The original strain of SARS-CoV-2 and variants used for sensitivity evaluation were obtained from the laboratory of the Zhejiang Provincial Center for Disease Control and Prevention (Zhejiang CDC), China. The original strain, SARS-CoV-2/E6/WGF/2020/ZJ8, had a titer of 3.76 × 10⁶ TCID50/mL; the Delta strain, SARS-CoV-2/Vero/LXG/2021/ZJ28, had a titer of 3.73 × 10⁵ TCID50/mL; and the Omicron strain, SARS-CoV-2/E6/Gabriol/2022/ZJ60, had a titer of 3.16 × 10⁵ TCID50/mL.

Seven common respiratory viruses and SARS-CoV-2 pseudovirus quantification reference material (Fantasiabio, Zhejiang, China; RFKSS001) were used to evaluate the specificity of RT-ddPCR. Seven common respiratory viruses were obtained from Zhejiang CDC, including influenza A (H1N1)pdm 09, Victoria lineage of influenza B virus, respiratory syncytial virus subtype A, human parainfluenza virus type III, adenovirus type 7, human coronavirus OC43, and human coronavirus 229E. The original strain and thirty SARS-CoV-2 variants, including the three variants of concern (VOCs) (Alpha, Delta, and Omicron) were used for specificity evaluation. The Omicron variants comprised 28 sublineages: BA.1.1, BA.2.12.1, BA.2.3, XBB.1.5.4, XBB.1.9.2, EG.5.1, EG.5.1.1, HK.3, XBB.1.16, FU.1, XBB.1.22, BA.4.1, BA.5.2, BF.7, BA.5.2.48, DY.2, BQ.1.1, BF.7.14, JN.1, JN.1.4.5, LB.1.2, JN.1.16, KP.2, KP.3.1.1, JN.1.67.1, XDV.1, XDV.1.5, and NB.1.

For quantitative comparison between RT-ddPCR and RT-qPCR, a total of 148 nasopharyngeal samples collected from fever clinics between June 2022 and December 2024 at three hospitals: Yiwu Central Hospital, Hangzhou First People’s Hospital, and Children’s Hospital, Zhejiang University School of Medicine, China. All samples were collected in 3 mL universal viral transport medium and transported to the laboratory immediately or storage at −80°C until use. Additionally, 50 environmental wastewater samples were obtained from the inlet of the Qianjiang sewage-treatment plant in Xiaoshan District, Hangzhou City, Zhejiang Province, China. The initial volume of each wastewater sample was 5 L.

The samples were transported to the laboratory at 4°C and concentrated through hyperfiltration. Firstly, 400 mL of the 5 L untreated wastewater was centrifuged at 2,500 × g for 20 min at 4°C after blending. The liquid supernatant was then placed in an aseptic bottle and concentrated with a tubular ultrafiltration membrane connected to an ultrafiltration device. Next, the viruses retained on the ultrafiltration membrane were eluted with 2 mL of 3% beef extract solution. The eluates were stored at −80°C until use.

RNeasy Mini Kit (Qiagen, Hilden, Germany; 74,104) was used to extract RNA according to the manufacturer’s instructions. A total of 50 μL viral RNA was extracted from 200 μL clinical and concentrated wastewater samples. The RNA was stored at −80°C until use.

SARS-CoV-2 specific primer-probe sets targeting the N and S genes were developed for RT-ddPCR and RT-qPCR are provided in Table 1. The primer-probe sets were synthesized by Shanghai Sangon Company (Shanghai, China). All primers and probes were prepared at a concentration of 20 μmol/L, and primer-probe mix of the N and S genes were prepared in working solutions with a volume of 2:2:1.

Absolute quantification of SARS-CoV-2 RNA was performed with the One-Step RT-ddPCR Advanced Kit for Probes (Bio-Rad, Hercules, USA; 1,864,021). The reaction mixture volume of 20 μL comprised 5 μL Supermix, 2 μL primer-probe mix, 2 μL reverse transcriptase, 1 μL 300 mM dithiothreitol (DTT), 8 μL nuclease-free water, and 2 μL RNA template. According to the manufacturer’s guidelines, the reaction mixture was used to produce droplets with a Bio-Rad Auto Droplet Generator (Bio-Rad, USA). Thermal cycling for all RT-ddPCR assays was performed with a T100 Thermal Cycler (Bio-Rad, USA) under the following amplification conditions: 45°C for 60 min; 95°C for 10 min; 40 cycles of 95°C for 30 s and 52°C to 60°C for 1 min; 98°C for 10 min; and storage at 4°C. Droplet signals were read in different channels using the QX200 Droplet Reader (Bio-Rad, USA), FAM channel for the N gene, and VIC channel for the S gene. Data were considered valid if the total number of droplets in each tube was ≥ 10,000. A sample was considered positive if the number of droplets exceeded three and negative if the number of droplets was three or fewer. Results are expressed as copies per reaction (copies/reaction).

