Identification of protein recognition elements within heparin chains using enzymatic foot-printing in solution and on-line SEC/MS

Abstract

Understanding molecular mechanisms governing interactions of glycosaminoglycans (such as heparin) with proteins remains challenging due to their enormous structural heterogeneity. Commonly accepted approaches seek to reduce the structural complexity by searching for “binding epitopes” within the limited subsets of short heparin oligomers produced either enzymatically or synthetically. A top-down approach presented in this work seeks to preserve the chemical diversity displayed by heparin by allowing the longer and structurally diverse chains to interact with the client protein. Enzymatic lysis of the protein-bound heparin chains followed by the product analysis using size exclusion chromatography with on-line mass spectrometry detection (SEC/MS) reveals the oligomers that are protected from lysis due to their tight association with the protein, and enables their characterization (both the oligomer length, and the number of incorporated sulfate and acetylate groups). When applied to a paradigmatic heparin/antithrombin system, the new method generates a series of oligomers with surprisingly distinct sulfation levels. The extent of sulfation of the minimal-length binder (hexamer) is relatively modest yet persistent, consistent with the notion of six sulfate groups being both essential and sufficient for antithrombin binding. However, the masses of longer surviving chains indicate complete sulfation of disaccharides beyond the hexasaccharide core. Molecular dynamics simulations confirm the existence of favorable electrostatic interactions between the high charge-density saccharide residues flanking the “canonical” antithrombin-binding hexasaccharide and the positive patch on the surface of the overall negatively charged protein. Furthermore, electrostatics may rescue the heparin/protein interaction in the absence of the canonical binding element.

Graphical Abstract

Introduction

Heparin is one of the most versatile biopolymers, capable of binding an impressive variety of other biomolecules: its interactome^1–3 consists of an astonishing number of proteins, by the most recent count approaching eight hundred.⁴ The vast majority of these interactions are not related to heparin’s biological functions (its raison d’être is maintaining tight packing of histamine and proteases within mast cells,^5,6); indeed, heparin’s versatility is largely a reflection of its structural similarity to another member of the glycosaminoglycan family (and, in fact, its next-of-kin), heparan sulfate.^7–9 However, this redundancy is hardly relevant vis-à-vis successful exploitation of the heparin’s ability to interact with and modulate the behavior of a range of proteins with high therapeutic value,^10,11 be it direct interaction with therapeutic targets¹² or utilization of heparin in drug delivery systems.¹³ Presently, activation of antithrombin (AT) by heparin is probably the best known (and certainly the most extensively studied) example of such therapeutically relevant interactions.¹⁴ Successful exploitation of heparin’s versatility for therapeutic purposes critically depends on the ability to characterize its interactions with the target proteins; however, the specific molecular mechanisms remain elusive outside of the very few extensively investigated systems.¹⁵ While the enormous degree of structural heterogeneity exhibited by heparin (and, to even a greater extent, heparan sulfate)¹⁶ is the main factor making structure/activity studies challenging, it is likely to be the key to versatility of these polysaccharides, allowing them to associate with a diverse range of targets.

Deciphering mechanistic details of such interactions critically depends on the ability to identify a set of segments within heparin chains that act as recognition elements for a specific protein partner. Although it might be tempting to view these elements as binding epitopes, they are not necessarily represented by unique structures. Even though recognition of some proteins may be driven primarily by well-defined structural motifs with a high degree of complementarity towards the binding interface on the protein surface, contribution of less specific interactions (such as long-range electrostatics) may result in multiple sequences having comparable affinities to a specific partner protein. Furthermore, the very notion of a “binding segment” may contain a certain degree of ambiguity, as it assumes negligible contribution to the binding process from neighboring structural elements. Given the highly polyanionic character of heparin and the long-range nature of electrostatic interactions, the validity of this assumption may be far from certain. In contrast to the deterministic views of protein/heparin interactions being driven by structural complementarity, a more nuanced view is now becoming increasingly accepted, with binding activity towards a particular target not necessarily residing in a unique heparin sequence, but rather being distributed among a limited number of potential binding segments.² Furthermore, the realization that the vast majority of the heparan sulfate client proteins also associate with structurally similar, but more extensively sulfated heparin gave rise to the notion of “hidden specificity” (implying that the critical sulfate groups – constituents of the “protein-binding epitopes” within the heparan sulfate chains - remain accessible to the protein, while the “surplus” of the sulfates do not interfere with the binding).^15,17

