Protein Phosphatase 1 Is Targeted to Microtubules by the Microtubule-associated Protein Tau*

Phosphorylation has been implicated in the regulation of microtubule (MT) stability and function by controlling the interactions between MTs and MT-associated proteins. We found previously that protein phosphatase inhibitors selectively break down stable MTs, suggesting that protein phosphatases may be involved in regulating MT stability. To identify the protein phosphatases involved, we examined purified calf brain MTs and found a protein phosphatase activity that copurified with MTs to constant stoichiometry. Western blot analysis and inhibitor profiles demonstrated that the MT-associated phosphatase was a type 1 protein phosphatase (PP1), which we named PP1MT. Recombinant PP1 catalytic subunit (PP1c) did not bind to MTs, whereas PP1MT did bind, suggesting the presence of proteins that target PP1 to MTs. By Sepharose CL-6B chromatography, the phosphatase activity of PP1MT eluted as a large protein complex of ∼400 kDa. High salt (2 m NaCl) treatment followed by CL-6B chromatography dissociated PP1MT into PP1c and the MT-targeting subunit(s). The MT-targeting subunit was shown to be the MT-associated protein tau by PP1 blot overlays and other assays. Also, recombinant tau reconstituted the binding of PP1c to MTs. These results identify PP1 as the first tau binding protein and suggest that tau is a novel PP1-targeting subunit.

Microtubules (MT) 1 are major cytoskeletal elements in eukaryotic cells and play essential roles in cell motility, division and differentiation. The assembly and functions of MTs are dependent on their interactions with microtubule-associated proteins (MAPs) (1)(2)(3)(4)(5). Specifically, MAPs including MAP1, MAP2, and tau have been found to control MT dynamics and stability and to contribute to neuronal cell functions in the nervous system (6 -10). Although the mechanisms by which MT function is regulated are not completely understood, phosphorylation of MAPs has been shown to regulate their interactions with MTs (1)(2)(3)(4). In general, increased phosphorylation of MAPs reduces the affinity of the MAPs for MTs (11)(12)(13)(14). Hy-perphosphorylation of tau may underlie the formation of paired helical filaments (PHF) found in neurofibrillary tangles, one of the hallmarks of Alzheimer's Disease (AD) brains (15)(16)(17)(18).
Protein kinases that phosphorylate MAPs have been intensely studied (19 -21). In contrast, much less is known about the protein phosphatases involved. However, several lines of evidence suggest that protein phosphatases may be rate-limiting enzymes in the regulation of MAP-MT interactions. In our previous study, phosphatase inhibitors okadaic acid (OA) and calyculin-A (CL-A) were found to induce the complete loss of stable MTs without significantly affecting the level of dynamic MTs in interphase cells (22). The dose response curves of OA and CL-A in this study suggest that protein phosphatase 1 (PP1) is involved in maintaining stable MTs. Treatment of interphase sea urchin egg extracts with OA induces mitoticlike MT dynamics by eliminating the rescue phase of MT dynamics (23). Incubation of mouse oocytes with OA leads to lengthening of mitotic spindles, disorganization of the metaphase plate, and phosphorylation of proteins associated with MTs (24). These results imply that protein phosphatases are necessary for dephosphorylation of MAPs, resulting in increased MT assembly and stability in vivo. However, the identity of the protein phosphatases involved, their specific in vivo substrates, and how are they regulated in response to internal and external cues are still unknown.
PP1 is one of the four major serine/threonine phosphatases found in eukaryotic cells. It plays important roles in diverse cellular activities, such as carbohydrate and lipid metabolism, calcium transport, gating of ionic channels, muscle contraction, nuclear organization, protein synthesis, mitosis, and meiosis (25)(26)(27)(28). The fact that numerous and diverse substrates are dephosphorylated by this single enzyme in various subcellular localizations suggests the existence of mechanisms to determine and regulate phosphatase specificity and activity. Emerging evidence has supported a general mechanism whereby PP1 is directed to various substrates by specific targeting subunits that also modulate the enzymatic activity of PP1 toward its substrates (29). Thus, for a particular cellular event regulated by PP1, it is essential to identify the targeting subunit that directs PP1 to protein substrates involved in the event.
