The Connecting Segment between Both Epidermal Growth Factor-like Domains in Blood Coagulation Factor IX Contributes to Stimulation by Factor VIIIa and Its Isolated A2 Domain*

The light chain of activated factor IX comprises multiple interactions between both epidermal growth factor-like domains that contribute to enzymatic activity and binding of factor IXa to its cofactor factor VIIIa. To investigate the association between factor IXa-specific properties and surface-exposed structure elements, chimeras were constructed in which the interconnection between the modules Leu84-Thr87 and the factor IX-specific loop Asn89-Lys91 were exchanged for corresponding regions of factor X and factor VII. In absence of factor VIIIa, all chimeras displayed normal enzymatic activity. In the presence of factor VIIIa, replacement of loop Asn89-Lys91 resulted in a minor reduction in factor IXa activity. However, chimeras with substitutions or insertions in the spacer between the epidermal growth factor-like domains showed a major defect in response to factor VIIIa. Of these chimeras, some displayed a normal response to isolated factor VIII A2 domain as a cofactor in factor X activation. Surprisingly, chimeras containing elongated inter-domain spacers from factor X or VII displayed reduced response to both complete factor VIIIa and the isolated A2 domain. Moreover, these chimeras still displayed effective association with immobilized A2 domain as assessed by surface plasmon resonance. We conclude that both sequence and length of the junction Leu84-Thr87 between both epidermal growth factor-like domains contribute to the enhancement of factor IXa enzymatic activity that occurs upon assembly with factor VIIIa.

The light chain of activated factor IX comprises multiple interactions between both epidermal growth factor-like domains that contribute to enzymatic activity and binding of factor IXa to its cofactor factor VIIIa. To investigate the association between factor IXa-specific properties and surface-exposed structure elements, chimeras were constructed in which the interconnection between the modules Leu 84 -Thr 87 and the factor IX-specific loop Asn 89 -Lys 91 were exchanged for corresponding regions of factor X and factor VII. In absence of factor VIIIa, all chimeras displayed normal enzymatic activity. In the presence of factor VIIIa, replacement of loop Asn 89 -Lys 91 resulted in a minor reduction in factor IXa activity. However, chimeras with substitutions or insertions in the spacer between the epidermal growth factor-like domains showed a major defect in response to factor VIIIa. Of these chimeras, some displayed a normal response to isolated factor VIII A2 domain as a cofactor in factor X activation. Surprisingly, chimeras containing elongated inter-domain spacers from factor X or VII displayed reduced response to both complete factor VIIIa and the isolated A2 domain. Moreover, these chimeras still displayed effective association with immobilized A2 domain as assessed by surface plasmon resonance. We conclude that both sequence and length of the junction Leu 84 -Thr 87 between both epidermal growth factor-like domains contribute to the enhancement of factor IXa enzymatic activity that occurs upon assembly with factor VIIIa.
Factor IX (FIX) 1 is a vitamin K-dependent serine protease precursor that participates in the process of blood coagulation (1,2). Factor IX is converted into an active serine protease upon cleavage by either factor XIa (FXIa) or the factor VIIa (FVIIa)tissue factor complex (3,4). Proteolysis occurs at restricted cleavage sites and results in the formation of the two-chain enzyme factor IXa (FIXa). FIXa is composed of a light chain that contains an N-terminal Gla domain followed by a short hydrophobic region and two epidermal growth factor (EGF)like domains (5). The heavy chain comprises the protease domain with the catalytic center. FIXa shares its typical domain structure with related serine proteases of the blood coagulation cascade, including FVIIa, factor Xa (FXa), and activated protein C (6,7). All of these enzymes contain two EGF-like domains that display considerable sequence homology with the EGF-like modules in factor IXa (8), indicating that these domains are similar in structure and function. Indeed, FIXa, FXa, and FVIIa are similar in that calcium binding to the first EGF-like module is essential for conserving the linkage between its N-terminal half and the N-terminal Gla domain in these proteins (for review, see Ref. 9). However, with respect to the relative orientations of the two EGF-like modules, the three-dimensional crystal structures reveal that this is different in each of these proteases. The connection between these modules is flexible within the structures of FVIIa and FXa (6, 10 -13), whereas in FIXa the relative orientation of both domains is much more restricted because of a variety of interdomain contacts (14). These include a salt bridge between residues Glu 78 in the first EGF-like domain and Arg 94 in the second EGF-like domain (15) and hydrophobic interaction between the regions Phe 75 -Phe 77 and Lys 106 -Val 108 at the interface between both domains (16). The salt bridge is crucial for interaction of FIXa with its cofactor, factor VIIIa (FVIIIa) (15), whereas the hydrophobic contact contributes to FIXa enzymatic activity (16).
