Microfibril-associated Glycoprotein-2 Interacts with Fibrillin-1 and Fibrillin-2 Suggesting a Role for MAGP-2 in Elastic Fiber Assembly*

Elastic fibers are composed of the protein elastin and a network of 10–12 nm microfibrils. The microfibrillar proteins include, among others, the fibrillins and microfibril-associated glycoproteins-1 and -2 (MAGP-1 and MAGP-2). Little is known about how microfibrillar proteins interact to support fiber assembly. We used the C-terminal half of MAGP-2 in a yeast two-hybrid library screen to identify relevant ligands. Six of 13 positive clones encoded known microfibrillar proteins, including fibrillin-1 and -2. Deletion analysis of partial fibrillin-1 and -2 clones revealed a calcium-binding epidermal growth factor repeat-containing region near the C terminus responsible for binding. This region is distinct from the region of fibrillin-1 reported by others to bind MAGP-1. The MAGP-2 bait was unable to interact productively with other epidermal growth factor repeats in fibrillin-1, demonstrating specificity of the interaction. Deletion analysis of the MAGP-2 bait demonstrated that binding occurred in a core region containing 48% identity and 7 conserved cysteine residues with MAGP-1. Immunoprecipitation of MAGP-2 from transfected COS-7 cells resulted in the coprecipitation of fibrillin. These results demonstrate that MAGP-2 specifically interacts with fibrillin-1 and -2 and suggest that MAGP-2 may help regulate microfibrillar assembly. The results also demonstrate the utility of the yeast two-hybrid system to study protein-protein interactions of the extracellular matrix.

like shape (6,7). Two fibrillins have been characterized (6,8). Both are 350-kDa molecules rich in 6-cysteine calcium-binding EGF 1 -like repeats and unique 8-cysteine repeats also found in the latent transforming growth factor-␤-binding proteins (9). The Marfan syndrome is an autosomal dominant connective tissue disorder linked to the fibrillin-1 gene on chromosome 15 characterized by vascular and skeletal abnomalities (10 -13). A similar disorder, congenital contractural arachnodactyly, is linked to the fibrillin-2 gene on chromosome 5 (10). Collectively, these disorders demonstrate the critical contribution of the fibrillins to tissue integrity.
Extraction of elastin-rich bovine nuchal ligament has resulted in the identification of a number of other microfibrillar proteins (14). Among others, these include microfibril-associated glycoprotein-1 (MAGP-1) and MAGP-2 (15,16). The requirement for reducing agents in their extraction suggests that intermolecular disulfide bonding is important for microfibril assembly. MAGP-1 and -2 share limited homology confined to a region containing 7 conserved cysteine residues located in the C-terminal half of the molecule. MAGP-2 also contains an RGD motif near its N terminus not conserved in MAGP-1, which mediates adhesion of various cell types (17). Ultrastructurally, microfibrils appear as beads on a string. Fibrillin is thought to comprise most or all of the mass of the "string," and immunolocalization studies have shown that MAGP-1 is found at the beads (18). Such studies have not been done for MAGP-2, although the structural homology with MAGP-1 implies that MAGP-2 may perform a similar function.
Various models have emerged regarding the arrangement of fibrillin molecules within microfibrils. Labeling studies have suggested that some cysteine residues in fibrillin are free outside the cell, as fibrillin becomes incorporated into a disulfidebonded network in the extracellular matrix (14,19). Disulfide bonding of fibrillin into dimers was suggested from immunoprecipitation studies (20), and it is now recognized that fibrillin molecules can form homodimers (21)(22)(23). Dimers can form between homologous regions of fibrillin, as well as between Nand C-terminal regions (24). Staining of microfibrils with monoclonal antibodies to fibrillin suggests that fibrillins are ordered in a head to tail parallel arrangement within microfibrils (7,25), which is supported by scanning transmission electron microscopy mass mapping and atomic force microscopy (26,27). Epitope mapping studies have suggested that stacking of fibrillin molecules within microfibrils occurs in an unstaggered arrangement (25). Alternatively, transglutaminase cross-link data have suggested a staggered arrangement of fibrillin molecules (28), and the ability of N-and C-terminal regions to dimerize supports this model (24). A staggered arrangement is also suggested from solution structure models (29). It is unknown whether both fibrillins can contribute to a single microfibril.
