Locations and immunoreactivities of phosphorylation sites on bovine and porcine tau proteins and a PHF-tau fragment.

Tau protein is a phosphorylated neuronal microtubule-associated protein. Tau protein is also present in the major pathological lesions of Alzheimer's disease in an insoluble hyperphosphorylated state as paired helical filaments (PHFs). We have investigated the phosphorylation state of control taus and a fragment of PHF-tau. Tau samples were digested with protease, separated by reversed-phase high-performance liquid chromatography, and analyzed by mass spectrometry and Edman microsequencing. The serine homologous with S404 of human tau 441 was phosphorylated on bovine and porcine tau and up to two phosphates were present on a peptide of amino acids 182-240 of bovine tau (193-251 of human tau 441). The serine within the KSPV motif was not phosphorylated on bovine or porcine tau. PHF-tau fragments, isolated from pronase-treated PHFs encompassed a 93-amino acid region within the microtubule binding domain. Enzymatic digestion and mass spectrometric analysis showed no phosphate was present and a second carboxyl terminus was identified at E380. Antibodies T3P and SMI34, which recognize PHF-tau and peptides phosphorylated at the sequence KSPV, both reacted with bovine and porcine tau even though the KSPV sequence was not phosphorylated. These data indicate that the 93-amino acid sequence of F5.5 tau from PHFs is not phosphorylated, and the serine equivalent to S404 of human tau is phosphorylated in bovine and porcine tau. Antibodies T3P and SMI34 react with phosphorylated epitopes that are not unique to PHF-tau and that are not necessarily at the KSPV site.

Tau protein is a phosphorylated neuronal microtubule-associated protein. Tau protein is also present in the major pathological lesions of Alzheimer's disease in an insoluble hyperphosphorylated state as paired helical filaments (PHFs). We have investigated the phosphorylation state of control taus and a fragment of PHF-tau. Tau samples were digested with protease, separated by reversed-phase high-performance liquid chromatography, and analyzed by mass spectrometry and Edman microsequencing. The serine homologous with S404 of human 7441 was phosphorylated on bovine and porcine tau and up to two phosphates were present on a peptide of amino acids 182-240 of bovine tau  of human 7441). The serine within the KSPV motif was not phosphorylated on bovine or porcine tau. PHF-tau fragments, isolated from pronasetreated PHFs encompassed a 93-amino acid region within the microtubule binding domain. Enzymatic digestion and mass spectrometric analysis showed no phosphate was present and a second carboxyl terminus was identified at E380. Antibodies T3P and SMI34, which recognize PHF-tau and peptides phosphorylated at the sequence KSPV, both reacted with bovine and porcine tau even though the KSPV sequence was not phosphorylated. These data indicate that the 93-amino acid sequence of F5.5 tau from PHFs is not phosphorylated, and the serine equivalent to 5404 of human tau is phosphorylated in bovine and porcine tau. Antibodies T3P and SMI34 react with phosphorylated epitopes that are not unique to PHF-tau and that are not necessarily at the KSPV site.
Tau protein is a microtubule-associated protein that induces microtubule assembly and stabilizes microtubules in vitro (Weingarten et al., 1975;Drubin and Kirschner, 1986). The amino acid sequences are known for human, mouse, and bovine tau proteins (Goedert et al., 1989;Lee et al., 1988;Himmler et al., 1989), but not for porcine tau. Six human tau isoforms have been identified, varying in length from 352 to 441 amino acids for the shortest isoform (7352) to the longest isoform (7441) (Goedert et al., 1989). These isoforms differ from each other depending on whether either a 29-or 58amino acid insert is present after K44 or a 31-amino acid insert is present after K216 of ~3 5 2 , in the tubulin-binding region (Goedert et al., 1989). Three tandem repeat regions are present in the tubulin-binding region of 7352, and the presence of the insert in this region results in four tandem repeats, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed. referred to as 3-repeat tau and 4-repeat tau, respectively. Tau protein is usually phosphorylated, introducing further heterogeneity (Lindwall and Cole, 1984;Butler and Shelanski, 1986). The sites phosphorylated in vitro by calcium-calmodulin-dependent protein kinase (Steiner et al., 1990), tau protein kinase (Ishiguro et al., 1991), protein kinase C (Correas et al., 1992), mitogen-activated protein kinase (Gustke et al., 1992), proline-directed protein kinase (Vulliet et at., 1992), and CAMP-dependent protein kinase (Scott et al., 1993) have been identified. The presence of phosphate groups on tau protein tends to decrease its ability to promote microtubule assembly (Lindwall and Cole, 1984;Hoshi et al., 1987;Yamamoto et al., 1985) and decreases the elasticity of tau (Hagestedt et ul., 1989). The locations of endogenously phosphorylated sites have not been precisely defined until recently. Hasegawa et al. (1992) have identified sites in both normal and PHF-tau by direct chemical analysis and have shown that at least three sites are phosphorylated on PHF-tau but not normal tau.