In the present study, we focused on other FIXa-specific elements that are exposed at the interface between both EGF-like domains, including the sequence Leu 84 -Thr 87 that connects both domains and the loop Asn 89 -Lys 91 at the N-terminal part of the second EGF-like domain. To this end, FIX chimeras were constructed in which these regions were exchanged for the corresponding residues of FX and FVII. The spacer sequence between both EGF-like modules in FIXa, FXa, and FVIIa varies not only in sequence but also in length (4, 6 and 8 residues, respectively). To evaluate whether the length of the connecting segment or the particular amino acid residues therein are important for FIXa function, two additional FIX/FX chimeras were constructed with partial replacements. In one chimera, the FIX residues were replaced by the corresponding number of FX residues, whereas in the second variant the two extra residues of FX were inserted immediately after the FIX spacer.
Activated FIX chimeras were characterized with particular reference to enzymatic activity and response to FVIIIa. Our results demonstrate that in particular the junction Leu 84 -Thr 87 between the two EGF-like modules and to some extent also the loop Asn 89 -Lys 91 contribute to respond to the cofactor FVIIIa. For chimeras comprising the complete spacer sequence of FX or FVII, impaired sensitivity to intact FVIIIa was accompanied by a reduced stimulation by the isolated FVIII A2 domain. Our present data support a model in which the various structural elements that comprise the interface between the EGF-like domains serve a variety of factor IXa functions, including the FVIIIa-induced enhancement of enzymatic activity.
Recombinant FIX-Plasmids encoding wt-FIX, FIX-R333Q, and FIX-E78K have been described previously (15,17,21). The wt-FIX plasmid has been used for the construction of the plasmids encoding chimeras FIX 85-87 /FX FTR , FIX 85-87 /FX ϩKL , FIX 85-87 /FX FTRKL , FIX 84 -87 /FVII, FIX 89 -91 /FX, and FIX 89 -91 /FVII chimeras (see also Fig. 1). Chimeric FIX DNA fragments were obtained by the overlap extension polymerase chain reaction mutagenesis method (18). Sequence analysis was performed to verify the sequence of the constructs. Transfection of FIXencoding plasmids to Madin-Darby canine kidney cells was performed as described previously (15). Cells expressing appropriate amounts of FIX were maintained cell factories Dulbecco's modified Eagle's medium supplemented with 2.5% fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin, 1 g/ml amphotericin B, 0.8 g/ml desoxycholate, and 5 g/ml vitamin K 1 . FIX-containing medium was filtered to remove cell debris and concentrated ϳ10-fold employing a hollow fiber cartridge (Hemoflow F5, Fresenius, Bad Homburg, Germany). Benzamidine was added to a final concentration of 10 mM, and concentrates were stored at -20°C. FIX was purified from medium by immuno-affinity chromatography using monoclonal antibody CLB-FIX 14 according to an established procedure (15,16,20,21). Activated FIX was prepared by incubation of 1.4 M FIX with 7 nM FXIa in 100 mM NaCl, 5 mM CaCl 2 , and 50 mM Tris (pH 7.4) for 1 h at 37°C. Activation was stopped by the addition of EDTA (10 mM) and benzamidine (10 mM). Activated FIX was loaded on Q-Sepharose FF and washed with 150 mM NaCl, 5 mM benzamidine, 50 mM Tris (pH 7.4) to remove FXIa. FIXa was eluted from the column by the addition of 500 mM NaCl, 5 mM benzamidine, and 50 mM Tris (pH 7.4). FIXa preparations were dialyzed against 100 mM NaCl, 5% glycerol, 50 mM Tris (pH 7.4), and subsequently against the same buffer containing 50% glycerol. Final FIXa preparations were stored at -20°C. In agreement with previous studies (20), electrophoretic analysis demonstrated that purified recombinant FIX and FIXa preparations displayed the same mobility as their plasma-derived counterparts, indicating that no propeptide-containing unprocessed FIX species were present. As previously documented, the expression system employed produces FIX with normal Ca 2ϩ -dependent properties (20). In accordance with this notion, Gla analysis (22) revealed that recombinant FIX and FIXa preparations used in this study contained 9 -12 mol of Gla/mol of protein, with an average Gla content (mean Ϯ S.D.) of 11.1 Ϯ 1.2 mol/mol of FIX(a).
