Nuclear Import of Cdk/Cyclin Complexes: Identification of Distinct Mechanisms for Import of Cdk2/Cyclin E and Cdc2/Cyclin B1

Interphase egg extracts and the ATP-regenerating cocktail added to these extracts were prepared as described previously (Murray, 1991; Smythe and Newport, 1991). Egg cytosol for protein binding and transport assays was produced by centrifugation of crude interphase extracts for 70 min at 200,000 g. Depletions of cytosol were carried out essentially as described by Görlich et al. (1994).

Binding assays were typically performed by mixing 100 μl of egg cytosol (in some cases diluted two- to fourfold with extract buffer) with 10–20 μl of recombinant protein linked to glutathione agarose or amylose beads at 4°C or room temperature for 40–60 min. ³⁵S-labeled proteins were produced by coupled in vitro transcription/translation (TNT system; Promega) using translabel (New England Nuclear). Reticulocyte lysates were usually mixed 1:20 with diluted egg extract for binding assays. Where necessary, biotinylated nucleoplasmin, NP-core, or NP-core-importin-β binding (IBB) were added to extracts at a final concentration of 20 μg/ml. In some experiments, low speed egg extracts or extracts from Sf9 cells overexpressing Cdks were used instead. Beads were spun out of the extracts by centrifugation for 3 min at 2,000 g at 4°C and then washed three times with 1 ml of XB (Murray et al., 1989).

Assays to detect interactions between purified factors were carried out in PAT buffer (PBS, pH 7.2, 1% [wt/vol] cas-amino acids, 0.1% [vol/vol] Tween 20) instead of diluted extract. Binding assays were carried out at 4°C with 15 μl of immobilized protein on beads (∼1 μg protein/μl packed beads) in final volumes of 200–500 μl. Where appropriate, importin-α was added to 70–100 μg/ml, cyclins to 15–20 μg/ml, and Cdk subunits to ∼10 μg/ml. After a 45-min incubation, beads were pelleted and washed six times with 1 ml of PAT buffer.

Bead-bound proteins were eluted with SDS-PA gel loading dye and separated on 10% SDS-polyacrylamide gels. For detection of ³⁵S-labeled proteins, gels were soaked in 0.5 M sodium salicylate in 20% methanol, 5% acetic acid then dried and fluorographed. Otherwise, the separated proteins were transferred to PVDF membranes. Biotinylated proteins were detected with streptavidin-HRP conjugate. Other proteins were detected with unconjugated mouse or rabbit primary antibodies followed by HRP-conjugated goat antibody directed against the primary species used. Primary antibodies used in this study included commercial murine monoclonals against human cyclins E and B1 and rabbit antibodies raised against importin-α and -β (a kind gift of D. Görlich). Enhanced chemiluminescence was used to detect HRP activity.

To express full-length importin-α, we first constructed pGEX-SB, a modified version of GEX-KG (a GEX2T derivative with a polyglycine linker at the thrombin cleavage site) which introduces a novel SphI site whose ATG encodes a methionine residue in frame with the GST fusion tag. This was constructed by digesting pGEX-KG with BamHI and EcoRI and inserting a pair of complementary oligonucleotides containing a SphI site (5′-gat cga gca tgc gcc cgg gat ccg-3′ and 5′-aat tcg gat ccc ggg cgc atg ctc-3′). pGEX-importin-α was then made by ligating the importin-α encoding SphI-BamHI fragment from pQE70-importin-α (Görlich et al., 1994) to similarly digested pGEX-SB. To construct pGEX-IBB, and pGEX-IBB55, one of two NcoI-HindIII–digested PCR products encoding the NH₂ terminus of importin-α was inserted into similarly digested pGEX-KG (Görlich et al., 1996b). The same forward primer was used in both PCR reactions (5′-cat atg cac cat cat cat cat cat gcc atg gtg ccg acc aca aat gaa gca gat a-3′), the reverse primers were 5′-aag ctt aag cat tct ttt ctg gag aga gag aat-3′ for pGEX-IBB and 5′-cgg tac aag ctt cac tct tct ggt aag cat aca tt-3′ for pGEX-IBB55.

To express a fusion protein consisting of GST linked to human importin-β, a 1.8-kb partial NcoI-HindIII fragment from pQE60 importin-β (Görlich et al., 1995a) was ligated to NcoI-HindIII–digested pGEX-KG. The resulting plasmid, pGEX-importin-β, was subsequently used as a template in PCR reactions in which sequences encoding importin-β residues 1–462 were amplified using the forward primer 5′-gag gag gga tcc atg gag ctg atc acc att ctc-3′ and reverse primer 5′-cga tgc gag ctc taa gca ctg aga ccc tca atc ag-3′. The PCR product was digested with BamHI-SalI and ligated to similarly digested pGEX-KG, yielding pGEX-importin-β.

