Genetic Diversity and Resistome Analysis of Campylobacter lari Isolated from Gulls in Croatia

Campylobacter lari is a thermotolerant bacterium that sporadically causes gastrointestinal diseases in humans and can be found in wildlife and the environment. C. lari is an understudied species, especially in wild birds such as gulls. Gulls are potentially good carriers of pathogens due to their opportunistic behavior and tendency to gather in large flocks. During winter and their breeding period, 1753 gulls were captured, and cloacal swabs were taken to be tested for the presence of C. lari. From isolated bacteria, the DNA was sequenced, and sequence types (ST) were determined. Sixty-four swabs were positive for C. lari, and from those, forty-three different STs were determined, of which thirty-one were newly described. The whole genome was sequenced for 43 random isolates, and the same isolates were tested for antimicrobial susceptibility using the broth microdilution method to compare them to WGS-derived antimicrobial-resistant isolates. All the tested strains were susceptible to erythromycin, gentamicin, and chloramphenicol, and all were resistant to ciprofloxacin. Resistance to ciprofloxacin was attributed to a gyrA_2 T86V mutation. Genes connected to possible beta-lactam resistance (blaOXA genes) were also detected.

Only 462 isolates of C. lari have been genotyped using MultiLocus Sequence Typing (MLST) and are available on the PubMLST database (https://pubmlst.org, accessed on 27 July 2023) [21]. Most of them (83%) were isolated from different species of wild birds (159) and shellfish (143) and humans (78). The rest of the isolates (17%) come from a variety of sources in the environment (fresh and marine water, soil, etc.).
Gulls are one of the most common birds in human surroundings. Although they are "known" as seabirds, many gull species are opportunistic and form huge flocks, while feeding on rubbish tips, even far inland. More than 10,000 birds from all over Europe can be seen in the winter feeding on a Zagreb rubbish tip [22,23]. Also, roosting and sleeping places are usually used by a large number of birds, and therefore, are a great opportunity for the circulation of different pathogens.
The aim of this study was to determine the prevalence of C. lari found in different gull species that occurin Croatia, its molecular characterization, and the antimicrobial susceptibility of isolates. Additionally, in this study, we aim to compare the phenotypic and genotypic resistances of C. lari strains. In Croatia, C. lari has only been studied in shellfish [24], and in general, it is still under-researched especially in wildlife; therefore, this is the first study conducted on gulls.

Results
A total of 1753 gulls from five species were captured in Croatia over a five-year period (2017-2022) in order to study the occurrence of C. lari.
There were sixty-four positive swabs for C. lari, mostly from YLG (54) and BHG (7), while two positive swabs were from Caspian Gulls, and one was from a Common Gull. The overall prevalence was 3.65%, with the highest isolation rate found for the Caspian Gulls (11.76%), followed by the YLG (4.98%) and Common Gulls (3.03%), while the lowest rate was for the BHG (1.13%).

Genotypic Determination of Antimicrobial Resistance (AMR)
The testing of ciprofloxacin, as one out of the six antimicrobial agents included in this study, showed genotypic and phenotypic resistance. A resistance to ciprofloxacin was detected in all the C. lari isolates carrying a gyrA_2 T86V mutation. The genes connected to Antibiotics 2023, 12, 1310 5 of 10 possible beta-lactam resistance (blaOXA genes) were often identified in the tested isolates; however, as far as we know, none of those are directly connected to a specific antimicrobial beta-lactam agent. Two different blaOXA genes were identified; blaOXA-493 gene was identified in thirty-three samples (76.7%), and blaOXA-518 was identified in three samples (7.0%).

Comparison between Phenotypic and Genotypic AMRs
The phenotypic and genotypic antimicrobial predictions were correlated in the case of ciprofloxacin, where all the tested strains showed phenotypic and genotypic AMRs. One strain showed a phenotypic resistance to tetracycline and ertapenem, but not a genotypic one. In total, one strain out of the forty-three tested (2.3%) on six different antimicrobial agents showed two discordances between the phenotypic and genotypic results.

