Expression of tau protein in non-neuronal cells: microtubule binding and
stabilization
GLORIA LEE and SUSAN L. ROOK
Program in Neuroscience, Harvard Medical School and Center for Neurologic Diseases, Department of Medicine (Division of Neurology),Brigham and Women's Hospital, Boston, MA 02115, USA
Summary
The microtubule-associated protein tau is a developmen-tally regulated family of neuronal phosphoproteins thatpromotes the assembly and stabilization of micro-tubules. The carboxy-terminal half of the proteincontains three copies of an imperfectly repeated se-quence; this region has been found to bind microtubulesin vitro. In addition, a fourth copy of the repeat has beenfound in adult-specific forms of tau protein. To examinethe structure and function of tau protein in vivo, we havetransiently expressed fetal and adult forms of tau proteinand tau protein fragments in tissue culture cells.Biochemical analysis reveals full-length products withheterogeneity in post-translational modification syn-thesized in the cells. Immunofluorescent staining oftransfected cells shows that, under our conditions,sequences on both sides of the repeat region are requiredfor in vivo microtubule co-localization. These additionalregions may be required either for enhancing micro-
tubule contacts or for proper protein folding in the cell.In our expression system, the bundling of cellularmicrotubules occurs only in transfections using four-repeat tau constructs; any four-repeat construct capableof binding is also able to induce bundling. Our datasuggest that the presence of bundles is correlated withenhanced microtubule stability; factors that increasestability such as higher levels of tau protein expression orthe presence of the fourth repeat, increase the fraction oftransfected cells showing bundles. Finally, the presenceof tau protein in the cell allows all interphase micro-tubules to become acetylated, a post-translational modi-fication usually reserved for a subset of stable cellularmicrotubules.
Microtubules are ubiquitous structures in eukaryoticcells and perform a variety of cellular functions. Theability of microtubules to achieve such functionaldiversity has been thought to be partially providedthrough the assistance of various associated proteins.Tau protein is a microtubule-associated protein foundprimarily in neuronal tissue. It promotes microtubuleassembly in vitro (reviewed by Olmsted, 1986) andstabilizes cellular microtubules when microinjected intocells (Drubin and Kirschner, 1986). Expression of tauprotein has been correlated with the extension ofneurites during the differentiation of pheochromo-cytoma cells (Drubin et al. 1985). Moreover, thepresence of tau antisense in primary neuronal cellculture appears to block the development and mainten-ance of axon-like processes, suggesting a role for tauprotein in the establishment of neuronal cell polarity(Caceres and Kosik, 1990; Caceres et al., 1991). In situ,tau protein is associated with axons (Binder et al., 1985;Kowall and Kosik, 1987; Brion et al., 1988; Trojanowskiet al., 1989).
The earliest purification of tau protein revealed afamily of related phosphoproteins (Cleveland et al.,1977). It is now clear that some of this heterogeneity isgenerated by alternative splicing (Himmler, 1989). Inhuman, cDNA clones for six tau isoforms and theirdevelopmental expression have been described(Goedert et al., 1988; Goedert et al., 1989a,b). Humanadult-specific forms are distinguished by the presence ofone or more of exons2, 3 or 10 (Goedert et al., 1989a,b;exon number as assigned by Himmler (1989) for bovinegene). The most striking feature of the primarystructure of tau protein as predicted from cDNA clonesis the presence of three evenly spaced, imperfectlyrepeated copies of an 18 amino acid sequence in thecarboxy-terminal half of the protein. One or morecopies of the repeat are able to bind to microtubules invitro (Butner and Kirschner, 1991; Himmler et al.,1989; Lee et al., 1989). Not surprisingly, increasing thenumber of copies of the repeat increased the efficiencyof microtubule binding (Butner and Kirschner, 1991;Lee et al., 1989). Synthetic one-repeat peptides alsopromote microtubule assembly; however, the requiredstoichiometry indicates that the peptides are much less
228 G. Lee and S. L. Rook
potent than intact protein (Ennulat et al., 1989; Joly eta!., 1989). The high molecular weight microtubule-associated protein MAP2 shares about 50% amino acidhomology with tau protein at the carboxy-terminal end220 residues and also contains three repeats (Lewis etal., 1988). As expected, a fragment of MAP2 from thisarea coassembles with microtubules in vitro (Lewis etal., 1988).
The adult-specific exon 10 encodes a 31 amino acidsequence that contains a fourth copy of the microtubulebinding repeat unit. The inclusion of this fourth repeatincreases the activity of tau protein in in vitromicrotubule assembly assays (Goedert and Jakes,1990). The expression of a four-repeat tau protein invivo results in the bundling of cellular microtubules(Kanai et al., 1989).
In this study, we examined the in vivo structure andfunction of fetal and adult tau protein. Transientlyexpressed tau protein and tau protein fragments werecharacterized using Western blot and immunoprecipi-tation protocols. The ability of the expressed protein toassociate with cellular microtubules and to inducemicrotubule bundling were determined by immunoflu-orescence. Our results indicate that, under our con-ditions, sequences adequate for in vitro binding are notadequate for in vivo co-localization to microtubules.Furthermore, the presence of adult-specific tau proteincan induce microtubule bundling; microtubule stabilityis also increased. Bundling and stability may be related,since no evidence for cross-linking of microtubules bytau protein was found. Lastly, the presence of tauprotein correlated with the acetylation of all cellularmicrotubules, indicating enhanced stability for allmicrotubules.
Materials and methods
Construction of expression plasmidsThe eukaryotic expression vector pECE (Ellis et al., 1986)was modified in the following manner to introduce a uniqueNcol cloning site into its multiple cloning site. First, the soleNcol site in pECE was removed by Ncol cleavage followed byfill-in and re-circularization. The plasmid was then cut withBglll, the first site in the polylinker site downstream from the
SV40 early promoter, filled in, then ligated to an Ncol linkerin order to introduce an Ncol site at this location. Thesequence of the new multiple cloning site in this vector,named pECEN+, was checked by DNA sequencing.
Most three-repeat human tau inserts were prepared bypolymerase chain reaction (PCR) using pl9tau plasmid DNA(Lee et al., 1989) as template. PCR primers containedrestriction sites that enabled insertion into pECEN+. pEn wasprepared by cutting and recircularizing pEnl23c. In someplasmids, translation was terminated by a stop codon in theprimer; in other plasmids, translation was terminated by thestop codons in the vector. Table 1 lists all plasmids used in thisstudy; tau sequences expressed and new sequences contribu-ted by primers, linkers or vectors are shown. The longestsequence contributed in this manner was 11 residues and itspresence did not correlate with any functional property of theexpressed proteins.
pEnl234c was constructed from pEnl23c by replacing thetau cDNA carboxy-terminal 500 bp with the analogous 593 bpfragment from an adult human tau four-repeat cDNA clone(Mori et al., 1989) kindly provided by Dr. H. Mori (TokyoMetropolitan Institute of Gerontology, Tokyo, Japan). Theresultant cDNA is a full-length four-repeat tau cDNAidentical to a clone previously identified by Goedert et al.(1989a). pE1234c(487) and pE1234(487) were constructedfrom pE123c(487) by similar replacements. pEnl23(4c) wasprepared by cutting and recircularizing pEnl234c. Theremaining human four-repeat tau inserts were prepared byPCR from pEnl234c. Polymerase chain reaction primers usedto prepare inserts for insertion into pECEN+ were identical tothose used for the three-repeat human tau inserts and,therefore, the starts and stops of expressed sequences areidentical.
