Intracranial Vessel Morphology in RNF213-Related Vasculopathy: A Quantitative Imaging Study in Ischemic Stroke Patients

Article information

J Stroke. 2026;28(3):393-404
Publication date (electronic) : 2026 September 30
doi : https://doi.org/10.5853/jos.2025.06282
1Office of Education and Human Resources Development, Samsung Medical Center, Seoul, Korea
2Department of Neurology and Stroke Center, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea
3Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University, Seoul, Korea
4URGENT Inc., Seoul, Korea
Correspondence: Woo-Keun Seo Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 06351, Korea Tel: +82-2-3410-0799 E-mail: mcastenosis@gmail.com
Co-correspondence: Oh Young Bang Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 06351, Korea Tel: +82-2-3410-0799 E-mail: neuroboy50@naver.com
Received 2025 December 7; Revised 2026 August 9; Accepted 2026 August 20.

Abstract

Background and Purpose

The RNF213 p.R4810K variant is a major susceptibility allele for moyamoya disease (MMD) in East Asians. This study investigated the relationship between RNF213 variants and intracranial vessel morphology in acute ischemic stroke.

Methods

Patients from the Samsung Medical Center (SMC) Stroke Registry (2017–2019) were included. Quantitative vessel features were extracted from magnetic resonance angiography, including the basal global tortuosity index (GTIb), internal carotid artery (ICA) area, and basal minimum-to-maximum diameter ratio (MMRb). Genome-wide association analyses of GTIb, ICA area, and MMRb were performed (n=945). Validation was conducted in an independent Korea University Medical Center (KUMC) cohort. Groupwise comparisons were performed across three predefined groups: MMD+/variant+, MMD-/variant+, and MMD-/variant-. Adjusted linear regressions for ICA area were performed. Receiver operating characteristic (ROC) analyses evaluated GTIb, ICA area, and a combined model for identifying RNF213 carriers.

Results

GTIb and ICA area showed genome-wide significance at the RNF213 locus (p.R4810K), whereas MMRb did not. In the KUMC cohort, ICA area differed between variant+ and variant- groups, while GTIb and MMRb did not. Across SMC groups (MMD+/variant+ [n=14], MMD-/variant+ [n=37], MMD-/variant- [n=888]), GTIb was highest and ICA area and MMRb smallest in MMD+/variant+, intermediate in MMD-/variant+, and lowest/largest in MMD-/variant- (P<0.001). ICA area differences remained significant after multivariable adjustment. In ROC analyses, ICA area provided significant incremental discrimination beyond clinical variables.

Conclusions

The RNF213 variant was consistently associated with smaller ICA area in patients with ischemic stroke. The association with intracranial tortuosity was not reproduced in the validation cohort.

Introduction

The RNF213 p.R4810K variant, formed by a single-nucleotide polymorphism (SNP) on chromosome 17, is a well-known susceptibility locus for moyamoya disease (MMD) in East Asian populations [1,2]. Recent studies suggest that RNF213 variants are not limited to classic MMD but may contribute to a broader spectrum of intracranial vasculopathies [1,3]. Beyond its association with MMD, RNF213 has been implicated in diverse cerebrovascular phenotypes such as intracranial arterial dissection and aneurysm [4,5], suggesting a pivotal role in vascular remodeling and endothelial dysfunction.

Previous reports have shown that carriers of RNF213 variants often have a lower atherosclerotic burden in the presence of intracranial stenosis [2] and that some patients with middle cerebral artery steno-occlusive disease harboring these variants can progress to MMD [6]. These findings indicate that RNF213 carriers may require careful monitoring for risk management and early prevention of stroke progression.

Apart from the importance of the RNF213 variant, analyzing vessel morphology helps stratify the risk of intracranial atherosclerosis (ICAS) and stroke [7]. Emerging evidence has revealed a link between wall shear stress and subsequent cerebrovascular outcomes [8,9], with wall shear stress being modulated by vessel morphology. Although the clinical relevance has been suggested, quantitative metrics of vessel morphology have not yet been incorporated into clinical practice. In addition, the genetic basis of vessel morphology—including tortuosity, caliber, and circularity—remains poorly understood.

While MMD and RNF213-associated vasculopathy are not primarily defined by vascular geometry including tortuosity [10], several experimental and computational studies have suggested a tight link between vessel morphology, wall shear stress, and vessel wall fragility [11]. Computational fluid dynamics analyses have shown that increased arterial tortuosity and curvature can locally augment wall shear stress and other hemodynamic stresses, which are implicated in cerebrovascular lesion formation and progression [12-14]. In parallel, recent work has demonstrated that loss of RNF213 function in brain endothelial cells disrupts endothelial barrier integrity and endothelial junctions [15,16] and renders endothelial cells unusually vulnerable to shear stress [17], supporting a role for RNF213 in mechanosensitive vascular homeostasis. Based on these observations, we hypothesized that RNF213 variants may secondarily influence the geometry of major intracranial arteries and that quantitative descriptors of tortuosity, luminal circularity, and caliber might serve as surrogate markers of RNF213-related vasculopathy and vascular event risk. Previous MMD studies have predominantly focused on the development of moyamoya collaterals, whereas potential RNF213-related remodeling of the major intracranial arteries themselves has not been systematically evaluated, despite its potential clinical implications.

In this study, we aimed to identify genetic variants influencing intracranial vessel tortuosity, luminal circularity—represented by mean minimum-maximum ratio of vessel diameter—and internal carotid artery (ICA) caliber, using quantitative imaging features derived from patients with acute ischemic stroke. Following the genome-wide association analysis, we further examined the association between RNF213 variants and intracranial vessel morphology and validated the findings in an independent cohort. Together, these analyses aimed to provide insight into how genetic variation influences intracranial vessel morphology.

Methods

Study design and participants

This retrospective study utilized data from the Samsung Medical Center (SMC) Stroke Registry, a prospective, hospital-based registry integrated into the electronic medical record system. Patients admitted for acute ischemic stroke between January 2017 and December 2019 were included. Patients without appropriate imaging sequences (time-of-flight [TOF] magnetic resonance [MR] angiography), those whose imaging was performed outside SMC, or those with failed vessel feature labeling were excluded.

The protocol for the SMC Stroke Registry was approved by the Institutional Review Board (IRB no. 2016-08-064), and written informed consent was obtained from all participants or their legal guardians. The present study was additionally approved by the SMC IRB (IRB no. 2021-04-072). The requirement for additional consent was waived due to the retrospective design and use of de-identified registry data.

For external validation, data were obtained from the Korea University Medical Center (KUMC) Stroke Registry, another prospective, hospital-based registry. Patients admitted for acute ischemic stroke between January and December 2014 were included. Patients without appropriate imaging sequences, with failed feature extraction or labeling, or without RNF213 genotype data were excluded. The study was approved by the KUMC IRB (IRB no. KUGH 2019GR0300).

