Investigation of the causal effect and molecular signatures of serum hydroxyvitamin D on atherosclerosis: a Mendelian randomization and transcriptomic approach

Abstract

Arterial atherosclerosis is a common cardiovascular disease with serious health impact. Growing evidence suggests that serum hydroxyvitamin D may influence its progression. Investigating the causal relationship and underlying mechanisms between vitamin D and atherosclerosis is essential for developing effective prevention and treatment strategies. We conducted Mendelian randomization analysis with stringent single-nucleotide polymorphism selection to assess causal relationships. We analyzed transcriptomic datasets to identify differentially expressed genes (DEGs) and genes related to vitamin D metabolism. Feature selection was performed using Least Absolute Shrinkage and Selection Operator regression, Support Vector Machine-Recursive Feature Elimination, and random forest algorithms. Receiver operating characteristic (ROC) curves were used to evaluate diagnostic performance, and a nomogram was constructed for predictive modeling. Immune cell infiltration was assessed via CIBERSORT, while Gene Set Enrichment Analysis (GSEA) was employed to explore key pathways. Quantitative real-time PCR validated gene expression. Mendelian randomization analysis confirmed that 25-hydroxyvitamin D acted a protective factor against coronary atherosclerosis. We identified 1,195 DEGs and 33 vitamin D-related genes, with four key genes demonstrating strong diagnostic accuracy in ROC curve analysis. Immune profiling revealed significant differences in nine immune cell types, and GSEA highlighted critical biological pathways involved in disease progression. The nomogram model showed high predictive performance. This multi-omics study establishes a causal relationship between serum hydroxyvitamin D levels and atherosclerosis while uncovering potential biomarkers and pathogenic mechanisms through analyses of gene expression, immune infiltration, and signaling pathways. These findings provide valuable insights into atherosclerosis research and may help guide future therapeutic strategies.

INTRODUCTION

Atherosclerosis is a major etiological factor in cardiovascular diseases, contributing to high incidence and mortality rates, which constitute a significant global health burden and place considerable strain on healthcare systems [1]. Cardiovascular diseases cause over 17 million deaths annually, accounting for 31% of global mortality. Among these, coronary heart disease causes approximately 74 million deaths, while stroke accounts for 6.7 million [2]. Moreover, over 75% of cardiovascular disease-related fatalities occur in low- and middle-income countries [3].

Serum hydroxyvitamin D plays a crucial role in various physiological processes in the humans. After transported to the liver, vitamin D is hydroxylated to 25-hydroxyvitamin D by CYP2R1, an essential step for maintaining calcium homeostasis [4]. Moreover, vitamin D regulates macrophage functions at multiple levels, influencing phenotypic polarization, anti-inflammatory activities, and broader modulation of the immune system [5,6]. Previous studies have demonstrated a strong association between vitamin D deficiency and various diseases [7]. In the context of cardiovascular health, the potential link between vitamin D and atherosclerosis has garnered considerable attention [8]. However, the exact causal relationship and the complex pathogenesis between vitamin D and atherosclerosis remain unclear and warrant further investigation.

This study addresses a critical scientific problem through a comprehensive multi-omics approach. The study conducts genetic analysis to identify key genes and signaling pathways, investigates the roles of immune cell roles and their interactions with genes, and applies Mendelian randomization (MR) to control for confounding factors, thereby ensuring accurate causal inference. The research aims to fill existing knowledge gaps by offering novel insights into early diagnosis, personalized treatment, and prognosis evaluation of atherosclerosis. Additionally, the study provides potential biomarkers for cardiovascular disease prevention, contributing to improved patient health outcomes.

METHODS

Data source

The GSE43292 and GSE100927 microarray datasets related to arterial atherosclerosis were obtained from the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/). The GSE43292 dataset comprised 32 normal samples and 32 case samples with GPL6244, while the GSE100927 dataset included 35 normal samples and 69 case samples with GPL17077. GSE100927 was used as the validation set. Vitamin D metabolism-related genes (VMGs) were identified from the Molecular Signatures Database (MsigDB; https://www.gsea-msigdb.org/), with the specific pathways outlined in Table 1. A total of 269 genes involved in various vitamin D metabolic pathways were retrieved.

Table 1.

