C-terminal Sequences Direct Cyclin D1-CRM1 Binding*

GSK-3β-dependent phosphorylation of cyclin D1 at a conserved C-terminal residue, Thr-286, promotes CRM1-dependent cyclin D1 nuclear export. Herein, we have identified a short stretch of residues adjacent to Thr-286 that mediates CRM1 association and thus cyclin D1 nuclear export. We found that disruption of this hydrophobic patch, stretching from amino acids 290 to 295 within cyclin D1, results in constitutively nuclear cyclin D1-CDK4 complexes with an increased propensity to potentiate transformation of murine fibroblasts. Our data support a model wherein deregulation of cyclin D1 nuclear export might contribute to human neoplastic growth.

by itself sufficient to induce a transformed cellular phenotype (17,18). In contrast, we found previously that neither the cyclin D1 mutant, cyclin D1-T286A, nor an alternatively spliced cyclin D1 variant lacking the fifth exon can be phosphorylated by GSK-3␤, and thus both are refractory to phosphorylation-dependent nuclear export and are capable of driving transformation of murine fibroblasts in the absence of a collaborating oncogene (17,20). This suggests that deregulation of cyclin D1 nuclear export results in increased cyclin D1 oncogenic capacity.
Herein we describe the identification of residues within cyclin D1 that direct CRM1 association. Our results demonstrate that hydrophobic residues adjacent to Thr-286 and within the C-terminal nine amino acids of cyclin D1 mediate CRM1 binding.
Expression and Purification of Proteins-FLAG-D1 mutants were cloned into the pVDL-1393 baculoviral expression vector (Pharmingen), and virus was isolated according to established procedures as described previously (15). After infection of insect Sf9 cells at high multiplicity, cells were lysed in EBC buffer and clarified by sedimentation in a microcentrifuge for 10 min. Recombinant proteins were then subjected to precipitation with epitope-specific antibodies and detected by immunoblot.
Protein Turnover Analysis-Equivalent numbers of NIH-3T3 cells overexpressing either FLAG-tagged wild type or mutant cyclin D1 were seeded in 10-cm dishes. The following day, cells were treated with the protein synthesis inhibitor cycloheximide (100 g/ml, Sigma), incubated for the indicated times, and harvested in SDS-sample buffer. Cyclin D1 was detected by Western blot analysis.

RESULTS
Cyclin D1 Residues 290 -295 Mediate CRM1 Binding-CRM1 directs nuclear export of target proteins via direct binding to leucine/hydrophobic stretches of amino acids referred to as a nuclear export signal (23). Because phosphorylation of Thr-286 is required for CRM1 association with cyclin D1, we reasoned that phosphorylation of this residue might result in a conformational change that allows CRM1 access to a region of cyclin D1 adjacent to this site of phosphorylation. A stretch of 10 highly conserved amino acids exists within the C terminus of cyclins D1 and D2 ( 286 TPT-DVRDVD(I/L) in cyclin D1) (24) with the hydrophobic residues (in boldface) being conserved in cyclin D3 as well (Fig. 1A). Since both the spacing of these residues closely resembles that of other known CRM1 substrates such as cyclin B1 (25,26) and a high degree of homology within this region is retained among all three D-type cyclins, we sought to evaluate the potential of these residues to mediate the cyclin D1-CRM1 association. We used site-directed mutagenesis to engineer a mutant D1-V290/295A, wherein the two valines and the isoleucine residue within residues 290 -295 of cyclin D1 (VRDVDI) were mutated to alanine (ARDADA). Alanine substitutions have previously been shown to abrogate CRM1 association (27,28).
We first determined whether alanine substitutions abrogated cyclin D1-CRM1 binding. NIH-3T3 cells were transfected with HA-tagged CRM1 along with either FLAG-D1 as a positive con-trol, FLAG-D1-T286A as a negative control, or FLAG-D1-V290/ 295A. Lysates prepared from these cells were analyzed by immunoblotting or subjected to precipitation with the 12CA5 monoclonal antibody, which recognizes the HA epitope, followed by immunoblotting with antibodies directed toward either the HA epitope or cyclin D1. Although wild type cyclin D1 co-precipitated with CRM1, consistent with previous analysis (17,29), neither D1-T286A nor D1-V290/295A was detected in the CRM1 precipitates (Fig. 1B). The inability of D1-V290/295A to associate with CRM1 was not due to reduced expression of this protein (lanes 4 -6), suggesting that alanine substitutions within this region disrupt CRM1-D1 interaction.
