Polynucleotide Binding to Macrophage Scavenger Receptors Depends on the Formation of Base-quartet-stabilized Four-stranded Helices*

Macrophage scavenger receptors exhibit unusually broad, but circumscribed, polyanionic ligand-binding specificity. For example, the polyribonucleotides poly(1) and poly(G) are ligands but poly(A) and poly(C) are not. To further investigate the molecular basis of this polynucleotide-binding specificity, we tested the capacity of various oligodeoxyribonucleic acids to in- hibit the scavenger receptor-mediated degradation of ’251-labeled acetylated low density lipoprotein by Chinese hamster ovary cells expressing the type I bovine scavenger receptor. A series of short oligodeoxy- riboguanines (dG,, where 5 5 n 5 37) were effective inhibitors. The dGs, dGlz, and dA6G37 members of this series were shown by circular dichroism and UV spectroscopy to be assembled into four-stranded helices stabilized by G-quartets. [3ZP]dAeGs7 bound directly to scavenger receptors. Partial or complete denaturation of the quadruplex structures of these oligonucleotides by boiling destroyed their inhibitory activity. Receptor activity was also inhibited by d(T4G&, a telomere-like oligonucleotide which forms an intramolecular quad- ruplex. In addition, conversion of the four-stranded potassium salt of poly(1) to the single-stranded lithium salt dramatically reduced its inhibitory activity. reformation of quadruplex and of


Polynucleotide Binding to Macrophage Scavenger Receptors Depends on the Formation of Base-quartet-stabilized Four-stranded Helices*
Alan M. Pearson, Alexander Rich, and Monty KriegerS From the Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 Macrophage scavenger receptors exhibit unusually broad, but circumscribed, polyanionic ligand-binding specificity. For example, the polyribonucleotides poly (1) and poly(G) are ligands but poly (A) and poly(C) are not. To further investigate the molecular basis of this polynucleotide-binding specificity, we tested the capacity of various oligodeoxyribonucleic acids to inhibit the scavenger receptor-mediated degradation of '251-labeled acetylated low density lipoprotein by Chinese hamster ovary cells expressing the type I bovine scavenger receptor. A series of short oligodeoxyriboguanines (dG,, where 5 5 n 5 37) were effective inhibitors. The dGs, dGlz, and dA6G37 members of this series were shown by circular dichroism and UV spectroscopy to be assembled into four-stranded helices stabilized by G-quartets.
[3ZP]dAeGs7 bound directly to scavenger receptors. Partial or complete denaturation of the quadruplex structures of these oligonucleotides by boiling destroyed their inhibitory activity. Receptor activity was also inhibited by d(T4G&, a telomere-like oligonucleotide which forms an intramolecular quadruplex. In addition, conversion of the four-stranded potassium salt of poly (1) to the single-stranded lithium salt dramatically reduced its inhibitory activity. Addition of KC1 to the Li+ salt resulted in the reformation of poly(1)'s quadruplex structure and restoration of its inhibitory activity. A variety of single-stranded and double-stranded oligo-and polydeoxyribonucleotides (e.g. dA3,, Hue111 restriction fragments of @X174) exhibited very little or no inhibitory activity. Thus, a base-quartet-stabilized four-stranded helix appears to be a necessary structural determinant for polynucleotide binding to and inhibition of scavenger receptors. This conformational requirement accounts for the previously unexplained polyribonucleotide-binding specificity of scavenger receptors. The spatial distribution of the negatively charged phosphates in polynucleotide quadruplexes may form a charged surface which is complementary to the positively charged surface of the collagenous ligand-binding domain of the scavenger receptor.
* This work was supported by National Institutes of Health, National Heart, Lung and Blood Institute Grant HL41484 (to M. K.), Department of Energy Grant DE-SG05-92ER79111 (to A. R.), and Office of Naval Research Grant N00014-90-5-4075 (to A. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ T o whom correspondence should be addressed Rm. E25-236, Biology Department, Massachusetts Institute of Technology, Cambridge, MA 02139. Tel.: 617-253-6793;Fax: 617-258-5851 or 617-258-6553. Macrophage scavenger receptors are trimeric integral membrane glycoproteins which have been implicated in the deposition of lipoprotein cholesterol in artery walls during the formation of atherosclerotic plaques (1,2) and in pathogen recognition for host defense (3). Two classes of macrophage scavenger receptors have been cloned from bovine, murine, human, and rabbit cDNA libraries (4-6, 8, 9).' The sequence of the type I bovine scavenger receptor cDNA predicts a 453amino acid protein with the following domains (4)': I, NH2terminal cytoplasmic (amino acid residues 1-50) ; 11, transmembrane (51-76); 111, spacer (77-150); IV, a-helical coiledcoil (151-271); V, collagenous (272-343); and VI, , which includes a 102-residue scavenger receptor cysteine-rich domain. The scavenger receptor cysteine-rich domain helped to define a previously unrecognized family of remarkably well conserved cysteine-rich protein domains (6,lO). The type I1 scavenger receptor is identical to the type I receptor, except that the 110-amino acid COOH terminus is replaced by a 6-amino acid COOH terminus ( 5 ) . Despite its truncated COOH terminus, the type I1 scavenger receptor mediates the binding and endocytosis of ligands with affinities and specificity similar to those of the type I receptor ( 5 , 11,12). The collagenous domain appears to play a critical role in ligand binding (3,13,14). 2 Unlike most cell-surface receptors, macrophage scavenger receptors exhibit unusually broad ligand specificity (1,3,(15)(16)(17). Their high affinity ligands are polyanions, including certain chemically modified proteins such as acetylated LDL (Ac-LDL)~ and maleylated BSA (M-BSA) and certain polysaccharides, phospholipids (12, 18), and polyribonucleotides (see below). Bacterial endotoxin (lipopolysaccharide) is also a ligand for scavenger receptors, which appear to play an important role in the rapid clearance of endotoxin from the circulation by the liver (12).
