Structurally and functionally distinct Ca2+ binding sites in the y ...

Eur. J. Biochem. 234, 293-300 (1995) 0 FEBS 1995

Structurally and functionally distinct Ca2+binding sites in the y-carboxyglutamic acid-containing domain of factor VIIa Egon PERSSON’ and Lars C. PETERSEN’ ’ Department of Hemostasis Research, Vessel Wall Biology, Health Care Discovery, N o w Nordisk N S , Gentofte, Denmark ’ Department of Leucocyte Research, Vessel Wall Biology, Health Care Discovery, Novo Nordisk A/S, Gentofte, Denmark (Received 30 June 1995) - EJB 95 106113

The structural and functional effects of Ca’+ binding to vitamin-K-dependent coagulation factor VIIa were investigated. Conformational changes with a midpoint around 0.7 mM Ca2+ quenched the intrinsic protein fluorescence of a fragment of factor VIIa comprising only the light chain and this coincided with an increase in factor VIIa amidolytic activity in the absence of tissue factor. Ca’+ binding to sites in factor VIIa and in the fragment with an apparent dissociation constant of 1.3- 1.4 mM induced binding to phospholipids. A similar Caz+ dependency was not observed with factor VIIa lacking the N-terminal 38 or 44 residues of the light chain and the observed effects could thus be attributed to y-carboxyglutamicacid-dependent Ca2+ binding. Mg” appeared to bind to the site(s) of relatively higher affinity since, although it was less efficient than Ca”, it stimulated the amidolytic activity and induced quenching of the intrinsic fluorescence. In contrast, Mg’+ did not induce expression of the phospholipid-interactive structure. The binding properties of two monoclonal antibodies that recognized epitopes in the y-carboxyglutamic-acid-rich domain of factor VIIa corroborated the occurrence of two Ca’ ’ -induced, sequential structural changes and only one of the antibodies recognized the Mg2’ -induced structure. Thus Ca2+ binding to the y-carboxyglutamic-acid-containingdomain appeared to result in at least two distinct structural transitions with different functional consequences. The two (sets of) sites responsible for the observed effects could be distinguished based upon differences in Ca’+ affinity and metal ion selectivity. The interaction between factor VIIa and tissue factor was monitored by means of a direct binding assay and an amidolytic assay. In both systems, half-maximal Ca’+ enhancement was observed at 0.25 mM. This coincided with a Ca’+-induced conformational change in factor VIIa associated with fluorescence quenching. The same effect on amidolytic activity was observed with the two N-terminally truncated forms of factor VIIa and it is presumably mediated by Ca” binding to a site located in the serine protease part. Keywords: factor VIIa; Gla domain ; Ca’+ binding; phospholipid binding ; intrinsic fluorescence.

Factor VIIa triggers the extrinsic pathway of blood coagulation when complexed to its membrane cofactor tissue factor (TF; Bsterud, 1990). Factor VIIa is capable of activating the zymogen factor VII (Pedersen et al., 1989; Nakagaki et al., 1991 ; Yamamot0 et al., 1992) and also converts factors IX and X to active enzymes that catalyze reactions down-stream in the clotting cascade (Bsterud and Rapaport, 1977, 1980). Factor VII belongs to a family of proteins recognized by a characteristic modular organization and requiring vitamin K for their biosynthesis (Furie and Furie, 1988). The amino-terminal membrane-binding domain contains y-carboxylated glutamic acid (Gla) residues, post-translationally modified by a carboxylase in a vitamin-K-dependent reaction (Vermeer, 1990); the carboxy-terminal part contains the serine protease domain. Two doCorrespondence ro E. Persson, Department of Hemostasis, Health Care Discovery, Novo Nordisk A/S, Hagedornsvej 1, HAB3.93, DK2820 Gentofte, Denmark Fax: +45 44 43 81 10. Ahhreviafions. Gla, ;yarboxyglutamic acid; EGF, epidermal growth factor; TF, tissue factor; Tris/NaCI/EDTA,50 mM Tris pH 8.0 containing 0.1 M NaCl and 2 mM EDTA; Hepes/NaCI, 50 mM Hepes pH 7.5 containing 0.1 M NaCI. Enzymes. Coagulation factor VITa (EC 3.4.21.21): trypsin (EC 3.4.21.4); cathepsin G (EC 3.4.21.20).

