Characterization of Gas6, a member of the superfamily of G domain-containing proteins, as a ligand for Rse and Axl.

Rse, Ax1, and c-Mer comprise a family of cell adhesion molecule-related tyrosine kinase receptors. Human Gas6 was recently shown to act as a ligand for both human Rse (Godowski et al., 1995) and human Ax1 (Varnum et al., 1995). Gas6 contains an NH2-terminal Gla domain followed by four epidermal growth factor-like repeats and tandem globular (G) domains. The G domains are related to those found in sex hormone-binding globulin and to those utilized by laminin and agrin for binding to the dystroglycan complex. A series of Gas6 variants were tested for their ability to bind to Rse and Ax1. The Gla domain and epidermal growth factor-like repeats were not required for receptor binding, as deletion variants of Gas6 which lacked these domains bound to the extracellular domains of both Rse and Axl. A deletion variant of Gas6 containing just the G domain region was shown to activate Rse phosphorylation. These results provide evidence that G domains can act as signaling molecules by activating transmembrane receptor tyrosine kinases. Furthermore, they provide a structural link between the activation of cell adhesion related receptors and the control of cell growth and differentiation by the G domain-containing superfamily of proteins.

Rse, Axl, and c-Mer comprise a family of cell adhesion molecule-related tyrosine kinase receptors. Human Gas6 was recently shown to act as a ligand for both human Rse (Godowski et al., 1995) and human Axl (Varnum et al., 1995). Gas6 contains an NH 2 -terminal Gla domain followed by four epidermal growth factor-like repeats and tandem globular (G) domains. The G domains are related to those found in sex hormone-binding globulin and to those utilized by laminin and agrin for binding to the dystroglycan complex. A series of Gas6 variants were tested for their ability to bind to Rse and Axl. The Gla domain and epidermal growth factor-like repeats were not required for receptor binding, as deletion variants of Gas6 which lacked these domains bound to the extracellular domains of both Rse and Axl. A deletion variant of Gas6 containing just the G domain region was shown to activate Rse phosphorylation. These results provide evidence that G domains can act as signaling molecules by activating transmembrane receptor tyrosine kinases. Furthermore, they provide a structural link between the activation of cell adhesion related receptors and the control of cell growth and differentiation by the G domain-containing superfamily of proteins.
Specific signals that control the growth and differentiation of cells in developing and adult tissues often exert their effects by binding to and activating cell surface receptors containing an intrinsic tyrosine kinase activity. We recently reported the human and murine complementary DNA sequences of a receptor tyrosine kinase we termed Rse (Mark et al., 1994). A complementary DNA sequence encoding a protein identical to human Rse was reported independently and called Sky (Ohashi et al., 1994). Versions similar or identical to murine Rse have also been reported (Tyro3, Lai and Lemke, 1994;Brt, Fujimoto and Yamamoto, 1994;Etk2, Biesecker et al., 1995).
Rse has about 31% sequence identity with the receptor tyrosine kinases Axl (O'Bryan et al., 1991;Janssen et al., 1991) and c-Mer (Graham et al., 1994). The extracellular domains of the Rse/Axl/c-Mer family are composed of two immunoglobulin-like repeats followed by two fibronectin type III repeats. Together, these proteins define a class of receptor tyrosine kinases whose extracellular domains resemble neural cell adhesion molecules (reviewed by Rutishauser (1993) and Brummendorf and Rathjen (1993)).
While Rse mRNA is expressed preferentially in the adult brain, it is expressed at lower levels in a number of tissues including kidney, ovary, and testis and in a variety of hematopoietic cell lines (Mark et al., 1994;Lai and Lemke, 1994). Axl and c-Mer are also widely expressed, but the highest levels of Axl and c-Mer mRNA are detected in the heart and skeletal muscle (Graham et al., 1995) and testis, ovary, prostate, lung, and kidney (Graham et al., 1994), respectively.
