Phosphorylation of the high molecular weight neurofilament protein (NF-H) by Cdk5 and p35.

The high molecular weight neurofilament protein (NF-H) is highly phosphorylated in the axon. The phosphorylation sites have been identified as KSP (Lys-Ser-Pro) repeats in the tail domain of NF-H. These KSP sequences are present more than 50 times in the NF-H tail, and most of these sites are normally phosphorylated in vivo. These KSP sites can be further divided into two separate consensus sequences, KSPXK and KSPXY (where Y is not K). The extensive phosphorylation of NF-H has been proposed to play a critical role in the determination of axonal diameter. Recent studies have shown that Cdk5, a kinase related to the cell cycle-dependent kinase Cdc2, is expressed in the brain and associates with the cytoskeleton. In vitro phosphorylation studies have shown that Cdk5 in conjunction with its activator, p35, is able to phosphorylate histone H1, dephosphorylated NF-H, as well as a synthetic peptide with the repetitive KSP motif. We have cloned the cDNAs for rat Cdk5 and p35 by reverse transcription-polymerase chain reaction and cDNA library screening and studied the phosphorylation of NF-H both in vivo and in vitro. By transient transfection assays, we have shown that NF-H can only be extensively phosphorylated in the presence of both Cdk5 and p35. This phosphorylation can be inhibited by a Cdk5-dominant negative mutant, an observation which further supports that Cdk5 is a kinase that is able to phosphorylate NF-H. By immunoprecipitating Cdk5 and p35 from the transfected cells, we have been able to show that the KSPXK repeats are the preferred phosphorylation sites for Cdk5, while the KSPXY repeats are not directly phosphorylated by Cdk5 and p35.

. The consensus sequence for phosphorylation has been identified as Lys-Ser-Pro (KSP) (7)(8)(9), which can be further divided into two separate consensus sequences, KSPXK and KSPXY (where Y is not K). There are about 50 potential phosphorylation sites in the rat, mouse, and human NF-H carboxyl-terminal tail domains (10), although the relative numbers of KSPXK and KSPXY repeats differ markedly among different species. Phosphate determinations have shown that most of these sites are phosphorylated in vivo (8,11,12). Rat NF-M contains only 5 KSP sites (13), and human NF-M has 12 KSP sites (14). Antibodies have been raised which can distinguish phosphorylated and nonphosphorylated KSP epitopes (15). Several studies have shown that after extensive dephosphorylation by alkaline phosphatase, NF-H migrates more rapidly on SDS-PAGE (5,16). This characteristic mobility shift upon phosphorylation of NF-H and the antibodies specific for the phosphorylated and nonphosphorylated KSP epitopes are therefore useful tools for studying its phosphorylation.
The functional role of neurofilament phosphorylation is still not completely clear. Early studies indicated that NF-H might function as a cross-linking protein whose phosphorylation was thought to be important for the formation of cross-linking bridges between neurofilaments (17,18). However, a later study showed that nonphosphorylated NF-H was also situated between intermediate filaments (19). More recently it has been suggested that the phosphate groups of NF-H may result in electrostatic repulsion, which in turn could increase axonal diameter (20,21). Phosphorylation of NF-H has also been correlated with its ability to interact with microtubules (22).
The kinase(s) which phosphorylate the KSP sequences on NF-H are still not completely identified. A number of different protein kinases, including casein kinase (23), cyclic AMP-dependent protein kinase (24), Ca 2ϩ -calmodulin-dependent protein kinase (25), and protein kinase C (26) are able to phosphorylate the neurofilament proteins. However, NF-M is a better substrate than NF-H for all these protein kinases, and the characteristic mobility shift on SDS-PAGE due to phosphorylation of NF-H is not observed. Other studies have revealed protein kinases which are associated with neurofilaments, including a protein kinase activity from bovine spinal cord neurofilament-enriched preparations (27) and a 115-kDa neurofilament-associated kinase isolated by affinity chromatography using bacterially produced NF-H (28). These protein kinases also only phosphorylate NF-H in a limited manner and there is no evidence that they phosphorylate NF-H at the KSP sites.
