Rapidly quantitative antimicrobial susceptibility testing for Klebsiella pneumoniae using Matrix-Assisted Laser Desorption Ionization-Time of Flight Mass Spectrometry under optimal conditions

Klebsiella pneumoniae infections, especially Carbapenem-resistant Klebsiella pneumoniae (CRKP), have become an “Urgent Threats” with high morbidity and mortality. Therefore, rapidly determining of the susceptibility and timely choosing an appropriate antibiotic were the important premises of the treatment of Klebsiella pneumoniae infections. The present study was rst to explore the tness of matrix-assisted laser desorption ionization-time of ight mass spectrometry (MALDI-TOF MS) in quantitative rapid antimicrobial susceptibility testing (RAST) with optimal conditions.

Klebsiella pneumoniae was an opportunistic pathogen of clinical importance and the CRKP was still listed as an "Urgent Threats" in the 2019 Antibiotic Resistance Threats Report by the Centers for Disease Control and Prevention (CDC) [1], which was one of the most prioritized threats of antibiotics resistance in the healthcare facilities, warranting urgent and aggressive action. The promptly increasing resistance and the resulting high mortality of CRKP that admit of no optimism in many countries [2][3][4]presented a global challenge. In particular, the mortality of the bloodstream infections (BSIs) caused by CRKP was as high as 71.9% [5], three times higher than infections in other sites. Therefore, rapidly determining of the MIC values and timely choosing an appropriate antibiotic were the important premises of the treatment of patients with Klebsiella pneumoniae infections.
However, the reporting time of the routine AST, such as E-test, disk diffusion, BMD and VITEK-2 system, can take up to at least one day[6], because these routine phenotypic assays required extensive proliferation to reach detection level. Technically, MALDI-TOF MS [7], integrated micro uidic device[8], biological small-molecule assays [9], diffraction-based method [10] and other innovations for RAST were ourishing to shorten the turnaround time and accelerate clinical diagnostics. The above MALDI-TOF MS not only was the rst-line tool [11] to identify pathogens with simplicity, speediness, cost-e ciency, highthroughput and high-resolution, but has been developed for rapid detection of antimicrobial resistance.
Furtherly, MALDI-TOF MS-based RAST about Klebsiella pneumoniae were mainly divided into the following three approaches: 1. Identify the MS peaks of speci c protein associated with resistance. The strong correlation between bla KPC -harboring pKpQIL-like plasmids [12][13]and the 11,109 m/z MS peak has been con rmed with excellent speci city (100%) and sensitivity (85.1%) [14], which succeeded in the real-time detection of KPC-producing CRKP during the routine clinical identi cation. Apart from carbapenemases, loss or modi cation of the major outer membrane porins (OMPs) OmpK35 and OmpK36 also aggravated the carbapenem-resistance of Klebsiella pneumoniae. Hu et al. [15] [19], MS ratio [20] and logarithm of the quanti cation of resistance (LogRQ) values [21] and other algorithms to semi-quantitatively differentiate sensitive and resistant strains. 3. Compare the bacterial growth by MALDI-TOF MS in the presence or absence of antibiotic. The most advantage of growth-based RAST was the universality, which independently from the speci c resistance mechanism. For Klebsiella pneumoniae, Lange et al. [22] rstly developed the MALDI Biotyper antibiotic susceptibility test rapid assay (MBT-ASTRA), extracting bacterial protein after incubation with meropenem breakpoint concentration, adding an internal standard, normalizing the maximum peak and calculating the relative growth value to semi-quantitatively detect the resistance.
And the newly developed direct-on-targt microdroplet growth asssy (DOT-MGA) [23] was also growthbased RAST but with 6µL micro-droplet and more simpli ed processing process.
Recently studies have achieved qualitative [24][25] and semi-quantitative AST[26] by MALDI-TOF MS, but there were few studies conducted quantitative tests. Here, we aimed at developing quantitative RAST that are based on MALDI-TOF MS technology implemented in the VITEK MS instrument with optimized assay conditions. According to European Committee on Antimicrobial Susceptibility Testing (EUCAST) or Clinical Laboratory Standards Institute (CLSI) recommendations [27], we adopted the growth-based phenotypic RAST to determine the susceptibility of the Klebsiella pneumoniae, independently from some speci c resistance mechanisms. And to our knowledge, this was the rst study to evaluate the in vitro diagnostic (IVD) mode of the VITEK MS under the routine conditions and the existing parameter settings.

