INTRODUCTION

Top Abstract Introduction Materials and methods Results Discussion References

Progress through the G₁ phase of the mammalian cell cycle is regulated by the ordered synthesis, assembly, and activation of distinct cyclin-CDK holoenzymes (45, 46). Cyclins D1, D2, and D3 are up-regulated as cells exit from quiescence and associate with their major kinase partners CDK4 and CDK6 (3, 29, 32, 53). These two kinase molecules are highly homologous and associate exclusively with the D-type cyclins (3). Numerous studies have implicated cyclin D-CDK4-CDK6 complexes as key regulators of the cell cycle up to a hypothetical point during late G₁ (24, 25), the restriction point, when hyperphosphorylation and inactivation of the retinoblastoma tumor suppressor gene product, pRB, occur (37, 44).

In contrast to mitotic cyclin-CDK complexes, the D-type cyclins do not automatically assemble into complexes with either CDK4 or CDK6. For example, when overexpressed in NIH 3T3 cells in the absence of serum, D-type cyclins and CDK4 do not interact efficiently (30). Hence, assembly of D-type cyclins and CDK4 and CDK6 into functional complexes in vivo is likely to depend on numerous factors, in particular, synthesis rates and stability of the various components. Indeed, the D-type cyclins possess canonical PEST sequences near their C termini and have short half-lives in vivo (4, 31).

Association of the D-type cyclins with CDK4 and CDK6 is also influenced by the INK4 family of CDK inhibitors (p15^INK4b, p16^INK4a, p18^INK4c, and p19^INK4d) (9, 10, 12, 18, 42). INK4 polypeptides bind to the catalytic subunits and inhibit the association of D-type cyclins (10, 38). Typically, human cell lines that lack functional pRB express very high levels of p16^INK4a (1, 35, 38). In such cells, CDK4 and CDK6 do not interact with D-type cyclins but are sequestered into long-half-life, binary complexes with p16^INK4a (38). These observations have led to a simple model whereby INK4 family members compete with the D-type cyclins for binding to their target CDKs, and uncomplexed D-type cyclins are rapidly degraded.

Members of a second family of CDK-regulatory molecules (p21^CIP1, p27^KIP1, and p57^KIP2) (8, 15, 22, 28, 40, 48) comprise a class of polypeptides thought to be broad-spectrum inhibitors of different cyclin-CDK complexes. The prototypic member is p21^CIP1, a molecule independently identified by several laboratories. Variously described as a p53-regulated cell cycle inhibitor and a marker induced during cellular senesence (7, 34), p21^CIP1 was also cloned biochemically by virtue of copurification with cyclin D1 from mammalian cell extracts (52). In vitro, p21 family members bind to and inhibit the kinase activities of many mammalian cyclin-CDK complexes (16). Recently, evidence has emerged that in addition to simply inhibiting kinase activity, members of the p21 family of molecules may have additional roles. For instance, in overexpression studies, p21^CIP1 seemed to play a role during assembly of cyclin D-CDK complexes (20).

Aberrant accumulation of active cyclin D-CDK complexes and the inappropriate phosphorylation of pRB are common events in a variety of human tumors (11). Cyclin D1 becomes amplified or overexpressed in many different tumor types (5, 21). Similarly, CDK4 is subject to amplification (17, 26, 41), as well as to point mutations that render it insensitive to INK4 inhibition (50). However, of all the CDK inhibitors, only p16^INK4a has been shown convincingly to be a tumor suppressor (19, 39). Consistent with this, p16^INK4a levels rise dramatically during cellular senescence (2, 13, 36). In this work, we address the mechanism of cyclin D-CDK assembly by examining specific biochemical properties of individual CDK inhibitors and their associated target molecules. The data suggest that the INK4 and p21 families of CDK-regulatory polypeptides play antagonistic roles during the formation of cyclin D-CDK4-CDK6 holoenzymes. Moreover, we find that the p16^INK4a tumor suppressor gene product is unique among the INK4 group, as it alone forms stable complexes with both CDK4 and CDK6 under proliferative conditions, suggesting that this molecule is a specialized member of the INK4 family.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and methods Results Discussion References

Cell culture. CEM and ML1 cells were maintained in RPMI 1640 medium with 10% fetal calf serum (FCS); EL4, WI38, 293, and MCF7 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% FCS. For adenovirus infections, 293 cells were split from the stock culture and grown to ~80% confluence. Adenovirus supernatants, both AdVec control and recombinant Adp21^CIP1 (49), were applied at multiplicities of infection of ~2. Twenty-four hours postinfection, cells were harvested prior to lysis.

