Diversified mcr-1-Harbouring Plasmid Reservoirs Confer Resistance to Colistin in Human Gut Microbiota

ABSTRACT Colistin is an ultimate line of refuge against multidrug-resistant Gram-negative pathogens. Very recently, the emergence of plasmid-mediated mcr-1 colistin resistance has become a great challenge to global public health, raising the possibility that dissemination of the mcr-1 gene is underestimated and diversified. Here, we report three cases of plasmid-carried MCR-1 colistin resistance in isolates from gut microbiota of diarrhea patients. Structural and functional analyses determined that the colistin resistance is conferred purely by the single mcr-1 gene. Genetic and sequence mapping revealed that mcr-1-harbouring plasmid reservoirs are present in diversity. Together, the data represent the first evidence of diversity in mcr-1-harbouring plasmid reservoirs of human gut microbiota.


RESULTS
Discovery of three mcr-1-positive clinical isolates from human gut microbiota. Given the fact that current knowledge on the genetic evolution of both the mcr-1 gene and the mcr-1harbouring vectors and plasmids is extremely limited (and has even lagged), we anticipated that genetic diversity of mcr-1carrying plasmid backbones/reservoirs is probably present in gut microbiota of animals as well as human beings. Here we report RESEARCH ARTICLE crossmark that this is the case. We analyzed fecal samples collected from diarrhea patients admitted to hospitals in Shenzhen City, China, in 2015. The patients involved in this study comprised neonates (3 months old), adults (18 to 55 years old), and older persons (83 years old) (not shown). Following routine examination procedures such as 16S sequencing, we investigated 48 isolates of Escherichia coli and 27 isolates of K. pneumoniae (not shown). PCR screening for the mcr-1 gene was carried out using a pair of specific primers covering the full length of the mcr-1 coding sequence (1,626 bp). As a result, 3 of 48 clinical E. coli isolates were shown to be mcr-1 positive in our PCR assays (Fig. 2c), whereas none of the human K. pneumoniae isolates were found to be PCR positive for mcr-1 (not shown). To further confirm the identity of the acquired 3 isolates, 16S-based phylogenetic analyses (Fig. 3a), as well as Gram staining assays, were conducted ( Fig. 3b to d). In light of the fact that Liu and coworkers successfully acquired a big plasmid (pHNSHP45; 64,105 bp in length) from a pig E. coli isolate (5), we attempted to subject the three mcr-1-positive strains (designated E15004, E15015, and E15017 in Fig. 2c) to plasmid isolation by employing manual extraction with the alkaline lysis method. It seems likely that different plasmids are present in those clinical human strains in that their profiles are completely different, at least in 0.7% agarose gel electrophoresis (Fig. 2d). As expected, all three of the acquired plasmids, namely, pE15004, pE15015, and pE15017, contained the mcr-1 gene, evidenced by our PCR-based determination with these plasmids as the templates (Fig. 2e). Di-rect DNA sequencing of the three mcr-1 genes (1,626 bp) that we obtained from human clinical isolates showed that they were 100% identical to all the mcr-1 genes of diversified origins with known sequences (such as those carried by pHNSHP45 [5] and even the human gut microbiota contig from N009A [6]), indicating that the mcr-1 colistin resistance gene itself is highly conserved and seems to be under low selective pressure right now.
In vivo evidence for MCR-1-mediated colistin resistance. Structural modeling of MCR-1 showed that its architecture is similar to that of Neisseria lipo-oligosaccharide phosphoethanolamine transferase A (Fig. 2c), implying similar enzymatic mechanisms by which bacterial lipid A modification proceeds. To address the function of mcr-1 in vivo, we employed two approaches, one of which was testing the colistin tolerance of the three clinical human Escherichia isolates with a natural mcr-1positive plasmid and the other of which was visualizing the effect of colistin-susceptible E. coli strain MG1655 by regulated expression of the pure mcr-1 gene. As we expected, the prototypical strain of E. coli MG1655 as the negative control grew on the Luria-Bertani agar (LBA) plates supplemented with no more than 2.0 mg/liter of colistin ( Fig. 4a) (note that the resistance of 2.0 mg/ liter refers to a diagnostic cutoff/breakpoint [13]). In contrast, the clinical human enteric strains (E15004, E15015, and E15017) consistently exhibited appreciable growth on LBA plates with up to 16 mg/liter of colistin, and the MIC was 32 mg/liter for colistin ( Fig. 4a), which is much higher than the 8 mg/liter seen with  mcr-1-harbouring plasmids. pHNSHP45 is a plasmid from the Chinese pig isolate of E. coli with a known genome sequence (5); the other three plasmids (referred to as pE15004, pE15015,and pE15017) were isolated from three clinical E. coli strains (E15004, E15015, and E15017) collected from diarrhea patients admitted to hospitals in Shenzhen City, China, in 2015. Arrows denote the known (and/or putative) genes. A mobile element, ISApl1-mcr-1 (or an