A total of 128 MTB clinical isolates were included in this study, comprising both FQ-susceptible and FQ-resistant strains. In addition, five common respiratory bacterial isolates—Haemophilus influenzae, Streptococcus pneumoniae, Staphylococcus aureus, Klebsiella pneumoniae, and Pseudomonas aeruginosa—were used to assess assay specificity. All isolates were obtained from the Chinese Center for Disease Control and Prevention (China CDC), National Institute for Communicable Disease Control and Prevention.

The 128 MTB isolates were collected from FQ resistance surveillance sites across China (specifically from Xinjiang, Hebei, and Hunan provinces) and originated from culture-positive TB patients. Following nucleic acid extraction and whole-genome sequencing, all isolates were confirmed to belong to the MTB complex. Based on genotypic profiling of FQ resistance-associated genes, 50 isolates were classified as FQ-susceptible, and 78 were identified as FQ-resistant.

Additionally, 88 archived clinical specimens from patients with suspected tuberculosis were collected from Hebei Provincial People’s Hospital, including 70 sputum samples and 18 bronchoalveolar lavage fluid (BALF) samples. These specimens were used to evaluate the clinical performance of the POCT-DO-RAP assay. The sample selection criteria and overall study workflow are illustrated in Figure 2.

The gyrA gene sequence (Gene ID: 887105) was retrieved from the NCBI GenBank database and used as the reference for primer and probe design. Oligo 7 software was employed to design candidate primers and probes for the RAA system, with primer lengths set at approximately 30 bp to minimize primer–dimer formation and to ensure an amplicon size of 100–200 bp. The specificity of all designed primers and probes was evaluated using the Primer-BLAST tool available on the NCBI website.

To enhance single-nucleotide discrimination, locked nucleic acid (LNA) modifications were introduced at the mutation sites of the gyrA probes. Primers used for qPCR and nested PCR were designed based on previously published literature (Singhal et al., 2016). All primers and probes were synthesized by Bioligo Biotechnology (Shanghai) Co., Ltd., and their detailed sequences are listed in Table 1.

The target sequence within the gyrA promoter region was cloned into the pUC57 vector to construct a recombinant plasmid. Plasmid synthesis and assembly were performed by Qingke Biotechnology Co., Ltd. The plasmid DNA concentration was quantified using the Qubit dsDNA High-Range Assay Kit together with a Qubit 2.0 fluorometer. The plasmid copy number per microliter was calculated according to the following equation: plasmid concentration (copy number/μL) = [6.02 × 1,023 × concentration (ng/μL) × 10−9]/ [plasmid size × 660]

A 10-fold serial dilution of the recombinant plasmid was then prepared in 1 × TE buffer to generate a standard range from 105 to 10⁰ copies/μL, and aliquots were stored at –80°C for future use.

To simulate bacterial loads similar to those encountered in clinical samples, the reference MTB strain H37Rv was cultured and subsequently subjected to 10-fold serial dilution in 1 × TE buffer, yielding bacterial suspensions ranging from 104 to 10¹ CFU/mL. All diluted bacterial standards were stored at –80°C until use.

The POCT-DO-RAP workflow was established based on our previous study (Han et al., 2025) and optimized using the commercially available real-time PCR based POCT device C10 (TargetingOne Corporation Ltd., Beijing, China). The cartridge was designed based on a magnetic bead–based nucleic acid extraction strategy and contains seven chambers, including lysis buffer, magnetic bead mixture, two wash buffers, waste reservoir, elution buffer, and mineral oil. In addition, one sample chamber and two PCR reaction tubes (WT and MT) are included (as illustrated in the schematic). The extraction procedure comprises four automated steps: lysis, magnetic bead binding, washing, and elution. Lysis and binding times were systematically adjusted under different conditions to achieve optimal nucleic acid recovery.