Because most constructs were made using polymerase chainreaction technology, we acknowledge the small probability ofa base change occurring during synthesis. However, most ofthe DNA fragments prepared were fairly small and thebiochemical characterizaton of expressed protein (see below)would detect the inadvertant production of a terminationcodon. Also, all of the constructs overlapped, so that if asingle amino acid replacement were to alter dramatically theproperties of the expressed protein, an inconsistency wouldbe noted.
TransfectionsCHO or 3T3 (NIH) cells were grown in oMEM sup-plemented with 10% fetal bovine serum (Hyclone, Logan,UT) and seeded onto 12 mm coverslips in 24-well plates at adensity of approximately 0.2 x 104 /well. CHO cells weretransfected by calcium phosphate precipitation using standard
techniques (Ausubel et al., 1989). 3T3 cells were transfectedby calcium phosphate precipitation as modified by Chen andOkayama (1988). A2ftg sample of plasmid DNA was addedper well.
Nocodazole treatment of the transfected cells was per-formed by adding nocodazole (Aldrich Chemical Co.,Milwaukee, WI) to a final concentration of 3.3 /iM, thenincubating for the times indicated. Triton extractions wereperformed by first washing cells with extraction buffer minusTriton at 37°C, then incubating with extraction buffer (80 mMPipes, pH 6.8, 1 mM MgCI2, 1 mM EGTA, 0.1% Triton-X100, 30% glycerol) at 37°C (Solomon et al., 1979).Extraction buffer was then removed and cells washed withextraction buffer minus Triton, then with PBS (137 mM NaCI,2.7 mM KC1, 10 mM PO4, pH 7.4) before fixation. The timeof extraction had been determined by inspecting the cells afterextraction using anti-tubulin staining and by immunoblottingTriton-soluble and -insoluble fractions with anti-tubulin. Overextraction (1 min for 3T3 or 2 min for CHO cells) resulted in alarge increase in Triton-soluble tubulin as detected byimmunoblot and ruptured cytoskeletal network as detected byimmunofluorescence. Therefore, extraction times of 0.5 minand 1 min were chosen for 3T3 and CHO, respectively. Cellswere judged to be fully extracted, since non-cytoskeletal taufragments were no longer detectable by immunofluorescence.
ImmunofluorescenceCells were fixed with 0.3% glutaraldehyde in 80 mM Pipes,pH 6.8, 1 mM MgCl2, 5 mM EGTA, 0.5% NP40 for 10minutes at room temperature (Drubin and Kirschner, 1986).After washing with PBS, cells were incubated with 10 mg/mlNaBH4 for 7 minutes, washed with PBS, then incubated with0.1 M glycine for 20 minutes. After washing with PBS, cellswere incubated with methanol for 10 minutes. The initial stepof this protocol combines permeabilization and fixation,because performing permeabilization first results in theextraction of tau protein from the detergent-insoluble cyto-skeleton. Since full-length tau protein remains bound tomicrotubules under this fixation protocol, we chose tocompare all constructs using these identical conditions. Fixedcells were washed with PBS, 10 mg/ml BSA, 0.1% Tween 20prior to antibody staining.
The antibody staining protocol entailed staining with thefirst primary antibody, washing with PBS, 10 mg/ml BSA,0.1% Tween 20 (5 X 2 min), staining with second primaryantibody, washing similarly, then staining with both labeledsecondary antibodies. The primary antibodies used were:affinity-purified polyclonal antibody against tau protein(Pfeffer et al., 1983) at 1:1000 for 15 minutes at roomtemperature, monoclonal antibody DMa-1 against tubulin(ICN Biochemicals, Inc., Irvine, CA) at 1:500 for 15 minutesat room temperature, and monoclonal antibody 6-11B-1against acetylated tubulin kindly provided by Dr. G. Pipernoand used at 1:10 overnight at 4°C. Secondary antibodies weregoat anti-rabbit IgG, rhodamine- or fluorescein-labeled(Boehringer Mannheim Corp., Indianapolis, IN), and goatanti-mouse IgG, rhodamine- or fluorescein-labeled (Boehr-inger Mannheim Corp., Indianapolis, IN). Secondary anti-bodies were incubated with cells for 30 minutes at roomtemperature. Coverslips were mounted in 1 mg/ml p-phenylenediamine, 90% glycerol (Johnson and Araujo,1981). Cells were photographed using x63 Neofluar lens on aZeiss Axioskop.
Bundling efficiency was determined by scoring 100-400transfected cells for bundles. A cell was scored positive forbundling if one or more obviously thick, dense bands ofmicrotubules were observed in the cell (see Fig. 5, below).
Western blotting and immunoprecipitationCell lysates for Western blot analysis were prepared fromtransfected 3T3 cells by using 60 y\ boiling of Laemmli SDSsample buffer (Laemmli, 1970) to scrape transfected cellsfrom 24-well plates. Lysates were electrophoresed on 7.5% to15% gradient gels, then transferred to Immobilon (MilliporeCorp., Bedford, MA). The blot was probed with affinity-purified antibody to tau protein and I-labeled goat anti-rabbit secondary antibody (ICN Radiochemicals, Irvine,CA).
An estimation of the average amount of tau proteinsynthesized per transfected cell was made on the basis of thetransfection efficiency and the amount of tau protein detectedby Western blotting. The quantity of tau protein present in aharvested cell lysate (2X104 cells) was calculated by excisingthe Immobilon band corresponding to the expressed protein,counting the 125I-labeled goat anti-rabbit secondary antibodybound, and converting the cts/min to protein, using Immobi-lon bands from adjacent lanes processed in parallel, whichcontained known amounts of fetal brain tau protein orEscherichia coli tau protein. The average amount synthesizedper transfected cell equaled quantity in lysate divided bynumber of transfected cells (2X104 x transfection efficiency;transfection efficiency was determined by a parallel transfec-tion processed for immunofluorescence). Because the taupolyclonal antibody used to probe the blot was derived againstbrain tau protein, the reactivity of the antibody to thestandard brain tau protein vas greater than the reactivity toE. co/i-synthesized tau protein, which lacks phosphorylatedepitopes (e.g. 1 ng E. coli protein corresponded to 420 cts/minwhile 1 ng brain protein corresponded to 1516 cts/min). Sinceit is not known whether the reactivity of 3T3 expressed tauprotein more closely resembles that of brain protein or E. coliprotein, the amount of protein expressed is given as a range,using E. coli standard for the higher estimate, brain proteinfor the lower estimate. While in theory, a monoclonalantibody would have reacted to tau from all sources equally,in practice this did not work, since the monoclonal we triedwas unable to yield a signal above background in detecting tauin transfected cell lysates. A comparison of the reactivity ofthe two antibodies to various quantities of E. coli tau proteinshowed the signal from the tau monoclonal (5E2) to be 5- to10-fold less than that from the affinity-purified tau polyclonalantibody.
