Structural basis of citrate-dependent and heparan sulfate-mediated cell surface retention of cobra cardiotoxin A3.

Anionic citrate is a major component of venom, but the role of venom citrate in toxicity other than its inhibitory effect on the cation-dependent action of venom toxins is poorly understood. By immobilizing Chinese hamster ovary cells in microcapillary tubes and heparin on sensor chips, we demonstrated that heparan sulfate-mediated cell retention of the major cardiotoxin (CTX) from the Taiwan cobra, CTX A3, near membrane surfaces is citrate-dependent. X-ray determination of a CTX A3-heparin hexasaccharide complex structure at 2.4 A resolution revealed a molecular mechanism for toxin retention in which heparin-induced conformational changes of CTX A3 lead to citrate-mediated dimerization. A citrate ion bound to Lys-23 and Lys-31 near the tip of loop II stabilizes hydrophobic contact of the CTX A3 homodimer at the functionally important loop I and II regions. Additionally, the heparin hexasaccharide interacts with five CTX A3 molecules in the crystal structure, providing another mechanism whereby the toxin establishes a complex network of interactions that result in a strong interaction with cell surfaces presenting heparan sulfate. Our results suggest a novel role for venom citrate in biological activity and reveal a structural model that explains cell retention of cobra CTX A3 through heparan sulfate-CTX interactions.

Anionic citrate is a major component of venom, but the role of venom citrate in toxicity other than its inhibitory effect on the cation-dependent action of venom toxins is poorly understood. By immobilizing Chinese hamster ovary cells in microcapillary tubes and heparin on sensor chips, we demonstrated that heparan sulfatemediated cell retention of the major cardiotoxin (CTX) from the Taiwan cobra, CTX A3, near membrane surfaces is citrate-dependent. X-ray determination of a CTX A3-heparin hexasaccharide complex structure at 2.4 Å resolution revealed a molecular mechanism for toxin retention in which heparin-induced conformational changes of CTX A3 lead to citrate-mediated dimerization. A citrate ion bound to Lys-23 and Lys-31 near the tip of loop II stabilizes hydrophobic contact of the CTX A3 homodimer at the functionally important loop I and II regions. Additionally, the heparin hexasaccharide interacts with five CTX A3 molecules in the crystal structure, providing another mechanism whereby the toxin establishes a complex network of interactions that result in a strong interaction with cell surfaces presenting heparan sulfate. Our results suggest a novel role for venom citrate in biological activity and reveal a structural model that explains cell retention of cobra CTX A3 through heparan sulfate-CTX interactions.
Citrate is present as a major component of snake, bee, scorpion, and ant venom and serves as counter ion for the basic polypeptides of many venoms (1). For instance, ϳ50 mM citrate is present in the Taiwan cobra (Naja atra) venom (2). The role of citrate as a divalent cationic chelator that inhibits calcium ion-dependent enzyme activity in venom storage glands has been established (2). The function of citrate in the envenomed host, however, remains elusive. Most interestingly, citrate serves as a buffer component and often mediates the monomer-dimer equilibrium of many proteases for regulation of their catalytic activity by either stabilizing the active dimeric form or, indirectly, the protease oxyanion hole (3). Citrate has been used in many protein crystallization studies and has been observed to be located at functionally important anionic binding pockets. For instance, co-crystallization of citrate with the membrane phosphatidylinositol 3-phosphate (PIP) 1 -binding protein of Hrs has been shown to directly promote PIP dimerization through two FYVE tandem domains. It suggests a structural model of PIP/Hrs based on citrate binding conformation (4).
Cell surface retention of biologically active ligands through heparin or heparan sulfate (HS) binding plays an important role not only in certain disease states but also in cell development (5). Heparin and HS are a class of glycosaminoglycans (GAGs) that are widely distributed in many tissues and cell types (6). They are negatively charged polymers composed of heterogeneous disaccharide repeating units of uronate-glucosamine (IdoUA/GlcUA-␣1-4-GlcN). An uneven sulfation pattern, especially N-substitution differences, accounts for heterogeneity (7) and provides the structural basis for the specific binding of extracellular regulatory proteins to exert their biological functions. Glycosaminoglycans have also been suggested to play biological roles as regulators of the transport and effector functions of numerous extracellular regulatory proteins, including growth factors, proteases, lipoprotein lipase, chemokines, morphogens, and viral proteins (8 -14).
