We designed a genetic association study by consecutively recruiting patients with first-ever acute ischemic cerebrovascular events. All participants were within 48 h of event onset and had a prior diagnosis of acute ischemic stroke or transient ischemic attack (TIA) upon admission to The First Affiliated Hospital of Chongqing Medical University from June 2019 to November 2020. Cerebral hemorrhage was excluded by computerized tomography scans. All participants were recruited from the Chinese Han population and were over 18 years of age. The exclusion criteria included: (1) a previous history of acute ischemic cerebrovascular accident; (2) a history of malignant tumor; (3) a history of autoimmune disease or chronic inflammatory disease; and (4) a diagnosis of cardioembolic stroke, stroke of other determined etiology, or a stroke of undetermined etiology based on TOAST classification (21).

The diagnosis was further reviewed after discharge. Participants with a final diagnosis of ischemic stroke or TIA and lacking the exclusion criteria listed above were included in subsequent evaluations. This study was approved by the Ethics Committee of The First Affiliated Hospital of Chongqing Medical University, and written informed consent was obtained from each recruited patient.

After recruitment, we preliminarily enrolled 231 patients into this study. Under review, 19 patients were excluded in accordance with the exclusion criteria, and 11 with incomplete clinical data were excluded. The remaining 201 patients were included in the final analysis.

To evaluate intracranial and extracranial vascular AS burden, computerized tomography angiography (CTA) of the head and neck was conducted with a multi-slice scanner (GE Discovery CT750HD; GE Healthcare, Chicago, IL). The scanning parameters were 100 kV, 120 mAs, 5-mm section thickness, and intravenous administration of 80 mL non-ionic contrast (rate, 5 mL/s). Reconstructed time-of-flight magnetic resonance angiography was selected if CTA was contraindicated. The extent of cerebrovascular AS-related stenosis was examined in the intracranial artery (bilateral anterior, middle, and posterior cerebral arteries; basilar artery, and intracranial portions of the internal carotid and vertebral arteries) and extracranial artery (extracranial portions of the internal carotid and vertebral arteries). Two experienced neurologists who were blinded to the clinical information independently assessed all images. Final decisions were reached by a consensus. Quantifying CAB was determined in accordance with the extent of the stenosis, as previously described (22). Stenosis scores were recorded as 0 for no stenosis, 1 for <50% stenosis, 2 for 50–74% stenosis, 3 for >75% stenosis, or 4 for occlusion. The sum of cerebral AS scores was calculated and categorized into quartiles, and advanced CAB was defined to include patients exhibiting the highest quartile range of the AS scores.

Pro-atherogenic risk factors, which included total cholesterol, total triglycerides, low-density lipoprotein cholesterol (LDL-c), high-sensitivity C-reactive protein (hs-CRP), uric acid (UA), white blood cell count, and platelet count, were included. The estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease formula [175 × (serum creatinine)^−1.154 × (Age)^−0.203 × (0.742 if female)]. An hs-CRP level > 3.0 mg/L was defined as a high vascular risk. Histories of hypertension, diabetes mellitus (DM), and smoking status were obtained for further multivariable regression analysis.

We chose 3,000-bp upstream of the RGMa transcriptional start site as the promoter sequence region. FuncPred (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html) was used to predict functional SNPs, and loci with altered transcription factor binding were identified (prediction results are shown in Supplementary Table 1) (23). From these results, we chose loci with minor allele frequencies > 0.05 in the Southern Chinese Han population from the 1,000 Genomes database (http://internationalgenome.org), and referred to previous related studies (13). Pairwise linkage disequilibrium was determined using the Linkage Disequilibrium Calculator to identify complete linkage disequilibrium in the specific population of the 1,000 Genome database in Ensembl (http://ensembl.org) (results are shown in Supplementary Table 2) (24). Three SNPs [rs4778099(G>A), rs10520720(G>A), and rs725458(C>T)] were selected for further analysis. Annotated SNP information was obtained from the dbSNP database (https://www.ncbi.nlm.nih.gov/snp/). Notably, the frequency of the minor allele T in rs725458 (66%), as annotated in the dbSNP, was higher than that of the major allele C (34%) in the Chinese Han population (1,000 Genome database). Tissue-specific expression quantitative trait loci (eQTL) information was obtained from the GTEx portal (http://www.gtexportal.org) and Blood eQTL Browser (https://genenetwork.nl/bloodeqtlbrowser/) (25, 26).

Samples (5 ml) of EDTA-anticoagulated blood were collected from all recruits, and whole blood genomic DNA was isolated using a blood DNA isolation kit (TIANGEN, Beijing, China). Kompetitive allele-specific PCR (KASP) technology was used to genotype the selected SNPs. KASP master mix and primers were supplied by LGC Bioresearch Technology (Middlesex, UK), and KASP reactions were run on a compatible real-time PCR instrument (CFX Connect; Bio-Rad, Hercules, CA). Five percent of all genomic DNA samples, which were randomly selected for quality control, were Sanger sequenced.

