In Australia, restrictions on live animal imports, geographic isolation, and sound antibiotic stewardship have limited the incorporation and spread of genes encoding resistance to antibiotics used to treat serious human infections among Enterobacteriaceae circulating in food animals (Turner, 2011; Abraham et al., 2015; Reid et al., 2017; Kidsley et al., 2018). Australian porcine Escherichia coli are, albeit at low frequency, known to carry bla_CMY−2, bla_CTX−M−14, and bla_CTX−M−9, possibly because of the use of ceftiofur as an off-label, last line antibiotic to treat serious disease (Abraham et al., 2015). First generation antibiotics; commonly tetracyclines, penicillins, and sulphonamides are otherwise used for the therapeutic treatment of bacterial infections in swine (Jordan et al., 2009). Both pathogenic and commensal E. coli sourced from the feces of pig carry class 1 integrons and are often multiple drug resistant (MDR) (Abraham et al., 2015; Wyrsch et al., 2015; Reid et al., 2017). Notably, the insertion element IS26 has infiltrated the genomes of commensal E. coli of porcine origin, where it has played a role in altering the genetic context of clinical class 1 integrons and facilitated the acquisition of further resistances (Reid et al., 2017). Furthermore, it is important to identify and characterize the genetic vehicles that can carry class 1 integrons and genes encoding resistance to first generation antibiotics which are widespread in commensal E. coli populations in Australian pigs (Abraham et al., 2015; Reid et al., 2017; Kidsley et al., 2018) because they are likely to acquire genes encoding resistance to clinically relevant antibiotics when resident in commensal enterobacterial populations in the gut of humans and companion animals.

Plasmids of the incompatibility group HI2 (IncHI2) have been linked with the carriage of tellurium resistance, plus a broad range of antimicrobial resistance genes, including bla_IMP, bla_NDM, and bla_VIM (carbapenemase resistance) (Abraham et al., 2016; Dolejska et al., 2016; Falgenhauer et al., 2017), bla_CTX−M (Garcia Fernandez et al., 2007), mcr-1 (plasmid-mediated colistin resistance) (Gilrane et al., 2017; Li et al., 2017; Zheng et al., 2017), oqxAB and aac(6')-Ib-cr (plasmid-mediated quinolone resistance) (Fang et al., 2016), and resistance genes effective against first generation antibiotics and disinfectants (Cain et al., 2010; Cain and Hall, 2012; Abraham et al., 2016; Dolejska et al., 2016; Gilrane et al., 2017). Notably, clinically-important resistance genes have been associated with IncHI2 plasmids that also encode resistance to copper, zinc and, arsenic residues (Fang et al., 2016). In Australia, IncHI2 plasmids carrying a diverse range of antimicrobial resistance genes have been found in MDR E. coli (Dolejska et al., 2016; Saputra et al., 2017), Salmonella enterica serovar Typhimurium (Cain et al., 2010; Cain and Hall, 2012; Billman-Jacobe et al., 2018), Enterobacter cloacae (Sidjabat et al., 2015), and Enterobacter hormaechei susbp. Oharae (Monahan et al., 2019).

