Binding stoichiometry and structural model of the HIV-1 Rev/importin β complex

Rev forms oligomers that are disrupted by point mutations V16D and I55N

Our initial efforts to study the Impβ/Rev interaction were hampered by the tendency of Rev to aggregate under the low to medium ionic strength conditions required for stable Impβ/Rev complex formation. Previous studies reported several point mutations that hinder Rev multimerization (13, 65, 85, 86, 87, 88, 89), including mutations V16D and I55N which destabilize the homotypic A–A and B–B interfaces, respectively (Fig 1C) (85, 89). In particular, the double V16D/I55N mutant was previously used for biophysical studies of Rev, including single-molecule fluorescence spectroscopy (90) and NMR (66) experiments. Accordingly, we generated Rev mutants V16D and V16D/I55N, as well as a truncated version (residues 4–69) of the V16D/I55N mutant lacking the C-terminal domain. Bacterially expressed Rev proteins were purified to homogeneity, and their oligomeric state verified by size-exclusion chromatography/multi-angle laser light scattering (SEC/MALLS) in a high salt buffer (HSB) required to keep Rev soluble (Fig 2A). Wild-type (WT) Rev was eluted as two separate peaks: the first contained molecular species ranging between 330 and 500 kDa in mass (suggesting Rev multimers comprising ∼24–36 monomers), whereas the second exhibited masses of 40–80 kDa (consistent with species comprising 3–6 Rev monomers), confirming the highly polydisperse nature of Rev. The V16D mutant was eluted later than WT Rev, yielding a broad peak with a mass distribution (15–45 kDa) consistent with a mixture of species comprising 1–3 Rev monomers. The mass observed at the maximum of this peak (∼23 kDa) indicated a prevalence of the dimeric form (27.6 kDa), in agreement with previous findings (85, 89). By contrast, both the full-length and truncated versions of the V16D/I55N mutant eluted as relatively narrow peaks with mass distributions centered around 13.3 and 7.5 kDa, respectively, consistent with a monomeric state (13.8 and 8.0 kDa, respectively). The monodisperse nature of the V16D/I55N mutant facilitated several of the interaction studies with Impβ (isothermal titration calorimetry [ITC], NMR, and SAXS experiments) described below. For convenience, we refer to the full-length and truncated versions of this mutant as Rev^OD (for oligomerization-deficient Rev) and Rev^ODΔ, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. HIV-1 Rev forms a stable complex with Impβ.

(A) Analysis of different Rev constructs by size-exclusion chromatography/multi-angle laser light scattering (SEC/MALLS) using a Superdex 75 10/300 GL column. Elution curves recorded at 280 nm and molar mass distributions determined by MALLS are shown for the following Rev constructs: WT (green), V16D (blue), V16D/I55N (purple), and Rev^4-69(V16D/I55N) (pink). The molar masses detected at the elution peaks are indicated. (B) SEC analysis of an Impβ/Rev^OD complex. Top: Elution profiles of samples containing Impβ (blue), Rev^OD (magenta), or a mixture of Impβ and Rev^OD (green). Fractions collected are indicated in brown. Chromatography was performed using a Superdex 200 5/150 Increase GL column. Bottom: SDS–PAGE analysis of the indicated fractions. (C) SEC analysis of an Impβ/Rev^WT complex. Elution profiles are shown for Impβ (blue), Rev^WT (magenta) and a mixture of Impβ and Rev^WT (green). Because most free Rev^WT forms insoluble aggregates in the buffer conditions used and hence is not detected in the elution, the elution profile of Rev^WT analyzed on a column pre-equilibrated in high-salt buffer was included as an additional reference (purple). (D) Native gel analysis showing the association of Impβ with one or more monomers of Rev^WT (top) and Rev^OD (bottom). (E) SEC/MALLS analysis of Impβ in the presence and absence of Rev. Elution curves recorded at 280 nm and molar mass distributions determined by MALLS are shown for His-tagged Impβ in the absence (blue) and presence of a either a twofold (light green) or fourfold (dark green) excess of Rev^WT. The observed masses of 124 and 137 kDa are consistent with those expected for 1:1 (116.2 kDa) and 1:2 (129.4 kDa) Impβ:Rev stoichiometry. Chromatography was performed using a Superdex 200 10/30 GL column. Injected concentrations were 20 μM Impβ and 40 or 80 μM Rev^WT.

Impβ binds up to two monomers of Rev

We next assessed the ability of Rev to associate with Impβ. SEC analysis revealed that unbound Impβ eluted as a single peak, whereas preincubating Impβ with a molar excess of Rev^OD led to the co-elution of both proteins within an earlier peak, indicating complex formation (Fig 2B). SEC analysis confirmed that Impβ also co-eluted with WT Rev, even though WT Rev was poorly soluble in the ionic strength conditions used for this experiment (Fig 2C). Impβ/Rev complex formation was confirmed by native gel analysis, which revealed a single band for unbound Impβ and a discrete shift upon the addition of increasing concentrations of either WT Rev or Rev^OD (Fig 2D, complex 1). Surprisingly, higher Rev concentrations resulted in a supershift, suggesting the formation of an additional species containing two or more Rev monomers (Fig 2D, complex 2). The similar results obtained with WT Rev and Rev^OD suggest that residues V16 and I55 are not critical for the stability of the observed Impβ/Rev complexes. SEC/MALLS analysis also revealed molecular masses for Impβ/Rev^WT complexes consistent with 1:1 and 1:2 stoichiometry (Fig 2E). A glutaraldehyde cross-linking experiment followed by mass spectrometry (MS) analysis corroborated the ability of Impβ to bind two Rev^OD monomers (Fig S1A and B).

• Download figure
• Open in new tab
• Download powerpoint

Figure S1. Glutaraldehyde cross-linking and MALDI-TOF mass spectrometry indicate that Impβ binds two Rev monomers.

(A) Impβ was incubated with Rev^OD in the presence of glutaraldehyde and the mixture was analyzed by native (top panel) and denaturing (bottom two panels) gel electrophoresis followed by Coomassie staining. The upper denaturing gel (containing 15% acrylamide:bis-acrylamide in a 37.5:1 ratio) allowed Impβ and Rev to be visualized on the same gel, whereas the lower gel (containing 10% acrylamide:bis-acrylamide) was used to resolve cross-linked species migrating close to Impβ. At Rev^OD concentrations yielding a single gel shift by native gel electrophoresis, SDS–PAGE analysis revealed an additional band that migrated just above Impβ (bottom gel, compare lanes 1 and 3). At higher Rev^OD concentrations that yielded a supershift on the native gel, a second more slowly migrating band was detected by SDS–PAGE (bottom gel, lanes 4–5). The vertical black line on the native gel indicates that lanes 1–2 and lanes 3–5 are taken from nonadjacent regions of the same gel. (A, B) MALDI-TOF mass spectrometry analysis of samples corresponding to lanes 3 (upper profile) and 5 (lower profile) from panel (A) was performed. MALDI MS spectra showed peaks consistent with the molecular mass of free Impβ (indicated by a circle), Impβ cross-linked to either one or two Rev^OD monomers (indicated by a circle attached to one or two diamonds, respectively), or cross-linked Rev^OD dimers (indicated by two diamonds).

