Clonal Composition and Community Clustering of Drug-Susceptible and -Resistant Escherichia coli Isolates from Bloodstream Infections

Abstract

Multidrug-resistant Escherichia coli strains belonging to a single lineage frequently account for a large proportion of extraintestinal E. coli infections in many parts of the world. However, limited information exists on the community prevalence and clonal composition of drug-susceptible E. coli strains. Between July 2007 and September 2010, we analyzed all consecutively collected Gram-negative bacterial isolates from patients with bloodstream infection (BSI) admitted to a public hospital in San Francisco for drug susceptibility and associated drug resistance genes. The E. coli isolates were genotyped for fimH single nucleotide polymorphisms (SNPs) and multilocus sequence types (MLSTs). Among 539 isolates, E. coli accounted for 249 (46%); 74 (30%) of them were susceptible to all tested drugs, and 129 (52%) were multidrug resistant (MDR). Only five MLST genotypes accounted for two-thirds of the E. coli isolates; the most common were ST131 (23%) and ST95 (18%). Forty-seven (92%) of 51 ST131 isolates, as opposed to only 8 (20%) of 40 ST95 isolates, were MDR (P < 0.0001). The Simpson's diversity index for drug-susceptible ST genotypes was 87%, while the index for MDR ST genotypes was 81%. ST95 strains were comprised of four fimH types, and one of these (f-6) accounted for 67% of the 21 susceptible isolates (P < 0.003). A large proportion (>70%) of both MDR and susceptible E. coli BSI isolates represented community-onset infections. These observations show that factors other than the selective pressures of antimicrobial agents used in hospitals contribute to community-onset extraintestinal infections caused by clonal groups of E. coli regardless of their drug resistance.

INTRODUCTION

Recently, much attention has been given to the worldwide dissemination of extraintestinal pathogenic Escherichia coli (ExPEC) strains belonging to a related lineage, such as the drug-resistant E. coli 025b:H4 multilocus sequence type (MLST) ST131 (1–7). Retrospective studies have shown that from as early as 2000, multidrug-resistant (MDR) ST131 strains were implicated in a large proportion of community-acquired urinary tract infections (UTI) and bloodstream infections (BSI) in different regions of the world (5, 6, 8–10). This global dispersion of ST131 has been suggested to be mediated by activities such as global food trade, international travel, and other modes of transmission (5, 6, 10, 11).

There are other ExPEC strains belonging to a limited number of lineages that circulate globally (2, 4–7, 12, 13). However, most published reports on these lineages have focused on drug-resistant strains. The epidemiologic observation that certain drug-resistant E. coli lineages predominate in community and institutional settings is unexplained. One explanation may be that most studies focus on drug-resistant infections and that such study results are more likely to be reported. Another explanation is that drug-resistant strains have selective advantage because of their drug resistance. Finally, it is possible that they become predominant because of epidemiologic or biologic factors unrelated to drug resistance. We reasoned that if drug-susceptible E. coli strains are also found to exhibit clonal distribution in community settings, this would indicate that factors other than drug resistance contribute to the clonal spread of ExPEC. This study was undertaken to compare the clonal compositions of drug-resistant and -susceptible E. coli clinical isolates from patients with BSI who were treated at a large public hospital in San Francisco.

MATERIALS AND METHODS

Strain collection.

The clinical microbiology laboratory of San Francisco General Hospital (SFGH) is one of the largest in the Bay Area, and the cultures originate from all the hospital wards and jail clinics and San Francisco's city outpatient clinics. All Gram-negative bacillus (GNB) isolates from BSI at the clinical microbiology laboratory of SFGH are stored on agar slants for up to 3 months. Every 3 months, these isolates were provided to us for analysis. We examined all consecutively collected GNB BSI isolates from inpatients admitted to SFGH between July 2007 and September 2010. This study was approved by the University of California, San Francisco, Committee on Human Research.

The information on the dates of admission and first blood culture that yielded the GNB pathogen was available for most of the isolates. In this study, we considered BSI to be community onset if the time period between the date of admission and the date of the first blood culture that grew a GNB pathogen was ≤48 h.

Strain identification and susceptibility testing.

