Serotypes, Antimicrobial Resistance Profiles, and Virulence Factors of Salmonella Isolates in Chinese Edible Frogs (Hoplobatrachus rugulosus) Collected from Wet Markets in Hong Kong

Salmonella is an important agent of gastrointestinal disease in humans. While livestock, such as cattle, poultry, and pigs, are well-recognised animal reservoirs of Salmonella, there is a lack of data on Salmonella in edible frogs, even though frog meat is a popular food worldwide. In this study, 103 live edible Chinese frogs (Hoplobatrachus rugulosus) were collected from wet markets throughout Hong Kong. After euthanasia, faeces or cloacal swabs were examined for Salmonella. Overall, Salmonella spp. were isolated from 67 (65%, CI: 0.554–0.736) of the samples. The serotypes included S. Saintpaul (33%), S. Newport (24%), S. Bareilly (7%), S. Braenderup (4%), S. Hvittingfoss (4%), S. Stanley (10%), and S. Wandsworth (16%). Many isolates were phylogenetically related. A high number of genes encoding for resistance to clinically relevant antimicrobials, and a high number of virulence determinants, were identified. Antimicrobial susceptibility testing (AST) identified multidrug resistance (MDR) in 21% of the isolates. Resistance to ampicillin, ciprofloxacin, nalidixic acid, and tetracycline was common. These results demonstrate that a high percentage of live frogs sold for human consumption in wet markets are carriers of multidrug-resistant Salmonella. Public health recommendations for handling edible frogs should be considered, to mitigate the risk of Salmonella transmission to humans.


Introduction
Non-typhoidal Salmonella (NTS) species are among the most important etiological agents of gastrointestinal diseases in humans worldwide, causing an estimated 180 million diarrhoeal illnesses each year [1]. Most human infections involve the Salmonella enterica subspecies enterica, which contains more than 2600 serotypes (serovars) [2]. Salmonella serotypes are assigned using the traditional antibody-based serological method, according to the Kauffmann-White-Le Minor scheme [3], or by applying DNA-sequenced-based typing methods, such as multilocus sequence typing (MLST) and genetic antigen prediction using whole genome sequencing (WGS) data [4,5].

Sample Collection
All officially registered wet markets in Hong Kong (a total of 94 markets distributed across Hong Kong, including Hong Kong Island, Kowloon, and the New Territories) [18] were visited twice by a research assistant within a period of three months, and markets that sold edible frogs were noted. Of these wet markets, a total of nine (designated A-I) were selected randomly, and a total of 103 live Chinese edible frogs (H. rugulosus) were collected during March and April 2022 ( Table 1). The country of origin of the frogs from each market was recorded from the vendors.
The collected frogs were taken to the Department of Infectious Diseases and Public Health, City University of Hong Kong. Physical examinations of the frogs were performed by a certified veterinarian, and age, sex, and weight were documented. For this study, all frogs were considered, regardless of their age and sex.
Anaesthesia was induced by MS-222 (Tricaine methanesulfonate, Syndel, Ferndale, WA, USA) bathing at a concentration of 2 g/L. The MS-222 was buffered by an equal concentration of sodium bicarbonate (Sigma-Aldrich, St. Louis, MO, USA). Once anaesthesia was confirmed, humane euthanasia was induced by an intravenous injection of pentobarbital (Dorminal 20%, Alfasan, Woerden, The Netherlands) at a dose of 100 mg/kg in the ventral abdominal vein. Decapitation was performed as a secondary method of euthanasia, following the guidelines of the American Veterinary Medical Association [19]. During post-mortem examination, faeces or cloacal swabs were aseptically collected, and placed into sterile tubes containing Amies agar gel (Thermo Fisher Scientific, Melbourne, Australia), until further processing.

