Introduction

Multidrug-resistant Salmonella enterica serotype 4,[5],12:i:-, a monophasic variant of S. enterica serovar Typhimurium, has become more prevalent and widespread in the last two decades. Colistin is the last resort antibiotic against the multidrug-resistant bacteria. Recently, mobile colistin resistance (mcr) genes have been identified in many species of Enterobacteriaceae family including Salmonella enterica, particularly S. 4,[5],12:i:-1. Reports from England2, Denmark3, Canada4, and USA5 revealed that several infected patients had traveled to China or Southeast Asian countries, including Thailand. This suggested a circulation of the colistin resistant S. 4,[5],12,i:- strains in the region, which might occasionally be transmitted internationally.

There are 10 mcr genes, designated as mcr-1 to mcr-10, all encoding phosphoethanolamine transferases, which modify lipid A at the outer membrane of Gram-negative bacteria. Several mcr genes, including mcr-1, -3, -6, -7, -8, and -9 have been reported in Escherichia coli in Thailand with mcr-1 being the most common, followed by mcr-96. In addition to E. coli7,8 mcr-1 was also reported from several other species of Enterobacteriaceae in Thailand9,10 including Klebsiella11, Edwardsiella12 and recently S. enterica serovar Cannstatt13. mcr-3 has been reported in Aeromonas veronii14, E. coli15, K. pneumoniae13,16, Enterobacter17 and recently S. enterica serovar 4,[5],12:i:-18,19,20,21,22.mcr-3 shares 45% homology to mcr-1. It is usually located upstream of diacylglycerol kinase gene (dgkA) and flanked by a truncated and intact ISKpn40. The ΔISKpn40-mcr3-dgkA-ISkpn40 segment is flanked by IS26 and IS15DI in IncHI2 or pWJ1 plasmid23. In IncA/C2 or IncHI2A/IncY plasmid, the fragment is flanked by two IS15DI elements2.

We screened for the presence of the mcr genes in whole genome sequences (WGS) of 53 isolates of S. 4,[5],12,i:- in Thailand and identified three isolates with mcr-3 but none of the other mcr genes. In order to gain more understanding on the transmission dynamics of mcr-3 mediated resistance, we compared the genomes of these three isolates with the available genomic sequences of other mcr-3-carrying isolates of S. 4,[5],12,i:- deposited in public databases. The analysis revealed that mcr-3 genes were carried by several plasmids and the similarity of plasmid profiles conformed partially with the core genome phylogeny. The findings conformed to the notion that some European patients might be infected by S. 4,[5],12,i:- in Thailand and suggest that the deletion of fljAB-hin has been mediated by recombination of IS26.

Results

Three samples of S. 4,[5],12:i:-, all isolated in 2010, harbored mcr-3.1. Two were from human stool and the other was from ready-to-eat frozen food. Phenotypic characterization confirmed that all three isolates were resistant to colistin, all with MIC and MBC of 8 µg/ml.

WGS of H1-012, H1-014 and H1-120 were analyzed together with the ones of 25 other isolates from several countries. WGS of all 28 isolates covered > 95% of the genome of the LT2 strain, as shown in Supplementary Table 1.

Long read sequence data statistics and the genome metrics of the H1-012, H1-014 and H1-120 are shown in Supplementary Table 2. Their chromosomal sizes ranged from 4,929,877 to 5,039,354 bp, and the GC content was approximately 52.1%.

Structural variations of the fljAB-hin regions

fljAB-hin region deletion is a hallmark of S. 4,[5],12:i:-. This region is generally replaced by drug resistance genes. We mapped the presence and absence of genes around the fljAB-hin region from STM2692 (hlyD) to STM2775 (iroD, enterochelin esterase) by blasting the gene sequences of the S. Typhimurium LT2 reference against the de novo assembled contigs of the short read sequences of each S. 4,[5],12:i:- isolate. The presence of some genes cannot be ascertained, including STM2768-STM2769, both probably encoding IS3 transposases. Their multiple presence in the genome might preclude accurate mapping. Many genes were clearly absent or present. However, the assembled contigs might also contain only some parts of the genes. We consider the results as ambiguous because the partial absence can be a result of incomplete assembly.

Of 28 isolates, mapping the fljAB-hin region of isolate 2008AR-0009 (Genbank Biosample Number SAMN07688908) from USA did not identify any deletions compared to S. Typhimurium LT2. Thus, the strain should still be considered as S. Typhimurium and was consequently excluded from further analysis. The other 27 isolates all lacked Fels-2 prophage genes (STM2693 to STM2739), present in S. Typhimurium LT2. Only STM2740, coding for phage integrase protein, consistently remained intact. Eight patterns of deletions were identified with the shortest one involving only fljAB-hin deletions while 70 genes were deleted in the longest one as shown in Supplementary Fig. 1. In all 27 isolates, the right borders of the deleted region were at the same position on the 5’ non-coding region of iroB side but variable on the other side. Identification of iroB revealed ambiguous results in three cases, though.

