Novel Strategy for Typing Mycoplasma pneumoniae Isolates by Use of Matrix-Assisted Laser Desorption Ionization–Time of Flight Mass Spectrometry Coupled with ClinProTools

Abstract

The typing of Mycoplasma pneumoniae mainly relies on the detection of nucleic acid, which is limited by the use of a single gene target, complex operation procedures, and a lengthy assay time. Here, matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) coupled to ClinProTools was used to discover MALDI-TOF MS biomarker peaks and to generate a classification model based on a genetic algorithm (GA) to differentiate between type 1 and type 2 M. pneumoniae isolates. Twenty-five M. pneumoniae strains were used to construct an analysis model, and 43 Mycoplasma strains were used for validation. For the GA typing model, the cross-validation values, which reflect the ability of the model to handle variability among the test spectra and the recognition capability value, which reflects the model's ability to correctly identify its component spectra, were all 100%. This model contained 7 biomarker peaks (m/z 3,318.8, 3,215.0, 5,091.8, 5,766.8, 6,337.1, 6,431.1, and 6,979.9) used to correctly identify 31 type 1 and 7 type 2 M. pneumoniae isolates from 43 Mycoplasma strains with a sensitivity and specificity of 100%. The strain distribution map and principle component analysis based on the GA classification model also clearly showed that the type 1 and type 2 M. pneumoniae isolates can be divided into two categories based on their peptide mass fingerprints. With the obvious advantages of being rapid, highly accurate, and highly sensitive and having a low cost and high throughput, MALDI-TOF MS ClinProTools is a powerful and reliable tool for M. pneumoniae typing.

INTRODUCTION

Mycoplasma pneumoniae is one of the most common pathogens that cause respiratory tract infections (1). The genotyping of clinical isolates is an important means for understanding the epidemiology of M. pneumoniae outbreaks. The 170-kDa protein encoded by the p1 gene is an important adhesion and antigenic factor in M. pneumoniae and is densely clustered at its terminal structure (2–4). The p1 gene contains two previously described repetitive regions, one located within the 3′ region (RepMP2/3) and another located within the 5′ region (RepMP4). RepMP2/3 and RepMP4 elements are present in the M. pneumoniae genome (5). M. pneumoniae clinical isolates can be categorized as type 1 or type 2 according to the sequence variation of the p1 gene (6–10). At present, among the techniques for laboratory typing to gain understanding of the epidemiology of M. pneumoniae, the most popular is molecular typing based on the p1 gene (7, 9). All M. pneumoniae isolates are classified as type 1 or type 2 according to the RepMP4 and RepMP2/3 repetitive sequences within the p1 gene. However, the genotyping of M. pneumoniae isolates based on a single gene limits our understanding of the biological characteristics of M. pneumoniae. Thus, more rapid and convenient methods for the identification and typing of M. pneumoniae are needed to perfect and supplement the present techniques. Silver nanorod array surface-enhanced Raman spectroscopy was use to detect and differentiate M. pneumoniae isolates with 95% to 100% specificity and 94% to 100% sensitivity (11). The peptide mass-fingerprinting technique in the mass range of 2 to 20 kDa, based on matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS), that has emerged in recent years for identifying and typing pathogenic microorganisms is broadly accepted as a new diagnostic gold standard for the identification of many microbial species, and it has the potential to replace conventional identification techniques (12–17). MALDI-TOF MS coupled with ClinProTools software (Bruker Daltonics) is an integrated set of tools provided by Bruker Daltonics for the preparation, measurement, and visualization of peptide and protein biomarkers (18–21). In the present study, MALDI-TOF MS ClinProTools was used for the first time to type M. pneumoniae strains.

MATERIALS AND METHODS

Strain selection and identification.

