DAMAGE, a Novel α-Dystrobrevin-associated MAGE Protein in Dystrophin Complexes*

Mice rendered null for α-dystrobrevin, a component of the dystrophin complex, have muscular dystrophy, despite the fact that the sarcolemma remains relatively intact (Grady, R. M., Grange, R. W., Lau, K. S., Maimone, M. M., Nichol, M. C., Stull, J. T., and Sanes, J. R. (1999) Nat. Cell Biol. 1, 215-220) Thus, α-dystrobrevin may serve a signaling function that is important for the maintenance of muscle integrity. We have identified a new dystrobrevin-associated protein, DAMAGE, that may play a signaling role in brain, muscle, and peripheral nerve. In humans, DAMAGE is encoded by an intronless gene located at chromosome Xq13.1, a locus that contains genes involved in mental retardation. DAMAGE associates directly with α-dystrobrevin, as shown by yeast two-hybrid, and co-immunoprecipitates with the dystrobrevin-syntrophin complex from brain. This co-immunoprecipitation is dependent on the presence of α-dystrobrevin but not β-dystrobrevin. The DAMAGE protein contains a potential nuclear localization signal, 30 12-amino acid repeats, and two MAGE homology domains. The domain structure of DAMAGE is similar to that of NRAGE, a MAGE protein that mediates p75 neurotrophin receptor signaling and neuronal apoptosis (Salehi, A. H., Roux, P. P., Kubu, C. J., Zeindler, C., Bhakar, A., Tannis, L. L., Verdi, J. M., and Barker, P. A. (2000) Neuron 27, 279-288). DAMAGE is highly expressed in brain and is present in the cell bodies and dendrites of hippocampal and Purkinje neurons. In skeletal muscle, DAMAGE is at the postsynaptic membrane and is associated with a subset of myonuclei. DAMAGE is also expressed in peripheral nerve, where it localizes along with other members of the dystrophin complex to the perineurium and myelin. These results expand the role of dystrobrevin and the dystrophin complex in membrane signaling and disease.

The dystrophin glycoprotein complex links the actin cytoskeleton to the extracellular matrix in skeletal muscle. In addition, recent studies (1)(2)(3)(4)(5)(6)(7) have shown that the dystrophin complex is critical for the localization of a variety of signaling molecules to the skeletal muscle sarcolemma. Thus, the dystrophin complex functions as both a structural link to the extracellular matrix and as a scaffold that concentrates and stabilizes functionally inter-related signaling proteins at the muscle cell membrane.
Part of the signaling function of the dystrophin complex may be mediated by the dystrobrevins. The role of the dystrobrevins in signaling is not fully understood; however, targeted disruption of the ␣-dystrobrevin gene in mice results in muscle pathology without the membrane fragility characteristic of dystrophin-deficient muscular dystrophy (8). These findings suggest that ␣-dystrobrevin is important for muscle function but in a non-structural way. Dystrobrevin has no enzymatic activity of its own; therefore, the involvement of dystrobrevin in signaling pathways is dependent on its interaction with other proteins.
The dystrobrevins are transcribed from two separate genes, designated ␣ and ␤, that share sequence homology with the C terminus of dystrophin and are composed of several protein interaction domains. These include two EF hands, a zinc binding domain (ZZ), two syntrophin-binding sites, a pair of coiledcoil domains that mediate the interaction of dystrobrevin with dystrophin/utrophin (9), and a dystrobrevin unique tail that can be phosphorylated (10 -12). In addition, alternative splicing of these protein interaction sites (13)(14)(15)(16) may regulate the association of dystrobrevin-interacting proteins.
