The extracellular matrix protein MAGP-2 interacts with Jagged1 and induces its shedding from the cell surface.

Elastic fibers are composed of the protein elastin and a network of 10-12-nm microfibrils, which are composed of several glycoproteins, including fibrillin-1, fibrillin-2, and MAGP1/2 (microfibril-associated glycoproteins-1 and -2). Although fibrillins and MAGPs covalently associate, we find that the DSL (Delta/Serrate/LAG2) protein Jagged1, an activating ligand for Notch receptor signaling, also interacts with MAGP-2 in both yeast two-hybrid and coimmunoprecipitation studies. Interaction between Jagged1 and MAGP-2 requires the epidermal growth factor-like repeats of Jagged1. MAGP-2 was found complexed with the Jagged1 extracellular domain shed from 293T cells and COS-7 cells coexpressing full-length Jagged1 and MAGP-2. MAGP-2 shedding of the Jagged1 extracellular domain was decreased by the metalloproteinase hydroxamate inhibitor BB3103 implicating proteolysis in its release. Although MAGP-2 also interacted with the other DSL ligands, Jagged2 and Delta1, they were not found associated with MAGP-2 in the conditioned media, identifying differential effects of MAGP-2 on DSL ligand shedding. The related microfibrillar protein MAGP-1 was also found to interact with DSL ligands but, unlike MAGP-2, was unable to facilitate the shedding of Jagged1. Our findings suggest that in addition to its role in microfibrils, MAGP-2 may also affect cellular differentiation through modulating the Notch signaling pathway either by binding to cell surface DSL ligands or by facilitating release and/or stabilization of a soluble extracellular form of Jagged1.

Underscoring the importance of microfibrils are human disease-associated mutations in microfibril components. The Marfan syndrome is an autosomal dominant disorder linked to fibrillin-1 (19) with manifestations that include dilatation and dissection of the aorta, mitral valve prolapse, tall stature, scoliosis, sternal deformities, ectopia lentis, and myopia. Congenital contractural arachnodactyly is linked to fibrillin-2 (20) and involves contractures of the large joints and abnormal pinnae formation.
The amino acid similarity between MAGP-1 and MAGP-2 (25%) is limited to a region within the C-terminal half of the protein containing 7 conserved cysteine residues. The N-terminal half of MAGP-2 shares no identity with MAGP-1, contains an RGD motif that has been shown to mediate adhesion and spreading of numerous cell types through the ␣ v ␤ 3 integrin (17), and perhaps mediates interaction of microfibrils with cell surfaces at specific stages of development and differentiation. We considered that the common functions of MAGP-1 and MAGP-2 are likely mediated through the conserved C-terminal region. Therefore, to identify interacting proteins for MAGP-2, we used MAGP-2 C-terminal sequences as bait in a yeast two-hybrid screen (21). Known elastic fiber proteins, including fibrillins-1 and -2 as well as fibulin-1, were identified in our screen, validating the use of this assay to identify relevant interacting proteins for MAGP-2.
We also identified the DSL 1 (Delta/Serrate/LAG-2) protein Jagged1 as a putative MAGP-2 interacting protein. Jagged1 belongs to a family of cell-surface, single-pass transmembrane ligands that activate Notch receptor signaling (22). Notch signaling directs cell fate specification in numerous cell types, and unlike many signaling systems, Notch mediates signaling between adjacent cells because of the transmembrane tethering of both Notch receptors and DSL ligands (23). The extracellular domain of Jagged1 is involved in receptor binding and consists mainly of 16 tandem epidermal growth factor-like (EGF-like) repeats. In this regard, we have demonstrated that fibrillin EGF-like repeats mediate the interaction between MAGP-2 and fibrillin. Consistent with the yeast two-hybrid interaction reported previously for MAGP-2 and Jagged1, many of the tissues and structures reported to express MAGP-2 (24) also express Jagged1 (25,26). Although Jagged1 is widely expressed during development, overlap between MAGP-2 and Jagged1 in blood vessels is particularly intriguing given the angiogenic remodeling defect displayed by Jagged1 knock-out mice (27). The expression patterns shared by MAGP-2 and Jagged1 suggest potential functional consequences of the previous yeast two-hybrid interaction reported for MAGP-2 with Jagged1.