For two-dimensional gel analysis, transfections were per-formed in 6-well plates using 5 ng DNA per well. Cells wereharvested by scraping with 80 jj\ boiling lysis buffer (0.5%SDS, 50 mM Tris-HCl, pH 7.6,150 mM NaCI). Samples wereelectrophoresed on NEPHGE gels (O'Farrell et al., 1977)using pH 3-10 Ampholines. The second dimension was 10%SDS/polyacrylamide gel, which was blotted and probed asdescribed above. E. coli lysate containing full-length tauprotein was prepared from E. coli BL21(DE3)LysS cellstransformed with pET-3d expression plasmid (Studier et al.,1990) harboring full-length human three-repeat tau cDNA.
Cell lysates for immunoprecipitation were prepared from3SS-methionine-labeled transfected 3T3 cells by using 80 fj\lysis buffer supplemented with 2 mM EDTA, 2 mM PMSF(phenylmethylsulfoxyl fluoride) to scrape transfected cellsfrom 6-well plates. (Each well had received approximately 175jtCi of 35S-methionine for 4 hours prior to harvesting.Harvesting was performed 48 hours after DNA addition.)Lysates were boiled and then immunoprecipitated as de-scribed by Drubin et al. (1988) with the following modifi-cations: lysate was precleared twice with Pansorbin and oncewith Protein A/Sepharose beads (Pharmacia, Piscataway,NJ), Protein A/Sepharose beads were used for adsorption of
230 G. Lee and S. L. Rook
antibody-antigen complexes, and beads were washed 5 timesin NET (0.5% NP40, 150 mM NaCl, 50 mM Tris-HCl, 5 mMEDTA) buffer prior to elution. Complexes were electrophor-esed on 15% to 25% or 10% to 20% gradient gels. Gels werefluorographed prior to autoradiography.
Results
Full-length tau protein and tau protein fragmentsexpressed in vivoTo study the in vivo structure and function of humantau protein, a panel of eukaryotic expression plasmidscontaining tau sequences was derived from both three-repeat and four-repeat tau cDNA clones (Fig. 1A,B).The four-repeat tau cDNA used was identical to thethree-repeat clone except for the inclusion of tau geneexon 10, which encodes a fourth copy of the repeat.Most of the deletion constructs were made in the samemanner for both the three-repeat and four-repeatconstructs so that the starting and ending sequences ofthe expressed proteins were identical (Table 1). Theseplasmids were used to express transiently tau protein in3T3 cells, which normally do not express tau protein.
To verify the presence of tau proteins in 3T3 cells, thetransiently transfected cells were analyzed by Westernblotting. Fig. 2 shows that detectable levels of full-
Mfcrotubule bindlnj
pEn123c
pEn
pEn1
pEn123(843)
pEn123(921)
pE123c(487)
pE123c<517)
pE123(517)
B
pEn1234c
pEn1234(92l)
pEn1234(843)
pEM23(4c)
pE1234c<487)
pE1234c<517)
pE1234(517)
pE1234(487)
— 307
173 —
173 307
Wcrotubule
Binding BundUng
— — — J3»
311
2tS
173 — —
173
M
3J3
331
331
| required byI all isoforms
281(312)
required only bythree repeat
\ \
307 352(338) (383)
• not required
length products are obtained from pEnl23c,pEnl23(921), pEnl23(843), pEn, PEnl234c,pEnl234(921) and pEnl234(843) (refer to Fig. 1 for tausequences expressed). The estimated molecular masseswere consistently higher than those calculated fromprimary structure data; this discrepancy may be due inpart to phosphorylation of the protein (Lindwall andCole, 1984) and has been noted for tau protein purifiedfrom brain as well. While multiple bands were seen inseveral lanes, it is unlikely that this reflected degra-dation, since there were no bands in common amongthe constructs, all of which overlapped in sequence.
To confirm the lack of degradation for the smallerconstructs from the carboxy-terminal end, it wasnecessary to label the cells with -^S-methionine andimmunoprecipitate tau protein as Western blotting wasnot sufficiently sensitive to visualize these proteins. Fig.3 shows the labeled products immunoprecipitated fromtransiently transfected cells. Plasmids pE123c(487),pE123c(517), pE123(517), pE1234c(487) andpE1234c(517) exhibited protein patterns that suggestthat intact protein is synthesized. The molecular massescalculated from size standards on the gel were muchhigher than those calculated from primary sequencedata. For example, the two proteins in pE123c(487)(Fig. 3A, lane 2) average 32.5 kDa while the primarysequence calculation yields 20 kDa for residues 164 to352. Therefore, besides indicating that intact protein is
Fig. 1. (A) Schematic diagram of expression plasmidinserts from three-repeat tau cDNA. Each plasmid isconstructed in expression vector pECEN+ as described inMaterials and methods, and is predicted to express theindicated amino acids from tau protein. The human full-length three-repeat clone used was first reported byGoedert et al. (1988). The three 18 amino acid repeatsidentified in mouse tau protein (Lee et al., 1988) arelocated at residues 198-215, 229-246 and 261-278 in human;each repeat is marked with a short line. Microtubulebinding capability, detected by immunofluorescence, isshown at the right. pE123(487), which encodes residues164-307, was not included because no transfected cells weredetected. (B) Expression plasmid inserts from four-repeattau cDNA. Plasmids were constructed as described above.The human full-length four-repeat tau cDNA used isidentical to the three-repeat cDNA above, except that itincludes exon 10 (Himmler, 1989), which results in theinsertion of 31 residues at residue 217 of the three-repeatsequence. The inserted sequence is shown by the heavy barand contains the fourth repeat. Therefore, amino acids 248-383 of the four-repeat protein correspond to the three-repeat protein residues 217-352. The repeats are located atresidues 198-215, 229-246, 260-277 and 292-309 and aremarked with short lines. Microtubule binding and bundlingcapability for each construct are shown at the right.Relative bundling efficiencies are discussed in Results.(C) Diagram showing the role of various tau proteinregions towards in vivo colocalization to microtubules.Numbers denote amino acid residues of three-repeat tauprotein; numbers in parenthesis denote analogous residuenumbers in four-repeat tau protein located downstreamfrom exon 10, which inserts at residue 217. R indicatesrepeat region with the first repeat starting at residue 198and the last repeat ending at 281(312).
