Activation of Glycogen Synthase Kinase 3β Promotes the Intermolecular Association of Tau

Tau is hyperphosphorylated and undergoes proteolysis in Alzheimer disease brain. Caspase-cleaved tau efficiently forms fibrillary structures in vitro and in situ. Glycogen synthase kinase 3β (GSK3β) phosphorylates tau and induces the aggregation of caspase-cleaved tau in situ. Given the hypothesis that increased association of tau precedes the formation of fibrillar structures, we generated a cell model to quantitate the extent of tau association in situ using fluorescence resonance energy transfer (FRET) microscopy. The cyan and yellow fluorescent proteins were attached to full-length (T4) and caspase-cleaved (T4C3) tau at either the N or C termini, and a pair of cyan and yellow fluorescent protein-tagged tau were co-transfected into human embryonic kidney cells. The FRET efficiency was examined in the presence of a constitutively active or a kinase-dead GSK3β. Active GSK3β significantly increased FRET efficiency with both T4 and T4C3, indicating that GSK3β activation resulted in an increase in the self-association of both T4 and T4C3, but interestingly only T4 is efficiently phosphorylated by GSK3β. There was no significant difference in FRET efficiency between T4 and T4C3, although only T4C3 in the presence of active GSK3β leads to the formation of Sarkosyl-insoluble inclusions. These FRET studies demonstrate that GSK3β facilitates the association of T4 and T4C3, and the presence of caspase-cleaved tau is necessary for the evolution of tau oligomers into Sarkosyl-insoluble inclusions even though it is not extensively phosphorylated. These data imply that increased association of tau should not be regarded as a direct indicator of the formation of insoluble tau aggregates.

The accumulation of polymeric filaments of the microtubule-associated protein tau as the intracellular neurofibrillary tangles (NFTs) 2 is one of the major neuropathological features of several diseases known as "tauopathies," which include Alzheimer disease (AD) and frontotemporal dementia with parkinsonism linked to chromosome 17, a group of autosomal dominant neurodegenerative diseases caused by mutations in the tau gene (1). NFTs are mainly composed of paired helical filaments (PHFs), which are formed from abnormally hyperphosphorylated tau (2,3). Although in AD, the abundance of NFTs correlates positively with the severity of cognitive impairment (4), the role of NFTs as toxic mediators of neuronal dysfunction and death is still not clear. Several animal models show cognitive deficits and impaired axonal transport in the absence of NFTs (5)(6)(7). Furthermore, suppression of tau expression in a transgenic mouse model restored memory function and stabilized neuronal cell populations, whereas NFTs continued to accumulate (8). These studies suggest that NFTs are not sufficient to cause cognitive decline or neuronal death, and small soluble oligomers may be the toxic species (9). Therefore, elucidating the early steps in the process of tau oligomerization is of fundamental importance.
Post-translational modifications of tau such as aberrant phosphorylation have been demonstrated to play a significant role in tau aggregate formation (10 -15). Tau is hyperphosphorylated at numerous sites in NFTs of AD brains (16), and in vitro studies showed that hyperphosphorylated tau can self-assemble into PHFs (17,18). Phosphorylation at specific sites can significantly increase the tendency of tau to aggregate. For example, pseudophosphorylation of Ser 396 and Ser 404 , as well as Ser 205 , Thr 205 , and Thr 212 , makes tau more fibrillogenic (19 -21), and tau protein in which Ser 422 is mutated to Glu shows a significantly increased propensity to aggregate (15). There is emerging evidence that glycogen synthase kinase 3␤ (GSK3␤) may play a major role in regulating tau phosphorylation in pathological conditions (22,23). Increased expression of GSK3␤ results in increased tau phosphorylation at pathological sites, including the PHF-1 epitope (Ser(P) 396/404 ) (24,25). Furthermore, there is evidence for increased activation of GSK3␤ in AD brain (26 -28).