In this study, we confirm that MAGP-2 interacts with Jagged1 in mammalian cells, and we identify the EGF-like repeats of Jagged1 as the basis for its interaction with MAGP-2. We also show that MAGP-2 increases the amount of soluble Jagged1 released from the cell surface in a metalloproteinase-dependent manner. Although we find that MAGP-2 has the capacity to interact with other members of the DSL family of Notch ligands, augmentation of shedding appears specific for Jagged1. In addition, whereas its structural relative MAGP-1 was found to bind DSL ligands, MAGP-1 was unable to effect shedding of Jagged1 from the cell surface. Our findings suggest that beyond a structural role in microfibrils, MAGP-2 may function as a biological modulator of Notch signaling, and potential mechanisms by which MAGP-2 could affect ligandinduced Notch signaling are discussed.

EXPERIMENTAL PROCEDURES
Constructs-Jagged1 deletion mutants were constructed using standard molecular biology techniques. In brief, the amino acids deleted in each of the mutants is as follows: ⌬CD, amino acids 1103-1213; ⌬CR, amino acids 863-957; ⌬DSL, amino acids 92-250; and ⌬EGFR/CR, amino acids 295-1037. Details are available upon request.
Antibodies-Rabbit polyclonal anti-Jagged1 and anti-Delta1 antibodies were generated by cloning Jagged1 extracellular sequence encoding amino acids 35-360 (PCR8), Jagged1 intracellular sequence encoding amino acids 1102-1219 (J59), or Delta1 extracellular sequence encoding amino acids 150 -350 (148G) into the pGEX vector (Amersham Biosciences), which creates a glutathione S-transferase fusion protein. Transformed bacteria were induced to produce fusion protein, and resultant proteins were isolated and purified by glutathione-agarose (Amersham Biosciences) and injected into rabbits. An antibody to the N-terminal half of murine MAGP-2 has been described previously (28). An anti-HA monoclonal antibody was also used for immunoprecipitations (Covance).
Radioimmunoprecipitations-COS-7 or 293T cells were plated at a density of 5 ϫ 10 5 cell per 60-mm dish. Cells were transfected with a total of 5 g of plasmid using Lipofectin (Invitrogen) following the manufacturer's instructions. When two constructs were used, 2.5 g of each was used, and controls in these cases included 2.5 g of test plasmid plus 2.5 g of control vector. Following transfection for 6 h, transfection media were replaced with normal growth media (Dulbecco's modified Eagle's medium ϩ 10% fetal calf serum). 31 h following transfection, cells were starved for 1 h in Dulbecco's modified Eagle's medium minus cysteine ϩ 5% dialyzed fetal calf serum. Cells were labeled overnight (18 h) in 1.5 ml of the same media containing 50 Ci/ml [ 35 S]cysteine. In experiments using PMA (100 M, Sigma P8139) or the metalloproteinase inhibitor BB3103 (5 M), each was added for the entire labeling period. Cell extracts or media were immunoprecipitated using 5 l of antibody and protein A-Sepharose (Zymed Laboratories Inc.) as described (21).
Immunoprecipitation of Jagged1 ICD Fragments-COS-7 cells were plated on 60-mm dishes and transfected via Lipofectamine (Invitrogen) using 1.6 g of total DNA and 4 l of Lipofectamine reagent following the manufacturer's instructions. Within these parameters, each transfection mixture contained 0.4 g of Jagged1 and either 0.4 g of MAGP-2 or vector control plasmid. The day after transfection, monolayer cultures were serum-starved overnight and then treated for 5 h with a proteasome inhibitor, MG132 (BioMol), at a final concentration of 10 M. Cell lysates were collected in lysis buffer as described for radioimmunoprecipitations and immunoprecipitated with J59 antiserum for 1 h on ice. Immunoprecipitates were collected on protein-Aagarose (Invitrogen), washed as for radioimmunoprecipitates, and Western-blotted using an anti-HA monoclonal antibody (12CA5).

