Tau interaction with microtubules in vivo

Neuronal microtubules (MTs) are essential for neurite extension, axonal guidance, and communication between the cell body and neuronal processes. Dynamic properties of individual MTs and the architecture of the MT array are regulated by MT-associated proteins (MAPs). One member of the MAP family, tau, has received particular attention since hyperphosphorylation of tau has been linked to the formation of abnormal filaments characteristic of Alzheimer's and other neurodegenerative diseases (`tauopathies') (Garcia and Cleveland, 2001; Lee et al., 2001).

A large body of experimental evidence indicates that tau promotes MT stability and induces MT bundling in vitro (Drechsel et al., 1992; Kanai et al., 1992). Expression of tau in non-neuronal cells is sufficient to induce formation of neurite-like processes that contain arrays of parallel MTs (Baas et al., 1991; Knops et al., 1991; Lee and Rook, 1992), while lowering the level of tau expression in neuronal cells inhibits axonal growth (Caceres and Kosik, 1990). Although mice deficient in tau are essentially normal (Harada et al., 1994), knockout of both tau and MAP1B leads to serious neuronal defects (Takei et al., 2000), suggesting functional redundancy between these (and, perhaps other) MAPs. Interestingly, in humans, mutations in tau linked to some neurodegenerative diseases reduce the ability of tau to promote MT assembly (D'Souza et al., 1999; Hasegawa et al., 1999). Collectively, these data predict that tau-induced stabilization of the MT cytoskeleton is essential for proper development of nerve processes.

Tau decoration of MTs appears to be uniform (Murphy and Borisy, 1975; Sloboda and Rosenbaum, 1979), and its binding to MTs is non-cooperative (Biernat et al., 1992). Electron microscopy studies suggest that tau binds to the outer surface of MTs and stabilizes tubulin-tubulin interactions along protofilaments (Al-Bassam et al., 2002). Binding of tau to MTs is phosphorylation-dependent and is regulated by several protein kinases (Drewes et al., 1998). Different isoforms of tau include three or four MT-binding repeats which are highly conserved between different species and different members of the MAP family (Goedert et al., 1996; Heidary and Fortini, 2001). 3R-tau and 4R-tau are developmentally regulated, and imbalances in their expression have been linked to neurodegenerative diseases (Hutton et al., 1998; Spillantini et al., 1998), suggesting a distinct function for each isoform. Further, even though both 3R- and 4R-tau induce neurite outgrowth in transfected cells, the onset of this outgrowth in 4R-tau-expressing cells is significantly more rapid (Yu et al., 2002).

Immunocytochemical evidence indicates that tau density progressively increases in a proximal-to-distal direction along the length of growing axons (Black et al., 1996). This somewhat paradoxical co-localization of tau with a dynamic subset of axonal MTs at the distal end of the axon has led to the suggestion that, in addition to its well-established function as MT stabilizer, tau may play a role in signaling during growth cone navigation (Shahani and Brandt, 2002). It has also been suggested that tau may determine the spacing between MTs (Chen et al., 1992), provide a link between MTs and other membrane or cytoskeletal structures (Brandt et al., 1995; Cunningham et al., 1997), alter mechanical properties of MTs (Felgner et al., 1996), and affect transport of membrane organelles along MTs (Ebneth et al., 1998; Stamer et al., 2002).

Progress towards understanding the functional properties of tau-MT interaction in neuronal cells has been limited by the technical difficulties of detecting individual MTs in living neurons. In this study, we took advantage of a unique morphological feature of Xenopus embryo neurons in culture. When the cultures are prepared on strongly adhesive substrate, individual tau-decorated MTs can be resolved with fluorescence microscopy throughout the entire axonal length. Using a combination of time-lapse microscopy and a photobleaching method, we examined the dynamics of tau association with MTs in living cells. We report that association of tau with individual MTs in neurons is very rapid, does not depend on MT location along the axon, and is strongly affected by MT curvature. Unexpectedly, we found that the MT-stabilizing drug taxol displaced tau from MTs. Our results reveal unexpectedly rapid and complex interaction of tau with MTs in live neurons.

pEGFP-Tub encoding a fusion protein consisting of EGFP and human α-tubulin was from Clontech (Palo Alto, CA). Preparation of GFP-tau constructs has been described previously (Yu et al., 2002).

