Identification of functional domains in the neuronal Cdk5 activator protein.

Cyclin-dependent kinase 5 (Cdk5) is activated by the neuronal-specific activator protein, p35. In contrast to the activation of typical CDKs by cyclin subunits, p35·;Cdk5 was not further activated by the CDK-activating kinase (CAK) and was neither phosphorylated nor inhibited by the Tyr-15-specific Wee1 kinase. The previously identified proteolytic active fragment of p35, p25 (residues 91-307) as well as the slightly smaller fragment containing residues 109-291, was found to be sufficient to bind and activate Cdk5. Other CDKs, including Cdk2, associated weakly with p25. However, their kinase activity was only activated to the low level observed for cyclin A·;Cdk2 without Thr-160 phosphorylation, and phosphorylation of Thr-160 in Cdk2 did not activate the p25·;Cdk2 complex further. We have identified distinct regions in p35 required for binding to Cdk5 or activation of Cdk5. Residues ∼150-200 of p35 were sufficient for binding to Cdk5, but residues ∼279-291 were needed in addition for activation of Cdk5 in vitro.

Cyclin-dependent kinase 5 (Cdk5) is activated by the neuronal-specific activator protein, p35. In contrast to the activation of typical CDKs by cyclin subunits, p35⅐Cdk5 was not further activated by the CDK-activating kinase (CAK) and was neither phosphorylated nor inhibited by the Tyr-15-specific Wee1 kinase. The previously identified proteolytic active fragment of p35, p25 (residues 91-307) as well as the slightly smaller fragment containing residues 109 -291, was found to be sufficient to bind and activate Cdk5. Other CDKs, including Cdk2, associated weakly with p25. However, their kinase activity was only activated to the low level observed for cyclin A⅐Cdk2 without Thr-160 phosphorylation, and phosphorylation of Thr-160 in Cdk2 did not activate the p25⅐Cdk2 complex further. We have identified distinct regions in p35 required for binding to Cdk5 or activation of Cdk5. Residues ϳ150 -200 of p35 were sufficient for binding to Cdk5, but residues ϳ279 -291 were needed in addition for activation of Cdk5 in vitro.
Cyclins and cyclin-dependent kinases (CDKs) 1 are key regulators of the eukaryotic cell cycle (1). Cdc2 is associated with B-type cyclins and regulates M phase (2). Cdk2 is associated with A-and E-type cyclins, and the respective complexes are believed to control the S phase and G 1 -S transition, respectively (3,4). Cdk4 and Cdk6 are associated with the D-type cyclins and are important for G 1 progression (3).
The activity of CDKs is tightly regulated by an intricate system of protein-protein interaction and phosphorylation (5). The activation of CDKs, by definition, requires the association with cyclin partners. Full activation of CDKs requires in addition the phosphorylation of Thr-161/Thr-160, which lies in the activating T-loop in the crystal structure of Cdk2 (6,7). The Thr-161/Thr-160 residue is phosphorylated by the CDK-activating kinase (CAK), which is composed of a cyclin H-Cdk7 complex and a RING finger protein subunit MAT1 (8). The activity of CDKs can be inhibited by phosphorylation of Thr-14 and Tyr-15 by the Wee1 and Myt1 protein kinases (9). Further-more, CDKs can be inactivated by binding to CDK inhibitors like those from the p21 cip1/WAF1 family (p21 cip1/WAF1 , p27 kip1 , and p57 kip2 ) and the p16 INK4A family (p16 INK4A , p15 INK4B , p18 INK4C , p19 INK4D ) (10).
It has become clear that not all cyclins and CDKs function in cell cycle control. Examples of cyclins and CDKs that function in non-cell cycle regulating events are mounting; for instance, cyclin H⅐Cdk7 and cyclin C⅐Cdk8 in the TFIIH subunit of the RNA polymerase II holoenzyme have possible roles in phosphorylating the C-terminal domain of RNA polymerase II (8), and p35⅐Cdk5 has a role in neurite outgrowth in postmitotic neurons (11).