mains homologous to the epidermal growth factor (EGF) separate the membrane-interactive and the enzymatically active regions in the linear sequence (Hagen et al., 1986). The roles of different domains in this family of multi-modular proteins are complex. Several regions of factor VIIa appear to contribute in a crucial manner to the interaction with TF. This includes the Gla module including residues 38-45 of the hydrophobic stack (Sakai et al., 1990; Ruf et al., 1991; Wildgoose et al., 1992; Petersen et al., 1994), the EGF-like region (Toomey et al., 1991) and sequences in the protease domain (Wildgoose et al., 1990; Kumar and Fair, 1993). There is also evidence suggesting the importance of an intact GlaEGF region (Kazama et al., 1993; Higashi et al., 1994). Interestingly, the Gla, EGF-like and protease domains all contain one or more Ca’+ binding sites (SorianoGarcia et al.. 1992; Schifldt et al., 1992; Wildgoose et al., 1993). Ca’+ binding to the Gla module is known to be essential for membrane binding (Stenflo and Suttie, 1977) and saturation of the site in the catalytic domain appears to be important for activity and tissue factor binding of factor VIIa (Wildgoose et al., 1993), but much is still unknown about the structural and functional effects of Ca2’ binding to the different sites. To further investigate the role of Ca” binding to various regions of factor VIIa, we have studied functional and structural properties of factor VIIa and derivatives thereof as a function of the Ca’+

294

Persson and Petersen (EUKJ . Biochem. 234)

concentration. In this report we describe the effects of two different Ca’+ binding events in the Gla domain.

MATERIALS AND METHODS Proteins. Human recombinant factor VIIa was expressed and purified as previously described (Thim et al., 1988). Des(138)-factor VIIa was produced by factor VlIa autoproteolysis (Sakai et al., 1990), whereas des(1-44)-factor VIIa was produced by cathepsin-G-mediated cleavage (Nicolaisen et al., 1992). Cathepsin G was purified from human neutrophiles by the method of Baugh and Travis (1976). TF-(1-218)-peptide was kindly provided by Dr W. Kisiel (University of New Mexico, Albuquerque) and TF (full-length) was from Corvas International. The factor VIIa Gla peptide (residues 1-38) was purified from autodegraded factor VTIa by a combination of ion-exchange (Mono Q, Pharmacia Biotech) and reverse-phase (C,) chromatography. A fragment containing residues 1- 144 and 248-266 of factor VII [factor VII-(l- 144)-(248-266)-peptide] was produced according to Kazama et al. (1993) with the following exceptions. To prevent cleavage at Lys residues, factor VIIa was treated with citraconic anhydride (Aldrich) prior to tryptic degradation (Persson et al., 1989) and the digestion time was 30 min. After the digestion, excess EDTA over CaCI2 was added, pH w justed to 8.0 and the digest was applied to a column of Q-Sepharose Fast Flow (Pharmacia Biotech) in SO mM Tris pH 8.0 containing 0.1 M NaCl and 2 mM EDTA (Tris/NaCl/EDTA). The column was washed with this buffer and the protein eluted with a linear 0.1 -0.6 M NaCl gradient. SDS/PAGE (Laemmli, 1970) on a 12.5% gel was used to localize fractions containing the factor VII-(1 - 144)-(248-266)-peptide. The lysine blocking groups were subsequently removed (Persson et al., 1989). The volume of the sample was reduced to 2 ml using a Centricon-I0 (Amicon) and final purification of the factor Vll-( 1 - 144)(248-266))peptide was achieved by gel permeation chromatography on a column (1.6X89cm) of Sephadex G-75 Superfine (Pharmacia Biotech) in Tris/NaCl/EDTA. Amino acid analysis after acid hydrolysis was performed as earlier described (Persson et al., 1989) and N-terminal amino acid sequences were determined on an Applied Biosystems 475A pulsed liquid sequencer according to the manufacturer’s instructions. Intrinsic protein fluorescence and terbium phosphorescence measurements. The intrinsic protein fluorescence and Tb” phosphorescence were measured at 25°C in a Perkin Elmer LS 50B spectrofluorimeter equipped with FLDM software. Prior to analysis, the proteins were transferred to 50 mM Tris pH 7.5, 0.1 M NaC1, treated with Chelex’“ 100 (Bio-Rad), using a NAP5 column (Pharmacia Biotech). See legends to Figs 2-4 for protein concentrations. In the fluorescence experiments, the excitation wavelength was 280 nm and the excitation and emission bandwidths were 5 nm. Fluorescence emission spectra were recorded between 290-450 nm in the absence of Ca’ ‘ and in the presence of 0.25 mM or 20 mM Ca2+. The Ca’ ’ dependencies of the intrinsic fluorescence of factor VIIa and factor VII-(l144)-(248-266)-peptide were studied in titration experiments using 5-p1 additions of stock solutions of CaCI2 or MgCL, ranging over 1 mM-2 M, to a 0.5-ml sample. The emission intensity at 340 nm (factor VIIa) or 350 nm [factor V11-(1-144)-(248266)-peptide] was measured 1.5 s after each addition of metal ion by averaging the signal during 40 s. The titrations were performed in duplicate. Corrections were made for successive dilution of the samples. The data were expressed as relative fluorescence (AFIAF,,,,), where A F is the intensity at each metal ion concentration minus the intensity at 20 mM Ca” and AF,,,, is the intensity in the absence of metal ions minus the intensity at 20 mM Ca’ +.

