The Heparin/Heparan Sulfate-binding Site on Apo-serum Amyloid A

Serum amyloid A isoforms, apoSAA1 and apoSAA2, are apolipoproteins of unknown function that become major components of high density lipoprotein (HDL) during the acute phase of an inflammatory response. ApoSAA is also the precursor of inflammation-associated amyloid, and there is strong evidence that the formation of inflammation-associated and other types of amyloid is promoted by heparan sulfate (HS). Data presented herein demonstrate that both mouse and human apoSAA contain binding sites that are specific for heparin and HS, with no binding for the other major glycosaminoglycans detected. Cyanogen bromide-generated peptides of mouse apoSAA1 and apoSAA2 were screened for heparin binding activity. Two peptides, an apoSAA1-derived 80-mer (residues 24–103) and a smaller carboxyl-terminal 27-mer peptide of apoSAA2 (residues 77–103), were retained by a heparin column. A synthetic peptide corresponding to the CNBr-generated 27-mer also bound heparin, and by substituting or deleting one or more of its six basic residues (Arg-83, His-84, Arg-86, Lys-89, Arg-95, and Lys-102), their relative importance for heparin and HS binding was determined. The Lys-102 residue appeared to be required only for HS binding. The residues Arg-86, Lys-89, Arg-95, and Lys-102 are phylogenetically conserved suggesting that the heparin/HS binding activity may be an important aspect of the function of apoSAA. HS linked by its carboxyl groups to an Affi-Gel column or treated with carbodiimide to block its carboxyl groups lost the ability to bind apoSAA. HDL-apoSAA did not bind to heparin; however, it did bind to HS, an interaction to which apoA-I contributed. Results from binding experiments with Congo Red-Sepharose 4B columns support the conclusions of a recent structural study which found that heparin binding domains have a common spatial distance of about 20 Å between their two outer basic residues. Our present work provides direct evidence that apoSAA can associate with HS (and heparin) and that the occupation of its binding site by HS, and HS analogs, likely caused the previously reported increase in amyloidogenic conformation (β-sheet) of apoSAA2 (McCubbin, W. D., Kay, C. M., Narindrasorasak, S., and Kisilevsky, R. (1988) Biochem. J. 256, 775–783) and their amyloid-suppressing effects in vivo (Kisilevsky, R., Lemieux, L. J., Fraser, P. E., Kong, X., Hultin, P. G., and Szarek, W. A. (1995) Nat. Med. 1, 143–147), respectively.

In response to tissue injury or infection, activated macrophages secrete cytokines (interleukin-1 and -6 and tumor necrosis factor) that induce liver synthesis of a number of acutephase (AP) 1 proteins (1,2). The function of most of the AP proteins is unknown, but it is widely accepted that their purpose is to enhance host survival by neutralizing infectious agents, contributing to tissue repair, and restoring homeostasis. One of these AP proteins is a novel HDL apolipoprotein, called serum amyloid A (apoSAA), which is encoded by a multigene family conserved from fish to humans (3)(4)(5). Maximum transcription rates are reached for two of the four known apoSAA isoforms (apoSAA1 and apoSAA2) 3 h after AP induction (6 -8). Their concentration in plasma increases 500 -1000fold (from 1-5 g/ml up to 1 mg/ml) within 18 -24 h and returns to near normal levels within 5-7 days of a single inflammatory stimulus (9). ApoSAA is found mainly associated with HDL 3 (10,11), where it has been postulated to modulate cholesterol transport during the AP response (12,13). But several other functions have also been proposed for apoSAA including immune regulation, as a chemoattractant and as an inhibitor of fever induction, platelet activation, and neutrophil oxidative burst (3).
ApoSAA was originally identified by its cross-reactivity with antisera raised against peptides isolated from inflammationassociated amyloid (AA amyloid) (14,15) which is a pathological tissue deposit associated with chronic inflammatory diseases. Amyloid is a generic term describing the primarily extracellular accumulation of fibrillar protein deposits that have unique tinctorial and structural properties and that cause the disruption of tissue architecture and function (16,17). ApoSAA and at least 17 other unrelated normally nonfibrillar proteins are known precursors of amyloid (18). Each is associated with a specific disease such as Alzheimer's disease, chronic hemodialysis, adult-onset diabetes, rheumatoid arthritis, and certain malignancies. Regardless of the underlying amyloid fibril protein/peptide or associated disease, isolated amyloids fibrils are composed of two or more 3-nm filaments twisted around each other forming nonbranching fibrils, 7-10 nm in diameter, with a crossed ␤-pleated sheet conformation. They stain with Congo Red, and when stained and viewed under polarized light they exhibit a red/green birefringence, a property diagnostic for amyloid.
It has been proposed that the deposition of amyloid requires the formation of a nidus or protofilament around which amyloid fibrillogenesis takes place, and the glycosaminoglycan (GAG) heparan sulfate (HS) plays an important role in this pathological process (17,19,20). GAGs are sulfated heteropolysaccharides that have been known to be associated with amyloid for over 30 years (21, 22) but attracted little attention until 1987, when Snow and co-workers (23) showed that the GAG component of mouse AA amyloid was deposited coincidentally with the amyloid protein and that the GAG was part of the HS proteoglycan, perlecan (24,25). By experimentally varying the induction speed of murine AA-type amyloidosis, it was possible to show that HS was both temporally and spatially deposited with the AA peptide (24 -26) and that splenic perlecan mRNA was increased prior to the histological detection of AA amyloid (26). In vitro experiments have also shown that of a number of different GAGs examined, only HS could increase the ␤-sheet content (the characteristic conformation of amyloid) for mouse apoSAA2 but not for the nonamyloidogenic isoforms apoSAA1 and apoSAA CE/J (27,28).