At SFGH, the BSI isolates were identified to the species level biochemically with API 20E (bioMérieux, Durham, NC) for fermenters or API 20NE for nonenteric bacteria. Antimicrobial susceptibility tests were performed by a MicroScan WalkAway Gram-negative panel (Dade Behring/Siemens USA, Deerfield, IL). This panel included the following 11 classes of antimicrobial agents (the drugs tested within each): aminoglycosides (amikacin, gentamicin, tobramycin), aminopenicillins (ampicillin), β-lactamase inhibitor combinations (ampicillin-sulbactam, piperacillin-tazobactam, ticarcillin-clavulanic acid), broad-spectrum 1st-generation cephalosporins (cefazolin), broad-spectrum 2nd-generation cephalosporins (cefotetan, cefuroxime, cefoxitin), extended-spectrum 3rd-generation cephalosporins (cefotaxime, ceftriaxone, ceftazidime), extended-spectrum 4th-generation cephalosporins (cefepime), fluoroquinolones (ciprofloxacin, levofloxacin, moxifloxacin), folate path inhibitors (trimethoprim-sulfamethoxazole), monobactams (aztreonam), and carbapenems (ertapenem, imipenem, meropenem). E. coli multidrug resistance (MDR) was defined as resistance to one or more agents in three or more classes of tested drugs (8, 14). Intermediate susceptibility was classified as resistant. Etest (bioMérieux, Marcy l'Etoile, France) was used to confirm susceptibility to carbapenems for select isolates.

DNA extraction and PCR amplification for gene detection.

The bacterial DNA was extracted by a freeze-boil method, and PCR amplification was carried out as previously described (15). Primer sequences for PCR analysis of fimH genotyping, β-lactamase genes, class 1 integron, and associated gene cassettes were used as previously described (Table 1).

Table 1.

PCR primers used in this study

Primer target	Primer namea	Sequence (5′–3′)	Expected product size (bp)	Reference
Class 1 integrase	201F	CCTCCCGCACGATGATC	280	1
	202R	TCCACGCATCGTCAGGC
Class 1 gene cassette, variable region	5′ CS	GGCATCCAAGCAGCAAG	Variable	25
	3′ CS	AAGCAGACTTGACCTGA
fimH SNP	fimH F	TCGAGAACGGATAAGCCGTGG	506	49
	fimH R	GCAGTCACCTGCCCTCCGGTA
16S ribosomal DNA, positive control	16S8F	AGAGTTTGATCCTGGCTCAG	800	1
	16S8R	GGACTACCAGGGTATCTAATCC
ESBL multiplex
TEM variants, including TEM-1 and -2	MultiTSO-T_for	CATTTCCGTGTCGCCCTTATTC	800	7
	MultiTSO-T_rev	CGTTCATCCATAGTTGCCTGAC
SHV variants, including SHV-1	MultiTSO-S_for	AGCCGCTTGAGCAAATTAAAC	713	7
	MultiTSO-S_rev	ATCCCGCAGATAAATCACCAC
OXA-1, -4, and -30	MultiTSO-O_for	GGCACCAGATTCAACTTTCAAG	564	7
	MultiTSO-O_rev	GACCCCAAGTTTCCTGTAAGTG
CTX-M universal	CTX-M Univ F	TTTGCGATGTGCAGTACCAGTAA	500	45
	CTX-M Univ R	CTCCGCTGCCGGTTTTATC
CTX-M families	CTX-M-1-all F	ATGGTTAAAAAATCACTGCG	876	45
	CTX-M-1-all R	TTACAAACCGTCGGTGACGAT
	CTX-M-9-F	GCAGATAATACGCAGGTG	392	45
	CTX-M-9-R	CGGCGTGGTGGTGTCTCT
CTX-M multiplex
Variants including CTX-M-1, -3, and -15	MultiCTXMGp1_for	TTAGGAARTGTGCCGCTGYA	688	7
	MultiCTXMGp1_rev	CGATATCGTTGGTGGTRCCAT
Variants including CTX-M-2	MultiCTXMGp2_for	CGTTAACGGCACGATGAC	404	7
	MultiCTXMGp2_rev	CGATATCGTTGGTGGTRCCAT
Variants including CTX-M-9 and -14	MultiCTXMGp9_for	TCAAGCCTGCCGATCTGGT	561	7
	MultiCTXMGp9_rev	TGATTCTCGCCGCTGAAG
Variants including CTX-M-8, -25, -26, -39, and -41	MultiCTXMGp8_for	AACRCRCAGACGCTCTAC	326	7
	MultiCTXMGp8_rev	TCGAGCCGGAASGTGTYAT
pabB	pabB F	TCCAGCAGGTGCTGGATCGT	347	3
	pabB R	GCGAAATTTTTCGCCGTACTGT

^aF, forward primer; R, reverse primer.

Sequence analysis.

