Cyclin A-CDK phosphorylation regulates MDM2 protein interactions

The product of the MDM2 gene interacts with and regulates a number of proteins, in particular the tumor suppressor p53. The MDM2 protein is likely to be extensively modified in vivo, and such modification may regulate its functions in cells. We identified a potential cyclin-dependent kinase (CDK) site in murine MDM2, and found the protein to be efficiently phosphorylated in vitro by cyclin A-containing complexes (cyclin A-CDK2 and cyclin A-CDK1), but MDM2 was either weakly or not phosphorylated by other cyclin-containing complexes. Moreover, a peptide containing a putative MDM2 cyclin recognition motif specifically inhibited phosphorylation by cyclin A-CDK2. The site of cyclin A-CDK2 phosphorylation was identified as Thr-216 by two-dimensional phosphopeptide mapping and mutational analysis. Phosphorylation of MDM2 at Thr-216 both weakens its interaction with p53 and modestly augments its binding to p19(ARF). Interestingly, an MDM2-specific monoclonal antibody, SMP14, cannot recognize MDM2 phosphorylated at Thr-216. Changes in SMP14 reactivity of MDM2 in staged cell extracts indicate that phosphorylation of MDM2 at Thr-216 in vivo is most prevalent at the onset of S phase when cyclin A first becomes detectable.


Introduction
The mdm2 gene was originally cloned from a spontaneously transformed mouse cell line 3T3DM (1). Its potential pro-oncogenic activity was supported by the observation that amplification or over-expression of MDM2 promotes tumorigenesis in NIH 3T3 or Rat2 cell lines (2). In cooperation with ras, MDM2 over-expression can also promote transformation of primary rodent fibroblasts (3). The finding that MDM2 is amplified in ~30% of human sarcomas (4) further suggests that it is a proto-oncogene that promotes cell proliferation.
One important function of the MDM2 protein is to down-regulate p53 protein levels and p53 activity (reviewed in 15,16). In response to various cellular stresses, active p53 protein regulates numerous effectors that are involved in cell cycle arrest, apoptosis or other cellular processes (reviewed in 17). The interaction between the N-terminus of MDM2 and p53 proteins is important for inhibiting p53-mediated transactivation (18,19) as well as for targeting p53 for degradation (20)(21)(22). In turn, p53 protein can bind to the promoter of MDM2 and activate MDM2 transcription (23). This feed-back loop between MDM2 and p53 fine-tunes the cellular responses after p53 activation (24). Another tumor suppressor protein, p19 ARF , can also bind to a 5 interactions of MDM2 with components of the basal transcription factor TFIID: the MDM2 Cterminal RING finger domain interacts with TAFII250 and the region close to the acidic domain interacts with TBP (10,11,40,41).
Cell cycle progression depends upon the formation of active cyclin-CDK complexes that are regulated by the synthesis and degradation of cyclins, their state of phosphorylation, and their association with inhibitory subunits (reviewed in [42][43][44][45]. Cyclin A, which forms complexes with CDK2 at the beginning of S-phase and CDK1 at the beginning of M-phase, is required for entry into S-phase, passage through G2 and mitosis (46)(47)(48)(49). The substrates of cyclin A-CDK2 include a number of proteins, including the p53 protein (50).
MDM2 is likely to be extensively regulated by phosphorylation. More than one third of the amino acids on MDM2 protein are either serine or threonine residues, and MDM2 protein is phosphorylated at multiple sites in vivo, especially in the N-terminus and the central acidic domain (51). It has been reported that ATM kinase (52), DNA-dependent protein kinase (DNA-PK)(53) and casein kinase 2 (CK2) (54,55) can phosphorylate MDM2 proteins and modulate MDM2 functions.
Murine MDM2 has at least one potential site for CDK phosphorylation.
We report here that cyclin A-CDK complexes are unique in their ability to efficiently phosphorylate MDM2 and that this phosphorylation affects the interactions of MDM2 with proteins. 7 polyacrylamide gels. The gels were either silver-stained or stained with Coomassie Blue to detect protein levels before exposure to X-ray film. Peptides used in the cyclin recognition motif experiments were synthesized by Synpep.

Two Dimensional Phosphopeptide Mapping
[ 32 P] labeled GST-MDM2 (1 µg) was phosphorylated by cyclin A-CDK2 as described above and then run on a 10% SDS-polyacrylamide gel and exposed to X-ray film. The phosphorylated GST-MDM2 full-length protein was then cut out and eluted from the gel slice in buffer containing 0.05M NH 4 HCO 3 (pH 7.3), 0.5% β-mercaptoethanol and 0.1% SDS. The eluted proteins were acid-precipitated and oxidized as described (59) A. Levine). The same blots were also probed with an anti-cyclin A polyclonal antibody H437 or an anti-cyclin B monoclonal antibody Ab-2 (both from Santa Cruz). Duplicate plates of cells at each time point were collected and processed for FACS analysis.