Recent results have pointed to the family of cyclin-dependent protein kinases (Cdks) as candidate kinases for the KSP sites on NF-H. The cell cycle kinase Cdc2 was shown to phosphorylate neurofilaments (29,30) and cause the characteristic gel mobility shift of NF-H upon phosphorylation (30). However, these results do not have much physiological relevance, since it is known that Cdc2 kinase is absent in terminally differentiated neurons. By using the polymerase chain reaction (PCR) with degenerate oligonucleotide primers corresponding to conserved regions of the Cdks, a family of novel Cdc2-related cDNA clones have been isolated (31). Among them, PSSALRE is expressed in neuronal cell lines and brain, as well as other cell lines and organs (31). This same kinase was also separately identified by another laboratory, shown to be able to bind cyclin D1 (32) and renamed as Cdk5 (cyclin-dependent kinase 5). A number of recent studies have shown that Cdk5 is associated with the cytoskeleton and able to phosphorylate NF-H in vitro. (33)(34)(35). An additional protein with a molecular mass of 23-25 kDa was observed in most of the Cdk5 preparations. cDNA cloning showed that this 23-25-kDa protein is a degradation product of a larger protein, p35, which acts as a neural-specific regulatory subunit of Cdk5 (36,37). p35 associates physically with and activates Cdk5 kinase. Although it has no homology with cyclins, it serves a similar function in modulating the activity of Cdk5.
In this study, we have cloned the cDNAs for both Cdk5 and p35 from rat brain and conducted NF-H phosphorylation studies by transient transfections and immunoprecipitation/kinase assays. We have demonstrated that NF-H can be phosphorylated only in the presence of both Cdk5 and p35. Furthermore, this phosphorylation is specific for the KSPXK sites.

MATERIALS AND METHODS
RT-PCR, PCR, cDNA Library Screening, and Plasmid Constructs-Rat Cdk5 cDNA was amplified from mRNA by the RT-PCR method. Briefly, 1 g of poly(A) ϩ RNA from rat brain (Clontech) was primed with random hexamers and reverse-transcribed with Moloney murine leukemia virus reverse transcriptase (RT) according to the manufacturer's protocol (Perkin-Elmer). For amplification of Cdk5 cDNA, the RT reaction was added to a PCR reaction mixture, containing reaction buffer, 2.5 units of Taq DNA polymerase, sense (5Ј-CGAAGCTTG-GACTCTTAGAACCGA-3Ј) and antisense (5Ј-TGGAAGCTTGGCTTA-AATAGGTCAGG-3Ј) oligonucleotide primers (Operon) corresponding to the published rat Cdk5 sequence. This reaction was heated to 94°C for 5 min, cooled to 72°C for 5 min, and cycled 30 times at 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min. The resulting rat Cdk5 PCR product was subcloned into the pGEM-7 (Promega) and pRSVi vectors and sequenced in its entirety from both directions.
Bovine p23 cDNA was obtained by the same approach with sense (5Ј-GGCATATGTCGTCCGTCAAGAAGG-3Ј) and antisense (5Ј-ATGC-ATATGGCTGGCGGGCTCACC-3Ј) primers corresponding to the published sequence of bovine p23 (34). The bovine p23 PCR product was subcloned into pGEM-5 and pRSVi vectors and used in screening a rat brain gt11 cDNA library (generously provided by Dr. David Colman at Mt. Sinai Medical Center). Standard cDNA library screening procedures were employed (38). Briefly, the 0.5-kb HindIII fragment containing the bovine p23 cDNA was purified from pRSVi-p23, labeled with 32 P by random primer synthesis (Amersham) and used as a probe for screening the amplified gt11 rat brain cDNA expression library (13).
After screening approximately 1 ϫ 10 5 plaques, 20 positive plaques were recovered and confirmed to correspond to be rat p35 by secondary and tertiary screening. The largest p35 cDNA clone was isolated from the phage, subcloned into the EcoRI site of pGEM-7 (Promega), and sequenced in its entirety from both directions.
