|Home | About | Journals | Submit | Contact Us | Français|
The microtubule-associated protein tau, in a hyperphosphorylated form, aggregates into insoluble paired-helical filaments (PHFs) in Alzheimer’s disease (AD) and other tauopathies. In AD, there are approximately 8 mol of phosphate per mol of tau distributed among approximately 30 PHF phosphorylation sites as compared to 2–3 phosphate per mol in normal brain. In AD, kinases such as glycogen synthase kinase-3β (GSK-3β) are believed to be involved in generation of hyperphosphorylated tau. However, the functional consequences of hyperphosphorylation on the microtubule binding and polymerization of tau are not well understood. To address this question, we have generated pseudo-hyperphosphorylation mutants consisting of six and seven sites in the proline rich region and carboxy-terminus of tau by amino acid substitution. In addition, several single, double and triple pseudo-phosphorylation mutants were also generated. Pseudo-phosphorylation of tau decreases its affinity for microtubules, and pseudo-hyperphosphorylated forms of tau do not have significantly decreased microtubule binding as compared to single and double-sites. Three pseudo-hyperphosphorylated forms of tau with altered SDS-PAGE migration have a greater effect on its inducer mediated polymerization, slowing the rate of nucleation and elongation. Based on the observations that pseudo-hyperphosphorylated tau has decreased affinity for microtubules and reduced inducer-initiated rates of nucleation and polymerization, we propose that this combination could be the cause for the increased cytotoxicity of hyperphosphorylated tau in Alzheimer’s disease and also explain the potentially beneficial role of tau polymerization and NFT formation.
Neurofibrillary tangles (NFTs) are a prominent hallmark of Alzheimer’s disease (AD) and other related neurodegenerative disorders, such as frontotemporal dementia and Parkinsonism linked to chromosome 17 (FTDP-17) (reviewed in (1)). Although the role of NFTs in Alzheimer’s disease is not clear, the location and amount of NFTs directly correlate with the type and severity of cognitive impairment (2, 3). NFTs contain large amounts of abnormal protein deposits in the form of straight and paired-helical filaments (4, 5) that are composed almost entirely of the microtubule-associated protein tau (6–9) in a hyperphosphorylated state (6). Tau protein is a soluble protein that normally functions to promote the assembly and stabilization of the microtubule cytoskeleton (10), and phosphorylation of tau is known to alter its functions (reviewed in (11)). However, it is still unclear how tau hyperphosphorylation is linked to its aggregation and neuronal degeneration observed in these diseases.
At least 30 phosphorylation sites in PHF tau, about half of which are proline-directed sites, have been identified by mass spectrometry and phospho-amino acid analysis or by phosphorylation-specific monoclonal antibodies. (reviewed in (11, 12)). The kinase(s) responsible for the hyperphosphorylation of tau in AD is not clear. It has been suggested that more than one combination of kinases and/or phospatases might be involved in converting normal tau to hyperphosphorylated tau (13). The proline-directed kinase GSK-3β is believed to be involved in this process since it phosphorylates tau at many of the same sites that are found to be phosphorylated in paired helical filaments of AD (14–17), and active GSK-3β is associated with the development of NFTs in AD (18). According to our previous observations, GSK-3β phosphorylates soluble and polymerized tau at a minimum of eleven and five sites respectively (19, 20). Phosphorylation at these sites does not change the amount of tau polymerization, but is sufficient to cause tau filaments to coalesce into tangle-like aggregates similar to those isolated from Alzheimer’s disease brain (19, 20).