Proteins-Human FVIII was purified as described previously (23). Stable FVIIIa preparations were prepared by thrombin activation and CM-Sepharose chromatography essentially as described elsewhere (16). FVIII A2 domain was isolated from FVIIIa employing S-Sepharose chromatography as described (24). Plasma-derived FX and FIX were purified as outlined previously (25). Monoclonal antibody CLB-FIX 14 has been described previously (23) and was purified from culture medium employing protein A-Sepharose as recommended by the manufacturer. Ovalbumin was obtained from NBS Biologicals (Huntingdon, UK). Purified antithrombin and human serum albumin were obtained from the Division of Products of CLB (Amsterdam).
Protein Concentrations-Protein was quantified by the method of Bradford (26) using human albumin as a standard. FVIII activity was assayed by a spectrophotometric method employing bovine coagulation factors and a chromogenic substrate specific for FXa (Coatest FVIII, Chromogenix, Mölndal, Sweden). Pooled plasma, which was calibrated against the World Health Organization standard 91-666, was used as a standard. The amount present in 1 ml of human plasma (1 unit/ml) was assumed to correspond to 0.4 nM. FVIIIa concentrations were determined using a 1-stage clotting assay (16). The concentration of FX was determined as described previously (25) assuming that 1 unit/ml corresponds to 0.14 M FX. Concentrations of wt-FIXa and FIXa chimeras were measured by active site titration using antithrombin as described (19). Active site titration revealed that more than 90% of the protein in the FIXa preparations represented active enzymes.
Amidolytic Activity-Hydrolysis of CH 3 SO 2 -LGR-pNA was essentially performed as described previously (19). Briefly, 50 l of a solution of CH 3 SO 2 -LGR-pNA (0 -8 mM) was added to 50 l of a solution containing 200 nM FIXa and 5 mM Ca 2ϩ in a microtiter plate (Costar, flat bottom type). Initial rates of substrate hydrolysis were measured by monitoring the absorbance at 405 nm in time. Absorbance values were converted into molar concentrations using a molar extinction coefficient of 9.65 ϫ 10 3 M Ϫ1 cm Ϫ1 for p-nitroanilide (pNA) and a path length of 0.35 cm for a 100-l volume.
FX Activation-FX activation studies were performed as described (15)  Surface Plasmon Resonance Studies-The interaction between FIXa and the isolated FVIII A2 domain was assessed by surface plasmon resonance employing a BIAcore TM 3000 biosensor system (Biacore AB, Uppsala, Sweden). Purified FVIII A2 domain was immobilized on a CM5 chip using the amine coupling kit as prescribed by the manufacturer at a density of 35.7 fmol/mm 2 . FIX or FIXa variants were applied in varying concentrations at a flow rate of 20 l/min in a buffer containing 20 mM Hepes (pH 7.4), 150 mM NaCl, 2 mM CaCl 2 , and 0.005% (v/v) Tween 20 at 25°C. Regeneration of the sensor chip surface was performed by incubation with in a buffer containing 20 mM Hepes (pH 7.4), 1 M NaCl, 10 mM EDTA, and 0.005% (v/v) Tween 20. Data were analyzed using BIAevaluation software 3.1.

