A Recurrent Mutation at Position 26340 of SARS-CoV-2 Is Associated with Failure of the E Gene Quantitative Reverse Transcription-PCR Utilized in a Commercial Dual-Target Diagnostic Assay.

Control of the ongoing severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic requires accurate laboratory testing to identify infected individuals while also clearing essential staff to continue to work. At the current time, a number of quantitative real-time PCR (qRT-PCR) assays have been developed to identify SARS-CoV-2, targeting multiple positions in the viral genome. While the mutation rate of SARS-CoV-2 is moderate, given the large number of transmission chains, it is prudent to monitor circulating viruses for variants that might compromise these assays. Here, we report the identification of a C-to-U transition at position 26340 of the SARS-CoV-2 genome that is associated with failure of the cobas SARS-CoV-2 E gene qRT-PCR in eight patients. As the cobas SARS-CoV-2 assay targets two positions in the genome, the individuals carrying this variant were still called SARS-CoV-2 positive. Whole-genome sequencing of SARS-CoV-2 showed all to carry closely related viruses. Examination of viral genomes deposited on GISAID showed this mutation has arisen independently at least four times. This work highlights the necessity of monitoring SARS-CoV-2 for the emergence of single-nucleotide polymorphisms that might adversely affect RT-PCRs used in diagnostics. Additionally, it argues that two regions in SARS-CoV-2 should be targeted to avoid false negatives.

INTRODUCTION

Coronavirus disease of 2019 (n class="Disease">COVID-19) originated in Wuhan, China, in late 2019 (1, 2), generating a global pandemic (3). As of 22 May 2020, there have been close to 5 million confirmed cases and more than 300,000 deaths reported worldwide (4). Metagenomic RNA sequencing revealed that COVID-19 is caused by a novel coronavirus, subsequently named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). SARS-CoV-2 is a close relative of SARS-CoV and Middle East respiratory syndrome-CoV (MERS-CoV) (2), coronaviruses that have both been responsible for large outbreaks of respiratory illness within the last 2 decades (5, 6). The release of the first SARS-CoV-2 genome sequence on 10 January 2020 spurred the development of quantitative real-time PCR (qRT-PCR) assays (7–9) and thereby enabled reliable laboratory diagnosis of infections. In addition, protocols have been developed to allow for rapid sequencing of the SARS-CoV-2 genome (10), with sharing of the resultant data (11) and phylogenetic analysis (12, 13).

Laboratory testing for SARS-CoV-2 is a corene">nerstone of the strategy to mitigate its spread, as it facilitates the ideene">ntification aene">nd isolation of n class="Disease">infected individuals, while negative tests can allow essential personnel to continue to work (14). In the context of SARS-CoV-2, due to its high transmissibility (15), false negatives could have particularly adverse effects on efforts to control its spread. As qRT-PCR oligonucleotides rely on binding to small ∼20-bp regions, mutations in these targets have the potential to impair efficient amplification or probe binding, thereby generating false negatives. In contrast to other RNA viruses, coronaviruses have a moderate mutation rate due their ability to carry out RNA proofreading (16). Nevertheless, given the large number of ongoing transmission chains, it remains prudent to monitor the integrity of qRT-PCR assays.

Here, we report the identification of a single-nucleotide polymorphism (SNP) in the E gene of SARS-CoV-2 that is associated with the failure of the qRT-PCR that targets the E geene">ne iene">n the cobas n class="Species">SARS-CoV-2 test (Roche). As this dual-target assay also detects a region in ORF1b, these samples were still correctly identified as SARS-CoV-2 positive. This observation highlights the necessity of targeting two regions in SARS-CoV-2 RT-PCR assays and shows the role sequencing can play in resolving and anticipating problems with the qRT-PCR assays in use.

MATERIALS AND METHODS

RNA extraction and real-time PCR.

The study was approved by the Comité d’Ethique Hospitalo-Facultaire Universitaire de Liège (reference number CE 2020/137). COVID-19 detection was routinely performed using the cobas 6800 platform (Roche). For this, 400 μl of nasopharyngeal swabs in a preservative medium (Amies or UTM) were first incubated at room temperature for 30 min with 400 μl of cobas PCR media kit (Roche) for viral inactivation. Samples were theene">n loaded on the cobas 6800 platform using the cobas SARS-CoV-2 assay for the detection of the ORF1ab and E genes.

