Rapid Sputum Multiplex Detection of the M. tuberculosis Complex (MTBC) and Resistance Mutations for Eight Antibiotics by Nucleotide MALDI-TOF MS

The increasing incidence of multidrug-resistant (MDR) and extensively drug-resistant (XDR) Mycobacterium tuberculosis (MTB) adds further urgency for rapid and multiplex molecular testing to identify the MTB complex and drug susceptibility directly from sputum for disease control. A nucleotide matrix-assisted-laser-desorption-ionization time-of-flight mass spectrometry (MALDI-TOF MS)-based assay was developed to identify MTB (MTBID panel) and 45 chromosomal mutations for resistance to eight antibiotics (MTBDR panel). We conducted a 300 case trial from outpatients to evaluate this platform. An MTBID panel specifically identified MTB with as few as 10 chromosome DNA copies. The panel was 100% consistent with an acid-fast stain and culture for MTB, nontuberculous mycobacteria, and non-mycobacteria bacteria. The MTBDR panel was validated using 20 known MDR-MTB isolates. In a 64-case double-blind clinical isolates test, the sensitivity and specificity were 83% and 100%, respectively. In a 300-case raw sputum trial, the MTB identification sensitivity in smear-negative cases using MALDI-TOF MS was better than the COBAS assay (61.9% vs. 46.6%). Importantly, the failure rate of MALDI-TOF MS was better than COBAS (11.3% vs. 26.3%). To the best of our knowledge, the test described herein is the only multiplex test that predicts resistance for up to eight antibiotics with both sensitivity and flexibility.

In clinical laboratories, the rapid identification of MTBC has mainly relied on acid-fast staining, which features a low recovery rate. For example, only 44% of new cases, including 15-20% in children, can be identified by the presence of acid-fast bacilli in sputum smears 3 . Therefore, many commercial or in-house molecular diagnostic methods have been developed to identify MTB. Among these assays, the COBAS TaqMan MTB test (Roche Diagnostics, Branchburg, NJ) is one of the most widely used tests in clinical diagnostics. Another promising diagnostic is GeneXpert MTB/RIF (Cepheid, Sunnyvale, CA), which can simultaneously detect MTB and rifampin resistance directly from sputum within 2 hours 4 . Both assays not only show good sensitivity and specificity but also exhibit certain improvements in the testing procedure, including convenience and turnaround time 4-6 . However, due to the growing incidence of TB cases caused by antibiotic-resistant MTBC, a novel strategy capable of detecting multiple-resistance mutations for different anti-TB drugs directly from raw specimens is necessary.
Previously, we developed a nucleotide matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) assay consisting of 25-multiplex probes to detect 50 types of four cancer-driving mutations [7][8][9] . Thus far, this approach has provided routine molecular diagnostics for more than 8000 cases under an ISO15189-certified central laboratory 10 . Here, we extend the MALDI-TOF MS application to tuberculosis diagnoses (see details in Methods) and successfully established a MALDI-TOF-based molecular assay for identifying MTB and resistance mutations for eight anti-TB drugs. In a parallel test using clinical patient sputum specimens, this assay showed better sensitivity and specificity than the COBAS TaqMan MTB test. Thus, the MALDI-TOF-based MTB assay is a promising new strategy for the rapid, simultaneous identification of MTB and drug resistance from sputum.