Image data acquisition and vessel feature labeling

The intracranial arteries were imaged using a 3.0 T Philips Achieva magnetic resonance imaging (MRI) scanner (Philips Medical Systems, Best, The Netherlands), equipped with a 32-element phased-array receiver head coil for the SMC stroke registries, and a 1.5 T Magnetom Sonata MRI scanner (Siemens Medical Solutions, Erlangen, Germany) equipped with a 16-channel phased-array head coil for external validation images of patients in the KUMC stroke registry. This study used whole-brain three-dimensional MR angiography images collected from each participant using the TOF protocol. For the SMC cohort, the voxel size was 0.284×0.284×0.284 mm3, with the following parameters: echo time, 4.59 ms; repetition time, 22 ms; flip angle, 23°; receiver bandwidth (RBW), 130 Hz/pixel; generalized autocalibrating partially parallel acquisitions factor, 3; and 32 reference lines. For the KUMC cohort, the voxel size was 0.612×0.612×0.612 mm³, with the following parameters: echo time, 4.97 ms; repetition time, 37 ms; flip angle, 25°; RBW, 65 Hz/pixel; and 16 reference lines.

Overall procedure for calculating cerebrovascular geometric features was reported previously [18]. In brief, we built our pipeline based on Vascular Modeling Toolkit, a vascular modeling toolkit, and made the following adaptations for brain vessel quantification. A candidate brain vessel mask was extracted using a region-growing algorithm [19]. The surface nodes and edges of the binary vessel mask were calculated using a marching-cube algorithm, and adjusted to isotropic triangles by applying spatial interpolation to the coordinates of each surface node [20]. Centerline tracking was performed using the edges of the Voronoi diagram from the bottom part of the ICA or basilar artery as the starting point and the termination of each brain vessel as the endpoint [21]. Various morphological characteristics of the brain vessels were calculated using the obtained centerline. From all the centerline points, the sectional area and minimum-to-maximum diameter ratio in the section perpendicular to the centerline direction were calculated. In addition, the curvature, a structural characteristic of the centerline, was calculated. Vessel segmentation and arterial labeling were conducted according to previously validated frameworks [22]. For ICA area analysis, all ICA chunks visually identifiable on TOF MR angiography, from the inlet of the carotid canal to the ICA terminus, were included, and the mean sectional luminal area across the extracted ICA segments, without the exclusion of stenotic portions, was used as the representative ICA area for each subject. ICA area was expressed in mm2. All cross-sectional measurements from the left and right ICAs were pooled before calculating the subject-level mean. The basal global tortuosity index (GTIb) was calculated as described in prior studies, standardizing tortuosity metrics of the Circle of Willis [7]. The basal minimum-to-maximum diameter ratio (MMRb) was computed using an identical standardization method to GTIb.

Calculation of intracranial atherosclerotic burden

ICAS burden was quantified based on previously described methods [7]. Briefly, two neurologists independently reviewed the TOF MR angiography images. Each intracranial arterial segment was evaluated for the presence of ≥50% stenosis or occlusion. ICAS burden was defined as the total number of affected intracranial arteries, yielding a cumulative score ranging from 0 to 7. ICAS burden data incorporated in the present analysis were obtained from prior studies and were therefore available only for a subset of participants [7].

Genome-wide association analysis

DNA samples were genotyped using the Illumina Global Screening Array MD BeadChip (Illumina, San Diego, CA, USA). Sample-and variant-level quality control procedures—including filters based on call rate, heterozygosity, relatedness, Hardy–Weinberg equilibrium, and minor allele frequency thresholds—were performed using an identical analytical pipeline as described previously [23]. The same genotyping method was implemented in both the SMC and KUMC cohorts. Genotypes were phased and imputed using the Michigan Imputation Server with the Haplotype Reference Consortium reference panel (r1.1 2016) [24].

Genome-wide association analyses were performed using linear regression models adjusted for age, sex, and the first four principal components, following the prior framework [23]. Age and sex were included given their influence on vascular structure [25]. The first four principal components were included as covariates to account for population stratification, consistent with the apparent elbow observed after the fourth component in the scree plot (Supplementary Figure 1) [26]. Variants with P<5.0×10-8 (Bonferroni-corrected genome-wide significance threshold) were considered genome-wide significant.

In the KUMC cohort, the RNF213 p.R4810K SNP was excluded during quality control. The genotype information for this SNP was imputation-derived and was obtained from the original imputed genotype dataset. Group comparisons for this SNP were performed using the Wilcoxon test rather than meta-analysis.

Subgroup analysis of vessel features

Patients were categorized into three groups according to the presence of RNF213 p.R4810K variant and MMD diagnosis: (1) variant-positive with MMD (MMD+/variant+), (2) variant-positive without MMD (MMD-/variant+), and (3) variant-negative without MMD (MMD-/variant-).

The presence of the RNF213 variant was determined based on genotype data. MMD was defined on the basis of the Diagnostic Criteria 2021 for MMD from the Research Committee on Moyamoya Disease [27], which require steno-occlusive changes at the terminal ICA or the proximal anterior/middle cerebral arteries together with the presence of basal collateral networks. Final classification in the stroke registry incorporated the physician’s judgment, taking into account noninvasive vascular imaging, digital subtraction angiography (DSA) findings when available, and the overall clinical context. Accordingly, both “suspected on noninvasive imaging” and “confirmed with DSA” cases recorded in the registry were included in the MMD+ group. Patients without findings compatible with the MMD diagnostic criteria were categorized as non-MMD. The non-MMD groups comprised individuals with diverse Trial of ORG 10172 in Acute Stroke Treatment (TOAST) etiologies—including large artery atherosclerosis (LAA), cardioembolism, small-vessel occlusion, and undetermined causes—reflecting the consecutive inclusion of all acute ischemic stroke patients irrespective of etiology.

Statistical analysis

All statistical analyses were performed in RStudio using R version 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were summarized as mean±standard deviation or median (interquartile range), and categorical variables as counts (percentages). Baseline characteristics across the three subgroups (MMD-/variant-, MMD-/variant+, and MMD+/variant+) were compared using Kruskal–Wallis tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables. Because of the small size of the MMD+/variant+ group, Fisher’s exact test was used where expected cell counts were <5. Group comparisons of vessel features were performed using the Kruskal–Wallis test with post hoc analysis (Bonferroni-adjusted pairwise Wilcoxon test). Normality and homogeneity of variance were assessed before testing.

Multivariable linear regression analyses were performed to evaluate the independent association between ICA area and RNF213 variant status. ICA area was analyzed following log transformation. The primary model included RNF213 variant status, stroke etiology (LAA vs. non-LAA), age, and sex. Additional analyses were restricted to patients without moyamoya disease (non-MMD). Stratified analyses were conducted according to LAA status within the non-MMD group. The interaction between RNF213 variant status and LAA was examined. ICAS burden was subsequently included as an additional covariate. To assess potential selection bias related to the availability of ICAS burden data, baseline characteristics were compared between non-MMD participants with and without available ICAS burden data using the Wilcoxon rank-sum test for continuous variables and the chi-square or Fisher’s exact test for categorical variables, as appropriate. Among participants with available ICAS burden data, the multivariable regression model was repeated both before and after additional adjustment for ICAS burden to evaluate the consistency of the association.

To evaluate potential laterality effects, ICA area on the ipsilateral and contralateral sides of the index lesion was additionally analyzed. Patients with bilateral lesions or incomplete laterality-specific ICA measurements were excluded. Laterality-specific ICA area values were analyzed following log transformation using linear regression models adjusted for age, sex, and LAA status.

