Tracing carriage, acquisition, and transmission of ESBL-producing Escherichia coli over two years in a tertiary care hospital

Abstract

Background

The impact of community carriage on the influx of extended-spectrum beta-lactamase-producing Enterobacterales (ESBL-E) into hospitals remains understudied. In this prospective 2-year single-centre study, we investigate the community ESBL-E influx and trace the colonisation, nosocomial acquisition, transmission, and infection dynamics of ESBL-producing Escherichia coli (ESBL-Ec) in non-ICU wards at a tertiary care hospital.

Methods

This study reports primary and post hoc outcomes of the clinical trial NCT01208519 in which hospitalised patients were screened for rectal carriage of ESBL-E. ESBL-Ec isolates from ≈50% of carriers, including all patients who developed infections, were sequenced and genotyped. Endogenous infection was defined as infection by the same strain (< 10 SNPs distance) as colonizing strain.

Results

Of 3703 screened patients, 456 (12.3%) were ESBL-positive-at-admission (PA-ESBL). Of the 2268 ESBL-negative-at-admission (NA-ESBL) patients with follow-up samples, 240 (10.6%) acquired ESBL-E (HA-ESBL), with an incidence density rate of 7.96 cases/1000 patient-day, notably higher in patients receiving antibiotics (P < 0.001). PA- and HA-ESBL patients developed significantly more ESBL-E infections than ESBL-free patients (P < 0.001). Sequenced ESBL-Ec showed high clonal diversity dominated by the multidrug-resistant and highly virulent ST131 clade, C2/H30-Rx. Among ESBL-Ec infections, 60% (18/30) were endogenous. Direct between-patients transmission clusters (n = 21) involved 23.9% (48/201) of patients and 23.0% (84/366) of ESBL-Ec isolates.

Conclusions

Our data show a high prevalence of nosocomial acquisition of ESBL-E in a non-ICU setting. The study provides genomic evidence that the endogenous reservoir is the main driver of ESBL-Ec infections underscoring the need for wide implementation of antibiotic stewardship programmes to reduce antibiotic pressure.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13073-024-01424-2.

Background

The prevalence of multidrug-resistant Gram-negative bacteria among hospital-associated and community-acquired infections has shown a sharp rise over the last two decades, mostly due to the emergence and spread of extended-spectrum beta-lactamase-producing Enterobacterales (ESBL-E) species, such as Escherichia coli and Klebsiella pneumoniae [1]. According to global surveillance data collected during 2018–2019 within the SMART programme, 14.4% and 21.5% of non-ICU infections were respectively attributed to ESBL-harbouring E. coli and K. pneumoniae in Western Europe, 29.6 and 32.8% in Eastern Europe, 35.1 and 40.9% in Middle East/Africa, and 15.9 and 9.9% in the USA [2]. Recent data from the USA show an increased incidence of ESBL-E infections from 2012 to 2017 by 53.3% (from 37.55 to 57.12 cases per 10,000 hospitalizations), primarily attributed to an increase in community-acquired infections [3]. Regarding intestinal carriage, ESBL-producing E. coli (ESBL-Ec) are more common than ESBL-producing K. pneumoniae in both community and nosocomial settings [4, 5]. Worldwide, intestinal colonizing ESBL-Ec increased threefold from 7% in 2001–2005 to 25.7% in 2016–2020 among inpatients and tenfold from 2.6 to 26.4% within community settings [1].

Gut carriage with ESBL-E, primarily ESBL-Ec, constitutes the primary reservoir for subsequent infections and nosocomial transmission [6, 7]. Thus, the community burden and influx of ESBL-E harbouring patients from the community upon hospital admission plays a crucial role in shaping the complex epidemiological landscape of ESBL-E within hospital settings [8]. While these data emphasize the need for active surveillance of ESBL-E carriage to prevent infection and nosocomial transmission, the lack of comprehensive knowledge of the dynamics of sustained intra-patient carriage versus inter-patient transmission has hindered the formulation of effective infection control measures to contain transmission and infections due to ESBL-E among hospitalised patients.

Compared to several other EU/EEA countries, Italy had the highest antimicrobial resistance-related mortality and DALYs (disability-associated life years) loss in 2015 [9]. Moreover, during the last two decades, Italy was consistently among the top ten European countries with the highest resistance rates to third-generation cephalosporin among invasive E. coli and K. pneumoniae [10]. In this prospective 2-year single-centre study, our objective was to perform genomic analysis to dissect the dynamics of ESBL-Ec carriage and infection among patients hospitalised in non-ICU wards at a tertiary care hospital in Rome, Italy.

Methods

Study design

To fulfil the study’s objectives, we analysed biobank samples collected from three wards at the Italian 1560-bed Università Cattolica Sacro Cuore hospital “Policlinico A. Gemelli” (UCSC, Rome) in a 2-year (2010–2013), multicentre, prospective, longitudinal, observational cohort study (NCT01208519), and also utilised related datasets including epidemiological and clinical data. The clinical study protocol is described elsewhere [11].

