p27^Kip1 Inhibits Cyclin D-Cyclin-Dependent Kinase 4 by Two Independent Modes

Abstract

Cell cycle progression is regulated by cyclin-dependent kinases (cdk's), which in turn are regulated by their interactions with stoichiometric inhibitors, such as p27^Kip1. Although p27 associates with cyclin D-cyclin-dependent kinase 4 (cdk4) constitutively, whether or not it inhibits this complex is dependent on the absence or presence of a specific tyrosine phosphorylation that converts p27 from a bound inhibitor to a bound noninhibitor under different growth conditions. This phosphorylation occurs within the 3-10 helix of p27 and may dislodge the helix from cdk4's active site to allow ATP binding. Here we show that the interaction of nonphosphorylated p27 with cdk4 also prevents the activating phosphorylation of the T-loop by cyclin H-cdk7, the cdk-activating kinase (CAK). Even though the cyclin H-cdk7 complex is present and active in contact-arrested cells, p27's association with cyclin D-cdk4 prevents T-loop phosphorylation. When p27 is tyrosine phosphorylated in proliferating cells or in vitro with the tyrosine Y kinase Abl, phosphorylation of cdk4 by cyclin H-cdk7 is permitted, even without dissociation of p27. This suggests that upon release from the contact-arrested state, a temporal order for the reactivation of inactive p27-cyclin D-cdk4 complexes must exist: p27 must be Y phosphorylated first, directly permitting cyclin H-cdk7 phosphorylation of residue T172 and the consequent restoration of kinase activity. The non-Y-phosphorylated p27-cyclin D-cdk4 complex could be phosphorylated by purified Csk1, a single-subunit CAK from fission yeast, but was still inactive due to p27's occlusion of the active site. Thus, the two modes by which p27 inhibits cyclin D-cdk4 are independent and may reinforce one another to inhibit kinase activity in contact-arrested cells, while maintaining a reservoir of preformed complex that can be activated rapidly upon cell cycle reentry.

Cyclin-cyclin-dependent kinase (cyclin-cdk) complexes drive progression through the different phases of the cell cycle by acquiring catalytic activity only at specific points (29, 36). These serine/threonine kinases phosphorylate the substrates that promote these transitions, and therefore, their activity must be tightly regulated to ensure orderly cell cycle progression. Cyclin-dependent kinase 4 (cdk4) and its homologue cdk6 serve as regulators of early G₁ and appear particularly important in the G₀-to-G₁ transition. Multiple steps are required for the activation of these kinases. cdk4 and cdk6 are catalytically inactive unless they partner with one of three cyclin monomers, D1, D2, or D3. Unlike other cyclins (cyclins A, E, and B) whose levels oscillate during the cell cycle, cyclin D levels are more constant but depend on the presence of mitogens. Cyclin D is localized in the nucleus only during the G₁ phase, thus preventing inappropriate activation of this complex (19). However, cyclin D and cdk4 do not readily assemble and appear to need a mitogen-dependent assembly factor to stabilize the complex (12). The cdk inhibitors p27^Kip1 and p21^Cip1 have been implicated in this role, although other factors may be able to compensate in their absence (5, 11, 25, 38). Cyclin D does not possess an obvious nuclear localization signal, and it is translocated into the nucleus primarily by its association with p27 or p21 (3).

Even the assembled, nuclear cyclin D-cdk4 complex requires further activation by phosphorylation on residue T172 by a cdk-activating kinase (CAK). In mammalian cells, CAK is itself a complex composed of a catalytic subunit (cdk7), a regulatory subunit (cyclin H), and the RING finger protein MAT1 (reviewed in reference 17). CAK phosphorylates the T-loops of multiple cdk's, but it is also a subunit of transcription factor TFIIH that phosphorylates the C-terminal domain of the large subunit of RNA polymerase II (17). CAK appears to be a constitutively expressed, nuclear holoenzyme, whose activity is not cell cycle regulated in an obvious way.

Both cyclin binding and CAK-mediated phosphorylation of the cdk subunit alter the three-dimensional structure of the cyclin-cdk complex. Cyclin A binding to cdk2 moves the T-loop from the “closed” conformation to the “open” conformation in which the T-loop becomes more accessible to solvent (32). Phosphorylation by CAK moves the T-loop further, stabilizing its structure (34) and widening the catalytic cleft. The three-dimensional structure of cyclin D-cdk4 has not been solved, but given the homology between cdk2 and cdk4/6 in this region, similar conformational changes might occur upon CAK-mediated phosphorylation of cdk4 or cdk6. T-loop phosphorylation of cdk4 and cdk6 has been demonstrated in vitro and in vivo, and mutation of residue T172 in cdk4 or T177 in cdk6 has been shown to render either kinase inactive (4, 7, 9, 16, 23, 24, 30, 31).

p27^Kip1 is expressed throughout most of the cell cycle, although its levels dramatically increase in response to certain antiproliferative signals, such as contact inhibition or serum starvation (14). Its levels decrease as cells exit the quiescent state, due to modulation of its ubiquitin-mediated degradation. Multiple regulatory phosphorylations of p27 are detected in vivo, including those that affect its cellular localization, ubiquitination, and cdk inhibitory activity (14).

While p27 has been implicated in both cyclin D-cdk4 assembly and nuclear localization, it may also directly regulate the catalytic activity of the complex (6). p27 is associated with cyclin D-cdk4/6 complexes both in growth-arrested cells, where it is responsible for inhibiting the complex, and in proliferating cells, where it appears to associate with the complex in a noninhibitory manner (22). Recently, we demonstrated that p27 could be either a cyclin D-cdk4 inhibitor or a noninhibitor, depending on the absence or presence, respectively, of Y phosphorylation in its 3-10 helix (residue Y88 or Y89) (22). Phosphorylation on residue Y88 or Y89 may prevent productive interaction between the 3-10 helix and the cdk subunit, pushing the C-terminal tail of p27 out of the active site of the kinase (13, 20). As p27 can bind to the cyclin subunit as well, p27 would remain associated with cyclin D-cdk4, permitting kinase activity while stabilizing the intrinsically weak complex. p27 is preferentially phosphorylated in proliferating cells with high levels of tyrosine kinase activity. An attractive model therefore postulates that p27 is not Y phosphorylated in contact-arrested cells due to the lack of tyrosine kinase activity, and this nonphosphorylated p27 inhibits cdk4 activity (6). Upon release from quiescence, p27 becomes Y phosphorylated, converting it to a bound noninhibitor. The tyrosine kinase responsible for this transition has not been determined, but several Y kinases known to be activated in proliferating cells, such as Abl, Lyn, or Src, appear to phosphorylate p27 in vitro or in vivo (13, 20, 22).

Others have suggested that, in addition to blocking cdk4's active site, p27 might inhibit cyclin D-cdk4 by preventing the activating phosphorylation on T172 (8, 9, 23). We investigated p27's ability to prevent CAK-mediated phosphorylation in proliferating or contact-arrested cells. Our data suggest that when Y-phosphorylated p27 associates with cdk4, it permits both phosphorylation by CAK and access by ATP to cdk4's active site. Loss of Y phosphorylation converts p27 into a cyclin D-cdk4 inhibitor, preventing both T-loop phosphorylation and ATP binding. Our data suggest that the two modes by which p27 blocks kinase activation are independent, potentially reinforcing mechanisms that stably inhibit cdk4 in contact-arrested cells, while maintaining a reservoir of cdk4 complexes that can be rapidly activated upon cell cycle reentry.

MATERIALS AND METHODS

Cell culture.

Mv1Lu cells were maintained in minimal essential medium supplemented with 10% fetal bovine serum. The Tetp27, TetHisp27, TetY74F, TetY88F, and TetY89F cell lines were previously described (7, 22). All Tet lines were maintained in minimal essential medium supplemented with 10% fetal bovine serum plus 0.5 mg/ml G418, 0.3 mg/ml hygromycin, and 1 μg/ml tetracycline. Asynchronously growing cells (A cells) were harvested from plates that were no greater than 60% confluent. Contact-arrested cells (G₀ cells) were harvested 5 days after visible contact arrest. Complete medium was replaced every other day, and fluorescence-activated cell sorting analysis confirmed that cells had a >95% G₁ content as previously described (22). Immunoprecipitation, immunoblot, and in vitro kinase assays were performed as described previously (7).

Antibodies.

