Genomic Microbial Epidemiology Is Needed to Comprehend the Global Problem of Antibiotic Resistance and to Improve Pathogen Diagnosis

Contamination of waste effluent from hospitals and intensive food animal production with antimicrobial residues is an immense global problem. Antimicrobial residues exert selection pressures that influence the acquisition of antimicrobial resistance and virulence genes in diverse microbial populations. Despite these concerns there is only a limited understanding of how antimicrobial residues contribute to the global problem of antimicrobial resistance. Furthermore, rapid detection of emerging bacterial pathogens and strains with resistance to more than one antibiotic class remains a challenge. A comprehensive, sequence-based genomic epidemiological surveillance model that captures essential microbial metadata is needed, both to improve surveillance for antimicrobial resistance and to monitor pathogen evolution. Escherichia coli is an important pathogen causing both intestinal [intestinal pathogenic E. coli (IPEC)] and extraintestinal [extraintestinal pathogenic E. coli (ExPEC)] disease in humans and food animals. ExPEC are the most frequently isolated Gram negative pathogen affecting human health, linked to food production practices and are often resistant to multiple antibiotics. Cattle are a known reservoir of IPEC but they are not recognized as a source of ExPEC that impact human or animal health. In contrast, poultry are a recognized source of multiple antibiotic resistant ExPEC, while swine have received comparatively less attention in this regard. Here, we review what is known about ExPEC in swine and how pig production contributes to the problem of antibiotic resistance.

global antimicrobial consumption will increase by 67% from an estimated 63,151 tons in 2010 to 105,596 tons in 2030 (Van Boeckel et al., 2015). Much of this rise in consumption is expected to come from China, India, Russia, South Africa, and Brazil; countries where meat consumption is growing due to rising incomes (Van Boeckel et al., 2015). Large scale animal production facilities generate huge volumes of animal waste that is contaminated with veterinary antibiotics. While estimates vary, it is thought that in the USA, 11-14 million kilograms of antibiotics are used annually in the production of animals for food (i.e., livestock; Institute of Medicine; National Research Council; Panel on Animal Health, 1999; Landers et al., 2012). Almost half of these are used for non-therapeutic purposes, which is significantly more than the estimated 1.4 million kilograms of antibiotics used in human medicine (Institute of Medicine; National Research Council; Panel on Animal Health, 1999). Many classes of antibiotics are used in animal production including β-lactams, sulfonamides, tetracyclines, streptogramins, macrolides, lincosamides, polyethers, quinoxalines, elfamycins, glycolipids, arsenicals, and polypeptides (Sarmah et al., 2006). How these compounds are modified within the animal, how they persist in soils and waterways after they are excreted, and their fate when animal slurries are used to fertilize pastures are complex and poorly understood (Sarmah et al., 2006;Zhu et al., 2013). Antimicrobial residues that contaminate groundwater tables, aquatic environments, and land used for crop production exert selection pressure on microbial populations that may promote the lateral transfer of resistance and virulence genes between populations, and contributes to the emergence of novel pathogenic profiles (Venturini et al., 2010(Venturini et al., , 2013Baquero and Tobes, 2013;Roy Chowdhury et al., 2015). It is perhaps unsurprising that commensal gastrointestinal bacterial populations are commonly resistant to more than one antimicrobial (Bettelheim et al., 2003;Diarrassouba et al., 2007;Bailey et al., 2010) and consequently the frequency of community-acquired infections with resistance to multiple antimicrobials are also increasing (Onken et al., 2015;Paniagua-Contreras et al., 2015;Yahiaoui et al., 2015). Furthermore, surveys have shown the presence of antibiotic resistance genes in the gastrointestinal tracts of humans and animals that have never received antibiotic interventions, specifically in: human feces from remote communities (Bartoloni et al., 2009;Ruppe et al., 2009;Eisenberg et al., 2012), migratory bird populations (Foti et al., 2011;Carroll et al., 2015;Jamborova et al., 2015;Stedt et al., 2015) and various wildlife species (Allen et al., 2010;Baez et al., 2015;Jobbins and Alexander, 2015;Katakweba et al., 2015). This demonstrates that resistance genes in host bacteria can readily spread to naïve microbial populations via waterways, air currents, road incursions into remote areas, animal and bird migration, and land clearance (Tsubokura et al., 1995;Middleton and Ambrose, 2005;Ewers et al., 2009). The application of manure onto pasture from ruminants that have not received recent antibiotic therapy is known to stimulate the expansion of resident flora encoding resistance to β-lactam antibiotics (Udikovic-Kolic et al., 2014). These observations demonstrate our limited understanding of the impact of the application animal waste to agriculture. The mechanisms by which mobile elements flux through microbial populations are not fully understood, but it is clear that efforts must be made to reduce environmental contamination sourced from human and animal effluent.
