OUP user menu

Genetics and ischaemic stroke

Ahamad Hassan , Hugh S. Markus
DOI: http://dx.doi.org/10.1093/brain/123.9.1784 1784-1812 First published online: 1 September 2000


Ischaemic stroke can be caused by a number of monogenic disorders, and in such cases stroke is frequently part of a multisystem disorder. Cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy (CADASIL), due to mutations in the Notch 3 gene, is increasingly appreciated as a cause of familial subcortical stroke. The genetics and phenotypes of monogenic stroke are covered in this review. However, the majority of cases of ischaemic stroke are multifactorial in aetiology. Strong evidence from epidemiological and animal studies has implicated genetic influences in the pathogenesis of multifactorial ischaemic stroke, but the identification of individual causative mutations remains problematic; this is in part limited by the number of approaches currently available. In addition, genetic influences are likely to be polygenic, and ischaemic stroke itself consists of a number of different phenotypes which may each have different genetic profiles. Almost all human studies to date have employed a candidate gene approach. Associations with polymorphisms in a variety of candidate genes have been investigated, including haemostatic genes, genes controlling homocysteine metabolism, the angiotensin-converting enzyme gene, and the endothelial nitric oxide synthase gene. The results of these studies, and the advantages and limitations of the candidate gene approach, are presented. The recent biological revolution, spurred by the human genome project, promises the advent of novel technologies supported by bioinformatics resources that will transform the study of polygenic disorders such as stroke. Their potential application to polygenic ischaemic stroke is discussed.

  • cerebrovascular diseases
  • Mendelian disorders
  • candidate genes
  • genetics
  • ACE = angiotensin-converting enzyme
  • ANF = atrial natriuretic peptide
  • CADASIL = cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy
  • EST = expressed sequence tag
  • MELAS = mitochondrial encephalomyelopathy, lactic acidosis and stroke-like episodes
  • QTL = quantitative trait locus/loci
  • SHR = spontaneously hypertensive rat


Ischaemic stroke can be the presenting feature of a number of single-gene disorders, but much more important on a population level are the sporadic forms of this disease. The aetiology in these cases is multifactorial and, whilst classical forms of inheritance cannot be demonstrated, recent evidence strongly suggests the importance of genetic factors. In the first part of this article we will focus on some of the single-gene disorders associated with ischaemic stroke. In the second part we will review the evidence that genetic factors are important in the pathogenesis of common ischaemic stroke, including evidence from both epidemiological studies and observations in animal models. We will then review candidate gene studies linking specific genes with the risk of ischaemic stroke. Finally, we will discuss potential molecular approaches, both those currently available and novel strategies, which may be suitable for the identification of stroke susceptibility genes.

Human single-gene disorders associated with stroke

Several rare Mendelian traits arising from a single-gene defect have been described in which stroke is a prominent feature (Natowicz and Kelley, 1987). These diagnoses should be considered in the differential diagnosis of any young patient presenting with stroke, or a young or middle-aged adult who lacks the usual risk factors, particularly in the presence of a family history of young-onset stroke. In these cases, diagnosis is frequently aided by characteristic clinical phenotypes, often including manifestations of co-existent systemic disease. Cerebral infarction can result from one of several pathophysiological mechanisms which might be influenced by a single-gene defect (Table 1). These include cardioembolism, large arterial disease, haematological disorders, small vessel disease, mitochondrial disorders, ion channel disorders, and connective tissue disorders leading to arterial dissection. Some familial conditions predispose to stroke by more than one mechanism.

View this table:
Table 1

Mendelian disorders associated with stroke

CADASIL = cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy; MELAS = mitochondrial encephalomyelopathy, lactic acidosis and stroke-like episodes
CardioembolicCardiomyopathies: primary/secondaryNatowicz and Kelley (1987)
Familial atrial myxoma
Familial dysrhythmias
homocysteinuriaSee text
dyslipidaemiasThird et al. (1984); Natowicz and Kelley (1987); Brown et al. (1983)
sickle cell diseasePowars et al. (1978); Wood (1978); Adams et al. (1992)
Prothombotic states:See text
protein S deficiency
antithrombin III deficiency
protein C deficiency
Small vessel diseaseCADASILSee text
Fabry's diseaseCrutchfield et al. (1998)
Mitochondrial disordersMELASSee text
Arterial dissectionMarfan syndromeSchievink et al. (1990); Spittell et al. (1993); Austin and Schaefer (1957)
Ehlers–Danlos syndromeSchievink et al. (1994)
ChannelopathiesFamilial hemiplegic migraineOphoff et al. (1996); Joutel et al. (1993)

Cardioembolic stroke

Cardiac dysfunction frequently leads to embolic stroke. Single-gene disorders in this category include familial atrial myxomas, hereditary cardiac conduction defects and inherited cardiomyopathies. The latter may present as a primary cardiac disorder (e.g. hypertrophic obstructive cardiomyopathy) or be secondary to a neuromuscular disorder (e.g. Duchenne muscular dystrophy) or one of the inborn errors of metabolism (e.g. Menkes disease). The associated risk varies widely according to the disorder. For example, recent evidence suggests that idiopathic autosomal dominant mitral valve prolapse may confer a negligible risk of cardioembolic stroke (Gilon et al., 1999), whilst patients with forms of autosomal dominant or recessive dysrhythmias and familial atrial myxoma may be at very high risk (Natowicz and Kelley, 1987).

Large artery disease

Metabolic disorders damaging the intra- or extracerebral vessels may lead to atherosclerosis and thromboembolism or haemodynamic insufficiency. Important single-gene disorders include homocysteinuria and dyslipidaemias. A number of conditions, such as homocysteinuria and sickle cell disease (see section headed Haematological disorders), may result in stroke via both arterial disease and a prothrombotic tendency.


Several autosomal dominant and recessive enzyme deficiencies exist which can lead to a high level of homocysteine in both plasma and urine, and this condition is referred to as homocysteinuria. It should be distinguished from milder hyperhomocysteinaemia, without homocysteinuria, which is a risk factor for ischaemic stroke on a population basis. The full homocysteinuria phenotype consists of mental retardation, ectopia lentis, skeletal deformities and thromboembolic vascular events. The most common underlying biochemical defect is the absence of the enzyme cystathione β-synthase, which converts homocysteine to cystathione, and this results in high plasma levels of homocysteine, methionine and homocysteine–cysteine mixed disulphide. More rarely, homocysteinuria may also result from a defect in the remethylation of homocysteine arising from a deficiency in methionine synthase or methylene tetrahydrofolate reductase. In patients with cystathione β-synthase deficiency, two forms of the phenotype can be distinguished according to the patient's responsiveness to treatment with the coenzyme precursor pyridoxine. Patients with the non-responsive form appear to have a more severe clinical presentation (Mudd and Levy, 1983). Homocysteine is believed to be toxic to endothelial cells and to predispose to a prothrombotic state, and it is associated with premature atherosclerosis. According to one study, one-half of the patients with inherited cystathione β-synthase deficiency suffered from a thromboembolic episode before the age of 29 years, and in 32% of these cases this was a cerebrovascular event (Mudd et al., 1985). The molecular genetic basis of homocysteinuria has been well characterized. The cystathione β-synthase gene maps to chromosome 21q22.3 (Kraus et al., 1993), and there are several relatively common mutations. The I278T and A114V mutations are widely distributed and frequently found in pyridoxine responders, whilst the G307S and A1224–2C mutations occur more frequently in northern European populations and are associated with a lack of pyridoxine responsiveness (Sebastio et al., 1995). Less frequently occurring mutations have also been described, and interestingly most of the disease alleles cluster in exons 3 and 8. Homocysteinuria usually occurs in individuals homozygous for the mutation, although it has been reported in the heterozygous state for some cystathione β-synthase mutations (Sebastio et al., 1985).


Hereditary dyslipidaemias associated with premature atherosclerosis may lead to early stroke. The relationship between these disorders and ischaemic stroke is less well defined than that with coronary artery disease, but a link has been reported in several disorders, including familial hypoalphalipoproteinaemia, familial hypercholesterolaemia (homozygous form), type II and type IV hyperlipidaemia and Tangier's disease (Brown et al., 1983; Third et al., 1984; Natowicz and Kelley, 1987).

Haematological disorders

Haemoglobinopathies, including sickle cell disease and various inherited coagulopathies, are associated with thrombotic cerebral infarction. Stroke is an important complication of sickle cell disease. It usually occurs in childhood, affecting ~8% of children with homozygous sickle cell disease in the first 14 years of life (Balkaran et al., 1992). Ischaemic stroke is infrequently seen in haemoglobin sickle cell disease and is extremely rare in sickle cell trait. Several mechanisms predispose sickle cell patients to an increased risk of ischaemic stroke; these include sickling of red blood cells during a crisis and thrombotic infarction of both small and large vessels. In addition, narrowing of the large extracranial and intracranial vessels occurs secondarily to fibrous intimal proliferation. The large-vessel stenotic lesions can be identified using transcranial Doppler ultrasound, and this technique is helpful in predicting which patients are at high risk of cerebral infarction; for such patients a programme of exchange transfusion is beneficial (Adams et al., 1998).

A prothrombotic state resulting from a deficiency of the natural anticoagulants protein C and protein S is a well-recognized cause of familial venous thrombosis. The association with arterial stroke is less strong. Proving a causal association in an individual patient is complicated by the fact that reduced levels of proteins C and S may occur transiently after stroke, while low levels may been seen in other conditions, including liver disease, disseminated intravascular coagulation and renal disease, or in patients on warfarin therapy. Therefore, it may be necessary to confirm an underlying gene defect by serial sampling of levels and screening other family members (Markus and Hambley, 1998). It appears that these prothrombotic tendencies are primarily a risk factor for arterial stroke in the young, and the relationship has been described in several case reports. In cases where there is also a family history of premature thrombosis in other family members, the association is likely to be causal (Kohler et al., 1990; Barinagarrementeria et al., 1994). In elderly individuals, the relationship between the levels of these natural anticoagulants and stroke appears to be weak (Markus and Hambley, 1998). In addition, the risk of venous thrombosis is much higher than that of arterial thrombosis, and the possibility of a patent foramen ovale leading to a right-to-left cardiac shunt should therefore be considered in patients with stroke. The risk of cerebral venous thrombosis is also increased. Antithrombin III deficiency, inherited as an autosomal dominant trait, is also primarily associated with venous thrombosis, although rare cases of arterial stroke have been reported (Vomberg et al., 1987).

