Introduction

Bacteraemia represents a major cause of death in industrialised countries, with large increases in incidence and mortality being seen over the past 20 years [1, 2]. European data indicate that the burden of bacterial bloodstream infection (BSI) has been increasing for five major bacterial species (Staphylococcus aureus, Escherichia coli, Streptococcus pneumoniae, Enterococcus faecium and Enterococcus faecalis) during EARSS surveillance in 27 European countries from 2002 to 2008 [3]. Notably, E. coli was the most frequent causative pathogen in all age groups, representing 47 % of the global reports. BSIs due E. coli experienced an 8.7 % annual increase from 2002 to 2008 [3]. In addition, rates of BSIs caused by antimicrobial-resistant bacteria are also increasing worldwide. Worrisome, these resistant infections increase the total burden of disease rather than replace infections caused by more susceptible bacteria [4, 5]. Antibiotic therapy for E. coli bacteraemia usually relies on third-generation cephalosporins, but E. coli resistance to these drugs increased up to 82 % in a recent report from some European areas in 2014 (http://www.who.int/drugresistance/documents/surveillancereport/en).

Escherichia coli has a chromosomal AmpC (bla c-AmpC) expressed at low basal levels that does not confer resistance to β-lactams. Its overexpression, however, may lead to resistance to most β-lactams except fourth-generation cephalosporins and carbapenems [6]. This resistance may also be due to the acquisition of other AmpC genes (bla ac-AmpC), from different genera of various microorganisms by horizontal transfer [7]. The prevalence of acquired AmpC β-lactamase producing E. coli isolates has been increasing [8, 9]. Remarkably, as in extended-spectrum β-lactamases (ESBLs), plasmids carrying genes for AmpC β-lactamases often carry multiple other antimicrobial resistance determinants [6, 8, 10]. Carbapenems are frequently the only therapeutic option for E. coli overproducing AmpC. The effect of an AmpC β-lactamase in addition to porin mutations of the outer membrane can reduce susceptibility to carbapenems [11–14].

Knowledge of clinical features and outcomes of AmpC overproducing E. coli bloodstream infections (AmpC-Ec-BSI) is limited, and infections caused by AmpC overproducing E. coli are an increasing therapeutic challenge. On the other hand, it is crucial to identify risk factors for AmpC-Ec-BSI in order to appropriately design empirical therapies. The aim of the present study was to describe the epidemiology and clinical features of AmpC-Ec-BSI as compared with cephalosporin-susceptible E. coli bloodstream infections (Cph-S-Ec-BSI).

Methods

A case–control study was conducted in three acute care hospitals [Hospital Universitari Mútua de Terrassa (HUMT), Hospital de la Santa Creu i Sant Pau (HSCiSP) and Consorci Sanitari de Terrassa (CST)] and their corresponding primary care centres. Cases were recorded prospectively from June 2010 to November 2011, covering an area with 1,300,000 inhabitants and included all E. coli clinical isolates. Controls were matched 1:1 by sex and age. Data on the whole cohort have been published elsewhere [10]. In the present study, we included all adult patients with bacteraemia due to E. coli with reduced susceptibility to third-generation cephalosporins.

Variables and definitions

Cases were patients with bla AmpC producing E. coli (including both bla ac-AmpC and bla c-AmpC: AmpC-Ec-BSI) isolated from one or more sets of blood cultures, in the clinical microbiology laboratories. In patients with E. coli isolated in more than one blood culture less than 1 month apart, they were considered to be the same episode, and only the first isolate was included for analysis.

Controls were patients with bacteraemia due to a third-generation-cephalosporin-susceptible E. coli (3GC-S-Ec-BSI). The study was approved by the Ethics Committee of the participating hospitals.

Clinical data were prospectively collected [patient demographics, comorbidities (Charlson score) [15], severity (defined as the presence of septic shock and Pitt score), antimicrobial use within the previous 3 months, urinary tract infections in the previous 12 months, urinary and biliary tract abnormalities, indwelling external devices (mechanical ventilation, tracheotomy, urinary catheter, nasogastric tube, nephrostomy, external biliary drainage, endobiliary prosthesis), manipulation of the urinary, biliary, respiratory and/or gastrointestinal tracts within the previous 4 weeks, source of infection and antimicrobial susceptibility pattern]. Healthcare-related BSI was defined as a hospital-acquired or healthcare-associated infection. Hospital-acquired infection was defined as an infection acquired during hospital care that was not present or incubating at admission (infections occurring 48 h after admission were considered nosocomial) or in a patient discharged from hospital in the previous 14 days. Healthcare-associated infection was diagnosed if the patient fulfilled at least one of the following criteria: (i) resided in a nursing home or long-term care facility in the 30 days before the episode; (ii) hospitalised in an acute care hospital for ≥48 h, 90 days before the episode; (iii) attended a hospital or haemodialysis clinic or received intravenous therapy, 30 days before the episode; and/or (iv) received intravenous therapy, wound care, enteral nutrition or healthcare at home, 30 days before the episode. Otherwise, the infection was considered as community-acquired, according to the definitions by Friedman et al. [16].

