1 (2)	1 (2)	0 (0)	<0.001	<0.001	0.003
	usp	155 (67)	61 (100)	42 (100)	11 (100)	<0.001	<0.001
	pic	21 (9)	0 (0)	0 (0)	0 (0)	0.002
Siderophores	iroN	58 (25)	3 (5)	0 (0)	0 (0)	<0.001	<0.001
	fyuA	189 (81)	61 (100)	42 (100)	11 (100)	<0.001	<0.001
	ireA	30 (13)	1 (2)	1 (2)	0 (0)	0.001	0.022
	iutA	114 (49)	57 (93)	40 (95)	11 (100)	<0.001	<0.001	<0.001
Protectins	kpsM II	154 (66)	46 (75)	27 (64)	11 (100)			0.02
	K1	53 (23)	2 (3)	1 (2)	0 (0)	<0.001	<0.001
	K2/K100	22 (9)	14 (23)	5 (12)	5 (46)	<0.001		0.002
	K5	35 (15)	22 (36)	15 (36)	3 (27)	<0.001	<0.001
	traT	160 (69)	58 (91)	39 (93)	11 (100)	<0.001	<0.001	0.02
	iss	21 (9)	3 (5)	0 (0)	0 (0)		0.02
Other	clbB	74 (32)	1 (2)	1 (2)	0 (0)	<0.001	<0.001	0.02
	clbN	74 (32)	1 (2)	1 (2)	0 (0)	<0.001	<0.001	0.02
	ibeA	34 (15)	10 (16)	0 (0)	0 (0)		0.001
	ompT	180 (77)	61 (100)	42 (100)	11 (100)	<0.001	<0.001	0.07
	malX	165 (71)	61 (100)	42 (100)	11 (100)	<0.001	<0.001	0.04
	H7 fliC	30 (13)	1 (2)	1 (2)	0 (0)	0.001	0.02
ExPECc	NAd	143 (61)	44 (72)	27 (64)	11 (100)	0.05		0.008

^aOf the 50 virulence genes tested, the 34 shown yielded P < 0.05 in at least one comparison: afa/draBC, Dr family adhesins; iha, adhesion siderophore; alleles GII and GIII, P-fimbriae adhesin variants; clbB and clbN, colibactin synthesis; cnf1, cytotoxic necrotizing factor; focG, SF1C fimbriae; fyuA, yersiniabactin receptor; hlyD, hyaluronidase; hra, heat-resistant agglutinin; ibeA, invasion-associated gene; ireA, siderophore receptor; iroN, salmochelin receptor; iss, increased serum survival gene; iutA, aerobactin receptor; K1, K2/K100, and K5, group 2 capsule variants; kpsM II, group 2 capsule; H7, fliC (flagella) variant; malX, pathogenicity island marker; ompT, outer membrane protein T (protease); papAH, papC, papEF, and papG, P-fimbrial structural subunit, assembly, tip pilins, and adhesion, respectively; pic, serine protease autotransporters; sat, secreted autotransporter toxin; sfa/focDE, S or F1C fimbriae; sfaS, S fimbriae; traT, serum-resistance associated; usp, uropathogenic-specific protein; and vat, vacuolating toxin. Genes associated positively with at least one ST131 clonal subset are shown in bold. Genes detected in ≥1 isolate each but not yielding P < 0.05, included (total prevalence): afaE8, variant Dr family adhesion (0.4%); astA, enteroaggregative E. coli toxin (4%); bmaE, M fimbriae (0.4%); cdtB, cytolethal distending toxin (3%); cvaC, microcin V (7%); hlyF, variant hemolysin (8%); kpsM III, group 3 capsule (2%); rfc, O antigen polymerase (4%); and tsh, serine protease autotransporters (5%). Genes sought but not detected included allele GI, P fimbriae adhesin variants; clpG, major subunit of CS31A fimbriae; F17, F17 fimbriae; gafD, G fimbriae; and K15, group 2 capsule variant.

^bP values (by Fisher's exact test) are shown where P < 0.05.

^cExPEC, extraintestinal pathogenic E. coli, defined operationally as the presence of >2 of (papA and/or papC, counted as one), sfa/focDE, afa/draBC, iutA, and kpsM II.

^dNA, not applicable.

Virulence scores in relation to ST131 and its subclones.

Aggregate virulence scores did not differ significantly between the ST131, H30, or H30Rx isolates and other isolates, either overall or specifically among ciprofloxacin-susceptible or ESBL-negative isolates (Table 6). In contrast, among ciprofloxacin-resistant and ESBL-positive isolates, the ST131, H30, and H30Rx isolates had significantly higher aggregate virulence scores than those of other isolates by 1 to 4 points (ciprofloxacin-resistant isolates) or 6 points (ESBL-positive isolates), according to the group medians (Table 6).

