Tau repeat regions contain conserved histidine residues that modulate microtubule-binding in response to changes in pH

Tau, a member of the MAP2/tau family of microtubule-associated proteins, stabilizes and organizes axonal microtubules in healthy neurons. In neurodegenerative tauopathies, tau dissociates from microtubules and forms neurotoxic extracellular aggregates. MAP2/tau family proteins are characterized by three to five conserved, intrinsically disordered repeat regions that mediate electrostatic interactions with the microtubule surface. Here, we used molecular dynamics, microtubule-binding experiments, and live-cell microscopy, revealing that highly-conserved histidine residues near the C terminus of each microtubule-binding repeat are pH sensors that can modulate tau–microtubule interaction strength within the physiological intracellular pH range. We observed that at low pH (<7.5), these histidines are positively charged and interact with phenylalanine residues in a hydrophobic cleft between adjacent tubulin dimers. At higher pH (>7.5), tau deprotonation decreased binding to microtubules both in vitro and in cells. Electrostatic and hydrophobic characteristics of histidine were both required for tau–microtubule binding, as substitutions with constitutively and positively charged nonaromatic lysine or uncharged alanine greatly reduced or abolished tau–microtubule binding. Consistent with these findings, tau–microtubule binding was reduced in a cancer cell model with increased intracellular pH but was rapidly restored by decreasing the pH to normal levels. These results add detailed insights into the intracellular regulation of tau activity that may be relevant in both normal and pathological conditions.

share conserved MT-binding motifs. These MAP2/tau family proteins associate with MTs in an extended conformation along the MT protofilament ridge (1,2), and binding is thought to be mediated largely through electrostatic interactions of basic amino acids distributed throughout the MAP2/tau family MT-binding repeats with the negatively charged MT wall. Tau (MAPT), the most intensely studied member of the MAP2/tau family, is expressed in both the developing and adult nervous system, and different tau splice variants contain either three or four MT-binding repeats (3). In adult neurons, tau-MT binding is restricted to the axon. Tau is thought to influence MT stability and mechanics and to regulate axonal transport potentially by organizing inter-MT spacing in the axonal MT bundle (4 -6), but its physiological function and regulation remain poorly understood.
Tau is also a key player in progressive neurodegenerative diseases, such as Alzheimer's disease and frontotemporal dementia. These and other tauopathies are characterized by a loss of axonal tau-MT association and an abnormal extracellular accumulation of aggregated tau (7,8), although it remains controversial if these tau neurofibrillary tangles are a cause or consequence of neurodegeneration. Both the spatial control of tau-MT binding in normal neurons and the pathological dissociation from MTs highlight the importance of understanding how tau-MT binding is controlled. Tau has many phosphorylation sites, and as is the case with other MT-binding proteins, tau phosphorylation decreases its affinity for MTs, and pathological tau aggregates are formed by all splice variants of hyperphosphorylated tau (8,9).
However, the intracellular environment is complex and dynamic, and many incompletely understood factors contribute to the control of protein interactions and functions (10). One such underappreciated regulator is intracellular pH (pH i ) (11). In contrast to previous static views of pH i homeostasis, we now know that pH i is dynamic and changes, for example, during cell cycle progression (12,13), cell adhesion (14,15), and migration (16 -18). In addition, dysregulated pH i is a feature of many diseases, including cancer (19,20) and neurodegenerative disorders (21,22). Although protein pH sensitivity is generally mediated by histidines with a pK a near neutral, other titratable residues can also play critical roles depending on the protein energy landscape and cooperativity (11). Consequently, several examples have emerged in which histidine protonation dynamics regulate protein-protein (23,24), protein-nucleotide (25), and protein-phospholipid interactions (26,27). Because all MAP2/tau family MT-binding repeats contain invariant histidine residues, we used in silico, in vitro, and in vivo methods to ask whether tau-MT binding is sensitive to changes in pH i . Consistent with the modulation of electrostatic interactions, we find that tau-MT binding is reduced at increased pH values, which open new directions for understanding tau-MT binding dynamics in normal and pathological cell behaviors.

