Dr David Habart, Institute of Hematology and Blood Transfusion, U nemocnice 1, Prague 2, 12820, The Czech Republic. Tel.: +42 022 1977218; fax: +42 022 1977412; e-mail: email@example.com
Summary. Detection of causal mutations is required for genetic counseling. Molecular modeling combined with patients' phenotype provides significant insight into structure–function relationship of factor (F)VIII molecule. Our objective was to identify defects in the gene of FVIII by a sensitive and simple scanning technique with high throughput in order to study molecular mechanisms by which novel amino acid substitutions may lead to hemophilia A. A cohort of 81 families with mild, moderate and severe hemophilia A negative in intron 22 inversion was studied. For detection of mutations in the FVIII gene a conformation sensitive gel electrophoresis (CSGE) was modified by multiplexing. Thirteen novel amino acid substitutions were studied by molecular modeling and a correlation with the cross-reactive material (CRM) phenotype was performed. In 74 families, 59 different mutations were detected. Six different mutations were recurrent in 21 unrelated families. Thirty-four novel mutations included 19 point mutations, four small insertions, nine small deletions and two complex mutations. Thirteen novel amino acid substitutions occurred at residues conserved in FVIII orthologs. Five of them were associated with a discrepancy between FVIII activity and antigen; another five with CRM reduced phenotype and one with undetectable FVIII antigen. Multiplexing of the CSGE significantly increased its throughput without substantial loss of sensitivity. Molecular modeling suggested mechanisms by which substitutions at residues 382 and 569, located outside the proposed FIXa-binding region, may influence FVIII/FIXa interaction. His2155 was predicted to participate in FVIII/VFW binding.
Hemophilia A is an X-linked (Xq28) bleeding disorder caused by absent, reduced or dysfunctional factor (F)VIII protein in circulation. The FVIII gene consists of 26 exons dispersed across 180 kb and encodes ∼9 kb mRNA. The only common gene defect in hemophilia A is an intron 22 inversion, which occurs in 45% of patients with severe phenotype . Recently, a prevalence of up to 5% in severe hemophiliacs was reported for an intron 1 inversion . In the remaining severely hemophilic patients, a variety of nonsense, missense, splice-site and frameshift mutations or large deletions/insertions has been detected. Moderate and mild cases usually result from missense mutations .
Mutation detection is required for carrier status determination and for prenatal testing. With respect to the size of the gene and variety of the mutations, different approaches have been described, ranging from direct sequencing  to diverse scanning methods that minimize the number of finally sequenced segments . Despite considerable improvements a direct sequencing remains unaffordable to many laboratories. The denaturing gradient gel electrophoresis (DGGE) is sensitive, predictable and highly reproducible, but its throughput cannot be increased easily . Single-strand conformation polymorphism (SSCP) analysis usually requires radioactive chemicals, which is unsuitable for routine usage. Heteroduplex analysis by means of dHPLC (denaturing high-performance liquid chromatography) is rapid and highly sensitive, but a costly apparatus is required . Williams et al. applied a conformation-sensitive gel electrophoresis (CSGE) for the screening of the FVIII gene . It is technically simple and has a potential for throughput increase.
The FVIII gene encodes a glycoprotein, which is composed of three A domains homologous to a copper-binding protein ceruloplasmin, two C domains homologous to the discoidin family including coagulation FV and a single B domain without any known homology. The FVIII protein is secreted into plasma in a form of non-covalently associated heterodimer consisting of heavy (domains A1, A2 and B) and light (domains A3, C1 and C2) chains . The FVIII circulates in plasma non-covalently associated with von Willebrand factor (VWF), which protects it against degradation . After specific proteolysis by thrombin, it is released from the VWF in the form of heterotrimer activated FVIII (FVIIIa; A1/A2/A3C1C2). The FVIIIa binds to the phospholipid bilayer, where it increases protease activity of FIXa in the tenase complex . The structure of the FVIII C2 domains has been solved by X-ray crystallography . The A domains are currently modeled on the basis of ceruloplasmin X-ray studies . The three-dimensional structure of membrane-bound FVIII has been obtained by electron crystallography . The study of naturally occurring mutations in hemophilia A patients has been successfully used to elucidate a structure–function relationship of the FVIII molecule [14–17].
In this report, we present a modification of the conformation sensitive gel electrophoresis that substantially increases its throughput without significant loss of sensitivity. We report on 59 different mutations detected by this technique. Based on molecular modeling correlated with phenotypic data, we propose molecular consequences of 13 novel amino acid substitutions.