CM r-- T-
o n N
V
s B CO CO C\J00 CO r-co co co
— 43 • —«3
- 2 9- 2 8
Expression of tau protein in 3T3 celb 231
N N N Nw m o oco co co co
O *d" CO x f COn ffl N (O N
B CO
8CO00 O
DC
"-43
"-29
-18
-14
-43
"-29
— * ' « •
Fig. 2. Immunoblot of transiently transfected 3T3 cells.Cells were transfected and harvested as described inMaterials and methods. Lysates were electrophoresed on a7.5% to 15% gradient gel, blotted and probed withaffinity-purified anti-tau followed by 125I-labeled goat anti-rabbit secondary. (A) Duplicate transfections usingpEnl23c (1-352), pEnl23(921) (1-307), PEnl23(843) (1-281)and pEn (1-163). Numbers at top indicate three-repeat tauresidues expressed. (B) Transfections using pEnl234c (1-383), pEnl234(921) (1-338) and pEnl234(843) (1-312).Numbers at top indicate four-repeat tau residuesexpressed. Molecular mass markers (in kDa) are asindicated.
present, these data also suggest that post-translationalmodification in the 3T3 cells causes dramatic shifts inthe gel mobility of these carboxy-terminal fragments. Itis likely that these shifts are caused by phosphorylation;it has been suggested that phosphorylaton may changethe conformation of the molecule such that SDSbinding is increased (Lindwall and Cole, 1984). Two-dimensional gel analysis shows in vivo expressed tauprotein to be much more acidic than the identicalprotein synthesized in E. coli (data not shown).
In Fig. 3A, cells transfected with pE123c(517) (lane3) express two forms that appear to differ in molecularmass by approximately 2.5 kDa. When the carboxy-terminal 45 amino acids was truncated (pE123[517],lane 5), heterogeneity was lost, since only one form wasdetected. This suggests that residues 308-352 contain asite that creates heterogeneity in the gel mobility of theprotein. Indeed, Steiner et al. (1990) have reported thatphosphorylation of serine residue 327 of tau proteinresults in a shift in the electrophoretic mobility of tauprotein. However, other sites with this property exist,since doublet bands are also exhibited by constructslacking this same region (e.g. pEnl23[921],pEnl234[921]).
The difference in gel mobility between three-repeatfragments 164-352 and 173-352 (Fig. 3A, lanes 2,3) islarger than one would expect from a nine amino acid
1 2 3 4 5 1
Fig. 3. Immunoprecipitation of 35S-labeled tau proteinfragments from transiently transfected 3T3 cells. Immunecomplexes from 3T3 transfected cell lysates were preparedas described in Materials and methods. (A) A 15% to 25%gradient gel displaying products from transfections withthree-repeat constructs. Lane 1 shows controlimmunoprecipitation from non-transfected cells. Lane 2contains immunoprecipitated products from pE123c(487)transfection expressing residues 164-352, lane 3pE123c(517) expressing residues 173-352, lane 4pE123(487), and lane 5 pE123(517) expressing residues 173-307. Positions of bands marked with arrows (lane 2)correspond to molecular masses of 34 kDa and 31 kDa;bands marked with asterisks (lane 3) 26.3 kDa and 23.8kDa; dark band in lane 5, 16 kDa. Molecular masscalculated from primary structure for 164-352 is 20 kDa,173-352 19 kDa, and 173-307 14.4 kDa. (B) A 10% to 20%gradient gel displaying products from transfections withfour-repeat constructs. Lane 1 shows immunoprecipitatedproducts from pE1234c(487) transfection expressingresidues 164-383, lane 2 pE1234c(517) expressing residues173-383, and lane 3 non-transfected control. Positions ofbands marked with arrows correspond to molecular massesof 35 kDa and 33 kDa. Molecular masses calculated fromprimary sequence data for residues 164-383 is 23.3 kDa andfor 173-383 is 22.3 kDa. Molecular mass standards are asshown.
difference. Since the 164-352 products (lane 2) areestimated to be 34 kDa and 31 kDa while the 173-352products (lane 3) are estimated to be 26.3 kDa and 23.8kDa, the inclusion of residues 164-172 in these three-repeat fragments results in an apparent increase inmolecular mass of at least 7 kDa. This implies thatresidues 164-172 may affect the conformation of theprotein, causing the retardation of the fragments in gelelectrophoresis. While the sequence of this region,EPKKVAWR, does not contain any serines orthreonines, residue 173 is a threonine, which raises thepossibility that phosphorylation of this residue couldaffect gel mobility or SDS binding (the loss of residues
232 G. Lee and S. L. Rook
164-172 could affect the recognition of residue 173 bykinases). The contribution of residues 164-172 towardscausing gel mobility shifts is diminished, however, whenfour repeats are present. Fig. 3B shows that four-repeatfragments 164-383 (lane 1) and 173-383 (lane 2) lookidentical.
We were unable to immunoprecipitate products forpE123(487) (Fig. 3A, lane 4), pE1234(487) andpE1234(517) (data not shown). Since these plasmidshad transfection efficiencies of 0-1% as assessed byimmunofluorescence (see below), we judged that verylittle protein was present. Plasmids whose productswere detectable by Western blotting had transfectionefficiencies of 20-40% while those detectable only byimmunoprecipitation had efficiencies of 5-10%. Sincethe necessity to immunoprecipitate correlated withplasmids encoding carboxy-terminal fragments, wesurmised that these protein fragments or their mRNAswere susceptible to cellular degradation.
An estimate of the amount of full-length tau proteinsynthesized in transfected 3T3 cells was made bycomparing the Western blot signal to the signal given byknown quantities of tau protein (see Materials andmethods). By dividing this number by the number oftransfected cells in the culture, as determined in aparallel transfection, we estimate that an average of0.5-2 picograms of full-length tau protein is synthesizedin a transiently transfected cell.
Requirements for co-localization of tau protein andmicrotubules in vivoThe in vivo function of tau protein and tau proteinfragments was assessed by examining transiently trans-fected CHO or 3T3 cells by immunofluorescence. Thecells were fixed 24-48 hours after DNA addition andstained with antibodies to tau protein and tubulin, usingdouble immunofluorescence to show the location of tauprotein relative to the microtubules. Because the cellstransiently express tau protein, the level of expressionvaries from cell to cell, depending on the number ofgene copies present. In looking at many cells, we foundthat only the intensity of the pattern of staining variedwith the level of expression; the pattern did not vary. Afraction of the transfected cells had no discerniblestaining pattern, perhaps due to the cell-cycle stage,and were not included in the analysis.
Fig. 4 shows tau protein and tubulin staining of CHOcells transfected with plasmids pEnl23c, pEn andpE123. As expected, full-length tau (pEnl23c) co-localizes with microtubules (Fig. 4a) while the amino-terminal fragment having no repeats (pEn) showscytoplasmic rather than cytoskeletal staining (Fig. 4b).Surprisingly, cells transfected with pE123 did not showtau protein co-localizing to microtubules (Fig. 4c); thisfragment contains the entire repeat region and wasshown to bind to taxol-stabilized microtubules in vitro(Lee et al., 1989). Interaction with the cytoskeleton wasconfirmed by extraction with 0.1% Triton prior tofixation and immunofluorescence (conditions for ex-traction were derived as described in Materials andmethods). Triton extraction removes soluble corn-
Fig. 4. Immunofluorescent staining of CHO cellstransfected with tau-expressing plasmids. Cells weretransfected, fixed and stained as described in Materials andmethods. (a,b,c) Anti-tau staining. The transfectingplasmids were: a, pEnl23c encoding amino acids 1-352;b, pEn encoding amino acids 1-163; and c, pE123 plasmidencoding amino acids 173-307. (d,e,f) Corresponding anti-tubulin staining for each transfection. Cells in a and d wereTriton-extracted prior to staining as described in Materialsand methods. Bar, 10 /.an.
ponents from the cytoplasm, leaving intact micro-tubules unobscured by cytoplasmic staining (Solomonet al., 1979). This procedure did not reveal any fractionof cytoplasmically localized tau fragments binding tomicrotubules. Also, none of the constructs exhibitingmicrotubule localization showed significant loss ofbinding after Triton extraction.