Overexpression of CRM1 can drive wild type cyclin D1 into the cytoplasm, whereas the phosphorylation-deficient D1-T286A mutant is refractory to the enforced nuclear export driven by CRM1 overexpression (17). It stands to reason that cyclin D1 mutants deficient in CRM1 binding, such as D1-V290/295A, should be refractory to CRM1-directed nuclear export and thus constitutively nuclear. To address this possibility, we examined the subcellular localization of D1-V290/295A in asynchronously proliferating cells. The localization of cyclin D1 varied, with ϳ50% of the cells exhibiting primarily nuclear localization, whereas in the remaining cells it localized either to the cytoplasm or to both the nucleus and the cytoplasm (Fig.  1, C and D). This distribution of nuclear cyclin D1 approximates the percentage of cells in G 1 phase in an asynchronous population (15). In contrast, D1-V290/295A was nuclear in more than 90% of the cells (Fig. 1, C and D). To assess the capacity of CRM1 to regulate localization, we determined whether ectopic CRM1 could drive wild type cyclin D1 or D1-V290/295A into the cytoplasm. Whereas expression of CRM1 shuttled wild type cyclin D1 out of the nucleus (Fig. 1, C and D (quantitation)), D1-V290/295A remained nuclear in the presence or absence of ectopic CRM1 (1, C and D (quantitation)).
FIG. 1. Cyclin D1 residues 290 -295 mediate association with CRM1. A, schematic of spatially conserved residues within the C terminus of D-type cyclins. Included is a proposed consensus motif (x ϭ nonconserved amino acid) that directs both CRM1 binding. B, abrogated CRM1 binding in the absence of either Thr-286 or an intact hydrophobic patch (V290/295A). NIH-3T3 cells were co-transfected with HA-CRM1 and either wild type cyclin D1, D1-T286A, or D1-V290/295A. Following immunoprecipitation of CRM1 using the 12CA5 antibody, proteins were resolved on a denaturing polyacrylamide gel, transferred onto a nitrocellulose membrane, and visualized by immunoblot. Whole cell extracts (WCE, right panel) show equal expression of co-transfected proteins. IP, immunoprecipitate. C, constitutive nuclear localization of phosphorylated D1-V290/ 295A. The effect of CRM1 overexpression in NIH-3T3 cells stably expressing either wild type D1 or D1-V290/295A was determined by immunofluorescence. Following transient transfection of either control vector (left panels) or CRM1 cDNA (right panels), the respective cell lines were fixed, and cyclin D1 localization was detected using monoclonal cyclin D1 antibody and fluorescein isothiocyanate-conjugated secondary antibody. D, quantitation of cyclin D1 immunofluorescence shown in C.
Although these results suggest that this region of cyclin D1 mediates CRM1 binding, its close proximity to Thr-286 could disrupt GSK-3␤-mediated phosphorylation and thereby indirectly disrupt CRM1 association. To examine the phosphorylation state of D1-V290/295A, we performed immunoblot analysis using an antibody specific for phosphorylated Thr-286. We have previously demonstrated that recognition of cyclin D1 by this antibody is strictly dependent upon phosphorylation of Thr-286 (17). FLAG-D1, FLAG-D1-T286A, or FLAG-D1-V290/ 295A was precipitated from cells with the M2 monoclonal antibody. Although the D1-T286A mutant was not phosphorylated, both wild type cyclin D1 and D1-V290/295A were phosphorylated on Thr-286 ( Fig. 2A). In fact D1-V290/295A phosphorylation was increased relative to wild type cyclin D1, suggesting decreased turnover of this mutant cyclin D1 protein (see Fig. 3). These results are consistent with the notion that hydrophobic residues between amino acids 290 and 295 direct cyclin D1-CRM1 association. Furthermore, these results demonstrate that phosphorylation of cyclin D1 at Thr-286 alone is insufficient to direct CRM1-mediated cyclin D1 nuclear export in the absence of these hydrophobic residues.