In 1993, a new form of familial thrombophilia was recognized, detectable on functional assays as a resistance to activated protein C (Dahlback et al., 1993). Subsequent DNA analysis revealed that, in the majority of cases, this was due to an A→G substitution at position 1691 (Q506 Leiden) of the selective coagulation factor V gene (Bertina et al., 1994). The heterozygous state is found in 3–8% of the normal population. This appears to be the commonest inherited cause of venous thrombosis and is an important risk factor for cerebral venous thrombosis, but large case–control studies have failed to find an increased frequency of the factor V Leiden mutation compared with published population frequencies and with age-matched controls (Table 3). However, there are a number of kindreds described with inherited activated protein C resistance and stroke occurring at a young age (Simioni et al., 1995; De Lucia et al., 1997). Similarly, a mutation in the non-coding 3′-terminal end of the prothrombin gene associated with increased protein expression (G20210A) has been linked to venous thrombotic events (Poort et al., 1996), although most of the data emerging on the role of this variant in ischaemic stroke are negative. The sole exception was a single case–control study in young patients (De Stefano et al., 1998). Several case reports have suggested that epistatic (gene–gene) interactions may increase the risk of arterial thrombosis synergistically in individuals carrying prothrombotic mutations and/or hyperhomocysteinaemia, as demonstrated in families with combined disorders (Franken et al., 1993; Koller et al., 1994; Mandel et al., 1996).

Although most cases of the anticardiolipin antibody and lupus anticoagulant syndrome are sporadic, several families with the syndrome have been described (Mackie et al., 1987). Similarly, Sneddon's syndrome, characteristically presenting with livedo reticularis and ischaemic cerebrovascular disease, has been described in several families and probably has an autosomal dominant pattern of inheritance (Pettee et al., 1994).

Small vessel disease

Small vessel disease is a pathological term used to refer to the structural alterations affecting the small penetrating end arterioles which supply the deep white matter and basal ganglia. The clinical phenotype arising from this type of injury may take one of several forms, most commonly specific lacunar stroke syndromes (Bamford et al., 1991). Multiple lacunar infarcts, with or without more diffuse ischaemic changes in the penetrating arterial territories, may present with cognitive impairment, pseudobulbar palsy and disorders of gait. These changes are best seen on MRI as focal or diffuse hyperintensities on T2-weighted sequences. Several single-gene disorders can be identified with this phenotype.

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy (CADASIL)

Cases of familial vascular dementia began to appear in the literature in 1977, notably descriptions by Sourander and Walinder (Sourander and Walinder 1977) and Stevens and colleagues (Stevens et al., 1977). By 1993, several similar families had been reported across Europe using a variety of terms, including hereditary multi-infarct dementia (Sourander and Walinder, 1977), autosomal dominant syndrome with stroke-like episodes and leucoencephalopathy (Tournier-Lasserve et al., 1991), chronic familial vascular encephalopathy (Stevens et al., 1977), familiare zerebrale arteriosclerose (Gerhard, 1980), demence sous-corticale familiale avec leucoencephalopathie arteriopathique (Davous and Fallet-Bianco, 1991), familial disorder with subcortical ischaemic strokes, dementia and leucoencephalopathy (Mas et al., 1992), and slowly progressive familial dementia with recurrent strokes and white matter hypointensities on CT scan (Salvi et al., 1992). In 1993, the acronym CADASIL, an abbreviation for cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy, was adopted to designate a form of this disorder as specified by genetic homogeneity (Tournier-Lasserve et al., 1993). This disease has now been reported from many regions worldwide, including Europe (Jung et al., 1995; Bergmann et al., 1996), North America (Hedera and Friedland, 1997; Desmond et al., 1998) and the Far East (Nishio et al., 1997).

The clinical phenotype is characterized by subcortical stroke-like episodes, which occur usually in mid-adulthood in the absence of normal vascular risk factors, and a stepwise subcortical dementia with pseudobulbar palsy (Chabriat et al., 1995b; Dichgans et al., 1998). Migraine and psychiatric disturbance are features which usually occur earlier in life and are the prominent feature in some families (Chabriat et al., 1995a; Verin et al., 1995). Epilepsy has also been reported (Malandrini et al., 1997; Dichgans et al., 1998). The average life expectancy of these patients is 65 years. MRI reveals leucoencephalopathy and small deep infarcts in all symptomatic patients, and these may be also found in asymptomatic individuals. Both focal lacunar infarcts and more diffuse leucoaraiosis are seen on T2-weighted images (Fig. 1), and are found in the deep white matter and periventricular regions (Chabriat et al., 1998). These changes have also been reported in the brainstem deep white matter but not the cerebellar cortex (Chabriat et al., 1999). The extent of the changes is proportional to age, and the lesion load appears to correlate with the level of disability and impairment of cognitive performance (Dichgans et al., 1999).

Fig. 1

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy (CADASIL). A T2-weighted MRI scan showing both confluent regions of periventricular and deep white matter high signal, referred to as leucoaraiosis, in addition to small punctate regions of high signal consistent with lacunar infarction. This patient had an exon 4 mutation in the Notch 3 gene.

The underlying pathology consists of a non-arteriosclerotic, non-amyloid angiopathy involving the media of the cerebral small vessels. Histopathological studies reveal concentric thickening of the arterial walls with extensive granular eosinophilic material deposited within the media and reduplication of the internal elastic lamina. In the white matter, there is diffuse and focal pallor of myelin staining with destruction of nerve fibres and accompanying gliosis. The frontal, parietal and occipital lobes are most commonly affected and show a symmetrical pattern. These changes are accompanied by multiple infarcts in the deep white matter, thalamus and basal ganglia at different stages of development. Electron microscopy of the vessels reveals patches of granular osmiophilic material in close association with focal destruction of the vascular smooth muscle (Baudrimont et al., 1993). The identity of this granular osmiophilic material is unknown, but it may represent abnormal condensation of vascular smooth muscle cytoplasm (Ruchoux and Maurage, 1997). Interestingly, these changes are not restricted to cerebral arterioles. Similar changes have been reported in the small vessels supplying the peripheral nerves, skin, muscle, and occasionally the viscera, indicating the presence of systemic arteriopathy (Ruchoux et al., 1995; Schroder et al., 1995). However, the destruction of vascular smooth muscle is less marked in the peripheral vasculature (Ruchoux and Maurage, 1997). These morphological observations suggest that the intrinsic angioarchitecture of the endothelial blood–brain barrier may be an important factor leading predominantly to subcortical involvement. Vascular smooth muscle cells in the cerebral small vessels are completely dependent on normally functioning endothelium, and it has been hypothesized that impairment of blood–tissue exchanges leads to severe cellular destruction. Certainly there is evidence of marked endothelial abnormalities in peripheral tissue which would be consistent with this hypothesis (Ruchoux and Maurage 1998). It is speculated that, later in the disease process, there may be breakdown of the blood–brain barrier leading to the formation of lacunae, but the role of the granular osmiophilic material in all these processes has yet to be resolved. By mechanisms similar to those proposed for the more commonly occurring form of ischaemic white matter disease (or leucoaraiosis) seen in hypertensive individuals, it is hypothesized that damage to cerebral small vessels leads to impaired autoregulation, poor cerebral perfusion and ischaemic demyelination of the white matter (Pantoni and Garcia, 1997).

The underlying genetic mutation was identified using a linkage strategy. Linkage analysis of two unrelated families allowed the disease locus to be assigned initially to 19q12 (Tournier-Lasserve et al., 1993). Using a similar approach, it was found that another Mendelian disorder, familial hemiplegic migraine, could be mapped to the same locus (Joutel et al., 1993), suggesting that allelism of the same gene could be responsible for both conditions. This was later disproved when linkage analysis of further pedigrees allowed the critical region to be further refined. A sequence of 8 cM (centimorgans) was defined in CADASIL (Joutel et al., 1996; Dichgans et al., 1996), within which the human equivalent of the mouse Notch 3 gene was identified. This was proposed as a candidate gene on the basis of similarity between another gene involved in Notch signalling, Sel 12, and S 182, a gene implicated in Alzheimer's disease (Levitan and Greenwald, 1995). Identification of deleterious mutations confirmed Notch 3 as the causative gene. Subsequently, mutations of the P/Q-type Ca2+ channel α1 subunit gene (CACNL1A4) were identified as being responsible for familial hemiplegic migraine (Ophoff et al., 1996). However, it is likely that other loci may be responsible in some families with this CADASIL phenotype, in a manner analogous to that seen in familial Alzheimer's disease. Recently, families with the typical phenotype have been reported in whom no mutations in the Notch 3 gene could be detected (Uyama et al., 1999).

The Notch 3 gene is believed to encode a transmembrane receptor containing several tandemly repeated copies of an epidermal growth factor-like repeat. Notch signalling is a highly conserved pathway involved in cell signalling and fate during embryonic development (Artavanis-Tsakonas et al., 1995), but the role of Notch 3 in normal adult smooth muscle physiology is unknown. Recent work has demonstrated that the protein is expressed mainly on normal vascular smooth muscle membranes (Joutel and Tournier-Lasserve, 1998), and one tentative hypothesis is that Notch signalling may be important in maintaining vascular smooth muscle in a terminal differentiated state. Numerous different mutations have been identified within affected families, but to date no genotype–phenotype correlations have been established. Studies in unrelated individuals reveal strong clustering of the mutations, 70% of them occurring in exons 3 and 4 (Joutel et al., 1997). It has also been found that the mutation can occur spontaneously in individuals (Jontel et al., 2000), an observation which raises important issues about diagnostic testing for Notch mutations in the wider stroke population. Almost all cases described to date are in individuals heterozygous for Notch 3 mutations, although one homozygous case has recently been described; the individual also had adult-onset disease, but possibly with a slightly earlier age of symptom onset (Tuominen et al., 1999).

Strikingly, mutations lead to either loss or gain of a cysteine residue. Such mutations might alter the conformation of the protein, interfering with ligand–receptor interaction, perhaps causing a toxic gain of function. Alternatively, mutations lead to unpaired cysteine residues, which might result in abnormal homodimers or heterodimers accumulating within vascular smooth muscle cells and disrupting them. Future studies will be directed towards understanding the normal function of Notch 3 and how disruption may lead to disease. This in itself may provide useful insights not only into the pathogenesis of CADASIL, but perhaps also into the more common forms of stroke and vascular dementia due to small vessel disease.

In individuals with a typical phenotype and family history, the most efficient approach to screening is to look initially for mutations in exons 3 and 4, and also to consider a skin biopsy. Although it is diagnostic if granular osmiophilic material is present, the skin biopsy can be normal (Ebke et al., 1997). If these tests are negative and the index of suspicion is high, screening of the remaining exons can be performed with single-strand conformation polymorphism methods and direct sequencing, although this is time-consuming. Alternative genetic screening methods, such as chemical cleavage methods using RNA, may prove more efficient (Rowley et al., 1995).