A common approach to identifying protein-binding segments within heparin and related glycosaminoglycans relies on partial degradation of the intact polysaccharide chains (depolymerization) to obtain shorter oligosaccharides, a step that is frequently followed by size separation to obtain fixed-length fractions. Fractionation of such relatively short oligoheparins with respect to their affinity towards a specific protein target produces subsets of oligosaccharides in which the degree of structural heterogeneity has been sufficiently reduced to enable meaningful structural analyses.¹⁸ This strategy had led to several successes, most notably in identifying the heparin segment with a high affinity to AT,^19,20 eventually leading to the design of a short (pentasaccharide) synthetic analog with high anti-thrombotic activity, GlcNS,6S-GlcA-GlcNS,3S,6S-IdoA2S-GlcNS,6S-Me²¹ (Scheme 1, top). However, despite this resounding success, the “divide and conquer” strategy is not without limitations. It is now realized that the reliance on relatively short oligosaccharides narrows the search to sequences representing only a limited subset of the structural variety exhibited by intact heparin.² For example, commercially available heparinases (enzymes that use the elimination mechanism to cleave the glycosidic linkage between hexosamines and uronic acids, and commonly employed for heparin de-polymerization) have a common feature of being less effective in lysing the glycosidic bonds in tetrasaccharides containing the 3-O-sulfo glucosamine reducing end of the scissile bond (Scheme 1, middle)²² Clearly, enzymatic depolymerization in this case introduces a structural bias, exemplified by finding abundant longer heparin segments carrying multiple 3-O-sulfo groups,²³ modifications that are considered to be rare in heparin. Similar selectivity can occur in chemical depolymerization.²⁴ Above and beyond the structural bias introduced during the de-polymerization stage, focusing on short heparin segments neglects interactions outside of the primary binding sites that may also be important for the protein/heparin association. For example, in the case of AT/heparin interaction, even though the pentasaccharide sequence binds to the protein and activates it, longer chains have higher AT affinity and are more efficient AT activators.¹⁴ Explanations of these phenomena are frequently sought in the contribution of the nearest neighbors, focusing on the possibility of additional saccharide/amino acid contacts stabilizing the entire complex.²⁵ However, emphasizing the short-range interactions (viewed through the prism of steric complementarity) may ignore the interactions that are less specific but nonetheless essential for the association of the two biopolymers. In particular, the polyanionic nature of heparin is frequently overlooked,²⁶ despite the available evidence highlighting the role of charge density as an important determinant of heparin affinity to a range of proteins, including AT.²⁷

Scheme 1.

Chemical structures of (from top to bottom) the synthetic pentasaccharide AGA*IA_M (GlcNS,6S-GlcA-GlcNS,3S,6S-IdoA2S-GlcNS,6S-Me), the canonical AT-binding hexasaccharide (ΔHex-GlcNAc,6S-GlcA(IdoA)-GlcNS,3S,6S-IdoA2S-GlcNS,6S), and the generic structure of heparin (R₁ = H or SO₃, and R₂ = H, OCH₃ or SO₃). The sulfate and acetylate groups of the hexamer highlighted in red and blue, respectively, are essential for AT binding, while the sulfate group highlighted in orange is not essential; green arrows indicate the heparinase II cleavage sites (the presence of the sulfate group at the 3-O position makes the digestion inefficient).

Clearly, the prevailing approaches to deciphering the heparin structure/bioactivity relationship need to be modified to move beyond the fragment-based discovery paradigm and account for multiple factors driving the recognition and binding of proteins by heparin (and, by extension, glycosaminoglycans in general). In this work we present a new method that allows binding-competent segments of different lengths to be identified simultaneously within long heparin chains. Application of this method to the paradigmatic heparin binder AT indicates that the minimal-length binders are hexasaccharides whose overall sulfation/acetylation levels are in agreement with the consensus structure of the pentasaccharide AT-binding core.²⁸ However, the analysis of the sulfation levels of longer binders provides evidence for the importance of electrostatic interactions in the protein recognition process. Based on the experimental results, we propose a model of protein/heparin interaction in which the protein-binding element of the latter is steered towards the cluster of positively charged residues on the surface of the protein by long-range electrostatic forces. Once the initial encounter is made, the structural fragments projecting the basic residues reorganize to offer a more compact presentation of the positive patch, and short-range interactions take over, with steric complementarity between the binding partners becoming critical for the stable complex formation. Long-range (and less specific) electrostatic interactions between the highly charged saccharide residues flanking the “canonical” AT-binding hexasaccharide core and the positive patch on the surface of the overall negatively charged protein provide additional reinforcement to protein/heparin complex. Furthermore, electrostatics may rescue the protein/heparin association even in the absence of the highly complementary binder (the pentasaccharide segment).

Materials and methods

Materials.