Prior to the work of Hasegawa et al. (1992), the locations of phosphorylation sites on either normal tau or PHF-tau had not been precisely determined, although attempts have been made to map these sites using antibodies. For example antibody T3P was raised to a synthetic peptide corresponding to amino acids 389-402 of 7441 phosphorylated at S396 (Lee et al., 1991). T3P reacts with PHF-tau but not normal tau, suggesting that S396 is phosphorylated in Alzheimer's disease. This serine occurs in the sequence, KSPV. This KSPV sequence also occurs in neurofilament peptides and antibody SMI34 (also referred to as 07-5) reacts with the phosphorylated version of neurofilaments and reacts with PHFs (Sternberger et al., 1985). Therefore, it too may recognize PHF-tau phosphorylated at S396 within the KSPV motif. This serine is near the tubulin-binding region of tau, and phosphorylation near this region has been suggested to cause a conformational change in this region which is required for SMI34 reactivity .
Another phosphorylation site in PHFs is postulated to be in or near amino acids 189-207 of 7441, the epitope for antibody tau-1 (Kosik et al., 1988). Tau-1 reacts with PHFtau or tangles only after they are dephosphorylated (Wood et al., 1986;Grundke-Iqbal et al., 1986b;Lee et al., 1991). The phosphate appears to occlude the antibody binding site or possibly induce a conformational change that the antibody does not recognize. Recently antibody AT8 was described, whose reactivity with PHF-tau seems to be localized to the phosphate that interferes with tau-1 binding .
The role of tau phosphorylation in the formation of PHFs is unknown. For example, it is not known whether the presence of phosphates at specific positions on tau converts tau into PHFs. This knowledge, however, could be valuable for designing therapeutic interventions for Alzheimer's disease. One way to approach this question is to determine whether the aberrant phosphorylation sites are unique to PHF-tau or whether they occur on tau molecules such as bovine or porcine tau, which are not known to form PHFs. Another issue that must be resolved in addressing this question is whether the antibodies that have been used to postulate the locations of specific tau phosphorylation sites are specific for phosphate at one particular location on tau or whether they can react with phosphates at several different locations. In the present study we address these issues by using mass spectrometry and Edman microsequencing to identify specific phosphorylation sites on bovine and porcine tau, as well as on a PHF-tau fragment. This fragment, isolated in the F5.5 fraction, binds tightly to PHFs and contains part of the tubulin-binding region of tau (Wischik et al., 1988). Several antibodies that react with PHF-tau in a phosphate-dependent fashion were also evaluated with these control tau protein preparations in order to determine whether they react with the phosphorylation sites that were identified on tau by direct chemical analysis.

MATERIALS AND METHODS
Bovine and Porcine Tau Proteins-Microtubules were purified from porcine brain and bovine brain using the large-scale method described by Murphy (1982). A phosphatase inhibitor mixture consisting of 20 mM NaF, 20 mM &glycerophosphate and 0.2 mM sodium orthovanadate was included in all buffer solutions. Tau was extracted from the microtubules using method three of Grundke-Iqbal et al. (1986a) and was further purified using size exclusion HPLC on a TSKgel G3000SW column pre-equilibrated in 50 mM Mes, pH 6.25, 150 mM NaCl, 5 mM dithiothreitol.
