Phosphorylations of Cyclin-dependent Kinase 2 Revisited Using Two-dimensional Gel Electrophoresis*

To control the G1/S transition and the progression through the S phase, the activation of the cyclin-dependent kinase (CDK) 2 involves the binding of cyclin E then cyclin A, the activating Thr-160 phosphorylation within the T-loop by CDK-activating kinase (CAK), inhibitory phosphorylations within the ATP binding region at Tyr-15 and Thr-14, dephosphorylation of these sites by cdc25A, and release from Cip/Kip family (p27kip1 and p21cip1) CDK inhibitors. To re-assess the precise relationship between the different phosphorylations of CDK2, and the influence of cyclins and CDK inhibitors upon them, we introduce here the use of the high resolution power of two-dimensional gel electrophoresis, combined to Tyr-15- or Thr-160-phosphospecific antibodies. The relative proportions of the potentially active forms of CDK2 (phosphorylated at Thr-160 but not Tyr-15) and inactive forms (non-phosphorylated, phosphorylated only at Tyr-15, or at both Tyr-15 and Thr-160), and their respective association with cyclin E, cyclin A, p21, and p27, were demonstrated during the mitogenic stimulation of normal human fibroblasts. Novel observations modify the current model of the sequential CDK2 activation process: (i) Tyr-15 phosphorylation induced by serum was not restricted to cyclin-bound CDK2; (ii) Thr-160 phosphorylation engaged the entirety of Tyr-15-phosphorylated CDK2 associated not only with a cyclin but also with p27 and p21, suggesting that Cip/Kip proteins do not prevent CDK2 activity by impairing its phosphorylation by CAK; (iii) the potentially active CDK2 phosphorylated at Thr-160 but not Tyr-15 represented a tiny fraction of total CDK2 and a minor fraction of cyclin A-bound CDK2, underscoring the rate-limiting role of Tyr-15 dephosphorylation by cdc25A.

The major events of the eukaryotic cell cycle depend on the sequential and ordered formation, activation (by phosphorylation and/or dephosphorylation) and then inactivation of different complexes of cyclin-dependent kinases (CDKs). 1 CDK2 plays an essential role in controlling the G 1 /S transition and the progression through the S phase: it participates in the inactivating phosphorylations of pRb and related p130 (1)(2)(3)(4), and it phosphorylates other key substrates whose activities are necessary to trigger and organize the DNA synthesis phase while preventing DNA re-replication (5)(6)(7)(8)(9). Because the critical roles of CDK2, both in positive and negative mitogenic controls of cell cycle and as the end point target of DNA damage checkpoint mechanisms, have been well established (10 -13), understanding the details of CDK2 regulation is of fundamental importance.
Based on the cdc2/CDK1 activation model (14 -16) and on structural (17) and enzymatic studies (18 -24), the consensual framework of CDK2 activation is the following. The binding of a cyclin partner (cyclin E at the G 1 /S transition, cyclin A during the S phase) confers a low basal activity to the cyclin⅐CDK2 complex and enables subsequent phosphorylation of CDK2 by the CDK-activating kinase (CAK/cyclin H-CDK7) on a conserved threonine residue in the activation loop (Thr-160), which is essential for proper alignment of the kinase domain. Further phosphorylation within the ATP binding region of CDK2, at Tyr-15 by the Wee1 kinase and, to a lesser extent, at Thr-14 inhibits the kinase activity, which can be restored through dephosphorylation by the dual-specificity cdc25A phosphatase (25,26). The activity of cyclin⅐CDK2 complexes is also negatively regulated by the binding of inhibitory proteins belonging to the Cip/Kip family (10) (p27 kip1 , rapidly degraded consecutively to Thr-187 phosphorylation by the cyclin E⅐CDK2 complexes (27), and p21 cip1 ). In vitro studies have indicated that p27 and p21 interfere with CAK-induced phosphorylation of CDK2 (21,28,29), which is supported by crystal structure analysis of the p27⅐cyclin A⅐CDK2 complex (30). p27 also inhibits the activity of cyclin A⅐CDK2 by other mechanisms, including insertion in the catalytic cleft of CDK2 (30). In human fibroblasts, low stoichiometry p21-binding was nevertheless found to support rather than prevent the catalytic activity of cyclin A⅐CDK2 complexes (31,32). This was not observed by others (33,34) and questioned by an in vitro investigation (35).
The precise relationship between the different inhibitory and activating phosphorylations of CDK2, and the influence of cyclins and CDK inhibitors on these phosphorylations, have never been investigated in the context of endogenous proteins in a normal human cell. Here, we introduce the utilization of the high resolution power of the two-dimensional gel electrophoresis, combined to Tyr-15-or Thr-160-phosphospecific antibodies, to demonstrate the relative proportions of the different CDK2 phosphoforms and their respective association with cyclin E, cyclin A, p21 and p27, during the mitogenic stimulation * This study was supported in part by grants from the National Fund for Scientific Research (FNRS, Belgium), the Belgian Fund for Scientific Medical Research, and the Fortis Bank cancer research fund. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ Fellows of the Fonds pour la Formation à la Recherche dans l'Industrie et l'Agriculture.