Conserved histidine residues contribute to tau-MT interactions
Recent cryo-electron microscopy (cryo-EM) data of MTbound tau at near atomic resolution (28) indicate that tau-MT interactions extend beyond the central KXGS MT-binding repeat motif. In particular, the structure of the second tau MTbinding repeat (R2) shows that a highly conserved histidine residue, His-299, near the R2 C terminus may contribute to tau-MT binding. His-299 points into a hydrophobic cleft defined by ␤-tubulin Phe-399 and Phe-395 near the interdimer interface, and His-299 is in a T-shaped configuration with Phe-399 ( Fig. 1, B and C).
Because of the negative electrostatic potential of the MT surface (Fig. 1A), the pK a of His-299 is estimated to be upshifted to about 7.0 when compared with the imidazole group in solution that has a pK a of 6.3 (29). A positively charged His-299 ϩ imidazolium cation is further stabilized by favorable electrostatic interactions with negatively charged ␤-tubulin Glu-442 and is within reach of other hydrophobic contacts (Ͻ7 Å) with ␤-tubulin residues near the inter-tubulin dimer cleft (Fig. 1C). Based on this analysis, we predicted that His-299 is positively charged MT-binding repeat from the MT outside based on the recent cryo-EM structure (PDB code 6CVN). The tubulin surface is colored by electrostatic potential highlighting the strongly negative surface charge or the protofilament ridge. B, last frames of 5-ns MD simulations of protonated His-299 ϩ (i.e. low pH; orange) and unprotonated His-299 0 (high pH; blue) compared with the cryo-EM structure (gray). A side view of the tubulin protofilament is shown with tau R2 on top. C, close-up views of the binding pocket surrounding His-299. D, sequence alignment of MT-binding repeats in all members of the human MAP2/tau family indicating the high degree of conservation of the histidine residue in the position equivalent to His-299 in tau R2 (asterisk).

pH-modulated tau-microtubule binding
at neutral pH. His-299 protonation could thus be regulated through pH changes that fall within the cytoplasmic pH range. In contrast, solvent-exposed lysine and arginine residues in the MT-binding repeats have pK a values of Ͼ10 and should always be protonated near neutral pH values. We therefore asked how positively charged His-299 ϩ or neutral His-299 0 could affect the interaction between tau R2 and MTs by in silico molecular dynamics (MD).
Based on the cryo-EM structure, we built models of the tau R2-MT complex with different His-299 protonation states. Both systems were then evaluated in short time MD simulations (5 ns). In the model containing positively charged His-299 ϩ , the T-shaped interaction with Phe-399 was maintained throughout the entire simulation ( Fig. 1C; Video S1), with distances not exceeding 5 Å and in agreement with the cryo-EM structure. In contrast, neutral His-299 0 turned away from the hydrophobic cleft and remained in this position during the MD simulation. This also caused higher deviation from the tau-MT cryo-EM structure (RMSD ϭ 2.9 Å) when compared with His-299 ϩ (RMSD ϭ 1.9 Å) as well as globally increased conformational fluctuations in both tubulin and tau (Fig. S1). Although our simulations analyzed only one of the tau MT-binding repeats for which a cryo-EM structure was available, the MTbinding repeat histidine residues in positions equivalent to His-299 are highly conserved (Fig. 1D) (1,2). In addition, 3R and 4R tau isoforms bind MTs in a similar manner utilizing the fulllength of the respective MT-binding repeat region (30). Because individually weak MT-binding repeats act together to support high-affinity tau-MT interactions (31), histidine deprotonation at increased pH in multiple repeats may thus cooperatively weaken the tau-MT interaction and consequently destabilize the tau-MT complex.

