Requirement of Glycosylphosphatidylinositol Anchor of Cripto-1 for trans Activity as a Nodal Co-receptor*

Cripto-1 (CR-1) has an indispensable role as a Nodal co-receptor for patterning of body axis in embryonic development. CR-1 is reported to have a paracrine activity as a Nodal co-receptor, although CR-1 is primarily produced as a glycosylphosphatidylinositol (GPI)-anchored membrane protein. Regulation of cis and trans function of CR-1 should be important to establish the precise body patterning. However, the mechanism by which GPI-anchored CR-1 can act in trans is not well known. Here we confirmed the paracrine activity of CR-1 by fluorescent cell-labeling and immunofluorescent staining. We generated COOH-terminal-truncated soluble forms of CR-1 based on the attachment site for the GPI moiety (ω-site), which we identified in the present study. GPI-anchored CR-1 has a significantly higher activity than COOH-terminal-truncated soluble forms to induce Nodal signal in trans as well as in cis. Moreover, transmembrane forms of CR-1 partially retained their ability to induce Nodal signaling only when type I receptor Activin-like kinase 4 was overexpressed. NTERA2/D1 cells, which express endogenous CR-1, lost the cell-surface expression of CR-1 after phosphatidylinositol-phospholipase C treatment and became refractory to stimulation of Nodal. These observations suggest that GPI attachment of CR-1 is required for the paracrine activity as a Nodal co-receptor.

Establishment of the basic body plan of vertebrates is strictly regulated by a number of morphogens and signaling molecules such as Wnts, Hedgehog, fibroblast growth factors, and members of transforming growth factor ␤ family ligands (1)(2)(3)(4). Among them, a transforming growth factor ␤ family ligand, Nodal, is known to be indispensable for early embryonic development. Nodal acts as a morphogen to initiate gastrulation and to establish the anterior-posterior and the left-right body axis (5,6). Nodal utilizes a shared signaling pathway with other transforming growth factor ␤ family ligands such as Activin, which activates Smad2/3 through phosphorylation to interact with Smad4. This oligomeric complex then translocates into nucleus to regulate transcription of target genes through a tran-scriptional co-factor, FOXH-1 (Fast-1) (5,7). Unlike other transforming growth factor ␤ family ligands, Nodal requires epidermal growth factor-Cripto-1/FRL-1/Cryptic (EGF-CFC) 2 family proteins to bind its type I receptor Activin-like kinase 4 (ALK4) (5,8). To regulate precise formation of body axis, the spatiotemporal activity of these extracellular molecules should be strictly regulated. Expression and localization of Nodal itself are known to be regulated by several mechanisms (5). For example, Nodal gene expression is strictly regulated by node-specific enhancer and left-side-specific intronic enhancer (9,10). In addition, laminar leftward flow, which is generated by constant movement of cilia of embryonic cells, contributes to the left-side localization of Nodal ligand (11,12). Requirement for EGF-CFC proteins as coreceptors for Nodal suggests that the activity of Nodal signaling could also be regulated by the spatiotemporal localization and function of EGF-CFC proteins.
EGF-CFC proteins contain several domains that include an NH 2 -terminal signal peptide, a variant EGF-like domain, a cysteine-rich CFC domain, and a hydrophobic membrane-associated domain at the COOH terminus. Most of EGF-CFC proteins including human/mouse Cripto-1 (CR-1/Cr-1) have been experimentally shown (13) or predicted (14) to possess a glycosylphosphatidylinositol (GPI) anchorage. Although EGF-CFC proteins are primarily synthesized as GPI-anchored membrane proteins, several studies have shown trans activity of EGF-CFC proteins in vitro and in vivo (15)(16)(17)(18)(19). A chimeric mouse model revealed that cells derived from Cr-1-null embryonic stem (ES) cells can contribute to the formation of mesoendoderm in the presence of Cr-1 wild-type cells (15,17). In addition, COOH-terminaldeleted forms of Cr-1 or recombinant Cr-1 protein could rescue the One eye pinhead (oep)-mutant phenotype in zebrafish (18) or could induce cardiomyocyte differentiation in Cr-1 null ES cells (19). However, the mechanism by which GPI-anchored forms of EGF-CFC proteins can act as soluble co-receptors for Nodal signaling is not known.
In the current study we generated several variants of human CR-1 including soluble and transmembrane forms based on the View, CA). Other expression vectors were described previously (20). Transfections were performed using Lipofectamine 2000 (Invitrogen).