HS has in fact been found in all amyloid deposits that have been investigated (19). High affinity binding between perlecan and three Alzheimer's amyloid (A␤) precursors, ␤PP-695, ␤PP-751, and ␤PP-770, could be inhibited with dextran sulfate and heparin but not chondroitin sulfate or dermatan sulfate (29). HS has also been found to enhance A␤ fibrillogenesis (30), and analogs of HS (aliphatic polysulfonates) were recently reported to block HS-induced A␤ fibrillogenesis, in vitro, and interfere with in vivo AA-type amyloid accumulation in mice (31,32). Congo Red (CR), a disulfonated acidic dye, which has long been used as a stain for amyloid (33), can inhibit AA amyloid in vivo (34). CR and a number of sulfated glycans have also been shown to prevent the accumulation of the protease-resistant prion protein (amyloid form) (35)(36)(37).
HS and closely related heparin are negatively charged polymers composed of disaccharide repeats that contain carboxyl and sulfate groups (38,39). Many proteins bind these GAGs through electrostatic interactions, and it has been demonstrated by substitution and chemical modification experiments that the protein binding is dependent on their positively charged (basic) residues (40 -42). Because a wide range of proteins bind specifically to heparin, it was expected that a common structural motif would be found. Cardin and Weintraub in 1989 (43) examined a series of heparin-binding sequences and found that the basic residues tended to be arranged on one side of an ␣-helix with the pattern XBBBXXBX (B, basic residue; X, nonbasic residues). A second pattern, XBBXBX, was proposed to align the basic residues on one side of a ␤-strand. Later, a third consensus sequence was published, XBBXXBBBXXBBX (44). However, there are many examples of heparin/HS-binding sequences lacking these consensus sequences, and recently Margalit et al. (45) reported that a common feature of heparin-binding sequences was that the outer basic residues were always 20 Å apart.
The co-localization of HS with amyloids, the induction of perlecan expression prior to the appearance of AA amyloid, its ability to promote conformational change in native amyloid precursors and amyloid fibrillogenesis in vitro, and the amyloid-blocking effects of HS analogs and other sulfonates are all consistent with the working hypothesis that HS-amyloid precursor associations take place in situ and are a critical early step leading to protofilament formation and/or amyloid fibrillogenesis. In an effort to understand the underlying mechanism of HS-dependent conversion of apoSAA into AA amyloid fibrils, we undertook this study to characterize the GAG binding activity of apoSAA. This was done by affinity chromatography with defined apoSAA peptides using columns to which different GAGs were covalently attached. The GAG-binding site on apoSAA was initially mapped testing the binding activ-ities of apoSAA CNBr fragments. This was followed with an assessment of the binding activities of a series of synthetic peptides corresponding to the smallest heparin/HS-binding CNBr fragment, containing specific residue substitutions, or a deletion. This approach allowed us not only to identify the heparin/HS-binding peptide sequence but also to rank the relative importance of the individual basic residues within the binding site. The demonstration of heparin/HS binding activity for apoSAA is consistent with some of the functions proposed for apoSAA. Characterization of the apoSAA-binding site also advances our understanding of amyloidogenesis and may assist in the design of therapeutic compounds.

EXPERIMENTAL PROCEDURES
Purification of Lipoproteins and ApoSAA Peptides-Plasma apoSAA concentrations were experimentally elevated in CD1 mice (Charles Rivers, Montreal, Quebec, Canada) by a subcutaneous injection of 0.5 ml of 2% (w/v) AgNO 3 (46) which resulted in a sterile abscess. After 18 -20 h, mice were sacrificed by CO 2 narcosis and exsanguinated by cardiac puncture preventing clotting with a small amount (50 l) of 7% EDTA. High density lipoprotein containing apoSAA (HDL-apoSAA) was isolated from plasma by density flotation (47). The density of the plasma was adjusted to 1.25 g/ml with NaBr and centrifuged at 250,000 ϫ g for 24 h at 10°C. The top layer was aspirated, pooled, and LDL and HDL-apoSAA were separated by gel filtration on a Sephacryl S-300 column (1.5 ϫ 45 cm) washed with 50 mM Tris, 150 mM NaCl, pH 7.5, at 20 ml/h. Normal HDL and LDL were also purified from uninflamed mice by this procedure.
ApoA-I and apoSAAs were purified from total lipoprotein by gel filtration as described previously (48). The lipoprotein preparations (3-5 ml) were dialyzed against 10% formic acid, pH 2.0 (2 liters), for 18 h at 4°C and then applied to a Sephacryl-S-100HR column (2.5 ϫ 110 cm) and eluted in the same buffer at 25 ml/h. The sample separated into two peaks which were collected, frozen in liquid N 2 , lyophilized, and stored at Ϫ20°C. The protein residue from the first peak containing apoA-I was delipidated with diethyl ether (49). The dried residue was resuspended in 4 M urea, 10% formic acid, and the apoA-I was purified by gel filtration on the Sephacryl S-100HR column.
