Binding of heparin/heparan sulfate to fibroblast growth factor receptor 4.

Fibroblast growth factors (FGFs) are heparin-binding polypeptides that affect the growth, differentiation, and migration of many cell types. FGFs signal by binding and activating cell surface FGF receptors (FGFRs) with intracellular tyrosine kinase domains. The signaling involves ligand-induced receptor dimerization and autophosphorylation, followed by downstream transfer of the signal. The sulfated glycosaminoglycans heparin and heparan sulfate bind both FGFs and FGFRs and enhance FGF signaling by mediating complex formation between the growth factor and receptor components. Whereas the heparin/heparan sulfate structures involved in FGF binding have been studied in some detail, little information has been available on saccharide structures mediating binding to FGFRs. We have performed structural characterization of heparin/heparan sulfate oligosaccharides with affinity toward FGFR4. The binding of heparin oligosaccharides to FGFR4 increased with increasing fragment length, the minimal binding domains being contained within eight monosaccharide units. The FGFR4-binding saccharide domains contained both 2-O-sulfated iduronic acid and 6-O-sulfated N-sulfoglucosamine residues, as shown by experiments with selectively desulfated heparin, compositional disaccharide analysis, and a novel exoenzyme-based sequence analysis of heparan sulfate oligosaccharides. Structurally distinct heparan sulfate octasaccharides differed in binding to FGFR4. Sequence analysis suggested that the affinity of the interaction depended on the number of 6-O-sulfate groups but not on their precise location.

The fibroblast growth factors (FGFs) 1 belong to a family of about 20 related polypeptides. They display biological activity toward cells of mesenchymal, neuronal, and epithelial origin and are involved in processes such as cell growth, organ development, and angiogenesis (1). The biological effects of FGFs are exerted through interactions with FGF receptors (FGFRs). The receptor family consists of four known members, FGFR1-4, with many isoforms (2). Upon ligand binding the receptor is thought to be activated through dimerization and phosphorylation by the intracellular tyrosine kinase domains (3). Heparan sulfate proteoglycans (HSPGs), abundant components of cell surfaces and the extracellular matrix, appear central to signaling through FGF⅐FGFR complexes (for reviews, see Refs. 4 -6). Cells lacking endogenous HSPGs respond poorly to FGF, whereas the response can be readily restored by addition of exogenous heparin (7,8). Accumulated evidence points to formation of biologically active complexes involving FGF, FGFR, and HSPGs, in which heparan sulfate interacts with both the FGF and FGFR components of the complex (9 -11). A direct interaction between HSPGs and FGFRs appears critical for FGFR activation (12). A heparin-binding domain identified in the second Ig-loop of the four FGFRs comprises sequence of about 20 amino acids toward the NH 2 terminus of the loop (12). Different splice variants of the receptors differ in affinity for heparin, such that the interaction may vary with the structure of the extracellular receptor domain (13).
Heparin/HS chains are initially synthesized as polymers of alternating glucuronic acid (GlcA) and N-acetylglucosamine (GlcNAc) units (for reviews, see Refs. 14 -16). In HS biosynthesis, the polymer is first modified by partial N-deacetylation/N-sulfation of GlcNAc residues. The further modification reactions, C5-epimerization of GlcA to iduronic acid (IdoA) units and O-sulfation at various positions (C2 of IdoA and GlcA, C3 and C6 of GlcN units), all occur in the vicinity of previously incorporated N-sulfate groups. Heparin, a highly specialized product of mast cells, is more extensively modified than HS and the modifications are more evenly distributed along the polymer (14).
The heparin/HS structures required for FGFR binding are poorly defined. However, both IdoA(2-OSO 3 ) and GlcNSO 3 (6-OSO 3 ) residues appear to be required for the FGF2 induced activation of FGFR1, whereas 2-O-sulfate groups alone are sufficient to mediate binding to FGF2 (11,17,18). While these findings suggest a role for 6-O-sulfated GlcNSO 3 residues in the interactions with FGFR1, it would seem likely that different FGFRs may bind structurally distinct HS species. Neuroepithelial HSPG thus preferentially bound FGFR1 at the cell surface although FGFR3 was also present (19). The importance of polysaccharide-FGFR interaction was underpinned by the finding that heparin could alone induce phosphorylation of FGFR4 in the absence of an FGF ligand (20).
In the present paper we describe FGF-independent binding of heparin and HS to the extracellular domain of FGFR4. We show that the interaction is mediated by N-sulfated octasaccharides that contain both IdoA  and GlcNSO 3 (6-OSO 3 ) residues, and provide sequence data for FGFR4-binding HS domains.

