Proline-directed Phosphorylation of Human Tau Protein*

The primary sequence of the microtubule-associated protein tau contains multiple repeats of the sequence -X-SerJThr-Pro-X-, the consensus sequence for the proline-directed protein kinase When phosphorylated by proline-directed protein ki- nase in vitro, tau was found to incorporate up to 4.4 mol of phosphate/mol of protein. Isoelectric focusing of the tryptic phosphopeptides demonstrated the presence of five distinct peptides with PI values of approximately 6.9, 6.5, 5.6-5.9,4.7, and 3.6. Mapping of the tryptic phosphopeptides by high performance liquid chromatography techniques demonstrated three dis- tinct peaks. Data from gas phase sequencing, amino acid analysis, and phosphoamino acid analysis suggest that proline-directed protein kinase phosphorylates tau at four sites. Each site demonstrates the presence of a proline residue on the carboxyl-terminal side of the phosphorylated residue. Two phosphorylation sites are located adjacent to the three-repeat microtubule- binding domain that has been found to be required

The primary sequence of the microtubule-associated protein tau contains multiple repeats of the sequence -X-SerJThr-Pro-X-, the consensus sequence for the proline-directed protein kinase (p34Cdc2/p58cYc'i" A ).
When phosphorylated by proline-directed protein kinase in vitro, tau was found to incorporate up to 4.4 mol of phosphate/mol of protein. Isoelectric focusing of the tryptic phosphopeptides demonstrated the presence of five distinct peptides with PI values of approximately 6.9, 6.5, 5.6-5.9,4.7, and 3.6. Mapping of the tryptic phosphopeptides by high performance liquid chromatography techniques demonstrated three distinct peaks. Data from gas phase sequencing, amino acid analysis, and phosphoamino acid analysis suggest that proline-directed protein kinase phosphorylates tau at four sites. Each site demonstrates the presence of a proline residue on the carboxyl-terminal side of the phosphorylated residue. Two phosphorylation sites are located adjacent to the three-repeat microtubulebinding domain that has been found to be required for the in vivo co-localization of tau protein to microtubules. Two other putative phosphorylation sites are located within the identified epitope of the monoclonal antibody Tau-1. Phosphorylation of these sites altered the immunoreactivity of tau to Tau-1 antibody. Since the neuronal microtubule-associated protein tau is multiply phosphorylated in Alzheimer's disease, and Tau-1 immunoreactivity is similarly reduced in neurofibrillary tangles and enhanced after dephosphorylation, phosphorylation at one or more of these sites may correlate with abnormally phosphorylated sites in tau protein in Alzheimer's disease.
Microtubules are essential for the structure and function of the neuronal cell (1,2). Among the factors that distinguish neuronal microtubules from those in non-neuronal tissues is their association with a collection of neuronal-specific microtubule-associated proteins (3). Within this collection is tau protein, which has been localized primarily to the axon in situ (4)(5)(6)(7) and is required for the development and maintenance of the axon during neuronal development in culture (8,9). Tau protein has been found in the neurofibrillary tangles, * This research was supported by National Institutes of Health Grants 1R01-NS28765-01 (to P. R. V.) and GM39300 (to G. L.) and National Science Foundation Grant BNS-8716846 (to P. R. V.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
As isolated from brain microtubules, the tau protein family migrates with an apparent molecular mass range of 55-62 kDa on SDS-polyacrylamide gel electrophoresis (12); heterogeneity is generated in part by alternative mRNA splicing mechanisms (13). Additional heterogeneity results from posttranslational modification (phosphorylation) of tau (12,14). Four different protein kinases, including cyclic AMP-dependent protein kinase ( E ) , calcium-and phospholipid-dependent protein kinase (16), calcium-and calmodulin-dependent protein kinase I1 (17), and casein kinase (15) are known to phosphorylate tau protein. The site of phosphorylation by calmodulin-dependent protein kinase has been identified as serine 327 (15) (all tau residue numbers given are in reference to the 352-amino acid isoform; Ref. 18); phosphorylation at this site was shown to cause a shift in tau's mobility on SDSpolyacrylamide gels. The apparent molecular mass increased by -4 kDa. An even greater phosphate-dependent increase in molecular size is observed in abnormal tau protein isolated from brains of AD patients. This abnormal tau protein (termed "A68") comigrates with normal tau after enzymatic and chemical dephosphorylation (19)(20)(21). A68 is thought to be a "hyperphosphorylated" form of tau (phosphorylated to a stoichiometry of greater than 1). The kinase(s) responsible for this hyperphosphorylation, the phosphorylation