We next determined whether these alanine substitutions specifically interfere with CRM1 binding or whether they have a general effect on cyclin D1 folding that would be reflected in its capacity to associate with other known cyclin D1 interacting proteins or, alternatively, in its capacity to support CDK4 catalytic activity. We first assessed the association of D1-V290/ 295A with GSK-3␤. We have previously demonstrated that kinase-defective GSK-3␤ will form stable complexes with cyclin D1 in insect Sf9 cells (15). Insect Sf9 cells were infected with baculoviruses encoding kinase-defective GSK-3␤ along with cyclin D1, D1-T286A, or cyclin D1-V290/295A. GSK-3␤-cyclin D1 complexes were isolated from lysates prepared from these cells by precipitation with a GSK-3␤-specific monoclonal antibody. Immunoblot analysis confirmed the presence of D1-V290/ 295A within GSK-3␤ precipitates (Fig. 2B).
Overexpression of GSK-3␤ will drive wild type cyclin D1 into the cytoplasm (15,17), whereas the phosphorylation-deficient D1-T286A mutant is refractory to this enforced nuclear export. Because D1-V290/295A retained both binding to and phosphorylation by GSK-3␤, it was important to determine whether D1-V290/295A was refractory to GSK-3␤-mediated nuclear export. We examined the subcellular localization of cyclin D1 and D1-V290/295A in asynchronously proliferating cells transfected with GSK-3␤. As shown in Fig. 2C, cyclin D1 was efficiently shuttled out of the nucleus and cytoplasmically localized in the presence of GSK-3␤. In contrast, D1-V290/295A remained nuclear and is thus refractory to GSK-3␤-mediated FIG. 2. Cyclin D1-V290/295A is subject to GSK-3␤-mediated phosphorylation. A, lysates prepared from cells expressing cyclin D1, D1-T286A, or D1-V290/295A were subjected to precipitation with the M2 antibody. Thr-286 phosphorylation was assessed by immunoblot using a phospho-specific antibody (pT286, upper panel), and total cyclin D1 was assessed using a monoclonal antibody that recognizes both phosphorylated and unphosphorylated protein (lower panel). IP, immunoprecipitate. B, GSK-3␤ co-precipitates with wild type and mutant cyclin D1 isoforms. Sf9 lysates co-expressing kinase-defective GSK-3␤ along with FLAG-D1, D1-T286A, or D1-V290/295A were subjected to precipitation with the M2 antibody. Cyclin D1 levels were confirmed by immunoblot with a cyclin D1-specific monoclonal antibody (lower panel), and co-precipitating GSK-3␤ was assessed by immunoblot (upper panel). Immunoprecipitation with IgG was used as a control. C, constitutive nuclear localization of phosphorylated D1-V290/295A. The effect of GSK-3␤ overexpression in NIH-3T3 cells stably expressing either wild type D1 (a and b) or D1-V290/295A (c and d) was determined by immunofluorescence. Following transient transfection of GSK-3␤ cDNA, the respective cell lines were fixed, and cyclin D1 localization in the presence of GSK-3␤ was detected using monoclonal cyclin D1 antibody and fluorescein isothiocyanateconjugated secondary antibody. Corresponding Hoescht DNA staining is also shown. D, CRM-1 binding-deficient mutant retains association with p21 Cip1 . Cyclin D1, D1-T286A, and D1-V290/295A were immunoprecipitated from NIH-3T3 cells stably overexpressing each respective FLAGtagged protein. Following precipitation with the M2 antibody, proteins were separated on a denaturing polyacrylamide gel and transferred onto a nitrocellulose membrane. The nitrocellulose membrane was processed for Western blot analysis for detection of cyclin D1 (upper panel) and co-precipitating p21 Cip1 (lower panel).
nuclear export (Fig. 2C, panels c and d). GSK-3␤ expression was confirmed by indirect immunofluorescence in parallel (data not shown).
We subsequently assessed the capacity of D1-V290/295A to associate with and activate CDK4 as well as to retain binding to p21 Cip1 . Similar to wild type cyclin D1 and D1-T286A, cyclin D1-V290/295A assembled with CDK4 and supported CDK4-dependent phosphorylation of Rb (data not shown). Moreover, D1-V290/295A also associated with p21 Cip1 (Fig. 2D), indicating that the alanine substitutions have not significantly perturbed the structural or functional integrity of cyclin D1-V290/ 295A. These data support the hypothesis that residues 290 -295 of cyclin D1 mediate CRM1 association and together with phosphorylated Thr-286 constitute the cyclin D1 nuclear export signal.
