Genetic Variants Associated with von Willebrand Factor Levels in Healthy Men and Women Identified Using the HumanCVD BeadChip


Corresponding author: Steve E. Humphries, Centre for Cardiovascular Genetics, Department of Medicine, University College London, 5 University St, London WC1E 6JF. Tel: 020-7679 6962; Fax: 020-7679 6212; E-mail:


We have used the gene-centric Illumina HumanCVD BeadChip to identify common genetic determinants of Von Willebrand factor (vWF) levels in healthy men and women.

The Whitehall II (WHII) study (n= 5592) and the British Women's Heart and Health Study (BWHHS) (n= 3445) were genotyped using the HumanCVD BeadChip. Replication was conducted in the British Regional Heart Study (n= 3897) and 1958 Birth Cohort (n= 5048).

We identified 48 single nucleotide polymorphisms (SNPs) in four genes/regions associated with vWF at P < 10−4. These included 19 SNPs at the ABO blood group locus with the lead variant being rs657152 (P= 9.7 × 10−233). The lead variant in the 24 VWF SNPs was rs1063856 (P= 2.3 × 10−20). SNPs at ESR1 (rs6909023) and NRG1(rs1685103) showed modest associations with vWF, but these were not confirmed in a meta-analysis. Using variable selection, five SNPs at the locus for ABO and two for VWF were found to have independent associations with vWF levels. After adjustment for age and gender, the selected ABO SNPs explained 15% and the VWF SNPs an additional 2% of the variance in vWF levels. Individuals at opposite tails of the additive seven SNP allele score exhibited substantial differences in vWF levels. These data demonstrate that multiple common alleles with small effects make, in combination, important contributions to individual differences in vWF levels.


Von Willebrand factor (vWF) is a multimeric cystine-rich circulating glycoprotein, with a mature subunit of 2050 amino acids, synthesised in endothelial cells and platelets. vWF is involved in the binding of platelets to endothelium and the formation of platelet plugs at sites of vascular injury. It also functions as a carrier protein for clotting factor VIII. Von Willebrand disease, a bleeding disorder of varying degrees of severity, arises from a deficiency in vWF. vWF is a risk factor for adverse cardiac events (Spiel et al., 2008), however, it is uncertain whether it is causally involved in cardiovascular disease.

vWF undergoes extensive post translational modification by sulphation and glycosylation with the same N-linked oligosaccharides that determine the ABO blood group (Matsui et al., 1992). The common alleles of the ABO system code for enzymes that add N-acetyl galactosamine and D-galactose respectively to a common side chain precursor, converting it into the A or B antigens. It has been consistently found that individuals with the O blood group have lower vWF levels on average than individuals with other blood groups, although the mechanism by which this occurs is unclear. Some studies (Moeller et al., 2001) have suggested that the ABO antigens influence the rate of proteolysis or clearance of the vWF molecule.

Apart from the ABO blood group, and possibly ADAMTS 13 (Majerus et al., 2005), the obvious gene for having a major involvement in determining plasma levels of vWF is the VWF gene itself located on chromosome 12 at 12p13.2 (Ginsburg et al., 1985). The gene comprises 52 exons and spans approximately 80 kb. As well as a number of point mutations and deletions identified in patients with vWF deficiency, a number of common polymorphisms have been studied, particularly several in the promoter region which, because of high linkage disequilibrium (LD) form a common haplotype. The most frequently studied is a C > G change at—1793 in the VWF 5’ flanking region, with the G allele being associated with higher plasma vWF levels (Harvey et al., 2000).

Smith et al. reported a meta-analysis for five genome-wide association studies—the CHARGE consortium with vWF levels (Smith et al., 2010). These studies imputed 2.5 million SNPs and identified a number of loci associated with vWF levels. We looked for SNPs in the HumanCVD BeadChip (also known as the IBC chip) that were proxies (based on highest r2) for the reported ones using the HapMap data (Rel 27, 2009) to check if they come out significant in our study. Where r2 is low in the table it is because no other SNPs with higher r2 were available on the HumanCVD BeadChip. A table with these results is presented in the Supplementary Table S6.