In an effort to identify new ␣-dystrobrevin-interacting proteins, we used the yeast two-hybrid system to screen a mouse cDNA library. By using this approach, we identified an X-chromosome-encoded protein that contains two MAGE homology domains (MHD) 1 and multiple 12-amino acid repeats, which we have named DAMAGE (dystrobrevin-associated MAGE protein). The presence of this protein in neurons, at the neuromuscular junction, in perinuclear structures in skeletal muscle, and in myelin and perineurium of peripheral nerve expands the possible role of ␣-dystrobrevin and the dystrophin complex in membrane signaling and disease.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening-The nucleic acid sequence encoding mouse ␣-dystrobrevin-3 was subcloned into the pPC86BD yeast twohybrid vector (Invitrogen) by using standard methods. This bait plasmid when transfected into the Y190 yeast strain caused no autoactivation of the histidine and ␤-galactosidase reporter genes. Library screens * This work was supported by grants from the Muscular Dystrophy Association and National Institutes of Health (to S. C. F.) and a National Research Service Award (to D. E. A.). 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  were performed by sequential transfection of a mouse brain cDNA library in the pACT2 vector (Clontech), on leu Ϫ /trp Ϫ /his Ϫ plates containing 10 mM 3-amino-1,2,4-triazole. The prey plasmid was isolated from positive clones by standard methods, and the clones were sequenced to determine their identities. Each positive clone was tested for specificity of interaction by retransformation into the Y190 yeast strain alone, with the empty bait vector, or with the bait vector containing the PDZ domain of spinophilin, and was tested for growth on leu Ϫ /trp Ϫ /his Ϫ plates and for ␤-galactosidase activity. The clones were also tested in the more stringent CG1945 yeast strain to confirm the interaction. Clones that passed these tests were considered "positive" and were characterized further.
Isolation of Full-length DAMAGE cDNA Clones-The DNA sequence of the longest yeast clone obtained was used to search the NCBI data base using the BLAST search program. Mouse clone 4943 (Japanese Collection of Research Bioresources at the National Institute of Infections Diseases, Tokyo, Japan) was identified, sequenced using standard methods, and found to contain a full-length mouse DAMAGE sequence including a small portion of 5Ј-UTR. This clone was generously provided by Katsuyuki Hashimoto (Division of Genetic Resources, National Institute of Infectious Diseases, Tokyo, Japan).
Antibodies to DAMAGE were raised against keyhole limpet hemocyanin-conjugated peptides with the sequences SLVSQNSRRRRGGRA-NARR (SLVS), RVARPFRRPLFAEVAPELD (RVARP), and EDEAN-RAEAGRRPLIVRNLR (EDEAN) that correspond to amino acids 2-20, 672-690, and 890 -918, respectively, of the mouse DAMAGE sequence. The conjugated peptides were injected into New Zealand White rabbits on a boost and bleed schedule that followed established protocols (Covance Immunoresearch). Immune serum was affinity-purified on columns containing the immunizing peptide.
Immunoprecipitations, Immunohistochemistry, and Immunoblots-Whole tissue samples were prepared in 20 volumes of reducing SDS gel sample buffer in a Dounce homogenizer and boiled for 2 min. Insoluble material was removed by centrifugation at 10,000 ϫ g before separation of proteins by SDS-PAGE.
Dystrophin protein complexes were immunoprecipitated from mouse brain as described previously (20). Briefly, 5 unstripped mouse brains (PelFreeze) were homogenized in 50 ml of HB (10 mM sodium phosphate, pH 7.8, 400 mM NaCl, 5 mM EDTA, 1 mM EGTA, 0.02% sodium azide) in a Waring blender. The insoluble fraction was collected by centrifugation at 12,000 ϫ g and homogenized in 50 ml of HB using a Vertishear homogenizer. This fraction was centrifuged again at 12,000 ϫ g for 10 min, and the pellet was homogenized in HB containing 1% Triton X-100. The detergent-insoluble material was removed by centrifugation at 55,000 ϫ g, and the final supernatant was divided into appropriate volumes. Immunoprecipitations were performed using 10 g of antibody or control IgG. The immunocomplexes were isolated with protein G-Sepharose beads (Sigma) and were washed twice in HB, 1% Triton X-100, twice in HB, 0.1% Triton X-100, and once with HB containing protease inhibitors. The immune complexes were then boiled in reducing sample buffer and loaded onto 9% polyacrylamide gels and separated by SDS-PAGE.
