Noncanonical junctions in subgenomic RNAs of SARS-CoV-2 lead to variant open reading frames

SARS-CoV-2, a positive-sense RNA virus in the family Coronaviridae, has caused the current worldwide pandemic, known as coronavirus disease 2019 or COVID-19. The definition of SARS-CoV-2 open reading frames is a key step in delineating targets for vaccination and treatment for COVID-19.

Here, we report an integrative analysis of three independent direct RNA sequencing datasets of the SARS-CoV-2 transcriptome.

We find strong evidence for variant open reading frames (ORFs) encoded by SARS-CoV-2 RNA. A variant transcript for the matrix protein (M) lacking its N-terminal transmembrane domain, initiated by a TTG start codon, is produced by a strong transcriptional regulatory sequence (TRS)-mediated junction within the M ORF and represents up to 19% of all M ORFs. Sporadic non-canonical junctions in the spike (S) ORF lead to N-terminal truncations that remove the N-terminal and receptor-binding domains from up to 25% of S ORFs. Surprisingly, nearly all ORFs from ORF1a identified in these transcriptome sequences were variant. These ORFs contain the first 200-800 amino acids of ORF1a and may represent a mechanism to regulate the relative abundance of ORF1a nonstructural proteins.

We show there is strong transcriptome and junctional support for variant ORF1a ORFs in independent direct RNA sequencing and short-read RNA sequencing datasets, and further show that up to 1/3 of these ORFs are expected to have C-terminal fusions with downstream genes. Finally, we show that currently unannotated ORFs are abundant in the SARS-CoV-2 transcriptome. Together, these analyses help to elucidate the diverse coding potential of SARS-CoV-2.