Rectal swabs were collected from 3703 consenting patients and screened for ESBL-E at hospital admission, during hospitalisation and at discharge (Additional file 1). Patients with positive cultures taken within 48 h of admission were classified into the ESBL positive-at-admission cohort (PA-ESBL, Fig. 1); meanwhile, negative patients were classified into the negative-at-admission cohort (NA-ESBL). After ≥ 48 h of admission, NA-ESBL patients who had positive ESBL-E cultures were classified into the hospital-acquired ESBL cohort (HA-ESBL), while patients who remained negative throughout all screenings were classified into the ESBL-free cohort.

Fig. 1.

Flowchart showing patient inclusion, classification, number of patients (N), and number of isolates (n) isolated. Patients were classified into PA-, NA-ESBL cohorts, ESBL-free, and HA-ESBL cohorts based on their positivity for colonizing ESBL-producing Enterobacterales (ESBL-E). Sequenced ESBL-producing Escherichia coli (ESBL-Ec) were illustrated in dash-lined boxes

Infections were diagnosed by the treating clinicians based on clinical signs, physical examination following the “CDC/NHSN Surveillance Definitions for Specific Types of Infections” guideline, and microbiological data, i.e., positive isolation of causative pathogens from the suspected infection site. The infection status of the patients was documented during hospitalisation. Patients were deemed admitted with an infection if clinical signs of infection, and the causative microorganisms were identified within ≤ 2 days of admission; otherwise, the infections were considered hospital-onset.

The primary objectives of the clinical trial are addressed in this study by performing genomic characterization of ESBL-Ec (i) to delineate the impact of carriage and nosocomial acquisition of ESBL-Ec on infection development, and (ii) to quantify nosocomial transmission of ESBL-Ec. Secondarily, as post hoc outcomes, we aim to measure the prevalence and incidence of ESBL-E colonisation and acquisition, respectively, and identify the clinical characteristics of patients vulnerable to ESBL-E colonisation.

Phenotypic ESBL detection and susceptibility testing

ESBL-E were screened by cultures of rectal swabs on selective chromogenic Brilliance ESBL Agar (Oxoid, UK), identified to species level, and confirmed by double disc diffusion synergy test. In vitro susceptibility to expanded-spectrum cephalosporins, ciprofloxacin, and six other classes of antimicrobial agents was determined by disk diffusion (Additional file 1). Isolates were considered multidrug-resistant (MDR) if they were resistant to at least one drug in at least three classes of antimicrobial agents [12].

Whole-genome sequencing

Among the identified ESBL-E isolates, ESBL-producing E. coli (ESBL-Ec) was the most predominant species and was the focus of the molecular investigation of this study. A subset of 366 ESBL-Ec isolates (Additional file 2) were selected for short-read sequencing (MiSeq, Illumina Inc., USA). To resolve the complete genomes of ESBL-Ec ST131, long-read sequencing was performed on nine isolates. Sequencing criteria, methods, and read processing are detailed in Additional file 1.

Genotypic analysis

E. coli phylogroups were determined in silico by ClermonTyper [13]. Multilocus sequence types (MLST), antibiotic resistance genes, and virulence genes including fimH, were identified using the BacPipe pipeline [14]. Virulence genes were classified into functional groups and virulence profiles compared between cohorts (Additional file 1). Point mutations in chromosomal gyrAB and parCE genes conferring quinolone resistance were identified using PointFinder [15]. IS elements and genomic islands were predicted by ISEScan [16] and IslandViewer 4.0 [17], respectively. Integrons were identified using IntegronFinder v2.0 [18] and the Integrall database [19].

Phylogenetic analysis

Phylogenetic analysis was performed on all sequenced ESBL-Ec to identify the population structures of ESBL-Ec at the local hospital. In addition, the phylogenetic structure of the local ST131 ESBL-Ec population was analysed in a global context involving 1733 publicly available ST131 genomes (Additional file 1).

Intra-and inter-patient microevolution analysis

To study the dynamics of ESBL-Ec within and between patients, all patients carrying ESBL-Ec of the same STs were included. Mash v2.2.2 [20] was used to identify reference genomes. Mobile genetic elements (MGEs) and phage regions, respectively identified by MGEFinder [21] and Phaster [22], were masked from the reference genomes. The pairwise SNP distance matrix was generated using the CFSAN SNP pipeline v2.2.0 [23]. A cut-off of 10 SNPs [24] was used to delineate related isolates of the same strain.

Patients were linked to a putative clonal transmission cluster if their isolates belonged to the same STs, exhibited no more than 10 SNPs distance, and displayed strong epidemiological links, demonstrated by overlapping hospitalisation durations and wards of stay between patients in a cluster.

Statistical analysis

The incidence density of ESBL-Ec acquisition was determined as the number of new cases for 1000 patient-days. Differences between cohorts were assessed by chi-squared or Fisher’s exact test for qualitative data. For quantitative data (age, BMI, hospitalisation length, and length of antibiotic treatments), pairwise comparisons between cohorts were calculated using Wilcoxon rank sum test and Bonferroni correction was used to account for multiple comparisons. Variables associated with the outcomes at a P value of < 0.25 in the univariate logistic regressions were included in the multivariate logistic regression as predictors. Statistical tests were performed in R v.4.0.5 with a significance level of 0.05.