Little is known about the involvement of the other microfibrillar proteins in fiber assembly, with few exceptions. MAGP-1 has been proposed to bind a C-terminal region of tropoelastin, the soluble precursor of elastin, as antibodies to either MAGP-1 or tropoelastin which disrupt this interaction prevent deposition of tropoelastin into the fiber (30,31). However, other studies (32) have indicated that although MAGP-1 can bind intact tropoelastin, N-and C-terminal halves of elastin generated by proteolytic digestion are unable to bind MAGP-1 alone, suggesting a complex mechanism of interaction. Both fibrillin-1 and fibrillin-2 interact with tropoelastin through a region near the N terminus of fibrillin (33). Chondroitin sulfate proteoglycans have been shown to associate with fibrillin at the beads, with chondroitinase treatment disrupting the bead structure (34). The chondroitin sulfate proteoglycan decorin has also been shown to interact with MAGP-1, and the two can form a ternary complex with fibrillin-1 (35). Versican can also interact directly with fibrillin in a calcium-dependent manner and has been suggested to form a bridge between fibrillin-containing microfibrils and other connective tissue networks (36). Certain fibrillin-1 peptides have been found to bind fibulin-2, another putative microfibrillar component (37).
In this study, we used the yeast two-hybrid system to identify ligands for MAGP-2, a recently described microfibrillar protein. The C-terminal half of MAGP-2 was found to interact with fibrillin-1 and -2, as well as fibulin-1, another component of elastic fibers. These findings suggest that MAGP-2 may be important in the assembly of microfibrils.

EXPERIMENTAL PROCEDURES
Construction of the MAGP-2 Bait Vector-An EST clone encoding mouse MAGP-2 (GenBank TM accession number AA153960) was identified and sequenced in its entirety (38). A 568-bp PstI fragment beginning at bp 279 in the cDNA and encoding the last 92 amino acids of the protein was subcloned in-frame into the yeast two-hybrid GAL4 DNA binding domain vector pGBT9 (CLONTECH).
Yeast Two-hybrid Library Screening-A mouse 17-day total embryo cDNA library in the two-hybrid GAL4 activation domain vector pGAD10 (CLONTECH ML4006AB) was amplified following the manufacturer's instructions. For each library screen, 1 mg of DNA binding domain/bait vector and 0.5 mg of activation domain/library vector were cotransformed into 8 ml of yeast competent cells. Transformation mixtures were spread on plates lacking tryptophan and leucine to select for the bait and library plasmids, respectively, and to determine the cotransformation efficiency. A total of ϳ250,000 cotransformed colonies were screened for bait/ligand interactions by plating on media lacking leucine, tryptophan, and histidine supplemented with 5 mM 3-amino-1,2,4-triazole, which results in selection of interacting clones by expression of the HIS3 reporter. Colonies growing on plates lacking all three amino acids were further screened for ␤-galactosidase activity (a second reporter) using a colony lift assay and following the manufacturer's instructions. Yeast DNA was harvested from double-positive clones and used to transform bacterial KC8 cells to recover GAL4 activation domain/library cDNA plasmids. Inserts from all isolated clones were sequenced from both ends using vector-specific primers. The identity of individual clones was determined using the BLAST program to screen the GenBank TM .
