Defining the interleukin-8-binding domain of heparan sulfate.

Interleukin-8, a member of the CXC chemokine family, has been shown to bind to glycosaminoglycans. It has been suggested that heparan sulfate on cell surfaces could provide specific ligand sites on endothelial cells to retain the highly diffusible inflammatory chemokine for presentation to leukocytes. By using selectively modified heparin and heparan sulfate fragments in a nitrocellulose filter trapping system, we have analyzed sequence requirements for interleukin-8 binding to heparin/heparan sulfate. We demonstrate that the affinity of a monomeric interleukin-8 molecule for heparin/heparan sulfate is too weak to allow binding at physiological ionic strength, whereas the dimeric form of the protein mediates binding to two sulfated domains of heparan sulfate. These domains, each an N-sulfated block of approximately 6 monosaccharide units, are contained within an approximately 22-24-mer sequence and are separated by a region of </=14 monosaccharide residues that may be fully N-acetylated. Binding to interleukin-8 correlates with the occurrence of the di-O-sulfated disaccharide unit -IdceA(2-OSO3)-GlcNSO3(6-OSO3)-. We suggest that the heparan sulfate sequence binds in horseshoe fashion over two antiparallel-oriented helical regions on the dimeric protein.

Interleukin-8, a member of the CXC chemokine family, has been shown to bind to glycosaminoglycans. It has been suggested that heparan sulfate on cell surfaces could provide specific ligand sites on endothelial cells to retain the highly diffusible inflammatory chemokine for presentation to leukocytes. By using selectively modified heparin and heparan sulfate fragments in a nitrocellulose filter trapping system, we have analyzed sequence requirements for interleukin-8 binding to heparin/heparan sulfate. We demonstrate that the affinity of a monomeric interleukin-8 molecule for heparin/ heparan sulfate is too weak to allow binding at physiological ionic strength, whereas the dimeric form of the protein mediates binding to two sulfated domains of heparan sulfate. These domains, each an N-sulfated block of ϳ6 monosaccharide units, are contained within an ϳ22-24-mer sequence and are separated by a region of <14 monosaccharide residues that may be fully N-

acetylated. Binding to interleukin-8 correlates with the occurrence of the di-O-sulfated disaccharide unit -Id-ceA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-. We suggest that the heparan sulfate sequence binds in horseshoe fashion over two antiparallel-oriented helical regions on the dimeric protein.
Glycosaminoglycans (GAGs), 1 i.e. the linear, sulfate-substituted carbohydrate constituents of proteoglycans, are ubiquitous components of cell surfaces. They are generally believed to exert their biological functions by interacting with proteins. Heparin and heparan sulfate (HS) are sulfated GAGs of alternating hexuronic acid and glucosamine residues that have been shown to bind a variety of enzymes, cytokines/growth factors, and extracellular matrix molecules (for review, see Ref. 1). Such interactions may be highly specific, as illustrated by heparin/HS and antithrombin. A defined pentasaccharide sequence in the GAG chain binds antithrombin and thus strongly promotes its function as an inhibitor of the serine proteases acting within the coagulation cascade (2). Other proteins, such as histones, bind GAGs due to their overall basic character, without any apparent need for a specific saccharide sequence. Several attempts have been made to identify minimal consensus sequences in proteins required to bind heparin or HS (3,4). Clusters of basic amino acid residues may provide binding sites that are either located within a short linear peptide stretch or composed of residues that occur topologically close together, but on different peptide loops (for review, see Ref. 5).
Sulfate groups on the GAG chains have been identified as important determinants of protein-binding sites. They are introduced during the process of polymer modification, through which the initial (GlcUA␤1,4-GlcNAc␣1,4) n polysaccharide chain is transformed into the final product (6). While each reaction, catalyzed by a distinct enzyme, generates the substrate for subsequent reactions, these are generally incomplete, which leads to progressive structural heterogeneity. Due to modulated activity of the GlcNAc N-deacetylase/N-sulfotransferase, i.e. the enzyme that initiates polymer modification in HS and heparin biosynthesis, variable proportions of GlcNAc units escape N-deacetylation/N-sulfation. These residual Nacetylated units are few and isolated in heparin and are more abundant and typically arranged in consecutive sequence in HS. The N-sulfate groups are prerequisite to substrate recognition by the GlcUA 5-epimerase and the O-sulfotransferases that catalyze the subsequent modification reactions. Consequently, heparin shows a high proportion of IdceA-containing, O-sulfated disaccharide units and a relatively homogeneous overall sulfation pattern, whereas HS is composed of alternating sulfated and nonsulfated sequences of variable length (7,8). A typical HS chain will thus contain essentially unmodified regions of up to 8 -10 consecutive GlcUA-GlcNAc disaccharide units. Clearly, the occurrence of such regions will profoundly influence the mode of interaction with proteins.