Antimicrobial resistance is largely an ecological problem (Summers, 2002(Summers, , 2006Stokes and Gillings, 2011;Landers et al., 2012;Berendonk et al., 2015;Gillings et al., 2015). In general, studies have focused on the emergence and carriage of resistance genes in certain key microbial populations, without an appreciation for the wider complexity of the problem, or its origins. Hence, there are numerous under-reported microbial sources of resistance genes. Antibiotic resistance genes often traffic in association with transposons that carry mercury resistance operons; these play an important role in transforming mercury to less biologically toxic forms, thereby regulating the availability of toxic mercury compounds in the environment. Mercury-resistance gene operons have evolved as key components of Tn3-family transposons and these transposons are globally disseminated. Mercury has been released into the environment via geological processes over millions of years and, although anthropogenic activity over the past two centuries has contributed significantly to the release of mercury into waterways and the atmosphere, it is likely that mercury levels were higher prior to the industrial revolution than they are now as a result of extensive volcanic activity (Osborn et al., 1997;Barkay et al., 2003). Despite this, large scale surveillance studies of antimicrobial resistance fail to screen microbial populations for the presence of mercury-resistance genes . Resistance genes can also accumulate in microbial populations in soils that have been fertilized with animal waste and treated effluent (Heuer and Smalla, 2007;Chee-Sanford et al., 2009;Heuer et al., 2011a,b;Kristiansson et al., 2011;Berendonk et al., 2015). Meat, seafood, and vegetables are also important and under-recognized sources of antimicrobial resistance genes in bacteria (Johnson et al., 2005a;Agerso et al., 2014;Kim and Woo, 2014;Ahmed and Shimamoto, 2015a,b;Ben Said et al., 2015;Le et al., 2015;van Hoek et al., 2015;Zhang et al., 2015;Zurfluh et al., 2015).
Anthropogenic industrial processes generate heavy metals including cadmium, lead, arsenic, copper, silver, and a wide range of synthetic organic compounds that pollute natural ecosystems. As such, heavy metal contamination represents yet another selective pressure for the development of antimicrobial resistance (Baker-Austin et al., 2006;Castillo et al., 2008;Li et al., 2010;Zhu et al., 2013). Genes coding for resistance to silver, copper, arsenic and antimony are associated with complex resistance gene loci in bacteria resistant to more than one antimicrobial and have great significance to human, livestock, and plant health (Berg et al., 2005;Seiler and Berendonk, 2012;Hobman and Crossman, 2015;Bondarczuk et al., 2016). The widespread occurrence of genes coding for resistance to heavy metals has, in part, been demonstrated in studies using information mined from publicly available genome sequences highlighting the value of genomic surveillance studies (Hobman and Crossman, 2015).
Genome-wide studies of genes coding for resistance to antibiotics, biocides and heavy metals have shed light on how these genes co-assemble on mobile genetic elements. In an analysis of 2522 fully sequenced bacterial genomes and 4582 plasmid sequences, Pal et al. (2015) were able to identify examples where antimicrobial resistance genes co-occurred. The authors found that the most likely co-selection scenario occurred in bacterial strains that carried plasmid-borne antibiotic resistance genes and biocide/heavy metal resistance genes on the same chromosome. A plasmid localized gene cluster coding for a cadmium/zinc resistance gene (cadD) with aminoglycoside and macrolide resistance genes was also identified (Pal et al., 2015). Resistance to these combinations of antimicrobials would benefit bacterial populations found in intensive animal production systems where cadmium and zinc are both used to promote growth and aminoglycosides and macrolides are frequently administered for the treatment of Gram-negative and Grampositive infections, respectively (Castillo et al., 2008;Li et al., 2010). It is clear that plasmids can carry combinations of genes coding for resistance to more than one microbial agent alongside genes coding for resistance to biocides and heavy metals. Invariably these plasmids are large, conjugative and commonly found in pathogens from hospital and intensive farming environments (Chen et al., 2007;Woodford et al., 2009;Venturini et al., 2010Venturini et al., , 2013Pal et al., 2015). Sub-lethal levels of antibacterial biocides and heavy metals found in mildly polluted aquatic and land-based ecosystems may be sufficient to maintain these mobile elements carrying multiple antibiotic, biocide and heavy metal resistance genes Gullberg et al., 2014).