Following SDS-PAGE, proteins were transferred to Immobilon nylon membranes (Millipore). Membranes were blocked in TBST containing 3% non-fat dry milk overnight at 4°C. Antigen detection was performed by overlaying blots with the appropriate antibody diluted to a final IgG concentration of 30 nM. Blots were then washed and incubated with either donkey anti-rabbit IgG or donkey anti-mouse IgG secondary antibodies conjugated to horseradish peroxidase (Jackson ImmunoResearch) diluted 1:8,000. Bands were visualized by enhanced chemiluminescence (ECL) using Super Signal West Pico ECL substrate (Pierce) and an AlphaInnotech gel documentation system (AlphaInnotech Corp.). For brain immunohistochemistry, C57 mice were sacrificed, and the brain was removed and grossly dissected into 2-5-mm 2 chunks that contained the hippocampus or cerebellum. These were incubated in 2% paraformaldehyde in PBS with 10% sucrose for 20 min and then in 25% sucrose overnight at 4°C. The tissue was then covered with OCT, frozen in liquid nitrogen, and cryosectioned at 20 m. For muscle and nerve immunohistochemistry, C57, mdx, and mice rendered null for ␣-dystrobrevin-1, -2, and -3 by targeted disruption of exon 3 of the ␣-dystrobrevin gene (␣Db Ϫ/Ϫ ) were sacrificed, and the quadriceps and soleus/ gastrocnemius muscles were rapidly removed, frozen in liquid nitrogencooled isopentane, and cryosectioned at 10 m. Cryostat sections were fixed in 2% paraformaldehyde, blocked with 3% bovine serum albumin, 1% fish gelatin, and 0.05% Tween 20 in PBS, and labeled with primary antibody at a concentration of 30 nM. The sections were then washed in PBS and incubated with goat anti-rabbit IgG conjugated to Alexa 568 diluted 1:500 or goat anti-mouse IgG conjugated to Alexa 488 diluted 1:500 (Molecular Probes), washed in PBS, and stained with Hoechst 33258. Labeled sections were viewed using a Zeiss Axioskop 2 equipped with a Zeiss Axiocam and Axiovision software or with a Leica TCS confocal microscope. Images were imported into Adobe PhotoShop for final cropping and adjustment of brightness and contrast.
Northern Blots-A mouse multiple tissue Northern blot purchased from Sigma was probed with a 32 P-labeled full-length DAMAGE cDNA. Hybridization and washing of the blot were done using PerfectHybe plus (Sigma) hybridization solution following the PerfectHybe plus protocol.

Isolation of DAMAGE cDNA and Properties of the Encoded
Protein-Yeast two-hybrid screening of a mouse brain cDNA library was performed using full-length mouse ␣Db-3 cDNA. Of 600,000 clones screened, 18 clones interacted with ␣Db-3 as determined by growth on leu Ϫ /trp Ϫ /his Ϫ plates and generation of ␤-galactosidase activity. Four of these clones interacted with the Gal4 binding domain alone or with the Gal4 binding domain fused to an unrelated sequence (the spinophilin PDZ domain). Sequencing of the remaining clones revealed that 9 of the 14 clones encoded overlapping parts of a protein that we have named DAMAGE.
A full-length clone of mouse DAMAGE, as well as several partial clones, were identified by a search of the GenBank TM EST data base. Similarly, the human DAMAGE orthologue was identified as an uninterrupted genomic sequence on the X chromosome. The human nucleotide sequence of ϳ3600 bp is located at Xq13.1 and likely encodes the entire mRNA for DAMAGE (see below). Furthermore, this sequence contains a single, uninterrupted open reading frame of 2874 nucleotides beginning at 208 ATG. Thus, DAMAGE appears to be encoded by an intronless gene. Although unusual, this trait is common among genes encoding MHD proteins (25). Additional mouse, human, and macaque genes encoding DAMAGE were found in GenBank TM under the accession numbers AB046807 (human), AB060233 (macaque), and AL672248.12 nucleotides 4959 -8423 (mouse). A mouse sequence (AF319978), called MAGE-E1, is identical to the mouse DAMAGE sequence but is missing 582 nucleotides of 5Ј-coding sequence (25). However, several other sequences named MAGE-E1 that encode alternatively spliced isoforms of a different MAGE protein are also present in GenBank TM (accession numbers AB056059, AB040527, AB040528, and AB040529) (49). Therefore, the name MAGE-E1 is ambiguous. In addition, hepatocellular carcinoma antigen-1 (HCA-1, AF490507) is identical to DAMAGE; however, this name is inadequate to describe a protein that is predominantly expressed in the central nervous system in non-pathological conditions (see below). The confusion generated by current nomenclature has prompted us to refer to the protein identified here as DAMAGE to emphasize its interaction with ␣-dystrobrevin and to distinguish it from MAGE-E1 cloned by Sasaki and co-workers (49).