To further verify the stoichiometry of the Impβ/Rev association we performed native MS, which preserves non-covalent interactions and allows the mass of intact macromolecular complexes to be determined (91). To ensure correct interpretation of native MS spectra, we first used liquid chromatography coupled to electrospray ionization MS (LC/ESI-MS) to measure accurate masses under denaturing conditions for the unbound Impβ and Rev^OD proteins; these were 97,300 and 13,264 Da, respectively, which are within 1 Da of the theoretical masses (Fig S2 and Table S1). Native MS analysis of unbound Impβ yielded a clear set of peaks whose m/z values correspond to a mass of 97,296 Da (Fig 3A and B), closely matching that determined by ESI-TOF. We next incubated Impβ with Rev^OD and measured the masses of the resulting species (Fig 3A, C, and D). Mixing Impβ with a twofold molar equivalent of Rev^OD led to the appearance of two additional sets of peaks: the first corresponds to a mass of 110,558 Da (Fig 3C, blue circles), matching the expected mass of an Impβ/Rev^OD heterodimer (110,563 Da), and the second to a mass of 123,829 Da (Fig 3C, green circles), consistent with a complex of Impβ bound to two Rev^OD monomers (123,827 Da). Mixing Impβ with Rev^OD in a 1:5 M ratio enhanced the intensity of this second set of signals (Fig 3D). Mixing Impβ with the truncated construct (Rev^ODΔ) instead of full length Rev^OD yielded an analogous set of peaks for Impβ/Rev^ODΔ complexes having a stoichiometry of 1:1 (mass: 105,330 Da) and 1:2 (113,366 Da) (Fig 3A and E). Taken together, these results confirm the ability of Impβ to bind up to two monomers of Rev^OD.

• Download figure
• Open in new tab
• Download powerpoint

Figure S2. Mass determination of Impβ and Rev proteins under denaturing conditions.

A, B) LC/ESI-TOF MS analysis of (A) Rev^OD and (B) Impβ. The upper mass spectrum in each panel represents the raw ESI data, whereas the lower panel shows the deconvoluted ESI spectrum. βME, β-mercaptoethanol; TFA, trifluoroacetic acid.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Native MS reveals that Impβ binds up to two Rev monomers.

(A) Summary of masses observed by native MS. (B, C, D, E, F, G) Native MS spectra. Peaks labeled by magenta circles correspond to unbound Impβ. Peaks labeled by blue circles with a single dot or by green circles with two dots correspond to Impβ/Rev complexes with 1:1 or 1:2 stoichiometry, respectively. (B) Spectrum of Impβ in the absence of Rev. (C, D) Spectra of Impβ incubated with a (C) twofold or (D) fivefold molar equivalent of Rev^OD. (E) Spectra of Impβ in the presence of the truncated construct Rev^ODΔ. (F, G) Spectra of Impβ incubated with a (F) twofold or (G) fivefold molar equivalent of Rev^WT.

We then investigated complex formation between Impβ and Rev^WT. We overcame the tendency of Rev^WT to oligomerize by exchanging its buffer into a high concentration of ammonium acetate before incubating with Impβ and using relatively low Rev^WT concentrations (10–20 μM). Upon mixing Impβ with a twofold molar equivalent of Rev^WT, we detected peaks consistent with the formation of a 1:1 Impβ:Rev^WT complex (110,476 Da; Fig 3A and F). With a fivefold molar equivalent of Rev^WT, we detected both 1:1 and 1:2 complexes (123,649 Da; Fig 3A and G). This confirms that Impβ can bind to either one or two molecules of WT Rev. Gel shift experiments (performed with either Rev^OD or Rev^WT) showed that RanGTP disrupted the formation of both the 1:1 and 1:2 complexes, indicating that RanGTP competes with both Rev monomers for binding to Impβ (Fig S3).

• Download figure
• Open in new tab
• Download powerpoint

Figure S3. RanGTP competes with both monomers of Rev for binding to Impβ.

(A, B) Samples containing a mixture of Impβ, RanGTP and either (A) Rev^OD or (B) Rev^WT were analyzed by native gel electrophoresis (top) or SDS–PAGE (bottom) followed by Coomassie blue staining. Protein concentrations used were 2.5 μM for Impβ, 5 μM for RanGTP and either 2.5 or 5 μM for Rev proteins, indicated by (+) and (++), respectively. Compared with unbound Impβ (lanes 1, 11 and 18), incubating Impβ with Rev resulted in shifted and supershifted bands corresponding to 1:1 and 1:2 Impβ/Rev complexes, marked by light and dark green arrowheads, respectively (lanes 4, 7, 9, 14, and 16). Incubating Impβ with RanGTP yielded a band that migrated with intermediate mobility (purple arrowhead, lane 5). When all three Impβ, RanGTP, and Rev proteins were co-incubated, the bands corresponding to the Impβ/Rev complexes were absent or only fainted detected whereas strong intensity was observed for the Impβ/RanGTP band, showing that RanGTP prevents Impβ from binding either monomer of Rev.

Impβ recognizes the Rev helical hairpin through a high- and a low-affinity binding site

We next characterized the Impβ/Rev interaction using ITC. We recorded ITC profiles for the binding of Impβ to either Rev^OD or Rev^ODΔ because poor solubility and the tendency to multimerize precluded the use of WT Rev (Figs 4A and B and S4). Fitting the resulting binding isotherms using a model consisting of a single class of binding sites resulted in a significantly poorer fit compared with a model consisting of two nonsymmetric classes of binding sites, further confirming the ability of Impβ to bind two Rev monomers (Fig S5). (We note that the uniphasic, rather than biphasic, appearance of the observed binding isotherms is compatible with the presence of two binding sites, as has been previously shown (92, 93)). These analyses yielded a dissociation constant (K_d) for the Impβ/Rev^OD interaction of ∼0.6 μM for the first site and 5 μM for the second site. Comparable K_d values (0.6 and 8 μM) were obtained for the Impβ/Rev^ODΔ interaction, suggesting that the C-terminal domain missing from Rev^ODΔ does not contribute significantly to binding affinity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Rev binds Impβ at two sites with sub- and low micromolar affinity.

(A, B) Representative isothermal titration calorimetry profiles of the binding of Impβ to (A) Rev^OD and (B) Rev^ODΔ. Top: Differential power time course of raw injection heats for a titration of Impβ into the Rev protein solutions. Bottom: Normalized binding isotherms corrected for the heat of dilution of Impβ into buffer. The solid line represents a nonlinear least squares fit using a model consisting of two nonsymmetric classes of binding sites. (C, D) Thermodynamic values obtained from isothermal titration calorimetry data for the binding of Impβ to (C) Rev^OD and (D) Rev^ODΔ. Binding to site 1 is characterized by a favorable enthalpy (ΔH ≈ −8 kcal/mol) and a negligible entropy change for both Rev^OD and Rev^ODΔ, whereas binding to site 2 is associated with an unfavorable entropy change that is offset by a large negative enthalpy change. Thermodynamic parameters (in kcal/mol) are as follows: Rev^OD: ΔG = −8.3 ± 0.1 and −7.2 ± 0.4, ΔH = −6.3 ± 2.4 and −26.5 ± 3.5, −TΔS = −2.1 ± 2.5 and 19.3 ± 3.9 for sites 1 and 2, respectively; Rev^ODΔ: ΔG = −8.5 ± 0.2 and −7.2 ± 0.8, ΔH = −8.3 ± 0.6 and −23.6 ± 3.6, −TΔS = 0.3 ± 0.8 and 16.5 ± 4.0 for sites 1 and 2, respectively. Data represent the mean ± SD from three independent replicates. All replicate profiles are shown in Fig S4.

• Download figure
• Open in new tab
• Download powerpoint

Figure S4. ITC analysis of the Impβ:Rev interaction.

(A, B) ITC profiles of the binding of Impβ to (A) Rev^OD and (B) Rev^ODΔ. Three independent replicates are shown for each assay. Top: Differential power time course of raw injection heats for a titration of Impβ into the Rev protein solutions. Bottom: Normalized binding isotherms corrected for the heat of dilution of Impβ into buffer. The solid line represents a nonlinear least squares fit using a model consisting of two nonsymmetric classes of binding sites. Profiles labeled “Exp1” are identical to the panels shown in Fig 4A and B.

• Download figure
• Open in new tab
• Download powerpoint

Figure S5. Isothermal titration calorimetry data are consistent with two classes of Rev binding sites on Impβ.

Representative isothermal titration calorimetry profile for the binding of Impβ to Rev^OD showing that a model consisting of two nonsymmetric classes of binding sites (right) yields a better fit to the normalized binding isotherm than a model consisting of a single class of binding site (left), as highlighted by the data points circled in red.

