Prospective study on human fecal carriage of Enterobacteriaceae possessing mcr-1 and mcr-2 genes in a regional hospital in Hong Kong

The goal of this surveillance study was to have a gross picture on (1) fecal carriage of mcr-1- and mcr-2-harboring Enterobacteriaceae among patients/ clients using our service, (2) characteristics of the carriers and (3) phenotypic and molecular traits of the positive strains. This information is important for infection control and risk assessment. From October 31 to November 25, 2016, all fecal specimens submitted for routine analysis were included in this surveillance study. These comprised 672 consecutive routine fecal specimens collected from 616 individuals, including 79 fecal specimens from 67 outpatients, 171 fecal specimens from 144 inpatients and 422 fecal specimens from 418 asymptomatic clients coming for health assessment. Forty out of 144 inpatients had been screened for vancomycin-resistant enterococci (VRE), which is routinely performed for inpatients who have been hospitalized in/ outside Hong Kong, underwent VRE screening in other hospitals, undergone surgical operation overseas or staying in a nursing home within past 6 months. After routine testing, leftover fecal specimens were used for surveillance screening.

The concept of Blackburn and coworkers’ method for screening carbapenem-resistant bacteria was adopted [9], with some modifications. Briefly, fecal specimens with approximate sizes of 10 μL inoculation loop were homogenized in 500 μL sterile, double-deionized water and centrifuged at 1,500 rpm for 30 s. An aliquot of 10 μL supernatant from each sample was inoculated on chromID® CPS® Elite agar (bioMérieux, Marcy, I’Etoile, France) and a piece of polymyxin B disc (PB300, Oxoid, Basingstoke, Hampshire, England) was applied on the area of inoculation, followed by overnight incubation at 35 ^oC in ambient air. Bacterial growth around polymyxin B disc, namely inside the inhibition zone of approximately 11 mm in diameter (taking the Clinical and Laboratory Standards Institute (CLSI) guidelines’ interpretative criteria of polymyxin B for Pseudomonas aeruginosa as reference), was subjected to polymerase chain reaction (PCR) targeting mcr-1 and mcr-2 genes.

Polymyxin B-resistant bacterial colonies from each sample were homogenized in 20 μL sterile, double-deionized water. A 1 μL aliquot of this suspension was added to PCR cocktail comprising 20 μL Platinum® PCR Supermix (Invitrogen, Carlsbad, CA) and 1 μL of each 10 μM primer (mcr-1 forward primer, 5’-ATGATGCAGCATACTTCTGTG-3’ [10] and reverse primer, 5’-CCCAAACCAATGATACGCAT-3’ [8], corresponding to amplicon size of 398 bp; mcr-2 forward primer, 5’-TGTTGCTTGTGCCGATTGGA-3’ and reverse primer, 5’-AGATGGTATTGTTGGTTGCTG-3’ [2], corresponding to amplicon size of 567 bp). PCR was performed using GeneAmp® PCR System 9700 (Applied Biosystems, Foster City, CA, USA) with initial cell lysis and denaturation at 95 ^oC for 10 min, three-step cycling for 40 times (95 ^oC for 10 s, 55 ^oC for 30 s and 72 ^oC for 30 s) and final extension at 72 ^oC for 5 min. The amplicons were electrophoresized in 2% agarose gel.

PCR-positive colony suspensions were subcultured on CPS agar and incubated at 35 ^oC overnight in ambient air. Recovered bacterial colonies were identified by matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS, Microflex LT, Bruker Daltonik, Bremen, Germany). Briefly, minute amount of bacterial colony was smeared on target plate and covered with 1 μL of 10 mg/mL matrix solution (α-cyano-4-hydroxycinnamic acid, Bruker Daltonik, Bremen, Germany). The dried samples were subjected to MS measurement after successful calibration with Bacterial Test Standard (Bruker Daltonik, Bremen, Germany). Spectra were obtained with an accelerating voltage of 20 kV in linear mode with m/z range of 2,000 to 20,000 Da, followed by analysis with MALDI Biotyper version 3.0 and reference library version 3.1.2.0 (Bruker Daltonik, Bremen, Germany).

