Báo cáo Y học: Interaction of bovine coagulation factor X and its glutamic-acid-containing fragments with phospholipid membranes A surface plasmon resonance study

Eur. J. Biochem. 269, 3041–3046 (2002) Ó FEBS2002 doi:10.1046/j.1432-1033.2002.02981.x Interaction of bovine coagulation factor X and its glutamic-acid-containing fragments with phospholipid membranes A surface plasmon resonance study Eva-Maria Erb1, Johan Stenflo1 and Torbjo¨rn Drakenberg2 1Department of Clinical Chemistry, University Hospital Malmo, Lund University, Malmo, Sweden; 2Department of Biophysical Chemistry, Lund University, Lund, Sweden The interaction of blood coagulation factor X and its Gla-containing fragments with negatively charged phos-pholipid membranes composed of 25 mol% phosphatidyl-serine(PtdSer)and75 mol%phosphatidylcholine(PtdCho) was studied by surface plasmon resonance. The binding to 100 mol% PtdCho membranes was negligible. The calcium dependence in the membrane binding was evaluated for intact bovine factor X (factor X) and the fragment con-tainingtheGla-domainandtheN-terminalEGF(epidermal growth factor)-like domain, Gla–EGFN, from factor X. Both proteins show the same calcium dependence in the membranebinding.Calciumbindingiscooperativeandhalf-maximum binding was observed at 1.5 mM and 1.4 mM, Bloodcoagulationfactor Xbelongstothefamilyofvitamin K-dependent proteins. It consists of an NH2-terminal c-carboxyglutamic acid (Gla)-containing domain, followed by two epidermal growth factor (EGF)-like domains and a serine protease (SP) domain [1]. The Gla-domain mediates Ca2+-dependent binding to biological membranes, for example the platelet membrane [2]. Binding of factor X and other Gla domain-containing coagulation factors is greatly enhanced after platelet activation, due to the exposure of negatively charged phosphatidylserine (PtdSer) on the cell surface. The crystal structure of the Ca2+-loaded formofprothrombinfragment1showedthatsixorsevenof theGlaresiduesligatefourtofiveCa2+ intheinteriorofthe protein and that three conserved residues with hydrophobic side-chains,Phe4,Leu5andVal8inbovinefactor X,forma hydrophobic patch on the surfase of the domain [3–5]. These residues are thought to mediate membrane-binding Correspondence to T. Drakenberg, Department of Biophysical Chemistry,Lund University, P.O. Box124,SE-221 00 Lund,Sweden. Fax: + 46 46 222 45 43, Tel.: + 46 46 222 44 70, E-mail: Torbjorn.Drakenberg@bpc.lu.se Abbreviations: PtdSer, phosphatidylserine; PtdCho, phosphatidtylcholine; Gla, c-carboxy glutamic acid; EGF-like, epidermal growth factor-like; Gla–EGFN, a fragment comprising the Gla domain and the first EGF domain of factor X; Gla-EGFNC, a fragment comprising the Gla domain, the first and the second EGF domain of factor X; RU, response units. Note: this work was funded in part by the EU Biotechnology program (contract no BIO4-CT96-0662). (Received 20 December 2001, revised 23 April 2002, accepted 7 May 2002) with the best fit to the experimental data with three cooperativelyboundcalciumionsforboththeintactprotein andthefragment.Thedissociationconstant(Kd)forbinding to membranes containing 25 mol% PtdSer decreased from 4.6 lM for the isolated Gla-domain to 1 lM for the frag-ments Gla–EGFN and Gla–EGFNC (the Gla-domain and both EGF-like domains) fragments and to 40 nM for the entireproteinaszymogen,activatedenzymeorintheactive-siteinhibitedform.Analysisofthekineticsofadsorptionand desorption confirmed the equilibrium binding data. Keywords: blood coagulation; membrane binding; calcium dependence; factor