Novel Phosphorylation Sites in Tau from Alzheimer Brain Support a Role for Casein Kinase 1 in Disease Pathogenesis*

Tau in Alzheimer disease brain is highly phosphorylated and aggregated into paired helical filaments comprising characteristic neurofibrillary tangles. Here we have analyzed insoluble Tau (PHF-tau) extracted from Alzheimer brain by mass spectrometry and identified 11 novel phosphorylation sites, 10 of which were assigned unambiguously to specific amino acid residues. This brings the number of directly identified sites in PHF-tau to 39, with an additional six sites indicated by reactivity with phosphospecific antibodies to Tau. We also identified five new phosphorylation sites in soluble Tau from control adult human brain, bringing the total number of reported sites to nine. To assess which kinases might be responsible for Tau phosphorylation, we used mass spectrometry to determine which sites were phosphorylated in vitro by several kinases. Casein kinase 1δ and glycogen synthase kinase-3β were each found to phosphorylate numerous sites, and each kinase phosphorylated at least 15 sites that are also phosphorylated in PHF-tau from Alzheimer brain. A combination of casein kinase 1δ and glycogen synthase kinase-3β activities could account for over three-quarters of the serine/threonine phosphorylation sites identified in PHF-tau, indicating that casein kinase 1δ may have a role, together with glycogen synthase kinase-3β, in the pathogenesis of Alzheimer disease.

In Alzheimer disease one of the primary neuropathological characteristics is the presence of intraneuronal neurofibrillary tangles assembled from paired helical filaments (PHF) 4 comprised of Tau protein in a hyperphosphorylated state (PHF-tau) (1)(2)(3). All six alternately spliced human brain Tau isoforms are present in tangles, and their increased phosphorylation may underlie the transformation of soluble Tau into PHF and tangle pathology in the tauopathies (2, 4 -7).
Because the development of Tau pathology is related to its phosphorylation state, it is important to determine the phosphorylation sites on PHF-tau that lead to the abnormal characteristics of insolubility and aggregation that define Tau in Alzheimer brain. To date, there have been only a few reports using direct sequencing methods such as Edman degradation or mass spectrometry (MS) to identify directly the phosphorylation sites on Tau in Alzheimer disease and Tau from control human brain (8 -11). Additional phosphorylation sites have been mapped using phosphorylation-dependent antibodies to Tau to label tangles and PHF-tau preparations from Alzheimer brain (12)(13)(14)(15)(16). In recent years, improvements in MS methods have resulted in increased sensitivity and specificity of detection of post-translational modifications on proteins. In particular, highly resolving high pressure liquid chromatography has been used to simplify the complex peptide mixtures present in protein digests, and this, combined with tandem MS/MS (LC-MS/MS), has resulted in greatly enhanced protein sequence coverage (17,18). We report here our exploitation of this powerful technology to determine close to the full complement of phosphorylation sites on PHF-tau.
Several proline-directed and non-proline-directed protein kinases have been suggested to have a role in the generation of PHF-tau in Alzheimer brain (19), including casein kinase 1 (CK1), but this enzyme has been relatively little studied as a Tau kinase. The role of CK1␦ as a potential Tau kinase is of particular interest because it has been reported that CK1␦ protein is increased more than 30-fold in the hippocampus of Alzheimer brain compared with equivalent controls (20), although its mRNA content is increased 24-fold (21) and CK1 has also been shown to be tightly associated with PHF (22). Recently, CK1␦ has been reported to phosphorylate Tau at two epitopes detected using phosphospecific monoclonal antibodies to Tau, and exogenous expression of CK1␦ in non-neuronal cells reduces binding of Tau to microtubules (23). Of note in the context of Alzheimer disease is a report that CK1 activity is stimulated by amyloid ␤-peptide (A␤), a component of the senile neuritic plaques that together with tangles characterize Alzheimer brain (24). Further evidence for the possible involvement of CK1 in Alzheimer disease comes from the reported influence of CK1 in the regulation of A␤ production in neurons (25).
In this study we present our findings of 11 previously unidentified phosphorylation sites in PHF-tau, all on serine and thre-onine residues. Ten of these 11 sites could be assigned unambiguously to particular hydroxyamino acids, the remaining one being on either of two closely spaced residues. Because only one of these new phosphorylation sites is proline-directed, we have investigated potential candidate Ser/Thr Tau kinases that are not dependent on Ser-Pro or Thr-Pro motifs. We report here new phosphorylation sites on Tau generated by the activities of casein kinase 1␦ (CK1␦), casein kinase 2 (CK2), cyclic AMP-dependent protein kinase (PKA), and glycogen synthase kinase-3 (GSK-3), individually and in combination. We have found that at least six of the new phosphorylation sites as well as many of the previously identified sites in PHF-tau can be generated by CK1␦. The finding of a significant number of phosphorylation sites in PHF-tau for which CK1 is a strong candidate kinase, including three for which it is the only known kinase, implies that CK1 may make an important contribution to the pathogenesis of Alzheimer disease.