Individual apoSAA isoforms were purified from the second gel filtration peak by reversed phase-high performance liquid chromatography (RP-HPLC) using a Waters (Millipore) HPLC system with a model 680 automated gradient controller, model 501 pump units, and a series 440 absorbance detector connected to a Waters 740 data module integrator. The lyophilized apoSAA powder was solubilized in 20% formic acid and then injected onto a semi-preparative C18 Vydac column (1 ϫ 25 cm), eluted at 3 ml/min with 0.1% trifluoroacetic acid, 10% acetonitrile for 5 min, and then developed with an acetonitrile concentration gradient, increasing 2.5%/min for 10 min followed by 1.0%/min for 20 min, and finally 4.3%/min for 10 min bringing the elution buffer to 98% acetonitrile by 45 min. The eluant was monitored continuously at 214 nm, and absorbance was plotted against retention time. Peaks were collected manually and dried by vacuum centrifugation. The dried residue was stored at Ϫ20°C. Apolipoproteins were identified by their mobility on SDS-urea-polyacrylamide gel electrophoresis (50).
Preparation of the ApoSAA Peptides-ApoSAA1 and apoSAA2 were cleaved at Met-X peptide bonds with cyanogen bromide (CNBr). Protein was dissolved in 70% formic acid at 1 mg/ml, to which 5.5 mg/ml CNBr was added (250 M excess over Met) plus free L-Trp (5 M excess over Met) to protect Trp residues. The reaction was carried out under nitrogen, at room temperature overnight, and then the solvent was evaporated by vacuum centrifugation. The reaction was evaluated, and peptides were purified by RP-HPLC. A semi-preparative C-18 Vydac column was equilibrated with 10% acetonitrile, 0.1% trifluoroacetic acid. Peptides were dissolved in 40% formic acid, filtered through a 0.2-m filter, or centrifuged at 10,000 ϫ g, and loaded onto the column, washed for 5 min at 3 ml/min, and then developed with a 1.5%/min acetonitrile linear concentration gradient for 30 min, followed by 4.3%/min acetonitrile for 10 min, bringing the elution buffer to 98% acetonitrile by 45 min. The separated peptides were identified by amino-terminal sequencing and molecular weight determination by mass spectroscopy carried out at the Alberta Peptide Institute (Edmonton, Alberta, Canada). Also, a series of apoSAA peptides corresponding to 1) mouse apoSAA2, residues 77-103 (m27-mer), with specific basic residues substituted with Ala, 2) mouse apoSAA2, residues 77-96 (a 20-mer missing Lys-102), and 3) human apoSAA2, residues 78 -104 (h27-mer), were synthesized by Multiple Peptide Systems (San Diego, CA).
Glycosaminoglycan, Congo Red, and Taurine Affinity Columns-Heparin/Affi-Gel media were purchased from Bio-Rad. Heparin, heparan sulfate, chondroitin sulfate, dermatan sulfate, and hyaluronan were all purchased from Sigma and coupled to either to Sepharose 4B (Amersham Pharmacia Biotech) based on the method of Smith et al. (51) or to Affi-Gel 102 (Bio-Rad) as per manufacturer's instructions. Sepharose 4B was washed with 20 bed volumes of water, resuspended in 1 volume of water, and transferred to a beaker with a stir bar on ice. GAGs were dissolved in water at 2 mg/ml and were also placed on ice. The two solutions were mixed, and while stirring the pH was adjusted to pH 11 with NaOH (5 N). Cyanogen bromide, 1 g/ml in N,N-dimethylformamide, was added dropwise to a final concentration of 31.3 mg/ ml. The pH was maintained at about 11 by adding NaOH for 15 min and then left stirring for 18 h at room temperature. The gel was then washed with 20 bed volumes of water followed by 1 M ethanolamine, pH 9.0, to block excess reactive groups. After further washing with 10 bed volumes of (i) water, (ii) 0.1 M sodium acetate, pH 4.0, and (iii) 0.1 M NaHCO 3 , pH 8.3, the column gel was equilibrated in 20 mM Tris-HCl, pH 7.5. Affi-Gel 102 (4 ml) was washed with 20 bed volumes of 50 mM acetate, pH 6.0, and then 8 mg of GAG in 4 ml of the same buffer was mixed with the gel. The coupling reaction was initiated by the addition of 32 mg of 1-ethyl-3-3-dimethylaminopropyl carbodiimide, adjusting the pH to 5 with 1 N HCl and allowing the reaction to proceed for 3 h. Heparin and HS-Sepharose 4B were also treated with carbodiimide that forms a stable adduct with GAG carboxyl side groups (52). The amount of GAG linked to columns ranged from 0.5 to 0.75 mg/ml gel as determined colorimetrically by the toluidine blue method (51). Taurine (2-aminoethanesulfonate) and Congo Red (CR) dye were also coupled to Sepharose 4B. Taurine at 30 mg/ml and CR at 5 mg/ml were reacted with the CNBr-activated matrix. Once the reaction was complete the CR-Sepharose 4B was washed with 10% ethanol followed by 2 M guanidine HCl, pH 7.5, to remove excess CR prior to equilibration with 20 mM Tris-HCl, pH 7.5. Based on the amount of CR that was washed from the column (absorbance at 480 nm), 3.8 mg of CR was coupled to the column. By including [ 3 H]taurine (15 Ci) with the cold taurine, the amount of taurine coupled to Sepharose 4B was estimated at 12.4 mg/ml by scintillography. The CR and taurine columns contained approximately equimolar amounts of sulfate groups.