We used the original strain to optimize the RT-ddPCR annealing temperature and DTT concentration. The annealing temperature of RT-ddPCR was optimized by testing eight temperatures (60.0, 59.4, 58.4, 56.9, 55.1, 53.5, 52.5, and 52.0°C) with the described method. The optimal annealing temperature was determined on the basis of the signal discrimination and nucleic acid copy number of each reaction.

In PCR systems, DTT is frequently employed as a protein reducing agent to preserve the sulfhydryl group of cysteine in proteins in a reduced state and to safeguard PCR reaction enzymes. Subsequently, with the optimal annealing temperature, a 300 mM DTT concentration in the RT-ddPCR reaction system was optimized. Two concentrations, 0.5 μL and 1 μL of 300 mM DTT, were tested in four repeated experiments, and the results were analyzed to determine the optimal reaction conditions.

An AgPath-ID™ One-step RT-PCR Kit (Thermo Fisher, Carlsbad, CA, USA; AM1005) was used to prepare a 25 μL RT-qPCR reaction system comprising 12.5 μL 2 × RT-PCR buffer, 1.5 μL primer-probe mix, 1 μL reverse transcriptase, 8 μL nuclease-free water, and 2 μL RNA template. The RT-qPCR reaction conditions were as follows: 45°C for 10 min; 95°C for 10 min; and 40 cycles of 94°C for 15 s and 53.5°C for 35 s. The cycle threshold (CT) value was derived from the amplification of the N gene and S gene of SARS-CoV-2 variants using the ABI 7500 Real Time PCR System (Applied Biosystems, USA).

We assessed the sensitivity of the RT-ddPCR method by determining the LOD for three strains of SARS-CoV-2 (original, Delta, and Omicron). The initial nucleic acid concentrations of these strains were quantified with identical primer-probe systems in a preliminary experiment. Samples were initially diluted by a factor of 10 and subsequently 2-fold serially diluted for samples with low nucleic acid concentrations. The diluted nucleic acid samples underwent RT-ddPCR detection to determine the positive detection rate. LOD was defined as the concentration corresponding to the 95% confidence interval of nucleic acid copies. A lower LOD indicated higher detection sensitivity.

Additionally, influenza A (H1N1)pdm 09, Victoria lineage of influenza B virus, respiratory syncytial virus subtype A, parainfluenza virus type III, adenovirus type 7, human coronavirus OC43, human coronavirus 229E, and SARS-CoV-2 pseudovirus quantification reference material were used to evaluate the specificity of the RT-ddPCR method developed in this study.

SARS-CoV-2 pseudovirus nucleic acid standards were utilized to extract nucleic acids and evaluate the repeatability of RT-ddPCR. The concentrations of positive quantitative reference materials ranged from 1 × 10² to 1 × 10⁶ copies/mL, accompanied by a negative control group. Each concentration group underwent 16 tests under optimal RT-ddPCR conditions. The resulting data were used to calculate the mean, standard deviation, and coefficient of variation to assess the repeatability of the experiment. A coefficient of variation less than 10% indicated good repeatability.

The original strain and thirty variants of SARS-CoV-2 were selected: Alpha (B.1.1.7), Delta (B.1.617.2), BA.1.1, BA.2.12.1, BA.2.3, XBB.1.5.4, XBB.1.9.2, EG.5.1, EG.5.1.1, HK.3, XBB.1.16, FU.1, XBB.1.22, BA.4.1, BA.5.2, BF.7, BA.5.2.48, DY.2, BQ.1.1, BF.7.14, JN.1, JN.1.4.5, LB.1.2, JN.1.16, KP.2, KP.3.1.1, JN.1.67.1, XDV.1, XDV.1.5, and NB.1. The RT-ddPCR and RT-qPCR methods were used for quantitative detection, and reaction systems were as previously described.

A total of 148 clinical specimens from three hospitals (Yiwu Central Hospital, Hangzhou First People’s Hospital, and Children’s Hospital, Zhejiang University School of Medicine, China) were analyzed with the RT-ddPCR method described previously. Additionally, 50 wastewater samples from a sewage treatment plant in Hangzhou underwent the same extraction and amplification methods. The results obtained via RT-ddPCR subsequently compared with those obtained via RT-qPCR.

RT-ddPCR data were analyzed using QuantSoft 1.7.4.0917 software (Bio-Rad). The LOD, the coincidence rate, and Kappa value were calculated through SPSS 26.0 software, with a threshold for positive detection set at 95%. Statistical significance was set at p < 0.05.

Eight gradient annealing temperatures (60.0, 59.4, 58.4, 56.9, 55.1, 53.5, 52.5, and 52.0°C) were tested, revealing that as the temperature decreases, the RT-ddPCR amplification efficiency gradually increases (see Table 2). Improved separation between positive and negative droplets was notably observed within the range of 53.5–52.0°C (see Figure 1). To ensure PCR specificity and optimize amplification efficiency, we identified 53.5°C as the optimal annealing temperature.