Antibodies. Peptides corresponding to the C-terminal regions of the following polypeptides were synthesized (Research Genetics), coupled to keyhole limpet hemocyanin (Pierce), and injected into rabbits (Pocono Rabbit Farms): murine-human cyclin D1 (CLACTPTDVRDVDI), murine-human cyclin D2 (CDPDQATTPTDVRDVDL), murine-human cyclin D3 (CGPSQTSTPTDVTAIHL), murine-human CDK4 (CALQHSYLHKEESDAE), murine-human CDK6 (CSQNTSELNTA), murine p15^INK4b (CGHRDIARYLHAATGD), human p15^INK4b (CGHRDVAGYLRTATGD), human p16^INK4a (CARIDAAEG PSDIPD), murine p18^INK4c (CSLMEANGVGGATSLQ), human p18^INK4c (CSLMQANGAGGATNLQ), murine p19^INK4d (CQNLMDILQGHMMIPM), human p19^INK4d (CQDLVDILQGHMVAPL), and human p21^CIP1 (CTDFYHSKRRLIFSKRKP). All peptide antisera were affinity purified against the cognate immunogen (Sulfolink; Pierce). Antibodies to glutathione-S-transferase (GST) fusion proteins of murine p21^CIP1 and human p27^KIP1 were also raised. These were partially purified on protein A columns (Pierce) and subsequently depleted of GST-reactive immunoglobulins by passage over agarose-immobilized GST. Following purification, all sera were either dialyzed against phosphate-buffered saline (PBS)-1 mM dithiothreitol (DTT) plus 20% glycerol or covalently coupled to protein A-Sepharose beads for use in immunoprecipitation analyses (14). Dialyzed sera were stored at 20°C, and antibody-bead slurries were maintained at 4°C. Antibody specificity was confirmed with specific in vitro-translated products and by immunoprecipitation, V8 proteolytic analysis, and Western blotting.

Gel filtration chromatography. Gel filtration chromatography was carried out by using a Superdex 200 HR 10/30 column with a fast-performance liquid chromatography system (Pharmacia). Samples (500 µl) containing 5 to 10 mg of whole-cell extract were loaded onto the column and separated in gel filtration buffer (50 mM HEPES [pH 7.5], 10 mM MgCl₂, 150 mM NaCl) at a flow rate of 0.3 ml/min for the first 5 ml, 0.4 ml/min for the next 10 ml, and then 0.5 ml/min for the final 10 ml. The molecular mass standards (Sigma) used to calibrate the column were blue dextran (2,000 kDa), thyroglobulin (669 kDa), apoferritin (443 kDa), -amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa) and cytochrome c (12.4 kDa). The void volume of different columns was between 8 and 8.5 ml, and this volume was the point at which fraction collection commenced. For each fractionation, 24 0.5-ml fractions were collected. Typically, fractions 2 to 19 were used in later experiments.

Immunoprecipitation and Western blotting analyses. Cells were washed once in PBS and lysed in Nonidet P-40 (NP-40) lysis buffer (50 mM HEPES [pH 7.5], 1% NP-40, 150 mM NaCl, 1 mM DTT, aprotinin [5 µg/ml], leupeptin [5 µg/ml], pepstatin [5 µg/ml], Pefabloc [5 µg/ml]) at 4°C for 30 min. Lysates were then clarified by centrifugation at 15,000 × g for 15 min. For kinase assays, lysates were prepared by the addition of Tween 20 lysis buffer (50 mM HEPES [pH 7.5], 10 mM MgCl₂, 150 mM NaCl, 0.1% Tween 20, 1 mM DTT, 25 µM ATP, aprotinin [5 µg/ml], leupeptin [5 µg/ml], pepstatin [5 µg/ml], Pefabloc [5 µg/ml]), followed by passage through a 19-gauge hypodermic needle and two rounds of freeze-thawing on dry ice. Immunoprecipitations were performed with 50 µl of bead-immobilized antibodies (20% slurries). Peptide blocks were performed by preincubating 5 µg of cognate peptide with antibody beads at 30°C for 30 min prior to the addition of lysate. Typically 500 to 1,000 µg of whole-cell extract was used in standard immunoprecipitations. Immune complexes were collected by rocking for 2 h at 4°C, followed by extensive washing with ice-cold lysis buffer. For Western blotting analysis, samples were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to Immobilon P membrane (Millipore). Membranes were probed with primary antibodies diluted in PBS-0.2% Tween 20-5% milk powder, either singly or in combination. Complexes were detected with horseradish peroxidase-linked secondary serum (Amersham) and enhanced chemiluminescence (ECL; Amersham).

Immunodepletions. Whole-cell extracts were prepared with NP-40 lysis buffer and subjected to five rounds of immunodepletion, using either normal rabbit immunoglobulin or a combination of anti-p21^CIP1 (p21^CIP1) and p27^KIP1 antibodies immobilized on beads. Following depletion, supernatants were immunoprecipitated with specific antisera as required.

