Extended-Spectrum-β-Lactamase- and Plasmid AmpC-Producing Escherichia coli Causing Community-Onset Bloodstream Infection: Association of Bacterial Clones and Virulence Genes with Septic Shock, Source of Infection, and Recurrence

Invasive infections due to extended-spectrum-β-lactamase- and pAmpC-producing Escherichia coli (ESBL/pAmpC-EC) are an important cause of morbidity, often caused by the high-risk clone sequence type (ST131) and isolates classified as extraintestinal pathogenic E. coli (ExPEC). The relative influence of host immunocompetence versus microbiological virulence factors in the acquisition and outcome of bloodstream infections (BSI) is poorly understood. Herein, we used whole-genome sequencing on 278 blood culture isolates of ESBL/pAmpC-EC from 260 patients with community-onset BSI collected from 2012 to 2015 in Stockholm to study the association of virulence genes, sequence types, and antimicrobial resistance with severity of disease, infection source, ESBL/pAmpC-EC BSI low-risk patients, and patients with repeated episodes.

ABSTRACT

Invasive infections due to extended-spectrum-β-lactamase- and pAmpC-producing Escherichia coli (ESBL/pAmpC-EC) are an important cause of morbidity, often caused by the high-risk clone sequence type (ST131) and isolates classified as extraintestinal pathogenic E. coli (ExPEC). The relative influence of host immunocompetence versus microbiological virulence factors in the acquisition and outcome of bloodstream infections (BSI) is poorly understood. Herein, we used whole-genome sequencing on 278 blood culture isolates of ESBL/pAmpC-EC from 260 patients with community-onset BSI collected from 2012 to 2015 in Stockholm to study the association of virulence genes, sequence types, and antimicrobial resistance with severity of disease, infection source, ESBL/pAmpC-EC BSI low-risk patients, and patients with repeated episodes. ST131 subclade C2 comprised 29% of all patients. Factors associated with septic shock in multivariable analysis were patient host factors (hematologic cancer or transplantation and reduced daily living activity), presence of the E. coli virulence factor iss (increased serum survival), absence of phenotypic multidrug resistance, and absence of the genes pap and hsp. Adhesins, particularly pap, were associated with urinary tract infection (UTI) source, while isolates from post-prostate biopsy sepsis had a low overall number of virulence operons, including adhesins, and commonly belonged to ST131 clades A, B, and subclade C1, ST1193, and ST648. ST131 was associated with recurrent episodes. In conclusion, the most interesting finding is the association of iss with septic shock. Adhesins are important for UTI pathogenesis, while otherwise low-pathogenic isolates from the microbiota can cause post-prostate biopsy sepsis.

INTRODUCTION

Bloodstream infection (BSI) caused by extended-spectrum-β-lactamase- and pAmpC-producing Escherichia coli (ESBL/pAmpC-EC) is a major clinical challenge. It is well established that host factors such as neutropenia and comorbidities and time to appropriate therapy are associated with severe outcomes and mortality, but the contribution of virulence factors (VFs) to the severity of disease in ESBL/pAmpC-EC BSI remains unclear. The primary aim of this study was to determine the association of patient characteristics and microbial virulence with severity of disease. In addition, we aimed to study the association of host and microbial determinants with other characteristics, such as various types of infection, i.e., post-prostate biopsy BSI (PPBS) and urinary tract infection (UTI), patients with a low risk of ESBL/pAmpC-EC BSI, and patients with recurrent BSI.

Many previous observations indicate that microbial properties influence the development and severity of BSI. The successful high-risk clone sequence type 131 (ST131) is responsible for a substantial share of clinical infections with ESBL/pAmpC-EC worldwide (1–3). A recent Swedish study reported that phylogroup B2 and ST131 were found significantly more frequently among isolates from bloodstream infections than among isolates from asymptomatic population carriers (4).

ST131 and especially clade C have also been associated with various virulence factors (VFs) (5). Isolates of E. coli with enhanced virulence potential for causing extraintestinal infection, such as BSI and urinary tract infection (UTI), are often referred to as extraintestinal pathogenic E. coli (ExPEC) as opposed to diarrheagenic E. coli (DEC) and normally low-pathogenic commensal gut E. coli. A molecular definition of ExPEC isolates associated with extraintestinal infection suggested by James Johnson (JJ) et al. is widely used (6). Isolates are classified as ExPEC if they are positive for ≥2 of the following VFs: any pap, sfa and/or foc, afa and/or dra, any kpsMT, and iutA (6, 7). Hereafter, the molecular definition is referred to as ExPEC_JJ to distinguish this definition from ExPEC in a wider sense. However, according to the same author, the distinguishing characteristics of ExPEC are still incompletely defined (7).

Some evidence exists that ExPEC_JJ strains and strains carrying other ExPEC VFs are indeed more virulent, not only by the association with infection but also by association with the severity of disease. In a murine sepsis model, Johnson et al. assessed isolates from both clinical and surveillance fecal samples (7). ExPEC_JJ isolates were associated with severity of disease, and the strongest individual predictors for severity of disease were fyuA and kpsM and K1 capsule gene. Conversely, it has been difficult to show a significant association of virulence factors with mortality in several clinical studies (5, 8, 9).

A recent prostate biopsy was one of the strongest predictors for BSI caused by ESBL-producing Enterobacterales (EPE) among patients with suspected community-onset infection in a previous study (10). In the present study, we aimed to analyze the clonal relation and virulence properties of isolates from PPBS. Several of the ExPEC-defining VFs are adhesins and have been associated with UTI in previous studies. Our hypothesis was that these virulence factors would be more common in UTI than in other type of infections.

In a previous study, as many as 50% of cases with EPE BSI lacked any of the three major risk factors identified in the study (prior EPE-positive culture, prior health care abroad within 6 months, prior prostate biopsy within 1 month), and 20% neither had any of these risk factors nor had been hospitalized in the previous year (10). We hypothesized that these patients were healthier and that these infections were caused by more virulent strains.

Recurring episodes of EPE BSI are common in some patients. For these patients, an association with microbial determinants was sought.

Herein, we describe the molecular epidemiology of community-onset bloodstream infection caused by ESBL/pAmpC-EC in Stockholm. Here, all strains were invasive as well as ESBL/pAmpC producing, and a high prevalence of virulence genes was expected. Association of the suggested virulence genes with the clinical severity of disease within this selected group would strengthen their virulence significance.

RESULTS

Overview of epidemiology and resistance.

A total of 260 isolates from unique patients with community-onset BSI due to ESBL/pAmpC-EC were available for whole-genome sequencing (WGS). ST131 was the most prevalent sequence type (ST), accounting for 47%. The ST131 subclades C2 (29% out of all isolates) and C1 (11%) clearly dominated (see Table 2; see also Table S1 in the supplemental material). Other common sequence types were ST38 (7%), ST648 (6%), ST10 (4%), and ST405 (4%). The remaining 86 isolates (33%) belonged to 43 different sequence types. The phylogenetic trees analyzed by single nucleotide polymorphism (SNP) and associated clinical data and resistance and virulence genes are shown in Fig. 1 and 2. ESBL production was genetically confirmed by detection of ESBL or pAmpC genes in all but four isolates. The most common CTX-M β-lactamase gene was bla_CTX-M-15 (present in 60% of isolates) followed by bla_CTX-M-27 (13%) and bla_CTX-M-14 (10%).

