Putative Reaction Intermediates in Crm1-mediated Nuclear Protein Export*

We discovered several novel interactions between proteins involved in Crm1-mediated nuclear export of the nuclear export signal containing human immunodeficiency virus type 1 protein Rev. First, a Rev/Crm1/RanGTP complex (where Ran is Ras-related nuclear protein) reacts with some nucleoporins (Nup42 and Nup159) but not others (NSP1, Nup116, and Nup1), forming a Nup/Crm1/RanGTP complex and concomitantly releasing Rev. Second, RanBP1 (or homologous proteins) can displace Nup and form a ternary RanBP1/RanGTP/Crm1 complex that can be disassembled by RanGAP via GTP hydrolysis. Third, and most surprisingly, RanBP1/RanGTP/Crm1 can be disassembled without GTP hydrolysis by the nucleotide exchange factor RanGEF. Recycling of a Ran/RanGEF complex by GTP and Mg2+ is stimulated by both Crm1 and Rev, allowing reformation of a Rev/Crm1/RanGTP complex. Based on these reactions we propose a model for Crm1-mediated export.

One of the major unresolved problems in bidirectional macromolecular traffic across the nuclear pore complex (NPC) 1 is the sequence of events in which soluble transport factors interact with a distinct subset of nucleoporins (nups) (a collective term for NPC proteins). These interactions between the mobile and the stationary phase of transport are thought to facilitate translocation of cargo across the approximately 200-nm distance of the NPC (reviewed in Ref. 1). Several distinct import and export pathways have been identified each mediated by a signal in the transport substrate and by cognate signal recognition factors (reviewed in Refs. 2 and 3). The latter are members of a structurally related family of proteins and are referred to as karyopherins (kaps) (also termed importins, exportins, transportins, Ran-binding proteins, or various three-letter terms in Saccharomyces cerevisiae). The key regulator crucial for all import and export pathways is the small GTPase Ran. Ran exists in a GDP-or GTP-bound form. The location of RanGAP primarily in the cytoplasm (4) and RanGEF primarily in the nucleus (5,6) is believed to generate a nucleocytoplasmic gradient of RanGTP (nucleoplasm) and RanGDP (cytoplasm) across the NPC, which is thought to be one of the determinants for directionality of transport (7). Another likely determinant is the asymmetric distribution of nups on either side of the NPC. Ran is further modulated by two small Ran-binding proteins as follows: RanBP1 (Ran-binding protein 1), which binds RanGTP and stimulates GTP hydrolysis by RanGAP but inhibits GTP exchange by RanGEF (8,9), and p10/NTF2 (nuclear transport factor 2), which only binds Ran in its GDP-bound form (10 -12). Both of these proteins play a role in transport (13)(14)(15)(16)(17).
Much of our understanding of nuclear transport comes from studies of nuclear protein import. In the classical import pathway a cNLS (classical nuclear localization signal) is recognized and docked to the NPC by the kap ␣/␤1 heterodimer (reviewed in Ref. 3). The cNLS-bound kap ␣/␤1 binds to several peptide repeat-containing nups in vitro (18 -20), which are localized along the 200-nm-long nucleocytoplasmic axis of the NPC from the tips of the cytoplasmic fibrils to the terminal ring collating the nucleoplasmic fibrils (reviewed in Refs. 21 and 22). Binding of a cNLS/kap ␣/␤1 complex to some of these repeat-containing nups in vitro results in release of the cNLS protein (18). However, release of the kap ␣/␤1 complex from the nup requires Ran in its GTP-bound form (18,23) and results in the formation of a kap ␤1/RanGTP complex that is protected against GTP hydrolysis by RanGAP (24 -26) or GTP exchange by RanGEF (25,27). Recycling of kap ␤1 and RanGTP requires kap ␣, a nup, and either RanGAP or RanGEF (27) and is further stimulated by RanBP1 (27,28). Together these data led to the proposal that RanGTP may function in repeated cycles of docking and undocking along the cytonucleoplasmic axis of the NPC (18,27,29). Alternatively, RanGTP may exert its dissociative abilities only at the end point of transport on the nucleoplasmic side of the NPC (26).