Increased pH decreases tau-MT binding
To test the hypothesis that increased pH weakens tau-MT binding, we first used an in vitro MT co-sedimentation assay. The standard PIPES buffer used in such assays has a pK a of 6.8, which is at the low end of the physiological pH range. Therefore, we instead incubated 100 nM purified 0N3R tau protein, a developmental tau isoform that contains three MT-binding repeats, with 0.5 M paclitaxel-stabilized MTs in a MOPSbased buffer (pK a ϭ 7.2) that provides better buffering capacity in the relevant range. MT-bound tau was then separated from soluble tau protein by centrifugation through a glycerol cushion. Nearly all tau bound to MTs at pH 7.1 (99.5 Ϯ 0.2%; mean Ϯ S.D.; Fig. 2A). In contrast, the amount of tau in the MT pellet was significantly reduced at pH 7.8 to 91.7 Ϯ 3.4%, and a detectable fraction of unbound tau remained in the supernatant.
Because this was a relatively small difference likely due to near-saturation of the binding reaction at these concentrations, we next determined the apparent dissociation constant (K d ) of tau-MT binding at pH 7.1 and 7.8. To quantify the amount of MT-bound tau protein more accurately than we were able to achieve by immunoblotting and densitometry, we instead used 0N3R sfGFP-tau that has a fluorescent tag. The fraction of MT-bound sfGFP-tau was determined by measuring the fluorescence of sfGFP-tau that remained in the supernatant and that co-sedimented with MTs, varying the MT concentration over approximately 2 orders of magnitude around the anticipated K d values (32). Using hyperbolic curve fitting to a binding isotherm, we calculated a K d of 0.11 Ϯ 0.014 M (mean Ϯ S.D.; Fig. 2B) at pH 7.1. This reflects a slightly higher 0N3R affinity for MTs compared with reported values (32), which may be due to our different and lower salt buffer conditions. Importantly however, at pH 7.8, the measured K d was significantly increased by ϳ2-fold (0.2 Ϯ 0.013 M) indicating that the affinity of tau for MTs is decreased at increased pH values within the cytoplasmic pH range.

Intracellular pH modulates tau-MT binding in cells
In vitro experiments cannot fully reconstitute the intracellular milieu or the effect of posttranslational modifications (10). In addition, taxanes used to stabilize MTs in vitro alter tau-MT The black line shows a hyperbolic fit through data from all experiments, and the gray area indicates the 95% confidence interval of the fit. The box plot shows dissociation constants obtained when each experiment was fitted independently (n ϭ 4). Box plots show median, first, and third quartile, with whiskers extending to observations within 1.5 times the interquartile range, and all individual data points. Statistical analysis was by Student's t test.

pH-modulated tau-microtubule binding
interactions (33). We therefore next asked whether changes in pH i alter tau-MT binding in human retinal pigment epithelial (RPE) cells transiently expressing mEmerald-tagged 0N3R tau (Em-tau) at a low level that did not induce MT bundling. pH i changes were measured in parallel in untransfected cells (34,35). In control HEPES pH 7.4 buffer, the pH i of RPE cells was ϳ7.4, and Em-tau clearly bound along MTs, although a substantial fraction of Em-tau remained in the cytoplasm. Acute increase of pH i to ϳ7.7 by addition of 20 mM NH 4 Cl resulted in a notable decrease in MT-bound tau ( Fig. 3A; Video S2). Because of low amounts of MT-bound tau at high pH i and movement of the MT network in these time-lapse sequences, direct measurements of MT-bound tau were difficult and unreliable. Therefore, we instead quantified Em-tau fluorescence intensity in the cytoplasm. It is important to note that the reported pK a of mEmerald is ϳ6.0, and mEmerald fluorescence is virtually insensitive to pH changes above 7.0 (36). Indeed, we did not observe an increase in total cell fluorescence when pH i was increased within the physiological range. Thus, the observed increase in cytoplasm Em-tau fluorescence in response to experimentally increased pH i reflects dissociation from MTs. Em-tau dissociation from MTs occurred within seconds and was rapidly reversed when the NH 4 Cl solution was washed out and replaced with control HEPES buffer returning pH i to its normal value.
In contrast, the MT network was not affected by these acute pH i changes as quantified by measuring the cytoplasm fluores-cence intensity of EGFP-tagged tubulin not incorporated into MTs (Fig. 3). This demonstrates that Em-tau dissociation from MTs at elevated pH i was not a consequence of MT depolymerization and that the overall polymerization state of the MT network is insensitive to short-term pH i changes within this range, although we did not directly evaluate effects on MT plus-end growth dynamics.
We observed a similar rapid and reversible disruption of Em-tau binding to MTs when pH i was acutely increased with 100 mM NaCl (Fig. 4), which osmotically activates proton efflux by the plasma membrane Na ϩ -H ϩ exchanger NHE1 (34,35). Tau belongs to the MAP2/tau family of MAPs that share the same repeat motif in the MT-binding domain (1), and the histidine residues at the C termini of MT-binding repeats in all these proteins are conserved (Fig. 1D). To test whether other MAP2/tau family proteins respond to intracellular pH i changes, we transiently expressed the MT-binding region of mouse MAP4 containing three MT-binding repeats in RPE cells. Like Em-tau, Em-MAP4 responded to the hyperosmotic pH i increase with a reversible decrease in MT binding ( Fig. 4) indicating that decreased MT binding at increased pH i may be a common characteristic of MAP2/tau family proteins.
The above experiments relied on acute changes in pH i that occur within seconds to minutes. To test how long-term pH i changes affected tau-MT interactions, we transiently expressed Em-tau in MCF10A mammary epithelial cells stably expressing an estrogen receptor-induced oncogenic H-RasV12 (ER-RasV12).