TABLE 2.

Univariate association between microbiological determinants and clinical outcome^a

Microbiological determinant	All patients (no. [%]) (n = 260)	No septic shock or death in <3 days (no. [%]) (n = 233)	Septic shock or death in <3 days (no. [%]) (n = 27)	Univariate OR (95% CI)	P valueb	Alive at 30 days (no. [%]) (n = 243)	30-day mortality (no. [%]) (n = 17)	Univariate OR (95% CI)	P valueb
ST131	121 (47)	112 (48)	9 (33)	0.54 (0.23–1.25)	0.151	114 (47)	(47) (7)	0.79 (0.29–2.15)	0.647
Subclade C1	29 (11)	28 (12)	1 (4)	0.28 (0.04–2.16)	0.222	27 (11)	2 (12)	1.07 (0.23–4.92)	0.934
Subclade C2	76 (29)	69 (30)	7 (26)	0.83 (0.34–2.06)	0.687	71 (29)	5 (29)	1.01 (0.34–2.97)	0.987
ST10	10 (4)	8 (3)	2 (7)	2.25 (0.45–11.18)	0.322	8 (3)	2 (12)	3.92 (0.76–20.09)	0.102
ST38	17 (7)	14 (6)	3 (11)	1.96 (0.52–7.29)	0.318	17 (7)	0 (0)	N.A.	0.612
ST405	10 (4)	10 (4)	0 (0)	N.A.	0.606	10 (4)	0 (0)	N.A.	1.000
ST648	16 (6)	15 (6)	1 (4)	0.56 (0.07–4.41)	0.581	16 (7)	0 (0)	N.A.	0.609
ST1193	6 (2)	6 (3)	0 (0)	N.A.	1.000	5 (2)	1 (6)	2.98 (0.33–27.01)	0.333
Combined phenotypic resistance ≥3	160 (62)	150 (64)	10 (37)	0.33 (0.14–0.74)	0.008	153 (63)	7 (41)	0.41 (0.15–1.12)	0.082
Combined genetic resistance ≥3	183 (70)	167 (72)	16 (59)	0.57 (0.25–1.30)	0.185	171 (70)	12 (71)	1.01 (0.34–2.97)	0.985
AMG phenotypic resistance	97 (37)	91 (39)	6 (22)	0.45 (0.17–1.15)	0.094	94 (39)	3 (18)	0.34 (0.10–1.21)	0.097
AMG genetic resistance	203 (78)	185 (79)	18 (67)	0.52 (0.22–1.23)	0.135	191 (79)	12 (71)	0.65 (0.22–1.94)	0.443
ExPECJJ	202 (78)	186 (80)	16 (59)	0.37 (0.16–0.84)	0.018	192 (79)	10 (59)	0.38 (0.14–1.05)	0.061
UPECHM	172 (66)	159 (68)	13 (48)	0.43 (0.19–0.97)	0.041	163 (67)	9 (53)	0.55 (0.21–1.48)	0.239
Virulence operons ≥18	55 (21)	50 (21)	5 (19)	0.83 (0.30–2.31)	0.724	53 (22)	2 (12)	0.48 (0.11–2.16)	0.337
Mean no. of virulence operons (95% CI)	14.7 (14.2–15.1)	14.8 (14.3–15.3)	13.3 (11.7–14.9)		0.055	14.8 (14.3–15.2)	13.5 (11.4–15.5)		0.181
Adhesin(s)
afa and/or draBC	47 (18)	42 (18)	5 (19)	1.03 (0.37–2.89)	0.950	45 (19)	2 (12)	0.59 (0.13–2.66)	0.489
sfa and/or foc	17 (7)	14 (6)	3 (11)	1.96 (0.52–7.29)	0.318	16 (7)	1 (6)	0.89 (0.11–7.12)	0.910
papACGH	105 (40)	101 (43)	4 (15)	0.23 (0.08–0.68)	0.008	102 (42)	3 (18)	0.30 (0.08–1.06)	0.061
fim	229 (88)	204 (88)	25 (93)	1.78 (0.40–7.90)	0.450	214 (88)	15 (88)	1.02 (0.22–4.67)	0.983
Immune evasion
kpsM	198 (76)	180 (77)	18 (67)	0.59 (0.25–1.39)	0.226	188 (77)	10 (59)	0.42 (0.15–1.15)	0.091
kfiC	49 (19)	43 (18)	6 (22)	1.26 (0.48–3.32)	0.636	45 (19)	4 (24)	1.35 (0.42–4.35)	0.611
iss	191 (73)	167 (72)	24 (89)	3.16 (0.92–10.86)	0.067	177 (73)	14 (82)	1.74 (0.48–6.25)	0.396
tcpC	11 (4)	10 (4)	1 (4)	0.86 (0.11–6.97)	0.886	11 (5)	0 (0)	N.A.	1.000
Invasion
aslA	240 (92)	217 (93)	23 (85)	0.42 (0.13–1.38)	0.153	223 (92)	17 (100)	N.A.	0.377
fdeC	254 (98)	227 (97)	27 (100)	N.A.	1.000	237 (98)	17 (100)	N.A.	1.000
ibeA	8 (3)	7 (3)	1 (4)	1.24 (0.15–10.49)	0.842	7 (3)	1 (6)	2.11 (0.24–18.19)	0.498
ompA	260 (100)	233 (100)	27 (100)	N.A.	1.000	243 (100)	17 (100)	N.A.	1.000
ompT	186 (72)	169 (73)	17 (63)	0.64 (0.28–1.48)	0.300	173 (71)	13 (76)	1.32 (0.41–4.17)	0.642
Siderophore, iron uptake
iuc	202 (78)	187 (80)	15 (56)	0.31 (0.13–0.70)	0.005	191 (79)	11 (65)	0.50 (0.18–1.41)	0.191
iutA	190 (73)	175 (75)	15 (56)	0.41 (0.18–0.94)	0.034	179 (74)	11 (65)	0.66 (0.23–1.85)	0.424
chuA	217 (83)	197 (85)	20 (74)	0.52 (0.21–1.32)	0.171	204 (84)	13 (76)	0.62 (0.19–2.01)	0.426
ent	260 (100)	233 (100)	27 (100)	N.A.	1.000	243 (100)	17 (100)	N.A.	1.000
iroN	33 (13)	28 (12)	5 (19)	1.66 (0.58–4.75)	0.341	32 (13)	1 (6)	0.41 (0.05–3.21)	0.398
Protease
pic	18 (7)	16 (7)	2 (7)	1.09 (0.24–5.00)	0.917	17 (7)	1 (6)	0.83 (0.10–6.65)	0.861
sat	156 (60)	146 (63)	10 (37)	0.35 (0.15–0.80)	0.013	147 (60)	9 (53)	0.73 (0.27–1.97)	0.540
vat	30 (12)	27 (12)	3 (11)	0.95 (0.27–3.38)	0.941	28 (12)	2 (12)	1.02 (0.22–4.71)	0.976
Toxin
hly	62 (24)	58 (25)	4 (15)	0.52 (0.17–1.58)	0.252	61 (25)	1 (6)	0.19 (0.02–1.44)	0.107
cnf1	48 (18)	45 (19)	3 (11)	0.52 (0.15–1.81)	0.306	47 (19)	1 (6)	0.26 (0.03–2.01)	0.198
Miscellaneous
yfcV	210 (81)	189 (81)	21 (78)	0.81 (0.31–2.14)	0.677	197 (81)	13 (76)	0.76 (0.24–2.43)	0.643
fyuA	233 (90)	212 (91)	21 (78)	0.35 (0.13–0.95)	0.040	220 (91)	13 (76)	0.34 (0.10–1.13)	0.078
hsp	153 (59)	141 (61)	12 (44)	0.52 (0.23–1.17)	0.113	144 (59)	9 (53)	0.77 (0.29–2.07)	0.610
malX	258 (99)	231 (99)	27 (100)	N.A.	1.000	241 (99)	17 (100)	N.A.	1.000

^aFirst episodes (n = 260).

