Importin-9 wraps around the H2A-H2B core to act as nuclear importer and histone chaperone

We report the crystal structure of nuclear import receptor Importin-9 bound to its cargo, the histones H2A-H2B. Importin-9 wraps around the core, globular region of H2A-H2B to form an extensive interface. The nature of this interface coupled with quantitative analysis of deletion mutants of H2A-H2B suggests that the NLS-like sequences in the H2A-H2B tails play a minor role in import. Importin-9•H2A-H2B is reminiscent of interactions between histones and histone chaperones in that it precludes H2A-H2B interactions with DNA and H3-H4 as seen in the nucleosome. Like many histone chaperones, which prevent inappropriate non-nucleosomal interactions, Importin-9 also sequesters H2A-H2B from DNA. Importin-9 appears to act as a storage chaperone for H2A-H2B while escorting it to the nucleus. Surprisingly, RanGTP does not dissociate Importin-9•H2A-H2B but assembles into a RanGTP•Importin-9•H2A-H2B complex. The presence of Ran in the complex, however, modulates Imp9-H2A-H2B interactions to facilitate its dissociation by DNA and assembly into a nucleosome.


Introduction
Eukaryotic chromatin is organized into nucleosomes, which are structural and functional units that are composed of 147 base pairs of DNA wrapped around two H3-H4 dimers and two H2A-H2B dimers (Luger et al., 1997). Nucleosomes are assembled in the nucleus during S-phase as new H2A, H2B, H3 and H4 proteins are synthesized in the cytoplasm (Adams and Kamakaka, 1999;Annunziato, 2012;Verreault, 2000). Newly translated histones are folded and assembled into H2A-H2B and H3-H4 dimers, which are then imported into the nucleus for deposition onto replicating chromatin. Despite their small sizes, histones do not diffuse into the nucleus but are transported by nuclear import receptors of the Karyopherin-b family termed importins (Baake et al., 2001;Jäkel et al., 1999;Johnson-Saliba et al., 2000;Mosammaparast et al., 2002b;Mosammaparast et al., 2001;Mühlhäusser et al., 2001).
Core histones H2A, H2B, H3 and H4 all contain disordered N-terminal tails followed by small histone-fold domains; H2A also has a disordered C-terminal tail (Luger et al., 1997). The N-terminal tails of histones contain many basic residues, somewhat resembling classical NLS motifs eLife digest Cells contain two meters of DNA which, if left to its own devices, would soon end up in a knot. To keep things organized, the genetic code is wrapped around protein 'spools' called histones, meaning it can all fit within a part of the cell known as the nucleus. The cell makes a copy of its DNA every time it divides, and this copy needs a new set of histones to keep it tidy. The machinery required to construct new histones sits outside the nucleus and getting the histones into position in the nucleus can be a challenge. Histones have a positive charge, which helps to keep the negatively charged DNA wound around the spool. Yet without supervision, histones can stick to other charged molecules in the cell and cause blockages.
The proteins responsible for histone transport are called importins. These proteins normally recognize their cargo by molecular patterns called "nuclear localization signals". These patterns work like a postal address, telling the importin to take the cargo into the nucleus. When they arrive at their destination, another protein called Ran interacts with the importins to release the cargo. Strangely, removing the predicted address pattern from histones does not stop them getting to the nucleus. To find out what was going on, Padavannil et al. solved the three-dimensional structure of an importin bound to a pair of histones via a technique called X-ray crystallography. This made it possible to see how the proteins fit together.
The structure revealed that, rather than interact with the predicted address pattern, the importin wraps around the core of the histones. This blocks the positive charges, stopping the histones sticking to other molecules on their way to the nucleus. The next challenge was to find out how the cell unhooks the histone cargo from the importin when it arrives in the nucleus; with the positive charges covered by the importin, the histones could not stick to the DNA. Yet, something changed when the levels of Ran were high. Rather than unhook the histone, Ran joined the importin-histone complex. This then made it possible for the histones to attach to DNA, helping them to get into position without sticking to the wrong molecules.
These findings form the first step in understanding how the cell transports sticky histones without getting in a knot. The next step is to find out whether these interactions, shown in test tubes, happen in the same way inside living cells. (Blackwell et al., 2007;Ejlassi-Lassallette et al., 2011;Johnson-Saliba et al., 2000;Marchetti et al., 2000;Greiner et al., 2004;Mosammaparast et al., 2001;Moreland et al., 1987). H2A and H2B tails are able to target heterologous proteins into the nucleus (Mosammaparast et al., 2001), but removal of the tails does not abolish localization of H2A-H2B in the nucleus (Thiriet and Hayes, 2001). Furthermore, analysis of seven different importins binding to H3 and H4 tails vs. full-length H3-H4 vs. H3-H4 . Asf1 chaperone complex suggested that specificities for importin-binding reside not only in the tail 'NLSs' but also in the histone folds and the bound chaperone (Soniat et al., 2016).
