Tau interacts with src-family non-receptor tyrosine kinases

Microtubules are dynamic structures in cells that must respond to cellular signals such as those that induce cell division, cell motility, and process outgrowth. The ability of microtubules to respond to diverse signals and to form different functional structures may be facilitated by microtubule associated proteins (MAPs). The microtubule associated protein tau plays an important role in neuronal differentiation and axonal development (reviewed by Mandell and Banker, 1996a; Kosik, 1997). Although its role in these processes has been attributed to its ability to promote microtubule assembly, much evidence now suggests that tau has other functions in addition to the stabilization of microtubules. These functions may differentiate tau from other MAPs and may also explain why cells express a variety of MAPs. The first evidence that tau was not strictly associated with axonal microtubules stemmed from tau’s presence in the somatodendritic compartment of neurons where tau was associated with ribosomes and polysomes as well as microtubules (Papasozomenos and Binder, 1987). Tau has also been found in the nucleolar organizing region (Loomis et al., 1990) and in the plasma membrane of neuronal cells (Brandt et al., 1995). Moreover, non-neuronal cells, such as astrocytes and oligodendrocytes, also express tau (LoPresti et al., 1995; Thurston et al., 1996; Migheli et al., 1988; Riederer and Innocenti, 1991; Gu et al., 1996; Cross et al., 1993). Non-microtubule associated tau has also been reported (Esmaeli-Azad et al., 1994; DiTella et al., 1994; Mandell and Banker, 1996b) as well as tau that associates with dynamic microtubules (DiTella et al., 1994; Black et al., 1996). Tau’s ability to activate phospholipase C-γ in vitro (Hwang et al., 1996) and to bind to phosphatase 2A in vitro (Sontag et al., 1996) may also represent new functions for tau.

Phosphorylation of tau affects its ability to bind to microtubules and may be an important factor in Alzheimer’s disease (AD) where abnormally phosphorylated tau is a prominent component of the neurofibrillary tangles of AD (reviewed by Billingsley and Kincaid, 1997; Goedert et al., 1995; Lee, 1995). Abnormal tau filaments have also been found in other neurodegenerative diseases (Spillantini et al., 1997; Bird et al., 1997) and recently, the genetic defect in a group of autosomal dominantly inherited dementias has been found to be in the tau gene (Spillantini et al., 1998; Hutton et al., 1998; Poorkaj et al., 1998). It has been hypothesized that abnormalities in tau may alter microtubule stability in diseased neurons. However, as the function of tau extends beyond stabilizing microtubules, further insights into tau’s functions may suggest new mechanisms involved in the neuropathogenesis of AD and other neurodegenerative diseases.

Recently, we reported that the amino terminus of tau (which is not required for interactions with microtubules) associates with the plasma membrane (Brandt et al., 1995). This report now identifies src-family non-receptor tyrosine kinases as membrane associated components that interact with tau. Of the src family tyrosine kinases, we focused on fyn because of the high level of expression of the brain-specific isoform of fyn in neuronal tissues (Cooke and Perlmutter, 1989) and because of specific neurological defects observed in fyn−/− mice (Grant et al., 1992). Moreover, the pattern of fyn expression in brain correlates well with that of tau. Fyn is localized to developing axonal tracts throughout fetal brain (Bare et al., 1993) and is enriched in isolated growth cones (Bare et al., 1993; Bixby and Jhabvala, 1993; Helmke and Pfenninger, 1995; Meyerson and Pahlman, 1993). Similarly, tau is localized to axons in situ (Trojanowski et al., 1989; Brion et al., 1988; Binder et al., 1985) and is in the growth cones of cultured neurons (Gordon-Weeks et al., 1989; Rocha and Avila, 1995; DiTella et al., 1994; Black et al., 1996; Kempf et al., 1996) and in growth cones isolated from fetal brain (Lohse et al., 1996). Strikingly, both fyn and tau are required for process outgrowth. Neurons from fyn−/− mice lack process outgrowth when cultured on specific cell substrates (Beggs et al., 1994) while primary neurons treated with tau antisense oligonucleotides do not develop axons (Caceres and Kosik, 1990; Caceres et al., 1991). In addition, both fyn and tau are expressed in oligodendrocytes where they have been implicated in myelinogenesis (LoPresti et al., 1995; Umemori et al., 1994). Lastly, fyn, but not other tyrosine kinases, appears to be upregulated in a subset of neurons in Alzheimer’s disease brain; these neurons also react with an antibody specific for abnormally phosphorylated tau (Shirazi and Wood, 1993).

