Domain and Functional Analysis of a Novel Platelet-Endothelial Cell Surface Protein, SCUBE1*

SCUBE1 (signal peptide-CUB-EGF domain-containing protein 1) is a novel, secreted, cell surface glycoprotein expressed during early embryogenesis and found in platelet and endothelial cells. This protein is composed of an N-terminal signal peptide sequence followed by nine tandemly arranged epidermal growth factor (EGF)-like repeats, a spacer region, three cysteine-rich repeat motifs, and one CUB domain at the C terminus. However, little is known about its domain and biological function. Here, we generated a comprehensive panel of domain deletion constructs and a new genetic mouse model with targeted disruption of Scube1 (Scube1Δcub/Δcub) to investigate the domain function and biological significance. A number of cell-based assays were utilized to define the critical role of the spacer region for membrane association and establish that the EGF-like repeats 7–9 are sufficient for the formation of SCUBE1-mediated homophilic adhesions in a calcium-dependent fashion. Biochemical and molecular analyses showed that the C-terminal cysteine-rich motifs and CUB domain could directly bind and antagonize the bone morphogenetic protein activity. Furthermore, genetic ablation of this C-terminal region resulted in brain malformation in the Scube1Δcub/Δcub embryos. Together, our results support the dual roles of SCUBE1 on brain morphogenesis and cell-cell adhesions through its distinct domain function.

SCUBE1 (signal peptide-CUB-EGF domain-containing protein 1) is the founding member of an evolutionarily conserved SCUBE gene family (1,2). To date, three distinct isoforms have been cloned and named SCUBE1 to SCUBE3 according to their order of discovery in mammals (1)(2)(3)(4). These genes coding for polypeptide molecules of about 1000 amino acids share an organized protein domain structure of at least 5 recognizable motifs: an N-terminal signal peptide sequence, 9 tandem repeats of epidermal growth factor (EGF) 3 -like modules, a large N-glycosylated spacer region followed by three repeated stretches of 6-cysteine residues with unique and regular spacing, and one CUB domain at the C terminus (see Fig. 1A). Our previous study demonstrated that the signal peptide sequence is sufficient to direct the expressed SCUBE1 protein into the secretory pathway and results in a secreted and surface-associated protein (1). However, little is known about the functional significance of the other four domains in SCUBE1.
SCUBE genes have been shown to be expressed predominantly in a variety of developing tissues, including gonads, the central nervous system, dermomyotome, digital mesenchyme, and limb buds during mouse embryogenesis (2,3,5), which implies that the proteins may play important roles in development. However, direct functional studies of the roles of the SCUBE genes during mammalian development are still lacking. In addition to its embryonic expression, SCUBE1 was found to be expressed in the endothelium and platelets (1,6). Our recent study showed that SCUBE1 is stored in the platelet ␣-granules and exposed to the cell surface upon platelet stimulation and activation (6). Yet, the molecular function for such targeted exposure of SCUBE1 on the cell surface has not been addressed.
Here, we investigated the domain and biological functions of SCUBE1 by utilizing a series of SCUBE1 domain-specific deletion constructs and employing a new genetic mouse model with targeted disruption of the Scube1 gene. Our results demonstrated, for the first time, that the N-terminal EGF-like repeats participate in the formation of adhesions in a Ca 2ϩ -dependent fashion and the EGF-like repeats 7-9 are sufficient for reciprocal and lateral interactions between SCUBE1 in homophilic adhesions. Moreover, gene-targeting study indicated that SCUBE1 is required during the early stages of central nervous system development, possibly through modulating the activity of members of the bone morphogenetic protein (BMP) family of proteins via its C-terminal 6-cysteine repeat motifs and the CUB domain.

EXPERIMENTAL PROCEDURES
Cell Culture, Construction of Expression Plasmids, and Transfection-A2058 melanoma cells, human embryonic kidney (HEK)-293T and GP2-293 retrovirus packaging cells were maintained in Dulbecco's minimal essential medium supplemented with 10% heat-inactivated fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in an atmosphere of 5% CO 2 . The epitope-tagged version of the SCUBE1 mutant was constructed as described (1). Cells were transfected by using Lipofectamine 2000 (Invitrogen).
Immunoprecipitation and Western Blot Analysis-Two days after transfection, cell lysates were clarified by centrifugation at 10,000 ϫ g for 20 min at 4°C. Samples underwent immunoprecipitation, followed by Western blot analysis as described (1).
Cell Aggregation Assay-A2058 stable cells were detached by trypsin (0.01%)-EDTA (2.5 mM) treatment and suspended in Dulbecco's minimal essential medium, either containing 5 mM CaCl 2 or Ca 2ϩ free, at 1 ϫ 10 6 cells/ml in polystyrene tubes. Then, tubes were incubated on a rotating platform (10 rpm) at 37°C for 1 h. The extent of cell aggregation was viewed microscopically and photographed. For quantitation, cell clusters of more than four cells were counted as being aggregated. The data are shown as the means Ϯ S.E. of three independent experiments in duplicate.