To learn more about the mechanism underlying the scavenger receptors' unusual broad binding specificity, we have examined the structural basis of the type I bovine scavenger receptor's specificity for polyribonucleic acids. Poly(1) and poly(G) are ligands and competitive inhibitors of the scavenger receptor whereas poly(C), poly(U), and poly (A) are not (1,16,19). It is striking that the polyribonucleotides which are ligands form four-stranded helices under normal physiological conditions while the polyribonucleotides which are not J. Ashkenas, M. Penman, E. Vasile, S. Acton, M. W. Freeman, and M. Krieger, submitted for publication. ligands do not form such quadruplexes reviewed i n Ref. 25).
A model for a four-stranded deoxyguanosine qvadruplex is shown in Fig. 1. This model is based on the 2.5-A resolution x-ray crystal structure of the quadruplex formed by a dimer of the telomere-like oligodeoxyribonucleotide d(G4T4G4) (26). Fig. 1A shows the hydrogen bonding between the four guanine bases of adjacent strands in this structure. The bases in this "G-quartet" are roughly planar (shaded region). In the quadruplex, the G-quartets are held together by the sugar-phosphate backbones and stack with a right-handed twist, resulting in the formation of a four-stranded helix. Fig. 1B shows a hypothetical model of a portion of a quadruplex which is based on the d(G,T4G4) crystal structure and the proposed quadruplex structures of poly(1) and poly(G) (20)(21)(22). Detailed features of quadruplex structures vary depending on the oligoand polynucleotide sequences and ionic conditions (25-38). For example, quadruplexes can be composed either of extended oligo-or polynucleotide strands stabilized by intermolecular base-quartets, or of strands which fold back on themselves to form intramolecular or mixed intra-and intermolecular base-quartets. In all cases, the four-stranded helices have very high negative charge densities due to the close proximity of the phosphate groups.
In the current study we have tested and confirmed the hypothesis that assembly of polynucleotide ligands into fourstranded helices is a requirement for their binding to and consequent inhibition of scavenger receptors. We have also demonstrated that other quadruplex forming molecules not previously known to inhibit scavenger receptor activity, including telomeric oligodeoxyribonucleotides, are scavenger receptor ligands.  I n the structure of the d(G4T4G4) dimer the oligonucleotide chains fold back on themselves to form an antiparallel quadruplex (not shown). As a consequence, the base conformations around the Gquartet alternate between syn and anti and the quadruplex has wide and narrow helical grooves (see "Discussion"). Panel B shows a hypothetical model of an extended parallel four-stranded helix, based in part on the x-ray fiber structures of poly(1) and poly(G) (20)(21)(22). T h e shaded planes represent the base-quartets (see text).

Scavenger Receptors 3547
Pharmacia LKB Biotechnology Inc. ( S Z O ,~ values 6.2 and 5.1; molecular masses, estimated by the manufacturer, 111 and 72.5 kDa, respectively). These are known to be four-stranded in a K+ solution (27). Stock solutions of the K+ salt of poly(1) (5 mg/ml) in deionized water (with or without diethyl pyrocarbonate treatment) were stored at 4 "C. Two double-stranded DNA preparations were used HaeIIIdigested fragments of @X174 (New England Biolabs) and pRc/CMVs-bSR-I (19), a plasmid derived from pRc/CMV (Invitrogen). Most oligonucleotides were synthesized on an Applied Biosystems Model 380B automated DNA synthesizer and deprotected in ammonia at 55 "C by the Biopolymers Laboratory at MIT. Deprotected oligonucleotides were dissolved in deionized water and de-salted over NAP10 or NAP25 columns (Pharmacia), or over Sephadex G-10 (Pharmacia) columns for oligonucleotides 15 residues or shorter. The de-salting step is not expected to remove counterions, such as Na+ and K+, which are tightly bound to the dG, oligonucleotides (32). The fractions containing oligonucleotides were concentrated by evaporation, resuspended a t 0.4-4 mg/ml in deionized water, and stored a t -20 "C. The concentrations of the oligonucleotides were determined from their A260 values (conversion factor: 30 pg/ml for A2W = 1.0 (I-cm path length) (39)). Due to the hypochromic shift accompanying structure formation, this method underestimates the concentrations of oligonucleotides which fold into four-stranded helices (40). The oligonucleotide RlOl was a gift from Jonathan Wallach (MIT), and the oligonucleotide osP24 was provided by Steven Podos (MIT).