MAGP-2
Interacts with the DSL Ligand Jagged1-In a previous study, we used the C-terminal MAGP domain shared by MAGP-1 and -2 as a bait in a yeast two-hybrid screen for interacting proteins. In addition to the microfibrillar proteins fibrillin-1 and -2, the screen also identified Jagged1 as a potential interacting protein with MAGP-2, even though Jagged1 is cell-associated and not known to be associated with the extracellular matrix. However, some evidence indicates that MAGP-2 can be cell-associated as well as microfibril-associated. In fact, MAGP-2 has an RGD motif in the N terminus that facilitates binding of the ␣ v ␤ 3 integrin expressed on the surface of cells (17), and cell-associated MAGP-2 has been seen in both immunohistochemistry of fetal bovine tissues and cultured rat lung fibroblasts (24,29). We therefore undertook studies to address the potential significance of MAGP-2 interacting with Jagged1.
We first verified the MAGP-2 interaction with Jagged1 in mammalian cells. 293T cells were cotransfected with MAGP-2 and Jagged1, metabolically labeled overnight with [ 35 S]cysteine, and then both whole cell lysates and conditioned medium were collected, because MAGP-2 is both secreted and cell-associated. Immunoprecipitation with an antibody specific for MAGP-2 also coimmunoprecipitated Jagged1 (Fig. 1A,  and 8). The antibody to MAGP-2 did not cross-react with Jagged1 in the absence of MAGP-2 ( Fig. 1A, lanes 3 and 4). In a converse experiment, immunoprecipitation of whole cell lysates expressing both Jagged1 and MAGP-2 with an antibody directed against a C-terminal HA tag fused to Jagged1 also coimmunoprecipitated MAGP-2 (Fig. 1B, lane 8). Together, these results indicate that full-length forms of Jagged1 and MAGP-2 can interact in mammalian cells.
The EGF-like Repeats of Jagged1 Are Essential for MAGP-2/ Jagged1 Binding-By having verified the interaction between MAGP-2 and Jagged1, we wanted to know what domain of Jagged1 was binding to MAGP-2. The extracellular domain of Jagged1 consists mainly of 16 tandem EGF-like repeats. The ligand binding domain is composed of an N-terminal domain and a degenerate EGF-like repeat termed the DSL domain, a hallmark of DSL family ligands ( Fig. 2A). Following the 16 tandem EGF-like repeats is a cysteine-rich domain found only in Jagged/Serrate-like ligands. A series of Jagged1 constructs was made to delete the cytoplasmic domain (⌬CD), the DSL domain (⌬DSL), the cysteine-rich region (⌬CR), or a region spanning 14 EGF-like repeats and the cysteine-rich region (⌬EGFR/CR) ( Fig. 2A). Each deletion construct was transfected into COS-7 cells with MAGP-2. After metabolic labeling, immunoprecipitation with anti-MAGP-2 antibodies was performed. All deletion constructs except ⌬EGF/CR were able to coimmunoprecipitate with MAGP-2 ( Fig. 2B). Expression of ⌬EGFR/CR was confirmed by immunoprecipitation with an anti-HA antibody directed against the HA tag in the cytoplasmic domain (data not shown). The loss of interaction between ⌬EGFR/CR and MAGP-2 but not between ⌬CR and MAGP-2 implicates the EGF-like repeats as the relevant binding domain for MAGP-2. Most interestingly, among the original clones identified in the yeast two-hybrid screen with MAGP-2, 8 of the 13 encoded proteins contain multiple copies of EGF-like repeats, and in particular, we localized the MAGP-2 binding domain in both fibrillin-1 and fibrillin-2 to a region near the C terminus consisting of seven tandem EGF-like repeats. Our results indicate that the EGF-like repeats in Jagged1 are essential for its interaction with MAGP-2 and that tandem EGFlike repeats are a common feature of MAGP-2-binding proteins.

MAGP-2 Facilitates the Shedding of Jagged1 from the Cell Surface in a Metalloproteinase-dependent Manner-Trans-
membrane-tethered Jagged1, as well as other DSL ligands, are substrates for a metalloproteinase-dependent cell-surface "sheddase" activity (30 -33). The molecular weight of Jagged1 found in MAGP-2 immunoprecipitates from the conditioned medium (Fig. 1A, lane 7) is consistent with a shedding mechanism, as it migrates slightly faster than full-length Jagged1 found in whole cell lysates (compare Fig. 1A, lane 7 with 8). We confirmed that the Jagged1 present in the conditioned medium is not full-length ligand by performing an immunoprecipitation with anti-HA antibodies from conditioned medium. Because Jagged1 is tagged at its C terminus with three HA epitopes, this experiment should reveal any full-length Jagged1 present in the supernatant due to nonspecific events such as membrane blebbing. No Jagged1 is seen under these conditions (Fig. 1B,  lane 7), indicating that a soluble extracellular domain fragment of Jagged1 associates with secreted MAGP-2.