Cdk5 was identified as a CDK-related protein PSSALRE (12) and one of the CDK partners of cyclin D in human normal diploid fibroblasts (13). There is no other indication, however, that Cdk5 is active or functions in the normal cell cycle control (14). Active Cdk5 was first purified from brain extracts as a proline-directed protein kinase (15). The purified protein kinase contains 33-and 25-kDa subunits; the 33-kDa subunit was later identified as the Cdc2-related kinase Cdk5 (16). Cdk5 was similarly identified as a neuronal protein kinase capable of phosphorylating the KSPXK sequence motif in neurofilament proteins NF-H, NF-M (17)(18)(19), and the tau protein (20). Phosphorylation of tau by Cdk5 is particularly interesting because abnormally phosphorylated tau is the major component of the paired helical filaments, which accumulate in the brains of Alzheimer patients (21). Cdk5 is able to phosphorylate tau on sites that are abnormally phosphorylated in Alzheimer's paired helical filaments (22,23). The 25-kDa subunit of the Cdk5 kinase (15,24) was later found to be a proteolytic fragment of a larger 35-kDa protein (p35) (14,25,26). An isoform of p35 that shares 57% amino acid identity to p35 has been identified, and its mRNA is predominantly expressed in the hippocampus (27).
The complex p35⅐Cdk5 is only active as a histone H1 kinase in postmitotic neurons (14,28). Moreover, p35⅐Cdk5 appears to be important for normal neuronal function; its kinase activity increases during neurogenesis, and it is essential for neurite outgrowth during neuronal differentiation (11). A fundamental role for p35⅐Cdk5 is supported by the findings that targeted disruption of the p35 gene in the mouse leads to severe defects in laminar organization of neurons in the neocortex and cerebellum, 2 and targeted disruption of Cdk5 gene leads to abnormal corticogenesis, lesions in the central nervous system, and perinatal death (29). Although p35 shares little sequence similarity to cyclin, computer modeling predicts that p35 may fold into cyclin-like tertiary structure and activate Cdk5 in a manner similar to a cyclin (30). This is reminiscent of the two repeating cyclin folds found in cyclin A, which lack sequence homology but nevertheless have an identical structure (31).
Even though p35 may adopt a cyclin-like structure and can activate Cdk5, the regulation of p35⅐Cdk5 appears to be very much distinct from other cyclin⅐CDK complexes. Activation of typical cyclin⅐CDK complexes requires phosphorylation of Thr-161 or equivalent residues, but p35⅐Cdk5 is active as a histone H1 kinase in the absence of phosphorylation by other protein kinases (32), despite the fact that the surrounding sequences of Ser-159 in Cdk5 (the Thr-161-equivalent residue) are similar to other CDKs. Moreover, no autophosphorylation of Cdk5 was observed under those conditions (32). Unlike most other cyclin⅐CDK complexes, p35⅐Cdk5 is not inhibited by the p21 cip1/WAF1 (33) or p27 kip1 (34) CDK inhibitors. There is some evidence, however, that a population of p35⅐Cdk5 exists as an inactive form within a macromolecular structure, suggesting p35⅐Cdk5 may bind to inhibitors in the brain (35).
Here we investigated the mechanism of activation of Cdk5 in vitro. We have defined distinct regions in p35 required either for binding to Cdk5 or activation of Cdk5. Residues ϳ150 -200 of p35 were sufficient for binding to Cdk5 in vitro, but residues ϳ279 -291 were in addition critical for activation of Cdk5. The regulation of p35⅐Cdk5 is likely to be very different from that of other cyclin⅐CDK pairs, because unlike cyclin A⅐Cdk2, p35⅐Cdk5 was not activated by CAK or inhibited by Wee1 under the same conditions.
