Suppression of Calpain-dependent Cleavage of the CDK5 Activator p35 to p25 by Site-specific Phosphorylation*

Cdk5 is a proline-directed Ser/Thr protein kinase predominantly expressed in postmitotic neurons together with its activator, p35. N-terminal truncation of p35 to p25 by calpain results in deregulation of Cdk5 and contributes to neuronal cell death associated with several neurodegenerative diseases. Previously we reported that p35 occurred as a phosphoprotein, phospho-p35 levels changed with neuronal maturation, and that phosphorylation of p35 affected its vulnerability to calpain cleavage. Here, we identify the p35 residues Ser8 and Thr138 as the major sites of phosphorylation by Cdk5. Mutagenesis of these sites to unphosphorylatable Ala increased susceptibility to calpain in cultured cells and neurons while changing them to phosphomimetic glutamate-attenuated cleavage. Furthermore, phosphorylation state-specific antibodies to these sites revealed that Thr138 was dephosphorylated in adult rat, although both Ser8 and Thr138 were phosphorylated in prenatal brains. In cultured neurons, inhibition of protein phosphatases converted phosho-Ser8 p35 to dual phospho-Ser8/Thr138 p35 and conferred resistance to calpain cleavage. These results suggest phosphorylation of Thr138 predominantly defines the susceptibility of p35 to calpain-dependent cleavage and that dephosphorylation of this site is a critical determinant of Cdk5-p25-induced cell death associated with neurodegeneration.

Cyclin-dependent kinase 5 (Cdk5) 2 is a unique member of the Cdk family. Its activity in postmitotic neurons is completely dependent upon association with one of two neuronal specific activators, p35 or p39. Cdk5/p35 is involved in a panoply of processes critical to central nervous system function both during development and throughout maturity including neuronal migration during corticogenesis, neurite outgrowth, regulation of the synaptic vesicle cycle, neurotransmitter release, and postsynaptic neurotransmitter receptor regulation and signaling (1)(2)(3). The mechanisms by which Cdk5 activity is normally regulated remains to be fully delineated. Furthermore, because aberrant Cdk5 activity has been implicated in the etiology of neurodegenerative diseases (4,5), identifying the biochemical mechanisms contributing to deregulation of Cdk5 is of substantial biomedical relevance.
Deregulation of Cdk5 results from removal of the first 98 amino acids of p35 by the Ca 2ϩ -dependent cysteine protease, calpain, leaving Cdk5 associated with the N-terminal truncated form p25. Cleavage of p35 to p25 changes the subcellular distribution of active Cdk5 from membranes to the cytosolic fraction (6,7), thereby altering substrate specificity. p25 accumulates in neurons undergoing various types of cell death (6 -9). Expression of Cdk5/p25 in cultured cells results in increased phospho-Tau levels in comparison to cells expressing Cdk5/ p35 (6). Furthermore, exogenous overexpression of p25 in transgenic mice results in a neurodegenerative phenotype including the formation of paired helical filaments, Tau aggregation, and neuronal loss similar to that observed in Alzheimer disease (10,11).
Cdk5/p25 has also been implicated in ischemia-induced neuronal loss in the hippocampus via increased phosphorylation of the NR2A subunit of the N-methyl-D-aspartic acid receptor (12). In addition, several recent reports indicate that Cdk5-p25 mediates cell death via translocation to the nucleus (13)(14)(15). p25 generation increases nuclear Cdk5 activity in cultured neurons, facilitating phosphorylation and inhibition of the pro-survival transcription factor MEF2 (13,15,16). Aberrant Cdk5 activity may also contribute to neuronal cell death via phosphorylation of other survival factors such as the tumor suppressor protein p53 (17) and retinoblastoma protein (14). Ca 2 -dependent activation of the cytoplasmic protease calpain is involved in apoptotic and necrotic cell death (18). Calpain generally recognizes motifs between conformational domains and cleaves substrate proteins in a limited manner, although the physiological function of calpain activity remains unclear. In some cases calpain cleavage is suspected to be a signaling process. Calpain-mediated cleavage of many proteins including neurofilament proteins (19), ␣II-spectrin (20), NR2 subunits of N-methyl-D-aspartic acid receptors (21), and ezrin (22) is suppressed through phosphorylation. However, how such signaling works and the mechanisms of phosphorylationdependent inhibition are unknown.
