Study on diagnostic genes and immune microenvironment disorder in comorbid atherosclerosis and Alzheimer disease

Abstract

Atherosclerosis (AS) and Alzheimer disease (AD) are globally prevalent chronic diseases with challenges of difficult early diagnosis and a lack of effective combined therapies. This study explored their shared molecular mechanisms, potential diagnostic biomarkers, and immune microenvironment disorders. Using AS dataset GSE100927 and AD dataset GSE97760, key differentially expressed genes (key DEGs) were identified via bioinformatics (differential expression analysis, protein–protein interaction network, machine learning). A risk prediction model was built and validated with receiver operating characteristic/area under the curve. Immune infiltration was compared between patient and control groups; miRNA–transcription factor–mRNA and key DEGs–chemical networks were predicted, and key DEGs–AS/AD causal relationships verified by Mendelian randomization. Venn analysis found 89 common DEGs (enriched in lipid metabolism, Wnt pathway, etc). Four key DEGs (early growth response 2 [EGR2], leukocyte-specific transcript 1 [LST1], membrane-spanning 4-domains subfamily A member 7 [MS4A7], and 2’–5’-oligoadenylate synthetase like [OASL]) were confirmed. The model had high accuracy (C-index = 0.982, area under the curve > 0.9). Mendelian randomization showed EGR2 was an AS risk/AD protective factor; LST1 an AS risk factor; MS4A7 an AD risk factor; no clear OASL–AD causality. EGR2, LST1, MS4A7, and OASL are novel diagnostic biomarkers and potential therapeutic targets for comorbid AS and AD.

1. Introduction

Atherosclerosis (AS) and Alzheimer disease (AD) are chronic diseases with high incidence rates worldwide. They not only pose a serious threat to patients’ lives and health but also impose a heavy burden on the social economy. AS is one of the major etiologies of cardiovascular disease, and AD is a key cause of cognitive decline in the elderly.^[1] In recent years, vascular diseases such as cognitive impairment, dementia, and AD have gained attention.^[2,3] Some studies have suggested that AS may induce nerve damage based on cumulative pathological changes in the blood vessels, which could contribute to the development of AD.^[4] At present, the diagnosis and treatment methods for AS and AD have many limitations, such as difficulty in early diagnosis and the lack of effective combined diagnosis and treatment strategies. These problems urgently require further research to be addressed.^[5] Therefore, exploring the common molecular mechanisms and diagnostic biomarkers of these 2 diseases is of great clinical significance.

AS is a disease involving pathological changes in the blood vessels of human tissues.^[6] AD is a neurodegenerative disease that is characterized by memory loss and mental dullness.^[7] The inflammatory response is regarded as an important link between AS and AD. The occurrence of AS is accompanied by chronic low-grade inflammation and impaired vascular endothelial function, which promotes lipid deposition in the vascular wall and plaque formation.^[8] This inflammatory state is not limited to large blood vessels but can also spread to cerebral microvessels, leading to insufficient cerebral blood perfusion, which in turn induces neuronal damage and cognitive decline.^[9] In addition, AS is often complicated by metabolic abnormalities, such as hypertension and diabetes. These factors promote the abnormal accumulation of β-amyloid (Aβ) and tau proteins in the brain by exacerbating oxidative stress and inflammatory responses, thereby driving the progression of AD pathology.^[10,11]

Some studies have indicated that metabolic abnormalities, particularly diabetes mellitus and lipid metabolism disorders, are important mechanisms underlying comorbid AS and AD. The prevalence of cognitive impairment is significantly higher in patients with diabetes mellitus. Diabetes mellitus affects cerebral Aβ metabolism and tau protein abnormalities by promoting chronic inflammation, oxidative stress, and insulin resistance, thereby increasing the risk of AD.^[8,12] Although abnormalities in lipid metabolism, especially high-density lipoprotein dysfunction, are closely associated with cardiovascular diseases, they have also been found to be involved in the pathological process of AD. The reduced functions of high-density lipoprotein (including anti-inflammatory, antioxidant, and reverse cholesterol transport effects) may promote neurodegeneration.^[13]

Although numerous studies have explored the molecular mechanisms of AS and AD, most have focused on the individual mechanisms of each disease. Insufficient attention has been paid to the shared genetic mechanisms between these 2 diseases and the correlations between their immune responses.^[14] Therefore, investigating the mechanism of action of common genes in comorbid AS and AD, as well as their role in immune dysregulation, not only helps to reveal the pathogenic essence of these 2 diseases but also provides a theoretical basis for early diagnosis and individualized treatment. Conducting a genetic risk assessment targeting the comorbid genes of AS and AD may be an effective approach to slow down the progression of comorbid AS and AD. A research flowchart is presented in Figure 1.

