Whole genome sequencing-based detection of antimicrobial resistance and virulence in non-typhoidal Salmonella enterica isolated from wildlife

A total of 103 Salmonella isolates were revived from archival cultures obtained from exotic pet or wildlife clinical specimens submitted to the Oklahoma Animal Disease Diagnostic Laboratory, Stillwater, Oklahoma during 1988–2003. The metadata for the samples used in this study is provided in Table 1 and the details of genome sequencing and assembly parameters are given in Additional file 1: Table S1. Isolates were streaked on Luria–Bertani agar slants and were transported to the Animal Disease Research and Diagnostic Laboratory, South Dakota State University, Brookings, South Dakota for WGS. Samples were streaked on Luria–Bertani agar plates upon arrival to the laboratory. A single bacterial colony from the agar plate was then inoculated to Luria–Bertani broth and cultured at 37 °C.

Genomic DNA was isolated from 1.0 mL overnight cultures using the Qiagen DNeasy kits (Qiagen, Valencia, CA, USA) according to manufacturer’s protocol. The quality of isolated DNA was analyzed using NanoDrop™ One (Thermo Scientific™, DE) and was quantified using Qubit^® 3.0 (Thermo Fisher Scientific Inc., MA) fluorometer and stored at − 20 °C until use. Whole-genome sequencing was performed on Illumina Miseq platform using V2 chemistry with 2 × 250 paired-end chemistry Briefly, the concentrations of genomic DNA samples were adjusted to 0.3 ng/µL concentration and were processed using Nextera XT DNA Sample Prep Kit (Illumina Inc. San Diego, CA). The libraries were normalized using bead-based procedure and pooled together at equal volume. The pooled library was denatured and sequenced using Miseq reagent version 2 (Illumina, Inc., CA).

The raw data files were de-multiplexed and converted to FASTQ files using Casava v.1.8.2. (Illumina, Inc, San Diego, CA). The FASTQ files were trimmed and assembled de novo using CLC Genomics workbench 9.4 (Qiagen Bioinformatics, CA). The antibiotic resistance genes in the assembled Salmonella genomes were identified by BLAST search against a local copy of the antibiotic resistance gene sequence data from ResFinder [9] and CARD [10]. The parameters used for BLAST search were ≥ 95% gene identity and 50% sequence length of the resistance gene. The virulence genes in the genomes were predicted using a similar approach. Salmonella virulence gene sequences were extracted from Virulence Factor Database [11] and Salmonella genome assemblies were searched against these sequences using BLAST with ≥ 90% gene identity and 50% sequence length cut off.

A total of 103 Salmonella isolates sampled between 1988 and 2003 from wildlife and exotic pets were included in the present study for determining the antimicrobial susceptibility using whole genome sequencing. Among 103 isolates, 52 isolates (50.48%) were from wild birds, 1 isolate (0.9%) was from fish, 25 isolates each (24.27%) were from reptiles and mammals (Table 1). The serovars of 96 isolates in this study were determined at the National Veterinary Service Laboratory, Ames, IA, and the remaining 6 serovars were predicted using Seqsero [12]. The serovar of one isolate (ADRDL-093) was not identified under Kauffmann-White classification. A total of 45 serovars were identified among the 103 isolates, of which Typhimurium (12.62%) was the most frequent serovar. Other serovars that had higher prevalence were Enteritidis (6.8%), Anatum (5.8%), Arizona (5.8%), Bredeney (3.9%) and Montevideo (3.9%). The presence of multiple serotypes in wildlife has also been reported from previous epidemiological studies. Nine Salmonella samples isolated from marine mammals and birds in California yielded 7 serovars [4]. Similar to our findings, Salmonella Typhimurium was reportedly the predominant serovar present in wildlife [13‐15] in various parts of the world.

