Characterization of Neuronal Tau Protein as a Target of Extracellular Signal-regulated Kinase*

Tau neuronal protein has a central role in neurodegeneration and is implicated in Alzheimer disease development. Abnormal phosphorylation of Tau impairs its interaction with other proteins and is associated with its dysregulation in pathological conditions. Molecular mechanisms leading to hyperphosphorylation of Tau in pathological conditions are unknown. Here, we characterize phosphorylation of Tau by extracellular-regulated kinase (ERK2), a mitogen-activated kinase (MAPK) that responds to extracellular signals. Analysis of in vitro phosphorylated Tau by activated recombinant ERK2 with nuclear magnetic resonance spectroscopy (NMR) reveals phosphorylation of 15 Ser/Thr sites. In vitro phosphorylation of Tau using rat brain extract and subsequent NMR analysis identifies the same sites. Phosphorylation with rat brain extract is known to transform Tau into an Alzheimer disease-like state. Our results indicate that phosphorylation of Tau by ERK2 alone is sufficient to produce the same characteristics. We further investigate the mechanism of ERK2 phosphorylation of Tau. Kinases are known to recognize their protein substrates not only by their specificity for a targeted Ser or Thr phosphorylation site but also by binding to linear-peptide motifs called docking sites. We identify two main ERK2 docking sites in Tau sequence using NMR. Our results suggest that ERK2 dysregulation in Alzheimer disease could lead to abnormal phosphorylation of Tau resulting in the pathology of the disease.

Tau is an intrinsically disordered protein whose primary sequence is divided into several functional domains: an N-terminal region, a proline-rich domain (PRD), 2 a microtubule binding domain (MTBD) constituted of partially repeated sequences R1 to R4, and a C-terminal region (see Fig. 1A). Both PRD and MTBD are involved in microtubule stabilizing activity of Tau (1,2).
Phosphorylation of Tau is an essential process to regulate its physiological function(s); however, abnormal phosphorylation of Tau has been linked to neuronal dysfunction (3)(4)(5). In Alzheimer disease (AD) Tau is hyperphosphorylated and aggregated (6). The longest Tau protein isoform (441 residues) has 80 threonine (Thr) or serine (Ser) residues. These residues are exposed as Tau is an intrinsically disordered protein subject to modification by numerous kinases (7). Mass spectrometry (MS) analyses have identified ϳ45 phosphorylated sites on Tau aggregates extracted from AD patients with a typical paired helical filament (PHF) morphology (8,9) compared with 15-30 phosphorylation sites in soluble Tau extracted from mice (10) or normal human brain (11). Monoclonal antibodies such as AT8 (recognizing Ser(P)-202/Thr(P)-205; Ref. 12) or PHF1 (Ser(P)-396/Ser(P)-404; Ref. 13) are often used to detect abnormal phosphorylation of Tau (5). Proline-directed kinases with a (S/T)P phosphorylation consensus motif, such as cyclin-dependent kinase 5 (CDK5; Refs. 14 and 15) with its activator protein p25, glycogen synthase kinase (GSK3␤; Ref. 16), stress-activated protein kinases (JNK; Ref. 17), and p38 (18) and extracellular-signal-regulated kinase (ERK1/2; Ref. 19), are considered as potential therapeutic targets to prevent Tau hyperphosphorylation (20). A clear understanding of the mechanism of abnormal phosphorylation of Tau leading to neuronal dysregulation is still lacking. Characterization of the kinases involved in the phosphorylation of Tau and the related specific patterns of phosphorylation are, therefore, of interest to develop new strategies to counteract the pathological processes.
