Length-dependent motions of SARS-CoV-2 frameshifting RNA pseudoknot and alternative conformations suggest avenues for frameshifting suppression

Conserved SARS-CoV-2 RNA regions of critical biological functions define excellent targets for anti-viral therapeutics against Covid-19 variants. One such region is the frameshifting element (FSE), responsible for correct translation of viral polyproteins. Here, we analyze molecular-dynamics motions of three FSE conformations, discovered by graph-theory analysis, and associated mutants designed by graph-based inverse folding: two distinct 3-stem H-type pseudoknots and a 3-way junction. We find that the prevalent H-type pseudoknot in literature adopts ring-like conformations, which in combination with 5′ end threading could promote ribosomal pausing. An inherent shape switch from “L” to linear that may help trigger the frameshifting is suppressed in our designed mutant. The alternative conformation trajectories suggest a stable intermediate structure with mixed stem interactions of all three conformations, pointing to a possible transition pathway during ribosomal translation. These observations provide new insights into anti-viral strategies and frameshifting mechanisms.


21
In less than two years, COVID-19 with its novel infectious 22 agent SARS-CoV-2 has already caused more than 266 mil-    Figure 1: Secondary structures of the three FSE conformations at different lengths we study, along with their arc plots and corresponding dual graphs. For 77-nt, the three conformations 3 6 pseudoknot, 3 3 pseudoknot, and 3 5 junction have common Stems 1 (blue) and 3 (green), while different Stem 2 (red). The two pseudoknots are classified as H-type, 14 where in 3 6 the loop region of Stem 1 binds with the external single-stranded 3 ′ end, and in 3 3 the Stem 1 loop binds with the 5 ′ end. For 87-nt, 10 upstream residues are added that include the 7-nt slippery site, and the 3 3 conformation contains an extra flanking stem SF (grey). For 144-nt, 37 upstream and 30 downstream residues are included, and extra stems (grey) are formed, including attenuator hairpin AH for 3 6 and SF for 3 3. Stems are represented as vertices in dual graphs, and loops as edges, with the central 3 6, 3 3, and 3 5 submotifs corresponding to the 77-nt FSE region highlighted in red, and the flanking vertices/edges corresponding to the extra stems/loops in grey. assign this pseudoknot motif as dual graph 3 6 ( Fig. 1)  S2/3 junction S1/2 junction Pseudoknot-stabilizing hydrogen bonds in 77-nt 3_6 Pseudoknot-stabilizing hydrogen bonds in 144-nt 3_6 Non-threaded S1 5 strand and 87-nt 5 end Threading of S1 5 strand and 87-nt 5 end S1 3 strand S2/3 junction S1/3 junction Ring 77-nt 3_6 Threaded Ring Conformation 77-nt 3_6 Non-threaded Ring Conformation to the other two conformations at 77-nt (Fig. S3, S4). A the "L" and the linear shape, via bending of Stem 3 (Fig. 7).

291
The pseudoknot complex (Stems 1 and 2) and the ring con- ing more substantially (Fig. S8). region for all lengths (Fig. 7). The unpaired 3 ′ end also ex- indicate that Stem 1 is the strongest, followed by Stem 3, 303 and lastly by Stem 2 (Fig. S9, S10). stretching caused by the bending of 3 ′ end and Stem 3 loop 306 (Fig. 7, Fig. S11). In this motion, Stems 1 and 2, especially 307 triplets that contain interactions from all three Stem 2 (pur-308 ple and red residues in Fig. 7), are stable and move in unison.

309
That these triplets are not transient suggests that they may be 310 part of the structural transition among alternative conforma-311 tions, as discussed above.

312
Comparing to 3 6, we see a higher RMSF peak value in the    linear shape (Fig. 8, Fig. S13). The major difference oc- base pairs to 6-7 (Table 1, Fig. S17). The three stems then 386 have similar sizes (Fig. 8). Stem 2 is no longer held around 387 Stem 3, but instead extends as a third helical arm. Coaxial 388 stacking of Stems 1 and 2, as well as a tilting motion of these 389 two stacked stems, are observed (Fig. S18).  (Fig. 8). The combined insights suggest three anti-viral 405 intervention avenues and a mechanism for frameshifting that 406 links our three alternative conformations (Fig. 9).

421
The second approach is to strengthen the 5 ′ end threading impede ribosomal translation (Fig. 9).

435
The third approach is to target the 3 6 pseudoknot structural 436 switch between an "L" shape (coaxially stacked Stems 1 and 437 2 and an extruding Stem 3) and a linear shape (vertical stack-438 ing of the 3 stems), revealed by our PCA analysis (Fig. 7).

439
In the mRNA-ribosome Cryo-EM structure captured during from the "L" to linear shape, residues in the Stem 2/3 junc-446 tion are exposed (Fig. 7); small molecules like MTDB 10,51 447 can thus block the switch and hamper frameshifting (Fig. 9).

448
Another option is to deploy our 3 6 mutant, which assures a 449 stabilized linear shape (Fig. 8).      Table S1). For 3D structure prediction 549 programs that gave multiple structures as output, the first 550 structure that retained the correct motif was selected for MD