Functional differences of tau isoforms containing 3 or 4 C-terminal repeat regions and the influence of oxidative stress.

We report functional differences between tau isoforms with 3 or 4 C-terminal repeats and a difference in susceptibility to oxidative conditions, with respect to the regulation of microtubule dynamics in vitro and tau-microtubule binding in cultured cells. In the presence of dithiothreitol in vitro, a 3-repeat tau isoform promotes microtubule nucleation, reduces the tubulin critical concentration for microtubule assembly, and suppresses dynamic instability. Under non-reducing conditions, threshold concentrations of 3-repeat tau and tubulin exist below which this isoform still promotes microtubule nucleation and assembly but fails to reduce the tubulin critical concentration or suppress dynamic instability; above these threshold concentrations, amorphous aggregates of 3-repeat tau and tubulin can be produced at the expense of microtubule formation. A 4-repeat tau isoform is less sensitive to the oxidative potential of the environment, behaving under oxidative conditions similarly to the 3-repeat isoform under reducing conditions. Under conditions of oxidative stress, in Chinese hamster ovary cells stably expressing either 3- or 4-repeat tau, 3-repeat tau disassociates from microtubules more readily than the 4-repeat isoform, and tau-containing high molecular weight aggregates are preferentially observed in lysates from the Chinese hamster ovary cells expressing 3-repeat tau, indicating greater susceptibility of 3-repeat tau to oxidative conditions, compared with 4-repeat tau in vivo.

Tau is principally a neuronal MAP 1 (1) and, in the brain, is composed of six isoforms generated by alternative splicing (2,3). The isoforms differ in the presence of one, two, or zero N-terminal inserts and the presence of either three or four C-terminal imperfect repeat domains, the latter being involved in binding of tau to tubulin. Tau is the principal component of PHF that compose the neurofibrillary tangles of Alzheimer's disease (AD) and is also a major component of several other inclusion bodies characteristic of some less common neurodegenerative diseases, collectively now described as tauopathies. The inclusion bodies in the different tauopathies contain different complements of tau isoforms (4 -12). For example, PHF in AD contain all 6 isoforms, whereas Pick bodies contain only 3-repeat tau (3R) isoforms, but progressive supranuclear palsy and frontotemporal dementia with Parkinsonism linked to chromosome 17 tangles can contain only 4-repeat tau (4R). Thus, despite the discovery of mutations in the tau gene (4 -7), since selective isoform aggregation occurs in the absence of a mutation and the combination of isoforms in aggregates is disease-related, the mechanism of aggregation must involve other unknown factors that may differentially affect tau isoforms. We have therefore undertaken to re-examine the differences in the microtubule assembly properties of two isoforms, deliberately choosing the shortest and longest isoforms in order to maximize any possible isoform-specific differences in tau properties.
Previous studies have implicated a major role for tau in the regulation of growth and shortening of MTs (13)(14)(15)(16)(17), promotion of MT nucleation (17)(18)(19)(20)(21), and the modulation of dynamic instability of MTs (18,22). Many neuronal MAPs have been shown to reduce the critical concentration of tubulin, C c , required for MT assembly (23,24). More recently, distinct differences have begun to emerge in the ability of individual MAPs to regulate MT behavior, and the simplified picture of a general suppression of dynamic instability and promotion of MT growth by MAPs is becoming more complicated (25)(26)(27)(28). This complex behavior of MAPs extends to the individual isoforms of tau, which, for example, have now been found to have distinct effects on MT behavior as shown by Trinczek et al. (26), Goode et al. (28), and by us as reported herein. There exist, however, some discrepancies regarding the major effects of tau on MT dynamics. For example, some studies (14) have emphasized effects of tau on changes in MT association rates and on the complete suppression of dynamic instability, whereas other studies (15,26,29) have focused on changes in MT dissociation rates and on varying degrees of suppression of dynamic instability.
The work reported here was initiated by our previous, somewhat surprising, observation that 3R promoted MT nucleation in PEM assembly buffer (PEM, microtubule assembly buffer containing 100 mM PIPES, 0.1 mM EGTA, 0.5 mM magnesium chloride, at pH 6.5) but was found not to have a significant effect on the apparent critical concentration, C c(app) , of tubulin (30). This finding was in contradiction to a report by Trinczek et al. (26), in which it was shown that 3R in assembly buffer containing a sulfhydryl-reducing agent does reduce the apparent critical concentration of tubulin, C c (app) . We have therefore characterized the effects of tau isoforms on C c(app) and other parameters of MT dynamics in further detail, including a variation in buffer conditions.
Oxidative damage is the earliest cytopathological marker of neuronal dysfunction in AD (31)(32)(33)(34), and the cytoskeleton is a known target for oxidative stress, highlighted by the aggregation of components such as actin (35). Oxidation/dimerization of tau is important because tau dimers enhance tau aggregation into PHF (36 -39). It has also been suggested that in conditions where tau may overload microtubule surfaces, due to microtubule decay or overexpression of tau, changes reminiscent of PHF could occur on the microtubule surface. This implicates tubulin, acting as a polyanionic inducer of tau aggregation, in the progression of tau pathology (36).
