The mutational spectrum of type 1 von Willebrand disease: results from a Canadian cohort study

Von Willebrand disease (VWD) is the most common inherited bleeding disorder known in humans, with prevalence estimates as high as 1%.^1,2  There are 3 main subtypes, classically characterized by excessive mucocutaneous bleeding, a positive family history, and abnormal von Willebrand factor (VWF) laboratory studies. Type 1 VWD is a partial deficiency of qualitatively normal VWF, type 2 VWD is caused by functionally abnormal VWF, and type 3 VWD represents a virtual absence of the VWF protein.

Von Willebrand factor (VWF) is a multimeric glycoprotein that plays an essential role in primary hemostasis. It is synthesized in endothelial cells and megakaryocytes. At sites of vascular injury, VWF is released and binds to the platelet glycoprotein Ib (GP Ib)/GP IX receptor complex to initiate platelet adhesion to the subendothelium. Platelet activation results in the exposure of the GP IIb/IIIa integrin receptor through which VWF and fibrinogen mediate platelet aggregation.³  VWF is also secreted into the circulation where it is bound to and stabilizes factor VIII (FVIII).

Cloning and characterization of the VWF gene, by 4 groups simultaneously in 1985,^4-7  have facilitated our understanding of the molecular genetic basis of VWD. Located on the short arm of chromosome 12 at 12p13.3, the VWF gene spans 178 kb and is composed of 52 exons that range in size from 1.3 kb (exon 28) to 40 bp (exon 50).⁸  The encoded VWF mRNA is 9 kb in length and the translated pre–pro-VWF molecule contains 2813 amino acids (AAs), comprising a 22-AA signal peptide, a 741-AA propolypeptide, and a 2050-AA secreted mature subunit that possesses all the adhesive sites required for VWF's hemostatic function.⁹  There is a partial, unprocessed pseudogene located on chromosome 22 that duplicates the VWF gene sequence for exons 23 to 34 with 97% sequence homology. As well, the VWF gene is highly polymorphic. Both of these factors combined with the large size of the VWF gene make investigating the genetic basis of VWD challenging.

Despite this, both type 2 and type 3 VWD have been relatively well characterized from a molecular perspective. However, an understanding of the genetic basis of type 1 VWD (which represents ∼ 80% of cases) has proved more elusive. Studies of a mouse model of type 1 VWD have shown locus heterogeneity, indicating that loci outside the VWF gene may play a role.^10,11  Nevertheless, the VWF gene itself has remained the primary target of investigation, and a small number of type 1 VWD mutations are currently listed on the ISTH VWF SSC homepage (http://www.vwf.group.shef.ac.uk/index.html) including the dominant-negative mutations C1130F and C1149R and the common missense mutation Y1584C.^12-15 

In this paper, we report the results of the Canadian Type 1 VWD Study. In conjunction with the European Union project (MCMDM-1VWD), these studies provide unique information concerning the molecular pathogenesis of type 1 VWD based on large cohorts of selected families. In addition to elucidating the genetic changes within the VWF gene that are associated with type 1 VWD, these studies also confirm the importance of additional genetic loci in the pathogenesis of type 1 VWD, particularly in the cases where the VWF:Ag level is more than 0.30 IU/mL.

A total of 194 families were submitted for enrollment to the Canadian Type 1 VWD Study from 13 academic health science centers across Canada. After repeat phenotypic analysis at both the referral clinical center and a central study laboratory, 150 families met the initial type 1 VWD diagnostic criteria used in this study. These families had an individual (the index case) with a personal history of excessive mucocutaneous bleeding and plasma levels of VWF:Ag and VWF:RCo between 0.05 and 0.50 U/mL obtained on at least 2 occasions. The absence of a family history for VWD was not an exclusion criterion for this study. Of these 150 families, 10 were excluded because of abnormal VWF multimers in the index case (loss of high-molecular-weight forms and/or abnormal triplet pattern) and an additional 11 families were excluded because of a mean VWF:RCo/VWF:Ag ratio less than 0.60 in the index case. Finally, 5 families were excluded because of the presence of individuals with type 3 VWD and type 1 VWD in the same family, and 1 family was excluded because the index case had a type 2N phenotype. This left 123 type 1 VWD families for analysis (Figure 1) All participants in the study were informed of the experimental nature of the study and gave informed consent. The study was approved by the Research Ethics Board of Queen's University and at each of the source institutions. Whole-blood samples were collected by phlebotomy in both 3.2% sodium citrate (at a ratio of 9:1 vol/vol) and EDTA tubes from the index case and available family members. A plasma sample from the index case was frozen and repeat VWD phenotypic studies were performed in a central laboratory to confirm the diagnosis of type 1 VWD.