Although plasmids can vary in size and coding ability, their size is generally positively correlated with the size of the host chromosome (Smillie et al., 2010). Plasmids conferring multipledrug resistance are usually large, as they code for a suite of resistance genes, alongside associated integrons, transposition factors, and genes for heavy metal resistance, which can be in excess of 20,000 bp or more in size. Small plasmids that acquire resistance to more than one antimicrobial are, however, not without precedent. Labar et al. (2012) described the sequence of a small plasmid (pASL01A; 27,072 bp) that codes for resistance to ampicillin, streptomycin, sulfonamides, trimethoprim, and mercury (Labar et al., 2012). Most of the plasmid (21,904 bp) is comprised of a Tn21 derivate mercury resistance transposon that carries a class 1 integron with a trimethoprim resistance dfrA7 gene cassette and Tn6029C (Roy . Tn6029C comprises a bla TEM−1 -IS26-repA/C-sul2-strA-strB gene cluster flanked by direct copies of IS26 (Labar et al., 2012;Roy Chowdhury et al., 2015). The plasmid backbone (5,168 bp) and related plasmids are widespread in enterobacterial populations in Africa (Labar et al., 2012).
Antimicrobials, including heavy metals, are used widely in animal production to treat clinical disease, to prevent disease outbreaks during critical or vulnerable periods, and to promote growth (McEwen and Fedorka-Cray, 2002;Landers et al., 2012). The use of antibiotics to promote growth is a phenomenon whereby animals receiving very low doses of antimicrobials appear healthier and grow larger (Roy Chowdhury et al., 2014). Although, the mechanism for this growth promoting effect is still poorly understood, the majority of pigs produced in the USA are exposed to tetracyclines and tylosin to prevent disease and promote growth (Apley et al., 2012). While these approaches have proved effective in controlling disease and aiding farms globally, the uninhibited use of antimicrobials also provides the selection pressure that drives antimicrobial resistance (Pulcini and Gyssens, 2013). As such, concerted efforts to reduce the use of antimicrobials in food production should be a priority.
Given that the persistence and spread of resistance is intrinsically linked with the presence and activity of antimicrobials, it is significant that 30-90% of ingested antibiotics are excreted in an un-metabolized or only partially metabolized form (Sarmah et al., 2006) into waste treatment plants and animal waste-holding facilities. Antibiotic residues are known to persist in secondary effluent despite the treatment process (Sarmah et al., 2006;Watkinson et al., 2007;Martinez, 2008;Le-Minh et al., 2010;Bondarczuk et al., 2016). Furthermore, the pH, temperature, nutrient concentration, and bacterial loads in waste treatment plants are conducive to the evolution and spread of antimicrobial resistance (and virulence) genes (Andersen, 1993;Dolejska et al., 2011). This is in part evidenced by an increase in the presence of the class 1 integrase gene, a reliable proxy for resistance to more than one antimicrobial in bacterial populations sampled in wastewater treatment plants (LaPara et al., 2011;Ma et al., 2013;Gillings et al., 2015), and an increase in the number of resistance genes in wastewater treatment plants and rural domestic wastewater treatment systems (Gao et al., 2012;Chen and Zhang, 2013;Ju et al., 2015;Mao et al., 2015). This indicates that controlling the use of antimicrobials is a prerequisite for slowing the development and spread of pathogens with multiple antimicrobial resistance mechanisms.

ANTIMICROBIAL RESISTANCE IN Escherichia coli
Escherichia coli is a widespread and abundant organism capable of causing a wide range of gastrointestinal and extra-intestinal diseases. It can survive and proliferate in a diversity of terrestrial and aquatic environments (van Elsas et al., 2011). As such, its genetic profile is shaped profoundly by horizontal gene transfer, and it is a useful marker species for understanding how antimicrobial resistance and virulence genes accumulate over evolutionary time. Horizontal gene transfer events are known to generate novel combinations of putative and established virulence factors, and generate novel pathotypes of E. coli Bielaszewska et al., 2014 Kaper et al., 2004;Brzuszkiewicz et al., 2011). EHEC are a subgroup of STEC, which colonizes the terminal regions of the gastrointestinal tract of ruminants. It is notable that different serotypes of E. coli preferentially colonize different ruminant hosts (Djordjevic et al., , 2004Hornitzky et al., 2002) and carry different Shiga toxin gene variants (Ramachandran et al., 2001;Brett et al., 2003a,b). These studies underscore the importance of understanding the ecology of disease-causing organisms and how niche adaptation influences lateral gene transfer. Notably, however, ruminants are not a recognized source of ExPEC (Manges and Johnson, 2015).
