Elena M. Faioni Haemophilia and Thrombosis Centre, Via Pace 9, 20122 Milan, Italy.
Two mutations in exons 3 and 9 of the protein C gene were identified by amplification and sequencing from symptomatic probands referred for venous thromboembolism and thrombophilia screening. The phenotype associated with the mutations is a type II protein C deficiency with normal amidolytic activity. In one family, the mutation in exon 3 (G3545→A), which predicts an R9 to H substitution in the Gla domain, was identified. A mutation in exon 9 (G10899→A), which predicts an R352 to W substitution in the catalytic site, was identified in the second family and has been reported previously in association with type II deficiency with low amidolytic activity. Western blotting of the purified proteins from the probands' plasma did not show any abnormal migratory pattern. Molecular modelling suggested a possible impairment in the recently described Na+ binding pocket for the R352→W mutant. No conclusions could be drawn relative to the R9→H mutant.
Deficiency in protein C (PC), the zymogen to the key regulatory enzyme of the coagulation system, activated PC, is relatively rare (2–4% of selected patients with venous thrombosis; De Stefano et al, 1996 ). It is associated with a lifelong increased risk of developing venous thrombosis, estimated as approximately five times that of individuals without the deficiency ( Koster et al, 1995 ). Both the gene and the protein have been well characterized. The genetic defects underlying PC deficiency have been resolved in part and have fostered several structure–function studies.
PC shares homologies with other vitamin K-dependent coagulation proteins as a result of a common evolutionary pathway ( Neurath, 1984). Its modular structure enables the differentiation of functional domains that correspond to the intron–exon organization of the gene ( Plutzky et al, 1986 ). The mature protein contains a Gla domain region (coded by exon 3), a connecting region (exon 4), two EGF-like modules (exons 5 and 6), the activation peptide (exon 7) and the catalytic domain (exons 8 and 9), with the conserved catalytic triad (H211, N256 and S360). A preproleader peptide (exons 2 and 3) is cleaved upon secretion of the mature protein. Among the post-translational modifications, γ-carboxylation of nine glutamic acid residues in the amino-terminal region (Gla domain: E residues 6, 7, 14, 16, 19, 20, 25, 26 and 29) has major relevance for calcium-mediated phospholipid binding and specific protein–protein interactions ( Colpitts et al, 1995 ).
Phenotypic classification of PC deficiency enables the identification of a type I deficiency (parallel reduction in concentration and function measured by amidolytic or anticoagulant methods) and a type II deficiency (normal or increased concentration and reduced function). The latter can be subdivided further in type II deficiency with both reduced anticoagulant and amidolytic activity and, in type II deficiency, with reduced anticoagulant activity only ( Reitsma et al, 1995 ). In the last published database of PC mutations, the majority of entries were related to type I deficiency (≈ 75%) and, of the remaining type II deficiencies, the vast majority were the subtype with reduced global function (≈ 95%). More than half the reported mutations are heterozygous missense mutations for both types ( Reitsma et al, 1995 ). In the rare type II phenotype with normal amidolytic activity, the largest number of mutations has been identified in exon 3 and exon 9, which code for the propeptide Gla domain and part of the catalytic site region respectively ( Gandrille et al, 1993 ; Reitsma et al, 1995 ). These natural mutants are very interesting from the point of view of the structure–function relationship of PC and allow insight into the biochemical role of specific residues.
In this study, we describe two thrombophilic families with type II PC deficiency with normal amidolytic activity. A mutation was identified in the PC gene of each of the probands and confirmed in the available family members. Electrophoresis of the purified proteins under denaturing conditions was not informative, whereas molecular modelling proposed an explanation for the functional impairment in one family. Although one of the mutations (in exon 3) is novel, the other one (in exon 9) has been associated with a different phenotype compared with the one reported here.
PATIENTS AND METHODS
The probands were referred to our centre for screening for inherited thrombophilia (see below) after a venous thromboembolic event. The proband from family A is a 41-year-old woman who had deep vein thrombosis of a lower limb complicated by pulmonary embolism in the postpartum in 1984; the proband from family B is a 25-year-old woman who had deep vein thrombosis of a lower limb after surgery and while on oral contraceptives in 1997.