* of normal human diploid fibroblasts. Some unexpected observations prompt a re-evaluation of the accepted sequence of regulatory events involved in the complex activation process of CDK2.

EXPERIMENTAL PROCEDURES
Cell Culture-Human diploid fibroblasts (IMR-90) (American Type Culture Collection, Manassas, VA) were obtained at population doubling 24 and grown for not more than 40 total population doublings in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with antibiotics and 10% fetal calf serum (FCS) as described previously (36). After reaching semiconfluency, they were synchronized by starvation in 0.2% FCS for 3 days. Quiescent cells were then growth stimulated by addition of FCS (20%). Bromodeoxyuridine (BrdUrd) labeling of DNAreplicating cells was obtained at the end of the mitogenic stimulation by a 30-min incubation in the presence of 10 Ϫ4 M BrdUrd and detected as described (37).
Alternatively, for the analysis of the two-dimensional gel electrophoresis pattern of the whole population of CDK2, cells in 10-cm Petri dishes were washed with phosphate-buffered saline and scraped in 200 l of denaturing lysis buffer (50 mM Tris-HCl (pH 7.5), 0.6% SDS, 10 mM DTT, 50 mM NaF, 100 M vanadate, and protease inhibitors), boiled for 5 min, and frozen. Cell lysates containing 300 -500 g of protein were cleared by centrifugation and diluted by adding five volumes of radioimmune precipitation assay buffer without SDS (10 mM Tris-HCl (pH 7.5), 1% sodium deoxycholate, 1% Nonidet P-40, 150 mM NaCl, 50 mM NaF, 100 M vanadate, and protease inhibitors). They were then subjected to immunoprecipitation as above, except that the M2 polyclonal antibody (Santa Cruz Biotechnology) was used for CDK2.
Gel Electrophoresis and Western Blotting-For two-dimensional gel electrophoresis separations, immunoprecipitated proteins were denatured in a buffer containing 7 M urea and 2 M thiourea. Proteins were separated by isoelectric focusing using the IPGphor apparatus from Amersham Biosciences and active in-gel rehydration as described (38) on immobilized linear pH gradient (pH 3-10) (IPG) strips (Amersham Biosciences). Isoelectric focusing was performed for a total of 30,000 Vh, starting at 200 V and gradually raising the voltage to 8000 V. The IPG strip was then equilibrated for 15 min in 50 mM Tris (pH 8.5), 6 M urea, 30% glycerol, 1% SDS, and 1% DTT, and for 15 min in 50 mM Tris, 6 M urea, 30% glycerol, 1% SDS, and 5% iodoacetamide before loading onto SDS-polyacrylamide slab gels (12.5%) for separation according to molecular mass. After transfer on polyvinylidene difluoride membranes, CDK2 was immunodetected using the phosphospecific-CDK2/CDK1 (Thr-160 or Tyr-15) antibodies from Cell Signaling Technology (Beverly, MA) or the M2 polyclonal CDK2 antibody (Santa Cruz Biotechnology); membranes detected with the phosphospecific antibodies were reprobed with the M2 polyclonal antibody for detection of total CDK2. In some cases, long exposures were required to detect minor phosphorylated forms of CDK2 from low abundance complexes, which markedly enhanced the appearance of diffuse "background" spots. These very characteristic spots were clearly nonspecific, because exactly the same spots were obtained when other polyclonal antibodies were used both for the immunoprecipitation and the immunodetection (e.g. CDK4 or p27 antibodies). In that case, to improve the visibility of the specific CDK2 spots, the program Adobe Photoshop was used to decrease the gray levels of the background spots in the respective figure panels (as indicated in the figure legends).
Whole cell extract proteins were separated according to molecular mass on SDS-PAGE (7, 10 or 12%) and immunodetected after Western blotting. Equal amounts of protein were loaded for each condition. The phosphospecific and total CDK2 antibodies were as above. The JG39 cyclin A polyclonal antibody and the HE-12 cyclin E and MO.1.1 CDK7 monoclonal antibodies were kind gifts of Julian Gannon and Tim Hunt, and Jiri Bartek, respectively. The p21, p27, and p38 MAP kinase polyclonal antibodies were from Santa Cruz Biotechnology, and the cdc25A DCS-121 monoclonal antibody was from NeoMarkers (Fremont, CA). All the experiments were reproduced at least twice with identical results.