Receiver operating characteristic (ROC) analyses were used to evaluate the predictive performance of GTIb, ICA area, and a combined logistic regression model for identifying RNF213 variant carriers. Nested ROC analyses were conducted by sequentially comparing the clinical model (including age, sex, and stroke etiology [LAA vs. non-LAA]), the clinical model plus ICA area, and the clinical model plus ICA area and GTIb. The area under the curve (AUC) and corresponding 95% confidence intervals were calculated. Differences in AUC between correlated ROC curves were assessed using DeLong’s test. To further evaluate the incremental discriminative value of ICA area and GTIb, paired bootstrap resampling (2,000 iterations) was performed to estimate the distribution of ΔAUC and its 95% confidence intervals.

Results

In the SMC cohort, a total of 945 patients were included after image- and genotype-level quality control. The KUMC validation cohort included a total of 632 patients after quality control and selection of those with available RNF213 genotype data. The inclusion and exclusion process is shown in Supplementary Figure 2, and baseline characteristics of the study population are summarized in Table 1. ICA-related analyses, including genome-wide association study and group comparisons, were conducted in 943 patients after excluding two individuals with missing ICA area measurements.

Descriptive statistics of clinical variables of the SMC and KUMC cohorts

In the genome-wide association analysis, both GTIb and ICA area demonstrated genome-wide significant associations at the RNF213 locus, annotated to RNF213-AS1 and the RNF213 p.R4810K variant (Figure 1 and Supplementary Figure 3). For ICA area, additional SNPs also reached genome-wide significance, as summarized in Supplementary Table 1. These variants were not the primary focus of the present study and were not further evaluated for independence or replication. These findings suggest that RNF213 variants influence both vascular tortuosity and arterial caliber. However, RNF213 variants were not significantly associated with MMRb, and no SNPs reached genome-wide significance for this feature.

Figure 1.

Manhattan plots of genome-wide association results for intracranial vessel features in the SMC cohort. Results for the GTIb (A), ICA area (B), and MMRb (C) are shown. The red horizontal line indicates the genome-wide significance threshold (P=5.0×10-8), and the blue line indicates the suggestive threshold (P=1.0×10-5). The genome-wide significant locus on chromosome 17 is highlighted in green, and the most significant single-nucleotide polymorphism, encoding RNF213 p.R4810K, is annotated. SMC, Samsung Medical Center; GTIb, basal global tortuosity index; ICA, internal carotid artery; MMRb, basal minimum-to-maximum diameter ratio.

Among the 945 patients in the SMC cohort, 14 were MMD+/variant+, 37 were MMD-/variant+, 888 were MMD-/variant-, and 6 were MMD+/variant-. Because the primary comparison was designed to evaluate vascular phenotypes associated with the RNF213 p.R4810K variant while accounting for the presence of MMD, the small MMD+/variant- subgroup (n=6) was excluded from the three-group comparison. The baseline characteristics of each group are summarized in Supplementary Table 2. The MMD+/variant+ group showed the highest GTIb values and the smallest ICA area and MMRb compared with the other groups. GTIb, ICA area, and MMRb differed significantly among the three groups (Kruskal–Wallis test, P<0.001 for all). In post hoc comparisons, both the MMD+/variant+ and MMD-/variant+ groups had significantly higher GTIb and smaller ICA area and MMRb than the MMD-/variant- group (Bonferroni-adjusted P<0.05) (Figure 2). The MMD+/variant+ group also had higher GTIb and smaller ICA area and MMRb than the MMD-/variant+ group, but these differences in GTIb and MMRb did not reach statistical significance (Figure 2). Age also differed significantly among the groups: the MMD+/variant+ group was the youngest, while the MMD-/variant- group was the oldest.

Figure 2.

Group-wise comparison of intracranial vessel features across the RNF213–MMD spectrum in the SMC cohort. Box-and-jitter plots show the GTIb (A), ICA area (B), MMRb (C), and age (D) across three predefined groups: MMD-/variant-, MMD-/variant+, and MMD+/variant+. Six patients in the MMD+/variant- group were not included in the comparison because of the small size of this subgroup. Boxes indicate the IQR with the median as a horizontal line; whiskers extend to 1.5×IQR. Kruskal–Wallis tests were followed by Bonferroni-adjusted pairwise Wilcoxon tests. n.s., not significant; GTIb, basal global tortuosity index; MMD, moyamoya disease; ICA, internal carotid artery; MMRb, basal minimum-to-maximum diameter ratio; SMC, Samsung Medical Center; IQR, interquartile range. *P<0.05; **P<0.01; ***P<0.001.

In the KUMC validation cohort, 19 patients carried the RNF213 variant and 613 did not. ICA area was significantly smaller in the variant+ group, whereas GTIb and MMRb did not differ significantly (Supplementary Figure 4).

To evaluate the independent association between ICA area and RNF213 variant status, multivariable linear regression analyses were performed adjusting for age, sex, and LAA (Table 2). Following log transformation, ICA area remained significantly associated with RNF213 variant status in both the overall cohort and the non-MMD cohort (P<0.001 for both models). This association was consistently observed across LAA-stratified analyses (P<0.001 for both subgroups).

Multivariable regression models for log-transformed ICA area

The interaction between RNF213 variant status and LAA was not statistically significant, indicating no evidence that the association between RNF213 variant status and ICA area differed according to LAA status.

ICAS burden data were available for 647 of the 923 patients included in the non-MMD analysis. Participants with and without available ICAS burden data were generally comparable, although the distribution of TOAST subtypes and ICA area differed between the groups (Supplementary Table 3). In the ICAS-available subset, the association between RNF213 variant status and ICA area was comparable to that observed in the full non-MMD cohort and remained materially unchanged after additional adjustment for ICAS burden.

Whether the association between RNF213 variant status and ICA area differed according to lesion laterality was additionally evaluated. ICA area on the ipsilateral and contralateral sides of the index lesion was calculated, and linear regression analyses were performed following log transformation (Supplementary Table 4). RNF213 variant status was significantly associated with both ipsilateral and contralateral ICA area (P<0.001 for both models).

ROC analyses were performed to evaluate the discriminative performance of GTIb, ICA area, and their combination for identifying RNF213 variant carriers in patients without MMD (Figure 3A). The AUCs were 0.659 for GTIb, 0.777 for ICA area, and 0.800 for the combined model (ICA area plus GTIb).

Figure 3.

ROC curves for discrimination of RNF213 variant carriers in the non-MMD group. (A) Unadjusted ROC curves for ICA area (dark blue), GTIb (light blue), and the combined logistic regression model incorporating both ICA area and GTIb (green). (B) Adjusted ROC curves based on multivariable logistic regression models including clinical covariates (age, sex, and stroke etiology [large artery atherosclerosis vs. non-large artery atherosclerosis]). The models depict the clinical model alone (black), clinical model plus ICA area (blue), and clinical model plus ICA area and GTIb (green). ICA, internal carotid artery; GTIb, basal global tortuosity index; AUC, area under the curve; ROC, receiver operating characteristic; MMD, moyamoya disease.

To assess the incremental contribution of ICA area and GTIb beyond clinical variables, nested ROC analyses were performed (Figure 3B and Table 3). ICA area provided a significant improvement in model discrimination, whereas the incremental effect of GTIb did not reach statistical significance, despite a numerical increase in AUC.

