Genomic and Molecular Characterization of Clinical Isolates of Enterobacteriaceae Harboring mcr-1 in Colombia, 2002 to 2016

ABSTRACT

Polymyxins are last-resort antimicrobial agents used to treat infections caused by carbapenem-resistant Enterobacteriaceae. Due to the worldwide dissemination of polymyxin resistance in animal and human isolates, we aimed to characterize polymyxin resistance associated with the presence of mcr-1 in Enterobacteriaceae and nonfermenter Gram-negative bacilli, using isolates collected retrospectively in Colombia from 2002 to 2016. A total of 5,887 Gram-negative clinical isolates were studied, and 513 were found to be resistant to the polymyxins. Susceptibility to colistin was confirmed by broth microdilution for all mcr-1-positive isolates, and these were further subjected to whole-genome sequencing (WGS). The localization of mcr-1 was confirmed by S1 pulsed-field gel electrophoresis (S1-PFGE) and CeuI-PFGE hybridization. Transferability was evaluated by mating assays. A total of 12 colistin-resistant isolates recovered after 2013 harbored mcr-1, including 8 Escherichia coli, 3 Salmonella enterica serovar Typhimurium, and 1 Klebsiella pneumoniae isolate. E. coli isolates were unrelated by PFGE and belonged to 7 different sequence types (STs) and phylogroups. S. Typhimurium and K. pneumoniae isolates belonged to ST34 and ST307, respectively. The mcr-1 gene was plasmid borne in all isolates but two E. coli isolates which harbored it on the chromosome. Conjugation of mcr-1 was successful in 8 of 10 isolates (8.2 × 10⁻⁵ to 2.07 × 10⁻¹ cell per recipient). Plasmid sequences showed that the mcr-1 plasmids belonged to four different Inc groups (a new IncP-1 variant and the IncFII, IncHI1, and IncH families). Our results indicate that mcr-1 is circulating in clinical isolates of colistin-resistant Enterobacteriaceae in Colombia and is mainly harbored in transferable plasmids.

INTRODUCTION

Antimicrobial resistance is a serious threat to global public health. The emergence of multidrug-resistant (MDR) Gram-negative bacteria represents a serious risk for hospitalized patients with severe infections. Unfortunately, in the last decades, development of new antibiotics has been scarce (1), and old antibiotics (such as the polymyxins) have been resurrected as last-resort options to treat infections caused by MDR Gram-negative bacteria (i.e., carbapenem-resistant Enterobacteriaceae and other Gram-negative rods) (2, 3). Furthermore, in 2012, the World Health Organization expert meeting on critically important antimicrobials for human health included colistin as a critically important antimicrobial for human medicine (4). Although resistance to polymyxins has been characterized as the result of nontransferable chromosomal mutations resulting in modification of lipid A (5), a recent Chinese report of transferable, plasmid-mediated colistin resistance caused by the mcr-1 gene (6) suggests that this resistance determinant can spread worldwide from isolates of human, environmental, and food sources (7). Interestingly, there is evidence that mcr-1 has been circulating since before the initial description in China, and it has now been reported in many countries around the world (8). The mcr-1 gene has been described for several countries of the Americas (5) and for fecal Escherichia coli isolates recovered from travelers visiting South America (9).

In May 2016, Colombia reported to the Pan-American Health Organization (PAHO) the presence of 4 Enterobacteriaceae isolates (3 Salmonella enterica serovar Typhimurium and 1 E. coli isolate) carrying mcr-1 (10). Following this report, we conducted a retrospective surveillance study with the aim of detecting and characterizing the presence of mcr-1 in Colombian clinical isolates of Enterobacteriaceae recovered since 2013.

RESULTS AND DISCUSSION

Presence of the mcr-1 gene among polymyxin-resistant Enterobacteriaceae and nonfermenter Gram-negative bacillus isolates in Colombia.

After screening a total of 5,887 clinical isolates belonging to the Enterobacteriaceae and nonfermenter Gram-negative bacilli, recovered in Colombia from 2002 to 2016, we found 513 (8.7%) isolates that exhibited polymyxin resistance according to the EUCAST colistin and CLSI polymyxin B criteria. Among these, 443 (7.5%) isolates were Enterobacteriaceae and 70 (1.19%) isolates were nonfermenter Gram-negative bacilli (see Data Set S1 in the supplemental material). The regions of Antioquia, Valle del Cauca, and Santander showed the highest frequencies of colistin resistance (ranging from 8.2% to 36.8%). Among the resistant isolates, only 12 (2.3%) Enterobacteriaceae isolates were positive for the mcr-1 gene by PCR. Note that a limitation of our study is that mcr-2 (11), mcr-3 (12), and mcr-4 (13) variants were not evaluated by PCR, and as a consequence, the number of colistin-resistant isolates harboring mcr gene variants may have been underestimated. Interestingly, the frequency of colistin resistance in Enterobacteriaceae from Colombia seems to be higher than that reported for Brazil (3.8%) (14). Indeed, our results contrast with those reported by Bradford et al. (15), which indicated that the prevalence of colistin resistance in Enterobacteriaceae in Latin America was 1.5%, with Colombia classified in the group with prevalences below 2%, in which only resistance in Enterobacter spp. and Klebsiella pneumoniae had been reported.

