Differential Phosphorylation of T-47D Human Breast Cancer Cell Substrates by D1-, D3-, E-, and A-type Cyclin-CDK Complexes*

The cyclin-dependent kinases (CDKs) promote cell cycle transitions in mammalian cells by phosphorylation of key substrates. To characterize substrates of the G1and S phase cyclin-CDK complexes, including cyclin D1-CDK4, cyclin D3-CDK4, cyclin D3-CDK6, cyclin E-CDK2, and cyclin A-CDK2, which are largely undefined, we phosphorylated T-47D breast cancer cell nuclear lysates partially purified by ion-exchange chromatography with purified baculovirus expressed cyclin-CDK complexes. A comparison of the substrates that were phosphorylated by the different cyclin D-CDKs revealed some common as well as specific substrates. Hence, cyclin D1-CDK4 specifically phosphorylated a 38-kDa protein while cyclin D3-CDK4 specifically phosphorylated proteins of 105, 102, and 42 kDa. A 24-kDa protein was phosphorylated by both complexes. Cyclin D3-CDK6 exhibited similar substrate preferences to cyclin D3-CDK4, phosphorylating the 105- and 102-kDa proteins but not the 24-kDa protein. Hence, both the cyclin D1 and D3 as well as CDK4 and CDK6 subunits can confer substrate specificity on the overall cyclin D-CDK complex. Cyclin E-CDK2 and cyclin A-CDK2 phosphorylated a greater number of substrates than the cyclin D-CDKs, ranging in size from 10 kDa to over 200 kDa. Twenty-two substrates were common to both complexes, while six were specific for cyclin A-CDK2 and only one protein of 34 kDa was specific for cyclin E-CDK2. These studies indicate that cyclins E and A modulate the specificity of CDK2 and have demonstrated substrates that may be important for the specific roles of these cyclin-CDKs during G1 and S phase progression. Protein sequencing of one of the cyclin-CDK substrates characterized in this study identified this protein as nucleolin, a previously characterized CDC2 (CDK1) substrate, thus indicating the utility of this approach in identifying cyclin-CDK targets. These results show that both the cyclin and CDK subunits can regulate the substrate specificity of the overall cyclin-CDK complex and have demonstrated numerous substrates of D-, E-, and A-type cyclin-CDK complexes potentially involved in regulating transit through the G1and S phases of the cell cycle.

of the cyclin-dependent kinases (CDKs). 1 Active CDKs consist of a protein kinase subunit, catalytic activity of which is dependent on association with a regulatory cyclin subunit (1). In mammalian cells both the kinase and cyclin subunits consist of families with numerous members, including cyclins A-H and at least eight different kinases (2). The association of different kinase subunits with different cyclin subunits controls progression through different stages of cell division. Therefore, following mitogenic stimulation, the D-type cyclins in association with either CDK4 or CDK6 are the first complexes activated and are important for progression during the G 1 phase of the cell cycle (3). Subsequently, transition through G 1 into S phase is controlled by cyclin E in association with CDK2 (4), followed by cyclin A association with CDK2 controlling S phase progression (5). Cyclin A-CDC2 is active during the G 2 phase of the cell cycle, and finally, transition throughout mitosis is regulated by cyclin B activation of CDC2 (6). Although cyclin binding is required for activation of the kinase subunit of the complex, other means of modulating the activity of cyclin-CDKs also exist. Hence, the activity of the different CDKs can also be controlled by both activating and inhibitory phosphorylation and dephosphorylation of key residues on the CDK subunit (7), as well as the binding of inhibitory proteins (8). Apart from their roles during normal cellular proliferation, emerging evidence indicates that perturbations of genes involved in cell cycle progression contributes to oncogenesis in numerous tissues (9).
Since cyclin-CDKs play pivotal roles in cell proliferation and their aberrant expression or activity can contribute to cellular transformation and tumorigenesis, an important issue relates to the substrate targets of these complexes. Although the activities of the different cyclin-CDK complexes throughout the different phases of the cell cycle has been characterized relatively well, the downstream targets of most of these complexes have not been extensively defined, particularly for the G 1 and S phase cyclin-CDKs. The necessity for different cyclin-CDK complexes during the different stages of the cell cycle is related to the distinct functions required during these stages, such as transcription during the G 1 phase, DNA synthesis during S phase, and chromosome condensation during mitosis. Results to date indicate that differential substrate phosphorylation confers the cell cycle stage-specific functions of the different cyclin-CDKs. For example, both cyclin A-CDC2 and cyclin B-CDC2 phosphorylate histone H1, but only cyclin A-CDC2 phosphorylates the retinoblastoma related protein, p107, indicating that the cyclin subunit plays an important role in substrate specificity (10). To date, a number of substrates have been identified for the mitotic cyclin B-CDC2 complexes, including histone H1, nuclear lamins, caldesmon, vimentin, myosin regulatory light chain, protein phosphatase 1, p60 src , casein kinase II, nucleolin, numatrin, and DNA polymerase ␣ (11)(12)(13)(14).
Unlike the mitotic substrates described for the cyclin B-CDC2 complexes, relatively few substrates for the G 1 and S phase-specific cyclin D-, E-, and A-associated kinases have been described. In mammalian cells, there are three D-type cyclins (D1, D2, and D3) that associate with either CDK4 or CDK6 and are expressed differentially in different cell types (3). The tissue-specific action of the different D type cyclins is exemplified by studies showing that generation of mice lacking cyclin D1 results in a lack of normal mammary gland development in adult female mice as well as retinopathy (15,16), whereas mice lacking cyclin D2 are infertile due to lack of development of ovarian granulosa cells (17). Numerous studies have shown that a major substrate of D-type cyclins, in association with either CDK4 or CDK6, is the retinoblastoma gene product (pRb) (3,18). This protein is a tumor suppressor, which inhibits progression through the G 1 phase of the cell cycle by sequestering transcription factors such as members of the E2F family. Phosphorylation of pRb by cyclin D-CDKs and cyclin E-CDK2 results in release of E2F from pRb, which can then bind to the promoters and thereby activate the transcription of genes required for progression through G 1 into S phase, such as DNA polymerase ␣, thymidylate synthase, proliferating cell nuclear antigen, ribonucleotide reductase, cyclin A, and CDC2 (19). Studies using either a human sarcoma cell line or mouse embryonic fibroblasts deficient for pRb have shown that cyclin D1 is not required for progression through G 1 phase in the absence of functional pRb, indicating that at least in these cells pRb is the major and possibly only target of cyclin D1 (20). However, in myoblasts, cyclin D1 prevents myogenic differentiation as a result of cyclin D1-mediated MyoD-phosphorylation independent of pRb phosphorylation, suggesting that these effects are mediated by phosphorylation of other cyclin D1-CDK substrates (21). Furthermore, a recent study using a yeast two-hybrid screen has identified a Myb-like transcription factor termed DMP1, which is phosphorylated by all three D-type cyclins in association with CDK4 in vitro, suggesting a role for cyclin D-CDK phosphorylation in transcriptional control (22). These emerging studies suggest that other substrates unrelated to pRb are important for cyclin D-CDK action.