The Cdk5 dominant negative construct was generated by two-step PCR (39). The first amplification was performed using an antisense primer corresponding to the amino acid residues KLAD*FGL (where * marks the point of a mutation, which contains a single G to A mismatch resulting in a Asp 3 Asn mutation (40)) and a sense primer corresponding to the 5Ј end of Cdk5 described above. The resulting PCR product was purified and used with the Cdk5 antisense primer described above for the second amplification. The final PCR product was subcloned into the pRSVi vector and sequenced to confirm the Asp 3 Asn mutation.
The 3.4-kb cDNA of rat NF-H encoding the full-length polypeptide was previously cloned into the pRSVi-HindIII vector (10), which contains the Rous sarcoma virus long terminal repeat (41). The full-length NF-H cDNA was also cloned into the expression vector pET-16b (Novagen Inc.). The truncated NF-H tail constructs which contain either the KSPXK or KSPXY sequences were generated by PCR. Two pairs of primers (Operon) were designed according to the NF-H sequence (10). Primers KSP1 (5Ј-GAACATATGTCTCCTGTGAAAGAA-3Ј) and KSP2 (5Ј-TCTGGATCCTTTATGGAGATTTTA-3Ј) were for the KSPXY-containing construct, whereas primers KSP3 (5Ј-GCCCATATGCCCGT-GAAGGAAGGT-3Ј) and KSP4 (5Ј-GAGGGATCCGGGCTATTCTGGG-TG-3Ј) were for the KSPXK-containing construct. PCR was carried out as described above, and the resulting PCR products were subcloned in pET16b at the NdeI and BamHI sites and sequenced.
Cell Culture and Transient Transfections-The SW13.cl.2Vim Ϫ cell line was the kind gift of Dr. Robert M. Evans (University of Colorado Health Science Center, Denver, CO) and was grown in Dulbecco's modified Eagle's/Ham's F12 medium (Life Technologies, Inc.) supplemented with 5% fetal bovine serum at 37°C in a humidified atmosphere of 5% CO 2 . All the transient transfections in this study were done using the calcium phosphate precipitation procedure previously described (44). The transiently transfected cells were lysed in 6.25 mM Tris buffer, pH 6.5 and 1% SDS, an equal amount of sample buffer was added, and the samples were boiled for 10 min.
SDS-PAGE and Immunoblot Analysis-SDS-PAGE was performed by the method of Laemmli (43) in different percentage vertical slab gels. For immunoblots, the proteins were separated by SDS-PAGE, transferred to nitrocellulose (Schleicher and Schuell) and probed with the different antibodies. ECL (Amersham) was performed according to the manufacturer's protocols, and the blots were exposed to x-ray film (Kodak).
Antibodies-Antibodies SMI36 and SMI32 were purchased from Sternberger Monoclonals Inc. The Cdk5 antibodies C-8 and DC-17 were purchased from Santa Cruz Biotech.
Expression and Purification of the Recombinant Proteins-The pET constructs encoding NF-H, "KSPXY" and "KSPXK" proteins were transformed into the Escherichia coli strain BL21(DE3). The transformed bacteria were cultured in 500 ml of LB containing 100 g/ml ampicillin to A 600 of 0.5 at 37°C. After adding isopropyl-1-thio-␤-D-galactopyanoside to a final concentration of 0.4 mM, the cultures were incubated another 4 h. The bacteria were harvested by centrifugation at 2000 ϫ g and the "KSPXY" and "KSPXK" proteins, which contained the histidine tag from pET16b were further purified according to the manufacturer's protocol (Novagen). Briefly, the bacterial pellets were sonicated in binding buffer containing 5 mM imidazole, 0.5 M NaCl, and 20 mM Tris-HCl, pH 7.9, and centrifuged at 20,000 ϫ g to collect the inclusion bodies and cellular debris, while leaving other proteins in solution. Repeating the previous step several times released more of the trapped soluble proteins. The final pellets were sonicated in binding buffer prepared in 6 M urea and incubated on ice for 1 h. The remaining insoluble material was removed by centrifugation at 39,000 ϫ g, and the supernatants were filtered through 0.45-micron membranes. For the "KSPXK" and "KSPXY" proteins, the filtered supernatants were loaded onto NiSO 4 columns, equilibrated according to the manufacturer's protocol, washed with binding buffer, and eluted with elution buffer. All the buffers used in chromatography were prepared in 6 M urea. The eluted fractions were dialyzed overnight against phosphate-buffered saline at 4°C, and the protein concentrations were determined by the Bradford assay (Bio-Rad). The NiSO 4 column was not effective for the purification of full-length NF-H, and we therefore purified full-length NF-H using an HTP column as described elsewhere (42); the protein concentrations were measured by the same method.