X; Gla-domain. by inserting their side-chains into the membrane. This hypothesis gained support from site directed mutagenesis studies. In protein C the Leu5 fi Gln mutation reduces membrane affinity and biological activity [5,6]. NMR studies have illustrated how Ca2+ induces a drastic conformational transition in the Gla domain [7]. The Gla-residues at positions 6, 7, 16, 20, and 29 (bovine factor X numbering),solventexposedintheabsenceofCa2+,turnto the inside of the domain where they coordinate Ca2+, whereas the three hydrophobic residues, Phe4, Leu5 and Val8, located in the interior of the domain in the absence of Ca2+, become solvent exposed and form the hydrophobic patch[7].Theseresults,aswellasstudiesutilizingasynthetic GladomainwithLeu6andPhe9(factorIX,residues5and8 in factor X) substituted for a hydrophobic photoactivable crosslinking agent, suggested that there is an important hydrophobic component in the interaction of Gla-contain-ing proteins with biological membranes [8]. Although the Gla domain sequence is highly conserved among the various hemostatic Gla-containing proteins, the dissociation constant (Kd) for binding to model membranes varies by as much as three orders of magnitude [9]. Presumably, this is caused by still poorly understood electrostatic interactions between the Ca2+-bound Gla domain and phosphate head groups in the phospholipid membrane. This notion also gains support from numerous studies where site-directed mutagenesis was employed to establishthefunctionalroleofindividualaminoacidsinGla domains [9–11]. Membrane binding of vitamin K-dependent coagulation factors has previously been studied by ellipsometry [12,13], light scattering [9,14–16] and fluorescence polarization [17]. The Kd values determined for the same coagulation factor 3042 E.-M. Erb et al. (Eur. J. Biochem. 269) under similar conditions by different methods varied by as much as two orders of magnitude [12,13,17]. We therefore decided to investigate membrane binding by surface plasmon resonance. With this method the kinetics of membrane interaction is measured in real time. Also, the proteins do not have to be labeled with fluorescent compounds as in, for instance, fluorescence energy transfer studies. We have previously characterized the surfaces generated by liposome binding to the Biacore L1 sensor chip [18]. This sensor chip consists of a dextran matrix to which hydrophobic residues are covalently bound. Our results indicate that the liposomes were captured on the modified dextran matrix and subsequently fuse to generate a homogeneous lipid membrane. Moreover, a flat mem-brane is favorable as compared to the curvature of the liposomes [19–21]. To elucidate the impact of domains other than the Gla domain on membrane binding, we have now investigated the membrane-binding properties of coagulation factor X and Gla domain-containing frag-ments of this protein. MATERIALS AND METHODS Materials The lipids 1-palmitoyl 2-oleoyl-sn-glycero-3-phosphocho-line and 1,2-dioleoyl-sn-glycero-3-[phospho-L-serine] were obtained from Avanti Polar Lipids (Alabaster, AL, USA), polycarbonatefilterswerefromSPIsuppllies(WestChester, PA, USA). All other reagents were obtained from Merck (Darmstadt,Germany)orSigma(StLouis,MO,USA).The peptide corresponding to the Gla domain (residues 1–46) of factor X, was chemically synthesized using standard Fmoc chemistry. The fragments Gla–EGFN (residues 1–86) Gla– EGFNC (residues 1–140, 154–183) were generated by digestion of bovine factor X with trypsin [22]. Bovine factor X, factor Xa