Pathways involved in vitamin D metabolism and corresponding genes

Name	Genes	Collections	Contributor
GOBP_NEGATIVE_REGULATION_OF_VITAMIN_D_BIOSYNTHETIC_PROCESS	5	C5 GO	Gene Ontology Consortium
GOBP_POSITIVE_REGULATION_OF_VITAMIN_D_RECEPTOR_SIGNALING_PATHWAY	6	C5 GO	Gene Ontology Consortium
GOBP_REGULATION_OF_VITAMIN_D_BIOSYNTHETIC_PROCESS	7	C5 GO	Gene Ontology Consortium
GOBP_REGULATION_OF_VITAMIN_D_RECEPTOR_SIGNALING_PATHWAY	33	C5 GO	Gene Ontology Consortium
GOBP_VITAMIN_D3_METABOLIC_PROCESS	7	C5 GO	Gene Ontology Consortium
GOBP_VITAMIN_D_BIOSYNTHETIC_PROCESS	33	C5 GO	Gene Ontology Consortium
GOBP_VITAMIN_D_METABOLIC_PROCESS	21	C5 GO	Gene Ontology Consortium
GOBP_VITAMIN_D_RECEPTOR_SIGNALING_PATHWA	12	C5 GO	Gene Ontology Consortium
GOMF_VITAMIN_D_BINDING	7	C5 GO	Gene Ontology Consortium
GOMF_NUCLEAR_VITAMIN_D_RECEPTOR_BINDING	15	C5 GO	Gene Ontology Consortium
HP_ABNORMALITY_OF_VITAMIN_D_METABOLISM	50	C5	The Jackson Laboratory (JAX)
HP_DECREASED_CIRCULATING_VITAMIN_D_CONCENTRATION	34	C5	The Jackson Laboratory (JAX)
REACTOME_VITAMIN_D_CALCIFEROL_METABOLISM	12	C2 CP	Reactome
WP_VITAMIN_D_RECEPTOR_PATHWAY	185	C2 CP	WikiPathways
WP_VITAMIN_D_METABOLISM	33	C2 CP	WikiPathways

MR

The serum hydroxyvitamin D data were obtained from the MRC IEU OpenGWAS dataset (GWAS ID: ieu-b-4808), which includes 441,291 samples. The selection of 25-hydroxyvitamin D as the exposure variable in this study was primarily based on its role as the primary circulating reservoir form of vitamin D. Its blood concentration was stable with a long half-life, enabling it to accurately reflect an individual's overall vitamin D status [9]. In contrast, although 1,25-dihydroxyvitamin D was the active form of vitamin D, its serum concentration was influenced by multiple metabolic regulatory factors, exhibited rapid dynamic changes, and was difficult to measure [10,11]. Consequently, it had poor representativeness and stability in large-scale population studies and was therefore not selected as the primary exposure indicator in this research. The atherosclerosis outcome data were also sourced from the MRC IEU OpenGWAS dataset, with further details provided in Table 2.

Table 2.

Data sources for Mendelian randomization analysis

Phenotype	Year	ID	Population	Control	Case
25 hydroxyvitamin D level	2020	ieu-b-4808	European	/	/
Atherosclerosis, excluding cerebral, coronary and PAD (no controls excluded)	2021	finn-b-I9_ATHSCLE_EXNONE	European	6,599	212,193
Atherosclerosis, excluding cerebral, coronary and PAD	2021	finn-b-I9_ATHSCLE	European	6,599	206,541
Cerebral atherosclerosis (I9_CERATHER)	2021	finn-b-I9_CERATHER	European	334	203,068
Cerebral atherosclerosis (I9_CEREBATHER)	2021	finn-b-I9_CEREBATHER	European	334	218,688
Cerebral atherosclerosis (no controls excluded)	2021	finn-b-I9_CERATHER_EXNONE	European	334	218,688
Coronary atherosclerosis	2021	finn-b-I9_CORATHER	European	23,363	187,840
Coronary atherosclerosis (no controls excluded)	2021	finn-b-I9_CORATHER_EXNONE	European	23,363	195,429
Peripheral atherosclerosis	2021	finn-b-DM_PERIPHATHERO	European	162,201	2,631

The selection criteria for instrumental variables were as follows:

Single-nucleotide polymorphism (SNP) selection: (1) SNPs were selected as instrumental variables for the MR analysis based on a genome-wide significance threshold of p < 5e-08. For the reverse MR analysis, with atherosclerosis as the exposure, a less stringent threshold of p < 1e-05 was applied to increase the number of instrumental variables. (2) Clumping: SNPs underwent clumping to minimize the impact of linkage disequilibrium (r²= 0.001, region length = 33,000 kb). (3) F-statistic: the F-statistic was used to assess instrument strength. SNPs with an F-statistic exceeding 10 were retained to exclude weak instruments. (4) Harmonization: SNPs were harmonized regarding allele direction, and palindromic SNPs or those with incompatible orientations were removed. (5) Outlier Test: outlier tests were performed using MR-PRESSO, and MR analysis was conducted after excluding the identified outliers.

