Abstract

The recent discovery of the plasmid-mediated mcr-1 gene conferring resistance to colistin is of clinical concern. The worldwide screening of this resistance mechanism among samples of different origins has highlighted the urgent need to improve the detection of colistin-resistant isolates in clinical microbiology laboratories. Currently, phenotypic methods used to detect colistin resistance are not necessarily suitable as the main characteristic of the mcr genes is the low level of resistance that they confer, close to the clinical breakpoint recommended jointly by the CLSI and EUCAST expert systems (S 2mg/L and R >2mg/L). In this context, susceptibility testing recommendations for polymyxins have evolved and are becoming difficult to implement in routine laboratory work. The large number of mechanisms and genes involved in colistin resistance limits the access to rapid detection by molecular biology. It is therefore necessary to implement well-defined protocols using specific tools to detect all colistin-resistant bacteria. This review aims to summarize the current clinical microbiology diagnosis techniques and their ability to detect all colistin resistance mechanisms and describe new tools specifically developed to assess plasmid-mediated colistin resistance. Phenotyping, susceptibility testing, and genotyping methods are presented, including an update on recent studies related to the development of specific techniques.

1. Introduction

Multidrug-resistant (MDR) bacteria are of a global concern, notably with the description of carbapenemase-producing Enterobacteriaceae [1]. Colistin is an old antibiotic that regained popularity as a last resort treatment to face the worldwide emergence of these pathogens [2]. Colistin is a polycationic and bactericidal drug that targets the lipid A moiety of lipopolysaccharide (LPS), moving its cationic charges, leading to cell wall lysis and bacterial death [3]. The increasing use of colistin has led to emerging resistance, a phenomenon that represents a clinical source of worry [4]. Enterobacteriaceae are Gram-negative bacteria that are often described as the pathogens responsible for human infectious diseases, particularly the Escherichia coli and Klebsiella pneumoniae species. Until recently, all mechanisms described were of chromosomal origin, mostly mediated by the two-component systems PmrAB and PhoPQ, leading to the addition of positively charged carbohydrates on the negatively charged lipid A (Figure 1), a phosphoethanolamine by a phosphoethanolamine transferase or a 4-amino-4-arabinose by surexpression of arnBCADTEF operon, leading to the loss of polymyxin affinity for the LPS [5]. In November 2015, Liu et al. reported the first plasmid-mediated gene which they named mcr-1 [6], which encodes for a phosphoethanolamine transferase, and this was followed by the description of variants (mcr-1.2, mcr-1.3,…) and the genes mcr-2, mcr-3, mcr-4, mcr-5, mcr-6, mcr-7, and mcr-8 [715]. This recent discovery raised concern about the increase and spread of resistance among the Enterobacteriaceae [16] and led to new recommendations for laboratory diagnosis and clinicians [17]. Specifically, the majority of these mcr-1 strains exhibited a low minimal inhibitory concentration (MIC) of colistin, around 4 µg/ml [6], which is close to the MIC breakpoint according to the EUCAST guidelines (susceptibility2 µg/ml and resistance>2 µg/ml) (http://www.eucast.org). Moreover, several studies have reported the detection of the mcr-1 gene in carbapenemase-producing Enterobacteriaceae strains, describing coproduction with other plasmid-mediated genes (blaNDM-1, blaNDM-5, blaNDM-9, blaKPC-2, blaKPC-3, blaOXA-48, and blaOXA-181) [1824].

The emergence of antibiotic resistance of clinical interest usually conduces to the development of new tools in clinical microbiology laboratories [25]. Currently, the detection of carbapenemase-producing bacteria is well determined, combining specific culture media, phenotyping testing, antibiotic susceptibility testing, and molecular biology [2628]. As colistin resistance is a recent global phenomenon, the implementation of rapid and reliable screening tools to detect and analyze colistin-resistant pathogens in such a way as to isolate the patient and adapt the treatment is a necessary approach [29]. Moreover, heteroresistance to colistin is a common phenomenon that is widely underestimated, requiring specific methods [3032]. Here, we propose an overview of all the screening and analysis methods developed to assess colistin resistance among bacterial pathogens causing infectious diseases in hospitalized patients. This review summarizes the current clinical microbiology diagnosis techniques and their ability to detect all colistin resistance mechanisms, and describes new tools specifically developed to assess plasmid-mediated colistin resistance [33].

