VIM‐1 carbapenemase‐producing Escherichia coli in gulls from southern France

Abstract Acquired carbapenemases currently pose one of the most worrying public health threats related to antimicrobial resistance. A NDM‐1‐producing Salmonella Corvallis was reported in 2013 in a wild raptor. Further research was needed to understand the role of wild birds in the transmission of bacteria resistant to carbapenems. Our aim was to investigate the presence of carbapenem‐resistant Escherichia coli in gulls from southern France. In 2012, we collected 158 cloacal swabs samples from two gull species: yellow‐legged gulls (Larus michahellis) that live in close contact with humans and slender‐billed gulls (Chroicocephalus genei) that feed at sea. We molecularly compared the carbapenem‐resistant bacteria we isolated through culture on selective media with the carbapenem‐susceptible strains sampled from both gull species and from stool samples of humans hospitalized in the study area. The genes coding for carbapenemases were tested by multiplex PCR. We isolated 22 carbapenem‐resistant E. coli strains from yellow‐legged gulls while none were isolated from slender‐billed gulls. All carbapenem‐resistant isolates were positive for bla VIM ‐1 gene. VIM‐1‐producing E. coli were closely related to carbapenem‐susceptible strains isolated from the two gull species but also to human strains. Our results are alarming enough to make it urgently necessary to determine the contamination source of the bacteria we identified. More generally, our work highlights the need to develop more bridges between studies focusing on wildlife and humans in order to improve our knowledge of resistant bacteria transmission routes.

This detection raised questions about the potential risk of the spread of resistance potentially associated with this wild reservoir.
Acquired carbapenemases currently pose one of the most worrying public health threats related to antibiotic resistance (Gupta, Limbago, Patel, & Kallen, 2011;. They confer resistance to carbapenems, but also to almost all β-lactams, the most widely used class of antibiotic, and are encoded by genetic elements that are transferable between bacteria (Krahn et al., 2016;Luca et al., 2017;Potron, Poirel, & Nordmann, 2014). The actual number of carbapenemase-producing bacterial isolates is rising, and the epidemiological status of these bacteria (sporadic versus local spread versus national endemicity) is progressively worsening worldwide (Glasner et al., 2013;WHO 2014).
Carbapenems represent the latest therapeutic innovation for β-lactams, but this innovation is old, the latest group of molecules having been approved for clinical use more than a decade ago. Yet they are currently our last effective defense against multiresistant Gramnegative bacteria (Woodford et al., 2014).
Our ability to limit the rise and spread of carbapenemase producers, which occur only at basal levels in many countries at present, should serve as a key performance indicator for the success or failure of the efforts that have been called for by international organizations and governments to reduce the impact of antibiotic resistance (Woodford et al., 2014). To meet this challenge, we need to investigate the role of any nonhuman reservoirs of carbapenemresistant bacteria, which could favor their further spread in human populations (Woolhouse, Ward, van Bunnik, & Farrar, 2015). To date, carbapenem-resistant bacteria have been isolated from water in some rivers and sewage plants as well as in a few pets and food animals (reviewed in Woodford et al., 2014). The resistant bacteria isolated from a black kite is so far the sole evidence of the presence of carbapenem-resistant bacteria in a wild species without any direct link with domestic animals or humans. The only other evidence of carbapenem resistance in wildlife is from a pig farm in Germany where a VIM-1 carbapenem-resistant Salmonella serovar Infantis was isolated in a mouse (Mus musculus) (Fischer et al., 2012). Verona integron-encoded metalloβ-lactamases (VIM) belong to class B carbapenemase and were first described in Italy in 1999 (Lauretti et al., 1999). Greece is now considered to be the epicenter of the spread of VIM-producing Enterobacteriaceae to other European countries where they have been subsequently detected including Spain, Italy and France (Canton et al., 2012;Mathlouthi, Al-Bayssari, Bakour, Rolain, & Chouchani, 2016). In the light of these data, further research is clearly needed to understand the potential role of some wild species in the spread of carbapenem-resistant bacteria. in a species that lives in close contact with humans following its recent colonization of urban habitats and that has subsequently experienced a strong demographic increase: the yellow-legged gull (YLG, Larus michahellis; Duhem, Roche, Vidal, & Tatoni, 2008). The focal YLG population was previously reported to carry high loads of extendedspectrum β-lactamase (ESBL)-producing E. coli (Bonnedahl et al., 2009). We also investigated the E. coli strains found in slender-billed gulls (SBG, Chroicocephalus genei) living in the same area. We chose to study both species since they share the same environment but their feeding habits differ. YLG are opportunistic, feeding on fresh fish, but, like the black kite, they also feed on refuse and carcasses, whereas SBG mainly feed on marine fishes (Flitti, Kabouche, Kayser, & Olioso, 2009). We investigated E. coli carried by chicks since, within the colonies we studied, they had no contact with humans and they could only be contaminated by bacteria brought by adults or already present in the colony. Thus, finding AMRB in those chicks would mean either that these bacteria have been transmitted from adults to chicks or that the environment (surrounding water or soil) is contaminated by them.
In France, as in most Western European countries, carbapenemaseproducing bacteria infections have so far been limited to hospital settings and represent only a few cases per year. Yet the number of those cases has significantly increased in recent years in many European

