All study protocols were reviewed and approved by the Institutional Review Board of CHA Bundang Medical Center and followed the recommendations of the Declaration of Helsinki. Study subjects were recruited from the South Korean provinces of Seoul and Gyeonggi-do between 2000 and 2008. The Institutional Review Board of CHA Bundang Medical Center approved this genetic study in June 2000 (IRB No. 2013-09-073) and informed consent was obtained from study participants.

The Department of Neurology at CHA Bundang Medical Center, CHA University, referred 585 consecutive patients with ischemic stroke. Ischemic stroke was defined as a stroke (a clinical syndrome characterized by rapidly developing clinical symptoms and signs of focal or global loss of brain function) with evidence of cerebral infarction in clinically relevant areas of the brain according to magnetic resonance imaging (MRI) scan finding. Based on clinical manifestations and neuroimaging data, two neurologists classified all ischemic strokes into four causative subtypes using the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria, as follows: (1) large artery disease (LAD), characterized by an infarction lesion ≥15 mm in diameter documented by MRI, and significant (>50%) stenosis of a major brain artery or a branch cortical artery documented by cerebral angiography with symptoms associated with that arterial territory; (2) small vessel disease (SVD), characterized by an infarction lesion <15 and ≥5 mm in diameter documented using MRI, and classic lacunar syndrome without evidence of cerebral cortical dysfunction or potentially detectable cardiac sources for embolism; (3) cardioembolism (CE) or arterial occlusions presumably due to an embolus arising in the heart, as detected by cardiac evaluation; and (4) undetermined pathogenesis, in which the cause of stroke could not be determined with any degree of confidence or involved >2 causes. Single and multiple (≥2 lesions) SVD cases were distinguished via brain MRI scans. The sizes and sites of cerebral infarctions were documented using MRI only. We selected 403 control subjects that were matched for sex ratio and age (within 5 years) in accordance with the patient group (Table 1). Controls were drawn from subjects visiting our hospitals during the same period for health examinations, including biochemical testing, electrocardiograms, and brain MRIs. Control subjects did not have a recent history of cerebrovascular disease or myocardial infarction. Exclusion criteria were the same as those used for the case group, as mentioned previously.

DNA was extracted from leukocytes using a G-DEX II Genomic DNA Extraction kit (Intron Biotechnology, Seongnam, Korea) according to the manufacturer’s instructions. The six best-studied single nucleotide polymorphisms (SNPs) in the miRNA biogenesis genes were determined through a documentary search that included 3′-UTR SNPs (DICER rs13078A>T, rs3742330A>G, DROSHA rs10719T>C, rs6877842G>C, RAN rs14035C>T, and XPO5 rs11077A>C). The miRNA biogenesis gene polymorphisms were analyzed by the polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method. The PCR conditions for miRNA biogenesis genes polymorphism analyses are presented in Supplementary Table 1. To validate RFLP findings, 30% of the PCR assays for each polymorphism were randomly selected and repeated, followed by DNA sequencing. Sequencing was performed using an ABI 3730×l DNA Analyzer (Applied Biosystems, Foster City, CA, USA). The concordance of these quality control samples with the RFLP results was 100%.

To evaluate the association between miRNA biogenesis gene polymorphisms and long-term prognosis after ischemic stroke, survival time from stroke onset to death was tracked. The dates of death for each stroke patient (n=585) were ascertained using death certificates from the Korean National Statistical Office. Patients who were alive on December 31, 2013 were excluded from the study.

Genotype and allele combination frequencies in ischemic stroke cases and controls were compared using multivariate logistic regression models and Fisher exact test, respectively. Allele frequencies were calculated to identify deviations from Hardy-Weinberg equilibrium using P=0.05 as a threshold. Odds ratios, adjusted odds ratios (AORs), and 95% confidence intervals (CIs) were used to measure the strength of association between various genotypes and ischemic stroke. The association between miRNA biogenesis gene SNPs and post-stroke mortality was evaluated using Cox proportional hazard regression. The proportional hazards assumption was tested using a log(-log[survival]) plot and interaction for follow-up time in a time-dependent Cox regression model, which was found to be satisfactory. For multivariate analyses, logistic regression analyses were used to adjust for possible confounders, including age, sex, hypertension, diabetes mellitus, hyperlipidemia, and smoking. Statistical significance was accepted at the P<0.05 level [28,29].

The demographic characteristics of the 585 stroke cases and 403 controls are presented in Table 1. Of the stroke and control samples, 41.5% and 41.7%, respectively, were men, and the mean ages of stroke cases and controls were 62.7±10.9 and 62.8±10.6 years, respectively.