The mcr-1-positive isolates were recovered in 9 hospitals located in 7 distinct regions of Colombia (mainly central and western areas that include Antioquia, Boyacá, Caldas, Cauca, Santander, and Valle del Cauca and the capital city, Bogotá) (Fig. 1). In our study, the 12 mcr-1-positive isolates included 8 E. coli (3.5% of 223 isolates), 3 S. Typhimurium (23% of 13 isolates), and 1 K. pneumoniae (1.5% of 66 isolates) isolate (Data Set S1). Although the prevalence of mcr-1 in clinical isolates has not been documented extensively, our results indicate that the frequency of mcr-1 in clinical isolates in Colombia is comparable to that reported by two studies in China, with values of 1% for E. coli and <1% for K. pneumoniae (16, 17). Note that isolates carrying mcr-1 were identified only since September 2013 (even though isolates in our collection date back to 2002), suggesting a recent introduction of this resistance mechanism into the country (Fig. 1A). The E. coli isolates RA229.16, RA346.16, and RA453.16 were recovered from different patients in the same hospital in Santander, while isolates RA381.16 and RA432.16 were recovered from the same patient in a hospital located in Antioquia (Fig. 1A; Data Set S1).

FIG 1.

Countrywide screening for detection of mcr-1-positive Enterobacteriaceae isolates. (A) Map depicting the 24 geographical regions (departments) included in polymyxin resistance surveillance in Colombia from 2002 to 2016. The colistin resistance prevalence (percentage) is indicated for each department, and numbers of resistant isolates harboring mcr-1 are indicated in orange in the pie charts. (B) Genetic relationships among the mcr-1-positive isolates were established using PFGE. The dendrogram was clustered by use of the UPGMA algorithm using Dice coefficients, with a 1.5% band position tolerance and optimization. In addition, the information for Klebsiella pneumoniae was described. The sequence type (ST) for E. coli was determined by use of the scheme of Wirth et al. (78), that for Salmonella according to the scheme of Achtman et al. (79), and that for K. pneumoniae according to the scheme of Diancourt et al. (80). Phylogroups were assigned following the methodology described by Clermont et al. (81). For E. coli, the cutoff for a clonal relationship was established at >90%, whereas that for S. Typhimurium was >80%.

Genomic characterization of mcr-1-positive clinical isolates.

We performed whole-genome sequencing (WGS) of all mcr-1-positive isolates. Species confirmation and sequence type (ST) and resistome determinations were performed in silico (Fig. 1B; Table 1). In order to determine alternative colistin resistance mechanisms, we explored the presence of amino acid changes in 13 predicted proteins previously associated with development of colistin resistance (Table S2) (18–20). We confirmed the presence of mcr-1 in all isolates, but we did not find any additional changes in proteins previously associated with colistin resistance (5, 21, 22), suggesting that mcr-1 was solely responsible for the increased MICs for these isolates.

TABLE 1.