ERK2 belongs to the mitogen-activated protein kinase (MAPKs) family and is activated by dual phosphorylation on a Thr-Xaa-Tyr sequence in its activation loop by MAP kinase kinases (MKK) (21)(22)(23). Human ERK2 is phosphorylated on Thr-183 and Tyr-185 by MEK kinase. MAP kinases have a classic kinase structure with an N-terminal lobe folded as a ␤ sheet associated to a larger C-terminal domain constituted of ␣-helices (24). The activation loop is located at the hinge between these domains and shows a large conformational change upon phosphorylation (25). Mechanisms regulating specificity and efficiency of kinases are not fully understood but involve the binding of kinase docking grooves to binding motifs on their substrates, made up of a linear peptide (26,27). Interaction regions distinct from the catalytic pocket were found to be involved in the recognition of both ERK1/2 upstream regulators and downstream substrates (28). The ERK docking site is targeted by proteins that contain a linear motif called the D-recruitment site (DRS or D-site), defined by a loose consensus of ϳ10 amino acid residues ⌿ 1-3 X 3-7 ⌽X ⌽ (⌿, ⌽, and X refer to positively charged, hydrophobic, or any intervening residues, respectively, and subscripts refer to the number of residues (30)). Interactions of the conserved D-recruitment site motifs with the MAP kinase docking site has been characterized at the atomic level in complexes of ERK2 with peptide fragments from several protein partners (30 -33). In addition, another docking motif was identified, called F-recruitment site (FRS or F-site) that recognizes a Phe-Xaa-Phe motif (34). ERK1/2 activation in AD brain is well documented (35)(36)(37). Results from previous studies indicate that ERK1/2 is able to transform Tau into an AD-like state, but this is mainly documented by the detection of the epitope recognized by the AT8 antibody (19). To better define the phosphorylation sites involved in the transformation of Tau as present in AD (19), we used NMR spectroscopy. We observed that 14 out of the 17 S/T sites, which are followed by proline ((S/T)P sites) and likely to be recognized by proline-directed ERK1/2, are phosphorylated by activated recombinant ERK2 in vitro. Another well known in vitro model of Tau hyperphosphorylation is obtained by incubation of Tau with rat brain extracts along with phosphatase inhibitor okadaic acid (38). We compared the phosphorylation pattern of Tau obtained using rat brain extract with that obtained using the in vitro activated recombinant ERK2. A similar pattern of Tau phosphorylation was indeed observed, which strongly suggests that ERK2 is sufficient for hyperphosphorylation of Tau. We then studied the Tau-ERK2 interaction. We mapped the interaction sites of ERK2 along the Tau sequence and identified two docking sites in the MTBD, which is distal to the PRD containing most of the ERK phosphorylated sites of Tau. We, therefore, conclude that Tau and ERK2 form a dynamic complex. Our study indicates that Tau could be the first example of an ERK2 substrate with multiple docking sites. These conclusions expand our view on the potential role of ERK2 in the neurodegeneration pathway in AD involving Tau and shed light on the molecular mechanisms leading to pathological Tau formation.
Recombinant His 6 -tagged p42 MAP kinase (ERK2) from Xenopus laevis was prepared by growing bacteria transformed with a ERK recombinant T7 expression plasmid in LB medium until induction by 1 mM isopropyl 1-thio-␤-D-galactopyranoside of the protein production for a period of 4 h at 37°C. His-ERK2 was purified using nickel-nitrilotriacetic acid-based affinity chromatography following the manufacturer's protocols. After analysis on SDS-PAGE, the pooled fractions were buffer-exchanged using a PD-10 column (GE Healthcare) against conservation buffer 50 mM Tris-HCl, 150 mM NaCl, 0.1 mM EGTA, 50% glycerol, pH 7.5, and kept frozen at Ϫ80°C. His 6 -tagged D321N/D324N ERK2 was prepared in the same manner. Activated-ERK2 was prepared as previously described by in vitro incubation with activated MEK (40).
To label Tau with CF3 attached on its native Cys residues (Cys-291 and Cys-322), 100 M [ 15 N]Tau or [ 15 N]Tau-(244 -372) in 50 mM ammonium bicarbonate were first thoroughly reduced by incubation for 2 h at 22°C with 1 mM Tris (hydroxymethyl) phosphine reducing agent. The protein samples were next incubated with 2 mM 2-iodo-N-trifluoroethyl acetamide (EnamineStore) in dimethyl sulfoxide at 22°C for 3 h. The excess of 2-iodo-N-trifluoroethyl acetamide was removed by desalting. The CF 3 -[ 15 15 N-double-labeled Tau was incubated in the same conditions overnight with the activated recombinant ERK2. Enzymatic incubation was terminated by heating the reaction mixture for 15 min at 75°C followed by centrifugation. The phosphorylation mixture was buffer-exchanged using desalting centrifugal devices (0.5-mlbed G25 resin, cutoff of 7 kDa, Thermo Scientific Zeba Desalting Columns) against NMR buffer (50 mM deuterated Tris, pH 6.65, 25 mM NaCl, 2.5 mM EDTA, 1 mM DTT and 10% D 2 O). Phosphorylation of the Tau-(165-245) by ERK2 followed the same protocol. For the time course of phosphorylation, 100 M [ 15 N]Tau was incubated for 15, 30, 60, 120, and 180 min with 1 M activated-ERK2 as described above. The reaction was stopped by heating the sample at 75°C for 15 min. The experiment was repeated three times.