Immunoprecipitation and in Vitro Kinase Assay-Transiently transfected cells were washed with phosphate-buffered saline and incubated in lysis buffer containing 50 mM Tris-HCl, pH 7.5, 250 mM NaCl, 5 mM EDTA, 5 mM dithiothreitol, 0.1% Nonidet P-40 (Sigma), and 1 ϫ proteinase inhibitors (5 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin) for 10 min at 4°C. Cells were then scraped into Eppendorf tubes and incubated on ice for 1 h. The cell suspensions were pelleted by centrifugation at 14,000 ϫ g for 10 min, and the supernatants were mixed with 1 g of antibody for 2 h at 4°C. 20 l of Protein G Plus-Agarose (Santa Cruz Biotech) was added, and the mixtures were subsequently incubated for 2 h at 4°C. The immune complexes were pelleted by centrifugation at 14,000 ϫ g and washed with cold lysis buffer four times. The kinase assays were initiated by mixing kinase buffer containing 50 mM Hepes, pH 7.4, 5 mM MgCl 2 , 5 mM MnCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 100 mM NaCl, 1 Ci of [␥-32 P]ATP (Amersham), and 2 mg of substrate with the final immune complex pellet. The reaction mixtures were then incubated in a 30°C water bath for 30 min and centrifuged at 14,000 ϫ g for 10 min. The supernatants were saved and the pellets were washed 3 times. An equal amount of 2 ϫ sample buffer (43) were added to the supernatants and pellets, and the samples were boiled for 10 min before SDS-PAGE.

RESULTS
Cloning of Rat Cdk5, p35, and GSK3␤ cDNAs-In order to study Cdk5 and its activator, p35, we first isolated their cDNAs by RT-PCR. To obtain the Cdk5 cDNA, reverse transcription was carried out using rat brain mRNA as template, and the resulting RT-PCR product was amplified by two primers complementary to the 5Ј and 3Ј ends of the published Cdk5 cDNA sequence (45). This PCR product was subcloned into the pGEM-5 vector, its sequence was determined and confirmed to be identical with the Cdk5 sequence from GenBank TM .
The full-length rat p35 cDNA was obtained by first isolating a cDNA corresponding to the partial sequence of bovine p35 (p23) by RT-PCR using bovine brain mRNA as template (46) and screening a rat brain cDNA library using the bovine p23 cDNA as a probe. The resulting p35 cDNA was subcloned into pGEM-5 and sequenced. The deduced amino acid sequence of rat p35 is nearly identical with those of human and bovine p35 (Fig. 1). There are only four amino acid differences between rat and human p35 (36) or between rat and bovine p35 (37). The Cdk5 dominant-negative (Cdk5dn) construct was generated by PCR and subcloned into the expression vectors described under "Materials and Methods." This dominant-negative Cdk5dn contains a point mutation resulting in a single amino acid residue change (Asp 144 3 Asn) in the kinase domain of Cdk5. This mutant form has previously been shown to result in an inactive kinase, which competes with the activator protein p35 and thereby inhibits Cdk5 kinase activity (40).
A full-length cDNA for glycogen synthase kinase 3␤ (GSK3␤) was also generated by RT-PCR using primers based on the published sequence (61). The resulting GSK3␤ cDNA was subcloned into pGEM-5 and sequenced. The sequence was found to be identical with the GSK3␤ sequence obtained from GenBank TM .