Affinity Chromatography on GAG-charged Columns-ApoSAA CNBr cleavage products were resuspended in 20 mM Tris-HCl, pH 7.5, and loaded onto a 6-ml heparin-Affi-Gel column equilibrated with the same buffer at 0.5 ml/min. The column was then washed with 4 bed volumes of buffer and then developed with a 0 -1 M NaCl concentration gradient. Fractions (0.6 ml) were collected and their absorbance measured at 214 nm. Heparin/Affi-Gel and other GAG-charged matrices were also packed into a 3-ml stainless steel column and equilibrated with 20 mM Tris-HCl, pH 7.5, at 0.5 ml/min using a Waters HPLC system. Samples (20 -80 g in 150 -200 l) were injected onto the column, washed with 3 bed volumes (18 min) of the same buffer, and then developed with a 0 -1.0 M NaCl linear gradient for 10 bed volumes (60 min). The eluate was monitored continuously at 214 nm, and the absorbance was plotted against the retention time (RT). Generally, unbound peptides/proteins eluted 6.5-7.0 min after loading, and based on the RTs for the bound peptides/proteins, the NaCl concentration at which desorption took place could be calculated as follows: desorption [NaCl] ϭ (RT Ϫ 6.5 min Ϫ 18 min)/60 min.

Cyanogen Bromide Cleavage of RP-HPLC-purified ApoSAA1
and ApoSAA2-The acute-phase serum amyloid A isoforms, apoSAA1 and apoSAA2, of mouse are 91% identical in amino acid sequence and have a single heparin-binding consensus sequence (XBBXBX, X, non-basic residue; B, basic residue) (43) located between residues 82 and 87 near the carboxyl terminus (Fig. 1A). A second potential GAG-binding sequence, rich in basic residues, is located between residues 18 and 46. Unfortunately, the direct testing of heparin binding activity of apoSAA was hampered by the insolubility of delipidated apoSAA under physiological buffer conditions in the absence of chaotropic agents (urea, SDS, and CHAPS). But two Met residues at positions 16 and 23 facilitated the removal of the amino-terminal amphipathic lipid binding domain (residues 1-24) (53) by CNBr cleavage (Fig. 1A). For apoSAA1 this reaction generated an insoluble 16-mer (apoSAA 1-16 ), and two soluble fragments, a 7-mer (apoSAA1 [17][18][19][20][21][22][23] ) and an 80-mer (apoSAA1 24 -103 ) which constituted about 78% of the native protein. For apoSAA2 a substitution of Ile with Met at residue 76 introduces an additional cleavage site allowing the 80-mer to be cleaved into a 53-mer (apoSAA2 24 -76 ) and a 27-mer (apoSAA2 77-103 ) (Fig. 1A). The GAG-binding consensus sequence was predicted to lie within the 27-mer. Both apoSAA1 and apoSAA2 were purified (Fig. 1B), and cleavage of the individual isoforms with CNBr yielded the expected fragments, with a minimum of secondary by-products (Fig. 1, C and D). The identity of the CNBr-generated peptides was determined by amino-terminal sequencing and molecular weight analysis by mass spectrometry (data not shown).
The Heparin-binding ApoSAA CNBr Peptides-Preliminary chromatographic analysis of the apoSAA1 and apoSAA2 CNBr cleavage products on a heparin/Affi-Gel column (6 ml) revealed that both preparations contained heparin binding activity (Fig.  2). The bound peptides were eluted from the column by increasing the NaCl concentration, suggesting that the association was primarily electrostatic in nature. The interactions involved only heparin since binding was not detected with the uncharged agarose Affi-Gel matrix alone (data not shown). Amino-terminal sequencing of the heparin-bound peptides repurified by RP-HPLC identified the peptides as apoSAA1 24 -103 (m80-mer) and apoSAA2 77-103 (m27-mer). The second potential GAG binding region contained on the m53-mer did not to have heparin binding activity. Further analysis of GAG binding activities was performed on matrices produced by covalently coupling different GAGs to Affi-Gel or Sepharose 4B and analyzed on a Waters HPLC apparatus. The two different coupling reactions linking the GAGs through their accessible carboxyl (Affi-Gel) or hydroxyl/amino groups (CNBr-Sepharose 4B) allowed us to sterically block different GAG chain motifs. We found that heparin-Sepharose 4B columns gave better resolution than the commercially available heparin/Affi-Gel or the heparin/Affi-Gel 102 we generated. The use of an HPLC apparatus also allowed for precise solvent delivery and gradient formation, resulting in highly reproducible retention times (Ϯ1-2%) and accurate determination of NaCl concentrations at which peptide desorption took place.
The Basic Amino Acid Residues Required for Heparin Binding-Since heparin binding activity was detected for the CNBr 27-mer peptide, we opted to identify the basic residues required for binding using wild type and mutant synthetic peptides corresponding to this 27-mer. Although our interest was initially drawn to the GAG-binding consensus sequence, an alignment of the mouse apoSAA sequence with 29 other apoSAAs from 12 different species indicated that the consensus sequence was not conserved (Fig. 3A). Four of the basic residues on this heparin-binding fragment were conserved, but only one was part of the consensus sequence. Therefore, the role of all the basic residues within this 27-mer was examined using nine synthetic peptides, two wild type 27-mers corresponding to both the mouse (m27-mer) and human apoSAA2 (h27-mer) and seven different mutant peptides, five 27-mers in which basic residues, Arg-83, His-84, Arg-86, Lys-89, and Arg-95, were replaced individually by Ala, one 27-mer in which Arg-83, His-84, and Arg-86 were substituted simultaneously with Ala, and a 20-mer missing the carboxyl-terminal Lys-102 (Fig. 3B).