Antibodies used in this study were as follows: anti-mouse p27, anti-mink cdk4, and anti-mouse cdk2 were a generous gift from J. Massagué. Anti-cyclin D1 (AHF0092) was from Biosource International. Anti-cyclin D1 (Sc-177), anti-cdk6 (Sc-7961), anti-cdk2 (Sc-6428), and anti-cdk7 (Sc-529) were from Santa Cruz Biotechnology. Anti-phospho-cdk2 T160 (2561S) was from Cell signaling. Antiactin antibody (A 2066) was from Sigma-Aldrich. PhosphoY89p27 antibodies were generated by immunization of rabbits with a phospho-specific p27 peptide (Invitrogen). Negative- and positive-affinity chromatography with nonphosphorylated and phosphorylated peptides, respectively, were performed to purify the antibody.

Generation of recombinant p27 and cyclin D-cdk4 source materials.

Purified, bacterial Hisp27 was generated as described previously (7). To generate Abl*p27, recombinant Hisp27 was incubated with Abl kinase buffer (NEB) and 400, 800, or 1,600 units of Abl kinase (NEB) for 1 h at room temperature. Hisp27 was recovered by metal agarose chromatography with Talon beads (BD Biosciences). Abl* p27 and mock*p27 were treated by incubating samples in 1× protein tyrosine phosphatase buffer (25 mM Tris-HCl, 50 mM NaCl, 2 mM Na₂EDTA, 5 mM dithiothreitol [DTT], 0.01% Brij 35) and 20 U of recombinant human T-cell protein tyrosine phosphatase (NEB, Calbiochem) at 30°C for 30 min. Recombinant His-cyclin D1-cdk4 was purified from High5 cells using metal agarose chromatography as described previously (22). To additionally purify cyclin D-cdk4 (form 3), the His-cyclin D1-cdk4 complex was passaged over a Superdex 200 gel filtration column (AKTA, GE Healthsciences), fractions were collected, and immunoprecipitation-immunoblot analysis and kinase assay were performed to detect the formation of complexes and catalytic activities of the fractions.

Generation and treatment of p27 from Mv1Lu cells.

AHisp27 and G0Hisp27 were purified from A and G₀ TetHisp27 cells in the presence of urea as previously described (22). To generate potato acid phosphatase (PAP)-Ap27 and PAP-G0p27, purified AHisp27 and G0Hisp27 were incubated in 100 μl of 50 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) (pH 6.0) and 1 mM DTT for 10 min at 30°C. Ten units of PAP (Roche) was added and incubated for 15 min at 30°C, followed by repurification on Talon beads (BD Biosciences) in the presence of urea and standardization by immunoblot analysis using recombinant p27 as a control.

CAK assay.

Recombinant cyclin H-cdk7 and Csk1 were purified from Sf9 insect cells as described previously (18). Briefly, baculoviruses expressing cyclin H or cdk7 (coinfection) or Csk1 (single infection) were used to infect Sf9 cells. Cells were harvested, centrifuged, and resuspended in 10 mM HEPES (pH 7.4) and 10 mM NaCl, followed by lysis in a glass Dounce homogenizer. The buffer was adjusted to contain 150 mM NaCl and 10 mM imidazole. A nickel-agarose column (Qiagen) was equilibrated in 25 mM HEPES (pH 7.4), 150 mM NaCl, and 10 mM imidazole. Lysate was loaded onto the column and washed with 10 column volumes of 25 mM HEPES (pH 7.4), 150 mM NaCl, and 10 mM imidazole. The protein was eluted with 25 mM HEPES (pH 7.4), 150 mM NaCl, and 200 mM imidazole. Fractions were collected, analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and pooled, and dialysis of the pooled protein was done against 25 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM DTT, and 10% glycerol.

To visualize direct cdk4 phosphorylation, immunoprecipitates or recombinant cyclin D-cdk4-p27 complexes were incubated in 50 μl of CAK buffer (50 mM HEPES [pH 7.4], 80 mM β-glycerophosphate, 20 mM EGTA, 15 mM MgCl₂, 5 mM DTT, 1 μg/ml aprotinin, 1 μg/ml leupeptin) with [γ-³²P]ATP (Amersham), 0.2 mM ATP, and 1 μg of cyclin H-cdk7 or 1 μg of Csk1 (or different concentrations as noted). Treated complexes were either directly immunoprecipitated with cdk antibodies or boiled in 1% SDS at 95°C for 5 min before increasing the volume to reduce the concentration of SDS to 0.1% and immunoprecipitating. To detect retinoblastoma (Rb) kinase activity, immunoprecipitates or recombinant cyclin D1-cdk4-p27 complexes were incubated in CAK buffer and nonradiolabeled ATP and either immunoprecipitated with cyclin D1, cdk4, cdk2, or p27 antibodies or assayed directly.

To determine the number of picomoles of p27 and cdk4 phosphorylated (see Fig. 6F), a standard graph plot of the picomoles of γ-³²P (x axis) versus the cpm (y axis) was generated. γ-³²P was diluted 1:100,000, 1:10,000, 1:1,000, and 1:100 and spotted onto Whatman filter paper. The filter paper was exposed to a phosphorimager screen and analyzed by phosphorimaging (Molecular Dynamics). Based on the number of curies/millimole, the number of picomoles of γ-³²P was determined that corresponded to the cpm counts. The standard equation was y = 4.0556x, with a R² value of 0.9872. Purified p27 was incubated with recombinant cyclin D1-cdk4 and immunoprecipitated with p27 antibody. Increasing amounts of Abl were added along with [γ-³²P]ATP for an hour at 37°C. The complex was washed thoroughly and then incubated with a constant amount of cyclin H-cdk7 and [γ-³²P]ATP for 1 h at room temperature. Samples are washed thoroughly and analyzed on a 14% gel to resolve the bands by SDS-PAGE. The cpm of phosphorylated p27 or cdk4 (y axis) and their corresponding picomole values (x axis) were determined from the standard equation.

FIG. 6.

Phosphorylation of p27 permits cyclin H-cdk7 phosphorylation of p27-cyclin D-cdk4. (A) Recombinant p27 was incubated with (+) or without (−) Abl kinase and then treated with (+) or without (−) protein tyrosine phosphatase (PTP) to detect tyrosine-phosphorylated p27 (p27*) by autoradiography as well as by immunoblot analysis with p27 antibodies (W). (B) (Top) No p27 (lanes 1 to 3), Abl-phosphorylated p27 (Abl*p27) (lanes 4 to 6), or mock phosphorylated p27 (mock*p27) (lanes 7 to 9) were incubated with cyclin D-cdk4 in the presence or absence of cyclin H-cdk7 or Csk1 and [γ-³²P]ATP. This material was reimmunoprecipitated with cdk4 antibodies to detect cdk4 phosphorylation (cdk4*) (lanes 1 to 9). Ip, immunoprecipitate. For the cdk4-associated Rb kinase panel, the treated p27-cyclin D-cdk4 complexes were incubated in the presence or absence of cyclin H-cdk7 or Csk1 and nonradiolabeled ATP, reimmunoprecipitated with cdk4 antibodies, and used in in vitro kinase assays (Rb*) (lanes 1 to 9). For the bottom two panels, the treated p27-cyclin D-cdk4 complexes were immunoprecipitated (Ip) with cdk4 antibodies to detect p27 phosphorylation (p27*) or total p27 by immunoblot analysis with p27 antibodies (W) (lanes 1 to 9). In lane 10, the p27-cyclin D1-cdk4 complexes were immunoprecipitated with normal rabbit serum (NRS) as a control (lane 10). (C) Recombinant cyclin D-cdk4 was incubated with Abl*p27 and 1 μg of cyclin H-cdk7 (lane 1) or p27 and 1, 2, or 4 μg of cyclin H-cdk7 (lanes 2, 3, and 4, respectively) in the presence of [γ-³²P]ATP and immunoprecipitated (Ip) with cdk4 antibodies to detect cdk4 phosphorylation (cdk4*). (D) Cyclin D-cdk4 was incubated in the absence (−) or presence (+) of p27 followed by immunoprecipitation with cdk4 antibodies (Ip cdk4). The cdk4 immunoprecipitates were incubated with (+) or without (−) Abl kinase in the presence of [γ-³²P]ATP to detect p27 phosphorylation (p27*) (bottom) and total associated p27 by immunoblot analyses with p27 antibodies (top). (E) Cyclin D-cdk4 was incubated in the presence (+) or absence (−) of p27 to form p27-cyclin D-cdk4 complex. Abl kinase was added to some of the complexes (+). (Top) The mock-treated and Abl-treated p27-cyclin D-cdk4 complexes were incubated in the presence (+) or absence (−) of cyclin H-cdk7 (cdk7) and [γ-³²P]ATP to detect cdk4 phosphorylation (cdk4*). (Bottom) The mock-treated or Abl-treated p27-cyclin D-cdk4 complexes were incubated in the presence (+) or absence (−) of cyclin H-cdk7 and nonradiolabeled ATP and used in in vitro kinase assays (Rb*). The first step was to immunoprecipitate cdk4 (Ip cdk4), the second step was to reimmunoprecipitate cdk4 (Re-Ip cdk4). (F) p27 was incubated with cyclin D1-cdk4, followed by immunoprecipitation with p27 antibodies. Increasing units of Abl were added with [γ-³²P]ATP and additionally incubated. The complex was washed thoroughly, and a constant amount of cyclin H-cdk7 was added along with additional [γ-³²P]ATP. The complex was washed and analyzed by SDS-PAGE, before analysis using a phosphorimager to determine the numbers of incorporated cpm and picomoles. The positions of radiolabeled p27-associated cdk4 and p27 (cdk4* and p27*, respectively) are shown to the left of the gel. p27 immunoprecipitates were also analyzed by direct cdk4 immunoblotting (W). (Middle top) The number of units of Abl kinase was plotted against the number of picomoles of phosphorylated p27 (p27*). (Right top) The number of picomoles of phosphorylated p27 (p27*) was plotted against the number of picomoles of phosphorylated of cdk4 (cdk4*). (Middle bottom) Standard plot of the number of picomoles of γ-³²P (x axis) versus the number of cpm (y axis) determined as described. The cpm and pmol values are listed in the table at the right.