The rat brain extract was prepared by homogenizing a brain (ϳ2 g) in 5 ml of homogenizing buffer (10 mM Tris⅐Cl, pH 7.4, 5 mM EGTA, 2 mM DTT, 1 M okadaic acid (Sigma) supplemented with 20 g/ml leupeptin and 40 mM Pefabloc (38). Ultracentrifugation was next performed at 100,000 ϫ g for 1 h at 4°C. The supernatant was directly used for its kinase activity. Total protein concentration was estimated to be 7 mg/ml by the Bradford colorimetric assay. The [ 15 N]Tau protein (1-1.5 mg) was dissolved at 10 M in 2.5 ml of phosphorylation buffer (2 mM ATP, 40 mM Hepes⅐KOH, pH 7.3, 2 mM MgCl 2 , 5 mM EGTA, 2 mM DTT complemented with a protease inhibitor mixture (Roche Applied Science) and 1 M okadaic acid (Sigma). The phosphorylation reaction was performed at 37°C for 24 h with 500 l of brain extract. Enzymatic incubation was terminated by heating the reaction mixture for 15 min at 75°C followed by centrifugation. Phosphorylation mixture was buffer-exchanged against 50 mM ammonium bicarbonate before lyophilization. Phosphorylation of the Tau-(165-245) by rat brain extract kinase activity followed the same protocol. The samples were prepared by solubilizing the lyophilized powder into NMR buffer. Each set of experiment was independently repeated two or three times.
NMR Spectroscopy-Analysis of the phospho-Tau samples by NMR spectroscopy was performed at 293 K or 298 K on Bruker 600-MHz and 900-MHz spectrometers equipped with a triple resonance cryogenic probehead (Bruker, Karlsruhe, Germany). Trimethylsilyl propionate was used as the reference (0 ppm). 100 M [ 15 N]Tau in a volume of 200 l in 3-mm tubes was sufficient to obtain the two-dimensional 1 H, 15 N heteronuclear single quantum correlation (HSQC) spectra with 32 scans. HSQC two-dimensional experiments to map the interaction between Tau protein and recombinant ERK2 were recorded at 293 K on a 900-MHz spectrometer with 64 scans per increment as a complex matrix of 2048 ϫ 416 complex points for 14 ϫ 25 ppm spectral widths in the proton and nitrogen dimensions, respectively. Three-dimensional HNCACB spectra were recorded for 13 C, 15 N-phospho-Tau obtained both with rat brain extract and recombinant ERK2 with a standard Bruker pulse sequence. A three-dimensional HNcaNNH spectrum (41) was additionally used in the assignment. Three-dimensional HNCACB experiments were acquired to confirm the assignment of the Tau residues in 13 C, 15 N-His-SUMO Tau-(220 -240) and 13 19 F spectra were recorded at 293 K with 1024 scans on a Bruker 600MHz Avance NMR spectrom-eter equipped with a 5-mm CPQCI 19 F/ 1 H-13 C/ 15 N cryogenic probehead (Bruker). The chemical shifts in one-dimensional 19 F spectra were referenced by using the fluor resonance of external Trifluoroacetic acid at Ϫ76.55 ppm.
NMR Data Processing-Spectra were processed with Bruker TopSpin 3.1 software. Data analysis, peak picking, and calculation of peak volumes were done with Sparky 3.114 software (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco). Assignment of resonances of Thr(P)-and Ser(P)-phosphorylated residues is based on the typical C␣ and C␤ chemical shifts of these residues (42). Ser(P) and Thr(P) resonances were assigned to a specific residues in phospho-Tau primary sequence using as additional information the 15 N C␣ and C␤ chemical shifts of their N-terminal residues (43-45) ( Table 1). Resonance integration was performed in two-dimensional spectra by summing all the intensities over a box enclosing the peak of interest. To account for variations between spectra, integration volumes were normalized to the integration value of the resonance of Glu-62 residue, which is isolated in the spectrum and not influence by phosphorylation(s). To estimate the level of phosphorylation at each site, we first estimated the level of phosphorylation on Thr-153 residue using the resonances of Ala-152. Ala-152 has two distinct resonances in the two-dimensional 1 H, 15 N spectrum depending on the phosphorylation status of Thr-153 residue. The ratio of the integration volumes of these Ala-152 resonances gives an estimation of the phosphorylation degree of Thr-153. We next used this value to correlate integration volumes of resonances with phosphorylation level of the corresponding phospho-residue.
In Vitro Aggregation Assay-Tau aggregation assays were performed at 0.25 mg/ml (5 M) Tau (incubated overnight with activated recombinant ERK2 kinase in absence of ATP) or phosphorylated Tau in MES buffer consisting of 100 mM MES, pH 6.9, 2 mM EGTA, 1 mM MgCl 2 , and 0.3 mM freshly prepared DTT. The samples were incubated at 35°C with shaking at 300 rpm. After 2 h of incubation, the insoluble fraction was collected by centrifugation at 16,500 ϫ g, and the pellet was suspended in 50 l.
