Tau Proteins: The Molecular Structure and Mode of Binding on Microtubules Nobutaka Hirokawa, Yoko Shiomura, and Shigeo Okabe Department of Anatomy and Cell Biology, Medical School, University of Tokyo, Hongo, Tokyo, 113 Japan Abstract. Tau is a family of closely related proteins (55,000-62,000 mol wt) which are contained in the nerve cells and copolymerize with tubulin to induce the formation of microtubules in vitro. All information so far has indicated that tau is closely apposed to the microtubule lattice, and there was no indication of do- mains projecting from the microtubule polymer lattice. We have studied the molecular structure of the tau fac- tor and its mode of binding on microtubules using the quick-freeze, deep-etch method (QF-DE) and low an- gle rotary shadowing technique. Phosphocellulose column-purified tubulin from porcine brain was poly- mefized with tau and the centrifuged pellets were processed by QF. DE. We observed periodic armlike elements (18.7 + 4.8 nm long) projecting from the microtubule surface. Most of the projections appeared to cross-link adjacent microtubules. We measured the longitudinal periodicity of tau projections on the microtubules and found it to match the 6-dimer pattern better than the 12-dimer pattern. The stoichiometry of tau versus tubulin in preparations of tau saturated microtubules was 1:,o5.0 (molar ratio). Tau molecules adsorbed on mica took on rodlike forms (56.1 5:14.1 nm long). Although both tau and MAP1 are contained in axons, competitive binding studies demonstrated that the binding sites of tau and MAP1A on the microtubule surfaces are mostly distinct, although they may partially overlap. M ICROTUBULES are one of the main cytoskeletal ele- ments in eukaryotic cells and are particularly abundant in nerve cells. It is well known that there are several microtubule-associated proteins (MAPs) ~ which copurify with brain tubulin during repetitive cycles of temperature-dependent assembly and disassembly. Among these proteins in neuronal tissues, high molecular weight proteins (MAP1 and MAP2), and tau factor are major spe- cies, and recently several minor proteins have also been identified (6, 7, 17, 27, 28, 33, 41). The high molecular weight microtubule-associated pro- teins (MAP1, MAP2) are flexible, rodlike structures "o100- 200 nm long and form armlike projections when attached to microtubule surfaces (22, 31, 39, 40, 42). Recent structural studies demonstrated that microtubule domains in nerve cells are composed of microtubules, associated cross bridges, and granular materials (11, 14, 31). High molecular weight MAPs (MAP1, MAP2, 270,000-mol-wt MAP) were proven to be components of these cross bridges associated with microtubules in vivo (12, 14, 32). Tau factor is composed of four to five polypeptides (55,000- 62,000 mol wt) which were shown to be closely related by both peptide mapping and amino acid analysis (6), and were assumed to represent a highly asymmetric molecule by hy- drodynamic data (7). This protein promotes polymerization 1. Abbreviations used in this paper: MAP(s), microtubule-associated pro- tein(s); PC, phosphocellulose column purified; PEM, 0.1 M Pipes, 1 mM EGTA, 1 mM MgCI2, pH 6.8. of tubulin, is heat stable (41), and is able to bind to calmodu- lin in the presence of calcium (34). Recent immunocyto- chemical studies have shown that tau is mainly localized in axons (3). It has also been revealed recently that highly phos- phorylated tau is a major element of paired helical filaments in Alzheimer's disease (8, 18, 23). However, the molecular structure of tau and the mode of binding of tau to microtubules has remained unclear. This study was designed to disclose the molecular structure oftau by the quick-freeze, deep-etch method and low angle rotary shadowing technique, and has allowed us to reveal for the first time that tau is a rodlike structure (56.1 -6 14.1 nm long) and associates with microtubules with armlike projections (18.7 ± 4.8 nm long). Materials and Methods Isolation of Tau Tau was prepared from porcine brain microtubules by the modification of a method described by Grundke-Iqbal et al. (8). Pellets of microtubule pro- teins obtained by cycles of temperature-dependent assembly and disassem- bly (30) were suspended in 3 vol of buffer containing 100 mM 2-(N-mor- pholino)ethane sulfonic acid M~, 0.5 mM MgCI2, 1 mM EGTA, 0.1 mM EDTA, 0.75 M NaC1, 2 mM dithiothreitol, and 0.1 mM phenylmethylsul- fonyl fluoride (PMSF), pH 2.7. The suspension was then heated at 95°C in boiling water for 5 min and centrifuged at 25,000 g at 4°C for 30 rain. The pH of the resulting supernatant was adjusted to 6.8, and it was then subjected to Bio gel A 1.5 m (Bio-Rad Laboratories, Richmond, CA) column chroma- tography and concentrated in order to prepare purified tau fractions (1.89 © The Rockefeller University Press, 0021-9525/88/10/1449/11 $2.00 The Journal of Cell Biology, Volume 107, October 1988 1449-1459 1449
`
`Exhibit 2067
`Page 01 of 11
`
`