Phenotyping, susceptibility testing, and genotyping methods are presented, including an update on recent studies related to the development of specific techniques.

2. Phenotypic Detection Methods

2.1. Selective Culture Media

Culture remains the benchmark method for isolating pathogens within clinical samples, and selective media are continuously developed to isolate specific bacteria [25]. Until recently, there was no specific culture medium for the detection of colistin-resistant strains, and current polymyxin-containing culture media were not able to detect low-level resistant strains because the concentrations of polymyxin in their composition are too high or because they contain other antimicrobial drugs [3461] (Table 1). Therefore, some in-house media have been developed for colistin-resistant strain screening studies, including strains carrying the mcr genes (Supplementary Table S1). These selective culture media were developed by adding low concentrations of colistin (2 or 4 mg/L) to LB nonselective agar or a MacConkey medium, which is selective of Gram-positive contaminants [62, 63]. The chromogenic and nonselective CHROMagar Orientation medium (Biomérieux, Marcy l’Étoile, France) was also used with 4 mg/ml of colistin [64]. They were used in studies to detect the growth of colistin-resistant isolates by directly culturing samples [6567] or following an enrichment step [68] which could also be selective with the addition of 2 mg/L of colistin to the broth medium [62, 64, 69]. Other anti-infective drugs could be added to avoid contaminants: vancomycin for Gram-positive contaminants [64, 66, 68] and/or amphotericin B for fungal pathogens [67, 68]. For some other studies, such media were developed to screen colistin resistance in bacterial isolates by subculturing them on agar with 2 mg/L of colistin: MH agar [9], COS medium [70], or MacConkey medium [65]. Wong et al. named their medium MHC1 for Mueller–Hinton colistin 1 [71]. Lastly, the selective CNA medium (colistin and nalidixic acid-containing agar), containing 10 mg/L of colistin, could detect mcr-1-positive isolates, one E. coli [72] and one K. pneumoniae [73], and was also used with CLED (cysteine lactose electrolyte deficient) medium (BioMérieux, Marcy l’Étoile, France) for screening samples that had or had not been precultured on Trypticase Soy Broth  ±2 mg/L of colistin [74].

More specifically, the SuperPolymyxin medium (Elitech Microbio, Signes, France) was developed and intended to specifically detect colistin-resistant strains, including those with a low MIC of colistin and harboring the mcr-1 gene [75]. The SuperPolymyxin medium has the advantage of facilitating the visualization of E. coli strains because it is composed of EMB agar, meaning that they exhibit a metallic green reflect. Its specificity is enabled by 3.5 µg/ml of colistin, 10 µg/ml of daptomycin, and 5 µg/ml of amphotericin B in its composition.

The CHROMagar COL-APSE medium was also developed to detect colistin-resistant strains and was compared to the SuperPolymyxin [76]. Its composition is not precisely described, based on commercial CHROMagar compounds containing colistin sulfate and oxazolidonone antibiotics. The CHROMagar COL-APSE medium presents the advantage to be chromogenic, with the capacity to differentiate colistin-resistant nonfermentative Gram-negative strains as well as Enterobacteriaceae.

The LBJMR medium was also developed to detect all the colistin-resistant bacteria, including those harboring mcr-1 genes [77]. The LBJMR medium presents the advantage of being versatile, combining colistin-resistant and vancomycin-resistant bacteria screening tools, conferred by 4 µg/ml of colistin sulfate and 50 µg/ml of vancomycin. In particular, the LBJMR medium can be used to detect vancomycin-resistant enterococci (VRE), which represents another emerging field of clinical concern. Both colistin-resistant Enterobacteriaceae and VRE strains are easy to detect on the LBJMR medium with the presence of bromocresol purple and glucose: fermentative strains exhibit yellow colonies on a purple agar. Lastly, it can be used to specifically detect pathogens that are often diagnosed in cystic fibrosis patient samples.

The sensitivities of these three media were excellent to detect colistin-resistant strains.

