Antibiotic resistant bacteria are one of the major threats to modern human and animal health care worldwide. Of major concern is the increasing trends of extended spectrum beta-lactamases (ESBL) and, plasmid-mediated AmpC β-lactamase (pAmpC)-producing Enterobacteriaceae and carbapenem-resistant Enterobacteriaceae (CRE) in both humans and animals (Bezabih et al., 2020; Hansen, 2021). In addition, these bacteria also occur in the environment, including wildlife, but the occurrence appears mainly to mirror the one described in human populations (Guenther et al., 2011; Atterby et al., 2017). Due to increasing trends of human infections caused by multidrug-resistant CRE, the antibiotic colistin, previously used mainly in food-producing animals, has received renewed attention as a treatment of human infections (Falagas et al., 2011). Resistance to colistin was previously thought only to be connected to chromosomal mutations, and despite the relatively extensive use in livestock settings, the proportion of colistin-resistance among Enterobacteriaceae from healthy animals appeared to remain low (Kempf et al., 2013). However, this changed in 2015 when increasing trends of colistin-resistant Enterobacteriaceae due to a transferable colistin resistance gene mcr-1 was described in China (Liu et al., 2016). Since the description of mcr-1 an additional nine mcr-genes have been described and occurrence has been shown in humans, animals, and the environment worldwide (Elbediwi et al., 2019).

The occurrence of ESBL- and pAmpC-producing Enterobacteriaceae, CRE and Enterobacteriaceae with mcr-genes in animals, primarily livestock, is of concern as animal populations can function as reservoirs for occurrence in humans, and transmissions from animals to humans have been indicated (Börjesson et al., 2016; Madec et al., 2017). The environment may also be a reservoir and dissemination routes for these resistant bacteria (EFSA, 2021). There also exists a risk that new strains of ESBL-, pAmpC-producing Enterobacteriaceae and CRE can emerge in the environment and bacteria may acquire previously unknown genes encoding antibiotic resistance in the environment. For example, the bla_CTX-M encoding ESBL and bla_OXA-48 encoding carbapenemase appear to be a progeny from genes in the environmental bacteria Kluyvera spp. and Shewanella spp., respectively (Humeniuk et al., 2002; Tacão et al., 2018).

Despite Sweden having a relativity low incidence of ESBL-, pAmpC-producing Enterobacteriaceae and CRE, increasing trends have been shown in human setting (Swedres-Svarm, 2020). In companion animals and livestock, except for broilers and laying hens, prevalence of ESBL and pAmpC-producing E. coli has remained low, and no carbapenemase-producing Enterobacteriaceae (CPE) has been detected (Swedres-Svarm, 2020). Today, the occurrence of ESBL/pAmpC-producing E. coli in broilers is also low but there was a high occurrence in 2010 to 2017 linked to introduction by and transmission from contaminated imported breeding stock (Nilsson et al., 2020). Occurrence of mcr-genes in Sweden appears uncommon with only a handful human cases and no cases in animals, except for mcr-9 which has been identified in clinical colistin susceptible ESBL-producing Enterobacteriaceae isolates from horses (Börjesson et al., 2020; Swedres-Svarm, 2020).

A recent EFSA report (EFSA, 2021) highlighted that feed can be contaminated by a range of resistant bacteria and is therefore a potential route for introduction of antibiotic resistant bacteria to the livestock population. Furthermore, as demonstrated by Crump et al. (2002), colonization and infections of Enterobacteriaceae in food-producing animals and outbreaks of Salmonella in humans can also be traced back to contaminated feed. Particularly imported soy products have been pointed out as high-risk (Wierup, 2017). In Sweden feed materials with high protein content, such as soya, are mainly imported from European countries, e.g., Italy or third countries like Brazil, India and China (Statistics Sweden, 2022). As ESBL-, pAmpC-producing Enterobacteriaceae, CRE and Enterobacteriaceae with mcr-genes are commonly detected in different compartments in these geographical areas, import of feed products from there might represent a potential introduction route for these bacteria to the Swedish livestock population (Sun et al., 2017; Li et al., 2019; García-Betancur et al., 2021; Lalhruaipuii et al., 2021). However, studies about the role of feed in spreading transferable resistance to extended spectrum cephalosporins, carbapenems or colistin in Enterobacteriaceae in the feed chain is lacking.