Xenopus eggs were fertilized and dejellied in vitro as described previously (Chang et al., 1998). At the two-cell stage, the eggs were injected with 10-15 nl of 20 μg/ml DNA plasmids using the air pressure injector Picospritzer II (General Valve, Fairfield, NJ). The diameters of the pipettes used for injection ranged from 9 to 18 μm. Injections were performed in Ringer's solution (115 mM NaCl, 2 mM CaCl₂, 2.5 mM KCl, and 10 mM HEPES, pH 7.6) supplemented with 5% Ficoll (Sigma). One hour after injection the embryos were transferred to 20% Ringer's solution containing 5% Ficoll, and 2-3 hours later they were transferred to 20% Ringer's solution. Development of embryos injected with the plasmid encoding GFP-tau23 or GFP-tau24 was slowed down by approximately 20% in comparison with control (uninjected or injected with vector encoding GFP alone) embryos. However, up to 25% of injected embryos developed normally (the actual number in a particular experiment strongly depended on the quality of eggs). The eggs were allowed to develop to stages 19-24 and were then used for the preparation of neuronal cultures.

Cy3-tubulin was a generous gift of Gary Borisy (Northwestern University, Chicago, IL). Xenopus embryos were injected with 10-25 nl of 10 mg/ml Cy3-tubulin as described elsewhere (Chang et al., 1998).

Xenopus embryo neuronal cultures were prepared according to previously reported methods (Popov et al., 1993; Spitzer and Lamborghini, 1976). Briefly, the neural tubes of embryos at stages 19-24 were dissociated in Ca²⁺- and Mg²⁺-free solution (115 mM NaCl, 2.6 mM KCl, 10 mM HEPES, and 0.4 mM EDTA, pH 7.6) for 20-30 minutes. Dissociated cells were plated on glass coverslips precoated with concanavalin A (0.1-1.0 μg/cm²; Sigma). The cultures were kept at 20°C in a culture medium consisting of 50% (v/v) Ringer's solution, 49% L-15 Leibovitz medium (GIBCO Life Technologies), and 1% fetal bovine serum (GIBCO). The neurons were used for experiments 14-20 hours after plating. Human 356 foreskin fibroblasts were kindly provided by T. Svitkina (Northwestern University) and were cultured in HAM's F-10 medium supplemented with 10% fetal bovine serum (GIBCO). Okadaic acid and calyculin A were from Sigma.

We obtained digital images of cells at room temperature using an inverted microscope (TE200, Nikon) equipped with a 60× PlanApo objective (NA 1.4) and a 100-W mercury lamp. The light passed through an infrared-blocking filter, neutral density filters, and a GFP-wide-band filter. Images were acquired with a charge-coupled device (CCD) camera (CoolSnapHQ, Roper Scientific) driven by IPLab imaging software (Scanalytics). The image was directly projected onto the CCD chip and the final magnification was ≈0.11 μm/pixel. Exposure time was 0.2-1.0 seconds, and images were collected at 1- to 3-second intervals. Typically, 100-200 frames were captured. The culture medium was supplemented with the oxygen-depleting preparation, Oxyrase (Oxyrase, Mansfield, OH), during observation, in order to reduce photodamage. Neither axonal morphology (as inspected with differential interference contrast microscopy) nor MT integrity was affected by exposure to the fluorescent light during image acquisition. Images were processed with IPLab and Photoshop (Adobe Systems). Quantification of data was performed with IPLab software.

The beam of the 150-mW multimode argon laser (Spectra-Physics, Fremont, CA) was channeled into the epi-illumination port of a Nikon 200 microscope. A cylindrical lens with a 700 mm focal distance and a 60×, 1.4 NA objective lens were used to produce a 4 μm×15 μm or 2 μm×15 μm focused beam in the specimen plane. Exposure was 50-1000 milliseconds. Laser irradiation of neurons under these conditions had no detectable effect on axonal growth or on MT assembly/disassembly. Cells were imaged before and immediately after photobleaching. In general, a 2- to 4-μm-wide region on a single MT or MT bundle was bleached by a single laser pulse and allowed to recover. Images were typically collected at 1-second intervals. To evaluate the rate of fluorescence recovery of cytoplasmic GFP-tau, ten images per second were collected without shutter opening/closure. To measure fluorescence intensity of a single MT, a small linear region of interest, 1×10 pixels (≈0.1×1 μm) in size, was selected along the MT length, and the average fluorescence intensity was calculated. The background fluorescence was measured in a similar 1×10 pixel MT-free region of the cytoplasm. Typically, background fluorescence was averaged over 3-4 regions. Background-subtracted MT intensity was plotted as a function of time and fitted with a single exponential function to calculate the half-time t_g (Bulinski et al., 2001).

