Phosphorylation of Microtubule-associated Proteins MAP2 and MAP4 by the Protein Kinase p110 mark PHOSPHORYLATION SITES AND REGULATION OF MICROTUBULE DYNAMICS*

The phosphorylation of microtubule-associated pro- teins (MAPs) is thought to be a key factor in the regulation of microtubule stability. We have shown recently that a novel protein kinase, termed p110 microtubule-affinity regulating kinase (“MARK”), phosphorylates mi- crotubule-associated protein tau at the K X GS motifs in the region of internal repeats and causes the detach- ment of tau from microtubules (Drewes, G., Trinczek, B., Illenberger, S., Biernat, J., Schmitt-Ulms, G., Meyer, H. E., Mandelkow, E.-M., and Mandelkow, E. (1995) J. Biol. Chem. 270, 7679–7688). Here we show that p110 mark phosphorylates analogous K X GS sites in the microtu- bule binding domains of the neuronal MAP2 and the ubiquitous MAP4. Phosphorylation in vitro leads to the dissociation of MAP2 and MAP4 from microtubules and to a pronounced increase in dynamic instability. Thus the phosphorylation of the repeated motifs in the microtubule binding domains of MAPs by p110 mark might pro-vide a mechanism for the regulation

The phosphorylation of microtubule-associated proteins (MAPs) is thought to be a key factor in the regulation of microtubule stability. We have shown recently that a novel protein kinase, termed p110 microtubuleaffinity regulating kinase ("MARK"), phosphorylates microtubule-associated protein tau at the KXGS motifs in the region of internal repeats and causes the detachment of tau from microtubules (Drewes, G., Trinczek In living cells, microtubules undergo transitions between stable and dynamic states. They are organized into stable cytoskeletal structures such as the processes of neuronal cells or the axonemes of cilia and flagella, but are also key players in dynamic events during cell morphogenesis or chromosome partitioning at mitosis. Microtubule stability is thought to be modulated by a variety of post-translational modifications of both tubulin and MAPs. 1 Structural MAPs are filamentous proteins which bind to microtubules in a nucleotide-insensitive way, forming elongated projections from the microtubule surface (for reviews, see Olmsted (1991), Hirokawa (1994), Schoenfeld and Obar (1994), and Mandelkow and Mandelkow (1995)). MAPs can control microtubule dynamics in vitro and in vivo (Drechsel et al., 1992;Pryer et al., 1992;Umeyama et al., 1993;Gustke et al., 1994;Brandt et al., 1994;Dhamodharan and Wadsworth, 1995;Trinczek et al., 1995). Tau and MAP2 are the most studied MAPs in the vertebrate nervous system; tau is abundant in the axon, whereas MAP2 is localized predominantly in dendrites (Binder et al., 1985;Riederer and Matus, 1985). MAP4 is not limited to the nervous system and is the predominant MAP in many types of cells and tissues (Bulinski and Borisy, 1980;Parysek et al., 1984;Aizawa et al., 1990). MAP2, tau, and MAP4 are grossly similar in domain structure, having N-terminal projection domains and C-terminal microtubule binding domains (Lee et al., 1988;Lewis et al., 1988;West et al., 1991;Chapin and Bulinski, 1991). The C-terminal part of these proteins displays considerable homology in a repeated sequence motif. The sequences in the Cterminal region are rich in basic amino acids which probably interact with the acidic sequence in the C terminus of tubulin (Littauer et al., 1986).
Several lines of evidence suggest that the binding of MAPs to microtubules is regulated by phosphorylation. MAPs isolated from tissue or cells are phosphoproteins (Sloboda et al., 1975;Vallee, 1980;Burns et al., 1984;Tsuyama et al., 1986;Brugg and Matus, 1991;Watanabe et al., 1993), MAPs are good substrates for many protein kinases in vitro (Theurkauf and Vallee, 1983;Lindwall and Cole, 1984;Mori et al., 1991;Drewes et al., 1992), and phosphorylation interferes with their microtubule stabilizing capacity (Brugg and Matus, 1991;Shiina et al., 1992;Drechsel et al., 1992;Biernat et al., 1993;Brandt et al., 1994;Ookata et al., 1995;Trinczek et al., 1995). In the case of tau protein, phosphorylation has been extensively studied, because aberrantly phosphorylated tau is involved in the neurofibrillar pathology of Alzheimer's disease (reviewed by Goedert (1993), Mandelkow and Mandelkow (1993), and Trojanowski and Lee (1994)). However, it has been difficult to establish the relationship between protein kinases, phosphorylation sites, and their effect on microtubule affinity, nucleation, and dynamic instability. Recently, we have used an approach which combined site-directed mutagenesis of recombinant tau and in vitro phosphorylation by a brain tissue extract to identify sites that are crucial for microtubule binding . We found that phosphorylation of tau at a single serine residue, located within the sequence KIGS 262 )) in the first repeat of the binding domain, strongly suppresses microtubule binding . The phosphorylation of sites outside the microtubule binding domain, which occurred mostly on Ser/Thr-Pro motifs, had a relatively weak effect. Subsequently, we characterized and purified from brain tissue a novel kinase of molecular mass 110 kDa, which effectively phosphorylated Ser 262 and displayed a pronounced specificity for all four KXGS motifs in the repeat domain of tau (Drewes et al., 1995). This kinase efficiently caused the loss of tau's affinity for microtubules, resulting in high dynamic instability, and was termed p110 mark (microtubule affinity regulating kinase). In this paper, we show that p110 mark phosphorylates MAP2 and MAP4 efficiently on their microtubule binding domains in vitro, and that the KXGS motifs within the conserved repeats are the major phosphorylation sites. Both MAP2 and MAP4 become detached from microtubules upon phosphorylation by the kinase, and the microtubules become unstable. The data suggest that phosphorylation of MAPs by p110 mark could be generally important in the MAP-mediated regulation of the dynamics and rearrangement of the microtubule network in cells.