The finding that Impβ binds one Rev monomer with submicromolar affinity and the other with ∼10-fold weaker affinity is consistent with the sequential appearance of complexes observed when Impβ is titrated with Rev (Figs 2D and E and S1A). The interaction of Impβ with both the first and second Rev monomers is enthalpically driven (Fig 4C and D), suggesting electrostatic and hydrogen bond interactions that presumably take place between the highly basic ARM of Rev and the acidic cargo-binding inner surface of Impβ. Interestingly, although the higher affinity interaction (site 1) is associated with a negligible entropy change, the lower affinity interaction (site 2) is characterized by an unfavorable entropy term, suggesting that the binding of the second Rev monomer might significantly reduce the conformational entropy of the Impβ HEAT-repeat array, which is dynamically highly flexible in the unbound state (94).

To further probe the regions of Rev involved in Impβ binding, we performed NMR spectroscopy of ¹⁵N-labeled Rev^OD in the presence and absence of unlabeled Impβ. Triple resonance NMR was used to assign the backbone resonances of ¹⁵N-labeled Rev^OD. The ¹H-¹⁵N-BEST-TROSY spectrum of free Rev^OD shows peaks concentrated between 7.5 and 8.5 ppm (Fig 5A), in agreement with a previous study (66) and consistent with the expectation that Rev contains many intrinsically disordered residues (95). In total, 67 of the 116 residues of Rev could be assigned. Assignment was hampered by resonance overlap caused by the high (14%) arginine composition of the Rev sequence and the large number (10) of prolines, as well as severe line-broadening in the C-terminus of the helical hairpin, possibly because of residual oligomerization at concentrations used for NMR measurement (90 μM). Analysis of the secondary ¹³C chemical shifts showed a strong helical propensity for residues within the helical hairpin domain, as expected, and significant helical propensity for residues 84–93 (Fig 5C). The latter residues are immediately downstream of the NES and their helical propensity might facilitate recruitment of the NES to the binding surface of CRM1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. NMR analysis of Rev binding by Impβ.

(A) ¹H,¹⁵N HSQC spectrum of free Rev^OD (600 MHz, 283K). (B) ¹H,¹⁵N-HSQC spectrum Impβ-bound Rev^OD (blue) superimposed on that of unbound Rev^OD (green). (C) Secondary chemical shifts (96) of unbound Rev^OD. The two point mutations are indicated by asterisks. (D) Ratio of peak intensities for Rev residues in the presence and absence of Impβ (600 MHz, 283K).

The addition of Impβ to Rev^OD had a dramatic effect on the spectrum, causing 28 peaks to disappear either completely (24 peaks) or nearly completely (4 peaks; >95% reduction in peak height) (Fig 5B and D). These peaks include all assigned peaks between residues 10 and 65, which delimit the helical hairpin domain. By contrast, peaks corresponding to N- and C-terminal regions of Rev (residues 2–9 and 73–116) were less affected by the presence of Impβ. This finding indicates that Impβ recognizes Rev primarily through its helical hairpin domain and interacts only weakly or not at all with the N-terminal extension and C-terminal domain, in agreement with the similar binding affinities measured by ITC for the interaction of Impβ with Rev^OD and Rev^ODΔ.

Impβ retains an extended conformation upon Rev binding

We next sought structural information on the Impβ/Rev complex using small-angle X-ray scattering, which was coupled with size-exclusion chromatography (SEC-SAXS) to ensure sample monodispersity. We measured data from Impβ alone or in complex with Rev^ODΔ and compared the observed scattering to that predicted for cargo-bound Impβ conformations (cargo coordinates removed) available from the Protein Data Bank (PDB) (Fig 6A). The scattering profile of unbound Impβ exhibits a shoulder at s ≈ 0.1 Å⁻¹ (Fig 6B), in agreement with previous studies of unbound murine and fungal Impβ proteins (49, 50, 97). The radius of gyration (R_g) and maximum dimension (D_max) of unbound Impβ determined from the pair distance distribution function were 37 and 125 Å, respectively, indicating that Impβ adopts a more extended conformation in solution compared with cargo-bound crystal structures (Fig 6A and C). Consistent with this finding, the observed scattering showed the best agreement (χ² values < 4) with scattering curves calculated for Impβ coordinates from PDB structures 1UKL and 3W5K, which exhibit the most elongated cargo-bound conformations (Fig 6C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Impβ retains an extended conformation upon binding Rev.

(A) Top: Representative crystal structures of Impβ illustrating different degrees of compaction of the HEAT-repeat array. Bottom: Model-independent parameters obtained from SAXS data compared with values calculated for available Impβ crystal structures using the programs CRYSOL (127) and PRIMUS (126). All structures are of human Impβ except for 1UKL which is of murine Impβ. Chain A in 3W5K lacks coordinates for Impβ residues 1–15. For the calculation of R_g, D_max, and χ² values, this chain was extended to include these residues (denoted A*) by replacing HEAT repeat 1 (residues 1–31) by the corresponding residues from 1UKL following local alignment of the two structures. (B) Scattering data from unbound Impβ (black) and the Impβ/Rev^ODΔ complex (red). (C) Scattering data from unbound Impβ (black) compared with profiles calculated from representative cargo-bound conformations exhibiting different degrees of elongation (colored lines).

The binding of Rev^ODΔ caused only modest changes in the scattering profile relative to that of unbound Impβ (Fig 6B), corresponding to a small decrease in R_g and D_max values (to 36.7 and 120 Å, respectively). These changes are consistent with the expected binding of Rev^ODΔ to the concave inner surface of the HEAT-repeat array, which appears not to undergo a dramatic conformational change relative to the unbound state. The small mass of Rev^ODΔ relative to Impβ and the large impact that changes in HEAT repeat conformation have on the predicted scattering curve preclude us from deducing a reliable structural model for the Impβ:Rev^ODΔ complex. Nevertheless, the data reveal that the complex retains an extended conformation resembling that of unbound Impβ and of the cargo-bound conformations 1UKL and 3W5K.

Rev binding sites localize to N- and C-terminal regions of Impβ

To obtain additional structural insights into how Impβ recognizes Rev, we performed cross-linking–mass spectrometry (XL-MS) on the Impβ/Rev complex using the amine-reactive cross-linker bissulfosuccinimidyl suberate (BS3). Although Rev contains two lysine residues (Lys20 and Lys115), BS3-mediated cross-links with Impβ were only detected for Lys20 and for the Rev N-terminal amino group (Fig 7A). Conversely, peptides containing BS3-modified Lys residues were identified for 28 of the 43 lysines present in Impβ. These residues are distributed over the length of Impβ and can be classified into three groups. Group 1 formed cross-links with both the N-terminus and Lys20 of Rev; group 2 formed cross-links with the flexible Rev N-terminus but not with Lys20; and group 3 formed monolinks (where only one end of BS3 was covalently bound to the protein) or cross-links with other Impβ residues but did not form detectable cross-links with Rev. Because Lys20 is located on Rev helix α1, group-1 lysines on Impβ are expected to localize close to Rev’s helical hairpin, whereas lysines in groups 2 and 3 are expected to be farther away. Interestingly, although group-3 residues primarily reside on the A helix of a HEAT repeat and localize to the outer convex surface of Impβ, nearly all group-1 residues reside on a B helix and localize to the inner concave surface (Fig 7B and C), consistent with Rev recognition by the inner surface of Impβ.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7. Cross-linking-MS localizes two Rev binding regions on Impβ.