EXPERIMENTAL PROCEDURES
Materials-All studies were performed using the soluble extracellular domain of FGFR4. The expression and purification of recombinant human FGFR4, containing the three extracellular Ig domains (Ser 25 -Arg 366 ), were as described earlier (21). Briefly, the His-tagged protein was expressed in Sf9 insect cells and purified directly from the culture medium by nickel and heparin affinity chromatography. Heparin from pig intestinal mucosa (stage 14, Inolex Pharmaceutical Division, Park Forest South, IL), was purified as previously described (22). It was used either unlabeled or radiolabeled by 3 H-acetylation of free amino groups (specific activity 75,300 dpm/nmol) as described (23). The selectively desulfated heparin preparations and oligosaccharides of bovine lung heparin (24,25) were a kind gift from Dr. Dorothe Spillmann (Uppsala University, Uppsala, Sweden). Heparan sulfate preparations from bovine aorta, kidney, lung, and intestine were generously provided by Dr. Keiichi Yoshida (Seikagaku Corp., Tokyo, Japan). N-Sulfated HS oligosaccharides were prepared from bovine intestinal mucosa heparan sulfate (a gift from Kabi AB, Stockholm, Sweden) and 3 H-labeled as described previously (26). Briefly, HS was N-deacetylated by hydrazinolysis followed by treatment with nitrous acid at pH 3.9, resulting in cleavage at the N-unsubstituted GlcN residues. The resistent N-sulfated oligosaccharides were recovered, reduced with NaB 3 H 4 (28 Ci/mmol, Amersham Pharmacia Biotech, Uppsala, Sweden), and separated by gel chromatography. The column materials, Sephadex G-15 and CH-Sepharose-4B, were obtained from Amersham Pharmacia Biotech, as were the PD-10 desalting and Superdex 30 columns. The Partisil-10 HPLC column (4.6 ϫ 250 mm) was from Whatman Inc., Clifton, NJ, and the Propac PA1 HPLC column was from Dionex, Surrey, United Kingdom.
Binding Studies-In the filter-trapping assay (27), radiolabeled glycosaminoglycans were incubated with FGFR4 in 8.1 mM Na 2 HPO 4 , 1.5 mM KH 2 PO 4 , 2.7 mM KCl, and 140 mM NaCl, pH 7.4 (PBS), containing 0.1 mg/ml bovine serum albumin, in a total volume of 200 l for 2 h at room temperature. The mixtures were rapidly passed through nitrocellulose filters (Sartorius, diameter 25 mm, pore size 0.45 m) using a vacuum suction apparatus, followed by washing with PBS. Proteins and protein bound saccharides remain on the filter whereas unbound saccharides pass through. The bound saccharides were released by 2 M NaCl and quantified by a ␤-scintillation counter.
Binding studies were also performed using a CH Sepharose-4B column, with immobilized FGFR4, that was prepared according to the instructions of the manufacturer. For preparation of 1 ml of the affinity matrix, ϳ0.5 mg of FGFR4 was used. Heparin (0.5 mg) was included in the immobilization reaction to protect the heparin-binding site on FGFR4. To avoid immobilization of heparin to the matrix, the heparin preparation used had been treated with HNO 2 at pH 3.9 (28) to destroy any N-unsubstituted GlcN residues, followed by recovery of the high molecular weight fraction by gel chromatography on Superdex 30. Samples of 3 H-labeled heparin/HS were applied to the column in PBS, with or without CaCl 2 supplementation, followed by washing with PBS and elution of the bound material with a linear gradient of NaCl (0.14 -1.0 M) in PBS at a flow rate of 0.5 ml/min. Fractions of 1 ml were collected and measured for radioactivity. The NaCl gradient was monitored by measuring the conductivity of every third fraction. A control column was prepared without immobilized FGFR4. This column did not bind any of the tested glycosaminoglycans (data not shown). To study the interaction of the antithrombin (AT) binding heparin domain with FGFR4, 3 H-labeled heparin decasaccharides were subjected to affinity chromatography on antithrombin-Sepharose as described (29). Bound decasaccharides were eluted with a step gradient of NaCl (0.14, 0.5, and 2.0 M NaCl in 50 mM Tris-HCl, pH 7.4). The high affinity (ϳ2% of total saccharide) and non-binding fractions recovered in the 2.0 and 0.14 M NaCl eluates, respectively, were tested for FGFR4 binding by affinity chromatography.