Two additional constructs for MAGP-2 were generated in the pGBT9 yeast two-hybrid DNA binding domain vector. The first, encoding a region containing the 7 cysteine residues conserved with MAGP-1, was constructed by digesting the MAGP-2 bait construct (described above) with StuI and MscI, removing the 339-bp fragment encoding the last 28 amino acids of the protein, and religating. A construct expressing these last 28 residues was generated by digesting the mouse MAGP-2 cDNA with StuI and SspI and subcloning the 442-bp fragment into the SmaI site of pGBT9. DNA binding domain and activation domain constructs were used to cotransform yeast following the manufacturer's instructions, selected on media lacking leucine and tryptophan, and assayed for ␤-galactosidase activity as outlined below.
Liquid ␤-Galactosidase Assays-Colonies from the original library screen that grew on media lacking histidine were assayed for ␤-galactosidase activity. Colonies were restreaked on triple dropout media. Three colonies from each restreaked plate (from one original colony) were grown overnight in liquid selection media (ϪLeu, ϪTrp). These cultures were used to reinoculate YPD media and grown to log phase the day of the assay. Cells were centrifuged and washed in Z buffer, lysed by three freeze/thaw cycles, and diluted in Z buffer containing 0.24% 2-mercaptoethanol. The reaction was started by the addition of O-nitrophenyl ␤-D-galactopyranoside (ONPG) to a final concentration of 0.67 mg/ml. Reactions were stopped by addition of 0.4 ml of Na 2 CO 3 after yellow color development, and supernatants were read at A 420 . ␤-Galactosidase activity was expressed in units, with 1 unit defined as the amount that hydrolyzes 1 mol of ONPG per min per cell.
MAGP-2 and Fibrillin-1 Mammalian Expression Constructs-The mouse MAGP-2 EST clone contained an XhoI cloning site at its 5Ј end. A C-terminal Myc epitope tag fusion was generated by amplifying with a T7 primer (flanking the 5Ј-cloning site) and the primer 5Ј-CTTACT-CTCGAGTCATTCATTCAAGTCCTCTTCAGAAATGAGCTTTTGCTC-CATCAGACCATCGGGTCTCTGCA-3Ј, which includes a Myc tag, a stop codon, and an XhoI site 3Ј to the Myc tag. The PCR product (ϳ600 bp) was digested with XhoI and subcloned into the pCAGGS expression vector (39). A human fibrillin-1 minigene construct (mFib-1) was used that contains an internal deletion of exons 16 -49. Transfection, Metabolic Labeling, and Immunoprecipitation-COS-7 cells were grown in Dulbecco's modified Eagle's medium containing 5% fetal calf serum. 5 ϫ 10 5 cells were plated into 60-mm dishes 16 h prior to transfection, and cells were transfected using a DEAE-dextran procedure followed by a 3-h treatment with 100 M chloroquine (40,41). Cells were transfected with either the MAGP-2 construct (1 g), fibrillin construct (10 g), both constructs (1 g of MAGP-2 ϩ 9 g of fibrillin), or mock-transfected (no DNA). Cells were metabolically labeled 60 h post-transfection. Briefly, cells were starved for 1 h in Dulbecco's modified Eagle's medium lacking cystine (Invitrogen) and labeled overnight in Dulbecco's modified Eagle's medium minus cystine supplemented with 5% dialyzed fetal calf serum (Invitrogen) and 50 Ci/ml [ 35 S]cysteine (ICN). Following labeling, media were collected, and 10ϫ immunoprecipitation buffer was added to bring the media to 1ϫ (25 mM Tris, pH 7.4, 150 mM NaCl, 1 mM CaCl 2 , 1% Triton X-100, 5 mg/ml bovine serum albumin). Media were precleared by incubation for 1 h at room temperature with 50 l of protein A-Sepharose (Zymed Laboratories Inc.), followed by incubation overnight at 4°C with either 5 g of anti-Myc monoclonal antibody (Invitrogen) or 5 l of antiserum raised against human fibrillin-1. Immune precipitates were collected by incubation with 50 l of protein A-Sepharose for 1 h followed by centrifugation for 30 s at 15,000 ϫ g. Immunoprecipitates were washed 5 times with wash buffer (0.4% Triton X-100, 10 mM EDTA, 1 mg/ml bovine serum albumin), eluted by boiling in 1ϫ SDS-PAGE sample buffer containing 2-mercaptoethanol, run on 10% SDS-PAGE, and analyzed by fluorography with Amplify (Amersham Biosciences) following the manufacturer's instructions.