Previously we demonstrated that Cdk5 phosphorylated p35, that p35 occurred as a phosphoprotein in neurons, and that the phosphorylation state of p35 affected its susceptibility to calpain cleavage (23). Phospho-p35 predominates in fetal rat brain and is resistant to the cleavage by calpain, whereas unphosphorylated p35 present during adulthood is more vulnerable to calpain-dependent cleavage. Here we report that Ser 8 and Thr 138 of p35 serve as the sites of Cdk5-dependent phosphorylation. Furthermore, phosphorylation at these sites reduces the susceptibility of p35 to calpain cleavage. Moreover, specific dephosphorylation of Thr 138 increases the susceptibility of p35 to the cleavage by calpain in adult rat brains, suggesting that the phosphorylation of this site is a particularly critical determinant of Cdk5-dependent neuronal cell death in neurodegenerative diseases.
Expression and Metabolic Phosphorylation of p35 in HEK293 Cells-HEK293 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 100 units/ml of penicillin, and 0.1 mg/ml streptomycin. Transfection of Cdk5 and p35 was performed by the Polyfect transfection reagent (Qiagen, Hilden, Germany) according to the manufacturer's instructions. For metabolic phosphorylation of p35, transfected cells were cultured in the presence of 3.7 MBq of [ 32 P]orthophosphate in phosphate-free Dulbecco's modified Eagle's medium for 2 h. In some cases, roscovitine (50 M) was added to inhibit Cdk5 activity 2 h before metabolic labeling. After washing with phosphate-buffered saline, cells were suspended in RIPA buffer (20 mM Tris-HCl, pH 7.5, 1 mM EGTA, 1 mM EDTA, 0.15 M NaCl, 1% Nonidet P-40, 0.1% SDS, and 0.1% sodium deoxycholate) containing 10 mM ␤-glycerophosphate, 5 mM NaF, 0.2 mM Pefabloc SC (Merck, Darmstadt, Germany), 1 g/ml leupeptin, and 1 mM dithiothreitol, and lysed by freezing and thawing. The extract was obtained as a supernatant of centrifugation at 17,000 ϫ g for 15 min. p35 was isolated from the extract by immunoprecipitation with C19 anti-p35 antibody. 32 P incorporation into p35 was detected by a BAS 2000 image analyzer (Fuji Film, Tokyo, Japan) after SDS-PAGE.
Neuronal Culture, Metabolic Phosphorylation of p35, and Adenoviral Expression of p35-Cerebral cortical neurons were prepared from rat brains at embryonic day 17 as described previously (28). Neuronal cultures were used 7 days after plating. In some experiments, 0.1 M okadaic acid was added to the culture medium for the final 2 h of a 4-h labeling period. MG132 (50 M) was also added 1 h before the addition of okadaic acid to suppress proteosome-dependent degradation of p35. After washing with phosphate-buffered saline, neurons were lysed with RIPA buffer. The supernatant of centrifugation at 17,000 ϫ g was immunoprecipitated with C19 and phosphorylation of p35 was detected as described above. Adenoviral expression vectors encoding p35 or p35 4A were infected into rat primary cortical neurons cultured for 5 days in vitro. Two days after infection, neurons were treated with 5 M Ca 2ϩ ionophore A23187 for 1 h, and the cleavage of p35 was detected by immunoblotting with anti-p35 antibody (C19).
