Study on carbapenemase-producing bacteria by matrix-assisted laser desorption/ionization approach

The development of new techniques for the detection of carbapenemase activity is of great importance since the increased incident of resistance against carbapenems represents a serious threat to global public health. In this context, the matrix-assisted laser desorption/ionization approach already demonstrated to be a reliable tool for rapid carbapenemase detection. As a newly developed test, there is still a lack of in-depth analysis of its robustness and possible wider application. The main goal of this study was to evaluate the potential for using the design MBT STAR-Carba assay as the pre-characterization method for Enterobacterales and P. aeruginosa strains in terms of the produced classes of carbapenemases using modified procedure parameters—various suspension densities and incubation times. Moreover, its usefulness for the in-depth analysis and characterization of metallo-β-lactamases (MBL) was tested by applying inhibition assays. In this study, the designed assay proved to be a sensitive tool for the detection of carbapenemase hydrolytic activity, which can be successfully used to partially classify the class of carbapenemase present. Additionally, the use of defined high concentration suspensions would allow to shorten the incubation time to 1 minute for certain strains. Considering that the assay was also suitable to investigate the effect of different inhibitors on the MBL activity, it demonstrates far higher discriminatory potential than only a rapid routine carbapenemase detection tool and could be used as a susceptibility assay.


Introduction
The increasing prevalence of drug-resistant bacteria is a serious threat to human health and life around the world. Carbapenems are amongst the latest developed β-lactam antibiotics still possessing a broad spectrum of bactericidal activity, therefore often used as last resort medication for the treatment of severe bacterial infections [1][2][3]. Therefore, the global rise of carbapenemase-producing bacteria (CPB) incidence represents an increasing threat to the healthcare system and patient safety and is of great concern worldwide [4,5]. Reported increase in resistance to carbapenems among Enterobacterales as well as Pseudomonas aeruginosa isolates is Diagnostics, Bruker Daltonik GmbH in Bremen (Germany), were used. All strains were previously characterized as β-lactamase-producing bacteria (BLPB) using validated routine PCR methods (TaqMan-probe realtime-PCR using the KAPA™ PROBE FAST qPCR MasterMix Universal (peqlab), Primer and TaqMan-Probes (both metabion) using an ABI 7500 Fast System) and showed the presence of different classes and types of β-lactamases (BLs) coding genes-including Ambler class A & D serine carbapenemases as well as class B β-metallo-carbapenemases ( Table 1).

Culture conditions
For each analysis, bacterial isolates grown overnight on BD TM Columbia Agar on 5% Sheep Blood plates (Becton Dickinson, Germany) incubated at 35 (±2˚C) under aerobic conditions were used. To obtain single bacterial colonies from primary cultures, the quadrant streaking technique was applied.

MBT STAR-Carba standard assay
For testing BLs activity in standard mode, the MBT STAR-Carba assay was performed according to the manufacturer's instructions provided with MBT STAR-Carba IVD Kit (Fig 1A).
For each sample and controls, one inoculation loop (1 μL) filled with sufficient material of bacterial colonies was suspended in the antibiotic solution (MBT STAR-Carba Antibiotic Reagent dissolved in 50 μL MBT STAR Buffer) and then incubated for 30 min. at 35 ± 2˚C with shaking (900 rpm) using a Thermomixer (Eppendorf AG, Germany). Subsequently, 1 μL of each sample was deposited on the MALDI Biotarget 96 (Bruker Daltonics) in duplicate, airdried, and overlaid with 1 μL of MBT STAR Matrix. For each analyzed MALDI Biotarget 96 prepared, one spot with 1 μL MBT STAR-ACS (antibiotic calibration mass standard containing dedicated masses below m/z = 1000) as well as negative (dedicated β-lactam non-hydrolyzing quality control strain-E. coli ATCC 25922) and positive controls (dedicated β-lactam hydrolyzing quality control strain-K. pneumoniae ATCC BAA-1705) were used. Table 1. List of investigated carbapenemase-producing bacteria with marked type of β-lactamases and hydrolytic activity value obtained using routine protocol of MBT STAR-Carba assay.