mcr-1-positive isolates and transconjugants analyzed in this study

Species	Isolate/donore	Colistin MIC (μg/ml)a		mcr-1 localization	Plasmid family or families	Antibiotic resistance gene(s)	Conjugation frequencyc	Transconjugant antibiotic resistance gene(s)	Transconjugant plasmid family or families	Related plasmidd
		Donor	Tcb
E. coli	CID6770	4		190-kb plasmid	Col156, IncFII, IncFIA, IncFIB, IncHI1A, IncHI1B, IncI2, IncX1, new IncP-1 clade	mcr-1, blaTEM-1, dfrA1, tetB	Nontransferable			pHNFP671
	RB634	4		190-kb plasmid	IncHI1A, IncHI1B, IncFIA, IncFIB, IncFII, IncI1, IncX1	mcr-1, aadA5, blaCMY-2, dfrA17, floR, sul2, tetA	Nontransferable			pH226B
	RA229.16	8	4	60-kb plasmid	IncFIB, IncY, new IncP-1 clade	mcr-1	2.83 × 10−3	mcr-1	New IncP-1 variant	pHNFP671
	RA381.16	4		Chromosome	IncFIA, IncFIB, IncFIC, IncFII, IncX1	mcr-1, blaTEM-1, qnrB19, tetA	Nontransferable
	RA346.16	4	4	90-kb plasmid	Col, IncFIB, IncFII, IncI1, IncQ1	mcr-1, aadA1, aadA5, blaSHV-12, blaTEM-1, catA1, dfrA17, qnrB19, strA (aph(3″)-Ib)), strB (aph(6)-Id), sul1, sul2	2.9 × 10−3	mcr-1, aadA1	IncFII	pO26
	RA372.16	4	4	50-kb plasmid	IncFIA, IncFIB, IncFII, IncQ1, new IncP-1 clade	mcr-1, aadA, blaTEM-1, dfrA1, floR, mph(B), strA, strB, sul1, sul2, tetA	2.07 × 10−1	mcr-1, aadA2	New IncP-1 variant	pHNFP671
	RA402.16	4	4	50-kb plasmid	Col (BS512), IncFIA, IncFIB, IncFII, IncX1, new IncP-1 clade	mcr-1, aadA1, dfrA1, strA, sul3, tetB	1.72 × 10−4	mcr-1, aac(6′)-Ib, aac(6′)Ib-cr	New IncP-1 variant	pHNFP671
	RA432.16	4		Chromosome	IncFIA, IncFIB, IncFIC, IncFII, IncX1	mcr-1, blaTEM-1, qnrB19, tetA	Nontransferable
S. Typhimurium	S-1257.15	4	2	50-kb plasmid	ColpVC, IncQ1, new IncP-1 clade	mcr-1, blaTEM-1, qnrB19, strA, strB, sul2, tetB	3 × 10−2	mcr-1	New IncP-1 variant	pHNFP671
	S-1454.15	8	4	220-kb plasmid	IncFIA, IncHI1A, IncHI1B	mcr-1, blaTEM-1, floR, qnrB19, tetA, tetB	8.2 × 10−5	mcr-1, blaTEM-1, floR, qnrB19, tetA, tetB	IncFIA, IncHI1A, IncH1B	pH226B
	S-356.16	8	4	220-kb plasmid	IncFIA, IncQ1, IncHI1A, IncHI1B	mcr-1, blaTEM-1, qnrB19, strA (aph(3″)-Ib)), strB (aph(6)-Id), sul2, tetB	2.1 × 10−4	mcr-1, tetB	IncHI1A, IncHI1B, IncF1A	pH226B
K. pneumoniae	RA453.16	4	4	300-kb plasmid	IncFIB	mcr-1, aac(3)-IIa, aac(6′)Ib-cr, aadA2, blaCTX-M-11, blaOXA-1, blaSHV-28, blaTEM-1, dfrA12, fosA, mph(A), oqxA, qnrB1, strA, strB, sul1, sul2, tetA	3.4 × 10−5	mcr-1, sul1, mph(A), dfrA12, aadA2, tet(A)	IncFIB	pNDM-MAR

^aDetermined by broth microdilution susceptibility testing.

^bTc, transconjugant.

^cEstimated efficiency of transfer to E. coli J53 Az^r, expressed as the number of transconjugants per receptor. Data are medians for three independent experiments.

^dBest-hit plasmid from BLASTN search of the contig harboring the mcr-1 gene in the transconjugant isolates.

^eCID, isolate recovered by CIDEIM surveillance; RB, RA, and S names, isolates recovered by INS surveillance.

