Abbreviations used : AD, Alzheimer's disease ; GSK-3, glycogen synthase kinase 3 ; MAP, mitogen-activated protein ; PHF, paired helical filament ; PKA, cyclic AMP-dependent protein kinase ; TBS, Trisbuffered saline ; TBS+, Tris-buffered saline containing 1% bovine serum albumin and 5% normal goat serum.
Address correspondence and reprint requests to Dr. P. Davies at Departments of Pathology and Neuroscience, Albert Einstein College of Medicine, Forcheimer 526, 1300 Morris Park Avenue, Bronx, NY 10461, U.S.A.
Abstract : Immunoaffinity-purified paired helical filaments (PHFs) from Alzheimer's disease (AD) brain homogenates contain an associated protein kinase activity that is able to induce the phosphorylation of PHF proteins on addition of exogenous MgCl2 and ATP. PHF kinase activity is shown to be present in immunoaffinity-purified PHFs from both sporadic and familial AD, Down's syndrome, and Pick's disease but not from normal brain homogenates. Although initial studies failed to show that the kinase was able to induce the phosphorylation of tau, additional studies presented in this article show that only cyclic AMP-dependent protein kinase-pretreated recombinant tau is a substrate for the PHF kinase activity. Deletional mutagenesis, phosphopeptide mapping, and site-directed mutagenesis have identified the PHF kinase phosphorylation sites as amino acids Thr361 and Ser412 in htau40. In addition, the cyclic AMP-dependent protein kinase phosphorylation sites that direct the PHF kinase have been mapped to amino acids Ser356 and Ser409 in htau40. Additional data demonstrate that these hierarchical phosphorylations in the extreme C terminus of tau allow for the incorporation of recombinant tau into exogenously added AD-derived PHFs, providing evidence that certain unique phosphorylations of tau may play a role in the pathogenesis of neurofibrillary pathology in AD.
Alzheimer's disease (AD), classified as a neurodegenerative disorder, is the most prevalent cause of dementia in the elderly population. The two main pathological hallmarks of AD are the intraneuronal neurofibrillary tangle, composed of paired helical filaments (PHFs), and the senile plaque, composed of abnormally processed aggregates of β-amyloid interwoven with dystrophic neurites containing PHFs (Khachaturian, 1985). The main constituent of PHF is the microtubule-associated protein tau, which is posttranslationally altered in AD (Wood et al., 1986 ; Kosik et al., 1988 ; Lee et al., 1991). As such, much research has focused on elucidating the pathological alterations in tau that may contribute to the formation of PHFs in AD. Tau in PHFs (PHF-tau) differs from normal adult tau in its state of phosphorylation, with PHF-tau being hyperphosphorylated at as many as 20 amino acid residues as compared with normal adult tau (Grundke-Iqbal et al., 1986 ; Wood et al., 1986 ; Lee et al., 1991 ; Goedert et al., 1992 ; Kanemaru et al., 1992 ; Morishima-Kawashima et al., 1995). Aberrant activation of certain neuronal kinases and inactivation of certain neuronal phosphatases have been postulated to be the proximal cause of PHF formation in AD (for review, see Trojanowski and Lee, 1994, 1995).
Recent evidence has suggested that the hyperphosphorylation of tau seen in AD is due to the inactivation of protein phosphatases that are normally functional and quite active in the postmortem interval and that many of the previously-thought-to-be-pathological phosphorylations of tau are present in biopsy-derived tissues from non-Alzheimer patients (Matsuo et al., 1994 ; Hasegawa et al., 1996). These discoveries have been the impetus for the renewed search for true Alzheimer-specific events. Several Alzheimer-unique phosphorylations have been identified in PHF-tau by either phosphopeptide mapping strategies or antibody detection (Hasegawa et al., 1992, 1996 ; Matsuo et al., 1994). It is likely that this new class of unique phosphorylation events plays a direct role in the pathological modifications of tau leading to the neurofibrillary degeneration characteristic of AD and requires further study.
Recently, a method for immunoaffinity-purifying PHF from AD brain homogenates has been developed using an isotype-switched form of the Alz-50 monoclonal antibody, p42, linked to Sepharose beads (Vincent and Davies, 1992). Standard methods of PHF purification from AD brain homogenates include the use of harsh denaturing agents that inactivate associated enzymatic activities (Gonzalez et al., 1992 ; Greenberg et al., 1992), but the methods used for immunoaffinity purifying PHFs are relatively mild and have been shown to preserve a PHF-associated kinase activity (Vincent and Davies, 1992). Immunoaffinity-purified PHFs appear to be quite soluble in standard buffers and presumably represent an early stage in the pathogenesis of PHF formation in AD (Vincent and Davies, 1992). Thus, enzymatic activities associated with these preparations may contribute to the pathogenetic mechanisms involved in PHF formation.