Here, we solved the crystal structure of Imp9 bound to the full-length H2A-H2B dimer to understand how histones are recognized for nuclear import. The superhelical Imp9 wraps around the histone dimer. Most of the N-terminal tails of both H2A and H2B are disordered, and only five residues of the H2B tail contact Imp9. Binding of Imp9 blocks DNA and H3-H4 sites on H2A-H2B, and Imp9 prevents H2A-H2B from aggregating on DNA, consistent with a histone chaperone-like activity for Imp9. Unlike other importin-cargo complexes, RanGTP does not dissociate Imp9 . H2A-H2B but binds the complex and enhances its dissociation by DNA. The Ran.Imp9.H2A-H2B complex is also able to promote H2A-H2B assembly into nucleosomes. Formation of the Ran . Imp9 . H2A-H2B complex appears to modulate importin-histone interactions to facilitate histone deposition to nuclear targets such as the assembling nucleosome.

Results
Structure of the Imp9.H2A-H2B complex The major nuclear importer for H2A-H2B in S. cerevisiae is Kap114 (Mosammaparast et al., 2002b;Mosammaparast et al., 2001). Imp9, the human homolog of Kap114, was previously shown to bind and import H2A-H2B (Jäkel et al., 2002a;Kimura et al., 2017;Mühlhäusser et al., 2001). We show Imp9-histone interactions in immunoprecipitation from the cytoplasmic fraction of a stable HeLa cell line expressing mCherry-H2B ( Figure 1A). We also show by fluorescence microscopy that Imp9 in these cells localizes mostly to the cytoplasm ( Figure 1B). Similar cytoplasmic localization of Imp9 was reported in the Human Protein Atlas (Thul et al., 2017;Uhlen et al., 2017). To understand how Imp9 recognizes histones for nuclear import, we solved the crystal structure of human Imp9 bound to full-length X. laevis H2A-H2B (dissociation constant, K D = 30 nM; Table 1 and Figure 1-figure supplement 1) by single wavelength anomalous dispersion to 2.7 Å resolution (Figure 1-source data 1).
Imp9 is made up of twenty tandem HEAT repeats, each containing two antiparallel helices A and B that line the convex and concave surfaces of superhelical-shaped protein, respectively ( Figure 1C and . The Imp9-bound H2A-H2B has a canonical histone-fold as in nucleosomes (151 Ca atoms aligned, r.m.s.d. 0.505 Å ; PDB ID 1AOI) (Luger et al., 1997). In our structure, the N-terminal and C-terminal tails of H2A (residues 1-16, 101-130) and H2B (125)(126), the first 14 residues of Imp9 and its H19loop (residues 936-996) were not modeled due to missing electron density.
The N-and C-terminal HEAT repeats of Imp9 (Interfaces 1 and 3) clamp the histone-fold domain while the inner surface of central HEAT repeats 7-8 (Interface 2) interacts with a five-residue segment of the H2B N-terminal tail ( Figure 2). Interface 1 on Imp9 comprises the loop that follows helix 2B and the last turns of helices 3B, 4B and 5B ( Figure 2B, Figure 2-figure supplements 1A, 2A and 3A). Hydrogen-bonding with H2A-H2B residues caps the C-terminal ends of these Imp9 B helices (Figure 2-figure supplement 1D). Of note is the end-to-end capping of the last turn of Imp9 helix 4B by the first turn of histone H2B helix a2. Interface 1 on the histones involves a2-L2-a3 of H2A and a1-L1-a2 of H2B, which constitute a significant portion of the basic DNA-binding surface found in nucleosomes. Although histones and Imp9 surfaces at this interface are electrostatically complementary (  Interactions between Imp9 and H2A-H2B in the cell and crystal structure of the Imp9 .H2A-H2B complex. (A) Coimmunoprecipitation (CoIP) studies of H2B mCherry from whole cell, cytoplasmic and nuclear fractions of the lysates from HeLa cells stably expressing H2B mCherry , followed by immunoblots with Imp9, Ran, RFP antibodies. PCNA and MAb414 antibodies are used as loading control antibodies. 10 mg of 1.5 mg lysates are analyzed as CoIP input. Blots are representative of three identical experiments. (B) Subcellular localization of Imp9 and Ran in Hela::H2B mCherry cells. HeLa cells were fixed, permeabilized, incubated with affinity-purified rabbit polyclonal Imp9 antibody and mouse monoclonal anti-Ran antibody, and visualized by confocal microscopy. The secondary antibodies were Alexa 488 conjugated anti-rabbit and Alexa 405 conjugated anti-mouse, respectively. The column on the right contains two-color merge images. (C). The crystal structure of human Imp9 (blue) in complex with X. laevis H2A (yellow)-H2B (red). DOI: https://doi.org/10.7554/eLife.43630.003 The following source data and figure supplements are available for figure 1: Source data 1. Data collection and refinement statistics, Imp9.H2A-H2B structure. DOI: https://doi.org/10.7554/eLife.43630.006 Interface 2 involves Imp9 helices 7B, 8B and the H8loop (connects helices 8A to 8B) binding to the short 28 KKRRK 32 segment of the H2B N-terminal tail ( Figure 2C and Figure 2-figure supplements 1B, 2B-C and and 3B). Electron densities for H2B 28 KKRRK 32 are weak (see Figure 2-figure supplement 2B-C) and atomic displacement parameters ('B-factors') for H2B residues 28-32 are also high (>100 Å 2 ), suggesting dynamic interactions. Charged H2B side chains make electrostatic interactions with several acidic Imp9 residues, while the aliphatic part of these basic side chains and their backbone participate in hydrophobic interactions.
