Inhibition or Activation of Apert Syndrome FGFR2 (S252W) Signaling by Specific Glycosaminoglycans*

Most Apert syndrome patients harbor a single amino acid mutation (S252W) in fibroblast growth factor (FGF) receptor 2 (FGFR2), which leads to abnormal FGF/FGFR2 signaling. Here we show that specific combinations of FGFs and glycosaminoglycans activate both alternative splice forms of the mutant but not of the wild-type FGF receptors. More importantly, 2-O- and N-sulfated heparan sulfate, prepared by a combined chemical and enzymatic synthesis, antagonized the over-activated FGFR2b (S252W) to basal levels at nanomolar concentrations. These studies demonstrated that specific glycosaminoglycans could be useful in treating ligand-dependent FGFR signaling-related diseases, such as Apert syndrome and cancer.

Apert syndrome is one of the most severe types of human craniosynostosis syndromes. The vast majority of Apert syndrome patients have one of two missense mutations in adjacent amino acids, S252W or P253R, of fibroblast growth factor receptor 2 (FGFR2) 2 (1,2), occasionally patients have an Alu-element insertion in the FGFR2 gene (3). The S252W mutation, found in two-thirds of Apert patients, has been associated with the most severe craniofacial anomalies and dermatological disorders (4,5). The P253R mutation is associated with more severe fusion of the bones in fingers and toes (4,5). Both crystallographic and biochemical studies of FGFR2 (S252W) and (P253R) mutants with various FGFs indicate that the mutations, in the highly conserved linker region between the immunoglobulin-like domains II and III, resulted in increased affinity and altered specificity of FGF ligand binding (6 -9). A mouse model supports the idea that the Apert S252W mutation in FGFR2 renders the mesenchymally expressed mutant FGFR2c abnormally susceptible to activation by mesenchymally expressed FGF7-like proteins and the epithelially expressed mutant FGFR2b abnormally susceptible to activation by epithelially expressed FGFs. Thus, these mutations circumvent the normal reciprocal epithelial-mesenchymal signaling restrictions (10).
Glycosaminoglycans (GAGs) are required for FGF/FGFR signaling at both the cellular and whole animal levels (11)(12)(13)(14)(15). Animals defective in GAG biosynthesis have a complete loss of FGF/FGFR signaling (16 -18). Two major types of GAGs are heparan sulfate (HS) and chondroitin sulfate (CS). Both HS and CS are highly sulfated linear GAG chains abundantly expressed on the cell surface and in the extracellular matrix in the form of HS, CS, or HS/CS hybrid proteoglycans. FGF⅐FGFR⅐GAG interactions depend to a large extent on specific GAG sequences, which can activate, inhibit, or have no effect on FGF/FGFR signaling (19 -23).
Apert FGFR2 (S252W) signaling cannot be explained solely by the additional interaction between the mutant receptor and FGF2 observed in the crystal structure (7). For example, FGF7 promotes mutant FGFR2 (S252W) signaling, but the binding affinity between FGF7 and mutant FGFR2c is negligible (8,24), possibly because specific GAGs mediate this interaction. The severity of the phenotype varies in Apert patients, although they carry the same FGFR2 (S252W) mutation. Furthermore, only certain tissues and organs are affected during development and in the adult. This may be due to ligand availability and to the modification of activity by specific GAG sequences.
GAG sequences are not directly encoded by genes but are assembled in the Golgi by enzymes encoded by over 40 genes. Because of the vast expression repertoire of the GAG assembly enzymes, GAGs may have a sulfation pattern, chain length, and fine structure unique to each cell (25). Cell surface GAGs turn over within one-eighth to one-third of a cell cycle (26). This means their quantity and fine structures are able to rapidly change in response to a variety of environmental factors.
Heparin and heparin oligosaccharides have been used in most studies of the effect of GAGs on FGF/FGFR signaling. Heparin, the most heavily sulfated HS, is made by mast cells and is usually stored in mast cell granules. The only obvious defect in heparin-deficient mice involves connective tissue-type mast cells (27,28). In contrast, mice deficient in HS and CS show defects in FGF signaling during gastrulation (15,17,18). Therefore heparin is not biologically relevant, but HS and CS are (22,23,29). Kidney HS has unique 3-O-sulfated sequences that are different from that of 3-O-sulfated heparin (30).