Strain
Type of β-lactamases Ambler class � Activity value (According to MBT STAR-Carba Kit protocol) with shaking (900 rpm) using Thermomixer (Eppendorf AG, Germany) (Fig 1B). Afterwards, MALDI Biotarget 96 was prepared according to the procedure described in the previous paragraph.

β-metallo-carbapenemases inhibition assay
For investigation of β-metallo-carbapenemases inhibition: (a) 3 mg/mL dipicolinic acid (DPA, Sigma Aldrich, Germany) and (b) 10 mM ethylenediaminetetraacetic acid (EDTA, Sigma Aldrich, Germany) stock solutions in MBT STAR Buffer adjusted to 8 pH were used. From overnight cultures of 9 investigated strains characterized by the presence of MBL, bacterial suspensions were prepared as described in the previous paragraph-the McF range used for each strain was selected individually based on previous suspension test results. 200 μL of each suspension were centrifuged (900 rpm, 3 min.), supernatants were discarded and remaining cell pellets were resuspended in (a) 200 μL of MBT STAR Buffer-control variant-and (b) 100 μL of MBT STAR Buffer + 100 μL of respective inhibitor stock solutions: DPA (final conc. 1.5 mg/mL) or EDTA (final conc. 5 mM)-inhibited variants (Fig 1B). Subsequently, 50 μL of prepared suspensions were added to the tubes containing dissolved MBT STAR-Carba Antibiotic Reagent, vortexed, incubated (30 min., 35 ± 2˚C, 900 rpm) and transferred onto MALDI Biotarget 96 according to procedure described in the paragraph 2.3. Additionally, all tested MBL strains were subjected to control test using non-MBL inhibitor addition-phenyl boronic acid (PBA) at final concentration 2 mg/mL.

MALDI-TOF MS analysis
Prepared MALDI Biotarget 96 plates were analyzed using a Microflex LT/SH instrument (Bruker Daltonics) and the MBT Compass platform with MBT STAR-BL Module (Bruker Daltonics). Each spot was analyzed in duplicate, finally obtaining 4 spectra for each sample. All the processes, spectra collection, data acquisition, and processing, were performed in a fully automated manner by the software using predefined methods provided by the manufacturer.
Obtained results are finally expressed as normalized logRQ values, meaning the calculated logarithm of the ratio of the summed signal intensities of hydrolyzed forms to the summed signal intensities of non-hydrolyzed forms of the antibiotic, normalized using the respective positive (hydrolyzing) and negative (non-hydrolyzing) control strains. The results are understood as a measure of hydrolysis efficiency (higher normalized logRQ = higher degree of antibiotic hydrolysis), with normalized logRQ values � 0.2 threshold-negative result; � 0.4 thresholdpositive results; values between these thresholds-unclear results. In this study, all result values described or depicted as "logRQ" refer to normalized values.

PLOS ONE
Study on carbapenemase-producing bacteria by matrix-assisted laser desorption/ionization approach . Before PCA, all investigated data have undergone standardization using Z normalization method in order to avoid outlier issues. Assigning ranks corresponding to the level of carbapenemase activity was performed based on the grouping of the strains on the first two PCs-plane.

