Studies on the in Vitro Phosphorylation of HSSB-p34 and -p107 by Cyclin-dependent Kinases CYCLIN-SUBSTRATE INTERACTIONS DICTATE THE EFFICIENCY OF PHOSPHORYLATION*

Cyclin-dependent kinases (Cdks) are required for cell cycle progression. Two potentially significant Cdk substrates in human cells are the human single-stranded binding protein (HSSB or RPA), which plays an essential role in DNA replication, repair, and recombination, and the tumor suppressor p107 which acts to negatively regulate cell growth. In this report we describe the in vitro phosphorylation of these two proteins by Cdks in an attempt to under- stand how cyclin-substrate interactions direct phosphorylation efficiencies. We show that cyclin A-Cdk2 effi- ciently phosphorylates the p34 subunit of HSSB (HSSB-p34) alone or as a part of the heterotrimeric complex. In contrast, cyclin E-Cdk2 that is active in phosphorylating histone H1, does not support the phosphorylation of the p34 subunit of HSSB. We provide evidence that this dif-ferential phosphorylation results from a specific inter- action between HSSB-p34 and cyclin A, but not cyclin E. Thus the observed cell cycle-dependent phosphoryla- tion of HSSB-p34 at the G 1 to S transition is most likely catalyzed by cyclin A-Cdk2 initiated by the direct inter- action between cyclin A and the HSSB-p34 subunit. These studies are consistent with our previous obser- vation that p107, which directly binds cyclin A, is efficiently phosphorylated by cyclin A-Cdk2 but not cyclin B-associated kinases. Here we further demonstrate that cyclin

Two important events of the eukaryotic cell cycle are the S and M phases, when chromosomes are replicated and when replicated chromosomes are segregated into dividing cells, respectively. Progression through the cell cycle is controlled by several factors including cyclin-dependent kinases (Cdks), 1 Cdk inhibitors, Cdk-activating kinase, Cdk phosphatases, and the checkpoint proteins (see Refs. 1-3 for reviews). It is likely that these mechanisms serve to regulate downstream pathways leading to coordinated DNA replication, DNA repair, and chromosomal segregation.
Key regulators of the G 1 to S phase transition in mammalian cells include three D-type cyclins and cyclin E (4). In the presence of growth factors, the D-type cyclins exhibit moderate oscillations during the cell cycle reaching a peak at the G 1 /S transition (5). Maximal levels of cyclin E-dependent kinase activity are also apparent at this stage (6). Once cells enter S phase, cyclin E is degraded and its kinase partner (Cdk2) complexes with cyclin A concomitantly with the onset of DNA synthesis. In order for cells to enter mitosis after DNA replication, cyclin B must accumulate above a threshold level and complex with Cdc2 (7,8). Following completion of division this cyclin is proteolyzed and the cell progresses to G 1 (9).
The mechanism by which cyclins activate Cdks has recently been revealed by the crystal structure of the cyclin A-Cdk2-ATP complex (10). Cyclin A binds to one side of the Cdk2 catalytic cleft inducing large conformational changes that realign active site residues and relieve the steric block at the entrance to the catalytic cleft.
It is thought that formation of each successive cyclin-Cdk complex is responsible for phosphorylating substrates essential for any given cell cycle event. Thus the G 1 /S phase cyclin E or A/Cdk2 complex may phosphorylate and activate proteins necessary for DNA replication, and the G 2 /M phase cyclin B/Cdc2 phosphorylates proteins necessary for mitosis and cytokinesis (see Ref. 3 for a review). How these cyclin/Cdk complexes activate replication at the G 1 /S transition and what their physiological substrates are, is not known. In this report we have utilized two Cdk substrates in human cells, the human singlestranded binding protein (HSSB) and the tumor suppressor protein p107, to investigate the mechanism by which a Cdk targets its substrate.
HSSB, a complex containing three polypeptides of 70, 34, and 11 kDa, was originally identified as a HeLa cell protein required for the in vitro replication of SV40 DNA (12). While the p70 subunit of HSSB binds to single-stranded DNA (16), the biological functions of the p34 and p11 subunits are not known. However a direct role for the HSSB-p34 and -p11 in this reaction is likely as antibodies against these subunits block SV40 DNA replication (16,17). The trimeric HSSB is an essential component of the replication initiation complex functioning to stabilize unwound regions of DNA as well as complexing with T antigen and DNA polymerase ␣-primase to catalyze RNA primer synthesis (18). HSSB also plays multiple roles in the elongation stages of replication stimulating DNA polymerases ␦ and ⑀ to catalyze elongation of DNA chains (19).
In addition to its role in replication, HSSB is essential for nucleotide excision repair (20 -25) and may play important roles in recombination, double-stranded break repair and transcription (26 -30).
The p34 subunit of HSSB is phosphorylated at the G 1 /S boundary of the cell cycle (12) or following exposure to DNA damage-inducing agents, including ultraviolet light (31) or ionizing radiation (32). We have demonstrated that cyclin A-dependent hyperphosphorylation of HSSB-p34 in G 1 extracts is due to the combined actions of Cdk-cyclin A and the DNA-dependent p350 protein kinase (DNA-PK) (33). During G 1 to S phase progression, cyclin A expression leads to the formation of active cyclin protein kinase A complexes that may initiate efficient phosphorylation of the p34 subunit of HSSB allowing further phosphorylation by DNA-PK that converts HSSB into a hyperphosphorylated form. The biological significance of these findings is currently not clear as the phosphorylation state of HSSB does not alter its in vitro activity in either SV40 DNA replication or nucleotide excision repair (34 -36). However it has been suggested that HSSB may be part of a cellular signal transduction mechanism for sensing that DNA synthesis is ongoing (34).