Analysis of immunoaffinity-purified PHFs has shown that a serine/threonine kinase activity (PHF kinase) is present in these preparations. This PHF kinase has been shown to be able to induce the phosphorylation of PHF constituents on addition of exogenously added Mg2+ and ATP (Vincent and Davies, 1992). In addition, the level of phosphate incorporation can be increased by prior incubation with alkaline phosphatase, suggesting that the PHF kinase phosphorylation sites are partially phosphorylated in PHFs and that this phosphorylation is at least partially responsible for the pathological modifications of PHF-tau seen in AD (Vincent and Davies, 1992). Many neuronal kinases and phosphatases have been implicated in the pathogenesis of PHF formation (for review, see Trojanowski and Lee, 1994, 1995) ; however, only the PHF kinase has been shown to copurify with this pathological hallmark of AD (Vincent and Davies, 1992).
The present study demonstrates that (a) PHF kinase activity is present in immunoaffinity eluents from all PHF-associated diseases but not from normal individuals, (b) only PKA phosphorylation of tau on Ser356 and Ser409 allows the PHF kinase to phosphorylate sequentially recombinant tau in vitro, and (c) the PHF kinase induces the phosphorylation of Thr361 and Ser412 on PKA-pretreated recombinant tau in vitro. These data illustrate the specificity of the PHF kinase for AD and other PHF-associated diseases, provide further support for the hypothesis that the sequential actions of multiple protein kinases are required to produce the hyperphosphorylation of tau seen in AD, and positively identify the major PHF kinase-induced phosphorylation of tau (Ser412) as a previously identified AD-specific event (Hasegawa et al., 1992). They also raise the possibility that the phosphorylation of Thr361 in PHFs may have been overlooked in previous mapping studies of PHF-tau and demonstrate the unique substrate specificity of the PHF-associated protein kinase on tau, providing some insight into the potential functional consequences of these unique phosphorylations on tau and the role they may play in the pathogenesis of neurofibrillary pathology in AD.
MATERIALS AND METHODS
Antibodies and immunoaffinity columns
Alz-50 is an IgM monoclonal antibody raised to AD basal forebrain homogenates (Wolozin et al., 1986). An isotype switched form of the Alz-50 antibody, p42, was conjugated to agarose beads to create an Alz-50 immunoaffinity column (Vincent and Davies, 1992). The MC-1 monoclonal antibody was raised to p42-immunoaffinity-purified PHF and has been shown to recognize a similar pathological conformation of the tau molecule that is recognized by Alz-50 (Jicha et al., 1997). MC-1, however, appears to have greater affinity and an increased specificity for pathological conformations of tau than Alz-50 (authors' unpublished data). To exploit these advantages, MC-1 antibody was purified from serum-free Hybridoma Medium SPF (Fisher Scientific) on protein A and coupled to Affi-gel 10 (Bio-Rad Laboratories) in 50 mM sodium phosphate (pH 8.1) at a ratio of 2-5 mg of protein/ml of gel. These conditions afforded binding of >95% of the antibody to the gel. Columns that contained a ratio of <2 mg of protein/ml of gel did not bind appreciable concentrations of antigen.
Immunoaffinity purification of PHFs
Purified, soluble PHF preparations were obtained according to the procedures of Vincent and Davies (1992). The MC-1 antibody/gel mixture was poured into columns with a length to diameter ratio of at least 20 and washed with 50 volumes of Tris-buffered saline (TBS ; 10 mM Tris and 150 mM NaCl, pH 7.4). Before purification of antigen, the columns were washed with 2 volumes of 3 M KSCN, followed with 5 volumes of TBS. Brain extracts were made by homogenizing 1 g of dissected cortical gray matter in 10 ml of TBS containing 1 mM phenylmethylsulfonyl fluoride using a PowerGen Polytron (Fisher Scientific) on maximal setting. Homogenates were centrifuged at 27,000 g for 20 min, and the supernatants were decanted. Pellets were rehomogenized in TBS using half of the original volume and recentrifuged. The supernatants then were pooled and filtered through a 4-cm-high, 3.5-cm-diameter column of Sepharose 400 superfine to prevent particulate matter from accumulating at the top of the MC-1 columns. The final extracts were eluted through the MC-1 columns at a flow rate of ~25 ml/h for up to 36 h, after which the columns were washed with TBS at flow rates of >100 ml/h for at least 18 h. Bound antigen was eluted with 3 M KSCN, and fractions were assayed for protein by the method of Bradford (1976). Samples containing at least 50 μg of protein/ml were pooled and dialyzed against TBS. The MC-1 columns were reused at least six times before replacement of the gel and demonstrated no reduction in the capability of the column to bind antigen.