Laboratory tests for VWF:Ag, VWF:RCo, and factor VIII coagulant activity (FVIII:C) were performed at the source clinic attended by the patient according to local methods. These tests were repeated on frozen plasma samples at the Clinical Hemostasis Laboratory at Kingston General Hospital (Canada) and all available laboratory results were averaged. VWF:Ag was measured by the IMUBIND VWF enzyme-linked immunosorbent assay (ELISA) kit according to the procedure supplied by the manufacturer (American Diagnostica, Greenwich, CT). The VWF:RCo was measured by platelet aggregometry using freshly prepared, washed normal platelets,¹⁶  and FVIII:C was measured using a one-stage assay.¹⁷  All measurements of VWF:Ag, VWF:RCo, and FVIII:C were made against a non-ABO–matched commercial reference plasma that had been calibrated against the 91-666 or 97-586 World Health Organization (WHO) Plasma Standard. VWF multimers were analyzed by electrophoresis using a 1.6% sodium dodecyl sulfate (SDS) agarose gel followed by electrotransfer to a nylon membrane, and the multimers were visualized using the chemiluminescent visualization kit from Amersham Pharmacia Biotech (Baie D'Urfé, QC).¹⁸ 

A blood sample was obtained from all of the index cases and genomic DNA was isolated from leukocytes using a salt extraction method.¹⁹  Full-length sequencing of the coding region was not performed in all individuals. DNA corresponding to approximately 1300 base pairs of the VWF promoter, exon 1, and exons 18 to 52 (including an average of 68-bp sequence at the splice acceptor site and 98-bp sequence at the splice donor site) were amplified by polymerase chain reaction (PCR) for all index cases. Nucleotide numbering in the 5′ noncoding region follows the Human Genome convention with the “A” of the translational ATG codon being +1 (the first nucleotide in exon 2 of the VWF gene) and all subsequent 5′ nucleotides being assigned negative numbers. For those index cases with no identified mutation in the VWF promoter, exon 1, or exons 18 to 52 and their adjacent splice sites, exons 2 to 17 were also sequenced. Primer sequences are available upon request. Using a DNA thermal cycler (Perkin Elmer Life Sciences, Shelton, CT) a T3 Thermocycler (Biometra, Montreal, QC, Canada), or an ABI GeneAmp (Applied Biosystems, Foster City, CA), DNA was amplified for 30 to 35 cycles of 45 to 60 seconds at 94°C, 45 to 60 seconds at 53° to 65°C, and 45 to 60 seconds at 72°C. The amplified products were purified using QIAGEN QIAquick PCR Purification Kit or Gel Extraction Kit and sequenced directly on an ABI model 373 automated sequencer (Mobix Laboratories, McMaster University, Hamilton, ON, Canada; or The Centre for Applied Genomics, Toronto, ON). The DNA sequences were compared with consensus VWF DNA sequences with the assistance of Vector NTI Suite software (InforMax, Bethesda, MD). When a putative mutation was identified, another template was PCR amplified from the index case, and the opposite DNA strand was sequenced to confirm the sequence variation. As well, all available family members were sequenced for the putative mutation to confirm familial transmission. In index cases with VWF:Ag levels less than 0.20 IU/mL in whom an initial round of sequencing showed no putative mutations, approximately 1300 bp of the VWF promoter and exons 1 to 52 and their adjoining splice sites were resequenced using the alternate DNA strand for analysis. All sequence alignments and every sequencing chromatogram were analyzed independently by 2 technologists.