We have examined the functional differences of 3-and 4R in vitro, which led to further study to identify whether oxidative stress conditions affected cells expressing tau. We have found that there are differences in isoform behavior in vitro such that under certain conditions 3R can behave disruptively, and that in vivo 3R and 4R have different susceptibility to oxidative stress.
These observations may be of pathological importance in Alzheimer's disease and other tauopathies in which the stoichiometry of isoform composition in neurons may change as well as degenerating neurons being subjected to oxidative stress; these two factors may then result in disruption of the cytoskeleton (40), potentially aided by abortive aggregation involving the shortest tau isoform.

EXPERIMENTAL PROCEDURES
Tubulin and Tau Purification-MT protein was isolated from porcine brain (41). Tubulin was purified from endogenous MAPs by phosphocellulose chromatography (Whatman P11). No contamination of the tubulin preparation with high molecular weight MAPs or tau was detected using SDS-PAGE (42), staining for total protein, and Western blotting probing with a polyclonal antibody to tau, TP70 (43). MT fragments for seeded assembly studies (EGS seeds) were prepared by cross-linking MTs with ethylene glycol bis(succinic acid N-hydroxysuccinimide ester) (EGS) (44). Immediately prior to experiments, tubulin was subjected to an additional cycle of assembly and disassembly. All experiments were carried out in PEM buffer supplemented with 1 mM GTP. For detailed methodology see Ref. 27, and for review of in vitro MT dynamic studies see Ref. 45.
The expression and purification of recombinant tau isoforms, 3R (0N, 3R) and 4R (2N, 4R), was based on previous procedures (30,46). 1 mM DTT was included throughout the purification procedure, and final tau stocks were either placed in PEM alone or PEM ϩ DTT (PEM ϩ DTT, PEM buffer containing 1 mM DTT). Tau concentrations were determined by the Bradford method (Bio-Rad protein assay).
In Vitro Microtubule Dynamics-Determining C c in order to determine the critical concentration for MT assembly, C c , tubulin (30 M) was assembled at 37°C in PEM or PEM ϩ DTT buffer containing 1 mM GTP in the absence or presence of 3 M tau. MTs were subsequently diluted to indicated concentrations, further incubated at 37°C, and sedimented by ultracentrifugation at 100,000g av for 15 min at 37°C (Beckman TLA 100 rotor, Beckman Instruments). The protein concentration in the resulting supernatants (termed C s ) was determined by the method of Bradford, using known concentrations of tubulin as standard. The contribution of the MAP to the total amount of protein was found to be 5%. C c is determined by plotting C s against the total tubulin concentration, C t , and extrapolating the plateau to C t ϭ 0 M, and the slope gives an indication of the fraction of assembly incompetent tubulin (47).
In our previous publication (30), tubulin (C t ϭ 15 M) was incubated with or without 3R (1.2 M) in PEM buffer for 60 min at 37°C. We determined the polymer mass, C p , by sedimenting the polymeric material by ultracentrifugation at 100,000g and determining the remaining tubulin concentration in solution, C s . C p was calculated as C p ϭ C t Ϫ C s . Single point measurements of the critical concentration are always referred to as "apparent" critical concentration, C c(app) . The quoted C c(app) values under "Results" are therefore the measured C s values of Ref. 30.
Electron Microscopy-Samples were taken following incubation of 21 M tubulin and 2.1 M tau prior to centrifugation. Samples were fixed in PEM buffer containing 0.25% (v/v) glutaraldehyde and were mounted on Formvar/carbon-coated copper grids (200 mesh) (Ted Pella Inc. Redding, CA), rinsed with cytochrome c (1 mg/ml) and then negatively stained using uranyl acetate (1% w/v). Images were observed using a Jeol 1200 EX at 80 kV. For the EM analysis using colloidal gold labeling, samples were taken following incubation of 21 M tubulin and 2.1 M tau for 30 min at 37°C, mounted onto grids as described above, and the grids incubated in PBS, followed by 0.1 M ammonium chloride/ PBS and then 0.5% (v/v) cold-water fish skin gelatin/PBS. The grids were then incubated with a polyclonal antibody to tau, TP70 (43), and a monoclonal antibody to ␣-tubulin (clone DM1A, Sigma) both diluted 1:100 in 0.5% (v/v) fish skin gelatin/PBS, followed by incubation with gold-conjugated goat anti-mouse IgG (10 nm) and anti-rabbit IgG (5 nm) (British Biocell, Cardiff, UK) both diluted 1:50 into 0.5% (v/v) fish skin gelatin/PBS. After washing, the grids were transferred to fixative (0.5% (v/v) glutaraldehyde/PBS) and either positively or negatively stained using uranyl acetate (1% w/v).

Microtubule Length Distribution and Growth
Rates-The effect of 3R and 4R on the growth rate of individual MTs (using EGS crosslinked MT fragments as seeds) was measured by dark-field video microscopy as described previously (27).