Immunofluorescence and Fluorescent Cell Labeling-Cells were seeded in 4-well-chambered coverglass (Nunc, Rochester, NY) that had been coated with poly-L-lysine (Sigma-Aldrich) to avoid detachment of cells. For cell-labeling and co-culture assay, 293T cells stably transfected with wild-type CR-1 or NTERA2/D1 cells that express endogenous CR-1 were labeled with CellTracker green 5-chloromethylfluorescein (Invitrogen) before co-culture. After co-culturing overnight in the regular growth medium, cells were serum-starved for 24 h, and a 200 or 50 ng/ml final concentrations of Nodal or Activin, respectively, were added without changing the medium and stimulated for 1 h. After washing with phosphate-buffered saline, cells were fixed in 4% paraformaldehyde and then permeabilized with 0.2% Triton X-100. After blocking with phosphate-buffered saline containing 10% bovine serum albumin and 10% normal goat serum, samples were incubated with 1:200 dilution of anti-total Smad2 antibody overnight at 4°C. Primary antibody was then labeled with Alexa Fluor 596-conjugated secondary antibody (1:200,Invitrogen). For co-staining of CR-1 and Smad2, CR-1 staining was carried out for 1 h at room temperature with 5 g/ml MAB2771 that had been labeled with Alexa Fluor 488 using Zenon tricolor mouse IgG1 labeling kit (Invitrogen). Counterstaining for nuclei was performed with DAPI. For quantification of nuclear Smad2-positive cells, images were taken with a fluorescent microscope equipped with a 40ϫ objective lens, and only cells with clear nuclear staining of Smad2 were counted as nuclear Smad2-positive cells. Counting was performed in blinded manner. For each experiment total 177-249 cells/field were counted in two different fields, and three independent experiments were performed.
For studies on the intracellular localization of CR-1 mutants, 293T cells were transiently transfected with each CR-1 mutant vector and the DsRed-Golgi expression vector. After 24 h of transfection, fixation and permeabilization were performed as described above, and CR-1 proteins were stained with 5 g/ml MAB2771 and Alexa Fluor 488-conjugated secondary antibody. Endoplasmic reticulum (ER) was visualized with anti-calnexin rabbit polyclonal antibody (Santa Cruz, Santa Cruz, CA) at 1:250 dilution and Alexa Fluor 596-conjugated secondary antibody (1:200, Invitrogen). Images were taken with a Zeiss LSM 510 NLO Meta confocal system (Carl Zeiss, Göttingen,

Mutant
Mutagenesis primer sequence S161L GATGAGCACCTCGTGGCTCTCAGGACTCCAGAACTACCACCG S161L/R162L GATGAGCACCTCGTGGCTCTCCTGACTCCAGAACTACCACCG S161L/R162L/T163L GATGAGCACCTCGTGGCTCTCCTGCTTCCAGAACTACCACCG S169L ACTCCAGAACTACCACCGCTTGCACGTACTACCACTTTTATGCTA Germany) with an Axiovert 200M inverted microscope equipped with a 63ϫ NA 1.4 Plan-Apochromat oil immersion objective lens. Images were collected with Zeiss AIM software using a multi-track configuration. Western Blot Analysis-Western blot analysis was performed using 16% (for detection of CR-1 protein) or 10% (for detection of phospho-and total Smad2) SDS-PAGE gels (Invitrogen). Depending on the experiments, 30 -50 g of total cell lysates or 40 l of conditioned medium were loaded. CR-1 protein was detected with B3F6 mAb at a 1:5000 dilution and anti-mouse IgG horseradish peroxidaseconjugated secondary antibody (1:3000; Amersham Biosciences). All images of Western blot analysis in this work were visualized, processed, and quantified with an Image Analyzer equipped with LabWorks software (Ultra Violet and Laboratory Products, Upland, CA).
Phase Separation-Phase separation by Triton X-114 was performed as previously described (23). Cells were lysed in TX-114 solution (20 mM Tris, pH 8.0, 150 mM NaCl, 2% TX-114) for 1 h on ice. Phase separation was carried out by warming up to 37°C and subsequent centrifugation at 3000 ϫ g at 25°C. The upper aqueous phase and the lower detergent phase were collected by micropipettes carefully. Before Western blotting, proteins were precipitated with chloroform-methanol (1:4) to remove the detergent.
Preparation of Conditioned Medium-293T transfectants were incubated with Opti-MEM serum-free medium (Invitrogen) for 24 h. Conditioned medium was collected and centrifuged at 3000 ϫ g for 5 min to remove cellular debris. Recombinant Nodal protein was then added at a final concentration of 200 ng/ml and immediately used for stimulation of cells.