Construction of Recombinant FIXa Variants-Recombinant
FIX/FVII and FIX/FX chimeras were constructed in which the sequence connecting the EGF-like domains (i.e. residues Leu 84 -Thr 87 ) and the loop Asn 89 -Lys 91 in the EGF2 domain were exchanged for the corresponding sequence of both human FVII and FX (Fig. 1). With regard to the spacer between the EGFlike domains, it should be noted that in FVII as well as in FX, the sequence and length differs from that in FIX. Consequently, functional effects of the replacements could be attributed to either the variation in length or to the difference in sequence. To distinguish between these possibilities, two additional constructs were made, one FIX chimera in which the sequence Leu 84 -Asp 85 -Val 86 -Thr 87 was exchanged for the FX sequence Leu-Phe-Thr-Arg (FIX 85-87 /FX FTR ) and one chimera in which the "extra" FX residues, Lys and Leu, were inserted immediately after the FIX sequence Asp 85 -Val 86 -Thr 87 (FIX 85-87 /FX ϩKL ). Recombinant FIX variants were expressed and purified by immunoaffinity chromatography (see "Experimental Procedures"). All FIX mutants were activated by FXIa under conditions similar to those for recombinant wt-FIXa. All FIX chimeras could completely be converted into the active form, FIXa. The final preparations of activated normal and chimeric FIX were more than 90% active as determined by active site titration using antithrombin.
Amidolytic Activity of Recombinant FIXa Chimeras-To explore whether substitutions in the FIXa light chain affect activity toward a small synthetic peptide substrate, hydrolysis of various concentrations of synthetic substrate CH 3 -SO 2 -LGR-pNA was monitored in the presence of FIXa chimeras and Ca 2ϩ ions. All chimeras displayed similar rates of substrate hydrolysis compared with that of wt-FIXa, as demonstrated by the calculated catalytic efficiency (k cat /K m ) ( Table I). These results indicate that replacement of regions Leu 84 -Thr 87 and Asn 89 -Lys 91 in the light chain of FIX for corresponding sequences of FVII and FX does not have major effects on FIXa amidolytic activity.
Enzymatic Activity of FIXa Chimeras toward FX-To investigate whether proteolytic activity of FIXa chimeras toward FX was affected, FX activation by FIXa was assessed in the presence of phospholipids and Ca 2ϩ ions but in the absence of FVIIIa. wt-FIXa and FIXa chimeras appeared to be similar in their ability to activate FX. Data were fitted into the Michaelis-Menten equation to calculate the apparent Michaelis-Menten constant (K m,app ) and apparent catalytic rate constant (k cat,app ). As summarized in Table I, apparent K m and k cat values were found to be in the same range for wt-FIX and FIXa chimeras. This suggests that replacement of FIX sequences by the corresponding sequences of FVII or FX in regions 84 -87 and 89 -91 has only minor, if any, effects on enzymatic activity toward the macromolecular substrate FX.