Discriminative performance of models for identifying RNF213 variant carriers in the non-MMD group

Paired bootstrap resampling (2,000 iterations) was subsequently performed to assess the robustness of the AUC difference given the small number of variant carriers. In these analyses, both ICA area and GTIb demonstrated a positive incremental effect on the discrimination of RNF213 variant status (Supplementary Figure 5).

ROC analyses in the overall study population yielded results consistent with those observed in the non-MMD subgroup (Supplementary Figure 5 and Supplementary Table 5). In the KUMC validation cohort, ICA area retained its incremental discriminative value, while the contribution of GTIb was comparatively modest (Supplementary Figure 6 and Supplementary Table 5).

The AUC of the combined model (clinical variables plus ICA area plus GTIb) was 0.813 in the non-MMD population, 0.855 in the overall population, and 0.739 in the KUMC population, indicating fair-to-good discriminatory performance for identifying RNF213 variant carriers.

Discussion

This study demonstrated an association between the RNF213 p.R4810K variant and quantitative intracranial vascular morphology among patients with acute ischemic stroke, most consistently reflected by smaller ICA area. Increased GTIb and lower MMRb were observed in the SMC cohort, although these findings were not replicated in the KUMC validation cohort. Accordingly, these findings should be interpreted within a stroke-manifesting population, and extrapolation to asymptomatic carriers or non-ischemic phenotypes warrants caution. Carriers of the variant exhibited increased vascular tortuosity and smaller ICA caliber, both reaching genome-wide significance. Although MMRb did not reach genome-wide significance, lower values were observed in RNF213 variant carriers in the group-wise analysis. These morphological alterations were consistently seen in RNF213 carriers even in the absence of clinically or radiologically confirmed MMD. The association between the RNF213 variant and ICA area was replicated in the KUMC validation cohort, whereas the associations with GTIb and MMRb were not.

The frequency of the RNF213 p.R4810K variant differed between the SMC and KUMC cohorts. This difference should be interpreted in the context of the distinct cohort compositions. The SMC cohort included patients with clinically diagnosed MMD, whereas patients with clinically diagnosed MMD were excluded from the KUMC cohort. Accordingly, these frequencies should not be interpreted as estimates of the variant prevalence in the general population.

Multivariable linear regression analyses showed that the association between RNF213 variant status and smaller ICA area persisted after adjustment for age, sex, and stroke etiology (LAA vs. non-LAA), and remained robust in sensitivity analyses additionally accounting for ICAS burden. The association was observed for both ipsilateral and contralateral ICA areas relative to the index lesion, supporting the interpretation that the finding is not solely driven by territory-specific effects of the acute event. Nevertheless, residual confounding related to stenosis severity and segment-level distribution cannot be fully excluded due to the absence of segment-specific analyses.

In nested ROC analyses, ICA area provided a significant incremental improvement beyond clinical variables, whereas the additional contribution of GTIb was modest and did not reach statistical significance by DeLong’s test despite a numerical increase in AUC. Given the class imbalance in variant status, paired bootstrap resampling replicated the positive incremental trend. However, the clinical utility of these metrics should be considered primarily within stroke populations rather than as a general-population screening tool.

The variant+ groups showed significantly younger age compared with the variant- group. Although approximately one-quarter of the cohort had a prior history of stroke or transient ischemic attack, this finding suggests that RNF213 carriers experience ischemic events at a younger age. This aligns with previous studies reporting the association between RNF213 p.R4810K and early-onset intracranial arterial disease [28,29]. Together, these results suggest that RNF213-related vasculopathy may predispose to earlier manifestation of stroke through increased susceptibility to vessel wall remodeling and hemodynamic stress, even in the absence of clinically evident MMD.

The lack of external replication for GTIb and MMRb may reflect limited measurement reproducibility or heterogeneity arising from differences in imaging protocols and acquisition parameters between hospitals. Alternatively, the discrepancy may be attributed to incomplete phenotypic penetrance—meaning that not all carriers exhibit measurable vascular remodeling in terms of tortuosity or luminal circularity.

In our study, RNF213 variant+ MMD accounted for approximately 1.5% of the cohort, and variant+ but non-MMD stroke accounted for 3.9%. These proportions differ from those reported in the earlier single-center study from our institution [30]. This discrepancy originates primarily from the different sampling frameworks. The previous study selectively enrolled patients who presented with ischemic events in the middle cerebral artery territory and had relevant intracranial arterial stenosis, thereby enriching the cohort for individuals with MMD or MMD-suspected ICAS. In contrast, our study included all consecutive acute ischemic stroke patients regardless of stroke etiology, encompassing cardioembolic, small-vessel, large-artery, and undetermined mechanisms. As a result, our cohort contained a substantially broader spectrum of vascular phenotypes, and the distribution of RNF213 carriers naturally differed from previous reports.

Recent experimental studies have begun to elucidate the biological mechanisms through which RNF213 variants influence cerebrovascular structure and function. RNF213 encodes a large AAA+ ATPase (ATPase associated with diverse cellular activities)–E3 ubiquitin ligase that participates in endothelial homeostasis and vascular remodeling. Loss or mutation of RNF213 in endothelial cells of the blood-brain barrier disrupts barrier integrity and tight-junction organization, leading to increased endothelial permeability and impaired angiogenic response [15]. In transcriptomic and fluid dynamics studies, RNF213-deficient endothelial cells show an abnormal response to shear stress, supporting the concept that this gene contributes to mechanosensitive regulation of the intracranial circulation [16,17]. These cellular alterations—impaired angiogenic signaling, compromised barrier function, and heightened vulnerability to hemodynamic stress—may collectively promote vessel-wall fragility and maladaptive remodeling, providing a biological basis for the morphological differences observed in RNF213 variant carriers.

While distal ICA stenosis is a well-known feature of MMD [31], the morphological correlates of RNF213-related vasculopathy have not been well characterized quantitatively in vivo. Our findings suggest that in acute ischemic stroke, RNF213 variants should be considered when decreased ICA area or increased diffuse vascular tortuosity is observed. Because ICA area was defined as the mean value across ICA chunks, whether the observed reduction primarily reflects focal stenosis or diffuse negative remodeling warrants further investigation. Given that this polymorphism has been associated with adverse outcomes including recurrent stroke and MMD progression [6,32], careful attention to vessel morphology may aid in risk identification and improve clinical management.

Although vascular tortuosity did not show a significant association with RNF213 status in the validation cohort, the SMC findings raise the possibility that tortuosity of the major intracranial arteries may represent an additional, exploratory feature of RNF213-related vascular remodeling. In the present study, analyses were restricted to the vessels of the Circle of Willis, excluding pial collaterals. MMD is classically characterized by tortuous collateral networks that develop as compensatory responses to progressive steno-occlusive changes in the Circle of Willis [33]. However, our data indicate that increased tortuosity in MMD and RNF213 carriers is not limited to collaterals but may also involve the major intracranial arteries of the Circle of Willis. Whether this tortuosity arises congenitally or results from progressive wall fragility and shear stress–induced deformation remains uncertain, but prior studies suggest that such alterations impose chronic mechanical burden on the intracranial circulation [7,34].