Altered cartilage and bone development are two of the primary phenotypes associated with the pathogenesis of Apert syndrome (10). Cartilage is enriched in extracellular matrix proteoglycans and contains both HS and CS. CS-A is the most abundant GAG in the growth plate and joints. 3 CS-B, also called dermatan sulfate, is enriched in skin and wound fluid (23). CS-E, a highly sulfated CS, interacts with chemokines and growth factors (31). CS-E inhibits the herpes simplex virus infection better than heparin in a cell-based model system (32).
Because of the essential role of GAGs in regulating FGF activity, we hypothesized that not only HS, but also the abundant CS-A, could differentially modulate the activity of wild-type and mutant FGFR2 (S252W). Because GAG regulates FGF/FGFR signaling in the extracellular environment, inhibitory GAGs that down-regulate FGF/FGFR signaling could be potentially useful drugs for treating Apert syndrome, especially the unique dermatological disorders (hyperhidrosis, hypopigmentation, and severe widespread acne) found in these patients. Based on the above observations, we predict that different GAGs can either activate or inhibit specific FGF/FGFR signaling. Here we test this in a cell-based model of Apert syndrome. We show that FGFs and GAGs are equally important partners in FGFR signaling in general and in Apert FGFR2 signaling in particular. Our results suggest that in the presence of multiple FGFs and FGFRs, cell-and tissue-specific GAGs select special FGF/FGFR pairs and regulate the degree of signaling.
FGF/FGFR-mediated Proliferation Assay-Wild-type and mutant FGFR activity was assayed in BaF3 cells. BaF3 cells are a lymphoid cell line that are dependent on the cytokine IL-3 for growth and have no intrinsic response to FGF. When stably transfected to express FGFRs, BaF3 cells will survive in IL-3-free medium only if FGF and heparin are both present. BaF3 cells die if only FGF or GAG is included in IL-3-free cell culture medium. Because BaF3 cells express no endogenous FGF, the background activity in the absence of added FGF is very low regardless of whether the cells express an FGFR. BaF3 cells make low levels of GAGs (ϳ74 ng/10 6 cells, ϳ20% HS (ϳ14 ng/10 6 cells), and ϳ80% CS (ϳ60 ng/10 6 cells)). 4 In contrast, most mammalian cells make 50 -500 ng/10 6 HS (33). Fetal bovine serum used in the cell culture also contains an appreciable amount of HS and CS. However, GAGs from either source have very low mitogenic activity. Therefore, BaF3 cells are a reliable system for detecting FGF-and GAG-dependent FGFR activities (34). The BaF3 cells ectopically expressing wild-type or FGFR2b (S252W) or -2c (S252W) have been previously described (8). The BaF3-FGFR cells were maintained in RPMI 1640 medium (Fisher) supplemented with 10% fetal bovine serum, 1 ng/ml IL-3 (PeproTech Inc., Rocky Hill, NJ), 2 mM L-glutamine, penicillin (50 IU/ml), streptomycin (50 g/ml), 50 M ␤-mercaptoethanol, and 400 g/ml active G418.
For mitogenic assays, BaF3 FGFR cells were washed three times with RPMI 1640 medium to remove IL-3 and were resuspended in assay medium without IL-3 with 5% fetal bovine serum. 30,000 live cells were plated per well in a 96-well plate in media containing GAGs, FGFs (PeproTech), or both in a total volume of 200 l. The cells were then incubated at 37°C for 40 h. Four to five hours before harvesting, an additional 50 l of assay medium containing 1 Ci of [ 3 H]thymidine (Amersham Biosciences) was added to each well. Cells were harvested by filtration through glass fiber paper (PerkinElmer). The incorporation of [ 3 H]thymidine into DNA was determined by scintillation counting.
