Opening and closing of a cryptic pocket in VP35 toggles it between two different RNA-binding modes
Figures

Crystal structures of the RNA-bound and unbound states of all three VP35 interferon inhibitory domain s (IIDs) are nearly identical.
(A) Crystal structures of Zaire IID alone (3FKE, brown) and bound to 8 bp dsRNA (3L25, gray) show that both the blunt end and backbone binding poses of Zaire are nearly identical to the unbound (apo) structure. (B) Overlay of Zaire IID bound to 8 bp dsRNA (3L25, gray) with unbound structures of Reston IID (3L2A, red) and Marburg IID (4GH9, blue).

Pairwise sequence comparisons of all homologs used in this study.
Residues known to make contacts with dsRNA from structural studies are shown in green. Identical residues are shown with a black background. Similar residues are shown with a gray background. P280 in Reston interferon inhibitory domain (IID) and A291 in Zaire IID are shown in a red box.

Simulations predict the probability of pocket opening is lowest in Reston and highest in Marburg.
Each curve is the probability distribution of the distance between two residues that serves as a proxy for pocket opening. Reston VP35 interferon inhibitory domain (IID) (red) shows the least probability of opening the pocket. Zaire VP35 IID shows a greater probability of opening the pocket as well as an increased maximum distance of pocket opening. Marburg VP35 IID shows the highest probability of pocket opening and the highest maximum distance of pocket opening. The shadow around each curve shows the distributions obtained for 100 Markov State Models (MSMs) constructed using random samples of the data chosen with replacement to indicate the statistical variability in the MSM construction. The solid line indicates the mean equilibrium probability of all the bootstraps. The structure with the largest pocket (i.e. largest distance between the two residues) for each variant is shown in ribbon with a transparent surface, using the same color scheme as for the probability distributions. Residues 225 (236 in Zaire) and 295 (306 in Zaire) are shown in yellow sticks.

Structure of Reston and Marburg VP35 interferon inhibitory domains (IIDs) with residues in the allosteric network colored according to the CARDs community they belong to.
Network representation of the coupling between communities of residues is shown below the corresponding structure colored as in the structures. Node size is proportional to the strength of coupling between residues within the community, and edge widths are proportional to the strength of coupling between the communities.

Implied timescales tests for the six slowest eigenvectors of Markov State Models (MSMs) of Reston interferon inhibitory domain (IID) (A) and Marburg IID (B).
Lag times of 6 ns were used for both MSMs.

Thiol labeling of C296 confirms that the cryptic pocket in Marburg has the highest probability of being open, while Reston has the lowest probability of opening.
(A–C) Plots of the distance between residues 225 and 295 vs. the Solvent Accessible Surface Area (SASA) of (A) C296 of Reston interferon inhibitory domain (IID), (B) C307 of Zaire IID, (C) C296 of Marburg IID. Each point on the plot represents a Markov State Model (MSM) center and is colored according to its equilibrium probability. (D) Observed thiol labeling rates for C296/C307 of Zaire IID (black circles), Marburg IID (blue squares), and Reston IID (red triangles) at a range of 5,5’-dithiobis-(2-nitrobenzoic acid) (DTNB) concentrations. (E–G) Plots of the distance between residues 225 and 295 vs. the solvent accessible surface area (SASA) of (E) C315 of Reston IID, (F) C326 of Zaire IID, and (G) C315 of Marburg IID calculated from our MSMs. The SASA of C296 is more correlated with the opening of the cryptic pocket than the SASA of C315, so we focus on thiol labeling of C296 to experimentally characterize pocket opening. (H) Observed thiol labeling rates for C315/C326 of Zaire IID (black circles), Marburg IID (blue squares), and Reston IID (red triangles) at a range of DTNB concentrations. Fits to the Linderstrøm–Lang model are shown in colored lines and the expected labeling rate from the unfolded state is shown as blue dotted lines for Marburg IID and red dotted lines for Reston IID. This rate is estimated from the stability and unfolding rate measured for these homologs shown in Figure 3—figure supplement 4. The mean and standard deviation from three measurements are shown for D and H. Error bars are smaller than the marker used.

Plots of solvent accessible surface area (SASA) of C264 of Reston IID (A) and ZAIRE interferon inhibitory domain (IID) (B) and C236 of Reston IID (C) and ZAIRE IID (D) against the distance between residues 225 and 295 calculated from our Markov State Models (MSMs).