Core genome phylogeny, fljAB-hin deletion and presence of plasmid replicons.

The single nucleotide variants (SNVs) in the core genome of 27 isolates were identified using S. Typhimurium LT2 as reference. A total of 369 of core genome SNVs were identified. All isolates belonged to ST34, with the SNV differences from S. Typhimurium LT2 ranging from 16–46 SNVs (average 31) as shown in Supplementary Table 3. Their pairwise SNV distances ranged from 2–79 with the average pairwise SNV distance of 45.1 and the median of 45. A maximum-likelihood phylogenetic tree was constructed and shown in Fig. 1. The bootstrap scores were calculated, and the bootstrap score of 100 was used to define a clade. The bacteria were separated into three clades.

Figure 1
figure 1

A core genome SNV ML phylogenetic tree of 27 mcr-3 carrying S. 4,[5].12:i:- isolates. The names of the samples sequenced in this study are colored in red. The country of origin, the isolation sources, the existing history of travel to Southeast Asia (SEA), the isolation year and the type of mcr are presented by the panel of colored squares on the right. The black circle represents the clade that has the bootstrap value of 100 and the gray circle indicates the bootstrap value of more than 80.

Full size image

The first clade comprised eight isolates from Denmark (3), UK (2), USA (1) and two isolates sequenced here (H1-012 and H1-014). All isolates were from human sources, except for the one from the USA whose source was unknown. Six isolates carried mcr-3.1 together with IncA/C plasmid replicon while the other two carried mcr-3.24 with IncA/C and mcr-3.20 with IncFII and IncX1 (Supplementary Table 4). Mapping the fljAB region revealed that all had the deletion of fljA-fljB (STM2770-2771) and hin (STM2772) with the conservation of iroB (Supplementary Fig. 1b). However, the 5’ ends of the deletions were distinct except for two pairs of isolates, which also had relatively small SNV distances. The first pair from the UK (712,472 and 717,816) had a pairwise distance of only two and both shared a large deletion from STM2703 to hin. The second pair of isolates from Thailand (H1-012) and Denmark (SSI-AC258) had a pairwise distance of 11. Both isolates similarly had complete absence of only fljAB and hin. SSI-AC258 might also lose a part of STM2769 upstream to fljA, though. These results indicated that the two UK isolates were epidemiologically linked and the Denmark isolate was probably the same clone as H1-012, which had been isolated a year earlier.

The second clade comprised three Japanese cattle isolates, all carrying IncFIB plasmid. Isolates L-4445 and L-4567 were close to each other, with the pairwise SNV distance being 14 while L-4605 was 26 and 41 SNVs distant from the first pair, respectively (Supplementary Table 3). All three had an identical fljAB deletion starting from STM2753 to hin, which was different from all other isolates in this study (Supplementary Fig. 1b).

The third clade included seven isolates from Denmark (1), UK (4), USA (1) and Australia (1), seven isolates from swine or swine production environment in Thailand and the food isolate sequenced here (H1-120) (Fig. 1). Interestingly, in contrast to the first clade, all isolates in this clade shared the same deletion from STM2760 to hin (Supplementary Fig. 1). The early branching isolates in the tree tend to have IncA/C plasmid while eight of the nine isolates in the terminal branch carry IncFII (Fig. 1). The latter included all the seven swine isolates from Thailand, isolated in 2011–2014. They were distributed into two clades with average pairwise SNV distances of 10.2 and 8.33 (Supplementary Table 3). Their average interclade pairwise SNV distance was 23.6. They were probably parts of a bigger outbreak in swine. The terminal branch also included two human IncFII-carrying isolates from Denmark and the UK, isolated in 2011 and 2017 respectively. The Denmark (SSI-AC259) and the UK (445,956) isolates were 13 and 17 SNVs different from the closest swine isolates respectively, which were shorter than the average interclade differences (Supplementary Table 3). SSI-AC259 was isolated from a patient with a travel history to Thailand in 20113, indicating the epidemiological linkage of SSI-AC259 and the swine outbreaks. The UK isolate, isolated much later in 2017, had a different deletion profile upstream the deleted STM2760-hin segment and was, therefore, more remotely linked to the swine outbreak (Supplementary Fig. 1).

Pangenome analysis revealed a total of 3,063 hard-core and 950 soft-core genes shard by 27 isolates as well as 1,575 accessory genes (Supplementary Fig. 2a). The number of shared genes in each clade is higher than core genes as shown in (Supplementary Fig. 2b–d).