Figure 1.

The flowchart of this study.

2. Methods

2.1. Dataset selection and preparation

The gene datasets selected for this study were sourced from Gene Expression Omnibus (GEO)^[15] (https://www.ncbi.nlm.nih.gov/geo/). To better reflect the actual clinical situation of patients, we selected datasets from patients with “atherosclerosis” (AS) and “Alzheimer disease” (AD). The AS dataset includes GSE100927^[16] and GSE28829,^[17] whereas the AD dataset includes GSE97760^[18]and GSE48350^[19] for research purposes. Among these, GSE100927 and GSE97760 served as the study sets, and GSE28829 and GSE48350 served as the validation sets. Specific information regarding the gene datasets was shown in Table 1.

Table 1.

Datasets information containing patients datas for AS and AD.

Disease	Datasets	Platform	Samples	Group
Atherosclerosis	GSE100927	GPL17077	69 patients and 35 controls	Rsearch
	GSE28829	GPL570	13 early patients and 16 advanced patients	Validation
Alzheimer disease	GSE97760	GPL16699	9 patients and 10 controls	Rsearch
	GSE48350	GPL570	80 patients and 173 controls	Validation

AD = Alzheimer disease, AS = atherosclerosis.

2.2. Gene expression levels and differentially expressed genes (DEGs) analysis

In this study, R software (version 4.2.1) was used to download, process, annotate, and analyze the datasets. The GEO query package^[20] (version 264.2) was used to download GSE100927, GSE28829, GSE9770, and GSE48350 from the GEO database. The Limma package^[21] (version 3.52.2) was used to analyze disease-related differential gene data from GSE10092 and GSE97760. The results of the differential analysis were visualized using a volcano plot, with the significance thresholds for differential genes set at |log2(fold change)| > 1 and a P-value < .05. Significantly expressed molecules were visualized in the form of heat maps using the ComplexHeatmap package (version 2.13.1).^[22]

2.3. Venn analysis of the intersection genes of DEGs between AS and AD and their enrichment analysis

This study utilized bioinformatics^[23](http://www.bioinformatics.com.cn/) to conduct a Venn analysis of DEGs between AS and AD, displaying the intersection of upregulated and downregulated genes in both diseases. Subsequently, enrichment analysis was performed on the intersecting DEGs, with the analysis species set to “Homo” as a reference, incorporating enrichment results with a P-value < .05 and visualizing them.

2.4. Identifying key DEGs

This study employed 3 machine learning methods combined with functional protein–protein interaction and molecular complex detection (PPI–MCODE) to identify key DEGs. First, PPI analysis of genes was conducted using the String 12.0^[24](https://cn.string-db.org/) database, and a PPI network was established. This was visualized using Cytoscape 3.10.0 software, and the key DEGs were filtered using the MCODE plugin. In addition, we employed 3 powerful machine learning methods to further identify key intersection DEGs: the least absolute shrinkage and selection operator (LASSO)^[25] algorithm, Support Vector Machine Recursive Feature Elimination (SVM-RFE),^[26] and Random Forest analysis.^[27] The key parameter settings for the LASSO algorithm were family = “binomial” and nfolds = 10. The “glmnet” package was utilized to optimize and implement LASSO logistic regression. The “1071” R package was used to perform SVM-RFE, combined with ten-fold cross-validation, to simplify the feature vector generated by SVM and improve the accuracy of the algorithm. A total of 500 decision trees were constructed to perform Random Forest analysis, selecting genes with importance scores >5 as the RF feature genes. Finally, the intersecting genes from the results of the 3 machine learning methods and the PPI–MCODE results were taken as key DEGs for subsequent research.

2.5. Diagnostic value model prediction and validation of key DEGs

To better align with clinical situations, this study utilized the R software package rms (version 6.4.0) to construct predictive diagnostic Nomograms^[28] for key DEGs related to AS and AD. Additionally, we evaluated the diagnostic model’s value using test datasets for AS and AD (GSE100927 and GSE97760), and validation datasets (GSE28829 and GSE48350). The predictive accuracy of the key intersection DEGs model was assessed using receiver operating characteristic (ROC) curves and the area under the curve (AUC).^[29] When AUC > 0.5, the closer the AUC is to 1, the better the diagnostic effect of the variable in predicting the outcomes. An AUC above 0.9 indicates high accuracy.