^bLogistic regression used for calculations, except for variables with 0 or 100% in the outcome group, for which Fisher's exact test was used to assess strength of association. N.A., not available.

FIG 1.

Maximum likelihood SNP phylogram with data on outcome, patient history, virulence operons, phenotypic AST, and resistance genes for 121 ST131 isolates and reference isolates with assemblies from EnteroBase. Reference genomes (from Petty et al. [40]) are those that lack metadata. Only selected reference genomes are shown in the figure. See Fig. S2 in the supplemental material for the full phylogenetic tree with all reference genomes used and Data Set S1 in the supplemental material for accession numbers. EC958 was the reference for SNP mapping. The tree scale is the number of substitutions per site.

FIG 2.

Maximum likelihood SNP phylogram with data on patient history, outcome, virulence operons, phenotypic AST, and resistance genes for 139 isolates with STs other than ST131 plus additional reference isolates from EnteroBase. Reference genomes are those that lack metadata. See Data Set S1 in the supplemental material for a list of reference genomes. EC958 was the reference for SNP mapping. The tree scale is the number of substitutions per site. Cx, clonal complex.

Isolates belonging to the ST131 subclade C2 were among the most resistant and virulent isolates. All ST131 subclade C2 isolates were classified as uropathogenic Escherichia coli according to the criteria of H. Mobley (UPEC_HM), 97% as ExPEC_JJ, and 41% had ≥18 virulence operons (see Table S1). All C2 isolates carried bla_CTX-M-15, as expected, and 82% were multidrug resistant (MDR). Isolates belonging to ST73 carried a high number of virulence operons (7/8 isolates had ≥18 virulence operons), and all were classified both as UPEC_HM and ExPEC_JJ. Similarly, the four ST12 isolates were all classified as UPEC_HM, ExPEC_JJ, and had ≥18 virulence operons.

Phenotypic fluoroquinolone resistance was common (72%) and strongly clonally distributed. Resistance to trimethoprim-sulfamethoxazole (68%) and aminoglycosides (37%) was more variably distributed between STs. Two isolates carried colistin resistance gene mcr-1; one was ST88 and one was ST95.

Association of host factors and microbiological determinants with severity of disease.

Septic shock or death within 3 days was used as the primary outcome as a measure of severe infection. This outcome, hereafter referred to as septic shock, affected 27 patients (10%). The frequency of the secondary outcome, all-cause 30-day mortality, was 6.5%. In univariate analysis, patient factors associated with septic shock (Table 1) were a history of hematologic cancer or transplantation, Charlson index of ≥4, and daily living activity of ≥2. Presence of the adhesin genes pap, aerobactin siderophore system genes iuc, iutA, the protease gene sat, the yersiniabactin receptor gene fuyA, and ExPEC_JJ and UPEC_HM status had a protective association in univariate analysis (Table 2). In the final model of the multivariable analysis, the highest odds ratios (ORs) were obtained for hematologic cancer or transplantation (OR, 16.34 [95% confidence interval {CI}, 4.73 to 56.49]; P < 0.001) and presence of the E. coli immune evasion gene iss (increased serum survival) (OR, 7.71 [1.8 to 33.07]; P = 0.006). iss was detected in 73% of the present isolates and clonally distributed. Isolates belonging to sequence types ST405, ST648, and ST1193 generally lacked iss (Fig. 1 and 2). Additional statistically significant factors in the multivariable analysis were reduced daily living activity, absence of phenotypic MDR, and absence of the genes pap and hsp (Table 3).

TABLE 1.

Univariate association between patient characteristics and clinical outcome (n = 260)

Characteristic	No septic shock or death in <3 days (no. [%]) (n = 233)	Septic shock or death in <3 days (no. [%]) (n = 27)	Univariate OR (95% CI)	P valuea	Alive at 30 days (no. [%]) (n = 243)	30-day mortality (no. [%]) (n = 17)	Univariate OR (95% CI)	P valuea
Hospital code
1	124 (53)	18 (67)	1.0 (Reference)		130 (53)	12 (71)	1.0 (Reference)
2	63 (27)	6 (22)	0.66 (0.25–1.73)	0.396	65 (27)	4 (24)	0.67 (0.21–2.15)	0.497
3	46 (20)	3 (11)	0.45 (0.13–1.60)	0.216	48 (20)	1 (6)	0.23 (0.03–1.78)	0.158
Index yr
2012	40 (17)	7 (26)	1.0 (Reference)		44 (18)	3 (18)	1.0 (Reference)
2013	63 (27)	3 (11)	0.27 (0.07–1.11)	0.070	63 (26)	3 (18)	0.7 (0.13–3.62)	0.669
2014	64 (27)	9 (33)	0.8 (0.28–2.33)	0.687	66 (27)	7 (41)	1.56 (0.38–6.34)	0.538
2015	66 (28)	8 (30)	0.69 (0.23–2.06)	0.508	70 (29)	4 (24)	0.84 (0.18–3.92)	0.823
Age category
0–49	24 (10)	4 (15)	1.0 (Reference)		27 (11)	1 (6)	1.0 (Reference)
50–59	38 (16)	4 (15)	0.63 (0.14–2.77)	0.542	40 (16)	2 (12)	1.35 (0.12–15.64)	0.810
60–69	80 (34)	7 (26)	0.52 (0.14–1.95)	0.335	84 (35)	3 (18)	0.96 (0.1–9.66)	0.975
70–79	45 (19)	5 (19)	0.67 (0.16–2.72)	0.572	48 (20)	2 (12)	1.13 (0.1–12.99)	0.925
≥80	46 (20)	7 (26)	0.91 (0.24–3.43)	0.893	44 (18)	9 (53)	5.52 (0.66–46.05)	0.114
Female	81 (35)	11 (41)	1.29 (0.57–2.91)	0.539	85 (35)	7 (41)	1.30 (0.48–3.54)	0.606
Charlson category
0–1	115 (49)	6 (22)	1.0 (Reference)		120 (49)	1 (6)	1.0 (Reference)
2–3	76 (33)	9 (33)	2.27 (0.78–6.64)	0.134	79 (33)	6 (35)	9.11 (1.08–77.15)	0.043
4–5	19 (8)	5 (19)	5.04 (1.4–18.18)	0.013	22 (9)	2 (12)	10.91 (0.95–125.55)	0.055
≥6	23 (10)	7 (26)	5.83 (1.79–18.96)	0.003	22 (9)	8 (47)	43.64 (5.2–366.45)	0.001
Daily living activity
0	119 (51)	7 (26)	1.0 (Reference)		124 (51)	2 (12)	1.0 (Reference)
1	50 (21)	6 (22)	2.04 (0.65–6.37)	0.220	52 (21)	4 (24)	4.77 (0.85–26.85)	0.076
2	52 (22)	11 (41)	3.6 (1.32–9.80)	0.012	53 (22)	10 (59)	11.7 (2.48–55.22)	0.002
3	12 (5)	3 (11)	4.25 (0.97–18.61)	0.055	14 (6)	1 (6)	4.43 (0.38–52)	0.236
Charlson heart disease	28 (12)	7 (26)	2.56 (0.99–6.61)	0.051	30 (12)	5 (29)	2.96 (0.97–8.99)	0.056
Charlson solid metastatic	15 (6)	5 (19)	3.30 (1.10–9.95)	0.034	13 (5)	7 (41)	12.38 (4.06–37.80)	<0.001
Chronic wound	17 (7)	4 (15)	2.21 (0.69–7.13)	0.184	18 (7)	3 (18)	2.68 (0.70–10.19)	0.148
Hematologic cancer	13 (6)	6 (22)	4.84 (1.67–14.04)	0.004	16 (7)	3 (18)	3.04 (0.79–11.68)	0.105
Other cancer	24 (10)	6 (22)	2.49 (0.91–6.77)	0.074	23 (9)	7 (41)	6.70 (2.33–19.27)	<0.001
Solid organ transplantation	3 (1)	3 (11)	9.58 (1.83–50.13)	0.007	5 (2)	1 (6)	2.98 (0.33–27.01)	0.333
Bone marrow transplantation	1 (0)	1 (4)	8.92 (0.54–146.93)	0.126	2 (1)	0 (0)	N.A.	1.000
Chronic kidney dysfunction	23 (10)	6 (22)	2.61 (0.96–7.12)	0.061	25 (10)	4 (24)	2.68 (0.81–8.86)	0.105
Obstructive urinary tract disease	105 (45)	7 (26)	0.43 (0.17–1.05)	0.063	105 (43)	7 (41)	0.92 (0.34–2.50)	0.870
Arrival from long-term care facility	39 (17)	7 (26)	1.74 (0.69–4.40)	0.241	41 (17)	5 (29)	2.05 (0.69–6.14)	0.198
Health care-associated infection	159 (68)	20 (74)	1.33 (0.54–3.28)	0.537	164 (67)	15 (88)	3.61 (0.81–16.19)	0.093

