Molecular Defect (Gla+14 +Lys) and Its Functional Consequences in a Hereditary Factor X Deficiency (Factor X “Vorarlberg”)*

From the $Departments of Medicine and Pathology, and Center for Thrombosis and Hemostasis, University of North Carolina, Chupel Hill, North Carolinu 27599, IjMerck, Sharp, cmd Dohme Reseurch Laborutories, West Point, Pennsylvunia 19486, YlDepurtment of Celhdur und Structural Biology, Uniuersity of Texas Health Sciences Center, Sun Antonio, Texas 78284, §First Department of Medicine, University of Vienna, Vienna A-1090 and **Municipal Hospital, Feldkirch A-6800, Austria

vation by factor VIIa/tissue factor and by factor IXa. The decrease is much more marked for the extrinsic than for the intrinsic pathway.
Factor X (FX) is a plasma glycoprotein required in both the intrinsic and extrinsic pathways of blood coagulation (1). It is synthesized in hepatocytes as a single polypeptide chain; following several post-translational modifications, including glycosylation, vitamin K-dependent carboxylation of the ycarbon of the first 11 glutamic acid residues, and cleavage of the leader sequence, the mature protein is secreted into the circulation. FX circulates as a two-chain zymogen composed of a light chain (Mr 16,200) and a heavy chain (Mr 42,000) joined by a single disulfide bond (2). The light chain of FX contains the 11 T-carboxyglutamic acid residues, 1 residue of /?-hydroxyaspartic acid, and two growth factor-like domains, which show sequence similarity with epidermal growth factor. The heavy chain contains the activation peptide and the catalytic domain. A single arginine-isoleucine bond is cleaved in the heavy chain of FX by both the intrinsic and the extrinsic pathways of blood coagulation, generating factor Xa (FXa) and a small activation peptide (3). This reaction is catalyzed by factor IXa (FIXa) and its cofactor VIIIa (FVIIIa) in the intrinsic system and by factor VIIA (FVIIa) and its cofactor tissue factor (TF) in the extrinsic system. Both reactions require calcium ions and phospholipids (4,5).
The amino acid sequence for human FX has been derived from FX cDNAs and direct amino acid analysis (6)(7)(8)(9). The gene has also been isolated and partially characterized. It has been mapped to chromosome 13q32-qter (lo), where it spans approximately 25 kilobases. Although still incompletely characterized, it consists of seven introns and eight exons and shows considerable structural homology with the genes encoding the other vitamin K-dependent clotting factors (11 precedes the stop codon by one nucleotide. The 5' end of the mRNA has not been characterized. Congenital FX deficiency is inherited as an autosomal recessive trait (12). Considerable phenotypic heterogeneity exists among factor X variants. This is evident both structurally, with patients reported to have normal, reduced, or absent antigen levels (13,14), and functionally, as evidenced by the variety of patterns seen following activation of abnormal Factor Xs in either the intrinsic or the extrinsic system (15

