Phosphorylation-dependent monoclonal Tau antibodies do not reliably report phosphorylation by extracellular signal-regulated kinase 2 at specific sites.

Analysis of phosphorylation of tau, the microtubule-associated proteins hyperphosphorylated in Alzheimer's disease, is often performed using phosphorylation-sensitive monoclonal antibodies thought to report the presence or absence of one or two specific phosphorylations (cognate sites). Using several such antibodies we found a much more complicated relationship between phosphorylation at specific sites, as monitored by two-dimensional phosphopeptide mapping, and antibody recognition of these sites. Multiple phosphorylation of tau in several stages by the brain extracellular signal-regulated kinase 2 isoform PK40 suggested that phosphorylation at cognate sites is sometimes necessary (but not sufficient) to induce a change of antibody reactivity and in some cases is not even necessary in the background of multiple phosphorylation at other sites. No single phosphorylation site was found to be responsible for any level of gel mobility shift associated with phosphorylation. Tau acquired its maximal gel mobility retardation and final immunochemical profile at substoichiometric phosphorylation of most sites. This suggests that many alternate phosphorylation patterns can produce the same conformational and immunochemical presentation on sodium dodecyl sulfate-gel electrophoresis. Although PK40erk2 prefers some phosphorylation sites, most notably Ser235, followed by Ser199 or Ser202 and Thr205, the phosphorylation of multiple Ser/Thr-Pro sites is not highly sequential. Ser396 is one of the least preferred sites and seems to require prior phosphorylation at Ser404.

Analysis of phosphorylation of tau, the microtubuleassociated proteins hyperphosphorylated in Alzheimer's disease, is often performed using phosphorylation-sensitive monoclonal antibodies thought to report the presence or absence of one or two specific phosphorylations (cognate sites). Using several such antibodies we found a much more complicated relationship between phosphorylation at specific sites, as monitored by two-dimensional phosphopeptide mapping, and antibody recognition of these sites. Multiple phosphorylation of tau in several stages by the brain extracellular signal-regulated kinase 2 isoform PK40 suggested that phosphorylation at cognate sites is sometimes necessary (but not sufficient) to induce a change of antibody reactivity and in some cases is not even necessary in the background of multiple phosphorylation at other sites. No single phosphorylation site was found to be responsible for any level of gel mobility shift associated with phosphorylation. Tau acquired its maximal gel mobility retardation and final immunochemical profile at substoichiometric phosphorylation of most sites. This suggests that many alternate phosphorylation patterns can produce the same conformational and immunochemical presentation on sodium dodecyl sulfate-gel electrophoresis. Although PK40 erk2 prefers some phosphorylation sites, most notably Ser 235 , followed by Ser 199 or Ser 202 and Thr 205 , the phosphorylation of multiple Ser/ Thr-Pro sites is not highly sequential. Ser 396 is one of the least preferred sites and seems to require prior phosphorylation at Ser 404 .
The microtubule-associated tau proteins, normally involved in organizing the neuronal cytoskeleton, are a subject of intense interest because of their involvement in the neurofibrillary degeneration of Alzheimer's disease (AD) 1 (for a review, see Ref. 1). Tau proteins associated with paired helical filaments (PHF) (2)(3)(4) and a substantial fraction of the soluble tau proteins from AD brains (5, 6) exhibit abnormal gel electrophoretic mobilities and abnormal immunochemical properties with phosphorylation-sensitive antibodies, indicative of a pathological hyperphosphorylated state. PHF-tau proteins are in-competent at binding and stabilization of microtubules and are believed to be responsible for the cytoskeletal abnormalities leading to neuronal dysfunction in AD.