Fig. 1 shows the microtubule localization results forour panel of constructs. From our data, we concludethat under our conditions of fixation, residues 164 to 307of the three-repeat tau protein are required formicrotubule association in the cell. For four-repeat tauconstructs, deleting residues upstream from residue338, which corresponds to three-repeat residue 307,also resulted in loss of binding. However, the requiredsequence at the amino-terminal side of the four-repeat
Expression of tau protein in 3T3 cells 233
region was different from that required for the three-repeat protein - a four-repeat construct that was deletedup to 173 still bound. Fig. 1C summarizes the tauprotein structural requirements for in vivo microtubuleco-localization. The necessity for the extra nine resi-dues in three-repeat constructs may be related tophosphorylation and/or protein folding, as suggested bythe immunoprecipitation results. While four-repeatfragments 164-383 and 173-383 co-migrate on gels (Fig.3B, lanes 1,2) and co-localize to microtubules in vivo,three-repeat fragments 164-352 and 173-352 differ bothin their gel migration patterns (Fig. 3A, lanes 2,3) andin their microtubule localization abilities. Fragment173-352 has increased gel mobility relative to 164-352and is unable to co-localize with microtubules.
Lastly, although our deletion analysis of three- andfour-repeat constructs identifies the unique sequencerequired in vivo as amino acids 164-307 for the three-repeat and 173-338 for the four-repeat protein, con-structs expressing just residues 173-338 or 164-338 didnot bind microtubules in vivo. However, since we couldnot detect intact protein by immunoprecipitation forthese two plasmids (pE1234[517] and pE1234[487]), wecannot rule out the possibility that their inability to bindwas due to degradation. Alternatively, these fragmentsmay not have been able to fold properly and requiredinclusion of neighboring regions to achieve functionalconformation. The localization of the three-repeatfragment 164-307 could not be determined because notransfected cells were detected by immunofluorescencefor this plasmid. Immunoprecipitation also failed todetect protein (Fig. 3A, lane 4).
Microtubule bundlingCells with full-length four-repeat tau protein(pEnl234c) exhibited bundled microtubules similar tothose reported in mouse L-cells by Kanai et al. (1989),where a rat full-length four-repeat tau cDNA was used.Bundles were most dramatically seen as sharp spikesemanating from the cell or as thick dense bands runningalong the edge of cells (Fig. 5). Kanai et al. (1989) havedefined bundles as microtubule structures with 10-100times the thickness of normal microtubules. In non-transfected 3T3 cells, microtubules did not coalesce intosuch structures. In contrast to the cytoplasmic versuscytoskeletal localizations described above, the occur-rence of bundles appeared to be correlated with thelevel of expression of the four-repeat protein; bundleswere never found in cells with faint tau staining.Transfections with four-repeat protein showed anaverage of 20% transfected cells containing bundledmicrotubules. In contrast, while the three-repeat con-struct consistently had a higher transfection efficiencythan the four-repeat construct, only 1% of cellstransfected with the three-repeat protein showedbundling. When non-transfected cells were scored forbundled microtubules, less than 0.25% were positive.
Using the collection of human tau constructs bearingfour repeats, we found that every tau protein fragmentcapable of binding to microtubules also exhibitedbundled microtubules (Fig. IB). Therefore, the car-
Fig. 5. Immunofluorescentstaining of 3T3 cellstransfected with four-repeattau cDNA showingmicrotubule bundling. Cellswere transfected, fixed andstained as described inMaterials and methods. Thetransfecting plasmids were:(a.b) pEnl234(921); and (c)pE1234c(517). Anti-taustaining is shown. Anti-tubulin staining pattern oftransfected cells was virtuallyindistinguishable from theanti-tau pattern. Bar, 10 /m\.
boxy-terminal 45 residues and the amino-terminal 172residues of tau protein are not required for this result.
Tau protein stabilized microtubules in cellsDrubin and Kirschner (1986) have shown that tauprotein, when microinjected into cells, stabilizes micro-tubules against drug-induced depolymerization. Wefound that cells transfected with four-repeat tau cDNAwere able to withstand more prolonged drug treatmentthan those transfected with three-repeat cDNA. Micro-tubule fragments were still visible in cells expressingfour-repeat protein after 1 hour of incubation in 3.3 /xMnocodazole (Fig. 6a) while none were present in thethree-repeat protein transfectants (Fig. 6b) and non-transfected cells (Fig. 6c). After 3 hours of drugtreatment, microtubules were absent in all cells.
The enhanced stability of tau-associated micro-tubules was further demonstrated by staining trans-fected cells with an antibody to acetylated tubulin(Piperno and Fuller, 1985). Microtubules containingacetylated-tubulin subunits have been identified as astable subset of microtubules that persist after low-levelnocodazole treatment (Piperno et al., 1987) and turnover less rapidly in the cell relative to dynamicmicrotubules (Schulze et al., 1987; Webster and Borisy,1989). The top cell in Fig. 7b shows a typical acetylatedtubulin pattern in a non-transfected cell. Looking attransfected cells that exhibited an interphase micro-
234 G. Lee and S. L. Rook
Fig. 6. Microtubules stabilized by the presence of tau protein against depolymerization by nocodazole. Transfected 3T3cells were treated with 3.3 fjM nocodazole for 1 hour prior to Triton extraction and fixation as described in Materials andmethods, (a) Full-length four-repeat tau-transfected cell; (b) full-length three-repeat tau-transfected cell; and (c) non-transfected cell. The cells shown are stained with anti-tubulin. Transfected cells nocodazole-treated and Triton-extractedwere readily distinguishable from non-transfected cells due to background material, presumably free tau and tubulin,located in and around the transfected cells that stained with tau and tubulin antibody. Bar, 10 fan.
linked to its acquisition of acetylated subunits, thepresence of tau in the cell does enable a larger set ofmicrotubules to become acetylated than would haveotherwise.