Discussion
Most of the samples included in the present study belonged to two species of gulls (YLG and BHG), and there was a great discrepancy in the isolation rates of C. lari between these two species (4.98% in YLG and 1.13% in BHG). One of the main ecological differences between these two species is their breeding habitat. While YLG mostly breed in marine habitats (mostly on rocky islets and sea cliffs), BHG mostly use freshwater habitats (mostly islands in lakes or wetlands) [26]. It is known that C. lari can tolerate a much higher concentration of NaCl than other thermotolerant species of the same genus can [27,28]. This is in accordance with published data on the prevalence of C. lari in different gull species; it appears higher to be in marine species, such as 3% in Herring Gulls [29] and 3.30% in YLG [30], and it is lower in freshwater species, such as 1.02% in BHG [31].
The facts that only 462 isolates of C. lari (as opposed to more than 80,000 C. jejuni or more than 20,000 C. coli isolates) were listed in the PubMLST database and 330 different STs were described (including isolates from this study) highlight how understudied this organism is [21]. Also, having 31 newly described ST (compared to the 158 STs described before this study) out of 64 isolates supports this statement, but it also shows the great diversity of the isolates [21]. Only 20 isolates from this study belong to the previously described STs. Four of the isolates from this study share the same STs with only one isolate from humans, four share STs with the gull isolates, two share STs with the human and shellfish isolates, one is shared with shellfish, and one is shared with Brent Geese (Branta bernicla) [21]. Unfortunately, the species names of wild birds are not always provided in databases, and it is not possible to link them to specific (i.e., marine) environments. If we exclude them from being a part of the marine environment analysis of the PubMLST collection of C. lari STs (Figure 1), this supports the fact that C. lari is a more halophilic bacterium than other Campylobacter species can, as most of the described STs (64.1%) are from shellfish, gulls, and marine water.
The most prevalent resistant gene determined in this study was the quinolone resistance gyrA_2 T86V gene (100%). This Thr-Val 86 substitution in the GyrA protein has been previously reported only in the C. lari group [32][33][34], and as it cannot be acquired due to a single point mutation, it is considered that this is the sequence of an intrinsically resistant organism [32]. Two known β-lactam resistance genes (blaOXA variants) were also detected: blaOXA-493 gene was identified in thirty-three samples (76.7%), and blaOXA-518 was identified in three samples (7.0%). The presence of the blaOXA-493 gene as the predominant β-lactamase coding gene in C. lari has previously been described [34,35]. Both blaOXA-493 and gyrA_2T86A were detected in a total of 33 isolates (76.7%). The authors of earlier studies found that the combination of these two genes is exclusively found in the C. lari group [34]. Regarding the blaOXA-518 gene, it has been described earlier for C. coli [35] and in C. lari (GenBank: Accession No. NG_049792, National Center for Biotechnology Information (NCBI); online: https://www.ncbi.nlm.nih.gov/nuccore/NG_049792.1, accessed on 20 July 2023, unpublished), and it belongs to the OXA-493-like subfamily, with 99.6% Antibiotics 2023, 12, 1310 6 of 10 similarity. It differs from OXA-493 due to a Gly-Asp 72 substitution. However, functional information about the OXA-493-like subfamily is not available [36].
described STs. Four of the isolates from this study share the same STs with only one isolate from humans, four share STs with the gull isolates, two share STs with the human and shellfish isolates, one is shared with shellfish, and one is shared with Brent Geese (Branta bernicla) [21]. Unfortunately, the species names of wild birds are not always provided in databases, and it is not possible to link them to specific (i.e., marine) environments. If we exclude them from being a part of the marine environment analysis of the PubMLST collection of C. lari STs (Figure 1), this supports the fact that C. lari is a more halophilic bacterium than other Campylobacter species can, as most of the described STs (64.1%) are from shellfish, gulls, and marine water . Most of the STs are fitted in six groups. Nineteen isolates (belonging to nineteen different ST's) were not taken in this analysis as the source was either not specified (e.g., "soil", "environmental water") or was too scarce to affect analysis (e.g., dog (n = 3); horse (n = 1); bat (n = 1)). Most of the STs are fitted in six groups. Nineteen isolates (belonging to nineteen different ST's) were not taken in this analysis as the source was either not specified (e.g., "soil", "environmental water") or was too scarce to affect analysis (e.g., dog (n = 3); horse (n = 1); bat (n = 1)).
The overall correlation rate between the WGS-based genotypic prediction and phenotypic resistance is 95.5%.
The discordances in the results included one strain showing a phenotypic resistance to tetracycline and ertapenem. There are no known or thus-far documented mutations that were found in the investigated genome.
Until now, a genotypic resistance to fluoroquinolones in Campylobacter species has been explained by the presence of different gyrA gene mutations. Thus far, mutations that correlate with a genotypic resistance to fluoroquinolones also present different levels of resistance, making it possible for strains with a certain mutation to be phenotypically susceptible, and vice versa [37]. Moreover, detected resistant determinants do not always confer a resistant phenotype. These findings can be explained by the existence of multiple mechanisms of resistance to antibiotics, which include a decrease in outer membrane permeability and efflux systems, and detected modifications and mutations [38]. Li et al. (2017) pointed out, after conducting the integrated genomic and proteomic analysis of chloramphenicol resistance in C. jejuni, that this integrated approach is the key to understanding AMR in its fullness. Often, there are many mechanisms in action, not just mutations of certain genes [39]. In this study, again, WGS proved to be a good starting tool for comprehensive AMR characterization that is highly concordant with phenotypic AST, and also, a great tool for quick outbreak investigations. This is especially true in cases like these, where information about AMR in C. lari is extremely lacking.
At the beginning of this study, it was emphasized that C. lari is widespread across multiple hosts and environmental samples. The authors of previous research have also pointed out that the role of C. lari and other Campylobacter species, such as C. concisus, C. ureolyticus, and C. upsaliensis, in human and animal diseases is underappreciated. Because of this, these species are known as ''emerging Campylobacter species" [40].
The fact that C. lari is isolated from such a wide range of habitats and its resistance to the fluoroquinolones class of antimicrobials may classify it as a reservoir of resistant gene Antibiotics 2023, 12, 1310 7 of 10 material. Also, this supports the importance of this emerging underestimated species as a potential pathogen.