2.6. Immune infiltration analysis

Using thesingle-sample Gene Set Enrichment Analysis algorithm^[30] provided in the R package GSVA and the markers of 24 types of immune cells provided in the Immunity article, we calculated the immune infiltration between the control group and the study group in AS and AD samples, and compared the expression levels of immune-related cells. In addition, the CIBERSORT algorithm^[31] was used to assess the composition and function of the immune cells. Through CIBERSORT analysis, the relative proportions of the 22 immune cell subpopulations in the samples were quantitatively estimated, further revealing differences in the immune microenvironment and its functional status. Subsequently, we analyzed the correlation between key DEGs and various immune cells and immune cell functions and visualized the results using heatmaps.

2.7. Key intersection DEGs–miRNA–transcription factor and key intersection DEGs–chemical compounds network

By setting the parameter species to “Homo,” we predicted the miRNAs of the key intersection DEGs using the starBase or ENCORI^[32] (https://rnasysu.com/encori/index.php), miRNAwalk^[33] (http://129.206.7.150/) and miRDB^[34] (http://www.mirdb.org/) databases. We predicted the transcription factor (TF) targets of the key intersection DEGs in the Chipbase v3.0^[35] (https://rnasysu.com/chipbase3/index.php) and KnockTF 2.0^[36] (http://www.licpathway.net/KnockTF/index.html) databases and constructed an miRNA–TF–mRNA regulatory network. In addition, we predicted and downloaded chemical drugs related to the key intersection DEGs from the Comparative Toxicogenomics Database^[37] (https://ctdbase.org). The constructed networks were visualized using Cytoscape 3.10.0.

2.8. Mendelian randomization validation

Mendelian randomization (MR) studies reflect the causal relationship between genetic variation as an instrumental variable and the occurrence of diseases.^[38] In this study, 2-sample MR analysis was performed to evaluate the causal relationship between key DEGs, AS, and AD. The Genome-Wide Association Study (GWAS) IDs of the key DEGs were obtained from the National Center for Biotechnology Information ^[39](https://www.ncbi.nlm.nih.gov/) database, and the expression quantitative trait locus summary level data were sourced from the IEU OpenGwas database^[40] (https://gwas.mrcieu.ac.uk/). The GWAS summary level data for diseases AS^[41] and AD^[42] were obtained from the GWAS Catalog^[43] (https://www.ebi.ac.uk/gwas/home) database.

2.9. Ethical

All analyses in this study were conducted using publicly available data from the GEO database, IEU open GWAS project, and GWAS catalog. All original datasets had been previously approved by the corresponding institutional ethics committees, and written informed consent had been obtained from all participants in the original studies. Therefore, additional ethical approval and informed consent were not required for the present bioinformatics analysis, MR study, and secondary data analysis. No human participants, human biological samples, or animal experiments were involved in this study.

3. Results

3.1. Analysis of DEGs between AS and AD

To identify DEGs between the disease and control groups, we conducted differential gene expression analysis on the AS dataset GSE100927 and AD dataset GSE97760. The Limma package was used to identify DEGs, with the parameters |log2(fold change)| > 1 and P-value < .05 set as filtering criteria. We found that the GSE100927 dataset contained 540 DEGs, of which 388 were upregulated and 152 were downregulated. The GSE97760 dataset contained 8071 DEGs, with 3720 upregulated and 4351 downregulated genes. We performed clustering analysis of these DEGs and visualized them using volcano plots (Fig. 2A, B). The heatmap in Figure 2C, D shows the expression levels of the DEGs, indicating the high reliability of sample clustering.

Figure 2.

DEG analysis on AS and AD. (A and B) Volcanic plots of gene expression of AS in GSE100927 (A) and AD in GSE97760 (B). Red represents upregulated DEGs, blue represents downregulated DEGs. P < .5, |log2(fold change)| > 1. (C and D) Heat maps of the AS in GSE100927 (C) and AD in GSE97760 (D) DEGs. Red color indicates upregulated genes, and blue indicates downregulated genes. AD = Alzheimer disease, AS = atherosclerosis, DEGs = differentially expressed genes.

