Tau Isoform Composition Influences Rate and Extent of Filament Formation^*

Materials

Recombinant His-tagged human Tau isoforms were prepared as described previously (23, 30). Aggregation inducer Thiazine red (Chemical Abstract Service registry no. 2150-33-6) was obtained from TCI America (Portland, OR). Formvar/carbon-coated copper grids (300 mesh), glutaraldehyde, and uranyl acetate were obtained from Electron Microscopy Sciences (Fort Washington, PA).

Fibrillization Assays

Tau preparations were incubated (37 °C) without agitation in assembly buffer (10 mm HEPES, pH 7.4, 100 mm NaCl, 5 mm dithiothreitol) in the presence of 100 μm Thiazine red inducer for up to 24 h (unless specified otherwise). Aliquots were removed at indicated time points, fixed with glutaraldehyde, and examined by transmission electron microscopy as described in Ref. 31. Adsorbed filaments >10 nm in length were quantified from at least three fields captured for each condition using Optimas imaging software (version 6.5, Media Cybernetics, Silver Spring, MD). The summed lengths of all completely resolved filaments per field are reported ± S.D.

Analytical Methods

K_crit values were estimated by linear regression analysis of the Tau concentration dependence of aggregation at interaction plateau (27, 28). The abscissa intercept (x̂) determined by inverse prediction was taken as K_crit. Higher order approximations of x̂ were obtained from the Taylor series expansion (32),

where μ_y ± σ_y is the ordinate intercept ± SEE, μ_x ± σ_x is the regression slope ± SEE, and r is the regression correlation coefficient. The variance (S²) was calculated as shown in Equation 2 (32).

To estimate dissociation rate constant k_e−, Tau filaments prepared as described above were diluted 10-fold into assembly buffer containing 100 μm Thiazine red and incubated at 37 °C. Aliquots were withdrawn as a function of time up to 5 h post-dilution and then assayed for filament length. The resultant disaggregation time series was fit to an exponential decay function to obtain k_app, the pseudo-first order rate constant describing the time-dependent decrease in filament length, and L₀, the total filament length at time zero. The rate constant k_e− was estimated from k_app, L₀, and the number of filaments at time zero as described previously (33, 34). The association rate constant for elongation, k_e+, was then obtained from the relationship (27).

Aggregation lag times were obtained by Gompertz regression of time series as described in Ref. 35.

Statistical Tests

Estimated kinetic parameters were assumed to resemble normally distributed random variables (X_i) with mean μ_i and known standard deviation s_i. As a global test of the null hypothesis H₀ (i.e. that all compared μ_i values were the same), the statistic T based on the maximum likelihood ratio test principle (32) was calculated,

where k is the number of kinetic parameters being compared, T is the 1 − α point of the Chi-square distribution having k − 1 degrees of freedom, w_i = 1/σ_i², and μ̂, the common mean under the null hypothesis, is the weighted sum of X_i.

If H₀ was true, then the probability (p) of obtaining more extreme values of T than actually observed is α.

For pairwise comparisons, the probability (p) of obtaining the observed results, assuming the null hypothesis, was assessed by z-test,

where x₁ ± S_x1 and x₂ ± S_x2 are the pair of estimates ± SEE being compared, z is the 1 − α point of the standard normal distribution, and p is 2α. All statistical analyses were carried out using JMP (version 9.0, SAS Institute, Cary, NC).

Human Tau Isoforms Differ in Aggregation Propensity

All six Tau isoforms aggregated in the presence of Thiazine red inducer, forming filaments with twisted ribbon morphology (supplemental Fig. 1). To quantify aggregation propensity of each isoform, the minimal concentration required to support filament formation in the presence of Thiazine red inducer was estimated using an electron microscopy assay for total filament length (31). In nucleation-dependent interactions, the minimal concentration is termed the critical concentration (K_crit) and approximates the dissociation equilibrium constant for elongation (K_e) (27). Results showed that K_crit values for the six Tau isoforms varied over more than an order of magnitude (Fig. 1B; Table 1). When subjected to a global statistical test (Equations 4 and 5), the null hypothesis was rejected at p < 0.0001. Furthermore, the Fieller 95% confidence intervals for any pair of isoforms did not overlap. These data indicate that naturally occurring human Tau isoforms differed in their ability to support fibrillization in the presence of Thiazine red and that our analytical methods were adequate to detect and quantify the differences.

To determine whether the observed variation in aggregation propensity resulted from differential sensitivity to Thiazine red inducer, the concentration-effect relationship for Thiazine red-mediated fibrillization was quantified for each isoform at constant Tau supersaturation (i.e. the net difference between bulk concentration and K_crit was held constant, so that the amount of aggregation at plateau was approximately the same for all six isoforms). Results showed that all isoforms resembled one another with respect to both the potency and efficacy of Thiazine red (supplemental Fig. 2). Together, these data indicate that Tau isoforms respond similarly to Thiazine red and that the observed variation in K_crit reflects intrinsic differences in aggregation propensity among Tau isoforms.