FX Activation by FIXa Chimeras in the Presence of FVIIIa-It has previously been established that substitutions in the light chain of FIXa may affect response to the cofactor, FVIIIa (15,16,20,27). Therefore, the ability of FVIIIa to stimulate the rate of FXa generation by FIXa chimeras was addressed. As expected, FX was activated by wt-FIXa in the presence of FVIIIa in a saturable and dose-dependent manner (Fig. 2). The rate of FX activation by chimera FIXa 89 -91 /FX or FIXa 89 -91 /FVII was also stimulated by FVIIIa, although to a lesser extent compared with wt-FIXa ( Fig. 2A, Table II). A reduced response to FVIIIa was further observed for chimeras with replacements in the EGF spacer sequence (Fig. 2B, Table  II). The most dramatic effect in this regard was observed for chimeras FIXa 85-87 /FX FTRKL and FIXa 84 -87 /FVII. As calculated from the data of Fig. 2B, these two mutants displayed a 3-4-fold higher apparent K d than wt-FIXa, in combination with a 5-20-fold lower maximal rate of FX activation. These findings are compatible with defective FVIIIa binding as well as with impaired activity of the FIXa-FVIIIa complex when formed. Interestingly, a less marked decrease in stimulation was observed for chimera FIXa 85-87 /FX ϩKL (Fig. 2B). This may suggest that not only the increase in spacer length but also the replacement of FIX residues Asp 85 -Thr 87 by FX residues Phe-Thr-Arg affects FVIIIa-mediated stimulation. Indeed, chimera FIX 85-87 /FX FTR , which has a normal length of the inter-EGF sequence, was more than 2-fold less efficient than wt-FIXa in activating FX in the presence of FVIIIa (Fig. 2B, Table II). Because all chimeras displayed normal proteolytic activity toward FX in the absence of FVIIIa (Table I), our data suggest that cofactor-dependent activation of FX involves the inter-EGF sequence Leu 84 -Thr 87 (Fig. 2B) and, to a minor extent, also loop Asn 89 -Lys 91 ( Fig. 2A).
FX Activation by FIXa Chimeras in the Absence of Phospholipids-It has been proposed that FIXa EGF-like domains primarily serve to position the FIXa protease domain above the membrane surface for optimal FVIIIa interaction (28). We considered the possibility that the reduced response to FVIIIa by FIXa chimeras could be explained by a disturbed alignment of the FIXa protease domain toward the phospholipid surface as a result of replacement of the EGF spacer sequence Leu 84 -Thr 87 . This possibility was addressed by investigating FX activation by FIXa chimeras in the presence of FVIIIa but in the absence of phospholipids. Activation of FX in the presence of FVIIIa was enhanced in a dose-dependent fashion by wt-FIXa (Fig. 3). Similar rates of FXa formation were observed for chimera FIXa 85-87 /FX FTR (Fig. 3). In contrast, chimeras FIXa 85-87 / FX ϩKL , FIXa 85-87 /FX FTRKL , and FIXa 84 -87 /FVII proved to be ϳ5-fold less efficient than wt-FIXa in activating FX. These chimeras differ from wt-FIXa and FIXa 85-87 /FX FTR in the number of amino acids that link both EGF-like domains. Apparently, also in the absence of phospholipids these chimeras display a reduced response to FVIIIa. This effect, however, is selectively observed when an increased number of amino acids is introduced between both EGF-like domains. (29) demonstrate that FIXa activity can be enhanced by the isolated A2 domain of FVIII (29). Because FIXa chimeras described in the present study display an impaired FVIIIa-mediated stimulation, the effect of isolated A2 domain on FIXa activity was determined. As shown in Fig.  4, the activity of wt-FIXa was increased by the A2 domain in a dose-dependent manner. It should be mentioned that A2 domain-mediated stimulation is 65-fold lower compared with intact FVIIIa (Table II). Stimulation by A2 domain has been proposed to involve the FIXa protease domain residues 333-339 and, in particular, residue Arg 333 (21,30). As expected, control experiments showed that the variant FIXa-R333Q was associated with a total lack of response to the A2 domain (Table  II). Surprisingly, the FIX chimera FIXa 85-87 /FX FTRKL also displayed a sharply reduced response to the isolated A2 domain and, as such, behaved similar to the FIXa-R333Q variant (Table II, Fig 4). The activity of both chimeras FIXa 85-87 /FX FTR and FIXa 85-87 /FX ϩKL was enhanced by the A2 domain at least as efficiently as observed for wt-FIXa, whereas the stimulation of the FIX 84 -87 /FVII was decreased to the level activation of The arrows indicate the C-terminal region of the EGF1 domain, the spacer sequence, and the N-terminal region of the EGF2 domain. Amino acid residue numbering is that of FIX. The lower part indicates the various FIX chimeras that have been constructed and the amino acid replacements made therein.