Luciferase Activity Assays-Human HepG2 cells (3 ϫ 10 5 cells per well) were seeded into 24-well plates and transfected on the following day with 0.4 g of the BMP-inducible luciferase reporter I-BRE-Luc and 0.01 g of the Renilla luciferase reporter vector used as an internal control. The transfected responding cells (HepG2) were stimulated with conditioned media derived from the signaling cells (HEK-293T) transfected with the BMP2 expression plasmid alone or in a combination of various SCUBE1 deletion constructs. Luciferase activity was measured following 24-h incubation by the use of the dual reporter system (Promega, Madison, WI). Data are expressed as relative luciferase activity (firefly luciferase activity divided by Renilla luciferase activity).
Targeted Disruption of the Scube1 Gene in Mice-The targeting vector obtained from the MICER website was digested with ClaI (see Fig. 8A). The ES cell line R1, established from a hybrid of two 129 substrain (129X1/SvJ and 129S1) F 1 male blastocysts (7), was transfected with linearized targeting vector, and selection involved use of puromycin. Of a total of 251 ES cell clones, we obtained 13 correctly targeted ES clones verified by Southern blot and/or PCR analysis. The correctly targeted ES cell clones were injected into blastocysts of C57BL/6 mice. The targeted allele was backcrossed with C57BL/6 females for at least three generations before phenotyping experiments.
Histological and Immunohistochemical Analysis-Embryos were dissected from pregnant mice at the indicated time, fixed, and embedded in paraffin. Sections were cut at 5 m and stained with hematoxylin and eosin. For SCUBE1 immunohistology, sections were probed with a monoclonal antibody specifically against the CUB domain of SCUBE1 (6).
Proliferation and Apoptosis Assays-Sections were stained with rabbit anti-mouse Ki-67 antibody (DakoCytomation) for active cell proliferation and counterstained with hematoxylin to detect all nuclei. Detection of apoptotic cells involved the terminal dUTP nick-end labeling reaction kit (Roche Applied Science) and counterstaining by 4Ј,6-diamidino-2-phenylindole.
Statistical Analysis-Data are expressed as mean Ϯ S.E. Differences between groups were analyzed by unpaired Student's t test. p Ͻ 0.05 was considered significant.

Molecular Dissection of the SCUBE1 Domain Responsible for
Secretion and Cell Surface Association-We previously showed that overexpression of SCUBE1 in HEK-293T cells resulted in a secreted glycoprotein tethered on the cell surface (1). However, the structure responsible for the surface targeting and membrane association was not precisely defined. To delineate the domain required for the secretion and cell surface anchorage, we then generated a comprehensive panel of mutated SCUBE1 expression constructs (D1-D7) with specific deletions of the EGF-like repeats, spacer region, cysteine-rich repeats, or CUB domain (Fig. 1A). The FLAG epitope tag was added at the N terminus for easy detection of these mutated SCUBE1 forms.
The SCUBE1 deletion proteins were expressed by means of transient expression in HEK-293T cells. Two days after transfection, the conditioned media and cell lysates were collected and subjected to Western blot analysis with anti-FLAG antibody. As shown in Fig. 1B, all SCUBE1 deletion constructs produced corresponding recombinant proteins in HEK-293T cells (bottom panel). Furthermore, immunoblotting of the conditioned media from transfected cells showed an effective secretion of all expression products, except for the SCUBE1-D2 mutant protein, which lacks the spacer region (Fig. 1B, top  panel). Adding back this region restored the secretion of the D2 mutant into the conditioned culture medium (data not shown). Together, these data suggest that the spacer region, located between the EGF-like domains and the cysteine-rich repeats, is capable of directing the expressed SCUBE1 protein through the secretory pathway for secretion, at least when overexpressed in HEK-293T cells.
The same set of transfected cells was detached and stained with anti-FLAG antibody to determine their surface expression by flow cytometry (Fig. 1C). Flow cytometric analysis showed SCUBE1-D2, -D4, -D5, and -D6 deletion mutants defective in membrane association, but SCUBE1-FL, -D1, -D3, and -D7 mutant protein, which all contain the spacer region, was expressed on the cell surface. Together, our results support the concept that the spacer region has at least two functions: to target SCUBE1 for secretion and anchor the molecule on the cell surface.
Establishment of Cell Lines Expressing a Series of SCUBE1 EGF-like Repeat Deletion Mutants-Our previous study showed that the preformed SCUBE1 protein stored in the platelet ␣-granules could be translocated to the surface upon platelet activation (6). However, the functional significance of such targeted exposure of SCUBE1 on the cell surface remained elusive. Because certain EGF-like repeats mediate homophilic or heterophilic protein-protein interactions (8,9), we next examined by suspension cell aggregation assay whether the N-terminal EGF-like repeats of SCUBE1 play a role on cell-cell interaction.