Several protein kinases and one phosphatase have been found to associate with MTs (30 -34). In all known cases, the kinases were found to associate with MTs through MAPs. MAP2 was found to bind both the regulatory subunit of Type II cAMP-dependent protein kinase and MAP kinase to MTs (31,32), and MAP4 was found to bind cyclin B, which may direct Cdc2 kinase to MTs (33). This suggests that regulatory enzymes associated with MTs may use members of the MAP family to mediate their association. This may be a general mechanism that also applies to MT-associated phosphatases, although the mechanism of their association has not been elucidated.
In this study, we report that there is a PP1 activity that remains tightly associated with MTs during their biochemical isolation from brain tissue. We provide evidence that this association is mediated by a MT-targeting protein. Using biochemical assays, we show that the MAP tau is predominantly responsible for binding PP1 to MTs. Our results suggest that tau is a novel PP1-targeting subunit and raise the possibility that phosphatases and kinases use different MAPs to associate with MTs.

EXPERIMENTAL PROCEDURES
Materials-Okadaic acid and CL-A were from LC Laboratories (Woburn, MA). Taxol was obtained from the Drug Synthesis and Chemical Branch of the Division of Cancer Treatment, NCI. All tissue culture media were from Life Technologies, Inc. Sepharose CL-6B, blue dextran, and catalase were obtained from Amersham Pharmacia Biotech.
Antibodies-A peptide antibody against PP1 was generated by immunizing rabbits with a synthetic peptide CDLQSMEQIRRIM (residues 180 -191 of PP1c with an additional Cys residue at the N terminus for coupling) corresponding to a region conserved among all PP1 isoforms. Rabbit antisera were produced by Pocono Rabbit Farms (Canadensis, PA) by immunizing rabbits with the peptide-keyhole limpet hemocyanin conjugate (50 g peptide/injection) (35). A peptide antibody against PP2A was generated as described previously (36). Tau antibody T46, which reacts with all tau isoforms (37), was a generous gift from Dr. J. Q. Trojanowski (University of Pennsylvania).
MT Assembly/Disassembly Cycles and Preparation of Tubulin and MAPs-MTs were prepared from calf brains by 3-4 cycles of warm assembly/cold disassembly as described (38). MAPs and tubulin were prepared from thrice-cycled MT pellets by DEAE Sephadex A-50 frozen chromatography (38), frozen in liquid nitrogen, and stored in aliquots at Ϫ80°C.
Gel Filtration Chromatography under Low and High Salt Conditions-All steps were performed at 4°C. A Sepharose CL-6B column (0.9 ϫ 90 cm) was equilibrated in PEM buffer (100 mM Pipes, pH 6.9, 1 mM MgCl 2 , 1 mM EGTA) (low salt) or PEM buffer containing 2 M NaCl (high salt). Blue dextran and catalase were used to calibrate the column. For low salt conditions, PP1c or MAPs were applied to the column and eluted with PEM buffer at 0.2 ml/min. Fractions of 0.5 ml were collected. For high salt conditions, PP1c or MAPs were incubated in PEM containing 2 M NaCl overnight at 4°C and then applied to the column and eluted with PEM containing 2 M NaCl. Fractions from the high salt column were dialyzed in PEM buffer overnight using a microdialysis system (Life Technologies, Inc.). Protein concentrations of the column fractions were estimated by A 280 .
Preparation of [ 32 P]Phosphorylase and Assay of Protein Phosphatase Activity-Phosphorylase b was purified from rabbit skeletal muscle as described (39) and labeled with [␥-32 P]ATP (40). Phosphatase assays were performed as described (40), except that 0.5 mM MnCl was included in the assay buffer. A PEM buffer blank was run in parallel, and the background was subtracted from each sample. One unit of activity is equal to one nmol of phosphate released per min.
The inhibitor profile of the MT-associated protein phosphatase was determined by adding inhibitors to the reaction mixture before adding the [ 32 P]phosphorylase a substrate. For experiments with the heat stable inhibitor I2, samples were preincubated for 15 min at 30°C prior to adding the substrate. I2 was purified as described previously (41). The rabbit skeletal muscle PP1c used for comparison in these experiments was purified to homogeneity as described earlier (42).
