Optimization of the MALDIxin test for the rapid identification of colistin resistance in Klebsiella pneumoniae using MALDI-TOF MS

Abstract Background With the dissemination of carbapenemase producers, a revival of colistin was observed for the treatment of infections caused by MDR Gram-negatives. Unfortunately, the increasing usage of colistin led to the emergence of resistance. In Klebsiella pneumoniae, colistin resistance arises through addition of 4-amino-l-arabinose (l-Ara4N) or phosphoethanolamine (pEtN) to the native lipid A. The underlying mechanisms involve numerous chromosome-encoded genes or the plasmid-encoded pEtN transferase MCR. Currently, detection of colistin resistance is time-consuming since it still relies on MIC determination by broth microdilution. Recently, a rapid diagnostic test based on MALDI-TOF MS detection of modified lipid A was developed (the MALDIxin test) and tested on Escherichia coli and Acinetobacter baumannii. Objectives Optimize the MALDIxin test for the rapid detection of colistin resistance in K. pneumoniae. Methods This optimization consists of an additional mild-acid hydrolysis of 15 min in 1% acetic acid. The optimized method was tested on a collection of 81 clinical K. pneumoniae isolates, including 49 colistin-resistant isolates (45 with chromosome-encoded resistance, 3 with MCR-related resistance and 1 with both mechanisms). Results The optimized method allowed the rapid (<30 min) identification of l-Ara4N- and pEtN-modified lipid A of K. pneumoniae, which are known to be the real triggers of polymyxin resistance. At the same time, it discriminates between chromosome-encoded and MCR-related polymyxin resistance. Conclusions The MALDIxin test has the potential to become an accurate tool for the rapid determination of colistin resistance in clinically relevant Gram-negative bacteria.


Introduction
Currently, antimicrobial resistance is at the top of the agenda for scientists and governments, while XDR organisms, such as carbapenemase-producing Enterobacterales (CPE) are rapidly emerging. The pipeline of new antibiotics is very limited, and colistin is now considered as one of the last-resort therapies for the treatment of infection caused by XDR Gram-negative bacteria. 1 In countries that are considered to be endemic for CPE (e.g. Greece, Italy), colistin is often used as empirical treatment for severe infection, such as bacteraemia. Unfortunately, this increased use of colistin in the therapeutic armamentarium has led inexorably to the development of resistance. [2][3][4][5][6] In Gram-negative bacteria, acquired resistance to colistin results mostly from modifications of the drug target, i.e. LPS. These modifications correspond to addition(s) of 4-amino-L-arabinose (L-Ara4N) and/or phosphoethanolamine (pEtN) to lipid A, the anchor of LPS. Addition of such cationic components leads to the repulsion of colistin (an old class of cationic antibiotic that targets polyanionic bacterial LPS and disrupts the bacterial outer membranes), resulting in the protection against outer-membrane disruption by the antibiotic. 7,8 The mechanisms involved in such modification of lipid A might be chromosome or plasmid encoded. Plasmid-encoded resistance to colistin involves the acquisition of an mcr-like gene encoding a specific pEtN transferase. 9 MCR producers have mostly been reported among Escherichia coli and Salmonella spp. 9 In contrast, in Klebsiella spp., chromosomeencoded resistance has been reported to be far more prevalent than mcr acquisition. The most prevalent chromosome-encoded mechanisms are mutations in genes encoding the PmrA/PmrB or PhoP/PhoQ two-component systems and (even more prevalent) alterations of the master regulator MgrB. 10 Although the epidemiology of acquired colistin resistance varies depending on the bacterial species and geographical area, rapid detection of such resistance is one of the key ways of improving the treatment of patients infected with MDR bacteria for which other alternatives are not available (e.g. MBL producers). Currently, detection of colistin resistance in Enterobacterales relies on MIC determination using broth microdilution, which is the gold standard for colistin susceptibility testing. 11 Recently, we developed a novel rapid approach using MALDI-TOF MS that detects colistin resistance directly on intact bacteria in <15 min, the MALDIxin test. 12 It has been validated for E. coli and Acinetobacter baumannii, for which it can discriminate between chromosome-and/or plasmid-encoded resistance (i.e. mcr), besides detecting colistin resistance. 12,13 Here we report an optimization of the MALDIxin test for the rapid detection of colistin resistance in Klebsiella pneumoniae.