Expression of Cdk5-p35 in COS-7 Cells, Cleavage of p35 to p25 by Calpain, and Dephosphorylation of p35-COS-7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml of penicillin, and 0.1 mg/ml of streptomycin. Transfection of Cdk5, p35, -calpain, and the small subunit of calpain was performed by the Polyfect transfection reagent according to the manufactur-er's instruction. The cleavage of p35 to p25 by calpain in COS-7 cells was induced by treatment of COS-7 cells with 10 M ionomycin at 37°C for 30 min. To increase the phosphorylation status of p35, COS-7 cells were treated with 0.1 M okadaic acid for 3 h before treatment with ionomycin. For inhibiting the Cdk5 activity, cells were treated with 50 M roscovitine for 2 h before treatment with ionomycin. For inhibiting the calpain activity, cells were treated with 10 M calpain inhibitor IV for 2 h before treatment with ionomycin. To increase the calpain activity, -calpain and the small subunit of calpain were coexpressed in COS-7 cells with Cdk5 and p35. Dephosphorylation of p35 expressed in HEK293 cells was performed in the cell extract (50 mM Tris-HCl, pH 8.5, 0.15 M NaCl, and 1% Nonidet P-40) by incubation with bacterial alkaline phosphatase (20 unit/ml, Wako, Osaka, Japan) at 37°C for 1 h.
Immunoprecipitation, SDS-PAGE, and Immunoblotting-Preparation of the rat brain extracts and immunoprecipitation of Cdk5-p35 were done as described previously (23,28). SDS-PAGE was performed using a 12.5% polyacrylamide gel. Proteins were transferred to polyvinylidene difluoride membrane (Millipore, Bedgord, MA). The membranes were blocked with 5% skim milk in TBST (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.05% Tween 20). The membranes were incubated with primary antibodies followed by horseradish peroxidase-or alkaline phosphatase-conjugated secondary antibody (DAKO). The reactions were detected using an ECL system for horseradish peroxidase or the 5-bromo-4-chloro-3-indolyl phosphate/ nitro blue tetrazolium phosphatase substrate system for alkaline phosphatase and quantified by densitometry using NIH Image software. All experiments in this study were performed more than three times with similar results and representative results are shown in the figures.

RESULTS
p35 Is Phosphorylated at Ser 8 and Thr 138 by Cdk5-Substantial evidence has been presented to suggest that Cdk5 autophosphorylates its activator p35 in vitro and in cultured cells (24, 30 -32). To further investigate this autophosphorylation, we expressed both Cdk5 and p35 in HEK293 cells and cultured them with [ 32 P]orthophosphate for 2 h in the absence or presence of the Cdk5 inhibitor, roscovitine. 32 P incorporation was readily detected in p35 immunoprecipitated from the cell extracts by autoradiography (Fig. 1A). The highest levels of radiolabeled p35 were immunoprecipitated from cells cotransfected with Cdk5 (Fig. 1A, lane 4), but not with kinase-dead Cdk5 (Fig. 1A, lane 6). Furthermore, a large portion of the radioactivity associated with p35 could be attenuated by treating the cells with roscovitine (Fig. 1A, lane 5). Interestingly radiolabeled p35 was detected in cells transfected with p35 alone (Fig. 1A, lane 2). Moreover, treatment of cells transfected with p35 alone with roscovitine reduced radiolabeled p35 to almost undetectable levels (Fig. 1A, lane 3), suggesting that the endogenous Cdk5 was responsible for phosphorylation of p35 in the absence of exogenous Cdk5.
Cdk5 is a proline-directed protein kinase and exhibits the requirement of Pro following the substrate Ser/Thr sites, with a preference for the consensus motif Ser/Thr-Pro-X-His/Lys/ Arg. There are only four Ser/Thr-Pro sites, Ser 8 , Thr 138 , Ser 170 , and Thr 197 , within the primary sequence of p35 (Fig. 1B). Only Thr 138 is encompassed within a consensus motif sequence ( 138 TPKR), although Ser 8 is also followed by two positively charged amino acid residues ( 8 SPSYRK). Based on these observations, site-directed mutagenesis was used to replace Ser 8 or Thr 138 with Ala residues and the resulting plasmid was used together with one encoding Cdk5 to cotransfect HEK293 cell cultures. Mutation of Ser 8 or Thr 138 resulted in a large loss in radiolabel associated with immunoprecipitated p35 (data not shown).