Cobra cardiotoxins (CTXs) are basic proteins with different GAG binding specificities conferred by a cationic belt of the conserved residues and the availability of specific Lys residues (15). They constitute ϳ50% of the venom by weight (16) and are composed of 60 -62 amino acids, with ␤-sheets forming three fingered loop structures (17). By binding to cell membranes, CTXs depolarize cardiomyocytes to cause systolic arrest in the envenomed victim. Most interestingly, heparin binding not only stabilizes the membrane-bound form of CTXs but also induces CTX aggregation (18 -20). Subcutaneous injection of cobra toxin in mice suggested that certain CTXs, unlike other venom components, remained at the sites of lesion by unknown molecular mechanisms (21,22). Considering that various homologous CTXs from the venoms of N. atra and other cobras, Naja nigricollis and Naja mossambica, have diverse potential biological targets in a wide range of cell membranes, it is of interest to investigate whether the interaction of GAGs with CTXs plays a role in mediating tissue retention of specific CTX types and whether other venom components are also involved in retention.

EXPERIMENTAL PROCEDURES
Materials-Crude snake venom (N. atra), heparin, and other chemicals were purchased from Sigma. CTXs were purified according to a published procedure (23) and were determined to be at least 99% pure by reverse phase-high pressure liquid chromatography. Protein concentrations were determined by BCA assay (Pierce). Anti-cardiotoxin antibody was prepared by immunoprecipitation of horse antiserum (Center for Disease Control Taiwan, Taipei, Taiwan) with CTXs.
For performing retention tests in microcapillary pipettes, trypsinized cells (5 ϫ 10 5 cells/ml) were pipetted into 160-l microcapillaries (VWR Scientific) that were maintained in tissue culture medium for 1 day prior to retention tests. A continuous flow system (10 l/min) consisting of a syringe pump (ISCO Inc., Lincoln, NE) connected to a microcapillary tube with silicon tubing was employed for retention assays. Cells in the microcapillary tube were washed with Hanks' balanced salt solution (HBSS) before CTX A3 application. Ten l of CTX (2 M in HBSS), segregated by two air bubbles to avoid dilution, were drawn through the microcapillary tube. The flow was stopped when the CTX A3 solution reached the outlet, and the cells were fixed with 4% (v/v) formaldehyde in PBS for 15 min followed by blocking with 1% (w/v) bovine serum albumin in PBS for 30 min. Control experiments with mouse anti-␤actin antibody and anti-mouse-IgG antibody conjugated to horseradish peroxidase were performed to ensure that the cells were evenly immobilized. An additional treatment with 0.1% (w/v) Triton X-100 in PBS was performed before the blocking step for control experiments. CTX A3 retention by the cells was detected by immunostaining with a horse anti-CTX-antibody (1:2000) and a secondary anti-horse IgG antibody (1:2000) conjugated with horseradish peroxidase (Abcam, Inc., Cambridge, MA). After washing away unbound antibody, a chemiluminescence detection reagent (ECL, Amersham Biosciences) was added.
The microcapillary tubes were then placed in contact with hyperfilm (Amersham Biosciences) for 5 min for detection of cell staining.
The distribution of staining along the microcapillaries was scanned and analyzed by image-editing software (Adobe Photoshop 7.01, Adobe Systems Inc., San Jose, CA) and brightness/contrast adjustment (Ϫ50/ ϩ50%). High background staining due to the nonspecific binding of antibodies could be readily subtracted to reveal the distribution of specific staining as shown in Scheme 1. The average trailing edge of staining from replicate experiments was estimated as the position of half-maximal intensity and was a measurement of the retention of CTX A3 on immobilized CHO cells.
Preparation of Heparin Derivatives-Biotinylated heparin was prepared by mixing high molecular weight (HMW) heparin with a 5-fold molar excess of NHS-LC-LC-biotin (Pierce). After 24 h at room temperature, the biotinylation reagent was removed by Sephadex G-15 chromatography (Amersham Biosciences).