The insertion sequence IS26 of the IS6 family plays a significant role in the capture, assembly and mobilization of drug resistance genes found on plasmids (Dionisi et al., 2009; Cain et al., 2010; Shahid, 2010; Venturini et al., 2010, 2013; Lai et al., 2013; Abbo and Hooton, 2014; Chavda et al., 2015; Reid et al., 2015; Garcia et al., 2016) and in the chromosome (Roy Chowdhury et al., 2015, 2018). IS26 may promote plasmid stability and persistence by mediating deletions of plasmid backbone sequence, the expression of which incurs a burden to the host (Porse et al., 2016). IS26 can also facilitate the generation of hybrid virulence/resistance plasmids, formed by co-integration of plasmids separately carrying resistance and virulence gene cargo (Mangat et al., 2017; Wong et al., 2017). Although IS26 has no target site specificity (Harmer et al., 2014), it is often observed localizing next to or within class 1 integrons in multiple drug resistant E. coli recovered from the feces of both healthy pigs, poultry and cattle with gastrointestinal disease (Dawes et al., 2010; Reid et al., 2017). IS26 can shape the structure of class 1 integrons by facilitating the addition of foreign DNA flanked by IS26, and by generating inversion and deletions of sequence within complex resistance structures. Consequently, many class 1 integrons have lost genes that reside within the often observed 3′-conserved sequence (3′-CS), particularly the sulphonamide resistance gene sul1 (Reid et al., 2017). However, sulphonamide resistance persists globally (Grape et al., 2003; Bean et al., 2005; Suhartono et al., 2017), despite restrictions on its use (Enne et al., 2001), with sul2 (Bean et al., 2005) and sul3 genes (Grape et al., 2003; Perreten and Boerlin, 2003; Zhou et al., 2014; Reid et al., 2017) found frequently in close association with class 1 integrons. A sul4 gene associated with a chromosomal locus containing the folate synthesis gene folK and a copy of ISCR20 has recently been described (Razavi et al., 2017). Emerging trends show that genes encoding resistance to last line drugs, such as mcr-1, bla_CTX−M, and carbapenemases are captured on plasmids already carrying genes encoding resistances to first generation antibiotics (Alonso et al., 2017; Botts et al., 2017; Delannoy et al., 2017; Poirel et al., 2017). Any one of a number of selection pressures including heavy metals and biocides may be sufficient to then facilitate the persistence and spread of multiple drug resistance plasmids (Argudín et al., 2019). This is further compounded by reports that agrichemicals can alter selection for drug resistant bacteria (Kurenbach et al., 2015, 2018). The potential for complex resistance structures to persist under multiple different selective pressures and across diverse environments underpin the importance of adapting a One Health approach to antimicrobial resistance gene surveillance in humans, food and companion animals, agriculture, effluent (municipal, hospital, and agricultural), and the environments impacted by effluent from diverse sources (Djordjevic et al., 2013; Wyrsch et al., 2016; Huang et al., 2017; Mir et al., 2018).

Here we used a combination of Illumina and PacBio SMRT sequencing to completely close two IncHI2:ST3 MDR plasmids from E. coli recovered from the feces of Australian swine. One of these, pSDE-SvHI2 is from a severe ETEC/ExPEC pathogen (E. coli O157 SvETEC) (Wyrsch et al., 2015), and the other, pSDC-F2_12BHI2, is from Escherichia coli F2_12B, a commensal E. coli ST93 (Reid et al., 2017). Phylogenetic and genomic comparisons were undertaken between these two plasmid sequences and pSTM6-275, recently isolated from a monophasic variant of Salmonella enterica in Australian pigs (Dyall-Smith et al., 2017), as well as all available IncHI2 ST3 plasmids on GenBank.

Unicycler hybrid assemblies resolved pSDE-SvHI2 as a 275,402 bp circular sequence and pSDC-F2_12BHI2 as a 288,288 bp circular sequence with 344 (140 hypotheticals) and 357 (134 hypotheticals) coding sequences (CDS) respectively. Both sequences typed as incompatibility group HI2, sequence type 3 (IncHI2 ST3) plasmids. The full map of each plasmid is presented in Figure 1, along with content comparisons to the most related ST3 plasmids, most of which are from Australia and China. The plasmids had a typical IncHI2 structure, including the RepHIA and RepHI2 replication operons, trh and tra transfer operons and tellurite resistance (ter operon) (Gilmour et al., 2004), plus each carried three variants of complex resistance loci, two of which are class 1 integron associated, one encoding sul1 in a Tn1721/Tn21 hybrid transposon, and one encoding sul3 from within a Tn7-like copper/silver resistance transposon.

Single nucleotide polymorphism analysis was performed on the set of all ST3 plasmids (Figure 2, Supplementary Table 2), and on this same set plus IncHI2 ST1 plasmids as reference to confirm tree topology (Supplementary Figure 1). Forty-three plasmid sequences available from 2006 to 2017 were included in the ST3-only analysis. Based on the availability of metadata, plasmid sequences were from Enterobacteriaceae of different sources in the pacific region, including Australian porcine production operations and multiple human, agricultural and environmental sources in China. The ST3 sequences formed a single clade with one exception, MH715960 from Taiwan, which separated with 125 core SNPs compared to the Australian reference sequence. Of these 125 core SNPs, 101 are within an ~1.5 kb region of the ter operon. The remaining sequences formed four major subclades, with closest relatives to the Australian reference ranging from two to 14 SNPs. One subclade, highlighted red in Figure 2, demonstrates relatedness between three Australian plasmids from the feces of pigs, pSDE-SvHI2, pSDC-F2_12BHI2, and pSTM-275, an apparently separate Australian plasmid lineage (pIncHI2-MU3) also from pig feces, and six other plasmid sequences from diverse sources in China. Content comparisons of this clade can be seen in Figure 1. Of the remaining plasmids, the most distant relative had 25 SNPs identified from conserved core sequence, suggesting a close evolutionary relationship between plasmid sequences reported from Australia and China.

