Pseudophosphorylation and Glycation of Tau Protein Enhance but Do Not Trigger Fibrillization in Vitro*

Alzheimer's disease is defined in part by the intraneuronal aggregation of tau protein into filamentous lesions. The pathway is accompanied by posttranslational modifications including phosphorylation and glycation, each of which has been shown to promote tau fibrillization in vitro when present at high stoichiometry. To clarify the site-specific impact of posttranslational modification on tau fibrillization, the ability of recombinant full-length four repeat tau protein (htau40) and 11 pseudophosphorylation mutants to fibrillize in the presence of anionic inducer was assayed in vitro using transmission electron microscopy and laser light scattering assays. Tau glycated with d-glucose was examined as well. Both glycated tau and pseudophosphorylation mutants S199E, T212E, S214E, double mutant T212E/S214E, and triple mutant S199E/S202E/T205E yielded increased filament mass at equilibrium relative to wild-type tau. Increases in filament mass correlated strongly with decreases in critical concentration, indicating that both pseudophosphorylation and glycation promoted fibrillization by shifting equilibrium toward the fibrillized state. Analysis of reaction time courses further revealed that increases in filament mass were not associated with reduced lag times, indicating that these posttranslational modifications did not promote filament nucleation. The results suggest that site-specific posttranslational modifications can stabilize filaments once they nucleate, and thereby support their accumulation at low intracellular tau concentrations.

Especially dramatic among these is phosphorylation, which achieves significantly greater stoichiometries in filamentous tau (averaging 7-8 mol/mol) relative to normal tau (averaging 2-3 mol/mol) (8,9). The phosphate is distributed into ϳ30 sites, most of which are clustered into two regions flanking the microtubule-binding domain (10,11). Because few of these sites are unique to filamentous tau, hyperphosphorylation mostly reflects increased occupancy of phosphorylation sites found in normal tau (12,13). At least 15 of these sites fill in hierarchical fashion as neurofibrillary lesions mature, suggesting a sequential correlation between site occupancy and neuritic lesion development (14,15). In addition to serving as a marker for diseased neurons, hyperphosphorylation neutralizes the microtubule binding activity of tau, yielding a cytosolic population available for aggregation (16 -19). Phosphorylation may also directly promote fibrillization, but in vitro this reaction requires stoichiometries of phosphorylation well above average levels observed in filamentous tau (20,21). Part of the experimental challenge stems from the diversity of tau phosphorylation sites, each of which may enhance or depress fibrillization (22), and the difficulty of recapitulating the precise pattern found in disease. To overcome these limitations, and to investigate the contribution of individual sites, phosphorylation has been mimicked by mutation of hydroxyamino acids into negatively charged Asp or Glu residues. The resultant "pseudophosphorylation" mutants imitate phosphorylation-induced changes in the structure and microtubule-binding function of tau (23)(24)(25)(26), react with many phosphoepitope-selective antibodies (27), and reportedly can promote or inhibit tau fibrillization depending on the sequence context investigated (24,25,27,28). Nonetheless, the mechanisms through which phosphorylation or phosphorylation mimicry modulates fibrillization have not been explored.
In addition to being phosphorylated, AD-derived tau protein is glycated (5, 29 -31). As with phosphorylation, glycation has been implicated in events that may contribute to tau fibrillization and neuronal death (32,33). In vitro glycation of tau can modify at least 13 sites (34), decrease its affinity for tubulin (5,30,35), and promote fibrillization (5,30,31). Again the mechanisms underlying the effect on tau aggregation have not been investigated.
Fibrillization of purified recombinant tau preparations is slow and essentially undetectable when assayed over a period of days (38,39), but can be greatly accelerated by the addition of exogenous inducers. Among these, anionic surfactants (including fatty acids) have emerged as especially useful agents because of their efficacy with full-length tau protein under near-physiological conditions of temperature, pH, ionic strength, reducing environment, and tau concentration (38,40). Anionic surfactants act in micellar form to stabilize assembly competent intermediates on their surfaces, which then accumulate to sufficient concentrations so that the barrier to filament nucleation is overcome (41). Using this approach, differences in fibrillization efficiency among individual tau isoforms can be quantified in as short as 5 h (39).
Here we extend the approach by investigating the fibrillization properties of glycated tau and pseudophosphorylation of tau mutants that mimic occupancy of sites known to be filled in disease (14) using quantitative transmission EM and static light scattering assays. The results show that glycation and phosphorylation mimicry influence tau aggregation through mechanisms that include filament stabilization.