Expression and Purification of Proteins-Expression of GST-and histidine-tagged proteins in bacteria and purification with GSH-agarose and Ni 2ϩ -nitrilotriacetic acid-agarose chromatography respectively were as described (41). Transcription of mRNA in vitro (42) and translation in reticulocyte lysate in the presence of [ 35 S]methionine (43) was as described. In some experiments, reticulocyte lysate was supplemented with 1 ⁄10 volume of Xenopus egg extract as described (38).
Binding Assays-GST-p25 or mutants (1 g) were incubated with reticulocyte lysate programmed with CDK mRNAs (5 l) at 23°C for 30 min. GST-p25 was then recovered with 15 l of GSH-agarose in 250 l of bead buffer (50 mM Tris-HCl, pH 7.4, 5 mM NaF, 250 mM NaCl, 5 mM EDTA, 5 mM EGTA, 0.1% Nonidet P-40, 2 g/ml aprotinin, 15 g/ml benzamidine, 1 g/ml leupeptin, 10 g/ml soybean trypsin inhibitor). After incubation at 4°C with end-to-end rotation for 45 min, the beads were washed three times with 250 l of bead buffer, transferred to new tubes, and washed twice more with bead buffer. The samples were then dissolved in 30 l of SDS-sample buffer, and the bound CDKs were detected by SDS-PAGE followed by PhosphorImager analysis (Molecular Dynamics).
Kinase Assays-Purified proteins (1-3 l) or immunoprecipitates equilibrated with kinase buffer (80 mM Na-␤-glycerophosphate, pH 7.4, 20 mM EGTA, 15 mM Mg(OAc) 2 , and 1 mM dithiothreitol) were mixed with 10 l of kinase buffer containing 50 M ATP, 1.25 Ci of [␥-32 P]ATP, and 1 g of histone H1. The samples were incubated at 23°C for 30 min, and the reactions were terminated by addition of 30 l of SDS-sample buffer. The samples were subjected to SDS-PAGE, and phosphorylation was detected with a PhosphorImager.
Immunological Methods-Anti-mouse Cdk7 (CAK) antibody and immunoprecipitation of CAK from Swiss 3T3 cells were as described (44). Immunoprecipitation and immunoblotting were performed as described (45).

Identification of Binding and Activating Domains of the
Cdk5 Activator Protein p35-To identify the regions in p35 that are important for the activation of Cdk5, we constructed truncation mutants of p35 and tested their ability to bind and activate Cdk5 in vitro. Full-length p35 contains 307 amino acids and the originally identified 25-kDa proteolytic fragment of p35 contains residues 99 -307. The 25-kDa proteolytic fragment is active in vivo (15,24) and in vitro when expressed in bacteria as a GST-fusion protein (26,32). A GST-fusion protein containing residues 109 -291 (designated as GST-p25 here) is also active as a Cdk5 activator (32). We sought to define the minimum region of GST-p25 that is required for Cdk5 binding and activation. GST-p25 and deletion mutants derived from it were expressed in bacteria and purified by GSH-agarose chromatography. We then tested their binding to human Cdk5 and to other CDKs and CDK-related proteins (Cdc2, Cdk2, Cdk3, Cdk4, Cdk5, Cdk6, Cdk7, and PCTAIRE1, a CDK-related protein) synthesized in a mRNA-dependent rabbit reticulocyte lysate in the presence of [ 35 S]methionine. Only Cdk5 bound strongly to GST-p25 under our assay conditions (Fig. 1). Cdc2, Cdk2, Cdk3, Cdk7, and PCTAIRE all bound weakly to GST-p25; interaction between GST-p25 and endogenous Cdk2 was also detected when GST-p25 was added to a cell extract (data not shown). The significance of these interactions will be explored below.
