Large Fecal Reservoir of Escherichia coli Sequence Type 131-H30 Subclone Strains That Are Shared Within Households and Resemble Clinical ST131-H30 Isolates

Abstract

Background

Emerging antimicrobial-resistant Escherichia coli represent mainly the nested (fluoroquinolone-resistant [FQR]) H30R and H30Rx subclones within sequence type 131 (ST131). Intestinal colonization and within-household transmission may underlie H30R’s emergence.

Methods

We screened fecal samples from 741 volunteers (383 veterans, 358 household members, including pets) for ST131 and FQR E. coli (FQREC) and used molecular profiling to resolve unique strains. Selected strains underwent PCR-based detection of phylogroups, sequence types (STs), H30, H30Rx, and 53 virulence genes (VGs). Within-household strain sharing was compared with household, host, and bacterial characteristics. Fecal isolates were compared with clinical isolates.

Results

Colonization prevalence was 5.1% for H30R, 8% for ST131 (67% FQREC), and 10% for FQREC (52% ST131). ST131 isolates exhibited more VGs than non-ST131 isolates. Strain sharing (27% of multisubject households, 18% of corresponding subjects) was associated with the elderly, FQREC, H30R, H30Rx, ST73, and specific VGs. Fecal ST131 and FQREC isolates resembled contemporaneous and historical clinical isolates according to all studied traits.

Conclusions

Veterans and their human household members commonly carry and extensively share FQREC, predominantly H30R, thereby likely facilitating the ST131 pandemic. Strain sharing corresponds with multiple bacterial characteristics, including FQ resistance and specific VGs, which may promote intestinal colonization and/or host-to-host transmission.

Uninfected veterans and their household members commonly carried and shared fluoroquinolone-resistant gut E. coli, predominantly ST131-H30R, likely facilitating the ST131 pandemic. Strain sharing corresponded with fluoroquinolone resistance and specific “virulence genes,” which may also promote intestinal colonization and/or host-to-host transmission.

Extraintestinal Escherichia coli infections are an ever-growing problem that is aggravated by the rising prevalence of multidrug resistance, including to fluoroquinolones (FQs), which historically have been the preferred therapeutic agents for such infections [1, 2]. Most of the recent increase in FQ resistance in E. coli reflects the global emergence and expansion of a single lineage, the H30 subclone within sequence type 131 (ST131), and specifically its FQ-resistant H30R subset, which comprises sister clades H30R1 (usually extended-spectrum β-lactamase [ESBL] negative) and H30Rx (usually CTX-M-15 ESBL positive) [3, 4].

Why H30R expanded so dramatically is poorly understood, notwithstanding its numerous virulence genes (VGs) and resistance traits [5]. Because the extraintestinal pathogenic E. coli (ExPEC) strains that cause most extraintestinal infections usually derive proximally from the host’s own gut microbiota [6, 7], widespread gut colonization with a particular clone conceivably could underlie epidemic clonal emergence. Indeed, abundant evidence—albeit mostly from case studies, small series, and surveys of institutionalized and other exposed individuals—implicates H30R as a prevalent and persistent colonizer of uninfected hosts, among whom it can be shared within households, sometimes leading to clinical disease [8–18]. However, the available data address minimally the broader population prevalence and clinical relevance of such phenomena.

Given the tremendous public health importance of the H30R pandemic, and the fecal reservoir’s largely unexplored contribution to this phenomenon, we surveyed uninfected veterans and their human and animal household members for fecal carriage of ST131 and FQ-resistant E. coli (FQREC), characterized extensively the resulting strains, and compared them with reference clinical isolates. Specifically, we sought to determine the prevalence of fecal carriage of H30R, other FQREC, and ST131 per se, and of within-household strain sharing, in relation to characteristics of the household, host, and strain, and to clarify the likely clinical relevance of the identified H30R and FQREC strains.

METHODS

Study Design

Fecal donors included inpatient and outpatient veterans at the Minneapolis Veterans Affairs Medical Center (MVAMC) and their household members, as recruited from 2014 through 2016, according to an Institutional Review Board-approved protocol [16]. Inpatients were veterans who had been hospitalized within 30 days before study enrollment/sample collection; outpatients were all veterans under care at MVAMC and not hospitalized in the past 30 days. Newly discharged inpatients and randomly selected outpatients from MVAMC’s computerized patient database were invited to participate.

Interested veterans were invited to refer for study participation all cohabiting household members, including children and pets. All human subjects (or their surrogates) provided informed consent. Subjects received materials to facilitate at-home consent form completion and fecal swab collection/submission. A short questionnaire captured basic demographic data, including participants’ status as veteran, other adult, child, or pet, and for adults, age and sex.

Specimen Collection and Processing

Subjects collected a fecal sample using standardized procedures and mailed it in a transport swab at ambient temperature to the research laboratory. There, samples were screened for FQREC by selective cultures (using Tergitol-7 agar plates, with and without ciprofloxacin 4 µg/mL), and for ST131 by population DNA polymerase chain reaction (PCR) involving a sweep of mixed growth harvested from the culture plate and a highly sensitive and specific PCR assay [16].