CDK2 Phosphorylations and CDK2
Regulators during Mitogenic Stimulation of Human Fibroblasts-A maximum of DNAsynthesizing cells was observed 20 h after serum stimulation of quiescent IMR-90 cells (Fig. 1A). CDK2 and CDK2 regulatory proteins were analyzed from whole cell extracts from quiescent unstimulated (control, Cont) cells or from cells that were stimulated by serum for 16 h or 24 h, i.e. when a maximum of cells were in early S-phase or late S-phase/G 2 phase, respectively (Fig. 1B). CDK2 was present in quiescent cells and gradually accumulated after stimulation. CDK2 activating (Thr-160) or inhibitory (Tyr-15) phosphorylations were detected using phosphospecific antibodies, which also recognize the corresponding Thr-161-or Tyr-15-phosphorylated forms of cdc2/CDK1 (Fig.  1B). The Thr-160 phosphorylation of CDK2 was already detected in quiescent cells and similarly increased after 16 and 24 h of stimulation, preceding the Thr-161 phosphorylation of cdc2 only observed at 24 h. By contrast, the Tyr-15 phosphorylation was almost undetectable in quiescent cells and appeared after cell stimulation on both CDK2 and cdc2, being maximum at 24 h. Cyclin E was detected in quiescent cells and was maximally expressed 16 h after serum stimulation. It thus preceded the accumulation of cyclin A, which was absent from quiescent cells and maximally expressed at 24 h. Like cyclin E, p21 cip1 was present in quiescent cells and transiently enhanced 16 h after serum stimulation. By contrast, the presence of p27 kip1 progressively decreased after cell stimulation. Cdc25A displayed several forms, which were increased or induced after serum stimulation. CDK7 was present in quiescent cells and slightly enhanced by serum stimulation. These different observations are in perfect agreement with previous reports from IMR-90 cells and other normal human fibroblasts (5, 20, 33, 36, 39 -41).
In serum (16 h)-stimulated IMR-90 cells, a very large fraction of CDK2 was associated with p21 ( Fig. 1C). A much smaller fraction of CDK2 was bound to p27. Serum increased not only the presence of cyclin E⅐CDK2 complexes, but also the association of CDK2 with p21 and p27 (Fig. 1C). Both the inhibitory Tyr-15 and activating Thr-160 phosphorylations of CDK2 were found not only in cyclin E⅐CDK2 complexes, but also in CDK2 bound to p21 or p27 (Fig. 1C).
Two-dimensional Gel Electrophoresis Analysis of the Phosphorylation Pattern of Total CDK2-CDK2 was immunoprecipitated (using the M2 CDK2 antibody) from quiescent unstimulated cells (Cont) or cells stimulated by serum for 16 or 24 h, and its different forms were resolved by isoelectric focusing (pH 3-10 linear gradient) followed by SDS-PAGE and then detected by Western blotting using either the M2 CDK2 antibody, or the Thr-160 or Tyr-15 phosphospecific antibodies (Fig.  2). After immunodetection with the phosphospecific antibodies, membranes were systematically reprobed with the total CDK2 M2 antibody to ascertain the identity of the different forms and the homogeneity of the blotting transfer. In stimulated cells, multiple forms of CDK2 were distributed within an interval of 2.3 pH units (assuming the linearity of the supplied pH gradient gels) (Fig. 2). The most basic, and most abundant, form of CDK2 (form 0) focused at approximately pH 9, which is close to the calculated isoelectric point of human CDK2 (8.80; as calculated according to Ref. 42  Thr-160ϩTyr-15). Considering that phosphate groups produce a two-charge isoelectric point shift above pH 7 (PO 4 2Ϫ ) or a one-charge shift below pH 6 (HPO 4 Ϫ ) (43, 44), CDK2 phosphoforms 2, 4, and 5/6 were expected to contain one, two, or three phosphorylations, respectively. The minor intermediate spots could be explained by a small proportion of phosphate group in the HPO 4 Ϫ form or by other enzymatic or non-enzymatic covalent modifications (44). Together, data on the isoelectric points and the detection by the phosphospecific antibodies allowed us to identify CDK2 forms 2T and 2Y as mono-phosphorylated at Thr-160 and Tyr-15, respectively; form 4T,Y as phosphorylated at both Thr-160 and Tyr-15; form 4Y as phosphorylated at Tyr-15 and presumably Thr-14; and minor forms 5T,Y and 6T,Y as 3-fold phosphorylated at Thr-160, Tyr-15, and presumably Thr-14. This determination is in perfect agreement with a previous investigation of CDK2 phosphorylations by 32 P i incorporation, tryptic peptide, and mutagenesis analyses (18), which did not detect any other important phosphorylations, and demonstrated that the major phosphorylated sites are Tyr-15 and Thr-160, whereas the Thr-14 phosphorylation is restricted to a fraction of Tyr-15-phosphorylated CDK2. An identical two-dimensional gel electrophoresis pattern of CDK2 was obtained from primary cultures of canine thyroid epithelial cells (not shown), indicating that the present identification of CDK2 phosphoforms might apply to a wide variety of systems.
In control quiescent IMR-90 cells, CDK2 was essentially not phosphorylated. 2 Only a small proportion of mono-Thr-160phosphorylated CDK2 was detected (Fig. 2). The inhibitory Tyr-15 phosphorylation was induced by serum and was mostly represented by the mono-phosphorylated 2Y form. Smaller proportions of twice phosphorylated CDK2 forms containing either both the Tyr-15 and Thr-160 phosphorylations (form 4T,Y) or the Tyr-15 phosphorylation but not the Thr-160 phosphorylation (form 4Y, putatively containing the Thr-14 phosphorylation), were also observed 16 h and even more 24 h after serum stimulation (Fig. 2). In control quiescent cells, the detection of a trace of the twice phosphorylated 4T,Y form required long exposures of the blots (Fig. 2). The activating Thr-160 phosphorylation was also strongly increased by serum stimulation. Surprisingly, it was essentially associated with Tyr-15 phosphorylation (4Y,T and other minor Y,T forms). The potentially active mono-Thr-160-phosphorylated form (2T) thus represented only a small fraction of Thr-160-phosphorylated CDK2 and an extremely minor fraction of total CDK2 (Fig. 2).