RESULTS
The yeast two-hybrid system was used to identify proteins within elastic microfibrils that bind MAGP-2, as well as other potential ligands. The MAGP-2 bait is shown in Fig. 1. The homology between MAGP-1 and MAGP-2 was confined to the C-terminal half of the molecule, which defines the bait (arrow). Based on the homology, we would predict that the two MAGPs might bind to common ligands through C-terminal domains and unique ligands through N-terminal domains. A 17-day mouse embryo cDNA library in the two-hybrid GAL4 activation domain vector pGAD10 (CLONTECH) was chosen for screening based upon in situ hybridization data indicating that many elastic fiber proteins, which represent likely potential ligands, are abundantly expressed at this stage in the developing mouse (not shown).
By using the C-terminal half of mouse MAGP-2 as the bait, 30 clones of 250,000 screened were found to interact with the bait based on the ability to activate the HIS3 reporter via the GAL4-responsive GAL1 promoter, resulting in growth on media lacking histidine. The yeast strain used, HF7c, also contains a lacZ reporter under the control of three GAL4-binding sites. Of these 30 clones, 13 were also visibly positive for ␤-galactosidase activity using a colony lift assay. Plasmids encoding the respective ligands, as GAL4 activation domain fusion constructs, were isolated from each of these yeast clones. Upon sequencing, 11 of the 13 double positive clones were identified as previously cloned cDNAs by BLAST sequence similarity searching of the GenBank TM data base. The identity of these cDNAs are shown in Table I. Six of the 13 clones encoded known microfibrillar proteins (fibrillin-1, fibrillin-2, and fibulin-1), with five of these encoding fibrillin-1 or -2. All of these proteins contain multiple copies of EGF-like repeats. Two of the other clones identified, jagged-1 and MEGF-6, also contain similar repeats, implying that MAGP-2 interacts with EGFlike motifs within these proteins.
The 5Ј and 3Ј boundaries of the fibrillin-1 and -2 clones identified in the screen were determined, and the region of fibrillin encoded by each clone is shown in Fig. 2A. In all cases, the clones corresponded to EGF repeat containing regions located near the C terminus of fibrillin. The fact that similar regions of fibrillin-1 and -2 were identified using the C-terminal MAGP-2 bait suggests that the site(s) on fibrillin involved in binding is highly conserved between fibrillin-1 and -2. Interestingly, none of the clones contained the C-terminal domain of fibrillin, composed of ϳ182 residues, 50% of which are completely conserved between fibrillin-1 and -2. This C-terminal domain is encoded by exons 64 and 65, the final two exons. We are not aware of any mRNA forms of either fibrillin-1 or -2 reported to terminate 5Ј to this C-terminal domain. Although it is possible that the 3Ј end of the fibrillin clones identified in the screen were the result of nuclease cleavage, the finding that all of these clones lack the material that encodes the C-terminal domain suggests that similar truncated forms of the fibrillins may exist in vivo.
