Binding of heparin-dependent antibodies to PF4 modified by enoxaparin oligosaccharides: evaluation by surface plasmon resonance and serotonin release assay

Authors


Yves Gruel, Department of Hematology-Hemostasis, University Hospital of Tours, 37044 Tours, France.
Tel.: +33 2 47 47 46 72; fax: +33 2 47 47 59 04.
E-mail: gruel@med.univ-tours.fr

Abstract

Summary.  Background: The minimal structural requirements of low-molecular-weight heparins that determine the risk of developing heparin-induced thrombocytopenia (HIT) are not fully defined.Objectives: The ability of enoxaparin-derived oligosaccharides (OS) to induce platelet activation and exposure of platelet-factor 4 (PF4) epitopes recognized by antibodies developed in HIT was studied by surface plasmon resonance (SPR) and serotonin release assay.Results: Decasaccharides with ≥ 11 sulfate groups induced platelet activation in the presence of plasma from patients with confirmed HIT. Serotonin release of > 80% without full inhibition at 100 μg mL−1 was achieved with decasaccharides containing 14 or 15 sulfate groups, 2 dodecasaccharides and 2 tetradecasaccharides. An SPR method was developed using purified PF4 immobilized on carboxymethylated dextran. Antibodies from all HIT samples bound to PF4/heparin in SPR assays with resonance units (RU) ratio of 109–173 with HIT plasma vs. 88–93 with control plasma. RU ratios > 100 were measured when PF4 was pre-incubated with OS with ≥ 10 saccharide units and one octasaccharide containing 10 sulfate groups. RU ratios > 140, similar to those measured when PF4 was pre-incubated with unfractionated heparin or enoxaparin, were obtained with purified dodeca- and tetradecasaccharides. RU values strongly correlated with the number of sulfate groups in the decasaccharides tested (r = 0.93, = 0.02).Conclusions: LMWHs with fragments > 10 saccharides and a large number of sulfate groups are more likely to be associated with a higher risk of HIT. These structure-activity relationships were independent of the ability of the OS to bind antithrombin.

Introduction

Understanding the mechanisms and consequences of interactions between proteins and heparin and its derivatives, such as low-molecular-weight heparins (LMWHs) and fondaparinux, is a major issue in the area of glycosaminoglycan (GAG) research [1]. Enoxaparin is the most frequently prescribed LMWH worldwide, and is obtained by alkaline hydrolysis of benzyl esters of unfractionated heparin (UFH). Like other LMWHs, enoxaparin is a complex mixture of polysaccharide fragments, of which about 20% contain the so-called antithrombin (AT)-binding sequence with high affinity for AT [2]. Apart from AT, numerous other heparin-binding proteins, including protease inhibitors, growth factors, lipid-binding proteins, adhesion proteins and chemokines, have been identified [3]. Among chemokines known to bind heparin, platelet-factor 4 (PF4) is the most clinically relevant because interactions between PF4 and heparin fragments are crucial for priming the synthesis of the platelet-activating antibodies responsible for heparin-induced thrombocytopenia (HIT). HIT is a severe and relatively frequent complication of treatments with heparin [4], which results from an atypical immune response involving the synthesis of immunoglobulin G antibodies that bind to PF4 epitopes exposed by conformational changes induced by interactions with sulfated polysaccharides [5,6].

Recently, the US Food and Drug Administration (FDA) approved the first generic LMWH [7]. However, some differences may exist between generics and biosimilars, and the original product in terms of molecular weight range, stability and immunogenicity [8]. HIT is a non-hemorrhagic adverse effect also associated with LMWHs, although the incidence is lower than with UFH. Therefore, in addition to defining five criteria to demonstrate the sameness of generics and the branded LMWH enoxaparin, the FDA also outlined the usefulness of evaluating the immunogenicity that may potentially lead to the synthesis of anti-PF4 antibodies and to HIT [9].

In this context, we designed a study combining surface plasmon resonance (SPR) analysis with the serotonin release assay (SRA) to define the minimum structural requirements for enoxaparin-derived oligosaccharides (OS) to induce the exposure of epitopes recognized by the antibodies to PF4 associated with HIT.