The ST3 plasmids carried multiple class 1 integron structures with different sul genes including sul1, sul2, sul3, and a sul3-associated mefB, encoding a macrolide efflux pump (Figure 2). One sequence (KX254341) of Chinese origin was observed with only 260 bp remaining of mefB (ΔmefB₂₆₀). The Australian IncHI2 ST3 plasmids explored here carry ΔmefB₁₁₁. The remaining sul3-positive plasmids carry a ΔmefB₃₃ signature, aside from CP026492 from China which completely lacks mefB. This includes both the pseudo-phylogenetically distant MH715960 from Taiwan and the earliest Chinese plasmid sourced from 2006. Interestingly, one single subclade of sequences from China and Hong Kong was universally negative for intI1 and sul3, but not for either sul1 or sul2. Many of these plasmids are also associated with various globally important resistance genes, including bla_CTX−M, oqxAB and mcr-1 (Supplementary Table 3).

pSDE-SvHI2, pSDC-F2_12BHI2, and pSTM-275 host a class 1 integron located within a Tn1721/Tn21 hybrid tetracycline resistance transposon (Figure 3). This integron structure is host to the only complete intI1 gene on the plasmids, and has acquired trimethoprim (dfrA12), orfF and streptomycin/spectinomycin (aadA2) resistance gene cassettes. Comparisons to similar Tn1721/Tn21 hybrid transposons show identical sequence across the hybridization point between tnpR₁₇₂₁and tnpM₂₁, however these sequences also show a modified transposition module, with a loss of Tn1721 tnpA sequence between two homologous 8 bp regions (CCAGGGCG), between the second ΔtnpA₁₇₂₁ and its neighboring predicted relaxase. The In2-like class 1 integron tniA gene is truncated to 436 bp by insertion of an IS26 element with no observable associated repeats, suggesting a complex evolutionary path to this final structure. As part of the class 1 integron 3′-CS, the structure encodes sulphonamide resistance (sul1), however in pSTM6-275 the terminal IS26 truncation lies within the sul1 gene, giving a distinguishable gene marker.

A large (58,759 bp), mosaic combination of a globally observed Tn7-like transposon (copper and silver resistance, sil/pco), and transposons mobilizing antimicrobial resistance genes (Figure 1) was also observed in this Australian plasmid lineage. An IS26-associated, class 1 integron-encoding structure has inserted near the Tn7-like transposition module (Figure 4). From the pSTM6-275 sequence, it appears this insertion was in a Tn3 associated gene, however in pSDE-SvHI2 and pSDC-F2_12BHI2 IS26 activity has altered sequences around the insertion site. The integron is a variant of a sul3-associated structure that has been described previously in Australian plasmid pCERC3 (KR827684), now including signature IS26-meditated deletions in intI1 (ΔintI1₇₀₅) and mefB (ΔmefB₁₁₁). The integron cassette array encodes estX, psp, aadA2, cmlA, and aadA1. The crossover site described near qacH reported in pCERC3 was identical, confirming it is a derivative of the same integron sequence encoding sul3-qacH-mefB, previously seen within a Tn21 background (Moran et al., 2016). We also noted an IS1203^* insertion into the qacH ORF, which may prove epidemiologically useful. The integron structure is followed by a short intermediate sequence encoding rop, then by a Tn5393 variant (encoding streptomycin resistance) with a novel IS903B^* insertion into IS1133.

From these sequences it is difficult to ascertain what originally mobilized the ΔintI1 module into the Tn7-like transposon. While IS26 flanks the captured sequence (ΔintI1 to sul3), no repeats have been identified to indicate a clean insertion location near the Tn7-like transposon, nor the bordering sequence near rop.