Many recent molecular surveillance studies on E. coli report the frequency of genes that code for extended-spectrum beta lactamase enzymes (ESBL's), which are indicative of resistance to penicillin, cephalosporin, and other clinically important antibiotics, but fail to screen for resistance to the older, first generation antibiotics (Bean et al., 2009;Ewers et al., 2012;Glasner et al., 2013;Roschanski et al., 2015). Genes that code for resistance to first generation antibiotics are often carried in bacteria containing class 1 integrons (Johnson et al., 2005b;Jakobsen et al., 2011;Hansen et al., 2014). From a diagnostic perspective, recent genomic studies on an outbreak of EAHEC and EAEC strain O104:H4, highlighted this issue (Ahmed et al., 2012). EAHEC is a recently emerged E. coli pathotype and EAEC strain O104:H4 has acquired the Shiga toxin gene stx2a, a diagnostic marker of EHEC. Outbreak strains were found to be resistant to ampicillin, streptomycin, sulfamethoxazole, trimethoprim, and tetracycline but little effort was given to determining how the genes that coded for these antibiotic resistance phenotypes were assembled (Ahmed et al., 2012). We were subsequently able to show that the genes coding for resistance to these antibiotics were assembled in Tn6029D, a transposon that abuts a class 1 integron in a Tn21/Tn1721 hybrid backbone (Roy . The resistance region resided in genomic island 3 which also carried the virulence gene ag43 encoding a self-associating serine protease auto-transporter, important for biofilm formation. This is an example of how antimicrobial resistance genes and virulence genes co-localize on the same mobile genetic element (Roy . Chromosomal islands are increasingly associated with carriage of class 1 integrons encoding resistance to multiple antibiotics of clinical significance (Roy Chowdhury et al., 2016).
Salmonella genomic island 1 (Boyd et al., 2001;Levings et al., 2005), Salmonella genomic island 2 (Levings et al., 2008) and several large plasmids in EHEC strain O26:H (Venturini et al., 2010), and an atypical EPEC, strain O111 (Venturini et al., 2013) are also examples of virulence and resistance genes being carried on the same mobile element. Strains of E. coli that carry genes that are markers for more than one E. coli pathotype pose a dilemma for diagnostic laboratories. There are numerous reports describing novel E. coli strains with combinations of virulence genes that are diagnostic of more than one E. coli pathotype; this suggests that newly emerging strains are increasingly likely to have complex resistance loci facilitated by the clustering of mobile element-associated resistance and virulence genes. In a study of 265 cases of E. coli causing UTI, 28 carried virulence genes typical of IPEC (Bielaszewska et al., 2014). Several other examples of hybrid E. coli strains have been described and their presence demonstrates the need to have genomic surveillance strategies in place to determine how antibiotic resistance and virulence genes assemble in emerging pathogens (Wallace-Gadsden et al., 2007;Abe et al., 2008;Bielaszewska et al., 2014;Mariani-Kurkdjian et al., 2014).
Extraintestinal pathogenic E. coli is a phylogenetically diverse group comprised of a broad range of E. coli sequence types (STs). ExPEC are fecal E. coli that rarely cause intestinal disease but have acquired a wide variety of virulence gene cargo that facilities their ability to cause disease at extraintestinal sites, particularly in the urinary tract. ExPEC are the most frequently isolated Gram negative bacterial pathogen infecting humans incurring significant mortality and morbidity. Notably, ExPEC are a leading cause of sepsis and are increasingly resistant to multiple antibiotics posing a serious health concern (Manges and Johnson, 2015;Poolman and Wacker, 2016). ExPEC have a fecal origin but rarely cause gastrointestinal disease. The acquisition of a diverse range of virulence factors by lateral gene transfer has armed ExPEC with an ability to cause life-threatening blood-borne, urinary tract, respiratory, skin and soft tissue infections, and meningitis in humans of all ages (Russo and Johnson, 2003;Poolman and Wacker, 2016). ExPEC vary widely in the combinations of putative virulence genes that they carry. Many genes encode for adhesins, iron acquisition systems and other putative virulence attributes and this genetic redundancy complicates efforts to develop broadly applicable molecular diagnostic tests to detect ExPEC (Johnson et al., 2001;Spurbeck et al., 2012). Importantly, livestock are important reservoirs for ExPEC (Ewers et al., 2007(Ewers et al., , 2009Manges et al., 2007;Vincent et al., 2010;Nordstrom et al., 2013;Manges and Johnson, 2015). Virulence genes are often shared among E. coli isolates from samples collected from humans, and the meat and feces of intensively reared livestock, particularly poultry (Rodriguez-Siek et al., 2005;Vandekerchove et al., 2005;Jakobsen et al., 2010a,b). Consistent with this view, ExPEC carry genes coding for resistance to antimicrobials frequently used in veterinary medicine. Hospital diagnostics laboratories do not screen for most of the antibiotics used in veterinary medicine and so resistance to these is underreported. In a Chinese study of 315 ExPEC isolates from various pig tissues, more than 63% exhibited resistance to 10 antibiotics including ampicillin, trimethoprim, sulfadimidine, tetracycline, neomycin, streptomycin, and kanamycin (Tan et al., 2012). The genes bla TEM−1 (ampicillin resistance); strAB (streptomycin resistance); sul1 and, sul2 (sulfonamide resistance); aphA1 (kanamycin and neomycin resistance); tetA, tetB, and tetG (tetracycline resistance); as well as various dfrA genes (trimethoprim resistance) coded for resistance to these antibiotics.