Terbium ion phosphorescence spectra were recorded using a pulsed Xe light source with a width at half peak intensity of less than 10 ps. The background due to stray light and intrinsic protein fluorescence was eliminated by insertion of a 390-nm cutoff filter in the emission beam and by measuring the phosphorescence intensity with a 50-ps delay. The protein-Tb” complex was excited at 285 nm (slit width 15 nm) and phosphorescence intensity was measured in the range between 440-640 nm (slit width 5 nm). Phospholipid binding assay. A 50-pg/ml phospholipid mixture containing 80% PtdCho (Sigma) and 20% PtdSer (Sigma) in ethanol was prepared and 40 p1 was added to each well in a microtiter plate (MaxiSorp, Nunc). The solvent was allowed to evaporate and the wells were blocked with 200 pl SO mM Hepes pH 7.5, 0.1 M NaCl (HepedNaCl) containing 10 mg/ml gelatin (EIA grade, Bio-Rad) for 3 h. The wells were washed twice with Hepes/NaCI containing 1 mg/ml gelatin and 1 mM EDTA, 0.05-20 mM CaCl, or 20 mM MgCI, (washing buffers). Factor VIla, des(l-38)-factor VIIa, des(1-44)-factor VIIa or factor VII-(1-144)-(248-266)-peptide ( S O ng/ml) was added together with a mAb (F7A2, 0.5 pg/ml) that recognizes a metal-ion-independent epitope in the EGF-like region of factor VIIa and peroxidase-labeled rabbit anti-mouse Ig (0.6 pg/ml, DAKO) in 100 p1 of the different washing buffers. The mixture was incubated for 1 h. After washing five times with washing buffers, 100 p1 ophenylenediamine (Novo Nordisk BioLabs), 0.8 mglml in 0.1 M sodium acetate pH 5.2 was added. The colour was allowed to develop for 15 min and the reaction was terminated by the addition of 100 pl 2 M H2S04.The absorbance at 492 nm was measured in an EAR 400 AT ELISA reader (SLT-Labinstruments) with a reference wavelength of 620 nm. Factor VIIa-tissue factor binding assay. A microtiter plate (MaxiSorp, Nunc) was coated for 2 h with streptavidin (3.3 pg/ nil, Vector Laboratories) in 1 0 m M sodium phosphate pH 7.5, 0.15 M NaCl. The wells were washed four times with 10 mM Tris pH 7.2 containing 0.15 M NaCl and 0.25% Triton X-70.5. The next step was a 1-h incubation with biotinylated factor VIIa (100 ng/ml) in 2 mM Tris pH 8.0, 0.15 M NaCI, 0.05% Tween 20 (assay buffer). Factor VIIa was biotinylated using a fivefold molar excess of biotin-X-NHS (Calbiochem). The washing step was repeated followed by a 2-h incubation with TF (100 ng/ml) and a peroxidase-conjugated anti-TF mAb (0.5 pg/ml) directed against the cytoplasmic tail in assay buffer containing 0- 10 mM CaCI,. After washing, 3,3’,5,5’-tetramethyIbenzidinedihydrochloride (0.2 mg/ml, Sigma) in 40 mM sodium acetate, pH 5.0, containing 3.2 mM sodium perborate was added and the colour developed for 20 min. The reaction was stopped with 4 M H,P04 and the absorbance measured at 4 5 0 n m as described above. Amidolytic assays. Factor VIIa, des(l-38)-factor VIIa or des(l-44)-factor VIla, final concentration 100 nM, and S-2288 (Chromogenix), final concentration 2 mM, was mixed i n Hepes/ NaCl containing 1 mg/ml gelatin and 1 m M EDTA, 0.0520 mM CaC1, or 20 mM MgC12. The absorbance at 405 nm was measured immediately and with 10-min intervals during a 30min incubation at ambient temperature. The absorbance increase was linear with time. The results [A,,,(30 min)-A,,,(O min)] were corrected for the absorbance development in a blank sample containing only S-2288 and given as a percentage of the result obtained at 1 0 mM CaCl,. An amidolytic assay was also performed in the presence of TF-(1-218)-peptide, in which 10 nM TF-(1-218)-peptide and 20 nM factor VIIa or 20 nM TF-(l-218)-peptide and 20 nM des(1 -38)-factor VIIa or des(l-44)-factor VIIa was mixed with S-2288 as described above. The absorbance was measured every 10 min during a SO-min incubation. The increase in absor-

