Hospital-diagnosed infections with Escherichia coli clonal group ST131 are mostly acquired in the community

The worldwide spread of E. coli ST131 has significantly contributed to the dissemination of E. coli producing extended-spectrum β-lactamases (ESBL). In a French University hospital, we assessed the molecular features of ESBL-producing E. coli and identified risk factors in patients for colonization or infection with E. coli ST131. Over a 2-year period (2015–2017), each patient with at least one clinical isolate or one screening isolate positive with ESBL-producing E. coli were included (n = 491). The ST131 clonal group accounted for 17.5% (n = 86) of all ESBL-producing E. coli and represented 57.3% isolates of phylogroup B2. FimH-based sub-typing showed that 79.1% (68/86) of ST131 isolates were fimH30, among which 67.6% (n = 46), 20.6% (n = 14) and 11.8% (n = 8) isolates harbored genes encoding the ESBL CTX-M-15, CTX-M-27, and CTX-M-14, respectively. The multivariate analysis identified two factors independently associated with ST131 ESBL-producing E. coli isolates: infection (Odds ratio [OR] = 1.887, 95% confidence interval [CI]: 1.143–3.115; p = 0.013) and community acquisition (OR = 2.220, 95% CI: 1.335–3.693; p = 0.002). In conclusion, our study confirmed the predominance of ST131 clonal group among ESBL-producing E. coli and the difficulty to identify common risk factors associated with carriage of this pandemic clonal group.

The worldwide spread of Sequence Type 131 (ST131) Escherichia coli has significantly contributed to the dissemination of E. coli resistant to fluoroquinolones (FQ) and producing extended-spectrum β-lactamases (ESBL) 1 . The molecular epidemiology of BLSE-encoding gene in E. coli is a typical model to decipher the spread of antibiotic resistance in pathogenic bacteria. Hence, it combines two levels of complexity: a first level related to the strain tested through the examination of core genes and the second level related the bla ESBL genes carried by the accessory genome and mobile genetic elements. These isolates exhibit the usual virulence factors associated with extra-intestinal pathogenic E. coli belonging to phylogroup B2 2 . Recent analyses of the whole genome sequences of large collections of ST131 established that multi-drug-resistant clades C1/fimH30-R and C2/fimH30-Rx emerged in the 1980′s, most likely in North America 3,4 . The evolutionary history suggests that fimH30 allele was acquired by a single FQ-susceptible ancestor, which later became resistant to this antibiotic family (subclone H30R). The further acquisition of bla CTX-M-15 created another distinct subclone (H30-Rx) which derived from a single common ancestor within H30R. FimH30 subclones have disseminated globally as a result of clonal expansion. Despite these advances in our knowledge of the epidemiology of ST131, there remain considerable gaps regarding reservoirs, mechanisms of transmission and patient risk factors of ST131 acquisition. In a previous study, we described the long-term epidemiology of this clonal group in our University hospital, and demonstrated that it was predominant, although it only accounts for less than 20% of ESBL-producing E. coli 5 . The objectives of this study were to assess the molecular features of ESBL-producing E. coli and to identify risk factors for colonization or infection with the clone ST131.
Risk factors associated with ST131 ESBL-producing E. coli. Univariate analysis revealed that, accommodation in a nursing home (p = 0.002) and hospitalization in a medical unit (p = 0.001) were significantly associated with the carriage of ST131 isolates (Table 2). In contrast, solid cancer/blood cancer, immunosuppression, home housing, hospitalization in intensive care and hematology units were protective factors (p < 0.05). ST131 E. coli isolates were more frequently associated with an infection than colonization (defined by the presence of clinical signs or healthy carriage, respectively) (p < 0.001). In addition, they were more frequently isolated from urinary samples (p = 0.005) and community acquired (p < 0.001). Rectal/stool samples were associated with non-ST131 E. coli (p < 0.001). Time between admission (p < 0.001) and sample collection and length of stay were shorter for patients with ST131 isolates (p = 0.018). Moreover, for these patients 30-day all-cause mortality was lower (p = 0.004). Multivariate analysis identified only two factors positively associated with ST131 ESBLproducing E. coli isolates: infection (p = 0.013) and community acquisition (p = 0.002) ( Table 3).