In this study, we report that tau interacts with src-family non-receptor tyrosine kinases. A PXXP motif in tau bound to the SH3 domains of fyn and src tyrosine kinases. Tau was associated with fyn in neuronal cells and was tyrosine phosphorylated. These findings suggest that tau is involved in signal transduction in neuronal cells.

GST fusion proteins were affinity purified on glutathione Sepharose beads (Pharmacia Biotech Inc., Piscataway, NJ) and quantified, as previously described (Reedquist et al., 1994). For binding assays using Escherichia coli synthesized tau protein, 10 μg of purified GST fusion proteins, noncovalently coupled to glutathione Sepharose beads, were incubated for 1 hour with tau with rocking at 4°C, in 0.5% Triton X-100, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM AEBSF. Beads were washed 6 times in wash buffer (0.5% Triton X-100, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl) and bound proteins were separated by SDS-PAGE and subjected to immunoblotting. For binding assays using transfected COS or SH-SY5Y cell lysates, 30 μg purified GST-fyn SH3 fusion protein was reacted with tau in lysates prepared from a 10 cm plate of cells (about 70% confluency) in the same lysis buffer with protease and phosphatase inhibitors (see below).

E. coli synthesized tau proteins were purified by DEAE cellulose chromatography as described (Brandt and Lee, 1993). Tau deletion mutant constructs pET-n(516) and pET-n(591) that encode tau amino acids 1-172 and 1-197, respectively (numbering according to 352 amino acid tau isoform), were constructed in pET-3d vector using low error PCR and standard techniques. Tau with an internal deletion of amino acids 170-178 (pET-tauΔ(170-178)) was constructed in pET-3d using low error PCR to synthesize an amino-terminal fragment and a carboxy-terminal fragment of tau encoding amino acids 1-169 and 179-352, respectively. The two fragments were joined using an AgeI site which resulted in the insertion of thr-gly between amino acids 169 and 179. DNA sequencing confirmed the sequence of the construct.

Transfections into NIH3T3 cells were performed as described (Lee and Rook, 1992). The vector expressing human tau 352 amino acid isoform was pRc/CMV (Invitrogen Corp., Carlsbad, CA). The vector expressing human fyn was Ap^rM8 (de Fougerolles et al., 1993). For immunofluorescence, cells were fixed as previously described (Lee and Rook, 1992) with the substitution of 0.1% saponin for 0.5% NP40. A 0.1% Triton X-100 permeabilization step (30 minutes) was included after fixation. Antibodies used for immunofluorescence were polyclonal affinity purified anti-tau or anti-fyn (FYN3; Santa Cruz Biotechnology, Santa Cruz, CA), tau14 (mouse monoclonal, Calbiochem-Novabiochem Corp., San Diego, CA) and anti-tubulin mouse monoclonal DMA1 (Sigma Chemical Co., St Louis, MO) or anti-tubulin rat monoclonal YL1/2 (Accurate Chemical and Scientific Corp., Westbury, NY). All secondary antibodies were obtained from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA). For triple staining, anti-mouse secondaries were double labeling grade with minimal cross reactivity for rat IgG. Affinity purified anti-tau was prepared from rabbit antiserum raised against bovine brain tau as previously described (Pfeffer et al., 1983), using an Affigel column coupled to E. coli synthesized tau.

Confocal microscopy was performed using a Bio-Rad MRC-600 attached to a Nikon Optiphot (Nikon, Inc. Melville, NY). Up to forty serial optical sections (∼0.5 μm section thickness) were collected for each cell. Individual channels of double-labeled cells were merged in Confocal Assistant (written by Dr Todd Breljie) or Adobe Photoshop. Projections were generated in Confocal Assistant using a maximum brightness algorithm. 3-D reconstructions from serial optical sections were performed using VoxBlast software (Vaytek, Fairfield, IA). Stereo pair was generated by reconstructing two views that differed by 7° in azimuth.

