In vivo platelet activation in atherothrombotic stroke is not determined by polymorphisms of human platelet glycoprotein IIIa or Ib

Authors


Dr David J Meiklejohn, Department of Medicine and Therapeutics, Polwarth Building, Medical School, University of Aberdeen, Foresterhill, Aberdeen AB25 2ZD, UK. E-mail: d.meiklejohn@doctors.org.uk

Abstract

Platelet membrane glycoprotein polymorphisms are candidate risk factors for thrombosis, but epidemiological data are conflicting. Thus, demonstration of a genotype-dependent alteration in function is desirable to resolve these inconsistencies. We investigated in vivo platelet activation in acute thrombosis and related this to platelet genotype. Frequencies of the 1b and 2b alleles of the HPA 1a/1b and HPA 2a/2b platelet glycoprotein polymorphisms were determined in 150 (52 men/98 women, mean age 58·3 years) patients with atherothrombotic stroke, and the influence of genotype on markers of platelet activation was assessed. Platelet P-selectin (CD62P) expression and fibrinogen binding was measured using whole blood flow cytometry within 24 h of stroke and 3 months later in 77 patients who provided a repeat blood sample. Results were compared with matched controls. Neither the 1b allele [allele frequency 0·11 vs. 0·13, odds ratio (OR) confidence interval (CI) 0·8 (0·5–1·3)] nor the 2b allele [0·09 vs. 0·07, OR (CI) 1·4 (0·8–2·4)] was significantly over-represented in patients. Increased numbers of activated platelets were found following stroke (acute mean P-selectin expression 0·64% vs. control 0·35%, P < 0·001; acute mean fibrinogen binding 1·6% vs. control 0·9%, P < 0·001). Activation persisted in the convalescent phase (P < 0·001 and P = 0·005 vs. controls for P-selectin and fibrinogen respectively). Expression of P-selectin and fibrinogen was not influenced by either the HPA 1a/1b genotype (P > 0·95 for each marker, Scheffe's test) or the 2a/2b genotype (P > 0·95 for each). Although persisting platelet activation is seen in atherothrombotic stroke, it is independent of HPA 1a/1b and 2a/2b genotypes. These data suggest an underlying prothrombotic state, but do not support the polymorphisms studied as risk factors for thrombotic stroke in this population.

Following atheromatous plaque rupture, platelets have a pivotal role in arterial thrombus formation (del Zoppo, 1998). Given the efficacy of anti-platelet therapy with aspirin in secondary prevention of acute arterial thrombosis (Antiplatelet Trialists' Collaboration, 1994) and the fact that enhanced platelet aggregation has been shown prospectively to predict mortality following myocardial infarction (MI) (Trip et al, 1990), enhanced platelet activation might be a risk factor for arterial thrombosis. Furthermore, genetic polymorphisms of platelet surface membrane glycoproteins (GPs), by an effect on platelet function, might contribute to that risk.

GPIIIa, which is part of the fibrinogen receptor with GPIIb, is central in the process of platelet aggregation (Nurden, 1995). Following platelet activation, the GPIIIa molecule undergoes conformational change and binds fibrinogen more avidly. Platelet fibrinogen binding (by GPIIIa) is therefore a marker of platelet activation. The 1b allele of the human platelet antigen (HPA) 1a/1b (PlA1/A2) polymorphism of the GPIIIa gene is common (Newman et al, 1989) and has therefore been widely investigated as a candidate risk factor for arterial thrombosis. Studies have failed to consistently link the 1b allele with ischaemic stroke (Ridker et al, 1997; Carter et al, 1998), but an association in younger patients has been proposed (Carter et al, 1998; Wagner et al, 1998). Some case–control studies (Weiss et al, 1996; Carter et al, 1997; Garcia-Ribes et al, 1998; Gardemann et al, 1998; Mikkelsson et al, 1999), but not others (Marian et al, 1996; Osborn et al, 1996; Herrmann et al, 1997; Ridker et al, 1997; Samani & Lodwick, 1997; Durante-Mangoni et al, 1998), have associated the 1b allele with an increased risk of MI. It has been suggested (Di Castelnuovo et al, 1999) that the 1b allele is not a risk factor for coronary heart disease per se, but is associated with MI at a younger age and thrombosis in situations of heightened risk, such as following ruptured coronary plaque (Zotz et al, 1998), angioplasty (Abbate et al, 1998) or coronary stenting (Walter et al, 1997; Kastrati et al, 1999). However, negative associations in this context have also been reported (Laule et al, 1999).

GPIb, as part of the Ib–V–IX complex, binds von Willebrand factor (VWF) and mediates platelet adhesion. The HPA 2a/2b (Koa/Kob) and variable number of tandem repeats (VNTR) polymorphisms of the GPIb gene are in linkage disequilibrium, and have been studied as candidate risk factors for arterial thrombosis, possibly mediated by altered VWF binding. Some reports (Murata et al, 1997; Gonzalez-Conejero et al, 1998; Sonoda et al, 2000), but not others (Carter et al, 1998; Corral et al, 2000), have linked the 2b allele of HPA 2a/2b and the C/B VNTR genotype with coronary artery and cerebrovascular disease.