Fabry's disease

This is an X-linked recessive disorder due to α-galactosidase A deficiency. Progressive accumulation of ceramidetrihexoside within the intima and media of blood vessels results in luminal narrowing and complications such as stroke and myocardial ischaemia. Other features of the phenotype include angiokeratomata, painful acroparaesthesiae and proteinuria. Stroke occurs most frequently from the third decade onwards. It may occur secondarily to large vessel disease, small vessel disease, or embolism from associated cardiac disease. Involvement of large vessels appears to affect the vertebrobasilar system preferentially, and therefore ischaemia is most common in this territory. The cerebral vasculopathy is well visualized on T2-weighted MRI, which initially demonstrates periventricular hyperintensities and lacunar infarcts indicative of small vessel disease, and subsequently progresses with age, affecting the larger vessels and leading to symptomatic infarcts which appear in the cortical grey matter (Crutchfield et al., 1998).

Mitochondrial disorders

Mitochondrial encephalomyelopathy, lactic acidosis and stroke-like episodes (MELAS) is a mitochondrial encephalopathy that is characterized by seizures, stroke-like episodes, migraine-like headaches, nausea, vomiting, lactic acidosis, external ophthalmoplegia, ptosis, sensorineural hearing loss and dementia. Cardiac abnormalities and endocrine disturbances can also occur (Graeber and Muller, 1998). Muscle biopsy usually reveals abnormal mitochondria and ragged red fibres. The cerebral lesions most commonly affect the occipital-parietal regions, often resulting in visual field defects, and frequently they do not respect the borders of typical vascular territories. They may show marked resolution on subsequent MRI, as shown in Fig. 2. Lactate can be demonstrated on proton nuclear magnetic resonance spectroscopy in the ischaemic lesions. During the acute phase, increased signal is seen on diffusion-weighted imaging, which is consistent with cytotoxic oedema. The underlying cause of this disorder is mutations in the mitochondrial DNA, which in man are transmitted maternally. MELAS, like many mitochondrial disorders, demonstrates the phenomenon of heteroplasmy. There is variation in the expression of mutated DNA within different tissues, which is thought to arise because of random replicative segregation of mitochondria during germ-layer differentiation early in embryonic development. It is believed to be an important cause of the phenotypic heterogeneity associated with mitochondrial mutations within the same family (de Vries et al., 1994), and such heterogeneity may result in the lack of a family history. MELAS mutations are usually, but not exclusively, missense and lie within the tRNA-leu gene (UUR). Most frequently reported is an A→G transition at position 3243 (Enter et al., 1991) and a T→C transition at position 3271 (Sakuta et al., 1993). It has been proposed that mutations at position 3243 may result in somatic mutations accumulating in the mitochondrial DNA, leading to progressive mitochondrial dysfunction (Kovalenko et al., 1996).

Fig. 2

Mitochondrial encephalomyelopathy lactic acidosis and stroke-like episodes (MELAS). (AD) A temporal sequence of T2-weighted MRI scans from a patient with the typical clinical phenotype of MELAS, showing the appearance and marked regression of high-signal lesions in the posterior temporoparieto-occipital regions. (A) Time zero: first neurological episode. The scan shows a left occipital `stroke'. (B) Time 2 months: second neurological episode. The scan shows a new right occipital `stroke' but the previous lesion has regressed almost completely. (C) Time 5 months. The scan is almost normal. (D) Time 3 years, after a number of neurological events. The scan now shows high signal in the bilateral basal ganglia, which is characteristic of advanced disease. (E) During a neurological episode at 7 months, proton magnetic resonance spectroscopy was performed with the voxel placed in the middle of the abnormal left parietotemporal region. (F) The spectrum shows a large lactate peak identifiable by its bifid nature.

Arterial dissection

Ischaemic stroke may occasionally be the consequence of arterial dissection, and infrequently this may be a complication of an underlying heritable connective tissue disorder. Defects in collagen synthesis can predispose individuals to spontaneous dissection of the extracranial carotid and vertebral arteries in Ehlers–Danlos syndrome type IV (Schievink et al., 1990), whilst in Marfan syndrome the most common neurovascular complication is extension of an aortic dissection into the common carotid artery (Spittell et al., 1993). Spontaneous dissection limited to the common or internal carotid artery has also been reported in patients with Marfan syndrome (Austin and Schaefer, 1957), and there are isolated case reports of vertebral dissection in osteogenesis imperfecta (Schievink et al., 1994). In contrast, patients with pseudoxanthoma elasticum and neurofibromatosis tend to develop vaso-occlusive disease due to fibrointimal proliferation, which is believed to be the mechanism underlying cerebral infarction.

Other monogenic disorders resulting in ischaemic stroke

Familial amyloid angiopathy usually presents with intracerebral haemorrhage rather than ischaemic lesions. However, familial British dementia, an amyloid angiopathy, may present with white matter lesions that are seen as high intensity on T2-weighted imaging, without haemorrhage. Clinically stroke-like episodes and a progressive dementia occur. Inheritance is autosomal dominant (Plant et al., 1990). A stop-codon mutation in the novel BRI gene has been found recently in a family with this disease (Vidal et al., 1999).

The genetic basis of common multifactorial ischaemic stroke

Most cases of stroke represent a multifactorial disorder or complex trait for which classical patterns of inheritance cannot be demonstrated. The major risk factors for stroke, such as hypertension, cigarette smoking and diabetes mellitus, fail to account for a large portion of the risk of ischaemic stroke. It has been estimated that up to 69% of the population-attributable risk may be unaccounted for by these three risk factors alone (Jamrozik et al., 1994). Subsequent twin and family-based studies, together with observations in animal models, which are described below, have provided evidence that genetic factors influence the risk of stroke. As in other complex traits, the genetic aetiology of ischaemic stroke is likely to be polygenic, i.e. to reflect the influence of many different loci modulating different pathophysiological processes. The spectrum of disease alleles is probably wide, each individual gene conferring a small relative risk. However, the influence of an isolated gene can still be considered important, for several reasons. First, as sporadic ischaemic stroke is a common disorder, the population-attributable risk may be high, even for genetic variants conferring a modestly increased relative risk. Secondly, on an individual level the penetrance of a gene may be dependent on certain permissive backgrounds. For example, the presence of several genes may increase the risk of disease in an additive manner (a gene dose effect), or a gene may interact with another risk factor and modulate its effects. This could be an environmental trigger, such as smoking or diet, or another gene (i.e. an epistatic interaction). Often, such combinations are synergistic, the increase in risk being multiplicative. The modulatory effect of genes on the extent and pattern of end-organ damage resulting from conventional risk factors may be particularly important. For example, certain hypertensive individuals develop cerebral small vessel disease without any evidence of large vessel atherosclerosis, while in others the converse occurs, or a combination of the two patterns may be present. The reasons for these different patterns are unknown but, by analogy, it has been demonstrated that genetic factors may determine whether hypertensive individuals develop other manifestations of end-organ damage, such as cardiac left ventricular hypertrophy (Celentano et al., 1999).

The heterogeneity of common ischaemic stroke

Ischaemic stroke is a syndrome and not a single disease state. Recent advances in imaging allow the underlying pathogenic stroke mechanism to be determined in many cases. Although there is limited epidemiological work comparing even the conventional risk factor profiles of different subtypes of stroke, it is likely that they will differ significantly, and particular genetic factors may predispose individuals to specific subtypes of stroke. Therefore, genetic epidemiological studies should determine the subtype of stroke in individual patients, although this may not be possible in as many as 30% of cases. Classification systems defined on the basis of common pathophysiological mechanisms, such as the TOAST (trial of org10172 in acute stroke treatment) classification system (Adams et al., 1993), have been used frequently. Phenotypes can be subdivided into stroke due to cardioembolism, stroke secondary to intracerebral and extracerebral large artery disease, and small vessel stroke arising from occlusion of the deep perforating arterioles. Another popular system in use is the Oxford Community Stroke Project Classification (Bamford et al., 1991). This is primarily a clinical rather than a pathophysiological classification and, although it may allow a pathological subtype to be inferred, it is insufficiently precise for accurate phenotyping. For example, a clinical lacunar syndrome can result from a striatocapsular infarct secondary to embolism from a carotid stenosis, and the correct phenotype of large artery disease can only be determined after brain and carotid artery imaging. However, with any current classification system a significant proportion of strokes are of unknown cause.

For each of these stroke subtypes, genetic factors may act either by predisposing to conventional risk factors, such as hypertension, by modulating the effects of such conventional risk factors on the end organs, or by a direct independent effect on stroke risk. The well-documented conventional risk factors, such as hypertension, hyperlipidaemia and diabetes, are themselves believed to be `synthetic traits' under genetic control (Havlik et al., 1979; Barnett et al., 1981; Hunt et al., 1989; Kiely et al., 1993). However, the emphasis of this review will be on stroke-causing genes that do not contribute directly to these phenotypes, but have either a modulatory or an independent effect on the risk of stroke.

Multiple sites of action of stroke susceptibility genes and the use of intermediate phenotypes

A cerebral infarct is the end result of a number of pathological processes, each of which may be under genetic influences. For example, in a patient with stroke secondary to carotid stenosis, genetic factors could influence the development of atherosclerosis, the processes leading to plaque disruption and thromboembolism, the patency of the circle of Willis, and therefore the effect of a carotid occlusion on middle cerebral artery flow, and the response of brain tissue to ischaemia. This complexity may make associations with specific genes difficult to detect, particularly if each step of the process is under the influence of a number of different genes. A logical step is the use of intermediate phenotypes. These represent specific components of the disease process, and therefore the number of genes involved in their pathogenesis is likely to be much reduced compared with those involved in ischaemic stroke. In vivo imaging has provided a number of potential intermediate phenotypes. Common carotid artery intimal medial thickness, determined by ultrasonography, has been used widely as a measure of early carotid atherosclerosis, while carotid plaque size can also be used as an estimate of more advanced disease. Carotid ultrasound has been used widely in this context to determine the role of a variety of conventional risk factors, novel risk factors such as chronic infection and inflammation, and genetic risk factors in the pathogenesis of carotid atherosclerosis (Crouse and Thompson, 1993). These estimates of intimal medial thickness and ultrasonically identified carotid plaques may be useful intermediate phenotypes for large artery stroke. Silent white-matter hyperintensities seen on T2-weighted MRI may provide a useful intermediate phenotype for small vessel disease stroke.

Evidence for the role of genetic factors in ischaemic stroke

Both epidemiological and animal-based studies provide strong evidence that genetic factors are important in the pathogenesis of stroke.