To further evaluate the relative contribution of the EGF-like repeats in the adhesion function of SCUBE1, we first constructed four deletion mutants consisting of the EGF-like modules 1-9 (E1-9), 4 -9 (E4 -9), 7-9 (E7-9), or only E7-9 deletion (⌬E7-9) into a retrovirus vector, pLNCX-eB7 (10). In addition, an HA epitope tag was added at the N terminus for easy detection, and the mutant constructs were further fused to the B7 extracellular and transmembrane domain (B7TM) to effectively target the chimeric protein on the plasma membrane ( Fig. 2A).
The melanoma A2058 cells lacking endogenous SCUBE1 expression were infected with recombinant retroviral particles and selected in G418 to obtain four independent A2058/ SCUBE1-E1-9, -E4 -9, -E7-9, or -⌬E7-9 stable transfectant cell lines. Western blot analysis with the anti-HA antibody demonstrated the SCUBE1 EGF-like repeat deletion mutant proteins to be produced at comparable levels with the expected molecular masses (Fig. 2B). Most importantly, flow cytometry using the anti-HA or ant-SCUBE1 antibody confirmed that all of these SCUBE1 EGF-like repeat mutants expressed by A2058 cells were transported to the cell surface (Fig. 2C).
The EGF-like Repeats 7-9 Are Sufficient, but Not Necessary, for Reciprocal (trans) or Lateral (cis) Interactions between SCUBE1 Molecules-To examine the involvement of SCUBE1 EGF-like repeats in intercellular adhesion, we evaluated the aggregating properties of the SCUBE1 EGF-like repeat mutant stable transfectant cell lines by cell aggregation assay. Monolayer cultures of the parental (control) or transfected stable cell lines were dissociated into single cells and allowed to aggregate in suspension culture. When the suspensions were gently shaken, A2058 cells expressing SCUBE1 EGF-like repeat mutants tended to aggregate (Fig.  3). The parental A2058 cells aggregated poorly in suspension (Ͻ3%), but a good proportion (ϳ8 -20%) of the SCUBE1 EGF-like repeat mutant-transfected clones aggregated after 1 h of shaking (Fig. 3).
The chelating of calcium by Ca 2ϩ -binding EGF-like domains is important for the structural integrity and function of several other EGF-like domain-containing proteins (11,12). Because six of nine EGF-like modules (i.e. repeats 1-3 and 7-9) contain Ca 2ϩ -binding consensus sequences (1, 6), we hypothesized that these EGF-like repeats may function in a Ca 2ϩ -dependent manner. When the same cell lines were dissociated with gentle pipetting in the presence of 5 mM EDTA or allowed to aggregate in the absence of Ca 2ϩ , no apparent aggregation was observed during 1 h of incubation for either the parental or EGF-like repeat deletion transfectants (Fig. 3A).
To further address the specificity of the EGF-like repeats, the recombinant GST fusion protein fragment containing the EGFlike repeats (GST-E4 -9) or spacer region (GST-Spacer) was added into the cell aggregation assay to block the cell-cell interactions. Soluble EGF-like repeat protein GST-E4 -9 clearly blocked A2058-SCUBE1-E7-9-mediated cell aggregation assay, whereas GST alone or recombinant spacer region protein A, domain structure of SCUBE1 full-length (FL) and its deletion constructs (D1-D7). A FLAG epitope tag was added immediately after the signal peptide sequence, at the N terminus for easy detection. FL, amino acids 1-988; D1, amino acids 1-789; D2, amino acids 1-411; D3, amino acids 402-988; D4, amino acids 628 -988; D5, amino acids 791-988; D6, amino acids 628 -788; D7, amino acids 402-627. SP, signal peptide; E, EGF-like repeat; Cys-rich, cysteine-rich motifs; CUB, CUB domain. B, the spacer region is essential for the secretion of the expressed SCUBE1 protein. HEK-293T cells were transfected with the indicated expression vector. Two days after transfection, samples from conditioned medium or cell lysates were immunoprecipitated by anti-FLAG antibody and separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes. Recombinant SCUBE1 proteins were detected by western blotting with anti-FLAG antibody. C, the spacer region is necessary for the surface expression of SCUBE1. 24 h after transfection, a set of transfected cells as above was detached and stained with anti-FLAG antibody to determine the cell surface expression. Experiments were repeated three times with similar results. IP, immunoprecipitation; WB, Western blot analysis.