Discussion
Regions flanking the repeat region contribute tomicrotubule co-localization in vivoThe structure and function of tau protein and MAP2protein have been studied in vitro using proteinfragments synthesized in E. coli or in rabbit reticulocytelysate systems (Butner and Kirschner, 1991; Himmler etal., 1989; Lewis et al., 1988; Lee et al. 1989). These datasuggest that sequences up to residue 190 on the amino-terminal side and sequences downstream from residue281 on the carboxy-terminal side could be deletedwithout abolishing in vitro microtubule binding activity.However, our in vivo experiments indicate that ad-ditional sequences on each side of the repeat region arenecessary for co-localization of tau protein to cellularmicrotubules. Butner and Kirschner (1991) also foundin their in vitro binding study that sequences outside ofthe repeats (residues 92-190 or 289-352, numbered as inthree-repeat tau) can increase the affinity of a taufragment for microtubules 10- to 20-fold. With MAP2,Lewis et al. (1989) also found that including residues145-193 or 292-352 qualitatively enhanced in vivomicrotubule binding. Therefore, while the presence ofthree- to four-repeat motifs can result in 80-100% of theinput tau fragment binding to microtubules in vitro(Butner and Kirschner, 1991), contributions made bythe flanking regions are important for in vivo micro-tubule co-localization. Our data more precisely definethe areas of tau protein that enhance microtubuleinteraction. As summarized in Fig. 1C, we find thatincluding up to residue 164 on the amino side or downto residue 307 on the carboxy side of the repeats issufficient to differentiate between in vivo binding andnot binding in our system.
The role of the contributions made by the flanking
Fig. 7. Immunofluorescent staining of acetylatedmicrotubules in transfected cells. Transfected 3T3 cellswere Triton-extracted before fixation and stained with anti-tau (a and c) and anti-acetylated tubulin (b and d) asdescribed in Materials and methods, (a and b) A non-transfected cell (top) and a cell expressing three-repeat tau(bottom), (c and d) A cell expressing four-repeat tau. Bar,10 ^m.
tubule array, many contained an acetylated tubulinpatterni that was virtually identical to the tau-stainedmicrotubule pattern (Fig. 7). This occurred using boththe three- and four-repeat proteins. Other transfectedcells had the typical acetylated tubulin patterns found innon-transfected cells (data not shown). Therefore,while the presence of tau on a microtubule is not tightly
Expression of tau protein in 3T3 celb 235
regions may be to enhance microtubule interaction byproviding additional contacts that by themselves areinsufficient to bind microtubules. These contacts maynot be needed for in vitro binding because in vitroassays contain a vast excess of stabilized microtubules,which may drive the binding reaction. Our in vivo datareflect the cellular conditions normally encountered bytau protein. Besides smaller amounts of microtubules,there may also be the presence of competing endogen-ous microtubule binding proteins. Alternatively, theflanking sequences may be necessary for efficientfolding of tau fragments. While this may not have beencritical for an in vitro assay with purified components, invivo, an improperly folded protein may bind elsewhereor lack the higher binding affinity required to yielddetectable localization to microtubules. While ourfixation conditions may reflect a 'stringent' bindingrequirement, the existence of these non-repeat regionsthat enhance binding may suggest the existence ofspecific functions of tau that require a higher bindingaffinity than that solely necessary to bind to micro-tubules in vitro.
Microtubule bundling in the presence of tau andMAP2Our results show that the expression of the fourth-repeat leads to a discernible functional difference in themicrotubule organization and stability in cells. In ourexpression system, every four-repeat microtubule-binding tau construct was more effective in inducingmicrotubule bundling than the analogous three-repeatconstruct. The immunofluorescent staining of cellstransiently transfected with a fetal mouse tau ex-pression plasmid, as reported by Lewis et al. (1989), isconsistent with our results in that very little bundling isexhibited by three-repeat forms. However, they foundthat MAP2, which also contains three repeats, inducesbundling. Since MAP2 and tau protein share only 50%homology in the carboxy-terminal 220 residues, it islikely that the sequence differences account for thedifference in bundling capability. Moreover, MAP2 ismore basic than tau protein in this region and, since theinteraction of these proteins with microtubules has beenpostulated to be one of an ionic nature (Vallee, 1982),MAP2 would be more effective than tau protein. It hasbeen reported that MAP2 promotes assembly, nu-cleates microtubules and stabilizes microtubules moreefficiently than tau protein (Sandoval and Vandekerck-hove, 1981).
Recently, Lewis and Cowan (1990) have reportedthat the area responsible for bundling microtubules inMAP2 is the area homologous to amino acids 313-335 offour-repeat tau protein (282-304 in three-repeat) andthat MAP2 lacking this region is still able to bindmicrotubules. In tau protein, this area is required notonly for microtubule bundling but also for binding. Thissuggests that the role of this area in bundling by MAP2might be to enhance binding, perhaps by influencing theconformation of the repeat region or by providingadditional interactions with microtubules. In agreementwith Lewis and Cowan (1990), we found no evidence
supporting the notion that the carboxy-terminal end oftau contains a short hydrophobic "zipper" sequencethat mediates microtubule bundling (Lewis et al. 1989).
Microtubule stabilization and bundlingThe deletion analysis of the four-repeat tau proteinindicates that sequences upstream from amino acid 173and downstream from 338 are not required formicrotubule bundling. We propose that bundling is verylikely the result of enhanced stabilization of micro-tubules, on the basis of the following reasons: (1) theareas of tau protein required for bundling correlate withthose involved in microtubule binding. (2) Bundledmicrotubules are seen mainly in cells with a high level oftau expression. (3) The presence of the required fourthrepeat increases the stability of the microtubules, asshown by nocodazole resistance. The number of cellsshowing bundled microtubules in a given transfectionwould be related to the association constant of thestabilizing protein and the level of expression. It isprobable that, for the induction of bundles, aheightened level of expression may compensate for alower association constant and vice versa. Kanai et al.(1989) have also suggested that bundling is increased incells with a high level of tau protein expression, on thebasis of a comparison of transient and stable expressiondata. In our study, the constructs with the highestbundling efficiencies were pEnl234c and pEnl234(921).Among the constructs capable of inducing bundles,these two also exhibited the highest transfectionefficiency and expressed picogram levels of protein incells, allowing for detection by Western blotting.Lastly, the propensity for microtubule bundles to formin the cell may also depend on the cell type employed,since bundling may require re-modeling of existingcytoskeleton; the ease with which the existing micro-tubules are re-arranged may vary from cell to cell type.The expression of tau protein in baculovirus-infectedSf9 cells provides an example of an extremely high levelof tau expression (greater than 10-fold of that oftransfected cells) achieved in the presence of a pool ofunpolymerized tubulin (Knops et al., 1991). In thissystem, microtubule bundles are formed in 100% of theinfected cells using either three- or four-repeat tau-expressing virus. This result stresses the difficulties incomparing tau protein's bundling activity where differ-ent expression systems have been used and indicatesthat valid comparisons between different isoforms ordifferent fragments must be made within a singleexpression system. Moreover, it also shows that aheightened level of protein expression can compensatefor a lowered binding affinity towards inducing micro-tubule bundling.