Any VWF sequence variation that was identified was noted. The ISTH SSC VWD homepage (http://www.vwf.group.shef.ac.uk/index.html) and the published literature were checked to see if the variation had been previously recorded and if the molecular mechanism had been elucidated. When we identified a new sequence variation, we evaluated the likelihood that the change could be pathogenic. If the sequence variation seemed likely to affect the subsequent protein either by causing an amino acid substitution or by affecting an invariant splice site (even if the variation had been previously reported as a polymorphism), that change is reported here as a putative mutation. In some instances, the differentiation between a mutation and a polymorphism cannot be resolved without significant additional investigation, and thus, for simplicity, in this report these variations are referred to as putative mutations. Polymorphisms that seem unlikely to affect either the biosynthesis or function of the protein are reported in summary in this paper.

ABO blood group nucleotide (nt) 261 (rs8176719) and nt 703 (rs8176743) SNPs were genotyped using the 5′ nuclease assay to distinguish classical ABO blood groups.^20,21  Two allele-specific probes were designed for each SNP, using Primer Express 1.5 (ABI) following the manufacturer's guidelines. One probe matched the wild-type sequence and the other matched the mutant sequence. Each probe was labeled at the 5′ end with a fluorescent reporter dye and at the 3′ end with a nonfluorescent quencher with a minor groove binder also present at the 3′ end. The reporter dyes used in this case were 6-carboxyflourescein for the wild type and VIC (ABI) for the mutant. The primer and probe sequences for the rs81767119 and rs8176743 SNPs have been previously published.²²  ABO genotypes were then generated from the results of these SNP analyses.

One hundred twenty-three families were evaluated in this study. There were a total of 387 individuals in these 123 families: 229 affected, 108 unaffected, and 50 of unknown phenotypic status (primarily parents of index cases who have never been tested). The mean number of individuals per family is 3.2, with the most common family structure being a 2-generation family (94 families). Ten families had 3 generations and 19 had 1 generation (either single index case or siblings; Table 1) Eighty-six of the index cases were female and 37 were male. The mean age of the index cases was 24 years (range, 1-65 years; Table 2) The majority of individuals collected for this study recorded their race/ethnic origin as Caucasian (69%). Ten percent recorded French Canadian and 3% recorded Asian/Oriental. There were small numbers of the following ethnicities also recorded in the study: Italian, Egyptian, Moroccan, Peruvian, Argentinean, Iranian, Caribbean, South American, Lebanese, African, Welsh, East Indian, Pakistani, and Hispanic. We had individuals from 3 families who were of mixed ethnic origin (African/Caucasian, Argentinean/Caucasian, and Peruvian/Moroccan).

All index cases gave a personal history of excessive mucocutaneous bleeding as documented by their clinical site hematologist. Among females older than 12 years, 75% (53/71) reported menorrhagia, the most common symptom reported. For all index cases, the next most commonly reported symptom was easy bruising, reported by 62% (77/123). This was followed by epistaxis, reported by 56% (69/123), then postdental procedure bleeding, reported by 26% (32/123). Both excessive bleeding from wounds and postoperative bleeding were reported by 24% (29/123). We are unable to draw conclusions, based on these data, between the relationship of bleeding symptoms and VWF level or mutation.

The mean VWF:Ag level for the index cases was 0.36 IU/mL (range, 0.07-0.50 IU/mL), and the mean VWF:RCo level was 0.34 IU/mL (range, 0.07-0.50 IU/mL). The mean FVIII level was 0.54 IU/mL (range, 0.09-1.36 IU/mL; Table 2). The VWF multimer pattern as judged by 3 independent observers was normal for all index cases and, specifically, showed no significant loss of high-molecular-weight multimers nor any abnormality of the multimer triplet pattern.

As expected, blood group O was overrepresented in the index cases. A total of 77 (62%) of the index cases were blood group O, with 34 (28%) being blood group A. Six (∼ 5%) of the index cases were blood group B and 3 (∼ 2%) index cases were blood group AB. In 3 (∼ 2%) cases, the ABO blood group was not obtained (insufficient sample for analysis). The relative proportion of blood group O in the index cases varies depending on the VWF level. For the index cases with VWF:Ag level of 0.30 IU/mL or less (32 individuals), 16 (50%) are blood group O. This is not significantly different from the frequency of blood group O in the general Canadian population (46%). However, for the index cases with VWF:Ag level more than 0.30 IU/mL (91 individuals), 61 individuals (66%) have blood group O. This difference in the prevalence of blood group O in those index cases with VWF:Ag level more than 0.30 IU/mL and the Canadian population is highly significant (P = .002, Pearson chi-square; Figure 2)