GAG Quantification-GAGs were quantified by a reversed-phase HPLC separation of fluorescent isoindole derivatives of the glucosamine (GlcN) and galactosamine (GalN) released by acid hydrolysis (35). We found that 1 g/ml of heparin equals 800 nM GlcN residue. Assum-ing the commercial heparin has an average molecular mass of 16,000 Da and the repeating disaccharide has a molecular mass of 600 Da, the average chain length should be 16,000/600 ϭ 27 disaccharides, which is equivalent to 27 GlcN residues. Thus, 1 g/ml of heparin equals 800 nM/27 ϭ 30 nM heparin. Because our chemically and enzymatically synthesized HSs were originally derived from the same heparin, 27 GlcN/chain were used to calculate the HS molar concentration. However, the average chain lengths of HS from bovine kidney, CS-A, -B, and -E are unknown. Their concentrations were either expressed as g/ml in Figs

The FGFR2 (S252W) Mutation Altered Specificity for Both FGFs and
GAGs-Apert mutant and wild-type FGFR2b-and FGFR2c-expressing cells were used to assay GAG-mediated FGFR signaling (8). Wild-type or mutant FGFR2b-expressing BaF3 cells ( Fig. 1) were treated with 2 nM FGF2, -7, or -9, and varying concentrations of GAGs (heparin, HS, CS-A, CS-B, or CS-E from 0 to 2 g/ml). In the absence of GAGs, FGF2, -7, and -9 were unable to activate either wild-type or mutant FGFR2b (see the zero concentration point in each plot, Fig. 1). Consistent with previous studies, wild-type FGFR2b was only activated by FGF7 (8). Relative to wild-type FGFR2b, FGFR2b (S252W) was more susceptible to activation induced by 0.125 g/ml of heparin and FGF7 (90% of the maximum response for the mutant versus 75% for wild-type, Fig. 1, B and E), consistent with increased affinity of FGF7 and mutant FGFR2b (S252W) (8). CS-B (2 g/ml) and FGF7 (Fig. 1B) also showed a detectable mitogenic activity on wild-type FGFR2b, which is consistent with previous reports (22,23). However, FGFR2b (S252W) preferentially responded to distinct GAGs in the presence of the three FGFs ( Fig. 1, D-F). Cell proliferation was stimulated by FGF9 plus heparin, FGF2 plus heparin or CS-B, and FGF7 plus heparin, CS-B, or HS.
Although the same FGFR2b-expressing BaF3 cells were used in the experiments shown in Fig. 1 (top, wild-type; bottom, mutant), the background [ 3 H]thymidine incorporation in the absence of added GAGs depended on the FGF ligand. FGF2 generated a higher background compared with that of FGF7 and FGF9.
Similar experiments with BaF3 cell lines expressing FGFR2c showed changes in GAG specificity in the presence of the wild-type receptor (Fig. 2, top) and the S252W mutant receptor (Fig. 2, bottom). Wild-type FGFR2c/FGF2 ( Fig. 2A) was most efficiently activated by heparin and CS-B, whereas FGFR2c (S252W)/FGF2 was activated by all five GAGs tested (Fig. 2D). In contrast, mutant (Fig. 2E), but not wild-type (Fig. 2B), FGFR2c/FGF7 was activated by heparin. Wild-type FGFR2c/FGF9 (Fig.  2C) was maximally activated by heparin and showed a modest response to CS-B (2 g/ml). In contrast FGFR2c (S252W)/FGF9 (Fig. 2F) was activated by heparin, CS-B, and HS from bovine kidney. CS-E (2 g/ml) also showed a detectable mitogenic effect on the mutant-expressing cell line (Fig. 2F). FGFR2c (S252W) required lower doses of GAGs for activation or was less selective with respect to GAGs than wild-type FGFR2c.
The same mutant FGFR2c-expressing BaF3 cells were used in the experiments shown in Fig. 2, D-F. However, the background [ 3 H]thymidine incorporation in the absence of added GAGs was high (FGF2, 32,000 cpm; FGF9, 22,000 cpm; and FGF7, 10,000 cpm). These results suggest that the proliferation was stimulated by intrinsic BaF3 and/or fetal bovine serum GAGs or resulted from FGF-dependent, but GAGindependent, activity. When BaF3 cells treated with both chondroitinase ABC and heparin lyases were assayed in AIM-V medium (which does not contain fetal bovine serum), the background was reduced by ϳ50% (data not shown). We suspect that other sugar structures, such as undigested GAG linkage oligosaccharides on proteoglycan core proteins, GM1 ganglioside (37), or other N-linked or O-linked sulfated sugars also contribute to the background mitogenic activity.   Figs. 1 and 2 show that FGFR2b (S252W) and FGFR2c (S252W) were activated by a wider spectrum of GAGs than wild-type FGFR2. This parallels previous discoveries that the S252W mutation uncouples ligand binding specificity of both splice forms of the S252W receptor (8). To investigate the role of FGFs versus GAGs in FGFR2 (S252W) activation, we compared the FGF and heparin dose response by either fixing the heparin concentration and varying the FGF concentration or vice versa (Fig. 3). FGFR2b (S252W)-expressing cells were not activated by FGF or heparin alone (see the zero concentration point in each plot). Both FGF and heparin exhibited dose-dependent activities. These results indicate that both FGFs and GAGs are required to activate FGFR2 (S252W).