EXPERIMENTAL PROCEDURES
Materials-Frozen post-mortem human brain (frontal and temporal cortex) from clinically and pathologically confirmed Alzheimer disease and control human brain was obtained from the MRC Neurodegenerative Diseases Brain Bank at the Institute of Psychiatry, King's College London, UK. Sequencinggrade trypsin and Asp-N were obtained from Roche Applied Science. The following purified recombinant protein kinases expressed in bacteria were obtained from New England Biolabs: CK1␦⌬317 (truncated at residue 317, which exhibits increased enzyme activity relative to wild-type CK1␦ (26)); CK2 from a clone expressing both ␣ and ␤ subunits; PKA catalytic subunit, ␣ isoform; and GSK-3␤, intact and unmodified. Purified recombinant cyclin-dependent kinase-5/p25, expressed in Sf21 insect cells, was obtained from Upstate.
Purification of PHF-tau from Alzheimer Brain-PHF-tau was prepared from human post-mortem Alzheimer brain, including phosphatase inhibitors, as described previously (10). Samples were analyzed on Western blots probed with antibody to Tau as described (10), except that the secondary antibody was coupled to horseradish peroxidase, and blots were developed using enhanced chemiluminescence (Amersham Biosciences).
Preparation of Tau from Human Control Brain-A heat-stable preparation of soluble Tau from control human brain was prepared in the presence of phosphatase inhibitors as described previously, omitting the chromatography stages (27).
Preparation and Purification of Recombinant Human Tau-A plasmid expressing the largest (2N4R) Tau isoform was used to prepare and purify recombinant human Tau as described previously (27). In-gel Proteolytic Digestion of Tau-PHF-tau or in vitro phosphorylated recombinant human Tau was separated on 10% (w/v) polyacrylamide gels and stained with colloidal Coomassie Blue G. Bands corresponding to Tau were excised, reduced, alkylated, and digested with trypsin or Asp-N as described previously (28).

Phosphorylation of Recombinant Tau by Protein Kinases-
Tandem MS/MS of Tau Peptides-Peptides obtained by proteolytic digestion were separated by reversed-phase chromatography using an Ultimate liquid chromatography system (Dionex) and sequenced by tandem MS/MS using orthogonal quadrupole time-of-flight electrospray ionization mass spectrometry (Q-Tof micro TM , Waters) as described previously (30). The mass spectral data were searched against a data base containing the sequences of all the human brain Tau isoforms using the Mascot searching algorithm (Matrix Science). Phosphopeptides and unphosphorylated peptides of Tau were identified based on the following search criteria: protease used (i.e. trypsin or AspN) with up to three missed cleavages; phosphorylation of Ser, Thr, or Tyr; carbamidomethylation of cysteine; oxidized methionine; deamidation of asparagine (Asn) and glutamine; and N-acetylation. All amino acid modifications were set as variable modifications. Phosphorylated residues were determined by manual inspection of the mass spectral MS/MS data. Confirmatory data were also obtained using multireaction monitoring (MRM) (29) on a hybrid triple quadrupole-linear ion trap mass spectrometer (QTRAP 4000 Applied Biosystems/ MDS Sciex). Briefly, peptide masses and major fragment ion masses for phosphopeptides of interest were obtained from previous LC-MS/MS experiments. These masses were used to allow the instrument to focus on phosphopeptides of interest thus increasing specificity and sensitivity. MRM transitions that were triggered were confirmed to be the correct peptide using conclusive MS/MS data, the loss of phosphoric acid (98 Da) from the parent, or the correct mass and charge state depending on the quality of the data.
Preparation and Treatment of Cortical Neurons with IC261-Primary cortical neurons were prepared from embryonic day 18 rats and cultured as described previously (30). Cells were maintained at a density of 10 6 cells/well (6-well plate) in Neurobasal medium (Invitrogen) supplemented with 2% (v/v) B27 supplement (Invitrogen), 2 mM L-glutamine, 100 IU penicillin, and 100 g/ml streptomycin for 10 days prior to treatment. The CK1 inhibitor 3-[(2,3,6-trimethoxyphenyl)methylidenyl]-indolin-2-one (IC261, Calbiochem) was dissolved in Me 2 SO and stored at Ϫ20°C as a 10 mM stock solution (31). Rat cortical neurons were treated with 10 M IC261 (or vehicle) for 4 h and (i) either harvested by removing the medium and scraping the cells into SDS-PAGE sample buffer (32), or (ii) the medium was replaced and neurons were incubated for a further hour in the absence of IC261 to assess recovery before harvesting into SDS-PAGE sample buffer.
Western Blot Analysis of Neuronal Tau Phosphorylation-Neuronal lysates were resolved on 10% (w/v) SDS-PAGE, electroblotted onto nitrocellulose, and probed with primary antibodies. Blots were developed with IRDye800-conjugated goat anti-rabbit (Rockland Immunochemicals, Inc.) or IRdye 680conjugated goat anti-mouse secondary antibodies (Molecular Probes). Labeled Tau was visualized using an Odyssey Infrared Imaging System (Li-Cor Biosciences).

RESULTS
Mass Spectrometric Analysis of PHF-tau-Combining the LC-MS/MS data generated from each of the enzymatic digestions of PHF-tau resulted in 90% sequence coverage of PHFtau, based on the amino acid sequence of the longest human Tau isoform in the central nervous system. This is equivalent to sequence coverage of 92% of the 85 phosphorylatable residues in Tau detected by MS/MS sequencing ( Fig. 1, sequenced residues in uppercase, and supplemental Table S1).