and DEGR-factor Xa were purchased from Haematologic Technologies Inc. (Burlington, VT, USA). All surface plasmon resonance experiments were performed on either a BIAcore X or a BIA2000 together with L1 pioneer sensor chips (Biacore AB, Uppsala, Sweden). Membrane generation Liposomes were prepared by the extruder technique and bound to the L1 sensor chip as described previously [18]. In brief, liposomes containing either 100 mol% PtdCho, 10 mol% PtdSer/90 mol% PtdCho, 25% mol% PtdSer/ 75 mol% PtdCho or 40 mol% PtdSer/60 mol% PtdCho were injected into a Biacore instrument equipped with a L1 sensor chip. The flow rate was 10 lLÆmin)1. Liposomes were captured on the sensor chip and spontaneously fused to generate a flat lipid membrane surface. Excess liposomes were removed by two 60 s pulses with 5 mM EDTA, pH 8.0 at a flow rate of 5 lLÆmin)1. The running buffer was then changed to 10 mM Tris/HCl, 150 mM NaCl, pH 7.4 (Tris buffer) containing 0.1% (w/v) bovine serum albumin (BSA). For titration experiments the buffer was made 0–10 mM in CaCl . For binding experiments, the Ca2+ concentration was 10 mM. All solutions used in the Biacore experiments were degassed and filtered through 0.22 lm filters. Ó FEBS2002 Ca2+-dependence of membrane binding Factor X and Gla–EGFN were diluted in the Tris buffer containing 0.1% (w/v) BSA, 0–10 mM CaCl2 to a final concentration of 39 nM and 2 lM CaCl , respectively. The running buffer always had the same Ca2+ concentration as the protein containing buffer. Association was followed for 180 s at a flow rate of 10 lLÆmin)1, followed by a 600-s dissociation phase using the same flow rate. The membrane was regenerated by two 60 s pulses with 5 mM EDTA pH 8.0 at a flow rate of 5 lLÆmin)1. The binding data were fitted to Eqn (1). Y ¼ R ½Ca2þn=ð½Ca2þn þ K0:5Þ ð1Þ whereRis the maximumresponse signal, n isthenumberof cooperatively bound Ca2+ ions needed for membrane binding and K0.5 is the Ca2+ concentration at which half-maximum binding occurs. Kinetics of membrane binding Membrane binding experiments on factor X, factor Xa, DEGR-factor Xa and the Gla-containing fragments of factor X were performed with membranes containing either 25 mol% PtdSer and 75 mol% PtdCho or 100 mol% PtdCho in the presence of 10 mM Ca2+. The Ca2+ concentration used here would be expected to almost completely saturate the Ca2+ binding sites in the Gla domain. The response signal, when using membranes containing 25 mol% PtdSer, was corrected for the back-ground binding to membranes composed of 100% PtdCho. Data were evaluated with the program BIAEVAL-UATION 3.0 using either the simple bimolecular interaction model or a two-step binding model as described by the following equations. The rate equation for the bivalent analyte model: A þ B );1 AB ð2Þ koff;1 AB þ B );2 AB ð3Þ koff;2 where d½B=dt ¼ 2kon;1½A½B þkoff;1½AB kon;2½AB[B] þ 2koff;2½AB2 ð4Þ d½AB=dt ¼ 2kon;1½A½B koff;1½AB kon;2½AB[B] þ 2koff;2½AB2 ð5Þ d½AB2=dt ¼ kon;2½AB½B 2koff;2½AB2 ð6Þ The rate equations for the conformational change model: A þ B kon;1 AB ð7Þ koff;1 AB kon;2 AB ð8Þ koff;2 where d½B=dt ¼ kon;1½A½Bþ koff;1½AB ð9Þ Ó FEBS2002 Membrane binding of coagulation factor X ( Eur. J. Biochem. 