Comparing the fljAB-hin deletion patterns with the SNV distances reveals that a pair of isolates with SNVs distance less than 11 always had the same fljAB-hin deleted segment while a pair of isolates with SNVs distances more than 55 always had different fljAB-hin deletion patterns as shown in Supplementary Fig. 1b. Summary of the fljAB-hin deletion patterns of the 27 mcr-3 carrying S. 4,[5],12:i:- isolates is shown in Table 1. Overall, the data indicated that the isolates with genetically related core genomes and similar fljAB-hin region deletion profiles were likely to share the same mcr-3.1 carrying plasmids as illustrated in Fig. 1.

Table 1 Summary of the fljAB-hin deletion patterns of the 27 mcr-3 carrying S. 4,[5],12:i:- isolates.
Full size table

fljAB-hin regions of H1-012, H1-014 and H1-120

The fljAB-hin deleted regions of H1-012, H1-014 and H1-120 were investigated using the hybrid-assembled sequences. Lack of the Fels-2 prophage (STM2693 to STM2739) was confirmed. The deleted region of H1-012, which was the shortest, was 5053 bp long corresponding to position 2,910,987–2,916,039 in the S. Typhimurium LT2 genome, which included only fljA, fljB and hin. The deleted region was replaced by an 820-bp-long segment carrying a transposase, with a pair of an inverted repeat (GGCACTGTTGCAAA) (Supplementary Fig. 3) at the ends on both sides, identified as IS26 (IS6 family). Nevertheless, there was no drug resistance gene inserted in the region (Fig. 2). There was an additional insertion of 1,213-bp-long ISKpn40 (IS3 family) at position 2,901,703 immediately downstream STM2761. The copy of ISKpn40 was flanked by a direct repeat CCGG.

Figure 2
figure 2

Comparative gene maps of S. Typhimurium LT2 (upper), S. 4,[5],12:i:- H1-012 (middle) and H1-014 (lower). H1-012 is characterized by deletion of the fljAB-hin segment with the replacement of IS26 and insertion of ISKpn40. H1-014 had a larger deletion and insertion of a few resistance genes, flanked on both ends by IS26, the 3’ one exactly at the same position as H1-012.

Full size image

The deleted region of H1-014 was 17,540 bp long corresponding to the position 2,898,500–2,916,039 in the S. Typhimurium LT2 genome, which included 14 genes, starting from nucleotide 772 of STM2759 (losing 909 nucleotides) to hin. The deleted segment was replaced by a 14,524-bp-long common DNA fragment containing repA, repC, sul2, aph(3″)-Ib, and aph(6)-Id, flanked by two copies of IS26 (Fig. 2). The 3’ copy of the IS26 of H1-014 resided at exactly the same nucleotide position as the one in H1-012. Moreover, a replicon sequence of IncQ1 was identified in the inserted segment.

The 15,715-bp-long deleted region of H1-120 included 13 genes from STM2760 to hin (position 2,900,325–2,916,039), which is different from H1-014, as it did not include a part of STM2759. Interestingly the deleted regions of all three isolates ended at the same nucleotide position. The deletion in H1-120 was accompanied by an inversion of a 150,626-bp-long segment (position 2,916,040–3,066,666), including 152 genes from STM2773 (iroB) to STM2924 (rpoS) and the last 111 bp of STM2925c (nlpD). The inverted segment was flanked on both sides by IS26. There were two inserted resistance gene regions on both sides of the large inverted region, between STM2759 and the last 111 bp of STM2925, as well as between iroB and the remaining part of STM2925 (nlpD) as shown in Fig. 3. The former segment contained only blaTEM-1, flanked by two IS26. The latter inserted region, also flanked on both sides by IS26, comprised two parts, separated by another copy of IS26. The major part contained several metabolic genes, two drug resistance genes (tetR(B) and tet) and a mer operon (merR, merT, merP, merC, merA, merD and merE). The metabolic genes co-integrated with the drug resistance gene included genes encoding DNA (cytosine-5-)-methyltransferase, thermonuclease, phospholipase D, LysR family transcriptional regulator, sodium/glutamate symporter, and antibiotic biosynthesis monooxygenase. The minor part was the aph(6)-Id-aph(3″)-Ib-sul2-repC-repA segment similar to the one found in H1-014 (Fig. 2).

Figure 3
figure 3

Comparative gene maps of S. Typhimurium LT2 and, S. 4,[5],12:i:- H1-120. H1-120 is characterized by deletion of the fljAB-hin segment and a large inversion of a DNA segment with insertions of two segments of multidrug resistance genes and mer genes (lower) which were flanked on both ends by IS26C.