Surface plasmon resonance analysis on a BiacoreX instrument (Bia-Core AB, Uppsala, Sweden) was a third means of studying saccharide: FGFR4 binding. Heparin (0.5-1 mg) was biotinylated by incubation in 0.1 M MES buffer (pH 5.5 with 50 mM biotin hydrazide (Calbiochem, San Diego, CA) and 10 mM N-ethyl-NЈ(dimethylaminopropyl)carbodiimide (Pierce Chemical Corp., Rockford, IL) for 5-6 h at room temperature. Biotinylated heparin was separated from excess reagent on a PD-10 column and immobilized to streptavidin-coated sensor chips (BiaCore AB). FGFR4 was incubated with saccharides for at least 10 min prior to injection over the heparin-coated surface. The running buffer used was PBS supplemented with 0.005% Tween 20. The concentrations of the interactants and flow rates were as indicated in the figure legends. A streptavidin surface without immobilized heparin was used as a control. The response from this surface was subtracted from the response of the heparin surface.
Compositional Analysis of Heparan Sulfate-N-Sulfated oligosaccharides from bovine intestinal HS were fractionated by binding to the FGFR4 affinity matrix. To deplete the unbound pool (ϳ80% of the total saccharide) of any remaining FGFR4 binding components, it was rechromatographed twice on the FGFR4 column (Ͻ4% of the material was bound to the matrix upon the second rechromatography step). The disaccharide composition of HS samples was determined as described (30,31). Briefly, saccharides were treated with nitrous acid (HNO 2 ) at pH 1.5, leading to deaminative cleavage of the saccharide chain at GlcNSO 3 units (28). The resultant terminal anhydromannose units were radiolabeled by reduction with NaB 3 H 4 (0.25-0.5 mCi/reaction) yielding 3 H-labeled 2,5-anhydromannitol ([ 3 H]aMan R ) residues. The labeled disaccharides were recovered by gel chromatography and further separated by anion-exchange HPLC on a Partisil-10 SAX column eluted with a step gradient of KH 2 PO 4 . The disaccharide peaks were identified by comparing their elution positions to those of standard heparin disaccharides. The proportions of non-O-sulfated disaccharides were determined by high voltage paper electrophoresis of the total labeled disaccharides in 0.83 M pyridine, 0.5 M acetic acid buffer, pH 5.3 (32).
Sequence Analysis-FGFR4 binding HS oligosaccharides, containing a reducing terminal [ 3 H]aMan R residue, were prepared for sequence analysis by anion-exchange HPLC on a Propac PA1 column in H 2 O, pH 3 (adjusted with HCl). The bound oligosaccharides were eluted with a linear gradient of NaCl (up to 1.5 M). The fractions containing the octasaccharides of interest were pooled, desalted, dried in a centrifugal evaporator, and sequenced through a combination of chemical and enzymatic degradation procedures as described (33). Samples were first subjected to partial HNO 2 (pHNO 2 ) cleavage (34) by treatment with 2 mM NaNO 2 in 20 mM HCl on ice. After incubation for various periods of time (30, 60, 90, 120, and 180 min), aliquots were removed and the reaction was stopped by addition of 200 mM sodium acetate, pH 6. The aliquots, containing the cleavage products from the different time points, were combined and subjected to enzyme digestion. The exoenzymes used for sequence analysis were iduronate-2-sulfatase (IdoA2Sase), ␣-L-iduronidase (IdoAase), and glucosamine-6-sulfatase (GlcN6Sase) (Oxford GlycoSciences, Abingdon, U.K.). The enzymes are recombinant human (IdoA2Sase and IdoAase) or caprine (GlcN6Sase) proteins produced in Chinese hamster ovary K1 cells. IdoA2Sase removes ester sulfates at C2 of nonreducing terminal IdoA(2-OSO 3 ) residues of heparin/heparan sulfate whereas IdoAase cleaves the ␣1-4 linkage between IdoA and GlcNR in heparin/heparan sulfate (R is -COCH 3 , -SO 3 Ϫ , or -H 3 ). GlcN6Sase cleaves ester sulfates at C6 of nonreducing GlcNR residues. The substrate specificities of the enzymes were confirmed using known heparin structures as substrates. IdoA2Sase and IdoAase were shown to be active on all potential substrates including heparin disaccharide deamination products. Notably, 6-O-sulfated aMan R is not a substrate for GlcN6Sase. 2 The amounts of enzyme used were 1 milliunit/reaction of IdoA2Sase/IdoAase, and 0.2 milliunits/reaction of GlcN6Sase. The samples were incubated with a single enzyme or combinations of the enzymes in a total volume of 25 l, in the incubation buffer provided with the enzymes (50 mM sodium acetate, pH 5.0, and 0.1 mg/ml bovine serum albumin) at 37°C for 13-15 h. In digestions involving several enzymes, the enzymes were added at 2-h intervals, in the order of action. The enzyme-treated saccharides were analyzed by Propac ion-exchange chromatography as described above. Sequence information was obtained by detecting shifts in the elution positions of the enzyme-treated subfragments (33,34).