Antibiotic susceptibilities were determined using Kirby-Bauer disk diffusion method with reference to CLSI guidelines. Minimum inhibitory concentration (MIC) of colistin was determined using broth microdilution (BMD) (SensiTest Colistin, Liofilchem® Diagnostics, L’ Aquila, Italy). Susceptibility results of VITEK® 2 AST cards (AST-N338, bioMérieux, Marcy, I’Etoile, France) and polymyxin B Etest strips (bioMérieux, Marcy, I’Etoile, France) served as additional reference.

Total nucleic acid of PCR-positive isolates was extracted using NucliSENS® easyMAG® automated system (bioMérieux, Marcy, I’Etoile, France) and full mcr-1 genes were amplified using primers published by Ye and co-workers [10]. Amplicons were purified enzymatically with ExoSAP-IT® (Affymetrix, Santa Clara, CA), followed by cycle sequencing using BigDye® Terminator v.1.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and purification of sequencing products using BigDye® XTerminator™ Purification Kit (Applied Biosystems, Foster City, CA, USA). Purified sequencing products were analyzed using 3130xl Genetic Analyzer/ 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA).

Briefly, DNA concentrations of total nucleic acid extracts were measured using Quant-it™ PicoGreen® dsDNA Assay Kit via Qubit® 2.0 Fluorometer (Invitrogen, Carlsbad, CA). DNA concentrations were adjusted to 0.2 ng/μL, followed by tagmentation and limited-cycle amplification using Nextera XT DNA Library Prep Kit (Illumina, San Diego, CA, USA). Indexed libraries were purified using Agencourt® AMPure® XP (Beckman Coulter, Beverly, MA, USA) and analyzed by 2 × 300 bp paired-end sequencing on MiSeq platform using MiSeq v3 Reagent Kit (Illumina, San Diego, CA, USA).

WGS short reads were assembled into contigs using SPAdes version 3.10.1 [11]. E. coli sequence types (STs) were determined using MLST databases of The University of Warwick [12] and PubMLST [13]. Acquired resistance genes were identified using ResFinder 2.1 [14]. Virulence genes were detected using VirulenceFinder 1.5 [15]. Alignment of contigs to appropriate reference sequences was performed using Mauve version 2.4.0 [16] followed by post-assembly gap filling by PCR and Sanger sequencing. Prokka version 1.11 [17] was used for plasmid annotation with manual editing. Plasmid replicons and insertion sequences were identified using PlasmidFinder 1.3 [18] and IS finder [19]. Plasmid genetic maps were generated using SnapGene® Viewer 4.1. Sequence comparisons between plasmids were generated using BLAST Ring Image Generator (BRIG) [20].

Seventy nine out of 672 fecal specimens revealed bacterial growth inside the inhibition zone of polymyxin B disc and subjected to colony PCR, with 14 specimens being mcr-1-positive (2.08%) and none being mcr-2-positive. Results of MALDI-TOF MS showed that all 14 isolates were E. coli (top identification scores ranged from 2.271 to 2.526) which were consistent with the appearance of bacterial colonies on CPS agars (red-colored). These E. coli strains were isolated from 14 separate individuals including 3 inpatients, 2 outpatients and 9 asymptomatic clients coming for health assessment, with age from 1 to 58 years (Table 1). Regarding available clinical/ epidemiological information of these mcr-1 carriers, Subject 1 and 10 had received antibiotic treatment other than colistin prior to fecal specimen collection. For others, none had been prescribed with antibiotics in our hospital. We could not access any information on antibiotic usage outside our hospital. Subject 3 had been hospitalized in another hospital within 6 months prior to admission. None of these mcr-1-possessing E. coli strains caused diseases in their human hosts. In general, no apparent correlation was observed between mcr-1 fecal carriage and the available clinical/ epidemiological information.