EXPERIMENTAL PROCEDURES
Materials-Stocks of alkyl sulfate detergent C 18 H 37 SO 4 Na (Research Plus, Bayonne, NJ) were prepared in 1:1 H 2 O/isopropyl alcohol. AA (Cayman Chemicals, Ann Arbor, MI) was dissolved in ethanol (333 mM) and stored at Ϫ80°C until used. Dithiothreitol was from Sigma, Dglucose was from Fisher Chemicals (Fair Lawn, NJ), and uranyl acetate, glutaraldehyde, and 300-mesh copper formvar/carbon-coated grids were from EM Sciences (Fort Washington, PA).
Tau Purification-Recombinant wild-type tau protein (htau40) and all tau mutants were purified as described previously (42). The quality of all preparations was assessed by SDS-PAGE, and final protein concentrations were calibrated on the basis of optical density (42).
Tau Glycation-Aliquots of recombinant C291A/C322A and assembly incompetent I277P/I308P double mutants (20 M) were incubated in the presence and absence of 10 mM D-glucose in assembly buffer (10 mM HEPES, 100 mM NaCl, pH 7.4) for 7 days, at 37°C, in the presence of protease inhibitors (aprotinin and leupeptin; 0.02 mg/ml each) and 0.02% NaN 3 (31). To determine aggregation behavior, aliquots of glycation reactions were diluted to 4 M tau and subjected to the tau aggregation assays described below. Where indicated, glycation products were separated from unincorporated glucose by desalting chromatography (Bio-Rad Econo-Pac 10DG) prior to dilution and assay.
To estimate glycation stoichiometry, tau was glycated as described above except that radiolabeled D-[U-14 C]glucose (10 Ci/mol) served as carbohydrate donor. Aliquots of the glycation reaction were removed as a function of time and precipitated with trichloroacetic acid (15% final concentration, 15 min on ice; Ref. 45). Trichloroacetic acid pellets were harvested by centrifugation (15,800 ϫ g for 10 min at 4°C), washed five times with 1 ml of 10 mM D-glucose, 15% trichloroacetic acid, then resuspended in 1 M NaOH and subjected to scintillation spectroscopy. Protein was estimated in parallel using the Coomassie Blue binding method with recombinant htau40 as standard (19).
Tau Aggregation Assays-All tau preparations (4 M) were incubated (25°C) in assembly buffer containing dithiothreitol (5 mM) and either AA (75 M) or C 18 H 37 SO 4 Na (50 M) inducer. Reaction products were analyzed by two methods. For analysis by transmission EM (46), aliquots (30 l) were removed, fixed with glutaraldehyde (2%), and adsorbed (1 min) onto 300-mesh formvar/carbon-coated copper grids. The grids were washed with water, stained (1 min) with 2% uranyl acetate, washed again with water, blotted dry, and visualized using a Phillips CM 12 microscope operated at 65 kV. Two to 10 random images from each experimental condition were captured on film at 22,000-fold magnification, digitized, calibrated, and imported into Optimas 6.5.1 for quantitation of filament length and number as described previously (46). Interfacial filament concentration (⌫ f ) is defined as the summed lengths of all filaments Ͼ50 nm in length per unit area of grid surface, and is reported in units of length/field Ϯ S.D. or as a percentage of wild-type htau40 control values Ϯ S.D.
Right angle laser light scattering (LLS) assays were performed as described previously (47), except that the contribution of micellization to scattering intensity was controlled in parallel reactions containing htau40 I277P/I308P , an assembly incompetent mutant of htau40 (48). When employed as light scattering control, htau40 I277P/I308P was preincubated in exact parallel with experimental conditions (including glycation and desalting procedures described above). We refer to this assay as the corrected LLS assay (43). Aggregation reactions (200 l) were transferred to quartz cuvettes (3 ϫ 3-mm internal diameter) and illuminated with light from an argon laser ( ϭ 488 nm; Edmund Scientific). The scattered light was captured at an angle of 90°to the incident light using an EDC1000HR digital camera operated at an aperture of f8 and exposure times ranging from 50 to 400 ms (47). Captured images were imported into Adobe Photoshop 5.0 and the intensity of scattered light was estimated as described previously (47). The scattering intensity captured at zero exposure time (dark current) was subtracted from all measurements, which were then normalized for exposure time so that the arbitrary units of light scattering intensity reported always corresponded to a 200-ms exposure at aperture f8. Scattering intensities from parallel reactions containing htau40 I277P/I308P were then subtracted from all values to yield I n (90°), the net right angle light scattering intensity. I n (90°) values are directly proportional to ⌫ f values determined by the transmission EM assay (43).
