CRM1-mediated Recycling of Snurportin 1 to the Cytoplasm

The following proteins were expressed in Escherichia coli BLR/Rep4 and purified as previously described: Xenopus importin α (Görlich et al., 1994), human importin α (Görlich et al., 1996a), NTF2 (Ribbeck et al., 1998), His-tagged RanQ69L and Ran wild type (Görlich et al., 1996c), Rna1p (Bischoff et al., 1995a), RanBP1 (Bischoff et al., 1995b), and CAS (Kutay et al., 1997). Snurportin 1 was cloned into the BamHI-XmaI sites of pQE30 (Qiagen), expressed with an NH₂-terminal His tag and purified on nickel-NTA agarose followed by dialysis in 20 mM Hepes-KOH, pH 7.5, 200 mM NaCl, 2 mM magnesium acetate, 250 mM sucrose. GST-snurportin 1 was expressed from pGEX T4 and purified on glutathione Sepharose 4B, followed by chromatography on MonoQ. zz-snurportin 1 (full length and fragments) were cloned into the NcoI-BamHI sites of pQE60 and expressed with a COOH-terminal His tag and with NH₂-terminal IgG binding domains from Staphylococcus aureus protein A. Human CRM1 was cloned into the NdeI-BamHI sites of pET3a and was expressed in BL21 DE3 without addition of IPTG at 30°C. This recombinant CRM1 is untagged and was purified on nickel-NTA agarose followed by chromatography on MonoQ. The cDNA coding for the HIV Rev protein was cloned into the NcoI-BamHI site of the 6z60 vector (Jäkel and Görlich, 1998). Expression was with an NH₂-terminal 6z tag and a COOH-terminal his tag, purification was on nickel agarose.

Labeling of snurportin 1 with fluorescein 5′ maleimide (Calbiochem) was performed at a 1:1 molar ratio in 50 mM Hepes-KOH, pH 7.5, 200 mM NaCl, 2 mM magnesium acetate, 250 mM sucrose for 2 h on ice. Free label was removed on a NAP5 column equilibrated with 50 mM Hepes-KOH, 150 mM NaCl, 2 mM magnesium acetate. Labeling of Xenopus importin α has been described before (Görlich et al., 1996a).

Antibodies against the following antigens have been described before: human importin β (Görlich et al., 1995b), Xenopus RanBP7 (Görlich et al., 1997), and human CAS (Jäkel and Görlich, 1998). Anti-CRM1 antibodies were raised in rabbits against a CRM1 peptide (cys-EKHKRQMSV). Antibodies were affinity-purified on sulfoLink (Pierce) to which the antigens had been coupled.

Permeabilization of HeLa cells and nuclear import reactions in suspension was performed essentially as described before (Adams et al., 1990; Görlich et al., 1996c; Kutay et al., 1997). The import buffer contained: 2 mg/ml nucleoplasmin core (to block nonspecific binding), 20 mM Hepes-KOH, pH 7.5, 140 mM potassium acetate, 5 mM magnesium acetate, 250 mM sucrose, 0.5 mM EGTA. Reactions were supplemented with an energy-regenerating system (0.5 mM ATP, 0.5 mM GTP, 10 mM creatine phosphate, 50 μg/ml creatine kinase), 5 μM RanGDP, and 1 μM NTF2.

zz-tagged RanQ69L and snurportin 1 prebound to IgG-Sepharose were used as affinity matrices. The z domain is the IgG binding domain from S. aureus protein A. 500 μl of cytoplasmic HeLa extract or 200 μl of E. coli lysate expressing the recombinant proteins was incubated with 20 μl of affinity matrix for 4 h at 4°C in binding buffer (50 mM Hepes-KOH, pH 7.5, 200 mM NaCl, 5 mM magnesium acetate, 0.005% digitonin). The beads were recovered by mild centrifugation and washed three times with 1 ml binding buffer. Elution was with 100 μl of 50 mM Tris-HCl, pH 7.5, 1 M magnesium chloride for 10 min at room temperature. Proteins were precipitated with 90% isopropanol (final concentration), dissolved in SDS sample buffer, and analyzed by SDS-PAGE.