Cyclin E-CDK2 also phosphorylates pRb during the G 1 -S phase transition (23). However, unlike cyclin D1, cyclin E is required for G 1 -S phase progression in pRb-deficient cells (24), suggesting that cyclin E-CDK2 phosphorylates other substrates. Evidence for the existence of other cyclin E-CDK2 substrates comes from a recent study, which showed that both cyclin E-CDK2 and cyclin A-CDK2 phosphorylate the transcriptional regulator Id2, thus inhibiting its ability to negatively regulate the binding of basic-helix-loop-helix transcription factors to DNA, which are critical for regulating cell proliferation and differentiation (25). Cyclin E-CDK2 and cyclin A-CDK2 also associate with and phosphorylate a p300 coactivator, thereby modulating the regulation of NF-B transcriptional activation (26). In addition to common substrates of these two cyclin-CDKs, specific substrates for cyclin A-CDK2 but not cyclin E-CDK2 have been demonstrated, including DP-1 and p53. Cyclin A-CDK2 phosphorylation of DP-1, a component of the E2F-1⅐DP-1 transcription complex inhibits the transcriptional activity of this complex (27,28), while phosphorylation of the tumor suppressor protein p53 increases the sequence-specific DNA binding ability of this protein (29). This differential substrate specificity of cyclins A and E in association with CDK2 is attributable to sequences in the cyclin box and C terminus of the cyclin subunits (30).
Several approaches for identifying substrates of protein kinases have been utilized previously. An oriented degenerate peptide library has been used to successfully determine the primary sequence preference for cAMP-dependent kinase as well as cyclin A-CDK2 and cyclin B-CDC2 (31). Other workers have employed the yeast two-hybrid screen to identify protein kinase substrates (22,32). More recently, a novel approach of using a phage cDNA expression library for solid phase phosphorylation of proteins expressed and bound to nitrocellulose filters has been used to identify substrates of the mitogenactivated protein kinase (33). In the present study, we sought to investigate endogenous substrates of the D-, E-, and A-type cyclin-CDK complexes to gain an insight into potential substrates involved in the regulation of G 1 -S phase progression, as well as the substrate preferences of these complexes. We utilized the approach of phosphorylating endogenous proteins from partially purified T-47D breast epithelial cell nuclear lysates with exogenous cyclin-CDKs. This work unveiled numerous in vitro substrates and revealed that substrate specificity can be modulated by both the cyclin and CDK subunit. One of the characterized substrates was identified as nucleolin, a previously described cyclin-CDK substrate (34), thus demonstrating the applicability of this approach in identifying cyclin-CDK substrates and suggesting that many of the substrates demonstrated in this study may have important roles during cell cycle progression.

Construction of Recombinant Cyclin and CDK Baculoviruses-Re-
combinant human cyclin proteins D1, D3, E, and A were expressed as glutathione S-transferase (GST) fusion proteins in Spodoptera frugiperda 9 (Sf9) insect cells. For this purpose, a GST cassette was amplified from pGEX-2T (Pharmacia Biotech Inc.) by polymerase chain reaction (PCR) using 5Ј-GGACTGCAGATGTCCCCTATACTAGGTTA-TTG-3Ј as the forward primer and 5Ј-CCTAGATCTCAGTCATGCACG-ATGAATTCC-3Ј as the reverse primer. The product was digested with PstI and BglII and cloned into the PstI and BamHI sites of the baculovirus transfer vector pVL1392 (PharMingen, San Diego, CA), resulting in the pVL1392-GST transfer vector. Subsequently, all cyclin DNA sequences were inserted into the BamHI and EcoRI cloning sites of this vector in-frame with the GST coding sequence. To create suitable subcloning ends, the different cyclin cDNAs were either first subcloned into pUC18 or pBluescript, or amplified by PCR as follows. Cyclin D1 cDNA was excised from plasmid pD1-H124 by partial NcoI and complete HindIII digestion and subcloned into XbaI-digested pUC18. Cyclin D3 cDNA was excised from plasmid pD3-H347 by NruI and BsaHI digestion and subcloned into SmaI-digested pBluescript II KS ϩ . Cyclin E cDNA was amplified by PCR using 5Ј-CCTGGATCCATGAAGGAGGA-CGGCG-3Ј as forward primer and 5Ј-CCTGAATTCACGCCATTTCCG-GCCC-3Ј as reverse primer. A 2.5-kilobase pair EcoRI fragment containing the cyclin E cDNA excised from pBluescript KS Ϫ served as template. Cyclin A cDNA was amplified by PCR using 5Ј-CCTGGATC-CATGTTGGGCAACTCTGCGC-3Ј as forward primer and 5Ј-CCTGAA-TTCTTACAGATTTAGTGTCTCTGG-3Ј as reverse primer and cyclin A cDNA inserted in pGEM4Z plasmid as template. All cyclins were subsequently digested with BamHI and EcoRI and the respective fragments subcloned into the BamHI/EcoRI sites of pVL1392-GST.
