Heparin interacts with the adhesion GPCR GPR56, reduces receptor shedding, and promotes cell adhesion and motility

G protein-coupled receptors (GPCRs) are medically important molecules, not only because they are the cause of many pathological disorders but because they are very druggable targets (Drews, 2000; Ma and Zemmel, 2002). One GPCR of great interest is the adhesion-class GPCR (adhesion GPCR) GPR56 (also known as ADGRG1) (Hamann et al., 2015; Liu et al., 1999). GPR56 was first identified in melanoma cells, where its gene expression level correlates inversely to the metastatic potential (Zendman et al., 1999). Nevertheless, it is best known as the receptor molecule solely responsible for bilateral frontoparietal polymicrogyria (BFPP), a rare human autosomal recessive disorder characterized by unique brain cortical malformations (Piao et al., 2004). More recent studies have underlined additional roles of GPR56 in male gonad development, myoblast fusion, muscle hypertrophy, oligodendrocyte development, and hematopoietic stem cell generation and maintenance (Ackerman et al., 2015; Chen et al., 2010; Giera et al., 2015; Saito et al., 2013; Solaimani Kartalaei et al., 2015; White et al., 2014; Wu et al., 2013). In the human immune system, GPR56 is expressed in a restricted manner in cytotoxic lymphocyte subsets, and its overexpression suppresses natural killer cell chemotaxis (Della Chiesa et al., 2010; Peng et al., 2011).

Unlike other GPCRs, adhesion GPCRs are characterized by an extended extracellular domain (ECD) upstream of the seven transmembrane (7TM) segment (Hamann et al., 2015; Stacey et al., 2000; Yona et al., 2008). Within the ECD, various cell-adhesion protein motifs are usually found in the N-terminal half, followed by a signature GPCR autoproteolysis-inducing (GAIN) domain that contains a consensus GPCR proteolysis site (GPS) motif (Araç et al., 2012). GPS-mediated autoproteolysis cleaves the receptor into an N-terminal ECD fragment (NTF) and a C-terminal 7TM-fragment (CTF), which then form a non-covalent complex on the cell surface. Hence, adhesion GPCRs exert their functions mostly by binding to cellular ligand(s) through the NTF and signaling through the CTF (Hamann et al., 2015; Krasnoperov et al., 2002; Lin et al., 2004).

Indeed, recent deorphanization studies of GPR56 have identified its many specific ligands. In melanoma progression, GPR56 inhibits tumor growth and metastasis by binding to and promoting the internalization and degradation of tissue transglutaminase 2 (TG2) (Xu et al., 2006; Yang et al., 2014). In brain development, the GPR56–collagen-III interaction is crucial for the integrity of the pial basement membrane and proper cerebral cortex lamination (Luo et al., 2011). In a previous report, we have described a new cellular protein ligand that interacts with the NTF of wild-type GPR56 but not with BFPP-associated point mutants (Chiang et al., 2011). In addition to the aforementioned ligands, which presumably interact with GPR56 in trans, examples of cis-acting binding partners have also been noted. As such, GPR56 has been shown to associate with the tetraspanins CD9 and CD81 (Little et al., 2004). Likewise, it was revealed that α₃β₁ integrin might function synergistically with GPR56 to fine-tune cerebral cortex development (Jeong et al., 2013). Taken together, the cellular functions of GPR56 are mediated, in part, by interacting with specific protein ligands (in trans and/or in cis) on the cell surface and/or extracellular matrix (ECM).

Alternatively, adhesion GPCRs can function through receptor shedding. Indeed, ectodomain shedding has been noted for several adhesion GPCRs, including CD97, brain angiogenesis inhibitor 1 (BAI1) and GPR124 (also known as ADGRA2) (Kaur et al., 2005; Vallon and Essler, 2006; Wang et al., 2005). Shedding is likely to generate distinct soluble receptor fragments that have different cellular functions from the membrane-bound proteins. Previously, we and others have detected the presence of soluble GPR56 (sGPR56; the NTF that has been shed from the membrane) in transfected cells, suggesting constitutive GPR56 shedding (Chiang et al., 2011; Jin et al., 2007; Yang et al., 2015). More intriguingly, expression of a truncated GPR56 receptor that lacked the NTF was shown to signal constitutively (Paavola et al., 2011). Similarly, deletion of the TG2-binding Ser–Thr–Pro-rich (STP) segment within the GPR56 NTF also enhances melanoma tumor growth and angiogenesis significantly (Yang et al., 2011). A more recent study has shown that GPR56 is activated by a cryptic tethered agonist peptide that is located at the most N-terminal end of the CTF and exposed following the dissociation of the NTF (Stoveken et al., 2015). Finally, binding of collagen III to GPR56 promotes the release of the GPR56 NTF from the CTF, which then relocates to the lipid raft microdomain for RhoA activation (Luo et al., 2014; Schoneberg et al., 2015). Taken together, the GPR56 NTF might act as a repressor of the constitutively active CTF, whose signaling activity is turned on by the exposure of a tethered agonist upon removal of the NTF (Liebscher et al., 2015; Stoveken et al., 2015). Thus, the signaling activity of GPR56 might be modulated through the interaction of its cellular binding partners with the NTF, or alternatively by shedding of the NTF.

