Multidrug-Resistant High-Risk Escherichia coli and Klebsiella pneumoniae Clonal Lineages Occur in Black-Headed Gulls from Two Conservation Islands in Germany

Multidrug-resistant (MDR) Enterobacterales, including extended-spectrum β-lactamase (ESBL)-producing Escherichia coli and Klebsiella pneumoniae, not only emerge in healthcare settings but also in other habitats, such as livestock and wildlife. The spread of these pathogens, which often combine resistance with high-level virulence, is a growing problem, as infections have become increasingly difficult to treat. Here, we investigated the occurrence of ESBL-producing E. coli and K. pneumoniae in fecal samples from two black-headed gull colonies breeding on two nature conservation islands in Western Pomerania, Germany. In addition to cloacal samples from adult birds (n = 211) and their nestlings (n = 99) during the 2021 breeding season, collective fecal samples (n = 29) were obtained. All samples were screened for ESBL producers, which were then subjected to whole-genome sequencing. We found a total of 12 ESBL-producing E. coli and K. pneumoniae consisting of 11 E. coli and 1 K. pneumoniae, and including the international high-risk E. coli sequence types (ST)131, ST38, and ST58. Eight of the investigated strains had a MDR genotype and carried a large repertoire of virulence-associated genes, including the pap operon, which is important for urinary tract infections. In addition, we identified many genes associated with adherence, biofilm formation, iron uptake, and toxin production. Finally, our analysis revealed the close phylogenetic relationship of ST38 strains with genomes originating from human sources, underlining their zoonotic and pathogenic character. This study highlights the importance of the One Health approach, and thus the interdependence between human and animal health and their surrounding environment.


Introduction
Antibiotics such as the β-lactams revolutionized medicine, as infections that had previously been fatal became easily treatable. Nevertheless, nearly 100 years after the discovery of penicillin, reports of severe infections caused by multidrug-resistant (MDR) bacteria have accumulated [1]. For example, the high prevalence of extended-spectrum β-lactamase (ESBL)-producing Enterobacterales (e.g., Escherichia coli and Klebsiella pneumoniae) is a major public health issue [2,3] and also a veterinary and environmental concern [4,5]. Livestock, in particular, represent an important reservoir for MDR pathogens, which can then be transmitted to humans and other animals, including wildlife, via direct contact, sewage,

The Presence of ESBL-Producing E. coli and K. pneumoniae in Two Black-Headed Gull Colonies
Sampling in both gull colonies took place during the breeding season (May-June 2021) on the islands of Böhmke and Riether Werder in Western Pomerania [27] (Figure 1). bacteria have accumulated [1]. For example, the high prevalence of extended-spectrum β-lactamase (ESBL)-producing Enterobacterales (e.g., Escherichia coli and Klebsiella pneumoniae) is a major public health issue [2,3] and also a veterinary and environmental concern [4,5]. Livestock, in particular, represent an important reservoir for MDR pathogens, which can then be transmitted to humans and other animals, including wildlife, via direct contact, sewage, manure, and the food chain [6][7][8]. It is therefore not surprising that we frequently find MDR E. coli and K. pneumoniae in various wildlife populations [9], including birds [10]. These Gram-negative bacteria have already been described in black-headed gull (Chroicocephalus ridibundus) colonies from Kalmar (Sweden) [11], Novi Sad (Serbia) [12], and different locations in the Czech Republic [13].
Here, we investigated the occurrence of ESBL-producing E. coli and K. pneumoniae in two remote black-headed gull colonies in Western Pomerania, Germany. Through wholegenome sequencing (WGS) and in-depth bioinformatics, this study aimed not only to determine (i) the prevalence of ESBL-producing E. coli and K. pneumoniae in the two gull colonies but also to characterize their (ii) phylogenetic background, as well as (iii) the resistance and virulence genes they carried.