In vitro binding assays. In vitro translation products were generated by using a coupled transcription-translation system (TNT; Promega) according to manufacturer's protocols. Equal amounts (typically 5 µl) of two freshly generated in vitro-translated products were mixed and incubated at 30°C for 30 min. The mixture was then diluted with 1 ml of lysis buffer containing 1% NP-40, 50 mM HEPES (pH 7.5), 500 mM NaCl, 3% bovine serum albumin, and protease inhibitors, followed by centrifugation at 16,000 × g for 10 min to remove any denatured material. A 950-µl aliquot of the supernatant was transferred to a fresh tube containing INK4-specific antibody-bead slurry. The sample was then processed in the same way as a normal immunoprecipitation.

Pulse-chase analyses. Approximately 10⁷ cells per time point were washed in DMEM without methionine supplemented with 10% dialyzed FCS and incubated in similar medium for 30 min at 37°C. Cells were pulsed-labelled for 4 h with ³⁵S cell labelling mix (Amersham) at 50 to 100 µCi/ml and then washed with warm PBS before being placed in complete DMEM. Cells were lysed as described above at hourly time points. Lysates were precleared overnight at 4°C with normal rabbit immunoglobulin and killed Staphylococcus aureus. Immunoprecipitations were performed by incubating lysates with immobilized antibodies, and immune complexes were washed and separated by SDS-PAGE. Immunoprecipitated proteins were visualized by fluorography using Amplify (Amersham).

Preparation of bacterially expressed recombinant protein. For use in CDK4 kinase assays, a C-terminal fragment of human RB (amino acids 773 to 928) fused to GST was utilized as an in vitro substrate (32). Briefly, BL21(DE3)pLysS Escherichia coli transformed with pGEX-RBCT were induced at room temperature with 0.1 mM IPTG (isopropyl--D-thiogalactopyranoside) for 3 h. Bacterial pellets were lysed by the addition of ice-cold Tween 20 lysis buffer (50 mM HEPES [pH 7.5], 10 mM MgCl₂, 150 mM NaCl, 0.1% Tween 20, 1 mM DTT, 25 µM ATP, aprotinin [5 µg/ml], leupeptin [5 µg/ml], pepstatin [5 µg/ml], Pefabloc [5 µg/ml]) and sonication. Soluble fusion proteins were purified by affinity chromatography on glutathione-agarose and eluted with 20 mM reduced glutathione. Positive fractions containing the fusion protein were pooled and dialyzed overnight against kinase assay buffer (50 mM HEPES [pH 8.0], 10 mM MgCl₂, 2.5 mM EGTA, 1 mM DTT, 50 µM ATP) plus 20% glycerol. Dialyzed GST RB was dispensed as aliquots and stored at 80°C. GST fusion proteins that were to be used as immunogens were dialyzed against PBS containing 1 mM DTT and stored at 80°C until required.

Kinase assays. Kinase assays were carried out according to an adaptation of a previously described protocol (30) using the GST RB C-terminal (GST-RBCT) construct described above as a substrate. Cells were lysed in Tween 20 lysis buffer and immunoprecipitated as described above. Immune complexes were washed three times with Tween 20 lysis buffer and twice with kinase reaction buffer (50 mM HEPES [pH 8.0], 10 mM MgCl₂, 2.5 mM EGTA, 1 mM DTT, 50 µM ATP). GST RB (2.5 µg) and 10 µCi of [-³²P]ATP (Amersham) were added to each sample, and the volume was made up to 50 ml with kinase reaction buffer. Reactions were incubated at 30°C for 20 min and stopped by adding 6× SDS sample buffer. Phosphorylated products were resolved on an SDS-12% polyacrylamide gel and analyzed by autoradiography at room temperature.

	RESULTS

Top Abstract Introduction Materials and methods Results Discussion References

CDK4 and CDK6 have different gel filtration profiles. We have previously shown by gel filtration that CDK6 from CEM cells is present in three major complexes, implying biochemical regulation of CDK6 interactions within cells (27). A 450-kDa complex comprises CDK6 complexes with chaperone molecules such as HSP90 and CDC37. A minor complex that migrates at approximately 150 to 170 kDa contains cyclin D-CDK6 complexes. When assayed in vitro with a GST RB substrate, this fraction of CDK6 is kinase active. Finally, CDK6 is present in a smaller, ~55-kDa complex that comprises independent binary complexes of CDK6 with INK4 proteins, as well as potentially monomeric forms of the kinase. To examine whether CDK4 regulation was similar to that of CDK6, we extended these analyses in a range of cell lines (Table 1) and compared the properties of CDK4 and CDK6 by gel filtration.

