Design of peptides with high affinities for heparin and endothelial cell proteoglycans.

Proteoglycan-binding peptides were designed based on consensus sequences in heparin-binding proteins: XBBXBX and XBBBXXBX, where X and B are hydropathic and basic residues, respectively. Initial peptide constructs included (AKKARA)(n) and (ARKKAAKA)(n) (n = 1-6). Affinity coelectrophoresis revealed that low M(r) peptides (600-1,300) had no affinities for low M(r) heparin, but higher M(r) peptides (2,000-3,500) exhibited significant affinities (K(d) congruent with 50-150 nM), which increased with peptide M(r). Affinity was strongest when sequence arrays were contiguous and alanines and arginines occupied hydropathic and basic positions, but inclusion of prolines was disruptive. A peptide including a single consensus sequence of the serglycin proteoglycan core protein bound heparin strongly (K(d) congruent with 200 nM), likely owing to dimerization through cysteine-cysteine linkages. Circular dichroism showed that high affinity heparin-binding peptides converted from a charged coil to an alpha-helix upon heparin addition, whereas weak heparin-binding peptides did not. Higher M(r) peptides exhibited high affinities for total endothelial cell proteoglycans (K(d) congruent with 300 nM), and approximately 4-fold weaker affinities for their free glycosaminoglycan chains. Thus, peptides including concatamers of heparin-binding consensus sequences may exhibit strong affinities for heparin and proteoglycans. Such peptides may be applicable in promoting cell-substratum adhesion or in the design of drugs targeted to proteoglycan-containing cell surfaces and extracellular matrices.

Proteoglycans (PGs) 1 are composed of a core protein to which are covalently attached one or more sulfated glycosaminoglycans (GAGs). PGs are ubiquitous components of cell surfaces and the extracellular matrix, and their GAG chains contribute to myriad biological functions, such as modulation of enzyme activities, regulation of cell growth, and control of assembly of the extracellular matrix (1). PGs are thus potential targets for therapeutic intervention. For example, heparin antagonists are needed to take the place of protamine, a heterogeneous, sometimes toxic protein mixture commonly used to neutralize the anticoagulant activity of heparin in humans (2,3); in the design of drugs to be targeted to PG-rich tissues, such as cartilage (4); and to be used to promote cell adhesion in a variety of situations, e.g. by promoting binding of cells that express abundant amounts of PGs, such as endothelial cells (5), to synthetic vascular graft surfaces. To develop a rationale for the design of such reagents, it is useful to examine known features of protein structure required for high affinity interactions with GAGs. Thus, analysis of the structural features of many known heparin-and heparan sulfate (HS)-binding proteins has revealed the presence of conserved motifs, through which GAG binding has been postulated to occur. Cardin and Weintraub (6) identified two clusters of basic charge in known heparin-binding proteins in which amino acids tend to be arranged in the patterns XBBXBX or XBBBXXBX, where B represents an amino acid with basic charge, usually arginine or lysine, and X represents an uncharged or hydrophobic amino acid. Molecular modeling of these consensus sites predicts the arrangement of amino acids into either ␣-helices or ␤-strands. This allows for the clustering of noncontiguous basic amino acids on one side of the helix, thus forming a charged domain to which GAGs could bind. Indeed, for some but not all of the heparin-binding proteins, disruption of the heparin-binding consensus sequences hinders heparin binding. For example, chemical modification of the heparin-binding consensus site in thrombospondin (7) or site-directed mutagenesis of a heparin-binding sequence in fibronectin (8) eliminates or diminishes heparin binding affinity. Others have proposed a necessary distance of approximately 20 Å between basic amino acids for heparin binding, regardless of protein tertiary structure (9).
To date, few attempts have been made to use these concepts regarding the structural specificities of GAG-protein interactions to develop families of high affinity GAG-or PG-binding peptides. Thus, here we describe the design and characterization of high affinity heparin-and EC PG-binding peptides that were modeled from the proposed heparin-binding consensus sequences of native heparin-binding proteins. tinal mucosa (Sigma) was tyramine end-labeled and radiolabeled with Na 125 I (Amersham Pharmacia Biotech) to an average specific activity of ϳ1.0 ϫ 10 7 cpm/g as described (10). Radiolabeled heparin was fractionated on Sephadex G-100 (Bio-Rad), and the final ϳ12% of material to elute was retained as the low M r material of Յ6000 (11,12).
