Five (5) selective nasopharyngeal swab specimens collected from non-hospitalized patients with respiratory infection, which were confirmed to be true-positive for SARS-CoV-2 Omicron variant by Sanger sequencing, were further analyzed by bidirectional Sanger sequencing of 3 genomic targets to show the effects of minor subvariant sequences with multi-allelic SNPs on sequencing-based variant diagnosis. Three (3) of these samples belonging to the Omicron BA.2 sub-lineage were collected in April, 2022, one belonging to the Omicron BA.4/BA.5 was collected in June, 2022 and one (1) belonging to the BA.1 sub-lineage was collected in January, 2022. Written consents were obtained from the 4 patients whose samples were positive for BA.2 or BA.4/BA.5 subvariant to allow their samples to be further analyzed for publication. The sample positive for BA.1 subvariant was commercially supplied with independent IRB certification.

As previously reported [25] [27], the cellular pellet derived from about 1 mL of the nasopharyngeal swab rinse along with 0.2 mL supernatant after centrifugation was first digested in a buffered solution containing sodium dodecyl sulfate and proteinase K. The digestate was extracted with phenol. The nucleic acid was precipitated by ethanol and re-dissolved in 50 µL of diethylpyrocarbonate-treated water.

The primary and nested RT-PCR conditions were described in detail previously [25] [26] [27]. Briefly, to initiate the primary RT-PCR, a total volume of 25 µL mixture was made in a PCR tube containing 20 µL of ready-to-use LoTemp^® PCR mix with denaturing chemicals (HiFi DNA Tech, LLC, Trumbull, CT, USA), 1 µL (200 units) of Invitrogen SuperScript III Reverse Transcriptase, 1 µL (40 units) of Ambion^TM RNase Inhibitor, 0.1 µL of Invitrogen 1 M DTT (dithiothreitol), 1 µL of 10 µmolar forward primer in TE buffer, 1 µL of 10 µmolar reverse primer in TE buffer and 1 µL of sample RNA extract. The ramp rate of the thermal cycler was set to 0.9˚C/s. The program for the temperature steps was set as: 47˚C for 30 min to generate the cDNA, 85˚C 1 cycle for 10 min, followed by 30 cycles of 85˚C 30 sec for denaturing, 50˚C 30 sec for annealing, 65˚C 1 min for primer extension, and final extension 65˚C for 10 minutes.

The nested PCR mixture was a 25 μL volume of complete PCR mixture containing 20 μL of ready-to-use LoTemp^® mix, 1 μL of 10 μmolar forward primer, 1 μL of 10 μmolar reverse primer and 3 μL of molecular grade water.

To initiate the nested PCR, a trace (about 0.2 μL) of the primary PCR products was transferred by a micro-glass rod to the complete nested PCR mixture. The thermocycling steps were programmed to 85˚C 1 cycle for 10 min, followed by 30 cycles of 85˚C 30 sec for denaturing, 50˚C 30 sec for annealing, 65˚C 1 min for primer extension, and final extension 65˚C for 10 minutes.

The crude nested PCR products showing an expected amplicon at agarose gel electrophoresis were subjected to automated Sanger sequencing without further purification.

About 0.2 µL of the nested PCR products was transferred by a micro-glass rod into a Sanger reaction tube containing 1 μL of 10 μmolar sequencing primer, 1 μL of BigDye^® Terminator (v 1.1/Sequencing Standard Kit), 3.5 μL 5× buffer, and 14.5 μL molecular-grade water in a total volume of 20 μL for 20 enzymatic primer extension/termination reaction cycles according to the protocol supplied by the manufacturer (Applied Biosystems, Foster City, CA, USA).

After a dye-terminator cleanup, the Sanger reaction mixture was loaded in an Applied Biosystems SeqStudio Genetic Analyzer for sequence analysis. Sequence alignments were performed against the standard sequences stored in the GenBank database by on-line BLAST. The sequences were also visually analyzed for nucleotide mutations and indels.

Seven (7) sets of primary RT-PCR primers and their corresponding nested PCR primers, which were used to generate the nested PCR products to be used as the templates for Sanger sequencing, are listed in Table 1. To maintain detection sensitivity, the maximum size of the primary RT-PCR amplicon was limited to 530 bp for routine diagnostics.

