A stark difference in the profiles of defective viral transcripts between SARS-CoV-2 and SARS-CoV.

Dear Editor,

Recently, Fantini et al. reported that mutations in the N-terminal (NTD) and receptor binding (RBD) domains of the SARS-CoV-2 spike protein act synergistically to optimize virus infection. However, genomic variants beyond the coding region of spike protein is poorly understood, especially the large structural variants within a single or between closely related coronaviruses.

SARS-CoV and SARS-CoV-2 are two closely related β coronaviruses that caused the global pandemics of SARS and COVID-19 in 2003 and 2019, respectively. Despite similarities in their receptor, tropism, and clinical manifestations, the two viruses demonstrate a drastic difference in their transmissibility, which remains largely unexplained. Defective viral transcripts are known to attenuate the replication of a parental virus by competing translational machinery in host cells.

To determine whether the two viruses produce a differential profile of defective transcripts, we performed direct RNAsequencing (dRNA-seq) of the transcripts derived from the Vero E6 cells infected with SARS-CoV, SARS-CoV-2 or MERS-CoV for 24 h as we did previously , . We chose this sequencing technology for its potential to generate ultra-long read. Notably, dRNA-seq generates reads from 3′ to 5′ end and depends on the presence of poly(A) tail at 3′ end, meaning that all reads produced with dRNA-seq will carry a poly(A) tail. We generated approximately 2.88, 0.86 and 2.04 million reads for the SARS-CoV, SARS-CoV-2 and MERS-CoV samples, respectively, with N50 sizes of approximately 1.9, 2.5 and 2.2 k nucleotides (nts), respectively (Table S1, Fig. S1). Approximately 87.6%, 84.6% and 59.8% of the total reads were viral reads in the samples infected with SARS-CoV, SARS-CoV-2 and MERS-CoV, respectively. We focused mainly on the full-length reads that carried both a leader at the 5′ end and a poly(A) tail at the 3′ end (Fig. 1 A). These reads represented 12.2%, 26.3% and 11.2% of the total viral reads of SARS-CoV and SARS-CoV-2, respectively. Mapping of these reads allowed us to unambiguously determine the global structure of both DVGs and dsgRNAs, because the reads without a leader may be subject to the uncertainty in terms of the absence of an unknown portion at the 5′ end.

Fig. 1

Full-length dRNA-seq read coverage reveals precise boundaries of sgRNAs. (A) Schematic representation of genomic and subgenomic organizations of SARS-CoV-2 RNAs and their defective formats. Full-length genomic, subgenomic RNAs and UTRs are depicted in scale and are differentially color coded. Name of ORF and non-structural protein (nsp) are indicated. Positions of TRS-B and TRS-L inferred from the presence of core TRS are indicated with red arrow. A magnified view of 5′ end (leader and UTR) and subgenomic RNAs are shown below. Also shown on the left bottom are two types of defective viral transcript, including DVG and dsgRNA. Color codes for ORF, nsp and structural proteins are used throughout. (B) Shown on the top is the coverage of full-length reads carrying a leader and a poly(A) tail derived from dRNA-seq of sense-strand RNA. Note the precise punctuation between the read coverage and existing ORFs except ORF 6. Also note the two imprecise coverage drops within the ORF of nsp1–3 (indicated with arrowhead). Shown on the bottom is the coverage of all antisense reads derived from dRNA-seq with a poly(A) tailing step. Note the precise punctuation between the read coverage and the existing ORFs and a sharp jump in the antisense leader. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

As expected, the coverages of the full-length reads demonstrated a precise demarcation that coincided precisely with the predicted ORF boundaries for all structural proteins except for the ORFs 6 and 10 (Fig. 1B). We were surprised to find that although few reads that contained a leader were mapped to the start of ORF 6, approximately 6000 reads were precisely mapped to the internal part of ORF M (Fig. S2), indicating that the ORF 6 sgRNAs do not begin from their own start codon, but from the transcriptional regulatory sequence (TRS) within the ORF M, which is consistent with the previous findings, in which the core sequence was identified but the global structure of ORF 6-specific sgRNAs was not clear , .