E. coli isolates harboring mcr-1 belonged to a variety of STs (ST10, -37, -101, -744, -1263, -3056, and -6627) grouped within clonal complex 10 (CC10) and comprised four phylogroups (A, B1, C, and E) (Fig. 1B). Note that CC10 is a widespread E. coli lineage that is found in animals and humans and causes infections (Fig. S1) (23) but was not previously reported for human clinical isolates in Colombia (24). However, ST10 (CC10) and ST101 have been found in E. coli isolates recovered from a Colombian poultry chain, suggesting possible dissemination of mcr-1 to humans through the food chain (25), similar to what has been reported before in Europe (26–31). In order to explore possible genetic relationships between the Colombian E. coli isolates and other mcr-1-carrying E. coli isolates reported worldwide, we built a phylogenetic tree using the entire genomes of the mcr-1-carrying isolates, including the currently available genomes of E. coli strains harboring mcr-1 in the NCBI database. Our phylogenomic analyses showed that the mcr-1-positive E. coli isolates are distributed in different lineages, suggesting that they are not closely related to each other (Fig. 2). Interestingly, we found three E. coli isolates that exhibited an MDR phenotype (Table 2): (i) E. coli RB634 was resistant to third- and fourth-generation cephalosporins and other beta-lactams, trimethoprim-sulfamethoxazole (SXT), and colistin; (ii) RA346.16 exhibited resistance to third-generation cephalosporins, ciprofloxacin (CIP), trimethoprim-sulfamethoxazole, gentamicin (GEN), and colistin; and (iii) RA372.16 showed resistance to ciprofloxacin, trimethoprim-sulfamethoxazole, and colistin. The bla_CMY-2 and bla_SHV-12 genes were identified in RB634 and RA346.16, and in addition, E. coli RB634 was also resistant to carbapenems and monobactams; however, we were unable to confirm production of carbapenemase by phenotypic assays, including the CarbaNP test, the modified Hodge test, and synergy tests with ethylenediaminetetraacetic acid (EDTA) and 3-aminophenylboronic acid (APB) disks. Note that the three isolates of E. coli belonged to different STs (ST3056, -744, and -1263, respectively) (Table 1). Furthermore, we did not identify genes coding for carbapenemases in the genomic analyses (Tables 1 and 2). The coexistence of mcr-1 with genes coding for extended-spectrum beta-lactamases (ESBLs) and plasmidic AmpC in the same isolate has previously been reported in the Americas (32–38).

FIG 2.

Maximum likelihood phylogenetic tree for E. coli isolates harboring mcr-1 from around the world, built with RAxML. Numbers on branches represent bootstrap support. Data shown in bold are for the Colombian isolates. ST, sequence type; SLV, single-locus variant.

TABLE 2.

Susceptibility profiles of mcr-1-positive clinical isolates as determined by Trek Diagnostic Systems^a

Species or isolate group	Isolate	MIC (μg/ml)
		CEP	FOX	CTX	CAZ	FEP	ATM	PTZ	SAM	IMP	MEM	ERT	DOR	CIP	GEN	AMK	SXT	TGC
E. coli	RB634	>32	>32	>64	>32	8	>32	>132/4	>32/16	2	≤1	>4	1	≤0.06	≤1	≤4	>8/152	≤0.5
	RA229.16	≤4	8	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	≤0.06	≤1	≤4	≤1/19	≤0.5
	RA346.16	>32	8	2	16	≤1	32	≤8/4	8/4	≤1	≤1	≤0.25	≤0.5	>4	>16	≤4	>8/152	≤0.5
	RA372.16	8	8	≤1	≤1	≤1	≤2	≤8/4	16/8	≤1	≤1	≤0.25	≤0.5	>4	≤1	≤4	>8/152	≤0.5
	RA381.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	8/4	≤1	≤1	≤0.25	≤0.5	0.25	≤1	≤4	≤1/19	≤0.5
	RA402.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	0.25	≤1	≤4	>8/152	≤0.5
	RA432.16	≤4	8	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	0.25	≤1	≤4	≤1/19	≤0.5
	CID6770	>32	8	≤1	≤1	≤1	≤2	>132/4	>32/16	≤1	≤1	≤0.25	≤0.5	≤0.06	≤1	≤4	≤1/19	≤0.5
S. Typhimurium	S-1257.15	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	16/8	≤1	≤1	≤0.25	≤0.5	0.5	≤1	≤4	≤1/19	≤0.5
	S-1454.15	8	16	≤1	≤1	≤1	≤2	≤8/4	32/16	≤1	≤1	≤0.25	≤0.5	1	≤1	≤4	≤1/19	1
	S-356.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	16/8	≤1	≤1	≤0.25	≤0.5	1	≤1	≤4	≤1/19	≤0.5
K. pneumoniae	RA453.16	>32	≤4	32	16	8	16	≤8/4	32/16	≤1	≤1	≤0.25	≤0.5	>4	16	≤4	>8/152	≤0.5
Transconjugants	Tc RA229.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5
	Tc RA346.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5
	Tc RA372.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5
	Tc RA402.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5
	Tc S-1257.15	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5
	Tc S-1454.15	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	8/4	≤1	≤1	≤0.25	≤0.5	0.25	≤1	≤4	≤1/19	≤0.5
	Tc S-356.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5
	Tc RA453.16	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≥8/152	≤0.5
E. coli recipient	J53 Azr	≤4	≤4	≤1	≤1	≤1	≤2	≤8/4	≤4/2	≤1	≤1	≤0.25	≤0.5	<0.06	≤1	≤4	≤1/19	≤0.5

^aCEP, cephalothin; FOX, cefoxitin; CTX, cefotaxime; CAZ, ceftazidime; FEP, cefepime; ATM, aztreonam; PTZ, piperacillin-tazobactam; SAM, ampicillin-sulbactam; IMP, imipenem; MEM, meropenem; ERT, ertapenem; DOR, doripenem; CIP, ciprofloxacin; GEN, gentamicin; AMK, amikacin; SXT, trimethoprim-sulfamethoxazole; TGC, tigecycline. Antimicrobial susceptibility was evaluated by Trek Diagnostic Systems. Bold values are nonsusceptible MICs (intermediate and/or resistant). Interpretation of MIC values was performed according to the CLSI 2016 guidelines (83) and, for tigecycline, according to FDA guidelines.