A comparison of the human, mouse, and macaque sequences reveals that all three are highly homologous in their C-terminal portions, sharing at least 85% amino acid identity to the human sequence (Fig. 1). However, the mouse protein is divergent in the N-terminal half of the protein, which is primarily composed of repeats. In both macaque and human the repeated sequence is very pronounced and highly conserved. However, the mouse sequence contains only occasional identifiable elements of the repeats (Fig. 1A). This divergence leads to a 73% DNA and 62% amino acid identity between the mouse and human sequences, whereas the human and macaque sequences ␣-Dystrobrevin Interaction with DAMAGE/MAGE-e1 7016 are 93% identical at both the DNA and protein levels. The homology between mouse and human is very high in the Cterminal half of the protein (85% identical at the amino acid level). The homology is particularly high in the two MHDs suggesting that they are highly conserved across species. Finally, the first 12 amino acids, including a string of 4 arginines, are identical in all three species (Fig. 1A).
The sequence of DAMAGE contains several interesting structural features ( Fig. 2A). First, DAMAGE has homology to a family of proteins known as the melanoma-associated antigens (MAGE) protein family that has 29 members (25,26), all of which contain motifs known as MAGE homology domains (MHDs). DAMAGE is one of only a few proteins that contain two MHDs, located in the C-terminal half of the protein, and is most closely related to two other type II MHD proteins named necdin and MAGE-H1 (26).
Another interesting feature of DAMAGE is the presence of 30 contiguous 12-amino acid repeats, beginning at residue 62, with the consensus sequence A(S/T)EGPSTSVPPT (Fig. 1B). This consensus sequence does not appear to be related to any other known repeat sequence. The repeats are very prominent in the human sequence but are much less pronounced in the mouse sequence with only small portions of the repeated sequence readily identifiable in the latter (Fig. 1A). The DAM-AGE repeat domain may be hyperphosphorylated on serine and threonine as there are 30 sites in mouse and 39 in human that have a consensus sequence recognized by either protein kinase C or casein kinase II. Finally, the repeat region of DAMAGE contains numerous prolines, serines, threonines, and acidic residues. Sequences rich in these amino acids can target the protein for degradation by either the 26 S proteasome or cal- pain (27). A score greater than 5 using the PESTfind program suggests that a sequence is a genuine PEST domain. The human/macaque sequence has 7 potential PEST sequences with scores of 8.46, 18.86, 8.45, 14.57, 6.19, 6.07, and 11.08. The mouse sequence, with its less prominent repeats, has three PEST domains with scores of 10.58, 5.05, and 10.57.
Finally, a region with a high concentration of basic amino acids (SRRRRGGRANARR in mouse and SRRRRRRVAKATA in human and macaque) near the N terminus of DAMAGE may be a nuclear localization sequence (NLS). The basic residues in the mouse sequence are interrupted by several amino acids suggesting that the mouse NLS is bipartite, whereas human and macaque have larger monopartite basic NLSs. Neither sequence has a proline or acidic amino acid directly preceding the basic residues, indicative of a classic NLS. However, the sequence upstream of the basic residues is identical among the three species, demonstrating that it is conserved and suggesting that they may be involved in the NLS or some other function of DAMAGE. Although the functionality of these putative NLSs has not been tested, it is intriguing that DAMAGE can be found in nuclear and perinuclear locations (see below).
Regions of Interaction between ␣-Dystrobrevin and DAM-AGE-In an effort to define the binding site between ␣-dystro-brevin and DAMAGE, we tested truncated forms of DAMAGE for interaction with ␣-dystrobrevin-3 by yeast two-hybrid. The minimal portion of DAMAGE required for the interaction with ␣-dystrobrevin-3 is a 214-residue sequence that includes the second MHD as well as the C terminus of the protein (Fig. 2A). Shorter segments of DAMAGE were inactive, possibly because ␣-dystrobrevin interacts with multiple sites within this region or because the smaller segments do not fold properly.
