Viruses | Free Full-Text | Structural Impact of the Interaction of the Influenza A Virus Nucleoprotein with Genomic RNA Segments

Not all RNA molecules spontaneously adopt their functional conformation in cells and viral particles, and some of them require the assistance of proteins to do so. By analogy with protein chaperones, RNA chaperones were initially defined as proteins that bind transiently and non-specifically to RNA and resolve kinetically trapped, misfolded conformers [41]. However, any protein that shows activity in any of the assays designed to test RNA chaperone activity [41] is usually considered an RNA chaperone even though its binding to RNA is not transient. Over time, numerous structural [31,33,42,43] and non-structural [32,44] viral proteins have been considered RNA chaperones, even though most of them do not spontaneously dissociate from RNA. The effect of such RNA chaperones can be assessed by comparing the RNA structure before the addition of the protein and after the removal of the protein by SDS or/and proteinase K treatment [27,32,42].

We thus analyzed the structure of the NS and M vRNAs before the addition of NP (No-NP condition, see Section 2) and after the removal of the protein by proteinase K treatment (ProtK condition) by SHAPE using NMIA [45,46]. NMIA modifies the ribose of flexible unpaired nts [45,46,47], and SHAPE reactivities can be implemented as pseudo-energies to improve RNA secondary structure predictions [48,49]. Mean SHAPE reactivity values of the NS and M vRNAs under the NoNP and ProtK conditions were obtained from highly correlated triplicate experiments (median = 0.89, range = 0.67–0.97) (Data S1 and Table S1).

SHAPE reactivity values of RNA were used to define three broad categories: nts with low reactivity (0.8), usually unpaired and thus present in apical loops, internal loops, bulges or single-stranded junctions; nts with intermediate reactivity values that can be part of unstable secondary structures or structured loops [36,48,49,50].

We thus first compared the reactivity of NS and M vRNAs under the NoNP and ProtK conditions (Table 2).

Table 2 shows that 79.7 and 75.4% of nts of the NS and M vRNAs, respectively, belonged to the same reactivity category under the NoNP and ProtK conditions; the reactivity of 18.0 and 22.8% of nts of the NS and M vRNAs, respectively, differed by one category (from intermediate to low or intermediate to high or vice-versa), and the reactivity of 2.3 and 1.8% of nts of the NS and M vRNAs, respectively, went from low to high or vice-versa. Notably, 9 and 37 nts of the NS vRNA and 31 and 50 nts of the M vRNA were highly reactive uniquely under the NoNP or ProtK condition, respectively (Table 2).

This analysis suggests that while the NS and M vRNAs both adopt similar structures under the NoNP and ProtK conditions, these structures differ to some degree. To test this hypothesis, we used the SHAPE reactivity data as constraints to model the secondary structures of these vRNAs using RNAStructure [49], revealing that most local secondary structure elements are common to the two structures (Figure S1a,b). Indeed, both structures share 13 stems capped by an apical loop, while two are unique to the NoNP structure, and one only exists solely in the ProtK structure (compare Figure S1a,b).

In order to confirm this, we compared the predicted NoNP and ProtK secondary structures of vRNA NS using RNAStructViz [39] (Figure 2). This analysis indicated that 550 and 530 nts out of the 890 nts of the NS vRNA are base-paired under the NoNP and ProtK conditions, respectively. Of these, 436 form identical base pairs under both conditions. Remarkably, almost all short-range interactions are maintained in the NoNP and ProtK conditions (Figure 2a). By contrast, the predicted long-range interactions that maintain the overall secondary structure of the NS vRNA drastically differ in the NoNP and ProtK conditions (Figure 2a).

We performed a similar analysis on the M vRNA. When we used the SHAPE data as constraints to model the secondary structure of the M vRNA under the NoNP and ProtK conditions, we observed that the predicted structures are more different than in the case of the NS vRNA (Figure S2). The two M vRNA structures share 13 stems capped by an apical loop in common, but three are unique to the NoNP structure, and seven only exist in the ProtK structure (Figure S2). Comparison of the NoNP and ProtK structures of the M vRNA using RNAStructViz [39] indicated that 618 and 604 nts out of the 1027 nts of the M vRNA are base-paired under the NoNP and ProtK conditions, respectively. Of these, 378 form identical base pairs under both conditions. These results indicated that the NP induced more permanent structural rearrangements in the M vRNA than in the NS vRNA (Figure 2). Moreover, unlike in the NS vRNA, NP induced mainly short-range and intermediate-range structural rearrangements in the M vRNA, while the long-range interactions were less affected (Figure 2b).

Next, we checked whether the differences in the proposed secondary structure models between the NoNP and ProtK conditions correlate with local SHAPE reactivity differences. To that aim, we reported the most pronounced reactivity changes on the secondary structures predicted under the NoNP condition for the NS and M vRNAs (Figure 3a and Figure 3b, respectively).