| Ethics statement
The study has been approved by the Scientific and Ethical Council

| Sampling, bacterial strains and antibiotic susceptibility testing
Cloacal swabs were collected from gull chicks, aged from 1 to 4 weeks, in two colonies. The yellow-legged gull colony was situated on an islet Twenty-five clinical isolates of carbapenem-susceptible nonpathogenic E. coli, recovered from human patient's stools in the same geographical area (Montpellier hospital) and during the study period, were used for phylogenetic comparison.

| Detection and identification of carbapenemresistant genes
The E. coli strains that were isolated from the media containing carbapenem were further analyzed to determine the mechanisms involved in resistance to carbapenems using a multiplex PCR ( Briefly, the presence of the most prevalent carbapenemase genes (including bla KPC , bla VIM , bla OXA-48 group, and bla IMP ) was assessed by multiplex PCR, as previously described (Dallenne et al., 2010) and bla NDM gene presence was investigated by PCR assay (Hornsey et al., 2011). Metalloβ-lactamase production was assessed by the Carbapenemase/Metalloβ-Lactamase Confirmation Identification Pack (Rosco Diagnostic Neo-Sensitabs ™ , Eurobio, Courtaboeuf, France).
All amplified PCR products were purified using the ExoSap purification kit (ExoSap-it, GE Healthcare, Piscataway, NJ, USA), and bidirectional sequencing was performed using the BigDye Terminator version 3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and an Applied Biosystems 3730 XL capillary sequencer. Each sequence was then compared with already known carbapenemase genes available in the NCBI database using the BLAST program. The detection of bla VIM-1 gene was confirmed by simplex PCR assay and sequencing using the following primers: VIM_F (5′-AGTGGTGAGTATCCGACAG-3′) and VIM_R (5′-TGCAACTTCATGTTATGCCG-3′).

| Molecular analyses
To determine the relatedness between the E. coli isolates sampled from YLG, SBG, and humans, we performed PCR phylotyping, discriminant single nucleotide polymorphism (SNP), multi locus sequence typing (MLST) and polymorphic variable number of tandem repeats loci (VNTR).
Multi locus sequence typing procedures were performed using Wirth et al. MLST scheme (Wirth et al., 2006