Cochran's Q test was employed to assess SNP heterogeneity. A p-value < 0.05 indicated the presence of heterogeneity, in which case the random-effects inverse variance weighted (IVW) model was applied. MR-Egger regression was used to evaluate horizontal pleiotropy of instrumental variables. A statistically significant MR-Egger intercept (p < 0.05) suggested the presence of horizontal pleiotropy. To assess the influence of individual SNPs on the causal association, a "Leave-one-out" sensitivity analysis was conducted. This method sequentially excluded each SNP to ensure that the MR results were not disproportionately affected by any single SNP.

Procedure of gene differential analysis

Gene differential analysis of the GSE43292 dataset was performed using the "limma" package (|log FC| > 0.5, adj.p.value < 0.05). The differentially expressed genes (DEGs) were then intersected with vitamin metabolism-related genes to identify differential vitamin D metabolism-related genes (DVMGs). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted on the DVMGs. A protein-protein interaction (PPI) network of the DVMGs (confidence > 0.4) was constructed using STRING (https://string-db.org/) and visualized with Cytoscape (v3.10.2). Additionally, single sample Gene Set Enrichment Analysis (GSEA) scoring of the GSE43292 dataset was performed using the "GSVA" package, based on the vitamin D metabolism gene set.

Strategy for feature gene selection

Firstly, the samples in GSE43292 were divided into a training set and a testing set at a ratio of 7:3. In the training set, feature selection was performed using three machine learning algorithms: Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine-Recursive Feature Elimination (SVM-RFE), and random forest (RF). Specifically, LASSO regression analysis was conducted using the "glmnet" package, with 10-fold cross-validation employed to determine the optimal penalty parameter, effectively eliminating highly correlated genes and reducing model complexity. The "caret" package was used to perform SVM-RFE analysis, and the "randomForest" package was applied for RF analysis. The top 10 most important genes were selected based on the MeanDecreaseGini importance ranking from the RF model. After feature selection, the optimized model was applied to the testing set for prediction, and its performance was evaluated using Precision, Recall, F1 score, and area under the curve (AUC) values.

Receiver operating characteristic (ROC) curve and expression quantitation for validating feature genes

The "rms" package was used to construct a nomogram and calibration curve for based on selected feature genes. To assess the diagnostic performance of both the nomogram and the feature genes, the "pROC" package was employed to generate the ROC curve. Additionally, the "rmda" package was utilized to generate the decision curve analysis (DCA) for both the feature genes and the nomogram. The Wilcoxon test was applied to assess the statistical significance of differences in feature gene expression levels.

Immune analysis approaches

Immune analysis was performed using the "IOBR" package for CIBERSORT analysis, generating bar charts that display immune cell abundance in control and atherosclerotic samples. The immune infiltration levels of the two groups were computed using the CIBERSORT algorithm and visualized in box plots. Additionally, the "ggcorrplot" package was used to create a correlation heatmap illustrating the associations between characteristic genes and the differentially infiltrated immune cells.

Operation of single gene GSEA

GSEA enrichment analysis of characteristic genes was performed using the c2.cp.kegg.v7.5.symbols.gmt reference dataset and the "clusterProfiler" package. The significance thresholds were set at adj.p.value < 0.05 and |Normalized Enrichment Score (NES)| > 1. For each characteristic gene, the top 5 most significantly enriched pathways, ranked by the lowest adj.p.value, were reported.

Cell culture and treatment

Human umbilical vein endothelial cells (HUVEC) cell lines were obtained from Wuhan Pricella Biotechnology Co., Ltd. (Wuhan, China). HUVEC cells were maintained in complete endothelial cell culture medium (ZQ1304, Zhongqiaoxinzhou Biotech), and were incubated at 37°C in a 5% CO2 incubator and passaged upon reaching approximately 80% confluence. Additionally, HUVEC were exposed to 100 μg/ml ox-LDL for 24 h to simulate an in vitro model.

RNA isolation and quantitative RT-PCR assay

Total RNA was isolated from cell lines using TRIzol reagent (Thermo Fisher Scientific). The complementary DNA (cDNA) was synthesized as per the manufacturer’s instructions, utilizing the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). qRT-PCR was performed with SYBR Green PCR kit (Takara Bio) on a StepOne Real-Time PCR system (Thermo Fisher Scientific). The relative gene expression levels were quantified by employing the 2^–△△CT method. The sequences of all primers were listed in Table 3.

Table 3.