2.2. Qualitative Detection of Colistin Resistance with Phenotypic Tests
2.2.1. Rapid NP Polymyxin Test for Enterobacteriaceae

The rapid polymyxin NP test (Elitech, Signes, France) is based on a simple pH test, and detection of colistin resistance is obtained by a color change within 2 hours [78, 79]. The test was evaluated on 200 isolates of Enterobacteriaceae and can be used directly on blood samples [80]. A recent review proposed a diagnosis plan integrating this phenotypic test to confirm colistin resistance of Enterobacteriaceae strains after their growth on a selective medium [29], and its reliability is discussed in several studies [81, 82]. Compared to the broth microdilution (BMD) susceptibility testing method, agreements were excellent to detect mcr-1 and mcr-2 strains [83, 84]. The rapid polymyxin test has a good sensitivity to detect Hafnia sp. colistin-resistant isolates [79] but failed to detect Enterobacter sp. isolates, surely due to their heteroresistance to colistin [85]. This test has to be evaluated with nonfermentative colistin-resistant strains, such as Acinetobacter baumannii and Pseudomonas aeruginosa.

2.2.2. Micromax Assay for A. baumannii

The Micromax assay (Halotech DNA SL, Madrid, Spain) is based on the detection of DNA fragmentation and cell wall damage in the presence of colistin [86]. Bacteria are incubated for 60 min with 0.5 µg/ml of colistin, trapped in a microgel, and then incubated with a lysis solution to remove weakened cell walls. The presence of DNA fragments is detected after staining by SYBR Gold fluorochrome and observed by fluorescence microscopy. Resistance corresponds to ≤11% of bacteria with cell wall damage. This method is rapid (3 h 30 min) and showed an excellent sensitivity for the detection of colistin resistance on the 70 A. baumannii tested isolates (50 susceptible and 20 resistant), but it is not specific for determining the resistance type.

2.3. Specific Phenotypic Screening Methods for the Detection of MCR-1
2.3.1. Matrix-Assisted Laser Desorption-Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS)

The detection of polymyxin-resistant bacteria by MALDI-TOF is a promising and costless approach, as the majority of clinical microbiology laboratories own the required equipment to routinely identify clinical isolates [87]. Currently, the use of MALDI-TOF for detecting the carbapenemase-producing bacteria is described, with the detection of carbapenem hydrolysis [8890]. As a specific peak was described for lipid A at 1796.2 m/z [91], the MALDI-TOF could be used for the detection of lipid A modifications [92]. Very recently, the MALDIxin test was developed for E. coli strains, based on the detection of phosphoethanolamine addition on lipid A, and could specifically detect the mcr-positive isolates [93]. Indeed, an additional peak at 1919.2 m/z was observed for all polymyxin-resistant strains, and a second additional peak at 1821.2 m/z was observed for all the mcr-positives. The MALDIxin test could detect polymyxin-resistant E. coli and also differentiate the chromosome- and plasmid-encoded resistance in 15 minutes, and should be evaluated on other species for which phosphoethanolamine addition is involved in polymyxin resistance.

2.3.2. Inhibition of MCR-1 Activity

Several studies on the structure of the catalytic domain of the MCR-1 protein have demonstrated that the phosphoethanolamine transferase is a zinc metalloprotein [9496], and that zinc deprivation could reduce the colistin MIC in E. coli isolates [97]. Screening tests were developed to specifically detect MCR-1, based on the difference of colistin susceptibility obtained in the presence or absence of chelators of zinc ion.

The colistin-MAC test consists of the addition of dipicolinic acid (DPA) in the BMD method, leading to a colistin MIC reduction of ≥ 8-fold in case of MCR-1-positive strain [98]. 74 colistin-resistant Enterobacteriaceae strains were tested, and 59 of the 61 strains carrying mcr-1-like genes were detected by the colistin-MAC test, while the 13 mcr-negative strains exhibited discrepancy in results (increase, maintain, or slow decrease) giving a sensitivity of 96.7% and specificity of 100%. Interestingly, the two mcr-1 strains that were negative with the colistin-MAC test were K. pneumoniae strains.

More recently, four assays were tested, based on inhibition by EDTA [99]. The specific detection of MCR-1 was assessed with the following tests: combined-disk method with diameter differences ≥3 mm, BMD with a reduction of colistin MIC of  ≥4-fold, modified rapid polymyxin NP test with the absence of color change, and the alteration of zeta potential RZP ≥ 2.5. These assays were performed on 109 Enterobacteriaceae including 59 mcr-1-positive E. coli and one mcr-1-positive K. pneumoniae. The modified rapid NP test and zeta potential methods showed excellent sensitivity and specificity and could be inexpensive and simple methods to detect the presence of the mcr-1 gene.