Following in-gel digestion and MS analysis of purified PHFtau, we detected phosphopeptides corresponding to 36 different phosphorylation sites (Fig. 1). In this study we did not detect phosphorylation at Ser 208 or Ser 356 , residues that have previously been reported to be phosphorylated on PHF-tau, despite obtaining full sequence coverage in these regions (Fig. 1). In the majority of cases, precise phosphorylation sites were identified, but we were unable to obtain sufficiently reliable MS/MS spectral data to ascertain the phosphorylation status of some residues, and as we found in our previous study (10), a short region close to the carboxyl terminus of Tau proved to be especially intractable to MS/MS analysis. In this region we frequently identified phosphopeptides with either one or two phosphorylation sites between residues Ser 412 and Ser 416 , a sequence that includes four closely spaced potential phosphorylation sites at Ser 412 , Ser 413 , Thr 414 , and Ser 416 , but it was not possible to identify the exact individual phosphorylated residues because of repeated poor fragmentation that yielded only weak b ions (see supplemental Table S1). We did not obtain sequence cov-erage of residues 127-163 in this study (Fig. 1). This region includes Thr 153 that has been suggested as a possible Tau phosphorylation site present in the Tau-immunoreactive smear characteristic of Alzheimer brain (9). A short eight-residue peptide corresponding to amino acids 291-298 was also not found, and this region contains a single potentially phosphorylatable residue at Ser 293 . In addition, we did not detect the amino-terminal methionine residue, which we presume is removed by post-translational modification because we found that Ala 2 is N-acetylated on PHF-tau.
Of the 36 phosphorylated residues we detected on PHF-tau, 11 are new sites not reported previously. This included phosphorylation on both Ser 184 and Ser 185 , one site on either Ser 184 or Ser 185 having been detected by us in an earlier study (10) (Fig. 1, phosphorylation sites indicated in boldface and new sites in black and gray boxes). Nevertheless, our improved Tau sequence coverage resulted in the discovery of new phosphorylation sites, which brings to 39 (37 Ser/Thr and two Tyr) the total number of directly identified phosphorylation sites in PHF-tau. We used MRM-MS to confirm the presence of some of the novel phosphorylation sites in brain tissue from three additional cases of Alzheimer disease. The PHF-tau used for these studies was prepared as described previously (10) and above, except that the insoluble Tau was analyzed after solubilizing in SDS-PAGE sample buffer. These results showed that at least seven of the new sites were present in other Alzheimer brains and at least three of these residues were phosphorylated in all cases examined. In comparison, Tau phosphorylated at Ser 422 , a site suggested to be specific to Alzheimer brain, was among several phosphorylated residues that we detected in only one of the three additional cases. All the data, taken together and including both directly identified phosphorylation sites and those sites identified with phosphospecific antibodies, suggest there may be as many as 45 different phosphorylation sites on PHF-tau, although not all sites are necessarily phosphorylated in sufficient amounts for detection in all cases of Alzheimer disease.
Phosphorylation in Different Regions of PHF-tau-We found two new phosphorylation sites within the microtubule-binding region of Tau (residues 244 -368), Ser 258 in the first repeat (exon 9) and Ser 289 within the second repeat (exon 10, only expressed in 4R isoforms), which is in an equivalent position to Ser 258 (see Fig. 2). This brings the total number of phosphorylated residues contained within the microtubule-binding domain of PHF-tau to four, because phosphorylation of Ser 262 in the first repeat and Ser 356 in the fourth repeat have been reported previously (9 -11). FIGURE 1. Mass spectrometric sequence coverage and phosphorylation sites identified on PHF-tau from Alzheimer brain. The amino acid sequence of the longest isoform of human brain Tau (2N4R) is shown. The residues in lowercase were not detected by MS/MS sequencing; the overall sequence coverage was 90%. Ala 2 is N-acetylated, indicating endogenous cleavage of the amino-terminal methionine in PHF-tau. The newly identified phosphorylation sites are shown in black or gray boxes, and previously reported sites identified by sequencing methods are indicated in boldface type. Phosphorylation was detected on both Ser 184 and Ser 185 , one site on either Ser 184 or Ser 185 having been detected by us in an earlier study (10). In one pair of residues (Thr 414 /Ser 416 , gray boxes) a single phosphorylation site was identified at only one of these two phosphorylatable amino acids. Phosphorylation of Ser 208 or Ser 356 was not detected in this study (double underlined). The most densely phosphorylated regions of Tau, which are rich in hydroxyamino acids, are the carboxyl-terminal region (394 -441) and the central region (175-238), flanking the microtubule-binding region. We discovered four new phosphorylation sites near the carboxyl terminus of PHF-tau, namely Thr 414 /Ser 416 (only one of these is phosphorylated, but the data do not allow unambiguous assignment to one or other residue), Thr 427 , Ser 433 , and Ser 435 . Phosphorylated Ser 416 has been reported in Alzheimer brains using a phosphospecific antibody (34), where it was primarily localized to the neuronal soma. The MS/MS spectra generated by peptides phosphorylated on Ser 433 and Ser 435 are shown in Fig. 3. Our results show that of the 13 Ser and Thr contained within the 50 most carboxyl terminal residues, 12 of these (92%) can be phosphorylated in PHF-tau.