The fecal carriage data further revealed asymptomatic carriage of mcr-1-possessing Enterobacteriaceae by both patients and healthy individuals. The mcr-1 positive rate estimated in this study (2.08%) was much higher than that reported by another local research group (0.4%) [8], which implies the prevalence was higher than expected in Hong Kong. This difference in mcr-1 positive rate might be explained by different screening targets adopted, with the study by Wong and coworkers focusing on clinical isolates whereas our study screening fecal specimens. As fecal specimens normally contain numerous strains of Enterobacteriaceae per sample, the chance of isolating mcr-1-harboring strains might be increased. Compared with human fecal carriage data of other regions, the mcr-1 positive rate estimated in this study was lower than that of mainland China (4.9 - 6.2%) [21, 22] and higher than that of several European countries (Switzerland, 0% [23]; Western France, 0% [24]; The Netherlands, 0.35% [25]). This is not surprising considering high prevalence of mcr-1-harboring Enterobacteriaceae among livestock animals in mainland China [26], which is a major supplier of food animals in Hong Kong. On the other hand, no mcr-2-carrying Enterobacteriaceae was isolated. In another large-scale study in Japan, mcr-2 gene was not detected among 9,306 E. coli strains isolated from healthy animals [27]. Considering these prevalence data together with reports of zero human fecal carriage from various European countries [23‐25], it appears that mcr-2-carrying Enterobacteriaceae has not been disseminated globally at present.

Antibiotic susceptibility results are summarized in Table 1. Colistin MICs ranged from 2 to 4 μg/mL by BMD and 2 to > 16 μg/mL by VITEK® 2 AST. The range of polymyxin B MICs was 4 to 12 μg/mL by Etest. The mcr-1-positive isolates displayed different patterns of susceptibility to a panel of 19 antibiotics for Enterobacteriaceae. With exception of Subject 8, all other isolates were intermediate or resistant to at least 1 antimicrobial agent, including penicillins (ampicillin, 71.4%), β-lactam/β-lactamase inhibitor combinations (amoxicillin-clavulanate, 50%; piperacillin/tazobactam, 14.3%; ticarcillin-clavulanate, 21.4%), cephems (cefepime, 14.3%; cefotaxime, 14.3%; ceftazidime, 14.3%; cefuroxime, 57.1%; cephalothin, 64.3%), aminoglycosides (amikacin, 28.6%; gentamicin, 21.4%; tobramycin, 28.6%), tetracyclines (minocycline, 35.7%), quinolones (ciprofloxacin, 35.7%; levofloxacin, 28.6%) and folate pathway inhibitors (trimethoprim/sulfamethoxazole, 78.6%). All these isolates were susceptible to carbapenems with 2 being extended-spectrum β-lactamase (ESBL) producers.

The range of colistin MICs in this study was 2 to 4 μg/mL by BMD, which appeared to be narrower than that of mcr-1-positive E. coli isolates causing human diseases (4 to 16 μg/mL by BMD) reported by various research groups [28‐31]. Colistin MICs determined by VITEK® 2 AST were consistent with that of BMD for 35.7% of the isolates, whereas MICs were 2- to > 4-fold higher than that of BMD for the remainder. Albeit underestimation of colistin MIC was not observed in this study, the suitability of using VITEK® 2 AST for colistin MIC determination awaits further evaluation due to its high very major error rate (36%) revealed by a recent study [32]. In addition, as 21.4% of the isolates displayed colistin MICs lower than the resistant breakpoint (> 2 μg/mL) published by The European Committee on Antimicrobial Susceptibility Testing (EUCAST), we echo the suggestion of using a lower susceptible breakpoint of < 1 μg/mL by Chew and coworkers [32] for improving detection of mcr-1-possessing Enterobacteriaceae. On the other hand, it was noticeable that more than half of the isolates (57.1%) displayed a ‘baseline’ insusceptibility to 5 or more antibiotics, and 2 of which were ESBL producers. The choice of usable antibiotics may be further limited for these isolates upon acquisition of antibiotic resistance genetic elements, for instance, carbapenemase genes.

The results are summarized in Table 2.

The mcr-1 gene sequences of 12 isolates were 100% identical to that of E. coli strain SHP45 (NG_050417), with exception of the isolate from Subject 4 harboring C>T substitution at nucleotide position 27 and the isolate from Subject 2 with mixed nucleotides (R) at position 1263, indicating possible coexistence of ‘wild type’ and 1263G>A variant in the same bacterium. Both of these nucleotide substitutions were silent mutations.

