Structural requirements for Ca2+ binding to the gamma-carboxyglutamic acid and epidermal growth factor-like regions of factor IX. Studies using intact domains isolated from controlled proteolytic digests of bovine factor IX.

Blood coagulation factor IX is composed of discrete domains with an NH2-terminal vitamin K-dependent gamma-carboxyglutamic acid (Gla)-containing region, followed by two domains that are homologous with the epidermal growth factor (EGF) precursor and a COOH-terminal serine protease part. Calcium ions bind to the Gla-containing region and to the NH2-terminal EGF-like domain. To be able to determine the structure and function of the Gla- and EGF-like domains, we have devised a method for cleaving factor IX under controlled conditions and isolating the intact domains in high yield, either separately or linked together. The Ca2+ and Mg2+ binding properties of these fragments were examined by monitoring the metal ion-induced changes in intrinsic protein fluorescence. A fragment, consisting of the Gla region linked to the two EGF-like domains, bound Ca2+ in a manner that was indistinguishable from that of the intact molecule, indicating a native conformation. The Ca2+ affinity of the isolated Gla region was lower, suggesting that the EGF-like domains function as a scaffold for the folding of the Gla region. The Gla-independent high affinity metal ion binding site in the NH2-terminal EGF-like domain was shown to bind Ca2+ but not Mg2+. A comparison with similar studies of factor X (Persson, E., Björk, I., and Stenflo, J. (1991) J. Biol. Chem. 266, 2444-2452) suggests that the Ca2(+)-induced fluorescence quenching is due to an altered environment primarily around the tryptophan residue in position 42.

Blood coagulation factor IX is composed of discrete domains with an NHz-terminal vitamin K-dependent y-carboxyglutamic acid (G1a)-containing region, followed by two domains that are homologous with the epidermal growth factor (EGF) precursor and a COOHterminal serine protease part. Calcium ions bind to the Gla-containing region and to the NHz-terminal EGFlike domain. To be able to determine the structure and function of the Gla-and EGF-like domains, we have devised a method for cleaving factor IX under controlled conditions and isolating the intact domains in high yield, either separately or linked together. The Caz' and M 2 + binding properties of these fragments were examined by monitoring the metal ion-induced changes in intrinsic protein fluorescence. A fragment, consisting of the Gla region linked to the two EGF-like domains, bound Ca2+ in a manner that was indistinguishable from that of the intact molecule, indicating a native conformation. The Ca2+ affinity of the isolated Gla region was lower, suggesting that the EGF-like domains function as a scaffold for the folding of the Gla region. The Gla-independent high affinity metal ion binding site in the NHz-terminal EGF-like domain was shown to bind Ca2+ but not Mg'. suggests that the Ca2+-induced fluorescence quenching is due to an altered environment primarily around the tryptophan residue in position 42.
Factor IX is a vitamin K-dependent plasma zymogen that upon activation by factor VIIa-tissue factor or factor XIa is converted to a serine protease active in blood coagulation (1-4). Together with the closely related factors VII, X, and protein C, factor IX forms a subgroup among the vitamin Kdependent clotting factors that is characterized by identical domain structure, a pronounced sequence homology, and identical positioning of introns separating the exons that code for the various domains (2, 5-9). The mature factor IX molecule *This investigation was supported by grants from the Swedish Medical Research Council (Projects 4487 and 4212), Albert PBhlsson's Foundation, Kock's Foundation, Osterlund's Foundation, Magnus Bergvall's Foundation, and Knut and Alice Wallenberg's Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
T To whom correspondence and reprint requests should be addressed.