Effect of Isolated FVIII-A2 Domain on FX Activation by FIXa Variants-Fay and Koshibu
the FIXa-R333Q variant. These results suggest that the orientation of the two EGF-like domains toward each other effects the enzymatic properties, in particular in the presence of FVIIIa. This is confirmed by experiments with a FIXa-E78K substitution mutant in which the salt bridge between both EGF-like domains is disrupted, which also becomes manifest by a defect in response to intact FVIIIa and the A2 domain (Table II, Fig. 4). The defective response to the FVIIIa cannot be attributed to under-carboxylation of the recombinant FIXa species, because the Gla content of individual mutants that lack response to the A2 domain varied (mean Ϯ S.D.) between 8.8 Ϯ 0.4 and 12.0 Ϯ 0.2 mol of Gla/mol of protein for FIX-E78K and FIX 84 -87 /FVII, respectively. These data thus are compatible with a mechanism in which optimal alignment of FIXa EGF-like domains is required for A2 domain-dependent enhancement of FIXa activity.
Association of FIXa Variants with Isolated FVIII A2 Domain-FIXa variants that lack response to the isolated FVIII A2 domain may either be defective in complex assembly with the A2 domain or may lack the rate enhancement that should occur upon A2 domain binding. To distinguish between these possibilities, the association of the FIXa variants with the FVIII A2 domain was assessed employing surface plasmon resonance technology. As shown in Fig. 5, wt-FIXa proved to be effective in A2 domain association, whereas the non-activated FIX precursor displayed only minor association with the FVIII A2 domain. The FIXa-R333Q variant, which lacks response to the A2 domain (Fig. 4) and has previously been reported to be defective in A2 domain binding (30), indeed displayed strongly reduced assembly with the A2 domain (Fig. 5), with a more than 5-fold reduced initial rate of association (data not shown). In contrast, the other FIXa variants that are lacking response to the FVIII A2 domain (Fig. 4) did assemble with the A2 domain, with an association rate close to that of wt-FIXa (Fig.  5). Apparently, the lack of response to the FVIII A2 domain was not due to a defect in A2 domain binding. This suggests that the mutants FIXa-E78K, FIXa 85-87 /FX FTRKL , and FIXa 85-87 /FVII are defective in that complex assembly with the FVIII A2 domain fails to induce the typical enhancement of FIXa enzymatic activity. DISCUSSION In the present study we have explored the role of two structure elements that are exposed at the interface between the two EGF-like domains, the small loop Asn 89 -Lys 91 at the N-terminal end of the EGF2 domain and the connection between the two EGF-like domains, residues Leu 84 -Thr 87 . Both elements are FIX-specific and are completely different in the related coagulation factors FVII and FX (Fig. 1). The rationale for directing our studies to FIX-specific elements at the junction between the two EGF-like domains was 2-fold. First, the crystal structures reveal that FIX is distinct from FX and FVII in that it contains a variety of inter-domain contacts. These restrict flexibility between the two EGF-like modules to a minimum (6,14). Second, we have previously demonstrated that such inter-domain contacts support the interaction of FIXa with its cofactor FVIIIa (15) or contribute to the enzymatic properties of the protease domain in a FVIII-independent manner (16). It thus seems reasonable to suppose that those con-

-Leu-Gly-Arg-pNA and activation of FX by FIXa in the absence of FVIIIa
Hydrolysis of various concentrations of CH 3 SO 2 -Leu-Gly-Arg-pNA (0 -8 mM) by FIXa (100 nM) and activation of FX (0 -500 nM) by FIXa (15 nM) in the absence of FVIIIa were assayed as described under "Experimental Procedures." Catalytic efficiency (k cat /K m ) for hydrolysis of CH 3 SO 2 -Leu-Gly-Arg-pNA was calculated as described (21). Data for FX activation were fitted into the Michaelis-Menten equation in order to obtain apparent k cat and K m values. The apparent catalytic efficiency (k cat /K m ) was inferred from k cat   tacts might exert their function by structure elements that are exposed at the interface between the two EGF-like modules.
