The QIAamp DNA Mini Kit and glass bead-based kit used for DNA extraction were purchased from Qiagen (Hilden, Germany) and CapitalBio Technology Co., Ltd. (Beijing, China). The TwistAmp Basic Kit used for isothermal amplification was purchased from TwistDx (Cambridge, United Kingdom). Graphene oxide (2 mg mL⁻¹) was purchased from Sigma Aldrich (MO, United States), and the CRISPR–Cas enzyme LbCas12a was obtained from GenScript Biotechnology Co., Ltd. (Nanjing, China). Primers, gRNAs, and ssDNA reporter molecules were synthesized by Qingke Biotechnology Co., Ltd. (Beijing, China). Lateral flow biosensors (LFB) were manufactured by HuiDeXing Biotechnology Co., Ltd. (Tianjin, China). DNA concentration was determined using the Nanodrop 2000 instrument (ThermoFisher Scientific, United Kingdom). Isothermal amplification and quantitative fluorescence PCR were performed using Eppendorf AG pro S Mastercycler (Eppendorf, Germany) and the AriaMx Real-Time PCR system (Agilent Technologies, CA, United States), respectively. The gel was imaged with the Gel Doc XR + Imaging System (Bio-Rad, CA, United States).

Two sets of RPA primers targeting IS6110 and IS1081 were designed with Primer Premier 5.0 (Table 1). A BLAST analysis of the GenBank nucleotide database was performed to confirm the specificity of the primers. The gRNAs (Table 1; Figure 1B) were designed against the 20 nt sequence following the protospacer-adjacent motif (PAM) sequences [5’-(T) TTN-3′ on the sense strand or 5′-NAA(A)-3′ on the antisense strand] that served as Cas12a recognition sites. We used two ssDNAs (a fluorescent reporter and a biotin-labeled reporter) (Table 1).

The MTBC reference strain H37Rv was utilized during the establishment and clinical application of the MCMD technique. Mycobacterium abscessus was used as a negative control (NC). A total of 27 bacterial strains were used to determine the analytical specificity of MCMD (see Section 2.8). Genomic DNA was extracted using the QIAamp DNA Mini Kit (bacteria strains) or glass bead-based kit (sputum specimens) according to the manufacturer’s instructions and stored at −80°C before use. DNA quantity and purity were determined using ultraviolet spectrophotometry at 260 and 280 nm.

The Singlex RPA assay for IS6110 (or IS1081) was performed according to the manual (TwistAmp Basic Kit). A 50 μL reaction mixture was prepared as follows: dry reaction pellet, 29.5 μL rehydration buffer, 0.48 µM forward and reverse primers, 1 ng μL⁻¹ target template 2 μL (5 μL for sputum specimen), and 14 mM magnesium acetate. It has been reported that RPA is an error-prone reaction and usually yields non-specific amplicons and primer dimers (Munawar, 2022). Different concentrations of GO (0, 2, 4, 8, 16, 24, 32 μg mL⁻¹) were added into the RPA mixtures at 38°C for 30 min to overcome the technical disadvantages of RPA. The reaction mixtures were incubated at 35.4°C–39.9°C (with 0.5°C–1.0°C intervals) for 20–40 min (with 10 min intervals) to determine the optimal reaction temperatures and amplification times for the RPA assay. Finally, the optimized operating conditions were used in a subsequent GO-assisted multiplex RPA assay, from which two sets of RPA primer (IS6110 and IS1081) were added in one reaction system at a total concentration of 0.48 µM (1:1). Mycobacterium abscessus and double-distilled water (DDW) were used as NC and blank control (BC), respectively. Agarose gel electrophoresis (2.5%) was used to confirm the amplification of the RPA assay.