Increased Stability of Constitutively Nuclear Cyclin D1 Mutants-Our previous work has suggested that cyclin D1 proteolysis is a cytoplasmic event. This conclusion was based on the increased stability of constitutively nuclear cyclin D1-T286A. However, it was not possible to determine the role of Thr-286 phosphorylation versus nuclear export in mediating cyclin D1 proteolysis. In characterizing D1-V290/295A, we noted that it accumulated as a Thr-286-phosphorylated protein (Fig. 2A,  compare lanes 1 and 3), suggesting that phosphorylation is not sufficient to trigger proteolysis of this nuclear protein. Using cycloheximide to inhibit nascent protein synthesis (30), we investigated the rate of D1-V290/295A degradation versus that of either wild type cyclin D1 or the D1-T286A mutant (Fig. 3A). Our results revealed that both D1-V290/295A and D1-T286A have increased protein stability in comparison with wild type cyclin D1. To confirm reduced Thr-286 proteolysis of phosphorylated D1-V290/295A, we measured decay of phosphorylated isoforms of wild type cyclin D1 and D1-V290/295A. Although the half-life of Thr-286 phosphorylated cyclin D1 was less than 15 min, the half-life of phospho-D1-V290/295A was greatly extended (Fig. 3B).
Moreover, to corroborate the increased stability of D1-V290/ 295A relative to wild type cyclin D1, we assessed ubiquitination of cyclin D1 versus D1-V290/295A. Either wild type or D1-V290/295A was co-transfected with HA-ubiquitin into NIH-3T3 cells. Transfected cells were treated with either vehicle or MG132 to inhibit the 26 S proteasome. Cells were then harvested, and lysates were subjected to immunoblotting with the cyclin D1 monoclonal antibody (Fig. 3C). Wild type cyclin D1 alone exhibited the characteristic poly-ubiquitin laddering (poly-Ub) recognized for ubiquitinated proteins in the presence of MG132 (Fig. 3C, lane 2). No higher molecular weight isoforms of D1-V290/295A were detected either in the presence or absence of MG132. These data demonstrate that disruption of CRM1 association promotes the accumulation of Thr-286 phosphorylated cyclin D1 that is inaccessible to and consequently not degraded via cytoplasmic 26 S proteasomes.
Loss of CRM1 Binding Increases the Transforming Potential of Cyclin D1-Cells engineered to overexpress cyclin D1 display a contracted G 1 interval (18). These cells, however, do not form foci, grow in soft agar, or promote tumor formation in immunocompromised mice (18). As shown previously in our laboratory, a cyclin D1 mutant that cannot be phosphorylated at Thr-286 and is thus constitutively nuclear promotes cellular transformation (17). This result suggests that disruption of cyclin D1 nuclear export is an oncogenic event. We therefore considered the possibility that the D1-V290/295A mutant, which unlike D1-T286A retains phosphorylation at Thr-286, might exhibit an increased potential to drive cell transformation relative to wild type cyclin D1. For these experiments NIH-3T3 cells were co-transfected with vectors encoding either wild type FLAG-tagged cyclin D1 or the specified cyclin D1 mutant and a vector encoding puromycin as a selectable marker. Transfectants selected in puromycin were then pooled to eliminate potential clonal variation when assessing transformation. NIH-3T3 cells (data not shown) or NIH-3T3 derivatives engineered to overexpress either wild type FLAG-tagged or mutant cyclin D1 isoforms were assayed for characteristics of cell transformation: foci formation (refractory to contact inhibition) and growth in soft agar (capable of anchorage-independent growth). Over the course of five independent experiments, all NIH-3T3 cells overexpressing the C-terminal cyclin D1 mutants D1-T286A and D1-V290/295A reproducibly formed numerous foci ( Fig. 4A; data not shown). Consistent with these results, unlike wild type NIH-3T3 (data not shown) and D1-3T3 cells, which were incapable of significant growth in soft agar, NIH-3T3 cell lines overexpressing constitutively nuclear C-terminal cyclin D1 mutants D1-T286A (used as positive control) and D1-V290/295A reproducibly grew in soft agar, forming numerous and robust colonies (Fig. 4, B and C (quantitation)). Our data support the model wherein mutations in the CRM1 binding site and/or loss of Thr-286 phosphorylation in wild type cyclin D1 leads to a constitutively nuclear cyclin D1 with increased capacity to drive neoplastic transformation and, as a result, a loss in G 1 /S homeostasis (Fig. 5). FIG. 3. Increased stability of constitutively nuclear cyclin D1 mutants. A, NIH-3T3 cells stably overexpressing the indicated cyclin D1 proteins were treated with cycloheximide (CHX) for the indicated intervals. Lysates prepared from the respective cells were subjected to Western analysis using the cyclin D1 monoclonal antibody. Results shown are representative of multiple independent experiments. B, wild type cyclin D1 or D1-V290/295A was precipitated from lysates prepared from cells treated with cycloheximide as indicated, and decay of phosphorylated cyclin D1 proteins was assessed by immunoblot with the phospho-T286 antibody. C, HA-ubiquitin co-transfected with either cyclin D1 or D1-V290/295A was immunoprecipitated from cells in the presence or absence of MG132, and ubiquitinated cyclin D1 was detected by Western blot. poly-Ub, poly-ubiquitin laddering.