We previously showed that human SCUBE1 proteins are capable of forming oligomeric complexes when overexpressed in HEK-293T cells (1). To further determine the relative contribution of which EGF-like repeat modules involved in lateral (cis) intracellular oligomerization, we transiently co-expressed three deletion constructs, HAtagged SCUBE1-E1-9, -E4 -9, -E7-9, or -⌬E7-9 with FLAG. SCUBE1-D2 containing only the EGF-like repeats 1-9 in HEK-293T cells. Immunoprecipitation of HA.SCUBE1-FIGURE 2. Generation of cell lines stably expressing the SCUBE1 EGF-like repeat deletion mutants. A, domain composition of EGF-like repeat mutants. The EGF-like repeats 1-9, 4 -9, 7-9, or only 7-9-specific deletion (E1-9, amino acids 33-412; E4 -9, amino acids 157-412; E7-9, amino acids 281-412; ⌬E7-9, same as E1-9 but lacking amino acids 282-397) were fused to the B7 transmembrane domain (B7TM) to target the chimeric protein on the plasma membrane. An HA epitope tag was added to monitor the expression of these recombinant proteins. B, Western blot analysis of the EGF-like repeat deletion constructs. A2058 melanoma stable lines expressing E1-9, E4 -9, or E4 -9 deletion constructs underwent Western blotting with the anti-HA antibody to verify the expression of these recombinant proteins. C, surface expression of the EGF-like repeat deletion mutant proteins. A2058 stable cell lines were analyzed by flow cytometry with the anti-HA or anti-SCUBE1 antibody to confirm the surface targeting of these mutant proteins. . The EGF-like repeats 7-9 are sufficient for homophilic reciprocal interactions of SCUBE1 in a Ca 2؉ -dependent manner. A, degree of aggregation in suspension of A2058 parental and stable lines. A2058 parental (Control) or EGF-like repeat mutant stable lines (E1-9, E4 -9, E7-9, or ⌬E7-9) were detached and allowed to aggregate in suspension culture in the presence or absence of Ca 2ϩ in aggregation medium for 1 h. Aggregates were defined as clusters of four or more cells. B, representative micrographs of aggregates formed by A2058 parental control or EGF-like repeat mutant stable lines. The experiments were performed three times in duplicate with similar results. *, p Ͻ 0.01 versus control cells. Original magnification, ϫ100. E1-9, -E4 -9, -E7-9, or -⌬E7-9 resulted in the co-precipitation of FLAG.SCUBE1-D2 (Fig. 5). Therefore, the EGF-like repeats 7-9 are also sufficient, but not necessary, for homophilic lateral SCUBE1 interactions within the same cells.
Molecular Analysis of the C-terminal Six-cysteine Repeat Motif and CUB Domain-Recent genetic analysis identified that the zebrafish orthologue of the mammalian SCUBE2 gene acts upstream of hedgehog ligands or through a parallel pathway (13)(14)(15). In addition, a nonsense mutation in the ty97 allele encodes a truncated null protein lacking the cysteine-rich repeat motif and the CUB domain, implying that this C-terminal region is essential for SCUBE2 function (13)(14)(15). From this genetic evidence, the SCUBE2 protein is postulated to play a critical role for transport or stability of hedgehog signaling in the extracellular environment (15), function as an endocytic receptor during reception of the hedgehog signal (13), or modulate the long range action of BMP-dependent signaling in the neural tube and somites (14), but the exact biochemical and molecular mechanism underlying the SCUBE2 action remains unclear.
To further explore the molecular function of the C-terminal region of SCUBE1, we generated one additional deletion mutant, SCUBE1-ty97 (Q623 to stop codon), mimicking the ty97 mutant allele in the zebrafish Scube2 gene by removing the 6-cysteine repeat motif and the CUB domain (Fig. 6A). Since one previous report demonstrated that BMP activity can be attenuated by the co-expression of SCUBE2 in the zebrafish (14), we next investigated whether the SCUBE1 C-terminal domain could directly interact with BMP2 protein and form a complex. HEK-293T cells were transfected with a Myc-tagged BMP2 expression plasmid alone or in combination with a series of FLAG-tagged SCUBE1 domain deletion constructs (Fig. 6A). Two days after transfection, cell lysates underwent immunoprecipitation with the anti-Myc monoclonal antibody, and the precipitates were analyzed by immunoblotting with anti-FLAG monoclonal antibody to determine the protein association. Immunoprecipitation with anti-Myc antibody resulted in a specific coimmunoprecipitation of the SCUBE1-D4 or -D5 deletion protein (Fig. 6B), which suggests that SCUBE1 could form a complex with BMP2 through its C-terminal CUB domain.