In contrast to the findings for E. coli, the three S. Typhimurium isolates belonged to the same genetic lineage. Indeed, all isolates belonged to ST34 and exhibited low genetic diversity by pulsed-field gel electrophoresis (PFGE) (similarities of ≥84%) (Fig. 1B). The PFGE patterns of these isolates were unrelated to those previously reported in the country and were recovered in different regions (39), suggesting a recent introduction of this clone. The three isolates of S. Typhimurium exhibited an MDR phenotype. Indeed, they were all resistant to tetracycline, ampicillin (AMP), nalidixic acid (NAL), and colistin (Data Set S1). Additionally, isolates S-1454.15 and S-356.16 were resistant to ciprofloxacin, and S-1454.15 was resistant to chloramphenicol, cephalothin (CEP), and cefoxitin (FOX). Using WGS, three common resistance genes were identified in the S. Typhimurium isolates, including bla_TEM-1, qnrB19, and tetB (Tables 1 and 2). Note that S. Typhimurium ST34 isolates containing mcr-1 have previously been reported among clinical isolates from England (40) and Portugal (41). Recently, dissemination of ST34 associated with mcr-1 was reported for animal samples in China (42). Some ST34 isolates are considered high risk for fluoroquinolone resistance due to the presence of the oqxAB and aac(6′)-Ib-cr genes (43, 44).

Finally, the only K. pneumoniae isolate found to carry mcr-1 belonged to ST307 (a non-CG258 ST), was positive for the ESBL test, and harbored bla_CTX-M-11 and bla_SHV-28. As MDR Klebsiella spp. are endemic in Colombia, this isolate exhibited resistance to at least 8 different families of antibiotics (Tables 1 and 2). Note that isolates belonging to this ST have been associated with strains producing carbapenemases (45–48). In Colombia, ST307 isolates producing the KPC carbapenemase were identified in 2013 and 2014, in two hospitals in Antioquia (49); however, bla_KPC was absent in the isolate from the present study.

Genetic environment and transferability of mcr-1.

To characterize the genetic localization of the mcr-1 gene, we performed PFGE of S1 and CeuI genomic DNA restriction digests, followed by Southern blotting and hybridization assays with an mcr-1 probe. Our results indicated that all but two isolates had mcr-1 located on plasmids (Fig. 3; Fig. S2). For the two other mcr-1-positive E. coli isolates, the gene was located in a chromosomal location, similar to what was previously reported for E. coli from food samples in Germany (50, 51) and for clinical isolates from China (17, 31). For the 10 isolates in which the mcr-1 gene was plasmid borne, eight of the plasmids were readily transferred through mating experiments (Table 1). We were not able to transfer mcr-1-containing plasmids from two E. coli isolates under the tested conditions. Plasmid profiles showed that all isolates carried more than one plasmid (up to four), with plasmid sizes ranging from 30 kb to 300 kb, and several incompatibility groups were identified (Table 1 and Fig. 3; Fig. S2). For the mcr-1 conjugative plasmids with successful transfer, the frequencies of transfer ranged from 10⁻¹ to 10⁻⁵ cell per recipient. No additional ESBL or AmpC resistance genes were cotransferred (Table 1).

FIG 3.

Genetic localization of mcr-1 in clinical isolates and their J53 transconjugants (TC). Plasmid locations were analyzed by S1 nuclease PFGE. (A) Arrows indicate where the mcr-1 probe hybridized. White arrows indicate plasmids, and red arrows indicate chromosomes. (B) Chromosomal locations were confirmed by I-CeuI endonuclease PFGE. (Left) Electrophoresis gel; (center) hybridization with 16S rRNA probe; (right) hybridization with mcr-1 probe.

