SARS-CoV-2 Delta–Omicron Recombinant Viruses, United States

To detect new and changing SARS-CoV-2 variants, we investigated candidate Delta–Omicron recombinant genomes from Centers for Disease Control and Prevention national genomic surveillance. Laboratory and bioinformatic investigations identified and validated 9 genetically related SARS-CoV-2 viruses with a hybrid Delta–Omicron spike protein.

To detect new and changing SARS-CoV-2 variants, we investigated candidate Delta-Omicron recombinant genomes from Centers for Disease Control and Prevention national genomic surveillance. Laboratory and bioinformatic investigations identified and validated 9 genetically related SARS-CoV-2 viruses with a hybrid Delta-Omicron spike protein.
To rule out Delta and Omicron co-infection, laboratory contamination, and bioinformatic error, we examined the raw read data from the 9 candidate recombinants created from molecular loop and amplicon-based sequencing strategies. Two of these specimens were readily available from the original diagnostic laboratory, and extracted RNA was shipped to CDC for confirmatory sequencing. We used Illumina (https://www.illumina.com) and PacBio (https://www.pacb.com) sequencing of 2 whole-genome amplicon strategies, as well as spike-gene amplification followed by Oxford Nanopore (https:// nanoporetech.com) sequencing (Appendix). All sequencing strategies yielded functionally identical consensus sequences compared with the corresponding original sequencing strategies.
Nextclade (11) classified the 9 whole genomes as 21K (Omicron/BA.1). We then split the genomes at position 22150 (within the predicted recombination site range). Nextclade classified the first 22150 base fragment as clade 21J (Delta) and the remainder as clade 21K (Omicron/BA.1). Pangolin version 3.1.20 (pangoLEARN 1.2.123, Scorpio 0.3.16, https:// cov-lineages.org) assigned a lineage of none to the full-genome sequences. Pangolin classified the first 22150 base fragment of each recombinant as AY.43 (Delta), although the call was not supported by Scorpio. Inspection of this region revealed closer homology to AY.119.2 (Delta) sequences because of mutations orf1ab:A2855V and orf1ab:A6248S, which are common to AY.119 lineages, and orf1ab:K4176N, which is found in a subset of AY.119.2 (Delta) sequences. The remaining sequence fragment from nt 22151 to the 3′ end was classified by pangolin as BA.1.1 (Omicron). This observation has been documented in the PANGO-designations repository (https://github. com/cov-lineages/pango-designation/issues/439) and is under review for potential lineage assignment.
Detailed sequence analysis confirmed the 2 resequenced specimens as true recombinants and indicated no evidence of co-infection or contamination. Comparison with a representative AY.119.2 (Delta) specimen indicated characteristic Delta mutations (C21618G, C21846T, G21987A, and deletion 22029-22034) at >99% frequency (>600× coverage for Oxford Nanopore, >1,800× coverage for PacBio, >1000× coverage for Illumina) in the 5′ end of the recombinant (Appendix Figure 1, panel B). The 2 BA.1.1 (Omicron) deletions at the beginning of the spike gene (21765-21770 and 21987-21995) and the characteristic Omicron 9-base insertion after nt 22205 were not present in read data, consistent with a Delta origin for the 5′ end of the spike gene. After position 22577, the mutation profiles mirrored that of a representative BA.1.1 (Omicron) specimen (Appendix Figure 1, panel B). Analysis of individual Oxford Nanopore reads showed characteristic Delta mutations co-occurring with Omicron singlenucleotide variants on the same reads (sharing Delta 22029-22034 deletion and Omicron 22673 T>C; Appendix Figure 2). The translated spike protein is a hybrid, containing characteristic amino acids from both Delta and Omicron parents with a breakpoint between the N terminal domain and receptor-binding domain of spike S1 protein (Appendix Figure 1, panel A).
To visualize the parents of the recombinant genomes, we split all candidate recombinant genomes at position 22150, within the predicted breakpoint, and used Nextclade (11) to place each genome fragment (1-22150 and 22151 through the 3′ end) onto a reference tree. We visualized the 2 trees as a tanglegram tree with Auspice (12). Nucleotides 1-22150 clustered with clade 21J (Delta) sequences, and the remaining fragment of the genome clustered with 21K (Omicron/BA.1) (Appendix Figure 3).

Conclusions
Our results provide evidence of a recombinant SARS-CoV-2 genome containing a hybrid spike protein derived from a Delta (AY.119.2)-Omicron (BA.1.1) recombination event. However, the ability to effectively identify and confirm additional recombinant viruses remains challenging because of the range of sequence quality available in the public domain. These limitations are a result of amplification inefficiency and consensus-calling algorithmic error, as well as cases of co-infection or potential sample contamination.
Comparative phenotypic characterization of virus isolates from the recombinant cluster was not possible because all specimens were chemically inactivated. In the spike protein, there are no additional amino acid substitutions within the receptor-binding domain compared with BA.1.1 (Omicron) lineage viruses. Recombinant viruses with this hybrid spike protein were detected over the course of 6 weeks, but the number of cases resulting from those viruses remains low. Most cases were identified within the mid-Atlantic region of the United States. However, epidemiologic linkage cannot be determined because CDC does not collect identifying information for these samples.
Systematic virus surveillance is essential for longterm monitoring of SARS-CoV-2 evolution. Given the potential public health consequences of new variants emerging from recombination, investigations involving laboratory and bioinformatic components, such as the one presented here, are critical for correctly identifying and tracking these viruses.