Crm1 (also known as exportin1, Xpo1, or Kap124)-mediated export of NES-containing proteins was the first nuclear export pathway to be discovered (30 -34). CRM1 was originally identified in S. cerevisiae as a gene involved in chromosomal region maintenance (35). Recently, the human homolog was found to be a soluble protein that interacts with the nucleoporin Nup214/CAN (36). Evidence that Crm1 is a kap came from inhibitor studies with the antifungal drug leptomycin B (LMB). In Schizosaccharomyces pombe LMB was shown to target Crm1 (37). When LMB was found to inhibit nuclear export of the HIV-1 protein Rev, Crm1 was suggested to be the kap that exports Rev from the nucleus (30).
Rev transports unspliced viral mRNA from the nucleus to the cytoplasm (38 -40). Rev contains an RNA-binding site that interacts with a specific Rev response element (RRE) in viral mRNA (41)(42)(43). In addition to a cNLS Rev also contains an NES, a leucine-rich stretch of amino acids that is required and sufficient for nuclear export of Rev (42,44,45). Similar NESs were also found in other proteins (46 -48), which suggested the existence of a receptor that recognizes an NES and exports the protein from the nucleus. Previous studies have shown direct binding of Crm1 to an NES peptide (31)(32)(33), which might involve RanGTP (31). It is now thought that Crm1 and NES interact in the absence of Ran but that a ternary complex is formed when RanGTP is present (49). The interaction between Crm1 and NES is sensitive to LMB (31)(32)(33), and recent data suggest that in particular the formation of the Rev/Crm1/ RanGTP complex is inhibited by LMB (49) explaining the inhibitory effect of LMB on Rev export (30). A Rev interacting protein, termed Rip1, has been previously identified (50,51). Rip1 is an FG peptide repeat containing nucleoporin (51) and has therefore been termed Nup42. The Nup42/Rev interaction is bridged by Crm1 (52). Furthermore, Crm1 was shown to export an NES reporter protein (34). Together these findings have defined a Crm1-mediated nuclear export pathway. However, the molecular mechanism of Crm1-mediated protein export is unknown.
We have investigated Crm1-mediated nuclear export using GTP hydrolysis and exchange assays, as well as a solution binding assay, all with recombinant yeast proteins. The export substrate we have used for our studies was the HIV-1 protein Rev. As previously proposed for NLS-mediated import, we suggest that NES-dependent nuclear export may involve multiple rounds of docking and undocking at the NPC proteins.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Yeast Ran and RanGAP were expressed and purified as described (24). Yeast RanGEF, RanBP1, Nup36, and the FXFG peptide repeat region of Nup1 were expressed and purified as described (27). HIV-1 Rev protein was purchased from Intracel.
Crm1 was expressed as a GST fusion protein. The gene encoding Crm1 was amplified by polymerase chain reaction (PCR) from S. cerevisiae genomic DNA (Promega). The PCR product was inserted as a BamHI/SalI fragment into vector pGEX-4T3 (Amersham Pharmacia Biotech), into which a Tev protease cleavage site had been inserted (this vector was kindly provided by Y. M. Chook). 2 The protein was expressed in Escherichia coli strain DH5␣. The GST fusion protein was purified on glutathione-Sepharose beads (Amersham Pharmacia Biotech), and the GST moiety was cleaved by incubation with 1,000 units of Tev protease (Life Technologies, Inc.) per 20 mg of protein for 12 h at 21°C. The purified protein was stored in frozen aliquots at Ϫ80°C.
To obtain recombinant Nup42 protein the gene encoding Nup42 was amplified from S. cerevisiae genomic DNA by PCR. The resulting PCR product was inserted into pGEX-4T3 as a BamHI/SalI fragment. The protein was expressed in E. coli strain DH5␣, purified, and cleaved with thrombin as described for other GST proteins (24). The protein was stored in frozen aliquots at Ϫ80°C.
Recombinant Nup116 and NSP1 were expressed as GST fusion proteins. The genes encoding Nup116 and NSP1 were amplified from S. cerevisiae genomic DNA by PCR. The PCR products were inserted as a BglII/XhoI or BamHI/SmaI fragment, respectively, into vector pGEX-4T3. The GST fusion proteins were expressed in E. coli strain BL21(DE3). The proteins were purified on glutathione-Sepharose beads (Amersham Pharmacia Biotech) and eluted with 10 mM glutathione as described for other GST fusion proteins (27). The purified proteins were stored in frozen aliquots at Ϫ80°C.