Along with aberrant phosphorylation, caspase cleavage has been also reported to play a role in the aggregation of tau (29 -31). It has been demonstrated that caspases play a critical role in A␤-induced neuronal apoptosis (32) and are activated in apoptotic neurons in AD brain (33). Although tau in AD is present predominantly as an intact full-length molecule (16), tau in AD NFTs has been shown to be truncated at Asp 421 (29). Truncated tau, albeit not a large amount, may play a significant role in the neuronal cell death and PHF formation, given that truncated tau has been shown to be associated with apoptosis in cultured cells (34,35) and has been demonstrated to play a significant role in the nucleation-dependent filament formation of tau (35). Therefore, it is important to investigate how GSK3␤-mediated phosphorylation and caspase cleavage of tau contribute to the initiation of the tau aggregation process. In a previous study, it was demonstrated that tau truncated at Asp 421 (T4C3), but not full-length tau (T4), partitioned into a Sarkosyl-insoluble fraction and formed thioflavin S-positive aggregates when co-expressed with GSK3␤ (36). This indicates that a combination of phosphorylation and cleavage of tau may coordinate in regulating tau aggregation.
Fluorescence resonance energy transfer (FRET) is a process by which energy is transferred from a donor fluorophore (CFP) to an acceptor fluorophore (YFP) in a distance-dependent manner (typically 20 -60 Å). FRET is one of the few tools available for measuring nanometer scale distances and changes in distances, both in vitro and in vivo (37)(38)(39). Because FRET efficiency decreases as the inverse sixth power of distance, small distance and orientation changes between the donor and acceptor fluorophores can dramatically affect the FRET efficiency (40). A number of applications using FRET to understand protein-protein interactions have been reported (38,(41)(42)(43). For example FRET has been used to detect the dimerization of a protein (44), indicating that FRET can be used to evaluate the self-association of a protein.
This study was undertaken to understand the intermolecular association of tau proteins in response to GSK3␤ activation and/or caspase cleavage. To achieve this goal FRET microscopy was used with full-length tau (T4) and caspase-cleaved (Asp 421 ) tau (T4C3) to which CFP and YFP were linked. To evaluate different tau orientations during the intermolecular interactions of tau, N-terminally linked CFP (CFP-tau) was co-transfected with N-terminally linked YFP (YFP-tau) or C-terminally linked YFP (tau-YFP) for both T4 and T4C3 in the absence or presence of active GSK3␤, and subsequent to quantitating FRET efficiency. The data from these studies demonstrate that activation of GSK3␤ results in an increase in the intermolecular association of both T4 and T4C3 tau. However, increased intermolecular association did not necessarily lead to the formation of insoluble tau aggregates, suggesting that although the increased intermolecular interaction is likely an early event in the formation of insoluble tau aggregates, tau may remain in an oligomeric state and not progress into an aggregate in the absence of other contributing cellular processes.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-To make the FRET constructs, PCR products of pECFP-C1 (Clontech) and pEYFP-C1 (Clontech) were cloned into NheI and KpnI sites of pcDNA3.1(ϩ) (Invitrogen) to make the N-terminally linked CFP and YFP vectors, nCFP and nYFP vector, respectively. The PCR product of pEYFP-N1 (Clontech) was cloned into XbaI and ApaI sites pcDNA3.1(ϩ) (Invitrogen) to make a C-terminally linked YFP vector, cYFP vector. To make a positive FRET control construct, a PCR product of pEYFP-N1 was cloned into XbaI and ApaI sites of nCFP vector (nCFP-YFP vector), and then a highly flexible linker domain, which consisted of Gly-Ser-Asp-Gly-Gly-Ser-Gly-Gly-Gly-Ser-Thr-Ser (45), was introduced into the vector to make nCFP-Linker-YFP. T4 containing human tau with four microtubule-binding repeats but without exons 2 and 3 and T4C3 with the last 20 amino acids deleted and thus mimicking caspase-cleaved tau (36) were introduced into KpnI and XbaI sites of the pcDNA3.1(ϩ) FRET vectors to yield nCFP-tau, nYFP-tau, and tau-cYFP constructs. To make the tau-cYFP vectors, the stop codons of T4 and T4C3 were replaced with GGA before cloning into FRET vectors. All the constructs were confirmed by DNA sequencing analysis.
Coverslip Preparation for FRET Microscopy-Forty eight hours after transfection, cells on the coverslips were washed with PBS and fixed with 3% paraformaldehyde in PBS for 15 min. The cells were washed with PBS and mounted. Coverslips were sealed with clear nail polish.