Salmonella Isolation and Identification
Salmonella spp. detection was performed, using the ISO 6579-1: 2017 method for isolation and identification of Salmonella [20]. Each swab was placed in 10 mL buffered peptone water (BPW; Thermo Fisher Scientific), and incubated at 37 • C for 18 h. Following incubation, 0.1 mL of cultured BPW was used to inoculate 10 mL Rappaport-Vassiliadis Soya Peptone broth (RVS broth; bioMérieux, Marcy-l'Étoile, France), with incubation at 41.5 • C for 24 h. Cultured RVS broth was streaked on Xylose Lysine Deoxycholate agar (XLD agar; Thermo Fisher Scientific), and incubated at 37 • C for 24 h. Salmonella Typhimurium ATCC 14028 was used as a positive control. No plating media, other than XLD, were used, and no other biochemical characterisations were done. Putative Salmonella colonies with black morphologies were subjected to species identification by Matrix-assisted laser desorption/ionisation time-of-flight mass spectrometry (MALDITOF; Bruker, MA, US), and analysed using the MALDI Biotyper (Bruker, MA, US). For calibration and internal quality control of the MALDI Biotyper System, the Bruker Bacterial Test Standard (BTS) containing extract of Escherichia coli DH5 alpha was used, according to the manufacturer's instructions (Bruker). Salmonella Typhimurium ATCC 14028 was used for internal evaluation.

Salmonella Serotyping
All isolates were serotyped at the Swiss National Reference Centre for Enteropathogenic Bacteria and Listeria (NENT), Switzerland. Serotyping was performed according to the Kauffmann-White-Le Minor scheme [3], using commercially available antisera (Sifin Diagnostics, Berlin, Germany).

DNA Extraction and Whole Genome Sequencing
Isolates were grown on sheep blood agar at 37 • C overnight, prior to DNA isolation using the DNeasy Blood & Tissue Kit (Qiagen, Hombrechtikon, Switzerland). DNA libraries were prepared, using a Nextera DNA Flex Sample Preparation Kit (Illumina, San Diego, CA, USA). Whole genome sequencing was performed on an Illumina MiniSeq Sequencer (Illumina). The Illumina reads files passed the standard quality checks, using the software package FastQC 0.11.7 (Babraham Bioinformatics, Cambridge, UK), and were assembled using the Spades 3.14.1-based software Shovill 1.0.4 [21], using default settings. The assembly was filtered, retaining contigs >500 bp, and annotated using the NCBI prokaryotic genome annotation pipeline [22].
Whole-genome-based Salmonella serotyping was performed, using SeqSero with standard settings [23]. Sequence types (STs) were determined using a 7-house-keeping genebased in silico MLST scheme and the Ridom SeqSphereC + software version 8.5 (Ridom GmbH, Münster, Germany). The genetic relatedness of the isolates was assessed by core genome MLST (cgMLST) analyses based on a 3002-locus cgMLST scheme using Ridom SeqSphere+. A minimum spanning tree (MST) was generated for visualisation, with the threshold for cluster identification set to ≤10 alleles between a pair of neighbouring isolates, according to the Ridom SeqSphere+ software. Antimicrobial resistance genes were determined by the Resistance Gene Identifier (RGI) 4.2.2, using database version 3.2.6 [24]. Virulence factors (VFs) were identified by a bi-directional best-hit approach, using Diamond with standard setting [25]. The Prokka-predicted proteomes were compared to the representative-proteins data set A of the virulence factor database VFDB, downloaded in May 2023 [26] The aggregate VF score was defined as the number of unique VF detected for each isolate.

Statistical Analysis
Comparisons of the proportions of samples containing Salmonella from frogs of different age categories (adult or juvenile) and of different sex (male or female) were performed by Fisher's exact test. The significance criterion was set at p ≤ 0.05. Calculations were performed using GraphPad (https://www.graphpad.com, accessed on 8 March 2023).

Ethics Statement
This study was approved of by the Animal Research Ethics Sub-Committee of the City University of Hong Kong (Internal Ref: A-0698).

Demographic Data of H. rugulosus Collected from Different Wet Markets
Of the 103 collected frogs, 76 (74%) were adult, and 27 (26%) were juvenile animals. A total of 81 (79%) were male, 21 (20%) were female, and 1 (1%) was a hermaphrodite. The median weight was 164 g (range 91-292 g). Frogs purchased in wet markets A-H were imported from mainland China, whereas those from market I were imported from Thailand.

Prevalence of Salmonella among H. rugulosus
Salmonella enterica subsp. Enterica was isolated from 67 (65%, 95% CI: 0.554-0.736) of the frogs (Table 1). Frogs harbouring Salmonella were collected from all nine wet markets, and the proportions of Salmonella-positive animals varied, from 17% of the animals collected from wet market C, to 100% of those from markets D, E, and H ( Table 1).

