Aberrant Tau Phosphorylation by Glycogen Synthase Kinase-3β and JNK3 Induces Oligomeric Tau Fibrils in COS-7 Cells*

Neurofibrillary tangles (NFTs) are found in a wide range of neurodegenerative disorders, including Alzheimer's disease. The major component of NFTs is aberrantly hyperphosphorylated microtubule-associated protein tau. Because appropriate in vivo models have been lacking, the role of tau phosphorylation in NFTs formation has remained elusive. Here, we describe a new model in which adenovirus-mediated gene expression of tau, ΔMEKK, JNK3, and GSK-3β in COS-7 cells produces most of the pathological phosphorylation epitopes of tau including AT100. Furthermore, this co-expression resulted in the formation of tau aggregates having short fibrils that were detergent-insoluble and Thioflavin-S-reactive. These results suggest that aberrant tau phosphorylation by the combination of these kinases may be involved in “pretangle,” oligomeric tau fibril formation in vivo.

Filamentous tau aggregates is the major component of neurofibrillary tangles (NFTs) 1 (1), the most common neuropathological hallmark in several neurodegenerative disorders, including Alzheimer's disease (AD). Discovery of the molecular mechanisms of NFT formation may provide more direct insight into the process of neurodegeneration in AD. NFTs consist of highly phosphorylated microtubule-associated protein tau that assembles to form fibrils with ␤-sheet structures within the cell body and dendrites of neurons (2,3). Several in vitro studies reveal that the repeat domain of tau aggregates more readily than full-length tau (4,5) and forms the core of tau fibrils in AD (6). Moreover, this aggregate formation is enhanced by the existence of a polyanion such as heparin (7,8) or by RNA (9) or fatty acids (10) in the absence of tau phosphorylation. However, these in vitro conditions may not be relevant to the mechanism underlying the formation of NFTs, because tau is always aberrantly hyperphosphorylated in AD. Therefore, it would seem necessary to consider the role of hyperphosphorylation of tau in the abnormal aggregation of filamentous tau.
The assembly of phosphorylated tau was also observed during the presence of 4-hydoroxy-2-nonenal (11), a lipid peroxidation by-product of oxidative stress. Increased oxidative stress is reported to occur in AD (12)(13)(14)(15). Interestingly, phospho-tau immunoreactive neurons are also stained positively with 8-hydroxy-oxyguanine, another marker for oxidative stress (16,17). These results suggest that oxidative stress may, in part, trigger the formation of NFTs. If oxidative stress does participate in NFTs formation in AD, such stress can lead to the activation of kinases that phosphorylate tau and stimulate NFT formation. One such candidate kinase is stress-activated protein kinase/ c-Jun N-terminal kinase (SAPK/JNK), a member of the mitogen-activated protein kinase family that is activated by several kinase cascades. Recent studies, employing antibodies against paired helical filaments (PHFs), have reported that activated phospho-JNK co-localized in neurons displaying PHF immunoreactivities (18,19). Phospho-JNK was also shown to translocate from the nucleus to the cytoplasm in NFT-bearing neurons (19). Furthermore, JNK has been shown to phosphorylate tau at Ser 422 (20), a site that is specifically phosphorylated in AD brains (21).
To understand the role of tau phosphorylation in NFT formation, we sought to develop a model that could reproduce the aberrant tau phosphorylation, including phosphorylation of Thr 212 and Ser 214 , key sites commonly referred to as the AT100 epitope. The AT100 epitope appears to be more specific for Alzheimer tau and required more than one kinase to be accomplished in vitro (22). In the present study, we used an adenovirus-mediated gene expression system to synergistically express tau, ⌬MEKK, JNK3, and GSK-3␤. This system enabled us to produce hyperphosphorylation of tau in cultured cells including that at AT100 and PS422. This hyperphosphorylated tau formed aggregates that were detergent-insoluble and Thioflavin-S-reactive and displayed relatively shorter fibrils than PHF. Our results suggest that the aberrant phosphorylation of tau can contribute to the formation of oligomeric tau fibrils and that other additional factors may be required for the growth step of tau fibrils.