Antimicrobial susceptibility test of 103 Salmonella bacterial isolates was performed using Sensititre NARMS gram-negative plate. The results were classified into 3 categories: resistant, intermediate, or susceptible. Fifty-two out of the 103 isolates (50.48%) showed resistance to at least one antibiotic (Fig. 1a). Resistance against the aminoglycoside streptomycin was most commonly observed. Forty-eight of the 103 isolates (46.6%) exhibited this phenotype. However, only three isolates (2.9%) showed resistance to gentamicin which also belonged to the aminoglycoside class of antibiotics. The isolates with resistance against gentamicin were also resistant to streptomycin. In the beta-lactam group, ampicillin resistance was the most common phenotype and was seen in 11 of the isolates (10.67%). Among these 11 isolates, few also shared resistance against other beta-lactams such as amoxicillin–clavulanic acid (4), cefoxitin (3), and ceftiofur (3). All the isolates were susceptible to ceftriaxone except one with intermediate resistance. The isolates that were susceptible to ampicillin were also susceptible to all other beta-lactams. Chloramphenicol resistance was observed for seven isolates (6.7%), trimethoprim–sulfamethoxazole resistance in 4 (3.88%) and tetracycline resistance in 19 (18.44%) of the isolates. All the isolates were susceptible to ciprofloxacin and all except one isolate was susceptible to nalidixic acid. Nine isolates were found to be multi-drug resistant having resistance against more than three antibiotics.

The presence of genes that could contribute to AMR was detected by BLAST searching the assembled Salmonella genomes against a local copy of Resfinder and CARD sequence data (Fig. 1b). Additional details on the query length and percentage of gene identity for the BLAST results are provided in Additional file 2: Table S2. Bacterial isolates showing “intermediate” resistance on antimicrobial susceptibility test was grouped with “susceptible” isolates for the calculation of sensitivity and specificity of AMR genotype. Twenty-two genes that provided resistance to aminoglycosides were detected and the genes were present in 100 isolates. The sensitivity was 100% and specificity was 5.45% for resistance against aminoglycosides. The low specificity was probably due to the lack of resistance genes being expressed in vitro. Genes responsible for resistance to beta-lactam antibiotics were detected in 11 isolates which were also resistant by antimicrobial susceptibility test. The plasmid-mediated cephalosporinase gene blaLAT-1, plasmid-borne class C beta-lactamase gene blaBIL-1, and blaCMY (Class C) genes were found together and were detected in three isolates. Genes belonging to blaTEM (class A) were found in eight isolates. Collectively, there were 280 beta-lactamase genes present in those 11 isolates. The sensitivity and specificity was 100% for beta-lactams. Phenicol resistance encoded by cat, catA1, and floR genes was present in 8 isolates. The sensitivity was 100% and specificity was 98.96% for phenicol resistance. dfrA1, dfrA10, dfrA12, sul1, sul2, and sul3 genes conferring resistance to trimethoprim–sulfamethoxazole drugs were present in 12 isolates. The sensitivity was 100% and specificity was 91.92% for trimethoprim–sulfamethoxazole. The sul1, sul2, and sul3 genes could also contribute to resistance against sulfisoxazole. However, a definite conclusion of genotype–phenotype correlation is lacking due to the absence of antimicrobial susceptibility test data that matches the CLSI recommended breakpoint for resistance against sulfisoxazole. Tetracycline resistance encoded by tet(A), tet(B), tet(C), and tet(D) genes for tetracycline efflux pumps were detected in 18 samples all of which were also resistant by antimicrobial susceptibility test. The sensitivity was 94.74% and specificity was 100% for tetracycline resistance. Two isolates carried the mph(A) gene which confers resistance to macrolides. However, the only macrolide that was tested was azithromycin and the genotype–phenotype relation could not be established due to lack of data from antimicrobial susceptibility test that matches with the breakpoint recommended by CLSI (> 32 mg/L).

Overall, the sensitivity for detecting AMR using genotype was 100% except for tetracycline where 1 isolate was phenotypically resistant even in the absence of the (tet) gene. The specificity for aminoglycosides had the highest degree of incongruence between genotype and phenotype. Fifty-two isolates that were positive for aminoglycoside resistance genes were phenotypically susceptible. Although not to the degree found in this study, a mismatch in phenotype-genotype correlation was also reported previously in E. coli and Salmonella for aminoglycoside resistance, especially for streptomycin [5, 16]. There was 100% phenotype-genotype correlation for beta-lactam resistance. Phenicols and tetracycline also had > 98% specificity, while trimethoprim–sulfamethoxazole had lower specificity (91.2%) because of four isolates that were genotypically resistant but were phenotypically susceptible. These results are also similar to those obtained in previous studies [5, 16] where correlation approaching 100% was obtained for antimicrobials other than aminoglycosides.

