Highly accurate and sensitive diagnostic detection of SARS-CoV-2 by digital PCR

ETHICS STATEMENT

Data collection of cases and close contacts were determined by the National Health Commission of the People’s Republic of China to be part of a continuing public health outbreak investigation and were thus considered exempt from institutional review board approval. The analysis was performed on existing samples collected during standard diagnostic tests, posing no extra burden to patients.

STUDY DESIGN

103 febrile suspected SARS-CoV-2 infected patients, 75 close contacts and 16 supposed convalescents were chosen in this study. Positive, negative and suspect results were all included in the chosen specimens according to the RT-qPCR test.

CLINICAL SAMPLES

Respiratory samples were obtained during February and March 2020 from patients hospitalized or close contacts tested by Beijing CDC (BJCDC), Wuhan CDC (WHCDC) and a government designated clinical test laboratory (Wuhan considerin laboratory for medical test, KXR). Samples were stored in viral transport medium (Yocon Biology) at 4□. RNA was extracted from clinical specimens within 24 hours by using the MagMAX-96 viral RNA isolation kit (Thermo Fisher Scientific). RNA extracts containing human coronaviruses (HCoV)-229E and (HCoV)-OC43 and avian influenza virus RNAs A/H3N2 and A/H1N1 Virus and Influenza B/Victoria Virus were provided by BJCDC. Extracted RNA was stored at -80□.

ONE STEP REVERSE TRANSCRIPTION DPCR (RT-DPCR)

The primer and probe sequences for detecting N and ORF1ab gene target of the SARS-CoV-2 published by Chinese CDC were used for this study(12). For detecting E gene target, primer and probe recommended by world health organization (WHO) was used(13). The 20 μL reaction mixture comprises5 μL of Supermix, 2 μL of reverse transcriptase, 1 μL of 300 mM DTT from One-Step RT-ddPCR Advanced Kit for Probes (Cat. #1864021, Bio-Rad), 1 μL of mixture of primers and probe and 11 μL of RNA template. Each reaction mix was converted to droplets using the QX200 Droplet Generator (Bio-Rad,), transferred to a 96-well plate (Cat. #120019285, Bio-Rad), heat sealed and amplified in a GeneAmp System 9700 thermal cycler (Applied Biosystems). The thermal cycling conditions were as follows: 45°C for 10 min (reverse transcription); 95°C for 5 min; and 40 cycles of 95°C for 15 sec, and 58°C for 30 sec. The cycled plate was then transferred to the QX200 Droplet Reader (Bio-Rad) and analyzed using the QuantaSoft droplet reader software (V1.7.4, Bio-Rad). Reactions containing more than 10,000 droplets were treated as effective and involved in data analysis.

LIMIT OF BLANK (LOB) AND DETECTION (LOD) OF RT-DPCR

To establish the limit of blank (LoB) (14), 60 blank measurements were obtained from 3 blank samples on three days. 70 to 76 measurements from 4-5 samples with low concentration (1 to 3 cp/reaction) were used to determine the limit of detection (LoD) according to the CLSI guideline of EP17-A(15).

RT-QPCR

Three different commercial RT-qPCR kits (H&R from Shanghai Huirui Biotechnology Co., Ltd, BioGerm from Shanghai BioGerm Medical Biotechnology and Daan from Daan Gene Co., Ltd) were used for the detection. Briefly, a 25-μL reaction containing 7.5 μL of PCR reaction buffer, 5 µL of primer and probe mixture and 5∼11 μL of RNA was prepared. Thermal cycling was performed at 50 °C for 15 min for reverse transcription, followed by 95°C for 5 min and then 45 cycles of 95 °C for 10 s, 55 °C for 45 s in ABI 7500 RT-PCR thermocycler. Data analysis was performed using software of ABI 7500 RT-PCR thermocycler.

DYNAMIC RANGE OF THE RT-DPCR ASSAY

The linear range was investigated by varying the mean copy number per droplet, denoted as λ(16). The precision or relative error of RT-dPCR is related to λ because of RT-dPCR relies on the Poisson statistics to account for droplets with multiple molecules(17). The upper limit of the linearity was 7.8 copies/partition tested by N gene assay. To determine the lower limit of all three assays, serial dilutions of each RNA transcript were prepared (Supplemental Table 1). The measured targets matched the anticipated values in each tested interval. A good linearity (0.93<slope<1.02, R² ≥ 0.9997) between the measured RNA target and the prepared value was observed over the range from approximately 10⁴ to 10⁰ copies/reaction (Fig. 1). Reactions containing a mean of 60 E, 66 N or 11 ORF1ab copies fulfilled the criterion for an LoQ with a CV lower than 25%.