Following nucleic acid extraction, the eluted DNA is automatically dispensed into the WT and MT reaction tubes. The POCT device C10 includes an integrated mechanical module capable of performing autonomous pipette mixing throughout the extraction process, as well as dedicated temperature-control and magnetic-rack units to ensure efficient lysis, binding, washing, and elution. After extraction is completed, the system transfers the eluted nucleic acid into the corresponding PCR tubes. In the final preparation step of the amplification reaction, the instrument aspirates Mg²⁺ solution from an independent magnesium reservoir and delivers it into each PCR tube, followed by automated sealing with mineral oil. Subsequently, the robotic arm closes the tube caps and transfers the reaction tubes to the PCR amplification module. The optimized thermal cycling protocol consisted of an initial RAA phase at 40°C for 10 min, followed by a PCR phase: pre-denaturation at 95°C for 30 s, and 30 cycles of denaturation at 95°C for 10 s and annealing/extension at 62°C for 30 s.

Because the only difference between WT and MT tubes lies in the probe sequences, optimization experiments were performed solely with the WT tube. To enhance analytical specificity and sensitivity, betaine was included in the reaction mixture to reduce secondary structure formation (Kang et al., 2005), and its optimal concentration was determined using 10 copies/μL plasmid templates. Under the same template concentration, different Mg²⁺ concentrations (8, 9, 10 mM) were evaluated to identify the optimal POCT-DO-RAP reaction conditions.

Quality Control: To prevent false-negative results due to PCR inhibition or extraction failure, the conserved insertion sequence IS1081 was detected in every reaction as an endogenous Internal Control (IC). For external quality assessment, a positive control (H37Rv genomic DNA) and a negative control (sterile water) were tested periodically (e.g., daily or prior to each new reagent batch) to verify system stability, rather than running them simultaneously with every clinical sample due to the limited throughput of the portable device.

To evaluate the analytical sensitivity of the optimized POCT-DO-RAP assay in comparison with conventional qPCR, recombinant plasmids containing either wild-type or mutant gyrA sequences (gyrA90-WT, gyrA94-WT, A90V, and D94G) were tested across a dilution range of 1–105 copies/μL. To assess the overall sensitivity of the integrated POCT workflow—including sample preprocessing and nucleic acid extraction—MTB reference strain H37Rv was spiked into sputum from healthy donors at concentrations of 10–104 CFU/mL to generate simulated clinical samples. A sputum liquefaction step was included prior to DNA extraction. Nuclease-free water was used as the negative control in all experiments. Each simulated sample series contained an internal control, and the assay sensitivity and reproducibility were evaluated by performing eight independent runs on different days. The optimized DO-RAP reaction conditions and workflow are described in Section 3.2.

Conventional qPCR was performed using the PN101 kit (Vazyme, Nanjing, China) in a total reaction volume of 20 μL, consisting of 4 μL of 5 × Taq Pro buffer, 0.2 μL of Taq Pro HS DNA polymerase (5 U/μL), 0.4 μL each of the forward and reverse gyrA primers (10 μM), 0.2 μL each of the gyrA-A90V-p-WT or -p-MT probes (10 μM), 0.2 μL each of the gyrA-D94G-p-WT or -p-MT probes (10 μM), 12.6 μL nuclease-free water, and 2 μL of DNA template. The qPCR cycling conditions were as follows: 95°C for 2 min, followed by 40 cycles of 95°C for 15 s and 52°C for 30 s with fluorescence acquisition.

The analytical specificity of the optimized POCT-DO-RAP assay was assessed using five common respiratory bacterial species (Haemophilus influenzae, Streptococcus pneumoniae, Staphylococcus aureus, Klebsiella pneumoniae, and Pseudomonas aeruginosa) as well as 128 clinical MTB isolates. All MTB isolates were genotypically characterized for FQ susceptibility or resistance by Sanger sequencing at the National Institute for Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention.

To assess the capability of the POCT-DO-RAP assay in detecting heteroresistance, recombinant plasmids containing both wild-type and mutant gyrA sequences (gyrA-D94G and gyrA-A90V) were mixed at defined proportions.

For each reaction, the total plasmid concentration was adjusted to 10² copies/μL.