Epidemiological studies

Epidemiological studies have used twin, affected sibling pair and family-based approaches. Twin studies provide the most robust evidence for genetic influences in stroke. The principle is the comparison of concordance rates between monozygotic and dizygotic twins for a disorder. It is assumed that, apart from genetic factors, monozygotic and dizygotic twins will be similar in other respects, such as environmental exposures. From the degree of concordance it is then possible to determine the heritability of a disorder, defined as the proportion of the phenotype that can be attributed to genetic factors (Hrubec and Robinette, 1984). Using this approach, Brass and colleagues (Brass et al., 1992) found a concordance rate of 17.7% in monozygotic twins as opposed to 3.6% in dizygotic twins, giving a relative risk of 4.3. However, as only a few twin pairs were studied it was not possible to calculate the heritability of stroke in this study. Furthermore, as the study was based on a questionnaire no conclusions could be drawn regarding the stroke phenotypes represented in the twin pairs. This cohort was re-evaluated a decade later, when genetic influences appeared to have less influence on the risk of stroke in the older population, in whom much of the variance was accounted for by environmental exposure (Brass et al., 1998). This is consistent with the role of genetic stroke risk factors being strongest in younger adults.

Twin and sibling studies have also shown that the intermediate phenotypes for stroke are under strong genetic control. A twin approach has recently been employed to determine the relative contributions of genetic influences on the volumes of white matter hyperintensities seen on MRI in healthy elderly individuals (Carmelli et al., 1998). They found concordance rates of 0.61 in monozygotic twins as opposed to 0.38 in dizygotic twins. In their model, 71% of the variation in white matter hyperintensity volume was accounted for by genetic factors behaving in an additive manner. In another study, the heritability of common carotid artery intimal medial thickness was found to be 92% (Duggirala et al., 1996). Interestingly, the contribution of genetic factors was not through conventional risk factors used as covariates in their analysis. However, this study was performed in sibs rather than twins, making it difficult to determine the relative roles of early shared environmental effects and genetic influences.

Family-based studies have examined the relationship between a family history of stroke amongst first-degree relatives and risk of stroke in probands (Table 2). Whilst a positive family history is consistent with the role of genetic factors, it must be remembered that alternative, but not necessarily exclusive, explanations, such as shared environmental influences, could be valid. With this proviso, the results of several large studies are consistent with a family history of stroke being an important independent risk factor. For example, although in the original Framingham cohort it was reported that a parental history of stroke did not confer an increase in the risk of stroke, amongst the offspring cohort there was an association with verified parental stroke history (Kiely et al., 1993). The adjusted relative risks in probands for paternal stroke and maternal stroke were 2.4 and 1.4, respectively. The presence of an atherothrombotic brain infarction in a sibling also conferred a relative risk of 1.8. However, the confidence intervals were wide because there were few strokes documented in this relatively young cohort. It will be interesting to re-evaluate the importance of family history in this cohort in the future. The importance of parental history was also seen in a prospective follow-up of a Finnish population (Jousilahti et al., 1997), in which a positive parental history of stroke led to a twofold increase in stroke in both men and women. The association between family and personal history of stroke appeared to be stronger in younger stroke patients. This was also found in a cross-sectional study (Howard et al., 1990), in which younger stroke victims were more likely to have an offspring who had a fatal coronary or cerebrovascular event. A positive stroke history was recorded in 47% of patients, compared with 24% in the study by Graffagnino and colleagues (Graffagnino et al., 1994). However, the difference was no longer significant after controlling for conventional risk factors. Furthermore, as hypertension and diabetes are known to have a strong genetic component (Havlik et al., 1979; Barnett et al., 1981; Hunt et al., 1989; Kiely et al., 1993), clustering of these risk factors within families may partly account for the familial aggregation of stroke. This has been suggested by the study of Diaz and colleagues (Diaz et al., 1986), in which it was found that siblings of patients with cerebral infarction or transient ischaemic attack were more likely to have multiple vascular risk factors than the siblings of spouse controls. There have also been some notable negative studies (Herman et al., 1983; Boysen et al., 1988). However, most studies suggest that a family history of stroke is an independent risk factor for stroke, and this is consistent with a genetic component operating outside the usual risk factors.

View this table:
Table 2

Epidemiological studies investigating the role of genetic factors in commonly occurring ischaemic stroke

Figures in parentheses denote 95% confidence intervals. MZ = monozygotic, DZ = dizygotic, WMH = white matter hyperintensities, CCA IMT = common carotid artery intimal medial thickness, OR = odds ratio, RR = relative risk, TIA = transient ischaemic attack.
Brass et al. (1992)Twin2722 twins identified from twin registry, 7 MZ twin pairs concordant, 65 discordant 1 DZ twin pair concordant, 53 discordantPositive4.3-fold increase risk in monozygotic twins
Brass et al. (1998)Twin follow-up study21 MZ twin pairs concordant, 205 discordant 23 DZ twin pairs concordant 204 discordantNegativeGenetic influences comparatively less in an older twin population
Kiely et al. (1993)Cohort4933 cohort individuals, comprising 604 sibships 2317 offspring cohort individualsPositiveIn offspring cohort:
adjusted RR if paternal history 2.4 (0.96–6.03)
adjusted RR if maternal history 1.4 (0.6–3.25)
Verified sibling atherothrombotic stroke adjusted RR 1.8 (0.68–4.94)
Khaw et al. (1986)Cohort1491 men, 1491 womenPositiveFamily history of stroke independent predictor of stroke mortality in women only; adjusted RR 2.3
Jousilahti et al. (1997)Cohort14 371 middle-aged men and womenPositiveAdjusted RR if paternal history 1.89 in men, 1.80 in women; association strongest in 25–49 age group
Howard et al. (1990)Cross-sectional55 probands selected from a stroke registry 243 offspringPositiveAn association was found between young onset of stroke in probands and fatal coronary or vascular events in the offspring
Welin et al. (1987)Cohort789 men born in 1913PositiveMaternal history of stroke mortality associated with adjusted RR of 3
Boysen et al. (1988)Cohort19 327 individualsNegative
Brass and Shaker (1991)Cross-sectional117 individuals discharged from hospital with TIAPositiveThere was an association between family history and stroke in patients aged >70 years
Liao et al. (1997)Cross-sectional3168 probands, 29 325 first degree relativesPositivePaternal history adjusted OR 2 (1.13–3.54) maternal history adjusted OR 1.41 (0.80–2.50)
Diaz et al. (1986)Case–control76 siblings of stroke patients; 55 siblings of patients' spousesPositiveClustering of 2 or 3 vascular risk factors amongst siblings of stroke patients
Graffagnino et al. (1994)Case–control85 patients, 86 controlsPositiveNot as independent risk factor, 47% of stroke patients had a first-degree relative affected compared with 24% of controls. Positive family history had a higher positive predictive value for presence of one or more vascular risk factors
Herman et al. (1983)Case–control132 stroke patients, 239 controlsNegative

Interestingly, amongst the positive studies there have been discrepant reports concerning how this increase in risk is transmitted. For example, from one study it was reported that a maternal history of stroke was associated with a threefold increase in stroke in a cohort of men followed up since 1913 (Welin et al., 1987). This is at odds with results from the Family Heart Study and the Framingham cohort (Kiely et al., 1993; Liao et al., 1997), in which it was found that individuals with a paternal history of stroke were more likely to have a stroke than those with a maternal history, which conferred a slightly lower risk (Liao et al., 1997). Interestingly, in the Rancho Barnardo study (Khaw and Barrett-Connor, 1986), a family history of stroke in any first-degree relative at baseline was an independent predictor of stroke mortality in women but not in men, which suggests a sex-specific interaction for the genetic risk of stroke.

There are a number of explanations for differences between studies, including difficulties in ascertaining family history, different methods, including the study of different populations, and in some cases small sample sizes. A further problem is the difficulty in ascertaining the exact stroke phenotype amongst affected family members. Only a few studies have attempted the detailed clinical evaluation of probands. Consequently, some will have incorporated cases of intracerebral haemorrhage as well as ischaemic stroke in their analyses, and differentiating between the two in dead relatives with stroke is often impossible.

An animal model of human stroke: the stroke-prone spontaneously hypertensive rat

Recent observations in animal models have provided strong evidence for the existence of stroke susceptibility genes. A well-established experimental tool in the study of hypertension has been the spontaneously hypertensive rat (SHR) (Okamoto and Aoki, 1963). However, whilst this is a good model for evaluating the response of blood pressure to a range of pharmacological agents and the development of left ventricular hypertrophy, these animals fail to develop stroke. An important exception was observed during the establishment of colonies of SHR rats, in which some animals tended to die at a very early age after the introduction of a stroke-permissive `Japanese' rat chow diet which was high in sodium but low in protein and potassium. These observations led to the development of a separate inbred strain of hypertensive animals, named the stroke-prone spontaneously hypertensive rat (Okamoto et al., 1974). The phenotypic difference between these two strains is believed to reflect the segregation of genes at an early stage during inbreeding which confer susceptibility to stroke but not hypertension. The isolation of these genetic factors is discussed in the next section.

Identifying genetic factors in ischaemic stroke

Determining the precise nature of genetic factors at a molecular level in complex traits is a difficult task (Lander and Schork, 1994). Linkage paradigms, which rely on predictable patterns of cosegregation of markers with phenotypes, have been employed successfully to identify the responsible gene mutations in large pedigrees of affected and unaffected individuals with single-gene disorders. In polygenic stroke the situation is more difficult because of several factors: (i) late onset: the late onset of stroke makes genetic comparisons between living relatives difficult; (ii) phenotypic heterogeneity: the variety of stroke subtypes or phenotypes is likely to reflect different aetiologies; (iii) genetic heterogeneity: mutations in any one of several genes might result in an identical phenotype; (iv) phenocopy: some individuals who do not inherit a predisposing allele will still manifest stroke because of random or environmental causes; (v) variable penetrance: some individuals who inherit a predisposing allele may not manifest the disease; causes of variable penetrance include gene dose, gene–environment interaction and epistatic phenomena; (vi) confounders: the presence of coexistent risk factors, such as hypertension and diabetes, may make the effects of a single gene difficult to assess in affected individuals.

Two major lines of investigation have been employed to determine the identity of the genes responsible; quantitative trait locus mapping in stroke-prone animals and candidate gene studies in man.

