Evaluation of the Autof MS1000 mass spectrometer in the identication of clinical isolates

Background: To evaluate the accuracy and performance of the Autof MS1000 mass spectrometer in bacteria and yeast identication, 2,342 isolates were obtained from microbial cultures of clinical specimens (e.g. blood, cerebrospinal uid, respiratory tract samples, lumbar puncture uid, wound samples, stool, and urine) collected in 2019 in Henan Provincial People’s Hospital. Repetitive strains from the same patient were excluded. We tested the Autof MS1000 and Bruker Biotyper mass spectrometry systems and the classical biochemical identication system VITEK 2/API 20C AUX. Inconsistencies in strain identication among the three systems were identied by 16S rDNA and gene sequencing. Results: At the species level, the Autof MS1000 and Bruker Biotyper systems had isolate identication accuracies of 98.9% and 98.5%, respectively. At the genus level, the Autof MS1000 and Bruker Biotyper systems were 99.7% and 99.4% accurate, respectively. The instruments did not signicantly differ in identication accuracy at either taxonomic level. The frequencies of unreliable identication were 1.1% (26/2,342) for the Autof MS1000 and 1.5% (34/2,342) for the Bruker Biotyper. In vitro experiments demonstrated that the coincidence rate of the Autof MS1000 mass spectrometer in the identication of ve types of bacteria was >93%, the identication error rate was <3%, and the no identication rate was 0. This indicates that the Autof MS1000 system is acceptable for identication. Conclusions: The Autof MS1000 mass spectrometer can be utilised to identify clinical isolates. However, an upgradation of the database is recommended to correctly identify rare strains.


Background
Matrix-assisted laser desorption ionisation-time of ight mass spectrometry (MALDI-TOF MS) is an emerging high-throughput technology with broad potential in clinical microbial identi cation because of its high resolution, speed, sensitivity, and accuracy [1][2][3]. Microorganism detection is based on databases of known bacteria. During detection, characteristic protein ngerprints are obtained, and these mass spectra are compared with the database for identi cation [4][5][6][7][8]. Many companies, such as Bruker Daltonics, bioMérieux, Shimadzu, Beijing Purkinje General Instrument Co., and Autobio Diagnostics, manufacture MALDI-TOF MS instruments. Recently, a new MS, the Autof MS1000 from Autobio Diagnostics, was developed for the identi cation of clinically important pathogenic bacteria. The Autof MS1000 has some advantages over the existing systems, such as a ion source vacuum (up to 10 -7 mPa) and a rapid identi cation module that delivers a sample result scan in 0.1 s, and can identify an entire target plate (96 isolates) in approximately 21 min. The mass spectrometer has been purchased by many laboratories in China, the United Kingdom, Italy, South Korea, and Thailand. The aim of this study was to evaluate the identi cation ability of the domestic Autof MS1000 in common clinical microbiology. A commercial Bruker Biotyper mass spectrometer (Bruker Daltonics, Bremen, Germany) was used as the control system. The results provide a reference for further assessment of this instrument in the medical device market.

Isolate identi cation
There were no statistically signi cant differences in the identi cation of the 2,342 strains between the two mass spectrometers at either the species or genus level. The Autof MS1000 and Bruker Biotyper systems had isolate identi cation accuracies of 98.9% and 98.5%, respectively, at the species level, and 99.7% and 99.4%, respectively, at the genus level. These results demonstrate that the Autof MS1000 and Bruker Biotyper mass spectrometers had equal ability to identify clinical isolates. Detailed results are shown in Figure 1 and Additional le 1. Common bacteria and yeast were routinely obtained from microbial cultures of clinical specimens, and Escherichia coli, Klebsiella pneumoniae, Acinetobacter baumannii, Staphylococcus aureus, and Pseudomonas aeruginosa were most common clinical isolates. The identi cation accuracy for common bacteria and yeast reached 98% on the Autof MS1000. The ability to identify these isolates is of great signi cance to the evaluation of the MS. Detailed results are shown in Figure 2 and Additional le 2.