In addition to interacting with protein ligands, adhesion GPCRs have also been shown to interact with non-protein molecules, such as polysaccharides, and phospholipids and glycolipids. These include phosphatidylserine (PS) and lipopolysaccharide (LPS) for BAI1, as well as glycosaminoglycans (GAGs) for CD97, EMR2 (also known as ADGRE2) and GPR124 (Das et al., 2011; Hochreiter-Hufford et al., 2013; Stacey et al., 2003; Vallon and Essler, 2006; Wang et al., 2005). Heparin and the closely related heparan sulfate, and chondroitin sulfate (also known as dermatan sulfate) are the most abundant GAGs attached to proteoglycans. They are defined by the composition of their amino sugars and the different degrees of modification on the sugars, such as epimerization and sulfation, which create structural complexities directly linked to functional variability (Bernfield et al., 1999; Carlsson et al., 2008; Esko et al., 2009). Herein, we report the identification of heparin as a GAG interacting partner of GPR56 and characterize the functional consequence of the GPR56–heparin interaction.

In addition to protein ligands, certain adhesion GPCRs such as CD97, EMR2 and GPR124 also interact with GAGs, the polysaccharide components of proteoglycans (Stacey et al., 2003; Vallon and Essler, 2006; Wang et al., 2005). In an attempt to define the new cellular ligand of GPR56 described in our previous report (Chiang et al., 2011), we examined and concluded that collagen III is unlikely to be the putative cellular ligand (Fig. S1). We hence decided to test the possibility of a GPR56–GAG interaction. We employed a modified enzyme-linked immunosorbent assay (ELISA) using plate-bound GAGs and detected strong binding of a GPR56 and mouse fragment crystallizable fusion protein (GPR56–mFc) to heparin, as well as a much weaker interaction with heparan sulfate and chondroitin sulfate-B (Fig. 1A). In contrast, binding of GPR56–mFc to chondroitin sulfate-A, chondroitin sulfate-C and hyaluronic acid was negligible. As expected, the control mFc protein showed no binding to any of the GAGs tested. Importantly, the GPR56–heparin interaction was inhibited in a dose-dependent manner by exogenously added heparin, thus confirming the specificity of the reaction (Fig. 1A,B). Likewise, an alternative heparin–agarose pulldown assay showed that only GPR56–mFc, but not mFc or EMR2–mFc protein, was captured and precipitated by the heparin–agarose beads (Fig. 1C). As EMR2 is a similar highly glycosylated adhesion GPCR with a unique binding affinity to chondroitin sulfate, these results further verify that the GPR56–heparin interaction is specific and not just the result of the ‘sticky’ nature of adhesion GPCRs. When tested at the same molar concentrations, GPR56 showed stronger binding than IL-8 – a well-known heparin-binding protein – to heparin, suggesting a potential higher heparin-binding reactivity of GPR56 (Fig. 1D). These results clearly indicate that heparin is a specific GAG binding partner of GPR56.

To characterize the GPR56–heparin interaction further, a GPR56–heparin binding assay was performed in the presence or absence of EDTA or EGTA with or without extra divalent cations (Ca²⁺ or Mg²⁺). As shown in Fig. 2A, the GPR56–heparin interaction was impaired by the addition of EDTA or EGTA. On the contrary, the interaction was greatly enhanced by the addition of Ca²⁺ or Mg²⁺. Likewise, GPR56–mFc protein that had been pulled down by heparin–agarose beads was efficiently uncoupled from the agarose beads by EDTA- or EGTA-containing buffer (Fig. 2B) (data not shown). We thus conclude that the GPR56–heparin interaction is dependent on divalent cations.

As GPR56 deficiency is a direct cause of BFPP, we next examined the heparin-binding characteristics of selected BFPP-associated GPR56 point mutants (Piao et al., 2004). In addition, the heparin-binding ability of human versus mouse GPR56 proteins was also compared because the two proteins have been shown previously to have different TG2-binding properties (Chiang et al., 2011). Interestingly, both solid-phase heparin-binding and heparin–agarose pulldown assays showed that the relevant BFPP mutants (GPR56-R38W, GPR56-Y88C and GPR56-C91S) displayed a much stronger heparin-binding ability than the wild-type protein (Fig. 2C,D). This result indicates that heparin (or heparan sulfate) is not the new cellular ligand of GPR56 that our studies have previously alluded to, and to which the BFPP mutants are unable to bind (Chiang et al., 2011). By contrast, mouse GPR56 showed a specific but weaker interaction with heparin in comparison to human GPR56 (Fig. 2C,E). This is true even when the binding was tested in an alternative ELISA-based assay whereby biotinylated heparin was incubated with plate-bound GPR56–mFc fusion proteins (Fig. 2F). We conclude that the heparin-binding activity is conserved in both human and mouse GPR56 proteins. In addition, the heparin-binding activity of GPR56 can be greatly modulated by single amino acid changes, as attested by the stronger heparin-binding BFPP-causing point mutants.

The specificity and affinity of protein–GAG interactions vary widely, and generally involve the electrostatic interaction of basic amino acids with the negatively charged GAGs (Hileman et al., 1998). Thus, the heparin-binding sites usually contain clusters of basic amino acids in a linear sequence and/or in specific conformational ensembles. To date, two well-known linear heparin-binding motifs, namely XBBXB and XBBBXXBX (B represents a basic amino acid, and X is any neutral or hydrophobic residue), are identified (Cardin and Weintraub, 1989). A survey of the human GPR56 NTF found a reverse XBBXB consensus sequence (R²⁶GHREDFR³³, basic residues in bold), and several XBBX clusters (Fig. 3A,B) (data not shown).

