Direct quantitation of SARS-CoV-2 using droplet digital PCR in suspected samples with very low viral load

Background: The reference test for SARS-CoV-2 detection is the reverse transcriptase real time PCR (real time RT-PCR). However, evidences reported that real time RT-PCR has a lower sensitivity compared with the droplet digital PCR (ddPCR) leading to possible false negative in low viral load cases. Methods: We used ddPCR for viral genes N1 and N2 on 20 negative (no detection) samples from symptomatic hospitalized COVID-patients presenting fluctuating real time RT-PCR results and 10 suspected samples (Ct value>35) from asymptomatic not hospitalized subjects. Results: ddPCR performed on RNA revealed 65% of positivity for at least one viral target in the hospitalized patients group of samples (35% for N1 and N2, 10% only for N1 and 20% only for N2) and 50% in the suspected cases (30% for N1 and N2, while 20% only for N2). On hospitalized patients’ samples, we applied also a direct ddPCR approach on the swab material, achieving an overall positivity of 83%. Conclusion: ddPCR, in particular the direct quantitation on swabs, shows a sensitivity advantage for the SARS-CoV-2 identification and may be useful to reduce the false negative diagnosis, especially for low viral load suspected samples.


INTRODUCTION
One year has passed since the novel coronavirus SARS-CoV-2 was isolated for the first time. The World Health Organization (WHO) declared COVID-19 a pandemic in March 2020, the state of emergency continues to be a serious and global challenge. SARS-CoV-2 belongs to the sub-family Coronavirinae, subgenus of Sarbecovirus, and it is a non-segmented, positive sense RNA and enveloped virus [1]. Currently, the detection of SARS-CoV-2 is performed on standardized molecular methods, usually in nasal/pharyngeal swabs [2]. The Centers for Disease Control and Prevention (CDC) developed the first clinical assay used worldwide for the SARS-CoV-2 detection [3]. As of today, innumerable different real time RT-PCR commercial kits are available for this purpose, but the sensitivity is often suboptimal for low viral load specimens [4] [5]. We previously performed a study on a cohort of 346 SARS-CoV-2 patients enrolled in the emergency room (ER) and underwent for the first time to a nasopharyngeal swab during the first pandemic wave (March-May 2020). Our study found significant differences in sensitivity using three real time RT-PCR tests including six different gene targets of the virus [6]. Among the molecular technologies, the droplet digital PCR (ddPCR) is an accurate and precise tool for the amplification of reaction based on the partitioning of the sample into thousands of micro-reactions of defined volume in aqueous droplets in oil [7]. This method showed a higher resistance to the amplification inhibitors, compared to the quantitative realtime PCR [5] [8] [9]. Recent studies reported the usage of ddPCR for the quantitation of SARS-CoV-2, showing higher sensitivity compared to real time RT-PCR, especially in low viral load specimens [4] [10] [11] [12]. In a previous study [13], we assessed a direct approach for the SARS-CoV-2 quantitation using ddPCR on the nasopharyngeal swab material without the RNA extraction showing a higher sensitivity compared to the quantitation performed on the extracted RNA of positive and negative samples. That study was mainly aimed to assess the analytical performance of the direct approach on a small group and using two different nasopharyngeal swab types of common use. In the present study, we wanted to extend the application and the performance evaluation of our ddPCR approach (direct compared with extracted RNA) focusing on the potentially false negative specimens from hospitalized COVID-19 patients and asymptomatic subjects diagnosed with SARS-CoV-2. We compared both the ddPCR approaches (direct and on the extracted RNA) to the real time RT-PCR used as our routine diagnostic tool. For our purpose, the analysis was performed on a subgroup of 20 symptomatic hospitalized COVID-19 patients with repeated nasopharyngeal swab testing and who presented at least one negative sample by real time RT-PCR followed by a positive one. Moreover, we analyzed 10 cases additional from asymptomatic subjects who were classified as SARS-CoV-2 infection suspected cases by our routine diagnostic assays based on real time RT-PCR results on nasopharyngeal swab.
In addition, we tested 10 samples resulted with suspicious SARS-CoV-2 infection and collected from asymptomatic not hospitalized subjects (AS 1-10) who were enrolled in a previous epidemiological study conducted on a cohort of 1,515 participants in Verona city (Italy) [14]. During the first wave, our molecular biology laboratory adopted different diagnostics methods for the SARS-CoV-2 detection according to WHO guidelines.
[15] Thus, two of ten samples were tested with CDC 2019-Novel Coronavirus (2019-nCoV) Real-Time RT-PCR Diagnostic Panel [3] (able to detect N1 and N2 genes) and showed a borderline Ct value > 35 at the real time RT-PCR only for one of the two targets.
The remaining 8 samples were screened with Corman et al. in-house protocol targeting the envelope protein gene (E) and the RNA-dependent RNA polymerase (RdRp) gene [16], and presented a low amplification curve below the threshold baseline. Anyway, these 8 samples were considered as positives as defined by our internal guidelines during the first pandemic wave in order to avoid a viral spread.
Moreover, in order to evaluate the performance of ddPCR, we analyzed additional 10 true negative samples from healthy subjects who were under screening by multiple nasopharyngeal swabs and real time RT-PCRs, and serology (IgM and IgG, ELISA) for SARS-CoV-2 infection (data not shown).
Since the study was conducted retrospectively on archived samples, the analysis was performed on different nasopharyngeal swab types available from our biobank: the ESwab 1 mL (COPAN) was used for the 20 hospitalized patients and the UTM 3 mL (COPAN) for the 10 asymptomatic suspected cases. Thus, we analyzed five of the true negatives in ESwab and five true negatives in UTM. All the biological materials were stored at -80°C for further use. For the purpose of our study, the aliquots of swab medium and RNA were thawed and immediately used for the quantitation by ddPCR. First we 4 tested the RNA and in case of negative result, we performed a direct quantitation on the stored swab medium ( Figure 1).