Fully sulfated heparin hexasaccharide used for crystallization studies was prepared by heparinase I treatment of HMW heparin (24). Heparinase digestion results in oligosaccharides at their nonreducing termini with a common structure of a 4,5-unsaturated uronic acid derivative from either D-glucuronic acid or L-iduronic acid. It has the advantage of being a chromophore with maximum absorbance at 232 nm although it loses its original chemical identity. Strong anion exchanger chromatography (4.6 ϫ 250 mm; Waters, Milford, MA) was performed with a 50-min linear gradient, 0.7-1.2 M NaCl, pH 3.5, at constant flow rate (1 ml/min) with absorbance monitoring at 232 nm. The fraction of fully sulfated heparin hexasaccharide was isolated, and its purity was characterized further on a single peak by repeated strong anion exchanger chromatography, and the obtained NMR spectra were greater than 99%.
Surface Plasmon Resonance Binding Studies-Surface plasmon resonance (SPR) experiments were performed on a Biacore X instrument (Biacore AB, Uppsala, Sweden). All experiments were conducted at 25°C under continuous flow using various buffers as described under SCHEME 1. Retention test for CTX binding to immobilized cells. "Results." Biotinylated heparin was immobilized on SA chips (Biacore AB, Uppsala, Sweden) at a flow rate of 5 l/min to give surface densities of 700 RU. One RU corresponds to ϳ1 pg of immobilized heparin per mm 2 . Streptavidin surfaces were used as references. Typically, toxins were injected over heparin surfaces at a flow rate of 40 l/min under different buffer compositions (10 mM at pH 7.4, as indicated under "Results"), with 150 mM NaCl and 0.005% Tween 20. There was not any significant nonspecific binding to reference surfaces. Surfaces were regenerated by 30-s injections of 1.0 M NaCl. The amount of CTX binding at steady state (R eq ) was used to calculate the dissociation constants from Scatchard plots. The maximum capacities of heparin surfaces for CTX binding (R max ) were also estimated from Scatchard plots. The retained fraction (%) indicated the fraction of bound CTXs as retained forms. It was defined as follows: retained fraction (%) ϭ bound CTXs after 6 min of dissociation/bound CTXs before dissociation. The dissociation rate constants were estimated by using BIAevaluation version 3.2 (Biacore AB; Uppsala, Sweden). Several recent reviews (28 -30) have indicated that the dissociation rate constant is often difficult to measure using surface-based techniques because of the existence of a layer of stationary liquid at the surface, which promotes rebinding. Nevertheless, numerical analysis for quantitative comparison is still possible, if certain assumptions are met.
Crystallizations-Crystallization mixtures contained CTX A3 and heparin hexasaccharide in a molar ratio of 1:3 in 10 mM HEPES, pH 7.4, containing 200 mM NaCl. Precipitation immediately occurred when protein and heparin were mixed, and the solution became clear, and NaCl was added at a proper concentration. Crystallization was achieved using the hanging-drop vapor diffusion method at 291 K with 1-l hanging drops containing equal amounts of protein solution (10 mg/ml) and a reservoir solution containing 18% (v/v) isopropyl alcohol, 18% (w/v) PEG4000, and 100 mM trisodium citrate dehydrate, pH 5.6. The crystals were visible after 2 days and reached a maximal dimension of 0.2 ϫ 0.2 ϫ 0.1 mm 3 in 2 weeks.
Crystallographic Data Collection and Processing-By taking advantage of the 18% (v/v) isopropyl alcohol in crystallization solution, the crystals were mounted on a glass loop (0.2-0.3 mm) and directly flashcooled in liquid nitrogen without any other cryo-protectant solution. X-ray diffraction data to 2.4 Å resolution were collected at 110 K using ADSC Q210 CCD detector at a wavelength of 1.127 Å on beamline BL17B2 at the National Synchrotron Radiation Research Center, Hsinchu, Taiwan. The data were indexed, integrated, and scaled using HKL2000 (31). The crystals belonged to the orthorhombic space group C222 1 with unit cell dimensions of a ϭ 43.74 Å, b ϭ 59.25 Å, and c ϭ 98.55 Å. There were two molecules per asymmetric unit according to estimated solvent content in a reasonable region. Details of the data statistics are given in Table I.