Results

Patient populations and incidence of ESBL-E nosocomial acquisition

From September 2010 to June 2013, 3824 patients were admitted to the recruiting wards. Of these, 3703 patients were included in the study and screened for ESBL-E. A total of 456 patients (456/3703, 12.3%) were rectally colonised with ESBL-E at admission and were classified into the ESBL positive-at-admission (PA-ESBL) cohort (Fig. 1). The remaining patients (3247/3703, 87.7%) were negative for ESBL-E screening at admission and belonged to the ESBL negative-at-admission (NA-ESBL) cohort.

Of the 2268 NA-ESBL patients with follow-up samples, who clocked a total of 30,147 hospitalisation days, 10.6% (n = 240) acquired ESBL-E and became HA-ESBL patients, resulting in an incidence density rate of ESBL-E nosocomial acquisition of 7.96 cases/1000 patient-days (95% CI 6.99 to 9.03). Available data on 227/240 HA-ESBL patients showed an average of 13 (median: 9, range 1–87) hospitalisation days prior to acquisition. The remaining 2028 NA-ESBL patients (89.4%) were consistently negative for ESBL-E throughout all screenings and were categorised as the “ESBL-free” cohort (Fig. 1). Detailed comparisons between PA- and NA-ESBL cohorts as well as between HA-ESBL and ESBL-free cohorts are shown in Additional file 3.

Compared to HA-ESBL patients who did not receive antibiotics (5.62 cases/1000 patient-days, 95% CI 4.35–7.15, n = 66, 27.5%), incidence density rate of ESBL-E nosocomial acquisition was higher for HA-ESBL patients who received antibiotics (9.45 cases/1000 patient-days, 95% CI 8.10–10.97, n = 174, 72.5%, P < 0.001), either before (10.12 cases/1000 patient-days, 95% CI 7.70–13.05) or after admission to the current hospital (9.15 cases/1000 patient-days, 95% CI 7.55–10.98).

In comparison to the ESBL-free cohort, ESBL-E rectal colonisation, both upon (PA-ESBL) and post (HA-ESBL) admission, was more prevalent in patients who were older than 80 years (P = 0.019), had a history of being ESBL carriers prior to admission (P < 0.001), recently hospitalised within the past 1 year (P < 0.05), received domiciliary assistance (P = 0.021), arrival from other care facilities (P < 0.05), and received antibiotic treatment (P = 0.004, Table 1). ESBL-E colonisation was also associated with the more frequent development of ESBL-E infection (P < 0.001), and the prevalence of ESBL-E infections was higher among PA-ESBL than HA-ESBL patients (8.6% versus 2.9%, P = 0.004). Antibiotic exposure during hospitalisation was more prevalent among HA-ESBL than the ESBL-free patients (72.5% versus 49.4%, P < 0.001, Additional file 3). At discharge, information of ESBL positivity was available for 522 (75%) of the 696 patients who were colonised. The loss of colonisation was higher among patients who admitted to the hospital with the colonisation (PA-ESBL patients, 121/304, 39.8%) than patients who acquired it during hospitalisation (HA-ESBL patients, 38/218, 17.4%, P < 0.0001).

Table 1.

Demographic characteristics of ESBL-E (PA-ESBL and HA-ESBL patients) and ESBL-free cohorts