Materials and methods
A cohort of 81 apparently unrelated index patients included severe hemophiliacs, negative in the intron 22 inversion and those who suffered from moderate and mild hemophilia A. FVIII:C was determined by a one-stage assay . FVIII:Ag was measured by ELISA using the Asserachrom FVIII:Ag kit (Stago, Asnières, France). Some patients have been sampled repeatedly over several years and the results in the tables represent a range of obtained values as a percentage of normal. Cross-reactive material positive (CRM +) patients were defined as those with more than 30% reduction of the FVIII activity as compared with the FVIII antigen . CRM reduced patients had a roughly comparable level of FVIII:Ag to FVIII:C . Plasma FVIII inhibitor activity was estimated according to  and expressed in Bethesda units (BU). Genomic DNA was isolated from peripheral blood leukocytes by a modified perchlorate method . The FVIII gene inversion status was determined by Southern blotting according to . Presence of the intron 1 inversion was tested by multiplex PCR as described in . Intron 19/Hind III polymorphism was tested as in . Microsatelite alleles in introns 13 and 22 were amplified with primers from  and separated by capillary electrophoresis with fluorescent detection on the Genetic Analyzer ABI 310.
FVIII gene amplification
Twenty-six exons with exon/intron boundaries including 5′-and 3′-untranslated regions were amplified in 35 segments with the same sets of primers as in  but with minor modifications. Exons 17 and 18 were amplified with primers published in . Twenty-two segments were amplified in 11 multiplex PCR reactions, while the remaining 13 segments were amplified individually (Table 1). The PCR reaction buffer (1 ×) consisted of 16 mmol L−1 (NH4)2SO4, 67 mmol L−1 Tris-HCl (pH 8.8) and 0.01% Tween-20. In 50 µL of final volume each PCR contained 300–500 ng of genomic DNA, 1.5 mmol L−1 MgCl2, 200 µmol L−1 dNTP and 1.25 U (or 2.5 U in multiplex PCR) of Taq polymerase (Promega, Madison, WI, USA) with an appropriate amount of primer (Table 1). Amplification was carried out on Peltier Thermal Cycler PTC-200 (MJ Research). After initial denaturation for 1 min at 94 °C, 35 cycles of 20 s at 93 °C, 20 s at an appropriate annealing temperature (Table 1) and extension for 1 min 30 s at 72 °C, a final extension for 10 min at 72 °C followed.
Table 1. Multiplex amplification and electrophoresis in M-CSGE
Fragment size (bp)
Final primer concentration (µmol L−1)
Annealing temperature (°C)
Fragments amplified in one multiplex PCR are marked ‘m’. *For amplification of exon 21 we ran 10 cycles at 50 °C, 10 cycles at 51 °C and 25 cycles at 52 °C.
Multiplex conformation sensitive gel electrophoresis (M-CSGE)
Heteroduplexes were prepared by mixing 5 µL of individual PCR products of an index patient with 5 µL of the control sample (healthy donor) and by heating them for 2 min at 98 °C, 10 min at 65 °C and 10 min at 37 °C. Twenty-four heteroduplex reactions from 24 separate PCRs were combined, commensurate with amplicon sizes, into 11 groups to be run on a CSGE gel. Each group was made by mixing 5 µL from 2–4 heteroduplex reactions with 2 µL of a loading buffer, thus combining 3–4 segments in a single line. The loading buffer contained 50% glycerol and 0.25% xylene cyanol in distilled water. Electrophoresis was carried out on the Model S2 apparatus (Life Technologies Inc., Paisley, UK) with gel dimensions (H × W × T) of 38.5 × 31 × 1 mm. A homemade 36-well comb allowed analysis of five patients on two CSGE gels. The gel consisted of 10% acrylamide (99 : 1 acrylamide (Sigma, St. Louis, MO, USA) : bis-acrolylpiperazine (BAB, Fluka, Buchs, Switzerland), 10% ethyleneglycol, 15% formamide (Sigma) in 0.5 × TTE buffer (1 × TTE = 89 mmol L−1 Tris, 28.5 mmol L−1 taurine (Sigma), 0.2 mmol L−1 EDTA). Ammonium persulfate (Bio-Rad) and N,N,N′,N′– tetramethylethylene diamine (TEMED, Serva, Heidelberg, Germany) were used to catalyze polymerization. Prior to loading, the gel was electrophoresed at 700 V for 1–2 h 5 µL of heteroduplexed amplicons were applied on the gel and separated at 9 W for 18 h. The gel was stained with ethidium bromide. Detection was carried out on the Fluoro-imager FLA2000 (Fuji Film) and digital images were created. Aida 2.0 software with black and white as well as color modes was used for detection of shifted bands (Fig. 1).