Sequences of primers for qRT-PCR

Gene name	Primer sequences (5’-3’)
ADRA1B	Forward: TTGGGCATTGTGGTCGGTAT
	Reverse: TGGAGAACAAGGAGCCAAGC
ID4	Forward: GCTGTCCAGGTGTGCGG
	Reverse: CCTCTCTAGTGCTCCTGGCT
MX2	Forward: TGAACGTGCAGCGAGCTT
	Reverse: GGCTTGTGGGCCTTAGACAT
SEMA4D	Forward: GTGGAAAGAAACCATGTGCTGA
	Reverse: GGGCCTCAGAAGAAATGCTGT

RESULTS

MR: causal link of 25 hydroxyvitamin D level and coronary atherosclerosis

As shown in Fig. 1, results from multiple MR methods were presented, with the IVW analysis as the primary reference. The exposure-related SNPs were provided in Supplementary Table 1. After outlier removal via MR-PRESSO analysis, the corresponding instrumental variables were listed in Supplementary Table 2. For the 25-hydroxyvitamin D level, a significant causal relationship was identified with coronary atherosclerosis (odds ratio [OR] = 0.67; 95% confidence interval [CI], 0.54, 0.83) and coronary atherosclerosis (excluding no controls) (OR = 0.68; 95% CI, 0.55, 0.83), suggesting its protective effect. Detailed MR analysis results for all the involved SNPs were provided in the Supplementary Table 3. The scatter plots (Supplementary Fig. 1A, Supplementary Fig. 2A) and forest plots (Supplementary Fig. 1B, Supplementary Fig. 2B) illustrated the causal effects of SNPs associated with serum hydroxyvitamin D on the risk of coronary atherosclerosis. Additionally, the funnel plots (Supplementary Fig. 1C, Supplementary Fig. 2C) and leave-one-out test results (Supplementary Fig. 1D, Supplementary Fig. 2D) further supported the reliability of the MR findings in this study.

Fig. 1. Analysis results of 25-hydroxyvitamin D level and atherosclerosis in Mendelian randomization.

Five methods for causal inference (inverse variance weighted, MR Egger regression, weighted median, weighted mode, and simple mode), with the inverse variance weighted method as the primary reference. SNP, single-nucleotide polymorphism; OR, odds ratio; CI, confidence interval. OR < 1, protective factor; OR > 1, risk factor.

In the Cochran's Q test for IVW analysis and MR-Egger regression, the p-value of the Q statistic was less than 0.05 (Supplementary Table 4), indicating heterogeneity in the causal relationship between 25-hydroxyvitamin D levels and both coronary atherosclerosis and coronary atherosclerosis without excluding the control group. As a result, the random-effects IVW method was employed for the MR analysis. The random-effects IVW method enhanced the robustness of the causal effect, making it less susceptible to heterogeneity. Importantly, the MR-Egger regression intercept test showed no statistically significant differences, indicating no evidence of horizontal pleiotropy (Supplementary Table 5), with all intercepts being non-significant.

In the reverse MR analysis, the results of the IVW method, as shown in Fig. 2, indicated no causal relationship between 25-hydroxyvitamin D levels and coronary atherosclerosis or coronary atherosclerosis (with no controls excluded). The SNPs associated with the exposure were listed in Supplementary Table 6. After removing outliers through MR-PRESSO analysis, the selected instrumental variables for this analysis were presented in Supplementary Table 7. Due to observed heterogeneity, the random-effects IVW method was applied for the MR analysis (Supplementary Table 8). Additionally, the MR-Egger intercept test revealed no evidence of horizontal pleiotropy, supporting the validity of the causal inference (Supplementary Table 9).

Fig. 2. Mendelian randomization results for 25-hydroxyvitamin D levels in relation to coronary atherosclerosis and coronary atherosclerosis (with no controls excluded).

Five methods for causal inference (inverse variance weighted, MR Egger regression, weighted median, weighted mode, and simple mode), with the inverse variance weighted method as the primary reference. SNP, single-nucleotide polymorphism; OR, odds ratio; CI, confidence interval. OR < 1, protective factor; OR > 1, risk factor.