Preparation of the Li+ and Li+/K+ Salts of Poly(Z)-The Li+ and Li+/K+ salts of poly(1) were prepared following the protocol of Howard and Miles (27,41). Briefly, stock solutions of the K+ salt of poly(1) (0.5 ml) were converted to the Li+ salt by successive dialysis against 4 liters of the following solutions: 0.5 M LiC1, 1 mM EDTA (3 h, room temperature); 0.5 M LiCl (2 h, 4 "C); 3 changes of 0.1 M LiCl (19-23 h total, 4 "C); and 4 changes of deionized water (74-95 h total, 4 "C). Any poly(1) molecules significantly smaller than the molecular mass cut-off of the dialysis tubing (6)(7)(8) were probably lost during dialysis. In some cases, the 1 2 salt of poly(1) was converted to the Li+/K+ salt by addition of KC1 (final concentration: 150 or 200 mM). These preparations were stored a t 4 "C for at least 4 days before use in '"1-Ac-LDL degradation assays. The poly(1) concentration of each sample was determined spectrophotometrically (conversion factor: 40 Fg/ml for AZG0 = 1.0 (1-cm path length) (39)). In several experiments, the concentrations of the poly(1) isoforms were also determined after digesting the specimens to monomers with Nuclease P1 (Pharmacia; 0.2 units of enzyme in a 20.~1 reaction volume, 4.5 h at 50 "C) and using an extinction coefficient for 5'-IMP of 12.2 ml/ pmollcm a t 249 nm (42) (not shown). This method for correcting errors due to the hypochromic shift did not significantly alter the results.
Gel Filtration Chromatography-Oligonucleotide samples  pg/ml) in buffer A (10 mM Tris-C1, pH 7.6, 146 mM NaC1, 3 mM KCl, 0.6 mM MgCl,, 0.3 mM CaCl2; salt composition similar to that of Ham's F-12 medium (45) were filtered through a 0.45-pm Millex HA filter (Millipore). The samples were analyzed a t 4 "C using a Pharmacia LKB 500 FPLC system with a Superose 6 HR 10/30 column (pre-equilibrated in buffer A) at a flow rate of 0.5 ml of buffer A/min. The absorbance at 254 nm of the eluted material was continuously recorded using a UV-M I1 optical unit with a 2-mm path length. There was some variability from preparation to preparation in the heterogeneity of the size distributions observed (see "Results"). Poly(1) isoforms (50 pg/ml) in buffer B (20 mM Tris-C1, pH 8.0, 150 mM NaC1,l mM CaCIZ) were subjected to similar FPLC analysis after clarification by centrifugation (13,000 X g, 15 S, 4 "C).

Quadruplex Binding to Scavenger Receptors
untreated K+ salt of poly(I), its Li+ salt, and its Li+/K+ salt (all at 25 pg/ml) were obtained at 247 nm using an Aviv Model 118 DS spectrophotometer. Absorbance was recorded from 10 to 90 "C in 2 "C steps, with an equilibration time of 1.5 min, and an acquisition time of 45 s. The absorbance values presented represent the differences between the values obtained for poly (1) in buffer C and for buffer C alone. Similar results were obtained for poly(1) in Ham's F-12 medium (data not shown).
Circular Dichroism (CD)-CD spectra of dG, oligonucleotides dissolved in water at either 35 pg/ml for dG,, and dAsGa7 (106 ~L M in nucleotide residues) or 15.3 pg/ml for dGs (46 p~) were collected at 37 "C on an Aviv Model 60DS equipped with a thermoelectric controller (1-cm cell, dynode voltage remained below 400 mV). Spectral scans of untreated samples were taken in 1-nm steps with an acquisition time of 5 s/datum. Each sample was then placed in boiling water for 15 min, cooled at room temperature for 30 min, and rescanned. The ellipticity values presented represent the differences between the values obtained for the oligonucleotide solutions and for a water blank and are corrected for specimen concentrations as previously described (46). UV spectra were recorded concurrently.
Cell Culture-All incubations with cells were performed at 37 "C i n a humidified 5% COP, 95% air incubator. Stock cultures of wildtype CHO cells were maintained in medium A (Ham's F-12 supplemented with 100 units/ml penicillin, 100 pg/ml streptomycin, 2 mM glutamine, and 5% (v/v) fetal bovine serum). The isolation of transfectants expressing the type I bovine scavenger receptor (CHO[bSR-I], clone I-B2) and the type I1 bovine scavenger receptor (CHO[bSR-111, clone 11-5) was described previously (11). Transfectants were maintained in MAC3 medium (medium A containing 3% newborn calf lipoprotein-deficient serum in place of the fetal bovine serum, and supplemented with 0.5 mg/ml geneticin (G418, Gibco), 250 ~L M mevalonate, 40 p~ compactin, and 3 pg of protein/ml of Ac-LDL), which provides selective pressure for the expression of scavenger receptor activity as previously described (11). Media components, including sera, were obtained or prepared as previously described (11,43).
Assays-Scavenger receptor activity at 37 "C was assessed by measuring cellular degradation of lz5I-Ac-LDL in 24-well culture dishes as previously described (15,43,44). Unless otherwise noted, CHO[bSR-I] cells were seeded a t a concentration of 60,000 cells/well in medium B (medium A supplemented with 0.5 mg/ml G418) on day 0. On day 2, 5 p g of protein/ml of lZ51-Ac-LDL (153-630 cpm/ng protein) were added to the wells in 0.5 ml of culture medium in the absence (triplicate determinations) or presence (duplicate determinations) of the indicated additions. The culture medium used for these assays was either medium A or medium C (medium A containing 10 units/ ml of Escherichia coli RNase inhibitor (Calbiochem) and 1% (v/v) ITS+ (an insulin, transferrin, selinium, and linoleic acid/albumin supplement from Collaborative Research) in place of fetal bovine serum). After a 5-or 5.5-h incubation at 37 "C, the amounts of "' I-Ac-LDL degradation products released into the media were measured and are presented as nanograms of '"I-Ac-LDL protein degraded per 5 (5.5) h/mg of cell protein. Cellular binding and uptake of ["PI dA,G,, after an incubation for 5 h a t 37 "C was measured by washing the monolayers 6 times at 4 "C as previously described (44), solubilizing each monolayer with 500 p1 of 0.1 N NaOH for 15 min at room temperature, neutralizing each sample with 50 pl of 1 N HC1, and counting 110-pl aliquots in 5 ml of Ultrafluor scintillation fluid (National Diagnostics). The binding plus uptake values are given as ng bound plus internalized per 5 h/mg of cell protein. Protein concentrations were determined by the method of Lowry et al. (47).