3.2. Identification of intersection DEGs and enrichment analysis

The intersection of AS–DEGs and AD–DEGs was identified using Venn analysis. A total of 89 intersection DEGs were screened (Fig. 3A), of which 21 were upregulated in both AS and AD, 23 were downregulated in both AS and AD, 31 were upregulated in AS, and 14 were upregulated in AD (Fig. 3B). We also performed Gene Ontology enrichment and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses of intersecting DEGs. The results of Gene Ontology enrichment analysis indicated that biological processes (Fig. 3C) were significantly enriched, showing positive regulation of lipid localization, regulation of plasma lipoprotein particle levels, regulation of lipid localization, and organization of actin–myosin structures. Cellular components (Fig. 3D) were significantly enriched, showing collagen-containing extracellular matrix, plasma lipoprotein particles, I bands, protein–lipid complexes, and Z discs. Molecular functions (Fig. 3E) were significantly enriched, showing phospholipid binding, glycosaminoglycan binding, heparin binding, actin filament binding, protein–lipid complex binding, and lipoprotein particle binding. Kyoto Encyclopedia of Genes and Genomes pathway (Fig. 3F) enrichment indicated that leukocyte transendothelial migration, cholesterol metabolism, phagosome, and Wnt signaling pathways were significantly enriched.

Figure 3.

Intersection DEGs and results of functional enrichment analysis. (A) Venn diagram of the intersection DEGs of AS and AD. (B) Venn diagram of the up- and down regulated intersection genes between AS and AD. (C–E) GO analysis of the intersection genes. The plots show the top 10 processes enriched by intersection genes in BP, CC, and MF. (F) KEGG enrichment analysis of the intersection DEGs. AD = Alzheimer disease, AS = atherosclerosis, BP = biological process, CC = cellular component, DEGs = differentially expressed genes, KEGG = Kyoto Encyclopedia of Genes and Genomes, MF = molecular function.

3.3. Identification of key DEGs through PPI network and machine learning

To further identify key DEGs, we conducted PPI analysis on 89 intersecting DEGs and constructed a network for visualization in Cytoscape 3.10.0 (Fig. 4A). The cytoscape–MCODE plugin was used to calculate and filter the key DEGs (Fig. 4B, C). In addition, 3 machine learning methods, LASSO (Fig. 4D, E), random forest (Fig. 4F), and SVM-RFE (Fig. 4G), were employed to filter the key DEGs. Finally, Venn analysis of the key DEGs identified by the 4 models resulted in 4 key DEGs, namely early growth response 2 (EGR2), leukocyte-specific transcript 1 (LST1), membrane-spanning 4-domains A7 (MS4A7), and 2’–5’-oligoadenylate synthetase like (OASL) (Fig. 4H).

Figure 4.

Screening results of key intersection DEGs. (A) PPI network of intersection DEGs. Red DEGs represent upregulated genes, blue represents downregulated genes; (B and C) core genes in the screened PPI network calculated using the MCODE plugin; (D) LASSO regression model screening for key intersection DEGs; (E) LASSO variable trajectory; (F) screening of key intersection DEGs using random forest with parameters set for heterogeneity analysis of 500 decision trees, importance score > 5; (G) screening of key intersection DEGs using the SVM-RFE method; (H) Venn diagram showing the common key intersection DEGs screened by 4 models: PPI–MCODE, diagnostic Lasso, random forest, and SVM-RFE. DEGs = differentially expressed genes, LASSO = least absolute shrinkage and selection operator, MCODE = molecular complex detection, PPI = protein–protein interaction, SVM-RFE = support vector machine recursive feature elimination.

3.4. Risk prediction model and diagnostic value assessment of key intersection DEGs in AS and AD

Based on the expression profiles of the test sets for AS and AD (GSE100927 and GSE97760), we analyzed the expression of key DEGs in AS and AD (Fig. 5A, B). Subsequently, we established and validated a risk prediction model for the key intersecting DEGs (Fig. 5C). The risk in the model was quantified by reflecting the expression score of each gene using a nomogram. The C-index for the discrimination ability of the model was 0.982, indicating high accuracy. The clinical diagnostic value of the key intersecting DEGs was verified using ROC and AUC analyses. ROC and AUC analyses were performed using the test and validation sets for AS and AD, respectively. When AUC > 0.5, the diagnostic efficacy improved as the AUC approached 1. AUC values exceeding 0.9 indicate high accuracy. In the AS test set GSE100927 (Fig. 5D, E) and the AD test set GSE97760 (Fig. 5H, I), the AUC values of the key intersection DEGs and the model AUC values were both above 0.9, indicating high diagnostic performance of the key intersection DEGs. In the AS validation set GSE28829 (Fig. 5F, G), the AUC values of the key intersection DEGs ranged from 0.452 to 0.904, whereas the model AUC was 0.952. In the AD validation set GSE48350 (Fig. 5J, K), the AUC values of the key DEGs ranged from 0.585 to 0.660, and the model AUC was 0.677, indicating that the 4 key DEGs still had good diagnostic value in the AS and AD validation sets.