Because of a large number of potential phosphorylation sites, a detailed analysis of normal or abnormal phosphorylation patterns is rather difficult. Therefore, the contribution of individual phosphorylation sites to tau function or dysfunction has largely remained obscure. Phosphorylation-sensitive antibodies appear to be convenient tools to assess the phosphorylation state of tau. Phosphorylated Ser/Thr-Pro sites are abundant in many cytoskeletal proteins and are very antigenic. Therefore, most antibodies are sensitive to these phosphorylation sites. This has directed much attention to the role of proline-directed tau phosphorylation, although other phosphorylations are clearly also important (7). Some widely used monoclonal antibodies (mAbs) seem to probe the phosphorylation state of one or at most two Ser-Pro or Thr-Pro sites, inviting the possibility that immunoassays could yield detailed (even quantitative) information about the phosphorylation state of ex vivo tau isolates. The experimental support for such mAb properties stems from site-directed mutagenesis experiments; conversion of single Ser or Thr residues into Ala led to dramatic alterations in immunoreactivity, either before or after phosphorylation in vitro. Monoclonal antibodies believed to indicate the phosphorylation state of individual sites are PHF-1, Tau-1, and AT8 (8 -10) and the neurofilament mAbs SMI31, 33, and 34, cross-reacting with tau due to shared Lys-Ser-Pro motifs (11,12).
There are several potential complications in such mutagenesis analyses. Conclusions are based on the immunochemical behavior of wild type versus mutant protein in very similar phosphorylation states, i.e. either completely unphosphorylated or fully phosphorylated with any of the available kinases. This may reveal that phosphorylation at a single site is necessary for the immunochemical change, although it may not be sufficient. Different results might be obtained if the phosphorylation states of wild type and mutant were very dissimilar (e.g. seemingly contradictory results regarding the cognate phosphorylation sites of SMI31, depending on the tau kinase used in Refs. 11 and 12). Another concern is that any mutation of the tau protein might alter the properties of distant domains in an allosteric manner. In that case any change in immunochemical properties could not be ascribed to the phosphorylation site that was mutagenized.
To complement existing studies of the relationship between tau phosphorylation and recognition by antibodies we used a correlative approach. The longest recombinant human tau isoform htau40 was phosphorylated to increasing levels with activated forms of ERK2. 32 P-Labeled tau populations in various states of phosphorylation were analyzed by quantitative two-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed: Institute for Dementia Research, Bayer Corp., 400 Morgan La., West Haven, CT 06516. 1 The abbreviations used are: AD, Alzheimer's disease; PHF, paired helical filaments; mAb(s), monoclonal antibody(ies); ERK, extracellular signal-regulated protein kinase; PAGE, polyacrylamide gel electrophoresis; PAA, phosphoamino acid analysis. dimensional phosphopeptide mapping. The results from this more direct technique were compared with conclusions derived from generally accepted interpretations of immunochemical data. PK40 erk2 was used in this study because forms of ERK2 phosphorylate tau in vitro to the highest stoichiometric level reported to date (14 -16 mol of PO 4 /mol of tau (13)(14)(15)). This provides an example to investigate the complex interactions of multiple phosphorylations.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant Human Tau and of PK40 erk2 -The longest human 441-amino acid tau isoform was expressed from the htau40 cDNA (gift of Dr. Goedert) in Escherichia coli and purified by heat treatment and Mono S fast protein liquid chromatography as described before (15). PK40 erk2 was purified from bovine brain according to previously published procedures and modifications (13,15).
Tau Phosphorylation-3 g of recombinant htau40 protein was incubated for 4 h at 37°C with increasing amounts of PK40 erk2 ranging from 0.002 to 0.2 g in 30 l of 50 mM HEPES, pH 7.0, 2 mM Mg 2ϩ , 1 mM dithiothreitol, 0.5 mM ATP containing 25 cpm/pmol [␥-32 P]ATP. For preparative purposes, 15 g of htau40 was phosphorylated with 0.2 g of PK40 erk2 under identical conditions. Reactions were terminated with 20 l of 50 mM EDTA or, in preparation for SDS-PAGE, by boiling with 20 l of sample buffer. For determination of phosphorylation stoichiometry, samples stopped with EDTA were applied to GF/C glass filters wetted with 10% trichloroacetic acid, 2% sodium pyrophosphate. After extensive washing with trichloroacetic acid and sodium pyrophosphate, 32 P incorporation was determined by liquid scintillation counting.