Phosphorylations of Cyclin E-bound CDK2-Phosphospecific and total CDK2 antibodies were applied to two-dimensional gel separations from cyclin E co-immunoprecipitations (Fig. 3). In control quiescent cells, the twice Thr-160ϩTyr-15-phosphorylated forms were the only CDK2 forms associated with cyclin E, as detected after long exposures of the blots. This complex should be present at a very low abundance, because the 4T,Y CDK2 form was faintly detectable only after long blot expo- 2 The most basic and most abundant CDK2 form 0 was therefore assumed to be unphosphorylated in agreement with (18), although this was not directly assessed in the present study.
FIG. 1. CDK2 and CDK2 regulatory proteins in human IMR-90 cells stimulated to re-enter the cell cycle. G 0 cells (Cont) were reactivated by addition of 20% serum and sampled at the indicated time points. A, DNA synthesis was monitored by counting the fraction (mean Ϯ range of duplicate dishes) of nuclei having incorporated BrdUrd during the last 30 min of the stimulation. B, Western blotting analyses of CDK2, P-Thr-160/P-Thr-161-phosphorylated forms of CDK2/cdc2 (P-T160 CDK2 and P-T161 CDK2), P-Tyr-15-phosphorylated forms of CDK2 and cdc2 (P-Y15 CDK2), cyclin E, cyclin A, p21, p27, cdc25A, and CDK7 from whole cell extracts. p38 MAP kinase was detected as a loading control. C, the presence of p21, CDK2, and its P-Thr-160 and P-Tyr-15 phosphorylations was detected by Western blotting in (co) immunoprecipitates (IP) of p21, p27, cyclin E, or CDK2.
Phosphorylations of Cyclin A-bound CDK2-In the present experiments, cyclin A was absent from control quiescent cells and maximally expressed 24 h after serum stimulation (Fig.  1B). The phosphorylations of cyclin A-bound CDK2 were thus only analyzed at this time point (Fig. 4). At variance with total CDK2 and cyclin E⅐CDK2 complexes, almost no unphosphorylated CDK2 was associated with cyclin A (Fig. 4). Although a small proportion of mono-Tyr-15-phosphorylated CDK2 (2Y) was detected, the immense majority of Tyr-15 phosphorylation was associated with the Thr-160 phosphorylation in cyclin A complexes (Fig. 4). The potentially active mono-Thr-160-phosphorylated form (2T, 1T) represented only a minor fraction of cyclin A-bound CDK2 (Fig. 4), in contrast to cyclin E⅐CDK2 complexes where this form was the most abundant one (Fig. 3).
Phosphorylations of p21-bound CDK2-A large fraction of CDK2 was associated with p21 as judged from comparisons of CDK2 amounts (co)precipitated by p21 and CDK2 antibodies (Fig. 1C). At variance with cyclin E/A⅐CDK2 complexes, p21bound CDK2 was clearly observed both in quiescent control and serum-stimulated IMR-90 cells (Fig. 5, A and B). Moreover, it was predominantly represented by the unphosphorylated 0 form, even in serum-stimulated cells (Fig. 5B). In control cells, the small proportion of Thr-160 phosphorylation seen in p21 complexes (Fig. 5, A and B) was similar to what was observed in total CDK2 (Fig. 2). Some p21⅐CDK2 complexes should thus exist without being associated with cyclin E or cyclin A. They could be associated with cyclin D1 (45), because cyclin D1bound CDK2 only consisted of the unphosphorylated form 0 and of a very small proportion of the mono-Thr-160-phosphorylated form 2T, both in unstimulated cells and 16 h after FIG. 2. The potentially active mono-Thr-160-phosphorylated CDK2 form represents an extremely minor fraction of total CDK2. Western blotting analysis of the two-dimensional gel electrophoresis pattern of total CDK2 (immunoprecipitated with an anti-CDK2 antibody) from quiescent unstimulated IMR-90 cells (Cont) or cells stimulated by 20% serum for 16 or 24 h. The immunodetection was performed using a CDK2 antibody (CDK2), or the Thr-160 or Tyr-15 phosphospecific antibodies (P-T160 or P-Y15, respectively). In the two 24 h panels, the same membranes were first detected using either the P-Thr-160 or P-Tyr-15 phosphospecific antibodies and then reprobed using the total CDK2 antibody. Two different exposures (1 min and 5 min) of the Cont panel are shown. The 1 min exposures of Cont, 16 h, and 24 h panels can be compared. To improve the visibility of the specific CDK2 spots, the program Adobe Photoshop was used to decrease the gray levels of the background spots in the P-T160 and P-Y15 panels (see "Experimental Procedures"). The different CDK2 forms were numbered according to their position relative to a computed scale (bottom) of isoelectric point shifts generated by the addition to CDK2 amino acid sequence of entire numbers (n) of negative charges (see "Results") and labeled according to their demonstrated phosphorylation(s) (T, Thr-160 phosphorylation; Y, Tyr-15 phosphorylation).

