Klebsiella pneumoniae is a major pathogen causing a variety of infections in humans. Colistin is the last resort antimicrobial agent to treat infections caused by K. pneumoniae including those with resistance to carbapenems. However, colistin-resistant K. pneumoniae have emerged worldwide (Olaitan et al., 2014a). A few mechanisms including both chromosomal and plasmid-borne ones have been identified to be responsible for resistance to colistin in K. pneumoniae (Olaitan et al., 2014b). Plasmid-borne colistin resistance genes including mcr-1 (Liu et al., 2016), mcr-2 (Xavier et al., 2016), and mcr-3 (Yin et al., 2017) have been found recently. In particular, mcr-1 has been identified in various species of the Enterobacteriaceae in many countries (Poirel et al., 2017).

Bacteriophages (phages) are viruses able to infect and replicate within bacteria. Phages mediate the transfer of genetic components between bacteria via transduction. Phages may have a lytic cycle or a lysogenic cycle or both. In the lytic cycle, phage genomes are replicated and are assembled into particles, which cause cell lysis and are then released. In the lysogenic cycle, phage genomes integrate into the chromosome of host bacterial cells to exist in a latent or dormant state without causing cell lysis (Feiner et al., 2015). The structure of phages typically consists of a protein head that encapsulates a DNA or RNA genome and a tail that attacks the bacterial host (Wurtz, 1992). Phage genomes vary remarkbly in form and size but usually encode products for host takeover, replication, virion assembly, or lysis (Black and Thomas, 2012). Some phages may integrate into plasmids and can therefore be transferred by the host plasmid (Oliver et al., 2005; Shin and Ko, 2015).

mcr-1 is commonly carried by plasmids of the IncI2 or IncX4 replicon type and has also been found on IncF, IncHI2, or IncP plasmids (Poirel et al., 2017). We have found a plasmid carrying mcr-1 and phage P7-like sequences, which is reported here.

Strain SCKP83 was resistant to colistin (MIC, 8 μg/ml) and had mcr-1 but no mcr-2 and mcr-3 genes. Whole genome sequencing of strain SCKP83 generated 5,247,124 clean reads, which were then assembled to 119 contigs (89 >1,000 bp) with a 50.38% GC content. Strain SCKP83 belonged to ST15, which is a relative common type of K. pneumoniae seen in China (Zhang et al., 2017b). The capsular type of strain SCKP83 was K41.

mcr-1 was carried by a 97.4 kb plasmid, pMCR_SCKP-LL83, which did not carry any additional known antimicrobial resistance genes. Despite repeated attempts, no colistin resistant transconjugants were obtained, suggesting that pMCR_SCKP-LL83 is not self-transmissible. In addition, the transformation of this plasmid into E. coli strain DH5α and a colistin-susceptible K. pneumoniae strain was unsuccessful. This suggests that this plasmid may be strain-specific or its transformation occurs at a low frequency, which could not be detected in our experiments. pMCR_SCKP-LL83 had a single pO111 plasmid replicon. Transformants containing repB and its ori were obtained. The presence of repB and ori allows the temperature sensitive vector pKC1139 to replicate at 37°C, suggesting that the replication module of pMCR_SCKP-LL83 indeed leads to the replication of this plasmid.

On pMCR_SCKP-LL83, mcr-1 was located downstream of a complete insertion sequence ISApl1. The phosphoesterase-encoding pho gene that is always located downstream of mcr-1 was truncated at its 3′-end with only 38 bp out of the 747-bp gene remaining. Surprisingly, immediate upstream of the complete ISApl1 (1,070 bp in length) lies another ISApl1 that is truncated at its 5′-end with the presence of 223 bp including an intact right-hand inverted repeat (IRR) (Figure 1). When we artificially subtract the ISApl1Δ-ISApl1-mcr-1-phoΔ region from pMCR_SCKP-LL83, the remaining artificially-joining sequence perfectly matched the ant1 gene, which encodes a putative antirepressor for antagonizing C1 repression by formation of Ant1/Ant2/C1 complex. It therefore becomes evident that the ISApl1Δ-ISApl1-mcr-1-phoΔ structure is inserted into ant1. It has been found that a single copy of ISApl1 is able to mobilize mcr-1 and pho together with itself (Li et al., 2017; Zhao et al., 2017). The insertion of ISApl1 would generate 2-bp direct target repeats (DR). However, no 2-bp DRs were present flanking the ISApl1Δ-ISApl1-mcr-1-phoΔ structure, suggesting that the formation of such a complex structure was not directly due to the insertion mediated by ISApl1. The mechanism responsible for generating the ISApl1Δ-ISApl1-mcr-1-phoΔ structure remains unclear but might have involved recombination.