MT Cosedimentation Assay-All procedures were performed at room temperature. Samples were first centrifuged at 50,000 ϫ g for 15 min at room temperature to remove aggregates. MTs were prepared by polymerizing DEAE-purified tubulin with 20 M taxol as described (43). Samples were incubated with taxol-MTs in PEM buffer for 20 min followed by centrifugation at 50,000 ϫ g for 10 min. The supernatants and pellets were recovered, and phosphatase activity was measured. In control experiments, PEM buffer was used instead of MTs to assess nonspecific sedimentation.
MT Targeting Assay-Fractions from the Sepharose CL-6B column eluted with 2 M NaCl were dialyzed and assayed for their ability to bind recombinant PP1c␣ to taxol-MTs. Purified recombinant PP1c␣ was a generous gift of Dr. E. Y. C. Lee (New York Medical School, Valhala, NY) and was prepared as described previously (44). Recombinant PP1c (0.13 g) was incubated with ϳ2 g of protein from each fraction for 10 min at room temperature. Samples were preclarified before incubation as described above. Taxol-MTs (20 g) were then added to the mixture and incubated for 20 min. Supernatants and pellets were prepared by centrifugation (50,000 ϫ g for 10 min) and then assayed for phosphatase activity. The pellets were solubilized in PEM buffer before assay. To measure the activation of PP1c activity by column fractions, PP1c was incubated for 10 min with each column fraction that was used in the MT targeting assay and then assayed directly for phosphatase activity. The activity of PP1c alone was used as a base line to calculate the magnitude of activation by each column fraction. The phosphatase activity in each MT pellet in the targeting assay was then normalized by the level of activation in the absence of MTs.
Blot Overlay and Western Blot Analysis-Protein samples were run on 4 -14% SDS-PAGE and, transferred to a nitrocellulose membrane, and blots were stained by Ponceau Red. The blot was blocked with NET buffer (150 mM NaCl, 50 mM Tris, pH 7.5, 5 mM EDTA, 0.05% Triton, 2.5 g/liter gelatin) containing 1% bovine serum albumin at room temperature for Ͼ 1 h. For blot overlays with PP1c, the membrane was incubated with 1.8 g/ml of recombinant PP1c in NET buffer for 2 h at room temperature followed by extensive washing with NET buffer. The membrane was then incubated with rabbit PP1 peptide antibody (at 1:2000 dilution) in NET buffer for ϳ2 h. The blot was washed and incubated in goat anti-rabbit horseradish peroxidase (Cappel, Organon Teknik Co.) at 1:20,000 dilution in NET buffer followed by detection with ECL solution. For Western blot analysis, blots were blocked with NET buffer containing 1% bovine serum albumin and incubated with the primary antibodies indicated under "Results." Blots were then developed with horseradish peroxidase-conjugated secondary antibodies and ECL.
Preparation of Recombinant Tau in Sf9 Cells-Sf9 cells were infected by recombinant tau baculovirus (provided by Dr. T. Frappier) as described (45). After 2 days of infection, the cells were collected, washed twice in phosphate-buffered saline buffer containing proteinase inhibitors chymostatin, leupeptin, antipain, and pepstatin (each at 0.1 mg/ ml) and phenylmethylsulfonyl fluoride (0.2 mM), resuspended in lysis buffer (50 mM Pipes, pH 6.9, 50 mM ␤-glycerol phosphate, 1 mM EGTA, 0.5 mM MgCl 2 , phenylmethylsulfonyl fluoride, and chymostatin, leupeptin, antipain, and pepstatin), and then sonicated. NaCl and ␤-mercaptoethanol were added to the cell extract at final concentrations of 0.4 M and 10 mM, respectively. The mixture was then boiled for 5 min and centrifuged at 30,000 ϫ g for 15 min. The supernatant containing heat stable tau was dialyzed in PEM buffer and stored at Ϫ80°C. The sample was Ͼ 90% tau as judged by quantitative SDS-PAGE.