Gas6 was initially identified as a product of a gene whose expression is increased in fibroblasts upon growth arrest (Manfioletti et al., 1993). We recently identified Gas6 as a ligand for human Rse (Godowski et al., 1995). Varnum et al. (1995) identified Gas6 as a ligand for human Axl. Gas6 has 46% amino acid identity to Protein S (PS), 1 an abundant serum protein and a negative regulator of the coagulation cascade. Stitt et al. (1995) reported that PS, but not Gas6, was a ligand for Rse. However, those conclusions were based on the analysis of the interactions of human PS and bovine Gas6 with murine Rse. While human PS does indeed bind to murine Rse, we found that even high concentrations of human PS failed to activate human Rse (Godowski et al., 1995). These results have recently been confirmed by Ohsahi et al. (1995). Thus, there are no published data indicating that PS is a physiologically relevant ligand for Rse.
Gas6 contains 678 amino acids, which may be divided into five domains (Fig. 1A). The Gla domain is rich in ␥-carboxyglutamic acid (Gla) residues. The corresponding Gla domain of PS mediates its Ca 2ϩ -dependent binding to negatively charged phospholipids (Dahlback et al., 1986;Hammond et al., 1987). The loop region, containing thrombin-sensitive cleavage sites in PS, but lacking such sites in Gas6, is followed by four epidermal growth factor (EGF)-like repeats. EGF-like repeats are found in a number of proteins which participate in diverse functions such as coagulation and fibrinolysis, cell adhesion, cell growth, and differentiation. These repeats are sufficient to bind to other receptor tyrosine kinases, and are believed to participate directly in protein-protein interactions (Cambell and Bork, 1993). The COOH-terminal regions of Gas6 and PS are homologous to the steroid hormone-binding globulin (SHBG) protein (Gershagen et al., 1987;Hammond et al., 1987) and contain tandem "globular" or G domains (G1 and G2; Joseph and Baker (1994)). G domains, first identified in laminin A chain, are present in a superfamily of proteins that include basement membrane proteins such as laminin A chain, agrin, merosin, and perlecan, as well as the Drosophila regulatory proteins Crumbs, Fat, and Slit (reviewed by Patthy and Nikolics (1993)).
As a first step in understanding the contributions of the Gla, EGF-like repeats and G domains of Gas6 to interaction with Rse and Axl, we analyzed the relative contributions of these domains in receptor binding and activation. Our data demonstrate the G domains of Gas6 are sufficient for receptor activation. These observations have implications for the mechanism by which other G domain-containing proteins may influence intercellular signaling.

EXPERIMENTAL PROCEDURES
Construction and Expression of Rse, Axl, Protein S, and Gas6 Derivatives-Human PS and Gas6 cDNAs were obtained by PCR using 1 g of human fetal liver or fetal brain cDNA (Clontech) as template with Pfu DNA polymerase (Stratagene) as described (Mark et al., 1994). The coding sequences of the cDNA clones used for the expression of human PS and Gas6 were identical to those available in GenBank (accession numbers Y00692 and L13720, respectively). Plasmids pRKN.hPS and pRKN.hGas6 were used to express human PS and Gas6, respectively, in stably transfected human fetal kidney 293 cells. Transfected cells were washed free of serum-containing media 4 h after transfection, and refed with media containing 2 g/ml vitamin K. Expression was verified by metabolic labeling of cultures with [ 35 S]Cys and [ 35 S]Met (Mark et al., 1992) and/or by Western blotting with a polyclonal anti-PS antiserum (Sigma) or a polyclonal anti-Gas6 antibody. 