A total of 68 Mycoplasma clinical isolates, including isolates of M. pneumoniae (n = 63; 9 ATCC strains, 54 clinical isolates), M. genitalium (n = 1), M. pirum (n = 1), M. penetrans (n = 1), M. hominis (n = 1), and M. fermentans (n = 1), were maintained by the Chinese Center for Disease Control and Prevention. Twenty-five M. pneumoniae strains (14 type 1, including 6 ATCC strains, and 11 type 2, including 2 ATCC strains) were used to construct the analysis model, and 43 Mycoplasma strains were used for validation (Table 1). Each strain was isolated in Mycoplasma-selective broth medium (Oxoid) based on color change. After subculture and purification on an agar medium, bacterial genomic DNA from each strain was extracted using the QIAamp DNA minikit (Qiagen). All the pure-cultivated M. pneumoniae clinical isolates were identified by real-time PCR (22, 23).

TABLE 1.

Classification results of Mycoplasma isolates by the GA model and the p1 gene

Species	Strain name	Typing by ClinProTools		Typing by p1 gene (class)
		Classified	Class
M. pneumoniae	105	True	1	1
	12066	True	1	1
	128	True	1	1
	151	True	1	1
	159	True	2	2c
	169	True	1	1
	21009	True	1	1
	21022	True	1	1
	21038	True	1	1
	21065	True	1	1
	21077	True	2	2c
	21101	True	1	1
	21108	True	1	1
	21109	True	2	2c
	21133	True	1	1
	217	True	1	1
	23	True	1	1
	370	True	1	1
	385	True	1	1
	388	True	1	1
	429	True	1	1
	434	True	1	1
	89	True	1	1
	90	True	1	1
	98	True	1	1
	F005	True	1	1
	F022	True	1	1
	F170	True	2	2c
	P005	True	1	1
	P118	True	2	2c
	P146	True	1	1
	P160	True	1	1
	P164	True	1	1
	P37	True	1	1
	P42	True	2	2c
	P54	True	2	2c
	P89	True	1	1
	U14	True	1	1
M. pirum	ATCC 25960		Excluded not recalibratable
M. penetrans	ATCC 55252		Excluded not recalibratable
M. genitalium	ATCC 33530		Excluded not recalibratable
M. hominis	ATCC 23114		Excluded not recalibratable
M. fermentans	ATCC 19989		Excluded not recalibratable

Genotyping of M. pneumoniae.

All 63 M. pneumoniae strains were genotyped by full-length sequencing of the p1 gene with the primers SeqP1-F (5′-ATGCACCAAACCAAAAAAACTGCCT-3′) and SeqP1-R (5′-CTAAGCGGGTTTTTTAGGTGGTTGC-3′) (24).

Sample preparation for MALDI-TOF MS.

Cultures were collected and centrifuged at 12,000 × g at 4°C for 10 min, and the resulting supernatants were discarded. The cell pellets were resuspended in sterile physiological saline and then centrifuged at 12,000 × g at 4°C for 10 min; the resulting supernatants were again discarded. Subsequently, the proteins were preextracted using the ethanol/formic acid method: the cell pellets were suspended in 200 μl of molecular-grade water and vortexed, and 600 μl of anhydrous ethanol (Sigma-Aldrich) was added. The samples were vortexed and centrifuged (13,000 × g) for 2 min. The supernatant was discarded, and 20 μl of 70% formic acid (Sigma-Aldrich) was added and mixed. Finally, 20 μl of acetonitrile (Sigma-Aldrich) was added, and the solution was carefully mixed. After centrifuging (13,000 × g) for 2 min, 1 μl of supernatant protein sample was dropped onto an MSP 96 ground-steel 600-μm sample target (Bruker Daltonics GmbH, Germany) or a FlexiMass 4- by 48-well steel sample target (Shimadzu-Biotech Corp., Kyoto, Japan) and allowed to dry before adding 1 μl of α-cyano-4-hydroxycinnamic acid (CHCA) (saturated matrix solution in 50% acetonitrile and 2.5% trifluoroacetic acid). Two spots were prepared for each sample.

MALDI-TOF MS data acquisition.

In this study, spectra were generated on a Microflex LT (Bruker Daltonics, Bremen, Germany) MALDI-TOF MS system operated in linear mode with a total of 500 laser shots. The Microflex LT was equipped with an N₂ laser (λ = 377 nm). The software used for data acquisition was FlexControl version 3.0 (Bruker Daltonics GmbH, Germany). The parameters used were: mass range, 2,000 to 20,000 Da; ion source 1, 20 kV; ion source 2, 18.5 kV; lens, 8.45 kV; pulsed ion extraction, 320 ns; and laser frequency, 20.0 Hz. Escherichia coli ATCC 8739 was used for mass calibration instrument parameter optimization.