The fragment of Nup159 containing its FG peptide repeat region was expressed as a His fusion protein as described (53). The protein was purified on a His-Trap column (Amersham Pharmacia Biotech) using a linear gradient of imidazole (5-500 mM) in buffer A (150 mM KOAc, 20 mM Hepes (pH 7.3), 2 mM Mg(OAc) 2 ). The protein eluted at 200 -250 mM imidazole. Fractions containing Nup159 were pooled, concentrated, and fractionated on a Superdex 200 FPLC column (Amersham Pharmacia Biotech) equilibrated with buffer A containing 1 mM dithiothreitol. Fractions containing Nup159 were pooled, and frozen aliquots were stored at Ϫ80°C.
To obtain GST-Ran the gene encoding Ran was amplified from S. cerevisiae genomic DNA by PCR. The PCR product was inserted as a BamHI fragment into vector pGEX-2TK (Amersham Pharmacia Biotech). The protein was purified as described for GST fusion proteins (27) in phosphate-buffered saline buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 (pH 7.3)). GST-Ran was found to be mostly nucleotide-free in phosphate-buffered saline buffer, as determined by a previously described method (24). GST-Ran could be loaded with GTP by incubation with 1 mM GTP in the presence of 5 mM Mg(OAc) 2 for 1 h at 21°C. Free nucleotide was removed by dialysis against buffer A containing 1 mM dithiothreitol. This resulted in GST-Ran that was mostly GTP-bound (80 -90%).
GTP Hydrolysis and Exchange Assays-Assays were performed as described previously (27). For hydrolysis assays 0.5 nM Ran-[␥-32 P]GTP was incubated with other proteins as indicated in the figure legends for 15 min at 21°C. Then 15 nM RanGAP was added, and the reactions were incubated for 20 min at 21°C. To determine the dissociation constants for RanGTP complexes, we also measured inhibition of GTP hydrolysis after 2 min reaction time. We did not observe any differences between the K D values measured after a 2-min or a 20-min reaction. For exchange assays 0.5 or 10 nM Ran-[␣- 32  Solution Binding Assays-Assay were performed essentially as described (27). GST-RanGTP was incubated with glutathione-agarose beads (Sigma) for 20 min at 21°C (7 g of GST-RanGTP per 10 l of beads) in 0.5 ml of binding buffer (150 mM KOAc, 20 mM Hepes (pH 7.3), 2 mM Mg(OAc) 2 , 1 mM dithiothreitol, 0.1% Tween 20, and 0.1% casamino acids). For the one-step assay the following procedure was followed: after washing with binding buffer, 10 l beads per reaction was incubated with other proteins as indicated in the figure legends in a final volume of 40 l for 30 min at 21°C. For the two-step assay the following procedure was followed: after washing with binding buffer, the immobilized GST-Ran was incubated with 1.7 M RanGEF for 20 min at 21°C. The beads were washed and 10 l of beads per reaction was incubated with GTP and other proteins as indicated in a final volume of 40 l for 10 min at 21°C. The unbound fraction was collected by removing 28 l from the meniscus after centrifugation at 2,000 ϫ g for 1 min. The beads were washed twice with 0.5 ml of binding buffer. Then 18 l of binding buffer was added; this constitutes the bound fraction. 6 l of 6ϫ Laemmli sample buffer with 2-mercaptoethanol was added to the bound and unbound fractions. After incubation at 95°C for 10 min, the proteins in 14 l of each sample were resolved by SDS-PAGE using 4 -20% acrylamide gradient gels and stained with Coomassie Blue.

RESULTS
A likely candidate for a Crm1 interacting nucleoporin is Nup42, based on previously reported in vivo data (50 -52). To investigate complex formation between Crm1, Nup42, Rev, and RanGTP, we immobilized GST-Ran that contained 80 -90% GTP (see "Experimental Procedures") on glutathione beads. The final concentration of GST-RanGTP was 3 M in these experiments. We found that Rev and Crm1 form a complex with RanGTP ( Fig. 1, lane 2), as has been previously described (49). RanGTP if present at a concentration of 1.5 M. If Nup42 was added together with Rev and Crm1, a Nup42/Crm1/RanGTP complex was formed (lane 3). In the presence of Nup42 the amount of Crm1 recruited to GST-RanGTP was significantly increased (compare lanes 1 and 3). However, Rev was excluded from this complex (lane 3). Nup42 did not bind to GST-RanGTP in the absence of Crm1 (lane 4). Together these results suggest that Rev is released when the Rev/Crm1/RanGTP complex binds to Nup42.