Recombinant tau constructs and protein expression
Clone htau40 was the generous gift of M. Goedert (Goedert et al., 1989). Clone 10 encoding for a fetal tau construct missing exons 4 and 10—an N-terminal insert and the second microtubule binding repeat, respectively—was cloned from a human fetal λgt11 expression library. Clone htau40 was digested with NdeI and EcoRI, blunted with Klenow enzyme, and ligated into SmaI-digested pQE-32 (Qiagen) to produce a histidine-tagged recombinant protein, TauW. The cDNA encoding TauW was digested with HindIII and religated to itself producing a construct encoding amino acids 1-342 of htau40, TauNW. Clone 10 and the cDNA encoding TauW were digested with PstI and ligated into PstI-digested pQE-31, to create cDNAs encoding for TauP and TauP + exon 10, respectively. A series of site-directed mutants were constructed by oligo-directed mutagenesis, including S356A (in TauP + exon 10), S409G (in TauP), S416G (in TauP), S412A (in TauP), S413A (in TauP), and S412A + S413A + T414A (triple mutant in TauP) individually. All constructs (Fig. 1) were chemically transformed into Escherichia coli strain MC-15 harboring the pREP4 plasmid. Cells were grown to an absorbance at 600 nm of 0.5, and protein expression was induced by addition of isopropyl β-d-thiogalactopyranoside (Boehringer-Mannheim) to 2 mM for 5 h. Cell cultures were then centrifuged at 5,000 g for 5 min, and the cell pellet was resuspended in 50 mM sodium phosphate and 300 mM NaCl (pH 8.0) and frozen overnight at -20°C. Cell suspensions were thawed on ice, sonicated three times each for 5 min to lyse cells, and spun at 5,000 g for 5 min. Supernatants were then boiled for 10 min at 100°C and recentrifuged for 5 min at 5,000 g. Supernatants were then applied to a column of Ni2+-conjugated agarose equilibrated in lysis buffer, washed with 50 column volumes of lysis buffer, washed with 200 column volumes of wash buffer (50 mM sodium phosphate, 300 mM NaCl, and 10% glycerol, pH 6.3), and eluted with wash buffer adjusted to pH 4.5. The eluted sample was then dialyzed overnight in three changes of TBS in preparation for phosphorylation reactions. Protein concentrations were assayed by the method of Bradford (1976), and concentrations were adjusted to 0.2 mg/ml.
Four micrograms of substrate was used in each reaction, and the total reaction volume was included in each subsequent step. PKA, protein kinase C, casein kinase II, p34-cdc2, MAP kinase (p44), and GSK-3 were all purchased from Upstate Biotechnology, and reactions were carried out according to the manufacturer's specifications. Activated cdk5 was the generous gift of Dr. J. Wang and was reacted as described previously (Paudel et al., 1993). All primary phosphorylation reactions were incubated for 16 h to allow for maximal phosphorylation and to aid in kinase inactivation in preparation for sequential phosphorylation reactions. PHF preparations containing the associated protein kinase activity were prepared and reacted as previously described (Vincent and Davies, 1992). In brief, PHF kinase reactions were carried out in TBS with addition of 5 mM MgCl2 and 200 μM ATP. PHF kinase phosphorylation reactions were carried out for 2 h at 25°C. Twenty microliters of PHF preparations was used for all lanes of Fig. 2. Six micrograms of casein was included in each sample to control for variability in substrate concentration in each individual PHF preparation. PHF preparations from sporadic AD cases containing 1 μg of total protein were used in all other phosphorylation reactions requiring the PHF kinase activity. All primary kinase reactions intended for use in sequential phosphorylation reactions were reacted with unlabeled ATP. Before PHF kinase sequential phosphorylation reactions, either kinases used for primary phosphorylations were inactivated by heating to 95°C for 10 min, or the phosphorylated tau was repurified on an Ni2+-conjugated agarose column. Inactivation of kinase activities used for primary phosphorylations was verified by the absence of labeled PHF proteins analyzed by autoradiography after addition of [γ-32P]ATP and MgCl2, incubation, and protein separation by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. All other reactions were performed with [γ-32P]ATP in preparation for autoradiogram and scintillation detection.