SDS-PAGE and Western Blotting-Aliquots of incubation mixtures stopped with sample buffer were separated on 10% SDS-PAGE (Novex) for Coomassie Blue R-250 staining or Western blotting. Gels were slightly overrun to provide for better separation of mobility isoforms of tau. Tau proteins were excised from the gel and prepared for tryptic digests as described below. For Western blotting, sample aliquots corresponding to 0.1-0.5 g of tau were transferred overnight at 4°C, 30 -50 mA onto nitrocellulose (16). Blots were blocked with 5% dry milk in 10 mM phosphate-buffered saline, Triton X-100, pH 7.2 (dilution buffer) and incubated for 2-4 h with mAbs Tau-1, 1:5,000; PHF-1, 1:100; and SMI31, 33, and 34 at 1:1,000 each in 5% dry milk/dilution buffer. Blots were stained with alkaline phosphatase-coupled goat antimouse polyclonal antibody (1:3,000) with a nitro blue tetrazolium staining kit (Life Technologies, Inc.).
Two-dimensional Tryptic Phosphopeptide Mapping-Tau protein bands excised from Coomassie Blue-stained SDS-PAGE were washed twice in 0.5 ml of 10% methanol and once each in 0.5 ml of water and 50 mM NH 4 HCO 3 . Gel slices were minced with a forceps and incubated overnight at 37°C with 10 g of bovine trypsin (Sigma, tosylphenylalanyl chloromethyl ketone-treated) in 100 l of 50 mM NH 4 HCO 3 . The supernatant was removed, and gel slices were extracted twice for several hours with 200 l each of 50 mM NH 4 HCO 3 . Extraction yield of 32 P-peptides was monitored by Cerenkov counting of gel slices and was generally Ͼ95%. Combined slice extracts were concentrated by lyophilization to about 100 l, and 5 g of fresh trypsin was added for a repeated digest over 6 h. Digests were evaporated to dryness and lyophilized three times with each 200 l of water. Residues were taken up in 50 l of peptide electrophoresis buffer (formic acid/acetic acid/ water 22:78:900) and evaporated again. Peptides were dissolved in 2 l of peptide electrophoresis buffer and applied to a 20 ϫ 20-cm cellulose plate (Merck). Peptide maps were developed as described (17). The first dimension was peptide electrophoresis buffer at 1,000 V for 50 min. The second dimension was isobutyric acid/n-butyl alcohol/pyridine/acetic acid/water 1,250:38:96:58:558 (thin layer chromatography buffer) overnight. After drying, plates were scanned in an Ambis radioscanner system for quantitative analysis.
Phosphoamino Acid Analysis (PAA) of Peptides-Spots of interest were scraped from the plates, extracted twice with 200 l of peptide electrophoresis buffer, and concentrated to dryness. Residues were treated with 6 N HCl for 90 min at 110°C. After repeated lyophilization with water and with 50 l of PAA buffer (acetic acid/pyridine/water 50:5:945 (17)) samples were taken up in PAA buffer, and aliquots containing comparable amounts of radioactivity (cpm) were loaded on a 20 ϫ 20-cm cellulose plate together with a mix of the three unlabeled phosphoamino acids. Electrophoresis in PAA buffer was at 1,000 V for 50 min. Plates were scanned, and in some cases quantified, in a Phos-phorImager (Fujix, BAS 1000).
Peptide Sequencing-After recovery from the two-dimensional plates, equivalent peptides (as shown in Fig. 2) were pooled for analysis.
In duplicate experiments, peptides were recovered from a preparative two-dimensional map (15 g of tau). Each peptide was redissolved in 50 l of 10% acetic acid, loaded onto a Hewlett-Packard miniature biphasic column, and washed with 2% trifluoroacetic acid. Sequencing was performed on a Hewlett-Packard G1005A protein sequencer using methods (version 3.0) and reagents of the manufacturer.
Radiosequencing of 32 P-labeled Peptides-32 P-Labeled peptides were covalently coupled to a polyvinylidene difluoride membrane that had been derivatized with arylamine groups (Sequelon-AA; Millipore) in the presence of carbodiimide according to the manufacturer's instructions.

FIG. 1. Correlation of stoichiometry of tau phosphorylation by PK40 erk2 , gel mobility shifts, and immunochemical changes.
