Molecular Structure of Microtubule-associated Protein 2b and 2c from Rat Brain*

Full length cDNA clones encoding microtubule-as-sociated proteins (MAP) 2b and 2c from rat brain have been isolated and sequenced. The cDNA fragments spanning the coding regions for both MAP2b and MAP2c were assembled and expressed in Escherichia coli. The mobility of these bacterial expressed proteins in sodium dodecyl sulfate gels is identical to that of MAPPb and MAP2c from rat brain. The protein se- quence of rat MAP2b has been compared to the full length sequence from mouse and the partial sequence from human high molecular weight MAPS. This comparison has revealed that MAP2b is composed of sev- eral highly conserved domains flanked by domains with extensive sequence divergence. Two of the conserved domains, found either at the NH2 or COOH terminus, overlap with the binding domain for the subunit of the CAMP-dependent protein nase and the microtubule-binding domain, A third homologous domain of lies in a central region of


Full length
cDNA clones encoding microtubule-associated proteins (MAP) 2b and 2c from rat brain have been isolated and sequenced. which extends from the microtubule surface is composed of an extensive number of cY-helices separated by small turns which may account for the extended yet flexible structure of MAP2. Interestingly, the 4000-base pair deletion from the middle of MAP2b which generates MAP2c not only removes these helices, but also this third highly conserved MAP2b domain. The vertebrate nervous system is composed of neurons with a large variety of different morphologies. One class of cytoskeletal elements which appear to play a crucial role in determining the shape of nerve cells are the microtubules and their associated proteins, called MAPS' (1). The latter appear to be important for both the assembly and stability of microtubules during neurite outgrowth (2), as well as for the overall shape and plasticity of neuronal processes (3)(4)(5). This appears to be accomplished by varying the affinities of MAPS for different cytoskeletal elements by phosphorylation (1,5) and by altering their composition during development (4)(5)(6). This * This work was supported by the Federal Government of Germany (Bundesministerium ftir Forschung und Technologie). The costs of publication of this article were defrayed in part by the payment of page charges. This  in turn effects the extent of cytoskeletal cross-linking (7). One major microtubule-associated protein that is specifically localized in dendrites is MAP2 (4). In rat brain, it is composed of at least three isoforms with different temporal patterns of expression (6). MAP2b is a protein of 280 kDa on SDS-polyacrylamide gels which is expressed throughout brain development.
MAP2a is slightly larger than MAPPb. It appears in rat cerebral cortex around postnatal day 12 and is present in adult neurons. A 70-kDa MAP2 variant, called MAP2c, is expressed during embryonic brain development and until postnatal day 10 (8). MAP2a and 2b are encoded by mRNAs of approximately 9 kilobase (9, lo), whereas MAP2c is encoded by a 6-kilobase mRNA (9). All three are transcribed from a single gene and arise via alternative splicing of a primary MAP2 transcript (9). The complete cDNA sequence for the mouse high molecular weight MAP2 was recently reported (ll), as was a description of two domains found at the COOH terminus of MAPS, that are considered to be involved in either microtubule binding (12) or bundling (13). In addition, two groups have reported that a short domain in the amino terminus of MAP2 is responsible for the binding of the regulatory subunit of the CAMP-dependent protein kinase II (14,15).
In this report we describe the full length cDNA cloning, sequencing, and expression in Escherichia coli of both rat brain MAP2b and MAP2c. We also present an analysis of their primary and secondary structures, and a model describing the role these two proteins might play in determining dendritic plasticity. In addition, we have compared the amino acid sequence of the entire rat MAP2b to the full length mouse high molecular weight MAP2 (12) as well as a partial human high molecular weight MAP2 sequence (16,17). This comparison has revealed that MAP2 is composed of several highly conserved domains, some of which overlap with previously identified functional domains, and others suggesting the existence of new functional domains. assembled, one from a postnatal day 5 rat brain poly(A+) RNA, described previously (9), and the second from glioma C6 cell mRNA.