(A) Graphical summary of cross-links. Impβ-Rev cross-links involving the N-terminus or Lys20 residue of Rev are shown in green and red, respectively. Rev–Rev and Impβ-Impβ cross-links and monolinks are shown in blue. Red, green, and blue inverted flags indicate Impβ residues that form cross-links with Rev Lys20 (group 1), with both Rev Lys20 and the Rev N-terminus (group 2), or for which no cross-links with Rev were detected (group 3), respectively. The 19 HEAT repeats (HR) of Impβ are indicated. (B) Impβ Lys residues modified by BS3 and detected in cross-links or monolinks. Group-1, -2, and -3 lysines are highlighted in red, green, and blue, respectively. For group-1 and -2 lysines, the number of peptide spectrum matches (PSMs) and the best pLink E-value score are indicated. (C) Surface representation of Impβ showing the location of cross-linked group-1, -2, and -3 Lys residues, colored red, green, and blue, respectively. Shading is from dark to light gray from N- to C-terminus. (D) Solvent-accessible surface distances (SASDs) between pairs of Impβ group-1 Lys residues. The upper and lower triangles show distances for the conformations of Impβ bound to the importin α IBB domain (PDB 1QGK) and to SREBP-2 (PDB 1UKL), respectively. SASDs over 70 Å are outlined in dark blue. Distances were calculated using the Jwalk webserver (101). The SAS distances shown here and in Fig S6B suggest that Lys23, Lys62, and Lys68 cross-link with Rev at the N-site and that Lys854, Lys857, Lys859, Lys867, and Lys873 cross-link with Rev at the C-site. (E) Spheres of radius 35 Å centered on the Cα atoms of group-1 residues Lys23 and Lys873 show that BS3 molecules bound to these two lysines cannot cross-link to the same Rev Lys20 position. (F) Localization of the Cα atom of Rev Lys20. Green and magenta volumes show the N- and C-terminal regions of space within cross-linking distance of group-1 residues in either HEAT repeats 1 and 2 (K23, K62, K68) or repeat 19 (K854, K857, K859, K867, and K873), respectively, defined by the intersection of 35 Å spheres centered on the Cα atoms of these residues. If the centroid of each envelope is used to estimate the position of the Rev K20 Cα atom, then the positional uncertainty, calculated as the rmsd of each grid point within the envelope (sampled on a 2 Å grid) relative to the centroid, is 22.8 Å for the N-site and 21.6 Å for the C-site for the 1UKL conformation. (C, E, F) The Impβ conformation shown in panels (C, E, F) is that of SREBP-2-bound Impβ (PDB 1UKL).

The 11.4 Å linker arm length of BS3 allows the Cα atoms of cross-linked lysines to be up to ∼24 Å apart; however, to allow for conformational dynamics, distance constraints of 30–35 Å are typically used for the purposes of 3D structural modeling (98, 99, 100). For such modeling, the solvent-accessible surface (SAS) distance between cross-linked Lys residues is more informative than the straight-line (Euclidean) distance separating them as the latter may trace a path sterically inaccessible to the cross-linker (101). Group 1 includes a cluster of lysines within HEAT repeats 1 and 2 and a second cluster within repeat 19 at the opposite extremity of Impβ. These two clusters are too far apart to form cross-links to the same Rev K20 residue position (Fig 7D). For example, if one considers the Impβ conformation shown in Fig 7E (from PDB 1UKL), the Cα atoms of residues Lys23 and Lys873 are separated by a SAS distance of 103 Å (Euclidean distance of 91.4 Å). Even for the most compact known conformation of Impβ (from PDB 3LWW), residues Lys859, Lys867, and Lys873 are separated from Lys23 by a SAS distance of over 70 Å, twice the BS3 cross-linking distance constraint (Fig S6A and B). These findings are consistent with the ability of Impβ to bind to two Rev molecules and suggest that the two clusters of group-1 lysines form cross-links with distinct Rev monomers, one bound close to the N-terminal end of Impβ and the other close to the C-terminal end. We refer hereafter to these two Rev binding sites on Impβ as the N-site and C-site, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure S6. Cross-linking-MS distance constraints localize two Rev binding regions on Impβ.

(A) Spheres of radius 35 Å centered on the Cα atoms of group-1 residues Lys23 and Lys873 show that BS3 molecules bound to these two lysines cannot cross-link to the same Rev Lys position regardless of the Impβ conformation. The four conformations shown are those of Impβ bound to the following cargos: the IBB domain of Snurportin1 (PDB 3LWW chain C), the IBB domain of importin α (PDB 1QGK), the SREBP-2 complex (PDB 1UKL) and Snail1 (PDB 3W5K). (B) Distances between pairs of Impβ group-1 Lys residues. The upper and lower triangles show distances for the conformations of Impβ bound to the IBB domain of importin α (PDB 1QGK) and the SREBP-2 complex (PDB 1UKL), respectively. Solvent-accessible surface distances (SASDs) over 70 Å are outlined in dark blue. Distances were calculated using the Jwalk webserver (101). (C) Localization of the Cα atom of Rev Lys20. Green and magenta volumes show the N- and C-terminal regions of space within cross-linking distance of group-1 residues in either HEAT repeats 1 and 2 (K23, K62, K68) or repeat 19 (K854, K857, K859, K867, and K873), respectively, defined by the intersection of 35 Å spheres centered on the Cα atoms of these residues.

Applying a 35 Å distance constraint from each residue in the two group-1 lysine clusters localizes the Lys20 Cα atoms of the two Rev monomers within an ellipsoidal volume next to the N- and C-terminal HEAT repeats with a positional uncertainty of ∼22 Å (Figs 7F and S6C). In contrast, group-1 residues Lys206 and Lys537 could potentially cross-link with either Rev monomer. The fact that both these residues localize to the concave inner surface of Impβ confirms that at least one of the two Rev monomers is bound to this surface.

Impβ recognizes peptides derived from Rev helix α2

To delineate the region(s) of Rev recognized by Impβ, we examined the ability of different Rev-derived peptides to interact with Impβ in a thermal shift assay. We used a set of 27 peptides (each comprising 15 residues) that span the entire Rev sequence, with an 11-residue overlap between consecutive peptides. We used differential scanning fluorimetry (DSF) to assess the ability of peptides to increase the thermal stability of Impβ. In the absence of peptides, the melting temperature (T_m) observed for Impβ was 35.4 ± 0.7°C (Fig 8A). Although most Rev peptides did not significantly increase this value (ΔT_m < 0.5°C), three consecutive Rev peptides (numbered 9 to 11) strongly stabilized Impβ (ΔT_m > 2°C), with peptide 10 (³⁷ARRNRRRRWRERQRQ⁵¹), which contains all but one of the ARM’s 10 Arg residues, yielding the strongest effect (Fig 8A and B). These peptides collectively span Rev residues 33–55 on helix α2 and the α1-α2 loop and nearly perfectly match the ARM motif (res. 35–50) (Fig 8C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8. Impβ recognizes Rev peptides derived from helix α2 and the α1-α2 loop.

(A) Examples of thermal denaturation curves measured by differential scanning fluorimetry of Impβ in the presence and absence of Rev peptides. Data are shown for Rev peptides 9–11. The melting temperature (T_m) and difference in T_m compared with unbound Impβ (ΔT_m) are listed as mean values ± SD from three independent experiments. (B) Summary of ΔT_m values determined by differential scanning fluorimetry analysis of Impβ for all 27 peptides spanning the Rev sequence. Details are shown for peptides 7–12. (C) Structure of the Rev N-terminal domain highlighting the residues in helix α2 and the α1-α2 loop spanned by Rev peptides 9–11. Atomic coordinates are taken from PDB 2X7L (69).

The increased stabilization observed for peptide 10 compared with peptide 9 (³³GTRQARRNRRRRWRE⁴⁷) suggests that the Rev ⁴⁸RQRQ⁵¹ motif includes an important binding epitope. Similarly, although Impβ was strongly stabilized by Rev peptide 11 (⁴¹RRRRWRERQRQIRSI⁵⁵), little or no stabilization was observed with peptide 12 (⁴⁵WRERQRQIRSISGWI⁵⁹), indicating that the tetra-arginine motif (⁴¹RRRR⁴⁴) is critical for Impβ binding. Finally, peptide 8 (²⁹PSPEGTRQARRNRRR⁴³), which contains three of these four Arg residues and a total of six closely spaced Arg residues, only modestly stabilizes Impβ (ΔT_m = 1.0 ± 0.7°C), indicating that clustered Arg residues are not sufficient for high-affinity binding and that sequence context also plays an important role.