MATERIALS AND METHODS
Proteins-A cDNA clone of the rat juvenile MAP2 isoform, MAP2c, was a gift of C. Garner (Kindler et al., 1990). Three point mutants of MAP2c were constructed: MAP2cA319, MAP2cA350, and MAP2-cA319ϩ350, in which serines at position 319 and/or 350 were mutated into alanine (corresponding to positions 1682 and 1713 in the full MAP2 numbering, Table I). Human tau cDNA clones were a gift of M. Goedert (Goedert et al., 1989). The numbering used here refers to the rat sequence of full-length MAP2 (1830 residues) and the biggest human tau isoform (clone htau40, 441 residues). The MAP4 construct MAP4-BDC was derived from a murine MAP4 clone (West et al., 1991;Olson et al., 1995) and comprises the microtubule binding region, from residue 640 to the C terminus, and carries an N-terminal hemagglutinin tag sequence (Field et al., 1988). The numbering used here refers to the full-length murine MAP4 sequence (1125 residues). Proteins were obtained either by expressing constructs in Escherichia coli using variants of the pET expression vector (Studier et al., 1990), or by purification from tissues. Brain MAP2 was prepared from porcine brain microtubule protein by heat treatment, Mono S FPLC (Pharmacia Biotech Inc.), and gel filtration as described by Wille et al. (1992). MAP4 was prepared from a mouse heart and lung tissue extract by ammonium sulfate precipitation, heat treatment, Mono S FPLC, and hydrophobic interaction chromatography on Phenyl-Superose (modified after Aizawa et al. (1989)). Phosphocellulose-purified tubulin was prepared from porcine brain following Mandelkow et al. (1985). The protein kinase p110 mark was prepared from porcine brain as described recently (Drewes et al., 1995). Using 50 M MAP2c as substrate and 1 mM ATP at 37°C, the preparation was determined to have an activity of 33 milliunits/ml by the phosphocellulose paper assay (1 unit is defined by the transfer of 1 mol of phosphate/min at 1 mM ATP at 37°C).
Phosphopeptide Mapping-Following phosphorylation reactions, the kinase was removed by boiling the samples in 0.5 M NaCl, 10 mM dithiothreitol and centrifugation. The heat-stable MAPs were precipitated with 15% trichloroacetic acid, cysteine residues were modified by performic acid treatment and the protein was digested overnight with trypsin (Promega, sequencing grade) using two additions of enzyme in a ratio of 1:20 (w/w). Two-dimensional phosphopeptide mapping by thin layer electrophoresis/chromatography was performed on thin layer cellulose plates (Macherey & Nagel, Dü ren, Federal Republic of Germany) according to Boyle et al. (1991). For identification of spots, the digests were run along with the HPLC-purified phosphopeptides (see below). For the mapping of phosphorylation sites by sequencing, recombinant MAP4 fragment or MAP2c (200 g) was phosphorylated with p110 mark and [␥-32 P]ATP (100 Ci/mol) for 2 h, the reaction terminated by a brief heat treatment, and cysteines oxidized with performic acid. After gel filtration on a Pharmacia "Fast Desalting" column ("Smart System") in 10 mM ammonium bicarbonate, pH 8.5, containing 0.1 mM CaCl 2 , incorporated radioactivity was measured by Cerenkov counting and the labeled protein was digested with trypsin (1:40). Separation of peptides was performed by two successive HPLC runs on a RPC C2/C18 SC 2.1/10 column (Smart System, Pharmacia). For peptides which did not bind efficiently to this column, we employed a Vydac 218TP52 column (The Separations Group, Hesperia, CA). The digest was fractionated by HPLC using a gradient of acetonitrile in 10 mM ammonium acetate (flow rate 0.1 ml/min, 0 -25% in 120 min, 25-50% in 20 min). Flowthrough fractions and radioactive peaks from this gradient were further purified using a gradient of acetonitrile in trifluoroacetic acid (flow rate 0.1 ml/min, 0% acetonitrile, 0.075% trifluoroacetic acid to 66% acetonitrile, 0.05% trifluoroacetic acid in 60 min). Sequence analysis of peptides was performed using a 476A pulsed liquid phase sequencer and a 120A on-line phenylthiohydantoin-derivative analyzer (Applied Biosystems). Phosphoserines were identified as the dithiothreitol adduct of dehydroalanine by gas phase sequencing (Meyer et al., 1993).