TABLE 6.

Distribution of aggregate virulence gene scores in relation to ST131, H30, and H30Rx status and ciprofloxacin and extended-spectrum β-lactamase (ESBL) phenotype

Population	Subgroup	No. of isolates in:		Virulence gene score (median [range]) in:		P valuea
		Subgroup	Others	Subgroup	Others
Total	ST131	61	172	10 (5, 17)	10 (0, 20)
	H30	42	191	10 (5, 15)	11 (0, 20)
	H30Rx	11	222	10 (10, 14)	10 (0, 20)
Ciprofloxacin susceptible	ST131	17	151	12 (8, 17)	12 (0, 20)
	H30	0	168	NAb	12 (0,20)	NDc
	H30Rx	0	168	NA	12 (0, 20)	ND
Ciprofloxacin resistant	ST131	44	21	10 (5, 15)	6 (1, 10)	<0.001
	H30	42	23	10 (5, 15)	8 (1, 10)	<0.001
	H30Rx	11	54	10 (10, 14)	9 (1, 15)	<0.001
ESBL negative	ST131	52	165	10 (5, 17)	10 (0, 20)
	H30	34	183	10 (5, 15)	11 (0, 20)
	H30Rx	5	212	10 (10, 10)	10 (0, 20)
ESBL positive	ST131	9	7	11 (10, 14)	5 (3, 8)	<0.001
	H30	8	8	11 (10, 14)	5 (3, 10)	0.001
	H30Rx	6	10	12 (10, 14)	6 (3, 10)	0.001

^aP values (by the Mann-Whitney test) are shown where P < 0.05.

^bNA, not applicable (no ciprofloxacin-susceptible H30 or H30Rx isolates).

^cND, not done (Mann-Whitney test not performed with empty groups).

DISCUSSION

In this molecular-epidemiological survey, we assessed the prevalence of multiple E. coli lineages, including the pandemic ST131 clonal group and its recently described H30 and H30Rx subclones, among 233 consecutive clinical E. coli isolates from three UM Medical Center hospital campuses over a 2-month period in 2013. We also studied the associations of these clonal groups with source, resistance phenotype, and virulence genotype.

Our main findings are 3-fold. First, ST131 was the dominant lineage both within group B2 and overall, and it was similarly prevalent (22% to 26%) across specimen types and hospitals. In contrast, the dominant H30 subclone of ST131 was negatively associated with the children's hospital (0% prevalence) and was most prevalent at the (high-acuity) East Bank facility. Second, ST131 and its subclones were the main carriers of fluoroquinolone resistance and ESBLs in the study population. Third, although aggregate virulence scores were comparable for ST131 and other E. coli isolates overall, within the fluoroquinolone-resistant and ESBL-producing subsets, ST131 isolates had a significant virulence score advantage over non-ST131 isolates. These findings indicate that ST131 E. coli comprises a significant population of virulent and resistant clinical isolates in our Midwestern locale and that its subclones have distinct ecological and, likely, epidemiological and clinical associations.

According to our PFGE analysis, the ST131 isolates comprise 38 different pulsotypes, the most prevalent of which were pulsotypes 968 (20%), 800 (7%), and 812 (10%). These pulsotypes were described previously as common E. coli ST131 U.S. strains, suggesting that our local ST131 population reflects part of the national (and global) ST131 ecology rather than being a localized regional phenomenon (28). Further, the absence of PFGE profile commonality across patients effectively excludes a local point source outbreak, favoring instead either the importation of diverse strains into the institutions from the community or clonally diverse within-institution reservoirs.

Our study uniquely compared multiple adjacent hospitals within the same health care system, which share a clinical microbiology laboratory, to assess the prevalence of ST131 and its H30 and H30Rx subclones. The three study sites constituted a spectrum from a children's hospital, to a lower-acuity primarily outpatient adult facility (West Bank), and to a higher-acuity adult tertiary care center (East Bank). The overall prevalence of ST131 (26%) corresponded closely with values noted in recent surveys of the U.S. veteran population (27.7%) and Olmstead County, MN, laboratories (27%) (9, 16), suggesting concordance with what is known about the nationwide prevalence of this E. coli lineage. Moreover, ST131 per se was fairly evenly distributed across the three study facilities, as were the major E. coli phylogenetic groups (except group A) and selected non-ST131 clonal groups.

In contrast, we found marked disparities in the across-facility distribution of the ST131 subclones. That is, although overall the H30 and H30Rx subclones accounted for 18% and 4.7% of all E. coli isolates and for 69% and 18% of ST131 isolates, respectively, they were absent from the children's hospital, at intermediate prevalence at the West Bank facility, and highly prevalent at the East Bank facility.