Both cyclins E and A are active during the cell cycle phase at which HSSB is phosphorylated indicating that cyclin E could also be involved in this phosphorylation event. However in vitro experiments presented here suggest that it is exclusively cyclin A/Cdk2 that comprises the cyclin/kinase complex responsible for phosphorylation of the p34 subunit of HSSB during the G 1 /S transition as a result of a specific interaction between cyclin A and its HSSB-p34 substrate. Cyclin E cannot participate in such a protein/protein interaction and is unable to phosphorylate HSSB. This finding is consistent with our previous observation that p107 is phosphorylated by cyclin A-CDK2 as the result of its interaction with cyclin A (42). We provide evidence that cyclin A only interacts with unphosphorylated p107 and following substrate modification becomes dissociated.

Preparation of Protein Reagents
GST-tagged cyclins A, D1, D2 and D3 (constructs were kind gifts from Dr. Yue Xiong, University of North Carolina) were overexpressed under the control of T7 RNA polymerase in Escherichia coli (BL21-DE3). Freshly transformed bacteria were grown overnight in Luria broth containing 0.4% glucose and 0.5 mg/ml ampicillin at 37°C. The overnight culture was diluted 1:100 with the same medium (1 liter) and grown to an A 600 of 0.6 at 37°C in the presence of ampicillin (0.5 mg/ml). The cells were then shifted to 25°C, and expression was induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside (IPTG) overnight. The cells were pelleted and lysed in 10 ml of buffer A (25 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.01% Nonidet P-40, 10% glycerol, 2 mM PMSF, 4 g/ml antipain, and 2 g/ml leupeptin) plus 0.1 M NaCl and 1 mg/ml lysozyme for 30 min at 0°C. Following sonication (six repetitive 20-s treatments) and centrifugation at 39,000 ϫ g for 15 min, the soluble supernatant protein was loaded onto glutathione-Sepharose beads, washed, and eluted as recommended by the manufacturer (Pharmacia Biotech Inc.).
E. coli cells (BL21-DE3-pLys S) transformed with a plasmid expressing GST-cyclin E (pGex-cycE and pGex-2T-cycE; generously provided by Dr. A. Koff, Sloan-Kettering Cancer Institute) were overexpressed as described above for the other cyclins except that the IPTG-induced expression step was carried out in the absence of ampicillin. The soluble GST-cyclin E protein (630 mg, 70 ml) was loaded onto a 1-ml glutathione-Sepharose column (1 ϫ 1.3 cm, 1 ml) equilibrated in buffer A plus 0.1 M NaCl. Following a wash step with buffer A plus 0.8 M NaCl (40 ml), 15 mg of GST-cyclin E were eluted from the column with 20 mM glutathione in 5 ml of the same buffer. However, GST-cyclin E was rapidly degraded and only approximately 10% of the glutathione eluted material was full-length, while 90% comprised a 30-kDa GST-cyclin E proteolyzed product that cross-reacted with the anti-cyclin E antibody.
In order to remove this degraded protein, the eluted material (15 mg) was loaded onto a 25-ml fast protein liquid chromatography-Superose 6 sizing column equilibrated in buffer A plus 0.5 M NaCl. Elution with the equilibration buffer resulted in the isolation of 0.3 mg of full-length GST-cyclin E.

Purification of DNA-dependent Kinase
Both components of the DNA-dependent kinase (DNA-PK) enzyme, the Ku autoantigen and the p350 catalytic subunit, were purified from Hela cell nuclear extracts. The assay used for the purification of each of these proteins was the phosphorylation of the serine residue present within the synthetic peptide (PK53) Glu-Pro-Pro-Leu-Ser-Gln-Glu-Ala-Phe-Ala-Asp-Leu Trp-Lys-Lys. The phosphorylation assay was carried out in a reaction (20 l) containing 50 mM HEPES, pH 7.5, 0.5 mM DTT, 10 mM MgCl 2 , 50 mM KCl, 0.1 mM EDTA, 100 M [␥-32 P]ATP, 0.2 g of bovine serum albumin, in the presence or absence of 50 ng of calf thymus DNA (average size, 300 base pairs). One unit of Ku was defined as the amount of enzyme that increased the 32 P incorporation into the PK53 peptide by 1 pmol in the presence of 100 units of p350 at 30°C after incubation for 30 min. One unit of p350 was defined as the amount required to catalyze the incorporation of 1 pmol of 32 P into the PK53 peptide at 30°C after incubation for 30 min. In the assay that scored for the p350 subunit of DNA-PK, each reaction contained 1500 units of Ku autoantigen (16 ng). In assays used for the measurement of the Ku autoantigen, 325 units of p350 protein were added. The reactions were then incubated for 30 min at 30°C before being stopped by the addition of 20 l of 30% acetic acid. Aliquots were spotted onto phosphocellulose filters that were washed four times with 15% acetic acid before being counted by liquid scintillation.