Our study utilises a chip that contains 48,000 SNP markers targeting candidate genes associated with cardiovascular disease and related phenotypes. These capture genetic diversity in more than 2000 genes associated with cardiovascular disease. The HumanCVD BeadChip was chosen because it contained many SNPs covering a number of the genes of major interest in determining vWF levels namely the VWF and the ABO gene loci. The aim of this study was to use this cardiochip to look for genetic variants associated with vWF levels in the Whitehall II sample of healthy men and women. This dense coverage enabled us to capture in detail the extent of the locus contribution to determining vWF levels. To increase power to detect modest effects, we have combined this analysis with data using the same chip from the British Women's Heart and Health Study (BWHHS) (Lawlor et al., 2003) and additional data from the 1958 Birth Cohort (Power & Elliott, 2006) and the British Regional Heart Study (Walker et al., 2004).


Cohorts and Measurement of Circulating vWF

Whitehall II study (WHII)

The Whitehall II study recruited 10,308 participants (70% men) between 1985 and 1989 from 20 London-based Civil Service departments (Marmot et al., 1991; Marmot & Brunner, 2005). Blood samples for DNA were collected in 2002–2004 from over 6000 participants Details of the design, participants, lipid measurements and genotyping of WHII are presented in Supplementary Table S1. Additional genotyping for an ABO SNP rs8176719, which determines the O and non-O blood groups was carried out by KBioscience using KASPar (Cuppen, 2007). Plasma vWF levels were measured in fasting blood samples taken at phase 3 of follow-up (Marmot & Brunner, 2005) and plasma samples were snap frozen and stored at –80°C until measurement with a double-antibody ELISA (Dako Ltd., High Wycombe, UK) using standards from the National Institute for Biological Standards and Control (Potters Bar, UK). The same standard pool of citrated, platelet-poor plasma from 120 donors was used throughout the study as a control calibrated against the fourth British Standard for Blood Coagulation Factors, Plasma Human 89/592 (established 1990) from the National Institutes for Biological Standards and Control. The coefficient of variation for the measurement of vWF levels was 16% (Kumari et al., 2000). vWF concentration was expressed in international units per deciliter.

British Women's Heart and Health Study (BWHHS)

BWHHS is a prospective cohort study of older British women. Between 1999 and 2001, 4286 women, aged 60–79 years, were randomly selected from 23 British towns. Methods used at baseline assessment have been previously described (Lawlor et al., 2003). Blood samples were taken after a minimum 8-h fast and immediately spun and snap frozen. They have been stored at –80°C since study initiation. Samples were defrosted for completing assays of vWF after a median storage time of 4 years and without any previous thawing. Plasma vWF antigen levels were measured using an in-house enzyme linked immunosorbent assay (ELISA), employing rabbit anti-human polyclonal antibodies obtained from DAKO Ltd, and the inter- and intracoefficients of variation for the measurement of vWF levels were 3.3% and 4.2% (May et al., 2007).

British Regional Heart Study (BRHS)

In 1978–1980, 7735 males aged 40–59 were randomly selected from general practice registers in each of 24 British towns and invited to a screening examination (response rate 78%). In 1998–2000, 4252 (77% of survivors) attended a further examination. Nurses administered questionnaires, made physical measurements, recorded an electrocardiogram (ECG) and collected fasting whole blood samples, which were stored at −70°C. Plasma levels of vWF antigen were measured with ELISA (Dako, High Wycombe, UK) as described above for BWHHS (Wannamethee et al., 2009).

1958 British Birth Cohort (1958BC)

The British 1958 Birth Cohort (1958BC) (Power & Elliott, 2006) is a prospective birth cohort, where for a week in the year all births (some 17,000) in England, Wales and Scotland were surveyed. Follow-up of study participants have been done regularly with information collected on a wide-range of factors related to health, lifestyle, growth and development. The biomedical survey was conducted when participants were aged 44–45 years over the period August 2002 to March 2004. Venous blood samples were obtained without prior fasting and posted to collaborating laboratories: vWF antigen was measured in stored plasma in Glasgow as described above for BWHHS (Rudnicka et al., 2007). DNA was extracted from 8000 participants.