Identification of a MT-associated Protein Phosphatase 1-To
identify protein phosphatases that might be involved in regulating MT stability, preparations of MTs purified from calf brain homogenates were obtained by cycles of warm assembly and cold disassembly using the standard protocol (38). After three or four cycles, the preparation yielded about 80% assembly competent tubulin and 20% presumptive MAPs. Using [ 32 P]phosphorylase, a standard substrate for both PP1 and PP2A, we detected a phosphatase activity that co-assembled with MTs throughout four cycles of MT polymerization/depolymerization (Table I). Approximately 5% of the total soluble phosphatase activity in brain extracts cosedimented with MTs during the first cycle. With additional cycles, about half of this activity was lost (compare activity in P1 with that of P2-P4), presumably because of the loss of phosphatases loosely associated with MTs or with contaminants that were removed in subsequent cycles. Nonetheless, 1-2% of the total extract activity cycled with MTs to constant stoichiometry from the second to fourth cycles. The phosphatase activity was also detected in DEAE-purified MAP fractions and could be dissociated quantitatively from MTs by 0.3 M salt treatment, which is a characteristic of conventional MAPs (data not shown).
To determine whether PP1 or PP2A were present in the cycled MT preparations and hence might be responsible for the phosphatase activity, we probed Western blots of the fractions from the MT assembly/disassembly cycles using peptide-spe-cific antibodies to either PP1 or PP2A. As shown in Fig. 1A, a single band of ϳ37 kDa, which comigrated with bacterially expressed recombinant PP1c, was detected in MT pellet fractions (P1-P4) probed by PP1 antibody. The levels of PP1 judged by the immunoreactive bands in the last three pellet fractions (P2-P4) were similar, which is consistent with the constant stoichiometry of phosphatase activity associated with MTs (Table I). Accordingly, there is little detectable PP1 in supernatant fractions in the last two cycles (S3 and S4). PP1 was also detected in the MAP fraction purified by DEAE-Sephadex chromatography of MT proteins cycled three times (Fig. 1A).
In contrast, PP2A was detected in the first MT pellet fraction but in decreasing amounts in subsequent MT pellets so that by the third or fourth cycle, no PP2A was detected in MT pellets (Fig. 1B). No significant amount of PP2A was detected in the DEAE-purified MAP fraction (Fig. 1B). Also, no substantial phosphatase activity was detected in purified MAPs when pnitrophenyl phosphate, a preferred substrate for PP2A but a much less sensitive substrate for PP1 (46), was used to assay phosphatase activity (data not shown). No increase of activity was observed when the PP2B activator Ca 2ϩ was included in the phosphatase assay (data not shown) (26). These results suggest that PP1, but not PP2A or PP2B, may be responsible for the phosphorylase phosphatase activity that coassembles with MTs to constant stoichiometry.
To demonstrate that the MT-associated phosphatase is indeed PP1, the inhibitor profile of phosphatase activity in the fourth cycle MT pellet (P4) was characterized. OA inhibits phosphatases at different concentrations with an IC 50 value of 0.1-1 nM for PP2A, ϳ50 nM for PP1, and Ͼ1 M for PP2B and PP2C (48). As shown in Fig. 2A, the dose response curve of the MT-associated phosphatase to OA was nearly identical to that measured for PP1c purified from skeletal muscle with an IC 50 of about 50 -60 nM (49). Similarly, the dose response curve of the MT-associated phosphatase in P4 toward CL-A closely resembled that of PP1c with an IC 50 of ϳ10 nM (Fig. 2B) (49). The greater sensitivity of the MT-associated phosphatase to CL-A compared with OA, is a characteristic of PP1 but not PP2A phosphatases (49). Moreover, both PP1c and the MT-associated phosphatase were inhibited by heparin and protamine, two specific inhibitors of PP1 but activators for PP2A (50). Most definitively, I2, a specific and diagnostic protein inhibitor of PP1, inhibited the activities of both the MT-associated phosphatase and PP1c to Ͼ90% (Fig. 2C). Taken together, the above results clearly indicate that the protein phosphatase that tightly associates with MTs through multiple cycles of assembly/disassembly is exclusively a PP1-type protein phosphatase. We therefore named it PP1 MT    I Association of protein phosphatase activity with MTs during assembly/disassembly cycles Following preparation of a brain extract (HSS), MTs were cycled four times between an assembled polymeric state and a disassembled soluble state. P1-P4 are sequential MT pellets that were resolubilized prior to assay. Protein concentration in each fraction was measured, and total protein was from one preparation using six calf brains. Equal amounts of protein from each fraction were assayed for phosphatase activity as described under "Experimental Procedures." One unit of activity is equal to 1 nmol of phosphate released per min. Similar results were observed in three separate preparations. There are three possible explanations for how PP1 MT associates with MTs. One is that PP1c binds to MTs via a MTtargeting subunit(s). Another is that some modification may regulate the MT binding ability of PP1c, and this is not exhibited by the recombinant PP1c. The third is that PP1 MT contains a specific PP1c isoform that differs from the recombinant PP1c␣, which was used in the experiment, and this isoform can bind to MTs directly.