Assembly-critical Concentration Determination-Tau samples (2-4.5 M, final concentrations) were added from serially diluted stocks to assembly buffer at room temperature and incubated for 26 h in the presence of 50 M anionic surfactant C 18 H 37 SO 4 Na and in the absence of agitation. Samples were analyzed by laser light scattering as described above. The net intensity of scattered light corresponding to fibrillization was plotted against tau concentration. Critical concentration for assembly was estimated from the abscissa intercept after least squares linear regression and expressed as Ϯ S.E. of the estimate.
Analytical Methods-Sigmoidal reaction progress curves were fit to the Gompertz function (49), where y is total filament length (50 nm cutoff) or light scattering intensity measured at time t, t i is the inflection point corresponding to the time of maximum growth rate, a is the maximum total filament length or light scattering at equilibrium, and b ϭ 1/k app , where k app is the proportional growth rate of the filament population in units of time Ϫ1 (49). Lag times, defined as the time when the tangent to the point of maximum polymerization rate intersects the abscissa of the sigmoidal curve (50), were calculated as t i Ϫ b (49).
Filament length distributions calculated as described previously (41,51) were fit to the exponential function, where y is the percentage of all filaments filling a bin of length interval x, and b is a constant (slope parameter) with units of length Ϫ1 . All parameters estimated by linear or non-linear regression are reported Ϯ S.E.

Impact of Phosphorylation Mimicry on Tau Fibrillization-
The distribution of phosphorylation sites on authentic ADderived tau protein reported from mass spectroscopic methods (10,11) is summarized in Fig. 1. Because of the large number of individual sites, and the even larger number of potential combinations, only a subset of these were modeled in a fulllength, htau40 isoform background by pseudophosphorylation mutagenesis and kinetic analysis (Table I). These 12 sites, which spanned the proline-rich, microtubule-binding, and Cterminal tau domains where phosphorylation sites cluster (Fig.  1), were selected because alone or in combination they compose phosphoepitopes that have been studied extensively by immu-  Table I. nohistochemistry and shown to be occupied in nearly all tauopathies investigated to date (1). Moreover, the temporal correlation between their occupancy and neuritic pathogenesis has been quantified (14,15). Therefore 11 mutant proteins containing selected single, double, and triple pseudophosphorylation mutations in 12 positions were prepared along with wild-type htau40 and characterized for fibrillization activity.
To determine whether low-stoichiometry site-specific pseudophosphorylation could replicate the effects of nonspecific high-stoichiometry phosphorylation, the ability of each of the mutant constructs to fibrillize under standard, near-physiological conditions of pH, ionic strength, and bulk tau concentration (4 M) was determined relative to wild-type recombinant htau40 in the presence of AA inducer. All preparations fibrillized with untwisted morphology as described previously (41), with no morphological differences apparent among samples except mutant 235, which formed a mixture of typical untwisted filaments and small (Ͻ50 nm in length) amorphous aggregates (data not shown). Despite similarities in morphology, significant differences in extents of fibrillization were apparent when reaction products were assayed by quantitative EM. Greatest differences were observed with mutants 212, 214, 212/214, and 199/202/205, which accumulated 2-3-fold as much filamentous tau as did wild-type htau40 and mutants 231, 262, 396/404, 409, and 422 ( Fig. 2A).
To confirm these findings, the amount of filamentous pseudophosphorylated tau at equilibrium relative to wild-type htau40 was estimated using a second inducer (C 18 H 37 SO 4 Na). Results showed that pseudophosphorylation mutants differentially fibrillized in response to C 18 H 37 SO 4 Na inducer as well. Indeed, when the equilibrium levels of fibrillization attained with AA and C 18 H 37 SO 4 Na inducers were compared, a linear relationship was found with nearly identical rank orders of mutant fibrillization levels (Fig. 2B). When equilibrium levels of fibrillization attained with C 18 H 37 SO 4 Na were compared by EM and corrected LLS (a solution-based assay method), again a consistent rank order of fibrillization levels was observed (Fig. 2C). The only exception was mutant 235, which formed both filamentous and nonfilamentous aggregates. Because the EM assay only measures filaments, whereas LLS detects all aggregates in solution, the LLS assay overestimated filament mass for this mutant. Together these results suggested that low-stoichiometry pseudophosphorylation could recapitulate the effects of nonspecific phosphorylation on tau fibrillization levels at equilibrium, and that they were additive with the action of anionic inducers.