To further evaluate whether the interaction between SCUBE1-D4/-D5 and BMP2 affects the long range signaling of BMP2, we co-cultured the conditioned media derived from HEK-293T cells co-transfected with BMP2 alone or in combination with SCUBE1 deletion constructs (signaling cells) added to the responding cells, HepG2 cells containing the BMP-inducible promoter luciferase reporter construct I-BRE-Luc (16) (Fig. 7A). As expected, BMP2 alone produced by the signaling cells could act as a long range signaling molecule by inducing an increase of ϳ10-fold in luciferase activity (Fig. 7B). Although the BMP2 proteins co-expressed with the SCUBE1-FL, -ty97, -D5, or -D6 triggered the BMPmediated transcriptional activation equally well, co-expres-   Fig. 1) was expressed together with HA.SCUBE1-E1-9, -E4 -9, or -E7-9, or -⌬E7-9 construct by transient transfection in HEK-293T cells. Detergent lysates were immunoprecipitated with the anti-HA antibody and then immunoblotted with the anti-FLAG antibody to determine the associated proteins. Cell lysates were also immunoblotted to confirm the protein expression level for each construct. IP, immunoprecipitation; WB, Western blot analysis.
sion with the SCUBE1-D4 mutant resulted in a marked attenuation of the BMP response (Fig. 7B).
Because the proteolytic processing of the large prepro precursor of BMP (proBMP) and its subsequent secretion into the extracellular environment are the required steps in the formation of the biologically active BMP ligand, we next investigated whether the inhibition of BMP2 signaling by SCUBE1-D4 occurs in the intracellular or extracellular space. HEK-293T cells were transfected with plasmids expressing proBMP2 alone or in combination with various SCUBE1 deletion constructs. Western blot analysis of the cell lysates and condition medium from these cultures revealed all deletion mutants with no effect on total proBMP2 synthesis (Fig. 7C), whereas only the SCUBE1-D4 mutant potently inhibited the secretion of mature BMP2 into the culture medium.
Targeted Disruption of the Scube1 Gene in Mice Produces a Neural Tube Closure Defect over the Midbrain-To further evaluate the biological function of the C-terminal domain of SCUBE1 in vivo, we disrupted the mouse Scube1 gene at the end of exon 16 by use of an insertional targeting vector obtained from a public resource (17). This vector was expected to result in the duplication of exons 11-16 and disruption of splicing to produce a mutant transcript coding for a truncated protein lacking the functional 6-cysteine repeat motifs (exons 16 -19) and the CUB domain (exons 20 and 21) (Fig. 8A). The correctly targeted embryonic stem cell clones and chimeric mice carrying germ line transmission of the mutant allele of Scube1 were confirmed by Southern blot (Fig. 8B) and genomic PCR analyses with one primer pair designed from the vector backbone and the other designed from outside the targeting construct (Fig. 8A). In addition, the expression of the mutant transcript in hetero-or homozygous mutant mice was confirmed by reverse transcriptase PCR analysis and direct sequencing of the PCR product (data not shown). Most importantly, the Scube1 ⌬cub/⌬cub mutant mice lacked the expression of the functional C-terminal region, because SCUBE1 immunoreactivity was not seen in the mutant brain sections stained with a monoclonal antibody specific against the CUB domain (6) (Fig. 8C). However, immunohistochemistry stained with the N terminus-specific monoclonal antibody could still detected the immunoreactivity in the mutant neural tissues, including forebrain (f), hindbrain (h), and cranial ganglia (asterisks), suggesting the targeted allele is not a true null allele (see supplemental Fig.  S1). Instead, a truncated form of SCUBE1 containing the N-terminal fragment is expressed in the mutant animal, as the protein fragment can be detected using an antibody specific for the N terminus. Thus, this allele is designated as Scube1 ⌬cub/⌬cub on the basis of the lack of the CUB immunoreactivity (Fig. 8C). FIGURE 6. The C-terminal region of SCUBE1 could specifically interact with BMP2. A, domain organization of constructs used in this experiment. All constructs are as described in Fig. 1, except a new deletion construct, SCUBE1-ty97 mutant (amino acids 1-622), was made by mimicking a null mutant allele (ty97) in the zebrafish Scube2 (13)(14)(15). B, interaction between SCUBE1 and BMP2. The BMP2 expression construct (Myc-tagged) was transfected alone or in combination with the expression plasmids encoding indicated FLAG-tagged SCUBE1 proteins in HEK-293T cells. Two days later, cell lysates underwent immunoprecipitation (IP) and Western blotting (WB) with antibodies as indicated to determine the protein-protein interactions. ProBMP2, precursor BMP2.
Mice heterozygous for the Scube1 mutation have a wild-type phenotype and normal fertility. However, a fraction of the Scube1 ⌬cub/⌬cub newborn pups died shortly after birth and manifested a phenotypic defect showing acrania and lack of brain tissue (Fig. 9A). In addition, mutant embryos recovered from E12.5 to E17.5 were alive (determined by active heart beating and/or limb movement) with exposed and progressively degenerating brain tissue ( Fig. 10B and data not shown).