Full size image

Comparison of the IS26 transposase sequences revealed that all copies of IS26 in H1-120 shared the same sequence, tentatively designated as IS26C. So was the 3’ copy of IS26 in H1-014. They were 2 bp different from the 5’ copy of H1-014 (designated as IS26B) and a bp different from the copy in H1-012 (IS26A) as shown in Supplementary Fig. 3 and Supplementary Table 5.

Comparison of plasmid sequences of mcr-3 carrying S. 4,[5],12:i:-

Plasmid replicons were identified from WGS data of all isolates using PlasmidFinder. Three and nine isolates carry plasmids IncFIB and IncFII, respectively. Eleven isolates, including H1-012 and H1-014, carry IncA/C ST3 plasmid while three isolates, including H1-120, carry IncHI2 plasmid (Fig. 1, Supplementary Table 4).

The plasmid sequences of H1-012, H1-014 and H1-120

To better understand the structural variation of mcr-3 carrying plasmids as well as their contributions to the spread of drug resistance, circular plasmid contigs were constructed from the combined Illumina/Nanopore sequencing data. The plasmids in H1-012, H1-014 and H1-120 were 180,560 bp, 179,719 bp and 222,107 bp long, respectively, and designated as pH1-012, pH1-014 and pH1-120.

The complete sequences of pH1-012 and pH1-014 were compared with the reference IncA/C-FII ST3 plasmid p16E08024 (180 kb, GenBank accession number: MN647788) isolated from a human sample in China in 2016 (Fig. 4). All three plasmids had comparable length and actually were hybrids of IncA/C and IncFII.

Figure 4
figure 4

Plasmid alignment of p16E080, pH1-014 and pH1-012. (a) Linear gene maps of p16E080, pH1-014 and pH1-012 plasmids starting from repB. The complete sequences of pH1-012 and pH1-014 were from the hybrid assemblies. The gene arrangement in the multidrug resistance and mcr regions were enlarged. (b) Short read mapping of other IncA/C carrying isolates in this study, using pH1-014 as reference. pH1-012 and the plasmid of SSI-AC258 were very similar.

Full size image

Using the repB as the first gene of the plasmids, all three plasmids contained two major parts. The segments on both sides of the repB genes (position 1–41,449 and 164,501–179,719 in pH1-014), which were parts of the original IncC sequence, were fairly conserved and had the same orientation. These segments contained three small drug resistance gene clusters namely, floR, tet(A)-tetR(A) and aph(6)-ld–aph(3″)-Ib–sul2.

The remaining parts of pH1-012 and pH1-014, which include the IncFII sequence segment and parts of IncC, were inverted while the orientation of the IncFII segment in H1-014 and p16E080 were the same. There were many small in/del, transposition and inversion, mainly in the drug resistance gene regions. The drug resistance genes in the region of all three plasmids were clustered into two subregions. The first subregions of H1-012 and H1-014 contain blaTEM-1, merD and merE flanked on both sides by IS15DI. Interestingly, the same region in p16E080 also contained blaCTX-M-55 in tandem to the blaTEM-1 segment and the entire set was duplicated in the same orientation.

The second subregions contained the mcr-3.1, accompanied by dgkA, ISKpn40 and flanked by IS15DI on both sides similar to the previous report2. There was aac(3)-IId and qnrS1 upstream to mcr-3.1 in all three plasmids. Interestingly, in H1-012 and H1-014, there is also blaCTX-M-55 downstream to qnrS1. However, their orientations of qnrS1-blaCTX-M-55 were inverted relative to mcr-3.1. Further upstream of mcr-3.1, there is catII in both H1-012 and H1-014 but not in p16E080. pslB is located distantly downstream of mcr-3.1 in both H1-012 and H1-014, with partially similar intervening sequences. However, pslB in p16E080 was split and present on both sides of this subregion. In short, pH1-012 and pH1-014 were similar albeit non-identical plasmids while p16E080 was more distantly related. blaCTX-M-55 was closer to the mcr-3.1 genes in pH1-012 and pH1-014.

The sequence of pH1-120 was compared to the reference pWJ1 plasmid, an IncHI2 plasmid, firstly reported as a mcr-3 containing plasmid from E. coli in China23. pH1-120 was shorter than pWJ1 with most genes in the same orientation (Fig. 5). The shorter pH1-120 was associated with the lack of two drug resistance gene regions (aac(6’)-lb-cr, blaOXA-1, catB3, arr3 and sul1; and, sul2 and floR) and a 14,095-bp-long segment which were duplicated in pWJ1 and had the same sequence as position 93 -14,032 in pH1-120. However, sul2 was present in the chromosome of H1-120 isolate (Figs. 3 and 5a).

