Identification of Cuproptosis‐Related Gene Clusters in Behçet's Disease and Its Immunological Profiles by Bioinformatics Analysis

ABSTRACT

Background and Aims

Behçet's disease (BD) is a chronic inflammatory vasculitis marked by immune cell abnormalities and clinical variability. The recently discovered type of programmed cell death, termed cuproptosis, appears to be involved in multiple disease mechanisms. Thus, this study aimed to elucidate the involvement of cuproptosis‐related genes (CRGs) in BD.

Methods

We obtained two bipolar disorder datasets from the gene expression omnibus repository and pinpointed differentially expressed genes linked to cuproptosis (CuDEGs) from a selection of 52 CRGs. Subsequently, machine learning methods were employed to identify hub CuDEGs. We analyzed 44 BD specimens to identify two unique subgroups derived from these hub CuDEGs. Additionally, we conducted gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes pathway enrichment studies, along with gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA), and assessed immune cell infiltration and immune function‐related cells.

Results

A total of 15 CuDEGs were discovered. Using machine learning algorithms, six hub CuDEGs were identified: ANKRD9, COX11, MT1G, MT2A, MT4, and TYR. These hub CuDEGs generally showed elevated expression levels in BD samples, leading to the identification of two CuDEGs‐related subclusters. Cluster 2 had higher expression of most hub CuDEGs. GSEA and GSVA results demonstrated that different hub CuDEGs were enriched in distinct pathways. Analyses of immune infiltration and function‐related cells revealed that hub CuDEGs were involved in various immune response processes, such as macrophage and neutrophil activation, highlighting significant differences in immune regulation among the hub CuDEGs.

Conclusion

This study demonstrates that CuDEGs are differentially expressed in BD and are associated with immune activity and pathway alterations. These findings suggest that cuproptosis may contribute to BD progression through the modulation of immune responses. Further experimental studies are needed to confirm these bioinformatics‐based findings and explore their therapeutic potential.

1. Introduction

Behçet's disease (BD) is a chronic and multisystem inflammatory disorder marked by recurrent oral ulcers, genital ulcers, skin lesions, and ocular inflammation [1]. The precise cause of BD is not fully understood; however, it is thought to stem from a complex interaction of genetic, immunological, and environmental factors [2]. The prevalence of BD is higher in regions along the historical Silk Road, including the Mediterranean, East Asia, and the Middle East, and it shows variations across different genders and ethnic groups [3]. Current treatments primarily aim to relieve symptom and reduce the frequency of flares. Common therapeutic approaches include glucocorticoids, immunosuppressants, biologics, and topical agents [4]. In recent years, increasing attention has been paid to the molecular mechanisms underlying BD. Understanding these mechanisms is essential for improving diagnosis and developing targeted therapies. In this context, our study investigates the role of cuproptosis—a newly identified, copper‐dependent form of programmed cell death—in BD.

Copper is an essential cofactor for many enzymes and is important for various bodily functions [5]. Under normal cellular conditions, copper levels are precisely controlled by active transport systems that prevent the buildup of free intracellular copper ions, thereby maintaining copper homeostasis [6, 7]. Copper ion carriers, which are small molecules that bind to copper, facilitate its intracellular transport and are instrumental in research to understand copper toxicity mechanisms [8, 9]. The cell death mechanism induced by these carriers is attributed to the intracellular accumulation of copper rather than the carriers themselves [10]. Disruption in copper balance can result in oxidative stress [11] and autophagy dysregulation [12], leading to various copper‐related diseases. Cuproptosis, a distinct form of copper‐dependent programmed cell death, differs from apoptosis, pyroptosis, necrosis, and autophagy. During cuproptosis, copper ions bind directly to lipoylated proteins that participate in the mitochondrial tricarboxylic acid cycle, leading to protein aggregation, disruption of iron‐sulfur clusters, proteotoxic stress, and eventually cell death. Although cuproptosis has been linked to mitochondrial dysfunction and immune responses, its relevance to BD remains largely unknown. Given that BD is driven by immune dysregulation, it is important to explore whether cuproptosis contributes to its development, especially through effects on immune signaling and oxidative stress pathways.