FIG. 3. Phosphorylations of cyclin E-bound CDK2. IMR-90 cells remained quiescent (Cont) or were stimulated by 20% serum for 16 h.
Lysates were subjected to co-immunoprecipitation with an anti-cyclin E antibody, separated by two-dimensional gel electrophoresis, and the Western blotting membranes were immunodetected first with the Thr-160 or Tyr-15 phosphospecific antibody (P-T160 or P-Y15, respectively), and then reprobed with the total CDK2 antibody (CDK2). Different exposure times were used for the Cont and 16 h panels. To improve the visibility of the specific CDK2 spots, Adobe Photoshop was used to decrease the gray levels of the background spots (see "Experimental Procedures"). Notice the enrichment of the potentially active mono-Thr-160-phosphorylated CDK2 form 2T at 16 h. serum stimulation (data not shown). Aside from the association of unphosphorylated CDK2 with p21, in serum-stimulated cells, p21 complexes contained cyclin E or cyclin A (Fig. 5A) and the pattern of p21-bound CDK2 phosphorylations (Fig. 5B) much resembled the CDK2 phosphorylations that are enriched in cyclins E/A complexes (Figs. 3 and 4). The activating Thr-160 phosphorylation was unexpectedly present in p21-containing complexes, associated or not with the Tyr-15 phosphorylation, and no mono-Tyr-15-phosphorylated CDK2 was detected (Fig.  5, A and B). As compared with the Thr-160ϩTyr-15-phosphorylated form 4T,Y, the mono-Thr-160-phosphorylated form (2T) was more abundant at 16 h than at 24 h (Fig. 5B). The distribution of p21-bound CDK2 phosphoforms was thus more similar to the CDK2 pattern associated with cyclin E at 16 h (Fig.  3) and to the pattern associated with cyclin A at 24 h (Fig. 4), consistently with the relative abundance of both cyclins at these time points (Fig. 1B).
Phosphorylations of p27-bound CDK2-We repeatedly failed to detect CDK2 in p27 immunoprecipitates from control unstimulated IMR-90 cells, as also mentioned by others (5). In serum-stimulated cells, p27⅐CDK2 complexes were also less abundant than p21⅐CDK2 complexes (Fig. 1C and Ref. 5). At least part of the p27 complexes contained cyclin E, as shown by reprobing with a p27 antibody the 2D gel blots of cyclin E immunoprecipitates (data not shown). The phosphorylation profile of p27-bound CDK2 16 h after serum stimulation (Fig. 6) was quite similar to the phosphorylation pattern of cyclin Ebound CDK2 (Fig. 3), with a predominant presence of the mono-Thr-160-phosphorylated form (2T). As in the case of p21bound CDK2, this surprising observation contrasted with the concept that Cip/Kip proteins inhibit the activity of CDK2 in part by preventing its phosphorylation by CAK. A small proportion of the mono-Tyr-15-phosphorylated form was also detected (Fig. 6), as in cyclin A complexes (Fig. 4).
Comparison of Total and Complexed CDK2 from Serumstimulated Fibroblasts-As illustrated in Fig. 7 and summarized in Table I, the main novel observations from this study are: (i) the enrichment of Thr-160-phosphorylated CDK2 forms, not only in cyclin E complexes and even more in cyclin A complexes (where unphosphorylated CDK2 was almost undetectable), but also in p21 and p27 complexes; (ii) the predominant presence in total CDK2 of the Tyr-15 phosphorylation in forms devoid of the Thr-160 phosphorylation (2Y, 3Y, and 4Y), whereas in the different complexes the Tyr-15 phosphorylation was restricted totally or almost totally to Thr-160-phosphorylated forms (3T,Y, 4T,Y, and 6T,Y). This last intriguing obser- FIG. 4. Phosphorylations of cyclin A-bound CDK2. Two-dimensional gel separation of a co-immunoprecipitation with a cyclin A antibody from IMR-90 cells stimulated with serum for 24 h. Western blotting membranes were immunodetected first with the Thr-160 or Tyr-15 phosphospecific antibodies (P-T160 or P-Y15, respectively), and then reprobed with the total CDK2 antibody (CDK2).