Purification of the p350 Subunit of DNA-PK-HeLa cell nuclear extract (840 mg, 76.4 ml, 1.65 ϫ 10 8 units) was loaded onto a phosphocellulose column (2.5 ϫ 4 cm, 20 ml) equilibrated in buffer D (50 mM Tris-HCl, pH 7.9, 1 mM EDTA, 0.02% Tween 20, 1 mM DTT, 1 g/ml each of pepstatin A, leupeptin, and soybean trypsin inhibitor, 10 g/ml PMSF and 5% glycerol) plus 0.1 M KCl. The column was washed with 15 column volumes of buffer D plus 0.1 M KCl and active fractions (as determined by the assay described above) were then eluted with 47 ml of buffer D containing 0.3 M KCl. This pooled fraction (122 mg, 47 ml, 1.77 ϫ 10 8 units) was diluted to 0.1 M KCl with buffer D and loaded onto a DEAE-cellulose column (2.5 ϫ 2 cm, 10 ml) equilibrated with buffer D plus 0.1 M KCl. Following a 10-column volume wash step using the same buffer, the p350 DNA-PK activity was eluted with buffer D plus 0.2 M KCl (31.9 mg, 82 ml, 9.3 ϫ 10 7 units) and applied to a 25-ml Superose 6 gel filtration column equilibrated with buffer D plus 0. Overall, 1.3% of the activity was recovered and 110-fold purification was achieved resulting in Ͼ90% homogeneity as assessed by SDS-PAGE analysis and silver staining (data not shown).
Purification of Ku-In the phosphocellulose chromatographic step used for the isolation of the DNA-PK p350 described above, all of the Ku activity was recovered in the 0. 10-ml linear gradient (0.9 mg, 11.25 ml, 8.4 ϫ 10 7 units). Overall, 29% of the Ku autoantigen activity was recovered, and 273-fold purification was achieved resulting in Ͼ90% homogeneity as visualized by SDS-PAGE analysis (data not presented).

Cloning and Purification of GST-p34 HSSB
PCR primers 5Ј (5Ј-TAAGACAAGGGATCCATGTGGAACAGTG-GATTCGAAAGCTATGGC3Ј) and 3Ј (5Ј-TCCGGAAGATCTGGTACCT-TATTCTGCATCTGTGGATTTAAAATGGTC-3Ј) to the coding sequence of the HSSB-p34 subunit, were designed to include 5Ј BamHI and 3Ј KpnI restriction sites (both underlined). The DNA product resulting from 35 cycles of amplification, after electrophoretic separation, was excised from a 1% agarose gel, digested with BamHI and KpnI, and ligated into a Pet 19b vector that included a GST fusion followed by thrombin, BamHI, and KpnI cleavage sites. The resulting plasmid was transfected into E. coli cells (BL21-DE3) and grown overnight in L broth containing 0.4% glucose and 0.5 mg/ml ampicillin. This culture was diluted 1:100 with the same medium and grown to an A 600 of 0.6 at 37°C and induced with 0.4 mM IPTG for 3 h. Cells from 1 liter of medium (5 g) were pelleted and then lysed in buffer A plus 1 mg/ml lysozyme for 30 min at 0°C. After sonication and centrifugation at 12,000 ϫ g for 15 min, the supernatant was loaded onto a glutathione-Sepharose column (0.7 ϫ 2.7 cm, 0.5 ml) which was washed and eluted as recommended by the manufacturer. The yield of the GST-p34 was 0.6 mg of protein from each L of cells.
Purification of Cyclin A/Cdk2 and Cyclin E/Cdk 2 G 1 extracts (5 mg) were incubated with human GST-cyclin A or GST-cyclin E (0.05 mg each) for 20 min at 30°C in the presence of 40 mM creatine phosphate (diTris salt, pH 7.7), 25 g/ml of creatine kinase, 7 mM MgCl 2 , 0.5 mM DTT and 4 mM ATP. The reaction mixtures (1 ml) were each loaded onto a glutathione-Sepharose column (0.7 ϫ 2.7 cm, 0.3 ml) equilibrated in buffer A (25 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.01% Nonidet P-40, 0.1 mM PMSF, 0.2 mg/liter antipain, and 0.1 mg/liter leupeptin) plus 0.2 M NaCl. After washing successively with 2 ml of equilibration buffer and 2 ml of buffer A containing 0.8 M NaCl, bound protein was eluted with 1.5 ml of buffer A plus 0.05 M NaCl and 20 mM glutathione. The resulting fractions contained cyclin A or E/Cdk2 activity as determined by immunoblot analysis and H1 kinase assays. One unit of kinase activity was defined as the amount of cyclin/kinase complex that catalyzed the incorporation of 1 pmol of 32 P into histone H1 after incubation for 30 min at 30°C.
Kinase Assays-For the phosphorylation of HSSB with G 1 extracts, reaction mixtures (10 l) contained 40 mM creatine phosphate (diTris salt, pH 7.7), 25 g/ml of creatine kinase, 7 mM MgCl 2 , 0.5 mM DTT, 4 mM ATP, and G 1 extracts (40 g) as well as human cyclin A or E, as indicated. Following preincubation of G 1 extracts with cyclin A at 37°C for 20 min, exogenous underphosphorylated HSSB (1.0 g) was added and the mixture incubated for 60 min at 37°C. The phosphorylation state of HSSB was then analyzed by examining electrophoretic mobility changes of HSSB-p34 as detected by immunoblot analysis using monoclonal antibodies against p34.
Conditions for the phosphorylation of histone H1 and quantitation of H1 kinase activity were as previously reported (42).