Dual Luciferase Assay-293T cells, which had been plated in 24-well culture plates, were transfected with an optimized amount of expression vectors ((n2)7-Luc, 50 ng/well; TK-renilla, 5 ng/well; mFast-1, 25 ng/well; mNodal-V5, 100 ng/well; ALK4-HA, 50 ng/well; each CR-1 mutant, 5-1000 ng/well). Depending on the experiments, recombinant Nodal or Activin B (R&D Systems) were used instead of mNodal-V5 expression vector. pCI neo or pEF6/V5-His empty vector was added to adjust for total amount of DNA. After 24 h of transfection or addition of ligands, dual luciferase assays were carried out using a kit provided by Promega according to the manufacturer's instructions.
Statistical Analysis-Student's t test was used to determine the statistical significance of the quantitative results. Results with a p value Ͻ0.05 were considered statistically significant.

RESULTS
Paracrine Activity of Wild-type CR-1-To confirm the paracrine function of human CR-1, we demonstrated the direct visualization of the trans activity of CR-1 in cell culture using immunofluorescence and fluorescent cell-labeling. Transient transfection of wild-type (WT) CR-1 into 293T cells that do not express CR-1 (16) achieved up to 50 -60% transfection efficiency resulting in the mixed population of CR-1 positive (transfected) and negative (untransfected) cells. We assessed the effect of Nodal and Activin on translocation of the Nodal/ Activin intracellular mediator, Smad2, into nuclei by immunofluorescent staining. The localization of Smad2 in 293T cells that had been transfected with empty vector (EV) was mainly cytosolic (data not shown), and these cells did not respond to Nodal stimulation (Fig. 1A). 293T cells transfected with WT CR-1 also showed cytosolic localization of Smad2 without stimulation (Fig. 1B). In contrast to EV-transfected cells, stimulation of WT CR-1-transfected cells with Nodal induced Smad2 nuclear translocation in ϳ70 -80% of the total population (Fig.  1C). Smad2 nuclear localization was observed both in CR-1staining-positive cells and in CR-1-staining-negative cells (Fig.  1C, arrowheads). Activin stimulated Smad2 nuclear translocation in almost 90 -100% of the WT CR-1-transfected cells (Fig.  1D). This effect of Activin was also observed in EV-transfected or untransfected 293T cells (data not shown), demonstrating the CR-1-independent activity of Activin on Smad2 nuclear translocation.
These data strongly indicate the paracrine activity of CR-1 in Nodal signaling. However, it was possible that WT CR-1-transfected cells, which did not show positive staining for CR-1, might express an undetectable but functional level of CR-1 protein. To exclude this possibility, we performed co-culture experiments of CR-1-stably transfected 293T cells and wildtype 293T cells (Fig. 1E) using a cell-labeling technique to segregate the CR-1-positive and negative cells (Fig. 1F). Consistent with the observation in transiently transfected 293T cells (Fig.  1A), a mixture of untransfected 293T cells and labeled stable transfectants of EV did not respond to Nodal (Fig. 1G). In contrast, Nodal stimulation of the mixture of untransfected 293T cells and labeled stable transfectants of WT CR-1 significantly induced Smad2 nuclear translocation not only in CR-1-positive cells but also in ϳ16% of the CR-1-negative cells even though only 10% of the mixed population was the labeled, WT CR-1 stable transfectants (Fig. 1, H and J). Activin was more potent than Nodal in inducing Smad2 nuclear translocation both in labeled CR-1-positive and in unlabeled CR-1-negative cells (Fig. 1, I and J). To ascertain if this observation could be reproduced in cells that express endogenous CR-1 protein, we performed co-culture experiments using CR-1-deficient 293T cells and human embryonal carcinoma NTERA2/D1 cells from which CR-1 was isolated and cloned and which express a relatively high level of endogenous CR-1 (14) ( Fig.  2A). CellTracker-labeled NTERA2/D1 cells responded to Nodal almost equally to Activin (Fig. 2, C-E). A significant population of unlabeled 293T cells responded to Nodal stimulation especially in the cells that were located proximally to NTERA2/D1 cells (Fig. 2, C, arrowheads, and E). From these results, we concluded that ectopically expressed or endogenously expressed wild-type CR-1 can act as a Nodal coreceptor in trans.
