Klebsiella pneumoniae and Enterobacter cloacae complex were selected as test microorganisms in this study, both pathogens are common causes of nosocomial infection outbreaks[4, 10, 20].
Traditional methods of bacterial strain typing are divided into three main categories according to their principles[5]. The first method is to type strains by comparing differences in DNA bands (fragments) produced after DNA amplification or cleavage[5]. The second method is to differentiate strains based on the sequencing of raw nucleotide sequences, which is highly reproducible compared to DNA banding pattern-based methods[5]. The third method is the hybridization of the DNA fragment (target) to be tested with a DNA fragment (probe) which is a known sequence[5]. This method is widely used to detect DNA mutations. Each of these genotyping methods has its advantages and disadvantages. PFGE is often considered as the gold standard for epidemiological studies due to its high resolution[21]. However, this method is cumbersome, has a long turnaround time, has insufficient reproducibility and no comparison results between laboratories, requires high DNA quality, and has insufficient ability to distinguish between bands of almost the same size[5, 22]. Unlike PFGE, MLST data can be easily stored in a database for data exchange within or between laboratories[23]. The disadvantages of this approach are that gene sequencing is more costly and time-consuming and cannot be integrated into daily work. WGS is based on the complete genetic information of the strain, and as a high-resolution tool, it could become a potential gold standard for future genotyping[4, 5]. The limitations of WGS are the high cost, redundant data, and difficult analysis[24].
Matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) is currently widely used for bacterial identification, but its ability in bacterial strain typing is still controversial. In this study, we collected 126 CRKP strains and 24 ECC strains and clustered them using different typing methods. The results showed that MALDI-TOF MS had a low match with MLST (ARI: CRKP, 0.403; ECC, 0.120) and changing the algorithm and parameters can only slightly increase the ARI (increased from 0.403 to 0.572; increased from 0.120 to 0.419). Khennouchi et al. analyzed 77 Enterobacter isolates and found that although MALDI-TOF MS was able to successfully distinguish isolates based on geographic distribution and species, the match with MLST-based phylogenetic tree results was relatively low[10]. We analyzed CRKP strains from the ICU using MALDI-TOF MS and PFGE and acquired an ARI of 0.078, indicating poor agreement between the results provided by both. Rodrigues et al. collected 83 multi-drug resistant K. pneumoniae strains and found that MALDI-TOF MS correlated poorly with the two reference typing methods (MLST and PFGE), probably due to the high similarity observed among the mass profiles of K. pneumoniae lineages[14]. However, Fang L et al. concluded that the results of MALDI-TOF MS and PFGE were similar and that MALDI-TOF MS could largely predict the results of PFGE[9]. In our study, the ARI of MALDI-TOF MS and PFGE was 0.727 when 10 ECC strains were used as the study background, and the similarity of the results was high. Whole-genome sequencing, as a possible future gold standard for genotyping, has also been used to compare with MALDI-TOF MS in many studies. The studies by Casanova et al. and Rödel et al. both showed that WGS confirmed the clustering results of MALDI-TOF MS[25, 26]. However, Dinkelacker et al. concluded that the results of MALDI-TOF MS and WGS have lower consistency compared with Fourier transform infrared spectroscopy (FTIR)[4]. We speculate that the reason for this discrepancy may be that the first two groups analyzed possible outbreak strains collected at the NICU with a more concentrated collection time, while the latter group was collected over a span of one year.
Although the ability of the analysis software accompanying the mass spectrometer to type bacteria is controversial, some groups obtained relatively good results by screening specific mass peaks and then performing mass spectrometric analysis. For example, in a study, the specific mass peaks of one standard Staphylococcus aureus and 57 methicillin-resistant S. aureus (MRSA) strains were screened and then validated with an additional collection of 24 clinical MRSA outbreak strains. These specific mass peaks were found to successfully identify strain clones and were in full agreement with PFGE results[13]. In addition, resistant isolates can also be identified by detecting resistance-related specific mass peaks[27].
Many authors have evaluated the ability of MALDI-TOF MS to detect outbreaks. Bar-Meir et al. found that MALDI-TOF MS could classify 33 outbreak strains into two clones that correlated with patient epidemiological characteristics and clustering results were consistent with cg-MLST[28]. Adrian Egli et al. reanalyzed six ESBL-producing Escherichia coli strains collected during hospital outbreaks and included two unrelated ESBL-producing control E. coli strains, and identified outbreaks by MALDI-TOF MS typing, with results similar to PFGE[11]. In this study, we found three possible outbreaks by grouping and clustering all isolates according to criteria such as strain type, department of origin, and patient ward. Our hospital had one possible outbreak of ECC strains carrying blaNDM−1, blaACC, blaTEM, blaSHV, acc(6')lb-cr in the pediatric, neonatal ward and PICU between December 7, 2018, and March 20, 2019. Outbreaks of ECC in neonatal wards have been reported nationally several times and may be transmitted with the help of medical devices and medical staff, while MALDI-TOF MS allows rapid detection of the relationship between samples from food and other sources and outbreak strains isolated from patients, thus guiding rapid clinical action[26, 29]. An outbreak of CRKP was also found in our neonatal ward, and all four strains carried blaNDM−1, three strains belonged to ST1887 and one strain belonged to ST6279. The strain characteristics were similar in the ICU unit due to the concentrated distribution of patients and long hospital stays, so after analyzing clusters according to the geographical location of the ward, an outbreak of CRKP carrying blaKPC−2 was found in wards A, B, and C, indicating that MALDI-TOF MS can also differentiate strains with high similarity to some extent, but the number of strains should not be too large. Previous studies have shown that NDM is the major carbapenemase in children, while KPC is the major carbapenemase of Klebsiella pneumoniae in Chinese adults[30, 31]. Since this study was retrospective, it was not possible to determine the specific route of transmission of the strains and whether the environment played a role in this.