Expression of GST-cyclin E required that the reading frame of the BamHI site in pGEX-KG be altered to be compatible with our cyclin E clone (pESP11-cyclin E, a kind gift of Dr. T. Hunt). pGEX-KG was digested with BamHI and a self-complementary oligo (5′-gat ctc cgg atc cgg a-3′) was inserted to generate pGEX-BamHIFS. A plasmid that encodes a fusion protein consisting of GST linked to Xenopus cyclin E (pGEX-cyclin E) was constructed by replacing the multiple cloning site sequences between the BamHI and HindIII sites of pGEX-BamHIFS with a BamHI-HindIII fragment from pEsp11XlE. pGEX-cyclin EΔCT and pEsp11- cyclin EΔCT, plasmids encoding truncation mutants of cyclin E that are incapable of binding to Cdk2 (residues 1–338), were made by replacing the BamHI-HindIII fragment containing the full-length cyclin E coding sequences with a fragment encoding cyclin E residues 1–338 prepared by BamHI-HindIII digestion of a PCR product prepared using pEsp11XlE as a template and primers 5′-gca aaa gca ggg tga cca aag aca-3′ (a sequence proximal to the start of the cyclin E coding sequence in the template) and 5′-ttt tcc aca agc tta tat ggc cat tgc aaa agg-3′.

To express a GST-cyclin B1 fusion protein, sequences encoding full-length Xenopus cyclin B1 were amplified by means of PCR (primers 5′-ggc aca cca tgg cgc tac gag tca cca gaa aca tg-3′ and 5′-ggc cgg aag ctt cac atg agt ggg cgg gcc at-3′, template pGEM-cyclin B1), cut with NcoI and HindIII, and inserted into similarly digested pGEX-KG to produce pGEX-cyclin B1 (the second residue in cyclin B1 was changed from a Ser to an Ala). A primer introducing BamHI and NcoI sites and a start codon at residue 121 of cyclin B1 (5′-cgg atc gga tcc atg gtt gat gct gat gat gat ggc aac-3′), along with primers introducing SalI and XhoI sites distal to the native stop codons and one introduced at residue 373 (5′-gca tcg gcc tcg agt cga ctc aca tga gtg ggc ggg cca ttt cca-3′ and 5′-gca tcg gcc tcg agt cga ctc agc tgc tag cat act tgt tct taa cag tca-3′, respectively), were used in PCR reactions with the same template to generate truncation mutants of cyclin B1 encoding residues 121–373 and 121–397 which could be inserted into both BamHI-SalI–digested pGEX-KG to produce GST-cyclin B1 fusion proteins and into an NcoI-XhoI–digested pSP64 variant for in vitro translation of the unfused cyclin.

To express nucleoplasmin (residues 1–200), as well as the nucleoplasmin core region that lacks the bipartite NLS (residues 1–149), we used pET16b-NEB, designed to express Xenopus nucleoplasmin with an NH₂-terminal His tag (a gift of Dr. D. Kalininch). To make pET16b-NPcore, a 550-bp fragment encoding residues 1–149 of Xenopus nucleoplasmin was amplified by means of PCR using pET16b-NEB (Kalinich et al., 1994) as a template and primers 5′-gaa ggt cgt cat atg gct aga att-3′ and 5′-gga ctt gga tcc tag gta ccc ccg ggc tag tcg aca gat tct gga tct tct tcc tct tcc tcc tcc tc-3′. The PCR product was digested with NdeI and BamHI and ligated to the large vector fragment derived from NdeI-BamHI digestion of pET16b-NEB. We also generated a plasmid, pET16b-NPcore/IBB, directing the expression of a His-tagged fusion protein consisting of the nucleoplasmin core linked to residues 10–55 of Xenopus importin-α (the IBB domain). A PCR reaction was performed with primers 5′-cca gaa tct gtc gac atg ccg acc aca aat gaa gca gat-3′ and 5′-gga ctt gga tcc tac tct tct ggt aag cat aca tt-3′ and pQE70 importin-α as template. The PCR product was digested with SalI and BamHI and ligated to the vector fragment of similarly digested pET16b-NPcore.

The plasmids encoding GST-Ran and GST-RanG19V fusion proteins (pGEX-Ran and pGEX-RanG19V) were described by Kornbluth et al. (1994). pMBP-Trn which encodes an MBP-human transportin fusion protein and pGEX-4T-M9 which encodes a fusion protein linking the 38– amino acid M9 NLS of human hnRNPA1 to the COOH terminus of GST were described by Fridell et al. (1997). The plasmids encoding the fusion proteins linking the NLS of SV-40 T antigen to the COOH terminus of GST (and the corresponding reversed NLS mutant) were a kind gift of Drs. M. Rexach and G. Blobel.