Identification of a Functional -Site of CR-1-We have recently demonstrated that the GPI moiety is attached to Ser-161 of CR-1 in the COOH-terminal region by mass spectrometric analysis of "shed" CR-1 in the conditioned medium (22). To confirm that Ser-161 is a functional -site for GPI attachment, we performed a mutational analysis of the COOH-terminal sequence of CR-1. Because Ser-169 of human CR-1 had also previously been predicted as a possible -site for GPI-anchorage (14) ("big-⌸ predictor"), we introduced point mutations in these two possible -sites and also at the ϩ1 and ϩ2 sites. Because previous reports have described that the -site should be a small hydrophilic amino acid such as Ser, Asn, Ala, Asp, Gly, and Cys (23), possible , ϩ1 , or ϩ2 sites were substituted by a hydrophobic amino acid, Leu (S161L, S161L/R162L, S161L/R162L/T163L, S169L, S169L/A170L, S169L/A170L/ R171L) or a small hydrophilic amino acid, Asn (S161N). Western blot analysis of cell lysates revealed that the WT CR-1 protein was found primarily as the 26-and 24-kDa forms and to a lesser extent as a 18-kDa form, which may be due to different glycosylated forms (Fig. 3A) (14,24). The 26-kDa form of CR-1 was not present in S161L series (S161L, S161L/R162L, and S161L/R162L/T163L) mutants, whereas the S169L series (S169L, S169L/A170L, S169L/A170L/R171L) and S161N mutants contained the 26-kDa form and showed similar band patterns to WT CR-1 (Fig. 3A). In the conditioned medium of the S161L series mutants, secreted CR-1 proteins were not detectable (Fig. 3B). On the other hand, similar amounts of the CR-1 protein to WT CR-1 were detected in the conditioned medium of both the S169L series and S161N mutants (Fig. 3B).
The size of the bands in the conditioned medium showed slight differences from the cell lysates (30-and 28-kDa forms, respectively). Higher amounts of the COOH-terminal-truncated mutant protein ⌬C (Ser-161) were detected in the conditioned medium. Triton X-114 phase partitioning revealed that although the WT and the S161N mutant were mostly partitioned into the detergent phase, the S161L mutant protein was found to be enriched in the aqueous phase (Fig. 3C), suggesting that the S161L mutant protein is relatively hydrophilic and lacks a GPI anchor. A transmembrane (TM) form of CR-1 was found both in the detergent phase and in the aqueous phase (Fig. 3C). We then performed FACS analysis to evaluate cell-surface expression of each CR-1 mutant in transiently transfected 293T cells (Fig. 4A). The S169L series and S161N mutants showed strong cell-surface fluorescence of PE-conjugated anti-CR-1 antibody similar to WT CR-1. In contrast, cells expressing the S161L series mutants were not positively stained compared with EV-transfected cells as a negative control, confirming that Ser-161 is required for expression of CR-1 on cell surface. To evaluate the intracellular localization of each of the CR-1 mutants, we performed immunocytochemical analysis on transiently transfected 293T cells. A Golgi localization as well as plasma membrane localization was observed in the WT, S161N, and S169L mutants as assessed by co-transfection with a Golgi marker, DsRed-Golgi (Fig. 4, B, D, and E). In contrast, the S161L mutant was not found either in the Golgi or on the plasma membrane in most cases but exhibited aberrant punctate intracellular cytoplasmic localization (Fig. 4C). Other S161L series mutants showed a similar aberrant localization to the S161L mutant, whereas other S169L series mutants were similar to the S169L mutant in Golgi localization (data not shown). The ⌬C (Ser-161) mutant, which had been generated by inserting a stop codon just after Ser-161, exhibited trans-Golgi network staining pattern without any staining of plasma membrane (Fig. 4F). The punctuate staining of S161L mutant protein co-localized with ER, which was labeled with the ER marker calnexin (Fig. 4G). These results are consistent with the previous reports which have demonstrated that uncleavable mutants of GPI-anchored proteins are retained in a post-ER compartment (23,25), suggesting that substitution of -site with a hydrophobic amino acid may interfere the ER-to-Golgi translocation.
To ascertain potential differences in biological activity among the various CR-1 mutants, we examined the ability of the CR-1 mutants to activate Nodal-dependent signaling in 293T cells. As a readout of Nodal signaling, we analyzed the activity of the Activin/Nodal-responsive (n2)7-luciferase reporter ((n2)7-Luc) which contains mouse Fast-1 (FOXH1) binding sites from the Nodal left side-specific enhancer (43) (Fig. 4H) and the phosphorylation status of Smad2 (Fig. 4I). For (n2)7-Luc reporter assay, we used TK renilla reporter as an internal standard. Mouse Fast-1 (mFast-1), mouse V5-tagged Nodal (mNodal-V5), and human HA-tagged ALK4 (ALK4-HA) were co-transfected at the concentrations described under "Experimental Procedures." The ALK4-HA expression vector was used to enhance the signal. Phosphorylated Smad2 was analyzed by Western blotting. S161L series mutants significantly lost the ability to induce Nodal signaling in these two assays, whereas S169L series and S161N mutants retained this ability (Fig. 4, H and I). These results suggest that Ser-161 is a functional -site for GPI-anchoring in CR-1 and that substitution of Ser-161 with a hydrophobic amino acid such as Leu, but not with a small hydrophilic amino acid such as Asn, affects the correct processing, localization, and function of CR-1 as a Nodal co-receptor.