koktion and ~haFacterization of the Mutant FX Gene
Southern blot analysis of a patient's (111/4) EcoRI-digested genomic DNA probed with a cDNA containing the whole coding sequence showed bands at 7.5,5.2,2.8, and 2.5 kilobase. The same bands were seen with genomic DNA from a nonrelated control (data not shown). Thus, there was no evidence of gross gene deletion or rearrangement in the patient's FX gene. The nucleotide sequence for all exons and for the exon/ intron junctions was determined from patient 111/4. Sequence analysis revealed two differences between this sequence and the normal FX cDNA sequence. One mismatch was found at bp 91 of exon II (bp 160 of the coding region of the cDNA) (Fig. 4). The G-A change at the first nucleotide of codon 14 results in the substitution of a lysine (AAA) for a glutamic acid residue (GAA), which normally undergoes y-carboxylation to form a Gla residue. The mutation abolishes a naturally occurring Z'aqI restriction site with the sequence 5'-TCGA-3' by changing the first base of the 3' palindromic sequence to an adenosine (5'-TCAA'3') ( Fig. 2). Thus, genomic exon II fragments amplified from the mutant allele cannot be cut with 7bqI. Enzymatically amplified and digested exon II fragments (Fig. 1) from the phenotypically homozygous family members III/l, 111/4, and III/9 uniformly show a single band at 185 bp representing the uncut exon II fragment. Therefore, these patients are genetically homozygous for the defect in exon II.
Phenotypically heterozygous family members (IV/l, IV/5, IV/g, IV/lo, and IV/ll) display three bands: a band of 185 bp which is also seen in homozygous patients and represents DNA amplified from a mutant allele, and bands of 84 and 101 bp representing the cut exon II fragment amplified from a normal allele. Thus, these patients are heterozygous for the defect in exon II. Complete cutting occurs when exon II is amplified from a normal control.
A second mismatch was detected at bp 57 of exon V (bp 424 of the cDNA coding sequence). The G-+A change at the first bp of codon 102 results in the replacement by lysine (AAG) of glutamic acid (GAG) (Fig. 3). This mutation abolishes a naturally occurring MnH site. The non-palindromic sequence within the recognition site for MnH (5'-(N7)GAGG-3') is changed to AAGG. Therefore, genomic exon V fragments amplified from the mutant allele cannot be cut with itf&I. Enzymatically amplified and digested exon V fragments from the phenotypically homozygous family members III/l, 111/4, and III/9 uniformly show three different bands: one band of 149 bp, representing uncut exon V fragments from a mutant allele and bands at 58 and 91 bp representing cut exon V fragments from a normal allele (Fig. 1). Therefore, these family members are heterozygous for the defect in exon V. The same bands were obtained when exon V fragments were amplified and digested from the phenotypically heterozygous family members. Thus, they too are heterozygous for the defect in exon V. Complete cutting occurs when DNA is amplified from a normal control.
Analysis of the pedigree (Fig. 1) shows that there are at least two mutant alleles in this family. One mutant allele carries the defect in exon II and the defect in exon V. The other mutant allele is defective in exon II but has a normal sequence in exon V. The possibility of a third mutant allele, normal in II but carrying the mutation in exon V, exists but cannot be proved or disproved based on the available kindred data. Family members who are phenotypically homozygous are genetically homozygous only for the defect in exon II. They are heterozygous for exon V.
Isolation of FX (Normal and Vorarlbergj FX Vorarlberg and normal FX were isolated from BaCIZprecipitated plasma using three different antibody columns: 1) a rabbit polyclonal anti-human FX antibody; 2) a monoclonal anti-FX antibody; and 3) a calcium-dependent rabbit polyclonal anti-human FX antibody. FX Vorarlberg and normal FX from any of the three antibodies ran as a single band on unreduced SDS-polyacrylamide gel electrophoresis with an apparent Mr of 70,000 and were both processed into their two-chain forms under reducing conditions, with an apparent Mr for the heavy chain of 47,000 (Fig. 5). Amounts of material loaded onto the gel were insufficient to allow visualization of the light chain, which stains poorly with silver stains. While not quantitated precisely, the yields from patient plasma for the three columns were similar, suggesting that factor X Vorarlberg exists as one population in patient plasma rather than two, i.e., one responsive to Ca'+ conformational changes and the other refractory to such changes. Still, to avoid isolating a subpopulation of factor X Vorarlberg, the non-Ca'+-dependent antibody columns were used for purification.

Gla Analysis
Amino acid analysis showed that neither the Gla content nor the /3-hydroxyaspartic acid content of FX Vorarlberg differed significantly from the content of normal FX (isolated by the identical procedures). FX Vorarlberg analyzed (as described under "Experimental Procedures") for 5.9 mol of Gla/mol of protein as compared with 6.3 mol of Gla/mol of normal factor X. The ratio of Glas in Vorarlberg to Glas in normal FX of 0.93 is almost exactly the value that would be expected (0.91) if a single Gla residue were removed from normal factor X. These results indicate that the substitution of Lys for Glu14 does not lead to a marked reduction in the Gla content of circulating factor X Vorarlberg.