Based on the identification of the EGF repeat-containing proteins jagged-1 and MEGF-6 as MAGP-2 ligands, it appeared likely that one or more EGF repeats in fibrillin-1 and -2 were responsible for MAGP-2 binding. We therefore generated deletion constructs of the fibrillin-1 and -2 clones isolated in the screen and tested these individually for the ability to bind MAGP-2 in the two-hybrid assay. The fibrillin-1 clone identified (clone 8) contained two discrete regions of EGF-like repeats, one with 5 tandem repeats and another with 7 tandem repeats. In addition, this clone contained two separate 8-cysteine domains that are unique to the fibrillins and latent transforming growth factor-␤-binding proteins. The smallest of the fibrillin-2 clones identified contained one 8-cysteine domain as well as the domain containing 7 tandem EGF-like repeats. When these regions were tested individually for the ability to bind MAGP-2, only the region containing the C-terminal 7 tandem EGF-like repeats conferred binding (Fig. 2B). This was true for both fibrillin-1 and -2. Although the mouse embryo library that was screened was primed with a combination of both oligo(dT) and random primers, we were unaware of whether or not clones encoding fibrillin fragments encompassing other parts of the molecule were present in the library. We therefore made additional fibrillin-1 constructs to investigate potential binding of MAGP-2 to other parts of fibrillin. Three constructs were made that encompass exons 5-15, 16 -38, and 39 -44 of mouse fibrillin-1. Together with the fibrillin-1 clone isolated in the library screen, these constructs cover more than 80% of the mature molecule. The relative ability of these fibrillin fragments to interact with the MAGP-2 bait was assessed in the yeast two-hybrid assay using the ␤-galactosidase reporter (Fig. 2B). Relative to the clone isolated in the library screen, none of the three new constructs were able to interact significantly with the MAGP-2 bait. These three constructs collectively contain 26 EGF-like repeats. The inability of the these constructs as well as the 5 EGF constructs of fibrillin-1 to bind MAGP-2 strongly suggests that the interaction with the 7 EGF region is specific. This region is distinct from the N-terminal region of fibrillin-1 reported by others to bind MAGP-1 (32).
As the MAGP-2 bait used to screen the library contains a  core region containing significant homology with MAGP-1, as well as a C-terminal region with no homology to MAGP-1, we investigated which region was responsible for binding. Upon expression of these regions separately in the two-hybrid assay with the fibrillin-1 and -2 clones identified in the screen, the binding activity is clearly localized to the region containing homology with MAGP-1 (Fig. 3). This suggests that MAGP-1 and -2 interact with fibrillin through this conserved region containing 7 common cysteine residues. In order to verify that MAGP-2 and fibrillin can interact in the context of mammalian cells, we coexpressed both proteins in COS-7 cells (Fig. 4). Fibrillin was expressed as a minigene lacking exons 16 -49 but retaining exons 50 -63 that were shown in the two-hybrid assay to contain MAGP-2 binding activity. Upon transfection of COS cells with this construct, antibodies to human fibrillin-1 were able to immunoprecipitate specifically a protein of the predicted size (ϳ160 kDa) following metabolic labeling with [ 35 S]cysteine (2nd lane). As antibodies to MAGP-2 were not available, we expressed MAGP-2 as a fusion protein containing a C-terminal Myc epitope tag. An anti-Myc epitope antibody was able to immunoprecipitate the MAGP-2/Myc fusion protein (4th lane), but was unable to immunoprecipitate fibrillin, as expected (3rd lane). Following cotransfection of the two constructs, immunoprecipitation of the MAGP-2/Myc fusion protein with the anti-Myc antibody resulted in the coprecipitation of fibrillin as well (5th lane). DISCUSSION Although the importance of the fibrillins in the biology of microfibrils and extracellular matrix has been well established (10 -13), the role of most other microfibrillar proteins in microfibril assembly, including MAGP-2, has not been defined. The goal of this study was to use the yeast two-hybrid system to identify proteins within microfibrils that interact with MAGP-2 in order to better understand the role of this protein in elastic fiber assembly. The C-terminal half of MAGP-2, which is the only region of the molecule that contains similarity with MAGP-1, was found to interact with fibrillin-1 and -2. Interaction between MAGP-2 and fibrillin-1 was confirmed by coexpression studies in COS-7 cells, where immunoprecipitation of MAGP-2 resulted in the coprecipitation of fibrillin.