pH-modulated tau-microtubule binding
We previously showed that tamoxifen-induced RasV12 expression increased pH i to almost 7.7 within 24 h (37). We found that Em-tau was predominantly cytosolic in tamoxifen-induced RasV12-expressing cells. However, acutely decreasing pH i with a pH 7.2 buffer containing the protonophore nigericin, which equilibrates extracellular and intracellular pH, resulted in a rapid rescue of Em-tau binding to MTs (Fig. 5). Together, these data demonstrate that an increase in pH i within the physiological cytoplasmic range weakens tau-MT binding, which in cells results in substantially decreased MT-associated tau.

Histidines confer pH sensitivity and are required for interactions with MTs
The location of His-299 within a hydrophobic pocket indicates that hydrophobic interactions of the histidine aromatic ring are important for MT binding, and it may explain the high  pH-modulated tau-microtubule binding degree of conservation of these histidine residues. To further investigate this, we sought to identify mutations that mimic His-299 0 and His-299 ϩ in a pH-independent manner. We therefore built a model in which His-299 was changed to a constitutively and positively charged lysine residue that we hypothesized could mimic His-299 ϩ . However, MD simulations showed that a lysine substitution behaved more like deprotonated His-299 0 (Fig. S1). Although protonated lysine cationinteractions are strong and commonly observed in proteins (38), due to the high entropic desolvation cost, the ␤-tubulin hydrophobic cleft near Phe-395 and Phe-399 cannot readily accommodate the hydrophilic lysine ammonium cation. In fact, Lys-299 turns away from the hydrophobic cleft, although it can maintain electrostatic interactions with Glu-422 (Fig. 6A).
To further evaluate the importance of hydrophobic contacts by in silico MD, we next analyzed a substitution of His-299 with alanine, a small noncharged amino acid that is not expected to contribute to either hydrophobic or electrostatic interactions. Like lysine, the alanine substitution also showed increased fluctuations of the tau-MT complex near the C terminus of tau R2 pH-modulated tau-microtubule binding ( Fig. S1), and throughout the MD simulation Ala-299 pointed away from the hydrophobic cleft ( Fig. 6A; Video S3).
Although we cannot directly infer binding free energy changes from MD, our in silico models support that the positively charged His-299 ϩ can uniquely form both an electrostatic interaction with Glu-422 and an energetically favorable interaction with Phe-399. In contrast, lysine at position 299 only maintains the electrostatic interaction, whereas an Ala-299 loses both resulting in partial dissociation of the tau peptide from the MT surface.
To directly test whether conserved histidines are required for tau-MT binding, we generated mutant tau 0N3R protein in which all four histidine residues in the MT-binding repeats (R3 has two adjacent histidines; Fig. 1D) were substituted with either lysine (H4K) or alanine (H4A). If MT binding relied only on electrostatic interactions with protonated histidines, tau(H4K) MT binding would be expected to be similar to WT tau, whereas tau(H4A) should show reduced MT binding. Instead, and consistent with our structural analysis, both variants displayed substantially reduced MT binding in vitro (Fig.  6B). Although tau(H4K) seemed to bind to MTs somewhat better than tau(H4A), in vitro this difference was not statistically significant.
If these conserved histidines are responsible for the pH sensitivity of tau-MT binding, because of the high pK a of lysine, residual tau(H4K) binding to MTs should no longer respond to pH changes within the physiological range. Because the K d values we measured for pH-modulated tau-MT binding were in the 100 -200 nM range (Fig. 2B), we tested this at a 20-fold lower tau to MT ratio that should be well-below saturation. At these low tau concentrations (10 nM), it was not possible to accurately determine the amount of tau remaining in the supernatant. Instead, we compared how much tau was recovered in the MT pellet at pH 7.8 and 7.1, which was reduced to 52.8 Ϯ 18.3% (mean Ϯ S.D.; Fig. 6C). As a control, the amount of MTs in the pellet remained unchanged indicating that this difference reflects tau-MT binding and not a decrease in MT stability at higher pH. In contrast to WT tau, MT binding of tau(H4K) was not decreased at increased pH (112.5 Ϯ 8.1%; Fig. 6C), and the pH 7.8 to 7.1 ratio was not significantly different from the amount of MTs recovered in the pellet.
Finally, we tested binding of fluorescently tagged tau variants to MTs in RPE cells expressing equivalent levels of WT or mutated Em-tau ( Fig. 7; Fig. S2). In cells, binding of Emtau(H4K) to MTs was significantly reduced compared with WT tau, and Em-tau(H4A) did not bind MTs at all. Together, these data confirm that the requirement of the tau MT-binding repeat histidine residues is not solely explained by electrostatics and that protonation of these histidines can be titrated to modulate tau-MT interactions.