NF-H, "KSPXY," and "KSPXK" Constructs-Rat NF-H con-FIG. 1. Nucleotide sequence and deduced amino acid sequence of rat p35. There are four differences between the predicted amino acid sequences of rat and human p35 (amino acids: Ala 68 , Ser 76 , Ile 133 , and Thr 134 ) and four differences between rat and bovine p35 (amino acids: Ala 68 , Ser 76 , Val 121 , and Ile 133 ). The cDNA corresponding to bovine p23 was isolated using RT-PCR and corresponds to amino acids 125-307. GenBank TM accession number: U50707. tains 52 KSP repeats, which can be divided into KSPXK and KSPXY repeats. After analyzing the protein sequence, we noticed that most of the KSPXY repeats are clustered in the amino-terminal part of the NF-H tail domain, while most of the KSPXK repeats are located after the KSPXY repeats. Using PCR, we were able to separate these 2 groups of repeats and prepare a "KSPXY" construct which contains 41 KSPXY repeats and no KSPXK repeats and a "KSPXK" construct which contains 7 KSPXK repeats (and 2 KSPXY repeats). These constructs were subcloned into the pET16b expression vector (Fig.  2). It should be noted that a single KSPXK is present in front of the cluster of the KSPXY repeats in the NF-H sequence and the "KSPXY" construct starts from the middle of this KSPXK motif.
Transient Transfections Show Cdk5 in Conjunction with p35 Phosphorylates NF-H in Cultured Cells-To study the phosphorylation of NF-H by Cdk5, transient transfections were performed with SW13cl2Vim Ϫ cells, a human adrenal carcinoma cell line without cytoplasmic intermediate filaments (47). We have previously shown by Western blot analysis and immunostaining that NF-H is predominantly in the nonphosphorylated form in transiently transfected fibroblasts (10). The same result is obtained from SW13cl2Vim Ϫ cells transiently transfected with NF-H alone (Fig. 3B). Cdk5 is expressed in a number of cell lines and tissues, but its activity has been shown to be present only in brain (48). Consistent with this observation, Cdk5 protein can be detected in SW13cl2Vim Ϫ cells by Western blot analysis (Fig. 3C), but the immunoprecipitated Cdk5 from cells lacking p35 does not have kinase activity on histone H1 (Fig. 4A), NF-H (Fig. 4B), "KSPXK" (Fig. 4C), or "KSPXY" proteins (Fig. 4D). Several recent reports have shown that p35 is only expressed in the central nervous system (36,37). Consistent with these reports, our Northern blot analysis showed that the SW13cl2Vim Ϫ cell line does not have any p35 mRNA (data not shown). These results indicate that the SW13cl2Vim Ϫ cell line is a good system to study the phosphorylation of NF-H by Cdk5 and its activator, p35.
For every transient transfection experiment, 20 g of each plasmid was used, and the resulting cell lysates were analyzed by Western blots using antibodies SMI36 and SMI32. SMI36 is specific for the phosphorylated KSP epitope on NF-H, while SMI32 recognizes this epitope in the nonphosphorylated form (15). Purified NF-H from rat spinal cord shows an apparent molecular mass of approximately 200 kDa on SDS-PAGE. After alkaline phosphatase treatment, the dephosphorylated NF-H migrates faster on SDS-PAGE (5,49). We used this mobility shift on SDS-PAGE and the immunoreactivity of NF-H with the two phosphorylation state-dependent antibodies as the criteria to determine whether or not the transfected NF-H was phosphorylated. As shown in Fig. 3, when NF-H is transfected by itself (lane 2) or co-transfected with Cdk5 (lane 3) into SW13cl2Vim Ϫ cells, the exogenously expressed NF-H protein is detected only by SMI32 and does not show the mobility shift characteristic of the phosphorylated NF-H, indicating the absence of any significant amount of NF-H phosphorylation (Fig.  3, A and B). However, when both Cdk5 and p35 are transfected along with NF-H (Fig. 3, lane 6), the expressed NF-H shows an apparent molecular