A beam from a 150-mW argon laser (Spectra-Physics) was steered through the single mode fiber into an Olympus total internal reflection fluorescence (TIRF) illuminator and focused at the back focal plane of the objective lens (Planapo 60×, 1.45 NA, Olympus). The focal point was moved off-axis to the most peripheral position that still allowed the beam to pass through the objective. Light underwent total internal reflection at the glass interface. Images were recorded with a CoolSnapHQ cooled CCD camera (Roper Scientific) driven by IPLab software.

Neurons expressing GFP-tau were extracted in a MT-stabilizing buffer (60 mM PIPES, 1 mM MgCl₂, 5 mM EGTA, 0.1% Triton X-100, 10 μM taxol, pH 6.8) and examined under the fluorescent microscope during extraction.

We made all position, intensity and length measurements on 12-bit digital images using the analysis function of the IPLab software. To determine significant differences between averages, unpaired t-tests assuming equal variance were performed.

CHO cells in DMEM/F12 (Invitrogen, Carlsbad, CA) containing 10% fetal bovine serum were cultured at 37°C with 5% CO_2. Cells were transfected with PcNA3 vector, wild-type tau, or GFP-tau plasmids. Twenty-four hours after transfection, cells were harvested and resuspended in G-PEM buffer (80 mM PIPES, 0.5 mM MgCl₂, 1 mM EGTA) containing Complete Protease Inhibitor Tablets (Roche Molecular Biochemicals, Indianapolis, IN). Cell suspensions were frozen and thawed twice and then homogenized with a teflon-glass homogenizer (Kontes Glass Company, Vineland, NJ) on ice. Unbroken cells and cellular debris were removed by centrifuging at 20,000 g at 4°C for 20 minutes. Each supernatant was divided into two 1.5 ml tubes and after incubation at 32°C for 3 hours, one tube of each group was treated with 5 μM taxol for 10 minutes. A high-speed centrifugation was applied to spin down microtubules at 146,000 g at 32°C for 30 minutes. Pellets were separated by 12% SDS-PAGE and transferred to Immobilon transfer membrane (Millipore, Bedford, MA). Blotting was done as described previously (Yu et al., 2002). Polyclonal antibody against the C-terminal of human tau and monoclonal antibody against α-tubulin were purchased from DakoCytomation (Carpinteria, CA) and Sigma (St Louis, MO), respectively.

Time-lapse sequences were linearly adjusted for brightness and contrast and converted to QuickTime movies (animation compression, medium quality, 12-bit depth) with IPLab software (see supplementary material).

The plasmid encoding GFP-tau23 fusion protein (GFP-tau) was injected into one of the blastomeres of two-cell-stage Xenopus embryos. Injected embryos were allowed to develop to stage 19-24 and then were used for the preparation of cell cultures. Approximately 20-25% of cells expressed GFP-tau as visualized with fluorescence microscopy. Experiments were performed on neurons and on fibroblast-like cells present in cell cultures.

Both in fibroblasts (Fig. 1) and in neurons (Fig. 2), GFP-tau decorated individual MTs and MT bundles. The background fluorescence in MT-free regions of the cytoplasm of transfected cells reflected the presence of free cytoplasmic GFP-tau; this background fluorescence, however, was sufficiently low to allow for the detection of individual MTs. In fibroblasts, the architecture of the MT array strongly depended on the level of tau expression. At the lower levels of expression, individual tau-labeled MTs could be resolved throughout the cytoplasm. We observed both non-centrosomal (Fig. 1A) and centrosomal (Fig. 1D) organization. At the higher expression levels, the majority of MTs were packed into tight bundles (Fig. 1B). In agreement with previous reports (Knops et al., 1991; Lee and Rook, 1992), expression of tau in fibroblasts often led to formation of neurite-like processes filled with MT bundles (Fig. 1C). Formation of such bundles could be observed in living cells at the leading edge of the fibroblasts, where individual GFP-tau-labeled MTs were easily detected.