FRET Microscopy-Intermolecular interaction of tau proteins was determined by the Acceptor Photobleaching method of FRET detection (37)(38)(39). A Leica SP2 confocal microscope (Leica) was used to record the fluorescence of CFP and YFP both before and after selective photobleaching of 70% of the YFP acceptor fluorophore. CFP was excited by 458 nm light, and the emission was collected through a 465-500 nm bandpass filter (Chroma Technology, Brattleboro, VT). YFP was excited by 514 nm light and the emission was collect through a 525-600 nm bandpass filter (Chroma Technology, Brattleboro, VT). Selective photobleaching of YFP was performed by repeatedly scanning a region of the specimen with the 514 nm laser line set at maximum intensity to photobleach at 70% of the original acceptor fluorescence. The fluorescence emission from the donor and the acceptor were collected using the Leica software. A Timed Bleach protocol was utilized to automate the acquisition of pre-bleach images, perform acceptor photobleaching of the acceptor, and acquisition of post-bleach images. The pre-and post-bleach images are saved as a single file that can be analyzed using measurement functions resident in the Leica software. Average fluorescence intensities of the donor are measured before and after bleaching, and the efficiency of FRET was calculated by E T ϭ 1 Ϫ(I DA /I D ) where I DA and I D represent the steady state donor fluorescence in the presence and the absence of the acceptor.
Sarkosyl Fractionation Assay-The assay was carried out essentially as described previously (36,53). HEK cells were cotransfected with tau FRET constructs and with either active GSK3␤ or kd-GSK3␤. Cells were rinsed with ice-cold PBS, scraped off the plate, collected by spinning at 200 ϫ g for 10 min at 4°C, resuspended in RAB buffer (100 mM Mes, 1 mM EGTA, 0.5 mM MgSO 4 , 750 mM NaCl, 20 mM NaF) containing 1 mM phenylmethylsulfonyl fluoride and 10 g/ml each of aprotinin, leupeptin, and pepstatin, and then homogenized with 30 strokes using a tissue grinder. The homogenates were centrifuged at 20,000 ϫ g for 20 min, and the supernatants were collected as the RAB fraction. The pellets were resuspended with RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 5 mM EDTA, 0.5% sodium deoxycholate, 0.1% SDS) containing 1 mM phenylmethylsulfonyl fluoride and 10 g/ml each of aprotinin, leupeptin, and pepstatin, sonicated briefly, and then incubated on ice for 30 min. The homogenates were centrifuged at 20,000 ϫ g for 20 min, and the supernatants were collected as RIPA fraction. The pellets were resuspended with RAB buffer (100 mM Mes, 1 mM EGTA, 0.5 mM MgSO 4 , 750 mM NaCl, 20 mM NaF) containing 1% N-lauroylsarcosine (Sarkosyl) in Mes buffer (20 mM Mes, pH 6.8, 80 mM NaCl, 1 mM MgCl 2 , 2 mM EGTA, 10 mM NaH 2 PO 4 , and 20 mM NaF) with protease inhibitors, vortexed for 30 min at room temperature, incubated overnight at 4°C, and then centrifuged at 200,000 ϫ g for 30 min (Airfuge) at room temperature. The supernatants (Sarkosyl-soluble fraction) were collected and diluted with 2ϫ protein loading buffer, and the pellets (Sarkosylinsoluble fraction) were resuspended in 2ϫ protein loading buffer and incubated in a boiling water bath for 5 min. Samples were separated by SDS-PAGE, transferred, and blotted with Tau5/5A6 antibodies.
Statistics-Data were analyzed using Student's t test, and values were considered significantly different when p was Ͻ0.05.