		ESBL cohort	ESBL-free cohort	Univariate analysis			Multivariate analysis
		n = 696 patients	n = 2028 patients	P-value	Odds ratio	95% CI	P-value	Odds ratio	95% CI
Demography
Female	n (%)	286 (41.1%)	805 (39.7%)	0.552	0.94	0.75 − 1.16
Age (years)	median (range)	60 (19 − 99)	57 (18 − 101)	0.632	1.00	0.99 − 1.01
Age > 80 (n = 691 ESBL and 2006 ESBL-free patients)	n (%)	88 (12.7%)	137 (6.8%)	0.039	1.58	1.02 − 2.42	0.019	1.55	1.07 − 2.22
BMI (kg/m2)	median (range)	24.04 (14.2 − 65.8)	24 (13.7 − 61.7)	0.423	1.01	0.99 − 1.03
Comorbidities
Diabetes	n (%)	115 (16.5%)	311 (15.3%)	0.369	0.87	0.64 − 1.17
Malignancy	n (%)	210 (30.2%)	577 (28.5%)	0.778	1.04	0.81 − 1.32
Immunodeficiency	n (%)	230 (33%)	628 (31%)	0.679	1.05	0.83 − 1.34
Liver diseases	n (%)	117 (16.8%)	379 (18.7%)	0.059	0.76	0.57 − 1.01	0.054	0.76	0.57 − 1
Chronic renal disease	n (%)	96 (13.8%)	234 (11.5%)	0.252	0.80	0.55 − 1.16
Chronic lung disease	n (%)	85 (12.2%)	137 (6.8%)	0.402	1.31	0.7 − 2.46
Chronic skin lesion	n (%)	16 (2.3%)	43 (2.1%)	0.230	0.67	0.34 − 1.26	0.240	0.68	0.35 − 1.26
Respiratory illness	n (%)	89 (12.8%)	149 (7.3%)	0.995	1.00	0.54 − 1.83
Medical procedures
Hospitalisation length (days)	median (range)	10 (0 − 374)	8 (0 − 407)	0.026	1.01	1 − 1.01	0.025	1.01	1 − 1.01
ICU stay	n (%)	30 (4.3%)	41 (2%)	0.220	1.51	0.77 − 2.92	0.147	1.59	0.84 − 2.98
Surgery	n (%)	193 (27.7%)	627 (30.9%)	0.318	0.88	0.67 − 1.13
Urinary catheter	n (%)	144 (20.7%)	267 (13.2%)	0.441	1.14	0.81 − 1.6
Central venous catheter	n (%)	124 (17.8%)	256 (12.6%)	0.181	0.79	0.56 − 1.11	0.315	0.85	0.62 − 1.16
Antibiotic treatment	n (%)	401 (57.6%)	1001 (49.4%)	0.004	1.49	1.14 − 1.95	0.002	1.45	1.15 − 1.84
Length of antibiotic exposure (days)*	median (range)	10 (1 − 96)	8 (1 − 195)	0.904	1.00	0.99 − 1.01
Patient conditions during hospitalisation
Patient bedridden	n (%)	72 (10.3%)	107 (5.3%)	0.677	1.09	0.72 − 1.65
Patient with diarrhoea	n (%)	28 (4%)	83 (4.1%)	0.659	0.90	0.55 − 1.43
Patient with open wound	n (%)	47 (6.8%)	114 (5.6%)	0.932	0.98	0.65 − 1.47
Patient with ESBL-E infections	n (%)	46 (6.6%)	22 (1.1%)	< 0.001	4.17	2.32 − 7.63	< 0.001	4.10	2.3 − 7.47
Patient with known ESBL-positive roommate	n (%)	25 (3.6%)	50 (2.5%)	0.307	1.33	0.76 − 2.26
Patient known as previous ESBL carrier	n (%)	28 (4%)	18 (0.9%)	< 0.001	3.65	1.9 − 7.1	< 0.001	3.92	2.07 − 7.53
Patients admitted from (n= 540 ESBL and 1532 ESBL-free patients)
Home	n (%)	400 (74.1%)	1238 (80.8%)	0.092	1.42	0.95 − 2.15	0.108	1.38	0.94 − 2.08
Acute care	n (%)	63 (11.7%)	97 (6.3%)	0.001	2.48	1.49 − 4.17	0.001	2.44	1.48 − 4.07
Extended care facility	n (%)	28 (5.2%)	29 (1.9%)	0.002	2.94	1.49 − 5.84	0.001	3.04	1.55 − 5.97
Institution	n (%)	10 (1.9%)	9 (0.6%)	0.018	3.41	1.22 − 9.64	0.014	3.52	1.28 − 9.79
Previous hospitalisation
Between 3 months and one year	n (%)	233 (33.5%)	527 (26%)	< 0.001	2.00	1.51 − 2.65	< 0.001	2.00	1.53 − 2.62
Between 1 and 3 months	n (%)	118 (17%)	291 (14.3%)	0.017	1.46	1.07 − 1.99	0.016	1.44	1.07 − 1.93
Previous hospitalisation in the past month	n (%)	100 (14.4%)	219 (10.8%)	0.010	1.68	1.13 − 2.49	0.013	1.55	1.09 − 2.19
Medical procedures prior to admission
Surgery within 12 months	n (%)	84 (12.1%)	210 (10.4%)	0.280	0.81	0.55 − 1.18
Invasive devices within 1 month	n (%)	188 (27%)	389 (19.2%)	0.869	0.97	0.72 − 1.32
Dialysis within 12 months	n (%)	26 (3.7%)	51 (2.5%)	0.053	1.89	0.98 − 3.59	0.151	1.49	0.85 − 2.56
Patient conditions prior to admission
Domiciliary assistance	n (%)	72 (10.3%)	98 (4.8%)	0.066	1.51	0.97 − 2.33	0.021	1.63	1.07 − 2.45
Patient admitted on antibiotics	n (%)	99 (14.2%)	242 (11.9%)	0.213	0.80	0.56 − 1.13	0.222	0.81	0.58 − 1.13

Significant P-values are in bold

^* Calculated for patients who received antibiotic therapies during hospitalisation at the current hospital

ESBL-Ec was the predominant ESBL-E isolated from both PA- and HA-ESBL cohorts

From the total of 696 patients who were colonised with ESBL-E (456 PA- and 240 HA-ESBL patients), 1144 ESBL-E isolates were isolated and ESBL-Ec was predominant, accounting for 71.9% (822/1144) of ESBL-E isolates (Fig. 1). Of these, 780 ESBL-Ec (72.4%) were colonizing isolates collected from rectal screening of 536/696 (77.0%) patients with no significant difference in prevalence between patients from the PA-ESBL (361/456, 79.2%) and HA-ESBL cohort (175/240, 72.9%, P = 0.072, Fig. 1). The remaining 42 ESBL-Ec were infecting isolates collected from thirty patients (25 PA- and 5 HA-ESBL, Fig. 1), of which, 12 (40.0%) were diagnosed with an infection upon admission to the current hospital, while 11 (36.7%) developed hospital-onset infections. Data on infection onsets for the remaining 7 patients (23.3%) was unavailable. Six types of infection were documented, all classified as extraintestinal infections, with urinary tract infection (UTI) being the most common (Additional file 4: Table S1).