Kinetic measurement of the RanGTPase was as described before (Bischoff et al., 1995a, 1995b; Kutay et al., 1997) with modifications described in the figure legends. The m₃G-cap (m3GpppAmpUmpA-oligonucleotide) was described previously (Huber et al., 1998) and the m⁷GpppG-cap dinucleotide was purchased from Pharmacia.

Oocyte injection and analysis of microinjected RNA by denaturing gel electrophoresis and autoradiography were performed as described (Jarmolowski et al., 1994). The mutant RNAs used (U1ΔSm, U5ΔSm) lack protein-binding sites required for the nuclear import of these RNAs (Hamm and Mattaj, 1990; Jarmolowski et al., 1994) and thus remain in the cytoplasm after export from the nucleus.

Snurportin 1 mediates the m₃G-cap–dependent import of U snRNAs (Huber et al., 1998), most likely in conjunction with importin β (Palacios et al., 1997; J. Huber and R. Lührmann, unpublished observations). To achieve multiple rounds of import, snurportin 1 has to return to the cytoplasm after delivering its import substrate to the nucleus. Since importin α and snurportin 1 both contain an IBB domain for binding to and import by importin β, it was of interest to know if snurportin 1 would also exit the nucleus in the same way as importin α, by using CAS for reexport (Kutay et al., 1997). Surprisingly however, we could not detect any interaction between CAS and snurportin 1 (see below), suggesting that snurportin 1 export is mediated by a distinct factor. To identify this factor, we immobilized snurportin 1 and tested which proteins from a HeLa extract it would bind (Fig. 1). Without further addition, three major bands were recovered in the bound fraction and identified by Western blotting as importin β, CRM1, and importin 7 (formerly called RanBP7). The binding of importin β to snurportin 1 was expected from previous studies. Importin 7 was apparently recovered only because it binds importin β (not shown, but see Görlich et al., 1997). Binding of the export receptor CRM1 to snurportin 1 might suggest that CRM1 mediates export of snurportin 1. Indeed, the pattern of snurportin 1–bound proteins changed when RanQ69L GTP was included in the binding reaction. This Ran mutant is GTPase-deficient, remains GTP-bound even in the presence of RanGAP1, and can therefore be used to mimic a nuclear environment in a cytoplasmic extract. Under these conditions, the binding of the two importins was greatly reduced. In contrast, binding of CRM1 was enhanced by the presence of RanGTP. Thus, CRM1 behaved as expected for a snurportin-specific export receptor.

The interaction between snurportin 1 and CRM1 could be either direct or mediated by an unidentified factor in the HeLa extract. To exclude the second possibility we repeated the binding experiments using E. coli lysates containing recombinant transport receptors. CRM1 bound to immobilized snurportin 1 (Fig. 2 A, lane 3), and this binding was enhanced by the addition of RanQ69L. Ran was also recovered in the bound fraction, indicating the formation of a trimeric snurportin/CRM1/RanGTP complex (Fig. 2 A, lane 2). Conversely, the trimeric complex could also be assembled using immobilized RanQ69L, free CRM1, and GST-snurportin 1 (Fig. 2 B). The high cooperativity of complex formation was evident from the observation that CRM1 binding to RanGTP was not detectable in the absence of snurportin 1 but was strong in its presence.

To obtain more quantitative data on the formation of the trimeric complex, we used a kinetic assay (Fig. 3, and see for example Kutay et al., 1997). Binding of an exportin to RanGTP prevents GTPase activation by RanGAP1. GTP hydrolysis can easily be quantified and used to calculate the proportion of Ran that is exportin-bound. From the dose dependence of the effects one can estimate apparent dissociation constants (K_d) for the complex formation. Without further addition, CRM1 bound RanGTP only very weakly (apparent K_d ≫ 1 μM, Fig. 3 A); In the presence of a saturating concentration of snurportin 1 (2 μM), however, the K_d shifted to 4–5 nM, emphasizing the highly cooperative nature of the snurportin/CRM1/RanGTP complex formation. The specificity of the effect was verified by two controls. First, the presence of importin α had no effect on the RanGTP/CRM1 interaction. Second, CAS responded only to its specific export substrate importin α, but not to snurportin 1 (Fig. 3 B).