Additional molecular testing

For fecal samples from households in which any subject carried FQREC and/or ST131 (hereafter, strains of interest [SOIs] and SOI households), and the first 62 remaining (ie, non-SOI) households, up to 10 E. coli colonies per sample (as available) were tested to determine the proportional contribution of ST131, FQREC, and other E. coli to the sample’s E. coli population [17]. Species identify (E. coli) was confirmed using indole and citrate phenotype. Strain identity was determined in a tiered manner, with an initial random amplified polymorphic DNA (RAPD) screen, then standardized Xbal pulsed-field gel electrophoresis (PFGE) analysis of 1 colony per RAPD profile per sample, an established and efficient clonal screening approach for deduplicating replicate colonies picked from fecal culture plates [17, 19]. Isolates were assigned to pulsotypes (strains) based on 94% PFGE profile similarity to reference profiles within a large private PFGE reference library (2382 pulsotypes, 7092 diverse-source E. coli isolates) [19].

A representative of each unique pulsotype per sample (1 sample per subject) underwent standardized PCR-based determination of major E. coli phylogroups, 12 key STs associated with ExPEC and/or antimicrobial resistance, H30 and H30Rx subclone status, and 53 putative VGs and associated variants [4, 16]. The VGs were used to calculate virulence scores (number of VG operons detected) and to determine presumptive ExPEC status (positive if ≥2 of [papAH and/or papC], sfa/focDE, afa/draBC, kpsM II, and iutA) [20].

Clinical Isolates

For a matched comparison of fecal and clinical isolates, 80 of the present fecal isolates and 80 contemporaneous unique-patient clinical E. coli isolates from the MVAMC microbiology laboratory, selected randomly per collection as 20 per ST131-by-FQR subset (ie, ST131/FQS, ST131/FQR, non-ST131/FQS, and non-ST131/FQR), were compared. Compared traits included PFGE profiles, antimicrobial resistance profiles for 22 agents (as determined by disk diffusion, using Clinical and Laboratory Standards Institute-specified procedures and interpretive criteria [21]), multidrug resistance (ie, resistance to ≥3 antimicrobial classes), ESBL production (as determined by double-disk diffusion), and PCR-determined phylogroups, STs, and VG profiles [4, 16].

Data Analysis

Data were analyzed using SPSS version 16 (IBM Analytics). Comparisons involving dichotomous variables were tested using Fisher exact test (2-tailed) or Χ² tests (if unpaired) or McNemar test (if paired); those involving continuous variables were tested using a t test (2-tailed) for 2-group comparisons and analysis of variance (ANOVA) for multigroup comparisons. Strain sharing, that is presence of a particular pulsotype in multiple members of a household, was assessed statistically at the level of the household, subject, and strain. Overall similarity relationships among isolates according to PFGE profiles, VGs, and resistance traits were depicted using (1) dendrograms, as inferred within Bionumerics (Applied Maths) according to the unweighted pair group method, and (2) a plot of the first 2 coordinates from a principle coordinates analysis of combined VGs and resistance markers, as implemented within R (The R Project for Statistical Computing: https://www.R-project.org/). Throughout, the significance criterion was P < .05.

RESULTS

Study Population

The 741 fecal donors included 383 veterans (171 inpatients, 212 outpatients) and 358 nonveteran household members (195 adults, 14 children, and 149 pets) (Table 1). They represented 383 households, including 178 (46%) single-subject households and 205 (54%) multisubject households. Of the multisubject households, 80 (39%) included at least 1 animal and 10 (5%) at least 1 child. Of the 578 adults, 381 (66%) were male and 342 (60%) were ≥65 years old (age range, 18–93 years).

Table 1.

Characteristics of Study Households and Subjects

Category	Characteristic	No. of Households or Subjects, Category (%)
Householdsa (n = 383)	Inpatient (vs outpatient) index veteran	171 (45)
	Single subject (vs multisubject)	178 (44)
	With pet	80 (21)
	With child	10 (3)
Subjects (n = 741)	From household of inpatient veteran	296 (40)
	From single-subject household	177 (24)
	Human	592 (80)
	Pet	149 (20)
	Adult	578 (78)
	Child	14 (2)
	Age ≥65 yearsb	342 (60)
	Race (for adults, n = 578)
	White	565 (98)
	Black	8 (1.3)
	Other	5 (0.7)
	Male sex (for adults, n = 545)c	381 (70)

^aHousehold size: mean, 1.9 (range, 1–10).

^bAge (for adults): mean, 65.2 years (range, 18–93 years).

^cData were available for 545/578 (95%) adult participants.

Prevalence of ST131

The overall by-subject ST131 colonization prevalence was 8% (58/741). Of the ST131 isolates, 39 (67%) were FQR, giving an FQR ST131 prevalence of 5% (39/741). ST131 prevalence varied significantly by host species (humans, 9% [56/592] vs pets, 1.3% [2/149; none FQR]; P < .001), but among humans did not differ by veteran status or, among adults, by sex or age (not shown).