Assay of Time Resolved Microtubule Length Distribution by Video Microscopy-Video microscopy of microtubules was done essentially as described (Trinczek et al., 1993). Briefly, 10 M phosphocellulose-purified porcine brain tubulin and MAP2 (1 M), MAP4 (1 M), or MAP2c (2 M) were mixed in 50 mM sodium-Pipes, pH 6.9, containing 3 mM MgCl 2 , 2 mM EGTA, 1 mM GTP, and 1 mM dithiothreitol. 1.0 l of the samples was put on a slide, covered with an 18 ϫ 18-mm coverslip, sealed, and warmed up to 37°C in a temperature-controlled air flow within 5 s. A constant temperature of 37°C was maintained by the air flow. The self-nucleated microtubules were recorded at time 2.5, 5, 10, 15, 20, 25, and 30 min after the temperature shift. For each condition and time, three sequential fields of a sample were scored spaced 10 s apart. Five to 10 experiments were analyzed, and the lengths of 500 -800 microtubules were measured. Only those microtubules which were clearly located within the focal plane were included in the data set. The depth of solution was 3-4 m, and the focal depth was 1-2 m. The microtubule number concentrations as dependent on the presence of MAP2c and its mutants (unphosphorylated or phosphorylated) were measured in the same buffer as above with 2 M MAP and 50 M tubulin by counting the microtubules per monitor field (117 m ϫ 86 m) 2-10 min after the temperature shift.
Binding Assays-The binding of MAPs to microtubules was determined as described previously (Gustke et al., , 1994. Briefly, microtubules were assembled and stabilized by taxol, MAPs were added in varying amounts, the bound and unbound fractions were separated by pelleting, and proteins were quantified by scanning of the bands in SDS gels stained with Coomassie Brilliant Blue. Sequence Comparisons and Terminology-In order to compare the three MAPs it is useful to describe their domains with a common nomenclature (Table 1, Fig. 1). The MAPs differ in size but contain regions of homologous sequences, and they have similar gross characteristics. Each of these MAPs has an acidic N-terminal region, followed by a basic region (containing the repeats), and a short acidic or neutral tail. The interaction with microtubules lies in the basic region. One can broadly distinguish between an N-terminal "projection" region and a C-terminal "assembly" region. This distinction is based on proteolytic cleavage which leaves the assembly domain attached to the microtubule wall while the projection domain is released (Murphy and Borisy, 1975;Vallee, 1980). A finer subdivision can be derived from the sequences, as follows.
In MAP2, the acidic N-terminal region extends from Met 1 to Ala 1423 , the basic region is Arg 1424 -Ser 1789 , and the neutral C-terminal tail is from Ser 1790 to Leu 1830 . The basic region contains a proline-rich domain (Leu 1548 -Leu 1663 ) and the repeats (Arg 1664 -Ser 1789 ); it also contains cleavage sites for trypsin (behind Lys 1528 and Arg 1664 , Wille et al. (1992) and thrombin (behind Arg 1630 , Ainsztein and Purich (1994)) which roughly separate projection and assembly domains. In MAP2c, the region Asp 152 -Thr 1514 (1363 residues) is spliced out. Most of the insert has acidic character, except for the last 90 residues (Arg 1425 -Thr 1514 ), and thus can be regarded as a large extension of the acidic N-terminal region (from 151 residues in MAP2c to 1424 in MAP2).
In MAP4, there is an acidic N-terminal domain (Met 1 -Thr 658 , includ-ing the acidic "KDM" domain Thr 243 -Lys 566 , a basic proline-rich domain (Asn 659 -Arg 895 ) which can be subdivided into "P" (Asn 659 -Ala 730 , prolinerich) and "SP" (Thr 731 -Arg 895 , rich in Ser-Pro motifs), the basic repeats (Ala 897 -Gly 1090 ), and an acidic C-terminal tail (A1091-I1125). Repeats "1a" and "2" can be absent due to alternative splicing (Chapin et al., 1995). The repeats are the most striking aspects of the three MAPs. They are typically 31-32 residues long, and are similar to one another within one MAP and between different MAPs. They can be subdivided into about Ϸ13 residues of lower homology (sometimes referred to as the linker region or inter-repeat region), and the Ϸ18 C-terminal residues of higher homology, the repeats proper, mostly ending with a PGGGX motif. The boundaries between the repeats can be chosen in different ways; we prefer the alignment shown in Table I because in this case the "second" repeat of tau coincides exactly with one alternatively spliced exon (number 10, Val 275 -Ser 305 ). The MAPs first cloned or sequenced contained three repeats (for tau, see Lee et al. (1988), for MAP2, Lewis et al. (1988) and Kindler et al. (1990), for MAP4, see Aizawa et al. (1989)). Later, other isoforms were found which contained four repeats (tau, Goedert et al. (1989) and Himmler et al. (1989); MAP2, Doll et al. (1993); MAP4, West et al. (1991) and Chapin and Bulinski (1991)). In addition it was realized that other stretches in the repeat domain were repeat-like, albeit with even lower homology (e.g. the 38-residue repeat following repeat "1" in MAP4 and the 32-residue repeat following repeat "4" in tau, Chapin and Bulinski (1992)). In order to unify the nomenclature we will denote the "classical" repeats (containing the higher degree of homology) as 1, 2, 3, and 4. One of these, 2, may be absent due to alternative mRNA splicing. The low homology repeat of MAP4 will be called 1a. The repeat following 4 will be "4a." Thus Table I shows 6 repeats, 5 of which are common to all three MAPs (1-4 and 4a), and 1a is specific for MAP4 (note that 4a of MAP4 shows less homology than the corresponding tau and MAP2 sequences). These repeats correspond to repeats 1-6 in Chapin and Bulinski (1992), and 4a was called RЈ in Gustke et al. (1994). The four classical repeats all contain a motif KXGS in repeats 1-4, with X ϭ Ile, Cys, or Val (the minor exception is KCVS in repeat 2 of MAP4). In repeat 1a of MAP4 a similar motif is KAAGS. Repeat 4a contains motifs KTDH (tau), RVDH (MAP2), or AGEE (MAP4) in equivalent positions. Although formally not homologous to KXGS, they have the character of "constitutively phosphorylated" KXGS.