To identify and verify the putative heparin-binding region(s), a series of truncated GPR56–mFc fusion proteins were engineered, and their relative heparin-binding activity was determined (Fig. 3A; Fig. S2). We found that the majority of the C-terminal truncation proteins still retained the heparin-binding activity, suggesting the presence of a dominant heparin-binding sequence at the most N-terminal region. Indeed, although heparin-binding was not observed in a mutant comprising the first 25 amino acids of [GPR56(1–25)–mFc], a weak binding signal was readily detected in GPR56(1–30)–mFc. Most interestingly, a much stronger heparin-binding signal was noted in GPR56(1–35)–mFc protein and other larger C-terminally truncated proteins. As the first 25 amino acid residues constitute the signal peptide of GPR56, the GPR56(1–25)–mFc protein was essentially the same as the negative control mFc-only protein. These results support the notion that the region between residues 26 and 35 of GPR56 constitutes a heparin-binding site (Fig. 3B,C).

To verify the role of the R²⁶GHREDFRFC³⁵ motif in heparin-binding, two GPR56(1–35)–mFc mutants, one with residue Arg33 mutated to Ala [GPR56(1–35/R33A)] point mutation and the other with three mutations of His28, Arg29 and Arg33 to Ala, were generated [GPR56(1–35/H28/R29/R33A)] (Fig. 3B; Fig. S3A–C). Interestingly, although the R33A mutation slightly reduced the heparin-binding activity of GPR56(1–35)–mFc, the GPR56(1–35/H28/R29/R33A) triple mutant almost completely abrogated heparin binding (Fig. 3C). These results confirmed that the R²⁶GHREDFR³³ motif is indeed the N-terminal heparin-binding domain of GPR56 with a significant contribution from His28, Arg29 and Arg33 residues.

Apart from the R²⁶GHREDFR³³ motif, additional heparin-binding regions were also present as positive heparin-binding activities were noted to similar extents in various N-terminally truncated GPR56–mFc fusion proteins (Figs S2 and S3D,E). Several potential heparin-binding XBBX clusters were noted in the C-terminal half of the GPR56 NTF and, hence, were investigated further. Among them, the L¹⁹⁰KHPQKASRRP²⁰⁰ region with two XBBX sequences was of interest because mutation of the basic residues to Ala greatly reduced the heparin-binding activity of the GPR56(1–340)–mFc protein, which contains most of the NTF, except the GPS motif (Fig. 3A,B,D; Figs S2, S3D,E). Combined mutation of the basic residues in the R²⁶GHREDFR³³ and the L¹⁹⁰KHPQKASRRP²⁰⁰ regions showed a synergistically diminished heparin-binding activity (Fig. 3B,D; Fig. S3D,E). Interestingly, similar abilities to bind to collagen III were detected in the wild-type GPR56(1–340)–mFc protein and the various mutant probes tested, indicating that these basic residues are not involved in the GPR56–collagen-III interaction (Fig. 3D). Taken together, we conclude the R²⁶GHREDFR³³ and L¹⁹⁰KHPQKASRRP²⁰⁰ clusters constitute the two major heparin-binding regions of GPR56.

As an important component of proteoglycans and ECM molecules, GAGs interact with a plethora of soluble cell surface and matrix proteins to modulate their functions either positively or negatively (Funderburgh, 2000; Rabenstein, 2002; Trowbridge and Gallo, 2002). In fact, the two known matricellular ligands of GPR56, collagen III and TG2, have both been reported to interact with heparin (Lortat-Jacob et al., 2012; San Antonio et al., 1994). More interestingly, a functional collagen-III-binding region has been identified within the region between residues 27 and 160 of GPR56, which overlaps with the heparin-binding R²⁶GHREDFR³³ motif (Luo et al., 2012). By contrast, the TG2-binding STP segment has been located in the region between positions 108 and 177 (Yang et al., 2011), which is distinct from but proximal to another heparin-binding L¹⁹⁰KHPQKASRRP²⁰⁰ region. The partial overlapping and close proximity of the heparin-binding motifs to the collagen III- and TG2-binding regions strongly suggests that heparin is likely to enhance or hinder the interaction of GPR56 and its protein ligands.

Hence, we examined the potential outcome of the ternary GPR56–heparin–protein-ligand interaction. As shown in Fig. 4A, pre-incubation of collagen III with heparin reduced the GPR56–heparin interaction significantly, but no such inhibitory effect was observed when collagen III was added simultaneously with GPR56–mFc. Similar results were obtained using agarose pulldown assays with both the heparin–agarose and collagen-III–agarose beads (Fig. S3F, lane 4). Interestingly, pre-incubation of GPR56 and heparin greatly enhanced the amount of GPR56 pulled down by collagen-III–agarose, but pre-incubation of GPR56 and collagen III did not further increase the GPR56 pulldown efficiency of heparin–agarose (Fig. S3F, lane 3). Hence, it seems that the strength and outcome of the ternary GPR56–heparin–collagen-III interaction might depend on the binding sequence among these three molecules. By contrast, no apparent difference was found in the GPR56–TG2 far-western binding assay when either probed sequentially with heparin followed by GPR56–mFc or by heparin and GPR56–mFc together at the same time (Fig. 4B).

