key: cord-0996193-mskkahwi
authors: Lu, Renfei; Wang, Jian; Li, Min; Wang, Yaqi; Dong, Jia; Cai, Weihua
title: SARS-CoV-2 detection using digital PCR for COVID-19 diagnosis, treatment monitoring and criteria for discharge
date: 2020-03-30
journal: nan
DOI: 10.1101/2020.03.24.20042689
sha: 218ff40b1e8c25b1b9339d6138d15b404036f4b8
doc_id: 996193
cord_uid: mskkahwi

Background: SARS-CoV-2 nucleic acid detection by RT-PCR is one of the criteria approved by China FDA for diagnosis of COVID-19. However, inaccurate test results (for example, high false negative rate and some false positive rate) were reported in both China and US CDC using the RT-PCR method. Inaccurate results are caused by inadequate detection sensitivity of RT-PCR, low viral load in some patients, difficulty to collect samples from COVID-19 patients, insufficient sample loading during RT-PCR tests, and RNA degradation during sample handling process. False negative detection could subject patients to multiple tests before diagnosis can be made, which burdens health care system. Delayed diagnosis could cause infected patients to miss the best treatment time window. False negative detection could also lead to prematurely releasing infected patients who still carry residual SARS-CoV-2 virus. In this case, these patients could infect many others. A high sensitivity RNA detection method to resolve the existing issues of RT-PCR is in need for more accurate COVID-19 diagnosis. Methods: Digital PCR (dPCR) instrument DropX-2000 and assay kits were used to detect SARS-CoV-2 from 108 clinical specimens from 36 patients including pharyngeal swab, stool and blood from different days during hospitalization. Double-blinded experiment data of 108 clinical specimens by dPCR methods were compared with results from officially approved RT-PCR assay. A total of 109 samples including 108 clinical specimens and 1 negative control sample were tested in this study. All of 109 samples, 26 were from 21patients reported as positive by officially approved clinical RT-PCR detection in local CDC and then hospitalized in Nantong Third Hospital. Among the 109 samples, dPCR detected 30 positive samples on ORFA1ab gene, 47 samples with N gene positive, and 30 samples with double positive on ORFA1ab and N genes. Results: The lower limit of detection of the optimize dPCR is at least 10-fold lower than that of RT-PCR. The overall accuracy of dPCR for clinical detection is 96.3%. 4 out 4 of (100 %) negative pharyngeal swab samples checked by RT-PCR were positive judged by dPCR based on the follow-up investigation. 2 of 2 samples in the RT-PCR grey area (Ct value > 37) were confirmed by dPCR with positive results. 1 patient being tested positive by RT-PCR was confirmed to be negative by dPCR. The dPCR results show clear viral loading decrease in 12 patients as treatment proceed, which can be a useful tool for monitoring COVID-19 treatment. Conclusions: Digital PCR shows improved lower limit of detection, sensitivity and accuracy, enabling COVID-19 detection with less false negative and false positive results comparing with RT-PCR, especially for the tests with low viral load specimens. We showed evidences that dPCR is powerful in detecting asymptomatic patients and suspected patients. Digital PCR is capable of checking the negative results caused by insufficient sample loading by quantifying internal reference gene from human RNA in the PCR reactions. Multi-channel fluorescence dPCR system (FAM/HEX/CY5/ROX) is able to detect more target genes in a single multiplex assay, providing quantitative count of viral load in specimens, which is a powerful tool for monitoring COVID-19 treatment.

1 0 8 clinical 28 specimens by dPCR methods were compared with results from officially approved RT-PCR 29 assay. A total of 109 samples including 108 clinical specimens and 1 negative control sample were 30 tested in this study. All of 109 samples, 26 were from 21patients reported as positive by officially 31 approved clinical RT-PCR detection in local CDC and then hospitalized in Nantong Third 32 following manufacturer's instruction. 89

The primers and probes targeted the ORF1ab and N of SARS-CoV-2 according to Chinese CDC. 91 Target For RT-PCR workflow, the primers and probes are from Liferiver Bio-Tech. A25-μlreaction was 116 set up containing 5 μ l of RNA, 19 μ l of reaction buffer provided with the one step RT-PCR 117 system and 1µl enzyme mix. Thermal cycling was performed at 45°C for 10 min for reverse 118 transcription followed by 95°C for 3 min and then 45 cycles of 95 °C for 15 sec, 58 °C for 30 sec 119 in SLAN 96P real time PCR system. 120

Analysis of the dPCR data was performed with analysis software GeneCount V1.60b0318 122 (RainSure Scientific). Concentrations of the target RNA sequences, along with their Poisson-123 based 95% confidence intervals were also provided by the software. Fluorescence channels of 124 FAM, HEX and Cy5 were scanned to detect ORF1ab gene, N gene and RPP30 gene respectively. 125

