Baculovirus-mediated expression of the epidermal growth factor-like modules of human factor IX fused to the factor XIIIa transamidation site in fibronectin. Evidence for a direct interaction between the NH2-terminal epidermal growth factor-like module of factor IXa beta and factor X.

Factor IX is a vitamin K-dependent procoagulant zymogen of a serine protease. In the presence of Ca2+ the active form of factor IX (factor IXa beta) forms a complex with factor VIIIa on suitable phospholipid surfaces such as aggregated platelets. This macromolecular complex rapidly activates factor X. We have previously provided data that suggest an interaction between the NH2-terminal epidermal growth factor (EGF)-like module of factor IXa beta and the substrate factor X. In an alternative approach to study this protein-protein interaction, we have expressed three recombinant baculovirus constructs encoding the EGF-like modules of human factor IX and a truncated form of fibronectin in a system based on the infection of insect cells (Spodoptera frugiperda 21). This strategy allows a simple one-step purification of the recombinant proteins on a gelatin-Sepharose column, followed by removal of the gelatin-binding part derived from fibronectin by proteolytic cleavage. The fusion proteins were isolated at yields of 20-50 micrograms/ml culture medium. The recombinant EGF-like modules contained 0.2-0.4 mol of erythro-beta-hydroxyaspartic acid/mol of protein, i.e. similar to the amount found in factor IX from human plasma, and appeared to be glycosylated at Ser-53. The NH2-terminal EGF-like module, which contained a transamidation acceptor site derived from fibronectin, was cross-linked by factor XIIIa in solution to intact and Gla-domainless factor X. There was no evidence of cross-linking to activated factor X or to factor X fragments containing only the gamma-carboxyglutamic acid module and the two EGF-like modules. The cross-linking results suggest a specific interaction between the NH2-terminal EGF-like module of factor IXa beta and the heavy chain of unactivated factor X. This interaction, albeit weak as judged by competition experiments, may be important for the targeting of factor X to the factor IXa beta-factor VIIIa complex on biological membranes and for the subsequent dissociation of factor Xa from the complex after activation.

Factor IX is a vitamin K-dependent procoagulant zymogen of a serine protease. In the presence of Ca2+ the active form of factor IX (factor IXM) forms a complex with factor VIIIa on suitable phospholipid surfaces such as aggregated platelets. This macromolecular complex rapidly activates factor X. We have previously provided data that suggest an interaction between the NH2-terminal epidermal growth factor (EGF)-like module of factor IXM and the substrate factor X. In an alternative approach to study this protein-protein interaction, we have expressed three recombinant baculovirus constructs encoding the EGFlike modules of human factor IX and a truncated form of fibronectin in a system based on the infection of insect cells (Spodoptera frugtprda 21 ). This strategy allows a simple one-step purification of the recombinant proteins on a gelatin-Sepharose column, followed by removal of the gelatin-binding part derived from fibronectin by proteolytic cleavage. The fusion proteins were isolated at yields of 20-50 pg/ml culture medium. The recombinant EGF-like modules contained 0.2-0.4 mol of erythro-&hydroxyaspartic acid/mol of protein, i.e. similar to the amount found in factor IX from human plasma, and appeared to be glycosylated at Ser-63. The NH2-terminal EGF-like module, which contained a transamidation acceptor site derived from fibronectin, was cross-linked by factor XIIIa in solution to intact and Gla-domainless factor X. There was no evidence of cross-linking to activated factor X or to factor X fragments containing only the y-carboxyglutamic acid module and the two EGF-like modules. The cross-linking results suggest a specific interaction between the NHz-terminal EGF-like module of factor IXM and the heavy chain of unactivated factor X. This interaction, albeit weak as judged by competition experiments, may be important for the targeting of factor X to the factor IXM-factor VIIIa complex on biological membranes and for the subsequent dissociation of factor Xa from the cQmplex after activation. *This investigation was supported by grants from the Swedish Medical Research Council ( roject 4487), Albert PHhlsson's Foundation, Kock's Foundation, dsterlund's Foundation, Knut and Alice Wallenberg's Foundation, and National Institutes of Health Grant HL 21644. 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 )I To whom correspondence should be addressed.
