Phylogenetic and antimicrobial resistance gene analysis of Salmonella Typhimurium strains isolated in Brazil by whole genome sequencing

Whole genome sequencing (WGS) has been used as a powerful technology for molecular epidemiology, surveillance, identification of species and serotype, identification of the sources of outbreaks, among other purposes. In Brazil, there is relatively few epidemiological data on Salmonella. In this study, 90 Salmonella Typhimurium strains had their genome sequenced to uncover the diversity of Salmonella Typhimurium isolated from humans and food, between 1983 and 2013, from different geographic regions in Brazil based on single nucleotide polymorphism (SNP) analysis. A total of 39 resistance genes were identified, such as aminoglycoside, tetracycline, sulfonamide, trimethoprim, beta-lactam, fluoroquinolone, phenicol and macrolide, as well as the occurrence of point mutations in some of the genes such as gyrA, gyrB, parC and parE. A total of 65 (72.2%) out of 90 S. Typhimurium strains studied were phenotypically resistant to sulfonamides, 44 (48.9%) strains were streptomycin resistant, 27 (30%) strains were resistant to tetracycline, 21 (23.3%) strains were gentamicin resistant, and seven (7.8%) strains were resistant to ceftriaxone. In the gyrA gene, it was observed the following amino acid substitutions: Asp(87)→Gly, Asp(87)→Asn, Ser(83)→Phe, Ser(83)→Tyr. Phylogenetic results placed the 90 S. Typhimurium strains into two major clades suggesting the existence of a prevalent subtype, likely more adapted, among strains isolated from humans, with some diversity in subtypes in foods. The variety and prevalence of resistant genes found in these Salmonella Typhimurium strains reinforces their potential hazard for humans and the risk in foods in Brazil.

Introduction Foodborne diseases have a major impact on the economy and public health worldwide. Nontyphoidal Salmonella (NTS) is one of the most common causes of bacterial foodborne illnesses [1,2]. It is estimated that NTS cause about 93.8 million annual cases of gastroenteritis and 155 thousand deaths per year worldwide [1].
Among the Salmonella enterica serovars, Salmonella Typhimurium (S. Typhimurium) is among the most frequent ones isolated worldwide [3]. From 2001 to 2007, this serovar was the most prevalent in the United States, Canada, Australia and New Zealand. In the same period, S. Typhimurium appeared as the second most prevalent serovar in Africa, Asia, Europe and Latin America, surpassed only by S. Enteritidis [3].
In Brazil, there are relatively little epidemiological data on Salmonella [4][5][6][7]. However, it is known that in the State of São Paulo, S. Typhimurium was the most commonly isolated serovar from human sources and the third most common from non-human sources before the 1990's [4]. After this period, S. Typhimurium declined becoming the third most commonly isolated serovar from human and non-human sources in the period of 1991-1995 in São Paulo State in Brazil, with S. Enteritidis being the most isolated serovar in both sources and, S. I 4, (5), 12:i:-and S. Havana the second most isolated serovar in human and non-human sources, respectively [5]. Between 1996 and 2000, the isolation of S. Typhimurium declined even more from non-human sources [6]. However, between 1996 and 2003, this serovar was ranked as the second most commonly isolated serovar from human sources [7].
Epidemiological studies have been crucial to verify the relationship among pathogenic strains isolated from different sources, to elucidate contamination routes and to differentiate strains isolated from outbreaks and sporadic cases. Investigative capabilities have been greatly enhanced with the development and increasing feasibility of WGS as a molecular epidemiological tool [8][9][10]. Over the last few years there has been a substantial reduction in the costs of WGS making this technology economically viable as a routine tool for molecular epidemiology. WGS has also been used for detection of antibiotic resistance determinants [11,12].
The use of antimicrobials is not recommended in cases of noninvasive Salmonella infections [13,14]. However, in some cases, the antibiotic therapy might be necessary. The drug of choice for the treatment of Salmonella infections is typically ciprofloxacin due to its broad spectrum antimicrobial activity [14].
The extensive use of antimicrobials has led to increasing numbers of non-typhoidal Salmonella strains that are resistant to quinolones and exhibited reduced susceptibility to fluoroquinolones [15][16][17]. This reduced susceptibility can lead to treatment failures in some cases [18,19]. Quinolone resistance is usually mediated by mutations in the quinolone resistance determining regions (QRDRs) of the gyrA, gyrB, parC, and parE genes that code for bacterial DNA gyrase leading to changes in the binding site of the antimicrobial to the enzyme [17,20,21]. Also, quinolone resistance may be due to the acquisition of plasmid-mediated quinolone resistance (PMQR) genes [22][23][24], such as the qnr genes that encodes a group of pentapeptide proteins that bind to DNA gyrase and prevent the action of quinolones, qepA gene, an quinolone efflux pump, aac(6')-Ib-cr gene that encodes to the aminoglycoside acetiltranferase that can reduce susceptibility to ciprofloxacin and oqxAB genes, a multidrug resistance efflux pump [25].
In previous studies of our group, we typed S. Typhimurium strains isolated from humans and food between 1983 and 2013 in Brazil by Pulsed-field gel electrophoresis (PFGE), multiple-locus variable number of tandem repeats analysis (MLVA), enterobacterial repetitive intergenic consensus PCR (ERIC-PCR), CRISPR-multi-locus virulence sequence typing (CRISPR-MVLST) and Multilocus sequence typing (MLST). Moreover, the frequency of 12 virulence markers was assessed by PCR and the resistance profile against 12 antimicrobials was verified [26][27][28].
In this present work, WGS is used to uncover the diversity of Salmonella Typhimurium isolated from humans and food, between 1983 and 2013, from different geographic regions in Brazil. Additionally, WGS is used to verify the presence of antimicrobial resistance genes, as well as, the occurrence of mutations points in the gyrA, gyrB, parC and parE genes.