Recombinant baculoviral CDKs were generated with 6 histidine residues in the N terminus. A pVL1392-6xHistidine cassette was first generated into which different CDKs were subcloned to create a pVL1392-6xHistidine-CDK baculoviral transfer vector. The pVL1392-6xHistidine cassette was generated by annealing the oligonucleotides 5Ј-GATGCACCATCACCATCACCATGGATCCCCGGGAATTCTGACT-GACT-3Ј and 3Ј-ACGTCTACGTGGTAGTGGTAGTGGTACCTAGGGG-CCCTTAAGACTGACTGACTAG-5Ј and ligating into PstI-and BamHIdigested pVL1392, resulting in destruction of the BamHI site of pVL1392. This cassette encodes for 6 histidine residues and BamHI and EcoRI cloning sites. Subsequently, all CDK cDNA sequences were ligated in frame into the BamHI and EcoRI sites immediately 3Ј to the 6-histidine coding sequence. CDK2 cDNA was amplified using 5Ј-CCT-GGATCCATGGAGAACTTCCAAAAGGT-3Ј as forward primer and 5Ј-CCTGAATTCAGAGTCGAAGATGGGGTA-3Ј as reverse primer, CDK4 cDNA was amplified using 5Ј-CCTGGATCCATGGCTACCTCTCGATA-TGAG-3Ј as forward primer and 5Ј-CCTGAATTCACTCCGGATTACC-TTCATC-3Ј as reverse primer, and CDK6 cDNA was amplified using 5Ј-CCTGGATCCATGGAGAAGGACGGCCTG-3Ј as forward primer and 5Ј-CCTGAATTCAGGCTGTATTCAGCTCCG-3Ј as reverse primer. The plasmids containing CDK open reading frames used as templates for amplification were pBluescript (CDK2 and CDK6) and pcD-PSK-J3 (CDK4). Following amplification by PCR, all CDKs were digested with BamHI and EcoRI, gel-purified, and subcloned into the pVL1392-6xHistidine vector. The identity of all final constructs was confirmed by dideoxy chain termination sequencing. The recombinant transfer vectors (2 g) were mixed with 0.5 g of Baculogold viral DNA (PharMingen) and co-transfected into Sf9 cells according to the manufacturer's instructions. After incubation for 5 days at 27°C, recombinant viruses were isolated and four single clones obtained by plaque assay. Each clone was used to infect Sf9 cells to express recombinant proteins, and the clone expressing the greatest level of protein as judged by immunoblotting was amplified and used for subsequent work.
Expression and Purification of Catalytically Active Cyclin-CDK Complexes-Sf9 cells were co-infected with the recombinant GST-cyclin and 6xHis-CDK recombinant baculoviruses and incubated at 27°C for 3 days. The cells were then harvested by centrifugation, resuspended in lysis buffer (50 mM HEPES, pH 7.5, 1 mM dithiothreitol (DTT), 0.1 mM vanadate, 5 mM sodium fluoride, 10 mM ␤-glycerophosphate, 500 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% Tween 20, 10 g/ml leupeptin, 10 g/ml aprotinin, 10 g/ml pepstatin, and 1 mM phenylmethylsulfonyl fluoride (PMSF)), and frozen in liquid nitrogen. The cells were freezethawed twice and then sonicated (two 10-s bursts). The lysate was centrifuged at 20,000 ϫ g for 20 min at 4°C. The supernatant was incubated with glutathione-agarose (Sigma) equilibrated in wash buffer (50 mM HEPES, pH 7.5, 1 mM DTT, 1 mM EDTA, 0.01% Tween 20) at 4°C for 1 h with gentle rotation. The glutathione-agarose beads were pelleted by centrifugation. The supernatant was then incubated with fresh glutathione agarose two more times for 1 h at 4°C. The glutathione-agarose beads from the three separate incubations were pooled and washed five times with ice-cold wash buffer. The bound GST-cyclin-CDK complexes were eluted by adding wash buffer containing 15 mM reduced glutathione, pH 7.5, followed by centrifugation to pellet the glutathione-agarose beads. After five elution steps, the eluates were pooled, concentrated to approximately 500 l and equilibrated in wash buffer using Centricon-30 (Amicon). The final concentrated cyclin-CDK enzymes were then stored in aliquots at Ϫ80°C.
Cell Culture and Preparation of T-47D Nuclear Lysates-T-47D breast cancer cells were maintained in RPMI 1640 medium containing 10 g/ml human insulin and 10% fetal calf serum as described previously (35). For preparation of nuclear lysates, 20 150-cm 2 tissue culture flasks of T-47D cells were grown to approximately 70% confluence, washed once with phosphate-buffered saline, removed from the flask using a cell scraper, and harvested by centrifugation. The cells were resuspended in 5 ϫ packed cell volume of hypotonic buffer (10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT, 0.2 mM PMSF, 10 g/ml leupeptin, 10 g/ml pepstatin, and 10 g/ml aprotinin) and harvested again by centrifugation. The cells were then resuspended in 3 ϫ packed cell volume of ice-cold hypotonic buffer and allowed to swell on ice for 10 min, followed by homogenization in a Dounce homogenizer on ice. The nuclei were collected by centrifuging the homogenate at 3,300 ϫ g, 15 min, 4°C. The packed nuclei were resuspended in 2 ml of ice-cold lysis buffer (20 mM HEPES, pH 7.9, 10% (w/v) glycerol, 500 mM KCl, 0.5% Tween 20, 0.2 mM EDTA, 0.5 mM DTT, 0.2 mM PMSF, 10 g/ml leupeptin, 10 g/ml pepstatin, and 10 g/ml aprotinin) and frozen in liquid N 2 . The nuclei were then freeze-thawed twice prior to sonication (two 10-s bursts). The nuclear homogenate was centrifuged at 20,000 ϫ g, 30 min, 4°C, and the supernatant collected as the nuclear lysate.
Ion-exchange Chromatography of T-47D Nuclear Lysates-The T-47D nuclear lysate was diluted 20-fold with ice-cold buffer A (20 mM Tris, pH 7.5, 10 mM ␤-mercaptoethanol, 1 mM EDTA, 1 mM EGTA) containing 0.3 mM PMSF, filtered through a 0.2-m filter (Sartorius), and then loaded onto a Mono Q (HR 5/5) anion-exchange column (Pharmacia) equilibrated in buffer A. The column was developed with a 200 -600 mM linear NaCl gradient at 0.5 ml/min, and 30 0.5-ml fractions were collected. The fractions where the majority of the protein eluted were combined into three separate pools: fractions 8 -14 (pool 1), fractions 15-20 (pool 2), and fractions 21-27 (pool 3). Each pool was diluted by the addition of 30 ml of ice-cold buffer B (20 mM HEPES, pH 7.0, 10 mM ␤-mercaptoethanol, 1 mM EDTA, 1 mM EGTA) and then loaded onto a Mono S (HR 5/5) cation-exchange column (Pharmacia) equilibrated in the same buffer. The column was developed with a 240 -600 mM linear NaCl gradient for pool 1, or 300 -600 mM NaCl gradient for pools 2 and 3 at 0.5 ml/min, and 30 0.5-ml fractions were collected. Eluted fractions were named according to which column they eluted from, followed by which fractions the protein(s) eluted in and the pool that was loaded on the column. For example, fraction 4 eluting off the Mono S column after loading pool 1 was named MS 4 (pool 1).