The presence of a complex containing both the Jagged1 extracellular domain (ECD) and MAGP-2 in the conditioned medium suggested that perhaps MAGP-2 plays a role in the cell surface shedding and/or stabilization of Jagged1. To determine whether shedding of Jagged1 is influenced by MAGP-2, we transfected Jagged1 alone or in combination with MAGP-2, and we then performed an immunoprecipitation with an antibody directed against the Jagged1 ECD to detect all Jagged1 ECD shed from the cells, and not just that proportion of Jagged1 found in complexes with secreted MAGP-2. COS-7 cells transfected with Jagged1 alone display a low level of ECD shedding that is augmented in the presence of MAGP-2 (Fig. 3A, lanes 1  and 3). The slower migrating band may represent multimerization of the shed Jagged1 ECD (asterisk in Fig. 3A, lanes 1  and 3). In contrast to the MAGP-2 pull downs that also detect Jagged1 ECD, MAGP-2 was not readily detected with the Jagged1 ECD pull downs. This inability to detect MAGP-2 under these conditions may be due to masking of the Jagged1 epitopes by complexed MAGP-2 or the possibility that not all shed Jagged1 ECD is complexed with MAGP-2. Nonetheless, if MAGP-2 induces Jagged1 ECD shedding then one would expect to detect an increase in the generation of cleavage products ) that remains cell-associated following the shedding process (Fig. 3B). Membrane-bound Jagged1 is further proteolytically processed by ␥-secretase (32,34,35) to yield a soluble intracellular cleavage fragment (J1HA ICD ), and MAGP-2 also enhanced detection of this cleavage product (Fig.  3B). Identification of the J1HA cleavage products was verified by using hydroxamate-based ADAM inhibitors and ␥-secretase inhibitors to block their production (data not shown).
LaVoie and Selkoe (32) have presented evidence that an ADAM-17-like activity can proteolytically process Jagged1. Although not direct proof that ADAM-17 is the relevant protease for Jagged1, they report that the protease that acts on Jagged1 has the characteristics of ADAM-17. For example, Jagged1 proteolysis is phorbol ester (PMA)-inducible, sensitive to both batimistat and TAPI-1, and resistant to the ADAM-10 inhibitor TIMP-1. Also as found here, the products of the ADAM-17-like activity are a soluble Jagged1 ECD fragment and a Jagged1 intracellular domain fragment. Because MAGP-2 increases the amount of Jagged1 shed from the cell surface, we asked whether the effect of MAGP-2 was also dependent on metalloproteinase activity. To address this question, we transiently transfected COS-7 cells with Jagged1 in combination with MAGP-2, and we used either PMA to induce metalloproteinase activity or BB3103, a hydroxamate-based inhibitor, to suppress metalloproteinase activity. We then measured the amount of soluble Jagged1 present in the conditioned medium by immunoprecipitation of metabolically labeled culture supernatant with anti-Jagged1 ECD antiserum. We first verified that PMA could induce Jagged1 shedding, and we found that MAGP-2 or PMA treatment resulted in similar levels of Jagged1 ECD in the conditioned medium (Fig. 4, lanes 1-3). Furthermore, we found that MAGP-2-dependent shedding of Jagged1 was greatly decreased in the presence of BB3103 (Fig. 4,  Reversal of the MAGP-2 effect on Jagged1 shedding by BB3103 indicates that MAGP-2-induced shedding is metalloproteinase-dependent. BB3103 also decreases PMA-induced shedding, confirming the previous reports of metalloproteinase-dependent Jagged1 shedding (32). Most interestingly, the effects of MAGP-2 and PMA appear to be additive (Fig. 4, lane 4).
