Crystal Structure of the Nuclear Export Receptor CRM1 in Complex with Snurportin1 and RanGTP

See allHide authors and affiliations

Science  22 May 2009:
Vol. 324, Issue 5930, pp. 1087-1091
DOI: 10.1126/science.1173388


CRM1 mediates nuclear export of numerous unrelated cargoes, which may carry a short leucine-rich nuclear export signal or export signatures that include folded domains. How CRM1 recognizes such a variety of cargoes has been unknown up to this point. Here we present the crystal structure of the SPN1⋅CRM1⋅RanGTP export complex at 2.5 angstrom resolution (where SPN1 is snurportin1 and RanGTP is guanosine 5′ triphosphate–bound Ran). SPN1 is a nuclear import adapter for cytoplasmically assembled, m3G-capped spliceosomal U snRNPs (small nuclear ribonucleoproteins). The structure shows how CRM1 can specifically return the cargo-free form of SPN1 to the cytoplasm. The extensive contact area includes five hydrophobic residues at the SPN1 amino terminus that dock into a hydrophobic cleft of CRM1, as well as numerous hydrophilic contacts of CRM1 to m3G cap-binding domain and carboxyl-terminal residues of SPN1. The structure suggests that RanGTP promotes cargo-binding to CRM1 solely through long-range conformational changes in the exportin.

Nuclear transport proceeds through nuclear pore complexes (NPCs) and supplies cell nuclei with proteins and the cytoplasm with nuclear products such as ribosomes and tRNAs. Most nuclear transport pathways are mediated by importin β–type nuclear transport receptors, which include nuclear export receptors (exportins), as well as importins (1, 2). These receptors bind cargoes directly or through adapter molecules, shuttle constantly between the nucleus and cytoplasm, and use the chemical potential of the nucleocytoplasmic RanGTP-gradient to act as unidirectional cargo pumps (where GTP is guanosine 5′ triphosphate and RanGTP is GTP-bound Ran) (3).

Exportins recruit cargo at high RanGTP levels in the nucleus, traverse NPCs as ternary cargo⋅exportin⋅RanGTP complexes, and release their cargo upon GTP hydrolysis into the cytoplasm. CRM1 (exportin1/Xpo1p) (4, 5) and CAS (Cse1p/exportin2) (6) are the prototypical exportins. Whereas CAS is specialized to retrieve the nuclear import adapter importin α back to the cytoplasm (6), CRM1 exports a very broad range of substrates from nuclei (4, 5, 711), including ribosomes and many regulatory proteins. It also depletes translation factors from nuclei and is essential for the replication of viruses such as HIV.

CRM1 has a dual function during biogenesis of spliceosomal U small nuclear ribonucleoproteins (snRNPs). It exports m7G-capped U small nuclear RNAs to the cytoplasm (4, 12), where they recruit Sm-core proteins and receive a 2,2,7-trimethyl (m3G) cap structure. The import adapter snurportin 1 (SPN1) and importin β then transport the mature m3G-capped U snRNPs into nuclei (13). To mediate another import cycle, SPN1 is returned to the cytoplasm by CRM1 (14).

Many CRM1 cargoes harbor a leucine-rich nuclear export signal (NES) that typically includes four characteristically spaced hydrophobic residues (7). Examples are the HIV-Rev protein (15) or the protein kinase A inhibitor (PKI) (16). In other cases, however, CRM1 recognizes not just a short peptide, but instead a large portion of the export cargo; here, SPN1 is the prototypical example (14). CRM1 binds SPN1 tighter than other export substrates, apparently because CRM1 must displace the imported U snRNP from SPN1 before export may occur.

The cytoplasmic dissociation of CRM1 from SPN1 is essential for multi-round import of U snRNPs. Hydrolysis of the Ran-bound GTP alone is insufficient to fully disrupt the interaction (Fig. 1, A to C) (14), but importin β can displace CRM1 from SPN1 (Fig. 1A). Thus, either the binding sites of SPN1 for CRM1 and importin β overlap, or importin β forces SPN1 into a conformation that is incompatible with CRM1 binding.