In general agreement with the dynamic instability model, individual MTs displayed periods of growth and shortening (Fig. 1E; see also Movie 1 in supplementary material). MTs appeared uniformly labeled along their length (Fig. 1E), and no gradient of fluorescence intensity was observed at the distal tip of MTs (Fig. 1F). `Collision' of MTs with each other typically led to the formation of small MT bundles (Fig. 1G). Such bundling of individual MTs is not detected in GFP-tubulin-transfected fibroblasts, suggesting that the lateral interaction between MTs that we observe is tau-mediated. In agreement with phosphorylation dependence of tau-MT interaction (Drewes et al., 1998), treatment of cells with the phosphatase inhibitor calyculin A (200 nM; 10-15 minutes) rapidly removed a significant fraction of tau from MTs (Fig. S1, supplementary material); a similar effect was observed after treatment of cells with another phosphatase inhibitor, okadaic acid (2 μM; data not shown). These results indicate that, in fibroblast-like cells in Xenopus embryonic cultures, GFP-tau is able to bind to individual MTs in a phosphorylation-dependent manner, and to induce MT bundling and formation of neurite-like processes. Our data are in general agreement with previous studies which demonstrated that fusion of GFP either to the N-terminus or to the C-terminus of tau does not significantly affect the functional properties of tau (Kaech et al., 1996; Yu et al., 2002).

In Xenopus neurons (Tanaka and Kirschner, 1991), as well as in other neuronal cell types (Kabir et al., 2001), individual fluorescently labeled MTs can be resolved at the growth cone region, whereas detection of single MTs in MT bundles along the axon is difficult or impossible at the light microscopy level. However, when Xenopus embryo neuronal cultures are prepared on a strongly adhesive substrate (Chang et al., 1999), MT bundles often splay into individual polymers at the lateral edges of the axonal shaft and/or in the vicinity of the branch points (Fig. 2A). This allows for the direct comparison of tau binding to MTs at different axonal segments (Fig. 2B-D). In contrast to previous immunocytochemical studies that documented a higher concentration of tau at the distal MTs in comparison with the proximal axon in rat sympathetic neurons (Black et al., 1996), we found that, in Xenopus neurons, the density of MT-bound tau did not depend on MT location along the axon (Fig. 2D).

To compare the concentrations of free cytoplasmic GFP-tau at the different axonal regions, we measured the intensity of fluorescence in flat MT-free regions of cytoplasm with TIRF microscopy. The fluorescence signal in this case is proportional to the concentration of cytoplasmic GFP-tau in the thin (≈100 nm) submembrane layer of the cytoplasm. The fluorescence intensities at the growth cone region and at the proximal axon (300-400 μm from the growth cone) were 43±5.5 and 54±7.3, respectively (arbitrary units; means±s.e.m., data from six different neurons). These data are consistent with the notion that the concentration of tau is uniform throughout the axoplasm.

Taken together, the results of these experiments indicate that, within the resolution limit of light microscopy, binding of tau to neuronal MTs is uniform along individual MTs, does not depend on MT location in the axonal shaft, and is independent of the dynamic state of the MT.

Local mechanical properties of a MT (such as flexural rigidity) on a submicrometer scale are likely to be affected by the local density of MT-bound MAPs. It is usually assumed that the concentration of MT-bound MAPs does not vary along the length of the MT. To test this assumption, we performed time-lapse imaging of fluorescent tau-labeled MTs in living fibroblasts and in neurons. We focused on the flat regions of the cells (leading edge of fibroblasts and growth cone-like protrusions at the lateral edge of the axonal shaft), where individual tau-decorated MTs could be resolved. Results of a typical experiment are presented in Fig. 3. At the lamella region of the cell and at the growth cone, MTs often buckled, presumably because of the activity of retrograde actin flow (Waterman-Storer and Salmon, 1997) or because of compressive forces generated by tubulin polymerization itself. As described above (Figs 1, 2), relatively straight MTs displayed a uniform pattern of GFP-tau labeling. However, for buckled MTs the concentration of GFP-tau was significantly higher at the regions of higher MT curvature (Fig. 3; see also Movie 2 in supplementary material). The density of MT-bound tau rapidly adjusted to MT shape (Fig. 3B-D). For quantitative analysis of the data we measured fluorescence intensity of MTs as a function of the radius of MT curvature (Fig. 3F). The dependence of tau density on the radius of MT curvature was similar for fibroblasts and neurons. Concentration of tau at the region with a radius of curvature of 0.5 μm was approximately twice as high as that at the straight segment of a MT. Similar quantitative analysis of individual MTs in GFP-tubulin-expressing neurons (Fig. 3F) indicated that fluorescence intensity of MTs remains uniform regardless of their curvature. This suggests that enhanced binding of tau to highly curved MT regions is not an optical artifact because of our CCD detection system.