2 NH 2 -terminally tagged versions of Gas6 or PS were constructed by linking the coding sequences for the gD signal sequence and epitope tag (Mark et al., 1994) by PCR to coding sequences immediately before the first amino acid of mature Gas6 (gD.Gas6; forward primer 5Ј-AGCTGCTCGAGGCGCTGTTGC-CGGCGC) or PS (forward primer 5Ј-AGCTGCTCGAGGCAAATTCTT-TACTTGAA), or amino acids 118 (gD.Gas6 118-C ; forward primer 5Ј-AGCTGCTCGAGGACCAGTGCACGCCCAACC) and 279 (gD.Gas6 279-C ; forward primer 5Ј-GCTGCTCGAGGACATCTTGCCGTGCGTG) of Gas6. The reverse primer for gD.Gas6 and gD.Gas6 118-C was 5Ј-CATG-GATCCTACCGGAAGTCAAACTCAGCTA. The reverse primers for gD.Gas6 279-C and gD.PS were 5Ј-GTCGGATCCGACAGAGACTGAG-AAGCC and 5Ј-CATTCATTTATGTCAAATTCA, respectively. Gas6.gD was constructed by fusing the coding sequences of Gas6 to the COOHterminal gD tag used for Rse.gD through an NheI site, which was added by PCR using the primers 5Ј-ATGGAGATCAAGGTCTG and 5Ј-CATCTTGAGGCTAGCGGCTGCGGCGGGCTCCAC. The cDNA encoding P N G C .gD was constructed in a two-step PCR reaction (Mark et al., 1994) to join the coding sequences of amino acids 1-283 of PS to SHBG homology regions (amino acids 279 -678) of Gas6.gD. The G N P C .gD expression vector was constucted in a similar fashion by fusing coding sequences for amino acids 1-278 of Gas6 to those encoding amino acids 284 -676 of PS.gD. The fusion junctions of P N G C .gD and G N P C .gD are: coding strand 5Ј-GAAGAGTTGT/GAGGACATCT and 5Ј-GGACAC-CTGT/GAGGTTGTTT, respectively, where the shill indicates the fusion junction. Rse-Fc (Godowski et al., 1995) and Axl-Fc 3 were purified as described (Godowski et al., 1995).
Protein Quantification-The concentration of epitope tagged Gas6 and variants in samples was determined by a competitive ELISA using anti-gD monoclonal antibody (5B6; Genentech, Inc., South San Francisco, CA) as a coat/capture antibody. Samples putatively containing epitope-tagged proteins were added to 5B6-coated and bovine serum albumin-blocked polystyrene immunoplates (Nunc, Roskilde, Denmark). Following a 2-h incubation at room temperature, biotinconjugated gD06 (a 6-amino acid peptide derived from the gD sequence; Genentech, Inc.) was added to each well and allowed to compete with the sample for the 5B6 coat for an additional 50 min. After washing the plate 10 times, the wells were incubated with peroxidase-conjugated streptavidin (Zymed Laboratories, Inc., South San Francisco, CA) for 30 min at room temperature. Plates were again washed and TMB Microwell peroxidase substrate (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD) was added to each well. Color was allowed to develop for 10 min, and the reaction was quenched with 100 l of H 3 PO 4 . Absorbance was measured at 450 nm using a Vmax kinetic microplate reader (Molecular Devices). Concentrations of gD-tagged Gas6 or variants were calculated from a standard curve using a purified standard (gDIIIBrgp120) in a range of 1.56 -100 nM.
Phosphorylation Assays-CHO.Rse.gD cells and NIH 3T3.hRse cells, and methods to detect receptor phosphorylation using the ELISA-based KIRA assay and by Western analysis using anti-phosphotyrosine antibodies have been described (Godowski et al., 1995).
Binding Assays-For binding to Rse-Fc or Axl-Fc fusion proteins, conditioned media containing 5-10 nM Gas6 or variant protein were precleared with pansorbin (Calbiochem) for 30 min at room temperature, then incubated with 5 g of the Rse-Fc fusion protein for 4 h at 4°C. Fusion proteins were immunoprecipitated with 20 l of Pansorbin, and the complexes were collected by centrifugation at 14,000 ϫ g for 1 min and then washed three times with PBS containing 0.1% Triton X-100. Precipitates were analyzed by SDS-PAGE under reducing conditions. Western blots of the SDS-PAGE gels were probed with antibody 5B6 as described (Mark et al., 1994).