Resistance Profiles of the Isolated Strains
WGS analysis using the AMRFinderPlus database [30] revealed that bla CTX-M-15 (n = 7) and bla CTX-M-14 (n = 2) were the predominant ESBL genes and that five different CTX-M family members were present in total ( Table 1). Note that the two ST38 strain as well as the ST58 strain carried different β-lactamase alleles, potentially indicating that they were unrelated. We then explored the context of the CTX-M-encoding genes further, as they are mostly carried by plasmids [31], but chromosomal integration has also been previously reported [32]. Interestingly, it appears that 4 of the 12 strains (PBIO3538, PBIO3541, PBIO3584, and PBIO3707) carried the ESBL gene chromosomally, with three of these strains belonging to the international high-risk clonal lineages ST38 (n = 2) and ST58 (n = 1), which is consistent with previous findings [33].
To determine whether the genotype matched the phenotype, we then performed phenotypic antimicrobial susceptibility testing (AST) against ciprofloxacin (quinolones), gentamicin (aminoglycosides), tetracycline (tetracyclines), and colistin (polymyxins). In summary, the results of the phenotypic AST were mostly consistent with the genotype ( Figure  2). Although PBIO3707 did not carry a typical mcr gene, it was phenotypically resistant to colistin (MIC, 4 µg/mL), similar to PBIO3583. We thus analyzed the genes related to lipopolysaccharide production, because it is known that modification of the respective enzymes can lead to colistin resistance [24]. Multiple sequence alignment of the protein sequences of PBIO3707 compared with the laboratory E. coli K-12 strain MG1655 (susceptible to colistin) revealed amino acid substitutions in PhoQ (Val386Leu) and EptA (Val336Met and Glu547Lys) that likely contributed to colistin resistance [37,38]. In addition, five strains (PBIO3678, PBIO3688, PBIO3691, PBIO3707, and PBIO3708) possessed heavy metal resistance genes, mainly associated with mercury resistance (mer; 4/5). The K. pneumoniae strain PBIO3691 demonstrated the largest set of heavy metal resistance genes, including resistance to arsenic (ars), copper (pco), mercury, and silver (sil). To determine whether the genotype matched the phenotype, we then performed phenotypic antimicrobial susceptibility testing (AST) against ciprofloxacin (quinolones), gentamicin (aminoglycosides), tetracycline (tetracyclines), and colistin (polymyxins). In summary, the results of the phenotypic AST were mostly consistent with the genotype (Figure 2). Although PBIO3707 did not carry a typical mcr gene, it was phenotypically resistant to colistin (MIC, 4 µg/mL), similar to PBIO3583. We thus analyzed the genes related to lipopolysaccharide production, because it is known that modification of the respective enzymes can lead to colistin resistance [24]. Multiple sequence alignment of the protein sequences of PBIO3707 compared with the laboratory E. coli K-12 strain MG1655 (susceptible to colistin) revealed amino acid substitutions in PhoQ (Val386Leu) and EptA (Val336Met and Glu547Lys) that likely contributed to colistin resistance [37,38]. In addition, five strains (PBIO3678, PBIO3688, PBIO3691, PBIO3707, and PBIO3708) possessed heavy metal resistance genes, mainly associated with mercury resistance (mer; 4/5). The K. pneumoniae strain PBIO3691 demonstrated the largest set of heavy metal resistance genes, including resistance to arsenic (ars), copper (pco), mercury, and silver (sil).

Virulence Profiles of the Isolated Strains
As the combination of MDR and high levels of bacterial virulence seems to be a prominent feature of international high-risk clonal lineages [33,39,40], we examined the genes associated with E. coli and K. pneumoniae virulence using the virulence factor database [41] ( Figure 3).