View larger version (64K): [in this window] [in a new window]   FIG. 1.   Gel filtration analysis of CDK4 and CDK6. EL4 whole-cell extract was separated and immunoprecipitated with CDK6 serum (A) or CDK4 serum (B) followed by blotting with the precipitating antiserum to reveal the different complexes. Blots were reprobed with D1 and D2 sera (C and D). (E) Column fractions were immunoprecipitated with D2 serum and immunoblotted with a mixture of CDK4 and CDK6 sera. This blot was then reprobed with D2 (F). (G) CDK4 kinase assays on similar EL4 fractions, with GST-RBCT used as a substrate. M, molecular mass markers; IP, immunoprecipitation.

Members of the INK4 family differ in ability to form complexes with both CDK4 and CDK6. To examine CDK4 association with INK4 family members, we surveyed INK4-CDK4 and -CDK6 interactions in a variety of cell lines by using specific INK4 sera (Table 2). Representative samples of the data are shown in Fig. 2A. Significantly, only p16^INK4a was found to form steady-state complexes with both CDK4 and CDK6, as shown by immunoprecipitation from WI38 diploid fibroblasts. In contrast, p15^INK4b, p18^INK4c, and p19^INK4d immunoprecipitations from asynchronously growing cells contained CDK6 but no detectable CDK4. Under similar conditions, CDK4 was detected in D2 complexes, suggesting that CDK4 functions normally in these cells (Fig. 2A). To address the stoichiometries of these interactions, whole-cell extracts were fractionated by gel filtration and subjected to immunoprecipitation with specific INK4 sera. Binary complexes of p16^INK4a with CDK4 and p16^INK4a with CDK6 were present in WI38 lysates (Fig. 2C). However, when assayed by this method the other INK4 family members formed 55-kDa complexes with CDK6 only (Fig. 2B, D, and E). The blots were reprobed with the precipitating antiserum to reveal the elution profiles of the various INK4 molecules. In the cell lines tested, all four INK4 family members coeluted with CDK4 or CDK6 in complexes of ~55 kDa. INK4 proteins were not detected in complexes outside of this size range (data not shown). Overall, an interesting correlation between absence of p16^INK4a and lack of 55-kDa CDK4 complexes was observed, suggesting that the majority of INK4 family members interact with CDK6 but not CDK4 under these conditions.

View larger version (69K): [in this window] [in a new window]   FIG. 2.   INK4-CDK4 and INK4-CDK6 interactions. (A) Cell lysates from EL4, WI38, and CEM cell lines were subjected to immunoprecipitation with specific INK4 serum with (+) or without () peptide block as indicated. Anti-CDK4, CDK6, and D2 immunoprecipitations were included as controls. Immune complexes were separated by SDS-PAGE and immunoblotted with a mixture of CDK4 and CDK6 sera. Similar analyses were then performed on fractionated whole-cell extracts. (B) Fractionated EL4 extract was immunoprecipitated with p15INK4b and immunoblotted with CDK4-CDK6. (C) Similarly, WI38 lysate was fractionated and immunoprecipitated with p16INK4a, followed by blotting with CDK4-CDK6. Likewise, EL4 and CEM fractions were immunoprecipitated with p18INK4c (D) and p19INK4d (E), respectively, and probed with CDK4-CDK6. IP, immunoprecipitation.

All INK4 family members associate with CDK4 and CDK6 in vitro. The differences in CDK4 and CDK6 binding found in vivo were unexpected, as members of the INK4 family are highly homologous in pairwise comparisons. One explanation for the data is that p15^INK4b, p18^INK4c, and p19^INK4d do not bind CDK4. Indeed, p18^INK4c was originally reported to be a specific CDK6 inhibitor (9). We performed in vitro binding assays using [³⁵S]methionine-labelled, in vitro translates of all INK4 family members and CDK4 and CDK6. Cyclin D1 and CDK2 were used as negative controls for interaction. Under these conditions, all four INK4 family members bound CDK4 and CDK6 with similar affinities (Fig. 3A). Samples of cyclin D1, CDK2, CDK4, and CDK6 in vitro translates were run on a separate gel to control for input levels (Fig. 3B). Viewed in combination with the immunoprecipitation data, these data provide evidence that all INK4 proteins are capable of association with both CDK4 and CDK6 in vitro. However, detectable, steady-state complexes between p15^INK4b, p18^INK4c, or p19^INK4d and CDK4 do not accumulate in proliferating cells.

View larger version (54K): [in this window] [in a new window]   FIG. 3.   In vitro binding of INK4 inhibitors to CDK4 and CDK6. (A) Freshly prepared D1, CDK2, CDK4, and CDK6 in vitro translates were mixed with individual INK4 in vitro translates, as indicated (A). Complexes were immunoprecipitated using specific antisera, separated by SDS-PAGE in a 12% gel, and visualized by autoradiography. (B) Aliquots of input in vitro translates were separated on another gel as a loading control.