COS7 cell transfections were performed in 100 mm dishes using Lipofectamine (Life Technologies, Inc., Gaithersburg, MD) according to the manufacturer’s directions. For immunoprecipitations or GST fusion protein binding assays, transfected COS7 were lysed using 1 ml lysis buffer per 100 mm plate (lysis buffer: 0.5% Triton X-100 (Fluka Chemical Corp., Milwaukee, WI), 50 mM Tris-HCl, pH 7.5, 150 mM sodium choloride, 1 mM AEBSF (Calbiochem-Novabiochem Corp., San Diego, CA), 10 μg/ml leupeptin, 10 μg/ml pepstatin, 2 μg/ml aprotinin, 1 mM sodium vanadate, 10 mM sodium fluoride). Immunoprecipitations were performed using affinity purified anti-tau and Protein A-Sepharose 4B (Pharmacia Biotech Inc., Piscataway, NJ). 30 μg GST-fyn SH3 fusion protein was used per 1 ml lysate. After washing, Protein A-Sepharose or glutathione Sepharose bound proteins were eluted and resolved by 8% SDS-PAGE. Transfer to PVDF membranes (Immobilon-P, Millipore, Bedford, MA) and visualization using enhanced chemiluminescence (ECL) detection was according to the manufacturer’s instructions (NEN Life Science Products, Inc., Boston, MA). Antibodies used for probing blots were tau46.1, tau14, tau1, 5A6 (Johnson et al., 1997), tau5 (Carmel et al., 1996), anti-phosphotyrosine monoclonal 4G10 (Upstate Biotechnology, Lake Placid, NY), and anti-fyn monoclonal (Transduction Laboratories, Lexington, KY). Secondary antibody was sheep anti-mouse Ig, horseradish peroxidase linked F(ab′)₂ fragment (Amersham Life Sciences, Inc. Arlington Heights, IL).

SH-SY5Y cells were grown in RPMI 1640 with glutamine, 20 mM Hepes, 8% fetal calf serum, non-essential amino acids, essential amino acids, sodium pyruvate, and 55 μM β-mercaptoethanol at 5% CO₂. Ten to fourteen 150 mm dishes of SH-SY5Y cells were used for fractionation. Cells were scraped into 1 ml Tris-saline supplemented with protease and phosphatase inhibitors (2 μg/ml aprotinin, 10 μg/ml leupeptin, 10 μg/ml pepstatin, 10 mM sodium fluoride, 1 mM sodium vanadate) and 0.25 M sucrose. After homogenizing (20 strokes using a Teflon pestle), the suspension was centrifuged at 1,300 g, for 5 minutes at 4°C. The supernatant was termed post-nuclear supernatant (PNS) and supplemented with Triton X-100 to 1% for immunoprecipitation. A nuclear extract (NE) was prepared from the 1,300 g pellet using a salt extraction protocol (Lee et al., 1994). For immunoprecipitation, the NE was adjusted to 150 mM NaCl and 1% Triton X-100. To prepare a Triton soluble extract from the 1,300 g pellet, the pellet was incubated with 1 ml Tris-saline supplemented with protease and phosphatase inhibitors and 0.5% Triton X-100, 5 minutes at 4°C. Under these conditions, lipids and soluble proteins are released from the pellet but nuclear tau is not released (Greenwood and Johnson, 1995). After centrifugation at 350 g for 5 minutes at 4°C, the supernatant was removed and termed the Triton soluble extract (TSE) from the 1,300 g nuclear pellet. The TSE was adjusted to 1% Triton X-100 for immunoprecipitation. For immunoprecipitation, 4 μg anti-fyn polyclonal (Santa Cruz Biotechnology, Santa Cruz, CA) or 4 μg purified rabbit IgG (Sigma Chemical Co., St Louis, MO) were used per 0.5 ml of PNS or TSE. Comparisons between PNS, nuclear extract, or TSE were always performed from the same batch of cells, thereby using equal cell equivalents in each immunoprecipitation. For cytochalasin treatment, eight 150 mm plates of cells were incubated in the presence of 10 μg/ml cytochalasin B (20 μM) for 1 hour at 37°C before collecting cells by centrifugation. Homogenization, fractionation, and immunoprecipitations were performed as above.

Non-equilibrium pH gradient gels were performed as described (O’Farrell et al., 1977) using pH 3-10 ampholytes (Pharmalyte, Pharmacia Biotechnology, Inc. Piscataway, NJ) and 8% SDS-PAGE in the second dimension. E. coli synthesized human fetal tau, purified as described (Brandt and Lee, 1993), was added to lysates and SH3 binding samples as a control and reference marker. Blots were stripped by incubation with 62.5 mM Tris-HCl, pH 6.7, 100 mM β-mercaptoethanol, and 2% SDS for 1 hour at 50°C. Adherent phosphorescent markers (Identi-Kit, Diversified Biotech, Boston, MA) located at blot corners aided in aligning different exposures and different antibody probings of the same blot.

In our earlier study, we described the association of tau with the plasma membrane through its amino-terminal region and showed that overexpression of this region blocked neurite outgrowth (Brandt et al., 1995). A high pH wash removed tau from the membrane, suggesting that the association was due to an interaction between tau and a membrane bound protein. Since the amino terminus of tau contains 7 potential SH3 binding motifs (PXXP; Cheadle et al., 1994; Sparks et al., 1994), we reasoned that membrane proteins containing SH3 domains would be candidate proteins responsible for tau’s membrane association. Src-family non-receptor tyrosine kinases and the membrane-cytoskeletal protein spectrin are prominent proteins containing SH3 domains that are localized to the membranes of neuronal cells. Since tau was reported to associate with spectrin (Carlier et al., 1984), these data led us to test tau for an interaction with the SH3 domain of spectrin and other proteins.