A 5-μl aliquot of each PCR product that appeared as a unique band when visualized by gel electrophoresis was cleaned with 2 μl of a 1:10 dilution of ExoSAP-IT (Affymetrix, Inc., Santa Clara, CA) and subjected to direct sequencing. Sequencing was done on an Applied Biosystems 3730 DNA analyzer (Applied Biosystems, Foster City, CA) at the University of California, Berkeley, DNA Sequencing Facility. The DNA sequences were visually inspected, edited, and assembled with BioEdit version 7.0.1. ClustalW was used to perform multiple alignment analyses of the sequences. Sequences were compared with those deposited in the National Center for Biotechnology Information (NCBI) database by an updated version of the BLAST program.

Genotyping of E. coli strains.

We genotyped E. coli isolates by two methods. One was a single nucleotide polymorphism (SNP) analysis of the fimH locus that encodes a mannose-binding subunit protein located at the tip of type 1 fimbriae, performed according to a previously published report (17). Designation of fimH allele types followed the classification system established by Dias et al. (21); the SNP analysis of fimH can further discriminate strains genotyped by multilocus sequence typing (MLST). The other method was MLST based on a protocol published on the University College Cork website (http://mlst.ucc.ie/mlst/). The MLST analysis is based on sequence comparison of a portion of the following seven housekeeping genes: adk, fumC, recA, mdh, purA, gyrB, and icd.

Allele sequences of these genes were submitted to the MLST database curator for sequence type (ST) and sequence complex assignment. Sequence complexes include STs that differ from each other by no more than two of the seven loci, as defined by Wirth et al. (22). Strains that were genotyped as ST131 were based on all seven MLST alleles if all of the targets yielded interpretable sequences or a combination of at least three MLST alleles and ST131 allele-specific PCR for the pabB gene as established by Clermont et al. (20).

Statistics.

Categorical variables were compared by a chi-square or Fisher exact test (2-tailed). Genotype diversity was analyzed by Simpson's diversity index (23).

Nucleotide sequence accession numbers.

The fimH gene sequences from this study that were found to be unique (partial coding sequence [CDS] corresponding to the bp positions 360 to 781 of the E. coli strain K-12 sequence [GenBank accession number GQ487190]) were deposited in GenBank under accession numbers KC020174 to KC020181.

RESULTS

Bacterial species identified from patients with BSI.

Of 539 Gram-negative bacterial BSI isolates collected between July 2007 and September 2010, five species comprised 76% of the population. E. coli was the most frequent species with 249 (46%) isolates, followed by 70 (13.0%) Klebsiella pneumoniae isolates, 35 (7%) Proteus mirabilis isolates, 31 (6%) Pseudomonas aeruginosa isolates, and 26 (5%) Enterobacter cloacae isolates. This study focuses on the E. coli BSI isolates.

Antimicrobial drug resistance.

Susceptibility results for 11 classes of antimicrobial agents were available for 246 (99%) E. coli isolates. The following results are based on these 246 isolates. Among these isolates, 74 (30%) were susceptible to all tested drugs, 43 (17%) were resistant to one or two classes, and 129 (52%) were MDR. Based on the number of discharges from the medical and surgical services of the study hospital in the study period of 2007 to 2010 (average of 12,200/year), the incidence (per 1,000 discharges/year) was estimated to be 6.1 for E. coli BSI, 1.9 for drug-susceptible E. coli BSI, and 3.2 for MDR E. coli BSI. The susceptibility results for isolates in the study period were reported prior to the 2010 Clinical and Laboratory Standards Institute (CLSI) antimicrobial susceptibility breakpoint changes for carbapenems, cephalosporins, and aztreonam. All isolates were susceptible to ertapenem, imipenem, and meropenem, 29 (12%) were resistant to aztreonam, 108 (44%) were resistant to trimethoprim-sulfamethoxazole, 66 (27%) were resistant to fluoroquinolones, 29 (12%) were resistant to cefotaxime, ceftazidime, and ceftriaxone, and 24 (10%) were resistant to cefepime (Table 2). All 25 CTX-M gene-positive isolates were MDR, comprising 19% of the 129 MDR E. coli isolates, and were significantly more likely to be MDR than other drug-resistant strains (P < 0.0001).

Table 2.