Interestingly, a previous study reported that RNF213 variant carriers exhibited lower ICA tortuosity [35]. These findings are not necessarily inconsistent, as the previous study evaluated local tortuosity of the ICA segment, whereas GTIb represents a global tortuosity index derived from the entire Circle of Willis. These results together suggest that RNF213-related vascular remodeling may be spatially heterogeneous, with structural alterations particularly pronounced beyond the distal ICA. The mechanisms underlying this heterogeneity remain incompletely understood. However, our findings raise the possibility that hemodynamic stress may contribute to mechanosensitive remodeling, particularly given the observed association between RNF213 variant status and vascular phenotypes such as ICA area.

MMRb, an index representing luminal circularity, does not inherently indicate a favorable or unfavorable vascular condition, as both higher and lower values may result from different biological processes. The lower MMRb observed in the variant+ groups, particularly in MMD+/variant+, may reflect underlying vessel wall remodeling or atherosclerotic deformation. Negative remodeling, including intimal fibromuscular hyperplasia described in RNF213-related vasculopathy [36], may contribute to luminal narrowing and altered vessel morphology. Additionally, irregular atherosclerotic plaque formation may lead to reduced circularity and a flatter lumen. A previous study reported that ICAS in RNF213 variant carriers more often exhibits tandem stenotic lesions compared with non-carriers [29]. Our findings and interpretation are consistent with those of the previous study. In contrast, the older MMD-/variant- group showed higher MMRb. Age-related vascular stiffening and calcification [37] may have contributed to this result, which can make the lumen appear more circular. Therefore, the observed MMRb pattern likely reflects a combination of remodeling processes in RNF213 carriers and aging-related changes in variant- individuals, although further imaging studies such as high-resolution MRI studies are required for confirmation.

Moreover, the ROC curve analyses suggest potential clinical utility of vessel morphology as a surrogate marker for identifying RNF213 carriers. Although this variant confers high risk of MMD, with odds ratios exceeding 50 in Japan and Korea but lower in Chinese cohorts [38-43], only about 1 in 150 carriers develops MMD [10]. It is also found in approximately 20% of patients with ICAS [36,44,45] and in 1%–2% of general East Asian population [40,46,47], indicating that RNF213 serves as a susceptibility rather than a diagnostic gene. ICA area alone or in combination with GTIb demonstrated fair discrimination of variant carriers. Combined with clinical variables, the AUC showed good discriminatory performance even in patients without MMD. In this context, quantitative vessel morphology, which may reflect underlying hemodynamic vulnerability, could serve as non-invasive markers for identifying individuals at increased risk of RNF213-related vasculopathy.

This study has several limitations. First, the study population consisted of patients with acute ischemic stroke; therefore, the distribution of RNF213 variants and vascular phenotypes may differ from those of the general population. Second, although the cohort size was relatively large for an imaging-genetic investigation, the study remained underpowered to detect additional loci with modest effect sizes. Third, external replication of the genetic associations was not performed by meta-analysis. Fourth, some individuals classified as MMD-/variant+ might represent preclinical or subclinical MMD rather than true non-MMD cases, because RNF213 carriers can develop MMD over time.

In addition, the vascular phenotypes were derived from a TOF MR angiography based segmentation and centerline pipeline, which may be susceptible to signal-related artifacts and discontinuities, particularly in severely stenotic segments. Moreover, ICAS burden data were available only for a subset of participants, and the distribution of TOAST subtypes and ICA area differed according to data availability; therefore, selection bias related to the ICAS-adjusted analysis cannot be fully excluded. Nevertheless, the association estimate in the ICAS-available subset was comparable before and after adjustment for ICAS burden.

Finally, the small number of RNF213 carriers limited statistical power for subgroup analyses and constrained the interpretability of comparisons between the MMD- and MMD+ groups. The KUMC validation cohort was also relatively small, with only 19 carriers available for comparison. Moreover, 141 patients were excluded due to the absence of appropriate image sequences. If these exclusions were non-random with respect to vascular phenotype or genotype, selection bias may have influenced the replication results. We were unable to fully assess whether exclusions differed according to stenosis severity or genotype.

In conclusion, the RNF213 p.R4810K variant was consistently associated with smaller ICA area in patients with acute ischemic stroke, including those without clinically evident MMD. Increased intracranial vascular tortuosity was observed in the discovery cohort but was not replicated externally and should therefore be considered exploratory. These findings suggest that quantitative ICA morphology may reflect aspects of RNF213-related vasculopathy beyond classic MMD, although its biological and clinical implications require confirmation in larger and more diverse populations.

Supplementary materials

Supplementary materials related to this article can be found online at https://doi.org/10.5853/jos.2025.06282.

Supplementary Table 1.

Genome-wide significant SNPs associated with GTIb and ICA area

jos-2025-06282-Supplementary-Table-1.pdf
Supplementary Table 2.

Baseline characteristics according to RNF213 variant and MMD status

jos-2025-06282-Supplementary-Table-2.pdf
Supplementary Table 3.

Baseline characteristics of participants without moyamoya disease according to the availability of intracranial atherosclerotic burden data

jos-2025-06282-Supplementary-Table-3.pdf
Supplementary Table 4.

Association between RNF213 variant status and laterality-specific ICA area in patients without moyamoya disease

jos-2025-06282-Supplementary-Table-4.pdf
Supplementary Table 5.

Discriminative performance of models for identifying RNF213 variant carriers in the overall study population of the SMC cohort and the KUMC cohort

jos-2025-06282-Supplementary-Table-5.pdf
Supplementary Figure 1.

Scree plot of genetic principal components. Eigenvalues of the first 10 principal components derived from the qualitycontrolled autosomal genotype data are shown. The eigenvalues declined more steeply across the first four components and showed a flatter pattern thereafter, indicating an apparent elbow after the fourth principal component.

jos-2025-06282-Supplementary-Figure-1.pdf
Supplementary Figure 2.

Flowchart of participant inclusion for the SMC cohort (A) and the KUMC cohort (B). SMC, Samsung Medical Center; KUMC, Korea University Medical Center.

jos-2025-06282-Supplementary-Figure-2.pdf
Supplementary Figure 3.

Quantile–quantile plots showing the observed versus expected −log10(P) distributions for the genome-wide association analyses of GTIb (A), ICA area (B), and MMRb (C). The red line indicates the null hypothesis of no association. The genomic inflation factors (λ) were 1.00 for GTIb, 0.99 for ICA area and 1.00 for MMRb, indicating no systematic inflation of test statistics. GTIb, basal global tortuosity index; ICA, internal carotid artery; MMRb, basal minimum-to-maximum diameter ratio.

jos-2025-06282-Supplementary-Figure-3.pdf
Supplementary Figure 4.

Validation of intracranial vessel feature distributions according to RNF213 variant status in the KUMC cohort. Box-and-jitter plots compare the GTIb (A), ICA area (B), MMRb (C), and age (D) between variant- and variant+ groups. RNF213 p.R4810K carriers showed smaller ICA area, consistent with the SMC cohort results. Statistical comparisons were performed using the Wilcoxon rank-sum test. n.s., not significant; GTIb, basal global tortuosity index; ICA, internal carotid artery; MMRb, basal minimum-to-maximum diameter ratio; KUMC, Korea University Medical Center; SMC, Samsung Medical Center. ***P<0.001.

jos-2025-06282-Supplementary-Figure-4.pdf
Supplementary Figure 5.