Modules' homologous to the epidermal growth factor (EGF)2 are found in many extracellular proteins including the vitamin K-dependent coagulation factors VII, IX, and X and the anticoagulant protein C (3, 4). Each of these proteins contains two EGF-like modules, the NH2 terminus of which has one or more postribosomal amino acid modifications. Erythro-#?-hydroxyaspartic acid, which has only been found in EGF-like modules, is formed by hydroxylation of an Asp residue. In contrast to factor X and protein C, which are fully hydroxylated, human factor IX is only partially hydroxylated (~0 . 3 mol/mol of protein) (5-7). O-Glycosidically linked disaccharide side chains with unique structures have been identified in the NH2-terminal EGF-like module of human factors VI1 and IX (8,9). The function of these amino acid modifications is unknown.
In the majority of the extracellular proteins, the functions of the EGF-like modules are unknown. However, in some proteins these modules have been implicated in protein-protein interactions (10-18). In addition, the NH2-terminal EGFlike module in factors IX and X binds one Ca2+ (19-21). Recently, we presented data suggesting that the NH2-terminal EGF-like module of factor IXa#? interacts directly with the factor 1 x 4 substrate, factor X (22). To study this interaction, we have now used baculovirus transfer vectors to express the EGF-like modules of human factor IX with or without the ahelical part of the Gla module. The modules were expressed as fusion proteins embedded in a truncated and deleted portion of fibronectin, GAP 1-5 (23). This approach offers two advantages compared with the expression of the factor IX modules alone. First, the gelatin-binding part of the truncated fibronectin, which can be removed by limited proteolysis, provides a simple purification procedure on a gelatin-sepharose column (24), and second the transamidation acceptor site(s) in fibronectin (25) can be used in cross-linking experiments employing factor XIIIa. Our results indicate an inter-action between the recombinant NHp-terminal EGF-like module of factor IXaP and the heavy chain of factor X but not of factor Xa.
Construction of Factor ZX-containing Gelutin-&i~ing Expression Pbmids-The following primers were synthesized in an Applied Biosystems 381 A DNA synthesizer to amplify the desired human factor IX cDNA sequences in a polymerase chain reaction (PCR) performed according to established procedures: GGCCATCCCCGT GGGAAAACACTGAAAGAACAACTG (sense for pGE-IXaN and SE-IXaNC, see below); GGCCATCCCCGTGGTTTTGGAAGCA GTATGTTGATG (sense for pGE-IXN); GGCCACGGGGATGGAT CTAATTCACAGTTCTTTCCTTC (antisense for pGE-IXaN and T G (antisense for pGE-IXaNC). Underlined sequences represent bases introduced at the 5' ends of the factor IX DNA to create a BstXI site for in-frame cloning into the gelatin-binding expression vector GE-lfpGEM4. This vector encodes the 32-amino acid preprosequence of rat fibronectin, the 18 NHz-terminal residues of the mature molecule as well as residues 243-572. It has a BstXI site within the codon for Pro-11 and is based on the GAP 1-5 construct from which residues 19-242, making up the 1st to the 5th type 1 homology region, had been deleted (23). The amplified DNAs were designated ZXaN, encoding residues 33-85 in mature human factor IX, ZXN, encoding residues 41-85, and IXaNC, encoding residues 33-127. These PCR products were precipitated, purified with standard techniques, and digested with BstXI. The fragments were then introduced into the BstXI site of GE-lfpGEM4 with T4 DNA ligase and subcloned in Escherichia coli (DH5a) to create the plasmids; pGE-IXaN, pGE-IXN, and pGE-IXaNC ( Fig. 1). Clones containing the insert in the correct orientation were identified by restriction enzyme analysis (EcoRI and BamHI) and shown not to contain mutations introduced by PCR by double-stranded DNA sequence andysis using Sequenase version 2 (U. S. Biochemical Corp.).
Construetion of Baculouirus Transfer Plasmids-The baculovirus transfer plasmid pAcIXaN was created by isolating the BamHIfragment of the pGE-IXaN construct by the Geneclean procedure (BIO 101, La Jolla, CA) according to the m~ufacturer's protocol. This fragment, encoding both the fibronectin mutant GAP 1-5 and the PCR-amplified human factor IX cDNA sequence, was then sub-pGE-1XN);GGCCACGGGGATGGTGCTGGTTCACAGGACTTC cloned into the baculovirus transfer vector pAcYMl (a gift from Dr. David Bishop, Institute of Virology, Oxford, UK) ( Fig. 1) (47). Clones containing the insert in the correct orientation were identified by restriction enzyme analysis (XhoI) and characterized as described above. The other two transfer plasmids, pAcIXN and pAcIXaNC, were created in the same way.