Protein interaction analysis using BIAcore™ instruments were performed on research grade BIAcore CM5 sensor chips. Running buffer was PBS (10 mM sodium phosphate, pH 7.4, 150 mM sodium chloride) with 0.05% Tween 20. The sensor chip was activated by injection of 20 l of 1:1 mixture of N-ethyl-NЈ-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide at 5 l/min flow rate. 20 l of Rse-Fc at 20 g/ml in 10 mM sodium acetate, pH 5.0, was injected over the sensor chip, followed by 30 l of ethanolamine. A total of 3340 response units of Rse-Fc was immobilized onto the sensor chip, corresponding to a density of 3.34 ng/mm 2 . Conditioned media containing tagged Gas6 or deletion variants expressed in 293 cells were concentrated in Amicon 10 Centriprep spin concentrators, and the buffer was changed to PBS with 0.05% Tween 20 using Pharmacia PD10 Sephadex columns. The concentrations of tagged Gas6 or deletion variants were measured by ELISA assays. In BIAcore protein interaction assays, 30 l of conditioned media containing tagged Gas6 or deletion variants was injected onto the Rse-Fc sensor chip at a flow rate of 10 l/min by the Kinject method. Proteins were allowed to dissociate for 20 min in the flow of PBS with 0.05% Tween 20. The sensor chip was regenerated by a short pulse of 2 l of 10 mM HCl, followed by 2 l of 10 mM NaOH before the next sample was injected. Sensorgrams were analyzed with BIAevaluation 2.1 software from Pharmacia Biosensor AB. Apparent dissociation rate constants (k d ) and association rate constants (k a ) were obtained by evaluating the sensorgram with A ϩ B ϭ AB type I fitting. Equilibrium dissociation constant K d was calculated as k d /k a .

Gas6/Protein S Chimeric Proteins Indicate the G Domains
Determine Receptor Specificity-While human Gas6 binds to and activates human Rse, the structurally related protein human PS does not. We constructed and expressed chimeric versions of PS and Gas6 to determine if we could correlate the presence of specific domains of Gas6 with the ability to bind to and activate Rse. These proteins also contain an epitope tag (gD) that allows for quantitative comparison of their properties. P N G C .gD contains amino acids 1-283 of PS which encode the Gla domain, loop region, and EGF-like repeats fused to amino acids 279 -678 of Gas6 encoding the tandem G domains of Gas6 (Fig. 1A). The complementary fusion protein G N P C .gD contains amino acids 1-278 of Gas6 and 284 -676 of PS, and contains the Gla, loop, and EGF-like repeats of Gas6 and the G domains of PS. The fusion junctions of these proteins correspond to intron/exon boundaries dividing the domains containing the EGF-like repeats with the region containing the tandem G domains. The epitope tagged versions of Gas6, PS, P N G C , or G N P C were expressed in stably transfected human embryonic kidney 293 cells and serum-free conditioned media collected from these cells were characterized by ELISA and Western blotting (Fig. 1B) as a means to monitor the size and expression levels of the proteins. Proteins of the expected size were observed, and the expression levels determined by ELISA correlated with the relative expression levels observed by Western blotting.
We compared the ability of tagged Gas6, PS, P N G C , or G N P C to bind to a soluble form of the extracellular domain of Rse termed Rse-Fc. Conditioned media from human 293 cells expressing the variants were incubated with Rse-Fc or, as a control, just the Fc portion of IgG. Proteins that bound to the Fc fusion proteins were recovered from the supernatant with protein A and tagged proteins were revealed by Western blotting of the resolved precipitates. As observed previously (Godowski et al., 1995), Gas6.gD bound to Rse-Fc. The epitope tag at the carboxyl terminus did not appear to influence receptor binding because Gas6.gD bound to Rse-Fc as efficiently as a version that contains an amino-terminal epitope tag (Fig. 3) or authentic, untagged Gas6 (data not shown). The binding was specific because Gas6.gD did not bind to an irrelevant Fc protein, and PS.gD did not bind to Rse-Fc. Interestingly, P N G C .gD, containing the Gla domain and EGF-like repeats of PS and the G domains of Gas6, bound to Rse-Fc. The complementary fusion protein G N P C .gD did not bind to Rse-Fc. Thus, the ability to bind to the extracellular domain of Rse correlated with the presence of the G domains of Gas6.
While tagged Gas6 and P N G C were capable of binding to the extracellular domain of Rse in vitro, it was important to compare their ability to activate Rse expressed on the surface of cells. We compared the ability of tagged Gas6, PS, G N P C , and P N G C to induce phosphorylation of Rse expressed in NIH 3T3 cells. Conditioned media from mock-transfected 293 cells or from cells expressing the variants were added to serum-starved NIH 3T3 cells expressing human Rse. Consistent with the results obtained in the binding assay, both Gas6.gD and P N G C .gD induced phosphorylation of Rse while PS.gD and G N P C .gD did not (Fig. 2).