Lowest panel, 3 g of recombinant htau40 protein was phosphorylated for 4 h at 37°C with increasing amounts of PK40 in the presence of [␥-32 P]ATP. The amount of PK40 is presented in terms of tau kinase activity (pmol/min) under initial rate conditions. Samples were precipitated by trichloroacetic acid on glass filters and counted for determination of stoichiometry. CB, aliquots of duplicate samples were analyzed on SDS-PAGE for mobility changes. White boxes designate mobility isoforms that were cut out separately from the gel for twodimensional peptide mapping analysis (Fig. 2). Upper panels, aliquots of samples analyzed by SDS-PAGE were subjected to Western blotting with Tau-1, SMI33, SMI34, SMI31, and PHF-1.
Membranes were then washed sequentially with 400 l of water, 1:1 (v/v) methanol/water, methanol, and water again. Sequencing was performed by automated Edman degradation on an Applied Biosystems model 477A protein sequencer with modified cycles to extract the unconverted anilinothiazolinone amino acid derivatives with neat trifluoroacetic acid followed by 1:1 (v/v) ethyl acetate:n-butylchloride containing 0.01% phosphoric acid. Fractions (1 ml) were collected, and radioactivity was determined in a scintillation counter.

Correlations of Phosphorylation Stoichiometry, Gel Mobility,
and Immunoreactivity-After incubation with increasing amounts of the ERK2 form PK40 (13, 15) recombinant tau protein showed a spectrum of changes of conformation (as indicated by gel mobility shifts) and of immunochemical properties. Species with reduced gel mobility became apparent at an average stoichiometry of about 2 mol of PO 4 /mol of tau, whereas species with maximal mobility shift appeared only at a phosphorylation stoichiometry of Ն 7 (Fig. 1). All tau species exhibited maximal gel mobility shift at a stoichiometry of 11.
Tau was probed with mAbs for epitopes that are eliminated by phosphorylation (Tau-1, SMI33) or induced by phosphorylation (PHF-1, SMI31, SMI34). Phosphopeptide and site-directed mutagenesis studies indicated that the reactivity of these mAbs with tau is regulated by one or at most two phosphorylation sites: Ser 199 and Ser 202 for Tau-1 (8), Ser 396 and Ser 404 for PHF-1 (9), Ser 235 for SMI33, Ser 396 for SMI31 (in some instances with a contribution by Ser 404 ) and SMI34 (in conjunction with unidentified phosphorylation sites (11,12)). The SMI mAbs were raised against neurofilaments but crossreact with tau because of the shared sequence motif Lys-Ser-Pro (18).
Assuming that the above epitope assignments were correct, the correlation of the gel mobility state of tau with its immunochemical reactivity in Fig. 1 would suggest the following.
• Tau  To test the validity of these conclusions derived from immunochemical analysis, 32 P-labeled mobility isoforms of tau as resolved by SDS-PAGE (Fig. 1) were separately subjected to tryptic phosphopeptide mapping. The major peptides and their phosphorylation sites were identified. As indicated in Fig. 1, tau species of lowest, intermediate, and highest gel mobility were analyzed. Of particular interest were phosphorylation states where two mobility isoforms coexisted (Fig. 1, 2a/2b, 5a/5b).
Identification of 32 P-labeled Peptides and Assignment of Phosphorylation Sites-32 P-Labeled peptides isolated from two-dimensional peptide maps were identified by Edman sequencing. Assignment of phosphorylation sites (Table I) was by a combination of the following criteria.  (17). 5. Product-precursor relationships between related peptides. 6. Radiosequencing in some cases. 7. Trypsin resistance due to phosphorylation and consequent deviations from the expected electrophoretic mobility due to charges resulting from additional Lys or Arg residues.
In some cases criterion 1 did not suggest phosphorylation clearly, possibly due to contamination; however, phosphorylation at the site in question was accepted when all other criteria combined strongly suggested phosphorylation.