The CRISPR–Cas12a-based trans-cleavage assay was adopted from previous studies (Chen et al., 2018; Li et al., 2018; Broughton et al., 2020) with some adjustments. Briefly, 41.7 nM of IS6110-gRNA and IS1081-gRNA (1:1) were preincubated with 33.3 nM LbCas12a in 1× NEBuffer 2.1 (NEB, MA, United States) at 37°C for 10 min to induce Cas12a-gRNA complexes that can be used immediately or stored at 4°C for up to 24 h. Subsequently, the trans-cleavage assay was conducted in a volume of 50 μL mixture, containing 25 μL 2 × NEBuffer 2.1, 13.5 μL Cas12a-gRNA complexes, 250 nM fluorescent reporter molecule, and 2 μL of the RPA products. This trans-cleavage assay was performed at 37°C, and the fluorescence signal was monitored for 20 min. The fluorescence signal increasing more than two-fold compared to BC at 10 min of the trans-cleavage assay was considered positive. A biotin-labeled reporter molecule was used instead of the fluorescent reporter for the LFB readout. After 8 min incubation at 37°C, the LFB assay was performed (see Section 2.6).

The LFB used for visual readout consists of several main components: sample pad, conjugate pad, nitrocellulose (NC) membrane (reaction region), and absorbent pad mounted on a backing card (Figure 2A). The streptavidin-gold nanoparticles (SA-GNPs) are adhered onto the conjugate pad and used as the indicator reagent, rabbit anti-fluorescein amide (anti-FAM) antibody and biotin-bovine serum albumin (biotin-BSA) are fixed onto the NC membrane and used as control line (CL) and test line (TL) capture reagents, respectively. Twelve microliters of trans-cleavage products and two drops of running buffer (100 mM phosphate-buffered solution, pH 7.4 with 1% Tween 20) were added to the sample pad. The detection results were visualized within 3 min as red bands on the NC membrane.

The limit of detection (LoD) was determined using 10-fold serial dilutions of genomic DNA from reference strain H37Rv, ranging from 40,000 to 0.4 copies μL⁻¹ to assess the analytical sensitivity of CRISPR–Cas12a-based singlex and multiplex detection for MTBC identification. DNA copy numbers per microliter were calculated using the following formula: (6.02 × 10²³) × (ng μL⁻¹ ×10⁻⁹)/(DNA length × 660). A volume of 2 µL from each DNA dilution was added to the RPA reaction mixture. Each dilution series was performed in triplicate. The lowest positive dilution (twice the fluorescence value of BC) in all replicates was considered the LoD.

Reactions were conducted with genomic templates extracted from different bacterial strains, including 1 MTBC reference strain H37Rv, 1 Mycobacterium bovis Bacilli Calmette-Guerin (BCG), 8 clinical MTBC strains isolated from TB patients, 10 non-tuberculous mycobacteria (NTM) strains, and 7 non-mycobacteria strains to determine the analytical specificity of MCMD (Supplementary Table S1). Each strain was tested at least twice.

This study was approved by the Ethical Committee of Beijing Children’s Hospital, Capital Medical University (2020-k-163). A total of 107 patients with suspected active pulmonary TB from Beijing Chest Hospital were enrolled in this study from May to June 2022. The need for informed consent was waived because the sputum specimens used in this study were leftover samples from clinical microbiology tests. According to the Chinese National Standard on Diagnosis for Pulmonary Tuberculosis (WS288-2017, National Health and Family Planning Commission of the People’s Republic of China, 2017), the patients were categorized according to the composite reference standard (CRS), which combines the clinical and laboratory diagnostic criteria: (a) definite TB/laboratory-confirmed TB, patients with bacteriological confirmation of MTB (culture, smear, nucleic acid detection, or histopathological evidence positive); (b) probable TB/clinical diagnosed TB, patients with radiologic findings suggestive of TB plus at least one of the following: TB clinical symptoms or signs, positive tuberculin skin test (TST) or IGRA, bronchoscopy or histopathology consistent with TB; and (c) non-TB, patients diagnosed as other diseases and improved in the absence of anti-TB treatment.

Sputum samples (2–3 mL) collected from patients were utilized for TB culture, Xpert, and MCMD assays. A tuberculosis culture was performed using Bactec MGIT 960 system (Becton Dickinson, MD, United States). Xpert (Cepheid, CA, United States) was performed following the manufacturer’s protocol. For MCMD, sputum samples were decontaminated and liquefied by adding an equal volume of 4% NaOH, and sputum DNA was extracted using the glass bead-based kit. Subsequently, 5 µL DNA solution was added to the RPA reaction mixture. All samples were detected in duplicate in multiple independent batches, and each batch included a positive control (PC) (H37Rv DNA as the template) and BC. The results from MCMD were compared to TB culture and Xpert to evaluate the diagnostic performance of this new technique.