The entire sample of recombinant phosphorylated tau fragments from each reaction (0.4 μg) was solubilized by heating in sample buffer [40 mM Tris (pH 6.8), 1% sodium dodecyl sulfate, and 50% glycerol] at 95°C for 5 min. Prepared samples were separated by polyacrylamide gel electrophoresis on 10% acrylamide gels. Proteins were then transferred to nitrocellulose (pore size, 0.45 μm) in preparation for western blotting and autoradiogram protocols. Blots were exposed to film for 12 h, and the resulting autoradiograms were developed. Blots were incubated in TBS with 5% milk for 30 min to block nonspecific antibody binding. Blots were incubated with TG-5 (Jicha et al., 1997), a monoclonal antibody raised in our laboratory recognizing amino acids 220-240 of the primary sequence of htau40, or with 6HIS, a monoclonal antibody raised in our laboratory recognizing the N-terminal histidine tag present on all recombinant tau constructs used in this study, at a 1:10 dilution in TBS with milk for 16 h at 4°C. Blots were then washed three times each for 5 min with TBS and incubated with horseradish peroxidase-labeled anti-mouse IgG1 or IgG2B antibody for 2 h at 25°C. Blots were washed three times each for 5 min, and reactive proteins were visualized by reaction with 4-chloronaphthal in peroxide. After visualization of protein bands, individual lanes were isolated and counted for radioactivity by scintillation on a Beckman model LS6000 spectrophotometer to determine stoichiometries of phosphate incorporation in each sample.
In preparation for peptide mapping studies, phosphorylated TauP was incubated with AP-1 protease at a 1:100 enzyme to substrate ratio for 16 h at 37°C in TBS with addition of 2 M urea as described previously (Masaki et al., 1981 ; Hasegawa et al., 1992). Trypsin and elastase digestions were performed in 10 mM Tris-base (pH 8.5) with addition of CaCl2 to 5 mM at a 1:100 by weight enzyme to substrate ratio for 16 h at 37°C.
Separations were achieved on an HP1090 chromatograph using Vydac C8 2.1-mm columns. All detection was at 214 nm. HPLC procedures consisted of a 20-min isocratic gradient of 100% solution A, followed by a 1% addition of solution B per minute for 60 min. Solution A contained 100% water with 0.1% trifluoroacetic acid, and solution B contained 100% acetonitrile with 0.08% trifluoroacetic acid.
Amino terminal sequence analysis of the digested peptide fragments was performed by automated Edman degradation with a model 477A sequencer (Applied Biosystems) using the manufacturer's standard programming and chemicals.
Ten microliters of PKA-reacted TauW, PKA-reacted TauW + PHFs, or PKA-reacted TauW + PHFs + Mg2+ + ATP was applied to Formvar carbon-coated, glow-discharged grids. The grids were treated with TBS containing 1% bovine serum albumin and 5% normal goat serum (TBS+) for 10 min. 6HIS monoclonal antibody was diluted 1:50 in TBS+, and the grids were stained for 1 h at 25°C. After washing with TBS+, 10-nm-diameter gold particles conjugated to goat anti-mouse IgG2B (Goldmark Biologicals) were diluted 1:40 in TBS+, and the grids were stained for 1 h at 25°C. Grids were then washed and fixed for 5 min with 2% glutaraldehyde in TBS. After fixation, the grids were stained with 5% uranyl acetate in water for 30 min. A JEOL Jem 100CX scope was used to visualize PHF samples at 30K magnification.
PHF-associated kinase activity in degenerative disease states
To investigate the specificity of the PHF kinase for all AD cases, a series of immunoaffinity purifications were performed on both familial and sporadic AD, Pick's disease, and normal adult brain homogenates. After addition of [γ-32P]ATP, Mg2+, and casein, sodium dodecyl sulfate-polyacrylamide gel electrophoresis separation and autoradiographic analysis demonstrated that immunoaffinity column eluents from sporadic AD, chromosome 14 familial AD (presenilin 1), chromosome 21 familial AD (amyloid precursor protein), Down's syndrome, and Pick's disease brain homogenates all contained the PHF kinase activity (Fig. 2). Immunoaffinity column eluents from a normal adult control brain homogenate were devoid of protein kinase activity (Fig. 2).