Comparative Analysis of Immunochemical Properties and Phosphopeptides of Tau at Various Stages of Phosphorylation-Superficial examination of the two-dimensional map of high mobility tau species (Fig. 2, panel 2a) showed that most phosphopeptides were already present at some level. Only peptide 10b (Ser 396 /Ser 404 phosphorylation, Table I) was restricted to a Assignment of underlined residues is tentative; amino acids in parentheses indicate unusually low recovery, presumably due to contaminating sequences.
b Initial coupling yield extrapolated back to cycle 0. c Underlined residues indicate that assignment is based on inference from a combination of the following criteria: relative migration properties, phosphoamino acid analysis, ERK2 consensus sites, abnormal tryptic cleavage.
d Peptides tend to colocalize on two-dimensional peptide maps. e Phosphoserine signals are from unidentified contaminant peptides (not reproducibly observed with peptide 5).
tau species with at least a partial gel mobility shift (Fig. 2,  panel 5a). However, no single phosphorylation site fulfilled all criteria to be solely responsible for any level of gel mobility alteration: (i) exclusive presence in a lower mobility isoform, and (ii) stoichiometric presence in reference to the most prominent peptide 1 (Ser 235 phosphorylation). Peptide 1 clearly contained the most preferred site of tau phosphorylation by PK40 erk2 . According to criteria 1-3, 6, and 7 (Figs. 2, 3, 4, and 5), Ser 235 was assigned as the phosphorylation site. At an average stoichiometry of 2 PO 4 /mol of tau, Ser 235 (Fig. 1) was already phosphorylated maximally (Fig. 6), and it also remained the most highly phosphorylated site of tau up to a stoichiometry of 12. Because of this fortuitous property, peptide 1 served as an internal reference to assess accurately the level of phosphorylation of all other sites. Comparison of immunochemical properties of tau species with their phosphopeptide maps showed that Ser 235 phosphorylation (see Fig.  6) and elimination of SMI33 immunoreactivity (Fig. 1) were not correlated. SMI33 reactivity was still strong when Ser 235 was already phosphorylated maximally. Elimination of this epitope clearly required phosphorylation at other sites.
Peptide 2 was the second most prominent peptide at low phosphorylation stoichiometries. Because of low recovery, se-quence information was sparse, but criteria 1-3 suggested phosphorylation at Ser 199 or Ser 202 , and Thr 205 , the sites relevant for mAb Tau-1. This assignment was supported further by relationships with triphosphorylated peptide 8, which contained the same phosphorylation sites: a product-precursor relationship evident in some experiments, and the diagonal migration shift (17) relative to the peptide 8 (criteria 4 and 5). Densitometric analysis of peptides 2 and 8 in Fig. 6 showed that phosphorylation at Ser 199 and Ser 202 did not reach the theoretical level in reference to Ser 235 phosphorylation; maximally only about 40% of the tau species became phosphorylated on both Ser 199 and Ser 202 and another 25% at either Ser 199 or Ser 202 . Nevertheless, the Tau-1 epitope was abolished completely (Fig. 1), suggesting the involvement of other phosphorylation sites.
Peptides 3b and 10b were relevant for the epitopes of PHF-1, SMI 31, and SMI 34. Although 3b colocalized with unrelated peptide 3a (see below), criteria 1-3, 6, and 7 identified it clearly as the Ser 404 monophosphorylated peptide cleaved between Lys 395 and unphosphorylated Ser 396 . It is prominent already at low stoichiometry of phosphorylation, in contrast to the exclusively Thr-phosphorylated peptide 3a (Fig. 3). Peptide 10b is the Ser 396 /Ser 404 -diphosphorylated homolog of peptide 3b ac- Two-dimensional phosphopeptide mapping of 32 P-labeled tau mobility isoforms. Mobility isoforms of PK40 phosphorylated tau isolated from SDS-PAGE as designated in Fig. 1 were subjected to tryptic digestion in the gel slice, and the resulting peptides were separated by two-dimensional peptide mapping (17). The maps shown are a typical representation of three similar experiments. Insets are from a different experiment, showing the separation of peptides 10a and 10b, which was not always achieved. Spot 12 was not reproducibly observed. In other experiments spot 2 was eliminated again at higher stoichiometries, supporting a product-precursor relationship with peptide 8. Not all minor spots exclusive to highly phosphorylated tau (panel 6) were identified because of low recovery. However, one of these proved to be a Ser 235 -phosphorylated peptide with the same NH 2 -terminal sequence as peptide 1, possibly indicating increased proteolytic resistance of highly phosphorylated tau. This peptide appeared to be formed at the expense of peptide 1 (see Fig. 6).