Human antithrombin (AT) was purchased from Haematologic Technologies (Essex Junction, VT), and heparinase II (2000 U /mL) from Bacteroides eggerthii was purchased from New England Biolabs (Ipswich, MA). Fixed-length heparin oligomers were purchased from Iduron (Alderley Edge, UK). The low-molecular weight heparin Tinzaparin^™ (Leo Pharma, Ballerup, Denmark) was provided by Prof. P.L. Dubin (UMass), and the synthetic pentasaccharide (Fondaparinux) was provided by Prof. Robert J. Linhardt (Rensselaer Polytechnic Institute, Troy, NY). All other chemicals and solvents used in this work were of analytical grades or higher.

Limited heparin digestion.

Heparinase II was diluted to 200 U/mL into reaction buffer (100mM NaCl, 1.5 CaCl₂ and 20 mM Tris-HCl, pH 7, provided by New England Biolabs). Aliquots were stored at −80°C prior to use. Limited digestion of long heparin chains (dp20) in the absence of AT was carried out by mixing 20 μg of a 2 mg/mL aqueous solution of dp20 with an appropriate amount of heparinase II (1 unit enzyme to 10 μg heparin oligomer), and adding this mixture to a 30 mM ammonium acetate solution (pH 6.9) that also contained 2mM CaCl₂, to a final volume of 100 μL. Incubation of this mixture in a water bath at 30°C for a specific period of time (40 min) was followed by addition of 40 μg of AT and then continuing the incubation for an additional 15 min at 21°C prior to SEC/MS analysis. Digestion of long-chain heparin oligomers (dp20 and Tinzaparin) bound to AT was initiated by preparing an AT/heparin oligomer mixture (approximate molar ratio 1:5) in a 30 mM CH₃CO₂NH₄/2mM CaCl₂ solution (pH 6.9) and incubating at room temperature for 15min prior to adding heparinase (1 enzyme unit per 10 μg of heparin oligomers). Incubation of this mixture in a water bath at 30°C for a specific period of time (from 40 min to 4 hrs) prior to SEC/MS analysis.

SEC-MS Measurements.

Separations of the enzymatically digested mixtures were carried out using an HP1100 (Agilent, Santa Clara, CA) HPLC system equipped with a TSKgel 2000SW_XL (Tosoh, Tokyo, Japan) size exclusion column. A 30 mM ammonium acetate solution (pH 6.9) was used as a mobile phase at a flow rate of 0.3 mL/min; a 1:10 flow splitter was used to introduce the eluate into the ESI source of the mass spectrometer. The measurements with post-column denaturation utilized a second HPLC pump (HP1100) to mix in H₂O/CH₃OH/HCO₂H (47:47:6, v:v:v) at a flow rate of 0.3 mL/min into the eluate prior to the flow split. All online MS measurements were carried out using a SolariX 7 (Bruker Daltonics, Billerica, MA) Fourier transform ion cyclotron resonance mass spectrometer equipped with a 7.0 T superconducting magnet and a standard ESI source. The following ion source parameters were used to minimize/eliminate AT/heparin oligomer complex dissociation and sulfate loss from heparin oligomers in the gas phase: dry gas temperature, 200°C; collision voltage, −10 V; DC extract bias, 0.4 V; skimmer 1 potential, 20 V. Negative ion mode was used to detect heparin oligomers following post-SEC column denaturation.

Molecular Dynamics simulations.

All computational procedures were carried out using Schrödinger software package v. 2016–1 (Schrodinger LLC, New York, NY). The model of the AT/dpX (X = 6 and 8) complex was prepared using a human AT/synthetic pentasaccharide complex (PDB 1E03) as a template. The dp6 model was created by extending the pentasaccharide with one saccharide unit from the non-reducing end and dp8 models are created by extending the dp6 with a fully sulfated disaccharide at reducing and non-reducing ends, respectively. Several variants of dp6 and dp8 were created by altering the number and location of the sulfate groups. Restricted minimization of the resulting complex (OPLS3 force field) provided a starting structure for subsequent simulated annealing experiments. The 6-ns simulation was set up using a neutralized system (with 11 and 14 Na⁺ ions placed in a close proximity to the sulfate groups of the hexasaccharides and octasaccharides, respectively) with explicit water and 30 mM NaCl ions at 300K. The final frame of the simulation was subjected to restrained minimization using the OPLS3 force field to provide a model for the AT/dpX complex. The MD trajectories were processed using Simulation Event analysis to provide energetic parameters.