Figure 5
figure 5

Plasmid alignment of pWJ1and pH1-120. (a) Linear gene maps of pWJ1and pH1-120 plasmids. pWJ1 was used as a reference for IncHI2 plasmids. Gene arrangement in multidrug resistance and mcr regions were enlarged. The order of genes is according to pWJ1 [23]. (b) Circular alignment of IncHI2 plasmids of pH1-120, p131681 and pWJ1. pH1-120 was used as a reference for short read mapping of other isolates.

Full size image

The genes surrounding mcr-3.1 in the IncHI2 plasmid were different from those of IncA/C plasmid. The mcr-3.1 was accompanied with dgkA, IsKpn40 and flanked by IS26 and IS15DI (Fig. 5a), similar to the previous report23.

IS26 plays an important role in disseminating antibiotic resistance genes. Three more variants of IS26 were identified in the plasmid, including IS15DI, which is only a bp different from IS26B. The other two were designated as IS26D and IS26E. The sequence relationship of all IS26 was summarized in Supplementary Fig. 4. IS15DI was present in all analyzed plasmids, particularly IncA/C (Figs. 4,5) but absent from fljAB-hin regions. IS26C was present not only in fljAB-hin region and pH1-014 and pH1-120 but also in p16E080 and pWJ1. IS26C has the same sequence as IS26 in Proteus vulgaris (Genbank accession number X00011.1).

To further explore the similarity between the plasmids, the short reads of the other IncA/C carrying isolates as well as p16E080 were mapped to the pH1-014 (Fig. 4b). The method did not allow the evaluation of the gene orders. All the IncA/C plasmids contained both the IncA/C and IncFII parts, similar to pH1-014. pH1-012 and the plasmid of SSI-AC258 were very similar. The mcr-3.1 and dgkA, flanked by IS15DI were present in all IncA/C plasmids, presumably in the IncFII part similar to pH1-014. The other antibiotic resistance genes present in all plasmids included qnrS1 and a group of antibiotic resistance genes composed of floR, tet(A), tetR(A), aph(6)-Id, aph(3″)-Ib and sul2 (Fig. 4b). blaCTX-M-55 and blaTEM-1 was present in 9 and 10 of the 12 plasmids respectively. The possible coexistence of mcr-3.1 and blaCTX-M-55 in the same plasmid conformed to the observations of the association between colistin resistance and extended spectrum beta-lactamase (ESBL)19.

In contrast to the isolates carrying IncFII, the ones carrying IncA/C had a variety of fljAB deletions, including four isolates with fljAB deletion similar to the ones carrying IncFII (Supplementary Fig. 1). This might signify the spreading potential of the IncA/C plasmids across various variant strains of S. 4,[5],12:i:-. Our results from de novo hybrid assembly of pH1-012 and pH1-014 show that in our IncA/C plasmids contain the replicon of IncFII. Both IncA/C and IncFII replicons contain their own tra region (Fig. 4a) which make the IncA/C plasmids have two sets of the tra regions. These might increase the chance of horizontal transmission.

Three isolates carried IncHI2 but did not form a clade in the phylogenetic tree. H1-120 carries a multi-replicon plasmid IncHI2/IncHI2A, belonging to ST3, similar to the plasmid pWJ1 (261 Kb, GenBank accession number KY92492823), originally identified as mcr-3 carrying plasmid in E. coli [23] (Fig. 5). A different IncHI2/IncHI2A plasmid, belonging to ST2, was identified to carry mcr-3.11 in an isolate from the UK (142,703) (Supplementary Table 4). The last isolate carrying only IncHI2 was from a swine in Thailand (A543010) and is phylogenetically close to the IncFII carrying isolates. Figure 5b shows short read mapping by using H1-120 as a reference. The isolates UK 142,703 (UK) and A543010 (Thailand) are not included in the figure because there belonged to different ST and had different plasmid replicon. p131681 (202 kb) was isolated from a patient in Canada, in April 20134. He had traveled to Thailand a month prior to the isolation of the bacterium. p131681 was include in the figure because it was identified as IncHI2/IncHI2A plasmid, ST3. From mapping, p131681 is the smallest plasmid.

Fourteen isolates carried IncQ1 replicon. IncQ1 is a small plasmid which has never been reported to carry mcr-3. As the IncQ1 replicon is identified in the chromosomes of H1-014 and H1-120, it is not clear whether there really was the IncQ1 plasmid in the other isolates. Nevertheless, it should be noted that there was an mcr-3 carrying isolate that did not carry any other plasmids apart from IncQ1.

