Identification of quantitative trait loci for fibrin clot phenotypes: The EuroCLOT study

Abstract

Objectives

Fibrin makes up the structural basis of an occlusive arterial thrombus and variability in fibrin phenotype relates to cardiovascular risk. The aims of the current study from the EU consortium EuroCLOT were to 1) determine the heritability of fibrin phenotypes and 2) identify QTLs associated with fibrin phenotypes.

Methods

447 dizygotic (DZ) and 460 monozygotic (MZ) pairs of healthy UK Caucasian female twins and 199 DZ twin pairs from Denmark were studied. D-dimer, an indicator of fibrin turnover, was measured by ELISA and measures of clot formation, morphology and lysis were determined by turbidimetric assays. Heritability estimates and genome-wide linkage analysis were performed.

Results

Estimates of heritability for d-dimer and turbidometric variables were in the range 17 - 46%, with highest levels for maximal absorbance which provides an estimate of clot density. Genome-wide linkage analysis revealed 6 significant regions with LOD>3 on 5 chromosomes (5, 6, 9, 16 and 17).

Conclusions

The results indicate a significant genetic contribution to variability in fibrin phenotypes and highlight regions in the human genome which warrant further investigation in relation to ischaemic cardiovascular disorders and their therapy.

Introduction

The EuroCLOT consortium is an EU-funded multi-centre collaborative study established to identify QTLs associated with fibrin clot phenotypes to shed light on the pathogenic mechanisms of ischaemic cardiovascular disease (stroke and myocardial infarction). In both myocardial infarction and ischaemic stroke, acute thrombosis is the key final step leading to tissue damage in the territory served by the vessel. Increased levels of haemostatic factors, including fibrinogen¹ and tissue plasminogen activator (tPA)² predict development of cardiovascular disorders and outcome.

Activation of the coagulation cascade results in thrombin generation and cleavage of fibrinogen to form fibrin monomers, which polymerise and are cross-linked by thrombin-activated factor XIII, to form the structural scaffold for thrombus formation. Fibrinolysis of fibrin clots is influenced by various factors including the fibrin binding characteristics of tPA/plasminogen, local concentrations of fibrinolysis inhibitors and the structure of the clot. For example, studies in vitro have shown that dense clots composed of thinner fibres lyse more slowly than less dense clots formed from thicker fibres^{3, 4}. Alterations in clot structure/function, including increased clot density and decreased clot lysis times, have been observed in subjects with arterial and venous thrombosis^{5, 6} although the association with clinical outcome in healthy volunteers is unclear at present. We have recently shown that genetic factors contribute to variance in turbidimetric measures of clot structure/function⁷ and numerous studies have demonstrated genetic influences on proteins involved in coagulation and fibrinolysis. Furthermore, genetic factors have been estimated to account for approximately 60% of the risk of thrombosis⁸. Consequently the identification of genetic loci influencing clot structure/function may further our understanding of the underlying factors predisposing to occlusive vascular diseases.

The aims of the current study from EuroCLOT were to 1) determine the heritability of fibrin phenotypes and 2) identify QTLs associated with fibrin phenotypes in healthy twin volunteers from the UK and Denmark.

Materials and Methods

Subject recruitment and sample handling

The Twins UK Adult Twin Registry⁹ and the Danish Twin Registries study on endophenotypes¹⁰ provided the subjects for this study (please see http://atvb.ahajournals.org.). Venous blood was drawn at the participating centres following a standard protocol for venepuncture and sample processing for citrated plasma as previously reported¹¹. EDTA-anticoagulated whole blood samples were also obtained for DNA extraction. Citrated plasma aliquots, for analysis of D-dimer and turbidimetric fibrin phenotypes, were snap frozen in liquid nitrogen then stored at −80°C. Samples were batched up and transported on dry ice to the University of Leeds for phenotyping, (see below). Genotyping for linkage analysis was performed by Gemini-Sequenom, Cambridge (UK samples) and by Genotyping Centre Helsinki, as part of GenomEUtwin (Denmark samples)¹². Genotyping and phenotyping raw data were sent to King’s College London for collation and analysis.

Laboratory methods

Fibrin Phenotypes

D-dimer levels were determined according to the manufacturers’ instructions using TintElize(r) D-dimer ELISA kit, (Biopool, Umea, Sweden). Inter- and intraassay CVs were 5.6 and 3.3% respectively.