TABLE I Sequence comparison of tau, MAP2, and MAP4
The sequence "tau" is that of the largest tau isoform in human CNS with 441 residues (htau40, Goedert et al., 1989). The sequence "MAP2c" is that of the juvenile MAP2c from rat (467 residues, Kindler et al. (1990)). The numbering follows that of the adult rat MAP2b which has 1830 residues and differs from MAP2c by a 1363 residue insert after residue 151 (position indicated by *). The sequence "MAP4" is that of murine MAP4 (1125 residues, West et al. (1991)). The construct used here starts at Ser 640 of the MAP4 sequence, preceded by 10 residues of the "hemagglutinin tag" (Field et al., 1988). Moreover, because the initial clone lacked the last 72 residues the C-terminal tail was actually derived from the human MAP4 sequence (West et al., 1991) which is highly homologous to the murine sequence. The residues where the human sequence differs are listed below the murine sequence in the last three rows. Phosphorylation sites by MARK are highlighted (black for KXGS motifs, grey for others).

RESULTS
Phosphorylation of MAP2 and MAP4 by the Protein Kinase p110 mark -Tau protein is phosphorylated efficiently on the KXGS motifs in its microtubule repeat region by a novel protein kinase, particularly at Ser 262 in repeat 1 (Drewes et al., 1995). This type of phosphorylation leads to the loss of taumicrotubule interaction, and therefore we termed the kinase p110 mark (microtubule affinity regulating kinase). This posed the question if p110 mark could play a more general role, by phosphorylating other MAPs in an analogous way. As shown in Fig. 2A, brain MAP2 and its juvenile isoform, MAP2c (lanes 2-5) and heart/lung MAP4 (lanes 8 and 9), are readily phosphorylated by p110 mark , with an efficiency comparable to the six tau isoforms (lanes 10 and 11). An E. coli expressed MAP4 fragment termed MAP4-BDC, comprising the C-terminal half which binds to microtubules (residues 640-1125), is also readily phosphorylated. As described previously for tau protein, phosphorylation of the other MAPs also leads to a small but significant shift toward higher M r . The time course of phosphorylation of tau, MAP2c, and MAP4-BDC is also similar, leading to a saturation of the incorporation of 32 P after 2-3 h (Fig. 2B). The final stoichiometry of phosphorylation was around 3 to 4 mol of phosphate/mol of MAP. The phosphorylation of recombinant MAP2c and tau by p110 mark is characterized by a similar K m value of around 30 -50 M, whereas the MAP4-BDC fragment is a somewhat better substrate with a K M of approximately 10 M (Fig. 2C).
As reported previously, p110 mark phosphorylated tau exclusively on serine residues. While phosphoamino acid analysis showed that the same is true for MAP2c, we found that MAP4 is also phosphorylated on threonine (Tables II and III).
Identification of the p110 mark Phosphorylation Sites on MAP2c and MAP2-For the determination of phosphorylation sites we used a strategy applied previously to tau protein in which two-dimensional phosphopeptide mapping by HV-TLE/ TLC is combined with HPLC purification and sequencing of peptides. This approach gives a clear representation of the relative amount of phosphorylation at different sites, and thereby allows a distinction between major and minor sites. The specificity of p110 mark for sites on MAP2 was examined by tryptic digestion of 32 P-phosphorylated MAP2 and MAP2c, the juvenile isoform, whose sequence is fully contained within the adult MAP2 isoforms. A comparison of the phosphopeptide maps obtained from recombinant MAP2c (Fig. 3A) and fulllength brain MAP2 (Fig. 3B) indicates that the majority of spots, including the most prominent ones, stem from sites located within the MAP2c sequence. This was confirmed by mapping a mixture of equivalent amounts of both samples (Fig.  3D). HPLC fractionation of the digest (Fig. 3E) allowed the isolation and subsequent sequence determination of four major labeled peptides. Each purified peptide was localized within the pattern of the MAP2c digest by TLE/TLC of the peptide in combination with a small aliquot of the digest (not shown). The results are compiled in Table II. The major phosphorylation site is Ser 1713 within the KCGS motif of repeat 3 (spot 2). A second major site is Ser 1682 within the KIGS motif of repeat 1. Minor sites are located outside the repeat region. Comparison of these data to the results obtained previously with tau shows that p110 mark displays a general specificity for the KXGS motifs in the microtubule binding domain. It is, however, inter-  Table I). esting that the preference for the individual repeats of MAP2 differs from that seen with tau. While the KXGS sites in repeats 1 and 4 were the major sites in tau, the main target in MAP2 was in repeat 3, followed by that in 1. Furthermore, while we did not observe significant phosphorylation of sites within tau's projection domain, we found several such sites in MAP2c, and based on the additional unidentified spots seen in Fig. 3B there appear to be additional sites in full-length MAP2. These minor sites did not correspond to KXGS motifs.