Multidomain interaction between HS and proteins has so far received little attention. Such interaction would apply in particular to oligomeric proteins composed of monomers carrying single "heparin-binding sites." To study such a system, we have chosen the heparin-binding cytokine interleukin-8 (IL-8). IL-8 is a proinflammatory cytokine of the CXC chemokine family that forms noncovalently linked dimers with a characteristic appearance (9). Crystallographic and NMR studies have provided insights into the shape of the molecule, which consists of a flat array of a ␤-pleated sheet with the ␣-helices of the two monomers arranged as antiparallel rods on top of this sheet (10 -12). The heparin-binding sites have been tentatively localized to the exposed positively charged residues on the top of these helices (13,14).
The results of this study define constraints critical to the interaction of a HS chain with the IL-8 dimer. A minimal sequence of 18 -20 monosaccharide units is required to span the two saccharide-binding sites on the dimer. The corresponding peptide-binding regions may be separated by a nonsulfated intervening stretch composed of up to 7 GlcUA-GlcNAc disaccharide units.

EXPERIMENTAL PROCEDURES
Materials-Recombinant human IL-8 was expressed in Escherichia coli and purified as described previously (15), except that a final step of affinity chromatography on heparin-agarose was added. Twenty mg of lyophilized IL-8 were dissolved in 10 ml of phosphate-buffered saline and passed over a 2-ml column of heparin-agarose (Sigma, type I) equilibrated in the same buffer. The column was developed with a linear salt gradient ranging from 0.15 to 2 M NaCl in 20 mM phosphate buffer, pH 8.2. The protein eluted as a single major peak at around 0.6 M NaCl. The pooled fractions were dialyzed against four changes of 50 mM ammonium bicarbonate and lyophilized. Bovine lung heparin (The Upjohn Co.) was purified as described (16). Heparan sulfate from bovine kidney was a gift of Seikagaku Corp. (Tokyo, Japan). Size-defined fragments of E. coli K5 polysaccharide were a generous gift of Dr. K. Lidholt (17). Heparinase from Flavobacterium heparinum (EC 4.2.2.7), which is specific for IdceA(2-OSO 3 ), was obtained from Sigma, and heparitinase from F. heparinum (heparitin-sulfate lyase, EC 4.2.2.8; a mixture of heparitinases I and II according to the information by the supplier), specific for N-acetylated sequences, was purchased from Seikagaku Corp. Bio-Gel P-10 (superfine) was from Bio-Rad. Sephadex G-15, Sephadex G-50, Superose 6, and Superose 12 (prepacked 1.5 ϫ 30-cm columns) and a prototype of a Superdex 30 column (2 ϫ 60 cm) for fast performance liquid chromatography were obtained from Pharmacia (Uppsala, Sweden). A Partisil-10 SAX anion-exchange column (4.6 ϫ 250 mm) and Whatman No. 3MM filter paper were purchased from Whatman Ltd. (Maidstone, Kent, Great Britain). NaB 3 H 4 (24 -28 Ci/ mmol) and [ 3 H]acetic anhydride (500 mCi/mmol) were obtained from Amersham International (Buckinghamshire, United Kingdom). Hydrazine hydrate was purchased from Fluka (Buchs, Switzerland). All other reagents were the best grade available.
Radiolabeling of Glycosaminoglycans-Heparin was N-3 H-acetylated at free amino groups as described (18) to a specific activity of 20,000 dpm of 3 H/g of hexuronic acid (ϳ0.2 ϫ 10 6 dpm/nmol of heparin). Bovine kidney HS (0.5 mg) was partially N-deacetylated by hydrazinolysis for 30 min at 96°C in 1 ml of hydrazine hydrate (ϳ30% water) and 1% hydrazine sulfate (19). The product was N-3 H-acetylated with 2.5 mCi of [ 3 H]acetic anhydride essentially as described (18) and extensively dialyzed against 1 M sodium acetate, water, and 0.25 M HIO 3 to eliminate any hydrazides found (20) and finally against water. A specific activity of ϳ0.4 ϫ 10 6 dpm/g of uronic acid was achieved (corresponding to an average of ϳ4.6 ϫ 10 6 dpm/nmol of HS based on an estimated molecular mass of 11.5 kDa for the HS chains labeled).