Table 2 provides the genotype distributions of the six miRNA biogenesis gene polymorphisms in ischemic stroke cases and controls. The DICER rs3742330A>G polymorphism was associated with greater odds of ischemic stroke (AA vs. GG: AOR, 1.459; 95% CI, 1.000 to 2.126; P=0.050; and AA vs. AG+GG: AOR, 1.360; 95% CI, 1.024 to 1.807; P=0.034). The DROSHA rs10719T>C polymorphism was also associated with greater odds of ischemic stroke (TT vs. CC: AOR, 2.038; 95% CI, 1.113 to 3.730; P=0.021; and TT+TC vs. CC: AOR, 2.001; 95% CI, 1.106 to 3.621; P=0.022). By contrast, the XPO5 rs11077A>C polymorphism was associated with lower odds of stroke (AA vs. CC: AOR, 0.101; 95% CI, 0.011 to 0.951; P=0.045; and AA vs. AC+CC: AOR, 0.669; 95% CI, 0.473 to 0.945; P=0.023). The frequency of the DICER1 rs13078A>T, DROSHA ABI 3730×l DNA Analyzer C>G, and RAN rs14035C>T polymorphisms was not significantly different between stroke cases and controls. To examine whether the effect of each polymorphism was confined to a specific subtype, stroke patients were separated into three subgroups (LAD, SVD, and CE) according to TOAST classifications (Table 3). Comparisons were also performed with control subjects and single versus multiple SVD patients (Supplementary Table 2). LAD was not significantly associated with any of the polymorphisms examined. However, the DICER1 rs3742330A>G polymorphism was significantly associated with SVD (AA vs. AG: AOR, 1.705; 95% CI, 1.073 to 2.710; P=0.024; and AA vs. AG+GG: AOR, 1.616; 95% CI, 1.041 to 2.509; P=0.032). In addition, the DROSHA rs10719T>C polymorphism was significantly associated with CE (TT vs. CC: AOR, 3.451; 95% CI, 1.264 to 9.422; P=0.016; and TT+TC vs. CC: AOR, 3.499; 95% CI, 1.348 to 9.082; P=0.010).

Stratified analysis of each clinical factor was performed to confirm the influence of clinical factors on the occurrence of ischemic stroke. However, no significant clinical factors were found to affect ischemic stroke risk (Supplementary Table 3). Therefore, a combined effect analysis was conducted to ascertain the effect of stroke and genotype on the prevalence of ischemic stroke. A synergistic effect was found for ischemic stroke prevalence between clinical factors (hypertension and diabetes mellitus) and miRNA biogenesis gene polymorphisms (Figure 1). The DROSHA rs10719 CC genotype was associated with stroke in individuals with hypertension (AOR, 4.781; 95% CI, 1.981 to 11.54). Diabetes mellitus combined with the DROSHA rs10719 CC genotype yielded the most significant association with stroke (AOR, 12.05; 95% CI, 1.541 to 94.19). Other gene-clinical factor combinations were not significantly associated with ischemic stroke (Supplementary Tables 4-6). The effects of miRNA biogenesis gene genotypes on blood coagulation status were evaluated by measuring platelet proportion, prothrombin time, activated partial thromboplastin time (aPTT), fibrinogen, and antithrombin (Supplementary Table 7). It was found that the DROSHA rs10719 CC genotype was significantly associated with elevated aPTT (TT vs. CC: P=0.007; TC vs. CC: P=0.019) (Supplementary Figure 1A) and antithrombin (TC vs. CC: P=0.039) (Supplementary Figure 1B). The other coagulant factors did not exhibit any statistically significant associations with any of the tested genotypes.

To evaluate the association between miRNA biogenesis gene polymorphisms and post-stroke mortality, Cox regression analysis was performed on the 585 patients with total ischemic stroke according to TOAST subtype (Figure 2 and Supplementary Table 8). During a mean follow-up of 4.80±2.11 years, 99 of the stroke patients died. In the multivariate Cox proportional hazard regression models, a significant association was found between RAN rs14035 and survival of LAD patients with ischemic stroke (CC vs. TT: adjusted hazard ratio [HR], 5.978; P=0.015; and CC+CT vs. TT: adjusted HR, 3.946; P=0.034) (Figure 2). A significant association was also found between RAN rs14035 and SVD in our analysis of ischemic stroke subtypes (CC vs. TT: adjusted HR, 9.403; P=0.015; and CC+CT vs. TT: adjusted HR, 5.223; P=0.039) (Figure 2). However, survival analysis was performed by Cox proportional-hazards regression based on the stepwise method for confirming covariant effect. A stepwise Cox regression analysis of ischemic stroke-related survival is shown in Supplementary Table 9. Mortality in SVD subgroup of ischemic stroke cases was associated with age and RAN rs14039 polymorphism status.

Gene-gene interaction analyses were performed for miRNA biogenesis gene polymorphisms to identify combinations that have synergistic effects on stroke risk (Supplementary Table 10 and 11). Some variants and allele combinations exhibited significant associations. However, the meaning of these associations should be interpreted with caution because the sample size is rather small.