Generation and Expression of FRET Tau Constructs-To
determine the intermolecular interaction of tau proteins using FRET microscopy, CFP and YFP constructs, which were linked to full-length (T4) or tau truncated at Asp 421 (T4C3) that mim-ics caspase-3 cleavage, were generated. To detect different orientations for the tau-tau interactions, YFP fluorophore was linked to the N terminus or C terminus of tau, whereas the CFP fluorophore was linked only to the N terminus of tau protein (Fig. 1). As a positive control for FRET, a flexible linker domain consisting of 12 amino acids was introduced between CFP and YFP (CFP-Linker-YFP in Fig. 1). HEK cells were transfected with these FRET constructs and immunoblotted to determine their expression levels (Fig. 2). The blots were probed with a green fluorescent protein antibody (Roche Applied Science), which also recognizes both CFP and YFP proteins (top panel in Fig. 2), and with the Tau5/5A6 antibodies, which recognize total tau (bottom panel in Fig. 2). Naive T4 not linked to a fluorophore was also run for comparison (Fig. 2). All the tau FRET constructs expressed at approximately equivalent levels. As expected, the T4C3 FRET proteins, which are 20 amino acids shorter than T4 FRET proteins, migrated faster than the T4 FRET proteins. Full-length tau (T4) and caspase-cleaved tau (T4C3) were linked to CFP or YFP fluorophores as indicated at either the C terminus or N terminus of tau. The T4 tau construct, which is a human tau, contains four microtubule binding repeats but without exons 2 and 3, and T4C3 contains the same as T4 except the last 20 amino acids were removed, which mimics caspasecleaved tau. The CFP fluorophore was only linked to the N terminus of T4 and T4C3, and the YFP fluorophore was linked to either the N terminus or C terminus of tau as indicated. To link YFP fluorophore at the C terminus of tau proteins, the stop codon of the tau constructs was mutagenized to GGA, which codes for glycine. The top bar shows a positive control FRET construct where the CFP fluorophore was linked with the YFP fluorophore using a flexible linker domain, composed of 12 amino acids as follows: Gly-Ser-Asp-Gly-Gly-Ser-Gly-Gly-Gly-Ser-Thr-Ser. Representative immunoblots with these antibodies showed that the different tau constructs were expressed at approximately equal levels, and linking of a fluorophore to tau reduced the electrophoretic mobility of the tau proteins as expected. The T4C3 FRET constructs migrated faster than the T4 FRET constructs. For molecular weight comparisons, naive T4, which was not linked to a fluorophore (far right band on blots), and CFP-linker-YFP, the positive FRET control construct (far left band on blots), were included in these blots.

Active GSK3␤ Increased Tau Intermolecular Interactions-
To determine the intermolecular association of T4 or T4C3, CFP-tau was transiently co-transfected with YFP-tau or tau-YFP in the presence of active GSK3␤ or kd-GSK3␤, and then FRET efficiency was measured. As a positive control in FRET microscopy, the CFP-linker-YFP construct was used (54). To determine the FRET efficiency, approximately half of a cell was selected as the region of interest to bleach, and the other half of the cell was used as background to obtain pure FRET efficiency. The net FRET values were calculated using the following equation: net FRET efficiency ϭ FRET efficiency in acceptor-bleached area Ϫ FRET efficiency in the other half nonbleached area. The FRET efficiency in the nonbleached area was negligible in all groups. Representative FRET images are shown in Fig. 3, and quantitative analyses of FRET efficiency, which were obtained from three independent experiments, are presented in Fig. 4. The positive control CFP-Linker-YFP showed a FRET efficiency of 15.81 Ϯ 1.46, and images of CFP and YFP with this construct completely overlapped as expected. Expression of active GSK3␤␤ resulted in significantly increased FRET efficiency with both T4 and T4C3 compared with kd-GSK3␤, which indicates that active GSK3␤ increased intermolecular associations of both T4 and T4C3 (Fig. 3 and Fig. 4). The arbitrary FRET efficiencies (mean Ϯ S.E.) were as follows: 3.61 Ϯ 0.21 and 5.89 Ϯ 0.30 for CFP-T4 with YFP-T4 in the presence of active and kd-GSK3␤, respectively; 3.63 Ϯ 0.29 and 5.73 Ϯ 0.52 for CFP-T4 with T4-YFP in the presence of active and kd-GSK3␤, respectively; 2.94 Ϯ 0.28 and 4.83 Ϯ 0.68 for CFP-T4C3 with YFP-T4C3 in the presence of active and kd-GSK3␤, respectively; and 2.98 Ϯ 0.50 and 5.29 Ϯ 0.70 for CFP-T4C3 with T4C3-YFP in the presence of active and kd-GSK3␤, respectively (Fig. 4). However, there were no significant differences in the FRET intensity between YFP-tau and tau-YFP, which suggests that there is equivalent parallel and anti-parallel selfassociation of tau in situ. Furthermore, no significant differences were observed between T4 and T4C3 either in basal conditions or in the presence of active GSK3␤. No FRET efficiency was observed with either CFP-tau or YFP-tau alone (data now shown).