Assembly studies of tubulin in the absence or presence of tau were carried out in a Varian Cary 3E-UV-visible spectrophotometer at 37°C. The turbidity of the solutions was monitored at 350 nm. Aliquots were taken and diluted in 0.25% (v/v) glutaraldehyde/PEM to fix assembled microtubules. The length distribution within each population was determined by dark-field microscopy (27). The corresponding MT polymer mass, C p , at steady state of assembly was measured by ultracentrifugation of MT samples assembled under similar conditions, and C p was calculated as C p ϭ C t Ϫ C s . The microtubule number concentration, C n , is calculated as C n ϭ C p /ϽLϾ ϫ 1625, where ϽLϾ is the average MT length and 1625 is the number of tubulin dimers per m of microtubule (see Ref. 30 for more detail).
For immunofluorescence microscopy the cells were washed, treated, and stained for tubulin using the monoclonal antibody against ␣-tubulin (DM1A, Sigma) as has been described previously (50). Staining using the monoclonal antibody against acetylated tubulin (611B1, Sigma) and the polyclonal antibody against tau (anti-rabbit tau, Dako, Glostrup, Denmark) was carried out at a dilution of 1:200 and 1:1000, respectively, followed by incubation with a secondary fluorescein-conjugated goat anti-rabbit antibody and a Texas Red-conjugated horse anti-mouse antibody, both at a dilution of 1:200 (Vector Laboratories, Burlingame, CA). Cells were viewed by fluorescence microscopy (Zeiss Axioskop), and images were collected via a CCD camera (Princetown Instruments) and displayed using Metamorph image analysis software. Tau immunofluorescence was quantitated by scoring cells in the following ways: 1) normal filamentous tau staining; 2) partial filamentous tau staining; and 3) no filamentous tau staining. Approximately 200 cells/ well were scored by an observer who was blind to the treatments, and the data were averaged from 3 to 4 independent experiments. Statistical analyses of the effect of oxidative stress compounds on filamentous tau staining were performed using one-way ANOVA.
For assessment of tau aggregation, cells were treated with the oxidative stress agents as described above, lysed in Laemmli sample buffer (without DTT (51)), and analyzed by 7.5% (w/v) SDS-PAGE and Western blotting probing with the polyclonal tau antibody, TP70 (43).

RESULTS
The Behavior of Tau Isoforms in Vitro-The behaviors of the smallest recombinant tau isoform containing three C-terminal repeat domains (3R) and an adult four-repeat tau isoform (4R) in regulating MT dynamics in vitro were investigated. These isoforms were chosen in order to extend the observations of our previous study on the effects of phosphorylation of 3R by glycogen synthase kinase-3␤ (30) and to enable comparison with the work of others (for example see Refs. 13-22, 26, and 28).
The Effect of 3R on MT Nucleation and Bulk MT Assembly in the Absence of Reducing Agents-Anomalous behavior of 3R in PEM was initially observed when tubulin was assembled in the presence of 3R in PEM or when 3R was added to steadystate MTs, giving a subsequent increase in turbidity that did not correspond to an increase in MT polymer mass (data not shown). Independent measurements of the apparent critical concentration, C c(app) , after ultracentrifugation of the MTs did not reveal a reduction in C c(app) in the presence of 3R in PEM (C t ϭ 15 M, [3R] ϭ 1.2 M, see Ref. 30 and "Experimental Procedures"). Thus, in the absence of tau, This suggested factors other than MT formation, such as e.g. oligomerization, were leading to an increase in turbidity. Incubation of 3R in the absence of tubulin, either in the presence or absence of DTT, did not give an increase in turbidity, hence any self-aggregation of 3R alone was not sufficient to explain the observed turbidity increase. Thus, measuring turbidity of MTs in the presence of tau highlights the need for a detailed knowledge of the assembly mechanism, including the particles involved and their scattering properties.
By taking a sample of the assembled bulk MT population, the average MT length of an assembled MT population can be determined by dark-field microscopy (see Ref. 30). Following determination of the MT polymer mass, the MT number concentration can be determined and is indicative of nucleating ability of the conditions employed (see under "Experimental Procedures" for derivation of parameters). Our previous studies highlighted that, under the conditions employed, 3R in PEM was a potent promoter of MT nucleation (Ref. 30 Comparable experiments carried out in the presence of 4R in PEM yielded C n ϭ 0.26 Ϯ 0.06 nM (based on three separate observations) indicating that the 3R isoform has a similar, if not stronger, ability to nucleate MT assembly as the 4R isoform.