The baculovirus transfer plasmids were cotransfected with A. cafifornica nuclear polyhedrosis virus DNA into Sf21 cells with lipofectin (Life Technologies, Inc.) as described (48). The supernatant (transfection mixture) was decanted 4 days later, centrifuged, and stored at 4 "C until use. Sf21 cells were infected with dilutions of the transfection mixture as described by Summers and Smith (49) and the plates visually screened for recombinant occlusion-negative plaques after 3-4 days. Putative recombinant plaques were then plaque-purified three times to generate recombinant viruses free of contaminating wildtype virus (48,49). The isolated recombinant viruses, designated as the rVIXaN, rVIXN, and rVIXaNC virus, respectively, were titered by the addition of 0.01% Neutral Red as described (50). Protein F r~~~n and P~r~~a~n -S f 2 1 cells (9 X 10' cells/l@mm dish) were infected with recombinant virus at a multiplicity of infection of 3-6 in medium containing 10% low fihronectin serum (49). This serum was prepared by passing fetal calf serum through a gelatin-Sepharose 4B column (0.9 x 6 cm) before use (24). After 72 h the medium was collected and phenylmethylsulfonyl fluoride added to a final concentration of 2 mM. The conditioned medium was centrifuged at ambient temperature and the supernatant then loaded at 60 mlfh onto a gelatin-Sepharose 4B column (1.6 X 10 cm), equilibrated in 20 mM Tris-HC1, pH 7.4, containing 0.15 M NaCl.
The column was washed with Tris buffer containing 1 M NaCl and eluted with 3 M guanidine HCL The protein-containing fractions were pooled and dialyzed against 50 mM Tris-HC1, pH 7.5, containing 0.1 M NaCl. The proteins, designated according to the recombinant viruses as the IXaNfGAP 1-5, IXNfGAP 1-5, and IXaNCfGAP 1-5 proteins, were characterized with respect to NHz-terminal amino acid sequence, amino acid c o m~i t i o n , and imm~ochemical properties.

Treatment of Sf21 Cells mith P y r i d i~-Z , 4 -d i c a~~l o t e -T h e
inhibitor pyridine-2,4-dicarboxylate was solubilized in deionized water and either added to the cells immediately postinfection or after 1 day. The medium was then harvested after 72 and 48 h, respectively. In experiments where the inhibitor was added after 1 day, the cells were preincubated for 7 h with TCl00 medium containing the inhibitor in the indicated concentrations before the addition of fresh inhibitorcontaining medium.
Zsolution of EGF Fragment Containing the Factor XZZZa Transamidatwn Acceptor Site(st-To prevent cleavage at lysine residues during trypsin cleavage, the purified recombinant proteins were citramnyfated as described (20). After dialysis against 50 mM Tris-HC1, pH 7.5, containing 0.1 M NaCl and 5 mM EDTA, the sample (0.3-0.4 mg/ ml) was digested with trypsin (0.1%) at 37 "C. The rate of digestion was followed by SDS-PAGE (Fig. 2). In preparative experiments 2-7 mg of the citraconylated proteins were digested and the reaction terminated by the addition of diisopropyl fluorophosphate (final concentration of 2 mM). The lysine-blocking groups were removed (20) and the sample chromatographed on a gelatin-Sepharose column (0.9 X 8 cm) as described above. The recombinant proteins were then gel filtered on a Sephadex G-50 column equilibrated in 0.1 M NHhHCOs. The amino acid sequence and composition of each protein fraction were determined.
Cross-linking Reactions-Cross-linking experiments were performed in 50 mM Tris-HC1, pH 7.5, containing 0.1 M NaCl and 10 m M Ca". The incubation mixture (100 pl) contained the recombinant fusion protains, the EGF-containing fragments, or IlzfpGE-l at a concentration of =2 PM, human a-thrombin (5 nM), human recombinant factor XI11 (40 pgfml), and either m o n~s y l c a~v e r i n e (0.7 mM) or protein ligands as indicated. After i n~b a t i o n for 2 h at 37 "C, the reactions were terminated by the addition of EDTA to 20 mM and 35 pl of a buffer, containing 3% SDS, 30% sucrose, and 0.01% bromphenol blue in 0.1 M Tris-HC1, pH 8.8. The dansylated proteins were reduced with 4 pl of 2-mercaptoethanol followed by heating for 3 min at 95 'C before electrophoresis (51).