To test these possibilities and determine whether PP1 MT exists as a protein complex consisting of PP1c and other proteins, both PP1c and PP1 MT were fractionated by Sepharose CL-6B gel filtration. Although PP1c migrated as expected as a protein of ϳ40 kDa (peak fraction 72), the PP1 MT activity migrated predominantly as a single peak (peak fraction 62) at a position corresponding to a globular protein of 400 -600 kDa (Fig. 3A). SDS-PAGE of the fractions eluted from the column showed that MAP1 and MAP2 eluted before PP1 MT , whereas tau proteins eluted at a similar position of 400 -600 kDa (Fig.  3B). The identity of these proteins was confirmed by Western blots with specific antibodies (data not shown). These results suggest that PP1 MT is a complex between PP1c and other proteins, which may be tau but probably not MAP1 or MAP2. The possibility that PP1 MT is composed of PP1c and targeting protein subunits was tested further below.
Dissociation of the Putative MT-targeting Subunit from PP1c-To test the possibility that PP1 MT is composed of PP1c and targeting protein subunits, the PP1 MT in a crude MAP fraction was treated with 2 M NaCl followed by Sepharose CL-6B gel filtration in the presence of 2 M NaCl. As shown in Fig. 3C, the majority of the PP1 activity eluted at the same position as PP1c, indicating that 2 M NaCl may have dissociated PP1c from other proteins in the PP1 MT complex. To test this, PP1 in peak fraction 72 from chromatography in high salt was tested after dialysis to see if it cosedimented with MTs. As a control, the PP1 MT in peak fraction 62 from chromatography in low salt was also tested. The results in Table III show that although the intact PP1 MT in fraction 62 (low salt) quantitatively bound to MTs, the PP1 in fraction 72 (high salt) did not bind to MTs. This suggests that the high salt treatment dissociated PP1c from a MT-targeting subunit. Furthermore, when dissociated PP1 in fraction 72 (high salt) was added to fraction 62 (high salt), which may still contain the putative MT-targeting subunit, the binding of PP1 to MTs was restored. These results strongly support the existence of a MT-targeting subunit in PP1 MT and show that the PP1c subunit can be dissociated from its MT-targeting subunit by interfering with ionic interactions.
Identification of Column Fractions Containing MT-targeting Subunits-To identify the MT-targeting subunit, we developed an assay to test the ability of fractions from Sepharose CL-6B (eluted with high salt) to reconstitute the binding of recombinant PP1c to MTs. When we added recombinant PP1c to an equivalent amount of protein from each of the CL-6B fractions, we detected an increase in the activity of PP1c in some of the fractions (data not shown). Therefore, to compare the relative ability of each fraction to bind PP1c to MTs, we normalized the amount of activity recovered in the MTs pellets by the total (activated) activity present in the incubation. This analysis revealed a major MT-targeting activity in fraction 62, the same position at which the undissociated PP1 MT migrated in low salt (Fig. 4). Lower activity was detected as a shoulder to this main peak (fraction 70). Fractions in the main peak contain mainly FIG. 2. Inhibitor profile of the MT-associated phosphatase. A, effect of OA on the phosphatase activity of brain MT-associated phosphatase. B, effect of CL-A on the phosphatase activity of brain MTassociated phosphatase. Protein (0.1 mg/ml) resolubilized from a MT pellet (P4) cycled four times was incubated with OA or CL-A at different concentrations and assayed for phosphatase activity as described. The activities of purified PP1c (15 ng/ml) in the presence of various concentrations of OA or CL-A were measured similarly. C, effects of PP1 inhibitors on the phosphatase activity of brain MT-associated phosphatase. Resolubilized, fourth cycle MT protein (10 g/ml) or purified PP1c (15 ng/ml) was assayed for phosphatase activity in the presence of various inhibitors at concentrations indicated. tau proteins as revealed by both Ponceau staining and Western blotting (Fig. 5, B and D). This result further demonstrates that PP1c binds to MTs via a MT-targeting subunit(s) and identifies tau as a candidate for this activity. Fractions comprising the major peak were chosen for further study because they had the highest MT-targeting