Genotyping and quality control

Full details of the genotyping and quality control have been published previously (Talmud et al., 2009). In summary, DNA from WHII was extracted from 6156 individuals from whole blood samples using magnetic bead technology (Medical Solutions, Nottingham, UK) and normalised to a concentration of 50 ng/μl. We used custom SNP arrays designed by the Institute of Translational Medicine and Therapeutics, the Broad Institute and the National Heart Lung and Blood Institute supported Candidate-gene Association Resource Consortium (HumanCVD BeadChip) (Keating et al., 2008) on 5592 of these samples. After quality control we were left with 5059 samples for further analysis.

For the BWHHS, DNA was extracted from K-EDTA whole blood samples using a salting-out procedure (Miller et al., 1988). Genotyping was successfully performed on 3445 of these samples using the same custom SNP array described above for WHII. The following “hard” limits for Illumina BeadStudio parameters were applied to data for WHII and BWHHS: cluster separation <0.3, call frequency < 0.95, AB R mean < 0.3, AB T mean <0.2 or > 0.8, heterozygote excess < –0.3 or > 0.1. SNPs at the borderline of these limits for each parameter were then checked and either reclustered or discarded as necessary. Principal components analysis confirmed self-reported ethnicity and 32 individuals were excluded on the basis of non-European ancestry, leaving 3413 samples for analysis.

The meta-analysis was carried out on a total of 8504 individuals. SNPs with minor allele frequency (MAF) > 0.001 and call rates greater than 0.98 were included in each of the two studies, with resulting genotypes available for each study: WHII equal to 38,477 SNPs, BWHHS equal to 36,530 SNPs and combined equal to 36,087 SNPs. The genomic inflation factors for all the analyses reported were close to 1 indicating negligible influence from population structure or differential genotyping errors.

Statistical analyses

The initial association analysis was carried out on WHII data. Results from this were later combined in a meta-analysis with data from the BWHHS (a total of 8504 individuals). In both studies, vWF was natural log-transformed to improve normality, and was expressed in study-specific SD units in order to overcome the difference in trait distributions between the cohorts. For the association analysis, we applied an additive genetic model in PLINK (Purcell et al., 2007) with and without age (and sex) adjustments, and used the corresponding P-values to construct quantile–quantile and Manhattan plots (Supplementary Figs. S1a and S1b). Analyses were not adjusted for use of lipid-lowering drugs as this has previously been found not to influence vWF levels (Kumari et al., 2000). We combined the WHII and BWHHS results in a fixed effects model meta-analysis of individual study beta-coefficients, weighted by the inverse of the variance, using the metafor library in R (The R Project for Statistical Computing, version 2.11, 2009). There was no evidence that prevalence of hypertension, smoking or alcohol intake was different by any of the 7 SNPs used in this score (Supplementary Table S2). The number of observations for use of lipid-lowering medication and diabetes was very low (< 1%), therefore we did not have enough power to calculate significant differences between the genotype classes. No adjustment for these potential confounders was carried out.

We report everything significant with P-value < 10−4. This is more liberal than the Bonferroni correction of 10−6 corresponding to 5% experiment wide Type I error for the 50,000 tests performed, and under the global null hypothesis of no association we would expect five SNPs to be declared as significant using this threshold. The significance threshold thus represents a trade-off between avoidance of false positive associations while taking into account the likely higher prior odds of association because of the nature of the array.

The CVD Chip does not include SNP rs8176719, which distinguishes most accurately the “O” blood group from “non-O,” and this SNP was determined additionally. This SNP was in strong LD (r2 0.98) with rs657152, which is in the chip, as shown in Supplementary Table S3 and Supplementary Figure S2. The minor allele frequency and the effect size associated with both SNPs are identical in these data, and therefore rs657152 was used as a proxy for the “O” and “non-O” genotype.