Using the above binding assay, we compared the affinity of Cdk5 for GST-p25 to various GST-fusion proteins containing C-terminal and N-terminal deletions of p35. The p35 deletion mutants and their relationship to p35/p25 are summarized schematically in Fig. 2. Deletion of p25 from the N terminus up to residue 122 (N⌬122) had little effect on its binding to Cdk5, and deletion up to residue 147 (N⌬147) only slightly impaired binding to Cdk5 (Fig. 3A). In contrast, deletion up to residue 200 (N⌬200) abolished binding to Cdk5. Similarly, deletions from the C terminus of p25 had little effect on the binding to Cdk5 up to residue 198 (C⌬198). There was a slight reduction in the affinity of C⌬179 for Cdk5. Further deletions from the C terminus resulted in greater reduction of Cdk5 binding (C⌬159) and finally a complete loss of Cdk5 binding (C⌬139).
In addition to assaying binding to in vitro translated Cdk5, we also assayed whether GST-p25 and its mutants could stimulate the kinase activity of bacterially expressed GST-Cdk5 (Fig. 3B). Deletion from the N terminus of p25 resulted in a progressive reduction of histone H1 kinase activity associated with Cdk5, resembling the decrease of Cdk5 binding activity shown above. In contrast, even the smallest deletion from the C terminus we tested (C⌬279) resulted in the complete loss of histone H1 kinase activity associated with Cdk5. These results were surprising because the C-terminal deletions of p25 up to residue 159 could still bind well to Cdk5 (Fig. 3A). Similar results were obtained by using bacterially expressed p25 mutants and in vitro translated Cdk5. 4 The Cdk5-binding and Cdk5-activating activities of p35/p25 and deletion mutants are summarized in Fig. 2. The indicated ability of full-length p35 to bind and activate Cdk5 has not been confirmed experimentally, because GST-p35 could not be produced in soluble form under the conditions used successfully for the expression of GST-p25 or other deletion mutants. In fact, it is unclear whether p35⅐Cdk5 complex is active as a kinase; there are indications that p25⅐Cdk5 is active but p35⅐Cdk5 is present in an inactive complex in the brain (35). Taken together, these results indi-cate that the region of p35 that interacts with Cdk5 is residues ϳ150 -200 and that the C-terminal region ϳ279 -291 is essential for the activation of Cdk5. Deletion of either the binding domain or the activating domain of p35 was sufficient to prevent the activation of Cdk5.

FIG. 2. Summary of GST-p35 deletion mutants.
The p25 was deleted from the N terminus or the C terminus as described under "Experimental Procedures." N⌬ or C⌬ denoted deletions from the N and C terminus, respectively, and the numbers indicate the amino acid residues in p35 where the mutants were deleted. Full-length p35 contains 307 amino acid residues, and p25 contains residues 109 -291. The relative Cdk5 binding affinity of p25 and its mutants and the ability to activate Cdk5 are indicated. The activity of p35 is only conceptual, because soluble GST-p35 could not be produced under the same conditions as GST-p25 or other deletion mutants. presence of increasing amounts of C⌬251, indicating that C⌬251 could compete with GST-p25 for binding to Cdk5. Nearly complete inhibition of Cdk5 activation was observed with a 10-fold excess of C⌬251 over p25.