Discussion
We used multiple approaches, including MD simulations and MT-binding assays in vitro and in cells, to show that interactions of MAP2/tau family proteins with MTs are sensitive to pH changes within the physiological intracellular range. We also demonstrate that histidine residues near the C-terminal end of MAP2/tau family MT-binding repeats are essential. Tau mutants with these histidines substituted with either lysine or alanine show reduced or completely absent MT binding in cells. In addition, at least in vitro these conserved histidines mediate pH sensitivity as replacing them with lysine residues, which are insensitive to pH changes in the physiological range, abolishes pH-modulation of MT binding in vitro. However, we were not able to test this rigorously in cells because of the already substantially reduced MT binding of the His 3 Lys mutant.
Although these MAP2/tau MT-binding repeat histidine residues are highly conserved, their importance has not been recognized because the intrinsically disordered nature of MAP2/ tau family proteins complicates structural analysis. Previous studies did not resolve MT-bound peripheral regions of these MT-binding repeats and predominantly found MAP2/tau family densities along the MT protofilament ridge (2). Although a high-definition model of native MT-bound tau is still not available, recent cryo-EM data of MT-bound constructs of human tau MT-binding repeat R2 show a close interaction of His-299 with the tubulin interdimer interface (28). This overall arrange- pH-modulated tau-microtubule binding ment of tau histidine residues near tubulin-tubulin interfaces is also supported by NMR and more recent EPR spectroscopy data (30,39).
At low pH, our structural and MD analysis defines a stable interaction of the protonated histidine with specific aromatic residues in ␤-tubulin ␣-helix H11. This requires the amphiphilic character of protonated histidine as deprotonation at high pH or substitution with either nonaromatic or noncharged amino acids substantially weakens tau-MT interactions. Thus, our data are consistent with older biochemical cross-linking data showing tau interaction with H11, H12, and the flexible loop connecting these helices (40) and new ultrastructural analysis of MT-bound MAP4 that indicates a strong contribution of the tubulin interdimer interface to the overall MAP4 -MT binding energy (41).
Cation-interactions between aromatic rings and positive charges are common in proteins, and histidine-aromatic complexes are energetically favorable in the hydrophobic protein interior (38,42,43). Despite its unusual arrangement for a cation-interaction with the edge of Phe-399 pointing toward the His-299 ring, we find that protonated His-299 ϩ maintains this proximity and geometry throughout 5-ns MD simulations. Given that the positive charge is not located above the system, we propose that this unusual geometry may be stabilized by additional electrostatics. The Amber force field used in our simulations does not explicitly compute cation-interactions, but it indirectly accounts for them by including both electrostatics and van der Waals energies. Our MD calculations are focused on how tau-MT interactions are modulated, but it is plausible that His-299 protonation influences the unbound tau conformation, which could contribute to tau-MT net binding affinities. However, quantitative estimates of such relative free energies are challenging and were not attempted here.
Because of its role in neurodegenerative disease, tau is the most widely studied member of the MAP2/tau family, and much research has focused on the role of tau aggregation in Alzheimer's disease and related tauopathies with over 14,000 publications listed in PubMed (3). Less is known about how MAP2/tau family proteins control the MT cytoskeleton in normal neurons. In vitro, tau promotes MT polymerization and protects MTs from disassembly likely by spanning multiple tubulin dimers and thus inhibiting their dissociation. Consequently, MAP2/tau family proteins are thought to stabilize MTs in neurons. Recent data of shRNA-mediated tau depletion in primary neurons indicate that tau promotes growth cone MT elongation (44). In