In the case of the NS vRNA, most increased reactivities in the ProtK condition compared to the NoNP condition were observed in single-stranded nts (e.g., in the 549–563, 580–592, 718–720, 733–745, 778–782, and 857–874 internal loops and the 613–620, 652–659, 786–796, 819–826 apical loops), whereas diminished reactivities were mainly observed in stems (e.g., in stems 250–255/276–281, 362–369/467–474, 480–485/520–525, 487–491/514–518, 497–502/507–513, and 663–666/729–732) (Figure 3a). These differences do not support the idea that the NS vRNA adopts different structures under the two conditions but suggest that most local structures are the same under both conditions (although the 3D structures might be different). However, a few nts that form stems in the NoNP secondary structure had a higher reactivity under the ProtK condition, namely the 23–25/646–648, 39–47/593–601, 196–199/218–222, 333–336/538–541, 341–347/528–534, 689–691/702–704 and 804–808/835–839 stems (Figure 3a), suggesting that these stems do not exist under the latter condition.

Indeed, the first five of these helices, which correspond to long-range interactions (except for the 196–199/218–222), were not predicted when the ProtK SHAPE data were used as constraints for secondary structure modeling (Figure S1 and Figure 2a). Despite their high relative SHAPE reactivity under the ProtK condition compared to NoNP, the two short-range stems 689–691/702–704 and 804–808/835–839 were predicted to exist under both conditions (Figure S1 and Figure 2a). Nevertheless, the ProtK model of vRNA NS does not contradict the experimental data, as the absolute SHAPE reactivities of these regions remained low and consistent with base-pairing under the ProtK conditions (Figure S1b).

Similarly, we observed that under the ProtK conditions some of the enhanced reactivities in the M vRNA conditions were located in unpaired regions (e.g., 529–532, 584–593, 869–877), whereas some diminished reactivities were observed in stems (e.g., in stems 128–133/146–151, 280–284/292–295, 358–364/455–461, 470–477/482–489, 570–577/598–606, and 617–627/817–827) (Figure 3b). However, compared to the NS vRNA, a greater proportion of increased reactivities in the M vRNA under the ProtK condition were observed in stems of the NoNP secondary structure (compare Figure 3b with Figure 3a). Moreover, in the M vRNA, several diminished reactivities under the ProtK conditions were located in regions that are unpaired in the NoNP secondary structure model (Figure 3b). Thus, a direct comparison of the SHAPE data suggested that the NP protein has a more pronounced RNA chaperone activity on the M vRNA than on the NS vRNA (Figure 3), consistent with the comparison of the secondary structure models predicted using the SHAPE data as constraints (Figure S2 and Figure 2). A comparison of the predicted secondary structure models of the M vRNA M the NoNP and ProtK conditions shows that three regions are largely remodeled under the ProtK conditions: (1) the region corresponding to stems 492–496/1001–1005, 46–53/347–354, 58–64/336–342, and 248–254/316–322 and the single stranded junctions between these helices, (2) the region between nts 568 and 832, and (3) the region encompassing nts 917–942 and 507–521 (Figure S2 and Figure 2). In each of these regions, at least one helix is destabilized under the ProtK condition compared to NoNP (Figure 3b), and the secondary structure models of the M vRNA exhibit no local disagreement with the SHAPE reactivity data (Figure S2), supporting the validity of our conclusions.

Next, we used NMIA to study the RNA chaperone-independent structural impact of NP on the vRNA structure. To that aim, we probed the vRNA/NP complexes (Comp condition) and compared the SHAPE values obtained under this condition with those obtained under the NoNP and ProtK conditions. The Comp SHAPE profiles of the NS (Figure S3a) and M (Figure S3b) vRNAs markedly differ from the NoNP and ProtK profiles, suggesting that NP substantially modifies the vRNA structure. This result was confirmed by comparing the classified (low, intermediate, or high) SHAPE values of the NS and M vRNAs between the NoNP and Comp (Table 3) or ProtK and Comp (Table 4) conditions.