Activation of Cdk5 by p35 Is Not Regulated by CAK-To investigate the activation of Cdk5 by p35 further, we explored whether the weak interactions between GST-p25 and other CDKs, like Cdk2, observed in Fig. 1 were of any significance relative to that between GST-p25 and Cdk5. As previously reported, the histone H1 kinase activity of bacterially expressed GST-Cdk2 was activated weakly by just mixing with bacterially expressed cyclin A (protein A (PA)-cyclin A fusion protein); the PA-cyclin A⅐GST-Cdk2 complex was further activated by phosphorylation of Thr-160 by CAK (36,46) (Fig. 5,  lanes 4 and 7). As expected, both the K33R (changing the lysine in the ATP-binding site of Cdk2 to arginine) and T160A (changing the activating Thr-160 to alanine) mutants of GST-Cdk2 were not activated by cyclin A in the presence of CAK (Fig. 5,  lanes 5, 6, 8, and 9). GST-p25 was also able to activate the kinase activity of GST-Cdk2 slightly, roughly to the level of activity of cyclin A⅐Cdk2 in the absence of Thr-160 phosphorylation (Fig. 5, lane 10). The T160A mutant but not the K33R mutant of Cdk2 was likewise weakly activated by GST-p25 (Fig. 5, lanes 11 and 12). In contrast to the activation with cyclin A, however, no further activation of GST-p25⅐Cdk2 was seen in the presence of CAK (Fig. 5, lanes 13-15). GST-Cdk2 can be phosphorylated by CAK even in the absence of cyclin subunit (36), and phosphorylation of GST-Cdk2 by CAK was also observed in the presence of GST-p25 (data not shown); hence, the lack of further activation of GST-p25⅐Cdk2 complex was not due to a block of CAK phosphorylation. Given the fact that the weak kinase activity of cyclin A⅐Cdk2 in the absence of Thr-160 phosphorylation may not have a physiological role, it seems unlikely that the weak activation of Cdk2 by p25 will be of physiological significance.
Given that the activity of p25⅐Cdk2 was not affected by the presence or absence of CAK, we next investigated whether the kinase activity of p25⅐Cdk5 complex could be regulated by CAK. As shown before, GST-Cdk2 is activated when bound to a bacterially expressed histidine-tagged cyclin A (cyclin A-H6) and is phosphorylated on Thr-160 by CAK (Fig. 6A, lane 6); as shown above Cdk2 was not activated by GST-p25 (lane 7). In contrast, GST-Cdk5 was not activated by cyclin A but was activated by p25 irrespective whether CAK was present or not. We also constructed a GST-cyclin A-p35 fusion protein containing the first cyclin-fold of cyclin A (residues 171-303) that interacts with Cdk2 (31) and the activating domain of p35 (residues 228 -291), in the hope that the fusion protein might bind to Cdk2 and activate its kinase activity in the absence of Thr-160 phosphorylation. But not too surprisingly, this cyclin A-p35 fusion construct turned out to be unable to activate either Cdk2 or Cdk5 (lanes 8 and 16). These data showed that p25⅐Cdk5 could not be further activated by CAK under the similar conditions when cyclin A⅐Cdk2 was activated. More-over, we did not observe any direct phosphorylation of the Thr-160 equivalent residue in Cdk5, Ser-159, by CAK in the presence or absence of p25 under conditions where GST-Cdk2 Thr-160 was phosphorylated (data not shown). Mutation of Ser-159 to an alanine (S159A) also did not affect the activity of p25⅐Cdk5, whereas mutation of the Lys-33 residue in the ATP-  -Cdk2 (lanes 1, 4, 7, 10, and  13), GST- Cdk2 K33R (lanes 2, 5, 8, 11, and 14), or GST-Cdk2 T160A (lanes 3, 6, 9, 12, and 15) (100 ng) were mixed with buffer (lanes 1-3), 100 ng of PA-cyclin A (lanes 4 -9), or GST-p25 (lanes 10 -15) 1-8) or GST-Cdk5 (100 ng) were mixed with buffer (lanes 1, 5, 9, and 13) or 100 ng of the following purified protein: cyclin A -H6 (lanes 2, 6, 10, and 14), GST-p25 (lanes 3, 7, 11, and 15), and GST-cyclin A-p35 (lanes 4, 8, 12, and 16). The samples were incubated with immunoprecipitates obtained using either normal rabbit serum (lanes 1-4 and 9 -12) or anti-Cdk7 (CAK) antibody (lanes 5-8 and 13-16) in the presence of 15 mM Mg 2ϩ and 50 M ATP at 23°C for 30 min. The histone H1 kinase activity was then assayed. Histone H1 phosphorylation was detected with SDS-PAGE and quantified by Phos-phorImager analysis. B, GST-p25 (100 ng) (lanes 1-3) or GST alone (100 ng) (lanes 4 -6) was incubated with 100 ng of GST-Cdk5 S159A (lanes 1  and 4), GST-Cdk5 K33R (lanes 2 and 5), or GST-Cdk5 (lanes 3 and 6) in the presence of 15 mM Mg 2ϩ and 50 M ATP at 23°C for 30 min. The histone H1 kinase activity was then assayed as in A.