Immunofluorescence.

A and G₀ cells were grown on coverslips. The cells were treated for 15 min with 2 ml of cold 4% paraformaldehyde-phosphate-buffered saline (PBS) (pH 7.4) and then washed three times with room temperature PBS. The cells were blocked for 1 h at room temperature with a 0.2% Triton X-100-5% bovine serum albumin (BSA) solution in PBS. Cells were incubated with primary antibodies against p27 or phosphoY89p27 in 0.2% Triton-1% BSA-PBS for 2 h at room temperature. Three washes in PBS were performed, followed by incubation with a fluorescence-labeled secondary antibody at a 1:1,000 dilution in 0.1% Triton X-100-1% BSA in PBS in a light-protected humidified chamber for 1 h. After one wash in PBS, the nuclei were stained using ToPro (Molecular Probe) Coverslips were mounted on a rectangular glass slide with fluoromount solution and visualized by using a confocal microscope.

RESULTS

Cyclin H-cdk7 is present and active in G₀ cells.

We and others have previously demonstrated that when Mv1Lu cells are grown to confluence, they arrest in the G₁ phase of the cell cycle, p27 levels increase approximately 10-fold, and p27-cyclin D-cdk4 complexes remain assembled but are catalytically inactive (7, 22, 33). Immunoblot analysis demonstrated that endogenous cdk7, the catalytic subunit of CAK, was present in both proliferating cells (A cells) and contact-arrested cells (G₀ cells) (Fig. 1A). cdk7 immunoprecipitated from either A or G₀ cells was able to phosphorylate recombinant cyclin D1-cdk4 complexes, purified from insect cells coinfected with baculoviruses expressing cyclin D1 and cdk4 (Fig. 1B, lanes 2 and 4). This demonstrated that endogenous cdk7 was active as a CAK under both A and G₀ conditions.

FIG. 1.

CAK is present in G₀ cells and able to phosphorylate recombinant cyclin D-cdk4 in vitro. (A) A and G₀ lysates were directly immunoblotted with cdk7 antibodies (W). Actin was used as a loading control. (B) A and G₀ lysates were immunoprecipitated with normal rabbit serum (NRS) (lane 5) or immunoprecipitated (Ip) with cdk7 antibodies to isolate A (lanes 1 and 2) or G₀ (lanes 3 and 4) CAK. These immunocomplexes were added to purified recombinant cyclin D-cdk4 (DK4) (lanes 2, 4, and 5) in the presence of [γ-³²P]ATP before cdk4 was isolated by immunoprecipitation with anti-cdk4 antibodies.

Cyclin D-cdk4 complexes in G₀ cells are resistant to cyclin H-cdk7 phosphorylation.

Others have suggested that cyclin D-cdk4 complexes are not phosphorylated by cyclin H-cdk7 in G₀ cells (9, 23). Antibodies specific for phosphorylated T172 are not available, so we were unable to confirm this result directly. Instead, we analyzed the ability of exogenous cyclin H-cdk7 to phosphorylate the p27-cyclin D-cdk4 complex isolated from A and G₀ cells. Cyclin H-cdk7 was purified from insect cells coexpressing cyclin H and cdk7. We isolated p27-associated complexes by p27 immunoprecipitation and added recombinant cyclin H-cdk7 in the presence of [γ-³²P]ATP (Fig. 2A to C). As p27 associates in vivo with cdk4, cdk6, and cdk2, the p27 immunoprecipitate contains a mixture of associated complexes. We separated this mixture after phosphorylation by boiling the p27-associated complexes in 1% SDS, increasing the volume to dilute the SDS, and reimmunoprecipitating with anti-cdk4, anti-cdk6, or anti-cdk2 antibodies. We were able to phosphorylate all of the p27-associated cdk's from A cells, but the cdk's isolated from G₀ cells were resistant to phosphorylation by cyclin H-cdk7 (Fig. 2A). Immunoblot analysis confirmed that each cdk immunoprecipitation was specific. For example, when cdk4 was reimmunoprecipitated, the immunoblot was probed with both cdk4 and cdk6 antibodies, but only cdk4 was detected (Fig. 2A, left panel). There is little cdk2 detected in the p27 immunoprecipitates from A cells (Fig. 2A, right panel) due to the fact that the bulk of p27 in proliferating Mv1Lu cells is associated with cdk4 or cdk6 (7, 22). A similar result was obtained when cyclin D1 was immunoprecipitated from A and G₀ cells and treated with recombinant cyclin H-cdk7. Cyclin H-Cdk7 was able to phosphorylate cyclin D1-associated cdk4 or cdk6 isolated from A cells but not G₀ cells (Fig. 2B). cdk2 is not associated with cyclin D1 and was not isolated in these immunoprecipitates. These results suggested that all of the cdk complexes isolated from the G₀ condition were resistant to phosphorylation by exogenous cyclin H-cdk7.

FIG. 2.

p27-associated complexes from G₀ cells are resistant to exogenous CAK phosphorylation in vitro while those isolated from A cells can be phosphorylated. (A) Lysates were immunoprecipitated (Ip) with p27 antibodies and treated with exogenous cyclin H-cdk7 (+cdk7) or not treated with exogenous cyclin H-cdk7 (−cdk7) in the presence of [γ-³²P]ATP. The complexes were boiled in 1% SDS and then reimmunoprecipitated (Re-Ip) with cdk4 (left), cdk6 (middle), or cdk2 (right) antibodies to isolate the individual cdk components (cdk*). Parallel immunoblot analysis confirmed that only the correct cdk was isolated in each reimmunoprecipitation (W). Normal rabbit serum (NRS) served as an immunoprecipitation control. (B) Lysates were immunoprecipitated with cyclin D1 antibodies and treated with cyclin H-cdk7 or not treated with cyclin H-cdk7 in the presence of [γ-³²P]ATP, boiled in 1% SDS, and reimmunoprecipitated with cdk4 (top), cdk6 (middle), or cdk2 (right) antibodies. (C) (Left) p27 immunoprecipitations were isolated as described above but treated with cyclin H-cdk7 or not treated with cyclin H-cdk7 in the presence of nonradiolabeled ATP and then used in Rb (Rb*) (top) or histone H1 (H1*) (bottom) kinase assays. cdk2 antibodies were used as a positive control. NRS was used as a control. (D) p27 immunoprecipitates from A and G₀ cells were probed for cdk2 and phosphoT160* by standard immunoblot analysis. NRS was used as a control. (E) Lysates from A or Tetp27 (Tet) cells, grown in the absence of tetracycline, were immunoprecipitated with p27 antibodies, and treated with exogenous cyclin H-cdk7 or not treated with exogenous cyclin H-cdk7 in the presence of [γ-³²P]ATP. The complexes were boiled in 1% SDS and reimmunoprecipitated with cdk4 (left) or cdk2 (right) antibodies to isolate the individual cdk components (cdk*). Parallel immunoblot analysis with cdk4 and cdk2 antibodies is shown.

