To the best of our knowledge, this is the first comprehensive association study of mtDNA variation with ischemic stroke risk in an European population. In a large population sample of ethnically-matched cases and controls, we found that haplogroup H1 is protective while haplogroups pre-HV/HV and U increase risk for ischemic stroke. Since these haplogroups are defined by the combination of several polymorphisms also present in other clades (e.g. allele A of m.3010G>A is a phylogenetic marker of subclades H1 and J1b), the observed haplogroup associations cannot be attributed to particular SNPs, but instead to their precise arrangement. To exclude the possibility that the observed associations are due to population stratification with study participants of African or non-West Eurasian ancestries, we performed the statistical analyses in the overall dataset excluding the individuals with haplogroups U6a, L, and "Others" (54 controls and 64 patients), and obtained the same associations (data not shown). Unlike H1, the pre-HV/HV, U and U5 haplogroups were found in a small number of individuals, and therefore their association with stroke risk is only suggestive. Low counts tend to inflate the qui-square values and lead to false-positive results. We did not study the association of mtDNA with stroke subtypes since a much larger sample size would be required to have a representative number of individuals in each subtype and haplogroup category. Stratification by sex was performed as there are clear differences between male and female ischemic patients [33
], and some of the associations (e.g. adjusted association of H1 in females) most likely did not reach statistical significance due to the relatively small sample sizes.
Earlier studies have addressed the contribution of mtDNA variation to stroke susceptibility. The m.12308A>G polymorphism defining haplogroups U and K, previously associated with occipital stroke in migraine [15
] and suggested to increase the risk of developing stroke in MELAS patients with the m.3243A>G mutation [34
], was not associated with ischemic stroke in our dataset. However, an association of the U5 subcluster with migrainous stroke has been reported [16
] and is consistent with our tentative association of the U5 haplogroup and its defining m.13617T>C polymorphism with ischemic stroke. m.5178C>A, associated with aging [4
] and cerebrovascular disorders (cerebral hemorrhage or infarction) in a small Japanese case-control sample [35
] and with intima-media thickness in carotid arteries of Japanese type 2 diabetic individuals [17
], could not be investigated here as it is Asian-specific [36
]. Haplogroup A, unlike its defining polymorphisms m.663A>G in the 12S rRNA gene and m.8794C>T in the ATPase 6 gene, was recently found associated with atherothrombotic cerebral infarction in 440 Japanese females after adjustement for significant co-variates [37
]. None of the three SNPs we studied in the 12S rRNA and ATPase 6 genes (m.709G>A, m.8701A>G, and m.9055G>A) were associated with ischemic stroke, suggesting that haplogroup A, but not its defining SNPs individually or other SNPs in the same genes, may constitute a risk factor for stroke in Japanese. Finally, we did not try to replicate the reported association with lacunar cerebral infarction of the m.16189T>C variant in the mtDNA hypervariable region [38
] as we only investigated SNPs in the coding region and this polymorphism is not restricted to any particular haplogroup [39
]. These discrepancies among reports highlight: i) the difficulty of finding reproducible mitochondrial genome associations with disease due to the continent-specificity of some mtDNA SNPs and clades, and ii) the necessity of performing association studies in very large samples so that even uncommon haplogroups are represented by a sufficient number of individuals. A power analysis of mitochondrial haplogroup association studies such as the present one (investigating 17 haplotypes) reveals that a sample of size similar to ours (515 cases and 515 controls) only provides 50% power to detect a change in haplogroup frequency from 0.251 in controls to 0.17 in cases (as observed here for H1) at a significance level of 0.05 [41
]. Even though we only had 50% power, we detected an association of H1 at a significance level of 0.001, and this association would survive a Bonferroni correction for the seventeen crude or adjusted association tests performed for haplogroups, suggesting that it is an important association. Much larger cohorts are required for less common clades or finer changes in haplogroup frequency, and therefore the present study provides preliminary evidence of association that requires further validation in independent cohorts.
Although the polymorphisms that characterize the phylogeny are thought to be evolutionarily neutral, they may cause subtle alterations in the encoded transcripts or proteins, which collectively and over time, influence the risk of a stroke event. Given that stroke is mostly a late-onset disorder, it does not affect the successful transmission of mtDNA alleles and their fixation in the population. Additionally, several reports have documented the tissue-specific accumulation of mitochondrial deletions with aging [42
], and it is conceivable that mtDNA polymorphisms or haplogroups which are neutral under normal circumstances become advantageous in post-mitotic tissues in the presence of acquired mutations.
The associated m.3010G>A non-coding polymorphism, located in the conserved 3' end of the 16S rRNA gene, lies near non-coding point mutations known to confer resistance to chloramphenicol, a prokaryotic and mitochondrial protein synthesis inhibitor [44
]. The synonymous m.7028C>T transition is located in the cytochrome c oxidase (COX) subunit I gene (COI) of complex IV. This protein complex is the terminal enzyme of the respiratory chain, which collects electrons from reduced cytochrome c and catalyzes the reduction of oxygen to water, and consists of 13 polypeptide subunits, 3 of which are mtDNA-encoded. m.11719G>A is a synonymous SNP in the ND4 gene. ND4 gene product is a subunit of the respiratory complex I which accepts electrons from NADH, transfers them to ubiquinone and uses the energy released to pump protons across the mitochondrial inner membrane. A mutation in ND4 (m.11778G>A) causing an arginine to histidine change at amino acid 340 [MIM 516003.0001] accounts for over 50% and 90% of all LHON cases among Caucasians and Asians, respectively. Interestingly, the penetrance of this mutation is higher within a J haplogroup background, but its effect is most prominent on the J2 subclade [8
]. The physical proximity of the associated polymorphism in ND4 to known mutations suggests that it lies in or close to important functional domains and has the potential to alter the protein's function.
It is interesting to notice that the majority of polymorphisms associated with stroke risk in the present report are localized in complexes I and IV, whose deficiencies are the most frequently observed abnormalities of the OXPHOS system. It would be of great interest to assess if stroke patients display complex I and IV deficits relative to matched controls, prior to their first stroke, and to identify phenotypic differences among haplogroups using transmitochondrial hybrid cell (cybrid) technology [45
]. In rats, a reduction in the aerobic capacity is concomitant with a decrease in the amount of proteins required for mitochondrial biogenesis and oxidative function in skeletal muscle, and with an increase in cardiovascular risk factors [46
The ethiopathogenic complexity of stroke is paralleled by that of mitochondrial disorders, probably in part due to their dual genetic control (mitochondrial and nuclear) and interplay with the environment. A small minority of complex I to IV subunits are mtDNA-encoded and produced, while the majority of subunits are nuclear-encoded and transported into the organelle. It is likely that mtDNA polymorphisms and haplogroups act synergistically with nuclear genetic factors and environmental components, and therefore mtDNA-encoded gene/nuclear-encoded gene and mtDNA-encoded gene/environment epistatic interactions may explain a larger fraction of the ischemic stroke heritability.