Due to the number and proximity of tau phosphorylation, in combination with the relative promiscuity with which most kinases act upon tau, the use of kinases to investigate the effects of specific sites on the function of tau is virtually impossible. It is therefore common for these investigations to employ site-directed mutagenesis to generate pseudo-phosphorylation variants of tau (21–28), although this approach is unlikely to completely mimic phosphorylative changes. We previously showed that the amino acid substitution of glutamic acid for serines or threonines to generate pseudo-phosphorylation mutations at S202 and T205 resulted in only subtle effects on the polymerization characteristics and microtubule binding of tau (28). However, it is not known whether adding additional pseudo-phosphorylation sites that are normally phosphorylated by GSK-3β to generate a “pseudo-hyperphosphorylation” version of tau has a greater effect on its function. To address this question, we have generated pseudo-hyperphosphorylated variants with six (6-phos) and seven (7-phos) pseudo-phosphorylation changes. The sites chosen for investigation include the five sites that are phosphorylated by GSK-3β when tau is in a polymerized state (S199, T205, T231, S396 and S404) (20). 6-phos also included pseudo-phosphorylation at S202. S202 was added to the core five sites since the AT8 antibody recognizes phosphorylation of tau at both S199 and S202 (29), and antibodies to S202 alone were not available for our previous study of GSK3β phosphorylation of tau. Similarly, S235, which is recognized in combination with T231 by the TG-3 antibody (30), was added to 6-phos to generate a 7-phos mutant. Single, double and triple pseudo-phosphorylation mutants were also investigated to control for the possibility that fewer sites could have similar effects to the pseudo-hyperphosphorylated versions of tau. In addition to S199, S202, T205, T231, S235, S396 and S404, other sites were also investigated. T212E was included since it has previously be shown to promote the in vitro assembly of tau to a high level (26). S208D, S210D and S208/210D were also included. These sites are not phosphorylated by GSK-3β, but are in the same proline-rich region of tau that the majority of GSK-3β phosphorylation sites are found. They are known to be phosphorylated by tau-tubulin kinase (TTK) (31) and are also phosphorylated in PHF tau (32). The pseudo-phosphorylation mutants were assayed for changes in SDS-PAGE migration, microtubule binding and arachidonic acid-induced polymerization. We have made two key observations: most pseudo-phosphorylation and pseudo-hyperphosphorylation variants decrease the microtubule binding of tau in a fashion similar to GSK-3β phosphorylated tau, and kinetic analyses of polymerization indicate that pseudo-hyperphosphorylated tau mimics with AD-like changes in SDS-PAGE electrophoretic mobility have increased lag times and reduced apparent rates of elongation compared to single or paired mimics. These results suggest a possible mechanism for the increased toxicity of hyperphosphorylated tau in vivo.
Arachidonic acid was obtained from Cayman Chemicals (Ann Arbor, MI), IPTG from Calbiochem (EMD Biosciences, La Jolla, CA), SDS-PAGE protein marker from Invitrogen (Gaithersburg, MD), urea from Bio-Rad (Hercules, CA), and uranyl acetate and formvar carbon coated grids from Electron Microscopy Sciences (Hatfield, PA). Wild-type, P301L and pseudo-phosphorylation mutant tau protein (441 amino acids) were expressed and purified as described previously (28). Pseudo-phosphorylation mutants were generated using the QuikChange site-directed mutagenesis kit from Stratagene (La Jolla, CA). S/T to E mutants and S/T to D mutants were generated at amino acids S199 and T231. S199E was found to have indistinguishable polymerization characteristics as S199D (not shown). T231E and T231D were also indistinguishable in their polymerization (not shown). Protein concentration was determined by a commercial BCA assay from Pierce Chemical (Rockford, IL) using bovine serum albumin (Pierce Chemical, Rockford, IL) as a standard.
Protein samples (1 μg per lane) were boiled for 5 min in sample buffer containing 2% SDS and 1% beta-mercapto-ethanol, fractionated on 15% SDS-PAGE and stained with Coomassie brilliant blue (33). For denaturing SDS-PAGE, protein samples (1 μg per lane) were boiled for 5 min in sample buffer containing 2% SDS and 1% beta-mercapto-ethanol and 6 M urea. Samples were fractionated on 15% SDS-PAGE with 6 M urea and stained with Coomassie brilliant blue (33).
2 μM tau protein (wild-type and pseudo-phosphorylation tau mutants) was incubated in polymerization buffer (10 mM HEPES pH 7.64, 100 mM NaCl, 0.1 mM EDTA and 5 mM DTT) at room temperature in the presence of 75 μM arachidonic acid (ARA) (in ethanol, final concentration: 3.75%) in the 5 × 5 mm optical glass fluorometer cuvettes (Starna Cells, Atascadero, CA) (34). The amount of polymerization was monitored by LLS at regular intervals for 20 hr with 5 mW solid state laser (λ= 475 nm, B & W Tek, Inc., Newark, DE). Images were captured at a right angle to the incident light using a SONY XC-ST270 digital camera. Captured images were imported into Adobe Photoshop 7.0.1 and the intensity of the scattered light was measured as described previously (28). The pictures were captured at an aperture of f5.6–8 or f8 and then normalized to the intensity corresponded to the exposure at f5.6–8. The data were fit to a non-linear Gompertz function:
where y is the value of laser light scattering measured at time t; a is the maximum amount of light scattering; ti is the point of inflection where the increase in scattering is at its maximum; and b is equal to 1/kapp. kapp is proportional to the rate of polymerization (26). The lag time for polymerization, or the amount of time required to observe significant amounts of polymerization, was calculated using the formula: ti − b (26). The Gompertz function has been used as a model for growth curves (35), including as a model for tau polymerization (26). Errors in the kinetics parameters were assessed by individually fitting three separate kinetics experiments to the Gompertz function to obtain three independent measures of LLSmax, kapp and lag times. These values were averaged and reported ± standard deviation.