In addition to fibrillin-1 and -2, the C-terminal MAGP-2 bait FIG. 2. Identification of the fibrillin domain, which interacts with MAGP-2. A, the domain structure analogous for both fibrillins is shown. The region encoded by each of the fibrillin-1 and fibrillin-2 clones isolated in the twohybrid library screen is shown below the domain structure (arrows). The regions encoded by 7 additional constructs used in B (exons 5-15, 16 -38, 39 -44, 8 cysteine domains, and tandem EGF repeats) are shown above the domain structure. B, yeast coexpressing the MAGP-2 bait with various fibrillin-1 or fibrillin-2 activation domain constructs were assayed for ␤-galactosidase reporter activity using a liquid assay. One unit of activity is defined as the amount of enzyme required to hydrolyze 1 mol of ONPG to O-nitrophenol and D-galactose per min per cell. Assays were done on three colonies expressing each hybrid pair, with error bars representing S.D. between these values. pTD1 ϩ pVA3 is a positive control for the yeast two-hybrid assay.
was also able to interact with three other proteins containing EGF-like repeats, including fibulin-1, jagged-1, and MEGF-6. We have localized binding of MAGP-2 to a C-terminal region of fibrillin containing 7 tandem calcium-binding EGF-like repeats. We believe this interaction is specific for two reasons. First, many other proteins are known to contain EGF-like repeats, and many of these are likely expressed in a 17-day mouse embryo, yet were not isolated in our yeast two-hybrid screen. More importantly, we made three additional fibrillin-1 constructs containing regions that were not part of the clones isolated in the screen. Collectively, these three constructs contained a total of 26 additional EGF-like repeats, and none were able to bind MAGP-2 effectively in the two-hybrid assay. A domain of fibrillin-1 containing 5 tandem EGF repeats was also unable to interact with the MAGP-2 construct. It is not easy to predict which of the EGF-like repeats in fibrillin-1 and -2 might be involved in binding MAGP-2, as all of the 7 repeats are highly conserved between the fibrillins (75-80%). Identification of the precise binding domain(s) within the fibrillins will be the focus of future experiments.
Deletion analysis of the MAGP-2 bait used in our studies revealed that the fibrillin binding domain was localized to a region containing 7 conserved cysteine residues with MAGP-1. Segade et al. (45) have localized the matrix binding domain of MAGP-1 to the same conserved region. This study investigated MAGP-1 domains responsible for incorporation of MAGP-1/ green fluorescent protein fusion proteins into the matrix elaborated by RFL-6 cells (rat fetal lung fibroblasts). They found that a MAGP-2/green fluorescent protein fusion protein was unable to be incorporated into this matrix, unless a critical cysteine residue from MAGP-1 was substituted into the MAGP-2 backbone. As this is not consistent with our finding that MAGP-2 can bind both fibrillin-1 and -2 through this conserved domain, we speculate that the MAGP-1 ligand within the RFL-6 matrix was not fibrillin but was perhaps a member of the latent TGF-␤-binding protein family or another potentially unique ligand for MAGP-1.
The observation that MAGP-2 can bind fibrillin raises interesting questions about the potential function of MAGP-2 in microfibril assembly. It is currently unknown whether the N-terminal half of MAGP-2, which shares no similarity with MAGP-1, may bind one or more microfibrillar proteins. However, this region of MAGP-2 does contain an RGD motif not found in MAGP-1 that has been shown to mediate its binding to cells through the ␣ v ␤ 3 integrin (17). It is important to note that several fibrillin-1 mutations giving rise to Marfan syndrome have been mapped to the 7 EGF repeats that we have shown bind MAGP-2 (46). It will be interesting to determine whether MAGP-2 colocalizes to microfibrils in these patients. At least three models have been proposed describing possible fibrillin arrangement in microfibrils, taking into account immunostaining patterns of a number of monoclonal antibodies that recognize various regions of fibrillin (46). In all cases, beaded structures are proposed at the C termini of fibrillin monomers/ dimers. As MAGP-1 and MAGP-2 share considerable homology within the C-terminal half of the molecule, and MAGP-1 localizes to beaded structures within microfibrils (18), we speculate that MAGP-2 will also be found to localize to the beads. This would be consistent with the location of some beads covering the C terminus of fibrillin and our observations that MAGP-2 binds this region of fibrillin. Others have shown (32) that

FIG. 3. Identification of the MAGP-2 domain that interacts with fibrillin.