FX-dependent Coagulnnt Activity of FX Vorarlberg Plasma
To determine coagulant activity of FX Vorarlberg plasma, human FX-deficient plasma was mixed with FX Vorarlberg (or normal) plasma and the PT, APTT, and RVV-times determined (Table II). Using the PT, FX Vorarlberg plasma has a FX clotting activity of 5% of normal. Using the aPTT, FX Vorarlberg plasma has a FX clotting activity of 25% of normal. When RVV is used to activate FX in plasma, FX Vorarlberg has a FX clotting activity of 15% of normal FX.  FX antigen was reduced to 20% of normal in the FX Vorarlberg plasma.

Coagukznt Activity of Purified FX VOFaF~beFg
Since the antigen levels were reduced in patient plasma, to determine the coagulant activity of the purified FX Vorarlberg, identical amounts of the purified FX Vorarlberg protein or normal FX were added to human FX-deficient plasma and the PT, the APTT, and the RVV-times of these plasmas were measured (Table II). Using the PT as a measurement of the FVIIa/TF-dependent, extrinsic pathway of blood coagulation, purified FX Vorarlberg has a clotting activity of 15% of normal. Using either the aPTT or RVV assays, purified FX Vorarlberg has a FX clotting activity of 100% of normal.

Activation of Purified FX vorarberg
The reduced activity of FX Vorarlberg in the PT could be due either to a defect in the activation of FX Vorarlberg by FVIIa/TF or to a defect of the activated factor thus formed. To characterize the interaction of purified FX Vorarlberg with the enzymes FVIIa, FIXa, and RVV, the initial rate of activation of purified FX Vorarlberg was compared with the initial rate of activation of normal FX using the chromogenic substrate Spectrozyme FXa. The increase in absorbance when FXa cleaves the chromogenic substrate is proportional to the FXa generated by the cleavage of the enzymes. All activations were performed at room temperature with Ca'+ concentration of 5 mM (the same Ca*+ concentration used in the coagulant assays).
Activation with FVZZa-Using a molar excess of human recombinant FVIIa in the presence of a tissue factor source (Ortho brain thromboplastin), the rate of activation of FX Vorarlberg was 15% of the rate of activation of normal FX (Table III).
Activation with FZXa-Using a molar excess of human FIXa/FVIIIa and phospholipid vesicles, the rate of activation of FX Vorarlberg was 75% of the rate of activation of normal FX (Table III).
Activation with RVV-When RVV was used to activate FX, no difference was observed in the rate of activation between FX Vorarlberg and normal FX (Table III). Ca2+ Dependence of FX Activation-Activation of FX by FIXa, FVIIa, and RVV is Ca*+-dependent.
Lowering the Ca*+ concentration impedes the rate of activation in a characteristic, non-linear fashion. The molecular defect in FX Vorarlberg affects a potential Ca*+-binding residue. We therefore compared the rate of activation of FX Vorarlberg and normal FX at different Ca'+ concentrations ranging from 0.5 to 5 mM. Fig. 6 shows the Ca*+ dependence of FX activation by FVIIa, FIXa, and RVV. The highest rate of activation of normal FX is set as 100%.
Activation immediate loss of activity below 3.5 mM Ca'+ (Fig. 6B). When the relative rates of the FX activation were used (normalized to 100% at 5 mM Ca*+), no difference in the Ca'+ dependence of FX Vorarlberg and normal FX was observed. Activation by FZXa-Beginning at a concentration of 5 mM Ca'+, normal FX showed a slightly increasing rate of activation until a Ca'+ concentration of 2 mM, with an immediate loss of activity below 2 mM Ca2+. FX Vorarlberg had a rate of activation that was slightly reduced compared with normal at 5 mM Ca*+ (Fig. 6A). Increasing the Ca*+ concentration to 7 mM resulted in a slight loss of activity. Below a Ca'+ concentration of 5 mM the rate of activation dropped immediately and was 10% at 3 mM Ca'+. Comparison of the relative rates of FX activation showed that the Ca'+ affinity of FX Vorarlberg is decreased.
Activation by RVV-Normal FX showed a slight increase in the rate of activation from 5 mM Ca'+ to 2 mM and lost activity immediately below that. FX Vorarlberg was normally active at 5 mM but lost activity below 5 mM ca*+ (Fig. 6C). The slopes of normal FX and FX Vorarlberg roughly paral-leled each other but were separated by a difference in the Ca*+ concentration of approximately 3 mM.