The following tests are routinely performed on all patients referred for thrombophilia screening: search for antiphospholipid antibodies and lupus anticoagulant, measurement of antithrombin, PC, protein S, detection of factor V Leiden and the prothrombin A20210 allele, baseline and post-methionine load levels of homocysteine. When factor V Leiden is present, functional PC is measured by an appropriate test, which has been previously reported ( Faioni et al, 1996 ), i.e. anticoagulant activity is measured after adsorption and elution of PC from barium citrate. The complete thrombophilia screening was carried out in family members as well. For all biochemical assays, a normal range was established by standard methods. Results from each assay were calculated by interpolation of a curve generated by a reference plasma calibrated against the international standards, when available (antithrombin, protein C, protein S).
All salts and other common reagents were of molecular biology or electrophoresis grade. Molecular weight standards were purchased from Boehringer Mannheim, Milan, Italy (marker VIII), MBI Fermentas, St Leon-Rot, Germany (φ174/HinfI), Bio-Rad Laboratories, Hercules, CA, USA (low and high molecular weight standards). Polyclonal anti-PC antibody was purchased from Dako, Glostrup, Denmark, monoclonal anti-PC light chain from Sigma-Aldrich, Milan, Italy, and anti-human IgG gold or alkaline phosphatase-conjugated from Bio-Rad Laboratories. Oligonucleotides were synthesized by Primm (San Raffaele Science Park, Milan, Italy). Restriction enzymes were from Bio-Rad Laboratories (MaeIII) and Celbio, Milan, Italy (HpAII). Taq polymerase and Taq-Gold were from Perkin-Elmer (Applied Biosystems Division, Foster City, CA, USA).
Strategy for search of mutations
As most mutations associated with PC deficiency with normal amidolytic activity are located in exons 3 or 9, we first amplified these exons, then the amplified fragments were analysed by single-strand conformation polymorphism (SSCP). The fragments that showed an abnormal migratory pattern on SSCP were sequenced. As both mutations that were identified delete a restriction site, restriction enzymes were used to search for the same mutation in family members.
Polymerase chain reaction (PCR) of exons 3 and 9
The following pairs of oligonucleotide primers were designed and used for the amplification of exons 3 and 9 in the PC gene (exon 9 was amplified as three overlapping fragments, 9A, 9B and 9C) and for sequencing: 5′-CTC ATG GCC CCA GCC CCT CTT AG-3′ and 5′-TGC TGG TTA CCA GCT CGC CCC TGA G-3′ for exon 3; 5′-TCC CCG CAG CCC ACT CTG ACT GTG-3′ and 5′-ACG AAG GTG CGG TTT CTC TTG GC-3′ for exon 9A; 5′-TGG GGC TAC CAC AGC AGC CGA G-3′ and 5′-CCC AGC CCT GCA GGG AGG GTC G-3′ for exon 9B; 5′-AGC CCC CCA GAA GAG CTG GGC ACC-3′ and 5′-CCT TCA GCA TCC CCG GCT TGC AC-3′ for exon 9C. The numbering of nucleotides of the PC gene and of amino acids was according to Foster et al. (1985 ).
For exons 3 and 9A. Target sequences were amplified in a 30-μl reaction mixture containing 1.5 mmol/l MgCl2, 15 pmol of oligonucleotide primers, 0.2 μl of Taq Gold, 100–1000 ng of genomic DNA, 20 mmol/l each deoxyribonucleotide in 50 mmol/l Tris-HCl (pH 8.8) and 2 mg/ml BSA. The fragments were amplified by touchdown PCR ( Heker & Roux, 1996) in 35 cycles in a thermal cycler as follows: initial denaturing at 95°C for 9 min; subsequent cycles: denaturing temperature of 95°C for 30 s; annealing temperature: two cycles at 72°C for 30 s, two cycles at 70°C for 30 s, two cycles at 68°C for 30 s, two cycles at 66°C for 30 s, 27 cycles at 64°C for 30 s, extension at 72°C for 30 s with a final extension time of 5 min.
For exons 9B and 9C. Target sequences were amplified in a 30-μl reaction mixture containing 1.5 mmol/l MgCl2, 5 pmol of oligonucleotide primers, 0.2 μl of Taq, 100 ng of genomic DNA, 20 mmol/l each deoxyribonucleotide in 50 mmol/l Tris-HCl (pH 8.8) and 2 mg/ml BSA. The fragments were amplified by touchdown PCR in 35 cycles in a thermal cycler as follows: initial denaturing at 94°C for 2 min; subsequent cycles: denaturing temperature of 94°C for 30 s; annealing temperature: two cycles at 72°C for 30 s, two cycles at 70°C for 30 s, two cycles at 68°C for 30 s, two cycles at 66°C for 30 s, 27 cycles at 64°C for 30 s, extension at 72°C for 30 s with a final extension time of 5 min. Amplified fragments were visualized by ethidium bromide staining after electrophoresis on 2% agarose gel.