To further characterize the phosphorylation of p35, radiolabeled protein immunoprecipitated from cotransfected cells was next analyzed by two-dimensional phosphopeptide mapping after trypsin digestion (Fig. 1C). Two radioactive spots representing two tryptic phosphopeptides were derived from wildtype p35. The S8A mutation resulted in disappearance of one of these phosphopeptides (spot 1), whereas the T138A mutation caused lose of the other phosphopeptide (spot 2). Taken together, these data indicate that p35 is phosphorylated by Cdk5 in transfected HEK293 cell cultures at two proline-directed sites, Ser 8 and Thr 138 .
Phosphorylation at Ser 8 or Thr 138 in p35 Suppresses the Calpain-dependent Cleavage of p35 to p25-Previously, we presented data suggesting that Cdk5-dependent autophosphorylation of p35 affected its susceptibility to cleavage by calpain (23). To further investigate the potential role of p35 phospho- rylation in controlling calpain cleavage, p35 and Cdk5 were next exogenously expressed in COS-7 cells. In preliminary analyses, cotransfected COS-7 cells were treated with various concentrations of the Ca 2ϩ ionophores, A23187 or ionomycin, or an inhibitor of intracellular Ca 2ϩ pumps, thapsigargin. For each of these treatments, a small but distinct amount of p25 was found to be produced by immunoblotting the cell lysates (data not shown). However, the largest amount of p35 cleavage was consistently derived using ionomycin. Therefore, 10 M ionomycin was selected as the method of choice for calpain activation in this paradigm. A time course for p25 generation in COS-7 cells treated with 10 M ionomycin is shown in Fig. 2A. After 30 min of ionomycin treatment, p25 can be clearly observed, with levels increasing thereafter until 24 h with mild attenuation of p25 levels at 3 h. Based on these results, the 30-min time point with 10 M ionomycin was chosen for all subsequent experiments as the minimal treatment necessary to induce the cleavage of p35 to p25.
To demonstrate that p25 production in response to ionomycin was indeed calpain-dependent, cells were either left untreated or treated with ionomycin in the absence or presence of 10 M calpain inhibitor IV (Fig. 2B, lanes 3 and 4). Treatment of cells with the calpain inhibitor resulted in a reduction in p25 levels both in the absence and presence of ionomycin. Further-more, p35 cleavage was increased substantially by coexpression of -calpain (Fig. 2B, lanes 5 and 6). This effect corresponded to increased active -calpain levels as assessed in immunoblots using an antibody to this cleaved form of the protease (27). Because the predominant form of endogenous calpain in COS-7 cells is m-calpain, active -calpain was detected only in cultures transfected with the corresponding plasmid encoding -calpain (Fig. 2B, second panel). Calpain activity was also assessed by immunoblot analysis of calpain-dependent cleavage of spectrin, a major component of the neuronal cytoskeleton (26). Spectrin cleavage closely mirrored p25 production in response to ionomycin and -calpain expression and was similarly blocked by calpain inhibition (Fig. 2B, third panel). These results demonstrate that p25 generation is tightly coupled to calpain activity.
Having established optimal conditions for p25 generation and demonstrated calpain dependence, the effect of phosphorylation on the cleavage of p35 was next assessed (Fig. 3A). Treatment of COS-7 cells expressing p35 and Cdk5 with ionomycin resulted in increased p25 levels, as observed earlier. Pretreatment with roscovitine resulted in increased basal levels of p25, as well as the levels achieved in response to ionomycin (Fig.  3A, compare lane 3 with lane 1 and lane 4 with lane 2). Furthermore, increased basal and ionomycin-induced levels were observed in cells transfected with p35 alone in comparison to those expressing both exogenous Cdk5 and p35. These results further confirm that calpain-dependent cleavage of p35 is inhibited via autophosphorylation by Cdk5.