Sampling and Campylobacter Isolation
In the period from 22nd January 2017 to 31st May 2022, gulls were captured in Croatia in order to test their cloacal swabs for the presence of C. lari. During the breeding season, the gulls were caught in bird colonies using "walk in" traps set up on nests, and during the non-breeding season, they were caught with cannon net on rubbish tips. Cloacal swabs were taken and stored in Amies transport medium with charcoal (Merck, Darmstadt, Germany) to prolong the vitality of the microorganisms. After collection, the samples were transported in cool boxes at 4-8 • C and examined within a maximum of 48 h of sampling. The detection of Campylobacter spp. was conducted according to the standard EN ISO 10272-1 method [41], and species determination was performed using multiplex PCR [42].
Briefly, each swab was added to the liquid enrichment medium (Bolton broth; Oxoid, UK). After incubation in a microaerobic atmosphere (CampyGen; Oxoid, UK) at 37 • C for 4-6 h, and then at 41.5 • C for 44 h, a loopful (10 µL) of the enrichment cultures were streaked on the surface of each of the two selective plating mediums, modified Charcoal Cefoperozone Deoxycholate agar (mCCD agar; Oxoid, UK) and CampyFood ID agar (CFA; BioMerieux, France). Selective solid media were incubated at 41.5 • C in a microaerobic atmosphere and examined after 44 h to detect the presence of suspect Campylobacter colonies. All of the presumptive Campylobacter colonies were examined for morphology and motilityand sub-cultured on a Columbia agar (Oxoid, UK) with 5% sheep blood (Biognost, HR) for further confirmation (the microscopic verification of the characteristic morphology and motility, the detection of oxidase activity, and an aerobic growth test at 25 • C).

