p107 and p130 Associated Cyclin A Has Altered Substrate Specificity*

We demonstrate that p107 and p130 immune complexes exhibit kinase activity. We have tested such immune complexes with four substrates commonly utilized to assay Cdk activity, including all three known members of the retinoblastoma family. Immunodepletion revealed this kinase activity could be abolished by removal of either cyclin A or Cdk2 but was unaffected by removal of Cdk4 or any D-type cyclin. The appearance of p107 associated activity followed the accumulation of p107 protein. In contrast, the kinase activity associated with p130 immune complexes became apparent after mid-G1, coincident with p130 hyperphosphorylation. GST-Rb, GST-p107, and GST-p130 (where GST indicates glutathioneS-transferase) were equally suitable substrates in p107 and p130 immune complex kinase assays, yielding activity equal to 25% of the cyclin A activity present. The p107 and p130 associated activity was unable to phosphorylate histone H1, suggesting the p107 and p130 associated cyclin A/Cdk2 may represent a distinct pool with a distinct substrate specificity. The p107 and p130 associated activity was released from the immune complexes upon incubation with ATP and Mg2+ and exhibited the same substrate preference observed with the untreated immune complex. Our data suggest that p107 and p130 recognize, or form by association, a distinct pool of cyclin A/Cdk2 that preferentially phosphorylates retinoblastoma family members.

The retinoblastoma family of proteins includes three family members, pRb, p107, and p130. The major similarity among these proteins is a core region, termed the "pocket," consisting of two homologous domains, A and B, separated by a variable spacer region. The pocket region dictates interactions of Rb 1 family members with a variety of viral oncoproteins and cellular proteins, including members of the E2F transcription factor family (1), cyclin/Cdks (2)(3)(4)(5), and c-Myc (6).
pRb was the first protein identified in this group and, consequently, is characterized in more detail than p107 and p130. The importance of pRb in decisions concerning cell fate has been documented with biochemical and genetic evidence as a key element in a regulatory pathway that includes cyclin D1/ Cdk4 (7) and E2F (8,9). It is well established that Rb is frequently mutated in a variety of human cancers (10 -13); these loss of function alterations gave rise to the notion pRb functioned as a growth suppressor. Such a role has been substantiated through studies in which ectopic expression or microinjection of pRb resulted in growth arrest (14 -16). Rb-mediated growth inhibition occurs, at least partly, through sequestration of the E2F family of transcription factors, preventing transcriptional activation of E2F target genes, such as cyclin E (reviewed in Ref. 17).
p107 and p130 may also function in a manner similar to pRb, and all three share the ability to inhibit the growth of Rb Ϫ/Ϫ human osteosarcoma cells (18 -20). All of these proteins have been found to exist in multiple states of phosphorylation, which dictates their ability to interact with E2F targets and correlates with their ability to function as growth inhibitors (21)(22)(23)(24). Although the Rb family members all seem to regulate E2F, each member appears to have a distinct binding pattern with respect to different E2Fs (21,(25)(26)(27)(28), an idea that is consistent with the appearance of specific complexes during cell cycle traverse. p130 exhibits association with E2F-4 during quiescence, which is lost upon mitogenic stimulation as a result of p130 phosphorylation. In such stimulated cells, E2F-4 is predominantly associated with p107, as p107 is induced during late G 1 and S phase (20,21,25,(27)(28)(29). Unlike E2F-4, which appears to remain constant during cell cycle traverse (30), E2F-1 and -3 are not induced until cells approach G 1 /S (27,30). Although p107 and p130 associate with E2F-4 and -5 (20,30,31), pRB is found to predominantly associate with E2F-1-3 (20,26,31). The association of E2Fs with p107 and the ability of cyclin/Cdks to phosphorylate one or both of these members has been postulated to serve as the basis for continuously monitoring and regulating E2F-dependent transcriptional activity (32).
In this study we have demonstrated that cyclin A, which has been found to be the major cyclin partner for p107 (30 -33), is also the major cyclin partner for p130 in stimulated Balb/c 3T3 cells. Furthermore, we also demonstrated that although cyclin A immune complexes can phosphorylate both histones and GST-Rb very efficiently, the p107 and p130 associated kinase is unable to efficiently utilize histone H1 as a substrate. Therefore, we conclude that there may be distinct cyclin A complexes that exhibit substrate specificity that may be distinguished by their association with p107 or p130.