Model construction and validation.

The ClinProTools (version 2.2) software was used for data analysis, which began with a raw data pretreatment, including baseline subtraction (top hat, 10% minimal baseline width), normalization (total ion current), recalibration (1,000 ppm maximal peak shift, 30% match to calibrant peaks, exclusion of spectra that could not be recalibrated), average spectra calculation (resolution, 800), average peak list calculation (signal-to-noise threshold, 5), peak calculation in the individual spectra, and normalization of peak lists. ClinProTools provides a number of highly sophisticated mathematical algorithms that generate models for differentiating between type 1 and type 2 samples. Mass spectra from the two model generation cohorts (14 type 1 strains and 11 type 2 strains) were analyzed using a genetic algorithm (GA) with the following parameters: no more than 60 generations, no more than 10 peaks in each model, a 0.2 mutation rate, a 0.5 crossover rate, no varying random seed, and 3 neighbors. The 43 clinical strains used to validate the optimized model were achieved by using the “classify” function.

Data analysis.

The ClinProTools software calculates cross-validation and recognition capability values, which are indicators of the model's performance and useful predictors of the model's ability to classify test isolates. Peak heights and peak areas are independently calculated in ClinProTools with Welch's t test to determine the statistical separation strength of all the peaks and to then generate an output file. According to the P value of the Anderson-Darling test (PAD), which can provide information about the normal distribution (<1, not normally distributed; >1, normally distributed), the P value of the t test (2 classes) or the analysis of variance (ANOVA) test (PTTA) (>2 classes) (preferable for normally distributed data) or the P value of the Wilcoxon (2 classes) or Kruskal-Wallis test (PWKW) (>2 classes) (preferable for abnormally distributed data) was used to confirm significant differences. If the PWKW or PTTA value was <0.05, the protein/peptide was confirmed to be significantly different. Principle component analysis (PCA) (a built-in feature of ClinProTools software) was also used to evaluate the distinguishing capability between type 1 and type 2 M. pneumoniae isolates based on their peptide mass fingerprints.

RESULTS

M. pneumoniae identification and genotyping.

After subculture and purification, each strain was positively identified by real-time PCR. The genotyping results, which were consistent with the results of the full-length sequence analysis of the p1 gene, indicated that 45 isolates (14 used for constructing the model and 31 used for validation) were type 1 and 18 isolates (11 used for constructing the model and 7 used for validation) were type 2.

Peptide profiling.

Approximately 200 peaks with signal-to-noise ratios of >3.0 were detected between 2,000 and 20,000 Da. The stack view and pseudo-gel view of 68 mass spectra used in this study are shown in Fig. 1. The spectra of M. pneumoniae show perfect whole correlation of the groups, providing a good foundation for our further research and analysis.

FIG 1.

Stack and pseudo-gel views of mass spectra used in this study. (A) Stack view of 68 mass spectra. Blue, red, and green denote Mycoplasma (non-Mycoplasma pneumoniae), type 1 M. pneumoniae, and type 2 M. pneumoniae, respectively. (B) Pseudo-gel view of the mass spectra, with all individual spectra shown in a density scale. There are obvious differences between M. pneumoniae spectra and Mycoplasma (non-Mycoplasma pneumoniae) spectra.

Classification ability of ClinProTools.

A GA model was constructed to distinguish type 1 and type 2 M. pneumoniae. The two parameters of cross-validation and recognition capability were calculated by ClinProTools. For the GA typing model, the cross-validation values, which reflect the model's ability to handle variability among test spectra, and the recognition capability value, which reflects the model's ability to correctly identify its component spectra, were all 100%. This model correctly identified 31 type 1 and 7 type 2 M. pneumoniae isolates from 43 Mycoplasma strains with a sensitivity and specificity of 100% (Table 1). The strain distribution map and PCA based on the GA classification model also clearly show that type 1 and type 2 M. pneumoniae isolates can be divided into one of two categories based on their peptide mass fingerprints (Fig. 2).