EXPERIMENTAL PROCEDURES
Generation of Recombinant Adenoviruses-All transcription units were ligated into cassette cosmid pAxCAwt (TaKaRa) digested with SwaI, so that the insert was transcribed under the control of the CAG promoter (23). Using PCR and Pyrobest DNA polymerase (TaKaRa) with pFLAG-CMV2-JNK3 (24) as a template, the mouse JNK3 open reading frame was amplified along with the FLAG tag primer (5Ј-ACCATGGACTACAAAGACGATGACGAC-3Ј) and the arm primer of these expression vectors (5Ј-GCACTGGAGTGGCAACTTC-3Ј). pcDNA3-FLAG-⌬MEKK (24), which encodes residues 1169 -1488 of mouse MEKK1, was used as a template to generate pAxCA-FLAG-⌬MEKK via PCR and the FLAG tag primer and ⌬MEKK-ter primer (5Ј-CTAC-CACGTGGTACGGAAG-3Ј). Human wild type GSK-3␤ cDNA, tagged with a Myc epitope at the carboxyl terminus, was cloned into the pCIneo vector (Promega). The GSK-3␤ open reading frame was excised by EcoRI digestion from pCIneo-GSK-3␤, blunt-ended, and cloned into pAxCAwt at the SwaI site (designated pAxCA-GSK-3␤). Human wild type Tau four-repeat (Tau4RWT) cDNA, tagged with a FLAG epitope at the amino-terminal and a Myc epitope at the carboxyl terminus, was cloned into pCIneo. The Tau4RWT open reading frame was excised by XhoI and NotI digestion from pCIneo-Tau4RWT, blunt-ended, and then inserted into pAxCAwt at the SwaI site (designated pAxCA-Tau4RWT). These resulting cosmid clones were used to generate recombinant adenoviruses (designated AxCA vector) using the COS-TPC method (25). Briefly, cosmid DNA was co-transfected with the EcoT221-digested DNA-terminal protein complex of Ad-dlX (26) into human embryonic kidney 293 cells to generate the recombinant viruses via homologous recombination. These recombinant viruses were propagated in 293 cells. After the third propagation, viruses were extracted from the 293 cells, purified using a double cesium step gradient purification (27), dialyzed against a vehicle solution containing 10% glycerol in PBS (pH 7.4), and stored at Ϫ80°C. Titers of the recombinant viruses were determined with the modified end point cytopathic effect assay and 293 cells (27). Titers were expressed in plaque-forming units. Positive expression of the inserted gene product was confirmed by Western blot analysis using COS-7 cells. Adenoviruses containing JNK3, ⌬MEKK, GSK-3␤, Tau4RWT, and LacZ were named AxCA-JNK3, AxCA-⌬MEKK, AxCA-GSK-3␤, AxCA-Tau4RWT, and AxCAi-LacZ, respectively.
For quantification of proteins on immunoblots, serial dilutions of cell lysates were loaded onto gels to obtain a calibration curve, which allowed reliable quantification. All bands were quantified using the proprietary software, Image Gauge (version 3.0) (Fuji Films Science Lab 97). Statistical analyses were performed using one-factor analysis of variance (p Ͻ 0.01), followed by post hoc pairwise comparisons using Fisher's protected least significant differences. In all bar graphs, error bars indicate the S.E. of three separate experiments, whereas ** and * represent significant differences of p Ͻ 0.001 and p Ͻ 0.01, respectively.
Isolation of Insoluble Tau from COS-7 Cells-COS-7 cells infected with recombinant adenoviruses were lysed with RIPA buffer containing 1% SDS. 2 mg of cell lysate were centrifuged for 20 min at 100,000 ϫ g at 4°C. The resulting pellet was washed four times with 300 l of RIPA buffer using a sonic homogenizer. The insoluble pellet was solubilized in 70% formic acid for use in the immunoblot analysis or resuspended in 100 mM Tris-HCl (pH 8.3) for examination using electron microscopy. Following centrifugation for 20 min at 100,000 ϫ g at 4°C, the formic acid fraction was collected, air-dried, and subjected to immunoblot analysis after suspension in SDS gel loading buffer. The samples were resolved on 8% SDS-PAGE gels, transferred onto a polyvinylidene difluoride membrane, and isolated, and insoluble tau was identified with the TauC antibody, as mentioned above.
Immunohistochemistry-For double immunohistochemistry, cell preparation and immunohistochemical staining was performed as previously described (33,34). Briefly, cells in culture dishes were fixed with 4% paraformaldehyde for 4 min and then with cold methanol-acetone (1:2) for 6 min. The fixed cells were incubated with anti-AT8 (1:400) for 3 h at 4°C followed by incubation with the secondary antibody, Texas Red-conjugated goat anti-mouse IgG (Capple) for 30 min at room temperature. The cells were subsequently double-labeled using standard Thioflavin-S histochemistry. The stained cells were mounted in PBSglycerin (1:9) and analyzed with a fluorescence microscope.