The equilibrium phase of tau fibrillization is attained when the rate of protein addition to filament ends equals the rate of protomer dissociation. It is characterized by a "critical" concentration corresponding to the maximal solubility of protein above which all additional assembly competent protein enters the polymeric phase (52). It is the highest concentration of protein that does not support fibrillization, and therefore can be estimated from the dependence of fibrillization on bulk protein concentration (53). To determine whether pseudophosphorylation modulated this equilibrium, the critical concentration of each mutant was estimated using corrected LLS and compared with values for wild-type htau40. The value for wildtype htau40 in the presence of the C 18 H 37 SO 4 Na inducer was 2.00 Ϯ 0.07 M (shown in Fig. 3 and summarized in Table II). Critical concentrations for pseudophosphorylation mutants ranged from this value to as low as 680 Ϯ 130 nM for mutant 212 ( Fig. 3; Table II). The relationship between critical concentration and equilibrium levels of fibrillization (both assayed with C 18 H 37 SO 4 Na inducer) was linear, with lowest critical concentrations yielding the greatest amounts of fibrillization at equilibrium (Fig. 4). These data suggest that the increased fibrillization at equilibrium produced by pseudophosphorylation resulted from increased affinity of filament ends for assembly competent tau.
Anionic surfactants trigger protein fibrillization in micellar form by shifting prenuclear equilibria from assembly incompetent to assembly competent conformations (43), which then serve as substrates for nucleation reactions on the micelle surface (41,51). The enhanced nucleation rate shifts sigmoidal reaction progress curves toward shorter lag times with increasing inducer concentrations (43,54). Because the increase in filament mass induced by high-stoichiometry phosphorylation has been attributed to creation of "nucleation centers" (55), a similar shift in reaction progress curves should accompany increased fibrillization. To test this hypothesis using pseudophosphorylation mutants, the total length of tau filaments formed from each mutant as a function of time in response to anionic surfactant inducer C 18 H 37 SO 4 Na was quantified using the EM assay. All resultant reaction progress curves were sigmoidal with clear lag, exponential growth, and equilibrium phases (Fig. 5). Kinetic parameters of lag time (a measure of nucleation rate; Ref. 50) and k app (the proportional growth rate) were derived from fitting these data to a 3-parameter Gompertz growth function as described under "Experimental Procedures." The Gompertz function was used because it fit time course data better than the 3-or 4-parameter logistic functions we have used in the past (51,54). Under these conditions, wild-type htau40 fibrillized with a lag time of 42.8 Ϯ 0.8 min and a k app of 1.33 Ϯ 0.07 h Ϫ1 , whereas pseudophosphorylation mutants fibrillized with lag times ranging from 38.9 to 61.6 min and k app values ranging from 1.00 to 1.40 h Ϫ1 (Table II). The nucleation rate as reflected by lag time was directly proportional to fibrillization at equilibrium, with longest lag times correlating with the greatest fibrillization at equilibrium (Fig.  6A). In contrast, no correlation between k app (which incorporates nucleation, elongation, and equilibrium terms) with fibrillization at equilibrium was apparent (Fig. 6B). These data suggest that pseudophosphorylation had statistically significant effects on tau fibrillization kinetics, but did not increase filament nucleation rates.
Despite having a critical concentration (ϳ2 M) low enough to support fibrillization under assay conditions (4 M), the wild-type htau40 preparation used here does not assemble spontaneously (i.e. in the absence of exogenous inducer) over time courses as long as 7 days at 37°C in the absence of nucleation inducers (38,39). To determine whether the assembly enhancing effect of pseudophosphorylation could promote spontaneous fibrillization, mutant 212 also was incubated (4 M) for 7 days at 37°C under assembly conditions. As with wild-type htau40, no spontaneous fibrillization was apparent over this period (data not shown). These data suggest that simply having a lower critical concentration for assembly competent conformations is insufficient to promote spontaneous fibrillization at physiological tau concentrations, and that low stoichiometry pseudophosphorylation alone cannot overcome the energy barrier for nucleation over the time periods studied. The nucleation and elongation phases of protein aggregation reactions compete for assembly competent species, resulting in their time-dependent depletion until the critical concentration is reached (56). In the case of tau aggregation, the exponential distribution of tau filament lengths observed at all time points indicates that the rate of elongation exceeds that of nucleation, with both occurring throughout the reaction time course (41,57). Under these conditions, lower critical concentrations could lead to decreased nucleation rates at the latter stages of the time course by consuming assembly competent tau. The resultant increase in filament elongation relative to nucleation should be detectable as shifts in filament length distributions toward longer lengths. For mutants with low critical concentrations, such as 212, the change in distribution was apparent on visual inspection of electron micrographs (Fig. 7). To test the normalized fibrillization lengths/field achieved in the presence of C 18 H 37 SO 4 Na (⌫ f C 18 H 37 SO 4 Na) relative to AA (⌫ f AA) inducer, whereas the line represents linear regression analysis of the data points (symbol key is located in panel C). The data show that the effect of mutations on tau fibrillization could be detected in the presence of both fatty acid and alkyl sulfate inducers. C, fibrillization assays were performed using 50 M C 18 H 37 SO 4 Na as aggregation inducer as described above except that fibrillization was quantified after 26 h by the static light scattering assay. Each point represents the net intensity of scattered laser light corresponding to fibrillization, I n (90°), normalized to wild-type tau relative to total filament lengths/field determined in panel B, whereas the solid line represents linear regression of the data points (excluding mutant 235, which produced amorphous aggregates as well as filaments). These data show that the effects of pseudophosphorylation on fibrillization equilibria were consistent by two distinct assay methods.  Table II), was modulated by sitespecific pseudophosphorylation. Although measured in quadruplicate, error bars on data points have been omitted for clarity. prediction, equilibrium length distributions were estimated after fibrillization of all tau constructs in the presence of AA, with the polydispersity of each population quantified by the slope parameter (b). Results (Fig. 8) confirmed that the increases in filament mass at equilibrium induced by pseudophosphorylation correlated with decreases in the slope parameter (i.e. shifts to longer length distributions). Together these data suggest that the principal effect of pseudophosphorylation on tau fibrillization is to reduce critical concentration, which leads to a shift in the length distribution to longer lengths, and a more stable filament population at low bulk tau concentrations.