For the phosphorylation of p107, three different sets of reactions were carried out and are referred to as samples I, II, and III (see Fig. 8A for a diagram of this experiment). Each reaction mixture (0.2 ml) contained 40 mM creatine phosphate (diTris salt, pH 7.7), 0.5 mM DTT, 25 g/ml creatine phosphokinase, 4 mM ATP, 7 mM MgCl 2 , 1 mg G 1 HeLa extract, and 10 g of cyclin A and were preincubated at 30°C for 60 min. One set of reactions (sample III) also contained 40 mM EDTA which acts to inhibit phosphorylation by chelating Mg 2ϩ ions required for kinase activity. GST-p107 (20 g) was added to the reactions referred to as samples II and III and GST (20 g) was added to sample I in a mixture (100 l) containing 4 mM PMSF, 11.2 g/ml antipain, 4.0 g/ml leupeptin, 20 mM Na 3 VO 4 , 210 mM NaF, and 2 M okadaic acid. The resulting mixture was further incubated for 60 min at 30°C. The samples I, II, and III were each loaded onto a glutathione-Sepharose column (0.7 ϫ 2.7 cm, 0.5 ml) equilibrated in buffer A plus 0.1 M NaCl. After washing with the equilibration buffer (10 ml), bound protein was eluted with buffer A plus 0.8 M NaCl (2.5 ml). These fractions were concentrated 10-fold using Amicon CF25 Centricon cones. Further bound protein was eluted with buffer A plus 0.1 M NaCl and 10 mM glutathione (2.5 ml). These fractions were dialyzed for 3 h at 0°C against 2 liters of buffer A plus 50 mM NaCl and concentrated 15-fold using Amicon cones. The proteins present in these fractions were visualized by both Coomassie staining and Western blot analyses with antibodies against cyclin A or p107 following SDS-PAGE.
In order to assay for the dephosphorylation of p107, the material eluted with glutathione (100 ng, 2 l) was added to reaction mixtures (10 l) containing 20 mM Hepes (pH 7.5), 7 mM MgCl 2 , 1 mM DTT, 0.1 M NaCl, and 0.05 unit of phosphatase PP2A (Upstate Biotechnology Inc.). After 20 min of incubation at 30°C, the phosphorylation state of p107 was analyzed by examining the alteration of its electrophoretic mobility by SDS-PAGE and Western blot analyses with polyclonal antibodies against p107.

Cyclin E, although Capable of Activating Cdk2 and Cdc2 in
Vitro, Cannot Support the Phosphorylation of HSSB, as Cyclin A Does-As both cyclins A and E are expressed during the G 1 /S transition phase of the cell cycle, either one or both of these cyclin-dependent kinase activities could be responsible for the phosphorylation of HSSB which occurs at the G 1 /S transition. In order to examine and compare the effects of these cyclins, as well as the D type cyclins, on the phosphorylation of HSSB, bacterial expressed GST-cyclins E, A, and D1-3 were solubilized and purified as described under "Materials and Methods." To ensure that these purified cyclins were active, they were examined for their ability to activate Cdk2 or Cdc2 kinases in the phosphorylation of histone H1. Cdk2 and Cdc2 were isolated from HeLa cell cytosolic extracts that harbor the Cdkactivating enzyme (CAK) that phosphorylates these kinases on residues Thr-160 or Thr-161, respectively (38). As shown in Fig. 1, whereas Cdk2 alone exhibited little H1 kinase activity, it was stimulated 350-fold following addition of cyclin A or 130-fold by the addition of cyclin E. Similar levels of stimulation were observed when Cdc2 was used in place of Cdk2 (data not shown). While reconstituted cyclin A or B/kinase complexes have been shown to phosphorylate p107 (42) and HSSB (33,34), the activity of cyclin E has not previously been reconstituted in vitro. Here we have shown for the first time that purified cyclin E is able to stimulate the activity of Cdc2 and Cdk2 in an in vitro reconstituted reaction.
The D type cyclins did not stimulate Cdk2 or Cdc2 kinase activities in the assays used here (data not shown), consistent with the notion that these cyclins form complexes with Cdk4 and Cdk6 which are either present at very low levels or not present in HeLa cell extracts.
Having shown that cyclin E was active in supporting histone H1 phosphorylation by Cdks, we then investigated whether this cyclin also stimulated the phosphorylation of HSSB. It has previously been shown in vitro that the p34 subunit of HSSB is phosphorylated by kinase activities present in G 1 extracts preincubated with cyclin A (33). This hyperphosphorylated HSSB product contained four species of p34 that migrated more slowly through denaturing polyacrylamide gels than the hypophosphorylated form. Following fractionation of the G 1 extracts, these hyperphosphorylated species of p34 were formed by the combined action of purified cyclin A-dependent Cdks and DNA-PK (33).
In order to ascertain whether cyclin E contributed to this phosphorylation we used similar assays. Addition of either GST-cyclin A or E stimulated the histone H1 kinase activity of G 1 extracts by ϳ4 -5-fold (from 2.7 pmol in the absence of cyclins to 12.5 or 14.2 pmol following cyclin A or E addition, respectively) indicating that active cyclin/kinase complexes were formed. The phosphorylation of the p34 subunit of HSSB by these cyclin E or A-activated G 1 extracts was then analyzed by examining the electrophoretic mobility changes of p34-HSSB using monoclonal antibodies against the p34 subunit. In the absence of cyclins, only limited amounts of the HSSB-p34 were converted into slow migrating forms (Fig. 2, lanes 1-3). However, following incubation for 60 min in the presence of cyclin A, nearly all of the p34 subunit had been converted into multiple slow migrating phosphorylated forms (lane 8), as reported (33). In contrast, incubation with cyclins D1, D2, D3, or cyclin E-activated G 1 extract resulted in conversion of limited amounts of p34 into slow migrating forms (lanes 4 -7, respectively). Furthermore, preincubation of cyclin A with any of cyclins D1, D2, or E did not significantly alter the cyclin A-dependent phosphorylated forms of HSSB observed (lanes 9 -11, respectively).