Only 54.9 (M vRNA) to 58.3% (NS vRNA) of nts belonged to the same reactivity category under the NoNP and Comp conditions; the reactivity of 30.0 (NS vRNA) to 34.8% (M vRNA) of nucleotides differed by one category (from intermediate to low or intermediate to high or vice-versa), and the reactivity of 10.3 (M vRNA) to 11.9% (NS vRNA) of nts went from low to high or vice-versa. Notably, 67 and 58 nts of the NS vRNA and 77 and 64 nts of the M vRNA were highly reactive uniquely under the NoNP or Comp condition, respectively (Table 3). Thus, the SHAPE reactivity values of the NS and M vRNAs are much more different between the NoNP and Comp conditions (Table 3) than between the NoNP and ProtK conditions (Table 2). Of note, the SHAPE reactivity differences between the ProtK and Comp conditions (Table 4) are very similar to those observed between the NoNP and Comp conditions (Table 3): 52.1 (M vRNA) to 57.1% (NS vRNA) of nts fell into the same reactivity category under the ProtK and Comp conditions, 29.5 (NS vRNA) to 35.3% (M vRNA) of nts differed by one category), and 12.6 (M vRNA) to 13.4% (NS vRNA) of nts differed by two categories (from low to high or vice-versa) (Table 4). Altogether, our SHAPE data indicate that NP has a major structural impact when it binds to the NS or M vRNA; most of this effect is lost when NP is removed by ProtK treatment.

There is no medium or high-resolution structure of the vRNA/NP interactions since, for instance, vRNA is not visible in the cryo EM structures of authentic vRNPs [3,51,52]. However, NP crystal structures revealed a putative RNA binding groove between the head and body domains of NP lined with conserved basic residues that most likely interact with the ribose phosphate moieties of RNA [12,15,16,17]. Furthermore, a recent sub-nanometric structure of a model helical IAV nucleocapsid identified RNA densities adjacent and between the positively charged NP surfaces [20]. Therefore, since NMIA, as all SHAPE reagents, modifies the 2′OH of unpaired and flexible nucleotides, it is expected that the binding of NP decreases the SHAPE reactivity of the nts interacting with the protein. Thus, unreactive nts can be either base-paired or interact with NP, while highly reactive nts are neither base-paired nor interact with NP.

In order to assess the impact of NP on the vRNA structure, we reported the differences in reactivity of the NS and M vRNAs between the NoNP and Comp conditions on the secondary structure models obtained using the NoNP SHAPE data as constraints (Figure 4). When comparing the NoNP and Comp reactivity values, it appears that, both for NS (Figure 4a) and M (Figure 4b) vRNAs, the addition of NP mostly induces SHAPE reactivity decreases in single-stranded regions, and reactivity increases in helices. In some cases, the protections induced by NP in single-stranded regions extend by one or a few base-pairs in the adjacent stem: e.g., nts 331-336 and 417-420 in the NS vRNA (Figure 4a) and 124–129, 139–144, 344–347, 416–418, 555–563, 583–595, 605–612, 632–635, 667–670, 728–734, for the M vRNA (Figure 4b). Furthermore, the stretches of contiguous nucleotides showing an increase in SHAPE reactivity upon NP binding are usually located in close vicinity to protected regions (Figure 4). This pattern of reactivity change suggests that NP usually binds to single-stranded regions of vRNAs and destabilizes the adjacent stems.

Next, in order to determine if NP binding displayed some sequence preference, we analyzed the nt composition of the regions protected from NMIA modification upon the addition of NP (Figure 5a).

Compared to their frequency in the NS and M vRNA sequences, G residues are underrepresented by 1.5-fold in the regions that became protected upon the addition of NP, while the frequency of C residues is slightly but significantly over-represented in the protected regions closely matched the frequency of theses nts in NS and M vRNAs (Figure 5a). Since protected regions varied from one to twelve nts (Figure 5b and Data S2), we compared the base composition of the short protected regions (one or two nts in length) with the longer ones (three to twelve nts in length). This distinction reveals that C residues are significantly over-represented in short-protected regions, whereas U residues are significantly over-represented in the longer-protected regions (Figure 5a). This suggests that, except for a preference for non-G residues, the global base composition is not the only determinant of NP binding and prompted us to analyze the frequency of dinucleotides in the regions protected by NP compared to their frequency in the NS and M vRNAs (Figure 6).

As a general trend, dinucleotides containing at least one G residue were less frequent in the protected regions than in the whole of the NS and M vRNAs, and this difference was significant for three of them (AG, GG, UG) (Figure 6). On the contrary, the dinucleotide UA was significantly more frequent in the protected regions compared to the whole NS and M vRNA sequences (Figure 6). Quite surprisingly, the frequency of the CU and UC dinucleotides was similar in the protected regions and the NS and M vRNA sequences (Figure 6), even though C and U residues were over-represented in the short and long-protected regions, respectively (Figure 5a).