The most prominent ESBL non-E. coli species colonizing patients in both PA- and HA-ESBL cohorts included Klebsiella pneumoniae and Proteus spp., isolated from 88 (12.6%) and 33 patients (4.7%), respectively.

Population structure of ESBL-producing E. coli reveals high genomic diversity in both PA- and HA-ESBL cohorts

We sequenced ESBL-Ec (366/822, 44.5%) isolates from 201 patients (99 HA- and 102 PA-ESBL patients) who were either colonised (n = 171) or colonised and infected (n = 30) (Fig. 1). We observed a high clonal diversity with 52 STs belonging to 8 phylogroups, of which, phylogroup B2, notorious for causing extraintestinal infections, was the most dominant, represented by more than half of the isolates (n = 196, 53.6%, Fig. 2A). Among the 52 STs identified, 15, including the most prevalent ones, were shared between both the HA-ESBL and PA-ESBL cohorts. In addition to these shared STs, 20 STs were unique to the HA-ESBL cohort, and 17 STs were unique to the PA-ESBL cohort, each represented by no more than 7 isolates (Fig. 2B). Five MLSTs, present in both cohorts with comparable prevalence, were identified in more than 10 isolates: ST131 (n = 178, 48.6%, phylogroup B2), ST10 (n = 25, 6.8%, phylogroup A), ST405 (n = 21, 5.7%, phylogroup D), ST648 (n = 19, 5.2%, phylogroup F), ST410 (n = 11, 3.0%, phylogroup C, Fig. 2C). The remaining 47 STs comprised 112 isolates (30.6%, range 1–9, average: 2 isolates). ST131 was the most dominant ST with no significant difference between the PA-ESBL cohort (119/230, 51.7%) and the HA-ESBL cohort (59/136, 43.4%, P = 0.151, Fig. 2C).

Fig. 2.

Population structure of 366 ESBL-Ec isolates. A The maximum likelihood phylogeny of 366 sequenced ESBL-Ec from HA- (n = 136) and PA-ESBL (n = 230) cohorts. In total, 8 phylogroups (ring 1) and 52 MLSTs (ring 2) were isolated from both PA- and HA-ESBL cohorts (ring 3). The first isolated isolates (n = 215) from 201 patients are denoted with black dots in ring 4 while infecting isolates are represented by red stars in ring 5. The wards and years of isolation are shown in rings 6 and 7, respectively. Bootstraps are shown by the thickness of branches. B ST types identified exclusively in each cohort and commonly in both HA- and PA-ESBL cohorts. Numbers of isolates belonging to each ST are indicated in the parentheses. C Prevalence and distribution of the only, the first and subsequently isolated isolates across sequence types identified from 366 sequenced ESBL-Ec from PA-ESBL and HA-ESBL cohorts. The five most dominant STs are shown separately while the other 47 STs are grouped into Others. In both cohorts, from patients where at least two isolates were isolated at different time points, the prevalence of sequence types was comparably distributed between the first-isolated and subsequently isolated isolate populations

The maximum likelihood tree of ESBL-Ec showed that isolates from the HA- and PA-ESBL cohorts were interspersed in the phylogeny without monophyletic clustering of more than 5 isolates. This structure suggested the absence of specific hospital-associated lineages (Fig. 2A).

Similar to the overall population, the cross-sectional analysis of the first ESBL-Ec isolates isolated from patients in both PA- and HA-ESBL cohorts showed high clonal diversity. Among these 215 first isolated isolates, we observed 8 phylogroups and 45 STs, with ST131 being dominant (Additional file 3).

ESBL-Ec from HA and PA-ESBL cohorts presented similar antibiotic resistance and virulence gene profiles

Overall, 62.6% (229/366) of all sequenced ESBL-Ec were phenotypically multidrug resistant without significant difference between the two cohorts: 60.0% in the PA-ESBL cohort (138/230 isolates from 75 patients) and 66.9% in the HA-ESBL cohort (91/136 isolates from 68 patients, P = 0.227). Moreover, the isolates from these two cohorts showed similar antibiotic resistance gene and virulence gene profiles (Additional file 3).

Ninety percent of the sequenced ESBL-Ec (333/366 isolates) harboured bla_CTX-M genes, while other less frequent ESBL genes were identified in the remaining isolates (Additional file 3). Except for bla_CTX-M-14 that was more prevalent in PA-ESBL (10%, 24/230) than in HA-ESBL (4%, 5/136, P = 0.021), the prevalence and distribution among STs of other ESBL genes were not significantly different between the two cohorts (Fig. 3). The association of ST131 and bla_CTX-M-15 was the most common combination (140/366, 38.2%, Fig. 3).

Fig. 3.