The N-terminal tip of Rev helix α2 is a major Impβ binding epitope

We next performed mutagenesis experiments to identify individual Rev residues critical for Impβ recognition. Because Rev binding is likely to involve a significant electrostatic component, we sought to destabilize the Impβ/Rev interface by introducing charge-reversal mutations on Rev. We first made mutants in which two or three basic residues were simultaneously replaced by Asp residues (mutants R1–R5; Fig 9A). We used a competitive fluorescence polarization (FP) assay to compare the ability of these proteins to bind Impβ and displace a peptide cargo. In this assay, the binding of Impβ to a fluorescently labeled Rev peptide (Rev-NLS) yields an FP signal that decreases when the peptide is displaced by the addition of unlabeled Rev. The half-maximal inhibitory concentration (IC₅₀) of the Rev protein tested (or its negative logarithm, pIC₅₀) provides an indirect measure of Impβ-binding affinity, with lower IC₅₀ (higher pIC₅₀) values corresponding to stronger binding.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9. Charge-reversal mutations identify an Impβ-binding epitope on Rev.

(A) Rev double and triple R/K→D substitution mutations. (B) Competitive fluorescence polarization (FP) inhibition assays showing the ability of WT and mutant forms of Rev to displace a fluorescently labeled Rev-NLS peptide from Impβ. A representative experiment is shown for each protein. Data shown are mean and SD values from two technical replicates. (C) Plot of IC₅₀ values derived from FP inhibition assays for WT Rev and double and triple mutants. ****P ≤ 0.001; **P ≤ 0.01. P-values were determined using an ordinary ANOVA test. (D) Plot of IC₅₀ values for WT Rev and single R→D point mutants. (E) Summary of IC₅₀ (mean ± SD) and corresponding pIC₅₀ values. N represents the number of biological replicates. ΔpIC₅₀ values are reported relative to WT Rev. (F) View of the Rev helical hairpin showing the Arg residues selected for single-point mutations. Carbon atoms are colored from blue to red in order of increasing ability of the R→D substitution to compromise Impβ recognition, as measured by ΔpIC₅₀ values.

Representative assays are shown in Fig 9B, and the results summarized in Fig 9C–E and Table S2. WT Rev competes with the Rev-NLS peptide with an IC₅₀ value of 0.5 μM, comparable to the K_d value of 0.6 μM obtained by ITC. Of the five mutants tested, three (R2-R4) yielded a large (10-fold) increase in IC₅₀, whereas the other two (R1, R5) gave more modest (3- and 1.7-fold) increases. The three mutants with the greatest effect localized to one end of the helical hairpin (the N-terminal half of helix α2 and the interhelical loop) and involved Arg clusters on both the A- and B-faces of Rev, whereas the mutants with more modest effects localized to the opposite end of the domain (helix α1 and the C-terminal half of helix α2) (Fig 9A), in line with our peptide-scanning data.

We then generated single-point mutations of the nine Arg residues altered in mutants R2-R4. Four of these (R35D, R38D, R39D, and R41D) significantly compromised the ability of Rev to compete with the Rev-NLS peptide, increasing the IC₅₀ by a factor of 1.6-1.9 (versus a factor of 1.1–1.3 for the other mutants), corresponding to a drop in pIC₅₀ of 0.20–0.27 (versus 0.04–0.11) (Fig 9D and E). Interestingly, the sum of the ΔpIC₅₀ values (−0.72) for mutations R35D, R38D, and R39D approached that observed for the corresponding R2 mutant (−1.0), indicating that the effect of the triple mutation primarily reflects the additive effects of the single mutations (Fig 9E). In contrast, the ΔpIC₅₀ values for triple mutants R3 and R4 were three to four times larger in magnitude than the summed ΔpIC₅₀ values of the corresponding single mutations, revealing that reversing the charge at these combined positions had a synergistic effect on disrupting binding. Notably, plotting the ΔpIC₅₀ values of the nine R→D mutations onto the structure of Rev reveals a progressively greater effect of the mutation on Impβ binding as one proceeds along helix α2 toward its N-terminus and the interhelical loop (Fig 9F), consistent with the spatial trend noted above for the triple mutants. These findings point to the N-terminal tip of Rev helix α2 as a major binding epitope recognized by Impβ.

Charge-reversal mutations on Impβ and Rev show compensatory effects

We next sought to identify Impβ residues implicated in Rev recognition by introducing charge-reversal mutations on the acidic inner surface of Impβ. Of the 57 Asp and Glu residues that localize to the protein’s inner surface, we selected 22 that were prominently exposed or clustered into acidic patches for substitution by Arg residues. To enhance the potential disruptive effect on Rev binding, we initially introduced these substitutions as double, triple, or quadruple mutations (mutants B1-B7, Fig 10A). We then used our FP inhibition assay to compare the ability of WT Rev to displace the Rev-NLS peptide from WT or mutant forms of Impβ. Only minor changes in the IC₅₀ value of Rev were observed when WT Impβ (IC₅₀ = 0.5 μM) was replaced by mutants B1-B7 (IC₅₀ = 0.43–0.55 μM) (Fig 10B and Table S2), consistent with the expectation that the Impβ mutations should not significantly alter Rev’s ability to compete with the Rev-NLS peptide. We then repeated the assay using Rev double and triple mutants R1-R5. Strikingly, although the IC₅₀ values of mutants R1 and R5 showed little variation in assays performed with WT or mutant Impβ, the IC₅₀ values of mutants R2-R4 varied considerably (Fig 10C and Table S2). Most notably, these values decreased from ∼5 μM when tested against WT Impβ to ∼1.6 μM against mutant B2, signifying that Rev mutants R2-R4 competed with the Rev-NLS peptide with ∼3 times greater efficacy when WT Impβ was replaced by the B2 mutant. More modest (1.5–1.8-fold) decreases in IC₅₀ were also observed with mutants B1, B3, and B4. These findings reveal that the Asp/Glu→Arg substitutions on Impβ partly compensate for the adverse effects of the Rev Arg→Asp mutations on the ability of Impβ to bind Rev, suggesting a possible electrostatic interaction between the mutated acidic and basic residues (Fig 10D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 10. Compensatory effects between charge-reversal mutants of Impβ and Rev.

(A) Multiple point mutants of Impβ in which 2–4 Asp or Glu residues are replaced by Arg residues. (B) IC₅₀ values obtained from FP inhibition assays measuring the ability of WT Rev to displace a Rev-NLS peptide from WT or mutant forms of Impβ. Results from individual assays are shown together with the mean. (C) The same assay performed with WT and charge-reversal mutant forms of Rev. Data for Rev mutants are colored as in Fig 8. Data shown are mean and SD values from 4 to 14 biological replicates (see also Table S2). (D) Compensatory effects between charge-reversal mutants. An interaction between an acidic residue on Impβ and an Arg residue on Rev is disrupted by a charge-reversal (R→D) mutation of Rev. The interaction is restored by a charge-reversal (D/E→R) mutation on Impβ. (E) Derivation of ΔΔpIC₅₀ values. Top: FP inhibition assays performed with WT Rev (black curves) or the R4 mutant (red curves) together with WT Impβ (circles) or the B2 mutant (diamonds). Bottom: The ability of WT or mutant Rev to displace the Rev-NLS peptide from WT or mutant Impβ is plotted as pIC₅₀ values and as the shift (ΔpIC₅₀) relative to the value observed when WT Rev is assayed with WT Impβ. The ΔΔpIC₅₀ value represents the shift in ΔpIC₅₀ observed when the assay is performed with an Impβ mutant instead of WT Impβ. A positive value of ΔΔpIC₅₀ indicates that the tested Rev mutant more potently displaces the Rev-NLS peptide from the indicated Impβ mutant than from WT Impβ. (F, G) Summary of ΔΔpIC₅₀ values for the indicated combinations of Impβ and Rev proteins. The dotted line at ΔΔpIC₅₀ = 0.1 (corresponding to a 21% decrease in IC₅₀ between the mutant and WT Impβ) indicates the threshold used to identify compensatory mutations. Compensatory effects satisfying this criterion involving single-point mutants of Impβ and Rev are marked by a gray arrow and numbered 1–8. For all eight of these mutant combinations, the mean shift in IC₅₀ value observed with the Impβ single-point mutant relative to WT Impβ was statistically significant according to a Dunnett’s multiple comparisons test (P-values for mutant combinations 1–8 were <0.0001, 0.0002, 0.0001, 0.0022, 0.0368, 0.0084, <0.0001 and 0.0127, respectively). (H) Summary of compensatory interactions involving single Impβ and Rev residues. (G) The numbered arrows correspond to the mutant combinations numbered 1–8 in (G). The thickness of arrows is proportional to the ΔΔpIC₅₀ values of the interacting mutants.