Activation of Cdk5 By p35 Is Not Regulated by Wee1-CDKs like Cdc2 (5), Cdk2 (47,48), and Cdk4 (49) are negatively regulated by phosphorylation of Thr-14/Tyr-15 by Wee1 or related kinases. The negative regulation of Thr-14/Tyr-15 phosphorylation is in general dominant over the positive regulators of CDKs like cyclins and Thr-160 phosphorylation. The kinase activity associated with Thr-160-phosphorylated cyclin A⅐Cdk2 complex was inhibited by Tyr-15 phosphorylation with recombinant Wee1 (Fig. 7B). In contrast, p25⅐Cdk5 was neither significantly phosphorylated nor inactivated by recombinant Wee1 under similar conditions (Fig. 7A), although both Thr-14 and Tyr-15 residues are present in Cdk5. Fig. 7A also shows that GST-Cdk5 had autophosphorylating activity and that GST-p25 was a substrate for GST-Cdk5, which is in contrast to what was observed by Qi et al. (32). The observed phosphorylation of GST-Wee1 was mainly due to autophosphorylation on tyrosine residues (data not shown).

DISCUSSION
Judging by sequence homology alone, Cdk5 is a typical CDKlike protein kinase with nothing in particular to indicate a major difference from the rest of the CDK family (12). Indeed, as with Cdk4 and Cdk6, Cdk5 has been found to associate with D-type cyclins in normal diploid fibroblasts (13). But as yet, there is no indication that cyclin D⅐Cdk5 functions in the normal cell cycle. Unlike dominant-negative mutants of Cdk2 and Cdc2, overexpression of a dominant-negative Cdk5 mutant does not cause cell cycle arrest (50), and no histone H1 kinase activity associated with Cdk5 can be detected except in postmitotic neuronal cells.
What then is the molecular basis of the uniqueness of Cdk5 compared with other CDKs? If Cdk5 function is important for neuronal differentiation, as indicated by recent evidence (11,29), then Cdk5 must be regulated by very different mechanisms compared with other cell cycle-regulating cyclin⅐CDK complexes. Cell cycle-regulating cyclins and CDKs are generally turned off during neuronal differentiation (51,52), since cell cycle arrest appears to be a prerequisite for differentiation. The activity of cell cycle regulatory cyclin⅐CDKs can be down-regulated by several mechanisms, including Thr-14/Tyr-15 phosphorylation by protein kinases such as Wee1, decreased Thr-160 phosphorylation, and binding of CDK inhibitors like p21 cip1/WAF1 and p27 kip1 . Cdk5 is immune to these inhibitory mechanisms because it is bound to an atypical partner p35/p25. The levels of p21 cip1/WAF1 and p27 kip1 are elevated during differentiation in tissues such as neurons and muscles and may play a role in turning off cyclin⅐CDK activities. (53,54). Consistent with a requirement for down-regulation of CDK activity in neuronal cell differentiation, expression of Cdk2 blocks nerve growth factor-induced differentiation of PC12 cells (55). During nerve growth factor-induced differentiation of PC12 cells, cyclin D1 expression is increased but cyclin D1-associated kinase activity is not induced, possibly due to inhibition by p21 cip1/WAF1 or related proteins (55)(56)(57). Moreover, overexpression of p27 kip1 alone is sufficient to induce neuronal differentiation in a mouse neuroblastoma cell line (52). The kinase activity of p35⅐Cdk5, however, is not inhibited by p21 cip1/WAF1 or p27 kip1 (33,34). We also found that the activity of bacterially expressed p25⅐Cdk5 is not inhibited by recombinant p21 cip1/WAF1 or p27 kip1 . 5 Not only is p25⅐Cdk5 not regulated by p21 cip1/WAF1 /p27 kip1 family, but as we show here p25⅐Cdk5 kinase activity is not dependent on CAK nor is it negatively regulated by Wee1. It is of course possible that neurons contain specific versions of CAK, Wee1, or inhibitors that are capable of regulating p25⅐Cdk5. This is particularly relevant because it has been observed that a population of p35⅐Cdk5 is not active in brain extracts (35). Moreover, an activity capable of phosphorylating Thr-14 in Cdc2 has been purified from thymus that can also phosphorylate and inhibit p25⅐Cdk5 (58).