Chemicals and Reagents-Biological buffers, detergents, and inorganic molecules were purchased from Sigma or Fisher. Antibiotics and protein A-agarose beads were from Life Technologies, Inc.. PDGF B homodimer (PDGF-BB) was purchased from Biosource (Camarillo, CA). Nitrocellulose paper and reagents for SDS-polyacrylamide gel electrophoresis were purchased from Bio-Rad. Anti-rabbit horseradish peroxidase, enhanced chemiluminescence reagents, and radioisotopes were purchased from Amersham Corp. Autoradiographic film was purchased from Eastman Kodak Co.
Preparation of Cell Lysates and Western Blotting-Balb/c 3T3 fibroblasts were washed twice with phosphate-buffered saline (PBS), harvested by scraping, and concentrated by centrifugation, and the pellets were stored frozen. For analysis of suspension cultures, confluent 3-day-old primary mouse keratinocytes were refed 12-15 h before cells were induced to undergo differentiation using suspension in semi-solid medium in 50-ml flasks coated with 0.4% polyHEMA (35). Cells in semi-solid medium were diluted 10-fold with PBS (4°C) and pelleted by centrifugation. Frozen pellets were thawed on ice, resuspended in immunoprecipitation buffer (50 mM HEPES, pH 7.2, 150 mM NaCl, 0.1% Tween 20, 1 mM EDTA, 1 mM EGTA, 0.1 mM orthovanadate, 0.5 mM NaF, 0.1 mM phenylmethylsulfonyl fluoride, 2.5 g/ml leupeptin, and 1 mM DTT), and sonicated for 5 s. Following centrifugation to remove cellular debris, protein concentrations were determined by the Bradford assay. The crude cell extract (20 g) was boiled 3 min in Laemmli buffer and separated on 11% discontinuous SDS-polyacrylamide gels, and the separated proteins were transferred to nitrocellulose membrane by electrophoretic blotting. Nonspecific binding was prevented by blocking the membrane in BLOTTO (5% dry milk in 1 ϫ phosphate-buffered saline/Tween 20 (PBS-T; 1 ϫ PBS, 0.1% Tween 20) and incubated with the primary antibody (1:1000 dilution in PBS-T) for 1 h at room temperature. After washing in PBS-T, the membranes were incubated with anti-rabbit-horseradish peroxidase (1:10,000) for 1 h, washed, and visualized by enhanced chemiluminescence, as recommended by the supplier. To identify p107 in specific cyclin complexes, immunoprecipitation with appropriate antibodies was performed on 200 g of protein followed by p107 analysis with Western blotting as described above, except we utilized protein A-horseradish peroxidase (1:2000 dilution in PBS-T) as a secondary antibody. Cell lysates depleted from various cyclins and Cdks (Fig. 2) were prepared by conducting immunoprecipitation on naive cell extracts with 1-2 g of the appropriate antibody. Resulting immune complexes were harvested on protein A-agarose beads, and the supernatant fraction was removed and subjected to a second immunoprecipitation with p107 or p130 antibodies. We have previously demonstrated this protocol results in the removal of D1, D2, D3, and Cdk4 (36).