Crystal Structure Determination and Refinement-The structure of the CTXA3-heparin complex was determined by molecular replacement (32) as implemented in CNS v1.1 (33) using the N. atra CTX A6 monomer structure (PDB code 1UG4) as a search model. Two CTX A3 molecules were located in the unit cell after rotation and translation function searches. All refinement procedures were performed using CNS v.1.1. The composite omit electron density maps with coefficients ͉2F o Ϫ F c ͉ were calculated and visualized using O v7.0 (34), and the model was rebuilt and adjusted iteratively as required. Throughout the refinement, a random selection (5%) of the data were set aside as a "free data set," and the model was refined against the rest of data with F Ն 0 as a working data set (35). The dimeric protein model was initially refined by rigid-body refinement using the data from 25.0 to 3.5 Å resolution, where the group B values were first restrained to 20.0 Å 2 . This refinement was followed by simulated annealing using a slow cooling protocol with a starting temperature 2500 K, provided in CNS, applied to data between 25.0 and 2.4 Å. The bulk solvent correction was then applied, and group B factors were adjusted. An elongated extra density stretching out from one CTX A3 molecule of the dimer was clearly visible and recognized as the hexasaccharide heparin in Aweighted ͉F o Ϫ F c ͉ maps. In order to enhance the density fitting, another x-ray diffraction data with a longer wavelength of 1.7412 Å was collected by the sulfur single-wavelength anomalous scattering (data were collected on the Taiwan contract beamline BL12B2 of SPring-8 in Japan; statistics not shown here). The sulfur anomalous difference Fourier maps showed clear sulfide peaks and allowed the heparin molecule to be adjusted and fitted into the electron density map. Moreover, considerable density was present between the dimer interface surrounded by 4 lysines (residue 23 and 31 from two molecules) and was well fitted with a citrate molecule that was used in crystallization buffer. The refinement then proceeded with another cycle of simulated annealing with a slow cooling protocol with a starting temperature of 1000 K. The heparin molecule was adjusted iteratively, and in the sugar unit the omitted electron density maps were followed by energy minimization and group B-factor refinement. Finally, the PICKWATER subroutine from CNS served to define peaks in difference maps (3 cut-off level) to locate 24 water molecules automatically.
Model Validation-The final model of CTX A3 and heparin complex contains 930 non-hydrogen protein atoms for the dimeric CTX A3, 105 hexasaccharide atoms for heparin molecule, and 13 citrate atoms. The refinement statistics are given in Table I. The correctness of stereochemistry of the model was verified using PROCHECK (36). The calculations of r.m.s. deviations from ideality (37) for bonds, angles, and dihedral and improper angles performed in CNS showed satisfactory stereochemistry. In a Ramachandran plot (38), all main chain dihedral angles were in the most favored and additionally allowed regions. The torsion angles for heparin hexasaccharide were also compared with available coordinates and were found to be consistent with the energy minima determined previously (39,40). The accessible surface area of CTX A3 was calculated by FastSurf (41). The van der Waals radius was assigned using a probe radius of 1.4 Å.
Coordinates-Atomic coordinates for the crystal structure of CTX A3 and heparin complex described in this paper have been deposited in the PDB (code 1XT3).

Retention of CTX A3 on Immobilized CHO Cells in Capillary
Tubes-CTX A3 binding to immobilized CHO cells was determined according to Scheme 1. As shown in Fig. 1, although there was no significant difference of the cell retention of CTX A3 by mutant cells (pgsA745 that are deficient in GAG biosynthesis) and wild-type cells (CHO-K1 expressing GAGs) in HBSS, the addition of 10 mM citrate significantly increased toxin retention by wild-type cells but not mutant cells. These results suggest that GAGs on CHO cell surfaces play a role in mediating the retention of CTX A3 in the presence of citrate.