Figure 5.

Risk prediction model and diagnostic ROC validation analysis results of key intersecting DEGs expression in AS and AD patients. (A and B) Expression of key intersecting DEGs in the AS test dataset GSE100927 and the AD test dataset GSE97760; (C) construction of a diagnostic Nomogram model for 4 key intersecting DEGs; (D–K) diagnostic ROC curves and AUC analysis of key intersecting DEGs in the AS test dataset GSE100927 (D and E), AS validation dataset GSE28829 (F and G), AD test dataset GSE97760 (H and I), and AD validation dataset GSE48350 (J and K). AD = Alzheimer disease, AS = atherosclerosis, AUC = area under the curve, DEGs = differentially expressed genes, ROC = receiver operating characteristic.

3.5. Immune infiltration in AS and AD patients

Comparative analysis of immune infiltration between patients with AS and the control group revealed immune dysregulation across 12 immune cell types, including B cells, CD8 + T cells, and cytotoxic cells (Fig. 6A). Furthermore, we observed imbalances in multiple immune functions in patients with AS, including naive and memory B cells, plasma cells, regulatory T cells, resting NK cells, mast cells, monocytes, and M0 and M2 macrophages (Fig. 6B). As shown in Figure 6C, key DEGs, EGR2, LST1, MS4A7, and OASL, demonstrated correlations with various immune cells and functions. Notably, EGR2 exhibited a significant positive correlation with activated mast cells, whereas LST1 and MS4A7 showed significant negative correlations with dendritic cells and NK cells. Comparative analysis of immune infiltration between patients with AD and the control group revealed potential immune dysregulation across 18 immune cell types, including CD8 + T cells, eosinophils, mast cells, NK cells, and T cells (Fig. 6D). Additionally, imbalances were observed in B-cell memory function, CD8 + T cells, neutrophils, activated NK cells, and immune functions of M0 and M2 macrophages in patients with AD (Fig. 6E). As shown in Figure 6F, key DEGs, including EGR2, LST1, MS4A7, and OASL, were correlated with various immune cells and functions. Notably, MS4A7 exhibited a significant positive correlation with CD8 + T cells and T helper cells and a significant negative correlation with mast cells and NK cells. OASL was significantly negatively correlated with CD8 + T cells, Tcm, and Tem.

Figure 6.

Immune infiltration status of AS and AD patients compared to the control group. ssGSEA analysis results of immune cells (A) and immune functions (B) in AS patients; (C) correlation of key DEGs with AS-related immune cells and immune functions; ssGSEA analysis results of immune cells (D) and immune functions (E) in AD patients; (F) correlation of key DEGs with AS-related immune cells and immune functions. AD = Alzheimer disease, AS = atherosclerosis, DEGs = differentially expressed genes, ssGSEA = single-sample Gene Set Enrichment Analysis.

3.6. Construction of miRNA–TF–mRNA regulatory network and key DEGs–chemical network

We performed a predictive analysis of miRNAs and TFs targeting key intersecting DEGs and constructed a visual regulatory network (Fig. 7A). This regulatory network comprises 4 genes, 90 TFs, and 186 miRNAs. Furthermore, we predicted chemicals potentially targeting the 4 key intersecting DEGs and visualized their interaction networks (Fig. 7B). Five chemical drugs were found to act on 2 or more key intersecting DEGs: Benzo(a)pyrene, Bisphenol A, Arsenic, and JP8 aviation fuel.

Figure 7.

miRNA–TF–mRNA and DEGs–chemical network. (A) The miRNA–TF–mRNA regulatory network of 4 key intersecting DEGs. Yellow represents key DEGs, purple represents miRNA, and pink represents TF; (B) the DEGs–chemical interaction network of 4 key intersecting DEGs. Yellow represents key DEGs, purple represents compounds that act on key DEGs, and red represents compounds that act on 2 or more key DEGs. DEGs = differentially expressed genes, TF = transcription factor.