We also identified new phosphorylation sites in the central region of Tau. Both Ser 184 and Ser 185 were found to be phosphorylation sites (previously only one phosphorylation site was identified that could be assigned to either Ser 184 or Ser 185 ) (10). This sequence of 64 amino acids in the center of Tau contains 21 potential phosphorylation sites of which 18 (86%) have now been identified chemically.
The phosphopeptides isolated from PHF-tau contained varying numbers of phosphates, showing that different PHF-tau molecules were differentially phosphorylated. From four short sequences of 7-10 amino acids (210 -217, 231-238, 394 -404, and 412-422), each containing four or five potential sites, we obtained peptides containing up to four phosphates (supplemental Table S1). Less phosphorylated peptides from these sequences were only seen in certain combinations, suggesting the possibility of "priming" or hierarchical phosphorylation (see "Discussion").
Other Post-translational Modifications in PHF-tau-In addition to phosphorylation, we detected N-acetylation and deamidation of PHF-tau, some of which has been reported previously by others (9,35). N-Acetylation of Ala 2 is consistent with the absence of the amino-terminal methionine in PHF-tau, and this modification is also found in Tau from normal human brain (see below) (8). We detected deamidated Asn at positions 167, 265, 279, 286, 359, and 381 on PHF-tau, supporting and extending a previous report of deamidation at residues 167 and 279 in smeared Tau from Alzheimer brain (35). Such deamidation is consistent with the suggestion that PHF-tau may be N-glycosylated on Asn residues because deglycosylation can result in deamidated asparagine that could result in increased degradation of Tau (36,37).
Phosphorylation of Tau from Control Human Brain-We obtained 85% sequence coverage of Tau from control adult human brain, based on the amino acid sequence of the longest human Tau isoform. This is equivalent to sequence coverage of 85% of the phosphorylatable residues in Tau (Fig. 4, sequenced residues in uppercase, and supplemental Table S3). Similar to PHF-tau, the initiator methionine was absent, and Ala 2 was N-acetylated as reported previously (8). We detected a total of eight phosphorylation sites in autopsy-derived normal human Tau, including five new sites (Ser 46 , Ser 199 , Ser 416 , and two out of three phosphorylatable residues in Ser 412 /Ser 413 /Thr 414 ) and FIGURE 2. Phosphorylation within the microtubule-binding repeat region of PHF-tau. Phosphorylation sites in the microtubule-binding repeat region PHF-tau are indicated in boldface type. Boxes indicate newly identified sites in this study, and phosphorylated residues reported previously are underlined. The remaining equivalent Ser residues in repeats 2, 3, and 4, not known to be phosphorylated in PHF-tau, are also numbered. The peptide mass and m/z of the fragment ions identify a phosphopeptide with the phosphate on Ser 435 of the longest isoform of human brain Tau. The y ion series (y7-y10) and the b ion series (b6 -b10) demonstrate that Ser 435 is dehydroalanine. Additionally, y7 ϩ P and y8 ϩ P fragment ions are observed that correspond to phosphorylated Ser 435 . three sites (Thr 181 , Ser 202 , and Ser 404 ) that have been reported previously (38). We did not detect phosphorylation of Thr 231 as had been reported by Morishima-Kawashima et al. (38), but this residue is contained within a short sequence of 18 amino acids for which we were unable to obtain MS/MS sequence coverage. Our results, combined with those of Morishima-Kawashima et al. (38), identified nine sites that can be phosphorylated in Tau extracted from normal adult post-mortem human brain. As with PHF-tau, we detected peptides corresponding to all six isoforms of Tau in control human brain (supplemental Table S4).
In Vitro Phosphorylation of Recombinant Human Tau with Individual Protein Kinases-With our discovery of 11 new phosphorylation sites on PHF-tau, 10 of which are not prolinedirected, we set out to determine which protein kinases are able to phosphorylate these sites in vitro. We selected four candidate kinases, CK1␦, CK2, PKA, and GSK-3␤, based on the following: 1) their ability to phosphorylate Ser/Thr residues in the absence of an adjacent Pro residue; 2) their possible involvement in either substrate priming or response to primed substrates; and 3) their potential involvement in Alzheimer pathogenesis. Each enzyme was assessed, both individually and in combination with the other three protein kinases, by investigating the phosphorylation sites generated on recombinant human brain Tau (2N4R) following incubation in the presence of ATP.
The sequence coverage and phosphorylation sites found on Tau generated by CK1␦, CK2, PKA, and GSK-3␤ are shown in Fig. 5. The phosphopeptides used to determine these sites are available as supplemental Tables S5-S9. The amino acid sequence coverage obtained for Tau after in vitro phosphorylation by these kinases and identified by MS/MS sequencing was 82-85% (Fig. 5), equivalent to 84 -100% when measured as the percentage of potentially phosphorylatable residues. Thus we were able to obtain complete or largely complete coverage of all potential phosphorylated residues under the phosphorylation conditions employed. In all cases, we did not detect the amino-terminal methionine of recombinant Tau, although Ala 2 was not acetylated, unlike in PHF-tau and Tau from control human brain.