To test whether phosphorylation by recombinant cyclin H-cdk7 was productive, anti-p27 immunoprecipitates from A and G₀ cells were incubated with nonradiolabeled ATP and cyclin H-cdk7 and then tested for Rb kinase activity in vitro (Fig. 2C). Although kinase activity was detected in anti-p27 immunoprecipitates from A cells, activity was increased when the immune complexes were treated with cyclin H-cdk7 before the addition of the Rb substrate (Fig. 2C, lanes 1 and 3). p27-associated kinase activity has been shown to be due specifically to cdk4 and cdk6 (7, 22). Kinase activity was not detected in anti-p27 immunoprecipitates from G₀ cells in the presence or absence of cdk7 (Fig. 2C). Therefore, recombinant cyclin H-cdk7 can activate p27-bound cdk4 and cdk6 from A cells (but not from G₀ cells), indicating phosphorylation on T-loop residue T172.

Using a cdk2 T160 phospho-specific antibody, we confirmed that p27-associated cdk2 isolated from A cells was phosphorylated in its T loop, while p27-associated cdk2 isolated from G₀ cells was not (Fig. 2D). To test whether the phosphorylation on cdk2 by recombinant cyclin H-cdk7 was productive, anti-p27 immunoprecipitates from A and G₀ cells were incubated with nonradiolabeled ATP and cyclin H-cdk7 and tested for histone H1 kinase activity in vitro (Fig. 2C). Histone H1 phosphorylation is a measure of cdk2 kinase activity, as cdk4 and cdk6 will not phosphorylate this substrate (7, 37). While p27-associated cdk2 isolated from A cells was phosphorylated by cyclin H-cdk7 (Fig. 2A), this increase in T-loop phosphorylation did not restore cdk2's ability to phosphorylate histone H1 substrates (Fig. 2C, lanes 1 and 3), presumably due to the continued association of p27 with this complex. This differs from the situation seen with p27-cyclin D-cdk4 and p27-cyclin D-cdk6 complexes where an increase in T-loop phosphorylation translated directly into increased activity (Fig. 2C, lanes 1 and 3). This suggests that p27's inhibition of cdk2 may be inherently different from its inhibition of cdk4 and cdk6. cdk2 immunoprecipitates from G₀ cells were unable to phosphorylate histone H1 (Fig. 2C, lane 8).

p27 levels increase approximately 10-fold in contact-arrested cells, and we wanted to determine whether the resistance to cyclin H-cdk7-mediated phosphorylation detected in G₀ cells was due to the increase in p27 levels in this condition. TetHisp27 cells contain the human p27 gene under the control of the tetracycline promoter and, in the absence of tetracycline in the culture medium, express histidine-tagged p27 to levels roughly 15-fold over basal levels, i.e., greater than those seen in contact-arrested cells (22). This increase in p27 arrests cells in the G₁ phase and inhibits cdk2 activity, but cdk4 and cdk6 complexes remain catalytically active (7, 22). TetHisp27 cells were grown in the absence of tetracycline, and lysates were immunoprecipitated with anti-p27 antibodies and treated with recombinant cyclin H-cdk7 and [γ-³²P]ATP (Fig. 2E). Cyclin H-cdk7 was able to phosphorylate cdk4 and cdk2 isolated from the TetHisp27 cells. Immunoprecipitation-immunoblot analysis demonstrated that cdk4 and cdk2 were associated efficiently with p27 (Fig. 2E). As the level of p27 in the TetHisp27 cells was greater than the level in the G₀ cells, the ability to block cdk7 phosphorylation was not dependent on the concentration of p27 but was a G₀-specific property.

cdk4 resistance to cyclin H-cdk7 is due to G₀ p27 itself.

We wanted to directly determine whether p27 isolated from A or G₀ cells differentially affected the ability of cdk4 to be phosphorylated by cyclin H-cdk7. Histidine-tagged p27 purified from TetHisp27 cells by metal agarose chromatography in the presence of urea is functional (i.e., able to bind and inhibit cdk's) in vitro (22). TetHisp27 cells were grown to confluence (G₀) or harvested as asynchronously growing populations (A) to generate Ap27 or G0p27, respectively (22). Previously, we demonstrated that Ap27 was capable of binding but not inhibiting recombinant cyclin D-cdk4 complexes (22). On the other hand, G0p27 was a strict cyclin D-cdk4-bound inhibitor. We wanted to determine whether association of Ap27 or G0p27 with cyclin D-cdk4 also affected the ability of the complex to be phosphorylated by cyclin H-cdk7. We incubated purified Ap27 and G0p27 with recombinant cyclin D-cdk4 in the presence or absence of exogenous cyclin H-cdk7 (Fig. 3C). To detect cyclin H-cdk7-mediated phosphorylation of cdk4 directly, [γ-³²P]ATP was included in the reaction mixture, and we immunoprecipitated cdk4 after incubation. To examine kinase activity of cyclin D-cdk4 with or without cyclin H-cdk7 treatment, incubations were performed in the presence of nonradiolabeled ATP, after which we immunoprecipitated cdk4-associated complexes and tested them for Rb kinase activity (Fig. 3C).

FIG. 3.

p27 purified from G₀ cells is sufficient to confer cyclin H-cdk7 resistance. (A) (Top) Increasing amounts of His-cyclin D-cdk4 (DK4) purified by metal agarose chromatography were added to a constant amount of Rb and [γ-³²P]ATP in an in vitro kinase assay (Rb*). (Middle) One microgram of cyclin H-cdk7 was added to increasing amounts of this cyclin D-cdk4 in the presence of [γ-³²P]ATP, followed by immunoprecipitation (IP) with anti-cdk4 antibodies to detect cdk4 phosphorylation (cdk4*). (Bottom) Increasing amounts of this cyclin D-cdk4 were added to a constant amount of p27 and then immunoprecipitated with p27 antibodies to detect associated cdk4 by immunoblot analysis (W). Normal rabbit serum (NRS) was used as a control. (B) Three forms of recombinant cyclin D-cdk4 were tested by Sypro-Ruby protein staining (Bio-Rad) (top), immunoprecipitation (IP)-immunoblot analysis (middle), or Rb kinase assays (bottom). Form 1, unpurified baculoviral lysate expressing cyclin D-cdk4; form 2, cyclin D-cdk4 purified via the histidine tag on cyclin D1; form 3, cyclin D-cdk4 purified first via the histidine tag on cyclin D1, followed by additional isolation of the 66-kDa fraction by Superdex 200 gel filtration chromatography (AKTA FPLC, GE Healthsciences). (Middle panels) All forms were immunoprecipitated with cdk4 antibodies and probed for associated cyclin D1 or cdk4. (Bottom) All forms were used in in vitro kinase assays. Form 2 was used in all of the experiments in the manuscript. Form 3 was unable to phosphorylate Rb, despite associated cyclin D-cdk4. (C) Form 3 was incubated with recombinant cyclin H-cdk7 (+) or not incubated with cyclin H-cdk7 (−). This material was used in an in vitro kinase assay. In the presence of cyclin H-cdk7, this cyclin D-cdk4 regained Rb kinase activity. (D) Purified G0p27 and Ap27 were incubated with recombinant cyclin D-cdk4. (Top) Cyclin H-cdk7 (cdk7) in the presence of [γ-³²P]ATP was added (+) or not added (−) to samples followed by immunoprecipitation (Ip) with cdk4 antibodies to detect radiolabeled cdk4 (cdk4*). (Bottom) Cyclin H-cdk7 in the presence of nonradiolabeled ATP was added (lanes 4 to 6) or not added (lanes 1 to 3) to the Ap27 and G0p27-cyclin D1-cdk4 complexes, followed by immunoprecipitation with cdk4 antibodies for use in Rb kinase assays (Rb*). (E) Ap27 and G0p27 were incubated with cyclin D-cdk4 followed by immunoprecipitation with p27 antibodies to probe for p27 and cdk4 (lanes 1 and 2). NRS served as an immunoprecipitation control (lane 3).

Recombinant cyclin D1-cdk4 complexes are generated by coinfection of High5 cells, and catalytically active cyclin D1-cdk4 can be detected in the unfractionated lysates (7, 9, 22, 24) or, in the present case, following partial purification of cyclin D1-cdk4 complexes containing a histidine tag on cyclin D1 by metal agarose chromatography (22). This partially pure cyclin D1-cdk4 was catalytically active (Fig. 3A, 6 and 12μl), but in our studies we used a concentration of cyclin D-cdk4 (3 μl) that resulted in little kinase activity, unless it was treated with exogenous cyclin H-cdk7 (Fig. 3A, middle panel). This enabled us to clearly see the effects of cyclin H-cdk7 treatment on cyclin D-cdk4. The metal affinity-purified cyclin D1-cdk4 can be further purified by gel filtration chromatography and is detected in the 66-kDa fraction (Fig. 3B, form 3). This material has no detectable activity unless treated with recombinant cyclin H-cdk7 (Fig. 3C).