Both Heparin and FGFs Are Necessary to Activate FGFR2b (S252W)expressing BaF3 Cells-The data in
Unique FGF and GAG Combinations Are Required for the Mitogenic Activity of Mutant FGFR2b (S252W)-Maximum FGFR signaling in the presence of 2 nM FGFs was induced by 0.5 g/ml heparin (Figs. 1 and 2).
The highest GAG concentration used was 2 g/ml. However, the physiological concentrations of GAGs are generally higher than those used in Figs. 1 and 2 and in most in vitro studies. For example, GAG concentrations in wound fluid ranged from 15 to 65 g/ml (29). CS concentrations in growth plates varied from 10 to 14 mg/ml. 3 The BaF3 proliferation assay was therefore used to examine the effect of more physiologically relevant GAG concentrations on FGF activity. GlcN or GalN equivalent concentrations of GAGs were 0, 0.1, 0.5, 2.5, 12.5, and 62.5 M. The highest heparin concentration was increased from 2 g/m1 (Figs. 1 and 2) to 78 g/ml.
Heparin activated every FGF (1, 2, 4, 7, 9, or 10) at low concentrations in FGFR2b (S252W)-expressing BaF3 cells (Fig. 4), but the dose-response curve differed from that of other GAGs (HS, CS-A, -B, and -E). FGF1, the universal FGF that activates all of the FGFRs, was activated by all the GAGs tested (at least at the highest concentrations). In contrast,  (19 -21, 38, 39). Data in Fig. 4 suggest that HS and CS-E can have inhibitory activity toward FGFR2b (S252W) in the presence of FGF2 and FGF7, respectively. However, both HS and CS-E are mixtures of GAG produced by different kinds of cells (in bovine kidney or squid cartilage). We therefore decided to search for structurally defined polysaccharide inhibitors. By a combined chemical and enzymatic approach, we synthesized 2SNS HS and 6S2SNS HS (36). Synthetic 2SNS HS inhibited FGFR2b (S252W) activity in a dose-dependant manner (Fig. 5). The first bar in both panels represents activation by intrinsic GAGs or other cofactors in combination with FGF2 (top) or FGF7 (bottom). When 2 nM FGF2 and 15 nM heparin were used in the assay (Fig. 5, top), 60 nM 2SNS HS reduced the [ 3 H]thymidine incorporation by 18%, whereas 120 nM 2SNS HS showed nearly complete inhibition (98%) of FGFR2b (S252W) activity. This is comparable to the [ 3 H]thymidine incorporation data in the absence of FGF2 (Fig. 5, heparin 15 nM), indicating that 2SNS HS not only inhibited the activity of heparin but also of other intrinsic GAGs or other cofactor activities that gave the higher [ 3 H]thymidine incorporation in the absence of heparin (Fig. 5, FGF2 2 nM). 6S2SNS HS had no inhibitory activity, even though it was made by 6-O-sulfation of 2SNS HS. In fact, 6S2SNS HS stimulated [ 3 H]thymidine incorporation at both 60 and 120 nM concentrations (Fig. 5, top).