Identification of p110 mark Phosphorylation Sites on MAP4 -For the identification of the phosphorylation sites of p110 mark on MAP4, an E. coli-expressed fragment, MAP4-BDC, which comprises the C-terminal microtubule binding domain of mouse MAP4 and the acidic tail (residues 640-1125), and fulllength MAP4 isolated from mouse tissue were analyzed. A comparison of the phosphopeptide map obtained from the basic fragment ( Fig. 4A) with that from the native MAP4 (Fig. 4B) showed that there is only minor phosphorylation of sites that lie outside this region (Fig. 4C). In light of these findings, we concentrated on isolating and sequencing the phosphopeptides from the recombinant fragment ( Fig. 4E and Table III). Major sites were identified as Ser 914 in the KVGS motif of repeat 1 (spot 3), and Ser 1046 in the KVGS motif of repeat 4 (spot 7). This pattern is similar to what we observed with tau (Drewes et al., 1995), but different from MAP2 where repeat 3 is the main target, as described above. An additional prominent target of p110 mark on MAP4 is Thr 802 (Fig. 4B, spot 4). This site is within the proline-rich region which flanks the repeats at the Nterminal side. It is noteworthy that this part of the molecule is also thought to be involved in microtubule binding Olson et al., 1995), but there is no pronounced homology to the proline-rich regions in MAP2 or tau (Fig. 1). The extent of phosphorylation at other N-terminal sites was minor.
Effects of MAP Phosphorylation on Dynamic Instability of Microtubules-Under certain conditions, microtubules show abrupt transitions between phases of rapid shortening ("catastrophe") and elongation ("rescue"), and are termed "dynamically unstable" (Mitchison and Kirschner, 1984). MAPs are able to suppress microtubule dynamic instability by decreasing the frequency of catastrophe or increasing rescue (Pryer et al., 1992;Drechsel et al., 1992;Panda et al., 1995;Trinczek et al., 1995)). Phosphorylation of MAPs affects this stabilizing capacity by lowering microtubule affinity, and as a result the mean length of microtubules decreases. This effect can be observed in a time resolved manner by video dark field microscopy of individual microtubules.
In the experiment shown in Fig. 5 the concentration of tubulin was 10 M to ensure that microtubules did not selfassemble. However, microtubules nucleated and grew upon addition of native MAP4 prepared from brain (Fig. 5A, open circles), native MAP2 (Fig. 5B, open circles), or recombinant MAP2c (Fig. 5C, open circles). In these control experiments, p110 mark was added together with the MAPs but without ATP so that phosphorylation could not proceed. In a parallel experiment under otherwise identical conditions, 1 mM MgATP was added together with the kinase. The effect of the phosphorylation by p110 mark appeared rapidly (Fig. 5, A-C, closed circles). After about 5 min, growth is largely inhibited. Some microtu- Peptides were obtained from tryptic digests of E. coli-expressed MAP2c and MAP2 isolated from porcine brain. Peptides from MAP2c purified by HPLC were matched with peptides from brain MAP2 using two-dimensional TLE/TLC. The sequences are derived from the main radioactive peaks. Listed are the amount of material which was used for sequencing after a second purification run, the sequence with the phosphorylated residue (identified as S-ethylcysteine) starred, the phosphorylation site, the molecular mass of the peptide obtained by MALDI-MS, and the identification of the peptide in the two-dimensional HV-TLE/TLC maps (see Fig. 3). 740,000 1000 CGS*LK a (KXGS in "3") 555 635 b S1713 2 470,000 500 IGS*TDNIK (KXGS in "1") 847 931 b S1682 3 320,000 300 LS*NVSS*SGS*INLLESPQLATLAEDVTXAX (''C'') 3100 3186 b S1798/1802/1805 c 1 220,000 800 DQGGS*GEGLS*R (Acidic) 1062 1064 d S37/42 4 800 SS*LPR e (Basic) S1547 4 a Cysteine was found as cysteic acid. b Corresponds to the molecular weight of the peptide phosphorylated on a single site. c These residues appear to be only partially phosphorylated. d The phosphorylated peptide was probably not separated from the unphosphorylated in both purification runs. e Peptide eluted together with the Ser37/Ser42 peptide in both purification runs.

TABLE III Phosphorylation sites on MAP4
Peptides were obtained from E. coli-expressed fragment MAP4-BDC by HPLC and matched with peptides from MAP4 (isolated from mouse heart and lung tissue) using two-dimensional TLE/TLC. The sequences are derived from the main radioactive peaks. Listed are the number of counts in radioactive labeled fractions of the the first HPLC run, the amount of material which was used for sequencing after a second purification run, the sequence with the phosphorylated residue (identified as S-ethylcysteine) starred, the phosphorylation site, the molecular mass of the peptide obtained by MALDI-MS, and the identification of the peptide in the two-dimensional HV-TLE/TLC maps (shown in Fig. 4). bule nucleation initially took place while the MAPs were not yet phosphorylated, but polymerization after this time was suppressed due to progressive phosphorylation of the MAPs. In another type of experiment, the MAPs were phosphorylated by p110 mark for 30 min prior to their addition to tubulin (Fig. 5, A-C, triangles). In this case, nucleation was also abolished. However, tubulin could still form polymers, as short microtubules of about 2 m length could be observed when axonemes were added to promote nucleation. Analysis of mutant forms of MAP2c shows that the loss of binding capacity depends on the phosphorylation of both Ser 1682 in repeat 1 and Ser 1713 in repeat 3 (Fig. 5D). If only one of these sites is mutated, microtubule growth is still induced by the phosphorylated mutants. The length histograms show the distribution of microtubule lengths at 5 min (Fig. 6, A-C) and 30 min (Fig. 6, D-F) after the addition of MAP and kinase. At 5 min, where the microtubules are still in the growing phase, the length distributions, peaking around 10 m, are comparable in the presence or absence of ATP (kinase active or inactive, Fig. 6, A-C, open and closed circles). After 30 min the distribution of the control microtubules (no ATP) has become broader (Fig. 6, D-F, closed circles). However, incubation with ATP strongly suppresses long microtubule and shifts the distribution to short lengths (open circles in Fig. 6, D-F).