To test tau for SH3 domain binding, recombinant tau was incubated with glutathione S-transferase (GST) fusion proteins of various SH3 domains that were non-covalently immobilized on glutathione Sepharose beads. Following incubation and washing, bead-bound tau was visualized by immunoblotting using anti-tau. While tau did not bind to the SH3 domain of spectrin, it did bind to the SH3 domains of the src-family non-receptor tyrosine kinases, fyn, lck, and src (Fig. 1A). The binding was specific since tau did not bind to the SH3 domains of abl, crk (Fig. 1A), or Grb2 (not shown), or with the SH2 domain of fyn. Moreover, a proline-rich peptide that binds to src-family SH3 domains with high affinity inhibited the binding of tau to the SH3 domain of fyn in a dose dependent manner (Fig. 1B).

Seven potential SH3 domain-binding PXXP motifs are located within the amino terminus of tau (Fig. 2, left). A panel of tau proteins with amino and carboxy-terminal deletions was used to locate the SH3 binding motif. Tau amino-terminal residues 1-163 (numbering according to the 352 amino acid isoform of tau) did not bind to fyn SH3 whereas tau residues 1-184 bound strongly (Fig. 2). This result suggested that the PXXP motif at residues 175-178 was the major fyn binding motif. However, tau sequences upstream of the PXXP motif also played a role in SH3 domain binding, as evidenced by the reduced SH3 binding of tau carboxy-terminal residues 173-352 (Fig. 2). To confirm that the PXXP motif at residues 175-178 was the major SH3 binding site, a tau mutant containing an internal deletion eliminating tau residues 170-178 was tested. As predicted, fyn SH3 binding to this mutant was reduced by over 90%, identifying the 175-178 PXXP motif as the major fyn SH3 binding site in tau. Similar experiments conducted with the src SH3 domain indicated that this PXXP sequence was also the src SH3 binding site (data not shown).

To examine the interaction between tau and fyn in a neuronal cell, we studied the human neuroblastoma cell line, SH-SY5Y, which expresses both tau and fyn (Pope et al., 1994; Meyerson and Pahlman, 1993; Tanaka et al., 1995; Smith et al., 1995). Tau in SH-SY5Y cells, as analyzed by RT-PCR, is predominantly the fetal 352 amino acids isoform with low level expression of the adult 383 amino acids isoform (Smith et al., 1995). The association of tau and fyn was examined by co-immunoprecipitation from SH-SY5Y fractions. Two fractions, a post-nuclear supernatant (PNS) and a nuclear extract of the nuclear pellet (NE), were prepared as described in Materials and Methods. While both tau and fyn were present in both fractions (Fig. 3A; tau left, fyn right), complexes of tau and fyn, as identified by co-immunoprecipitation by anti-fyn, were found only in the nuclear extract (Fig. 3B, lane 2). Since under our homogenization and buffer conditions, the nuclear pellet might also contain cellular components other than nuclei, the nuclear pellet was further fractionated by extracting the pellet with 0.5% Triton X-100. This procedure extracts lipids and detergent soluble proteins but does not extract nuclear tau (Greenwood and Johnson, 1995). We found complexes of tau and fyn in the Triton soluble extract (TSE) prepared from the nuclear pellet (Fig. 3B, lane 8). Stripping and re-probing the blot with anti-fyn confirmed the specific immunoprecipitation of fyn from the PNS and the TSE (Fig. 3B, lanes 10 and 12).

Among the cellular components that might be present in the nuclear pellet is the cell’s microfilament network. Moreover, Triton treatment of the pellet could release proteins associated with the microfilament network. Therefore, to test if tau-fyn complexes might be associated with the microfilament network, we treated SH-SY5Y cells with cytochalasin which disrupts the microfilament system. The cells were then fractionated similarly and co-immunoprecipitations were performed. We now found tau-fyn complexes in the PNS (Fig. 3C), suggesting that the tau-fyn complex is located in the actin network of SH-SY5Y cells. Since fyn is membrane associated, the complex may reside in the cell’s submembranous cytoskeleton.