The plausibility of any relationship between candidate genes and thrombosis is dependent upon evidence of altered platelet reactivity associated with individual alleles and interpretation of these conflicting epidemiological data is hindered by a lack of corresponding functional data in most studies. The Framingham Offspring Study found the 1b allele associated with greater aggregation in response to adrenaline (Feng et al, 1999), but others reported reduced aggregation of 1b platelets in response to activation of the thrombin receptor (Lasne et al, 1997). We previously found no association between the 1a/1b polymorphism and platelet fibrinogen binding (Meiklejohn et al, 1999), and the HPA 1a/1b genotype has not been associated with increased plasma concentrations of platelet activation markers (Carter et al, 1998). Furthermore, a paradoxically enhanced sensitivity of 1b platelets to aspirin has been reported (Cooke et al, 1998). The HPA 2a/2b and VNTR GPIb polymorphisms are not associated with density of surface receptor expression (Corral et al, 2000) or with plasma markers of platelet activation (Carter et al, 1998). Because of these deficiencies and inconsistencies in reported data, we performed a combined epidemiological and functional study to determine the role of these polymorphisms in acute thrombotic stroke. Evidence of in vivo platelet activation was sought, and the influence of HPA 1a/1b and 2a/2b genotypes determined.

Patients and methods

Recruitment Consecutive patients with a cerebrovascular accident (CVA) were considered for participation within 24 h of admission to Aberdeen Royal Infirmary between June 1997 and December 1998. This is the sole primary referral centre for the Grampian region of Scotland (population approximately 600 000); patients are therefore representative of the general stroke population. Stroke was defined as a sudden loss of global or focal cerebral function that persisted for more than 24 h, and patients were approached on admission and formal written consent obtained. In an attempt to study a single pathological cause of CVA, patients with a history of atrial fibrillation, valvular heart disease or connective tissue disease were not recruited. A computerized tomography (CT) brain scan was performed on all patients and those with evidence of intracranial haemorrhage or alternative intracranial pathology were excluded. A clinical assessment was then made to exclude a previously undetected cardiac source of thrombus. Patients with evidence of valvular heart disease or thrombus on echocardiography, or of atrial fibrillation following recruitment, were subsequently excluded. Additional diagnoses and drug therapies were recorded, and those taking oral anticoagulants were not recruited. Those considered to have suffered a transient ischaemic attack (TIA, symptoms resolving within 24 h) were included, provided that a cardiac source of embolus was considered improbable based on the clinical assessment. Duplex ultrasound examinations were performed to identify evidence of carotid atheroma. Healthy age- and sex-matched controls were obtained from the list of a local general practice that cares for a population from a large area of the City of Aberdeen and consists of a similar racial and social-class mix to the patient cohort. Those born in same year as subjects with no history of stroke, TIA, peripheral vascular or ischaemic heart disease were recruited after written informed consent. The study was approved by the Grampian Regional Ethical Committee.

Follow-up of subjects Patients were invited to reattend for repeat sampling in the convalescent period at least 3 months after the acute event. Patients were contacted by letter containing an appointment for reattendance. Those willing to provide a further sample who were unable to attend for venepuncture were visited at home or in a long-term care establishment. Repeat sampling was not therefore dependent on stroke severity. Repeat measurements of platelet activation status and plasma fibrinogen concentration were obtained, and the use of anti-platelet drugs was recorded.

Sampling Venepuncture was performed with subjects supine for at least 5 min. A 21G needle was inserted into an antecubital vein with the cuff applied to the upper arm. The cuff was removed and 10 ml of blood obtained, after discarding the first 5 ml to avoid ex vivo platelet activation. Samples were collected in 1:10 3·2% sodium citrate and a 50-μl aliquot was diluted in 450 μl of HEPES Mg buffer. DNA was extracted from the citrate sample (DNA extraction kit, Nucleon Biosciences, Coatbridge, UK).