Surprisingly, we found that the interaction of GPR56 with the new cellular ligand that had been previously described by us could be modulated bilaterally with GAGs (Fig. 4C,D). As such, it was noted that heparin inhibited, whereas heparan sulfate and chondroitin sulfate-B enhanced, GPR56 binding to the cellular ligand if GAGs and the GPR56–mFc probe were added simultaneously (Fig. 4C). If cells were allowed to incubate with GAGs first, the GPR56–cellular-ligand interaction was enhanced greatly by heparin but only slightly by heparan sulfate and chondroitin sulfate-B (Fig. 4D). Interestingly, other GAGs, such as chondroitin sulfate-A, chondroitin sulfate-C and hyaluronic acid, had no effects on the GPR56–cellular-ligand binding in both conditions. These results imply that GPR56 mediates the interaction with heparin and the novel cellular ligand through similar or overlapping binding region(s). Furthermore, as with collagen III and TG2, the new cellular ligand itself might be able to bind to specific GAGs as well. Taken together, these data suggest that heparin is able to modulate the interaction of GPR56 with its protein ligands selectively and vice versa.

GPR56 has been shown previously to regulate cell migration, but the precise mechanism is not fully understood. A well-established role of heparin (or heparan sulfate) is to interact with various growth factors, chemokines and ECM proteins (Rabenstein, 2002; Salmivirta et al., 1996). This action generally leads to phenotypic changes in cellular adhesion and motility. We therefore tested the effect of heparin on GPR56-medited cell adhesion and migration. For the cell adhesion assay, GPR56^dim A375 melanoma cells (expressing low levels of surface GPR56) were engineered to overexpress GPR56 and incubated with various forms of plate-bound heparin. As shown in Fig. 5, stronger cell adhesion was detected in GPR56-expressing A375 (A375-GPR56) cells than in the control A375-Neo cells. Interestingly, no significant differences were found in the adhesion of A375-GPR56 cells to low- or high-molecular-mass heparins, or to unfractionated heparins (Fig. 5). For the cell migration assay, highly mobile HT1080 cells stably expressing GPR56 (HT1080-GPR56) were subjected to the Boyden chamber-based chemotaxis assay using fetal calf serum (FCS) as a chemoattractant. HT1080-Neo cells were included as a control. The extent of cell migration was compared without or with heparin added in the upper or lower chamber. In the absence of heparin, both stable cell lines displayed comparable migration abilities toward a range of FCS concentrations (2% and 10%), indicating similar basal chemotactic motilities (Fig. 6A). When heparin was placed in the upper chamber, HT1080-GPR56 cells showed a ∼twofold increase over HT1080-Neo cells in chemotactic migration toward 2% and 10% FCS (Fig. 6A). In contrast, no significant difference in cell migration was observed in both cell lines if heparin was placed in the lower chamber (Fig. 6B). Likewise, a similar enhancing effect of heparin on cell invasion was observed only when heparin was placed in the upper chamber (Fig. 6C). These results suggest that heparin can somehow promote cell chemokinesis (random cell migration) but not chemotaxis (directional cell migration) through binding to GPR56.

Previous studies by Luo et al. have shown that the GPR56–collagen-III interaction inhibits the migration of neural progenitor cells (NPCs) (Luo et al., 2011). It has been further demonstrated recently that binding of collagen III to GPR56 enhances the dissociation (shedding) of the GPR56 NTF (sGPR56) from its CTF, which then relocates to the lipid raft microdomains and induces RhoA activation (Luo et al., 2014). In melanoma cells, GPR56 has been shown to inhibit VEGF production through a protein kinase Cα (PKCα)-mediated signaling pathway (Yang et al., 2011). Furthermore, TG2 binding to GPR56 promotes receptor-mediated internalization and subsequent degradation of TG2, and secreted GPR56 NTF competes and blocks the binding and internalization of TG2 by GPR56 (Yang et al., 2014). To investigate the potential mechanism(s) involved in the heparin-enhanced cell adhesion and motility, the effect of heparin on the cellular distribution of GPR56 receptor subunits, activation of signaling molecules and GPR56 receptor shedding were examined. Unlike collagen III, the interaction of heparin with GPR56 did not promote the relocation of the CTF to the lipid raft microdomains (Fig. 7A), nor the activation of RhoA (Fig. 7B). Likewise, no significant PKCα activation was identified in heparin-treated GPR56-expressing cells (Fig. 7C). Interestingly, however, an inhibitory effect on collagen-III-induced RhoA activation was observed in GPR56-expressing cells that had been either pre-incubated with heparin or incubated simultaneously with heparin and collagen III (Fig. 7D; Fig. S3G). This inhibitory effect was not found when heparin was added after GPR56-expressing cells had been treated with collagen III (Fig. 7D; Fig. S3G). More interestingly, sGPR56 that had been derived from transiently transfected HEK-293T cells was reduced significantly and in a dose-dependent manner in the presence of heparin (Fig. 7E). Moreover, C32 melanoma cells that expressed endogenous GPR56 also shed less sGPR56 when cultured in the presence of heparin (Fig. 7F). Finally, similar data were obtained in stable GPR56-expressing HT1080 cells (Fig. 7G). These results indicate that heparin binding to GPR56 reduces GPR56 receptor shedding without affecting the membrane distribution patterns of the GPR56 NTF and CTF. Moreover, here, no evidence of GPR56-induced signaling by heparin has been found, rather it seems that heparin acts as an inhibitory modulator preventing collagen-III-induced GPR56 activation by hindering the collagen-III–GPR56 interaction. In this sense, heparin could be considered an antagonist of the GPR56 receptor. It remains to be demonstrated whether the heparin–GPR56 interaction induces intracellular signaling.