The synthetic version of the apoSAA2 77-103 peptide (m27mer) bound heparin-Sepharose 4B desorbing at 0.27 M NaCl (RT ϭ 40.4 min) (Fig. 4A). Although the columns were normally washed with 3 bed volumes before starting the NaCl gradient, washing for up to 20 bed volumes did not elute the peptide (data not shown) indicating that it was immobilized with a partition coefficient (␣) of Ն0.95, where the bed volume ϭ 1/1 Ϫ ␣ (54). Purified apoSAA1 m80-mer peptide was also retained by the column desorbing at 0.28 M NaCl (RT ϭ 41.0 min) (Fig. 4A). The wild type peptides did not have an affinity for the agarose matrix alone, Affi-Gel 102, or CNBr-activated Sepharose-4B blocked with ethanolamine. As additional controls, carbonic anhydrase and bovine serum albumin lacked binding activity for the heparin columns, whereas fibronectin and LDL, known heparin-binding proteins, both bound the heparin columns. The binding characteristics of all the synthetic peptides were also tested on heparin/Affi-Gel and found to be similar to heparin-Sepharose 4B with the exception of the Lys-102 Ϫ peptide (described below).
The first series of mutant peptides were designed to investigated the Cardin and Weintraub (43) proposed GAG-binding consensus sequence XBBXBX, residues 82-87 of apoSAA (Fig.  4A). Of the individual replacements, R86A showed the greatest reduction in heparin-Sepharose 4B binding, eluting at 12.4 min (0 M NaCl), followed by R83A which eluted at 17.4 min (0 M NaCl). H84A reduced heparin binding only slightly (RT ϭ 38.1 min, 0.24 M NaCl), and these replacements in combination (R83A/H84A/R86A) had a cumulative effect on the inhibition of binding (RT ϭ 7.1 min). The second series of mutant peptides tested the involvement of Lys-89, Arg-95, and Lys-102 for heparin binding (Fig. 4B). Peptides K89A and R95A eluted from heparin-Sepharose at 16.6 (0 M NaCl) and 24.0 min (0 M NaCl), respectively, indicating impaired binding activity. Deletion of the carboxyl-terminal residue Lys-102 (Lys-102Ϫ) caused only a minor reduction in heparin binding activity with the Lys-102 Ϫ peptide eluting at 37.9 min (0.23 M NaCl). Human apoSAA2 78 -104 (h27-mer), which lacks a typical GAG-binding consensus sequence, bound heparin with an avidity comparable to that seen with the m27-mer (RT ϭ 39.5 min, 0.26 M NaCl). HDL-apoSAA, isolated from acute-phase mouse plasma, did not appear to have any affinity for heparin (Fig. 4B) suggesting that the heparin-binding site was cryptic on HDL-associated apoSAA. By omitting NaCl in the starting buffer, it was possible to rank the relative importance of all the basic residues of  (40.4 min) . The first four mutant peptides had very low affinity for heparin eluting at 0 M NaCl. When the carboxyl-linked heparin column (heparin-Affi-Gel) was used a similar hierarchy was observed (data not shown), but Lys-102 became more important for binding, with Lys-102 Ϫ desorbing at 0.047 M NaCl (RT ϭ 26.8 min) versus 0.23 M NaCl (RT ϭ 37.9 min) on heparin-Sepharose 4B (Fig. 4B).
GAG Binding Specificity of ApoSAA-The binding specificity of the 27-mers was further investigated by affinity chromatography on a series of columns charged with either heparan sulfate (HS), chondroitin sulfate (CS), dermatan sulfate (DS), or hyaluronan (HA). Binding of the wild type m27-mer (Fig. 5A) and h27-mer (Fig. 5B) was not detected with CS, DS, or HA columns. However, the 27-mers and the m80-mer peptide bound to HS-Sepharose 4B, desorbing at approximately the same ionic strength as that seen with heparin-Sepharose 4B. The Lys-102 Ϫ peptide was only slightly retained by HS (RT ϭ 16.5 min), suggesting that the Lys-102 residue contributed significantly to apoSAA-HS interactions. Since carboxyl-linked heparin showed a reduced avidity for Lys-102 Ϫ , and HS has one-third the sulfate content of heparin and thus a lower net charge, we were interested to see if the carboxyl groups of HS played a more prominent role than those of heparin in apoSAA interactions. Little or no binding was detected between HS/ Affi-Gel and the m27-mer, Lys-102 Ϫ , and h27-mer peptides. However, fibronectin (Fig. 5B) and LDL (not shown) still bound to this matrix, indicating that only the HS domain(s) containing the apoSAA-binding site(s) were blocked. The importance of the HS-carboxyl groups for apoSAA binding was further demonstrated when binding was destroyed after blocking carboxyl groups on HS-Sepharose 4B with 32 mg of carbodiimide, equivalent to the amount used in the Affi-Gel coupling reaction (Fig. 5B).
Hierarchy of Mutant Peptide Binding to HS-The m27-mer mutant peptides were also tested for HS binding (elution profiles not shown), and their relative importance was compared with that for heparin binding (Fig. 6). Each individual substi-tution/deletion of the basic residues, except His-84, had a greater impact on binding to HS than to heparin. The net reduction in RT for HS-Sepharose 4B ranged from 28.2 min for R86A to 23.6 min for Lys-102 Ϫ . For heparin-Sepharose the range was greater, 28.0 min for R86A to 2.5 min for Lys-102 Ϫ . The Lys-102 and Arg-95 residues appeared to be much more important for HS than for heparin binding.