FACS analysis The method of analysis was adapted from that described previously (Warkentin et al, 1990). Samples were treated in a standardized manner within 1 h of collection as follows: 40 μl aliquots of whole blood in HEPES Mg buffer were incubated with monoclonal antibodies for 30 min at room temperature. Platelets were labelled with fluorescein isothiocyanate (FITC)-conjugated anti-CD61 (Dako, Glostrup, Denmark), which binds to GPIIIa. We measured P-selectin expression using dual staining with FITC-anti-CD61 and phycoerythrin (PE)-conjugated anti-CD62P (Immunotech, Marseilles, France). Fibrinogen binding was measured by single-colour flow cytometry using FITC-conjugated rabbit polyclonal anti-human fibrinogen antibody. An optimal final concentration of 2 ng/μl of each antibody was used. Reactions were stopped by the addition of 1 ml of phosphate buffered saline (PBS). Samples were analysed within 4 h of preparation using a Coulter XL-MCL flow cytometer (Coulter Electronics, Luton, UK), although we found antibody binding to be stable for up to 6 h if stored at 4°C (data not shown). Platelets were gated by their side and forward light scatter characteristics and enclosed in an electronic bitmap. Listmode data were stored and processed using a personal computer and System II software version 1·0 (Coulter Electronics), and converted to scatterplots and histograms. Ten thousand events were analysed for fluorescence and the results were expressed as the percentage of platelets positive for P-selectin and binding fibrinogen. Mean cell fluorescence as a measure of the average density of antibody binding per platelet was recorded. Basal activation status was assessed in resting samples and platelet reactivity following incubation with adenosine diphosphate (ADP) at a final concentration of 1 × 10−5 mol/l for 5 min (Sigma, St. Louis, USA). The flow cytometer was aligned daily with ‘Flowcheck’ and ‘Immunobrite’ beads (Coulter Electronics) to calibrate light scatter and fluorescence parameters respectively. The flow cell was cleaned thoroughly between individual subject samples to exclude carry-over of platelets to subsequent analyses using Coulter Clenz cleansing solution (Coulter). As all antibodies except the anti-fibrinogen antibody were of murine origin, FITC- and PE-conjugated isotype controls from that species were used to adjust for non-specific antibody binding.

HPA 1a/1b genotyping Using the following primers (Jin et al, 1993), a 266-bp product was obtained using the polymerase chain reaction (PCR) of genomic DNA: sense, 5′-TTCTGATTGCTGGACTTCTCTT-3′; antisense, 5′-TCTCTCCCCATGGCAAAGAGT-3′.

Products were incubated overnight with 20 units of Msp 1 (MBI Fermentas, Vilnius, Lithuania), which yields a 221-bp fragment from the HPA 1a allele and a 177-bp fragment from the 1b allele (Weiss et al, 1996). The alleles were separated using electrophoresis on 2% agarose gel and visualized in ultraviolet light using ethidium bromide staining. Samples of a known genotype were included in each experiment and random samples were subjected to repeat genotyping to confirm the reproducibility of this method.

HPA 2a/2b genotyping This was performed using PCR sequence-specific priming (PCR-SSP) as described previously (Cavanagh et al, 1997). The following primers, reflecting the difference between 2a and 2b alleles at the 3′ end, were used: 2a, 5′-GCCCCCAGGGCTCCTGAC-3′; 2b, 5′-GCCCCCAGGGCTCCTGAT-3′.

Each sample was subjected to two PCR reactions, with each of the specific primers plus a common primer: 5′-TCAGCATTGTCCTGCAGCCA-3′. Products were electrophoresed on agarose gel and visualized following ethidium bromide staining. The identification of a 258-bp product with a given primer indicated the presence of the corresponding allele and the absence of an allele was indicated by the failure to obtain such a product. Each reaction included primers for the human growth hormone gene, which resulted in a 429-bp positive control for the amplification process. Control samples of a known genotype were used in each experiment.

Plasma fibrinogen concentration Whole blood (4·5 ml) was anticoagulated in 3·9% trisodium citrate and collected in a Vacutainer sample tube (Becton Dickinson, Cedex, France). Derived plasma fibrinogen concentration was measured on an ACL 3000 coagulometer (Instrumentation Laboratory, Warrington, UK).

Mean platelet volume (MPV) and platelet count Tripotassium EDTA anticoagulated whole blood (4·5 ml) was collected in a Vacutainer sample tube (Becton Dickinson, Oxford, UK) and analysed in a Bayer Technicon H3 autoanalyser (Bayer, Puteaux, France). All samples were processed within 1 h of collection.

Statistical analysis Calculations were performed using spss for Windows version 8·0 statistical software. The distributions of genotypes, allele frequencies and clinical risk factors for arterial disease in cases and controls were compared using Chi-square tests. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using standard formulae. Skewed continuous variables were normalized by log10 transformation and geometric means were calculated. In order to avoid the influence of artefactual platelet activation, we excluded values greater than three standard deviations from the mean. Numbers quoted in individual groups are therefore less than the total number analysed. Mean differences between patients and controls were analysed using Student's t-test, and between acute and convalescent patient variables using paired t-tests of transformed data. Unpaired skewed continuous variable medians were compared with the Mann–Whitney U-test. A two-tailed P-value of < 0·05 was considered significant. The effect of genotype on expression of activation markers was assessed by analysis of variance (anova) of transformed data and by Scheffe's post hoc analysis for multiple comparisons. We calculated that a sample size of 150 cases and controls was required to detect a twofold difference in HPA 1b allele frequencies between cases and controls with 80% statistical power and significance at the 5% level, assuming a control frequency of 19% based on published data (Weiss et al, 1996).