Independence of association signals To assess the best genetic predictors at each locus, a variable selection algorithm was implemented as follows. SNPs from two genes ABO and VWF were filtered according to MAF > 0.001. The set of markers thus obtained was augmented for any missing genotype data using the software fastPHASE (Scheet & Stephens, 2006) and a stepwise selection scheme using the Akaike's Information Criterion (AIC) (Akaike, 1974) was implemented separately for each chromosome. The genetic model assumed an additive effect on the appropriate scale, adjusting for gender and age as in the univariate analyses.

Effect size We first assessed the proportion of the trait variance explained by SNPs both singly and in combination. The estimates for the proportion of the total trait variance explained (r2) were obtained by the regression of the SNP on the trait of interest after adjustment for covariates. To obtain the r2 for all loci showing significant association with the trait we used SNPs selected from the stepwise regression (variable selection) and a linear model was fitted between age, gender and the trait. The residuals of this model were then used to obtain the percentage of residual variance in the trait explained by all the SNPs characterising a single gene or a gene cluster. We also assessed the effect size as the beta coefficient of the linear regression of individual SNPs on trait level using a per-allele model. Finally, to assess the joint effect of carriage of multiple independently associated SNPs, we developed a gene score and computed the odds of lying within the 10% tail of the risk factor distribution, based on the quartile position in the frequency distribution of gene score values. To calculate the gene score, the SNPs remaining in the model after variable selection were recoded with the allele having the detrimental effect as 1 and the other as 0, and each individual was assigned a score of 0, 1 or 2 according to their genotype for each SNP. The gene score was then calculated by summing the scores across all selected SNPs for each individual.


The anthropometric characteristics and lipid and lipoprotein profiles of the 5059 white men and women from WHII and the 3413 women from the BWHHS after quality control are presented in Table 1. Mean vWF levels were not significantly different between men and women in WHII but were considerably higher in the BWHHS (P < 0.0001). Since the absolute vWF levels are significantly different between the two cohorts, probably in part due to the 20 years age difference (Kadir et al., 1999), we chose to conduct pooled analyses on a standardised (study-specific mean = 0, standard deviation = 1) rather than absolute scale.

Table 1.  Lipid and anthropometric characteristics of the Whitehall-II participants taken at phase 3, and the British Women Heart and Health Study at baseline
Men (n= 3720)Women (n= 1338)Women (n= 3445)
  1. *Median (IQR).

Age (yrs)49.065.9449.596.1068.875.50
BMI (kg/M2)25.033.0825.314.7027.584.94
Hip (cm)96.886.1497.039.53105.1010.35
Waist (cm)87.159.1674.4111.5986.2112.06
DBP (mmHg)80.728.9276.149.3179.4111.87
SBP (mmHg)121.5212.77116.6413.65147.0225.12
Chol (mmol/l)6.451.116.431.176.641.21
LDL-C (mmol/l)4.430.994.1991.074.151.08
ApoB (g/L)1.300.291.190.30NANA
TG (mmol/l)1.551.201.130.691.871.24
HDL-C (mmol/l)1.330.351.720.431.660.45
ApoA1 (g/L)2.060.322.350.38NANA
vWF IU/dl104.639.0104.938.5147.6443.87
vWF IU/dl*98(77–125)97(78–123)144(114–176)

SNP Associations with vWF

In the combined data set of WHII and BWHHS, a total of 48 SNPs with a MAF > 0.001 and Hardy-Weinberg P-values > 0.0001 were found to be associated with vWF levels at the prespecified significance level of P < 10−4 (Table 2 and Fig. 1). A majority were located at two loci (Supplementary Fig. S3), 24 SNPs being on chromosome 12 in the region of the VWF gene and 19 on chromosome 9 in the region of the ABO gene. Three SNPs in NRGI on chromosome 8 and two in the ESR1 gene on chromosome 6 exhibited weak association with vWF in Whitehall II and BWHHS. The lead SNP at each locus was analysed for association with vWF levels in two additional studies: British Regional Heart Study and 1958 Birth Cohort (n= 16,041 subjects, see Supplementary text for details). The two SNPs from NRG1 and ESR1 were chosen based on minimum P-values. However, while P-values for the two SNPs in ESR1 were very similar (Table 2), rs6909023 (ESR1) with larger P-value was chosen based on genotype availability in the cohorts. As shown in a meta-analysis in Supplementary Figure S4, while overall the effect associated with each SNP was statistically significant (NRG1 rs1685103 beta =–0.06 [–0.102 to –0.021, P= 0.003(0.02)], ESRI rs6909023 beta =–0.098 [–0.160 to –0.037, P= 0.002(0.03)], this was mainly driven by the WHII study, and after exclusion of the WHII data neither effect was statistically significant (not shown).