Hierarchical tau phosphorylation
Because the major phosphorylated species identified in Fig. 2, other than the exogenously added casein, resembles the PHF-tau pattern demonstrated by many anti-tau antibodies in AD cases, the PHF kinase, PKA, protein kinase C, casein kinase II, MAP kinase, GSK-3, p34-cdc2, and cdk5 were all assayed for activity against unmodified recombinant TauW. All kinases previously implicated in the hyperphosphorylation of tau seen in AD, including PKA, protein kinase C, casein kinase II, MAP kinase, GSK-3, p34-cdc2, and cdk5, were able to phosphorylate recombinant TauW in vitro, but the PHF kinase was not (Fig. 3A), suggesting that it is not one of the kinases previously implicated as being responsible for the hyperphosphorylation of tau seen in AD.
Because previous studies have suggested that the hyperphosphorylation of tau seen in AD may be due to the sequential actions of multiple kinases acting in a synergistic fashion (Yang et al., 1993a,b ; Blanchard et al., 1994 ; Singh et al., 1995b), PHF kinase activity was assayed against recombinant tau previously phosphorylated in vitro by a panel of protein kinases. Prior phosphorylation by protein kinase C, casein kinase II, MAP kinase, GSK-3, p34-cdc2, and cdk5 did not allow the PHF kinase to phosphorylate TauW further. In contrast, PKA-phosphorylated TauW was shown to be a substrate for the PHF kinase (Fig. 3B).
Identification of PHF kinase phosphorylation sites on tau
To identify the region(s) of the tau molecule that are sequentially phosphorylated by the PHF kinase, TauNW (encoding for amino acids 1-342) and TauP (encoding for amino acids 242-440 but missing exon 10) were assayed for sequential phosphorylation by the PHF kinase after PKA pretreatment. TauNW and TauP were shown to be substrates for PKA phosphorylation (Fig. 4A), but only PKA-phosphorylated TauP was a substrate for the PHF kinase (Fig. 4B). These data show that the PHF kinase phosphorylation site(s) reside in the C-terminal 99 amino acids of tau, a region conserved in all six tau isoforms (Goedert et al., 1989).
To identify the exact PHF kinase phosphorylation sites and the PKA sites that direct these phosphorylations, PHF kinase sequentially phosphorylated TauP was digested with either AP-1 protease (cleaves C-terminal to lysine residues) (Fig. 5A) or AP-1 protease plus trypsin plus elastase and purified by HPLC separation on a C8 column. Peaks were collected and counted by scintillation for incorporation of the 32P label. The PHF kinase was shown to induce the phosphorylation of both Thr361 (a minor site) and additional site(s) in the C-terminal fragment consisting of residues 396-438 (Fig. 5).
Because sufficient quantities of PHF kinase sequentially phosphorylated fragments were not recovered from the HPLC system to allow exact detection of the PHF kinase phosphorylation site(s) in the C terminus of tau, additional site-directed mutagenic strategies were used. Site-directed mutagenesis of the previously identified PKA phosphorylation sites (Scott et al., 1993) S356A, S409G, and S416G demonstrates that the major sequential phosphorylation of tau by the PHF kinase is directed by the PKA-induced phosphorylation of Ser409 (Fig. 6A). An additional site-directed mutant, S412A + S413A + T414A (triple mutant in TauP), positively identified the PHF kinase phosphorylation sites as amino acids Ser412, Ser413, or Thr414 in htau40 (Fig. 6B). Further analysis of the PHF kinase using S412A and S413A mutants demonstrates that the primary PHF kinase phosphorylation site is Ser412 in the numbering system of htau40 (Fig. 6B).
Our calculation of PHF kinase-induced phosphate incorporation into PKA-pretreated tau is 0.18 mol of phosphate/mol of TauP (Table 1). Analysis of the S409G and S412A mutants demonstrated an overall reduction of 0.22 mol of phosphate/mol of tau for PKA phosphorylations (S409G mutant) and an overall reduction of 0.16 mol of phosphate/mol of tau for PHF kinase sequential phosphorylations (S409G and S412A mutants ; see Table 1), suggesting an overall stoichiometry for PHF kinase sequential phosphorylation at Ser412 of 0.73 mol of phosphate/mol of Ser409-phosphorylated TauP (using the equation [TauP — S412A]PHF kinase/[TauP — S409G]PKA). Analysis of the S356A mutant demonstrated an overall reduction of 0.30 mol of phosphate/mol of tau for PKA phosphorylations and an overall reduction of 0.03 mol of phosphate/mol of tau for PHF kinase sequential phosphorylations (Table 1), suggesting an overall stoichiometry for PHF kinase sequential phosphorylation at Thr361 of 0.1 mol of phosphate/mol of Ser356-phosphorylated TauP (using the equation [TauP — S356A]PHF kinase/[TauP — S356A]PKA). This analysis demonstrates that the optimal consensus sequence for the PHF kinase is Ser(P)-X-X-Ser (amino acids 409-412 in tau) but that the PHF kinase also recognizes the consensus sequence Ser(P)-X-X-X-X-Thr (amino acids 361-369), albeit to a lesser extent.