To assess the size of this compensatory effect and facilitate comparison across different combinations of Impβ and Rev mutants, we used the parameter ΔΔpIC₅₀, whose derivation is illustrated for mutants B2 and R4 in Fig 10E. In this figure, replacing WT Rev (solid black circles) by mutant R4 (solid red circles) causes the pIC₅₀ to drop from 6.3 to 5.3, yielding a ΔpIC₅₀ value of −1, whereas replacing both WT Rev and WT Impβ by mutants R4 and B2, respectively (red diamonds), results in a ΔpIC₅₀ of only −0.5. Thus, compared with the former assay, the latter assay causes ΔpIC₅₀ to shift from −1 to −0.5, yielding a ΔΔpIC₅₀ value of +0.5. The ΔΔpIC₅₀ values for selected combinations of Impβ and Rev mutants are shown in Fig 10F and G.

To pinpoint potentially interacting Impβ and Rev residues, we generated single charge-reversal mutations of the nine Impβ residues concerned by mutants B2-B4 and tested these in FP inhibition assays against Rev triple mutants R2-R4 and the corresponding Rev single-point mutants shown in Fig 9D. The results revealed several combinations of Impβ and Rev mutants that gave notable compensatory effects (Figs 10F and G and S7). These effects were generally more significant when either Impβ or Rev contained multiple mutations, as observed for example between Rev mutants R2-R4 and Impβ mutant D288R, between Rev mutant R3 and Impβ mutant D339R, and between Rev mutant R48D and Impβ mutants B3 and B4. Of the 51 pairwise combinations of single-point mutants tested, the strongest compensatory effects were seen for Impβ mutant D288R with Rev mutants R42D and R46D (ΔΔpIC₅₀ = 0.41 and 0.33, respectively) and to a lesser extent R43D (ΔΔpIC₅₀ = 0.17) (Figs 10G and S7A–C). ΔΔpIC₅₀ values exceeding 0.1 were also observed for Impβ mutant E299R with Rev mutants R42D, R43D, and R46D, for Impβ E289R with Rev R46D, and for Impβ E437R with Rev R48D (Figs 10G and S7D–G). These findings identify Impβ acidic residues on the B helices of HEAT repeats 7 and 10 that are likely to be spatially proximal to basic residues on Rev helix α2 (Fig 10H), providing important clues into the relative binding orientations of Impβ and Rev.

• Download figure
• Open in new tab
• Download powerpoint

Figure S7. Compensatory effects between charge-reversal mutants of Impβ and Rev.

Panels (A, B, C, D, E, F, G) show representative individual FP inhibition assays performed with WT Rev (black curves and symbols) or the indicated Rev mutant (colored curves and symbols) together with WT Impβ (circles) or the indicated Impβ mutant (squares, triangles or diamonds).

Molecular docking simulations associate certain compensatory mutations with the C-site of Impβ

The mutual proximity of Rev residues Arg42 and Arg46 on helix α2 (Fig 10H) and the short distances (<18 Å) separating Impβ residue Asp288 from residues Glu299 and Glu437 relative to the width of the Rev helical hairpin (20–25 Å) make it likely that the compensatory effects summarized in Fig 10H are all mediated by the same Rev monomer. However, whether that Rev monomer localizes to the N- or C-site on Impβ is unclear. To explore this question, we performed molecular docking simulations to assess whether certain distance restraints could be attributed reliably to either site. We docked the Rev helical hairpin onto the Impβ surface (using the extended 1UKL conformation as suggested by our SAXS data) with the program HADDOCK, which exploits experimentally derived interaction restraints to guide the docking procedure (102). (For clarity, in the following description of interaction restraints, a superscript is used to indicate whether a residue belongs to either Rev or Impβ). In initial simulations, we specified distance restraints derived from compensatory mutations involving residue pairs Asp288^Impβ:Arg42^Rev and Asp288^Impβ:Arg46^Rev and combined these with BS3 cross-linking distance restraints for Rev bound to the N-site (experiment 1) or C-site (experiment 2) (Table S3A). Both simulations yielded similar scoring statistics (e.g., HADDOCK score, cluster size, buried surface area) for the top-ranked cluster of solutions (Table S3B). These solutions positioned Rev on the inner surface of Impβ next to HEAT repeats 1–8 (experiment 1) or repeats 7–19 (experiment 2), with Rev helix α2 oriented roughly anti-parallel or parallel, respectively, to the Impβ superhelical axis (Fig S8A and B). (We refer hereafter to these two generic Rev orientations as “anti-parallel” or “parallel,” respectively). In both experiments, most of the docking solutions closely resembled the top-ranked solution with only minor variations in Rev orientation, indicating a relatively limited region of configurational space consistent with the docking parameters used (Fig S9A and B).

• Download figure
• Open in new tab
• Download powerpoint

Figure S8. Electrostatic interactions suggested by compensatory mutagenesis data are more likely to involve Rev bound at the C-site than at the N-site.

Panels (A, B, C, D) show the results of HADDOCK docking experiments 1–4, respectively, described in the text. In each case, the lowest-energy structure (LES) for the top-ranking cluster of docking solutions is shown, with Impβ colored from blue to cyan from N- to C-terminus and Rev in violet. In panel (B), the LES from the second ranked cluster is also illustrated, with Rev shown in gray. The pink and blue arrows indicate the orientation (from N- to C-terminus) of Rev helix α2 and the Impβ superhelical axis, respectively. BS3 cross-linking and electrostatic interaction restraints included in each experiment are summarized (see Table S3A for details). (E) Evidence supporting the conclusion that electrostatic interactions deduced from our compensatory mutagenesis data are more compatible with the C-site of Impβ than with the N-site. Experiments 3 and 4 are identical to experiments 1 and 2, respectively, except that the Glu437^Impβ:Arg48^Rev interaction was included as an additional distance restraint. (B, D) Experiments 2 and 4 yielded similar results (Table S3B), except that experiment 4 resulted in a greater number of clustered solutions (189 versus 177) that were more uniform in configuration (compare Fig S9D with Fig S9B) and yielded a top-ranked solution in which residues Glu437^Impβ and Arg48^Rev were closer together (compare panel (D) versus (B)). In contrast, experiment 3 yielded a different outcome compared with experiment 1 (Table S3B), resulting in fewer clustered solutions (84 versus 143) that were more diverse in configuration (compare Fig S9C with Fig S9A). (A, C, D) Indeed, several clusters of solutions (including the top-ranked) positioned Rev next to HEAT repeats 7–19 with helix α2 parallel to the Impβ superhelix, that is, closely resembling the C-site solution for experiment 4 (panels [A, C, D]), even though the BS3 cross-linking restraints used were those associated with the N-site. Moreover, all the solutions that placed Rev in the N-site (clusters ranked second and fourth) positioned residues Glu437^Impβ and Arg48^Rev far apart. These findings suggest that the set of electrostatic interactions used as distance restraints in experiments 3 and 4 are more compatible with the C-site than with the N-site.

• Download figure
• Open in new tab
• Download powerpoint

Figure S9. Results of docking experiments with program HADDOCK.