The antimicrobial treatment regimen administered was recorded, including the agent or agents administered and the duration of treatment. Empirical therapy was considered to be appropriate if an active antimicrobial agent determined by in vitro susceptibility testing was administered at the usual recommended dose for AmpC-Ec-BSI or 3GC-S-Ec-BSI. Isolates were considered multidrug-resistant when they were resistant to three or more antimicrobial classes. The clinical outcome was evaluated as a cure rate at the 14th day of antibiotic and mortality until the 30th day.

Microbiological studies

Each centre selected E. coli isolates that met the following criteria: strains with resistance or exhibited reduced susceptibility (according to CLSI breakpoints and standard methodology [17]) to amoxicillin/clavulanate (<18 mm, >8/4 mg/L) and cefotaxime (<26 mm, >1 mg/L), or ceftazidime (<21 mm, >4 mg/L) or aztreonam (<21 mm, >4 mg/L). Strains with ESBLs were discarded, except those that were resistant to cefoxitin (<18 mm, >8 mg/L) or amoxicillin/clavulanate.

In each hospital, microbiological identification and susceptibility studies were determined by using Vitek [HUMT (bioMérieux Vitek, Hazelwood, MO, USA)], Microscan [CST (Dade Behring MicroScan, West Sacramento, CA, USA)] or the disc diffusion method (HSCiSP) [17, 18].

Phenotypic detection of acquired bla acAmpC (ac-AmpC) was performed by using the cefotetan/cefotetan–cloxacillin Etest (bioMérieux), if deemed appropriate. Visual examination of the antibiogram was carried out to detect colonies in the vicinity of the edge of the haloes of inhibition of cefotaxime, ceftazidime and aztreonam [19, 20].

Molecular detection of bla ac-AmpC was done by using the multiplex PCR described by Pérez-Pérez and Hanson [21]. Depending on the outcome of the multiplex PCR, additional primers were used to characterise the entire gene, as previously described [10]. Overproduction of bla c-AmpC was studied by quantitative reverse transcription PCR (qRT-PCR). Alterations of the bla c-AmpC promoter was analysed by PCR and, subsequently, sequenced, as previously described [21].

The clonal relationship of isolates was determined by pulsed-field gel electrophoresis (PFGE) [10]; phylogenetic data were obtained by multilocus sequence typing (MLST) using the Institut Pasteur tools (http://bigsdb.web.pasteur.fr/ecoli/ecoli.html).

Statistical analysis

To describe the epidemiology and clinical features of AmpC-EC-BSI as compared with cephalosporin-susceptible E. coli, the Mann–Whitney U test was used to compare continuous variables and the χ2 test or Fisher’s exact test was used to compare categorical variables. The SPSS (version 15.0) software package was used for these analyses.

Results

Among the 841 E. coli BSI episodes diagnosed during the study period, 17 (2 %) strains fulfilled the inclusion criteria and were tested by PCR and qRT-PCR. All 17 carried bla ac-AmpC or bla c-AmpC. Eleven AmpC-Ec-BSI (64.7 %) harboured ac-AmpC enzymes [81.1 % bla CMYt (six CMY2 and one CMY7) and 18.1 % bla DHAt (all DHA1)]. The remaining six isolates were AmpC overproducers. Two isolates presented a promoter with an identical mutation profile (−88; −82; −42; −18; −1; +58; +81), resulting in a promoter displacement. Two isolates presented a modified spacer region, showing a mutation at the −28 position and an insertion between positions −15 and −14. The remaining two isolates presented different mutations in the attenuator region [22].

No clonal relationship was found between isolates. Four cases were analysed by MLST, two of them carrying bla c-AmpC, being ST477 and ST494, and two carrying bla ac-AmpC, which were ST39 and ST48. Among all 17 isolates, eight belonged to less virulent phylogroups (seven to group A and one to group B1) and nine to more virulent phylogroups (four to group B2 and five to group D). Six out of eleven (54.6 %) bla ac-AmpC producing E. coli corresponded to phylogenetic group A and among bla c-AmpC, three belonged to group B, two to group D and only one to group A. No differences were found among bla CMYt and bla DHAt producing isolates in terms of phylogeny.

The demographic characteristics, epidemiological features and comorbidities are shown in Table 1. Among the whole cohort, the mean age was 65.6 years and 71 % were men. Cases were more frequently hospital-acquired than controls (p = 0.007).

Table 1 Demographics and comorbidities
Full size table

Table 2 summarises AmpC-Ec-BSI risk factors and clinical features. Indwelling external devices, presence of abnormalities in the urinary or biliary tracts, and recent manipulation within the last 4 weeks prior to the episode were identified more frequently among cases as compared with controls. Almost all cases presented at least one recognised risk factor. Cases had longer length of stay (LOS) and higher severity than controls but showed similar cure rates and mortality. The sources of bacteraemia are also shown in Table 2. The biliary tract was the most frequent source of bacteraemia among cases, followed by the urinary tract.