Gold Immunolabeling of RIPA-insoluble Fraction and Electron Microscopy-For immunoelectron microscopy, the RIPA-insoluble pellet that was resuspended in 100 mM Tris-HCl (pH 8.3) (see above) was absorbed onto glow-charged supporting membranes placed on 300-mesh grids. After three washes with 100 mM phosphate buffer, treated samples were incubated overnight at 4°C in primary antibody solution containing either AT8 (1:20) or AT100 (1:40) or no antibody (for negative control) diluted in 100 mM Tris-HCl (pH 8.3). After washing, the samples were incubated with colloidal-gold conjugated secondary antibody (5 nm in diameter; 1:50, British BioCell International) for 2 h, fixed with 2% glutaraldehyde in 100 mM phosphate buffer for 5 min, negatively stained with neutralized 2% sodium phosphotungstic acid, and examined with an electron microscope (LEO 912AB, LEO Electron Microscopy, Ltd.).
Immunoelectron Microscopy of Tau-expressing Cells-The cell pellets were suspended in 4% paraformaldehyde, 0.2% glutaraldehyde in 0.1 M phosphate buffer for 4 h at 4°C. After washing, cells were embedded in 3% agarose, and then the samples were cut into 0.5-mm 3 blocks. The sample blocks were cryoprotected by immersion in graded concentrations of glycerol (10, 20, and 30%) in 0.1 M phosphate buffer. The slices were plunged rapidly into liquid propane (Ϫ184°C) by using a Leica EM CPC rapid freezing unit (Reichert, Vienna, Austria). The sections were dehydrated in 0.5% uranyl acetate dissolved in anhydrous methanol for 48 h at Ϫ90°C followed by raising of the temperature in steps of 4°C/h from Ϫ90 to Ϫ45°C in a Leica AFS cryosubstitution unit (Reichert). After washing with anhydrous methanol, the sections were infiltrated with Lowicryl HM20 resin (Polyscience, Inc., Warrington, PA) with a progressive increase in the ratio of resin to methanol and polymerized with ultraviolet light (360 nm) for 24 h at Ϫ45°C. For immunostaining, ultrathin sections (70 -80 nm) mounted on nickel grids were incubated with saturated sodium (meta)periodate for 20 min. After washing, the sections were incubated with 5% normal goat serum in washing buffer (0.1 M Tris-buffered saline, pH 7.4, containing 1% bovine serum albumin and 0.1% gelatin) for 20 min and then with primary antibody (anti-TauC, 1:50 dilution) overnight at 4°C. The sections were then washed with washing buffer and incubated with 5-nm colloidal gold-conjugated secondary antibody for 2 h. After a thorough wash, the sections were counterstained and examined electron-microscopically (LEO 912AB, LEO, Oberkochen, Germany).

Activated JNK3 Can Phosphorylate Tau in COS-7 Cells-
Because JNK is activated through ⌬MEKK (35-38), we first investigated the ability of JNK3 to phosphorylate tau in the presence or absence of ⌬MEKK (Fig. 1). Four different combinations of cDNAs (longest human tau with JNK3 and/or ⌬MEKK or with LacZ) were introduced into COS-7 cells by using the adenovirus-mediating gene transfer method (Fig. 1).
All of the cells expressed significant levels of tau, JNK3, ⌬MEKK, and/or LacZ (100% expression efficiency was obtained when the MOI of each construct used was over 10; data not shown). Cells co-expressing tau and JNK3 exhibited only negligible levels of activated forms of JNK (Fig. 1B, lane 1). In contrast, cells co-expressing tau, JNK3, and ⌬MEKK exhibited significant levels of activated JNK (Fig. 1B, lane 2). JNK activation by ⌬MEKK was associated with the reduced mobility of tau in SDS-PAGE and the reduced Tau-1-immunoreactivity (Fig. 1, C and D, lane 2) when compared with those in cells co-expressing tau with JNK3 alone, ⌬MEKK alone, or LacZ (Fig. 1, lanes 1, 3, and 4). These results suggest that ⌬MEKK activation is required for JNK3 to phosphorylate tau in COS-7 cells.