SSCP of PCR products
SSCP analysis was performed as outlined by Orita et al. (1989 ). PCR products were analysed on non-denaturing 10% polyacrylamide gels (50:1 acrylamide:bisacrylamide) at 20°C and 160 V for 6 h. DNA bands were visualized by silver staining.
Both forward and reverse DNA strands of amplified PCR products were sequenced by Primm on a 373A fluorescent automatic DNA sequencer (Perkin-Elmer, Applied Biosystems Division).
These were carried out by restriction analysis. The 3545 G→A substitution deletes a restriction site for Mae III in exon 3, whereas the 10899 C→T substitution deletes an HpaII restriction site in exon 9. The amplified products from the probands and their relatives were digested as suggested by the enzyme manufacturer. Digestion was checked on 3% and 2% agarose gels respectively.
PC was purified from the probands' plasma (1.5 ml) by batch adsorption on Affi-Gel-linked HPC4, a calcium-dependent monoclonal antibody directed against the activation peptide of PC (a gift from C. T. Esmon, Howard Hughes Medical Institute). Conditions for adsorption and elution have been reported elsewhere ( Faioni et al, 1988 ). Anticoagulant and amidolytic activities and antigen were measured in all eluates.
PC purified from the probands' plasma was analysed by polyacrylamide gel electrophoresis (30:1 acrylamide:bisacrylamide), in both reducing and non-reducing conditions and then transferred to nitrocellulose membranes (0.45 μm pore size) according to the method of Towbin et al. (1989 ). After blocking of non-specific sites, the membranes were incubated with rabbit anti-PC polyclonal or mouse monoclonal anti-PC light chain-specific antibodies. Further amplification and detection of the signal was obtained by incubation with goat anti-rabbit or rabbit anti-mouse (as appropriate) alkaline phosphatase-conjugated antibodies. Purified PC from a commercial source (Enzyme Research Laboratories, South Bend, IN, USA) and prestained molecular weight standards were run in parallel.
Operations were performed with the InsightII, Biopolymer and Discover programs (Molecular Simulation). The mutant R352→W was modelled by replacement of the R by W in position 352 (corresponding to 187 of the C chain of the PC crystallographic structure) ( Mather et al, 1996 ) available in the Brookhaven Data Bank ( Bernstein et al, 1997 ). The tryptophan lowest energy rotamer was chosen, and the whole C chain was subjected to energy minimization consisting of 500 steps of steepest descent and 500 steps of conjugate gradient method to relax steric overlaps and optimize the structure. For modelling operations relating to the R9→H substitution, as the PC crystallographic structure does not include the Gla domain, this was obtained by homology modelling from the corresponding region of factor VIIa ( Banner et al, 1996 ). In the interval from residue 1 to 45, the two sequences show 60% identical residues, including R9, with no insertions or deletions (Fig 1, bottom). The non-conserved residues were replaced in the factor VIIa chain by the corresponding ones of PC (A3→S, P10→H, G11→S, K18→I, Q21→I, S23→D, R28→K, K32→Q, D33→N, A34→V, E35→D, R36→D, K38→L, L39→A, I42→S, S43→K, Y44→H and S45→V); the structure was energy minimized to optimize the position of the side-chains. The mutant substitution R9→H was then executed according to the same procedure as for the R352→W mutant.
Phenotypes of the probands and family members
1 Table I illustrates the results of assays for PC. The proband of family A shows an abnormal phenotype, with low anticoagulant but normal amidolytic activity. No other causes for this pattern (namely acquired transient causes) were determined to explain this result, which was confirmed after 1 month on another plasma sample. As the mother was on oral anticoagulants at the time of blood sampling, PC deficiency was hypothesized based on the discrepancy observed between anticoagulant and amidolytic activities. The mother had a cerebral venous thrombosis at the age of 64 years, whereas the father, who had PC levels within the normal reference range in all assays, was asymptomatic. No other family members were available for testing.