10-l samples were applied on 400-mesh hexagonal Formvar-coated copper grids for 45 s. The sample-loaded grids were washed twice by ion-free water and drained. The grids were next negatively stained with 2% urinyl acetate solution for 60 s. The observation was performed by transmission electron microscopy (Hitachi H7500 at 80 kV and JEOL JEM-2100 at 200 kV. Interaction of ERK2 with Tau, Tau Fragments, and Tau Peptides-ERK2 buffer exchange after storage was done using desalting centrifugal devices (1.5-ml-bed of G25 resin, cutoff of 7 kDa; Thermo Scientific Zeba Desalting Columns). ERK2 protein in Tris NMR buffer was then concentrated to a volume of 200 l using a centrifugal concentrator (0.5 ml, 30 K membrane, Amicon Ultra) up to 250 M. The concentrated enzyme solution was directly used to suspend lyophilized [ 15

Results
Identification of Tau Amino Acids Phosphorylated in Vitro by ERK2-We used NMR spectroscopy to identify phosphorylation sites of Tau modified in vitro by activated ERK2. On incubation with MEK-activated recombinant ERK2 (double-phosphorylated on Thr-183/Tyr-185), several additional resonances were observed in the 1 H, 15 N two-dimensional spectrum for the resulting phospho-Tau compared with the spectrum of the unphosphorylated Tau (compare Fig. 1, B and C). These resonances are typical for the HN amide resonances of Ser(P) and Thr(P) residues (42,44), indicating that Tau is indeed phosphorylated at many sites by ERK2. Phosphorylation sites were identified using three-dimensional triple resonance NMR spectroscopy on a 13 C, 15 N-labeled Tau sample phosphorylated in vitro by activated ERK2. Of the 17 (S/T)P sites, 14 sites along the Tau sequence were found to be phosphorylated by ERK2 (Fig. 2, A  and B; Table 1). ERK2 phosphorylated Thr-50, Thr-153, Thr-175, Thr-181, Thr-205, Thr-231, Ser-235, Ser-404, and Ser-422 with a stoichiometry of Ͼ50%; estimation was based on peak integration (45) (Fig. 2B, Table 1). In addition, a minor phosphorylation was observed for Ser-191, which is not a Pro-directed site (Fig. 2, Table 1). Phosphorylation by ERK2 of a Tau-(165-245) fragment encompassing the PRD of Tau allowed confirmation of the identification of the phosphorylation sites located in this domain (Fig. 2C). A comparison with previous assignments of the NMR amide cross-peaks corresponding to Tau proline-directed phosphorylation sites, generated by cyclin-dependent kinase 2/CycA3 or glycogen synthase kinase 3 (GSK3) (46,47), further confirmed identification of the ERK2 phosphorylation sites located in the PRD and C-terminal region. Kinetic experiments were next performed to define the order of the multiple phosphorylations of Tau by recombinant ERK2. SDS-PAGE analysis of Tau during a time-course of phosphorylation by activated ERK2 showed a gradual decreased mobility associated with a global increase of Tau phosphorylation level (Fig. 3A). A series of two-dimensional spectra was then acquired corresponding to various incubation times to get insight at site-specific time-dependent modification of Tau by ERK2. The level of modification at each identified phosphorylation site was estimated by the integral of the corresponding resonance. The fastest observed phosphorylation was found for Ser-235, Ser-404, and Ser-422 followed by Thr-50, Thr-181, and Thr-205 (Fig. 3, B-D). The time-dependent phosphorylation increase for these residues can be fitted with an exponential function, with rate constants ranging from 10 to 85 min. Slower modification was observed for residues Ser-46, Thr-69, Thr-153, Thr-175, Ser-191, Ser-199, Thr-231, and Ser-396, which showed a linear trend of phosphorylation increase during the 3-h time course of Tau incubation with activated-ERK2.
