Whole-genome characterisation of multidrug resistant monophasic variants of Salmonella Typhimurium from pig production in Thailand

Monophasic S. Typhimurium isolate collection

Monophasic Salmonella Typhimurium were isolated during the period of November 2011 through July 2014 in Chiang Mai and Lamphun provinces, as decribed in the studies of Tadee et al. (2015) and Patchanee et al. (2016). The studies were conducted under ethical approval reference number R18/2554 by the Animal Care and Use Committee committee of Faculty of Veterinary Medicine, Chiang Mai University (FVM-ACUC). Briefly, samples were collected at three stages of pig production: on the farm (n = 10); at the slaughterhouse (n = 6) and at retail from local markets and supermarkets (n = 3) (Table 1).

Antimicrobial susceptibility assays

All isolates were tested for their susceptility to antimicrobials using disk diffusion methods following Clinical and Laboratory Standards Institute guidelines (Clinical & Laboratory Standards Institute (CLSI), 2011) by the WHO National Salmonella and Shigella Center, National Institute of Health, Department of Medical Science, Nonthaburi, Thailand. Ten different antimicrobial agents were tested, including ampicillin (AMP) 10 µg, amoxicillin-clavulanic acid (AUG) 20/10 µg, chloramphenicol (C) 30 µg, ciprofloxacin (CIP) 5 µg, cefotaxime (CTX) 30 µg, nalidixic acid (NA) 30 µg, norfloxacin (NOR) 10 µg, streptomycin (S) 10 µg, sulfamethoxazole-trimethoprim (SXT) 23.75/1.25 µg and tetracycline (TE) 30 µg (Table 1).

Whole-genome sequencing

Genomic DNA of 19 monophasic S. Typhimurium isolates was extracted, using a QIAamp DNA mini kit (Qiagen, Crawley, UK) and sequenced on an Illumina MiSeq genome sequencer (Illumina, Cambridge, UK). Nextera XT libraries (Illumina, California, USA) were prepared and short paired-end reads (300 bp) trimmed using TRIMMOMATIC (version 0.33; Bolger, Lohse & Usadel, 2014) and assembled de novo using SPAdes software, with the-careful command (St. Petersburg genome assembler, St. Petersburg, Russia; version 3.10.035) (Bankevich et al., 2012). The average number of contigs was 400 (range: 187–777) for an average total assembled sequence size of 5.02 Mbp (range: 4.87–5.20) with an average genome coverage of ×34 (range 16–46). The average N50 contig length (L50) was 33,091 bp (range: 10,047–84,409) and the average GC content was 52.2% (range: 52.0–52.5). Short read data are archived on the National Center for Biothechnology Information (NCBI) SRA, associated with BioProject accession PRJNA573746 (individual accession numbers in Supplemental File S1). FASTA files of all contiguous assemblies and Supplemental Files are archived at the public data repository figshare (https://doi.org/10.6084/m9.figshare.11305727).

Multilocus sequence typing

Nucleotide sequences for seven housekeeping gene loci, including aroC (chorismate synthase), dnaN (DNA polymerase III subunit beta), hemD (uroporphyrinogenIII cosynthase), hisD (histidinol dehydrogenase), purE (phosphoribosy laminomidazole), sucA (alpha-ketoglutarate dehydrogenase) and thrA (aspartokinase I/homoserine dehydrogenase) were identified. Alleles and subsequent sequence types (ST) were designated using the web-based MLST 2.0 method at the Center for Genomic Epidemiology (https://cge.cbs.dtu.dk/services/MLST/) (Larsen et al., 2012).