Phosphorylation with CK1␦ generated 33 sites on Tau, three of which were identified as one of two closely clustered alternative sites ( Fig. 5 and supplemental Table S5). This is the first study using chemical identification of CK1␦ phosphorylation sites on Tau. We did not detect CK1␦ phosphorylation of Ser 202 /Thr 205 or Ser 396 /Ser 404 , sites previously identified using the phospho-Tau antibodies, AT8 and PHF-1, respectively (23). Because we obtained complete sequence coverage of these residues following CK1␦ phosphorylation, our data suggest that the stoichiometry of phosphorylation at these epitopes may be low such that corresponding phosphopeptides could not be detected by our MS protocol. CK1␦ phosphorylation generated nine sites on Tau that have not been shown to be phosphorylated by any other kinases (namely Thr 17 , Thr 95 , Thr 101 /Thr 102 , Ser 113 , Thr 169 , Ser 238 , Thr 263 , Ser 341 , and Ser 433 ). Fifteen of the CK1␦ sites identified in Tau are also phosphorylated on PHF-tau. At least six of the newly identified Ser/Thr phosphorylation sites on PHF-tau are also phosphorylated by CK1␦, and three of these (Ser 113 , Ser 238 , and Ser 433 ) are not known to be phosphorylated by any other kinase, indicating that CK1␦ may have an important role in neurofibrillary pathology. Of the nine phosphorylation sites found on normal brain Tau, four are potential CK1 sites, three of which are in the carboxyl-terminal region.
After phosphorylation by CK2, we detected only six phosphorylation sites on Tau (Fig. 5 and supplemental Table S6), only one of which (Thr 39 ) has been reported previously (39). Two of these additional sites were not pinpointed to a single residue. We found only three sites on Tau (Ser 199 , Ser 400 , and a site within Ser 412 -Ser 416 ), which are targets of in vitro phosphorylation by CK2 and are also phosphorylated on PHF-tau. Thus, in contrast to CK1␦, CK2 does not appear to favor Tau as a primary substrate and may have a less significant role than CK1␦ in PHF generation.
PKA phosphorylated 16 sites on Tau, two of which were localized to peptides with more than one possible phosphorylation site (Fig. 5 and supplemental Table S7). Ten of the PKA sites are also present on PHF-tau, with a further two sites that may also be present (e.g. Thr 217 is phosphorylated on PHF-tau, but we were able to identify one phosphorylation site at either Thr 217 or Thr 220 after PKA phosphorylation). Seven of the PKA sites that we identified have not been reported previously (40 - . Mass spectrometric sequence coverage and phosphorylation sites identified on Tau from control human brain. The amino acid sequence of the longest isoform of human brain Tau (2N4R) is shown, and the residues in lowercase were not detected by MS/MS sequencing. The overall sequence coverage for Tau from adult human brain was 85%. Ala 2 is N-acetylated, indicating endogenous cleavage of the amino-terminal methionine in Tau from control human brain. The five newly identified phosphorylation sites are shown in black or gray boxes, and previously reported sites identified by sequencing methods are indicated in boldface type (residues Thr 181 , Ser 202 , Thr 231 , and Ser 404 ). Phosphorylated Thr 231 was not detected in the present study. In one set of residues containing three phosphorylatable amino acids (Ser 412 /Ser 413 /Thr 414 , gray box), two phosphorylation sites were identified. Sequences corresponding to alternatively spliced regions of Tau (exons 2, 3, and 10) are boxed. AUGUST 10, 2007 • VOLUME 282 • NUMBER 32 43), and at least four of these are also present on PHF-tau. Our results therefore support previous suggestions that PKA phosphorylation of Tau, possibly in conjunction with other kinases, might also have a significant part to play in tangle pathogenesis.

New Phosphorylation Sites on PHF-tau
We found 28 phosphorylation sites generated on Tau by GSK-3␤, two of which were not identified uniquely (Fig. 5 and supplemental Table S8). Of the GSK-3␤ sites detected in this study, 13 have not been identified previously by direct means and 15 sites are also present on PHF-tau, including at least two of the newly identified PHF-tau sites (supplemental Table 10). When all of the new GSK-3␤ phosphorylation sites on Tau identified here are combined with those detected in previous studies (discussed in Ref. 44), it becomes apparent that there are at least 36 residues in Tau that are targeted by this enzyme, and 27 of these sites are likely to be phosphorylated on PHF-tau.
Thus GSK-3␤ continues to represent a good candidate for the aberrant phosphorylation of Tau in Alzheimer disease.