Failure rates
The Autof MS1000 incorrectly identi ed or failed to identify 1.1% (26/2,342) of the isolates. Of these, 20 strains were identi ed at the genus level. The Bruker Biotyper incorrectly identi ed or failed to identify 1.6% (37/2,342) of isolates, 21 of which were accurately identi ed at the genus level. The two strains of Burkholderia pseudomallei were identi ed correctly by the Autof MS1000, while the Bruker instrument failed to identify them. B. pseudomallei can cause melioidosis, making it an important strain with clinical signi cance [9,10]. This is a major error and should be noted. The Autof instrument identi ed nine strains of Salmonella spp. and accurately identi ed a strain of Salmonella enteritidis to the species level. The Bruker instrument identi ed eight strains of Salmonella spp. and failed to identify one strain of Salmonella paratyphi A. Neither machine can be used for serotype identi cation; therefore, Salmonella spp. identi ed by mass spectrometry will require further serological typing before deciding whether to report an infectious disease. Most other identi cation errors were minor, such as Citrobacterfreundii and Raoultella planticola being erroneously identi ed as Citrobacter braakii and Raoultella ornithinolytica, respectively (Table 1). Fortunately, these results will not affect clinical diagnosis or treatment decisions. We evaluated 229 strains of Gram-negative Enterobacteriaceae, Gram-negative non-Enterobacteriaceae, Gram-negative fastidious bacteria, Gram-positive aerobic bacteria, and anaerobic bacteria, as well as yeasts and yeast-like microorganisms, according to the recommendations for the in vitro performance veri cation of commercial instruments in the Clinical and Laboratory Standards Institute (CLSI) M52 standard [11]. We compared the agreement, discrepancy, and unidenti ed isolates between the two mass spectrometers. The agreement values of both instruments were >93%, their discrepancies were <3%, and <2% of isolates were not identi ed ( Figure 3, Additional le 3). These are all acceptable values, indicating that the Autof MS1000 is a reliable system for isolate identi cation.

Discussion
Bacterial identi cation is of great clinical signi cance, helping clinicians select antibiotics, accurately treat patients, and improve cure rates. To our knowledge, this is the rst assessment of the identi cation of multiple bacteria using a Chinese mass spectrometer in central China. There were no major differences in the identi cation of multiple bacteria between the Chinese instrument and an imported mass spectrometer.
MALDI-TOF MS has advanced rapidly in recent years and is gradually replacing biochemical methods as the preferred tool for clinical bacterial identi cation [12][13][14][15]. The accuracy of MALDI-TOF MS identi cation depends on the collection of protein ngerprint data for all possible strains in a database [16]. The Autof MS1000 has a database of 9,050 strains and 2,727 species, and the Bruker Biotyper database has 5,989 strains and 2,371 species. Comparing the accuracy of strain identi cation is primarily a function of comparing strain databases; therefore, construction of the database (coverage, type, etc.) is critical [17].
For Gram-negative bacilli isolates requiring species-level identi cation (n =1,449), the analytical accuracies of the two systems were similar (99.0% and 98.7% for the Autof MS1000 and Bruker Biotyper systems, respectively (p=0.490)). However, certain closely related microorganisms cannot be distinguished from one another using MALDI-TOF MS, such as Aeromonas, Raoultella, Enterobacter, Acinetobacter, and Citrobacter spp. Similar conclusions have been reported by several other researchers, who could not distinguish these closely related species [18][19][20][21][22]. Hence, for closely related species or subspecies, MALDI-TOF MS should be used in combination with biochemical and molecular methods. For Salmonella spp. identi cation, the limitations of MALDI-TOF MS must be considered [23]. Biochemical and serological tests will still be required to accurately identify Salmonella spp.
In this experiment, a coagulase-negative staphylococcus was isolated from blood cultures. Species-level reporting is sometimes essential to determine the clinical signi cance of culture isolates of coagulasenegative staphylococci [24]. The Autof MS1000 allowed better identi cation of Staphylococcus hominis and Staphylococcus haemolyticus than the Bruker Biotyper, suggesting that the Autof MS1000 has increased speci city for the identi cation of these species. However, this will require further veri cation with increased sample size and additional species. Viridans streptococci isolates were all correctly identi ed by the Bruker Biotyper, whereas two strains of Streptococcus constellatus were not identi ed to the species level by the Autof MS1000. Using the Bruker Biotyper, one strain of Leifsonia shinshuensis, one strain of Staphylococcus gallinarum, two strains of Burkholderia pseudomallei, and one strain of Kazachstania servazzii could not be identi ed, as these strains were not included in the Bruker Biotyper database (v5.0 5898). Database updates may resolve the di culties in distinguishing these species.
Among the rare strains that were misidenti ed, Mycobacterium spp., Nocardia spp., and Actinomyces spp. were not correctly identi ed by either system. MALDI-TOF MS does have limitations in the identi cation of mycobacteria, Nocardia spp. and other aerobic actinomycetes found in the clinical microbiology laboratory [25]. This is due to the presence of multiple strains, which are not fully represented in the database. Although some strains are included, they cannot be accurately identi ed even with repeated operations. It may be that the protein pro les they produce are inconsistent with the characteristic pro le in the database. In that case, the strain diversity of the database should be increased. Another limitation in the use of MALDI-TOF MS with slowly growing Mycobacterium spp. and Actinomyces spp. is the potential occurrence of the two species in mixed cultures, which will be recognised as the colonies on the culture plate mature, but are misidenti ed by MALDI-TOF MS. In addition, the sample preparation method may be an important factor for successful identi cation, particularly for species that are di cult to lyse, such as Mycobacterium spp. and Nocardia spp. A twostep cell disruption protocol combining the use of 0.5-mm diameter silica/zirconia beads and sonication for 15 min greatly improves the e cacy of mycobacterial identi cation by MALDI-TOF MS [26].
This study has some limitations. First, the sample size should be increased, and the species detected should be expanded to include more rare bacteria. Second, we did not evaluate the identi cation of lamentous fungi. Therefore, we will increase the sample size and analyse lamentous fungi identi cation in subsequent evaluations.