Because the levels of endogenous tau and fyn are relatively low in undifferentiated SH-SY5Y cells, clear co-localization of tau and fyn could not be detected by immunofluorescence microscopy. However, when human tau and full length human fyn were co-expressed in NIH3T3 cells by transient transfection, the co-localization of tau and fyn could be detected in the periphery of doubly-transfected cells by immunofluorescence microscopy. In cells transfected with tau alone (Fig. 4A), tau was localized to cytoplasmic microtubules, as has been previously reported (Lee and Rook, 1992). In cells transfected with fyn alone (Fig. 4B), fyn was observed to localize to the plasma and internal membranes, consistent with the myristoylation-dependent association of fyn with cellular membranes (Silverman et al., 1993). (Neither protein was detectable in untransfected 3T3 cells.) Notably, NIH3T3 cells expressing either tau or fyn alone exhibited no significant alterations in the organization of their microtubule cytoskeletons. In contrast, cells expressing high levels of both fyn and tau often exhibited a rounded morphology uncharacteristic of NIH3T3 cells (Fig. 4C). Quantitation of rounded cells among those expressing both tau and fyn (>200 cells examined in 2 experiments) revealed that approximately 70% exhibited a rounded morphology, compared with only 4% of cells expressing fyn alone, or 13% of cells expressing tau alone.

The round morphology of cells expressing both fyn and tau made it difficult to examine the distributions of tau, fyn, and tubulin by conventional immunofluorescence microscopy. Therefore, we examined these cells by confocal microscopy. Optical sections of a 3T3 cell expressing fyn and tau revealed that fyn was associated with the plasma membrane and with numerous internal membranous vesicles (Fig. 5A). In the same cell, tau was observed in numerous punctate structures in the cortical and deeper cytoplasm (Fig. 5B). Superposition of the two images (Fig. 5C; fyn in green, tau in red) revealed considerable overlap in the distributions of tau and fyn (arrows in Fig. 5A-C). However, there were also clear examples of structures stained by anti-fyn (the plasma membrane and some vesicles) but not by anti-tau, and others stained by anti-tau but not by anti-fyn (arrowheads in Fig. 5A-C).

Confocal microscopy of cells expressing both fyn and tau, and stained with anti-tubulin and either anti-tau (Fig. 5D-F) or anti-fyn (Fig. 5G-H) revealed a dramatic reorganization of the interphase microtubule array. Individual microtubules and microtubule bundles were often observed to form a band associated with the cell cortex (shown in projection in Fig. 5D-F). When individual microtubules could be resolved, they were often observed to stain with anti-tau (arrows in Fig. 5D-F). However, examination of individual optical sections (not shown) revealed that a large fraction of the tau appeared to be dispersed throughout the cytoplasm, as in Fig. 5B, and not associated with microtubules (red in Fig. 5F). In some cases, cortical MTs formed a densely packed bundle or closed ring, just under the plasma membrane (Fig. 5G shows a single optical section; Fig. 5H shows a stereo reconstruction of the same cell).

Interaction of cellular proteins with the SH3 domains of src family kinases can serve to recruit these proteins to the catalytic domain of the enzyme, facilitating their tyrosine phosphorylation (Weng et al., 1995; Richard et al., 1995). To determine if tau was tyrosine phosphorylated in the presence of fyn, cDNAs encoding tau and fyn were transfected into COS cells. When expressed in COS cells, the 352 amino acid isoform of human tau exhibited 4-5 molecular species (50-60 kDa) that represent differential phosphorylation of tau on serines and/or threonines (Medina et al., 1995) (Fig. 6A, left). When expressed in the presence of fyn, the mobility of the various species of tau appeared unchanged, although the intensity of the fastest migrating species was lower (Fig. 6A, left). All tau species, expressed either in the presence or in the absence of fyn, bound to the GST-fyn SH3 fusion protein (Fig. 6A, right). To test for the tyrosine phosphorylation of tau, tau was isolated from transfected cell lysates by immunoprecipitation and subsequently tested for reactivity with an anti-phosphotyrosine antibody. Tau expressed in the presence of fyn was tyrosine phosphorylated whereas tau expressed in the absence of fyn was not (Fig. 6B). Similar results were obtained after tau and fyn co-transfection into 3T3 cells (data not shown). Thus, in the presence of fyn, tau was tyrosine phosphorylated.