Phosphorylation of Recombinant Human Tau by Combined Kinases-A total of at least 52 phosphorylation sites (three being ambiguous) is obtained by combining the data for each of the four kinases (CK1␦, CK2, PKA, and GSK-3␤) when used separately. To compare this with simultaneous exposure to these protein kinases, as may occur in neurons, phosphorylation with these four kinases together was carried out. The combination of CK1␦, CK2, PKA, and GSK-3␤ produced a total of 32 phosphorylation sites on Tau, of which three were ambiguous ( Fig. 5 and supplemental Table S9). Therefore, this combination of kinases generated in total fewer sites than the kinases had individually, but one additional phosphorylated residue was found, at a site not detected with these kinases used indi- vidually, namely Thr 111 , a proline-directed site that has not been found in PHF-tau. The combination of kinases acting in concert, rather than individual kinases acting alone, may be responsible for Tau phosphorylation at this residue. However, Thr 111 was the only phosphorylation site in a long peptide (supplemental Table S5), and so if this phosphorylation was because of priming it would suggest that this was at a distant site. Because CK1 and PKA have been shown previously to prime some substrates for subsequent phosphorylation by GSK-3␤, it is possible that these three kinases acting in concert may also be responsible for generating several of the phosphorylation sites on PHF-tau.
Supplemental Table 10 shows a summary of all the in vitro phosphorylation sites identified in Tau by MS/MS sequencing following incubation with the individual and combined protein kinases in this study. The PHF-tau phosphorylation sites determined by direct sequencing and antibody identification are included for comparison.

Inhibition of CK1 in Rat Cortical Neurons Decreases Endogenous Tau Phosphorylation-Cultures of rat cortical neurons
were treated with the CK1␦/⑀ inhibitor IC261 (10 M for 4 h), and endogenous Tau was analyzed by Western blotting with two unrelated phosphorylation-independent anti-Tau antibodies (Fig. 6, A and B). The intensity of the upper Tau band was decreased by IC261, whereas that of the lower bands was increased. This shift strongly suggests that the inhibitor is causing a decrease in Tau phosphorylation. This view was supported by probing blots with a phosphorylation-dependent antibody recognizing phospho-Ser 262 on Tau (Ser(P) 262 ). Phosphorylation at Ser 262 was reduced by IC261, in agreement with our MS analysis showing that Tau is phosphorylated on this residue in vitro by CK1␦. Washing out the inhibitor (right-hand lanes in Fig. 6) restored the electrophoretic mobility and Ser 262 phosphorylation of endogenous Tau. These results suggest that the activity of endogenous CK1 was sufficient to partially phosphorylate Tau in neurons.

DISCUSSION
Improved Identification of Phosphorylation Sites on PHF-tau and Tau from Control Human Brain-Using LC-MS/MS we obtained 90% Tau sequence coverage and identified 11 previously undetected phosphorylation sites on PHF-tau. This is a marked improvement on previous assessments of PHF-tau phosphorylation in which we and others obtained up to 60% sequence coverage and together identified a total of 28 phosphorylation sites (9 -11). These improvements are because of the increased sensitivity of detection obtained using on-line LC-MS/MS and orthogonal quadrupole time-of-flight electrospray ionization mass spectrometry, in contrast to nanoelectrospray and precursor ion scanning used previously, as well as the use of more advanced data acquisition and searching algorithms. Our new findings, together with previous results from this and other laboratories, bring to 39 the number of directly identified phosphorylation sites on PHF-tau from Alzheimer disease brain. Thus, more than 45% of the 85 potentially phosphorylatable residues in PHF-tau can be phosphorylated. With our new analysis of purified PHF-tau, including almost complete sequence coverage, it is likely that few phosphorylation sites on PHF-tau remain to be identified. However, the existence of six additional sites, indicated by their immunoreactivity with phosphospecific antibodies to Tau, suggests that overall more than half of all possible sites may be phosphorylated on PHF-tau. There have been few studies of Tau from autopsyderived control subjects, and together these report only four Tau phosphorylation sites (9,38). Here we report an additional five sites, bringing the total number of reported phosphorylation sites in normal brain Tau to nine, suggesting that, in contrast to PHF-tau, less than 12% of phosphorylatable residues are detectably phosphorylated in autopsy Tau from control adult human brain.
Alternatively Spliced Tau Isoforms in PHF-tau-Three of the new phosphorylation sites that we discovered on PHF-tau, Ser 68 , Thr 69 , and Thr 71 , are present in the first of the two alternatively spliced, tandem amino-terminal inserts in Tau (Fig. 1). Thus these residues are present in 1N and 2N but not 0N Tau isoforms that lack these inserts. We detected these phosphorylation sites in phosphopeptides derived specifically from 1N, but not 2N, Tau isoforms, indicating that residues 68, 69, and 71 are either not phosphorylated on 2N Tau isoforms or, because the 2N isoforms are expressed only at relatively low levels compared with the other four brain Tau isoforms, phosphopeptides originating from 2N Tau may be below the detection limit of the MS/MS methods used here. We were able to identify peptides specific for 0N, 1N, and 2N, as well as three (3R) and four (4R) microtubule-binding repeats (supplemental Table S2). These results provide direct support for the previous observations of all six isoforms of Tau in PHF-tau that have been determined by specific antibody labeling (15) and by comparison of bands of dephosphorylated PHF-tau with the six recombinant human brain Tau isoforms (3,45).