The aim of the present study was to investigate if Escherichia spp., Enterobacter spp., Klebsiella spp., or Raoutella spp. with transferable resistance to extended cephalosporins, carbapenems or colistin could be detected in the environment of feed material intake in Swedish feed mills. Any identified strains were characterized genotypically and phenotypically, and their relatedness to previously described isolates from animal and human sectors in Sweden was thereafter investigated.

The current study showed that Enterobacteriaceae with transferable resistance to extended cephalosporins, carbapenems and colistin are uncommon in the environment of Swedish feed mills, with only a single E. asburiae carrying the bla_IMI-2 gene isolated. However, there was growth observed of the investigated genera on the plates selective for extended-spectrum cephalosporin resistance (C3GR) and for colistin resistance (COL-APSE). This was not unexpected on the C3GR agar as it is designed not to inhibit growth of AmpC beta-lactamase producing strains and that the investigated genera can all carry intrinsic genes encoding beta-lactamase activity which can facilitate growth. In fact, the most commonly detected genus among the false-negatives were Enterobacter spp. (Table 1) which all carry different AmpC intrinsically (Jacoby, 2009). Furthermore, it was not also unanticipated to identify false-negative growth on the COL-APSE either, as Enterobacter spp. also can show reduced susceptibility to colistin, which was also observed for the E. asburiae carrying the bla_IMI-2 gene isolated in the current study (Landman et al., 2013; Band et al., 2016; Hong et al., 2018). Resistance to colistin and extended-spectrum cephalosporin among the investigated genera might also been influenced by chromosomal mutations which were not investigated. In addition, the false-negative growth might have been influenced by growth by bacteria (data not shown) commonly occurring in the environment, which are intrinsically resistant to the investigated antibiotics (Cantón, 2009; Cox and Weight, 2013; Gogry et al., 2021).

This is to our knowledge the first time a carbapenem resistant Enterobacteriaceae has been described in a feed mill and the first description of such a bacterium in an environment linked to animals in Sweden. When reviewing the available literature, Enterobacteriaceae with the bla_IMI gene group appear uncommon both among humans and domesticated animals globally, and the same is seen in Sweden where only seven human cases so far has been reported (Swedres-Svarm, 2020; Bonnin et al., 2021). Nevertheless, the bla_IMI-1 was described already in 1996 in two E. cloacae from patients in US, with the oldest isolate being from 1984, and since then, 21 additional variants have been described (Rasmussen et al., 1996; Bonnin et al., 2021; National Library of Medicine (NIH), 2021). As in this study, most previous studies have detected the bla_IMI genes in isolates mainly belonging to the E. cloacae complex (Bonnin et al., 2021). In addition, most descriptions are related to the outer environment and not clinical isolates, and primarily from or connected to aqueous environments. In fact, the bla_IMI-2 was also described in 1996 in an E. asburiae isolated from a US river and, as in this study, the gene was located on a transferable plasmid (Aubron et al., 2005). Three additional bla_IMI-2 E. asburiae isolates from three distant rivers were identified in the study, and since then, the bla_IMI-2 gene has been described in E. cloacae complex isolates from, Spanish river sediments and vannamei white shrimp (Litopenaeus vannamei) originating from India (Piedra-Carrasco et al., 2017; Brouwer et al., 2019). Additionally, human clinical cases with E. cloaace complex carrying the bla_IMI-2 have been reported from Czech Republic (Rotova et al., 2017), China (Yun-Song et al., 2006), Austria (Hartl et al., 2019) and South Africa (Gqunta et al., 2015). In the case report from Austria it was also reported that the infection was developed after a thermal bath visit. In addition, a recent French study also described a case related to a near-drowning in a river (Laurens et al., 2018). That study was also able to isolate E. asburiae with a bla_IMI-2 from the river water in which the near-drowning had occurred and showed that these isolates were closely related to the isolates from the patient. The bla_IMI-2 in E. cloacae complex appears therefore mainly linked, directly or indirectly, to environmental samples and appears uncommon in clinical settings. Thus, it is likely that the environment is the natural reservoir for these bacteria and the mobile elements carrying the bla_IMI-2. In our study the bla_IMI-2 isolate from the feed mill sample was compared with two E. mori and one E. ludwigii carrying bla_IMI-2 from human clinical cases, but there was no information available about any environmental connection for these patients. However, as bla_IMI-2 is located on transferable plasmids there is still a risk that this gene could be transferred to other species, or strains of E. cloacae, more prone to cause infections and spread in clinical settings, although this appears to happen very infrequently and the reason for this remains to be further elucidated. Plasmids with bla_IMI-2 have nonetheless been identified in clinical E. coli isolates in Spain and China, in Klebsiella variicola in the United Kingdom and in Pseudomonas aeruginosa in Spain (Rojo-Bezares et al., 2012; Hopkins et al., 2017; Zhang et al., 2017; Cabrera et al., 2022).