Results

Subjects

The outcome of patient recruitment is summarized in Fig 1. One hundred and seventy-nine consecutive patients with CVA or TIA consented to participate. Twenty-nine patients were subsequently excluded: 10 with intracranial haemorrhage, seven with atrial fibrillation, four with cardiac thrombi or valvular heart disease, four with intracranial tumours, and four who were considered to have an alternative diagnosis to stroke. We determined HPA 1a/1b and HPA 2a/2b genotypes of 150 cases (52 men/98 women) and 150 controls (52 men/98 women), and data on platelet activation from all patients and from 112 controls (52 men/60 women). The mean age (range) of patients and controls was 58·3 years (25–70 years) and 56·9 years (24–72 years) respectively. Seventy-seven (51·3%) patients were recruited within 24 h of stroke onset, a further 53 (35·3%) within 48 h, 15 (10%) patients within 72 h, and the remaining five (3·3%) patients were recruited within 96 h. One hundred and forty patients (93·3%) were considered to have suffered a cerebral infarction and 10 (6·7%) had suffered a TIA. One hundred and nine out of 150 patients (72·7%) had a carotid scan. Of these, 42 (38·5%) had evidence of carotid atheroma. Seventy-seven (48 men/29 women) patients provided a repeat sample in the convalescent period. The mean time to follow up was 100·3 d. All but one patient who attended 68 d after stroke were seen at least 90 d later (range 90–270 d). Seventeen of these patients were visited either at home or a nursing home. By July 2000, 15 patients had died.

Figure 1.

Outcome of study recruitment.

Established risk factors

Personal risk factors for ischaemic stroke and routine laboratory variables are summarized in Table I. Current smoking and hypertension were confirmed as risk factors for atherothrombotic stroke, but the odds ratios for diabetes mellitus and a positive family history did not reach statistical significance. A history of treatment for hyperlipidaemia or fasting total, low-density lipoprotein (LDL) cholesterol or triglyceride concentrations were also not associated with stroke. However, an apparent protective effect of high-density lipoprotein (HDL) cholesterol was detected, the mean concentration being significantly greater in controls (P < 0·001). The mean acute-phase plasma fibrinogen concentration (SD) was significantly higher than in controls (4·3 ± 1·1g/l vs. 3·3 ± 0·9, P < 0·001, Student's t-test) and this difference persisted into the convalescent phase (4·1 ± 1·1 vs. 3·3 ± 0·9, P < 0·001). In 77 patients who attended for follow-up, the mean fibrinogen concentration was higher immediately after stroke than in the convalescent period, but this difference did not reach statistical significance (4·3 ± 1·0g/l vs. 4·1 ± 1·0g/l, P = 0·06, paired t-test). Acute-phase fibrinogen concentration correlated with that at follow-up (r2 = 0·493, P < 0·001). We found no difference in mean platelet count or mean platelet volumes in acute and convalescent phases, or between patients and controls (Table I).

Table I.  Distribution of clinical risk factors, fasting lipid measurements and haematological variables in patients and controls.
Risk factor150 subjects n(%)109 controls n(%)Odds ratio (95% CI)P-value
  • *

    Student's t-test.
      

  • Paired t-test.
      

  • Mann Whitney U-test.

  • A diagnosis of hypertension, diabetes mellitus or hyperlipidaemia was defined as receiving current treatment for, or a past history of,the condition. A family history was defined as an arterial thrombotic event in a first-degree relative before the age of 55 years.

Current smoker77 (51·3) 26 (23·8)3·4 (2·0–5·8)
Hypertension65 (43·3) 16 (14·7)4·4 (2·4–8·3)
Family history45 (30) 27 (24·8)1·3 (0·7–2·3)
Diabetes mellitus12 (8)  4 (3·6)2·3 (0·7–7·2)
Hyperlipidaemia18 (12) 26 (23·8)0·4 (0·2–0·8)
No risk factor11 (7·3) 25 (22·9)0·2 (0·1–0·6)
Total serum cholesterol (mmol/l) mean (SD) 5·8 (1·3)  6·0 (1·1)0·22*
LDL cholesterol (mmol/l) mean (SD) 3·8 (1·2)  3·8 (1·0)0·80*
HDL cholesterol (mmol/l) mean (SD) 1·2 (0·4)  1·5 (0·3)< 0·001*
Triglyceride (mmol/l) median (range) 1·4 (0·3–8·2)  1·3 (0·4–4·2)0·12
Plasma fibrinogen (g/l) mean (SD) 4·3 (1·1)  3·3 (0·9)< 0·001*
Mean platelet volume (fl) mean (SD)acute 9·1 (0·8)
convalescent 9·1 (1·3)
  9·2 (0·9)acute vs. control P = 0·42*
convalescent vs. control P = 0·72*
acute vs. convalescent P = 0·87
Platelet count ( × 109/l) mean (SD)acute 252 (77·0)
convalescent 266 (59·3)
252 (66·4)acute vs. control P = 0·93*
convalescent vs. control P = 0·12*
acute vs. convalescent P = 0·23