To more specifically investigate the regulation of calpain cleavage of p35 by Cdk5-dependent phosphorylation of Ser 8 and Thr 138 , a site-directed mutagenesis approach was again  3 and 4), or under the coexpression of -calpain (lanes 5 and 6). The cleavage of p35 to p25 was examined by immunoblotting. The active -calpain expression and spectrin fragment generation was also examined by immunoblotting with antibodies specific for active (cleaved form) -calpain and ␣II-spectrin, respectively. An immunoblot of Cdk5 is shown in the lower panel. taken (Fig. 3B). The ability of ionomycin to induce p25 generation was substantially enhanced by the 2A (Ser 8 and Thr 138 to Ala) double mutation. On the other hand, the 2E (Ser 8 and Thr 138 to Glu) phosphomimetic double mutation substantially attenuated this effect. These observations indicate that the ability of Cdk5-dependent autophosphorylation to protect p35 from calpain cleavage occurs through phosphorylation of Ser 8 and/or Thr 138 .
Phosphorylation-dependent cleavage was also examined in neurons. p35 or p35 4A was expressed in primary cultured neurons and the cleavage was compared by immunoblotting with C19 after Ca 2ϩ ionophore A23187 treatment for 1 h. Upper bands of p35 and p25 exogenously expressed p35 species with FLAG tag at the C-terminal. p35 4A was cleaved more than the wild type of p35 (compare p25-F in lanes 4 and 6 of Fig. 3C). Exogenously expressed p35 appeared to be more resistant to calpain cleavage than endogenous p35 by unknown reasons.
The individual contribution of each of these two sites to this regulatory mechanism was further evaluated by site-directed mutagenesis (Fig. 3D). Interestingly, mutation of either Ser 8 or Thr 138 to Ala enhanced p25 generation in response to ionomycin treatment of COS7 cells cotransfected with Cdk5 and each of the respective forms of p35, suggesting phosphorylation of both of these sites contributes to protection of p35 from calpain cleavage.
Phosphorylation at Ser 8

and Thr 138 of p35 in Cultured Cortical Neurons and Rat
Brains-To further assess the physiological role of Cdk5-depedent phosphorylation of these two sites in regulating the cleavage of p35 to p25, the phosphorylation state of Ser 8 and Thr 138 was evaluated in cultured cortical neurons and intact brain tissue. Phosphorylation of p35 was first examined in cultured cortical neurons metabolically labeled with [ 32 P]orthophosphate by two-dimensional phosphopeptide map analyses (Fig. 4A). Interestingly, maps of radiolabeled p35 immunoprecipitated from cultured neurons produced a single major spot that corresponded to the phosphopeptide (spot 1 of Fig. 1C) associated with phospho-Ser 8 . Treatment of the neuronal cultures with the Ser/Thr protein phosphatase 1 and 2A inhibitor okadaic acid (0.1 M) together with the proteosome inhibitor MG132 (50 M, to prevent phospho-p35 from being degraded, Refs. 31 and 32) prior to immunoprecipitation of p35 resulted in the appearance of a second spot corresponding to the phosphopeptide (spot 2 of Fig. 1C) associated with phospho-Thr 138 . The identity of these two phosphopeptides were confirmed by an additional phosphopeptide map, where tryptic digests of p35 immunoprecipitated from cultured neurons treated with okadaic acid and HEK293 cells were mixed and the two spots comigrated to the exact same positions (Fig. 4A, right  panel). These results indicate that Ser 8 is phosphorylated in cultured cortical neurons and Thr 138 becomes phosphorylated in the presence of okadaic acid. These results suggest that whereas both Ser 8 and Thr 138 are phosphorylated in transfected HEK293 cells, only Ser 8 is phosphorylated in cultured cortical neurons under basal conditions.