Experimental crossbreeding and quantitative trait locus mapping

Early studies suggested that susceptibility to infarcts after middle cerebral artery occlusion in stroke-prone SHR rats was related to a single gene, transmitted in an autosomal recessive manner, which affected collateral flow (Coyle and Heistad, 1991). However, heterogeneity between animal strains precluded any definitive comparison of genetic differences. One approach to overcoming this problem was to establish a hybrid line of animals from the mating of stroke-prone SHR rats with SHR rats. It was hoped that the offspring of these experimental crosses would retain homogeneity for hypertensive alleles whilst stroke susceptibility genes would tend to cosegregate. This hypothesis proved to be correct, the F2 hybrids from these crosses showing marked variation in stroke susceptibility despite concordance for hypertension. By generating large numbers of progeny from experimental crosses, genome-wide screening approaches enable identification of alleles which cosegregate accordingly with a quantitative trait. The position of quantitative trait loci (QTL) can then be inferred from known genetic maps. QTL mapping is a powerful approach that has been used by several investigators with various modifications (Ikeda et al., 1996; Rubattu et al., 1996; Jeffs et al., 1997). Three major QTL, designated STR1, STR2 and STR3, were reported to influence stroke risk, measured as time to stroke onset in rats fed a stroke-permissive diet (Rubattu et al., 1996). These accounted for 28% of the overall variance in the risk of stroke. Comparison with human and mouse genetic maps revealed no obvious candidate genes localizing to the regions mapping to STR1 and STR3 (chromosomes 1 and 4). However, a protective locus, STR2, was found to map closely to a marker derived from the gene for atrial natriuretic factor (ANF), located on chromosome 5. The presence of one or two at-risk alleles at STR2 was associated with increased stroke risk in an additive manner. A further interesting observation in this study was that of the existence of epistatic interaction between alleles at the STR1 and STR2 loci, suggesting that the phenotype studied was a good model of human polygenic stroke.

A slightly different approach has been employed by Ikeda and colleagues, who obtained F2 hybrids by crossing stroke-prone SHR rats with Wistar Kyoto rats (Ikeda et al., 1996). They found that, in hybrids developing stroke, the brain was always heavier as a result of oedema, and therefore brain weight was used as the quantitative parameter for linkage analysis. They found no evidence of markers contributing to stroke on chromosome 5, but found a gene on chromosome 4, distinct from STR3, that contributed to the phenotype independently of hypertension. A further study determined the severity of stroke, estimated as infarct volume after a standard focal ischaemic insult, rather than the risk of stroke itself (Jeffs et al., 1997). Again, F2 hybrids derived from similar crosses were used to isolate the genetic component responsible and to exclude the influence of hypertension. Again, a highly significant QTL was localized to chromosome 5 in the vicinity of genes encoding ANF and brain natriuretic factor, and it accounted for 67% of the phenotypic variance. No QTL were localized to chromosomes 1 and 4, and, surprisingly, the QTL that mapped to chromosome 5 conferred increased susceptibility to cerebral ischaemia rather than a protective effect. Further work is required to explain the different conclusions of these studies, although they may relate to the use of different phenotypes and the different genetic backgrounds of the animals used in crossbreeding.

The identification of the gene for ANF as a putative candidate gene is consistent with the known circulatory effects of the factor (Levin et al., 1998) and the finding of elevated levels of ANF in acute ischaemic stroke (Estrada et al., 1994). Earlier crossbreeding studies also demonstrated impaired endothelium-dependent vasodilatation in response to ANF as an important determinant of stroke in the Heidelberg colonies of stroke-prone SHR rats (Russo et al., 1998). Analysis of DNA has revealed functional mutations in the ANF gene in these F2 hybrids that alter the expression of peptide levels (Rubattu et al., 1998). Interestingly, a recent study in patients with stroke found an increased prevalence of a G664A polymorphism in the human ANF gene (Rubattu et al., 1999). No information is given in the paper about stroke subtype, and whether both cerebral haemorrhage and infarction were included. In contrast, recent structural and functional studies appeared to exclude the genes for ANF and brain natriuretic factor as candidate genes responsible for the ischaemic sensitivity trait studied in the Glasgow colonies of stroke-prone SHR rats (Brosnan et al., 1999).

Human studies: the candidate gene approach

The current mainstay of genetic studies of stroke in man is the association or candidate gene approach. This involves first identifying a molecular variant within a functionally relevant gene, and then determining its role in conferring stroke risk by looking for an association with the phenotype using a case–control or cohort method. A limitation of such studies is that they are based on the availability for testing of candidate genes a priori. However, the large number of potential candidate genes recently described has made this a practical approach. On its own, a positive association does not necessarily imply a true causal relationship but may merely represent linkage disequilibrium due to close proximity between the locus under test and the disease-causing locus. False association may also result from confounding bias (population stratification), a result of genetic heterogeneity within different ancestral populations. To avoid such bias, this method is dependent on careful case–control matching, and the results may need to be reproduced in several populations before a firm association can be established. Furthermore, multiple hypothesis testing may lead to positive associations simply through chance, a problem that may be exacerbated by publication bias in favour of positive results.

Candidate gene studies in stroke can be considered as belonging to two broad categories: (i) those investigating the role of genes which may influence stroke risk, and (ii) those investigating genes which determine infarct size after vessel occlusion by influencing vascular reactivity and collateral supply, and neuronal responses to injury. It should be remembered that these two categories are not mutually exclusive because certain genes may both predispose to stroke and affect stroke outcome. Furthermore, genes resulting in increased neuronal injury after an episode of ischaemia may themselves be associated with an increased incidence of stroke. However, whilst genes in the second category may be equally important in all forms of ischaemic stroke, those falling into the first category may be important only in certain subtypes of stroke. Furthermore, some candidate genes may be important only in young individuals. These considerations suggest that, to maximize the chance of detecting an association, analysis of data according to stroke subtype and/or separate studies of young individuals may be required. The role of a large number of candidate genes has been investigated in stroke, in many cases following the demonstration of an association with ischaemic heart disease. They can be conveniently be described in five groups, affecting (i) haemostasis, (ii) the renin–angiotensin system, (iii) nitric oxide production, (iv) homocysteine metabolism and (v) lipid metabolism.


Most larger case–control studies have failed to find an association between prothrombotic states, such as activated protein C resistance or the underlying Leiden factor V mutation, and ischaemic stroke in older individuals (Table 3). These gene defects may be responsible for stroke in some younger individuals, as discussed earlier under single-gene disorders, but on balance these prothrombotic states are unlikely to be important causes of multifactorial stroke in middle-aged and elderly patients.

View this table:
Table 3

Association studies in ischaemic stroke: haemostatic system

OCSP = Oxford Community Stroke Project Classification; TOAST = trial of org10172 in acute stroke treatment; PICH = primary intracerebral haemorrhage; APC= activated protein C resistance; NIDDM = non-insulin-dependent diabetes mellitus; IMT = intima media thickness; CHD = coronary heart disease; MTHFR = methylene tetrahydrofolate reductase; DVT = deep vein thrombosis; PE = pulmonary embolism; TIA = transient ischaemic attack.
Factor VCatto et al. (1995, 1996a)Q506 LeidenCase–control: 348 cases, 247 controlsIschaemic stroke/stroke mortalityNegative
Forsyth and Dolan (1955)Q506 LeidenCase–control: 45 patients, population controlsIschaemic stroke <45 yearsNegativeMutation found only in 1/45 cases
Kontula et al. (1995)Q506 LeidenCase–control: 236 cases, 137 controlsIschaemic strokeNegative
Lalouschek et al. (1995)Q506 LeidenCross-sectional: 58 patientsTIA or minor ischaemic strokeNegativeNo difference in clinical characteristics or risk factor profile between mutation negative or positive cases
Ridker et al. (1995)Q506 LeidenNested case–control: 209 cases, 209 controlsIschaemic stroke/PICHNegativeAssociation only with DVT and PE in the original cohort
Albucher et al. (1996)Q506 LeiidenCase–control: 30 cases, 75 controlsIschaemic strokePositive3/30 patients heterozygous versus 1/75 controls
Chimowitz et al. (1996)Q506 LeidenCase–control: 53 cases, 397 controlsIschaemic stroke 18–50 yearsPositiveOnly in cryptogenic infarction
Fisher et al. (1996)Q506 LeidenCase–control: 63 cases, 31 controlsIschaemic strokeNegativeIncreased APC resistance detected, not Q506Leiden
Martinelli et al. (1997)Q506 LeidenCase–control: 155 cases, 155 controlsIschaemic strokeNegative
Press et al. (1996)Q506 LeidenCase–control: 116 cases, 54/161/287 controlsIschaemic stroke in elderly/elderly controls with and without risk factors/young controlsNegative
van der Bom et al. (1996)Q506 LeidenCase–control: 112 cases, 222 controlsIschaemic strokeNegativeIncreased APC resistance independent of factor V
Sanchez et al. (1997)Q506 LeidenCase–control: 66 cases, 66 controlsIschaemic strokeNegative
Longstreth et al. (1998)Q506 LeidenCase–control 106 cases, 391 controlsIschaemic stroke young women aged 18–44 yearsNegative
Halbmayer et al. (1997)Q506 LeidenCase–control: 229 cases, 71 controlsIschaemic strokeNegative
Lalouschek et al. (1999)Q506 LeidenCase–control: 81 cases, 81 controlsTIA or minor ischaemic strokePositive trend
Markus et al. (1996)Q506 LeidenCase–control: 180 cases, 80 controlsIschaemic strokeNegative
De Lucia et al. (1997)Q506 LeidenCase–control: 14 cases, 75 controlsYoung ischaemic strokes from 3 familiesPositive6/14 were heterozygotes versus 1/75 controls
ProthrombinLongstreth et al. (1998)G20210ACase–control: 106 cases, 391 controlsIschaemic stroke young women aged 18–44 yearsNegative
De Stefano et al. (1998)G20210ACase–control: 72 cases, 198 controlsIschaemic stroke <50 yearsPositive
Halbmayer et al. (1998)G20210ACase–control: 20 cases, 20 controlsIschaemic stroke young individualsNegative
Poort et al. (1996)G20210ACase–control: 104 cases, 104 controlsIschaemic strokeNegativeAssociation with DVT only
Martinelli et al. (1997)G20210ACase–control: 155 cases, 155 controlsIschaemic strokeNegative
Ridker et al. (1999)G20210ANested case–control: 259 cases, 1744 controlsIschaemic stroke/PICHNegativeWeak association with DVT
Factor VIINishiuma et al. (1997)R353QCase–control: 137 cases, 83/97 controlsSymptomatic stroke or MRI silent lacuneae in hypertensives. Hypertensive or normotensive controlsNegative
Heywood et al. (1997)R353QCase–control: 286 cases, 198 controlsIschaemic stroke/OCSP subtype/stroke mortalityNegativeNo association with stroke subtype or mortality
Corral et al. (1998)R353Q/323A2Case–control: 104 cases, 104 controlsIschaemic strokeNegative
FibrinogenNishiuma et al. (1998)G455ACase–control: 85 cases, 85/84 controlsHypertensive strokes. Hypertensive/normotensive controlsPositive
Kessler et al., 1997G455ACase–control: 227 cases, 225 controlsIschaemic stroke/TOAST subtypePositiveHomozygotes only in large vessel stroke
Schmidt et al. (1998a)β148(C/T)Cross-sectional: 399 casesCarotid artherosclerosisPositiveHomozygotes (T/T) genotype
Carter et al. (1997)β448(1/2)Case–control: 305 cases, 197 controlsIschaemic stroke/OCSP subtypePositiveGenotype distribution different amongst females only
PAI 1Catto et al. (1997)4G/5GCase–control: 421 cases, 172 controlsIschaemic stroke/OCSP subtype and mortalityNegativeNo association with stroke subtype or mortality
Factor XIIICatto et al. (1998)Val34LeuCase–control: 529 cases, 437 controlsIschaemic stroke/OCSP subtypeNegativeWeak association with PICH. No association with ischaemic stroke subtype
GpIIb/IIIaCarter et al. (1998)P1A2Case–control: 505 cases, 402 controlsIschaemic stroke/OCSP subtype and mortalityPositveAssociation with atherothrombotic stroke in non smokers and patients <50 years
Wagner et al. (1998)P1A2Case–control: 63 cases, 122 controlsIschaemic stroke young women aged 15–44 yearsNegative
Ridker et al. (1997)P1A2Nested case–control: 209 cases, 209 controlsIschaemic stroke/PICHNegative
Carlsson et al. (1997)HPA1/HPA3Case–control: 218 cases, 165/321 controlsIschaemic strokeNegative
GpIb/IXGonzalez-Conejero et al. (1998)HPA2/VNTRCase–control: 104 cases, 104 controlsIschaemic strokePositiveVNTR C/B genotype, HPA2b
GpIa/IiaCarlsson et al. (1999)HPA5Case–control: 218 cases, 165/321 controlsIschaemic strokeNegative
Carlsson et al. (1997)C807TCase–control: 227 cases, 170 controlsIschaemic strokePositiveT807 allele associated with stroke in patients aged <50 years