Here we show that a deletion of the C terminus of p25 (beyond residue 291) abolished its ability to activate Cdk5, despite the fact that it could still associate with Cdk5. The C-terminal deletion may either disrupt the conformation of the whole activating domain without affecting the binding domain, or the residues at the C terminus are directly required for Cdk5 activation. However, residues 279 -291 in p35 do not show any similarity to any other known protein including the cyclin family. Interestingly, the other isoform of p35, p39, is very similar in sequence to p35 up to amino acid residue 291, after which there is a 24-amino acid insertion in p39 (27). The ability of some p35 mutants described here to bind but not activate Cdk5 is in marked contrast to the situation with cyclin A, where all reported mutants that bind Cdc2 and Cdk2 also activate their kinase activities (38,59), suggesting a fundamental difference in the activation of Cdc2/Cdk2 by cyclin A and Cdk5 by p35. Indeed, the p25 truncation mutants characterized here are the first dominant-negative mutants of a CDK activating subunit to be described. The C⌬251 p35 mutant could prove useful in conjunction with Cdk5 mutants in probing p35⅐Cdk5 function in vivo.
Another group 6 have recently obtained results similar to ours and shown that a small deletion of the C-terminal region of p25 is sufficient to prevent Cdk5 activation. However, in contrast to our results, they found that a minimal deletion of the C-terminal region of p25 also abolished Cdk5 binding. This difference may due to the fact that we used a slightly longer version of p25 at the N terminus (from residue 109) than they did (from residue 145) or that the 12 vector-derived residues at the end of our fusion proteins have some effect; it is also possible that it is due to the more stringent binding assay conditions used in their studies.
Despite the fact that there is very little sequence similarity between p35 and cyclins, it has been predicted that p35 may have a cyclin-like structure (30). The cyclin box region of cyclin A has an ␣-helical fold comprised of five ␣-helices, and this fold is repeated in the C-terminal region following the cyclin box but shares little sequence similarity with the N-terminal fold (30,31). The region in p35 that is predicted to have a cyclin-fold structure is between residues ϳ135-227, which covers the region we found to be important in binding to Cdk5 (ϳ150 -200), although the minimal Cdk5 binding domain would lack the predicted helices 1 and 5. The C-terminal Cdk5-activating domain would be predicted to lie outside the "cyclin box" region and may have an interesting structure that distinguishes p35 from other cyclins. The fact that p35⅐Cdk5 activity does not require phosphorylation of Ser-159 in the Cdk5 activation loop implies that p35 interacts with this loop more extensively than the cyclins do, acting instead of phosphorylation to hold the activation loop in an active conformation. The regions in p35 that diverge in structure from other cyclins are presumably responsible for the different mechanism of Cdk5 regulation. The lack of similarity between p35 and cyclins may well account for the lack of binding to and inhibition by p21 cip1/WAF1 and p27 kip1 , since a key element of the inhibitory mechanism requires binding of the inhibitor protein to a groove in the cyclin (60). A specialized mechanism of p35 association with the Cdk5 activation loop could explain the lack of Cdk5 phosphorylation by CAK and Wee1.