`

`Figure 1. SDS-PAGE (7.5% running gel) of PC tubulin (lane 1 ), purified tau fac- tor (lane 2), supernatant (lane 3), and pellet (lane 4) of suspensions of PC tubulin (1.0 mg/ml) and tau (0.4 mg/ml). The pellet was resuspended with PEM in a volume equal to the original sus- pension, and the same volume of the sample was applied (lanes 3 and 4) on the gel. mg/ml). Tau proteins were dialysed against PEM buffer (0.1 M Pipes, 1 mM EGTA, 1 mM MgCI2, pH 6.8). Isolation of Tubulin Tubulin was prepared from porcine brain by phosphocellulose chromatogra- phy (41). Protein Determination Protein concentrations were estimated by the assays of Bradford et al. (5) and Itzhaki and Gill (19) using BSA as a standard. Quick-Freeze, Deep-Etch Electron Microscopy of Tubulins Polymerized with Tau Phosphocellulose column-purified (PC) tubulin was mixed with tau fraction (final concentration: tau 0.4 mg/ml, tubulin 1.0 mg/ml) and incubated at 37°C in the presence of 1 mM GTP for 30 min. As a control, PC tubulin without tau was incubated at 37°C in the presence of 20 ~tM taxol and 1 mM GTP (38). The resulting polymers were centrifuged at 19,000 rpm at 30°C for 30 min, and pellets were quick frozen and deep etched as described previously (9-11, 14). In some cases pellets were resuspended with a small amount of PEM buffer (0.1 M Pipes, 1 mM EGTA, 1 mM MgC12, pH 6.8) (four times the pellet volume), chilled on ice for 30 min, supplemented with taxol to 20 ~tM, and rewarmed at 37°C for 15 min. The suspension was quick frozen without centrifugation. The samples were dissolved in Purelox (Ohylox Corp., Tokyo, Japan), and replicas were washed with distilled wa- ter, and put on Formvar-carbon-coated grids. Quick-Freeze, Deep-Etch Electron Microscopy of Axons Sciatic nerves were dissected out of rats and incubated for 30 min at room temperature in 0.1% saponin, 70 mM KCI, 5 mM MgCl2, 3 mM EGTA, 30 mM Hepes, pH 7.4, 10 ~tM taxol, and 0.1 mM phenylmethylsulfonyl fluoride as described previously (14). They were quick frozen as described previously (9-11, 14). Low Angle Rotary Shadowing of Tau Molecules Tau fractions were suspended in glycerol (50% glycerol PEM) at 50 ~g/ml protein concentration and sprayed on freshly cleaved mica flakes as de- scribed by q~ylor and Branton (36) and dried by vacuum evaporation. The samples were rotary shadowed with platinum in a freeze-fracture machine (model 301; Balzers, Hudson, NH) at an angle of 6 °. The replicas were detached from mica with hydrofluoric acid, washed with distilled water, and collected on Formvar-carbon-coated grids. Electron Microscopy and Measurement Replicas were examined with a JEOL 2000EX electron microscope at 100 kV and photographed. Micrographs were printed with their contrast reversed. The length of tau projections in the tau-saturated microtubule pellets and the length oftau molecules on mica were measured under a mag- nifying glass. The center to center distances between adjacent arms in replicas of tau-saturated microtubules were measured in the same way. Stoichiometry of Tubulin and Tau PC tubulin was mixed with an excess amount of tau fraction and incubated at 37°C in the presence of 1 mM GTP with or without 20 IxM taxol for 30 min. As a control, PC tubulin without tau was incubated at 37°C in the pres- ence of 20 I.tM taxol and 1 mM GTP. The suspensions were centrifuged at 19,000 rpm at 30°C for 30 min. The resulting pellets were resuspended in PEM, chilled thoroughly on ice, and homogenized thoroughly. Protein con- centration of the pellets and supernatants was determined by the assays of Bradford et al. (5) and Itzhaki and Gill (19). SDS-PAGE of the pellets and