Genetic variants in other components of the coagulation cascade, such as factor VII and fibrinogen, have also been examined after prospective studies which have demonstrated the role of these proteins in arterial disease (Meade et al., 1986; Heinrich et al., 1994; Smith et al., 1997). A factor VII gene polymorphism (R353Q) has been associated with higher levels of factor VII:C (Green et al., 1991), but in a later study no association was found between levels of factor VII:C or between the R353Q variant and ischaemic cerebrovascular disease (Heywood et al., 1997). This is consistent with the results of a separate large study of arterial thrombotic events (Corral et al., 1998). Another study examining factor VII polymorphisms in hypertensive small vessel disease was also negative (Nishiuma et al., 1997).

Homozygosity for a G→A substitution at position 455 of the β-fibrinogen gene, which is associated with higher fibrinogen levels, has been found to be increased in large vessel stroke (Kessler et al., 1997). This observation is consistent with reports that a variant in complete linkage disequilibrium (b148) may be a specific risk factor for carotid atherosclerosis (Schmidt et al., 1998). A separate group (Carter et al., 1997) examined the influence of a different fibrinogen β-chain variant (b448) and found an association in women but not in men. The basis of this sex-specific interaction remains unclear. Raised fibrinogen levels may predispose to stroke both by accelerated atherosclerosis and prothrombotic mechanisms.

Studies in myocardial infarction suggest that the formation of abnormal fibrin structures may be important in arterial thrombosis (Fatah et al., 1996). In normal physiology this is dependent on both the function of factor XIII, which is involved in fibrin crosslinking, and the activity of the fibrinolytic system. Raised levels of plasminogen activator inhibitor 1 have been demonstrated in acute ischaemic stroke and in the convalescent phase. However, no correlation could be demonstrated between ischaemic stroke and an insertion deletion polymorphism (4G/5G), which is itself associated with higher levels of plasminogen activator inhibitor 1 (Catto et al., 1997). A polymorphism in the factor XIII gene (Val 34 Leu) has also been examined in ischaemic stroke after reports that this polymorphism was protective in myocardial infarction. It has been suggested that this variant may be associated with weaker fibrin structures, but there was a lack of association with ischaemic stroke or ischaemic stroke subtype. Interestingly, a weak association with intracerebral haemorrhage was found in the same study (Catto et al., 1998).

The role of platelet glycoprotein receptor polymorphisms has also been studied extensively in patients with ischaemic stroke. These molecules are members of the integrin family and, when activated, bind fibrinogen, von Willebrand factor or collagen, and therefore promote platelet aggregation and thrombosis. The P1A2 variant of the platelet fibrinogen receptor Gp IIa/IIIb has been reported as a risk factor for acute coronary syndromes specifically in young patients (Carter et al., 1996; Weiss et al., 1996). Subgroup analysis in a case–control study has suggested that the P1A2 allele may also be an important risk factor in stroke patients aged less than 50 years (Carter et al., 1998). Another study, however, failed to find an overall association between this polymorphism and cerebral infarction in young women (Wagner et al., 1998). Conflicting genotype–phenotype correlations have also been found with the HPA2 (human platelet antigen 2) and VNTR (variable number of tandem repeats) variants of the platelet von Willebrand factor receptor, Gp Ia/IIa. It has been reported recently that a silent point mutation (GpIa C807T), correlating with increased expression of the collagen receptor in vitro, is an independent risk factor for stroke in young patients (Carlsson et al., 1999). However, association studies of different polymorphisms in this gene have revealed a lack of association (Carlsson et al., 1997).


The production of angiotensin II and the catabolism of bradykinin are important effects of angiotensin-converting enzyme (ACE), and these peptides have important functions at the local vascular level, including the regulation of vascular tone and endothelial function, and smooth muscle proliferation. The ACE gene is probably the most extensively investigated candidate gene in ischaemic stroke (Table 4), after an initial study by Cambien and co-workers which suggested that an intron 16 insertion/deletion polymorphism was associated with myocardial infarction (Cambien et al., 1992). A number of studies have reported an association with stroke, with a relative risk usually of the order of 1.5–2.5 (Table 4), but other studies have failed to find a significant association. A meta-analysis has evaluated the risk of stroke in 1918 subjects versus 722 controls from seven studies (Sharma, 1998). It was concluded that the ACE genotype conferred a small but modest effect, with an odds ratio of 1.31 (95% confidence interval 1.06–1.62), according to a dominant model of inheritance. A weaker association was seen under a recessive model. Unfortunately, methodological differences between studies precluded subtype analysis. A criticism of such meta-analyses is that, whilst the power to detect a significant disease causing allele is increased, the method is highly subject to publication bias. Negative candidate gene studies may not always be submitted or accepted for publication. In such meta-analyses, attempts need to be made to identify unpublished studies and estimate the magnitude of publication bias.

View this table:
Table 4

Association studies in ischaemic stroke: renin angiotensin pathway

I = insertion allele; D = deletion allele; ACE = angiotensin-converting enzyme.
ACESharma et al. (1994)I/DCase–control: 100 cases, 73 controlsIschaemic strokeNegativeTrend towards DD genotype associated with young stroke
Markus et al. (1995)I/DCase–control: 100 cases, 137 controlsTOAST subtypePositiveLacunar stroke (no association with carotid atheroma)
Catto et al. (1996b)I/DCase–control: 418 cases, 231 controlsIschaemic stroke, OCSP subtype and mortalityNegativeNo association with stroke subtype. D allele associated with early mortality
Pullicino et al. (1996)I/DCase–control: 60 cases, published controlsLacunar strokeNegative
Nakata et al. (1997)I/DCase–control: 55 cases, 61 controlsIschaemic strokePositiveDD genotype
Margaglione et al. (1996)I/DCase–control: 101 cases, 109 controlsIschaemic strokePositiveDD genotype
Ueda et al. (1995)I/DCase–control: 488 cases, 188 controlsIschaemic stroke and OCSP subtypeNegativeDD genotype in hypertensives versus hypertensive controls
Doi et al. (1997)I/DCase–control 181 cases, 271 controlsIschaemic stroke (atheroembolic/lacunar)PositiveYoung strokes (possibly post-stroke mortality also)
Elbaz et al. (1998)CT 2/3Case–control: 510 cases, 510 controlsIschaemic stroke and subtypePositive2/2 genotype associated with lacunar stroke
Kario et al. (1996)I/DCase–control: 228 cases, 90/104 controlsSymptomatic stroke or MRI silent lacunae in hypertensives. Hypertensive/normotensive controlsPositiveAssociation of D allele with clinical stroke or silent lacunae
Zee et al. (1999)I/DNested case–control: 348 cases, 348 controlsIschaemic stroke/PICHNegativeNo association following stratification for low risk
Sharma (1998)I/DMeta-analysis: 1918 cases, 722 controlsIschaemic strokePositiveModerate increase in risk associated with the D allele under a dominant model of inheritance
Watanabe et al. (1997)I/DCross-sectional: 169 casesCarotid atherosclerosis/asymptomatic lacunar strokePositiveCarotid atherosclerosis only
Aalto Setala et al. (1998)I/DCross-sectional: 234 casesCarotid atherosclerosis in patients with ischaemic stroke aged <60 yearsNegative
Castellano et al. (1995)I/DCross-sectional: 199 casesIMT patients aged 50–64 yearsPositive
Hosoi et al. (1996)I/DCross-sectional: 288 casesIMT NIDDM patientsPositive
AngiotensinogenBarley et al. (1995)M235TCase–control: 100 cases, 45 spouse controlsIschaemic stroke/IMT/carotid atheromaNegative
Nakata et al. (1997)M235TCase–control: 55 cases, 61 controlsIschaemic strokeNegativeTT-positive interaction with ACE DD genotype

The conflicting results of studies of the ACE gene insertion/deletion polymorphism and stroke reflect a number of methodological difficulties and illustrate some of the problems associated with candidate gene studies in human stroke. Many of the studies were small and underpowered; frequently the relative risk associated with the deletion allele was found to be of similar magnitude in smaller, statistically negative studies and larger, statistically positive studies. The avoidance of bias due to population stratification requires careful case–control matching, and a variety of control groups have been used in the different studies, varying from randomly selected population controls to non-randomly selected hospital patients with other diseases. Such problems can be reduced by the study of population cohorts that are followed prospectively. A nested case–control study was performed on the US Physicians Health cohort and a negative result was reported (Zee et al., 1999). Although reducing the risk of false-positive results due to selection bias, such prospective cohort studies tend to suffer from poor stroke phenotyping; in this study the cases included both ischaemic and haemorrhagic stroke, and no subtyping of ischaemic stroke subtypes was performed. Such analyses will fail to detect a selective association with a particular stroke phenotype, and this may be of particular importance with the ACE gene. A number of studies have reported an association that was strongest or exclusively with lacunar stroke (Markus et al., 1995; Elbaz et al., 1998), and these findings are consistent with reported associations between the deletion allele and MRI-detected silent small vessel disease in hypertensives (Kario et al., 1996). In a recent study, a weak association was found between the deletion polymorphism and all ischaemic stroke cases in Han Chinese in Taiwan. When a further study was performed with more detailed investigations allowing recruitment of only lacunar stroke patients, a much stronger positive association was found (Lin and Yueh, 1999). This study illustrates the need for careful assessment, by both brain imaging and the exclusion of carotid stenosis, in all patients in order to identify the small vessel phenotype. On current evidence, it is reasonable to conclude that the ACE deletion/insertion polymorphism is not a major risk factor in an unselected group of patients with ischaemic stroke, but that it may be a risk factor for small vessel disease; further studies are required in this stroke phenotype.