Other Methods-To test for 0-linked carbohydrates, the recombinant EGF-containing fragments were dialyzed against 0.1 M N&HC03. One or two nmol of each construct was evaporated to dryness and incubated in 30 pl of 0.1 M NaOH containing 0.1 M NalS03 for 24 h at 4 "C (52). The cysteic acid content was determined on a Beckman 6300 amino acid analyzer.
SDS-PAGE, amino acid analyses, and amino acid sequence determinations were performed as previously described (22). Gels containing dansylated proteins were visualized by using an ultraviolet light source. Western blot analysis of the proteins after SDS-PAGE was performed by transfer of the proteins to an Immobilon membrane and blocking of unoccupied sites with 5% defatted dry milk in 10 m M Tris-HC1, pH 8.0, containing 0.15 M NaCl and 0.05% Tween 20 (quenching buffer). The immobilized proteins were visualized by using either the monoclonal antibody M7 against the NH2-terminal EGFl i e module in human factor IX or an affinity-purified rabbit polyclonal anti-light chain bovine factor X antibody (5 pg/ml in quenching buffer), followed by an anti-mouse or an anti-rabbit immunoglobulin fraction conjugated to alkaline phosphatase (DAKO) and staining according to the manufacturer's instruction.
The effects of the recombinant EGF fragments on factor X activation by factor IXa@ were measured in a linked enzyme assay as described in the accompanying paper (53).

Expression and Characterization of Recombinant Proteins-
Construction of the recombinant baculovirus rVIXaN is outlined in Fig. 1. The other two recombinant viruses, rVIXN and rVIXaNC, were created in the same way. Sf21 cells were infected with the purified viruses for 72 h in the presence of leupeptin to minimize the internal cleavage a t Arg-259 that otherwise was present in 5-20% of the molecules (Fig. 3). The amount of fusion protein secreted into the culture medium at 72 h from a monolayer culture was estimated after purification on a gelatin-Sepharose column by amino acid analysis to be approximately 50 pg/ml with rVIXN, 40 pg/ml with rVIXaN, and 20 pg/ml with rVIXaNC. The recombinant fusion proteins were designated IXN/GAP 1-5, IXaN/GAP 1-5, and IXaNC/GAP 1-5. In all three proteins two NHz-terminal sequences were identified Thr-Glu-Thr-Gly-Lys-Ser, corresponding to residues -8 to -3 in rat fibronectin and Gln-Ala-Gln-Gln-Ile-Val, corresponding to residues 1-6. The relative amount of the first sequence, which probably reflects the beginning of the propeptide, varied between 10 and 70%. However, the amount of material without putative propeptide may have been underestimated due to cyclization of the NHzterminal Gln to pyroglutamic acid (54).
The trypsin-sensitive site at Arg-259 (51) was used to remove the gelatin-binding part of each recombinant protein (residues 260-572 in rat fibronectin) (Figs. 2 and 3). To protect Lys-43 (residues numbered as in factor IX), the proteins were reversibly citraconylated before tryptic cleavage. Thrombin was unable to cleave the fusion proteins at Arg-259, although this cleavage has been described for native fibronectin (55). The material eluted with guanidine HC1 from the gelatin-Sepharose column had a sequence corresponding to residues 260-265 in rat fibronectin (Thr-Ala-Ile-Tyr-Gln-Pro) and an amino acid composition corresponding to residues 260-572 (not shown). The flow-through fractions had an NHzterminal sequence corresponding to residues 1-6 in rat fibronectin (see above). The amount of propeptide-containing material was reduced to less than 10%. The amino acid compositions were in good agreement with fragments consisting of the first 12 amino acids of mature fibronectin, including the factor XIIIa acceptor site at Gln-3, followed by residues 33-85 of human factor IX in aEGFN, residues 41-85 in EGFN, residues 33-127 in aEGFNc, and a 26-amino acid residue of human factor IX was amplified by PCR using human factor IX cDNA in the pUC 8 vector as a template. The resulting purified PCR fragment was digested with BstXI and ligated into the BstXI site of the GE-l/pGEM4 vector described under "Experimental Pmedures." The coding sequence of truncated fibronectin lacking the first five type I modules (GAP 1-5) is hatched. The cloned pGE-IXaN construct was digested with BamHI and the insert ligated into the BarnHI site of the baculovirus transfer vector pAcYM1. The created plasmid designated pAcIXaN, thus contained the PCR-amplified human factor IX DNA linked to the DNA coding for the fibronectin mutant Gap 1-5 under the control of the viral polyhedrin promoter. This plasmid was then cotransfected with wild-type A. californica nuclear polyhedrosis virus DNA in S a l to allow the cloning of the recombinant virus rVIXaN. The sequences amplified by PCR were sequenced to insure that no mutations had been introduced.