Based on the sequencing strategy that we outlined in “DNA sequencing,” we performed full-length VWF sequencing on 68 index cases. The other 55 index cases had sequencing of approximately 1300 bp of the VWF promoter, exon 1, and exons 18 to 52 (plus intron/exon boundaries) performed only (not exons 2-17). Using these sequencing data, we identified VWF gene mutations in 63% (n = 78) of the 123 type 1 VWD families that were studied. No VWF mutations were identified in 37% (n = 45) of index cases. We were more likely to identify a VWF mutation in index cases with VWF:Ag level of 0.30 IU/mL or less (n = 32). In this group, we found mutations within the VWF gene in 24 (75%) of the index cases compared with 45 (49%) of the index cases with a VWF:Ag level more than 0.30 IU/mL (n = 91) (P = .114, Pearson chi-square). We identified a total of 50 different VWF mutations in the 123 type 1 VWD families (see Tables 3-4 for the complete list; Figure 3) Fourteen of these have been previously reported and 36 are new. Of the 50 mutations identified in this study, the majority (31 or 62%) are missense mutations, changing amino acids in either the VWF mature protomer or the VWF propolypeptide. Eight (16%) mutations involve changes in the VWF regulatory region, and 5 (10%) mutations involve consensus splice site sequences. There are 5 (10%) small deletion or insertion mutations ranging in size from 1 to 6 nucleotides, and 1 nonsense mutation. Although 57 (46%) of the index cases had only 1 mutation identified, 17 (14%) index cases had 2 mutations identified, 2 (∼ 2%) had 3, and 1 (∼ 1%) had 4 and 5 putative mutations. The mutations were found in all regions of the VWF gene from the 5′ promoter sequence (HGVS, nt −2774) to the carboxyl-terminal cysteine knot domain. However, a substantial number (11 or 22%) of the changes were found within exon 28, the largest exon within the VWF gene, representing approximately 14% of the VWF coding region.

Additionally, many previously published polymorphisms within the VWF gene were identified. None of the previously published polymorphisms (except those specifically addressed in this report [ie, Y1584C and R924Q]) were identified at frequencies that were significantly different from the published frequencies. We did identify, however, 15 new sequence variations within the VWF gene. These are reported in Table 5 along with their frequencies in this population.

Although a total of 50 different mutations were identified in this study, 12 of the mutations occurred in multiple index cases. These 12 mutations were found in 63 (51%) of the index cases (Table 6) and, with one exception, these were missense mutations. Of these 12 mutations, 10 have previously been reported and in vitro expression studies have been performed on 6 of the missense mutants. The most common mutation found in our population was the Y1584C missense mutation in the VWF A2 domain. This was found in 18 (15%) families. The pathogenetic mechanism associated with this mutation, in which a new solvent-exposed thiol is produced, likely involves both increased intracellular retention of the mutant protein and an enhanced susceptibility to ADAMTS13-mediated proteolysis.^15,23  Family-based association studies (reported separately)²²  confirm significant association of the 1584 codon variant and the type 1 VWD phenotype. Of interest, many of the Y1584C families were recruited from Quebec, Canada; however, all but 2 (who recorded their ethnicity as French Canadian) recorded their ethnicity as Caucasian. Of note, there are 4 mutations listed in Table 6 (V1229G, N1231T, P1266L, and V1279I) that possibly represent a gene conversion event from the partial pseudogene sequence on chromosome 22. The first 2, V1229G and N1231T, and the last 2, P1266L and V1279I, are always coinherited and are linked (on the same haplotype).

Of interest, 2 of the index cases (V-113 and V-381) who have both the P1266L and V1279I mutations share a common haplotype, and both mutations lie on that chromosome. It is possible that there is a common, shared founder for these 2 families. In contrast, for the other 2 families with the P1266L/V1279I mutations, the mutations seem to have arisen spontaneously as there is not a common haplotype between them. There is not a common haplotype among individuals with the V1229G/N1231T mutations.

Formal linkage and association studies have been performed on our study population and, as mentioned, are reported separately.²²  As explained in our linkage and association publication, the lod scores for individual families cluster reasonably tightly around zero as a result of the structure and relatively small size of the families in this study. We are, therefore, unable to draw any significant conclusions about the VWF mutations identified in either linked or nonlinked families in this study.