Results

Our top-down approach to identifying the protein-binding segments within heparin chains was informed by the classical foot-printing method of localizing protein-binding segments within nucleic acids.²⁹ The glycosidic bonds connecting the protein-bound disaccharide units were expected to be inaccessible to heparinase and, therefore, protected from the lysis, while the regions outside of the protein-bound segment should not be afforded such protection. Interestingly, the idea of using the enzymatic degradation of protein-bound heparin chains as a means of identifying the protein recognition segments³⁰ precedes the introduction of the DNA foot-printing method. However, the analytical technologies of the day were inadequate for performing the task, and the foot-printing method had been abandoned in favor of the now-classic approach where the binders are identified within oligoheparin libraries produced by either chemical or enzymatic degradation of the intact glycosaminoglycan chains.

Our initial experiments in identifying the protein-binding segments within the heparin chains were conducted with relatively long fixed-length heparin oligomers (eicososaccharides, or dp20). Increasing the exposure time of the AT/dp20 mixture to heparinase II (HepII), the most promiscuous of commercially available heparinases,³¹ resulted in a noticeable change of the SEC retention time for the protein/heparin oligomer complex (Figure 1A). This behavior clearly indicates continuous degradation of the heparin oligomer chains that remain bound to the protein, although it is impossible to determine the “survivor” chains’ size(s) and sulfation level(s) based on the elution time alone. In order to characterize the AT-bound heparin oligomers, we supplemented SEC separation with on-line detection by native mass spectrometry (SEC/MS).³² A mass spectrum averaged across the early-elution chromatographic peak following the shortest (10 min) digestion contains a large number of poorly resolved features, which is indicative of a wide distribution of AT-bound heparin species (Figure 1B). Increasing the digestion time to 40 min gives rise to several discernable features in the mass spectrum, and a series of shorter fragments can be identified ranging from a hexamer (6,6,1) to dodecamer (12,15,1) using Henriksen’s nomenclature, with the numbers in each triad (X,Y,Z) designating the oligosaccharide length (X), its sulfation level (Y) and N-acetylation count (Z).³³ Importantly, the structure of these four oligomers can be represented as (6+2n,6+3n,1), where n = 0 – 3, i.e. all longer-chain survivors relate to the minimal-length binder (a hexasaccharide with one acetyl and six sulfate groups) by addition of disaccharides containing three sulfate groups. The extracted ion chromatograms of AT complexes with these four heparin oligomers (inset in Figure 1A) show elution profiles that are sufficiently different to rule out gas-phase fragmentation of glycosidic bonds as a possible cause for the appearance of this series of ions in the mass spectra.

Figure 1.

SEC/MS analysis of the products of the AT/dp20 mixture incubation with heparinase II. A: UV chromatograms of the mixture following 10-min (wheat), 40-min (gray), 2-hr (red) and 4-hr (blue) exposure to the enzyme (the reference chromatogram of heparinoid-free AT is shown in black). The inset shows extracted ion chromatograms for free AT and its complexes with heparin fragments of varying lengths following 40-min lysis. The chromatographic peaks within the 28–31 min elution window correspond to tetra-disaccharide fragments. B: on-line mass spectra (averaged across the 19–21 min elution window) of the AT/dp20 mixture following 10-min (wheat), 40-min (light gray from 19–20 min and dark gray from 20–21 min), 2-hr (red) and 4-hr (blue) exposure to heparinase II (the reference mass spectrum of heparinoid-free AT is shown in black). The inset shows zoomed views of several representative mass spectra (z = +16). AT/(6,6,1) complexes produced by different protein glycoforms are labeled with blue circles. Designation of the heparin oligomers follows Henriksen’s nomenclature.³³

Longer exposure of the AT/dp20 mixture to HepII (2 and 4 hrs) leads to a near-complete disappearance of all but one AT/heparin oligomer adducts, with (6,6,1) remaining the “sole survivor” still bound to the protein (the AT complexes with the (8,9,1) and (10,12,1) heparin species were barely detectable, while the (12,15,1) adduct could not be detected). No protein complexes with shorter chains, such as di- or tetra-saccharides were ever detected in the AT-bound form. Similar results were obtained when the fixed-length heparin oligomer dp20 was substituted with commercial low-molecular weight heparin, Tinzaparin. The latter is a mixture of heparin oligomers of varying length;³³ its average molecular weight (ca. 6,500 Da³⁴) corresponds to heparin chains containing 22–26 saccharide units). Despite exhibiting dramatically higher structural heterogeneity compared to dp20, SEC/MS analysis of the products of a 4-hr digestion of AT/Tinzaparin mixture with HepII also reveals the presence of only two abundant ionic species, free protein and its complex with the (6,6,1) heparin species (Supplementary Material).