Recombinant GST-importin-α and GST-importin-β proteins were produced in Escherichia coli strain BL21 by induction with 1 mM IPTG for 4–5 h at 25–30°C. Unfused GST and fusion proteins GST-IBB, GST-T antigen-NLS, GST-TNLS(Rev), GST-M9, and GST-Ran were similarly produced, except that inductions were performed at 37°C. GST-Xenopus cyclin fusion proteins were also produced in BL21 cells (induced with 0.6 mM IPTG at 18°C for 16–18 h). His-tagged nucleoplasmin derivatives were produced by inducing BL21(de3)pLysS cells with 2 mM IPTG at 37°C for 3 h. In each case, cells were grown at 37°C to an A600 of 0.5–0.9 and then cooled if necessary before addition of IPTG. GST proteins were purified by the method of Solomon et al. (1990) from cells lysed by lysozyme and detergent treatment. Cells containing MBP and nucleoplasmin proteins were lysed by sonication (after lysozyme treatment) and the proteins purified using the fusion tags as described (Riggs, 1990; Kalinich and McClain, 1994). Where appropriate, cleavage of GST-fusion proteins with thrombin (Pharmacia) was performed according to the manufacturer's instructions. This was required for the use of Ran, NTF2, importin-α, and importin-β in import assays.

Sf9 cells were grown at 27°C in spinner flasks in Grace's insect cell medium. They were infected with a multiplicity of infection of 50 of supernatant from a lytic viral culture and grown for a further 40–48 h at 27°C to allow the induction of recombinant human HA-tagged Cdc2 and Cdk2, and His-tagged cyclins B1 and E (all these viruses were a kind gift of Dr. D.O. Morgan). Cells were harvested and stored at −80°C. Extracts, cyclins, and Cdk/cyclin complexes were prepared as described by Kumagai and Dunphy (1995).

For nuclear import assays, the GST-Xenopus cyclin fusion proteins, as well as GST, GST-NLS, and GST-IBB as controls, were dialyzed into PBS (without any sulfhydryl reducing agents) then mixed with a 40:1 molar ratio of fluorescein maleimide (dissolved in DMSO as a 20 mM stock). The mixtures were incubated at room temperature in the dark for 2 h to allow for the formation of fluorescein conjugates, unbound reactive dye was then quenched with 50 mM β-mercaptoethanol and removed by dialysis against a buffer containing 20 mM Hepes-KOH, 100 mM KCl, 200 mM sucrose, 1 mM DTT, and a protease inhibitor cocktail, pH 7.5. The conjugates were flash frozen in liquid nitrogen and stored at −80°C. The GST-M9 fusion protein was labeled with FLUOS as described by Fridell et al. (1997).

For oocyte transport assays, reticulocyte lysates containing ³⁵S-labeled in vitro translated Xenopus cyclins E, A1, full-length B1, or a truncated B1 mutant (residues 121–397) were mixed with ¹⁴C-labeled BSA. Stage VI oocytes were then injected with 40 nl of this radiolabeled protein mix and then manually dissected into cytoplasmic and germinal vesicle (nuclear) fractions immediately after injection and after 4 or 9 h of incubation at 18°C. Oocytes which had been injected into the nucleus (as judged by the presence of the hemoglobin pigment in this fraction) were discarded. The cytoplasmic and nuclear fractions from five cytoplasmically injected oocytes were then separately pooled and separated on 10% SDS-polyacrylamide gels, along with a total lysate derived from five cytoplasmically injected oocytes. The gels were soaked in 0.5 M sodium salicylate, dried, and then the radiolabeled proteins were detected by fluorography.

For the digitonin semipermeabilized mammalian cell assay (Adam et al., 1990; Görlich et al., 1994), suspension grown HeLa cells (1 liter, 5–7 × 10⁵ cells/ml) were harvested by centrifugation at 200 g, washed once with 200 ml of PBS, then resuspended in 50 ml of cell dissociation solution (nonenzymatic; Sigma Chemical Co.). After 5 min at 37°C, cells were pelleted at 200 g and resuspended in a final volume of 12 ml of PBS + 10% DMSO. Cells were frozen slowly at −70°C in 0.5-ml aliquots and stored in liquid nitrogen (they were usable for import assays for at least 6 mo). For an import experiment, an aliquot was thawed rapidly at 37°C, washed twice with 10 ml PBS, then incubated in 2.5 ml of cell dissociation solution at 37°C for 5 min. Cells were then pelleted at 200 g, resuspended in 1 ml of transport buffer (20 mM Hepes-KOH, 100 mM potassium acetate, 0.5 mM magnesium acetate, 0.25 mM EGTA, and 2 mM DTT, pH 7.4; Adam et al., 1990) then mixed with 1 ml of transport buffer containing 100 μg/ml digitonin (Calbiochem) to give a final concentration of 50 μg/ml. After 8 min on ice, 10 ml of ice-cold transport buffer was added to stop permeabilization and the cells were pelleted at 200 g. They were then washed twice with 10 ml of transport buffer, in each case allowing 10 min on ice before repelleting to allow endogenous transport factors to diffuse out of the cells. The permeabilized cells were then resuspended in 5 ml of PBS, layered on top of a 3-ml sucrose cushion (50% wt/vol) in PBS and pelleted at 500 g for 5 min. After a final wash in 10 ml transport buffer, the permeabilized cells (∼3 × 10⁶) were resuspended in 0.3 ml transport buffer. Nuclear import assays contained 10⁵ nuclei, an ATP regenerating system consisting of 10 mM creatine phosphate, 1 mM ATP, 500 μM GTP, and 50 μg/ml creatine kinase, 30–100 μg/ml fluorescent import substrate, and, in complete assays, 4 mg/ml egg cytosol or reticulocyte lysate and 6 mg/ml BSA. For assays using recombinant factors, the cytosol was omitted and the BSA concentration increased to 10 mg/ml BSA. Recombinant import factors were added at a final concentration of 100 μg/ml. In some experiments we added wheat germ agglutinin at 0.5 mg/ml or replaced the energy regeneration system with an ATP depletion mix consisting of 1 U/μl hexokinase and 50 mM glucose to block energy-dependent nuclear transport pathways using the nuclear pores. In other experiments we added 350 μg/ml (50 μM) thrombin-cleaved IBB domain (importin-α residues 1–64), 50 μM of a 20–amino acid nucleoplasmin-NLS peptide (CKRPAATKKAGAAKKKKLDK), or 250 μg/ml thrombin-cleaved M9 NLS (50 μM) to competitively inhibit import through the importin and transportin pathways. Assays were started by the addition of nuclei and carried out in the dark at room temperature for 30 min: certain negative control reactions were instead incubated on ice. The nuclei were then fixed by mixing 1:1 with a formaldehyde/sucrose-based fixative containing Hoechst 33258 and examined by fluorescence microscopy.