FIG . 5. Phosphorylations of p21-bound CDK2. IMR-90 cells remained quiescent (Cont) or were stimulated by 20% serum for 16 or 24 h. Lysates were subjected to co-immunoprecipitation with an anti-p21 antibody and separated by one-dimensional (A) or two-dimensional gel electrophoresis (B). In B, the Western blotting membranes were immunodetected first with the Thr-160 or Tyr-15 phosphospecific antibody (P-T160 or P-Y15, respectively), and then reprobed with the total CDK2 antibody (CDK2). To improve the visibility of the specific CDK2 spots in the Cont panels, Adobe Photoshop was used to decrease the gray levels of the background spots (see "Experimental Procedures"). vation implies that the Tyr-15 phosphorylation induced by serum was not restricted to cyclin⅐CDK2 complexes and thus did not depend on cyclin binding, unlike the Thr-160 phosphorylation. It might also imply that in the different complexes the Tyr-15-phosphorylated forms could be preferentially recruited for Thr-160 phosphorylation. DISCUSSION This study constitutes the first analysis of CDK2 phosphorylations using two-dimensional gel electrophoresis. In onedimensional SDS-PAGE, the activating Thr-160 phosphorylation generates a downward electrophoretic shift (18), and thus it has been easily investigated in numerous studies, which recently identify it as a direct target of mitogenic controls (46 -48). By contrast, the inhibitory Tyr-15 phosphorylation does not affect the SDS-PAGE migration of CDK2 (18), which explains in part the paucity of published information concerning the relationship of this phosphorylation with the other regulatory mechanisms of CDK2, despite the strong evidence of the crucial role of the Tyr-15 phosphorylation emerging from the investigation of the cell cycle function of cdc25A (25) and the elucidation of DNA damage checkpoint mechanisms (49). The present characterization of the two-dimensional gel electrophoresis pattern of CDK2 using phosphospecific antibodies is fully consistent with the previous identification of major phosphorylation sites of human CDK2 (18). It allows one to distinguish the different Thr-160-and/or Tyr-15-phosphorylated forms of CDK2 and thus to separate for the first time the potentially active CDK2 phosphorylated at Thr-160 but not Tyr-15 (nor Thr- 14), which was demonstrated here as a very minor fraction of CDK2 in human fibroblasts. As discussed below, the visualization of the relative stoichiometry of the different (un)phosphorylated forms of CDK2 associated or not with cyclins and CDK inhibitors alters the current concept of the sequence of regulatory events leading to CDK2 activation.
Thr-160-phosphorylated CDK2 Is Associated with Cyclins and CDK Inhibitors-CAK (cyclin H-CDK7) has been initially reported to phosphorylate monomeric as well as cyclin-bound CDK2 in vitro (50). Furthermore, monomeric Cak1p-type CAKs from yeasts and plants exhibit a preference for cyclin-free CDKs (51). Nevertheless, the crystal structure study suggests that cyclin A-binding exposes the T-loop of CDK2 making it a better substrate for phosphorylation by CAK (52), in agreement with a more recent study clearly showing that mammalian CAK preferentially phosphorylates cyclin⅐CDK2 complexes (21), as previously demonstrated for cdc2 (14). The nuclear translocation of CDK2 (40,53) associated with cyclin⅐CDK2 complex formation should also favor CDK2 phosphorylation by permitting the colocalization with nuclear CAK (54). Moreover, cyclin binding protects Thr-160-phosphorylated CDK2 from dephosphorylation by the CDK-associated phosphatase KAP (55). The strong enrichment of Thr-160-phosphorylated forms of CDK2 in cyclin E/A complexes observed in the present study ( Fig. 7 and Table I) thus fully confirms in vivo the dependence of Thr-160 phosphorylation upon prior cyclin binding. Interestingly, cyclin binding might be the sole rate-limiting factor for phosphorylation by CAK in cyclin A complexes (because they contained almost exclusively CDK2 phosphorylated at Thr-160; Fig. 4) but not in cyclin E complexes from stimulated IMR-90 fibroblasts (which contained a significant proportion of unphosphorylated CDK2; Fig. 3), which is consistent with previous observations from these cells (33).
Because CAK is reported to be constitutively active (54,56), Thr-160 phosphorylation within cyclin⅐CDK2 complexes is generally believed to be limited by their association with p21 cip1 or p27 kip1 , which interfere with phosphorylation by CAK in vitro (21,28,29), possibly by hindering T-loop access (30). The present observations from intact cells challenge this concept essentially derived from in vitro experiments performed using bacterially produced p21 and p27, which are presumably unphosphorylated and possibly misfolded. Unexpectedly, Thr-160-phosphorylated CDK2 was largely associated with the CDK inhibitors p21 and p27 (Table I). Moreover, although a large proportion of cyclin A⅐CDK2 complexes contain p21 in proliferating human fibroblasts (32,39), almost all the CDK2 was phosphorylated at Thr-160 in cyclin A complexes (Fig. 4). Similarly, the inactive cyclin E⅐CDK2 complexes from unstimulated IMR-90 cells contain p21 (5) and/or p27 (40) (although we repeatedly failed to detect p27-bound CDK2 from quiescent cells, in agreement with Ref. 5), but the entirety of CDK2 from these cyclin E complexes was phosphorylated at Thr-160 (Fig.  3). Our results are thus more consistent with the paradoxical observation by some authors that the vast majority of active CDK2 is associated with p21 in normal human fibroblasts (31,32), implying that, at least at low stoichiometry binding, p21 should permit CDK2 phosphorylation by CAK. The abundant association of Thr-160-phosphorylated CDK2 with p27 ( Fig. 6) was even more intriguing. This is not to say that p27 is not a strong CDK2 inhibitor (10), because p27 also inhibits the activity of Thr-160-phosphorylated CDK2 bound to cyclin E (57), in agreement with the multiplicity of p27 inhibitory mechanisms suggested from the structural study of cyclin A⅐CDK2⅐p27 complexes (30). At this stage, we cannot formally exclude the possibility that p27 might have joined the cyclin E⅐CDK2 complex after CDK2 phosphorylation by CAK, because p27 associates with only a fraction of CDK2 complexes. Nevertheless, at variance with the in vitro evidence obtained with bacterially produced p21 and p27, our results clearly indicate that, at least in intact human fibroblasts, p21 and possibly p27 do not prevent the activity of CDK2 complexes by impairing the phosphorylation of CDK2 by CAK.