Automated RNA extraction
RNA was isolated from 200 μl of nasopharyngeal swab medium by Microlab Nimbus workstation (Hamilton) coupled to a Kingfisher Presto System (Thermo Fisher). The MagnaMax Viral/Pathogen extraction kit (Thermo Fisher) was used according to the manufacturer's instructions. Samples were eluted in 100μL of elution buffer. The isolated RNA was used for real time RT-PCR (routine diagnostics) and then stored at -80 °C for further ddPCR analysis.

ddPCR-one step reverse transcriptase
We performed the ddPCR analysis according to the manufacturer's instructions of the 2019-nCoV CDC ddPCR triplex probe assay (dEXS28563542, Bio-Rad) as previously described [13].
We used a negative control (no template control, NTC) and a positive control (mixture of synthetic viral target N1&N2, and the human gene RPP30 as control of amplification). The analyses were performed on QX200 ddPCR system (BioRad). The reactions with less than 7,000 droplets were repeated. Data were analyzed using the QuantaSoft™ v1 AnalysisPro Sofware (Bio-Rad) and expressed as Log10 (copies/µL).

Data statistical analysis
The performance analysis of real time RT-PCR and ddPCR was performed by MEDCALC (https://www.medcalc.org/calc/diagnostic_test.php).

Setting of the study for the hospitalized patients
From March to May 2020, 213 patients affected by COVID-19 were hospitalized in our hospital.
Nineteen of them (8 females and 11 males; median age 72 and 67 years respectively) were selected for the present study because during hospitalization their nasopharyngeal swabs monitoring presented fluctuating results by real time RT-PCR (after a confirmed positivity, they presented negative and then results positive again, Figure 2). Clinical characteristics and treatments of the subjects are 5 reported in Table 1S. The first negative test was registered on average after the 20 th day from the diagnosis, but then they resulted to be positive until 45 days on average ( Figure 2). The real time RT-PCR results on the first positive swab after the negative one showed a range of Ct value from 20 to 39 (SD=5.19); data are show in Figure 2. Thus, in order to assess if the negative result could be due to the presence of undetectable viral loads by real time RT-PCR, we performed on the first negative sample the ddPCR on RNAs and swab medium. Table 1

Direct ddPCR analysis on swab-derived material for the hospitalized patients
After the ddPCR analysis on the RNA as described above, we used the direct approach [13] on the swab medium of samples resulted negative at ddPCR for both viral targets or positive only for one. So, the direct quantitation was performed on 14 out of 20 samples (Table 1) with those from the analysis on RNA described above, we detected 17/20 (85%) positives.

Setting of the study for the suspected cases
During the pandemic, we often faced with a not clear real time RT-PCR result. In particular, here we analyzed 10 suspected cases of SARS-CoV-2 infection in asymptomatic subjects. In these specimens the real time RT-PCR of our routine diagnostics detected suspicious Ct values (> 35) only in one

Direct ddPCR analysis for the suspected cases
After the ddPCR analysis on the RNA, we used the direct approach [13] as described for the samples. Thus, if we combine the results obtained using the direct ddPCR with those from the analysis on RNA described above, we detected 8/10 (80%) positivity.