To further characterize the phosphorylation of Ser 8 and Thr 138 in intact brain tissue, two rabbit polyclonal phosphorylation state-specific antibodies were generated for these sites. The specificities of these antibodies were defined by immuno-blot analyses (Fig. 4B). First lysates of HEK293 cells cotransfected with Cdk5 and p35 were prepared and either left untreated or treated with bacterial alkaline phosphatase prior to immunoblotting. Both phosho-Ser 8 and phospho-Thr 138 p35 were detected in untreated lysates and the signal for each of these sites was completely removed by phosphatase treatment (Fig. 4B, lane 2). In contrast, addition of the phosphatase had no effect on total p35 levels, although it did result in increased electrophoretic mobility (Fig. 4B, lanes 1 and 2 of the lowest  panel), consistent with loss of phosphate moieties and our earlier observations in cortical neurons (23). The specificity of the antibodies was further demonstrated by experiments where addition of the phospho-Ser 8 or phospho-Thr 138 peptides to the corresponding antibodies in blotting solution resulted in almost complete lose of signal due to preadsorption (Fig. 4B,  lanes 3 and 4). Finally, the phosphorylation site specificity was further confirmed by immunoblot analysis of lysates of HEK293 cells transfected with S8A or T138A mutant forms of  1 and 2 (left) lysates from HEK293 cells expressing p35 were either untreated or treated with bacterial alkaline phosphatase (BAP) prior to immunoblotting with phospho-Ser 8 (pS8), phospho-Thr 138 (pT138), or total p35 antibodies. Lanes 3 and 4 (middle) show the effect of preabsorption with the phosphopeptide used for immunization. Lanes 5-7 (right) show the effect of mutation of each site to Ala on specific detection by the corresponding antibody. Immunoblotting of control HEK293 cell extract without expressing p35 is shown in lane 8. C, p35 immunoprecipitated from fetal (E17) or adult brain (18 weeks) extracts immunoblotted phospho-Ser 8 , phospho-Thr 138 , and total p35. D, immunoblotting of p35 immunoprecipitated from rat brains of postnatal days 1, 5, 10, 15, and 20 with phospho-Ser 8 (pS8), phospho-Thr 138 (pT138), and total p35. p35. Mutation of each site to Ala resulted in complete loss in the ability of the corresponding antibody to detect p35 (Fig. 4B,  lanes 7 and 8). On the other hand, the sensitivity of the phospho-Ser 8 and phospho-Thr 138 antibodies were unaffected by Ala mutations at the non-targeted site, Thr 138 and Ser 8 , respectively. These results demonstrate that the phospho-Ser 8 and phospho-Thr 138 antibodies are able to selectively detect phopho-p35 and are quite specific for the phosphorylation sites on p35 to which they were raised.
Previous findings have suggested that the phosphorylation state of p35 changes during brain development (23). Therefore, the novel phosphorylation state-specific antibodies were next employed to examine the phosphorylation states of these two sites in fetal and adult brains (Fig. 4C). p35 immunoprecipitated from fetal or adult brain extracts was immunoblotted for phospho-Ser 8 , phospho-Thr 138 , and total p35 levels. Fetal p35 was phosphorylated at both Ser 8 and Thr 138 , whereas adult p35 was phosphorylated only at Ser 8 . To further characterize this observation, the dephosphorylation of Thr 138 was examined across postnatal development using p35 immunoprecipitates prepared from rat brains of postnatal days 1 to 20. Dephosphorylation began at postnatal days 5-10 and was almost completed by day 20 (Fig. 4D). These results indicate that p35 is phosphorylated at Ser 8 and Thr 138 in fetal brains as in HEK293 cells, and phosphorylation at Thr 138 decreases with aging.