The remarkable homogeneity of the minimal-length survivors with respect to the extent of sulfation is surprising, as the “canonical” hexameric binder has seven sulfate groups; even though one of them does not participate in AT binding (the 6-O-sulfate of unit F, using the commonly accepted nomenclature²¹), its presence does not have a detrimental effect on binding.²¹ Therefore, we expected to see (6,7,1) among the minimal-length survivors alongside (6,6,1). Furthermore, longer-chain survivors also exhibit remarkable homogeneity with respect to the number of sulfate groups: for example, the octameric AT binders contain nine sulfate groups, which contradicts the earlier findings by Guerrini et al. who identified (8,10,1) and (8,11,1) as effective AT binders.²³ One might argue that the non-essential sulfate groups (that are not involved in protein binding) may be lost in the course of MS analysis due to sulfate shedding in the gas phase.³⁵ Indeed, the integrity of the non-covalent complexes in the gas phase is preserved at a price of having to select mild desolvation conditions in the ESI interface region.³⁶ Ineffective desolvation may lead to adduct formation and concomitant loss of resolution in the mass spectra; enhancing the effectiveness of ion desolvation in ESI (e.g., by applying collisional activation to ions in the ESI interface region) may not only result in dissociation of the non-covalent complexes, but in some cases lead to fission of labile covalent bonds. Glycosaminoglycans are particularly vulnerable in this respect due to lability of O-sulfate bonds;³⁵ and even partial sulfate dissociation in the gas phase would result in misidentification of AT binders.

To enhance the mass measurement resolution for shortest binding-competent heparin oligomer while ensuring the absence of sulfate loss in the gas phase, a parallel set of measurements was carried in which the SEC column eluate was mixed with denaturing solvent prior to the on-line MS detection. Post-elution denaturation results in AT/heparin oligomer complex dissociation, but the AT-binders can still be readily identified based on the anomalous elution time. SEC/MS analysis with post-column denaturation applied to the AT/dp20 mixture exposed to HepII for 40 min (Figure 2) reveals the presence of multiple oligomers eluting within the 24–31 min time window. The most abundant species are anions representing dimers (2,3,0) and tetramers (4,3,1); the most abundant hexasaccharide is (6,5,1), with the number of sulfation groups within fragments of this length ranging from four to nine (complete sulfation), and acetylation present in all moderately sulfated hexasaccharides. Despite the cornucopia of the hexasaccharide fragments present in the digest, only one of them gives rise to a peak that coincides with the protein elution (20–22 min). This hexasaccharide is identified as (6,6,1) based on the high-resolution mass measurements (Figure 2), and has the same composition as the shortest AT binder detected by native MS (vide supra), providing strong evidence that gas-phase processes do not deplete the repertoire of binding-competent heparin segment.

Figure 2.

Identification of the AT-bound hexameric heparin fragment (produced by incubating the AT/dp20 mixture with heparinase II for 40-min) using SEC/MS with post-column denaturation. The inset shows ionic signals (z = −2) for the heparin fragment co-eluting with the protein (purple) and a representative fragment with a late elution profile (green).

Another concern in designing a robust top-down protocol was the possible influence of steric hindrance on the spatial resolution. Even though the enzymatic degradation of the protein-bound heparin chain is carried out in solution without immobilizing the target protein on the solid support, the physical dimensions of the 85 kDa enzyme³¹ may prevent effective cleavage of the glycosidic bonds connecting disaccharides proximal to the AT- binding segment. This would be caused by the heparin-bound protein hindering access of such bonds to the heparinase catalytic site; the situation would be further exacerbated by the relative inflexibility of heparin (its persistence length of 20–21 ų⁷ corresponds to ca. five monosaccharide units). Identification of the hexamer as the minimal-length AT binder (in agreement with a large number of previous studies that relied on different experimental approaches¹⁸) suggests that steric hindrance introduces no bias in the top-down measurements described in this work. However, application of this approach to other heparin-binding proteins (for which there is no structural information obtained by orthogonal methods of analysis) may require an independent means of verification. In order to conclusively demonstrate the absence of the steric hindrance-related bias in the top-down experiments, a bottom-up scheme can be used for identification of the binding-competent heparin fragments. This orthogonal approach would entail limited digestion of the heparin chains without any protection afforded by the client protein, followed by incubation of all such fragments with the protein in question and identification of all binding-competent oligomers directly in solution using native MS or SEC/MS.