Chemical Modification of Heparin-Selective chemical desulfation of heparin was performed as described previously (21) (see Table I). A preferentially 6-O-desulfated sample obtained by treatment with dimethyl sulfoxide/methanol (9:1, v/v) for 2 h at 93°C was used as starting material for chemical depolymerization as described below.
Chemical Depolymerization of Polysaccharides-Limited deaminative cleavage of native and modified heparin was performed as follows to create size-defined fragments. To cleave chains at N-sulfated glucosamine units, 10 mg of native or 2-O-desulfated heparin (which still contains most of the N-sulfates) were dissolved in water adjusted to pH 1.5 with H 2 SO 4 (final volume, 1 ml). Following the addition of NaNO 2 (140 g in 10 l), the solutions were kept on ice for 3 h. The reaction was stopped by adjusting the pH to 9 with Na 2 CO 3 , and the cleavage products were reduced with 2.5 mCi of NaB 3 H 4 overnight at room temperature, followed by 10 mg of unlabeled NaBH 4 for 2 h. The resultant labeled oligosaccharides were desalted on a column (1.2 ϫ 66 cm) of Sephadex G-15 in 0.2 M NH 4 HCO 3 and separated by gel chromatography on a column (1 ϫ 146 cm) of Bio-Gel P-10 in 0.5 M NH 4 HCO 3 . Preferentially 6-O-desulfated heparin was cleaved at some of the Nunsubstituted amino groups of glucosamine residues that had been created by the preceding desulfation step. A 10-mg sample was dissolved in water, and the solution was adjusted to pH 4.0 with CH 3 COOH and rendered 0.1 M in NaNO 2 (final volume, 1 ml). The sample was incubated on ice for 30 min. The reaction was terminated, and the sample was 3 H-labeled as described above. All O-desulfated heparin fragments were re-N-sulfated as described (21) before final separation by gel chromatography on Bio-Gel P-10. The specific activities of the 3 H-labeled modified heparin fragments ranged between 0.25 and 0.5 ϫ 10 6 dpm/nmol of oligosaccharide.
Enzymatic Depolymerization of Polysaccharides-Size-defined HS fragments were produced enzymatically. Ten g of N-3 H-acetylated bovine kidney HS (specific activity, 0.4 ϫ 10 6 dpm/g of uronic acid, ϳ4.6 ϫ 10 6 dpm/nmol of HS) were digested with 0.04 units of heparinase in 200 l of 5 mM sodium P i , pH 7.0, 150 mM NaCl, and 0.1 mg/ml bovine serum albumin for 1 h at 37°C. The digest was separated on a Bio-Gel P-10 column as described for the heparin fragments. HS fragments of approximate 20 -24-mer size (by reference to heparin oligosaccharide standards) were isolated and used in binding experiments with IL-8.
In a protection approach, the N-3 H-acetylated HS (ϳ1 g) was incubated with 0.1-1.0 milliunit of heparitinase (a mixture of heparitinases I and II specific for N-acetylated glucosamine units) in the presence or absence of 10 g of IL-8 in 200 l of 50 mM Hepes buffer, pH 7.0, and 1 mM CaCl 2 . Digestion was maintained for 60 min at 43°C and was then stopped by heating at 96°C for 5 min. The digests were separated on a Superose 12 column in 50 mM Tris, pH 7.5, and 1 M NaCl, and effluent fractions were analyzed by scintillation counting and pooled as indicated.
Binding Assays-Radiolabeled GAGs or size-defined fragments were incubated with protein in 10 mM phosphate buffer, pH 7.4, containing different amounts of salt (see figure legends) as described (22). Protein along with any bound oligosaccharides was trapped on a nitrocellulose filter (25-mm diameter unless otherwise stated). The filter was washed with the incubation buffer, protein-bound oligosaccharides were dissociated from the membrane in 2 M NaCl, and the radioactivity was measured or the sample was desalted for further analysis. In preparative experiments, 1 nmol each of the indicated 3 H-labeled modified heparin 18-mers was incubated with 10 nmol of protein in a final volume of 1 ml of phosphate-buffered saline. The bound fractions were isolated by filter absorption over two 45-mm filters and recovered as described above. The amounts of bound material were calculated based on recovered 3 H and the specific radioactivity of the respective preparations.