Methods

Plasma samples from HIT patients and healthy controls

Plasma samples were obtained from 10 patients with confirmed HIT. All these samples contained significant levels of antibodies against PF4/polyvinyl sulfonate complexes (median A405 nm 2.945, range 0.898–3.690; cut-off value 0.4; PF4 Enhanced®; GTI Diagnostics, Brookfield, WI, USA) and had been tested positive in an SRA [10]. HIT plasma from three patients, for whom enough material was available, was also used for platelet activation tests by SRA and for studying the binding of HIT antibodies to PF4 by SPR.

Normal plasma samples were collected from 10 healthy subjects who were not receiving heparin. Each normal plasma control was tested by a PF4/polyvinyl sulfonate ELISA, and no significant levels of heparin-dependent antibodies were measured (median A405 nm 0.118, range 0.041–0.316).

Blood samples were collected in 0.109 m sodium citrate (9:1) as an anticoagulant, centrifuged twice (2500 ×g, 15 min), and platelet-free plasma stored at −80 °C until assayed.

Purification of OS

Twenty OS, either affine (= 14) or non-affine for AT (= 6), and containing the 1,6 anhydro moiety (= 5) or not (= 15), were purified by combining AT-affinity chromatography, strong anion-exchange on Carbopack AS11 (Dionex, Sunnyvale, CA, USA) and cetyl trimethyl ammonium strong anion-exchange chromatography on a semi-preparative scale, starting from gel permeation chromatography (GPC) fractions of enoxaparin. GPC and desalting of fractions were performed as described previously [11] (Supporting information).

The chemical structure of each OS was checked by nuclear magnetic resonance imaging and high-performance liquid chromatography techniques [12]. Structural characterization of OS was based on key features, including OS size, AT binding ability, and the presence of the 1,6 anhydro moiety.

Serotonin release assay

The ability of HIT antibodies to induce platelet activation in the presence of enoxaparin-derived OS was studied by SRA, as previously described [10]. All tests were performed using plasma from three patients with confirmed HIT and with washed platelets from three subjects previously identified as good responders to HIT antibodies.

For each series of assays, 14C-serotonin labeled platelets were incubated with the HIT plasma samples and increasing concentrations of UFH, enoxaparin, fondaparinux or purified OS (0, 0.1, 1, 10 and 100 μg mL−1). To determine if platelet activation induced in the presence of HIT antibodies was dependent on FcγRIIa receptors, 10 μg mL−1 of the monoclonal antibody IV.3 (Stem Cell, Grenoble, France) was pre-incubated with washed platelets for 10 min at 37 °C as a control assay.

SPR analysis

Real-time binding SPR experiments were performed using a Biacore® T100 biosensor instrument (GE Healthcare, Velizy, France). SPR enables the analysis of molecular interactions between one reactant bound on a surface and a second reactant in solution. During the assay, changes in refractive index occurring during complex formation or dissociation are measured and the resulting SPR signal is expressed in resonance units (RU). PF4 (Hyphen Biomed, Neuville-sur-Oise, France) was immobilized on a CM5 sensor chip coated with a carboxymethylated dextran matrix (GE Healthcare). Analyses were conducted at 25 °C using a standard HEPES-buffered saline buffer (HBS; 10 mm HEPES pH 7.4, 150 mm NaCl) with 0.05% (v/v) Tween 20 (Sigma-Aldrich, Saint-Quentin Fallavier, France).

A three-step SPR procedure, based on the use of two flow cells (FCs), one test FC and one control FC, was defined as follows: the CM5 sensor chip was stabilized, then PF4 protein (6 μg mL−1) was injected at a rate of 10 μL min−1 for 120 s in the test FC. A solution of the test GAG (i.e. UFH [0.08 IU mL−1], enoxaparin [0.08 IU mL−1], fondaparinux [0.8 μg mL−1], or enoxaparin-derived OS [0.5 μg mL−1]), was then injected into each FC. Finally, HIT or normal plasma diluted in HBS (1:20) was injected into each FC. After dissociation (120 s), the sensor chip surface was regenerated with a solution of 50 mm NaOH injected for 20 s at 30 μL min−1. The binding of rabbit polyclonal anti-PF4 antibodies (Hyphen-BioMed) injected at 10 μg mL−1 was also evaluated to check that PF4 remained bound to the sensor chip after injection of each heparinoid into the FC.