Phosphorylation and Protein Sequencing of Potential Cyclin-CDK Substrates-Phosphorylation of partially purified T-47D nuclear fractions separated by ion-exchange chromatography was performed in a final volume of 30 l consisting of 50 mM HEPES, pH 7.5, 1 mM DTT, 10 mM MgCl 2 , 25 M ATP, 10 Ci of [␥-32 P]ATP, 10 mM NaF, 1 mM vanadate, and 10 mM ␤-glycerophosphate for 15 min at 37°C in the presence or absence of cyclin-CDK. Prior to phosphorylation of nuclear fractions, the different cyclin-CDK complexes were normalized for activity using GST-pRb 773-928 as substrate to ensure equivalent amounts of kinase activity were added. GST-pRb 773-928 was prepared as described previously (36). The reactions were terminated by the addition of 15 l of 3 ϫ sodium dodecyl sulfate (SDS) stop buffer (187 mM Tris-HCl, pH 6.8, 30% (w/v) glycerol, 6% SDS, 15% ␤-mercaptoethanol). The samples were then heated at 100°C for 2 min, centrifuged in a microcentrifuge, and 25 l was loaded and electrophoresed on a 20-cm length 7-15% linear sodium dodecyl sulfate polyacrylamide gel (SDS-PAGE) (Protean II; Bio-Rad). Following electrophoresis, the proteins were stained with 0.5% Coomassie Brilliant Blue R, dried under vacuum, and exposed to Kodak Biomax MR x-ray film for autoradiography. For protein sequencing of cyclin-CDK substrate protein, P105, the protein was first separated by SDS-PAGE and Coomassie-stained, and, after drying, the polyacrylamide gel strip containing the protein was excised. The protein was eluted from the polyacrylamide and then passively absorbed onto polyvinylidene difluoride prior to amino acid sequencing on a Applied Biosystems Procise Sequencer (model 494).

Generation of Purified Catalytically Active Cyclin-CDK Com-
plexes-To investigate substrates of the G 1 and S phase cyclin-CDK complexes, we constructed recombinant baculoviruses expressing N-terminal GST fusion proteins for cyclins D1, D3, E, and A as well as N-terminal 6-histidine-tagged proteins for CDKs 2, 4, and 6. Following co-infection of insect Sf9 cells with the appropriate combination of cyclin and CDK, the cyclin-CDK complexes were purified using glutathione-agarose resin. The expression of each cyclin and CDK was confirmed by immunoblotting (data not shown). To evaluate the catalytic activity of the different combinations of cyclin-CDK complexes, we utilized a recombinant GST-pRb fusion protein comprising the C-terminal fragment of the retinoblastoma gene product (GST-pRb 773-928 ), which contains phosphorylation sites for CDK2 and CDK4 (37,38). Co-infection of Sf9 cells with both the cyclin and CDK subunits, followed by purification of the complexes with glutathione-agarose, resulted in catalytically active complexes for all the combinations of cyclin-CDKs, namely cyclin D1-CDK4, cyclin D3-CDK4, cyclin D3-CDK6, cyclin E-CDK2, and cyclin A-CDK2 (Fig. 1). There was no phosphorylation of GST alone, indicating that the kinase complexes were phosphorylating the pRb 773-928 portion of the fusion protein, and as expected purified monomeric cyclin or CDK subunit alone resulted in no kinase activity toward GST-pRb 773-928 (data not shown). In the absence of added GST-pRb 773-928 , autophosphorylation of both the cyclin and CDK subunits was observed (Fig. 1, lanes 1-5). Interestingly, the level of autophosphorylation of the GST-cyclin D3 subunit (65 kDa) and the associated CDK4 (35 kDa) or CDK6 (40 kDa) subunits was significantly greater than the GST-cyclin D1 subunit (65 kDa) associated with CDK4 (35 kDa) (Fig. 1), even though the level of the protein subunits in these complexes was similar as judged by Coomassie staining. Furthermore, both cyclin E-CDK2 and cyclin A-CDK2 exhibited a significantly greater specific activity toward the GST-pRb 773-928 substrate compared with all of the cyclin D-CDK complexes as determined by the significantly lower protein levels of these complexes required to obtain an equivalent level of phosphorylation (data not shown).

Demonstration of Cyclin-CDK Substrates in Partially Purified T-47D Nuclear
Lysates-After characterizing the kinase activity of the cyclin-CDK complexes against GST-pRb 773-928 , we established conditions for investigation of endogenous cellular substrates of these kinases. Since the G 1 and S phase cyclin D-CDK (39), E-CDK (24), or A-CDK (5) complexes are localized to the nucleus, we used nuclear lysates of T-47D human breast cancer epithelial cells as our source of potential substrates. These cells exhibit the normal cell cycle phasespecific sequence of expression of cyclins D1, D3, E, A, and B1 during progression through the cell cycle following mitogenic stimulation (35). To investigate cyclin-CDK substrates, T-47D nuclear lysates were phosphorylated by the addition of exogenous cyclin-CDK complex in the presence of [␥-32 P]ATP and the resulting phosphoproteins separated by polyacrylamide gel electrophoresis, followed by autoradiography for analysis of protein phosphorylation.