This second library was constructed in an identical manner as the rat library (9) in X gtll. The rat liver genomic library, constructed in EMBL3 (Stratagene) was a generous gift from G. Schuetz (Deutscher Krebs Forschung Zentrum, Heidelberg). Genomic and cDNA clones were identified by plaque hybridization as described by Maniatis (45).
Positive clones were plaque purified and cDNA inserts subcloned into pUC18 (46) before sequencing with T7 polymerase (Pharmacia) as described by the manufacturers. Expression ATG P 'H BK P K H x " " P G R c x

RESULTS
cDNA Clones Encoding Rat Brain MAP2b-cDNA clones encoding pieces of MAP2b and MAP2c were initially isolated from a random primed X gtll expression library screened with MAP2 antibodies (9). The cDNA inserts from two of these clones, 38a and 19a (Fig. l), were then used to rescreen the same library to isolate a series of overlapping clones. The layout of these clones and the strategy used to determine 100% of their sequences from both strands is shown in Fig. 1.
One set of these clones (14b, 19a, and 4.2) contains an open reading frame encoding a protein of 1830 amino acids (Fig. 4). The methionine at position 1 conforms to the eukaryotic translation start site sequence (18) and is preceded by two inframe stop codons (19). The molecular weight of the encoded protein is calculated to be 200,000. This is about 80 kDa smaller than that predicted for high molecular weight MAP2 on SDS gels. In order to resolve this discrepancy, we assembled this group of clones and expressed the encoded protein in E. coli. As shown in Fig. 2, lane 1, the E. coli expressed protein has an apparent molecular weight on SDS gels of 280,000. A comparison of the mobility of the E. coli expressed protein to that of MAP2a and MAP2b from adult rat brain on SDS gels (Fig. 2, lane 2) demonstrates clearly that our E. coli expressed protein has an identical mobility to MAP2b (Fig. 2, lane 3). It would therefore appear that this collection of clones encode MAP2b and not MAP2a.
Genomic and cDNA Clones Encoding MAP2c-The cDNA sequence analysis of both strands of clone 38a revealed that this clone was composed of nucleotide sequences found at both the 5' and 3' end of the first set of clones (Fig. 1). The deduced amino acid sequence from 38a contained an open reading frame which was identical with amino acids 50-152 and the last 316 amino acids of the MAP2b sequence (underlined sequence in Fig. 4).
Since this and earlier data (9) suggested that clone 38a might encode a portion of MAP2c, we sought to answer two questions. First, does 38a actually encode MAP2c (see next section), and second do MAP2c and MAP2b contain identical amino-terminal sequences? We were able to answer the second of these two questions by two separate approaches. The first was to isolate and sequence cDNA clones that extended upstream from the 5' end of 38a. This was accomplished by screening a glioma C6 random primed cDNA X gtll expression 148 Yaa Peptide Cys library with clone 38a. This cell line was chosen since we had previously observed that C6 cells express only MAP2c and a 6-kilobase mRNA thought to encode MAP2c (9). A 500-base pair clone, C6.11, was identified ( Fig. 1) which overlapped with the 5' end of 38a and extended the sequence of 38a by 350 nucleotides in the 5' direction. The sequence of this clone was found to be identical with the first 500 nucleotides of our MAP2b clones and the first 150 nucleotides from the 5' end of 38a (Fig. 1). The sequence included the two in-frame stop codons followed by the initiating methionine found in MAP2b. This suggested that both MAP2b and MAP2c have an identical amino terminus, however, to verify this, we also isolated genomic clones from a rat liver genomic library constructed in an EMBL3 X vector which hybridized to a 5' EcoRI fragment from MAP2b, called 14bII (Fig. l), and the 5' end of 38a. Several clones were found that hybridized to 14bII. When mapped, it was found that the first 500 base pairs of MAP2 were encoded on 5 exons spanning 50 kilobases of genomic DNA (data not shown). One of these exons was found to be 200 nucleotides long and to be colinear with both the proposed MAP2b initiating codon and the 5' end of 38a The underlined sequence of MAP2b indicates the amino acids that compose MAPZc. The tubulin and regulatory subunit of the CAMP-dependent protein kinase II binding domains are in boxes, as well as the proposed calmodulin-binding domain. Proposed CaM kinase and CAMP-dependent kinase phosphorylation sites are indicated as ouerlines. Dashes indicate amino acids not present in a given sequence.