Antimicrobial resistance (AMR) gene search

The AMR genes of H1-012, H1-014 and H1-120 were identified by ResFinder and found in both chromosomes and plasmids. Twenty antibiotic resistance genes (6 genes in the chromosomes and 18 genes in the plasmids) related to resistance to eight antibiotic classes were found in the isolates. Five and three genes were related to resistance to aminoglycosides beta-lactams (aac(6’)-Iaa, aac(3)-Iid, aph(3’)-Ia, aph(3’’)-Ib and aph(6)-Id) and beta-lactamase (blaCTX-M-55, blaTEM-1B, and blaTEM-216.) respectively. Most antibiotic resistance genes in H1-012 (10 genes), H1-014 (9 genes) and H1-120 (11 genes) were found more in the plasmids than in the chromosomes, where there were 1, 5 and 6 genes, respectively (Supplementary Table 6). Most of antibiotic resistance genes in the chromosome were found in the fljAB-hin deletion region. Only aac(6')-Iaa was found outside the region in all three isolates, between genes coding for Doer/GlpR family DNA-binding transcription and lactate oxidase.

We compared the antibiotic resistance phenotype (AMRP) and predicted antibiotic resistance phenotype (PAMRP) from ResFinder. The result shows that all AMRP of the three isolates agreed with PAMRP. (Supplementary Table 7).

Discussion

S. 4,[5],12,:i:- is multidrug resistant with increasing report of colistin resistance1,19. Here we reported the complete genomes of three colistin resistant S. 4,[5],12:i:- and investigated their core genome phylogeny, fljAB-hin deletion profiles and plasmids in comparison with WGS of 24 other isolates of mcr-3 carrying S. 4,[5],12,:i:- deposited in NCBI. All isolates belonged to ST34 with the differences of the core genomes not more than 71 SNVs, but had variable fljAB-hin deletions. The similarity of the deletion profiles and plasmids correlated with the shorter SNV distances between isolates. This suggests that combination of the information helps recognizing epidemiologically linked isolates.

Transmission of S. 4,[5],12:i:- is evidenced by the presence of a few pairs or groups of isolates with a small numbers of SNV differences, the identical fljAB-hin deletions and the identical mcr-3 carrying plasmids replicon types, indicating epidemiological linkages within each pair or group. The bacterial isolates of a number of patients in developed countries who had a history of travel to Thailand were closely related to isolates from swine or food suggesting the transmission mechanism as food-borne24.

This study illustrates the usefulness of open genomic repository for identifying possible transmission routes of Salmonella. In order to do so, a SNV difference cutoff may be needed to infer which pairs of isolates were epidemiologically related. Various cutoffs for Salmonella enterica were proposed25. A previous study on outbreaks of S. Typhimurium in Australia revealed that most of the outbreak isolates had not more than two SNV differences26. This is a very strict criterion compared to other bacteria or other serovars of Salmonella and probably suitable for local outbreaks but not for identification of international transmission. The study here indicated that a more relaxed SNP cutoff criterion of 11 may be useful for screening for possibly epidemiological linkage of S. 4,[5],12,:i:- pairs. The suspected linkage can be further strengthened by the similarity of the fljAB-hin deletion profiles and plasmid replicon types. In reverse, the fact that core-genome related isolates can harbor different plasmids as shown here demonstrates both the plasticity of the plasmids as vehicles for transferring drug resistance as well as the limitation of the plasmid profiling for inferring transmission.

S. 4,[5],12:i:- is characterized by the deletion of the fljAB-hin region, usually replaced by drug resistance gene cassettes. Nevertheless, the complete genome sequence of H1-012 demonstrated that the replacement by drug resistance genes is not mandatory. The replacement of fljAB-hin of H1-012 by a single copy of IS26, which is in the same genomic position as the other sequenced isolates supported the previous suggestion that the IS26 transposition was an initial step for the development of the monophasic S. Typhimurium27. The fact that the IS26 in H1-012 inserted at the same nucleotide position as the 3’ copy of IS26 in H1-014 and H1-120 indicates that they descended from an ancestor harboring that particular copy.

The absence of identical short direct repeat at both ends of the IS26 in H1-012 suggests that the copy was a result of recombination of two original IS26 on both sides of the fljAB-hin region. We hypothesize that the variation of fljAB-hin deletions are the results of the insertion of the second copy of IS26 at different positions upstream of the fljAB resulted in the deletions of different sizes after homologous recombination. The multiple events of IS26 recombination resulting in the deletion of fljAB-hin suggest an evolutionary benefit for the loss of the fljAB-hin genes, which results in the loss of phase 2 flagella. The appearance of the antibiotic resistance genes was likely to be a result of subsequent recombination with a resistance gene cassette flanked by IS26.