Of the 1,030 bp sequence by rop, 257 bases nearest the Tn5393 repeat are associated with the tetracycline resistance transposon Tn1721 [AJ634602.1] while the remaining 773 bp nearest the IS26/sul3 end matched IncN plasmid backbone [HF545433.1]. These remnant sequences may give indications as to the sources of these structures should more pertinent references become available. Further, the structures we have described have undergone individual insertions and inversion events. In pSDE-SvHI2, an IS26 insertion into tnsA has led to the generation of new direct repeats (DRβ) and has subsequently led to an IS26 mediated inversion. In pSDC-F2_12BHI2, an IS26 insertion near the tnsA ORF has generated DRγ, and an inversion event has again occurred, flipping the structure between ΔintI1 and sul3. A separate inversion event has then occurred between the IS26 elements nearest to rop, re-inverting this sequence to match the original orientation. Importantly, the insertion of a complex Tn1721/Tn2/IS26 structure into Tn5393 within pSTM6-275 demonstrates further resistance consolidation occurring in the 7-year gap between pSDE-SvHI2/pSDC-F2_12BHI2 and pSTM-275 isolations. This insert encoded tetAB and bla_TEM. Unfortunately, this insertion has removed most of IS1133, and we cannot determine if pSTM6-275 carried the ΔIS1133-IS903B^* insertion described above. The other Australian plasmid, pIncHI2-MU3, encodes the Tn7-like sil/pco heavy metal resistance transposon and is positive for Tn3 and intI1, but lacks any sul gene.

Finally, an IS15DI (IS6-family element; 3 SNPs compared to IS26) mobilized resistance region with a highly recombined and varied structure (Figure 5) has also been identified in pSDE-SvHI2 and pSDC-F2_12BHI2. The original IS15DI insertion into the plasmid backbone has generated 8 bp direct repeats (AACAGCGT) that remain flanking the structure. This composite transposon appears to mobilize neomycin resistance (neoR), and a truncated bleomycin resistance (ble) gene. It has also acquired IncN backbone (tra genes) and the relBE toxin/anti-toxin system, alongside various other whole and truncated IS elements. An internal IS15DI element is present in both pSDE-SvHI2 and pSTM6-275 and is replaced by a Tn6010-like element (similar to KT716391.1, mobilized by IS15DI) carrying oqxABR in pSDC-F2_12BHI2. there is an IS15DI mediated inversion of the tra associated region in pSDC-F2_12BHI2. Plasmid pSDE-SvHI2 has an IS15DI insertion into IS1294, generating DRδ. This has been followed by a rearrangement event leading to the movement of DRδ and neoR toward the terminal IS element, including an inversion, and a loss of sequence through to the ΔtnpA_Tn3.

Plasmids belonging to the incompatibility group HI2 are carriers of antimicrobial resistance genes globally. With the increased availability of IncHI2 sequences, a di-locus sequence typing scheme was established (Garcia-Fernandez and Carattoli, 2010) to aid in tracking plasmid lineage development and dissemination. In Australia, most IncHI2 plasmids have been observed mobilizing bla_IMP−4, the predominant carbapenemase-encoding gene within Enterobacteriaceae on the eastern seaboard of Australia (Sidjabat et al., 2015). Here we report a comprehensive analysis of two Australian IncHI2 ST3 plasmids from E. coli, and make comparisons to plasmid pSTM6-275 from S. enterica serotype 1,4,[5],12:i:- (Dyall-Smith et al., 2017). Phylogenetic analyses of the current available IncHI2 ST3 dataset showed these three Australian porcine IncHI2 ST3 plasmids are nested within a subclade alongside another Australian porcine ST3 plasmid sequence, pIncHI2-MU3 (Abraham et al., 2018), which hosts neither of the class 1 integrons identified here, but does host the metal resistance genes as part of the Tn7-like transposon. This suggests various lineages of ST3 plasmid may be circulating within Australian porcine agricultural operations.

The IncHI2 ST3 plasmids sequenced to date have been resolved from multiple Enterobacteriaceae including Escherichia coli, various Salmonella enterica serovars, Shigella flexneri, Klebsiella pneumonia, and Raoultella ornithinolytica. Additionally, these Enterobacteriaceae isolates were taken from different sources, including pathogens and non-pathogens. As these ST3 plasmids carry diverse resistance gene cargo and may be found in numerous members of the Enterobacteriaceae family they will likely be important targets for future antimicrobial resistance surveillance (Fang et al., 2018). Our analyses also demonstrate evidence of differential gene acquisition by these plasmids, continuing to expand and alter the antimicrobial resistance gene repertoire which they are associated with internationally. Particularly, sul3-associated integrons carrying different ΔmefB fragments were identified here.