Electrophoretic Analysis of Binding of Heparin and Human Umbilical Vein Endothelial Cell (HUVEC) PGs to Peptides-Binding of radiolabeled heparin and HUVEC PGs to peptides was studied by ACE as detailed elsewhere (13), because the heparin-protein binding affinities revealed by ACE match reasonably well with those obtained by other well established quantitative techniques for measuring binding interactions, e.g. (14 -17). Briefly, peptides were dissolved in 1ϫ ACE running buffer, 50 mM MOPSO (Sigma)/125 mM sodium acetate, pH 7.0, and serially diluted in running buffer at 2ϫ concentrations. Peptides were then mixed 1:1 with 2% agarose/1% CHAPS (Roche Molecular Biochemicals) and loaded into wells of a 1% agarose gel. Radiolabeled heparin or HUVEC PGs were then loaded in a slot on the anode side of the gel and electrophoresed through the peptide-containing wells, toward the cathode. Gels were dried, and PG mobility was measured with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA) by scanning each protein lane and determining the relative radioactivity content per 88-m pixel through the length of the lane. Retardation coefficient (R) measurements, binding isotherm curve fittings, and apparent K d value determinations were calculated as detailed previously (10,13).
Some peptides were also analyzed by ACE for heparin binding under reducing conditions. Thus, after serial peptide dilution, ␤-mercaptoethanol was added at 5% to each peptide sample, and these were mixed 1:1 with 2% agarose/1% CHAPS/5% ␤-mercaptoethanol and added to the ACE gel sample wells as usual.
Binding analysis of peptides to enzymatically or chemically degraded PGs (see below) was carried out by ACE as detailed, except that PG samples included 6 M urea to denature any residual enzymes.
Radiolabeling and Isolation of Total HUVEC PGs and GAGs-Exponentially growing, subconfluent HUVECs were labeled with 35 Ci/ml [ 35 S]Na 2 SO 4 (ICN Pharmaceuticals, Costa Mesa, CA) in normal culture medium minus heparin for 12 h. Culture medium and cell layers were harvested separately. After removal of the medium, cells were washed with 2.0 ml of phosphate-buffered saline plus Ca 2ϩ -Mg 2ϩ . Medium and rinses were pooled and brought to 6 M urea, 10 mM EDTA, 1 mM phenymethylsulfonyl fluoride, 5 mM N-ethylmaleimide, 50 mM 6-aminocaproic acid, 5 mM benzamidine, and 1 g/ml pepstatin A. Samples were stirred for 15 min at room temperature and then centrifuged at 10,000 rpm for 30 min to remove insoluble materials.
To the cell layer was added 2.0 ml of extraction solution, 6 M urea, 100 mM NaCl, 0.2% Triton X-100, 50 mM Tris-HCl, pH 7.0, and protease inhibitors as described above. Cells were scraped off the dishes, and the extracts were pooled and stirred for 5 min at room temperature and then centrifuged as described above.
Samples were concentrated on DEAE columns (DEAE Bio-Gel A agarose, Bio-Rad) equilibrated with low salt buffer, 0.1 M NaCl, 6 M urea, and 50 mM Tris-HCl, pH 7.0. Columns were rinsed with 10 ml of low salt buffer; flow-through was discarded. Bound PGs/GAGs were eluted with 3 ml of high salt buffer, 1.5 M NaCl, 6 M urea, and 50 mM Tris-HCl, pH 7.0. Eluted samples were dialyzed against distilled water, lyophilized, and stored at Ϫ20°C until binding analysis.
Enzymatic Digestions of PGs-The contributions of PG GAG chain components to peptide binding affinities were assessed by selective enzymatic degradation of GAG chains prior to ACE analysis. [ 35 S]Na 2 SO 4 -labeled HUVEC PGs were digested with chondroitinase ABC or heparatinase I (Seikagaku America, Ijamsville, MD). PG samples were resuspended in 100 l of enzyme buffer (chondroitinase buffer: 50 mM Tris-HCl, 30 mM sodium acetate, pH 8.0, 0.1 mM pepstatin A, 0.5 mg/ml bovine serum albumin, 10 mM N-ethylmaleimide, 1 mM phenylmethylsulfonyl fluoride, and 5 mM EDTA; heparatinase buffer: 50 mM Tris-HCl, 5 mM calcium acetate, pH 7.0, 0.5 mg/ml bovine serum albumin, and 1 mM phenylmethylsulfonyl fluoride). Samples were digested with 0.05 units/ml chondroitinase ABC at 37°C for 3 h. Fresh enzyme was then added to the same concentration, and the incubation was continued for an additional 1 h. Heparatinase I was added to samples at 0.01 units/ml. Samples were incubated for 3 h at 43°C, fresh enzyme was then added to the same concentration, and incubation was continued for an additional 1 h. All samples were then stored at Ϫ20°C until binding analysis.