The enormous size of the Xenopus oocyte has made it one of the systems of choice for nuclear transport studies. Microinjection of radiolabeled transport cargo into either the nuclear or cytoplasmic compartment, followed by manual oocyte dissection, allows rapid quantitation of nuclear import and export rates. As a starting point for examining the pathways responsible for nuclear import of Cdk/cyclin complexes, we followed the fate of radiolabeled cyclins injected into the cytoplasm of stage VI oocytes. Since Cdks are present in excess in the oocyte, each injected cyclin should bind to its cognate Cdk. Full-length ³⁵S-labeled, in vitro translated Xenopus cyclins E, A1, and B1 were individually injected into the cytoplasm of stage VI oocytes along with ¹⁴C-labeled BSA as a control. The nuclear and cytoplasmic fractions were then manually separated immediately, 4 or 9 h after injection. As shown in Fig. 1 A, BSA remained in the cytoplasm for at least 9 h after injection, while cyclins E and A1 translocated to the nucleus. Within 4 h, virtually all of the cyclin E and 50% of the cyclin A1 had translocated to the nucleus; the great majority of cyclin A1 was nuclear by 9 h. In contrast, cyclin B1 injected into the cytoplasm remained there for at least 18 h, consistent with its normally interphase cytoplasmic localization. A truncated form of cyclin B1 protein, which eliminates the NES (cycB1 121–397), entered the nucleus, albeit more slowly than cyclins E and A1; somewhat less than 50% of the injected protein had accumulated in the nucleus 9 h after injection. These data suggest that the nuclear import rates of all three cyclins are different, raising the possibility that they are recognized by the import machinery with different affinities or that distinct import pathways are used. Indeed, when we added fluorescein-conjugated, recombinant GST-cyclins B1 and E to Xenopus egg extracts containing ∼1,000 nuclei/μl, fluorescent cyclin E was clearly transported into nuclei more efficiently than cyclin B1, even after removal of the cyclin B1 NES (data not shown). We selected these cyclins, the fastest and slowest for nuclear import, for further study. Interestingly, these cyclins form complexes with different catalytic subunits. Moreover, in combination, Cdk2/cyclin E and Cdc2/cyclin B1 complexes are fully capable of driving multiple rounds of S phase and M phase in Xenopus egg extracts.

In a first step towards analyzing the mechanism of cyclin nuclear import, we labeled bacterially produced cyclins B1 (121–397) and E with fluorescein maleimide and monitored the nuclear import of these fluoresceinated cyclins using interphase egg cytosol as a source of soluble import factors and digitonin-permeabilized HeLa cells as a source of intact nuclei. As for the positive control GST-NLS protein, import of cyclins required incubation at room temperature (Fig. 1 B) and was blocked by the lectin wheat germ agglutinin (data not shown). The negative control GST protein was not detectably imported into nuclei.

To investigate the possible involvement of known soluble transport factors in cyclin import, we preincubated the interphase egg cytosol with competitor peptides designed to prevent importins or transportin from interacting with their cargo. When the interphase egg cytosol used in the digitonin-permeabilized cell import assay was preincubated with peptide encoding the classical NLS from nucleoplasmin (50 μM), fluoresceinated cyclin B (121–397), cyclin E, and control GST-NLS proteins were all prevented from entering nuclei (Fig. 2). These data suggested that cyclins might use the well-known importin-α/β pathway for nuclear import. However, when we used a peptide encoding the IBB domain of importin-α (importin-α residues 1–64) in similar experiments, we found that import of cyclin E, but not cyclin B1 (121–397), was impaired. A peptide encoding the M9 NLS blocked import of only GST-M9 protein and did not interfere with cyclin import. These data suggest that nuclear import of both cyclins E and B1 proceed by importin-dependent, transportin-independent pathways. However, the fact that GST-NLS and GST-IBB proteins interfered differentially with import of cyclins E and B1 suggested, again, that these cyclins might use different pathways for nuclear import.