The positive populations for each target gene were identified using positive and negative controls 126 with single primer-probe sets for each fluorescence channel. The concentration reported by 127 GeneCount has the unit of copies of template per microliter of the final 1× dPCR reaction, which 128 was also reported and used in all the subsequent analysis. 129

3.1. Comparison of the lower limit of detection between dPCR and the standard RT-PCR 131 L ower limit of detection (LLoD) of RT-PCR and dPCR was compared using serial dilution of clinical 132 specimen. The starting clinical specimen showed Ct value of 35 in RT-PCR. The specimen was diluted 133 using virus storage solution. Each dilution was 5 fold. A total of 7 dilutions (8 samples S1-S8 134 including the starting stock) were tested by both RT-PCR and dPCR assays. As shown in Figure 1 and 135 Table 2 RT-PCR failed to detect S3, while dPCR was able to detect S3 and S4. dPCR showed 136 negative results for S5 through S8. dPCR assay showed at least 10 times lower LLoD than RT-PCR 137 assay. However. LLoQ (lower limit of quantification) was estimated to be larger than the viral 138 concentration in S3. 139 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Results from dPCR targeting ORF1ab gene were also consistent with that from RT-PCR result (Table  144 4). In particular, 4 positive samples from 3 patients based on dPCR assay were not detected by PCR. The computerized tomography (CT) results of the 3 patients showed 1 patient's lung texture 146 thickening, no pneumonia imaging characteristics; 1 patient had two lungs infections,; and 1 patient 147 showed inflammation of the right lung and lower lobe. But RT-PCR assay failed to detect these 4 148 samples, and generated negative results. 149

Digital PCR targeting N gene exhibited higher sensitivity than that targeting ORF1ab gene. As shown 150 in to be positive based on other factors. One was a suspected case. One of them were of unknown clinical 153 classification. In particular, one was actually discharged based on RT-PCR assay result, although a 154 retrospective dPCR test detected the virus. In contrast, as long as a sample is deemed positive by 155

ORF1ab gene, all but one were positive by N gene. 156

Digital PCR also provides a window to monitor the progression and treatment of disease more 158 consistently than RT-PCR. We analyzed positive samples from 3 patients based on dPCR assay that 159 were not consistently detected by Patient 33 was first identified as SAR-CoV-2 positive by RT-PCR from pharyngeal swab with 161

ORF1ab gene Ct of 32.3 and N gene Ct of 33.2 ( Table 6) . dPCR showed ORF1ab concentration of 162 28.3 copies/µl, N gene concentration of 35.2 copies/µl and internal reference gene concentration of 163 32.8 copies/µl on January 27, 2020 (Figure 2 and Table 6 ). The patient was treated in hospital under 164 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. These examples support our conclusion that dPCR offers improved sensitivity and consistency when 205 testing specimens from patients during the course of treatment. dPCR is also able to detect low viral 206 load in asymptomatic infection patients and suspected patients. dPCR can check if the negative result 207 was caused by insufficient RNA loading by quantify the copy number of internal reference gene 208 RPP30. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020. . https://doi.org/10.1101/2020.03.24.20042689 doi: medRxiv preprint using RT-PCR and dPCR targeting ORF1ab genes, but dPCR targeting N gene reported 2 positive 217 results. From dPCR results, it clearly indicates viral load in patients is throat > stool > blood. 218

The samples with ORF1ab positive detected in throat for dPCR is 6 more than . 219

The samples with ORF 1ab positive detected in blood and sputum for dPCR is 1 more than RT-PCR (1 220 vs 0). While the samples with ORF1ab positive detected in stool for dPCR is 2 less than RT-PCR (2 vs 221 4). Therefore, dPCR showed less sensitivity than RT-PCR in specimens from stool, higher sensitivity 222 than RT-PCR in specimens from throat, blood and sputum. Different detection sensitivity of dPCR 223 assay for different sample types may require further experiments to draw a conclusion. 224

Stool specimen from patient 4 on Feb. 24 th was tested as negative by both However, the internal reference gene result by dPCR is 0 suggesting no RNA was loaded in either tests 227 or the PCR reactions were inhibited. The samples need to be re-collected to re-run the tests. The 228 internal reference gene of dPCR assay serves as a quality control to ensures no PCR inhibition 229 happened and RNA extraction was successful. Without such a control in RT-PCR assay, the patient 230 could have been discharged prematurely, putting those in close contact at risk of being infected. 231 232

We first demonstrated that the LOD of dPCR assay is at least 10 times better than that of RT-PCR 234 assay using serial dilution of the same clinical sample. Higher sensitivity, along with more reliable 235 quantification of viral load, provides valuable information to help clinicians choose the appropriate 236 treatment plan 16 . To demonstrate its application in clinical settings, we performed head-to-head 237 comparison of RT-PCR and dPCR assays using a cohort of 39 patients totaling 109 samples obtained 238 at different stage of the treatment and from different locations.. 239