GST-MDM2 is preferentially phosphorylated by cyclin A containing CDK complexes
Examination of the sequence of murine MDM2 protein indicated that there are two possible CDK sites in the protein at residues 216 (TPSH) and 384 (TPLS). In fact, T216-P217 is ( Figure 2C, lane 5-8) even though, as expected, both kinases phosphorylated histone H1 to roughly the same extent. Since cyclin A has been shown to form a complex with CDK1 at the onset of mitosis (47), we then compared the abilities of purified cyclin A-CDK1 and cyclin B-CDK1 complexes to phosphorylate GST-MDM2 in vitro ( Figure 2D). Indeed, cyclin A-CDK1 phosphorylated GST-MDM2 and, like cyclin A-CDK2, did so more effectively than cyclin B-CDK1. Therefore, cyclin A containing CDK complexes are the most effective in phosphorylating MDM2 protein.

Cyclin A-CDK2 phosphorylates MDM2 at T216
To identify the CDK site(s) in MDM2, either T216 or T384 was mutated to alanine in full-length GST-MDM2, and wild-type and T216A or T384A MDM2 proteins were purified from bacteria and treated with cyclin A-CDK2 in mixtures containing [ 32 P]ATP. We then performed 2-D mapping of phosphopeptides generated from V8 protease digestion of phosphorylated wild-type or mutant proteins. In the map of the phosphorylated wild-type protein, a predominant phosphopeptide was observed ( Figure 3A, arrow). This spot was diminished when the CDK2 specific inhibitor roscovitine was added to the kinase reaction mixture ( Figure 3B), supporting the possibility that it is specifically phosphorylated by cyclin A-CDK2. Importantly, this phosphopeptide was absent in the 2-D map of the T216A mutant ( Figure 3C) but was present in the map of the T384A mutant ( Figure 3D). The additional minor phosphopeptides that were in the 2-D maps are most likely the result of cryptic phosphorylation by the CDK complex since they are largely suppressed by roscovitine. We conclude that cyclin A-CDK2 phosphorylates predominantly T216 in murine MDM2.

MDM2 monoclonal antibody SMP14 cannot recognize GST-MDM2 phosphorylated by cyclin A-CDK2 or cyclin A-CDK1 at T216
During an ELISA experiment intended as a control for the effect of phosphorylation of MDM2 on protein-protein interactions, we unexpectedly discovered that Mab SMP14, an MDM2 specific monoclonal antibody, displays reduced binding to cyclin A-CDK2 phosphorylated GST-MDM2 protein. This was surprising because this antibody was generated using a peptide spanning human MDM2 (HDM2) residues 154-167. Nevertheless, reactivity with this antibody was clearly decreased after murine MDM2 protein was phosphorylated by cyclin A-CDK2 ( Figure 5A). To eliminate the possibility that the effect was caused by either ATP binding or the presence of the kinase complex, we substituted ATP with a non-hydrolysable ATP analog, AMP-PNP, which was inactive in the kinase reaction. Mab SMP14 reactivity in this case was the same as with the unphosphorylated protein, indicating that the loss of recognition by this antibody was caused by cyclin A-CDK2 phosphorylation of MDM2 protein. In a separate ELISA experiment, we also discovered that Mab SMP14 could not recognize the T216A mutant protein ( Figure 5B). Thus, the SMP14 monoclonal antibody can differentiate between an un- and Mab 3G5 reactivity when cyclin B-CDK1 was used to phosphorylate GST-MDM2 ( Figure   5E).