AS-related microarray dataset searches were conducted manually on NCBI's GEO database. Datasets containing artery samples or peripheral blood samples in independent cohorts were collected. For the RGMa differential expression analysis, the collected datasets with matched normal controls were selected (GSE7074, GSE23746, GSE20129, GSE57691). The RGMa expression array data in the AS group and normal control group were used in this analysis. Data in GSE7074 and GSE23746 datasets were obtained from independent carotid AS cohorts and their matched controls. Data in GSE20129 were obtained from an independent coronary artery AS cohort and matched controls. Data in GSE57691 were obtained from an independent aorta AS cohort and matched controls. For the AS plaque RGMa-related correlation analysis, only the datasets where the size of the target sample was >30 were selected (GSE125771, GSE24495), and the expression array data for RGMa, platelet-derived growth factor subunit B(PDGFB), actin alpha 2(ACTA2), and phosphatase and tensin homolog (PTEN) were used for further analysis. GSE125771 contained 40 independent AS carotid samples. GSE24495 contained 113 independent AS carotid samples.

SPSS (v26.0; IBM, Armonk, NY) was used for statistical analysis. We expressed continuous variables as the mean ± standard deviation (SD), and categorical values as frequencies (percentages). Student's t-tests and chi-square tests were performed for continuous values and categorical values, respectively. Hardy-Weinberg equilibrium analysis and the chi-square tests were performed for each locus. To evaluate the contribution of genotypes to CAB, multivariable logistics regression analysis was adjusted for common vascular risks (Model 1: age, sex, hypertension, DM, and smoking; Model 2: Model 1 plus eGFR, UA, LDL-c, and hs-CRP level). Haplotype analysis (rs4778099-rs10520720-rs725458) was performed using SNPstats (http://snpstats.net) (27). Our analysis of the association between haplotype and CAB was adjusted for age, sex, hypertension, smoking, eGFR, and hs-CRP level. GEO data correlation analysis was performed using Spearman's rank test. P-values of <0.05 were considered statistically significant. All tests were two-tailed.

All enrolled patients were grouped according to their calculated CAB, and their baseline clinical data are shown in Table 1. Of the 201 enrolled patients, 55 were in the advanced CAB group and 146 were not considered to have advanced CAB. No significant differences among baseline clinical characteristics were found between the two groups, with the exception of the hs-CRP level (>3.0 mg/L), a well-known clinical marker for cerebral AS (28). The advanced CAB group contained a significantly higher proportion of patients with abnormal hs-CRP levels [Advanced CAB+: 29 (52.7%) vs. Advanced CAB–: 36 (24.7%), p < 0.001].

Table 2 shows the allele and genotype frequencies of the target polymorphisms from our research. We also compared the genetic frequency between the enrolled patients and the Han population in Southern China (CHS), a population of age and sex matched healthy subjects published by the 1,000 Genome project. We found no significant difference in the target polymorphisms' allele and genotype frequency.

Using the public GEO database, we analyzed the RGMa expression patterns in various tissues and cells between AS and non-AS states. Data from GSE20129 and GSE23746 indicated no significant difference in the RGMa levels in peripheral blood or peripheral monocytes between patients with AS and non-AS controls. Nevertheless, RGMa levels in atherosclerotic aortas or in macrophages isolated from plaques were much lower than those in normal aortas or macrophages from normal tissue (Figure 1). We also investigated whether RGMa levels were correlated with AS-related molecule expression. We found a negative correlation between RGMa levels and PDGFB, a crucial pro-atherogenic gene (Figure 2). RGMa levels were also positively correlated with the transcription levels of ACTA2 and PTEN, which are regarded as markers of normal quiescent vascular smooth muscle cells and vascular protective factor, respectively (Figures 3, 4) (29, 30).

We investigated whether associations between target SNPs and CAB were present in the patients. Table 3 shows the frequency of all SNPs in each group. Compliance with Hardy-Weinberg equilibrium (p > 0.05) was tested. Logistic regression models were used in accordance with additive, recessive and dominant inheritance models in this analysis. We found the rs725458 T allele (in the additive or dominant models) was independently associated with increased CAB after adjusting for common vascular risk factors (adjusted OR = 1.66, P = 0.046; adjusted OR = 7.18, P = 0.010, respectively). The rs4778099 AA genotype was inversely associated with CAB in the recessive model (adjusted OR = 0.10, P = 0.027). rs10520720 was not significantly associated with CAB in this analysis.

To further substantiate our results, we queried the GEO database to verify the association between our target SNPs and atherosclerosis from publicly available SNP-array data in related studies. Data in GSE90073 were collected from an independent cohort of 106 American patients (African ancestry, 24; European ancestry, 82) who were undergoing clinically indicated diagnostic cardiac catheterization (Supplementary Table 3). Coronary artery atherosclerosis burden was quantified by an ordinal coronary artery disease score (range 0–4) according to the angiographic stenosis distribution (31). The patient's genome-wide SNP identification was obtained via an SNP array. Complete genotyping data and atherosclerosis quantification data were in public access. In this analysis, we found a similar genotype-atherosclerosis burden distribution (high atherosclerosis burden is defined as coronary artery disease score equal to 3 or 4) with our study (Figure 5). We also tested the association in the logistic regression model. It supported the rs725458 T allele (in the additive or dominant models) was associated with high atherosclerosis burden (OR = 2.14, P = 0.021; OR = 2.49, P = 0.029, respectively) and the rs4778099 A allele was inversely associated with atherosclerosis burden in the additive model (OR = 0.53, P = 0.032). These findings suggested the consistent effect of our target SNPs on atherosclerosis across ancestries.

The estimated haplotype distribution from our analysis of the selected SNPs is shown in Table 4. GGT was the most frequent haplotype in both groups (55.7 and 43.5%, respectively). In the haplotype association analysis, no haplotype exhibited a significant association with CAB after adjustment.