GSK3␤ Phosphorylates T4 More Efficiently than T4C3-To determine the relative effectiveness of GSK3␤-mediated phosphorylation of these FRET tau constructs, samples were immu- FIGURE 3. Representative images of increased intermolecular association of tau by GSK3␤ activation using FRET microscopy. FRET tau constructs were transiently co-transfected with either active or kd-GSK3␤ into HEK cells. The CFP-linker-YFP was also transfected as a positive control. 48 h after transfection, cells on coverslips were fixed, mounted onto slides, and examined. Images shown are of pre-photobleached and post-photobleached images of acceptor (YFP) and donor (CFP), respectively, and the image of FRET efficiency. Intensity of the acceptor (YFP) was adjusted to slightly lower than the saturation intensity given the fact that the acceptor would be bleached to 30% of its beginning intensity, whereas the beginning intensity of donor (CFP) was adjusted to less bright intensity compared with that of pre-bleached acceptor (YFP) in order to better visualize the increased intensity of donor (CFP) after bleaching. To detect FRET efficiency, approximately half of the cell was selected as region of interest and the other half of the cells as control area. The acceptor (YFP) was bleached in the selected area, and then the FRET efficiency was obtained. The positive control in which the two fluorophores were directly connected via a flexible 12-amino acid stretch (CFP-linker-YFP) exhibited a high FRET efficiency (top panel), whereas there was no FRET efficiency with either CFP or YFP alone (data not shown).
In the presence of active GSK3␤, FRET efficiencies between tau proteins were significantly increased compared with those of kd-GSK3␤ in both T4 and T4C3. However, no differences between CFP-tau and YFP-tau or tau-YFP were observed with both the T4 (A) and T4C3 (B) groups. AUGUST 10, 2007 • VOLUME 282 • NUMBER 32

JOURNAL OF BIOLOGICAL CHEMISTRY 23413
noblotted with the phosphospecific antibodies PHF-1 (Ser(P) 396 / 404 ) or AT-180 (Thr(P) 231 ) (Fig. 5). Co-expression of active GSK3␤ with T4 resulted in a significant decrease in electrophoretic mobility compared with the mobility when kd-GSK3␤ was used indicating that the phosphorylation state of tau was increased with active GSK3␤ (Fig. 5, Total tau). In contrast to what was observed with T4, the electrophoretic mobility shift of T4C3 was negligible in the presence of active GSK3␤ (Fig. 5, Total tau). The extent of phosphorylation of T4 by active GSK3␤ at both the PHF-1 and AT-180 epitopes was substantially greater than that of T4C3 (Fig. 5). This is consistent with a previous finding showing that truncation of tau at Asp 421 attenuates GSK3␤-mediated phosphorylation (36).
GSK3␤ Causes T4C3 but Not T4 to Form Sarkosyl-insoluble Aggregates-To examine the role of GSK3␤ on the formation of Sarkosyl-insoluble aggregates of these tau constructs used for the FRET, the partitioning of fluorophore-tagged tau into the Sarkosyl-insoluble fractions was investigated. Lysates were prepared from cells expressing the tau constructs in the presence of either active or kd-GSK3␤ and then separated into RAB, RIPA, Sarkosyl-soluble, and Sarkosyl-insoluble fractions (Fig.  6). In the presence of kd-GSK3␤, none of the tau FRET constructs localized to the Sarkosyl-insoluble fractions. In accordance with previous findings (10,36), expression of active GSK3␤ resulted in the presence of tau in the Sarkosyl-insoluble fractions with the fluorophore T4C3 constructs but not the T4 constructs (Fig. 6). Interestingly, T4C3 was more abundant in the RIPA fraction than T4, which indicates T4C3 is less soluble in basal conditions compared with T4.