supernatants was performed according to the method of Laemmli using 7.5 % acrylamide (25). Gels were stained with Coomassie Brilliant Blue, scanned, and the areas of peaks were measured by densitometry (model CS 9000; Shimadzu Corp., Kyoto, Japan). Tau-MAP1 and MAP1-MAP2 Displacement Experiments MAP1A was purified by affinity chromatography on a cyanogen bromide- activated Sepharose 4B column containing our monoclonal antibody against MAP1A as described previously (31). After the application of crude extracts of rat brain, the column was washed with PEM containing 0.75 M NaCI and 0.5% NP-40. Bound polypeptides were eluted with 3 M MgCI2 in 0.1 M PEM buffer, and peak fractions were dialyzed against 20 mM PEM buffer, and subjected to PAGE to check their purity. PC tubulin was mixed with MAP1A at a weight ratio of 1:2.4 in the pres- ence of 20 ~tM taxol and 1 mM GTP. Suspensions were incubated at 37°C for 30 min and were then layered onto 20% sucrose cushions containing tau at a weight ratio of 1 tubulin (in suspension) per 2 tau (in the cushion) or at a ratio of 1:0.2, 20 gM taxol and 1 mM GTP. Microtubules were cen- trifuged through the sucrose cushions at 30,000 g for 1 h at 30°C. Sucrose cushions lacking tau proteins were used as control. In reciprocal experiments, PC tubulin was mixed with tau at a weight ra- tio of 1:0.4 in the presence of 20 gM taxol and 1 mM GTP. After incubation at 37°C for 30 min, the suspensions were centrifuged through 20% sucrose cushions containing MAPIA at a weight ratio of 1 tubulin (in suspension) per 2 MAP1A (in the cushion) or at a ratio of 1:1, 20 gM taxol, and 1 mM GTP. In these cases, the sucrose cushions of control samples contained no MAP1A. In addition PC tubulin was mixed with tau and MAPIA at weight ratios of 1:1:1.2 in the presence of 20 ~tM taxol and 1 mM GTP. PC tubulin plus tau (1:1), PC tubulin plus MAPIA (1:1.2), and PC tubulin alone were also examined. After incubation at 37"C for 30 min, the suspensions were cen- trifuged through 20% sucrose cushions containing 20 I~M taxol and 1 mM GTP. After their surfaces were washed with PEM, the pellets were resuspended in PEM. SDS-PAGE of the pellets and supernatants was per- formed according to the method of Laemmli using 7.5% acrylamide gels (25). Gels were stained with Coomassie Brilliant Blue, scanned with a den- sitometer (model CS 9000; Shimadzu Corp.), and the areas of peaks were measured. MAP1 and MAP2 displacement experiments were performed similarly using MAP2 instead of tau proteins. MAP'2 was purified from rat brains using a Superose 6 prepgrade gel filtration column (Pharmacia Fine Chemi- cals, Uppsala, Sweden) as described previously (15). PC tubulin was mixed with MAPLA at a weight ratio of 1:2.4 in the presence of 20 IXM taxol and 1 mM GTP. Suspensions were incubated at 37°C for 30 min and then layered onto 20% sucrose cushions containing MAP2 at a weight ratio of 1 tubulin (in suspension) per 2.4 MAP2 (in the cushion), 20 gM taxol and I mM GTP. Microtubules were centrifuged through the sucrose cushions at 30,000 g for 1 h at 30°C. Sucrose cushions without MAP2 were used as controls. Reciprocal experiments were also carded out. In additional experiments PC tubulin was mixed with MAP1A and MAP2 at a weight ratio of 1:2:2 in the presence of 20 p.M taxol and 1 mM GTP. Tubulin plus MAP1A (1:2), tubulin plus MAP2 (1:2), and tubulin alone were also examined. After incubation at 37°C for 30 min the suspensions were centrifuged through 20% sucrose cushions containing 20 I.tM taxol and 1 mM GTP. Microtubules were cen- The Journal of Cell Biology, Volume 107, 1988 1450
`
`Exhibit 2067
`Page 02 of 11
`
`