consists of an NHz-terminal y-carboxyglutamic acid (G1a)'containing region, followed by two domains that are homologous to the epidermal growth factor (EGF) precursor and a COOH-terminal serine protease part. In factor IX, as well as in prothrombin and factor X, the Gla region has an ordered structure in the presence of Cap+ (10-12). It can easily be removed by limited proteolytic cleavage in factor IX and related proteins, suggesting that it is quite mobile relative to the EGF-like domains, at least in the absence of Cap+ (13-18). The EGF-like domains are independently folded structural units, a conclusion that derives from the observations that synthetic EGF-like domains fold spontaneously into their native conformation and that the isolated EGF-like domains of factors IX and X and protein C retain a Cap+ binding site (19)(20)(21)(22)(23)(24). The NH2-terminal EGF-like domain of bovine factor IX (BfIX) has one P-hydroxyaspartic acid (Hya) residue, whereas the corresponding residue of human factor IX is only partially hydroxylated (25,26). In addition, this domain has a unique 0-linked carbohydrate side chain in both bovine and human factor IX (27)(28)(29).
The Gla region of factor IX binds Cap+ and phospholipid and is also involved in binding of factor IX to a putative endothelial cell receptor (30)(31)(32)(33). The NHp-terminal EGF-like domain binds one Cap+ and also seems to interact with the same endothelial cell receptor (20,21,33). The serine protease part of activated factor IX has been implicated in the interaction with the cofactor, factor VIIIa (34), although indirect evidence suggests that there is also a weak interaction between the COOH-terminal EGF-like domain and factor VIIIa (35).
To elucidate the function of the EGF-like domains of factor IX, proteolytic fragments that consist of one or two EGF-like domains linked to the Gla region, to ensure phospholipid affinity, would be most useful. We have therefore exploited the putative interdomain mobility, the compact structure of the EGF-like domains, and the fact that the highly anionic Gla-containing region is quite resistant to degradation by proteolytic enzymes and developed a procedure to cleave factor IX under controlled conditions and to isolate the various domains. In contrast to methods that depend on the expression of recombinant proteins in pro-or eukaryotic 1. systems or the chemical synthesis of the domains, the current approach has the advantage that the disulfide bond pairing and the postribosomal modifications do not pose a problem. Furthermore, considerable amounts of the highly purified intact domains can be obtained if bovine proteins are used as starting material.

BfIX-Gla
In this communication, we describe the isolation and characterization of a fragment, containing the two EGF-like domains, from a controlled proteolytic digest of bovine factor IX, as well as a fragment containing these two domains linked to the Gla region. We also demonstrate that the Gla region, linked to the EGF-like domains, has a native conformation as judged by its normal Ca2+ binding properties. Moreover, we report that the Gla-independent high affinity Ca2+ binding site does not bind M e .

RESULTS
Isolation and Characterization of Factor I X Domains-The strategy for the isolation of fragments of bovine factor IX (BfIX) is shown in Fig. 1. Cleavage of BfIX with chymotrypsin was monitored by SDS-PAGE, which revealed the formation of several discrete fragments (Fig. 2). One of these fragments, with an apparent molecular weight of 26,000, gave two se- quences in equimolar amounts: Tyr-Asn-Ser-Gly-Lys-Leu, corresponding to residues 1-6; and Val-Thr-Pro-Ile-Cys-Ile corresponding to residues 286-291 of intact BfIX (45). This fragment was isolated from a digest of BfIX (50-90 mg) by chromatography on a column of Q-Sepharose Fast Flow followed by gel filtration on a Sephadex G-75 column (Figs. 3 and 4). Sequence analysis indicated that the isolated fragment (BfIX-GlaEGFNc) was at least 95% homogeneous and gave no evidence for internal peptide bond cleavage (less than 5%). The recovery was between 25 and 35% in three experiments. The isolated BfIX-GlaEGFNc was reduced and alkylated and the two peptides were separated by HPLC. The long peptide, derived from the NH2-terminal part of the molecule, contained 11.8 mol of Gla and 1.1 mol of Hya per mol of protein and had an amino acid composition corresponding to residues 1-144 of intact BfIX (Table I). The small peptide, derived from the serine protease part, had a composition in agreement with residues 286-296. The absorption coefficient (A:% at 280 nm) of the intact fragment was determined to be 10.5, based on a molecular weight of 18,170 for the apoprotein (calculated from the amino acid composition).