addition, tau inhibits EB1 association with growing MT ends (45) suggesting multiple roles of tau in controlling neuronal MT polymerization dynamics. High levels of MAP2/tau family proteins induce MT bundles in non-neuronal cells with inter-MT spacing similar to neurite MT bundles (6). Binding of MAP2/tau family proteins along MTs also inhibits MT motor movements (5). However, these functional data rely mostly on in vitro experiments or MAP2/tau protein overexpression, and to what extent these reflect physiological neuronal tau functions remains unclear. For example, MT-based transport appears unaffected in tau knockout mice (46), and moderate tau expression levels have little effect on MT network organization. In summary, much remains to be learned about how MAP2/tau proteins control physiologically normal neuronal MT cytoskeleton function. What we do know though is that neurons are morphologically complex and thus the activity of neuronal cytoskeleton proteins must be controlled precisely.
Characteristic of many intrinsically disordered proteins are their complex phosphorylation patterns. MAP2/tau proteins are no exception (3,47), and phosphorylation within the MT-binding repeats generally reduces tau MT binding, consistent with electrostatic repulsion between negatively charged phosphates and the strong negative electrostatic field of the MT surface. We propose that the upshifted pK a value of the conserved MT-binding repeat histidine residues allows them to act as pH sensors and modulate tau-MT electrostatic interactions within the physiological pH range. Consistent with a decreased tau-MT interaction strength, individual tau molecules also show increased diffusion along MTs at increased pH values (48). Thus, pH i changes may add another layer of control to MAP2/tau MT binding. Observed differences in the magnitude of pH-modulated tau MT binding between our experiments in vitro and in cells may be due to partial phosphorylation of tau in cells. However, tau binding to MTs is also sensitive to subtle changes in MT lattice geometry (33). Although we find that the total MT polymer is insensitive to short-term pH i changes, we cannot rule out that pH-mediated conformational changes of the MT wall contribute to pH-sensitive tau-MT interactions.
Currently, we can only speculate as to the physiological relevance of pH modulating tau-MT binding. Because neurons are morphologically differentiated with highly compartmentalized cytoplasm, it is possible that developmental or spatial control of pH i in developing neurons contributes to spatiotemporal control of MAP2/tau activities, but potential local differences in neuronal cytoplasmic pH i remain poorly characterized (49,50). However, acidification of the neuronal cytoplasm has been associated with and speculated to be a causative agent in neurodegeneration (21). Our current data would predict enhanced tau-MT binding in acidified neurons, which is inconsistent with tau aggregation due to weakened tau-MT interactions in Alzheimer's disease. Alternatively, increased binding of protonated tau to negatively charged phospholipid membranes may increase toxic effects of aggregated tau (51).
Finally, ectopic tau expression has been described in many types of cancer, and elevated tau levels appear to correlate with aggressive metastatic cancer phenotypes and resistance to taxane chemotherapy (52)(53)(54). Cancer cells generally have elevated pH i (19,20), although a large variability in different types of cancers is likely. Our data indicate that tau-MT binding is greatly reduced in a breast cancer model expressing oncogenic RasV12 that has an increased pH i . Tau-MT binding is rapidly rescued in these RasV12-expressing cells by returning pH i to a physiological normal level. Thus, taxane sensitivity may positively correlate with pH i in tau-expressing cancers, which could serve as a diagnostic tool for taxane chemotherapy. In conclusion, even though we do not yet understand the functional relevance of pH-modulated tau-MT interactions in normal or pH-modulated tau-microtubule binding pathological conditions, this study adds another facet to the already complex regulation of MAP2/tau cell biology.