Prevalence of FQREC

The overall by-subject FQREC colonization prevalence was 10% (75/741) and 52% (39/75) of FQREC isolates were ST131. (The ST distribution of the non-ST131 FQREC isolates is presented below.) As with ST131, FQREC prevalence varied significantly by host species (humans, 12% [71/592] vs pets, 3% [4/149; none ST131]; P < .001) but, among humans, did not vary significantly by veteran status or, among adults, by sex or age (not shown).

Household-Level Analysis

The by-household colonization prevalence was 14% for ST131 (55/383 households; 9% [16/178] if single subject; 19% [39/205] if multisubject; P = .006), 17% for FQREC (67/383 households; 15% [26/178] single subject; 14% [41/285] multisubject; P > .10), and 9% for FQR ST131, a proxy for H30R (see below) (34/383 households; 8% [14/178] single subject; 10% [20/205] multisubject; P > .10). FQREC was associated negatively with presence of pets (pets, 13% [10/80] vs no pets, 24% [30/125]; P = .04), but not children (P > .10), whereas ST131 was unassociated with either pets or children (P > .10 for each).

Strain Analysis

Of the 741 fecal samples, 296 (40%) underwent clonal analysis of multiple E. coli colonies to resolve discrete strains. The resulting 425 unique strains (per-sample mean, 1.2 strains; range, 1–4) included 61 (14%) ST131 strains (39 [64%] FQR) and 77 (18%) FQREC strains (39 [51%] ST131) (Table 2 and Table 3). The 61 ST131 strains included 41 (67%) H30 strains—nearly all H30R (93%, 38/41), with H30R1 (73%, 30/41) greatly outnumbering H30Rx (20%, 8/41) (P < .01, McNemar test)—and 20 (33%) non-H30 strains (11 O25b, 4 O16, 5 others; 19 FQS, 1 FQR). The total H30R prevalence was thus 5.1% (38/741) overall and 6.4% (38/591) among humans.

Table 2.

Molecular Characteristics in Relation to Sequence Type 131 (ST131) Status Among 425 Fecal Escherichia coli Strains

Characteristic		Prevalence of Characteristic, No. (Column %)			P Value
Category	Specific Traita,b,c	Total (n = 425d)	ST131 (n = 61)	Non-ST131 (n = 364)
Adhesins	papAH	85 (20)	5 (8)	80 (22)	.007
	sfa/focDE	69 (16)	2 (3)	67 (18)	.001
	sfaS	21 (5)	0 (0)	21 (6)	.04
	afa/draBC	11 (3)	5 (8)	6 (2)	.01
	iha	109 (26)	48 (79)	61 (17)	<.001
	hra	92 (22)	7 (12)	85 (23)	.02
	yfcV	207 (49)	56 (92)	151 (42)	<.001
Toxins	hlyD	65 (15)	3 (5)	62 (17)	.007
	sat	106 (25)	47 (77)	59 (16)	<.001
	pic	30 (7)	0 (0)	30 (8)	.008
	vat	129 (30)	1 (2)	128 (35)	<.001
	astA	27 (6)	0 (0)	27 (7)	.01
Siderophores	iroN	97 (23)	7 (12)	90 (25)	.01
	fyuA	268 (63)	60 (98)	208 (57)	<.001
	ireA	41 (10)	1 (2)	40 (11)	.01
	iutA	135 (32)	52 (85)	83 (23)	<.001
	chuA	306 (72)	61 (100)	245 (67)	<.001
Protectins	kpsM II	231 (54)	47 (77)	184 (50)	<.001
	K1	81 (19)	3 (5)	78 (21)	.001
	K5	42 (10)	25 (41)	17 (5)	<.001
	K2/K100	15 (4)	6 (9)	9 (3)	.01
	rfc	23 (5)	0 (0)	23 (6)	.03
	traT	184 (43)	47 (77)	137 (38)	<.001
Miscellaneous	usp	203 (48)	61 (100)	142 (39)	<.001
	ompT	280 (66)	57 (93)	223 (61)	<.001
	H7 fliC	38 (9)	0 (0)	38 (10)	.002
	malX	201 (47)	57 (93)	144 (40)	<.001
	clbB/N	81 (19)	1 (2)	80 (22)	<.001
Combination	ExPEC	178 (42)	43 (70)	135 (39)	.007

^aTraits shown are those that yielded P < .05 (by Fisher’s Exact Test) for the 2-group comparison. Boldface indicates higher prevalence group. Definitions: papAH, P fimbriae major subunit; sfa/focDE, S and F1C fimbriae; sfaS, S fimbriae adhesin; afa/draBC, Dr-binding adhesins; iha, adhesin-siderophore receptor; hra, heat-resistant agglutinin; yfcV, putative chaperone-usher fimbria; hlyD, alpha hemolysin; sat, secreted autotransporter toxin; pic, protein associated with intestinal olonization; vat, vacuolating toxin; astA, enteroaggregative E. coli toxin; iroN, salmochelin receptor; fyuA yersiniabactin receptor; ireA, catecholate siderophore receptor; iutA, aerobactin receptor; chuA, heme-binding outer membrane; kpsM II, group 2 capsules; K2, K5, K2/K100, group 2 capsule variants; rfc, O4 lipopolysaccharide; traT, serum-resistance associated; usp, uropathogenic-specific protein; ompT, outer membrane protease; H7 fliC, flagellar variant; malX, pathogenicity-associated island marker; clbB/N, colibactin (polyketide) synthesis; ExPEC, extraintestinal pathogenic E. coli.