sites which constitute the hyperphosphorylated state, and the role of this hyperphosphorylation in the formation of neurofibrillary tangles are unknown. Evidence exists that a serine residue equivalent to Ser-307 is abnormally phosphorylated in AD (19,20). However, it is postulated that other sites remain to be identified. The primary amino acid sequence of tau contains 14 potential sites for the proline-directed protein kinase (PDPK), a recently discovered growth factor-activated protein kinase identified to be a heterodimer of p58'y''i" "/ ~34'~"'. This protein kinase recognizes the minimal consensus sequence of -X-(Ser/Thr)-Pro-X- (22)(23)(24). We report here that PDPK hyperphosphorylates tau and identify the phosphorylated sites. The relevance of this phosphorylation to AD is discussed. Purification of Proline-directed Protein Kinase-The proline-directed kinase was purified from mouse FM3A cells by a slight modification of the method of Hall et al. (24). Modifications include: isolation from FM3A cells maintained in spinner flasks in logarithmic growth in HEPES-buffered RPMI-1640 medium supplemented with 10% calf serum and chromatography on Q-Sepharose FF (Fast Q ) , instead of DEAE-cellulose. The column was washed with equilibration buffer containing 10 p~ cyclic AMP and then equilibration buffer containing 150 mM NaCl. PDPK activity was eluted from Fast Q with equilibration buffer containing 350 mM NaC1. Throughout the various purification steps, the kinase activity was assayed using a highly selective synthetic peptide substrate derived from tyrosine hydroxylase (TH2-16) as described by Vulliet et al. (22). PDPK prepared by this method had a specific activity of 5-20 unitslpg protein (1 unit of activity is defined as the amount of kinase necessary to transfer 1 pmol of phosphate/min from ATP to TH2-lfi peptide when assayed under standard assay conditions: 100 mM Tris-acetate, pH 7.6, 100 p M peptide substrate, 100 p M ATP, and 10 mM magnesium acetate).

Proline-directed Phosphorylation
Phosphorylation of Tau with PDPK and Phosphopeptide Mapping-Typically 2.5 pg of purified human recombinant tau protein was phosphorylated by PDPK under standard assay conditions using 3-5 units of PDPK and incubation for 60 min at 30 "C in a final volume of 50 pl. The reaction was stopped by addition of 5.5 pl of 70% perchloric acid. Following centrifugation for 5 min at 13,000 X g, the pellets were washed three times with 1 ml of 25% trichloroacetic acid, twice with 1 ml of cold acetone, and the incorporated phosphate estimated by Cerenkov counting. The acid precipitate was then resuspended in 25 p1 of SDS sample buffer and separated on 10% SDSpolyacrylamide gels by the method of Laemmli (26).
For phosphopeptide mapping, the pellets were resuspended in 100 pl of 0.1 M N-morpholine acetate, pH 8.3, and trypsin added a t a ratio of 1:lO trypsin:protein (w/w). Following digestion at 37 "C overnight, the samples were analyzed on a Gilson gradient high performance liquid chromatograph equipped with a 0.46 X 25-cm Vydac CIR reversed-phase column equilibrated in 0.1% trifluoroacetic acid and developed with a water/acetonitrile gradient (from 0 to 50% acetonitrile also containing 0.1% trifluoroacetic acid) a t a flow rate of 1.0 ml/min. One ml fractions containing the phosphopeptides were quantified by Cerenkov counting and concentrated by evaporative centrifugation (Savant SpeedVac).
The peptides were resuspended in 10-20 pl of water and analyzed on isoelectric focusing gels (Serva Precote gels, pH 3-10) by the manufacturer's directions. The radioactive peptides were identified in the dried gels by autoradiography. For purification of the phosphopeptides, bands containing significant amounts of radioactivity were excised from the gel and the phosphopeptides were extracted from the gel into 0.1% trifluoroacetic acid by gentle rocking at 25 "C for 2 h. Phosphoamino acid analysis was performed by resuspending the dry pellets in 200 pl of 5.7 M HCI followed by incubation in a sealed container for 2 h at 110 "C (27). The HC1 was removed by evaporative centrifugation, and the amino acids were resuspended in 10 p1 of pH 1.9 electrophoresis buffer (88% formic acidglacial acetic acid:water, 50:156:1794) containing authentic phosphoamino acid standards. Samples containing a similar amount of "P counts were spotted onto a thin-layer cellulose plate (Kodak 13255). These samples were subjected to horizontal electrophoresis on an LKB Multiphor I1 a t 1000 V for 1 h at 10 "C. The plate was sprayed with 0.2% ninhydrin in acetone and developed at 65 "C to visualize the phosphoamino acid standards. The radioactive areas of the cellulose plate were determined by autoradiography. Gas-phase sequencing of HPLC-purified phosphopeptides was performed on an Applied Biosystems model 470A sequenator using the manufacturer's recommended methods. The sequence of each peptide was verified by amino acid content analysis.