Okadaic Acid-dependent Suppression of the Cleavage Was Due to the Phosphorylation at Thr 138 -The observation that inhibition of protein phosphatases resulted in increased phospho-Thr 138 levels, as assessed by phosphopeptide map analysis (Fig. 4A) and that phospho-Thr 138 levels were selectively reduced in adult tissue (Fig. 4C), together with our previous observations suggesting phosphorylation protects fetal p35 from calpain, prompted further studies of the specific contributions of each of these sites to calpain susceptibility. In these experiments COS-7 cultures cotransfected with Cdk5 and p35 were either left untreated or treated with okadaic acid. The levels of phospho-Ser 8 and phospho-Thr 138 were then compared by quantitative immunoblot analysis of lysates using the phosphorylation state-specific antibodies. As suspected, phospho-Thr 138 , but not phospho-Ser 8 , was increased significantly by treatment with okadaic acid (Fig. 5A). Next, these experimental conditions were employed to investigate the relative contribution of each of the two phosphorylation sites to the vulnerability of p35 to calpain cleavage (Fig. 5B). The effect of okadaic acid on p25 production in response to ionomycin was evaluated in lysates of COS-7 cells cotransfected with Cdk5 and wild-type, S8A, or T138A. Okadaic acid, which selectively increased phospho-Thr 138 levels, substantially decreased the ability of ionomycin to induce p25 production in cells transfected with either wild-type or S8A p35. In contrast, okadaic acid had no discernible effect on the ability of ionomycin to induce p35 production in cells expressing T138A p35. These observations strongly suggest that the critical determinant of the susceptibility of p35 to calpain cleavage is Cdk5-dependent phosphorylation of Thr 138 . These results are in agreement with observations that fetal p35, which is highly phosphorylated at Thr 138 is more resistant to the cleavage than adult p35 (23), which exhibits comparatively low phospho-Thr 138 levels. In other words, Thr 138 is the phosphorylation site most directly involved in the age-dependent cleavage of p35 to p25.

DISCUSSION
Calpain cleavage of p35 is thought to untether Cdk5 from the plasma membrane, resulting in aberrant activity that contributes to neuronal cell death and a variety of neurodegenerative disorders. Previously we have implicated reduction in autophosphorylation of p35 in age-dependent susceptibility to p25 generation (23). In the present study, we further characterized phosphorylation of p35 by Cdk5 and investigated the role of this important regulatory mechanism in preventing p35 cleavage by calpain. A combined site-directed mutagenesis/cell culture expression approach was used to identify the phosphorylation sites as Ser 8 and Thr 138 . Inhibition of Cdk5 or mutation of either of these sites to Ala enhanced p25 production in cultured non-neuronal cells and primary neurons in response to ionophore-induced increased intracellular Ca 2ϩ concentrations. Whereas p35 was phosphorylated at Ser 8 and Thr 138 in fetal brain tissue, only Ser 8 was phosphorylated in primary cultured cortical neurons and adult brain. Phosphorylation at Thr 138 was induced by the protein phosphatase inhibitor, okadaic acid, suggesting the potential of this site in particular to function as a regulatory signaling mechanism. Furthermore, loss of phosphorylation at Thr 138 , not Ser 8 , rendered p35 more susceptible to cleavage by calpain in the presence of okadaic acid. Taken together these observations suggest that phosphorylation of Thr 138 of p35 by Cdk5 is a critical regulatory mechanism that prevents p25 formation during brain development.
Whereas p25 generation has been most highly associated with neuronal cell loss, growing evidence suggests Cdk5 is involved in the mechanisms of synaptic plasticity underlying learning and memory (33)(34)(35)(36). Recently, transient overexpression of p25 has been suggested to contribute to enhanced synaptic plasticity and memory before resulting in deleterious effects (37). Given these recent observations, a physiological role for p25 cannot be completely ruled out. On the other hand, calpain has also been thought to be a signaling protease, which generates limited proteolysed fragments that possess altered properties in comparison to their parent molecules (18). Cdk5-p25 is different from Cdk5-p35 with regard to solubility, cellular localization, stability, and kinase activity (6 -9, 38). Thus, it is important to understand how proteolytic cleavage of p35 to p25 is regulated. Calpain has no distinct consensus cleavage sequences, and is thought to target linker regions between conformational domains within proteins (18,39). The cleavage site between Phe 98 and Ala 99 of p35 appears to correspond to the linker connecting the N-terminal membrane-binding region and C-terminal Cdk5 activation globular domain (40). In fact, the cleavage of p35 may be suppressed by deletion of about 20 amino acids including the cleavage site but not by mutation at several amino acids prior (i.e. N-terminal) to the cleavage site. 3 Calpain cleavage of a number of other proteins has been shown to be regulated by phosphorylation (19 -22). The cleavage of p35 was also suppressed by phosphorylation at Ser 8 or Thr 138 of p35. Considering that these two phosphorylation sites located on opposite sides of the cleavage site were differently phosphorylated, it is possible that each may suppress p35 cleavage using different or independent mechanisms.