Cleavage of BfIX-GlaEGFNc (4-8 mg) with lysyl endopeptidase in an EDTA-containing buffer gave a fragment with an apparent molecular weight of 20,000 (Fig. 11B). This fragment, which was isolated as described above (Fig. 5), gave two sequences in equimolar amounts; Gln-Tyr-Val-Asp-Gly-Asp, corresponding to residues 44-49 of intact BfIX, and one sequence corresponding to residues 286-291 (see above) (45). The fragment was at least 95% homogeneous. Approximately 10% of the material was cleaved in the COOH-terminal EGFlike domain, at Lys-96. The fragment (BfIX-EGFd contained no detectable Gla (<0.02 mol/mol of protein), but 1.2 mol of Hya per mol of protein, and had an amino acid composition corresponding to residues 44-144 and residues 286-296 of BfIX (Table I). The recovery relative to BfIX-GlaEGFNc was 30-40%. The absorption coefficient (Ai%, at 280 nm) was determined to be 9.5 based on a calculated molecular weight of 12,390.
The Gla peptide (BfIX-Gla), which was eluted together with BfIX-GlaEGFNc from the ion exchange column (Fig. 5), was isolated by gel filtration on a Sephadex G-75 column (not shown). Sequence analysis of the isolated BfIX-Gla corresponded to residues 1-6 of intact BfIX (see above) and indi-

TABLE I Amino acid composition of BflX-GlaEGFNc, BflX-EGFNC, and BflX-Gla
The residue number in intact Bfl X is given below the name of each fragment. The composition according to the sequence is shown in parentheses. Thr"
Fluorescence emission spectra showed a significant quenching of tryptophan fluorescence in all proteins, except BfIX-EGFNc, on addition of saturating concentrations of Ca", whereas only a small blue shift was observed (Fig. 7). In addition, a small increase of the fluorescence of BfIX-Gla-EGFNc was seen at an intermediate Ca2+ concentration of 0.2 mM (Fig. 7 c ) .
On titration, monitored by the fluorescence emission, of BfIXap' and Gla-domainless BfIXaP' with Ca2+, there was an initial decrease of the intrinsic tryptophan fluorescence of about 5% with half-maximum quenching at approximately 25 p~ Ca2+ (Fig. 8). At higher Ca2+ concentrations, an additional fluoresence quenching, to approximately 15%, was observed in BfIXaP' (Fig. SA), due to binding of Ca2+ to low affinity sites (half-maximum at approximately 1 mM). This fluorescence quenching was not seen in Gla-domainless BfIXaP' (not shown). On addition of Mg"', no initial fluorescence quenching could be observed with either fragment (Fig. 9). At higher Mg"' concentrations, there was a 15% fluorescence quenching in BfIXaP' due to binding of Mg"' to sites that were half-saturated at 3-3.5 mM M e , whereas only a small (2-3%) quenching was observed in Gla-domainless BfIXaP' (Fig. 9).
Titration of BfIX-GlaEGFNc with Ca2+ resulted in an initial 10% increase of the intrinsic fluorescence due to binding of Ca2+ to sites that were half-saturated at approximately 40 p~ Ca2+ (Fig. 8B), i.e. with an affinity similar to that of BfIXaP' and Gla-domainless BfIXaP'. At higher Ca2+ concentrations, there was a 45-50% quenching that was half-saturated at about 0.8 mM Ca2+ (Fig. 8A), to be compared with the halfsaturated quenching at approximately 1 mM Ca2+ for BfIXab'. A similar pattern was observed on addition of M$+, although the binding affinity was lower (Fig. 9). The initial increase of the fluorescence was approximately 17%, with half-saturation To further characterize the ion binding properties of the factor IX molecule, the Gla residues in BfIX-GlaEGFNc were decarboxylated. After 16 h of heating, the fragment was intact as judged by sequence analysis, but retained only 13% of the Gla residues. On titration with Ca2+, there was a 6% initial increase in tryptophan fluorescence, due to binding of Ca2+ to sites that were half-saturated at approximately 60 ~L M (Fig.   lo), with no additional fluorescence change being observed at higher Ca2+ concentrations (up to 2.5 mM). In contrast, Mg2+ did not induce any significant change in the fluorescence emission of the fragment (Fig. 10).