Platelet genotypes

The distributions of platelet genotypes and allele frequencies are summarized in Table II. The 1b allele was not over-represented in patients and there was no difference in distribution of the HPA 1a/1b or HPA 1b/1b genotypes between cases and controls. There was no significant difference in 1b allele distribution (allele frequency) when analysis was restricted to subjects under the age of 60 years [16 out of 79 (0·20) vs. 22 out of 90 (0·24), OR (CI) 0·8 (0·4–1·6)]. There was no over-representation of the HPA 2b allele in the entire study cohort or in those under 60 years [n = 80, 16 out of 160 (0·10) vs. 16 out of 184 (0·09), OR (CI) 1·01 (0·5–2·1)]. The number of patients heterozygous or homozygous for the HPA 2b allele did not differ statistically significantly from controls and there were no differences when those under 60 years were studied.

Table II.  Genotype distributions and allele frequencies.
GenotypePatients (n = 150)Controls (n = 150)Significance
  1. Heterozygous and homozygous carriers of HPA1b or 2b were combined for chi square analysis. df, degrees of freedom; CI, 95% confidence intervals.

 1a/1a118112χ2 = 0·67, P = 0·41
 1a/1b 31 361 df
 1b/1b  1  2 
 2a/2a122128χ2 = 0·86, P = 0·35
 2a/2b 27 221 df
 2b/2b  1  0 
Allele frequency  Odds ratio (CI)
 1a267/300 (0·89)260/300 (0·87)0·8 (0·5–1·3)
 1b 33/300 (0·11) 40/300 (0·13) 
 2a271/300 (0·91)278/300 (0·93)1·4 (0·8–2·4)
 2b 29/300 (0·09) 22/300 (0·07) 

Platelet activation

Data on platelet expression of P-selectin and fibrinogen binding are summarized in Table III. There were significantly more platelets expressing P-selectin in acute stroke than in control samples (P < 0·001, Student's t-test) and this difference persisted into the convalescent period (P < 0·001, Student's t-test). The proportion of platelets binding fibrinogen was also greater in acute and convalescent stroke patients than in controls (P < 0·001 and P = 0·002, respectively, Student's t-test). We found no statistical difference in mean percentage P-selectin expression between the acute and convalescent periods in patients who provided a sample at both times (P = 0·67, paired t-test), but the number of platelets binding fibrinogen decreased from a geometric mean of 2·1% in the acute period to 1·4% in the convalescent period (P = 0·02, paired t-test). Values obtained immediately after stroke correlated with follow-up values (r2 = 0·431, P < 0·001 for log10% CD62P expression and r2 = 0·231, P = 0·048 for log10 fibrinogen binding). There was also evidence of a persisting greater density of anti-GPIIIa binding sites in stroke patients, as evidenced by greater mean cell fluorescence in acute and convalescent patients than in controls (P < 0·001 and P = 0·008, respectively, Student's t-test). There was no alteration in mean cell fluorescence between the acute and convalescent periods (P = 0·26, paired t-test). We found no difference in mean cell fluorescence for P-selectin expression or fibrinogen binding between acute, convalescent and control samples (data not shown). There were no differences in results when analyses were restricted to the 112 patients for whom a matched control was available (data not shown). CD62P expression (r2 = −0·09, P = 0·12) and platelet fibrinogen binding (r2 = −0·08, P = 0·58) did not correlate with plasma fibrinogen concentration.

Table III.  Platelet activation markers in acute and convalescent stroke.
GroupAcute n = 147Control n = 110Convalescent n = 74Significance
  • *

    Student's t-test.
      

  • Paired t-test.

% CD62P expression
geometric mean (range)
0·64 (0–4·4)0·35 (0·1–6·5)0·62 (0·1–4·3)Acute vs. control P < 0·001*, convalescent vs. control P < 0·001*,
 acute who also attended for follow-up, n = 74, geometric mean (range) 0·62 (0·1–4·4) vs.
 convalescent P = 0·67
% fibrinogen binding
geometric mean (range)
1·6 (0·2–11·0)0·9 (0·2–10·2)1·4 (0·2–8·7)Acute vs. control P < 0·001*, convalescent vs. control P = 0·005*,
 acute who also attended for follow-up, n = 74, geometric mean (range) 2·08 (0·2–4·4) vs.
 convalescent P = 0·02
Density of fibrinogen receptor
 (Anti-CD61 mean cell
 fluorescence units) mean (SD)
9·5 (2·3)8·4 (1·5)9·1(1·6)Acute vs. control P < 0·001*,convalescent vs. control P = 0·008*
Acute who also attended for follow up, n = 74, mean (SD) 8·7 (1·8) vs.
 convalescent, P = 0·26
% fibrinogen binding post
ADP Mean (SD)
55·2 (17·2)64·2 (16·5)60·9 (17·4)Acute vs. controls P < 0·001*, convalescent vs. control P = 0·43*
Acute who also attended for follow up, n = 74 mean (SD) 60·8 (18·4) v.
 convalescent, P > 0·95

We found evidence of reduced platelet reactivity immediately after stroke (Table III), as the percentage of platelets binding fibrinogen following stimulation with ADP was lower in the acute phase than in controls (P < 0·001). The density of fibrinogen binding sites after ADP stimulation did not significantly differ between the three groups (acute geometric mean (range) mean cell fluorescence 8·45 (2·0–39·8) units vs. 7·66 (2·3–37·2) in controls, P = 0·15, Student's t-test; convalescent mean 7·69 (2·1–46·8) P = 0·1 vs. controls, P > 0·95 vs. acute, paired t-test).