^aLogistic regression used for calculations, except for variables with 0 or 100% in the outcome group, for which Fisher's exact test was used to assess strength of association. N.A., not available.

TABLE 3.

Multivariable analysis septic shock or death in ≤3 days (n = 260)

Characteristic	No septic shock or death ≤3 days (no. [%]) (n = 233)	Septic shock or death ≤3 days (no. [%]) (n = 27)	Univariate OR (95% CI)	P value	Multivariate OR (95% CI)	P value
Hematologic cancer or transplantation	17 (7)	9 (33)	6.35 (2.48–16.27)	<0.001	16.34 (4.73–56.49)	<0.001
Daily living activity ≥2	64 (27)	14 (52)	2.84 (1.27–6.38)	0.011	3.85 (1.47–10.11)	0.006
iss	167 (72)	24 (89)	3.16 (0.92–10.86)	0.067	7.71 (1.8–33.07)	0.006
papACGH	101 (43)	4 (15)	0.23 (0.08–0.68)	0.008	0.17 (0.05–0.61)	0.006
hsp	141 (61)	12 (44)	0.52 (0.23–1.17)	0.113	0.36 (0.14–0.96)	0.040
Combined phenotypic resistance ≥3	150 (64)	10 (37)	0.33 (0.14–0.74)	0.008	0.31 (0.12–0.82)	0.018

For all-cause 30-day mortality, only a Charlson index of ≥4 (mainly due to solid metastatic cancer) and daily living activity of ≥2 had a statistically significant association. Sequence types were not associated with septic shock or 30-day mortality.

Association of microbiological determinants with PPBS and UTI as sources of infection.

PPBS is principally different from other types of BSI, as enteric bacteria are iatrogenically transferred via the prostate into the bloodstream through the procedure. Ciprofloxacin prophylaxis is routinely given to minimize the risk of BSI. Prostate biopsy was the source of infection in 40 patients (24% of male patients), none of whom had septic shock nor died within 30 days.

Hence, the association of microbial determinants with PPBS was analyzed in male patients and showed that ST131 clades A and B, ST131 subclade C1, ST648, and ST1193 together accounted for 63% (25 cases) of PPBS compared to 22% (28 cases) of other causes (Table 4). These groups were combined into a variable called “high-risk STs” that had an OR of 5.71 (95% CI, 2.42 to 13.47; P < 0.001) for PPBS.

TABLE 4.

Association of microbiological determinants and prostate biopsy in men (n = 168)