These tests should be performed on other species harboring the mcr-1 gene, in particular K. pneumoniae, and also on strains harboring other mcr genes, to validate their ability.

3. New Recommendations on Polymyxins Susceptibility Testing

Polymyxin susceptibility testing is challenging, as these large and cationic molecules poorly diffuse into the reference cation-enriched Mueller-Hinton (MH2) agar, giving discrepant results, and much more since the description of the mcr genes that confer low MICs. Moreover, even in MH2 broth medium, the concentration of cation could largely influence the polymyxin MIC results [64], notably by interacting with the acquired resistance mechanisms of the tested isolates. Defining a reference method for colistin susceptibility testing is a priority, along with the increasing use of polymyxin as last-line therapies.

3.1. Reference Method

Broth microdilution (BMD) is the only approved method for colistin MIC determination by both the European Committee on Antibiotic Susceptibility Testing (EUCAST) and the Clinical and Laboratory Standards Institute (CLSI) [100, 101]. BMD has to be performed with colistin sulfate in untreated polystyrene plates without addition of any surfactant (polysorbate 80) (Table 2). The Mueller–Hinton broth medium has to be cation-adjusted, with a final composition of 20–25 mg/L of calcium and 10–12.5 mg/L of magnesium [102]. EUCAST and CLSI joined their recommendations on the polymyxin breakpoint for MIC of Enterobacteriaceae, P. aeruginosa and Acinetobacter spp. isolates: susceptible (S) if≤2 µg/ml and resistant (R) if >2 µg/ml [100, 103]. In 2017, EUCAST added a new quality control (QC) strain that has to be used to control the performances of a colistin susceptibility method: E. coli NCTC 18853 that harbors the mcr-1 gene, in addition to E. coli ATCC 25922 and P. aeruginosa ATCC 27853 [104] (Table 2).

Dilution methods consist of adding colistin to the culture medium in such a way as to obtain twofold dilutions and are prepared according to the CLSI guide M07-A10 [101] and ISO 20776-1 standard (International Standard Organization). Broth macrodilution is performed in tubes when reference broth microdilution (BMD) is performed in 96-well trays. Only colistin sulfate can be used and particular care is required, as the powder is expressed in IU/mg, meaning that the concentrations need to be adjusted according to the CLSI M100 and the manufacturer’s instructions [103]. The antibiotic is suspended in sterile water and then diluted in MH2 broth medium before its distribution into 96-well trays. The final bacterial inoculum is 5 × 105 CFU/ml (colony-forming unit) or 5 × 104 CFU/well for the BMD method, prepared using the 0.5 McFarland standard (corresponding to approximately 1 to 2 × 108 CFU/ml) [101]. Trays are then incubated at 35 ± 1°C for 18 ± 2 hours [100, 102]. Results are read visually or with a spectrophotometer.

Broth microdilution is a time-consuming and fastidious way to assess MIC in clinical routines [105, 106]. Many errors can occur, such as an incorrect colistin concentration or dilution. This technique is not well suited to clinical microbiology routines and needs to be automated. Moreover, this method exhibits limitations for assessing heteroresistance. Indeed, the presence of resistant subpopulations can give uninterpretable results due to the presence of skipped wells and has been described for the Enterobacter species, as presented in a study of 114 Enterobacter cloacae [107]. Population analysis profiling is recommended to confirm heteroresistance [108]. For now, heteroresistance to polymyxins is not correlated with the presence of mcr genes.

3.2. Comparative Evaluations of Polymyxin Susceptibility Testing Methods

Evaluating antimicrobial susceptibility testing (AST) methods is performed using a comparison with the reference method, as per the ISO 20776-2 standard [109]: a categorical agreement (CA) is obtained when the strain is in the same clinical category (R, I, S), and an essential agreement (EA) is obtained when the MIC is within plus or minus one doubling dilution from the reference MIC. A very major error (VME) corresponds to a false susceptibility result and is calculated using the resistant strains tested, and a major error (ME), in the case of false resistance, is calculated on the number of susceptible strains. Finally, a minor error (MiE) occurs when a strain is classified as Intermediate (I) instead of S or R, or S or R instead of I. A reliable method will obtain the following scores: CA ≥ 90%, EA ≥ 90%, VME ≤ 3%, and ME ≤ 3% [109]. The results of all the comparative studies performed on colistin susceptibility testing are summarized in Table 3 (in Table S2 for polymyxin B). MIC50 and MIC90 correspond to the MIC that inhibits 50 or 90%, respectively, of the tested strains of the same species.