CDK4 forms unstable complexes with p18^INK4c in vivo. To test the hypothesis that the stability of INK4-CDK4 and -CDK6 interactions differ among family members, we performed pulse-chase experiments with two cell lines, EL4 and ML1. Cells were labelled with [³⁵S]methionine and chased in nonradioactive medium over a 4-h time course. Lysates were prepared, and p18^INK4c immunoprecipitations were performed at hourly time points. Under these conditions, newly synthesized CDK6 and CDK4 were found to associate with p18^INK4c at time zero, indicating that both kinases do indeed associate with p18^INK4c, which confirms the in vitro binding experiments. However, during the chase in nonradioactive medium, ³⁵S-labelled CDK4 disappeared rapidly from p18^INK4c complexes in both cell lines analyzed (Fig. 4), whereas CDK6 remained associated with p18^INK4c throughout the duration of the experiment. In control experiments, both CDK4 and CDK6 were found to possess long half-lives when measured by specific immunoprecipitation (data not shown).

View larger version (75K): [in this window] [in a new window]   FIG. 4.   Stability of p18INK4c-CDK complexes. Logarithmically growing cultures of EL4 (A) and ML1 (B) cells were pulse-labelled with [35S]methionine and chased with medium containing unlabelled amino acids. At the time points indicated, cell extracts were prepared and subjected to immunoprecipitation (IP) with specific p18INK4c sera. Immune complexes were separated by SDS-PAGE in a 12% gel.

CDK4 is found in steady-state, ternary complexes with cyclin D and p21^CIP1-p27^KIP1. CDK4 does not form stable complexes with p18^INK4c in EL4 and ML1 cells. In other metabolic labelling experiments, newly synthesized CDK4 was also detected in p15^INK4b and p19^INK4d immunoprecipitations (data not shown), implying that CDK4 forms similar short-lived complexes with these INK4 molecules. However, CDK4 is detectable by Western blotting as one component of a 150- to 170-kDa moiety, suggesting that in this context CDK4 is part of a stable complex. The D-type cyclins also contribute to this complex (Fig. 1). Other likely candidates for cyclin D-CDK interacting proteins were members of the p21 family of CDK inhibitors. To examine the distribution of p21 family molecules, EL4 whole-cell extracts were fractionated by gel filtration and immunoprecipitated with specific p21^CIP1 or p27^KIP1 sera. Immune complexes were separated by SDS-PAGE and immunoblotted with D2 and CDK4 sera. p21^CIP1 immunoprecipitations revealed a 150- to 170-kDa, p21^CIP1-D2-CDK4 complex (Fig. 5A, upper panel). When reprobed with p21^CIP1, it was apparent that p21^CIP1 was not detectable in complexes outside of this molecular mass range (Fig. 5A, lower panel). Virtually identical results were obtained for p27^KIP1-containing complexes when p27^KIP1 serum was used as the precipitating antibody (Fig. 5B). Furthermore, similar profiles were obtained when p21^CIP1-D2-CDK6 and p27^KIP1-D2-CDK6 complexes were analyzed by gel filtration (data not shown). To confirm the ternary nature of these complexes, similar gel filtrations were performed, and fractions were immunoprecipitated with D2 sera, separated by SDS-PAGE, and immunoblotted with p21^CIP1 (Fig. 5C, upper panel) or p27^KIP1 (Fig. 5C, lower panel). Thus, the majority of D-type cyclins and p21^CIP1 and p27^KIP1 are associated in readily detectable 150- to 170-kDa complexes. To rule out the possibility that p21^CIP1 and p27^KIP1 reside in the same complex simultaneously, p21^CIP1 immunoprecipitations were immunoblotted with p27^KIP1 sera, and vice versa. We could find no evidence for such quarternary interactions (data not shown). To estimate the fraction of the 150- to 170-kDa cyclin D-CDK complex that was associated with p21 family members, immunodepletion experiments using lysates prepared from two cell lines, EL4 (T lymphoma) and MCF7 (breast adenocarcinoma), were performed. Whole-cell extracts were subjected to immunodepletion with either normal rabbit immunoglobulin G or a combination of both p21^CIP1 and p27^KIP1 sera. The depleted extracts were then divided equally and immunoprecipitated with D1-D2-D3, p21^CIP1-p27^KIP1, or normal rabbit immunoglobulin G. The resulting immune complexes were separated by SDS-PAGE and immunoblotted with a mixture of D1-D2-D3 sera. Significantly, all three D-type cyclins were removed from EL4 and MCF7 extracts upon immune depletion of p21^CIP1 and p27^KIP1 (Fig. 5D). There was no diminution of the D-type cyclin signal in D1-D2-D3 or p21^CIP1-p27^KIP1 immunoprecipitations performed with mock-depleted extracts of either cell line. Although not quantitative, these data strongly suggest that the majority of detectable cyclin D-CDK complexes contain either p21^CIP1 or p27^KIP1.