Chemical Degradation of PGs-The contributions of PG GAG chain components to peptide binding affinities were further assessed by selective chemical degradation. Total secreted HUVEC PGs were subjected to nitrous acid degradation as detailed (20), which selectively degrades HS GAG chains. Binding analysis to peptides was then measured by ACE.
␤-Elimination of PGs-GAG chains were released from PG core proteins by alkaline borohydride reduction as detailed in Ref. 21.
Circular Dichroism Spectroscopy-CD spectra were recorded at 22°C using a JASCO J-500C spectropolarimeter interfaced to a 486 PC. The path length of the CD cells was 0.5 mm, and the CD was expressed in terms of ellipticity [⌰] in degree⅐cm 2 ⅐dmol Ϫ1 . Samples were initially dialyzed to water to remove residual synthesis contaminants, lyophilized, and resuspended in water at 1 mg/ml, and the pH was adjusted to 7.0. Peptide concentrations of 0.1 or 0.2 mg/ml were analyzed. Typically, two scans were averaged for each spectrum. CD spectra of peptides in ␣-helical conformations were recorded in the presence of trifluoroethanol (TFE), which served as a positive control of peptides in ␣-helical conformations. To determine the effects of heparin on peptide conformation, solutions containing peptide plus heparin were prepared at various peptide:heparin ratios (w/w).
Analysis of Peptide Concentration-Concentrations of peptides used in CD were verified by one-dimensional NMR spectroscopy based on an internal 2,2-dimethylsilapentane-5-sulfonate standard. NMR experiments were recorded on a Bruker AMX 600 NMR spectrometer equipped with a 5 mm broadband inverse probe, using the XwinNMR 2.1 software package run on a Silicon Graphics INDY work station. One-dimensional proton spectra were acquired at 303 K using a 4 s relaxation delay and were processed with 0.5 Hz exponential line broadening. 250 l of a 1 mg/ml peptide solution (as determined by weight) was lyophilized and dissolved in 405 l of D 2 O containing 0.123 mM 2,2-dimethylsilapentane-5-sulfonate. The degenerate arginine ␦CH 2 resonances at 3.2 ppm, ascertained by a total correlation spectroscopy experiment, were integrated and compared with the internal standard.
Peptide Length Calculations-Simple random walk statistics, neglecting contributions due to swelling, charge, and self-avoidance, were assumed to apply to peptides of type (AKKARA) n and (ARKKAAKA) n , where n ϭ 1-6, and used to calculate the average spatial separation of terminal amino acids and to estimate overall peptide size. Peptide size was calculated using the average end-to-end separation using the equation, where ϽrϾ ϭ average end-to-end distance, n ϭ number of amino acid residues, l o ϭ average bond length, ϭ average bond angle, and ϭ rotation angle (22). The rotation angle was taken to equal either 0°for a fully extended molecule or 120°for bond rotation leading to a coiled structure. A perfect ␣-helix was assumed subsequent to binding with heparin, such that helical wheel diagrams, representing an 18/5 (residues/rotation) ␣-helix (23), could be constructed to visualize the position of and calculate the spacing for basic amino acids along the helix.