Interface 3 involves the last three HEAT repeats of Imp9, specifically the last turn of helix 18A and the short loop that follows, the H18-19loop, the C-terminal half of helix 19A and the first turn of helix 20B ( Figure 2D, Figure 2-figure supplements 1C, 2D and 3C). Instead of the typical basic H2A-H2B residues interacting with the acidic Imp9 residues, charges at Interface 3 are reversed (Figure 1-figure supplement 2B). A basic patch formed by the Imp9 H18-19loop and nearby helices complement an acidic surface on the histones formed by residues from H2A helices a2 and aC, and the C-terminal half of H2B that comprises a2-a3-aC. Of note here are salt bridges between Imp9 residue Arg898 and several acidic residues of H2A ( Figure 2D). Many hydrophobic contacts are also found at this interface, and several helices (Imp9 H18A, H19A and histone H2B a2) are capped through hydrogen-bonding with partner proteins ( Distribution of binding energy in the Imp9 . H2A-H2B complex We analyzed the distribution of binding energy of the extensive Imp9-H2A-H2B interface through mutagenesis of the N-terminal histone tails and several long Imp9 loops and determined K D s of the mutants using isothermal titration calorimetry (ITC; Table 1 and Figure 1-figure supplement 1). Imp9 binds full-length H2A-H2B with high affinity (K D = 30 nM). We did not make mutations to Interface 1 because of the many main-chain interactions found there (Figure 2-figure supplement 1D). Interface 2 involves the H8 loop of Imp9 and the N-terminal tail of H2B, both of which are convenient for deletion mutagenesis. Similarly, two long Imp9 loops (H18-19loop and H19loop) in Interface 3 are convenient for deletion mutagenesis. H2A-H2B mutant assembled with the core of H2A (residues 14-119) and full-length H2B, hence named H2ADTail-H2B, has similar binding affinity (K D = 40 nM) as full-length H2A-H2B. This result is consistent with structural observations that H2A residues in its N-and C-terminal tails are disordered and likely do not contact Imp9. Removal of the H2B tail (deleting residues 1-35), generating mutant H2A-H2BD(1-35), also did not affect binding affinity (K D = 40 nM). This is not surprising given the weak electron density and high B-factors of H2B 28 KKRRK 32 bound to Imp9 in Interface 2 (Figure 2- figure supplement 2). A H2A-H2B mutant dimer with only the core domain (H2A residues 14-119 complexed with H2B residues 25-123; named H2ADTail-H2BDTail) also bind as tightly to Imp9 as the full-length histones (K D = 40 nM). Removal of the Imp9 H8loop (Imp9DH8loop), which forms part of the binding site for H2B 28 KKRRK 32 , also did not decrease binding (K D , Imp9DH8loop = 10 nM; Table 1, Figure 1-figure supplement 1E). The histone tails thus do not contribute much binding energy for interactions with Imp9.
At Interface 3, the basic H18-19loop of Imp9 contacts the acidic patch of the histones while the nearby H19loop is mostly disordered and its contribution to histone binding is uncertain. Removal of the H18-19loop reduced the affinity 15-fold (K D = 450 nM; Table 1, Figure 1-figure supplement 1F). We note the endothermic binding reaction that occurred upon truncation of this 40-residue loop. This result suggests substantial contribution of Interface three to the total binding energy. Removal of the H19loop did not affect affinity (K D = 40 nM; Table 1, Figure 1-figure supplement 1G), suggesting that this disordered loop does not participate in H2A-H2B binding.
Histone chaperones are a class of functionally, structurally and mechanistically diverse histonebinding proteins that 'chaperone' histones to protect them from promiscuous DNA-histone interactions (Elsässer and D'Arcy, 2012; Mattiroli et al., 2015) in many different contexts surrounding the formation of nucleosomes (Laskey et al., 1978). The observation that Imp9 buries more surface area on H2A-H2B than well-characterized histone chaperones raises the question of whether Imp9 might also function as a histone chaperone. This function is manifested biochemically by the protein outcompeting DNA from non-nucleosomal DNA . H2A-H2B complexes (Andrews et al., 2010;Andrews et al., 2008;Hondele et al., 2013;Hong et al., 2014). To test if Imp9 can compete H2A-H2B from DNA like histone chaperone Nap1, we performed native gel-based competition assays. Titration of Nap1 or Imp9 against DNA . H2A-H2B complexes leads to the release of free DNA as Nap1 or Imp9 binds H2A-H2B ( Figure 3D,E). These results suggest that Imp9 can act as a histone chaperone by shielding H2A-H2B from promiscuous interactions while it accompanies the histones from the cytoplasm to the nucleus.