activity and eluted at a position similar to that of endogenous PP1 MT . The Major MT Targeting Activity Is Heat Stable-If tau was responsible for binding PP1c to MTs, then the targeting activity should be heat stable. Accordingly, fraction 62 (from CL-6B chromatography in high salt) was boiled to obtain heat stable proteins. The heat stable proteins in fraction 62 were found to be predominantly tau protein as revealed by Coomassie staining and Western blotting with tau antibody (data not shown). As shown in Table IV, the ability of the heat-treated fraction 62 to bind recombinant PP1c to MTs was similar to that of the untreated fraction 62. Thus, the putative PP1-targeting protein is heat stable. Similar heat stable MT-targeting activity was obtained when total MAPs were heat-treated (data not shown). Because MAP2 and tau are the two major heat stable MAPs in cycled brain MTs and little MAP2 could be detected in fraction 62 (Fig. 5B), this result further suggests that tau is the MTtargeting subunit for PP1c.
Tau Interacts with PP1c in Vitro-To directly identify candidate proteins that can bind to PP1c and to test whether tau exhibits PP1 binding, we performed blot overlays using recombinant PP1c as a probe to detect binding proteins separated by SDS-PAGE and transferred to nitrocellulose. When blot overlays were performed on the fractions from Sepharose CL-6B chromatography in high salt, we detected a group of bands that had apparent molecular weights between 45,000 -66,000 in fractions corresponding to the peak MT-targeting activity (fractions 58 -68 in Fig. 5A). Both Ponceau staining (Fig. 5B) and Western blotting with tau antibody T46, which recognizes all tau proteins (Fig. 5D), demonstrated that these bands are tau proteins. In contrast, no tau bands were detected in the control experiment in which PP1c was omitted from the overlay incu-   . 4. MT targeting activity of fractions from Sepharose CL-6B chromatography in high salt. Equal amounts of recombinant PP1c were incubated with equal amounts of protein from each CL-6B fraction and then tested for MT binding using a cosedimentation assay. Activation of PP1c activity by the column fractions was assessed independently (see "Experimental Procedures"). To compare the relative ability of each fraction of the CL-6B column to target PP1c to MTs, the PP1c activity in the MT pellet was normalized by dividing the activity by the fold activation and then plotting the result as relative activity. Shown here is a representative determination. Similar results were obtained in three independent experiments. bation (Fig. 5C). Also, many proteins appearing on the blots (e.g. MAP1 and MAP2 heavy chains) did not show appreciable PP1 binding activity. These results clearly show that tau protein can specifically interact with PP1c in vitro. In addition, we observed specific reactivity to a distinct band of ϳ50 kDa in the shoulder peak (fraction 70) to the main MT-targeting peak at fraction 62. The identity of the 50 kDa band is unknown. Finally, in fractions 34 -50, which contained lower levels of MT targeting activity (Fig. 4), the blot overlay revealed 3-4 bands varying from 31-43 kDa (Fig. 5A). The identities of these proteins are unknown, but two of these bands may be MAP1 light chains, because they have the appropriate size (31 and 33 kDa) and elute with the MAP1 heavy chains.
Recombinant Tau Can Target PP1 to MTs-To further demonstrate that tau protein is a MT-targeting subunit of PP1 MT , recombinant tau expressed in Sf9 insect cells was tested in a MT-targeting assay. Because there is a 3-4-fold activation of PP1 activity by recombinant tau (data not shown), we converted the amount of MT-bound PP1 activity to a percentage of the total PP1 activity measured in the binding assay. As shown in Fig. 6, recombinant tau protein quantitatively bound PP1c to MTs (40% of PP1c tested bound to MTs in the experiment), whereas PP1c alone or a mixture of tau and PP1c did not pellet by themselves. The lower percentage of PP1c brought down with MTs by recombinant tau compared with an equivalent level of tau from the Sepharose CL-6B columns (Table III and IV) may be because of modifications of tau in Sf9 cells that reduced its activity. Nevertheless, the ability of recombinant tau to bind PP1c to MTs confirms that tau is a MT-targeting subunit of PP1 MT .