Citrate-induced Retention of CTX A3 on Heparin Surfaces-As we have demonstrated previously that CTX A3 binds preferably to GAG heparin/HS, we carried out a more sensitive and quantitative binding analysis of the CTX-heparin interaction by surface plasmon resonance. As shown in Fig.  2A, the profiles obtained for CTX A3 binding to heparin surfaces is buffer composition-dependent. In HEPES and Tris buffers, binding reached equilibrium at higher toxin concentrations, and the toxin dissociated relatively rapidly with reasonable fitting to a monoexponential decay model, giving an observed dissociation rate constant of ϳ2 ϫ 10 Ϫ2 s Ϫ1 . In citrate and phosphate buffers, however, an initial fast association was followed by a much slower second phase of bind-ing when 0.1-6 M CTX A3 was injected over an immobilized heparin surface at a flow rate of 40 l/min. The dissociation phases in phosphate and citrate buffers were distinctly biphasic and could be best fitted to a double exponential decay model that gave apparent dissociation rate constants of 6 ϫ 10 Ϫ2 s Ϫ1 for fast dissociation components and 5 ϫ 10 Ϫ4 s Ϫ1 for slow dissociation components. The huge difference (ϳ100folded) in rate constants for the two dissociation phases indicates the presence of two distinct populations of CTX A3 molecules on the heparin surface.
Plots of the accumulated responses versus log concentra- Venom Citrate-dependent Cell Retention of Toxins tion (Fig. 2B) were characterized by points of inflection indicating that the heparin surface was saturable and that the accumulation of CTX A3 was associated with heparin binding. Scatchard analyses of the binding data were highly unusual with the lower concentration data giving positive slopes for all four buffers. The fact that equilibrium was only approached in some instances does not significantly contribute to these unusual profiles. The positive slopes are indicative of positive cooperativity observed at lower CTX A3 concentrations. The negative slope regions of the Scatchard plots indicated that phosphate and citrate buffers increased the observed dissociation constants for CTX A3 binding to heparin. Based on the least square linear fitting of the data, the observed dissociation constants are estimated to be 0.37 Ϯ 0.03 and 2.01 Ϯ 0.09 M in the presence of HEPES and citrate buffer (10 mM), respectively. The observed slow dissociation rates are interpreted as being due to the citrate-induced oligomerization of CTX A3. However, we found no evidence for citrate-induced oligomerization of CTX A3 in the absence of heparin by diffusion constant measurements by both NMR and analytical ultracentrifugation of CTX A3 in the presence of citrate (data not shown). Because we have demonstrated previously that the interaction of heparin with CTX A3 stabilized the membrane-bound conformation of CTX A3 (18), citrate-induced oligomerization probably occurs only in the presence of heparin binding through heparin-induced CTX A3 conformational change. Further SPR experiments were therefore performed in the presence of citrate to understand the molecular mechanism and biological significance.
Characterization of CTXA3 Retention in the Presence of Citrate-The retention of CTX A3 on heparin surfaces was measured as a function of citrate concentration. As shown in Fig. 3A, the amount of CTX A3 retention (as reflected by the slow dissociation phase indicated by the arrows shown in Fig. 3A) and the retained fraction relative to total binding significantly increased with increasing citrate concentration. However, in the presence of citrate, higher CTX A3 concentrations are needed in order to achieve similar amounts of CTX A3 binding  N. atra, Chinese cobra (Southeastern Asia); N. mossambica mossambica, Mozambique cobra (Eastern Africa); Naja kaouthia, monocled cobra (Southern Asia); Naja naja naja, Indian cobra (Southern Asia); N. nigricollis, spitting cobra (Central and Eastern Africa); Naja haje haje, Egyptian cobra (Northern Africa).

Venom Citrate-dependent Cell Retention of Toxins
as evidenced by the increasing CTX A3 concentrations needed to achieve ϳ2000 RU of binding. This observation is consistent with the Scatchard plot analyses of the data for binding in different buffers, indicating that citrate ions reduced the binding of CTXA3 to heparin (Fig. 2C).
Reducing the flow rate for the injection of a constant volume and constant concentration of CTX A3 over heparin surfaces in citrate-buffered saline increased the amount of CTX A3 retention, indicating that retention was time-dependent (Fig. 3B). To rule out that this was a flow rate effect, increasing volumes of a constant CTX A3 concentration were injected over a heparin surface at constant flow rate (Fig. 3C). Increasing amounts of retention were observed with increasing volume, again indicating that retention was time-dependent.