DNA sequencing and data analysis
The FVIII gene fragments displaying unmatched shift in M-CSGE were directly sequenced using the BigDye Terminator Cycle Sequencing Kit (Applied Biosystems) and analyzed by the capillary sequencer Genetic Analyzer ABI PRISM 310 (Perkin Elmer) with the Sequencing Analysis 3.0 software working on the Macintosh platform. Both DNA strands of each sample from two separate PCR reactions were sequenced. The sequences were compared with the wild type using an online multiple sequence alignment engine at NCBI database . The numbering of FVIII gene codons is according to .
Sequences for alignment of the homologous proteins were taken from the Swissprot database (http:/us.expasy.org/sprot). The accession numbers are as follows: FVIII (human P00451, pig P12263, canine O18806 and mouse Q06194), FV/FVIII (fish Q90 × 47), FV (human P12259, pig Q9GLP1, bovine Q28108) and ceruloplasmin (human P00450). Missense mutations were studied based on a model of B-domainless FVIII , which was downloaded from the HAMSTeRS database . For the C2 domains the mutations were modeled based on the crystal structure published in . The potential effects of amino acid substitutions were modeled by SwissPdb Viewer software  and the What If program [31,32]. The model quality was checked with What If validation tools . The final images were created by ViewerPro 4.2 software (Accelrys, http://www.accelrys.com). Buried residues were defined by solvent accessibility below 10%.
A total of 59 different mutations were detected in 74 apparently unrelated families. All the identified mutations are presented in Tables 2–5 according to their type. Among 59 different gene defects, there were 37 point mutations (Table 2), seven small insertions (Table 3), 13 small deletions (Table 4) and two complex changes (Table 5). Thirty-four novel mutations included 19 point mutations, four insertions of 1–5 bp, nine deletions of 1–16 bp and two complex defects. Non-CpG transitions and transversions predominated among novel point mutations (16/19). One novel missense mutation involved a double substitution of two adjacent nucleotides 1564–5 (Table 2).
Table 2. FVIII point mutations (missense, nonsense and splice site). The consensus of native residues in FVIII, FV, FV/FVIII and ceruloplasmin (Cp) from different species (h, human; p, porcine; c, canine; m, mice; b, bovine; f, fish) is shown. The resulting stop codons are designed as AMB (amber), OCH (ochre) and OPA (opal)
Amino acid change
Novel mutation; +, inhibitor detected; r, recurrent mutation; d, de novo mutations;
value taken from another center; B, buried residue; E, exposed residue; –, value not determined. Protein domains (A1–C2) are stated.
Novel mutation; d, de novo mutations; subst/ins, single nucleotide substitution combined with a single nucleotide insertion; del/subst, single nucleotide substitution combined with a single nucleotide deletion. Protein domains (A1–C2) are stated.
The native residues at which novel amino acid substitutions occurred were compared with the FVIII proteins from four different species (human, porcine, canine and mice) and to homologous proteins of FV (human, porcine and bovine), zebrafish FV/FVIII and human ceruloplasmin (Table 2). Thirteen native residues were conserved in FVIII and six were also conserved in FV. Based on the FVIII models , seven native residues were buried (Table 2). Eight substituted residues were located to loops, four to β-sheets and one mutation targeted a cysteine residue predicted to form a disulfide bond. The CRM status was determined for 11 novel amino acid substitutions.
CRM positive phenotype
Trp382Gly was present in a mild/moderate hemophiliac. Trp382 is conserved in all compared homologous proteins except for ceruloplasmin (Table 2). Its side chain forms an amino–aromatic interaction  with the backbone of Gln561. Gln561 is located on the loop 558–565 which was shown to be critical in the modulating of FIXa activity . Gln561 lies adjacent to Arg562, which was proposed to directly contact the protease domain of FIXa . The replacement of the aromatic side chain by glycine predicts loss of the intramolecular stabilization (Fig. 2a,b). The ensuing conformation change of the FIXa binding region may explain the reduction of the cofactor activity.