Since the basic NLS peptide blocked cyclin import, we wished to investigate whether importin-α was required for nuclear import of either cyclin B1 or cyclin E. For this purpose, nuclear import assays were performed in which the egg cytosol was replaced with cytosol depleted of importin-α (by passage through a column of Ni-NTA-agarose; see Görlich et al., 1994) (Fig. 3 A) or with control cytosol mock-depleted with Sepharose CL-6B. GST-cyclin E import was abolished in Ni²⁺NTA-agarose–treated extracts, whereas cyclin B1 (121–397) import was not severely affected (Fig. 3 B). Readdition of recombinant importin-α to Ni²⁺NTA-agarose–depleted extracts restored GST-cyclin E nuclear import, though not quite to the levels observed in the mock-depleted extract. These data indicate that cyclin E nuclear import proceeds via an importin-α–dependent mechanism, whereas, surprisingly, given its inhibition by NLS peptide, cyclin B1 import does not.

Since the above data indicated that cyclin E nuclear import required importin-α, we wished to determine whether cyclin E could interact physically with the importin-α/β heterodimer used to import classical NLS-containing proteins. For this purpose, resins linked to GST or GST- importin-β were incubated in interphase Xenopus egg cytosol supplemented with in vitro translated, radiolabeled cyclins, pelleted, washed, and resolved by autoradiography. As shown in Fig. 4 A, cyclin E behaved like the NLS-containing control protein, nucleoplasmin (biotinylated for detection with HRP-streptavidin), binding specifically to importin-β. Similar interactions were observed using a nucleoplasmin variant in which the NLS had been removed and replaced with the IBB domain of importin-α (NPcore-IBB), whereas the NPcore alone, or luciferase, both lacking nuclear localization signals, did not bind any transport factors. The interaction of nucleoplasmin with GST-importin-β beads indicates that importin-α binding proteins are capable of interacting (presumably indirectly via importin-α provided by the extract) with immobilized importin-β in this system. Indeed, given the requirement for importin-α in cyclin E nuclear import (Fig. 3 B), we suspected that importin-α from the extract mediated cyclin E–importin-β interactions (a point we will return to below). Since importin-α did not appear to be required for cyclin B1 nuclear import, we were surprised to find that the importin-β resin was also able to retrieve radiolabeled cyclin B1 from egg extracts (Fig. 4 A). These data suggested that cyclin B1 might bind directly to importin-β.

To determine whether the interaction between cyclins and importin-β was direct, we produced recombinant His-tagged variants of cyclins E and B1. The purity of these preparations is shown in Fig. 4 B; any contaminating bound factors were likely present at very low stoichiometries. We found that GST-importin-β bound directly to His-tagged cyclin B1 incubated in the absence of cytosol (Fig. 4 C); cyclin B1 actually interacted less well with importin-β in the presence of excess importin-α (data not shown). In contrast, binding of purified cyclin E to importin-β was vastly increased in assays supplemented with recombinant importin-α (Fig. 4 D). Addition of recombinant Cdc2 or Cdk2 did not enhance the importin-β binding of their cognate cyclins (Fig. 4 C and data not shown). These data show that cyclins E and B1 interact with the importins in distinct manners and demonstrate that cyclin-importin interactions are direct; no piggyback from a cross-linking factor is required. They also suggest that Cdks are not required for cyclin–importin interactions.

It remained possible that Cdks could also interact independently with importins. To determine if this was the case, baculovirus-produced HA-tagged Cdk2 or Cdc2 proteins were incubated with importin-β beads (in the presence of importin-α) in the absence of cytosol. Neither Cdk2 nor Cdc2 interacted significantly with importins under these conditions, but addition of the cognate cyclins dramatically increased these interactions (Fig. 4 E).

The experiments above suggested that nuclear import of Cdk/cyclin complexes might be promoted by direct interaction of the cyclin subunits with importins. However, previously published studies have indicated that the ability of mutant cyclins A and B3 to be imported into nuclei correlated with their ability to interact with Cdks (Maridor et al., 1993; Gallant and Nigg, 1994). Thus, it remained possible that Cdks might be in some way required for effective cyclin nuclear import. To determine if this was the case, we produced a recombinant mutant of cyclin B1 previously reported to lack the ability to interact with Cdc2 (cyclin B1 121–373; Stewart et al., 1994) and an analogous mutant of cyclin E unable to bind Cdk2 (cyclin E 1–338; Fig. 5 A). Both of these cyclins bound to importins as efficiently as their wild-type counterparts (data not shown). Moreover, fluoresceinated variants of these proteins were imported efficiently into nuclei of digitonin-permeabilized cells (Fig. 5 B; for the purposes of comparison, note that the control cyclin B1 and cyclin E import assays from this experiment are shown with the same exposure time in Fig. 1 B). These data suggest that Cdk binding is entirely dispensable for cyclin nuclear import, in accordance with our observation that the major determinants of importin binding reside within the cyclin subunit of the cyclin/Cdk complex. As predicted by these data, we have found that depletion of all of the Cdk2 and >90% of the Cdc2 from egg cytosol using p13 beads had no deleterious effect on nuclear import of either cyclin E or cyclin B1 (data not shown).