In an attempt to assess the phenotypic penetrance of the VWF mutations that we have identified, we have evaluated the presence of the mutant sequences in all available family members whose phenotypic status was known. A systematic and inclusive collection of family samples was not undertaken as part of this project, and thus, the results from this analysis may well be influenced by ascertainment biases. Nevertheless, with this limitation in mind, the following mutations showed complete penetrance (note that both affected and unaffected family members available were for analysis, all affected subjects have the mutation, and all of the unaffected individuals do not have the mutation): the nonsense mutation W642X; the splice site mutations c3378 +1 G>A, c2685 +2 T>G, and c3537 +1 G>A; the frameshift mutation c5180_5181insTT; and 2 missense mutations R1315C and R1205H (Table 7) The mean VWF:Ag level of the index cases in the families demonstrating complete penetrance is 0.23 IU/mL. The mean VWF:Ag level of the index cases in all of the remaining families is significantly higher at 0.38 IU/mL (P < .001, t test).

Within our study population, we had 25 families with both parents available for investigation. In 2 of these families, the mutation identified in the index case was not identified in either parent. In each of these 2 index cases, only one mutation was identified: exon 28 1405 to 1408 del 1 lysine (AAG) in V-31 and c3537 +1 G>A in V-508. It is possible that both of these represent new or spontaneous mutations.

In this study, we had 3 index cases with VWF:Ag levels less than 0.20 IU/mL for whom we did not identify VWF gene mutations (Table 8) After an initial round of VWF gene sequencing that had shown no mutations in these cases, new PCR amplicons were generated and the VWF gene from the promoter through exons 1 to 52 and accompanying splice sites was resequenced off the opposite DNA strand. Once again, after this second round of sequencing, with an independent evaluation of the sequencing chromatograms, no candidate mutations were found in any of these cases. In 2 of these families, the linkage information is not informative, however in 1 (V-433) the lod score is −1.45, suggesting nonlinkage.

The 31 missense mutations found in this type 1 VWD population are disseminated throughout the VWF protein from codon 129 in the D1 domain of the propolypeptide to codon 2647 of the C2 domain in the mature VWF protomer. Eight (27%) of the missense changes occur within the D3 and A1 to A3 domains encoded by exon 28 of the VWF gene, the largest exon in the gene (1.3 kb) and, thus, the largest mutational target. Intra- and intermolecular disulfide linkages are critical for the efficient biosynthesis, secretion, and function of VWF, and this is reflected by the fact that 9 (29%) of the amino acid substitutions involve either the loss or gain of a cysteine residue.

Two of the missense changes are located in the propolypeptide at codons 129 and 576, and 2 of the substitutions in the D′/D3 domain at codons 854 and 924 have previously been associated with the type 2N VWD phenotype in the ISTH Database.²⁴  In the index cases in which these type 2N mutations were found, no additional VWF mutations were found to explain the type 1 VWD phenotype. Finally, missense mutations were found at 9 (18%) arginine codons, which contain the hypermutable CpG dinucleotide sequence.

The influence of these changes on the pathogenesis of type 1 VWD is likely to be complex to unravel. While some of the missense changes are undoubtedly of pathogenic significance, others may be neutral polymorphisms. To date, the characteristics of some of the variant proteins have been studied in vitro, and in silico assessments of the nature of the substitutions can also be evaluated. In this report, we present data from 2 protein comparison algorithms, SIFT (http://blocks.fhcrc.org/sift/SIFT.html)²⁵  and PolyPhen (http://www.bork.embl-heidelberg.de/PolyPhen/)²⁶  (Table 3). While these probabilistic comparisons show agreement about the significance of the amino acid substitutions in some instances (ie, all substitutions involving cysteines are predicted to be deleterious by both algorithms), there is disagreement about the influence of some other changes (ie, P1266L, R2185Q, T2104I, S2497P, and T2647M). Furthermore, there are also instances where substitutions with known pathogenic significance, such as R854Q and R1205H, have been labeled benign by one of the algorithms. Thus, caution must be exercised in interpreting these in silico results.