The results of applying this bottom-up approach to the AT/dp20 system (used as a model) are shown in Figure 3. Here a range of heparin fragments were produced enzymatically (40-min dp20 digestion with HepII) in the absence of the protein, followed by their 15-min incubation with AT. While the heparin chain fragmentation/protein binding steps followed the commonly accepted scheme, the protein binders were identified by allowing the oligoheparin fragments to interact with a free protein in solution followed by the SEC/MS analysis of the resulting protein/polysaccharide complexes. This should eliminate any biases caused by using the salt gradient elution in traditional affinity chromatography² and immobilization of the target protein (AT). Despite the abundance of di- and tetra-saccharide heparin fragments in the digest, no AT-bound dp2 or dp4 species were detected by SEC/MS: hexamers were the shortest binding-competent fragments. Importantly, out of a wide range of dp6 species produced by digestion, only (6,6,1) was found to form complexes with AT (Figure 3). Among longer fragments, both (8,9,1) and (10,12,1) were shown to bind AT; the only binding-competent fragment detected by the top-down method that was not identified using the bottom-up scheme was (12,15,1), most likely due to its insufficient amount in the dp20 digest. Therefore, no binding-competent fragments detected by the bottom-up method eluded detection with the top-down approach; in fact, the latter provided a slightly expanded repertoire of the protein binders.

Figure 3.

Identification of the binding-competent heparin fragments produced by digesting dp20 with heparinase for 40-min. A: representative mass spectra averaged over several adjacent early elution windows in SEC/MS of AT mixed with the products of limited digestion of dp20 (the UV chromatogram is shown in the inset). Mass shifts due to binding of (6,6,1), (8,9,1) and (10,12,1) are indicated with purple, brown and orange arrows, respectively. B: mass spectrum of the products of the limited digestion of dp20 with heparinase II prior to their mixing with AT.

The remarkable agreement between the results of the studies carried out with two different approaches provides a clear indication that no bias is introduced by the steric hindrance effect vis-à-vis identification of the minimal-length protein binders. In order to provide further insight into this phenomenon, the top-down approach was also applied to the AT/pentasaccharide (GlcNS,6S-GlcA-GlcNS,3S,6S-IdoA2S-GlcNS,6S-Me, see Scheme 1) complex. All saccharide units comprising the latter are involved in AT binding,³⁸ and the K_d of the complex is estimated to be in the tens of nM.^39,40 A two-hour incubation of the AT/pentasaccharide complex in the presence of HepII, with 1 unit enzyme to 10 μg pentasaccharide, followed by the SEC/MS analysis of the products showed the presence of the pentasaccharide-free protein (which remained undetectable prior to the complex incubation with the enzyme) alongside the surviving AT/pentasaccharide complex (Supplementary Material). Analysis of the low-molecular weight components of the AT/pentasaccharide mixture following its exposure to HepII revealed the presence of complementary tri- and disaccharide fragments in solution; however, their association with AT was not detected. The ability of the enzyme to degrade the AT-bound pentasaccharide indicates that the enzyme can act on the substrate during its transient dissociation from the protein. Likewise, in the case of longer heparin oligomers, the protection that might be afforded by the steric hindrance to the saccharide units that are proximal to the binding site would be lifted during the transient dissociation events. The apparent susceptibility of the AT-bound pentasaccharide to lysis by heparinase raises an important question of how tight the binding should be in order to be detectable by the methodology presented in this work. The K_D value of the AT/pentasaccharide complex is ca. 30 nM,⁴⁰ but the k_OFF value for this complex is likely to be a more relevant parameter in this respect. We estimate k_OFF of the AT/pentasaccharide to be below 0.3 sec⁻¹ (based on its K_D value, and the lower limit of the k_ON value of 10⁷ M⁻¹s⁻¹, which had been reported for a structurally similar anticoagulant⁴¹). However, the half-life of the complex (> 3 sec according to these rough estimates) is not the only parameter determining the applicability of the top-down method for identification of heparin binding elements for a specific protein (not every dissociation event results in digestion of the “liberated” heparin oligomer; the actual rate of digestion in this case depends on the competition between the pentasaccharide re-association with AT and formation of the Michaelis complex with the enzyme). Reliable estimates are not feasible at this time due to unavailability of the key kinetic parameters, but it seems that variation of the digestion time and/or heparinase concentration should allow this technique to be applied even to systems where the protein/heparin association is considerably weaker than the AT/heparin binding.

Discussion

Identification of specific “protein recognition elements” within heparin chains remains elusive for the majority of heparinome.¹⁵ Furthermore, an argument is often made in favor of considering less specific interactions (such as long-range electrostatics) as the primary forces guiding recognition and binding.^26,27 In fact, AT remains one of the very few proteins for which this debate was settled following the discovery of a unique protein-binding element, the pentasaccharide.⁴² However, even in the case of AT/heparin binding certain observations appear to be in conflict with the prevailing dogma. For example, the pentasaccharide sequence is relatively rare and is found in only about one third of the chains in commercial heparin preparations;⁴³ however, a single heparin chain can accommodate multiple AT molecules.⁴⁴ Longer heparin chains also have higher AT affinity and are more efficient AT activators even when no bridging within the protease/serpin complex is required.⁴⁵ Lastly, short heparin oligomers whose structure is inconsistent with the pentasaccharide sequence were shown to be capable of binding to AT,⁴⁶ and in fact even the GAG backbone does not appear to be required for the modulation of hemostasis driven by AT activation.^47,48