Antimicrobial resistance and selected β-lactamase genes identified among BSI E. coli isolates

ST (total no. of isolates)	No. (%) of isolates resistant to each classa						blaCTX-M type(s) (no. of isolates)	No. of isolates with:		blaKPC type (no. of isolates)
	QUIN	SXT	CTX	FEP	ATM	CARB		blaTEM	blaOXA-1
ST131 (51)	46 (90)	36 (71)	17 (33)	15 (29)	16 (31)	0 (0)	-15 (13), -1 (3), -14 (2)	3	14
ST95 (40)	1 (3)	7 (18)	1 (3)	1 (3)	2 (5)	0 (0)	-14 (1)	0	0	-2 (1)
ST73 complex (20)	0 (0)	7 (35)	1 (5)	0 (0)	1 (5)	0 (0)
ST69 (19)	2 (11)	12 (63)	0 (0)	0 (0)	0 (0)	0 (0)
ST12 complex (13)	0 (0)	9 (69)	3 (23)	3 (23)	3 (23)	0 (0)	-15 (1b), -14 (2b)	3	0	-2 or -3 (1)
ST10 complex (9)	1 (11)	6 (67)	1 (11)	1 (11)	1 (11)	0 (0)	-15 (1)	1	0
ST1576 (2)	2 (100)	0 (0)	2 (100)	1 (50)	2 (100)	0 (0)	-15 (1), -14 (1)	0	0
ST224 (1)	1 (100)	1 (100)	1 (100)	0 (0)	1 (100)	0 (0)		0	0
ST624 (1)	1 (100)	1 (100)	1 (100)	1 (100)	1 (100)	0 (0)	-14 (1)	0	1
ST405 complex (4)	1 (25)	3 (75)	0 (0)	0 (0)	0 (0)	0 (0)
ST998 (1)	0 (0)	0 (0)	1 (100)	1 (100)	1 (100)	0 (0)		0	0
ST964 (1)	1 (100)	1 (100)	0 (0)	0 (0)	0 (0)	0 (0)		0	0
ST38 complex (4)	1 (100)	4 (100)	0 (0)	0 (0)	0 (0)	0 (0)		1	0
Other cluster ST (39)c	6 (15)	10 (26)	1 (3)	1 (3)	1 (3)	0 (0)
Unique ST (15)c	0 (0)	1 (7)	0 (0)	0 (0)	0 (0)	0 (0)
No ST (26)	3 (12)	10 (38)	0 (0)	0 (0)	0 (0)	0 (0)
All (246)	66 (27)	108 (44)	29 (12)	24 (10)	29 (12)	0 (0)	26b	8	15	2

^aQUIN, fluoroquinolones; SXT, trimethoprim-sulfamethoxazole; CTX, broad-spectrum cephalosporins (including cefotaxime, ceftazidime, and ceftriaxone); FEP, cefepime; ATM, aztreonam; CARB, carbapenem.

^bOne strain had both bla_CTX-M-14 and bla_CTX-M-15.

^cCluster ST, 2 or more isolates; unique ST, 1 isolate per group.

Identification of genotypes by MLST and fimH sequence analysis.

MLST was done to genotype 220 (89%) isolates, which fell into 41 distinct STs or ST complexes. MLST results were not able to be obtained for 26 (11%) isolates due to repeated failure of PCR amplification or interpretable sequence results of an insufficient number of the seven alleles to place them in a ST complex. Only five STs accounted for 143 (65%) of 220 typed E. coli BSI isolates (Table 3). ST131 comprised the largest ST cluster with 51 (23%) isolates, followed by ST95 identified in 40 (18%) isolates. Others included the ST73 complex with 20 (8%) isolates, ST69 with 19 (9%), and the ST12 complex with 13 (6%).

Table 3.