Microtubule-associated
Proteins 2b and 2c ( Fig. 1). No other genomic clones or fragments were found which contained these sequences, demonstrating that both MAP2b and MAP2c utilize an identical initiating methionine located within the same exon. A summary of the relationship of the MAP2c sequence to MAP2b is diagrammed in Fig. 3. This demonstrates that MAP2c is composed of identical sequences from the first 152 and last 316 amino acids of MAP2b.
Clone 38a Encodes MAPZc-Since the combined sequences from clone 38a and C6.11 predict a protein of about 43 kDa, whereas the predicted molecular mass of MAP2c was 70 kDa (20), we wished to verify that the protein encoded by these clones was MAP2c. This was accomplished by two separate approaches. In the first, we assembled the cDNA fragments predicted to encode MAP2c and expressed them in E. coli. In Fig. 2, we show Western blots stained with an anti-MAP2 monoclonal antibody comparing the mobility of our E. coli expressed protein to rat brain expressed MAP2c. As seen in lane 3, postnatal day 7 rat brain expresses several immunoreactive MAP2 bands running at 70 kDa. Since these multiple bands are due to protein phosphorylation, we pretreated our preparation with calf intestinal alkaline phosphatase overnight (lane 4). As observable in lane 5, our E. coli expressed construct has an identical mobility to the dephosphorylated MAP2c (lane 4) suggesting that our clones encode MAP2c. The identity of this protein was further corroborated by immunizing rabbits with an &amino acid peptide, coupled to thyroglobulin, that spans the proposed splice junction identified in 38a (Fig. 3). Sera from one of these rabbits reacted specifically with a 70-kDa protein on Western blots from postnatal day 7 rat brain Sl phosphatase-treated supernatant and from the E. coli expressed MAP2c (Fig. 2, lane 7) and not with MAP2b (lane 8). These data demonstrate that MAP2c is a 42.3-kDa protein which runs at 70 kDa on SDS gels and that it is missing 1363 amino acids present in the middle of MAP2b.
Comparison of Rat MAP26 to Mouse and Human High Molecular Weight MAPZ-Presented in Fig. 4 is a comparison of the amino acid sequence from rat MAP2b to the full length amino acid sequence from mouse high molecular weight MAP2 described by Lewis et al.,(12) and the partial human high molecular weight MAP2 (16, 17). The rat MAP2b sequence shows a 92% overall homology to the mouse and 76% to the human. This, however, only tells part of the story since as seen in Fig. 4, the non-homologous amino acids are clustered and separated by domains of nearly complete homology. Interestingly, several of these highly conserved regions have recently been identified as domains involved in microtubule binding (amino acids 1650-1800) (11,13) and in the binding of the regulatory subunit of CAMP-dependent kinase (amino acids 75-125) (14, 15). Two additional highly conserved domains lie between amino acids 650 and 940 (see next section) and between amino acids 1370 and 1650 (Fig. 4). The latter domain has three very interesting features: first, it is very rich in proline; second it contains many of the proposed phosphorylation sites for both Ca*'/calmodulin (CaM)-dependent protein kinase type II (CaM kinase) and CAMPdependent kinase; and third it contains a stretch of positively charged amino acids whose arrangement resembles the CaMbinding domains found in other proteins (Fig. 5). These characteristics, as well as the fact that it lies just 5' of the microtubule-binding domain, suggest that this region is a hinge domain of MAP2b and 2c whose functional properties can be altered by phosphorylation.