mass of 200 kDa, comparable to that of the phosphorylated NF-H. Furthermore, the protein is recognized by SMI36 (Fig. 3A), which is specific for the phosphorylated NF-H, but not by SMI32 (Fig. 3B), which detects the nonphosphorylated KSP sequences on NF-H. From these results, it is quite clear that the NF-H expressed in the transfected cells is extensively phosphorylated. When NF-H is co-transfected with only p35 (Fig. 3, lane 5), a slight mobility shift of the NF-H band is observed on the immunoblots and this broadly diffuse band is recognized by both SMI36 and SMI32 (Fig. 3, A and B). These results are consistent with partial phosphorylation of NF-H by the endogenous Cdk5 activated by the transfected p35. However, the endogenous Cdk5 is apparently not present in sufficient quantity to phosphorylate NF-H to the same degree as the exogenous Cdk5 (Fig. 3, A and B, lanes 5 and 6). When the dominant-negative Cdk5dn is introduced into the cells along with NF-H, p35, and Cdk5 (lane 7), the large mobility shift of NF-H is not observed; instead the exogenously expressed NF-H shows a similar electrophoretic mobility pattern as the one from cells transfected with only NF-H and p35 (Fig. 3, A and B). These results indicate that Cdk5dn competes with Cdk5 for p35 and thereby inhibits the phosphorylation of NF-H by Cdk5. Taken together, all the results from the transfection experiments suggest that Cdk5 in conjunction with p35 can extensively phosphorylate NF-H in cultured non-neuronal cells.
In order to show the levels of Cdk5 and Cdk5dn in the transfected cells, the same Western blots were immunostained with the Cdk5 antibody CD17 (Fig. 3C). The endogenous Cdk5 can be observed along with the transfected Cdk5 or Cdk5dn. The mutant can be distinguished readily from the endogenous Cdk5, because its mobility is different from the endogenous Cdk5 due to the presence of a hemagglutinin (HA) tag (Fig. 3C). From the intensity of the Cdk5 band, it is obvious that the level of the exogenous Cdk5 is much higher than the endogenous Cdk5. Like Cdk5, GSK3␤ has been shown to phosphorylate SP sites on the microtubule-associated protein, , and to associate with the cytoskeleton. Recent transfection studies have shown that overexpression of GSK3␤ and in non-neuronal cells resulted in phosphorylation on SP sites, which have been associated with the abnormal phosphorylation of in Alzheimer's disease.
To determine if NF-H could serve as a substrate for GSK3␤, we co-transfected GSK3␤ and NF-H in the SW13 cells (Fig. 3, lane  8). The results show that no significant amount of SMI36reactive NF-H can be observed in the cells transfected with GSK3␤ (Fig. 3A) and there is also no apparent shift observed with SMI32 (Fig. 3B). These experiments indicate NF-H phosphorylation is not induced to a significant extent by the overexpression of GSK3␤.
Immunoprecipitation Experiments Show That Cdk5 Is Able to Phosphorylate Only the KSPXK Sequence on NF-H-It has been noted that NF-H has two different consensus phosphorylation KSP sequences, KSPXK and KSPXY. In an effort to determine which one of these sequences is the phosphorylation site for Cdk5 and to show that the results in Fig. 3 are not due to another kinase activated by Cdk5/p35, we performed immunoprecipitation experiments followed by kinase assays. Immunoprecipitations were conducted by using the Cdk5 polyclonal antibody C-8 on protein extracts from cells, which had been transiently transfected with various combinations of Cdk5, p35, and Cdk5dn constructs. After the immunoprecipitated Cdk5 complexes were immobilized on Protein G beads, kinase buffers containing the different substrates, histone H1, NF-H, "KSPXY" and "KSPXK" proteins were mixed with the beads. The mixtures were incubated in the presence of [␥-32 P]ATP and separated on SDS-PAGE, and the gel was subsequently exposed to x-ray film.