Molecular dynamics
The recently reported cryo-EM structure of MT-bound tau MT-binding repeat 2 (amino acids 274 -300; PDB code 6CVN) was used as starting point for MD simulations (28). The pK a of MT-bound His-299 was predicted using the PROPKA3 software (29). Two different protonation states for His-299 were modeled, a neutral His-299 0 protonated on the ⑀ nitrogen, as predicted by the Maestro Protein Preparation package, and a double protonated form, His-299 ϩ . His-299 was also mutated to lysine or alanine using the "swappa" command as implemented in Chimera (55). Preceding MD simulations, the tau-MT model structures were submitted to 1,000 steps of steepest descent minimization, followed by 1,000 ps of equilibration. The SHAKE algorithm was used to constrain all bonds containing hydrogen atoms. A cutoff of 12 Å was used for longrange interactions. These energy-minimized models were then submitted to short-time MD simulations (5 ns) using the NPT ensemble with the Amberff12SB force field (56). The Berendsen pressure coupling scheme was used for keeping the pressure constant at 1 atm. The temperature of the production was constant at 310 K using the Langevin thermostat.

Plasmids and protein production
Mammalian expression plasmids mEmerald-MAPTau-C-10 (encoding full-length human tau isoform 4; accession number NM_016841) and mEmerald-MAP4 -C-10 (containing the MAP4 MT-binding domain; Addgene plasmid no. 54152) were from the Michael Davidson plasmid collection. Tau(H4K) and (H4A) mutations were constructed as gene blocks (Integrated DNA Technologies) and inserted into the WT pmEmerald-MAPTau-C-10 plasmid using Gibson assembly. For bacterial expression, full-length 0N3R tau was subcloned into a pET28a bacterial expression vector using Gibson assembly containing an N-terminal His 6 -tag. sfGFP-tau was constructed with an N-terminal His 6 -tag and tandem Strep-tags connected by a GS-linker, followed by sfGFP, a precision protease cleavage site, and the human 0N3R tau sequence.

In vitro MT-binding assays
Paclitaxel-stabilized MTs were prepared by polymerizing 25 M purified porcine or bovine tubulin in 80 mM K-PIPES, pH 6.8, 5 mM MgCl 2 , 1 mM EGTA, 33% glycerol, and 1 mM GTP at 37°C. After 30 min, the MTs were stabilized by addition of 50 M paclitaxel.
To test MT binding of histidine mutants, 100 nM tau protein was incubated with 1 M MTs in 250 l of BRB80 (80 mM K-PIPES, pH 6.8, 1 mM MgCl 2 , 1 mM EGTA) with 1 mM DTT, 10 M paclitaxel, and 10 g/ml BSA. For testing pH-sensitive MT binding in vitro, we instead used a MOPS-based buffer with improved buffering capacity in the physiological pH range (20 mM MOPS, 0.1 mM EDTA, 0.1 mM EGTA, 5 mM magnesium acetate, 50 mM potassium acetate) with 1 mM DTT, 10 M paclitaxel, and 10 g/ml BSA. The MOPS buffer pH was adjusted to 7.2 and 8.0, respectively. pH values reported on the figures were measured after adding buffer equivalent to the amounts of MTs and tau added, which lowered the pH in the binding reactions by 0.1-0.2 pH units. For reactions with 10 nM tau, the volume was doubled to 500 l. After 20 min of incubation at room temperature, MT-binding reactions were overlaid onto 400 -600 l of BRB80 or the MOPS-based buffer with 10 M paclitaxel containing 60% w/v glycerol and centrifuged in a Beckman TLA 100.2 rotor at 50,000 rpm for 20 min at 25°C. After centrifugation a portion of the supernatant was saved for analysis; the cushion was carefully washed with double-distilled H 2 O to minimize contamination of the pellet and removed, and the pellet was resuspended in SDS-PAGE sample buffer. Equivalent amounts of supernatant (S) and pellet (P) were analyzed by SDS-PAGE on 4 -12% NuPAGE gels (Invitrogen) and immunoblotted with a tau antibody (E-4, sc-515539; Santa Cruz Biotechnology). Coomassie-stained gels and chemiluminescent immunoblots were imaged with a FluorChem Q gel documentation system (92-14116-00, Alpha Innotech). The MT-bound tau fraction was calculated as (I P Ϫ I bkg )/(I S ϩ I P Ϫ I bkg ), in which I bkg is the intensity of a background ROI adjacent to the quantified protein bands.
To determine dissociation constants using sfGFP-tau fluorescence, MTs were diluted in a 1:2 dilution series into 10 l of BRB80 with 10 M paclitaxel and mixed with 50 nM sfGFP-tau in 200 l of MOPS buffer with DTT, paclitaxel, and BSA as above in 500-l Eppendorf tubes with the lids cut off. After 20 min of incubation at room temperature, the tubes with the binding reactions were set into TLA 100.2 polycarbonate tubes and centrifuged as above, but without a glycerol cushion. 150 l of supernatant was then directly transferred from the centrifuge tube into a 100-l Sub-MicroQuartz Fluorometer Cell (Starna Cells, Inc.) to minimize sfGFP-tau loss through unspecific adhesion to plastic. After removing the remainder of the supernatant, MT pellets were resuspended in 200 l of ice-cold MOPS buffer without paclitaxel. sfGFP-tau fluorescence (F) was measured with a SpectraMax M5 plate reader (Molecular