Table 1. Stoichiometry of phosphate incorporation for PKA, control, and PHF kinase sequential phosphorylations on TauP and the site-directed mutants fromFig. 6Data are in mol of phosphate/mol of tau construct.
TauP, PKA-phosphorylated TauP, and PHF kinase sequentially phosphorylated TauP were all purified by HPLC separation on a C8 column (Fig. 7A-C). Sequential phosphorylations of TauP lead to a decreased recovery of protein from the HPLC column (Fig. 7D), suggesting either (a) that the addition of phosphates to the tau molecule results in an increased exposure of hydrophobic residues that bound irreversibly to the hydrophobic matrix of the column or (b) that sequential phosphorylation leads to the formation of high-molecular-weight aggregates that are not recoverable from the HPLC system. To investigate the possibility that the PKA and PHF kinase sequential phosphorylations of tau facilitate the aggregation of recombinant tau, immunoelectron microscopy using the 6HIS antibody was performed on PKA-reacted TauW, PKA-reacted TauW + PHF, and PKA-reacted TauW + PHF + Mg2+ + ATP. Figure 8 demonstrates that PKA-reacted TauW was ineffective at inducing filament formation. Addition of PHF in the absence of Mg2+ and ATP allowed for a slight incorporation of PKA-reacted TauW into existing PHFs in what may be a nucleation or seeding event (Fig. 8). In contrast, PKA-reacted TauW + PHF + Mg2+ + ATP resulted in an enhanced aggregation of histidine-tagged recombinanut tau to existing PHFs, suggesting that enzymatic modifications of tau may accelerate the process of filament formation in AD.
The recent development of a method to immunoaffinity purify conformationally altered (Jicha et al., 1997) pathological tau molecules under relatively mild conditions (Vincent and Davies, 1992) has allowed the opportunity not only to study the pathological alterations in tau seen in AD, but also to study proteins that are associated with pathological tau. Initial analysis of the immunoaffinity column eluent demonstrated that a serine/threonine kinase was present in the collected sample (Vincent and Davies, 1992). The PHF kinase was able to induce the phosphorylation of the PHF constituents found in the eluent on addition of exogenous ATP and Mg2+ (Vincent and Davies, 1992).
All of the kinases previously postulated to be responsible for the pathological phosphorylations of tau seen in AD, including PKA, protein kinase C, casein kinase II, MAP kinase, GSK-3, p34-cdc2, and cdk5, are able to phosphorylate the recombinant tau fragments used in this study. However, the PHF kinase is unable to induce the phosphorylation of unmodified recombinant tau in vitro, despite the observation that PHF constituents are an excellent substrate for this kinase (Vincent and Davies, 1992). These observations suggest that certain phosphorylations of the tau molecule found in PHFs may be required to create or expose the PHF kinase substrate recognition sites in tau. This phenomenon has been termed hierarchical phosphorylation (Roach, 1991) and is known to play a key role in the substrate specificity of both GSK-3 and casein kinase I, two kinases that have been widely implicated as responsible for the hyperphosphorylation of tau seen in AD.
Initial characterization demonstrated that the PHF kinase shares many similarities with the casein kinase I family of protein kinases (Vincent and Davies, 1992). A more recent study using immunoaffinity PHF preparations similar to ours has suggested that the PHF kinase activity can be attributed to 36- and 39-kDa isoforms of casein kinase I as determined by in-gel renaturation assays, western analysis, inhibitor selectivity, and synthetic peptide substrate specificity (Kuret et al., 1997). Additional experiments using a polyclonal antibody to casein kinase Iα (Santa Cruz Biotechnology) demonstrate that the 39-kDa isoform is present in our PHF preparation (data not shown) and is likely to be responsible for the PHF kinase activity characterized in this study. In the previous study by Kuret et al. (1997) identification of phosphorylated residues on tau was not performed but was assumed to include Ser396 and Ser404, which have previously been demonstrated to be phosphate acceptors for casein kinase I (Singh et al., 1995a,b). Although the present study did not identify Ser396 and Ser404 as phosphate acceptors for the PHF kinase activity, it does provide further support for the conclusion that the kinase activity associated with PHF is a member of the casein kinase I family of kinases. Additional experiments (data not shown) demonstrate that exogenous casein kinase I, but not the PHF kinase, is capable of eliciting enhanced reactivity of the PHF-1 monoclonal antibody (recognizes phospho-Ser396 and phospho-Ser404) on sequentially phosphorylated recombinant tau and PHF-tau. This discrepancy suggests that the substrate specificity of casein kinase I may differ between members of this family of kinases or that the kinase itself may be altered in AD, perhaps as a result of its tight association with endogenous PHF constituents. Further studies attempting to elucidate the factors responsible for the unique substrate specificity of the PHF kinase are currently underway.