Purified, bacterially expressed NF-H, "KSPXK," and "KSPXY" proteins were not phosphorylated as determined by Western blot analysis (data not shown). Histone H1 was used as a positive control in these experiments, since it is a common kinase substrate for most Cdks and has previously been used for Cdk5 in vitro assays. Immunoprecipitates of the Cdk5 and p35 co-transfections yielded the most dramatic phosphorylation of histone H1, NF-H, and the "KSPXK" proteins ( Fig. 4, A,  B, and C, lane 5). There was no phosphorylation of the "KSPXY" protein by Cdk5 and p35 under the same conditions used for the other substrates (Fig. 4D). Cdk5 or Cdk5dn alone could not phosphorylate any of the substrates (Fig. 4, lanes 2  and 3). A long exposure of the autoradiograms showed that some phosphate incorporation on histone H1, NF-H, and "KSPXK" protein occurred in the immunoprecipitation/kinase reaction from the p35-transfected cells (Fig. 4, lane 4). Thus, from the in vitro kinase assay results, we conclude that Cdk5 in conjunction with p35 is able to phosphorylate NF-H, and its phosphorylation site is on the KSPXK sequence in the NF-H tail. Furthermore, these results are consistent with the in vivo phosphorylation data obtained from the transfection experiments (Fig. 3). DISCUSSION Neurofilament phosphorylation has been postulated to play an important role in maintaining axonal caliber (1). The highly FIG. 4. In vitro kinase assays. Extracts from cells, which had been transiently transfected with combinations of Cdk5, p35, and Cdk5dn constructs were immunoprecipitated with the Cdk5 antibody C-8 and used for the in vitro kinase assays. Histone H1 was used as a positive control (A). Bacterially expressed NF-H (B), "KSPXK" (C), and "KSPXY" (D) proteins were purified and used as substrates for the in vitro kinase assays. Lane 7 shows the Coomassie Blue stain of the substrates. phosphorylated carboxyl-terminal tail domains of NF-H and NF-M are unique among intermediate filaments and are presumed to mediate neuron-specific functions, including maintaining the shape of the axon, which presumably requires a specialized cytoskeleton. The kinase(s) responsible for the phosphorylation of the NF-H tail domain have not been definitively identified. Several lines of evidence have pointed to Cdk5 as a candidate kinase for the phosphorylation of NF-H. Cdk5 is a kinase which was originally identified by homology to the cyclin-dependent kinase, Cdc2 (31). Several other reports identified a protein kinase activity in the central nervous system, which was co-isolated with the cytoskeleton and upon further characterization turned out to be Cdk5 (45,50,51). In a number of these studies, NF-H or a synthetic peptide containing the KSP repeats was used as the substrate for these kinase assays. In addition, the microtubule-associated protein, , was shown to be a substrate for Cdk5 (52). It was noted from early studies that Cdk5 and its mRNA are expressed in a number of cell lines and tissues. However, immunoprecipitation experiments showed that Cdk5 kinase activity was only detected in brain (48) suggesting the presence of a brain-specific activator. Two different activator proteins have been proposed for Cdk5. One of these, p35, is only expressed in brain (36,37), and phosphorylation of histone H1 by Cdk5 is observed when p35 and Cdk5 are co-transfected into non-neuronal cells.
In vitro kinase assays using bacterially produced Cdk5 and p35 also showed that p35 is able to activate Cdk5 to phosphorylate histone H1 (36). A second putative activator, p67, has been identified in purified kinase preparations containing Cdk5 and shown to activate bacterially expressed Cdk5 in in vitro kinase experiments (53). p67 is identical to nsec-1 (rbsec1 or munc-18) and plays a role in exocytosis (54), but has not been reported to be a kinase activator. In this report, we have focused our attention on the first of these activators, p35.
In order to show that Cdk5 in the presence of p35 is able to phosphorylate NF-H in vivo, we performed transient transfections with various combinations of the Cdk5, p35, and NF-H expression constructs into the SW13 cell line. We have previously shown that NF-H is not significantly phosphorylated when transfected in fibroblasts. In the present study, we have also determined that in SW13 cells, NF-H remains predominantly in the nonphosphorylated state (Fig. 3B). Although Cdk5 is expressed in untransfected SW13 cells, the endogenous Cdk5 does not show any kinase activity as determined by the lack of phosphorylation of NF-H or histone H1, a commonly used substrate for cyclin-dependent kinases (Fig. 4A). By Northern blot analysis, we have determined that p35 mRNA is not expressed in SW13 cells, a result consistent with the observed lack of kinase activity from the endogenous Cdk5.