We compared the Imp9.H2A-H2B structure with the structures of different importins bound to RanGTP, to predict the Ran-binding site on Imp9. In these structures, RanGTP is always sandwiched between N-terminal and either central or C-terminal HEAT repeats of the importins (Figure 4-figure supplement 4). Importin-RanGTP interactions at the first four HEAT repeats of importins (binding Switch 1, Switch two and a3 of RanGTP) appear to be structurally conserved even though the interface on the opposite side of RanGTP involves different central or C-terminal HEAT repeats in different importins (Chook and Blobel, 1999;Kobayashi and Matsuura, 2013;Lee et al., 2005;Tsirkone et al., 2014;Vetter et al., 1999). Structural alignment of HEAT repeats 1-4 of Imp9 with HEAT repeats 1-4 of Importin-b(1-462) . RanGTP (PDB ID 1IBR (Vetter et al., 1999); r.m.s.d. of 152 Cas in the alignment is 3.27 Å ), Kap95.RanGTP (2BKU (Lee et al., 2005) r.m.s.d. of 152 Cas in the alignment is 3.20 Å ), Kapb2.RanGTP (1QBK (Chook and Blobel, 1999)  Structural alignment of HEAT repeats 1-4 of Imp9 and Kap121.RanGTP allows us to predict the RanGTP binding site at the N-terminus of Imp9 (Figure 4-figure supplement 6A,B). The prediction is supported by an Imp9 mutant with HEAT repeats 1-3 removed that no longer binds RanGTP (Figure 4-figure supplement 6C-E). This likely Ran-binding site at the N-terminus of Imp9 appears separate from but adjacent to the H2A-H2B binding site (Figure 4-figure supplement 6A,B). The GTPase can most likely access Imp9 without dislodging H2A-H2B but proximity of RanGTP to the histones could modulate Imp9-histones interactions especially the kinetics of binding.

Discussion
The solenoid-shaped Imp9 wraps around the folded globular domain of the H2A-H2B dimer, leaving most of the N-terminal tails of H2A, H2B and the C-terminal tail of H2A disordered in the complex. Only the 5-residue 28 KKRRK 32 segment of the H2B tail contacts Imp9 even though weak electron density and high atomic displacement parameters of the H2B tail suggests that these interactions are dynamic. Our structural observations that Imp9 binds mostly to the globular domain of the H2A-H2B are also consistent with the lack of effect in Imp9 binding when either or both histone tails are deleted (Table 1), and with the previously reported nuclear localization of a mutant of H2A-H2B that lacks both its N-terminal tails (Thiriet and Hayes, 2001). However, very weak dynamic/fuzzy longrange electrostatic interactions between Imp9 and histones tails may still exist -we had previously reported very weak and dynamic interactions between an importin-cargo pair by NMR that could not be observed by X-ray crystallography or detected in mutagenesis/ITC experiments (Yoshizawa et al., 2018). Nevertheless, H2A-H2B thus belongs to a small category of nuclear import cargos that mostly use surfaces of folded domains rather than extended linear nuclear import/localization motifs to bind their importins (Aksu et al., 2016;Bono et al., 2010;Cook et al., 2009;Grünwald and Bono, 2011;Grünwald et al., 2013;Matsuura and Stewart, 2004;Okada et al., 2009).
Gö rlich and colleagues proposed in 2002 that negatively charged importins act as chaperones toward positively charged cargo proteins like histones (Jäkel et al., 2002b). Others also suggested importins acting as chaperones (Lusk et al., 2002). We provide structural evidence to support this proposal as the mostly negatively charged Imp9 indeed shields the mostly positively charged histone-fold domain of H2A-H2B, and perhaps also dynamically shields the extended basic histone tails. The ability of Imp9 to chaperone H2A-H2B, however, goes beyond charge shielding. Despite overall charge complementarity, there are only a few salt bridges at the Imp9 . H2A-H2B interface, which also employs hydrophobic interactions and hydrogen bonds, many involving main chains of both proteins. Imp9 also shields many hydrophobic patches on H2A-H2B. The interaction further involves a charge reversal where a basic surface at the C-terminal end of Imp9 interacts with the acidic patch of H2A-H2B. The extensive and persistent interactions that allows Imp9 to surround and shield H2A-H2B also differ significantly from the recently revealed chaperoning interactions of another importin, that of Kapb2 (or Transportin-1) with the Fused in Sarcoma protein (FUS). Kapb2-FUS interactions are anchored through high affinity binding at the 26-residue PY-NLS linear motif of FUS that then enable weak, distributed and dynamic interactions with multiple mostly intrinsically disordered regions of FUS, to block formation of higher-order FUS assemblies and liquid-liquid phase separation (Yoshizawa et al., 2018).
The way the Imp9 solenoid wraps around H2A-H2B leaves the predicted N-terminal Ran-binding site of Imp9 accessible and ready to bind RanGTP. We showed by pull-down, electrophoretic mobility shift, size exclusion chromatography, analytical ultracentrifugation and SAXS experiments that Imp9 binds both the histones and RanGTP simultaneously and stably, suggesting that unlike most importin-cargo complexes, Imp9 . H2A-H2B is unlikely to be dissociated by RanGTP alone upon entering the nucleus. This finding is not without precedence as Pemberton and colleagues previously showed an assembly that contains Kap114, H2A-H2B, RanGTP and the histone chaperone Nap1 (Mosammaparast et al., 2002a). Unlike the Pemberton study, which found the complex containing RanGTP, histones and importin to be intact in the yeast nucleus, we do not detect interactions between Imp9 and H2A-H2B in the nucleus even though that interaction is easily observed in the cytoplasm ( Figure 1A,B). Imp9 is likely dissociated from histones soon after the complex enters the nucleus.