Tau Glycation Promotes Tau Filament Formation-The effects of glycation on tau assembly were examined by incubating tau with D-glucose for up to 7 days (31), after which time samples were analyzed by EM-or LLS-based assays. Double mutant C291A/C322A was used for these experiments because it lacked Cys residues that could potentially react under glycating conditions and complicate tau aggregation kinetics (58). On the basis of EM assays using 75 M AA as inducer, the presence of up to 2 mM glucose in the assay had no significant effect on fibrillization levels at equilibrium (Fig. 9). After 7 days incubation at 37°C with 10 mM glucose, however, statistically significant increases in fibrillization at equilibrium appeared (Fig. 9). On the basis of parallel reactions containing 14 C-labeled glucose, these conditions led to the incorporation of ϳ1 mol/mol radiolabel. Desalting of the preparation prior to assay did not significantly diminish the stimulatory effect, confirming that the modification was stable and that carryover of unincorporated glucose was not a contributory factor (Fig. 9).
To determine the mechanism underlying these effects, the full reaction progress curves for both glycated-and non-glycated C291A/C322A were estimated using the corrected LLS assay and C 18 H 37 SO 4 Na as a second fibrillization inducer. All glycated samples were desalted prior to assay in these studies. C291A/C322A incubated at 37°C for 7 days in the absence of glucose assembled with a lag time of 27.4 Ϯ 4.3 min and k app of 2.15 Ϯ 0.17 h Ϫ1 (Fig. 10A). Although C291A/C322A incubated 7 days in the presence of glucose assembled with similar   SO 4 Na was assayed as a function of incubation time at room temperature. Each data point represents the average filament lengths/field calculated from EM images, whereas each curve represents the best fit of the data points to a three-parameter Gompertz growth function (Equation 1). Although pseudophosphorylation had only minor effects on fibrillization kinetics as estimated by lag time or first-order growth rates (summarized in Table II), it greatly modulated total filament mass at equilibrium relative to wild-type htau40. kinetics as the non-glycated sample (lag time ϭ 30.9 Ϯ 2.6 min; k app ϭ 1.86 Ϯ 0.09 h Ϫ1 ), the amount of fibrillar tau at equilibrium was 1.8 Ϯ 0.1-fold greater (Fig. 10A). Filament accumulation was accompanied by a shift in length distribution to longer lengths as indicated by a decrease in the slope parameter to 76 Ϯ 14% of control values (data not shown). To clarify the source of increased filament mass at equilibrium, critical concentrations of both glycated and non-glycated samples were determined by LLS and found to be 1.16 Ϯ 0.09 and 1.86 Ϯ 0.14 M, respectively (Fig. 10B). These characteristics resemble the effects of pseudophosphorylation on tau assembly, and indicate that low-stoichiometry glycation enhances fibrillization by lowering critical concentration and stabilizing filaments.