The specific cyclin-activated kinases present in the G 1 extracts following addition of each cyclin were examined. For this purpose cyclin A and E/kinase complexes were purified by selecting for proteins bound to each of the cyclins in the G 1 extracts by virtue of their GST fusions (as detailed under "Materials and Methods"). This procedure yielded both cyclin A-and E-specific Cdk2 kinase complexes (as shown in Fig. 3A). Cdk2 but not Cdc2 was detected in the cyclin/kinase complexes isolated from G 1 extracts. The mobilities of both cyclins A and E were retarded following incubation with G 1 extracts due to their phosphorylation. Calf intestinal phosphatase treatment converted the slower migrating forms of each of these cyclins into a single fast migrating, hypophosphorylated form (data not shown). Both the kinases responsible for phosphorylating these cyclins and the physiological significance are, however, unknown.
The isolated cyclin A-Cdk2 and cyclin E-Cdk2 complexes were examined for their ability to phosphorylate histone H1. As   FIG. 2. Cyclins A-, E-, D1-, D2-, or D3-dependent phosphorylation of HSSB-p34. As described under "Kinase Assays", reaction mixtures (10 l) containing 250 ng of GST-cyclins A, E, D1, D2, or D3, were preincubated with 50 g of G 1 extract for 30 min at 30°C; 1.0 g of exogenous underphosphorylated HSSB was added to each reaction followed by a further 60-min incubation at 37°C. An aliquot (2.0 l) of each reaction was subjected to immunoblot analysis using monoclonal antibodies against the p34 subunit of HSSB. In lanes 9 -11, 0.75 g of each of the cyclins E, D1, or D2 was preincubated with 250 ng of cyclin A and 50 g of G 1 extract prior to HSSB addition. The positions at which the unphosphorylated and phosphorylated species of HSSB-p34 migrate are indicated to the right.
FIG. 3. Immunoblot analysis of purified GST-cyclin/kinase complexes. G 1 extract (5 mg) was incubated with GST-cyclin A or GST-cyclin E (50 g each) for 20 min at 30°C, as described under "Purification of Cyclin A/Cdk2 and Cyclin E/Cdk2". The reaction mixtures (1 ml) were loaded onto 0.3 ml of glutathione-Sepharose columns equilibrated in buffer A plus 0.2 M NaCl. After washing with buffer A containing 0.8 M NaCl (10 ml), the column was eluted with buffer A plus 0.05 M NaCl and 20 mM glutathione. Aliquots (100 ng of protein) of both the material loaded onto each column (L), as well as the glutathione eluted material (E), were subjected to immunoblot analysis using antibodies against cyclin E, cyclin A, or Cdk2, as indicated.

FIG. 4. Phosphorylation of histone H1 and HSSB-p34 by cyclin A/Cdk2 or E/Cdk2
. Reactions were carried out as described under "Kinase Assays" and as previously published (42). Following separation of the reaction products by SDS-PAGE, the 32 P-phosphorylated forms of histone H1 or HSSB resulting from incubation with cyclin A/Cdk2 or cyclin E/Cdk2, were visualized by Coomassie Brilliant Blue, excised, and counted by liquid scintillation. Quantitation of the radiolabeled phosphate incorporated into histone H1 (A) and HSSB-p34 (B) is shown. indicated in Fig. 4A, the addition of 2.0 l of the cyclin E/Cdk2 eluate resulted in the incorporation of 15 pmol of phosphate into histone H1, while the addition of 1.0 l of the cyclin A/Cdk2 eluate resulted in a similar amount of histone H1 phosphorylation. To ascertain whether HSSB was also a substrate for cyclin E-dependent kinase activity, the amount of 32 P-labeled phosphate incorporated into the HSSB-p34 subunit by cyclin A/Cdk2 and cyclin E/Cdk2 was quantitated and compared. As shown in Fig. 4B, the addition of 1.0 l of cyclin A/Cdk2 (which catalyzed the incorporation of 15 pmol of 32 P into histone H1) resulted in the transfer of 400 fmol of 32 Plabeled phosphate into the HSSB-p34 subunit; the addition of 2.0 l of cyclin E/Cdk2 (which catalyzed the incorporation of 15 pmol of 32 P into histone H1) resulted in the incorporation of 13 fmol of 32 P into the HSSB substrate. Thus, cyclin E/Cdk2 phosphorylated HSSB-p34 about 30-fold less efficiently than cyclin A/Cdk2. Cyclin A/Cdk2, but Not DNA-PK, Efficiently Phosphorylates the HSSB-p34 Subunit Alone-The influence of the HSSB-p70 and -p11 subunits on the phosphorylation of the p34 subunit was examined. For this experiment, the HSSB-p34 subunit was cloned into a vector that overexpresses this protein in the absence of the other two subunits as a GST fusion protein in bacteria. As shown in Fig. 5A, lane 3, incubation of cyclin A-activated G 1 extracts with the HSSB-p34 subunit resulted in the extensive conversion of the p34 subunit into multiple slow migrating forms, similar to that seen with the trimeric