The ability of tau to follow rapidly changing MT shape suggests that interaction of tau with individual MTs is very rapid. To measure the dynamics of tau-MT interaction, we performed photobleaching experiments on single MTs and MT bundles both in fibroblasts and in neurons expressing GFP-tau. Small regions of cells were illuminated with a brief flash of argon laser and time-lapse imaging of cells was performed to measure t_g of fluorescence recovery of MTs (the corresponding rate constant for the photobleaching recovery is identical to the dissociation rate constant, k_off) (Bulinski et al., 2001). Fig. 4 illustrates results of a typical photobleaching experiment performed on fibroblasts (Fig. 4A,C) and on neurons (Fig. 4B,D). In both cell types, some recovery of fluorescence along single MTs (Fig. 4A,B) was visible within the first few seconds and recovery appeared to be complete within 5-10 seconds after the bleaching. Fluorescence recovery along neurite-like processes produced by fibroblasts (Fig. 4C) was significantly slower. To address a potential problem of our approach with regard to the full functionality of human tau in a heterologous Xenopus system, we repeated photobleaching experiments in human 356 fibroblasts. As with Xenopus cells, in 356 fibroblasts we observed rapid fluorescence recovery of GFP-tau-decorated single MTs after photobleaching (Fig. 4E) with a characteristic t_g≈5.8±0.9 s (n=7). This rate of recovery was comparable with that observed in Xenopus cells (see Table 1 below).

Previous studies have demonstrated that photobleaching does not significantly affect cytoskeletal dynamics; in agreement with these data, brief exposure to the laser beam in our experiments had no detectable effect on growth cone motility or the integrity of axonal MTs. Moreover, t_g did not depend on the duration of the pulse or the intensity of illumination, and multiple cycles of tau binding could be observed after repeated photobleaching of the same axonal regions (data not shown). Therefore, it seems unlikely that photobleaching significantly affects tau-MT interaction.

Two processes contribute to the rate of fluorescence recovery of individual MTs after photobleaching. The first is the diffusion of free cytoplasmic GFP-tau into the bleached zone; the second is the dissociation of bleached GFP-tau from MTs. Only the second (but not the first) process reflects the true dynamics of tau-MT interaction. To estimate the contribution of diffusion of cytoplasmic tau to the fluorescence recovery of individual MTs, we repeated photobleaching experiments on GFP-tau-expressing neurons incubated overnight at 4°C and pretreated with nocodazole (10 μg/ml, 1 hour). Under these experimental conditions, some of the well-spread regions of the axonal shaft were devoid of MTs (Fig. 5). Photobleaching of a small (≈4 μm) region in these neurons was followed by rapid recovery, with a characteristic t_g≈400±52 ms (n=7). Therefore, for the typical bleached area (≈4 μm), fluorescence recovery of cytoplasmic GFP-tau is 10-50 times faster than that of individual MTs and MT bundles (see below). In this situation, the experimentally measured t_g of the fluorescence recovery of individual MTs reflects dynamics of tau-MT interaction with an accuracy of 2-10%.

In each photobleaching experiment performed on GFP-tau-decorated MTs, we measured t_g for single MTs or MT bundles and the extent of fluorescence recovery (the intensity of fluorescence in the bleached segment ≈1-5 minutes after photobleaching normalized to the initial fluorescence signal). The results of these measurements are summarized in Table 1. As suggested by the visual analysis of time-lapse sequences after photobleaching, association/dissociation of tau along single MT was dynamic both in neurons and in fibroblasts, with t_g≈2-3 seconds. The slowest recovery of tau fluorescence was observed in neurite-like processes produced by fibroblasts (t_g≈16 seconds). In this case, as well as for MT bundles in the axonal shaft, fluorescence recovery was incomplete (≈55% recovery). This suggests that a fraction of MT-bound tau is tightly associated with MTs. Somewhat unexpectedly, no significant difference was found between 3R-tau and 4R-tau in their association with MTs (Table 1).

Biochemical and structural studies concerning interaction of tau with MTs are typically performed in the presence of the MT-stabilizing drug taxol (Gustke et al., 1994; Littauer et al., 1986; Melki et al., 1991). Recent studies have demonstrated that in the presence of taxol, the dynamics of association of ensconsin (E-MAP-115) with MTs is slowed down by an order of magnitude (Bulinski et al., 2001). To investigate whether interaction of tau with MTs can be modified by exposure to taxol, we attempted to conduct photobleaching experiments on GFP-tau-expressing cells in the presence of taxol (1 μM). Unexpectedly, within minutes of taxol incubation, both in fibroblasts and in neurons (Fig. 6; see also Movie 3 in supplementary material) most single MTs decorated with tau became undetectable with fluorescence microscopy. Simultaneously, we observed an increase in the level of background fluorescence in the cytoplasm. In contrast, bundles of cytoplasmic MTs were still visible after taxol incubation although the intensity of their fluorescence had decreased. Incubation of Xenopus cell cultures with vinblastine (1 μM) for 10 minutes had no visible effect on association of GFP-tau with MTs (Fig. 6C,F). These results suggest that taxol induces rapid dissociation of tau from single MTs.