Colistin resistance is a serious problem28. A recent study indicated that the majority of mcr in Salmonella was mcr-3, which was found in 4 of 26 investigated S. 4,[5],12:i:- in Thailand19. It was also reported that mcr-3 sometimes co-propagate with an ESBL blaCTX-M-5519. Here we confirmed that both genes co-existed in the IncFII part of the same conjugable IncA/C-FII hybrid plasmid. This is worrisome as the propagation of both important resistance genes may occur easily among many enteric bacterial species.

The fact that isolates that had closer core genome phylogenetic relationship and similar fljAB-hin deletion profiles tended to carry similar plasmids suggests that the spread of mobile colistin resistance in S. 4,[5],12:i:- is more attributable to the spread of the host strains than by the spread of plasmids. Nevertheless, several plasmid replicons involved in the transmission of mcr-3.1 warrant further surveillance of the plasmids in other Enterobacteriaceae.

It should be noted that the first reported plasmid replicon carrying mcr-3.1, IncHI2, contributed only to the minority of isolates in this study. In contrast to mcr-1, the data here, therefore, do not support the hypothesis that IncHI2 might be a more efficient vessel to disseminate mcr genes than other plasmids29.

In addition to IncHI2/IncHI2A, originally reported to carry mcr-3.1, three other plasmids namely IncC, IncFII, and IncFIB were also implicated. Interestingly, mcr-3 has not been reported from other common Enterobacteriaceae such as E. coli or Klebsiella in Thailand where the mcr genes in E. coli are carried mostly by IncX4 and IncI213. The origins of the mcr-3 carrying plasmids in Salmonella in Thailand are not clear and remain to be further studied.

Conclusion

This study provided the evidences supporting the international transmission of mcr-3-carrying isolates of S. 4,[5],12:i:-. This demonstrates the benefits of a global genomic database of Salmonella, which would be an important tool for identifying and hopefully preventing the international transmission of Salmonella. In this study we also found mcr-3 in IncA/C plasmid more often than the other groups of plasmids which is different from mcr-1 gene.

The complete genome information suggests the important role of IS26, including the closely related IS15DI, in the antibiotic resistance gene exchange among bacterial chromosomes and plasmids, as suggested from the distribution of multiple copies of IS26 in the area around multidrug resistance genes. Many inversions of gene-order were observed due to the presence of insertion sequences suggesting that only data from short read sequencing might not be optimal to study the evolution of drug resistance genes.

Materials and methods

Samples

Twenty-two human isolates and 31 non-human isolates of Thai S. 4,[5],12,i:- collected during 2009–2012 were previously sequenced30. The presence of mcr genes were identified using ResFinder31. Among all mcr genes, only mcr-3 was identified in three isolates i.e., H1-012, H1-014 and H1-120. The three samples were additionally sequenced by Oxford Nanopore.

We also included available short read data of mcr-3 harboring S. 4,5,[12]:i:- isolates that were already published before 2020 or available in the Pathogen Detection database (https://www.ncbi.nlm.nih.gov/pathogens/). Seven isolates were from swine farms in Northern Thailand32. The origins of the other samples were Australia (1), Denmark (4), Japan (3), Thailand (1), United Kingdom (6) and USA (3). The information regarding the samples was shown in Supplementary Table 8.

Phenotypic and genetic profiles of the three isolates

H1-012 and H1-014 isolates were isolated from humans in 2010 and had similar antimicrobial resistance (AMR) patterns i.e., resistant to ampicillin (Amp), cefotaxime (Ctx), chloramphenicol (C), ciprofloxacin (Cp), streptomycin (S) and tetracycline (T)33. H1-120 was isolated from frozen food in 2010 and demonstrated resistance to Amp, C, S, T, and sulfamethoxazole/trimethoprim (Sxt)33. Genetic studies of the three isolates were previously reported30,33. Their PCR patterns indicated the absence of fljAB-hin region. The PFGE patterns of H1-012 and H1-014 were the same but different from that of H1-12033.

Antibiotic susceptibility testing by microdilution method

Colistin resistance phenotypes in the three isolates were confirmed by determining the minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) using a broth microdilution method in cation-adjusted Muller Hinton broth (CA-MHB). Eight concentrations (0.5, 1, 2, 4, 8, 16, 32 and 64 µg/ml) of colistin were used34.

WGS analysis

Short read sequencing

The short read sequencing was previously done30. The accession number and the numbers of reads of all selected isolates were shown in Supplementary Table 8.

For long read sequencing, Salmonella isolates were cultured in Trypticase Soy broth (TSB) at 37 °C for 16 h. Genomic DNA was extracted using the QIAmp DNA mini kit (Qiagen, Hilden, Germany). The DNA quality and quantities were assessed with Nanodrop DenoVix® and the Qubit 3.0 fluorometer. The total input of DNA was about 400 ng for each flow cell. Separation and determination of the sizes of linear DNA fragments were carried out by electrophoresis through submerged 1.0% horizontal agarose gels.