Clustering of Tau Phosphorylation Sites and Possible Sequential Phosphorylation of PHF-tau-The phosphorylation sites on PHF-tau appear to occur in clusters along the Tau molecule giving multiple phosphorylated peptides with possible mutually exclusive combinations of phosphorylation. It is therefore feasible that such sequences may represent regions in Tau where sequential phosphorylation might occur, with initial phosphorylation of a specific Ser/Thr that primes Tau for secondary phosphorylation at a nearby site. For example, in the case of residues spanning Ser 210 -Thr 217 , mono-phosphorylated peptides were identified corresponding to Ser 214 and Thr 217 , but phosphorylation at Ser 210 or Thr 212 was detected only when phosphorylation at Ser 214 or Thr 217 was also present. Similarly, Thr 231 was detected in the mono-phosphorylated form, whereas Ser 235 , Ser 237 , and Ser 238 were identified only in combination with Thr 231 . This is an important observation because dual phosphorylation of Thr 212 and Ser 214 would generate the "Alzheimer-specific" epitope on Tau for the monoclonal antibody AT100 (46), and the Alzheimer-specific antibody PHF-27 requires phosphorylation of both Thr 231 and Ser 235 (47). Our results therefore agree with previous suggestions of a temporally ordered addition of phosphate to Tau in which phosphorylation of specific residues leads to subsequent phosphorylation at nearby amino acids in brain (48 -50).
Tau Phosphorylation in Vitro, Individual and Sequential Action of Protein Kinases-Recombinant Tau was phosphorylated in vitro with individual candidate kinases to identify potential kinases for phosphorylating particular sites, and combinations of kinases to identify possible priming. Many in vitro studies of Tau phosphorylation have been reported, with in excess of 20 different protein kinases being suggested as potential candidates for Tau phosphorylation (19). The challenge now is to identify the protein kinases responsible for the pathological phosphorylation of Tau that generates PHF-tau and tangles in Alzheimer disease. Among the Tau candidate kinases, GSK-3␤ and cyclin-dependent kinase-5 are the most frequently cited contenders for a pathological role in the production of PHF-tau, and the results presented here suggest that CK1 may also play an important part. Notably, three sites that are phosphorylated in PHF-tau (Ser 113 , Ser 238 , and Ser 433 ) are generated only by the action of CK1␦ and have not been reported to be phosphorylated by any other kinases. Taken together with previous reports of increased CK1␦ expression in Alzheimer brain, the co-localization of CK1␦ with tangles (22,51), and the possible role of CK1 in regulation of A␤ secretion (25), our new data further endorse previous suggestions that CK1 may have a role to play in the generation of PHF-tau and tangle pathology in Alzheimer disease.
Both GSK-3␤ and CK1 can phosphorylate certain of their substrates more efficiently when they have been primed by prephosphorylation with other kinases (52,53). Such substrates include ␤-catenin (54,55), phosphorylation of which targets it for proteolysis, as well as other components of the Wnt signaling pathway (56,57). CK1 can more efficiently phosphorylate some substrates, such as the tumor suppressor protein, p53, and cubitus interruptus in the Hedgehog signaling pathway, when they are pre-phosphorylated by a different priming kinase (58 -60); this suggests that hierarchical phosphorylation may be mandatory for certain substrates.
Several studies point toward Tau phosphorylation being partly sequential or hierarchical. A neurofibrillary tangle-associated kinase phosphorylates Thr 361 and Ser 412 in Tau (61) only after pre-phosphorylation of Tau by PKA at the nearby residues Ser 356 and Ser 409 , consistent with the substrate specificity of CK1. This PHF-tau kinase was described as having similarities to the CK1 family, and we also detected phosphorylation of Thr 361 by CK1␦ in the present study. Previous reports showed that CK1 purified from bovine brain phosphorylated recombinant human 2N3R Tau at epitopes recognized by antibodies M4 (Thr 231 ) and SMI-31 (Ser 396 and Ser 404 ), but not PHF-1 (Ser 396 /Ser 404 ) (62), and that recombinantly expressed CK1␦(⌬317) phosphorylated the epitopes for antibodies AT8 (Ser 202 /Thr 205 ) and PHF-1 in recombinant human 2N4R Tau (23). However, we were unable to detect CK1␦-catalyzed Tau phosphorylation at any of these five sites, despite obtaining sequence coverage of each. It is possible that phosphorylation at these epitopes may create strong antigenic determinants, but the stoichiometry of CK1␦ phosphorylation at these residues may not have been sufficient for detection using the MS technologies used here.
Our experiments to examine the potential role of CK1␦ in priming substrates for sequential phosphorylation identified two possible instances, involving Thr 111 and Thr 205 . Thr 111 was phosphorylated when four kinases (CK1␦, CK2, GSK-3␤, and PKA) were used together. Because Thr 111 is adjacent to a proline residue, it is likely that GSK-3␤ was responsible, possibly after priming of Tau at another site (or activation of GSK-3␤) by one of the other kinases. Thr 111 is not a known PHF-tau site, and this was not investigated further. Thr 205 , a PHF-tau site identified by antibodies, has been shown to be weakly phosphorylated by GSK-3␤ alone (63), yet in our hands we only observed its phosphorylation after pretreatment with CK1␦. 5 We therefore propose that CK1␦-catalyzed phosphorylation might make Tau a better substrate for GSK-3␤, i.e. increase its rate of phosphorylation. Residues 175-238 and 392-441, the regions of Tau that flank the microtubule-binding repeats, contain the majority of the phosphorylation sites, and approximately threequarters of the PHF-tau sites in these regions can be phosphorylated in vitro by CK1␦, GSK-3␤, or PKA. CK1␦ and GSK-3␤ are well known to recognize primed substrates, which could account for why many of the peptides recovered from these flanking regions in PHF-tau contain multiple phosphates. Indeed, priming may also be important for Tau phosphorylation in the normal brain; a peptide derived from the carboxylterminal region contained three phosphates although their exact locations could not be determined (supplemental Table  S3), even though the overall phosphorylation level was much lower than in PHF-tau. Quantitative measurements on multiphosphorylated peptides will need to be developed to study this further.