Sugar Specificity Involved in FGF/FGFR Signaling-Of all the GAGs we tested, heparin possesses the structural features that initiate maximum BaF3 cell proliferation (Figs. 1-4). But biologically irrelevant sulfated polysaccharides, such as dextran sulfate and fucoidan, work better than heparin in initiating specific FGFR signaling (40). Usually octaoligosaccharides and larger heparin oligosaccharides are required for FGF/FGFR signaling, but disaccharides or trisaccharides have been reported to have activity in FGF/FGFR signaling (41,42). Cell membrane ganglioside GM1, another type of sugar, was found to facilitate FGF2/FGFR1c-mediated cell proliferation (37). Without detailed knowledge of assay systems, especially the specific FGF or FGFR involved in the referenced studies, the contribution of sugar specificity is difficult to define. At physiological concentrations of GAGs, each FGF showed a unique receptor activation capability (Fig. 4). Bovine kidney HS promoted mutant FGFR2b/FGF1 signaling (Fig. 4, FGF1) but did not activate FGFR2b (S252W)/FGF2 signaling (Fig. 4, FGF2). CS-A and CS-B could attain the activation level of heparin, but HS could only attain 60% of the activation level of heparin even at the highest concentrations tested (Fig. 4, FGF7). These observations revealed that specific GAG, FGF, and FGFR combinations determine the level of ligand-dependant FGF signaling and suggest that cell and tissue specific GAG structures and their local concentrations are important regulators of FGF/FGFR activity.
Our Data and the Current FGF⅐FGFR⅐GAG Models-The discovery that HS forms a ternary complex with FGF and FGFR (12,13,43) led to numerous models of the molecular mechanism by which a GAG enables FGF and FGFR to assemble into an active dimeric complex (44). Two models have dominated the field because two FGF/FGFR/GAG co-crystal structures were solved. One model proposed a symmetric structure consisting of 2 FGF2:2 FGFR1c:2 heparin non-reducing ends. The co-crystal was obtained by diffusing heparin decasaccharides into the binary FGF2/FGFR1c crystal (45). The other model proposed an asymmetric structure composed of 2 FGF1:2 FGFR2c:1 heparin (46). This co-crystal was directly formed from a solution containing FGF1, FGFR2c, and heparin decasaccharide. The difference between the two co-crystals was that in the first crystal, FGF2 and FGFR1c interact and form a binary crystal prior to GAG addition, whereas in the second co-crystal, GAG facilitates FGF1 and FGFR2c interaction because FGF1 and FGFR2c do not interact with each other (47). In the co-crystal structures, proline (253), in the highly conserved linker region of FGFR,  FGF (1, 2, 4, 7, 9, or 10)  adopted a trans-conformation in the symmetric two ends co-crystal and a cis-conformation in the asymmetric co-crystal.
Much experimental data supports the symmetric "two ends" model in which "the heparin-binding sites of FGF and FGFR merge to form a basic canyon that recruits two HS for binding. Within this canyon, the HS molecules primarily act to orchestrate and fortify multivalent and cooperative protein-protein contacts within the dimer that are the foundations of dimerization" (48). The high affinity interaction between FGF and FGFR forms the basis of the two ends model. In this model, the role of GAG is secondary, so even a disaccharide (octasulfated sucrose) or HS trisaccharide can activate FGF/FGFR-mediated mitogenic activities (41,42). The co-crystal structure of Apert syndrome FGFR2 (S252W)/FGF2 was consistent with the FGFR1c/FGF2 co-crystal structure on which the two ends model was based (7).
Because of the increased FGF/FGFR2 (S252W) affinity, the two ends model would predict a lower specificity for GAG structures. The observation that the mutant FGFR can be activated not only by heparin, but also by CS-A, -B, -E, and HS, is consistent with the two ends model. However, the model cannot explain all of the data. The observation that mutant FGFR2c was activated by FGF7 and heparin (Fig. 2E) is consistent with previous observations (8) but is inconsistent with the very low affinity between FGF7 and the mutant FGFR2c (S252W) (24). In fact, the high degree of GAG dependence is more consistent with the "the asymmetric model," where despite the low affinity between FGF1 and FGFR2c, high mitogenic activity was observed (47). Secondly, FGF7 has lower affinity (K d 1.19 ϫ 10 Ϫ6 M) for the mutant FGFR2b than does FGF9 (K d 7.62 ϫ 10 Ϫ7 M) (24). Based on affinity alone, FGF9 should be more accommodating of different GAG structures as activators. However FGF7 was activated more efficiently by heparin, CS-A, -B, and HS than FGF9 (Fig. 1, E and F, and Fig. 4, FGF7 and FGF9). Thirdly, FGF7 had lower affinity (K d 1.04 ϫ 10 Ϫ5 M) for the wild-type FGFR2b than did FGF4 (K d 5.34 ϫ 10 Ϫ7 M) (24). However, FGF7 activated wild-type FGFR2b (Fig. 1B), but FGF4 did not activate FGFR2b (49). These results suggest that high affinity FGF/FGFR interactions are not a prerequisite for forming an active FGF⅐FGFR⅐GAG signaling complex.