In Table 1, the PCR primers were grouped according to the nested PCR amplicons that they collectively generated. The N gene C04/C03 nested PCR amplicon was used as the sequencing template to verify the presence of a SARS-CoV-2 genomic nucleic acid in a given sample. The S gene RBD S9/S10 and S gene NTD SB12/SB8 nested PCR amplicons were used as the templates for Sanger sequencing to detect the key mutations in the RBD and NTD for the diagnosis of variants (or subvariants), including the BA.2 and BA.4/BA.5 subvariants. The SB7/SB8 nested PCR primers pair was effective in amplifying the NTD segment for all variants except those of the BA.2 and BA.4/BA.5 lineages. The S gene NF3/NR4 nested PCR amplicon was used as an alternative to detect RBD mutations in case significant mutations involving the S9 and S10 primer-binding sites leading to an RBD RT-PCR amplification failure. The last two sets of primers were designed to amplify the junctional region between the RBD and the NTD of the S gene, using the 214EPEins sequence (GAGCCAGAA) as a discriminator

in the 3’ end of the forward primers to selectively amplify the BA.1 sequence in case there was a co-existing BA.1 RBD sequence in a sample containing a dominant Omicron BA.2 subvariant sequence.

Since crude nested PCR products were used for DNA sequencing, the sample used for Sanger reaction might contain multiple templates, including a dominant allele sequence and a number of minor subvariant sequences with multi-allelic SNPs.

All the data presented in this paper except Figure 19 were detected and analyzed in Milford Molecular Diagnostics Laboratory by the author.

When the SARS-CoV-2 nucleic acid samples containing little minor subvariant sequences with multi-allelic SNPs are sequenced for variant determination, the diagnostic test is straightforward. Bidirectional Sanger sequencing of an RT-PCR amplicon of the RBD and an RT-PCR amplicon of the NTD of the S gene are adequate for accurate molecular diagnosis of a SARS-CoV-2 Omicron BA.2 or a BA.4/BA.5 subvariant.

As demonstrated in Section 3.1., 11 - 14 key amino mutations in the RBD of the Omicron variant can be detected in one forward unidirectional sequencing of a 445-bp PCR amplicon. A unidirectional NTD sequencing can verify the absence

of A67V, Δ69-70, T95I, and Δ143-145 mutations or the presence of Δ24-26 and A27S mutations for further confirmation of a BA.2 or a BA.4/BA.5 sub-lineage. However, some samples positive for an Omicron variant may contain a large number of subvariant sequences with multi-allelic SNPs that may cause uncertain or questionable base calling in Sanger sequencing. One of such examples is illustrated by the sequencing data on a nasopharyngeal swab specimen collected on 25 April 2022 (patient W). The electropherogram of a forward sequencing of the RBD is presented in Figure 9.

In order to prove that the multicolored low peaks in the electropherogram

presented in Figure 9 were reproducible in their specific positions of the sequence and not random sequencing noise artefacts, aliquots of the nucleic acid extract used to generate the electropherogram of Figure 9 were reamplified by 3 sets of nested RT-PCR for re-sequencing the N gene, the RBD and the NTD of the S gene in one single run to reduce possible sequencing artefacts introduced by between-run technical and reagent variations. The electropherograms of the repeated bidirectional Sanger sequencing showed no unambiguous sequences in the N gene segment and in the NTD segment for the diagnosis of an Omicron BA.2. However, the presence of minor subvariant sequences may affect correct base calling as demonstrated below.

In a nasopharyngeal swab specimen collected on 3 April 2022 from a symptomatic adult patient in Connecticut, U.S.A., identified as sample M22-75, bidirectional

sequencing of the N gene nested RT-PCR products demonstrated a 398-base SARS-CoV-2 N gene sequence with R203K and G204R mutations. However, bidirectional Sanger sequencing of the S gene RBD and the NTD nested RT-PCR products revealed that the sample actually contained an Omicron BA.2 variant with two different NTD sequences. One of them be-longs to the BA.1 sub-lineage and the other to the BA.2 sub-lineage. Since an Omicron BA.2 subvariant containing a co-existent BA.1 S gene NTD mutation profile has not been reported in the world literature, the relevant sequencing findings are presented as follows.