Secondary Structure Prediction of MAPZb-Using two different computer programs which utilize the principles established by 22) to predict the secondary structure of proteins, we have analyzed the helix, coil, and turn distribution in MAP2b. These programs predict that MAP2b is composed of 12% turn, 62% helix, and 26% coil structure. However, as diagramed in Fig.  6 these predicted secondary structures are clustered into several domains. Beginning with the COOH terminus and ending with the MAP2c splice junction, one finds very little secondary structure predicted (Fig. 6). The 3' end of this domain has previously been identified by several groups as the domain essential for microtubule binding (13, 23). Near the MAP2c junction is a region very rich in prolines which could be part of a hinge domain from which the remainder of MAP2b extends from the microtubule surface. The next domain begins at the MAP2c junction and extends two-thirds of the distance to the NH2 terminus. This region is predicted to be composed of some 19 helices in a row separated by very short turns. In principle this domain might be predicted to be a very flexible rod. Adjacent to this is a sequence of amino acids, again with very little secondary structure followed by several more regions rich in helices before coming to the NH2 terminus. It should be noted that this second coil domain overlaps with the highly conserved domain spanning amino acids 650-920 as described earlier, as well as a second clustering of possible CaM and CAMPdependent kinase sites (Figs. 4 and 6). DISCUSSION In this paper we have described the cDNA cloning, sequencing, structural characterization, and expression of two isoforms of MAP2. The cDNA clones spanning the rat brain high molecular weight MAP2 were found to encode a protein of 200,000. Although this is some 80 kDa smaller than previous estimates on SDS-polyacrylamide gels (24), it is in accord with estimates of 220 kDa from sedimentation equilibrium centrifugation and gel filtration experiments (25), as well as with the predicted molecular mass of the cloned mouse high molecular mass MAP2 (12). When clones encoding the high molecular weight rat brain MAP2 were assembled and expressed in E. coli, the encoded protein was found to be cm Tubulin ---- However, we and others have never been able to generate MAP2b from MAP2a by phosphatase treatment so that it is still not clear whether MAP2a is generated via some other post-translational modification or alternative splicing.* A second set of clones were found to encode a 43-kDa protein which is composed of sequences from the 5' and 3' ends of the high molecular mass MAP2. Antibodies generated against an 8-amino acid peptide spanning the splice sequence ( Fig. 2) in this protein were found to be immunoreactive with a 70-kDa MAP which co-migrates with MAP2c. The E. coli expressed form of this protein was also found to co-migrate on SDS-polyacrylamide gels with dephosphorylated rat brain MAP2c and was immunoreactive with known anti-MAP2 monoclonal antibodies demonstrating that these clones encode MAP2c.
A comparison of the protein sequences for MAP2b and MAP2c ( Fig. 4) demonstrates that both proteins possess identical amino-terminal sequences which have recently been shown to bind the regulatory subunit of CAMP-dependent kinase (14,15) and identical carboxyl-terminal domains where both bind to microtubules.
These data indicate that MAP2c is lacking the central domain of MAP2b thought to be involved in intra-cytoskeletal cross-bridging (44) and is generated as a result of alternative splicing of a primary MAP2 transcript (9). Structural Analysis of MAP2b and MAPSc-Previous structural analysis of the high molecular mass form of MAP2 and the 65-kDa microtubule-associated protein 7 by 'H NMR, 14C NMR, and circular dichroism (25,(27)(28)(29) suggested that these proteins contain very little secondary structure (7% a-helix and 90% unordered conformation) (25). Interestingly, when Woody et al. (27) studied the changes in flexibility of high molecular weight MAP2 and 7 before and after binding to microtubules by 'H NMR, they observed that whereas the binding of MAP2 to the microtubule surface had very little effect on its degree of flexibility, T'S flexibility decreased after binding. For MAPB, the measured flexibility was shown to lie in its long extended arm (27). MAP2c not only has a similar molecular weight to 7, but also possesses a nearly identical microtubule-binding domain (12). It might therefore be predicted that much of MAP2c interacts with microtubules and that a small projection domain interacts with the regulatory subunit of the CAMP-dependent kinase II (14). These biophysical data of high molecular weight MAP2 stand somewhat in contrast to both rotary shadowing and electron microscopic data which suggest that the arm of high molecular weight MAP2 is an extended semirigid rod-like molecule (7, 30) which might be folded back on itself (31). Perhaps the answer is somewhere in between since our computer-generated secondary structure predictions suggest that although a portion of MAP2 exists as a random coil, one cannot fail to notice that beginning near the MAP2c junction and extending toward the amino terminus is a very significant stretch of helices separated by short turns. This motif could theoretically permit MAP2 to exist in an extended flexible conformation yet perform a structural role in the neuronal cytoskeleton.