Analyses of predicted mcr-1 plasmid sequences showed that four different incompatibility families were present: a new IncP-1 variant, IncFII, IncHI1, and IncH. Four of the eight transferable mcr-1 plasmids (three from E. coli and one from S. Typhimurium) exhibited high DNA sequence identity (95% over about 50 kb) to pHNFP671. This plasmid belongs to a new IncP-1 variant (accession number KP324830) (Table 1; Fig. S4A) (52), was recovered from an E. coli isolate (FP671) in Guangzhou, China, and did not harbor mcr-1. For the remaining four transferable plasmids that carried mcr-1 (two in S. Typhimurium), we found high sequence identity (96% over 167 kb) to an E. coli (strain H226) plasmid carrying mcr-1 (designated pH226B), previously recovered in Europe from imported vegetables from Thailand (53). E. coli RA346.16 carried mcr-1 on an IncFII plasmid similar to the virulence plasmid pO26 of enterohemorrhagic E. coli (EHEC) (accession number GQ259888) (54), with a DNA sequence identity of 97% (over 59 kb). pO26 is a multiantibiotic-resistant EHEC virulence plasmid isolated from an Australian patient in 2010 and, to the best of our knowledge, has not been associated with mcr-1 (Fig. S4C). Finally, the mcr-1 plasmid in the K. pneumoniae isolate was similar to the recently described pNDM-MAR plasmid (DNA identity of 95% over 187 kb), recovered from patients in Morocco with two novel variants of the IncFIB and IncHI1B incompatibility groups (accession number JN420336.1) (55) (Table 1; Fig. S4D). We inferred by in silico prediction that the Inc groups of nontransferable plasmids belonged to the IncHI1 family (RB634) and the new IncP-1 variant (CID6770). In addition, a comparative BLAST-based analysis using the contigs obtained from transconjugant (Tc) assemblies against reference plasmids suggested that the backbone and many of the putative conjugative genes are conserved among our isolates and the reference plasmids (Fig. S4). Note that the Inc groups of the mcr-1-harboring plasmids identified in Colombian isolates were different from the IncX4 and IncI2 types previously reported in neighboring countries, such as Brazil (33), Venezuela (32), and Argentina (56), and in China (16, 17). Our data suggest that mcr-1 can be carried by different plasmids and is likely to disseminate horizontally among Enterobacteriaceae in distinct genetic platforms in Colombia.

The genetic environment of mcr-1 on pHNFP671-related plasmids indicated the presence of the complete element ISApl1-mcr1-pap2 flanked by the corresponding inverted repeats (IRL and IRR, respectively) (Fig. 4) (6). In one plasmid (pRA372.16-190kb) (Fig. 4), ISApl1 was absent, similar to previous reports (57). A variation in the nucleotide sequence of IRR2 that was not previously reported (58) was found in pS1257-50kb. Interestingly, a new point mutation at position 689 (T to A) was identified in the ISApl1 insertion element in the pH226B-like mcr-1 plasmid. Taken together, our results suggest that highly transposable elements carry mcr-1 within Colombian Enterobacteriaceae.

FIG 4.

Genetic environment of the mcr-1 gene. The schematic representation shows the mcr-1 element (gray arrows) inserted into the pHNFP671 backbone. The insertion sites for the mcr-1 element are shown with colored vertical bars, and sequences are indicated with the same colors, as follows: blue, IRR2***, reported previously (57); green, IRL (57, 82); red, IRR (57); and yellow, IRR2****, described in this study.

Conclusions.

Overall, our results show a relatively high frequency of colistin resistance in clinical Enterobacteriaceae isolates in Colombia, explained in part by the presence of the transferable resistance gene mcr-1 since 2013. Our findings indicate that transposition of mcr-1 is the mechanism of mobilization among strains of different genetic backgrounds, raising the possibility of rapid dissemination of this gene. Our results underline the importance of continuous surveillance of colistin-resistant Gram-negative isolates, including in intensive care unit (ICU) patients and animal and food products, in a country where carbapenem resistance is endemic.

MATERIALS AND METHODS

Countrywide surveillance, screening of polymyxin resistance, and mcr-1 detection.

Two collections of Gram-negative clinical isolates recovered from independent countrywide surveillance networks (Grupo de Microbiología, Instituto Nacional de Salud [INS], and Grupo de Resistencia Bacteriana y Epidemiología Hospitalaria, International Center for Medical Research and Training [CIDEIM]) were screened for polymyxin resistance. A total of 5,887 Gram-negative isolates recovered between 2002 and 2016 in 173 hospitals were evaluated for colistin or polymyxin B resistance (Fig. 1A; see Data Set S1 in the supplemental material). Species that exhibited intrinsic resistance to colistin, such as Proteus spp., Morganella spp., Providencia spp., Serratia marcescens, and the Burkholderia cepacia complex, were excluded from the study.