Of 146 patients with evaluable data, 48 (33%) were taking aspirin at the time of their stroke. Those not taking aspirin had significantly greater numbers of platelets binding fibrinogen than those who did not [median (range) 1·5% (0·2–5·5) vs. 1·2 (0·2–4·4), P = 0·05, Mann–Whitney U-test]. There was also a tendency for those not on aspirin to have more platelets expressing CD62P, but this was not statistically significant [0·70% (0·1–1·85) vs. 0·53 (0·1–1·81), P = 0·1, Mann–Whitney U-test]. In a subset of the study cohort (n = 35), we assessed the effect of commencing aspirin following stroke on platelet activation. We found no significant difference in the expression of CD62P [median (range) 0·7% (0·1–1·85) before aspirin vs. 0·5% (0·1–1·81) after aspirin, P = 0·2, Wilcoxon ranked signs test] or in fibrinogen binding [1·4% (0·4–5·3) before aspirin vs. 1·2% (0·2–4·3) after aspirin, P > 0·95, Wilcoxon ranked signs test).

Platelet genotype and activation

We investigated the effect of the 1b and 2b alleles in the acute phase, the convalescent phase, and in controls separately by anova and Scheffe's test, for multiple comparisons. Results are summarized in Table IV. No subgroup of 1b or 2b subjects exhibited a greater % fibrinogen binding, % CD62P expression or surface density of fibrinogen receptors than their 1a/1a or 2a/2a counterparts. These findings were not appreciably altered by inclusion of the statistical outliers, that had been excluded as described above, in the analysis (data not shown). There were no differences in % fibrinogen binding following stimulation with ADP.

Table IV.  Subgroup analyses of the influence of HPA 1a/1b and 2a/2b genotypes on the expression of markers of platelet activation using anova and Scheffe's test.


Group (number)
% expressing CD62P
Geometric mean
(range)


P
% binding fibrinogen
Geometric mean
(range)


P
Density of fibrinogen
receptor (anti-GPIIIa
fluorescence) mean (SD)


P
% binding
fibrinogen post
ADP Mean (SD)


P
Fibrinogen binding Mean
cell fluorescence post ADP
Geometric mean (range)


P
1a acute (108)0·63(0·1–4·17)> 0·951·59(0·2–10·9)> 0·959·53(2·41)> 0·9554·5(16·7)0·88 8·2(2·3–39·8)0·77
1b acute (30)0·69(0·2–4·37) 1·65(0·4–10·0) 9·30(1·82) 59·1(18·3)  9·7(3·0–24·0) 
1a control (86)0·36(0·1–6·46)> 0·951·01(0·2–10·2)> 0·958·48(1·33)> 0·9564·8(16·9)> 0·95 8·0(3·0–37·1)0·72
1b control (26)0·34(0·1–1·2) 0·92(0·4–2·4) 8·68(1·50) 59·7(17·4  6·6(2·3–13·2) 
1a convalescent (62)0·58(0·1–4·27)0·820·83(0·2–8·3)0·109·05(1·71)> 0·9558·4(17·1)0·18 7·2(2·1–46·8)0·33
1b convalescent (13)0·98(0·4–1·82) 2·57(0·5–8·7) 9·50(1·02) 72·6(14·2) 10·7(5·8–21·9) 
2a acute (121)0·63(0·1–4·37)> 0·951·77(0·2–18·2)> 0·959·55(2·26)0·9260·1(19·2)> 0·95 8·5(2·0–39·8)> 0·95
2b acute (28)0·74(0·2–2·57) 1·49(0·3–10·0) 9·09(2·44) 61·5(15·8)  8·1(4·6–23·4) 
2a control (93)0·37(0·1–6·46)> 0·951·06(0·2–15·49)> 0·958·56(1·37)> 0·9563·0(16·9)0·69 7·5(2·3–37·2)> 0·95
2b control (15)0·33(0·1–1·2) 1·04(0·4–2·29) 8·34(1·34) 71·6(11·7)  8·5(3·7–14·8) 
2a convalescent (58)0·68(0·1–7·76)> 0·951·56(0·2–25·7)0·949·23(1·65)> 0·9560·0(18·1)> 0·95 7·6(2·1–46·8)> 0·95
2b convalescent (14)0·83(0·3–7·59) 1·17(0·4–3·98) 8·80(1·44) 64·3(14·6)  8·0(3·3–18·6) 

We investigated whether allele-dependent altered sensitivity to aspirin was a potential confounding factor in our study. There was no difference between 1a and 1b platelets or 2a and 2b-containing platelets in expression of activation markers in those who were taking aspirin at the time of sampling and those who were not (data not shown).