mcr-1-containing mobile element), is highlighted with a yellow background. The gray arrow illustrated with plasmid pE15004 (in panel a) represents a DNA fragment of 1,189 bp that shows 100% identity to the equivalent sequence of Salmonella enterica serovar Heidelberg plasmid pSH146-32, whereas the mcr-1 gene of plasmid pE15004 (and/or plasmid pE15015 plus pE15017) is 100% identical to the counterpart gene of E. coli SHP45 strain plasmid pHNSHP45 (5). The mcr-1-harbouring mobile element candidate from the clinical pE15017 plasmid exhibited 99% identity to that of E. coli plasmid pHNSHP45 (5). The broken arrow (tnpA*) denotes the intergenic sequence that is closely next to the tnpA gene at the 3-terminus determined in our trials, and the dots indicate that point mutations are present in tnpA* and a hypothetical protein-encoding locus (hp gene). PCR-based fine mapping suggested that the pE15017 plasmid carries all four genes (nikB, pilP, virD4, and virB4), similarly to E. coli plasmid pHNSHP45, but that such is not the case for the other two clinical mcr-1-harbouring plasmids (pE15004 andpE15015) that we have reported here. The results validated the idea that the diversified plasmid background is linked to a mcr-1 colistin resistance gene. pHNSHP45 (5). This discrepancy might be in part due to different evaluation methods. When we engineered the mcr-1 gene into an arabinose-inducible expression vector, pBAD24, for E. coli, similar scenarios were observed in our experiments. Although basal expression of mcr-1 (without the addition of any colistin into growth media) confers resistance of the recipient MG1655 strain to 4 mg/liter colistin on LBA plates, induced expression of mcr-1 in the presence of 0.2% arabinose increased its colistin tolerance to 16 mg/liter (Fig. 4b).
The results showed that all of the mcr-1 genes in these clinical isolates from gut microbiota of diarrhea patients are functional in colistin resistance. Unexpected diversity in mcr-1-harbouring plasmid reservoirs. To further figure out whether or not our plasmids harbour key genetic elements similar to those of pHNSHP45, the only mcr-1-harbouring plasmid with a sequenced genome, we designed six more pairs of primers (Table 1) targeting the following six genes (Fig. 2a): first, nikB (Fig. 2f) and pilP (Fig. 2g) separately encode a relaxase for transposal function and a type IV pilus biogenesis protein; second, virD4 (Fig. 2h) and virB4 (Fig. 2i) are two genes encoding two components (VirD4 and VirB4 with an ATPase activity) of a type IV secretion system; third, the two neighboring DNA fragments adjacent to the mcr-1 gene correspond to the tnpA transposase-encoding gene (Fig. 2j) and to a gene encoding a hy-pothetical protein (hp) (Fig. 2k), respectively. To our surprise, PCR assays showed that only the pE15017 plasmid is positive for the first four genes of those six genes (nikB [ Fig. 2f], pilP [Fig. 2g], virD4 [ Fig. 2h], and virB4 [Fig. 2i]), suggesting that pE15017 probably has high homology to pHNSHP45, whereas the remaining two plasmids (pE15004 and pE15015 [ Fig. 2c]) exhibit a different plasmid backbone in that at least those four genes are lacking ( Fig. 2f to i). A similar scenario was also noted in the PCR amplification of the hp gene adjacent to 3= end of the mcr-1 gene (i.e., the pE15017 plasmid is the only hp gene-positive one, as shown in Fig. 2k). This speculation was also supported by subsequent direct DNA sequencing of the four genes (nikB, pilP, virD4, and virB4) amplified from the pE15017 plasmid in that all of them are 100% identical to the counterparts of pHNSHP45 (Fig. 2a). In contrast, the DNA region covering tnpA, another gene that neighbors the mcr-1 gene, exhibited a differential profile in that the PCR amplicon (~1.5 kb) is present in the pE15004 plasmid, a shortened version of around 0.5 kb (tnpA*) appears in the pE15017 plasmid, but no such band exists in the pE15015 plasmid (Fig. 2j). Further sequence assembly of the DNA fragments overlapping the mcr-1 gene revealed that unexpected diversity is present the mcr-1harbouring plasmid backbones/reservoirs (Fig. 2a). First, as in the case of the mcr-1-containing mobile element (note that only a   (1,189 bp) in the pE15004 plasmid is 100% identical to that in S. enterica serovar Heidelberg plasmid pSH146-32, whereas the remaining mcr-1 part completely matches the counterpart of pHNSHP45 (5) (Fig. 2a). Second, the adjacent sequences of mcr-1 similar to those of pHNSHP45 are absent in the pE15004 plasmid (Fig. 2a). Third, the mcr-1- barboring mobile sequence (note that only a 3,286-bp sequence is acquired) in pE15004 showed 99% identity to the equivalent part of pHNSHP45 (5) in that point mutations are present in the shortened version of tnpA (namely, tnpA*) and in the hp gene (Fig. 2a). Our findings provided direct molecular evidence that the mcr-1carrying plasmid backbones/reservoirs present in the gut microbiota of diarrhea patients are of diversified/hybrid origins, which is somewhat consistent with an in silico speculation by Tse and Yuen (17) in a study of Salmonella and validates a similar hypothesis by Arcilla et al. (13).