The high throughput turbidimetric assay of clot structure using customised software was performed as described elsewhere⁷. The following variables were analysed: Lag_C, (which represents the time at which sufficient protofibrils have formed to enable lateral aggregation, was taken as the time point at which an exponential increase in absorbance occurred), MaxAbs_C (a measure of clot density reflected by the absorbance at which 3 consecutive readings were identical corrected for the Lag_C absorbance), Lys50_t0 (calculated as the time from initiation of clot formation to the time at which a 50% fall in absorbance from MaxAbs_L occurred) and AUC (area under the curve, reflecting the balance between coagulation and fibrinolysis) The inter-assay CVs for Lag_C, MaxAbs_C, Lys50_t0 and AUC were 8%, 4%, 7.0% and 16%, respectively.

Genotyping

Standard methods were used for linkage genotyping (http://atvb.ahajournals.org.).

Statistical analysis

Heritability Estimates

For each phenotype (D-dimer, LagC, MaxAbsC, Lys50t0 and AUC) data were transformed using a box cox transformation to make their distributions approximately normal. Observations that were more than 3 standard deviations away from the mean were considered outliers and excluded from the analyses ¹³. Modelling for heritability assumes phenotypic variance to be due to three latent factors namely additive polygenic effects (A), common environment (C) and specific individual effects and measurement error (E). Please see http://atvb.ahajournals.org.

Linkage Analysis

Genotype and phenotype data were available on 447 UK DZ pairs and 199 DZ Danish pairs. Phenotype data (D-dimer, Lag_C, MaxAbs_C, Lys50_t0 and AUC) were analyzed using R¹³. Data were transformed to optimize closeness to normality using a Box-Cox transformation using the box.cox.power function in the car package^{14, 15}, to reduce the type I errors while preserving power for phenotypes having skewed distribution¹⁶. For details of joint linkage analysis please see http://atvb.ahajournals.org.

Results

Initially 2422 UK and 429 Danish twins were recruited to the study. Following genotyping and phenotyping and the removal of outliers as described above, complete data were available on 1814 UK female twins (447 DZ pairs and 460 MZ pairs), and 199 unselected Danish DZ twin pairs (Table 1).

Table 1.

Characteristics of UK and Danish twins and their phenotypes

Population	UK twins	Denmark twins
Sample size	1814	398
% females	100	50.3
No MZ pairs	447	0
No DZ pairs	460	199
Age/mean (range) yr	57 (48 – 64)	36 (29 – 49)
Ddimer/median (IQ range)ng/ml	77 (56 – 112)	64 (48 – 87)
Lys50t0/median (IQ range)	1716 (1531 – 1975)	1377 (1270 – 1542)
AUC/median (IQ range)	366 (250 – 533)	302 (224 – 415)
Lagc/median (IQ range)	343 (297 – 394)	282 (250 -321)
MaxAbsc/median (IQ range)	0.41 (0.35 – 0.47)	0.31 (0.25 – 0.37)

MZ represents monozygotic, DZ dizygotic, yr years, IQ interquartile

Phenotypes are defined in the Methods

Heritability Estimates

Narrow sense heritability (proportion of phenotypic variance attributable to additive genetic factors) requires data from both MZ and DZ twins so the UK twin sample was used. Examination of the ACE model and its sub-models revealed, in each phenotype, that the ACE model provided the best explanation of the observed data (data not shown). Heritability estimates obtained from the ACE model were in the range 17-46% (Table 2). These results indicate a significant contribution of additive genetic factors to variance in fibrin clot structure and function as measured by the 5 phenotypes (Table 2, column A).

Table 2.

Estimates for additive genetics effects (A), common environment (C) and specific individual effects and experimental error (E) and 95% confidence intervals for UK twins

Phenotype	A	C	E
D-dimer	25% (6% - 44%)	18% (3% - 33%)	57% (50% - 64%)
Lys50t0	23% (11% – 35 %)	47% (37%- 55%)	30% (27% – 35%)
AUC	28% (15% - 40%)	41% (30% - 51%)	31% (28% - 37%)
Lagc	17% (8% - 25%)	62% (55% - 69%)	21% (18% - 25%)
MaxAbsc	46% (30% - 62%)	15% (2% - 28%)	39% (34% - 45%)

Estimates shown are derived from the ACE model which was found to be the best fitting model for each phenotype