3.7. Validation of relationships between four key DEGs with AS and AD through MR study

Utilizing expression quantitative trait locus data from the 4 key DEGs, we conducted MR studies to analyze their causal relationships with AS and AD. MR study results indicated a correlation between EGR2 and AS. Inverse variance weighted (IVW) analysis showed an OR = 3.31 (95% CI = 1.22–8.96), β = 1.20 (P < .05) (Fig. 8A), suggesting that elevated EGR2 expression increases AS disease risk. The heterogeneity test results (Q_P > .05) indicated no heterogeneity among the instrumental variables (Fig. 8B), thus neglecting the impact of heterogeneity on causal effect estimation. MR-Egger regression analysis (Fig. 8C) for pleiotropy testing yielded an intercept of 0.016083 (P = .265 > 0.05), demonstrating a weak likelihood of pleiotropy. Meanwhile, the scatter plot shows that the slopes of both the MR-Egger and IVW results are greater than zero, indicating that EGR2 acts as a risk factor for AS. Leave-one-out analysis confirmed that the overall estimate remained unaffected by any individual SNP (Fig. 8D).

Figure 8.

Mendelian randomization analysis results. (A) MR analysis results of EGR2 with AS. (B) Funnel plot of EGR2–AS causality demonstrates no heterogeneity among SNPs. (C) Scatter plot of EGR2–AS causality indicates no horizontal pleiotropy among SNPs. (D) Leave-one-out analysis of EGR2–AS causality. (E) MR analysis results of EGR2 with AD. (F) Funnel plot of EGR2–AD causality. (G) Scatter plot of EGR2–AD causality. (H) Leave-one-out analysis of EGR2–AD causality. AD = Alzheimer disease, AS = atherosclerosis, EGR2 = early growth response 2, MR = Mendelian randomization.

The MR study results demonstrated a correlation between ERG2 and AD. IVW analysis indicated an OR of 0.502 (95% CI = 0.276–0.91), β = −0.69, P < .05 (Fig. 8E), suggesting that EGR2 may exert a protective effect against AD. Elevated EGR2 expression correlates with a reduced risk of AD. Heterogeneity test results (Q_P > 0.05, Fig. 8F) indicated the absence of heterogeneity in the MR findings. MR-Egger regression analysis (intercept = 0.0029683, P = .699 > 0.05, Fig. 8G) was employed for pleiotropy testing, revealing no significant overall horizontal pleiotropy. A slope of <0 indicates that EGR2 serves as a protective factor against AD. Leave-one-out analysis demonstrated that the overall estimates were unaffected by any individual SNP (Fig. 8H).

The MR study results confirmed causal relationships between LST1 and AS (Figure S1, Supplemental Digital Content 1), as well as between MS4A7 and AD (Figure S2, Supplemental Digital Content 2). LST1 acts as a risk factor for AS, and elevated LST1 expression increases susceptibility to AS. MS4A7 functions as a risk factor for AD, and increased MS4A7 expression increases the risk of AD.

The results indicated no causal relationship between OASL and AD (P > .05) (Figure S3, Supplemental Digital Content 3).

4. Discussion

AS is an important risk factor for cardiovascular disease.^[44] AD is a neurodegenerative disease characterized by a delayed response and decreased cognitive ability, with pathological features in the brain Aβ and excessive phosphorylation of tau deposition.^[45] Recently, increasing attention has been paid to the influence of AS on cognitive impairment, dementia, and the occurrence and development of AD.^[46] Some epidemiological studies^[47,48] have reported that there is an independent interaction between the incidence rates of AS and AD in some populations. There is a molecular mechanism underlying overlapping interactions between AS and AD. Some studies^[49–51] have shown that multiple risk factors or pathogenic genes associated with AS may be associated with the risk of future cognitive decline in individuals without dementia. An imaging study used the measured thickness of the carotid artery in patients with carotid AS as a measurement index and found that the greater the thickness of the carotid artery middle layer, the faster the cognitive decline in exchange.^[52] Previously, it was generally believed that intracranial AS might be related to cerebral infarction, but a study on intracranial AS showed that the risk factors related to intracranial AS increased the risk of AD by influencing the manner of vascular changes.^[53] Further research has verified that Aβ toxicity accumulation, cerebrovascular lesions, oxidative stress, and neuroinflammation serve as the core shared pathological mechanisms linking vascular diseases and AD progression, which fundamentally explains the pathological correlation and comorbidity of the 2 diseases.^[54]