Mutant and wild-type templates were combined to generate mixtures containing 0, 1, 5, 20, 50, and 80% mutant alleles, and each mixture was tested using the MT tube. This design enabled evaluation of the assay’s ability to identify low-abundance mutant alleles within mixed genetic backgrounds and determine the detection limit for heteroresistance under low template conditions.

A total of 88 clinical specimens were used to evaluate the clinical performance of the POCT-DO-RAP assay for detecting MTB drug resistance. The results obtained by POCT-DO-RAP assay were compared with those from conventional qPCR. The discrepant results were resolved by nested PCR followed by Sanger sequencing.

Nested PCR was performed in two rounds. In the first round, primers gyrA RS-F and gyrA PRR-R were used; in the second round, primers gyrA PRR-F and gyrA RS-R were employed (primer sequences listed in Table 1). Nested PCR was conducted according to the instructions of the PN101 kit. Both rounds of amplification were carried out under identical conditions: initial denaturation at 95°C for 2 min, followed by 45 cycles of 95°C for 15 s, 60°C for 30 s, and 72°C for 30 s. Nuclease-free water was included as a negative control in each round. The nested PCR products were sent to Sangon Biotech (Shanghai, China) for Sanger sequencing.

Statistical analyses were performed using IBM SPSS Statistics 21 (IBM, United States). Agreement between different detection methods was assessed using the Kappa statistic.

The POCT-DO-RAP workflow was systematically optimized using the MTB reference strain H37Rv as the template. The finalized nucleic acid extraction protocol required approximately 30 min from the onset of lysis. The optimized extraction system consisted of 200 μL of liquefied sputum or BALF sample, 400 μL of lysis buffer, 300 μL of magnetic bead suspension, two washing steps with 400 μL wash buffer each, 100 μL of elution buffer, and 50 μL of mineral oil.

Following extraction, 2 μL of the eluted nucleic acid was automatically dispensed into both the WT and MT PCR tubes. The system then aspirated 2 μL of Mg²⁺ solution from an independent reservoir and added it to each reaction tube. Subsequently, 20 μL of mineral oil was overlaid to seal the reaction chambers. The robotic arm then closed the reaction tube caps and transferred the tubes into the PCR module to initiate downstream amplification.

To achieve optimal amplification performance, the effects of betaine concentration and Mg²⁺ concentration on amplification efficiency and fluorescence signal stability were systematically evaluated.

Optimisation experiments revealed that both components exerted significant influence on the DO-RAP reaction. For betaine, a concentration of 0.4 M produced the most stable and robust amplification curves (Supplementary Figure 1A). Reducing the concentration to 0.2 M resulted in insufficient amplification efficiency, whereas increasing it to 0.6 M caused attenuated or distorted fluorescence signals. These adverse effects at higher concentrations are likely due to excessive lowering of DNA melting temperature, impaired DNA polymerase activity, disruption of recombinase–SSB–DNA complex formation during the RAA phase, altered ionic strength, and increased primer–dimer or nonspecific annealing events.

Similarly, Mg²⁺ concentration had a pronounced impact on amplification performance. A final concentration of 9 mM provided the optimal balance between polymerase activity and primer–template specificity (Supplementary Figure 1B). At 8 mM, insufficient Mg²⁺ limited polymerase activity and weakened primer annealing stability, resulting in delayed Ct values. Conversely, at 10 mM, excess Mg²⁺ may stabilize nonspecific secondary structures or compromise polymerase fidelity, leading to reduced fluorescence intensity and elevated background noise.

Collectively, these results demonstrate that 0.4 M betaine and 9 mM Mg²⁺ constitute the optimal reaction conditions for the POCT-DO-RAP assay.

Based on these optimization experiments, the final WT-tube POCT-DO-RAP reaction mixture (20 μL total volume) consisted of: 8 μL RAA reaction buffer and lyophilized enzyme mix, 0.2 μL PN101 DNA Taq polymerase (5 U/μL), 0.6 μL gyrA-F (10 μM), 0.6 μL gyrA-R (10 μM), 0.2 μL gyrA90-WT-p (10 μM), 0.2 μL gyrA94-WT-p (10 μM), 0.6 μL IS1081-F (10 μM), 0.6 μL IS1081-R (10 μM),0.1 μL IS1081-p (10 μM), 1.6 μL betaine (4 M), and 3.3 μL nuclease-free water. Following nucleic acid extraction, the instrument automatically added 2 μL of 90 mM Mg²⁺ solution (yielding a final Mg²⁺ concentration of 9 mM), along with 2 μL of the extracted nucleic acid template.