Contributions of Insert Sequences

Visual inspection of Fig. 1B revealed that 4R-containing Tau isoforms aggregated with lower K_crit than 3R isoforms. Moreover, the rank order of K_crit within 3R and 4R isoforms consistently was 0N > 2N > 1N, suggesting that the differences resulted from the presence or absence of alternatively spliced segments. To test this prediction, the ratio of K_crit values observed in the presence and absence of each alternatively spliced segment was calculated for isoform pairs. For example, pairwise comparison of 0N4R with 1N4R, and of 0N3R with 1N3R, gave two estimates of the contribution of the N-terminal segment encoded by exon 2 to K_crit, whereas comparison of 1N4R with 2N4R and of 1N3R with 2N3R provided two estimates of the effects of the exon 3-encoded sequence. Similarly, comparison of 0N4R with 0N3R, 1N4R with 1N3R, and 2N4R and 2N3R provided three estimates of the contribution of the alternatively spliced microtubule binding repeat to Tau K_crit. Replot of K_crit ratios showed that the presence of the exon 2-encoded insert lowered K_crit ∼ 2-fold in both 3R and 4R backgrounds (Fig. 1C). In contrast, exon 3-encoded sequence increased K_crit, with stronger effects in 3R than in 4R background (Fig. 1C). The largest effects, however, associated with the exon 10-encoded microtubule repeat, which on average lowered K_crit 6-fold (Fig. 1C). These data indicate that aggregation propensity is increased by exon 2- and exon 10-encoded segments but antagonized by the segment encoded by exon 3 in a manner that is sensitive to the number of microtubule binding repeats.

Mechanism of Critical Concentration Effects

K_crit approximates the ratio of dissociation (k_e−) and association (k_e+) rate constants for filament elongation (Equation 2). Thus, modulation of K_crit may result from changes in k_e− (i.e. filament stability), in k_e+ (i.e. the efficiency of monomer association with filament ends), or both. To distinguish these possibilities, k_e− was estimated for each Tau isoform by diluting preassembled filaments below K_crit and estimating the initial rate of filament shortening in the electron microscopy assay. Loss of filament length followed first order kinetics as predicted for endwise depolymerization from a Poisson-like length distribution (27, 34) (supplemental Fig. 3). On the basis of the relationship between Tau mass and filament length established for wild-type 2N4R Tau (27), the dissociation elongation constant k_e− was derived from the disaggregation rate of each isoform. Rate constant k_e+ was then calculated from estimates of k_e− and K_crit for each isoform through Equation 2. Estimated k_e− and k_e+ values for all isoforms are summarized in Table 1. Pairwise comparisons among isoforms showed that neither exon 2- nor exon 3-encoded segments exerted their effects at the level of dissociation rate constant k_e− (Fig. 2). Rather, the aggregation promoting effects of the exon 2 encoded insert resulted from promotion of extension rate through increase in k_e+, whereas the additional presence of exon 3 partially antagonized this effect (Fig. 2). In contrast, the strong aggregation promoting effects of the fourth microtubule binding repeat resulted from coordinated depression of k_e− and increase in k_e+ (Fig. 2). These data indicate that alternatively spliced segments modulate K_crit through differing mechanisms that can synergize or compete with one another.

Isoform Structure Influences Nucleation Rate

In the presence of Thiazine red inducer, Tau aggregation approximates an equilibrium nucleation-elongation interaction (27), where assembly-competent monomer rapidly equilibrates with a thermodynamically unstable species termed the nucleus (36). Once the critical nucleus cluster size is reached, subsequent additions to the nascent filament ends are favorable energetically, and elongation proceeds efficiently. As a result, aggregation rate depends not only on the rate of filament elongation (k_e− and k_e+) but on the efficiency of the nucleation step as well. To assess the effects of primary structure on nucleation rate, the time course of aggregation was quantified for each Tau construct at constant supersaturation. Under these conditions, differences in interaction rates primarily reflect differing rates of nucleation and protein concentrations (37). All resultant progress curves were sigmoidal with lag, exponential growth, and equilibrium phases (Fig. 3A). Lag times, which vary inversely with nucleation rate (38), were obtained ± SEE after fitting each time series to a three-parameter Gompertz growth function as described under “Experimental Procedures.” Resulting values ranged from 0.23–1.36 h, with 4R isoforms having the shortest lags (Table 1). To determine the contribution of each alternatively spliced segment to nucleation rate, the ratio of lag times observed in the presence and absence of exons 2, 3, and 10 encoded segments was calculated and averaged for isoform pairs. Replot of lag time ratios showed that the presence of a fourth microtubule binding repeat depressed lag time >4-fold at p < 0.01 (Fig. 3B), indicating that nucleation rate was promoted strongly by this segment. N-terminal inserts encoded by exons 2 and 3 also weakly decreased lag time (Table 1), but their effects did not reach statistical significance at p < 0.05 in pairwise comparisons (Fig. 3B). However, when 0N, 1N, and 2N isoforms were analyzed in separate 3R and 4R backgrounds, linear correlations between lag time and net Tau charge at assay pH were apparent (Fig. 4). In contrast, no correlation was detected (r² < 0.2) when isoform lag time was plotted against grand average of hydropathy (calculated by the Kyte and Doolittle algorithm; data not shown). These data are consistent with all three alternatively spliced segments accelerating nucleation rate, with the weak effects of the acidic N-terminal inserts potentially being mediated by decreases in isoelectric point and net protein charge (39).