Results of MBT STAR-Carba standard protocol
The principle of designed approach relies on the mass spectrometric evaluation of the carbapenem molecule after a short co-incubation with the bacterial strains, monitoring the distinct mass peaks of the hydrolyzed and non-hydrolyzed forms of the antibiotic resulting from carbapenemase-dependent β-lactam ring degradation [24,25]. Due to the low signal intensities of the typically observed hydrolyzed molecular forms (254 m/z), the internal standard provided with the MBT STAR Matrix is used as hydrolysis mass reference to calculate the hydrolyzation ratio. On presented spectra (Fig 2) it can be observed signals (green lines) representing native imipenem (300.1 m/z) and its adducts with matrix (489 m/z) which intensity vary depend on carbapenemase activity-the lower signal the higher hydrolytic activity. The design MALDI TOF assay gave correct results for all investigated strains, however, the analysis of results did not reveal a grouping of the investigated strains because of the class of carbapenemases. Mostly, Enterobacterales strains representing the same genus had similar activity values-two Enterobacter strains, E. kobei and E. cloacae (0.96, each), two C. freundii strains (0.98 and 1.00), and two E. coli strains (1.03, each). More variance was observed for the Klebsiella genus (0.94 to 1.18) ( Table 1). The use of standard protocol allowed division of hydrolyzing strains into two main groups in view of their carbapenemases activity: (1) strains with considerably lower logRQ values-3 Pseudomonas aeruginosa and (2) those demonstrated higher logRQ values-Enterobacterales members. Within the first group logRQ values ranged from 0.43-0.87 obtaining the lowest values for GES carbapenemase-producing P. aeruginosa 1 strain. LogRQ values in the second group ranged from 0.94 (multi-carbapenemase-producers K. pneumoniae 12) to 1.18 (GIM carbapenemase-producing K. pneumoniae 6).

Impact of suspension density on the MBT STAR-Carba results
The tested McF density series of bacterial suspensions ranging from McF 1 to 7 enabled the indication of a present carbapenemase activity in all respective strains except for GES producing P. aeruginosa 1, for which an increase in logRQ values with increasing suspension density has been observed; however, none exceeded the 0.2 threshold (Table 2; S1 Data).
Examples of graphs obtained using MBT Biotyper Prototype software during results generating are presented in Fig 3A and 3B. The analysis of the results revealed a different effect of bacterial density on the logRQ values depending on the bacterial order and class of carbapenemase.
Regarding Enterobacterales, division of the strains into two groups could be observed (Fig 4A).
Group I was mostly comprised of β-metallo-carbapenemase-producing strains (6 out of 7) demonstrating hydrolytic activity in the whole McF range and recognized as the most active group among investigated bacteria. In 4 cases-E. coli 7, K. pneumoniae 6/12 as well as E. kobei

PLOS ONE
Study on carbapenemase-producing bacteria by matrix-assisted laser desorption/ionization approach 11-positive detection of carbapenemases activity was possible already with the lowest suspension density (1 McF) and differences between 1 and 7 McF suspensions did not exceed 0.25 logRQ. Group II was mostly represented by class A and D carbapenemase producers (4 out of 5 strains) characterized by high hydrolytic activity only at a higher McF range-positive results of the test started from 3-5 McF depending on the strain ( Table 2).
For all investigated Enterobacterales strains, elevated logRQ values were observed as suspension density increased. In turn, for two β-metallo-carbapenemase-producing P. aeruginosa strains (4 and 5), a considerably different behavior was observed. For both strains, the increase of hydrolytic activity in the lower suspension density range was followed by a gradual activity Regardless of the family investigated, as the bacterial suspension density increased, the logRQ values of the tested strains approached the levels obtained when using the standard protocol, except for the strain P. aeruginosa 4 in which standard method normalized logRQ values placed between results noted for 4 and 5 McF suspension ( Table 2).

Effect of incubation time on the MBT STAR-Carba results
Based on the suspension experiments, bacterial suspensions with McF 4 and 7 were chosen for investigation of the incubation time. Strain P. aeruginosa 1 was excluded from investigation due to lack of positive results in investigated suspensions range.

PLOS ONE
Study on carbapenemase-producing bacteria by matrix-assisted laser desorption/ionization approach Regarding lower suspension density, the analysis revealed that in case of all OXA-like carbapenemase producers (class D) as well as P. aeruginosa 3 strain time needed to achieve positive results (H 4McF) was longer compared to the class A and B producers except for C. freundii 9 (no positive results at 4 McF noted) (Group II, Fig 4B; S2 Data). The same applies to time to obtain maximum hydrolyzation level (H max 4 McF). Strains clustered as Group I on the PCA graph were characterized by the shortest time needed to achieve positive as well as maximum results of hydrolytic activity and was represented by all class A producers as well as almost half of the β-metallo-carbapenemases producing strains (4 out of 9). Analysis revealed that those strains required no more than 3 minutes of incubation time to give positive results when OD 4 McF was used while in the case of class D producers 10 and more minutes were needed (S1 Table; S2 Data).

