The Kozak sequence polymorphism of platelet glycoprotein Ibα and risk of nonfatal myocardial infarction and nonfatal stroke in young women

The glycoprotein (GP) Ib-IX-V complex is the major platelet surface receptor for von Willebrand factor (vWF). GPIb-IX-V plays a key role in the adhesion of platelets to injured vascular subendothelium and mediates shear-induced platelet activation.1,2 Two polymorphisms in the GPIbαgene affect the structure of the protein. The first is a size polymorphism resulting from a variable number of tandem repeats (VNTR) of a 13-amino acid sequence with designated alleles A, B, C, and D. The second is a Thr/Met dimorphism at position 145 located in the vWF binding domain of GPIbα, which defines the human platelet alloantigen system HPA-2. The HPA-2 and VNTR polymorphisms are in strong linkage disequilibrium.3-5 

A third polymorphism of the GPIbα gene that has been described more recently involves a −5T/C dimorphism in the translation initiation sequence (Kozak sequence).6,7 From previous studies of the translation initiation site, the nucleotides preceding the AUG initiator codons are thought to be important in translation efficiency in eukaryotic cells.8 The original consensus Kozak sequence contains nucleotide C at position −5, and therefore the switch to T at the −5 position of GPIbα may theoretically decrease translation efficiency of the receptor. Afshar-Khargan et al6 reported an association between the C allele of the Kozak dimorphism and increased platelet GPIb-IX-V receptor levels both in vitro and in vivo.

Because GPIb-IX-V receptor density may influence platelet adhesion to exposed vascular subendothelium after atherosclerotic plaque rupture, and thus affect risk of thrombosis, we assessed the relationship between the C allele of the Kozak sequence dimorphism ofGPIbα and risk of nonfatal MI and stroke among young women in a population-based case-control study. To assess potential gene-environment interactions, we also performed analyses stratified by other known cardiovascular risk factors. Finally, because of the possible association between the HPA-2 and VNTR polymorphisms and risk of atherothrombotic disease,9-11 we also considered the potential confounding effects of these other GPIbαpolymorphisms in our analysis.

Approval for these studies was obtained from the University of Washington's institutional review board. Informed consent was provided according to the Declaration of Helsinki. We used data and specimens from a population-based case-control study among women ages 18 to 44 years living in 3 contiguous counties of western Washington state between July 1, 1991, and February 28, 1995.12-14Potential MI and stroke cases were initially identified by chart review of discharge diagnoses from all 34 hospitals and Emergency Medical Service incident reports in the study region. Eligible MI cases were women who had a probable MI diagnosed using criteria of the Cardiovascular Health Study (CHS).15 Stroke cases were defined as women with new neurologic deficits lasting more than 24 hours or resulting in death in less than 24 hours and were classified by the study neurologist as ischemic, hemorrhagic, venous, or “other” (including arterial dissections) based on brain imaging studies and lumbar puncture results. Two hundred eight eligible MI patients and 249 stroke patients were identified, and 161 of MI patients and 198 of stroke patients were living at the time recruitment was initiated. One hundred twelve MI and 149 stroke patients were recruited and interviewed. Six hundred eighty-four age-matched female control subjects without history of cardiovascular disease (CVD) were identified from the same geographic area by random digit telephone dialing. Five hundred twenty-six control subjects were recruited and interviewed for the study (Table 1).

Case patients and control subjects were interviewed in person regarding histories of cardiovascular risk factors, including hypertension, diabetes mellitus, hypercholesterolemia, cigarette smoking, height and weight, menstrual history, contraceptive practices, history of MI in first-degree relatives, and demographic characteristics. All questions elicited information from the period before the MI or stroke in each case patient or an equivalent date for control subjects. A venous blood sample was collected from 79 MI cases, 106 stroke cases, and 391 control subjects into EDTA-treated vacuum tubes. Aliquots of plasma and buffy coat were frozen at −70°C. Samples were sent to Leiden, The Netherlands, where DNA was extracted as described by Miller et al.16 Analyzable DNA was available for 78 MI cases, 106 stroke cases, and 384 controls.