In addition to the well-known protein ligands collagen III and TG2, we have identified heparin (and heparan sulfate) as the GAG interacting partner of GPR56 (Figs 1 and 2). The GPR56–heparin interaction is dependent on divalent cations, is conserved in human and mouse receptors, and can be specifically blocked by exogenous heparin. Through the use of truncated and mutated GPR56 protein probes, the major heparin-binding regions were mapped to two distinct basic-residue-rich motifs in the GPR56 NTF, which overlap partially with the collagen-III- and TG2-binding sites (Fig. 3). Our finding expands the repertoire of endogenous binding partners of GPR56 and suggests that the ubiquitous heparin (heparan sulfate) GAG might regulate GPR56 receptor function.

Heparin and heparan sulfate are closely related GAGs with variably sulfated repeating disaccharide units. The major disaccharide unit of heparin comprises 2-O-sulfated iduronic acid (IdoA2S) and 6-O-sulfated N-sulfated glucosamine (GlcNS6S), whereas the most common disaccharide repeat of heparan sulfate is a glucuronic acid moiety (GlcA) linked to N-acetylglucosamine (GlcNAc). Further complexity and variability arises when the length, arrangement and degrees of modification of the disaccharide units are considered (Bernfield et al., 1999; Esko et al., 2009). Although heparin is naturally produced and stored in the granules of mast cells and heparan sulfate is ubiquitously present on proteoglycans, it is possible to have ‘heparin-like’ and ‘heparan-sulfate-like’ GAGs on the same proteoglycan (Bernfield et al., 1999; Carlsson et al., 2008). The heparin-like and heparan-sulfate-like GAGs are known to have different affinities for specific proteins, hence variable functional outcomes. Our finding of the stronger GPR56 binding to heparin over that to heparan sulfate might suggest a preferential affinity of the GPR56 protein to certain modified disaccharides of heparin-like GAGs. Most interestingly, GPR56 is shown to bind equally well to heparin species of both low- and high-molecular masses, confirming its unique affinity to the heparin-like GAG structure (Fig. 5). It is worth noting that GPR56 protein does not bind to other GAGs, such as chondroitin sulfate and hyaluronic acid, indicating again that the heparin-binding ability of GPR56 is specific and not due to the ‘sticky’ nature of adhesion GPCRs (Fig. 1). This is especially true when we compare GPR56 and EMR2, another well-known chondroitin-sulfate-binding adhesion GPCR (Fig. 1C).

In fact, various adhesion GPCRs, such as CD97, EMR2 and GPR124, have been shown previously to interact with GAGs (Stacey et al., 2003; Vallon and Essler, 2006). Indeed, binding of chondroitin sulfate-B with CD97 and EMR2 is thought to play a role in the cellular interaction of activated T cells, dendritic cells and macrophages with B cells (Hamann et al., 1999; Stacey et al., 2003). The combinatorial interaction of CD97 with α₅β₁ integrin and chondroitin sulfate synergistically enhances endothelial cell invasion (Wang et al., 2005), whereas the GAG-bound soluble GPR124 promotes endothelial cell survival by interacting with the α_vβ₃ integrin (Vallon and Essler, 2006). Hence, the interaction of GAGs and adhesion GPCRs is prevalent and functionally important.

The identification of the two major heparin-binding R²⁶GHREDFR³³ and L¹⁹⁰KHPQKASRRP²⁰⁰ regions in GPR56 is consistent with the notion that clusters of basic residues, providing the electrostatic interaction, are important for heparin and heparan sulfate binding (Fig. 3). Interestingly, sequence comparison has shown that mouse GPR56 contains a similar basic residue composition (L¹⁹⁰QHPQKAAKRP²⁰⁰) at the homologous region between positions 190 and 200, but lacks the reverse XBBXB consensus sequence (G²⁶SPREDFR³³, basic residues in bold) as found in the most N-terminal sequence of human GPR56. This might explain the weaker heparin-binding activity of the mouse GPR56 protein. Because residual heparin-binding activity was still detected in the GPR56 probe where the seven basic residues [GPR56(1–340/7A)] had been mutated (Fig. 3D), the contribution to heparin binding of other basic residues in the NTF could not be completely ruled out.

It is known that residues proximal to the GAG-binding domains can influence the strength of GAG–protein interactions (Hileman et al., 1998; Lortat-Jacob et al., 2012). The stronger heparin-binding activity detected in BFPP-associated point mutants (R38W, Y88C and C91S) seems to reinforce this notion (Fig. 2). However, earlier molecular analyses have confirmed the tested BFPP point mutants are mostly intracellular misfolded proteins with a loss-of-function in collagen III binding (Luo et al., 2012). Hence, heparin-binding probably does not play a role in the development of BFPP. Rather, the stronger heparin-binding activity of these BFPP mutants is most likely due to conformational misfolding.

