Functional organization of microtubule-associated protein tau. Identification of regions which affect microtubule growth, nucleation, and bundle formation in vitro.

Tau protein is a microtubule-associated protein that is almost exclusively expressed in the brain and is enriched in the axon. Determination of tau's sequence has revealed three to four tandem repeats that have been shown to constitute the microtubule binding site. In order to study the functional organization of tau, we prepared a series of truncated tau fragments using an Escherichia coli expression system. We assayed each fragment's activity in promoting growth of microtubules and in nucleating free microtubules. We found that tau's ability to nucleate microtubules requires the presence of additional sequence amino-terminal to that required for growth. We demonstrate that tau's carboxyl and amino termini differentially affect microtubule growth and nucleation. Finally, we show that in vitro microtubule bundle formation occurs when tubulin is assembled in the presence of an amino- and carboxyl-terminally truncated tau protein, whereas almost no bundling is observed in the presence of full-length tau or tau fragments that contain the amino terminus in addition to the repeat domain. We conclude that although the presence of the repeat domain promotes the growth of microtubules, the structural requirements for nucleation activity are more stringent. The differentiation between the growth promoting and nucleation activities on the structural level makes it possible for the two activities to be differentially regulated in vivo.

alternative mRNA splicing (for reviews see Ref. 10 and 11). The carboxyl-terminal half of tau protein is conserved across species and is also homologous to MAPB, whereas the amino terminus is more variable. The carboxyl-terminal half contains the repeat domain, which consists of stretches of 18 residues that are imperfectly repeated three times in the fetal and four times in the adult specific form; the repeats are separated by 13-or 14-residue spacer regions. The repeat domain constitutes the microtubule binding region with a single repeat being sufficient for interacting with tubulin (12-16). Synthetic 18-mer copies of a repeat can promote microtubule polymerization, although the molar ratio of peptide required is 2 orders of magnitude higher than that of fulllength tau (12, 13, 15). The adult form of tau promotes assembly of microtubules more actively than fetal forms (17- 19), probably reflecting a higher binding affinity due to the additional repeat (16).
Although it is clear that the repeat domain constitutes the microtubule binding site, it is less clear what roles the aminoand carboxyl-terminal parts of tau play. It has been reported that sequences adjacent to the repeat domain increase the binding affinity to taxol-stabilized microtubules in vitro (16) and are required for binding of tau to microtubules in cells (20). In addition, evidence exists that regions beyond the repeats might affect the formation of microtubule bundles i n vitro and in vivo (19)(20)(21)(22)(23).
In order to study the functional organization of tau protein and the effects of sequences adjacent to the repeats on microtubule assembly, we prepared a series of truncated fragments of tau using a bacterial expression system. We tested the fragments for activity to promote growth of microtubules by length measurements of centrosome-nucleated microtubules and activity to nucleate free microtubules using a newly developed ELISA-based assay and immunofluorescent microscopy. In addition we studied the role of the amino-and carboxyl-terminal parts of tau protein in the formation of microtubule bundles in vitro using electron microscopy.

EXPERIMENTAL PROCEDURES
Materials-All reagents, unless otherwise specified, were obtained from Sigma.
Construction of Expression Plasmids-Prokaryotic expression plasmids were constructed in PET-3d (24). Inserts for all plasmids except for n fragment and n[ 1841 were prepared by polymerase chain reaction (PCR) using pl9tau and p2ltau cDNA (14,20). PCR primers contained restriction sites and start and stop codons as needed. The inserted sequences are shown in Fig. 1. Plasmids for n fragment and n [184] were constructed by cutting and re-circularizing the plasmid containing n123c; amino acids contributed at the carboxyl terminus of these tau fragments are given in the legend of Fig. 1.
Tubulin was isolated from calf brain by two assembly-disassembly cycles and phosphocellulose chromatography as described previously (25), applying less than 3 mg of protein/ml of resin. Peak fractions were pooled, brought to 80 mM K-PIPES, 1 mM EGTA, 1 mM MgC12, 0.1 mM GTP, pH 6.8, and stored in small aliquots at -135 "c.
Immunoblotting using affinity-purified tau antibody and '251-labeled secondary antibodies failed to detect contaminating levels of tau protein in the tubulin preparation.