Table 2.  SNPs associating with vWF levels in the combined WHII and BWHHS that reached P < 10−4 in the meta-analysis using the fixed effects model. Variable selected SNPs are shaded
SNP descriptionWHII (n= 38,477)BWHHS (n= 36,530)Meta analysis (n = 36,087)
  1. n= number of individuals, QEp =P-value for heterogeneity.

Figure 1.

Manhattan plots showing the associations of 48,032 SNPs by chromosome for vWF levels versus–log10P-value, in the combined WH-II and BWHHS data. The horizontal line indicates a P-value threshold of 10−4. The quantile–quantile plots for the test statistics of the observed association P-values plotted as a function of the expected SNP— association P-values are inset.

Associations at the ABO and VWF Loci

At the ABO locus the most significant association was with rs657152 (P= 9.7 × 10−233), with 16 other SNPs having significance values of P < 10−20. For VWF the most significant association was with rs1063856 (P= 2.3 × 10−20), and 10 other SNPs also had effects of P < 10−10.

Refinement of Association Signals Using Variable Selection

In the presence of multiple signals of association in regions of LD, we undertook a stepwise regression analysis using the AIC to identify independently predictive SNPs for each associated gene or region in the WHII data. For the ABO locus, five SNPs were retained (rs657152, rs651007, rs8176722, rs512770, rs638756) and for VWF two SNPs (rs1063856, rs216320). Conditional P-values are in Supplementary Table S4. All SNPs retained by variable selections are individually highly significant, with meta-analysis P-values < 10−16.

Effect Size

In the WHII sample, age and gender explained ∼2% of the sample variance in vWF levels, and after adjustment for these factors, the proportion of the variance explained by each single SNP was typically <5%. However, the overall variance jointly explained by the common associating SNPs was 15% for ABO with an additional 2% for VWF. The SNP having the largest individual effect was the ABO SNP rs657152 that distinguishes “O” from “non-O.” As shown in Supplementary Figure S5 the “non-O” group subjects consistently had 22%–30% higher vWF levels than “O” subjects. For VWF, there was a 14 IU/dl mean difference in vWF levels between homozygous individuals for the SNP with the strongest effect rs1063856, (Supplementary Table S5). In our study the A2 tagging SNP rs8176704 did show a non-significant trend across vWF levels by genotype for WHII (84.2, 103.2, 105.0 IU/dl, P= 0.02) and for BWHHS (143.7, 145.5, 147.9 IU/dl, P= 0.48) possibly pointing to a minor secondary influence of this allele.

We derived VWF-specific gene scores comprising five SNPs for ABO and two for VWF. The gene score distribution and effect on vWF levels are shown in Figure 2. Only 10% of individuals in the lowest quartile of the gene score distribution (1–5 raising alleles) were in the upper tertile of vWF levels (> 114 IU/dl), while 45% of subjects in the top quartile of the score were in the top vWF tertile; similar results were seen in BWHHS (Table 3).

Figure 2.

Frequency distribution of the gene count score for VWF (a) WH-II, (b) BWHHS. The fitted line represents the proportion of subjects in the top tertile of vWF levels, i.e., the effect of gene score on trait level, with increasing score. Below each gene score histogram is a plot of the odds ratio for being in either the top or bottom 10% of the trait distribution at different cut points of the respective gene score.

Table 3.  Distribution of vWF and gene score
StudyGene score distribution (Number of raising alleles)vWF (%)
Lower tertileMiddle tertileUpper tertile
  1. For WHII P-value = 1.1 × 0−11, for BWHHS P-value = 2.4 × 10−7.