In the current study the bla_IMI-2 was located on plasmids belonging to the IncFII-family in all four investigated isolates. This mirrors the results of earlier published studies where the bla_IMI-2 appears exclusively linked to IncFII plasmids, primarily IncFII(Yp), including those where the bla_IMI-2 has been identified in isolates belonging to the E. cloacae complex species and in other Enterobacteriaceae (Rojo-Bezares et al., 2012; Hopkins et al., 2017; Piedra-Carrasco et al., 2017; Zhang et al., 2017; Brouwer et al., 2019). We therefore compared the pSB89A/IMI-2 plasmid identified in this study to earlier published plasmids and although it was shown that they shared some similar traits, the IncFII-plasmids varied greatly in both gene content and size (Figures 2, 3; Table 4). The same was true when comparing the pSB89A/IMI-2 to the plasmids identified in the three Swedish human clinical isolates. In contrast to the other plasmids the pSB89A/IMI-2 also contained multiple IncF-replicons, was the largest in size and it also contained sequence segments which showed similarity to an IncHI2 plasmid p48880_VIM-4 carrying a mcr-9 and bla_VIM-4 (Figure 3). However, the ISEae1, or a truncated version of it, was identified on all investigated bla_IMI-2 plasmids in the bla_IMI-2-imiR region and all had the toxin/antitoxins relBE/stbDE (Figures 2, 3). It was also shown that all four plasmids sequenced in this study had a variety of different genes which could possibly give the isolates fitness advantages in different environmental settings such as genes encoding amino acid productions, arsenic and metal resistance, but any function by these genes needs to be further investigated. The pSB89A/IMI-2 and the pEn701 showed the greatest similarity with most overlapping genes including that both carried the parED1 toxin/antitoxin, in addition to the relBE/stbDE (Figures 2, 3). These two plasmids also had an almost identical bla_IMI-2-imiR region with an identical ISEae2 identified, thus indicating a potential common origin (Figure 2). Of the previously published plasmids two also had highly similar bla_IMI-2-imiR regions to pSB89A/IMI-2 and pEn701, with the difference being primarily in the ISEae1 (Figure 2). Interestingly, both of these plasmids originated from South-Asia and were isolated from shrimps (Table 4). However, no information was available if the patient or the positive feed mill had any contact with South Asia. The bla_IMI-2-imiR regions in the remaining plasmids showed higher dissimilarity and did not carry the ISEae2 (Figure 2). The pGA45 and the bla_IMI-2 plasmid pJF-787 did have an ISEc36 which is related to the ISEae2 present in most other bla_IMI-2 plasmids, with the exception of the pEn542 and pEn718. So it is possible that the ISEae2 family could play a role in the transposition of bla_IMI genes, and the study describing the pJF-787 also suggested that ISEc36 could have an essential role in the dissemination of bla_IMI genes (Dang et al., 2016). ISEae1was not present in pGA45 which is present in all bla_IMI-2 plasmids included in the current study, so this insertion sequence likely plays an essential role in understanding the dissemination of the bla_IMI-2. Unfortunately, not much appears known about this IS except its connection to the E. cloace complex and that it appears to be first detected in this complex.¹ In Sweden, ISEae1 has also been identified in connection with a pathogenicity island in a group B Salmonella enterica isolated from porpoise (Phocoena phocoena) (Sandholt et al., 2021).