The Effects of 3R and 4R on the Critical Concentration of Tubulin for MT Assembly and the Generation of Tau-Tubulin
Aggregates-The failure of 3R to reduce the C c(app) as documented in Utton et al. (30) is contradictory to perceptions on the effects of MAPs on MTs; therefore, we investigated the effect of 3R on the critical concentration of MTs in further detail. This investigation included comparison with the effects of the largest 4R isoform under the conditions of our previous study (30), in addition to an investigation into the effects of the buffer conditions employed. The critical concentration of MT assembly was determined by measuring the apparent critical concentration (as given by the concentration of soluble tubulin, C s , after ultracentrifugation from steady state) as a function of the total tubulin concentration ( Fig. 1). Regression analysis of the plateau values of the concentration of soluble tubulin, C s , plotted against the formal tubulin concentration, C t , gave in the absence of tau in PEM C c ϭ 5.6 Ϯ 1.2 M (slope ϭ 0.078 Ϯ 0.029); in the absence of tau in PEM ϩ DTT, C c ϭ 5.3 Ϯ 1.2 M (slope ϭ 0.047 Ϯ 0.046); in the presence of 3R in PEM, not determined (observed effect on C c does not represent typical MT assembly, as is confirmed by the lack of MTs in the sample when viewed under EM, hence the derivation of C c is not applicable); in the presence of 3R in PEM ϩ DTT, C c ϭ 3.6 Ϯ 1.9 M (slope ϭ 0.089 Ϯ 0.064); in the presence of 4R in PEM, C c ϭ 4.6 Ϯ 1.2 M (slope ϭ 0.066 Ϯ 0.039); and in the presence of 4R in PEM ϩ DTT, C c ϭ 3.4 Ϯ 0.6 M (slope ϭ 0.099 Ϯ 0.023) (C c values calculated from regression analysis on the raw data).
There was a slight but insignificant difference between the behavior of tubulin alone in the presence or absence of DTT (Fig. 1a). It is clear that both 3R in PEM ϩ DTT and 4R in PEM, or PEM ϩ DTT, reduced the C c for MT assembly. The absence of DTT, however, appears to affect the MT-stabilizing properties of both tau isoforms. Thus, the ability of 4R to lower the critical concentration for MT assembly was reduced in the absence of DTT, whereas in the presence of 3R there was a . 30 min after the start of selfassembly, the samples were diluted to the indicated concentrations, C t , and samples were incubated at 37°C for a further 30 min before ultracentrifugation at 100,000g av at 37°C for 15 min. The tubulin concentration in the supernatant, C s , is plotted as a function of C t , and C c is determined by extrapolation of the plateau to C t ϭ 0 M. d, data shown are from a single representative experiment using an identical tubulin sample; hence any inactive tubulin was identical in all samples; f, control in PEM; E, ϩ 3R in PEM ϩ DTT; q, ϩ 3R in PEM; OE, ϩ 4R in PEM. Regression analysis of the plateau values is given in the text. tendency for anomalous behavior to occur, indicated by the marked increase in slope of C s against C t . Fig. 1d shows the results of a parallel experiment on the same tubulin sample. This clearly indicates that the observations cannot be explained by the state of the tubulin or its sensitivity to oxidation (52). The tubulin preparations used in this study, in the absence of exogenous MAPs, behaved as expected, in that the tubulin readily self-assembled into MTs at C t Ͼ C c(app) . It is likely that the anomalous observations with 3R are related to the fact that tau and tubulin were allowed to react at high concentrations after which the stock solution was gradually diluted to the indicated concentrations. As mentioned earlier, when 1.2 M 3R is added to preassembled MTs in PEM buffer at steady state of assembly (C t ϭ 15 M), MTs remain and the apparent critical concentration is hardly affected. In other words, in the case of 3R in PEM the results in Fig. 1c most likely show a kinetic side effect of the nucleation of MTs rather than a true steady-state balance.
To investigate the anomalous behavior of 3R in PEM in further detail, analysis by EM was carried out (Fig. 2). Samples assembled in the absence of tau typically showed MTs of varying lengths (Fig. 2a). MTs assembled in the presence of 3R in PEM ϩ DTT (Fig. 2b (large arrow)) or 4R in the absence (Fig.  2d) or presence (data not shown) of DTT show the existence of MT bundles. Immunogold labeling demonstrated that 3R (Fig.  2b, inset, 5 nm gold particles) was associated with MTs (large arrowhead, Fig. 2b, inset, 10 nm gold particles), much of which was seen in discrete clusters along the MT surface (small arrow, Fig. 2b, inset). Clustering of tau on the microtubule surface has also been implicated recently by Ackmann et al. (36). By contrast Fig. 2c shows amorphous aggregates generated in the presence of 21 M tubulin and 2.1 M 3R in the absence of DTT (arrows). The inset shows a large aggregate (open arrow), co-labeled by both the tubulin (large arrowhead) and the tau antibodies (small arrows). At a lower 3R concentration in the absence of DTT, however, 3R did promote the nucleation and assembly of tubulin into normal MTs (Fig. 2e, also see Figs. 3 and 4). Interestingly, Fig. 2e shows MTs and small structures that are reminiscent of the nucleation constructs postulated by Erickson and Pantaloni (53), suggesting that a major effect of 3R on MTs may be related to the nucleation of the MT lattice (see below).
Recombinant tau samples used for the microtubule assembly studies were analyzed by SDS-PAGE (using no reducing agent in the SDS sample buffer (data not shown)). Both 3R and 4R preparations stored in the absence of DTT show similar levels of high molecular weight tau species (100 kDa), which reduced on incubation with DTT. Pre-aggregation differences between 3R and 4R are therefore unlikely to be the cause for the anomalous behavior of 3R under certain non-reducing conditions.