We constructed a TM form of CR-1 in which the transmem-brane domain of the EGF receptor-related receptor ErbB4 was attached just after Ser-161. We also prepared a ⌬C (Ser-169) construct in which a stop codon was inserted after Ser-169. The - WT S161L ∆C (S161) S169L G2 G1 G3 S161L Nodal-dependent (n2)7-Luc reporter assay was performed using these CR-1 variants. WT and TM forms of CR-1 significantly induced the (n2)7-Luc activity, whereas both ⌬C mutants of CR-1 did not activate Nodal-responsive reporter at a detectable level after co-transfection with mFast-1, ALK4-HA, and mNodal-V5 expression vectors (Fig. 5B). We also confirmed the dose-dependent effect of ligands using either recombinant Nodal or Activin instead of mNodal-V5 expression vector (Fig. 5, C and D). Recombinant Nodal was only able to induce (n2)7-Luc activity dose-dependently in cells transfected with WT or TM CR-1 but not in ⌬C (Ser-161)-transfected cells (Fig. 5C). In contrast, Activin was able to induce the (n2)7-Luc activity regardless of the type of CR-1 expression vectors (Fig. 5D), confirming the CR-1-independent activity of Activin. To address the interactions of the CR-1 variants with ALK4 or Nodal, co-immunoprecipitation experiments were performed using the CR-1 variants with epitope-tagged ALK4 and Nodal (Fig. 6). Although all three CR-1 variants could bind to ALK4-HA to a similar extent (Fig. 6A), the ability of ⌬C (Ser-161) to form a com-plex with mNodal-V5 on the cell surface was markedly lower than that of WT or TM forms of CR-1 as assessed by coimmunoprecipitation after reversible cross-linking with the membrane-impermeable cross-linker DTSSP (Fig. 6B). A similar result was observed in the ⌬C (Ser-169) form of CR-1 (data not shown). We then directly compared the trans activity of GPI-anchored and soluble forms of CR-1. Smad2 phosphorylation status after stimulation with recombinant Nodal was significantly enhanced in WT CR-1-transfected 293T cells as compared with EV-transfected cells (Fig. 7A). However, ⌬C (Ser-161)-transfected cells did not show any increase of Smad2 phosphorylation over EV-transfected cells (Fig. 7A), which is consistent with the results of (n2)7-Luc assay (Fig. 5). Conditioned medium obtained from WT CR-1-transfected 293T cells significantly induced Smad2 phosphorylation in CR-1-deficient wild-type 293T cells in the presence of recombinant Nodal, but the conditioned medium from ⌬C (Ser-161)-transfected cells did not show any difference compared with the conditioned medium from the EV-transfected cells even though Nodal Signaling and Cripto-1 DECEMBER 7, 2007 • VOLUME 282 • NUMBER 49 the amount of immunoreactive ⌬C (Ser-161) form of CR-1 protein in the conditioned medium was ϳ30-fold higher than WT CR-1 protein found in the conditioned medium (Fig. 7B). Conditioned medium of WT CR-1-transfected 293T cells could not induce Nodal signaling in the absence of Nodal (data not shown). Results that were similar to the ⌬C (Ser-161) were also observed with the ⌬C (Ser-169) form (data not shown). These results were also confirmed by the (n2)7-Luc reporter assay (Fig. 7C). To compare differences in physiochemical properties between the released GPI-anchored form and COOH-terminal-deleted soluble form in the conditioned medium, Western blot analysis under reducing and non-reducing conditions was carried out (Fig. 7D). Because the amount of the GPI-anchored form released in the conditioned medium is markedly lower than that of the soluble form (Fig. 7B), concentrated medium was analyzed for the GPI-anchored form of CR-1. As shown in Fig. 7D, the released GPI-anchored proteins were detected as major two species of 30-and 28-kDa, which were also found in the PI-PLC-treated samples of the cell lysates. On the other hand, the ⌬C (Ser-161) protein in the conditioned medium was found as a major band of 22-24 kDa. There is a clear difference in size and pattern between the released WT and ⌬C (Ser-161) forms of CR-1 in the conditioned medium even though the released GPI-anchored form and ⌬C (Ser-161) form should theoretically contain the identical amino acid sequence. Treatment with a reducing agent did not affect the SDS-PAGE mobility of each band in both WT and ⌬C (Ser-161) of CR-1. However, the antigenicity of the released WT of CR-1 was completely lost under non-reducing conditions, whereas the ⌬C (Ser-161) protein in the conditioned medium was still detected under the same conditions. This finding may suggest the presence of structural or conformational differences between the released GPI-anchored form and the soluble form of CR-1 in the conditioned medium.