Characteristic	Other origin of infection (no. [%]) (n = 128)	Prostate biopsy <1 mo (no. [%]) (n = 40)	Univariate OR (95% CI)a	P value	OR adjusted for ST (95% CI)a	P value
Sequence type
Other STs	61 (48)	9 (23)	1.0 (Reference)
ST131 subclade C1	14 (11)	12 (30)	5.81 (2.05–16.46)	0.001
ST131 subclade C2	36 (28)	5 (13)	0.94 (0.29–3.03)	0.919
ST131 clades A and B	6 (5)	4 (10)	4.52 (1.06–19.18)	0.041
ST405	3 (2)	1 (3)	2.26 (0.21–24.14)	0.500
ST648	6 (5)	6 (15)	6.78 (1.79–25.64)	0.005
ST1193	2 (2)	3 (8)	10.17 (1.49–69.43)	0.018
Combined STs (adjusting variable)
Other STs	64 (50)	10 (25)	1.0 (Reference)
ST131 C2	36 (28)	5 (12)	0.89 (0.28–2.80)	0.841
High-risk STs (ST131 clades A, B, subclade C1, ST648, ST1193)	28 (22)	25 (63)	5.71 (2.42–13.47)	<0.001
FQ phenotypic resistance	90 (70)	34 (85)	2.39 (0.93–6.17)	0.071	2.03 (0.71–5.81)	0.185
Combined phenotypic resistance ≥3	80 (63)	26 (65)	1.11 (0.53–2.34)	0.775	1.00 (0.44–2.25)	0.994
Combined genetic resistance ≥3	86 (67)	28 (70)	1.14 (0.53–2.46)	0.740	1.34 (0.58–3.11)	0.498
blaCTX-M-15	82 (64)	18 (45)	0.46 (0.22–0.94)	0.034	0.93 (0.38–2.25)	0.871
blaCTX-M-27	15 (12)	13 (33)	3.63 (1.55–8.51)	0.003	1.14 (0.40–3.23)	0.804
FQ and AMG resistance aac(6')-Ib-cr	36 (28)	9 (23)	0.74 (0.32–1.71)	0.484	1.10 (0.37–3.27)	0.868
FQ genetic resistance	7 (5)	1 (3)	0.44 (0.05–3.72)	0.453	0.90 (0.10–8.25)	0.929
ExPECJJ	91 (71)	34 (85)	2.30 (0.89–5.95)	0.085	1.54 (0.52–4.53)	0.435
UPECHM	80 (63)	32 (80)	2.40 (1.02–5.63)	0.044	0.95 (0.23–3.99)	0.944
Virulence operons ≥18	33 (26)	1 (3)	0.07 (0.01–0.56)	0.012	0.09 (0.01–0.73)	0.024
Mean no. of virulence operons (95% CI)	14.5 (13.8–15.2)	14.2 (13.3–15.1)		0.696
Adhesion (all)	121 (95)	32 (80)	0.23 (0.08–0.69)	0.008	0.24 (0.07–0.82)	0.023
pap and/or sfa and/or foc	60 (47)	3 (8)	0.09 (0.03–0.31)	<0.001	0.12 (0.03–0.43)	0.001
Adhesin(s)
afa and/or draBC	17 (13)	6 (15)	1.15 (0.42–3.15)	0.783	1.54 (0.51–4.61)	0.442
sfa and/or foc	12 (9)	0 (0)	N.A.	0.071
papACGH	55 (43)	3 (8)	0.11 (0.03–0.37)	<0.001	0.13 (0.04–0.47)	0.002
fim	116 (91)	30 (75)	0.31 (0.12–0.79)	0.014	0.30 (0.10–0.88)	0.028
Immune evasion
kpsM	88 (69)	36 (90)	4.09 (1.36–12.27)	0.012	2.71 (0.76–9.63)	0.124
kfiC	25 (20)	14 (35)	2.22 (1.01–4.85)	0.046	1.11 (0.44–2.81)	0.822
iss	93 (73)	26 (65)	0.70 (0.33–1.49)	0.354	1.31 (0.54–3.15)	0.546
tcpC	9 (7)	0 (0)	N.A.	0.117
Invasion
aslA	114 (89)	40 (100)	N.A.	0.024
fdeC	126 (98)	37 (93)	0.20 (0.03–1.22)	0.080	0.11 (0.02–0.82)	0.031
ibeA	5 (4)	2 (5)	1.29 (0.24–6.94)	0.763	2.01 (0.32–12.53)	0.453
ompA	128 (100)	40 (100)	N.A.	1.000
ompT	87 (68)	30 (75)	1.41 (0.63–3.17)	0.400	0.72 (0.24–2.14)	0.555
Siderophore, iron uptake
iuc	93 (73)	35 (88)	2.63 (0.96–7.27)	0.061	1.60 (0.50–5.10)	0.430
iutA	87 (68)	35 (88)	3.30 (1.20–9.04)	0.020	2.24 (0.71–7.05)	0.168
chuA	101 (79)	36 (90)	2.41 (0.79–7.35)	0.123	1.09 (0.28–4.26)	0.896
iroN	19 (15)	2 (5)	0.30 (0.07–1.36)	0.118	0.52 (0.10–2.56)	0.418
Protease
pic	10 (8)	1 (3)	0.30 (0.04–2.44)	0.262	0.60 (0.07–5.27)	0.645
sat	70 (55)	28 (70)	1.93 (0.90–4.14)	0.089	1.04 (0.35–3.09)	0.944
vat	16 (13)	4 (10)	0.78 (0.24–2.48)	0.671	0.86 (0.24–3.02)	0.808
Toxin
hly	38 (30)	0 (0)	N.A.	<0.001
cnf1	31 (24)	0 (0)	N.A.	<0.001
Miscellaneous
yfcV	106 (83)	34 (85)	1.18 (0.44–3.14)	0.746	0.35 (0.09–1.37)	0.131
fyuA	107 (84)	40 (100)	N.A.	0.004
hsp	74 (58)	25 (63)	1.22 (0.59–2.52)	0.599	0.72 (0.26–1.96)	0.518
malX	127 (99)	39 (98)	0.31 (0.02–5.02)	0.408	0.89 (0.05–15.00)	0.935

^aORs and CIs calculated with logistic regression, except for variables with either 0 or 100% counts in the prostate biopsy group, where Fisher's exact test was used for measure of association. N.A., not available.

Isolates from patients with PPBS carried fewer virulence factors than other isolates. In univariate analysis, when analyses were performed without adjusting for ST, five factors had a significant negative association with prostate biopsy. These were adhesins, particularly pap, sfa and/or foc, and fim, and a virulence operon score of ≥18. In addition, none of the PPBS were caused by isolates harboring the toxins hly, cnf1 (Fisher’s exact test, P < 0.001), or immune evasion gene tcpC (P = 0.117). Conversely, all PPBS isolates carried fyuA and aslA. Capsule group 2 (kpsM), iutA, and UPEC_MH statuses were also positively associated with PPBS in univariate analysis. Phenotypic fluoroquinolones (FQ) resistance was numerically more common in isolates from BSI after prostate biopsy (85%) than in other infections (70%) (P = 0.071).

In multivariable analysis, when adjusting for the sequence types with a strong association with PPBS, only the traits adhesin (especially pap) and a virulence operon score of ≥18 remained significant with a negative association.

UTI was the most common cause of infection, occurring in 64% of patients (Table 5). Factors significantly associated with UTI in multivariable analysis were urinary catheterization at admission, diabetic end-organ damage, kidney stones, gender, chronic neurological disease, prior endoscopy, and E. coli carrying the adhesin pap. When excluding PBBS from the analysis, urinary catheterization, diabetic end-organ damage, kidney stones, and pap remained with a significant positive association.

TABLE 5.

Association between patient characteristics, virulence genes, and UTI (n = 260)