DISCUSSION
Phosphorylation and the Tau Aggregation Pathway-Sitespecific phosphorylation is an early and ongoing event in neurofibrillary lesion formation, suggesting a potential role in the process (14,15,59). Although phosphorylation could potentially modulate multiple aspects of tau biology (e.g. turnover, localization, etc.), it has been hypothesized that at least one major function in disease is to directly promote the fibrillization reaction (1). One established mechanism through which it may do so is the neutralization of the microtubule-binding activity of tau, thereby fostering an accumulation of freely soluble tau in the cytoplasm (16,19). This reaction, which may represent the first step in the fibrillization pathway, has been replicated both in situ and in vitro using various phosphotransferases and also with pseudophosphorylation mutations (16 -19, 23-26). But whether it represents a commitment toward fibrillization is not clear. In vivo phosphorylation of fetal tau at similar sites (12,13) and nearly the same stoichiometry (60) as PHF-tau does not trigger fibrillization. In situ, tau hyperphosphorylation leads to formation of thioflavin S-reactive tau species with primarily reticular macroscopic morphology (61), consistent with the formation of thioflavin S-reactive intermediates but only modest amounts of filaments (51).
In contrast to these findings, experimentation in vitro suggests that nonspecific, high stoichiometry phosphorylation can directly promote tau aggregation, leading to increased amounts of amorphous and fibrillar aggregates at equilibrium (20,21,55). This phenomenon has been proposed to stem from incorporation of negative charge, which may neutralize the net positive charge of native tau protein and lower the energy barrier for self-association (62,63). Thus the incorporation of negative charge accompanying phosphorylation has been proposed to create nucleation centers that directly trigger tau fibrillization (55).
Although the pseudophosphorylation mutants studied here recapitulated the effects of hyperphosphorylation with respect to increased filament mass at equilibrium, neither spontaneous nucleation nor increased nucleation rate in the presence of FIG. 6. Filament mass at equilibrium correlates with fibrillization lag times. A, lag times; and B, k app values for wild-type htau40 and pseudophosphorylation mutants estimated in Fig. 5 were plotted against total filament lengths at equilibrium estimated in Fig. 2C (% wild-type tau control Ϯ S.D.) and analyzed by linear regression. Horizontal error bars were omitted for clarity. Total filament lengths at equilibrium were directly proportional to lag times but not k app , consistent with a decrease in critical concentration. anionic inducers was observed. Rather, the mechanism involved a shift in assembly equilibrium toward the filamentous state as indicated by site-dependent decreases in critical concentration. Lower critical concentrations reflect decreases in the free energy of the filamentous state, and therefore increases in the driving force for filament nucleation and subsequent elongation (64). But as shown here and previously by others (27, 38, 39, 65, 66), simply raising tau high above its critical concentration for assembly (i.e. creating a supersaturated solution) is insufficient to trigger fibrillization over incubation periods spanning 1 week. Moreover, supersaturated solutions of soluble recombinant tau protein do not efficiently polymerize onto exogenous filamentous seeds in the absence of anionic inducers (38,62), and only small amounts of filaments form in situ even when tau is expressed at supraphysiological levels (67). These data suggest that while recombinant tau is natively unfolded (68), the equilibrium conformation is not necessarily assembly competent. Thus, despite being energetically favorable, filament nucleation proceeds slowly if at all until assembly competent conformations are attained. Under these conditions, protein turnover rates may well exceed tau nucleation rates, and the increased cytosolic levels of supersaturated tau protein engendered by phosphorylation may not lead to fibrillization.
These considerations suggest that assembly competent conformations must be populated before fibrillization can proceed efficiently. In vitro, the barriers to conformational change and nucleation can be overcome by anionic condensation agents such as anionic surfactants or polymers (40, 41, 51, 69, 70). Increasing concentrations of anionic inducers lead to decreasing lag times (i.e. increasing nucleation rates) and increasing FIG. 8. Pseudophosphorylation modulates equilibrium length distributions. Wild-type and mutant htau40 preparations (4 M) were incubated (4 h at room temperature) in assembly buffer containing 75 M AA and then examined by transmission EM. The total lengths/field of all filaments Ն50 nm in length adsorbed onto grids (⌫ f ) were measured from the resultant digitized images. Distributions of lengths that segregated into consecutive intervals (50-nm bins) were also calculated and quantified by a slope parameter as described under "Experimental Procedures." All data points are presented as percentages of normalized values, with wild-type control htau40 defined as 100% on both axes. The solid line represents the best fit of the data points to an exponential function. Increases in filament mass at equilibrium produced by pseudophosphorylation were accompanied by decreases in slope.