Export complexes form in the nucleus and need to be disassembled after export in the cytoplasm. In the case of CAS and exportin-t, this disassembly is brought about by the concerted action of RanBP1 and RanGAP1. However, there is a complication in respect to CRM1, namely that RanBP1 and RanGAP1 apparently also contain leucine-rich NESs (Richards et al., 1996; Matunis et al., 1998) and are thus also potential export substrates for CRM1 which could stabilize a CRM1/RanGTP complex. Therefore, it was quite surprising that RanBP1 rendered also the snurportin/CRM1/RanGTP complex sensitive to RanGAP and thereby caused disassembly (see corresponding curves in Fig. 3 A).

The observation that CRM1 binds snurportin 1 with high affinity and specificity strongly suggests that CRM1 mediates nuclear export of snurportin. To test this directly, we performed the pulse-chase export experiment shown in Fig. 4. Nuclei of permeabilized HeLa cells were first loaded for 10 min with fluorescein-labeled snurportin 1 and Texas red–labeled importin α in the presence of importin β, Ran, NTF2, and an energy-regenerating system. The sample was then split into four. The first aliquot was fixed immediately and confocal fluorescence microscopy confirmed that snurportin 1 and importin α had both efficiently accumulated in the nuclei. The three remaining aliquots were supplemented with either CRM1, CAS, or buffer and incubated for another 10 min before these samples were fixed also. The addition of CRM1 caused a significant depletion of the nuclear pool of snurportin and an increased NPC signal that probably represented import and export intermediates. However, CRM1 did not change the localization of importin α. Addition of CAS had exactly the opposite effect. It led to nuclear export of importin α but had no effect on snurportin. Therefore, we can conclude that CRM1 specifically mediates export of snurportin. The experiment also confirms our previous data that CAS is the importin α–specific exportin (Kutay et al., 1997) and contradicts the suggestion by Boche and Fanning (1997) that importin α would exit the nucleus by virtue of a Rev-like NES, i.e., in a CRM1-dependent manner.

Productive import cycles require snurportin 1 to enter the nucleus with cargo and to exit without cargo. Although nuclear retention of the import substrate might contribute to this directionality, transport would be most efficient if the affinity of snurportin 1 for m₃G-capped U snRNPs was high during nuclear entry and low during reexport. β-Family transport receptors use RanGTP to regulate interactions with their transport substrates. This mechanism is not available to snurportin 1 because it does not bind Ran (Fig. 2 B, lane 5). However, it uses importin β for nuclear entry and CRM1 for exit from the nucleus and the binding of the two transport receptors might regulate snurportin's affinity for the m₃G-cap. Therefore, we tested for a cross-talk between the binding sites for the m₃G-cap and CRM1. In the absence of RanGTP, CRM1 bound weakly to snurportin 1 (Fig. 5 A). A 10-fold excess of a m₃G-cap oligo prevented CRM1 binding completely, suggesting that m₃G-cap and CRM1 binding to snurportin 1 are mutually exclusive. However, the m₃G-cap oligo could only reduce, but not prevent, CRM1 binding in the presence of Ran-GTP, i.e., when the exportin was in its high affinity form.

The antagonism of CRM1 and m₃G-cap binding was also clearly evident in the kinetic assay that measures snurportin/CRM1/RanGTP complex formation (Fig. 5 B). The presence of 5 μM m₃G-cap shifted the equilibrium and reduced the apparent affinity of snurportin 1 for the CRM1/RanGTP complex ∼40-fold. The effect was highly specific as verified by two controls. First, a m⁷G-cap analogue, which does not bind snurportin 1, had no effect on the snurportin/CRM1/RanGTP interaction. Second, the m₃G-cap did not interfere with the interaction between CRM1 and BSA conjugated to HIV Rev NES peptides and is thus snurportin 1–specific (Fig. 5 C).