Analysis of GAGs-GAGs were quantified by colorimetric determination of hexuronic acid using the meta-hydroxydiphenyl method (23) with GlcUA as a standard. A factor of 3 was arbitrarily employed to convert values to saccharide mass.
To ensure complete N-substitution following chemical N-sulfation, saccharide samples were tested for N-unsubstituted glucosamine residues by treatment with nitrous acid at pH 4 (24), followed by gel chromatography on a column (1 ϫ 150 cm) of Sephadex G-50 in 1 M NaCl. Size separation of selectively desulfated heparin preparations on a Superose 6 column gave an average molecular mass for all preparations of 10 -11 kDa. Thus, no appreciable cleavage had occurred during chemical processing.
The composition of native or modified heparin chains or oligosaccharides was determined after complete deamination at pH 1.5 for Nsulfated chains or at pH 4 following hydrazinolysis of N-acetylated chains (4 h at 96°C) (25). The resultant disaccharides were reduced with NaB 3 H 4 , recovered by gel chromatography on Sephadex G-15 (1 ϫ 200 cm), and separated by anion-exchange high performance liquid chromatography on a Partisil-10 SAX column. The proportions of nonsulfated disaccharides were determined by high-voltage paper electrophoresis on Whatman No. 3MM paper in pyridine acetate buffer, pH 5.3.

Minimal Size of Heparin Oligosaccharide Binding to IL-8 -
The smallest oligosaccharide fragment able to bind to IL-8 was identified by in-solution binding studies using radiolabeled size-defined heparin fragments. Under physiological buffer conditions, the smallest fragment with appreciable affinity for IL-8 was a heparin 18-mer (Fig. 1A). The interaction between heparin and IL-8 was previously shown to be of relatively low affinity, with a K d of ϳ6 ϫ 10 Ϫ6 M (26), and we therefore considered the possibility of a charge-dependent interaction involving a shorter saccharide sequence that had escaped detection due to the ionic conditions of the assay. Indeed, washing of the nitrocellulose filter at lower ionic strength (10 mM phosphate buffer without added NaCl) revealed a remarkably distinct binding profile with a heparin 8-mer as the smallest binding species (Fig. 1A). In accord with this finding, small heparin oligosaccharides were able to compete with full-sized radiolabeled heparin for binding to IL-8 at physiological ionic strength. The smallest oligosaccharide that inhibited the binding of labeled heparin for 50% or more was a 6-mer (Fig. 1B).
Sulfate Dependence of Heparin Binding to IL-8 -The importance of different sulfate groups in the binding of heparinrelated saccharides to IL-8 was evaluated using a competition assay with variably modified heparin chains. Full-length radiolabeled heparin was allowed to interact with IL-8 in the presence of unlabeled native or selectively modified heparin ( Fig. 2 and Table I) . Native bovine lung heparin showed the highest inhibitor capacity, above that of any of the modified heparins. Affinity Fractionation of Partially Desulfated Heparin Oligosaccharides-To further define the role of O-sulfate groups, 3 H-labeled 18-mers derived from selectively O-desulfated heparins (see "Materials") were affinity-fractionated, and the disaccharide composition of IL-8-bound fractions was compared with that of the corresponding unfractionated material. Application of a preferentially 6-O-desulfated 18-mer to this procedure resulted in retention of 6% of the material on the nitro-cellulose filter, complexed with IL-8, whereas the 2-Odesulfated fraction yielded 15% of IL-8-bound material.