Given the above data demonstrating direct interactions between cyclin B1 and importin-β, it was unclear why, in our initial experiments (Fig. 2), the NLS (presumably complexed to importin-α from the egg cytosol), but not the IBB peptide, could interfere with cyclin B1 nuclear import. Indeed, we also found that addition of the NLS peptide to egg cytosol impaired retrieval of added radiolabeled cyclin B1 on an importin-β resin (data not shown). One interpretation of these data is that cyclin B1 binds to a site on importin-β distinct from that used for binding the IBB domain of importin-α. If so, two possible models could account for the fact that NLS, but not IBB peptide, inhibited productive cyclin–importin-β interactions: the small IBB peptide and cyclin B1 are able to bind simultaneously to importin-β, but the larger 60-kD NLS/importin-α complex, binding to the same site as the IBB peptide, sterically prevents simultaneous binding of cyclin B1; or, the importin-α and cyclin B1 binding sites on importin-β cannot be occupied simultaneously. Under these circumstances, the relative affinities of the NLS/importin-α, cyclin B1, and IBB proteins for importin-β would dictate which of these proteins would bind preferentially. Since the binding of an NLS to importin-α stimulates the importin-α/β interaction (Rexach and Blobel, 1995), it is possible that the intact importin-α molecule bound to the NLS peptide would compete more effectively with cyclin B1 for importin-β binding than would the isolated IBB domain.

The most direct way to test the hypothesis that cyclin B1 and importin-α bound to different sites on importin-β was to identify an importin-β mutant that could bind to one protein, but not the other. Kutay et al. (1997) have shown that the IBB domain of importin-α requires residues 286– 876 of importin-β for binding, while Ran-GTP and certain nucleoporins bind to the NH₂-terminal half (residues 1–462) of the importin-β protein. As shown in Fig. 6 A, cyclin B1 bound to both full-length and 1–462 importin-β. Consistent with a requirement for importin-α in the cyclin E–importin-β interaction, the 1–462 mutant could bind neither importin-α from the egg cytosol nor cyclin E (Fig. 6 A). These data confirm that cyclin B1 binds to a different site on importin-β from that used by the IBB domain of importin-α. Very recently, Jäkel and Görlich (1998) identified an NLS in ribosomal protein L23 that binds in a Ran-GTP–sensitive manner to the first 462 amino acids of importin-β, indicating that such interactions can be relevant to nuclear transport processes.

To determine whether the IBB domain and cyclin B1 could interact with importin-β simultaneously, we immobilized the recombinant GST-IBB domain on beads, incubated these beads in egg cytosol supplemented with ³⁵S-labeled cyclin B1 (121–397), and asked whether cyclin B1 could bind to the IBB domain resin. Indeed, as shown in Fig. 6 B, we found that cyclin B1 was able to bind to the GST-IBB beads, presumably via the importin-β that was also bound from the extract. As expected, the GST-IBB beads did not bind importin-α, nor did they bind radiolabeled cyclin E. Identical results were obtained using a variant of the IBB domain lacking a potential Cdc2/cyclin B phosphorylation site at residues 60–64 (GST-IBB55), arguing against the possibility that cyclin B1–IBB interactions were mediated via enzyme–substrate interactions of the cyclin B1 binding partner, Cdc2 (Fig. 6 B). These data suggest that cyclin B1 and an IBB peptide can bind simultaneously to importin-β. Interestingly, cyclin B1 (121–397) did not bind to the importin-α/importin-β complexes that associated with the immobilized GST-NLS resin. When a basic NLS–importin-α complex is bound to importin-β, steric hindrance may make the additional binding site unavailable for cyclin B1. These binding assays thus provide a plausible explanation for the ability of the NP-NLS to block cyclin B1 nuclear import.

The binding assay data, peptide competitions, and permeabilized cell assays lead to the conclusion that cyclin B1 is imported into nuclei via direct interactions with importin-β and that cyclin E uses the importin-α/β heterodimer. Moreover, these import reactions are not dependent upon Cdks. As a final, definitive test of these conclusions, we performed nuclear transport assays using digitonin-permeabilized cells and entirely recombinant import factors. All assays contained recombinant Ran and its accessory factor, NTF2. Fluoresceinated GST-cyclin E, like GST fused to the NLS of SV-40 T antigen, was imported into nuclei only in the presence of both recombinant importin-α and -β, while cyclin B1 (121–397) required only importin-β (Fig. 7). The full lysate (here reticulocyte lysate) containing all factors imported both cyclins; recombinant transportin could support nuclear import of only the GST-M9 protein. These results establish that cyclin B1 and cyclin E use distinct importin-mediated pathways for nuclear import.