View larger version (68K): [in this window] [in a new window]   FIG. 5.   Analysis of p21CIP1 and p27KIP1 by gel filtration. EL4 whole-cell extracts were prepared and fractionated over Superdex 200. (A) Fractions were immunoprecipitated with p21CIP1 sera and immunoblotted with a mixture of D1-D2 and CDK4 sera (upper panel). The blot was then reprobed with p21CIP1 (lower panel). (B) Immunoprecipitates from a similar EL4 fractionation using p27KIP1 sera were immunoblotted in the same way. (C) -D2 serum was used to immunoprecipitate complexes following gel filtration. After SDS-PAGE, the immune complexes were immunoblotted with sera specific for p21CIP1 (upper panel) or p27KIP1 (lower panel). (D) Mock-depleted and p21CIP1-p27KIP1-immunodepleted extracts of EL4 and MCF7 cells were subjected to immunoprecipitation with D1-D2-D3, p21CIP1-p27KIP1 or normal rabbit serum, as indicated. Immune complexes were separated by SDS-PAGE and immunoblotted with a mixture of D1/D2/D3 sera. IP, immunoprecipitation.

Both p16^INK4a and p21^CIP1 form stable complexes with CDK4. In pulse-chase experiments, CDK4 has a long-half-life interaction with p16^INK4a (38), an observation supported by the relative ease with which both CDK4 and CDK6 can be detected in p16^INK4a immunoprecipitations. Other INK4 molecules seem less able to form stable complexes with CDK4 (Fig. 2 and 4) in the cell lines tested. Conversely, the abundance of CDK4 present in steady-state p21^CIP1 complexes implies that cyclin D-CDK4-p21^CIP1 ternary complexes have long half-lives. To compare the duration of CDK4 in p16^INK4a- and p21^CIP1-containing complexes, we performed pulse-chase analyses with WI38 primary fibroblasts, a cell line that expresses both proteins (Table 1). At each time point, lysates were prepared and immunoprecipitated with antisera specific for p16^INK4a or p21^CIP1. The resulting immune complexes were separated by SDS-PAGE, and the turnover rate of coprecipitated CDK4 was assessed by fluorography.

View larger version (49K): [in this window] [in a new window]   FIG. 6.   Relative stabilities of p16INK4a and p21CIP1-bound CDK4. Logarithmically growing cultures of WI38 cells were pulse-labelled with [35S]methionine and chased with medium containing unlabelled amino acids. At the time points indicated, extracts were prepared and immunoprecipitated with specific p16INK4a (A) or p21CIP1 (B) sera. Immune complexes were separated by SDS-PAGE in a 12% gel. (C) Metabolically labelled CDK4 reimmunoprecipitated from an p21CIP1 pulse-chase. IP, immunoprecipitation.

CDK4 forms stable complexes with p16^INK4a and p18^INK4c in cells that lack cyclin D-CDK4-p21^CIP1 ternary complexes. Pulse-chase analyses with WI38 cells suggested that p21 family members influence the interactions of CDK4 via competitive stabilization, thus antagonizing the effects of INK4 molecules. We were interested in examining the distribution of CDK4 in cells without ternary cyclin D-CDK-p21^CIP1 complexes, such as human cell lines that lack functional pRB (4). In such cells, CDK4 and CDK6 are found associated with p16^INK4a, and 293 is a cell line that conforms to this paradigm. 293 cells express p18^INK4c and undetectable levels of both p21^CIP1 and p27^KIP1 (data not shown) on a background of extremely high p16^INK4a expression. This allowed analysis of different INK4-CDK4 interactions in a setting independent of p21^CIP1.

View larger version (63K): [in this window] [in a new window]   FIG. 7.   CDK4 complexes in 293 cells. (A) 293 whole-cell extract was immunoprecipitated with the antisera shown. Immune complexes were separated by SDS-PAGE and immunoblotted with a combination of CDK4 and CDK6 sera. Asynchronously growing cultures of 293 cells were pulse-labelled with [35S]methionine and chased with medium containing unlabelled amino acids. At the time points indicated, cell extracts were prepared and immunoprecipitated with specific p16INK4a (B) or p18INK4c (C) sera. Immune complexes were separated by SDS-PAGE in a 12% gel. IP, immunoprecipitation.