Incubation of Xenopus neurons with low concentrations of taxol does not change the total amount of polymeric tubulin (Zakharenko and Popov, 1998); hence it seems unlikely that the observed effect of taxol is because of rapid depolymerization of MTs. To test the effect of taxol on the integrity of single MTs directly, we repeated experiments on cells loaded simultaneously with Cy3-tubulin (to observe the architecture of the MT array) and with GFP-tau. As expected, single MTs could easily be detected in the cytoplasm after exposure to taxol (Fig. 7D). In contrast, no association of GFP-tau with MTs was observed (Fig. 7C). As with Xenopus cells, incubation with taxol induced rapid dissociation of tau from MTs in human 356 fibroblasts (Fig. 8; see also Movie 4 in supplementary material). This suggests that the effect of taxol is not specific to the heterologous Xenopus expression system.

The taxol-induced dissociation of tau from MTs in living cells was totally unexpected. We wanted to investigate whether a similar phenomenon can be observed after detergent extraction of cells in a MT-stabilizing buffer containing taxol (Tanaka and Kirschner, 1991). Such extraction is known to preserve individual MTs while rapidly removing free tubulin dimers and other cytoplasmic proteins. Within seconds after the onset of extraction, single GFP-tau-labeled microtubules became significantly dimmer and acquired a characteristic `speckled' pattern (Fig. 9A; see also Movie 5 in supplementary material), indicating rapid dissociation of tau from individual MTs (Waterman-Storer et al., 1998). Interestingly, both in GFP-tau-transfected fibroblasts (Fig. 9B) and in neurons (Fig. 9C) a significant fraction of GFP-tau remained associated with MT bundles for as long as 30 minutes after extraction. This effect was particularly robust for MT bundles in fibroblasts (Fig. 9B). These data are in general agreement with the results of photobleaching experiments, which indicated that a significant fraction of tau is tightly associated with MT bundles (Table 1).

Results from living cell studies showed taxol-induced dissociation of GFP-tau from MTs, which was found both in Xenopus cells and in human 356 fibroblasts. It is not feasible to observe the effect of taxol on wild-type tau association with MTs in living cells using fluorescence techniques. To estimate further the dissociation of endogenous tau from MTs promoted by taxol treatment, CHO cells were transfected with wild-type tau or with GFP-tau and homogenized. Standard cold-heating cycle was used to induce MT polymerization and MT pellet from the cellular supernatant was isolated on SDS-PAGE gels. Western blots indicated that both wild-type tau and GFP-tau expressed in CHO cells co-precipitated with microtubules (Fig. 10A). Ten-minutes taxol (5 μM) treatment dramatically decreased the amount of wild-type tau or GFP-tau co-precipitation with MTs (Fig. 10A). Taxol treatment increased microtubule formation in CHO cells that did not express tau protein. However, this did not occur in CHO cells expressing either wild-type tau or GFP-tau (Fig. 10B). It thus appears that wild-type tau and GFP-tau are displaced from microtubules by tau in vitro. This is consistent with the results of experiments in living cells (Figs 7, 8).

Tau is a major neuron-specific MAP implicated in MT stabilization. Tau binds to MTs with affinity in the low micromolar range and demonstrates neither enzymatic activity (e.g. nucleotide hydrolysis) nor high affinity binding for nonpolymeric tubulin (Mandelkow and Mandelkow, 1995). As a step towards understanding tau-MT interaction in living cells, we expressed a fetal isoform of tau (tau₂₃) fused to GFP in Xenopus embryo neurons in culture and performed imaging of fluorescent tau using a combination of time-lapse microscopy and photobleaching. We were able to detect individual MTs decorated with tau both in neurons and in non-neuronal cells. We interpret fluorescence intensity of single MTs in GFP-tautransfected cells as a quantitative measure of tau binding to MTs. We also assume that photobleaching studies provide information on the dynamics of tau-MT interaction in living cells. The validity of our experimental approach is based on several assumptions. First, the fusion of GFP to the N-terminus of tau does not affect the functional properties of tau-MT interaction. Second, human tau used in this study is functional in Xenopus cell cultures. Third, photobleaching does not significantly alter the dynamics of tau-MT interaction. To a large extent, our rationale behind all these assumptions is based on several previous studies in which similar assumptions have been made and tested (e.g. Chang et al., 1998; Dehmelt et al., 2003; Kaech et al., 1996; Murphy and Borisy, 1975; Ozer and Halpain, 2000; Rodionov et al., 1994; Yu et al., 2002). In agreement with these findings, we further provide evidence that in our specific experimental conditions, GFP-tau binds to individual MTs, does not interfere dramatically with MT assembly/disassembly, induces MT bundling in non-neuronal cells, and binds to MTs in a phosphorylation-dependent manner (Figs 1, 2; see also Fig. S1 in supplementary material). We provide evidence indicating that the major results of this study can be reproduced in human cell line (Figs 4, 8) and in vitro (Fig. 10). Finally, as discussed in Results, we performed several conventional controls for potential phototoxic effects of photobleaching. Therefore, it seems likely that our video microscopy approach provides reliable information concerning tau-MT interaction in living cells.