4R Tau Can Drive Aggregation of 3R Tau

The above experiments show that 4R Tau isoforms aggregate more efficiently and at faster rates than 3R isoforms. Because human brain expresses both 3R and 4R isoforms in certain neurons (11), the potential for isoform co-assembly (40) exists. To test whether 4R Tau can promote aggregation of 3R Tau, the critical concentrations of 2N4R/0N3R mixtures were determined at varying isoform ratios. These two isoforms were chosen for analysis because they differed by nearly an order of magnitude in K_crit (Table 1). Measured K_crit values were then compared with theoretical values for a fully dominant interaction (i.e. where all 3R Tau aggregated like 4R Tau) or no interaction (i.e. where K_crit values reflected the individual, noninteracting 3R and 4R components). Electron microscopy was used as the assay modality so that measured K_crit values corresponded to filament formation rather than nonspecific trapping of monomers during aggregation. At high (3:1) 4R:3R ratios, K_crit approximated the dominant interaction scenario, although it differed from the interaction scenario only at p = 0.07 (Fig. 5). At equimolar and low (1:3) 4R:3R ratios, however, K_crit shifted to a value intermediate between the fully dominant and non-interacting scenarios (statistical differences when compared with both boundary conditions: p < 0.05 at 1:1 4R:3R ratio, and p < 0.01 at 1:3 4R:3R ratio; Fig. 5). These data indicate that the high aggregation propensity of 4R Tau can be partially dominant over 3R Tau, suggesting that the 4R species can recruit less assembly prone Tau isoforms into aggregates in the submicromolar free Tau concentration regime.

Correlation of Aggregation Propensity with Disease

Animal models have shown that neurodegenerative phenotypes can result from high level overexpression of Tau sequences irrespective of filament formation (49, 50). In sporadic AD, however, total levels of all Tau transcripts do not increase relative to cognitively normal cases (51), and neurodegeneration correlates both spatially and temporally with Tau aggregation (52, 53). The discovery of familial forms of FTLD linked to MAPT (reviewed in Ref. 9) has further extended the correlation between Tau aggregation and disease to the level of Tau isoform composition. Intronic mutations that increase exon 10 inclusion are associated with Mendelian inheritance of certain forms of FTLD (54), whereas sub-haplotypes of H1 that increase both exon 10 inclusion and total Tau transcript levels (55) are associated with increased risk for PSP and corticobasal degeneration (6–8). Shift in isoform distribution toward 4R forms also has been observed in mild cognitive impairment and AD (51, 56, 57). On the basis of aggregation propensity investigated herein, exon 10 inclusion will bias Tau isoform distribution toward aggregation-prone species, whereas increases in bulk Tau concentration will accelerate all association steps in the pathway. In particular, the strong increase in aggregation propensity associated with the fourth microtubule binding repeat more than compensates for the depletion of 1N isoforms in AD (56). In contrast, haplotype H2, which is protective against PSP (7, 8), supports increased inclusion of exon 3 as well as decreased inclusion of exon 10 relative to H1 (47, 55). Moreover, levels of 2N forms of Tau are depressed in the Tau aggregates that accumulate in PSP (58–60). These data are consistent with the lower aggregation propensity observed with the insert encoded by exon 3. Overall, the aggregation behavior of naturally occurring Tau isoforms is consistent with the law of mass action in this subset of tauopathies (61).

The aggregation behavior identified herein for unmodified Tau monomers does not account for tauopathies that accumulate aggregates composed primarily of 3R Tau (reviewed in Ref. 1). In myotonic dystrophy, 3R aggregation reflects primarily an abnormal MAPT expression pattern favoring exclusion of exons 2, 3, and 10 (62). In contrast, 3R aggregation in Pick disease occurs in certain cell populations (63), perhaps reflecting selective vulnerability of cells expressing primarily 3R Tau (e.g. neural precursor cells (64)). Post-translational modifications that selectively promote 3R Tau dimerization, including disulfide (21, 65) and dityrosine bond formation (66), may foster 3R Tau aggregation in these cells. These examples indicate that native aggregation propensity is not the only factor underlying neurofibrillary lesion formation in tauopathic neurodegenerative diseases.

In summary, we have provided evidence that naturally occurring Tau isoforms differ in aggregation propensity owing to the contributions of inserts encoded by alternatively spliced exons 2, 3, and 10. The segments differentially modulate rate-limiting steps in the aggregation pathway and can synergize with the effects of missense and pseudophosphorylation mutations. The aggregation propensity of Tau isoforms may influence how Tau misfunction leads to clinically and histopathologically distinct diseases.