Heparin and heparan sulfate are known to greatly modulate the binding and function of many growth factors and cytokines to their cognate receptors (Pellegrini, 2001; Spivak-Kroizman et al., 1994; Yayon et al., 1991). As collagen III and TG2 themselves are also heparin-binding proteins (Lortat-Jacob et al., 2012; San Antonio et al., 1994), the overlapping and close proximity of the heparin-interacting domains with the collagen-III- and TG2-binding sites strongly suggest that heparin might modulate the interaction of GPR56 with its protein ligand(s) (Fig. S4). Indeed, this was subsequently confirmed by the differential effects of heparin on the ternary GPR56–heparin–protein-ligand interaction (Fig. 4A,B; Fig. S3F). It is not apparent why the GPR56–heparin interaction was inhibited by pre-incubation with collagen III and heparin, but not simultaneous addition of collagen III and heparin. However, it is interesting to note that although the heparin-binding and collagen-III-binding regions overlap, mutations of heparin-binding basic residues did not affect the collagen-III-binding activity of GPR56 (Fig. 3D). In sum, the differential ligand-binding effect is likely to be determined by multiple factors, such as the binding site occupancy, affinity and avidity of different individual GPR56–ligand interactions. Finally, the unexpected finding of the differential effects of GAGs on the interaction of GPR56 and the new cellular ligand further signifies that the new cellular protein ligand previously identified by us is distinct from collagen III (Fig. 4C,D). It will be of great interest to reveal its identity in the future.

The selective inhibition of the GPR56–heparin interaction by collagen III is more intriguing when the distinct effects of collagen III and heparin on GPR56 receptor shedding and GPR56-mediated cell migration are considered. As the ligand–receptor pair responsible for BFPP, the collagen-III–GPR56 interaction enhances GPR56 shedding, promotes the relocalization of the CTF to lipid raft microdomains and inhibits NPC migration by recruiting G_α12 and G_α13 proteins and promoting RhoA activation (Luo et al., 2011,, 2014). Furthermore, mice deficient in Col3a1 display neuronal over-migration and a cobblestone-like cortical malformation similar to that seen in BFPP.

By contrast, our results show that the heparin–GPR56 interaction reduces GPR56 shedding, and promotes cell adhesion and migration without apparent activation of known GPR56-mediated signaling molecules, such as RhoA and PKCα (Figs 5–7). The heparin–GPR56 interaction also does not promote CTF relocalization to lipid raft microdomains (Fig. 7A). As a highly expressed GAG molecule in the mammalian central nervous system (CNS), heparan sulfate is known to bind to a number of growth factors and morphogens, and to play a vital role in the embryonic brain development (Ford-Perriss et al., 2003; Inatani et al., 2003). Indeed, CNS-specific heparin-sulfate-deficient mice are known to display many developmental CNS defects owing to selective inhibition in midline axon guidance (Inatani et al., 2003). More recently, GPR56 has been found to be highly expressed in oligodendrocyte precursor cells (OPCs) and to act as an essential regulator of OPC proliferation and maturation to oligodendrocytes, which is important for CNS myelination (Ackerman et al., 2015; Giera et al., 2015). Interestingly, OPCs and oligodendrocytes express abundant levels of proteoglycans that are decorated by heparan sulfate (HSPGs) or chondroitin sulfate (CSPGs), some of which are known to be crucial in modulating OPC development and oligodendrocyte-mediated myelination positively or negatively (Harlow and Macklin, 2014; Kucharova and Stallcup, 2010; Pendleton et al., 2013; Properzi et al., 2008; Stringer et al., 1999; Szuchet et al., 2000). In summary, these findings are highly suggestive of a possible role of heparin and heparan sulfate in GPR56-regulated oligodendrocyte development.

Recent studies have revealed that shedding or removal of the NTF of adhesion GPCRs exposes a hidden Stächel sequence at the N-terminus of the CTF, which then acts as a tethered agonist for receptor activation (Liebscher et al., 2014; Stoveken et al., 2015). Our results show that heparin or heparan sulfate binding to GPR56 inhibits the dissociation of the GPR56 NTF from the CTF without inducing the known signaling activities (Fig. 7B–G). Most importantly, an inhibitory effect on collagen-III-induced RhoA activation was observed after heparin binding to GPR56 (Fig. 7D; Fig. S3G). This interesting feature is reminiscent of a receptor antagonist. Hence, it is possible that heparin and heparan sulfate act as a suppressor (antagonist) of GPR56 receptor function, in part, by inhibiting GPR56 receptor shedding, thereby trapping and masking the tethered Stächel peptide agonist and preventing the activation of the signaling pathways.

Outside the CNS, GPR56 also plays a pivotal role in many distinct biological systems, including male gonad development, myoblast fusion, muscle hypertrophy, hematopoietic stem cell generation and maintenance, tumor growth and metastasis, and cytotoxic immune cells. Heparin and heparan sulfate are present abundantly in many, if not all, of these cells, tissues and organs. Hence, the heparin–GPR56 interaction is expected to have more widespread functional implications. Our results provide the first clue for a possible role of heparin in GPR56-mediated functions in these systems.