We mixed the recombinant cyclin D-cdk4 with Ap27 or G0p27 prior to the addition of cyclin H-cdk7 and found that cdk4 could be phosphorylated only when it was associated with Ap27 (Fig. 3D, lane 5); cdk4 associated with G0p27 was resistant (Fig. 3D, lane 6). The addition of Ap27 to recombinant cyclin D-cdk4 in the absence of cyclin H-cdk7 increased the kinase activity of the complex slightly, possibly by enhancing its stability (Fig. 3D, lane 2). Ap27 and G0p27 appeared to bind cdk4 to a similar extent (Fig. 3E), as previously demonstrated (22). This suggested that the association of p27 isolated from A cells permitted T-loop phosphorylation, while the association of p27 isolated from G₀ cells rendered the complex resistant to CAK.

Phosphatase treatment converts Ap27 to an inhibitory form.

Purified Ap27 and G0p27 were treated with PAP to remove S, T, and Y phosphorylations that had occurred in the mammalian cell. The PAP-treated p27 was then repurified by metal agarose chromatography in the presence of urea to remove and inactivate PAP. Immunoblot analysis demonstrated that PAP treatment shifted the p27 isolated from both A and G₀ conditions, consistent with the removal of phosphates that had been added in vivo (Fig. 4A). Mock- and PAP-treated Ap27 and G0p27 continued to associate with cyclin D-cdk4, as demonstrated by p27 immunoprecipitation followed by cdk4 immunoblot analysis (Fig. 4B). Each purified p27 isoform was incubated with cyclin D-cdk4 in the presence or absence of recombinant cyclin H-cdk7 and [γ-³²P]ATP, and cdk4 was then immunoprecipitated with anti-cdk4 antibodies to detect cdk4 phosphorylation (Fig. 4C). In the absence of p27, cdk4 was phosphorylated by cyclin H-cdk7 (Fig. 4C, lane 6). Association of mock-treated Ap27 with cyclin D-cdk4 permitted phosphorylation by cyclin H-cdk7 (Fig. 4C, lane 7), while PAP-treated Ap27 reduced cdk4's ability to be phosphorylated (Fig. 4C, lane 8), suggesting that phosphorylation of Ap27 was sufficient to influence the activity of p27. Both mock- and PAP-treated G0p27 prevented phosphorylation of cdk4 (Fig. 4C, lanes 9 and 10). To ensure that residual PAP had not been carried over into this reaction mixture from the pretreatment of p27, we first incubated recombinant cyclin D-cdk4 with cyclin H-cdk7 to phosphorylate cdk4 (Fig. 4D, lane 2), and then added PAP-treated Ap27 to this already T172 phosphorylated complex (Fig. 4D, lane 3). Similar levels of cdk4 phosphorylation persisted in the presence or absence of this material. The inability to phosphorylate cdk4 in the presence of PAP-treated p27 is therefore unlikely to be due to residual PAP in the reaction mixture. Thus, the reduction in CAK-mediated phosphorylation seen in Fig. 4C, lane 8, was due to the association of PAP-treated (or dephosphorylated) p27 with the complex. This suggested that the ability of p27 to confer resistance to cyclin H-cdk7 is a characteristic of p27 isolated from G₀ cells, but not from A cells, and that phosphatase treatment of Ap27 could convert p27 to a more “G₀-like” isoform.

FIG. 4.

Phosphatase treatment of AHisp27 converts it to an inhibitory form. (A) Ap27 and G0p27 were treated with PAP (lanes 2 and 4) or mock treated (lanes 1 and 3) and then repurified. Immunoblot analysis was performed with p27 antibodies (W). (B) Recombinant cyclin D-cdk4 was incubated with Ap27, PAP Ap27, G0p27, or PAP G0p27 (lanes 1 to 4) and immunoprecipitated (Ip) with p27 antibodies to detect cdk4 association by immunoblot analysis. Normal rabbit serum (NRS) served as a control. (C) Cyclin D-cdk4 (DK4) was mock incubated (lane 1) or incubated with Ap27, G0p27, PAP Ap27, and PAP G0p27 (lanes 2 to 5 and 7 to 10). Recombinant cyclin H-cdk7 (cdk7) in the presence of [γ-³²P]ATP was added (+) or not added (−). Samples were immunoprecipitated with cdk4 antibodies to detect radiolabeled cdk4 (cdk4*). (D) Recombinant cyclin D-cdk4 was incubated with [γ-³²P]ATP and cyclin H-cdk7 (lanes 2 and 3). PAP Ap27 was added to half of this phosphorylated cyclin D-cdk4, followed by immunoprecipitation (Ip) with cdk4 antibodies to detect cdk4 phosphorylation (cdk4*) (lanes 2 and 3).

Nonphosphorylatable p27 mutant Y89F renders cyclin D-cdk4 resistant to phosphorylation by cyclin H-cdk7.

It has been shown by two-dimensional isoelectric focusing analysis that p27 is preferentially phosphorylated in proliferating Mv1Lu cells and that this phosphorylation prevents bound p27 from inhibiting cyclin D-cdk4 activity (22). Previously, we mutated all three of p27's tyrosines to nonphosphorylatable phenylalanine residues (22). These mutants were stably transfected into Mv1Lu-tTA cells to generate the tetracycline repressible cell lines TetY74F, TetY88F, and TetY89F. In the absence of tetracycline in the medium, wild-type Wtp27 and mutants Y74F, Y88F, and Y89F were overexpressed (Fig. 5A). Endogenous p27 is detectable only when the blot is overexposed, as the mutants are expressed roughly 10-fold over basal p27 levels (data not shown). It was shown that, even in proliferating cells, nonphosphorylatable mutant Y89F, and to a lesser extent Y88F, were potent cdk4 inhibitors, whereas Wtp27 and Y74F remained bound cdk4 noninhibitors (22). This suggested that in Mv1Lu cells, Wtp27 and Y74F variants could be phosphorylated and function as noninhibitors, but lack of phosphorylation on residue Y89, or to a lesser extent on Y88, held p27 constitutively in its inhibitory form.

FIG. 5.

Mutant Y89F is resistant to cyclin H-cdk7 but sensitive to Csk1 phosphorylation. (A) TetHisp27 (Wtp27), TetY74F, TetY88F, and TetY89F cells were grown in the presence [+Tet (Off)] (lanes 1 to 4) or absence [−Tet (On)] (lanes 5 to 8) of tetracycline and immunoblotted directly with p27 antibodies (W). (B) Immunofluorescence by confocal microscopy on A and G₀ cells using p27 (panels 1) or phosphoY89p27 (panels 2) antibodies. The nucleus was stained with ToPro (panels 3). Panels 4 show the merged images. (C) p27 immunoprecipitates (Ip) from TetHisp27, TetY74F, TetY88F, and TetY89F were treated with cyclin H-cdk7 (+cdk7) or without cyclin H-cdk7 (−cdk7) and [γ-³²P]ATP and reimmunoprecipitated with cdk4 antibodies to detect cdk4 phosphorylation (cdk4*). Immunoprecipitation-immunoblot analysis with p27 antibody to probe for cdk4 binding is shown in the bottom panel (W). Immunoprecipitation with normal rabbit serum (NRS) served as a control (lane 9). (D) (Top) Lysates from TetHisp27, TetY74F, or TetY89F were immunoprecipitated with p27 antibodies and were mock treated (lanes 1 to 3), treated with Csk1 (lanes 4 to 6), or treated with cyclin H-cdk7 (lanes 7 to 9) in the presence of [γ-³²P]ATP. The complexes were boiled in 1% SDS and reimmunoprecipitated with cdk4 to detect cdk4 phosphorylation (cdk4*). In the p27-associated Rb kinase panel, p27 immunoprecipitates mock treated (lanes 1 to 3), Csk1 treated (lanes 4 to 6), or cyclin H-cdk7 treated (lanes 7 to 9) were used in in vitro kinase assays (Rb*). In the bottom two panels, p27 immunoprecipitates were analyzed by immunoblot analysis with cdk4 and p27 antibodies. Immunoprecipitation with normal rabbit serum (NRS) served as a control (lane 10).