Determination of Tau Phosphorylation Sites in COS-7 Cells Expressing Activated JNK3-We further analyzed phosphorylated tau epitopes in cells co-expressing tau, JNK3, and ⌬MEKK using the well characterized phosphorylation dependent anti-tau antibodies PS199, PT205, PT231, PS262, PS396, PS404, PS422, AT8, AT100, AT180, and AT270 (Fig. 2). The level of tau phosphorylation at each site was quantified as a ratio of immunoreactivity of each antibody to that of phosphoindependent tau antibody TauC (Fig. 2a). Images of Western blots are shown in the lower panels of Fig. 2, and the graphs showing the corresponding quantification are shown in the upper panels. The levels of phosphorylation at PT205, PT231, PS396, AT8, AT180, AT270, PS422, and AT100 sites were significantly increased in cells expressing activated JNK3 (Fig.  2, b-i). In particular, tau phosphorylation at PT205, AT8, and PS422 was dramatically increased in response to JNK3 activation. Activation of JNK3 by ⌬MEKK (MOI ϭ 100) resulted in a 77-fold increase in PT205 immunoreactivity, an 81-fold increase in AT8 immunoreactivity, and a 330-fold increase in PS422 immunoreactivity (Fig. 2, b, e, and h, lane 3, respectively) when compared with those from cells co-infected with LacZ (MOI ϭ 100). AT100 immunoreactivity in these cells displayed an 18-fold increase over that in control cells expressing LacZ (Fig. 2i, lane 3). Interestingly, AT100 as well as PS422 epitopes are known to be more specific in Alzheimer tau (21,22). The immunoreactivities of PT231, PS396, AT180, and AT270 in activated JNK3-expressing cells (Fig. 2, c, d, f, and g,   lane 3, respectively) exhibited 3-5-fold increases over those in control cells. PS199, PS262, and PS404 immunoreactivities increased only 1-1.8-fold, suggesting that these epitopes are not major phosphorylation sites for JNK3 (Fig. 2, j-l, lane 3, respectively). Thus, with JNK3 activation, the Ser 202 , Thr 205 , and Ser 422 sites were most susceptible, the Thr 212 and Ser 214 sites were moderately susceptible, and the Thr 181 , Thr 231 , Ser 235 , and Ser 396 sites were relatively less but significantly susceptible to JNK3-mediated phosphorylation. The Ser 199 , Ser 262 , and Ser 404 sites appear to be unsusceptible to JNK3 phosphorylation.
PHF-like Aberrant Tau Phosphorylation by GSK-3␤ and Activated JNK3-JNK3 phosphorylates tau at most of the pathological sites except Ser 199 , Ser 262 , and Ser 404 . These JNK3sensitive or -insensitive sites overlapped with those reported in a previous in vitro study using recombinant proteins (39). In the same study, GSK-3␤ was reported to phosphorylate some of the JNK-insensitive sites. Motivated by this finding, we next attempted to simultaneously express tau, ⌬MEKK, JNK3, and GSK-3␤ in COS-7 cells to determine the effect of simultaneous overexpression of activated JNK3 and GSK-3␤ on tau phosphorylation in our model. We refer to this co-expression as quadruple expression in this paper. Fig. 3a shows a result from the Western blot analysis probed with the phospho-independent anti-tau antibody TauC. With quadruple expression, tau migrated more slowly almost as a single band than tau phosphorylated solely by activated JNK3 (Fig. 3a, lanes 1 and 2), suggesting that additional overexpression of GSK-3␤ to activated JNK3 causes additive effect on tau phosphorylation in COS-7 cells. The Western blot analysis using the PS199, PS262, and PS404 antibodies indicated the enhanced phosphorylation at these JNK3-insensitive sites (Fig. 3, b-d). Moreover, other sites also displayed enhanced phosphorylation in response to additional GSK-3␤ expression. For example, the immunoreactivities of AT8 (Fig. 3g), AT180 (Fig. 3e), and AT100 (Fig. 3i) with quadruple expression increased 2-3.4-fold compared with those with activated JNK3 expression alone. In contrast, additional GSK-3␤ expression had no significant additive effect on the AT270 (Fig. 3f) and PS422 (Fig. 3h) immunoreactivities. Taken together, activated JNK3 and GSK-3␤ synergistically phosphorylated tau at most of the pathological phosphorylation sites of tau in COS-7 cells.