We have previously used protection of RanGTP from Ran-GAP-induced GTP hydrolysis to measure RanGTP complex formation (24). We therefore investigated whether the Nup42/ Crm1/RanGTP complex is protected against RanGAP (Fig. 2). Preincubation of 0.5 nM RanGTP with increasing amounts of Nup42 in the presence, but not in the absence, of 1 M Crm1 (compare Fig. 2, open circles to open squares) resulted in complete inhibition of RanGAP-stimulated GTP hydrolysis. Incubation of RanGTP with 1 M Crm1 alone did not result in detectable levels of RanGAP inhibition (not shown). Together, the results from Fig. 1 and Fig. 2 suggest that RanGTP, Crm1, and Nup42 might form a cooperative complex. A constant for RanGTP dissociation from Nup42/Crm1/RanGTP complex is estimated to be 60 nM from these experiments. Crm1 and RanGTP were also found to form a complex with Nup159*, a fragment of the nucleoporin Nup159 containing its FG repeat region (53) (closed circles). However, the affinity of this complex for RanGTP was lower, and a K D of 100 nM was estimated. These estimates reflect average dissociation constants, since the stoichiometry of the Nup42/Crm1/RanGTP or Nup159/ Crm1/RanGTP complex is not known. Crm1 and RanGTP did not form a complex with the nucleoporins NSP1, Nup116, or Nup1*, a fragment of Nup1 containing its FXFG repeat region. Complex formation of Crm1, RanGTP, and Nup42 was also found to inhibit GTP exchange by RanGEF (Fig. 3). In the presence of 1 M Crm1 and increasing amounts of Nup42 GTP exchange on Ran by 20 nM RanGEF was inhibited (Fig. 3, open circles). Incubation of RanGTP with 1 M Rev and 1 M Crm1 did not inhibit RanGAP-induced GTP hydrolysis or GTP exchange by RanGEF. 3 Neither did incubation of RanGTP with 1 M Rev, 1 M Crm1, and increasing amounts of Nup42 further stimulate the inhibition seen in the presence of Crm1 and Nup42 (not shown). Inhibition of RanGEF-induced exchange of GTP was also seen in the presence of Crm1 and Nup159* (not shown). Together these results indicate that Crm1, RanGTP, and Nup42 or Nup159* form a complex independently of an NES-containing export substrate.
How are Crm1 and RanGTP released from Nup42? Since complex formation of RanGTP with Crm1 and Nup42 results in inhibition of RanGAP-induced GTP hydrolysis, we assayed for complex disassembly by measuring release of RanGAP inhibition. Addition of the Ran-binding protein RanBP1 to the Nup42/Crm1/RanGTP complex completely restored RanGAPinduced GTP hydrolysis (Fig. 4A). For these experiments 0.5 nM RanGTP was preincubated with 0.7 M Crm1 and 0.25 M Nup42. Addition of increasing amounts of RanBP1 in the presence of 15 nM RanGAP completely restored RanGAP-induced GTP hydrolysis. 50% RanGAP-induced GTP hydrolysis occurred in the presence of 0.7 nM RanBP1. The concentration of RanGTP used in this experiment is too high to allow an accurate measurement of the interaction between RanBP1 and Nup42/Crm1/RanGTP complex. However, an estimate for the free concentration of RanBP1 at 50% GTP hydrolysis of 0.2 nM is close to the K D for the RanBP1/RanGTP interaction, which has been determined to be between 0.1 and 0.6 nM for mammalian proteins (26,54,55).
To understand further the mechanism of Crm1/RanGTP release from Nup42 by RanBP1, we investigated whether Crm1, RanGTP, and RanBP1 form a complex. For these experiments GST-RanGTP was immobilized (Fig. 4B). Addition of RanBP1 in the presence of Crm1 resulted in binding of RanBP1 to RanGTP and an increase in Crm1 binding to RanGTP compared with Crm1 alone (Fig. 4B, compare lanes 1 and 2). This result suggests the formation of a cooperative complex between Crm1, RanGTP, and RanBP1. Together these results indicate that RanBP1 dissociates the Crm1/RanGTP complex from Nup42, thereby forming a Crm1/RanGTP/RanBP1 complex. Interestingly, we found that Nup36, also known as Yrb2, can disassemble Crm1/RanGTP from Nup42. 3 Nup36 is one of two RBH (RanBP1-homologous) domain containing proteins in S. cerevisiae, the other being Nup2 (11, 56 -58). We found that Nup36 also forms a complex with RanGTP and Crm1 in the GST-binding assay (Fig. 4B, lane 3). These data indicate that Nup36 and RanBP1 might have overlapping functions. However, the exact function of a Crm1/RanGTP/Nup36 complex remains to be determined.