Although phosphorylations on Ser356 and Thr361 have not previously been identified on PHF-tau, phosphorylations in this region of the tau molecule composing the fourth microtubule binding repeat have been shown to influence the association between tau and microtubule arrays (Goedert and Jakes, 1990 ; Drechsel et al., 1992 ; Raffaelli et al., 1992 ; Biernat et al., 1993 ; for review, see Lee, 1993). It is possible that phosphorylation of these sites exists in PHF-tau and have simply been overlooked in past studies. Previous studies identifying phosphorylation sites on PHF-tau have relied almost exclusively on phosphopeptide mapping strategies, which can be inherently limited as demonstrated by the lack of recovery and inability to determine the exact PHF kinase sites in the extreme C terminus of tau seen in this study. Alternate strategies aimed at identifying phosphorylation sites in PHF-tau have relied on the creation and empirical reactivity of phosphorylation-dependent antibodies raised against PHF-tau. Although this approach has yielded important information regarding certain immunodominant phosphorylations found in PHF-tau, the usefulness of antibodies recognizing epitopes in the microtubule binding regions of tau has been limited by the extremely low levels of reactivity seen against PHF-tau (Dickson et al., 1992). This lack of reactivity is assumed to be due to the compact nature of PHF-tau, the core of which is thought to be composed of the microtubule binding repeats of tau (Novak et al., 1993), and the limited access of antibodies to their antigenic counterparts in these regions (Dickson et al., 1992). Future studies will attempt to determine if Ser356 and/or Thr361 are phosphorylated in PHF-tau, or if their identification in this study is merely an artifact of the in vitro conditions used.
Ser409 in tau has previously been shown to be phosphorylated in all PHF preparations analyzed to date (Hasegawa et al., 1992). In vitro analysis of recombinant tau reacted with various kinases has shown that only PKA (Scott et al., 1993) and a 35/41-kDa kinase (Biernat et al., 1993) are capable of inducing this phosphorylation. Subsequent studies on the 35/41-kDa kinase identified the 41-kDa fragment as the catalytic subunit of PKA (Drewes et al., 1995), suggesting that only PKA is capable of inducing the phosphorylation of Ser409 in tau. This observation may explain why PHF constituents are such excellent substrates for the PHF kinase. Tau in PHF is already phosphorylated at Ser409, bypassing the need for PKA pretreatment that is required for PHF kinase sequential phosphorylations of recombinant tau. These data suggest that the phosphorylation of tau on Ser409, although not necessarily pathological in itself, may be one of the earliest modifications of tau leading to PHF formation in AD.
It is possible that the phosphorylation of Ser409 does not contribute directly to the pathogenesis of PHF formation in AD, but rather represents a ubiquitous phosphorylation of the tau molecule found in all tau-expressing cells, and that only those cells destined for PHF assembly pathologically express the PHF kinase. Although plausible, this possibility is highly unlikely in that the PHF kinase is completely ineffective in inducing the phosphorylation of normal brain-derived tau (Vincent and Davies, 1992). It is also possible that PKA-induced phosphorylation of Ser409 directly contributes to the pathogenesis of PHF formation in AD as has been reported previously (Pierre and Nunez, 1983 ; Steiner et al., 1990 ; Litersky and Johnson, 1992 ; Scott et al., 1993 ; Singh et al., 1994).
Phosphorylations of tau in the C-terminal tail by both PKA on Ser409 and Ser416 and Ca2+/calmodulin-dependent protein kinase on Ser416 impose conformational restraints on the tau molecule that significantly retards migration by sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis (Steiner et al., 1990 ; Scott et al., 1993). Although Scott et al. (1993) assumed that the PKA-induced phosphorylation of Ser416 was responsible for the retarded migration of tau seen by sodium dodecyl sulfate-polyacrylamide gel electrophoresis analyses, our S409G and S416G site-directed mutants clearly demonstrate that this is not the case. PKA phosphorylation of the S416G mutant demonstrates the typical shift in electrophoretic mobility seen in nonmutated tau, but this shift is completely blocked in the S409G tau mutant. Also, an S→E409 tau mutant constitutively exhibits this apparent molecular weight shift regardless of phosphorylation state (data not shown). This same aberrant migration pattern is seen in all preparations of PHF-derived tau (Flament et al., 1989 ; Greenberg and Davies, 1990 ; Lee et al., 1991 ; Goedert et al., 1992), suggesting that C-terminal conformational restraints may be required for the pathogenesis of PHF assembly in AD.