Pro521Thr was associated with a moderate hemophilia with mild reduction of FVIII antigen. Pro521 is a buried residue conserved in all tested homologous proteins (Table 2). Proline is known as a protein-backbone breaker and its substitution by threonine is predicted to introduce a novel hydrogen bond with Thr522. A conformation change of the loop 517–527 bearing three residues directly interacting with FIXa (Val517, Lys523 and Arg527)  may explain the decrease of the cofactor activity.
Asp569Tyr was found in a clinically severe hemophiliac with moderate CRM + phenotype. Asp569 is a buried residue conserved in all compared homologous proteins. Substitution of Asp569 for bulky tyrosine is supposed to change local conformation in the vicinity of residues Glu557, Gln561 and Arg562 proposed to directly interact with FIXa .
Substitution Gly701Ser was detected in a severe hemophiliac with moderately reduced FVIII antigen. Gly701 is not conserved in FV. The atypical CRM + phenotype (Table 2) in this patient seems to spring from the inability of the mutant FVIII to accommodate the bulkier side chain affecting protein stability. The misfolded mutant FVIII secreted into plasma is apparently non-functional.
His1961Pro was found in a moderate hemophiliac with mild reduction of FVIII antigen. His1961 is a buried residue conserved only in the FVIII orthologs and ceruloplasmin (Table 2). His1961 is predicted to form hydrogen bonds with the backbone of Pro1990 and Lys1991, which are lost upon replacement by proline.
CRM reduced phenotype
Leu307Pro was found in moderate hemophilia. Leu307 is conserved in FVIII across different species, but not in FV. Replacement of leucine by proline probably changes the conformation of the β-sheet S16.
Substitution Glu1885Asp was associated with moderate phenotype. Glu1885 is a buried residue conserved in all compared homologous proteins. It is predicted to form two hydrogen bonds with conserved residues of Phe1730 and Asn1894, both within the A3 domain. No naturally occurring mutations have been reported at these residues. Aspartic acid substituting for the glutamic acid results in the loss of both predicted hydrogen bonds, which may cause protein instability (Fig. 2c,d).
His2155Asp was associated with moderate/mild CRM-reduced phenotype. The residue of His2155 is exposed on the surface of the β-sheet 18 within the C1 domain. The replacement of neutral His2155 by negatively charged aspartic acid is predicted to change local surface charge suggesting alteration of a protein–protein interaction (Fig. 2e,f).
The substitution Leu2210Phe was found in five unrelated families with mild hemophilia. Leu2210 is a partially exposed residue conserved across the discoidin family . The introduction of bulkier phenylalanine slightly modifies the surface of the C2 domain, which is consistent with mild phenotype.
CRM negative phenotype
The only substitution associated with CRM negative phenotype was Cys2326Ser. This can be explained by protein misfolding resulting from the destruction of a conserved disulfide bond stabilizing the C2 domain.
Undetermined CRM status
Cys310Phe was detected in a severe hemophiliac. Cys310 is a buried residue conserved in FVIII orthologs and ceruloplasmin only (Table 2). Its substitution results in a loss of the proposed copper binding site in the A1 domain, which is required for FVIII stability . Ile305 is a buried residue and its substitution for tyrosin may cause protein instability due to poor accommodation of the hydroxyl group.
His1047Tyr was found in a severe hemophiliac who developed a high responding FVIII inhibitor. His1047 is not conserved in the porcine B domain. No homology model of the B domain is currently available to allow the prediction of the structural consequences of this substitution.
The CSGE was modified by multiplexing the steps of amplification and electrophoresis (M-CSGE). The modification significantly increased its throughput without substantial loss of sensitivity. It allowed parallel testing of five index patients on two CSGE gels. The estimated detection rate was 90% on a panel of 30 different previously known mutations (data not shown). In the cohort of 81 index patients, M-CSGE failed to detect a mutation in five severe cases. Two of these later tested positive and two tested negative for the intron 1 inversion (in one case DNA was not available for testing). No mutation was found in two additional mild hemophiliacs in whom type 2 N von Willebrand disease was excluded by normal VWF:FVIII binding (data not shown). Out of 59 different mutations identified in 74 apparently unrelated families, 34 were novel including 14 amino acid substitutions. Eleven of them were stratified by the CRM status.