269) 3043 Table 1. Kinetic constants for binding of factor X and its Gla-containing fragments to membranes containing 25 mol% PtdSer in the presence of 10 mM Ca2+ obtained by evaluation of association and dissociation phases (I) and equilibrium binding data (II) as described in Materials and methods. Gla Gla–EGFN Gla–EGFN,C Factor X Factor Xa DEGR-factor Xa kon (MÆs))1 (8.0 ± 2.2) · 103 (4.5 ± 1.1) · 104 (6.7 ± 2.1) · 104 (8.3 ± 1.9) · 105 (4.5 ± 0.8) · 105 (5.3 ± 1.3) · 105 koff (s)1) (3.7 ± 0.2) · 10)2 (3.8 ± 0.2) · 10)2 (4.3 ± 0.2) · 10)2 (3.2 ± 0.2) · 10)2 (3.6 ± 0.2) · 10)2 (3.7 ± 0.2) · 10)2 Kd (M) (I) (4.6 ± 1.3) · 10)6 (8.4 ± 2.1) · 10)7 (6.4 ± 2.0) · 10)7 (3.9 ± 0.9) · 10)8 (8.0 ± 1.5) · 10)8 (8.0 ± 1.5) · 10)8 Kd (M) (II) (9.4 ± 1.4) · 10)6 (1.7 ± 0.3) · 10)6 (2.0 ± 0.3) · 10)6 (3.7 ± 0.6) · 10)8 (5.2 ± 0.8) · 10)8 (6.2 ± 0.9) · 10)8 d½AB=dt ¼ kon;1½A½B koff;1½AB kon;2½AB þ koff;2½AB ð10Þ d½AB =dt ¼ kon;2½AB koff;2½AB ð11Þ The concentrations at t ¼ 0 are [B]0 ¼ Rmax, Rmax ¼ response at full saturation, [AB]0 ¼ 0 and [AB2]0 ¼ 0. The total response signal is the sum of the initial response signalRi plusthesignalsfromthecomplexesABandAB2 or AB* for the bivalent model or for the conformational change model, respectively. Equilibrium response signals Equilibrium response signals were plotted vs. the protein concentration. The Kd values were determined by fitting the data to Eqn (2) assuming a single class of binding sites: serine protease domain alter those Ca2+-binding properties of factor X that are relevant to membrane binding. Experi-ments using membranes containing either 10 mol% PtdSer/ 90 mol% PtdCho or 40 mol% PtdSer/60 mol% PtdCho showed the same Ca2+-dependence as 25 mol% PtdSer/ 75 mol% PtdCho for binding intact factor X and Gla– EGFN (data not shown). Kinetics of membrane binding The kinetics of binding to PL membranes of the zymogen factor X, activated factor X (factor Xa) and the active site inhibited form DEGR-factor Xa as well as the the factor X peptides were studied with surface plasmon resonance. The Ca2+ concentration was 10 mM to ascertain that the Ca2+ binding sites of the Gla domain were completely satur-ated. Figure 2 presents the binding of factor X to the saturation ¼ ½protein=ð½proteinþ KdÞ: ð12Þ The equilibrium response signal is the sum of the signals from the intermediate complex AB and the final complex AB2. However, the contribution of the second binding step to the total response is about 15%, and therefore the evaluation of the equilibrium response signals by Eqn (2) gives a good approximation for the Kd values of the first binding step. The uncertainties given in Table 1 are therefore set to 15%. RESULTS Ca2+-dependence of membrane binding The Ca2+ concentration dependence of membrane binding was determined by measuring the equilibrium response signal at different Ca2+ concentrations. Factor X and the fragmentGla–EGFN werebound to membranes containing 25 mol% PtdSer/75 mol% PtdCho at a concentration of 39 nM and 2 lM, respectively. Binding of both species to membranes composed of 100 mol% PtdCho was less then 5% of the binding to membrane containing 25 mol% PtdSer. The Ca2+ titration curves of factor X and Gla– EGFN binding indicate cooperative binding (Fig. 1). Half-maximal binding occurred at a calcium concentration of 1.5 and 1.4 mM for factor X and Gla–EGFN, respectively, whichisclosetotheconcentrationoffreecalciuminbloodof 1.2 mM. The best fit to the data in Fig. 1 was obtained assumingthreecooperativelyboundCa2+ ions.Asshownin Fig. 1themembranebindingofintactfactor XandtheGla– EGFN fragment, showed very similar Ca2+-dependencies, indicating that neither the second EGF domain nor the Fig. 1. Ca2+-dependence in the membrane binding of factor X (A) and the fragment Gla–EGFN (B) as determined by surface plasmon reson-ance. Binding experiments were performed on 25 mol% PtdSer-con-taining membranes (solid symbols) and 100 mol% PtdCho-containing membranes (open symbols). The solid curve is the best fit to the experimental data points obtained by Eqn (1), assuming n ¼ 3 (c2 ¼ 359.2); the dotted line assuming n ¼ 4 (c2 ¼ 595.7); the dashed line assuming n ¼ 2 (c2 ¼ 715.3). 