Colistin is a polymyxin classified among the Highest Priority Critically Important Antimicrobials for human medicine by the World Health Organization (WHO), and it is considered a last resort antimicrobial for the treatment of infections by carbapenem-resistant Enterobacteriaceae in humans (Poirel et al., 2017).

In 2015, colistin resistance mediated by mcr-1, a phosphoethanolamine transferase gene located on a transferable plasmid was first reported in Escherichia coli from animals, food and patients from China (Liu et al., 2016). Since then, at least 32 countries from the five continents have found mcr-1 in E. coli isolates from different sources including humans, animals and foods (Xavier et al., 2016).

In Europe, the presence of mcr-1 was first detected in E. coli from poultry meat and humans in Denmark (Hasman et al., 2015). Subsequently, this gene was found in Enterobacteriaceae from different sources in almost all European countries, including Italy (Battisti, 2016; Cannatelli et al., 2016). In 2016, a new mcr gene, mcr-2, was described in E. coli in calves and piglets in Belgium (Xavier et al., 2016), followed by the description of three additional mobile colistin resistance genes in 2017, namely mcr-3 (Yin et al., 2017) in E. coli isolated from pig in China, mcr-4 (Carattoli et al., 2017) in Salmonella and E. coli from pigs in Italy, Spain and Belgium, and mcr-5 (Borowiak et al., 2017) in Salmonella Paratyphi B from poultry and environmental sources from Germany.

At present, eleven mcr-1 variants (KP347127, KX236309, KU934208, KY041856, KY283125, KY352406, KY488488, KY683842, KY964067, KY853650, and LC337668), one mcr-2 variant (LT598652), six mcr-3 variants (KY924928, NPZH01000177, FLXA01000011, MF463699, MG214533, and MG489958), three mcr-4 variants (MF543359, MG581979, and ERS1801979) and mcr-5 (KY807921) have been described in Enterobacteriaceae according to GenBank records (last accessed 21st February 2018).

It is clear that the epidemiology of transferable mcr-mediated colistin resistance is evolving rapidly and timely information on prevalence and molecular epidemiology of mcr-positive isolates is needed to enhance surveillance, and implement measures to prevention and to control further spread of colistin resistance. In the European Union (EU), the harmonized antimicrobial resistance (AMR) monitoring and reporting in poultry and livestock, which includes colistin susceptibility testing in E. coli and Salmonella, ensures that prevalence of colistin-resistant bacteria in a representative proportion of the food-animal population is reported from each Member State (MS) (Decision 2013/652/EU). However, the lack of molecular data limits the epidemiologic monitoring of colistin resistance (Schrijver et al., 2017).

The aim of this study is to determine the prevalence of colistin resistance, and the molecular epidemiology of mcr-mediated colistin resistance genes and their genetic environment in commensal E. coli, Extended Spectrum Beta-Lactamase (ESBL)/AmpC-producing E. coli, and Salmonella spp. in food-animals in Italy in 2014-2015.

The high diversity of transferable colistin resistance mediated by mcr genes and alleles, quickly spreading across pathogenic enterobacteria globally (Kluytmans, 2017), is an emerging challenge for treatment of Gram-negative infections due to increased occurrence of Healthcare-Associated Extremely Drug Resistant bacterial pathogens (EARS-Net, 2015).

In the present study, we found high prevalence (∼25%) of colistin resistance in both indicator commensal E. coli and ESBL/AmpC-producing E. coli in turkeys in Italy. In other Italian primary productions such as broilers, fattening pigs and bovines <12 months, colistin-resistant E. coli occurred at relatively lower, though still relevant levels.

From a (Veterinary) Public Health perspective, the spread of colistin resistance across production types is of major concern also considering that it was often associated with resistance to multiple drugs and notably also to fluoroquinolones. For instance, MDR E. coli exhibiting co-resistance to three Critically Important Antimicrobials for humans were detected in approximately 23% of turkey flocks investigated.

By using WGS, the vast diversity of mcr types and variants was evident within and across the major production types. At least three mcr-1 alleles were identified in turkeys, pigs and bovines <12 months in a 2-year-period, one of which has not been previously described. The majority of isolates harbored mcr-1.1, but mcr-1.2, mcr-1.13, mcr-3.2, mcr-4.2 and mcr-4.3 were also detected.

The core genomes of the E. coli and Salmonella isolates harboring mcr-1 were quite variable as shown by diversity of MLST types, in agreement with what has been observed in different contexts previously (Hasman et al., 2015; Bernasconi et al., 2016; Cannatelli et al., 2016; Zhao et al., 2017).