Platelet genotyping for the GPIbα Kozak polymorphism was performed by polymerase chain reaction (PCR) amplification of genomic DNA, followed by restriction enzyme digestion, according to the method of Kaski et al.7 Each reaction contained 100 ng DNA, 4 pmol of each primer, 200 μM dNTP, 0.4 units of Taq polymerase, 2 μL 10× reaction buffer, and 16 μL water in a total volume of 20 μL. For the amplification, the samples were heated to 93°C for 5 minutes, followed by 35 cycles of denaturation at 95°C for 30 seconds, annealing at 65°C for 30 seconds, and extension at 72°C for 1 minute, ending with a final extension step of 72°C for 5 minutes. The amplification products were digested in a mixture of 5 units of the restriction enzyme HaeIII, 1 μL 10× buffer, 3.5 μL water, and 5 μL PCR product, which was incubated for 2 hours at 37°C. Samples were analyzed by electrophoresis on a 2% agarose gel and stained with ethidium bromide. Thirty-eight samples were randomly selected for duplicate testing using the methods previously described with consistent results. In addition, 5 samples of each genotype (T/T, T/C, and C/C) were selected for direct nucleotide sequencing with confirmation of accuracy on all 15 samples.

Platelet genotyping for the HPA-2 (Thr/Met¹⁴⁵)17 and VNTR⁴polymorphisms of GPIbα were performed as previously described.

Current smokers were defined as subjects who reported smoking currently and regularly (at least 5 cigarettes per week for at least 6 consecutive months), and all others were classified as nonsmokers. Obesity was defined as a body mass index (BMI) 27.3 kg/m²or greater. Hyperhomocysteinemia was defined as a serum homocysteine level greater than 12.6 μM. A woman was classified as hypertensive, hypercholesterolemic, or diabetic if she reported ever receiving the diagnosis by a physician. Women were classified as premenopausal if they (1) reported still having menstrual periods, (2) were currently pregnant or nursing, or (3) had undergone hysterectomy but had at least one remaining functioning ovary. Family history of early onset MI was defined as having at least one first-degree relative who had an acute MI before the age of 55.

The association of the C allele of the GPIbα Kozak sequence polymorphism with MI and stroke was examined by calculation of the age-adjusted odds ratio (OR) and 95% confidence interval (CI) using unconditional logistic regression. The extent to which the association of the C allele with risk of MI or stroke was modified by other cardiovascular risk factors (smoking, hypertension, diabetes mellitus, hypercholesterolemia, obesity, and homocysteinemia) was assessed through analyses stratified by these other risk factors. To test for the presence of interaction between the Kozak C allele and these other cardiovascular risk factors, multiplicative terms were introduced into the logistic regression model, and the Pvalue was computed for the likelihood ratio test comparing the model containing the interaction term with the model lacking the interaction term. To control for the potential confounding effects of theGPIbα HPA-2 and VNTR polymorphisms on the risk of MI and stroke due to linkage disequilibrium with the Kozak sequence C allele, we also performed separate logistic regression models that included either (1) an indicator variable encoding the presence (one copy or more) of the less common HPA-2b (Met¹⁴⁵) allele or (2) 3 indicator variables encoding the presence (one copy or more) of the VNTR-B, VNTR-C, and VNTR-D alleles. All data were analyzed using the SAS package (Cary, NC).

The characteristics of the MI patients, stroke patients, and control subjects are summarized in Table2. The 78 women with acute MI had a median age of 39.7 years (range 23 to 44) and the 106 women with stroke a median age of 36.6 years (range 18 to 44). Most of the study patients and control subjects were white. There was a higher proportion of African Americans among the hemorrhagic stroke and MI cases than in the control group. The frequency of cardiovascular risk factors such as smoking, obesity, hypercholesterolemia, hypertension, diabetes mellitus, and hyperhomocysteinemia was higher in the MI patients than control subjects. For both ischemic and hemorrhagic subtypes, the frequency of smoking, hypertension, and hyperhomocysteinemia was higher in the stroke patients than control subjects. In contrast, higher frequencies of obesity and diabetes were confined to the ischemic stroke cases.

The distribution of the GPIbα-5C/T genotypes among the study subjects is shown in Table 3. Among the control subjects, the GPIbα-5C/T genotype distribution was in Hardy-Weinberg equilibrium. MI patients were less likely to have at least one copy of the C allele compared with the control subjects (14.1% vs 23.7%), although the frequency of the −5C/C genotype was similar in both groups (1.3% vs 1.6%). The age-adjusted odds ratio for women carrying at least one copy of the C allele (0.53, 95% CI = 0.27-1.05) was virtually identical to the unadjusted OR (0.53; 95% CI = 0.27-1.04). When adjusted additionally for obesity, homocysteine level, and diagnosis of hypertension, diabetes mellitus, or hypercholesterolemia, the OR was 0.46 (95% CI = 0.22-0.96). With the analysis restricted to the white population, the OR for the age-adjusted model was 0.58 (95% CI = 0.28-1.20).