295

Persson and Petersen (EUKJ. Biochem. 234)

bance was linear with time. To specifically obtain the TF-induced enhancement, the results [A,,,,(50 min)-A,,,,(O min) J were corrected for the absorbance development in samples containing only the respective form of factor Vtta and substrate at the different metal ion concentrations and given as a percentage of the result obtained at 10 mM CaCI,. Antibody binding experiments. Microtiter plates (MaxiSorp, Nunc) were coated overnight with a mAb (F1A2 or F6A4) against factor VII, 10 pg/ml i n 100 pI 0.1 M NaHCO, pH 9.8. Remaining binding capacity was blocked by incubation for 3 h with Hepes/NaCI containing 0.25% Tween 20 and 10 mg/ml gelatin. The wells were washed three times with Hepes/NaCI containing 0.02% Tween 20, 1 mg/ml gelatin and 1 mM EDTA, 0.05-20 mM CaC1, or 0.1 -20 mM MgCI, (washing buffers). Factor VIIa, des(l-38)-factor Vlla, des(l-44)-factor VIIa or factor VII-(1-144)-(248-266)-peptide, 100 ng/ml in 100 pl of the different washing buffers, was added followed by incubation for 1 h. The wells were washed five times with the different washing buffers before the incubation for 1 h with 100 pl biotinylated rabbit anti-(human factor VII) (1.5 pg/ml) in the washing buffer with 10 mM CaCI2.The plate was washed three times with this buffer and then incubated for 1 h with 100 p1 avidinconjugated peroxidase (Kem-En-Tec), 0.5 pg/ml in the same buffer. The plates were washed five times, developed for 4 min and results measured as described for the phospholipid binding assay. In a competition experiment, the factor VII Gla peptide, 1 or 10 pg/ml, was incubated together with factor VIIa at 10 mM CaC1,. For the production of biotinylated rabbit anti-(human factor VIIa), a polyclonal antiserum was purified on protein ASepharose (Pharmacia Biotech) and subsequently biotinylated with N-hydroxysuccinimidobiotin (Sigma).