A schematic of the full-length MAGP-2 molecule, the back half used as bait in the two-hybrid screen, the region containing 7 cysteine residues conserved with MAGP-1, and the unconserved C-terminal (C-term) region is shown at the bottom. Yeasts coexpressing either the fibrillin-1 (number 8) or -2 (number 30) GAL4 activation domain clones with MAGP-2 GAL4 DNA binding domain constructs were assayed for ␤-galactosidase reporter activity using the liquid assay described in Fig. 3. Assays were done on three colonies expressing each hybrid pair, with error bars representing S.D. between these values. pTD1 ϩ pVA3 is a positive control for the yeast two-hybrid assay. MAGP-1 binds to an N-terminal region of fibrillin, also consistent with all three models of fibrillin alignment. Future experiments will address whether MAGP-1 and -2 colocalize within individual beads or whether they localize to alternating beads.
Based on analogy to MAGP-1, it is interesting to speculate on other ternary complexes that might be formed via bridging interactions of MAGP-2. MAGP-1 has been shown to bind both tropoelastin and type VI collagen, although the binding domain on MAGP-1 involved appears to located near the N terminus, in a region that shares no homology with MAGP-2 (30,31,47). Chondroitin sulfate proteoglycans have been shown to associate with fibrillin (34), and MAGP-1 has been shown to bind the chondroitin sulfate proteoglycan decorin as well (35). Decorin could not inhibit the interaction of MAGP-1 and type VI collagen, implying that MAGP-1 does not bind decorin through its N terminus (47). As the C-terminal halves of MAGP-1 and -2 are largely conserved, this raises the possibility that MAGP-2 could potentially bind proteoglycans as well, and that a ternary complex might be formed between MAGP-2, fibrillin, and proteoglycans within microfibrils. The presence of an integrinbinding RGD motif near the N terminus of MAGP-2 raises the possibility that MAGP-2 may serve a unique function within microfibrils as a bridging protein between microfibrils and the cell surface (17).
Interactions between MAGP-2 and other proteins identified here could have physiological relevance as well. The interaction with fibulin-1 is obviously relevant, as fibulins have been suggested to be microfibrillar components in certain cases (37,48,49). Jagged-1 and glycogenin are expressed in skeletal muscle, a place where MAGP-2 is prominently expressed (38). The interaction with von Willebrand factor is particularly interesting, as platelets have been shown to interact with the subendothelial matrix through vWF (50,51). What is a matter of controversy is whether this interaction is through type VI collagen, thrombospondin-associated microfibrils, or fibrillinassociated microfibrils. Interaction between MAGP-2 and vWF supports a role for fibrillin-associated microfibrils in this process. We are currently trying to verify these interactions in mammalian expression systems.
As MAGP-2 and the fibrillins are extracellular matrix proteins, it was somewhat surprising that they could interact within the nucleus of yeast, as it has been suggested that post-translational modifications such as glycosylation and sulfation, which occur in the secretory pathway, are important for interactions between microfibrillar proteins (35). The yeast two-hybrid system has been used successfully to identify interactions with other extracellular proteins including type IV collagen (52), thrombospondin 1, the laminin ␤3 chain, and the NCI domain of type VII collagen (53), as well as the microfibrillar protein emilin (54). Thus, the yeast two-hybrid system should prove useful in identifying interactions between other microfibrillar proteins as well.