^bTraits detected in ≥1 isolate but not yielding P < .05 (% of 425): focG (F1C fimbriae adhesin, 7%), gafD (G fimbriae, 1%), fimH (type-1 fimbriae adhesin, 97%), hlyF (variant hemolysin, 7%), cnf1 (cytotoxic necrotizing factor 1, 12%), cdtB (cytolethal distending toxin, 6%), kpsMT III (group 3 capsules, 3%), cvaC (microcin V, 4%), ibeA (invasion of brain endothelium A, 15%), iss (increased serum survival, 7%).

^cTraits sought but not detected in any isolate: afaE8, F17, clpG, bmaE (variant mannose-resistant adhesins).

^dThe 425 strains were selected to include all strains of interest, all other strains from strain-of-interest households, and all strains from the first 62 non-strain-of-interest households.

Table 3.

Molecular Characteristics in Relation to Resistance Category Among 425 Fecal Escherichia coli Strains

Strain Characteristic		No. of Strains (Column %)			P Value
Categorya,b,c	Trait	Total (n = 425f)	FQR (n = 77)	FQS (n = 348)
Adhesin genes	papAH	85 (20)	5 (7)	80 (23)	<.001
	sfa/focDE	69 (16)	0 (0)	69 (20)	<.001
	sfaS	21 (5)	0 (0)	21 (6)	.01
	focG	31 (7)	0 (0)	31 (9)	.002
	iha	109 (26)	53 (69)	56 (16)	<.001
	fimH	411 (97)	71 (92)	340 (98)	.01
	hra	92 (22)	9 (12)	83 (24)	.01
	yfcV	207 (49)	54 (70)	153 (44)	<.001
Toxin genes	hlyD	65 (15)	0 (0)	65 (19)	<.001
	hlyF	30 (7)	0 (0)	30 (9)	.002
	cnf1	51 (12)	0 (0)	51 (15)	<.001
	cdtB	27 (6)	0 (0)	27 (8)	.004
	sat	106 (25)	57 (74)	49 (14)	<.001
	pic	30 (7)	0 (0)	30 (9)	.002
	vat	129 (31)	7 (9)	122 (35)	<.001
	tsh	22 (5)	0 (0)	22 (6)	.01
Siderophore genes	iroN	96 (23)	0 (0)	96 (28)	<.001
	fyuA	267 (63)	68 (88)	199 (58)	<.001
	ireA	41 (10)	2 (3)	39 (11)	.01
	iutA	134 (32)	59 (77)	75 (22)	<.001
	chuA	304 (72)	66 (86)	238 (67)	.001
Protectin genes	K5	42 (10)	24 (31)	18 (5)	<.001
	iss	27 (6)	0 (0)	27 (8)	.004
	rfc	23 (5)	0 (0)	23 (7)	.009
	cvaC	17 (4)	0 (0)	17 (5)	.03
	traT	183 (43)	45 (58)	138 (40)	.002
Miscellaneous virulence genes	usp	203 (48)	49 (64)	154 (45)	.002
	ibeA	65 (15)	2 (3)	63 (18)	<.001
	ompT	279 (66)	58 (75)	221 (64)	.04
	H7 fliC	38 (9)	0 (0)	38 (11)	<.001
	malX	201 (48)	57 (74)	144 (42)	<.001
	clbB/N	81 (19)	1 (1)	80 (23)	<.001
Pathotype (gene combination)	ExPEC	178 (42)	45 (58)	133 (38)	.001
Phylogroupd	B1	60 (14)	5 (7)	55 (16)	.03
	B2	201 (48)	48 (62)	153 (44)	.004
Clonal groupe	STc12	21 (5)	0 (0)	21 (6)	.02
	ST131	61 (15)	39 (51)	22 (6)	<.001
	ST131-non-H30	20 (5)	1 (1)	19 (5)	<.001
	ST131-H30	41 (7)	38 (49)	3 (1)	.001
	ST131-H30Rx	8 (2)	8 (10)	0 (0)	.001
	ST648	5 (1)	3 (4)	0 (0)	.006
	ST73	28 (7)	0 (0)	28 (8)	.004
	STc95	18 (4)	0 (0)	18 (5)	.048

Abbreviations: ExPEC, extraintestinal pathogenic Escherichia coli; FQR, fluoroquinolone resistant; FQS, fluoroquinolone susceptible; ST, sequence type; STc, ST complex (group of closely related STs).

