Dissemination of Escherichia coli with CTX-M Type ESBL between Humans and Yellow-Legged Gulls in the South of France

Extended Spectrum β-Lactamase (ESBL) producing Enterobacteriaceae started to appear in the 1980s, and have since emerged as some of the most significant hospital-acquired infections with Escherichia coli and Klebsiella being main players. More than 100 different ESBL types have been described, the most widespread being the CTX-M β-lactamase enzymes (bla CTX-M genes). This study focuses on the zoonotic dissemination of ESBL bacteria, mainly CTX-M type, in the southern coastal region of France. We found that the level of general antibiotic resistance in single randomly selected E. coli isolates from wild Yellow-legged Gulls in France was high. Nearly half the isolates (47,1%) carried resistance to one or more antibiotics (in a panel of six antibiotics), and resistance to tetracycline, ampicillin and streptomycin was most widespread. In an ESBL selective screen, 9,4% of the gulls carried ESBL producing bacteria and notably, 6% of the gulls carried bacteria harboring CTX-M-1 group of ESBL enzymes, a recently introduced and yet the most common clinical CTX-M group in France. Multi locus sequence type and phylogenetic group designations were established for the ESBL isolates, revealing that birds and humans share E. coli populations. Several ESBL producing E. coli isolated from birds were identical to or clustered with isolates with human origin. Hence, wild birds pick up E. coli of human origin, and with human resistance traits, and may accordingly also act as an environmental reservoir and melting pot of bacterial resistance with a potential to re-infect human populations.


Introduction
The levels of antibiotic resistance in Escherichia coli have reached a point where they pose severe clinical challenges to humans. E. coli is frequently used for studies of resistance status and development, and displays a relatively rapid transfer of resistance between strains, particularly if plasmid-borne [1,2]. It is also a major player in the spread of the recently emerged resistance to blactam antibiotics through the action of CTX-M b-lactamases, enzymes that have their origin in chromosomally located genes of other members of the Enterobacteriaceae family, i.e. Kluyvera sp. and possibly others [1]. Enterobacteriaceae that produce extended spectrum b-lactamases (ESBLs) started to appear in the 1980s, and have since emerged as some of the most significant hospitalacquired infections [1]. ESBL-producing bacteria, mainly E. coli and Klebsiella, are rapidly increasing among human isolates and today more than 100 different ESBL types have been described, with the most widespread type being the CTX-M b-lactamase enzymes (hereafter CTX-M) encoded by the bla CTX-M gene. The first CTX-M was characterized in 1989, but today CTX-M have a large geographical distribution [1,3]. In France reports of CTX-M emergence arose with the new century and several outbreaks were described in northern France 2002-2003. In southern France, other types of ESBL accounted for .90% of the ESBLs in a 2002-2003 survey. In this region, CTX-M started to appear later but have from 2004 been identified in several hospitals [3]. The CTX-M has also spread to other microorganisms, including community species such as Salmonella, Shigella and Vibrio cholera [3].
Migrating birds seem to act as transporters, or as reservoirs, of resistant bacteria and could therefore have an important epidemiological role in the dissemination of resistance, as well as being mirrors of the spectrum of pathogenic micro-organisms present in humans. A human-bird transfer of E. coli has been shown by the finding of a higher E. coli presence in wild birds with human association than in birds lacking human contacts [4]. Furthermore, although animals isolated from human activities previously displayed lower levels of resistant bacteria [5], more recent studies now show the presence of resistant E. coli in wild birds regardless of isolation. We saw e.g. that E. coli in Arctic birds sampled at the Siberian tundra carried antibiotic resistance, an indication of worldwide resistance dissemination [6]. In Maryland, US, migratory geese were carriers of resistant E. coli, as were Black-headed Gulls in the Czech Republic [7,8]. E. coli harboring b -lactamases of CTX-M, TEM and SHV type have previously been detected in birds of prey and seagulls in Portugal [9,10].
We investigated the presence of antibiotic resistance, with special focus on CTX-M ESBL bacteria, in Yellow-legged Gulls (Larus michahellis) in two populations in Camargue, southern France ( Figure 1). This species belongs to the Herring Gull-complex, a set of large size gulls including several species present in the northern hemisphere. Many of these gulls, including the Yellow-legged Gull, have increased in numbers and distribution due to their custom of living in proximity to humans. They are often opportunistic marine feeders, seeking their food (fish, bivalves, mollusks, eggs, birds etc) along the shoreline or offshore, but also readily utilizing the food sources provided by humans, especially garbage. The two Yellow-legged Gull populations are located in a region with a dense human population, and were chosen in order to take into account any variations in resistance patterns resulting from differences in the availability of human-associated food sources. The gulls breeding at Port Saint-Louis Carteau colony feed mainly at Marseille city dump, whereas the gulls at Aigues-Mortes colony feed mainly offshore ( Figure 1). Our results show that resistance levels of randomly selected E. coli (one strain from each sample) are high when measured against a panel of six antibiotics, and that wild birds in this region carry high levels of ESBL producing bacteria displaying genotypes highly related with French (human) clinical samples. Further, ESBL producing E. coli from these birds shared phylogenetic diversity with human E. coli, coherent with that these bacteria are established zoonotic agents.