Centrosomes were isolated from Chinese hamster ovary cells as described previously (25). Fractions containing centrosomes were identified by performing regrowth assays in the presence of 0.7 mg/ ml twice cycled microtubule protein; fractions were stored in small aliquots at -135 "C.
Centrosome-mediated Microtubule Regrowth Assay-The assay was essentially performed as described previously (25,26). The incubation mixture was prepared at 0 "C and contained, in 50 pl of BRB80, 1 mM GTP, 15 p~ tubulin, centrosomes, and, when specified, 0.3 p M tau fragment. Incubation was for 10 min at 37 "C. Reaction was stopped by addition of 200 p1 of 1% (v/v) glutaraldehyde in BRB80, containing 1 mM GTP, at room temperature followed 3 min later by the addition of 1 ml of BRB80. The mixture was sedimented for 15 min at 25,000 X g through a 5-ml cushion of 25% (v/v) glycerol in BRB80 onto polylysine-treated coverslips. The supernatant was aspirated, and the interface washed with 1% Triton X-100. The coverslip was removed and postfixed with methanol at -20 "C for 5 min. Immunofluorescence was performed with DMlA anti-tubulin monoclonal antibody and rhodamine-conjugated goat anti-mouse antibody (Boehringer). Coverslips were mounted in Aquamount (Lerner Laboratories, Pittsburg, PA). Fluorescent microscopy employed a Zeiss Axioskop Neofluar 63X lens. For length measurements, microtubules were photographed and lengths determined by projection of 35-mm negatives onto a digitizing tablet interfaced with a Macintosh computer. For each condition, 100 microtubules were traced; the mean length and standard error calculations were computer-aided. To determine the number of microtubules nucleated per centrosome, microtubules nucleated by 10 randomly chosen centrosomes were counted from the immunofluorescence photographs and averaged.
Competitive Enzyme-linked Immunosorbent Assay (ELISA)-The incubation mixture contained, in 50 pl of BRB80, 1 mM GTP, 15 p~ tubulin, and tau fragment as specified. Incubation was for 10 min at 37 "C. Reaction was stopped by addition of 200 pl of 1% (v/v) glutaraldehyde in BRB80, containing 1 mM GTP, at room temperature followed 3 min later by the addition of 1 ml of BRB80. Polymerized tubulin was pelleted for 30 min at 100,000 X g, and aliquots of the supernatant were assayed using biotinylated tubulin as a competitor. Residual glutaraldehyde was found not to interfere with the assay. Biotinylation was performed as previously described (27) at a ratio of 250 pg of biotin-N-hydroxysuccinimide (Zymed, San Francisco, CA) per mg of tubulin for 4 h at room temperature. ELISA was conducted on Microtiter plates (Nunc, Kamstrup, Denmark) coated with 100 pllwell 1:5,000 DMlA anti-tubulin monoclonal antibody and evaluated using a biotin-avidin-alkaline phosphatase-system (Vector, Burlingame, CA). The linear range of the ELISA was determined as 0.2-2 pg of tubulin. performed as described above (see "Competitive ELISA") but in a Determination of Microtubule Number-Microtubule assembly was total volume of 25 pl. The reaction was stopped after 2, 5, 10, and 30 min at 37 'C by the addition of 100 pl of 1% glutaraldehyde in BRB80, containing 1 mM GTP, at room temperature followed 3 min later by the addition of 500 pl of BRB80. Microtubules were collected by sedimenting 1% or 0.1% aliquots, diluted in a total of 500 p1 of BRBSO, onto polylysine-treated coverslips for 1 h at 25,000 X g. Immunofluorescence and fluorescent microscopy was performed as described above (see "Centrosome-mediated Microtubule Regrowth Assay"). The number of microtubules was determined by counting microtubules from five randomly chosen frames (175 X 120 pm); average number per frame and standard error were calculated. Blot Overlay Assay-1 pg of tau protein and fragments of tau were separated by SDS polyacrylamide gel electrophoresis (28) with 15% acrylamide. Electrophoretic transfer onto Immobilon-P (Millipore, Bedford, MA) was carried out in 0.2 M glycine, 250 mM Tris, 20% (v/ v) methanol overnight at constant current (100 mA). Blots were washed and incubated with tubulin (25 pg/ml) as previously described (29). Immunodetection was done with DMlA anti-tubulin monoclonal antibody, alkaline-phosphatase-conjugated goat anti-mouse antibody (Promega, Madison, WI), and developed using a 5-bromo-4-chloro-3indolyl-phosphate/nitro blue tetrazolium phosphatase substrate system (Kirkegaard and Perry, Gaithersburg, MD).