Identifying MTBC and Non-Tuberculosis Mycobacteria (NTM) Using MALDI-TOF MS.
The TB-specific mce3B gene or NTM common gyrA gene was amplified using PCR specific primers (Supplementary Table 1) followed by a single nucleotide extension reaction using "TB"-or "NTM"-specific probes (Fig. 1A,B and Supplementary Table 2). Probes without a single nucleotide incorporated are referred to as unextended probes (UEPs) and did not produce a mass-shifted signal. Due to the significant differences in the probe masses, the two probes can be detected in a single reaction (Reaction #1 in Supplementary Table 2) (Fig. 1C). To evaluate the specificity of the platform, genomic DNA from several reference strains, including M. tuberculosis H37Ra, 7 strains of Mycobacterium spp. and 5 non-Mycobacterium bacterial stains, were assayed in the pilot study (Supplementary Table 3). As shown, only the M. tuberculosis genomic DNA spectrum showed shifted signals for both the TB and NTM probes (Fig. 1C, lower left panel). In contrast, the Mycobacterium fortuitum and other NTM genomic DNA spectra exhibited shifted signals for the NTM probe only (Fig. 1C, upper right panel and Supplementary Fig. 1A). Moreover, the K. pneumoniae and other non-mycobacterium bacterial genomic DNA spectrum did not exhibit shifted signals for either the TB probe or the NTM probe, which indicates that the assay is highly specific for mycobacteria (Fig. 1C, lower right panel and Supplementary Fig. 1B). Taken together, MALDI-TOF MS can specifically distinguish M. tuberculosis, non-tuberculosis mycobacteria and non-mycobacterium bacteria in multiplex reactions.

Limit of Detection Test for MALDI-TOF MS.
According to the results of serially diluted MTB genomic DNA testing, as shown in Fig. 2A, the TB probe's shifted signal can be detected with as low as 5 MTB genomic DNA copies. The signal intensity correlates highly with the MTB genomic DNA copies, as demonstrated by the linear correlation value (R squared) at 0.902 (Fig. 2B).
Comparing Results from Acid-fast Staining, Bacterial Culture, and MALDI-TOF MS. We further performed a double-blind test with clinical isolates cultured from sputum specimens using an MTBID panel, acid-fast staining and traditional bacterial culture to identify MTB and NTM ( Table 1). The results indicate that MALDI-TOF MS can detect the culture-positive MTBC cases of both TB and NTM with 100% consistency. However, acid-fast staining failed to identify 3 of 5 and 3 of 6 culture-confirmed TB and NTM cases, respectively.

Detecting Drug Resistant Mutations Using MALDI-TOF MS.
In addition to MTB identification, we also designed an MTBDR panel with specific probes for detecting mutations in several resistance-associated genes, including rpoB (codon 513, 516, 522, 526, 531 and 533), katG (codon 315), and inhA (promoter -15 nucleotide) ( Table 2). To evaluate the feasibility and accuracy of the MTBDR panel, twenty MDR-MTB strains with known antibiotic-resistance mutations were obtained from the Taiwan CDC and assessed using the MALDI-TOF MS platform and Sanger sequencing in a double-blind study. As shown, all mutations detected using the MALDI-TOF MS assay were also identified through Sanger sequencing. The results were 100% consistent.
Furthermore, 64 clinical MTB isolates with known antibiotic susceptibility, including seven XDR-MTB, five MDR-MTB, 8 fluoroquinolone-resistant MTB, 15 isoniazid-resistant MTB, 2 rifampin-resistant MTB and 27 all-sensitive MTB, were obtained from the NTUH for the MTBDR panel validation (Fig. 3A). Among 45 probes for 8 first-and second-line anti-MTB drugs, no mutations were detected among the 27 all-sensitive isolates. MALDI-TOF MS had 90.6% concordance with the drug susceptibility test. Four isoniazid-resistant isolates (#3, #13, #37, and #42) and 2 ofloxacin-resistant isolates (#10 and #33) did not test positive for any of the selected probes. Moreover, the XDR-MTB clinical isolate (#14) mutation spectra were clearly distinguished from the all-sensitive spectra in the rpoB, katG, embB, and gyrA genes ( Fig. 3B-E (Table 3). Concerning the 11 smear AFB-positive cases, the MALDI-TOF MS vs. COBAS sensitivities and specificities were comparable. In addition to MTB identification, MTBID panel-positive samples from the MALDI-TOF MS assay were further assessed using the MTBDR panel.
No drug-resistance mutations were detected among these samples, which was consistent with the drug susceptibility test. Overall, the COBAS vs. MALDI-TOF MS sensitivities and specificities were 52.6% vs 72.4% and 96.0% vs. 92.4%, respectively (Table 3).