The surfactant polysorbate 80 was previously added to trays to limit polymyxin adherence to polystyrene and is not yet recommended; however, it could induce VME and mcr strains might not be detected [31, 110112]. Albur et al. demonstrated that the polystyrene trays used also have an influence: using tissue-culture-coated round-bottom trays gave a 5.3-fold increase in MIC values compared to noncoated V-bottom trays [113], for the material used [106] (Table S3). A very recent study compared polystyrene coated trays to glass coated trays and also showed very few differences (Table 3) [114].

Until 2013, many comparative studies used agar dilution (AD) as the reference method for polymyxins susceptibility testing, a method that differs from the BMD only because the polymyxins are added to a solid MH2 medium [31, 32, 115126]. Compared to BMD, agar dilution generally gave concordant results for colistin and polymyxin B [31, 110, 127, 128]. VMEs were very uncommon with AD, and this pointed to the AD’s potential role in screening, as it presents the advantage to test several strains on the same plates [117, 129]. A recent study compared agar dilution to broth macro- and microdilution on 8 strains and concluded that agar dilution was the most reproducible method, with an excellent distribution of colistin in agar, but that colistin-containing agar plates could be only stored for 7 days [130].

Diffusion methods based on the antibiotic diffusion in agar, whether with the Kirby–Bauer disk diffusion [131] or with the gradient strips, are not reliable for polymyxin testing and should not be used as a large number of studies have obtained high rates of VMEs or poor EA [32, 120, 125, 128, 132135]. Some studies showed good results but contained only susceptible strains [136138]. The influence of MH2 agar composition was assessed: agreement was not affected with agar dilution, but important differences were highlighted with Etest (BioMérieux, Marcy l’Étoile, France) [31, 139]. The advantage of the agar diffusion method is the detection of heteroresistance: colonies present within the inhibition zone correspond to resistant subpopulations [140]. One study compared disc diffusion to Etest method, and a large rate of minor errors occurred [141]. The ColiSpot test consists of replacing the disk of colistin by a drop of a calibrated colistin solution (8 µg/ml). Colistin resistance is revealed in the absence of the inhibition zone. This technique was evaluated with 89 colistin-resistant and 52 colistin-susceptible strains and was developed for veterinarian laboratories [142].

3.3. Commercial Devices Based on Broth Microdilution Reference Method

Several commercial devices based on BMD reference methods were developed to easily assess the reference method by offering ready-to-use systems. Their advantage is the elimination of critical preparation steps of MH2 medium and antibiotic dilutions. These systems were used to detect mcr-1 strains and were evaluated by EUCAST, giving correct results, with essential agreement ranging from 82% to 96%, and few MEs or VMEs (http://www.eucast.org/ast_of_bacteria/warnings/) [143].

3.3.1. UMIC Colistine (Biocentric, Bandol, France)

UMIC colistine consists of unitary tests composed of 12-well polystyrene strips, one for growth control and 11 containing dehydrated colistin, with concentrations ranging from 0.06 to 64 µg/ml, provided with unitary MH2 tubes. Inoculation is performed simply, after diluting the 0.5 McFarland suspension by 200-fold into the MH2 tubes, by distributing 100 µL of this diluted suspension into the 12 wells of the strip, leading to the rehydration of the antibiotic. The strips are then incubated at 35 ± 1°C using the UMIC box to avoid any desiccation. One comparative evaluation on 71 A. baumannii isolates and one on 92 Gram-negative isolates including 76 Enterobacteriaceae highlighted the reliability of the UMIC colistine kit [144, 145].

3.3.2. MIC Strip Colistin (MERLIN Diagnostika Bornheim-Hersel, Germany)

MIC Strip Colistin also consists in unitary 12-well strips with concentrations of dehydrated colistin ranging from 0.06 to 64 µg/mL, and Micronaut-S is a panel composed of different antibiotics on standard 96-well trays. Those systems can be automated with Micronaut ASTroID that concomitantly performs dilution for antimicrobial susceptibility testing (AST) and deposits on the MALDI-TOF target, simultaneously identifying the same colony being tested.