Electron Microscopy-Microtubule assembly was performed as described above (see "Competitive ELISA") but in a reaction volume of 100 p l . The reaction was stopped after 10 min at 37 "C by the addition of 400 p1 of 1% glutaraldehyde in BRB80, containing 1 mM GTP, at room temperature. Microtubules were pelleted either on a polylysinetreated coverslip as described above (see "Centrosome-mediated Microtubule Regrowth Assay"), followed by postfixation for 5 min at -20 "C in methanol or in Beem embedding capsules (Fullam Inc., Latham, NY) for 1 h at 100,000 g. Samples were rinsed in cacodylate buffer (0.1 M sodium cacodylate, pH 7.4), stained with 2 mg/ml tannic acid in cacodylate buffer for 5 min, fixed for 20 min with 1% OsOl (Electron Microscopy Sciences, Ft. Washington, PA) in cacodylate buffer containing 0.8% (w/v) potassium ferricyanide (Electron Microscopy Sciences) dehydrated in ethanol series, and embedded in quetol653 (Ted Pella, Redding, CA). Polymerization of the resin was for 24 h at 60 "C. Sections of the sample were stained with lead citrate and viewed by electron microscopy.
Other Methods-Protein concentrations were determined by the method of Bradford (30) using bovine serum albumin as a standard. SDS-polyacrylamide gel electrophoresis was performed as previously described (28) and stained with Coomassie Blue.

RESULTS
Purification of Tau Fragments Expressed in E. coli-We constructed 10 recombinant plasmids encoding amino-and carboxyl-terminally truncated tau fragments as shown in Fig.  1. The fragments were expressed and purified as described under "Experimental Procedures." All purified fragments were reactive to polyclonal anti-tau antibody (data not shown) and were pure according to SDS-polyacrylamide gel electrophoresis ( Fig. 1). Bands at molecular weights lower than the full-length protein were due to limited proteolysis during expression, as confirmed by immunoreaction.
Microtubule Growth-promoting Activity of Expressed Tau Fragments-It has been reported previously that isolated centrosomes nucleate microtubule assembly in uitro when incubated with phosphocellulose-purified tubulin (25). In this assay, the use of centrosomes gives microtubule nucleation a kinetic advantage by anchoring and stabilizing microtubules and provides a population of microtubules whose growth can be studied in uitro. As previously reported, the presence of tau protein increases both microtubule growth and the average number of microtubules nucleated per centrosome, presumably by reducing microtubule instability (26). We used the centrosome-mediated microtubule regrowth assay to assess the effect of tau fragments on the promotion of microtubule growth and on centrosome-mediated nucleation activity.
The minimal concentration of tubulin heterodimers for centrosome-nucleated microtubule growth in the absence of tau was 9.5 pM for our preparation, which is in the range of previous reports (25, 26, 31). Increasing the tubulin concentration led to a linear increase in the mean length of centrosome-nucleated microtubules as measured after 10 min of incubation (data not shown). In addition, when the mean length of centrosome-nucleated microtubules was determined during a time course ranging from 2 to 10 min, the increase was linear (data not shown), indicating that the mean length of the assembled microtubules at 10 min reflects the microtubule growth rate. We routinely used 15 p~ tubulin and an incubation time of 10 min in all assays; under these conditions centrosome-nucleated microtubules, assembled in the absence of any tau, were long enough to be measured from immunofluorescence photographs. Pilot experiments revealed that tau increased the mean length of microtubules a t tau:tubulin ratios up to 0.04 in the presence of 15 p~ tubulin. At higher tau:tubulin ratios, spontaneous assembly of free, i.e. not centrosome-nucleated, microtubules was observed. Therefore, all regrowth assays were carried out a t a tau:tubulin ratio of 0.02 (tau concentration of 0.3 p~) .