RESULTS
Binding of Heparin to the Extracellular Domain of FGFR4 -To assess the binding of the recombinant extracellular domain of FGFR4 to heparin, increasing amounts of [ 3 H]he-parin were incubated with FGFR4 at physiological ionic strength. The protein-polysaccharide complexes formed were trapped on nitrocellulose filters and the bound saccharide was quantified by scintillation counting (27). The results indicated that heparin bound FGFR4 in a dose-dependent and saturable manner ( Fig. 1A), whereas the binding was completely abolished by addition of excess cold heparin (data not shown). A K D value of 0.3-0.4 M was determined for the FGFR-heparin interaction (assuming an average molecular weight of 10 kDa for heparin) by fitting a hyperbolic function using nonlinear regression analysis to the data visualized by a Scatchard plot (inset, Fig. 1A). The interaction was also studied by affinity chromatography of [ 3 H]heparin on immobilized FGFR4, as well as by surface plasmon resonance (BiaCore) measurements of soluble FGFR4 binding to immobilized heparin. In affinity chromatography, [ 3 H]heparin was found to require 0.25-0.50 M NaCl for elution from the FGFR4 column (Fig. 1B). Surface plasmon resonance analysis showed saturable binding of FGFR4 to the heparin-coated surface whereas little or no binding was seen to a control surface without heparin (data not shown). Together, these data indicate that the FGFR4 ectodomain is capable of binding heparin in a FGF-ligand independent fashion, in agreement with previous results (20,21,35).
An important role for divalent cations such as Ca 2ϩ in the binding of heparin to FGFR1, another member of the FGFR family, were proposed by McKeehan and co-workers (36,37). To examine whether calcium ions affect the heparin-FGFR4 interaction, we tested the binding of [ 3 H]heparin to the FGFR4 affinity matrix in the presence of 1.3 mM Ca 2ϩ , which corresponds to the physiological Ca 2ϩ concentration of extracellular fluids. Under these conditions, the peak elution of [ 3 H]heparin occurred at ϳ0.40 M NaCl, as compared with a peak elution position corresponding to ϳ0.35 M NaCl in the absence of Ca 2ϩ ions (Fig.  1B). These results suggest that Ca 2ϩ ions may slightly enhance, but are not required for, the heparin-FGFR4 interaction.
Structural Requirements for Binding of Heparin to FGFR4 -To identify the minimal size of the FGFR4 binding heparin domain, even numbered, 3 H-end-labeled heparin oligosaccharides were incubated with FGFR4 in solution, after which the binding was assessed by the filter-trapping method (see "Experimental Procedures"). Octasaccharides were the shortest oligosaccharides with appreciable FGFR4 binding capacity ( Fig. 2A). The binding of decasaccharides and longer fragments to FGFR4 increased gradually with increasing fragment length, but without any striking differences in binding between the consecutive fragments of the series. Surface plasmon resonance was also used to define the minimal FGFR4binding heparin domain by assessing the ability of heparin oligosaccharides to inhibit binding of FGFR4 to biotinylated full-length heparin, immobilized on the chip surface (Fig. 2B). Octasaccharides were the smallest fragments with substantial inhibitory capacity, whereas decasaccharides and longer fragments had still higher inhibitory effect. Taken together, the above data implicate a minimal FGFR4-binding site within a sequence encompassing eight monosaccharide units.
We next studied the importance of the N-, 2-O-, and 6-Osulfate groups of heparin in FGFR4 binding, by testing the ability of selectively desulfated heparin preparations to inhibit binding of [ 3 H]heparin to FGFR4 in solution (Fig. 3A). Whereas low concentrations (1-5 g/ml) of unlabeled, native heparin

FGFR4-binding Heparan Sulfate Domains
blocked the binding almost completely, corresponding amounts of the various selectively desulfated heparin preparations showed little inhibitory capacity. However, each of the preparations resulted in 50 -75% inhibition at high concentrations (50 -100 g/ml) (Fig. 3A). In Biacore studies (Fig. 3B), FGFR4 was incubated with the desulfated heparin preparations prior to injection of the mixture over the heparin-coated surface. The results were in agreement with the data from filter trapping assays, such that each of the selective desulfation treatments led to a dramatic decrease in inhibitory capacity (Fig. 3B). Collectively, these results suggest that the N-, 2-O-, and 6-Osulfate groups of heparin all contribute to binding FGFR4.