Virulence Profiles of the Isolated Strains
As the combination of MDR and high levels of bacterial virulence seems to be a prominent feature of international high-risk clonal lineages [33,39,40], we examined the genes associated with E. coli and K. pneumoniae virulence using the virulence factor database [41] ( Figure 3).  Although PBIO3541 and PBIO3687 belong to the well-known ExPEC lineages ST38 and ST131, they carried enterotoxins typically found in obligately pathogenic E. coli that are uncommon in ExPEC strains. Here, PBIO3708 (ST58) showed the highest number of VAGs (n = 174 genes), followed by the two ST38 strains PBIO3541 (n = 165 genes) and PBIO3584 (n = 156 genes). In contrast, the K. pneumoniae strain PBIO3691 had a Kleborate virulence Antibiotics 2022, 11, 1357 6 of 14 score of 0 because it lacked genes for yersiniabactin, colibactin, and aerobactin [43]. It carried a total of 102 VAGs, revealing the lowest level in our sample set.

Phylogenetic Relationship
Finally, to elucidate the phylogenetic relationships from a global perspective and to investigate the zoonotic potential of two of our most resistant and virulent strains, we constructed a phylogeny of 30 closely related, publicly available ST38 genomes in comparison with PBIO3541 and PBIO3584 (Figure 4). the P fimbriae adhesin cluster (pap operon; 50.0%, 6/12), the Dr family of adhesins (8.3%, 1/12), and Type I fimbriae (fim; 100.0%, 12/12) enable adhesion to uroepithelial cells and are thus often associated with strains that cause urinary tract infections [42].
Although PBIO3541 and PBIO3687 belong to the well-known ExPEC lineages ST38 and ST131, they carried enterotoxins typically found in obligately pathogenic E. coli that are uncommon in ExPEC strains. Here, PBIO3708 (ST58) showed the highest number of VAGs (n = 174 genes), followed by the two ST38 strains PBIO3541 (n = 165 genes) and PBIO3584 (n = 156 genes). In contrast, the K. pneumoniae strain PBIO3691 had a Kleborate virulence score of 0 because it lacked genes for yersiniabactin, colibactin, and aerobactin [43]. It carried a total of 102 VAGs, revealing the lowest level in our sample set.

Phylogenetic Relationship
Finally, to elucidate the phylogenetic relationships from a global perspective and to investigate the zoonotic potential of two of our most resistant and virulent strains, we constructed a phylogeny of 30 closely related, publicly available ST38 genomes in comparison with PBIO3541 and PBIO3584 (Figure 4). with publicly available genomes (n = 30). The genomes included here were selected on the basis of the most shared k-mers using a Mash-based approach. The phylogenetic tree was constructed using a maximum likelihood-based approach and based on a core single nucleotide polymorphism alignment, with PBIO3541 as the reference. The tree was midpoint-rooted and the circumference of the circles on the branches indicates bootstrapping support of ≥50%. Labels indicate the accession numbers (except for PBIO3541 and PBIO3584, which are highlighted in bold). In addition, the three-letter country code indicates the sample's origin (https://www.iso.org/obp/ui/#search, accessed on 19 September 2022). Different colors indicate the collection date and source. Further information includes the encoded ESBL genes and either their chromosomal or plasmid location. In the case of an unclear assignation, both the chromosome and plasmid locations are indicated (n = 6). NA, not available (i.e., no metadata provided); ST, sequence type. The phylogeny showed two major clades with several subclades, including the two ST38 strains and publicly available genomes originating from humans, domestic and wild animals, food, and the environment. PBIO3541 was most closely related to a strain from a sewage treatment plant in Japan (DRR199218). In addition, we found related genomes from humans (SRR19561363, SRR9924593). Interestingly, SRR19561363 originated from a bloodstream infection of an Australian patient, pointing towards its clinical significance.
PBIO3584, as part of the second major clade, showed a close phylogenetic relationship to a strain derived from a fecal sample from a healthy cow (DRR102958 [44]). In addition to this apparent association with animal husbandry, PBIO3584 clustered with strains originating from hospitalized patients from different geographic locations (SRR9113461, SRR11348017 [45]) but not with PBIO3541, again indicating our strains' zoonotic potential and the non-clonality of PBIO3584 and PBIO3541.