Fig. 1

(A) Effects of RanGTP and importin β on the SPN1⋅CRM1 interaction. 1 μM SPN1 was mixed with an Escherichia coli lysate containing 200 mM NaCl, 1 μM CRM1, and the indicated combinations of 3 μM RanGTP and 1, 2, or 3 μM importin β (Impβ) (21). Complexes were retrieved by immunoglobulin G (IgG)–Sepharose via the zz-tag of SPN1. SPN1-ligands were eluted with 1.5 M MgCl2 (upper panel); the remaining baits (zz-SPN1; lower panel) were eluted with SDS. Analysis was by SDS–polyacrylamide gel electrophoresis/Coomassie staining. Note that RanGTP enhanced CRM1-binding to SPN1; however, this interaction was also detectable in the absence of Ran. This residual CRM1⋅SPN1 interaction could be suppressed by importin β that binds the IBB domain of SPN1. MW, molecular weight standard. (B) Met1, Leu4, Leu8, Phe12, and Val14 of SPN1 are all required for high-affinity binding to the CRM1⋅RanGTP complex. zz-tagged CRM1 immobilized on IgG-Sepharose was incubated with an E. coli lysate containing 200 mM NaCl and indicated combinations of 3 μM RanGTP and 1 μM untagged wild-type SPN1 or the specified mutants. CRM1-ligands were eluted with MgCl2 and analyzed as described in (A). Note that mutating Leu4, Leu8, Phe12, or Val14 to Ser (left) or deleting Met1 (right) abolished or substantially impaired SPN1 binding to CRM1⋅RanGTP, whereas mutating Leu28 did not (left). (C) The N terminus of SPN1 contains export determinants that allow autonomous, RanGTP-stimulated binding to CRM1. Indicated zz-tagged SPN1 derivatives or the PKI-NES immobilized on IgG-Sepharose were incubated with an E. coli lysate containing 1 μM CRM1 and 3 μM RanGTP as specified. Bound ligands were eluted with MgCl2 and analyzed as described in (A). At low NaCl concentration (50 mM, upper panel), SPN12–360 bound CRM1⋅RanGTP nearly as efficiently as full-length SPN11-360; however, a clear decrease in binding was observed without Ran. SPN11–21 recruited CRM1 in a strictly RanGTP-dependent manner. This CRM1 binding was lost when SPN1Met1 was deleted. Even though SPN11–21 contains five hydrophobic residues, it bound CRM1 considerably weaker than the classical PKI-NES with only four hydrophobic residues. This difference was particularly apparent at 200 mM NaCl (lower panel).

Two functional domains in SPN1 have been described: (i) the m3G cap-binding domain (SPN97–300) (17) and (ii) the N-terminal importin β–binding (IBB) domain (SPN140–65) (14, 18, 19), which confers binding to and import by importin β (20). A multiple alignment of SPN1 from various species revealed another conserved region that precedes the IBB domain and includes the hydrophobic residues Leu4, Leu8, Phe12, and Val14. Mutating any of those residues to serine or deleting Met1 strongly impaired the interaction with CRM1, in particular at higher salt concentrations (Fig. 1B and fig. S1). Even though the SPN1 N terminus (with its conserved hydrophobic residues) resembles a classical NES, there are clear differences: foremost that CRM1 binds the isolated SPN1 N terminus (SPN11–21) considerably weaker than, for instance, the PKI-NES (Fig. 1C). In the context of full-length SPN1, however, this difference is more than compensated by the contribution of the m3G cap-binding domain to the CRM1 interaction.

We then assembled, purified, and crystallized an export complex containing full-length human SPN11–360, full-length mouse CRM11–1071, and GTP-RanQ69L1–180, a C-terminally truncated and GTPase-deficient form of human Ran (21). The resulting crystals contained two complexes per asymmetric unit. The structure was solved by molecular replacement with the use of known structures of GTP-bound Ran7–176 (22), SPN197–300 (17), and a short human CRM1707–1027 fragment (23). The final model, refined at a resolution of 2.5 Å, includes residues 12 to 1055 of CRM1 and Ran9–179, as well as SPN11–360. CRM167–69 and four regions of SPN1 appear to be disordered (table S1) (21).

As expected from previous sequence analysis (23, 24), CRM1 is built from so-called HEAT repeats (Fig. 2, fig. S2, and table S2), which include two consecutive helices (A and B) that pack in antiparallel orientation against each other and against the adjacent repeat (25). However, previous structure prediction (23) failed to predict the correct number and exact positions of the 21 repeats. This reflects the highly degenerate nature of some of the repeats, which even leads to an inverted topology of helices at the C terminus of CRM1 (fig. S2).