A variant of the angiotensinogen gene (M235T) has also been implicated in vascular disease, but its evaluation in stroke so far suggests that it does not behave as an important risk factor (Table 4). However, it has recently been proposed that an epistatic interaction with the ACE gene may exist (Nakata et al., 1997).

Nitric oxide

The activity of the l-arginine/nitric oxide synthase system is an important mediator of endothelial function. It has diverse effects, including the regulation of the tone, integrity, growth and thrombogenic properties of the vessel wall. Strong evidence from animal and human studies indicates that the activity of this system is under genetic control. Work in the stroke-prone SHR rat has suggested that impaired endothelial dysfunction is an important predisposing factor leading to stroke (Russo et al., 1998). In addition, knockout mice deficient in endothelial nitric oxide synthase are highly sensitive to focal cerebral ischaemia (Samdani et al., 1997) and have marked vessel wall abnormalities (Rudic and Sessa, 1999). An earlier study (Wang et al., 1996) had demonstrated that a functional variant of nitric oxide synthase (ecNOS 4a) was associated with increased risk of significant coronary artery disease and myocardial infarction in smokers. However, neither this nor another variant with unknown functional significance (Glu298Asp) has been shown to be an important risk factor for ischaemic cerebrovascular disease (Table 5).

View this table:
Table 5

Association studies in ischaemic stroke: endothelial nitric oxide, homocysteine and lipid metabolism

I = insertion allele; D = deletion allele; ACE = angiotensin-converting enzyme.
eNOSYahashi et al. (1998)ecNOS 4a/bCase–control: 127 cases, 91 controlsAtherothrombotic/lacunar/silent strokeNegative
Macleod et al. (1999)Glu298AspCase–control: 265 cases, 293 controlsIschaemic strokeNegative
Markus et al. (1998)Glu298AspCase–control: 361 cases, 236 controlsIschaemic stroke/TOAST subtypeNegativeNo association with carotid atheroma
MTHFRMarkus et al. (1997)C677TCase–control: 345 cases, 161 controlsIschaemic stroke/TOAST subtypeNegativeNo association in folate deficient patients, young individuals or carotid atheroma
Lalouschek et al. (1999)C677TCase–control: 81 cases, 81 controlsTIA or minor ischaemic strokeNegativePossible synergism between MTHFR and Q506 alleles
Nakata et al. (1998)C677TCase–control: 48 cases, 105 controlsIschaemic strokeNegative
De Stefano et al. (1998)C677TCase–control: 72 cases, 198 controlsIschaemic stroke <50 yearsNegative
Reuner et al. (1998)C677TCase–control: 91cases, 182 controlsIschaemic stroke Negative
Harmon et al. (1998)C677TCase–control: 174 cases, 183 controlsIschaemic stroke CT proven >60 yearsInconclusiveOR 1.59 (CI = 0.85–2.97)
apo EKessler et al. (1997)apoε2/ε3/ε4Case–control: 227 cases, 225 controlsIschaemic stroke/ TOAST subtypePositiveε4 association with large vessel disease
Ferruci et al. (1997)apoε2/ε3/ε4Cohort study: 1664 subjectsIschaemic stroke >71 yearsPositiveε2 protective in patients 70–79 years
McCarron et al. (1998)apoε/ε3/ε4Cohort study: 714 subjectsIschaemic stroke survivalPositiveε4 associated with a favourable
Basun et al. (1996)apoε2/ε3/ε4Cohort study: 1077 subjectsIschaemic stroke in those >75 years at baselineNegative.No association with apo E genotype. Reduced ε3/ε4 frequency in patients with previous stroke at baseline
Kuusisto et al. (1995)apoε2/ε3/ε4Cohort study: 1067 subjectsIschaemic and haemorrhagic stroke 65–74 years at baselineNegative
Margaglione et al. (1998)apoε2/ε3/ε4Case–control 100 cases, 108 controlsIschaemic strokePositiveε4 allele a risk factor, ε3/ε3 homozygotes protected
Couderc et al. (1993)apoε2/ε3/ε4Case–control: 69 cases, 68 controlsIschaemic stroke or TIAPositiveε3/ε3 protective, ε3/ε2 risk factor
Pedrobotet et al. (1992)apoε2/ε3/ε4Case–control: 100 cases, 100 controlsIschaemic strokePositiveε4 a risk factor
Nakata et al. (1997)apoε3/ε3/ε4Case–control: 55 cases, 61 controlsIschaemic strokeNegative
Schmidt et al. (1997)apoε2/ε3/ε4Cross-sectional: 280 casesSilent white matter diseasePositiveε2/ε3 genotype risk factor
De Andrade et al. (1995)apoε2/ε3/ε4Case–control: 145 cases, 224 controlsCarotid atherosclerosis age 45–64 yearsPositiveε2/ε3 genotype risk factor
Terry et al. (1996)apoε2/ε3/ε4Cross-sectional: 260 casesIMT in patients with and without CHDPositiveε2 protective
Aolto-Setala et al. (1998)apoε3/ε3/ε4Cross-sectional: 234 casesCarotid atherosclerosis in patients with ischaemic stroke <60 yearsNegative
apo A1/CIIIAolto-Setala et al. (1998)SstICross-sectional: 234 casesCarotid atherosclerosis in patients with ischaemic stroke <60 yearsNegative
Patsch et al. (1994)XmnICross-sectional: 268 casesIMT in groups with different lipid profilesPositive6.6 kb allele a risk factor in group with elevated cholesterol and triglyceride
apo BAolto-Setala et al. (1998)XbaICross-sectional: 234 casesCarotid atherosclerosis in patients with ischaemic stroke <60 yearsNegative
Lipoprotein lipaseHuang et al. (1997)A291GCase–control: 125/56 cases, 95 controlsIschaemic stroke/carotid atherosclerosisNegative
Paraoxonase 1Cao et al. (1998)glu192argCross-sectional: 197 casesIMT in NIDDMNegative
Schmidt et al. (1998b)glu192arg met54leuCross-sectional: 316 casesCarotid atherosclerosis 44–75 yearsPositiveleu54leu genotype

The genes encoding both the neuronal and the inducible form of nitric oxide synthase are potential candidate genes for stroke. In animal models, their inhibition reduces infarct size, which is also smaller in knockout mice. Both genes have been cloned and common polymorphisms described. The results of association studies in human stroke are awaited.

Homocysteine metabolism

The knowledge that inborn errors of homocysteine metabolism can lead to severe homocysteinaemia and arteriosclerosis has generated significant interest in mild homocysteinaemia as a risk factor for vascular disease. Subsequent cross-sectional, case–control and prospective studies have established that mild to moderate elevations of serum homocysteine (>15 mmol/l), at levels below those associated with homocysteinuria, are also associated with increased risk (Brattstrom et al., 1998). The aetiology of hyperhomocysteinaemia within this range is likely to be multifactorial. Environmental factors such as folate intake are believed to be important, in combination with genetic factors. The prevalence of heterozygosity for the cystathione β-synthase gene is estimated at between 0.5–0.15% of the population (Mudd and Levy, 1983), heterozygous individuals possessing ~30% of the normal enzyme activity. This has led to several studies to determine whether allele heterozygosity is itself a significant risk factor for polygenic ischaemic stroke. Two studies found an increased frequency of heterozygotes in patients with occlusive cerebrovascular disease compared with controls (Boers et al., 1985; Clarke et al., 1991), and this has also been shown in patients with asymptomatic carotid artery atherosclerosis (Rubba et al., 1990). In contrast, no association was found between heterozygosity for the cystathione β-synthase gene and either carotid intima media thickness (de Valk et al., 1996) or asymptomatic carotid atherosclerosis (Clarke et al., 1992). These inconsistencies may reflect the different ages of the patients examined, mechanisms other than wall disease through which homocysteine acts, or the influence of multiple risk factors on homocysteine levels and carotid artery damage in carriers.

Very rarely, patients with homocysteinuria have complete deficiency of methylene tetrahydrofolate reductase, a folate-dependent enzyme catalysing the rate-limiting step in the methylation of homocysteine to methionine. In 1988, a common thermolabile variant of methylene tetrahydrofolate reductase associated with decreased enzyme activity and mildly elevated plasma homocysteine levels was identified (Kang et al., 1988). The underlying genetic basis of this is a common mutation (a C→T substitution at position 677), and the role of this polymorphism has been studied in the context of ischaemic stroke. Several case studies, with one possible exception, indicate that variation at this locus does not correlate with stroke risk (Table 5). These conclusions are consistent with a recent meta-analysis which found that the C677T variant was associated with mild homocysteinaemia but not increased vascular risk (Brattstrom et al., 1998). There is an interaction with folate, and it is still possible that the methylene tetrahydrofolate reductase polymorphism may be a risk factor in younger individuals with low folate intake, but further studies are required in these populations.