Histological and Biochemical Analysis of Tau in COS-7 Cells Expressing Tau, ⌬MEKK, JNK3, and GSK-3␤-To determine whether this quadruple expression in COS-7 cells induces aberrant tau aggregates, cells were stained with the phospho-dependent anti-tau antibody AT8 (Fig. 4b) and Thioflavin-S histochemistry (Fig. 4a). Thioflavin-S is known as a marker for the insoluble protein aggregates with ␤-pleated sheets, as is shown in Fig. 4g, displaying an NFT-bearing neuron in an AD brain. AT8 staining was confined mainly to the cytoplasm of quadruple expressing cells. Some of these cells exhibited Thioflavin-S staining in the cytoplasm, probably showing a partial aggregation of hyperphosphorylated tau in these cells (Fig. 4, a-c). We never found Thioflavin-S staining in tau/JNK3/⌬MEKK-, tau/ GSK-3␤-, and tau/LacZ-expressing cells (Fig. 4, d-f). These data suggest that some cells simultaneously expressing tau, activated-JNK3, and GSK-3␤ occasionally form aggregates of hyperphosphorylated tau with ␤-sheet structures.
We further investigated biochemically whether hyperphosphorylated tau formed insoluble fibrillar aggregates. Lysates from cells co-expressing tau, ⌬MEKK, JNK3, and/or GSK-3␤ were dissolved in RIPA (containing 1% SDS) buffer, and both RIPA-soluble and -insoluble fractions were analyzed. Fig. 5a shows an immunoblot of the RIPA-soluble fractions probed with the phospho-independent anti-tau antibody TauC. Al- though the total amounts of tau recovered in the RIPA-soluble fractions were almost constant, mobility of tau in each lane changed depending on the combination of kinases expressed. RIPA-insoluble materials were next solubilized using 70% formic acid and dissolved into Laemmli sample buffer after lyophilization. Only tau from quadruple-expressing cells was recovered in the formic acid fraction (Fig. 5b), suggesting that tau from quadruple-expressing cells becomes partly insoluble in RIPA (containing 1% SDS).
The ultrastructure of immunolabeled RIPA insoluble materials from quadruple expressing cells and AD brains were investigated by using the electron microscope. We found RIPAinsoluble tau filaments from AD brains identified with the AT8 and AT100 antibodies (Fig. 5, c and e, respectively) that had two different shapes with different diameters: short fibrils with 10-nm diameter (Fig. 5c) and relatively longer fibrils with 20-nm diameter (Fig. 5e). Consistent with this observation, in the RIPA-insoluble fractions of quadruple-expressing cells, the AT8-and AT100-positive tau aggregates were identified that contain short fibril-like structures with about 10-nm diameters (Fig. 5, d and f, respectively). The similar tau aggregates were found in the perinuclear regions of the quadruple expressing cells (Fig. 5g) but not in control cells expressing tau and LacZ (Fig. 5h). These data suggest that hyperphosphorylation of tau by GSK-3␤ and activated JNK3 could enforce the formation of tau aggregates sometimes containing short fibrils. (h), and AT100 (i) antibodies, respectively. Because the introduced tau cDNA was tagged with FLAG and Myc, the molecular weight of tau expressed in COS-7 cells was higher than that of main tau bands identified in the PHF-tau smears.

DISCUSSION
In this report, we demonstrate that activated JNK3 phosphorylates tau in COS-7 cells. This kinase increases levels of tau phosphorylation at the epitopes of PT205, PT231, PS396, PS422, AT8, AT100, AT180, and AT270. This is almost consistent with the results obtained from the in vitro studies (40). The combination of activated JNK3 and GSK-3␤ cooperatively phosphorylates 12 Ser and Thr residues of tau in vivo. These include the AT100 site, which is the specific phosphorylation site in PHF tau. There are more than 19 phosphorylation sites identified in PHF tau (41). Most phosphorylation-dependent tau antibodies, however, also have been known to recognize a fraction of tau from biopsied normal adult or fetal brains, although the levels of phosphorylation are less than those recognized in AD brains (42). Among these antibodies, AT100 never recognizes normal adult or fetal tau but does specifically recognize PHF tau (22,42). Thus, the AT100 epitope is a unique and specific phosphorylation site in PHF tau. In this sense, tau in quadruple expressing cells is in a similar phosphorylation state to that in AD.