PLOS ONE
Study on carbapenemase-producing bacteria by matrix-assisted laser desorption/ionization approach Considering the highest bacterial density (7 McF), the analysis did not reveal differences between the class of produced carbapenemases since in all cases except for C. freundii 9 (4 min) the first time point which revealed positive results was 1 min (S1 Table). Instead, differences were noted in case of time needed to achieve maximum hydrolyzation level, where, 4 of class B producing strains-P. aeruginosa 4, C. freundii 8/9, and E. kobei 11, reached the highest logRQ after at least 15 min. of incubation. In the case of the rest strains H mas 7 McF did not exceeded 5 minutes.

Ranking the strains
Based on the PCA results, strain ranking related to the levels of hydrolytic activity in comparison with results of the standard procedure was performed ( Table 3).
Based on the total ranking, the analysis showed good compliance between tested and standard protocol in 53% of cases and decreased to 33% if only time ranking is accounted for. Ranking based on the suspension assay was similar to standard ranking; however, for 40% of the strains only weak compliance was revealed. The best rank matching was obtained for Pseudomonas aeruginosa, E. coli, and Salmonella sp. 10 strains while the highest discrepancies were detected in the case of multi-carbapenemase-producer K. pneumoniae 12 ranked as 1 and 12 the most active strain according to the experimental test and standard procedure, respectively.

Results of β-metallo-carbapenemases inhibition assay
The use of the MBT STAR-Carba test to track the effect of carbapenemase inhibition in the tested range of cell suspension density revealed great differences between the type of inhibitor used as well as type of β-metallo-carbapenemases (Table 4). Exemplary graphs created via MBT Biotyper Prototype software are presented in Fig 3C and 3D. Moreover, a control test with the use of a non-MBL inhibitor-2 mg/mL PBA-did not reveal any significant inhibitory effects (S2 Table), thereby confirming that the noted observation actually reflects a DPA and EDTA influence and that they are not an artifact. Considering inhibitors, tested strains proved to be typically more sensitive to 1.5 mg/ml DPA addition-with a significant decrease in logRQ values in all strains regardless of the suspension density used (Table 4;   was observed for NDM-like carbapenemase producing E. coli 7 and K. pneumoniae 12 as well as in case of C. freundii 9; however, the latter one demonstrated relatively low overall hydrolytic activity which was starting only from a 5 McF suspension.