^aGenes shown are those that yielded P < .05 (by Fisher’s Exact Test) for the 2-group comparison. Boldface, higher prevalence group. Gene definitions: papAH/C/EF/G, P fimbriae; sfa/focDE, S and F1C fimbriae; sfaS, S fimbriae adhesin; focG, F1C fimbriae adhesin; afa/draBC, Dr-binding adhesins; iha, adhesin-siderophore receptor; fimH, type-1 fimbriae adhesin; hra, heat-resistant agglutinin; yfcV, putative chaperone-usher fimbria; hlyD, alpha hemolysin; hlyF, variant hemolysin; cnf1, cytotoxic necrotizing factor 1; cdtB, cytolethal distending toxin; sat, secreted autotransporter toxin; pic, protein associated with intestinal colonization; vat, vacuolating toxin; tsh, temperature-sensitive hemagglutinin; iroN, salmochelin receptor; fyuA, yersiniabactin receptor; ireA, catecholate siderophore receptor; iutA, aerobactin receptor; chuA, heme-binding outer membrane; K5, group 2 capsule variants; iss, increased serum survival; rfc, O4 lipopolysaccharide; cvaC, microcin V; usp, uropathogenic-specific protein; traT, serum resistance-associated; ibeA, invasion of brain endothelium A; ompT, outer membrane protease; H7 fliC, flagellar variant; malX, pathogenicity island marker; clbB/N, colibactin (polyketide) synthesis.

^bGenes detected in ≥1 isolate but not yielding P < .05 (% of 425): afa/draBC (Dr-binding adhesins, 3%), gafD (G fimbriae, 1%), astA (enteroaggregative E. coli toxin, 6%), kpsM II (group 2 capsules, 54%), kpsMT III (group 3 capsules, 3%), K2/K100 (group 2 capsule variants, 4%)

^cGenes sought but not detected: afaE8, F17, clpG, bmaE (variant mannose-resistant adhesin).

^dPhylogroups detected in ≥1 isolate but not yielding P < .05 (% of 425): A (13%): C (.2%), D (18%), E (2%), F (4%). Phylogroup was indeterminate for 1% of isolates.

^eClonal groups detected in ≥1 isolate but not yielding P < .05 (% of 425): ST127 (3%), ST141 (3%), ST144 (1%), ST372 (2%), STc405 (2%), STc69/CGA (5%). Each of the 12 clonal groups sought was detected at least in one isolate.

^fThe 425 strains were selected to include all strains of interest, all other unique strains from strain-of-interest households, and all strains from the first 62 non-strain-of-interest households.

Virulence Genotypes

Among the 425 strains that underwent virulence genotyping, ST131 strains (vs non-ST131 strains) exhibited similar numbers of positive and negative associations with VGs (14 vs 17) (Table 2), whereas FQR strains (vs FQS strains) exhibited fewer than half as many positive as negative associations (11 vs 24) (Table 3). Accordingly, compared with non-ST131 strains, ST131 strains had higher VG scores (mean, 9.9 vs 7.0; P < .001) and more often qualified as ExPEC (56% vs 39%; P = .007), whereas compared with FQS strains, FQR strains had only slightly higher VG scores (mean, 7.7 vs 7.4; P = .62), although more often qualified as ExPEC (58% vs 38%; P = .001). Accordingly, FQR ST131 strains (nearly all H30R) had significantly higher VG scores (mean, 9.2) than non-ST131 FQR strains (mean, 5.5; P = .01), and comparably high scores to FQS strains, whether ST131 or non-ST131 (not shown).

Phylogenetic/Clonal Distribution of Non-ST131 Fecal FQREC

Among the FQR strains, H30R was by far the most prevalent lineage (51%, n = 39), outnumbering entire STs by more than 5-fold, and entire phylogroups by more than 3-fold (P < .05, McNemar test). Specifically, the 38 non-ST131 FQREC strains were divided between groups D (29% [n = 11; 3 STc69, 3 ST405, 3 STc31, 2 other STs]), B2 (24% [n = 9; 7 ST1193, 2 other STs]), F (18% [n = 7; 5 ST648, 2 other STs]), A (16% [n = 6]), and B1 (13% [n = 5]).

Abundance

The mean abundance of individual strains within each sample’s E. coli population was 68% and did not differ significantly by ST131 status or FQR phenotype (P > .05, ANOVA).

Strain Sharing

Strain sharing appeared in 27 (27%) of the 100 assessed multisubject households, which included 248 total subjects. By household, sharing (generically) was unassociated with presence of children or pets, household size, or SOI status. By subject, it involved 44 (18%) of 248 subjects, was more frequent among subjects from SOI households (32% [22/69] vs 19% [34/179]; P = .04), and among adults was associated with greater host age (median age, 70 vs 65 years; P = .004); by contrast, it was unassociated with species, adult/child status, or sex.

By strain, 59 (17%) of 350 unique strains were shared. Sharing frequency was associated significantly with FQREC, H30R, ST131-H30Rx, and ST73 (Table 4), but not phylogroup or ST131 status per se (Table 4). Compared with nonshared strains, shared strains had a significantly higher prevalence of 12 VGs and a lower prevalence of none (Table 4), hence had significantly higher VG scores (mean, 8.9 vs 7.4; P = .01) and more frequently qualified as ExPEC (58% vs 38%; P = .001).