`

`Figure 2. Higher magnification views of quick-frozen, deep-etched microtubule pellets polymerized with (A and B) and without tau (C). Although only tightly packed microtubules are observed in C, numerous projections (arrows) exist on the microtubules polymerized with tau (A and B). Bar, 100 nm. Hirokawa et al. Molecular Structure of Tau 1451
`
`Exhibit 2067
`Page 03 of 11
`
`

`

`Z I00 CO m 5o 10" R S.I I lEO m 10.1 15.1 20.1 I I I 15.0 20,0 25.6 LENGTH I---1 253 30.1 35.1 I I I 30.0 35.0 40.0 ntn Figure 3. A histogram showing the length-frequency distribution of tau projections on microtubules, n = 292. trifuged at 30,000 g for 1 h at 30°C. SDS-PAGE of the pellets and superna- tants was performed. Results Fig. 1 shows SDS-PAGE analysis of PC tubulin (lane 1), purified tau factor (lane 2), the supernatant (lane 3), and pel- let (lane 4) of a suspension of PC tubulin plus tau incubated at 37°C for 30 min in the presence of 1 mM GTE The tau factor from porcine brain prepared by heat treatment at a low pH was composed of five bands (molecular weight of 50,000-65,000) (lane 2). Boiling at a low pH removed the high molecular weight MAPs while tau remained soluble (8). The presence of tau promoted polymerization of PC tubulin, and thus, in the pellet we found tubulin plus tau. Tau Forms Armlike Projections on Microtubule Surfaces Fig. 2 displays high magnification views of the tau-micro- tubule pellet processed by the quick-freeze, deep-etch method. As can be seen, the pellet contained numerous par- allel microtubules, each with numerous armlike projections (<20 nm long) attached to the microtubule surfaces. The projections were short and straight and appeared to cross- link adjacent microtubules. When PC tubulin was polymer- ized with tau in the presence of 10 llM taxol and 1 mM GTP, we observed similar projections on the microtubules. Be- cause samples which contained only PC tubulin exhibited bare surfaces, (Fig. 2 C), we could conclude that tau forms straight, short, armlike projections on the surfaces of microtubules. Fig. 3 shows a histogram of the length-frequency distribu- tion of tau projections on microtubules. The average length was 18.7 + 4.8 (SD) nm. The projections appeared to cross-link adjacent microtu- bules. To check the possibility that centrifugation induced the patterns we observed, we resuspended the tau-saturated microtubules in a small volume of PEM, incubated them on ice for 30 min, and after homogenization rewarmed the sus- pension in the presence of 20 pM taxol and 1 mM GTP. The suspension was then quick frozen without centrifugation. In such preparations most of the microtubules tended to run Figure 4. Quick-frozen, deep-etched suspension of microtubules saturated with tau proteins. Note the frequent cross bridges (arrows) be- tween the microtubules. Bar, 100 nm. The Journal of Cell Biology, Volume 107, 1988 1452
`
`Exhibit 2067
`Page 04 of 11
`
`