In the absence of FVIII, none of the FIX variants that we examined displayed any significant defect in enzymatic activity (Table I). This implies that neither the loop Asn 89 -Lys 91 nor the spacer Leu 84 -Thr 87 supports the enzymatic properties in manner as described for the hydrophobic contact between the regions Phe 75 -Phe 77 and Lys 106 -Val 108 (16). In the presence of FVIII, however, the various FIX chimeras indeed displayed reduced activity, varying from only a minor reduction for chimeras with replacements of loop Asn 89 -Lys 91 (Fig. 2A) to a major defect for chimeras with replacements in the spacer region Leu 84 -Thr 87 (Fig. 2B). The chimera FIX 84 -87 /FVII, containing the entire spacer sequence from FVII, displayed the most prominent defect. By lacking a response to FVIIIa, this chimera has a phenotype similar to that of FIXa mutants in which the salt bridge between residues Glu 78 and Arg 94 was disrupted, thus affecting one of the contacts between the two EGF-like modules (14,15).
Previous studies provide evidence that assembly of the FIXa-FVIIIa complex involves two distinct interactions, one between the FIXa protease domain and the FVIII A2 domain (21,28,29) and the other between the FIXa light chain and the FVIII A3 domain (23,31). In view of the known K d values for the individual interactions, we previously proposed that FIXa-FVIII complex assembly is driven by the interaction between the respective light chains of FVIII and FIXa on the phospholipid membrane (32). Our present study (Fig. 3) demonstrates that the same mutants that display reduced activity in the presence of the isolated A2 domain are also defective in response to complete FVIIIa under lipid-free conditions. This suggests a predominant role of the A2 domain in complex assembly when lipids are absent. Further, it is striking that FIXa activity toward FX in the presence of the isolated FVIII A2 domain involves the residues Leu 84 -Thr 87 (Fig. 4). The length of this interdomain spacer, but not its sequence, determines the response to FVIIIa in the absence of lipids (Fig. 3). This is in contrast with the data obtained in the presence of phospholipids, which suggest that both spacer length and sequence contribute to response to FVIIIa (Fig. 2B). It is of further interest to compare the cofactor effect of intact FVIIIa with that of the isolated FVIII A2 domain in quantitative terms. In the presence of lipids, the isolated A2 domain is more than 20-fold less effective as a cofactor than complete FVIIIa (Table II). This demonstrates that stimulation of FIXa activity is greatly de-  Subsamples were drawn at various time points, and the amount of FXa generated was quantified as described under "Experimental Procedures." Rates of FX activation were calculated from at least three measurements within the initial 10 min of activation. Data represent the mean of two independent experiments. pendent on interaction with a site beyond the FVIII A2 domain, for instance with the major FIXa binding site in the A3 domain of the FVIII light chain (23).
One question is whether or not FVIIIa directly interacts with the loop Asn 89 -Lys 91 and the residues Leu 84 -Thr 87 at the EGF1-EGF2 domain interface. It is evident that both these structure elements contribute to FVIII-dependent FX activation in the presence of phospholipids (Fig. 2). In comparison with the domain linker Leu 84 -Thr 87 , however, the contribution of loop Asn 89 -Lys 91 is only minor. This opens the possibility that this loop serves another function than FVIII binding. In this regard it is interesting that a FIX chimera with FVII residues in positions 88 -99 in the EGF2 domain displays decreased binding to platelets in the presence and absence of FVIII, suggesting a role in assembly of the factor X-activating complex on the platelet membrane (33). The effect of replacements in the domain linker Leu 84 -Thr 87 is particularly prominent when mutation involves the introduction of extra amino acid residues (Fig. 2B). These findings do not exclude an indirect effect of the spacer on assembly with FVIIIa, for instance by maintaining the orientation of the EGF1 domain, and perhaps also the Gla domain, in the FIXa light chain. However, our data are also compatible with a direct role of the spacer in FVIIIa binding, which is apparently lost in the absence of phospholipids (Fig. 3). Recently, cross-linking studies demonstrate a direct interaction between a FIXa fragment of residues 68 -94 with a synthetic peptide of the FVIII residues 1804 -1818 (34). We previously identified this A3 domain region as a major FIXa binding site in the FVIII light chain (23,35). It is therefore conceivable that an interaction between the FVIII A3 domain and the FIXa EGF domain interface provides the molecular basis of the lipid-dependent assembly of the enzymecofactor complex.