Linkage Analysis

Genome-wide linkage analysis using the combined datasets identified evidence of linkage at a number of regions (Table 3). Regions with LOD>3 included Lag_c on chromosome 6p22.2-q16.1 with peak LOD score 3.6 (Figure 1 panel B) at marker D6S426; on chromosome 9q22.1-q22.33 with maximum LOD=3.53 at marker D9S287 (Figure 1 panel C); chromosome 16p13.2-p13.13 with peak LOD score 4.57 (Figure 1 panel D) at marker D16S404. The first of these peaks, on chromosome 6, also showed linkage for Lys50_t0 with a partially overlapping peak with LOD 4.2 at marker D6S462 (Figures 1 panel B). Evidence of linkage was also observed for AUC on chromosome 17q22-q24.3 with peak LOD score 3.51 (Figure 1) at marker D17S944 and for d-dimer on chromosome 5q14.1-q21.3 with LOD=3.51. Suggestive linkage¹⁷ was observed for Lag_c on chromosomes 1, 3 and 17; for Lys50_t0 on chromosomes 4, 5, 6, 9, 12 and 17; and for D-dimer on chromosome 14 (Table 3).

Table 3.

Details of the linkage peaks having LOD>2 for the DZ twins from UK and Denmark

Chromosome	Trait	Flanking Markers*	Peak-LOD
1p13.2-1q21.3	Lagc	WIAF_3798 - WIAF_3272	2.27
3p26.3-3p25.3	Lagc	D3S4559 - D3S1263	2.17
4p16.1-4p15.1	Lys50t0	D4S2935 - WIAF_3923	2.01
5q14.1-q21.3	d-dimer	WIAF_1721 - WIAF_3198	3.51
5q23.1-5q31.3	Lys50t0	WIAF_3210 - WIAF_2528	2.57
6p22.2-q16.1	Lagc	D6S1691 - WIAF_577	3.60
6q14.1-16.1	Lys50t0	D6S460 - WIAF_577	4.20
6q26-q-ter	Lys50t0	D6S1599 - q-ter	2.63
9q22.1-q22.33	Lagc	WIAF_2079 - D9S1690	3.53
9p21.2-9p13.1	Lys50t0	D9S171 - WIAF_975	2.55
12q24.32-q-ter	Lys50t0	D12S1679 - q-ter	2.45
14q24.1- 14q32.13	D-dimer	WIAF_2677 - WIAF_1602	2.62
15q13.3-q21.3	MaxAbsC	WIAF_1283 - D15S117	2.18
16p13.2-p13.13	Lagc	D16S423 - D16S3075	4.57
17p13.3- 17p13.1	Lagc	D17S831 - D17S945	2.76
17q22-q24.3	AUC	D17S787 - D17S949	3.51

^*1cM confidence interval

Figure 1.

Linkage plots for the UK and Danish twins over 5 phenotypes

Plot A represents chromosome 5, plot B chromosome 6, plot C chromosome 9, plot D chromosome 16 and Plot E chromosome 17. X axis represents distance along chromosome in cM, Y axis LOD score.

Discussion

Over the last 20 years there has been considerable progress in understanding the contribution of abnormalities in haemostasis to the pathogenesis of vascular disease. Alterations in individual components of the coagulation and fibrinolytic pathways have been related to both cardio- and cerebrovascular disease in case control and prospective studies^{1, 2, 18}. Evidence indicates genetic contributions to most aspects of these pathways^{19, 20} including a substantial genetic component to the structure and function of fibrin^21,22. Taken together, these observations support the view that genetic factors contribute ultimately to vascular risk.

The heritability study shows varying degrees of genetic influence on fibrin phenotypes, with the strongest effect on MaxAbs_c, a measure of clot density, at 46%. The heritabilities obtained are similar to those we have previously described in twin²¹ and family⁷ studies. These results confirm that the intermediate fibrin-related phenotypes are heritable and that a genome-wide approach offers a tractable strategy for the elucidation of underlying genetic factors. Multipoint linkage analysis of fibrin phenotypes revealed 6 chromosomal regions with significant linkage peaks (LOD scores >3.0) and a further 9 regions of suggestive linkage (LOD scores 2.0 – 3.0). The peaks having significant linkage did not harbour any obvious known structural genes for coagulation or fibrinolytic factors that might influence fibrin phenotypes.