For qRT-PCR control and sequencing analysis, RNA was extracted from clinical samples (300 μl) on a Maxwell 48 device using the Maxwell RSC viral RNA kit (Promega) following a viral inactivation step using proteinase K according to the manufacturer’s instructions. RNA elution occurred in 50 μl RNase-free water, aene">nd 5 μl was used for the RT-PCR. Reverse traene">nscription aene">nd RT-PCR were performed on a LC480 thermocycler (Roche) based on the Cormaene">n et al. (9) protocol for the detection of n class="Gene">RdRP and E genes using the TaqMan fast virus 1-step master mix (Thermo Fisher). Primers and probes (Eurogentec, Belgium) were used as described by the authors (9).

SARS-CoV-2 whole-genome sequencing.

Reverse transcription was carried out using SuperScript IV VILO master mix, and 3.3 μl of RNA was combined with 1.2 μl of master mix and 1.5 μl of H2O. This was iene">ncubated at 25°C for 10 miene">n, 50°C for 10 miene">n, aene">nd 85°C for 5 miene">n. PCRs used the primers aene">nd conditions recommeene">nded iene">n the n class="Species">nCoV-2019 sequencing protocol (17). Primers from version 3 of the Artic Network were used and were synthesized by Integrated DNA Technologies. Samples were multiplexed using the Oxford Nanopore native barcoding expansion kits 1 to 12 and 13 to 24, in conjunction with a ligation sequencing kit. Sequencing was carried out on a Minion using R9.4.1 flow cells. Data analysis followed the nCoV-2019 novel coronavirus bioinformatics protocol of the Artic network (17). The resulting consensus viral genomes have been deposited at the Global Initiative on Sharing All Influenza Data (GISAID) (11).

Sanger sequencing.

Reverse transcription was carried out as described above. The primers nCoV-2019_87_LEFT aene">nd n class="Species">nCoV-2019_87_RIGHT from the Artic Network nCoV-2019 amplicon set (17) were used to amplify the regions between positions 26198 and 26590. The resultant PCR product was purified using Ampure XP beads (Beckman Coulter), sequenced using a BigDye Terminator cycle-sequencing kit (Applied Biosystems), and run on an ABI PRISM 3730 DNA analyzer (Applied Biosystems).

Phylogeny.

SARS-CoV-2 geene">nomes and the associated metadata were downloaded from GISAID (https://www.gisaid.org/) on 25 May 2020. Viral geene">nomes marked by GISAID as complete (>29,000 bases) and high coverage (<1% Ns, <0.05% unique amino acid mutations, and no insertion/deletions unless verified by submitter) were selected, leaving 20,386 viral geene">nomes. We also downloaded viral geene">nomes using a less stringeene">nt cutoff, requiring the virus to be complete (>29,000 bases) and excluding viruses with low coverage (>5% Ns); in this case, 29,699 viral geene">nomes remained. Viruses carrying a variant at position 26340 were ideene">ntified with SeqKit (18) using the following grep command and motif eene">ncompassing the variant (underlined): “seqkit grep -s -i -p TTACACTAGCTATCCTTACTG.” The viruses containing the variant were added to the list of viruses to include in the Nextstrain build. Viruses from nonhuman hosts were excluded from the analysis. Nextstrain phylogenetic trees were generated for both data sets using the default configuration (https://github.com/nextstrain/ncov).

The SARS-CoV-2 genomes were assigned to a lineage via pangolin (https://github.com/hCoV-2019/pangolin), which used the virus nomenclature proposed by Rambaut et al. (19).

RESULTS

Real-time PCR.

The cobas system (Roche) implements a dual-target assay to detect SARS-CoV-2, with qRT-PCRs targetiene">ng both the n class="Gene">ORF1ab region and the E gene (see Fig. S1 in the supplemental material). During the course of routine SARS-CoV-2 testing, we observed eight samples that were negative for the E gene qRT-PCR but positive for the ORF1ab qRT-PCR (Table 1). These represented 0.2% of the SARS-CoV-2 positive samples we identified between 23 March and 25 May 2020 (during this time period, 25,619 tests were carried out on the system, and 3,398 [13.3%] were positive). Four of these samples were retested using the Corman et al. (9) SARS-CoV-2 assay that targets the RdRP and E genes. In this instance, both the RdRP and E gene qRT-PCRs were positive in all four samples (Table 1). All came from Belgian health care workers in the same service, with sampling dates that ranged between 23 March and 17 April 2020 (Fig. 1A). As the samples were positive for the ORF1ab qRT-PCR, all samples were correctly classified as positive by the cobas system (Roche).