Core Genome Multilocus Sequence Types (cgMLST) and Phylogenetic Relatedness
Based on WGS data, the 67 isolates were assigned to nine different STs, whereby ST50 (n = 22; 33%) and ST1498 (n = 11;16%) were predominant ( Table 2). Isolates from the same serotype were assigned to the same ST, except for the S. Newport isolates, which showed three different MLST patterns corresponding to ST31 (n = 8), ST45 (n = 7), and ST46 (n = 1), respectively ( Table 2). The genetic relatedness of the isolates was visualised by constructing a cgMLST-based phylogenetic tree. The isolates grouped according to serotypes and STs, which were separated by clear allelic distances (Figure 1). Isolates belonging to S. Bareilly, S. Braenderup, and S. Hvittingfoss clustered tightly within their respective serotypes and STs. By contrast, the S. Newport ST31 isolates formed two different subclusters (subcluster 1 and subcluster 2 in Figure 1), which consisted of five and three isolates, respectively, while the S. Newport ST45 isolates were indistinguishable by cgMLST ( Figure 1). Isolates belonging to S. Saintpaul ST50 appeared in two subclusters, with one major subcluster containing 18 isolates, and one minor subcluster formed by two isolates (subcluster 3 and subcluster 4 in Figure 1). Two further S. Saintpaul isolates were observed at distinct positions in the tree ( Figure 1). Likewise, the S. Stanley ST29 isolates formed two clusters consisting of four and two isolates, respectively (subcluster 5 and subcluster 6 in Figure 1). Finally, the S. Wandsworth ST1438 isolates comprised two subclusters containing five and four isolates, respectively (subcluster 7 and subcluster 8), and two distinctly separated isolates ( Figure 1).        Table 2.
The metallo-beta-lactamase gene bla IND-6 was found in one S. Wandsworth ST1498 (isolate F33) ( Table 2). Notably, vanX was identified in one S. Wandsworth ST1498 (isolate F86), but was not related to vancomycin resistance, because Gram-negative bacteria are insensitive to vancomycin [31]. All isolates carried either aac(6 )-Iy or aac(6 )-Iaa, which are cryptic genes that do not confer phenotypic aminoglycoside resistance [29,30]. Furthermore, all isolates contained a large number of genes encoding MDR efflux pumps, and genes encoding for antibiotic target alterations (Supplemental Table S1): among the latter, the gyrA (S83F) mutation was detected in one S. Newport ST46 (isolate F77) (Supplemental Table S1). Additional data are provided in supplemental File S1.
None of the isolates were resistant to cefotaxime, cefepime, fosfomycin, or nitrofurantoin (Table 2 and Supplemental Table S2). There was a correlation between phenotypic antimicrobial resistance and the presence of at least one ARG encoding resistance to aminoglycosides, amoxicillin/clavulanic acid, folate pathway inhibitors, phenicols, macrolides, and tetracycline, but no concordance between phenotypic and genotypic resistance to extended-spectrum beta-lactams and nalidixic acid ( Table 2).