Sulfonate Spacing Is Important for ApoSAA Binding-Taurine (2-aminoethylsulfonate), a cysteine catabolite, and Congo Red (sodium diphenyldiazo-bis-␣-naphthylaminesulfonate, CR), an acidic diazo dye used to stain amyloid deposits, are mono-and disulfonate compounds, respectively. The sulfonate groups in CR are spaced 19.3 Å apart (Fig. 7A). Both taurine and CR also have amino groups which facilitated their coupling to Sepharose 4B and oriented their sulfonate groups away from the matrix surface. For taurine-Sepharose 4B, the sulfonates would be randomly distributed, but for CR-Sepharose 4B they would be organized in regularly spaced pairs. m27-mer binding was only detected for CR-Sepharose and was retained through electrostatic interactions since desorption required increased NaCl concentration (Fig. 7B). Neither the Lys-102 Ϫ nor the R83A/H84A/R86A triple mutant peptides bound to CR-Sepharose, consistent with the CR-m27-mer interaction being de-  3. Protein sequence alignment of the carboxyl-terminal 27 residues for apoSAAs. A, the sequences for the carboxyl-terminal 27 residues for the apoSAAs were retrieved from the GenBank TM data base (National Center of Biotechnology Information) and compared with that of mouse apoSAA2 (m27-mer); the conserved basic residues are shown in bold; dashes represent residues identical with the m27mer, and the GAG-binding consensus sequence is shaded. B, synthetic peptides used to characterize the GAG-binding site, with the Ala substitutions underlined.
pendent on basic residues. This also indicated that the spacing of the anionic groups on CR was important for binding and possibly mimics the sulfate spacing in the apoSAA-binding motif on HS.
GAG-HDL/HDL-ApoSAA Interactions; ApoA-I Binds HS-HDL-apoSAA from inflamed mouse plasma did not demonstrate any heparin binding activity which suggested that the heparin-binding site is inactive when apoSAA is associated with HDL (Fig. 8). However, HDL-apoSAA did bind to HS-Sepharose 4B and HS/Affi-Gel columns. To determine if apoSAA was solely responsible for HDL-apoSAA-HS binding HDL from normal plasma, lacking acute-phase apoSAA, was purified and tested for heparin/HS binding. Its binding pattern was found to be identical to that of acute-phase HDL-apoSAA, suggesting that in addition to apoSAA, other HDL apoproteins could also contribute to the HDL-apoSAA-HS interactions (Fig.  8). ApoA-I, is the most abundant apoprotein of HDL, and although apoSAA can comprise up to 80% of the apoproteins on acute-phase HDL-apoSAA, the remainder is mostly apoA-I (55). ApoE is known to bind heparin/HS, but as a minor component of both normal and acute-phase HDL-apoSAA (Ͻ1% of HDL 3 apoproteins) (56), it would be found on a small subfraction of HDL and is unlikely to be responsible for such a high proportion of the HDL binding to the heparin/HS columns.  [77][78][79][80][81][82][83][84][85][86][87][88][89][90][91][92][93][94][95][96] ) which lacked the carboxyl-terminal Lys-102, and the h27-mer (human apoSAA2 78 -104 ). The elution profile of the Lys-102 Ϫ peptide on heparin/ Affi-Gel is also shown. All peptides were chromatographed separately, and their corresponding peaks are labeled on the graph. Peptides (25-50 g) were applied to a heparin-Sepharose 4B column (3 ml), pre-equilibrated with 20 mM Tris-HCl, pH 7.5, at 0.5 ml/min with continuous monitoring of the eluant at 214 nm. After washing with 3 bed volumes, the column was develop with a 0 -1 M NaCl linear concentration gradient (10 bed volumes). The m80-mer was generated by CNBr cleavage of apoSAA1 and purified by RP-HPLC. The peptides were run separately, and composite representative elution profiles are shown. The minor peaks present between 35 and 40 min may represent small amounts of impurities present in the synthetic peptides. Hence, apoA-I was purified, its GAG binding activity was tested and found to parallel that of HDL's, binding specifically to HS, and not to heparin or chondroitin sulfate (data not shown). These results demonstrate that HDL-HS and acutephase HDL-apoSAA-HS interactions could be mediated through apoA-I, with apoSAA possibly contributing to the latter interaction.

DISCUSSION
Heparin and heparan sulfate (HS) belong to a family of linear heteropolysaccharides called glycosaminoglycans (GAGs), which are normally synthesized linked to a protein core (proteoglycans) (38,39). The GAG chains consist of up to a hundred disaccharide repeats composed of a hexuronic acid and a hexosamine. The hexuronic acid is either ␤-D-glucuronic (GlcUA) acid or the C-5 epimer ␣-L-iduronic acid (IdceA), and the hexosamine is glucosamine (GlcN), with N-and O-sulfation at various positions in the disaccharide repeats. During its biosynthesis, heparin is extensively modified by a C-5 epimerase and three different sulfotransferases, producing a mature chain in which the hexuronic acid is Ն80% IdceA and is almost completely sulfated, with 2 to 3 sulfates per disaccharide unit. By comparison, HS is less extensively modified (IdceA Յ50%) containing a more varied structure with regions of high and low sulfation, 2 to 0.2 sulfates per disaccharide. Their distribution in tissue is also very different, with heparin being confined to cytoplasmic granules of mast cells, whereas HS is found ubiquitously on cell surfaces and within the extracellular matrix. Given the limited anatomic distribution of heparin, binding studies with HS may be the more relevant.
Due in part to their net negative charge, heparin/HS can associate with a variety of proteins (57). These interactions have profound physiologic and pathologic importance since they modulate the function of numerous proteins such as growth factors, serine protease inhibitors, extracellular matrix proteins, plasma lipoproteins, and lipolytic enzymes. Among the apolipoproteins, apoE and apoB have well characterized heparin/HS binding activities (58 -61). To this list we can add two other apolipoproteins, apoSAA and apoA-I. By affinity chromatography of CNBr fragments and synthetic mutant peptides of apoSAA, we were able to map the heparin/HS-binding site to the carboxyl-terminal end for both mouse (residues 77-103) and human (residues 78 -104) apoSAA. The basic residues, Arg-83, Arg-86, Lys-89, and Arg-95 of the mouse peptide  8. HDL, HDL-apoSAA (acute-phase), and apoA-I binding activities on heparin-Sepharose 4B and HS-Sepharose 4B were investigated and compared with the mouse apoSAA m80-mer peptide. HDL (asterisks) was found to have very similar elution profile to HDL-apoSAA and for simplicity is not shown here.