We showed that RanGTP changes the interactions between Imp9 and H2A-H2B as it forms the RanGTP.Imp9.H2A-H2B complex. DNA competes effectively with RanGTP.Imp9.H2A-H2B to produce Imp9 . RanGTP and DNA . H2A-H2B even though it is unable to extract H2A-H2B from Imp9.H2A-H2B. Furthermore, RanGTP.Imp9.H2A-H2B is better at promoting H2A-H2B deposition to assemble nucleosome than either Nap1 . H2A-H2B or no chaperone, while Imp9 alone cannot deposit H2A-H2B. The GTPase in the RanGTP.Imp9.H2A-H2B complex appears to modulate Imp9-H2A-H2B interactions to facilitate histone release and nucleosome assembly. Accessibility of the N-terminal HEAT repeats of Imp9 in the histones complex may allow formation of the RanGTP.Imp9.H2A-H2B complex, but proximity of the Ran and histones binding sites coupled with the flexibility of the HEAT repeats architecture of Imp9 and the propensity for conformational changes likely changed the kinetics of Imp9-histone binding.
Although histones can be deposited by RanGTP.Imp9.H2A-H2B onto DNA or the tetrasome, it remains unclear how H2A-H2B is released from Imp9 in cells. Assembling nucleosomes may release H2A-H2B from RanGTP.Imp9.H2A-H2B or the histones may be passed to another histone chaperone or nucleosome assembly factor as part of a chaperone hand-off cascade in the nucleus. These questions and the one regarding potential additional roles for Imp9 in the cytoplasm are topics for future studies.
Wild type and mutant Xenopus histones H2A, H2B proteins were expressed individually in E.coli BL21 DE3 plysS cells, which were lysed by sonication. The lysate centrifuged at 16000 rpm and the washed pellet was resuspended in unfolding buffer (7 M guanidinium HCl, 20 mM Tris HCl, pH 7.5, 10 mM DTT) and dialyzed overnight in SAU-200 buffer (7 M urea, 20 mM sodium acetate, pH 5.2, 200 mM NaCl, 1 mM EDTA, 5 mM b-mercaptoethanol). The unfolded histone protein samples were further purified with cation exchange chromatography in SAU buffer (200-600 mM NaCl) followed by dialysis overnight in cold water. Mutant H2BD(1-35) was purified as described above. Mutant proteins H2ADTail (contains residues 14-119 of H2A) and H2BDTail (contains residues 25-123 of H2B) used for Isothermal titration calorimetry were obtained from The Histone Source (Colorado, United States).
Ran (Gsp1 (1-179, Q71L)) and MBP-Ran were expressed in E.coli BL21 (DE3) cells as His 6 -tag proteins (induced with 0.5 mM IPTG for 12 hr at 20˚C). Harvested cells were lysed with the Emulsi-Flex-C5 cell homogenizer (Avestin, Ottawa, Canada) and the proteins purified by affinity chromatography on Ni-NTA column. Eluted proteins were loaded with GTP, and RanGTP and MBP-RanGTP were further purified by cation exchange chromatography followed by exchanging into buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, 4 mM magnesium acetate, 1 mM DTT, 10% glycerol (Chook and Blobel, 1999;Fung et al., 2015). Imp9 . H2A-H2B complex assembly, crystallization, crystal structure determination Purified Imp9 was mixed with 10-fold molar excess of H2A-H2B in gel filtration buffer (20 mM HEPES (pH 7.3), 110 mM potassium acetate, 2 mM magnesium acetate, 2 mM DTT, 15% glycerol). Imp9.H2A-H2B was separated from excess histones by size-exclusion chromatography and concentrated to 18 mg/ml for crystallization. Selenomethionyl-labeled Imp9 was expressed as described previously (Doublié, 1997) and purified as for Imp9. Selenomethionyl-Imp9 . H2A-H2B complex was assembled as for the native complex. Initial crystals were obtained by the sitting drop vapor diffusion method from commercial screens (reservoir solution -40 mM MES pH 6.5, 3 M potassium formate, and 10% glycerol) and were further optimized by the hanging drop vapor diffusion method. Crystals were cryoprotected in reservoir solution that was supplemented with 15% glycerol, and flash frozen in liquid nitrogen. Selenomethionyl-Imp9.H2A-H2B crystals were obtained in the same conditions as native crystals and were prepared similarly for crystallographic data collection.
Imp9 . H2A-H2B native crystals diffracted to a minimum Bragg spacing (d min ) of 2.70 Å and exhibited the symmetry of space group P2 1 2 1 2 with cell dimensions of a = 127.4 Å , b = 223.3 Å , c = 131.8 Å and contained two heterotrimers per asymmetric unit. All diffraction data were collected at beamline 19-ID (SBC-CAT) at the Advanced Photon Source (Argonne National Laboratory, Argonne, Illinois, USA) and processed in the program HKL-3000 (Minor et al., 2006) with applied corrections for effects resulting from absorption in a crystal and for radiation damage Otwinowski et al., 2003), the calculation of an optimal error model, and corrections to compensate the phasing signal for a radiation-induced increase of non-isomorphism within the crystal (Borek et al., 2010;Borek et al., 2013). These corrections were crucial for successful phasing and stable model refinement. Crystals of Imp9.H2A-H2B displayed mildly anisotropic diffraction. To minimize radiation damage and maximize redundancy, native data was collected in two separate scans of 125 degrees for a total of 250 degrees by translating a single crystal in the X-ray beam. Analysis of the self-Patterson function calculated with the native data revealed a significant off-origin peak at approximately (1/2, 1/2, 1/2) and 27% the height of the origin peak, indicating translational pseudosymmetry.