Dysfunctional FVIII protein associated with CRM + phenotype is an uncommon cause of hemophilia A (5%) [37,38]. We report here on two mutations (Trp382Gly and Asp569Tyr), which displayed a typical CRM + phenotype (Table 2, Fig. 2). The other three mutations (Pro521Tyr, Gly701Ser and His1961Pro) were associated with moderately decreased FVIII antigen, but the activity was reduced to a much greater extent, suggesting a dysfunctional FVIII molecule (Table 2). Except for His1961Pro, the substituted residues are proposed to interact with, or they occur at or in the vicinity of, the loops within the FVIII molecule, which are in direct contact with FIXa in the tenase complex. Surprisingly, His1961Pro, which is buried in the groove between A3 and C1 domain, also displayed dysfunctional phenotype. The mechanisms leading to different phenotype as compared with Tyr1961  are unclear.
CRM reduced phenotype
Four novel CRM reduced amino acid substitutions included Leu307Pro, Glu1885Asp, His2155Asp and Leu2210Phe. A decreased level of functional FVIII in circulation may result either from its decreased secretion due to misfolding, from decreased stability or from loss of the protective interaction of FVIII with VWF. Substitution Leu307Pro is adjacent to Phe309, which was shown to play a crucial role in the FVIII–immunoglobulin-binding protein interaction . The loss of two stabilizing intramolecular bonds in Asp1885 mutant FVIII suggests misfolding and possibly poor secretion. On the other hand, the altered surface charge in the Asp2155 FVIII mutant is consistent with loss of protein–protein interaction. Jacquemin et al. has shown in plasma from patients and by expression studies, that substitutions Ile2098Ser, Ser2119Tyr, Arg2150His and Arg2150Cys, are associated with decreased VWF binding . His2155 is located on the same surface as the above-mentioned residues. We therefore propose that His2155 is involved in FVIII/VWF binding and that the Asp2155 mutant may have a decreased affinity to the VWF. Testing of the hypotheses was beyond the scope of this study.
Undetermined CRM status
Cys310 was proposed to form a binding site for a copper ion, which plays an auxiliary role in FVIII heterodimer reconstitution studies and is likely to contribute FVIII specific activity . It would have been of interest to compare the results obtained from expression studies  with plasma from a hemophiliac carrying Cys310Phe substitution, but the patient had died. The causality of the substitution His1047Tyr remains unclear since His1047 is not conserved in porcine FVIII B domain, which itself is not essential for FVIII function. Although the intron 1 inversion was excluded in the patient, a possibility remains that a causal mutation had escaped M-CSGE detection. On the other hand, the B domain plays an important role in FVIII biosynthesis  and six different substitutions at conserved as well as non-conserved residues of the B domain have already been reported to the HAMSTeRS database .
Recurrent and de novo mutations
The frequency of recurrent mutations in our cohort was 28% (21/76). Six mutations were recurrent in 21 apparently unrelated families. A non-CpG transition 6685C > T (Leu2210Phe) was detected in five families who share the same haplotype, suggesting a founder effect. Families carrying 5303G→A and 5399G→A transitions differed in their respective haplotypes.
The presence of 19 mutations in DNA isolated from the peripheral blood leukocytes derived from mothers of sporadic hemophiliacs was tested by CSGE and direct sequencing. Three mutations (marked ‘d’ in the Tables 2, 4 and 5) were not detected, suggesting de novo mutations. It has been demonstrated by others that somatic mosaicism in hemophilia A may be a common event, particularly in the case of CpG transitions . Our techniques were not sensitive enough to exclude maternal mosaicism, but none of the proposed de novo mutations involves CpG transition.
In conclusion, we reported on a modified screening technique M-CSGE, which allowed an effective detection of mutations in 74 unrelated families with intron 22 inversion negative hemophilia A. Thirteen novel amino acid substitutions were further studied by molecular modeling and correlations with phenotype were performed. Molecular modeling suggested mechanisms by which substitutions at residues 382 and 569, located outside the proposed FIXa-binding region, may influence the FVIII–FIXa interaction. His2155 was predicted to participate in FVIII/VWF binding.
The authors are grateful to Drs H. Fischlova, M. Matyskova, S. Kralova, J. Lastuvkova, E. Rimanova, P. Smejkal, J. Sulovska and I. Votrubova for referring their patients to the study. The authors appreciate technical advice from I. Williams and control samples provided by Dr A. Goodeve. The authors are delighted to thank A. Janouskova, I. Zelezna, M. Zemanova, H. Novotna and J. Dudlova for outstanding technical assistance. The authors thank Drs I. Hrachovinova and P. Turek for careful reading of the manuscript. The graphic works of G. Misarova and J. Buriankova was much appreciated. This study was supported by a grant from IGA MZCR: NE/6062–3.