Several studies have reported elevated levels of copper in patients with BD. Notably, patients with BD, especially those exhibiting severe symptoms, showed elevated serum copper concentrations [13]. This elevation in copper levels correlated with increased oxidative stress markers such as malondialdehyde and ceruloplasmin, suggesting that copper might contribute to oxidative damage in BD [14, 15]. Furthermore, research indicated a positive association between copper concentrations and specific clinical indicators of BD. For example, serum copper levels were observed to be linked with the frequency of oral ulcers, implying a role in mucocutaneous inflammation [13]. Additionally, elevated levels of copper and ceruloplasmin in plasma have been linked to impaired polymorphonuclear leukocyte functions and increased oxidative enzyme activities, which contribute to tissue damage in BD [15]. These findings suggested that copper‐induced oxidative stress may exacerbate inflammatory responses and tissue injury in BD. Copper‐related enzymes, such as Cu‐Zn superoxide dismutase (Cu‐Zn SOD), exhibited variable activity in patients with BD. For instance, Taysi et al. documented increased erythrocyte Cu‐Zn SOD activity, while Tüzün et al. found no significant changes in Cu‐Zn SOD levels across different phases of the disease [16, 17]. Furthermore, elevated copper levels exacerbated oxidative stress by promoting reactive oxygen species (ROS) production or impairing antioxidant defenses [14, 15, 16]. Evidence indicated that copper dysregulation, alongside other oxidative stress markers, represented a potential therapeutic target in BD. Antioxidant therapies designed to normalize copper levels or counteract its oxidative effects could have enhanced clinical outcomes in these patients [16, 18]. Recent studies had underscored the critical role of copper homeostasis in immune cell infiltration [19]. Recent studies have linked cuproptosis to immune modulation, including T cell activation and macrophage polarization, as well as to the pathogenesis of autoimmune diseases such as rheumatoid arthritis and systemic lupus erythematosus. However, its specific role in BD has not yet been explored [20, 21]. Therefore, we hypothesize that cuproptosis‐related genes (CRGs) are differentially expressed in BD and may serve as potential biomarkers or therapeutic targets.

This study employed the gene expression omnibus (GEO) database to examine copper‐associated genes (CRGs) in individuals with BD and in healthy controls. Employing various machine learning techniques, we pinpointed key differentially expressed cuproptosis‐related genes (CuDEGs). Following this, BD patients were divided into two separate categories according to the expression patterns of these key CuDEGs, which play a role in copper‐triggered cell death. Subsequently, we examined the differences in immune cell infiltration among these groups. Additionally, we conducted gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) to pinpoint the signaling pathways linked to each CuDEG. The research additionally examined the relationships among hub CuDEGs, immune cell infiltration, and immune functionality. Furthermore, a regulatory network of competing endogenous RNAs (CeRNAs) was developed to provide fresh insights into the molecular mechanisms underlying BD pathogenesis. By elucidating the role of cuproptosis in BD and identifying potential therapeutic targets, this study aims to provide novel insights into disease mechanisms and support the development of future treatment strategies.

2. Materials and Methods

2.1. Acquisition of Datasets and CRGs

We employed the “GEOquery” package in R to retrieve two raw datasets, GSE17114 and GSE209567 [22], from the GEO database, containing gene expression data from BD patients and controls. The GSE17114 data set comprises 15 BD and 14 control samples, while the GSE209567 data set includes 29 BD and 15 control samples. The detailed information for these two datasets was presented in Table S1. Given that these datasets share the same platform, they were merged into a single data set for the training cohort. The GSE209567 data set was designated as the validation cohort. Raw data were normalized, annotated with background subtraction, and batch effects were corrected using the “SVA” package. CRGs were identified from the Molecular Signature Database (MsigDB) v7.0 [23] and supplemented with gene sets associated with cuproptosis from a previous study [10, 24]. After eliminating duplicates, we identified 52 CRGs.

2.2. Identification of CuDEGs in BD

We performed differential gene expression analysis employing the Wilcoxon rank‐sum test, setting a significance level of p < 0.05 to detect differentially expressed genes (DEGs) between BD and healthy samples in the merged data set. The results of the DEGs were visualized using boxplots and heatmaps. The interactions between DEGs were illustrated using the “circlize” and “corrplot” R packages. Furthermore, the “RCircos” R package was utilized to identify the chromosomal positions of each DEG. To explore the biological roles of the DEGs, we utilized the “clusterProfiler” package in R to conduct gene ontology (GO) enrichment and kyoto encyclopedia of genes and genomes (KEGG) pathway analyses.

2.3. Machine Learning

To accurately identify CuDEGs specific to BD, we utilized three machine learning methods‐least absolute shrinkage and selection operator (LASSO), support vector machine (SVM), and random forest (RF)‐to narrow down the list of candidate genes for BD diagnosis. These methods were chosen for their complementary strengths:

The LASSO is a regression technique employed for regularization, improving prediction precision and model clarity by choosing pertinent variables while avoiding overfitting. It is particularly effective in high‐dimensional datasets, where feature selection is critical to prevent the inclusion of redundant or irrelevant genes.