WHII0–25% (1–5 alleles)392610
26–50% (5–6 alleles)402114
51–75% (7–8 alleles)133131
76–100% (8–12 alleles)82245
BWHHS0–25% (1–6 alleles)473419
26–50% (7–7 alleles)373429
51–75% (8–9) alleles243342
76–100% (10–12) alleles133254


This study reports a meta-analysis across two British cohorts, totalling 8504 European British participants for 36,087 SNPs, with relevant replication tests in a further 8945 participants. The study confirms and extends known associations of genetic polymorphisms in genomic regions encompassing VWF and ABO with vWF levels, for example, a recent study performed by Smith et al. (2010). The density of SNPs representing these regions allows insight into likely causal sites for these effects. The combined effect of five SNPs at the ABO locus explained 15% of the sample variance in the WHII study, with two SNPs at the VWF locus explaining an additional 2%. Two further novel potential loci influencing vWF levels, ESR1 and NRG1, were identified and subjected to replication analyses. However, their effects were very small and did not convincingly replicate when combined with three extra cohorts, nor were they reported in the GWAS conducted by Smith et al. (2010).

vWF Levels and Gene Score

Although the sample has limited power to detect gene–gene interactions, we found no evidence that the combined effects of the ABO and VWF SNPs on vWF levels were greater than additive. By combining five SNPs for ABO and two for VWF identified by variable selection we derived a VWF gene score, and estimated the proportion of individuals in the upper tertile of vWF levels (> 114 IU/dl). Whincup et al. (2002) reported a meta analysis where individuals with vWF levels in the top tertile had an odds ratio of CHD of 1.5 (95% CI 1.1–2.0). In WHII, 10% of individuals in the lowest quartile of the gene score distribution (4–9 raising alleles) had vWF levels in the top tertile, while 65% of subjects in the top quartile of the score (12–16 raising alleles) had levels above this cut-off, while in BWHHS these figures were respectively 17% and 51%. However, the ABO blood group, which has been a primary target for genetic association analyses for many decades, has shown no clear evidence of CHD association. The score described here and its components may therefore be useful in further testing whether there is a causal relationship between vWF levels and CHD. We do not have sufficient CHD cases to explore the effect of the SNPs or the gene score on risk of CHD in a way that is statistically robust. This sort of analysis would require several thousand cases and controls and will be the subject of future studies.

VWF Locus

Of 24 SNPs at the VWF locus associated with vWF levels at P < 10−4, the most significant effect by orders of magnitude was with rs1063856 (P= 2.3 × 10−20). rs1063856 codes for a Threonine to Alanine amino-acid change at codon 789, with subjects homozygous for the Alanine variant having vWF levels roughly 13% higher than subjects homozygous for the Threonine allele (Supplementary Table S5), as has been reported previously (Lacquemant et al., 2000). This polymorphism is located in the domain that binds factor VIII, but it is a relatively conservative amino acid difference, which is not predicted to be damaging using the PolyPhen program ( In addition, the expression in COS cells of recombinant vWF encoding the Alanine or Threonine variants showed no difference in vWF expression level nor its interaction with FVIII (Kroner et al., 1996), suggesting that the SNP is most likely to be a nonfunctional marker for a functional variant elsewhere in the gene that is affecting expression levels.

Considering the very large size and range of functional domains of VWF, and the wide mutational spectrum and diverse phenotypic forms of von Willebrand disease, it might be anticipated that more than one common polymorphic SNP effect might also be associated with this gene. Following variable selection, only one other of the 24 VWF SNPs associated at P < 10−4 remained in the combined model. This further reinforces the primacy of rs1063856 as candidate for causality considering the relatively dense SNP coverage. The other SNP, rs216320 (conditional analysis on rs1063856, P= 8.4 × 10−8) is in intron 20 in a region of LD also encompassing exon 18 (Laird et al., 2007). In Europeans, its minor allele frequency is low compared with that of rs1063856 (0.09 vs. 0.342 in HapMap CEPH Europeans, 0.09 vs. 0.37 in WHII and 0.09 vs. 0.39 in BWHHS). Taken together, these data further implicate the important functions encoded in this region of LD representing the D2/D’/D3 domains of the protein. These two SNPs may jointly tag an undiscovered, causal SNP. LD between the two SNPs is low (r2= 0.025, D’= 0.688 for HapMap CEU, and r2= 0.05, D’= 0.94 for WHII), as well as conditional P-value P= 8.4 × 10−8 in WHII, further suggesting that they may mark independent effects. rs216320 appears to be in complete LD with other intronic SNPs such as rs216339 and rs216335, and further functional inference is limited at this stage.