Discussion
Here, we isolated 12 ESBL-producing E. coli and K. pneumoniae strains from blackheaded gull colonies breeding in two nature reserves in Western Pomerania in 2021. Their distribution is interesting, as 11 of the 12 strains originated from Böhmke Island and only one was from Riether Werder Island. Due to the shorter distance to the mainland and tourist hotspots (e.g., Neppermin), the birds from Böhmke Island apparently not only feed from the surrounding sea but also on remains from tourist sites, which might imply a higher risk of acquiring clinically relevant strains through closer contact with humans. In general, nature areas near human settlements are more affected by anthropogenic impacts, such as contaminated food/landfills, agriculture, and polluted surface waters that subsequently serve as the gulls' habitat [46]. However, there is also evidence that individual birds move between the two locations [27], contributing to the occurrence of MDR bacteria in both colonies. Nevertheless, this is likely to be of minor importance, given the uneven distribution and unrelatedness of the strains across the islands.
ESBL-producing Enterobacterales have been detected worldwide in wildlife populations, including wild birds. For example, in Greece, 3.3% of fecal samples from wild birds contained ESBL-producing E. coli [47]. In pristine areas of Brazil, ESBL-producing E. coli were isolated from 2.4% of cloacal swabs from various seabird species, including the international high-risk clonal lineage ST131 [48], while a Nigerian study found a 100% rate in fecal samples from herons [49]. In the United Kingdom, 50% of herring gulls and black-backed gulls carried MDR pathogens [50], and a Swedish study found a prevalence of 8-55% in different Alaskan colonies, which was dependent on the predominant gull species at the sampling site [51]. However, the emergence of ESBL producers is not limited to wild bird populations. In Mecklenburg-Western Pomerania, we previously reported a rate of 1.2% in wild boars and wild ruminants [9], with a phylogenetic background similar to that of human isolates [11]. This underlines the potential of wildlife as a reservoir and a potential distributor of zoonotic pathogens.
The occurrence of five clinically relevant representatives belonging to international highrisk clonal lineages among only 12 ESBL-producing strains is remarkable. ST131 has not only been reported as the major E. coli clone in the clinical setting [32] but is often also found in veterinary, food, and environmental settings [52,53]. While previous studies have frequently found MDR ST58 strains in farm animals and, to a lesser extent, in wild birds [54,55] and as contaminants of retail meat [21], ST38 has been detected in several animal species, including broilers, rats, and black-headed gulls [12,56], with the latter matching our results. ST744 and ST617 have also been previously detected in German birds of prey and in rooks wintering in Europe, fitting into the pattern of the locally settled wild birds [16,57]. Taken together, our results and those of others highlight the omnipresence of international high-risk clonal E. coli lineages in wildlife species, particularly in wild birds. Moreover, ST290 represents a phylogenetic background of MDR K. pneumoniae that is not only present in wildlife [58] but has previously led to lethal outbreaks in human settings [59,60]. Other K. pneumoniae STs have been detected in wild birds, such as ST37 in Canadian ravens and ST548 in Australian silver gulls [61,62]. Thus, the occurrence of ST290 seems to be new in wildlife; however, other resistant K. pneumoniae STs are a known phenomenon.
A major contributor to the global dissemination of bacterial pathogens is sewage water, which originates from private and public facilities, including hospitals, and which is insufficiently treated before entering surface waters such as rivers and seas [63]. In addition, contaminated manure on fertilized agricultural fields exacerbates this situation [64,65]. Spillover to environmental habitats, where the wildlife acquires MDR pathogens and migrating wild birds in particular carry them for long distances, happens frequently [66]. This leads to the global dissemination of MDR bacteria, even in the most remote regions, again supporting the necessity of the One Health approach. None of the major public, veterinary, and environmental health disciplines should be considered separately [67].
Another interesting finding in our study was the chromosomal location of the ESBL genes in the four strains. It is well known that mostly plasmids drive the emergence of cephalosporine resistance in Enterobacterales [31]. Note that both ST38 strains carried the bla CTX-M genes chromosomally, which is consistent with our previous results, where the ST38 E. coli strains also originated from wild birds [19]. Chromosomal integration of previous accessory features represents a stable "fixation" of resistance for the bacterium, as the gene can no longer be lost as easily through plasmid loss [68], potentially leading to long-term benefit. Apart from non-susceptibility to antibiotics, bacterial heavy metal resistance occurs increasingly frequently, mostly driven by anthropogenic pollution and enhanced use, for example, in livestock feed [69]. In the case of the co-localization of heavy metal and antibiotic resistance genes on mobile genetic elements, heavy metals may promote the selection of MDR E. coli and K. pneumoniae in the environment [70], potentially leading to their even more accelerated dissemination.
Finally, our genomic analysis revealed that the ESBL-producing strains carried multiple virulence factors, likely providing an advantage in addition to antimicrobial resistance. For example, the ability to adhere to surfaces and form biofilms is important to help bacteria resist external stressors such as antibiotic treatment, UV radiation, and low temperatures [71]. Note, however, that we did not perform phenotypic virulence assays, which leads to a limited understanding of the bacteria's pathogenicity in vivo. We will address these experiments in the future. Nevertheless, it is well known that successful international high-risk clonal lineages such as the identified ST38, ST58, and ST131 usually combine MDR with high-level phenotypic virulence [22]. This, plus the fact that we detected germane strains that were clinically relevant, suggests their zoonotic and pathogenic potential, at least for ST38 and likely also for the other clonal lineages.
In summary, here, the occurrence of multiple high-risk clonal lineages in wild birds demonstrates the importance of One Health as a paradigm approach that recognizes the interdependence of human, animal, and environmental health. Both humans and animals depend on sustainable ecosystems, where small changes impact the health of all, which is also true in a global context.