Unexpected sequence mutation in primer-binding sites is a well-known cause of

PCR failure in nucleic acid-based test for SARS-CoV-2 [32]. But the mechanism leading to failure of the NTD amplification may be different from that responsible for the RBD RT-PCR failure in one sample.

As previously reported, testing a group of nasopharyngeal swab samples collected in October, 2020 in the U.S. prior to the emergence of any variant of concern by sequencing a 398-base N gene showed that there is a hypervariable 79-base stretch of nucleotide sequence corresponding to the position 28821 to 28899 (GenBank reference sequence NC_045512.2) [26]. This 79-base region is flanked by two segments of conserved sequence, which were used to design the PCR primers (referred to as Co1, Co8, Co4 and Co3 in Table 1) for the diagnostic N gene-sequencing assay. The N gene of the Omicron variant strains invariably harbors the R203K and G204R mutations. But these two mutations were already sporadically reported in nasopharyngeal swab specimens collected in the early part of 2020 [42] long before the appearance of the Omicron variant. Therefore, variant determination based on N gene sequencing alone is not reliable. Nevertheless, compared to the S gene RBD and NTD, this N gene target is more conserved, especially in the primer-binding sites, and is associated with a much lower level of homeologous minor sub-variant sequences with multi-allelic SNPs, which may suppress RT-PCR amplification [33] [34].

After the SARS-CoV-2 has circulated for more than 2 years from population to population since its outbreak, numerous mutated subvariant sequences with multi-allelic SNPs [9] have been accumulated in the virus and show up in some of the emerging variant isolates displaying genetic diversity within single infected hosts [43] [44]. The number of these minor subvariant sequences with multi-allelic SNPs in some clinical samples positive for Omicron may be extremely high. Most surveillance studies relying on testing high-viral-load samples with NGS only consider a single consensus sequence for each infected person for statistic purpose and ignore the co-existing subvariant sequences. Bioinformatic analysis of the NGS data is often based on alignment, or mapping, of reads against a reference sequence followed by the consensus extraction by majority voting. If the studied virus sequence is divergent from the chosen reference sequence, the reads covering the regions of divergence could not be aligned correctly or might be discarded, which will bias the resulting consensus [45]. It has been reported that the key sites of the S gene known to harbor mutations of interest, such as the K417N/T, E484K and N501Y in the RBD, are in the regions of low read coverage when NGS is used and may need target Sanger sequencing to recover [46]. For diagnostic purpose, routine target Sanger sequencing of the RBD is a better approach to determine variants [18]. However, Sanger sequencing can generate a dominant sequence for accurate mutation analysis only if the level of minor subvariant sequences with multi-allelic SNPs in the sample is not high enough to suppress the PCR amplification of the dominant target sequence to be used as the sequencing template or to interfere with the Sanger sequencing process. The failure of RT-PCR amplification of a target RBD or NTD sequence may be due to mutations affecting the primer-binding sites in the SARS-CoV-2 genome or due to the presence of an over-whelming amount of minor subvariant sequences with multi-allelic SNPs. If a target PCR amplicon band is visualized at agarose gel electrophoresis, the failure to generate a readable electropherogram may be mitigated by performing a bidirectional sequencing (Figure 10 and Figure 11). If no S gene target PCR amplicon is visualized at gel electrophoresis when the N gene sequencing of the sample is positive for SARS-CoV-2, a new PCR primer set for the S gene amplification may be required as illustrated in Figure 24 and Figure 25, and in Figure 26 and Figure 27.

Target Sanger sequencing can reveal important information that NGS misses, as demonstrated in Section 3.3.