These results are at variance with the previous reports that have demonstrated the activity of different COOH-terminal-truncated forms of mouse Cr-1 as summarized in Fig. 8A and Table 2. Therefore, we introduced similar COOH-terminal truncations into human CR-1 as had been introduced into mouse Cr-1 in the previous reports (18,19,26,27) (Fig. 8A and Table 2). Western blot analysis of cell lysates revealed that ⌬C (T174) and ⌬C (T174)-V5His constructs generated some additional species with slower mobility in SDS-PAGE (Fig. 8B). These species were likely to be highly glycosylated forms in the COOH-terminal extension beyond the -site since there are several possible O-linked glycosylation sites including Ser-161 (the -site for GPI-anchored form), Ser-169, and Thr-172 to Thr-174. These highly glycosylated species of ⌬C (T174) and ⌬C (T174)-V5His mutants were also found in the conditioned medium (Fig. 8C). The GPI-V5His mutant in which V5-His 6 tag had been added at the end of COOHterminal hydrophobic domain showed an expected SDS-PAGE mobility from its amino acid sequence (Fig. 8A) and was difficult to detect in the conditioned medium (Fig. 8B).
We then tested the activity of these COOH-terminal variant forms of CR-1 in Nodal signaling in transiently transfected 293T cells. None of the COOH-terminal-truncated variants of CR-1 such as ⌬C (Ser-161), ⌬C (Ser-161)-V5His, ⌬C (T174), or ⌬C (T174)-V5His were able to mediate significant Nodal signaling, as assessed by (n2)7-Luc assay (Fig. 8D) and by Smad2 phosphorylation (data not shown). On the other hand, GPI-V5His retained some reduced activity to induce Nodal signaling in transiently transfected 293T cells with co-transfection of mFast-1, mNodal-V5, ALK4-HA expression vectors (Fig. 8D). The GPI-V5His construct corresponds to the COOH-terminal FLAG-tagged construct of mouse Cr-1 which was used in a previous study and was shown to interact with ALK4 and Nodal (26). GPI-V5His was also able to induce significant activities of the (n2)7-Luc reporter with exogenous recombinant Nodal protein (Fig. 8E).
Reduced Activity of Transmembrane Forms of CR-1 in Nodal Signaling-We then assessed the difference in the activity of the GPI-anchored and transmembrane forms of CR-1 to induce Nodal signaling. As shown in Fig. 5, the activity of the TM form was reduced up to ϳ30 -40% compared with WT CR-1. We confirmed this reduction of activity by titration of the DNA amount used for transfection (Fig. 9A). The reduced activity of this artificial TM construct of CR-1 could be due to the distance and/or flexibility of linker region between the transmembrane domain and the EGF-CFC domain. Therefore, we introduced several different lengths of flexible poly-glycine-serine (poly-GS) spacer linkers ((GGGGS) n ) between the functional -site (Ser-161) and the ErbB4 transmembrane domain (Fig. 9B). These TM variants with different lengths of the poly-GS linker did not show any significant difference in their activity to induce Nodal signaling (Fig. 9C), suggesting that the length and flexibility of the linker region may not be critical for interaction with Nodal and ALK4. We then tested if the TM forms of CR-1 could induce Nodal signaling without ALK4 overexpression in transiently transfected 293T cells. (n2)7-Luc assays without overexpression of ALK4-HA revealed that transmembrane forms of CR-1 but not WT CR-1 require ALK4 overexpression to induce Nodal signaling (Fig. 9D).