When stratified according to the presence or absence of known cardiovascular risk factors, the presence of the C allele was not associated with increased risk of MI even in women with known cardiovascular risk factors (Table 4). In women with obesity, the age-adjusted OR was 0.81 (95% CI 0.33-1.94), with diabetes, hypercholesterolemia, or hypertension was 0.83 (95%CI 0.35-1.99), and with hyperhomocysteinemia was 0.92 (95% CI 0.34-2.48). By contrast, in certain subgroups of women without traditional cardiovascular risk factors, the risk of MI associated with the C allele was particularly low. The age-adjusted OR in women without obesity was 0.20 (95% CI 0.05-0.87), without diabetes, hypercholesterolemia, or hypertension was 0.22 (95% CI 0.05-0.96), and without hyperhomocysteinemia was 0.36 (0.12-1.05). However, none of these interactions between traditional cardiovascular risk factors and the Kozak −5C allele were statistically significant (Table 4).

Examination of the GPIbα genotypes in cases and controls revealed strong linkage disequilibrium between the Kozak −5C allele and the GPIbα HPA-2a/VNTR-C haplotypes. For example, analysis of the 551 subjects (1002 haplotypes) who are genetically informative for the HPA-2 and −5C/T dimorphisms (ie, homozygous for one or both dimorphisms) revealed that all 120 of the haplotypes that contain the −5C allele also contain the HPA-2a allele. In contrast, all 83 HPA-2b haplotypes were associated with the −5T allele. In the remaining 899 haplotypes, the −5T allele was associated with HPA-2a. Because the GPIbα HPA-2 and VNTR polymorphisms have been associated with risk of atherothrombotic disease in other population studies,9-11 we reanalyzed the risk of MI associated with the Kozak sequence polymorphism in our subjects, while separately controlling for the effects of the HPA-2 and VNTR polymorphisms. The OR for MI associated with the presence of the Kozak C allele was 0.53 (95% CI 0.27-1.04) after adjusting for HPA-2, and 0.50 (95% CI 0.25-0.99) after adjusting for VNTR. Furthermore, there was no evidence of effect modification of the risk of MI associated with the Kozak C allele when the subjects were stratified according to presence or absence of the HPA-2b allele (Table 4).

Stroke patients overall were as likely as control subjects to have at least one copy of the C allele (23.6% vs 23.7%). The age-adjusted OR associated with ischemic stroke for women who possessed at least one copy of the C allele was 0.99 (95% CI = 0.59-1.65). When analyzed according to stroke subtype, there was no association between carrying at least one copy of the C allele and risk of hemorrhagic or ischemic events (Table 3). Additional adjustment for the HPA-2 and VNTR polymorphisms did not appreciably change the results (data not shown). The 3 stroke cases who possessed the C/C genotype all had thromboembolic events (2 had ischemic strokes and one had a cerebral venous thrombosis). The frequency of the C/C genotype in the ischemic stroke cases (4.9%) was higher compared with controls (1.6%), but the difference was not statistically significant. The age-adjusted OR associated with ischemic stroke for women who possessed 2 copies of the C allele was 3.21, but the confidence interval is wide (0.62-16.54), reflecting the small numbers of subjects.

The nucleotide −5C Kozak sequence variant of GPIbαwas recently associated with increased translational efficiency and increased platelet surface density of the GPIb-IX-V receptor on platelets as well as transfected cells expressing the complex.6 These findings support the possibility of the Kozak sequence C allele as a candidate genetic susceptibility marker for atherothrombotic disease. However, the role of theGPIbα Kozak sequence polymorphism in the pathophysiology of arterial thrombotic disease has not been well characterized. We have assessed the GPIbα Kozak polymorphism in young women, a study sample that we have previously used to demonstrate the importance of other genetic prothrombotic factors in atherothrombotic disease and their potential interaction with other cardiovascular risk factors.12 13 We found no increased risk of MI or stroke in young women who carry one or more copies of the C allele of the Kozak sequence of GPIbα.

MI and stroke are unusual events in young women, thus our study population is unique and well suited for analyzing potential prothrombotic risk factors. The incidence of cardiovascular risk factors such as smoking and obesity are quite high in our population, allowing us to analyze potential interactions between these known risk factors and new genetic factors. We have previously shown that factor V Leiden and prothrombin 20210 G to A each increase the risk of MI in young women and interact synergistically with other cardiovascular risk factors such as smoking.12 13 It is therefore important to note that we find no association between the GPIbα-5C allele and increased risk of MI in the same study population even in subgroups of women with other cardiovascular risk factors.