The G Domain-Containing Region of Gas6 Is Sufficient to Bind Rse-The results of the experiments with chimeric variants of Gas6 and PS suggested that the G domains of Gas6 determined the specificity for binding to and activating Rse. We next determined if the G domain region of Gas6 was sufficient for receptor binding and activation. We expressed a series of amino-terminal deletion derivatives of Gas6 (Fig. 3A). Gas6 118-C .gD is truncated for the Gla domain and loop region but contains the EGF-like repeats and G domains. Gas6 279-C .gD contains the region of Gas6 that comprises the tandem G domains. These variants also contain a portion of gD at their amino terminus, which provided an epitope tag as well as a signal sequence for secretion. Proteins of the ex-  Fig. 1 legend. In contrast to PS, the Gas6 derivatives were bound by Rse-Fc and Axl-Fc. The binding was specific to the extracellular domains of Rse and Axl because the epitope-tagged proteins were not precipitated by control human Fc (Fc). The input lanes represent 20% of the material used for binding. pected size were observed by Western blotting of conditioned media from cells expressing the variants. Full-length Gas6 as well as both deletion variants were bound by Rse-Fc but not by control Fc fusion protein (Fig. 3B). Human Gas6 has also been shown to bind to and activate the Rse-related receptor Axl (Varnum et al., 1995). We analyzed the ability Gas6 deletion derivatives to interact with the extracellular domain of Axl. As observed with Rse-Fc, Axl-Fc also efficiently bound both Gas6 118-C and Gas6 279-C (Fig. 3B).
These results show that the G domains of Gas6 are sufficient to bind to the extracellular domains of both Rse and Axl and that the Gla, loop, and EGF-like repeats are not absolutely required for this interaction. We then attempted to express the individual G1 and G2 domains of Gas6 in 293 cells. However, we could not detect significant levels of these variants in conditioned medium of transfected cells (data not shown).
Analysis of the Binding Affinities of Gas6 Deletion Derivatives for Rse-To characterize better the binding of Gas6 and deletion variants to Rse, we compared the kinetics of their interaction with Rse-Fc that had been immobilized on a BIAcore biosensor chip. Representative sensorgrams obtained from such an experiment are shown in Fig. 4. When purified Gas6 was injected over the Rse-Fc chip, we observed rapid association and slow dissociation. From the sensorgrams, we determined the on-and off-rates and the dissociation constant for the binding of Gas6 to Rse-Fc (Table I). The K d for interaction of Gas6 with Rse-Fc (4.2 nM) was similar to that reported for the binding of Gas6 to the extracellular domain of Axl (Varnum et al., 1995). The sensorgrams and the calculated on-and off-rates of epitope-tagged Gas6 and the deletion derivatives for Rse-Fc were very similar to those observed using purified Gas6.
Activation of Rse by Gas6 Deletion Derivatives-We compared the ability of the deletion variants to induce tyrosine phosphorylation of human Rse expressed in NIH 3T3 cells. Treatment of NIH 3T3.hRse cells with either gD.Gas6 118-C or gD.Gas6 279-C resulted in phosphorylation of Rse (Fig. 5). The dose response of activation of Rse by the variants was deter-mined in a quantitative KIRA assay (Fig. 6). Using this assay, the EC 50 values for activation of Rse by epitope-tagged Gas6, Gas6 118-C , and Gas6 279-C were 12, 11, and 26 nM, respectively.

DISCUSSION
Gas6 is secreted ligand with structural homology to members of a superfamily of basement membrane proteins implicated in the growth and differentiation of many cells. A series of Gas6 variants were expressed to begin to define the relative roles of the Gla domain, EGF-like repeats, and G domains in receptor binding and activation. Our deletion studies show that the Gla and EGF-like repeats are not absolutely required for receptor binding or activation, and that the G domain region is sufficient for these activities. The on-rates and off-rates for binding of either full-length Gas6, Gas6 118-C , and Gas6 279-C to the extracellular domain of Rse were similar. This maps the receptor binding domain of Gas6 to the G domain-containing region. Furthermore, it shows that the Gla and EGF repeats contribute little to receptor binding.