Thr-72 of mouse Cr-1, which corresponds to Thr-88 in human CR-1 has been reported as a site of O-fucosylation and is necessary for Nodal signaling (24,28). Phe-78 of mouse Cr-1, which corresponds to Phe-94 in human CR-1, has also been reported to be necessary to rescue the oep-mutant phenotype in zebrafish (18). We, therefore, substituted these two residues with Ala by point mutations in the GPI-anchored (WT) and TM forms of human CR-1 (T88A and F94A, respectively, Fig. 9, E-G). GPI-anchored, WT CR-1 protein was found primarily as 26-and 24-kDa forms (Figs. 3A and 9E). In the TM forms of CR-1, a similar pattern was observed with the 34-and 32-kDa forms (Fig. 9E). The F94A mutant forms showed an identical band pattern to the non-mutated forms of both the GPI-anchored and TM forms of CR-1. On the other hand, in T88A mutant forms the larger bands (26-kDa in GPI-anchored forms or 34-kDa in the TM forms) were weaker than the non-mutated forms (Fig. 9E). FACS analysis revealed that the F94A mutants exhibited comparable cell-surface expression of immunoreactive CR-1 as compared with the non-mutated forms of both WT and TM forms (Fig. 9, F and G). However, the cell-surface expression of the T88A mutants was decreased by ϳ50% (Fig. 9,  F and G). We then assessed the activity of these mutants to induce Nodal signaling using (n2)7-Luc assay in 293T cells cotransfected with mFast-1, mNodal-V5, and ALK4-HA. In accordance with previous reports (24,28), the T88A mutant of the GPI-anchored form completely lost the ability to induce Nodal signaling (Fig. 9H), whereas the F94A mutant of the GPI- anchored form still retained some activity. Because the EGFlike domain of CR-1 is known to bind to Nodal (14), these results suggested that the T88A mutant was not able to bind to Nodal, and the F94A mutant still retained some ability to bind Nodal but possibly with weaker affinity. In contrast to the GPIanchored form, the F94A mutant of the TM form completely lost the ability to induce Nodal signaling, suggesting that the TM form of CR-1 could induce Nodal signaling only when the Nodal binding capacity is totally intact. These results suggest that the TM form of CR-1 requires high ALK4 expression as well as an intact Nodal binding capacity to function.
To ascertain the importance of GPI attachment of endogenous CR-1, we examined the effect of PI-PLC treatment on Nodal signaling using NTERA2/D1 cells.  Table 2. B-C, Western blot analysis of tagged or untagged COOH-terminal variants of CR-1 in total cell lysates (B) and conditioned medium (C). Immunoblotting was first performed for CR-1, and the same blots were reprobed and re-analyzed for V5-tag. 10A). We then tested the ability of Nodal and Activin to induce Smad2 phosphorylation in NTERA2/D1 cells treated under the same conditions (Fig 10B). PI-PLC treatment significantly decreased the ability of Nodal to induce Smad2 phosphorylation but PI-PLC treatment did not affect the activity of Activin. This result suggests that GPI attachment of endogenous CR-1 is critical to induce Nodal signaling.

DISCUSSION
In the present study we confirmed the paracrine activity of GPI-anchored CR-1 in transfected 293T cells, which is consistent with the previous report (16). We further demonstrated that cells which express endogenous CR-1 are also able to induce Nodal signaling in CR-1-deficient cells by co-culture. The fact that the inability of Cr-1-null mouse ES cells to form mesoendoderm can be rescued in a chimeric mouse of Cr-1null ES cells and wild-type embryonic cells (15,17) strongly suggests that the trans activity of GPI-anchored CR-1 is not negligible. However, previous reports are controversial as to the activity of different artificially generated soluble forms of Cr-1 (18,19,27) (Table 2). To design a suitable soluble human CR-1 construct, we identified the functional -site Ser-161 by structural and mutational analysis (22). Therefore, the ⌬C (Ser-161) protein should have an identical amino acid sequence with GPIanchored CR-1 after cleavage of the GPI signal. However, the ⌬C (Ser-161) protein almost completely lost the ability to induce Nodal signaling. Our results are consistent with a previous report which demonstrated the inability of a COOH-terminal-truncated form of mouse Cr-1 using a similar (n2)7-Luc reporter assay in Xenopus embryos (27), although their conclusion that Cr-1 functions solely as a "cell-autonomous" factor in Nodal signaling may not be correct. There are still some discrepancies between the present study and some of the previous reports which have demonstrated that COOH-terminal-deleted forms of mouse Cr-1 or zebrafish oep still retained biological activities (18,19,29,30). In these reports the activity of truncated forms of Cr-1 (or oep) were mainly measured by rescue experiments of the oep-mutant phenotype in zebrafish (18,29,30) or of the lack of the ability of mouse Cr-1-null ES cells to differentiate into cardiomyocytes (19). It might be possible that soluble forms of CR-1 could have an additional or different function(s) other than as a Nodal co-receptor. Additionally, this discrepancy might be explained by the difference in the sensitivity of the various assays used for the functional analysis ( Table 2). For example, the presence of extracellular matrix proteins that contain heparin sulfate-containing proteoglycans and that may sequester or locally concentrate a soluble form of CR-1 (14) and/or expression of endogenous inhibitors of CR-1 such as lefty proteins (21) might account for these differences. However, the present findings at least demonstrate that the deletion of GPI-signal sequence can reduce the activity of CR-1 as a Nodal co-receptor. Our conclusion is also supported by the clinical observation of human Cryptic (hCFC1), a homologue of human CR-1. A frameshift mutation at the beginning of the hydrophobic COOH-terminal domain in CFC-1 is related to human left-right laterality defects (31). In fact, this hCFC1 mutant was absent from the cell surface and failed to rescue a zebrafish oep-mutant phenotype (31) or to mediate Nodal signaling in vitro (16).