Our proposal that tau induces microtubule bundlingby stabilizing microtubules is also supported by ourfinding that the minimal unique tau sequence requiredfor binding and bundling is too small to span the 20-25nm distance between bundled microtubules, especiallyif the entire repeat region is bound to a singlemicrotubule. The possibility that a single repeat regionbinds to two microtubules would require that only a
236 G. Lee and S. L. Rook
single repeat be in contact with each microtubule; giventhe much reduced binding affinity of single repeat units(Butner and Kirschner, 1991), it seems unlikely thatsuch a configuration would be physiologically relevant.The report that taxol produces similar microtubulebundles (Chapin et al., 1991) is consistent with ourfindings. With the increased stability of microtubules intaxol-treated or tau-transfected cells, other cellularfactors may participate in the bundling process. Vallee(1990) has raised the possibility that cells transfectedwith MAP2 may over-express other MAPs that may beinvolved in bundling. Another possibility is thatstabilized microtubules may have new physicochemicalproperties at their surfaces that make them self-adherent.
If microtubule bundling does not require otherproteins, it should be reproducible in vitro. While wehave been able to obtain microtubule bundles in vitrousing E. co/i-synthesized tau protein in a centrosome-mediated microtubule regrowth assay (Brandt and Lee,unpublished data), the relationship between thesebundles and those seen in 3T3 cells remains to beshown.
The presence of tau protein is correlated with theacetylation of all cellular microtubulesThe acetylation of microtubules has been postulated tobe the consequence of microtubule stabilization(Piperno et al. 1987; Schulze et al., 1987; Webster andBorisy, 1989). Therefore, one might expect micro-tubules stabilized by tau protein to be acetylated. In ourtransfections, we found many transfected cellsexhibiting an acetylated tubulin staining pattern coin-ciding with the general cytoplasmic interphase micro-tubule array as detected by tau staining. This suppliesnew evidence that tau protein stabilizes microtubules invivo, leading to the acetylation of all microtubules inthe cell; other lines of evidence have been based on theuse of microtubule depolymerizing drugs (Drubin andKirschner, 1986). However, transfected cells withacetylated tubulin staining patterns resembling thosefound in non-transfected cells were also present; inthese cells, only a subset of microtubules were acety-lated. This suggests that microtubule acetylation isregulated by other factors in addition to stability, asconferred by tau protein association. Webster andBorisy (1989) have found that acetylated microtubulesare absent among prophase microtubules and that theirreappearance lags behind that of the cytoplasmicmicrotubules after the completion of mitosis. Thissuggests that acetylation may be cell cycle regulated andthat our transfected cells have a distribution ofacetylated tubulin patterns due to asynchrony in theculture.
ConclusionsIn summary, our results show that regions outside therepeat region of tau protein influence the ability of theprotein to bind to microtubules in vivo. We defineregions flanking the repeats that are required for in vivobinding under our conditions. The requirement for
additional sequence on each side of the repeat regionmay be related to protein folding and/or to additionalmicrotubule contact sites. The presence of tau proteinin the cell is correlated with the acetylation of allcellular microtubules, possibly in a cell cycle-dependentmanner. When tau protein with four-repeats wasexpressed, thick dense bands of microtubule bundleswere observed in cells. Deletion analysis did not revealany unique sequence associated with bundling; four-repeat constructs capable of binding also inducedmicrotubule bundling. Also, microtubules associatingwith four-repeat tau in vivo were found to have greaterstability, as assessed by nocodazole resistance, thanthose associating with three-repeat tau. Our datasuggest that microtubule bundling is correlated withincreased stability of microtubules. Since the fourthrepeat is found only in adult isoforms of tau protein, ourwork provides evidence that the developmental regu-lation of tau protein heterogeneity bears functionalimplications for cellular microtubules.
We sincerely thank Dr. Frank McKeon for help at variousstages of this work and Drs. Alfredo Caceres, Kenneth Kosikand David Gard for useful suggestions and discussion. Wealso thank Dr. Gianni Piperno for antibodies and Dr. HiroshiMori for his cDNA clone. The contributions of Ms. BelindaChang towards DNA sequencing are also gratefully acknowl-edged. Lastly, we thank Drs. David Gard, Garth Hall andRoland Brandt for critical reading of this manuscript. Thiswork was supported by the National Institute of GeneralMedical Sciences (no. GM39300-O4).
References
Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman,J. G., Smith, J. A. and Struhl, K. (1989). Current Protocols inMolecular Biology. John Wiley and Sons, New York, NY.
Binder, L. I., Frankfurter, A. and Rebhun, L. I. (1985). Thedistribution of tau in the mammalian central nervous system. / . CellBiol. 101, 1371-1378.
Brion. J. P. , Guilleminot, J. , Couchic, D. , Flament-Durand, J. andNunez, J. (1988). Both adult and juvenile tau microtubule-associated proteins are axon specific in the developing and adult ratcerebellum. Neuroscience 25, 139-146.
Butner, K. A. and Kirschner, M. W. (1991). Tau protein binds tomicrotubules through a flexible array of distributed weak sites. J.Cell Biol. 115, 717-730.
Caceres, A. and Kosik, K. S. (1990). Inhibition of neurite polarity bytau antisense oligonucleotides in primary cerebellar neurons.Nature 341, 461-463.
Caceres, A., Potrebfc, S. and Kosik, K. S. (1991). The effect of tauantisense oligonucleotides on neurite formation of culturedcerebellar macroneurons. J. Neurosci. 11, 1515-1523.
Chapin, S. J., Bulinski, J. C. and Gundersen, G. G. (1991).Microtubule bundling in cells. Nature 349, 24.
Chen, C. C. and Okayama, H. (1988). Calcium phosphate-mediatedgene transfer: a highly efficient transfection system for stablytransforming cells with plasmid DNA. Biotechniques 6, 632-638.
Cleveland, D. W., Hwo, S.-Y. and Kirschner, M. W. (1977). Physicaland chemical properties of purified tau factor and the role of tau inmicrotubule assembly. J. Mol. Biol. 116, 227-247.
Drubin, D. G., Feinstein, S. C , Shooter, E. M. and Kirschner, M. W.(1985). Nerve growth factor-induced neurite outgrowth in PC12cells involves the coordinate induction of microtubule assembly andassembly-promoting factors. /. Cell Biol. 101, 1799-1807.
Drubin, D. G. and Kirschner, M. W. (1986). Tau protein function inliving cells. J. Cell Biol. 103, 2739-2746.
Expression of tau protein in 3T3 cells 237
Drubin, D., Kobayashi, S., Kellogg, D. and Kirschner, M. (1988).Regulation of microtubule protein levels during cellularmorphogenesis in nerve growth factor-treated PC12 cells. J. CellBiol. 106, 1583-1591.
Ellis, L., Clauser, E., Morgan, E. O., Edery, M., Roth, R. A. andRutter, W. J. (1986). Replacement of insulin receptor tyrosineresidues 1162-1163 comprises insulin-stimulated kinase activity anduptake of 2-deoxyglucose. Cell 45, 721-732.
Ennulat, D. J., Llem, R. K. H., Hashim, G. A. and Shelanski, M. L.(1989). Two separate 18-amino acid domains of tau promote thepolymerization of tubulin. J. Biol. Client. 264, 5327-5330.
Goedert, M. and Jakes, R. (1990). Expresson of separate isoforms ofhuman tau protein: correlation with the tau pattern in brain andeffects on tubulin polymerization. EMBO J. 9, 4225-4230.