Panels (A, B, C, D) show the results for experiments 1–4, respectively. For each docking experiment, the 200 lowest-energy binding configurations were grouped into clusters according to structural similarity. Clusters were then ranked according to the average energy of the four lowest-energy structures in each cluster. The four lowest energy structures are shown for each cluster. All members of each cluster are colored according to the color scheme shown at the left. The downward gray arrow indicates the direction of the Impβ superhelical axis. For each cluster, the approximate binding orientation of Rev is designated by indicating the orientation of Rev helix α2 as roughly antiparallel (↑), parallel (↓) or perpendicular (← or →) relative to the Impβ superhelical axis. Rev monomers with parallel or antiparallel orientations are shown as ribbon diagrams, and those with perpendicular orientations are shown as Cα line tracings. Although the inclusion of the additional distance restraint between Glu437^Impβ and Arg48^Rev in experiment 3 results in a greater diversity of cluster configurations relative to experiment 1, its inclusion in experiment 4 results in more uniform clusters relative to experiment 2.

The above experiments included only two of the eight possible distance restraints suggested by our mutagenesis data (Fig 10H) and in particular omitted the distance restraint between residues Glu437^Impβ and Arg48^Rev. Strikingly, however, experiment 2 (with Rev in the C-site) yielded solutions in which these two residues were relatively close to each another, including the second ranked solution which positioned these residues <5 Å apart (Fig S8B). In contrast, all solutions from experiment 1 (with Rev in the N-site) placed these residues far apart and sterically inaccessible to each other (Fig S8A). Repeating these simulations with the Glu437^Impβ:Arg48^Rev interaction explicitly included as an additional distance restraint yielded significantly worse docking results for Rev bound to the N-site (experiment 3) and considerably improved results for Rev bound to the C-site (experiment 4) (Figs S8C and D and S9C and D and Table S3). These findings suggest that the set of electrostatic interactions used as distance restraints in experiments 3 and 4 are more compatible with the C-site than with the N-site (Fig S8E).

To confirm this hypothesis, we performed exhaustive rigid-body sampling of Rev binding orientations on the surface of Impβ within the N-site and C-site envelopes depicted in Fig 7F. The Impβ structure was held fixed, whereas the Rev helical hairpin was rotated and translated in a full 6-dimensional search that maintained Lys20^Rev within BS3 cross-linking distance of the group-1 lysines in HEAT repeats 1 and 2 (in the case of the N-site) or in repeat 19 (in the case of the C-site), yielding >3.5 and >2.8 million configurations, respectively. After excluding configurations with a severe steric overlap between Impβ and Rev, the remaining configurations were used to measure Cβ-Cβ distances between specific Impβ and Rev residues. These measurements showed that the putative interactions illustrated in Fig 10H are individually compatible with Rev bound to either the N- or C-site because docking configurations were identified for both sites that brought the relevant charged residues within salt-bridge distance of each other (Fig S10A). However, no configurations were identified for the N-site in which interactions involving residue pairs Arg42^Rev:Asp288^Impβ and Arg48^Rev:Glu437^Impβ could simultaneously occur (i.e., configurations in which both Cβ-Cβ distances were <14 Å), unlike the case for the C-site, where >80 such configurations were identified (Fig S10B). These findings confirm that the putative interactions depicted in Fig 10H have a much greater probability of being associated with the C-site than with the N-site.

• Download figure
• Open in new tab
• Download powerpoint

Figure S10. Analysis of distances between Impβ and Rev residues in rigid-body docking simulations.

(A) Boxplots of Cβ-Cβ distances between the indicated Impβ and Rev residues for docking simulations involving the N-site (green) or C-site (purple). (B) Scatter plot of docking configurations showing the distribution of Cβ-Cβ distances between residues D288^Impβ and Arg42^Rev (horizontal ordinate) and between residues E437^Impβ:Arg48^Rev (vertical ordinate) for docking simulations involving the N-site (left panel) or C-site (right panel). The number of configurations satisfying the indicated distance cutoffs is indicated in blue. (C) Scatter plot of docking configurations showing the distribution of Cβ-Cβ distances between residues Lys206^Impβ and Lys20^Rev (horizontal ordinate) and between residues Lys537^Impβ:Lys20^Rev (vertical ordinate) for docking simulations involving the N-site (left panel) or C-site (right panel). The number of configurations satisfying the distance cutoffs (indicated by dashed blue lines) is shown in blue. Because the Lys20^Rev Cβ atom was centered on each grid position before applying the rotational search, each point in the plot may correspond to several different (up to 4,056) Rev orientations.

In addition, the rigid-body sampling analysis also shed light on two of the BS3 cross-links (between Lys20^Rev and Lys206^Impβ or Lys537^Impβ) that we previously were unable to assign reliably to either the N- or C-site. The minimal Cβ-Cβ distance measured between residues Lys20^Rev and Lys206^Impβ was 30.1 Å for all C-site configurations, compared with 12.5 Å for N-site configurations (Fig S10A), thereby assigning this cross-link with high probability to the N-site. Among the ∼48,000 N-site configurations compatible with a Lys20^Rev:Lys206^Impβ cross-link, only ∼670 were compatible with a long cross-link between Lys20^Rev and Lys537^Imp (Cβ-Cβ distance between 28.9 and 30 Å) and none with a shorter cross-link (Cβ-Cβ distance <28.9 Å) (Fig S10C). In contrast, the minimal Cβ-Cβ distance for this residue pair was 13.7 Å for C-site configurations (Fig S10A), making it highly likely that the Lys20^Rev:Lys537^Impβ cross-link is associated with the C-site.

Structural model of Rev bound to the Impβ C-site

We next filtered the results of our rigid-body docking experiment for configurations of Rev bound to the Impβ C-site that were compatible with our BS3 cross-linking and compensatory mutagenesis data (Fig 11A). This resulted in a highly uniform set of parallel Rev configurations (Fig 11B and Table S4). To confirm this result, we ran HADDOCK using the BS3 cross-linking and compensatory mutagenesis data as interaction restraints to guide the docking procedure. Nearly all (97%) of the solutions obtained in the final refinement stage of docking yielded parallel configurations belonging to seven very similar clusters (Fig 11C and Table S5) that closely resembled the configurations obtained by rigid body docking (Fig 11D). As expected, both docking approaches yielded models that agreed well with our BS3 cross-linking data (Fig S11A) and placed residues related by compensatory charge-reversal effects in close proximity (Fig S11B). To represent the entire ensemble of configurations obtained by rigid-body sampling and HADDOCK, we calculated an “average” configuration that minimizes the change in orientation and position relative to each of the individual docking solutions obtained (Fig 11E). The resulting model closely resembles the top-ranked configurations obtained by rigid-body sampling and by HADDOCK (Fig S12). Comparing the individual docking configurations to the average model shows that the predicted location (i.e., translational coordinates of the centroid) of Rev on Impβ (Fig 11E, lower inset) is relatively well defined, as are two of the rotational angles (yaw and pitch), whereas the angle about the long axis of Rev (roll angle) is somewhat more variable. A SAXS curve calculated from the average model agreed reasonably well (χ² = 1.86 and 1.17 using Impβ conformations 1UKL and 3W5K, respectively) with the SAXS data measured from the Impβ/Rev^ODΔ complex (Fig S13). Overall, these findings support a C-site Rev binding mode whereby the Rev helical hairpin is embraced by the C-terminal half of Impβ with the long axis of Rev roughly parallel to the Impβ superhelical axis and the Rev CTD positioned near the C-terminal end of the HEAT-repeat superhelix.

• Download figure
• Open in new tab
• Download powerpoint

Figure 11. Structural model of Rev bound to Impβ at the C-site.