MDM2 is phosphorylated at T216 in vivo at the G1/S transition
The fact that Mab SMP14 reactivity is sensitive to phosphorylation of T216 could The p19 ARF interacting region on MDM2 within its acidic domain between residues 210 and 304 is potentially close to the cyclin A-CDK2 phosphorylation site (27,58). We therefore also tested whether CDK phosphorylation may also modulate the interaction of MDM2 with p19 ARF . Here we used the MDM2-interacting N-37 fragment of p19 ARF protein, which was purified from bacteria. When tested by ELISA, contrary to its effect on p53-MDM2 interaction, cyclin A-CDK2 phosphorylation reproducibly produced a modest stimulation of the interaction between MDM2 and p19 ARF -N-37 at lower concentrations of MDM2 ( Figure 7C-I). In the control ELISA experiment using mutant MDM2 T216A protein, we observed no significant difference ( Figure 7C-II). Taken together our data suggest the possibility that phosphorylation at T216 serves to regulate the interaction of MDM2 with other proteins. The observation that MDM2 is selectively phosphorylated by cyclin A containing complexes also led to the identification of a cyclin recognition motif on MDM2. The MDM2 CRM resembles that of pRb in that a stable interaction between the MDM2 and cyclin A could not be detected (67). We were able to demonstrate its importance in the functional kinase assay, however, because a peptide containing the MDM2 CRM sequence was able to inhibit cyclin A-CDK2 phosphorylation of MDM2 at low concentrations. Since this peptide can also inhibit pRb phosphorylation by cyclin A-CDK2, it will be of interest to explore its effect on the phosphorylation of other CRM-containing proteins and whether it has any physiological effect in vivo, as has been shown for the CRM of p21 protein (75).
It was unexpected that phosphorylation of MDM2 at T216 by cyclin A-CDK2 causes loss of reactivity with the MDM2-specific monoclonal antibody Mab SMP14. Mab SMP14 was generated by using a peptide spanning amino acids 154-167 on human HDM2 protein (7,76).
Our experiments support the likelihood that this region on HDM2 is the recognition site for Mab SMP14 since an N-terminally deleted HDM2 lacking residues 1-166 does not react with this antibody (data not shown). The human Mab SMP14 epitope is not completely conserved on murine MDM2, however, and in fact we also observed that murine MDM2 is much more weakly recognized by Mab SMP14 than its human counterpart (data not shown). When a 12-residue peptide spanning the CDK site (sequence shown in Figure 4) was tested for its ability to block SMP14 recognition of MDM2 by ELISA assay, no significant inhibition was observed (data not shown). It is possible that this peptide is not long enough or does not have the right conformation to be recognized by SMP14. Alternately the SMP14 epitope may be located elsewhere in MDM2, and its recognition by the monoclonal antibody is negatively affected by either phosphorylation or substitution of T216. This is not the first example of a monoclonal antibody that fails to recognize phosphorylated MDM2: ATM phosphorylated HDM2 is not recognized by the monoclonal antibody Mab 2A10 (52).  (77). We are currently examining whether cyclin G-bound PP2A plays a role in the dephosphorylation of MDM2 protein in the cell. Alternately, CDK inhibitors such as p21, another p53 transcription target, may be involved. P21 has been reported to bind to cyclin A-CDK2 when the cell passes S-phase (78)(79)(80). Therefore, despite the increasing cyclin A expression in the cell cycle, the cyclin A-CDK2 complex may no longer be active to phosphorylate the newly synthesized MDM2.
It would be interesting to investigate whether cyclin A, whose promoter is transcriptionally regulated by MDM2, is differentially regulated by phosphorylated MDM2 protein. TBP interacts with MDM2 at a region adjacent to the acidic domain. This interaction is implicated in the cyclin A promoter activation (10), as well as transcriptional repression of p53 responsive elements (11). Deletion analysis indicated that the TBP-MDM2 interaction domain is close to T216 (10,11). When we examined whether cyclin A-CDK2 phosphorylation of MDM2 affects its interaction with TBP, however, the small negative effect we observed was not statistically significant (data not shown). It is therefore difficult to predict what the effect of cyclin A-CDK2 phosphorylation would be at the transcriptional level. Further in vivo experiments will be employed to address this question.
Our results support the likelihood that phosphorylation of MDM2 at T216 affects its interactions with proteins. We found that the p53-MDM2 interaction is reduced after phosphorylation by cyclin A-CDK2. However, the reduction is not as dramatic as reported with DNA-PK phosphorylated N-terminal fragment of MDM2 (53). This is most likely because the DNA-PK phosphorylation site lies within the N-terminal domain of MDM2 that interacts directly with p53, while T216 is outside of the MDM2-p53 interaction domain. Phosphorylation at T216 must therefore have an auxiliary or indirect effect on p53-MDM2 interaction, perhaps through altering the conformation of the MDM2 N-terminus. On the other hand, the interaction between CDKs, it has also been shown to be able, at higher levels, to inhibit cyclin A and cyclin B containing complexes as well (78,85). Thus we propose the existence of a feed-back loop that controls the activity of both cyclin A-CDK2 and p53 when the cells go into S-phase.
Most of our research was done using murine MDM2 and mouse cells. T216 is well conserved in human and mouse MDM2 but neither the CDK consensus sequence nor the CRM is present in human HDM2. Human HDM2 was a much worse substrate than murine MDM2 for cyclin A-CDK2 in our assay (data not shown). We can speculate that the T218 in human HDM2 corresponding to the T216 in murine MDM2 might be recognized by other proline-directed kinases in human cells. There is a cluster of conserved serines or threonines in the vicinity of murine T216 (SESTETPS) and some of these residues may be sites for other serine/threonine  Table 1.
A. Western blots of serum released cell extracts with antibodies as indicated.
B. Ratio of SMP14-reactive MDM2 to 3G5-reactive MDM2 protein at different time points after serum release. The readings for the reactivities were obtained from densitometry analysis on the Western blots shown in part A and blots from a duplicate experiment.