The class 1 integron is widespread in clinical settings. Bacteria that are resistant to more than one antimicrobial often carry class 1 integrons and their presence is a reliable indicator of multiple antimicrobial resistance (Leverstein- van Hall et al., 2003). The class 1 integron structures identified in most clinical samples are thought to be derived from the capture of a chromosomally located class 1 integron, prevalent within environmentally abundant bacterial species belonging Frontiers in Microbiology | www.frontiersin.org to the broad β-proteobacterium group, by a Tn402 family transposon (Gillings et al., 2008;Gillings, 2014). Global spread of the class 1 integron is in part a result of the propensity for the Tn402 family of transposons to target res sites found on a wide variety of conjugative and mobile plasmids, including the widespread mercury-resistance transposon, Tn21, and the broader Tn3 family (Liebert et al., 1999;Gillings et al., 2008Gillings et al., , 2015. For this and a variety of other reasons described below, resistance genes cluster on mobile elements. Clinical class 1 integrons have subsequently evolved, typically by the loss of most of the Tn402 tniB transposition gene, and partial fusion of the qacE gene (coding for an efflux pump for disinfectants, Hall et al., 1994;Partridge et al., 2001) with the sulfonamide resistance gene sul1. The most commonly observed structure observed in clinical isolates ( Figure 1A) is comprised of a 5 conserved segment (5 -CS), a 3 -conserved segment (3 CS) and a variable region inserted in between them. The 5 -CS contains the intI1 gene encoding the class 1 integrase; the Tn402 inverted repeat, the two promoters, one required for the transcription of the integrase and the other for the gene cassettes and the attI1 recombination site. The variable region contains a diverse array of resistance gene cassettes (Partridge et al., 2009) and the associated attC sites. The 3 conserved segment (3 -CS) most frequently contains qacE 1 and sul1 genes, remnants of the transposition module and a terminal Tn402 inverted repeat. Most gene cassettes found in clinical class 1 integrons from nosocomial and livestock settings code for resistance to trimethoprim (dfrA) and aminoglycosides (aadA); despite this over 130 different gene cassettes have been described (Partridge et al., 2009) and, to date, 410 annotated variants are recognized (Tsafnat et al., 2011).
Further variation in the structure of the 3 -CS has also been described (Pan et al., 2006;Cain et al., 2010;Dawes et al., 2010;Saenz et al., 2010;Venturini et al., 2010Venturini et al., , 2013Mattiello et al., 2015). In one instance, the 3 -CS was almost completely lost and had been replaced with the compound transposon, Tn6026 (Venturini et al., 2013). The Tn6026 structure contains the betalactamase bla TEM−1 gene; the sulfonamide resistance gene, sul2; the streptomycin resistance gene strAB; the neomycin/kanamycin resistance gene, aphA1 and it is flanked by direct copies of IS26 ( Figure 1B). Tn6026 is located between attI, the insertion point of the gene cassette, and the defective tni module of a clinical class 1 integron, replacing all but the first 24 bp of the 3 -CS (Cain et al., 2010;Dawes et al., 2010;Venturini et al., 2010Venturini et al., , 2013. A sul3 variant structure associated with IS26 has also emerged over the last decade from human-and animal-sourced isolates ( Figure 1C). The first report of sul3 was from Swiss pigs and humans in Grape et al. (2003) and Perreten and Boerlin (2003). Since then, sul3 has been reported in samples from a number of countries (Pan et al., 2006;Saenz et al., 2010;Tang et al., 2011;Mattiello et al., 2015;Wyrsch et al., 2015;Moran et al., 2016). These atypical class 1 integrons (Figure 1) are often components of larger complex resistance regions, hosted on plasmids that also carry virulence genes (Venturini et al., 2010(Venturini et al., , 2013. Small mobile elements, including insertion sequences (IS1, IS26, ISEcp1c, and IS6100 are of note), Insertion Sequence Common Regions (ISCRs) and composite transposons formed from two or more IS elements, play an important role in reshaping the 3 -CS of clinical class 1 integrons (Daly et al., 2005;Dionisi et al., 2009;Venturini et al., 2010). The genes blaSHV-11, and blaSHV-12 (coding for β-lactam resistance; Ford and Avison, 2004;Dionisi et al., 2009); qnrB19 (coding for quinolone resistance; Dionisi et al., 2009); and aphA1 (coding for kanamycin/neomycin resistance; Cain et al., 2010) have been shown to be flanked by direct copies of IS26. The mechanism of action of IS26 has been the subject of several recent studies (Harmer et al., 2014;Harmer and Hall, 2015;He et al., 2015). These studies show that resistance loci containing IS26 can be hotspots for the capture of further resistance genes and other genetic cargo flanked by IS26 and demonstrates the important role of IS26 and other mobile elements in the assembly and re-assembly of novel complex resistance loci.