Tau polymerization reactions were diluted 10-fold with polymerization buffer then fixed with glutaraldehyde (2% final concentration) for 5 min. Ten microliters of fixed reactions were placed on formvar-carbon coated grids (1 min), washed with water, blotted, washed with 2% uranyl acetate, blotted, stained with 2% uranyl acetate (1 min), then blotted dry. Grids were viewed with a TECNAI G2 20 electron microscope (FEI Company, Hillsboro, OR) and digital images were captured with the Gatan Digital Micrograph imaging system.
The microtubule binding assay was performed in general tubulin buffer (80 mM PIPES, pH 7, 1 mM MgCl2, 1 mM EGTA). Tau protein ranging from 0.125 μM to 10 μM were mixed with paclitaxel stabilized microtubules at a final concentration of 1.62 μM tubulin dimer in a 50 μL reaction. Samples were incubated at room temperature for 30 min, and centrifuged in a Beckman Optima TLX ultracentrifuge at 100,000 × g for 5 min to sediment the microtubules and tau bound to the microtubules. The pellets were resuspended in SDS sample buffer and microtubules were separated from tau bound to microtubules by SDS-PAGE. The concentration of tau bound to the microtubules was determined by the intensity of tau band by Adobe Photoshop and normalized to the intensity of tubulin band. The tau band was normalized to the intensity of the tubulin bands in each lane to account for any differences in centrifugation or resuspension of samples because the amount of tubulin should be constant in all binding reactions. The amount of free tau was determined by subtracting the amount of bound tau from the total amount of tau. The concentration of bound tau were plotted versus the concentration of free tau in GraphPad Prism and fit to a one site binding (hyperbola) equation.
Hyperphosphorylated tau in Alzheimer’s disease (AD) is phosphorylated at a molar ratio of approximately 5–9 mol phosphate per mol of tau whereas normal tau contains only 2–3 mol phosphate per mol of tau (36). However, the role of this hyperphosphorylation is poorly understood. To better understand the effects of tau hyperphosphorylation, pseudo-phosphorylation mutants were constructed. We chose to focus on the five sites known to be phosphorylated by GSK-3β in vitro when arachidonic acid-induced tau filaments are used as the substrate (S199, T205, T231, S396 and S404) (20). Pseudo-phosphorylation changes at S202 were added to generate the 6-phos, and both S202 and S235 were added to generate 7-phos (Figure 1A). These sites were included since S202 is often found phosphorylated along with S199, and S235 is often found phosphorylated along with T231. Single, paired and triplet mutants were investigated to determine whether fewer sites might also have an effect on the function of tau (Figure 1A). We also investigated a GSK-3β site in the same proline rich region but not among the sites above (T212) as a control. As further controls, non-proline directed sites S208, S210 and S208/S210 were investigated. The disease causing P301L tau mutation was included in the analyses as a positive control.
Alzheimer’s disease hyperphosphorylated tau and recombinant tau phosphorylated by GSK-3β in vitro have an upward band shift when analyzed by SDS-PAGE (19, 37). This upward band shift is the definitive characteristic of hyperphosphorylated tau in neurodegenerative disorders (1). To determine whether the pseudo-phosphorylation mutants can cause this upward band shift, the pseudo-phosphorylation forms of tau were analyzed by coomassie-stained SDS-PAGE electrophoresis. S199/S202/T205E, S396/S404E, 6-Phos and 7-Phos had an upward shift in mobility compared to wild-type tau (Figure 1B). The apparent molecular masses for wild-type, S199/S202/T205E, S396/S404E, 6-Phos and 7-Phos were 70 kDa, 72 kDa, 75 kDa, 78 kDa and 78 kDa, respectively. P301L tau and all other pseudo-phosphorylation mutants did not show a change in mobility (data not shown). The mobility shift of pseudo-phosphorylation mutants was reduced in the presence of 6M urea (molecular weights for wild-type, S199/S202/T205E, S396/S404E, 6-Phos and 7-Phos were 70 kDa, 70 kDa, 71 kDa, 72 kDa and 72 kDa, respectively, Figure 1C). This phenomenon of altered mobility in SDS that is abolished in urea has been described in the literature as a SDS-resistant change in conformation (33, 38–41). Because S199/S202/T205E, S396/S404E, 6-Phos and 7-Phos all demonstrated an AD-like shift in mobility as a result of phosphorylation-like changes, we conclude that they have the characteristics of hyperphosphorylated tau. These mutants will therefore be referred to as pseudo-hyperphosphorylated tau throughout the manuscript.