Western Blotting of Phosphorylated and Non-phosphorylated Tau-Protein samples were prepared and analyzed by SDS-polyacrylamide gel electrophoresis as described above. Following the electrophoresis, the proteins were transferred to nitrocellulose sheets (Schleicher & Schuell BA85, 0.45 pm) by the method of Towbin et al. (28). The blots were stained with Ponceau S red and the gels were stained with Coomassie Blue R to confirm the complete transfer of protein from the polyacrylamide gels. Following removal of the Ponceau S red dye, the blot was probed with Tau-I and developed using the alkaline phosphatase/bromochloroindoyl phosphate/nitro blue tetrazolium system as described in Ref. 29.

RESULTS AND DISCUSSION
Purified recombinant human tau was incubated in the presence of PDPK (22-24) and then analyzed on SDS-polyacrylamide gel (Fig. lA). Use of recombinant tau in these studies circumvents the intrinsic heterogeneity of brain protein and also provides a non-phosphorylated substrate for this protein kinase. The stained gel (Fig. lA) shows that tau incubated in the absence of PDPK (Fig. lA, lune 1 ) has a slightly greater mobility than tau incubated in the presence of the protein kinase (Fig. lA, lane 2). This result is consistent with previous reports demonstrating a decreased mobility of phosphorylated forms of tau (15, 19-21). The autoradiogram of the stained gel ( Fig. 1B) confirms that 32P is incorporated into tau only in the presence of the protein kinase. PDPK was found to transfer 4.4 mol of 32P-labeled phosphate/mol of tau protein under conditions in which this kinase would transfer 0.9 mol of phosphate/mol into synapsin (24).
The ability of PDPK to phosphorylate tau to a stoichiometry greater than one suggested that this protein might contain more than one proline-directed phosphorylation site. To examine this possibility, tau was incubated in the presence of PDPK and [-y-"P]ATP and then digested with trypsin. The analysis of the tryptic phosphopeptides by isoelectric focusing (IEF) revealed five distinct radioactive bands (labeled A-E) with respective isoelectric points of 6.9, 6.5, 5.6-5.9, 4.7, and 3.6 ( Fig. 2A). The percentage of the total incorporated radioactivity in each phosphopeptide was as follows: A, 40%; B, 26%; C, 7%; D, 22%; E, 5%. HPLC analysis of a similar tryptic digest showed the presence of only three major peaks of radioactivity (Fig. 2B). When each IEF band was isolated and analyzed by the HPLC, the chromatogram revealed only a single peak of radioactivity, suggesting that each IEF band contained a single phosphopeptide (data not shown). The retention times of the individual IEF phosphopeptides are indicated in the HPLC chromatogram with the labeled arrows (Fig. 2B). Tryptic phosphopeptide pair A and C and pair D and E do not resolve under these chromatographic conditions. The peptides were also subjected to phosphoamino acid analysis (27). Peptide B demonstrated the presence of phosphoserine, while peptides A, D, and E contained both phosphoserine and phosphothreonine (Fig. 2C).