Discussion
The nucleotide MALDI-TOF MS test is a simple, rapid and multiplexed method with high specificity, sensitivity and flexibility in many research fields, including mutation and SNP analysis, sequencing, and microorganism detection [11][12][13][14][15][16][17][18] . Compared with other detection systems, MALDI-TOF MS acquires the absolute mass value, which represents an intrinsic property of a molecule, while others depend on signals of a relative electrophoretic mobility or a hybridization event. In addition, the molecular weight is a significantly more informative signal than fluorescent label-based methods. This study applied this platform to establish a rapid multiplexed test to detect MTB and its drug susceptibility in sputum. When testing for MTB at an early stage, the sensitivity, specificity, positive prediction value (PPV), and negative prediction value (NPV) are comparable with other currently used methods (Table 3). Compared with the limit of traditional testing methods such as the sensitivity of microscopic examination (22-65%) or bacillus culture (approximately 10-20% unsuccessful), MALDI-TOF MS provided a solution for disease control and management 15,19,20 . Furthermore, the rapid diagnosis of MTB infections is critical for successful treatment. However, due to the limit of detection, the low sensitivity impedes rapid laboratory MTB diagnosis in patients. Many PCR-based molecular diagnostic methods have been developed in recent years. However, the sensitivity of the assays remains unsatisfactory, particularly for acid-fast stain negative samples in which only 43-74% of samples were identified 21,22 . Our results show that the MALDI-TOF MS detection limit is less than 10 MTB copies (Fig. 2). This characteristic facilitates MTB detection in raw specimens, including sputum, and significantly reduces the turnaround time for diagnosis.
Another advantage of MALDI-TOF MS is the panel flexibility for drug-resistant MTB prediction based on customized requirements, which are widely considered serious threats to global TB control. In this panel, we simultaneously predicted resistance for anti-TB drugs, including, rifampin, isoniazid, pyrazinamide, streptomycin, ethambutol, fluoroquinolone, pyrazinamide and ethionamide. This test can identify drug-resistant MTB more comprehensively than the rifampin-resistant MTB determined using GeneXpert MTB/RIF. Among 64 clinical isolates, all sensitive or drug-resistant strains could be identified using our system (Fig. 3), which could provide a rapid diagnosis compared with the traditional six-to eight-week bacterial culture process. Because there was no mutation detected in six resistant isolates (Fig. 3, #10, #33, #3, #13, #37, and #42), it was possible that other rare mutations may not be included in our panel, and any other unknown resistance-mechanisms may not be detected [23][24][25] . The limitation of this platform is that MALDI-TOF MS cannot perform de novo mutation identification; however, adding novel drug resistant mutations into the multiplex panel is a convenient process. Any newly identified mutations can be detected by designing a specific probe and coordinating with other probes in the spectrum. Thus, MALDI-TOF MS has a greater capacity compared with other molecular diagnostic methods, such as the INNO-LiPA Rif TB kit (LiPA) (Innogenetics, Zwijndrecht, Belgium) 26 . Furthermore, due to the single nucleotide extension reaction in the biochemical process, this assay can detect various alterations within one nucleotide residue that other molecular methods cannot detect, such as ARMS or COBAS 27 .
The other greatest challenge for MTB identification is early diagnosis and short turnaround time. Regarding the time-consuming culture process, many well-established sensitive systems have been developed to improve detection rates at early stages, such as COBAS and GeneXpert MTB/RIF. With the development of new methods, sputum specimens have been routinely used in clinical diagnoses. One important issue should be considered that may cause false negative results using sputum for MTB identification: the copy number of MTB in the sputum, especially in smear-negative sputum. The overall sensitivity and specificity of COBAS ranged from 66.9% to 82.4% and 97.7% to 100%, respectively 5,6,28 . However, the sensitivity for smear-negative, culture-positive samples was relatively low (34.4-73.6%). For the smear-negative cases in this study, MALDI-TOF MS showed 61.9% sensitivity and 92.4% specificity, which is better than COBAS and comparable to other systems 5,29,30 . Methodologically, despite the concern of multiplex detection, TaqMan QPCR-based COBAS required relatively high quality nucleic acids, not only in large amounts but also of good integrity, for probe annealing. That means samples with highly degraded nucleic acid or reaction inhibitors may result in a poor reaction and an undetermined/invalid result. In addition, the unpredicted nucleotide alteration within the probe binding region may affect detection. This may be the cause of the failure rate of MALDI-TOF MS and COBAS in this study (11.3% vs 26.3%) ( Table 3). The sensitivity of GeneXpert MTB/RIF ranged from 43.4% to 76.9% in smear-negative, culture-positive specimens 4,5,29,31-34 . MALDI-TOF MS was also competitive with this method. However, for clinical utility, using GeneXpert MTB/RIF might be restricted by the limited shelf-life of the cartridges, strict environmental considerations or high-level hardware requirements 35,36 . Thus, MALDI-TOF MS may provide a more friendly platform for TB management than other methodologies.
Finally, the PPV and the NPV are also important for clinical practice. In this study, the PPV of the pooled specimens was unsatisfactory (Table 3) due to a low prevalence rate in the unselected patient population. It should be noted that only 3.6% of samples were smear-positive in this study. Although this is similar to another study that examined a large number of specimens, we think a more comprehensive prospective study is necessary for further clinical validation 37 . MALDI-TOF MS may be suitable for low-prevalence developing countries. Taken together, we developed a MALDI-TOF MS-based highly sensitive molecular testing method that combines MTB identification and drug-resistance predictions. It should be emphasized that the flexibility will allow clinicians to rearrange or modify the detection panels based on clinical needs. Additionally, to the best of our knowledge, this is the only platform that comprehensively features multiplex detection for most MTB drug-resistant mutations, not only for rifampin resistance. Further, the turnaround time was less than 48 hours from sample receipt to data analysis, which is also comparable to other methods 38 . This method provides a powerful and an alternative strategy for MTB disease management and control.