Because FTDP-17 mutations (42) and pseudo-phosphorylation mutations (28) in tau can alter its interactions with microtubules, we measured the microtubule binding affinity of P301L tau and fifteen pseudo-phosphorylation mutants using a centrifugation assay (43). Pellets containing microtubules and bound tau were resuspended in SDS-sample buffer and analyzed by coomassie-stained SDS-PAGE (Figure 2A). The concentration of tau bound to the microtubules was determined by the intensity of tau band and normalized to the intensity of tubulin band (Figure 2A). Normalization of the amount of tau relative to tubulin was performed to take into account any differences in sedimentation or resuspension of the pellets. The concentration of free tau was calculated by subtracting the measured bound tau from the total tau added to the reaction. The amount of bound tau was plotted against free tau. Two representative curves (wild-type tau and 7-Phos) are shown in Figure 2B. The data were fit to a simple one-site binding equation to determine the affinity of binding (Kd) and the maximal amount of bound tau (Bmax) (Figure 2C and 2D). All proteins examined, with the exception of S199E, S202E and S208D, had significantly lower affinities for microtubules than was observed with wild-type tau (Figure 2B). The decreases in binding affinity were all in the 2–3 fold reduction range, which is in general agreement with previously published reports of GSK-3β phosphorylation of tau (44) and the P301L mutation (42). The stoichiometry of binding to microtubules was not greatly affected (Figure 2C).
To determine whether pseudo-phosphorylation mutations of tau have an effect on the arachidonic acid (ARA) induction of tau polymerization, each protein at a final concentration of 2 μM was incubated in the presence of 75 μM ARA. This protein to inducer ratio was chosen since it has been shown to be the optimal condition for the arachidonic acid induction of wild-type tau polymerization (34). The kinetics of polymerization of each mutant were followed by right angle laser light scattering and compared to wild-type tau. Representative curves are shown in Figure 3A. The data were fit to a Gompertz function to determine the maximal amount of light scattering (LLSmax) (Figure 3B), the apparent proportional growth rate (kapp, Figure 3C), and the lag time of polymerization (Figure 3D). No significant increases in polymerization were observed for the pseudo-phosphorylation mutants (Figure 3B), although the FTDP-17 mutation P301L did have a significant increase in the extent of polymerization compared to wild-type, as expected (45). Several pseudo-phosphorylation variants (S202/T205E, S199/S202/T205E, S208/S210D, T231E, S396/S404E, 6-Phos and 7-Phos) had significantly less maximal polymerization as compared to wild-type.
The apparent proportional growth rate, kapp, was not significantly altered for most of the mutants (Figure 3C). The FTDP-17 mutation P301L has been shown to have the most dramatic effects in increasing the apparent rate of polymerization as compared to wild-type tau protein (45), and had a significantly greater kapp than wild-type tau and all the pseudo-phosphorylation mutants. Mutants S210D and T231D also had significantly greater kapp values than wild type, but still less than P301L tau (Figure 3C). S199/S202/T205E, 6-phos and 7-phos all had significantly lower kapp than wild type tau (Figure 3C).
Nucleation-elongation polymerization pathways are also typically characterized by a lag time in the polymerization process. Wild type tau had a lag time of 19 minutes (Figure 3D). P301L tau had a significantly reduced lag time, as expected (45). S210D, S208/S210D, T231E and T231E/S235D also had significantly reduced lag times, but greater than P301L tau (Figure 3D). 6-phos had a significantly longer lag time than wild type tau (Figure 3D). On average, S199/S202/T205E and 7-Phos had longer lag times than wild type, but these differences did not reach statistical significance (Figure 3D).