Currently, genomic epidemiological surveillance relies heavily on whole genome and metagenomic sequence information generated using short-read (<400 bp) sequencing platforms. A fall in the cost of high throughput, next generation sequencing with platforms that generate short-read lengths means they have become the leading technology for characterizing microbial evolution and uncultured microbial consortia, and to catalog antimicrobial resistance and virulence gene carriage. However, short-read sequencing technologies are unable to assemble complex resistance gene loci, or generate complete genome sequences, without which it is difficult to characterize mobile genetic elements rapidly and reliably. These limitations severely restrict efforts to comprehend the role of lateral gene transfer in the evolution of microbial pathogens and antimicrobial resistance, particularly in regards to the microevolutionary events that lead to the generation of novel complex resistance loci. Long-read sequencing technologies have improved significantly over the past 5 years and are beginning to address this shortfall, but are currently either cost prohibitive for large-scale genomic epidemiological surveillance or require further improvements in base call accuracy (Loman and Pallen, 2015).
Genes that code for resistance to older-generation antibiotics are often incorporated within complex resistance regions. In part, this provides a plausible explanation for why these genes persist despite enforced restrictions on prescribing of older-generation antibiotics. In the UK, strict limitations on the prescription of sulfonamide drugs were enforced in Anonymous (1995). However, functional sul genes persist in clinically important Gram negative bacteria and their frequency has not declined since the restriction was imposed (Enne et al., 2001;Bean et al., 2005Bean et al., , 2009. A similar scenario unfolded in Finland when trimethoprim was withdrawn from use for the treatment of UTI (Heikkila et al., 1990). In the Enterobacteriaceae, resistance to sulfonamides, trimethoprim, ampicillin, tetracycline, and fluoroquinolones remains high (Enne et al., 2001;Nozarian and Abdollahi, 2015;Paniagua-Contreras et al., 2015;Uzun et al., 2015;Yahiaoui et al., 2015). Carriage rates for sul1, sul2, or sul3 genes in E. coli isolated from retail meat, livestock, healthy humans, and humans with clinical disease also remains a problem (Enne et al., 2001;Blahna et al., 2006;Frank et al., 2007;Altalhi et al., 2010;Bailey et al., 2010;Saenz et al., 2010;Vogt et al., 2014;Yahiaoui et al., 2015). Carriage of sul genes is linked with the presence of both typical and atypical class 1 integrons, including genes that code for resistance to ampicillin, streptomycin, and trimethoprim (Bettelheim et al., 2003;Bean et al., 2009;Bailey et al., 2010;Soufi et al., 2011;Ewers et al., 2014;Yahiaoui et al., 2015). Combinations of genes coding for resistance to first generation antibiotics contribute significantly to the multiple antimicrobial resistance phenotypes observed in commensal E. coli (Bettelheim et al., 2003;Vinue et al., 2008Vinue et al., , 2010Bailey et al., 2010). Genetic studies that have examined resistance to a broad spectrum of antibiotics frequently report the presence of bla TEM−1 , sul1, sul2, strAB, tetA(A), and a range of aadA and dfrA genes (Soufi et al., 2011;Labar et al., 2012;Ewers et al., 2014;Yamamoto et al., 2014;Yahiaoui et al., 2015). The streptomycin resistance gene, strAB, and the sul2 sulfonamide resistance gene are also spread via small plasmids such as RSF1010 (Yau et al., 2010), and IncQ plasmids that are related to RSF1010 (Tietze et al., 1989). The strAB gene and the adjacent inverted repeat IR R , as seen in RSF1010, probably had their origins in Tn5393c, while sul2 was recruited from a CR2containing element (Yau et al., 2010). These molecular events describing the evolution of the sul2-strAB configuration observed in RSF1010 have been described previously (Yau et al., 2010). The sul2-strAB gene cluster and flanking sequences from RSF1010 have been captured onto Tn21 derivative transposons, on large and small plasmid backbones, and in chromosomal islands; they are often in association with IS26 and are distributed widely (Daly et al., 2005;Cain et al., 2010;Venturini et al., 2010Venturini et al., , 2013Labar et al., 2012;Reid et al., 2015;Roy Chowdhury et al., 2015).

E. coli IN INTENSIVE PRODUCTION ANIMALS Porcine Enterotoxigenic Escherichia coli (ETEC)
Enterotoxigenic Escherichia coli (ETEC) are the causative agents of three distinct diseases in young pigs: neonatal, pre-weaning, and post-weaning diarrhea (Fairbrother et al., 1988;Noamani et al., 2003). These diseases vary in severity and are categorized by the age at which the pig develops symptoms. Diarrhea caused by ETEC strains, generally known as enteric colibacillosis, is a problem in all countries involved in pig production, and various methods have been employed to protect pigs from ETEC infection and any subsequent diarrheal disease. These methods include, but are not limited to, treatment with antibiotics, vaccination, and feed modulation (Fairbrother et al., 2005;Virdi et al., 2013). While these protective steps have helped reduce levels of ETEC infection, particularly in neonatal and preweaning colibacillosis, antimicrobial resistance has become a global challenge that is severely limiting antimicrobials as a treatment option.