For the first collection of isolates (CIDEIM, Cali, Colombia), a total of 2,649 clinical isolates recovered from 17 hospitals in seven regions of Colombia (Atlántico, Valle del Cauca, Antioquia, Santander, Risaralda, Tolima, and Bogotá [capital city]) were screened using two main strategies. First, 654 isolates which lacked colistin MIC data were screened using Mueller-Hinton agar supplemented with 4 μg/ml of polymyxin B. For the second screening strategy, we used MIC data available by the broth microdilution method (Sensititre customized panels; Trek Diagnostic Systems) (1,995 isolates). MICs were interpreted using the EUCAST colistin/polymyxin B breakpoint (2 μg/ml) (59) and CLSI polymyxin B criteria (the latter for Acinetobacter baumannii). All isolates capable of growing on Mueller-Hinton agar supplemented with polymyxin B at 4 μg/ml (n = 214) or exhibiting MICs of >2 μg/ml (n = 47) were tested by PCR for the presence of the mcr-1 gene by use of primers CLR5-F (5′-CGGTCAGTCCGTTTGTTC-3′) and CLR5-R (5′-CTTGGTCGGTCTGTAGGG-3′), using the conditions described previously by Liu et al. (6). E. coli OPS229 was used as a positive control for the PCR assay. The second collection encompassed 3,238 isolates collected by the Instituto Nacional de Salud from May 2013 to August 2016 in 171 hospitals in 24 Colombian regions. The colistin MICs were determined by use of different automated systems (Phoenix, Vitek, or Microscan). The isolates were screened using available MIC data. A total of 252 (7.8%) isolates were evaluated for the presence of the mcr-1 gene by the PCR method (6), using the aforementioned strategy.

Phenotyping and genotyping of isolates carrying the mcr-1 gene.

MICs of colistin were confirmed by the broth microdilution method, and the results were interpreted according to EUCAST criteria. Species identification was performed using automated Microscan, Phoenix, and Vitek-2 systems. MICs of cephalothin (CEP), cefoxitin (FOX), cefotaxime (CTX), ceftazidime (CAZ), cefepime (FEP), aztreonam (ATM), piperacillin-tazobactam (PTZ), ampicillin-sulbactam (SAM), imipenem (IMP), meropenem (MEM), ertapenem (ERT), doripenem (DOR), ciprofloxacin (CIP), gentamicin (GEN), amikacin (AMK), trimethoprim-sulfamethoxazole (SXT), and tigecycline (TGC) were determined by the broth microdilution method, using a customized Sensititre system (Trek Diagnostic Systems, Westlake, OH), and were interpreted according to CLSI parameters (60). Multidrug-resistant isolates were defined as those exhibiting nonsusceptibility to at least one agent in three or more antimicrobial categories (61). An ESBL test was performed for isolates resistant to third-generation cephalosporins, following CLSI guidelines. For the carbapenem-resistant isolates, carbapenemase production was evaluated by the CarbaNP test (62), the modified Hodge test (following CLSI guidelines), and double-disk synergy tests using disks with the inhibitors 3-aminophenylboronic acid (APB) and ethylenediaminetetraacetic acid (EDTA) (Britania) disks. The Salmonella serotypes were determined by use of the Kaufman-White-Leminor scheme (63), and additional susceptibilities to tetracycline, chloramphenicol, nalidixic acid (NAL), and ampicillin (AMP) were determined by Kirby-Bauer and Microscan testing and interpreted according to CLSI criteria.

Molecular typing of mcr-1-positive isolates was performed by PFGE, using the PulseNet protocols described on the CDC website (https://www.cdc.gov/pulsenet/pathogens/pfge.html). The macrorestriction patterns were obtained using the enzyme XbaI (Promega, USA). Salmonella strain Braenderup H9812 was included as a molecular weight marker for the gels. The banding patterns generated were analyzed using Gel Compare 2 V.4.0 software (Applied Maths, Saint Martens-Latem, Belgium). Similarity percentages between electrophoretic patterns were calculated by use of the Dice coefficient, and the unweighted-pair group method using average linkages (UPGMA) was used to construct a dendrogram with a 1.5% tolerance and optimization settings. PFGE patterns were compared against the country's database of unique patterns (39).

Transferability of the mcr-1 gene and plasmid analysis.

Isolates carrying the mcr-1 gene were used as donors in conjugation experiments using sodium azide-resistant E. coli strain J53 as the recipient. Briefly, donor and recipient strains were grown in brain heart infusion (BHI) broth until they reached the exponential growth phase (optical density at 620 nm [OD₆₂₀] of 0.6). Conjugation experiments were performed on BHI agar (Becton Dickinson) plates, using a 1:1 donor-recipient mix, and plates were incubated at 37°C for 2 h. The conjugation mixture was suspended in 1 ml of phosphate-buffered saline, and dilutions were plated on BHI agar containing colistin (4 μg/ml) and colistin with sodium azide (100 μg/ml) for selection. Transfer frequencies were calculated as numbers of transconjugants per recipient. In E. coli-E. coli conjugations, transconjugants of E. coli J53 were confirmed by repetitive extragenic palindromic PCR (REP-PCR; using a DiversiLab system) (Fig. S3) and by phylogenetic comparisons (see the next section).