To determine whether filament morphology (number and length of filaments) was changed by pseudo-phosphorylation, samples from polymerization reactions were prepared and viewed by TEM (Figure 4). Filaments from six fields of view were combined into a single data set and then ranked according to their length (examples are shown Figure 5A). Particles less than 15nm in their longest dimension were not included in the analysis because particles smaller than this cutoff value are smaller than the average filament width of arachidonic acid induced filaments (19), and therefore difficult to distinguish from background. In general, pseudo-phosphorylation mutants tended to form fewer but longer filaments than wild-type tau (Figure 5B and 5C). The standard deviation of the average lengths was roughly equal to the value of the average lengths for most proteins, as is expected for exponentially distributed data (46). The standard error of the mean is also shown, as this is an estimate of the accuracy of the determination of the true mean. Among these mutants, the three with the most phosphorylation sites formed the longest filaments (S199/S202/T20E5, 6-Phos and 7-Phos) (Figure 5C). However, the overall amount of filament formation observed in six fields was relatively similar for most all proteins (Figure 5D), with the possible exceptions of T231E, S202/T205E, T205E and S396/404E which seemed to have somewhat reduced levels of polymerization compared to the other proteins (Figure 5D). In general, pseudo-phosphorylation of tau resulted in fewer, but longer filaments as compared to wild-type (Figure 4 and and5).5). In contrast, the rapidly polymerizing P301L formed more, but shorter filaments than wild-type tau, resulting in an overall increase in the amount of polymerized material (Figure 5A–D).
Quantitative analysis of differences between the filaments is difficult due to the differences in the number of filaments, the variability in filament distribution on the TEM grid, the biological variability in filament lengths, and the exponential distribution of filament lengths. Therefore, it is possible that the apparent differences in filament length distributions (Figure 5) could be due to differences in sampling. We therefore measured 500 filaments for each mutant and determined the number-average length (Ln=(ΣNiLi)/(ΣNi), Figure 6) for each mutant (47). With the exception of T231, all pseudo-phosphorylation mutants had greater average filament lengths than wild type protein. This analysis also confirmed that S199/S202/T205E, 6-Phos and 7-Phos had the greatest tendencies for forming long filaments.
Hyperphosphorylation of the microtubule-associated protein tau is a pathological hallmark of Alzheimer’s disease and other tauopathies. Many kinases have been shown to be involved in this phosphorylation process, and GSK-3β is believed to be a major contributor to tau/PHF phosphorylation (reviewed in (11)). In previous studies, we found that GSK-3β phosphorylation of pre-formed tau filaments at sites including S199, T205, T231, S396 and S404 can cause tau filaments to coalesce into NFT-like structures (20). Because these five sites are recognized by specific AD-associated antibodies, we sought to test whether pseudo-phosphorylation at these sites would have an impact on the function of tau (summarized in Table 1).
Phosphorylation of tau at specific sites has been suggested to correlate with the formation of neurofibrillary tangles (NFTs), raising the possibility that a major function of tau phosphorylation is to promote tau polymerization (48). Under conditions that we have determined to be optimal for wild-type tau polymerization (34), pseudo-phosphorylation of tau mildly reduced its polymerization, which is consistent with previously published reports using in vitro GSK-3β phosphorylated tau (19, 49). The kinetics of polymerization for most of the single and double phosphorylation variants were not significantly changed (Table 1). However, the pseudo-hyperphosphorylated version of tau (S199/S202/T205E, 6-Phos and 7-Phos) had a drastically decreased rate of elongation and a pronounced lag time. This suggests a deficiency in the nucleation step for the pseudo-hyperphosphorylation mimics. The observation of fewer but longer filaments with pseudo-hyperphosphorylation mutants agree with this idea.
The effect on filament nucleation and filament length was greatest with the “pseudo-hyperphosphorylated” versions of tau. 6-phos and 7-phos were generated from the combination of multiple pseudo-phosphorylation sites in both the proline-rich and C-terminal regions of tau. The effects on polymerization of 6-phos and 7-phos seem to be greater than those observed by the heparin (49) or arachidonic acid (19) induction of GSK-3β phosphorylated tau polymerization. However, GSK-3β phosphorylation in vitro only results in 2–4 mol phosphate incorporation per mol of tau distributed unequally over as many as fourteen sites (19, 20, 49–51). 6-phos and 7-phos are homogeneous in their modifications and therefore are more highly modified than the average GSK-3β phosphorylated tau protein.