To determine if tau in neuronal cells was tyrosine phosphorylated, we examined tau from SH-SY5Y cells, using two-dimensional gels electrophoresis to uniquely identify tau. Tau from SH-SY5Y cells appeared as a single species on one-dimensional gel electrophoresis (Fig. 3). However, using two-dimensional non-equilibrium pH gradient gel electrophoresis, two tau species, a major species (‘acidic tau’) and a minor species (‘basic tau’), were identified in SH-SY5Y cell lysates (Fig. 7A, panel 1). The major and minor tau species are likely to be the 352 and 383 amino acid isoforms of tau, respectively, as previously described (Smith et al., 1995). The major tau isoform, the 352 amino acid isoform, is more acidic than the corresponding E. coli synthesized tau (included as a control), indicating a highly phosphorylated state. The minor species is more basic than the 352 amino acid isoform, consistent with the amino acid composition of the 383 amino acid isoform. Moreover, the molecular mass of the minor species is comparable to that of the more acidic species, making it less likely to represent a less phosphorylated form of the 352 amino acid isoform (a less phosphorylated form would have had a downward shift in molecular mass, as shown by the non-phosphorylated E. coli synthesized tau relative to the major tau isoform in SH-SY5Y).

To further characterize SH-SY5Y tau, the tau species that bound to the GST-fyn SH3 fusion protein was similarly resolved using two-dimensional non-equilibrium pH gradient gel electrophoresis. Interestingly, the more basic of the two species (Fig. 7A, panel 2, ‘basic tau’) bound the fyn SH3 domain much more efficiently than the more acidic species. The identity of the SH3 binding tau species was verified by mixing the SH3 binding tau species with the whole cell lysate. The same two tau species were observed, but now each had a similar intensity (Fig. 7A, panel 3; longer exposures did not reveal any additional species). This confirmed that the SH3 binding tau species corresponded to the basic tau species in the SH-SY5Y lysate.

To determine if tau from SH-SY5Y cells was tyrosine phosphorylated, we examined tyrosine phosphorylated proteins from SH-SY5Y cells. Whole cell lysates were resolved by two-dimensional non-equilibrium pH gradient gel electrophoresis and probed with an anti-phosphotyrosine antibody. While Coomassie blue staining showed more than 50 distinct proteins on the gel, very few proteins were labeled by anti-phosphotyrosine and two of the labeled proteins had mobility characteristics identical to the major and minor tau species identified above (Fig. 7B, panel 1). The major species was confirmed as tau by stripping and re-probing the blot with anti-tau (Fig. 7B, panel 2). To confirm that the minor species (‘basic tau’) was tyrosine phosphorylated, the fyn SH3 binding tau species was similarly analyzed. The basic tau species was strongly labeled by the anti-phosphotyrosine antibody (Fig. 7B, panel 3). The identity of this species as tau was also confirmed by stripping and re-probing the blot with anti-tau (not shown). Thus, tau is constitutively tyrosine phosphorylated in a neuronal cell line suggesting that tyrosine phosphorylation of tau takes place in vivo. Examination of SH-SY5Y cells differentiated with TPA also revealed tyrosine phosphorylation of tau (data not shown). Lastly, while both tau species in SH-SY5Y cells are tyrosine phosphorylated, the fyn SH3 domain interacted most efficiently with the less abundant and more basic tau species.

We have identified a PXXP motif in the proline rich region of tau that interacts with the SH3 domain of the src-family non-receptor tyrosine kinases. The expression of the src-family kinase, fyn, with tau resulted in the tyrosine phosphorylation of tau. Consistent with their physical interaction, these two proteins co-localized when transfected into recipient cells and their co-expression was associated with greatly altered cell morphology. Tau from a neuronal cell line that naturally expresses fyn and src, was also tyrosine phosphorylated and co-immunoprecipitated with fyn. These data provide evidence for the in vivo interaction between tau and fyn and suggest that this interaction may allow signals transduced through fyn and other src-family non-receptor tyrosine kinases to alter the microtubule cytoskeleton in neuronal cells.

Our analyses suggest that factors exist in neuronal cells that regulate the interaction between tau and fyn. In our co-immunoprecipitation and two-dimensional gel results, we found that only a subpopulation of tau expressed in SH-SY5Y cells interacted with fyn. Among the factors that could affect tau’s ability to interact with the fyn SH3 domain are phosphorylation and alternative splicing. Phosphorylation may regulate the interaction between tau and fyn as suggested by our comparison of E. coli synthesized fetal tau (352 amino acid isoform; Fig. 1) and tau in the cell lysate of the SH-SY5Y human neuroblastoma cell line (Fig. 7). While E. coli synthesized tau bound to the fyn SH3 domain efficiently, the highly phosphorylated SH-SY5Y tau (also the 352 amino acid isoform) did not bind efficiently. Alternative splicing may also regulate the interaction between tau and fyn since the more efficient SH3 binding tau species in SH-SY5Y cells is likely to be an adult tau isoform of 383 amino acids (Smith et al., 1995) (Fig. 7, ‘basic tau’) that is generated by alternative splicing. Since alternative splicing can alter the phosphorylation of tau in vitro (Singh et al., 1997; Greenwood et al., 1994) and phosphorylation alters the conformation of tau (Hagestedt et al., 1989), the conformation of tau may determine the accessibility of the PXXP region to binding partners, thereby affecting the ability of tau to interact with the fyn SH3 domain.