Fig. 2

Structure of the SPN1⋅CRM1⋅RanGTP nuclear export complex. Two views of the complex are depicted. Color-codes for Ran, SPN1, and the 21 consecutive HEAT repeats of CRM1 are indicated. Except for HEAT 21, A helices of the HEAT repeats are located at the outer surface of the CRM1 toroid and B helices at the inner surface (see also fig. S2). RanGTP is engulfed by the toroid-shaped structure of CRM1 and fixed by the so-called acidic loop (shown in the lower panel in gray), which is part of HEAT repeat 9. SPN1 is bound on the outer surface of CRM1, far away from the Ran molecule.

In contrast to importin β (26), transportin (27), and CAS/Cse1p (22, 28), the overall CRM1 structure shows remarkably little superhelical twist (Fig. 2). However, it is bent to a distorted toroid structure, with HEAT 21 touching helices 2B and 5A, as well as the loop between HEATs 4 and 5 (Fig. 2 and fig. S2). Ran is enclosed into this toroid and stabilizes the ring closure by extensive contacts. In contrast to the IBB–importin β interaction (18, 19, 26), the cargo SPN1 is not enveloped by CRM1 but instead rests on the outside of the CRM1 toroid (Fig. 2). This different binding topology might reflect the fact that CRM1 carries cargoes, such as ribosomal subunits, that are too large to be engulfed by an exportin. In addition, the outside of the torus provides a larger surface area and possibly also a greater variety of binding sites for cargo recognition than does the inner face that already accommodates the Ran molecule.

The structure of m3G cap-bound SPN197–300 was previously solved (17) and remained essentially unaltered in the SPN1⋅CRM1⋅RanGTP complex (root mean square deviation = 0.67 Å). However, several residues of SPN1 as well as of CRM1 HEATs 12 and 13 protrude into the m3G cap-binding pocket (fig. S3). With the physiological import cargo of SPN1 (fully assembled U snRNPs), the clashes would be even more severe, because the RNA would run into the CRM1 molecule. Thus, SPN1 cannot simultaneously bind its import cargo and its export receptor, which agrees with previous data (14). This ensures that only cargo-free SPN1 is returned to the cytoplasm and allows SPN1 to mediate unidirectional transport of m3G-capped U snRNPs into nuclei.

SPN1 binds CRM1 through an elaborate contact area (2330 Å2), which includes three parts: (i) the N terminus (SPN11–35), (ii) the m3G cap-binding domain (SPN197–300), and (iii) a C-terminal region, SPN1349–360 (Fig. 3A). This is consistent with biochemical data that revealed strong contributions of SPN11–21 and the cap-binding domain to CRM1 binding (Fig. 1, B and C, and fig. S1) (14) and a weaker contribution of SPN1286–360 (14).

Fig. 3

The nuclear export signature of SPN1 involves a large interface formed by residues from all three domains of SPN1. (A) Three domains of SPN1 contact CRM1. These include N-terminal residues of SPN1 (orange), the cap-binding domain (gray), and the C-terminal residues (SPN1349–360, yellow). The IBB domain of SPN1, which forms a straight helix within the importin β complex (18, 19), is here partially unwound and bent (green). White dashed lines mark unresolved stretches. (B) The N-terminal hydrophobic residues of SPN1 (Met1, Leu4, Leu8, Phe12, and Val14) dock into a hydrophobic cleft of CRM1. Carbons of SPN1 are shown as orange, oxygens as red, and nitrogens as blue sticks. The side chains of the hydrophobic residues are depicted as spheres. CRM1 is shown as a surface representation (blue indicates hydrophilic areas, white denotes hydrophobic areas). The yellow patch marks the sulfur of Cys528, which is covalently modified by the CRM1-specific inhibitor LMB (29).

All N-terminal residues that were found to be critical for CRM1-binding (SPN1Met1, Leu4, Leu8, Phe12, Val14) (Fig. 1B and fig. S1) dock into a hydrophobic cleft that is formed by helices 11A and 12A and the intervening helical linker between 11B and 12A of CRM1 (Fig. 3B and fig. S4). The side chain of CRM1Lys534, which is positioned by a salt bridge to CRM1Glu575, closes the cleft and introduces a sharp kink into the SPN1 chain between SPN1Val14 and SPN1Ser15. There are several additional contacts in this area, such as electrostatic attraction between the negatively charged N-terminal helix of SPN1 and basic regions on the CRM1 surface, as well as hydrogen bonds between SPN1Ser15 and CRM1Glu575 and between SPN1Tyr35 and CRM1Glu529 (fig. S4). The CRM1 inhibitor leptomycin B (LMB) covalently modifies CRM1Cys528 (29). Cys528 is located within the hydrophobic cleft (Fig. 3B), which explains plausibly why LMB-modified CRM1 cannot bind export cargoes that rely on this cleft.