In summary, the results show that phosphorylation by p110 mark has similar dramatic effects on the function of MAP2c, MAP2, and MAP4. Microtubule stabilization is progressively impaired when the kinase and ATP are added together with the MAP to the tubulin sample. Moreover, prephosphorylated MAPs are not able to support microtubule growth or even nucleation. These effects are comparable to the previously reported effects of p110 mark phosphorylation on tau (Drewes et al., 1995).
Effects of MAP Phosphorylation on MAP-microtubule Binding-To determine whether the effects of MAP phosphorylation on dynamic instability was due to a reduced interaction with microtubules, we performed binding studies with taxol-stabilized microtubules. As we had found previously for tau protein coli. B, full-length MAP2 (isolated from porcine brain). C, diagram of the more prominent spots with identification of the HPLC-purified and sequenced phosphopeptides (see Table II). Spot 2 contains Ser 1713 (KXGS in repeat 3), spot 3 Ser 1682 (KXGS in repeat 1). D, mixture of A and B showing that the major sites on MAP2 and MAP2c are identical. E, separation of the tryptic digest of phosphorylated MAP2c by reversed phase HPLC (C 18 ). Radioactive fractions were purified by a second HPLC run (not shown) and sequenced. The identification of phosphorylated peptides is compiled in Table II, e.g. spot 2 contains the CGS motif of repeat 3, spot 3 contains the IGS motif of repeat 1, spots 1 and 4 are peptides outside the repeats. C, diagram of the more prominent spots with identification of the HPLC-purified and sequenced phosphopeptides (see Table III). Spot 3 contains Ser 914 (KXGS in repeat 1), spot 7, Ser 1046 (KXGS in repeat 4). D, a mixture of A and B showing that the major sites on MAP4 are localized within the MAP4-BDC construct. E, separation of the tryptic digest of phosphorylated MAP4-BDC by reversed phase HPLC (C 18 ). Radioactive fractions were purified by a second HPLC run (not shown) and sequenced. The identification of phosphorylated peptides is compiled in Table III, e.g. spot 3 contains the VGS motif in repeat 1, spot 7 contains the VGS motif of repeat 4. (Gustke et al., 1994;Trinczek et al., 1995), these data demonstrate that loss of affinity (or increase in K d value) correlates very well with increased dynamics. As shown in Fig. 7, nonphosphorylated, recombinant MAP2c binds tightly to taxolstabilized microtubules with an apparent K d around 0.25 M (open circles). In the non-phosphorylated state, three mutants (MAP2cA319, MAP2cA350, and MAP2cA319ϩ350, having Ser 1682 and Ser 1713 in the KXGS motifs of repeats 1 and 3 mutated to Ala individually or both) show the same behavior (data not shown). However, after phosphorylation with p110 mark only the binding capacity of the wild-type MAP2c is dramatically affected. It decreases to 10% of the original value, and the K d thereby increases at least hundredfold (closed circles). With a single mutation of either Ser 1682 or Ser 1713 to Ala, the binding becomes weakened about 30-fold (K d increases to about 7 M upon phosphorylation, open and filled squares). When both phosphorylation sites are mutated, the binding characteristics after phosphorylation remain almost unchanged as compared to the non-phosphorylated protein (triangles).

DISCUSSION
Phosphorylation Sites on MAPs-Microtubules are dynamic polymers, and their dynamic behavior is regulated by cells according to their needs. MAPs and their phosphorylation state have a pronounced effect on microtubule dynamics, and indeed changes in MAP phosphorylation patterns accompany major rearrangements of microtubules during the cell cycle or during differentiation (Vandre et al., 1991;Dinsmore and Solomon, 1991;Halpain and Greengard, 1990;Preuss et al., 1995;Ookata et al., 1995). Our previous studies have focused mainly on the kinases and phosphatases regulating the phosphorylation of the neuronal tau protein, with emphasis on the hyper- For each condition 500 -800 microtubules were analyzed, and the mean length were plotted against time. Tubulin concentration was 10 M in all cases, the concentration of MAP4 and MAP2 was 1 M, that of MAP2c, 2 M. In control experiments, ATP was omitted (-ATP). Open circles in A-C: the MAPs were preincubated for 30 min with 2.5 milliunits/ml p110 mark (final concentration), but without ATP. By adding 10 M tubulin, microtubules were nucleated and the mean microtubule length increased up to about 20 m within 30 min. If ATP was present no self-nucleation occurred, showing that the phosphorylation of the MAPs prevented microtubule formation. Short microtubules of about 2 m length could only be observed by adding axonemes (10 -100 fmol) to promote seeded nucleation (open triangles in A-C). Closed circles in A-C: tubulin and MAP were mixed at 4°C with 2.5 milliunits/ml of p110 mark (final concentration) and 1 mM ATP, and the temperature was shifted immediately to 37°C (so that initially the MAPs were unphosphorylated). Microtubule growth was promoted in all three cases, but the final mean microtubule length was only about half of that observed for the unphosphorylated MAPs (compare to open circles). D, the effect of phosphorylation site point mutations of MAP2c. All proteins were preincubated with kinase and ATP as described above. Triangles, wild type MAP2c; closed circles, MAP2cA319 (KXGS in repeat 1 mutated to KXGA); squares, MAP2cA350 (KXGS in repeat 3 mutated to KXGA); closed squares, MAP2cA319/A350 (KXGS in both repeats mutated to KXGA).  (Table I). Open circles, non-phosphorylated wild-type MAP2c. The binding is tight (K d about 0.25 M) and saturates around 17 M ligand (Ϸ1 MAP2c molecule per 2 tubulin dimers). Closed circles, wild-type MAP2c, phosphorylated previously with p110 mark (2.5 milliunits/ml) for 2 h. Note that there is essentially no binding. Closed and open squares, MAP2cA319 and MAP2cA350, phosphorylated previously with p110 mark (2.5 milliunits/ ml) for 2 h. The affinity to microtubules has decreased markedly (K d Ϸ 7 M) although the stoichiometry remains similar to the wild type MAP2c. Triangles, MAP2cA319/A350, phosphorylated previously with p110 mark (2.5 milliunits/ml) for 2 h. The binding is similar to the unphosphorylated protein, showing that the sensitivity to phosphorylation has disappeared. phosphorylation that plays a role in the neurofibrillary pathology of Alzheimer's disease. It was intriguing to note that many "abnormal" phosphorylation sites were in Ser-Pro or Thr-Pro motifs and can be phosphorylated by proline-directed kinases (such as MAPK, cdk5, or GSK-3) which are important in cellular signal transduction (for review, see Mandelkow and Mandelkow (1993)). Indeed, using phosphorylation-sensitive antibodies we observed that kinases of this class phosphorylate both MAP2 and tau in cells in a cell-cycle or differentiation dependent fashion (Berling et al., 1994;Preuss et al., 1995). Similar observations were made in other laboratories (e.g. Dinsmore and Solomon (1991) and Ookata et al. (1995)). There was, however, the puzzle that proline-directed phosphorylation had only a comparatively mild effect on microtubule stability; it was dwarfed by the much larger effect of another kinase activity which phosphorylated mainly KXGS motifs in tau Trinczek et al., 1995). The search for the kinase lead to a 110-kDa protein which was termed MARK because of its regulation of the affinity of tau to microtubules (Drewes et al., 1995). Given the specificity of the enzyme it was natural to ask whether the kinase would also phosphorylate related MAPs such as MAP2 and MAP4, and whether this would have similar consequences on microtubule stability. We show here that this is indeed the case.
Our results imply that the role of MARK in regulating MAP interactions with microtubules may be more general than expected. Because tau, localized primarily to axons, and MAP2, distributed in dendrites, are both substrates, MARK or related kinases could be active in different neuronal cell compartments. An even more general role is implied by the results with MAP4, since this ubiquitous MAP has been inferred to affect microtubule stability in dividing cells (Bulinski and Borisy, 1980;Parysek et al., 1984;Chapin and Bulinski, 1994;Olson et al., 1995). Thus far it has been difficult to determine what combination of MAPs, phosphorylation sites, kinases, and other factors are responsible for the pronounced increase in microtubule dynamics during mitosis. MAP4 was considered a likely candidate, as well as other related ones (e.g. XMAP from Xenopus eggs, Faruki and Karsenti (1994)). Regarding kinases, cdc2 and MAP kinases were suggested as potential triggers of microtubule reorganization (Gotoh et al., 1991;Verde et al., 1992;Lieuvin et al., 1994;Ookata et al., 1995). However, it remains to be seen whether these kinases act directly or via other intermediate steps. The weak effect of proline-directed phosphorylation on microtubule dynamics makes us believe that other kinases, such as MARK, may be involved. In this regard it is interesting that MARK is itself activated by phosphorylation, pointing to other kinase(s) upstream in the signaling pathway. 2 The phosphorylation of MAPs has been studied by a number of authors, and it is pertinent to ask how the results compare with ours. In most cases it was concluded that phosphorylated MAPs bound less tightly to microtubules and supported their assembly less efficiently (although exceptions were also noted, see Brugg and Matus, 1991). However, in the majority of studies, the phosphorylation sites involved in the regulation were not known, and indirect information, such as kinase consensus motifs, are not reliable (as illustrated for tau and CaM kinase by Steiner et al. (1990), or for MAP4 and cdc2 by Ookata et al., 1995). There are, however, a few cases where phosphorylation sites have been determined directly. Examples include the sites in MAP2 altered by PKC (Ainsztein and Purich, 1994), or the sites on tau phosphorylated by several kinases (PKA, PKC, Ca/calmodulin dependent kinase II, and the proline-directed kinases MAPK, GSK-3, cdc2, and related kinases, see below). While additional parameters need to be measured for the influence of site-specific modifications on microtubule dynamics to be rigorously assessed, the observations with tau (the MAP studied most comprehensively) allows a distinction to be made between the sites within and outside the repeat domain. The sites outside the repeats examined so far have either no effect on microtubule binding and dynamics, or only a moderate effect, reducing the stabilizing power of tau from "high" to "medium" (in the classification of Trinczek et al. (1995)). This includes the many Ser-Pro or Thr-Pro sites (phosphorylated by proline-directed kinases), as well as PKA or Ca/calmodulin dependent kinase II sites (Steiner et al., 1990;Scott et al., 1993;Brandt et al., 1994). Inside the repeats there are the KXGS motifs affected by MARK. One of these (Ser 262 in repeat 1) eliminates the stabilizing power of tau, the others have only a modulatory influence. The KXGS motifs of tau can also be phosphorylated to some extent by PKC (Ser 324 in repeat 3, Correas et al., 1992), PKA (mostly Ser 324 and Ser 356 in repeats 3 and 4, Scott et al. (1993) and Drewes et al. (1995)), and GSK-3 when activated by heparin (Ser 262 in repeat 1, Song and Yang (1995)). 3 In vivo the phosphorylation at KXGS motifs is normally low (Seubert et al., 1995), consistent with its tight association with microtubules. This implies that the kinases affecting KXGS motifs are normally down-regulated.