MAGP-2 Binds Other DSL Ligands but Does Not Facilitate Their Shedding-Other DSL family members contain EGF repeats that could have MAGP-2-binding sites as found for Jagged1. To determine whether MAGP-2 could bind to Delta1 and Jagged2, COS-7 cells were cotransfected with each ligand plus MAGP-2 and metabolically labeled, and both conditioned media and whole cell lysates were collected. Immunoprecipitation of whole cell lysates with anti-MAGP-2 antiserum revealed the presence of proteins consistent with the molecular weight of Delta1 and Jagged2 (Fig. 5, lanes 4 and 6), indicating that these DSL family members also contain MAGP-2-binding sites. Identification of the coprecipitating bands as J1HA, J2HA, and D1HA was done by transfection and immunoprecipitation of these constructs alone compared with untransfected cells, and it was verified that the MAGP-2 antibody does not recognize these ligands in the absence of MAGP-2 (data not shown). However, in the conditioned medium, unlike the interaction between Jagged1 and MAGP-2 (Fig. 5, lane 1), neither Delta1 nor Jagged2 was immunoprecipitated by anti-MAGP-2 antibodies (Fig. 5, lanes 3 and 5). We surmised that the inability of these two DSL family members to be found in a complex with secreted MAGP-2 could be a result of MAGP-2 not being active in the shedding of Jagged2 or Delta1, or that the ligands were  shed but had a low affinity for MAGP-2 and therefore did not remain associated with MAGP-2. To investigate further these possibilities, we set up experiments similar to those in Fig. 3. COS-7 cells were transfected with Delta1 alone or in combination with MAGP-2, and the levels of Delta1 ECD shedding were determined by immunoprecipitation with Delta1 ECD antiserum. Similar to Jagged1, the basal level of Delta1 shedding is very low (Fig. 6, lanes 1 and 5). However, MAGP-2 does not increase the amount of Delta1 ECD released into the conditioned medium to the same extent as Jagged1 ECD (Fig. 6,  lanes 3 and 7), indicating that at least for Jagged1 and Delta1, MAGP-2 can differentially affect shedding of DSL ligands.
The Related Microfibrillar Protein MAGP-1 Binds DSL Ligands but Does Not Induce Jagged1 Shedding-Given that binding of MAGP-2 to Jagged1 is mediated by a domain that shares similarity with the related protein MAGP-1, we investigated whether MAGP-1 could form similar complexes with DSL ligands and potentially effect their shedding from the cell surface. We found that MAGP-1, like MAGP-2, was coimmunoprecipitated with both Jagged1 and Jagged2 from cell lysates (Fig. 7A). However, unlike MAGP-2, MAGP-1 was not able to augment shedding of Jagged1 from the cell surface, using the Jagged1 ECD antibody for immunoprecipitations (Fig. 7B). Thus, augmentation of Jagged1 shedding appears to be restricted to MAGP-2. DISCUSSION MAGP-2 is one of a number of low molecular weight glycoproteins associated with the microfibril component of elastic fibers present in the extracellular matrix. Together with MAGP-1, MAGP-2 contains a matrix binding domain in the C-terminal half of the protein that mediates binding to microfibrillar proteins such as fibrillin-1, fibrillin-2, and fibulin and is characterized by a unique pattern of cysteine residues (21). However, neither MAGP-1 nor MAGP-2 are exclusively localized to microfibrils. Two alternatively spliced forms of MAGP-1 lack a signal peptide and in transfection studies are found intracellularly (29). In the same study, MAGP-2 was found to be secreted but remained cell-associated. These differences between MAGP-1 and MAGP-2 imply that their respective functions in the cellular compartment may be nonredundant. In particular, it may have implications for our data indicating that only MAGP-2 can potentiate Jagged1 shedding. Perhaps the specificity of the MAGP-2 effect on Jagged1 can be attributed to unique functions in the N-terminal half of MAGP-2 not shared with MAGP-1 which allow MAGP-2 to be tethered to the cell surface, such as the RGD integrin binding domain that has been shown to mediate the interaction of cells with MAGP-2 (17). The interactions detected here between DSL ligands and MAGP-2 potentially provide a separate mechanism whereby MAGP-2 may affect cellular signaling. Specifically, we find that MAGP-2 physically interacts with the cell-surface protein, Jagged1, a ligand for the Notch signaling pathway, and potentiates the shedding of the Jagged1 extracellular domain.