PHF-Tau Protein Fragment (F5.5 Tau)-Pronase-treated PHFs were extracted with formic acid, yielding the F5.5 fraction as described previously (Wischik et al., 1988). This F5.5 (formic acid extract) fraction contains a fragment of tau protein from the tubulin-binding region (referred to as F5.5 tau in this study). Two F5.5 tau samples were prepared from brains of two severe Alzheimer cases, referred to as 76 and 77 and described elsewhere (Harrington et al., 1991;Mutaetova-Ladinska et al., 1992). Formic acid was used at 88% to prepare the first sample (case 76) and at 2% for the second sample (case 77). The F5.5 samples were provided by Dr. C. Wischik.
One F5.5 sample (case 76) was further purified by affinity chromatography. A 200-pl aliquot was applied to an affinity column (monoclonal antibody 6.423-linked CNBr-Sepharose), which binds with F5.5 tau (Wischik et al. 1988). The column was washed with 40 mM Tris, pH 8.0, and the sample was eluted from the column with 40 mM acetic acid, pH 3.0, followed by a second elution step employing 40 mM Tris, 0.5 M sodium chloride, 0.05% Tween. By amino acid analysis the yield of purified protein was 20% of the original sample. NO protein was recovered in the second elution step.
Peptide fragments in the porcine tau and F5.5 tau digests were separated using a Waters HPLC system with two 510 pumps, a 490E UV detector and Maxima software, and a Waters DELTA PAK C18 300 A column. The same sequence of eluent samples was run as described above for the SMART system.
A 100-pg aliquot of F5.5 tau (case 77) was digested with trypsin (Worthington) at an enzyme-to-substrate ratio of 1:100, in 100 mM ammonium hydrogen carbonate, pH 8.5, for 2 h a t 37 "C. The resulting digest was chromatographed using the method described above for the other F5.5 tau sample.
In all cases, fractions were dried down to 5-10 pl on a Savant vacuum centrifuge. One-third of each fraction was submitted for mass spectrometry, and, where necessary, one-third of each fraction was submitted for Edman sequencing.
Mass Spectrometry-Fast atom bombardment mass spectrometry was performed on a VG ZAB2-SE mass spectrometer, using a thioglycerolglycerol matrix (1:1, v/v) (Sigma) containing l% trifluoroacetic acid. Electrospray mass spectrometry was measured using a VG instruments Bio-Q mass spectrometer. Samples were redissolved and introduced into the source in methanol (Analar grade):water (l:l, v/ v) containing 1% acetic acid (Analar).
Amino Acid Sequencing and Deriuatization of Phosphorylated Peptides with Ethanethiol-Amino-terminal sequence analysis was measured on an AB1 475 gas-phase sequencer and the phenylthiohydantoin amino acids were measured with an on-line AB1 120 analyzer. Peptides containing phosphorylated residues were derivatized with ethanethiol (Aldrich Chemical Co.) by the method of Meyer et al. (1986). The derivatized amino acid was detected using Edman sequencing.
Alkaline Phosphatase Treatment-Tau proteins were treated with alkaline phosphatase where indicated, prior to immunoassay. Ten pl of tau plus 1 unit of Escherichia coli alkaline phosphatase (Sigma) in 45 pl of 0.1 M glycine buffer, pH 10.5, 1 mM MgCIz, 0.1 M phenylmethylsulfonyl fluoride, and 5 mM benzamidine was incubated for 6 h at 37 "C. One unit of phosphatase was then added along with 1 pl of 200 mM phenylmethylsulfonyl fluoride and 5 p1 of 1 M benzamidine and incubated for an additional 18 h. 5 pl of bovine tau was combined with 5 units of calf intestine alkaline phosphatase (Boehringer Mannheim, molecular biology grade) in 50 pl of 30 mM ethanolamine buffer, pH 7.6, with 3 M NaC1, 1 mM MgCl,, and 0.1 mM ZnClz, further diluted to 300 p1 with 50 mM Tris buffer, pH 8, and incubated overnight at 37 "C.