Aside from conformational considerations, the ability of tau to interact with fyn may also be affected by phosphorylation at the PXXP motif. The SH3 binding PXXP motif in tau is located in the sequence, Arg-Thr-Pro-Pro-Lys-Ser-Pro-Ser-Ser. The threonine residue immediately preceding the PXXP (corresponding to thr231 in the human 441 amino acid tau isoform) and the serine residue in the PXXP (ser235) are known sites of in vivo phosphorylation for tau that are developmentally regulated in rat, with adult tau being less phosphorylated than fetal tau (Watanabe et al., 1993). SH-SY5Y tau has been reported to be phosphorylated at Thr231 (Tanaka et al., 1995). Thus, serine/threonine phosphorylation of tau not only affects tau’s ability to bind to microtubules (reviewed by Billingsley and Kincaid, 1997), but may also alter tau’s interaction with the SH3 domains of src-family tyrosine kinases.

The PXXP motif is located upstream of the microtubule binding repeats, bordering a region that enhances the interaction of tau with microtubules (Goode et al., 1997; Lee and Rook, 1992; Brandt and Lee, 1993). Moreover, when tau containing an internal deletion of the fyn interaction site (tauΔ(170-178)) was expressed in fibroblasts, free cytoplasmic tau was visible (unpublished data). Therefore, the binding of src-family non-receptor tyrosine kinases to this region of tau may affect tau’s interactions with microtubules. While we noted that tau-fyn double transfected cells had a substantial amount of tau not associated with microtubules, it is difficult to assess whether this result reflects a loss of tau’s microtubule binding activity or whether there is an indirect effect of fyn on tau’s interactions with microtubules.

The mechanisms that lead to the rounded cell morphology of the tau-fyn double transfected cells have yet to be elucidated. The distinctive microtubule structure in the rounded cells is reminiscent of the marginal band of erythrocytes, which is known to contain tau (Murphy and Wallis, 1985; Sanchez and Cohen, 1994). Tau is thought to promote microtubule assembly and to bundle microtubules in the marginal band. The marginal band in erythrocytes is also thought to associate with the plasma membrane (Murphy et al., 1986). While bundled microtubules organized in circles have been observed in flat cells over-expressing either MAP2 or tau (Weisshaar et al., 1992, and our unpublished data), the co-expression of fyn greatly increases the percentage of cells containing such structures and also correlates with a rounding of the cell. The presence of fyn may link tau to the membrane, facilitating the formation of microtubule bundles adjacent to the membrane that then coalesce into one band, leading to the rounding of cells. Consistent with this hypothesis, it has been suggested that the characteristic morphology of erythrocytes is driven by the formation of the marginal band (Winckler and Solomon, 1991).

Complexes of tau and fyn may be associated with the actin cytoskeleton in SH-SY5Y cells, since cytochalasin treatment of these cells caused the complexes to shift from an insoluble to a soluble state. The association of tau with the actin cytoskeleton has been suggested by in vitro studies that show the binding of tau to actin in a phosphorylation dependent manner (Griffith and Pollard, 1982; Selden and Pollard, 1983). In addition, in vivo associations between tau and microfilaments have been described (Cross et al., 1993; DiTella et al., 1994; Kempf et al., 1996). The association of fyn with the actin cytoskeleton has been suggested by the location of both fyn and c-src to focal adhesions, structures that connect actin stress fibers to the plasma membrane. Focal adhesions mediate signals from the extracellular matrix that affect cell adhesion, cell morphology, and cell migration (reviewed by Burridge and Chrzanowskawodnicka, 1996). Moreover, the ability of c-src to translocate to the actin cytoskeleton and to alter the actin cytoskeleton following receptor stimulation is well documented (reviewed by Thomas and Brugge, 1997).

Our immunofluorescence data from transfected cells is consistent with an association between tau-fyn complexes and the actin cytoskeleton. Co-localization between tau and fyn occurs just beneath the plasma membrane, where actin filaments are localized to form the cell’s actin cortex. Furthermore, the presence of tau and fyn led to a change in cell shape, suggesting that their interaction resulted in alterations in the actin cortex. Our data support a role for tau in src-family non-receptor tyrosine kinase signalling, with the site of action being the submembranous actin cytoskeleton.