Phases were obtained from a single wavelength anomalous dispersion (SAD) experiment using the selenomethionyl-Imp9 . H2A-H2B protein with data to 2.65 Å . Fifty-four selenium sites were located, phases improved and an initial model containing over 50% of all Imp9.H2A-H2B residues was automatically generated in the AutoBuild routine of the Phenix (Adams et al., 2010) program suite. Completion of this model was performed by manual rebuilding in the program Coot (Emsley et al., 2010). Positional and isotropic atomic displacement parameter (ADP) as well as TLS ADP refinement of native Imp9.H2A-H2B with NCS restraints was performed to a resolution of 2.70 Å using the Phenix program suite with a random 2.1% of all data set aside for an R free calculation. The final model for Imp9 . H2A-H2B (R work = 20.9%, R free = 24.0%) contained 2275 residues and 356 waters. The relatively high R work and R free values are likely due to the presence of translational pseudosymmetry. A Ramachandran plot generated with the program MolProbity  indicated that 97.1% of all protein residues are in the most favored regions and 0.1% in disallowed regions. Illustrations were prepared with PyMOL (Schrodinger LLC, 2015). Data collection and structure refinement statistics are summarized in Figure 1-source data 1.

Quantification of binding affinities by isothermal titration calorimetry (ITC)
Imp9 and mutant Imp9 proteins were expressed and purified as described above. The wild type fulllength H2A, H2B and mutant H2BD(1-35) proteins were purified as described above. Mutant H2ADTail and H2BDTail proteins were obtained from Histone Source (Colorado, USA). H2A-H2B, H2ADTail -H2B, H2A-H2BD(1-35) and H2ADTail-H2BDTail heterodimers were reconstituted and purified as described above. Imp9 or mutant Imp9 proteins and H2A-H2B or H2A-H2B mutant dimers were dialyzed in ITC buffer containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM TCEP and 5% glycerol. ITC experiments were carried out using ITC-200 calorimeter (Microcal, LLC, Northampton, MA, USA) at 20˚C with 0.035 mM of Imp9 or mutant Imp9 protein in the sample cell and 0.35 mM H2A-H2B or mutant H2A-H2B protein in the syringe. All samples were thoroughly degassed and then centrifuged at 16000 g for 10 min to remove precipitates. 21 injections each of 1.9 ml except for the first (0.5 ml) were sequentially made in each experiment. The injections were mixed at 300 rpm and consecutive injections were separated by 300 s to allow the peak to return to baseline. All experiments were carried out in triplicates. Data were integrated and baseline corrected using NITPIC (Keller et al., 2012). The baseline corrected and integrated data were globally analyzed in SEDPHAT (Houtman et al., 2007;Zhao et al., 2015) using a model considering a single class of binding sites. SVD-reconstructed thermogram provided by NITPIC, the fit-isotherms and the residuals from SEDPHAT were all plotted using GUSSI (Brautigam, 2015). Individual experiments in the triplicate sets are differently color-coded in Figure 1-figure supplement 1A. For error reporting, we used F-statistics and error-surface projection method to calculate the 68.3% confidence intervals of the fitted data (Bevington). The K D (nM), DH (kCal/mol), DS (Cal/mol.K), DG (kCal/mol) and the Imp9 local concentration correction factors for each set of triplicate experiments are reported in the Table 1.

Pull-down binding assays
Pull-down binding assays were performed by immobilizing purified MBP-Imp9 or MBP-RanGTP (S. cerevisiae Gsp1(1-179/Q71L) on amylose resin (New England BioLabs, Ipswich, MA). 40 ml of 100 mM MBP-Imp9 or MBP-RanGTP was immobilized on 200 ml of amylose resin (50% slurry) in binding assay (BA) buffer containing 20 mM HEPES pH 7.3, 110 mM potassium acetate, 2 mM magnesium acetate, 2 mM DTT and 15% glycerol. 100 ml of~20 mM of immobilized MBP-Imp9 resin was incubated with 100 ml of 400 mM of purified H2A-H2B in a total reaction volume of 400 ml for 30 min at 4˚C, followed by five washes each with 400 ml BA buffer. 100 ml of~20 mM of MBP-RanGTP resin were incubated with 100 ml of 50 mM of purified Imp9 . H2A-H2B in a total volume of 400 ml for 30 min at 4˚C, followed by five washes each with 400 ml BA buffer.