Since crude nuclear lysates contain many endogenous kinases, a key aspect in demonstrating in vitro substrates of exogenously added cyclin-CDK complexes was the reduction in background phosphorylation. To overcome this problem, we employed a strategy of partially purifying these lysates by anion-and cation-exchange chromatography, before phosphorylation with exogenous cyclin-CDK complexes, similar to the approach used to demonstrate substrates of the cGMP-dependent protein kinase in rat brain (40). The rationale of this approach was that the endogenous background phosphorylation in individual fractions eluting from the ion-exchange columns would be lower than in the crude nuclear lysate since the endogenous kinases would differentially elute into different fractions from the columns, thus allowing detection of substrates of the exogenously added kinase. To test the applicability of this approach, we initially prepared a T-47D nuclear lysate, which was fractionated over a Mono Q (MQ) anionexchange column and the crude lysate and fractions eluting from the column were then phosphorylated with exogenous cyclin D3-CDK4. Phosphorylation of crude nuclear T-47D lysate in the presence of exogenous cyclin D3-CDK4 yielded no detectable additional specific substrates when compared with lysate where no cyclin D3-CDK4 was added ( Fig. 2A). However, phosphorylation of fractions eluting from the Mono Q column revealed an in vitro substrate migrating at 24 kDa. This pro-tein eluted in  where the level of background phosphorylation was relatively low (Fig. 2B). This result indicated that in vitro substrates of exogenous cyclin-CDKs may be detected in crude nuclear lysates by employing this approach. However, it was apparent that partial purification on the Mono Q column alone did not sufficiently reduce the background phosphorylation in most of the eluates to allow identification of more than a minority of potential substrates, i.e. MQ 8 -24 (Fig. 2B).
To obtain further decreases in endogenous kinase phosphorylation and hence reduction in background phosphorylation, fractions eluting from the Mono Q column were combined into three separate pools, fractions 8 -14 (pool 1), fractions 15-20 (pool 2), and fractions 21-27 (pool 3). Each pool was then separated on a Mono S (MS) cation-exchange column prior to phosphorylation with cyclin D3-CDK4. Further purification unveiled additional substrates. Hence, after resolving pool 3 (Mono Q eluates, fraction 21-27) over the Mono S column, proteins migrating at 42 kDa in MS 2-10 as well as 105 kDa in fractions MS 16 -24 (Fig. 3B) were phosphorylated by cyclin D3-CDK4 that were previously undetectable after phosphorylation of Mono Q eluates (Fig. 2B). The 42-kDa protein was also present in MS 2-10 after resolving pool 2 (Mono Q eluates, fraction 15-20) over the Mono S column (Fig. 4C). The 42-kDa protein was present in both pools 2 and 3, probably because its elution from the Mono Q column overlapped fractions combined for these two pools. These results demonstrated that after two purification steps the level of endogenous kinase activity can be reduced significantly, hence allowing detection of substrates of exogenously added kinases. The further discussion of cyclin-CDK phosphorylation includes only substrates that were relatively prominently and reproducibly phosphorylated.
Differential Substrate Specificity of Cyclin D1-CDK4 and Cyclin D3-CDK4 -Current evidence suggests that the three D type cyclins (D1, D2, and D3) in association with either CDK4 or CDK6 are differentially expressed in different tissues; however, nothing is known as to whether these different combinations of complexes have identical or differing substrate specificities, which may be an important issue relating to whether they have redundant or differing functional roles. We therefore chose initially to investigate the role of the D-type cyclin subunit in conferring substrate specificity on the holoenzyme complex. Cyclins D1 and D3 have the lowest sequence homology in the human cyclin D family (41) and are expressed at different stages of G 1 (35). These two D-type cyclins are thus most likely to exhibit potential differences, and we therefore compared the phosphorylation of partially purified T-47D lysates with cyclins D1 and D3 both in association with CDK4. To ensure that equivalent levels of kinase activity were added to the phosphorylation reactions in this and all further experiments when comparing different cyclin-CDKs, we normalized the activity of the complexes using GST-pRb 773-928 as substrate. Phosphorylation of MS 14 (pool 1) revealed preferential phosphorylation of a 38-kDa protein by cyclin D1-CDK4 while cyclin D3-CDK4 preferentially phosphorylated a protein of 102 kDa not phosphorylated by cyclin D1-CDK4 (Fig. 4B), thus demonstrating differential substrate specificity of these two kinase complexes. Two other major substrates were exclusively phosphorylated by cyclin D3-CDK4 and not cyclin D1-CDK4, including the 42-kDa protein in MS 2-10 (pools 2 and 3) (Fig. 4C) and the 105-kDa protein in MS 16 -24 (pool 3) (Fig. 4D). The 24-kDa cyclin D3-CDK4 substrate in MQ 26 -30 (Fig. 2B) was phosphorylated by both kinases, albeit to a greater extent by cyclin D3-CDK4 (Fig. 5D). These studies indicate that cyclin D1-CDK4 and cyclin D3-CDK4 as well as sharing common substrates also phosphorylate specific substrates. Thus, the differ-  1 and 7), cyclin D3-CDK4 (lanes 2 and 8), cyclin D3-CDK6 (lanes 3 and 9), cyclin E-CDK2 (lanes 4 and 10), and cyclin A-CDK2 (lanes 5 and 11) were incubated with [␥-32 P]ATP in either the absence (lanes 1-5) or presence of 3 g of GST-pRb 773-928 (lanes 7-11) for 15 min at 37°C. Lane 6 contains GST-pRb 773-928 alone with no cyclin-CDK added. Following phosphorylation, the phosphoproteins were separated on a 10% sodium dodecyl sulfate-polyacrylamide gel and visualized by autoradiography. The position of standard molecular mass marker proteins in kDa is shown on the left, and phosphorylated GST-pRb 773-928 is indicated with an arrow on the right. ent D-type cyclin subunits can modulate the substrate specificity of the CDK4 subunit.