cording to criteria 1-3 and 7 (Figs. 4 and 5). Ser 404 phosphorylation is apparently required as a prerequisite for Ser 396 phosphorylation (criterion 5), since Ser 396 monophosphorylated peptides were not recovered. Criterion 4 is not applicable, since the migration relationship between peptides 3b and 10b is atypical due to trypsin resistance induced by Ser 396 phosphorylation. Ser 396 /Ser 404 diphosphorylation occurred only at relatively high tau phosphorylation stoichiometries associated with partial gel mobility shift (Fig. 2). However, it was not coincident with induction of PHF-1, SMI 31, or SMI 34 immunoreactivity which required full gel mobility retardation (Fig.  1). The SMI mAbs required even higher phosphorylation levels of low mobility tau than PHF-1.
Other ERK2 Phosphorylation Sites of Tau-Phosphorylation of Thr 50 (peptide 5), Thr 69 (peptide 11), and Thr 153 (peptide 7) was deduced from criteria 1-3, except for peptide 5, where sequencing through the inferred site Thr 50 was not achieved. Lack of phosphorylation of the ESP motif in peptide 5 is consistent with previous data on ERK consensus sites (20).
Thr 212 phosphorylation in peptide 3a was identified according to criteria 1-3, 6, and 7. However, assignment of Thr 217 phosphorylation in peptide 3a rests on criterion 4 (migration relationship to the homologous peptide 6), since contamination of the sequence data by the more abundant peptide 3b and a trypsin-derived peptide interfered with the interpretation of Edman cycle 8 (Figs. 4 and 5). In the homologous peptide 6 Thr 212 phosphorylation was suggested by criteria 1-5 and 7 in spite of a weak Thr signal in the relevant Edman cycle (Fig. 4). Thr 217 was clearly not phosphorylated since it was recovered in the Edman sequence.
Finally, permutations of phosphorylation patterns involving the motif KT 175 PPAPKT 181 P complicated the assignment of peptides 4, 9, and 10a. Criterion 7 but not criterion 4 was important for the analysis of the phosphorylation state of this motif, since tryptic cleavage was altered by phosphorylation on each site. Peptide 4, tentatively identified by criteria 1-3 and 7 (Table I), occurred only at low levels of tau phosphorylation (Յ3 mol of PO 4 /mol of tau) and is derived from tau species phosphorylated only at KT 181 P. At higher levels of tau phosphorylation this peptide was eliminated ( Fig. 2; criterion 5) in favor of the Thr 175 /Thr 181 -diphosphorylated peptide 10a (criteria 1-3 and 7). Peptide 9 persisted up to the highest levels of phosphorylation (Fig. 2) and represented tau phosphorylation only at KT 175 P, but not KT 181 P, by criteria 1-3 and 7.

DISCUSSION
Phosphorylation-dependent monoclonal antibodies are frequently used to assess the phosphorylation state of specific sites of tau from cell cultures or brain. We tested whether immunochemical analysis of tau with several established Ser/ Thr-Pro phosphorylation-dependent mAbs reliably reflects the phosphorylation state of specific sites in a background of multiple phosphorylations. For the first time a correlational analysis was used taking advantage of the ability of ERK2 to FIG. 3. Phosphoamino acid analysis of peptides isolated from two-dimensional phosphopeptide maps. 32 P-Labeled peptides identified on the two-dimensional maps of Fig. 2 were recovered and subjected to PAA (17). Upper panel, PAA of peptides obtained from highly phosphorylated tau (12 mol of PO 4 /mol). Quantitative densitometric analysis yielded a higher Ser/Thr ratio for peptide 8 over peptide 2, which assisted in identification as tri-and diphosphorylated forms of the same peptide. Lower panel, PAA of spots 3 and 5, obtained from tau phosphorylated to a lower stoichiometry (4 -5 mol of PO 4 /mol).