We observed that the result of dPCR targeting ORF1ab gene is consistent with RT-PCR targeting the 240 same. In particular, dPCR assay detected 4 positive samples that were determined to be negative by 241 RT-PCR assay. Although there were also 3 positive samples from RT-PCR assay that were deemed 242 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020. . The dPCR results also showed higher sensitivity when the primer and probe are designed against N 248 gene. It may be explained by the higher copy number of RNA for N gene arising from the replication 249 process of the virus 17,18 . Interestingly, RT-PCR assay didn't exhibit significance difference between the 250 two targets. Further study is warranted to elucidate the underlying mechanism. It is well recognized 251 that dPCR assays are less susceptible to the existence of PCR inhibitors, and the results are thus more 252 reliable in general 11, 19 . 253

With dPCR assay, we were able to track the progress of the treatment by monitoring the viral load 254 from samples obtained on different dates. RT-PCR suffered from sporadic appearance of positive 255 result which puzzled clinicians. dPCR results, in contrast, faithfully reflected the onset and healing of 256 the disease, when examined together with relevant radiological evidence and treatment history. dPCR 257 showed evidences of higher sensitivity to detect low virus load in patients who showed mild symptoms 258 or have been treated for COVID-19 than We were also able to compare the viral load in different organs thanks to higher sensitivity of dPCR 260 assay. The dPCR assay provides quantitative information on the viral load of specimens collected from 261 different locations of the same patient. In all but one cases, the viral load is the highest in pharyngeal 262 samples, lower in stool samples and the lowest in serum. Interestingly, a considerable amount of virus 263 was found in the phlegm of one of the patients whose pharyngeal sample was negative. These 264 observations may provide valuable insight into the pathology of this emerging disease 20 . 265

Digital PCR shows improved lower limit of detection, sensitivity and accuracy, enabling COVID-19 267 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020 . . https://doi.org/10.1101 detection with less false negative and false positive results comparing with RT-PCR, especially 268 for the tests with low viral load specimens. We showed evidences that dPCR is powerful in 269 detecting asymptomatic patients and suspected patients. Digital PCR is capable of flagging the We are grateful to RainSure Scientific for their great support to this work. We are grateful to Drs. 287 (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020. . The clinical specimen showed RT-PCR Ct value at 35. The specimen was diluted using virus 380 storing solution 5 for each dilution. Each dilution is 5 fold. Total of 7 dilutions (8 concentrations) 381

were tested by both RT-PCR and dPCR assays. Orange color is internal reference, green is N gene, 382 blue is ORF1ab for both RT-PCR and dPCR. S1 is original specimen (left RT-PCR, right dPCR). 383 S2 is S1 specimen 5X dilution (left RT-PCR, right dPCR). S3 is S2 specimen 5X dilution (left 384 RT-PCR, right dPCR). S4 is S3 5X dilution (left RT-PCR, right dPCR). S5-S8 are 5X series 385 dilutions from S4 specimen. The dPCR assay showed at least 10 fold lower LLoD than RT-PCR 386 assay. RT-PCR failed to detect at S3 dilution, dPCR was able to detect S3 and S4 dilution. 387 However. LLoQ (lower limit of quantification) is above S3 concentration. 388 S1 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020 . . https://doi.org/10.1101 389 S4 S3 S2 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020 . . https://doi.org/10.1101 (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020 . . https://doi.org/10.1101 A B All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020 . . https://doi.org/10.1101 (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 30, 2020 . . https://doi.org/10.1101 

293 2. WHO. WHO Director-General's opening remarks at the media briefing on

Emerging coronaviruses: Genome structure, replication, and 297 pathogenesis

RNA based mNGS approach identifies a novel human 299 coronavirus from two individual pneumonia cases in 2019 Wuhan outbreak

Accessed March 15, 2020. 304 6. China public health commission. Clinical guidance for the diagnosis and treatment of

Characteristics of and Important Lessons From the Coronavirus 307 Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the 308

Chinese Center for Disease Control and Prevention

Principle and applications of digital PCR

High-throughput droplet digital PCR system for 315 absolute quantitation of DNA copy number

Absolute quantification by droplet digital PCR 318 versus analog real-time PCR

Quantitation of HBV cccDNA in anti-HBc-positive 320 liver donors by droplet digital PCR: A new tool to detect occult infection

Detection and quantification of HDR and 323 NHEJ induced by genome editing at endogenous gene loci using droplet digital PCR

Droplet-based digital PCR 326 and next generation sequencing for monitoring circulating tumor DNA: a cancer diagnostic 327 perspective

Digital droplet PCR (dPCR) for the precise 329 quantification of human T-lymphotropic virus 1 proviral loads in peripheral blood and cerebrospinal 330 fluid of HAM/TSP patients and identification of viral mutations