`

`Figure 5. A stereopair of an axoplasm in a rat sciatic nerve quick-frozen and deep-etched after saponin extraction. Frequent short cross bridges (short arrows) are found between microtu- bules. Longer strands are also asso- ciated with microtubules (long ar- rows). Bar, 100 nm. randomly and granular structures were attached on the microtubule surfaces; however, when microtubules ran par- allel and close to each other we found frequent cross-links between adjacent microtubules as shown in Fig. 4. These results suggest that tau proteins cross-link microtubules when microtubules are in very close proximity to each other. Furthermore, we examined microtubule domains in the axons of rat peripheral nerves after saponin extraction. As shown in a stereopair in Fig. 5, microtubules are linked with each other via fine short cross bridges exactly like those found in the microtubules saturated with tau in vitro (See Fig. 4 in reference 11). Longer strands tending to form net- works were also associated with microtubules (Fig. 5). Longitudinal Periodicity of Tau Projections We measured distances between adjacent projections on in- dividual microtubules. Fig. 6 is a histogram of longitudinal spacing of adjacent arms on the same microtubules. The dots and arrows indicate spacings predicted by a 12-dimer super- lattice model and a 6-dimer superlattice model, respectively (1, 20). Fig. 7 shows a schematic diagram of the 6-dimer su- perlattice model. The numbers 11, 15, 22, 27, 33, 37, and 48 on Fig. 7 indicate predicted spacings by a 6-dimer superlat- tice model. As shown in Fig. 6, the longitudinal spacings ob- served in tau-saturated microtubule pellets match the 6-di- mer pattern better than the 12-dimer pattern. The actual data fit with spacings of 11, 15, 22, 26, 33, and 48 nm, while there Hirokawa et al. Molecular Structure of Tau 1453
`
`Exhibit 2067
`Page 05 of 11
`
`

`

`150- • W. ~ 4P 9P _l- -1- ~100- o D. o "~50. .r 10- 0 10 2"0 3"0 41) 51) 6! nm Longitudinal spacing Figure 6. A histogram showing the frequency of longitudinal spac- ings between adjacent arms on the same microtubules saturated with tan. Arrows and dots indicate spacings predicted by the 6-dimer superlattice and 12-dimer supperlattice models, respectively, n = 1,605. were additional peaks at 19, 30, 38, 42, and 46 nm. There- fore, the pattern of longitudinal spacing resembles, but is not identical to, the sites defined by the 6-dimer superlattice. Stoichiometry of Tubulin and Tau PC tubulin was mixed with a saturating amount of tau pro- teins. After incubation at 37°C for 1 h in the presence of 1 mM GTP with or without 20 I~M taxol, the resulting pellet A and supernatant were subjected to SDS-PAGE. We estimated the amount of proteins in the pellets and supernatants by the assay of Bradford (5). Peak areas corresponding to tubulin and tau proteins were measured by densitometry. As shown in Fig. 1 and 8 A, tau proteins from rat brains were composed of five distinct bands. Because in the microtubule pellets saturated by tau the third, fourth, and fifth bands overlap the tubulin bands, we estimated the area of the third plus fourth plus fifth peaks relative to the area of the first plus second peaks. As a result the ratio of the area of the first plus second peaks to the third plus fourth plus fifth peaks was 1:2.2. Since this ratio did not change in the supernatants of tau-saturated microtubules polymerized in the presence of taxol, we esti- mated the binding affinities of each five bands to tubulin to B be equal. Thus, we estimated the total amount of tau proteins in the tau-saturated microtubule pellets based on these assump- 48 Figure 7. A schematic drawing of a 6-dimer model for the arrange- ment of tan projections. The opened-out lattice as viewed from the outside. The circles represent tubulin monomers. The squares indi- cate the binding sites of tan. The numbers indicate predicted spac- ings (in nanometers) between binding sites. tions; the total area oftau peaks are 3.2 times the area of the first plus second peaks. Accordingly, the area of tubulin peaks was estimated as the third peak plus a large tubulin peak -2.2 times (first plus second peaks of tau proteins). The molar ratio of tau/tubulin ranged from 1:3 to 1:6, using an average molecular weight of 60,000 for tau. The average was l:,x,5. An another approach, we calculated the molar ratio of tau to tubulin from the amount of proteins in the pellets. The PC tubulin was polymerized with or without tau in the presence of taxol and GTP. The amount of tubulin in the tau-saturated microtubules was estimated from that in the pellet containing only tubulin. From this approach the molar ratio of tau to tubulin was also found to be 1:(cid:127)5. J T Figure 8. Densitometric scans of SDS gels of tau proteins (A) and microtubules saturated with tau proteins (B) equivalent to the gels shown in Fig. 1. The Journal of Cell Biology, Volume 107, 1988 1454
`
`Exhibit 2067
`Page 06 of 11
`
`