One way of disassembling the Crm1/RanGTP/RanBP1 complex is via RanGAP-induced GTP hydrolysis (Fig. 4A). We found an alternative mechanism of Crm1/RanGTP/RanBP1 complex disassembly by the exchange factor RanGEF. Preincubation of about 6 nM RanGTP with increasing amounts of RanBP1 inhibited GTP exchange by 20 nM RanGEF (Fig. 5A,  closed circles). When RanGTP was preincubated with 1 M Crm1 and increasing amounts of RanBP1, a shift in the inhibition curve suggested the formation of a Crm1/RanGTP/ RanBP1 complex (Fig. 5A, open circles). This ternary complex might be formed in a cooperative manner, since in the presence of Crm1 less RanBP1 is required for inhibition of RanGEFstimulated GTP exchange on Ran (Fig. 5A, compare closed  circles to open circles). Surprisingly, if the concentration of RanGEF was increased to 80 nM, this Crm1/RanGTP/RanBP1 complex was no longer protected against RanGEF-induced GTP exchange (Fig. 5B, open circles). Since the concentrations of Crm1 and RanBP1 were sufficient to allow for Crm1/ RanGTP/RanBP1 complex formation under these conditions (compare with Fig. 5A, open circles), this result might indicate that the Crm1/RanGTP/RanBP1 complex, which is formed when all three components are preincubated, is disrupted by 80 nM RanGEF. Interestingly, a RanGTP/RanBP1 complex that is formed at higher concentrations of RanBP1 was not sensitive to increased concentrations of RanGEF (compare Fig. 5A, closed circles, to Fig. 5B, closed circles). All the experiments shown in Fig. 5 were done at least three times. These data might suggest that Crm1/RanGTP/RanBP1 complex can be attacked by RanGEF if present at sufficiently high levels. However, since in this experiment only release of GTP was measured, we could not determine whether the complex is disassembled by RanGEF. Most likely, however, release of GTP weakens the affinity of Crm1 and RanBP1 for RanGTP, resulting in disassembly of the complex and formation of an intermediate Ran/ RanGEF complex.
We found that the Crm1/RanGTP/Nup36 complex is also sensitive to increased concentrations of RanGEF. In the presence of 1 M Crm1 and increasing amounts of Nup36 GTP exchange on Ran by 20 nM RanGEF was inhibited (Fig. 5C, open circles). However, in the presence of 80 nM RanGEF the Crm1/RanGTP/Nup36 complex was attacked by RanGEF (Fig.  5D, open circles). We did not find complex formation between RanGTP and Nup36 in the absence of Crm1 at the concentrations of Nup36 used in these experiments (Fig. 5, C and D,  closed circles).
Nucleotide exchange by RanGEF is thought to occur through release of the nucleotide from Ran upon binding to RanGEF (59). This results in the formation of an intermediate nucleotide-free Ran/RanGEF complex. Is dissociation of this intermediate complex and rebinding of GTP to Ran in the presence of Mg 2ϩ stimulated by other factors? For these experiments GST-Ran was immobilized and preincubated with RanGEF, which resulted in the formation of a Ran/RanGEF complex (Fig. 6,  lane 1). Addition of GTP in the presence of 2 mM Mg(OAc) 2 stimulated release of RanGEF from Ran as expected (lane 2). Interestingly, addition of Crm1 in the presence of GTP further stimulated release of RanGEF from Ran (lane 3). If Rev was added together with Crm1 all the bound RanGEF was released (lane 4). However, addition of Nup42 together with Crm1 did not stimulate release of RanGEF over the levels seen with Crm1 alone (lane 5). If Rev was added together with Nup42 and Crm1, release of RanGEF was restored (lane 6). The differences in the amount of RanGEF released in the presence of different proteins were small but were reproducibly seen in three independent experiments. Together these data suggest that binding of GTP and release of RanGEF from Ran is stimulated by Crm1 and Rev. DISCUSSION We have analyzed the interactions between proteins involved in Crm1-mediated nuclear protein export. A model for the mechanism of protein export is shown in Fig. 7. Export is presumably initiated by the formation of an NES/Crm1/ RanGTP complex in the nucleus (see Fig. 7; Fig. 1, lane 2), where RanGTP concentrations are thought to be high due to the predominantly nuclear localization of RanGEF (5, 6) and the low concentrations of RanGAP in the nucleus (4,60). It has recently been described that a ternary Rev/Crm1/RanGTP complex is formed (49). This complex is also formed when Rev is bound to an RRE containing RNA (49,61). Similar export complexes