Peptide-Heparin
Interactions-To design small peptides that exhibit high affinities for heparin and for the GAG components of PGs, peptide sequences were modeled after proposed heparin-binding consensus sequence motifs. Thus, a collection of peptides containing one of two consensus sequence motifs, either XBBXBX or XBBBXXBX, as well as various modifications of these, were synthesized (Table I). As peptides containing a single heparin-binding sequence often show little or no affinity for heparin (24), a strategy used here was to include consensus sequences in multiple copies within peptides. In initial studies, we selected for synthesis the sequences (AKKARA) n or (ARK-KAAKA) n , where n ϭ 1-6. Alanine was included in the hydropathic positions because of its stabilizing activity on ␣-helices (25), and the basic amino acids were chosen to represent those with the highest probability of occurrence in each basic position in the heparin-binding consensus sequences of native heparinbinding proteins (6). When single copies of either sequence were tested for heparin binding by ACE, no affinities were detected. In contrast, peptides containing two copies of the consensus sequence exhibited weak but detectable affinities for heparin (Ͻ6 M), and peptides of higher molecular weight containing 4 -6 copies of a consensus sequence showed a marked increase in heparin binding affinity (40 -150 nM) (  Table I).
To define the sequence and conformational features of the tandem repeat peptides that confer their high affinity heparin binding characteristics, peptides containing variants of one of the consensus sequences first tested, (ARKKAAKA) 3 , were synthesized. These included those in which alanines were replaced by other hydropathic residues, the spacings between consensus sequences were altered by removal or addition of alanine residues, or the potential of the peptides to form stable ␣-helices was inhibited by including proline residues at various positions. It was found that peptide affinity for heparin was decreased when alanine was replaced by glycine in all the hydropathic positions ((ARKKAAKA) 3 , 3 , K d Х 200 nM; p Ͻ 0.01); less conservative substitutions had varying effects on heparin binding affinity, i.e. for (LRKKLGKR) 3 , K d Х 105 nM, affinity was unaffected; for (TRKKLGKI) 3 , K d Х 740 nM (p Ͻ 0.01), affinity was decreased (Table I).
Two peptides were synthesized in which the spacings be- I-tyramine-heparin through peptides were determined from ACE gel electrophoretograms and are plotted against peptide concentration as detailed under "Experimental Procedures." Smooth curves represent nonlinear leastsquares fits to the equation R ϭ R ϱ /(1ϩ (K d /[peptide]) n ). Peptides containing single consensus sequences (AKKARA (q) or ARKKAAKA (E)) do not bind heparin with a measurable affinity; in contrast, significant heparin binding was seen with peptides containing multiple heparinbinding consensus sequences and increased as a function of peptide M r . ϫ, (AKKARA) 2  tween adjacent consensus sequences were altered. Both increasing(ARKKAAKA-AAAA-ARKKAAKA-AAAA-ARKKAA-KA) and decreasing (ARKKAAKA-RKKAAKA-RKKAAKA) the distance between consensus sequences resulted in decreased heparin binding affinity (K d Х 250 and 450 nM, respectively). Inclusion of prolines also decreased the heparin binding affinity, the degree of which was influenced by their position and number. Thus, the heparin binding affinity decreased to 360 nM when prolines were present in each tandem repeat in place of an alanine ((ARKKPAKA) 3 ); however, a weaker affinity was obtained when a single proline was substituted in the center of a series of three heparin-binding consensus sequences (ARK-KAAKA-ARKKPAKA-ARKKAAKA, K d Х 730 nM; Table I).
Other peptides synthesized and studied include sequences native to the mouse (YPARRARYQWVRCKP) and human (YPTQRARYQWVRCNP) serglycin (SG) core proteins, which contain a single and a partial consensus sequence, respectively. These showed significant affinities for heparin (K d Х 200 -900 nM; Table I and Fig. 2), despite their small sizes (about 2000 Da). To elucidate the basis for the strong heparin binding features of these peptides, the ability of the basic residues to sustain high affinity binding was tested by studying a peptide that contained all of the basic residues of the mouse sequence in their native positions but in which all other residues were changed to alanines (AAAR-RARAAAARAKA). A 350-fold decrease in heparin binding affinity (K d Х 72 M) for this peptide indicated that the number and arrangement of basic residues in the mouse sequence was not sufficient for high affinity binding and suggests the importance of one or more of the other nonbasic residues (Fig. 2). We next tested whether the C-terminal cysteine in the mouse SG peptide may promote peptide dimer formation, thereby influencing heparin binding affinity. Thus, heparin binding was tested by ACE under reducing conditions, and it was found that this treatment yielded negligible heparin binding. Likewise, when the cysteine residue was replaced by an alanine in the native mouse SG sequence, YPARRARYQWVRAKP, heparin binding affinity was again negligible; both results are consistent with the potential cross-linking function of the cysteine residues (Fig.  2).