| Phylogenetic and statistical analyses
Exact Fisher tests were used to test for differences in E. coli phylogroup distributions among host populations.
The multiple alignments of all complete MLST sequences were conducted using ClustalW in BioEdit version 7.0.9.0. software. Maximum likelihood (ML) tree construction was based on the MLST sequences and the best-fitting ML model under the Akaike information criterion was GTR (general time reversible) + Γ (gamma distribution) for nucleotides as identified by ModelTest (Posada & Crandall, 1998). The most likely DNA tree and corresponding bootstrap support values were obtained by PhyML using Mega 5.0 software (Tamura et al., 2011) with nearest-neighbor interchange branch swapping and 100 bootstrap replicates. All positions containing gaps and missing data were eliminated.
Genetic polymorphism of VNTR data was measured in terms of allelic richness per locus and sample (Rs) (Mousadik & Petit, 1996), gene diversity was measured in terms of expected heterozygosity per locus and sample (Hs) using the unbiased estimator adapted to haploid data (Nei & Chesser, 1983), and the genotypic linkage disequilibrium was tested by the log-likelihood ratio G-statistic after Bonferroni's correction using FSTAT version 2.9.4 software (Goudet, 2003).
Multi locus sequence typing, single nucleotide polymorphism, and variable number of tandem repeats loci data were analyzed with BioNumerics software version 7.0 (Applied Maths, Sint-Martens-Latem, Belgium). Based on allelic profiles, the evolutionary relationship between isolates was assessed by a minimal spanning tree (MST) implemented in BioNumerics. The MST is a graphical tool that links the nodes by unique minimal paths in a given dataset: the total summed distance of all branches is minimized. The Prim's algorithm calculated a standard MST with single-and double-locus variance priority rules.

| Detection of carbapenem-resistant Escherichia coli
Twenty-two E. coli strains isolated from yellow-legged gulls (Table 1) were resistant to most of the β-lactam antibiotics, except aztreonam, and were metalloβ-lactamase producers. Those carbapenemaseproducing E. coli originated from 18 of the 93 YLG chicks on which we sampled cloacal swabs. Two birds carried two carbapenemaseproducing isolates and one gull carried three. All 22 carbapenemaseproducing E. coli isolates were positive for bla VIM-1 gene. Conversely, no carbapenem-resistant bacteria were isolated from the samples collected on 65 slender-billed gulls. Table 1 indicates the number of isolates sampled from each host group that were used for each type of analysis (Phylogroup, SNP, MLST and VNTR). A1 (81.8%). This proportion was significantly higher than that found in the susceptible isolates as a whole (p < .001 Fisher's exact test).

| Single nucleotide polymorphism
In the SNP analysis, 26 distinct haplotypes were detected among the 88 E. coli isolates studied (reference strains were not included, see  Table S1).
F I G U R E 1 Phylogenetic relationships among the 92 Escherichia coli isolates studied based on concatened MLST sequences. The circle tree was constructed using maximum likelihood methods. Bootstrap values greater than or equal to 60% are indicated at the nodes, and those relating to the five clusters containing carbapenem-resistant isolates are shown in blue. The E. coli strains were isolated from yellow-legged gulls (YLG), slender-billed gulls (SBG), and humans

| Variable number tandem repeat
The VNTR gave complete results for 79 isolates only (

| Phylogenetic analysis based on MLST, SNP, and VNTR
The minimum spanning tree presented in Figure 2 is based on genetic sequence similarity according to MLST (7 fragments of gene sequences shortened and aligned to the reference sequences in the MLST Databases at the University of Warwick (freely available at: http://mlst.warwick.ac.uk/mlst/) (i.e., 3,423 nucleotides in total)), SNP and VNTR analyses. The phylogenetic structure highlighted is similar to that underlined in Figure 1. Isolates of the same phylogroup tend to cluster together. Carbapenem-resistant isolates do not form a distinct cluster. Conversely, they are grouped in three clusters that also F I G U R E 2 Minimum spanning tree of 79 of the studied Escherichia coli strains based on MLST, SNPs and VNTR. The 13 strains for which part of the VNTR data was missing were excluded. The phylogroups are shown as ovals. Clonal complexes are indicated by symbols proportional in size to the number of strains within them. Black lines connecting strains indicate that they differ at (1) least by one VNTR (bold thick lines) to two VNTR and (2) seven markers (five MLST genes and two SNPs (the thinnest lines)) include carbapenem-susceptible strains isolated from the different host species (YLG, SBG, and humans). In addition, only one resistant isolate presents a unique sequence, the others are grouped in five clonal complexes including 2-6 strains.
In order to test for potential artifacts in the deep branches caused by VNTR which tend to have a very rapid dynamics, we led the same analysis without VNTR, i.e., MLST + SNP. The results obtained show the same network topography; the only minor difference concerns the B2 group position, but all strains of this group cluster together as observed in Figure 2 (see Supplementary Information- Figure S1).