Results
Antibiotic resistance-levels in E. coli from Yellow-legged Gulls In order to determine a general picture of resistance levels in Yellow-legged Gulls we tested one randomly selected E. coli isolate per sample, to a panel of six antibiotics. Nearly half (47.1%; 72/153 isolates) of the E. coli isolates were resistant to at least one antibiotic, and there was no difference in overall resistance rates between the two gull colonies (33/75 E. coli isolates from ''city dump'' colony and 39/78 isolates from ''off shore'' colony). Resistance to two or more antimicrobial agents was widespread, and found in approximately 1/3 of the isolates. The most widespread resistance phenotypes were those resistant to tetracycline, ampicillin and streptomycin. Resistances to the remaining compounds were less frequent ( Table 1). Four strains displayed resistance to cefadroxil (Table 1), an indication of potential ESBL carriage.

Selective screen for ESBL producing bacteria
ESBL producing bacteria were initially identified by phenotypic tests. 17 isolates (16 E. coli and one Enterobacter cloacae, all from separate samples) were found that exhibited disc diffusion synergy test and cefepime MIC concordant with ESBL production, as well as growth on cefpodoxime containing chromID TM ESBL plates. The presence of bla CTX-M , bla TEM and bla SHV was then determined by PCR. Ten isolates were confirmed to harbor CTX-M-1-group enzymes, and sequencing defined genotypes to 9 isolates (including the E. cloacae) with bla CTX-M-1 and one with bla CTX-M-15 (Table 2). bla TEM was found in combination with bla CTX-M in two isolates and with bla SHV in two isolates, and separately in five isolates. All isolates, containing any of the blagenotypes, displayed ESBL positive results in Etests with cefotaxime, ceftazidime and cefepime combined with clavulanic acid inhibition. Three of the four fecal samples containing cefadroxil resistant, randomly selected, E. coli also generated ESBL producing bacteria in the ESBL selective screen (E. coli with sample ID 63638, 63654, and 63725; Table 2).

Multi locus sequence typing and phylogenetic group determination of the ESBL producing isolates
Based on sequencing of 3423 nucleotides at seven housekeeping genes, the sixteen ESBL-producing E. coli were assigned to 13 different sequence types (STs) ( Table 3). Nine isolates were assigned to seven novel STs, i.e., allelic profiles that were not found in the multi locus sequence typing (MLST) database; STX1 (1 isolate), STX2 (2 isolates), STX3 (1 isolates), STX4 (1 isolate), STX5 (2 isolates), STX6 (1 isolate), and STX7 (1 isolate) ( Table 3). Seven isolates were assigned to six previously reported STs; single isolates to ST156, ST90, ST351 or ST746, respectively, and two isolates to ST681. The 13 STs representing the 16 ESBLproducing E. coli isolates were compared with 273 STs representing full global diversity of the species (Figure 2) [11]. It was apparent that the 16 ESBL isolates covered a broad range of genetic diversity. Consequently, a large fraction of the currently known E. coli genetic diversity was present among the ESBLproducing isolates and they were not of recent common ancestry. Specifically, only those isolates that are found in a clonal complex or share an ST are likely to have a recent common ancestry, i.e., isolates in STX2, in STX5, or in ST681.
According to the classification method by Clermont et al. [12], later modified by Escobar-Páramo et al. by addition of subgroups [13], seven isolates belonged to phylogenetic group A0, five to group D, one to group B1, two to group B2, and one to group A ( Table 3). The seven isolates classified as A0 were not closely related as determined by MLST, they were frequently found interleaved among isolates of other phylogroups and were widely dispersed in the minimum spanning tree (Figure 2). For other phylogenetic groups, there was some correspondence with the position of a corresponding ST in the minimum spanning tree, e.g. isolates of STX5 shared phylogroup D with other neighbor isolates and ST681 shared the B2 phylogroup with its neighbors (Figure 2). Phylogroup assignments within STs were consistent, isolates assigned to the same ST always shared phylogenetic group. The bla CTX-M harboring E. coli were widely dispersed out in the minimum spanning tree showing that these resistance genes are present across the full E. coli genetic diversity.