CD-The intrinsic structural properties of the peptides were explored using CD spectroscopy. Short peptides of known heparin-binding proteins containing heparin-binding consensus sequences have previously been shown to fold into ␣-helical conformations. In doing so, the basic amino acids locate to one face of the helix and thus are potentially exposed for binding. Peptides that displayed weak ((AKKARA) 2 ), moderate ((AKKARA) 4 ), and strong ((AKKARA) 5 and (AKKARA) 6 ) heparin binding affinities were analyzed by CD to characterize their degree of ␣-helical contents and propensities to form an ␣-helix. All peptides exhibit very similar spectra with peaks at 195 and 216 nm and a crossover at 210 nm (for example, see Fig. 3, (AKKARA) 6 , 1:0 (q), and Fig. 4, (AKKARA) 2 , 1:0 (q)). These spectra are indicative of an extended charged coil conformation that was previously reported for charged poly-Llysines and poly-L-arginines (26).
Intrinsic CD of the peptides shows that they do not adopt ␣-helical conformations. To explore the conformational repertoire of the peptides and to record CD spectra for the ␣-helical conformations, peptides were analyzed by CD in the presence of the nonpolar solvent TFE. Nonpolar solvents are known to increase the degree of ␣-helicity of a peptide in solution by enhancing hydrogen bonding and electrostatic interactions (27). CD of (AKKARA) 6 at 0.1 mg/ml containing 0, 10, 20, 30, 40, and 50% TFE (v/v) was measured. At TFE concentrations Ͼ30%, with an apparent maximal effect induced at 40% TFE, the peptide assumes an ␣-helical conformation, with classic ␣-helical peaks at 206 and 220 nm and a crossover at 197 nm (data not shown).
The CD spectra of (AKKARA) 6 recorded in the presence of increasing amounts of heparin (Fig. 3) demonstrate that a change from a charged coil conformation displayed in the absence of heparin (1:0) occurs upon heparin addition (1:0.25, 1:0.50, and 1:1). Heparin induces a similar ␣-helical conformation at a 1:1 peptide:heparin ratio that was obtained in the presence of Ͼ30% TFE, with classic ␣-helical peaks at 190, 207, and 222 nm. At higher heparin concentrations (1:2 or 1:4) the ␣-helical form is lost, and the spectrum resembles that of a random coil structure. This ability of excess GAG to disrupt the  Table I). Peptide AAARRARAAAARAKA (f) displayed negligible heparin binding affinity (K d Х 75 M), indicating the importance of the nonbasic residues to heparin binding. YPARRARYQWVRCKPheparin binding in the presence of ␤-mercaptoethanol (YPAR-RARYQWVRCKP ϩ ␤-mercaptoethanol (ϫ)) was decreased by over 20-fold (K d Х 4 M). Replacement of cysteine by alanine in the mouse SG peptide (YPARRARYQWVRAKP (Ⅺ)) further reduced heparin binding affinity (K d Х 36 M).
FIG. 3. CD spectroscopy of (AKKARA) 6 in the presence or absence of low M r heparin. CD spectra measurements of (AKKARA) 6 in the absence of heparin (1:0 (q)) reveal peaks at 195 and 216 nm and a crossover at 210 nm, indicative of an extended charged coil conformation. Upon heparin addition (1:0.25 (E) or 1:0.50 (ϫ)), the peptide conformation is altered, and at a 1:1 peptide-heparin ratio (f), the peptide becomes ␣-helical with characteristic ␣-helical peaks at approximately 190, 207, and 222 nm. Excess heparin (1:2 (Ⅺ) or 1:4 (OE)) disrupts this interaction, and the spectra resemble that of a protein in a random coil conformation. Spectra are heparin and/or blank (water) corrected.
␣-helical conformation of a polypeptide in solution has been reported previously (26).