Since the amino acid sequence of human tau is known (18,30), Edman sequencing of these phosphopeptides was employed to locate the phosphorylation sites within the tau protein. Table I lists the results from gas phase sequence analysis of purified peptides A, B, and D. A single phenylthiohydantoin-derivative (PTH-derivative) at each cycle during sequencing of the peptide confirmed that only one peptide species was present. Although peptide D could not be sequenced to completion, the remaining amino acids can be deduced from the amino acid composition determined for each of the peptides (data not shown). Definitive sequence data could not be obtained from peptide C or peptide E. Tau was phosphorylated to 4.4 mol of phosphate/mol of tau as described in Fig. 1. The tryptic phosphopeptides were purified by a combination of isoelectric focusing and HPLC techniques. The fractions containing the tryptic phosphopeptides were extracted from IEF gel and purified by HPLC chromatography as described in Fig.  2. The fractions containing radioactive counts were concentrated by evaporative centrifugation and subjected to automated gas phase sequencing on an Applied Biosystems model 470A sequenator as per manufacturer's directions. The amount of each amino acid residue is expressed in pmol. Parentheses denote amino acids whose presence was established by amino acid analysis and whose position is assumed from tau's deduced cDNA sequence. Comparisons of the peptide sequences to the human tau sequence as predicted from cDNA showed that peptide A corresponded to residues 167-182 and peptide B corresponded to residues 173-182 (18,30). Since phosphoamino acid analysis demonstrated that peptide A contained both phosphothreonine and phosphoserine (Fig. 2C), threonine 173 must be phosphorylated as inspection of the sequence reveals only 1 threonine in this peptide. Peptide B, which lacks a phosphothreonine on phosphoamino acid analysis and exhibits a PTH-Thr equivalent to threonine 173 on gas phase sequencing, contains threonine 173 in the non-phosphorylated state. The phosphate present on Thr-173 in peptide A would be expected to render tau protein resistant to tryptic cleavage at Arg-172 (31). Consequently, peptide B most likely results from incomplete phosphorylation of tau at Thr-173 allowing trypsinolysis at Arg-172 to yield a peptide containing residues Protein sequencing by Edman degradation depends on the quantitative recovery of the PTH-derivative at each cycle. However, when a phosphorylated serine or threonine residue is encountered during Edman sequencing, little or no PTHderivative is recovered at that cycle. The low recovery of an expected serine or threonine derivative at a particular cycle followed by adequate recoveries of the next amino acid in subsequent cycles of degradation is characteristic of a phosphorylated residue. On this basis, the data in Table I reporting the recovery of the amino acid derivatives independently suggest that Thr-173 in peptide A is phosphorylated. Similarly, the sequencing data for peptide B suggest that of the 3 serines in this peptide, only Ser-177 is phosphorylated.

Residue Amount Residue Amount Residue Amount
The protein sequencing and amino acid composition data indicate that Peptide D corresponds to tau residues 137-151 (18,30). As it contained both phosphoserine and phosphothreonine, Thr-147 can be identified as a phosphorylation site since it is the only threonine in the peptide. The identity of the phosphorylated serine in peptide D is more difficult from the sequencing results alone due to the low yields. However, when the data from gas phase sequencing, amino acid analysis, and phosphoamino acid analysis are taken together, the identity of the serine residue can be deduced. Since PDPK requires the -X-(Ser/Thr)-Pro-X-motif (22)(23)(24) and the three phosphorylated residues identified thus far all possess this motif, the modified serine residue in peptide D most likely would have the same motif. Only 2 of the 5 serines in peptide D fit this criteria, Ser-141 and Ser-144. From Table I, the recovery of PTH-derivatives from cycles corresponding to the serines at positions 137, 140, and 141 suggest that these serines are not phosphorylated. Given that no other protein kinase activities have been detected in this preparation of PDPK and that no other peptides co-purify with peptide D, it can be inferred that Ser-144 is the phosphorylated serine residue since no PTH-AA is detected at the cycle corresponding to Ser-144 and its location is in the requisite motif for prolinedirected phosphorylation. Although the data supporting this conclusion is indirect, serine 144 is the most probable serine residue phosphorylated by PDPK in peptide D.
The phosphopeptides sequenced account for 87% of the total radioactivity recovered from the isoelectric focusing step, indicating that the major PDPK phosphorylation sites were identified. The remaining sites are not likely to have been phosphorylated with high efficiency. Interestingly, not all of the fourteen potential proline-directed phosphorylation sites (i.e. the -X-(S/T)-P-X-motifs) in tau are phosphorylated by PDPK, indicating that this motif is a necessary but not sufficient "discriminator" for this protein kinase. Comparison of the adjacent residues on both sides of all actual and potential PDPK phosphorylation sites does not indicate any pattern in the primary structure other than the known requirement of a carboxyl-side proline (22)(23)(24); the answer awaits elucidation of tau's tertiary structure to determine accessibility of potential phosphorylation sites to the protein kinase.
Two of the four identified phosphorylation sites (Ser-144 and Thr-147) are located within the region of tau identified as the epitope for the monoclonal antibody Tau-1 (residues 131-149) (32). Fig. 3 shows an immunoblot with equal amounts of PDPK-phosphorylated and non-phosphorylated tau protein (lanes 1 and 2, respectively) probed with Tau-1 antibody. There is a distinct quantitative difference in the immunoreactivity. Tau-1 demonstrates reduced affinity for PDPK phosphorylated tau, as identified by its characteristic retarded gel mobility. If the reaction resulted from trace amounts of non-phosphorylated tau in the PDPK reaction, the immunoreactive protein would have migrated as the dephosphorylated form of tau. The phosphorylation of serine residue equivalent to serines 141 and 144 was recently reported to alter the epitope of Tau-1 antibody and the AT8 antibody (33).