Specific interactions on the surface of the sensor chip during each step of the procedure were assessed by continuous monitoring of sensorgrams with RU reflecting the extent of binding and dissociation of every molecule onto the CM5 sensor chip. The binding of HIT antibodies to heparinoid-modified PF4 was evaluated using the last part of the sensorgram in each assay. The RU values recorded from the control FC were subtracted from the test FC RU values to control for non-specific interactions. Results are expressed as an RU ratio calculated by:

image

Statistical analysis

SPR results were analyzed using the supplied computer software (BIAevaluation™, version 3.1; GE Healthcare). Correlation analyses were performed using the Spearman test with StatView™ software (SAS Institute Inc. Cary, NC, USA).

Results

Characteristics of enoxaparin OS

Twenty purified OS that contained 8–14 monosaccharide residues were isolated. Fourteen of these OS contained an AT-binding pentasaccharide with different structural characteristics (Table 1 and Supporting Information). The majority of OS purified from enoxaparin were decasaccharides, and contained between 10 and 15 sulfate groups. Two decasaccharides (OS 10/14 and OS 10/15) did not bind to AT. Moreover, two AT-binding decasaccharides (OS 10/12A and OS 10/12B), as well as one non-AT-binding decasaccharide (OS 10/14), displayed the 1,6 anhydro moiety, which is characteristic of enoxaparin. The smallest OS obtained were three AT-binding hexasaccharides, which contained seven or eight sulfate groups. One (OS 6/7B) exhibited a lower affinity for AT compared with the two others due to lack of 6-O sulfation of the reducing end glucosamine. Five different AT-affine octasaccharides were purified, containing 8 (= 1), 9 (= 1) or 10 (= 3) sulfate groups. The largest OS isolated were two dodecasaccharides and two tetradecasaccharides, with and without the 1,6 anhydro moiety, none of which bound to AT.

Table 1.  Primary characteristics of the 20 enoxaparin OS studied. All were tested in SRA and 15 were evaluated in SPR experiments performed with three HIT plasma samples and one healthy control plasma sample
GAGSaccharide units, nSulfate groups, nAT affinityRU ratio
HIT plasma samplesNormal plasma
HIT #1HIT #2HIT #3
  1. AT, antithrombin; GAG, glycosaminoglycan; HIT, heparin-induced thrombocytopenia; ND, not determined; OS, oligosaccharide; RU, resonance units; SPR, surface plasmon resonance; SRA, serotonin release assay; UFH, unfractionated heparin. Values >100 (cut-off) are in bold.

OS 6/7A67+94718087
OS 6/7B67+ (low)NDNDNDND
OS 6/868+NDNDNDND
OS 8/888+NDNDNDND
OS 8/989+98748489
OS 8/10A810+ 106 738991
OS 8/10B810+NDNDNDND
OS 8/10C810+NDNDNDND
OS 10/101010+ 112 789891
OS 10/111011+ 110 769591
OS 10/12A1012+ (low) 118 80 104 92
OS 10/12B1012+ 123 80 110 93
OS 10/13A1013+ 129 85 114 92
OS 10/13B1013+ 122 82 110 93
OS 10/141014 139 108 133 107
OS 10/151015 138 113 138 102
OS 12/171217 145 121 146 90
OS 12/181218 141 121 149 89
OS 14/201420 140 123 153 88
OS 14/211421 138 122 153 88
UFH  + 143 131 171 87
Enoxaparin  + 141 NDND92
Fondaparinux58+85NDND89

Platelet activation induced by HIT antibodies and enoxaparin OS

HIT plasma samples induced a strong 14C-serotonin release in the presence of UFH or enoxaparin as expected, with complete inhibition of platelet activation achieved using a high concentration of heparinoid (Fig. 1). In contrast, SRA was consistently negative with fondaparinux at all concentrations tested.