DISCUSSION
Cyclin D1 localization is driven by the competing processes of nuclear import and export. Although the mechanisms of cyclin D1 nuclear import remain poorly characterized, it is now clear that CRM1-dependent nuclear export drives cyclin D1 complexes into the cytoplasm during S phase. CRM1 directs nuclear export of target proteins via direct binding to leucine/ hydrophobic stretches of amino acids. However, the exact spacing of the leucine/hydrophobic patch is variable making it difficult to identify CRM1 binding sites by scanning the primary sequence of a given putative substrate. Although two stretches of residues closely conformed to previously documented nuclear export signal motifs, residues 87-94 (RFLSLEPL) and residues 290 -295, the 290 -295 sequence was an attractive target given its proximity to the site of GSK-3␤-mediated phosphorylation, Thr-286, and its highly conserved nature among all three D-type cyclins. Mutation of valines and isoleucines to alanines in this region abrogates CRM1 binding and promotes nuclear retention without perturbing phosphorylation of Thr-286. This mutant, D1-V290/ 295A, accumulates in the nucleus as a Thr-286-phosphorylated protein. In contrast, the phosphorylated form of wild type cyclin D1 is highly labile and is detectable in the cytoplasm (9). Taken together, these data identify residues 290 -295 as the CRM1 binding site in cyclin D1.
Our data demonstrate that a hydrophobic patch within the C terminus of cyclin D1 mediates CRM1 binding. Mutations within this region do not perturb GSK-3␤-mediated phosphorylation of cyclin D1 at Thr-286, a modification that increases the propensity for CRM1 recognition and binding. As our results indicate, however, phosphorylation in itself is insufficient for cyclin D1 nuclear export or degradation, which is consistent with cytoplasmic proteolysis of cyclin D1. Thus, in the absence of an intact CRM1 binding site, CRM1 fails to associate with cyclin D1, and the resulting protein is more stabile and constitutively nuclear.
These data suggest that the C terminus of cyclin D1 is a critical modulator of cyclin D1 nuclear export and might therefore represent a potential "hot spot" for the activation of mutations in cyclin D1 that ultimately contribute to cancer genesis. In support of this notion, we have recently identified a splice variant of cyclin D1, referred to as cyclin D1b, which lacks the C-terminal residues that direct GSK-3␤ phosphorylation and CRM1-dependent nuclear export (20). As with D1-V290/295A described herein, cyclin D1b is constitutively nuclear and can drive cell transformation (20,31). Furthermore, we have found that this protein is expressed in a significant fraction of primary cancers, highlighting the significance of maintaining temporal control of cyclin D1 subcellular localization for normal cell growth and proliferation (20). FIG. 4. Constitutively nuclear cyclin D1 mutants transform murine fibroblasts. Cells were plated in 6-well dishes in complete medium containing 5% FCS. Cells were grown for 21 days and stained with Giemsa to visualize foci. B, NIH-3T3 cells engineered to stably express the indicated cyclin D1 proteins were plated in semisolid medium and allowed to proliferate for 21 days. Colonies were visualized by 0.01% neutral red stain. C, quantitation of colonies scored in A.
FIG. 5. Cell transformation results from inhibition of cyclin D1 nuclear export. Wild type cyclin D1 is phosphorylated at Thr-286 (circled "P") in late G 1 phase, allowing for CRM1 recognition and binding that triggers the rapid nuclear export of cyclin D1-CDK4 complexes in S phase (top). Mutation of phosphorylatable Thr-286 or mutations directly within the CRM1 binding site of cyclin D1 results in abrogated CRM1 binding and inhibition of cyclin D1 nuclear export. Expression of constitutively nuclear C-terminal cyclin D1 mutants results in cellular transformation.