HSSB (Fig. 2). This finding indicates that both the cyclin A-dependent kinase and DNA-PK present in these extracts efficiently phosphorylated the HSSB-p34 subunit in the absence of the p11 and p70 subunits. This experiment was repeated using both purified cyclin A/Cdk2 and DNA-PK. As indicated in Fig. 5B, the level of cyclin A/Cdk2 catalyzed 32 P incorporation into trimeric HSSB and the p34 subunit alone, were similar (0.65 and 0.5 pmol, respectively). Thus, HSSB-p34 phosphorylation by cyclin A/Cdk2 is independent of the p70 and p11 subunits. Purified DNA-PK catalyzed efficient phosphorylation of the p34 subunit in the presence of DNA (Fig. 5C, lane 10). However, in contrast to the efficient phosphorylation of the p34 subunit alone catalyzed by DNA-PK present in extracts, purified DNA-PK catalyzed phosphorylation of the HSSB-p34 subunit poorly in the absence of the other HSSB subunits (Fig. 5C, lanes 3 and 4; Fig.  5D, quantitation of results). Efficient phosphorylation of the HSSB-p34 subunit was restored following addition of the p34⅐p11 complex isolated after coexpression of these two subunits (Fig. 5C, lane 8). Thus a conformational change in the p34 subunit brought about by its interaction with the p11 subunit (with which it has been shown to interact, prior to the addition of the p70 subunit) (40,41) may result in a more suitable substrate for phosphorylation by DNA-PK.
Cyclin A, but Not Cyclin E, Interacts Specifically with HSSB-p34 -The results outlined above indicate that cyclin A/Cdk2 phosphorylates HSSB-p34 while cyclin E/Cdk2 is unable to catalyze this reaction. To test the hypothesis that this substrate specificity is dictated by the ability of a cyclin to directly interact with its substrate, we examined whether cyclin A directly interacted with the HSSB-p34 protein (see scheme in Fig. 6A).
As seen in Fig. 6B, cyclin A was retained on a GST-p34 column at salt concentrations up to 0.4 M NaCl, suggesting that there was an interaction between this cyclin and the GST-p34 subunit. Furthermore, this GST-p34-cyclin A complex was efficiently phosphorylated following addition of human Cdc2 (data not shown). In order to determine whether this interaction was specific, a control column was used in which the GST protein was coupled to glutathione-Sepharose beads. When FIG. 5. Phosphorylation of the p34 subunit of HSSB by cyclin A-activated G 1 extracts. A, as described under "Kinase Assays", 250 ng of cyclin A were preincubated with 40 g of G 1 extract. HSSB-p34 (1 g) was added followed by a further 60 min incubation. An aliquot of the reaction (2 l) was subjected to immunoblot analysis using monoclonal antibodies against the p34 subunit of HSSB (lane 3). Lane 1 shows the signal detected with the HSSB-p34 substrate, and lane 2 is a control that shows the effects of incubating G 1 extracts with HSSB-p34 in the absence of cyclin A. B, comparison of the phosphorylation of trimeric HSSB and the HSSB-p34 subunit by cyclin A/Cdk2. The cyclin A/Cdk2 complex was added (in amounts as indicated) to reactions (10 l) containing 40 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, and 50 M ATP with either 1 g of trimeric HSSB (9 pmol) or 300 ng of the HSSB-p34 subunit (9 pmol), as indicated. Following incubation for 45 min incubation at 37°C, the products were separated by SDS-PAGE and the 32 P-phosphorylated forms of p34 were visualized by Coomassie Brilliant Blue staining, excised, and counted by liquid scintillation. Quantitation of the results is shown. C, comparison of the DNA-PK catalyzed phosphorylation of trimeric HSSB and the HSSB-p34 subunit. Assay conditions used were as described under "Materials and Methods." Equivalent amounts of substrate were added, i.e. 0.3 g of HSSB (2.7 pmol), 79 ng of HSSB-p34 subunit (2.7 pmol), 29 ng of HSSB-p11 subunit (2.7 pmol), or 0.12 g of the HSSB-p34-p11 complex (2.7 pmol) in the presence or absence of DNA, as indicated. Quantitation of the results are summarized in D.
loaded onto this GST column, cyclin A was detected only in the flow through fraction (Fig. 6C, lane 2), suggesting that cyclin A interacts specifically with the p34 subunit of HSSB.
In contrast to the retention of cyclin A by the GST-p34, cyclin E was detected in the flow through fraction only when loaded onto the GST-p34-coupled glutathione-Sepharose beads (Fig.  6D, lane 2), suggesting that this cyclin does not form a complex with HSSB-p34. Thus, there appears to be a correlation between the ability of cyclin A/Cdk2 to phosphorylate HSSB and the ability of cyclin A to interact with this substrate suggesting that a specific interaction between these proteins may be necessary prior to HSSB phosphorylation.