To investigate if the mcr-1 gene was located on the chromosome or on plasmids, S1/I-CeuI-PFGE was performed using previously described protocols (64, 65), followed by hybridization using a digoxigenin (DIG)-labeled mcr-1 probe (DIG High Prime DNA labeling and detection starter kit II; Roche, Germany) according to the manufacturer's instructions.

Whole-genome sequencing and bioinformatic analyses.

Genomic DNA was extracted from overnight cultures of all the clinical strains and transconjugants (Tc) by use of a DNeasy blood and tissue kit (Qiagen, Hilden, Germany). DNA libraries were prepared using a NexteraXT DNA sample preparation kit and multiplexed with a NexteraXT index primer kit (Illumina, San Diego, CA). Genomic libraries were sequenced on a MiSeq desktop sequencer (Illumina, San Diego, CA) to obtain 250-bp paired-end reads.

De novo genome assemblies were created using CLC Genomics Workbench assembler, version 8.5 (CLCbio, Aarhus, Denmark). Genome annotation was carried out using the Rapid Annotation using Subsystem Technology (RAST) server, version 2.0 (66). For molecular typing, assembled genomes were analyzed with the MLST 1.8 (67) and PlasmidFinder 1.3 (68) server tools, hosted at the Center for Genomic Epidemiology. Phylotyping of the E. coli isolates was performed by searching the assembled genomes against the arpA (accession number AFST01000005.2), yjaA (accession number 556503834), chuA (accession number U67920), and tspE4.C2 (accession number AF222188) genes by use of BLASTN (69) (identities of >95% and alignment coverage above 80%). Acquired antibiotic resistance determinants were found in the assembled genomes by use of BLASTX (69) and protein sequences from a translated ResFinder database (70) (identities of >95% and alignment coverage above 80%). Furthermore, amino acid changes associated with colistin resistance in previously reported proteins (5, 21) (Table S1) were examined in all the assembled genomes. Protein sequences were searched by use of BLASTX (identities of >95% and alignment coverage above 80%) against the reference sequences listed in Table S1, and changes were identified after multiple-sequence alignment with Muscle (71).

To obtain the sequences of the transferred plasmids, the reads from the transconjugants were mapped against the Escherichia coli J53 reference genome (accession number AICK01), and the unmapped reads were assembled de novo by use of CLC Genomics Workbench v8.5 with default parameters. Identification of candidate plasmids associated with transfer of the mcr-1 gene was performed with BLASTN, using the contigs harboring mcr-1 in the transconjugants after WGS. In cases in which the contig contained only the mcr-1 gene (average length, 2,700 bp), we used the longest contig in the transconjugant assembly instead and queried it against the Refseq database at the NCBI website (72). Subsequently, the best hits were selected and used as references for a following BLASTN search of the complete assemblies against the candidate plasmids. All hits with identity values of >90% were selected, and average identity and coverage percentages were calculated for the corresponding alignments.

Genomic sequences from the other 18 E. coli isolates harboring mcr-1 were obtained (Table S2) from the NCBI genome database (72). Single nucleotide polymorphisms of each isolate relative to the genome sequence of E. coli K-12 strain MG1655 (accession number NC_000913) were detected using mummer (73) for draft genomes and BWA (74) and SamTools (75) for available reads. A maximum likelihood phylogenetic tree was created by use of RAxML (76), with selection of the best tree out of 20 runs, using the general time-reversible model with a gamma distribution and the Lewis ascertainment bias; additionally, 100-bootstrap resampling was done. The tree was plotted using Itol (77).

Accession number(s).

Whole-genome shotgun sequences have been deposited in GenBank by use of the NCBI Prokaryotic Genome Annotation Pipeline (PGAP), under accession numbers MVPR00000000, MVPQ00000000, MVPP00000000, MVPO00000000, MVPN00000000, MVPM00000000, MVPL00000000, MVPK00000000, MVPJ00000000, MVPI00000000, MVPH00000000, MVPG00000000, MVPF00000000, MVPE00000000, MVPD00000000, MVPC00000000, MVPB00000000, MVPA00000000, MVOZ00000000, MVOY00000000, and MVOX00000000 (see Table S3 in the supplemental material for further details).