The MCMD integrated GO-assisted multiplex RPA assay that simultaneously targets IS6110 and IS1081 with a CRISPR–Cas12a-based trans-cleavage assay at fixed temperatures for MTBC nucleic acid detection. First, the extracted MTBC DNA templates are pre-amplified with three proteases [recombinase, single strand DNA binding protein (SSB), and DNA polymerase] at 37–42°C within 30 min (Figure 1A). Next, in the trans-cleavage stage, the PAM site in the RPA amplicon can guide the CRISPR–Cas12a-IS6110(or IS1081)-gRNA complex to its location (Figures 1B, C), activating the CRISPR–Cas12a effector. The activated Cas12a has trans-cleavage activity against ssDNA reporter. Finally, the digestion of reporter molecules can be detected via fluorescence or LFB, which confirms the presence of the target genome MTBC. The entire MCMD assay, including rapid DNA extraction (15 min, STEP 1), GO-assisted multiplex RPA reaction (30 min, STEP 2), CRISPR–Cas12a-based trans-cleavage assay (8–10 min, STEP 3), and result readout (immediately in fluorescence and 3 min in LFB, STEP 4), can be completed within 60 min.

For fluorescence readout, the fluorescent reporter, labeled at the 5′-end with a FAM fluorophore (6-carboxyfluorescein) and at the 3′-end with a black hole quencher (BHQ1), is cleaved by the activated Cas12a and released from its quencher, increasing fluorescent signaling (Figure 1C STEP 3, STEP 4). Thus, CRISPR–Cas12a-based trans-cleavage and fluorescence readout can be performed simultaneously.

For the LFB readout, when the cleaved products are added to the sample pad followed by the running buffer, the running buffer travels along the biosensor through capillary action, rehydrating the indicator reagent (SA-GNPs) in the conjugate pad (Figure 2B). In the negative sample, the uncleaved biotin-labeled reporter molecule (5′-Biotin-TATTATTATTATTT-FAM-3′) binds SA-GNP (via biotin at the 5′ end of the reporter), and it is captured by anti-FAM immobilized on the CL (via FAM at the 3′ end of the reporter), then CL is displayed in red for visualization. In the positive sample, the biotin-labeled reporter cleaved by activated Cas12a binds SA-GNP. It is then captured by biotin-BSA immobilized on the TL, and then TL is visualized. Therefore, the principle of visualization in the LFB assay is that the biotin–SA-GNPs complexes seized by capture reagents (anti-FAM on CL or biotin-BSA on TL) develop red bands (Figure 2C).

According to our previous study, different concentrations of GO ranging from 4 μg mL⁻¹–32 μg mL⁻¹ were added to the GO-assisted singlex RPA assays to improve the specificity of RPA (Wang Y et al., 2020). As shown in Supplementary Figure S1A and S1B, GO concentrations ranging from 8 μg mL⁻¹–16 μg mL⁻¹ exhibited good specificity performance with no obvious decrease in RPA product yield. Concentrations less than 4 μg mL⁻¹ showed no improvement in amplification specificity, while concentrations above 24 μg mL⁻¹ partially inhibited the amplification reactions. Therefore, we recommend 8 μg mL⁻¹ as the optimal concentration for GO-assisted RPA assays.

According to Supplementary Figures S1C, S1D, the optimal reaction temperatures for IS6110-RPA and IS1081-RPA primers were 37.4°C–39.4°C and 37.9°C–38.9°C, respectively. Thus, an ideal temperature range of 37.9°C–38.9°C was recommended for the subsequent GO-assisted multiplex RPA assays, using both IS6110 and IS1081 RPA primers in one reaction. As can be seen form Supplementary Figure S1E, RPA products increased with the duration of amplification. Considering both a shorter test time and a higher amplification yield, the optimal amplification time was recommended to be 30 min. Figure 2D shows that specific amplification products were slightly reduced in the GO-assisted multiplex RPA assay compared to the two GO-assisted singlex RPA assays. This occurrence is presumed to be due to interference among the primers, particularly when multiple pairs of RPA primers are present. Whether this could lead to an analytical loss of sensitivity requires further experiments (see Section 3.5).