Heat map showing the similar prevalence of ST-ESBL gene combination in PA- and HA-ESBL cohorts. We identified eight bla_CTX-M genes belonging to three groups: CTX-M group 1 (bla_CTX-M-1, bla_CTX-M-15, bla_CTX-M-15-like_, bla_CTX-M-32, bla_CTX-M-189), group 2 (bla_CTX-M-2) and group 9 (bla_CTX-M-14, bla_CTX-M-27). bla_CTX-M-15 associated with ST131 was dominant in both cohorts with the prevalence of 38.3 and 36.0%, respectively. bla_CTX-M-15*: novel variant of bla_CTX-M-15 (c.602C > T, p.A201V) detected in two HA-ESBL-Ec and in one PA-ESBL-Ec

Local ST131 reflect the epidemic properties of global ST131

As ST131 was the most dominant ST identified among ESBL-Ec of both cohorts, we further investigated the population structure of this clone. The majority of the local ST131 ESBL-Ec belonged to subclade C2/H30-Rx (135/178 isolates, 75.8%) which is characterised by the presence of fimH30, chromosomal mutations on gyrA and parC genes conferring fluoroquinolone resistance, and mostly associated with bla_CTX-M-15 (132/135, 97.8%, Fig. 4). Long-read sequencing revealed that up to three bla_CTX-M-15 copies could be present in a genome and located not only on plasmids but also on chromosomes (Additional file 3). Compared to the non-ST131 ESBL-Ec, the ST131 ESBL-Ec exhibited higher prevalence of virulence genes encoding increased serum survival (iss, function: protection factor against phagocytosis [25]) found either on IncF plasmid or chromosomes, and chromosomal pathogenicity islands (PAIs) carrying IrgA homologue adhesin (iha, function: iron‐regulated‐gene‐homologue adhesin [25]) and secreted autotransporter toxin (sat, function: proteolytic toxin with cytotoxic effect [25]; Additional file 3).

Fig. 4.

Maximum likelihood phylogeny of 178 ST131 ESBL-Ec isolates. The isolates were isolated from HA- (n = 59 isolates from 45 patients) and PA -ESBL (n = 119 isolates from 62 patients) cohorts. Inside green lines represent connections between colonizing and infecting isolates from the same patients

The core-genome-based maximum likelihood phylogeny of the 178 local ST131 ESBL-Ec suggested no evidence of hospital-specific ST131 sub-lineages (Fig. 4), demonstrated by the absence of monophyletic clusters containing ST131 isolates isolated from more than 5 HA-ESBL patients.

In the context of global ST131, we observed that ST131 genomes from Italy, either isolated from the studied hospital (n = 178) or retrieved from the public database (n = 14), were distributed across the phylogeny. We observed five monophyletic branches (CL-I to CL-V, Fig. 5) with apparent clustering of Italian isolates with isolates from various European countries, including Denmark, France, Sweden, Germany, and the Netherlands, suggesting limited or no barriers to the spread across countries.

Fig. 5.

Maximum likelihood phylogeny of 178 local ST131 ESBL-Ec isolates within the global genetic context. The global genomes included 14 Italian ST131 genomes that were not isolated from the studied hospital and 1719 publicly available ST131 genomes. Different bla_CTX-M genes and ST131 clades are shown in rings 0 and 1, respectively. Year, continent, and country of isolation are shown in rings 2, 3, and 4, respectively. Italian ST131 genomes, including 178 genomes isolated from the studied hospital and 14 from Enterobase, are shown as green bars in ring 5, while ring 6 shows only ST131 genomes from the studied hospital and the cohorts from which they were isolated. Five clusters (CL-I–CL-V) which contained ST131 genomes from other countries are denoted on the outermost ring and shown in more detail on the side. Of these five clusters, two involved ST131 clade C1/H30-R carrying bla_CTX-M-27 (CL-I, II) and three involved ST131 clade C2/H30-Rx carrying bla_CTX-M-15 (CL-III, IV, V)

Patient-to-patient putative clonal transmissions occurred on the ward level in both cohorts

In the non-outbreak setting of our study, most of the isolates representing a clonal population from the same patient differed by no more than 10 SNPs (Fig. 6A). Therefore, we adhered to the recommended cut-off of 10 SNPs [24] to define putative clonal transmission clusters between different patients that carried the same STs.

Fig. 6.

Patient-to-patient putative clonal transmission in the studied hospital. A SNP distance between ESBL-producing E. coli isolates isolated within patients and between different patients. B Putative transmission clusters identified among ST131 and other STs by SNP distance ≤ 10 and supported by strong epidemiological links: patients stayed in the same wards (represented by the colours of patient ID) and with overlapping periods of hospitalisation durations represented by blue and pink bars for PA- and HA-ESBL patients, respectively

Supported by strong epidemiological links, we identified 21 putative direct transmission clusters involving 23.0% (84/366) ESBL-Ec isolated from 23.9% (48/201) patients (Fig. 6B). The prevalence of PA-ESBL (25/102, 24.5%) and of HA- (23/99, 23.2%) patients involved in these 21 clusters were not significantly different (P = 0.870). The median size among these 21 clusters was 2 patients and 4 isolates, with three patients (IT1327, IT927, IT931) each noticeably involved in two clusters. These 21 clusters involved 8 STs and 8 ESBL genes with ST131 and bla_CTX-M-15 being the ones forming the most clusters as well as the largest cluster (cluster 4: 6 patients with 9 isolates, Fig. 6B). Most of the clusters (n = 15) involved both HA- and PA-ESBL patients, except for 4 clusters of only PA-ESBL patients (clusters 6, 8, 9, 13) and two clusters of HA-ESBL patients (clusters 5, 19; Fig. 6B). The majority of the clusters were identified in general surgery and infectious diseases wards (8 clusters each). However, the largest cluster (cluster 4) involved six patients hospitalised in the general medicine ward.