`

`Figure 9. The SDS gel of a tau and MAP1A displacement experiment. Lane 1, Supema- tant (S) and pellet (P) of a suspension con- mining tau and PC tubulin after centrifuga- tion through 20% sucrose containing 20 laM taxol and 1 mM GTP only. Lane 2, su- pernatant (S) and pellet (P) of a suspension containing tau and PC tubulin after cen- trifugation through 20% sucrose containing 20 ~tM taxol, 1 mM GTP, and MAP1A. Lane 3, supernatant (S) and pellet (P) of a suspension containing tau and PC tubulin after centrifugation through 20% sucrose containing 20 I.tM taxol, 1 mM GTP, and MAP1A (concentration of MAP1A is half of that in the 20% sucrose cushion of lane 2). Lane 4, MAP1A purified by affinity chro- matography on a cyanogen bromide-acti- vated Sepharose 4B column using our mono- clonal antibody against MAPIA. Lane 5, supernatant (S) and pellet (P) of a suspension containing MAPIA and PC tubulin after centrifugation through 20% sucrose containing 20 lt M taxol and I mM GTP only. Lane 6, supernatant (S) and pellet (P) of a suspension containing MAP1A and PC tubulin after centrifuga- tion through 20% sucrose containing 20 IxM taxol, 1 mM GTE and tau. Lane 7, experiment similar to lane 6, except that the concentration of tau in the 20% sucrose cushion was one-tenth of that in the cushion of lane 6. Lane 8, microtubule proteins from rat brain. Binding Sites of Tau and MAP1A on Microtubules Both tau and MAP1 have been found to be localized in the axons of neurons in brain (3, 4, 14, 17). Concerning the binding sites of tau proteins on microtubules, it has been shown that tau and MAP2 bind competitively to the same sites (21, 29). Therefore, we attempted to analyze whether tau and MAP1 bind to microtubules competitively or not. MAP1A was purified by an affinity column using a monoclo- nal antibody against rat brain MAP1A. We performed tau-MAP1 displacement experiments according to a proce- dure used by Kim et al. (21). First, tubulin was polymerized with an excess amount of the first MAP (tau or MAP1A). Then the suspension was centrifuged through 20% sucrose cushions containing a large amount of the second MAP (MAP1A or tau). In control experiments tubulin was poly- merized with the first MAP and was centrifuged through 20 % sucrose cushions containing no second MAP. As shown in Figs. 9 and 10, the amount of the first MAP (tau or MAP1A) relative to tubulin in the pellets was not significantly changed regardless of the presence or absence of the second MAP in the cushions. If the amount of the second MAP in the cushions was decreased, a smaller amount of the second MAP bound to tubulin, but the ratio of tubulin to the first MAP remained unchanged. We carried out other experiments as well. Tubulin was polymerized with excess amounts oftau, MAP1A, or tau plus MAP1A. Then the suspensions were centrifuged through 20% sucrose cushions. Fig. 11 shows SDS-PAGE of superna- tants and pellets of these experiments. After scanning with a densitometer, we found that the amounts of tau relative to tubulin in the pellets of tau plus tubulin and tau plus MAP1A plus tubulin had not significantly changed. However, the amount of MAP1A in the pellet of tubulin plus tau plus MAP1A was '~81% of that in the pellet of MAP1A plus tubu- lin. From these experiments, then, we could conclude that the binding sites of tau and MAP1A on the microtubule sur- faces are mostly distinct, but may be partially overlapping. A B Figure 10. Densitometric scans of SDS gels of a tau and MAP1A displacement experiment. (A) Pellet of microtubules saturated with MAP1A equivalent to lane 5P in Fig. 9. (B) Pellet of microtubules saturated with MAPIA after centrifugation through 20% sucrose containing 20 laM taxol, 1 mM GTP, and an excess amount of tau equivalent to lane 6P in Fig. 9. The ratio of MAP1A to tubulin does not change after tau binds to microtubules. Hirokawa et al. Molecular Structure of Tau 1455
`
`Exhibit 2067
`Page 07 of 11
`
`