Pathogenic E. coli express a wide variety of functional molecules including toxins, adhesins, autotransporters, and invasins (Nataro and Kaper, 1998;Kaper et al., 2004), which enable the E. coli cells to damage host tissues. In porcine ETEC, these molecules, referred to as virulence factors, include heatstable (ST) and heat-labile (LT) enterotoxins, both of which can cause diarrhea via different molecular pathways within the pig's gastrointestinal tract (Nataro and Kaper, 1998). Various adhesins have evolved in ETEC and are expressed on the bacterial cell surface, allowing adhesion to the host's gastrointestinal tract. In pigs, the K88 (F4), K99 (F5), 987P (F6), F18 and F41 fimbrial adhesins all provide host intestinal cell binding, with K88, K99 and F18 fimbriae being particularly prevalent (Nataro and Kaper, 1998;Nagy and Fekete, 1999;Koh et al., 2008). Interestingly, the K88 and F18 fimbrial adhesins are often associated with pre-and post-weaning diarrhea, whilst the K99 fimbria has been associated with neonatal diarrhea in pigs (DebRoy and Maddox, 2001). These binding effects are thought to be supplemented by other common adhesins coded by genes such as fimH (Bouckaert et al., 2006) and eaeH (Sheikh et al., 2014), which are known to provide additional post-attachment binding. Chapman et al. (2006) employed multiplex PCR assays targeting 58 virulence genes to screen 52 clinical E. coli isolates from episodes of porcine neonatal and post-weaning diarrhea, and 23 commensal E. coli isolates from healthy pigs. Of the 58 genes, 17 were useful in distinguishing commensal from clinical E. coli, nine of which (iha, hlyA, aidA, east1, aah, fimH, iroN E.coli , traT, and saa) were identified for the first time in clinical porcine isolates in Australia (Chapman et al., 2006).
In an attempt to combat E. coli expressing these and other virulence genes, antimicrobials have been employed as both curative and prophylactic medicines on farms worldwide. This has placed a selection pressure on both targeted and surrounding microbial populations, and as such resistance to these antimicrobials has become extremely prevalent. For pathogenic E. coli in general, the definition and interpretation of virulence has been tempered based on results from large-scale microbial genome sequencing projects (Falkow, 1988;Pallen and Wren, 2007). ExPEC reside in the gut of their host, often in a commensal state, but have the genetic repertoire of virulence genes that enables them to cause life-threatening extraintestinal infections. Due to the degree of redundancy inherent in many ExPEC virulence genes, it remains a challenge to define virulence in ExPEC. These ExPEC-associated virulence genes have also been found in isolates sourced from humans, poultry and swine but rarely in cattle and other ruminants (Johnson et al., 2001;Tan et al., 2012;Manges and Johnson, 2015). Poultry are a wellrecognized reservoir of ExPEC that have the potential to cause human disease but the role of swine in this regard has not been investigated sufficiently.

Current Status of Global Antimicrobial Use and Its Influence on the Porcine E. coli Resistome
Rationalization of farming practices has had profound effects on the spread of resistance to antimicrobials, and consequently both human and animal health -and the problem continues to grow (FAO, 2009). An estimated 210 million kilograms of antibiotics are produced annually in China, making it the largest producer and consumer of antibiotics globally. Approximately half of the antibiotics produced are used in livestock production (Hvistendahl, 2012;Zhu et al., 2013). China is also the largest producer and consumer of pork; 690 million pigs were produced in 2011 and 38.1 kg of pork are consumed per person per year (Krishnasamy et al., 2015), now reportedly almost double the pork consumption of the USA population. China is said to contain half the world's pig population, with five times the production rate of the USA (the former largest production center) and a higher per capita meat consumption than any developing nation other than Brazil and Latin America (FAO, 2009). Trends are similar for poultry production in China (Krishnasamy et al., 2015). An estimated 34 million kilograms of antibiotics, particularly tetracyclines, sulfonamides, macrolides, and penicillins were used in Chinese pig production during 2012 (Krishnasamy et al., 2015). Combination therapies of chlortetracycline with sulfathiazole and penicillin, and chlortetracycline with sulfamethazine and penicillin, are used widely. Other individual antibiotics that are used less extensively include tylosin, chlortetracycline, oxytetracycline, bacitracin, and bambermycin amongst others (Krishnasamy et al., 2015). Furthermore, pig production in China is responsible for the production of approximately 618 billion kilograms of manure per annum , much of which is contaminated with antibiotic residues and heavy metals, particularly tetracyclines, sulfonamides and copper; some residues are present at levels in the order of 100s of milligrams per kilogram (Pan et al., 2011;Qiao et al., 2012;Zhu et al., 2013). Elevated levels of zinc and arsenic have also been detected in these studies. Of 149 unique antimicrobial resistance genes 63 were detected significantly more frequently in microbes from manure at different stages of processing on three large pig production farms in China, compared with microbes in manure from control animals that received no antimicrobials; this included genes coding for resistance to antibiotics used to treat humans (Zhu et al., 2013). Samples of pig manure from China were also rich in microbes carrying transposase genes, notably those belonging to the IS6family, predominantly IS26 transposase (Zhu et al., 2013).