Recently, the AT-binding pentasaccharide motif of heparin, containing a critical 3-O-sulfated GlcNSO 3 residue was implicated in binding to FGFRs (38). We decided to reassess this proposal by separating 3 H-labeled heparin decasaccharides with regard to affinity for AT, and then test the resultant high and low affinity fractions for ability to bind FGFR4. About 2% of the starting material bound with high affinity to immobilized AT, in good agreement with previous findings (29) and this fraction was quantitatively retained by the FGFR4 column (Fig. 4A), in accord with the proposal by McKeehan et al. (38). However, a major portion of the fraction with low affinity for AT also bound to the immobilized FGFR4, and elution of this material from the FGFR4 column required the same NaCl concentration as that needed to displace the decasaccharide with high affinity for AT (Fig. 4A). This finding is in disagreement with the notion that the AT binding sequence is essential for FGFR binding. Indeed, isolation of the two FGFR4-binding fractions followed by analytical AT-Sepharose chromatography confirmed that one of the fractions, as expected, showed high affinity for AT whereas the other did not (Fig. 4B).
Binding of Heparan Sulfate to FGFR4 -The major physiological polysaccharide ligand for FGFR4 is presumably HS rather than heparin, that is essentially confined to the mast cell. The interactions of 3 H-labeled HS samples from bovine lung, aorta, and kidney with FGFR4 were studied by affinity chromatography (Fig. 5). All HS species tested were retained by the column and required 0.2-0.3 M NaCl for elution. Generally, the binding was somewhat weaker than that of heparin, judging from the higher NaCl concentration required to displace heparin compared with HS. The FGFR4 binding profiles of the various HS species differed such that a substantial portion of kidney HS emerged at NaCl concentrations Ͼ0.35 M, whereas aorta HS contained only minor amounts of such highaffinity material.
Disaccharide Composition of FGFR4 Binding Heparan Sulfate Decasaccharides-We next proceeded to characterize HS domains with affinity toward FGFR4. The experiments with selectively desulfated heparin suggested that the binding of HS to FGFR4 would require highly sulfated structures, of the type represented by the N-sulfated domains of the polysaccharide. These domains are composed of consecutive N-sulfated disaccharide units and contain most of the O-sulfate groups of the HS chain. To isolate such domains, we used bovine intestinal HS that bound to the FGFR4 affinity column (data not shown) similar to the lung HS preparation shown in Fig. 5. N-Sulfated oligosaccharides were prepared as described under "Experi-   (Table I, Fig. 6). The 6-O-sulfate groups occurred mainly in trisulfated -IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )disaccharide units. Similar results were obtained upon compositional disaccharide analysis of FGFR4 bound and unbound dodecasaccharides (not shown). Together, these findings suggest that the FGFR4-HS interaction is mediated by highly sulfated domains containing both IdoA(2-OSO 3 ) and Glc-NSO 3 (6-OSO 3 ) residues.
Sequence Analysis of FGFR4-binding Heparan Sulfate Domains-To obtain more detailed information of the FGFR4binding HS domain, we employed a novel method for sequencing of end-labeled, N-sulfated HS oligosaccharides (33). Affinity chromatography of [ 3 H]octasaccharides (i.e. the smallest fragments containing the FGFR4-binding domain; see Fig. 2) derived from bovine intestinal HS on FGFR4, yielded a fraction (ϳ15% of the starting material) of labeled components that remained bound to the receptor at physiological ionic strength (Fig. 7).
Approximately 60% of the bound octasaccharides were eluted from the FGFR4 affinity column at 0.2 M NaCl (in the following denoted as the 0.2 M fraction) whereas ϳ30% required 0.3 M NaCl (0.3 M fraction) for displacement (Fig. 7). Further selection of target fractions for sequence analysis within each affinity class aimed for components with the lowest charge density; this approach was adopted to pinpoint the minimal structural features required for FGFR4 binding, without redundant O-sulfate groups. The octasaccharides thus recovered (ϳ10% of the labeled material) from the 0.2 M fraction by preparative Propac PA1 chromatography (not shown) yielded a somewhat heterogeneous peak upon analytical rechromatography (Fig. 8A), at an elution position corresponding to an Nsulfated octasaccharide sequence with three O-sulfate groups.