Discussion
Although frog meat is a popular food worldwide [35], there is a lack of information regarding the occurrence of Salmonella in frogs sold for human consumption. In the current study, the overall prevalence of Salmonella carriage among H. rugulosus obtained from wet markets in Hong Kong was 65%, and included a variety of Salmonella serotypes, with many isolates featuring genotypic and phenotypic resistance to clinically relevant antimicrobials.
Among the isolates, several serovars, including S. Hvittingfoss, S. Newport, S. Stanley, and S. Wandsworth, had been described previously in H. rugulosus farmed in Thailand [36]. Furthermore, S. Saintpaul and S. Wandsworth had been isolated from H. rugulosus used as feeder frogs for captive reptiles in Thailand [37]. These reports are suggestive of frogs as natural hosts for these serotypes. Some serotypes in this study are infrequently reported, but have nevertheless been associated with human cases of salmonellosis. For example, the rare serotype S. Hvittingfoss ST446, found in three frogs in this study, was recently linked to an outbreak of salmonellosis in travellers returning from Hong Kong [38]. Likewise, S. Wandsworth, identified in 11 frogs, was associated with infections in small children in the USA in 2009, and with peritonitis in a patient in Hong Kong in 2022 [39,40]. The identification of this rare serotype among 16% of the frogs indicates that S. Wandsworth may be widespread among frogs, and that frogs may represent an important reservoir for this serotype. Other serotypes found in this study are more frequent among human infections, and have also been implicated in recent disease outbreaks in the EU (S. Bareilly, S. Braenderup, S. Newport, S. Stanley) [6], or are endemic in Asia but less frequent in Europe, such as S. Stanley [41].
Phylogenetic analysis showed that many of the Salmonella STs in this study were closely related by cgMLST, with subclusters observed among isolates from frogs obtained from different markets. Despite the lack of details on the acquisition of the frogs sold at the wet markets, these subclusters may be considered to reflect a common origin of the frogs, e.g., farm or location of capture. In particular, the clonality of the MDR S. Newport ST45 isolates is suggestive of a lineage deriving from a common source.
Multiple genes conferring resistance to antimicrobials that are listed as critically important, or very important to human health, were detected [42]. One of the clinically relevant AMR genes included bla DHA-1 , identified among six MDR S. Newport ST45. DHA-1 belongs to the family of plasmid-mediated AmpC beta-lactamases that confer resistance to the betalactamase inhibitor clavulanic acid [43]. Accordingly, all DHA-1-producing S. Newport in this study were phenotypically resistant to ampicillin and to amoxycillin/clavulanic acid. The bla DHA-1 gene is infrequently reported in Salmonella and other Enterobacterales [44]. To the best of our knowledge, this is the first report of bla DHA-1 in S. Newport isolated from amphibians, although one DHA-1-producing human clinical S. Newport was described in Spain between 2004 and 2009 [45]. One of the most prevalent AMR genes in this study was bla TEM-60, which encodes an extended-spectrum beta-lactamase (ESBL); however, in this study, bla TEM-60 -carriers failed to exhibit a detectable ESBL phenotype. Likewise, the bla IND-6 -harbouring S. Wandsworth did not exhibit resistance to carbapenems. By contrast, the abundance of the plasmid-mediated quinolone resistance (PMQR) genes aac(6 )-Ib-cr6, qnrB4, qnrS1, and qnrS2 among the isolates was reflected in a high proportion of nalidixicacid-and ciprofloxacin-resistant isolates. Markedly, none of the isolates in this study were fully susceptible to ciprofloxacin. PMQR among Salmonella is of concern, because it has the potential to spread horizontally, and, although PMQR genes only confer low-level resistance to quinolones, they promote step-wise development to high-level fluoroquinolone resistance [46,47]. The presence of PMQR genes in human pathogens isolated from animals destined for human consumption is of concern, because Salmonella with decreased ciprofloxacin resistance carrying PMQR genes have been linked to outbreaks, as previously described, in the UK with S. Virchow [48], in the USA with S. Newport [49], and in Spain with S. Corvallis [50].
With resistance to fluoroquinolones increasing worldwide, azithromycin is currently considered an important alternative for the treatment of invasive salmonellosis [42]. Azithromycin resistance in human NTS is currently rare in the USA (0.5%) and Europe (0.8%) [51,52]. Moreover, resistance to azithromycin was either not detected, or was identified only at very low levels in food-producing animals, including calves (0%), broilers (1.9%), pigs (0.3%), and turkeys (1.2%), in the EU, between 2019 and 2020 [52]. By contrast, azithromycin resistance, mainly associated with the presence of mphA, has been increasing over time, and seems to be more prevalent in human NTS in Europe than in Taiwan (3.1%) [51], and in Salmonella from humans, animals, and food in China (3.9%) [53]. The observation, in our study, that azithromycin resistance and the mph gene occur more frequently in Salmonella of amphibian origin (5.7%) than in Salmonella of livestock or human origin may suggest the potential of H. rugulosus as a source of dissemination of azithromycin-resistant Salmonella.
Animals including amphibians may acquire antimicrobial-resistant bacteria from other animals, their environment, or their feed. Little information is available regarding frog farming and the usage of antimicrobial agents in frog farms in Asia; however, a recent report from Vietnam indicates that antimicrobials, such as amoxicillin, enrofloxacin, ciprofloxacin, sulfamethoxazole-trimethoprim, doxycycline, oxytetracycline, tetracycline, florfenicol, and rifampicin, are commonly applied empirically by frog farmers to treat frog diseases [16,54]: thus, the use and misuse of large amounts of different antimicrobials may select for resistant strains in farmed frogs. Although information on the origins of the frogs analysed in this study was available at country level, data on the aquacultural or natural settings were lacking, and it cannot be excluded that many of the frogs may have been farmed under comparable conditions. Similarly to our study, a recent investigation on wet markets in Hong Kong recently reported a high prevalence (42%) of Salmonella, and a high proportion of AMR isolates, among edible freshwater turtles, further underlining the potential risk of Salmonella and AMR in non-traditional aquaculture [55].
The presence of genotypic and phenotypic resistance characteristics in Salmonella in food animals is worrying, and indicates that first-line antibiotic ciprofloxacin, as well as ampicillin, amoxycillin/clavulanic acid, and azithromycin, used against such strains, may fail, should infectious transmissions to humans occur. Likewise, isolates showing intermediate resistance, i.e., reduced susceptibility, are of concern for public health. Reduced susceptibility is important, because it facilitates the selection of isolates with higher-level resistance, and contributes to the development and spread of resistance to critically important antimicrobial agents, such as ciprofloxacin [46,47].
The in silico screening of the VFs revealed an abundance of VFs in all the isolates whilst, concurrently, the differences in VF content among the various serotypes was rather unremarkable: these findings are in agreement with previous studies, in which VFs were reportedly conserved among NT Salmonella serotypes [56]. Based on the distribution of the VFs, and the lack of variation of median aggregate VF scores among the serotypes, there was no specific Salmonella serotype in this study that showed a distinct virulence profile; however, the universal presence of fimbrial genes, genes encoding secretion systems, and toxins among the isolates was an indication that these Salmonella serotypes have the potential to cause disease in humans.
This study had some limitations. Firstly, the interpretation of the results should take into consideration that sampling was restricted to H. rugulosus because this is the only frog species sold in wet markets in Hong Kong: thus, the results cannot be generalised to other frog species that may be available for human consumption elsewhere. Secondly, there were unequal sample sizes among the different wet markets from which the frogs were obtained; therefore, there remains the possibility of unintended overrepresentation of the proportion of positive samples from individual markets, which were randomly selected. Thirdly, the study was challenged by the lack of any previous data on the prevalence of Salmonella in frogs, and lack of data on frog import and retail volumes in Hong Kong; therefore, essentially, this study was a pilot study, to collect some baseline data. Finally, the sampling period was short: it should be considered that the prevalence of Salmonella among the frogs may have been subject to unknown factors-for example, fluctuations among the suppliers of the wet markets-and that such differences would have remained undetected.