• Download figure
• Open in new tab

Fig 1. Validated range of the RT-dPCR assays for E, ORF1ab and N gene.

Evaluation of linearity of samples containing E, ORF1ab and N gene molecules over the extended λ range (0.0002 <λ< 7.83). Data are shown in mean with standard deviation for each dilution series (3≤n≤10).

ESTABLISHMENT OF LOB AND LOD FOR RT-DPCR ASSAY

Sixty blank measurements obtained from 3 blank samples were analyzed to determine the LoB. As the distribution of the 60 blank measurements is skewed (Supplemental Fig. 1), the LoB was estimated nonparametrically as the 95th percentile of the measurements. The 15 highest blank values for each target are displayed in Supplemental Table 2. The 95th percentile corresponds to the 57.5 ordered observation (=60*(0.95/60+0.5))(15). Linear interpolation between the 57th and 58th observation yields a LoB estimate of 1.6, 1.6, and 0.8 copies/reaction for E, ORF1ab and N, respectively.

For determining the LoD of ORF1ab gene assay, 76 measurements were performed on five samples in 3 different runs on three different days to ensure the total assay variation is reflected. The distribution of the 76 measurement results from low concentration samples is not Gaussian (Supplemental Fig. 2A) and so that nonparametric statistics was used according to the guideline of EP17-A. Consequently, the LoD is determined to be 2 copies/reaction, the lowest level material where the β-percentile is 5 %.

• Download figure
• Open in new tab

Fig 2. Diagnostics of SARS-CoV-2 by RT-qPCR and RT-dPCR on 103 febrile suspected patients and 75 close contacts.

25 samples positive, 29 negative and 49 suspected were reported by RT-qPCR (A). 90 positive, 6 negative and 7 equivocal were determined by RT-dPCR (B).10 positive, 49 negative and 16 suspected were reported by RT-qPCR (C). 26 positive, 48 negative and 1 equivocal were determined by dPCR (D).

To determine the LoD of N and E assay, 83 measurements of E assay on 5 samples and 71 measurements of N assay on 4 samples were performed in 4 different runs. Similar to ORF1ab gene, the distribution of the 71 measurements for N gene and 83 measurements are not Gaussian (Supplemental Fig. 2B and 2C), and so that nonparametric statistics was used. Consequently, the LoD is determined to be 2 copies/ reaction.

SPECIFICITY TESTING

The Specificity of the assays for ORF1ab and E gene has been tested in a previous report (13). To further validate the specificity of all assays, Influenza virus and other human coronavirus were collected. All assays were tested on human clinical nucleic acid samples at National institute of Metrology, China. All tests returned negative results (see Supplemental Table 3).

COMPARISON BETWEEN RT-QPCR AND RT-RT-DPCR ON FEBRILE SUSPECTED PATIENTS

103 pharyngeal swabs were collected from febrile suspected SARS-CoV-2 infected patients and the relevant information is listed in Table 1. Among the 103 specimens, 81 (P1 to P81) were tested at KXR with the H&R qPCR kit and 7 (P82-88) were tested at WHCDC by the Daan qPCR kit. Firstly, the criteria claimed by the H&R kit manufacturer are: Ct value≤35 are positive, Ct value >39.2 are negative, and 35<Ct<39.2 are equivocal. The criteria of the Daan qPCR kit are: Ct >40, negative, Ct ≤40, positive, and equivocal if only one gene with Ct ≤ 40 and no amplification for another gene. According to such criteria, 14 positive, 25 negative and 49 suspected SARS-CoV-2 infections were determined and this was reported to the clinical doctors. For RT-dPCR, three targets are tested in parallel and the determination of a positive result should meet the following criteria: quantification of any one of the three gene targets is ≥2 copies/ reaction. If no positive droplet was detected in FAM channel but positive droplets were detected in VIC indicating RNAseP positive for human reference control(11), the sample can be judged negative. If 0 < result < 2, it should be attributed to equivocal and need further check. According to such criteria, 44 out of 49 suspects and 17 out of 25 negatives were corrected to be positive by RT-dPCR and the positive rate significantly increased. No positive droplet was detected for the 6 negatives and copy numbers were quantified under the established LoD for 7 suspects infections, due to either no virus sampled or ultra-low viral load in these specimens.

