The RNA Architecture of the SARS-CoV-2 3'-Untranslated Region

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for the current COVID-19 pandemic. The 3' untranslated region (UTR) of this β-CoV contains essential cis-acting RNA elements for the viral genome transcription and replication. These elements include an equilibrium between an extended bulged stem-loop (BSL) and a pseudoknot. The existence of such an equilibrium is supported by reverse genetic studies and phylogenetic covariation analysis and is further proposed as a molecular switch essential for the control of the viral RNA polymerase binding. Here, we report the SARS-CoV-2 3' UTR structures in cells that transcribe the viral UTRs harbored in a minigene plasmid and isolated infectious virions using a chemical probing technique, namely dimethyl sulfate (DMS)-mutational profiling with sequencing (MaPseq). Interestingly, the putative pseudoknotted conformation was not observed, indicating that its abundance in our systems is low in the absence of the viral nonstructural proteins (nsps). Similarly, our results also suggest that another functional cis-acting element, the three-helix junction, cannot stably form. The overall architectures of the viral 3' UTRs in the infectious virions and the minigene-transfected cells are almost identical.

Keywords: 3′ UTR; COVID-19; DMS; DMS-MaPseq; DREEM; SARS-CoV-2; ShapeKnots; minigene; pseudoknot; three-helix junction.

Figure A1

The workflow of DMS-MaPseq in this study. (a) BSR T5/7 cells were transfected with SARS-CoV-2 minigene for in vivo DMS-MaPseq. Infectious virions were collected for in-virion DMS-MaPseq. (b) Adenines and cytosines in the unpaired region of the transcriptome have a higher probability to be methylated by DMS. (c) The DMS-modified RNA was extracted. (d) The A- or C-methylated RNA can induce single-point mutations at the methylated nucleobase during the reverse transcription. Amplicons of 895 and 1291 base pairs were amplified by PCR for in-virion and in vivo minigene DMS-MaPseqs, respectively. (e) The band was purified by agarose gel electrophoresis. (f) The amplicons were ligated with Illumina sequencing adapters and sequenced in at a paired-end mode (150 bp × 2). (a,e,f) are generated by BioRender.com.

Figure 1

(a) The SARS-CoV-2 minigene construct containing the 5′ and 3′ untranslated regions (UTRs) and a short poly(A) tail. (b) The amplicon location (highlighted in yellow) in the SARS-CoV-2 genome for the in-virion DMS-MaPseq. The arrows indicate the primer-binding sites for the amplicons.

Figure 2

(a) The in vivo minigene, in-virion, and in vitro normalized dimethyl sulfate (DMS) activity (ΔnDMS) profiles of the SARS-CoV-2 3′ UTR. Compared to the in vivo minigene, the significantly higher or lower ΔnDMS activities are labeled in red and green, respectively (>95% confidence in Z-factor test and >1.5-fold ΔnDMS activity change, see Table S1 for details). The activity data are not available for the primer-binding regions, i.e., nts 319–327, 302–327, and 311–327 in the in vivo minigene, in-virion, and in vitro DMS-MaPseqs, respectively. (b) The correlations between the in vivo minigene and in-virion or in vitro ΔnDMS activities. The out-of-range data points are omitted in the figure. The in vitro DMS-MaPseq was re-analyzed using the data from Ref. [12] with the ShapeMapper2 pipeline (see Methods).

Figure 3

Arc plots of the most possible base-pairing pattern for SARS-CoV-2 3′ UTR. A colored arc represents a base pair with the following probability: 100% > green ≥ 80% > blue ≥ 30% > yellow ≥ 10%. By definition, ΔnDMS is high when ΔnDMS > 0.85, medium 0.40 < ΔnDMS ≤ 0.85, or low ΔnDMS ≤ 0.40, and the corresponding nucleotide symbol is colored red, orange, or black, respectively. The nucleotides of G and U, or within PCR primer-binding regions, where the DMS data is not available, are colored grey. The arc plots were generated by the SuperFold software package [26] from the experiments (a) in vivo minigene DMS-MaPseq, and (b) in-virion DMS-MaPseq. See Figure S2 for full arc plots with all possible base-pairing possibilities.

Figure 4

A consensus structure derived from the in-virion and in vivo minigene DMS-MaPseqs. The ΔnDMS reactivity for each nucleotide from the in vivo minigene DMS-MaPseq is colored according to the legend on the RNA backbone, with the in-virion DMS activity drawn aside if it belongs to a different reactivity group. The occurrences of the base pairing for BS-B–L1 and S3-A–S3-B (dotted lines) are not supported in this study.

Figure 5

(a) Two predominant conformations of bulged stem-loop (BSL) and stem-loop 1 (SL1) uncovered by the DREEM analysis on the DMS-MaPseq reads that cover >95% of the region. The illustrated structures were derived from replicate 2 data set. (b) Correlation in ΔnDMS activity profiles for the same structural cluster between the replicates and between two clusters.

Figure 6

Comparison of the 3′ UTR sequences among SARS-CoV-2 (RefSeq NC_045512.2), SARS-CoV (RefSeq NC_004718), MERS-CoV (NC_019843), BCoV (U00735), and MHV (NC_048217). Major RNA elements, BSL, SL1, and HVR, are boxed and annotated in black. Shorter RNA segments are annotated in blue above a bar aligned with the sequences. Conserved nucleotides are shaded in cyan. 717 nucleotides at the 5′ of MERS 3′ UTR are omitted for clarity.