The SVM is a robust classification algorithm, ideal for creating decision boundaries to distinguish between categories and providing high accuracy in binary classification tasks. Its strength lies in its ability to handle non‐linearly separable data by mapping it into a higher‐dimensional space using kernel functions, making it suitable for complex biological datasets. In our analysis, we used a radial basis function kernel and performed 10‐fold cross‐validation to optimize hyperparameters, specifically the cost (C) and gamma parameters, using a grid search strategy.

The RF, a powerful ensemble learning technique, is well‐suited for handling high‐dimensional data and is widely recognized for its high accuracy, sensitivity, and specificity in classification and variable selection tasks. As an ensemble method, RF mitigates overfitting by aggregating multiple decision trees, thus providing stable and interpretable results in gene selection. We built RF models with 500 trees (ntree = 500) and used the mean decrease in Gini index to evaluate variable importance. The number of variables randomly selected at each split (mtry) was optimized through fivefold cross‐validation using the caret R package.

By integrating these three techniques, we leveraged their respective advantages to enhance the robustness and reliability of CuDEG identification. LASSO ensures effective feature selection, SVM provides strong classification capabilities, and RF captures intricate patterns within high‐dimensional data. This multi‐method approach enhances the reliability of the findings, reducing the likelihood of false positives and ensuring that the selected genes are highly relevant for BD diagnosis.

Analyses using LASSO regression, SVM, and RF were conducted with the R packages “glmnet,” “e1071,” and “randomForest,” respectively. Hyperparameter tuning was carried out using built‐in functions such as cv. glmnet for LASSO, tune.svm for SVM, and train from the caret package for RF. To define “hub CuDEGs,” we cross‐referenced the genes identified by all three machine learning techniques using the “VennDiagram” R package. The overlap of these results was considered “hub CuDEGs,” as this intersection represents genes consistently identified by multiple, independent methods.

However, despite the advantages of these methods, there are certain limitations that should be acknowledged. LASSO, while effective in feature selection, can be sensitive to the regularization parameter (lambda), which may lead to inconsistent gene selection under different parameter settings. SVM, particularly with nonlinear kernels, can be computationally intensive and may require careful tuning of hyperparameters to achieve optimal performance. RF, although robust to overfitting, can sometimes produce less interpretable results due to the complexity of decision tree ensembles. Furthermore, all three methods rely on the quality and representativeness of the input data, meaning that any biases or limitations in the datasets could influence the results.

Line plots were created to illustrate the expression levels of hub CuDEGs between BD and healthy controls. Hub CuDEGs violin plots were created utilizing the “ggpubr” package in R. To assess the model's capability to differentiate BD from healthy individuals, a receiver operating characteristic (ROC) analysis of key CuDEGs was conducted utilizing the “pROC” R package. The GSE209567 data set was also used to validate the hub CuDEGs through violin plots and ROC analysis.

2.4. Subcluster Analysis With Hub CuDEGs

A hierarchical clustering analysis without supervision was conducted on 44 BD samples utilizing the “ConsensusClusterPlus” R package and the mRNA expression data of key CuDEGs. The k‐means method was utilized for a total of 1000 runs, with the cluster count capped at k = 9. The ideal cluster count was identified using the cumulative distribution function curve, the consensus matrix, and the cluster consistency score. The geometric distances between subclusters were visualized using a PCA plot. The distribution of these hub CuDEGs within the clusters was further analyzed using boxplots and heatmaps. The CIBERSORT algorithm was employed to evaluate the proportions of 22 distinct immune cell types in each cluster, with a p value of less than 0.05 signifying notable differences.

2.5. Exploration of the Ciological Roles of Hub CuDEGs

To explore the biological roles of CuDEGs in more detail, we performed GSEA and GSVA analyses utilizing the “clusterProfiler” and “GSVA” libraries in R. GSEA is a computational method that determines whether predefined sets of genes exhibit statistically significant differences between two biological states. In contrast, GSVA is a non‐parametric, unsupervised method that estimates variation in pathway activity over a sample population, allowing for sample‐specific enrichment analysis. These complementary methods were employed to ensure a comprehensive evaluation of the functional implications of CuDEGs. For our analysis, we utilized the predefined ontology gene sets “c5.go.Hs.symbols.gmt” and “c2.cp.kegg.Hs.symbols.gmt” from the MSigDB collection. The “c5.go” gene set focuses on GO categories, including biological processes, molecular functions, and cellular components, while the “c2.cp.kegg” gene set represents curated pathways from the KEGG database, offering insights into cellular signaling and metabolic processes. During the GSEA analysis, pathways were categorized based on their normalized enrichment score (NES) and adjusted p value. Pathways with a positive NES (NES > 0) and an adjusted p value below 0.05 were considered significantly enriched, indicating upregulated biological functions. Conversely, pathways with a negative NES (NES < 0) and an adjusted p value below 0.05 were identified as depleted, reflecting downregulated pathways or functions. Similarly, GSVA provided enrichment scores for each pathway at the sample level, enabling us to evaluate the variability in pathway activity across samples and identify key pathways associated with CuDEGs. The integration of GSEA and GSVA allowed us to identify critical biological pathways and processes related to CuDEGs, providing valuable insights into their roles in the pathogenesis of BD.