Overexpression of p21^CIP1 in 293 cells restores p21^CIP1-cyclin D-CDK4 ternary complexes. The experiments outlined above demonstrate that CDK4 contributes to several independent complexes within cells and suggest that CDK4 may be involved in a dynamic equilibrium between INK4-bound and p21^CIP1- or p27^KIP1-bound states. To test this hypothesis, 293 cells were infected with recombinant adenovirus vectors expressing p21^CIP1 under the control of a cytomegalovirus promoter (Adp21^CIP1), or with vector control viruses (AdVec) (49). At 24 h postinfection, lysates were prepared and equivalent amounts were subjected to immunoprecipitation with D2, p21^CIP1, or normal rabbit serum. Following SDS-PAGE, the immune complexes were immunoblotted with D2 sera. In AdVec-infected lysates, low levels of cyclin D2 were detected in the D2 immunoprecipitate, and no cyclin D2 was detectable in p21^CIP1 immunoprecipitates (Fig. 8A). Following infection with Adp21^CIP1, the total level of cyclin D2 detected in D2 immune complexes appeared elevated, suggesting that the introduction of p21^CIP1 affected the steady-state levels of this protein (Fig. 8A). Significantly, in the p21^CIP1 immunoprecipitate from Adp21^CIP1-infected lysate, cyclin D2 was present at a level similar to that seen in the adjacent D2 immune complex, providing evidence that the majority of the cyclin D2 present in these cells was now associated with p21^CIP1. Next, we tested for the presence of CDK4 and p21^CIP1 in D2 and D3 immunoprecipitates from similarly infected lysates. As expected, in AdVec-infected cells no CDK4 was detectable in D2 or D3 immune complexes, whereas in lysate prepared from Adp21^CIP1 infected cells, CDK4 was efficiently coimmunoprecipitated with cyclins D2 and D3 (Fig. 8B, upper panel). Aliquots of whole-cell extract separated on the same gel demonstrate that equivalent levels of CDK4 were present in both lysate preparations. The blot was subsequently reprobed with p21^CIP1 sera (Fig. 8B, lower panel). As expected, p21^CIP1 was readily detected in Adp21^CIP1- but not AdVec-infected whole-cell extracts. Significantly, in Adp21^CIP1 lysates, p21^CIP1 was coimmunoprecipitated with cyclin D2-CDK4 and cyclin D3-CDK4 complexes. These data suggest that p21^CIP1 is able to stimulate the formation of cyclin D-CDK4-p21^CIP1 ternary complexes, even in the presence of excess p16^INK4a. To confirm this, whole-cell extract prepared from Adp21^CIP1-infected 293 cells was fractionated by gel filtration and immunoprecipitated with p21^CIP1 sera. Immune complexes were separated by SDS-PAGE and immunoblotted with D2. Cyclin D2 was found associated with p21^CIP1 in a 150- to 170-kDa complex (Fig. 8C), which is in good agreement with similar complexes detected previously in EL4 cells (Fig. 5).

View larger version (42K): [in this window] [in a new window]   FIG. 8.   Stimulation of cyclin D-CDK4 interaction by p21CIP1. (A) 293 cells were infected with AdVec or Adp21CIP1 adenoviruses, and lysates were immunoprecipitated with D2, p21CIP1, or normal rabbit sera (NR), as indicated. Following SDS-PAGE, the immune complexes were probed with D2 sera. (B) Similar infections were immunoprecipitated with D2, D3, or normal rabbit sera. Immune complexes were separated by SDS-PAGE alongside equivalents of whole-cell extract prepared from the same samples. Following transfer, the membrane was immunoblotted with CDK4 serum (upper panel) or p21CIP1 serum (lower panel). (C) Adp21CIP1 293 cell lysate was separated on Superdex 200. Fractions were collected and immunoprecipitated with p21CIP1 sera. Immune complexes were separated by SDS-PAGE and immunoblotted with D2 serum. IP, immunoprecipitation.

	DISCUSSION

Top Abstract Introduction Materials and methods Results Discussion References

Gel filtration analyses of CDK4 and CDK6. Both CDK4 and CDK6 form large multimeric complexes with chaperone molecules such as HSP90 and CDC37 that elute at ~450 kDa following gel filtration (6, 27, 47). Similarly, both kinases coelute with the peak of cyclin D in complexes with a molecular mass of 150 to 170 kDa that are capable of efficiently phosphorylating a GST RB substrate in vitro. A major difference between the CDK4 and CDK6 gel filtration profiles was the apparent lack of a 55-kDa CDK4 complex. We had previously demonstrated that the 55-kDa CDK6 complex comprises CDK6-p19^INK4d complexes (27). The gel filtration observations implied that CDK4 did not associate with the INK4 proteins present in the cell lines analyzed. Consistent with this, CDK4 was not found in p15^INK4b, p18^INK4c, or p19^INK4d immune complexes isolated from a variety of cells, indicating that although these inhibitors bound CDK6, they did not form complexes with CDK4 that were detectable on Western blots. A common feature of the immortalized cell lines analyzed is a lack of p16^INK4a expression. When we examined WI38 primary diploid fibroblasts, CDK4 was identified in association with p16^INK4a in 55-kDa complexes that had properties similar to those of the various 55-kDa CDK6-INK4 complexes. This correlation suggests that p16^INK4a alone can titrate CDK4 into inactive binary complexes in asynchronously growing cells.