On titration of BfIX-Gla with Ca2+, an initial plateau was observed, followed by a decrease of the intrinsic fluorescence to about 40% at >20 mM Ca2+ (Fig. 8 A ) . Half-saturation occurred at approximately 4 mM Ca", i.e. at a concentration 4 times higher than that observed for BfIXaP' and BfIX- GlaEGFNc. All fluorescence changes were reversible on addition of EDTA.
Metal Ion Dependence of Cleavage at Lys-43"During development of the procedure for isolation of BfI X-GlaEGFNc, it was noticed that the rate of cleavage, as well as the cleavage pattern, depended on the presence of Ca2+ in the buffer. Fig.  11A demonstrates that lysyl endopeptidase (0.2%, w/w) cleaves BfIXaP' completely within 2 min in the presence of EDTA, whereas no cleavage at all can be seen after 20 min in a Ca2+-containing buffer. The same protection against cleavage was achieved with a buffer containing M e (not shown). This Ca2+ and M e dependence was also observed when the enzyme cleaved BfIX-GlaEGFNc to produce BfIX-Gla and BfI X-EGFNc, indicating a similar metal ion-induced conformational transition in the fragment and in intact BfIXaP' (Fig. 11B).

DISCUSSION
Binding of Ca2+ to BfIX-GlaEGFNc, which has 2 Trp residues, results in changes of intrinsic protein fluorescence indicative of at least two binding processes, one with high affinity (half saturation at ~4 0 p~ Ca2+) and one with low affinity (half-saturation at -0.8 mM Ca2+). The high affinity Ca2+ binding is observed as an increase in fluorescence emission, whereas Ca2+ binding to the low affinity sites results in a fluorescence quenching. A similar initial increase in fluorescence also occurs upon binding of Ca2+ to the high affinity sites in the GlaEGF fragments from protein C and factor X (47, 48). As judged from the increase in the fluorescence emission, a high affinity Ca2+ binding site is also present in the decarboxylated form of BfIX-GlaEGFNc, and thus this site seems to be Gla-independent. On the other hand, no change in intrinsic tryptophan fluorescence was observed in decarboxylated BfIX-GlaEGFNc on addition of M e , indicating that the Gla-independent site does not bind M e . The Ca2+-induced alteration of fluorescence emission of the GlaEGF fragments from factors IX and X and protein C must be caused primarily by changes in the environment of Trp-42 in BfIX (45) and Trp-41 in factor X (49) and protein C (50), since this is the only conserved Trp residue in all three fragments. Moreover, in factor X it is the only Trp residue in the light chain. The initial increase in tryptophan fluorescence, that takes place on titration of intact BfIX-GlaEGFNc with Ca2+, probably does not solely depend on binding of Ca2+ to the Gla-independent site, since a similar increase was observed on titration with M e , which does not bind to this site. Hence, it is likely that part of the observed increase in tryptophan fluorescence of intact BfIX-GlaEGFNc on addition of Ca2+ is due to binding of Ca2+ to the Gla-dependent binding sites and part to the Gla-independent site. The Gla dependence of the low affinity Ca2+ binding sites was clearly demonstrated by the absence of fluorescence changes in the decarboxylated fragment. In this context, it is noteworthy that factor IX has two classes of M e binding sites (Fig. 9). However, the affinity for binding of Ca2+ is higher and, most importantly, there appears to be no Gla-independent binding of Mg+.