Results
Optimal conditions for MALDI-TOF MS-based quantitative RAST.
Compared with previously published study [25], (i) we directly incubated the diluted bacterial suspension into the 200 µL EP tubes with or without imipenem (IPM), rather than incubating in the at-bottomed 96well plates and transferring into the Eppendorf (EP) tubes and (ii) we skipped the washing step, although some studies washed the precipitable bacteria with pure water and 70% alcohol after the rst time centrifugation. We found that no washing had no signi cant effect on strain identi cation (Fig. 1). Conversely, transferring into the EP tubes and washing the precipitable bacteria would cause unnecessary loss of bacteria, which led to MIC values identi ed by MALDI-TOF MS (MS-MIC) decline to lower concentrations. Further research demonstrated that adding 3 µL of FA/ACN was enough to split bacteria and extract protein in the preprocessing step. Excessive addition may diluted the protein concentration to reduce sensitivity. And the optimal volume of the supernatant on the VITEK MS target plate was 2 µL. Using this optimized combination, the validity for the 25-fold, 50-fold and 100-fold dilution reached 100% in the 2hr, 3hr and 3hr, respectively.

Consistency between MS-MIC and BMD-MIC.
The results of the BMD showed that the ranges of the MIC values in CRKP and CSKP strains were 32 256 µg/mL (MIC 50/90 , 128/256 µg/mL) and 0.25 2 µg/mL (MIC 50/90 , 1/1 µg/mL), respectively. In this study, all strains were rstly tested in the 100-fold dilution ( nal concentration 5×10 5 CFU/mL) as the same as Subsequently, we applied the initial inoculum size, which was higher than CLSI-recommended inoculum (5×10 5 CFU/mL with an acceptable range of 2×10 5 CFU/mL to 8 ×10 5 CFU/mL) as a precautionary measure to reduce the shifting to a lower concentration gradient and compensate the reduction in incubation time. In the pre-experimental phase, we con rmed the dilution of 50-fold ( nal concentration 1×10 6 CFU/mL ) and 25-fold ( nal concentration 2×10 6 CFU/mL ), which initial biomass could not reach the limit of detection (LOD) of MALDI-TOF MS, but could after a short-term incubation. the high concentration gradient (256, 128, 128, 128, 128 µg/mL, respectively), while the latter was lower (64, 32 µg/mL, respectively). Along with the time extended to 3 hr, CRKP strains had higher MS-MIC values compared with BMD-MIC values, which accounted for 33.33%(10/30). For CSKP strains, although the consistency of incubation at 25-fold dilution for 2hr cannot hold a candle to the consistency of CRKP strains, it created the highest consistency of the 66.67%(20/30) in all protocols. And the main reason of inconsistency was that the MS-MIC values transferred a higher concentration gradient but no major error (ME, false resistance) in susceptibility interpretation occurred. In addition, the consistency of the MIC values detected by MALDI-TOF MS and BMD of the incubation at 50-fold dilution for 4hr in CRKP strains was nearly the same as that at 25-fold for 2hr but with a large time cost. However, shortened the time to 3hr, the consistency in CRKP strains suddenly dropped to 46.67% and the consistency showed no statistical signi cance (P=0.584) in CSKP strains. Meanwhile, the results of three repeatability tests indicated that the MS-MIC value distributions of CSKP strain was more stable than that of CRKP strain (Fig. 3).