Comparison of ERK2-phosphorylated Tau with a Hyperphosphorylation Model of Tau-Phosphorylation of S/T residues corresponding to 12-15 sites has similarly been reported after incubation of Tau with rat brain extract (12,38,48). The sitespecific, precise pattern of the hyperphosphorylation of Tau by rat brain extract was, however, never investigated. We, therefore, used NMR spectroscopy to establish the map of Tau amino acid residues phosphorylated by kinases in rat brain extract and compared with those phosphorylated by ERK2. Rat brain extracts were used to in vitro phosphorylate recombinant 15 N-labeled Tau. The corresponding 1 H, 15 N two-dimensional spectrum of the phosphorylated Tau showed additional resonances in the region for Thr(P) and Ser(P) residues compared with the spectrum of the unphosphorylated Tau (Fig. 1, com- pare panels D and B). Comparison of two-dimensional spectra of Tau phosphorylated by either activated ERK2 or rat brain extracts showed that many resonances assigned to the ERK2 phosphorylated sites have a match in the brain extract phos-phorylated Tau (Fig. 2, A and B). To confirm that the similarity in the phospho-Tau spectra reflects a shared phosphorylation pattern, phosphorylation sites were identified using three-dimensional triple resonance NMR spectroscopy on a 13 C, 15 Nlabeled Tau phosphorylated in vitro by rat brain extract. The major modified sites of Tau correspond to Ser(P)-46, Thr(P)-50, Thr(P)-153, Thr(P)-175, Thr(P)-181, Ser(P)-202, Thr(P)-205, Ser(P)-208, Thr(P)-231, Ser(P)-262, Ser(P)-356, Ser(P)-396, Ser(P)-404, and Ser(P)-422 (Fig. 2). This analysis confirmed that the overlapping resonances in the two-dimensional spectra indeed correspond to the same phosphorylated residues. The level of phosphorylation for these residues is also similar between Tau phosphorylated by the rat brain extract or the activated ERK2 (Fig. 2B). A few additional Tau phosphorylation sites are also identified exclusively from the incubation with the rat brain extract including Ser(P)-208, Ser(P)-262, and Ser(P)-356 that are not proline-directed sites (Fig. 2B). Phosphorylation with the brain extract of Tau-(165-245) fragment, corresponding to the PRD, resulted in detection in the corresponding two-dimensional 1 H, 15 N HSQC of resonances corresponding to phosphorylated residues. These phosphorylated sites matched those found in the PRD embedded in the fulllength phosphorylated Tau protein (Fig. 2D), confirming the assignment. The data showed that the pattern of phosphoryla-FIGURE 2. Comparison of Tau phosphorylation by activated recombinant ERK2 and rat brain extracts. A, overlaid details of 1 H, 15 N HSQC two-dimensional spectra of phosphorylated Tau by activated ERK2 (in red) or by rat brain extract (in black). The enlarged region corresponds to the boxed region in Fig. 1. C and D, resonances corresponding to assigned phosphorylated Tau residues are annotated. B, level of phosphorylation for each Ser(P)-/Thr(P)-Tau residue presented as black bars for Tau phosphorylated by the rat brain extract and red bars by activated ERK2. Each experiment was repeated twice. S.D. around average value is shown by error bars. The level of phosphorylation is estimated based on the integrals of resonance peaks as described under "Experimental Procedures." C and D, 1 H, 15 N HSQC two-dimensional spectrum of Tau-(165-245) modified by incubation at 37°C for 3 h with activated recombinant ERK2 (C) or overnight with rat brain extract (D). The annotated resonances were assigned to phosphorylation sites by comparison with the 1 H, 15 N HSQC two-dimensional spectrum of Tau protein phosphorylated by activated recombinant ERK2 or rat brain extract in the same conditions.

TABLE 1
Sequence specific assignment of resonances corresponding to phosphorylated residues in ERK phospho-Tau 1 H, 15 N, CA, and CB columns correspond to the chemical shifts (in ppm) of atoms from the assigned phospho-residues indicated in the first column. The amino acid type at the direct N terminus of each phospho-residue is displayed in column i-1 residue type and the chemical shifts of the corresponding CA and CB atoms in CA-1 and CB-1 columns. tion obtained solely by ERK2 activity is similar to the one of the hyperphosphorylated Tau obtained by incubation with rat brain extract. Aggregation Propensity of Tau Phosphorylated by Rat Brain Extracts or Activated ERK2-Tau phosphorylated by rat brain extract was shown to be able to self-assemble without the addition of an exogenous compound such as heparin (48). Given the similarity in the phosphorylation pattern of Tau phosphorylated by the rat brain extract or activated ERK2, we next evaluated the aggregation propensity of the later. Phosphorylated Tau samples by either rat brain extract or ERK2 were incubated at 35°C, and the insoluble fraction of the sample was observed by electron microscopy. In both cases fibers of phospho-Tau were observed already after a 2-h incubation (Fig. 4). Filaments with diameters of 15-20 nm and lengths of several m long were observed in both ERK2 or rat brain extract-phosphorylated Tau-aggregated samples. The fibers showed the typical helical appearance reported for PHF (49,50). (Fig. 4, A and C). The aggregation is specific for the phosphorylated Tau samples as only rare isolated oligomers can be found in the control Tau sample incubated at 35°C (Fig. 4E). However, the aggregated phosphoprotein only corresponded to a small fraction of the incubated samples and was, therefore, not positive for thioflavin detection, a fluorophore that detects amyloid fibril formation. We conclude that ERK2 by itself can generate a phosphorylated protein whose aggregation properties resemble those of the rat brain-phosphorylated protein. However, under our experimental conditions a robust aggregation of the rat brain extract-generated phospho-Tau was not observed.