INK4-CDK4 interactions. The INK4 protein, p18^INK4c, was originally described as a CDK6-specific inhibitor because of a lack of association with CDK4 in CEM cells (9). However, p18^INK4c overexpression is sufficient to arrest cells that express CDK4 as well as CDK6 in a pRB-dependent manner, implying that p18^INK4c is capable of inhibiting the activity of CDK4 under certain circumstances (9, 18). The apparent inability of certain INK4 family members to bind CDK4 in vivo is surprising, since in vitro, all four INK4 family members associate with CDK4 and CDK6 equally well (Fig. 3). When we analyzed interactions between newly synthesized components in pulse-chase experiments, it became apparent that INK4 molecules such as p18^INK4c do indeed associate with CDK4 in cells, but such complexes are short-lived in comparison to analogous CDK6-containing complexes. The transient nature of p18^INK4c-CDK4 association in these cells provides a possible explanation for the lack of detectable interactions on Western blots. Significantly, when we examined p18^INK4c interactions in cell lines such as 293 that lack cyclin D-CDK complexes, p18^INK4c was found associated with both CDK4 and CDK6 in stable binary complexes indistinguishable from similar p16^INK4a complexes (Fig. 7). It is possible, therefore, that cellular context influences the distribution of CDK4 and CDK6, with CDK4 being particularly sensitive to changes in subunit concentration. The exact mechanism that determines the half-lives of the different INK4-CDK4 complexes is unclear at present, although it seems likely that CDK4 interactions are strongly influenced by intracellular levels of D-type cyclins and p21 family members.

p21^CIP1 as a holoenzyme stabilizer. Although CDK4 does not normally form long-half-life associations with p18^INK4c, it is found in steady-state complexes containing D-type cyclins in a variety of proliferating cell lines that retain functional pRB (Fig. 1). Upon analysis, such complexes were also found to contain p21 family members (Fig. 5). Moreover, following immunodepletion experiments it was apparent that the majority of cyclin D-CDK4 complexes are associated with p21^CIP1 and p27^KIP1. These findings extend previous observations (20) and imply that all detectable cyclin D-CDK complexes in asynchronously growing cells are associated with either p21^CIP1 or p27^KIP1 in 150- to 170-kDa complexes that are kinase active (Fig. 1G) (33). Additionally, pulse-chase analyses demonstrated that CDK4 has a long-half-life interaction with complexes containing p21^CIP1 (Fig. 6), suggesting that p21 family members might influence ternary cyclin D-CDK interactions via stabilization. In agreement with this, expression of p21^CIP1 can reconstitute detectable cyclin D-CDK4-p21^CIP1 ternary complexes in 293 cells, even in the presence of excess p16^INK4a (Fig. 8). We attempted to discover whether enforced expression of p21^CIP1 led to a concomitant decrease in the level of p16^INK4a-CDK4 or p18^INK4c-CDK4 complexes, but were unable to definitively demonstrate such a decrease by immunoprecipitation-Western blot analyses or pulse-chase approaches (data not shown). It is likely that the extremely high level of INK4 activity (p16^INK4a in combination with p18^INK4c) in 293 cells continually interferes with cyclin D-CDK-p21^CIP1 assembly, resulting in a process that is detectable but intrinsically inefficient. We estimate that the majority of CDK4 remains associated with p16^INK4a-p18^INK4c in this experimental system. Also, because p21^CIP1 can interact with many other cyclin-CDK complexes, the specific concentration available to contribute to cyclin D-CDK assembly may actually be very low.

Subunit arrangement and tumorigenesis. CDK inhibitors play pivotal regulatory roles during the mammalian cell cycle by restraining the activity of cyclin-CDK complexes that might otherwise lead to inappropriate hyperphosphorylation of the RB tumor suppressor gene product. Because of this, all CDK inhibitors have been considered as candidate tumor suppressor genes. However, data compiled by numerous research groups from many different tumor types have demonstrated that p16^INK4a alone is commonly affected in human cancers. The apparent inability of INK4 family members other than p16^INK4a to suppress cyclin D-CDK4 association during normal proliferation, coupled with observations that p21 family members promote the formation of active cyclin D-CDK4 complexes during a mitogenic response, provides potential explanations as to why p16^INK4a alone is specifically targeted during tumorigenesis.