Kinase Assays-To measure p107, p130, and cyclin A associated histone H1 kinase activity, specific antisera to either p107, p130, or cyclin A were added to 30 g of protein from sonicated cell extracts, in a volume of 0.3 ml, and the mixture was rocked for 2-10 h at 4°C, followed by the addition of 20 l of protein A-agarose beads and an additional 30 min of mixing. Immune complexes were collected by centrifugation and washed twice with immunoprecipitation buffer, once with histone kinase buffer (50 mM Tris-Cl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT), and resuspended in histone kinase buffer containing 10 Ci of [ 32 P]ATP, 10 M cold ATP, and 100 g/ml histone H1 (Boehringer Mannheim). Reactions were incubated at 30°C for 5 min, mixed with an equal volume of loading buffer, heated to 95°C for 4 min, and separated on an 11% SDS-polyacrylamide gel. To measure p107, p130, and cyclin A associated Rb kinase activity, immunoprecipitations with specific antisera to either p107, p130, or cyclin A were performed as described above. Immune complexes were collected by centrifugation and washed twice with immunoprecipitation buffer, once with Rb kinase buffer (50 mM Tris-Cl, pH 7.5, 10 mM MgCl 2 , 5 mM MnCl 2 , 10 mM DTT), and resuspended in Rb kinase buffer containing 10 Ci of [ 32 P]ATP, 3 M cold ATP, and 0.5 g of GST-Rb, p107, or p130 (purified from bacterial sources and kindly provided by D. Cress, Moffitt Cancer Center). Reactions were incubated at 30°C for 6 -8 min, mixed with an equal volume of loading buffer, heated to 95°C for 4 min, and separated on a 11% SDS-polyacrylamide gel. Phosphorylated histone H1 and GST-Rb was visualized by autoradiography and quantitated by exposure on a Phos-phorImager (Molecular Dynamics).

RESULTS
The demonstration that p107 can be found in complexes containing cyclin A and E2F-4 (30 -33) and p130 associates with cyclins A and E (22,29,37) led us to examine if either p107 or p130 complexes exhibited kinase activity. We tested for this possibility by conducting kinase assays on p107 or p130 immune complexes using GST-Rb as a substrate, and we found immune complexes of both proteins exhibited kinase activity (Fig. 1A). Parallel immunoprecipitations with p107/p130-specific serum blocked with immunizing peptide did not yield complexes exhibiting kinase activity, demonstrating the activity was specific for p107 and p130 (Fig. 1A). To determine whether the p107/p130 associated activities displayed cell cycle-specific regulation, we examined immune complexes prepared from density-arrested cultures of Balb/c 3T3 fibroblasts FIG. 1. Induction of p107 and p130 expression and their associated kinase activities. A, density-arrested Balb/c 3T3 cells were stimulated with 10% serum and 10 ng/ml PDGF-BB. After stimulation for 24 h, extracts were analyzed for in vitro Rb kinase activity as described under "Materials and Methods." To determine if antibodies were specific, we added 3 l of anti-p107 or anti-p130 and 1 g of peptide p107 (sc-318) or p130 (sc-317, Santa Cruz Biotechnology) to 30 g of lysate and rocked for 1 h, and then proceeded with kinase assay. B, quiescent Balb/c 3T3 cells were stimulated to proliferate for various lengths of time. Cell extracts were assayed for the presence of p107 and p130 associated kinase activities utilizing glutathione S-transferase (GST)-Rb, p107, and p130 fusion proteins as substrates. C, quiescent Balb/c 3T3 cells were stimulated to proliferate for various lengths of time. Cell extracts were assayed for total cyclin A, p107, and p130 by Western blotting. harvested various times after mitogenic stimulation. Fig. 1B shows p130 associated kinase activity appears at 6 h and remains fairly constant thereafter. The activity profile obtained when GST-p107 and GST-p130 were used as substrates was similar to that obtained with GST-Rb, demonstrating this activity was not specific for Rb.
A similar cell cycle analysis of p107 associated kinase activity revealed a significantly different profile than the one obtained with p130 associated kinase activity (Fig. 1B). The p107 kinase activity was not detected until 18 h after stimulation and increased thereafter. The p107 associated activity, like the p130 associated activity, displayed the same pattern of activity using either Rb, p107, or p130 as substrates.
We quantified the relative level of p107 and p130 proteins at different times during cell cycle traverse to determine if the activity we observed could be explained by periodic fluctuations in the amount of protein (Fig. 1C). p107 protein was induced 12 h after stimulation, and by 24 h the protein level had increased about 5-fold, suggesting that p107 activity may be limited by p107 levels. In contrast, p130 protein is present throughout the cell cycle, but 12 h after stimulation the mobility of the protein is altered, probably due to phosphorylation. Thus, the appearance of p130 associated activity is not dependent upon the level of protein but appears to be related to its phosphorylation, which is first detectable between 0 and 6 h. This suggests perhaps that the phosphorylation status of p130 may regulate not only its interactions with transcription factor targets but also with components of the cyclin machinery. It may also be observed from these data that the level of p130 protein appears to decrease upon its phosphorylation, and this decreased level persists, as does its associated kinase activity.