Considering a stoichiometry of ~1 NP per 24 nts [7,8], there should be ~37 and ~43 NP binding sites in the NS and M vRNAs, respectively. We observed 59 and 96 stretches of protected nucleotides in the NS and M vRNAs, respectively (Data S2), but several of these protected stretches lie in close proximity to each other in the vRNA secondary structures models (Figure 4a,b), suggesting that in some cases two or more protected stretches could correspond to a single NP binding site. For instance, in the NS vRNA, nts 195/218–219, 226/246–248/287, 257/273–274; 360–365/473, etc., might correspond to single NP binding sites (Figure 4a) and the same is true for nts 124–129/153–157; 203–205/208/211–212, 254–255/313, etc., in the M vRNA (Figure 4b). Furthermore, large junctions that connect several helices often display several protected nt stretches and are possibly protected by binding of a single NP protein: e.g., junctions 348–355, 397–407/437–439, 549–563/580–592, and 778–782/857–874 in the NS vRNA (Figure 4a), and junctions 201–218, 246–247/323–335, 355–357/462–469/490–491/1006, 365–370/394–398/428–439, 540–551/607–616/828–832, and 647–648/713/792–798, etc. in the M vRNA (Figure 4b). This suggests that, while protections can theoretically reflect either direct binding of NP or new base-pairings induced by NP binding, most protections likely correspond to NP footprints.

To test this hypothesis, we used the GUUGle v1.2 software [53] to identify sequences that could base-pair with the longest protected nt stretches. Of the eight protected sequences that are at least 8 nts in length, four protected regions have no partner sequence to form a stable helix. Amongst the four remaining protected regions, only one can potentially base-pair with a region that partially becomes protected upon the addition of the NP protein: nts 588–595 in the NS vRNA can potentially base-pair with nts 507–514, but the resulting helix would consist of 4 G-U and 4 A-U base-pairs and would therefore be fairly unstable. This reinforces the idea that a large fraction of the protections observed upon the addition of NP corresponds to NP footprints rather than the formation of new base pairs. Interestingly, the number of nts that became protected upon NP addition is similar to the number of nts that showed increased reactivity: in the NS and M vRNAs, respectively, 178 and 260 nts showed decreased reactivity, while 186 and 247 nts displayed increased reactivity upon NP addition (Figure 4). Since increased reactivity reflects the destabilization of helixes, whereas, as discussed above, most reduced reactivity does not reflect the formation of new base pairs, our results indicate that NP strongly destabilized the secondary structures of the NS and M vRNAs.

The common way to incorporate SHAPE probing data during secondary structure modeling is to implement them as pseudo-energies [48,49]. However, in the case of the vRNA/NP complexes, both base-pairing and NP binding can protect nts from modification by NMIA. Using the complete set of SHAPE reactivities as pseudo-energies to model the RNA secondary structure will likely favor the base-pairing of nts that are bound to NP, resulting in incorrect secondary structure models (Figures S4a and S5a). Nevertheless, SHAPE data contain useful information as nts with a SHAPE reactivity >0.8 are most likely unpaired. Therefore, we tested two different approaches to account for the high reactivity SHAPE values. In the first one, termed “partial pseudo-energies,” we only treated SHAPE values > 0.8 as pseudo-energies (Figures S4b and S5b) [36]. Alternatively, we used SHAPE values > 0.8 as hard constraints, forcing the corresponding nts to be unpaired in the secondary structure models (Figures S4c and S5c). Next, we compared the secondary structure models obtained using these three different approaches to take the SHAPE data into account using RNAStructViz [39] (Figure 7).

This analysis indicated that in 564, 482, or 454 out of the 890 nts, the NS vRNA are base-paired when the SHAPE data are treated as pseudo-energies, partial pseudo-energies, or hard constraints, respectively; for vRNA M, the corresponding figures are 614, 598, and 508, respectively. As expected, treating the high SHAPE reactivities as hard constraints resulted in the secondary structures with the lowest number of base pairs (Figures S4 and S5, and Figure 7). Figures S4 and S5, and Figure 7 reveal that the predicted secondary structures strongly depend on the way SHAPE data are used to constraint RNA modeling. Indeed, only 146 nucleotides of vRNA NS (Figure S4 and Figure 7a) and 226 nucleotides of vRNA M (Figure S5 and Figure 7b) form the same base-pairs in the secondary structure models generated by treating SHAPE reactivities as hard constraints, partial pseudo-energies, or pseudo-energies. Interestingly, these common base pairs correspond almost exclusively to local structures (Figure 7), with only one common helix in each vRNA involving nucleotides ~100 nucleotides apart: helix 362–364/472–474 in vRNA NS (Figure S4) and helix 359–364/455–460 in vRNA M (Figure S5). Of note, these local structures were also predicted to exist under the NoNP and ProtK conditions (compare Figure 2 with Figure 7), suggesting that they are particularly likely to exist. By contrast, different long-range interactions were predicted, depending on how the SHAPE data were used during RNA modeling (Figure 7a,b). Of note, using partial pseudo-energies instead of pseudo-energies did not reduce the number of highly reactive nts that were predicted to be base-paired (40 nts versus 30 nts for NS vRNA, and 30 nts versus 35 for M vRNA) (Figures S4a,b and S5a,b). This suggests that more information on NP binding sites is needed to predict the secondary structure of vRNA complexed with NP with high confidence.