Representative time traces from (A) three repeats of a thiol labeling experiment (red) performed on Reston interferon inhibitory domain (IID) at 100 μM 5,5’-dithiobis-(2-nitrobenzoic acid) (DTNB) and a quadruple exponential fit (black) and (B) one repeat of a thiol labeling experiment (black) performed on MARV IID at 100 μM DTNB and a double exponential fit (red).
The data are background subtracted (the average absorbance from three runs with DTNB but no protein were subtracted) to account for spontaneous hydrolysis of DTNB.

Kobs vs [5,5’-dithiobis-(2-nitrobenzoic acid), DTNB] plots from thiol labeling of Reston interferon inhibitory domain (IID) wild-type (WT) (A), Reston IID C236S/C264S (B), Marburg IID WT (C), and Marburg IID C296S.

Stability of the homologs obtained from a two-state fit to urea denaturation observed via intrinsic tryptophan fluorescence (A) Reston wild-type (WT) (light red triangles, solid lines), Reston P280A (dark red triangles, dashed lines), (B) Marburg WT.
Unfolding rates at 0 M urea estimated by measuring unfolding rates at higher urea concentrations for (C) Reston WT, (D) Reston P280A, and (E) Marburg WT. The observed rate for the unfolded fraction is calculated using the Linderstrøm-Lang model using the unfolding rate and the stability of each protein (see Materials and methods).

Binding to different length dsRNAs suggests closed conformations preferentially bind dsRNA blunt ends while open conformations prefer binding the backbone.
(A–C) Binding to fluorescently labeled 8 bp (dashed lines) and 25 bp RNA (solid lines) with and without a 3’ 2 nucleotide overhang of (A) wild-type (WT) Reston interferon inhibitory domain (IID) (B) WT Zaire IID (C) WT Marburg IID. The anisotropy was measured via a fluorescence polarization assay, converted to anisotropy, and fit to a one-dimensional lattice binding model. The mean and standard deviation from three replicates is shown but error bars are generally smaller than the symbols. (D) Comparison of binding affinities obtained from the global fits. The mean and standard deviation from fits to each of the three replicates are shown. (E–F) Fraction of the backbone of 25 bp RNA covered by the open states (empty markers) and closed states (colored markers) calculated from the binding parameters obtained from the fits for (E) Reston WT IID and (F) Zaire WT IID.

Equilibrium constants for pocket opening and cooperativities between binding modes obtained from the global fit to the binding model.
(A) Comparison of obtained from the global fits to RNA-binding data with for C296 (C307 in Zaire interferon inhibitory domain, IID) exposure obtained from 5,5’-dithiobis-(2-nitrobenzoic acid) (DTNB) labeling experiments shown in Figure 3D. (F) Comparison of cooperativity between backbone binding modes obtained from the global fits. Subscript o stands for open state and c stands for closed state. For example, is the cooperativity between the open and the closed states binding in that order along the backbone.

Single amino acid substitutions at residue 280 (291 in Zaire) modulate the probability of pocket opening.
(A) Probability distribution of the distance between residues 225 and 295 obtained from Markov State Models (MSMs) built from FAST adaptive sampling simulations of Reston interferon inhibitory domain (IID) wild-type (WT) (solid red), Reston IID P280A (dashed red), Zaire IID WT (solid black), and Zaire IID A291P (dashed black). The shadow around each curve shows the distributions obtained for 10 MSMs constructed using random samples of the data chosen with replacement to indicate the statistical variability in the MSM construction. The solid line indicates the mean equilibrium probability of all the bootstraps. (B) Observed labeling rates of C296 for Reston IID WT (transparent solid red) and Reston IID P280A (dark dashed red) Zaire IID WT (transparent solid black), and Zaire IID A291P (dashed black).

Probability distributions of solvent exposure of Cysteine 296 and Cysteine 315 in Reston P280A and Zaire A291P IIDs.
(A) Probability distribution of the distance between residues 236 and 306 obtained from Markov State Models (MSMs) of FAST pockets simulations of Zaire interferon inhibitory domain (IID) WT (solid black) and Zaire IID A291P (dashed black). (B) C307 SASA as a function of distance between residues 236 and 306 in Zaire IID A291P (C) Observed labeling rates of C307 for Zaire IID WT (transparent solid black) and Zaire IID A291P (dark dashed black). (E) C307 SASA as a function of distance between residues 236 and 306 in Zaire IID A291P. (F) Observed labeling rates of C326 for Zaire IID WT (transparent solid black) and Zaire IID A291P (dark dashed black).