FIG. 4. Amino acid yields from Edman sequencing of phosphorylated tau peptides. 32 P-Labeled tryptic peptides 1-11 eluted from two-dimensional maps were directly subjected to NH 2 -terminal sequencing. Data of peptides with relatively low recovery (peptides 2, 4, and 5) and many tentative assignments are not shown (see Table I). Ser or Thr residues in parentheses indicate that these amino acids are probably either not or only partially derived from the sequence presented, e.g. in peptide 8 the Ser signal in cycle 5 is believed to be derived from a contamination, as Ser was also elevated in cycles 6,9,15, and 17, and other evidence supports phosphorylation of Ser 199 . Rarely, trypsin fragments were a source of contamination, e.g. in peptide 3 a Thr signal in critical cycle 8 was contributed by a trypsin fragment as well as by peptide 3b. phosphorylate a higher number of Ser/Thr-Pro sites than any other single tau kinase examined so far. An overview of the findings is provided in Table II.
Phosphorylation of recombinant tau by the ERK2 form PK40 was not a sequential process, although some sites were preferred. Rather, many sites were phosphorylated concomitantly. Although this does not necessarily predict a similar lack of specificity in biological systems, it fits well into the notion of enormous heterogeneity of tau phosphoisoforms (21) and partial phosphorylation of virtually all Ser/Thr-Pro sites in rat fetal tau to varying degrees. 2 In spite of this general trend, the motif KS 235 P was clearly preferred. This site is believed to be recognized in its unphosphorylated state by the neurofilament mAb SMI33 (11,12). In our correlational study, however, complete phosphorylation of Ser 235 did not suffice to eliminate the SMI33 epitope. Thus, an unphosphorylated KS 235 P motif is not necessary for mAb binding, and its phosphorylation may be necessary but not sufficient for epitope elimination. Ser 199 / Ser 202 in conjunction with Thr 205 , a site previously implied in Tau-1 and AT8 recognition (10,22), was the second preferred phosphorylation motif of PK40 erk2 . Nevertheless, both Ser 199 and Ser 202 remained partially unphosphorylated in some otherwise highly phosphorylated tau species, although the Tau-1 epitope was abolished. Thus, phosphorylation at Ser 199 /Ser 202 is not strictly required for epitope elimination, in contrast to earlier suggestions based on site-directed mutagenesis studies (8). The complex relationship between the Tau-1 epitope and phosphorylation has also been indicated by a study with multiple tau phosphopeptides (23).
The two examples of mAbs with epitopes eliminated by phosphorylation (Tau-1, SMI33) suggest that interpretations of immunochemical data with this kind of antibody are particularly problematic. Conclusions about the phosphorylation state of specific sites could even be qualitatively incorrect. The failure of SMI33 to report reliably the phosphorylation state of the KS 235 P motif is particularly compelling; this motif is likely to be directly involved in the binding of mAb SMI33 since it is the only shared sequence motif between tau and neurofilaments, the original antigen of the SMI series of mAbs.
Monoclonal antibodies SMI31/34 and PHF-1 represent cases where the epitope is induced by phosphorylation. Previous work based on site-directed mutagenesis and phosphopeptide studies had defined one or two specific phosphorylation sites as sole determinants of the epitopes (9,11,12). However, in this study phosphorylation at those specific sites could be identified in tau species not recognized by these mAbs. The discrepancy was more pronounced with the SMI mAbs than with PHF-1. Hence, phosphorylation at these specific sites may at best be necessary but is not sufficient for epitope formation. However, immunostaining by this type of mAb (epitope induced by phos-phorylation) may yield at least qualitatively correct information, since epitope formation has only been observed in conjunction with phosphorylation at the determinant sites. On the other hand, the absence of immunostaining is uninterpretable. Furthermore, a linear relationship between epitope formation and phosphorylation of specific sites should not be assumed.