Immunoassays-Aliquots of the same bovine and porcine tau protein samples that were subjected to mass spectrometry were diluted in 25 mM Tris buffer, pH 7.1 (porcine tau) or 50 mM sodium bicarbonate, pH 9.6 (bovine tau), and analyzed by ELISA as previously described (Caputo et al., 1992b). Control wells for T3P and tau-1 antibodies were plated with 0.7% bovine serum albumin in phosphatebuffered saline, and control wells for SMI34 samples were plated with 0.2% non-fat dry milk in borate-buffered saline. The color reagent used was 2,2'-azino-di-(3-ethylbenzthiazoline-b-sulfonic acid) and color development was allowed to proceed for up to 60 min. Antibody SMI34 was purchased from Sternberger Monoclonals Inc., and antibody tau-1 was generously provided by Dr. Lester I. Binder. Antibodies tau-14, which recognizes amino acids 141-178 of ~4 4 1 independent of the phosphorylation state of tau (Kosik et al., 1988), and T3P were generously provided by Dr. Virginia Lee. Statistical significance was determined using the Student's t test.
Bovine and porcine tau proteins were applied to 4-20 and 10-20% Daiichi gradient gels, respectively, electrophoresed, electroblotted, and subjected to immunoassay as previously described (Caputo et al., 1992b). was assigned by matching calculated and experimentally derived masses together with limited Edman sequencing ( Table   I). Coverage of the 4-repeat bovine tau sequence from this mapping experiment was approximately 60% (Fig. 2).

Proteolysis, rp-HPLC, and Mass
Two phosphorylated peptides were identified in this bovine tau digest (Table I). One phosphorylated peptide was present in fraction 2 and corresponded to amino acids 391-406 of 4repeat bovine tau (402-417 of human 7441). The mass spectral data for this peptide indicated that it was present only as the monophosphorylated peptide (Fig. 3A). This peptide contains seven potential phosphate acceptor sites (5 serine and 2 threonine residues). To identify the specific residue within this peptide that was phosphorylated, the fraction containing this peptide was subjected to tryptic digestion and reanalyzed by FAB mass spectrometry. A protonated ion of mass 1101.6 Da was observed, corresponding to the nonphosphorylated peptide containing bovine amino acids 396-406. This result indicates that the phosphorylation site was within the first five amino acids, DTSPR, of the parent peptide. The peptide of bovine amino acids 391-406 was dehydrated, derivatized with ethane-thiol to form an ethylcysteine residue, and Edman-microsequenced. The phosphorylation site was assigned to S393 of bovine tau (S404 of human 7441) based on the elution of phenylthiohydantoin-ethylcysteine at the corresponding Edman cycle. The identification of nonphosphorylated peptide 396-406 indicates that S405 of bovine tau (S416 of human 7441) was not phosphorylated in uiuo. This site is known to be phosphorylated in vitro by calcium-calmodulindependent protein kinase (Steiner et aZ., 1992) and CAMPdependent protein kinase (Scott et al., 1992) resulting in a shift in tau mobility upon electrophoresis.
The second phosphorylated peptide was recovered in fraction 10 (Table I) and corresponds to amino acids 182-240 of bovine tau (193-251 of human 7441). Based on mass spectral data, this peptide existed in the nonphosphorylated, monophosphorylated, and diphosphorylated states (Fig. 4). Too little sample remained of this peptide to allow precise assignment of the phosphorylated residues. No other peptide identified in the bovine digest was phosphorylated (Table I), including the peptide containing amino acids 376-390 of bovine tau (387-401 of human 7441). This peptide was isolated in fraction 5 and contains the sequence KSPV. The mass spectrum for this peptide is shown in Fig. 3C.
Porcine tau was also digested with asp-N and the digestion products were separated by rp-HPLC and analyzed by mass  Fraction numbers refer to the elution position on rp-HPLC (Fig.  1). The numbers in parentheses refer to the amino acid positions in bovine tau (Fig. 2). The molecular ion observed in fraction 1 could not be assigned. Edman sequencing of this fraction did not yield any data. DRSGYSSPGSPGTPGSspectrometry and limited Edman sequencing. The sequence coverage was 50% of the human 4-repeat (7441) tau sequence. One phosphorylation site was identified, which was located on the peptide corresponding to amino acids 402-417 of human 7441 (Table 11). The monophosphorylated and nonphosphorylated forms of this peptide were detected (Fig. 3B). A serine in porcine tau homologous with S404 of human 7441 was determined to be the residue phosphorylated in this peptide based on the analysis of the tryptic digest of this peptide by FAB mass spectrometry, as well as the Edman microsequencing results of the ethane-thiol-derivatized peptide as described for bovine tau, above. The peptide containing amino acids 394-408 of human 7441 which contains the KSPV motif was recovered in only the nonphosphorylated form as was the case with bovine tau. No peptide containing the sequence corresponding to human 7441 amino acids 193-251  -
In addition, a peptide from the digest of porcine tau corresponding to amino acids 358-379 (human 7441) was detected. This peptide ends in arginine and was therefore not expected to be produced by asp-N cleavage. Whether this carboxylterminally truncated species was generated during isolation or represents a naturally occurring species is not clear. However, it should be noted that the PHF-tau fragments contain a species which is terminated at E380, 1 residue downstream from this terminus (see below). Also Hasegawa et al. (1992) report that their preparation of PHF-tau contains a carboxylterminally truncated species as determined by a lack of reactivity with carboxyl-terminally specific antibodies BR134 and C6. While it was not possible to quantitate the levels of this species accurately, it represents a minor component (less than 10% of native tau in our sample). There were barely detectable levels of this species in the bovine tau digests as determined by mass spectrometry (data not shown).