Gene screening and analysis in vitamin D metabolism

In the GSE43292 dataset, a total of 1,195 DEGs were identified, including 530 down-regulated and 665 up-regulated genes. Fig. 3A presented a volcano plot of these DEGs, while Fig. 3B showed the heatmap of the top 20 DEGs ranked by the magnitude of |log FC|. In Fig. 3C, the GSVA-based scores for case and normal samples were illustrated, with case samples exhibiting significantly higher scores than normal samples. By intersecting the DEGs with genes related to vitamin D metabolism, 33 DVMGs were identified, as shown in Fig. 3D. The PPI network of these 33 DVMGs, consisting of 22 nodes and 46 edges, was depicted in Fig. 3E. Enrichment analyses of these genes revealed significant findings, with KEGG analysis highlighting tuberculosis, acute myeloid leukemia, and pertussis pathways (Fig. 3F). Although these pathways involved different diseases, they were all closely associated with immune and inflammatory responses, reflecting the complex immunological and inflammatory mechanisms underlying atherosclerosis. The GO enrichment analysis yielded 246 significant terms, including 235 biological processes, 6 cellular components, and 5 molecular functions. Notably, these genes were predominantly enriched in biological processes such as leukocyte migration, leukocyte chemotaxis, and cell chemotaxis (Fig. 3G). The above results suggested that while some disease pathways might exhibit non-specific enrichment, the overall enrichment pattern reflected the active involvement of immune and inflammatory pathways in atherosclerosis. Furthermore, Supplementary Fig. 3 presented the expression heatmap and GO enrichment analysis of 33 intersecting genes. Upregulated genes primarily related to immune responses against lipopolysaccharides and bacterial molecules, while downregulated genes were enriched in cardiac function regulation, valve development, and extracellular matrix organization. This indicated that up- and down-regulated genes participated in distinct biological processes: immune defense and cardiac structural maintenance, respectively. The findings above revealed specific pathways for up- and down-regulated genes in biological processes, suggesting a potential regulatory relationship between immune responses and cardiac structure and function.

Fig. 3. Screening and preliminary analysis of genes associated with atherosclerosis diagnosis.

(A) Volcano plot of differential genes. (B) The heatmap shows the differential expression of the top 20 DEGs ranked by log FC. (C) The difference in ssGSEA scores based on VMGs between Case and Normal. (D) The upset plot presents the intersection of the DEGs and VMGs. (E) Construction of the PPI network for DVMGs. (F) KEGG pathway enrichment analysis for the DVMGs. (G) GO enrichment analysis of DVMGs. DEGs, differentially expressed genes; ssGSEA, single-sample gene set enrichment analysis; VMGs, vitamin D metabolism-related genes; PPI, protein-protein interaction; DVMGs, differential vitamin D metabolism-related genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes. ****p < 0.0001.

Multi-algorithm feature gene screening for atherosclerosis

To further screen for characteristic genes, three machine learning algorithms were employed to identify important characteristic genes based on the 33 intersecting genes obtained previously. LASSO analysis was performed using the 'glmnet' package, identified eight characteristic genes (Fig. 4A, 5B). The residual distribution plot of the RF model was presented in Fig. 4C. The error decreases as the number of subtrees increases. Using the MeanDecreaseGini metric, the top 10 genes by feature importance were identified (Fig. 4D). Fig. 4E illustrated the accuracy of the SVM classifier relative to the number of features, and the SVM model using eight features was ultimately selected. The genes selected by the three machine learning methods were listed in Supplementary Table 10. The validation set performance of the three models, measured by precision, recall, F1 score, and AUC, was illustrated in Fig. 4F. The intersection of features obtained from LASSO, SVM-RFE, and RF analyses yielded four key genes: ADRA1B, ID4, MX2, and SEMA4D (Fig. 4G). These four genes were considered the final characteristic genes.

Fig. 4. Identification of potential biomarkers for atherosclerosis diagnosis.

(A) LASSO coefficient regression path. (B) LASSO cross-validation error plot, where the horizontal axis represents the logarithmic values of the regularization parameter λ, and the vertical axis represents the cross-validation error. A smaller error indicates better predictive performance. (C) Residual distribution plot for random forest (RF), with the abscissa representing the number of subtrees and the ordinate representing the error. As the number of subtrees increases, the error gradually decreases. (D) Feature importance plot for RF. (E) Relationship between generalization error and the number of features in SVM-RFE. (F) Precision, Recall, F1 score and AUC value of the three algorithms on the test set. (G) Venn diagram showing the intersection of genes identified by integrating the three algorithms. LASSO, Least Absolute Shrinkage and Selection Operator; SVM-RFE, Support Vector Machine-Recursive Feature Elimination; AUC, area under the curve.

Fig. 5. Establishment and evaluation of the atherosclerosis diagnostic model.

(A) Comprehensive risk assessment nomogram. (B) The findings of the ROC curve analysis for assessing the diagnostic performance of the nomogram. (C) Visualization of the DCA curve results. (D) Calibration curve assessment outcomes. ROC, receiver operating characteristic curve; DCA, decision curve analysis.