Cyclin A Interacts Only with the Unphosphorylated Substrate (p107)-Having demonstrated that cyclin A directly interacts with HSSB prior to phosphorylation, we investigated whether this interaction was maintained following substrate phosphorylation or whether cyclin A disengaged after phosphorylation. For this purpose similar experiments to those detailed above were carried out with the tumor-suppressor protein p107 as the substrate for phosphorylation by cyclin A-activated G 1 extracts. This protein substrate was used since it was previously shown that cyclin A binds to p107 very tightly. The phosphorylation state of the p107 in these experiments, however, was unknown (47,48). The phosphorylation of p107 by cyclin A-activated G 1 extracts was analyzed by examining the electrophoretic mobility changes of GST-p107 using polyclonal antibodies against p107. Incubation of GST-p107 with G 1 extracts in the absence of cyclin A did not alter the migration of the protein (Fig. 7, compare lanes 1, 2, and 5). However, following a 60 min incubation in the presence of cyclin A-activated G 1 extracts, nearly all of the GST-p107 was converted into a slower migrating form (lane 6).
In order to isolate p107 in its phosphorylated or unphosphorylated form and examine whether these p107 species interact with cyclin A, GST-p107 was incubated with G 1 extracts in presence of cyclin A, with or without EDTA, a Mg 2ϩ chelating agent that acts to inhibit kinase activity. GST-p107 was reisolated after each reaction and its phosphorylation status as well as its association with cyclin A were analyzed as diagrammed in Fig. 8A. As shown in Fig. 8, B and D, GST-p107 eluted from the glutathione-Sepharose column exclusively in the phosphorylated slow migrating form following incubation with cyclin A-activated G 1 extracts in the absence of EDTA (Fig. 8, B, lane  7, and D, lane 3). However, in the presence of EDTA, reisolated GST-p107 migrated through SDS-polyacrylamide gels in the same position as the unphosphorylated starting material (Fig.  8, B, lane 8, and D, lane 4).
To confirm that the SDS-PAGE mobility alterations of p107 result from its phosphorylation, aliquots of the GST-p107 glutathione eluates were treated with phosphatase PP2A (Fig.  8E). Following phosphatase treatment, the slower migrating GST-p107 (Fig. 8E, lane 1) was converted to the faster migrating species (Fig. 8E, lane 2). Incubation of phosphatase PP2A with EDTA-treated GST-p107 did not alter its mobility (Fig.  8E, compare lanes 3 and 4). Thus, the change in mobility of p107 following incubation with cyclin A-activated G 1 extracts was due to the phosphorylation of p107, and the addition of EDTA inhibited this modification.
In order to investigate whether cyclin A was stably associated with p107 before and/or after p107 phosphorylation, the material isolated following incubation of GST-p107 with cyclin A-activated G 1 extracts was analyzed for protein by Coomassie staining (Fig. 8B) and for cyclin A by immunoblot analysis (Fig.  8C). Only limited amounts of cyclin A were retained on the glutathione columns from reactions carried out with either GST alone (Fig. 8C, lane 2) or reactions that generated phosphorylated GST-p107 (Fig. 8C, lane 3). In contrast, cyclin A present in the extracts was retained with unphosphorylated GST-p107 and eluted from the glutathione column with high salt (Fig. 8C, lane 4) or with 10 mM glutathione (Fig. 8C, lane  7). This suggests that cyclin A interacted with unphosphorylated p107, but following p107 phosphorylation, cyclin A was no longer bound. . Cyclin E (100 g) was loaded onto a 0.5-ml glutathione-Sepharose column coupled to 0.6 mg of GST-p34 (test column). The column was washed as above and aliquots of each fraction were subjected to SDS-PAGE and immunoblot analysis using anti-cyclin E antibodies (D) .   FIG. 7. p107 is hyperphosphorylated by cyclin A-activated G 1 extracts. Anti-p107 immunoblot, lanes 1 and 2, 100 ng of GST-p107 incubated in the absence and presence of cyclin A, respectively; lane 3, 1 l from the 200-l reaction mixture containing 1 mg of G 1 extract; lane 4, 1 l from the 200-l reaction mixture containing 10 g of cyclin A and 1 mg of G 1 extract; lanes 5 and 6, 1 l from the 200-l reaction mixture containing 20 g of GST-p107 and 1 mg of G 1 extract in the absence or presence of 10 g of cyclin A, respectively. The positions at which the phosphorylated and unphosphorylated species of p107 migrate are indicated to the right.

DISCUSSION
Reconstituted cyclin A or B/kinase complexes have been shown to phosphorylate p107 (42,52) and HSSB (33) but the activity of cyclin E and the D-type cyclins has not been reconstituted in vitro, and no substrates for these complexes have been identified. Results described here demonstrate that in vitro purified cyclin E is able to activate Cdk2 or Cdc2 kinases to phosphorylate histone H1 by 130-fold, which is 3-fold less than the observed stimulation by cyclin A/Cdk2. However, cyclin E was unable to stimulate the phosphorylation of HSSB or to affect cyclin A-dependent HSSB phosphorylation in vitro. Thus, cyclin A/Cdk2 is most likely the major cyclin-dependent kinase activity responsible for phosphorylation of this protein at the G 1 /S phase transition.