peptide of fibronectin at the COOH-terminal end of each fragment ( Fig. 3; Table I). The proteins were found to contain 0.2-0.4 mol erythro-p-hydroxyaspartic acid/mol of protein.
The isolated EGFN fragment, subjected to NaOH/NazSOs treatment, was found to contain 1.2-1.4 mol of cysteic acid/ mol of protein. When chemically synthesized factor X-EGFN was subjected to the same treatment, a background of 0.2-0.4 mol of cysteic acid/mol of protein was obtained. The results indicate the presence of 0-linked carbohydrate side chain(s) in the recombinant factor IX fragment. Moreover, the serine residue in position 53 in the recombinant NH2-terminal EGFlike module, known to be glycosylated in human factor IX (8), was undetectable on sequencing, whereas the following two residues were found in the expected amount, suggesting that Ser-53 was modified. Before reduction all three fusion proteins bound a monoclonal antibody recognizing the NH2terminal EGF-like module of human factor IX in Western blotting experiments (Fig. 4). The antibody did not bind the reduced proteins. Attempts to visualize the isolated EGF module-containing fragments by Western blotting were unsuccessful as they did not bind to the membrane. Functioning factor XIIIa transamidation acceptor site(s) were identified in all three fusion proteins, as well as in the isolated EGFlike fragments by cross-linking with the fluorescent amine monodansylcadaverine (Fig. 5).

Effects of Recombinant EGF-like Fragments on Factor X
Actiuation-In previous papers, we have described the inhibitory effects of proteolytic fragments of bovine factor IX on factor X activation by factor IXa8 (22,53). The effects of the recombinant EGF fragments were investigated using the same systems in the absence of factor VIIIa and phospholipid. The aEGFNc fragment was the most potent inhibitor of factor Xa formation with a K, of =2 ~I M (Fig. 6). The fragments aEGFN and EGFN also inhibited factor Xa formation but more weakly (K,, =8 and 20 ~I M , respectively). No inhibition was found with intact fibronectin. The inhibition data were best fit to a substrate depletion model in which the fragments bind to the substrate, factor X, forming a complex that is not available for activation by factor IXa& The EGF-like fragments had similar effects on the activation of Gla-domainless BfX (not shown).

Cross-linking of Recombinant EGF-like Modules to Factor X-Cross-linking experiments with the
fluorescent amine monodansylcadaverine demonstrated that the recombinant proteins contained functioning factor XIIIa transamidation acceptor site(s) (Fig. 5). Small amounts of cross-linked complexes were also found. It is noteworthy that, although trypsin was efficient in cleaving the recombinant native constructs, cleavage progressed much more slowly after incorporation of the dansyl group (not shown). This may be due to additional acceptor sites close to Arg-259 ( Fig. 3) that have a sequence corresponding to that of reactive glutamines in 8-casein (57).
T o shed light on the nature of the putative interaction between the NH2-terminal EGF-like module of factor IXa8 and the substrate factor X (22, 53), we have attempted to cross-link the recombinant EGF-like modules to proteolytic fragments of factor X and visualize the products by Western blotting. The recombinant EGF-containing fragments alone did not bind to the Immobilon membrane, a feature also observed for EGF-like fragments from factor X. The immunoblot, using a monoclonal antibody against the NH,-terminal EGF-like module of factor IX, would therefore only be positive if this module had been linked to another molecule that binds to the membrane. Employing affinity-purified p o lyclonal antibodies against the light chain of bovine factor X, complexed and uncomplexed factor X (indicated by arrows in Figs. 7 and 9) were identified. The recombinant ECFN fragment was cross-linked to intact as well as to Gla-domainless factor X, both in the presence and absence of bovine serum  The composition according to the expected sequence is shown in parentheses.