Bacterial isolates
A collection of 81 K. pneumoniae clinical isolates from Belgian and French national reference centres for antimicrobial resistance were used in this study (Table 1), including 49 colistin-resistant isolates (45 with chromosome-encoded resistance, 3 with MCR-related resistance and 1 with both mechanisms) and 32 colistin-susceptible isolates.

Optimized MALDIxin test
The MALDIxin procedure was performed as previously described, 12 with the addition of a short mild-acid hydrolysis step, which was crucial for K. pneumoniae ( Figure S1, available as Supplementary data at JAC Online). Briefly, a single colony cultured on Mueller-Hinton agar (bioMérieux, La Balme-les-Grottes, France) was resuspended in 200 lL of distilled water, washed three times with double-distilled water and resuspended in 100 lL of double-distilled water. A 50 lL aliquot was then submitted to mild-acid hydrolysis by the addition of 50 lL of 2% acetic acid in double-distilled water and heating for 15 min at 100 C. The hydrolysed cells were spun, the supernatant was discarded and the pellet was suspended in 25 lL of doubledistilled water. An aliquot of 0.4 lL of the bacterial solution was loaded onto the target and immediately overlaid with 0.8 lL of a super-2,5-dihydroxybenzoic acid matrix (Sigma-Aldrich, Gillingham, UK) used at a final concentration of 10 mg/mL in chloroform/methanol (90:10, v/v). Bacterial solution and matrix were mixed directly on the target by pipetting and the mix was dried gently under a stream of air (<1 min). MALDI-TOF MS analysis was performed on a 4800 Proteomics Analyzer (Applied Biosystems, Foster City, CA, USA) using the reflectron mode. Samples were analysed by operating at 20 kV in the negative ion mode using an extraction delay time set at 20 ns. MS data were analysed using Data Explorer version 4.9 (Applied Biosystems).

Statistical analysis
All experiments were carried out on three independent bacterial cultures. Data were compared two-by-two using the unpaired Welch's t-test. P values <0.05 were considered statistically different.

Results
Detection of colistin resistance markers in K. pneumoniae using the MALDIxin test In polymyxin-susceptible K. pneumoniae isolates, the mass spectrum is dominated by two sets of peaks centred at m/z 1840 and m/z 2078 (Figure 1a). The ions at m/z 1824 and m/z 1840 are assigned to a bis-phosphorylated, hexa-acylated lipid A molecule containing or not containing a hydroxylation on the C 0 -2 fatty acyl chain. The ions at m/z 2062 and m/z 2078 are assigned to a bisphosphorylated, hepta-acylated lipid A molecule either containing or not containing a hydroxyl group, respectively, on the C 0 -2 fatty acyl chain and resulting from a palmitoylation at the C-1 acyl-oxoacyl position of the molecule at m/z 1824 and m/z 1840, respectively.
In chromosome-encoded colistin-resistant K. pneumoniae isolates, the mass spectrum exhibits two sets of peaks centred at m/z 1971 and m/z 2209, corresponding to the previously observed m/z !131 shifts of mass unit caused by the addition of L-Ara4N to the hexa-and hepta-acylated lipid A structures at m/z 1840 and m/z 2078, respectively ( Figure 1b).
In MCR producers, the mass spectrum exhibits two sets of peaks centred at m/z 1963 and m/z 2201, corresponding to the previously observed m/z !123 shifts of mass unit caused by the addition of pEtN to the hexa-and hepta-acylated lipid A structures at m/z 1840 and m/z 2078, respectively ( Figure 1c).
In colistin-resistant isolates that exhibit both plasmid (mcr)and chromosome-encoded resistance, the mass spectrum exhibits three sets of peaks centred at m/z 1963, m/z 2201 and m/z 2209, corresponding to the previously observed m/z !123 shifts of mass unit caused by the addition of pEtN to the hexa-and heptaacylated lipid A structures at m/z 1840 and m/z 2078, respectively, and m/z !131 shifts of mass unit caused by the addition of L-Ara4N to the hepta-acylated lipid A structures at m/z 2078 ( Figure 1d).
To further support this observation, we validated the MALDIxin test on 81 K. pneumoniae clinical isolates, including 45 colistinresistant and 36 colistin-susceptible isolates. The percentage of modified lipid A corresponding to the sum of the intensities of the peaks associated with pEtN modification (m/z 1963 and m/z 2201) and L-Ara4N modification (m/z 1971 and m/z 2209) divided by the intensities of the peaks assigned to the native lipid A (m/z 1824, m/z 1840 and m/z 2062) allows accurate distinction between colistin-susceptible and colistin-resistant isolates ( Figure 1e). Of note, the peak at m/z 2078 was not taken into account in the calculation of the percentage of modified lipid A since this native peak observed in colistin-susceptible isolate spectra might potentially correspond to a peak of modified lipid A resulting from the addition of pEtN plus L-Ara4N to the native bis-phosphorylated, hexaacylated lipid A (m/z 2078=m/z 1824 ! m/z 123 ! m/z 131). The percentage of modified lipid A was found to be 0 for all colistinsusceptible K. pneumoniae isolates, while it was >5 for all colistinresistant isolates (Table 1 and Figure 1e).
Detection of colistin-resistant K. pneumoniae using MALDIxin      (Figure 1f). Accordingly, using our isolate collection, arbitrary cut-off values at 20% and 80% for % L-Ara4N and % pEtN, respectively, might be suggested to easily discriminate chromosome-encoded resistance from MCR production and coexpression of both mechanisms.