Alternate Explanation for Activation of FGFR2 (S252W) Signaling in Apert Syndrome-The amino acid residues in the linker region (RSPHRP) are highly conserved in almost all FGFRs. Apert mutations yield new linker peptide sequences of RWPHRP and RSRHRP. The Apert peptides could enhance interactions with GAGs because of a stretch of basic and aromatic amino acid residues. The interaction might induce conformation changes in FGFR and affect the cis-or trans-conformation of the FGFR observed in two distinct FGF⅐FGFR⅐GAG co-crystals (45,46). However, FGF, but not GAG, interacts with the FGFR linker region in both FGF⅐FGFR⅐GAG co-crystal structures. This suggests a model in which FGF⅐FGFR⅐GAG ternary complex formation parallels the heparin/thrombin/antithrombin anticoagulant system (50). Heparin interacts with both thrombin and antithrombin. This interaction causes conformational changes and accelerates thrombin⅐antithrombin complex formation by 1000-fold. However, once the binary thrombin and antithrombin complex is formed, heparin dissociates from the complex (50). In our model, the Apert mutation may accelerate the GAG and FGFR2 interaction because such an interaction has been proposed to be the rate-limiting step in forming a functional FGF⅐FGFR⅐GAG signaling complex (51). This interaction may cause a conformation change in FGFR2 and allow a functional FGFR2⅐FGF⅐GAG ternary complex to form. However, experiments are needed to test our hypothesis.
Molecular Basis for the Inhibitory Effect of the 2-O-and N-Sulfated HS-One study indicated that certain cellular HS interact with FGFR1 in a cation-dependent manor. Such an interaction may inhibit FGFR1/ FGF interactions (52). Based on this study, 2SNS HS inhibition might be through interaction with mutant FGFR2b.
We and others (44,53) previously showed that although HS does not interact with FGF or FGFR alone, it forms a FGF⅐FGFR⅐HS ternary complex and has mitogenic activity. Although 2SNS HS has a lower affinity for both FGF and FGFR, we cannot eliminate the possibility that a ternary complex incapable of FGFR dimerization was formed in the presence of 2SNS HS.
Further studies are needed to test each possible inhibitory mechanism imposed by 2SNS HS. Understanding such inhibitory mechanisms is critical for rational drug design. Developing a GAG Drug That Regulates FGF/FGFR Signaling-The FGF complex represents an excellent target for chemotherapy of Apert syndrome, cancers, and other related diseases because of hyper-or hypo-FGF signaling. FGF⅐FGFR⅐GAG interactions are highly specific and regulated, so an agent could be administered with a lower risk of side effects. Additionally, the site of action of a GAG inhibitor is outside the cell, which removes the difficulty of crossing a biological membrane and reduces the likelihood of drug resistance. However, even with a good model of FGF signal transduction, the rational design of GAG or other inhibitors is a process fraught with difficulties. N-Desulfated, 2-Odesulfated, 6-O-desulfated heparin, and other chemically modified heparins have been tested in different FGF/FGFR signaling assays as signaling promoters/inhibitors or as an antiangiogenesis drugs (19 -21, 38, 60). However, because of different chemical desulfation levels, these modified heparins can either promote or inhibit the mitogenic activity of the same FGF/FGFR pair, e.g. FGF2/FGFR1C, in biological assays (19,39). We showed that 2-O-and N-sulfated HS derived from heparin by a combined chemical and enzymatic synthesis approach inhibits FGFR2b (S252W) signaling. The degree of sulfation imposed by sulfotransferase is moderate, which gives specificity and avoids over-sulfation-induced nonspecific interactions with other proteins.

2-O-and N-Sulfated HS Resembles Physiologically Relevant HS
Conclusion-The regulation of FGF/FGFR signaling by GAGs is an important but poorly studied area. The inhibition or activation of FGFR (S252W) by specific combinations of FGFs and GAGs reflects the complexity of FGF/FGFR signaling. Pursuing structurally defined GAGs will be the key to understanding this unique signaling pathway. Such GAGs should be very useful in treating human diseases caused by ligand-dependent FGFR signaling malfunctions.