FIG. 6. A comparison of the relative importance of the basic residues of the m27-mer peptide for heparin and HS binding.
The same series of peptides as in Fig. 4 were analyzed on an HS-Sepharose 4B column, and the net reduction in RTs (wild type m27-mer RT minus mutant peptide RTs) was plotted comparing heparin and HS-Sepharose 4B columns. The sequence of m27-mer is shown above the graph. Open bars, heparin-sepharose; closed bars, heparan sulfate-sepharose.
(m27-mer) were found to be necessary to maintain full heparin binding activity, and for HS binding Lys-102 residue was also required. The binding hierarchy of the basic residues differed between heparin and HS possibly reflecting differences in GAG ligand structure. Differences in basic residue composition and spacing between heparin and HS binding sequences have also been observed by others (62)(63)(64). Heparin-binding sequences tend to be richer in Arg residues which are arranged in clusters. Arg residues have been shown to have a higher affinity than Lys residues for heparin and sulfate groups. For HSbinding sites, Arg residues are less common, and basic residues are generally more widely spaced (64).
The other major GAGs (CS, DS, or HA) normally found in tissue showed no detectable affinity for the apoSAA peptides. These data not only showed that the GAG binding activity of apoSAA was specific for heparin/HS but also provided clues as to the nature of the apoSAA-binding motif on HS. The structures of DS and HA both contain elements resembling HS. HA is composed of unsulfated GlcUA/GlcN units which are similar to a large portion of the unmodified HS chain. The lack of m27-mer-HA binding suggests that heparin/HS binding is probably dependent on IdceA and/or sulfates. DS and CS have a higher sulfate content than HS suggesting that apoSAA-HS binding was not dependent exclusively on the net cationic charge of HS and that the appropriate spacing of the carboxylates and sulfates is probably necessary.
The binding activities of heparin and HS were influenced by the type of coupling reaction used to link them to the column matrix, i.e. carboxyl linkage on Affi-Gel versus hydroxyl/amino linkage on Sepharose 4B. For heparin/Affi-Gel only the binding avidity of the Lys-102 Ϫ peptide was reduced; however, with HS/Affi-Gel neither the wild type m27-mer nor the Lys-102 Ϫ bound possibly because of a higher dependence on HS carboxyls for binding. This interpretation was supported by the observation that binding was inhibited after treating HS-Sepharose with carbodiimide which blocks carboxyls (52). Alternatively, the differences in binding between heparin/Affi-Gel and HS/ Affi-Gel may be explained by preferential coupling of the IdceA C-5 carboxyl to the column matrix. If the apoSAA-binding site on heparin and HS is located in a region rich in IdceA, then heparin with its greater IdceA content would be predicted to have a greater number of binding sites. Hence the apoSAAbinding sites on heparin/Affi-Gel may be less likely to be sterically blocked than those on HS/Affi-Gel. In order to understand better this difference in binding affinities, identification of the apoSAA-binding oligosaccharide on both heparin and HS is required.
An influential paper by Cardin and Weintraub in 1989 (43) identified two consensus sequences, XBBBXXBBX and XB-BXBX (X, hydropathic; B, basic residues) by comparing 12 heparin-binding sequences from a series of heparin-binding proteins. Later a third consensus sequence was reported by another group (XBBXXBBBXXBBX) (44). For mouse apoSAA an XBBXBX sequence at residues 82-87 initially attracted our attention; however, inspection of other available apoSAA sequences (from 12 species) revealed that this consensus sequence was not conserved. There are many examples of heparin/HS-binding protein or peptides that also lack these apparent heparin-binding "consensus" sequences including antithrombin III, residues 124 -145 (65), platelet factor 4, residues 46 -70 (66), apoE, residues 202-242 (58), and basic fibroblast growth factor, residues 25-46 and 111-141 (67). This would suggest that the primary sequence alone cannot define the heparin/HS-binding site. Protein conformation may also play a role in placing critical basic residues into energetically favorable positions juxtapositioning them with cationic groups on the GAG chain. In support of this view, Margalit and colleagues (45) have found that by comparing the spatial distribution of the basic residues for 18 known heparin binding domains, for which three-dimensional structures were available, two basic residues were always about 20 Å apart (20.3-23.5Å). In binding sites with an ␣-helical or ␤-strand conformation, the two basic residues were separated by 13 and 7 residues, respectively, without a discernible consensus sequence. Furthermore, it was demonstrated that a distance of 20 Å between the two outermost basic residues of a heparinbinding site could accommodate docking of the binding site to a heparin pentasaccharide, the size sequence of heparin that binds to antithrombin III (68).
The secondary structure for apoSAA is unknown, however, based on the spacing of basic residues observed by Margalit et al. (45), if the m27-mer conformation were in an ␣-helix then Lys-89/Lys-102 are at the correct distance apart (13 residues) and if in a ␤-strand then Arg-95/Lys-102 would have the appropriate spacing (7 residues). None of the four basic residues absolutely required for heparin binding (Arg-83, Arg-86, Lys-89, and Arg-95) fit the 7 (␤-strand) and 13 (␣-helix) spacing pattern. Interestingly, an His or Arg residue is found at position 93 in all species, except mouse, and it is possible that Arg/His-93 functionally replaces the mouse Arg-83. The Arg-86/Arg-93 or Arg-86/His-93 are separated by seven residues and could be spaced correctly on a ␤-strand.