Kobs vs [5,5’-dithiobis-(2-nitrobenzoic acid), DTNB] plots from thiol labeling of Reston interferon inhibitory domain (IID) P280A (A), Reston IID P280A/C236S/C264S.

Kobs vs [5,5’-dithiobis-(2-nitrobenzoic acid), DTNB] plots for C315 in Reston interferon inhibitory domain (IID) P280A (A) and Zaire IID A291P (B).
Both cysteines label faster than expected from the unfolded fraction (shown in dotted lines).

Single amino acid substitutions that alter the probability of pocket opening also induce differential binding to blunt ends and the backbone.
(A–D) Binding of Zaire A291P interferon inhibitory domain (IID) (empty black circles), wild-type (WT) IID (solid red triangles), WT Zaire IID (solid black circles), Reston P280A IID (empty red triangles), WT Marburg IID (blue squares) to fluorescently labeled (A) 8 bp blunt ended RNA (B) 8 bp RNA with two nucleotide overhangs on 3’ ends (C) 25 bp blunt ended RNA (D) 25 bp RNA with two nucleotide overhangs on 3’ ends. The anisotropy was measured via a fluorescence polarization assay, converted to anisotropy, fit to a one-dimensional lattice binding model. The mean and standard deviation from three replicates are shown but error bars are generally smaller than the symbols. Lines indicate the global fits. (E) Comparison of obtained from the global fits to for C296 exposure obtained from 5,5’-dithiobis-(2-nitrobenzoic acid) (DTNB) labeling experiments shown in Figure 3D. (F) Comparison of dissociation constants obtained from the global fits. (G) Comparison of cooperativity between backbone binding modes obtained from the global fits. (H) Comparison of cooperativity between end and backbone binding modes obtained from the global fits. The mean values of fits to three individual replicates are shown for E, F, G, and H. Standard deviations are shown as error bars.

Global fits to the binding model of (A) Reston P280A interferon inhibitory domain (IID) and (B) Zaire A291P IID.

Fraction of the RNA backbone covered by the open and closed states of the various homologs and mutants (in increasing probability of pocket opening from left to right) for 8 bp, 25 bp, and 100 bp blunt ended RNA calculated from the binding parameters obtained from the global fits.

Sum of squared residuals for fits of the RNA binding of all five variants of interferon inhibitory domain (IID) used in this study for various combinations of binding site sizes for the three binding modes.
(A–F) show all values of sum of squared residuals (colorbar) for every combination of binding site sizes tested. Horizontal axis in each panel is the binding site size of the open state and vertical axis is the binding site size of the closed state. Each panel shows the matrix for a single end binding site size ranging from 1 to 6 (A–F), respectively. (G) Data from A-F plotted to obtain the condition where sum of squared residuals is minimum. End binding site size is plotted on the horizontal axis. Circle, diamond and plus markers indicate backbone binding site size of the open state being 3, 4, 5 nucleotides, respectively. Blue, yellow, and orange markers indicate backbone binding site size of the closed state being 3, 4, and 5, nucleotides respectively.

Fits and resulting parameters with the backbone binding site size of the closed state of three nucleotides, backbone binding site size of the open state of four nucleotides and the end binding site size as one nucleotide.

Fits and resulting parameters with the backbone binding site size of the closed state of four nucleotides, backbone binding site size of the open state of three nucleotides and the end binding site size as one nucleotide.

Fits and resulting parameters with the backbone binding site size of the closed state of three nucleotides, backbone binding site size of the open state of three nucleotides and the end binding site size as one nucleotide.

Fits and resulting parameters with the backbone binding site size of the closed state of four nucleotides, backbone binding site size of the open state of four nucleotides and the end binding site size as four nucleotides.

List of different statistical weights for each base pair in the presence of the open and the closed states.
Additional files
-
Supplementary file 1
Parameters for the Linderstrøm-Lang model obtained from fits of the Thiol-labeling experiments.
- https://cdn.elifesciences.org/articles/104514/elife-104514-supp1-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/104514/elife-104514-mdarchecklist1-v1.docx