involving the substrate, the export kap, and RanGTP have been described for the export of kap ␣/importin ␣ via CAS/Kap109 (62), the human homolog of yeast Cse1 (63,64), as well as for tRNA export via exportin-t (65), the human homolog of the yeast kap Los1 (66,67). Rev/Crm1/RanGTP complex has a low affinity for RanGTP, since RanGAP-induced GTP hydrolysis on Ran is not inhibited in the presence of 1 M Crm1 and 1 M Rev. 3 Recently, it has been reported that a RRE/Rev/Crm1/RanGTP complex is protected against Ran-GAP-induced GTP hydrolysis (49). We do not currently know whether the presence of an RRE containing RNA increases the affinity of Rev for Crm1 and RanGTP, so that RanGTP is protected against RanGAP at micromolar concentrations of Rev and Crm1. This complex was at least partially disassembled in the presence of RanBP1 and RanGAP (49). Disassembly of an export cargo/kap/RanGTP complex by RanBP1 and RanGAP has also been shown for the kap ␣/CAS/RanGTP complex (62).
The NES/Crm1/RanGTP complex is likely to be the complex that docks at the NPC in vivo. We have shown that Rev no longer binds to Crm1/RanGTP in the presence of Nup42, suggesting that binding of Rev/Crm1/RanGTP complex to Nup42 results in release of Rev and formation of a ternary Nup42/ Crm1/RanGTP complex (Fig. 7, step 1; Figs. 1-3). Release of the export substrate upon binding of an NES/Crm1/RanGTP complex to a nup is reminiscent of release of a cNLS protein upon binding of a cNLS/kap ␣/␤1 complex to certain nups (18). We do not currently know whether release of the substrate, i.e. the cNLS or NES protein, upon docking at the NPC occurs in vivo. However, one might speculate that release of the substrate upon docking is favorable, if the rate-limiting step in transport is diffusion of the substrate across the NPC. Dissociation of the substrate upon docking would immediately allow its diffusion to a next docking site independently of the release of its kap.
Nup42 was originally identified as a Rev-interacting protein in a yeast two-hybrid screen or a copper resistance assay (50,51), and the interaction was shown to be mediated by Crm1 (52). Our data indicate that Rev is released when Rev/Crm1/ RanGTP binds to Nup42. This finding suggests that the Rev/ Nup42 interaction seen in the yeast two-hybrid assay might constitute an intermediate complex between Rev, Crm1, RanGTP, and Nup42. This quaternary complex would be shortlived under physiological conditions but might be stabilized under the conditions of the two-hybrid assay. Alternatively, another factor might stabilize a Rev/Crm1/RanGTP/Nup42 complex in vivo. Interestingly, we found that binding of Crm1 to a nup occurs in the presence of RanGTP (Figs. 1-3). This is in contrast to nuclear protein import, where RanGTP prevents binding of kap ␤1 to some repeat-containing nups (18,27) and releases a kap ␣/␤1 complex from the nup (18). However, recent findings indicate 4 that binding of kap ␣/␤1 to a nup can also occur in the presence of RanGTP, if the nup contains an RBH domain like mammalian Nup358 (68). Nup42 is an FG peptide repeat containing nucleoporin (51), but its exact localization at the NPC remains elusive. Involvement of Nup42 in export of Rev from the nucleus in vivo was suggested by microinjection experiments in Xenopus oocytes (69). Injection of a fragment of Nup42 containing its FG repeat region into the nucleus of Xenopus oocytes blocked nuclear export of Rev. This result suggested that a Rev-containing export complex might be able to bind Nup42 in vivo. Nevertheless, the gene encoding for Nup42 is not essential, and a Nup42 deletion in yeast does not result in mislocalization of an NES-green fluorescent protein-NLS reporter protein (34). This finding suggests that other nups might take over the function of Nup42 in its absence. Interestingly, Nup42 is highly homologous to yeast Nup159 (51), which contains FG repeats and has been localized to the cytoplasmic fibrils of the NPC (53). We found that Crm1 and RanGTP form a complex with the FG repeat region of Nup159 (Fig. 2). If multiple docking steps occur during transport of an NES protein across the NPC, one could speculate that Nup159 constitutes one of the last docking sites at the cytoplasmic side of the NPC. Interestingly, injection of a fragment of Nup159 containing its FG repeat region into the nucleus of Xenopus oocytes also blocked Rev export (69). Furthermore, Crm1 has been found in a complex with