We identified 8 changes within the 5′ regulatory region of the VWF gene (Table 4, nucleotide numbering as per the Human Genome convention with the “A” of the initiator ATG at the start of VWF exon 2 being +1 and all 5′ nucleotides being assigned negative numbers). In prior studies, regulatory sequence single nucleotide polymorphisms (SNPs) have been identified at nts −2525 (previous numbering −1051), −2659 (previous numbering −1185), and −2708 (previous numbering −1234). Two of these polymorphic sites have been shown to bind nuclear proteins derived from endothelial cells. Of interest, our association studies (reported separately)²²  showed a weak association of the −2659 (−1185) polymorphism and the quantitative traits VWF:Ag, VWF:RCo, and FVIII:C, providing further support for the importance of the role that VWF 5′ sequence elements play in regulating VWF levels. In this study, we report 8 novel regulatory sequence variations: 1 involving a 13-bp deletion mutation 48 nts 5′ of the transcriptional start site (−1522), and the remaining 7 involving single nucleotide substitutions. Two of these substitutions occur within 5 and 22 nts of the previously characterized SNPs at −2525 and −2708, respectively. The functional significance of all these variants requires further study.

In one index case, an intronic change was found 20 nucleotides before the start of exon 38 at c6599-20 A>T. This change appears to be at the splicing branch site invariant “A” nucleotide and could potentially significantly interfere with splicing. The branch site score of the motif containing the “A” (calculated using http://www.genet.sickkids.on.ca/%7Eali/splicesitefinder.html) is strong at 78.4 (minimum branch site score, 65; maximum, 100). The type 1 VWD index case “T” substitution at that position results in the program being unable to identify either the acceptor splice site or the branch site providing reasonable evidence that this change could significantly affect splicing. Further in vitro studies to investigate the influence of this change on VWF splicing are also required.

As mentioned, there were 21 index cases with more than one change identified within the VWF gene. These do not appear to be more severely affected cases, and it is likely, in many instances, that some of the changes listed are polymorphic rather than pathogenic (Table 9)

Type 1 VWD is the most common subtype of the most common inherited bleeding disorder known in humans. Despite this, establishing the diagnosis is often a significant challenge. The mucocutaneous bleeding symptoms characteristic of type 1 VWD are often underestimated or overlooked by both patients and clinicians alike. There are temporal variations in the laboratory parameters used to make this diagnosis, and environmental factors such as estrogens or stress are known to influence these parameters as well.²⁷  The effects of ABO blood group and ethnicity are well described,^28,29  and the phenotype is known to exhibit both incomplete penetrance and variable expressivity, further complicating matters.³⁰  Recently, there has also been considerable debate and attention drawn to the issue of the definition of type 1 VWD, raising the possibility that some of the heterogeneity seen in this condition may be as a result of inaccurate diagnosis or overdiagnosis.³¹ 

Along with the contemporaneous European Union Study, this is the first large cohort study addressing the molecular genetic pathology of type 1 VWD. The original diagnosis of type 1 VWD was made, in every case, in a tertiary care academic center by a hematologist with experience in the diagnosis and management of inherited bleeding disorders. The spectrum of family structures and the range of phenotypes seen are therefore likely to be representative of families regularly referred to tertiary care clinics. We intentionally did not recruit large, severely affected families primarily in an attempt to avoid biasing our results toward highly penetrant disease variants that would represent only a minority of families. This fact must be considered when interpreting and comparing our data with other studies.

The laboratory coagulation phenotypes in the index cases in this study were all consistent with type 1 VWD. In particular, cases in which the VWF:RCo/VWF:Ag ratio was repeatedly less than 0.6, and the VWF multimer patterns (as assessed by 3 independent evaluators) were normal, were excluded as type 2M VWD variants. Preliminary results from the European Union type 1 VWD study have reported a significant number of patients with subtly abnormal VWF multimer patterns, mostly losses of the highest-molecular-weight forms. In the current study, we did not appreciate these changes.

In this study, we systematically investigated type 1 VWD index cases and their families for sequence variation within the VWF gene. Our results support 2 important conclusions. First, sequence variations within the VWF gene are relatively common within a type 1 VWD population, but are clearly not the sole genetic determinants of the type 1 VWD phenotype in some cases. Second, the relative contribution of VWF sequence variation appears to be most important in more severely affected individuals, and the contribution of factors such as ABO blood group becomes more significant in milder cases. Taken together, these conclusions suggest a model to consider when evaluating the spectrum of the genetic pathology of type 1 VWD (Figure 4) In more severe cases, genetic changes are common within the VWF gene and are highly heritable. In milder cases, where heritability is variable, the genetic determinants are more complex and more likely to involve contributions from other genetic and possibly environmental factors.