Distribution of E. coli genotypes by MLST

MLST	No. (%) of isolates that are:			P valuec	Total no. (%) of isolates typedd	fimH type(s)	Dominant f-type (no. in each type [% of total])
	Susceptible	Resistanta	MDRb
ST131	0 (0)	4 (8)	47 (92)	<0.0005	51 (23)	f-1, -6, -9, -11, -17	f-9 (44 [86])
ST95	21 (53)	11 (27)	8 (20)	0.001	40 (18)	f-1, -6, -9, -47	f-6 (17[43]), f-1 (16[40])
ST73 complex	6 (30)	7 (35)	7 (35)	0.806	20 (8)	f-3, -8, -9, -60	f-8 (12[52])
ST69	3 (16)	4 (21)	12 (63)	0.198	19 (9)	f-1	f-1 (19[100])
ST12 complex	1 (8)	1 (8)	11 (84)	0.116	13 (6)	f-1, -6, -51, -52	f-1 (6[46])
ST10 complex	2 (22)	1 (11)	6 (67)	1.0	9 (4)	f-1, -4, -6, -20, -28	f-6 (3[33])
ST38 complex	0 (0)	0 (0)	4 (100)	0.319	4 (2)	f-12, -57	f-12 (3[75])
ST405 complex	0 (0)	0 (0)	4 (100)	0.319	4 (2)	f-1, -4	f-1 (2[50])
ST127	2 (67)	1 (33)	0 (0)	0.216	3 (1)	f-2	f-2 (3[100])
ST1278	0 (0)	3 (50)	3 (50)	0.182	6 (2)	f-9	f-9 (6[100])
ST14 complex	0 (0)	2 (67)	1 (33)	0.556	3 (1)	f-1, -9	f-1 (2[67])
ST144	2 (67)	0 (0)	1 (33)	0.216	3 (1)	f-4	f-4 (3[100])
ST155 complex	3 (100)	0 (0)	0 (0)	0.026	3 (1)	f-7	f-7 (3[100])
ST23 complex	1 (33)	1 (33)	1 (33)	1.0	3 (1)	f-1, -6	f-6 (2[67])
ST31 complex	0 (0)	1 (25)	3 (75)	0.556	4 (2)	f-4	f-4 (4[100])
ST372	3 (100)	0 (0)	0 (0)	0.026	3 (1)	f-3	f-3 (3[100])
ST1841	0 (0)	0 (0)	2 (100)	1.0	2 (1)	f-7	f-7 (2[100])
ST1576	0 (0)	0 (0)	2 (100)	1.0	2 (1)	f-1	f-1 (2[100])
ST420	2 (100)	0 (0)	0 (0)	0.09	2 (1)	f-29	f-29 (2[100])
ST568 complex	2 (100)	0 (0)	0 (0)	0.09	2 (1)	f-2	f-2 (2[100])
ST62	0 (0)	1 (50)	1 (50)	1.0	2 (1)	f-5	f-5 (2[100])
ST59 complex	2 (67)	0 (0)	1 (33)	0.512	3 (1)	f-47, 2 NTe	f-47 (1[50])
Unique ST	12 (63)	2 (11)	5 (26)	0.001	19 (9)	f-1, -2, -3, -5, -7, -13, -27, -39, -49, -50, -59	f-7 (4[21])
Total isolates typed	62 (28)	39 (18)	119 (54)		220
Not typed	12 (46)	4 (15)	10 (39)		26
Total	74 (30)	43 (17)	129 (52)		246

^aResistant to one or two classes of drugs.

^bResistant to three or more classes of drugs.

^cP value compares the susceptible column and any resistance.

^d% of those typed by MLST (n = 220).

^eNT, not typed.

Of 246 E. coli isolates, 236 (96%) had the fimH gene, comprised of 34 different fimH SNP types (17, 21). Only three fimH types (f-1, -6, and -9) accounted for 138 (58%) of 236 typed E. coli isolates; 115 (83%) of these belonged to the five most frequently detected STs (Table 4). The three most frequent fimH types comprised 49 (96%) of the ST131 isolates, 36 (90%) of the ST95 isolates, 19 (100%) of the ST69 isolates, 4 (20%) of the ST73 complex isolates, and 7 (54%) of the ST12 complex isolates. The eight most prevalent STs contained from one to five fimH types, with one or two types found to be dominant per ST (Table 3). Strains containing the f-9 type were more likely to be MDR (P = 0.001), and 44 (86%) of them were ST131, while strains containing the f-6 type were more likely to be susceptible (P = 0.02) and 17 (43%) of them were ST95.

Table 4.