ABO Locus

The highly statistically significant association of vWF levels with the ABO region reaffirms the well-established observation (Gill et al., 1987) that vWF levels are on average 25% to 30% lower in blood group O individuals compared with groups A and B. While the earliest studies used ABO phenotyping that cannot distinguish AA from AO or BB from BO, ABO genotyping can resolve the six combinations. The HumanCVD BeadChip contains rs657152, which is in very high LD (r2= 0.94) with rs8176719, a single base deletion representing the O blood group. rs657152 showed the strongest statistical association with vWF levels, consistent with a main effect determined by the absence of the glycosyltransferase function included by the ABO locus. Thus the study confirms the major effect on vWF level deriving from O/non-O status.

The BeadChip also contains rs8176704 that represents the A2 subgroup of A alleles. This showed no statistically significant effect on vWF levels. While rs8176746, representing group B (exon 7 M266L) is not in the BeadChip, rs8176722 that is in HapMap CEU and shows an r2 of 0.818 with rs8176746, is present, and rs8176722 showed an association with vWF levels at P∼10−17, and was retained in the model with rs657152 following variable selection. While rs8176722 is in modest LD (r2= 0.102) with SNPs marking the O allele, its retention alongside the group O marker suggests that there might be a minor effect related to group B alleles. In variable selection, several additional ABO SNPs other than rs8176722 (rs651007, rs512770 and rs638756) also remained in the final model with rs657152. These may reflect a greater phenotypic spectrum of effect than yet resolved, contingent on the considerable allelic complexity of the ABO locus beyond the basic O, A and B alleles. The effect size of these SNPs is 15%, of which rs657152 accounts for 1%.

Previously Reported Associations and Functional Candidates

The transcriptional control of expression of the VWF gene has been identified by previous studies to be located in the promoter region –487 to +247 base pairs, which is an endothelial-specific promoter (Ardekani et al., 1998) and in a DNAase 1 hypersensitive region in intron 51 of the VWF gene that appears to confer specificity for expression in lung and brain vascular endothelial cells (Kleinschmidt et al., 2008). While no common SNPs have been identified in either of these regions, several studies have identified association between SNPs in the promoter region at position—1793 (rs7965425), which tagged the haplotype of several other nearby SNPs (Keightley et al., 1999). Although one of these SNPs was included in the HumanCVD BeadChip (rs7965425, MAF < 0.0001) our data suggest that these variants are not themselves the functional variants determining VWF expression and that other sequence changes remain to be found, the LD structure of VWF including these SNPs is in Supplementary Figure S2.

The study by Smith et al. (2010) identified the following loci associated with vWF levels with P < 5 × 10−8: 6p24 (STXBP5), 8p21 (SCAR5), 12q23 (STAB2), 12q24.3, (STX2), 14q32 (TCN2) and 19p13.2 (CLEC4M) summarised in Supplementary Table S6. With the exception of SNPs in ABO and VWF, there were no SNPs in the HumanCVD BeadChip that tagged the reported associations from Smith et al. (2010) (r2 > 0.25); we therefore can neither confirm nor refute these reported loci from our analysis.

Van Schie et al. (2010) report an association study for 421 acute CHD patients and 409 healthy controls using tagging SNPs in the VWF gene. They report a novel association between SNP rs4764478 and vWF levels. Even though this SNP is included in the HumanCVD BeadChip, in our data, rs4764478 showed no significant association: WHII, P= 0.21, BWHHS, P= 0.16, and meta-analysis of both studies, P= 0.15. Our work differs from that of van Schie et al. as their study comprised 421 cases ascertained on CHD, whereas our study was carried out on a large number of healthy individuals.