The identification of Ser-144 and Thr-147 as residue(s) that affect Tau-1 reactivity is relevant to the tau protein in neurofibrillary tangles of AD. The first indication that tau in AD was abnormally phosphorylated followed from the observation that phosphatase treatment of AD brain sections was necessary for Tau-1 reactivity (34, 35). A68 also requires phosphatase treatment for Tau-1 reactivity (19)(20)(21). The classification of these phosphorylated sites as abnormal would be premature, since phosphatase-dependent Tau-1 reactivity has also been described in normal adult rat brain tissue (36). However, an increase in phosphatase-dependent Tau-1 reactivity has been reported in AD brain extracts, which may reflect axonal disruption and the dislocation of tau from the axon into the somatodendritic region (37). Therefore, it can  (26), and transferred to nitrocellulose sheets (28). The nitrocellulose blot was probed with Tau-1 and developed using the alkaline phosphatasefiromochloroindoyl phosphate/nitro blue tetrazolium system as described in Ref. 29. A similar blot probed with SMI-51, a control tau monoclonal antibody, failed to detect any difference between reactivity in the two lanes (data not shown). be argued that in AD, Ser-144 and Thr-147 are abnormally phosphorylated in the sense that they are phosphorylated to a higher stoichiometry than normal.
Proline-directed protein kinase is known to phosphorylate other neuronal structural proteins including synapsin I (24), neurofilament (38), and MAP-2 (39). The phosphorylation sites Thr-147, Thr-173, and Ser-177 are conserved relative to the microtubule-binding domains in MAP-2; Ser-144 corresponds to a conservatively substituted threonine in MAP-2 (40). These sites lie in two small regions of homology between MAP-2 and tau, surrounded by stretches of non-homology (Fig. 4). While none of the sites are in the repeat domain of tau which constitutes the microtubule binding sites, Thr-173 and Ser-177 lie in an area that enhances the binding of tau to microtubules in vitro and in uiuo (41): Although phosphorylation has been reported to affect tau and MAP-2 activity (14,43), functional effects mediated by PDPK phosphoryl-ation remain to be shown. None of the PDPK phosphorylation sites in tau and MAP-2 are conserved in the non-neuronal MAP-4 sequence, although this protein contains a microtubule-binding repeat domain homologous to that of tau and It is very likely that other proline-directed protein kinases, in addition to PDPK, will be able to hyperphosphorylate tau. The relationship of PDPK to two tau protein kinases recently isolated from bovine brain is unknown (46)(47)(48). Tau protein kinase 11, which appears to require an adjacent proline on the carboxyl terminus of the phosphorylated residue was identified to phosphorylate Ser-144, Thr-147, Ser-177, and Ser-315 in bovine brain tau (47). The similarity of the phosphorylation sites in tau modified by both of these protein kinases suggest that they are closely related. Tau protein kinase I1 was noted t o be activated by the presence of tubulin in the assay (46,48), whereas PDPK is activated by treatment of PC12 cells with nerve growth factor (22) and A 431 cells with epidermal growth factor (23). The catalytic subunit of tau protein kinase I1 was identified to have a molecular mass of approximately 30 kDa (48). Since the catalytic subunit of PDPK is the ~34'~'' kinase, it appears that these kinases may share structural properties as well as substrate selectivity. Both of these tau kinases readily phosphorylate histone H1, MAP-2, and tau protein and do not phosphorylate casein (23,48).
Members of the MAP-2 kinase/ERK family, which are also known to phosphorylate target residues with carboxyl-side prolines (49), are expressed in high levels in normal neuronal tissues and in transformed cell lines derived from neuronal cells (50). Recent experiments have demonstrated that an ERK-like protein kinase activity in bovine adrenal chromaffin cells (51) will also phosphorylate tau p r~t e i n .~ It will be important to establish the utilization of the proline-directed phosphorylation sites in uiuo and the circumstances under which this phosphorylation/dephosphorylation occurs. Recently, mass fragmentation of abnormal tau protein peptides isolated from neurofibrillary tangles has revealed that Ser-173 and Thr-177 are both phosphorylated further establishing a role of proline-directed phosphorylation in the formation of neurofibrillary tangles (52). Since it is known t h a t ~3 4 '~' ' kinase is normally found in proliferating cells and is down-regulated upon neuronal differentiation (42, 53), ~3 4 '~"~ would not normally be expected to phosphorylate tau in adult brain. It may be that the pathogenesis of neurodegenerative diseases involves an induction or activation of ~3 4 "~" kinase or some other proline-directed protein kinase. The identification of proline-directed phosphorylation sites and the specific phosphate-dependent epitope for Tau-1 will further investigations of the biochemical mechanisms controlling these disease processes.