Distribution of E. coli genotypes by fimH SNP analysis

fimH typea	No. (%) of isolates that are:			P valued	No. (%) of isolates among 5 most prevalent STs	Total no. of fimH type isolates
	Susceptible	Resistantb	MDRc
f-1	10 (19)	13 (24)	31 (57)	0.024	43 (80)	54
f-2	7 (88)	1 (13)	0 (0)	0.001		8
f-3	6 (75)	0 (0)	2 (25)	0.01	3 (38)	8
f-4	3 (27)	1 (9)	7 (64)	1.0		11
f-5	0 (0)	1 (25)	3 (75)	0.319		4
f-6	14 (54)	4 (15)	8 (31)	0.005	21 (81)	26
f-7	7 (78)	0 (0)	2 (22)	0.004		9
f-8	2 (17)	6 (50)	4 (33)	0.519	12 (100)	12
f-9	4 (7)	9 (16)	45 (77)	<0.0005	51 (88)	58
f-10	1 (33)	0 (0)	2 (67)	0.512		3
f-11	1 (50)	0 (0)	1 (50)	0.512	1 (50)	2
f-12	0 (0)	0 (0)	3 (100)	0.556		3
f-13	1 (25)	1 (25)	2 (50)	0.556		4
f-27	2 (100)	0 (0)	0 (0)	0.09		2
f-29	2 (100)	0 (0)	0 (0)	0.09		2
f-47	2 (40)	2 (40)	1 (20)	0.638	4 (80)	5
f-51	2 (33)	1 (17)	3 (50)	1.0	4 (67)	6
f-52	0 (0)	0 (0)	2 (100)	1.0	2 (100)	2
f-53	2 (100)	0 (0)	0 (0)	0.09		2
Unique f-type	4 (27)	3 (20)	8 (53)	1.0	2 (13)	15
Total typed	70 (30)	42 (18)	124 (52)			236
Total untyped	4 (40)	1 (10)	5 (50)		0 (0)	10
Total	74 (30)	43 (17)	129 (52)		143 (58)	246

^afimH types (f-types) in bold are the three most prevalent fimH SNP types.

^bResistant to one or two classes of drugs.

^cResistant to three or more classes of drugs.

^dP value compares the susceptible column and any resistance.

fimH SNP types were used to further differentiate ST95 strains. ST95 strains were comprised of 4 fimH SNP types (f-1, -6, -9, and -47). Fourteen (67%) of the 21 drug-susceptible ST95 strains belonged to one fimH SNP type (f-6), whereas 16 (84%) of 19 resistant ST95 strains belonged to three other fimH types (P = 0.003) (Fig. 1). The f-6 type was the most prevalent SNP type, accounting for 17 (43%) of the ST95 strains. None of these were MDR (P < 0.05). The f-1 type comprised 16 (40%) of the ST95 strains, and 12 (75%) of these were resistant to at least one class of antimicrobial agents while 6 (38%) were MDR (P > 0.05). The f-9 and f-47 types comprised 3 and 4 ST95 strains, respectively, with one MDR strain in each.

Fig 1.

Distribution of fimH SNP types among ST95 strains by drug susceptibility.

Diversity of genotypes.

Among 220 E. coli isolates typed by MLST, 41 STs or ST complexes were identified. Of these, 201 (91%) belonged to one of 22 ST groups (Table 5). Nineteen isolates belonged to unique STs. Simpson's diversity index for all STs was 89% (95% confidence interval [CI], 86% to 91%). The index for drug-susceptible ST genotypes was 87% (95% CI, 80% to 94%), while the index for MDR ST genotypes was 81% (95% CI, 74% to 87%). The index for any drug-resistant (MDR and resistance to one or two classes of drug) ST genotypes was 85% (95% CI, 81% to 89%). A higher index represents greater diversity.

Table 5.

Simpson's diversity index by MLST and by antimicrobial resistance status

Resistance level (no. of isolates)	No. (%) of isolates MLST typed	No. of MLST cluster types	No. (%) of isolates in MLST cluster typesa	No. of isolates in MLST unique typesa	Simpson's diversity index (%) (95% CI)
Susceptible (74)	62 (84)	13	50 (81)	12	87 (80–94)
Resistant (43)b	39 (91)	12	37 (95)	2	86 (80–93)
MDR (129)c	119 (92)	17	114 (96)	5	81 (74–87)
Any resistance (172)d	154 (90)	18	147 (95)	7	85 (81–89)
All E. coli isolates (246)	220 (89)	22	201 (91)	19	89 (86–91)

^aCluster, 2 or more isolates; unique, 1 isolate per group.

^bResistant to 1 or 2 antimicrobial agents.

^cMultidrug resistant (≥3 classes).

^dResistance to ≥1 class.

Three of the five most prevalent E. coli clonal groups, ST131, ST69, and ST12 complex, were significantly more likely to be MDR than were others (P = 0.047). These three clonal groups comprised 70 (54%) of the MDR strains (Table 3). The other MDR strains were comprised of 19 distinct STs (10 strains were not typed). On the other hand, ST95 strains were more likely to be susceptible (P = 0.001). In fact, drug-susceptible strains belonging to this clonal group accounted for 28% of all 74 drug-susceptible isolates in the study. Of 119 typed MDR E. coli strains, 114 (96%) belonged to one of 17 STs represented by 2 or more members per ST. Of 62 typed susceptible strains, 50 (81%) belonged to one of 13 STs represented by 2 or more members per ST. A large proportion (63%) of the 19 unique STs were fully susceptible.