Strengths and Limitations

Like other association studies, our study has limitations. The IBC HumanCVD BeadChip covers approximately 2000 genes related to cardiovascular traits, with added coverage in high priority CVD-related loci including VWF and ABO known to influence vWF plasma levels. Our conclusions are restricted to the scope of the IBC chip, and there are likely to be additional genes missing from the chip that also impact on levels of vWF. However, our use of variable selection and the development of an allele score comprising SNPs from both ABO and VWF provide further evidence that carrying multiple common trait-modifying alleles can make an important contribution to the individual differences in vWF levels.

Since both the main studies used for the analysis of the relationship between genotype and vWF levels were carried out in fasting blood samples, and samples from BWHHS, the British Regional Heart Study and the British Birth Cohort were all determined in Glasgow by Professor Gordon Lowe and his colleagues using the same method, we do not believe that differences in blood sampling is a potential confounder.

In summary, we report a set of SNPs marking effects on vWF levels. Von Willebrand disease is a bleeding disorder displaying a heterogeneous phenotypic spectrum, ranging from severe mutations through milder mutations to asymptomatic states. Apart from known risk factors influencing penetrance and recognition, such as age and menstruation, these SNPs are all likely also to impact on penetrance in a fashion dependent on “allele score.” Such a score will also be relevant to the likely influence of vWF and factor VIII on arterial and venous thromboembolism, and plausibly to other disease risks in which vWF level may play a part.


WHII: Professor Humphries is a British Heart Foundation (BHF) Chairholder. The UCL Genetics Institute supports DZ, and SS. The work on WH-II was supported by the BHF PG/07/133/24260, RG/08/008, SP/07/007/23671 and a Senior Fellowship to Professor Hingorani (FS/2005/125). Dr. Kumari's and Professor Kivimaki's time on this manuscript was partially supported by the National Heart Lung and Blood Institute (NHLBI: HL36310). The WH-II study has been supported by grants from the Medical Research Council (MRC); British Heart Foundation; Health and Safety Executive; Department of Health; National Institute on Aging (AG13196), US, NIH; Agency for Health Care Policy Research (HS06516); and the John D and Catherine T MacArthur Foundation Research Networks on Successful Midlife Development and Socioeconomic Status and Health. BWHHS: The British Women's Heart and Health Study has been supported by funding from the BHF (PG/07/131/24254) and the UK Department of Health Policy Research Programme. We thank all of the participants and the general practitioners, research nurses and data management staff who supported data collection and preparation. The BWHHS is co-ordinated by Shah Ebrahim (PI), Debbie A Lawlor and Juan-Pablo Casas, with HumanCVD BeadChip work funded by the BHF PG/07/131/24254, PI Tom Gaunt, co-PIs Ian Day, Debbie A Lawlor, Shah Ebrahim, George Davey Smith, Yoav Ben-Shlomo and Santiago Rodriguez. Tom Gaunt, George Davey Smith, Debbie A Lawlor and Ian Day work in a Centre that receives funds from the UK Medical Research Centre (G0600705). BRHS: The British Regional Heart Study is a British Heart Foundation Research Group. The measurements and laboratory analyses reported here were supported by British Heart Foundation Project Grant PG97012. DNA extraction was supported in part by British Heart Foundation Senior Research Fellowship FS05/125. 1958BC: EH is partly funded by Department of Health (UK) Public Health Career Scientist Award. DNA collection and statistical analyses in the 1958BC were funded by the MRC (grants G0000934 and G0601653). WTCCC was funded by Wellcome Trust grant 068545/Z/02. The Type 1 Diabetes Genetic Consortium is a collaborative clinical study sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases, National Institute of Allergy and Infectious Diseases, National Human Genome Research Institute, National Institute of Child Health and Human Development, and Juvenile Diabetes Research Foundation International and supported by U01 DK062418. None of the funders influenced data collection, analysis, interpretation or the decision to publish these findings. The views expressed in this paper are those of the authors and not necessarily those of the funders.