These bound fragments were dissociated from nitrocelluloseadsorbed protein by sodium chloride and desalted. In analytical assays, aliquots of the bound fractions were compared with the respective unfractionated starting material for binding to IL-8. Both preparations, independent of the type of modification, showed increased binding to IL-8 as compared with the unfractionated starting material. Under similar interaction conditions, 1. Domain Structure of IL-8-binding Heparan Sulfate-The interaction studies using native or modified heparin oligomers demonstrated that the IL-8 dimer may be spanned by exclusively N-sulfated, extended saccharide sequences. However, since the physiological GAG ligand for IL-8 at the cell surface is presumably HS rather than heparin, we proceeded to study HS-derived IL-8-binding oligosaccharides, with the explicit aim of assessing to what extent such species may contain N-acetylated domains. N-[ 3 H]Acetyl-labeled kidney HS was first degraded by limited digestion with heparinase to cleave glucosaminidic linkages between GlcNSO 3 (Ϯ6-OSO 3 ) and IdceA(2-OSO 3 ) units, thus generating HS oligomers with internal (labeled) N-acetylated sequences flanked on both sides by N-sulfated, presumably IL-8-binding regions (Fig. 4A). After size separation, a fraction corresponding to a heparin ϳ20 -24mer was selected for binding to IL-8 using the nitrocellulose trapping system. Following incubation with the protein, the bound (ϳ10% of added 3 H under conditions of low-salt wash) and unbound fractions were recovered and treated exhaustively with nitrous acid at pH 1.5. The resulting labeled, exclusively N-3 H-acetylated fragments were separated by gel chromatography (Fig. 5). The IL-8-unbound fraction yielded a spectrum of variously sized oligomers, with a predominant component larger than a heparin 12-mer or a K5 16-mer (see legend to Fig. 5). This component was largely lacking in the deamination products from the IL-8-bound fraction. Instead, these products showed a larger proportion of smaller labeled oligosaccharides, which ranged in size from heparin 4-to 12mers. This result indicates that an IL-8-binding region in a HS These conclusions were corroborated in enzyme protection experiments involving heparitinase, which will preferentially cleave nonsulfated (GlcUA-GlcNAc) n sequences of the HS chain ( Fig. 4B). N-[ 3 H]Acetyl-labeled kidney HS was incubated with various amounts of enzyme in the presence or absence of IL-8 as described in the legend to Fig. 6. Gel chromatography of the digests obtained in the absence of cytokine indicated a fairly homogeneous distribution of oligosaccharides, with peak elution positions shifting corresponding to the amounts of enzyme added (Fig. 6B). Cleavage patterns in the presence of IL-8 appeared biphasic, such that a minor fraction was protected from extensive degradation while the majority of labeled material was degraded, in fact more extensively than the corresponding material in the absence of IL-8 (Fig. 6A). To characterize the two pools, unprotected and protected digests ( 18-mers. Peaks were identified by reference to the elution position of standard disaccharides as follows: peak 1, GlcUA-aMan R /IdceA-aMan R ; peak 2, GlcUA-aMan R (6-OSO 3 ); peak 3, IdceA-aMan R (6-OSO 3 ); peak 4, IdceA(2-OSO 3 )-aMan R ; peak 5, IdceA(2-OSO 3 )-aMan R (6-OSO 3 ). Peaks marked with a bracket under an asterisk indicate tetrasaccharides, partly due to "anomalous" ring contraction (24), and were not included in the quantification of disaccharides presented in Table II.

Interleukin-8/Heparan Sulfate Interaction
of the IL-8-protected fragment (Fig. 6, C, pool I; and D) is clearly larger than a standard 20-mer, degradation by HNO 2 at pH 1.5 results in fragments of various sizes from 4-mers up to ϳ16-mers, in accord with the notion that a sequence of up to ϳ7 contiguous N-acetylated disaccharide units may be inserted between the sulfated IL-8 monomer-binding regions. 2 In contrast to these fragments, the prominent fraction of low-molecular mass degradation products (pool II), which presumably reflected facilitated enzymatic attack of unprotected stretches of GlcNAc-containing saccharides, was clearly smaller (Fig. 6E) than pool I material and was fairly similar to the total pool created in the IL-8-free digest (Fig. 6F). Both these preparations were completely degraded to short oligosaccharides by HNO 2 .

DISCUSSION
A number of cytokines and chemokines, including granulocyte/macrophage colony-stimulating factor (27), interleukin-3 (28), IL-8 (9), macrophage inflammatory protein-1␤ (29), and platelet factor-4 (22,30), have been shown to bind to proteo-glycans/GAGs. While the functional implications of these interactions are not fully understood, it has been proposed that the proteoglycans may act as storage sites for the small highly diffusible molecules or aid in their presentation to receptors (28,(31)(32)(33)(34). A model study involving various chemokines and heparin suggested that GAG chains might provide specificity to chemokine action also in vivo (26), yet little information has been available regarding the structural properties of these chains as required to mediate protein binding.