ROC curves and expression levels of characteristic genes in training and validation sets

Fig. 6 presented the ROC curves for the training set (GSE43292) and validation set (GSE100927), along with the box plots of the expression levels of the characteristic genes. All AUC values were greater than 0.8. Fig. 6A showed the ROC curve for the training set, where the AUC values for the four characteristic genes exceeding 0.8. Fig. 6B displayed the ROC curve for the validation set, with AUC values for the four characteristic genes above 0.8. Fig. 6C, D presented the box plots of the expression levels of the four characteristic genes in the training set and the validation set, respectively. Notably, ADRA1B and ID4 were significantly downregulated in diseased samples in both datasets, while MX2 and SEMA4D showed higher expression in diseased samples.

Fig. 6. Assessment of diagnostic performance and expression profiles of feature genes.

(A) ROC curve evaluation of feature genes in the training set GSE43292. (B) ROC curve analysis of the diagnostic performance of feature genes in the validation set GSE100927. (C) Expression patterns analysis of feature genes in the training set. (D) Comparative analysis of feature gene expression levels in the validation set. ROC, receiver operating characteristic curve. ****p < 0.0001.

Nomogram construction for atherosclerosis risk prediction

A diagnostic nomogram was developed to predict atherosclerosis risk by integrating the expression levels of the four characteristic genes ADRA1B, ID4, MX2, and SEMA4D. The nomogram presented the expression scores for each gene (Fig. 5A). The nomogram exhibited an AUC value of 0.911, demonstrating superior performance compared to the individual genes (Fig. 5B). To enhance the practicality and effectiveness of the diagnostic model for atherosclerosis, DCA (Fig. 5C) and clinical calibration curves (Fig. 5D) were constructed. Overall, the results strongly supported the high predictive capability of the diagnostic model based on these four characteristic genes.

Immune cell infiltration and gene association in atherosclerosis

Based on the GSE43292 cohort training set, we performed an immune infiltration analysis. The infiltration of immune cells within the immune microenvironment played a crucial role in understanding and managing atherosclerosis. We utilized the CIBERSORT algorithm to assess the infiltration levels of 22 immune cell types in both atherosclerotic and normal samples (Fig. 7A). We then compared the infiltration levels of these immune cell types between the atherosclerotic and control samples. The analysis revealed significant differences in 9 immune cell types: regulatory T cells (Tregs), CD8+ T cells, activated memory CD4+ T cells, plasma cells, activated natural killer (NK) cells, monocytes, M0 macrophages, activated dendritic cells, and naive B cells (Fig. 7B).

Fig. 7. Characterization of the immune landscape in atherosclerosis.

(A) Infiltration levels of 22 immune cell types in normal and atherosclerosis tissues were analyzed using the CIBERSORT algorithm. (B) The box plot presents a comparative analysis of immune cell infiltration levels across different sample groups. (C) Spearman’s correlation analysis was conducted to evaluate the relationship between feature genes and immune cell infiltration. (D) Correlation matrix of 9 significant immune cells, red represents positive correlation and bule represents negative correlation, ns indicates no significant difference, * signifies p < 0.05, ** denotes p < 0.01, and *** means p < 0.001.

Additionally, we explored the correlations between the characteristic genes ADRA1B, ID4, MX2, and SEMA4D and differentially expressed immune cells using Spearman’s correlation analysis. The results revealed that the correlation coefficients between ID4 and activated NK cells, as well as between ID4 and CD8+ T cells, and those between ADRA1B and naive B cells, along with ADRA1B and CD8+ T cells, were all greater than 0.6 (Fig. 7C). Fig. 7D illustrated the correlations among differentially expressed immune cells. CD8+ T cells were significantly and positively correlated with activated NK cells, monocytes, and naive B cells. It should be noted that due to the limited sample size, these results only suggested possible associations between these genes and immune cell subsets. The significant correlations among immune cell subsets might reflect the complexity of immune regulatory networks. These findings provided clues for further research, indicating that these genes might influence immune cell functions and thus contributed to the development of atherosclerosis. However, the specific mechanisms still required experimental validation.

GSEA analysis of feature genes in atherosclerosis pathways

GSEA was performed to identify key pathways associated with the characteristic genes and to elucidate their underlying biological functions (Supplementary Fig. 4). A significant negative correlation was observed between ID4 and the KEGG_LYSOSOME pathway, with a NES of –2.56. In contrast, MX2 showed a significant positive correlation with the KEGG_hematopoietic cell lineage pathway, yielding an NES of 2.51. Similarly, ADRA1B was significantly negatively correlated with the KEGG_LYSOSOME pathway (NES = –2.54), while SEMA4D demonstrated a significant positive correlation with the KEGG_hematopoietic cell lineage pathway (NES = 2.50).