Despite the large amount of new information derived from this study, it is critical to admit that in many cases, the molecular mechanism by which the genetic changes result in reduced VWF levels has either never been examined or is poorly understood. Thus, while the pathogenic nature of the consensus splice site mutations, the deletion/insertion mutations, and the single nonsense mutation is relatively clear, the large number of missense changes and the variants in the 5′ upstream sequence will require much further study to confirm their functional significance. Even in cases where in vitro expression and evaluation of the variant protein have been performed, results are sometimes conflicting, as for example, in the cases of Y1584C and R924Q. Although other groups have confirmed the Y1584C missense mutation as the most frequent within other type 1 VWD populations (ie, the European population), the exact pathogenic mechanism mediated by this change, and its contribution to the type 1 VWD phenotype, is still debated.^15,23,32  Additionally, the R924Q change is controversial. Although this variation was originally reported as a polymorphism,³³  it is currently listed on the ISTH SSC VWF Database (http://www.vwf.group.shef.ac.uk/index.html)²⁴  as a type 2N VWD mutation and is scored as a benign variant by both of the in silico protein algorithms used in this study. Similarly, the R1315C missense mutation is listed in the VWF Mutation Database²⁴  as either a type 2M or unclassified mutation. However, this substitution is clearly associated with the type 1 VWD phenotype in our population.

Another significant finding in this study has been the documentation of several mutations in type 1 VWD patients that have previously been reported as causing other subtypes of the disease. Thus, 3 index cases were found to possess the R854Q missense mutation that, in its homozygous state, results in type 2N disease, and 3 families were documented with the R1205H Vicenza variant. Of interest, in all 6 of these families either the R854Q or the R1205H mutation was the only apparently pathogenic VWF change identified. The association of these changes with the type 1 VWD phenotype in this population is unclear, although the Vicenza variant is recognized to undergo accelerated clearance from the circulation.

It is possible, and in fact likely, that some of the putative mutations we report in this paper are not pathogenic. Some of these variants may simply be linked to the pathogenic change and act as markers of the disease allele (ie, R924Q). Additionally, in some cases, it is unlikely that there is just one pathogenic change; some of these mutations may be acting in combination with other factors, both within the VWF gene itself and at other sites to cause the type 1 VWD phenotype. The 3 index cases with a VWF:Ag level less than 0.20 IU/mL that have had no changes identified within the VWF gene may provide an opportunity to perform genome-wide scanning in an attempt to identify other important genetic loci influencing plasma VWF levels, particularly the family of V-433 (lod score, −1.45).

Two limitations of our study deserve mention. First, we performed full-length VWF sequencing in only 68 index cases. In 55 index cases, exons 2 to 17 were not sequenced. It is therefore possible that we missed additional mutations within the D1 or D2 regions in these cases. Second, our method of directly sequencing the VWF gene, although widely accepted as the best initial method of genetic investigation in these cases, does restrict our findings to mutations obvious only at the level of genomic DNA (Southern blot analysis was not performed). This method of investigation would not identify small or large gene deletions, or splicing variants not caused by changes at the invariant splice sites. Additional analysis including examination of platelet-derived VWF mRNA is essential in some cases to investigate these possibilities.

In conclusion, our study provides some of the first population-based information of the changes found within the VWF gene in a relatively unselected group of type 1 VWD families. In many cases, changes within the VWF gene are identifiable; however, much additional work is needed to completely understand the contribution that these and other changes make to the type 1 VWD phenotype. These results, considered alongside the results of the European Union Type 1 VWD Study, provide the basis on which an approach to genetic testing for type 1 VWD could begin to be generated.

The authors wish to acknowledge the invaluable contributions of the Canadian Association of Nurses in Hemophilia Care and of Wilma Hopman, who performed the statistical analysis.

This work was supported by a CIHR (Canadian Institutes for Health Research) Operating Grant (MOP-42467) and funds from ZLB Behring. D.L. holds a Canada Research Chair in Molecular Hemostasis and is a Career Investigator of the Heart and Stroke Foundation of Ontario.