RESULTS
To test the hypothesis that polynucleotide ligands bind to and inhibit scavenger receptors when the ligands are assembled into four-stranded helices, we synthesized and analyzed the properties of an oligodeoxyribonucleotide, dA6G37. Guanine-rich oligo-and polynucleotides have been shown to adopt four-stranded, or quadruplex, structures via the formation of inter-and/or intramolecular G-quartets (22,38,(48)(49)(50). The 5 adenosine residues at the 5' end were included to facilitate 32P-end-labeling with polynucleotide kinase (see below). Fig.  2 shows that this oligonucleotide inhibited scavenger receptor activity, as measured by the degradation of lZ51-Ac-LDL by CHO cells expressing the bovine type I scavenger receptor (CHO[bSR-I] cells). The 50% inhibitory dose (ID5J was 1.1 f 0.5 pg/ml (average f S.D. from seven experiments) and virtually all of the scavenger receptor activity was inhibited at an oligonucleotide concentration of 25 pg/ml. The control molecule d A 3 7 , which is not expected to form quadruplex structures, did not significantly inhibit '"I-Ac-LDL degradation at this concentration (95 f 11% of control activity, from four experiments).
These inhibition data suggested that dA5G37 might bind directly to the scavenger receptor and competitively inhibit '251-Ac-LDL degradation. This conclusion was supported by experiments which showed that dAsG37, but not d A 3 7 , could also block the specific binding of a soluble form of the type I bovine scavenger receptor to M-BSA beads (Ref. 19, and data not shown). Thus, it is unlikely that the dA5G37 was interfering with lZ5I-Ac-LDL degradation because of binding to lZ5I-Ac-LDL rather than to the receptor.
T o more directly examine oligonucleotide binding to the receptor, we end-labeled dA5G37 with [32P]phosphate and measured its binding and uptake by CHO[bSR-I] cells. Fig.  3A shows that specific [32P]dA5G37 binding and uptake (open circles) a t low concentrations ( 4 0 pg/ml) was saturable and was characterized by a dissociation constant of approximately 1 pg/ml. This Kd was similar to the ID50 determined by inhibition of '2'I-Ac-LDL degradation. The specific binding plus uptake shown in Fig. 3A was calculated by subtracting the "nonspecific" cell-associated [32P]dA5G37 measured in the presence of a large excess of the scavenger receptor inhibitor M-BSA (400 pg/ml, solid squures, Fig. 3B) from the total cellassociated [32P]dA5G37 (Fig. 3B, solid circles). The nonspecific binding plus uptake was relatively high and increased linearly with increasing [32P]dA5G37 concentrations. The relationship of this nonspecific binding plus uptake to the previously reported cellular uptake of oligonucleotides is not clear (51,52). To verify that the specific cell association of [32P]dA5G37 ( Fig. 3A) was due to scavenger receptor expression, we measured the binding plus uptake of [32P]dA5G37 by untransfected CHO cells, which express essentially no scavenger receptor activity (11,12,15). The untransfected CHO cells exhibited little or no M-BSA-competable [32P]dAsG37 binding and uptake over the concentration range shown in Fig. 3 (data not shown). Fig. 3C shows the transfection-dependent binding plus uptake of [32P]dA5G37 (open triangles), which was calculated by subtracting the values for the total binding plus uptake for untransfected cells (panel D, solid triangles) from those for the cells expressing the type I bovine scavenger receptor (panel D, solid circles). The shapes of the curves for the transfection-dependent data (panel C) and the M-BSAcompetable ("specific") data (panel A) were remarkably similar. Taken together, these data clearly establish that ["PI dA6G37 bound to scavenger receptors in a saturable manner with high affinity.
We next examined the effect of varying the length of dG, oligonucleotides on their inhibition of lZ5I-Ac-LDL degradation by CHO[bSR-I] cells. The monomers 5'-dGMP and 3'-dGMP and the dimer dGpG at concentrations of up to 100 pg/ml failed to inhibit receptor activity (data not shown). However, all dG, oligonucleotides tested with n 2 5 were inhibitors (Fig. 4A, dGs and dG4 were not examined). Essentially complete inhibition was observed a t 53 pg/ml for n 2 12. The level of maximal inhibition decreased as the oligonucleotide length decreased from n = 10 to n = 5. However, the apparent affinity for the receptor for at least one of these shorter oligonucleotides, dGs, did not appear to be substantially less than that for dGlz or dA5G37 (Fig. 4B). The molecular basis for this length dependence of maximal inhibitory activity has not yet been determined. We have recently observed that a form of bacterial endotoxin, lipopolysaccharide, can only partially inhibit scavenger receptor-mediated degradation of lZ5I-Ac-LDL by transfected CHO cells expressing the bovine type I1 or murine type I scavenger receptors.' The relationship, if any, between the partial inhibition of activity by the short dG,'s and the partial inhibition of activity by lipopolysaccharide has not yet been established.