A mutual displacement of CRM1 and m₃G-cap from snurportin 1 should occur twice per import cycle. When the trimeric export complex has been exported and RanGTP removed, a still quite stable CRM1/snurportin 1 dimer ends up in the cytoplasm. Binding of the import substrate to snurportin 1 displaces the low affinity form of the exportin (Fig. 5 A). Conversely, following nuclear entry, the U snRNPs need to be displaced from snurportin 1 by the high affinity form of CRM1 in the presence of RanGTP. Because the m₃G-cap binds snurportin 1 very tightly, displacement would only be efficient if the CRM1/RanGTP complex would bind with at least comparable affinity. In fact, Fig. 5 D shows that CRM1 binds snurportin 1 50 times more tightly (K_d ≈ 10 nM) than the HIV Rev protein (K_d ≈ 0.5 μM). The difference is even more dramatic when compared with the isolated NES from HIV Rev. A peptide comprising the Rev activation domain plus a few flanking residues bound CRM1 700 times weaker (K_d ≈ 7 μM), and the minimum Rev activation domain itself even 5,000-fold weaker (K_d ≈ 50 μM) than snurportin 1.

The 5,000-fold higher affinity of CRM1 for snurportin 1 compared with the affinity for the minimum Rev activation domain (Rev-NES) raised the question as to the nature of snurportin's export signal. Snurportin 1 contains an IBB domain for importin β binding (residues 26–65), while m₃G-cap binding has been attributed to the part of the protein that is COOH-terminal to the IBB (Huber et al., 1998). Analysis of the amino acid sequence of snurportin 1 did not reveal the existence of a single amino acid sequence matching the consensus leucine-rich NES; rather, several regions were found which distantly resemble it. To delineate the binding site for CRM1 on snurportin 1 we expressed NH₂- and COOH-terminal snurportin 1 deletion mutants in E. coli, immobilized them, and tested their capacity to bind CRM1 in the presence or absence of RanQ69L (Fig. 6 A). As a control, we tested the same mutants for binding to importin β and found, as expected, that all deletion mutants containing the IBB domain (residues 26–65) also bound importin β (Fig. 6 B).

Deletion of as few as 26 amino acids from the NH₂ terminus prevented CRM1 binding and also nuclear export in the permeabilized HeLa cell assay (Fig. 6 A and data not shown). A snurportin 1 fragment lacking 74 amino acids from the COOH terminus (residues 1–285) retained the ability to bind CRM1 in the solution binding assay (Fig. 6 A); however, the kinetic assay indicated that the deletion lowered the affinity for CRM1 by 60% (not shown). A smaller fragment comprising residues 1–159 lost CRM1 binding entirely. Therefore, it appears that binding to CRM1 is sensitive to deletions from both the NH₂ and the COOH termini. Thus, in contrast to the short export signals characterized previously, snurportin 1 uses a large domain of at least 159 residues, but probably >285 residues, to bind CRM1. Such a domain could make more contacts to CRM1 and thereby bind more tightly than a simple NES. However, the main advantage might be that a conformational change in this domain can allow regulation of affinity.

We next wanted to know if snurportin 1 would also use CRM1 for reexport to the cytoplasm in Xenopus oocytes. As an initial experiment, we found that snurportin 1 indeed gets rapidly exported when injected into nuclei of Xenopus oocytes (not shown). If this rapid export is mediated by CRM1, then one would expect that snurportin 1 also competes other CRM1-mediated export pathways. To test this, we injected a mixture of ³²P-labeled RNAs consisting of DHFR mRNA, histone H4 mRNA, U1ΔSm RNA, U5ΔSm RNA, U6Δss RNA, and tRNA_i^Met into nuclei of Xenopus oocytes. After 180 min, the mRNAs, the U1 and U5 RNAs, and the tRNA had been efficiently exported to the cytoplasm (Fig. 7). U snRNA export is known to require CRM1 (Fornerod et al., 1997a) and when GST-snurportin 1 was coinjected, export of U1 and U5 was severely inhibited, but export of tRNA and the two mRNAs remained unaffected. Thus, one can conclude that snurportin 1 accesses the CRM1 export pathway also in this cell type. Furthermore, the very high affinity for CRM1 should make snurportin 1 a quite useful tool to characterize export of other substrates, and sensitivity towards competition by snurportin 1 should be a stringent test for a CRM1 requirement.

Transport between nucleus and cytoplasm appears to be largely mediated by adapter molecules and Ran-binding transport receptors that shuttle between the two compartments. Multiple rounds of import require a cyclic process that results not only in the transport of the cargo to the other side of the nuclear membrane, but also in the subsequent restoration of the original state of the transport system (recycling).