Conversion of Prothrombin by FXa
To determine the activity of activated FX Vorarlberg (FXa Vorarlberg) toward prothrombin, normal FX and FX Vorarlberg were activated with FVIIa, FIXa, or RVV. Identical amounts of the activated normal FX or FX Vorarlberg were then used to convert prothrombin to thrombin. At a Ca*+ concentration of 5 mM, the activity of FX Vorarlberg was 100% of normal FX, irrespective of the method of activation of FX (Table III).

Ca'+ Dependence of the Conversion of Prothrombin by FXa
To determine the activity of FXa Vorarlberg and normal FXa at different Ca*+ concentrations, FX (Vorarlberg and normal) was activated with RVV and then incubated with prothrombin at various Ca*+ concentrations (Fig. 7). Normal FXa showed a slight increase in the rate of activation of prothrombin from 5 to 1 mM Ca'+. Below a Ca'+ concentration of 1 mM, normal FXa showed an immediate loss of activity. FXa Vorarlberg showed a slight increase in the rate of activation of prothrombin from 5 to 2 mM Ca*+. Below a Ca*+ concentration of 2 mM normal FXa showed an immediate loss of activity. The slopes of normal FX and FX Vorarlberg paralleled each other. However, FXa Vorarlberg had an apparent Caz+ affinity which was 0.5 mM lower than that of normal FX.