Validation of expression of feature genes that comprised the risk model by RT-qPCR

To validate the expression of the identified characteristic genes, four genes involved in the risk signature were selected for further validation. Our results revealed that the expression of characteristic genes ADRA1B and ID4 was significantly downregulated in atherosclerosis patients, whereas MX2 and SEMA4D were significantly upregulated (Fig. 8). These findings align well with our bioinformatic predictions. Therefore, we hypothesised that aberrant expression of these genes likely contribute to the malignant progression of atherosclerosis.

Fig. 8. Validation of expression of feature genes that comprised the risk model by RT-qPCR.

qRT-PCR analysis of ADRA1B, ID4, MX2, and SEMA4D. ADRA1B and ID4 are downregulated in atherosclerotic patients. Conversely, MX2 and SEMA4D exhibit upregulation. * signifies p < 0.05.

DISCUSSION

Atherosclerosis is a major health risk, and understanding its relationship with serum hydroxyvitamin D is crucial for advancing cardiovascular research. This study integrates multi-omics approaches to elucidate potential mechanisms linking these two factors, providing valuable insights into their interaction.

Vitamin D has been shown to impede macrophage transformation into foam cells within atherosclerotic plaques, thereby slowing atherosclerosis progression. This is achieved through suppression of endoplasmic reticulum stress, downregulation of scavenger receptor expression [12], and inhibition of autophagy through the PTPN6/SHP-1 pathway [13]. Additionally, vitamin D modulates the immune system by reducing the overactivation of inflammatory cells [14], decreasing the release of inflammatory mediators, and alleviating vascular wall inflammation, all of which contribute to the deceleration of atherosclerosis. Using MR analysis, this study identified 25-hydroxyvitamin D as a protective factor against coronary atherosclerosis. This finding aligns with previous research suggesting that vitamin D has a protective effect on cardiovascular health. However, differences in outcomes across distinct atherosclerosis types suggest that vitamin D's protective mechanisms may exhibit regional specificity. As the primary source of myocardial blood supply, the coronary artery possesses distinct local microenvironments, hemodynamics, endothelial cell characteristics, and vitamin D receptor (VDR) expression compared to cerebral and peripheral arteries. These variations likely account for the notable selectivity of vitamin D's effects [15,16]. Additionally, differences in sample size, statistical power, and relevant genetic backgrounds may also influence the analytical results across distinct vascular beds. The pathological progression of atherosclerosis exhibits high heterogeneity, and the development of coronary atherosclerosis is closely associated with myocardial ischemia [17,18]. Therefore, the protective effects of vitamin D may be more readily detectable in coronary artery lesions. Future studies should further investigate the cellular and molecular mechanisms of vitamin D action in different vascular beds to elucidate its site-dependent biological basis.

At the genetic level, four characteristic genes related to atherosclerosis have been identified: ADRA1B, ID4, MX2, and SEMA4D. The ADRA1B gene likely participates in vascular tone regulation, and its abnormal expression may disrupt vasoconstriction, thereby promoting the progression of atherosclerosis [19]. Angiogenesis is a key factor contributing to plaque rupture during atherosclerotic plaque development [20]. In breast cancer cells, elevated ID4 expression has been associated with the activation of a pro-angiogenic program in macrophages [21], suggesting its potential role in the angiogenesis of atherosclerotic plaques. MX2 is an interferon-stimulated gene (ISG) [22]. Interferon (IFN), a cytokine with both pro-inflammatory and immunomodulatory properties, induces the upregulation of ISGs [23,24]. High levels of MX2 expression are indicative of a robust inflammatory response, and given the crucial role of inflammation in the pathogenesis of atherosclerosis [25], MX2 may play a significant role in the development of this disease. Endothelial dysfunction in lesion-prone regions of the arterial vasculature is a major contributor to the pathobiology of atherosclerotic cardiovascular disease [26]. Research has shown that SEMA4D/PLEXINB1 induces endothelial cell dysfunction via the mDIA1 pathway [27]. Additionally, through GSEA analysis, we found that ID4 and ADRA1B were significantly negatively correlated with the lysosomal pathway, suggesting they may regulate lysosomal function. Lysosomes serve as crucial intracellular degradation systems responsible for the breakdown and recycling of damaged proteins and organelles [28]. In atherosclerosis, dysregulation of lysosomal function in macrophages and vascular smooth muscle cells leads to lipid deposition and exacerbated inflammation, thereby promoting plaque formation [29]. On the other hand, MX2 and SEMA4D showed positive correlations with the hematopoietic cell lineage pathway, indicating their involvement in immune cell development and regulation. The hematopoietic lineage encompasses various immune cells such as macrophages and T cells, which participate in the immune inflammatory response of atherosclerosis [30]. MX2 and SEMA4D may influence the development and function of immune cells, thereby regulating the immune microenvironment and affecting the progression of atherosclerosis. Taken together, these genes may influence the development of atherosclerosis by modulating processes such as angiogenesis and inflammation. However, the precise mechanisms underlying their interactions remain to be further explored. We also note that these genes may be associated with vitamin D metabolism and signaling pathways. For example, Lu et al. [31] identify significant enrichment of the SEMA4D gene in biological processes and signaling pathways related to the VDR through pathway enrichment analysis. VDR signaling is closely associated with the renin-angiotensin system and sympathetic nervous function, potentially indirectly regulating the physiological environment of adrenergic receptors such as ADRA1B [32,33]. ID4 [34] and MX2 [35] also appear to function in VDR-related pathways. These associations are mainly based on data mining and pathway enrichment, lacking direct experimental proof that vitamin D/VDR directly regulates these genes. More experiments are needed to clarify their exact roles and mechanisms.