Sampling Strategy and Bacterial Isolation
During the 2021 breeding season (May-June), we collected 339 different samples from two black-headed gull colonies on Riether Werder and Böhmke Islands, both located in Western Pomerania, Germany (Figure 1). Individual cloacal swabs (Sarstedt, Nümbrecht, Germany) were taken from adults and newly hatched nestlings. At Böhmke Island, 100 cloacal swabs were taken from adult birds and 99 swabs from nestlings, while at Riether Werder, 111 swabs were obtained from adults and none from nestlings. In addition, collective fecal samples were collected by pooling 5-10 individual fecal samples in 50 mL polypropylene tubes (Sarstedt, Nümbrecht, Germany). We thus included 14 collective fecal samples from Böhmke Island and 15 from Riether Werder Island. All samples were screened for ESBL-producing E. coli and K. pneumoniae. For this purpose, some material from the cloacal swabs or the mixed collective fecal sample was added individually to 5 mL Antibiotics 2022, 11, 1357 9 of 14 of lysogeny broth (LB; Carl Roth, Karlsruhe, Germany) supplemented with 4 µg/mL cefotaxime (Cayman Chemical, Ann Arbor, MI, USA) and incubated overnight under shaking conditions (130 rpm) at 37 • C. One hundred microliters of the cultures was then plated on chromogenic agar plates (CHROMagar orientation, Mast Diagnostica, Reinfeld, Germany), spiked with cefotaxime. The plates were incubated overnight at 37 • C. According to the manufacturer's protocol, the E. coli colonies were pinkish-purple, while the K. pneumoniae colonies had a metallic blue color. Subsequently, species identity was confirmed with API ID strips (bioMérieux, Marcy-l'Étoile, France). All confirmed strains were stored at −80 • C in LB containing 20% (v/v) glycerol (anhydrous; Merck, Darmstadt, Germany). Fresh overnight cultures were streaked on LB agar plates before use.