Overall Structure of CTX A3 Dimer-In order to understand the action mechanism of citrate-dependent, heparin-mediated cell retention, CTX A3 was co-crystallized with fully sulfated heparin hexasaccharide in the presence of citrate ions, and the three-dimensional structure of CTX A3 in complex with hexasaccharide was determined at 2.4 Å resolution. As shown in Fig.  4A, there are two CTX A3 molecules with pseudo 2-fold symmetry, one citrate and one heparin hexasaccharide per asymmetric unit. The crystal structure of CTX A3 shows the general three-fingered CTX folding and contains five ␤-sheets compris-ing residues 2-4 (␤1), 11-13 (␤2), 20 -26 (␤3), 35-39 (␤4), and 49 -54 (␤5), tightened by four disulfide bonds (Table II). Three functional loops are formed by residues 4 -11 (loop I), 26 -35 (loop II), and 39 -49 (loop III) (Fig. 4A). Dimer formation buried ϳ1100 Å 2 of accessible surface area of the two monomers with ϳ300 Å 2 of polar and 800 Å 2 of nonpolar contact areas. The major contacts of the dimer are from the hydrophobic residues at both the loop I and loop II regions, namely Leu-6, Val-7, Pro-8, Leu-9 of loop I and Val-32 and Pro-33 of loop II.
One citrate ion (an essential component in the crystallization buffer that contained 0.1 M sodium citrate) resided at the interface of two monomers near Lys-31 and Lys-23 of monomers A and B (Fig. 4B). The citrate ion occupied a strongly cationic pocket generated by CTX A3 dimerization (Fig. 4C). Both Lys-23 and Lys-31 stabilized the citrate ion with electrostatic interaction distances of ϳ3 Å. Although Lys-23 is conserved through all the CTX homologues (Table III), CTXs can be classified into two groups based simply on the presence and absence of Lys-31 near the tip of loop II. In our previous crystal structure of CTX A3 in complex with SDS, Lys-31 is the only positively charged residue of all three CTX molecules within the asymmetric unit that was observed to bind SDS (43). The significance of this observation will be discussed later in connection with the suggestion that the citrate-binding pocket is FIG. 5. Conformational changes induced in CTX A3 by heparin binding. A, structural superposition of two CTX monomers to show the effect of heparin binding on CTX conformation. Both convex (left) and concave (right) sides of the CTX molecules are shown for clarity. B, plot of structural variation (differences in r.m.s. derivation in Å) between two monomers as a function of amino acid positions. Main chain variation occurred mostly around Cys-14 as also indicated in Table II. The side chain variations were determined as the differences between the r.m.s. derivations of the overall structure and the main chain. also a potential binding site for negatively charged lipids. 2 Two positively charged clusters on the CTX A3 dimer (Fig.  4D), identified previously by an NMR study of heparin disac-charide binding to CTX A3 (44), constitute a potential long chain heparin-binding site (Fig. 4A). However, the heparin hexasaccharide only makes contact with monomer A at the convex heparin-binding site in each unit cell (Fig. 4A). The other ligand-free positively charged cluster is located on the same side of the CTX A3 dimer at a distance of 25-30 Å.  1AXM, 1AZX, 1BFB, 1BFC, 1E0O, 1E03, 1FQ9, 1G5N, 1GMN, 1GMO, 1HPN, 1NQ9, 1QQP, and 1TB6)  Because our SPR studies use high molecular weight heparin (typically with average chain length of ϳ250 Å), a further extension of heparin chain length by 6 -7 disaccharide repeats could help explain the observed citrate-dependent retention of CTX binding.
Heparin-induced Conformational Change of CTX A3-Comparison of the two CTX A3 monomers shows an overall r.m.s. deviation of 1.17 Å over 60 residues. The most significant structural variations of the backbone main chains between the two monomers (monomer A with bound heparin and monomer B without bound heparin) occur at CTX core surrounding Cys-14 and Cys-38. In fact, a heparin-induced conformational change is observed for the Cys-14 -Cys-38 disulfide bridge as reflected by the significant change (Ͼ30°) of the disulfide bond torsion angles (Table II). Previously, by comparing the nuclear Overhauser effect intensity change in a 1 H NMR study of heparin binding to the CTX A3, we showed that a significant heparin-induced conformational and/or dynamic change occurs near the Cys-38 core region (44). Our current x-ray structure further suggests that heparin-induced conformational change of CTX A3 at the core region indeed occurs perturbing the ␤-sheet structure.