The N-terminal part of snurportin’s export signature (with its five critical hydrophobic residues) resembles a classical NES and binds CRM1 in a conformation where residues Met1 to Ser11 form an α helix (Fig. 3A and fig. S4). The classical NES from the HIV Rev protein (15) must be recognized differently for three reasons: (i) The spacing of the hydrophobic residues is different, (ii) the intervening prolines would not allow such a helix to form, and (iii) this classical NES contains only four critical hydrophobic residues (15). Nevertheless, we cannot exclude the possibility that the same hydrophobic cleft also accommodates some or all of the key hydrophobic residues from classic NESs.

The interaction between the SPN1 m3G cap-binding domain and CRM1 is dominated by polar contacts. SPN1349–360, the third part of the export signature, binds to helices 14A, 15A, and 16A of CRM1 (Figs. 2 and 3A).

Importin β can displace CRM1 from the rather stable Ran-free SPN1⋅CRM1 complex (Fig. 1A) and therefore restore m3G cap-binding of SPN1 in the cytoplasm (fig. S3) (14). This antagonism between CRM1 and importin β is not caused by an overlap of the respective binding sites; rather it appears to be caused by a combination of conformational changes in SPN1 and volume extrusion. The IBB domain binds importin β as a straight helix (18, 19). However, within the CRM1 complex, the central part of this IBB helix is unwound, and the remaining helix-fragments are kinked by ≈80° (shown in green in Fig. 3A). This distortion of the IBB helix appears to be enforced by contacts of the 35 N-terminal residues of SPN1 with CRM1. Thus, straightening of the IBB helix by importin β is likely to break crucial contacts between CRM1 and SPN1.

The structure of Ran in the SPN1⋅CRM1⋅RanGTP complex is virtually identical to that in other transport receptor⋅RanGTP complexes (22, 30, 31). Ran is almost completely engulfed by the CRM1 toroid and contacts four distinct areas of CRM1 (Figs. 2 and 4, A to C, and movie S1). The first area is located within the region that is most conserved between nuclear transport receptors (24, 32). HEATs 1 to 3 bind switch II of Ran, whereas HEATs 4 and 5 pack against Ran helix 3 and the so-called “basic patch” (30), respectively (Fig. 4, B and C). The second Ran-binding region (HEATs 7 to 9) also contacts the basic patch and extends to β strand 6 of Ran. Analogous interactions occur in RanGTP complexes with CAS, transportin, and importin β (22, 30, 31). In contrast, the long “acidic loop” within HEAT 9 (region 3) binds Ran in an unprecedented manner. It forms a β hairpin, touches HEAT helices B12 to B15, and reaches through the entire central “hole” of the CRM1 toroid (Figs. 2 and 4, A and B, and figs. S2 and S5). The acidic loop locks RanGTP closely to the N- and C-terminal HEAT repeats and binds Ran37 from switch I, as well as Ran127,129,155 from the loops involved in guanine recognition. The fourth C-terminal Ran-binding region (HEATs 17 and 19) was not anticipated by sequence similarity or previous structures. It contacts both switch regions of Ran.

Fig. 4

Molecular details of the CRM1⋅RanGTP interaction (see also table S3). (A) CRM1 is shown as a gray backbone tube, and Ran is depicted as a surface representation. HEAT helices 11A and 12A, forming the cargo-binding hydrophobic cleft, are shown in green, and the acidic loop is shown in red. (B) RanGTP contact areas on CRM1. Orientation of CRM1 is as in (A), but Ran has been removed, and Ran-binding residues of CRM1 (distance < 3.6 Å) are shown as orange sticks. Note that the Ran-binding site includes four distinct areas (labeled 1 to 4). See also movie S1. (C) Contacts of RanGTP to CRM1. Ran is depicted as a ribbon diagram. Orientation is as in (A). CRM1-binding residues are shown in orange, switch I (Ran30–47) is shown in red, and switch II (Ran65–80) is depicted in blue. CRM1 contacts both switches. The basic patch (Ran139–142, dark blue) shows extensive contacts to CRM1 regions 1 and 2 [see panel (B)]. Secondary structure elements are numbered as in (33). GTP is depicted as gray sticks; the Mg2+ ion is shown as a green sphere. (D) Model for conformational states of CRM1. CRM1 switches between a relaxed cytoplasmic (top) and a strained nuclear conformation (bottom). In the hypothetical cytoplasmic conformation, the contact sites for RanGTP inside the CRM1 toroid are too far apart to bind Ran with high affinity. Also, the hydrophobic cleft on the outer side of the toroid is closed. Rigid body movements allow transition to the nuclear conformation. Here, the Ran-binding sites are close enough to bind Ran simultaneously and thus with high affinity. The conformational change also alters the curvature of the toroid near the cargo-binding site, opens the hydrophobic cleft, and allows the export cargo to dock. For details, see main text.