␣-Dystrobrevin-3 was also truncated to determine the region involved in its association with DAMAGE. The only dispensable sequence was ϳ75 residues of the C terminus of the protein downstream of the ZZ domain (Fig. 2B). Thus, the ZZ domain may be involved in the association of the two proteins. However, because a large portion of ␣-dystrobrevin-3 is required for interaction (286 residues), tertiary structure may play an important role in the association of ␣Db-3 with DAMAGE.
Expression Pattern of DAMAGE mRNA-Analysis of DAM-AGE mRNA on a multiple mouse tissue Northern blot revealed a single transcript of ϳ3.6 kb (Fig. 3), consistent with a coding region of 2757 bp plus a 3Ј-UTR of 562 bp and a 5Ј-UTR estimated to be 300 bp. This is very similar to the expected size of the full-length transcript encoded on the human X chromosome. The transcript is expressed most abundantly in brain but

␣-Dystrobrevin Interaction with DAMAGE/MAGE-e1
is present at lower levels in all of the other tissues examined including heart, liver, kidney, spleen, testes, lung, thymus, placenta, and skeletal muscle.
Characterization of Anti-DAMAGE Antibodies-For studies of DAMAGE association with the dystrophin complex, we produced three antibodies against peptides derived from the mouse DAMAGE sequence. Affinity-purified antibodies were tested to determine whether they react with DAMAGE expressed in CHO cells, and secondarily with a protein of the expected size in brain and sciatic nerve homogenates (Fig. 4). All three of the antibodies recognize a 100-kDa protein in DAMAGE-transfected CHO cells, but the RVARP antibody reacts non-specifically with an unidentified 40-kDa protein present in untransfected cells (Fig. 4A). These experiments also demonstrate that the antibodies recognize smaller bands that are proteolytic fragments of DAMAGE because they are seen only in the transfected cells. In whole lysates from brain, all three antibodies recognize a protein of the size expected for DAMAGE (100 kDa) (Fig. 4B). Unfortunately, the RVARP antibody recognizes both DAMAGE and another protein of higher molecular weight (185 kDa, tentatively identified as periaxin) in peripheral nerve (Fig. 4C). Therefore, the SLVS antibody was used for the studies on peripheral nerve.

Association of DAMAGE with the Dystrophin-Dystrobrevin-Syntrophin Complex in Vivo-Immunoprecipitation and immunoblotting experiments were performed to determine whether DAMAGE associates with the dystrophin complex in vivo.
Brain extracts were used for these experiments because DAM-AGE mRNA is most abundant in this tissue. These studies were based, in part, on the knowledge that isoforms of syntrophin are tightly associated with ␣-dystrobrevin (17)(18)(19). Thus, proteins associated with ␣-dystrobrevin should be present in immune complexes isolated with antibodies to syntrophin or ␣-dystrobrevin. For these experiments, we used pan-specific mAb 1351 that recognizes ␣-, ␤1-, and ␤2-syntrophins (20) and mAb Db-23 against ␣-dystrobrevin. As expected, complexes isolated with either of these mAbs contain ␣-dystrobrevin (Fig.  5A). These protein complexes also contained a band of ϳ100 kDa recognized by the DAMAGE anti-peptide (RVARP) antibody. The anti-DAMAGE antibody also detects a lower band in these extracts that is probably a breakdown product of the full-length protein. We also performed the reverse experiment in which anti-DAMAGE was used to isolate protein complexes. Protein complexes isolated with anti-DAMAGE contained ␣-dystrobrevin (Fig. 5B). In each case, no immunoreactive band was seen in the control lane containing material immunoprecipitated with an unrelated, control antibody. In addition, preincubation of the DAMAGE antibody with a 5-fold excess of the immunizing peptide eliminated detection of the DAMAGE band, whereas incubation with a 5-fold excess of an unrelated peptide had no effect (Fig. 5C). Taken together, these results provide strong evidence that complexes of DAMAGE and ␣-dystrobrevin exist in vivo.