The bla_IMI-2-imiR region in pSB89A/IMI-2 and pEn701 was also located on a putative prophage. The same prophage was also identified in this study in the plasmid, p3442-IMI-2, isolated in Netherlands from an Indian vannamei white shrimp (Brouwer et al., 2019). Interestingly, similar prophages have previously been linked with the Shiga toxin Stx1 in E. coli (Ogura et al., 2015). In addition, the tehA gene which previously has been identified in the chromosome of E. coli was also found on the prophages in this study (Taylor et al., 1994; Turner et al., 1997). These results can indicate that prophages may play a role in the spread of bla_IMI genes in the natural environment. Although not extensively studied it has been shown that transduction can play a role in the dissemination of genes encoding resistance (Colavecchio et al., 2017; Kondo et al., 2021; Wendling et al., 2021). In the case of STEC/EHECs prophages play an important role in the transfer and expression of the stx-genes, but they also transfer other virulence factors, and have been described to increases fitness of the host (Bondy-Denomy and Davidson, 2014; Wendling et al., 2021). Despite the bla_IMI-2 gene being both connected to prophages and transferable plasmids spread to other bacterial species than the E. cloacae complex isolates appears infrequent, as previously discussed. The exact reason for the limited transmission beyond environmental E. cloacae complex strains of the bla_IMI-2 gene and its corresponding IncFII-plasmids needs to be elucidated. However, as most other narrow-hosts IncF-plasmids they are likely dependent on both host-encoded and self-encoded factors for replication and the toxin/antitoxin systems also likely influences the limited dissemination (Kopotsa et al., 2019; Fraikin et al., 2020).

In addition to being a CRE due to the bla_IMI-2 the E. asburiae isolate from the feed mill also displayed a hetero-phenotype for colistin. The isolate was shown on several retests being sometimes resistant and sometimes susceptible to colistin, with a skip-well phenotype also being observed (Table 3). However, this sort of hetero-phenotype has previously been described in E. cloacae complex isolates and is likely related to a subpopulation with an increased expression of arnB and eptA (Landman et al., 2013; Band et al., 2016; Hong et al., 2018). These two genes are responsible for the modification of lipid A, the molecular target of polymyxins, and were also detected in the chromosome of the SB89A isolate. As both arnB and eptA are located on the chromosome and are not transferable, this phenotype was therefore not further investigated in the current study.

Even though uncommon, the present study shows that the CRE occurs in the environment of feed mills and therefore feed materials may be a potential source for introduction of CREs to livestock and humans. The impact of antibiotic resistant bacteria in feed materials is likely reduced by the heat treatment that most Swedish compound feed for livestock is subjected to. However, it has been described that heat treatment performed on feed does not eliminate all bacteria (Cox et al., 1983) and occurrence of antibiotic resistant Enterobacteriaceae has been described in processed feed (da Costa et al., 2007; Fraiture et al., 2021). Since feed cannot be excluded as a route of introduction it ought to be considered as a material to screen for antibiotic resistant bacteria.

Screening of the environmental samples from the feed material intake at Swedish feed mills revealed that transferable resistance to extended spectrum cephalosporins, carbapenems or colistin has a very low prevalence. A single carbapenemase producing E. asburiae was isolated during the study and was shown to encode a bla_IMI-2 carbapenemase gene on an IncFII plasmid that had similarities, but no direct molecular epidemiological links, to bla_IMI-2 encoding plasmids from Swedish human clinical and global clinical and environmental isolates could be established. Thus, this and other studies indicate that the environment may be a reservoir for the bla_IMI-2 gene and that this gene is present on a diverse set of IncF-plasmids.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://www.ebi.ac.uk/ena, PRJEB53023.

SB, JoE, and LE conceptualized the study with all authors contributing to the final design. SB, MB, JoE, and LE secured funding for the project. SB, MB, and JeE performed the laboratory work, with SB, MB, EÖ, and OK performing the bioinformatic analyses. SB, MB, EÖ, OK, and LE were responsible for the data curation. SB wrote the first draft of the manuscript after which all authors contributed text. SB and LE finalised the manuscript. All authors contributed to the article and approved the submitted version.

This work was conducted in the framework of the Full Force project, supported by funding from the European Union’s Horizon 2020 Research and Innovation program under grant agreement no 773830: One Health European Joint Programme. In addition, internal funds were used at the National Veterinary Institute (SVA) in Sweden.

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.