Table 1. Table I. Laboratory phenotype of members of the two families with type II protein C deficiency. Values outside the normal reference range are in bold.* After adsorption ( Faioni et al, 1996 ).† Protein C gene mutation.‡ On oral anticoagulants at the time of blood sampling.§ Heterozygous for factor V Leiden.
In family B, where the factor V Leiden was present in the proband and two family members, PC anticoagulant activity was measured after barium citrate adsorption, as stated in Methods. The proband and one brother (brother 1) of family B showed a type II PC deficiency, with low and normal amidolytic activity ( 1 Table I). The mother had borderline levels, whereas the father, one brother (brother 2) and the sister had levels of activity well within the normal range. Only the proband, who also carried factor V Leiden, was symptomatic.
Results of genetic analysis
By the mutation search strategy described in Methods, a mutation in exon 3 of proband A, 3545 G→A, which predicts an amino acid change in position + 9 of the mature protein, R→H was found. A mutation in exon 9 of proband B, 10899 C→T, which predicts an amino acid change in position + 352 of the mature protein, R to W was also detected. Both missense mutations delete a restriction site, for Mae III and HpAII respectively. This enabled us to search for the mutation in the other family members available for study (Fig 2 A and B). In family A, the mother carried the same mutation as the proband. In family B, the mother and one brother (brother 1, 1 Table I) also carried the mutation.
Electrophoretic analysis of purified PC from the probands
After affinity purification on HPC4, PC was electrophoresed, transferred to nitrocellulose membranes and incubated with either rabbit polyclonal anti-PC antibodies (not shown) or monoclonal anti-PC light chain antibody (Fig 3). No abnormalities could be shown in the mobility of purified PC from the probands or from their family members compared with PC from a normal individual (Fig 3). To verify that the Ca2+-dependent monoclonal antibody HPC4 recognizes the variant PCs, measurement of PC antigen, amidolytic and anticoagulant activity was also performed on the affinity-purified PCs. The results obtained showed that the variant PC was co-eluted with the normal protein (not shown).
R9 is located at the N-terminus of the light chain, about 9 Å apart from the Ca2+ binding site. In the factor VIIa structure, as well as in the activated PC model, it forms an ion pair with the γ-carboxylated glutamic acid 6, which is deleted by the substitution with the histidine. However, as the histidine is in an external, solvent-exposed position, it is difficult to hypothesize from the model any implication for the Ca2+-binding efficiency or a local misfolding effect. R9 is a highly conserved residue across the vitamin K-dependent proteins in humans and across species (Fig 1).
R352 is located in the catalytic domain of the activated PC on the C-terminus side of the loop spanning from glycine 347 to glutamine 353 (Fig 4A). The residue is largely in a solvent-accessible position and does not appear to have interactions with surrounding residues. It is about 20 Å from the catalytic triad region and at the opposite side with respect to the light chain interface. The replacement by tryptophan is responsible for a change in the surface charge of the molecule, because of the loss of a positive charge, and for a reduction of about 3% in the total solvent-accessible surface. However, it does not appear to have a dramatic effect on molecule folding, as it does not seem to disrupt important electrostatic contacts and can easily be accommodated in the surrounding space. Interactions with areas known for a specific functional activity, such as the substrate binding pockets (amino acid residues 311–325, 376–384 and 390–404) ( Mesters et al, 1991 , 1993; Marchetti et al, 1993 ) as well as with the Ca2+ and thrombomodulin binding areas (amino acids 225–235) ( Vincenot et al, 1995 ) were also examined, but no detrimental effects were observed as a result of the presence of the substitution. It might be of some importance that the loop in which R352 is located is adjacent to loop 385–390, which has recently been found to be responsible for the binding of Na+ and other cations to PC (Fig 4B) ( He & Rezaie, 1999).
Family A, R9→H mutation
The most frequent mutations associated with the infrequent phenotype of family A are found in exon 3, which codes for the propeptide and the Gla domain regions. This is of course not surprising, as these regions are implicated in calcium-mediated membrane binding, which is essential for anticoagulant activity ( Zhang et al, 1992 ). Previously reported mutations involve various amino acid residues, mainly those clustering around the prepropeptide region and the glutamic acid residues ( Poort et al, 1993 ; Gaussem et al, 1994 ; Reitsma et al, 1995 ; Aiach et al, 1997 ). Substitutions affecting R residues in the propeptide modify the conformation of the Gla domain, and R15→W is also associated with reduced anticoagulant activity ( Gandrille et al, 1993 ). Similarly, it is possible that the substitution of R9 in family A modifies the conformation of the Gla domain. This would not be evident upon gel electrophoresis in denaturing conditions, unless gross abnormalities of the protein were present, such as that reported for the R1→C in which the free cysteine allowed a complex to be formed with α1-microglobulin ( Wojcik et al, 1996 ). Neither would it necessarily be evident upon modelling of this region, especially considering that direct crystal resolution of the Gla domain of PC has not been accomplished. Also, inferences of the effect of amino acid substitutions are made from modelling performed in the absence of solvent and especially in the absence of Ca2+, the most relevant cation ligand of this region. It must be added that R9 is a highly conserved residue across species and within the vitamin K-dependent proteins, as shown by the alignment in Fig 1.