Sustained colonisation and endogenous infections play important roles in shaping the population structure of ESBL-Ec

To study the intra-patient clonal dynamics of ESBL-Ec, we investigated 128 patients from whom more than one isolate were isolated and sequenced (average of 2 isolates per patient, range 2–6). Per patient, a maximum of three sequence types (STs) were identified; although, from the majority of these patients (94/128, 73.4%), only one single ST was recovered. Thirty-one (24.2%) and three patients (2.3%) each carried 2–5 isolates belonging to 2 STs and 3 STs, respectively. Pairwise SNP analysis was performed among pairs of ESBL-Ec isolates belonging to the same STs.

Intra- and inter-patient dynamics of ESBL-Ec over time in colonised patients

Among 112 patients who provided at least two colonizing isolates, continuous colonisation over time by the same clone was commonly observed in both cohorts (Fig. 7). Specifically, colonisation by related isolates was observed in 72.3% (81/112) of patients overall, with no significant difference between the PA- (75.6%, 65/86) and HA-ESBL cohorts (61.5%, 16/26, P = 0.211). Only 5.4% of patients acquired a new ESBL-Ec, mostly of a different ST type, via putative direct transmissions with significantly higher prevalence in HA- (19.2%, 5/26) than in PA-ESBL (1.2%, 1/86, P = 0.003) cohorts. We also observed eight complex cases wherein patients showed mixed colonisation by different STs with endogenous and nosocomial origins via putative transmissions.

Fig. 7.

Clonal dynamics between ESBL-Ec isolates throughout screenings. The clonal dynamics between ESBL-Ec isolated from A 99 PA-ESBL patients and B 29 HA-ESBL patients from whom more than one ESBL-Ec was isolated (solid arrows indicate chronological events from the isolation of the first to subsequent and to the last colonizing ESBL-Ec. The dash lines to infecting isolates indicate that these isolates were not necessarily isolated in this time order)

Intra- and inter-patient dynamics of colonizing and infecting ESBL-Ec

Upon pairwise analysis of 30 colonised and infected patients, it was commonly observed that the infecting isolates and the colonizing isolates from the same patients were of the same strains (i.e., carried less than 10 SNPs distance), suggesting endogenous infection rather than nosocomial acquisition from an exogenous source. In particular, endogenous infections by the same strain (Fig. 7) were recorded in 60% (18/30) of colonised and infected patients and commonly found in patients presenting with both admission-onset (n = 10/12) and hospital-onset (n = 8/11) infections. Infections caused by a new ESBL-Ec ST type were documented in two patients (one PA- and one HA-ESBL patients, 6.7%). The source of infections could not be identified in the remaining 10 patients.

Discussion

Our study, conducted in non-outbreak, non-ICU hospital wards in a high ESBL-E prevalence setting, offers a comprehensive understanding of ESBL-E epidemiology. We were able to assess the relative impact of patients’ gut carriage of ESBL-Ec at admission and nosocomially acquired ESBL-Ec on infection and transmission rates. Our data suggest that ESBL-Ec carriage is highly associated with endogenous infections in hospitalised patients. Direct ESBL-Ec transmission between patients was infrequent, concurring with other reports [26], documented in one-fifth of patients in 21 clusters. This observation aligns with our finding that 19.2% of HA-ESBL patients acquired new ESBL-Ec via direct patient-to-patient transmission, as verified by genomic analysis. However, the colonisation prevalence in PA-ESBL (12.3%) and incidence in HA-ESBL (10.6%) cohorts were comparable, indicating that carriage on admission and acquisition of ESBL-E in the hospital occurred at a similar rate. Even when considering ESBL-Ec colonisation only, the prevalence was 9.7% (361/3703) in PA- and, comparably, 7.7% (175/2268) in HA-ESBL cohorts. Given the high sensitivity of the Brilliance ESBL agar used for screening [27], the role of additional reservoirs and transmission vectors in the ESBL-Ec transmission chain that were not studied here cannot be discounted. These reservoirs may include selection and horizontal transfer of ESBL genes between non-E. coli and E. coli species within patients’ guts, as well as transmission of ESBL-Ec by healthcare workers, and visitors. Further, the persistence of E. coli on floors, mattresses, clothing fabrics, and plastics has been observed for up to a period of > 8 weeks [28] and could form a continuous seeding source of transmission if adequate inactivation procedures are not in place. Thus, indirect transmissions via the environment and healthcare equipment might also be important but understudied routes. Our study design also did not allow monitoring the mobility and contact of the studied patients; therefore, we could not construct a complete inter- and intra-ward transmission network and, consequently, could have underestimated the transmission rate.