DNA extraction and quantification
The genomic DNA extraction methods followed Campioni and Falcão [29]. The quality of the DNAs were checked using NanoDrop 1000 (Thermo Scientific, Rockford, IL), and the concentrations were determined using Qubit double-stranded DNA BR assay kit and Qubit fluorometer (Life Technologies, Grand Island, NY) according to each manufacturer's instructions.

Genome sequencing, assembly, and annotation
All isolates were prepared using the Nextera Sample Preparation Kit (Illumina, San Diego, CA) and then sequenced on Illumina NextSeq (Illumina) for 2 x 151 cycles. De novo assemblies were generated from all raw sequence data. The Illumina reads were assembled with SPAdes 3.0 with the following parameters: only contigs of length !500 bp were included; mismatch (MM) 3.28; the genome fraction was 96.157; and number of mis-assemblies (MA) was 2 [30]. The contigs for each isolate (draft genome) were annotated using NCBI's Prokaryotic Genomes Automatic Annotation Pipeline (PGAAP) [31]. The draft genome sequences of S. Typhimurium strains are publicly available in GenBank, with accession numbers listed in S1 Table. The presence of resistance genes, as well as points mutation in the QRDR of the gyrA, gyrB, parC, and parE genes, were determined using ResFinder (Center for Genomic Epidemiology, https://cge.cbs.dtu.dk/services/ResFinder/) with settings of threshold of 90%, and minimum length of 60% [32].

Antimicrobial susceptibility testing
Antimicrobial susceptibility of the 90 S. Typhimurium strains were tested by the disc diffusion method of the Clinical and Laboratory Standards Institute (CLSI) [33]. The majority of these results were previously published in Almeida et al. (2015) for 12 antimicrobials including: cefotaxime; cefoxitin; ceftazidime; aztreonam; cefepime; amoxicillin-clavulanic acid; ampicillin; nalidixic acid; levofloxacin; trimethoprim-sulfamethoxazole; chloramphenicol; and ciprofloxacin (Oxoid). However, five additional antimicrobials were tested in this study including: gentamicin; streptomycin; tetracycline; sulfonamides; and ceftriaxone. Additionally, the minimum inhibitory concentrations (MIC) of fluoroquinolones in the nalidixic acid resistant and susceptible strains were evaluated using Etest1 following the Clinical and Laboratory Standards Institute (CLSI) guidelines. Strains with MIC 0.06 μg/mL were considered sensitive and ! 1 μg/mL resistant.

Phylogenetic analysis
In addition to the 90 S. Typhimurium strains sequenced in this study, four additional S. Typhimurium strains (the sequencing reads were downloaded from NCBI with run accessions of SRR1060710, SRR1963606, SRR6325339, and ERR1556230 for strain DT104, LT2, 14028s, and SL1344, respectively) were added into the phylogenetic analysis for diversity purpose. The genomic analysis was performed using the CFSAN SNP Pipeline that generated the SNP matrix, which was then used to infer the maximum likelihood tree using GARLI [34] with 200 maximum likelihood replicates and 1000 bootstrap iterations. Three samples were included as outgroups including: Salmonella enterica serovar Saintpaul CFSAN000611; Salmonella enterica serovar Saintpaul CFSAN000614; and Salmonella enterica serovar Heidelberg CFSAN000443 [35]. The SNP matrix included 59,130 and 11,176 SNPs, with or without the three outgroups sample, respectively.
Macrolide resistance genes. Only one macrolide resistant gene (mphA) was detected in one food isolate.