Mass Spectrometry of the PHF-Tau Fragment-Affinitypurified F5.5-tau was digested with asp-N protease, and the digestion products were separated by rp-HPLC (Fig. 5). The peptides that were identified in these fractions by mass spectrometry and Edman sequencing covered 95% of a 93-amino acid region from amino acid 299 to 391 of human 7441 (Table   111). Fractions 1-5 contained peptides with overlapping sequences corresponding to amino acids 50-93 of this sequence. Fraction 1 contained a peptide corresponding to amino acids 50-59. A similar peptide was identified in fraction 2, except that it was one mass unit higher, suggesting deamidation of glutamine. This deamidation may have resulted from extraction of the PHF preparation with 88% formic acid to isolate F5.5 tau.
The 93-amino acid region terminates with the sequence DHGAE. The peptides in fractions 3 and 5 ended at residue 82 with the sequence LTFRE. Both of these termini end with glutamic acid and were not predicted from asp-N cleavage.
A peptide was recovered in fraction 6 which contained, based on Edman sequencing, amino acids 1-44 of the 93amino acid region from 4-repeat tau (Table 111). The first seven amino acids, HVPGGGS, are from the insert in the tubulin-binding region of 4-repeat tau. The corresponding sequence in 3-repeat tau, HQPGGGK, was also detected in this fraction. No phosphorylated residues were detected on any of the digestion products from this F5.5 sample.
The second F5.5 preparation was digested with trypsin. The digest was separated into eight fractions by rp-HPLC (Fig.  6). Mass spectrometric analyses and Edman sequencing of these fractions resulted in the identification of peptides that covered 70% of the 93-amino acid region of 4-repeat tau described above (Table IV) digest contained peptides from the amino-terminal region of 3-repeat and 4-repeat tau, respectively. Intact ubiquitin was identified in fraction 8.
The peptides identified in both the asp-N and trypsin digests provided 100% coverage of the 93-amino acid region. Amino acids 299-391 of human 7441 is 100% homologous with amino acids 288-380 of bovine tau, and 81% of this sequence was covered in the analysis of bovine tau and no phosphate was detected. For porcine tau, 97% of the corresponding sequence (241-333 of ~3 8 3 ) was recovered, which was also 100% homologous with human tau and contained no phosphate.
Together these F5.5 data indicate that no phosphorylation sites were detected within the 93-amino acid region identified in the F5.5 fraction. Therefore phosphate-dependent antibodies to PHF-tau must react with epitopes outside of this 93amino acid region of F5.5 tau. Jakes et al. (1991) have previously Edman sequenced F5.5 and identified three fragments. One fragment corresponded to the end of repeat one plus repeats 3 and 4 from 3-repeat tau. The other fragments were derived from 4-repeat tau and corresponded to the end of repeat 2 plus repeats 3 and 4, and the end of repeat one plus repeats 2 and 3. This study confirmed the presence of the first two fragments. The third contains the amino acids 275-298 of 7441. We did not find any evidence of this sequence. In addition two carboxyl termini were identified, corresponding to E380 and E391 of human 7441, which were not predicted based on asp-N cleavage. The presence of ubiquitin in the F5.5 sample that was not affinity-purified confirms previous results from antibody studies (Cole and Timiras, 1987;Grundke-Iqbal et al., 1988) in which ubiquitin was inferred to be associated with PHFs.