The phosphorylation of Ser412 by the PHF kinase may play an important role in the pathogenesis of PHF formation in AD. To date, the phosphorylation of Ser412 has only been identified in PHF-tau, but not in normal adult or fetal brain-derived tau, suggesting that this phosphorylation is entirely specific for PHF-tau. Also, no known kinases have been shown to be able to induce the phosphorylation of Ser412 in vitro or in vivo. Most of the known phosphorylation sites in tau can be modified by multiple protein kinases, complicating the issue of which kinase(s) are responsible for these phosphorylations in vivo, but the PHF kinase is clearly responsible for the phosphorylation of Ser412 seen in PHF-tau.
Previous studies of tau phosphorylations in the amino acid 412 region have focused on the GSK-3-induced phosphorylation of Ser413 (Ishiguro et al., 1992, 1995). Although these studies have provided valuable information, the phsophorylation of Ser413 has also been shown to exist in normal adult and fetal tau, raising questions about the role this phosphorylation plays in the pathogenesis of PHF formation in AD (Ishiguro et al., 1992, 1995). It is interesting that the phosphorylation of Ser413 in these studies resulted in a lack of recovery from HPLC, precluding the use of phosphopeptide mapping strategies in its identification (Ishiguro et al., 1992, 1995). This same effect is seen in the present study where phosphopeptide mapping strategies were ineffective at elucidating the most C-terminal PHF kinase phosphorylation sites in recombinant tau. These observations suggest that phosphorylations of the tau molecule in this region may promote the formation of high-molecular-weight aggregates responsible for the lack of recovery from the HPLC. Also, attempts at mass determinations of PHF kinase sequentially phosphorylated C-terminal recombinant tau fragments identified only high-molecular-weight hypermasses that were not interpretable (data not shown).
Additionally, phosphorylations by both PKA and the PHF kinase resulted in a substantial loss of recovery from the HPLC column for both the monomeric and dimeric forms of TauP. Previous analyses of PHF-tau from AD brain demonstrated this same low recovery of C-terminal fragments of tau unless the PHF-tau was first subjected to alkaline phosphatase treatment (Hasegawa et al., 1992), suggesting that the PKA- and PHF kinase-induced phosphorylations play an important role in the transition of normal tau to the pathologically altered form of PHF-tau seen in AD. Immunoelectron microscopy further demonstrates the effects these unique phosphorylations have on the in vitro aggregation of recombinant tau molecules. Recent analyses of in vitro PHF formation have suggested that phosphorylation of tau is not a necessary prerequisite for PHF formation (Goedert et al., 1996), but the present data demonstrate that certain unique phosphorylations of the tau molecule can drastically alter the equilibrium between soluble and aggregated forms of tau. To our knowledge, this is the first demonstration that the hyperphosphorylation of tau seen in AD may play a functional role in the pathogenesis of PHF formation in AD. A panel of monoclonal antibodies is currently being raised against both the PKA and PHF kinase phosphorylations identified in this study with the hopes that they will provide further insights into the roles these unique phosphorylations play in the pathogenesis of neurofibrillary pathology in AD.
Although it is clear from recent genetic findings that the etiological factors responsible for AD can be quite varied, the eventual development of neurofibrillary pathology in all AD cases suggests that it may be a crucial point of convergence of multiple degenerative cascades that provides an ideal target for potential diagnostics and therapeutic interventions in AD. Analysis of the immunoaffinity column eluent from various forms of AD, other degenerative diseases, and normal brain homogenates demonstrates that the PHF kinase is entirely specific for all disease states associated with neurofibrillary degeneration. Because hyperphosphorylation of the tau molecule has been shown to dissociate tau from microtubules (Goedert and Jakes, 1990 ; Drechsel et al., 1992 ; Raffaelli et al., 1992 ; Biernat et al., 1993 ; for review, see Lee, 1993), leading to disruption of the neuronal cytoskeleton, interference with cellular transport mechanisms, and a reduction in its rate of degradation (Vincent et al., 1994), and to facilitate aggregation, the PHF kinase may play an important role in the development of neurofibrillary pathology and the resultant neuronal dysfunction and loss seen in all cases of AD.