`

`Figure 11. The SDS gel of a competition-experiment of tau and MAPIA. Lane 1, pellet (P) and supernatant (S) of microtubules saturated with tan and MAP1A. Lane 2, pellet (P) and supernatant (S) of microtubules saturated with tau. Lane 3, pellet (P) and su- pernatant (S) of microtubules saturated with MAP1A. Figure 12. The SDS gel of a MAP1A and MAP2 displace- ment experiment. Lanes I and 2, supernatant (1) and pellet (2) of micmtubules saturated with MAP2. Lanes 3 and 4, supernatant (3) and pellet (4) of a suspension containing MAP1A and PC tubulin after centrifugation through 20% containing 20 p,M taxol and 1 mM GTP only. Lanes 5 and 6, supernatant (5) and pellet (6) of a suspension containing MAP1A and PC tubulin after centrifugation through 20% sucrose containing 20 p.M taxol, I mM GTP, and MAP2. The ratio between MAP1A and tubulin is not significantly changed after centrifugation through an excess amount of MAP2. Upper and lower ar- rows in lane 6 indicate MAP1A and MAP2, respectively. In these experiments we also estimated the molar ratio of MAP1A to tubulin in the MAP1A saturated microtubules. The molar ratio was 1:12 (MAP1A/tubulin), In addition, we also performed MAP1A and MAP2 displacement experi- ments. As shown in Fig. 12, the ratio of tubulin to MAP1A was not changed regardless of the presence or absence of MAP2 in the cushion. When tubulin was polymerized with MAP1A alone, MAP2 alone or MAP1A plus MAP2, amounts of MAP1A and MAP2 relative to tubulin were not signifi- cantly changed. This indicates that MAP1A and MAP2 prob- ably have distinct binding sites on microtubules. Tau Is a Short Rodlike Molecule Fig. 13 demonstrates a gallery oftau molecules adsorbed on mica. Low angle rotary shadowing revealed that tau (~50 nm long) is a rodlike structure much shorter than the high molecular weight MAPs (100-200 nm long). We measured the length of tau molecules adsorbed on mica (Fig. 14) and found the average length to be 56.1 + 14.1 (SD) nm. Discussion Molecular Structure of Tau The present study for the first time demonstrates that tau pro- teins are rodlike molecules ('~50 nm long) and bind to mi- crotubules as periodic, short, armlike projections (<20 nm long). Previous examination by thin section, and negative staining methods and metal shadowing did not reveal their features (35, 42). This is probably because tau forms very short arms on the microtubules when microtubules are close to each other, making their visualization very difficult by conventional methods. When high molecular weight MAPs such as MAP2 or 270,000-mol-wt MAP were incubated with PC tubulin and processed similarly, long projections (longer than 30 nm) were observed on the microtubule surface (12). They some- times took on the appearance of anastomosing cross bridges between microtubules, while tau always appeared as a straight, short, rodlike component and never formed anastomosing networks. Since tau arms on the microtubules were shorter than 20 nm, certain parts of the molecule could comprise the binding domain associated with the microtubule surface. It is well known that limited chymotryptic digestion of MAP2 pro- duced a fragment with a molecular weight of 35,000 which promotes microtubule assembly and corresponds to the mi- crotubule-binding domain of MAP2 (37). Recently we found that the 190,000-mol-wt MAP from adrenal medulla is also a long, rodlike protein, and forms armlike projections on microtubules (26). A 27,000-mol-wt fragment of the 190,000- mol-wt MAP produced by limited chymotryptic digestion was also observed to stimulate tubulin polymerization and to correspond to the binding domain that associates with micro- tubules (1). Because both tau and MAP2 promote tubulin polymerization, are rodlike molecules, and bind competi- tively to microtubules, their binding domains may have a common nature. This binding domain of tau may play an ac- tive role when tau promotes the polymerization of tubulin. Concerning the molecular structure of MAPs, most of the high molecular weight MAPs, including MAP1A and B, MAP2, 2"/0,000-mol-wt MAP, and 190,000-mol-wt MAP ex- hibit long rodlike forms (12, 26, 31, 40). However, recently we found that one of the main MAPs in the mitotic spindle of sea urchin eggs, which has a molecular weight (75,000) close to tau, is a globular molecule, and covers the microtu- bule surface as round buttonlike structures (13, 16). We ini- tially thought that tau might be a similar molecule to this 75,000-mol-wt MAP (buttonin), but buttonin was not heat stable and did not bind to calmodulin in the presence of cal- cium (13). So buttonin was determined as expressing charac- teristics distinct from tau. In addition, the present study clearly shows that tau possesses quite a different molecular shape from buttonin (rodlike vs. spherical). So far, microtu- The Journal of Cell Biology, Volume 107, 1988 1456
`
`Exhibit 2067
`Page 08 of 11
`
`