Conclusions
In summary, both the Autof MS1000 and Bruker Biotyper meet the clinical requirements for bacterial identi cation. However, for some closely related bacteria, accurate identi cations should be obtained by combining morphological, phenotypic, and molecular characteristics. A lack of diversity in database strains is also a major factor affecting the ability to identify bacteria by MALDI-TOF MS [27]. MALDI-TOF MS databases are constantly expanding, and instrument databases should be regularly updated to ensure optimal isolate identi cation. Identi cation using the Vitek 2 Compact and API 20C AUX system Based on colony morphology and staining results, a corresponding identi cation card was selected for each isolate. Identi cation results were automatically interpreted by the system according to the product manual, using the established algorithm. When the isolate was properly assigned to a given species or identi ed with low discrimination but resolved by supplemental tests, the identi cation was considered reliable.

Bacterial identi cation by MALDI-TOF MS
MS quality control and operation were performed according to the CLSI M58 standard [28] and the Chinese Expert Consensus for Clinical Microbial Mass Spectrometry Application [29]. Deposit preparation and analysis were similar for both systems. For the Autof MS1000, protein spectra were analysed with Autof Acquirer version 1.0.55 software and library v1.1.0 9050. The manufacturer's interpretation criteria were applied, with identi cation scores ≥9 considered positive at the species level, scores of 6-9 considered positive at the genus level, and scores <6 de ned as not identi ed.
On the Bruker Biotyper, extraction procedures were performed according to the product manual. Protein spectra were analysed with Bruker Biotyper 3.1 software and library v5.0 5898. The manufacturer's interpretation criteria were applied, with identi cation scores ≥2.0 considered positive at the species level, scores of 1.7-2 considered positive at the genus level, and scores <1.7 de ned as not identi ed.

Sequencing
For certain strains, when both mass spectrometry identi cations and the biochemical identi cation were inconsistent at the species level, the isolate was sent to Beijing Ruiboxing Co., Ltd. (Beijing, China) for con rmation by sequencing. If the mass spectrometry identi cations at the species and genus levels were inconsistent with the results of 16S rDNA sequencing, then the mass spectrometry results were considered incorrect. The 16S rRNA genes of all bacteria were sequenced, along with dnaJ, sodA, tuf, or ropB for Gram-positive cocci [30,31]; ropB, gyrB, recA, or cpn60 for Gram-negative bacteria [30,32]; and ropB, gyrB, SecA1, or hsp65 for Gram-positive bacilli [30,33]. For yeasts, the internal transcribed spacer located between the nuclear 18S and 26S rRNA genes was sequenced [30].

Selection principles for the performance evaluation strains
Non-reference methods were used for comparison according to the CLSI M52 standard for the veri cation of the in vitro performance of commercial instruments [11]. Five kinds of bacteria (Gram-negative Enterobacteriaceae, Gram-negative non-Enterobacteriaceae, Gram-negative fastidious bacteria, Grampositive aerobic bacteria, and anaerobic bacteria), and yeast-like fungi were evaluated using three parameters: the agreement (% agreement), identi cation error (% discrepancy) and unidenti ed species (% not identi ed) rates. Methods with ≥93% agreement, <3% discrepancy, and <2% of species not identi ed were considered acceptable.

Declarations
Ethics approval and consent to participate This study was approved by the Ethics Committee of Henan Provincial People's Hospital, Henan, China (20190050); the need for written informed consent was waived because de-identi ed retrospective data were used. The permission of the corresponding author was required to access the raw data/samples.

Not applicable
Availability of data and materials The data sets used during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This study was supported by the Joint Programs of the Henan Provincial Government and Healthy Ministry (SB2018084, SB20190318, and LHGJ20190611) and Henan Provincial Key Programs in Science and Technology (181200211900 and 202102310355). The funders had no role in the design of the study, the collection, analysis, and interpretation of data, or in the writing the manuscript.
Authors' contributions QM, YY, and YL contributed to the writing of the manuscript; SL, SW, and JX analysed and interpreted the experimental data; WY, JZ, QZ, and MG analysed the mass spectrometry results; YY and QM performed statistical analyses and data plotting. All authors read and approved the nal manuscript.