Comparison of Cyclin D3-CDK4 and Cyclin D3-CDK6 Substrate Phosphorylation-Apart from CDK4, the D-type cyclins can also bind and activate CDK6 during the G 1 phase of the cell cycle (42). We were therefore interested in determining the substrates phosphorylated by CDK6 and comparing these to CDK4 substrates to determine if the CDK subunits can confer substrate specificity. After normalization of activity against GST-pRb 773-928 , we compared substrates phosphorylated by cyclin D3-CDK4 and cyclin D3-CDK6. These studies showed that the 105-kDa protein in MS 16 -26 (pool 3) (Fig. 5B) and a 102-kDa protein in MS 14 (pool 1) (Fig. 5C) were phosphorylated to an equivalent level by cyclin D3-CDK4 and cyclin D3-CDK6. However, the 24-kDa cyclin D3-CDK4 substrate in MQ 26 -30 was not phosphorylated by cyclin D3-CDK6 (Fig. 5D). We were unable to determine if the 42-kDa cyclin D3-CDK4 substrate in MS 2-10 (pools 2 and 3) was also phosphorylated by cyclin D3-CDK6, since autophosphorylated CDK6 migrated close to this size and therefore masked any potential phosphorylation of this protein. Therefore, along with common substrates, there is also differential phosphorylation of one protein between cyclin D3-CDK4 and cyclin D3-CDK6, indicating that apart from the cyclin D subunit, the CDK4 and CDK6 subunits can confer substrate specificity on the holoenzyme complex.
Differential Substrate Specificity of Cyclin E-CDK2 and Cyclin A-CDK2-We further investigated substrates phosphorylated by cyclin E-CDK2 and cyclin A-CDK2 as these kinases are important for cell cycle progression through late G 1 (Fig. 6C), a 42-kDa protein in MS 2-10 (pools 2 and 3) (Fig. 6D), and a 23-kDa protein in MS 28 (pool 1) (Fig. 6E). Many of these cyclin A-CDK2 substrates were also phosphorylated by cyclin E-CDK2, such as 107-, 52-, 44-, and 10-kDa proteins in MQ 28 (Fig. 6B) and 42-kDa in MS 2-10 (pools 2 and 3) (Fig. 6D). However, many were preferentially phosphorylated by cyclin A-CDK2, such as the 33-and 17-kDa proteins in MQ 28 (Fig. 6B), as well as the 23-kDa protein in MS 28 (pool 1) (Fig. 6E). The only preferential substrate observed for cyclin E-CDK2 was a 34-kDa protein in MS 14 -16 (pool 1) (Fig. 6F). Thus, both complexes phosphorylate a 102-kDa protein in MS14 (pool 1) to an equivalent level, demonstrating a common substrate, whereas cyclin E-CDK2 and cyclin A-CDK2 clearly preferentially phosphorylate 34-kDa and 78-kDa proteins, respectively. Therefore, phosphorylation of this fraction provides a good illustration of the differential substrate specificity exerted by the E and A cyclin subunits on CDK2. Interestingly, phosphorylation of the 42-kDa protein(s) in MS 2-10 (pools 2 and 3) by either cyclin E-CDK2 or cyclin A-CDK2 (Fig. 6D), resulted in several bands retarded in mobility in the polyacrylamide gels by up to 4 kDa as judged by silver staining of the gel (data not shown). Moreover, although the 42-kDa protein(s) was also phosphorylated by cyclin D3-CDK4, there was no significant retardation of mobility (Fig. 4C), suggesting that the different cyclin-CDKs may phosphorylate this protein(s) on different sites resulting in its differing electrophoretic behavior. Alternatively, the 42-kDa protein(s) phosphorylated by cyclin D3-CDK4 may be distinct from the protein(s) of similar molecular mass phosphorylated by cyclin E-CDK2 and cyclin A-CDK2. To investigate this possibility, two-dimensional gel electrophoresis following phosphorylation of the 42-kDa protein(s) by these three different cyclin-CDKs was performed. Two-dimensional gel electrophoresis revealed that the 42-kDa protein(s) migrated at a similar isoelectric point (pI) following phosphorylation by these different cyclin-CDKs, with approximate pI values of 5.5 following phosphorylation with cyclin D3-CDK4 and 5.2-5.5 following phosphorylation by cyclin E-CDK2 and cyclin A-CDK2 (data not shown). These results strongly suggest that the 42-kDa protein(s) phosphorylated by the different cyclin-CDKs are the same. Ultimately, unequivocal demonstration that a substrate of one particular cyclin-CDK is also a substrate for other cyclin-CDKs will require its identification, followed by phosphorylation of the homogeneous protein with the different cyclin-CDKs.
We were unable to detect phosphorylation of pRb with any of the cyclin-CDK complexes used. pRb eluted in MS 16 -26 (pool 1) as determined by Western blotting (data not shown) and migrates as a ladder of bands from approximately 105-110 kDa due to endogenous phosphorylation on several residues. The reasons for this are unclear, since the bacterially expressed C-terminal truncated pRb GST fusion protein (GST-pRb 773-928 ) used in our initial characterization of the kinase activity of these complexes served as a very good substrate (Fig. 1). It is possible that due to the diffuse migration of endogenous pRb on the gels it is difficult to demonstrate phosphorylation of a distinct protein band or that the level of pRb extracted was too low to observe significant phosphorylation. Furthermore, there was still a significant level of background phosphorylation in these eluates, which may have masked any pRb phosphoryla-tion (see Fig. 6C). A complete summary of substrates demonstrated in this study following phosphorylation by the different cyclin-CDKs is provided in Table I.
Identification of P105 as Nucleolin-To assess the potential utility of this methodology in identifying cyclin-CDK substrates, we performed N-terminal protein sequencing of one of the substrates characterized in this study. P105 (Figs. 3B, 4D, and 5B), which was phosphorylated by all the cyclin-CDKs apart from cyclin D1-CDK4 (Table I), was chosen for this purpose, since after only two ion-exchange purification steps this protein was relatively pure, as judged by Coomassie staining of the polyacrylamide gel. Following transfer of P105 to polyvinylidene difluoride and N-terminal protein sequencing, the following sequence was obtained: VKLAKAGKNQGDPKK-MAPPPK. A search of the Swiss-Prot protein sequence data base revealed that this sequence was a perfect match for the N-terminal sequence of nucleolin. Nucleolin has previously been described as a CDK1 (CDC2) substrate (34); hence, its identification as a substrate of cyclin-CDKs in our study indicates the general applicability of this approach for identifying cyclin-CDK substrates. DISCUSSION This study has focused on the characterization and identification of substrates of the cyclin-CDK complexes active through the G 1 and S phases of the cell cycle. A number of different approaches for identifying protein kinase substrates have previously been utilized, each with some limitations. Phosphorylation of peptide libraries is a powerful technique that determines the optimal primary consensus sequence of protein kinases (31), which may then be used to search sequence data banks for potential substrates. However, this approach does not take into account secondary or tertiary structure of proteins, which can be important for substrate recognition (43). Moreover, potential novel substrates not present in sequence data banks cannot be identified. Another approach that has recently been used to identify protein kinase substrates is the yeast two-hybrid interaction screen (22,32). A limitation of this approach is that this method may not be sensitive enough to identify many kinase-substrate interactions, since these interactions may be only transient, with the kinase dissociating from the substrate after phosphorylation. Furthermore, this method can identify proteins which interact with the protein kinase that are not necessarily substrates but are regulatory subunits or binding proteins. Other approaches have involved phosphorylating proteins expressed from phage cDNA expression libraries on nitrocellulose filters with mitotic extracts of HeLa cells and then probing these filters with a phosphoprotein-specific monoclonal antibody to identify protein substrates (44,45). A recent modification of this approach has involved phosphorylating proteins expressed on nitrocellulose filters with purified ERK1 mitogen-activated protein kinase in the presence of [␥-32 P]ATP, which identified a novel substrate termed MNK1, itself a serine/threonine kinase (33). Phosphorylation of protein expression libraries with purified kinases overcomes many of the problems already described, since potential protein substrates expressed from the library contain secondary and tertiary structure and only substrate and not just interacting proteins will be detected. Despite limitations, the methods described have nevertheless identified novel protein substrates.