When tubulin and centrosomes were incubated with fulllength tau protein (n123c), the mean microtubule length was more than 10-fold that found in the absence of tau (Fig. 2). n123 fragment did not differ significantly in its activity, whereas 1232 fragment was approximately 30% less active. Fragments that contained four repeats were generally more active compared to the corresponding three-repeat constructs. The n fragment was not active. These results indicate that the repeat domain is necessary to promote microtubule growth with the number of the repeats and the amino-terminal flanking region (residues 1-172) moderately affecting its activity.
In addition, the presence of the fragments that contained the repeat domain increased the number of microtubules nucleated by the centrosomes (average of 22.0 5 5.6 microtubules/centrosome); this represented a 4-5-fold increase over that found in the absence of tau or in the presence of the amino terminus (4.0 f 0.8 and 3.8 f 0.7 microtubules/centrosome, respectively). Since the fragments' activities to promote microtubule growth and to promote centrosome-mediated nucleation correlated, the increase in the number of microtubules per centrosome might result from the reduction of microtubule instability by tau rather than from a direct effect of tau on the centrosome.
Microtubule Nucleation Activity of Expreswd Fragments-It has been previously reported that MAPS promote the nucleation of microtubules in vitro (32,33). In order to assess the ability of tau fragments to nucleate microtubules, we used an ELISA-based assay to quant,itate the amount of polymerized tubulin in small reaction samples (see "Experimental Procedures"). In developing this assay, we had found by examining the products by immunofluorescent microscopy, that the appearance of free microtubules correlated with measurable microtubule mass (data not shown). In the absence of tau no microtubule mass was measurable after 30 min of incubation and tubulin concentrations up to 80 p M . Since all fragments that contained the repeat domain promoted microtubule growth as measured by the centrosomemediated microtubule regrowth assay (Fig. 2), accumulation of microtubule mass in the absence of centrosomes is indicative of tau's ability to nucleate free microtubules. The number of free microtubules correlated with the amount of polymerized tubulin as measured by the ELISA-based assay (data not shown), indicating that the ELISA data reflect primarilv nucleation rates rather than growth rates.
When 15 p~ tubulin was incubated for 10 min with increasing amounts of tau protein, an increase in the amount of polymerized tubulin was observed (Fig. 3a). This is consistent with previous results obtained by turbidometric studies showing that tau protein promotes tubulin polymerization in a dose-dependent manner (17-19, 34-36). In the presence of 11123 and n12.14 fragment a similar increase in the amount of polymerized tubulin was observed (Fig. 3c). In the presence of the 123c and 1234c fragment no polvmerization of tubulin was observed (Fig. 3e). At high tau:tubulin ratios (>O.l), some assembly of microtubules could be detected bv immunofluorescent microscopy; however, since extensive hundling of microtubules was observed under these conditions, it is likely that this assembly differs from tau-specific nucleation of microtubules (discussed below). In order to determine whether part or all of the amino-terminal region is necessary for tau's ability to nucleate microtubules, we tested an additional construct which extended 19 residues into the amino-terminal region ([154]123c). In the presence of this fragment a similar increase compared to full-length tau in the amount of polymerized tubulin was observed (Fig. 3g), indicating that a small amino-terminal region is sufficient to restore the ability to nucleate microtubules to 123c fragment.
In order to investigate the activity of microtubule nucleation, we determined the number of microtubules as a function of reaction time (Fig. 3, 6, d, f , and h). Using a tau:tubulin ratio of 0.1, the number of microtubules was maximal after 5-10 min. The decrease of microtubule number a t longer reaction times is probably due to some dynamic instability under conditions of lowered concentrations of free tubulin, since the polymerization of tubulin, as judged by ELISA, was almost complete after 10 min (data not shown). The number of microtubules polymerized in the presence of full-length tau (n123c) was almost maximal (>95%) after 2 min. When tubulin was incubated with n123 fragment, the number of microtubules was maximal after 5 min, with 60% maximal after 2 min, indicating a lower nucleation activity compared to full-length tau. In the presence of 123c fragment almost no microtubules were observed. When tubulin was incubated with [154]123c fragment, many microtubules were observed, but the number of microtubules increased more slowly compared to full-length tau, indicating a lower nucleation activity.