Tyr-15 Phosphorylation Is Not Restricted to Cyclin-bound CDK2-By analogy with the situation demonstrated for cdc2/ CDK1 (58), the Tyr-15 phosphorylation of CDK2 has been inferred to depend on cyclin binding, like the Thr-160 phosphorylation by CAK (18). Nevertheless, to our knowledge, this question has not been investigated directly. Here, the Tyr-15 phosphorylations of CDK2 and cdc2 were strongly induced by serum, but, unlike the Thr-160 phosphorylation of CDK2, the Tyr-15 phosphorylation of CDK2 was clearly independent of prior cyclin-binding. Indeed, Tyr-15 phosphorylation was FIG. 6. Phosphorylations of p27-bound CDK2. Lysates from 16h-serum stimulated IMR-90 cells were subjected to co-immunoprecipitation with anti-p27 antibodies and separated by two-dimensional gel electrophoresis, and Western blotting membranes were immunodetected first with the Thr-160 or Tyr-15 phosphospecific antibody (P-T160 or P-Y15, respectively), and then reprobed with the total CDK2 antibody (CDK2). To improve the visibility of the specific CDK2 spots in the phosphospecific antibody detections, Adobe Photoshop was used to decrease the gray levels of the background spots (see "Experimental Procedures"). mostly represented in total CDK2 by the mono-Tyr-15-phosphorylated form 2Y (Fig. 2), which was faint or undetectable in CDK2 complexed to cyclins E/A or Cip/Kip proteins (Table I).
The marked increase of Tyr-15 phosphorylation of CDK2 thus likely depends on the regulation of the Wee1 kinase (59), which is induced during cell cycle progression (20,60).
Tyr-15-phosphorylated CDK2 Is Preferentially Recruited for Thr-160 Phosphorylation-In HeLa cells, almost all of the phosphorylation of CDK2 on Tyr-15 occurs on CDK2 molecules that are also phosphorylated on Thr-160 (18). We have observed a different situation in IMR-90 fibroblasts. Although mono-Tyr-15-phosphorylated forms were abundantly detected in total CDK2 from stimulated cells, almost all of the Tyr-15 phosphorylation was associated with the Thr-160 phosphorylation in CDK2 complexed to cyclins or CDK inhibitors ( Fig. 7 and Table I). In CDK2 complexes associated with cyclin E and/or p21 or p27, it is thus apparent that the phosphorylation by CAK recruited the entirety of Tyr-15-phosphorylated CDK2. This contrasts with the persistence in these complexes of a significant proportion of unphosphorylated CDK2, which thus was less efficiently engaged by Thr-160 phosphorylation ( Fig. 7 and Table I). Moreover, the mono-Thr-160-phosphorylated CDK2 form might be generated, not through phosphorylation by CAK of unphosphorylated CDK2, but most likely through dephosphorylation of the Thr-160ϩTyr-15-phosphorylated forms by a cdc25 phosphatase. Indeed, the addition of cdc25A to co-immunoprecipitated cyclin E/A⅐CDK2 complexes considerably increases their catalytic activity (41), and the rapid disappearance of cdc25A following cell irradiation prevents most of CDK2 activity without affecting cyclin E⅐CDK2 complex formation (26), which demonstrates that CDK2 activity critically depends on its Tyr-15 dephosphorylation. The mechanism restricting phosphorylation by CAK to Tyr-15-phosphorylated CDK2 is unknown. Its physiological importance would be clearly to prevent a premature activation of cyclin⅐CDK2 complexes by CAK from bypassing the tight control by cdc25-dependent checkpoint pathways, such as those activated by DNA damage (12).
Activation of CDK2 during Mitogenic Stimulation-As designed, the present experiments only provide snapshots of the FIG. 7. Comparison of total and complexed CDK2 from serum stimulated fibroblasts. Same lysates from 16-h or 24-h serum-stimulated IMR-90 cells were subjected to (co-) immunoprecipitation with anti-CDK2, -cyclin E, -cyclin A, -p27, or -p21 antibodies, separated by twodimensional gel electrophoresis, and the resulting Western blots were immunodetected using the total CDK2 antibody.