the mammalian nucleoporin Nup214/CAN in vivo (36). Nup214/CAN contains FG repeats and is the closest mammalian homolog of yeast Nup42 and Nup159 (51). We did not find complex formation of Crm1 and RanGTP with the nucleoporins NSP1, Nup116, or the FXFG repeat region of Nup1, when the nups were present at 1 M concentrations in the GTP hydrolysis assay (Fig. 2). Neither was complex formation with these nups found in the presence of 1 M Crm1 and 1 M Rev. 3 Interestingly, previous studies on Rev interactions with different nups using a copper resistance assay identified interactions of Rev with Nup42, the FG repeat region of Nup159 and the GLFG repeat region of Nup100, respectively (51). However, the FXFG repeat regions of NSP1 or Nup1 or the GLFG repeat region of Nup116 did not interact with Rev in these assays. These results are in agreement with our data on complex formation in vitro. Our data suggest that only a subset of nups interact with the export kap Crm1. However, it cannot be excluded that complex formation of RanGTP and Crm1 (with or without Rev) with other nups occurs only at higher nup concentrations.
We show here that Crm1/RanGTP is released from Nup42 by the Ran-interacting protein RanBP1 (Fig. 7, step 2; Fig. 4A). Release by RanBP1 results in formation of a Crm1/RanGTP/ RanBP1 complex (Fig. 4B, lane 2, and Fig. 5A). RanBP1 has been postulated to contain an NES (47). This might have interesting implications for the mechanism by which RanBP1 releases Crm1/RanGTP from Nup42. However, since the Crm1/ RanGTP/RanBP1 complex has a much higher affinity for RanGTP than the Rev/Crm1/RanGTP complex as judged by their resistance to RanGEF-induced GTP exchange, 3 it is unlikely that RanBP1 interacts with Crm1 and RanGTP only through its NES. The Crm1/RanGTP/RanBP1 complex is susceptible to RanGAP-stimulated GTP hydrolysis (Fig. 7, step 3;  Fig. 4A). GTP hydrolysis is likely to disassemble the complex. Disassembly of the Crm1/RanGTP/RanBP1 complex via Ran-GAP might occur at the cytoplasmic side of the NPC, where RanGAP concentrations are high (4). However, RanGAP-stimulated GTP hydrolysis on Ran is not required for single turnover export, as previous results from microinjection experiments have suggested (70). Injection of the RanG19V mutant, which cannot interact with RanGAP, into nuclei of BHK21 cells did not affect export of an NES reporter protein. This result indicated that transport across the NPC does not require Ran-GAP-stimulated GTP hydrolysis on Ran. We have discovered an alternative way to disassemble the Crm1/RanGTP/RanBP1 complex that does not require RanGAP but the exchange factor RanGEF (Fig. 7, step 4; Fig. 5). Crm1/RanGTP/RanBP1 complex was found to be susceptible to increased concentrations of RanGEF resulting in release of GTP (Fig. 5B), which might weaken the interaction of Ran with Crm1 and RanBP1 and lead to disassembly of the complex. RanGEF might thereby form a binary complex with nucleotide-free Ran. It was also shown previously that Ran, RanBP1, and RanGEF form a complex that is nucleotide-free (8). Therefore, it might also be possible that an intermediate RanBP1/Ran/RanGEF complex is formed, when RanGEF interacts with Crm1/RanGTP/RanBP1. Our data on release of GTP from Crm1/RanGTP/RanBP1 complex by RanGEF were surprising, since a RanGTP/RanBP1 complex was resistant to RanGEF at the concentrations tested (Fig. 5B) in agreement with a requirement for RanGEF in export of U snRNA (72), which has been shown to be Crm1-mediated (31) and to compete with Rev export (45). Our data also shed new light on the observation that microinjection of the RanT24N mutant into nuclei of cells impairs export of an NES reporter protein (70) or export of U snRNA (7). The RanT24N mutant has a decreased affinity for nucleotide and an increased affinity for RanGEF (59,73). If RanT24N merely blocked export, because it would not allow formation of an NES/Crm1/RanGTP complex due to its decreased affinity for GTP, co-injection of RanGTP together with RanT24N should restore export. However, when RanG19V was injected together with RanT24N, export was not restored (70). This might indicate that the RanT24N mutant sequesters RanGEF. If RanGEF were necessary to disassemble the Crm1/RanGTP/ RanBP1 complex, depletion of RanGEF from the system by RanT24N would diminish disassembly of the Crm1/RanGTP/ RanBP1 complex. This would prevent recycling of Crm1 and RanGTP for a new round of docking.