Discussion
Design spectrometric approach proved to be a universal, rapid, and highly sensitive tool for detecting present hydrolytic activity amongst the range of carbapenemases tested, including class A, B, and D. Also, class B GES type carbapenemases were detected, albeit well known to exhibit relatively low hydrolytic activity and therefore to challenge many current colorimetric tests (e.g. β-Carba Test, Biorad) [26,27]. Although the results obtained by the standard procedure revealed great differences in activity levels between individual Enterobacterales and Pseudomonas strains-with typically P. aeruginosa strains being characterized by lower logRQ values-the test does not allow to ascribe differences in activity to different types or classes of carbapenemases. In this study, carbapenemase-type specific hydrolytic differences were revealed by modifying the standard test procedure regarding the amount of bacterial material, incubation time, and application of inhibitors. In most cases, there is a clear relationship of the modified parameters and the detected activity with the standard procedure being typically in an optimal range for sensitive detection. In most cases, logRQ values of 7 McF variant closed to that obtained using 1 uL inoculation loop which suggests transferring the similar number of cells. Indeed, 7 McF suspension density corresponds to ca. 3 x 10 8 colony forming units (cfu) and theoretical concentration in 1 uL inoculation loop ranged 3-6 x 10 8 CFU/mL as followed manufacturer guidelines. For GES-harbouring P. aeruginosa strain, only the standard procedure was detecting the carbapenemase and false-negative results were noted when the modified procedure was applied what resulted from the lower amount of bacterial cells and indicated the reduced activity of this type of carbapenemases. Comparing other types, most metallo-type class B carbapenemase-producing bacteria demonstrated relevant hydrolytic activity also in a wider McF range compared to the class D producers tested, a class of β-lactamases also known to be difficult for rapid detection [28]. Moreover, strains for which positive results were observed already at the lowest density tested (1 McF) and thus classified as the most active, also belonged to the class B producers. Burckhardt et al. [29] noted that strains producing carbapenem-hydrolyzing class D (OXA-type) beta-lactamases demonstrate weaker imipenem hydrolytic activity compared to the MBL producers. Our findings are in agreement with the general statement that strains producing class D β-lactamases, especially OXA-48 producers, are considered as difficult-to-detect strains since they are characterized by low carbapenem MICs and often go undetected by most phenotypic methods [30,31]. Nevertheless, the use of design spectrometric assay can effectively help not only to reliably detect the activity of oxacillinases, but also in their further classification. Similar effect was achieved by Á lvarez-Buylla et al. [32] using dipicolinic acid and Zn 2+ supplementation. Interestingly, similar classification of the strains into the metallo-and OXA-like producers can be performed based on the time needed to present relevant hydrolytic activity, where the oxacillinases-producing strains happened to need longer time-to-result (Fig 4B).
The observed lower hydrolytic activity of P. aeruginosa strains compared to that in Enterobacterales has previously been reported in the literature and was associated, among others, with lower quality of carbapenemase producing P. aeruginosa spectra [20]. Results of our studies suggest that the explanation of such findings could also be a different behavior of P. aeruginosa strains under increasing suspension density compared to Enterobacterales members. Possible explanation of such phenomenon could be the fact that P. aeruginosa has the genetic capacity to express a wide repertoire of carbapenems resistance mechanisms-besides enzymatic antibiotic degradation also expression of efflux pumps and low outer membrane permeability [33,34]. Moreover, Potron et al. [35] claim that carbapenem resistance in P. aeruginosa is mostly related to porin deficiency. In this context, the observed gradual decrease of hydrolytic activity with increasing suspension density in higher OD range may be due to the superiority of the efflux resistance mechanism and lowering membrane permeability leading to the limitation of the intracellular penetration of carbapenems. A similar observation was noted in the work of Miltgen et al. [36], in which disruption of the outer membrane by high concentrations of colistin and a freezing-thawing step was proposed as a solution for improving detection performance. P. aeruginosa is also known for their ability to synthesize significant amounts of siderophores demonstrating an affinity for Zn 2+ [37] which could affect the availability of this metal for carbapenemases. However, the available literature lacks any work that addresses the impact of siderophore synthesis on carbapenemases activity detection. Due to restricted number of investigated P. aeruginosa strains drawing of general findings is limited, nevertheless, our findings prove a great need to deepen studies on the carbapenemases activity within Pseudomonas species conducted on the bigger scale and applying various conditions like different amount of both Zn 2+ and biomass.
