Ischaemic stroke is a complex trait, with genetic influences likely to work through various risk factors for cerebrovascular disease. Hypertension and dyslipidemia are known to influence atherosclerosis and thus predispose to ischaemic stroke. Although lipids are a weaker risk factor for stroke than blood pressure or age, hyperlipidemia contributes to stroke risk 1, while lipid lowering therapy significantly reduces the risk of stroke 2. Low HDL has also been reported as a risk factor for ischaemic stroke 3. Genes involved in lipid metabolism have been previously studied in the stroke population. Some studies suggest APOE*ε4 and APOE*ε2 alleles are associated with higher risk and lower risk of ischaemic stroke respectively, although this has not been confirmed in all studies 45.

Paraoxonase, a 45 kDa glycoprotein, is well known for its ability to detoxify organophosphates 6. The process of atherosclerosis is thought to begin with inflammatory changes at the endothelial level, and oxidated LDL is believed to be pro-inflammatory. Oxidated lipids have also been associated with progression of carotid atherosclerosis 78. As an HDL associated protein, paraoxonase protects LDL against oxidation 9, and thus may protect against atherosclerosis 10.

The paraoxonase gene maps to chromosome 7 (q21-23) and has three isoforms, PON1, PON2 and PON3 11. Some studies suggest PON polymorphisms might predispose to CHD or ischaemic stroke 121314, while other studies are less conclusive 15. PON1 has two common polymorphisms at codon 192 [A/G: Gln (Q)/Arg (R)] and 55 [T/A: Leu(L)/Met (M)]. These have been shown to correlate with enzyme activity 1617. Increased enzyme activity has been associated with higher HDL levels in some populations 1819, but not in all studies 20. The PON 192 and 55 polymorphisms have been associated with risk of ischaemic stroke in a small study of younger adults 13 and older patients respectively 21. There are two common polymorphisms of the PON2 gene, PON2-148 [C/G: Ala (A)/Gly (G)] and PON2-311 [C/G: Cys (C)/Ser (S)], and the PON2-148 "G" (G variant) is associated with higher fasting blood glucose in NIDDM patients 22. The PON2-311 "G" (S allele) has been associated with coronary artery diseases in different populations 23. PON3 is thought to have a similar role to PON124.

The purpose of this study was to assess the distribution of polymorphisms of the PON1, PON2 and PON3 genes in well-phenotyped patients with ischaemic stroke and matched control subjects, and to expand on previous studies by examining haplotype distribution in these samples. We also examined the association between genotype and full lipid profile for a subgroup of controls.

Demographic parameters are shown in Table 2. Allele frequency for the six PON alleles investigated is shown in Table 3.

There was no significant difference between cases and controls by sex in genotype and allele frequencies. There was no significant difference in genotype and allele frequencies between cases and controls in terms of age stratification (≥65 and <65 years).

There was no significant difference in genotype or allele frequencies between cases and controls in the SVD group or LVD group, other than a small association between the GG genotype of PON3 and females with LVD (p = 0.03, OR = 2). PON2-311 "CC" genotype (SS variant) was higher in male cardioembolic (CE) stroke patients than in male controls (p = 0.03, OR = 2.05).

There was no association between lipid profile (LDL, HDL, total cholesterol/HDL ratio or triglycerides) and PON genotype in the stroke population. We also studied the association between PON genotype and full lipid profile in 181 of the controls. No differences were found in VLDL-Cholesterol, IDL, LDL, HDL2, HDL3, and total cholesterol concentrations in females or males between different genotypes of PON1-192, PON1-55, PON2-148 or PON3-1. VLDL-Triglyceride was higher in females with "GG" PON3-3 genotype (VLDL: AA genotype 69.8, "AG" genotype 77.8, "GG" genotype 116.0; p = 0.036).

Haplotype analysis was performed using PHASE version 2.02 and EHPLUS programs. The PHASE program 27, resulted in reconstruction of 16 different haplotypes in cases and controls. Overall haplotype frequencies in cases and controls were not significantly different (p = 0.19). However, some individual haplotypes were more frequent in one group than the other. Overall, the 112211 haplotype (1= frequent allele, 2= rare allele) was present at a higher frequency among cases than controls (16.62% vs 12.35%, p-value = 0.015, OR = 1.42 (1.07–1.87)). In contrast, 111121 was less frequent in cases than in controls (2.27% vs 4.81%, p = 0.006, OR = 0.46 (0.26–0.81)). There were also some differences in haplotype combination (genotype distribution) between cases and controls as follows: the "112211, 112221" haplotype (cases 10.6%, controls 5.2%: p = 0.005), the "111111, 112221" haplotype (cases 4%, controls 7.9%: p = 0.023).