pH-modulated tau-microtubule binding
Dynamics, Sunnyvale, CA) at Ex ϭ 480 nm and Em ϭ 520 nm. The fraction of MT-bound sfGFP-tau (f bnd ) for each MT concentration was calculated as the average of (F tot Ϫ F S )/F tot , in which F tot is the fluorescence in the supernatant with no MTs, and F P /(F S ϩ F P ). Curve fitting with a binding isotherm f bnd ϭ b max ⅐[MT]/(K d ϩ [MT]) was done in MATLAB (MathWorks).

Cell culture, live cell imaging, and analysis
RPE cells were maintained in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal bovine serum (Atlanta Biosciences) and 1% Glutamax (Invitrogen) at 37°C, 5% CO 2 . MCF10A cells stably expressing ER-RASV12 were maintained as described (37). For transient protein expression and microscopy, cells were plated in 35-mm glass-bottom dishes (Mat-Tek), transfected after 24 h using FuGENE 6 (Promega) and 1 g of plasmid DNA, and used for experiments 24 h later after replacing the transfection medium. RPE cells stably expressing EGFP-tubulin were provided by Julia Rohrberg and Andrei Goga.
To determine the pH response of tau MT binding in cells, the tissue culture medium was replaced with 25 mM Na-HEPES, pH 7.4, 140 mM NaCl, 5 mM KCl, 10 mM glucose, 1 mM MgSO 4 , 1 mM K 2 HPO 4 /KH 2 PO 4 , pH 7.4, and 2 mM CaCl 2 . Cells were imaged on an environmentally controlled spinning disk confocal system as described (57,58). After 3-5 images in the control HEPES pH 7.4 buffer, the buffer was replaced with the same buffer either containing 20 mM NH 4 Cl or 100 mM NaCl. After 5-10 min, the buffer was once again replaced with control HEPES buffer. Intracellular pH changes with these treatments were determined in parallel in untransfected cells loaded with 2,7-biscarboxyethyl-5(6)-carboxyfluorescein (BCECF) essentially as described (34,35). BCECF fluorescence at Em ϭ 530 nm at two different excitation wavelengths ( Ex ϭ 440 nm and 490 nm) was acquired using the SpectraMax M5 plate reader and calibrated to cells treated with 10 M of the proton ionophore nigericin (Invitrogen) at pH 7.5 and 6.6.
To indirectly analyze the amount of MT-bound tau at different pH i values, we measured the mEmerald-tau fluorescence in the cytoplasm, which increases if less tau is MT-bound. mEmerald fluorescence intensity was measured in three small regions-of-interest (ROIs) per cell at three different time points on the unprocessed 16-bit images per buffer condition, and as necessary, the ROIs were moved slightly between images to avoid MT movements. ROI intensities were corrected for photobleaching or potential pH-dependent changes in mEmerald fluorescence by normalizing to the total cell fluorescence in the first captured image. Normalization to the first time point also preserved variability between measurements. The relative amount of MT-bound tau mutants was quantified as described (33). All image analysis was done in NIS Elements version 4.3 or higher (Nikon). Significance of multiple comparisons was calculated by Tukey-Kramer honest significant difference (HSD) test in Analyze-It for Microsoft Excel. Figures were assembled in Adobe Illustrator CS5, and videos were made using Apple QuickTime Pro. Box-and-whisker plots show median, first and third quartile, observations within 1.5 times the interquartile range, and all individual data points.