(A) BS3 cross-links and electrostatic interactions used as distance constraints for rigid body docking and as interaction restraints in program HADDOCK. (B) Configurations of the Impβ/Rev complex obtained by rigid body docking. The mean variation in Rev orientation and centroid position are 33° and <5 Å, respectively, relative to the top-ranked configuration (see also Table S4). (C) Configurations of the Impβ/Rev complex obtained using HADDOCK. For clarity, only 28 (the four lowest-energy structures in each of the seven clusters) of the 194 clustered solutions are shown. Solutions are colored according to the rank of the corresponding cluster. The mean variation in Rev orientation and centroid position are 34° and 4 Å, respectively (see also Table S5). (D) Superposition showing agreement between docking solutions obtained by rigid body sampling and by HADDOCK. (E) “Average” configuration of the Impβ/Rev complex that minimizes the deviation from models obtained by rigid-body and HADDOCK docking experiments. Compared with the average configuration, the Rev orientation in the individual rigid-body and HADDOCK configurations shows the greatest variation (by a maximum of 52°) in the angle about the long axis of Rev (roll angle) and smaller variations (up to 25–35°) about the two orthogonal angles (yaw and pitch). The approximate location of the Rev CTD is indicated. Upper inset: Principal axes and corresponding rotational angles of Rev. Lower inset: Centroids of Rev monomers obtained in rigid-body (green spheres) and HADDOCK (violet spheres) docking experiments. These centroids differ from that of the average model by a maximum of 5 or 6 Å along each of the principal axes of Rev.

• Download figure
• Open in new tab
• Download powerpoint

Figure S11. Top-ranking configurations from docking experiments.

(A) Plot of solvent-accessible surface (SAS) distances calculated for the top-ranking configuration obtained by rigid-body sampling (top) or HADDOCK (bottom) analysis. SAS distances were calculated between Rev residue Lys20 and Impβ lysine residues belonging to group 1 (red), group 2 (green), or group 3 (blue), as defined in Fig 7B. Group 1 lysines associated with the N- or C-site of Impβ are indicated by open and closed red circles, respectively. SAS distances were calculated using the Jwalk server (101). (B) Orthogonal views of the top-ranking configuration obtained by rigid-body sampling (left) or HADDOCK (right) analysis. Acidic Impβ residues and basic Rev residues hypothesized to be in close proximity on the basis of compensatory charge-reversal mutation data are shown in stick representation. In the bottom panels, Rev helix α1 is omitted for clarity.

• Download figure
• Open in new tab
• Download powerpoint

Figure S12. Comparison of docking solutions.

(A) Top: Structures of the top-ranking solution from the rigid-body docking and HADDOCK experiments are shown and the “average” configuration of all solutions obtained by the two docking methods. Bottom: Table summarizing the relative shift in the centroid position of Rev and its relative rotation (polar angle κ) in pairwise comparisons of the three structures, calculated using program Lsqkab from the CCP4 suite (128). (B) Superposition of the three Rev docking poses shown in (A).

• Download figure
• Open in new tab
• Download powerpoint

Figure S13. Comparison of structural models of the 1:1 Impβ/Rev complex with SAXS data.

(A) Scattering data from the Impβ/Rev^ODΔ complex compared with the scattering profile calculated from the “average” docking configuration of Rev bound to the Impβ C-site. (B) Replacing the Impβ conformation from PDB 1UKL with the slightly more elongated conformation from PDB 3W5K yields an improved agreement between the observed and calculated scattering profiles. (C) Comparison of the two Impβ conformations of PDB 1UKL with 3W5K. The 3W5K conformation is more extended by ∼3 Å along the HEAT-repeat superhelical axis.

Implications for Rev bound at the Impβ N-site

We next wondered how the second Rev monomer might bind to Impβ. Because Rev is known to multimerize through homotypic “A–A” and “B–B” interactions, we structurally aligned both types of Rev homodimer with our average model of Rev bound to the Impβ C-site to see whether either dimer was compatible with the model (Fig S14A and B). Both alignments resulted in an implausible placement of the second Rev monomer: residue Lys20 on Rev helix α1 was very far from group-1 lysines in HEAT repeats 1 and 2, in disagreement with the BS3 cross-linking data; there was insufficient space to accommodate the Rev CTD; and, in the case of the Rev B–B homodimer, a severe steric clash occurred with Impβ (Fig S14B). These findings make it highly unlikely that the second Rev monomer binds Impβ by forming an A–A or B–B interface with the first bound Rev monomer. This conclusion is strongly supported by the ability of Impβ to bind two monomers of the Rev^OD mutant (Figs 2D and E, 3B–D, and 4B), whose ability to form A–A and B–B dimers is compromised (Fig 2A).

• Download figure
• Open in new tab
• Download powerpoint

Figure S14. Configurations for the binding of Rev monomer 2.

(A, B) Unlikely configurations in which Rev monomer 2 (cyan) interacts with Rev monomer 1 (magenta) via an A–A (head-to-head) or B–B (tail-to-tail) interaction. Predicted distances between residue Lys20 on Rev monomer 2 and Impβ residues Lys23, Lys62, and Lys68 are roughly twice the BS3 cross-linking distance constraint. (C) Results of the rigid body docking analysis showing the centroids of Rev monomer 2 that are consistent with BS3 cross-linking constraints associated with the N-site and sterically compatible with Impβ bound to Rev monomer 1 via the C-site. Centroids of Rev molecules within and beyond contact distance of Rev monomer 1 are indicated by blue and green spheres, respectively. (D) Two representative orientations of Rev monomer 2 in which the N-and C-termini of the Rev helical hairpin are near the N-terminal end of Impβ (N-ward) or are located midway between the N- and C-terminal ends (sideward). (E) Hypothetical configuration showing that the binding of Rev to the N- and C-sites of Impβ is compatible with a C–C interaction between the two Rev monomers as observed in PDB 5DHV (67).

For additional insights, we further exploited our rigid-body docking data. We structurally aligned the average model of Rev bound to the Impβ C-site with all Rev docking configurations that were compatible with BS3 cross-links involving the N-site (including the K20^Rev-K206^Impβ cross-link; ∼48,000 configurations; Fig S10C) and retained those exhibiting no, or at most only a mild, steric overlap between the two Rev monomers (∼18,000 configurations). As expected, the resulting Rev monomers all localize to the N-terminal half of Impβ, with approximately one-third located within contact distance of the Rev monomer bound at the C-site (Fig S14C). The latter subset is characterized by two predominant configurations in which the two Rev monomers interact through their α1-α2 loops, with the second Rev monomer adopting either an anti-parallel or a roughly perpendicular (“transverse”) orientation (Fig S14D). Interestingly, in the latter case, the relative orientation of the two Rev monomers closely resembles the C–C homodimer arrangement recently described for Rev (67). Indeed, aligning the structure of a Rev C–C homodimer (PDB 5DHV) with our average model of Rev bound to the C-site yields only a minor steric clash with Impβ that can be relieved by a small rigid-body rotation of the dimer (Fig S14E), suggesting that Impβ could conceivably bind two Rev monomers sharing a C–C interface. A non-exhaustive examination of docking configurations revealed that several models of the 1:2 Impβ:Rev complex, including those with a C–C homodimer or an antiparallel arrangement for the two Rev monomers, are consistent with the SAXS data measured from the Impβ/Rev^ODΔ complex, supporting the plausibility of these models (Fig S15). However, additional data are required to establish the specific binding configuration of the second Rev monomer and whether it interacts with the first.

• Download figure
• Open in new tab
• Download powerpoint

Figure S15. Comparison of structural models of the 1:2 Impβ/Rev complex with SAXS data.

(A, B) Scattering data of the Impβ/Rev^ODΔ complex were compared with scattering profiles calculated for representative docking configurations of the 1:2 Impβ/Rev complex, using Impβ conformations from (A) PDB 1UKL and (B) PDB 3W5K. The placement of Rev monomer 1 is that of the “average” configuration shown in Fig 11E. The placement of Rev monomer 2 corresponds to that of the antiparallel, C–C homodimer and transverse arrangements shown in Fig S14D. For all three arrangements, the 3W5K conformation yielded a better fit compared with 1UKL, consistent with an elongated conformation of the complex. (B) Comparing the models in (B) with that in Fig S13B shows that the presence of a second Rev monomer in the specific antiparallel or C–C dimer arrangement examined yielded an improved or unchanged χ² value relative to that of the corresponding 1:1 Impβ/Rev complex, supporting the plausibility of these configurations for the 1:2 complex, whereas the specific transverse arrangement analyzed gave a poorer fit. However, the possibility that other transverse configurations might also yield a good fit cannot be excluded because only a small fraction of the 18,000 sterically allowed configurations for Rev monomer 2 were examined.