Several recent results have highlighted the importance of the nucleocytoplasmic distribution of Cdk/cyclin complexes to cell cycle control. Three recent studies using egg extracts imply that accumulation of Cdk2/cyclin E in the nucleus is required both for initiating DNA replication (Chevalier et al., 1996) and preventing re-replication (Hua et al., 1997; Walter et al., 1998). The ability of Cdc2/cyclin B1 to drive M phase entry in oocytes is dependent on phosphorylation of serine residues on cyclin B1 that promote its nuclear localization (Li et al., 1995, 1997). Moreover, altering the normal subcellular localization of the Cdc2/cyclin B1 complex has been shown to abrogate the DNA damage checkpoint (Jin et al., 1998). Therefore, elucidating the mechanisms used for Cdk/cyclin nucleocytoplasmic trafficking has been an important objective for cell-cycle researchers. Here, we have used a combination of solution binding and nuclear import assays to uncover mechanisms for the nuclear import of Cdc2/cyclin B1 and Cdk2/cyclin E.

At the start of this work, we expected to find that Cdk subunits would provide essential information for the recognition of Cdk/cyclin complexes by the nuclear import machinery. Two studies by the Nigg laboratory had shown that cyclin A and B3 mutants that failed to bind Cdk2 also lacked the nuclear localization of the wild-type proteins (Maridor et al., 1993; Gallant and Nigg, 1994). We also anticipated that Cdk/cyclin complex recognition by the import machinery might not be due to bona fide NLSs on the Cdk/cyclin complexes themselves, but rather, might require a piggyback ride from an interacting factor such as p21^CIP or Cdc25. Indeed, it has been demonstrated recently that elevating p21 expression could promote Cdk4 nuclear localization (Diehl and Sherr, 1997).

Surprisingly, however, our results indicate that the two Cdk/cyclin complexes we examined interact directly with the best understood nuclear import factors, importin-α and -β. In fact, the Cdk subunits are dispensable for cyclin E and B1-importin binding and nuclear import. Moreover, neither Cdk2 nor Cdc2 can interact with components of the importin heterodimer in the absence of these cyclins. Cyclins E and B1 (121–397) thus act as nuclear targeting subunits for their partner kinases. Remarkably, though both cyclins E and B1 are imported via importin-dependent pathways, the molecular details of the interactions are very different. Cyclin E behaves as if it contains a classical basic NLS, using the α adaptor subunit to interact with importin-β; in contrast, cyclin B1 binds directly to importin-β, but at a site distinct from that used to bind importin-α. It is possible that other cyclins (e.g., A, B3) use distinct import mechanisms, perhaps ones that are dependent upon Cdks.

If cyclin E contains a basic NLS, as suggested by its behavior in solution binding and import assays, which sequences constitute the NLS, and why has it not been detected before? A cursory examination of the Xenopus cyclin E sequence reveals no lysine-rich sequences reminiscent of either the SV-40 T antigen or nucleoplasmin NLS, though the NH₂-terminal 55 amino acids of cyclin E appear basic in character. Basic regions are also present in the NH₂ termini of A and B type cyclins but these have been found to be involved in cyclin degradation, rather than nuclear import. If both lysines and arginines are considered, then the sequence RSRKRK between residues 24 and 29 of all vertebrate cyclin E proteins might be considered a candidate NLS. This sequence is very similar to the basic NLS that controls the nuclear localization of the v-jun oncogene (Chida and Vogt, 1992). However, in a recent alanine scanning mutagenesis of human cyclin E, it was found that conversion of three adjacent basic residues within this sequence to alanine did not perturb the nuclear localization of cyclin E (Kelly et al., 1998). Thus, either that sequence does not serve as an NLS or the determinants of cyclin E nuclear localization are more complex; perhaps multiple redundant sequences or piggyback mechanisms can contribute to cyclin E localization in vivo. We are currently performing experiments to address these issues.

Solution binding assays, peptide competition studies, and permeabilized cell transport assays all support the idea that cyclin B1 is imported into nuclei via a direct interaction with importin-β at a site distinct from that used by the IBB domain of importin-α. Since cyclin B1 can interact with importin-β in the absence of Cdc2, it may be possible to define the sequences on cyclin B1 which comprise the cyclin B1 NLS. Indeed, sequences critical for nuclear export lie entirely within the cyclin B1 component of the Cdc2/cyclin B1 complex and these have been identified precisely (Hagting et al., 1998; Yang et al., 1998). However, it is possible that the cyclin B1 sequences which contact importin-β do not lie within a discrete, contiguous amino acid stretch; if so, it may be difficult to identify the primary sequences specifying cyclin B1 nuclear localization. While IBB-like NLSs that interact with importin-β tend to have some common features (e.g., a concentration of arginine residues), our data suggest that cyclin B1 interacts with importin-β in a manner different from that of the IBB domain. Very recently, Jäkel and Görlich (1998) have identified a 32–amino acid sequence in the ribosomal protein L23 that, like cyclin B1, interacts with a site present in the NH₂-terminal half of importin-β (residues 1–462). This NLS is very basic in character, a feature that is not shared by any part of the cyclin B1 sequence between residues 121 and 373. In addition, the L23 NLS also interacts with transportin, RanBP5, and RanBP7, all of which can stimulate its import into permeabilized cells. Therefore, whether or not cyclin B1 and the L23 NLS bind to the same site on importin-β, the L23 NLS probably uses a broader array of import mechanisms than cyclin B1, since cyclin B1 import can be abolished by perturbations of the importin-dependent pathway.