Furthermore, substrate structure may also determine the efficiency of cyclin/Cdk-mediated phosphorylation. We have shown that the cyclin A-dependent HSSB-p34 phosphorylation is independent of the p70 or p11 HSSB subunits using both cyclin A-activated G 1 extracts and purified cyclin A/Cdk2 preparations. Incubation with cyclin A-activated G 1 extracts results in the appearance of multiple slow migrating phosphorylated forms of HSSB-p34 due to the combined actions of cyclin A/Cdk and DNA-PK (33). Following incubation of cyclin A-activated extracts with either trimeric HSSB or the p34 subunit, all the slower migrating p34 bands that result from phosphorylation by both cyclin A/Cdk and DNA-PK were observed. This indicates that under these conditions the p11 and p70 subunits are not required for efficient phosphorylation by either the cyclin A/Cdk2 or DNA-PK present in these extracts. However, when purified DNA-PK was used in these experiments, a paradox FIG. 8. Isolation of phosphorylated and unphosphorylated p107 and analysis of its association with cyclin A. A, diagrammatic representation of the experiment detailed under "Kinase Assays". B, 7.5% SDS-polyacrylamide gel, Coomassie stained, lane 1, 100 ng of cyclin A; lane 2, 100 ng of GST-p107; lanes 3, 4, and 5, 100 ng of protein eluted with 0.8 M NaCl following passage of samples I, II, and III through glutathione-Sepharose columns, respectively; lanes 6, 7, and 8, 100 ng of protein eluted with glutathione following passage of samples I, II, and III through glutathione-Sepharose columns, respectively. Molecular mass markers are indicated to the left and the positions of the phosphorylated and unphosphorylated species of p107 and of cyclin A are indicated at the right. C, anti-cyclin A immunoblot, lane 1, 45 ng of cyclin A; lanes 3, 4, and 5, 100 ng of protein eluted with 0.8 M NaCl following passage of samples I, II, and III through glutathione sepharose columns, respectively; lanes 6, 7, and 8, 100 ng of protein eluted with glutathione following passage of samples I, II, and III through glutathione-Sepharose columns, respectively. The molecular mass marker is indicated to the right, and the position of cyclin A is indicated at the left. D, 7.5% SDS-polyacrylamide gel, silver stained, lane 1, 160 ng of GST-p107; lanes 2, 3, and 4, 100 ng of protein eluted with glutathione following passage of samples I, II, and III through glutathione-Sepharose columns, respectively. The positions of the phosphorylated and unphosphorylated species of p107 are indicated at the left. E, anti-p107 immunoblot, lanes 1 and 3, 100 ng of the glutathione eluates obtained following GST-p107 incubation with cyclin A-activated G 1 extracts in the absence or presence of EDTA, respectively. Lanes 2 and 4, 0.05 unit of phosphatase PP2A was added to 100 ng of the same glutathione eluates in the absence or presence of EDTA, respectively. The molecular mass marker is indicated at the right and the positions of the phosphorylated and unphosphorylated species of p107 are indicated to the left. was apparent. Purified DNA-PK, in the presence of Ku and DNA, can efficiently phosphorylate trimeric HSSB, but was unable to phosphorylate the HSSB-p34 subunit in the absence of the HSSB-p11 and -p70 subunits. Efficient phosphorylation of the HSSB-p34 subunit was observed when the p34 subunit was complexed with the p11 subunit. It is unclear why DNA-PK present in extracts is able to phosphorylate the HSSB-p34 subunit in the absence of the p11 and p70 subunits while purified DNA-PK is not. It is possible that a conformational change in the p34 subunit brought about through its interaction with the p11 subunit may be necessary in order for the HSSB-p34 to constitute a suitable substrate for phosphorylation by DNA-PK. The interaction of the HSSB-p34 with other proteins present when extracts are used may allow it to be in the correctly folded form for phosphorylation by DNA-PK in the absence of either the HSSB-p70 or -p11 subunits.
The increasing number of cyclin and Cdk family members identified indicates that multiple distinct cyclin-Cdk complexes may be assembled in vivo. The role of each of these complexes is thought to be the phosphorylation of a specific cellular substrate(s) but the identity of very few of these is known. During S phase, cyclin A forms a multiprotein complex with Cdk2, the transcription factor E2F, and a retinoblastoma gene product (pRB)-related protein, p107 (43)(44)(45)(46). Cyclin A has been shown to directly bind p107 (47,48), and when overexpressed, cyclin A/Cdk2 hyperphosphorylates pRB overcoming its ability to suppress proliferation (49). Cyclin E also forms a complex with p107 and E2F (50), while cyclin D1 has been shown to associate with Cdk4, proliferating cell nuclear antigen (a protein essential for replication and repair) and a kinase inhibitor p21 (51).
Experiments carried out in our laboratory (42) and by Peeper et al. (52) have illustrated that both the cyclin and the catalytic subunit are important factors in determining kinase substrate specificity. These authors have proposed that cyclin activated kinases target their cellular substrates through cyclin mediated protein-protein interactions. The demonstration that cyclin A and HSSB-p34 form a complex in the absence of kinase or ATP is consistent with this hypothesis. Cyclin E did not participate in such a protein/protein interaction and was unable to phosphorylate HSSB.
Using Sf9 cells coinfected with baculovirus vectors encoding the product of the human retinoblastoma gene (pRB) and murine D-type cyclins, Kato et al. (53) demonstrated that cyclins D2 and D3 bind pRB. However, introduction of a vector encoding Cdk4 induced pRB hyperphosphorylation and the dissociation of cyclins D2 and D3. Consistent with this, we have shown that cyclin A interacts with unphosphorylated p107, but following phosphorylation, the association between these two proteins was markedly reduced. Thus the cyclins may target Cdks to their substrates, bringing about substrate phosphorylation and destabilization of the complexes. In this way cyclins can be recycled and become available for interactions with other proteins.