Structure-function analysis of the interaction between Jagged1 and MAGP-2 indicates that the proteins associate via the C-terminal MAGP domain of MAGP-2 and the extracellular EGF-like repeats of Jagged1. Because all DSL ligands possess tandem arrays of EGF-like repeats, it is not surprising that two other DSL ligands, Delta1 and Jagged2, also interact with MAGP-2. However, even though MAGP-2 binds to both of these DSL ligands, only shedding of Jagged1 is increased by MAGP-2. This is surprising given that both Delta1 and Jagged2 are substrates for metalloproteinase-dependent shedding (34) and suggests that DSL ligand binding to MAGP-2 alone does not lead to an increase in shedding. Most importantly, our findings suggest that MAGP-2-enhanced shedding is specific for the DSL ligand Jagged1. Likewise, Jagged1 shedding appears to specifically require MAGP-2, as MAGP-1 binds cell-associated DSL ligands but is not found complexed with Jagged1 in the media. Although it remains possible that the interactions between MAGP-2 and full-length forms of the DSL ligands at the cell surface have functional consequences, the detection of complexes containing both MAGP-2 and Jagged1 ECD in conditioned media is more consistent with a role for MAGP-2 in regulating Jagged1 activity through shedding.
The increased shedding of Jagged1 in the presence of MAGP-2 is likely mediated by metalloproteinases as reported for the proteolytic processing of other DSL ligands (30 -33). This supposition is based on our findings that shedding of the Jagged1 ECD is specifically inhibited by hydroxamate inhibitors BB3103 and BB94. 2 Although these are broad-based metalloproteinase inhibitors, Jagged1 has been reported to be proteolytically cleaved by the ADAM-17-like activity, and thus it is possible that this metalloproteinase is responsible for the MAGP-2-induced shedding of the Jagged1 detected here. Therefore, it seems likely that MAGP-2 does not induce a novel proteolytic cleavage within Jagged1 but rather modulates the previously characterized metalloproteinase cleavage of Jagged1. However, both the regulation and role of the metalloproteinase-dependent cleavage of DSL ligands is not well understood in mammals. In fact, other than one report of Notchresponsive cleavage of Delta in flies (35), no biological modifiers of ADAM processing for DSL ligands in mammalian cells have been reported. Given our findings that MAGP-2 both binds to and augments the production of soluble Jagged1, we propose that MAGP-2 is a novel biological modulator of Jagged1 shedding.
Future studies will need to address how binding of MAGP-2 to Jagged1 leads to increases in Jagged1 detected in the culture media. We can envision at least two mechanisms by which this could occur. First, MAGP-2 binding to Jagged1 could potentiate the ADAM-like cleavage of Jagged1 by inducing a conformational change in that would expose the ADAM binding and/or cleavage site to facilitate proteolysis of Jagged1. In this scenario, MAGP-2-induced Jagged1 shedding from cells would lead to a reduction of cell-surface Jagged1 and consequently a decrease in Notch signaling between cells. Another possibility is that MAGP-2 binding to Jagged1 could stabilize Jagged1 shed constitutively from the cell surface, allowing it to accumulate in the conditioned medium. MAGP-2 bound to Jagged1 could protect the Jagged1 ECD from spurious cleavage by an extracellular protease. In this regard, Notch2 EGF-like repeats, which are very similar to DSL ligand EGF-like repeats, have been shown to be susceptible to cleavage by neutrophil elastase (36). Our data do not allow us to differentiate between a direct effect of MAGP-2 on ADAM-induced shedding of cellsurface Jagged1 or that of MAGP-2 binding to Jagged1 EGFlike repeats to stabilize the shed form of Jagged1. However, given that MAGP-2 does not increase the detection of shed Jagged2 or Delta1 even though MAGP-2 interacts with both DSL ligands suggests that MAGP-2 likely enhances shedding of Jagged1.