Possible Functional Effects of Tau Hyperphosphorylation-
Because phosphorylation within the microtubule-binding repeat domain (Gln 244 -Asn 368 ) of Tau decreases its binding to microtubules, phosphorylation in this region is likely to have a critical function in regulating Tau binding to microtubules in neurons. Hence the presence of four phosphorylation sites identified in the microtubule-binding repeat region in PHF-tau but not control brain Tau (i.e. Ser 258 , Ser 262 , Ser 289 , and Ser 356 ), including two new PHF-tau phosphorylation sites detected in this study, is highly significant. The protein kinases most closely matching this pattern of Tau phosphorylation are GSK-3␤ and CK1␦, because each of these kinases phosphorylates all four of the sites detected in this region of PHF-tau. Consistent with this finding, we and others have found that Tau binding to microtubules is reduced when either GSK-3␤ or CK1␦ is exogenously expressed with Tau in non-neuronal cells 6 (23,64). In comparison, microtubule-affinity regulating kinase, which also disrupts Tau-microtubule binding in transfected cells, phosphorylates six sites in the microtubule-binding domain of Tau (Ser 262 , Ser 293 , Ser 305 , Ser 320 , Ser 324 , and Ser 356 ), only two of which (Ser 262 and Ser 356 ) correspond to PHF-tau phosphorylation sites (65).
Our inhibition studies of endogenous CK1 (Fig. 6) suggest that CK1␦ and/or ⑀ is an active physiological Tau kinase in neuronal cells. We have also found that Tau exhibits a similar shift in electrophoretic mobility when phosphorylated in vitro with CK1␦, 7 and our finding of phosphorylation in Tau from normal control brain near its carboxyl terminus (a modification that decreases the mobility of Tau on gels), at potential CK1 sites, supports this interpretation. Although this result would be consistent with CK1 phosphorylating Tau directly, it is also possible that CK1 could activate another kinase or inactivate a phosphatase. The strong elevation of CK1␦ reported in Alzheimer brains (20,21) points to a risk of severe and pathological disruption of Tau-microtubule binding, and possibly of other intracellular interactions (66).
Kinases have been identified that are capable of phosphorylating most of the chemically determined Ser/Thr residues on PHF-tau. All but five or six of these can be phosphorylated in vitro by a combination of CK1␦ and GSK-3␤. The exceptions are Ser 68 , Thr 71 , Ser 185 , Ser 412 , Ser 422 , and possibly Thr 414 (not a confirmed site). Of these, Ser 185 can be phosphorylated by p38 MAPK (44), and Ser 412 can be phosphorylated by PKA (this study) and by "PHF kinase" (61). Ser 422 is a substrate for MAPKs and also for Tau-tubulin kinase 1, a neuron-specific member of the CK1 family that phosphorylates seven residues on Tau, all of which are present in PHF-tau (67)(68)(69)(70). Tau-tubulin kinase 1 has been reported to reduce Tau solubility when exogenously expressed in human neuroblastoma cells, and this could be due to phosphorylation at Ser 422 or other residues targeted by this kinase (70). Phosphorylation at Ser 422 may be important in pathology as it has been reported to be necessary for the generation of A␤-induced PHF-like Tau filaments in neuroblastoma cells (71).
There remain on PHF-tau two definite phosphorylation sites and one possible site for which no kinase has been identified (orphan sites). All of these phosphorylation sites correspond to newly identified PHF-tau sites reported herein, i.e. Ser 68 and Thr 71 , with Thr 414 if this is subsequently shown to be responsible for the phosphorylation identified as Thr 414 /Ser 416 ; Ser 416 being phosphorylated by CK1␦, PKA, calcium-calmodulin kinase II, and Tau-tubulin kinase 1 (70).
Our findings therefore support the view that PKA and CK1␦ might be important for subsequent Tau phosphorylation by GSK-3␤. Taken together with the roles of CK1 in the regulation of the intracellular trafficking of ␤-site amyloid precursor protein cleaving enzyme (72,73) and in phosphorylation of presenilin (74), our data suggest that CK1 could have a role in the pathogenesis of Alzheimer disease. It remains to be determined which sites on PHF-tau prove to be unique determinants of Alzheimer pathology, but one possibility is that phosphorylation of some or all of these orphan sites requires the concerted action of several different protein kinases to generate all of the known phosphorylation sites on PHF-tau.
Further studies are now warranted to determine the effects of Tau phosphorylated by CK1 on Tau aggregation and/or cellular toxicity as has been shown previously with Tau tubulin kinase 1 (70), Congo Red (75), and A␤ (71).