One aim of this study was to define the role of sulfate substituents in GAG binding to IL-8 as reflected through the interaction with fully sulfated, 3 H-labeled heparin. The interpretation of such experiments is complicated by the potential redundancy of sulfate groups in saccharide sequences shown to be involved in binding. A striking example is the interaction of basic fibroblast growth factor with a fully sulfated heparin hexasaccharide region, in which only one out of two or three IdceA 2-O-sulfate groups and none of the GlcN 6-O-sulfate residues were found to be essential for binding (35,36). One approach to the analysis of sulfate requirement is to apply selectively desulfated heparin derivatives in competition experiments with fully sulfated, radiolabeled heparin and the protein ligand. Preferential removal of N-, 2-O-, or 6-O-sulfate groups was thus found to appreciably impede the interaction with IL-8, suggesting that all types of sulfate substituents contribute to binding. This conclusion was verified in direct binding experiments utilizing partially O-desulfated, 3 H-labeled heparin 18-mers. Preferential 2-O-and 6-O-desulfation both yielded products from which minor fractions could be sequestered by binding to IL-8. While the overall degree of sulfation of each IL-8-binding species did not differ significantly from that of the corresponding parent preparation, the two binding fractions differed markedly in composition from each other (Table II). Clearly, it must be assumed that neither all the IdceA 2-O-sulfate groups present in the IL-8-binding fraction derived from the preferentially 6-O-desulfated heparin 18-mer nor all the GlcN 6-O-sulfate groups in the corresponding species obtained following preferential 2-O-desulfation could be essential for the interaction. However, both fractions of binding components were appreciably enriched in the minor di-O-sulfated -IdceA(2-OSO 3 )-GlcNSO 3 (6-OSO 3 )-disaccharide unit; this finding applied in particular to the preferentially 6-O-desulfated material (Table II). We conclude that the 2-O-  5. Binding of [ 3 H]HS fragments to IL-8. N-3 H-Acetylated HS was cleaved by heparinase digestion, and a fraction corresponding to an ϳ20 -24-mer heparin standard was applied to IL-8 binding. A sample of 0.1 ϫ 10 6 dpm was incubated with 10 nmol of IL-8 under standard conditions, and bound (ϳ10%) and unbound fractions were recovered following nitrocellulose filter trapping. Both fractions were then desalted and subjected to exhaustive deamination at pH 1.5. The products derived from IL-8-bound (q) and -unbound (E) preparations were separated on Superdex 30 as described under "Experimental Procedures." The elution positions of standard, size-defined, 3 H-labeled oligosaccharides derived from heparin and from E. coli K5 polysaccharides are indicated by arrows and open arrows, respectively. Heparin fragments are highly sulfated, whereas the bacterial saccharides are nonsulfated. HS chains contain only a few sulfate groups within the nitrous acidresistant N-acetylated domains and therefore presumably elute more closely to the K5 standards. and 6-O-disulfated disaccharide structure promotes binding. On the other hand, each 18-mer contains two binding sites for IL-8 (see below), whereas neither of the two IL-8-binding fractions contained more than 1 mol of di-O-sulfated disaccharide unit/mol of 18-mer. Therefore, it must be assumed that some of the mono-O-sulfated disaccharides may substitute for a "missing" di-O-sulfated disaccharide unit. The precise and optimal arrangement of variously sulfated sugar units in the IL-8binding site remains to be defined.
The interaction of heparin with IL-8 is relatively weak (26) and was found in this study to require fragments Ն18-mers for measurable binding under physiological ionic conditions (Fig.  1). However, under conditions of reduced ionic strength, a distinct 8-mer lower size limit for binding was detected using the nitrocellulose filter trapping assay. Considering the cleavage specificity of the nitrous acid reaction used to randomly depolymerize the heparin starting material, the actual proteinbinding site might be as small as a pentasaccharide sequence within the recovered 8-mer (cf. the analogous antithrombinbinding site in heparin (37)). Accordingly, heparin oligosaccharides Ͻ10-mers could efficiently compete with full-sized radiolabeled heparin for binding to IL-8 (Fig. 1). Taken together with the known propensity of IL-8 to form dimers (12,38), these findings are readily interpreted in terms of a trimeric complex involving an extended sequence that spans two small binding sites, one on each IL-8 monomer. Indeed, dimerization of IL-8 upon binding to heparin was recently demonstrated (39).