The development of occlusive thrombotic disease is dependent on the formation of a cross-linked fibrin mesh which in the arterial system supports incorporation of platelets and platelet particles. The mechanisms involved in regulation of fibrin structure/function are complex and involve coagulation and fibrinolytic proteins, other plasma components and regulatory factors. Fibrin formation itself occurs following activation of the coagulation system with consequent thrombin generation, fibrinogen cleavage and activation of coagulation FXIII, all of which contribute to the final fibrin phenotype. That fibrin formed from plasma ex-vivo differs markedly from that formed by recombinant or purified fibrinogen²³ suggests that additional plasma components contribute to the clot phenotype – mediated, perhaps, by diverse mechanisms. Preliminary data from the fibrin proteome indicate a range of immunoglobulins, inflammatory proteins and other plasma components associates with the fibrin matrix and could alter structure/function^{24, 25}. Finally, regulatory factors influencing gene transcription, post-translational modification and protein degradation modify all processes leading to fibrin clot formation. For example, PAI-1 levels are related to a PAI-1 gene promoter polymorphism²⁶, modulated by triglyceride levels²⁷ and regulated by the transcriptional factor PPARγ^{28, 29} and components of the molecular clock³⁰. Similarly, there are substantial data indicating that fibrinogen is regulated by complex genetic influences²².

The above should be borne in mind when interpreting the results. Although several highly significant linkage peaks were identified, there was little to suggest the peaks covered structural genes coding for coagulation or fibrinolytic proteins. There was, however, a suggestive linkage peak for Lys50_t0 (6q26-q-ter) in proximity to genes encoding plasminogen and lipoprotein(a) - which have clearly defined roles in fibrinolysis³¹.

A variety of potentially important regulatory genes were identified. For the wide AUC 17q22-q24.3 peak a number of putative candidates were identified, including HLF (hepatic leukaemia factor) and GNA13 (guanine nucleotide binding protein α-13). HLF transactivates the genes encoding coagulation factors FVIII and FIX³² and GNA13 is involved in platelet signalling and deficiency in mice leads to decreased response to thrombin, thromboxane A2 and collagen and decreased thrombus formation under flow³³. Furthermore, this region has been linked to clustering of metabolic syndrome components³⁴. In addition, the 14q24.1-q32.13 D-dimer peak harbours a SERPIN cluster which includes SERPINA10 which inhibits Factors Xa and XIa and has been associated with thrombosis in a mouse model³⁵. The 3p26.3-p25.3 peak for Lag_C contains SNPs associated with FVII and fibrinogen in the Framingham Heart Study³⁶.

For several other linkage peaks putative candidate genes for thrombosis and/or cardiovascular disease were identified. The highest LOD scores were observed for Lag_c and AUC (4.57 and 3.51) on chromosomes 16 and 17 respectively. The 16p13.2-p13.13 peak (Lag_c) has been linked to type 1 diabetes (T1DM)³⁷ and carotid artery calcified plaque in T2DM³⁸. The 1p13.2-q21.3 peak includes 2 SNPs associated with LDL³⁹. The 12q24.32-q-ter (Lys50_t0) peak corresponds to a bivariate linkage locus for cholesterol and HDL⁴⁰ and a locus for insulin resistance⁴¹ and contains a SNP associated with vWF³⁶. The 6p22.2-q16.1 (Lag_C) peak is very wide and partially overlaps with the 6q14.1-q16.1 Lys50_t0 peak. The former includes the lymphotoxin-α (LTA) gene - LTA SNPs have been associated with MI⁴². This region also corresponds to a QTL for LDL in a Samoan sample⁴⁰ and 2 SNPs associated with vWF in the Framingham Heart Study³⁶. Of particular note, the 9p21.2-p13.1 Lys50_t0 peak corresponds with replicated loci for cardiovascular disease⁴³ and T2DM⁴⁴. Together, these results suggest mechanisms for linking genetic loci to cardiovascular risk through effects on fibrin structure/function.

This study contained an excess of women, present for historical reasons. There are significant gender differences in vascular risk, mediated by hormonal influences and environmental exposure and it is possible that a study of men might produce different results. Another potential limitation is that the samples comprised twins. However, twins have been shown to be comparable to singletons for many common traits and diseases^45,46 and same sex dizygotic twin pairs actually offer considerable advantages in linkage studies because intra-pair age and sex differences are eliminated, and the effect of the individual-specific random environment is reduced.

Several steps need to be taken in EuroCLOT to link the current findings to vascular outcomes. These include fine mapping, prospective and case control studies of cardiovascular disease and functional studies of the genes and their encoded proteins. Ultimately, this work will further understanding of the mechanisms underlying the pathogenesis of occlusive vascular disease and could potentially provide novel therapeutic targets for the amelioration of the effects of these common disorders.