We used mice lacking ␣-dystrobrevin (␣Db Ϫ/Ϫ ) (8) to provide additional evidence for a specific interaction between DAM-AGE and ␣-dystrobrevin. Protein complexes were isolated with the Pan-syntrophin antibody from brain extracts prepared from wild type or ␣Db Ϫ/Ϫ mice. DAMAGE co-purified with syntrophin, presumably via their mutual association with ␣-dystrobrevin, from wild type extracts, but was absent from complexes isolated from ␣Db Ϫ/Ϫ brain (Fig. 6A). The amounts of syntrophin isolated were similar in the two cases, and the levels of DAMAGE in whole lysates from wt and ␣Db Ϫ/Ϫ brain are similar (Fig. 6B). Therefore, association of DAMAGE with the syntrophin-␣-dystrobrevin complex requires the presence of ␣-dystrobrevin. Furthermore, ␤-dystrobrevin co-purified with syntrophins, as expected from previous work (16). However, despite the presence of ␤-dystrobrevin, DAMAGE was absent. Thus, DAMAGE does not appear to interact with ␤-dystrobrevin. These results, taken together with the yeast twohybrid results, strongly support a direct and specific association of ␣-dystrobrevin and DAMAGE, mediated by a region of ␣-dystrobrevin upstream of the syntrophin binding domains and a site near the second MHD of DAMAGE.
DAMAGE Localization in Brain, Muscle, and Peripheral Nerve-In brain, DAMAGE is present in a select population of neurons. In particular, neurons of the dentate gyrus and the pyramidal cells of the hippocampus stain brightly for DAM-AGE, as do the Purkinje neurons of the cerebellum (Fig. 7). The staining is intense in the cell bodies and extends into the dendrites of the neurons. Identical staining patterns were observed with both the SLVS and EDEAN antibodies, and both were blocked by preincubation with the immunizing peptide. Unfortunately, our attempts to perform colocalization studies with DAMAGE and ␣-dystrobrevin in brain failed due to inconsistencies in antibody labeling with isoform-specific dystrobrevin antibodies. However, it has been well established that dystrophin complexes are present in neurons (28 -38), including those of the hippocampus and cerebellum that contain DAMAGE.
In order to determine whether DAMAGE and ␣-dystrobrevin co-localize, we turned to skeletal muscle where the distribution of ␣-dystrobrevin has been thoroughly characterized (8, 10, 12, 39 -42). Although ␣-dystrobrevin is highly expressed on the sarcolemma, we were not able to detect significant levels of FIG. 6. A, immunoprecipitations (IP) from C57 and ␣Db Ϫ/Ϫ mouse brain. Protein complexes were isolated using anti-syntrophin (1351) or a control IgG. Pan-Dystrobrevin, DAMAGE (RVARP), ␣-syntrophin, and dystrophin antibodies were used for detection. B, whole brain homogenates from C57 and ␣Db Ϫ/Ϫ mice demonstrate that C57 and ␣Db Ϫ/Ϫ mice have similar levels of DAMAGE.

␣-Dystrobrevin Interaction with DAMAGE/MAGE-e1
DAMAGE on the muscle membrane. This is consistent with the low expression level of DAMAGE in skeletal muscle. However, in muscle cross-sections, DAMAGE antibodies stain peripheral nerve intensely (Fig. 8A). This staining was particularly bright in the perineurium, a layer of epithelial cells that surrounds the nerve and forms the blood-nerve barrier. ␣-Dystrobrevin is also present in this structure (Fig. 8A), as is the dystrophin homologue utrophin (43,44).
Within the nerve, ␣-dystrobrevin and DAMAGE co-localize in a patchy, crescent-shaped distribution on the Schwann cell plasma membrane (Fig. 8A). The dystrophin short form, Dp116, as well as DRP2 are also present on the abaxonal Schwann cell plasma membrane (43). Based on its distribution, DAMAGE may associate with the dystrophin complex in multiple structures of peripheral nerve. DAMAGE is also present at neuromuscular synapses (Fig.  8A) where both ␣-dystrobrevin-1 and -2 are concentrated (40,42). Here, DAMAGE appears not to colocalize with acetylcholine receptors but instead occupies a submembrane compartment directly beneath the synapse (Fig. 9). In addition, the DAMAGE antibodies stain a subset of nuclei in muscle sec-tions, most of which are myonuclei (Fig. 8B). This is particularly apparent in the central nuclei of regenerating fibers of ␣Db Ϫ/Ϫ muscle (Fig. 8B). Few nuclei of invading cells of dystrophic muscle are labeled for DAMAGE. It is noteworthy that the Hoechst staining (used to reveal nuclei) and the DAMAGE immunostaining are not precisely coincident. Thus, DAMAGE may reside within nuclei and/or in perinuclear compartments.