Discussion
In the Mediterranean region, including France, the usage of antibiotics in human and veterinary medicine is at a comparatively high level [14], and consequently we found a high general level of antibiotic resistance in E. coli from Yellow-legged Gulls in the Marseille region. We believe that the influence of antibiotics in the region has been such that wild bird populations with varied  feeding habitats carry organisms displaying comparable resistance levels. In fact, the resistance patterns are similar, with high level resistance especially to tetracycline, streptomycin and ampicillin. These are common resistance types in domesticated and wild animals in contact with human activities [2,9]. Of 17 (9.4%) ESBL type b-lactam resistant isolates, all 10 CTX-M type isolates (nine E. coli, one E. cloacae) belonged to the CTX-M-1 group, specifically nine of bla CTX-M-1 (including the E. cloacae) genotype and one bla CTX-M-15 . This is consistent with French clinical (human) isolates, where the CTX-M-1 group is also the major type [15], and both bla CTX-M-1 and bla CTX-M-15 genotypes have been isolated [1]. One isolate was E. cloacae, often a clinically relevant species as well [15,16]. Nine bla TEM and two bla SHV harboring isolates were present, all positive in phenotypic tests for ESBL. Two isolates included both bla CTX-M and bla TEM , and two both bla TEM and bla SHV .
We found that fecal samples from gull colonies contained a high prevalence of E. coli with antibiotic resistance traits and that ESBL producing isolates exhibited great genetic diversity. This suggests that dense bird colonies, such as the Yellow-legged Gulls in our study, might serve as melting pots for creation of new E. coli variants carrying antibiotic resistance, or alternatively, that these birds are remarkably effective samplers of genetically diverse ESBL-carrying E. coli from other sources, e.g. humans. The distribution pattern of ESBL producing isolates within the genetic structure of E. coli shows that not only a few sporadic imports of specific clones carrying antibiotic resistance has occurred. Regardless of where the ESBL-producing E. coli have evolved originally, the feeding habit of gulls creates opportunities for antibiotic resistance genes to be mixed with many different E. coli genetic backgrounds. We noted a lack of association between phylogenetic group and the allele based population structure analysis by MLST (Figure 2). Our findings are consistent with recent findings of other workers. Gordon et al. found that E. coli isolates that fail to produce any PCR amplicon at the three genes targeted in the triplex PCR assay (group A0 isolates [13]) are genetically heterogeneous [17]. Accordingly, our seven A0 isolates were genetically diverse illustrating that the A0 phylogenetic group assignment is a poor predictor of genetic relatedness.
Isolates assigned to the same STs or ST complex are likely to have a common recent ancestor. Our findings in bird samples of E. coli of STs that have been recovered from humans might therefore indicate transmission between man and birds. We found an ESBLproducing member of the ST23 clonal complex (Figure 2), which has previously been reported for several pathogenic E. coli isolates of human origin. Isolates of the ST23 clonal complex were reported from urinary tract infections in California, Toulouse in France, and from the northwest of England [18,19]. The isolate from England was, similar to our bird isolate, resistant to expanded-spectrum cephalosporins. Similarly, we recovered an ESBL producing E. coli of ST533, an ST previously reported for an isolate from a human urinary tract infection in Brazil and exhibiting bla tem-1 and bla CTX-M-2 genes. This data together with data from other studies support the fact that E. coli with or without antibiotic resistance traits are transmitted between humans and birds [20,21].
To summarize, we believe that a high general resistance level in Yellow-legged Gulls in the south of France is a result of long-time direct or indirect exposure to human activities in an area with relatively high antibiotic pressure. The presence of ESBL producing Enterobacteriaceae is high (9.4%) and in particular the proportion of the CTX-M-1 group (6%), which in this region is a very recently emerged and rapidly increasing ESBL type. The fact that ESBL producing bacteria are phylogenetically dispersed among E. coli isolates of human clinical and animal origin, underlines our hypothesis that these bacteria are both zoonotic and anthropogenic in their behavior, as are the resistance types found in them. Plasmidborne resistance, such as CTX-M type ESBL, is obviously rapidly disseminating since it is already present in high levels in this type of animal reservoir that the birds represent. Hence, our data together with other recent studies indicates that birds could act as important environmental bio-indicators, and reservoirs, of medically important pathogens and resistance genes, as well as a potential melting pot for the development of new resistance types.

Materials and Methods
Collecting of fecal samples 90 samples per colony were collected from juvenile non-fledged birds. Each sample consisted of a cotton swab that had been rotated in the cloacae of a bird to obtain fecal material which was then inoculated into bacterial freeze medium (Luria broth; BD, Sparks, USA, phosphate buffered saline containing 0.45% Nacitrate, 0.1% MgSO 4 , 1% (NH 4 ) 2 SO 4 , and 4,4% glycerol). After sampling in the field, the samples were transported to the laboratory and frozen at 270uC for later examination.