TABLE 1

Real-time PCR C values observed in the 8 patients

Patient ID	CT result from:
	cobas		Corman et al. (9)
	RdRp	E gene	RdRp	E gene
ULG-0184	21.53	0
ULG-10088	26.65	0	28.66	26.51
ULG-10089	22.64	0	25.02	22.85
ULG-10095+	24.62	0	23.73	22
ULG-10095*	31.94	0
ULG-10096	20.45	0	22.89	20.33
ULG-10169+	28.64	0
ULG-10169*	28.55	0
ULG-10145	27.19	0
ULG-10146	25.44	0

Two patients were tested twice, 2 weeks apart: +, initial test; *, 2-week test. Four of the samples were also tested using the SARS-CoV-2 assay of Corman et al. (9); these values are in boldface. C, cycle threshold.

FIG 1

(A) Screen shot from IGV (Integrative Genomics Viewer; http://software.broadinstitute.org/software/igv/home) shows the VCFs (lines represent SNPs) for the eight viruses as well as a BAM file showing the reads and coverage for one virus. The sampling date for each virus is indicated. Six of the viruses have identical patterns of SNPs and are underlined. The zoomed-in section shows the SNP at 26340, and the blue rectangle labeled E_Sarbeco_P1 corresponds to the region covered by the E gene probe in Corman et al. (9). (B) Sanger sequencing of four samples carrying the C-to-U transition at position 26340. The top chromatogram shows a virus carrying the wild-type sequence. (For convenience, U is displayed as T.)

Two patients were tested twice, 2 weeks apart: +, iene">nitial test; *, 2-week test. Four of the samples were also tested usiene">ng the n class="Species">SARS-CoV-2 assay of Corman et al. (9); these values are in boldface. C, cycle threshold.

We speculated that these samples carried a common variant that interfered with the E gene qRT-PCR and carried out whole-genome sequencing of the viruses using the Artic Network protocol (17). The consensus genomes generated showed six individuals to be infected with a geene">netically ideene">ntical virus (Fig. 1A). The remaiene">niene">ng two viruses shared the same SNPs as the previous six but had accumulated additional mutations (suggestiene">ng contiene">nued spread of the liene">neage iene">n the area). In two cases we also had a 2-week follow-up sample from the same n class="Species">patient; in each case, the consensus viral genomes generated were identical (Fig. S2). The six identical viruses (derived from different patients) deviated from the MN908947.3 reference isolated in Wuhan at only three positions (Fig. 1A). The first two SNPs were toward the 5′ end of the virus at positions 1,440 and 2,891, respectively. The third SNP, a C-to-U transition at position 26340, is within the E gene of the virus and was validated by Sanger sequencing in four samples (Fig. 1B). This SNP overlaps the E gene probe used in the Corman et al. (9) RT-PCR assay; however, as was mentioned above, it does not appear to affect the performance of this assay in our hands. Unfortunately, the position of primers and probes utilized in the cobas E-gene assay (Roche) are not publicly available, nevertheless it is parsimonious to assume that this SNP is the cause of the failure of the E-gene qRT-PCR implemented in the cobas system.

Out of the 229 SARS-CoV-2 geene">nomes we have sequeene">nced at the time of writiene">ng, eight carry the SNP at position 26340. To see if the same variaene">nt was circulatiene">ng more widely, we examiene">ned the n class="Species">SARS-CoV-2 sequences deposited in GISAID for a variant at the same position. When only complete, high-coverage genomes are considered (20,386 genomes), 18 were found to carry a C-to-U transition at position 26340 (0.09%). Eight of these were sequenced by us and seven were isolated in England, two in Switzerland, and one in Turkey. As can be seen in Fig. 2, viruses isolated in the same country cluster together; however, they do not cluster with other viruses carrying the SNP at position 26340. We also classified the viral genomes according to the nomenclature proposed by Rambaut et al. (19). Table 2 shows that samples isolated in the same country belong to the same lineage, with no overlap in lineage between countries. As a consequence, it appears that this variant has arisen multiple times in different transmission chains (homoplasic site).

FIG 2

(A) Phylogenetic tree generated by Nextstrain (12) with the viruses carrying the U allele at 26340 highlighted in yellow. The six viruses with identical patterns of SNPs are underlined. (B) Expanded view of the eight Belgian samples. (C) Expanded view of the seven English samples.