Bootstrap ΔAUC analyses for the non-MMD cohort and ROC curves for the overall population. (A) Bootstrap distributions of the incremental change in the AUC (ΔAUC) within the non-MMD cohort. The left panel illustrates the ΔAUC associated with adding ICA area to the clinical model, while the right panel illustrates the ΔAUC associated with adding GTIb to the clinical plus ICA area model. Solid lines represent kernel density estimates, and dashed vertical lines indicate the null hypothesis (ΔAUC=0). (B and C) ROC curves illustrating the discriminative performance of cerebrovascular geometric features for identifying RNF213 variant carriers. (B) Unadjusted ROC curves derived from the overall study population. The curves depict the diagnostic performance of ICA area (dark blue), GTIb (light blue), and the combined logistic regression model incorporating both ICA area and GTIb (green). (C) Adjusted ROC curves derived from multivariable logistic regression models including clinical covariates (age, sex, and stroke etiology [large artery atherosclerosis vs. non-large artery atherosclerosis]). The models represent the clinical model alone (black), clinical model plus ICA area (blue), and clinical model plus ICA area and GTIb (green). AUC, area under the curve; MMD, moyamoya disease; ICA, internal carotid artery; GTIb, basal global tortuosity index; CI, confidence interval; ROC, receiver operating characteristic.

jos-2025-06282-Supplementary-Figure-5.pdf
Supplementary Figure 6.

ROC curves for discrimination of RNF213 variant carriers in the KUMC validation cohort. (A) Unadjusted ROC curves for ICA area (dark blue), GTIb (light blue), and the combined logistic regression model incorporating both ICA area and GTIb (green). (B) Adjusted ROC curves based on multivariable logistic regression models including clinical covariates (age, sex, and stroke etiology [large artery atherosclerosis vs. non-large artery atherosclerosis]). The models depict the clinical model alone (black), clinical model plus ICA area (blue), and clinical model plus ICA area and GTIb (green). ICA, internal carotid artery; GTIb, basal global tortuosity index; AUC, area under the curve; ROC, receiver operating characteristic.

jos-2025-06282-Supplementary-Figure-6.pdf

Notes

Funding statement

This research was supported by the Bio&Medical Technology Development Program of the National Research Foundation (NRF) funded by the Korean government (MSIT) (No. RS-2023-00265393).

Conflicts of interest

Woo-Keun Seo is affiliated with URGENT Inc. The vascular quantification pipeline used in the present study was a pre-existing research pipeline developed independently of the pipeline currently used by URGENT Inc. URGENT Inc. had no role in the study design, data analysis, interpretation of the results, or preparation of the manuscript. The authors declare no conflicts of interest relevant to this study.

Author contribution

Conceptualization: Woo-Keun Seo, I Joon Han. Study design: Woo-Keun Seo, I Joon Han. Methodology: Ha-Na Song, Seong-Hoon Lee, Yun-Hwa Ha, In-Young Baek, Ji-Eun Lee, Seongvin Ju. Data collection: Hyung Jun Kim, Hyun Kyung Kim, Jong-Won Chung, Gyeong-Moon Kim, Oh Young Bang, Woo-Keun Seo. Investigation: I Joon Han. Statistical analysis: I Joon Han. Writing—original draft: I Joon Han. Writing—review & editing: Woo-Keun Seo, I Joon Han. Funding acquisition: Woo-Keun Seo. Approval of final manuscript: all authors.