The relationships of the structures of several dG, compounds to their inhibitory activities were assessed using CD spectroscopy at 37 "C and denaturation by boiling in deionized water. Fig. 5 shows the CD spectra of dGs, dG12, and dA5G37, before and after boiling. All three spectra of the untreated materials show maxima centered near 261 and 209 nm, min-

FIG. 4. Effect of dG. length on the inhibition of '''I-AcLDL degradation by CHO[bSR-I] cells. Panel A , CHO[bSR-I] cells
were plated on day 0 (30,000 cells/well). On day 3, the cells were incubated at 37 "C for 5.5 h in medium A containing 5 pg of protein/ ml of lZ51-AcLDL in the absence (triplicate incubations) or presence (duplicate incubations) of the indicated dG, oligonucleotides at concentrations of either 5.3 pg/ml (open circles) or 53.0 pg/ml (solid circles). Degradation was measured as described under "Experimental Procedures." For n = 37, dA5G37 was used in place of dG37. Panel B , cells were plated on day 0 at 60,000 cells/well. On day 2, the cells were incubated at 37 "C for 5 h in medium A containing 5 pg of protein/ml of '*'I-Ac-LDL in the absence (triplicate incubations) or presence (duplicate incubations) of the indicated concentrations of oligo dG6, dGIZ, or dA5G37. Degradation was measured as described under "Experimental Procedures." ima near 241 nm, and distinctive shoulders near 230 nm (Fig.  5A). The spectra of these untreated oligo dG,'s were similar to previously reported spectra of quadruplex polynucleotides, including dGs (48), dGs (50), and dG12 (38), poly(G) (23), poly(dG) (48), and telomeric (34-36)4.5, and model telomeric sequences (32, 54). Furthermore, the location of the maxima near 261 nm suggests that these oligo dG, quadruplexes were composed of parallel rather than antiparallel oligonucleotide chains, presumably with all anti-glycosidic bond conformations (see "Discussion"). Boiling in water for 15 min reduced the size of the peaks, shifted the minima near 241 nm and the maxima near 261 nm to shorter wavelengths by 2-5 nm, and eliminated the shoulder near 230 nm (Fig. 5B). These effects, which were most pronounced for dGs, have previously been seen in similar molecules as a result of heating, and have been ascribed to the melting of G-quartet stabilized four-stranded structures into single-stranded oligonucleotides (32, 34, 35, 48). Concurrent UV scans showed that the absorbances were significantly greater at 290-310 nm for the untreated specimens than for the boiled samples, and that there were significantly more pronounced shoulders near 275 nm in the spectra of the boiled samples than those near 280 nm in the spectra of the unboiled samples (data not shown). Similar results Lu, M., Guo, G., and Kallenbach, N. R. (1993) Biochemistry, in ' Q. Guo FIG. 5. CD spectra of untreated and boiled dG, oligonucleotides. The indicated oligonucleotides were dissolved in 3 ml of deionized water at 15.3-35 pg/ ml (46-106 pM in nucleotide residues). CD spectra for each oligonucleotide were recorded at 37 "C before (panel A ) and after (panel B ) boiling as described under "Experimental Procedures." have been reported for dG, (48), poly(G) (55), and for telomeric four-stranded G-quartet structures (56). Therefore, these CD and UV spectroscopic results provide strong evidence that dG6, dGIP, dABG37, and presumably the other dG, compounds, formed G-quartet-stabilized parallel stranded quadruplexes, and that boiling for 15 min in deionized water, followed by re-equilibration for 30 min at room temperature, resulted in the partial or complete denaturation of the quadruplexes.
Further characterization of these oligo dG,s by FPLC analysis supported these conclusions. Superose 6 chromatography at 4 "C ( Fig. 6) showed that the untreated dG6, dGlz, and dA5G37 preparations exhibited extremely heterogeneous size distributions and included material with very large Stokes radii (small elution volumes near the void volume of the column). Similar FPLC profiles were seen at 37 "C (data not shown). In every case, boiling in deionized water substantially reduced the apparent sizes of these oligonucleotide complexes (data not shown). Therefore, it is likely that intermolecular G-quartets noncovalently cross-linked the oligonucleotides into large and heterogeneous aggregates. These results are consistent with the previously reported G-quartet-mediated formation of aggregates by guanine-rich oligonucleotides (48, 57-59; however, see Ref.

50).
Denaturation of the four-stranded dGs, dGI2, and dA6GZ7 complexes by boiling dramatically reduced their capacity to inhibit scavenger receptor activity ( Table I). We assume, but have not shown, that the denatured oligonucleotides did not renature in the culture medium during the assay of receptor activity. Taken together, these spectroscopic, chromato-   Cells were plated on day 0 (20,000-60,000 cells/well) and scavenger receptor activity ('"I-AcLDL degradation) was assayed on day 2 or 3 as described under "Experimental Procedures." The 100% control values measured in the absence of added oligonucleotides ranged from 565 t o 2640 ng/5 h/mg of protein. For each oligonucleotide, the values are the averages k S.D. determined in 5-7 experiments, using 2 or 3 independently synthesized preparations of each oligonucleotide.
graphic, and inhibition data, along with structural data from other laboratories, provide strong support for the proposal that the formation of G-quartet-stabilized four-stranded structures was responsible for dG, binding to and inhibition of the scavenger receptor.