There are four unexpected findings in this report: (a) association of tau with MTs is very fast, with a t_g of 3-5 seconds; (b) binding of tau to axonal MTs does not depend on MT location in the axonal shaft or on MT assembly/disassembly status; (c) binding of tau to MTs varies with MT curvature; and (d) treatment of cells with taxol induces rapid dissociation of tau from individual MTs in living cells.

There is now an abundance of evidence suggesting rapid interaction of MAPs with MTs. Thus, previous photobleaching studies reported MAP-MT interactions with t_g ranging from 44 seconds to 5 minutes for different members of the MAP family (Bulinski et al., 2001; Olmsted et al., 1989; Scherson et al., 1984). Recent measurements of t_g for ensconsin, a protein expressed predominantly in epithelial tissue, suggested a remarkably rapid interaction of this protein with MTs (≈4 seconds) in non-neuronal cells (Bulinski et al., 2001). Despite the central role of tau in the regulation of the MT array, direct measurements of the dynamics of tau association with MTs in living neurons have not been reported. This is probably partly because of the difficulties in detecting individual MTs in living neurons. Indirect evidence based on the differential rates of slow axonal transport of tau and MAP2 indicates that the association of tau with MTs is dynamic (Mercken et al., 1995).

Both in neurons and in fibroblast-like cells present in Xenopus embryo cell cultures, we were able to detect individual MTs labeled with GFP-tau. In agreement with previous studies (Kaech et al., 1996), the GFP-tau fusion protein appeared to retain functional activity (Fig. 1), making it a faithful reporter of tau binding to MTs in living cells. In this study, we have not attempted to estimate the concentration of exogenous (GFP-labeled) tau in the cytoplasm. However, a strong fluorescence signal associated with individual GFP-tau-decorated MTs and low background fluorescence (because of unbound cytoplasmic GFP-tau) suggests that the concentration of exogenous tau was significantly lower than K_d (which is in the low micromolar range for tau and other members of MAP family).

Systematic analysis of tau-MT binding with the photobleaching method indicated that tau association with single MTs in Xenopus cells is very dynamic, with t_g of the order of 3 seconds (see Table 1). Similar kinetics of association with single MTs has been reported for ensconsin (Bulinski et al., 2001) and for tau-MT binding in vitro (Makrides et al., 2004). Although esconsin does not affect MT dynamics (Faire et al., 1999), tau is believed to be the major promoter of MT polymerization. It may seem paradoxical that the association of this MT stabilizer with MTs is so transient. Although the functional significance of such rapid tau-MT interaction remains unclear, it is tempting to speculate that it may be advantageous for proper activity of the kinesin-based axonal transport system. Indeed, in vitro tau reduces the probability of attachment of kinesin to MTs (Trinczek et al., 1999), and overexpression of tau in N2A neuroblastoma cells inhibits transport of vesicles and mitochondria (Ebneth et al., 1998). The transient nature of tau-MT interaction is likely to decrease the probability of irreversible steric interference between kinesin or its cargo and tau and, thus, to facilitate transport activity. It should be noted that photobleaching measurements (Table 1) indicate that fluorescence recovery of tau-decorated MT bundles is incomplete. This suggests that a significant fraction of tau is tightly associated with MT bundles (but not with individual MTs). This conclusion is further supported by the results of detergent extraction experiments (Fig. 9). Therefore, the dynamics of tau interaction with single MTs and with MT bundles appear to be significantly different.