We have investigated here a recycling reaction for U snRNP import, namely the reexport of snurportin 1 back to the cytoplasm (Fig. 8). Snurportin 1 binds m₃G-capped U snRNP in the cytoplasm (Huber et al., 1998) and is imported by importin β into the nucleus (Fig. 4; Palacios et al., 1997; Huber et al., 1998). There, importin β binds RanGTP, which destabilizes the interaction with snurportin 1 (Fig. 6). Interestingly, the snurportin/importin β/RanGTP complex is considerably more stable than the corresponding complex with importin α and can even be isolated in small amounts (Figs. 2 B and 6). This could indicate that the actual dissociation of snurportin 1 from importin β does not occur immediately after NPC passage, but further inside the nucleus. Snurportin 1, then, has to release the m₃G-capped import substrate before it can form a trimeric export complex with CRM1 and RanGTP. The formation of this complex is highly cooperative and we estimate that snurportin 1 increases the affinity of CRM1 for RanGTP ∼1,000 fold (Fig. 3 A). The export complex would then be transferred to the cytoplasm, where it must be disassembled.

This disassembly apparently occurs in several steps, involves GTP hydrolysis, and requires RanBP1 and RanGAP1 (Fig. 3 A). CRM1-bound RanGTP resists GTPase activation by RanGAP. Therefore, RanBP1 probably causes an initial release of Ran in the form of a RanBP1/ RanGTP complex in which the RanGTPase can be activated by RanGAP1 (Bischoff and Görlich, 1997). GTP hydrolysis is the irreversible step of the disassembly because RanGDP cannot rebind to CRM1. Removal of RanGTP could alternatively be accomplished by the RanBP2/ SUMO-RanGAP1 complex that is located at the cytoplasmic filaments of the NPC (Mahajan et al., 1997; Saitoh et al., 1997).

Although Ran-free CRM1 is the low affinity form, the remaining CRM1/snurportin 1 heterodimer is still quite stable. However, binding of an m₃G-capped substrate displaces CRM1 (Fig. 5). The snurportin/importin β complex would then form and mediate import of the next U snRNP.

For productive transport cycles, carrier molecules need to have different affinities for their substrates during nuclear entry and exit. Ran-binding transport receptors can regulate their affinity for cargo with the aid of the RanGTPase. Adapter molecules do not directly bind RanGTP and, instead, they apparently use differential binding of transport receptors to coordinate the interaction with their cargoes. The first example for that has been import by importin α. Binding of importin β on the way into the nucleus increases the affinity importin α for the NLS (Rexach and Blobel, 1995; Görlich et al., 1996c). Conversely, CAS preferentially binds and exports NLS-free importin α (Kutay et al., 1997). A second precedence is the nuclear CBC which binds 7 monomethyl-cap structures and promotes export of m⁷G-capped RNAs (Izaurralde et al., 1995). After export, CBC needs to release the RNA before it is reimported by the importin α/β heterodimer. CBC can simultaneously bind a m⁷G-cap and importin α. However, binding of importin β to the importin α/CBC/ RNA complex causes the displacement of the RNA, allowing CBC to reenter the nucleus without its export substrate (Görlich et al., 1996b).

Here we show that the affinity of snurportin 1 for its import substrate is regulated by CRM1. The molecular basis for the mutual exclusivity of CRM1 and m₃G-cap binding to snurportin 1 is still unknown. The two binding sites in snurportin 1 are not identical because the 65 NH₂-terminal residues are dispensable for m₃G-cap binding (Huber et al., 1998) but are required for the interaction with CRM1 (Fig. 6). However, the binding sites could overlap and thereby exclude simultaneous binding. Alternatively, snurportin 1 might adopt two distinct conformations, one with a high affinity for a m₃G-cap, but weak CRM1 binding, and the other with the opposite preference. U snRNP release from snurportin 1 could occur spontaneously and rebinding could be prevented by CRM1. Alternatively, the release would be more efficient if CRM1 could actively displace m₃G from snurportin, i.e., not only shift the equilibrium, but also increase the off-rate. In addition, it should be noted that formation of spliceosomes would also retain the mature U snRNP in the nucleus.