Discussion
MBT-ASTRA [22] and DOT-MGA [23] were the emerging technologies of the MALDI-TOF MS-based qualitative AST, but the former was limited to the speci c brand of equipment to calculate the MS peak area under curve (AUCs) with a classi cation between susceptibility and resistance. The application of the latter was universality and simplicity, but merely 'touching' of the microdroplets with a tissue wipe to removed the broth would exert an in uence on the results. In this study, we combined the MBT-ASTRA and DOT-MGA and successfully applied the MALDI-TOF MS to quantitatively detect MIC values of 30 CRKP strains and 30 CSKP strains under the optimized conditions within 2hr and further verify the reproducibility.
Compared with previous studies, we optimized the sample processing process. One of the reasons why BMD used at-bottomed 96-well plates was facilitated to visual reading the turbidity after the 18-20hr incubation to identity the MIC values. However, the incubation time of the MALDI-TOF MS-based AST was too short to generate visual differences between the bacterial suspensions incubated with different concentration of IPM. Directly incubated in the EP tubes and skipped the washing step can not only simplify the procedure and and shorten the reporting time, but reduce the unnecessary loss of bacteria biomass during the processing stage, which prevented the MS-MIC values from declining to a lower concentration and improved the sensitivity.
The prospect of the growth-based phenotypic RAST was partly hindered by the duration of the lag phase[28], which was the time needed for strains to cope with the constraints of their environment before proliferation. Within a population, the intensity of respiration repression [29] caused the considerable heterogeneity in the lag duration and growth rate between individuals. On the other hand, it was found that the main problem of the incubation with 100-fold dilution of BMD for a short time was the phenomenon of delayed phenotypic resistance. Some strains can express susceptibility at the early stages of incubation, caused the shifts of MS-MIC values to lower concentration gradients. After a long incubation, the strains gradually grew to express the correct MIC values in the presence of high concentrations of antibiotics. In this study, 50% (n=15) of the CRKP strains and 30% (n=9) of the CSKP strains were signi cantly affected by this phenomenon. Wang et al. [30] conducted a RAST for Klebsiella pneumoniae using a suspension of 5×10 5 CFU/mL bacterial cells incubating for 4hr by MALDI TOF MS Research-User-Only (RUO) mode. The 43.33% of the CRKP strains and 13.33% of the CSKP strains were in uenced by the phenomenon of delayed phenotypic resistance to show lower MIC values detected by MALDI-TOF MS compared to those derived from the BMD. The differences of inconsistencies possibly attributed to the acquisition system, algorithm and identi cation rules [31] of RUO and IVD mode respectively. At the same time, 10% (n=3) of the CSKP strains showed an increase in initial biomass. Our nding was consistent with Li et al. [25], who indicated that 8.60% of the susceptibility strains presented the MS-MIC values higher than the corresponding MIC values.
Inoculum effect (IE) [32], i.e. the initial inoculum of bacteria applied on AST may affected the interpretation of MIC values, demonstrated that minor, allowable deviations in inoculum would change the MIC values and thereby potentially affect therapeutic decision points. Therefore, determining the initial inoculum and matched incubation time of bacteria were key factors of implementing MALDI-TOF MS-based quantitative RAST. Compared with incubation with 100-fold dilution for 4hr, incubation with 25fold dilution for 2hr greatly improved the turnaround times and effectively reduced the inconsistency caused by the phenomenon of delayed phenotypic resistance in CRKP strains and CSKP strains by 33.33%(n=10) and 30.00%(n=9), respectively. However, improving the initial inoculum size inevitably led to an increase of 23.33% (n=7) in the inconsistency of CSKP strains. Finally, the limitation in our study was the MIC values of CRKP strains not fell on the breakpoints for resistant categories, which can not obtain a complete resistance pro le of IPM.

Conclusion
With the development of technology and clinical databases, the tness of MALDI-TOF MS for clinical microbiological applications will further expand, especially in RAST. There were considerable potentials for MALDI-TOF MS to provide universal, rapid and mechanism-indepent AST with the optimal conditions, which appeared feasible to drastically accelerate the clinical diagnostics and treatment .

Bacterial strains
A total of 60 random non-replicate clinical Klebsiella pneumoniae strains were collected from the blood of BSI patients at a tertiary hospital in Tianjin, China, between Jan 2018 and Jun 2021, including 30 CRKP and 30 CSKP. Klebsiella pneumoniae clinical strains were identi ed using VITEK MS (Biomérieux, France) and the AST was performed using an automated VITEK 2 Compact Antimicrobial Susceptibility card (Biomérieux, France). Based on the PCR results, all CRKP strains were indenti ed as carbapenemase producer and were positive for bla KPC (n=29) and bla NDM (n=1), corresponded to the phenotypic expression of resistance to imipenem (IPM).

Reference resistance determination
According to CLSI recommendations [27], MIC values were determined by BMD. Brie y, 0.5 McF standard suspension was diluted 100-fold (to give a nal CFU/mL of 1 x 10 6 ) by cation-adjusted Mueller-Hinton broth (CAMHB) and were inoculated triplicately into equal 100 μL of CAMHB in at-bottomed 96well plates (Corning, USA) containing two-fold serial dilutions of IPM giving nal concentrations ranging from 256 to 0.125 μg/mL. Wells with no IPM and wells with CAMHB alone were served as the positive and negative controls, respectively. Before and after incubation at 37℃ overnight, the optical density (OD) value of each well was measured by a micro-plate reader (Biotek Synergy HT, USA) at 600 nm. The BMD-MIC values was considered as the lowest concentration of the IPM that completely inhibited the bacterial growth with the OD value increase ≤ 0.050 [33].