The sul3 gene was first described in pigs (Perreten and Boerlin, 2003) and has since been reported widely in association with Enterobacteriaceae in humans, farm animals and their waste—particularly swine (Antunes et al., 2007; Phuong Hoa et al., 2008; Byrne-Bailey et al., 2009; Curiao et al., 2011; Moran et al., 2016). In Australia and elsewhere, sul3 is identified frequently in commensal E. coli from the feces of swine (Guerra et al., 2004; Bischoff et al., 2005; Reid et al., 2017) but less frequently in avian pathogenic E. coli (APEC) from diverse poultry production systems in Australia (Cummins et al., 2019). These observations suggest that swine production, and food animal production more broadly, plays a major role in the evolution of IncHI2 ST3 plasmids that carry sul3. Plasmid pSTM6-275 was shown to be thermostable at 44°C, and would readily conjugate at 27°C but not at 37°C, highlighting the IncHI2 plasmid families propensity to transfer under environmental conditions (Garcia-Fernandez and Carattoli, 2010; Billman-Jacobe et al., 2018).

Genetic signatures noted in this study; a combination of the IncHI2 ST3 di-locus typing alleles and mefB₁₁₁ have been seen in commensal E. coli short read assemblies from both the farm were strain F2_12B was sourced, and a separate Australian porcine operation circa 2007 (Reid et al., 2017). Combined with the isolation of pSTM6-275 in 2014, our data suggests these plasmids are purveyors of drug resistance in porcine agricultural settings, likely aided by the use of copper feed additives and the widespread use of first generation antibiotics (Jordan et al., 2009). It is notable that the IncHI2 plasmids sequenced in our study carry a unique IS6 family mobilized transposon encoding neomycin resistance, and additional fluoroquinolone resistance in pSDC-F2_12B. Also, they are host to a globally observed Tn7-like heavy metal resistance transposon (Fang et al., 2016), that has in turn become host to a variant of a sul3-mefB class 1 integron, which appears unique to Australian samples at this time. The ΔmefB₁₁₁-associated class 1 integron was carried only by the Australian plasmids pSDE-SvHI2, pSDC-F2_12BHI2 and pSTM-275, while ΔmefB₃₃ was observed throughout the remainder of the ST3 plasmid clade. Notably, ΔmefB₃₃ variants are not associated with the presence of the Tn7-like metal resistance transposon, so characterizing their distribution and methods of mobilization will be critical to monitoring complex resistance locus diversification.

Genomic sequence data has shown that IS26 is playing an important role in shaping the context of MDR islands in MDR fecal E. coli in Australian commercial pigs. Recently, we reported high carriage rates (101/103 isolates; 98%) of IS26 among commensal E. coli carrying class 1 integrons (Reid et al., 2017), however we were unable to determine the genetic context of some class 1 integrons. Nonetheless, the frequency of carriage of sul3 and IncHI2 plasmids in our earlier study was significant in two separate commercial swine production facilities. Here we show that IS26 (and the IS15DI variant) has played a pivotal role in the evolution of several porcine IncHI2 ST3 plasmids both as a means of creating deletions and inversions and altering antibiotic resistance gene content.

Finally, the epidemiological analysis of Enterobacteriaceae has been heavily influenced by the spread of antimicrobial resistance and the development of MDR pathogens, leading to a focus on the detection of clinically-relevant resistance genes (bla_IMP, mcr-1, and bla_CTX−M are examples of this). To further understand the true variety and dissemination of multiple drug resistance, sampling from agricultural and environmental sources impacted by antimicrobials will be important.

The datasets generated for this study can be found in GenBank, MH287084 and MH287085.

EW performed genomic analyses, generated figures, and drafted the manuscript. TC provided curated E. coli collections for the study. PRC assisted with data interpretation. ML, EW, and CR prepared sequencing samples. MD performed genome assemblies. SD conceived the study and together with EW and PRC wrote the manuscript. All authors provided edits, read, and approved the manuscript.