In addition to the conformational change of the main chain, there are also several heparin-induced changes in side chain orientation as indicated in Fig. 5. Basically, two major types of structural perturbation are observed, one associated mainly with residues, such as Lys-18, Lys-44, Lys-23, and Lys-31, involved in electrostatic interaction with either heparin or citrate (Fig. 6) and the other associated with residues involved in hydrophobic side interactions such as Leu-9 and Met-26 located at the tips of loops I and II. Most interestingly, the NH chemical shifts of Leu-9, Met-26, and Val-34 were also observed to undergo the most significant heparin-induced changes in a previous NMR study (18). Because both x-ray and NMR methods detect a similar heparin-induced conformational change in both the main chain and side chains, our results strongly suggest that the heparin-induced conformational change observed in the crystalline state may also occur in solution (18).
Bound Conformation of Heparin Hexasaccharide-The electron density of fully sulfated hexasaccharide at 2.4 Å resolution (Fig. 6A) is clear enough to allow us to determine its molecular orientation and atomic contacts relative to CTX A3 (Fig. 6C). The heparin hexasaccharide binds to monomer A through the first three carbohydrate residues from the nonreducing end, which is consistent with the fact that the reducing end of heparan sulfate is linked to the core protein of proteoglycans. The overall crystal packing showed, however, additional hexasaccharide contacts, at both the reducing and nonreducing ends, with neighboring CTX A3 dimers (Fig. 6C). The ring conformations of GlcN residues in heparin hexasaccharide are all 4 C 1 . The first and second IdoUAs are in the 2 S 0 and 1 C 4 conformation, respectively, whereas the terminal uronate adapts a 1 H 2 conformation. All the ring conformations fall into the predicted energy minima of conformational equilibria (45). The plot of torsional angles (⌽ and ⌿) of the determined heparin conformation allows a further check of the obtained structure (Fig. 6B). By taking 26 available heparin coordinates from 14 protein data bank structures, we obtained the torsion angles for 55 GlcN-UA and 57 UA-GlcN linkages. The average of torsion angles (⌽ and ⌿) with standard deviation was found to be 83 Ϯ 20 and 98 Ϯ 18 for GlcN-UA and Ϫ73 Ϯ 13 and 133 Ϯ 16 for UA-GlcN linkage. Our heparin-bound conformation is therefore consistent with most conformations determined by either NMR or x-ray methods.
Close examination of electrostatic interactions showed the basis of the specificity of the heparin-CTXA3 interaction (Fig.   6C). Of the sulfate groups, one N-sulfate (on GlcN3 as labeled in Fig. 6C) interacted simultaneously with Lys-18, Lys-12, and Lys-35 of monomer A and Lys-44 of monomer B. The 2-Osulfate groups were also important because three contacts existed for Lys-18, Lys-12, and Lys-35 of monomer A. Only one interaction was observed for 6-O-sulfate with Lys-5 of neighboring monomer B. The carboxylate groups also play an important role in CTXA3-heparin interaction as depicted in the figure.
Sulfate Specificity as Investigated by SPR Competition-The aforementioned binding specificity suggested by crystal structure was investigated in the solution state by SPR competition experiments using chemically modified heparins. As shown in Fig. 7, in the absence of citrate, a dramatic decrease (ϳ80-fold by comparing the IC 50 values) in competitive activity was observed for de-N-sulfated heparin, reconfirming the importance of N-sulfation in the CTX A3-heparin interaction. However, in the presence of citrate, the apparent binding selectivity of CTX A3 for N-sulfation was reduced because only an ϳ33-fold reduction in IC 50 was observed for de-N-sulfated heparin.