The acrania phenotype was more clearly revealed by staining E14.5 skeletons with Alcian blue (cartilage) and alizarin red (ossified bone) (Fig. 9, B and C). Despite no apparent defects in the limb and skeletal structures, the cranial vault is missing in the Scube1 ⌬cub/⌬cub skulls (Fig. 9, B and C). The interparietal and supraoccipital bones were missing, and only a small portion of the frontal and parietal bones were present at their lateral basal levels. Thus, the defects in cranial bone formation occurred early in development.
Close examination and histological analyses of E12.5 Scube1 ⌬cub/⌬cub embryos revealed that the brain malformation between the diencephalon and myelencephalon (Fig. 10B, between arrows), which resulted in exencephaly restricted to the midbrain because of overproliferation of neural tissue and everted cranial neural folds (Fig. 10D, asterisks). Further histological sections at the level of the eye uncovered distinct pattern abnormalities, including thickening of the forebrain neuroepithelium wall (Fig. 10F) and hyperplasty in the cranial ganglia, espe-cially the trigeminal and vestibulocochlear ganglia (gV or gVIII in Fig. 8C). However, the hindbrain neural tube seemed less affected (Fig. 8C).
Given the increase in forebrain neuroepithelial wall thickness in Scube1 ⌬cub/⌬cub embryos (Fig. 10F), we then analyzed the mitotic activity in this tissue. Interestingly, increased proliferation rate was observed in the forebrain neural tube, as evidenced by the number of nuclei stained positive for Ki-67, a cell proliferation marker (61.3 Ϯ 2.9% for Scube1 ⌬cub/⌬cub versus 48.5 Ϯ 2.5% for the wild type; p Ͻ 0.05, n ϭ 3 animals for each genotype). However, the proliferation rate of the hindbrain neural tube was not affected (11.1 Ϯ 3.1% for Scube1 ⌬cub/⌬cub versus 12.3 Ϯ 2.0% for the wild type; p Ͼ 0.05). Furthermore, terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling assay did not reveal any discernable difference in the number of apoptotic cells at E12.5 in the forebrain or hindbrain neuroepithelium (data not shown).

DISCUSSION
In the present study, we analyzed the domain and the biological function of SCUBE1, the first member of a small secreted, cell surface SCUBE protein family. In vitro cell-based assays and in vivo gene ablation experiments support the dual roles of the Scube1 gene product during early embryonic development and postnatal cell-cell adhesions. In addition, our results suggest that the SCUBE1 protein is organized in a modular fashion and has a unique function associated with each domain. For example, it not only could function as an antagonist for the BMP signaling through its C-terminal domain but may also act as a cell adhesion molecule through Ca 2ϩ -dependent interactions mediated by EGF-like repeat modules.
Moreover, our mapping of the secretion and membrane-association domain to the spacer region allowed us to further assign one additional specific function to a structural domain of SCUBE1 (Fig. 1). Because targeting of some proteins to the cell membrane takes place through post-translational modifications by N-glycosylation or glycosylphosphatidylinositol linkage (18 -20), we examined whether these modifications underscore the membrane association by the spacer region. However, flow cytometry and Western blot analysis confirmed that neither site-directed mutagenesis of the only glycosylation site (N466Q) residing within the spacer region nor treatment of transfected cells with phosphatidylinositol-specific phospholipase C affected the surface expression of SCUBE1-D7 mutant protein in HEK-293T cells (data not shown). Thus, N-linked FIGURE 7. The C-terminal region of SCUBE1 acts as a BMP antagonist. A, design of a co-culture assay to assess the effect of SCUBE1 co-expression on the BMP2 signaling activity. Conditioned media from signaling cells (HEK-293T) co-transfected with BMP2 alone or together with various SCUBE1 expression plasmids as indicated were added to the responding cells (HepG2) that contained the BMP-responsive luciferase reporter plasmid I-BRE-Luc (16). B, coexpression of SCUBE1-D4 mutant attenuates the BMP2 signaling activity. BMP2 protein expressed alone or in combination with SCUBE1-FL, ty97, D5, or D6 in conditioned media of transiently transfected HEK-293T cells are active in this signaling assay. However, BMP2 activity can be attenuated by co-expression of the SCUBE1-D4 mutant. C, SCUBE1-D4 mutant inhibits BMP2 precursor processing. HEK-293T cells were transfected with indicated expression plasmids for BMP2 and SCUBE1 mutants. Western blotting of the cell lysates and of the conditioned medium shows that transfected BMP2 precursor is cleaved into active, mature BMP2, which is markedly inhibited in the presence of the SCUBE1-D4 mutant, but not with all other mutant SCUBE1 forms. ProBMP2, precursor BMP2. glycosylation or glycosylphosphatidylinositol linkage is unlikely to be involved in the spacer region-mediated cell surface anchorage of SCUBE1. Although the spacer region contains no classic transmembrane segment, helical plotting analysis identified an amphipathic ␣-helix (residues 567-589) similar to the membrane-associated, lipid-embedded amphipathic helix structures described in apolipoproteins (21) and Cubilin (22). However, whether this amphipathic ␣-helix or other mechanisms contribute to the membrane anchorage of SCUBE1 requires further investigation.