Nup36 forms a ternary complex with Crm1 and RanGTP ( Fig. 4B and Fig. 5C), which is disassembled by high concentrations of RanGEF (Fig. 5D). Nup36 was further found to release Crm1 and RanGTP from Nup42 in the GTP hydrolysis assay. 3 These data might indicate that Nup36 could replace RanBP1 function. However, the constant of RanGTP dissociation from Crm1/RanGTP/Nup36 complex seems to be higher than from Crm1/RanGTP/RanBP1 complex (compare Fig. 5, A and C). Nup36 has been implicated in Crm1-mediated nuclear export in vivo and has been suggested to exist in a complex with Crm1 (74). Nup36 has further been shown to interact genetically with RanGEF and was also proposed to be associated with RanGEF (75). It is not clear to date whether Nup36 is a nucleoporin (11) or a nuclear factor (75) that might be transiently associated with the NPC. However, it is intriguing to speculate that Nup36 might play a role in the intranuclear phase of protein export.
Nucleotide exchange on Ran by RanGEF is thought to occur through destabilization of the Mg 2ϩ ion coordination to Ran by RanGEF, in analogy to recent findings for other small GTPases and their respective GEFs (76,77). This results in release of the nucleotide and formation of an intermediate nucleotide-free Ran/RanGEF complex (59). This Ran/RanGEF complex is then dissociated in the presence of nucleotide and Mg 2ϩ (59,78). We have shown that Crm1 and Rev stimulate dissociation of RanGEF from Ran in the presence of GTP and Mg 2ϩ (Fig. 7,  step 5; Fig. 6). The concentrations of GTP required for release of RanGEF from immobilized GST-Ran are much higher than GTP concentrations required for disassembly of the Ran/ RanGEF complex in solution (59,78). This might be due to the high concentrations of protein on the glutathione beads. Crm1 and Rev might stimulate release of RanGEF by stabilizing the newly formed RanGTP. However, Crm1 and Nup42 did not stimulate release of RanGEF from Ran over the levels seen with Crm1 alone (Fig. 6, lane 5). This result was surprising, since the Nup42/Crm1/RanGTP complex is likely to have a higher affinity for RanGTP than the Rev/Crm1/RanGTP complex, as judged by the resistance of these complexes to Ran-GAP. 3 Therefore, these data might indicate that Rev and Crm1 form a complex with RanGTP that is better protected against RanGEF than the Nup42/Crm1/RanGTP complex (Fig. 6). This would ensure that the newly formed Crm1/RanGTP complex is substrate-bound before it docks at a next nup.
A large fraction of Ran in the nucleus might be present in complex with RanGEF (78). One could therefore speculate that dissociation of Ran from RanGEF by NES protein bound Crm1 but not in the absence of an export cargo (Fig. 6) might be the event that initiates export. The subsequent docking and release reactions would result in the formation of complexes with increasing affinities for RanGTP (Fig. 7). The affinities we have measured in the GTP hydrolysis assay are a first approach to a quantitative treatment of these interactions. However, interactions of soluble transport factors with nucleoporins in the context of the NPC might be different than in solution. It should also be noted that local concentrations of transport factors at the NPC might be very high. The high affinity Crm1/ RanGTP/RanBP1 complex that is formed when RanBP1 releases Crm1/RanGTP from the nup can be disassembled via RanGAP-stimulated GTP hydrolysis. Alternatively, RanGEF can disassemble the complex by releasing GTP. Release of GTP may convert a high affinity complex into a lower affinity complex, leading to its disassembly. We propose an export mechanism that does not require GTP hydrolysis but rather nucleotide release by RanGEF. GTP hydrolysis might occur only on the cytoplasmic side of the NPC. Accumulation of RanGTP on the cytoplasmic side of the NPC would be detrimental to protein import (26,79) and presumably is tightly controlled.