Since its discovery as the receptor for leucine-rich NESs (Fornerod et al., 1997b; Stade et al., 1997), CRM1 has been implicated in nuclear export of a number of substrates including MAPKs (Ferrigno et al., 1998), cyclin B1 (Hagting et al., 1998; Yang et al., 1998), NFAT (Kehlenbach et al., 1998), actin (Wada et al., 1998), Dsk1p (Fukuda et al., 1997), and Pap1p (Kuge et al., 1998; Toone et al., 1998). The involvement of CRM1 in the export of the majority of these substrates was suggested by the identification of a leucine-rich NES and/or their sensitivity to the drug leptomycin B, a CRM1 inhibitor (Nishi et al., 1994). A direct physical interaction between CRM1 and its potential export substrates has not always been demonstrated and it is still possible that additional bridging factors are involved. Apart from mediating the export of RRE-containing viral RNAs through Rev, CRM1 is also believed to participate in the export of U snRNPs (Fischer et al., 1995; Fornerod et al., 1997a) and possibly in mRNA export in yeast (Segref et al., 1997; Stade et al., 1997; Watkins et al., 1998). Therefore, it appears that from the three exportins identified so far CRM1 has the broadest range of substrates. In contrast, CAS (Kutay et al., 1997) and exportin-t (Arts et al., 1998; Hellmuth et al., 1998; Kutay et al., 1998; Sarkar and Hopper, 1998) appear to have each specialized in the export of a single extremely abundant class of export substrates, the members of the importin α family and tRNA, respectively.

The question of whether CRM1 is involved in mRNA export in higher eukaryotes has been a controversial issue recently (reviewed by Stutz and Rosbash, 1998). In contrast to the experimental evidence against a role of Crm1p in mRNA export provided by Fischer et al. (1995) and Fornerod et al. (1997a), a later study arrived at the opposite conclusion (Pasquinelli et al., 1997). The fact that mRNA export from Xenopus oocytes nuclei resists competition by snurportin 1 (Fig. 7) confirms the conclusions by Fornerod et al. (1997a) and Fischer et al. (1995) and makes it very unlikely that mRNA export in higher eukaryotes is mediated by CRM1.

CRM1 substrates other than snurportin 1 have been suggested to interact with CRM1 by virtue of a leucine-rich stretch of 8–10 amino acids, a prototype being the Rev activation domain (Fischer et al., 1995; Wen et al., 1995; Bogerd et al., 1996). Surprisingly, however, the isolated activation domain binds CRM1 100 times weaker than the full-length Rev protein (Fig. 5 D). This could have two explanations. First, a leucine-rich export signal might need an appropriate protein context to adopt the conformation required for high-affinity CRM1 binding. In addition, our data indicate that residues flanking a given NES can also significantly contribute to its affinity for CRM1 (Fig. 5 D).

CRM1 binds snurportin 1 still 50-fold stronger than the full-length Rev protein. The probable reason for this difference is that CRM1 has to displace the avidly binding m₃G-capped import substrate. The structural basis for the high affinity of the snurportin/CRM1 interaction is still unclear, but nearly the entire snurportin 1 molecule appears to be required for it. The amino acid sequence of snurportin 1 does not contain a single perfect match to the consensus leucine-rich NES, but several regions that appear to resemble it. The affinity might be that high because several of these imperfect NESs interact synergistically with the same CRM1 molecule. However, the interaction of CRM1 with snurportin 1 might also be completely different from that with the prototype NES. This would not be the first case for a transport receptor recognizing very different signals. Transportin, for example, can bind the M9 import signal, which does not have a single essential basic residue, and it also binds the extremely basic import signal from the ribosomal protein L23a (Pollard et al., 1996; Jäkel and Görlich, 1998).

We would like to thank Petra Schwarzmaier for excellent technical help, Dr. Sekine for the kind gift of the m₃G-cap oligonucleotide, Drs. Maarten Fornerod and Iain W. Mattaj for a Crm1 clone, and Dr. Howard Fried for critical reading of the manuscript.

This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFBs 352 and 286) and the Human Frontier Science Programme Organization (grant RG0198/1998M).

CBC

cap-binding complex

GAP

GTPase-activating protein

IBB domain

importin β–binding domain

NES

nuclear export signal

NLS

nuclear localization signal

NPC

nuclear pore complex

U snRNP

uridine-rich small nuclear RNP