We also quantitated the effects of deletion of the Gla and EGF-like repeats of Gas6 on activation of Rse phosphorylation. In the KIRA assay, we observed that the EC 50 for activation of Rse by Gas6 279-C was reduced approximately 2-fold compared to intact Gas6. Thus, while the Gla and EGF repeats are not absolutely required for receptor activation, they may contribute directly or indirectly to this process. Furthermore, either the Gla or EGF-like domains may modulate Gas6 activity in vivo. For example, Gas6 binds to cell membranes in a Ca 2ϩ -dependent fashion. 4 Ca 2ϩ -dependent binding to membranes and phospholipids is characteristic of Gla domain-containing ␥-carboxylated proteins. The Gla domain of Gas6 may function to promote the establishment of a local concentration gradient of the ligand. Alternatively, the Gla domain might allow Gas6 to be concentrated on the surface of receptor-bearing target cells.
The deletion analysis of Gas6 localized the receptor binding 4 R. G. Hammonds and H. Raab, personal communication.

TABLE I Kinetic constants for binding of Gas6 and deletion variants to Rse-Fc
On-rates and off-rates were determined as detailed under "Experimental Procedures." Data were evaluated using BIAevaluation 2.1 software (Pharmacia Biosensor).
domain to a region that contains tandem G domains. First characterized as five repeating globular domains of approximately 180 residues found at the COOH terminus of laminin A chain (Patthy, 1991), G domains are found in a superfamily of basement membrane proteins as well as in integral membrane proteins such as Drosophila Crumbs, Fat, and Slit. The individual G domains of Gas6 are most closely related to those found in PS and SHBG and are more distantly related to those found in agrin and laminin. SHBG, which binds to both estrogens and androgens with high affinity, has been reported to regulate intracellular signaling in prostatic stromal cells by binding to cell surface receptors (Hryb et al., 1985;Hryb et al., 1989;Nakhla et al., 1994), although the specific receptors involved have not yet been identified. Interestingly, steroids are reported to regulate the ability of SHBG to bind to and activate its receptor. Steroid-free SHBG is required for binding to the receptor, but activation of the receptor requires subsequent binding of SHBG by a specific steroid ligand. Based on sequence similarity, and our observation that the SHBG-like region of Gas6 is sufficient to bind Rse and Axl, we speculate that SHBG will interact with a Rse-related molecule.
A number of G domain-containing proteins have been shown to play a role in regulating cellular growth and development. Crumbs, Fat, and Slit are involved in ectodermal differentiation processes such as neurogenesis and epithelial polarization (Mahoney et al., 1991;Rothberg et al., 1990;Tepass et al., 1990). The ability of laminin to promote epithelial cell polarization and neurite outgrowth are dependent upon the G domains. Similarly, the G domains of perlecan have been suggested to mediate the neurite outgrowth activity of perlecanlaminin complexes (Noonan et al., 1991). Aggregation of acetylcholine receptors on skeletal muscle fibers is induced by the binding of the G domains of agrin and laminin to dystroglycan-␣, a component of the dystrophin receptor (Gee et al., 1993;Campanelli et al., 1994;Gee et al., 1994;).
Our results demonstrate that the G domains of Gas6 are necessary and sufficient for a functional interaction with both Rse and Axl and provide direct evidence that G domains can activate receptor tyrosine kinases. Furthermore, they suggest that other G domain-containing proteins may exert their effects by binding to cell adhesion molecule-related proteins and subsequently activating intracellular signaling pathways. FIG. 5. Activation of Rse phosphorylation by Gas6 deletion derivatives. Conditioned media from mock-transfected cells, or containing 40 nM Gas6.gD, gD.Gas6, gD.Gas6 118-C , or gD.Gas6 279-C were added to serum-starved NIH 3T3.hRse cells at 37°C for 5 min. SDS-PAGE, immunoprecipitation, and Western blotting analysis are described in the Fig. 2 legend. Molecular sizes are indicated on the left (in kilodaltons).
FIG. 6. Dose response of activation of Rse by Gas6 and deletion derivatives. The amount of phosphorylation of Rse in response to the indicated concentrations of either Gas6, Gas6 118-C , or Gas6 279-C was determined using the KIRA assay. F, gD.Gas6; f, gD.Gas6 118-C ; å, gD.Gas6 279-C .