In addition to the genes that confer AMR, we also analyzed the genes that could confer multi-drug resistance (Fig. 1b). The golS gene is a promoter for multidrug efflux pump, mdsABC [17] and was detected among 84.46% (n = 87) isolates. Similarly, mdsABC (multidrug transporter of Salmonella) complex which is made up of mdsA, mdsB, and mdsC units, was found in all isolates that had golS gene except one isolate which lacked mdsB and mdsC genes. The mdsABC complex is known to provide resistance against a variety of drugs and toxins and is involved in Salmonella virulence and pathogenicity [17, 18]. The mdtK gene, a multi-efflux pump which could provide resistance against norfloxacin, doxorubicin and acriflavin [18] and sdiA, a regulator for multi-drug resistance pump AcraB [19], were present in 84.46 and 86.41% of the isolates respectively. The presence of these genes could contribute to the virulence and pathogenicity of these Salmonella isolates and also indicates the potential for these isolates to resist various antibiotics and toxins.

The genes that are associated with virulence among 103 wildlife Salmonella isolates were analyzed (Fig. 2) using CLC workbench 9.4. The parameters used were the minimum identity of 90% and minimum length of 50%. Additional details on the query length and percentage of gene identity for the BLAST results are given in Additional file 3: Table S3. A total of 197 virulence genes were detected by BLAST search against a local copy of the Virulence Factor Database. The virulence-associated determinants collectively were grouped under 9 categories: fimbrial adherence determinants, macrophage inducible genes, determinants associated with magnesium uptake, nonfimbrial adherence determinants, genes associated with secretion system, serum resistance determinants, stress proteins, toxins, and two-component regulatory systems.

Among fimbrial adherence determinants, the genes belonging to two csg operons csgBAC and csgDEFG were present universally in all isolates. These genes encode for curli fimbriae or thin aggregative fimbriae and mediate binding to various serum and tissues matrix proteins [20]. Another gene cluster that was ubiquitously present were the fim genes that encodes for type 1 fimbriae. This cluster is comprised of the fimAICDHF operon and three regulatory genes fimW, fimY, and fimZ and mediates adherence to eukaryotic cells [21]. However, the fimY gene was not detected in ten isolates at the BLAST search cut-off level we used.

The genes belonging to type III secretion system (TTSS/T3SS) encoded by Salmonella pathogenicity island-1 (SPI-1) and -2 (SPI-2) were also predominantly present among the isolates. This included SPI-1 regulator genes hilACD, and SPI-1 encoded inv/spa, and prg/org operons that were detected in all the isolates. Similarly, SPI-2 regulatory gene ssrB, chaperone protein-encoding genes—sscA and sscB, and ssa genes that encode for T3SS2 apparatus were also present among 103 isolates. However, the sse genes which encode for the effectors were observed only in fewer isolates. Another set of genes that were present in all isolates were the genes that respond to magnesium level in the extracellular environment [22]. This included mgtc, which mediates magnesium uptake and phoP–phoQ genes that are regulators of the two-component system.

The least abundant virulence determinants were the tcf, sta, and pef fimbrial operons and spv gene cluster. These genes belonging to the fimbrial adherence determinants category were detected in less than 25% of the isolates. Additionally, rck gene that provides protection against the complement-mediated immune response of the host was also found in low abundance. There were 16 isolates that possessed fewer than 50% of the total virulence genes in the database (Fig. 2). These isolates include ADRDL-002, -003, -019, -020, -021, -022, -023, -024, -032, -033, -042, -052, -060, -061, -075, and -076. Importantly, these isolates also had a lower abundance of genes that contributed to multi-drug resistance (Fig. 1). However, these isolates come under various serotypes and were isolated from different host species. Therefore, a common factor responsible for the observed low abundance of virulence genes is not evident. The universal presence of fimbrial genes and the genes encoded by pathogenicity islands 1–2 among the isolates we report here indicates that these isolates could potentially cause disease in humans. Therefore, the genomes we report here could be a valuable reference point for future traceback investigations in instances where wildlife may be considered as a potential source of human Salmonellosis.