MLST
Using the available research funding, whole genomes were sequenced from 43 randomly selected isolates. Extracted DNA was prepared according to the instructions of MicrobesNG (Birmingham, UK) and sent there for sequencing. The sequencing was conducted using an Illumina platform with a 250 bp paired-end output. The results were obtained as raw trimmed reads and assembled as fasta files. Sequence types (ST) were determined using the WGS-MLST plug-in in BioNumerics 8.1.1 version (BioMerieux, Applied Maths, Sint-Martens-Latem, Belgium).
The other 21 isolates were sequenced according to Miller et al. [43]. PCR products were sequenced at Macrogen Europe (The Netherlands). Sequences were edited using BioEdit software (version 7.2.5.). STs were determined using the Campylobacter multilocus sequence typing website (https://pubmlst.org/campylobacter/, accessed on 27 July 2023) sited at the University of Oxford [21].

Antimicrobial Susceptibility Testing
Antimicrobial susceptibility testing (AST) was conducted on 43 isolates that had their whole genome sequenced using the broth microdilution method according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [44] recommendations for fastidious microorganisms. Briefly, the isolates were recovered from freezer stocks (Tryptic soy broth with 20% glycerol, −80 • C) and incubated overnight on blood agar supplemented with 5% sheep blood in a microaerobic atmosphere. AST was carried out on EUCAMP3 microplates (Sensititer, Trek Diagnostic Systems Ltd. East Grinstead, West Sussex, RH19 1XZ, UK). Mueller-Hinton broth was used for the preparation of 0.5 McFarland solution, and the inoculum was made by adding 100 µL of initial solution in 11 mL of cationadjusted Mueller-Hinton broth supplemented with 5% lysed horse blood and 20 µg/L β-NAD. Microplates were incubated in a microaerobic environment at 41 • C for 24 h. The susceptibility to chloramphenicol (CHL; 2-64 mg/L), erythromycin (ERY; 1-512 mg/L), Antibiotics 2023, 12, 1310 8 of 10 ciprofloxacin (CIP; 0.12-32 mg/L), and tetracycline (TET; 0.5-64 mg/L) was determined using EUCAST C. coli epidemiological cut-off values (ECOFFs), while for ertapenem (ERTA 0.12-4 mg/L) and gentamicin (GEN; 0.25-16 mg/L), the European Food Safety Authority (EFSA) cut-off values were used, as there are no available data from the EUCAST [25]. Reference strains C. jejuni ATCC 33560 and Staphylococcus aureus ATCC 29213 were used for quality control.

Whole-Genome Sequencing Data Analysis
Basic bioinformatics analysis was conducted by MicrobesNG (Birmingham, UK). All the obtained reads were put through a standard analysis pipeline. The closest available reference genome was identified using Kraken, and the reads were mapped to this using BWA-MEM (Burrows-Wheeler Aligner) to assess the quality of the data. A de novo assembly of the reads was obtained using SPAdes, and the reads were mapped onto the resultant contigs, again using BWA mem to obtain more quality metrics. Also, automated annotation was performed using Prokka.
All but five samples were de novo assembled with a genome size of approximately 1.5 Mb and a GC content of around 30%. Samples 1c5, 71, 175, 176, and 266 were interspecies-contaminated, but we used reference mapping (reference sequence: CP0000932) to isolate only the C. lari sequences.

Antimicrobial Resistance
The WGS-derived antimicrobial resistance (AMR) was analyzed for de novo assemblies using the publicly available service ResFinder 4.1 at https://cge.food.dtu.dk/services/ ResFinder (accessed on 10 July 2023) provided and curated by the Center for Genomic Epidemiology. We analyzed 43 assembled C. lari genomes for chromosomal point mutations (also for all unknown mutations) with a 98% threshold for %ID and 80% minimum length and acquired antimicrobial resistance genes using the same restrictions.

Genotypic-Phenotypic Comparisons
The WGS-derived AMRs were compared to the results of in vitro AST for six clinically relevant antimicrobial agents (erythromycin, ciprofloxacin, tetracycline, gentamicin, chloramphenicol, and ertapenem).
The concordance between the methods was determined by comparing the genotypic detection of known resistance determinants against the phenotypic susceptibility results of each strain at a concentration equal to the ECOFF described by the EUCAST and EFSA [25,44].
Major errors were classified as those instances where a strain was predicted to be resistant to the detection of an AMR determinant in the genome, but was phenotypically susceptible. Very large errors were classified as those instances where a strain was predicted to be susceptible by the absence of an AMR determinant in the genome, but was phenotypically resistant.