Binding of MAGP-2 to the soluble Jagged1 ECD in the extracellular space may not only serve to stabilize the shed pro-tein but it may also regulate the activity of shed Jagged1. Soluble forms of DSL ligands have been reported to either activate or inhibit Notch signaling in worms, flies, and mammalian systems (37)(38)(39)(40). The differences in activity detected for soluble DSL ligands in different systems may reflect the multimeric state of the particular soluble ligand. For example, in tissue culture-based assays, the engineered soluble forms of DSL ligands fused to human Fc require clustering with an Fc antibody to activate Notch signaling and regulate cellular differentiation (39,41,42). Despite these reports, secreted forms of Jagged1 (43) as well as a Jagged1 DSL domain peptide (44) also activate Notch signaling in the absence of apparent clustering. However, it is possible that the high protein concentrations used in these studies produce Jagged1 aggregates that bind and activate Notch. Consistent with this idea, we and others have shown that monomeric forms of DSL ligands function to block ligand-induced Notch signaling and that the particular multimeric state of the soluble DSL ligand determines whether it functions as an activator or inhibitor of Notch signaling (39,45). In this regard, MAGP-2 bound to shed Jagged1 may allow the complexed ligand to bind and activate Notch rather than blocking signaling as found for monomeric soluble forms of Jagged1.
An indication of the biological significance of soluble Jagged1 is the natural occurrence of alternatively spliced mRNA species in both mammalian keratinocytes (43) and endothelial cells (46) that encode secreted forms of Jagged1 containing only the extracellular domain. This secreted form of Jagged1 is biologically active in both keratinocytes and fibroblasts and produces effects similar to that achieved by expression of a constitutively active form of Notch1. In a study by Lindner et al. (47), overexpression of secreted Jagged1 in NIH3T3 fibroblasts affected both cell-matrix interactions and cell-cell adhesion. Cell-matrix interactions were decreased in secreted Jagged1-expressing fibroblasts as measured by the reduction of focal adhesion plaque markers such as focal adhesion kinase phosphorylation and vinculin staining. Cell-cell adhesion, however, was strengthened in secreted Jagged1-expressing fibroblasts as indicated by increased expression of cell adhesion molecules such as cadherin and ␤-catenin at the cell-cell interface. These effects in fibroblasts negatively impacted the motility of secreted Jagged1-expressing cells in a wound-filling assay.
Along with the shedding of the extracellular domain of Jagged1, another consequence of MAGP-2-induced cleavage of Jagged1 is that the remaining transmembrane-bound fragment of Jagged1 is primed for a constitutive ␥-secretase cleavage event that releases the intracellular domain from its membrane tether (32,34,35). Evidence for a functional role of the soluble intracellular domain of Jagged1 (JICD) is more tenuous than that for the extracellular domain of Jagged1, but overexpression of constructs encoding only JICD have produced effects in both mammalian cell culture and Xenopus embryos (32,48). A major caveat to these studies is that neither attempted to show that JICD generated from the full-length Jagged1/ Serrate1 functioned as described for the truncated intracellular domain constructs. Nonetheless, these reports raise the possibility that the MAGP-2-induced shedding of Jagged1 reported here could lead to the production of a biologically active intracellular domain fragment.
There is growing evidence of Jagged1 involvement in cellmatrix interactions in both normal development and disease. In addition to our findings for MAGP-2 and Jagged1, the extracellular matrix protein thrombospondin-1 has also been reported to interact with Jagged1 (43). However, the functional consequences of the interactions between thrombospondin-1 and Jagged1 were not investigated. Nevertheless, the interac-tion of Jagged1 with two extracellular matrix proteins suggests that in addition to its role in signaling through cell-cell interactions that Jagged1 may also mediate cell-matrix interactions. Severe vascular defects such as hemorrhaging and decreased angiogenic remodeling have been identified for Jagged1 mutant mice (27) and may reflect defects in cell-cell as well as cell-matrix interactions. Furthermore, expression of Jagged1 is strongly up-regulated in endothelial cells stimulated to migrate following injury of the carotid artery in a rat model system (47). It has been suggested that Jagged1 may be biologically relevant to the movement of smooth muscle cells, either into the denuded area of vessel injury or through the elastic lamina to the intimal compartment of vessels (47). Most interestingly, we have found both MAGP-2 and Jagged1 expressed in a vascular smooth muscle cell line. 2 Although purely speculative at this point, the presence of both MAGP-2 and Jagged1 in these cells may allow for MAGP-2 potentiation of Jagged1 shedding and perhaps suggest that vascular smooth muscle cells can regulate levels of cell-surface Jagged1 through MAGP-2-induced shedding of Jagged1.