Family B, R352→W mutation
Mutations in the catalytic site are more rarely associated with the phenotype of family B and, by definition, they do not affect the catalytic activity, as amidolytic activity is conserved. However, they could alter the tridimensional structure of the binding sites for the physiological substrates (factors Va and VIIIa) or of the thrombin–thrombomodulin or the Ca2+ binding sites. This has been shown for mutants created by site-directed mutagenesis, but also for some natural mutants. For example, Arg-229→Gln, Gly-381→Ser, Met-343→Ile and Ser-252→Asn are associated with a type II deficiency with normal amidolytic activity, and the mutations affect the putative calcium-binding loop or substrate-binding pockets or limit the access of the substrates ( Marchetti et al, 1993 ; Poort et al, 1993 ; Reitsma et al, 1995 ). R352, however, does not seem to have direct interactions with any of these areas. The tryptophan residue seems to be well accommodated, and its side-chain does not interact with any other residue. Wacey et al. (1993 ) reported that W352 is very close to D172, and this could be a detrimental interaction that would lead to molecular instability. Their model, however, was not based on the crystal structure of PC, which became available only in 1996. We could not confirm this interaction.
The loop in which R352 is located is adjacent to loop 385–390, which binds Na+ and other cations in PC, thereby enhancing its catalytic efficiency ( He & Rezaie, 1999). In thrombin, where the Na+ binding site was first identified ( Wells & Di Cera, 1992), R187 (corresponding to R352 in the PC numbering used in this study) participates in Na+ binding by stabilizing the pocket through the formation of a double ion pair with D221 (corresponding to 385 in the PC numbering used in this study) and D222 (corresponding to 386 in the PC numbering used in this study) on the neighbouring loop ( Di Cera et al, 1997 ). In PC, R352 has a different orientation, which is more exposed to solvent, probably because of the absence of the aspartic acid charges, replaced by glycine and leucine respectively. Although there are no structural data on the Na+ localization in activated PC or on its entry way, the presence of the bulky hydrophobic side-chain of W352 in this area could prevent Na+ access to its binding site, or limit it. However, this should also result in an impairment of amidolytic activity. In fact, the original paper reporting the R352→W mutation identified a family with a type II PC deficiency with reduced amidolytic activity ( Doig et al, 1994 ). Although patients from family B who were heterozygous for the R532→W substitution had levels of amidolytic activity above the lower limit of the reference range, their amidolytic activity was lower than that of the family members who do not carry the mutation (67%, 73% and 72% vs. 110%, 95% and 107%; Table I). Tests specific for the Na+-induced conformation, to be performed with the purified proteins obtained from carriers of the mutation, are clearly needed to establish the exact phenotype and to prove the altered Na+ binding.
In family A, the mutation co-segregates with the clinical phenotype because both the proband and the mother, who carry the mutation, have thrombotic symptoms. In family B, only the proband was symptomatic, so that no comment can be made regarding the causative role of the PC mutation. However, the proband also carried another prothrombotic mutation, namely factor V Leiden, which was also carried by the sister and the father, who were asymptomatic and did not have a PC gene mutation. It could be argued, therefore, that the proband had a thromboembolic event because of the presence of a double thrombophilic defect.
In conclusion, we identified two point mutations in the protein C gene of patients and family members carrying a type II protein C deficiency phenotype, leading to amino acid substitutions in the Gla domain and catalytic regions. No gross abnormalities could be shown in protein C purified from the plasma of heterozygous carriers of the mutations. Molecular modelling studies did not reveal any detrimental interactions of the substituted residues, with the possible exception of impairment by 352W of sodium access to the sodium-binding loop.
This work was supported by institutional grants from IRCCS Ospedale Maggiore Policlinico.