Antimicrobial susceptibility testing
A total of 65 (72.2%) out of 90 S. Typhimurium strains studied were resistant to sulfonamides, 44 (48.9%) strains were streptomycin resistant, 27 (30%) strains were resistant to tetracycline, 21 (23.3%) strains were gentamicin resistant, and 7 (7.8%) strains were resistant to ceftriaxone. In our previously published paper (26), 34 strains were resistant to nalidixic acid (Nal R ). In this study we evaluated the reduced susceptibility to fluoroquinolones of 34 strains Nal R and 12 strains susceptible to nalidixic acid (Nal S ). All the 12 Nal S strains and 21 Nal R strains studied were sensitive to ciprofloxacin (MIC 0.06 μg/ml), whereas 11 Nal R strains presented intermediate resistance to this drug (MIC 0.12-0.5 μg/ml) and two Nal R strains were resistant to ciprofloxacin. All the antimicrobial susceptibility test results were presented in Table 1. gyrA, gyrB, parC and parE genes and of the  presence of qnr, qepA, oqxAB and aac(6')

-Ib-cr genes
A total of 33 (36.7%) out of 90 strains studied presented mutation points in the gyrA gene, with all being resistant to nalidixic acid ( Table 2). The nonsynonymous points of mutation in the gyrA gene included: aspartate/glycine, Asp(87)!Gly in 21 strains; aspartate/asparagine, Asp (87)!Asn in 7 strains; serine/tyrosine, Ser(83)!Tyr in 4 strains; and serine/phenylalanine, Ser(83)!Phe in one strain. None of the strains had more than one mutation point ( Table 2). One strain (5934/06 isolated from swine) Nal R did not show mutation in the gyrA gene. Seven (7.8%) strains presented synonymous nucleotide mutation, and these strains were Nal S (data not shown) suggesting undiscovered mutations. Thirty-two (35.6%) strains presented synonymous nucleotide mutation in the parC gene and 10 of those strains were Nal R with, two strains resistant to ciprofloxacin (data not shown). No strains presented mutations in the parE gene.
The qnrB88 gene was found in 1 (1.1%) Brazilian strain that previously had been reported both in Klebsiella pneumoniae (GenBank: KX118608) and under another gene (qnrE1) found in Klebsiella pneumonia (GenBank: KY781949). Additionally, one strain had the qnrB2 gene present in Salmonella Bredeney (GenBank: FJ844401). The oqxAB gene was found in 4 (4.4%) strains. However, these genes diverged in having 6 mutations compared to the oqxAB of Salmonella Derby (GenBank: FN811184). The aac(6')Ib-cr gene was identified in 5 strains isolated from humans.

Phylogenetic analysis
The 90 S. Typhimurium strains studied were distributed into 2 major clades (designated A and B, Fig 1).

Discussion
In this study 90 S. Typhimurium strains isolated from food and humans in Brazil were sequenced by next generation sequencing technology to evaluate their antimicrobial resistance Table 2. Quinolone resistance profiles of the 90 Salmonella Typhimurium strains studied isolated from humans and food in various States between 1983 and 2013 in Brazil.