Distribution of β-lactamase genes.

The bla_CTX-M gene was detected among 25 (10%) E. coli strains in five ST groups and only one nonclonal ST (Table 2). Fifteen had bla_CTX-M-15 (group 1), 6 had bla_CTX-M-14 (group 9), and one (an ST12 complex isolate) had both. For three strains with the bla_CTX-M-1 group gene, CTX-M variants were not able to be further differentiated. ST131 accounted for 18 (72%) of the 25 isolates carrying bla_CTX-M. Two carried bla_CTX-M-14, and16 had bla_CTX-M-1 group genes, of which 13 were bla_CTX-M-15. Fifteen CTX-M gene-positive isolates also had bla_OXA-1. PCR analysis of bla_TEM and bla_SHV was performed on 36 E. coli isolates. Three ST131 isolates were bla_TEM gene positive by PCR, and two of these coharbored bla_CTX-M. Isolates carrying bla_TEM were also found among three ST12 complex isolates and one each of ST10 complex and ST38 complex isolates. Only one of 40 ST95 isolates was extended-spectrum β-lactamase (ESBL) gene positive. This strain was resistant to cefotaxime and carried bla_CTX-M-14 and bla_KPC. One ST12 complex isolate was also found to carry both bla_CTX-M-14 and bla_KPC. Both KPC gene-positive isolates were sensitive to all carbapenems tested. Susceptibility to imipenem and ertapenem was confirmed by Etest, with MICs for both strains of 0.25 and 0.023 for imipenem and ertapenem, respectively. No ESBL genes were found among the ST69 or ST73 complex strains. The bla_SHV gene was not detected in any of the tested strains.

Strain distribution by time of first positive blood culture results.

Of 244 (99%) E. coli isolates with known time of first positive blood culture results, 198 (81%) were cultured from blood <48 h after patient admission (Table 6). Most of the cultures were done at the emergency department (ED) before admission. Among 73 fully susceptible and 128 MDR isolates with culture result data, 62 (85%) and 101 (79%), respectively, were isolated from blood <48 h after patient admission (P > 0.05). Large proportions (>70%) of each of the 5 most common ST strains were isolated from blood <48 h after patient admission. Based on the number of ED visits during the study period, the incidences of susceptible E. coli BSI and MDR E. coli BSI were 0.12 and 0.30 per 1,000 ED visits, respectively (P > 0.05).

Table 6.

Distribution of E. coli MLSTs by timing of the first collection of blood culture that grew E. coli

ST or ST complex	No. (%) of isolates for each collection time
	>48 h	<48 h
ST131	14 (27)	37 (73)
ST95a	4 (10)	36 (90)
ST73 complex	4 (20)	16 (80)
ST69a	3 (17)	15 (83)
ST12 complexa	1 (8)	11 (92)
Other STsa	13 (18)	61 (82)
Untyped	6 (23)	20 (77)
Totala	46 (19)	198 (81)

^aData on the timing of the first positive isolate collection are available for 244 of 249 E. coli isolates.

DISCUSSION

Of 246 E. coli isolates consecutively collected from patients with BSI in one San Francisco hospital over a 3-year period, we found that only five genotypes by MLST accounted for 146 (66%) of the typed E. coli isolates; these were ST131, ST95, ST73 complex, ST69, and ST12 complex. Interestingly, MDR strains belonging to ST131 and ST69 accounted for nearly half (46%) of all MDR isolates while drug-susceptible strains from ST95 and ST73 complex accounted for more than one-third (36%) of all drug-susceptible strains in this study. Thus, both MDR and drug-susceptible E. coli strains have clonal distribution and accounted for a large proportion of bloodstream isolates from patients admitted to a public hospital in San Francisco.

All of these STs, with the exception of ST69, which belongs to phylogenetic group D, are reported to belong to phylogenetic group B2. Both phylogenetic groups are typically associated with human ExPEC infections (13, 24, 25). Strains belonging to these clonal groups are also frequently isolated from food-producing and companion animals and from retail meats (10, 11, 26–35) and cause serious infections in otherwise healthy individuals (6, 13). Isolates of the enterobacterial repetitive intergenic consensus 2 (ERIC-2) PCR pattern CgA, later characterized as ST69 by MLST, were the most common uropathogenic E. coli genotype isolated from college students with UTI (36, 37). ST95 contains members of avian pathogenic E. coli (APEC) as well as human uropathogenic E. coli (UPEC) and neonatal meningitis E. coli (NMEC) (25). ST95 subgroups share common serotypes and virulence factors, including carriage of pap genes encoding P-type fimbriae which enhance colonization in avian and human epithelial cells. These characteristics may promote an opportunity for zoonotic transmission (38–41).