One intriguing finding is that the interdomain spacer Leu 84 -Thr 87 in FIXa plays a role in the enhancement of enzymatic activity by the isolated FVIII A2 domain (Fig. 4) without a concomitant role in the association with the A2 domain (Fig. 5). The predominant role of the helix Arg 333 -Ser 339 in A2 domain binding (21,28,30) is supported by our observation that the mutant FIXa-R333Q combines a lack of response to the isolated FVIII A2 domain (Fig 4) with a defect in association with the A2 domain (Fig. 5). In contrast, defective rate enhancement is combined with normal A2 domain binding in FIX variants with mutations in the light chain. Variants of this type do not only comprise the chimeras FIXa 84 -87 /FVII and FIXa 85-87 /FX FTRKL with their elongated domain spacer but also the substitution mutant FIX-E78K, which lacks the salt bridge between the two EGF-like domains (Fig. 4, Table II). Elongation of the domain spacer may be incompatible with maintaining the salt bridge between residues 78 and 94, which supports the interaction with the FVIII light chain (15) and the rate enhancement upon assembly with A2 domain (Table II, Fig 4). It may seem confusing that, whereas contacts between the two EGF-like domains are crucial for assembly with FVIII, the entire EGF1 domain may be replaced by that of FVIIa or protein C without major functional implications (27,30). However, those chimeras have conserved both the short spacer of FIX and the residue Glu 78 in the EGF1 domains of FVII and protein C (15). Because the interdomain spacer in the FIXa light chain contributes to FVIIIa-induced activity enhancement, it seems reasonable to suppose that contacts between the protease domain and the light chain contribute to this phenomenon. According to the FIXa crystal structure, at least two molecular contacts might link these domains (Fig. 6). These involve residues Asn 92 and Phe 98 in the FIXa light chain and Tyr 295 and Phe 299 in the protease domain (Fig. 6). The latter site is closely associated with the FVIII-interactive region Lys 301 -Gly 303 (21) and, through hydrophobic contact between Leu 300 and Leu 330 , with the FVIII A2 domain FVIII-interactive region Leu 330 -Ser 339 (21,28). We propose that the FVIIIa-dependent activity of the FIXa protease domain is under control of an allosteric mechanism that involves the interdomain spacer in the FIXa light  (14). The backbone structure is shown in ribbon format. The space-filling residues represent the FVIII interactive sites in the protease domain comprising residues 300 -303 (dark gray), 330 -339 (black), and the inter-EGF domain spacer residues 84 -87 (light gray). The salt bridge connecting both EGF-like domains (Glu 78 and Arg 94 ) and the side chains involved in contact between the EGF2-domain (Asn 92 and Phe 98 ) and the protease domain (Tyr 295 and Phe 299 ) are indicated in black. In the crystal structure, the distance between Asn 92 and Tyr 295 is 2.4 Å, whereas Phe 98 is located at ϳ5 Å from Tyr 295 in a hydrophobic pocket that also comprises residues Phe 299 and Phe 302 (14). Residues Glu 78 in the EGF1 and Arg 94 in the EGF2 domain forming the inter-domain connection by a salt bridge are separated by 2.73 Å. chain as well as contacts between the light chain and the FIXa protease domain.