Immunological analysis revealed significant differences in the infiltration of nine immune cell types, including Tregs and CD8+ T cells, in atherosclerotic samples. These characteristic genes were strongly correlated with immune cells such as CD8+ T cells, activated NK cells, monocytes, and naive B cells. Atherosclerosis is a chronic inflammatory disease involving various immune cells [36], with T cells playing a key role in its pathogenesis [37]. Alterations in the number or function of Tregs can disrupt immune balance [38]. In atherosclerosis, reduced Tregs may impair immune suppression and exacerbate the inflammatory response. CD8+ T cells produce pro-inflammatory cytokines like IFN-γ, perforin, and TNF-α, which further amplify inflammation [39]. The strong correlation between characteristic genes and immune cells suggests that these genes may influence the immune microenvironment of atherosclerosis by regulating immune cell function and migration. Recent studies indicate that PD-L1@NV can increase Tregs and reduce CD8+ T lymphocytes, alleviating myocardial infarction-related immunopathology and promoting myocardial repair, highlighting the crucial roles of Tregs and CD8+ T cells in myocardial infarction. In previous enrichment analyses, VMGs differentially expressed in atherosclerosis were primarily enriched in the leukocyte migration process, highlighting the crucial role of immune cell migration in the immune environment of cardiovascular disease. Leukocyte migration encompasses multiple immune cell subsets, potentially involving the migration and functional alterations of monocytes, macrophages, and various T cell subtypes (including CD8+ T cells) [40,41]. Immunomics data indicate a decline in CD8+ T cell numbers, potentially reflecting their migration from peripheral blood to the lesion site or depletion due to immunoregulatory mechanisms [42]. These findings reflect the complex dynamics and functional specialization of immune cell subsets within the atherosclerotic pathological process. This underscores the importance of immune factors in cardiovascular diseases. However, the mechanism by which vitamin D regulates these gene expressions and subsequently influences atherosclerosis through immune modulation remains hypothetical. The specific regulatory mechanisms between these immune cells, serum hydroxyvitamin D, and characteristic genes warrant further investigation.

This study has several limitations. Despite rigorous SNP screening and the use of multiple testing approaches in the MR analysis, undetected confounding factors and latent pleiotropic effects may still exist. Additionally, the data predominantly derive from the European population, which may introduce ethnic biases and limit the generalizability of the findings. Variations in gene frequencies related to vitamin D metabolism, genetic susceptibility to atherosclerosis, and environmental exposures across different ethnic groups may further impact the applicability of the results. For example, genetic differences in vitamin D-binding protein vary among ethnic groups and affect serum 25(OH)D levels and its activity [43]. Since the validation dataset did not specify participants' ethnicity, this may cause variability in results. Future studies should include diverse populations for broader validation. Despite limitations, this study supports a link between serum 25(OH)D and atherosclerosis, providing insights into disease mechanisms and prevention.

This study employed multi-omics approaches to explore the relationship between serum hydroxyvitamin D and atherosclerosis, identifying 25-hydroxyvitamin D as a potential protective factor for coronary atherosclerosis via MR. Four characteristic genes linked to angiogenesis and inflammation were identified. Immunological analysis further revealed correlations between these genes and immune cells. However, the study has limitations, including the potential influence of confounding factors, ethnic biases in the data, and the need for validation in diverse populations. Future research should focus on elucidating the underlying mechanisms and strengthening the evidence for clinical applications.

SUPPLEMENTARY MATERIALS

Supplementary data including ten tables and four figures can be found with this article online at https://doi.org/10.4196/kjpp.25.278