TABLE IV
The major MT targeting activity is stable to heat treatment Sepharose CL-6B fraction 62 was boiled in the presence of salt and tubulin as described (43), and the soluble supernatant was dialyzed in PEM buffer. Similar amounts of fraction 62 before and after boiling treatment were incubated with same amount of PP1c and tested in a MT targeting activity assay. The percentage of activity in the supernatant or pellet was calculated against total PP1 activity that was used in each assay. Similar results were obtained in three independent experiments.

Reconstitution of PP1 MT Complex by Recombinant PP1c and
Tau-To test whether PP1c and tau can form a stable protein complex similar to the endogenous PP1 MT , excess recombinant tau was incubated with recombinant PP1c and applied to a Sepharose CL-6B column in low salt conditions. The elution of recombinant PP1c was monitored by phosphatase activity. As a control, the same amount of recombinant PP1c alone was chromatographed on Sepharose CL-6B. As shown in Fig. 7, although PP1c alone migrated at a position corresponding to a single protein of ϳ40 kDa as previously shown (Fig. 3A), the PP1c-tau mixture eluted as a major peak at ϳ400 kDa. This is the same position (fraction 62) as that of the endogenous PP1 MT complex from crude MAPs (compare Fig. 7 with Fig.  3A). There were several minor peaks of PP1c activity, which may be because of a small amount of PP1c associated with aggregated tau that eluted earlier in the column (fractions 50 -58). Thus, PP1c and tau can form a stable complex with the same gel filtration characteristics as PP1 MT , confirming that the PP1 MT complex is composed of PP1c and tau. DISCUSSION In this paper, we identified a protein phosphatase activity that tightly associates with brain MTs through multiple rounds of MT assembly/disassembly, a distinguishing characteristic of MAPs (51). We showed that the MT-associated phosphatase was a member of type 1 protein phosphatases based on its reactivity with PP1 specific antibodies, its activity characteristics and its inhibitor profile. Based upon these characteristics, we have named this phosphatase, PP1 MT . PP1 MT was found to be a large protein of 400 -600 kDa by gel filtration chromatography, suggesting that it may be a complex of PP1c and associated subunits that target PP1c to MTs. By using a MT targeting assay, we showed that the associated subunits were responsible for binding PP1 MT to MTs and that the major associated subunit was the MAP tau. We further demonstrated the MT targeting ability of tau for PP1 by showing that purified recombinant tau reconstituted the binding of PP1c to MTs and reconstituted a protein complex with PP1c that eluted from gel filtration columns at the same molecular size as endogenous PP1 MT . Taken together, our results show that PP1 MT is a novel PP1 enzyme that is tightly associated with MTs during their biochemical isolation and is composed principally of PP1c and tau.
Several lines of evidence support our conclusion that the association of PP1c with tau reflects the fact that tau is a targeting subunit for PP1c rather than merely a substrate. First, we showed that tau and PP1c form a stable protein complex on blot overlays and on gel filtration columns, whereas protein phosphatases are not known to form stable complexes with their substrates (26 -28). Second, the PP1 activity in PP1 MT (or in the complex with recombinant tau) was increased compared with PP1c; if tau was simply a substrate of PP1c, the activity of the complex would be reduced. In the MT-targeting activity assay, certain fractions (including those containing tau) increased the activity of the added recombinant PP1c 3-4-fold. Similarly, PP1 activity seemed to be stimulated in the PP1 MT complex relative to free PP1c, because we recovered 3-4-fold less activity after high salt dissociation of PP1 MT (data not shown). This suggests that tau stimulates the phosphorylase phosphatase activity of PP1c, which is a characteristic of a regulatory interaction of PP1c with a targeting subunit rather than with a substrate. Further experiments will be needed to determine whether PP1 MT has elevated phosphatase activity toward its suspected endogenous substrates, such as MAP1, MAP2, and tau. Nevertheless, these results strongly suggest that tau is a novel member of the growing family of PP1targeting subunits. The PP1-targeting family includes the well characterized glycogen-targeting subunit (52) and myosin-targeting subunits (53) as well as the more recently described p53 binding protein p53BP2 (54), nuclear localization subunit sds22 (55), yeast glucose repressor REG1 (56), and other proteins such as pRB (57), NIPP-1 (58), L5 (59), and RIPP-1 (60). Many of these proteins have a consensus binding domain, (K/R)(V/I)XF, that may be responsible for the interaction with PP1c (61), although other domains (VXF or VXW) are also thought to be important for the interaction (62). Tau does not have one of these consensus sequences, although it does have three sequences that resemble the consensus sequence but are missing the V/I at the second position (i.e. (K/R)XXF). It will be interesting to determine whether these sequences are involved in the tau-PP1 interaction.