Lipid metabolism

Individuals with higher levels of plasma cholesterol, increased HDL (high-density lipoprotein) and decreased LDL (low-density lipoprotein) have a higher risk of premature atherosclerosis. The phenotype may arise not only from single gene disorders, as discussed above, but also from a number of genetic and environmental factors, including polymorphic variants of genes encoding the apolipoproteins, lipoprotein receptors and the key enzymes of plasma lipoprotein metabolism. Apolipoprotein E is a glycoprotein that mediates the binding of lipid particles to specific lipoprotein receptors, and three major isoforms arising from different amino acid substitutions and encoded by the different alleles, ε2, ε3 and ε4, have been identified. The ε4 variant has been associated with higher total serum cholesterol and LDL cholesterol levels, and has been postulated as an important risk factor in ischaemic stroke. However, studies to date have produced conflicting results as to the importance of apolipoprotein E alleles in predisposition to ischaemic stroke (Table 5). In small case–control or cross-sectional studies, both the ε2/ε3 genotype (Couderc et al., 1993; de Andrade et al., 1995; Schmidt et al., 1997) and the ε4 allele (Pedro-Botet et al., 1992; Kessler et al., 1997; Margaglione et al., 1998) have been over-represented in patients with ischaemic stroke. Other groups have examined the role of the apolipoprotein E genotype in modulating the outcome of cerebral infarction as this lipoprotein appears to be an important regulator of lipid turnover within the brain and of neuronal membrane maintenance and repair. Studies in patients with head injury and intracerebral haemorrhage have indicated that the ε4 allele is a predictor of poor outcome in terms of death and disability (Alberts et al., 1995; Teasdale et al., 1997), and this is consistent with studies of cognitive decline in ε4 carriers with cerebrovascular disease (Kalmijn et al., 1996). Surprisingly a study by McCarron and colleagues found a favourable effect of ε4 on stroke outcome (McCarron et al., 1998), but this may reflect the very broad-based measure of stroke outcome used in their analysis. The role of other lipoproteins and enzymes in relation to stroke and carotid artery disease is considered in Table 5.

Future directions

Conventional candidate gene association studies remain a powerful means of investigating the genetics of ischaemic stroke. The success of the approach will be increased by focusing on specific pathogenic stroke subtypes, perhaps by studying younger populations of stroke individuals, and by careful attention to case–control matching. The study of prospectively followed population cohorts will be important, but improved stroke subtyping of cases in these studies is required if they are to be sensitive to associations with particular stroke phenotypes. In practice this is difficult, because stroke cases will present without warning to a number of hospitals. In this respect, case–control studies allow more rigorous stroke phenotyping. The problems associated with inadequate case–control matching can be overcome by the use of internal family control-based association methods and the use of statistical analyses based on the transmission disequilibrium test, as discussed below under the heading Linkage analysis. Family trios, usually comprising one or both parents and one or more sibs, are recruited, and preferential transmission of disease-causing alleles to affected, as opposed to unaffected, children is sought. Such techniques have been applied to the study of other complex diseases, such as diabetes (Rogus et al., 1998) and hypertension (Niu et al., 1999), and these methods would be suitable for application to stroke, although no studies using these techniques in stroke patients have been published to date. However, the candidate gene approach is inherently limited by its reliance on the existence of known potential candidate genes, and it does not allow the identification of novel genes. Of the 75 000–100 000 potential genes, the functions of only perhaps 5–10% are known, which places a theoretical limit on these strategies for gene detection in ischaemic stroke. A number of alternative approaches are currently available, or will be available shortly.

Linkage analysis

Model-free methods of linkage analysis, based on the comparison of allele-sharing between affected relatives with that predicted by random segregation alone, are being used increasingly in the study of polygenic disorders (Lander and Schork, 1994). The simplest form of this method is affected sib-pair analysis. Under random segregation, two sibs would be expected to be identical for any locus with a 25, 50 and 25% distribution for 0, 1 and 2 identical alleles. The actual distribution of a marker might be significantly different from chance if it was linked to a locus associated with disease. Using a dense collection of polymorphic markers plus computational methods (genome-wide screening), one can infer that a chromosomal region contains the gene responsible for the disorder. Variants of this method can be applied to larger pedigrees containing large numbers of affected and unaffected members. These approaches have been used successfully in diseases such as type I diabetes (Owerbach and Gabbay, 1994). One advantage of this approach is that a hypothesis need not be developed concerning the identity of a gene, which is inferred according to its physical position on the chromosomal map. The main problem in applying this method in multifactorial ischaemic stroke is that, since it is predominantly a disease of the middle-aged and elderly, the identification of large numbers of affected siblings or the identification of large pedigrees is extremely difficult. As an example, in our database of 750 consecutive patients recruited with ischaemic stroke whose mean age was 68 years, we identified a positive history of stroke in a first-degree relative in only 80 cases. Thirty of these were alive at the time of the stroke in the index case. Therefore, studies will require cooperative networks of investigators to obtain sufficient numbers of affected sib pairs.

As an alternative, one might compare an intermediate phenotype, such as carotid intimal medial thickness, between siblings. Linkage models have been developed which specify that similarity between relatives correlates with the number of alleles shared at a trait-causing locus. Genome-wide screening methods can be used to map these quantitative trait loci, an approach that has been applied successfully in animal studies (Ikeda et al., 1996; Rubattu et al., 1996; Jeffs et al., 1997).

More recently, interest has focused on a combined linkage/association-based approach, the main advantage of which is that population stratification is no longer an issue. These methods use internal family controls instead of population-based controls to determine the association of candidate genes with disease. In the conventional transmission disequilibrium test, both parents and an affected sibling are available for marker genotyping (Spielman et al., 1993). However, in many instances parent controls may not be available, and an alternative method is the comparison of affected and unaffected siblings, the sibling transmission disequilibrium test (Spielman and Ewens, 1998). An attractive feature of using parent controls is that it avoids the problem of overmatching, which can occur where unaffected siblings are used who may be non-penetrant carriers of the susceptibility allele. Furthermore, the use of parent controls rather than unaffected sibling controls has theoretical advantages in that it allows the differential effects of paternal versus maternal transmission to be distinguished. In reality, however, it is likely that data from both types of family structure will be available, and these data can be combined in an overall test. Sample power calculations for the transmission disequilibrium test, using a variety of population allele frequencies and odds ratios associated with the disease allele, are shown in Table 6, and it can be seen that the studies still require large populations and are likely to be multicentre collaborative ventures.

View this table:
Table 6

Estimated number of families which would be required for transmission disequilibrium testing

Baseline allele frequencyOdds ratioSample size (trios)

Genome-wide association studies

To date, linkage analysis of complex polygenic disorders has resulted in only modest success. Recent technological developments enabling the identification and scoring of individual sequence variation have led to an increase in the number of polymorphisms available, which has rekindled interest in the use of association studies in the study of polygenic disease. It has been proposed that association studies could be extended to include a systematic search through the entire human genome for single-nucleotide polymorphisms in linkage disequilibrium with a disease-causing allele. The prospect of genome-wide association studies has become a real possibility after recent analytical considerations (Risch and Merikangas, 1996). The power expected from a whole-genome linkage study versus a genome-wide association study indicates that association-based methods are particularly efficient in detecting variants that have a small effect on the risk of disease. These calculations translate into a dense map of ~100 000 single-nucleotide polymorphisms being sufficient to put this strategy into place. Advances in identifying and cataloguing sequence variation, such as the identification of large numbers of single-nucleotide polymorphisms from expressed sequence tags (ESTs) using gel-based and novel technologies has made this goal a realistic proposal. Once technological constraints permit, genome-wide association could readily be applied to existing DNA stroke databases.

Human genome mapping project

In 1990 the human genome project was initiated, with the ultimate goal of mapping the entire sequence of the human genome, which comprises approximately 3 billion base pairs. Since most of the human DNA sequence appears to be non-coding, the defining and sequencing of regions of the genome encoding proteins was proposed as an intermediate goal by sequencing clones from cDNA libraries (ESTs) (Adams et al., 1991). As a rough estimate, it is believed that the list of ESTs contained within the dbEST database (www.ncbi.nlm.nih.gov/dbEST/index.html) constitutes ~90% of expressed genes and defines a considerable proportion of the human genetic diversity contributing to both normal physiological variation and pathophysiological states. Parallel projects have been initiated to define protein products, to catalogue sequence variation (http://www.ncbi.nlm.nih.gov/SNP) and to map the physical positions of ESTs (www.ncbi.nlm.nih.gov/genemap98). It is expected that all of these endeavours will facilitate the identification of genes involved in the pathogenesis of complex polygenic disorders.

Expression profiling

To realize the potential of EST databases fully, it would be of great interest to assess rapidly the patterns of expression of these genes both at the RNA and at the protein level. For example, if a linkage study identified a chromosomal region linked to ischaemic stroke susceptibility, one would next want to know whether there were any candidate genes within this region. As more ESTs are physically mapped, several candidate genes might lie within a defined interval, and knowledge of the patterns of expression of ESTs will be helpful in determining the course of future investigation. Approaches include differential display, random sequencing of subtracted and normalized cDNA libraries, and the use of microassay technologies. Some of these methods offer the possibility of novel transcript discovery, or provide insights into novel functions of known genes in the setting of stroke. The potential of these approaches was demonstrated recently in a study in which a novel clone was found to be upregulated in an animal model of focal cerebral occlusion, and a hitherto unsuspected function of tissue inhibitor of metalloproteinase 1 was discovered (Wang et al., 1998).

Physiological genomics

Candidate genes tested in association studies are usually chosen because of their physiological relevance to the pathophysiology of stroke. Increasingly, however, the functional significance of identified genes on the phenotype will be determined in in vitro or in vivo systems, for example in classical expression studies in cell culture and the use of transgenic technology and gene transfer methods. By selectively deleting nitric oxide synthase genes in knockout mouse models, it has been possible to determine the role of the different isoforms in the setting of focal ischaemia. For example, nitric oxide produced by endothelial nitric oxide synthase may be protective, but that produced by neuronal nitric oxide synthase may be neurotoxic (Samdani et al., 1997). Transgenic mice have been generated which carry different forms of the human apolipoprotein E allele, and these animals develop markedly different infarct volumes after middle cerebral artery occlusion (Sheng et al., 1998). In vivo gene transfer methods have also been used to determine the effects of upregulating gene expression on phenotype. For example, a recent study revealed that ANP gene delivery appeared to reduce stroke mortality in Dahl salt-sensitive rats (Lin et al., 1999).


Ischaemic stroke is a major cause of death and disability throughout the world, but attempts to define the genetic basis of this condition have lagged behind studies in other polygenic disorders. Strong evidence from epidemiological and animal studies has implicated genetic influences, but the identification of individual causative mutations in human polygenic stroke remains problematic, being limited in part by the number of approaches currently available. The recent biological revolution spurred by the human genome project promises the advent of novel technologies supported by bioinformatics resources that will transform the study of disorders such as stroke. The ultimate goal of these endeavours will be not only to provide new avenues for prevention but also to provide insights into factors that influence the outcome of stroke, and new therapeutic targets when preventative strategies have failed.


We wish to thank Dr Pak Sham and Dr John Powell for helpful discussions. Dr Hassan received support from Dr John Bamford and the St James's Trust for Nervous System Diseases.


View Abstract