Core genome genealogy

An alignment file was constructed from concatenated gene sequences of all core genes (found in ≥95% isolates) from the reference Salmonella Typhimurium genome, SO4698-09 (accession# LN999997.1) using MAFFT (Katoh & Standley, 2013) on a gene-by-gene basis (Méric et al., 2014; Sheppard, Jolley & Maiden, 2012) (size: 4,166,027 bp; Table S1). A gene was considered present in a given genome when its sequence aligned to an SO4698-09 locus with more than 70% sequence identity over at least 50% of sequence length using BLAST. Maximum-likelihood phylogenies were built in IQtree v1.6.8 (Center for Integrative Bioinformatics Vienna, Vienna, Austria) using the GTR + F + I + G4 substitution model. A consensus tree constructed from 1,000 bootstrap trees (Hoang et al., 2018) and visualised on microreact (Argimón et al., 2016).

Whole-genome multilocus sequence typing

Core genome multilocus sequence typing (cgMLST) and pan genome multilocus sequence typing (pgMLST) were performed using BioNumerics version 7.6 (Applied Maths, Ghent, Belgium). A total of 2,978 and 5,022 loci were identified as composing the cgMLST and pgMLST scheme, respectively. Cluster analysis of categorical differences of allelic numbers for cgMLST and pgMLST was applied to generate dendrograms, using the complete linkage algorithms.

Identification of antimicrobial resistance, virulence and plasmid associated genes

Nucleotide sequences of all 19 monophasic S. Typhimurium isolates in FASTA format were compared to antimicrobial resistance genes (ARGs), putative virulence factors and known plasmid genes using ABRICATE 0.9.8 (GNU, Boston, MA, USA), using Comprehensive Antibiotic Resistance Database (CARD) (Alcock et al., 2020), the NCBI (Feldgarden et al., 2019), ResFinder 3.0 (Zankari et al., 2012). The virulence factor database (VFDB) (Chen et al., 2016) and PlasmidFinder 2.1 (Carattoli et al., 2014) databases. All databases were accessed at 10 September 2019.

Identification of Salmonella Pathogenicity Islands

The Center for Genomic Epidemiology was also used to identify SPIs in our Salmonella isolates. SPIFinder 1.0 (https://cge.cbs.dtu.dk/services/SPIFinder-1.0/) with default settings of 95% threshold over 60% minimum length were used.

Pig isolates from Thailand are related to a globally disseminated clone

All monophasic S. Typhimurium isolates were sequenced, genomes assembled and typed using an in silico MLST scheme. Each of the 19 isolates were categorised as sequence type 34 (ST-34), with the alleic profile: aroC_10/dnaN_19/hemD_12/hisD_9/purE_5/sucA_9/thrA_2. Isolate genomes were mapped to the S. Typhimurium ST-34 reference, SO4698-09 and homologues of all core genes (Table S2) aligned and used to construct a maximum-likelihood phylogenetic tree (Fig. 1). Isolates were closely related and the sum of all internal branch lengths represented less than 5% of the tree length, indicating that all isolates were closely related to the globally disseminated, livestock-associated ST-34 clone.

Extended whole genome typing analysis (wgMLST) of the 19 monophasic S. Typimurium genomes was performed for core- (cgMLST; 2,978 genes present in 100% of isolates) and pan- (pgMLST; 5,022 genes) genomes. All isolates were closely related, with a maximum 130 and 170 locus differences between isolates, for cgMLST and pgMLST analyses respectively. Isolates A541011 and A541013 were the most closely related isolates and were collected from the same farm (PD farm), on the same sampling date (15 November 2011) and only differed in 18 core loci. When accessory genes were included in the pan-genome anlaysis, isolates A541006 and A541012 were the most closely related, with 26 loci differences. Visual comparison of the core- and pan-genome MLST dendrograms demonstrates similarity in the clustering of isolates (Fig. S1).