MCMD was performed in parallel with two singlex detections (MCSD-IS6110 and MCSD-IS1081). The high fluorescent signals in Figure 2E indicated that these assays could identify MTBC by targeting IS6110 (or IS1081) or both. Fluorescence signals from MTBC CRISPR–Cas12a detections on various RPA products were detectable in 1 min, increased rapidly within 8 min, and decreased slowly after 12 min (Figure 2E). Finally, an optimized reaction time of 8–12 min was recommended for the trans-cleavage stage.

Serial dilutions of H37Rv genomic DNA (40,000 to 0.4 copies μL⁻¹) were used for analytical sensitivity determination of MCSD and MCMD. Electrophoresis and fluorescence results showed that two singlex detections (MCSD) and the multiplex detection (MCMD) were all sensitive with the LoDs of 4 copies μL⁻¹ (Figures 3A–F). Figure 3G showed that the MCMD LFB assay also had a similar LoD value. Furthermore, the electrophoretic band brightness shown in Figures 3A, C, E indicates a dependency on the target amount in the RPA assay. However, such typical trend was not observed in the fluorescence intensity shown in Figures 3D, F. Two reasons may explain this result. First, this may be attributed to the differences in the targets of RPA assay and fluorescence detection. The RPA assay targeted serial dilutions of H37Rv DNA, whereas the fluorescence detection targeted the RPA products, which were not serial dilutions of DNA. Second, the cycle threshold (Ct) value (rather than the fluorescence intensity) is an inverse measure of the template load. The trans-cleavage reactions initiate almost instantly after the mixing of the reaction components at room temperature, resulting in the inability to read Ct values of RPA products in the resulting figures. While the curve height in Figures 3B, D, F represents the fluorescence intensity, it does not directly correspond to the concentrations of the templates. Altogether, MCMD that can juggle multiplex and unimpaired sensitivity was recommended for the subsequent experiments, including analytical specificity testing and clinical application.

Genomic DNA extracted from 27 bacterial strains shown in Supplementary Table S1 was used to validate the analytical specificity of MCMD. As shown in Supplementary Figure S2, MTBC strains, including H37Rv strain (as a PC), BCG strain, and clinical strains from TB patients, gave positive results. In contrast, NTM strains, non-mycobacteria strains, and DDW (as a BC) showed negative results, demonstrating the absence of cross-reactions. Only MTBC genomic DNA could be detected using MCMD, indicating the extremely high specificity (100%) of this method.

A total of 107 pulmonary TB patients and 40 non-TB patients were included in this study. The proportion of males in the TB group was 64.5% (69/107), and the mean age was 50.84 years. The TB group comprised 79 laboratory-confirmed and 28 clinically diagnosed TB patients. In the non-TB group, 57.5% (23/40) were males, and the average age was 68.30. The diagnostic specificity of MCMD was evaluated in these non-TB patients, which consisted of 30 patients with bacterial pneumonia, seven with non-infectious inflammatory diseases, and three with malignancies. All samples were sputum specimens.

As shown in Table 2, 107 TB patients were classified into two groups, the culture-positive group (67/107) and the culture-negative group (40/107). For the culture-positive samples, the diagnostic sensitivity of Xpert and MCMD was 85.1% (57/67) and 95.5% (64/67), respectively. For the culture-negative samples, the diagnostic sensitivity of Xpert and MCMD was 27.5% (11/40) and 40.0% (16/40), respectively. Among all TB patients in this study, the pooled diagnostic sensitivity of culture, Xpert, and MCMD was 62.6% (67/107), 63.6% (68/107), and 74.8% (80/107), respectively. These results indicate that MCMD exhibits higher sensitivity compared to culture and Xpert, making it a more promising diagnostic tool for TB. In addition, the diagnostic specificity was 100% (40/40) for Xpert and MCMD when testing samples from non-TB patients.