Peptide-PG Interactions-The interactions between consensus sequence peptides and PGs were also examined. For these experiments, total PGs were isolated from HUVEC cultures, because HUVECs have been shown to express a variety of types of HS and CS PGs, including, for example, syndecans, perlecan, glypican, and biglycan (28,29). Thus, cell layer-associated and secreted [ 35 S]SO 4 -radiolabeled PGs were purified by extraction with urea, and those PGs retained on DEAE after a 0.1 M NaCl rinse were studied for their binding to (ARKKAAKA) 4 by ACE (Fig. 5A, EC PGs). This peptide exhibited significant affinity for secreted HUVEC PGs, although the average affinity was somewhat weaker than that exhibited by the peptide for heparin (PG K d Х 300 nM; heparin K d Х 50 nM). Similar affinities were obtained for cell layer-associated PGs (data not shown). Inspection of ACE gels in which secreted PGs were fractionated through peptides demonstrated the presence of at least two populations of PG evident as two distinct bands of radiolabeled material migrating through the peptide lanes with different mobilities (Fig. 5A, EC PGs). This difference in migration rate could indicate heterogeneity of the PG in size or charge. In contrast to the heterogeneity seen in Fig. 5A, Fig. 5B shows that heparin migrates as a single band of radiolabeled material.
Thus, to ascertain which GAG chains, as well as which PG component (i.e. core protein, GAG chains, or both), were responsible for peptide binding, total HUVEC PGs were subjected to various chemical and enzymatic degradations. Samples were then tested for their ability to bind to (ARKKAAKA) 4 . PGs in which HS GAGs were chemically degraded by nitrous acid or enzymatically degraded by heparatinase I were able to maintain comparable affinity for the peptide as was displayed by the total PG sample (Fig. 5A, EC PGs/NA, and Fig. 6). PGs in which CS GAG chains were digested with chondroitinase ABC were also able to maintain comparable affinity for the peptide. Release of GAG chains from cores by borohydride reduction resulted in a 3-4-fold diminished affinity (Fig. 6). DISCUSSION The goal of this study was to design high affinity heparinand PG-binding peptides; the strategy we used was to incorporate into their structure copies of sequences proposed to bind heparin in native proteins. Our approach was also based on the fact that truncation of peptide structure without loss of activity FIG. 4. CD spectroscopy of (AKKARA) 2 in the presence or absence of low M r heparin. The CD spectra measurements of (AKKARA) 2 in the absence of heparin (1:0 (q)) indicate a charged coil conformation (peaks at 195 and 216 nm, crossover at 210 nm). However, in contrast to the heparin-induced conformational change seen for the high affinity heparin-binding peptide (AKKARA) 6 , (AKKARA) 2 , which binds heparin weakly, remains a charged coil in the presence of heparin (1:0.25 (E), 1:0.50 (ϫ), 1:0.75 (f), and 1:1 (Ⅺ)).

FIG. 5. ACE analysis of the interactions between peptides containing heparin-binding consensus sequences and HUVEC PGs.
ACE gel images as obtained by a PhosphorImager in which EC PGs/ GAGs (A) or heparin (B) was fractionated through peptides. In A, at least two populations of high affinity PG/GAG, seen as two bands of radiolabeled material migrating with different mobilities, are visible at peptide concentrations of Յ50 nM. At a peptide concentration of 250 nM (near the K d Х 300 nM), a separation of the PG/GAG species is evident as a broad smear throughout the lane and as a sharp band that migrates approximately half way down the lane, indicating heterogeneity in size, charge density, and/or peptide binding interactions of the PG/ GAG population. PG/GAG samples in which HS PGs have been chemically degraded by nitrous acid (EC PGs/NA) also displayed high binding affinity (K d Х 300 nM), implying that chondroitin/dermatan sulfates that remain in the sample bind the peptide strongly. In contrast to the heterogeneity seen in A, B shows that heparin migrates as a single broad band of radiolabeled material.
FIG. 6. Affinity of (ARKKAAKA) 4 for HUVEC PGs and PG components. The peptide was analyzed for binding affinity to HUVEC PGs/GAGs by ACE, and the K d values of the peptide-PG/GAG interactions were calculated from binding plots as detailed under "Experimental Procedures." Similar affinities (ϳ300 nM) were obtained for total PGs, for PG samples devoid of HS GAGs via nitrous acid treatment (NA) or heparatinase I digestion (H), and for PGs devoid of CS GAGs via chondroitinase ABC digestion (ABC). Liberation of GAG chains from the core protein by borohydride reduction (BH) of total PGs caused a 3-fold reduction in affinity (K d Х 1200 nM). can sometimes be achieved by constraining or manipulating peptide conformation (30). In the case of apolioprotein E and apolipoprotein B-100, heparin-binding sites are believed to form ␣-helices upon heparin binding, and molecular modeling illustrates that basic amino acids in the binding sites align to one side of the helix to form a region of high positive charge through which heparin binding occurs (6). Thus, in our design of heparin-binding peptides, we also incorporated structural features conducive to stable ␣-helicity.