Conclusions
This study demonstrates that a high proportion of live frogs sold for human consumption in wet markets are carriers of Salmonella, including serotypes that are frequently linked to cases of human disease, and Salmonella that contain clinically relevant AMR genes. To the best of the authors' knowledge, no official advice or guidelines exist regarding the handling of live edible frogs. To this end, the results presented in this study offer useful information for specialists in the public health and food safety sectors. Raised awareness and public health recommendations for hygiene requirements at the stages of handling and processing of edible frogs and their meat should be considered, to mitigate the risk of transmission of susceptible and resistant Salmonella to humans. The role of edible frogs and derived meat products, e.g., frog legs, as sources of antimicrobial-resistant and susceptible Salmonella and other foodborne pathogens, needs to be further investigated in future.
Supplementary Materials: The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/foods12112245/s1. Table S1: Antibiotic efflux pump genes and target alteration resistance mechanisms detected by RGI software in 67 Salmonella from Chinese edible frogs (Hoplobatrachus rugulosus) collected from wet markets in Hong Kong. Table S2: Antimicrobial susceptibility determined by disk diffusion of 67 Salmonella isolated from Chinese edible frogs (Hoplobatrachus rugulosus) from wet markets in Hong Kong. Table S3: Distribution of virulence genes detected via VFDB software in 67 Salmonella from Chinese edible frogs (Hoplobatrachus rugulosus) collected from wet markets in Hong Kong. File S1: Metadata of antimicrobial resistance genes collected by whole genome sequencing of 67 Salmonella isolates.   Table 2.