A previous study successfully demonstrated the interaction of Cr-1 with Nodal and its receptors using a construct of mouse Cr-1 that contained an epitope tag at the end of the GPI signal (26). A similar construct of human CR-1 that we generated in the present study (GPI-V5-His) retained some activity to induce Nodal signaling. However, the reduced activity of this COOH-terminal-tagged construct may not reflect the full physiological activity of wild-type CR-1 in mediating Nodal signaling. This is especially critical for immunoprecipitation assays that were performed using an epitope tag introduced after the GPI-signal sequence (26), since the hydrophobic COOH-terminal signal of wild-type CR-1 should be cleaved off during processing to the mature form of GPI-anchored CR-1.
The activity of the GPI-anchored form of CR-1 to induce Nodal signaling in trans was significantly higher than that of soluble forms of CR-1, since a nearly 30-fold larger amount of a soluble form of CR-1 in the conditioned medium failed to induce Nodal signaling. There are several possibilities that could explain the differences in biological activity to function as a Nodal co-receptor between the secreted GPI-anchored and soluble forms of CR-1. First, post-translational modification such as glycosylation of a soluble CR-1 protein might be different from that of the GPI-anchored form. The difference in the SDS-PAGE mobility as well as the loss of immunological detection of the secreted GPI-anchor form of CR-1 and the retention of immunoreactivity of the soluble form under non-reducing conditions may suggest that structural differences exist between these two forms. Indeed, differences in the glycosylation status and biological activity between GPI-anchored and soluble forms have been described for GPI-anchored proteins such as CD59 and prion protein (32,33). Second, GPI-anchoring of CR-1 might be necessary for the trans activity of CR-1. For example, potential CR-1-containing microvesicles released from the cell membrane might have a functional activity. Several reports have described that intact GPI-anchored proteins can be released in the form of membrane vesicles (34) and can transfer between cellular membranes (35)(36)(37). The fact that NTERA2/D1 cells, which express endogenous CR-1, lost their responsiveness to Nodal after PI-PLC treatment may support the possibility of the "microvesicle" hypothesis. To delineate the biological difference in activity between the GPI-anchored and soluble form of CR-1, we are now attempting to analyze potential structural differences between these two forms including post-translational modifications such as glycosylation and/or fatty acid acylation.
In contrast to the soluble forms of CR-1, the activity of the TM constructs to induce Nodal signaling was partially retained. In these transmembrane constructs the COOH-terminal sequence downstream of -site (Ser-161) was completely replaced with the transmembrane sequence of ErbB4. This strongly suggests the requirement of a membrane attachment signal of CR-1 to function as a Nodal co-receptor. However, the activity of the TM CR-1 constructs in Nodal signaling was reduced in comparison to the GPI-anchored form of CR-1. The different properties of these two membrane anchors may account for the different activities of these two forms. For example, GPI-anchored proteins predominantly localize in lipid raft microdomains on the plasma membrane, and the GPIanchored form of CR-1 is cleavable by the activity of mammalian phospholipases including GPI-phospholipase D, whereas transmembrane forms are not (22). Recent reports have shown that GPI-signal sequences and glycosylation can play a critical role as determinants of biological activity (38,39), suggesting that the structure of the GPI membrane anchor in conjunction with glycosylation could affect the biological function of membrane proteins. For example, N-glycosylation of mouse Nodal increases stability of the mature secreted form and thereby enhances the biological activity in vitro (40).
We have recently found that the same COOH-terminal-deleted constructs ⌬C (Ser-161) and ⌬C (Ser-169) can induce Nodal-independent, c-Src/MAPK/PI3k-Akt-dependent signaling and can promote endothelial cell migration (22). This suggests that the mechanism by which CR-1 acts in Nodal-dependent and Nodal-independent signaling pathways may be different. CR-1 is known as a multifunctional molecule and is important both in embryonic development and in tumor progression (14). Recent studies have also shown that CR-1 is a potential candidate as a tumor marker and as a target for tumor immunotherapy (8,41,42). This study provides additional information for understanding the molecular mechanisms of the action of CR-1.