Selection of the optimal conditions
Comparative study of the optimal conditions for MALDI-TOF MS-based quantitative RAST included the volume of FA/ACN, the volume of supernatant spotted on the target plate, incubation time and dilution. Firstly, we optimized the operation and methodology based on the formic acid extraction method, which was recommended prior to MALDI-TOF MS analysis. The volume of FA/ACN and supernatant spotted on the VITEK MS target plate were determined to optimize test performance from 3μL, 5μL, 10μL and 1μL, 2μL, respectively. Subsequently, referring to the preliminary experimental results, a 0.5 McF standard suspension was respectively diluted 25-fold, 50-fold and 100-fold and inoculated into equal 100 μL CAMHB without IPM. After incubated for 1, 2, 3, 4 hr at 37℃, the appropriate incubation time of each dilution was preliminary identi ed when the validity reach 100%, which was de ned as the percentage of strains were successfully identi ed by the MALDI-TOF MS.
MALDI-TOF MS-based quantitative RAST 1. VITEK MS-based analysis. Each dilution was detected in the same way as the BMD with an established incubation time. Different from BMD, bacterial suspensions were directly incubated in 200 μL Eppendorf (EP) tube and centrifuged at 13,000g for 2min [34]. Subsequently, discarded the supernatant, lysised the precipitable bacteria by the optimal volume of 70% formic acid for 10 min and extracted protein by the equal volume of 100% acetonitrile. Centrifuged, spotted optimal volume of the supernatant on the VITEK MS target plate, dried and overlaid with 1μL of α-cyano-4hydroxy-cinnamic acid (α-CHCA). After desiccation, the extractions were analyzed by the IVD mode on the VITEK MS (bioMérieux, France) calibrated with a quality-control strain of Escherichia coli ATCC8739 at a laser frequency of 200Hz, in the mass range of 2,000-20,000 Da.
2. Data capture and analysis. Data analysis was performed using the software Myla 3.2.0 (Biomérieux, France). The growth of bacteria treated with a certain concentration of IPM was quanti ed by the con dence level intended for identi cation. If the strains was not signi cantly inhibited at a certain concentration, the strains will be successfully identi ed with the con dence level of ≥90%. Conversely, con dence level of < 90% or non-identi cation was interpreted as growth inhibition. The MS-MIC values were considered as the lowest concentration of the IPM at which con dence level was < 90% or had no identi cation.

Statistical Analysis
The percentage of observed agreements and the Cohen's kappa coe cient with its corresponding 95% interval con dence were used to evaluate the consistency between the MALDI-TOF MS-based quantitative RAST and BMD of data and the reproducibility of the optimal combination between dilution and incubation time. Kappa values were strati ed as follow: 0.8<Kappa≤1 indicated almost perfect agreement, 0.60<Kappa ≤ 0.80 substantial agreement, 0.4< Kappa ≤0.60 moderate agreement, 0.20<Kappa ≤0.4 fair agreement and 0<Kappa ≤0.20 slight agreement [35]. Statistical signi cance was de ned as <0.05 in a two-tailed test. All statistical analyses were performed using the SPSS 23.0.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.

Availiability of data
Data sets for this study are available on reasonable request from corresponding authors.

Ethical Approval
The experiment has been approved by the Ethics Committee of Tianjin Medical University General Hospital (Ethical NO. IRB2021-WZ-151). All methods were performed in accordance with the relevant guidelines and regulations by including a statement in the Ethics approval and consent to participate section.

Consent to participate
The patient's written informed consent was obtained in accordance with the Declaration of Helsinki.

Consent for publication
Published written informed consent can be obtained from all participants.

Con ict of interest
The authors declared that they had no con ict of interest. Figure 1 The differences in mass spectra of three sample preprocessing protocol. a Direct smear method. b Formic acid/ acetonitrile extraction method without washing: successfully identi ed as Klebsiella pneumoniae with the high background noise. c Formic acid/ acetonitrile extraction method with washing: successfully identi ed as Klebsiella pneumoniae with the mass spectra of higher quality.

Figure 3
Consistency of three repeatable tests in CRKP and CSKP strains.