RESULTS Purified factor VIIa derivatives. Three forms of factor VIIa, with N-termini of the light chain at residue 1, 39 or 45, were used in this study. Amino-terminal sequence analyses confirmed the N-termini Ala-Asn-Ala-Phe-Leu, Leu-Phe-Trp-Ile-Ser and Ser-Asp-Gly-Asp-Gln, respectively (Hagen et al., 1986). All three forms were shown to contain the sequence Ile-Val-GlyGly-Lys, which is the N-terminal of the heavy chain corresponding to residues 153-157. Analysis of the factor VII-(l-144)(248 - 266)-peptide gave the sequences Ala-Asn-Ala-Phe-Leu (residues 1 -5) and Leu-His-Gln-Pro-Val (residues 248-252) in equal amounts, demonstrating that a heavy-chain-derived peptide was disulfide-linked to the GlaEGF region. No other sequences were observed, showing that citraconylation prevented the intenial cleavage after LyslO9 described in an earlier report (Kazama et al., 1993). Amino acid analysis agreed with a fragment comprising residues 1 - 144 disulfide-linked to residues 248-266 (not shown). SDSPAGE of the purified proteins is shown in Fig. 1. Effects of metal ions on the intrinsic protein fluorescence. The Ca”-induced changes of the intrinsic fluorescence of various factor VlIa derivatives were investigated. The effect on the fluorescence spectrum of factor VIIa of adding 0.25 mM or 20 mM Ca” is shown in Fig. 2A. Binding of Ca” to a highaffinity site caused quenching at 0.25 mM without change in fluorescence peak position, whereas subsequent increase of the CaL+concentration to 20 mM caused a visible blue-shift from 342 nm to 340 nm in addition to further quenching. The effect of Ca” on the intrinsic fluorescence of des(l-38)-factor VlIa, des(l-44)-factor VIIa and the factor VII-(1-144)-(248-266)peptide was studied to characterize the Ca2‘ -induced fluores-

43 -

30 20 A

B

C

D

Fig. 1. SDSPAGE of the factor VIIa derivatives. The factor VII-(1144)-(248-266)-peptide (21 pg, lane A), factor VIIa (24 pg, lane B), des(l-38)-factor VIIa (28 pg, lane C) and des(l-44)-factor VIIa (12 pg, lane D) were subjected to SDWPAGE on a 10- 15 % gradient gel after reduction with dithiothreitol. The gel was stained with Coomassie brilliant blue R-250. The molecular masses of marker proteins are shown on the left in kDa.

B

0. 300

340

.

380

Em (nm)

-

0

420 300

____.__

340

380

420

Em(nm)

Fig.2. Ca” effects on the fluorescence spectra of factor VIIa and factor VI1-(1-144)-(248-266)-peptide. The emission spectra of factor VIIa (A) and the factor VII-(l-144)-(248-266)-peptide (B) after excitation at 280 nm are shown. The protein concentrations were 1 pM and 2 pM, respectively. The spectra were r rded in the absence of Ca’+ (-) or in the presence of 0.25 mM ( ..) or 2 0 m M (---) Ca2’.

cence changes in factor VIIa in more detail. As previously reported (Schiedt et al., 1992), Ca2+ caused approximately 1 0 % fluorescence quenching (apparent K