Discussion

When investigating a candidate gene as a risk factor for arterial thrombosis it is desirable to demonstrate a plausible functional effect of a polymorphism to corroborate any epidemiological suggestion of risk (Nurden, 1997). The problem of interpreting conflicting data from epidemiological studies of platelet polymorphisms has given rise to the view that genetic influences are modified by environmental factors, so that only certain patients may be at increased risk (e.g. smokers, those of a young age or female), and subgroup analysis may be necessary to identify these risk groups (Carter et al, 1998; Wagner et al, 1998; Bray, 1999). However, epidemiological studies of modest size that conduct subgroup analyses probably overestimate the risk, especially given the publication bias associated with positive studies (Ridker & Stampfer, 1999). Thus, it has been suggested that only large population-based studies should be undertaken to study the association of genetic polymorphisms with arterial disease (Ridker & Stampfer, 1999; Keavney et al, 2000). In this modestly sized study, epidemiological data was combined with an investigation of the influence of polymorphisms on platelet function. Evidence of persisting platelet activation in atherothrombotic stroke was found, but there was no association between the HPA 1a/1b and 2a/2b polymorphisms of GPIIIa and GP1b, respectively, and expression of these markers in either healthy controls or patients in the acute and convalescent periods of stroke. There was no significant epidemiological association with genotype and atherothrombotic stroke in the study cohort, even when analysis was restricted to younger patients. It should be stated that our study is inadequately powered to determine whether these candidate genes are weak epidemiological risk factors for stroke. However, our primary intention was to investigate the functional effect of these polymorphisms on platelets from patients with thrombosis, in order to evaluate possible mechanisms by which increased risk might be mediated.

Given the fact that GPIIIa is part of the fibrinogen receptor and that fibrinogen binding is dependent on the conformational change of GPIIIa on activation, it has been suggested that HPA 1b might mediate an increased risk of thrombosis by enhancing platelet fibrinogen binding (Bray, 1999). We did not confirm this in a previous study of platelets from healthy subjects (Meiklejohn et al, 1999), and replicate those findings in the current study, despite evidence of increased numbers of platelets binding fibrinogen following stroke. It is acknowledged that platelet genotype may influence function by another mechanism and it has been suggested that the HPA 2a/2b polymorphism may result in altered VWF binding in the conditions of high shear stress seen in arteries narrowed by atheroma (Bray, 1999), but data to support this are lacking. However, both P-selectin expression and fibrinogen binding occur in the later stages of platelet activation (del Zoppo, 1998) and, although GPIIIa and GPIb polymorphisms may not influence these directly, any effect on platelet activity should ultimately be reflected in these end-points.

Until recently, the only study to demonstrate enhanced 1b platelet activity compared to 1a recruited a much larger cohort than other functional studies (Feng et al, 1999) and did not study in vivo activation during an acute thrombotic event. It is recognized that our study may be underpowered to detect a more modest effect, but it is unclear whether such an effect would be sufficient to influence a clinical event such as thrombotic stroke. A recently published study of 20 HPA 1a/1b heterozygotes, 16 1b/1b homozygotes and 20 HPA 1a/1a homozygotes (Michelson et al, 2000) investigated allele-dependent platelet function using platelet aggregometry and whole blood flow cytometry in healthy blood donors. A greater density of platelet surface P-selectin expression, fibrinogen binding and response to ADP was observed in 1b platelets than in 1a platelets. This effect was dose dependent, with 1b/1b platelets displaying greater functional differences than heterozygous platelets. It was not possible to investigate this in the Aberdeen study as only one such subject was identified. However, it was considered justifiable to search for an effect in 1a/1b platelets, as heterozygous subjects comprise the majority of those claimed to be at risk of thrombosis in some clinical studies (Weiss et al, 1996; Carter et al, 1997; Garcia-Ribes et al, 1998; Gardemann et al, 1998; Mikkelsson et al, 1999). In addition, as this study was conducted in healthy subjects, it remains unclear whether these differences result in a clinical effect. These observations were not confirmed in our study and it remains unclear whether allele-dependent effects observed in healthy subjects are sufficient to cause disease, as they were not apparent in patients with acute thrombosis. It could be postulated that small allele-dependent physiological differences observed in normal people (Feng et al, 1999; Michelson et al, 2000) or in patients with stable ischaemic heart disease (Goodall et al, 1999; Michelson et al, 2000) might be overcome by the increase in platelet activity that is observed following stroke, so that an effect of platelet genotype on risk is not apparent. These allele-dependent differences in healthy subjects therefore require prospective evaluation to determine whether they result in an increased risk of thrombotic events.