The in vitro phosphorylation of AT100 epitope was reported by using the combination of protein kinase A and GSK-3␤ (22). According to this study, the phosphorylation of the AT100 epitope required the sequential phosphorylation of GSK-3␤ and protein kinase A, although the other combination may be due to the phosphorylation of these epitopes. This combination may not be relevant to AD tau formation, because neither GSK-3␤ nor protein kinase A can phosphorylate Ser 422 , a site that is highly phosphorylated in AD brain (20,21). In our study, activated JNK3 phosphorylated tau both at the Ser 422 and AT100 epitopes. However, activated JNK3 did not phosphorylate the Ser 199 , Ser 262 , and Ser 404 sites, phosphorylation of which required the additional expression of GSK-3␤. Furthermore, immunoreactivity for AT100 epitopes is much more enhanced in cells co-expressing both activated JNK3 and GSK-3␤. It is also possible that the formation of AT100 epitopes in cells expressing only activated JNK3 may result from the synergism between JNK3 and endogenous GSK-3␤. Thus, the combination of GSK-3␤ and activated JNK3 can phosphorylate tau at most of the phosphorylation sites documented in AD. Recent studies reported that activated phospho-JNK or GSK-3␤ co-localized in NFT-bearing (18,19) or AT8-positive neurons (19,43,44), and amyloid ␤ treatment activated JNK (45,46) and GSK-3␤ in cultured cells (47)(48)(49)(50). Taken together, activated JNK and GSK-3␤ may be a strong candidate kinase combination involved in the mechanisms of tau pathology in AD.
It has been a question for a long time whether the aberrant phosphorylation of tau is a cause of NFT formation or just a consequence of some other unidentified process. One hint was  4). The blots were stained with the phosphorylation-independent antibody TauC. The arrowheads indicate the tau protein bands. The RIPA-insoluble pellets were resuspended in 100 mM Tris-HCl (pH 8.3) by brief sonication and prepared for immunogold electron microscopy. PHF-tau filaments in RIPA-insoluble pellets obtained from an AD brain are shown in c and e. The fibrillar tau aggregates in the RIPA-insoluble pellets were also isolated from quadruple expressing COS-7 cells (d and f). Aggregates were immunolabeled with AT8 (a and b) or AT100 (c and d) antibodies followed by 5-nm gold particle-conjugated secondary antibodies. Similar immunogold-positive tau aggregates (arrow in g; a larger image is shown in the inset of g) were observed in quadruple-expressing COS-7 cells (g) but not in control cells expressing tau and LacZ (h). The asterisks indicate nuclei. Scale bars (c-f), 20 nm; g-h, 100 nm.  ; green). The merged image is represented in c (yellow). As a positive control, a sagittal AD brain section was also stained with Thioflavin-S (g). The staining intensities of COS-7 cells co-infected with AxCATau4RWT, AxCAFLAGJNK3, and AxCAFLAG⌬MEKK (d); with AxCATau4RWT and AxCAGSK-3␤ (e); or with AxCATau4RWT and AxCAiLacZ (f) were below the threshold of detection at the exposure times used for the photography. Scale bar, 20 m.
provided by a recent in vitro study that showed that hyperphosphorylation of tau in AD brain induces self-assembly of tau into PHF (51). Our present results indicate that the aberrant phosphorylation of tau results in the formation of short fibrils that are RIPA-insoluble. These fibrils were relatively shorter than reported PHF in AD brains. Since similar types of fibrils were also recovered from AD brains, tau fibrils observed in the quadruple expressing COS-7 cells might be a good model to study tau aggregation in a cellular environment.
The formation of NFTs is mainly divided into three stages. First is the pretangle stage, which exhibits PHF epitopes and becomes Gallyas silver-positive. Second is the mature tangle stage, which shows PHF epitopes, Gallyas silver staining, and Thiazine Red staining, which recognizes fibrillar structures. The last is the ghost tangle stage. Our histochemical study revealed AT8 and Thioflavin-S-positive staining; Thioflavin-S is more sensitive than Thiazine Red when recognizing oligomeric fibrils with ␤-sheet structure (52). However, overall the staining pattern did not show a clear fibrillar structure, suggesting that COS-7 cells expressing both activated JNK3 and GSK-3␤ may mimic the neurons in the pretangle stage but not in the mature tangle stage. Thus, the aberrant phosphorylation of tau probably promotes tau oligomer formation such as seen in the pretangle stage. Other factors or certain intracellular environment are apparently required for the development of this stage to the mature one. For example, antiapoptotic factors might be such candidates, since activated JNK3 and GSK-3␤ have been known to activate a cascade for apoptosis. Further studies will be required to identify this tau fibril growth factors.