Additional evidence for the importance of four-stranded structures in mediating the binding to and inhibition of scavenger receptors by polynucleotide ligands was obtained in studies of the structure and activity of the polyribonucleotide poly(1). X-ray fiber diffraction and other methods have shown that, under normal physiologic conditions, poly(1) chains form parallel four-stranded helices held together by hydrogenbonded inosine-quartets (20-22, 60). Inosine bases lack the N2 group of guanine; therefore, unlike G-quartets which are stabilized by 2 hydrogen bonds between each pair of bases (Fig. lA), inosine quartets contain only one hydrogen bond between each pair of bases. Formation of poly(1) quadruplexes has a striking dependence on the type of monovalent cation available. Na' and K+ promote the formation of and stabilize the four-stranded helical structure by forming coordination complexes with the bases. In contrast, the Li' salt of poly (1) in the absence of such stabilizing cations is primarily singlestranded (27,61). Miles and co-workers (27) have shown that preparations of poly(1) can be cycled between the "denatured" single-stranded form (prepared by extensive dialysis with buffers containing Li+ as the only monovalent cation) and the four-stranded form (by subsequent addition of Na+ and/or K') .
The untreated K' salt of poly(1) exhibited a heterogeneous size distribution and included material with very large Stokes radii as determined by FPLC (Fig. 7A). Extensive dialysis  (1) were prepared in buffer C and the temperature-dependent UV spectra were recorded as described under "Experimental Procedures." against LiCl in water converted the poly(1) to a homogeneous preparation of presumably single-stranded molecules with substantially smaller apparent mass (smaller stokes radius) than the K+ salt form (Fig. 7B). Based on the elution volume, the single-stranded Li+ salt of poly(1) was approximately 19 nucleotides in length. Therefore, it appears that the large apparent mass and the heterogeneity of the untreated K' salt was due to the formation of inosine quartet-stabilized aggregates, as was the case for the G-quartet-stabilized oligo dG, molecules described above. After addition of 200 mM KC1 to the Li+ salt of poly(I), and subsequent incubation at 4 "C for several days, the heterogeneous mixture of large aggregates was reformed (Fig. 7C), although the average apparent size was somewhat less than that of the initial K+ form prior to the salt exchange (Fig. 7A).
To confirm that the structures of these poly(1) preparations were undergoing the expected cation-dependent quadruplex/ monomer transitions, we examined the temperature dependence of their UV absorbances at 247 nm (27,41). As shown in Fig. 8, both untreated poly(1) (solid circles) and the Li+/K+ salt (open diamonds) displayed cooperative melting transitions between 30 and 45 "C in a buffer containing a variety of salts at physiologic concentrations (see "Experimental Procedures''). In contrast, the Li+ salt of poly (1) (open circles) did not exhibit a similar temperature-dependent shift in absorbance at 247 nm between 10 and 75 "C; indeed, its absorbance at all temperatures was similar to those of the high temperature (melted) forms of the untreated and Li+/K+ salts of poly (1). Together with the FPLC data described above, these results indicate that our preparation of the Li' salt of poly (1) consisted predominantly of denatured single-stranded molecules, while the untreated and Li+/K+ preparations were predominantly aggregated quadruplexes. Fig. 9 shows a comparison of the abilities of the quadruplex and predominantly single-stranded poly(1) isoforms to inhibit lZ5I-Ac-LDL degradation by CHO[bSR-I] cells. The untreated K+ salt of poly(1) (open squares) inhibited scavenger receptor activity with an IDso of -1.0 pg/ml, and the inhibition was essentially complete at 4 pg/ml. In contrast, the Li+ salt of poly (1) (open triangles) was a much less efficient inhibitor; there was 50% inhibition at -7 pg/ml and the activity was not fully inhibited even at the highest concentrations of poly(1) tested (18% of control activity at 71 pg/ml). Although we were unable to detect, using FPLC, the formation of poly (1) aggregates during the course of this experiment (data not shown), the relatively low residual activity of the Li+ salt may have been due in part to the presence of small amounts of residual quadruplex structure after dialysis and addition to the assay medium (27, 61). The addition of 200 mM KC1 to the Li' salt of poly(1) (solid squares) restored its inhibitory activity: the ID50 was -1.2 pg/ml, and inhibition was essentially complete at 6 pg/ml. Three other independent Li' dialysis/KCl addition experiments, each using either one or two different commercial preparations of poly(I), gave similar results. The inhibitory activity of the Li+/K+ salt of poly(1) a t 37 "C was somewhat greater than might be expected based on the melting curve in Fig. 8. This may be due to differences in the experimental conditions of the spectroscopic and receptor activity assays which might alter the stability of the fourstranded form of poly (1). For example, the Na+ concentration in the degradation assay was approximately 146 mM, while it was 132 mM in the spectroscopic assay. Small differences in ion concentrations can have a significant effect on poly(1) stability (27). Also, BSA and other proteins were present in the Ham's F-12 medium used in the receptor activity assays but not in the UV melting experiments (see "Experimental Procedures").
In the experiment shown in Fig. 9, an RNase inhibitor from E. coli was used to protect all three poly(1) isoforms and poly(C) (used as a negative control, not shown) from extensive RNase degradation during the 5-h incubation with cells. We have found that the inclusion of an RNase inhibitor to the cell culture assay media was essential to prevent the virtually complete hydrolysis of poly(C) to mono and dinucleotides. The quadruplex polyribonucleotides appeared to be resistant to this degradation, as assessed by FPLC analysis (data not shown). The enzymatic and chemical lability of polyribonucleotides such as poly(C) should be considered when using these molecules as negative controls in scavenger receptor studies.