Overall, the above observations reveal that, at sufficiently high concentrations of tubulin (C t Ͼ 20 M) and in the absence of DTT, 3R can behave in a manner that is highly disruptive to MT assembly, producing aggregates, composed of tau and tubulin, at the expense of MT formation. At lower concentrations of tau and tubulin (C t ϭ 15 M), 3R appears to function normally as MTs were formed in the presence of 3R in the absence of DTT. 4R, however, appeared to be less affected by the alteration in reducing conditions, generating MTs at all concentrations tested in the absence or presence of DTT.
The Different Effects of 3R and 4R on Individual MT Behavior-The above observations show that both tau isoforms have a major effect on the nucleation of MTs. In addition to the effects on C c , the effect of tau on MT growth and dynamic behavior at steady state was further examined by dark-field video microscopy. These experiments were largely intended to obtain a qualitative view of the effects of 3R and 4R on the growth and dynamic behavior of single MTs and therefore should not be seen as a full characterization of the effects of tau on MT dynamic instability. Typical length versus time growth profiles are shown in Fig. 3. Without the presence of a stabilizing MAP, dynamic instability was observed at concentrations up to 10 M tubulin (see profiles for control Ϫ DTT (in the absence of tau) at 8 and 10 M). By increasing the concentration of tubulin to 13 and 16 M, dynamic instability was overcome (data not shown). In the presence of 3R under non-reducing conditions, dynamic instability was still observed at 8 M tubulin (molar ratio of tau:tubulin, 1:10), hence it would appear that the stabilizing ability of 3R under these conditions was modest. At 10 M tubulin in the presence of 3R in PEM, dynamic instability was suppressed. By contrast in the presence of 3R in PEM ϩ DTT or 4R in PEM, dynamic instability was suppressed at 8 M tubulin, shown by the continuous growth profiles. In the presence of DTT, growth of single MTs was observed at as low as 4 -4.5 M tubulin. These data show that the stabilizing ability of 3R and 4R is different under non-reducing conditions. The stabilizing ability of 3R was compromised under non-reducing conditions, even though these were conditions where non-productive aggregation was not prevalent, whereas the stabilizing ability of 4R was still significant under the same conditions. Considering the observed potential of 3R to disrupt MT assembly, an interesting phenomenon was seen when the MT growth rates, R g , were measured as a function of [tau] at C t ϭ 10 M (Fig. 4). In agreement with earlier reports (14), the MT growth rate was increased in the presence of 3R, both in PEM and in PEM ϩ DTT, up to a concentration of 0.35 M 3R. Although in the presence of DTT growth rates could no longer be measured owing to extensive production of self-assembling MTs obscuring visualization of the MT ends, in the absence of DTT individual MT ends were still measurable. No self-assembly of MTs was observed above 0.35 M 3R, but the solution contained particulate matter, most likely 3R-tubulin aggregation similar to those illustrated in Fig. 2c. On further addition of 3R in PEM, up to 1.2 M, R g reverted to that in the absence of tau. Therefore, under non-reducing conditions, 10 M tubulin, and concentrations of 3R above 0.35 M, both MTs and 3R-tubulin aggregates are likely to co-exist. Aggregation effectively reduces the concentration of tubulin and tau available to participate in MT growth; hence the reversion of the MT growth rate. Such effects could eventually lead to a displacement of the normal polymer-dimer balance of the MT system toward the disassembled state. The latter possibility is supported by the tendency toward a higher value of C c(app) upon the addition of 3R to preassembled MTs (30) and confirmed by a small, but significant, increase of C c(app) upon further addition of 3 M 3R to MTs at steady state. 2 Overall, the in vitro results showed a striking potential of the 3R isoform to disrupt MT dynamics, whereas the 4R isoform seems to be less prone to such effects. It would appear that the underlying cause of this effect resides in the ability of tau to nucleate the MT lattice, and it appears there are conditions where this process can malfunction. The results also indicate that oxidative conditions may be an important factor in modulating the behavioral effects of tau on MT dynamics. The latter observation may not be conclusive, however, because the above results are not a full characterization of the mechanism and the thermodynamics of the interaction between tubulin and tau. Therefore, we decided to test the possibility that 3R and 4R are affected differently by oxidative stress in vivo.
The Behavior of Tau Isoforms in Vivo-CHO cells stably expressing either 3R (0N, 3R) or 4R (0N, 4R) were treated with known oxidative stress agents, menadione and HPT. Menadione causes oxidative stress by generating reactive oxygen species through its redox cycling, and these free radicals are detoxified subsequently at the expense of intracellular thiol homeostasis (34). HPT decomposes on irradiation to generate hydroxyl radicals (48). Following immunofluorescence staining for tau and total tubulin, differences were easily observed between cells expressing 3R and those expressing 4R. Representative images and the corresponding quantification of the oxidative effect is shown in Fig. 5 (top panel).