Dystrobrevin-associated proteins, such as syncoilin and dysbindin, are up-regulated in dystrophin-deficient, mdx, mouse muscle (21,23). We examined the distribution of DAMAGE in mdx as well as ␣Db Ϫ/Ϫ muscle to determine whether there is any change in its distribution. No major difference in the distribution of DAMAGE at neuromuscular synapses or in peripheral nerves of either mdx or ␣Db Ϫ/Ϫ mice were found (Fig. 9).

DISCUSSION
In this study, we have identified a new dystrophin-associated protein, DAMAGE, that interacts with the N-terminal region of ␣-dystrobrevin. The association with ␣-dystrobrevin meets several criteria of specificity, including interaction by yeast twohybrid, co-immunoprecipitation, and co-localization in peripheral nerve and at NMJs. In addition, DAMAGE does not interact with ␤-dystrobrevin, a closely related homologue of ␣-dystrobrevin. Although the function of DAMAGE is unknown, the domain structure of DAMAGE provides several clues to its function. DAMAGE has two type II MAGE homology domains (MHDs), 30 contiguous 12-amino acid repeats, and a potential nuclear localization signal. Further analysis of these domains will be critical for determining the function of DAMAGE and its contribution to the signaling potential of dystrophin complexes in nerve, brain, and muscle.
MAGE Homology Domains-The MAGE proteins were first identified as intracellular proteins that contain a peptide sequence presented by T-cell receptors on the surface of melanoma cells (45). Many members of the MAGE protein family are expressed during development and are virtually absent in adult tissues. The exceptions in adults are the male germ cells and abnormally proliferating cells such as melanomas and other tumors. As a result, studies of the MAGE protein family have focused on their potential as tumor-specific antigens that could be exploited to fight cancer. The number of MHD proteins has grown to ϳ30 members, yet the function of MHD proteins is still obscure. However, recent findings demonstrate that some members of the MAGE protein family are expressed in adult tissues, and characterization of these proteins has uncovered some interesting functions of MAGE family members.
Studies of two MHD proteins, Necdin and NRAGE, have been particularly important in providing insight to the function of the MAGE protein family and suggest clues to the function of DAMAGE. NRAGE is a p75 NTR-binding protein (46) that regulates cell cycle progression and promotes apoptosis by enhancing the apoptotic effect of the p75 NTR (47,48). NRAGE also interacts with the Msx and Dlx families of homeodomain proteins that are involved in craniofacial, limb, and nervous system development (49). This interaction is mediated by repeat structures upstream of the MHD in NRAGE. Although the repeats in NRAGE are unrelated to those found in DAMAGE, the overall similarity in structure between DAMAGE and NRAGE suggests that they have common functions. A second MAGE protein, Necdin, is expressed in post-mitotic neurons and prevents cell cycle progression by inhibiting the transcription factor E2F1 (50). Necdin also has anti-apoptotic effects mediated by its interaction with p53 (51). In addition, necdin and MAGE H1, both of which have type II MHDs, FIG. 7. Immunofluorescence microscopy showing the localization of DAMAGE in neurons of the hippocampus and cerebellum in mouse brain. The staining pattern is identical with both the SLVS and EDEAN antibodies, and both are blocked by preincubation with the immunizing peptide. Scale bars, 100 m.

␣-Dystrobrevin Interaction with DAMAGE/MAGE-e1
interact with the type II death domain of p75 NTR. This result has raised speculation that type II MHDs, like those found in DAMAGE, may bind type II death domains in other proteins (52). These findings suggest that transcriptional regulation and control of cell differentiation and death may be common themes among MAGE proteins, and provide a starting point for future experiments that may reveal the function of DAMAGE.