A new phosphorylation-dependent mAb, AP422, has recently been described to report the phosphorylation state of the DS 422 P motif in tau (24). Interestingly, only mitogen-activated protein kinase, but not cdk5 or GSK-3, could induce recombinant tau immunoreactivity with this mAb in vitro. However, Ser 422 is preceded by an acidic residue and is therefore not necessarily a favored phosphorylation site for ERK2 (20). In our study neither ES 46 P nor DS 422 P phosphorylated peptides were recovered, although PK40 was able to induce the AP422 epitope strongly. Thus Ser 422 is either only a minor site for 2 H. M. Roder and F. J. Hoffman, manuscript in preparation.
FIG. 5. Radiosequencing of spots 1 and 3. 32 P-Labeled peptides in spots 1 (left panel) and 3 were eluted from twodimensional maps and directly subjected to Edman degradation. Radioactivity of eluates from each cycle was assessed by liquid scintillation counting. Cycle 0 is a wash cycle. According to indirect criteria, peptide 3a (the minor constituent compared with peptide 3b) should have yielded a signal in cycles 3 and 8. Since the signal at cycle 3 is only 2-fold over background, the expected signal at cycle 8 may have escaped detection because of declining Edman degradation yields.
FIG. 6. Relative stoichiometry of tau phosphorylation at specific sites as a function of exposure to increasing amounts of PK40 erk2 (amount defined as activity, see Fig. 1). Panel A, recovery of the Ser 235 -phosphorylated peptide on two-dimensional maps, preferentially formed by PK40 erk2 and assumed to represent a maximal phosphorylation stoichiometry of 1. The slight decrease at higher phosphorylation levels is complementary to formation of a minor peptide with the same NH 2 terminus and site of phosphorylation. Panel B, levels of formation of a Ser 199 or Ser 202 /Thr 205 -diphosphorylated peptide and a Ser 199 /Ser 202 /Thr 205 -triphosphorylated peptide relative to the maximal yield of the Ser 235 phosphopeptide. Panel C, combined relative yields of Ser 404 and Thr 212 /Thr 217 -phosphorylated peptides. Panel D, relative yield of a Ser 396 /Ser 404 -diphosphorylated peptide. ERK2, or phosphorylation at this site occurs only at stoichiometric ratios beyond 12.
The precise nature of the relationship between the actual phosphorylation state of tau and the phosphorylation-dependent presentation of epitopes is of importance for the analysis of normal and pathological tau phosphorylation states ex vivo. Immunochemical profiles are often used to evaluate potential biological model systems for PHF-tau-like phosphorylation (e.g. the similar immunochemical properties of tau proteins from fetal brains and AD brains has led to the suggestion that similar neuronal phosphorylation environments exist in these two situations). More direct analyses of tau phosphorylation states are often cumbersome and fraught with technical difficulties, depending on the source of the protein. The present study suggests that the relationship between tau phosphorylation and immunochemical properties is sufficiently complex to place some limits on the use of tau as a reporter protein for physiological phosphorylation environments. An additional restriction arises from evidence that phosphorylation at sites other than Ser/Thr-Pro either modulates proline-directed phosphorylation or alters the phosphorylation/epitope relationship (25)(26)(27). PHF-like electrophoretic mobility alterations are also not completely unambiguous; tau species with most Ser/Thr-Pro sites only partially phosphorylated can display full retardation in electrophoretic mobility. Consequently, many alternative phosphorylation patterns must exist which translate into maximally reduced electrophoretic mobility.
In summary, the analysis of tau as a reporter protein for AD-like phosphorylation events is valuable in providing exclusion criteria for model systems, if carried out carefully. A more detailed understanding of normal and pathological neuronal phosphorylation homeostasis will require a comparative analysis of various cell biological circumstances that fulfill the PHF-tau criteria but also match other criteria, such as biological context or regulatory features of kinases/phosphatases in adult central nervous system neurons.
a When PK40 is used as the tau kinase, mutation of Ser 396 to Ala alone is sufficient to eliminate SMI31 reactivity (12). With crude brain extract, Ser 404 has to be mutated also (11). b A second determinant distant to Ser 396 (and Ser 404 ) had already been implicated in tau mutagenesis studies (11,12).