SARS-CoV-2 diagnosis
Among the total 30 samples, 10 from suspected and 20 from hospitalized COVID-19 cases and reported as ambiguous or negative by real time RT-PCR of routine diagnostics, we detected 25 positives and 5 negatives for SARS-CoV-2 by our analysis with ddPCR, according to the above criteria for N1 and N2 gene targets (Tables 1 and 2). After the ddPCR analysis on the RNA as

DISCUSSION
The detection of SARS-CoV-2 is of paramount importance not only for diagnosis, but also for decision related to the infection control [19]. It is well known that the ddPCR method can significantly reduce the proportion of false negative results compared to real time RT-PCR, in particular in specimens with a low viral load [5]. Using a more sensitive molecular method, SARS-CoV-2 presence was highlighted in several "difficult" samples, such as fluctuating discordant and also false negative results [19] [20] [21] [22]. As of today it is still not clear how to interpret this incongruous findings. For instance, it is not so rare the case in which a patient presenting to the emergency room with high clinical suspicion of COVID-19, initially presents a negative nasopharyngeal swab result that needs to be confirmed by repeated testing. This could be due to a delay in the recognition of symptoms by the patients, leading to a development of the disease in the lower respiratory tract, with very low colonization remaining in the upper respiratory tract at that point [23].
In the present study, we wanted to evaluate the performance of ddPCR technique in identifying SARS-CoV-2 positive samples with very low viral load and resulted negative (no amplification detected by RT-PCR) from hospitalized COVID-19 confirmed patients with fluctuating results. We evaluated the ddPCR on RNA and also as direct quantitation on the swab-derived material to assess a potential false-negativity of real time RT-PCR test. We performed the ddPCR using the CDCapproved primers/probes (2019-nCoV CDC ddPCR triplex probe assay, BioRad) and we identified a 65% positivity on extracted RNA. Furthermore, we observed that the direct quantitation on swab medium (without the RNA extraction) allowed the detection of virus also in some additional samples, detecting the SARS-CoV-2 signal on the 85% of the samples. We also evaluated some suspicious samples from asymptomatic not hospitalized subjects collected during a previous epidemiological study [14]. These suspicious samples were borderline at the real time RT-PCR and according to the CDC-approved COVID-19 diagnostic panel, a borderline Ct value >35 is not diagnostic and a confirmatory analysis is needed [3] [24]. Thus, the real time RT-PCR results from our suspected cases needed confirmation. We tested these samples using the ddPCR and the analysis on RNA allowed us to detect the 50% of the suspected cases as positive with a lower limit of Log10 copies/µL 2.55 and 3.29 for N1 and N2, respectively. In addition, using the direct ddPCR on swab, we revealed the virus in three additional samples.
Overall, comparing ddPCR applied to extract RNA and directly on the primary swab samples, we noted an improvement in sensitivity using the direct approach (85% and 80% of positivity, respectively for the hospitalized patients and for the suspected cases). These findings confirm our previous data [13] and show that the direct approach is more effective in detecting very low viral  [28]. However, this is a still a debated issue. Indeed, the virus growth was observed from biological specimens presenting Ct values =32 as well as in samples collected up to 22 days after the first positive diagnosis [19]. Nevertheless, the presence of replicative virus for very low viral load samples has not been demonstrated.

Limitations
Our study has some limitations: i) the small number of samples analyzed; ii) the use of only one primers/probes set from CDC by ddPCR, which could not represent all the possible panels used for diagnostics; iii) the lack of additional confirmatory testing of viral replication (i.e. sub-genomic targets [29] and in vitro culture).

CONCLUSIONS
In the present study, we confirmed not only the higher diagnostic potential of ddPCR for complementing real time RT-PCR in case of very low viral load samples, but also the improvement of ddPCR sensitivity using the direct approach. Indeed, our results support the use of the ddPCR especially for highly suspected patients with negative diagnostics results or for ambiguous samples.
These results provide valuable information to other clinical laboratories. The ddPCR and real time RT-PCR have similar time-costing workflow and costs for reagents (the pre-analytical phases until the cDNA synthesis, the same primers and probes used for PCR), although the ddPCR instrument cost is still high. Further studies to compare the efficiency of different primers need to be performed, in order to improve the diagnostic accuracy of SARS-CoV-2 detection by ddPCR.

Funding
This work was supported by the Italian Ministry of Health "Fondi Ricerca corrente -L1P6" to IRCCS Ospedale Sacro Cuore -Don Calabria.

Author contributions
MD and EP conceived and designed the analyses. GLM, ML and DT contributed to the collection of samples. AB, LM and GPC contributed to the collection of clinical data. MD performed the molecular experiments, MD and EP analyzed the data. MD and EP draft the paper. AM, CP and ZB contributed to the revision of the draft. All the authors read and approved the final manuscript.

Data availability
All data generated or analyzed during this study are included in this published article (and its supplementary information files).

Conflicts of Interest
The authors declare no conflict of interest.