Since multiple cyclin/Cdk associations have been identified for both p107 and p130, we performed immunodepletions to identify those cyclins responsible for the immune complex associated activity we observed. It is clear that depletion of cyclin A or Cdk2 resulted in quantitative removal of p130 associated activity ( Fig. 2A), whereas depletion of D-type cyclins, the Cdk inhibitors p27kip1 and p21Cip1, or Cdk4 had no discernible effect. This result is consistent with previous studies that demonstrated p130 is associated with cyclin A/Cdk2 (22,29,37).
We also identified the cyclin partner associated with p107 Rb kinase activity in a similar manner and found immunodepletion of cyclin A, but not D-type cyclins, resulted in the removal of p107 kinase activity (Fig. 2B). These results are consistent with several reports demonstrating the association of p107 and cyclin A (30 -33). The activity profile for the induction of p107 associated kinase, observed after mitogenic stimulation of density arrested Balb/c 3T3 cells, is almost identical to the cyclin A activity profile (Fig. 3A), and both proteins are induced with similar kinetics (Fig. 1C). To determine the amount of p107 that was associated with cyclin A, we immunoprecipitated cyclins A, D1-3, and E, separated the proteins within the immune complex by SDS-polyacrylamide gel electrophoresis, and determined the amount of p107 associated with the cyclins by Western blotting. As shown in Fig. 3B, approximately 25% of the total p107 is bound to cyclin A in extracts from proliferating fibroblasts, and a very small amount is bound to cyclins E and D3.
Cyclin A kinase activity assays were performed using either GST-Rb or histones as substrates, and these substrates showed similar induction of activity (Fig. 3A). When the kinase activity found in immune complexes of cyclin A, p107, and p130 was compared, utilizing GST-Rb as a substrate, p107 and p130 FIG. 2. p107 and p130 are associated with active cyclin A in proliferating cells. Density-arrested Balb/c 3T3 cells were stimulated with 10% serum and 10 ng/ml PDGF-BB. After stimulation for 24 h, extracts were analyzed for in vitro Rb kinase activity. To perform depletions, lysates were first subjected to immunoprecipitation ("Materials and Methods") with the indicated antibodies. After addition of protein A-agarose, immune complexes were harvested by centrifugation, and the supernatant fraction was removed to a new tube and subjected to a second immunoprecipitation with either anti-p107 or anti-p130, as indicated and then assayed for Rb kinase activity as described under "Materials and Methods." A, p130 associated kinase activity was measured as described above. B, p107 associated kinase activity was measured as described above.

FIG. 3. Cyclin A-Cdk2 can efficiently phosphorylate histones and Rb.
A, quiescent Balb/c 3T3 cells were stimulated to proliferate by the addition of 10% calf serum and 10 ng/ml PDGF-BB. Cell extracts were assayed for the presence of cyclin A associated kinase activity utilizing histones or GST-Rb as substrates. B, extracts from Balb/c 3T3 cells stimulated for 24 h were used to immunoprecipitate p107 and indicated cyclins. Immune complexes were resolved on a 8% SDSpolyacrylamide gel, transferred to nitrocellulose, and probed with antibodies to p107. contained 2-fold less Rb kinase activity than cyclin A (Fig. 4A). Since cyclin A was the major cyclin partner associated with both the p107 and p130 associated activities, we tested the ability of p107 and p130 immune complexes to phosphorylate histones. Interestingly, p107 and p130 had almost undetectable histone kinase activity, less than 2% of that exhibited by cyclin A immune complexes (Fig. 4A). We considered the possibility that the inability of p107 and p130 to phosphorylate histones was due to the association of one of the cyclin inhibitors. To test this hypothesis, we depleted p27kip1 and p21Cip1, using specific antibodies, and used the immunodepleted supernatants to isolate remaining p107 and p130 complexes for kinase assays. However, the p107 and p130 immune complexes thus obtained retained the ability to phosphorylate Rb, yet were unable to phosphorylate histones (data not shown).