Identification of the ERK2 Docking Site(s) on the Tau Protein-MAP kinases recognize their substrates not only by the phosphorylation (S/T)P motifs but also by docking motifs. We thus next investigated the mechanism of Tau recognition by ERK2. To map potential docking sites, two-dimensional spectra of [ 15 N]Tau mixed with recombinant ERK2 were compared with a reference spectrum from the [ 15 N]Tau protein free in solution (Fig. 5, A and D). The resonances in these spectra are sensitive to their chemical environment and the local protein dynamics, an interaction affecting their chemical shifts and/or intensities. At a 1:5 [ 15 N]Tau:ERK2 molar ratio, an interaction indeed translates into an important broadening for numerous resonances compared with the reference spectrum (Fig. 5, A  and D). Assignment of the Tau protein resonances has been previously completed by us and others (51,52) and was used to map the interaction region(s) based on the two-dimensional spectra perturbations. Peak intensities were compared for 187 resonances corresponding to residues dispersed along the 441amino acid residue Tau sequence (Fig. 5F). The data showed interaction of ERK2 localized mainly in the MTBD of Tau (Tau-(244 -372); also called K18; Ref. 1). To increase the resolution, the interaction experiment was repeated with Tau fragments, Tau-(165-245), corresponding to the PRD, and Tau-(244 -372), corresponding to the MTBD (Figs. 5 and 6). The addition of ERK2 to [ 15 N]Tau-(165-245) resulted in minor perturbations of the resonances in the corresponding two-dimensional spectrum (Fig. 5B), whereas an interaction of ERK2 with Tau-(244 -372) is clearly observed as the intensity of several resonances in its spectrum is affected (Figs. 5, C and E, and 6A). The intensity of the resonances corresponding to amino acids included in Tau-(274 -288) and Tau-(306 -318) segments strongly decreased in the presence of ERK2 (Fig. 6A, I/I0 Ͻ40%). These ERK2 docking sequences also correspond to, respectively, the PHF6* ( 275 VQIINK 280 ) and PHF6 ( 306 VQI-VYK 311 ) peptides described as nuclei of Tau aggregation (53). Only the sequence of Tau-(274 -288) can accommodate a canonical D-site 274 KVQIINKKLDL 284 (Fig. 6B). Resonance intensities of residues in the Tau-(346 -358) segment also showed a decrease upon ERK2 interaction but to a lesser extent (Fig. 6A, I/I0 below 60%). This region could correspond to a secondary binding site of interaction, although it is not clear if it fits a classical ERK2 docking site.
Because the PHF sequences, respectively, located in R2 and R3 repeated regions of the MTBD share similarities, it is likely that these peptides would have the same binding mode to ERK2. Yet only the PHF6* peptide can be matched to a D-site Data are fitted with a mono-exponential function 1-(e (Ϫt/T0) ) with time t and rate constant T0 in minutes (B). Correlation was calculated between the fit and the average experimental data. D, detail of 1 H, 15 N HSQC two-dimensional spectrum of Tau incubated with activated ERK2 for 60 min. The enlarged region corresponds to the boxed region in Fig. 1, C and D. Resonances corresponding to assigned phosphorylated Tau residues are annotated.
sequence. To verify whether Tau interaction with ERK2 indeed involves a canonical D-site, a mutated ERK2 was used in our NMR binding assay. Replacement of two aspartic residues (D321N/D324N) in the D-site peptide binding pocket of ERK2 by asparagine residues disrupts the recognition of the positively charge residue of the D-site docking peptide. The mutated D321N/D324N ERK2 has consequently a reduced affinity for the D-site peptides (28). However, we did not observe major differences in the NMR intensity profiles of [ 15 N]Tau-(244 -372) in presence of ERK2 or D321N/D324N-mutated ERK2 (Fig. 6A), indicating no major impact of the mutations on the interaction.
Apparent K D values were calculated to further characterize the Tau/ERK2 interaction using fluor NMR spectroscopy (54) to confirm that the main binding is located in the MTBD (Fig.  7). The Tau protein and Tau-(244 -372) (MTBD) were labeled with fluor attached on the native Cys residues Cys-291 and Cys-322 along the sequence (Fig. 6B). Chemical shift value perturbation of the fluor signal along titration of the ERK2 protein on the labeled Tau or Tau fragment showed a saturation behavior typical of an interaction (Fig. 7). The similar apparent K D values of 73 M for Tau and 179 M for Tau-(244 -372) fragment indicated that the main binding indeed occurs within the Tau-(244 -372) segment. The data demonstrate that the main binding sites of Tau for ERK2 are localized in the MTBD of Tau, in the Tau-(244 -372) segment, outside of the regions containing ERK2 phosphorylation sites. The interaction is not mediated by a classical D-site interaction and shows a moderate  affinity similar to the interaction of ERK2 by a minimal docking sequence reported for some of its protein partners (55).