Discussion
Here we optimized the MALDIxin test for the detection of colistin resistance in K. pneumoniae. The procedure used included a preliminary short (15 min) mild-acid hydrolysis step, which allowed the rapid identification of L-Ara4N-and pEtN-modified lipid A, which are known to be the real triggers of colistin resistance. In K. pneumoniae, chromosome-encoded resistance is more frequent than MCR plasmid-encoded resistance. It mainly involves alteration of MgrB, leading to activation of the arn operon and subsequent addition of L-Ara4N to the native lipid A. 10 This modification results in an m/z !131 shift of the native lipid A-related peaks. In contrast, expression of MCR enzymes results in the addition of pEtN to the native lipid A. 14 Accordingly, a shift of m/z !123 is observed. Using the optimized MALDIxin test, we could (i) easily predict colistin resistance in K. pneumoniae by checking whether any modified (L-Ara4N or pEtN) lipid A is present in the bacterial membrane, but also (ii) discriminate between chromosome-and mcr-encoded resistance by looking at the percentage of L-Ara4N or pEtN modification in the modified lipid A. As expected, 100% L-Ara4N modification was observed in the case of chromosomeencoded resistance while close to 100% modified lipid A was related to pEtN addition in the case of MCR expression. Although only one isolate was available, detection of concomitant mechanisms (MCR production ! disruption of chromosome-encoded MgrB) repeatedly resulted in a mixture of pEtN-and L-Ara4Nmodified lipid A (about 50%/50%). In addition, despite the fact that the MALDIxin test was able to accurately detect colistinresistant isolates, there was no strong correlation between the modification level of lipid A and the resistance level of colistin in terms of MIC.
In the context of MCR-related colistin resistance, molecular biology is widely used for the detection of MCR-producing isolates. 15 However, the increasing number of mcr variants (mcr-1 to mcr-9) that do not share a strong nucleotide identity will inexorably lead to false-negative results. By targeting the pEtN modification of lipid A, which corresponds to the result of all MCR variants, the MALDIxin test might be an accurate screening test for the identification of a new MCR variant.
To the best of our knowledge, this is the first MALDI-TOF MSbased method that allows the rapid detection of colistin resistance and at the same time discrimination between chromosomeencoded and MCR-related polymyxin resistance in K. pneumoniae Dortet et al.
without necessitating any complex lipid extraction steps. Indeed, Liang et al. 16 recently described another MALDI-TOF MS-based method that has the ability to differentiate colistin-susceptible from colistin-resistant K. pneumoniae, but that requires fastidious sample preparation of membrane lipids with incubation in a special buffer, incubation in cooled ice, washes in ethanol and final extraction in chloroform/methanol/water (12:6: We should acknowledge that MALDI-TOF MS analysis was performed on a 4800 Proteomics Analyzer, which is not commonly available in clinical microbiology laboratories. In addition, samples were analysed by operating in the negative ion mode of the mass spectrometer, which is not currently and widely usable on routine mass spectrometers. Accordingly, a few optimization steps are still needed to implement this test in routine use.  Figure S1 is available as Supplementary data at JAC Online.