3.3.3. Sensitest Colistin (Liofilchem, Roseto degli Abruzzi, Italy)

It consists of a compact panel of 4 tests containing 7 twofold dilutions of dehydrated colistin (0.25–16 µg/ml). It showed excellent correlation with BMD when tested on 353 isolates, including 259 Enterobacteriaceae, 83 harboring the mcr-1 gene [146]. Recently, a combined Sentitest colistin/piperacillin-tazobactam was developed, with the same design, providing a unitary test for testing both antibiotics, with colistin concentrations ranging from 0.008 to 128 µg/ml.

3.3.4. The Sensititre System (Thermo Fisher Scientific, Waltham, MA, USA)

It presents different antibiotics on 96-well trays with a customizable plate layout. Inoculation, incubation, and reading (based on fluorescence) steps can be automated. Chew et al. [147] recently evaluated a Sensititre GNX3F panel containing both polymyxin B and colistin (0.25–4 mg/L) and presented a sensitivity of 95.2% and 100% in detecting the 21 mcr-1 isolates tested, respectively.

3.4. Automated Systems

Automated systems were developed to shorten result timeframes by increasing sensitivity, and also to avoid manipulation bias [148], with incubation and real-time reading. However, by combining several antibiotics, the number of concentrations tested is limited, and they cannot give a real MIC (Table 3).

3.4.1. MicroScan WalkAway (Beckman Coulter, San Diego, CA, USA)

It uses standard trays that are manually inoculated, and reading is based on fluorometry, with results obtained in 3.5–7H. It is not available on polymyxin B, and essential agreement cannot be evaluated between techniques as the NM44 panel proposes only ≤4 and ≥4 µg/ml for colistin. In the recent study of Chew et al., this panel was able to detect all mcr-1 tested isolates and presented only one VME on 76 Enterobacteriaceae isolates tested. It also evaluated 213 Acinetobacter species and presented 99.1 % categorical agreement against the agar dilution method [115].

3.4.2. Vitek 2 (BioMérieux, Marcy l’Étoile, France)

It is a semiautomated system that uses reagent cards containing dehydrated antibiotics and other reagents in a 64-well format. It combines rapid identification and AST using an extrapolated growth algorithm. Various comparative studies performed to evaluate Vitek 2 have returned discrepant results with high rates of VME. In their recent evaluation, Chew et al. [147] demonstrated the efficacy of Vitek 2 in assessing both polymyxin B and colistin MIC with only one VME for each and 96.1% and 93.9% EA, respectively, but it was only able to detect mcr-1 strains with polymyxin B.

3.4.3. BD Phoenix™ (Becton Dickinson, Le Pont de Claix, France)

It performs identification and AST in parallel in 84-well specific plates. Reading is based on an oxidation-reduction indicator in 6–16 hours. One study showed excellent agreement on 109 P. aeruginosa strains, but only colistin-susceptible strains were tested [149]. Vourli et al. [129] have shown concerning results testing Acinetobacter baumannii strains with very high VME rates (41.4%) despite the study including 24.8% of colistin-resistant strains. This was explained by the majority of errors occurring near the breakpoint (2 instead of 4 µg/ml). Lastly, in the study by Jayol et al. [83], the Phoenix system was able to detect all mcr-carrying bacteria, even those with a colistin MIC of 4 µg/ml, but the high rates of VMEs obtained prove its inability to assess heteroresistance.

4. Genotyping and Molecular Screening

4.1. PCR Amplification and Sequencing to Detect Gene Mutations

Molecular biology methods are the most sensitive for determining antibiotic resistance by assessing the presence of resistance genes or mutations conferring resistance. These methods are complementary to the phenotypic techniques and confirm the resistant status of bacterial isolates. The main mutations for Enterobacteriaceae species are located on genes coding the two-component systems PmrA/PmrB and PhoP/PhoQ (Figure 1). Specifically, mutations in the mgrB gene—the negative feedback regulator of PhoPQ—notably with the presence of insertional sequences, appeared to be the main resistance mechanism observed in K. pneumoniae strains. These colistin resistances are not based on drug-modifying enzymes or the acquisition of a resistance gene which could be easily detected. Screening of potential mutations on these chromosomal genes is done by amplification and sequencing, takes 3 days, and requires that all genes are tested. Sequenced amplicons are then compared by the BLAST tool against the NCBI database to screen possible mutations compared to wild-type genes.