FIG 2.

Strain distribution maps corresponding to type 1 and type 2 Mycoplasma pneumoniae. (A) ClinProTools 2D peak distribution diagrams; (B) strain distribution in the principle component analysis. Red and green denote type 1 and type 2 M. pneumoniae, respectively.

Analysis of discriminating peaks.

ClinProTools provided the output file for the statistical separation strength of the peaks. The PAD values of all the peaks were <1; therefore, the data were abnormally distributed, and the PWKW was used to confirm significant differences. The output file contained 38 discriminating mass peaks with statistical significance (P < 0.05) in the range of 2,000 to 20,000 Da (Table 2). In ClinProTools, the peaks with high separation power by Welch's t test are used to generate a biomarker pattern model. Here, we used GA to determine the peaks with the highest separation power and to generate a typing biomarker model. ClinProTools provided a model constructed with 7 peaks that were included in the 38 discriminating peaks between type 1 and type 2 M. pneumoniae (Table 2), including m/z 3,318.8, 3,215.0, 5,091.8, 5,766.8, 6,337.1, 6,431.1, and 6,979.9. These 7 peaks (Fig. 3, red lines) displayed significant differences in peak areas and peak height between type 1 and type 2 M. pneumoniae.

TABLE 2.

Thirty-eight discriminating m/z peaks between type 1 and type 2 Mycoplasma pneumoniae isolates

m/z	P
4,908.3	0.00000327
6,490.6	0.00000327
9,813.7	0.00000327
6,431.1a	0.00000327
9,837.7	0.00000327
5,091.8a	0.00000327
3,245.0	0.00000327
3,215.0a	0.00000327
5,352.5	0.00000327
5,766.8a	0.00000327
6,337.1a	0.00000327
6,979.9a	0.00000327
9,540.0	0.00000327
6,578.7	0.00000327
4,764.1	0.00000538
5,779.2	0.00000538
6,967.8	0.00000538
8,716.6	0.0000102
4,082.8	0.0000289
5,006.8	0.000211
4,868.8	0.000211
5,377.0	0.000277
4,324.0	0.000277
2,989.1	0.000955
5,584.7	0.00162
8,971.1	0.0019
5,392.7	0.0019
3,318.8a	0.0019
6,597.2	0.0023
8,752.4	0.0023
4,886.1	0.00287
4,486.3	0.00709
4,023.4	0.0083
5,645.8	0.0083
4,799.4	0.00968
3,289.9	0.00968
5,686.5	0.0354
2,190.8	0.0481

^aPeaks were selected by Welch's t test and were used to generate a biomarker pattern model.

FIG 3.

Seven peaks (m/z 3,318.8, 3,215.0, 5,091.8, 5,766.8, 6,337.1, 6,431.1, and 6,979.9) were determined by the GA to provide the highest separation power to generate a classification model. Settings for the GA: maximal number of peaks in model, 10; maximal number of generations, 60; and number of neighbors, 3. Red and green denote type 1 and type 2 Mycoplasma pneumoniae, respectively.

DISCUSSION

Currently, conventional methods for nucleic acid detection, such as PCR and restriction fragment length polymorphism (RFLP), are still used for the identification and typing of M. pneumoniae (7, 9, 25–27). Conventional identification and typing methods for M. pneumoniae have several shortcomings, including the detection of single targets resulting in less information, long detection times, complex procedures, and high cost. Rapid and accurate typing methods are essential for appropriate therapeutic management and timely intervention for infection control. Recently, a study using MALDI-TOF MS to identify and type Mycoplasma genus members found that 96% of Mycoplasma strains can be identified at the species level. For typing purposes, MALDI-TOF MS was shown to cluster M. bovis and M. agalactiae isolates by their year of isolation and genome profile, respectively, and M. pneumoniae isolates by their adhesion p1 type using the MALDI-TOF Biotyper software (28). Here, a novel strategy for typing type 1 and type 2 M. pneumoniae strains using the MALDI-TOF MS ClinProTools system was developed. Our study shows that the typing results using the GA model constructed by ClinProTools were completely consistent with conventional p1 genotyping and that the sensitivities and specificities of the GA model were consistently 100%. Biotyper and ClinProTools are products of the Bruker company, and the two sets of software are usually installed in the same mass spectrometer, such as the Microflex, Microflex LT, and Autoflex. After acquiring spectra from the test strains, a reference spectrum should be constructed using these spectra and added to the Biotyper database. Subsequently, the MSP dendrogram acquired using the MSP function was analyzed to determine the type of test strain (14). Using the ClinProTools system, the original spectrum of the test strain can be directly analyzed by the classification model, and the typing results are provided immediately without requiring technician interpretation. Therefore, using ClinProTools as a typing method is more rapid and intuitive.