Tau Peptides Interaction with ERK2 Kinase-Two peptides were next chosen to confirm whether they interact individually or not with the ERK2 docking site. The first peptide derived from the Tau sequence, Tau-(220 -240) in the PRD, contains several ERK2 phosphorylation sites but no interaction site for ERK2. On the other hand, Tau-(271-294) contains the Tau-(274 -284) sequence here above defined as an ERK2 docking site. As the Tau-(271-294) peptide showed a limited solubility (PHF6*), fusions with the SUMO protein were used both for the recombinant expression and two-dimensional spectra acquisition. SUMO being a small folded protein, the overlap with the Tau peptide signals was limited and, therefore, allowed a correct analysis of Tau resonances (Fig. 8). Overlay of the twodimensional 1 H, 15 N HSQC of the fusion peptides alone or mixed with ERK2 confirms that the resonance perturbations affected only those resonances corresponding to Tau peptides and not to SUMO fusion protein (Fig. 8). Resonances corresponding to Tau residues in SUMO-Tau-(220 -240) are only slightly affected by the addition of ERK2 (Fig. 8, A and C). To the contrary, the addition of the kinase to the SUMO-Tau-(271-294) affected peak intensities of most of the resonances and principally residues 274 to 284 (Fig. 8, B and D). The peptides Tau-(220 -240) and Tau-(271-294) behaved in their interaction with ERK2 in the same manner embedded in the protein or isolated from their context. The data confirmed the sequence Lys-274 to Leu-284 as a docking site of ERK2. Despite the presence in the Tau-(220 -240) of a predicted D docking site consensus sequence (Fig. 6B), no binding to ERK2 was observed.

Discussion
Using NMR spectroscopy, we identified Tau phosphorylation sites modified in vitro by the activated ERK2 and rat brain extracts. We demonstrate that Tau phosphorylation patterns observed with in vitro phosphorylation by ERK2 and rat brain extracts are similar, with most of the 17 (S/T)P motifs modified. Incorporation of 14 -16 phosphates on (S/T)P sites per Tau molecule was reported by the MAPK activity purified from the brain extract (19). Our previous analysis of Tau phosphorylation patterns obtained in vitro by several other kinases showed that none has the capacity to modify such a large number of phosphorylation sites (43,46,47,56). Phosphorylation by brain extracts is a method to obtain in vitro phosphorylated Tau (12,38) with AD-like characteristics (48) such as a reduced electrophoretic mobility on SDS-PAGE and positive immunodetection of the AT8 epitope (12,38) and the Ser(P)-396/Ser(P)-404 epitope (57). Our work shows that ERK2 by itself also has the potential to modify Tau into a phosphorylation state resembling Tau present in pathological AD state.
Early work by Iqbal and co-workers (48) showed that a Tau protein purified from AD brain or hyperphosphorylated in vitro by rat brain extract on 12-15 sites can form PHF-like aggregates. We observed fibers with a morphology typical of the helical pattern described for the PHF in both the activated ERK2 and rat brain extract-phosphorylated Tau samples incubated at 35°C for 2 h. Tau phosphorylated solely by ERK2 on 15 phosphorylation sites, dispersed along the Tau sequence, behaved similarly as the rat brain extract-phosphorylated Tau in the aggregation assays. This aggregation propensity is specific to the phosphorylated Tau sample as we did not detect any PHFlike structures in the control Tau samples. However, the amount of fibers observed in our samples is small and could only be detected by electron microscopy performed on the pellet of our aggregation samples. Compared with the aggregation induced by the addition of heparin (58), the aggregation of the in vitro phosphorylated Tau affects only a small fraction of the sample. In agreement with this observation, a Tau protein phosphorylated on ϳ10 or ϳ20 phosphorylation sites in insect sf9 cells was recently reported to form oligomers and only a small fraction of fibrils (59). That the in vitro aggregation propensity under near physiological conditions of the in vitro phosphorylated Tau is low does not exclude that it could be a trigger in a cellular context, in which the aggregation process takes place on a long time scale (60). Nevertheless, there are only a few studies here above discussed that have explored the relationship between phosphorylation and aggregation. That phosphorylation is a causative event of Tau aggregation should be stated with more caution in regard of the data presented here by us and recently by others (59).