Potential confounding by aspirin use was investigated, as it has been suggested that sensitivity to aspirin may be influenced by platelet HPA 1a/1b genotype (Cooke et al, 1998). There was no difference in expression of activation markers between 1a and 1b platelets or 2a and 2b platelets in either subjects who were taking aspirin or those who were not. The observation of greater platelet activation in those subjects not taking aspirin on admission and the subsequent failure to reduce this when aspirin was commenced after stroke suggests that a subgroup of patients might benefit from additional anti-platelet therapy. However, we may have failed to detect an effect of subsequent aspirin therapy on platelet activation because the subgroup studied was small; a larger study is therefore required to confirm these data.

The study of stroke pathophysiology is associated with particular difficulties. Stroke arises from numerous processes including intracranial haemorrhage, cardiac embolization, atherothrombosis (rupture of either large vessel atheroma with cerebral embolism or of small vessel atheroma with thrombotic occlusion) and vasculitis. Most studies have failed to distinguish between these diverse stroke types (Couch & Hassanein, 1976; Dougherty et al, 1977; Hoogendijk et al, 1979; Shah et al, 1985; van Kooten et al, 1997, 1999) and any individual risk factor might influence only one of these processes. Subjects with a probable single underlying pathophysiological process were therefore studied, although it is acknowledged that we may not have excluded all cases of cardioembolic stroke because we did not perform transoesophageal echocardiography. An association of atherothrombotic stroke with smoking, hypertension and plasma fibrinogen concentration was confirmed, but there was no association with diabetes mellitus, hyperlipidaemia or family history of arterial events. This lack of association between family history of thrombosis and stroke, and the observation of no association between platelet genotype and stroke, in our cohort further supports the view that genetic influences on cardiovascular disease are weak in comparison to environmental risk factors (Ridker & Stampfer, 1999).

In vitro measurements of platelet function may not necessarily reflect in vivo activity. In particular, assays of plasma beta thromboglobulin, platelet factor 4 (Shah et al, 1985) and platelet aggregometry (Couch & Hassanein, 1976; Dougherty et al, 1977; Konstantopoulos et al, 1995) are susceptible to artefactual platelet stimulation as a result of centrifugation and stirring procedures (Michelson, 1996). Platelet function was studied directly using whole blood flow cytometry to avoid this effect and was related to polymorphisms of receptors that have a major role in platelet recruitment in acute thrombus formation (Nurden, 1995). Artefacts were further minimized by careful blood sampling and by excluding values greater than three standard deviations from the mean from statistical analysis. However, it should be acknowledged that this study assessed function in circulating platelets. This may not necessarily reflect localized platelet activity within the thrombus in the acute phase of stroke and, as such, remains an indirect assessment of in vivo platelet function in arterial thrombosis. This point is illustrated by the observation that the absolute numbers of platelets expressing activation markers is a fraction of the total platelet count in patients (Table III).

Prospective data on platelet activity as a risk factor for future stroke development are lacking, but this study found evidence of persisting platelet activation in stroke, indicated by circulating platelets expressing P-selectin and binding fibrinogen, and by greater density of GPIIIa binding in both the acute and convalescent phases. Given that a substantial proportion of patients had significant pre-existing arterial disease (28% had a history of ischaemic heart disease and 26% had a history of previous stroke or TIA) and the observation that small absolute numbers of platelets expressed activation markers, it could be suggested that these results reflect the ongoing plaque inflammation and thrombosis observed in diffuse arterial disease. Furthermore, platelet activity immediately after stroke may reflect a secondary response to tissue injury (Dougherty et al, 1979; Robertson et al, 1980). However, the observation of persisting platelet activation may indicate an underlying prothrombotic tendency. This is supported by the observation that platelet activation markers did not mirror the plasma fibrinogen concentration and are therefore not merely markers of inflammation. Reduced platelet responsiveness in acute stroke was observed, indicated by reduced numbers of platelets binding fibrinogen following stimulation with ADP. This is consistent with circulation of activated platelets that are no longer susceptible to further stimulation, or removal from the circulation of those platelets that have formed platelet–platelet and platelet–leucocyte aggregates upon activation. These changes are clearly subtle, as they are not reflected in altered numbers of circulating platelets or in their mean volume.

In conclusion, there is evidence of platelet activation in acute atherothrombotic stroke, which persists months after the acute event. This suggests that platelet activation is a marker of a prothrombotic state and may be involved in the pathogenesis of acute stroke. However, the HPA 1b or HPA 2b alleles are not associated with stroke and they do not appear to influence in vivo platelet activation. These data do not support these polymorphisms of platelet membrane GPIb and GPIIIa as risk factors for atherothrombotic stroke in the Grampian population.

Acknowledgments

This study was funded by Chest, Heart and Stroke (Scotland) and by Aberdeen Royal Infirmary Endowments Fund.

Ancillary