As a shared pathological basis of AS and AD, a chronic inflammatory response acts as a crucial bridge connecting the 2 diseases by activating the complement system and inflammatory mediators, thereby promoting vascular endothelial injury and neuroinflammation.^[55] Studies have shown that abnormal activation of the immune system is observed in both AS and AD. Changes in T-cell subsets are closely associated with vascular endothelial function, suggesting that immunoinflammation may contribute to the development of comorbidities between the 2 diseases.^[56] Clinical evidence further indicates that the dysregulation of peripheral T lymphocyte proportion serves as a vital immune biomarker for evaluating disease progression and prognosis of AD, and immune homeostasis imbalance can significantly aggravate AD pathological damage and poor clinical outcomes, which further highlights the essential regulatory role of immune disorders in the comorbid progression of AS and AD.^[57] In addition, immune cells, such as macrophages and microglia, exhibit similar inflammatory phenotypes in AS and AD.^[58,59]

Furthermore, some inflammatory mediators are significantly upregulated in both AS and AD.^[60] Vascular inflammation promotes intracerebral inflammatory responses by damaging the blood–brain barrier, whereas systemic inflammation activates intracerebral immune cells via circulating mediators. These processes form a positive feedback loop for inflammation, which drives the progression of AD pathology.

Although existing research data strongly indicate that AS has an impact on AD, the experimental methods and number of related in vivo studies are limited. Therefore, it is necessary to explore and apply new, effective research methods and models. In this study, bioinformatics, machine learning, and other methods were employed to investigate the key genes commonly involved in the comorbidity of AS and AD, as well as their biological mechanisms of action. We identified 4 key DEGs: EGR2, LST1, MS4A7, and OASL.

EGR2 is a typical zinc finger transcription factor belonging to the C2H2 type zinc finger protein family. EGR2 is involved in important biological processes such as inflammatory responses and cell apoptosis.^[61] Research shows that with increasing age, the expression of EGR2 in macrophages in the AS lesion area is upregulated, particularly in alternatively activated anti-inflammatory macrophages (CD163 + and CD206+).^[62] EGR2 is a key transcription factor that plays an important role in the development and maintenance of the nervous system. Changes in the expression of EGR2 affect neuron survival and synaptic plasticity. Studies have shown that in AD animal models, the EGR2 gene exhibits significant expression changes in the early stages of the disease, suggesting that it may serve as an early biomarker for AD.^[63]

LST1 is a small transmembrane adapter protein mainly expressed in myeloid leukocytes (e.g., monocyte–phagocyte lineage) and plays a vital role in the immune system.^[64] LST1 influences the occurrence and development of AS by regulating macrophage migration and release of inflammatory factors.^[65] A recent study using whole-genome sequencing and single-cell sequencing data analysis^[66] revealed that LST1 is significantly overexpressed in unstable AS plaques. Additionally, LST1 is involved in the activation of multiple signaling pathways related to inflammation and oxidative stress, such as the PI3K–Akt pathway, platelet activation, and hypoxia-inducible factor regulation. The abnormal activation of these pathways collectively promotes AS progression. In the occurrence and development of AD, LST1 may be involved in neuroinflammatory responses by regulating immune activation in the brain. One study indicated that changes in LST1 expression are significantly associated with the burden of neurofibrillary tangles in AD brain tissues, which proves that LST1 may participate in the progression of AD by affecting the expression profile of inflammation-related genes in AD brain tissues.^[67]

MS4A7 belongs to the transmembrane protein family.^[68] Proteins of this family are widely distributed in various immune cells, with particularly high expression in the monocyte–macrophage lineage. During the progression of AS, MS4A7 regulates macrophage lipid uptake, clearance of apoptotic cells (apoptotic bodies), and release of inflammatory mediators, thereby affecting lipid accumulation in plaques and the inflammatory microenvironment.^[69] Additionally, MS4A7 is involved in the interaction between vascular endothelial cells and immune cells and regulates the response of endothelial cells to inflammatory signals, as well as the adhesion and migration of immune cells, which are of great significance for plaque formation and stability. During the progression of AD, microglia stimulated by pathological factors become activated and release a variety of inflammatory factors, which are involved in the process of intracerebral inflammatory responses.^[70] Changes in MS4A7 expression may affect the activation level of microglia, thereby regulating the intensity and duration of inflammatory responses.^[71]