Immunoreactivities of Bovine and Porcine tau Proteins-The same bovine and porcine tau protein preparations that were analyzed by mass spectrometry were also probed with antibodies whose immunoreactivities are known to be dependent on the phosphorylation state of tau. First bovine and porcine tau proteins were treated with alkaline phosphatase and electrophoresed to assess whether proteases that might contaminate the phosphatase preparation would degrade tau protein. Slight increases in the mobilities of the various tau bands were observed after incubating with alkaline phosphatase, consistent with dephosphorylation, but no major shifts indicative of proteolysis were observed (Fig. 7A).
Porcine tau proteins, with and without alkaline phosphatase treatment, were assessed for reactivity with antibody T3P. T3P is an antiserum produced against peptide T3 (GAEIVYKSPVVSGD) that was phosphorylated on the serine residue within the KSPV motif (S396 of 7441; Lee et al., 1991). Although this antiserum may detect more than one epitope on peptide T3, it is selective for the phosphorylated form of the peptide (Lee et al., 1991). Substantial T3P reactivity was detected on Western blots of porcine tau (Fig. 7 B ) , even though no phosphate was chemically detected within its proposed epitope (see mass spectrometric data). T3P reactivity was abolished upon alkaline phosphatase treatment, indicating that the T3P immunoreactivity was dependent on the presence of phosphate.
ELISA was used to assess further the reactivities of porcine and bovine tau proteins with several PHF-relevant antibodies. Bovine tau was reactive with T3P by ELISA, and both tau proteins were reactive with monoclonal antibody SMI34 (Figs.  8 and 9). These reactivities were lost upon alkaline phosphatase treatment of tau, indicating that both antibodies reacted with phosphorylated sites on both proteins.
Alkaline phosphatase from E. coli was more effective than alkaline phos- phatase from calf intestine in diminishing the immunoreactivity of porcine tau (Fig. 8). These results indicate that both T3P and SMI34 reacted with bovine and porcine tau proteins in a phosphate-dependent manner in two types of immunoassays. However, the proposed epitopes for both antibodies, which includes the phosphorylated KSPV sequence, was not phosphorylated in either protein sample. On the other hand, the serine homologous with S404 of human 7441 was phosphorylated in each sample.
Antibody tau-1 also reacted with both porcine and bovine tau proteins, although no significant enhancement of reactivity occurred after phosphatase treatment of tau. Several concentrations of tau protein and several dilutions of antibody were used to ensure that any enhancement of immunoreactivity would be detected. Thus the phosphates that inhibit tau-1 binding to PHF-tau appear to be absent from these bovine and porcine tau protein samples.

DISCUSSION
The results of the present study indicate that both bovine and porcine tau proteins are phosphorylated at the serine homologous to human S404. This residue can be phosphorylated in uitro by tau protein kinase (Ishiguro et al., 1991) and mitogen-activated protein kinase (Gustke et al., 1992). Since these tau proteins were not assembled into PHFs, phosphorylation a t 54404 is not by itself adequate for tau to form into PHFs. A study using rat tau suggested this residue is phosphorylated in a developmentally regulated fashion (Kanemaru et al., 1992).
The KSPV sequence was not phosphorylated on bovine or porcine tau. Therefore antibodies T3P and SMI34 could not have reacted with the phosphorylated KSPV site, their assumed epitope, even though both antibodies were shown to

Peptides identified from an Asp-N digest of F5.5 extract
The peptides were assigned to human tau sequences on the basis of mass spectrometry and Edman sequencing. Fraction numbers refer to the elution position on rp-HPLC (Fig. 5). The numbers in parentheses refer to the amino acid positions within the fragment of 4repeat human tau shown below, homologous to amino acids 299-391 of human 1441. Fraction 6 was assigned by Edman sequencing only.  * Peptides assigned using Electrospray mass spectrometry.

Time (min)
FIG. 6. Peptide map of FS.6 extract after digestion with trypsin. 100 pg of F5.5 was digested with trypsin, and the resulting digest was separated by rp-HPLC as described under "Materials and Methods." Numbers indicate fractions that were analyzed by mass spectrometry and Edman sequencing. react with both tau proteins in a phosphate-dependent manner. Instead they may recognize the phosphate on S404. Nonetheless, neither T3P nor SMI34 react with exclusively PHF-specific tau phosphorylation sites.