`

`Figure 13. A gallery of tau mole- cules adsorbed on mica revealed by the low angle rotary shadow- ing technique. Bar, 100 nm. bule-associated proteins of neuronal origin mostly tend to take on rodlike forms. This fact is in agreement with the cytoskeletal structure in the nerve cells in vivo. With the quick-freeze deep-etch technique we found various kinds of strands associated with microtubules in nerve cells (11, 14, 31, 32), and proved that MAPIA and B, and MAP2 are com- ponents of these strands (14, 15, 31, 32). When the microtu- bules run close together in the axon, the distance between them is 15-20 nm. This spacing is similar to the length of tau molecules bridging adjacent microtubules. In fact, we found short straight cross bridges between microtubules in the axons exactly like those found in the tau microtubule pellets. So it is very likely that tau forms short cross bridges between microtubules in vivo. Periodicity of Tau Arms on Microtubules From our data, spacings between adjacent arms on microtu- bules overlap with the sites defined by the 6-dimer superlat- tice, but are not identical. Jensen and Smaill studied the arrangement of MAP2 on microtubules by microdensitome- ter-computer correlation techniques and proposed that MAP2 projections are arranged in a "saturated 12-dimer, unsatu- rated 6-dimer" superlattice (20). Because MAP2 and tau compete for binding sites on microtubules in vitro (21), our results about tau binding sites coincide well with Jensen and Smaill's model about MAP2-binding sites. The binding sites of MAP2 and tau showed some tendency to overlap. Our stoichiometry data indicated the molar ratio of tau versus tubulin to be 1:~5. This value is in agreement with a previous report by Kim et al. (21), who used a different ap- proach. From their data the molar ratio of tau/tubulin in tau- saturated microtubules ranged from 1:3 to 1:5. The spacings between adjacent arms in the tau-saturated microtubules overlap with the 6-dimer supperlattice, while there are also additional spacings. From these data, if we accept the molar ratio oftau to tubulin as being 1:5, a simple possibility would be that one arm is composed of one tau. In the present study the tau arms appeared to cross-link with adjacent microtu- bules. Because the microtubules were cross-linked even in suspensions and because we found similar short cross bridges between microtubules in vivo, we suppose that tau proteins are able to cross-link microtubules only when the latter are in close proximity to one another. Concerning the further question of how many molecules it would take to form a cross bridge between microtubules, there are two possibilities. One is that if one arm has two tubulin-binding sites, one arm could form a cross bridge. The other possibility is that one cross bridge is composed of two arms from adjacent microtubules. From our data it is also supposed that a tau molecule may possess a strong bind- ing site to tubulin and other weak binding site to tubulin or tau. It may be possible to resolve this question in future studies using monoclonal antibodies. Binding Sites of Tau Proteins on Microtubules Previous immunocytochemical studies revealed that al- though MAP1 localizes in neuronal dendrites, cell bodies and axons, MAP2 is located mainly in dendrites and cell bodies and tau exists mainly in the axon (3, 4, 14, 15, 17, 24). The intracellular mechanisms for the sorting and transport of these proteins to their destinations is unknown. In this re- gard the binding sites of these MAPs on the microtubules are In 30.1 ~.o 411.1 ~.I 60.I LENGTH 70,1 II1.1 H.I ~.o ~ ,o~., rim Figure 14. A histogram of