Intrinsic Fluorescence Quenching
The intrinsic fluorescence quench of FX (normal and Vorarlberg) is shown in Fig. 8. Addition of Ca'+ to normal FX resulted in a continuous loss of intrinsic fluorescence until the Ca'+ concentration reached 2.5 mM. No further change in fluorescence was observed above this Ca*+ concentration. FX Vorarlberg showed the same loss of intrinsic fluorescence as normal FX until the Ca*+ concentration reached 1.5 mM. While further additions of Ca*+ decreased the fluorescence of normal FX, no further change was observed in FX Vorarlberg. DISCUSSION Analysis of the gene coding for FX Vorarlberg disclosed that a point mutation in exon II, which encodes the Gla domain, generates the coagulation defect in these patients. As the mutation is located within a naturally occurring TaqI restriction site, a clear distinction can be made between homozygous and heterozygous genotypes by checking for the TaqI restriction site in the amplified exon II fragments. We were able to show that family members who are genetically homozygous for the defect display a much more severely affected phenotype (i.e. homozygous phenotype) than family members who are genetically heterozygous. Thus, the two distinct classes of phenotype correspond to defects in either one or both alleles of exon II. The point mutation in exon V, present on one allele, does not appear to be the causative mutation. First, its appearance bears no relationship to the affected phenotype. Second, the affected residue (amino acid 102), in contrast to the one in the Gla domain, is not conserved at all among the vitamin Kdependent coagulation factors. The most likely possibility is that this amino acid change represents a polymorphism.
The mutation on exon II affects the codon for Gl@, a ycarboxylated glutamic acid (Gla) residue, and results in the substitution of a lysine at this position. Gla analysis indicates that this substitution results in the loss in circulating FX Vorarlberg of only a single Gla residue. Thus, residue 14 in factor X does not appear to be critical for carboxylation of the remaining Gla residues.
As is the case for the other vitamin K-dependent procoagulant proteins, T-carboxylation of glutamic acid residues in the amino-terminal portion of the molecule is required for the protein to have coagulant activity (1). It is not yet clear, however, how much of the Gla domain must be intact in order to preserve normal function; nor is it clear whether the intrinsic and extrinsic systems, which both result in the cleavage of a single bond in the heavy chain to generate FXa, depend in an equivalent manner on a fully carboxylated Gla region of X. It is known that complete removal of the Gla domain from factor X results in a species that cannot be activated by either VIIa/TF or IXa/VIIIa (35). Studies of the Gla domain of prothrombin have shown that loss of as few as 3 Gla residues results in a deficient Ca*+-dependent conformational change associated with poor phospholipid binding (36). Hiskey et al., modifying specific Gla residues in prothrombin fragment 1, also reported poor phospholipid binding (37). The study of factor X Vorarlberg provides a method for defining the functional characteristics of a single specific Gla residue.
Our initial characterization of the mutant protein was by clotting assays on patient plasma. Even this relatively crude measure demonstrated that activity in the extrinsic system (5% of normal), as measured by the PT, was markedly reduced compared with the amount of protein present (20% of normal). In contrast, the activities as measured in the intrinsic system (15%) and by Russell viper venom time (25%) were roughly equivalent to the antigen level. The reduction in activity in the extrinsic system might be due to either slowed activation or to reduced activity of the activated material. However, as shown in Table II, FXa Vorarlberg was fully active in prothrombin activation at 5 mM Ca*+, regardless of the pathway used to generate the activated factor X Vorarlberg. Thus, the lower activity seen in the coagulant assays based on the PT (at 5 mM Ca*+) must be due to a reduced rate of activation of purified FX Vorarlberg compared with normal (see Table III), confirming that the defect indeed occurs in activation.
Since the structural defect in factor X Vorarlberg occurs at Glai4, it seemed likely that the functional defect of slowed activation (by VIIa/TF) might be mediated through an alteration in Ca2+ ion binding by the defective factor X molecule. We therefore pursued two methods of examining Ca2+ binding to the mutant protein, the change in intrinsic fluorescence as a function of the Ca2+ concentration and Ca'+ dependence of the rate of activation of factor X Vorarlberg by the three different systems (intrinsic, extrinsic, RVV).
In the presence of Ca2+, factor X undergoes a conformational change required for activity in prothrombin activation. This Ca'+-dependent conformational change can be followed by measuring changes in intrinsic fluorescence of the molecule as a function of Ca'+ concentration. As shown in Fig. 8, the fluorescence quench of FX Vorarlberg is less than that exhibited by normal FX, suggesting that the conformational change in FX Vorarlberg is incomplete compared with that which normally occurs.
The data on rates of activation as a function of Ca2+ concentration provide further evidence that the slowed activation by VIIa/TF at 5 mM Ca'+ is mediated through an alteration in the Ca2+-dependent rates of activation. While activation rates are in a physicochemical sense a more indirect means of studying conformation than fhrorescence quenching, these rates are physiologically perhaps more relevant. Two conclusions can be drawn from the data in Figs. 6 and 7. In the plots of activation by RVV, IXa-VIIIa, and the activation of prothrombin, it is clear that FX Vorarlberg requires increased calcium concentration in order to reach the same activation rates as normal FX. Stated differently, the conformation required for full activity can be achieved but only at a higher Ca2+ concentration. The plot for VIIa/TF, on the other hand, represents a different situation. Even at optimal Ca2+ concentrations, the rate of FX Vorarlberg activation is only 15% of normal. Raising the Ca2+ concentration does not overcome this defect. Thus, our experiments suggest that carboxylation at Glar4 and the resulting conformational change are more critical for VIIa/TF-mediated activation of X than for IXa-VIIIa-catalyzed activation. This in turn suggests that the two complexes almost certainly have different sites of interaction with FX, despite the fact that they cleave FX at the same bond. It is not possible to determine from these data whether the reduced rate of activation by VIIa-TF depends specifically on the absence of the Gla at residue 14, or whether it results because Ca*+ binding is less than maximal (i.e. a point mutation at any other Gla residue would have a similar effect). Further study of this and of other proteins with mutations in the Gla region should prove informative from this standpoint.