We repeated analyses using EHPLUS 28 to compare haplotype arrangements in cases and controls. With this method, there was a significant difference (p = 0.005, df, 63) between groups. The most frequent haplotype in cases was 112211 and in controls 111111, while with PHASE2 the commonest haplotype for both groups was 112221. The EHPLUS program generated 64 different haplotypes both in cases and controls but only 32 had a frequency of more than 1%.

PON1 and PON2 are possible candidate genes for atherosclerotic diseases, especially coronary heart disease (CHD), but findings are inconsistent. In a recent meta-analysis of 43 studies 29, the 5 largest studies estimated the per-allele risk ratio at 1.05 for PON1-192 R (B) allele and combined analyses of studies of the PON1-55 M and PON2-311 C variants showed no significant overall associations with CHD. This differs from a recent study 30, not included in the meta-analysis, which suggested that the picture remains complicated. While there have been fewer studies in ischaemic stroke, there is some evidence of a role for the PON1-192 RR (BB) genotype and also for the PON1 (C-107T) allele in younger ischaemic stroke patients 131431. A recently published study of 81 stroke patients and 2053 controls from the CARE study also suggested a role for PON1-192 in predisposing to stroke, with no association with polymorphisms of the other PON genes. In contrast to this study, haplotype analysis or stroke subtyping was not performed 32. There is evidence that paraoxonase (PON1) phenotype may predict vascular disease more accurately than genotype alone 33. Due to logistical constraints in the current study it was not possible to measure PON1 phenotype, but future vascular studies should consider this.

As there are ethical difficulties in gaining consent from incapacitated individuals, this is not a study based on incident stroke. It is acknowledged therefore that patients with more severe stroke may not be included. However, the TOAST classification has similar distribution to other studies. The distribution of other vascular risk factors shows that these are high risk individuals. They form a subgroup of a Scottish stroke population which has been demonstrated to have an association with 5-lipoxygenase activating protein, an important mediator of the activity of cellular 5-Lipoxygenase (5-LO) which is a key enzyme in the biosynthesis of leukotrienes, and may thus be involved in atherosclerosis by proinflammatory mechanisms 34. While a large number of individuals were unclassifiable by TOAST criteria, they all had CT evidence of ischaemic stroke. Individuals with cardioembolic stroke were also included. While these individuals may have a different underlying mechanism for their event, the role of oxidative stress through lipids driving atherothrombosis is likely to underlie the cardiac disease in most of these individuals.

Other than weak associations in small subgroups (females with LVD stroke, males with CE stroke), our study found no significance in overall allele or genotype frequency, although haplotype analysis suggests that the combination of certain alleles may increase or decrease the risk of ischaemic stroke in the whole population or in some specific subtypes. These results should be interpreted with caution due to the small numbers of individuals with specific haplotypes, but reinforces the need for very large scale studies to provide a definitive answer.

Although there are several programs based on different methods that can infer haplotypes, there is no consensus as to which method is the best 35. Most programs, including EHPLUS, a modified version of EH developed by Ott, use the EM (Expectation-Maximisation) algorithm 2835. The program ignores all partial or missing genotype data so that the algorithm uses the full set of available data.

The PHASE program is based on coalescent models and uses Markov chain Monte Carlo methods 2735. These algorithms use a population genetics model to estimate different haplotype patterns, so a haplotype that is more similar to the commonly observed haplotype patterns is more likely to be inferred.

While the haplotype analysis is an important component of association studies in complex diseases 36, and the different computational methods have their individual strengths and weaknesses, significant haplotype differences in the absence of genotypic or allelic differences are difficult to interpret 35. Our study highlights this further since one method (EHPLUS) suggests a statistically significant difference while PHSE fails to demonstrate an overall significant difference. It remains unclear which method is superior, but PHASE is believed to be more reliable for larger numbers of markers 35. However, the level of significance is modest and could have arisen by chance.

In some genetic isolates certain PON genotypes can be associated with variation in plasma lipoproteins levels including LDL, HDL and total cholesterol 373839. In addition the physical relationship of PON1 with HDL and the existence of cholesterol pathway regulatory elements at the PON1 locus suggest a further relationship of PON1 with lipoproteins, which may contribute to its role of vascular diseases 40. However, apart from a minor correlation between VLDL-Triglyceride and PON3 genotypes in female, we did not replicate these differences in our out-bred Scottish population.

In conclusion, we found no significant genotypic or allelic frequency differences between stroke cases and controls for any of the structural polymorphisms of the PON genes tested. We identified specific haplotypes which may influence stroke risk in our population. Further work is required in large cohorts to determine whether these differences are real or have arisen by chance.

The author(s) declare that they have no competing interests.

AP participated in study design, carried out the genotyping and the statistical analyses and drafted the manuscript. HRA and JC helped with sample preparation LW and DS contributed to the study design, sample collection and helped draft the manuscript. AC helped with study design and commented on the manuscript. MJM contributed to study design, coordinated recruitment and drafted the manuscript. Authors read and approved the final manuscript.