Other groups have phosphorylated crude cellular lysates with exogenous kinases to identify potential substrates. A major problem with this approach is the presence of high background endogenous kinase activity, which makes it difficult to demonstrate substrates of exogenously added kinase, as documented in this study. To overcome this problem, Kwon and colleagues (46) heat-treated cellular lysates to inactivate endogenous kinases. The heat-treated lysates were then phosphorylated by immunoprecipitated cyclin-CDK complexes. Although this method demonstrated three proteins of 88, 44, and 18 kDa that were phosphorylated by all three D type cyclins in association with either CDK4 or CDK6 as well as cyclin E-CDK2, a major problem of this approach is that heat treatment may lead to subunit dissociation as well as protein denaturation. Denaturation of proteins may actually create a substrate from a protein not phosphorylated in its native conformation, presumably by exposing a substrate recognition motif that is otherwise inaccessible in the native conformation or, conversely, denaturation may render a substrate protein incapable of being phosphorylated since secondary and tertiary structure can play significant roles in substrate recognition (43). Therefore, to overcome many of the potential problems already described, the approach we undertook was to phosphorylate nuclear lysates that had been partially purified by ion-exchange chromatography with purified cyclin-CDK complexes. The key aspect of this approach is that by partially purifying the lysates by ion-exchange chromatography a significant reduction in background phosphorylation due to endogenous protein kinases is achieved. This allowed detection of endogenous substrates of exogenously added cyclin-CDKs, similar in principle to the method used to characterize substrates of cGMP-dependent protein kinase in rat brain (40). This method has a number of advantages over previously published methods. Most importantly, unlike phosphorylation of peptides or heat-inactivated samples, proteins are phosphorylated in their native conformation thus precluding artifactual phosphorylation or lack of phosphorylation due to protein conformational changes. Furthermore, of relevance to the present study, this approach allowed the analysis of substrate specificity of various cyclin-CDKs. Thus, unlike library phosphorylation screening or the yeast two-hybrid screen where individual substrate clones are first isolated and characterized, by first phosphorylating a complex mixture of proteins, it was possible to compare the substrate preferences of different kinases with several different substrates simultaneously. Apart from the advantages enunciated, clearly our approach as well as the others discussed all have limitations and no one particular method will identify all substrates of a protein kinase. Therefore, for example, low abundance substrate proteins may be difficult to detect without further purification and reduction in background phosphorylation. Thus, although we were unable to demonstrate phosphorylation of endogenous pRb, nevertheless phosphorylation of numerous other endogenous substrates was demonstrated.
Utilizing this approach, we investigated substrate phosphorylation by the cyclin-CDKs active in G 1 and into S phase of the cell cycle. We initially investigated the cyclin D-CDKs, since these are the first to be activated in G 1 phase following mitogenic stimulation (3). Although it has clearly been established that a major substrate target of the D-type cyclin-CDKs during G 1 -S phase progression is pRb (20), we were interested in determining if other substrates may also exist that may be important for the action of these kinases. Recent evidence for other cyclin D-CDK substrates comes from a study using a two-hybrid screen, which isolated a Myb-like transcription factor, DMP1, phosphorylated in vitro by all three D-type cyclins associated with either CDK4 or CDK6 (22). Another important question pertains to why three different D-type cyclins and two CDKs are required if their sole purpose is pRb phosphorylation and hence whether the three D-type cyclins in association with CDK4 or CDK6 have entirely redundant roles during G 1 -S phase progression in various tissues. Although the emerging evidence indicates that the different D-type cyclins are differentially active in specific tissues (15)(16)(17), in cells expressing more than one D-type cyclin, differences in their temporal expression exist suggesting cell cycle stage-specific roles. Thus, following mitogenic stimulation, cyclin D3 expression peaks later than cyclin D1 in HeLa cells, normal mammary epithelial cells (47), and T-47D breast cancer cells (35). Furthermore, 32D myeloid cells ectopically overexpressing cyclin D3 but not cyclin D1 fail to differentiate in the presence of granulocyte colony-stimulating factor, showing that in these cells the roles of these two cyclins are not redundant (48). Moreover, cyclins D2 and D3 but not cyclin D1 are capable of activating CDK2 (3,49,50), which may explain some of the different effects of these cyclins in vivo. Based on these and other studies, we postulated that apart from activating different CDKs, some of these differential effects may be due to D-type cyclins modulating the substrate specificity of a particular CDK. Our results showed that while cyclin D1-CDK4 and cyclin D3-CDK4 both phosphorylated a common substrate of 24 kDa, this protein was phosphorylated to a significantly greater extent by cyclin D3-CDK4. Specific substrates for each kinase were also shown to exist, such as a 38-kDa protein for D1-CDK4 and 42-and 105-kDa proteins for D3-CDK4. Therefore, these results show that cyclins D1 and D3 can confer differential substrate specificity on CDK4 and provide a biochemical rationale for some of their different effects in vivo. We further went on to investigate the relative contribution of the CDK subunit in conferring substrate specificity. Comparison of substrates phosphorylated by cyclin D3-CDK4 and cyclin D3-CDK6 showed that these two complexes appeared to have similar specificities both phosphorylating proteins of 102 and 105 kDa; however, the 24-kDa cyclin D3-CDK4 substrate was not phosphorylated by cyclin D3-CDK6. Therefore, for the cyclin D-CDKs tested in this study, both the cyclin and CDK subunits can modulate substrate specificity although it appears that at least for cyclins D1, D3, and CDKs 4 and 6, the cyclin subunit plays a greater role in conferring specificity. These studies suggest that, while the different combinations of D-type cyclins associated with CDK4 or CDK6 are important for mediating progression through the G 1 phase of the cell cycle by phosphorylation of pRb, other targets of the individual complexes also exist that may be important for other specific functions. By targeting common as well as specific substrates, it may be envisaged that a greater degree of flexibility for the more specialized functions of the different cyclin D-CDKs may be achieved in particular tissues or during different phases of the cell cycle.