Thirteen isolates were assigned to 12 E. coli STs by Achtman or Pasteur scheme, including ST442 (Achtman), ST201 (Pasteur), ST88 (Achtman)/ ST66 (Pasteur), ST1716 (Pasteur), ST155 (Achtman)/ ST21 (Pasteur), ST10 (Achtman)/ ST2 (Pasteur), ST34 (Achtman)/ ST638 (Pasteur), ST226 (Achtman)/ ST486 (Pasteur), ST5995 (Achtman), ST95 (Achtman)/ ST1 (Pasteur), ST746 (Achtman) and ST48 (Achtman, 2 ESBL-producing strains). The strain from Subject 4 could not be assigned to currently known ST by both Achtman and Pasteur schemes. Generally, the STs of these strains were diversified. Albeit both the strains from Subject 12 and 13 were assigned to ST48, their antibiotic susceptibility patterns, genetic context and Pasteur ST profiles were different, which appeared that they were divergent over years rather than a short period of time. Among these STs, ST10, ST34 and ST48 have been reported to possess mcr-1 gene [33], with ST10 being the most common among mcr-1-possessing E. coli [7]. On the other hand, ST95, ST155, ST226, ST442 and ST746 have been associated with human infection [34‐38].

Virulence genes were detected in 8 isolates, including astA (21.4%), which encodes EAST-1 heat-stable toxin; cma (14.3%), which encodes colicin-M; iroN (21.4%), which encodes enterobactinsiderophore receptor protein; iss (35.7%), which correlates with increased serum survival of E. coli; lpfA (28.6%), which encodes long polar fimbriae; mchF (7.1%), which encodes ABC transporter protein; sfaS (7.1%), which encodes S-fimbrial adhesion protein and tsh (7.1%), which encodes temperature-sensitive hemagglutinin. These genes are associated with mild virulence of E. coli.

Eight plasmid replicon types were identified in 10 isolates, with IncI1 (21.4%), IncX4 (21.4%) and IncN (21.4%) being the predominant incompatibility groups, followed by IncQ1 (14.3%), IncP1 (7.1%), IncR (7.1%), IncI2 (7.1%) and IncHI2 (7.1%). Previous study has shown that IncX4, IncI2 and IncHI2 incompatibility groups comprised 90.4% of identified mcr-1 plasmids [7]. Significant geographical clustering was observed with regional spread of IncI2 and IncHI2 types of mcr-1 plasmids in Asia and Europe, respectively, while there was no significant difference between distribution of these major replicon types from different sources of isolation [7]. In this study, albeit further work is needed to confirm the genomic locations of mcr-1 genes, plasmid replicon types were diversified and not confined to IncI2, which was detected in a single isolate from Subject 11. Further study with larger cohort of isolates from different sources is needed to determine local characteristics of mcr-1 plasmid incompatibility groups.

With exception of mcr-1, acquired antibiotic resistance genes were identified in 13 out of 14 isolates. Aminoglycoside resistance-related genes were the most commonly observed, including aadA1 (50%), aadA2 (50%), aadA5 (14.3%), aph(3’)-Ia (35.7%), strA (28.6%) and strB (28.6%). This was followed by genes related to sulphonamide resistance (sul1, 14.3%; sul2, 21.4%; sul3, 57.1%), quinolone resistance (aac(6’)Ib-cr, 7.1%; oqxA, 28.6%; oqxB, 28.6%; QnrS1, 21.4%; QnrS2, 7.1%), β-lactam resistance (bla_OXA-1, 7.1%; bla_TEM-1B, 42.9%; bla_CTX-M-9-like, 14.3%), trimethoprim resistance (dfrA12, 50%; dfrA17, 14.3%), tetracycline resistance (tet(A), 14.3%; tet(B), 21.4%), macrolide resistance (mph(A), 21.4%), rifampicin resistance (AAR-3, 7.1%) as well as fosfomycin resistance (fosA3, 7.1%). Carbapenemase genes were not detected in these mcr-1-possessing E. coli isolates.