3044 E.-M. Erb et al. (Eur. J. Biochem. 269) Fig. 2. Adsorption and desorption kinetics of factor X to 25 mol% PtdSer containing membranes. Experiments were performed using 10 mM Tris/HCl, pH 7.5, 150 mM NaCl, 10 mM CaCl2, 0.1% (w/v) BSA as running buffer at a flow rate of 10 lLÆmin)1. Factor X was diluted in the same buffer to the final concentration of 44 nM (h), 22 nM (j), 11 nM (n), 5.5 nM (m), 2.8 nM (s) and 1.4 nM (d). The protein was injected at t ¼ 0 and binding to the membrane is apparent during the association phase (180 s). The protein-containing buffer was then replaced by running buffer, resulting in dissociation of the protein from the membrane. The solid curves were calculated using equations 4–6. phospholipid membrane at various protein concentrations. Similar sensorgrams were obtained for the other forms of factor X and fragments, although with different concentra-tions for half maximum binding (data not shown). In a first attempt the association and dissociation processes were treated as simple one step processes. However, with this approach it was not possible to obtain a reasonable agree-ment between observed and calculated sensorgrams. Mod-els with two on-rates and two off-rates improved the fit significantly. Moreover, a model including a conformation-al change and a model including a bivalent analyte both gavegoodfitstotheexperimentaldata.Theresultsobtained withthebivalentanalytemodelisshowninFig. 2.Inallcases there is a dominating fast process with an almost constant off-rate for all the proteins (3.2–4.8 10)2Æs)1). The difference in binding affinity is therefore the resultof different on-rates (Table 1). The isolated Gla domain (the fragment with the lowest molecular mass, about 5 kDa) shows the lowest on rate, even though from thermodynamic aspects it would be expectedtoshowahigheronrate.Thismaybeexplainedby assuming that only a small fraction of the fragment has a conformation that is commensurate with membrane-bind-ing.Theon-ratesforGla–EGFN andGla–EGFNC areabout a factor of five higher than for the Gla-domain. This can presumably be attributed to a stabilizing effect of the N-terminal EGF domain on the Gla domain [7]. The entire proteinhasanon-ratethatistwoordersofmagnitudefaster than for the Gla-domain presumably due to a further stabilization of the structure of the Gla-domain, indicating that less than 1% of the free isolated Gla-domain has a conformation that is appropriate for membrane binding. Equilibrium binding isotherms Theconcentrationdependenceoffactor Xbindingisshown in Fig. 2. It is apparent that the adsorption is rapid and that Ó FEBS2002 Fig. 3. Equilibrium isotherms of factor X and its Gla-containing frag-ments binding to membranes containing 25 mol% PtdSer in the presence of 10 mM Ca2+. The measured equilibrium binding signal is plotted against the solution phase concentration of factor X (d), factor Xa (m),DEGR-factor Xa(n),Gla–EGFNC (e),Gla–EGFN (r)andGla (.). Solid lines indicate the least-square fit of the Langmuir model to thisdataasdescribedinMaterialsandmethods.Theestimatedbinding parameters are listed in Table 1. a plateau is reached within 100–200 s. Figure 3 shows the binding isotherms of factor X and its peptides. Their mem-brane binding affinities increase in the order Gla nguon tai.lieu . vn