Figure 1.

 Platelet activation evaluated by serotonin release assay and induced by heparin-induced thrombocytopenia plasma samples (= 3) in the presence of increasing concentrations of unfractionated heparin (UFH), enoxaparin, fondaparinux or oligosaccharides (OS).

No significant 14C-serotonin release was measured when OS composed of ≤ 8 saccharide units, including fondaparinux, were incubated with HIT plasma, regardless of sulfation pattern (Fig. 1). Significant platelet activation was induced in the presence of 11 of the 12 OS that contained ≥ 10 saccharide units. 14C-serotonin release > 40% was measured with 10 μg mL−1 of OS 10/11, whereas it remained < 20% with OS 10/10. The most potent platelet activation (close to 80% release) was induced with a low concentration (1.0 μg mL−1) of the decasaccharides containing 14 or 15 sulfate groups, the two dodecasaccharides and the two tetradecasacharrides, without complete inhibition at 100 μg mL−1.

Incubation with IV.3, which specifically blocks FcγRIIa receptors, fully abolished the serotonin platelet release induced by HIT plasma in the presence of OS (data not shown). No OS induced any significant release when labeled platelets were incubated with normal control plasma.

Evaluation by SPR of the binding of HIT antibodies to PF4 modified by UFH

To validate the feasibility of studying the interactions of HIT antibodies with H/PF4 by SPR, we first immobilized purified PF4 onto the polyanionic CM5 sensor chip by adsorption. After injection of PF4 into the test FC, the median RU value obtained after stabilization was relatively high, with variations between experiments that did not exceed 10–15%. When UFH was then injected into the FC, the resonance signal sharply decreased because of a partial desorption of PF4 from the sensor chip, although enough protein remained bound to the matrix, as demonstrated by separate experiments evaluating the binding of rabbit polyclonal antibodies to PF4. The sensorgram was normalized before injection of plasma to start from a baseline RU value of 0. Representative sensorgrams obtained after injection of HIT and control plasma samples into test FC previously exposed to UFH are depicted in Figure 2. SPR binding assays demonstrated significant binding of HIT antibodies present in all HIT plasma samples to immobilized H/PF4 (Fig. 2A). This binding resulted in an increase in RU values (median RU 254, range 101–717) when HIT samples were injected after heparin in the test FC. In contrast, control plasma samples showed a decrease of the resonance signal (median RU −103, range −73 to −111; Fig. 2B), suggesting minor removal of H/PF4 complexes from the sensor chip.

Figure 2.

 (A) Analysis by surface plasmon resonance (SPR) of the binding of heparin-induced thrombocytopenia (HIT) antibodies (= 10) to immobilized purified platelet-factor 4 (PF4) after previous interaction with unfractionated heparin (UFH). (B) SPR sensorgrams obtained with normal plasma samples (= 10) and purified PF4 modified by UFH. (C) Resonance units (RU) ratio values measured in SPR assay after interaction of HIT (○, •) or normal (□, bsl00001) plasma samples with immobilized PF4 before (white symbols) and after (black symbols) injection of UFH.

RU ratios measured after injection of HIT plasma samples in the test FC with H/PF4 were always > 100 (median RU ratio 128, range 109–173; Fig. 2C). The binding of HIT antibodies to PF4 alone (i.e. without any prior interaction with UFH) was also evaluated, and yielded RU ratios < 100 (median 77, range 60–122) in 8 of 10 cases. With the plasma from two HIT patients, the RU ratio measured with PF4 unexposed to heparin was slightly > 100 (113 and 122), and increased to 173 and 170, respectively, with PF4 modified by UFH. In contrast, RU ratios obtained with control plasma samples always remained lower than 100 (median 90, range 88–93), whether immobilized PF4 had previously interacted with heparin or not.