In our initial experiments, families of peptides were synthesized that contained single or multiple copies of heparin-binding consensus sequences. When their heparin binding was examined by ACE, peptides containing single sequences showed no measurable affinity for heparin. This result is as expected because peptides carrying single heparin-binding sequences found in native proteins often fail to display significant heparin binding (24), but they may contain multiple consensus sequences that come into proximity upon protein folding or multimerization, thereby enhancing heparin binding through cooperativity (31). In contrast, the affinity of peptides (AKKARA) n or (ARKKAAKA) n ranged from weak (K d Х 6 -40 M) at n ϭ 2 to strong (K d Х 50 -100 nM) at n ϭ 3-6. These latter affinities are in the range of those displayed by heparinbinding proteins such as basic fibroblast growth factor (K d Х 10 nM) or type I collagen (K d Х 100 -200 nM) (10). However, the fact that the peptides are roughly 4 times smaller than basic fibroblast growth factor and 100 times smaller than type I collagen highlights their significant heparin binding abilities. The affinity appeared to plateau at n Ն 5, or 36 -40 amino acids, suggesting that peptides of approximately 30 -32 amino acids were of sufficient length to occupy all available binding sites on low M r heparin and that additional amino acid residues beyond this did not contribute to heparin binding due to a lack of available ligand. However, this hypothesis cannot be tested without knowing the M r distribution of the heparin used in these experiments. Alternatively, one can speculate on the basis of coil and helix calculations that basic residue spacings within the peptides may account for both the increase and plateau in binding affinity observed for the larger peptides. The random walk calculations confirm that peptides of 20 residues are on the order of the same size as low M r heparin, with an average 6,000 M r , and have sufficient length to form a stable ␣-helix. Helical wheel constructions (not shown) illustrated that in peptides of Ն24 amino acids, basic residues occupy locations primarily to one side of the helix. The basic residues show three consistent spacings, estimated at approximately 10 -12-, 20 -24-, and 45-50-Å separations, with uncertainty being introduced primarily by the assumption of an ideal 18/5 helix. These spacings between basic amino acids could facilitate binding to sulfated heparin disaccharides, based on a 5.5-Å spacing between monosaccharide units, a distance predicted from spacing that occurs in CS (32). This is consistent with work by others who report a necessary 20 -24-Å distance between basic amino acids for optimal heparin binding (9). Peptides of Ͼ36 amino acids do not exhibit a further increase in the number of basic residues aligning to one side of the helix in this 10-or 20-Å spacing; thus, the additional basic amino acids in the larger peptides may not contribute to a further increase in heparin binding affinity.
Other experiments examined heparin binding by peptides including sequences native to proteins that contain a single or partial heparin-binding consensus sequence. Results again suggested the critical nature of peptide M r and number of consensus sequences to heparin binding. Thus, surprisingly, a strong heparin binding affinity was displayed by a peptide corresponding to the mouse SG proteoglycan core protein con-taining a single consensus sequence, YPARRARYQWVRCKP (K d Х 200 nM). However, the affinity was diminished over 200-fold by disulfide reduction or replacement of the cysteine with alanine, thus implying that its strong heparin binding relies on peptide dimerization and that the other residues flanking the consensus sequence were of little consequence. Indeed, others have shown that inclusion of cysteines near peptide termini to promote disulfide bond formation may improve peptide-ligand binding (30); our results suggest this to be a simple strategy to greatly enhance the affinity of peptides for heparin. SG, cerebroglycan (with PRRLRL) (33), and perlecan (with TRRFRD) (34) are among the few PGs that contain heparin-binding consensus sequences on their core proteins. Interestingly, the SG core protein, which carries many heparin chains, migrates at twice its predicted molecular weight on PAGE gels under reducing conditions (35), suggesting dimerization. This could result from GAG chains of one PG binding to the core protein of another or from core-core associations through disulfide bonding. The potential physiological function of such PG-PG interactions remains to be explored.