A 90.9-kb region of the 97.4-kb pMCR_SCKP-LL83 was predicted as an intact phage. Neither the appearance of lysis nor the formation of plaques was observed in the UV induction. In mitomycin C induction, no plaques were formed at the tested concentrations and intervals. These results suggest that pMCR_SCKP-LL83 was indeed a plasmid. Nonetheless, the phage region on pMCR_SCKP-LL83 had 47.79% GC content, encoded 105 proteins and contained three tRNAs, i.e., tRNA-Asn, tRNA-Thr, and tRNA-Met (Table 1). pMCR_SCKP-LL83 is highly similar (72% coverage and 98% identity) to the 101.7-kb Enterobacteria phage P7 (GenBank accession no. AF503408). Phage P7 (previously called as φamp) was isolated from E. coli of human fecal flora (Smith, 1972) and exists as a nonintegrated autonomous circular plasmid that constitutes a unique compatibility group (Hedges et al., 1975). Compared with P7, pMCR_SCKP-LL83 did not have the bla_TEM−1-carrying transposon Tn3, the type I restriction-modification system EcoP7, a 4-kb invertible C-segment and a few genes, most of which encode proteins of unknown function (Table S1 in the Supplementary file and Figure 2). C-segment contains several genes encoding phage tail fibers and also determines the host specificity of the phage (Iida, 1984). In contrast, pMCR_SCKP-LL83 had a few extra genes including an unnamed type I restriction-modification system, mcr-1 and a 5-kb putative invertible C-segment (Table S1), which is highly similar (92% coverage and 99% identity) to the multiple DNA inversion region min on plasmid p15B of E. coli 15T (Sandmeier et al., 1991).

It is well known that phages can transfer genetic components between bacterial isolates, but the role of phages in disseminating antimicrobial resistance genes is still a matter of debate (Colavecchio et al., 2017; Enault et al., 2017). Nonetheless, some studies have found that phages are able to transfer genes conferring resistance to aminoglycosides (aadA, aphA1, strA, strB), β-lactams (bla_CMY−2, bla_CTX−M−9, bla_OXA−2, bla_OXA−20, bla_PSE−1, bla_TEM), chloramphenicol (floR), or tetracycline (tet(A), tet(B), tetG, tetO, tetW) via transduction (Zhang and LeJeune, 2008; Colomer-Lluch et al., 2014; Bearson and Brunelle, 2015; Ross and Topp, 2015; Shousha et al., 2015; Anand et al., 2016). In addition, a recent study has identified that two E. coli phages could promote the transformation of plasmids carrying antimicrobial resistance genes (Keen et al., 2017).

During the process of this work, mcr-1 in either complete or interrupted version has been found on plasmids containing similar phage sequences including pHYEC7-mcr1 (GenBank accession no. KX518745), pSLK172-1 (GenBank accession no. CP017632) (Bai et al., 2017), and pMCR-1-P3 (GenBank accession no. KX880944) (Zhang et al., 2017a). All of these plasmids have been recovered from E. coli at various locations of China and are highly similar (75–79% coverage, 97–99% identity, identified by BLAST; Figure 3) to pMCR_SCKP-LL83. This suggests that the phage sequence-containing plasmids represent new vehicles, which may have circulated in China, to mediate the spread of mcr-1 in addition to plasmids of IncI2, X4, F, HI2, and P types. The identification of pMCR_SCKP-LL83 from a K. pneumoniae is worrisome, suggesting that the P7 phage-like plasmids are not restricted to E. coli and may involve in the inter-species spread of mcr-1. The various locations of mcr-1 on these plasmids suggest that these plasmids may have acquired mcr-1 independently.

In the previous study on the ability of E. coli phages to promote the transformation of plasmids carrying antimicrobial resistance gene, phages, and plasmids are separate entities (Keen et al., 2017), which are different from the phage-like plasmid in the present study. As mentioned above, the conjugation and transformation of pMCR_SCKP-LL83 were unsuccessful. Among phage-like plasmids carrying mcr-1, pMCR-1-P3 was not self-transmissible and there are no data about whether it can be transferred by transformation (Zhang et al., 2017a), while pSLK172-1 was self-transmissible (Bai et al., 2017). This suggests that some phage-like plasmids may have lost the conjugative module and are therefore not self-transmissible. It is possible that these plasmids acquire genes encoding the conjugative module to become self-transmissible.

In conclusion, we identified and characterized a mcr-1-carrying P7 phage-like plasmid from a K. pneumoniae clinical isolate. Such phage-like plasmids may represent new types of vehicles to mediate the spread of mcr-1.

ZZ: designed the experiments, analyzed the data, and wrote the MS. LL: performed the experiments and analyzed the data. WZ and YF: analyzed the data.