CFSAN n˚Isolate Name CIP E-test QRDRs mutations
gyrA mutation gyrB mutation parC mutation parE mutation  Whole genome sequencing of Salmonella Typhimurium in Brazil gene profiles and phylogenetic diversity. This is the first study of S. Typhimurium strains isolated in Brazil that used WGS to access the genetic diversity and the molecular bases of antimicrobial resistance. In previous studies, the same strains were typed by PFGE, MLVA, ERIC-PCR, CRISPR-MVLST and MLST [26][27][28].
In this study, 47 (52.2%) strains presented phenotypic resistance to gentamicin and/or streptomycin. Streptomycin is not frequently used to treat Salmonella enterica infections; but, it has been commonly used as a growth promoter in food-producing animals and for this reason may serve as a marker for resistant strains moving through the food supply [11].
Our results confirm McDermott et al's. [11] observations of discrepancies between phenotypic resistance and genotypic resistance of aminoglycoside resistant genes. We observed 35 isolates carrying streptomycin resistance genes, but these isolates were phenotypically susceptible to the drugs. It is unclear why the genes while present in the genomes were not expressed to provide phenotypic resistance. Presence of the known streptomycin resistance genes does not predict phenotypic resistance well for this class.
The tetracycline resistance genes were found in 32 (35.5%) strains. Interestingly, 2 strains that were phenotypically resistant to tetracycline did not present any known tetracycline resistance genes suggesting a possible alternative mode of resistance. In contrast, seven strains that presented tetracycline resistance genes were phenotypically susceptible. Of these seven, six strains had two tetracycline resistance genes and one strain had only one tetracycline resistance gene. Tetracycline has been used commonly as an antibiotic in swine husbandry [36]. Brazil is a major producer of pigs with 3.73 million tons of pork produced and exported in 2016 [37,38]. The Salmonella Typhimurium serovar usually does not cause severe disease in pigs and sometimes it is asymptomatic in these animals, which may be a serious public health problem, since it may be an important source of contamination of carcasses in slaughterhouses. In addition, the contamination by S. Typhimurium may not be detected while the pigs are on the farm, which may eventually lead to human contamination [36,39].
Cefoxitin resistance has been used to indicate certain types of beta-lactamases production by Salmonella and E. coli. First and second-generation cephalosporin susceptibility results are not reported in clinical medicine for Salmonella, because the drugs may appear active in vitro, but are not therapeutically effective [33]. Regarding the beta-lactam resistance genes found in Brazil, the most common was bla TEM-1B gene presented in 16 (17.8%) isolates (6 humans, 10 foods). The bla TEM-1B gene has been associated with ampicillin resistance and 32 (35.6%) strains were phenotypically resistant to the ampicillin. The bla CTX-M-8 and bla CTX-M-2 genes have been more closely associated to cephalosporin resistance and 7 strains were resistant to ceftriaxone (CRO), third generation cephalosporin, but only 3 strains presented a bla CTX allele. The most common resistant gene was aac(6')Ib-cr found in 5 (5.6%) human isolates followed by oqxA and oqxB found in 4 (4.4%) food isolates. The qnrB2 and qnrB88 genes were found each in 1 (1.1%) food isolate.
Some of the discrepancies observed when a resistance gene is present but no phenotypic resistance in bacterial growth is observed, or when the phenotype is present but no known resistance gene is observed, is likely due to new unidentified resistance genes or mutations conferring resistance in undiscovered genes. Therefore, it is important to study any discrepancy as each represents new ways that bacteria are acquiring resistance as was reported for a new mechanisms discovered for Campylobacter gentamicin resistance [40]. Pribul et al. [41]  https://doi.org/10.1371/journal.pone.0201882.g001 evaluated the prevalence of PMRQ genes in 129 isolates of non-typhoidal Salmonella from Brazil by PCR amplification. Qnr genes were found in 15 (11.6%) isolates (8 qnrS, 6 qnrB, and 1 qnrD), and the aac(6 0 )-Ib gene was found in 23 (17.8%) isolates. Regarding mutation points in the QRDRs, gyrA mutation was the only one found among the strains studied. Thirty-three (36.7%) of nalidixic acid resistant strains presented mutations in the gyrA gene (22 human, 11 foods).
McDermott and colleagues [11] used WGS technology to identify known antimicrobial resistance genes among 640 non-typhoidal Salmonella strains for 43 different serotypes and correlated these with susceptibility phenotypes to evaluate the utility of WGS for antimicrobial resistance surveillance. Overall, genotypic and phenotypic resistance correlated in 99.0% of the cases. They concluded that WGS is an effective tool for predicting antibiotic resistance in nontyphoidal Salmonella [11]. Regarding QRDR mutations and PMQR genes, 21 isolates had either QRDR mutations or PMQR genes, all of which were from human clinical cases. In contrast, in this study QRDR mutations were found in both human and food isolates.
Food isolates were distributed in Clades A and B in relatively similar numbers suggesting that there is more than one subtype in circulation, in foods in Brazil. Human's isolates were more prevalent in the Clade B suggesting the existence of a prevalent subtype. Genomic and phenotypic testing results suggest clinical strains isolated before the mid-1990s presented more antimicrobial resistance compared to later strains. The diversity and prevalence of resistant genes found in Brazilian Salmonella Typhimurium is an alert of their potential hazard for food safety and public health.
Supporting information S1