Discussion
So far, ten studies on risk factors associated with ST131 have been reported worldwide. The main results of these studies are summarized in Table 2. They identified many risk factors for acquisition of ST131: elderly, long-term-residency at care facility or nursing homes, diabetes mellitus, absence of urinary catheter, previous exposure to antibiotics or absence of prior antibiotic treatment, recent surgery, use of proton pump inhibitor, or cancer. Risk factors were not consistent from one study to another, presumably due to variation in study designs, patient populations, or bacterial populations considered (i.e. all ST131 isolates or only ESBL producers). It may also reflect the variability of ST131 prevalence according to countries or populations. In our study, multivariate analysis associated ST131 isolates with infection, certainly reflecting the higher virulence of the phylogroup B2  www.nature.com/scientificreports/  www.nature.com/scientificreports/ and in line with the fact that ST131 infection are more frequently acquired in the community 1 . Overall, identification within the human hosts of risk factors associated with acquisition of ESBL-producing ST131 is not that simple. Our interpretation is that the global spread of ST131 relies more on bacterial factors than on host characteristics. A recent outstanding study have demonstrated that the success of ST131 is related to the ability of ST131 subclades to evolve toward separate ecological niches and to select genes involved in anaerobic metabolism and human colonization 6 . This adaptation certainly accounts for the long-term intestinal colonization and high transmissibility rate in both the community and hospital settings of E. coli ST131 7 . The present study confirms previous data on ST131 epidemiology in our hospital 5 . First, the clone ST131 accounted for 17.5% but predominated among ESBL-producing E. coli. Secondly, the fimH typing revealed that 68 of the 86 clonal complex CC131 isolates (79.1%) were of H30Rx subclone and that pandemic clusters C2-M15 (n = 44; producing CTX-M-15) and C1-M27 (n = 13; producing CTX-M-27) were over-represented in the H30Rx subclone. H41 isolates (n = 9), very likely belonging to O16 serotype clade A, represented the second most frequent fimH type. H22 clade B (the predecessor of fimH30) was not detected although present in 2010-2012 in our hospital 5 . Thirdly, ESBL-producing ST131 should be considered as a cluster of distinct clonal subsets with specific characteristics rather than as a unified entity.
Phylogroup B2 accounts for a large subset of EXPEC strains that are responsible for the majority of human extra-intestinal infections globally 2 . Until recently, ST131 remained the only phylogroup B2 clone that successfully evolved under antibiotic resistance by acquiring chromosomal resistance to fluoroquinolones and plasmids harboring bla CTX-M genes 1 . Unexpectedly, more than 40% of our collection of ESBL-producing phylogroup B2 isolates were not ST131. We found historical clones responsible for UTIs (ST73, ST95) as well as emerging clones (ST127, ST141). Birgy and colleagues also recently described the presence of ST73 and ST95 in a large collection of ESBL-producing E. coli isolated from UTIs in French children 8 . ST141, which was the most frequent clone after ST131 in our collection of phylogroup B2 ESBL-producing E. coli, was considered as a Shiga toxin-producing E. coli (STEC) of serotype O2:H6 9 . All E. coli ST141-B2 isolates (n = 13) have already been tested in another study focusing on ST141 clonal group (manuscript in preparation). The sequencing of their full genomes revealed the absence of stx genes. More recently, an additional study suggests that ST141 isolates may be a hybrid positioned between Shiga toxin-producing and uropathogenic E. coli, serving as a "melting pot" for pathogroup conversion 10 . The emergence of ESBL in the clone ST141, as observed in our study, is worrisome and the investigation of phylogeny and the identification of virulence and resistance determinants are ongoing.
This study has some limitations. It would have been interesting to include all E. coli isolates, independently of the antibiotic resistance profile. However, ESBL production in E. coli is the main concern of public health and this justified the design of our study 11 . Although our study lacks novelty, our data are valuable since produced from a large cohort of more than 500 hospitalized patients over 2 years, with a systematic inclusion of patients with at least one clinical or screening positive E. coli isolate producing ESBL. This differentiates our study from previous ones. To conclude, our study confirms both the predominance of ST131 among ESBL-producing E. coli and the difficulty to identify common risk factors associated with this pandemic clonal group.

Methods
Setting, study period and patients included. We conducted a prospective observational cohort study for a 2-year period (February 2015-January 2017) in the Besançon University hospital, a 1200-bed hospital with approximately 50,000 admissions and 320,000 patient-days annually. Over the study period, all patients admitted in our hospital with at least one clinical isolate or one screening isolate positive with ESBL-producing E. coli were included (with the exception of consultations and day-care).
Bacterial isolates. All the isolates were identified as E. coli by MALDI-TOF MS Microflex LT (Bruker Daltonik GmbH, Bremen, Germany) according to the manufacturer's recommendations and routinely tested for ESBL production using the synergy test 12 . The prevalence of E. coli BLSE in clinical isolates was 7.5% among total E. coli. All ESBL-producing E. coli isolates were stored at the Centre de Ressources Biologiques Filière Microbiologique, Besançon (CRB-FMB, Biobanque BB-0033-00090). We retained only the first isolate of ESBLproducing E. coli for patients with multiple samples positive with this species.
Microbiological analysis. E. coli isolates were typed by phylogrouping as previously described 13 and by MLST according to Achtman scheme for phylogroup B2 isolates 14 . ESBL-encoding genes (bla CTX-M group 1 , bla CTX-M group 9, bla CTX-M group 2, bla SHV and bla TEM ,) were identified by PCR and sequencing 15 . For ST131 isolates, fimH sub-typing was determined by PCR and sequencing as previously described 16 . Among them, the distribution of the fimH30 isolates into sub-clades C1 and C2 was further determined by multiplex PCR 17 .
Patient data collection. Data were collected prospectively from patient medical records. We collected the following data for each patient: (i) demographical data (age, sex, housing), (ii) comorbidities and medical data (i.e. diabetes, immunodeficiency, cancer, high blood pressure, renal failure, liver failure, organ transplantation), (iii) medical history in the preceding year (hospitalizations, surgical interventions, antibiotic treatments), (iv) data about the current hospitalization (dates of admission and discharge, hospitalization ward, surgery, antibiotic treatment), (v) date and type of sample positive with ESBL-producing E. coli, (vi) data about the colonization or infection with ESBL-producing E. coli (type, site, origin) and (vii) patient outcome (length of stay, mortality at 30 days). We classified infections due to ESBL-producing E. coli as either community-or hospital-acquired when the first positive sample culture was collected ≤ 2 days or > 2 days after admission, respectively. Informed consent was obtained from the patients for the use of their clinical samples and data for this study. The patients whose anonymized data (age; risk factors) were given the following information: "Use of samples and microorganisms for research purposes: samples (blood samples, biopsies, surgical specimens) can be taken to establish a diagnosis and to adapt your treatment or that of your child. Some of these samples or the microorganisms they contain may be stored for diagnostic or research purposes. Your samples are anonymized. The medical data associated with the samples and the microorganisms are collected in a computer file authorized by the CNIL. You have the right to access and rectify the data entered. You may at any time reconsider your decision without any consequences for your care (or that of your child) and oppose the use of biopsies and operating documents for research by contacting the Franche-Comté Regional Tumorotheque