EGFN
uEGFN UEGFNC ASP 9.5 (9) 10.1 (10) 17.9 (16) T h r 0.9 (1) 3.6 (4) 5.8 (6) Ser 9.6 (11) 9.8 (11) 11.0 (14) Glu 12.9 (12) 15.9 (15) 22.5 (21) Pro ND" (7) ND" (7) ND" (8) G~Y 7.9 (7) 8.6 (7) 11.0 (9) Ala 3.8 (4) 3.8 (4) 6.2 (7) CY8 ND" (6) ND" (6) ND" (12) Valb 4.8 (5) 4.7 (5) 7.5    albumin (Fig. 7). Cross-linking to human and bovine proteins gave identical results. The other recombinant EGF fragments, CYEGFN and CYEGFNC, were cross-linked in a similar manner (not shown). Excess amounts of GlaEGFNc from factor IX inhibited the reactions, whereas GlaEGFN from factor X did not (Fig. 8). There was no evidence of cross-linking to BBX-GlaEGFNc or to the factor X fragments, BfX-GlaEGF, and BfX-GlaEGFNc. Interestingly, no cross-linking was found to factor Xa or DEGR-inactivated factor Xa a t concentrations corresponding to those of the zymogen (Fig. 9). This could be due to a change in the environment of the cross-linking site rather than to an abolished interaction with the active enzyme. However, as the factor XIIIa transamidation acceptor site(s) have different positions relative to the interacting NHZterminal EGF-like module in the three recombinant fragments and they all cross-link in a similar manner to the zymogen, factor X, we think this is a remote possibility. The NHz-terminal EGF-like module did not cross-link to factor IXaB', protein Z, prothrombin fragment 1, or bovine serum albumin. However, the interaction was not entirely specific for factor X because a weak interaction was observed with prethrombin 1 (not shown). Recombinant I12/pGE-1, which is a fusion protein identical to those described here but with the II2 module of fibronectin introduced into GAP 1-5 instead  1-5 (lane 3 ) , and IXaNC/GAP 1-5 ( l a n e 4 ) fusion proteins were applied to a 10-15% SDS-PAGE gradient gel without prior reduction. In lane I , 2 pg of purified human factor IX was applied. After transfer onto an Immobilon membrane, the proteins were incubated with a monoclonal antibody against the NH2-terminal EGF-like module of human factor IX, followed by an alkaline phosphatase-conjugated rabbit anti-mouse I& fraction. The membranes were developed as described under "Experimental Procedures." The high molecular weight bands in lanes 2-4 correspond to multimers of the recornbinant proteins and the band with an apparent molecular weight of approrimately 60,000 in lane I to activated factor IXaS.
of the human EGF-like modules, did not cross-link to factor X. We conclude that the complex is not formed by interaction of the fibronectin-derived residues in the recombinant EGF fragments with factor X.

DISCUSSION
The baculovirus expression system has been used successfully to express several multidomain human proteins.
We have now used this system to express the EGF-like modules of human factor IX and the adjacent a-helical part of the Gla module as fusion proteins with a truncated fragment of mature fibronectin that lacks the five NH2-terminal type 1 homology units (GAP 1-5) (23). This procedure endows the recombinant proteins with three important properties derived from fibronectin: secretory processing by insect cells, a gelatin-binding domain, which allows simple one-step purification of the recombinant proteins (24) and then can be removed by limited tryptic cleavage (51), and transamidation acceptor site(s), which allow factor XIIIa-mediated cross-linking (25).
The occurrence of a sequence in the recombinant proteins beginning at residue -8 in fibronectin probably reflects incomplete proteolytic removal of the propeptide and that the insect cells process fibronectin in the way proposed based on the cDNA sequence, Le. removal of a signal peptide followed by removal of the propeptide at the canonical KSKR/Q sequence (Fig. 3). The sequences LCLG/T and LGTS/V have been proposed as likely sites for removal of the signal peptide (58). Our data, however, indicate that the site is SVRC/T. Signal peptidase cleavage carboxyl terminus of a Cys residue in the prepropeptide has also been described in other plasma proteins (59). We have three reasons to believe that the recombinant EGF-like modules were correctly folded: 1) the proteins were exocytosed, suggesting a native conformation; 2) a monoclonal antibody against the NH2-terminal EGF-like module in human factor IX bound to the unreduced recombinant proteins but not to the reduced form (thie antibody recognizes native plasma factor IX but not the reduced and after tryptic cleavage. In lone I the intact recombinant fusion protein IXN/GAP 1-5 is shown. The heterogeneity of EGFN is probably due to the carbohydrate side chain(s) and the weaker bands with higher molecular weights presumably correspond to multimers of the fragment. The samples were reduced before electrophoresis. The fluorescent band at the bottom of the gel represents unincorporated monodansylcadaverine. amount similar to that found in the native human factor IX molecule. As the Asp @-hydroxylase requires a correctly folded EGF-like module as substrate, this implies a native conformation (60). Pyridine-2,4-dicarboxylate, an inhibitor of the Asp/Asn @-hydroxylase in human cells (56). inhibited the hydroxylation apparently without affecting the expression level. Recently, Monkovic and colleagues (61) purified an Asp/Asn @-hydroxylase from insect cells that had a cofactor requirement and substrate specificity similar to that of the human enzyme. In addition, the recombinant NH2-terminal EGF-like module seemed to contain an 0-linked carbohydrate moiety at Ser-53, a position where Iwanaga and co-workers (8) have found a disaccharide side chain with a unique structure.
Competition studies suggest that the NH2-terminal EGFlike module of factor IXa@ interacts directly with the substrate, factor X, either through an EGF-EGF interaction or through an interaction between the EGF-like module and the serine protease module of factor X (22, 53). We have now used the transamidation acceptor site(s) in the recombinant proteins in conjunction with factor XIIIa to further investi- 1 mg/ml polyethylene glycol 6OOO. Increasing concentrations of the fragments were added and the initial rate of factor Xa formation measured through the hydrolysis of the chromoEenic substrate S-2765. The data have been expresaed as the percent of control activity, which corresponds to the ratio of v in the presence of I to v in the absence of I times 100. The solid lines were drawn using the substrate depletion model described in the accompanying paper (53). The dashed line representa no effect of inhibitor.  FIG. 8. Effects of GlaEGF fragments from factors IX and X on cross-linking of recombinant E G F N to factor X. Recombinant EGFN-fragment was cross-linked to bovine factor X as described in the legend to Fig. 7, both in the absence (lones I and 3 ) and presence (lanes 2 and 4 ) of GlaEGF fragments. The factor X concentration was 8 p~ and that of BfIX-GlaEGFNc ( l a n e 2) and BfX-GlaEGFN ( l a n e 4 ) 25 pM, respectively.  l a n e 3), and the factor X concentrations were 5 p~ ( l a n e 4 ) and 30 p M ( l a n e 5).
the formation of a heterodimer between prothrombin and factor X (62). However, the EGFN and aEGFN fragments had almost identical inhibitory effects on factor X activation, suggesting that the Gla module in factor IXaS is not required for the EGF module to interact with factor X. The NH2terminal EGF-like module was cross-linked to both intact and Gla-domainless factor X. On the other hand, no binding to the GlaEGF fragments of factor X was seen, suggesting that the observed inhibitory effects are due to a direct interaction between the EGF-like module of factor IXaS and the serine protease module of factor X. This notion gains further support from the inability of the light chain modules of factor X to inhibit cross-linking of the factor IX module to intact factor X. Accordingly, the possibility that the EGF-like module interacts with the corresponding module in the GlaEGF fragments, but that these fragments lack the lysine residue to which cross-linking occurs, seems remote. The interaction is not completely specific for factor X, as there is also some cross-linking to prethrombin 1. A low EGF module-associated specificity has also been observed in factor X, e.g. a chimeric factor X molecule with the Gla and NH2-terminal EGF-like modules exchanged for those in factor IX has only a moderately reduced biological activity (16). The observation that the recombinant NH2-terminal EGF-like module did not bind to factor Xa further supports the notion that the interaction is with the serine protease module, as this is the only part undergoing a major conformational change upon activation (37).
It is possible that the interaction between the EGFN module of factor IXaS and a region in the serine protease module in factor X enhances the rate of factor X activation. The inter-action, albeit weak as judged by competition experiments, appears to be specific for the zymogen form of factor X. This may be important for the targeting of factor X to the active site of factor IXaS in complex with factor VIIIa on biological membranes, i.e. the EGF-like modules of factor IXaS can be regarded as part of the cofactor. The apparent lack of affinity of the EGF-like module for the serine protease module once factor X is activated suggests that this interaction is physiologically relevant. It is noteworthy that there is a precedence for a direct interaction between ECF-like modules and a serine protease module. In thrombomodulin two EGF-like modules bind thrombin with high affinity and even change the substrate specificity of the enzyme (10, 11).