References

1. Bang OY, Chung JW, Kim DH, Won HH, Yeon JY, Ki CS, et al. Moyamoya disease and spectrums of RNF213 vasculopathy. Transl Stroke Res 2020;11:580–589.
2. Ohara M, Yoshimoto T, Okazaki S, Gon Y, Todo K, Sasaki T, et al. RNF213 p.R4810K variant carriers with intracranial arterial stenosis have a low atherosclerotic burden. J Atheroscler Thromb 2022;29:1655–1662.
3. Sadhukhan D, Mitra P, Mishra S, Roy A, Podder G, Ray BK, et al. Arg4810Lys mutation in RNF213 among Eastern Indian non-MMD ischemic stroke patients: a genotype-phenotype correlation. Neurol Sci 2024;45:315–319.
4. Li Y, Liu J, Hu C, Luo C, Zhou J, Li B, et al. Association of rare RNF213 variants and intracranial aneurysm risk in a Chinese population. Ann Transl Med 2022;10:1336.
5. Zedde M, Grisendi I, Assenza F, Napoli M, Moratti C, Pavone C, et al. RNF213 polymorphisms in intracranial artery dissection. Genes (Basel) 2024;15:725.
6. Sasagasako T, Mineharu Y, Funaki T, Fushimi Y, Chihara H, Park S, et al. RNF213 mutation associated with the progression from middle cerebral artery steno-occlusive disease to Moyamoya disease. Transl Stroke Res 2025;16:1146–1155.
7. Eun MY, Song HN, Choi JU, Cho HH, Kim HJ, Chung JW, et al. Global intracranial arterial tortuosity is associated with intracranial atherosclerotic burden. Sci Rep 2024;14:11318.
8. Lan L, Liu H, Ip V, Soo Y, Abrigo J, Fan F, et al. Regional high wall shear stress associated with stenosis regression in symptomatic intracranial atherosclerotic disease. Stroke 2020;51:3064–3073.
9. Woo HG, Kim HG, Lee KM, Ha SH, Jo H, Heo SH, et al. Wall shear stress associated with stroke occurrence and mechanisms in middle cerebral artery atherosclerosis. J Stroke 2023;25:132–140.
10. Ihara M, Yamamoto Y, Hattori Y, Liu W, Kobayashi H, Ishiyama H, et al. Moyamoya disease: diagnosis and interventions. Lancet Neurol 2022;21:747–758.
11. Dorschel KB, Wanebo JE. Physiological and pathophysiological mechanisms of the molecular and cellular biology of angiogenesis and inflammation in moyamoya angiopathy and related vascular diseases. Front Neurol 2023;14:661611.
12. Kliś KM, Krzyżewski RM, Kwinta BM, Łasocha B, Brzegowy P, Stachura K, et al. Increased tortuosity of basilar artery might be associated with higher risk of aneurysm development. Eur Radiol 2020;30:5625–5632.
13. Nagargoje MS, Asif S, Nekkanti RK, Krishnakumar K, Kannath SK, Sudhir BJ, et al. Influence of carotid tortuosity on the hemodynamics in cerebral aneurysms. Phys Fluids 2023;35:061908.
14. Wang Y, Chen B, Song L, Li Y, Xu M, Huang T, et al. Effect of siphon morphology on the risk of C7 segment aneurysm formation: a case-control CFD study. Clin Neuroradiol 2024;34:485–494.
15. Roy V, Ross JP, Pépin R, Cortez Ghio S, Brodeur A, Touzel Deschênes L, et al. Moyamoya disease susceptibility gene RNF213 regulates endothelial barrier function. Stroke 2022;53:1263–1275.
16. Zhang L, Rashad S, Zhou Y, Niizuma K, Tominaga T. RNF213 loss of function reshapes vascular transcriptome and spliceosome leading to disrupted angiogenesis and aggravated vascular inflammatory responses. J Cereb Blood Flow Metab 2022;42:2107–2122.
17. Matsuo M, Nadanaka S, Soga M, Sugiyama T, Serigano S, Shimano K, et al. Vulnerability to shear stress caused by altered peri-endothelial matrix is a key feature of Moyamoya disease. Sci Rep 2021;11:1552.
18. Cho HH, Kim J, Na I, Song HN, Choi JU, Baek IY, et al. Predicting cerebrovascular age and its clinical relevance: modeling using 3D morphological features of brain vessels. Heliyon 2024;10e32375.
19. Adams R, Bischof L. Seeded region growing. IEEE Trans Pattern Anal Mach Intell 1994;16:641–647.
20. Heiden W, Goetze T, Brickmann J. Fast generation of molecular surfaces from 3D data fields with an enhanced “marching cube” algorithm. J Comput Chem 1993;14:246–250.
21. Grélard F, Baldacci F, Vialard A, Domenger JP. New methods for the geometrical analysis of tubular organs. Med Image Anal 2017;42:89–101.
22. Hong SW, Song HN, Choi JU, Cho HH, Baek IY, Lee JE, et al. Automated in-depth cerebral arterial labelling using cerebrovascular vasculature reframing and deep neural networks. Sci Rep 2023;13:3255.
23. Jung Y, Kim B, Kim CK, Won HH, Chae SH, Oh K, et al. Long-chain polysaturated fatty acid in atrial fibrillation-associated stroke: lipidomic-GWAS study. Thromb Haemost 2025;125:1127–1140.
24. McCarthy S, Das S, Kretzschmar W, Delaneau O, Wood AR, Teumer A, Kang HM, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat Genet 2016;48:1279–1283.
25. Deshpande A, Zhang LQ, Balu R, Yahyavi-Firouz-Abadi N, Badjatia N, Laksari K, et al. Cerebrovascular morphology: insights into normal variations, aging effects and disease implications. J Cereb Blood Flow Metab 2025;45:1249–1264.
26. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 2006;38:904–909.
27. Kuroda S, Fujimura M, Takahashi J, Kataoka H, Ogasawara K, Iwama T, et al. Diagnostic criteria for Moyamoya disease - 2021 revised version. Neurol Med Chir (Tokyo) 2022;62:307–312.
28. Kamimura T, Okazaki S, Morimoto T, Kobayashi H, Harada K, Tomita T, et al. Prevalence of RNF213 p.R4810K variant in early-onset stroke with intracranial arterial stenosis. Stroke 2019;50:1561–1563.
29. Kim HJ, Choi EH, Chung JW, Kim JH, Kim YS, Seo WK, et al. Role of the RNF213 variant in vascular outcomes in patients with intracranial atherosclerosis. J Am Heart Assoc 2021;10e017660.
30. Bang OY, Ryoo S, Kim SJ, Yoon CH, Cha J, Yeon JY, et al. Adult Moyamoya disease: a burden of intracranial stenosis in East Asians? PLoS One 2015;10e0130663.
31. Research Committee on the Pathology and Treatment of Spontaneous Occlusion of the Circle of Willis, ; Health Labour Sciences Research Grant for Research on Measures for Infractable Diseases. Guidelines for diagnosis and treatment of moyamoya disease (spontaneous occlusion of the circle of Willis). Neurol Med Chir (Tokyo) 2012;52:245–266.
32. Jeong DY, Kim BJ, Bae JH, Kim C, Kwon SU. Impact of RNF213 polymorphism in isolated intracranial arterial steno-occlusive disease. J Clin Neurol 2025;21:173–181.
33. Scott RM, Smith ER. Moyamoya disease and Moyamoya syndrome. N Engl J Med 2009;360:1226–1237.
34. Kliś KM, Wójtowicz D, Kwinta BM, Stachura K, Popiela TJ, Frączek MJ, et al. Association of arterial tortuosity with hemodynamic parameters-a computational fluid dynamics study. World Neurosurg 2023;180:e69–e76.
35. An S, Kim T, Oh CW, Bang JS, Lee SU, Heo J. Vascular tortuosity of the internal carotid artery is related to the RNF213 c.14429G > A variant in Moyamoya disease. Sci Rep 2019;9:8614.
36. Liu W, Morito D, Takashima S, Mineharu Y, Kobayashi H, Hitomi T, et al. Identification of RNF213 as a susceptibility gene for moyamoya disease and its possible role in vascular development. PLoS One 2011;6e22542.
37. Kohn JC, Lampi MC, Reinhart-King CA. Age-related vascular stiffening: causes and consequences. Front Genet 2015;6:112.
38. Kamada F, Aoki Y, Narisawa A, Abe Y, Komatsuzaki S, Kikuchi A, et al. A genome-wide association study identifies RNF213 as the first Moyamoya disease gene. J Hum Genet 2011;56:34–40.
39. Kim EH, Yum MS, Ra YS, Park JB, Ahn JS, Kim GH, et al. Importance of RNF213 polymorphism on clinical features and long-term outcome in moyamoya disease. J Neurosurg 2016;124:1221–1227.
40. Miyatake S, Miyake N, Touho H, Nishimura-Tadaki A, Kondo Y, Okada I, et al. Homozygous c.14576G>A variant of RNF213 predicts early-onset and severe form of Moyamoya disease. Neurology 2012;78:803–810.
41. Wang X, Wang Y, Nie F, Li Q, Zhang K, Liu M, et al. Association of genetic variants with Moyamoya disease in 13 000 individuals: a meta-analysis. Stroke 2020;51:1647–1655.
42. Wu Z, Jiang H, Zhang L, Xu X, Zhang X, Kang Z, et al. Molecular analysis of RNF213 gene for Moyamoya disease in the Chinese Han population. PLoS One 2012;7e48179.
43. Zhang Q, Liu Y, Zhang D, Wang R, Zhang Y, Wang S, et al. RNF213 as the major susceptibility gene for Chinese patients with Moyamoya disease and its clinical relevance. J Neurosurg 2017;126:1106–1113.
44. Bang OY, Chung JW, Cha J, Lee MJ, Yeon JY, Ki CS, et al. A polymorphism in RNF213 is a susceptibility gene for intracranial atherosclerosis. PLoS One 2016;11e0156607.
45. Miyawaki S, Imai H, Takayanagi S, Mukasa A, Nakatomi H, Saito N. Identification of a genetic variant common to moyamoya disease and intracranial major artery stenosis/occlusion. Stroke 2012;43:3371–3374.
46. Jang MA, Shin S, Yoon JH, Ki CS. Frequency of the moyamoya-related RNF213 p.Arg4810Lys variant in 1,516 Korean individuals. BMC Med Genet 2015;16:109.
47. Liu W, Hitomi T, Kobayashi H, Harada KH, Koizumi A. Distribution of Moyamoya disease susceptibility polymorphism p.R4810K in RNF213 in East and Southeast Asian populations. Neurol Med Chir (Tokyo) 2012;52:299–303.

Article information Continued

Figure 1.

Manhattan plots of genome-wide association results for intracranial vessel features in the SMC cohort. Results for the GTIb (A), ICA area (B), and MMRb (C) are shown. The red horizontal line indicates the genome-wide significance threshold (P=5.0×10-8), and the blue line indicates the suggestive threshold (P=1.0×10-5). The genome-wide significant locus on chromosome 17 is highlighted in green, and the most significant single-nucleotide polymorphism, encoding RNF213 p.R4810K, is annotated. SMC, Samsung Medical Center; GTIb, basal global tortuosity index; ICA, internal carotid artery; MMRb, basal minimum-to-maximum diameter ratio.