Additional consensus sequence peptides were designed to determine other aspects of peptide structure important to heparin binding. Including glycine in place of alanine in the hydropathic positions weakened heparin binding, and peptides in which arginine was included in all basic positions displayed higher affinity for heparin than did those containing arginines and lysines. The latter is consistent with work showing a higher affinity interaction of arginine-heparin and arginine-HS than lysine-heparin or lysine-HS (36). This suggests that the heparin binding characteristics of the peptides developed here may rely on amino acid type and arrangement in addition to ionic interactions. Inclusion of prolines within or between consensus sequence motifs weakened affinity for heparin, possibly as a result of alterations in peptide secondary conformation; this issue was investigated in our CD experiments. Finally, changing the spacing between consensus motifs weakened affinity for heparin; however, sequence orientation did not appear to influence binding ability as long as the motifs were contiguous and in one orientation.
Molecular modeling of consensus sequences in native heparin-binding proteins predicts their presence within ␣-helical regions (6). Additionally, GAG-directed conformational changes on polypeptides such as poly-L-lysine and poly-L-arginine have been identified (26,37,38). Aqueous solutions of these polypeptides at neutral pH were shown by CD to adopt charged coil conformations and to display ␣-helical conformations in the presence of heparin. Our results showed that peptides of the type (AKKARA) n have charged coil conformations at neutral pH. In the presence of heparin, however, a peptide that showed high affinity for heparin, (AKKARA) 6 , underwent a conformational change to an ␣-helix. In the presence of excess heparin, a further conformational change produced a random coil structure. In contrast, a peptide that displayed weak heparin binding, (AKKARA) 2 , failed to undergo any conformational change. Thus, the solution conformation of a peptide and its propensity to change conformation in the presence of heparin may be an indication of its ability to bind to heparin strongly. These data and those from experiments examining the effects of including prolines in peptides, which are known to disrupt the ␣-helical conformation, suggest that peptide secondary structure facilitates heparin binding.
Here we also examined the interaction between the high affinity heparin-binding peptide (ARKKAAKA) 4 and EC PGs. Results showed that ECs secreted several types of PGs/GAGs that displayed significant affinities for (ARKKAAKA) 4 (K d Х 300 nM). ACE gel images revealed the resolution of multiple PG/GAG species after their migration through the peptidecontaining lanes, suggesting heterogeneity in PG/GAG charge, size, and/or binding affinities. It was found that the CS PGs or HS PGs likely bind the peptide similarly, because affinity was maintained even after treatment of total PGs with nitrous acid, which selectively degrades HS GAGs, heparatinase I, or chondroitinase ABC. The free GAG chains had 3-4-fold lower affinity than the intact PGs. Thus, the core proteins of certain EC PGs may either contribute to binding directly or act as a tether to bring multiple GAGs into proximity for cooperative binding. Similar observations have been made previously for cartilage PG-type II collagen interactions (39) and SG-type I collagen interactions (40,41). Our results are inconsistent with carbohydrate sequence selectivity in the binding of these peptides with EC PGs, because similar affinities for peptides were displayed by either total EC PGs or its CS PG fraction.
Of note is that the heparin-binding peptides designed here incorporate concatamers of heparin binding consensus sequences, which should rarely, if ever, appear in native proteins. Nonetheless, the proposed characteristics of heparin-binding motifs in proteins, as set forth by Cardin and Weintraub (6) based on their theoretical analysis of putative heparin-binding domains of native proteins, hold true with our model peptides. Thus, our data suggest that peptides containing the Cardin and Weintraub heparin-binding consensus sequences may show a selective advantage in heparin binding over certain other sequences that do not fit their criteria.
In summary, optimally active heparin-binding peptides should include multiple sequences of the types (XBBXBX) n and (XBBBXXBX) n . Sequence number and peptide M r are the most critical features; peptides should be of at least approximately 30 residues, which could be decreased to 15 if cysteine is included near either terminus to promote dimerization. Peptides should contain contiguous sequence arrays, without intervening residues between sequences. Alanine, which stabilizes ␣-helical conformation, should occupy the hydropathic residue positions, and arginine should occupy the basic positions. The high affinity PG-or GAG-binding peptides developed here, or derivatives thereof, could prove useful as tools for the promotion of cell-substratum attachment of PG-expressing cells, in the targeting of drugs to PG-expressing cells and PG-rich extracellular matrices, or as antagonists of GAG-mediated actions, e.g. neutralization of the anticoagulant activity of heparin.