Our previous studies have shown that the presence of arachidonic acid drastically reduces the energetic barrier for the nucleation of polymerization (34). The observation that pseudo-hyperphosphorylation mutants are less likely to nucleate could be explained by diminished interactions with arachidonic acid as a result of altered conformations of the proteins conferred by pseudo-phosphorylation. In support of this, the pseudo-hyperphosphorylated proteins had altered SDS-PAGE electrophoretic mobility that was abolished in the presence of urea. This phenomenon has been described as an SDS-resistant change in conformation of tau (33, 38–41), and indicates that pseudo-phosphorylated proteins could interact differently with SDS or with arachidonic acid.
It should be mentioned that a previous report found that pseudo-phosphorylation of tau at several sites (including S199, S199/S202/T205 and T212) enhanced polymerization by 2–3 fold over non-modified tau (26). We obtain similar results when using concentrations of arachidonic acid that inhibit wild type tau polymerization (125–150 μM, data not shown). It is not clear whether these results at high inducer to protein ratios have physiological relevance, but they are consistent with the hypothesis that pseudo-phosphorylation of tau can decrease its interactions with arachidonic acid.
Most pseudo-phosphorylation forms of tau significantly decreased microtubule binding by 2–3 fold without significantly changing the saturation levels (Table 1), which agrees with previous studies investigating the effects of tau phosphorylation at proline-directed phosphorylation sites (49). However, results from multiple site constructs were not consistently additive. For example, both T231E and S235D had lower affinities than wild-type for microtubules. However, the T231E/S235D double mutant had an affinity similar to the individual mutations. And while S202/T205E had a lower affinity for microtubules than the individual mutants, the addition of S199E to form S199/S202/T205E actually improved microtubule binding (albeit at a lower affinity than wild-type tau). The constructs with six or seven pseudo-phosphorylation sites did not show dramatically different levels of microtubule binding than single or double mutants. We conclude that microtubule binding is regulated in a complicated site-specific fashion, and the levels of phosphorylation are less of a determinant of microtubule binding than the actual sites involved.
Since the establishment of the correlation between the number of NFTs with the degree of dementia in AD, NFTs have been widely accepted as a toxic species in AD (reviewed in (52)). However, recent observations in tau-expressing cultured cell, transgenic Drosophila and mouse models suggest that the toxicity is independent of NFT formation and NFTs may form as a result of cellular attempts to reduce toxicity (reviewed in (53)). NFTs contain large amounts of abnormal hyperphosphorylated tau (6) in the form of straight and paired-helical filaments (4, 5). The abnormal hyperphosphorylation of tau is one of the post-translational modifications of tau known to be related to toxicity, filament formation and the formation of NFTs (reviewed in (54)). The apparent phosphorylation-dependent toxicity of tau could be due to the disruption of microtubule dynamics through sequestration of normal tau (55–57), and could be reduced through the polymerization of phosphorylated tau (58). Additionally, the presence of soluble hyperphosphorylated tau is correlated with cognitive deficiencies in a transgenic mouse model (59). These results suggest that tau aggregation could protect against phosphorylation-induced microtubule disruption and subsequent neurodegeneration.
Our results provide a potential biochemical mechanism for the potential cytotoxicity of hyperphosphorylated tau. Phosphorylated tau binds to microtubules less tightly, so that there would be an increase in cytoplasmic tau and increased microtubule instability. It has also been shown through other work that hyperphosphorylated tau may sequester normal tau (56), further disrupting microtubule dynamics. Pseudo-hyperphosphorylated tau also has a decreased rate of polymerization, potentially increasing the amount of time it is soluble in the cytoplasm where it could sequester normal tau. It is also possible that the fibrillization of abnormally phosphorylated tau could protect cells from toxic effects. We have previously shown tau phosphorylation by GSK-3β is sufficient to induce the bundling tau filaments into NFT-like structures in a cell-free in vitro system (20). These observations are consistent with data from transgenic mouse models showing that NFT formation is not necessarily the toxic element of neurodegeneration and may instead play a protective role (59–61).
We thank Carolyn Rankin and Kellen Voss for critical reading of the manuscript. We also thank Mike Branden for protein production. This work was supported by NIH AG022428 (TCG) and NIH AG025898 (TCG).
†Support was provided by NIH AG022428 (TCG) and NIH AG025898 (TCG).