The activity we observed with p107 and p130 immune complexes could result from either a stable interaction with cyclin A/Cdk2 or a transient association resulting from enzyme-substrate interactions. In the presence of divalent ion chelators, and the concomitant absence of ATP, the conditions under which cell lysis and subsequent manipulations were performed, catalysis might be inhibited, perhaps resulting in the "trapping" of bound substrates. Exposure to magnesium and ATP would be expected to release bound substrates but not stably associated proteins, allowing the discrimination of these two possibilities. Therefore, we assayed p107 immune complex associated activity before and after exposure to ATP, Mg 2ϩ , or both. As shown in Fig. 4B, the p107 associated kinase activity decreased significantly when the immunoprecipitates were exposed to ATP in the presence of divalent ions. The requirement for divalent ions was demonstrated by inclusion of EDTA (lane 2), which prevented the loss of activity from the immune complex. These results indicate that p107 associated kinase activity may have been released or inactivated by exposure to cofactors required for kinase activity. These possibilities could be distinguished since the former would result in the activity appearing in the supernatant of the ATP/Mg 2ϩ wash, while the latter would result in the absence of activity. From the data shown in Fig. 5A, it is clear that after immunoprecipitates have been preincubated with ATP and Mg 2ϩ , the majority of the activity is found in the supernatant (Fig. 5A). A similar result was also obtained with p130 immune complexes, although recovery of kinase activity was less efficient. To demonstrate that the released kinase was cyclin A/Cdk2, ATP and Mg 2ϩ supernatants from both p107 and p130 immune complexes were re-immunoprecipitated with anti-cyclin A or non-immune serum. The resulting immune complexes were then utilized in Rb kinase reactions, revealing that the activity released from p107 and p130 complexes was cyclin A/Cdk2 (Fig. 5C). This result is consistent with the hypothesis that p107 may be acting as a substrate for cyclin A although further studies are required to confirm this notion.
Thus, our data revealed the presence of kinase activity associated with p107 and p130 in proliferating fibroblasts. The relatively low level of such activity in extracts prepared from quiescent cells suggests these activities and/or their association with p107 and p130 must be reduced upon cell cycle exit. In this regard, we have previously found that when primary mouse keratinocytes were placed into suspension to induce differentiation, cyclin A, D1, and E kinase activity rapidly declines, and after 6 h these activities are almost undetectable (38). To determine if p107 and p130 associated kinase activity followed similar kinetics, we placed keratinocytes into suspension for various times and performed kinase assays with the resulting extracts. As shown in Fig. 5D, we found that both p107 and p130 associated kinase activity decreased about 2-fold after 3 h in suspension and was barely detectable after 6 h in suspension. In this experiment, adherent cells have growth potential while cells placed in suspension lose growth potential. Therefore, the p107 and p130 associated kinase activity displayed kinetics very similar to G 1 and S phase cyclin-associated kinases and were reduced upon cell cycle exit. Thus, these experiments demonstrate these activities are not restricted to cultured cell lines but are also present in primary epithelial cells. DISCUSSION The recognition of the interactions that occur between E2F family of transcription factors, Rb family members, and cyclin/ Cdk components has led to the current beliefs that this group of proteins plays a crucial role in rendering decisions during G 1 and G 1 /S that ultimately allow DNA synthesis and control proliferation (17). In this paper, we have presented evidence that demonstrates p107 and p130 may serve not only in a regulatory capacity to control activities of E2F members but may also regulate cyclin/Cdk activity through either modulation or direction of cyclin activity by acting as a repository for a specific segment of the cyclin/Cdk pools. This control appears to be Cdk2-specific and suggests that these proteins play a role not only during G 1 and at G 1 /S but perhaps during S phase as well.