Conjugation experiments and WGS data analysis confirmed that mcr-1.1 and mcr-1.2 were located on IncX4 plasmids in E. coli and Salmonella spp. from turkey. In Europe, IncX4 plasmids have been described in association with mcr-1 in isolates of animal origin in Belgium (Xavier et al., 2016; Garcia-Graells et al., 2017) and from humans in Italy (Di Pilato et al., 2016).

The mcr-1.2 gene was first described on an IncX4 plasmid in a Klebsiella pneumoniae isolated from a human sample in the Tuscany Region (Italy) (Di Pilato et al., 2016). In our study, this variant was found on IncX4 plasmids in intestinal Salmonella and E. coli spread throughout the Italian fattening turkey production. Our observations suggest the transmission of mcr-positive IncX4 plasmids between different bacterial species, with the possibility of transmission from animals to humans, or vice versa.

The mcr-3 gene was first described on an IncHI plasmid (Yin et al., 2017), whereas the mcr-3.2 variant has been described from the whole genome shotgun sequence of a Shigella sonnei strain (NG_055523; NPZH01000177) isolated in the United States, but to our understanding, no formal description has been published to date. In our study, four ESBL/AmpC-producing E. coli isolates from bovines <12 months presented mcr-3.2. The WGS data analysis performed on these isolates did not allow us to determine the genomic location of mcr-3.2. However, only one of these isolates harbored an IncHI plasmid. A ST heterogeneity of E. coli isolates harboring the mcr-3.2 gene was evident. Further conjugation or transformation experiments are needed in order to elucidate if the mcr-3.2 found in E. coli in Italian cattle is plasmid-borne and transferable.

One E. coli from cattle was simultaneously positive for mcr-3.2 and mcr-1.1. In this isolate, mcr-1.1, was located on an IncX4 plasmid. An E. coli carrying both mcr-1.1 and mcr-3.2 and also isolated from cattle has been recently described in Spain (Hernández et al., 2017). Interestingly, all four mcr-3.2-positive E. coli isolated detected in bovines <12 months in Italy carried the ESBL bla_CTX-M-55 gene, apparently an emerging variant in veal calves in Italy. The same observation has been made by other authors (Hernández et al., 2017; Roer et al., 2017). Taken together, these features could be suggestive of a genetic linkage between the two genes. Our study, however, provides preliminary data that the mcr-3.2 and the bla_CTX-M-55 genes may not be linked on the same genetic element, and that this aspect needs further investigation.

The mcr-4.2 gene was described recently from two S. enterica Typhimurium (monophasic variant) isolated from human samples in Italy (Carretto et al., 2018). So far, this variant has not been associated to any plasmid. In our study, the mcr-4.2 gene identified in a ESBL(CTX-M-1)-producing E. coli isolated from a fattening pig, was located on a ColE plasmid, exactly the same type of plasmid in which mcr-4.1 was described by Carattoli et al., 2017. The mcr-4.3 variant was first described from a Salmonella Kedougou isolated from pigs in Spain (Rebelo et al., 2018). In our study, mcr-4.3 was harbored by an indicator E. coli from a bovine caecal sample. All the mcr-4 variants described so far have been found in Salmonella or E. coli from animals and from humans, despite their localisation on non-conjugative plasmids, similarly to what has been found with mcr-4.1 by Carattoli et al. (2017).

The mcr-1.13 gene was located on a mobile genetic element inserted in the chromosome. The mcr cassette found is different from the chromosomal mcr cassettes described so far, because of the presence of ISs other than ISplA (Snesrud et al., 2016; Li et al., 2017). By considering the high rate of self-transferability of the IncX plasmids (Sun et al., 2017) and that IncX4 is dominant in the Italian primary productions surveyed, this finding suggests that the mcr cassette may have been acquired from an IncX4 plasmid which lacks ISplA insertion sequence (Bernasconi et al., 2016; Matamoros et al., 2017; Sun et al., 2017). It is likely that the insertion occurred in different moments, since the two E. coli isolates derived from two different animal productions and presented different genome content. The ability of the mcr cassette to jump into several types of plasmids (IncI2, IncX4, IncHI2, and IncP) or into the chromosome is highly concerning also in view of the possibility of insertion into plasmids already present in MDR isolates, or even of the creation of a new mega-plasmid as occurred in Salmonella Infantis (Franco et al., 2015). This would imply that colistin resistance could be co-selected by the use of others antimicrobials. Indeed, multidrug resistance in colistin-resistant, mcr-positive isolates is a constant feature we have observed in these population-based studies. These findings highlight how rapidly mobile genetic elements can be acquired by Enterobacteriaceae and how genes can mutate in environments with high selective pressure as that occurring in the intensive farming systems. In this respect, the selection pressure exerted by the use of colistin and the complex co-selection mechanisms triggered by the overall high use of several antibiotic classes in the meat-producing industry is likely to have played a major role. For instance, the overall exposure to antibiotics in food-producing animals in Italy in 2014 was estimated around 360 mg/Population Correction Unit (PCU, i.e., mg per kilogram of biomass of farmed animals), and the exposure to colistin only was estimated around 29 mg/PCU (ESVAC, 2016). The mean, median and range of total sales in Europe (29 countries) were 108, 66, and 3 – 419 mg/PCU, respectively, while 22 out of 29 countries reported sales of colistin ≤5 mg/PCU (range of the 29 EU countries: 0.06–36.10 mg/PCU). It is well known that the intensively farmed, meat-producing animals (turkeys, broilers, pigs, bovines <12 months) are the population categories at risk of exposure to higher amounts of colistin in modern farming systems, and the results of this study strongly support the general concept that the spread of mcr-mediated colistin resistance has been favored by the semi-continuous and high exposure to colistin in these production chains.