Z 60

.-0

-x I

40

0

0.01

0.1

1

0.01

10

Fig. 6. The amidolytic activities of intact and N-terminally truncated factor VIIa. Amidolysis of S-2288 catalyzed by factor VIIa (H),des(138)-factor VIIa (0)and des(1-44)-factor VIIa (A), all 100 nM in Hepes/NaCl containing 1 mglml gelatin at the indicated concentrations of Ca'+, was measured. Each point represents the mean of two duplicate experiments. The amidolysis rate for each enzyme at 10 mM Ca2+was set to 100%.

of C P stimulated the amidolytic activity of factor VIIa in the absence of TF. Half-maximal enhancement was obtained at 0.6 mM Ca'+ and, under our experimental conditions, saturating concentrations of Ca" gave a 20-fold increase in activity compared with the activity measured in the presence of EDTA (Fig. 6). This is in good agreement with earlier reports (Wildgoose et al., 1993; Neuenschwander and Morrissey, 1994; Butenas et al., 1994). Mg?+ also stimulated, although less efficiently than Ca'+, the amidolytic activity of factor VIIa. The effect of 20 mM Mg2+ was comparable to that of 1 mM Ca'+. The activities of des(1-38)- and des(l-44)-factor VIIa displayed different Ca?' dependencies than did factor VIIa and the N-terminally truncated forms of factor VIIa reached maximal activity at much lower Ca" concentrations. This suggests an involvement of Gla-dependent Ca" binding sites in achieving full factor VIIa amidolytic activity in the absence of TF. Several studies have shown that TF enhances the activity of factor VIIa in a Ca"-dependent manner. This effect has been assumed to reflect the formation of a factor VIIa-TF complex. To further investigate this, we compared measurements of factor VIIa-TF amidolytic activity and complex formation as a function of the Ca2' concentration. Based on the determination of the dissociation constant (3 pM) for the interaction of factor VIIa with T F in the absence of Ca?' (Neuenschwander and Morrissey, 1994), we assumed that Ca'+ binding to factor VIIa would be a prerequisite for factor VIIa-TF complex formation at our experimental conditions. In the amidolytic assay in the presence of TF, factor VIIa exhibited half-maximal activity around 0.25 mM Ca" (Fig. 7). Very similar results were obtained with des(1-38)- and des(l-441-factor VIIa (not shown). Therefore, the TF-induced activity enhancement appeared to depend on CaZ+binding to site(s) outside the Gla domain. Results presented herein and those obtained with factor VIIa mutants Glu220-+Lys and Glu22O+Ala (Wildgoose et a]., 1993), suggest that this site is located in the serine protease domain. At our protein concentrations, T F increased the factor VIIa activity approximately 100-fold at saturating concentrations of Caz+, whereas the activities of the truncated forms of factor VIIa were enhanced 10- 15-fold. This supports earlier findings that the Gla domain is essential for an optimal interaction between factor VIIa and TF (Ruf et al., 1991). The direct binding assay showed a Ca' ' dependency of factor VIIa-TF complex assembly identical to that of the TF-induced enhancement of factor VIIa amidolytic activity (Fig. 7), i.e. binding of factor VIIa to T F appeared to cause the activity increase.

0.1

10

1

[ C a z + ] (mM)

[Ca2+] (mM)

Fig. 7. The Ca2+-inducedbinding of factor VIIa to tissue factor. Amidolysis of S-2288 by factor VIIa in the presence of 10 nM TF-(1-218)peptide (H) and factor VIIa-TF complex formation monitored in the direct binding assay (A) at different Ca" concentrations are shown (see Materials and Methods). Each data point represents the mean of duplicate experiments. In the amidolytic assay, the activity in the absence of TF-(1 -218)-peptide was subtracted and the activity at 10 mM Ca2' was set to 100%.In the binding assay, the signal at 10 mM Ca'+ was defined as 100%.

100 h

5

._ E 80

i E (c

s

60 40

C

p

20

0 0.1

1

10

[MeZ+] (mM)