OASL is a protein whose expression is induced by interferons. The main functions of OASL are reflected in its antiviral and immunomodulatory roles.^[72] During the development of AS, OASL primarily regulates immune responses and inflammation.^[73] OASL can affect the production of inflammatory factors and apoptosis in macrophages, thereby regulating the local inflammatory microenvironment^[74] and plaque stability.^[75] In the progression of AD, OASL mainly influences the intracerebral immune response by regulating the interferon signaling pathway and thus participates in neuroinflammation and neuronal damage.^[76] Meanwhile, the activated interferon-related pathway enhances the clearance capacity of viruses and abnormal proteins; however, its excessive activation may also trigger chronic inflammatory responses, cause neuronal damage, and promote AD progression.^[77]

Chronic inflammatory response plays a crucial role in the local pathological processes of AS and AD. In AS, the infiltration and activation of immune cells, particularly macrophages, T cells, and neutrophils, within the arterial wall drives the persistence of inflammatory responses and the instability of AS plaques.^[78,79] Meanwhile, in patients with AD, the abnormal activation of microglia (brain-resident immune cells) and peripheral immune cells, as well as the release of inflammatory factors, are involved in the occurrence of neuronal damage and cognitive impairment.^[80–82]

A growing body of evidence indicates a high degree of overlap between AS and AD in terms of molecular pathways, inflammatory signals, and immune cell subsets. For instance, the complement system, NLRP3 inflammasome, and TREM2 signaling pathway play key roles in both diseases.^[80,82–84]

Furthermore, changes in immune cell function promote mutual exacerbation and comorbid progression of the 2 diseases through mechanisms such as systemic inflammatory responses, cytokine networks, and immune cell migration.^[82,84] As a case in point, the monocyte–macrophage system contributes significantly to both AS plaque formation and intracerebral inflammation in AD; the imbalance of T-cell subsets is also regarded as a common pathogenic factor for both conditions.^[80,85]

In conclusion, by systematically analyzing the overlapping DEGs between AS and AD, this study revealed the common pathological mechanisms underlying these 2 diseases and provided a solid foundation for the construction of diagnostic models. These findings not only enrich the understanding of AS and AD but also offer new directions for future clinical research and the development of therapeutic strategies. By integrating multi-omics data and validating them through wet laboratory experiments, it is expected to further advance progress in this field.

5. Limitations

The limitations of this study are mainly reflected in the constraints of data sources. Although we utilized multiple sets of gene expression data for comprehensive analysis, reliance on results from public databases may lead to sample heterogeneity and batch effects, which could affect the accuracy of our DEG identification. In addition, despite employing a variety of bioinformatics methods to screen for key genes, wet laboratory validation was not performed. This leaves uncertainties regarding the biological functions of key DEGs and their specific roles in the pathogenesis of AS and AD. Furthermore, the lack of validation data from diverse populations might compromise the practicality and accuracy of the model. Future studies should integrate multi-omics data, combine basic experiments with clinical research, and further validate the functions of the key DEGs. This will enable a more comprehensive understanding of the common pathological mechanisms of AS and AD and facilitate clinical translation.

6. Conclusion

This study systematically identified the common DEGs between AS and AD, successfully screened 4 key genes – EGR2, LST1, MS4A7, and OASL – constructed a highly efficient disease risk prediction model, and revealed the close association between these genes and immune infiltration. Through MR analysis, we further confirmed the causal relationship between key genes and the 2 diseases.

These findings provide a novel perspective for understanding the comorbidity mechanism of AS and AD and lay a foundation for the development of early diagnosis methods and immune-related therapeutic strategies. Although this study has certain limitations, its results point out the direction for future research, which is conducive to promoting clinical translation in this field and the advancement of precision medicine, thereby offering new ideas and schemes for the early diagnosis of patients at high risk of AS and AD.

Acknowledgments

The authors gratefully acknowledge the data provided by the patients and researchers participated in GEO, IEU open GWAS project and GWAS Catalog.

Author contributions

Conceptualization: Pengyun Ni.

Data curation: Pengyun Ni.

Formal analysis: Bingbing Zhao.

Funding acquisition: Pengyun Ni.

Investigation: Hao Xv.

Methodology: Pengyun Ni, Hao Xv.

Project administration: Hao Xv.

Resources: Pengyun Ni, Hao Xv.

Software: Pengyun Ni, Bingbing Zhao.

Supervision: Hao Xv.

Validation: Pengyun Ni, Bingbing Zhao, Hao Xv.

Visualization: Pengyun Ni.

Writing – original draft: Pengyun Ni.

Writing – review & editing: Bingbing Zhao.