This relationship parallels the observed distribution of resistance phenotypes. Ciprofloxacin resistance and ESBL carriage were associated with the non-children's facilities and the East Bank facility, respectively, and with ST131-H30 and ST131-H30Rx. Although these associations with facility were statistically significant for H30 only, the trends for H30Rx were numerically similar. Thus, future studies of ST131 epidemiology should distinguish among these ST131 subclones, which have distinctive resistance and epidemiological correlates. An analysis of ST131 as a unitary entity, which is the traditional approach, may fail to detect important associations and lead to erroneous conclusions. Additionally, ecological context appears to influence significantly the antimicrobial resistance prevalence and clonal mix of the E. coli population, even within a given locale.

The negative association between ST131-H30 and the relatively small number of children's hospital isolates corresponds with the findings of Banerjee et al. (9) from their Olmstead County cohort. There, although ST131 was present in all age cohorts, non-H30 ST131 isolates predominated in the 11-to-20-year-old cohort, within which nearly all ST131 isolates were non-H30. Isolates from children <10 years included slightly more H30 than non-H30 isolates. This association of non-H30 ST131 E. coli with older children likely reflects multiple factors, including distinctive patterns of antibiotic exposure and comorbidities. An additional patient characteristic association was recently reported by Burgess et al. (32), whose data suggest that the H30 subclone is overrepresented among long-term care facility residents, and that a subset of these isolates have the H30Rx ESBL phenotype. While our West Bank facility includes a long-term rehabilitation facility, there are multiple other care services at this site. Our study does not allow for age-specific analysis; larger studies in both children and long-term care facility residents are needed to examine these issues.

Also notable were the associations of ST131-H30 and ST131-H30Rx with the East Bank facility. The East Bank facility patients tend to be severely ill, extensively antibiotic exposed, and multiply compromised (our unpublished data). This would favor extensively resistant pathogens while not requiring as much intrinsic virulence as for infections in intact hosts. Although ST131-H30Rx has been proposed as an aggressive sepsis-causing pathogen (8, 10), recent data suggest that it behaves more as a multidrug-resistant opportunist (33). More refined epidemiological analyses are needed to clarify the host correlates and associated clinical outcomes of the different ST131 subclones.

Regarding our third main finding, virulence gene content may contribute to the success of the ST131-H30 and ST131-H30Rx subclones. Here, although aggregate virulence scores of the ST131, H30, and H30Rx isolates did not differ significantly from those of other E. coli generally, they were significantly higher than those of non-ST131 ciprofloxacin-resistant and ESBL-producing E. coli isolates, similar to previous findings by our group and others (2, 16, 17). Additionally, many individual virulence genes were significantly associated with ST131 and its subclones. Nine such genes were overrepresented among the ST131 isolates, including specific adhesins, toxins, and siderophore receptors. These lineage-specific patterns of virulence gene carriage coincide closely with previous findings from other locales (9, 10, 14, 16), suggesting that the present ST131 study isolates are part of the global ST131 expansion and not a distinctive local population. Further, while the ST131 subclones may have nonsuperior virulence overall, they may be more virulent than other resistant strains, promoting their rise to predominance within the resistant subpopulation (1, 7).

Our study has several limitations. First, although consecutive sampling reflected the number of ST131 isolates originating from each facility, the relatively few isolates from the children's hospital limited the statistical power for certain comparisons. Second, given the relatively small sample size, conclusions must be interpreted cautiously. Third, we lacked patient-specific data and relied instead on characteristics of the specimens (facility, inpatient versus outpatient source, and specimen type). Third, for phylogenetic group assignments and inferences regarding virulence potential, we relied on PCR-based detection of SNPs and virulence genes, which have well-recognized limitations at the individual strain level (18, 34).

The study strengths include our ability to examine E. coli diversity within three adjacent hospitals in the same health care system that share a clinical microbiology laboratory and yet serve very different patient populations. Our findings underscore the importance of understanding intraregional differences when drawing conclusions about how microbiologic niches may differ between regions. Another strength was the extensive molecular characterization of the E. coli isolates, including phylotyping using the updated Clermont system (19) and the detection of multiple individual clonal groups.

In summary, we documented a high overall and individual hospital prevalence of ST131 among consecutive clinical E. coli isolates from three adjacent Midwestern hospital campuses. The H30 and H30Rx subclones were strongly associated with ciprofloxacin and CTX-M-mediated cephalosporin resistance and varied significantly in prevalence by hospital, being notably absent from the children's facility. Although these subclones are part of a global clonal expansion, any virulence advantage (as inferred from virulence genotypes) was found only in comparison with other resistant isolates. Our findings demonstrate the importance of delineating the ST131-H30 and ST131-H30Rx subclones in epidemiologic surveys and the effect of host characteristics (as inferred here from hospital of origin) in determining the composition of the local clinical E. coli population.