In order to test whether the region that is additionally required for tau's ability to nucleate microtubules (residues 154-173), acts independently from the repeat domain, we prepared an n fragment containing additional residues 164-184 (n[184], Fig. 1). n[184] fragment had no microtubule growth promoting activity as tested by the centrosome-mediated microtubule regrowth assay (data not shown). When coincubated with 123c fragment no tubulin polymerization could be observed (data not shown), indicating that residues 154-173 restore the ability to nucleate microtubules only when part of the same molecule that contains the repeat domain.
Since tau's ability to nucleate microtubules might require binding of tubulin heterodimers, we tested tubulin binding by a blot overlay assay (29). When tubulin heterodimers were incubated with immobilized tau fragments, all fragments containing the repeat domain-bound tubulin (Fig. 4). Interestingly, although tubulin did not bind to the n fragment, a slight binding to n(1841 fragment was detectable (Fig. 4, lone 91, suggesting a weak tubulin-binding site a t positions 164-184.

Electron Microscopic Obseruatwn of Assembled Microtubules and Effect of Expressed Fragments on the Formation of Microtubule
Bundles-It has been reported previously that microtubules assembled in vitro in the presence of synthetic peptides encoding part of the repeat domain are bundled (22); the term "bundle" was used to describe the appearance of several microtubules being closely attached to each other as observed by electron microscopy. In order to study the effect of the amino and carboxyl termini of tau on bundle formation, we prepared a tau fragment containing mainly the repeat domain (Fig. 1, fragment 123). This fragment behaved unlike larger fragments that contained the repeat domain and two anomalies were observed. 1) When tubulin was incubated with this fragment, a substantial amount of microtubule assembly was observed despite the absence of sequences upstream of 173. 2) Free single microtubules were never observed, and electron microscopy revealed that the microtubules were bundled (Fig. 5, a and 6). We speculated that in the presence of this fragment, microtubules assembled abnormally with nucleation occumng through a mechanism other than that employed by the other tau fragments.
In order to quantitate the extent of bundle formation,  n123 ( l o n e 2), [154]123c (&ne 3). 123c   (lane 4 ) , 123 ( l o n e 5 ) , n1234c (lane 6 ) , n1234 (&ne 7), 1234c (lone   8), n[184] (lane 9), n (&ne IO). and tubulin as positive control (lane 11) were separated by SDS-polyacrylamide gel electrophoresis, transferred to Immobilon-P, and incubated with tubulin heterodimers. The extent of hundled microtuhules ( a ) and the average numher of microt.uhules per hundle ( b ) were determined from the electron microscopy results shown in Fig. 5 (h-a). At least 100 microtubules in cross-sections of microtubule pellet were evaluated. Closely attached microtuhules with no detectable intermicrotuhule distance were considered to he "hundled." microtubules were polymerized in the presence of fragments, fixed, pelleted, and analyzed by electron microscopy (Fig.  5 ,  h-e, and Fig. 6). Since this procedure requires assembly of free microt,ubules prior to analysis of bundle formation, only those fragments capable of nucleating free microtubules were tested. When microtubules were polymerized in the presence of full-length tau (n123c), almost no microtubules in bundled formation were observed (Fig.  5c). In the presence of 12.7 fragment almost all microtubules were bundled (Fig.  5 h ) . When tubulin was incubated with [ 1.541 123c, both the extent of bundle formation as well as the average number of microt,ubules per bundle were reduced in comparison to 123 frag-ment (Fig. 6). In the presence of n123 fragment, the extent of bundling was similar to that of full-length tau. In all cases, the individual microtubules appeared morphologicallv normal with a diameter of 25 nm, indicating that the 123 fragment is sufficient to assemble normal microtubules.