Table 4.

Characteristics Associated with Strain Sharing Among 350 Escherichia coli Isolates from Multisubject Households

Strain Characteristics		Trait Prevalence, No. of Strains (Column %)			P value, Shared vs Nonshared
		Total (n = 350)	Shared (n = 59)	Nonshared (n = 291)
Category	Trait
Resistance	Fluoroquinolone	51 (15)	16 (27)	35 (12)	.004
Virulence genesa,b,c	iha	84 (24)	20 (34)	64 (22)	.04
	gafD	2 (1)	2 (3)	0 (0)	.03
	yfcV	170 (49)	37 (63)	133 (46)	.01
	sat	78 (22)	20 (34)	58 (20)	.02
	pic	27 (8)	9 (15)	18 (6)	.02
	fyuA	218 (62)	48 (81)	170 (58)	.001
	iutA	107 (31)	27 (46)	80 (28)	.005
	kpsM II	191 (55)	41 (70)	150 (52)	.008
	K5	31 (9)	12 (20)	19 (7)	.002
	usp	165 (47)	35 (59)	130 (45)	.03
	H7 fliC	34 (10)	10 (17)	24 (8)	.04
	malX	163 (47)	34 (58)	129 (44)	.04
	ExPEC	178 (51)	45 (76)	133 (38)	.001
Clonal group(s)d	ST131	45 (13)	10 (17)	35 (12)	.20
	ST131-H30R	25 (7)	8 (14)	17 (6)	.049
	ST131-H30Rx	7 (2)	4 (7)	3 (1)	.02
	STc73	25	9 (15)	16 (5)	.02

Abbreviations: ExPEC, extraintestinal pathogenic Escherichia coli; FQREC, fluoroquinolone-resistant E. coli; FQSEC, fluoroquinolone-susceptible E. coli; ST, sequence type; ST131, sequence type 131; STc, ST complex (group of closely related STs); VG, virulence gene.

^aVGs shown are those that yielded P < .05 (by Fisher’s Exact Test) for the 2-group comparison. Boldface, higher prevalence group (all prevalence differences favored the shared strains). Traits definitions: iha, adhesin-siderophore receptor; gafD, G fimbriae; yfcV, putative chaperone-usher fimbria; sat, secreted autotransporter toxin; pic, protein associated with intestinal colonization; fyuA, yersiniabactin receptor; iutA, aerobactin receptor; kpsM II, group 2 capsules; K5, group 2 capsule variants; usp, uropathogenic-specific protein; H7 fliC, flagellar variant; malX, pathogenicity island marker.

^bVGs detected in ≥1 isolate but not yielding P < .05: papAH/C/EF/G (P fimbriae: 21%–24%, respectively), sfa/focDE (S and F1C fimbriae, 18%), sfaS (S fimbriae adhesin, 5%), focG (F1C fimbriae adhesin, 9%), afa/draBC (Dr-binding adhesins, 3%), fimH (type-1 fimbriae adhesin, 98%), hra (heat-resistant agglutinin, 23%), hlyD (alpha hemolysin, 17%), hlyF (variant hemolysin, 8%), cnf1 (cytotoxic necrotizing factor 1, 13%), cdtB (cytolethal distending toxin, 7%), vat (vacuolating toxin, 31%), astA (enteroaggregative E. coli toxin, 7%), tsh (temperature-sensitive hemagglutinin, 6%), iroN (salmochelin receptor, 25%), ireA (catecholate siderophore receptor, 11%), chuA (heme-binding outer membrane, 73%), kpsMT III (group 3 capsules, 3%), K1, K2, K2/K100 (group 2 capsule variants, 20%, 51%, 4%, respectively), iss (increased serum survival, 7%), rfc (O4 lipopolysaccharide, 6%), cvaC (microcin V), traT (serum resistance-associated, 44%), ibeA (invasion of brain endothelium A, 16%), ompT (outer membrane protease, 66%), clbB/N (colibactin (polyketide) synthesis, 20%).

^cTraits sought but not detected in any isolate: afaE8, F17, clpG, bmaE (variant mannose-resistant adhesins).

^dClonal groups detected in ≥1 isolate but not yielding P < .05 (% of 350): STc12 (5%), ST127 (2%), ST141 (4%), ST144 (1%), ST372 (1%), STc405 (2%), STc648 (1%), STc95 (5%), STc69/CGA (5%). Each of the 12 clonal groups sought was detected in at least one isolate.