Binding Motifs and MAP2 Phosphorylation-Post-transla-' B. Schwanke, B. Schulz, and C. C. Garner, unpublished observations. tional modifications have been suggested to play a role in regulating MAP interaction with the neuronal cytoskeleton. Much of these data come from MAP2 which has been found to exist in a variety of phosphorylation states (32) which are thought to influence its affinity for microtubules as well as for other proteins (micro-and neurofilaments (26, 33, 34)). Several studies have localized two domains within MAP2 which can be efficiently phosphorylated (Fig. 6) by either the CAMP-dependent kinase or the CaM kinase (26,32,35,36). The binding of CAMP-dependent kinase has recently been shown to occur via the regulatory domain somewhere between amino acids 75 and 125 (14, 15). Although it is not known if the CaM kinase binds directly to MAPB, Lee and Wolff (37) have shown that CaM binds near the tubulin-binding domain. We have analyzed the MAP2 sequence for a motif which conforms to those previously described for CaM (38) and have identified a very basic region that lies near the tubulin-binding domain of MAP2 (Fig. 4). Fig. 5 shows a comparison between the proposed site in MAP2 to that described for both the skeletal and smooth muscle myosin light chain kinase (38). The principle features of a CaM-binding site conserved in MAP2 are basic residues at positions 11, 12, 13, and 17 and anywhere from 1 to 3 hydrophobic residues between amino acids 13 and 17 (38). Whether this is in fact the actual location remains to be demonstrated.
One additional aspect of the MAP2 structure which we wish to note relates to a highly conserved domain in MAP2 spanning amino acids 650-930. Although the function of this domain is not known, it seems more than coincidental that this region which is conserved between human, mouse, and rat, lies between the major helix domains and is the second major phosphorylation domain of MAP2. Like the hinge and tubulin-binding domain, this region contains a set of consensus sequences which might be phosphorylated by either CAMP or CaM kinases (Figs. 4 and 6) (39, 40). The results from Burns and Islam (31) and Hernandez et al. (36) suggest that this domain of MAP2 is phosphorylated in vitro. These features suggest that this domain plays a role in regulating the interaction of MAP2 with its surroundings. MAP2b and MAP2c in the Developing Nervous System-What then is the role of MAP2b and MAP2c in the developing nervous system? At present it has been demonstrated that MAP2b is composed of several domains. Beginning with the NH, terminus, there exists within the first 100 amino acids a binding domain for the regulatory subunit of the CAMPdependent protein kinase type II. This is followed by a large domain of 1600 amino acids which has been shown by rotary shadowing of MAP2 microtubules in vitro (30) and in situ by immunoelectron microscopy (7) to be a filamentous arm of MAP2 which is either involved in spacing microtubules or in cross-bridging various components of the cytoskeleton (33, 34, 44). Between amino acids 1600 and 1800 lies the microtubule-binding domain (12, 23) and in the last 100 amino acids of MAPS, Lewis et al. (13) have identified a region that might be involved in microtubule bundling. Since MAP2c is missing this long filamentous arm, one might predict that the role of MAP2c is to prevent microtubules from interacting with the cytoskeleton by occupying potential MAP2b microtubule-binding sites. Several lines of evidence support this hypothesis. First MAP2c has been shown to be expressed slightly before MAPXb appears in neurons (41). Second, it is present in developing central nervous systems only during neurite outgrowth and not during process stabilization (4,6). Third, the disappearance of MAP2c from neurons corresponds with both the accumulation of MAP2a (8) and the beginning of synap-