Phosphorylation of Ser 8 may serve to repress basal calpain activity toward p35. Ser 8 was constitutively phosphorylated when p35 was expressed with Cdk5 in either cultured cells or brains, although the exact stoichiometry of phosphorylation has not been determined. p35 associates with membranes via N-terminal myristoylation (6). 3 Ser 8 is close to the Gly 2 myristoylation site. Suppression of the cleavage by Ser 8 phosphorylation may be related to the interaction of p35 with membranes. There are several examples of myristoylated proteins whose phosphorylation in the N-terminal region alters the association with membranes (41,42). Calpain is also translocated to membranes when activated and cleaves proteins on membranes (18), although the manner in which calpain interacts and associates with membranes is unknown. The catalytic site of calpain may be positioned at a particular distance from the surface of membranes. Conceivably, phosphorylation at Ser 8 changes the interaction of p35 with membranes so as to render the cleavage site unavailable to the catalytic site of calpain.
Phosphorylation at Thr 138 seemed to be the more critical determinant and appeared much more subject to regulation in cultured cells and intact brain tissue. Several possibilities may be considered regarding how p35 phosphorylation prevents calpain cleavage. (i) Phosphorylation at Thr 138 may reduce the affinity to calpain by changing the structural conformation around the cleavage site. (ii) The binding of other proteins to the phosphorylated form of Thr 138 may mask the region recognized by calpain. Several p35-binding proteins have been reported (43), and it is possible that some of them facilitate phosphorylation of Thr 138 during development but not in the adult Cdk5-p35 complex. (iii) It is also possible that phospho-Thr 138 simply causes p35 to be degraded so rapidly that this form of the protein is never exposed to active calpain for any length of time. Although this study has focused on the calpaindependent cleavage, the proteosomal degradation of p35 is also regulated by phosphorylation (31,32). It has previously been indicated that phosphorylation at Thr 138 is required for degradation of p35 (31). In any case, phosphorylation at Thr 138 plays an important role as the determinant for the proteolytic fate of p35 and serves as a key point in preventing neurodegeneration.
The question remains of how phosphorylation of Thr 138 is differentially regulated in fetal versus adult tissue despite autophosphorylation by Cdk5-p35. Protein phosphorylation is determined by the balance of enzymatic activities between protein kinases and protein phosphatases. Both the activity of Cdk5-p35 activity and the phosphatase(s) responsible for dephosphorylating p35 may be developmentally regulated. Cdk5-p35 activity peaks during early postnatal days of rat brain development (24,44), and Cdk5-p35 activity in the fetal brain is greater than that in the adult brain extract. 4 These developmental differences in Cdk5-p35 activity appear, at least in part, to depend on the interaction of the protein kinase complex with membranes (45). On the other hand, the increased phosphorylation of p35 observed in cultured neurons and COS-7 cells in the presence of a phosphatase inhibitor suggests that Thr 138 is actively dephosphorylated by PP1 or PP2A under basal conditions, raising the modulation of these enzymes as possible mechanisms for regulating the susceptibility of p35 to calpain cleavage. Such putative okadaic acid-sensitive protein phosphatase regulatory mechanisms were also suggested by our previous observations (23). How phosphorylation of p35 is regulated remains an excellent subject for future study.