All of the quadruplex-forming oligonucleotide and polyribonucleotide molecules described above aggregated into heterogeneous higher order structures as determined by FPLC analysis (for example, see Figs. 6 and 7). To determine if oligonucleotides which form small homogeneous, quadruplex structures could also inhibit scavenger receptor activity, we tested the inhibitory activities of several telomere-like oligonucleotides. Unlike the heterogeneous four-stranded dG, and poly(1) molecules described above, many model telomeres fold into well-defined intramolecular G-quartet-stabilized monomeric structures (26, 28, 29, 33, 34). Two preparations of d(T4G4)4 were obtained in the monomeric intramolecularly folded four-stranded form from two independent sources (28, 34). Both of these oligonucleotides exhibited moderate scavenger receptor-inhibitory activity (the average receptor activity in the presence of 25 pg of oligonucleotide/ml was 43% of control), which was similar to those of dGs and dG6 (Fig. 4, and data not shown). We synthesized the related oligonucleotide d(G4T4)5 and found using FPLC (not shown) that it was composed of a mixture of two discrete species with Stokes radii approximately equivalent to those expected for d A z z and d A s 7 . The d(G4TJ6 was an effective inhibitor of scavenger receptor activity (22 f 9% of control activity at 25 pg/ml, average from three experiments), with activity approaching that of dG12. We therefore conclude that at least some oligonucleotides with model telomeric sequences can inhibit scavenger receptors, presumably due to their binding directly to the receptors. Thus, short oligonucleotides with well-defined G-quartet structures, as well as longer polynucleotides and the heterogeneous dG, complexes described above, can be effective inhibitors of scavenger receptors.
We have examined the capacity of several other oligo-and polynucleotides to inhibit scavenger receptors. Table I1 shows that those molecules not expected to form quadruplexes, including double-stranded DNA, were very poor inhibitors of scavenger receptor activity, while those expected or previously shown to form base-quartet stabilized four-stranded molecules were effective inhibitors. In combination with previous reports of the polynucleotide specificity of scavenger receptors (1,16), the current studies provide strong support for the proposal that the formation of quadruplex structures is an important determinant in the binding to and inhibition of scavenger receptors by oligo-and polynucleotides.

DISCUSSION
The broad ligand binding specificity of macrophage scavenger receptors has been well defined but is poorly understood. A large number of polyanions inhibit scavenger receptormediated binding, uptake, and degradation of Ac-LDL (1, 3,  12,15-18). These inhibitors include certain modified proteins, carbohydrates, lipids, and polyribonucleic acids. In many cases, these inhibitors have been shown to be ligands for the receptor and thus competitive inhibitors. However, there are many other polyanionic proteins, carbohydrates, and polyribonucleic acids that do not inhibit Ac-LDL degradation. For example, poly(1) and poly(G) are good inhibitors but poly(A),

Effective and ineffective oligo-ann' polynucleotide inhibitors of macrophage scavenger receptors
The inhibition of scavenger receptor activity was measured using cellular '2SI-Ac-LDL degradation assays. All of the compounds listed were examined in this study except for poly (A), poly(U), poly(G), polyriboxanthinylic acid, and GTP, which, along with poly(I), Dolv(c). were tested Dreviouslv (1. 16).

Effective inhibitors: four-stranded
Ineffective inhibitors: one-or molecules" two-stranded moleculesb POlY ( poly(C), and poly(U) are not. After determining the primary structures of the type I and type I1 bovine macrophage scavenger receptors, we suggested that their common extracellular collagenous domain, which is highly cationic and well conserved among mammalian scavenger receptors,* was a likely site of polyanionic ligand binding (4, 5 ) . Recent experiments have confirmed this proposal (13, 14).' It seems likely that the spatial distribution of the positively and negatively charged side chains in this domain plays a critical role in distinguishing between polyanionic ligands and polyanions that do not bind (62).
In the current study, we have extended the list of scavenger receptor inhibitors to include specific types of oligodeoxyribonucleic acids, and we have determined the structural basis of the type I bovine scavenger receptor's specificity for particular polyribonucleic and oligodeoxyribonucleic acids. Short dG, oligonucleotides, where 5 5 n 5 37, and certain telomerelike oligonucleotides (e.g. d(G4T4)6) effectively inhibited type I bovine scavenger receptor activity. Experiments using transfected CHO cells expressing the type I1 bovine scavenger receptor gave similar results, although oligonucleotide inhibition of the type I1 receptors was generally not as potent as was inhibition of the type I receptors? Cell binding studies with [32P]dA6G37 established that this oligonucleotide, and presumably the other dG,s, bound directly to the type I scavenger receptor and suggest that these molecules functioned as competitive inhibitors. These oligonucleotides and all of the previously described polyribonucleotide inhibitors can form four-stranded helices stabilized by either inter-or intramolecular hydrogen-bonded base-quartets (e.g. G-quartets see Fig. 1) (1, 16,20-24, 28,34,49,50,60).
The inhibitory activities of these molecules were dramatically reduced when their four-stranded helical structures were disrupted, either by boiling in deionized water in the case of the dG,s or by Li' denaturation in the case of poly (1). The coordinate denaturation and loss of activity of poly(1) were reversible; its quadruplex structure and its inhibitory activity were restored by the addition of KC1 to the Li' salt. A set of control oligo-and polynucleotides which do not form, or are not expected to form, four-stranded helices (e.g. dA37, GXDNA) did not inhibit scavenger receptor activity. Thus, the formation of a base-quartet-stabilized four-stranded helix appears to be a necessary structural determinant for the activity of oligo-and polynucleotide inhibitors of scavenger receptors.