Sequence analysis involved pHNO 2 treatment followed by Propac chromatography of the resultant, labeled even-numbered oligosaccharide fragments (di-, tetra-, and hexasaccharides from a parental octasaccharide, along with some uncleaved material). These oligosaccharides were then modified further, by stepwise incubations with IdoA2Sase, IdoA2Sase/ IdoAase, and IdoA2Sase/IdoAase/GlcN6Sase, the products of each step being analyzed by Propac chromatography (33). Sequence assignment was based on (i) the effects on each oligosaccharide of the various enzyme treatments and (ii) the observed elution positions of the oligosaccharides generated by pHNO 2 , as related to the behavior of standard oligosaccharides with known numbers of O-sulfate groups. 2 The effects on the oligosaccharides of the enzymes were highly reproducible and involved large, predictable shifts in elution position due to removal of sulfate groups, and smaller shifts after removal of nonreducing terminal IdoA residues. Notably, the residual parental oligosaccharides terminate with a GlcA unit (adjacent to a GlcNAc residue in the intact HS chain, hence not subject to C5-epimerization during HS biosynthesis (39)) and therefore resist the enzyme treatment. Even sequences of major components in mixtures could be resolved, based on the identification of distinct equal sized pHNO 2 oligosaccharides and the known (identical) number of sulfate groups in the corresponding intact saccharides (tri-O-sulfated octasaccharides in Fig. 8). In cases where deduced structures were not entirely represented by resolved degradation intermediates, the uncertainty in structural assignment has been indicated.
Application of these procedures to the recovered [ 3 H]octasaccharide sample (Propac chromatography) from the 0.2 M fraction (FGFR4 affinity chromatography) yielded the two tri-Osulfated sequences 8a and 8b shown in Fig. 10. These structures were deduced as follows. A hexasaccharide fraction (6b) appeared following pHNO 2 treatment, somewhat retarded compared with the parental octasaccharide fraction (Fig. 8B). This increase in charge density reflected the loss of a non-Osulfated GlcA-GlcNSO 3 disaccharide sequence (units 1 and 2 in Fig. 10; octasaccharides 8a and 8b). The hexasaccharide peak was eliminated by IdoA2Sase digestion, and replaced by two less anionic, incompletely resolved peaks, thus identifying an IdoA(2-OSO 3 ), unit 3. One of these components lost an additional sulfate group by IdoAase/GlcN6Sase treatment, whereas the other remained unaffected (Fig. 8, D and E). Unit 4 thus was 6-O-sulfated in one of the sequences (octasaccharide 8a) but not in the other (octasaccharide 8b). The elution positions of the two major tetrasaccharide products (4a and 4b in Fig. 8B) of pHNO 2 treatment suggested a difference in composition by one sulfate group. Both tetrasaccharides were susceptible to Ido2Sase (Fig. 8C) and IdoAase (Fig. 8D), but resistant to GlcN6Sase (Fig. 8E), and therefore both contained a nonreducing-terminal IdoA(2-OSO 3 )-GlcNSO 3 -disaccharide sequence (units 5 and 6 in Fig. 10, octasaccharides 8a and 8b). The additional O-sulfate group in octasaccharide 8b was located to the terminal aMan R residue by analysis of the labeled disaccharide fractions (2a and 2b in Fig. 8B). None of the disaccharides were affected by IdoA2Sase (Fig. 8C). Fraction 2a, that appeared at the elution position of nonsulfated HexA-aMan R , was partially cleaved by IdoAase, yielding nonsulfated labeled monosaccharide; however, the major portion remained unchanged, indicating a -GlcA-aMan R -sequence for units 7 and 8 in octasaccharide 8a (Fig. 10). Disaccharide 2b, at the elution position of monosulfated HexA-aMan R , was quantitatively converted to labeled aMan R (6-OSO 3 ) by digestion with IdoAase ( Fig. 8D; the peak coinciding with that of the 4aЈЉ product). This sulfated monosaccharide is not a substrate for GlcN6Sase (Fig.  8E). Octasaccharide 8b thus terminates (units 7 and 8) with an -IdoA-aMan R (6-OSO 3 )-structure (Fig. 10).