Table 2 Risk factors and clinical features
Full size table

Previous antimicrobial exposure and empirical therapy is summarised in Table 3. Previous antimicrobial exposure in the last 3 months was more frequent among cases and showed a trend towards statistical significance (p = 0.06). No specific antimicrobial class was found to be associated with the risk of AmpC-Ec-BSI.

Table 3 Previous antimicrobial exposure and empirical treatment
Full size table

Antimicrobial resistance patterns are shown in Table 4. All cases were susceptible to carbapenems, fosfomycin, amikacin and tobramycin. Higher rates of multi-resistance were detected among E. coli producing plasmid-mediated AmpC. Rates of inappropriate empirical therapy (IET) were higher among AmpC-EC-BSI.

Table 4 Antimicrobial resistance
Full size table

Discussion

The present cohort shows a low but not negligible prevalence of AmpC-Ec-BSI in a Spanish geographical area. The previously reported prevalence ranged from 1.5 to 4 % in different studies conducted in the USA [23], Europe [10, 24, 25] and Asia [26–28].

Remarkably, one-fifth of the analysed BSI due to AmpC producing E. coli were community-onset cases. This represents a therapeutic challenge, as empirical therapy for infections where E. coli should be the pathogenic agent must consider the presence of this antimicrobial resistance mechanism, depending on the local prevalence.

AmpC was mainly plasmid-mediated. A decade ago, this fact was considered a potentially epidemic threat. However, we have shown that, in our area, the prevalence of this resistance mechanism is low and stable. CMY2 was the most common ac-AmpC enzyme, as has been previously reported [6, 22, 29].

The epidemiological features of patients with BSI due to E. coli producing AmpC were somehow similar to those found among patients harbouring ESBL-producing E. coli [26, 30–32]. Cases presented significantly higher proportions of indwelling external devices, of abnormalities in the urinary or biliary tracts, recent manipulation of the urinary, biliary, respiratory and/or gastric tracts, and a trend towards a higher antimicrobial exposure. These findings suggest that bacteraemia due to AmpC-Ec occurs more frequently in patients with some comorbidities. Nevertheless, mortality was similar among cases and controls. Some authors have hypothesised that previous antibiotic use would select for these probably less virulent multidrug-resistant strains [33], which, hence, may cause invasive infections in predisposed patients [34, 35].

AmpC E. coli isolates included in the present cohort were mostly from phylogenetic groups D and A. Phylogenetic group D has been consistently recognised by many authors [36–39] as the most prevalent type together with group B1 among E. coli producing AmpC. On the other hand, in studies on BSIs caused by E. coli, phylogenetic group A has been found with increased frequency in nosocomial BSI, compromised hosts and in BSI caused by antibiotic-resistant isolates, a profile of patients that resembles that of the included cases in the present cohort [24, 40, 41]. Remarkably, the majority of E. coli isolates harbouring plasmid-mediated AmpC were identified as belonging to phylogenetic group A. Similar findings have been reported with ESBL-producing E. coli [42]. This low virulent phylogenetic group has been able to produce BSI in patients with intrinsic risk factors predisposing BSI and also in those with previous antibiotic exposure, which would have selected for AmpC-Ec because of their multidrug-resistant nature, regardless of their virulence profile.

The urinary and biliary tracts were the most common sites of infection [26, 30]. Remarkably, more than two-thirds of AmpC-Ec-BSI were of abdominal origin. On the contrary, urinary tract infections were more frequent among controls. We have no explanation for this finding. No epidemiological link was found among the included cases to suspect an outbreak in a surgical unit.

More than two-thirds of cases received an inadequate empirical therapy. This rate is similar to the data published by previous reports [34, 43]. This fact may be explained firstly because β-lactams are the first-line empirical therapy for urinary and abdominal infection and, secondly, because AmpC producing isolates were associated with higher rates of co-resistance to other antimicrobials. In Europe, some authors have reported the dissemination of plasmid-mediated quinolone resistance genes in Enterobacteriaceae, which can be associated with ESBLs or acquired AmpC β-lactamases. One mechanism for this association is incorporation on the same plasmid of qnr and genes for ESBLs or AmpC-type β-lactamases [44, 45]. In the era of carbapenem resistance, this fact represents an important therapeutic challenge, as carbapenems are frequently the only therapeutic option for E. coli producing AmpC.

The strengths of our study are the prospective and the geographically comprehensive design, including both community and hospital patients, the availability of detailed clinical information and the presence of a control group of cephalosporin-susceptible E. coli.

Our study also has several limitations. It is limited to a geographical area. Similar studies in other areas are needed to confirm our results. On the other hand, due the specific syndrome evaluated (BSIs due to AmpC producing E coli), a small sample size was recruited.

To summarise, the AmpC-Ec-BSI rates in our area are low and stable. AmpC-Ec-BSI affects mainly patients with intrinsic risk factors predisposing BSI and those with previous antibiotic exposure. Acquired enzymes are the main mechanisms implicated. Worrisome, a high proportion of patients receive IET. Despite the prevalence of AmpC-Ec still being low, in those patients with high suspicion or in those institutions with high prevalence, empiric therapy with agents active against AmpC strains may be warranted.