Several lines of evidence presented in this study show that Cdk5, in conjunction with p35, is capable of phosphorylating NF-H in vivo. Co-transfections of Cdk5, p35, and NF-H resulted in highly phosphorylated NF-H as shown by the significant mobility shift of NF-H on SDS-PAGE (Fig. 3A). In addition to the mobility shift, NF-H, which has been co-transfected with Cdk5 and p35 is readily recognized by an antibody specific for the phosphorylated form of NF-H, but not by an antibody to the nonphosphorylated form of NF-H. Since it is possible that the phosphorylation of NF-H is not directly by Cdk5, but rather by a kinase, which is in turn activated by Cdk5, we performed the immunoprecipitation experiments. These experiments show that immunoprecipitated Cdk5 can phosphorylate NF-H provided p35 is present (Fig. 4). These experiments also show that of the two possible consensus sequences for NF-H phosphorylation, Cdk5 is able to phosphorylate only one sequence (KSPXK) and not the other (KSPXY). Another line of evidence for the identification of NF-H as a substrate for Cdk5 comes from the experiment with the dominant-negative Cdk5dn, which can compete with Cdk5 and reduce its kinase activity on NF-H.
The data presented in this paper, as well as data from others using synthetic peptides, have demonstrated that the KSPXK repeats on NF-H are the preferred sites of phoshorylation by Cdk5. The lack of reactivity of SMI32 against exogenously expressed NF-H from cells co-transfected with Cdk5 and p35 is therefore somewhat surprising, since the KSPXK motif accounts for only ϳ20% of the KSP sites of rat NF-H and might indicate that the SMI32 antibody is specific for this motif in the nonphosphorylated form. Alternatively, additional phosphorylation of NF-H on the KSPXY sites by Cdk5 (or by a different kinase) could occur only after prior phosphorylation of the KSPXK sites by Cdk5. It is therefore interesting to note that recent studies have shown that in vitro phosphorylation of recombinant by Cdk5 and by glycogen synthase kinase 3␤ (GSK3␤) resulted in the phosphorylation of some of the same sites as from Alzheimer's disease-paired helical filaments (57,58). This abnormal phosphorylation of has also been observed in cultured cells in which GSK3␤ and were overexpressed by transient transfections (59) In vitro experiments have shown that becomes a better substrate for GSK3␤ after it is first phosphorylated by Cdk5 (60). In our experiments, we could not detect any significant level of phosphorylation of NF-H by transient co-transfection with GSK3␤ (Fig. 3). However, it is possible that in the cells co-transfected with NF-H, Cdk5, and p35 constructs, further phosphorylation of NF-H by the endogenous GSK3␤ occurred after the KSPXK sequences were phosphorylated by Cdk5/p35. It is also interesting to note that although is a substrate of Cdk5 it contains a KSPXY sequence, but does not have a KSPXK sequence.
Even though Cdk5 is related to the cyclin-dependent kinase family, it has several unique features, such as its expression in terminally differentiated neurons and the fact that its activator, p35, has nearly no homology to cyclins. We have observed no differences in the ability of rat p35 and bovine p23 (which corresponds to the carboxyl-terminal portion of p35) to activate Cdk5 in the co-transfection experiments. These results are consistent with the report showing kinase activity from the bacterially produced Cdk5 and truncated p35 (55) and support the notion that the carboxyl-terminal portion of p35 is sufficient to activate Cdk5. What role, if any, the amino-terminal part of p35 plays in regulating the Cdk5 activation remains an open question. Recently, a p35 isoform, which is able to activate Cdk5 in vitro, has been isolated from human brain (56). It will be interesting to see whether p35 belongs to a larger protein family and whether Cdk5 or other related kinases can be activated by different members of this family. The restricted expression of p35 to the central nervous system also raises the possibility that in the peripheral nervous system, other protein kinases and/or activators are present which phosphorylate NF-H.