TABLE I Summary of the main observations
The different CDK2 phosphoforms detected in the different (co-) immunoprecipitations (first column) from IMR-90 cells serum-stimulated for 16 and/or 24h or not (Cont), are divided into four categories: 0 (not phosphorylated at Thr-160 or Tyr-15), PY15 (phosphorylated at Tyr-15 but not Thr-160), PT160 (phosphorylated at Thr-160 but not Tyr-15), PT160 ϩ PY15 (phosphorylated at both Thr-160 and Tyr-15). For each CDK2 complex, the relative abundance of the main representative form (as indicated in boldface) of each category was scored from Ϫ (undetectable) to ϩϩϩ (the most abundant form). In quiescent unstimulated cells, CDK2 was mainly present as an unphosphorylated, uncomplexed form. A proportion of unphosphorylated CDK2 and the minor fraction of mono-Thr-160-phosphorylated CDK2 were also associated with p21 (Fig.  5) and cyclin D1 (not shown), but not with cyclins E/A. An even tinier fraction of CDK2, consisting only of the Thr-160ϩTyr-15-phosphorylated forms, was complexed with cyclin E and possibly p21. It is likely to be totally inactive (5,40). At variance with cyclin E-bound CDK2, CDK2 associated with cyclin D1 and p21 was only faintly phosphorylated at Thr-160, even in serum-stimulated cells (not shown), in agreement with previous reports indicating that cyclin D1 is an inhibitor for CDK2 that prevents its phosphorylation by CAK (45,61,62).
Upon serum stimulation, the abundance of cyclin E and CDK2 increases (Fig. 1B) and new cyclin E⅐CDK2 complexes are formed, perhaps depending in part on CDK2 nuclear import (40). The most impressive effect of serum was the appearance of the inhibitory Tyr-15 phosphorylation in both uncomplexed CDK2 and CDK2 associated with cyclins E/A and p21 or p27. The activating Thr-160 phosphorylation also increased. Unlike the Tyr-15 phosphorylation, it likely depended on the formation of cyclin E/A⅐CDK2 complexes, which may comprise p21 or p27. Within these complexes, CAK phosphorylates the entirety of Tyr-15-phosphorylated CDK2, but possibly not unphosphorylated CDK2 (as discussed above) (Fig. 8), thus constituting the important reservoir of pre-activated Thr-160ϩTyr-15-phosphorylated forms. Depending on serum stimulation, likely through activation of cdc25A (as possibly reflected by the appearance of an undetermined up-shifted form of this phosphatase (Fig. 1B)), Thr-160ϩTyr-15-phosphorylated CDK2 was then dephosphorylated in cyclin E complexes to generate the potentially active mono-Thr-160-phosphorylated CDK2 (Fig.  8), which became the predominant form associated with this complex at the G 1 /S transition (but still represents only a very minor fraction of total CDK2).
In contrast, paradoxically when the activity of cyclin A⅐CDK2 is maximum in cells in S-or G 2 phases (36), cyclin A-bound CDK2 and p21-bound CDK2 at 24 h remained largely phosphorylated at both Thr-160 and Tyr-15 with a surprisingly small proportion of mono-Thr-160-phosphorylated form (Figs. 4 and  5). This intriguing observation is consistent with the abrupt dephosphorylation at mitosis of most Tyr-15 residues in both cdc2 and CDK2 described by Gu and collaborators (18), which might be explained by the strong stabilization of cdc25A in mitotic cells (63). Though cdc25A can equally activate CDK2 bound to cyclin E or cyclin A in vitro (41), this might suggest that cyclin A⅐CDK2 complexes could be more resistant to dephosphorylation by cdc25A, perhaps due to an increased association with p21, which competes with cdc25A for binding to the same cyclin site (64). On the other hand, at S/G 2 phases (24 h after stimulation), cyclin A⅐CDK2 complexes are likely to be more abundant than the cyclin E⅐CDK2 complexes observed at the G 1 /S transition (16 h after serum stimulation), relative to unchanged levels of cdc25A (Fig. 1B). Even during unperturbed S phase progression (65,66), cdc25A presence and thereby activity are critically restricted by the checkpoint kinase Chk1, which is induced at the S to M phase and maximally activated 24 h after serum stimulation in human fibroblasts (67). Cdc25A might thus be especially rate-limiting for the dephosphorylation of cyclin A⅐CDK2. A fraction of cyclin A⅐CDK2 associates with DNA replication sites (68 -70) to phosphorylate various pre-replication complex proteins (9, 71) such as cdc6 (6,72), and thus to monitor ongoing DNA replication and control rereplication (8,73). We surmise that to tune the dynamic processivity of chromosome replication, the modulation of cyclin A⅐CDK2 activity by the S-phase checkpoint Chk1-cdc25A cascade (65,66) and/or through p21 release (13) might well need to be tightly restricted spatiotemporally, thus engaging at a given time only a minor fraction (as observed in Fig. 4) of the reservoir of pre-activated cyclin A⅐CDK2 complexes that would accumulate in a pan-nuclear manner through Tyr-15 phosphorylation, cyclin A-binding, and Thr-160 phosphorylation.
To conclude, the identification of the various CDK2 phosphoforms by two-dimensional gel electrophoresis has allowed for the first time the ability to distinguish the potentially active Thr-160-phosphorylated CDK2 from its Thr-160ϩTyr-15-phosphorylated precursor forms. Whereas Thr-160 phosphorylation appears to depend on both cyclin binding and prior Tyr-15 phosphorylation, the Tyr-15 dephosphorylation is underscored as the rate-limiting step of the complex activation process of cyclin⅐CDK2 complexes. This methodology should prove a valuable experimental tool in any future investigation of CDK2 regulation.