Based on the three-dimensional structure of p27 bound to cyclin A-cdk2, it has been suggested that residues 88 and 89 are part of the 3-10 helix that interacts with the cdk active site (14, 20). Phosphorylation on these residues may cause p27's tail to swing out of the active site, permitting ATP access and activity (20). Residues Y88 and Y89 are found within two consecutive Src homology 2 domains that serve as putative phosphorylation sites for non-receptor-bound Y kinases. To demonstrate that p27 is preferentially phosphorylated in A cells, we generated a phospho-specific antibody to p27 residue Y89. We examined A and G₀ Mv1Lu cells by immunohistochemistry with the phospho-Y89 antibody (Fig. 5B). The G₀ cells did not stain with the phospho-Y89 antibody, despite the increase in total p27 detected in these contact-arrested conditions. However, phospho-Y89 staining was detected in the A cells. The level of staining detected with the phospho-Y89 antibody was variable from cell to cell. As the A cell pool is an asynchronous population, this may reflect cell cycle-regulated Y89 phosphorylation (M. K. James and S. W. Blain, unpublished data).

To determine whether loss of Y phosphorylation also affected p27's ability to permit or prevent phosphorylation of cdk4 by cyclin H-cdk7, p27-associated complexes immunoprecipitated from TetHisp27, TetY74F, TetY88F, or TetY89F cell lysates were mixed with recombinant cyclin H-cdk7 in the presence of [γ-³²P]ATP in vitro, as described above. We found that cyclin H-cdk7 was able to phosphorylate the p27-associated cdk4 isolated from Wtp27, TetY74F, and TetY88F cells, but not the TetY89F-associated cdk4 (Fig. 5C, lane 4). We showed previously that mutant Y89F inhibited already T172-phosphorylated cyclin D-cdk4 complexes in vitro and in vivo, presumably by blocking the cdk4 active site (22). Here we have shown that Y89F also prevented cyclin H-cdk7-mediated phosphorylation of residue T172 de novo.

Activating phosphorylation of cdk's can be catalyzed by two structurally distinct types of activating kinases, exemplified by the cyclin H-cdk7 complex in metazoans and the single-subunit Cak1 in budding yeast. Metazoans appear to contain only the cyclin H-cdk7 complex, despite extensive searching by multiple groups for a Cak1 homologue (reviewed in reference 17). Fission yeast has both types: Mcs6-Mcs2, the cdk7-cyclin H homologue, and Csk1, a single-subunit kinase related to Cak1 (17, 28). We immunoprecipitated p27 from each of the different Tet cell lines and attempted to phosphorylate the associated cdk4 in vitro with purified Csk1 (Fig. 5D). We reimmunoprecipitated cdk4 and found that Csk1, like cyclin H-cdk7, was able to phosphorylate both wild-type and Y74F-associated cdk4 complexes (Fig. 5D, top panel, lanes 4 and 5). However, Csk1, but not the cdk7 complex, could phosphorylate Y89F-associated cdk4 (Fig. 5D, top panel, compare lanes 6 and 9). This suggested that the yeast CAK was able to access the cdk4 T-loop, even when p27 was in the “locked” or nonphosphorylated conformation.

To determine whether phosphorylation by Csk1 occurred on residue T172 and was competent to activate cdk4, anti-p27 immunoprecipitates preincubated with either cdk7 or Csk1 were tested in Rb kinase assays (Fig. 5D, p27-associated Rb kinase panel). Phosphorylation by cyclin H-cdk7 or Csk1 increased the kinase activity of both wild-type and Y74F-associated cdk4 complexes (Fig. 5C, lanes 4, 5, 7, and 8), suggesting that Csk1 was productively phosphorylating the T-loop. However, even though Csk1 was able to phosphorylate Y89F-associated cdk4, the p27-associated kinase activity of this complex was undetectable (Fig. 5D, lane 6). Phosphorylation by Csk1 was therefore not sufficient to overcome the mutant Y89F's block to cdk4 catalytic activity. This suggested that Y89F p27 inhibited cdk4 activity by two different modes, by physically disrupting the cdk4 catalytic site and by blocking access by cdk7 to residue T172.

Non-Y-phosphorylated p27-cyclin D-cdk4 ternary complexes can be Y phosphorylated and reactivated.

The results with the phosphorylation-defective p27 mutant Y89F suggested that, when p27 was not phosphorylated in the 3-10 helix and therefore bound in the “locked” conformation, it prevented both cdk7-mediated phosphorylation and cdk4 active site access. To directly demonstrate this point, we incubated recombinant p27, purified from bacteria, with purified Abl kinase, to generate Y-phosphorylated p27 in vitro (Fig. 6A, lane 1). Incubation of Abl phosphorylated p27 with protein tyrosine phosphatase removed the phosphate (Fig. 6A, lane 3). Mock*p27 was not treated with Abl (Fig. 6A, lane 2). Previously, we demonstrated that Abl primarily phosphorylated residue Y88 in the p27 3-10 helix in vitro and that this phosphorylation converted p27 into the noninhibitory form (22). In vitro, mutant Y88F potently inhibits recombinant cyclin D-cdk4, while mutant Y89F, which we have shown is still weakly phosphorylated by Abl, permits cyclin D-cdk4 kinase activity. These residues are found within contiguous Src homology 2 regions within p27's 3-10 helix, and the differential phosphorylation seen in different conditions suggests that phosphorylation of either residue may suffice to alter the tertiary structure of p27.

We mixed the Abl-phosphorylated p27 (Abl*p27) or mock-phosphorylated p27 (mock*p27) with recombinant cyclin D-cdk4 and then added cyclin H-cdk7 or Csk1 (Fig. 6B). We immunoprecipitated cdk4 and assayed for direct cdk4 phosphorylation (cdk4*) (Fig. 6B, top panel) or for cdk4-associated Rb kinase activity (Fig. 6B, cdk4-associated Rb kinase panel). In the absence of p27, both cyclin H-cdk7 and Csk1 phosphorylated cdk4 and increased the kinase activity of the cyclin D-cdk4 complex (Fig. 6B, lanes 1 to 3), demonstrating that like cyclin H-cdk7, Csk1 is productively phosphorylating cdk4 in the T-loop. Abl*p27's association with cdk4 permitted cyclin H-cdk7 and Csk1 phosphorylation, which translated into increased Rb kinase activity (Fig. 6B, lanes 4 to 6). Mock*p27's association with cdk4 blocked phosphorylation by cyclin H-cdk7 (Fig. 6B, lanes 7 to 9), suggesting that Y phosphorylation of p27 was required to permit cyclin H-cdk7 phosphorylation. Cyclin H-cdk7 was unable to phosphorylate mock*p27-associated cdk4, even when increasing amounts of CAK were added to the reaction mixture (Fig. 6C). In contrast, Csk1 was able to phosphorylate both the mock*p27 and the Abl*p27-associated cdk4 complexes (Fig. 6B, lanes 6 and 9), but only the cyclin D-cdk4 complex associated with Abl*p27 regained activity (Fig. 6B, lane 6). The association of cdk4 with phosphorylated or nonphosphorylated p27 was similar as measured by cdk4 immunoprecipitation followed by p27 immunoblot analysis (Fig. 6B, third panel from the top). This suggested that non-Y-phosphorylated p27 blocked both T172 phosphorylation by cyclin H-cdk7 and access to the cdk4 catalytic site. Csk1 was able to access T172 in the non-Y-phosphorylated p27-cyclin D-cdk4 complex, possibly reflecting its different mode of substrate recognition (26), but the non-Y-phosphorylated p27 still disrupted the cdk4 catalytic site.

The previous experiment tested the ability of phosphorylated p27 to bind and permit T172 phosphorylation and activation of cyclin D-cdk4 complexes. To determine whether a non-Y-phosphorylated p27-cyclin D-cdk4 ternary complex could become Y phosphorylated and reactivated, we treated preformed p27-cyclin D-cdk4 complexes with Abl. Cyclin D-cdk4 was incubated with p27, and cdk4-associated complexes were isolated by immunoprecipitation with anti-cdk4 antibodies. Treatment with recombinant Abl led to phosphorylation of p27 in the ternary complex (Fig. 6D, lane 4). We then isolated preformed p27-cyclin D-cdk4 complexes by immunoprecipitation with anti-cdk4 antibodies and treated these complexes with Abl, before adding cyclin H-cdk7 and assaying for cdk4 T172 phosphorylation and Rb kinase activity (Fig. 6E). In the absence of p27 or Abl, incubation with cdk7 increased both T172 phosphorylation and associated cyclin D-cdk4 kinase activity (Fig. 6E, lane 2). When cyclin D-cdk4 was incubated with p27 before cyclin H-cdk7 treatment, both T172 phosphorylation and Rb kinase activity were blocked (Fig. 6E, lanes 3 and 4). However, when cyclin D-cdk4-p27 complexes were incubated with Abl and ATP in the absence of cyclin H-cdk7, cdk4 was not phosphorylated, and only a low level of Rb kinase activity was detected, demonstrating that Abl cannot directly phosphorylate either cdk4 or Rb (Fig. 6E, lane 5). Cyclin H-cdk7 treatment of the Abl-treated p27-cyclin D-cdk4 complex resulted in T172 phosphorylation and increased Rb kinase activity (Fig. 6E, lane 6). This suggested that phosphorylation by Abl could convert p27 already bound to cyclin D-cdk4 from an inhibitor to a noninhibitor, permitting cdk4 phosphorylation and activation.