Comparison With Clinical Isolates

Matched comparisons were made between 80 of the present fecal study isolates and 80 contemporaneous clinical (mainly urine) E. coli isolates from the MVAMC microbiology laboratory, as selected randomly from the respective parent collections within 4 subsets as defined by a combination of ST131 and FQR status. According to PFGE profiles, VGs, and resistance markers, the 4 (ST131 vs FQR) subsets differed extensively from one another, whereas within a given (ST131 vs FQR) subset the fecal and clinical isolates differed minimally (Figure 1, Supplementary Tables 1–2, and Supplementary Figure 1). For example, for resistance markers, 15/42 (36%) of comparisons across the 4 (ST131 vs FQR) subsets yielded P < .05, as compared with only 3/84 (4%) of comparisons between same-subset fecal and clinical isolates (P < .001) (Supplementary Table 1). Similarly, for VG prevalence, 58/92 (63%) of comparisons across the 4 (ST131 × FQR) subsets yielded P < .05, vs only 16/200 (8%) of comparisons between same-subset fecal and clinical isolates (P < .001) (Supplementary Table 2).

Figure 1.

Principal coordinates analysis of virulence genes and resistance markers. The 160 total Escherichia coli isolates included, per (ST131 vs FQR) subset, 20 each contemporaneous clinical isolates (Minneapolis Veterans Affairs Medical Center) and fecal isolates (this study). The plot is based on the first 2 coordinates (X1 and X2), which collectively capture 24% of variance in the dataset. The axes are unitless. On the X-Y grid, the 4 (ST131 vs FQR) subsets overlap minimally, whereas within each subset the clinical and fecal isolates overlap extensively. Abbreviations: FQR, fluoroquinolone-resistant; FQS, fluoroquinolone-susceptible.

Fecal and clinical isolates also overlapped considerably with respect to phylogroup, ST, and ST131 clonal subset (metadata, Supplementary Figure 1). Among non-ST131 strains, shared STs included ST73, ST95, and ST127 (if FQS), and ST14, ST31, and ST405 (if FQR). Likewise, among ST131 strains, shared clonal subsets included O16 and O25b non-H30 (if FQS), and H30R and H30Rx (if FQR).

For a broader assessment, pulsotypes were compared between all 425 present pulsotyped fecal isolates (332 pulsotypes) and 925 arbitrarily selected diverse-source clinical isolates (336 pulsotypes) from a large private PFGE library, stratified by (ST131 × FQR) subset. Within each subset, multiple pulsotypes among the fecal isolates—including 2 or 3 of the 3 top pulsotypes—occurred also among the clinical isolates. Overall, 14% of fecal isolates matched a clinical isolate (90% if FQR/ST131, 37% if FQR/non-ST131, 27% if FQS/ST131, and 5% if FQS/non-ST131), whereas 54% of clinical isolates matched a fecal isolate (67% if FQR/ST131, 41% if FQR/non-ST131, 12% if FQS/ST131, and 10% if FQS/non-ST131).

DISCUSSION

This large cross-sectional prevalence survey yielded 4 main findings, each with potentially important implications for the current H30R and FQREC pandemic. First, FQREC—predominantly H30R—commonly colonized veterans and their (mostly adult) human household members, evidence of a large community reservoir of such strains. Second, colonization with ST131 was unassociated with host sex or veteran status, illustrating the breadth of the at-risk population. Third, fecal and clinical FQR ST131 isolates were similar according to VG content, resistance profiles, subclones, and PFGE profiles, supporting the fecal E. coli population as an important source of such MDR pathogens. Indeed, comparative prevalence values suggested that, relative to other FQREC, H30R is likely both a better gut colonizer and more capable of causing infections in colonized hosts. Fourth, within-household strain sharing was especially common for FQR, H30R, and H30Rx strains, and was associated with higher VG scores, ExPEC status, and older hosts, suggesting possible contributors to strain dissemination. Thus, colonization fitness, virulence, and within-household strain sharing—all of which may be promoted by H30R’s extensive array of VGs—may underlie H30R’s epidemic success and could be targets for preventive interventions.

Our first main finding was the substantial prevalence of fecal colonization with FQREC (10% overall, 12% among humans) and, among fecal FQREC, the large ST131 fraction (51%), nearly all from the H30R subclone (total H30R prevalence among humans, 6.4%). These values are substantially higher than documented in the few prior ST131 fecal surveys that did not focus on ESBL production, infected or institutionalized individuals, or contacts of a known source case (eg, for FQR ST131, overall prevalences of 0%, 0.8%, and 1.2%) [22–24]. These prior studies were from before 2012, suggesting recent expansion of H30R as a population-wide fecal colonizer, paralleling its expansion among clinical isolates. Here, H30R was by far the leading FQREC lineage among fecal E. coli isolates, mirroring its prominence among contemporaneous clinical E. coli isolates at MVAMC, other VAMCs nationally, and non-VA US centers [4, 25, 26].

Notably, however, H30R’s reported relative prevalence among clinical FQREC isolates (71%–78%) has typically been even higher than we observed among fecal FQREC isolates (51%) [4, 25, 27]. This clinical-to-fecal disparity suggests that, as compared with other FQREC, H30R not only is a more successful gut colonizer but, when colonizing the gut, also is more likely to be encountered clinically at an extraintestinal site. This supports a 2-stage model for the H30R pandemic that unites the “prevalence” and “special pathogenicity” hypotheses for UTI pathogenesis: superior gut colonization ability creates a large H30R fecal reservoir, then superior ability to exit the gut and cause extraintestinal infection produces a disproportionately large H30R disease burden.