Our data demonstrate, for the first time, that the association between p107 or p130 and cyclin A/Cdk2 contains active cyclin-Cdk complexes. A report describing kinase activity in p130 immune complexes has been recently published; however, the component(s) responsible for this activity were not identified. Immunodepletion of all G 1 cyclins revealed that removal of D-type cyclins did not have an effect on the p107 or p130 kinase activity. It is somewhat surprising that depletion of D-type cyclins did not affect the kinase activity of 107 or p130, since both contain the pocket region, which is known to interact specifically with D cyclin-Cdk4 complexes. It is possible that this association may be less stable than that with Cdk2 complexes. Preincubation with ATP in the presence of Mg 2ϩ releases or inactivates a component(s) required for phosphotransferase activity; the appearance of activity in the supernatant of such washes suggests the former. Since the activity released from p107 and p130 complexes can be recovered specifically with cyclin A antibodies, we believe the activity recovered in p107 and p130 complexes is due to the presence of cyclin A/Cdk2. However, we cannot rule out the possibility that an unidentified kinase controlled by cyclin A/Cdk2 may be involved in the Rb kinase activity we have measured. A more detailed study of the protein complex is therefore being conducted. Our observations are consistent with the idea that this association of the components in the p107 and p130 immune complex reflects a transient enzyme/substrate interaction. We have not mapped the regions of p107 or p130 that are involved in these interactions, but it has recently been demonstrated that cyclin-Cdk2 complexes interact with p107 in a region distinct from the pocket, between residues 655 and 666 (33). Interestingly, a similar sequence is found in p130 between residues 633 and 644.
The p107 protein and associated kinase activity is induced during the same time frame that cyclin A protein and its associated kinase activity is induced. We have found that approximately 25% of the total p107 can be recovered in cyclin A immune complexes. The remaining p107 that can be obtained in p107 immune complexes from cyclin A-depleted extracts has no detectable kinase activity, thus the remaining p107 must be free of at least cyclin A-Cdk2 complexes. We do not yet know the relative abundance of other proteins known to bind p107, such as E2F and p21Cip1; however, it is certainly possible to conclude from the data we have presented that p107 may exist in distinct pools, perhaps with distinct regulatory functions. The p130 associated kinase, on the other hand, is activated 6 h after stimulation. Since cyclin A is not present at this time, the activity probably stems from the association of cyclin D3 with p130, which has been previously demonstrated. 2 However, the kinase activity from cells during S phase is clearly associated with Cdk2 complexes and not depleted by D-type cyclins. It is possible, therefore, that the kinase activity associated with p130 could come from multiple cyclin/Cdk pairs, which are determined in a cell cycle-dependent manner.
Cdk2 complexes formed with either cyclin E or cyclin A have been found to efficiently phosphorylate either histone H1 or Rb protein, in vitro (39 -42). It is clear from our work, however, that these specificities may be the aggregate activity of distinct cyclin A-Cdk2 complexes. This conclusion is based on our demonstration that histones are unable to serve as credible substrates for either the p107 or p130 associated activity. Therefore, it is likely that either cyclin A specificity is altered when p107 or p130 are associated with the complex or that these two proteins specifically interact with a subpopulation of Cdk2 complexes that do not have histone H1 kinase activity. We do not yet understand how such a shift in substrate specificity occurs. If cyclin A exists in more than one distinct complex, it is possible that in a subset of the complexes, the p107-or p130containing portion, one of the associated proteins may confer the ability to exclude certain substrates like histones, perhaps 2 F. Dong and W. J. Pledger, manuscript submitted for publication.