Unless otherwise specified, general reagents were obtained from Sigma-Aldrich (Gillingham, Dorset, UK) or BDH-Merck (Poole, Dorset, UK). Oligonucleotide primers were supplied by Tri-I Biotech (Taipei, Taiwan). DNA and protein reagents were obtained from Invitrogen (Carlsbad, CA), Qiagen (Valencia, CA) and Amersham (GE Healthcare). The CG-2 and CG-3 monoclonal antibodies (mAb) against GPR56 have been described previously (5 µg/ml) (Yang et al., 2015). FITC-conjugated goat anti-mouse IgG was from Jackson ImmunoResearch (West Grove, PA). Mouse IgG₁ isotype control (clone 11711, 1:100) was from R&D Systems (Minneapolis, MN). Anti-TG2 mAb (clone CUB 7402, 1:200) was from Thermo Scientific (Fremont, CA). Anti-β-actin mAb (clone C4, 1:2000) was from Chemicon (Temecula, CA). Anti-7TM polyclonal antibodies were generated in rabbits immunized with the GPR56 cytoplasmic peptide NSDSARLPISSGSTSSSR. The anti-CD71 antibody (H68.4, 1:1000) was from Zymed Laboratories (San Francisco, CA), and the anti-caveolin-1 antibody (7C8, 1:1000) was from Upstate (Lake Placid, NY). Heparin (H4784) and GAGs (C3788, C4384, H7630 and H7640) were from Sigma-Aldrich. High-molecular-mass heparin (HEP001) and low-molecular-mass heparin (HO30) were from Amsbio (AMS Biotechnology, Oxon, UK). Human collagen III (ab73160) was from Abcam (Cambridge, UK). Human collagen IV (C5533) was from Sigma-Aldrich. Rat tail collagen I (354236) was from Becton Dickinson (Bedford, MA). Polyclonal PKCα antibody (C-20, 1:200) was from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Heparin–agarose (H6508) and collagen–agarose beads (C0286) were from Sigma-Aldrich. All culture media were from Invitrogen (Carlsbad, CA) and were supplemented with 10% heat-inactivated FCS, 2 mM L-glutamine, 50 IU/ml penicillin and 50 μg/ml streptomycin. All cell lines used in this study were purchased from the American Type Culture Collection (Manassas, VA) and cultured in conditions as suggested. Briefly, HEK-293T and HT1080 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM), whereas C32 melanoma and U87MG cells were cultured in minimum essential medium (MEM) containing non-essential amino acids and 1 mM sodium pyruvate.

The various truncated and mutated GPR56–mFc constructs were generated by PCR using GPR56–mFc cDNA as a template. The primer pairs used are listed in Table S1. The pFB-Neo retroviral construct encoding the full-length GPR56 receptor has been described previously (Hsiao et al., 2014). All expression constructs were sequenced to confirm their sequence fidelity.

The mFc-fusion proteins were purified from conditioned media using protein-G–Sepharose affinity chromatography as described previously (Chiang et al., 2011; Lin et al., 2005; Stacey et al., 2003). Briefly, conditioned media from transiently transfected HEK-293T cells were passed through the protein-G-conjugated Sepharose (nProtein-G Sepharose 4 Fast Flow, GE Healthcare), followed by extensive washes with washing buffer (50 mM Tris-HCl pH 7.4, 10 mM CaCl₂, 150 mM NaCl). The purified mFc-fusion proteins were stored at −80°C until use.

HEK-293T packaging cells in 10-cm dishes were transfected with 3 µg each of the pFB-Neo expression construct, pVPack-VSV-G and pVPack-GP vectors (Stratagene) with 25 µl of Lipofectamine in OPTI-MEM medium as recommended by the supplier. Virus-containing supernatant was harvested, and a final concentration of 5 µg/ml of polybrene solution was added. HT1080 and A375 cells (∼40–50% confluence) in 6-well plates were infected with 1 ml of viral supernatant. The infected cells were selected in medium containing 1 mg/ml G418. G418-resistant cells were collected after ∼2 weeks of selection and confirmed using appropriate analysis.

Protein–collagen-III binding was examined in 96-well tissue culture plates (Corning^® 1×8 Stripwell™), whereas protein–GAG interactions were assessed in 96-well BD™ heparin-binding plates (BD Biosciences) coated with various GAGs. Briefly, collagens (5–30 μg/ml) and GAGs were dissolved in PBS and incubated in wells (100 μl/well) overnight at room temperature. The plate was washed extensively, then blocked with 0.2% gelatin blocking solution for 1 h at 37°C (for GAGs) or 1% BSA blocking solution for 1 h at room temperature (for collagens). Plates were incubated with the indicated protein probes for 2 h at room temperature. The plate was then washed extensively and incubated with mouse anti-human IL-8 mAb (1:250) (PeproTech, Rocky Hill, NJ) for the detection of IL-8, and/or with horseradish peroxidase (HRP)-conjugated goat anti-mouse Fc mAb (1:10,000) (Sigma-Aldrich) for the detection of mFc-fusion proteins in blocking buffer for 1 h at room temperature. The binding was revealed by incubation with tetramethylbenzidine substrate for 20 min and measured at OD₄₅₀.

Conditioned medium of HEK-293T cells that had been transiently transfected with the mFc-fusion-protein expression constructs was collected by centrifugation at 4°C for 10 min. Pulldown assays were performed using protein-G–agarose or heparin–agarose beads. Briefly, agarose beads (30 μl) were washed and then incubated with 500 μl blocking buffer (1% BSA in PBS) for 4 h at 4°C. Conditioned medium (3 ml) containing mFc-fusion protein was then incubated with the agarose beads at 4°C overnight with constant rotation. The beads were then washed extensively and boiled for 5 min in equal volumes of 2× SDS-PAGE sample buffer. The eluate was separated by SDS-PAGE and analyzed by western blotting.

A modified cell-binding assay was performed essentially as described previously (Chiang et al., 2011). In brief, cells were subjected to the standard flow cytometry analysis using GPR56–mFc protein (10 μg/ml) as a probe. Cells were subsequently incubated with FITC-conjugated goat-anti mouse IgG (1:2000 in blocking buffer). Purified mFc protein (10 μg/ml) was used as a negative control. When necessary, protein ligands of GPR56, such as collagen III, were included as indicated. Binding was analyzed using a FACSCalibur Instrument (BD Bioscience). Data were analyzed by FlowJo 7.6.5 (Tree Star Inc., San Carlos, CA).