SPR analysis of HIT antibody binding to PF4 modified by enoxaparin OS

The potency of enoxaparin-derived OS in promoting the binding of HIT antibodies to immobilized PF4 was then evaluated by testing all OS previously identified by SRA to induce significant platelet activation with HIT plasma (= 11). In addition, SPR assays were performed with four other OS for which serotonin release was nil (OS 6/7, 8/9 and 8/10A) or negligible (OS 10/10). Each OS was injected at 0.5 μg mL−1 with subsequent reduction in RU due to a partial desorption of PF4. The binding of polyclonal rabbit antibodies to PF4 was therefore measured in SPR to check that sufficient PF4 remained bound to the dextran surface, and RU ratio values obtained were always > 130 (median 156, range 130–178), regardless of the OS previously injected in the FC.

One HIT plasma (#1) was tested by SPR after immobilized PF4 had interacted with 15 enoxaparin-derived OS, and RU ratio values > 100 were measured for 13 of these 15 OS, including one octasaccharide (OS 8/10A) and all OS with ≥ 10 saccharide units (Table 1, Fig. 3). No significant increase in RU was measured when PF4 had been pre-exposed to fondaparinux or the smallest OS tested (OS 6/7A and OS 8/9). With this plasma, the highest RU ratio values (> 140 and similar to those recorded with UFH and enoxaparin) were measured when PF4 had been pre-exposed with the dodeca- and tetradecasaccharides. The binding of HIT antibodies to PF4 after pre-incubation with each of the eight decasaccharides tested progressively increased according to the number of sulfate groups present, which varied from 10 to 15 (Table 1, Fig. 3), exhibiting a positive correlation with RU values (r = 0.93, = 0.02).

Figure 3.

 Resonance units (RU) ratio values measured in surface plasmon resonance after injection of heparin-induced thrombocytopenia plasma #1 (•, = 3–4 experiments) on immobilized platelet-factor 4 previously modified by unfractionated heparin, enoxaparin, fondaparinux or 15 different oligosaccharides (OS). Values obtained with normal plasma are also shown (○).

SPR analyses were also performed with two other HIT plasma samples, and RU values > 100 were recorded when PF4 had been exposed to OS containing ≥ 10 saccharide units and 12 or 14 sulfate groups (Table 1). As expected, no significant increase in RU was recorded when normal plasma was tested with 13 of the 15 OS, UFH or enoxaparin (Fig. 3). A slight increase in RU ratio > 100 was recorded when PF4 had interacted with the two decasaccharides having the highest number of sulfate groups (OS 10/14 and 10/15).

Discussion

In this study we devised a method based on SPR technology, a powerful tool for studying antigen-antibody interactions without labeling or washing procedures [13], to evaluate the effects of chemically well-characterized enoxaparin OS on the binding of HIT antibodies to immobilized PF4.

We used a sensor chip coated with a dextran matrix that allowed the direct binding of purified PF4 without chemical modification of the protein, which could alter its structure and antigenicity. As expected, the injection of therapeutic heparinoids (UFH, enoxaparin or fondaparinux), or purified OS in the FC was associated with partial removal of PF4 from the sensor chip. However, given that significant resonance signals were always measured after injection of either a rabbit polyclonal antibody to PF4 or the plasma of patients with confirmed HIT, enough protein remained bound to dextran to enable evaluation of the binding of anti-PF4 antibodies.

Data obtained with HIT plasma samples exhibited a significant increase in resonance signal with all samples tested, indicating that our SPR technique could be sufficiently sensitive to detect and study the binding of HIT antibodies to immobilized PF4 previously incubated with complex mixtures of GAGs, such as UFH or LMWH, as well as by pure OS such as those isolated from enoxaparin. This binding of HIT antibodies to modified PF4 as measured by SPR also appeared to be specific as no increase in resonance signal was detected with any of the control plasma samples tested.

Heparins behave differently in their ability to induce HIT, and LMWHs (approximately 15 saccharide units) are less likely than UFH (approximately 45 saccharide units) to trigger the synthesis of anti-PF4 antibodies and to cause clinical HIT [14]. These differences in the immunogenicity of heparins are related to polysaccharide chain length because heparin molecules wrap around positively charged PF4 tetramers to form the immunogenic complexes that cause HIT [15]. This interaction alters the three-dimensional structure of PF4 by exposing the neoepitopes that elicit the formation of HIT antibodies [5,6], and was already considered several years ago to mainly involve heparin fragments longer than octasaccharides [16].