Our second main finding, the presence of ST131 and FQREC within all human subsets, corresponded with our identification of few epidemiological correlates of such colonization. Thus, notwithstanding such strains’ clear predilection for the elderly and for humans over pets, they colonize broadly, consistent with well-documented community-acquired, pediatric, and veterinary infections and colonization involving such strains [9, 15, 17, 18, 28–31]. Although the several risk factors that have been identified for colonization and infection with ST131 and FQREC [13, 18, 26, 32, 33] may offer opportunities for targeted interventions, an overarching perspective might be that these strains have become ubiquitous, obliging community-wide responses.

Our third main finding, that is the considerable similarity of fecal and clinical isolates of FQREC and ST131, supports that, despite their obvious niche differences, fecal and clinical strains represent overlapping bacterial populations, with clinical isolates likely emerging from the gut reservoir. We demonstrated this commonality in multiple ways. First, with respect to VG content, resistance profiles, and PFGE profiles, the present fecal closely resembled contemporaneous MVAMC clinical E. coli isolates, stratified by FQR and ST131 status. Second, the present fecal isolates (especially if FQR) exhibited extensive pulsotype commonality with reference clinical isolates from diverse locales and time periods [19]. Third, the present fecal ST131 isolates exhibit broad VG profile commonality with published ST131 clinical isolates, including a >80% prevalence of a distinctive set of VGs (eg, iha, sat, fyuA, iutA, kpsM II, traT, usp, ompT, and malX) and a <10% prevalence of others (eg, sfa/foc, hra, hlyD, pic, vat, iroN, ireA, K1, H7 fliC, and clbB/N) [4, 25, 34, 35].

This clinical vs fecal commonality indicates that studies of fecal FQREC and ST131 likely address the substrate for clinical infections caused by these strains. This makes such studies potentially critical for understanding and intervening on pathogenesis, transmission, and clonal emergence, phenomena of obvious clinical importance.

Our fourth main finding involved strain sharing, which preferentially involved FQREC, H30R, and H30Rx strains. It suggests that host-to-host transmission likely contributes to dissemination of such strains within the population, offering a potential point of attack on the problem. Previous studies have documented within-household sharing of ST131 strains among contacts of ST131-infected patients (eg, [15]) and within healthy maternal-child pairs [18]. Our study provides what we believe is the first assessment of such strain sharing generally within households without a known clinical case or newborn, and in relation to characteristics of the household, subject, and strain.

In that regard, multiple VGs were associated with shared, ST131, and FQREC strains, evidence suggesting that the corresponding gene products (or linked traits) possibly promote not just virulence but also gut colonization [36, 37] and host-to-host transmission [38–40]. Such traits potentially could serve as targets for preventive measures, for example vaccines [41], to reduce carriage and sharing of potential pathogens, thereby decreasing the risk of both infection and clonal dissemination.

This study has multiple strengths, including its large sample size, systematic recruitment of veterans and household members (including animals and children), extensive isolate characterization, and comparisons of fecal isolates with contemporaneous and archival clinical isolates—a notable gap in prior work. It also has limitations, including its focus on veterans, single-center nature, certain small subgroups (eg, children), limited epidemiological data (including regarding recent antibiotic use), reliance on nonsequence-based typing methods, observational (ie, correlative, noninterventional) nature, and cross-sectional design. Of note, a longitudinal follow-on study is ongoing.

In conclusion, as an intestinal colonizer of uninfected veterans and their human household members, H30R was much more prevalent than any other FQREC strain, largely independent of the studied host characteristics. Comparative prevalence values suggested that, relative to other FQREC, H30R is likely both a better gut colonizer and more capable of causing infections in colonized hosts, phenotypes possibly promoted by H30R’s extensive repertoire of virulence factors, which may also represent colonization factors. Thus, colonization fitness, virulence, and within-household strain sharing may all underlie H30R’s epidemic success and could be targets for preventive interventions.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. Gary Dunny (University of Minnesota), Ruth Lynfield (Minnesota Department of Health), and Dimitri Drekonja (Minneapolis VA Medical Center [MVAMC] and University of Minnesota) provided helpful suggestions. The MVAMC clinical microbiology laboratory provided clinical isolates. Billie S. Slater provided study coordinator assistance.

Financial support. This work was supported by the Office of Research and Development, Department of Veterans Affairs (grant numbers 1 I01 CX000920-01 and 2I01CX000920-04 to J. R. J.); and the National Institute of Allergy and Infectious Diseases, National Institutes of Health (grant number UM1AI104681 Antibacterial Resistance Leadership Group, to J. R. J.).

Potential conflicts of interest. J. R. J. reports grants or consultancies from Achaogen, Allergan, Janssen/Crucell, Melinta, Merck, Shionogi, Syntiron, and Tetraphase, and has patent application for tests to detect specific E. coli strains. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.