Goedert, M., Spillantini, M. G., Jakes, R., Rutherford, D. andCrowther, R. A. (1989b). Multiple isoforms of human microtubule-associated protein tau: sequences and localization in neurofibrillarytangles in Alzheimer's disease. Neuron 3, 519-526.
Goedert, M., Spillantini, M. G., Potier, M. C , Ulrich, J. andCrowther, R. A. (1989a). Cloning and sequencing of the cDNAencoding an isoform of microtubule-associated protein taucontaining four tandem repeats: differential expression of tauprotein mRNAs in human brain. EMBO J. 8 393-399.
Goedert, M., Wischik, C. M., Crowther, R. A., Walker, J. E. andKlug, A. (1988). Cloning and sequencing of the cDNA encoding acore protein of the paired helical filament of Alzheimer disease:Identification as the microtubule-associated protein tau. Proc. Nat.Acad. Sci. USA 85, 4051-4055.
Himmler, A. (1989). Structure of the bovine tau gene: alternativelyspliced transcripts generate a protein family. Mol. Cell. Biol. 9,1389-1396.
Himmler, A., Dreschel, D., Kirschner, M. W. and Martin, D. W., Jr(1989). Tau consists of a set of proteins with repeated C-terminalmicrotubule-binding domains and variable N-terminal domains.Mol. Cell. Biol. 9, 1381-1388.
Johnson, G. D. and Araujo, G. M. de C. N. (1981). A simple methodof reducing the fading of immunofluorescence during microscopy.J. Immunol. Meth. 43, 349-350.
Joly, J. C , Flynn, G. and Purlch, D. L. (1989). microtubule bindingfragment of microtubule-associated protein 2: location of theprotease- accessible site and identification of an assemblypromoting peptide. J. Cell Biol. 109, 2289-2294.
Kanai, Y., Takemura, R., Oshima, T., Mori, H., Ihara, Y.,Yanagisawa, M., Masaki, T. and Hlrokawa, N. (1989). Expressionof multiple tau isoforms and microtubule bundle formation infibroblasts transfected with a single tau cDNA. J. Cell Biol. 109,1173-1184.
Knops, J., Koslk, K. S., Lee, G., Pardee, J. D., Cohen-Gould, L. andMcConlogue, L. (1991). Overexpression of tau in a nonneuronalcell induces long cellular processes. J. Cell Biol. 114, 725-733.
Kowall, N. W. and Koslk, K. S. (1987). Axonal disruption andaberrant localization of tau protein characterize the neuropilpathology of Alzheimer's disease. Ann. Neurol. 22, 639-643.
Laemmli, U. K. (1970). Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature 227, 680-685.
Lee, G., Cowan, N. and Kirschner, M. (1988). The primary structureand heterogeneity of tau protein from mouse brain. Science 239,285-288.
Lee, G., Neve, R. L. and Kosik, K. S. (1989). The microtubulebinding domain of tau protein. Neuron 2, 1615-1624.
Lewis, S. A. and Cowan, N. (1990). Microtubule bundling. Nature345, 674.
Lewis, S. A., Ivanov, 1. E., Lee, G.-H. and Cowan, N. J. (1989).Microtubule organization in dendrites and axons is determined by a
short hydrophobic zipper in microtubule-associated proteins MAP2and tau. Nature 342, 498-505.
Lewis, S. A., Wang, D. and Cowan, N. J. (1988). Microtubule-associated protein MAP2 shares a microtubule binding motif withtau protein. Science 242, 936-939.
LindwaU, G. and Cole, R. D. (1984). Phosphorylation affects theability of tau protein to promote microtubule assembly. /. Biol.Chem. 259, 5301-5305.
Mori, H., Hamada, Y., Kawaguchi, M., Honda, T., Kondo, J. andIhara, Y. (1989). A distinct form of tau is selectively incorporatedinto Alzheimer's paired helical filaments. Biochem. Biophys. Res.Commun. 159, 1221-1226.
O'Farrell, P. Z., Goodman, H. M. and O'Farrell, P. H. (1977). Highresolution two-dimensional electrophoresis of basic as well as acidicproteins. Cell 12, 1133-1142.
Olmsted, J. B. (1986). Microtubule-associated proteins. Annu. Rev.Cell Biol. 2, 421-457.
PfefTer, S. R., Drubin, D. G. and Kelley, R. B. (1983). Identificationof three coated vesicle components as a- and /J-tubulin linked to aphosphorylated 50,000-dalton polypeptide. J. Cell Biol. 97, 40-47.
Piperno, G. and Fuller, M. T. (1985). Monoclonal antibodies specificfor an acetylated form of a^tubulin recognize the antigen in ciliaand flagella from a variety of organisms. J. Cell Biol. 101, 2085-2094.
Piperno, G., LeDizet, M. and Chang, X. (1987). Microtubulescontaining acetylated o^tubulin in mammalian cells in culture. J.Cell Biol. 104, 289-302.
Sandoval, I. V. and Vandekerckhove, J. S. (1981). A comparativestudy of the in vitro polymerization of tubulin in the presence of themicrotubule-associated proteins MAP2 and x. J. Biol. Chem. 256,8795-8800.
Schulze, E., Asal, D. J., Bulinski, J. C. and Kirschner, M. W. (1987).Posttranslational modification and microtubule stability. J. CellBiol. 105, 2167-2177.
Solomon, F., Magendantz, M. and Sal/man, A. (1979). Identificationwith cellular microtubules of one of the co-assembling microtubule-associated proteins. Cell 18, 431-438.
Steiner, B., Mandelkoow, E.-M., Biernat, J., Gustke, N., Meyer, H.E., Schmidt, B., Mieskes, G., Soling, H. D., Drechsel, D.,Kirschner, M. W., Goedert, M. and Mandelkow, E. (1990).Phosphorylation of microtubule-associated protein tau:identification of the site for calcium-calmodulin dependent kinaseand relationship with tau phosphorylation in Alzheimer tangles.EMBO J. 9, 3539-3544.
Studler, F. W., Rosenberg, A. H., Dunn, J. J. and DubendorfT, J. W.(1990). Use of T7 RNA polymerase to direct the expression ofcloned genes. Meth. Enzymol. 185, 60-89.
Trojanowski, J. Q., Schuck, T., Schmidt, L. and Lee, V. M.-Y.(1989). Distribution of tau proteins in the normal human centraland peripheral nervous system. J. Histochem. Cytochem. 37, 209-215.
Vallee, R. B. (1982). A taxol-dependent procedure for the isolation ofmicrotubules and microtubule-associated proteins (MAPs). J. CellBiol. 92, 435-442.
Vallee, R. B. (1990). Molecular characterization of high molecularweight microtubule-associated proteins: some answers, manyquestions. Cell Motil. Cytoskel. 15, 204-209.
Webster, D. R. and Borisy, G. G. (1989). Microtubules are acetylatedin domains that turn over slowly. J. Cell Sci. 92, 57-65.
{Received 23 January 1992 - Accepted 9 March 1992)