Along with cyclin D-CDKs, substrates phosphorylated by cyclin E-CDK2 and cyclin A-CDK2 were also investigated to determine if differential substrate phosphorylation by these kinases may contribute to their cell cycle-specific actions. Previous studies suggest that as with the cyclin D-CDK complexes, pRb is a major substrate target for both cyclin E-CDK2 and cyclin A-CDK2 (51, 23). However, apart from pRb phosphoryl-ation, important differences between cyclin D-CDKs and cyclin E/A-CDKs exist. Most notably, unlike cyclin D1-CDK4, which is required for G 1 phase progression only in cells expressing pRb (20), cyclin E-CDK2 is also required in cells lacking pRb for G 1 -S phase progression (24), implying that other substrates are also required. This notion is supported by work showing that cyclins D1 and E have additive effects in shortening the G 1 interval prior to entering S phase compared with either cyclin alone (52). Moreover, a recent study, where pRb was phosphorylated in vitro by either cyclin D1-CDK4, cyclin E-CDK2 or cyclin A-CDK2 and then microinjected into SAOS-2 cells, demonstrated that only phosphorylation by cyclin D1-CDK4 alleviated the cell cycle arresting effects of pRb, implicating other substrates for cyclin E-CDK2 and cyclin A-CDK2 action (38). Our studies showed that cyclin E-CDK2 and cyclin A-CDK2 phosphorylated significantly more substrates than the cyclin D-CDKs, with cyclin A-CDK2 exhibiting the greatest range, phosphorylating 28 prominent substrates, and therefore support the concept that these two cyclin-CDKs mediate their cell cycle effects by phosphorylation of numerous targets. Most of the substrates phosphorylated by cyclin E-CDK2 were also phosphorylated by cyclin A-CDK2 while proteins of 22, 23, 24, 33, 36, and 78 kDa were specifically phosphorylated by cyclin A-CDK2 but not cyclin E-CDK2 (Table I). On the other hand, only one prominent substrate of 34 kDa specific for cyclin E-CDK2 was demonstrated. To our knowledge, this is the first protein substrate shown to be preferentially phosphorylated by cyclin E-CDK2 compared with cyclin A-CDK2 and may there-TABLE I Summary of cyclin-CDK substrates in T-47D breast cancer cell nuclear lysates Proteins phosphorylated or not phosphorylated by a particular cyclin-CDK are indicated as ϩ and Ϫ, respectively, and ? denotes inability to determine if the protein is phosphorylated by the cyclin-CDK due to autophosphorylation of the GST-cyclin or CDK subunit. The abbreviations used for the different cyclin-CDK complexes are: cyclin D1-CDK4 (D1/K4), cyclin D3-CDK4 (D3/K4), cyclin D3-CDK6 (D3/K6), cyclin E-CDK2 (E/K2), and cyclin A-CDK2 (A/K2). kDa fore represent a key substrate important in mediating the cell cycle stage-specific effects of cyclin E-CDK2 during the G 1 -S phase transition. Phosphorylation of a 42-kDa substrate(s) by either cyclin E-CDK2 or cyclin A-CDK2, but not by cyclin D3-CDK4, resulted in a retarded mobility of this protein(s) in the polyacrylamide gel. This is reminiscent of the retardation in mobility of pRb upon phosphorylation by these two cyclin-CDKs (38), suggesting conformational changes of this protein(s) upon phosphorylation by these cyclin-CDKs. Ultimately, the identity of the cyclin-CDK substrates demonstrated in this study will need to be determined to evaluate if their in vitro phosphorylation by cyclin-CDKs also occurs in vivo and the functional consequences of these phosphorylations during cell cycle progression. In preliminary work, protein sequencing of one of the cyclin-CDK substrates characterized in this study (P105) identified this protein as nucleolin, a known substrate of cyclin B-CDC2 during mitosis (34). Nucleolin is believed to be important in the transcription and processing of ribosomal RNA and therefore protein synthesis, as well as nucleolar structure. Therefore, these results now raise the possibility that the function of nucleolin is controlled by cyclin D3-CDK4/6, cyclin E-CDK2, and cyclin A-CDK2 phosphorylation, prior to mitosis. Further work identifying the sites phosphorylated on nucleolin by the different cyclin-CDK complexes both in vitro and in vivo, in conjunction with site-directed mutagenesis studies will need to be performed to determine the role of cell cycle dependent phosphorylation of nucleolin. Importantly, the ability to identify a previously characterized CDK substrate illustrates the utility of this approach in identifying cyclin-CDK substrates and suggests that many of the substrates demonstrated in this study are likely to have important roles in CDK action.
In conclusion, the work presented here demonstrates that both the cyclin and CDK subunits can modulate the substrate specificity of the holoenzyme complex. Through different combinations of associations of the G 1 and S phase cyclin and CDK subunits, phosphorylation of both common and some specific substrates for the different cyclin-CDKs in breast epithelial cells has been demonstrated. These studies suggest that phosphorylation of some common as well as key specific substrates by the different cyclin-CDKs is important for transition through specific stages of G 1 and S phase of the cell cycle. The identification and characterization of selected cyclin-CDK substrates demonstrated in this study with a view to determining their roles in cell cycle progression is currently under further investigation.