Characteristic	Other source of infection (no. [%]) (n = 94)a	UTI only (no. [%]) (n = 166)a	Univariate OR (95% CI)	P value	Multivariate OR (95% CI)	P value
Female gender	26 (28)	66 (40)	1.73 (1.00–2.99)	0.051	2.40 (1.23–4.70)	0.011
Age group					Not included
0–39	2 (2)	7 (4)	1.0 (Reference)
40–49	7 (7)	12 (7)	0.49 (0.08–3.04)	0.444
50–59	14 (15)	28 (17)	0.57 (0.1–3.12)	0.518
60–69	44 (47)	43 (26)	0.28 (0.05–1.42)	0.124
70–79	13 (14)	37 (22)	0.81 (0.15–4.42)	0.811
≥80	14 (15)	39 (23)	0.8 (0.15–4.3)	0.791
Daily living activity ≥2	21 (78)	57 (34)	1.8 (1.02–3.25)	0.044	n.s.b
Charlson diabetic end-organ damage	7 (7)	33 (20)	3.08 (1.31–7.28)	0.010	3.20 (1.18–8.68)	0.023
Chronic neurological disease	11 (12)	44 (27)	2.72 (1.33–5.57)	0.006	2.83 (1.21–6.65)	0.017
Charlson hemiplegia	1 (1)	10 (6)	5.96 (0.75–47.32)	0.091	n.s.
Chronic kidney dysfunction	6 (6)	23 (14)	2.36 (0.92–6.02)	0.073	n.s.
Kidney stones	1 (1)	10 (6)	5.96 (0.75–47.32)	0.091	10.11 (1.13–90.46)	0.039
Arrival from nursing home	8 (9)	38 (23)	3.19 (1.42–7.17)	0.005	n.s.
Any prior endoscopy	2 (2)	18 (11)	5.59 (1.27–24.67)	0.023	7.35 (1.48–36.59)	0.015
Any prior surgery	4 (4)	15 (9)	2.24 (0.72–6.94)	0.164	n.s.
Any prior EPE-positive culture	15 (16)	56 (34)	2.68 (1.42–5.08)	0.002	n.s.
Urinary catheterization at admission	8 (9)	60 (36)	6.0 (2.76–13.41)	<0.001	8.21 (3.38–19.97)	<0.001
ST131 subclade C2	20 (21)	56 (34)	1.88 (1.04–3.40)	0.035	n.s.
ExPEC	73 (78)	129 (78)	1.00 (0.55–1.84)	0.992	not included
UPEC	65 (69)	107 (64)	0.81 (0.47–1.39)	0.443	not included
Virulence operons ≥18	12 (13)	43 (26)	2.39 (1.19–4.80)	0.015	n.s.
papACGH	19 (20)	86 (52)	4.24 (2.36–7.64)	<0.001	5.27 (2.70–10.29)	<0.001
fim	78 (83)	151 (91)	2.06 (0.97–4.40)	0.060	n.s.
kfiC	24 (26)	25 (15)	0.52 (0.28–0.97)	0.040	n.s.
iss	65 (69)	118 (71)	1.10 (0.63–1.90)	0.743	not included
aslA	90 (96)	150 (90)	0.42 (0.14–1.29)	0.128	n.s.
hly	9 (10)	53 (32)	4.43 (2.07–9.48)	<0.001	n.s.
cnf1	7 (7)	41 (25)	4.08 (1.75–9.51)	0.001	n.s.

^aUTI only, final source urinary tract infection; other source of infection includes prostate biopsy, abdominal infection, pneumonia, other source, and unknown.

^bn.s., not significant in multivariate analysis.

Isolates from UTI-derived BSI were clearly different from those from PPBS. In isolates from UTI, any adhesin was detected in 96% and pap in 52% of isolates. In isolates from PPBS, any adhesin was detected in 80% and pap only in 8% of isolates.

Association of microbiological determinants with patients without risk factors for EPE BSI.

In a previous study, we saw that 20% of community-onset ESBL-producing Enterobacterales (EPE) BSI occurred in patients without any major risk factors for EPE BSI (prior EPE-positive culture, prior health care abroad ≤6 months, prior prostate biopsy) and no hospitalization in the previous year (10). In the present study, 59 patients (23%) belong to this EPE BSI low-risk group, lacking the mentioned risk factors (see Table S2 in the supplemental material). There was a univariate negative association with EPE BSI low-risk status for the microbial determinants FQ phenotypic resistance (58% versus 76%; OR, 0.44 [95% CI, 0.24 to 0.81]; P = 0.008) and phenotypic MDR (46% versus 66%; 0.43 [0.24 to 0.78]; P = 0.005), kfiC (7% versus 22%; 0.25 [0.09 to 0.73]; P = 0.011) but a positive association with pap (53% versus 37%; 1.90 [1.06 to 3.41]; P = 0.032) and the protease pic (14% versus 5%; 3.00 [1.12 to 7.98]; P = 0.028). Patient factors with an association were female gender, age <60, low comorbidity (Charlson, 0 to 1), and completely able in daily living activity. These patients also had a low sequential organ failure assessment (SOFA) score (51% had a SOFA of ≥2 versus 74%; P < 0.05).

Patients with repeated episodes.

Distinct repeated episodes of EPE BSI (>30 days apart) were recorded for 40 patients (15%). A history of hematologic cancer, diabetes mellitus, chronic wound, and chronic neurological disease was associated with repeated infection (see Table S2). None of the patients with PPBS had repeated episodes. Sequence types ST131 (25 isolates, 63% of patients with repeated episodes) and ST405 (4 isolates, 10%) were associated with repeated infection. For ST131, the univariate OR was 2.8 (95% CI, 1.3 to 6.0; P = 0.008), and for ST405, it was 7.5 (1.7 to 29.2; P = 0.006). For 16 out of the patients with repeated episodes of BSI, more than one isolate was available for WGS. For 15 out of these 16 patients, the repeated episode was caused by a closely related isolate of the same ST (Fig. 3). For 13 patients, there were ≤10 SNPs between different isolates sampled from the same patient. Two patients with isolates belonging to ST354 had 11 to 100 SNPs between different isolates sampled from the same patient.

FIG 3.

SNP tree of strains from 16 patients with repeated episodes from whom several isolates were available for WGS. The number of SNPs between isolates from the same patient and the number of SNPs between the reference genome for each ST and the patient isolates are graded according to legend. Isolates from the same patient are colored with the same color in the isolate identification column. Tree scale is the number of substitutions per site.

DISCUSSION

Herein, we have performed an exploratory analysis of the association of various patient characteristics and microbiological determinants with severity of disease (septic shock and 30-day mortality), source of infection (PPBS and UTI), patients without risk factors for EPE BSI, and patients with repeated episodes of EPE BSI in a cohort of patients with community-onset BSI caused by ESBL- and pAmpC-producing E. coli. The most important finding was the association of the serum resistance-associated gene iss with septic shock in humans.

ST and resistance gene distribution.

The global dominance of ST131 among ESBL-producing E. coli is well described, with reported proportions of 20 to 60% from ESBL-EC BSI (2, 11). Here, the distribution of STs among the included patients with ESBL/pAmpC-EC community-onset BSI was very similar to that found in a previous Swedish nationwide study of ESBL/pAmpC-EC BSI, supporting the generalizability of our findings (4). Isolates belonging to the FQ-resistant ST131 subclade C2 were the most common and carried more virulence genes and resistance traits than isolates belonging to many other STs, which is in concordance with previous studies (5, 8). Another subclade, ST131 subclade C1-M27 (carries bla_CTX-M-27), is predominant in Japan but also reported globally (11). Here, 66% of ST131 subclade C1 isolates belonged to the subclade C1-M27. These represented 7% of all isolates, and ST131 subclade C1-M27 was the overall second most common group after ST131 subclade C2.

Association of host factors and microbiological determinants with severity of disease.

We used septic shock or death within 3 days as the primary outcome to assess association with virulence factors. The overall all-cause 30-day mortality was low, only 6.5%, compared to the following comparable studies: community-onset E. coli BSI in England, 10.5% (12); ESBL-EC BSI in Taiwan, 28.1% (8); ESBL-EC BSI in Spain, 24.6% (9); E. coli BSI in Sweden, 12% (13); E. coli UTI-derived BSI in Denmark, 11.7% (14); and E. coli BSI Brazil, 37.5% (15). However, this was expected since our study, contrary to the other publications, only included community-onset BSI. In addition, blood cultures are routinely sampled even with low clinical suspicion of BSI in Sweden, which might result in less severe BSI episodes being included.