The new method for identifying protein-binding elements within heparin chains presented in this work provides a unique opportunity to address these controversies by allowing the protein-binding preferences to be explored without compromising the enormous structural diversity of heparin. A minimal binding segment is readily identified as a mono-acetylated hexa-saccharide with a relatively modest level of sulfation, and its composition (6,6,1) is consistent with the known structure of the high-AT affinity fragments obtained by either enzymatic or chemical degradation of heparin followed by fractionation on immobilized AT.²¹ Even though the latter has seven sulfate groups (Scheme 1), one of these sulfates (6S at 6S-IdoA2S) is not essential for AT binding.^21,42 It is remarkable that both top-down and bottom-up analyses of AT-heparin binding clearly identify (6,6,1) as the only minimal-length binder, despite the abundance of other hexamers, including the highly sulfated ones (Figures 2 and 3). This surprising finding may have important ramifications vis-à-vis the “hidden specificity” concept,^15,17 which is based on the assumption that the high sulfate density within heparin (compared to heparan sulfate) endows it with the ability to interact with a wide range of heparan sulfate’s client proteins, without any negative affect from the surplus of non-essential sulfate groups. In fact, the surprising homogeneity of the minimal binding segments with respect to the overall sulfation and acetylation levels appears to reinforce the notion of the structural complementarity being the major determinant of the AT/heparin binding. At the same time, we note that all “survivors” larger than a hexamer relate to the minimal-length binder by addition of fully sulfated, acetyl-free disaccharides, i.e. their common formula can be written as (6+2n,6+3n,1). An argument can be made that this maximized charge density within the extension segments of the AT-binding core simply reflects the high extent of sulfation typical of heparin (the manufacturer’s specifications indicate that tri-sulfated disaccharides (2,3,0) constitute ca. 70% of all dp20-derived disaccharide fragments, a notion confirmed by SEC/MS profiling of the products of extensive lysis of dp20 – see Supplementary Material). However, the 70% abundance of the (2,3,0) disaccharide units within the dp20 chains means that random selection of a tetrasaccharide fragment from this chain would result in a fully sulfated tetramer (4,6,0) having ca. 50% (0.7²) probability of occurrence, while the probability of randomly selecting three fully sulfated disaccharide building blocks to construct a (6,9,0) hexamer would drop below 35% (0.7³). However, only the fully sulfated extensions of the binding core are observed experimentally within the surviving heparin decamer and dodecamer fragments, indicating an important role played by electrostatic interactions in facilitating protein/heparin association outside of the canonical binding site.

Although the mass measurements alone are not sufficient to conclude whether the fully sulfated disaccharides extend the core hexasaccharide at the reducing end, non-reducing end, or both, we note that the surface charge distribution near the pentasaccharide-binding interface region displays a continuous positive patch extending beyond the groove-like region accommodating the non-reducing end of the pentasaccharide (Figure 4). The major contributors to this “extension” of the positive patch on the surface of the overall negatively charged protein are Arg¹³², Lys¹³³ and Lys¹³⁶. The size of this extension appears to be sufficient to accommodate several additional disaccharide units at the non-reducing end of the minimal-length binder, (6,6,1). Not surprisingly, MD simulations of AT/(8,9,1) complex demonstrate that addition of a fully sulfated disaccharide at the reducing end of the minimal-length binder is energetically inferior to extending the (6,6,1) segment with a fully sulfated disaccharide at the non-reducing end (Figure 4, see also Supplementary Material). Although no MD simulations were carried out with heparin oligomers exceeding octasaccharides, dimensions of the entire positively charged basin on the surface of AT appears to be sufficient to accommodate chains as long as dodecasaccharides, consistent with the experimental observation that the longest distinct heparin fragment that remains bound to AT following a 40 min-long digestion is (12,15,1) (Figure 1B).

Figure 4.

MD simulations of AT association with: (A) octamer constructed by extending the core hexamer (6,6,1) with a fully sulfated dimer (2,3,0) at its non-reducing end, (B) octamer constructed by extending the core hexamer (6,6,1) with a fully sulfated dimer (2,3,0) at its reducing end.