Wild-type CHO cells stained for ␣-tubulin showed a filamentous MT network (Fig. 5a). CHO cells stably expressing either 3R or 4R showed characteristic bundling of MTs (54), showing co-localization of tau and tubulin staining as expected. Fig.  5, b and d, shows filamentous 3R staining, co-localizing with ␣-tubulin (c) and acetylated tubulin (e). Fig. 5, f and h, shows filamentous 4R staining, co-localizing with ␣-tubulin (g) and acetylated tubulin (i). Wild-type CHO cells, stained for tubulin following HPT treatment, still show the presence of a filamentous MT network (Fig. 5j), indicating that the oxidative stress 2 A. Vandecandelaere, unpublished data.  Fig. 5k illustrates an example of partial filamentous tau staining, indicating that a proportion of the tau had been removed from the microtubules. Fig. 5l shows no filamentous tau staining, only cytoplasmic staining. The corresponding ␣-tubulin staining of these cells shows tubulin in a filamentous network (Fig. 5m). To determine the presence of a stable MT network under these conditions, 3R-expressing cells were also stained for tau (n) and acetylated tubulin (o), which confirms the presence of a stable MT network. It therefore appears that the tau was no longer bound to the MTs, giving the cytoplasmic tau staining. On the other hand, HPT treatment of CHO cells stably expressing 4R appeared to maintain tau-tubulin colocalization in many of cells (Fig. 5, p-s). Filamentous 4R staining is shown in Fig. 5, p and r. Staining these cells for ␣-tubulin (q) and acetylated tubulin (s) reveals the presence of a stable MT network. Similar immunofluorescence images were observed with treatment of menadione at concentrations of 5 and 10 M (images not shown, data quantified in Fig. 5b). Menadione at concentrations of 200 M is known to disrupt the cytoskeleton (35). The concentration used here is far lower hence no visible disruption of the MT network was observed on tubulin staining, showing that the effect on tau binding to the MTs was a more subtle effect of the menadione treatment. Thus under the oxidative conditions employed, the MTs appeared to remain intact, and 3R had a greater tendency to dissociate from the MTs than 4R.
To quantify the effects of the oxidative stress treatments (Fig. 5, bottom panel), cells were assessed according to the state of the filamentous tau staining, showing either no filamentous tau staining (example shown in Fig. 5n), partial filamentous tau staining (example shown in Fig. 5k), or total filamentous tau staining (example shown in Fig. 5r). Approximately 200 cells were assessed per group per experiment, and the mean numbers of cells in each category are shown (Ϯ S.E. pooled from 3 to 4 independent experiments). Untreated cells showed ϳ80% of normal filamentous tau staining and less than 10% with no filamentous staining, as expected for healthy untreated cells. No differences were observed for the cells expressing 3R or 4R in the absence of oxidative stress (e.g. on comparing the difference in the percentage of cells with no filamentous tau staining for untreated 3RCHO and 4RCHO cells, p ϭ 0.7, showing no statistical difference, as determined by one-way ANOVA). On treatment with 2 mM HPT, the proportion of cells under each category changes, with 3R CHO cells showing Ϸ50% with no filamentous tau staining and 4R CHO cells showing Ϸ28% with no filamentous staining. On treatment with 5 mM HPT, the difference in effect on 3RCHO and 4RCHO cells became more prominent with Ϸ70% of 3RCHO cells showing no filamentous tau staining, whereas only Ϸ40% of 4RCHO cells exhibited no filamentous tau staining. The differences shown in the percentage of cells in each category were statistically significant, for example p Ͻ 0.0005 using one-way ANOVA, when comparing the percentage of 3RCHO and 4RCHO cells with total filamentous staining following 5 mM HPT treatment. This behavior was also found with the menadione treatment (Fig. 5b), although the difference between 3RCHO and 4RCHOs was not as pronounced. These data showed that differences were observed in cells expressing 3R and 4R when placed under conditions of oxidative stress. The data suggest, as do the in vitro data, that 3R was more susceptible than 4R to conditions of oxidative stress, shown by the increased tendency of 3R to dissociate from the MT network under these experimental conditions.
Given the tendency of 3R to aggregate in vitro under certain conditions (37,38), it was of interest to determine if tau aggregates were formed in tau-expressing CHO cells under conditions of oxidative stress. Following treatment, cells were scraped directly into non-reducing sample buffer and analyzed by SDS-PAGE and Western blotting with a polyclonal antibody to tau, TP70 (Fig. 6). Untreated 3R-expressing cells (Fig. 6a,  lane 1) showed a predominant tau band at the position corresponding to monomeric tau molecular weight with virtually no higher molecular weight species with tau immunoreactivity, whereas treatment with 5 mM HPT markedly increased the amount of high molecular weight 3R species in the form mainly of a smear (Fig. 6a, lane 2). On treatment with 10 M menadione (lane 3), no increase in high molecular weight tau species was seen, whereas with 200 M menadione treatment, again extensive higher molecular weight tau species were observed (lane 4). The marked increase in higher molecular weight tau species was not seen when cells expressing 4R underwent the same treatment (Fig. 6b). Wild-type CHO cells were also treated and analyzed by Western blotting with TP70, and no difference was observed in the background staining of wild-type CHO cells under the same treatments (data not shown); thus, the higher molecular weight smears in Fig. 6a   FIG. 4. The effect of tau on individual microtubule growth rates at constant tubulin concentration. The growth rate, R g , at 10 M tubulin as a function of tau concentration, as determined from monitoring growth of several individual MTs using dark-field video microscopy. The average growth rates of fast growing ends of MTs are shown here with bars to represent the standard deviation. q, in the presence of 3R in PEM buffer; E, in the presence of 3R in PEM ϩ DTT. were not the result of nonspecific labeling by TP70 but were due to tau immunoreactivity.