The 108 virulence genes from 26 virulence operons examined in this study were genes with previously documented association with ExPEC. Twenty operons were listed as associated with UPEC and/or neonatal meningitis-associated E. coli (NMEC) in the Virulence Factor of Pathogenic Bacteria database (VFDB), and an additional six operons were selected from other studies. The evidence for these traits being virulence factors for extraintestinal disease are generally generated from studies comparing isolates from invasive disease with commensal or environmental isolates and, in some cases, from experimental infection models in mice (16). Many virulence factors of ExPEC are genes facilitating invasion of the urinary tract and from there into the bloodstream. All of the present isolates are from patients with BSI, which restricts the possibility to study the virulence potential of the isolates from this aspect. However, the same virulence factors could also be important for the severity of disease.

The association of iss with septic shock is an interesting observation (Table 3) since this gene is associated with the ExPEC subtype avian pathogenic E. coli (APEC) and NMEC. All septic shock episodes occurred with isolates carrying iss or affected a patient with hematologic malignancy or transplantation. iss and the closely related gene bor are cell membrane-bound lipoproteins, and both have been reported to confer complement resistance (17). Several prior studies indicate that iss is important for the severity of disease. Studies on poultry have shown an association between iss and colibacillosis (18). Vaccination with recombinant Iss protein induced a humoral response, and vaccinated birds had lower lesion severity scores (19, 20). In one other study on human BSI with ESBL-EC, iss was associated with 30-day mortality (OR, 3.36; P = 0.030) (8).

These results suggest that iss might be of importance for septic shock in E. coli BSI in humans. As a surface-expressed protein, which is accessible for monoclonal antibodies, Iss could potentially be a useful target for antivirulence therapy with monoclonal antibodies. Further studies experimentally exploring the potential of Iss in this respect are warranted, since the observed association could be due to low sample size and the association might not be causal. Nevertheless, the mentioned studies on poultry makes this finding intriguing.

The low rate (6.5%) of 30-day mortality could explain the apparent lack of an association between microbiological determinants and 30-day mortality, as the power of these analyses is quite low. This result should not directly be interpreted as evidence for no association, as larger studies are needed for this to be clarified.

Association of microbiological determinants with source of infection (PPBS and UTI).

Adhesion genes, and especially pap, were associated with UTI, but the presence of adhesion factors is not necessary to cause post-prostate biopsy sepsis, which has also been shown previously (21–23). Two studies that also included susceptible E. coli, not specifically ESBL-producing isolates, had similar results. Isolates related to PPBS were low in pap and/or sfa and/or foc, iroN, and malX (both studies), and in one study, there was also a negative significant association with hly, cnf1, and several other virulence genes (22, 23). The lower virulence gene presence in isolates from PPBS is logical, as enteric bacteria are inadvertently transferred from the rectum to the bloodstream during the biopsy procedure. This means that isolates that otherwise have relatively low pathogenicity nevertheless can cause BSI after prostate biopsy, such as isolates belonging to the STs ST648 and ST1193. Isolates of these two STs generally lacked iss, which might mean that they are less probable to cause severe infection, which might contribute to the absence of septic shock/death in BSI after prostate biopsy. Conversely, for severity of disease, patient history such as hematologic cancer and other comorbidities were more strongly associated than the presence of iss (Table 3).

Significance of ExPEC_JJ and UPEC_HM status.

E. coli isolates qualified as ExPEC_JJ have previously been associated with UTI compared to commensal isolates. Of the present BSI isolates, 78% qualified as ExPEC_JJ. ExPEC_JJ had a negative association with septic shock but did not have any statistically significant association with repeated episodes, source of infection, or patients with low risk of EPE BSI. pap seemed to be the most important part of the ExPEC_JJ association with septic shock. Of ST131, 100% qualified as UPEC_HM while 97% of both ST131 and ST131 C2 qualified as ExPEC_JJ. The results for ST131 are consistent with what have been shown previously in a study by Kanamori et al. (5), in which 95% of ST131 and 100% of ST131 C2 fulfilled criteria for ExPEC_JJ.

The alternative definition UPEC_HM had a negative univariate association with septic shock and a positive association with PPBS, the latter mainly caused by the presence of fuyA.

Repeated episodes.

ExPEC_JJ status was not associated with repeated infections. However, there was a statistically significant correlation with ST, as ST131 and ST405 together comprised 73% of patients with repeated episodes compared to 47% of single episodes. This supports the previous findings that ST131 clade C2 is associated with persistent infection and subsequent new infections (24). Isolates from repeated episodes within the same patient were commonly very closely related.

Limitations.

There are several limitations to this study. The retrospective nature means that some clinical data might be incomplete. SOFA score was calculated based on available data, which in many cases lacked information, particularly of bilirubin levels. However, this did not affect the classification of septic shock, as all patients with a need of vasopressor treatment had a SOFA score of ≥2. Although the presence of virulence genes is indicative of a virulent phenotype, this was not experimentally confirmed in a virulence model. Additionally, there might be many other virulence genes of importance that were not studied. Since the study population was geographically limited to Stockholm, the results might not be generalizable globally.

The genome sequence quality was varying, and two different extraction methods were used. Detection of virulence genes with WGS also has several general limitations. Illumina sequencing generates short contigs. A gene of interest might not be detected if the gene is divided into two contigs, and alternative assembly methods might give different results. For detection of virulence genes with WGS, there is no consensus on cutoffs for similarity and gene coverage. We applied a threshold of 80% for both similarity and gene coverage, since we saw that some virulence genes, particularly iutA and fimA, in many cases had as low as 89% similarity. fimA is known to have a high allelic diversity; this could be due to strong selection for antigenic variation under immune pressure (25). This threshold has also been used previously (5). However, for other more conserved genes, this cutoff might be too low and cause false positives. We do not believe that these limitations together greatly influence the main conclusions of the study, but this limits the possibility of even more detailed analysis of very closely related genes.

This was an exploratory study, analyzing the association of a relatively high number of different host and microbial factors with different outcomes and types of infection. When analyzing many variables, there is a possibility that some results may be due to chance, which is why the results should be interpreted with care and should be repeated in future studies. However, for the main conclusions, there is support in prior studies which has been discussed above.

Conclusion.

We found that the fluoroquinolone-resistant high-risk clone ST131 subclade C2 comprised 29% of all patients in community-onset BSI caused by ESBL/pAmpC-EC in Stockholm from 2012 to 2015. The risk factors associated with septic shock were patient host factors (reduced daily living activity, hematologic cancer, or transplantation), but the presence of the E. coli virulence factor iss, increased serum survival, was also significant. Adhesins, particularly pap, were associated with UTI-derived BSI, while isolates from post-prostate biopsy sepsis had a low overall number of virulence operons, low adhesin occurrence, and belonged to sequence types ST131 clades A, B, and subclade C1, ST1193, and ST648. ST131 and ST405 were associated with recurrent episodes, and the repeated isolates from the same patient were often closely related.

MATERIALS AND METHODS

Setting and patients.

Patients attending four emergency departments (ED) in Stockholm between 1 January 2012 and 31 December 2015 were eligible for inclusion if they were aged ≥18 years and had ESBL/pAmpC-EC isolated from a blood culture drawn at the ED. Patients were excluded if they were transferred from another hospital. The patients were identified through the laboratory information system of the Karolinska University Laboratory (see Fig. S1 in the supplemental material). The patients were a subset of patients described in a previous study on prediction of community-onset BSI caused by ESBL-producing Enterobacterales (10).