This study is the first to use well-characterized OS purified from enoxaparin, an LMWH used worldwide. These provided a unique opportunity to directly investigate the impact of an individual structure inducing conformational changes in PF4 and subsequent HIT antibody binding. The key features of the OS are: presence or absence of affinity to AT; the distinct position of sulfate groups in the OS; and the presence or absence of the bicyclic 1,6 anhydro ring at the OS non-reducing end. The SPR results confirm that exposure of PF4 epitopes able to bind HIT antibodies depends on previous conformational changes that are only induced by OS of sufficient length and containing ≥ 10 sulfate groups.

No binding of HIT antibodies on PF4 could be detected in SPR after injection of fondaparinux or OS with 6 or 8 saccharide units. Additionally, SRA indicated that platelet-activating immune complexes were formed in the presence of HIT antibodies in sufficient amounts with decasaccharides that contain > 10 sulfate groups, such as OS 10/14 and OS 10/15, and these OS did not inhibit release. Analysis of the sensorgrams obtained in SPR with these OS also indicated that dissociation of HIT antibodies from immobilized PF4 was less rapid (data not shown), and that PF4/OS complexes were more stable. Therefore, we would predict that heparins with ≥ 10 saccharide units and > 13 sulfate groups will be more efficient at forming the macromolecular complexes responsible for triggering the synthesis of antibodies to PF4 and the development of HIT.

Our structure-activity relationship analysis did not reveal an impact of specific sulfate positions on the ability of OS to form with PF4 antigen complexes recognized by HIT antibodies. Therefore, the strength of PF4/OS interactions does not seem dependent on specific patterns of sulfation, but rather on the charge density and OS size. These findings are thus in contrast to those obtained in previous studies focusing on specific interactions of OS with AT, which demonstrated that the precise OS structure exerted a critical impact on binding strength [12,17].

UFH is a heterogeneous mixture of sulfated OS of variable length, and HIT antibody-related platelet activation varied according to the molecular weight of heparin, as shown in 14C-serotonin release assays. Greinacher et al. [18] defined the optimal chain length for making HIT antigen as a hexadecasaccharide after evaluating five different heparin fractions with SRA, and reported that higher concentrations of heparin were necessary to induce platelet activation when smaller fragments were tested. This observation was then confirmed by Walenga et al. [19], who also showed that heparin fractions in the molecular weight range of 3500–4500 Da induced platelet activation at 1 μg mL−1, which was not inhibited at concentrations as high as 100 μg mL−1. Our data obtained after testing purified OS by both SRA and SPR are in full agreement with these results, confirming that the molecular weight of OS is a crucial factor in the antigenicity of H/PF4.

A strong correlation was also found between the number of sulfate groups within heparin chains and the RU ratio, which reflects the extent of HIT antibody binding on PF4 immobilized on the sensor chip. This apparent key role of heparin charge is consistent with a previous study in which desulfated heparin fractions were unable to induce HIT antibody-mediated platelet aggregation compared with the parent UFH [20].

In conclusion, this study further sheds light on the specific structural requirements underlying interactions between heparin-derived OS and PF4, and supports the hypothesis that LMWHs with fragments > 10 saccharides and a large number of sulfate groups are associated with higher risk of undesirable immune response with the generation of antibodies to PF4. These antibodies are not necessarily associated with a significant clinical risk of HIT, as supported by recent studies on fondaparinux [21]. However, HIT remains a possible severe adverse risk to be considered for all new heparin-derived antithrombotic drugs, and SPR analysis combined with the SRA may be a useful method for the preclinical evaluation of the risk of HIT associated with generic LMWHs.

Acknowledgements

We thank R. Spice, Excerpta Medica, for editorial assistance in the preparation of this article, funded by Sanofi.

Disclosure of Conflict of Interests

This study was supported by the Institut de Recherche sur la Thrombose et l’Hémostase (IRTH), and by a subvention from Sanofi. C. Viskov, F. Herman and P. Mourier are employees of Sanofi.

Ancillary