Extension of the hexasaccharide core by addition of fully sulfated disaccharides highlights the importance of electrostatics in enabling heparin association with AT beyond its canonical pentasaccharide binding interface. The long-range nature of electrostatic interactions may lead to the protein/heparin association scenarios where the extension regions of the heparin’s AT-binding core segment are not necessarily locked in a tight binding to the protein surface (making them more labile vis-à-vis HepII digestion compared to the core hexasaccharide). Nonetheless, the additional electrostatic interactions are likely to strengthen the binding by decreasing k_OFF value. It also seems likely that the long-range electrostatics is important at the initial stages of AT/heparin binding, where the highly sulfated extension segments would help steer the core hexasaccharide towards the positive patch on the surface of the overall negatively charged protein.

Above and beyond the auxiliary role played by electrostatics in heparin/AT association, these relatively non-specific forces become the primary driver of the association process if there is a deficiency of the pentasaccharide binding elements within the heparin sample. For example, the commercial dp6 sample appears to be enriched in high charge-density species, with (6,6,1) being a minor component. This bias results in the preferential binding of the highly sulfated hexasaccharides to the protein (see Supporting Information), a phenomenon that had been observed previously for the heparin hexamers lacking moderately-sulfated species⁴⁶. Both experimental data and the results of MD simulations provide clear evidence that electrostatic interactions are capable of driving the binding process by offsetting the enthalpic losses caused by the absence of the key recognition elements (such as the N-acetyl group at the reducing-end disaccharide unit of the core hexasaccharide). In fact, MD simulations indicate that replacing the canonical hexasaccharide (6,6,1) with the highly sulfated hexasaccharide (6,9,0) lacking the essential N-acetyl and 3-O-sulfate groups decreases the AT/heparin oligomer binding energy only 10 kcal/mol (with the loss of van-der-Waals contributions being partially offset by gains in electrostatic interactions for highly sulfated species, see Supporting Information). Clearly, the fully sulfated hexasaccharide is a viable (but not the optimal) binder, consistent with the association of the highly sulfated species with AT in the absence of the “optimal binder” observed in this work and elsewhere.⁴⁶ Contributions of the long-range electrostatic interactions beyond the hexasaccharide core are also important, as they provide additional gains in binding energy without the need to lock the rest of the heparin oligomer in a conformation that would be dictated by the short-range interactions. Indeed, locking an extended segment of the heparin chain in a specific conformation would be entropically unfavorable due to its relative inflexibility.⁴⁹ Thus, engagement of the long-range electrostatic interactions affords stronger protein/heparin binding by providing notable enthalpic gains while avoiding/minimizing the entropic penalties, consistent with the results of MD simulations (see Supporting Information).

Conclusions

Despite being the first polysaccharide introduced into modern medical practice, heparin remains underutilized in the clinic beyond the common anti-thrombotic/hemostatic applications. Successful exploitation of heparin’s unique ability to interact with a wide range of proteins hinges upon the ability to identify its specific structural elements that enable such interactions. However, the enormous structural heterogeneity (the very same feature that allows heparin to interact with a large number of partners in vivo⁵⁰) makes the search for the specific protein “recognition elements” an extremely challenging undertaking. The new method introduced in this work allows such structural elements to be identified within long heparin chains without compromising their structural diversity. Application of this new method to the paradigmatic AT/heparin system correctly identifies a hexasaccharide with a single N-acetyl group and six sulfates as a unique minimal-length binder; analysis of longer-chain binders yields evidence for the relatively non-specific electrostatic interactions acting outside of the hexasaccharide core to provide additional stabilization of the protein/heparin complex. This nuanced view highlights the importance of both specific and non-specific interactions in protein/heparin (and, more broadly, protein/glycosaminoglycan) association, reconciling the two opposing views that emphasize either the dominance of structural complementarity similar to that encountered in protein/protein interactions⁷ or the less specific (but also longer-range) electrostatic forces.^26,27,51 Application of the new method to heparin interactions with proteins outside of the blood coagulation cascade will provide structural information that will undoubtedly catalyze progress in diverse fields ranging from rational design of heparin-based scaffolds for delivery of therapeutic proteins to fabricating inhibitors for pathogenic heparin-protein interactions (such as those encountered in heparin-induced thrombocytopenia). The new method presented in this work is superior to the recently introduced gas-phase foot-printing methodology,⁵² as it is free of the artefacts related to the complex activation in the gas phase (such as asymmetric charge partitioning and dissociation of labile carbohydrate chains of glycoproteins). It should also be applicable to systems beyond 1:1 protein/heparin complexes, e.g. glycosaminoglycan-assisted assembly of protein/receptor complexes.⁵³ The information content provided by the new top-down method of heparin/protein interaction analysis would be expanded even further via incorporation of gas-phase heparin oligomer sequencing (e.g., negative ion/electron transfer dissociation⁵⁴) into the experimental scheme to enable precise localization of all sulfate and acetyl groups without raising the specter of sulfate shedding during ion activation.⁵⁵