The Repeat Domain-The 30 12-residue repeats in DAMAGE compose most of the N-terminal half of the protein. The DAM-AGE repeats have more than 30 consensus sequences for phosphorylation by protein kinase C and casein kinase II, suggesting that DAMAGE could become hyperphosphorylated. In addition, this region is highly enriched in the amino acids proline, glutamic acid, serine, and threonine and, as a consequence, has a high probability of being a PEST domain. PEST domains target proteins for rapid degradation by either the 26 S proteasome or calpain (27) and are typically present in proteins that are rapidly turned over. Even though the repeats are not highly conserved in the mouse sequence, the corre- sponding regions nonetheless score high as PEST sequences. Thus, the strong PEST sequence combined with the possibility of phosphorylation at many sites suggest that post-translational modification by kinases and phosphatases, as well as proteolysis, may play an important role in regulating DAM-AGE function.
Nuclear Localization Signal-The presence of a potential nuclear localization signal suggests that DAMAGE may translocate between the dystrophin-dystrobrevin complex on the membrane and the nucleus. The presence of a possible NLS combined with our observation that DAMAGE resides within nuclei and/or in a perinuclear compartment are consistent with this hypothesis. In addition, the precedent that other MAGE proteins interact with transcription factors (50,53,54) and regulate cell cycle progression and apoptosis (46 -48, 51, 55) suggests that DAMAGE may have similar functions.
DAMAGE Distribution in mdx and ␣Db Ϫ/Ϫ Mice-The distribution of DAMAGE in mdx and ␣Db Ϫ/Ϫ mice revealed no difference in the localization of DAMAGE to either the NMJs or peripheral nerves in either mutant. The rationale for examining the distribution of DAMAGE in mdx mouse muscle is based on the precedent that other dystrobrevin-associated proteins are up-regulated in mdx muscle (21,23). Although there was no increase in DAMAGE expression in mdx muscle, it is not surprising that DAMAGE still localizes to the NMJ and structures of peripheral nerves in mdx mice, because they still express utrophin as well as dystrophin short forms. However, the find-ing that DAMAGE localization is not altered in the ␣Db Ϫ/Ϫ mice suggests that DAMAGE may interact with other proteins that determine its distribution. This hypothesis is supported by the fact that DAMAGE is restricted to the NMJ in normal muscle even though there is an abundance of ␣-dystrobrevin on the sarcolemma of the muscle fiber. Thus it is likely that DAMAGE has multiple binding partners, only one of which is ␣-dystrobrevin.
Implications for Human Disease-Determining the function of dystrobrevin and its interacting proteins is particularly important in light of findings that implicate them in various pathological conditions. The gene encoding dysbindin has been linked to schizophrenia (56) and Hermansky-Pudlak syndrome type 7 (57). Syncoilin is up-regulated in mdx muscle (21) as well as in human desmin-related myopathy (58). Changes in ␤-dystrobrevin expression have been reported in heart failure (59), and a point mutation (P121L) in ␣-dystrobrevin has been linked to left ventricular noncompaction or Barth syndrome (60). In addition, mutations in the genes that encode dystrobrevin-interacting proteins are strong candidates for congenital muscular dystrophies in which dystrophin, the sarcoglycans, and the laminins are normal (61). These findings emphasize the importance of identifying new dystrobrevininteracting proteins, such as DAMAGE, and determining how they contribute to the function of dystrobrevin and the dystrophin complex.
The gene encoding DAMAGE is located on the human X chromosome in a region containing loci linked to mental retardation.

␣-Dystrobrevin Interaction with DAMAGE/MAGE-e1
Potential candidates include Wieacker-Wolff syndrome (62,63) and Miles-Carpenter syndrome (64), both of which exhibit a combination of mental retardation and distal muscle atrophy. In addition, defects in periaxin, a protein that associates with DRP2 in peripheral nerve, cause demyelination in mice and humans (43). Therefore, other members of the dystrophin complex in peripheral nerve, such as DAMAGE, may also be candidate genes for inherited peripheral neuropathologies.
Conclusions-The identification of all the components of the dystrophin complex will ultimately be critical for determining its full biological function. In this study, we have demonstrated a specific interaction between ␣-dystrobrevin and a novel MHD protein, DAMAGE. The structural elements of DAMAGE and its strong similarity to other MAGE proteins suggest that it augments the signaling capabilities of the dystrophin complex. In addition, the identification of DAMAGE as a dystrophin complex member, its presence in the central nervous system and peripheral nerves, and its localization on the X chromosome make DAMAGE a candidate gene for uncharacterized neuropathologies.