DISCUSSION
In summary, it is reasonable to surmise that diversified transfer of plasmid-mediated mcr-1 colistin resistance might be present, and confirmation requires further epidemiological investigations. Given that (i) colistin as a veterinary medicine is extensively applied in agricultural (poultry and livestock) production worldwide, (ii) widespread distribution of the mcr-1 colistin resistance gene by transposal elements/plasmids occurs in animals, meat/ food samples, and human gut microbiota, (iii) the relatively widespread occurrence of mcr-1 in isolates from no fewer than 10 countries is known right now, and (iv) colistin resistance represents a major breach in our last line of defense against multidrugresistant bacterial pathogens, it might greatly necessary to clinically monitor the diversified mcr-1-harbouring plasmids/ reservoirs, reevaluate the efficacy (safety) of colistin in veterinary use, and formulate a comprehensive strategy to fight against plasmid-mediated mcr-1 colistin resistance, especially in pandrug-resistant Gram-negative bacterial strains (19).

MATERIALS AND METHODS
Bacterial isolations and identification. Fecal samples were routinely collected from diarrhea patients admitted to hospitals in Shenzhen City, China, in 2015. Luria-Bertani (LB) liquid media (and/or solid agar plates) were applied to isolate the enterobacterial species. To initially determine the bacterial identity, the acquired bacteria were routinely subjected to biochemical tests as well as to colony PCR assays with 16S-specific primers. Gram staining assays were routinely performed. As a result, either Escherichia coli or Klebsiella pneumoniae was predominantly assigned to the gut microbiota of the diarrhea patients.

DNA manipulations.
To probe the presence of the mcr-1 gene, all of the bacterial isolates harvested (75 in total [48 for E. coli and 27 for K. pneumoniae]) were screened via PCR-based diagnostics with mcr-1specific primers ( Table 1). The mcr-1-positive bacteria were further subjected to plasmid isolations with the manual alkaline lysis method. Then, the plasmid-harbouring isolates were dissected finely using multiplex PCR with six pairs of specific primers targeting six different loci (such as virD4 and virB4, encoding a type IV secretion system) ( Table 1). All of the PCR-amplified DNA fragments were verified by direct DNA sequencing and analyzed by ClustalW-aided multiple-sequence alignments. The mcr-1 gene was cloned into the arabinose-inducible expression vector pBAD24 in E. coli (20), giving the recombinant plasmid pBAD24::mcr-1.
Determination of colistin MIC. Mid-log-phase bacterial cultures in a dilution series were spotted on LBA plates supplemented with various levels of colistin (0, 0.5, 1.0, 2.0, 4.0, 8.0, 16.0, and 32.0 mg/liter) and kept at 37°C overnight. In the assays of inhibitory colistin concentrations, the findings for all three of the mcr-1-containing strains (E15004, E15015, and E15017) were confirmed, whereas the negative control corresponded to MG1655, the prototypical strain of E. coli. To further examine the function of the single mcr-1 gene in vivo, we monitored the resistance of the colistin-susceptible MG1655 strain conferred by expression of arabinoseinducible, pBAD24::mcr-1-borne mcr-1.
Bioinformatics analyses. The acquired 16S sequences in full length were subjected to phylogenetic analyses using the webserver of Clustal Omega. The coding sequences of the mcr-1 gene and the other target genes such as virD4 and virB4 were aligned with their counterparts with known sequences using BLAST.