FIG. 9. Glycation increases inducer-mediated tau fibrillization. The fibrillization of htau40 C291A/C322A (4 M) under standard conditions (assembly buffer with 75 M AA) was quantified by transmission EM. Total lengths/field of all filaments Ն50 nm in length (⌫ f ) were calculated as percentages of normalized control values Ϯ S.E., with untreated htau40 C291A/C322A (glucose Ϫ) defined as 100%. The presence of free D-glucose (2 mM) in the assay (glucose ϩ; glycation time ϭ 0 d) did not significantly change the extent of fibrillization relative to the no glucose control, confirming that the presence of free glucose in the assay at low millimolar concentrations did not modulate tau assembly under these conditions. In contrast, preincubation of htau40 C291A/C322A under glycating conditions (glycation time ϭ 7 days at 37°C) resulted in a nearly 2-fold increase in the amount of filamentous material at equilibrium. The enhanced aggregation activity was stable to desalting by gel filtration prior to dilution and assay confirming that it resulted from a stable modification of tau protein and not from free glucose or other small molecules carried over into the assay. Significance at p Ͻ 0.01 (**) was estimated by t test. quantities of filamentous tau at equilibrium without changes in critical concentration (43). Similar data has been obtained with micelle-mediated fibrillization of ␣-synuclein (54). The ability of micellar inducers to increase levels of filamentous tau at equilibrium appears to derive in part from changes in prenuclear equilibria, including the conversion of assembly incompetent to assembly competent conformations (43).
On the basis of these considerations, we propose two general classes of tau assembly modulators. Those agents or posttranslational modifications that modulate pre-nuclear equilibria by stabilizing assembly competent conformations can act to trigger fibrillization. Examples include anionic membranes (41), which because of their association of tau filaments in biopsy specimens of AD tissue (71), may be pathophysiologically relevant sources of nucleating activity. The second category includes modifications that affect postnuclear equilibria such as critical concentration. These enhance fibrillization by stabilizing filaments and increasing the driving force for nucleation without necessarily triggering it or increasing its rate. The negative charges introduced by pseudophosphorylation exemplify this second category. Because of their different underlying mechanisms, anionic inducers and pseudophosphorylation act synergistically in vitro. Similarly, multiple fibrillization triggers and enhancers may combine to promote tau fibrillization in disease.
Molecular Mechanism-The extent of fibrillization of tau and pseudophosphorylation mutants at equilibrium was found linearly related to the critical concentration, which is defined as (53), where a crit is the critical concentration, and k Ϫ and k ϩ are the rate constants for protein monomer dissociation from and association with filament ends, respectively. Lowering of critical concentration implies a decrease in k Ϫ , an increase in k ϩ , or a combination of both. Although pseudophosphorylation clearly modulates filament elongation relative to nucleation rate, we could not directly test whether elongation rates (i.e. k ϩ ) change in response to pseudophosphorylation because of the inability to seed fibrillization reactions in the absence of condensing agents (38,62). The relationship between critical concentration and filament nucleation in nucleation-elongation reactions has been derived as (72), where k init represents the operational nucleation rate from n monomers (i.e. n ϭ nucleus size) and C is a constant. Thus, decreases in critical concentration can associate with decreases in k Ϫ , increases in k init , or decreases in nucleation molecularity (n). It has been shown in other systems that these parameters can change in combination in response to decreases in critical concentration (73). In the case of tau pseudophosphorylation, however, the absence of increased nucleation rates suggests that increases in k init can be ruled out.
The relationship between critical concentration and length distribution has been derived as (72), where a is the concentration of free monomer and x j is the concentration of polymers composed of j protomers. Thus, decreases in critical concentration are associated with lengthening of filament length distributions at equilibrium. This effect was observed for all pseudophosphorylation mutants that decreased critical concentration. These data suggest that the kinetic and equilibrium effects of pseudophosphorylation can be rationalized from changes in critical concentration, and that in addition to destabilizing tau-tubulin interactions, a second effect of pseudophosphorylation and potentially phosphorylation is to stabilize filaments once nucleated. This mechanism could be significant at early stages of disease by supporting filament accumulation at submicromolar intracellular concentrations.
Comparison with Previous Work-The effects of pseudophosphorylation on tau fibrillization has been contentious with many conflicting studies. Much of this stems from differences in the methodology used to assay fibrillization. Previously we showed a strong increase in fibrillization tendency with the pseudophosphorylation mutant 396/404 (28). However, these results were based on an LLS assay that was not corrected for inducer micellization, which makes a major contribution to scattering intensity (43). Eliminating the micelle background as described here brings the corrected LLS assay into harmony with the EM assay, and reveals that the increases in fibrillization at equilibrium accompanying pseudophosphorylation at 396/404 are statistically significant but modest. Others have assayed pseudophosphorylation mutants using fluorescent reporter probes such as thioflavin S (27). But this reagent fluoresces in the presence of tau intermediates as well as filaments, and therefore is not specific for fibrillization when employed under the conditions used here (41). Similarly, centrifugationbased assays pellet both fibrillar and non-fibrillar tau aggregates, potentially masking fibrillization-specific effects. This limitation is exemplified by mutant 235, which formed both filamentous and amorphous aggregates, and therefore yielded higher signals in LLS assays compared with EM assays. Finally, results are highly dependent on tau concentration. The effects described here, which are readily apparent at near physiological levels of tau protein, would be difficult to detect at the higher tau concentrations (40 to 100 M) used in other assembly paradigms (22,27).