While importin-β is responsible for transporting both Cdk2/cyclin E complexes and Cdc2/cyclin B1 complexes across the nuclear pore, the cyclin subunits interact with importin-β in distinct manners. Why do different Cdk/cyclin complexes use distinct import pathways? In all likelihood, the different cell cycle roles of these complexes impose different constraints on their nucleocytoplasmic transport. One implication of the recent finding that cytoplasmic retention of cyclin B1 relies on nuclear export (Hagting et al., 1998; Toyoshima et al., 1998; Yang et al., 1998) is that, during interphase, the rate of cyclin B1 nuclear export must exceed its nuclear import rate. Addition of a strong, importin-α–binding NLS to cyclin B1 results in its nuclear localization, apparently overwhelming the capacity of export mechanisms to maintain cyclin B1 in the cytoplasm (Pines and Hunter, 1994; Li et al., 1997; Jin et al., 1998). At the time of entry into mitosis, the cyclin B1 export rate is decreased and cyclin B1 accumulates in nuclei; the rate of cyclin B1 nuclear import must not be so slow that this relocalization does not occur (Yang et al., 1998). These biological imperatives predict that the import rate of cyclin B1 will be moderate, a result in accordance with our finding that cyclin E import is considerably faster than that of cyclin B1 (approximately fivefold faster based on quantitation of data in Fig. 1 A). Indeed, altering cyclin B1 subcellular localization interferes with proper DNA damage–induced checkpoint regulation, suggesting that localizing Cdc2/cyclin B1 to the cytoplasm in interphase helps to prevent premature mitosis. Thus, a weak NLS (which competes poorly with importin-α-NLS for importin-β binding) may be necessary to maintain moderate rates of cyclin B1 import and proper cell cycle regulation.

We see no compelling reason why a weak NLS would need to bind directly to importin-β, rather than to the importin-α subunit. However, it is possible that there are cooperative binding reactions between importin-α, -β, and transport substrates which make these complexes form more readily than dimeric importin-β–transport substrate complexes.

The importin pathways may not be solely responsible for nuclear import of Cdk2/cyclin E and Cdc2/cyclin B1 under all circumstances. Interestingly, in the course of performing our binding assays, we were able to detect interactions between transportin and full-length cyclin B1 (data not shown). Because the cyclin B1 121–397 mutant used in most of our assays did not bind to transportin (data not shown) or enter nuclei in the presence of recombinant transportin (Fig. 7), we do not think transportin interactions are critical for cyclin B1 import. However, since transportin is a shuttling factor, it could conceivably contribute to cyclin B1 nuclear export. Indeed, we did detect some residual cyclin B1 export after inhibition of CRM1-dependent export with leptomycin B (Yang et al., 1998). With regard to alternative transport pathways for cyclins, we note that elevation of p21^Cip levels within cells can promote the nuclear localization of not only Cdk4/cyclin D complexes (Diehl and Sherr, 1997; LaBaer et al., 1997), but also Cdc2/cyclin B complexes (Winters et al., 1998). This suggests that piggyback mechanisms may play some role in Cdk/cyclin nuclear import in vivo. We speculate that the precipitous accumulation of cyclin B1 in the nucleus of vertebrate cells just before mitosis might result from both a downregulation of nuclear export (Yang et al., 1998) and an upregulation in the rate of nuclear import; perhaps a piggyback mechanism involving binding to a basic NLS–containing protein (such as the Cdc25 phosphatase) might provide a boost in the slow importin-β binding mechanism described in this paper. The physiological importance of these mechanisms remains to be determined and will likely require the identification and mutation of cyclin NLSs.

We thank Dr. B. Cullen for his generous help. We thank E. Bardes for figure preparation and D. Lew for critical comments on the manuscript. We would like to thank Dr. T. Hunt for pESP-Xlcyclin E and Dr. D.J. Donoghue for the Xenopus cyclin B1–containing vectors used for in vitro translation. Recombinant baculoviruses encoding human cyclins and Cdks were generous gifts from Dr. D.O. Morgan. Clones of import factors and antiimportin sera were kindly provided by Dr. D. Görlich. Drs. J. Kalinich and D. McClain provided the original nucleoplasmin expression vector and Dr. C. Feldherr antinucleoplasmin antibodies. Finally, Drs. R. Hartley and J. Maller provided us with anti–cyclin B1 antisera. Dr. J. Pines is to be thanked for sharing unpublished observations.

This work was funded by a grant from the American Cancer Society and a postdoctoral fellowship awarded to J.D. Moore by the Leukemia Research Foundation. S. Kornbluth is a Scholar of the Leukemia Society of America.

Cdk

cyclin-dependent kinase

IBB

importin-β binding

NES

nuclear export sequence

NLS

nuclear localization sequence