FIG. 5. p107 and p130 associated kinase activity appears in the supernatant fraction after preincubation with cold ATP. A, after stimulation for 24 h, Balb/c 3T3 cell extracts were assayed for the presence of p107 or p130 associated kinase activities utilizing GST-Rb as a substrate. p107 and p130 immunoprecipitates from stimulated cell extracts were preincubated with 20 l of Rb kinase buffer containing 2 M cold ATP for 5 min at 30°C, then 0.5 g GST-Rb, and 10 Ci of [ 32 P]ATP was added (Total). Parallel p107 and p130 immune complexes were preincubated as above and separated by centrifugation. The supernatant was carefully removed with a 0.2-mm pipette tip and transferred to a fresh tube. 20 l of Rb kinase buffer containing 2 M cold ATP, 0.5 g of GST-Rb, and 10 Ci of [ 32 P]ATP was added to the pellet (Pellet) and 0.5 g of GST-Rb and 10 Ci of [ 32 P]ATP (in 2 l) was added to the supernatant (Supernatant). A third parallel immune complex, after preincubation and removal of supernatant, was washed once with 750 l of Rb kinase buffer lacking ATP to remove the residual liquid remaining in the pellet. The washed pellet was then assayed with Rb kinase buffer containing 2 M cold ATP, 0.5 g of GST-Rb, and 10 Ci of [ 32 P]ATP. B, the gel autoradiographed and shown in A was also exposed and visualized on a Molecular Dynamics PhosphorImager to quantify the relative activities. The values were normalized to the activity obtained for the reactions labeled Total. C, supernatants generated from p107, p130, or non-immune complexes (as indicated above individual lanes) were prepared as described in A. The supernatant mixtures were assayed for Rb kinase activity after immunoprecipitation (IP) with anti-cyclin A, non-immune serum, or without further treatment (None), as noted above individual panels. The phosphorylation reactions were separated by SDS-polyacrylamide gel electrophoresis and visualized on a Molecular Dynamics PhosphorImager. D, confluent 3-day-old primary mouse keratinocytes were trypsinized and harvested (Plate) or placed into suspension in semi-solid medium for various times to undergo differentiation. Cell extracts were assayed for the presence p107 and p130 associated kinase activity utilizing GST-Rb as a substrate. as a result of steric hindrance. The kinase activity released from p107 and p130 immune complexes by incubation in the presence of ATP and magnesium, however, exhibits a substrate preference that is identical to the untreated immune complex. 3 Thus, if the differential substrate specificity is the result of steric hindrance, the excluding factor must also be released in the ATP/Mg 2ϩ wash. Definitively ruling out this possibility awaits more defined studies with well characterized Cdk2 substrates, which are less likely to be subject to such a problem. At first glance, our findings would seem to contradict those of Mayol et al. (23), who demonstrated that p130 associated complexes could phosphorylate histones. We also found that p107 and p130 could phosphorylate histones, but the activity is 50-fold less when compared with cyclin A/Cdk2, whereas p107 and p130 phosphorylate GST-Rb almost as well as cyclin A/Cdk2. We note two recently published studies that have revealed parallel observations made for CAK activity (43,44). Both studies demonstrated a marked substrate preference for Cdk7cyclin H complexes that depended on whether it was assayed as a "free" complex or bound to TFIIH, a multi-subunit basal transcription factor containing Cdk7 as one of its subunits. The TFIIH bound CAK exhibited a preference for an RNA polymerase II C-terminal domain-derived substrate, whereas free CAK exhibited a preference for Cdk2. These reports and our study demonstrate that Cdk activity may be regulated by substrate binding and illustrate how Cdk substrates may participate in dynamically modulating Cdk activity.
According to current models, G 1 and G 1 /S control is mediated through interactions between Rb family members and E2Fs, which in turn is regulated by Cdk activity. An important aspect of this circuit is a release of E2F-4 from p130 as cells leave quiescence and undergo G 1 traverse. We have demonstrated that the association of p130 with an active kinase begins at 6 h after stimulation and persists, a pattern that is consistent with the appearance of the hyperphosphorylated form. The p107 protein and its associated kinase activity, in contrast, does not appear until S phase, at which time it also associates with members of the E2F family (20,21,25,(27)(28)(29). It has been suggested that one function served by independent binding domains for E2F and cyclins may be to cause dissociation of the E2F from p107 under conditions that support continued proliferation (33). We have considered the possibility that E2F immune complexes may also bind cyclin-Cdk complexes utilizing p107 as a bridging molecule; however, we found a very low level of associated activity. 3 Our data reveal that the interactions that have been found between cyclin/Cdks and at least one important physiological class of substrates, members of the Rb family, are sufficiently stable to allow isolation of active kinase activity from p107 and p130 complexes. Although all the data we have presented is consistent with the notion that the kinase activity is isolated as a "trapped" enzyme/substrate interaction, we cannot conclusively exclude the possibility that p130 or p107 are integral members of the active complex and, as we have shown, modify the cyclin A/Cdk2 activity. Nonetheless, further examination of the activity before and after release from the immune complex and characterization of the proteins that are involved in the activities we have observed, as compared with those present in the remaining cyclin/Cdk pool(s), should allow interesting studies on the means through which substrate specificity is controlled.