GAGs (10 μg/ml) were diluted in Hank's balanced salt solution (HBSS) and coated on 96-well BD™ heparin-binding plates (100 μl/well) overnight. The plate was washed with acetate buffer (100 mM NaCl, 50 mM NaOAc, 0.2% Tween-20, pH 7.2) before blocking with 0.2% gelatin blocking solution for 1 h at 37°C. A375 cells in suspension (1×10⁶ cell/ml) were added to each well (100 μl/well) and incubated at 37°C for 30 min. After extensive washes, adherent cells were fixed with 100 μl of 1% glutaraldehyde (diluted in 0.1 M cacodylate buffer) per well for 30 min before staining with 100 μl of 1% Methylene Blue in 0.01 M borate buffer for 30 min. The excess dye was washed out, and 100 μl/well of ethanol was added. The dye dissolved in ethanol was transferred to another 96-well plate for measurement at OD₅₉₅.

Cell chemotaxis and invasion analyses were performed using a QCM Chemotaxis Cell Migration Assay kit and QCM ECMatrix Cell Invasion Assay Fluorimetric kit (Millipore, Bedford, MA), respectively, as described previously (Huang et al., 2012). In short, cells were serum-starved and resuspended at a density of 5×10⁵ cells/ml in serum-free medium. Cells (100 μl/well) were placed in the upper chamber equipped with an 8-µm pore filter membrane. The bottom chambers were filled with medium with or without FCS (2% or 10%). The plates were kept in an incubator at 37°C for 6 h. Cells that had migrated were dissociated from the bottom of filter membrane and stained with CyQuant GR dye. Fluorescence intensity was measured in a fluorescence plate reader (Molecular Devices) with a excitation 480-nm and emission 520-nm filter set.

The GTP-Rho pull-down assay was performed using A375-GPR56 stable cells. Cells at approximately 30% confluence were serum starved for 18 h, followed by the addition of heparin (20 μg/ml) for 30 min. Cells were then lysed in 150 μl of lysis buffer (50 mM Tris pH 7.5, 10 mM MgCl₂, 0.5 M NaCl and 2% Igepal). 400 μg of lysate proteins were incubated with 50 μg of beads conjugated with the GST-tagged Rho-binding domain of rhotekin (GST–RBD) (Cytoskeleton) at 4°C for 60 min. The beads were washed twice with wash buffer (25 mM Tris pH 7.5, 30 mM MgCl₂, 40 mM NaCl). Bound Rho proteins were eluted with Laemmli sample buffer and detected by western blotting using anti-RhoA mAb (Cytoskeleton, ARH03, 1:500).

All procedures were performed on ice. Cells were washed, then lysed in 150 μl of ice-cold TNET buffer (25 mm Tris-HCl, pH 7.5, 150 mm NaCl, 5 mm EDTA, 1 mm EGTA, 1% Triton X-100, 10 mm NaF, 1 mm sodium pyrophosphate, 1 mm Na₃VO₄ and 1× protease inhibitor mixture) for 30 min. Total cell lysates were passed through a 25-gauge needle 20 times on ice, and then centrifuged at 1000 g for 10 min to collect the supernatant. Typically, 200 μl of cell lysate (∼500 μg protein) were mixed with 400 μl of 60% OptiPrep™ gradient medium and then placed at the bottom of a 5 ml polyallomer ultracentrifuge tube (Beckman). Samples were overlaid sequentially with 3400 μl of 30% and 200 μl of 5% ice-cold OptiPrep™ gradient medium diluted in TNET buffer, and subjected to ultracentrifugation (Beckman) at 200,000 g for 16 h at 4°C. Following ultracentrifugation, seven equal fractions (∼600 μl/fraction) were collected from the top of the tube for western blotting analyses.

For preparation of total cell membranes, cells were lysed in isotonic buffer without detergent (20 mM Tris–HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 5 mM EGTA, 20 mM sodium fluoride, 5 mM sodium pyrophosphate, 1 mM sodium vanadate, 1 μM okadaic acid and a cocktail of protease inhibitors). Cell lysates were passed through a 25-gauge needle 20 times on ice and then centrifuged at 1000 g for 10 min at 4°C. The supernatant was further centrifuged at 300,000 g for 1 h at 4°C. The supernatant after the second centrifugation was designated as the cytosolic fraction, whereas the pellet encompassing total membrane proteins was analyzed by western blotting.

sGPR56 was detected using a newly developed sandwich ELISA with minor modifications (Chen et al., 2011; Yang et al., 2015). Briefly, plates were coated with anti-GPR56 CG2 mAb (2 µg/ml) overnight at 4°C, followed by blocking for 1 h. Culture supernatant (100 µl/well) was added for 2-h incubations at 37°C. Following extensive washes, bound sGPR56 was detected using HRP-labeled CG3 mAb (5 µg/ml) and the peroxidase substrate. The absorbance was measured at 450 nm. All samples were analyzed in triplicate.

Quantifications were based on at least three independent experiments. Data were shown as means±s.e.m. Statistical analysis of data was performed using the Student’s t-test using Prism 5 software. P-values less than 0.05 were considered as significant.

We thank Dr Martin Stacey (University of Leeds, UK) for critical reading of the manuscript.