The results on the other MAPs echo those of tau. Phosphorylation sites outside the repeats may reduce the interaction with microtubules, but they do not eliminate it. For MAP4, this includes the sites Ser 667 and Ser 760 in the proline-rich domain (our numbering, see Table I) which are potential targets of cdc2 (Ookata et al., 1995). Sites inside the repeats include the KGXS motifs which cooperate to eliminate the interaction with microtubules (see Fig. 7). They also include reported PKC sites in MAP2 at serines 1705, 1713, and 1730 (our numbering, Fig. 1; see Ainsztein and Purich, 1994). The second of these is in the KXGS motif of repeat 3. The example illustrates how PKC could exert a modulatory effect by phosphorylating one KXGS motif, while MARK would eliminate microtubule interactions by phosphorylating two motifs (in 1 and 3). In the case of MAP4, point mutants of KXGS motifs are not available, but in analogy with MAP2 and tau we expect that the full inhibition of microtubule binding by MARK resides in repeats 1, 4, or both. Other phosphorylation sites of MAP4 have not been determined thus far.
Most studies on MAPs emphasize their role as microtubule stabilizers, but it is worth noting that they have at least two additional functions. One is their role as "spacers" between microtubules and other cellular components (Chen et al., 1992). This is achieved mainly by the acidic N-terminal domain which may be short (as in MAP2c or tau) or long (as in MAP2 or MAP4). A third function is that of a docking site for cellular enzymes, including kinases and phosphatases or their cofactors (PKA, cdc2, PKC-, MAP kinase, PP-1, see Obar et al. (1989), Mandelkow et al. (1992), Baumann et al. (1993), Ookata et al. (1995), Lehrich and Forrest (1994), Reszka et al. (1995), and Sontag et al. (1995)). It is intriguing to speculate that the docked kinases may be activated by some signaling cascade and then phosphorylate their host protein or others nearby, thus modulating their association with the microtubule cytoskeleton.
Structural Implications-We conclude by commenting on possible structural implications of the phosphorylation by MARK. The main sites affecting microtubule binding and dynamics are in the repeat domain although this domain, taken by itself, interacts only rather weakly with microtubules (Ennulat et al., 1989;Joly and Purich, 1990;Butner and Kirschner, 1991;Gustke et al., 1994). On the other hand, the domains flanking the repeats, particularly the basic and proline-rich domains, cause strong binding to microtubules Lee and Rook, 1992;Olson et al., 1995;Ookata et al., 1995); however, their phosphorylation has only a limited effect on microtubule interactions. We view the flanking domains as "targeting" domains which place the MAPs on the microtubule surface which then allows the "catalytic" repeat domains to affect microtubule stability, presumably by providing extra bonds between subunits and protofilaments (Gustke et al., 1994). This view would overcome the apparent discrepancy between the weak binding of the repeat domain to microtubules and the strong effect of phosphorylation in the repeats.
It still needs to be explained why certain sites in the KXGS motifs can have a major effect (e.g. in repeat 1 of tau, or in 1 and 3 of MAP2), independently of whether other KXGS motifs are phosphorylated as well. This conceptual difficulty could be overcome if one abandons the traditional models of MAP-microtubule interactions. These models have assumed that the repetitive elements in the MAP sequence correspond to the repetitive nature of the microtubule lattice, i.e. each repeat was thought to interact with a different tubulin subunit, but in an equivalent manner. In this picture one would expect that the phosphorylation of one repeat might release it from its subunit but leave the other repeats in place, and consequently one would expect that several phosphorylation sites would have to be combined before the MAP detaches from the microtubule.
There is, however, an alternative view: the repeats could be folded into a coherent structure, or "buttoned up," such that they could link tubulin subunits in adjacent protofilaments on the microtubule surface. This interaction could be disrupted if the folding of the repeat domain were destroyed. Imagine a folded structure formed by several repeats of the MAPs in which the "button" was formed by several unphosphorylated KXGS motifs which could be "unbuttoned" by phosphorylation. Within this model, the positions of the most critical residues might be variable and depend on details of the surrounding sequence. This might explain the predominant role of the phosphorylation in repeat 1 of tau, compared with the cooperation between 1 and 3 in MAP2 or between 1 and 4 in MAP4. There is increasing evidence for a folded structure in the repeats: the two cysteines in repeats 2 and 3 of tau are in close proximity (Schweers et al., 1995), and the reaction of certain antibodies can only be explained by discontinuous epitopes involving a folded repeat domain . Interestingly, the accessibility of the core of Alzheimer PHFs to proteases is also explained if one assumes that the repeats of tau are folded up; the result is that the resistant core is formed by peptides roughly equivalent to three repeats, but in different combinations (end of 1 plus 3, 4, 4a, or end of 1 plus 2, 3, 4, Jakes et al. (1991)). The details of this structure are not yet known, but it will be crucial for an understanding of the MAPmicrotubule interaction in molecular terms.