DISCUSSION
This study demonstrates for the first time using FRET microscopy that the kinase activity of GSK3␤ results in increased intermolecular association of tau, which is likely an early step in the process of the formation of insoluble tau aggre- HEK cells were transfected with a combination of tau FRET constructs in the presence of active or kd-GSK3␤, and the blots were probed with the indicated antibodies as follows: 5A6 for total tau, At-180 for phosphorylated tau at Thr 231 , and PHF-1 for phosphorylated tau at Ser 396 /Ser 404 . T4 FRET constructs were phosphorylated more efficiently by active GSK3␤ than the T4C3 FRET constructs based on the slower electrophoretic mobility of the T4 constructs when expressed with active GSK3␤, which was not the case with the T4C3 constructs (5A6 immunoblot, top panel). This was also evident with the phosphospecific tau antibodies (AT-180 [Thr(P) 231 ] and PHF-1 [Ser(P) 396/404 ]; 2nd and 3rd panels, respectively). There were no significant differences in the expression levels of active and kd-GSK3␤ (bottom panel). Y-and -Y represent YFP-and -YFP, respectively.  bleaching should be significantly increased compared with that of prebleaching donor (CFP) upon acceptor (YFP) photobleaching. In this study, the acceptor (YFP) was bleached to 30% of its original intensity, and the mean values of FRET efficiencies were calculated from at least five cells from three independent experiments using the Leica FRET software and displayed as mean Ϯ S.E. The presence of active GSK3␤ resulted in a significant increase in the FRET efficiencies of both T4 and T4C3 compared with those in the presence of kd-GSK3␤. However, no significant differences between the two orientation groups were observed. Furthermore, there was no significant difference between T4 and T4C3 either in the presence of active or kd-GSK3␤. Y-and -Y represent YFP-and -YFP, respectively. *, p Ͻ 0.05.
gates. In addition, the data suggest that although increased interactions of tau may be an initial step, it does not necessarily predict the efficiency of tau aggregate formation.
Although PHFs are one of the characteristic hallmarks of AD, the role of aggregates as toxic mediators of neuronal dysfunction and death is still not clear. Indeed, the correlation between NFT presence and the incidence of disease does not necessarily dictate a causal relationship (5)(6)(7)(8), and therefore the focus has been shifting toward identifying the toxic tau species during the entire fibrillogenic process from the soluble monomers through oligomers to the insoluble mature tangles (NFTs). Whether the monomers, oligomers, or NFTs are the important toxic mediators, the question of what causes tau to aggregate and the kinetic profile of this process must be addressed. Therefore, in this study we used FRET microscopy to examine the initial steps of tau self-association, which then subsequently contribute to aggregate formation.
Although the confocal images of co-expressed CFP-tagged and YFP-tagged tau proteins appeared to be overlapping in the presence of kd-GSK3␤, FRET efficiency data confirmed that in the basal condition CFP-tagged and YFP-tagged tau proteins were not within close proximity, and therefore a mere overlap of fluorescence is not necessarily indicative of an interaction. However, expression of active GSK3␤ resulted in an increase in FRET efficiency, indicative of an increased association between the tau proteins, accompanying the apparent overlapping images of the two proteins. Therefore, FRET microscopy provides a significant advantage in obtaining information about protein-protein interactions compared with confocal microscopy (55). To detect different orientations of intermolecular association of tau, N-terminally CFP-tagged tau protein (CFPtau) was co-transfected with either N-terminally YFP tagged tau (YFP-tau) or C-terminally YFP tagged tau (tau-YFP). Active GSK3␤ resulted in increased FRET efficiencies in with both orientations of T4 and T4C3, and there was no significant difference in FRET efficiency between the two orientations, suggesting that either orientation occurs when tau self-associates in situ. Although the intermolecular interaction of tau was significantly increased with GSK3␤ activation, the FRET efficiency was not increased as high as the positive control with both T4 and T4C3, indicating that intermolecular association of tau protein at this stage may not be as tight as that of the mature aggregates.
Activation of GSK3␤ resulted in the increased intermolecular association of both full-length (T4) and caspase-cleaved tau (T4C3) in the present study, an initial step in the oligomerization and aggregate formation of tau. It has been suggested that the fibrillization pathway of tau can be subdivided based on key steps (56). First, the microtubule binding function of tau must be neutralized so that tau protein accumulates in an assemblycompetent form (57), which suggests that aggregation of tau may be concentration-dependent. Second, tau molecules selfassociate through their microtubule binding repeat regions to form the ␤-sheet-enriched filaments. When the critical concentration of tau molecules is achieved, unfolded monomer tau molecules, which have no substantial secondary structures, oligomerize leading to a conformational change to a ␤-sheet enriched structure (57,58). The earliest secondary structure detectable with fluorescent dyes corresponds to tau aggregates associated with membranous structures (59), suggesting that the folding of tau protein into ␤-sheet-containing species may be facilitated by interaction with intracellular membranes and organelles. The final step involves the nucleation of tau filaments and formation of mature NFTs. In the transition, NFTs undergo proteolytic modifications (57) and become highly insoluble (58).