During neuronal differentiation, the src and fyn signalling pathways are involved in neurite growth through the involvement of the cell adhesion molecules, L1 and NCAM, respectively (Beggs et al., 1994; Ignelzi et al., 1994). Moreover, src-family non-receptor tyrosine kinases and tyrosine phosphorylation have been found in the growth cone (Bixby and Jhabvala, 1993; Helmke and Pfenninger, 1995; Wu and Goldberg, 1993) and associations between the src-family non-receptor tyrosine kinases and the growth cone cytoskeleton have been shown (Helmke and Pfenninger, 1995). We and others have observed a subpopulation of tau that is enriched in the distal portion of the axon and growth cone (Mandell and Banker, 1995, 1996b; Kempf et al., 1996; Black et al., 1996). This tau subpopulation had a distinct phosphorylation state relative to tau in the remainder of the cell (Mandell and Banker, 1995, 1996b). The specific localization of this tau was altered by microtubule and microfilament inhibitors (Kempf et al., 1996). Interestingly, this localization was also disrupted by a tyrosine phosphatase inhibitor, suggesting a role for tyrosine phosphorylation in this localization (Mandell and Banker, 1996b). We speculate that fyn and tau interact in the growth cone in a phosphorylation dependent manner, resulting in a distinct phosphorylation state for tau in the growth cone. Furthermore, this subpopulation of tau is membrane associated through its interaction with the SH3 domain of src-family non-receptor tyrosine kinases.

Tau’s unique role in axonal development may be related to its interaction with fyn. In response to extracellular signals that promote axonal growth (Lochter et al., 1994; Lochter and Schachner, 1993), a complex of activated fyn and tau in the actin rich growth cone might alter the local actin environment to facilitate invasion by dynamic microtubules. Tau may then capture microtubules, alter microtubule dynamics, and/or organize microtubules in the growth cone, thereby promoting the forward advance of the growth cone (Gordon-Weeks, 1997). As a role for tyrosine kinases and phosphatases in axonogenesis has been anticipated (reviewed by Mandell and Banker, 1996a; Maness et al., 1996; Bixby and Bookman, 1996; Walsh and Doherty, 1997; Desai et al., 1997), our data provides a possible link between the cytoskeletal events and upstream signalling pathways.

Recently, cells from fyn−/− mice were shown to be defective in completing cell division (Yasunaga et al., 1996), suggesting that fyn is activated during mitosis. Moreover, tyrosine phosphorylation is involved in regulating cell growth and development (reviewed by Hunter, 1996). Therefore, if a cell responding to a neurodegenerative influence mounts a regenerative or proliferative response, this might include a mitogenic response whereby tyrosine phosphorylation is used to initiate cell proliferation and recapitulate development. Supporting this hypothesis, several mitotic cell epitopes have been identified in the paired helical filaments of Alzheimer’s disease (AD), suggesting that degenerating cells up-regulate mitotic kinase activities normally absent in post-mitotic neurons (Vincent et al., 1996; Kondratick and Vandre, 1996). Given that fyn levels are increased in some neurons in AD (Shirazi and Wood, 1993), we speculate that fyn may be activated during neurodegeneration, leading to phosphorylation events characteristic of mitosis. Furthermore, exposure of neuronal cells to the Aβ peptide, a key component in the senile plaques in AD, upregulates tyrosine phosphorylation of several serine-threonine kinases (Zhang et al., 1994; Luo et al., 1996; Sato et al., 1997), supporting the hypothesis that Aβ activates tyrosine kinases (Luo et al., 1995). Since tyrosine phosphorylation activates serine-threonine kinases, these events could lead to the abnormal phosphorylation of tau and the formation of paired helical filaments.

In summary, our data demonstrate that src-family non-receptor tyrosine kinases such as fyn can interact with the microtubule associated protein tau. This association now directly links signal transduction pathways in neuronal cells to the microtubule cytoskeleton and suggests a mechanism for coupling extracellular signals to the cytoskeletal system. Tau and other MAPs may function not only to stabilize microtubule arrays, but also to direct cytoskeletal changes in response to extracellular signals.

We gratefully acknowledge Dr Craig Morita for valuable suggestions and discussions. We thank Drs Virginia M.-Y. Lee, Gail V. W. Johnson, and Lester I. Binder for generously providing antibodies tau46.1, 5A6, tau5, and tau1. We also thank Dr Gail V. W. Johnson for providing SH-SY5Y cells to us. Drs Garth Hall, Roland Brandt, and Athena Andreadis provided critical comments on the manuscript. G.L. was supported by grants from NIH (NS32100) and NSF (IBN9222446). G.P. was supported by an NIH institutional training grant. D.L.G. was supported by grant MCB9506051 from the NSF.