Bacterial isolation and antibiotic resistance testing
Each sample was plated on Juhlin 32 agar [22] for isolation of putative E. coli isolates. E. coli identity was confirmed by biochemical testing. Susceptibility of one isolate per sample (with 75/90 isolated from the ''city dump'' colony and 78/90 from the ''offshore'' colony) was tested against a set of antibiotic agents: tetracycline, ampicillin, streptomycin, chloramphenicol, nalidixic acid and cefadroxil. These antibiotics were selected to represent commonly used agents in E. coli infections in human and veterinary medicine. Susceptibility was determined by antibiotic disk diffusion on Iso-Sensitest agar (Oxoid, Basingstoke, UK) in accordance with the recommendations of the Swedish Reference Group for Antibiotics (SRGA) [22], using E. coli ATCC 25922 as a reference strain (as in all assessments).

Isolation of ESBL producing bacteria and phenotypic characterisation
For detection of ESBL producing bacteria, samples were plated onto on MacConkey agar (Oxoid, Basingstoke, UK) with two antibiotic discs containing cefotaxime (5 mg) and ceftazidime (10 mg), respectively. Isolates showing reduced susceptibility to one or both antibiotics were identified by biochemical testing. Phenotypic confirmation of ESBL-production was performed by disc diffusion synergy test according to SRGA [23].Cefepime MICs were determined using Etest (bioMérieux, Marcy l'Etoile, France) in isolates with a negative synergy test. ESBL production was also confirmed with ESBL Etest for cefotaxime, ceftazidime and cefepime including clavulanic acid (bioMérieux), and growth on cefpodoxime containing chromID TM ESBL plates (bioMérieux, Marcy l'Etoile, France).

Genetic characterization of ESBL producing bacteria
Isolates with ESBL-production were further analyzed for presence of CTX-M, TEM and SHV type ESBL enzyme genes (bla CTX-M ) using previously described methods [24,25]. The bla CTX-M gene was sequenced from CTX-M type positive isolates, using previously described primers [26] and GenScript BacReady PCR system (GenScript Corporation, Piscataway, NJ).

Multilocus sequence typing of ESBL producing E. coli isolates
We used the MLST scheme developed by Wirth et al. [11](http:// web.mpiib-berlin.mpg.de), utilizing seven housekeeping genes: adenylate cyclase (adk), fumarate hydratase (fumC), DNA gyrase (gyrB), isocitrate/isopropylmalate dehydrogenase (icd), malate dehydrogenase (mdh), adenylosuccinate dehydrogenase (purA) and ATP/GTP binding motif (recA). The PCR reactions were as follows; initial denaturation at 95uC for 2 min, 30 cycles of; denaturation at 95uC for 1 min; annealing at various temperatures for 1 min; elongation at 72uC for 2 min, and a final elongation at 72uC for 5 min. Reactions were performed using a T Gradient Thermocycler (Biometra GmbH, Goettinger, Germany). For sequencing reactions a commercially available sequencing kit was used (Big DyeH Terminator v3.1 cycle sequencing kit, Applied biosystem, Foster City, CA, USA) and analyzed in a 3730xl DNA Analyzer (Applied Biosystems, Foster City, CA, USA) according to manufacturers' instructions. Primers sequences and annealing temperatures can be found at http://web. mpiib-berlin.mpg.de. All sequence traces were imported, aligned, trimmed and quality controlled aided by functions of the software Bionumerics v.5.1 (Applied Maths NV, Sint-Martens-Latem, Belgium). Re-sequencing was performed if required and MLST allele designations were determined via the electronic MLST database at http://web.mpiib-berlin.mpg.de. Novel ST designations were provided by the curator of the database (Table 3).

Allele-based population genetic analysis of ESBL producing E. coli
The genetic relationships of ESBL producing E. coli were assessed by calculations based on MLST allelic profiles for isolates and comparison with existing profiles in the MLST database. We used the minimum spanning tree algorithm implemented in Bionumerics v.5.1 (Applied Maths NV) with priority rules set at first link genotypes that have maximum numbers of single-locus variants and then maximal numbers of single-locus variants and double-locus variants. Sequence types (STs) were assigned members of clonal complex if they shared alleles in six out of seven house-keeping genes. We choose ST1 to ST278 of the MLST database as a comparative dataset for analysis of isolates characterized in this study. ST1 to ST278 represent the known diversity of E. coli in terms of geography, pathogenicity, human and animal origins. Four STs of the MLST database were excluded from analysis since representing extreme outgroup isolates (ST102, ST125, ST133) or a lack of a phylogentic group designation (ST103) [11].

Phylogenic analysis of ESBL producing bacteria E. coli
Phylogenetic groups were determined using the triplex PCR method developed by Clermont et al. [12], and later modified by addition of subgroups by Escobar-Páramo et al [13], but with modified PCR conditions [27].