To knot or not to knot: Multiple conformations of the SARS-CoV-2 frameshifting RNA element

Figures

• 

  Figure 1:. The FSE sequence and three relevant 2D structures for the SARS-CoV-2 84 nt frameshifting element (residues 13462–13545) emerging from this work that combines 2D structure prediction, SHAPE structural probing, and thermodynamic ensemble modeling. The −1 frameshifting alters the transcript UUU-UU*A(Leu)-AAC(Asn)-GGG at the second codon (asterisk) to backtrack by one nucleotide and start as AAA-CGG(Arg) instead, so that translation resumes at CGG. At top is the FSE sequence, with the attenuator hairpin region and the 7 nt slippery site highlighted and A13533 labeled (C in SARS). The ORF1a end and ORF1b start codons for the overlapping regions are marked. For each 2D structure, H-type 3_6 pseudoknot, HL-type 3_3 pseudoknot, and three-way junction 3_5 (unknotted RNA), corresponding dual graphs, 2D structures, and corresponding arc plots are shown, with color coded stems and loops labeled.

• 

  Figure 2:. Multiple sequence alignments (MSA) of coronavirus frameshifting elements found by the Infernal covariance model42 shown for 16 top-scored sequences among 182 unique homologues. Arrows at top and bottom illustrate the FSE expansion from 77 to 84 nt (+7 slippery site nt), 144 nt (+30 nt on both ends), 156 nt (+12 upstream nt), and 222 nt (+66 upstream nt). Sixteen top scored coronaviruses are aligned with the SARS-CoV-2 222 nt FSE region (insertions are hidden), with sequence similarities shown for both the whole genome and the FSE region. Nucleotides are colored based on sequence conservation. The consensus sequence is written below with a sequence logo (at each position, the overall stack height indicates sequence conservation level, and the height of an individual letter within indicates the relative frequency of that nucleotide). Stems are marked based on our analysis here: black for Stems 1 and 3, red/green/purple for Stem 2 of 3_6/3_3/3_5, consistent with Fig. 1, and grey for Alternative Stem 1 (AS1) and upstream stems. The covarying base pairs detected by R-scape43 are colored by nucleotide identity: green A, blue U, orange C, and red G.

• 

  Figure 3:. Predicted optimal structures for the frameshifting element using PKNOTS, NUPACK, IPknot, and ProbKnot (see text). For each program, 4 different sequence lengths are used: 77, 84, 144, and 156 nt. The common 77 nt subsequence is aligned, the slippery site is colored orange, and the attenuator hairpin AH is magenta. The predicted structures are shown as arc plots, with Stems 1 and 3 in black, and Stem 2 of 3_6, 3_3, and 3_5 in red, green, and purple, respectively. An upstream hairpin that blocks 3_3 Stem 2 and Alternative Stem 128–31 are labeled UH and AS1, respectively. Corresponding dual graphs 3_6 (red), 3_3 (green), 3_5 (purple), and 2_1 (black) are highlighted as graphs or subgraphs of larger motifs.

• 

  Figure 4:. SHAPE reactivity analysis for SARS-CoV-2 frameshifting element for 77 and 144 nt (Replicate 1). (A) The SHAPE reactivity for the 77 nt construct is plotted by bars, with red/yellow/black representing high/medium/low reactivity. The arc plot at top shows the dominant 3_6 pseudoknot predicted by the ShapeKnots energy landscape (98% of conformational space), and at bottom is the minor 3_5 (2% of landscape). Stems are labeled, and the Gibbs free energy (kcal/mol) and Boltzmann distribution probabilities are given. (B) SHAPE reactivity and ShapeKnots predictions for 144 nt construct. (C) Reactivity differences between the two constructs are shown for two enlarged key regions highlighted in A,B, with positive/negative differences indicating less flexibility in the 144/77 nt construct. Base pairs in the 144 nt 3_3 conformation are plotted by arcs at top, and 77 nt 3_6 or 3_5 at bottom. Critical residues for reactivity comparisons are highlighted.

• 

  Figure 5:. Design and SHAPE analysis for (A) 77 and 144 nt 3_6 pseudoknot-strengthening mutant (PSM), (B) 77 nt 3_3 PSM, and (C) 77 nt 3_5 Mutant. For each mutant, we show the design flow, where we use RAG-IF and multiple 2D structure prediction program screening to determine mutations that stabilize the 3_6 pseudoknot, 3_3 pseudoknot, or 3_5 junction. Mutations are highlighted in blue. The SHAPE reactivity bar plots, and arc plots of the structure predicted by ShapeKnots, with alternative Stem 2 positions are shown. See Fig. S8 for reactivity differences between the mutants and the wildtype for two boxed key regions.

• 

  Figure 6:. Conformational landscape of the frameshifting element for different sequence lengths predicted by ShapeKnots using reactivities from the 144 nt construct. For each length, probabilities of all structures containing independently folded 3_6 or 3_3 pseudoknots are individually summed. The compositions are colored red (3_6) and green (3_3), respectively. (Top) Landscape for adding upstream nucleotides only to the 77 nt FSE (asymmetric expansion). The optimal sequence length of 87 nt for the 3_3 pseudoknot is in dashed black. (Bottom) Landscape for adding both upstream and downstream nucleotides to the 77 nt FSE (symmetric approach). At 90 nt (87 + 3 downstream nt) the landscape is almost all 3_3.

• 

  Figure 7:. Effects of 5 mutations tested for frameshifting efficiency by Bhatt et al.36 on 2D structure predictions of the 77 nt and 87 nt frameshifting element. (A) (Left) The 77 nt FSE 3_6 pseudoknot with mutation regions labeled in blue. Two weak base pairs for Stem 2 are indicated using dotted lines. (Right) A table showing 2D prediction results for the wildtype and the mutants. The upper half is for 4 programs: 3_6, 3_5, and 3_3 predictions are in red, purple, and green, respectively, with their corresponding Stem 2 lengths (in bold if structure change). The lower half is for ShapeKnots, showing probabilities of 3_6, 3_5, and 3_3. (B) (Left) The 87 nt FSE 2D structure by ShapeKnots, a 4_21 structure made of the 3_3 pseudoknot and a flanking Stem SF, with mutation regions in blue. (Right) A table showing prediction results using 4 programs and ShapeKnots.

• 

  Figure 8:. Three-dimensional models of the 87 nt 3_6 and 3_3 pseudoknot. Initial 3D structures are predicted by RNAComposer,63 Vfold3D,64 SimRNA,65 and iFoldRNA,66 and subjected to 1–1.5 μs MD using Gromacs.67 The last 500 ns are used for clustering analysis, and the most populated cluster center by RNAComposer (red)/iFoldRNA (green) is shown here for each system. The 88 nt cryo-EM 3_6 structure derived by Zhang et al. (blue)29 (PDB: 6XRZ) is aligned using Rclick68 for comparison. The three shaded stems of the cryo-EM structure align well with our 3_6 model.

• 

  Figure 9:. Three avenues for frameshifting interference, with cartoon models for the tertiary systems as modeled by molecular dynamics62 (created with BioRender.com).