4.2. Real-Time qPCR to Detect the Presence of mcr Genes

The discovery of the acquired gene mcr-1 justifies the use of molecular detection with RT-PCR, a rapid quantitative technique to detect the presence of the gene. A systematic screening of the gene in colistin-resistant strains was performed [150] (Figure 2). For such purposes, scientists have used the primers of the original study [6], or have designed their own primers for standard PCR [132, 151160] or RT-PCR, based on SYBR Green assays [64, 151, 161], TaqMan probe [66, 72, 152, 162], or other FAM-labelled probe [9, 71, 163, 164] or HEX-labelled probe [165] (Table 4).

Xavier et al. designed primers to screen mcr-2 [7], giving a 567 bp product [166]. Some designed their own primers for standard PCR [167169], and one study developed a TaqMan assay for qPCR [170]. Interestingly, three studies went further by designing a universal primer to detect both mcr-1 and mcr-2 genes by standard PCR [166, 171] and a generic primer and a probe to detect them by qPCR [74], but these have not yet been tested on other mcr genes. Lastly, primers were designed for detecting mcr-3 [10], mcr-4 [11], and mcr-5 genes [12] by standard PCR (Table 4). A recent study described a multiplex SYBR Green real-time PCR assay for the simultaneous detection of mcr-1, mcr-2, and mcr-3 genes [172]. Finally, a multiplex PCR assay for detection of the five mcr genes: mcr-1, mcr-2, mcr-3, mcr-4, and mcr-5, was developed in order to obtain sequential amplicons with a size difference of 200 bp, allowing their fast and simultaneous detection on agarose gels [173].

4.3. PCR to Detect Plasmid Carrying mcr Genes

mcr-1 is a 1626-base pair-long gene located on a 2607 bp common region flanked on both ends by an ISApl1 insertion sequence (IS) in some plasmids [174]. This sequence may form a composite transposon that can potentially move as one complete unit [155, 175, 176]. This insertion sequence appears to be a key component of the mobilome, and its presence is not systematic. Furthermore, only the upstream region can contain ISApl1 [165]. Li et al. identified the ability of the Tn6330 transposon (ISApl1-mcr-1-orf-ISApl1) to generate circular ISApl1-mcr-1-orf [177]. Specific primers were developed to screen the upstream presence of this IS transposon by PCR and Sanger sequencing [178180]. Others have also designed their own system to directly screen on plasmid carrying mcr-gene type IncX4 [9, 181], but these methods also exhibit limitations, as a wide distribution has been observed for mcr-1 among different plasmids (IncI2, IncX4, IncHI2, IncY, IncF, IncP, IncH1, and IncX3), demonstrating the great ability of mcr genes to spread.

4.4. Microarray

Microarray technology allows scientists to analyze dozens of genes at the same time. The Check-MDR CT103 microarray system (Check-Points, Wageningen, the Netherlands) was developed to screen the presence of extended-spectrum beta-lactamase (ESBL) genes (TEM and SHV) and carbapenemase genes (OXA-48, KPC, NDM…) in the same assay and can assay 24 samples at the same time, with an effective detection in 6.5 hours. Recently, a study evaluated this assay for detecting mcr genes: sensitivity and specificity were excellent for mcr-1 and its variants (from mcr-1.2 to mcr-1.7 and mcr-2 genes), but it was not able to detect the new gene mcr-3 [182]. mcr-4 has not been assayed yet.

4.5. Loop-Mediated Isothermal Amplification (LAMP)

The Eazyplex SuperBug mcr-1 kit (Amplex Biosystems Gmbh, Giessen, Germany) was developed to assess the presence of the mcr-1 gene within 20 minutes [183]. It was effective on 104 microbial strains but needs to be assayed directly on clinical samples. As it was developed before the description of the other mcr genes, it can only detect the mcr-1 gene and the mcr-1.2 variant.

More recently, another LAMP-based assay was developed to detect mcr-1 gene and evaluated as a screening tool on 556 multidrug-resistant Enterobacteriaceae [184]. Seven isolates were positive by both standard PCR and LAMP-based assay (6 E. coli and 1 K. pneumoniae). The results can be assessed by chromogenic visualization. This test constitutes a rapid, specific, and cost-effective tool that exhibits a higher sensitivity than PCR (10-fold). It has to be assayed on clinical samples; as for now, only spiked tools were used.