Figure 2.

Group-wise comparison of intracranial vessel features across the RNF213–MMD spectrum in the SMC cohort. Box-and-jitter plots show the GTIb (A), ICA area (B), MMRb (C), and age (D) across three predefined groups: MMD-/variant-, MMD-/variant+, and MMD+/variant+. Six patients in the MMD+/variant- group were not included in the comparison because of the small size of this subgroup. Boxes indicate the IQR with the median as a horizontal line; whiskers extend to 1.5×IQR. Kruskal–Wallis tests were followed by Bonferroni-adjusted pairwise Wilcoxon tests. n.s., not significant; GTIb, basal global tortuosity index; MMD, moyamoya disease; ICA, internal carotid artery; MMRb, basal minimum-to-maximum diameter ratio; SMC, Samsung Medical Center; IQR, interquartile range. *P<0.05; **P<0.01; ***P<0.001.

Figure 3.

ROC curves for discrimination of RNF213 variant carriers in the non-MMD group. (A) Unadjusted ROC curves for ICA area (dark blue), GTIb (light blue), and the combined logistic regression model incorporating both ICA area and GTIb (green). (B) Adjusted ROC curves based on multivariable logistic regression models including clinical covariates (age, sex, and stroke etiology [large artery atherosclerosis vs. non-large artery atherosclerosis]). The models depict the clinical model alone (black), clinical model plus ICA area (blue), and clinical model plus ICA area and GTIb (green). ICA, internal carotid artery; GTIb, basal global tortuosity index; AUC, area under the curve; ROC, receiver operating characteristic; MMD, moyamoya disease.

Table 1.

Descriptive statistics of clinical variables of the SMC and KUMC cohorts

Variable SMC (n=945) KUMC (n=627)
Age (yr) 67.1±13.4 66.0±12.7
Female sex 358 (37.88) 241 (38.44)
Smoking status
 Never smoker 545 (57.67) 370 (59.01)
 Ex-smoker 215 (22.75) 88 (14.04)
 Current smoker 185 (19.58) 169 (26.95)
Alcohol consumption, bottle per week 0.9±2.4 Not provided
Hypertension 679 (71.85) 377 (60.13)
Diabetes mellitus 307 (32.49) 175 (27.91)
Dyslipidemia 585 (61.90) 41 (6.54)
Previous stroke or TIA 222 (23.49) 83 (13.24)
Congestive heart failure 35 (3.70) 8 (1.28)
Peripheral arterial occlusive disease 16 (1.69) Not provided
Coronary artery disease 112 (11.85) 61 (9.73)
Initial NIHSS 4.2±4.8 3.7±4.5
Previous mRS 0 (0–0) 0 (0–0)
mRS at 3 months 1 (0–2) 1 (0–2)
TOAST classification
 Large artery atherosclerosis 257 (27.20) 191 (30.46)
 Cardioembolic 199 (21.06) 159 (25.36)
 Small artery occlusion 210 (22.22) 143 (22.81)
 Other 107 (11.32) 92 (14.67)
 Undetermined 172 (18.20) 42 (6.70)

Continuous variables are presented as mean±standard deviation, and categorical variables as frequency (percentage). mRS values are presented as median (interquartile range). For characteristics with missing values, the number of subjects is indicated. Five patients in the KUMC cohort (total enrolled n=632) did not provide baseline characteristics and were therefore excluded from the table.

SMC, Samsung Medical Center; KUMC, Korea University Medical Center; TIA, transient ischemic attack; NIHSS, National Institutes of Health Stroke Scale; mRS, modified Rankin scale; TOAST, Trial of ORG 10172 in Acute Stroke Treatment.

Table 2.

Multivariable regression models for log-transformed ICA area

Overall model (n=943) Non-MMD model (n=923) Non-MMD (non-LAA) (n=675) Non-MMD (LAA) (n=248) Interaction model (n=923) ICAS-available model (n=647) ICAS-adjusted model (n=647)
Intercept 2.960*** (2.885, 3.035) 2.975*** (2.900, 3.050) 2.945*** (2.864, 3.027) 2.993*** (2.809, 3.177) 2.974*** (2.900, 3.049) 3.023*** (2.943, 3.103) 3.024*** (2.944, 3.103)
RNF213 variant+ -0.264*** (-0.324, -0.204) -0.203*** (-0.271, -0.134) -0.181*** (-0.262, -0.101) -0.254*** (-0.384, -0.123) -0.183*** (-0.266, -0.100) -0.219*** (-0.293, -0.146) -0.209*** (-0.282, -0.136)
LAA (vs. non-LAA) -0.098*** (-0.128, -0.067) -0.098*** (-0.128, -0.067) -0.095*** (-0.126, -0.064) -0.091*** (-0.122, -0.059) -0.077*** (-0.110, -0.045)
Age 0.001** (0.000, 0.002) 0.001* (0.000, 0.002) 0.002** (0.001, 0.003) -0.000 (-0.003, 0.002) 0.001* (0.000, 0.002) 0.001 (0.000, 0.002) 0.001* (0.000, 0.002)
Sex 0.057*** (0.029, 0.085) 0.052*** (0.024, 0.079) 0.058*** (0.026, 0.089) 0.031 (-0.028, 0.090) 0.051*** (0.023, 0.079) 0.042** (0.012, 0.071) 0.039** (0.010, 0.068)
RNF213×LAA -0.061 (-0.206, 0.085)
ICAS burden (0–7) -0.020*** (-0.032, -0.009)

Regression coefficients (β) with 95% confidence intervals are shown. ICA area was analyzed after log transformation. RNF213 variant status was modeled as variant+ versus variant-. LAA was defined by TOAST classification. The interaction model included the RNF213 variant×LAA term. The ICAS-available model was restricted to non-MMD patients with ICAS burden data, but did not include ICAS burden as a covariate. The ICAS-adjusted model additionally included intracranial atherosclerotic burden (0–7). Sex was included as a categorical covariate.

ICA, internal carotid artery; MMD, moyamoya disease; LAA, large artery atherosclerosis; ICAS, intracranial atherosclerosis; TOAST, Trial of ORG 10172 in Acute Stroke Treatment.

*

P<0.05;

**

P<0.01;

***

P<0.001.

Table 3.

Discriminative performance of models for identifying RNF213 variant carriers in the non-MMD group

Model AUC (95% CI) ΔAUC P (DeLong)
Clinical model 0.630 (0.544, 0.716)
Clinical+ICA area 0.783 (0.706, 0.860) +0.153 <0.001***
Clinical+ICA area+GTIb 0.813 (0.746, 0.880) +0.030 0.087

The clinical model included age, sex, and stroke etiology (large artery atherosclerosis vs. non-large artery atherosclerosis). ΔAUC values represent the incremental change in the area under the curve compared with the preceding model. P-values were calculated using DeLong’s test for correlated receiver operating characteristic curves.

MMD, moyamoya disease; CI, confidence interval; AUC, area under the curve; ICA, internal carotid artery; GTIb, basal global tortuosity index.

*

P<0.05;

**

P<0.01;

***

P<0.001.