Methods
Study Cases. This study was conducted at the National Taiwan University Hospital (NTUH), which is a 2500-bed medical centre in Taipei, Taiwan. A total of 300 consecutive and non-duplicate sputum samples that were submitted to the Mycobacteriology Laboratory of NTUH for mycobacterial isolation from Dec-2010 to July-2011 were evaluated. All experimental procedures with the patients' specimens and information were approved by the institutional review board of the National Taiwan University Hospital, Taipei, Taiwan   The results of individual specimens that passed the criteria of the negative control, positive control, and internal control were defined as a valid run. MTB was determined to be detected or not detected only in a valid, but not invalid, run.

Mutation Detection by MALDI-TOF MS.
All nucleotide mass spectrometry assays utilized the MassARRAY ® System (Cat. No. 10411, SEQUENIM, San Diego, CA acquired by Agena Bioscience, http://agenabio.com/, San Diego, CA at 2014). It combined mass spectrometry and sensitive and robust chemistry in genetic testing. Briefly, for mutation detection, the desired region of interest was PCR-amplified using specific primers followed by shrimp alkaline phosphatase (SAP) treatment to remove excess dNTP. An iPLEX ® Pro (Cat. 10217, Agena Bioscience, San Diego, CA) single nucleotide extension was performed using the home-designed probe to distinguish the mutant nucleotide from the wild-type due to the mass difference. The iPLEX ® biochemical reaction products were dispensed onto the SpectroCHIP ® Array using the MassARRAY ® Nanodispenser RS1000, a silicon chip with a pre-dispensed matrix crystal. Finally, the chip was placed into the MassARRAY ® Analyzer 4

MALDI-TOF MS and analysed.
Biochemical Reaction for MALDI-TOF MS Analysis. TB identification and antibiotic-resistance mutation detection were performed using MassARRAY ® System and iPLEX ® Pro following the manufacturer's protocol. Briefly, a total volume of 5 μl of a mixture containing 10 ng of bacteria or sputum DNA, 0.5 unit HotStarTaq  Data were collected and analysed using Type4 software (SEQUENOM). The signal in the correct mass position (corresponding to products), which passed the criteria of the signal to noise ratio with an acceptable probability, was interpreted as a positive result by the Type4 software. To confirm that the testing run was valid in sputum, the probe for the human epidermal growth factor receptor (EGFR) gene previously used was spiked into each test as an internal reaction control 8,9 . The "undetermined" result representing MTB detection was uncertain due to the poor reaction or other confounding factors.