Binding of Ca2+ to Gla-domainless BfIXaP', which lacks Trp-42, is accompanied by a small quenching of intrinsic protein fluorescence, also described by Morita et al. (13,46). A similar quenching is observed on Ca2+ binding to Gladomainless bovine protein C (51). In contrast, Ca2+ binding to the Gla-domainless bovine factor X, which has no Trp residue in the light chain, results in no or a very small fluorescence quenching (49, 52, 53). These observations suggest that changes in the environment of Trp-72 in factor IX and Trp-84 in protein C are responsible for the fluorescence quenching that occurs in the Gla-domainless proteins, in spite of the fact that no fluorescence quenching can be seen on Ca2+ binding to B~IX-EGFNC and the corresponding fragment from bovine protein C (47). The latter finding would imply that quenching of the fluorescence from Trp-72 in factor IX and Trp-84 in protein C only occurs when these domains have an adjacent serine protease region. However, it cannot be excluded that the fluorescence quenching that accompanies Ca2+ binding to Gla-domainless BfIXaB' and protein C is caused by Trp residues in the heavy chain of the molecule.
It should be emphasized that both the increase and decrease of intrinsic protein fluorescence occur at the same Ca2+ concentration in BfIX-GlaEGFNc and intact BfIXap', indicating that the fragment has a native conformation. In contrast, much higher Ca2+ concentrations are needed to induce a fluorescence quenching in BfIX-Gla, which suggests that this region alone cannot attain a native conformation. It appears, therefore, that the EGF-like domains in factor IX provide a scaffold for the normal folding of the Gla region. Furthermore, a comparison with the factor X fragments suggests that only the NH2-terminal EGF-like domain is required for this purpose (48).
The large quenching of fluorescence on addition of high concentrations of Ca2+ to BfIX-GlaEGFNc indicates a marked change in the environment of Trp-42. This is reflected in the different accessibility of Lys-43 to cleavage by lysyl endopeptidase in metal ion-and EDTA-containing buffers (Fig. 11). This peptide bond is cleaved very rapidly in both BfIX-GlaEGFNc and BfIXap' in the absence of Ca2+ and M e , but is resistant in the presence of these ions. These ions thus induce a large conformational change not only in the Gla region but also in the connecting peptide. The latter peptide has a cluster of aromatic amino acids and is coded on a small exon in factors VII, IX, X, protein C, and prothrombin ( Fig.   12). It is noteworthy, that the crystal structure of prothrombin fragment 1 has revealed that in prothrombin the corresponding region forms an a-helix (11).
Certain patients with hemophilia B and normal antigen levels have circulating factor IX molecules with very low biological activity due to point mutations in the NH2-terminal EGF-like domain (3, 54,55). Some of these mutated amino acids have been implicated in the Gla-independent Ca2+ binding site of factor IX and protein C, i.e. Asp-47 and -49 and Hya-64 in factor IX and Hya-71 in protein C (56, 57). Recently, it was also demonstrated that a Gla-independent Ca2+ binding site in factors IX and X is located in the NH2-terminal EGF-like domain (20)(21)(22). These findings, which suggest an important role of the NH2-terminal EGF-like domain in the clotting activity of factor IX, have to be reconciled with the observations made by Lin et al. (35), who have exchanged the EGF-like domains between recombinant factors IX and X, expressed factor IX in mammalian tissue culture, and studied its activity in clotting assays. The evidence obtained so far suggests that the NH2-terminal EGF-like domain can be exchanged between the two clotting factors with only little loss of biological activity in factor IX. On the other hand, when the COOH-terminal EGF-like domain was exchanged, the activity was dramatically reduced (residual activity <5%). It is conceivable that the NH2-terminal EGF-like domain of factor X can function as a scaffold not only for the folding of the factor X Gla region, but also for the folding of the factor IX Gla region in the hybrid protein. Moreover, Ca2+ binding to these domains seems to confer a native conformation to the serine protease part and full biological activity. Our results are entirely consistent with the notion that the Gla region in factor IX is folded in a similar manner as the Gla region in prothrombin fragment 1 as recently determined by x-ray diffraction methods by Soriano-Garcia et al. (11).
Fluorescence emtssion spectra were recorded between 300 and 400 nm In the absence and presence 01 Ca2*. The excctatlon wavelength was 280 nm and the excitatm and emass1on bandwidths 2 and 8 nm, respectwely. All spectra were normalized to a vaiue lor the Spectra ~n the absence of C s + 01