2.1. Evaluation of Specific PCR Approaches for the Detection of SARS-CoV-2 RNA
In several laboratories, non-specific PCR products in both SARS-CoV-2-negative patient samples and in the non-template control (NTC) using the WHO-recommended SARS-CoV E-gene-specific PCR [
6] have been reported [
8] (personal communication). During a surveillance pool testing of SARS-CoV-2 using RNA prepared from routine respiratory samples [
9], we observed discordant results for E-gene and RdRP-gene qPCR analysis (
Supplementary Table S1). We then individually analyzed the pool sets in order to identify samples giving positive signals (
Table S1). In total, four out of the 48 (8.33%) tested swab samples showed a positive signal in the E-gene PCR, which, however, could not be confirmed in the RdRP-specific PCR. In order to confirm and further characterize these non-specific products in the negative samples, we carried out additional PCR-based analyses from the four E-gene-positive pharyngeal swab samples (Pat. 20–23,
Table S1) with the WHO-recommended qRT-PCR protocol using E (
Figure 1a)- and RdRP (
Figure 1b)-gene-specific primers (
Table 1). As a positive control, we used RNA extracted from a throat swab of one mildly symptomatic passenger (Pat.1) returning from Wuhan after initial quarantine (
Figure 1,
Table 2), described previously [
4]. Again, all four samples showed a positive signal in the E-gene PCR; however, the confirmatory RdRP-gene PCR was negative except for the positive control (
Figure 1b). These data indicate that using the two-step WHO PCR protocol might complicate the interpretation of the test results, especially for cases with higher Cq-values.
Since the RdRP-gene PCR is more specific than the E-gene PCR but is also described as less sensitive [
8], we aimed to develop a novel approach using specifically designed primer and probe pairs. For optimal primer design, consensus sequences were aligned to the reference SARS-CoV-2 genome sequence and analyzed for primers binding in the E, N, Orf1, M and S regions (
Table 1,
Figure S1, Table S3) that would allow SARS-CoV-2 but not SARS-CoV amplification. For experimental confirmation, seven different CoV-2 isolates were propagated and characterized for this study (
Table 2). Briefly, patient samples were passaged in Caco-2 cells as described previously [
10], and the progeny viruses (FFM1–7) were purified from culture supernatants for subsequent PCR and next generation sequencing (NGS) analysis. In addition, we included RNA from the human SARS-CoV strain Frankfurt 1 (NC_004718) to the analysis. We then compared the PCR performance using serial dilutions; however, the Cq values for the E- and N-gene-specific primers revealed limited linearity and thus were excluded from further analysis (
Figure S1a). The S- and M-gene-based PCRs were more sensitive than Orf-1 (
Figure S1b) and proved to be suitable for linear SARS-CoV-2 RNA detection, including for samples with low viral loads (~Cq 40), while only exceptionally high SARS-CoV samples were cross-reactive with high Cq values (
Figure S1c). Since the M-gene PCR was superior with an approximately Cq 1.73 +/− 0.26 earlier detection (Cq 1.48 +/− 0.15 for SARS-CoV,
Figure S1d), we continued with the M-gene PCR and further characterized the limit of detection.
In order to further assess the M-gene PCR’s efficiency, the M- and RdRP-gene-specific PCR primers and probe binding sites were analyzed. The alignment of over 4300 full-length SARS-CoV2 genomes including FFM1–7 isolates revealed single nucleotide polymorphisms (SNPs) in 13 different positions for RdRP (total 0.33%) and eight positions for the M-gene (total 0.61%) within the primer/probe binding sites (
Figure S3, Table S4). Among seven, a C/U mismatch within the probe binding site at Position 27,046 of the reference genome was the most common SNP with a 0.43% frequency. At the time of the initial submission of this manuscript, analysis with 165 full-length genomes revealed a 100% identity for both primers, and only one isolate of 165 (0.6%) had a mismatch to the consensus sequence C/U. Beside this particular locus, the M-gene region seems to be less polymorphic compared to the RdRP region, and thus, this may overall result in better PCR efficiency.