The physiological saccharide ligand for the IL-8 dimer presumably is not heparin, but rather HS proteoglycans at the surface of the vascular endothelium (32,34). Witt and Lander (26), using affinity co-electrophoresis, noted that HS bound to IL-8 with about the same affinity as heparin having a much higher overall degree of sulfation. This finding was attributed to the orientation of the Glu-63 residues in the IL-8 homodimer, which would create a patch of negative charge between the two clusters of basic amino acid residues that form the individual "heparin-binding sites" on each IL-8 monomer. It was hypothesized that this arrangement would favor binding of polysaccharide chains containing sulfated regions interspersed by a nonsulfated sequence of appropriate length that would minimize charge repulsion due to the Glu units. Indeed, this study revealed an internal sequence of up to 7 consecutive N-acetylated disaccharide units in HS fragments capable of binding to IL-8.
Combining structural data for the protein and polysaccharide moieties defines the constraints of potential interaction models. Crystallographic and NMR studies of the IL-8 dimer point to a 2-fold symmetrical arrangement of the monomers, with the two ␣-helices in antiparallel orientation (12,40). The heparin-binding sites have been tentatively localized to the positively charged amino acid residues exposed on these ridges (13,14). The dimensions of the IL-8 dimer have been estimated to 40 ϫ 42 ϫ 32 Å, with the two ␣-helices being 12-14 Å apart. Since the length of a heparin disaccharide unit is estimated to ϳ8.5 Å (41-43), a fully extended HS 22-24-mer would measure ϳ100 Å and thus would fit in horseshoe-like fashion across the dimer. According to this model, the postulated internal Nacetylated domain (Յ12-14-mer) would bridge the two ␣-helices, whereas the two adjacent sulfated regions (5-6-mer each) would each run alongside one helix (Fig. 7). Notably, the symmetry of the protein dimer would allow the HS chain to accommodate each of the two interaction sites with similar polarity.
The functional role of the intervening saccharide sequence remains to be clearly defined. Variations within this structure could conceivably modulate the protein/HS interaction. Conversely, a comparison of different chemokines with similar structures revealed distinct structural differences in the polypeptide domain facing the GAG ligand that were potentially implicated in the discrimination between different polysaccharide ligand subspecies (26). N-Acetylated sequences bridging separated sulfated protein-binding domains were found also in HS species binding to dimeric interferon-␥, another cytokine not related to IL-8 (44), and to tetrameric platelet factor-4 (30). The bridging domain in the case of platelet factor-4 seems to be similar in size to the one identified for IL-8 (30), although the domain is suggested to stretch over the intersection between two dimers and not, as in the suggested model for IL-8, over two monomers. In the case of interferon-␥, the bridging domain was considerably longer (15-16 disaccharide units) than that associated with IL-8 binding and therefore was thought to extend in a more loosely knit loop structure. Such extended N-acetylated regions, while clearly present in the bovine kidney HS investigated, were essentially absent from the IL-8-associated fragments recovered following digestion with either heparinase or heparitinase (Figs. 5 and 6). The significance of this difference in domain structure as related to the roles of HS in cytokine function is not understood. Conceivably, the variability could be exploited to modulate the efficiency of chemokine retention on different cell surfaces or to regulate the selection between different cytokine targets.
It has been proposed that HS promotes IL-8-mediated attraction of neutrophils by providing a trap on the endothelium or in the extracellular matrix that sequesters the secreted chemokine and prevents its diffusion from an inflammatory site (32,34) as well as its clearance from the blood stream through nonspecific chemokine receptors on erythrocytes (45). Moreover, HS and heparin had a positive effect on IL-8-mediated chemotactic activation of neutrophil migration in vitro (14), although haptotactic activation could not be excluded in this study. A secondary effect of added GAGs was inactivation of elastase released from the neutrophils. HS thus could promote chemoattraction by enhancing the local concentration of IL-8, either by retaining the chemokine on the HS chain or by protecting it from rapid degradation by the released protease (46,47). It has been noted that whereas HS appears to preferentially bind the dimeric form of IL-8 under physiological conditions (39), monomeric IL-8 has been claimed to be fully active in eliciting a functional response (38,48). The IL-8/HS interaction may therefore primarily serve to retain the cytokine at its site of production and secretion.