Our analytical characterization of Tau phosphorylation showed that ERK2 is promiscuous, as most of Tau (S/T)P motifs were phosphorylated. The question of how MAP kinases recognize specific substrates is only partially answered. We investigated the interaction of ERK2 with Tau and identified that it is mediated by two main docking sites. To our knowledge Tau is the first example of ERK2 protein substrate to contain a combination of docking sites. A similar combination of three D-docking sites was described in the regulatory disordered N-terminal region of MKK7 (MAP kinase kinase 7) used for the specific recognition by the JNK MAP kinase (61,62). Another type of a modular system of recognition was previously proposed consisting of a combination of the D-docking site and F-docking site, although in this case it corresponds to two distinct binding grooves of ERK2 (34). Two binding sites on the substrates for one recognition groove of ERK2 allows the formation of a dynamic complex (63), defined as involving more than two transient interfaces between the binding partners. The presence of a high number of interaction sites on a flexible ligand such as Tau, termed allovalency, increases the probability of the rebinding of the protein partner (64,65).
Comparison of the Tau interacting peptides with the degenerate ERK2 recognition sequence ⌿ 1-3 X 3-7 X (with ⌿ for Arg/Lys residues, X for any residue, and for hydrophobic residues, the indices corresponding to the number of residues (29,30) shows that only the first peptide Tau-(274 -284) corresponds to predicted classical D-docking (Fig. 6B). We additionally showed that a mutated ERK2 with two Asp residues replaced by Asn residues in its docking pocket (D321N/D324N) to compromise interaction with the ⌿ residue in the D-peptide (28) still binds Tau-(244 -372) in the same manner as the wild type ERK2 (Fig. 6A). Some docking interactions of ERK2 have been reported to be limited to the hydrophobic pocket in the D-peptide binding site (31).This is the case of the PEA-15 protein, as seen in the crystal structure of PEA15-ERK2 complex. The structure also reveals a C-to-N reverse binding compared with the classical D-peptide interaction. This type of docking interaction is also characterized by a weaker affinity, with a dissociation constant of 18 M reported for the PEA15/ERK2 interaction (31). This value is closer to the 10 M range that we here found for the Tau/ERK2 interaction than the submicromolar range values reported for the canonical docking interaction of ERK2 with D-site peptides (30,55). It was proposed that the minimal interaction mediated by the hydrophobic pocket of ERK2 is a hallmark of moderate affinity regulatory interactions (31). We showed that the Tau-(165-245) fragment does not contain a docking site for ERK2. Nevertheless, phosphorylation of Tau-(165-245) by activated ERK2 showed a similar pattern and level of phosphorylation as with the full-length Tau substrate (Fig. 2). This suggests that the enzymatic activity is not affected by the docking and that the phosphorylation of the PRD might not require the ERK2 docking in the MTBD, at least in the equilibrium conditions of our experiments. The essentiality of docking sites for ERK2 to exert its efficient kinase activity is not established (29,66,67). Kinetic studies demonstrate that docking sites that interact with either the D-recruitment site or the F-recruitment site have little effect on the intrinsic catalytic activity of ERK2 (66). In a proteomic study, a D-docking peptide was found only on 17% of the substrates within 20 amino acids of the phospho site (67). However, a peptide-based study shows to the contrary, that a docking site is necessary to mimic a ERK2 substrate (29). It might be in the case of Tau that the docking sites are crucial to manage formation of regulatory complexes in a cellular context rather than to stimulate activity toward the protein partners (31).
Activation of ERK1/2 is increased in AD neurons (68), is found in association with abnormally phosphorylated early Tau deposits (69), and is linked to the progression of the neurofibrillary degeneration through the Braak stages of AD (35,60). Activation of ERK1/2 also responds to fibrillar amyloid ␤ deposits in mature hippocampal neuronal culture (70) and to increased activity of inositol-trisphosphate 3-kinase B (37) and oxidative stress (71), all affecting Tau phosphorylation. ERK1/2 phosphorylation of Tau could thus take place in cells upon stress signaling or once the phosphorylation fails to be counteracted by an efficient dephosphorylation. ERK is thus accordingly considered as a Tau kinase that could be involved in AD pathophysiology. We have reinforced this view by showing that ERK2 has the capacity by itself to phosphorylate Tau on many sites. These results support the hypothesis that ERK activation under stress conditions might have a detrimental effect for Tau function and participate in AD physiopathology.
Author Contributions-H. Q., S. P., and I. L. conducted most of the experiments, F.-X. C. conducted most NMR data acquisition, H. Q. and I. L. performed NMR data analysis, B. C. prepared the rat brain extract and advised on the manuscript, I. L., S. P., G. L., J. G., and H. Q. wrote the manuscript, and I. L., S. P., G. L., and J. G. conceived the idea for the project. All authors reviewed the results and approved the final version of the manuscript.