4.6. Novel Approach with Direct Resistome Analysis

Genomic screening is an alternative approach for studying resistance and providing a better understanding of the behavior of bacterial isolates [185]. The development of next-generation sequencing has led to lower costs, reduced screening delay, and increased sequencing speeds combined with updated databases providing access to a large amount of information. The mcr-1 gene was initially discovered by whole genome sequencing during an active livestock monitoring program in China [6]. A considerable number of retrospective studies analyzing previously recorded genomic sequences have since been carried out, showing the global dispersion of the gene [9, 10, 20, 23, 24, 72, 74, 157, 158, 174, 177, 180, 186204] (Figure 2).

The technologies used to completely sequence the bacterial genome are Illumina (Illumina Inc., San Diego, CA, USA), which produces short sequences (300 bp) and requires several days, and PacBio RS II (Pacific Biosciences, Menlo Park, CA, USA), which produces a single real-time molecule producing long sequences (60kb) in a few hours [205]. The use of Illumina sequencers is not suitable for covering bacterial genomes with multiple repetitive elements because too many sequence pieces are obtained after assembly, whereas PacBio RS II delivers a single sequence without missing regions [174]. Sekizuka et al. performed a hybrid analysis using the two technologies to analyze three Inc2 plasmids and found that they were highly conserved with the exception of the shufflon region, meaning that the combination of the two methods enables to analyze rearragements in highly recombinant regions [174].

The sequences obtained are assembled, the genome is annotated, and then a mapping is carried out against a reference plasmid, in general pHNSHP45 for mcr-1 [6]. The aligned sequences are then compared to one of the resistance gene databases: Antibiotic Resistance Gene-Annotation [206], ResFinder [207], Comprehensive Antibiotic Resistance Database [208], and Antibiotic Resistance Genes Database [209]. They could also be compared to the plasmid genome, with GenEpid-J [210] or PlasmidFinder, which enabled the discovery of the mcr-4 gene [11] and presents the advantage of screening multiple genes and detecting the coexistence of several genes including carbapenemases. Lindsey et al. proposed a whole protocol for plasmid sequencing [211]. More specifically, PointFinder was developed to detect chromosomal point mutations associated with antimicrobial resistance [212].

Whole genome sequencing combined with new bioinformatic tools improves our ability to detect several resistance genes at the same time [186, 205] but presents the same limitations than PCR: new genes are not recognized by these techniques, which require the continuous updating of databases [175, 213] that should be merged into a single reference database [213].

5. Conclusion

The recent description of plasmid-mediated colistin-resistant genes has generated concern among the global scientific community about the lack of new antibiotics to treat infections caused by multidrug-resistant pathogens. This concern was raised by the worldwide screening that demonstrated the global spread of bacterial strains harboring the mcr-1 gene from diverse human and animal origins. Thus, it is necessary to implement an adapted protocol to effectively detect colistin-resistant strains in clinical microbiology laboratories.

Phenotypic methods indicate to the microbiologist the presence of polymyxin-resistant strains but do not define the mechanism involved and the risk of transmission. Molecular methods are rapid and more sensitive but are specific to the resistance genes examined and faced with the large number of molecular mechanisms conferring resistance to polymyxins, should only be used to screen mcr genes in clinical microbiology laboratories. Genomic analysis enables the complete screening of resistance genes in genetically identified bacteria from clinical samples but remains an in silico study which enables predictions but not resistance observation, as the presence of a resistance gene in a genome does not mean that the corresponding isolate is resistant, supported by studies that identified polymyxin-susceptible bacteria carrying the mcr-1 gene [92, 165, 213]. Thus, phenotypic and molecular methods are complementary in detecting colistin-resistant pathogens in order to analyze the behavior of the clinical isolate, and it is important to carry them out in parallel [148] (Figure 3). All these techniques and their detection characteristics are summarized in Table 5.

In conclusion, these new techniques need to be combined for a complete understanding of colistin resistance, in particular for strains carrying mcr genes, so clinicians can rapidly adapt treatments or isolate the carrier patient in the hospital.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

This work was supported by the French Government under the “Investissements d’avenir” (Investments for the Future) program managed by the Agence Nationale de la Recherche (ANR, fr: National Agency for Research) (reference: Méditerranée Infection 10-IAHU-03).

Supplementary Materials

Table S1: home-made polymyxin-containing agar. Table S2: comparison of different polymyxin B susceptibility testing methods to detect polymyxin B resistance in Gram-negative clinical isolates. Table S3: other comparative AST methods. (Supplementary Materials)