As a whole, Omicron subvariants have a high number of mutations in the spike protein gene and these mutations mainly occur in the NTD and RBD of the S gene [47]. Determination of the mutations in the NTD and RBD in any given sample usually generates enough information for accurate diagnosis of the key Omicron subvariants. However, there are exceptions. For example, in one specimen, M22-75, Sanger sequencing showed S371F, S373P, S375F, T376A, D405N, R408S, K417N, N440K, S477N, T478K, E484A, Q493R, Q498R, N501Y and Y505H (Figure 12 and Figure 13)-the 15 typical mutations in the RBD for a BA.2 subvariant, but A67V, Δ69-70, T95I, G142D, and Δ143-145 (Figure 14 and Figure 15)-a profile of mutations in the NTD characteristic of a BA.1 subvariant. In addition, there were 3 single nucleotide mutations, one nucleotide deletion and one overlapping single-nucleotide polymorphism (SNP), causing 3 novel mutations of F65L, T114N, and Q115K in the NTD and converting the wildtype base “A” to “G” to create a new nonsynonymous N74S mutation in a competing allele sequence.

After next-generation sequencing of a split sample M22-75 yielded a consensus sequence of Omicron BA.2 (Figure 19), repeated bidirectional Sanger sequencing of the original sample confirmed the previous finding of a BA.1 NTD sequence with all the mutations and deletion (Figure 17 and Figure 18). In addition, a set of new SB12/SB8 PCR primers were used to generate templates for Sanger sequencing to confirm that the M22-75 sample contained a BA.2 NTD sequence (Figure 20 and Figure 21) and a BA.2 RBD (Figure 23), but did not harbor a BA.1 RBD sequence (Figure 22). Therefore, the co-existence of a BA.1 NTD and a BA.2 NTD in sample M22-75 was not due to co-infection by an Omicron BA.1 and an Omicron BA.2 subvariants, as reported by others [48].

In other words, on sample M22-75 target partial Sanger sequencing has demonstrated that one host was infected by at least 3 Omicron subvariants, which share one BA.2 RBD sequence, but harbor 3 different S gene NTD sequences, namely a typical BA.2 NTD sequence, an atypical BA.1 NTD sequence and an atypical BA.1 NTD sequence with an extra N74S mutation. The two atypical NTD sequences with 3 novel amino acid mutations do not fully match with any SARS-CoV-2 genomic sequences annotated in the Gen-Bank database (Figure 16), and as allele sequences their clinical significance remains unknown. However, there is no direct sequencing evidence that these BA.1 NTD sequences are in fact connected to a BA.2 RBD in one single PCR amplicon of about 1,500 bp in size. Demonstration of such connection in one template would be the irrefutable proof for a BA.1/BA.2 S gene recombination, but is difficult to accomplish in diagnostic RT-PCR with patient samples.

Another challenge of using routine Sanger sequencing to determine variants is the possibility of detecting undescribed mutations, such as L84I, which changes the 84th amino acid from leucine to isoleucine, as demonstrated in an Omicron BA.4/BA.5 subvariant (Figure 7 and Figure 8). Since an L84I mutation has not been reported in any known SARS-CoV-2 variants, it is uncertain if these isolates should be grouped under the BA.4/BA.5 subvariant, or as a new Omicron subvariant.

Accurate diagnostic methods for determination of the mutations in the RBD and NTD of the S gene of the SARS-CoV-2 are needed in selecting therapeutics for COVID-19 patients and in evaluation of the transmissibility of the infecting virus. For examples, the current standard care in antiviral treatment for moderate to severe COVID-19 includes the use of the monoclonal antibody combination REGN10933 (casivirimab) and REGN10897 (imdevimab) [49]. However, in certain Omicron subvariants, the K417N, E484A, S477N, and Q493R mutations would lead to loss of electrostatic interactions with REGN10933 whereas a mutation of G446S would lead to steric clashes with REGN10987 [50], causing neutralization escapes [51]. The Q493R and Q498R mutations are known to introduce additional electrostatic interactions with ACE2 residues Glu35 and Asp38, respectively, whereas S477N enables hydrogen-bonding with ACE2 Ser19. Collectively, these latter mutations strengthen ACE2 binding, and could be a factor in the enhanced transmissibility of Omicron relative to previous variants [49]. Deletions of NTD amino acid sequences, such as Δ69-70, Δ141-144 and Δ146 are known to be associated with immune escape in certain patients because these deletions may hinder NTD recognition by neutralizing antibodies from convalescent plasma [20] [52].