In addition, although the Scube1 gene is expressed in various embryonic tissues in the mouse, including the ventral forebrain, limb buds, somites, and developing gonads (2), we observed abnormality only in the forebrain development in Scube1 ⌬cub/⌬cub mice. A simple explanation for the requirement of SCUBE1 in only the ventral forebrain development is that other SCUBE members or factors with redundant function might compensate for SCUBE1 in other regions of the Scube1 ⌬cub/⌬cub embryo.
For the four EGF-like repeat deletion constructs tested, we consistently detected reduced aggregation between the A2058-SCUBE1-E1-9 cells as compared with A2058-SCUBE1-E4 -9 or -E7-9 cells (Fig.  3). One explanation for this reduced aggregation could be that the SCUBE1-E1-9 construct simply does not express as much protein at the cell surface as the SCUBE1-E4 -9 or -E7-9 construct (Fig. 2C), thereby diminishing the strength of the interaction. Alternatively, the difference in strength of interaction may indicate a fundamental functional difference between SCUBE1-E1-9 and SCUBE1-E4 -9/-E7-9. For example, the first three EGFlike repeats 1-3 may negatively regulate the cell adhesive function of SCUBE1 mediated by the EGF-like repeats 7-9 in vivo. Clearly, additional studies are needed to clarify this issue.
In addition, our deletion analysis demonstrated that the EGF-like repeats 7-9 appear to function as an independent molecular unit that is sufficient for the formation of SCUBE1-mediated homophilic adhesions between adjacent cells, even in the absence of most of the EGF-like repeats (Fig. 3). However, the A2058 cells expressing the deletion mutant lacking only the EGF-like repeats 7-9 (⌬E7-9) could still form significant numbers of aggregates as compared with the parental cells (Fig. 3). Together, these results reveal that the EGF-like repeats 7-9 of SCUBE1 are sufficient, but not necessary, for trans-interactions between SCUBE1 molecules on adjacent cells in a Ca 2ϩ -dependent fashion. In accordance with our previous study (6), these findings further support the adhesive and pathological role for the surface-exposed platelet SCUBE1 in mediating the platelet-platelet agglutination under thrombotic conditions.
The EGF-like motif is characterized by a conserved arrangement of six-cysteine residues that form three pairs of disulfide bonds (23). Furthermore, most of these repeats (6 out of 9, repeats 1-3 and 7-9) in SCUBE1 have calcium-binding consensus sequences: (D/N)X(D/N)(E/Q)X m (D/N*)X n (Y/F) (where m and n are variables and the asterisk indicates possible ␤-hydroxylation) (23). The Ca 2ϩ -binding EGF-like repeat is a common structural module used by cell surface proteins and can be found in extracellular proteins involved in mediating proteinprotein and protein-carbohydrate binding interactions (23,24). They are also important in functionally diverse processes, such as cell adhesion, receptor-ligand interaction, cell fate decision, blood coagulation, cholesterol uptake, and maintenance of extracellular matrix architecture (11).
Although we have not yet demonstrated that these EGF-like repeats actually bind Ca 2ϩ ions, the Ca 2ϩ -dependent nature of the SCUBE1 EGF-like repeat-mediated reciprocal interactions is consistent with the role of Ca 2ϩ in the biological functions of other EGF-like repeat-containing proteins (11). NMR studies have shown that the Ca 2ϩ ion plays a key role in maintaining the conformational rigidity and binding surfaces of calcium-binding EGF-like repeats (25,26). However, whether the calciumbinding EGF-like repeats in SCUBE1 also undergo a conformational rearrangement upon calcium binding and how these Ca 2ϩ -induced conformational changes account for the homophilic reciprocal interactions between SCUBE1 proteins require further investigation.