The phosphate(s) that prevent antibody tau-1 from binding PHF-tau were absent from bovine and porcine tau, based on the pattern of tau-1 reactivity with them. Although several

#MI-84
Tau-1 11800 1rlOOO 1:SOO lrlOOO 111000 tau but the homologous peptide was not observed by mass spectrometric analysis.) These results leave open the possibility that a phosphate on tau that alters tau-1 reactivity is specific for PHF-tau. However, recent studies of human tau expressed in cell cultures suggests that the tau-1 epitope can also be occluded on non-PHF tau. ' Recently, Hasegawa et al. (1992) published the first direct analysis of normal and PHF-tau by mass spectrometry and enzyme sequencing. They found amino acids 396-438 of normal human 7441 to contain one phosphate and to exist in the phosphorylated and nonphosphorylated forms. We have refined the assignment further to S404 of human 7441 (by analogy to porcine and bovine tau) and have shown the degree of phosphorylation to be variable. It has also been shown (Hasegawa et al. 1992) that T231, S235, and S262 are phosphorylation sites specific to PHF-tau and that there is also a phosphorylation site in the tau-1 region, amino acids 191-225. All sites (except S235 which was not covered by mass spectrometric mapping of normal tau) were shown not to be phosphorylated in normal human tau. Our data show that amino acids 193-251 of normal human ~4 4 1 (by homology with bovine and porcine tau proteins) contains two phosphorylation sites, which would agree with the data of Hase- No phosphorylation sites or other post-translational modifications were identified in the tubulin-binding region of PHF-tau (F5.5) in the present study. Our data agrees with and extends the data of Hasegawa et al. (1992) for this region of PHF-tau. This region of tau seems very important for several reasons. Unlike its binding to microtubules, binding of this region to the rest of the PHF structure is very strong, which renders this region of tau resistant to pronase treatment. The present results show that this binding is not due to phosphorylation, suggesting that it may be a property of non-PHF tau as well.
The unusual binding capacity of F5.5 tau may be integral to PHF formation. This almost irreversible binding of F5.5 tau to PHFs may be detrimental to cells, because it may sequester tau in PHFs, rendering it unavailable for its normal cellular functions. This region of tau, but not full length tau, has been shown to assemble into PHF-like fibrils (Wille et al., 1992). This region can also induce polymerization of the carboxyl terminus of @-amyloid precursor protein into PHFlike fibrils (Caputo et al., 1992a).
The absence of post-translational modifications on F5.5 tau indicates that the reactivities of two antibodies specific for F5.5 tau, 6.423 and 728 (Wischik et al., 1988;Caputo et al., 1992b), are not based on recognition of PHF-specific phosphorylation sites. The possibility remains open that they recognize a conformation of F5.5 tau that is not shared by other tau proteins or may recognize the amino or carboxyl terminus.
Phosphorylation at a site outside of the tubulin-binding region of tau may be sufficient to either induce tau assembly into PHFs or to release tau from microtubules, where it may be more susceptible to polymerization. These sites will have to be identified by direct chemical analysis of intact tau from PHFs, as no antibody has been shown to be selective for only a single phosphate site exclusive to PHF-tau. However large quantitites of intact PHF-tau have not been isolated to date in a homogeneous state suitable for mass spectrometric analysis. Analysis of PHF-tau phosphorylation sites may not reveal a single PHF-unique site. Instead the total number of phosphorylation sites may prove to be more important than phosphorylation at any one particular site. Alternatively phosphorylation may occur after PHFs are formed or may be independent of it. Distinguishing among these possibilities will be useful in designing therapeutic approaches to Alzheimer's disease.
This study has identified several phosphorylation sites on non-PHF tau proteins that were phosphorylated in vivo and are not candidates for inducing PHF formation. This study also demonstrates that a 93-amino acid segment of the tubulin-binding region of PHF-tau is not phosphorylated and that ubiquitin is present in the preparation. Finally, several Alzheimer-relevant antibodies were shown to react with non-Alzheimer tau in a phosphate-dependent manner, although probably not at the epitopes they were originally thought to recognize.