TABLE 2

Lineages of the high-coverage viruses using the scheme of Rambaut et al. (19)

Virus name	Lineage	Virus name	Lineage
Switzerland/GE0304/2020	B.1.5	Belgium/ULG-0184/2020	B.3
Switzerland/GE2453/2020	B.1.5	Belgium/ULG-10088/2020	B.3
England/SHEF-C02F7/2020	B.2.1	Belgium/ULG-10089/2020	B.3
England/SHEF-C04F1/2020	B.2.1	Belgium/ULG-10095/2020	B.3
England/SHEF-C053A/2020	B.2.1	Belgium/ULG-10096/2020	B.3
England/SHEF-C0691/2020	B.2.1	Belgium/ULG-10145/2020	B.3
England/SHEF-C088C/2020	B.2.1	Belgium/ULG-10146/2020	B.3
England/SHEF-CE1DE/2020	B.2.1	Belgium/ULG-10169/2020	B.3
England/SHEF-D14DF/2020	B.2.1	Turkey/HSGM-8992/2020	B.4

Finally, we relaxed the filtering of viral genomes, selecting genomes of >29,000 bases in length and with less than 5% Ns (we no longer required the virus to be classified as high coverage). This added 9,313 genomes (29,699 in total) and revealed eight additional viruses carrying a C-to-U transition at 26340 (0.09%) (Fig. S3). Of these, six were isolated in England, four clustered with the previous English samples, while the other two fell in different parts of the tree. Of the remaining two viruses, one was isolated in Australia and the second was sequenced in Luxembourg. Interestingly, the Luxembourg virus clustered with the samples identified by us and was assigned to the same B.3 lineage, suggesting it is part of the same cluster of infections.

DISCUSSION

As the positions of the primers and probes used in the cobas (Roche) E gene qRT-PCR have not been disclosed to us upon request, we cannot definitively conclude that the C-to-U transition at position 26340 of the SARS-CoV-2 geene">nome causes the failure iene">n the E geene">ne qRT-PCR iene">n the n class="Species">patients examined. However, given the available data, causality appears likely. The cobas E gene qRT-PCR may use a primer-probe combination that is more sensitive to the presence of the SNP than the Corman et al. (9) E gene assay. Alternatively, it may target the same positions, but differences in reagents used and cycling conditions may prevent binding of the probe in the presence of the SNP.

The E gene qRT-PCR implemented in the cobas assay is intended to facilitate pan-Sarbecovirus detection (20). However, a U is found at the same relative position in the SARS-CoV-1 MA15 isolate P3pp5 (Geene">nBank accession no. FJ882961.1) and in the bat coronavirus Cp/Yunnan2011 (GenBank accession no. JX993988.1). This suggests that similar variability occurs at this position in other coronaviruses, which could impair the effectiveness of the assay for pan-Sarbecovirus detection.

It should be stressed that despite the failure of the E gene qRT-PCR in these patients, the cobas assay correctly called these iene">ndividuals as positive for n class="Species">SARS-CoV-2 due to the ORF1ab qRT-PCR. This highlights the prudence of targeting more than one position in the viral genome in a diagnostic assay. The Corman et al. (9) protocol recommends the use of its E gene assay as a first-line screening tool, with confirmatory testing using the RdRp gene assay (9). This SNP does not affect the Corman et al. (9) E gene qRT-PCR in our hands; however, our results highlight how a mutation in the virus can generate a false negative in a single qRT-PCR. In most cases these mutations will be rare; however, as our examination of the GISAID data have shown, such mutations have the potential to arise independently in separate transmission chains.

Recently, Vogels et al. (21) examined the efficiency as well as frequency of variants impacting a number of the qRT-PCRs commonly used for SARS-CoV-2 testing. They found a number of variants that fell within the primer and probe binding sites, with the majority preseene">nt at a low frequeene">ncy and involving only a single base. A promineene">nt exception involved a GGG→AAC mutation at geene">nome positions 28881 to 28883 that overlaps the first three bases of the 5′ eene">nd of the Chinese CDC N geene">ne forward primer (7). This mutation is found in approximately 25% of the viruses on GISAID (accessed 25 May 2020). As the Chinese CDC assay also includes an ORF1ab qRT-PCR, viruses carrying this variant will still be detected, even if this variant impairs the N gene qRT-PCR. Nevertheless, given the high frequency of this variant, it would appear prudent to avoid using this qRT-PCR primer.

This work shows the danger of relying on an assay targeting a single position in the viral genome. It also highlights the utility of combining testing with rapid sequencing of a subset of the positive samples, especially in cases where one of the qRT-PCRs fails. The sequencing allowed us to pinpoint the likely reason behind the failure of the E gene qRT-PCR. The identification of viruses carrying additional mutations as well as the clustering of the Luxembourg virus with the Belgian viruses also suggests that only a fraction of the virus carrying this variant came to our attention. This emphasizes that while the variant is at a low frequency globally, at the local level it could be much higher. This example shows that it remains prudent to continue monitoring viral genomes for variants that can negatively impact this and other diagnostic assays. Finally, it would be preferable if manufacturers were transparent about the primer and probes used, as this would allow problematic variants to be more readily identified from the available viral sequences.