ClinProTools offers the ability to generate classification models from large numbers of spectra in a relatively rapid and flexible manner. The aim of model generation is to determine a common signature among the spectra of each of the model generation classes such that the spectra from test isolates can be classified by the model (20, 29). In this study, ClinProTools automatically selected the seven peaks with the highest separation power (m/z 3,318.8, 3,215.0, 5,091.8, 5,766.8, 6,337.1, 6,431.1, and 6,979.9) to build a classification model. This model resulted in a sensitivity of 100% and a specificity of 100% for M. pneumoniae typing. Among the seven peaks, 3,318.8 and 6,979.9 are obviously upregulated in type 1 M. pneumoniae, and 3,215.0, 5,091.8, 5,766.8, 6,337.1, and 6,431.1 are upregulated in type 2 M. pneumoniae (Fig. 3). These findings indicate that the seven corresponding proteins have the power to distinguish between type 1 and type 2 M. pneumoniae. Therefore, the differentially expressed proteins of type 1 and type 2 M. pneumoniae can be analyzed in depth and serve as a series of marker proteins for the manufacture of a rapid typing detection kit. Furthermore, these differential proteins may contribute to our understanding of the mechanism of M. pneumoniae typing. Unfortunately, due to their low abundance and limits in instrumentation, these seven peptides were not identified.

There are five p1 gene variants, V2a to V2d and V1 (24, 30–34). In this study, only V2a and V2c isolates were detected. Based on our results, the peptide mass fingerprinting showed no significant differences among V2a, V2c, and the traditional type 2 strain. The genome sequences of the type 1 strain and the V2a, V2c, and traditional type 2 strains were analyzed (data not shown). We found that there were many insertions or deletions of single nucleotide polymorphisms (SNPs), simple sequence repeats (SSRs), small nucleotide fragments, and reported insertions or deletions of large fragment genes, including MPN130, MPN137/138 (35, 36), MPN457 to MPN459 (37), and MPN586 (38). With respect to these genes, all type 2 strains, including the traditional type 2 strains and the variants V2a and V2c, showed the same genome sequence. Sequence differences were observed only between the type 1 and type 2 strains, in accordance with the results of MALDI-TOF MS.

The conventional PCR-based method for M. pneumoniae typing involves DNA extraction, which requires large volumes of culture and a 3- to 10-h procedure. The MALDI-TOF MS ClinProTools-based typing can be completed in 30 min, albeit using a lower inoculum that is more rapidly obtained in culture. This technique required only 3 h after the culture was finished, and 96 isolates were tested simultaneously. Compared with conventional nucleic acid detection technologies, which require several hours for one sample and cost >10 times higher, the MALDI-TOF MS ClinProTools method developed in this study is superior.

This study developed a novel strategy for M. pneumoniae typing. However, this method has some limitations. First, the method is based on culture rather than on the direct detection of a clinical sample, such as a throat swab. In addition, the seven peaks with the highest separation power were not identified; therefore, the proteins that can be used to distinguish type 1 and type 2 M. pneumoniae remain unknown. Current methods for M. pneumoniae typing are based on pure culture. Therefore, with the obvious advantages of being rapid, highly accurate, and highly sensitive and having a low cost and high throughput, MALDI-TOF MS ClinProTools is a powerful and reliable tool for M. pneumoniae typing.