ST131, the most prevalent MLST genotype reported globally, was also the most common ST found in this study, accounting for 51 (23%) of typed E. coli isolates, 47 (36%) of all MDR isolates, and 18 (72%) of CTX-M gene-positive isolates. This clonal group was comprised of five fimH SNP types, for which 39 (76%) strains were f-9, and all but four of these were MDR. A large proportion of these ST131 infections appear to be community onset, since the blood cultures of 37 (73%) isolates grew E. coli within 48 h of hospital admission. This clonal group has worldwide distribution as a common cause of UTI and BSI (3–6, 42). ST131 is remarkable for its carriage of a wide array of virulence factors (8, 10, 43, 44) in addition to being a major producer of CTX-M-type ESBLs (1, 2, 7, 9, 10) and, more recently, for its acquisition of the carbapenemase-encoding genes bla_KPC, bla_VIM, and bla_NDM-1 (45–48). These drug resistance determinants are often located on plasmids and associated with integrons and IS elements (3, 10, 49–51). A recent report found that by pulsed-field gel electrophoresis (PFGE), ST131 strains isolated from human clinical sources belonged to pulsotypes distinct from those obtained from foods or animal food sources (52).

ST95 was the second largest clonal group identified in this study. Unlike ST131, the majority (53%) were susceptible to all tested antimicrobial agents. Like ST131, a large proportion (90%) of ST95 isolates were classified as representing community-onset infection. ST95 was found in our study population with prevalence nearly equal to that of ST131 in spite of having significantly fewer drug-resistant strains (P = 0.001). In fact, among four fimH SNP types comprising ST95, one (f-6) was responsible for most (67%) of the drug-susceptible strains. Such an observation suggests that drug resistance is not a prerequisite for clonal dissemination.

The observation that drug-susceptible ST95 disseminates clonally is not new. Manges et al. found several ST95 isolates in their report of endemic and epidemic lineages of E. coli causing UTI in Montreal and California (13). All except one of the strains were susceptible to all drugs tested. Vincent et al. detected ST95 in human clinical isolates and in one case of ready-to-eat food product (11). All but two of these isolates were susceptible. ST95 was found to be the most prevalent cause of bacterial peritonitis and BSI in France (24). All ST95 strains isolated from UTI in a study in the northwest of England were susceptible to all tested antimicrobial agents (12). Our study, based on consecutively collected, population-based samples from one large hospital, confirms that drug-susceptible E. coli strains can disseminate clonally in community settings.

Near the end of the study period (April 2010), one ST95 isolate collected from a patient within 48 h of admission carried both bla_CTX-M and bla_KPC, which, to our knowledge, is the first example of acquisition of these genes by ST95. This ST95 isolate, an f-1 fimH SNP type, was distinct from the other ST95 isolates in that it was resistant to 12 drugs within nine classes of antimicrobial agents and harbored the dfrA17 and bla_CTX-M-14 genes. No ESBL genes were found among any of 19 ST69 strains, while more than two-thirds of 25 CTX-M gene-positive strains belonged to ST131. These observations suggest intrinsic differences in the frequency of distinct E. coli lineages acquiring different types of mobile drug resistance genes or that these genes remain within the lineages after their “founder” strains acquire them.

Because we consecutively collected and analyzed all BSI isolates from one hospital in San Francisco over a period of 3 years, we were able to determine the relative proportion of infections caused by major clonal groups of both drug-resistant and -susceptible strains. Strain diversity was higher in drug-susceptible strains than in MDR strains, but some of the drug-susceptible strains were indeed clonal by MLST, by fimH SNP type, and by both combined. Furthermore, more than two-thirds of MDR as well as drug-susceptible E. coli infections were found to have community onset. We do not have additional information about prior health care exposures of these patients, so it is possible that some of these community-onset BSI cases had contact with health care settings prior to their infection. Our observation, of course, is limited to one geographic setting, and thus, a similar study should be performed at other sites to determine the generalizability of our findings. Nevertheless, our data suggest that the overrepresentation of both MDR and susceptible E. coli strains causing community-onset BSI may be explained by their enhanced capacity to cause extraintestinal infection rather than selection by antimicrobial agents used in hospitals.