The binding characteristics of the apoSAA m27-mer with Congo Red (CR) seems to agree with the findings of Margalit and co-workers (45). CR is an acid diazo-disulfonate dye used as a diagnostic stain for amyloid deposits (33) and has been reported to binding amyloid precursors and block fibrillogenesis (34 -37). We found that the m27-mer peptide bound through ionic interactions to CR-Sepharose 4B. Peptides missing the basic residues at either end of the m27-mer (Arg-83, Arg-86, or Lys-102) lacked any CR binding activity. As a control, taurine, a monosulfonate, linked to Sepharose 4B had little or no affinity for m27-mer. A major difference between these two columns was that with taurine-Sepharose 4B, the sulfonates were randomly distributed, whereas for CR the sulfonates were organized into pairs 19.3 Å apart (69).
From a pathological perspective, the apoSAA-HS binding activity is consistent with a view that most or all amyloid types are formed by an initial fibrillogenic amyloid precursor-HS interaction (17,19,20). It may also explain a number of previously reported observations concerning AA amyloidogenesis, namely the co-localization of HS with AA amyloid fibrils, the HS-induced ␤-sheet conformational change in apoSAA2 (27), and the inhibition of AA amyloid by alipathic sulfonates designed to simulate HS (31). In addition to apoSAA and apoA-I, a mutant form of which can deposit as amyloid in the peripheral nervous system (familial polyneuropathy) (70), heparin/HS binding activity has been demonstrated for three other amyloid precursors, A␤ and ␤PP (29,71), prions (72,73) and amylin (74,75).
If the linear distance between the two outer basic residues of most HS-binding sites are arranged in 20-Å intervals, as Margalit and co-workers suggest (45), this information may prove useful in the design of anti-amyloid drugs that would be effective for a variety of amyloids. The effectiveness of CR in binding to the apoSAA heparin/HS-binding site and preventing both AA (34) and prion amyloid deposition in vivo (35)(36)(37) may be due to the spacing of its sulfonates. This agent which potentially aligns with two basic residues within the binding site allows electrostatic interactions to take place blocking potential amyloidogenic interactions with HS. Further support for this idea comes from a recent report in which the efficacy of different polysulfonates for preventing AA amyloid was tested (31). Of the polysulfonates examined, polyvinylsulfonate was found to be the most effective at preventing AA amyloid formation in vivo. Computer modeling of the energy-minimized structure for the repetitive subunit of polyvinylsulfonate predicts the spacing of the sulfonates to be about 5.4 Å, placing every fourth sulfonate approximately 20 Å apart (76).
To our knowledge HDL, or its major apoprotein apoA-I, has not previously been reported to bind HS. The physiological significance of the HDL-HS interaction is unclear at present. For low density and remnant lipoproteins, HS proteoglycan facilitates their binding to the cell surface, contributing to their processing by lipoprotein lipase and hepatic lipase, and mediates receptor-lipoprotein internalization (77,78). Cholesterol efflux is believed to require HDL contact with the cell surface (79), and candidate protein receptors for apoA-I have recently been identified (80 -82). The hepatic lipase-mediated uptake of HDL by hepatocytes has also been reported to be dependent on HS proteoglycan (83), and binding of both HDL and acutephase HDL-apoSAA to macrophage could be destroyed by the removal of surface HS with heparinase (84). Therefore, HDL may associate with cell surfaces through an apoA-I-HS interaction possibly mediating its cholesterol efflux activity or clearance from plasma. The physiologic significance of the apoSAAheparin interactions is also at present unknown. Most heparinbinding proteins can also bind HS, which is believed to be the natural ligand in most cases (85). Hence, the value of heparin may be limited to locating HS-binding sites and may not be the best ligand for the detailed molecular analysis of HS-binding sites. Interestingly, HDL-apoSAA did not bind heparin suggesting the binding site is masked when apoSAA is associated with lipid, a situation that has also been observed for one of the heparin-binding sites on apoE, residues 192-243 (86).
There is evidence that apoSAA-HS interactions maybe involved in a number functions proposed for apoSAA. The HDL binding capacity of macrophage has been shown to increase in the presence of apoSAA, possibly for the purpose of redirecting HDL-dependent reverse cholesterol transport during the acutephase response (12,13). Saturable, high affinity interactions (K d ϭ 8 nM) between human apoSAA1␣ and neutrophils have also been detected and may be responsible for the apoSAAmediated inhibition of the oxidative burst response by activated neutrophils (87,88). The binding to neutrophils could be blocked with a carboxyl-terminal peptide of human apoSAA1␣, residues 77-104, a region that shows high homology to the corresponding human apoSAA2 (h27-mer) peptide exhibiting HS binding activity. The apoSAA1␣ peptide is also missing an Arg or Lys at position 83 (the residue present in the mouse apoSAA2 GAG-binding consensus sequence) which confirms that Arg-83 is not required for HS binding of non-mouse apoSAAs. Human apoSAA1␣ has also been reported to have monocyte and leukocyte chemoattracting activity but only after its release from the HDL possibly by limited proteolysis near the site of injury (89). This activity could be mediated by apoSAA peptide-HS interaction on cell surfaces and the extracellular matrix. Although the precise function of apoSAA has not yet been demonstrated, the identification of an evolutionarily conserved HS binding domain suggests that apoSAA-HS interactions on cell surfaces plays a role in the execution of apoSAA function.