An example of these considerations in the context of the surfactant-induced tau aggregation system employed here is illustrated in Fig. 4. The relationship between critical concentration and total fibrillization at equilibrium was found to be linear with an extrapolated ordinate intercept of 2.8 Ϯ 0.1 M htau40. These data indicate that tau preparations with critical concentrations above this level cannot fibrillize under our standard bulk tau concentrations (4 M), and may explain why certain tau isoforms fail to assemble in the presence of AA (39) but readily do so at high concentration (40 -100 M) in response to heparin (66). Similarly, the abscissa intercept of this plot shows that the maximum increase in fibrillization theoretically detectable at "zero" critical concentration and 4 M bulk htau40 concentration is ϳ3.6-fold. Relative increases will appear smaller at higher bulk tau concentrations, illustrating the importance of working at physiological tau concentration for these studies.
Nonspecific, high-stoichiometry phosphorylation of recombinant tau in the presence of crude extracts has been found to enhance aggregate formation with both amorphous and filamentous morphologies (20,55). The requirement for high stoichiometry (i.e. Ն10 mol/mol, or ϳ1.5 S.D. above the mean stoichiometry observed in autopsy-derived filamentous tau) may result from nonspecific incorporation of anionic charge into multiple sites, each of which may differ in the ability to enhance fibrillization (as found here). Although these data have been interpreted as evidence for the phosphorylation-dependent creation of nucleation centers, nonphosphorylated re-combinant full-length htau40 also aggregated spontaneously in these studies (20). As discussed above, the latter result is highly unusual, and suggests the presence of exogenous aggregation inducers that are absent from most in vitro fibrillization experiments. Quantitative experiments with purified components will be required to establish whether high stoichiometry phosphorylation can also directly accelerate nucleation rate and promote spontaneous assembly.
Sites of Pseudophosphorylation-On the basis of immunohistochemical studies, tau phosphorylation sites fill hierarchically during the course of AD (14,15). Early modifications predominating in the pre-tangle stage include sites 231, 235, 262, and 409. Pseudophosphorylation mutants corresponding to single residue changes at these positions did not differ significantly from wild-type htau40 with respect to assembly kinetics or final equilibrium position. Sites more fully occupied in classic intracellular NFTs, such as those detected by monoclonal antibodies AT100 (212/214) and AT8 (199/202/205), however, were associated with the largest changes in fibrillization behavior. Although both 212 and 214 single mutants greatly modified fibrillization characteristics, the effects of these closely spaced mutations were not additive in the 212/214 double mutant. Mutants corresponding to sites enriched in late-stage disease, including those identified by PHF1 (396/ 404) and site 422, were not substantially different from wild type. Correlations with lesion stage must be interpreted with caution, however, because staging data depends on the differential reactivity of monoclonal antibodies and is potentially biased toward sites that are stable to the proteolysis (74) and changes in tau conformation (75) that accompany lesion maturation. Rather than maturation stage, the rank order of pseudophosphorylation efficacy may more closely reflect their location in the polypeptide chain, with activating mutations clustering in residues 199 -214 of the proline-rich domain. This region flanks the microtubule-binding domain and is required for tau-induced tubulin nucleation in vitro (76). It also promotes interaction of tau with the microtubules (37). The ability of tau to induce microtubule nucleation is lost upon phosphorylation of this region (36). Together, these observations support the hypothesis that phosphorylation within the prolinerich domain critically affects the physiological function and pathological aggregation of tau.
Role of Glycation-As shown previously (5,30,31), and confirmed here, in vitro glycation can increase the equilibrium levels of tau fibrillization at constant bulk tau concentration. The reaction must proceed for several days to yield these differences, however. Like pseudophosphorylation mutants, glycation depressed critical concentration, indicating that filament stabilization is one mechanism underlying the phenomenon. Despite its ability to enhance tau fibrillization, glycated tau did not fibrillize spontaneously or increase the nucleation rate, again suggesting that the glycation conditions used here (i.e. yielding low-stoichiometry modification) did not promote the accumulation of assembly competent conformations.
Together these data are consistent with at least two key roles for tau modifications in promoting tau fibrillization in early stage disease. The first is to modulate tau/microtubule equilibrium to regulate intracellular concentrations of free tau. The second is to lower the critical concentration for assembly thereby stabilizing filaments and increasing the driving force for nucleation. The approach outlined here will be useful in characterizing the ability of diverse tau modifications to modulate tau aggregation kinetics.