To function as an effective, unidirectional cargo-pump, CRM1 must strongly discriminate between GTP- and guanosine diphosphate (GDP)–bound Ran. CRM1 can sense the nucleotide state of Ran because it contacts switches I and II; i.e., the regions that differ most between GDP- and GTP-Ran (Fig. 4, B and C, and fig. S5). The structure of RanGDP (33, 34) is incompatible with CRM1 binding, because switch I of RanGDP would clash with CRM1 HEAT 1 and switch II would collide with HEAT 19.

Ran switches the affinity of importin β–type transport receptors for their cargoes and thereby provides energy for the transport cycles. In the case of Cse1p, RanGTP increases the affinity of the exportin for its cargo importin α by directly interacting with both Cse1p and importin α (22). There are, however, no direct contacts between Ran and cargo in the SPN1⋅CRM1⋅RanGTP complex (Fig. 2). The ≈1000-fold increase in the affinity of CRM1 for RanGTP by SPN1 and the equally large strengthening of the CRM1⋅SPN1 interaction by RanGTP (14) must therefore be caused solely by conformational changes in the CRM1 molecule.

Both RanGTP and SPN1 obviously select the same conformation of CRM1 for high-affinity binding (here referred to as the nuclear conformation), whereas the free cytoplasmic conformation of CRM1 has only a low affinity for the two ligands (Fig. 1 and fig. S1). To explain this cooperativity, one must assume that the nuclear conformation is under considerable tension and that this tension is compensated for by the released binding energies of the Ran⋅CRM1 and SPN1⋅CRM1 interactions. In other words, Ran apparently promotes SPN1-binding by stabilizing the strained nuclear conformation of CRM1 and vice versa.

Ran and SPN1 are ≈55 Å apart in the export complex (Fig. 2). The conformational changes in CRM1 that coordinate their binding must therefore be transmitted over a considerable distance, probably through rigid body movements along the HEAT repeats. The splitting of the Ran-binding site on CRM1 into distinct regions is probably crucial for driving these movements, because even small changes in their distances will greatly affect the binding of Ran. Ran-binding regions 2 and 4, for example, are located on opposite sides of the CRM1 toroid (Fig. 4B), and it is quite possible that they are too far apart in the relaxed CRM1 conformation to contact Ran simultaneously (Fig. 4D). Hence, a low affinity for Ran would result. The transition to the nuclear conformation would bring these interfaces closer together and allow high-affinity binding of Ran.

Conformational changes in CRM1, which favor the CRM1⋅Ran interaction, must also activate the cargo binding site(s). Thus, we suggest that the hydrophobic cleft is also controlled by these transitions. The cleft might be closed in the cytoplasmic, relaxed conformation of CRM1 (Fig. 4D). Putting the CRM1 toroid under tension to bind Ran with all interfaces should also change the curvature of the CRM1 molecule around the cargo-binding site. This might stretch the contacts between the outer A helices of HEATs 11 and 12 and consequently open the hydrophobic cleft. The observation that CRM1 binding of the isolated SPN1 N terminus and, thus, docking into the hydrophobic cleft is efficient only in the presence of RanGTP (Fig. 1C) strongly supports this model.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Tables S1 to S3


Movie S1

  • * These authors contributed equally to this work.

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank the staff of synchrotron beamlines at BESSY (Berlin), the European Molecular Biology Laboratory/Deutsches Elektronen-Synchrotron (Hamburg), and the European Synchotron Radiation Facility (Grenoble) for assistance during data collection; D. Deichsel for excellent technical help; and S. Frey and S. Güttler for comments on the manuscript. We received funding from the Deutsche Forschungsgemeinschaft (grant SFB523 to R.F. and T.M.), the Max-Planck-Gesellschaft (to D.G. and T.G.), the Alfried Krupp-Foundation (to D.G. and T.G.), and the Boehringer Ingelheim Fonds (to T.G.). Coordinates and structure factors have been deposited in the Protein Data Bank with accession code 3GJX.
View Abstract

Stay Connected to Science

Navigate This Article