Multiple studies evaluated the time needed for reliable carbapenemases activity detection via MALDI-TOF MS [29,32,38]. However, all of them involve homebrew, non-standardized protocols. Our studies revealed that for the vast majority of bacteria tested (13 out of 15) incubation in the presence of the benchmark antibiotic could be shortened to 1 min if McF 7 suspension is used, a value also reached by the standard procedure. This applies to both Enterobacterales and P. aeruginosa strain. Considering that in almost half of the investigated cases, 1 min of incubation was enough to reveal the maximum logRQ value, we can suppose that enzymatic degradation of carbapenem molecules occurred immediately after mixing bacterial suspension with antibiotic solution. Therefore, the spectrometric assay could be potentially used for reliable carbapenemases detection in many cases of Enterobacterales and Pseudomonas strains, almost omitting the incubation step making this test even shorter. Similarly, Chong et colleagues [39] revealed complete imipenem hydrolization right after reaction mixture preparation (T = 0) for 85% of class A carbapenemase-producers, however, the majority of class B and D carbapenemase-producers did not fully hydrolyze imipenem until the 2 h of incubation. Such discrepancies may result from different reaction buffer composition and the applied mass spectrometric detection compared to that in our studies. As Knox and Palombo [30] observed, the addition of zinc ions is essential for consistent detection of many class B carbapenemases producers, like NDM-and VIM-harbouring strains and generally should be a requirement during the investigation of metallo-carbapenemase producers in these assays.
This approach has become particularly important in regard of increasing use of MALDI--TOF MS in clinical microbiology. Results of the works from the last few years regarding the use of MALDI TOF MS analysis for investigation of different carbapenemase activities-under addition of MBL inhibitors (e.g. ertapenem hydrolysis by IMP-and VIM-harbouring enterobacteria and P. aeruginosa using 50 mg/mL EDTA [40], by NDM-producing K. pneumoniae using 0.05 mg/mL DPA [3] or by VIM-and NDM-like K. pneumoniae as well as VIM-like P. aeruginosa using 1.5 mg/mL DPA [41]) makes this methodology a good option as a tool for the evaluation of new effective combinations of β-lactams and their inhibitors in any modern laboratory routine. Nevertheless, the authors referred to the use of ertapenem, which implicates the need to conduct a relatively long incubation to obtain correct results-up to 2 hours. Similarly to our work, recent studies of Oviano et colleagues [42,43] showed that the use of imipenem gives the opportunity to obtain reliable results of susceptibility testing in a simple way already after 30 minutes as it was in the case of combinations: imipenem-avibactam and imipenem-relebactam. Although results of mentioned works showed high sensitivity (up to 98%) and specificity (up to 100%) in detecting KPC-producing Enterobacterales susceptible to imipenem/relebactam [43] as well as for detecting KPC and OXA-48-type isolates through avibactam inhibition [42], authors faced the obstacles with the repeated indeterminate results for certain strains or inability to distinguish between the types of carbapenemases. As we suppose, those limitations could be associated with the use of 1 uL microbial loop containing only an approximate number of cells. Contrary to this, in our studies, we used the range of bacterial suspension densities which enabled revealing differences in susceptibility between types of βmetallo-carbapenemases and inhibitors. Thus, conducting tests in a wide range of certain suspension densities opens the possibility of sensitivity comparisons between strains on a given inhibitor depending on the type of carbapenemase produced, solve the problem of indeterminate results as well as has the potential to facilitate interlaboratory comparisons. Considering the fact that special emphasis is currently placed on discovering new clinically significant MBL inhibitors like VNRX-5133 (novel boronic acid, phase I), ANT431 (novel pyridine-2-carboxylic acid, preclinical), or potentially also avibactam (phase III) the development of which is at different phases of clinical trials [44], our proposed use of design assay demonstrates high potential as a surrogate test for susceptibility before the microdilution results become publicly available. In the future investigation, inhibition effect of the other classes of carbapenemases (A, D) using the spectrometric assay is planned, since our findings suggest, as it was in the case of MBL, it may turn out to be a reliable and fast tool for clinical diagnostic laboratories also in this field.

Conclusions
The new spectrometric assay has become known as a highly sensitive and robust tool for a very rapid detection of carbapenemases, including the broad spectrum of β-metallo-carbapenemases. The outcome of our research indicates that, besides using the standard workflow, providing highly sensitive results, procedural modifications of this assay, like varying suspension densities and incubation times, demonstrate great potential in further sub-characterization of samples. The investigation of activity level differences based on normalized logRQ values allowed not only the differentiation of various bacterial species on the order level but also a correlation of different classes of carbapenemases, even within the class B metallo β-lactamases. It also proved that the application of a defined suspension density gives the opportunity to shorten incubation time and thus the total time-to-result of analysis along with further increasing knowledge about differences in the antibiotic hydrolysis level and sensitivity to inhibitors addition of an investigated strain. Such information could be useful during initial adaption towards a proper antibiotic therapy as well as during development of new MBL inhibitors.
Supporting information S1 Table. The