15 samples (from P89-P103) were tested at BJCDC with BioGerm qPCR kit and assays recommended by Chinese CDC. Ct values were not available and only negative or positive information were reported. Single gene target positive was determined to be SARS-CoV-2 positive based on parallel test with a commercial kit and Chinese CDC assays. Therefore, these 15 samples were reported positive by BJCDC. 8 qPCR negatives for ORF1ab were positive tested by RT-dPCR, showing high sensitivity for ORF1ab by RT-dPCR. Only 3 negatives for ORF1ab which can be complemented by E gene targets.

Among the 103 specimens, 29 positive, 25 negative and 49 suspected were reported by RT-qPCR. However, 61 samples including 17 negative and 44 suspected tested by qPCR were confirmed to be positive by RT-dPCR, thus 90 patients in total whose SARS-CoV-2 nucleic acid were positive tested can be diagnosed with COVID-19. All the 103 patients were confirmed SARS-CoV-2 infection according to a follow-up survey. The sensitivity of SARS-CoV-2 detection was significantly improved from 28.2% to 87.4% for the 103 patients (Fig. 2A and 2B).

COMPARISON ON CLOSE CONTACTS AND CONVALESCENT

75 specimens were collected from contacts and close contacts. 48 specimens from contacts were reported negative based RT-qPCR test by BJCDC on Feb 6 and were confirmed by RT-dPCR on Feb 7 in Supplemental Table 4. According to a follow-up survey, all of them were in good health and isolation was lifted after 14 days.

27 specimens (Table 2 and Fig. 2C and 2D) were detected at WHCDC by RT-qPCR with a kit from Daan gene on March 2, 4 and 6. According to RT-qPCR result, 10 positive, one negative and 16 suspect were reported. It is very difficult to detect the SARS-CoV-2 nucleic acids due the low viral load at the early stage for the close contacts. However, 15 out of 16 equivocal and one negative can be determined positive by RT-dPCR. The suspect rate was significantly decreased from 21% down to 1% according to the detection of RT-dPCR. Consequently, except 6 patients can not be tracked, the rest 10 RT-dPCR positive were confirmed as SARS-CoV-2 infected patients based on a follow-up survey.

Furthermore, among the 16 specimens corrected by RT-dPCR, 6 persons (P14,18-21and P23 in Supplemental Table 5) were directed for secondary testing following an initial negative test 2 to 10 days prior. Based on RT-qPCR results, further isolation and observation was still needed to be conducted as the testing result is suspect or negative and no clinical symptoms were observed for them. However, if based on RT-dPCR, all the six patients can be diagnosed with COVD-19 infected by SARS-CoV-2 and treatment could be conducted earlier. This indicates RT-dPCR is more sensitive and suitable for low viral load specimens from the patients under isolation and observation without clinical symptoms, which is in agreement with the very recent online report (18).

Additionally, 16 pharyngeal swabs were collected from convalescent patients (Table 3). 12 positive, 3 suspect and 1 negative were reported by qPCR. However, all of these 16 patients are diagnosed to be positive by RT-dPCR, indicating that all of them still need to be observed in hospital. Regardless the sampling time, correlation between age and the RNA virus copy number was analyzed (Fig.3). Interestingly, except P15, with increasing age, the copy number of viral load was much higher, which indicates a longer observation in the hospital is needed. We set up the threshold of 15, 20 and 25 copies/reaction for ORF1ab, N and E, respectively. The ORF1ab, N and E gene copy number were higher than their threshold for 100% patients elder than 60 and 75% (6 out 8 patients) elder than 55 (the median).

• Download figure
• Open in new tab

Fig 3. Scatter plot of copies/reaction values of 15 patients detected by RT-dPCR for ORF1ab (A), N (B) and E (C).

The samples are arranged by ages (from younger to older). The blue curves fitted by lasso regression show the distribution of copies/reaction values. The shading that underlies the lasso curves denotes the 95% confidence intervals. The black curve represents the underlying relationship between copies/reaction values and ages of the convalescent patients, which was fitted by the methods of liner models. The liner model was labeled on the left-top of picture. The horizontal line represents the threshold of 15, 20 and 25 copies/reaction for ORF1ab, N and E, respectively. The ORF1ab, N and E gene copy number were higher than their threshold for 100% patients elder than 60 and 75% (6 out 8 patients) elder than 55 (the median).