A total of 293 episodes from 271 patients fulfilled the inclusion criteria. Bacterial isolates were available for whole-genome sequencing for 278 episodes from 260 patients. The first episode from each patient was included in the main analysis, and a separate analysis was performed on the 16 patients with repeated episodes for whom two or more isolates from distinct episodes >30 days apart were available.

The study was approved by the Regional Ethical Review Board in Stockholm (Dnr 2014/277-31). Patient data were handled in accordance with European General Data Protection Regulations (GDPR).

Clinical data.

Patient characteristics and clinical data were collected by review of medical records. Clinical data included Charlson comorbidity index (26), assessment of disease severity by sequential organ failure assessment (SOFA) score (27), mortality, antibiotic treatment, and suspected and final source of infection. SOFA score was calculated on available data for all patients. For the cardiovascular part of the SOFA score, information regarding noradrenaline/adrenaline doses was incomplete. Thus, any use of vasopressors was coded as 3. When calculating the SOFA score, the baseline SOFA score of the patient before the current infection episode was subtracted. Septic shock was defined according to the Third International Consensus Definitions for Sepsis and Septic Shock as an increase of SOFA of ≥2 from the baseline and serum lactate level of >2 mmol/liter and need of vasopressor treatment (27). The primary outcome variable for severity of disease was defined as septic shock or death within 3 days. Daily living activity before the onset of infection was graded into the following four categories: 0, completely able; 1, limited activities; 2, needs help with activities of daily living; 3, bedridden.

Detection of ESBL/pAmpC production and antimicrobial susceptibility testing.

Results from the routine microbiological laboratory were used for species identification, antimicrobial susceptibility testing (AST), and phenotypic ESBL detection. The microbiological methods used are described in detail in the supplementary material. Isolates were defined as ESBL/pAmpC producers if the E. coli demonstrated a positive phenotypic test indicating production of classic ESBL or the presence of a genetically verified plasmid AmpC β-lactamase according to the case definitions of the Swedish Public Health Agency and EUCAST recommendations (28, 29).

Phenotypic multidrug resistance (MDR) was defined as resistance to more than two of the following antibiotic classes: extended-spectrum cephalosporins, aminoglycosides (AMG), fluoroquinolones (FQ), or trimethoprim-sulfamethoxazole (SXT).

Whole-genome sequencing.

Extraction of genomic DNA was performed with MagNA Pure 96 (Roche Diagnostics) on 220 isolates and with the EZ1 Advanced XL system (Qiagen) on 58 isolates of ESBL/pAmpC-EC. The quantity of the extracted DNA was measured using a Qubit double-stranded DNA (dsDNA) assay kit (Life Technologies Europe). Extracted DNA was sequenced on Illumina HiSeq 2500 at the Science for Life laboratory (SciLifeLab, Solna, Sweden), generating 2 × 100 paired-end sequences. Sequencing quality control showed that 10× coverage was >97% for 97% of isolates, and 30× coverage was >90% for 74% of isolates. Mean duplication rate was 6.7% (standard deviation [SD], 3.1), and mean median insert size was 168 (SD, 37). All isolates had a mean sequencing depth of >30×. Assembly of raw reads and the identification of resistance and virulence genes were performed through the in-house bioinformatic pipeline microSALT (https://github.com/Clinical-Genomics/microSALT).

Phylogenetic analysis.

For phylogenetic analysis, SNP analysis was performed on assembled genomes using Snippy v4.4.3 (https://github.com/tseemann/snippy) (30) with EC958 as a reference strain. Next, Gubbins v2.3.4 (31) was used to remove recombination events with a threshold set to filter out taxa with more than 30% missing data. Snp-sites v2.4.1 (32) was used to extract variant sites, and snp-dists v0.6.3 (https://github.com/tseemann/snp-dists) was used to convert this data into an SNP matrix. RAxML v8.2.12 (33) was used to create randomized accelerated maximum likelihood (RAxML) trees using the GTRGAMMA model with a bootstrap of 1,000. Finally, phylogenetic trees were annotated using R v3.5.1 (34) with the following additional packages: ggtree v1.14.6 (35) and ape v5.3 (36). Separate trees were created for ST131 and non-ST131 isolates. The interactive tree of life (iTOL) software v5.5 (37) was applied for visualization of phylogenetic trees and metadata. The reference genomes used for ST131 were previously published (38–41). Reference genome assemblies were downloaded from EnteroBase https://enterobase.warwick.ac.uk (42, 43). Data Set S1 in the supplemental material lists all study and reference genomes used.

Resistance and virulence gene detection.

The resistance gene definitions were based on microSALT’s resistance gene database, which in turn used the ResFinder database v2.1 (44). For positive identification of resistance genes, a match with 97% similarity and 90% gene coverage was required.

The virulence gene definitions consisted of genes with previously described association to UPEC and ExPEC isolates. Gene sequences associated with UPEC and NMEC were downloaded from the Virulence Factor of Pathogenic Bacteria database (VFDB) (45). In addition, the genes malX, ompT, yfcV, hcp, fyuA, iss, and kfiC were included (5, 7, 8, 46). See the supplementary material for a full list of the 108 virulence genes, representing 26 virulence operons, examined. For virulence genes, a match with 80% similarity and 80% gene coverage was required. Isolates were defined as ExPEC_JJ if they were positive for ≥2 of any pap, any sfa and/or foc, any afa and/or dra, any kpsMT, and iutA (5, 6). UPEC_HM strains were defined by the presence of ≥3 of the following genes: chuA, fyuA, vat, and yfcV (46).

Statistical analysis.

The first isolate from each patient was included in the statistical calculations. Logistic regression was used to compute the odds ratio (OR) and 95% confidence interval (CI). For variables with 0 or 100% count in the outcome group, Fisher’s exact test was applied.

Variables included in multivariate logistic regression had to be biologically sound and have a P value of <0.2 in the univariate analysis. Variables were removed from the model in a manual backwards stepwise fashion, removing predictors with the highest P value until all were <0.05. Following each removal, the current model was compared with the previous one using the likelihood ratio test. If the test indicated that the current model was significantly different, the latest removed variable was reintroduced.

The analyses of the association of virulence factors with PPBS were performed both with and without adjusting for the STs that were strongly associated with PPBS. Based on the results of univariate analysis, the sequence types/clades that had the strongest association with PPBS (ST131 clades A, B, and subclade C1, ST648, and ST1193) were combined into a category named “high-risk ST.” The combined ST variable, with 3 levels (other ST, ST131 C2, and high-risk ST), was then used as an adjustment factor for univariate analysis of all other microbiological determinants in men. As virulence factors are often clonally distributed and coexist, adjusting for ST could clarify which virulence factors were most strongly associated with PPBS in an independent way.

Stata statistical software release 13 (StataCorp LP, College Station, TX) was used for all statistical calculations.

Data availability.

Genomic sequences for the included isolates have been deposited at NCBI under BioProject accession number PRJNA612606 and BioSample accession numbers SAMN14378527 to SAMN14378804 (see Data Set S1 in the supplemental material). Additional data and analysis files are available upon request.