The sulfate dependence of the CTX A3-heparin interaction showed the trend of N-sulfate Ͼ 2-O-sulfate ϳ 6-O-sulfate. Removal of 2-O-sulfate and 6-O-sulfate reduces the apparent binding affinity as indicated by an ϳ3-fold increase in IC 50 . The addition of citrate increases its binding selectivity as indicated by the relatively larger increase in IC 50 (ϳ6-fold). The fact that citrate could confer more selectivity for 2-O-and 6-O-sulfate suggests that the binding mode between heparin and CTX A3 may be different in the presence and absence of citrate. The conclusion is also consistent with our studies on heparin binding by various CTX homologues, suggesting that charged residues near the tip of loop II are involved in the binding specificity between CTX and heparin (20). In these competition experiments, we could not distinguish between the two populations of CTX A3, i.e. the fast and slow association/dissociation components. This probably explains why the binding specificity of 2-O-sulfate over 6-O-sulfate as suggested by crystal structure could not be verified in these competition experiments. DISCUSSION Despite the presence of high concentrations of citrate in most animal venoms, there are no reports on the potential involvement of venom citrate in the action of toxins present in venoms. In this study, we developed a new method for studying the cell retention behavior of animal toxins, and we showed that citrate played a role in the HS-mediated cell retention of CTX A3. Presumably, cell retention mediated by HS and citrate allows the toxin to remain on the cell surface long enough to exert its action, whereas binding specificity mediated by HS recruits different toxin homologues to different locations. The combined usage of the SPR method and the retention test developed in this paper provides a useful approach to address the biological function of protein-HS interaction near the membrane surface.
CTX-Heparin-Citrate Lattices-The cell binding, SPR, and structural data presented here provide a clear understanding of how several weak interactions in combination lead to an almost irreversible retention of CTX A3 on the cell surfaces of snake bite victims. A single heparin chain can bind multiple CTXA3 monomers. CTXA3 binding to heparin induces a conformational change in CTX A3 and results in citrate binding and CTX A3 dimerization. Each CTXA3 dimer can simulta-neously engage up to five different heparin chains observed in crystal packing. The end result of these binding events is the establishment of a CTX A3 lattice on cell surfaces. SPR indicated that positive cooperativity played a role in lattice development and that a threshold level of toxin was required. SPR analyses also demonstrated that the initial CTX binding was random and that rearrangement of CTX A3 molecules led to the formation of the lattice over several minutes. Two Types of CTXs-As shown in Table III, cobra CTXs can be simply classified into two classes depending on the nature of the specific amino acid residues near the tip of the functionally important loop II region. In fact, preliminary cell retention and SPR studies of various CTX homologues indicate that the second group of CTXs in Table III, in which substitution of Lys-31 with Th-r31 or Met-31 introduces a continuous patch near the tip of loop II, exhibits citrate-independent cell retention behavior. 3 Previous x-ray crystallography studies of toxin-␥ further revealed that Met-31 residues are involved in the protein contact of toxin-␥ in the crystal (46). It is therefore possible that our proposed model of HS binding to CTX dimer may also be applicable to other CTXs without Lys-31 in this region. A single amino acid residue, i.e. Lys-31 or Met-31, or Thr-31, in CTXs may play an important role in deciding the specific types of cell target depending on the dimerization capability of the CTX homologues and the availability of citrate.
Heparin-induced CTX Dimerization-As shown in the schematic diagram of Fig. 8A based on the determined x-ray structure, heparin hexasaccharide derivatives can bind to one CTX FIG. 8. Representations of cis and trans binding of heparin by CTX A3 dimers and of trans complex bound to SDS. A, cis form of heparin-induced CTX A3 dimer based on asymmetry unit of CTX A3-heparin-citrate complex. CTX, heparin, and citrate are shown in blue, red, and pink, respectively. B, trans form of heparin-induced CTX A3 dimer based on crystal packing of CTX A3-heparin-citrate complex. The colors of molecules are the same as in A. C, trans form of heparin-CTX A3 dimer complex formation based on D1 dimer packing in CTX A3-SDS complex (PDB code 1H0J). Modeling of CTXheparin-SDS complexes were performed on a Silicon Graphics O2 work station with Insight II 2000 (Accelrys Co., San Diego). Fully sulfated heparin hexasaccharide was adapted from a published PDB structure (PDB code 1PHN). Energy calculations were performed with modified Amber force field with sulfate parameter extension (42). Initial position of heparin hexasaccharide was manually docked with the evaluation of lowest intermolecular energy by Docking module of Insight II. The sulfate and carboxylate groups of hexasaccharide at reducing and nonreducing ends were located at heparin-binding pockets of CTXs. Energy minimization was then performed by Discover. CTX, heparin, and SDS are shown in blue, red, and green, respectively. The membrane interface was arbitrarily picked based on the hydrophobic region of CTX A3 surrounded by bound SDS in D1 dimer of CTX A3-SDS complex.