For RanGTP dissociations assays, a gradient of 10 ml, 20 ml, 40 ml or 60 ml of approximately 500 mM purified RanGTP was added to 50 ml of~20 mM of immobilized MBP-Imp9 that were pre-bound with H2A-H2B, in a total reaction volume of 400 ml. These binding reactions contain 12.5 mM, 25 mM, 50 mM or 75 mM of RanGTP added to 2.5 mM MBP-Imp9 . H2A-H2B. Binding was followed by five washes each with 400 ml of the BA buffer. From each of the reactions, 30 ml of beads after final wash was suspended in 30 ml of BA buffer. 10 ml of the bead slurry sample was analyzed on 12% SDS-PAGE gels and stained with Coomassie Blue for visualization. A control experiment involving immobilized GST-Kapb2, MBP-PY-NLS (PY-NLS of hnRNP A1), and a gradient of 12.5 mM, 25 mM, 50 mM and 75 mM of RanGTP (prepared as described above for the MBP-Imp9.H2A-H2B experiments) was carried out similarly to show that RanGTP dissociates PY-NLS bound to Kapb2. 2% of the input Ran-GTP added in each of the binding reactions and approximately 2% of flow-through is also shown in the Coomassie-stained SDS-PAGE gels.
Pull-down binding assay to probe Ran binding to Imp9 versus Imp9D1-144 were performed by immobilizing GST-Imp9 or GST-Imp9D1-144 on Glutathione Sepharose 4B resin (GE Healthcare Life Sciences). 12.5 ml of lysate from 500 ml cell culture (OD 600 = 1) pellet of E. coli expressing GST-Imp9 or GST-Imp9D1-144 (containing~8 mg/ml of GST-Imp9 protein) were incubated on 1 ml of 50% Glutathione Sepharose 4B slurry in BA buffer. The GST-Imp9 or GST-Imp9D1-144 bound resin was washed five times, each with 6 ml BA buffer, before the binding assay. 200 ml of 50% slurry GST-Imp9 or GST-Imp9D1-144 resin (~12 mM proteins) was incubated with 10 ml of~500 mM RanGTP in a total reaction volume of 400 ml for 30 min at 4˚C, followed by five washes (each with 400 ml BA buffer). After washing, 30 ml of 50% beads slurry was suspended in 30 ml BA buffer. 10 ml of the resulting bead slurry sample was analyzed by Coomassie-stained SDS-PAGE. A control experiment using empty GSH sepharose beads and RanGTP was performed as described above.

Analytical ultracentrifugation
The sedimentation coefficients of individual proteins and protein complexes in the mixture were estimated by monitoring their sedimentation properties in a sedimentation velocity experiment carried out in Beckman-Coulter Optima XL-1 Analytical Ultracentrifuge (AUC). The individual proteins and mixtures of proteins were analyzed in AUC buffer containing 20 mM HEPES pH 7.3, 200 mM sodium chloride, 2 mM magnesium chloride, 2 mM TCEP and 8% glycerol (details below). Protein samples (450 ml) and AUC buffer (450 ml) were loaded into a double sector centerpiece and centrifuged in an eight-hole An-50Ti rotor to 50000 rpm at 20˚C. The double sectors were monitored for absorbance at 280 nm (A 280 ). A total of 140 scans were collected and the first 80 scans were analyzed. Buffer density, viscosity of the buffer and partial specific volume of the protein was estimated using SEDN-TERP (http://www.rasmb.bbri.org/software/windows/sednterp-philo/). Sedimentation coefficient distributions c(s) (normalized for absorption differences) were calculated by least squares boundary modeling of sedimentation velocity data using SEDFIT program (Schuck, 2000). Sedimentation coefficients sw (weighted-average obtained from the integration of c(s) distribution) and frictional ratios f/f 0 were obtained by refining the fit data in SEDFIT (Schuck, 2000). For error reporting, F-statistics and Monte-Carlo for integrated weight-average s values were used (Bevington). Data were plotted using GUSSI (Brautigam, 2015).

Native gel shift assays Electrophoretic Mobility Shift Assays
One protein component was held constant at 10 mM and the other was titrated. Samples were separated by 5% polyacrylamide gel electrophoresis. Gels were run for 100 min at 150 V at 4˚C in 0.5x TBE (40 mM Tris-HCl pH 8.4, 45 mM boric acid, 1 mM EDTA). Gels were stained with Coomassie Blue.