The genomic DNA was ligated according to the manufacturer’s instructions. Libraries were sequenced with qualified FLO-MIN106 flow cells (R9.4.1, active pore number > 800) for approximately 72 h on MinION (Oxford Nanopore Technologies, Oxford, UK). Base-calling was performed in real time using Guppy with a base calling model modified for 6 mA dam/5 mC dcm and CpG, which was integrated in the MinKNOW software v3.5.40 installed on MinION. The long read sequencing data of H1-012, H1-014, and H1-120 are under BioProject accession numbers PRJNA675488 and PRJNA808666.

SNV calling and core genome phylogeny

The fastq files of all 27 isolates were quality-trimmed using Trimmomatic v0.3935 with the following parameters: sliding-window trimming with a window size of 4 with read quality threshold of 30 and minimum read length of 70 bp. The remaining high-quality reads were mapped to the complete genome of S. Typhimurium LT2 (GenBank accession number: NC_003197.2) using BWA program36 with the mem algorithm, skipping seed (-c) 100, marking split hits as secondary mapping (-M), and only reporting read with a minimum 50 score (-T 50). Per-sample single nucleotide variants (SNVs) were called using GATK HaplotypeCaller v4.1.6.037 with a base quality score >  = 20 and haploid model. Joint SNV calling of all 27 isolates was performed using GATK GenotypeGVCFs v4.1.6.0. A single nucleotide variant (SNV) alignment was made and filtered by using VariantFiltration and SelectVariants functions (GATK) with quality by depth (QD) >  = 2 or root mean square of mapping quality (MQ) >  = 40. The SNV positions having less than 50% calling bases from the whole population were excluded. This resulted in 1273 SNVs across 27 genomes.

The SNV alignment was filtered with the core genome position list38 resulting in 490 SNV sites. Potential recombinant regions within the alignment were checked using Recombination Detection Program 439. The recombination regions detected by more than four programs were excluded. Three hundred and sixty-nine nucleotides remained after the removal of the recombination regions.

A maximum-likelihood (ML) phylogenetic tree of 27 isolates was inferred using IQ-TREE v240 with ultrafast bootstrap supports from 1000 replications. The best-fit nucleotide model was determined to be K3P + ASC + R2 under the Bayesian information criterion by ModelFinder41. The S. Typhimurium LT2 (GenBank accession number: NC_003197.2) was used as an outgroup for rooting the tree, which was visualized with Figtree v1.43 (http://tree.bio.ed.ac.uk/software/figtree/). Pairwise SNV distances were calculated using MEGA742.

Short read assembly

De novo short read assembly was performed using Unicycler v.0.4.8.043. The Illumina short reads were quality checked using FastQC v0.11.944 and trimmed using Trimmomatic v.0.3835 with average quality per read >  = 30.

Hybrid de novo assembly and annotation

De novo hybrid assembly was done using the approach of Unicycler v0.4.8.043. The raw reads of Illumina were quality checked using FastQC v0.11.944 and then trimmed using Trimmomatic v0.38.035. The mean read quality of the raw long reads of MinION was scored using Nanoplot v1.0.045. The adapters on the end of the raw reads were trimmed using Porechop v0.2.446. Trimmed reads with the minimum length of 1000 bp were used for subsequent assembly using Filtlong V.0.2.047. The complete genomes were annotated using Prokaryotic Genome Annotation Pipeline (PGAP) by NCBI48. The de novo hybrid assembly of each isolate resulted in two circular contigs: a chromosome and a plasmid.

Plasmid identification and comparison

Plasmid replicon types were analyzed using PlasmidFinder v2.1 online tool49,50,51. Plasmid comparisons were performed using Easyfig 2.2.5 program52.

Genome identification and drug-resistance detection

Multi-locus sequence typing (MLST) and plasmid MLST (pMLST) were identified using PubMLST online tools53 with Salmonella spp. as the interested organisms and the plasmid MLST database. Antibiotic resistance genes were detected using ResFinder v4.1 online tool31,50,51,54 with Salmonella spp. as the interested organisms.

Pangenome

The analysis included a total of 27 isolates. Sequence data of isolates H-012, H-014, and H-120 were from hybrid de novo assembly results and only represent chromosomal DNA. The sequence data of the Japan isolates (L-4445, L-4567, and L-4605) were from the NCBI Rep-Seq database and, again, represent only chromosomal DNA. The remaining 21 isolates were obtained from short-read de novo assembly and represent both chromosomal and plasmid DNA. Contigs from the de novo short-read assembly that were less than 1,000 bp in length were excluded from the analysis. To perform the pan-genome analysis, we utilized the Panaroo program version 1.3.2 with strict mode55.