The CUB domain is a 110-amino acid-long, widespread module found in developmentally regulated proteins (27) and is known to form a ␤-barrel (28). The acronym CUB (complement C1r/C1s, Uegf, and Bmp1) is derived from the first letters of the initial founding members of this protein group (27). The CUB domains have been implicated in protein-protein and protein-carbohydrate interactions (27) and may confer specificity to ligand binding; for example, different CUB domains in Cubilin have been shown to bind to and promote subsequent endocytosis of specific ligands such as intrinsic factor-cobalamin complex, albumin, immunoglobulin light chains, and high density lipoprotein, and transferrin (29). Our study is the first demonstration of BMP protein as a specific ligand for the CUB domain (Fig. 6). It is of interest to further examine whether the CUB-mediated BMP binding/antagonism could be extended to other members of the TGF-␤ superfamily. A, lateral view of wild-type (ϩ/ϩ) and Scube1-mutant (⌬cub/⌬cub) neonates. Note the absence of the cranium (between arrows) and the lack of brain tissue. B and C, skull defects in the Scube1 ⌬cub/⌬cub mice. Lateral view of wildtype (B) and Scube1 ⌬cub/⌬cub (C) skulls stained with Alcian blue for cartilage and alizarin red for ossified bone. Note that cranial vault is missing in the mutant (asterisk). Only remnants of the frontal, pariental, squamosal, and alisphenoid bones remain in the mutant. The interparietal and supraoccipital bones are absent. The asterisk denotes exencephaly in the midbrain region. Our deletion analysis and biochemical/molecular data are broadly in line with genetic evidence (14), thus confirming that the 6-cysteine repeat motifs and CUB domain are essential for the SCUBE function in attenuating BMP activity when co-expressed in the same cells (Fig. 7). Furthermore, the permissive nature for the full-length SCUBE products may require as-yetunidentified cofactors, possibly proteases, to proteolytically release the active C-terminal fragment for its anti-BMP activity when present in the same cells. The signaling activities of BMPs could be antagonized by a diverse group of binding proteins, including Noggin and Chordin (30). The three-dimensional structure of the Noggin-BMP-7 complex revealed that Noggin binds as a dimer to a BMP dimer, occluding the interfaces of the binding domains for its receptors (31). Because of no sequence homology between the SCUBE1 C-terminal CUB domain and Noggin, a different mechanism may operate in the case of the SCUBE1-BMP2 complex. However, the precise mechanism by which SCUBE1 exerts its suppressive effects on BMP signaling remains to be further investigated.
Consistent with its in vitro anti-BMP activity, the Scube1 gene by targeted disruption in the mouse results in the brain patterning defects (Figs. 8 -10). Similar defects are seen in mice lacking the BMP antagonist Noggin with enhanced signaling through BMPs, central regulators of neural tube patterning (32,33). At E12.5, we found the SCUBE1 protein predominantly expressed in the cranial neural tube, with a lower level in the head mesenchyme, which implies its role in craniofacial development (Fig. 8C). Consistently, Scube1 ⌬cub/⌬cub embryos were characterized by abnormal hyperplasty in the cranial ganglia and a thick-walled neuroepithelium in the forebrain. Furthermore, the increase in neuroepithelial wall thickness coincided with excessive cell proliferation in the forebrain neuroepithelium, collapsing the third ventricle into a small canal-like structure (Figs. 8C and 10F). Because complex and dynamic interactions between the neuroepithelium and the head mesenchyme are required for cranial neural tube morphogenesis, defects in one region of the developing neural tube may alter the neurolation process in other cranial regions, eventually affecting the entire cascade of craniofacial development (34). Thus, the forebrain defects may block the initiation of neural tube closure at the forebrain/midbrain boundary. Despite this morphogenetic failure, the midbrain neuroepithelium continues to proliferate in exencephalic Scube1 ⌬cub/⌬cub embryos and becomes a large convoluted mass of brain tissue protruding from the head (Fig. 10B).
Because exencephaly could result from defects in neural crest emigration, impaired closure of the cranial neural tube, altered neuroepithelial proliferation and/or programmed cell death of neurons or neuronal precursor cells (34), it is possible that hyperproliferation of neuroepithelium seen in the Scube1-mutant E12.5 embryos might be a secondary phenotype caused by above-mentioned developmental processes. Further analysis of earlier E8.5 to E10.5 embryos is needed to elucidate the primary defect in developing brain caused by the disruption of the Scube1 gene.
Human cranial neural tube closure defects (NTDs) are among the most common types of lethal birth defects, with a frequency of 0.2-3.5 per 1000 births depending on ethnicity and geographical location (35). Studies of mouse mutant models with various NTDs have provided some potential cellular and molecular mechanisms underlying these birth defects; however, no mutations identified in mice have yet been shown to represent major genes for NTDs in humans. Thus, SCUBE1 may be a candidate for human NTDs, and the Scube1 mutant mice could provide a novel animal model for studying the cellular, molecular, and developmental mechanisms involved in the generation of NTDs.
In summary, the present molecular dissection of the SCUBE1 molecule has clearly defined the specific domains critical for membrane association, cell-cell adhesion, and BMP binding/ antagonism, respectively. Further structural and functional studies of these domains will provide insights to better understand the molecular action and regulation of this novel multifunctional, multidomain surface protein SCUBE1 during development or postembryonic biological processes. A conditional allele of the Scube1 gene is important to further investigate the stage-specific role during the brain development and the tissuespecific function in adulthood.