Our negative findings with respect to the C allele of theGPIbα Kozak sequence and increased risk of atherothrombotic events in young women are consistent with several recent reports involving predominantly middle-aged to older men. In a case-control study of 539 patients with nonfatal MI, Croft et al18 found no association with carriership of the C allele either overall (OR 1.03) or in subgroups of patients younger than 55 years old or defined by other cardiovascular risk factors. In a smaller hospital-based case-control study, Corral et al19 observed no association between Kozak genotype and acute coronary or cerebrovascular disease. Similarly, a preliminary report indicated no association with unstable angina.20 In contrast to these findings, another preliminary report noted an increased frequency of the C variant of the GPIbα Kozak sequence in men younger than 55 and women younger than 65 years of age with a family history of early onset MI, but only when individuals with a history of diabetes mellitus, hypertension, and hypercholesterolemia were excluded.21 

In addition to the absence of an association with increased risk of atherothrombotic disease in young women, our data suggest that the C allele may even be associated with reduced risk of MI in this population. However, because of the relatively small numbers of subjects and multiple subgroup comparisons, these findings should be interpreted cautiously. This trend toward an inverse association between the Kozak C allele and the risk of MI in young women clearly requires confirmation in larger studies.

The mechanism for the possible inverse relationship of the C allele of the GPIbα Kozak sequence is not readily apparent. One may speculate that the “high expression” C allele is associated not only with higher levels of the receptor on the platelet surface,6 but also with increased circulating levels of the soluble extracellular portion of GPIbα glycocalicin. Alternatively, the C allele may have functional consequences forGPIbα on another cell type (eg, endothelial cells) that may influence the development of atherothrombotic disease via a causal pathway distinct from increased platelet adhesion. Finally, because we analyzed only nonfatal MI cases, it is theoretically possible that the apparent inverse association results from selection bias due to a strong effect of the Kozak C allele on case fatality (discussed below).

As previously reported in the Spanish population,5 our results indicate strong linkage disequilibrium between theGPIbα Kozak C allele and the GPIbαHPA-2a/VNTR-C haplotype. The less common HPA-2b/VNTR-B haplotype, which appears to be linked to the Kozak T allele, has been associated with increased risk of MI and acute cerebral ischemia in some case-control studies9-11 but not in others.19,22-24 Thus, it is possible that the apparent inverse association between the Kozak −5C allele and risk of MI could result from its association with the “low risk” HPA-2a/VNTR-C haplotype. In our study population of young women, there was a trend toward an increased risk of ischemic stroke associated with HPA-2b,25 but no evidence of an association between HPA-2 and risk of MI (manuscript submitted). Furthermore, adjustment for HPA-2 and VNTR did not significantly alter the estimates of effect for the GPIbα Kozak C allele in our current analysis. Although Afshar-Kharghan et al6described a dose-dependent increase in GPIb-V-IX receptor levels associated with the C allele among healthy platelet donors, Corral et al19 were unable to confirm any difference in receptor levels between T/T and T/C individuals in a larger number of subjects. Thus, another possible explanation for a possible inverse association with MI is that the Kozak polymorphism may be linked to a nearby gene that is important in atherosclerosis or thrombosis.

Our study has several limitations, including a small number of subjects and multiple subgroup comparisons, which can potentially lead to spurious associations arising by chance. The small sample size also lacks the statistical power to detect some associations, including the effect of the C/C genotype. Lastly, the cases from whom blood specimens for DNA analysis were obtained included only women who survived at least 3 months after their MI or stroke. This resulted in exclusion of approximately 20% of the cases initially identified by chart review. Thus, our study design can only assess the relationship between genotype and nonfatal events, and an association may have been missed if the C allele is strongly associated with a high case fatality rate immediately after the acute event. However, the case-control study design is the only practicable approach, because MI and stroke among women under the age of 45 are very rare. Furthermore, it is highly improbable that early mortality is strongly influenced byGPIbα Kozak genotype. However, this issue can be resolved empirically only by genotype analysis of blood or tissue from fatal cases.

In conclusion, our results suggest that young women carrying the C allele of the Kozak sequence polymorphism of GPIbα are not at increased risk of MI or stroke. Paradoxically, the C allele may even be associated with a reduced risk of MI in this population. These findings require confirmation in studies that involve larger numbers of subjects and fatal events, as well as further elucidation of the phenotypic consequences of the GPIbα Kozak sequence variant.

We thank Fran Chard, Karen Graham, Carol Handley-Dahl, Judy Kaiser, Marlene Bengeult, Carol Ostergard, Denise Horlander, Barb Twaddell, Sandy Tronsdal, and Jill Ashman for assistance with the study.