The amplification protocol was as follows: an initial RAA phase at 40°C for 10 min, followed by a PCR phase consisting of 95°C for 30 s for pre-denaturation, and 30 cycles of 95°C for 10 s (denaturation) and 62°C for 30 s (annealing/extension).

The analytical sensitivity of the POCT-DO-RAP assay was evaluated using serially diluted recombinant plasmids and H37Rv-spiked healthy sputum specimens. The results demonstrated that the limit of detection (LOD) for both the WT and MT reaction tubes was 1 copy per reaction (Figure 3) and 10 CFU/mL (Figure 4). In comparison, the LOD of qPCR was 10 copies per reaction (Supplementary Figure 2A) and 100 CFU/mL (Supplementary Figure 2B), indicating that POCT-DO-RAP achieved approximately 10-fold higher sensitivity than conventional qPCR.

The specificity assessment showed that all 50 wild-type MTB isolates generated fluorescence signals exclusively in the WT tube. Among the mutant strains, gyrA-A90V (n = 29), gyrA-D94G (n = 43), and dual-mutation isolates (n = 2) produced distinct fluorescence signals in their corresponding mutation-specific tubes. The DO-RAP results were fully concordant with Sanger sequencing. For other mutation types without corresponding mutation-specific probes (gyrA-D94A, gyrA-D94N, gyrA-S91P, gyrA-G88A), only WT-tube fluorescence was observed, consistent with their probe-design characteristics. These results collectively demonstrate the high specificity of DO-RAP for distinguishing wild-type and mutant alleles.

Furthermore, no cross-reactivity was detected with any of the five common respiratory bacterial species tested—Haemophilus influenzae, Streptococcus pneumoniae, Staphylococcus aureus, Klebsiella pneumoniae, and Pseudomonas aeruginosa—all of which yielded negative results (Table 2). This confirms that the POCT-DO-RAP assay possesses excellent target specificity and does not cross-react with non-mycobacterial respiratory pathogens.

To determine the ability of the POCT-DO-RAP assay to detect FQ heteroresistance, mixtures of recombinant plasmids containing wild-type and mutant gyrA sequences were prepared at total template concentrations of 100 copies per reaction.

Across mutant proportions ranging from 1 to 80%, the MT tube consistently generated positive amplification signals for the mutant allele.

The assay successfully detected mutant DNA present at as low as 1% within a predominantly wild-type background (Supplementary Figure 3), demonstrating its high analytical sensitivity for identifying low-frequency resistant subpopulations. These results indicate that POCT-DO-RAP is capable of reliably detecting clinically relevant heteroresistance even when the mutant population is present at very low abundance.

A total of 88 clinical sputum specimens were tested using both the POCT-DO-RAP assay and conventional qPCR. According to the qPCR results, 41 samples were classified as susceptible (positive), 5 yielded borderline results (CT > 38), and 42 were negative; no resistant isolates were detected. In comparison, POCT-DO-RAP identified 50 samples as susceptible and 38 as negative, also without detecting any resistant isolates. Notably, 9 samples that were categorized as borderline or negative by qPCR were confirmed as positive by both DO-RAP assay and nested PCR followed by Sanger sequencing, indicating a superior detection capability of the POCT-DO-RAP assay for low-abundance targets.

Using nested PCR followed by Sanger sequencing as the reference standard, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and Cohen’s kappa coefficient for qPCR were 82, 100, 100, 80.9, and 0.797%, respectively. In contrast, the POCT-DO-RAP assay achieved 100% sensitivity, specificity, PPV, and NPV, with a kappa value of 1.0 (Table 3), demonstrating perfect agreement with the reference method and superior clinical performance over conventional qPCR.