An octasaccharide fraction was recovered by preparative Propac chromatography, corresponding to ϳ10% of the labeled material eluted from the FGFR4 column with 0.3 M NaCl (not shown). Analytical Propac assessment showed a peak at an elution position consistent with the occurrence of N-sulfated octasaccharides containing four O-sulfate groups (Fig. 9A; cf. elution position of analogs with three O-sulfate groups in Fig.  8A). Of the fragments generated upon pHNO 2 treatment (Fig.   9B), hexasaccharide 6c, tetrasaccharide 4c, and disaccharide 2c could be assigned to the parent octasaccharide 8c shown in Fig.  10. Briefly, the elution position of 6c reflects the loss of a terminal mono-O-sulfated disaccharide residue, defining units 1-2 as GlcA-GlcNSO 3 (6-OSO 3 ) (see above). Treatment with IdoA2Sase produced major shifts in the positions of subfractions 6c and 4c, indicating loss of a 2-O-sulfate group from each fragment (Fig. 9C). Digestion with IdoAase led to the predicted, less prominent, shifts for 6cЈ, 4cЈ, and 2c, as expected for nonreducing-terminal IdoA residues (Fig. 9D). The elution position of the intact disaccharide 2c is that of IdoA-aMan R (6-OSO 3 ) (Fig. 9B). None of these three components were further affected by digestion with GlcN6Sase (Fig. 9E).
Additional fragments created by pHNO 2 treatment of the tetra-O-sulfated octasaccharide fraction included the two tetrasaccharides 4d and 4e, and the disaccharide 2d. Due to lack of any detectable corresponding hexasaccharide fragments, the sequences of the parental octasaccharides could not be directly resolved. However, fragments 4d and 2d appeared to derive from a common parent structure, since 4d was identified as a mono-O-sulfated tetrasaccharide with nonreducing terminal IdoA(2-OSO 3 )-saccharide, and 2d as a nonsulfated disaccharide (Fig. 9, B-E). Presumably, the parent octasaccharide 8d contains three additional O-sulfate groups that would be distributed among units 2-4, as shown in Fig. 10. Fragment 4e emerged like a tri-O-sulfated heparin tetrasaccharide (Fig. 9B), and the effects of enzyme digestions indicated a nonreducingterminal IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-disaccharide sequence ( Fig. 9, C-E), thus a structure not represented by any of the four identified octasaccharide sequences 8a-d (Fig. 10). Octasaccharide 8e thus would contain one additional O-sulfate group at an undetermined site within units 2-4.

DISCUSSION
In this study we have resolved octasaccharide sequences ( Fig. 10) with affinity toward FGFR4. The structures represent the first HS sequences shown to bind FGF receptors. To the best of our knowledge, they also represent the first application of the exoenzyme-based sequencing technology (33) to HS oligosaccharides with known protein binding activity.
Both heparin and HS were shown to bind to the extracellular domain of FGFR4 independently of FGF ligands. The binding of heparin was saturable, with a K D of 0.3-0.4 M, thus close to the affinity proposed for binding of heparin to FGFR2-IIIb (0.2 M) (40). The FGFR-heparin interactions appear to be considerably weaker than those between FGFs and heparin, suggesting that heparan sulfate in vivo would be more prone to bind free FGFs and FGF⅐FGFR complexes than free FGFR. The K D values for binding of FGF1 and FGF2 to heparin have been reported to be in the nanomolar range (ϳ100 and ϳ2 nM, respectively) (41,42).
Previous studies have shown that the minimal structural requirements for interaction of heparin/HS chains with FGF1 and FGF2 are specific and distinctly different (see "Introduction"). In this study we have evaluated the requirements for FGFR4 binding. Highly stringent structure/function relations were proposed by McKeehan et al. (38), who claimed that a number of FGFR species bind exclusively to the AT-binding sequence of heparin. The present results demonstrate that heparin decasaccharides with high and low affinity for AT can bind equally well to FGFR4 (Fig. 4). The reason for this discrepancy is unclear. Assessment of a FGF2:FGFR1:heparin decamer crystal suggested that the N-, 2-O-, and 6-O-sulfate groups of the saccharide are all involved in binding to the receptor component, the 6-O-sulfate groups appearing to be of importance (9). These findings agree with the present compositional analysis of HS decasaccharides fractionated with re-gard to affinity for FGFR4, that indicated appreciable enrichment of 6-O-sulfate groups in the bound fraction (Table I), and with previous functional studies pointing to a role for 6-Osulfate groups in HS-FGFR1 binding (11). Moreover, competitive binding of [ 3 H]heparin and various desulfated unlabeled heparin preparations to FGFR4 clearly implicated all three kinds of sulfate substituents in the interaction (Fig. 3). The question then arises whether binding to FGFR4 requires any specific disposition of these residues. All FGFR4-binding HS octasaccharides defined with regard to sequence (Fig. 10) were fully N-sulfated and showed the same distribution of IdoA(2-OSO 3 ) groups, invariably present in both internal disaccharide units. Whether both of these 2-O-sulfate groups are indeed required for FGFR4 binding or simply occur in most or all of the isolated octasaccharides remains unclear. By contrast, the distribution of GlcN(6-OSO 3 ) residues