It has recently been reported that dGlo aggregates, which were presumably four-stranded, but not dClo, dAlo, or boiled dGlo, can bind to topoisomerase 11 and inhibit its activity (59). A comparison of the bovine scavenger receptor amino acid sequence with that of the human (63) or yeast (64) topoisomerases showed no significant h~mology.~ The topoisomerase I1 sequences do contain many clusters of basic residues which might interact with the phosphates in aggregated dGlo.
Three other proteins have been reported to bind to guaninerich oligonucleotides. The Oxytricha telomere-binding protein (65), the Oxytricha telomerase (56), and an avian telomerebinding protein (66) bind to oligonucleotides with telomeric sequences (e.g. d(T4G4)4). The two invertebrate proteins appear to differ from the scavenger receptor since they bind to the unfolded form of this telomeric oligonucleotide, rather than to the folded quadruplex form (56, 65). It has been suggested that the vertebrate binding protein recognizes G-G base pairs in telomere-like oligonucleotides (66). The scavenger receptor may provide an additional tool for the analysis of telomeric structures in natural nucleic acids. For example, the scavenger receptor may be able to detect quadruplex conformations in chromosomes.
It is important to note that the detailed three-dimensional structures of quadruplex oligonucleotides and polynucleotides can vary. The quadruplexes can be stabilized by intermolecular base-quartets between either parallel (20-22, 25, 60, 67)' or antiparallel (25, 32, 68, 69) extended polynucleotide strands. Alternatively, the strands can fold back on themselves to form quadruplexes stabilized by intramolecular or mixed intra-and intermolecular base-quartets (25, 26,28, 29, 33, 49). The orientations of the bases with respect to the sugar-phosphate backbones can be all anti in the case of parallel strands (20-22); or they can alternate between syn and anti when the strands fold back on themselves and/or are antiparallel (25,26,33,69, 70). The four polynucleotide chains are equidistant from each other when the strands are parallel. However, when the chains fold back on themselves, some of the sugar-phosphate chains are closer together and others are further apart, as in the d(G4T4G4) dimer (26) (Fig. lA), so that the helical grooves are alternately wide or narrow. Quadruplex structures can also vary depending on the ionic environment, the method of sample preparation, and sample concentration and history (compare Refs. 25-38, 49, 56, 67-69).",5,s For example, the telomeric oligonucleotide d(G4T4G4) has been reported to form a quadruplex with Na+ counterions whose structure differs in a number of ways from that of the quadruplex formed with K' counterions (26, 33).
The detailed structures of the quadruplexes used in this study varied considerably. Some consisted of large aggregates of noncovalently associated oligo-and polynucleotides, while others were clearly small monomeric structures. The poly(1) quadruplex comprises four individual and parallel polyribonucleotide strands (20-22, 60). It is likely that the individual strands are not in register. As a consequence, the ends of the quadruplexes might be free to fold back to form intramolecular antiparallel structures.
The telomere-like oligonucleotide d(T4G4)4 repeatedly folds back on itself to form an intramolecular quadruplex containing G-quartets and T loops (28, 34). Adjacent guanine segments are therefore oriented in an antiparallel fashion in this quadruplex. The telomere-like oligonucleotide d(G4T4)5 apparently forms at least two different quadruplex species whose detailed structures have not been determined.
The dG, oligonucleotides used in this study could have formed either parallel quadruplexes, like poly(1) and poly(G1, or folded back antiparallel quadruplexes, like the telomeric oligonucleotides, or perhaps a mixture of both. Recent studies of telomere-like oligonucleotides have demonstrated that circular dichroism provides a sensitive technique for distinguishing between parallel and antiparallel quadruplex conformations (36, 38):~~ A strong positive band centered near 260-265 nm in the CD spectrum correlates with a parallel quadruplex conformation (32, 34-36,38, 54):s' On the other hand, a strong positive band centered near 295 nm accompanied by a weak negative band near 265 nm correlates with an antiparallel conformation (26, 28, 32-34, 36, 38, 68, 69).4,5 These recent results are supported by CD studies of dG5 and poly(G) (23, 48), as well as more recent studies of several oligo dG,' s (50). Thus, our CD results indicate that the oligo dG, molecules used here were predominantly parallel-stranded quadruplex structures.
We have not yet determined whether all or only a subset of the many possible quadruplex conformations can inhibit scavenger receptor activity. Therefore, even though a fourstranded structure appears to be necessary for nucleic acid inhibition of the scavenger receptor, it may not be sufficient. There may be quadruplex-forming oligonucleotides which cannot inhibit scavenger receptor activity.
The results of this study suggest the possibility that the spatial distribution of negatively charged phosphates in oligoand polynucleotide quadruplexes can form a charged surface which is complementary to the positively charged surface of the collagenous domain of the scavenger receptor. This could account for the previously unexplained striking polynucleotide binding specificity of this receptor (1, 16) and our recent finding that the collagenous domain of complement factor Clq exhibits polyanionic ligand binding whose specificity is similar, but not identical, to that of scavenger receptors (7, 13). These results suggest that further analysis of the known structures of collagen and quadruplexes will provide additional insights into the bindmg properties of scavenger receptors.