It has been suggested that post-translational modifications of tau such as aberrant site-specific phosphorylation and caspase cleavage contribute to the formation of NFTs in the earliest stages. Several studies have provided evidence that phosphorylation of key sites on tau has a strong impact on the normal function of tau and likely contributes to its pathological role, including the tendency of aggregate formation. The role of phosphorylation on tau assembly was previously hypothesized (60). According to their hypothesis, tau self-assembles mainly through the microtubule binding domains; however, regions of tau molecule N-terminal and C-terminal to the microtubule binding domains inhibit tau association in the naive state. Hyperphosphorylation of tau neutralizes these inhibitory domains enabling self-association of tau. In accordance with this hypothesis, phosphorylation at both Ser 396 /Ser 404 and Ser 422 enhances tau aggregation (11). It also has been reported that pseudophosphorylation of tau at Ser 205 , Thr 205 , and Thr 212 enhances polymerization of tau into filaments (20). In addition, phosphorylation of tau at Thr 231 (46) or Ser 262 (61) negatively regulates tau-microtubule interactions. However, phosphorylation at Ser 262 may decrease the ability of tau to polymerize (11,15,19), suggesting that abnormal phosphorylation of this site on tau does not play a facilitatory role in the pathological aggregation into PHFs (11). The data in this study show that extensive phosphorylation, including phosphorylation at the PHF-1 epitope, does not necessarily lead to the aggregate formation of full-length tau (T4), whereas the aggregates were observed with truncated tau (T4C3), which was not phosphorylated extensively, suggesting that hyperphosphorylation is not sufficient to cause tau to form insoluble aggregates.
In addition to aberrant phosphorylation, caspase cleavage of tau has been reported to influence fibril formation of tau (62). Tau is a substrate for caspase-3 in vitro, and tau truncated at Asp 421 , the caspase-3 cleavage site in tau, aggregates more rapidly than full-length tau (29,34,35). In addition, the presence of Asp 421 -truncated tau in pretangle neurons suggests a role for caspase cleavage in the initiation of polymer formation (63). In accordance with previous reports, the present data demonstrated that T4C3 formed Sarkosyl-insoluble aggregates in the presence of active GSK3␤. Given the present data, either phosphorylation or caspase cleavage alone may not be sufficient to cause tau aggregation. Although T4 was extensively phosphorylated by active GSK3␤ and intermolecular interaction was significantly increased as shown in FRET microscopy, no T4 was present in the Sarkosyl-insoluble fractions indicating that T4 did not form aggregates. In contrast, even though T4C3 is minimally phosphorylated by GSK3␤, T4C3 formed Sarkosylinsoluble aggregates in the presence of active GSK3␤, an event that was concurrent with an increase in the intermolecular association of T4C3 as determined by FRET microscopy. Given the data from the present study and other studies, the aggregation of tau appears to involve an ordered series of phosphorylation and cleavage events (62).
In this study, the seemingly incongruent finding that although expression of active GSK3␤ caused an increase in T4 self-association this did not lead to the formation of Sarkosylinsoluble tau aggregates suggests that other factors might be necessary to facilitate aggregate formation by full-length tau. The availability of nucleation seeds, which may occur with truncated tau proteins such as T4C3, and the attainment of excessive amounts of phosphorylated tau over the critical concentration for aggregation might be conditioning factors. Furthermore, the data that GSK3␤ results in the increased selfassociation of T4C3, leading to the aggregate formation in the absence of extensive phosphorylation, also sets forth the possibility that GSK3␤ may phosphorylate other intracellular mediators, which in turn facilitate the aggregate formation of tau, given the numerous roles of GSK3␤ in cells (23). However, it is still possible that the amount of phosphorylated T4C3, although a small portion of the total amount of T4C3, may act as a nucleation seed for the remaining unphosphorylated T4C3, which is more fibrillogenic compared with full-length tau. It has been suggested that a combination of phosphorylation events and caspase cleavage play together in tau aggregation and partitioning into the Sarkosyl-insoluble fraction (36). The data from this study and other studies suggest that phosphorylation of tau by GSK3␤ likely precedes caspase cleavage, because Asp 421 -truncated tau is not efficiently phosphorylated by GSK3␤, and phosphorylation of tau by GSK3␤ does not inhibit cleavage by caspase-3 (10,36).
Taken together the present study demonstrates that both phosphorylation and truncation play an important role in a cooperative manner in the process of aggregate formation of tau, and FRET microscopy may be a valuable tool to examine the early step in the aggregate formation of tau.