These data further indicated the differences in behavior of 3R and 4R in vivo and suggested that under certain conditions 3R had a greater tendency than 4R to dissociate from the MT network and to form aggregates. DISCUSSION It is important to define the properties of different tau isoforms because there is selective involvement of isoforms in the inclusions characteristic of certain neurodegenerative diseases, and this may result from different isoforms responding to disease-specific pathogenic stimuli. The present work provides evidence that tau isoforms, containing either 3 or 4 C-terminal repeats, modulate MT properties differently and also appear affected differently by changes in the oxidative potential of the environment.
Effects of 3R and 4R on MT Dynamics in Vitro-The evidence accumulated in this and previous work (30) indicates that both 3R and 4R are potent nucleators of MT assembly, the shortest isoform possibly having the strongest effects on the nucleation of MTs. In addition, both isoforms have a significant stabilizing effect on the MT lattice as judged by the reduction of the critical concentration (Fig. 1) and the suppression of dynamic instability behavior of single MTs (Fig. 3) in reducing buffers. The stabilizing effect of both isoforms is affected by the lack of a reducing agent in the environment. In the absence of DTT, the effect of 4R on the critical concentration was reduced (Fig. 1), and 3R completely failed to lower the apparent critical concentration of preassembled MTs (30) and to suppress dynamic instability (Fig. 3). Moreover, at sufficiently high concentrations of tubulin and 3R, aggregates are produced that fail to assemble into MTs (Figs. 1 and 2). It would appear, therefore, that oxidative stress may be a major factor in determining the functional properties of tau.
Being a strong nucleator but having a weaker effect on suppression of dynamic instability, as occurs in the presence of low concentrations of 3R in non-reducing conditions, would appear to be contradictory but has also been found for MAP1B assessed in the following way: 1) no filamentous tau staining; 2) partial filamentous tau staining; 3) normal filamentous tau staining. Data show the mean number of cells analyzed as described above Ϯ S.E. pooled from 3 to 4 independent experiments. *** indicates a highly significant difference (p Ͻ 0.001), and * indicates a significant difference (p Ͻ 0.05), between the percentage of cells in each given category (i.e. no filamentous staining, partial filamentous staining or total filamentous staining) when comparing 3Rand 4RCHO cells under the same experimental conditions (one-way ANOVA test). (27). It is a logical consequence of the hydrolysis of GTP upon the assembly of tubulin into MTs and the resulting kinetic heterogeneity of the MT lattice. As a result, the actual critical concentration, C c , is raised relative to the critical concentration of the pure tubulin-GTP lattice, C c,T (often referred to as "critical concentration of elongation"), and MT nucleation is no longer related in a straightforward manner to the dynamic behavior of the MTs at steady state of assembly (see Ref. 27 for detailed explanations). Considering that MT growth is unlikely at C t Ͻ C c , the lowest value of C t at which MT growth can still be observed by video microscopy can be seen as a practical estimate of C c(app) . Applied to the data of Trinczek et al. (26) it can be deduced that these authors observed a reduction of the critical concentration from 10 M in the absence of tau to 7.8 M and 3.8 M in the presence of 1:5 and 1:1 4R:tubulin, respectively, and to 5 M in the presence of 1:2 3R:tubulin. Intrapolating the latter to 1:5 3R:tubulin yields C c(app) ϭ 8 M, indicating that both tau isoforms reduce the critical concentration to a similar extent in the presence of DTT. Thus, in qualitative terms, the observations of Trinczek et al. (26) agree with the effects of both tau isoforms on the critical concentration seen here in the presence of DTT. (Numerical differences are not uncommon between laboratories.) Inspection of the growth rates of single MTs as a function of 3R revealed an optimum in R g [tau] at a molar ratio of 0.035 (Fig. 4) under reducing and non-reducing conditions alike. Interestingly, at a similar tau:tubulin ratio, Brandt and Lee (21) observed a transition from growth promotion of existing MTs off centrosomes to promotion of the nucleation of new MTs. In PEM ϩ DTT, self-assembly of new MTs predominating over growth of pre-existing microtubules was evident above 0.04 tau:tubulin. A likely explanation for the observed optimum is therefore that 3R-tubulin oligomers are produced and, at ratios over 0.04, either accumulate or nucleate new MTs, effectively reducing the free tubulin concentration in solution.