These data suggest that the ability to be phosphorylated on residue T172 is governed by the fraction of p27 that is tyrosine phosphorylated, so that more p27 Y phosphorylation should result in more T172 phosphorylation. To directly demonstrate this, we isolated recombinant p27-cyclin D-cdk4 complexes by p27 immunoprecipitation and added increasing amounts of Abl kinase and [γ-³²P]ATP to phosphorylate p27. After incubation, we removed the Abl and ATP by washing the samples and then added a constant amount of cyclin H-cdk7 and [γ-³²P]ATP to each sample. We analyzed the results by SDS-PAGE and determined the amount of Y-phosphorylated p27 and T172-phosphorylated cdk4 generated (Fig. 6F). We found that an increase in the amount of p27 that became Y phosphorylated resulted directly in a proportional increase in the amount of cdk4 phosphorylation by a constant amount of cyclin H-cdk7 (Fig. 6F). In fact, a near 1:1 molar ratio in the amount of Y-phosphorylated p27 to cdk4 phosphorylation was detected. This suggested that the level of cdk4-bound, Y-phosphorylated p27 directly regulated the ability of this complex to be phosphorylated by cyclin H-cdk7. The number of units of Abl kinase was plotted against the number of picomoles of phosphorylated p27 (p27*). The number of picomoles of phosphorylated p27 (p27*) was plotted against the number of picomoles of phosphorylated cdk4 (cdk4*). The number of picomoles were determined from the equation generated using a standard curve (Fig. 6F, left bottom panel). p27 immunoprecipitates were also analyzed by direct cdk4 immunoblotting (Fig. 6F, left).

p27-cyclin D-cdk4 complexes from G₀ cells can be phosphorylated by Csk1 but remain catalytically inactive.

p27 isolated from G₀ cells is poorly Y phosphorylated (Fig. 5B) (22) and is therefore similar to p27 isolated from the TetY89F cells (Fig. 5C, lane 6). To test whether Csk1 could phosphorylate these complexes, we immunoprecipitated cyclin D1-associated complexes from A and G₀ cells and treated them with recombinant cyclin H-cdk7 or Csk1 (Fig. 7). We then performed cyclin D1-associated Rb kinase assays (Fig. 7A, top panel), or reimmunoprecipitated cdk4 directly from the reaction mixture (Fig. 7A, middle panel). Cyclin H-cdk7 was able to phosphorylate and increase the kinase activity of cyclin D1-associated complexes isolated from A cells (Fig. 7A, lane 3), but cdk4 complexes isolated from G₀ cells were resistant to phosphorylation and remained catalytically inactive (Fig. 7A, lane 4). Csk1 was able to phosphorylate cdk4 complexes isolated from both A and G₀ cells (Fig. 7A, lanes 5 and 6), but cyclin D1-associated kinase activity was not detected after Csk1-mediated phosphorylation of G₀ cdk4 complexes (Fig. 7A, lane 6). These results suggested that the G₀ complexes were resistant to phosphorylation by cyclin H-cdk7 but not Csk1. However, they remained catalytically inactive despite T172 phosphorylation, presumably due to p27's blockage of the catalytic site, arguing for two independent modes of inhibition.

FIG. 7.

Csk1 phosphorylates p27-cyclin D-cdk4 complexes from G₀ cells, but they remain catalytically inactive. (A) Lysates from A and G₀ cells were immunoprecipitated with cyclin D1 antibodies to detect associated cdk4 (lanes 5 to 7) (W). Immunoprecipitated complexes were treated with exogenous cyclin H-cdk7 (lanes 3 and 4) or exogenous Csk1 (lanes 5 and 6) to detect cyclin D1-associated Rb kinase activity (Rb*) and cdk4 phosphorylation (cdk4*). Normal rabbit serum (NRS) served as an immunoprecipitation control (lane 7). Ip, immunoprecipitated; Re-Ip, reimmunoprecipitated. (B) Lysates from G₀ cells were immunoprecipitated with cyclin D1 antibodies and treated with (+) or without (−) Abl kinase, followed with cyclin H-cdk7 (+) or Csk1 (+) to detect cyclin D1-associated Rb kinase activity (Rb*, middle panel) or associated cdk4 by immunoblot analysis (W). Treated cyclin D1 immunoprecipitates were boiled and reimmunoprecipitated with cdk4 antibodies to detect direct cdk4 phosphorylation (cdk4*) (top panel). (C) Model of T-loop sensitivity conferred by p27. Cyclin H-cdk7 is able to phosphorylate p27-associated cdk4 isolated from growing cells, where p27 is Y phosphorylated. Cyclin H-cdk7 is unable to phosphorylate p27-associated complexes isolated from G₀ cells or from Y89F-overexpressing cells, where p27 is not Y phosphorylated. Csk1, however, is able to phosphorylate p27-cdk4 complexes from G₀ cells or from Y89F-overexpressing cells. Despite this T-loop phosphorylation, the cdk4 complex does not regain Rb kinase activity. This suggests that p27 can inhibit cyclin D-cdk4 complexes by two separable modes: blocking the catalytic site of cdk4 and preventing cyclin H-cdk7 access to the T-loop.

p27-cyclin D-cdk4 complexes may be catalytically inactive in G₀ cells due to both the lack of T172 phosphorylation and the blockage of the cdk4 active site by non-Y-phosphorylated p27. Reactivation of inactive ternary complexes may be dependent on prior Y phosphorylation of p27, followed by subsequent cyclin H-cdk7 phosphorylation. To determine whether this is physiologically relevant to the reactivation of cyclin D-cdk4 complexes that occurs upon release from the G₀ state, we attempted to reactivate inactive cyclin D-cdk4 complexes isolated from G₀ cells. Cyclin H-cdk7 was unable to phosphorylate or restore the kinase activity of cyclin D immunoprecipitates isolated from G₀ cells (Fig. 7B, lane 2). However, if Abl kinase in the presence of ATP was added first to the cyclin D1 immunoprecipitates before the addition of cyclin H-cdk7, cdk4 was phosphorylated, and associated kinase was detected (Fig. 7B, lane 4). The addition of Abl alone did not have an effect on cdk4 phosphorylation or activity (Fig. 7B, lane 3). The addition of Csk1 to the cyclin D immunoprecipitates did result in cdk4 phosphorylation, but this was not accompanied by a restoration of kinase activity (Fig. 7B, lane 6). However, if the cyclin D immunoprecipitates were first treated with Abl kinase in the presence of ATP before the addition of Csk1, cdk4 phosphorylation and associated kinase activity were detected (Fig. 7B, lane 8). This suggested the following temporal order for the reactivation of inactive p27-cyclin D-cdk4 complexes: p27 must be Y phosphorylated first, and this directly permits cyclin H-cdk7 phosphorylation of residue T172 and the subsequent restoration of kinase activity.

DISCUSSION

In contrast to the situation in mitogen-depleted cells where cyclin D levels decrease, cyclin D-cdk4 complexes are still intact in contact-arrested cells but are catalytically inactive due to their association with the inhibitory, non-Y-phosphorylated, form of p27 (6, 22). We now demonstrate that this non-Y-phosphorylated p27 also blocks cyclin H-cdk7 access to the T-loop (Fig. 7C). Tyrosine phosphorylation of p27, which occurs preferentially in proliferating cells, would permit the p27-cyclin D-cdk4 complex to convert to an “open” conformation, permitting cdk7 access. Tyrosine kinases could phosphorylate preformed p27-cyclin D-cdk4 complexes, rendering both the T-loop and the catalytic site accessible. This suggests that the Y phosphorylation state of p27 may serve as a switch to govern cyclin D-cdk4 activity by two different modes.