Our discovery of the interaction of PP1 with tau and the association of the PP1 MT complex with MTs is the first molecular interaction described for tau in which tau forms a stable complex and maintains its binding to MTs. Other proteins are known to bind to tau, including PP2A (63) and GAPDH (64); however, these interactions are not known to result in the binding of the proteins to MTs. In fact, recent studies have shown that PP2A binds to the MT-binding repeat region of tau and prevents tau (or PP2A) from binding to MTs. 2 We only detected PP2A in the early cycles of our MT preparation, suggesting that it is either only weakly associated with MTs under our conditions or that it is associated with a contaminant that is lost during subsequent purification of the MTs.
Our identification of tau as a MT-targeting subunit of PP1 suggests a novel function for tau, in addition to its established role of stimulating MT polymerization and stabilizing MTs against disassembly (1)(2)(3)(4)65). The association of PP1 with MTs may lead to dephosphorylation of MAPs, which typically enhances their association with MTs and promotes MT stability (1)(2)(3)(4). Given the association of PP1 with tau, it is highly possible that tau itself is a PP1 substrate and that once PP1 is anchored to MTs by tau, it could actively dephosphorylate other tau molecules. This in turn would increase the affinity of tau for MTs and further increase MT stability. Thus PP1 may play an important role as the phosphatase capable of maintaining low levels of phosphorylation on tau and perhaps other MAPs 2 G. S. Bloom, personal communication. while they are bound to MTs. This may contrast the role played by PP2A, which binds to tau when it is not on the MTs. PP2A could instead serve as a cytosolic tau phosphatase, dephosphorylating soluble tau so that it could rebind to MTs. Having two phosphatases regulate tau phosphorylation may also allow far more complex regulation of the tau-MT interaction because these phosphatases are regulated differently by cellular signal transduction pathways. MAP2 and MAP4 have been shown to target kinases (or their regulatory proteins) to MTs (see the Introduction). Our identification of tau as a MT-targeting subunit for PP1 represents the first MAP identified that is capable of directing a protein phosphatase to MTs. This raises an interesting possibility that specific MAPs may interact with different classes of regulatory molecules: thus, tau may be a phosphatase-targeting MAP, whereas MAP2 and MAP4 may be kinase-targeting MAPs. By adjusting the level of the two types of MAPs on MTs, the level of phosphorylation of other MT interacting proteins would be regulated. It will be interesting to determine whether other MAPs, such as MAP1, X-MAP, and E-MAP also direct kinases or phosphatases to MTs.
The importance of phosphatases in regulating MAP phosphorylation and MT stability also implies that abnormal phosphorylation of MAPs caused by dysfunction of either protein kinases or phosphatases may alter MT functions and thus contribute to disease. Evidence from studies of AD has shown that hyperphosphorylated tau accumulates to form insoluble PHFs. PHFs are a major component of neurofibrillary tangles, which together with senile plaques represent the pathological hallmarks of AD (15, 16). Thus, studying factors that control tau phosphorylation in normal neurons and their potential alteration in AD may help to elucidate PHF formation and possibly the causes of AD (17,66). Recently, it has been found that protein phosphatase activities may be greatly reduced in AD brain, and all four major phosphatases have been implicated in dephosphorylating some of the sites on PHF tau in one or more in vitro studies (18,47). Our discovery that tau is a MT-targeting subunit of PP1 indicates that PP1 may be an important endogenous tau phosphatase. It will be important to determine whether tau is a substrate of PP1 MT in vivo and whether PP1 MT is active toward abnormally phosphorylated PHF tau.