Given that 4R has at least a 2-fold higher affinity for taxolstabilized MTs than 3R (16,36,(55)(56)(57)(58), the degree of saturation of the MTs with 3R is expected to be lower than with 4R, particularly since recent evidence (36) indicates that 3R and 4R bind with the same stoichiometry. The similarity of the effects of 3R (in PEM ϩ DTT) and 4R on C c and the similar nucleating ability of both isoforms (see Fig. 1 and Ref. 30), therefore, suggest that 3R may affect MT dynamics in a more potent manner than 4R. This could mean that 3R interacts with the MT lattice in a different manner to 4R, although current evidence suggests that the binding of single 3R and 4R molecules to the MT is essentially similar (16,56,57,59). It is also possible that the potent effect of 3R is related to an ability to bind to MTs in clusters (as seen in Fig. 2b, inset, and shown by Ackmann et al. (36)), as is seen with MAP2 (17, 60 -62), leading to localized strengthening of tubulin-tubulin interactions in the MT lattice. This hypothesis is supported by the fact that dimers of 3R have been recently reported to bind with a 10-fold increased affinity and double stoichiometry to that of monomeric 3R (Ref. 36 and see Ref. 27 for discussion of different types of lattice stabilization and their effects on MT dynamic instability). The view that emerges from our work is that particularly the shortest tau isoform facilitates the nucleation of the MT lattice by interacting with tubulin in a manner that is sensitive to and can be distorted by environmental conditions.
From the evidence presented here and from other data, it has been shown that 3R has a stronger tendency than 4R to self-aggregate under oxidizing conditions (37,38), potentially due to the presence of one cysteine in 3R allowing intermolecular bonding, whereas 4R has two cysteines allowing intramolecular bonding. Evidence also exists for a higher proportion of 3R than 4R in PHF (63,64). It is possible that, under our experimental conditions, 3R may readily self-aggregate, and the resulting 3R-3R aggregates may be able to bind to tubulin in a manner that is less appropriate for the generation of MTs. In the presence of DTT, 3R is less likely to form cysteine-dependent dimers and may therefore bind to tubulin in a more productive manner for MT formation. Assuming a more potent nucleating ability of 3R compared with 4R, when affinity for the MT lattice is considered, it is possible that any disruption of this nucleating ability of 3R may have more severe consequences on the formation of MTs than effects on 4R. The ability of 4R to form aggregates under oxidative conditions was not prevalent under our conditions but raises the question whether if such aggregation occurred, it does so productively, i.e. while retaining their ability to nucleate bona fide MTs.
Evidence by Garcia de Ancos et al. (65) suggests that large aggregates of tau can be formed under non-reducing and reducing conditions, indicating that regions other than the cysteine-containing regions could contribute to tau self-aggregation. In addition, serine to alanine mutations in 3R profoundly affect C c(app) under the same non-reducing conditions (30). Thus, disulfide bonds between 3R molecules may not be the definitive reason for the anomalous 3R behavior seen here but could be a contributing factor for 3R to behave abnormally.
Effects of 3R and 4R in a Cellular Model of Oxidative Stress-The central nervous system is especially vulnerable to free radical damage, and cumulative oxidative damage over time could account for the late life onset and slow disease progression in most neurodegenerative cases (66). Evidence also suggests that oxidative damage not only occurs to the proteins comprising the lesions of AD but also precedes lesion formation in neurons at risk of death. The oxidative state of degenerating neurons is likely to be altered due to disrupted mitochondrial activity, producing a more oxidized environment than that in healthy neurons (31,32).
Oxidative stress is known to influence tau phosphorylation (e.g. Refs. 67 and 68) and tau degradation (49), although there exists controversy over the importance of oxidation as a key event in tau aggregation and PHF formation. Certain data highlight the importance of tau oxidation/dimerization in PHF assembly (36 -39) suggesting that the redox potential in vitro and in neurons is crucial for PHF assembly. Other data, however, suggest that while oxidation may contribute to final PHF morphology, it is not a prerequisite for efficient nucleation or elongation of tau filaments (69), and further evidence suggests that cross-linked tau readily promotes tubulin polymerization so it is unlikely that tau dimers are the driving force for PHF formation (70).
Cross-linking of cytoskeletal proteins such as actin in response to oxidative stress have been reported in hepatocytes (35). Neuroblastoma cells, however, under oxidative stress conditions did not show high molecular weight aggregates of actin or tau (68). This suggests the possibility that cytoskeletal elements in different cell types can be differently affected by oxidative stress. From our in vitro data, it is highly likely that effects are dependent on the absolute and relative concentrations of tau and tubulin, which are likely to differ in different cell types due to the expression of endogenous tau or higher levels of tau due to transfection.
Elucidating how the different exons influence tau properties, including susceptibility to extraneous factors such as oxidative stress, is likely to be essential in understanding the molecular mechanisms involved in different tauopathies. In agreement with Ackmann et al. (36), we feel that if the concentration of tau is either increased relative to MTs due to MT decay or an increase in tau levels, it is possible for tubulin to act as an inducer of aggregation and that oxidation/dimerization of tau is highly important because such dimers strongly enhance PHF aggregation. If the oxidation state of neurons in AD is altered so that not only the likelihood of the removal of 3R from the MTs and its tendency to self-aggregate is increased but, in addition, that tau can sequester tubulin in a non-productive manner, along with sequestering other MAPs, as has been reported for Alzheimer-tau (71), then these processes could aid in the breakdown of the cytoskeleton seen in AD (72,73).