varied conspicuously. The two octasaccharides (8a and 8b) recovered by elution of the FGFR4 column with 0.2 M NaCl both contained a single 6-Osulfate group, but in different positions (units 4 and 8, respectively). If indeed 6-O-sulfate groups contribute to the interaction their precise location thus does not appear to be critical. Analysis of the octasaccharide species eluted with 0.3 M NaCl, hence with higher affinity for the receptor, reinforces this impression. In all, three components were considered, each containing a total of four O-sulfate groups. Even if only octasaccharide 8c could be pursued through the complete sequencing protocol, the information regarding octasaccharides 8d and 8e sufficed to demonstrate that the 0.3 M fraction contained octasaccharide species with highly diverse 6-O-sulfation patterns (Fig. 10). Yet these 6-O-sulfate groups clearly contribute to the higher overall ligand affinity of the 0.3 M fraction. We conclude that this increase in affinity is due to charge interactions at various locations of the saccharide:FGFR4 interface. Whether this diversity reflects non-FIG. 10. FGFR4 binding heparan sulfate octasaccharide sequences. Octasaccharides 8a and 8b were eluted from the FGFR4 column with 0.2 M NaCl, octasaccharides 8c, 8d, and 8e with 0.3 M NaCl. Sequences 8a-c were completely resolved, whereas sequences 8d and 8e were indirectly deduced from the resolved structures of the corresponding tetrasaccharides, and the known overall degree of O-sulfation (4 residues per molecule) of the parent octasaccharides. The locations of the three O-sulfate groups marked with asterisks in structure 8d (units 2-4) are tentative but highly plausible, since tetrasaccharide 4d was shown to contain only one O-sulfate residue (unit 5). Conversely, only one of the three positions occupied by unspecified R residues in sequence 8e carries an O-sulfate group, the remaining three O-sulfate residues being assigned to units 5, 6, and 8. In some of the octasaccharides unit 7 can be either GlcA or IdoA; the sequences shown represent the predominant structures. The hydrogen and hydroxyl groups are not shown.

FGFR4-binding Heparan Sulfate Domains
specific polyelectrolyte effects or selective interactions with distinct basic amino acid residues remains to be determined through more refined analysis at the molecular and atomic levels.
What is the functional significance of interactions between FGFRs and heparin-related GAGs? Notably, highly sulfated polyanions such as heparin and synthetic polysulfonates, but not HS that is less sulfated, can activate FGFR4 in the absence of a FGF ligand (20). This effect of heparin is not well understood, but could conceivably be of importance upon heparin release from degranulating mast cells in tissues with abundant FGFR4 expression. However, the major role in vivo of the "heparin-binding domain" of FGFR4 is presumably to mediate the formation of ternary complexes with HS proteoglycans and FGFs, leading to receptor activation. The interpretation of the present data is hampered by a lack of information regarding the functional and topological organization of HS proteoglycan species in the intact tissue, here intestinal mucosa, used as source of HS for oligosaccharide generation. Thus we do not know whether the HS sequences found to bind FGFR4 were actually designed for such interaction in vivo. The relatively modest affinities observed do not argue against an essential functional role in promoting formation of ternary complex (47). However, for productive interactions the FGFR-binding sequences must be contiguous with (or partially overlap) the FGF-binding domains, hence the need for an overall deca-to dodecamer N-sulfated region of the HS chain to accommodate both binding sites. Such regions account for only a small proportion, usually a few percent, of the total HS chains (48). The structural characteristics required for productive interactions of a particular FGF and its receptor with an N-sulfated HS domain remain unclear. At least some of the members of the FGF family, such as FGF1 and FGF2, show highly distinct requirements for HS ligand structure (26,49), but these features appear to be complemented by less exacting terms for FGFR binding. N-Sulfated domains that are too short, or otherwise structurally unfit to simultaneously interact with both FGF and its receptor, may store the growth factor in the pericellular environment and actually inhibit receptor activation (50,51). Heparin, frequently used as a substitute for HS in experimental work, consists largely of trisulfated -IdoA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-disaccharide units, and thus will basically cover most or all combinations of N-and O-sulfate groups required for interactions with different FGFs and FGFRs. By contrast, HS oligosaccharides with closely similar composition can be fractionated into species that either activate or inhibit FGF2 signaling in HS-deficient cells (18). Moreover, the preference for growth factor target (FGF1 or FGF2) could be shifted by subtle changes in HS structure (52)(53)(54). These findings emphasize the need for further analysis, not only of HS structures in relation to growth factor action, but also of the mechanisms that control the generation of such structures during their biosynthesis.