DISCUSSION
In this report, we demonstrate that tau's activitv to promote growth of microtubules and its activity to nucleate microtubules are mechanistically distinct and involve different regions of tau. This is consistent with previous studies, which have demonstrated that microtubule assemblv occurs by a condensation polymerization mechanism consisting of distinct nucleation and microtubule growth steps with MAPS affecting both (32,871. Although the repeat domain is required for both steps, the structural requirements for nucleation are more stringent and require additional sequence. Since this region (residues 154-173) cannot confer initiation activitv alone and cannot complement nucleation activity if not contiguous to the repeat domain, it might either contribute an additional tubulin-binding site, as suggested by the blot overlay result, or be required for protein folding. According to Chou-Fasman secondary structure prediction (38, 39), residues 168-173 would contribute an additional /$-sheet structure to the protein. The presence of this region might also he important for tau's function in oioo, since the inclusion of residues 164-172 is required for the 123c fragment to localize to microtubules in cells (20). The presence of tau's carboxyl terminus (residues 308-352 for the three-repeat form) affects microtubule nucleation activity, whereas its effect on microtubule growth is small, if any. The presence of the fourth repeat in the adult specific form does not change tau's structural requirements for nucleation and growth of microtubules but confers a higher activity in promoting growth of microtubules.
The differentiation at the structural level between tau's activity to promote microtubule growth and to nucleate microtubules makes it possible for the two activities to be independently regulated. The area required for tau's abilitv to nucleate microtubules (residues 154-307) is conserved in all isoforms and species of tau protein sequenced and is also found in MAPS. In vivo tau is posttranslationallv modified, and phosphorylation has been shown to affect tau's activity in vitro (40) and MAP2's localization in oioo (41). Tau can be phosphorylated in vitro by various protein kinases (42)(43)(44)(45)(46)(47)(48). Interestingly, all phosphorylation sites so far identified are located within or close to the region required for promotion of microtubule growth and nucleation. Since it has been shown that phosphorylation affects structural properties of tau (49), it is possible that phosphorylation induces conformational changes that might affect tau's activity in wavs similar to those brought about by deletions.
As previously reported (16), the presence of regions flanking the repeat domain increase hinding to h x o l stabilized microtubules. Increased hinding to microtubules might contribute to the increased growth-promoting activitv of the constructs containing amino-terminal flanking regions and the fourth repeat. However, the effect of amino-terminal deletions on tau's ability to nucleate microtubules is more complex and cannot he explained by the binding data; tau-dependent nucleation of microtubules may involve steps that are not affected by its binding affinity to microtubules.
Our study does not assess tau's effect on the dynamic instabilit,y of individual microtubules (50). However, such a study has been recently completed by Drechsel ct nl. using real-time video microscopy (.51). Thev find that at 3 p~

Functional Organization
of T a u 3419 tubulin and 0.3 PM tau, there is essentially no evidence for dynamic instability of microtubules. Since we work at higher tubulin concentrations, this further lessens the catastrophe frequency in our system. Therefore, it is likely that the mean microtubule growth rate measured in our assays reflects the elongation rate. Furthermore, it is unlikely that dynamic instability would account for the absence of nucleation activity, since under our conditions, microtubules exhibit net growth at both ends even in the absence of any MAPs (52). It has been proposed that microtubule stabilization and bundling result from the neutralization of the acidic carboxylterminal region of tubulin, which can be achieved by the binding of MAPs or a variety of other treatments (22). However, it has been reported that MAPs suppress bundling of microtubules in uitro (53,54). In this report, we show that a fragment lacking the amino and carboxyl termini (fragment 123) induces bundling of microtubules, whereas full-length tau and a fragment that contains the amino-terminal region and the repeat domain (fragment n123) do not. A fragment containing the carboxyl terminus and the repeat domain has intermediate activity. Therefore, the carboxyl-and aminoterminal regions of tau might hinder microtubules from becoming spatially close to one another; the large number of prolines in the amino-terminal region and the negatively charged amino acids at the extreme ends of tau suggest an extended structure that protrudes from the microtubule. This is consistent with the observation of tau as a projection from the microtubule wall (55). Since fragment 123 assembles microtubules despite the absence of residues upstream of 173 and since extensive bundling was observed in the presence of this fragment, we suggest that these microtubules are assembled by a mode similar to that achieved by non-MAP related conditions which may involve neutralization of the carboxylterminal region of tubulin (22,53,54,(56)(57)(58). A non-repeatcontaining MAP4 fragment also exhibits similar behavior (23). Although these activities have mainly been described in uitro, we speculate that an important feature of tau may be to suppress similar activities in uiuo, leaving the nucleation of microtubules to be regulated by tau alone.