Using plasmid DNA constructs that harbor the conserved SARS-CoV-2 amplicon sequences, we generated standard curves and compared the M-gene PCR (
Figure 2b) with the established WHO RdRP-gene PCR approach (
Figure 2a). Using the same dilution series of plasmid samples, the M- gene-based method allowed earlier detection (
Figure 2c). To determine the analytical sensitivity of the M-gene-based approach, we additionally used in vitro-transcribed RNA standards and tested four replicates to determine the limit of detection (
Figure 2d,e,g). To rule out the possibility that different RNA preparations might cause a bias in the comparison, we have also generated an in vitro-transcribed RNA standard that harbored both target sequences. Using an in vitro diagnostic (IVD)-certified test kit (Roche, Basel, Switzerland)
, we confirmed the M-gene PCR to be more sensitive than RdRP (
Figure 2f). SYBR green-based melting curve analysis revealed a melting point at 80 °C for all the samples (
Figure S2). To further characterize the capacity of the M-gene PCR to specifically detect high viral loads, we performed SYBR green-based PCR and quantified high loads of virus in cell culture supernatants (
Figure S2d).
In conclusion, our M-gene-based qRT-PCR detection of SARS-CoV-2 RNA was at least as specific as the RdRP PCR recommended for confirmation by the WHO but showed a significantly higher sensitivity. Importantly, non-specific signals, as observed in the E-gene PCR, were not detected.
2.2. Detection of SARS-CoV-2 in Clinical and Research Samples Using E-, RdRP-, and M-Gene-Specific Protocols
In order to validate our method, we re-tested clinically relevant samples that had been qualitatively tested as positive for SARS-CoV-2 RNA during routine diagnostics (
Table S2). WHO-recommended RdRP primer pairs were used for confirmation. As negative controls, we included eight negative samples (
Figure 1,
Table S1). As described above, non-specific E-gene amplicons were detected in a test kit-specific manner, possibly due to the reagents used [
8]. Therefore, we additionally compared the performance of two research kits (New England Biolabs, Ipswich, MA, USA) and one in vitro diagnostic (IVD) certified test kit (Roche Viral Multiplex RNA Kit) using M-gene and RdRP-gene primers. Overall, we observed lower Cq values with all three kits for the M-gene when compared to the RdRP-gene PCR (
Figure 3. Using the Luna OneStep Probe kit (Luna Universal Probe One-Step RT-qPCR Kit, NEB), the M-gene PCR was significantly more sensitive with all the tested kits, with a difference of approximately 10 Cq values with the Luna Universal Probe One-Step RT-qPCR Kit in comparison to PCR with the RdRP-specific primer pairs. With the SYBR green-based kit (Luna Universal One-Step RT-qPCR Kit) and Roche IVT kit (LightCycler
® Multiplex RNA Virus Master), we observed four and one Cq value difference(s), respectively, confirming that the detection of the M-gene is superior to the RdRP-gene PCR.
To further evaluate whether the newly developed method is also suitable for the detection of intracellular virus RNA, we infected Vero and Caco-2 cells with SARS-CoV-2 isolate FFM1 (MT358638). Using M-gene PCR, we were able to detect very low copy numbers of viral RNA in infected Vero cells, while RdRP-gene PCR was limited in sensitivity (
Figure 4a). Furthermore, we infected Caco-2 cells, generated an intracellular replication curve, and monitored viral replication and genome copy numbers, respectively. Viral RNA was detectable at the time points of 3, 6, 12 and 24 h post infection, with a linear trend (
Figure 4b,c).
In conclusion, non-specific PCR products and limited sensitivity pose a major problem when using the WHO protocol for SARS-CoV-2 detection and could lead to numerous unnecessary confirmation tests. Our newly developed PCR protocol is suitable for saving time and resources since pre-screening is no longer necessary. Thus, this approach might be used as a cost-effective alternative to the E- and RdRP-based protocol for research purposes and for diagnostics in resource-limited settings.