Whether animals are an important source for human extraintestinal pathogenic E. coli (ExPEC) infections in humans is still a matter of debate (Bélanger et al., 2011). Few studies provide information and comparison on STs and plasmids harboring ESBLs or transferable AmpCs (Leverstein-van Hall et al., 2011) for both human and food-producing animal isolates. Although conclusive epidemiological evidence is still lacking, it has been proposed that some human ExPEC infections could arise from poultry and pig ExPEC reservoirs (Jakobsen et al., 2010). That being said, it is interesting to notice that at least one-fourth of E. coli isolates described in our study belong to the same STs as isolates associated with human ExPEC infections in Europe. In some cases (Leverstein-van Hall et al., 2011; Mavroidi et al., 2012; Brolund et al., 2014) they also share the same ESBL (e.g., ST69 and CTX-M-14, ST410 and CTX-M-1). As for colistin-resistant E. coli, one of the mcr-1.1 positive isolates from turkeys here described belongs to ST131, although it lacks the virulence gene markers and ESBL generally associated with the globally spread human clinical clone. Indeed, a mcr-1-positive, ESBL-negative ST-131 E. coli was also described in human bloodstream infections in Italy in 2017 (Corbella et al., 2017). Overall, these observations should be taken very cautiously, since methods for genome analysis and parameters for assessing relatedness among both core genomes and accessory genomes are quickly evolving, which implies that it may be necessary to re-evaluate any earlier conclusions on relatedness or source attribution based on partial molecular characteristics, as previously shown (De Been et al., 2014).

In conclusion, harmonized cross sectional studies at slaughter like the ones implemented by the EU represent a very important tool for a deep insight into trends and emergence of antimicrobial resistance traits and patterns in major food-borne pathogens and commensal opportunistic bacteria. Especially when occurring at high prevalence, the spread of transferable colistin resistance in E. coli (both indicator commensal and ESBL/AmpC-producing isolates) is to be considered a concern per se. Additionally, as a general principle, the high spread of resistance increases the probability of transfer of specific resistance traits also to major zoonotic pathogens, such as Salmonella spp. The hypothesis that horizontal transfer, so far, has played a major role in spread of colistin resistance among bacteria in Italian meat-producing animals is supported by the observed heterogeneity of mcr-positive E. coli. Indeed, at least in the Italian turkey productions, we demonstrated that the same transferable determinant of colistin-resistance is being carried on the same conjugative plasmid in both E. coli and major Salmonella serotypes detected in the same intensive-farming industry.

For the above reasons, quick and thorough action should be taken by the farming industry and by the Competent Authorities to drastically reduce the use of colistin in food-producing animals, especially in turkeys, following the recommendations of the European Medicines Agency (≤5 mg/PCU). EU Member States were encouraged to set stricter national targets, ideally below 1 mg/PCU colistin. We also strongly recommend reducing the overall use of all other classes of antibiotics at primary production level, in order to mitigate the effects of the complex mechanisms behind co-selection and multidrug resistance toward Critically Important Antimicrobials, in a “Consumer Protection” and a “One Health” perspective.

PA, PL, AF, RH, VB, and AB conceived and designed the experiments. PA, PL, FF, AI, and FS performed the experiments. PA, AF, FF, AI, RH, VB, and AB analyzed the data. PA, PL, AF, AC, RH, VB, and AB contributed reagents, materials, and analysis tools. PA, AF, RH, VB, and AB wrote the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The conclusions, findings and opinions expressed in this scientific paper reflect only the view of the authors and not the official position of the European Food Safety Authority.