The results comparing the optimal volume of 300 mM DTT in the reaction system, between 0.5 μL and 1 μL, were replicated four times each (see Figure 2). Using of 0.5 μL 300 mM DTT in a 20 μL reaction system enabled clear differentiation between positive and negative droplets. Consequently, 0.5 μL is considered the optimal concentration of 300 mM DTT. Therefore, the final optimized RT-ddPCR reaction system consisted of 20 μL comprising 5 μL SuperMix, 2 μL primer-probe mix, 2 μL reverse transcriptase, 0.5 μL DTT (300 mM), 8.5 μL nuclease-free water, and 2 μL RNA template.

Three strains of SARS-CoV-2, original, Delta, and Omicron, were initially diluted to concentrations of 10⁵, 10⁴, and 10⁴ copies/reaction, respectively, and followed by 2-fold serial dilution for RT-ddPCR analysis. Each dilution underwent 16 repetitions to achieve a 95% positive detection rate. Probit regression analysis was performed on the average copy numbers and positive detection rates of the N and S genes across various dilutions for each strain. The fitted curves demonstrated that the p values from Probit analysis for all three strains were <0.05, indicating statistical significance. The LOD for the original strain was 4.26 copies/reaction (95% CI: 3.12–9.89) for the N gene and 3.87 copies/reaction (95% CI: 2.77–7.75) for the S gene. For the Delta strain, the LOD was 4.65 copies/reaction (95% CI: 3.28–9.64) for the N gene and 6.12 copies/reaction (95% CI: 4.33–15.59) for the S gene. For the Omicron strain, the LOD was 4.07 copies/reaction (95% CI: 3.11–6.26) for the N gene and 4.58 copies/reaction (95% CI: 3.43–7.40) for the S gene. Detailed results can be found in Table 3.

The specificity of RT-ddPCR was evaluated with influenza A (H1N1)pdm 09, Victoria lineage of influenza B virus, respiratory syncytial virus subtype A, parainfluenza virus type III, adenovirus type 7, human coronavirus OC43, human coronavirus 229E, and SARS-CoV-2 pseudovirus quantification reference material (see Figure 3). Positive droplets were observed for only the SARS-CoV-2 pseudovirus quantification reference material, whereas no positive detection was observed for the other viruses. This funding indicates that the RT-ddPCR method exhibits specificity for SARS-CoV-2 detection without cross-reacting with other respiratory viruses.

The repeatability evaluation results of SARS-CoV-2 pseudovirus quantitative reference materials using the RT-ddPCR method were presented in Table 4. These results were compared with the target values specified in the instructions: L1: 1 × 10² copies/mL, L2: 1 × 10³ copies/mL, L3: 1 × 10⁴ copies/mL, L4: 1 × 10⁵ copies/mL, and L5: 1 × 10⁶ copies/mL. The errors ranged from 1.38 to 11.38%, thus demonstrating consistency between the quantitative results and the target values of the reference materials. For nucleic acid copy numbers >73.50 copies/reaction, the coefficients of variation, calculated from the mean and standard deviation, were all below 10%, thereby indicating that the RT-ddPCR had excellent repeatability within the range of 73.50–7,500 copies/reaction, thus ensuring the reliability experimental outcomes.

The RT-ddPCR results for the original strain and 30 variants of SARS-CoV-2 revealed copy numbers ranging from 3 to 46,010 copies/reaction for the N gene, and from 1.4 to 54,300 copies/reaction for the S gene. Correspondingly, the N gene CT values in RT-qPCR ranged from 37.69 to 23.55, whereas the S gene CT values ranged from 37.47 to 22.51. These findings are detailed in Table 5 and illustrated in Figure 4. The consistency between the results of the two methods across the SARS-CoV-2 variants suggested that RT-ddPCR was well-suited to detection of SARS-CoV-2 variants.

We compared the established RT-ddPCR method with RT-qPCR in the detection of 148 clinical specimens and 50 wastewater samples. The RT-ddPCR analysis consistently detected more than 10,000 effective droplets per sample, thus ensuring data reliability. Among the 148 clinical specimens, both RT-ddPCR and RT-qPCR identified 128 positives for dual gene targets and 20 negatives, and the positive rate were 86.49% between two methods. There were 146 samples with consistent results, with a concordance rate of 98.65% and a Kappa value of 0.94. At 95% confidence interval, the sensitivity was 99.22% (95.02–99.96%) and the specificity was 95.00% (73.06–99.74%). RT-ddPCR quantification results ranged from 1.8 to 61,600 copies/reaction for the N gene and from 1.4 to 60,600 copies/reaction for the S gene. Corresponding RT-qPCR CT values ranged from 18.90 to 38.72 for the N gene and 18.51 to 38.86 for the S gene (see Supplementary Table 1).

In the analysis of low-concentration wastewater samples, RT-ddPCR detected 50 positives, whereas RT-qPCR identified 21 were positive for both the N and S genes, 25 were positive for the S gene, and 4 were negative. RT-ddPCR quantification results ranged from 0.6 to 163 copies/reaction for the N gene, and from 0.6 to 136 copies/reaction for the S gene. In contrast, in RT-qPCR, the CT values ranged from 37.92 to 32.01 for the N gene, and from 38.16 to 30.33 for the S gene. Overall, RT-ddPCR had slightly superior quantification performance to that of RT-qPCR in detecting low viral loads (see Table 6).