All protein samples for SAXS were expressed and purified as described above. Purified Imp9 was exchanged into SAXS buffer (20 mM HEPES pH 7.3, 110 mM potassium acetate, 2 mM magnesium acetate, 2 mM DTT, and 10% glycerol) by SEC and concentrated to 5 mg/ml (43 mM of Imp9) for SAXS analysis. The Imp9.H2A-H2B was purified as described above and then exchanged into SAXS buffer by SEC and concentrated to 5 mg/ml (35 mM of Imp9.H2A-H2B) for SAXS analysis. To prepare the Imp9 . RanGTP complex, previously purified Imp9 was first mixed with 5-fold molar excess of RanGTP for SEC to separate the Imp9 . RanGTP complex from excess Ran. This Imp9 . RanGTP complex was then buffer exchanged into SAXS buffer in another round of SEC and concentrated to 5 mg/ml (37 mM of Imp9.RanGTP) for SAXS. To prepare the RanGTP.Imp9.H2A-H2B complex, previously purified Imp9 . H2AH2B was mixed with 5-fold molar excess of RanGTP in SAXS buffer for SEC to separate RanGTP . Imp9 . H2A-H2B from excess RanGTP. Fractions containing RanGTP . Imp9 . H2A-H2B were pooled, concentrated and subjected to a second round of SEC in SAXS buffer, after which the complex was concentrated to 5 mg/ml (31 mM of RanGTP.Imp9.H2A-H2B) for SAXS. The 10% glycerol in the SAXS buffer protects the protein samples from radiation damage during X-ray exposure (Kuwamoto et al., 2004) and our early studies show that low glycerol concentrations (5-20%) do not affect protein compaction (Yoshizawa et al., 2018). All solutions were filtered through 0.1 mm membranes (Millipore) to remove any aggregates. The SAXS profiles were collected at protein concentrations ranging from 0.5 to 5.0 mg/ml. 20 one-second exposures were used for each sample and buffer maintained at 15˚C. Each of the resulting diffraction images was scaled using the transmitted beam intensity, azimuthally integrated by SASTool (SasTool, 2013) and averaged to obtain fully processed data in the form of intensity versus q [q = 4psin(q)/l, q = one half of the scattering angle; l = X ray wavelength]. The buffer SAXS profile was subtracted from a protein SAXS profile. Subsequently, the mean of the lower concentration (0.5-1.5 mg/ml) profiles in the smaller scattering angle region (q < 0.15 A˚À 1 ) and the mean of the higher concentration (2.0-5.0 mg/ml) profiles in the wider scattering angle region (q > 0.12 A˚À 1 ) were merged to obtain the final experimental SAXS profiles that are free of the concentration-dependent aggregation or polydispersity effect (Kikhney and Svergun, 2015).
The merged SAXS profiles were initially analyzed using the ATSAS package (Petoukhov et al., 2012) to calculate radius of gyration (R g ), maximum particle size (D max ), and pair distribution function (P(r)) (

Co-immunoprecipitation and immunoblotting
HeLa cells expressing H2B mCherry  (gift from Prof. Hongtao Yu, UT Southwestern). The HeLa Tet-ON cells (Cellosaurus Accession: HeLa Tet-On (CVCL_IY74)) stably expressing H2B-mCherry were originally created (with cell identity confirmation carried out by STR profiling) in Dr. Hongtao Yu's lab at University of Texas Southwestern Medical Center, Dallas, Texas USA. Mycoplasma negative status of the cell line was confirmed using the LookOut Mycoplasma PCR Detection kit, Sigma MP0035-1KT. The cells were grown to 80% confluency, and total-cell lysate was prepared by suspending the cells in TB buffer containing 20 mM HEPES-KOH pH 7.3, 110 mM potassium acetate, 2 mM magnesium acetate, 5 mM sodium acetate, 0.1 mM EGTA, 1 mM DTT and protease inhibitor cocktail (Kimura et al., 2017) on ice for 15 min, sonicating three times (5 s pulse, 10 s rest), then centrifuging the lysed cells at 15,000 g for 20 min at 4˚C. Nuclear and cytoplasmic fractions were prepared using the NE-PER Nuclear and Cytoplasmic Extraction reagents (Thermo Scientific) as per manufacturer's instruction. Protein concentration was quantitated using the Bradford protein assay kit (BioRad). The RFP-Trap (high quality Red Fluorescent Protein (RFP) binding protein coupled to a monovalent magnetic matrix, ChromoTek GmbH) was incubated with the cell lysates for 2 hr at 4˚C. The matrix was first washed with TB buffer supplemented with 200 mM NaCl, and then once with TB buffer supplemented with 150 mM NaCl. The proteins bound to the beads were dissolved in SDS sample buffer for immunoblot analysis.

Confocal microscopy imaging
Cells (5 Â 10 4 cells per chamber) were seeded into collagen coated culture coverslip (BD Falcon) The next day, cells were rinsed with ice-cold PBS and fixed with 4% paraformaldehyde for 10 min at room temperature followed by permeabilization with 0.1% sodium citrate plus 0.1% Triton X-100. The cells were subjected to immunofluorescence staining using rabbit polyclonal antibody against Imp9 (1:250 dilution, Cat no. A305-474A-T, Bethyl Laboratories, Inc) and mouse monoclonal antibody against Ran (1:250 dilution, Cat no. 610340, BD Biosciences), for 2 hr at room temperature. The cells were then washed with cold PBS three times for 1 min each and incubated with Alexa 480labeled anti-rabbit secondary antibody (1:800) (Invitrogen) and Alexa 405-labeled anti-mouse secondary antibody (1:800) (Invitrogen) at room temperature for 1 hr. Subsequently cells were washed with cold PBS three times for 1 min each and mounted with ProLong Gold Antifade Mountant (Invitrogen).
Image acquisition was performed with a spinning disk confocal microscope system (Nikon-Andor) with a 100 Â oil lens and the MetaMorph softwar. Images were acquired from randomly selected fields as a z-stack with step size of 0.1 mm to give a total of 196 slices. For each selected field of view, three images were taken, an Alexa488 (Imp9) image, and Alexa405 (Ran The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Data availability Diffraction data have been deposited in PDB under the accession code 6N1Z.

Author contributions
The following dataset was generated: