Structural determinants of the factor IX molecule mediating interaction with the endothelial cell binding site are distinct from those involved in phospholipid binding.

Previous studies have indicated that Factor IX/IXa interacts in a specific and high affinity manner with a binding site on the endothelial cell surface. In this study, the contributions of the gamma-carboxyglutamic acid-containing (GLA) and growth factor domains to the finding of Factor IX to the endothelium were assessed. While GLA-containing peptides from Factors IX, X, and prothrombin were inhibitors of 125I-Factor IX-endothelial cell binding, the GLA peptide from Factor IX was about 250-800-fold more effective than those from prothrombin and Factor X, respectively. In contrast to its relative efficacy as an inhibitor of Factor IX-cell surface interaction, the Factor IX-GLA peptide neither bound to lipid vesicles nor inhibited Factor IX-lipid interaction. A synthetic peptide comprising the entire first epidermal growth factor (EGF) exon was also an inhibitor of 125I-Factor IX-endothelial cell binding, although it did not interact with lipid vesicles. Experiments with synthetic peptides comprising each of the three loops of the first EGF domain or the entire first EGF region with specific substitutions indicated the importance of determinants in both the first and probably third loops for Factor IX-endothelial interaction. In contrast, the second loop of the first EGF domain and the first loop of the second EGF exon are probably not involved in Factor IX-endothelial interaction based on their inability to block 125I-Factor IX binding to cells. These results indicate that determinants in both the GLA and the first EGF domain contribute to the specific binding of Factor IX to the endothelial cell surface and that structural requirements for Factor IX-cell surface interaction are distinct from those for Factor IX binding to lipids.

Previous studies have indicated that Factor IXflXa interacts in a specific and high affinity manner with a binding site on the endothelial cell surface. In this study, the contributions of the y-carboxyglutamic acidcontaining (GLA) and growth factor domains to the finding of Factor IX to the endothelium were assessed. While GLA-containing peptides from Factors IX, X, and prothrombin were inhibitors of '"1-Factor IXendothelial cell binding, the GLA peptide from Factor IX was about 250-800-fold more effective than those from prothrombin and Factor X, respectively. In contrast to its relative efficacy as an inhibitor of Factor IX-cell surface interaction, the Factor IX-GLA peptide neither bound to lipid vesicles nor inhibited Factor IXlipid interaction. A synthetic peptide comprising the entire first epidermal growth factor (EGF) exon was also an inhibitor of '261-Factor IX-endothelial cell binding, although it did not interact with lipid vesicles. Experiments with synthetic peptides comprising each of the three loops of the first EGF domain or the entire first EGF region with specific substitutions indicated the importance of determinants in both the first and probably third loops for Factor IX-endothelial interaction. In contrast, the second loop of the first EGF domain and the first loop of the second EGF exon are probably not involved in Factor IX-endothelial interaction based on their inability to block '"1-Factor IX binding to cells. These results indicate that determinants in both the GLA and the first EGF domain contribute to the specific binding of Factor IX to the endothelial cell surface and that structural requirements for Factor IX-cell surface interaction are distinct from those for Factor IX binding to lipids.
Factor IX/IXa binds to the endothelial cell surface with high affinity (apparent Kd = 2 nM) and specificity, as other * This work was supported by grants from the United States Public Health Service (HL 34625, HL 41935, and HL 28433), the Council for Tobacco Research (CTR 1971), and Hoffmann-La Roche. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
' Recipient of a Genentech-American Heart Association Established Investigator Award during tenure of which this work was done. vitamin K-dependent coagulation factors (such as Factor X and prothrombin) do not compete (1-3). This contrasts with the lower affinity interaction of Factor IX with phospholipids and the ability of one vitamin K-dependent coagulation factor to displace another from the lipid surface (4-6). In this context, cross-linking studies have implicated the involvement of a cell surface protein, M, = 140,000, in mediating, at least in part, the binding of Factor IX to endothelium (7).
These considerations have led us to examine domains of the Factor IX molecule involved in recognizing the endothelial binding site (8,9) and to compare their role i n Factor IX-cell surface interaction with their effects on Factor IX binding to lipid vesicles. The results indicate that both the y-carboxyglutamic acid (GLA)'-containing and first epidermal growth factor (EGF) domains are involved in Factor IX-endothelial interaction and highlight the differences between the structural requirements for the binding of Factor IX to cellular surfaces and synthetic phospholipids.

Preparation of Coagulation Proteins and GLA Peptides-Bovine
Factors IX (230 units/mg) and X (100 units/mg) and human prothrombin fragment 1 were purified to homogeneity by previously described methods (10-12). Factor IX was radiolabeled by the lactoperoxidase method (13) using Enzymobeads (Bio-Rad) and lZ5I-Factor IX isolated as described previously (14). The specific radioactivity of these preparations was 4.6 X 10' cpm/ng over 15 iodinations, and the radioactivity profile of reduced 1251-Fa~tor IX on sodium dodecyl sulfate-PAGE showed a single peak at mass of -60 kDa. Factor IX modified by cleavage of the GLA domain, GLA-domainless Factor IX, was prepared by limited proteolysis with lysyl endopeptidase as described by Morita et al. (15,16).
GLA peptides were prepared from Factors IX, X, and prothrombin fragment 1 by incubating with Nu-p-tosyl-L-lysine chloromethyl ketone-treated a-chymotrypsin (Sigma) at an enzyme-to-substrate ratio of 1:200 in the absence of calcium followed by chromatography on QAE-Sephadex to separate the peptides from the other components of the reaction mixtures, as described previously (15)(16)(17)(18)(19). In the case of Factor X, this produces a peptide comprising residues 1-44. For human prothrombin fragment 1, a mixture of two peptides (by analogy with bovine prothrombin fragment 1 (20)), one including residues 1-41 and the other residues 1-44, could be produced. By using the results of amino acid analysis and protein determination (21), as described by Pollock et al. (20), the latter two peptides were identified. Human prothrombin fragment comprising residues 1-44 accounted The abbreviations used are: GLA, y-carboxyglutamic acid EGF, epidermal growth factor; HPLC, high performance liquid chromatography.

20283
for 40% of the material present in fractions pooled from the QAE-Sephadex column, the remainder being the peptide derived from residues 1-41. A similar analysis of material eluted in peak fractions from the Factor IX-chymotrypsin cleavage mixture indicated that the Factor IX-GLA peptide consisted predominantly of the fragment corresponding to residues 1-42.
Preparation and Characterization of Synthetic Peptides-The human Factor IX-EGF domain (residues 45-87) (22) was synthesized manually by the solid-phase method (24). Details of the synthesis, refolding, and characterization of the latter peptide and related peptide derivatives containing amino acid substitutions have been reported (25). These peptides have been found to be appropriately refolded and have recently been shown to be a suitable substrate to assess the /3-hydroxylation of Asp-64 by a 2-oxoglutarate-dependent dioxygenase (25).
Synthetic peptides comprising each of the loops of the first Factor IX-EGF exon (corresponding to residues 45-64, 55-73, and 72-85 from the bovine molecule) (23) and the first loop of the second Factor IX-EGF exon (residues 85-101) were prepared by solid-state methodology (manual method) (24). Peptides were purified to homogeneity by HPLC and characterized by amino acid analysis, and the molecular weight was confirmed by Fast Atom Bombardment (FAB/mass spectroscopy). For cyclization, linear peptides (100 pg/ml) were dissolved in 8 M urea containing K,Fe(CN)6 (0.18 mM). The solution was stirred at room temperature for 260 h, and cyclization was monitored by analytical HPLC. Then the solution was acidified with 40% aqueous trifluoroacetic acid (final pH = 2.5) and purified on a Waters Bondapak column (1.9 X 30 cm). Fractions containing the desired product were cooled, evaporated, and lyophilized. The product was shown to be homogeneous by analytical HPLC and gave a negative Ellman test (for free sulfiydryls).
Human and murine EGF were purchased from Calbiochem and Sigma, respectively. Cell Culture and Binding Assays-Bovine aortic endothelial cells were grown from aortae of newborn calves as described (26) and were used from passages 3-26. Experiments were carried out within 24 h after the cells achieved confluence using 0.32-cm' wells. At confluence, there were 1.0-1.5 x lo6 cells/cm2. The procedure for carrying out Factor IX binding assays has been described previously (2, 14). Previous studies from our (2)' and other laboratories (3) have shown that human and bovine Factor IX interact with endothelium in a very similar manner. Thus, use of peptides with sequences from the human and bovine molecules should not introduce an additional variable with respect to cell surface interaction. Data showing inhibition of 1251-Factor IX binding to endothelium by peptides was fit to the equation (

RESULTS
To investigate the role of Factor IX-GLA domain in the binding of this coagulation factor to the endothelial surface, peptides comprising GLA domains were prepared from Factors IX, X, and prothrombin by limited proteolysis. When these peptides were added to mixtures containing '251-Fa~tor IX and cultured endothelial cells, only the GLA peptide from Factor IX was highly effective as an inhibitor of binding (Fig.  M). In an attempt to determine if this binding and inhibition represented equilibrium measurements, inhibition of Factor IX binding at three concentrations of '251-Fa~tor IX was employed. Although the relationship fits closely to a model of competitive inhibition (27) with Ki = 0.06 p~ (Fig. 1A, inset), this value is viewed as approximate. The GLA peptides from Factor X and prothrombin were also inhibitors of '251-Fa~tor IX binding to endothelium. However, they were less effective ( Fig. 1, B and C ) . While the data employing the Factor X and prothrombin-GLA peptides did not extend to full inhibition, the partial inhibition was highly reproducible, suggesting that the observed inhibition was of significance. When analyzed by equilibrium binding inhibition, the data suggested competitive inhibition with Ki values of =50 p~ and ~1 5 p~, for the Factor X and prothrombin-GLA peptides, respectively (Fig.   1, B and C). Consistent with an important role for the GLA domain in mediating the interaction of Factor IX with endothelium, GLA-domainless Factor IX neither bound to endothelium nor competed with '251-Fa~tor IX for the cell surface sites up to a concentration of =1 p~ (data not shown).
Our previous studies suggested a role for the EGF domain in Factor IX-endothelial interaction (8,9). Other studies have suggested the involvement of the EGF domain in the interaction of other proteins, such as urokinase, with their cellular receptors (28). We therefore performed experiments with synthetic peptides spanning the entire first EGF exon of Factor IX (residues 45-87). This peptide appeared to inhibit binding of '251-Fa~tor IX to endothelium, and the data obtained ( Fig. 2 A )  The preliminary findings with the entire EGF domain led us to further examine portions of this structure. Two approaches were employed for these studies: 1) experiments with synthetic peptides comprising each of the loops of the first EGF domain; and 2) experiments with synthetic peptides comprising the whole first EGF exon substituted in specific areas. For the first approach, the effect of cyclized peptides corresponding to the first, second, and third loops of the first EGF domain on lZ5I-Factor IX-endothelial interaction was examined. The peptide corresponding to the first loop was the best inhibitor ( Fig. 2B and Table I,  Our other approach for assessing determinants of the first EGF domain mediating the interaction of Factor IX with endothelium led us to use peptides spanning the entire first EGF domain with specific substitutions in the first and third loops (Table I, lines 5-10). The latter substitutions were selected since these regions of EGF may be involved in its binding to the EGF re~eptor.~ A peptide comprising the first EGF domain with substitution of residues from the sequence of murine EGF for those in Factor IX in the first loop (see Table I  of the first EGF exon with those from human EGF or the laminin binding domain also rendered the peptide inactive as an inhibitor of Factor IX binding to endothelium (lines 8 and 9, respectively). If only the Asn a t position 81 was replaced by an Arg (this Asn is highly conserved in native EGF), the peptide retained its ability to inhibit Factor IX binding (line 10 and Fig. 2C). The latter peptide's ability to inhibit lZ5I-Factor IX-endothelial binding fit to a model of competitive inhibition (27) with Ki = 12 FM. The specificity of the peptides derived from the EGF domain of Factor IX for inhibition of '251-Fa~tor IX-endothelial binding was further confirmed by the finding that neither murine nor human EGF had any measurable effect on Factor IX-cell surface binding (lines 11 and 12).

DISCUSSION
Studies in our and other laboratories (1-3) have defined the presence of a selective binding site for Factor IX on endothelium. These observations have led to efforts to define the nature of the determinants mediating the interaction of Factor IX with the endothelial cell surface and how they may differ from those structures involved in Factor IX-phospholipid interaction. The experiments presented here demonstrate that the interaction of Factor IX with its binding site EGF domains Binding studies to assess inhibition of Iz5I-Factor IX binding by the peptides were carried out by incubating confluent monolayers of endothelium at 4 "C for 2 h with 'z51-Fa~tor IX at two or three different concentrations (0.13, 0.32, and 0.87 nM) in the presence of a range of peptide concentrations (0.5 to 250 p~) .
Cultures were then washed, and bound radioactivity was eluted with EDTA-containing buffer. This protocol for radioligand binding assays followed that referred to in the text. Each experiment was reDeated at least four times.

+ --
Peptides correspond to cyclized loops of the Factor IX EGF domain (lines 1-4) or peptides with/without substitutions from the full first exon of the EGF domain (lines 5-10). Lines 1-4, peptides 1-4 correspond to loops from the bovine Factor IX (22) EGF domain as follows: peptide 1 (residues 45-64, loop 1, disulfide formed between Cys-51 and -59 to close the loop, Cys at 56 (underlined) replaced by Ala); peptide 2 (residues 55-73, loop 2, disulfide formed between Cys at 56 and 71, Cys at 73 (replaced) replaced by Ala); peptide 3 (residues 72-85, loop 3, disulfide formed between Cys at 73 and 82); peptide 4 (residues 85-101, loop 1 of second EGF exon, disulfide formed between Cys at 88 and 99, Cys at 95 (underlined) replaced by Ala). Lines 5-10, peptide 5 is composed of residues 45-87 from human Factor IX (23) and the reformed disulfide bonds link Cys at residues 41 to 62,56 to 71, and 73 to 82 to form the three loops. The substituted amino acids in the other peptides (peptides 2-6) are underlined and derived from human/murine EGF (29), or the laminin binding domain (30), as follows (amino acid numbers in parentheses refer to location of that amino acid in murine/human EGF or Factor IX, and Cys residues linked by reformed disulfides are numbered based on arbitrarily assigning the number 45 to the first residue of the EGF domain or chimeric peptides): peptide 6, chimera of murine EGF (residues 6-18, underlined and constituting loop 1) and human Factor IX EGF domain (residues 45-50 and 61-87) with reformed disulfides linking Cys residues at 50 to 64,58 to 73, and 75 to 84; peptide 7, chimera of human EGF (residues 1-16, underlined and constituting loop 1) and human Factor IX (residues 59 to 87) with reformed disulfides linking Cys residues at 50 to 64; peptide 8, chimera of human EGF (residues 34-41, underlined and comprising loop 3) and human Factor IX (residues 45-73 and 82-87); peptide 9, a chimera of the laminin binding domain (from the laminin B1 chain, Tyr-Ile-Gly-Ser-Arg, underlined and substituted into the loop 3) and human Factor IX (residues 45-76 and 82-87); peptide 10 is the same as peptide 1 except that Asn (residue 81) has been substituted by Arg.
* If a peptide inhibited 1251-Factor IX binding to endothelium, a (+) follows that entry and the actual data are shown in the figure which is stated in the brackets. If a peptide had no effect on 'z51-Fa~tor IX binding, then a (-1 follows that entry. The limit of detection in the binding assay was less than 10% displacement. on endothelium involves structural features present in the GLA and first EGF domain of the molecule. Our data, however, do not rule out that other regions of the molecule may also be involved. The GLA domain has an important role in mediating the interaction of Factor IX with endothelium. In our preliminary report (9), we showed that the GLA-peptide derived from Factor IX was an inhibitor of '251-Fa~tor IX binding to endothelium. These data could be confirmed by Derian et al. (3), who noted an even higher inhibitory potency of the GLA peptide. Furthermore, the latter work demonstrated that recombinant decarboxy-Factor IX did not bind to endothelium. In the present study, the GLA peptide derived from Factor IX inhibits the binding of lZ5I-Factor IX to endothelium considerably more effectively than similar peptides derived from Factor X or prothrombin (Fig. 1). Although initially we reported in a pilot study that Factor IX modified by limited proteolysis to cleave the GLA domain retained some (but considerably reduced) binding activity (8), this was found to be due to a small amount of contaminating native Factor IX, and our results now agree with those recently reported by Derian et al. (3), i.e. Factor IX lacking the GLA domain does not bind to endothelium.
The GLA domain is traditionally thought of as mediating Ca2+-dependent interaction of vitamin K-dependent coagulation proteins with phospholipid surfaces (4-6). The GLA peptide from Factor IX (residues 1-42) employed in our study did not bind to phospholipids and did not inhibit the interaction of Factor IX with phospholipid (32). Lack of phospholipid binding by Factor IX GLA peptide from which residues 43-45 were removed supports the observations of Pollock et al. (20) that GLA peptides from bovine prothrombin spanning residues 1-42 do not have phospholipid binding properties. The ability of our GLA peptide (residues 1-42) to block Factor IX-endothelial interaction (although it did not block Factor IX-phospholipid interaction (32)) with greater potency than the GLA peptides from Factor X and prothrombin (both of which do bind to phospholipid (32)) emphasizes that different structural features in the GLA domain are important in mediating interaction of Factor IX with the cellular binding site versus a lipid surface.
The first EGF domain of Factor IX also appears to contribute to the interaction of this coagulation factor with the cell surface. The full length EGF peptide failed to interact with synthetic phospholipids and did not affect the binding of Factor IX to lipids (32). The results in Fig. 2 and Table I suggest that amino acid residues, especially in the first and possibly in the third loops of the first EGF exon, are involved in binding of Factor IX to endothelium. However, Derian et al. (3) found that recombinant decarboxy-Factor IX did not bind to endothelium. On the one hand, this result would appear to contradict our findings, that the first EGF domain or portions of this domain are capable of inhibiting Factor IX-endothelial cell binding, since these portions of the molecule should be intact in the recombinant protein. Nevertheless, lack of inhibition by the recombinant protein might arise from any of several factors, such as interference by the uncarboxylated GLA region, or by its effect on secondary/ tertiary structure. A requirement for interaction of two or more domains of Factor IX (such as the EGF and GLA domains) for the molecule to bind to the cell surface might also underlie the failure of GLA-domainless Factor IX to bind to endothelium. In the case of GLA-domainless Factor IX, the modified molecule has an additional internal cleavage (residue 142/143) (16), and this may affect recognition of the cellular binding site. Further studies will be needed to determine the basis for failure of recombinant Factor IX and GLAdomainless Factor IX to inhibit binding of Factor IX to endothelium.
In relation to the high affinity binding constant of Factor IX for the endothelium (Kd = 2 nM), the relatively high concentrations of Factor IX GLA-peptide and synthetic peptides necessary to effect inhibition of lZ5I-Factor IX binding to the endothelium in all likelihood reflects the fact that Factor IX interaction with the endothelial cell surface involves the cooperative effects of several determinants. In addition, these peptides probably exist in solution as random conformers in equilibrium with a conformer that assumes the native conformation of that segment of the Factor IX molecule (31). Presumably, it is this "native" conformer that associates with the cell surface Factor IX binding site to result in the inhibition of lZ5I-Factor IX binding, and, accordingly, this interaction may be very weak in the absence of other determinants that contribute to the overall energy of binding. Overall, the studies reported here are consistent with a model in which binding of Factor IX to endothelium involves contributions by several domains of the molecule. Because of this, it is difficult to localize with confidence the binding region of the Factor IX molecule, and there are probably multiple, complex structural features involving more than one loop which enable Factor IX to recognize the endothelial binding site. In addition, our results emphasize differences in determinants mediating interaction of Factor IX with lipids versus interaction with the endothelial cell binding site.