Axonal MTs at the growth cone region are generally believed to be dynamic, whereas those along the axonal shaft are relatively stable (Baas et al., 1993). This difference in MT stability is reflected by different posttranslational modifications of tubulin at different axonal regions. In addition, the distal tips of growing MTs are distinct from the bulk of the MT lattice both structurally and biochemically (Desai and Mitchison, 1997). These regional differences in MT structure/molecular composition may lead to preferential association of tau with various axonal segments. It should be noted, however, that the studies concerning distribution of tau along nerve processes were primarily based on immunocytochemistry. Although this method generally reliably reports localization of MAPs such as MAP2, in the case of tau results strongly depend on a particular fixation/extraction protocol (Black et al., 1996). The ability to detect individual tau-decorated MTs in live Xenopus neurons allows for a direct comparison of the amount of MT-bound tau at different axonal regions. Our results indicate that the concentration of cytosolic tau (as measured with TIRF microscopy) and the binding of tau to MTs are similar in different axonal regions. Furthermore, results of our photobleaching experiments suggest that technical difficulties associated with immunocytochemical localization of tau may be related to the unusually fast dynamics of tau-MT interaction. It is possible, however, that the difference in the results obtained with photobleaching and immunocytochemical methods can be related to different experimental systems.

One intriguing possibility concerning the functional significance of rapid tau-MT interaction is suggested by the results of experiments in which we performed time-resolved measurements of tau association with buckling MTs. Previous experimental evidence (Felgner et al., 1996) suggests that tau and other MAPs are evenly distributed along the length of MTs. Moreover, biochemical evidence indicates that binding of tau to MTs is non-cooperative. Therefore, observations of strong preferential binding of tau to MT regions with higher curvature (Fig. 3) came as a surprise. Previous biophysical measurements of MT flexural rigidity (Felgner et al., 1996; Gittes et al., 1993; Kis et al., 2002) were done under the assumption that MTs can be modeled as isotropic and homogeneous beams; in this case their flexural rigidity (or bending stiffness) can be characterized by a single parameter, Young's modulus. Our results indicate that association of tau with buckling MTs is highly non-uniform. Because the association of MAPs with MTs has a profound effect on the flexural rigidity of MTs, we hypothesize that transient association of tau with highly curved MT segments may increase local flexural rigidity of the MT to resist compressive forces acting on the MT lattice. This is likely to contribute to a relatively straight shape of neuronal MTs and to prevent formation of MT `tangles'. It is noteworthy that the rapid interaction of tau with MTs that we observe with photobleaching is crucial for tau's ability to adjust to a changing MT shape. The molecular mechanisms of preferential association of tau with highly curved MT segments remain to be investigated.

A large number of previous biochemical and structural studies have investigated the interaction of MAPs with MTs; these studies have typically been performed on taxol-stabilized MTs. Under these experimental conditions, tau binds to the outer surface of a MT (Al-Bassam et al., 2002; Santarella et al., 2004). Because taxol binds to a specific site on β-tubulin located inside a MT (Nogales et al., 1999), rapid dissociation of tau from individual MTs after incubation with taxol (Figs 6, 8; see also Movies 3 and 4 in supplementary material) came as a complete surprise to us. Taxol may influence tau binding to MTs indirectly by affecting the phosphorylation state of several proteins (Shtil et al., 1999; Yujiri et al., 1999) which may be involved in the regulation of tau-MT interaction. Although at present we cannot completely exclude this hypothesis, the very rapid effect of taxol on the association of tau with MTs makes this possibility unlikely. Therefore, we hypothesize that taxol binding induces a conformational change of a MT lattice that is sufficient to affect binding of tau to the MT. Alternatively, tau and taxol may compete for the same binding site on a MT (Kar et al., 2003). Regardless of the molecular mechanism of interaction between tau and taxol, our results demonstrate that within minutes after the onset of treatment with 1 μM taxol, tau largely dissociates from single MTs in live cells. This is likely to change the mechanical properties of the MT array and contribute to neuropathy which is a common complication of taxol-induced toxicity (Peltier and Russell, 2002).

Taken together, the results of our experiments reveal unexpectedly rapid dynamics of tau-MT interaction in live cells and suggest that this phenomenon may contribute to local changes in the mechanical properties of MTs. The ability to visualize tau-MT interaction in live cells should provide further information on the involvement of this MAP in axonal growth and neuronal plasticity.

We thank Tanya Andric and Barry Alpher for help with the manuscript and Andrey Tsvetkov for plasmid preparation. Supported by the National Institutes of Health grants NS3350 and NS39417 to S.V.P.