Phenotypic Antimicrobial Susceptibility Testing
We performed phenotypic antimicrobial susceptibility testing to investigate the occurrence of additional clinically relevant resistance phenotypes. Briefly, a single colony was picked and streaked on chromogenic agar plates containing antibiotics at their clinical breakpoint concentrations according to the published breakpoints of the Clinical and Laboratory Standards Institute [72]. We used the antibiotics tetracycline (16 µg/mL; VWR International, Radnor, PA, USA), gentamicin (2 µg/mL; Carl Roth, Karlsruhe, Germany), and ciprofloxacin (1 µg/mL; VWR International, Radnor, PA, USA). The minimum inhibitory concentration of colistin was determined using the MICRONAUT MIC-Strip Colistin Assay (Merlin Diagnostika, Bornheim, Germany) according to the manufacturer's instructions.

Whole-Genome Sequencing
Total DNA was extracted using the MasterPure DNA Purification Kit for Blood, Version 2 (Lucigen, Middelton, WI, USA), according to the manufacturer's instructions. The isolated DNA was quantified fluorometrically using the Qubit 4 fluorometer and the corresponding dsDNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). DNA was shipped to the SeqCenter (Pittsburgh, PA, USA) and, following library preparation using the Illumina DNA Prep Kit and IDT 10 bp UDI indices (Illumina, San Diego, CA, USA), whole-genome sequencing was carried out on an Illumina NextSeq 2000, producing 2 × 151 bp reads. Demultiplexing, quality control, and adapter trimming were performed using bcl-convert Version 3.9.3 [73].
To determine whether the bla CTX-M genes were encoded chromosomally or in plasmids, contigs were compared with the NCBI Nucleotide Database using blastn Version 2.13.0. The top 50 hits (to different subjects, where available) were used to manually determine the origin of the gene. Queries with all subjects annotated as chromosomes were classified as having a chromosomal origin. Accordingly, queries with hits to subjects annotated as plasmids and to subjects annotated as chromosomes were classified as ambiguous.

Phylogeny
For the creation of a core single nucleotide polymorphism (SNP)-based phylogeny of closely related isolates, Enterobase [81] was searched for E. coli ST38 strains (n = 2910) and the available assemblies were downloaded on 5 July 2022. Mash Version 2.3 [85] was used to generate sketches (sketch size, 1,000,000) of all downloaded genomes and the draft genomes of PBIO3541 and PBIO3584. Afterwards, the pairwise distance between PBIO3541 or PBIO3584 and the sketch archive was estimated. For a better overview, the 15 assemblies with the shortest distances, excluding the hits of PBIO3541 or PBIO3584 to itself (most shared k-mers), for which paired Illumina read data were available were selected. The data were downloaded from the public European Nucleotide Archive SRA FTP server (ftp://ftp.sra.ebi.ac.uk/vol1/, accessed on 25 July 2022). For these accessions, read trimming, genome assembly, and analyses were performed as described above. The trimmed reads were mapped against the draft genome of PBIO3541 using snippy Version 4.6.0 (https://github.com/tseemann/snippy, accessed on 27 July 2022) to create a whole-genome alignment of 32 sequences. The alignment was processed using Gubbins Version 3.2.1 [86] to filter out regions with SNPs that were likely the result of recombination. The alignment of sequences was processed using SNP-sites Version 2.5.1 [87] to retain only the alignment positions containing A, C, G, or T exclusively. A maximum likelihood tree was inferred with RAxML-NG Version 1.1.0 [88] using GTR+G by first parsing the alignment. The final alignment (containing 32 sequences and 1261 sites) was then processed by searching 500 parsimonious trees and 500 random starting trees, and performing 1000 bootstrap repeats. The best-scoring maximum likelihood tree was midpoint-rooted by iTOL Version 6.5.7 [89] and visualized with the bootstrap support values and metadata.