The two mechanisms by which p27 inhibits cyclin D-cdk4 in arrested cells are independent but linked and might not be simply redundant. A dual activation mechanism may exist to reactivate inactive p27-cyclin D-cdk4 complexes in G₀ cells. Our model suggests that following release from quiescence, p27 must be Y phosphorylated, allowing the C-terminal tail of p27 to vacate the cdk4 active site. Only then will cyclin H-cdk7 be able to phosphorylate the T-loop, conferring catalytic activity on the complex. We confirmed this by demonstrating that inactivated p27-cyclin D-cdk4 complexes could be reactivated by CAK only if Abl kinase was first added to the reaction mixture and permitted to phosphorylate p27. The amount of T172-phosphorylated cdk4 was directly proportional to the amount of phosphorylated p27 added to the reaction mixture, demonstrating that Y phosphorylation of p27 regulates cdk4 activation. Studies based on metabolic labeling and two-dimensional gel electrophoresis have suggested that cdk4 T172 phosphorylation is lost in arrested cells (9, 31). Our model predicts this, because the associated non-Y-phosphorylated p27 renders cdk4 resistant to phosphorylation by CAK. Association with p27 in G₀ cells keeps cyclin D-cdk4 inactive but also stabilizes it, obviating the need for new protein synthesis to generate the active complex. Dual phosphorylation in the 3-10 helix of p27 and in the T-loop of cdk4 would rapidly activate a preexisting pool of preformed p27-cyclin D-cdk4 complexes. Dual modes of inhibition might thereby make cdk4 inactivation more durable during a prolonged G₀, and more rapidly reversible upon cell cycle reentry.

This model implies that cyclin D-cdk4 assembly and activation might not occur simultaneously. Studies that suggested a mitogen-dependent assembly factor requirement may need to be reevaluated (12). In contact-arrested cells, the p27-cyclin D-cdk4 complex is assembled but inactive (22). Signaling by mitogens or the loss of contact inhibition would be required to convert p27 itself into a noninhibitory form. The nonphosphorylatable Y89F mutant also facilitates assembly of the cyclin D-cdk4 complex but prevents both its phosphorylation by cyclin H-cdk7 and its catalytic activity, suggesting that assembly and activation can be dissociated.

The identity of the proliferating cell-specific Y kinase that phosphorylates p27 is currently unknown. Abl, Lyn, Src, and Yes kinases have been shown by several groups to phosphorylate p27 on residues Y74, Y88, and Y89 (14). Both Y88 and Y89 are found within the 3-10 helix of p27, and structural studies with p27 and cdk2-associated complexes have suggested that these residues may form hydrophobic interactions with the cdk subunit, which might be blocked by phosphorylation. Which residue is phosphorylated might not be important. Rather, phosphorylation anywhere in this loop and the consequent disruption of those hydrophobic interactions may promote the “open” conformation. While Abl appears to phosphorylate residue Y88 better in vitro (22), we cannot rule out the possibility that this kinase also phosphorylates Y89 in vivo. As shown previously and in this current study, mutation of Y89 converted p27 into a constitutively inhibitory form in vivo and prevented T172 phosphorylation of immunoprecipitated cdk4 complexes, suggesting that this was the primary site of phosphorylation in this cell type. However, we also cannot rule out the possibility that under different conditions in this cell type, residue Y88 might also become phosphorylated.

Several studies had previously suggested that p27 might inhibit cyclin D-cdk4 by blocking the activating T-loop phosphorylation on the cdk subunit (reviewed in reference 8). Cyclic AMP causes growth arrest in macrophages by a p27-dependent blockage of cdk4 T-loop phosphorylation (23). Conversely, serum stimulation activated preformed cyclin D-cdk4 complexes by promoting T172 phosphorylation (9). The possibility that T-loop phosphorylation was directly regulated by extracellular mitogenic or antimitogenic factors appeared unlikely, given that cyclin H-cdk7 was constitutively active throughout the cell cycle. Instead, our results indicate that accessibility of the T-loop to CAK might be regulated by mitogenic or antimitogenic factors acting through p27. The fact that p27 could block phosphorylation by cdk7 in some contexts but not others led to the idea that the stoichiometry of binding was important (7). However, we have demonstrated that the ability of p27 to block both T-loop phosphorylation and catalytic activity (22) was dependent on the growth state but not on the concentration. The Y phosphorylation model may therefore reconcile previous studies, by postulating that in different cell types with different levels of Y kinases or phosphatases, p27 may permit or prevent cdk4 activation by cdk7. Modulation of Y kinase activity would influence cyclin D-cdk4 kinase activity without the need for protein synthesis or alteration of cyclin H-cdk7 levels.

It is unclear why Csk1 is able to phosphorylate the T-loop even in the locked complex, but this may reflect its ability to recognize the T-loop sequence directly, which cdk7 cannot do (26). Phosphorylation by Csk1 is able to activate recombinant cyclin D-cdk4 complexes, which suggests that phosphorylation occurs on the T-loop. It is indifferent to the state of p27 phosphorylation, however, so it is more indiscriminate about which complexes it phosphorylates and so, given a mixed population of complexes containing phosphorylated and unphosphorylated p27, might also catalyze “nonproductive,” but nonetheless accurate T172 phosphorylation. Perhaps for this reason, activation of recombinant p27-cyclin D-cdk4 complexes by Csk1 was less efficient that activation by cyclin H-cdk7 (Fig. 7A).

Whereas the requirement for T-loop phosphorylation is conserved across species, CAK itself has diverged (reviewed in reference 17). In metazoans, the cyclin H/cdk7/Mat1 trimer can phosphorylate most or all cyclin-cdk complexes, but in Saccharomyces cerevisiae, the Cdk7 ortholog Kin28 does not have CAK activity (15) and T-loop phosphorylation is provided by Cak1, a monomeric kinase distantly related to cdk's. In Schizosaccharomyces pombe, the cdk7 ortholog Mcs6 (10) and the Cak1 ortholog Csk1 both activate cdk's (21, 28, 35). No Csk1 homologue in mammalian cells has ever been found. Cdk7 is active throughout the cell cycle and is capable of phosphorylating all cdk's, possibly obviating the need for another CAK (17). A chemical-genetic analysis in human cells recently demonstrated that cdk7 was indeed the CAK responsible for activating both cdk1 and cdk2 in vivo and seemed to dispel the persistent notion that cdk2 was activated by a different kinase (27). In fact, data that were interpreted to suggest the existence of an additional cell cycle-regulated CAK in mammalian cells can be explained by the Y-phosphorylated p27 switch. Csk1 appears to phosphorylate cdk4's T-loop irrespective of the state of p27 Y phosphorylation, and so the loss of cdk4 T172 phosphorylation in G₀ cells with low levels of p27 Y phosphorylation can even be taken as indirect evidence against the existence of a Csk1-like cdk4-activating kinase in mammalian cells.

Others have suggested that Y-phosphorylated p27 binds cdk2 in a noninhibitory manner in certain cancers that overexpress Y kinases (13, 20). We have shown that p27-associated cdk2 is only competent for T-loop phosphorylation when isolated from nonarrested (A) cells, in which p27 is Y phosphorylated (Fig. 2A, right panel). However, unlike the situation with cdk4, p27-cdk2 complexes from both A and G₀ cells are catalytically inactive, even when phosphorylated on T160 by exogenous cyclin H-cdk7 (Fig. 2C) (7, 22). These results suggest that p27's association with the cdk2 complex may be enough to inhibit catalytic activity, irrespective of the state of p27 Y phosphorylation. Besides blocking ATP access to the catalytic site and T-loop phosphorylation, p27 may inhibit certain cyclin-cdk complexes by an additional mechanism. Both p27 and some cdk substrates, such as Rb and p107, appear to bind to a common motif present on the cyclin subunit, and p27's association has been shown to prevent substrate access (1, 2). This may be more applicable to cdk2 complexes, because cyclins A and E have better-defined p27 and substrate binding domains. Cyclin D-cdk4 might associate with Rb by a different mode, independent of the cyclin-binding domain (39), but p27 binding to cyclin D-cdk4 has not been fully defined due to the lack of structural information about the ternary complex. Y-phosphorylated p27, which “opens” the active site, might permit phosphorylation of a cdk2 substrate that does not depend on the cyclin targeting domain or of a prebound substrate, such as p27 itself. In fact, Y phosphorylation of p27 may increase its ability to be phosphorylated by cdk2 and targeted for ubiquitin-mediated degradation (14).

In conclusion, our data suggest that Y-phosphorylated p27 can inhibit cyclin D-cdk4 complexes by two independent mechanisms: blocking access to the T-loop and disrupting the cdk4 active site directly. Our model suggests that p27 Y phosphorylation is a molecular “switch” that would help turn cdk4 activity on or off. Modulation of Y kinase activity would permit activation of preformed, inactive p27-cyclin D-cdk4 complexes by cdk7 and may be used to regulate cdk4 activity throughout the cell cycle.