| DISCUSSION
We highlighted the presence of VIM-1 carbapenem-resistant E. coli strains in yellow-legged gulls in southern France. Our results confirm that gulls represent a bird group that frequently carries antimicrobialresistant bacteria, as was previously shown in several studies (e.g., Čížek Our findings are all the more worrisome if we consider that gulls live in close contact with human populations since they feed on waste matter, notably in cities, and thus represent a bridge species for pathogens between wildlife and humans. Moreover, young yellow-legged gulls can fly large distances from their native colony to their wintering sites that include the whole of the Rhone Valley and the French Atlantic coast (Sadoul & Pin, 2009 (Crémet et al., 2012, Cuzon, Naas, Lesenne, Benhamou, & Nordmann, 2010INVS 2014), which underscores the rarity of carbapenemresistant infection originating in France. Furthermore, the resistance gene we detected (bla VIM-1 ) is uncommon in France, where it caused only 5% of the reported infections due to carbapenem-resistant enterobacteria, the OXA-48-and OXA-48-like genes being the most frequent in the country (74%) (INVS 2014). bla VIM-1 is an integron-borne metalloβ-lactamase gene which was first reported in Pseudomonas aeruginosa in Italy in 1996 (Lauretti et al., 1999). It encodes for a class B carbapenemase which also hydrolyzes all β-lactams except monobactams, and evades all β-lactamase inhibitors. VIM-1-bearing bacteria have been reported from clinical samples in Greece although they are beginning to spread in southwestern Europe, notably in Spain and Italy, while France seems, for now, to be less affected (Canton et al., 2012;Mathlouthi et al., 2016). We report here the second isolation worldwide of carbapenemresistant enterobacteria from wild birds and the first detection in gulls. In-depth molecular analysis of blaVIM genetic surroundings and features will be necessary in the future to fully understand the downstream impacts of our findings. Yet, here our aim was not to fully characterize the genetic background of carbapenem-resistant E. coli carried by wildlife but rather to warn over the potential role of wild birds as carbapenem-resistant bacteria carriers and spreaders. Our results are alarming enough to justify an urgent call for further studies.
They also make it a matter of urgency to determine the contamination source of the bacteria we identified. The next step could be to extend our work to several other YLG and SBG colonies, including some in which the two species are breeding in neighboring islets, as well as to repeat sampling other time to improve our knowledge of VIM-1 carrying bacteria circulation in French gull populations. Such extended sampling could also help in disentangling the respective roles of feeding habits and environmental pollution in carbapenem-resistant bacteria contamination. To complete this analysis, it would also be necessary to search for resistant bacteria within the environment, especially in water, since it has previously been shown that rivers and coastal areas can contain carbapenem-resistant bacteria even when those pathogens are rare in the surrounding human populations (Aubron, Poirel, Ash, & Nordmann, 2005;Montezzi et al., 2015). More generally, our work highlights the urgent need to develop more bridges between studies focusing on wildlife and humans in order to improve our knowledge of resistant bacteria dynamics. Such bridges will be a key factor in enabling us to efficiently face the challenge of antimicrobial resistance in the future.

ACKNOWLEDGMENTS
We sincerely thank the municipality of Port-Saint-Louis, the Communauté d'Agglomération Toulon Provence Méditerranée, Aurélien Audevard and Frédérique Gimond for allowing us to access the gull colonies. We also thank all the people who helped us in collecting the samples, especially the bird ringers who supervised bird handling: Thomas Blanchon and Yves Kayser, as well as Prof. Richard Bonnet who kindly provided the four phylogroup reference E. coli strains.

DATA ACCESSIBILITY
Molecular data will be archived in the following database: http:// enterobase.warwick.ac.uk/.