In this setting, 12.3% of patients admitted to the hospital were ESBL-E carriers. This prevalence was higher than that of other European countries such as the Netherlands (5.7%) and France (8%), but lower than that of Turkey (39.8%) [5, 29, 30]. Meanwhile, in the background of applied contact precautions and cohorting whenever possible for detected ESBL-E carriers, a comparable prevalence (10.6%) of ‘newly’ detected colonisation during hospitalisation was observed, resulting in an incidence density rate of ESBL-E nosocomial acquisition of 7.96 cases/1000 patient-days. The acquisition prevalence in our study was double the pooled prevalence (5.65%) reported for European non-ICU settings [26]. As previously shown in our clinical study (NCT01208519) that focused on patients negative at the time of admission for ESBL-E carriage, antibiotic exposure was strongly associated with ESBL-E nosocomial acquisition [31]. In this study, we also showed that the ‘new’ colonisation, expectedly, was associated with patients receiving antibiotics. However, when stratifying data by individual antibiotic class, we did not observe any significant correlation between antibiotic class and the acquisition of ESBL-E colonisation (Additional file 4: Table S2). On the other hand, when this association was explored in a larger cohort across three hospitals, including the Italian hospital studied here, in the published data from Tacconelli E. et al. [31], cephalosporins and quinolones were among the antibiotics strongly associated with the acquisition of ESBL-GNB colonisation.

The results from our clinical study corroborate with the model prediction by Lewis et al., reinforcing the role of antibiotic usage and hospitalisation in driving the escalation of ESBL-mediated resistance [32]. However, unlike the study by Lewis et al., which included healthy subjects from the community who had limited healthcare contact, in the present study we investigated community-associated patients who were admitted to the hospital and directly influenced the hospital epidemiology of ESBL-Ec. Remarkably, we observe a high degree of similarity in clonal diversity, antibiotic resistance, and virulence gene profiles between the community (PA-ESBL) and hospital-associated (HA-ESBL) patients. We show instances of ESBL-Ec influx from the community leading to nosocomial transmission, although these were sporadic, small transmission clusters. Additionally, it could be argued that the PA-ESBL patients had previous contacts with other healthcare settings, i.e., admission to the current hospital following care at other care facilities and institutions (n = 79/364 PA-ESBL patients from whom this information was available). However, it is worth noting that this subset of patients accounted for only 21% of the PA-ESBL cohort in our study.

The population structure of ESBL-Ec at the institutional level was highly diverse, despite the fact that we, as per the limitation of the study, only collected morphologically unique colonies, which might have resulted in an underestimation of the diversity. In line with other studies [33, 34], we identified more than one and up to three STs per patient, although the occurrence of one ST per patient was most common. Of particular concern was the predominant circulation of the pandemic, multidrug-resistant ESBL-Ec ST131 which harbour a rich content of virulence genes, including iss, iha and sat, which are known to be strongly associated with UTI and sepsis [35]. This clone is notorious for its association with increased infections and mortality [36] and was reported to be 3.2 (95% CI 2.0–5.0) times more pathogenic than non-ST131 [37]. Therefore, proactive intervention, such as implementing active screening for ESBL-E, might offer substantial benefits in managing the risk of infections. However, instead of universal screening, our data allows the development of an efficient, and potentially cost-effective, risk-stratified approach for application in high-volume non-ICU settings. The screening could target patients at higher risk of colonisation and/or infection as identified by our study: being 80 years of age or older, with a history of previous ESBL carriage, hospitalisation within the past 1 year, previous residence in an extended care facility, or receiving domiciliary assistance, and exposure to antibiotic treatment upon admission or during hospitalisation. This approach not only places less constraints on the screening staff and patients but, importantly, can be utilised to avoid empirical use of last-line antibiotics such as carbapenems, a common practice in settings where ESBL-E are endemic.

Conclusions

In conclusion, we provide a comprehensive understanding of the factors linked to ESBL-E colonisation and the trajectories of ESBL-Ec transmission in a non-ICU, non-outbreak setting. The genotyping analysis highlighted the importance of the endogenous reservoir of ESBL-E as the main driver of infections in hospitalised patients whether colonised at time of admission or acquired during current hospitalization. These results are important to further support the need for wide implementation of antibiotic stewardship programmes to reduce the antibiotic pressure on hospitalised patients, and therefore, the incidence of ESBL-E infections.

Authors' contributions

Conceptualization: S.M.-K., E.T., Y.C., H.G. ─ Data collection: B.P.G., G.D.A., E.T. ─ Sequencing: M.N.N., X.B.B., C.L. ─ Data analysis: M.N.N., M.B., X.B.B., S.K.S., S.M.-K. ─ Writing: M.N.N., Y.G., S.K.S., S.M.-K. ─ Review: B.P.G., M.B., X.B.B., Q.L., S.V.P., H.G., S.K.S., E.T. ─ All authors read and approved the final manuscript.

Data availability

The sequencing data of this study is available in the NCBI repository under the BioProject number PRJNA1082372. The dataset supporting the conclusions of this article is included within the article and Additional file 1.docx, Additional file 2.xlsx.

Declarations

Ethics approval and consent to participate

The study received approval from the Catholic University Ethics Commission in Rome, Italy (protocol P/291/CE/2010 approved on 06 April 2010). Written informed consents to participate were obtained from all included patients. This research conformed to the principles of the Helsinki Declaration.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.