We were able to recover two distinct categories of defective viral transcripts that contain both a 5′ leader and a 3′ poly(A) tail. The first category carries an extended 5′ and 3′ end sequence separated by a large deletion, which is referred to as defective viral genome (DVG). The second category consists of defective subgenomic RNAs (dsgRNAs) that lack various portions of the 3′ co-terminal end (Fig. 1A). Strikingly, we found that full-length profiling of viral transcripts revealed a stark difference in the profiles of sgRNAs between SARS-CoV and SARS-CoV-2. Despite the high abundance of dsgRNAs produced by SARS-CoV and MERS-CoV, few dsgRNAs were detected in SARS-CoV-2 (Fig. 2 A-G, Figs. S3 and S4). dsgRNAs were observed for all structural proteins with various abundances in the case of SARS-CoV. For example, dsgRNAs accounted for as many as 60% of all full-length reads mapped to the ORF S that bear the longest full-length sgRNA, and dsgRNAs accounted for approximately 25% of all full-length reads mapped to the ORF N that bear the shortest full-length sgRNA (Fig. 2A-G). Apparently, the longer the ORF, the greater the abundance of dsgRNA, which suggests poor processivity of the RNA-dependent RNA polymerase (RdRp) of SARS-CoV in the generation of long transcripts. Sequencing of the antisense RNAs of SARS-CoV-2 revealed that the abundance of antisense RNAs was approximately 1000 times lower than that of sense RNAs, thus indicating the great efficiency of sgRNA synthesis.

Fig. 2

A stark difference in the profiles of defective viral transcripts between SARS-CoV-2 and SARS-CoV. (A-G) Abundant defective sgRNAs in SARS-CoV but not in SARS-CoV-2. (A-C) Coverage of ORF-specific reads from SARS-CoV-2 for S (A), 3a (B) and N (C), respectively. Diagrams showing the full-length sgRNA are depicted on the top. (D-F) Coverage of ORF-specific reads from SARS-CoV for S (D), 3a (E) and N (F), respectively. Diagrams showing the both the full-length and defective sgRNA are depicted on the top. (G) Quantification of full-length and defective sgRNAs for each ORF in SARS-CoV and SARS-CoV-2. (H-K) Selective retention of full-length ORF N in the DVGs of SARS-CoV. (H-I) Shown are the coverages of DVGs that contain approximately 10k nts of sequences at the 5′ end and various parts at the 3′ end in SARS-CoV-2 (H) and SARS-CoV (I). (J) and (K) 5′ and 3′ junctions of the DVGs in (H) and (I) respectively. Curved lines represent the 5′ and 3′ locations of the junctions in SARS-CoV-2 (J) and SARS-CoV (K).

In addition to the dsgRNAs, our full-length transcript profiling revealed the presence of DVGs in both SARS-CoV and SARS-CoV-2 (Fig. 2H-K, Fig. S4). The 5′ UTR extended into the ORF 1ab up to the ORF of nonstructural protein (nsp) 3 in DVGs. These DVGs lacked most coding regions, including those encoding the remaining nsps, and those encoding various ORFs of structural proteins at their 3′ ends (Fig. 1A). One important finding was that the two viruses demonstrated a significant difference in their 3′ ends. Specifically, nearly a half of the junction site between the 5′ and 3′ parts of DVGs began precisely at the beginning of the ORF N in the case of SARS-CoV and MERS-CoV (Fig. 2I, K, and Fig. S4B), whereas such a bias was absent in the DVGs produced by SARS-CoV-2 (Fig. 2H and J). These results indicate a differential DVG-generation mechanism between SARS-CoV-2 and other coronaviruses. Sequencing data also showed that overall SARS-CoV-2 transcripts had substantially longer poly(A) tails than SARS-CoV transcripts (Fig. S5A), and this was also the case in terms of DVGs (Fig. S5B) and dsgRNAs (Fig. S5C). Functional relevance of the shorten poly(A) tail warrants further investigation.

In summary, our dRNA-seq results indicate that SARS-CoV-2 has evolved a unique capability to generate full-length sgRNAs but has lost the ability to retain the full-length ORF of N in its DVGs, which may have implications for its transmissivity. The extremely low abundance of antisense strand of SARS-CoV-2 genome makes these RNAs an ideal target for development of inhibitory agents.

Ethics

Ethical approval was not required for this service evaluation and audit of practice.

Declaration of Competing Interest

No conflicts of interests declared by an author.