Clonal isolates shared different ARGs

The genome sequences of all 19 monophasic S. Typhimurium isolates were compared using ABRICATE 0.9.8 and known antibiotic resistance genes and alleles from CARD, ResFinder and NCBI databases (Table S3). As there is redundancy in these antibiotic resistance gene sequence databases, we show only results from the ResFinder comparison (Fig. 1). Twenty-five resistance genes were identified, from ten antibiotic classes (Table 2). Eight detected genes (8/25; 32%) were grouped in aminoglycoside classes. All monophasic S. Typhimurium genomes (19/19; 100.00%) harboured aminoglycoside and beta-lactam class resistance genes. Sulphonamide and tetracycline resistance genes were the next most prevelant (18/19; 94.73%), followed by fluoroquinolone (10/19; 52.63%) and phenicol (7/19; 36.84%). Fusidic acid, glycopeptide, nitroimidazole, oxazolidinone and rifampicin associated-resistance genes were not detected in any of the genomes. Putative resistance genes were also detected for colistin (7/19; 36.84%), phosphomycin (2/19; 10.52%) and macrolide-lincosamide-streptogramin B (3/19; 15.78%), although no phenotypic susceptibility tests were performed to support these results.

Predicted resistance generally matched the phenotypic susceptibility profiles (Table 3). Nearly half of the antimicrobial classes tested (2/5; beta-lactam and tetracycline) were 100% concordant. In other comparisons, some isolates were predicted to be resistant, but no phenotypic resistance was observed (n = 2 aminoglycoside; n = 9 fluoroquinolone; n = 5 phenicol). Conversely, one isolate was phenicol phenotypically resistant, but no antimicrobial resistance determinant was identified.

Agricultural reservoir isolates contribute to public health risk

Fewer ARG determinants were found in the isolates as pigs passed though the production process (farm: n = 10, mean = 11.8; slaughterhouse n = 6, mean = 8.5; marketplace n = 3, mean = 8.3), although with only a small number of samples strong statistical associations cannot be made. Samples collected from retail pork products were not predicted to confer resistance to macrolides, fosfomycin, fluoroquinolone, diaminopyrimidine or colistin (Fig. 2). Typically, isolates collected from farm animals contained ARGs conferring resistance to a broader range of antimicrobials. Comparison with the Plasmid finder database identified the ColRNAI plasmid (accession: DQ298019) in all marketplace isolates (3/3; 100%), nearly two-thirds (4/6; 66%) of slaughterhouse isolates and only a fifth of farm isolates (2/10; 20%) (Fig. 1). In addition, the IncQ1 plasmid (accession: M28829.1) was identified in multiple samples (13/19; 68%).

Identification of Salmonella Pathogenicity Island genes

ABRICATE was also used to identify known virulence genes in our monophasic S. Typhimurium isolates with Virulence finder databse (VFDB). In total, 137 different virulence genes were identified, including those from several Salmonella Pathogenicity Islands (SPIs) (Table S3). Seven out of 14 previously described SPIs were detected (Fig. 1A), supported by additional investigation of previously described SPIs using the SPIFinder 1.0 database. Isolates carried three to six SPIs, with more than half (12/19) carrying three islands. The TP3 isolate obtained from a retail pork sample at a local market carried six previously decribed SPIs (SPI-1, SPI-2, SPI-3, SPI-5, SPI-13 and SPI-14) and eleven antibiotic resistance gene determinants from six different antibiotic classes. All the monophasic S. Typhimurium isolates harboured SPI-5, SPI-13 and SPI-14 (100.00%; 19/19 strains), followed by SPI-3, which was carried by nearly one-third of them (31.57%; 6/19 strains).

The large Salmonella genomic island, SGI-4 (up to 87 genes) is often found in ST-34 clones from pigs. Gene-by-gene comparison of our isolates with the Typhimurium reference strain SO4698-09 identified many of the genes linked to carriage of the SGI-4 element (Table S3), including the mercury resistance genes merEDACPTR, which are found near an arsenate resistance operon (aseR, arsD, arsA1, arsB, arsenate reductase, bar N-acetyltransferase, arsA2, arsD2) and silver-copper resistance element (silESRCFBAGP pcoG silE pcoABCDRSE2). Variation has been observed in the SGI-4 genomic island, which has been found to contain up to 87 loci, with 48 described as core. In our collection, at least 84 loci (97%) of the SGI-4 island were identified in all isolates (Table S3).