In the USA, estimates suggest that the usage of tetracyclines is as high as in China, both in the combinations described above, and as stand-alone treatments. Additionally, bacitracin, carbadox, lincomycin, neomycin, penicillin, tiamulin, tilmicosin, tylosin, and viginiamycin have all been used in the USA pig production industry (Apley et al., 2012). The USA and China combined produce the majority of the world's pigs through intensive farming practices (FAO, 2009), with China producing up to five times more than the USA (Krishnasamy et al., 2015). However, other countries also have pig production systems that rely heavily on antibiotic usage. In Alberta, Canada, the same suite of antibiotics is used as in the USA, with the addition of dimetridazole (Rajic et al., 2006). A similar pattern of use has also emerged in an Australian survey of pig production, with the use of apramycin and neomycin, penicillins, macrolides, sulfonamides, tetracyclines, lincomycin and spectinomycin, tiamulin, olaquindox, ceftiofur, and dimetridazole reported (Jordan et al., 2009). A number of these antibiotics, including lincomycin, spectinomycin, and dimetridazole are also used in the treatment of infectious human diseases. In an Australia-wide study of 114 porcine-derived E. coli, resistance to numerous antibiotics was common: tetracycline (88.6%), ampicillin (71.05%), trimethoprim/sulfamethoxazole (67.5%), streptomycin (69.3%), chloramphenicol (44.74%), neomycin (35.96%), apramycin (34.21%), gentamicin (28.95%), florphenicol (26.32), cefalotin (24.56%), and spectinomycin (21.93%). Resistance to imipenem and amikacin was not detected . Of the 114 isolates evaluated 79% were classified as resistant to antibiotics from three or more different classes. Resistance to extended-spectrum cephalosporins (3%) and fluoroquinolones (1%) was detected in isolates belonging to some E. coli lineages (ST117, ST744, ST10, and ST1) albeit infrequently .
Much needs to be done to reduce the reliance on antimicrobials in food production. There is a growing body of evidence suggesting that food production may be a key driver in the evolution of multiple drug resistance in a number of bacterial species of clinical significance to human health. Livestock represent a reservoir of E. coli pathotypes with resistance to more than one antimicrobial. While pigs and poultry are significant sources of ExPEC, cattle are the major reservoirs for Shiga toxin producing E. coli (EHEC; Manges and Johnson, 2015) and atypical EPEC (Hornitzky et al., 2005;Dawes et al., 2010). The consequences for the environment of large-scale intensive animal production and a growing human population, estimated to be 9 billion by 2050 (United Nations, 2015) are not fully understood. Much remains to be done to determine how antimicrobial resistance genes are assembled on mobile genetic elements and to identify their unique genetic features. Diagnostics tests that target unique features will be useful in tracking the movement of mobile elements that are carrying complex resistance loci and to identify hotspots where they reside. Metagenomic approaches will have a major role to play in understanding the scale of the antimicrobial resistance problem, detecting and tracking emerging pathogens, and the role of gut flora in human and animal health. Studies on microbial populations from the gastrointestinal tracts of intensively reared livestock are needed to identify and characterize novel pathotypes that will undoubtedly emerge in response to the selection pressures that are commonplace in large-scale animal production facilities. Genomic surveillance studies will have an important role to play in identifying complex resistance loci, particularly how they move through different production systems and natural ecosystems (aquatic and terrestrial), and the impact they have on the carriage of multiple antimicrobial resistance genes in bacteria that colonize and infect humans and animals. Currently, there are just three complete porcine-derived E. coli chromosomal sequences published, as well as a variety of complete plasmid sequences, the majority of which have been derived from the study of isolates with closed genome sequences ( Table 2). This is at odds with the significant role that intensive pig production plays in the release of antibiotic residues into the environment, and we expect more focus to be placed on the sequencing of pig-associated E. coli isolates in the near future.

AUTHOR CONTRIBUTIONS
EW drafted an early version of the manuscript and produced the figure and tables. PR and IC contributed to several sections in the manuscript. TC and JH edited drafts of the manuscript. SD wrote most the manuscript.

FUNDING
The research in this article is part of the Ausgem partnership and supported by an Australian Research Council Linkage Project grant LP150100912.

ACKNOWLEDGMENT
We wish to acknowledge J. K. Pell Consulting for copyediting the manuscript.