2.6. Immune Cell Analysis

Using the CIBERSORT algorithm with the LM22 signature matrix and gene expression data, the proportion of 22 different immune cell types in each sample was calculated. CIBERSORT employs monte carlo sampling to derive a P‐value for the deconvolution of each sample. Samples were deemed to have precise immune cell fractions only if P was less than 0.05, ensuring that the total of the 22 immune cell types in each sample equaled. Additionally, we analyzed the correlations between CuDEGs and hub CuDEGs with the 22 types of immune cells. Furthermore, the expression differences of hub CuDEGs in various immune functional cells were examined using 29 immune function‐related gene sets.

2.7. lncRNA‐miRNA‐mRNA CeRNA Network Construction

Interactions between miRNA and mRNA were predicted using the TargetScan (http://www.targetscan.org), miRDB (http://www.mirdb.org/), and miRanda (http://www.microrna.org/) databases, which provided complementary predictions to ensure robust identification of miRNA‐mRNA interactions. To enhance reliability, only interactions consistently identified by at least two of these databases were retained. miRNA‐lncRNA interactions were identified using the SpongeScan database (http://spongescan.rc.ufl.edu/). The identified miRNA‐lncRNA interactions were merged with the corresponding miRNA‐mRNA interactions to construct a ceRNA network. This integration ensured that the network effectively captures the regulatory relationships among lncRNAs, miRNAs, and mRNAs. Utilizing the “reshape2” and “ggalluvial” R packages, a ceRNA network was developed from these interactions. The diagram represented lncRNAs, miRNAs, and mRNAs as nodes and their interactions as edges, highlighting the regulatory dynamics within the ceRNA network. Specific colors were assigned to each node category to enhance visualization, and annotations were added to indicate key components of the network. This ceRNA network underscored the potential interplay between lncRNAs, miRNAs, and mRNAs, offering valuable insights into their roles in posttranscriptional regulation and the pathogenesis of BD.

3. Results

3.1. CRGs Expression in BD Patients

To investigate the biological roles of CRGs in BD, we analyzed the expression patterns of 52 CRGs in both BD patients and healthy controls utilizing the combined data set. Among these, 15 CuDEGs were identified, exhibiting significant differential expression between BD samples and controls. Specifically, the expression levels of ANKRD9, ARF1, ATOX1, BACE1, MAP1LC3A, MT1G, MT2A, MT4, NFE2L2, SNCA, and TYR were significantly upregulated in BD samples, while COX11, CP, LIAS, and SOD1 were downregulated (Figure 1A,B). To further explore their roles, we analyzed the chromosomal locations of the 15 CuDEGs (Figure 1C) and performed correlation analysis to identify potential interactions. Interestingly, synergistic effects were observed between some CuDEGs, such as MT1G and MT2A, which are closely linked to metal ion homeostasis. In contrast, significant antagonistic effects were noted between COX11 and MAP1LC3A (Figure 1D,E), suggesting distinct regulatory mechanisms in BD pathogenesis.

Figure 1.

CRGs Expression in BD Patients. (A) Boxplots show the expression levels of 52 CRGs between BD patients and healthy controls. Red asterisks indicate statistically significant differences. (B) Heatmap illustrates the expression patterns of the 52 CRGs across all samples. (C) The chromosomal locations of 15 CuDEGs identified between BD and control samples. These CuDEGs are potential key regulators in BD pathogenesis. (D) Circular plot showing pairwise correlations among the 52 CRGs; the thickness and color of the lines represent correlation strength and direction. (E) Corrplot of Pearson correlation coefficients among CRGs; positive and negative correlations are indicated by red and green, respectively. (F) Bubble plot of Gene Ontology (GO) enrichment analysis of the 15 CuDEGs, showing enriched biological processes related to metal ion response and oxidative stress. (G) Circle plot of GO enrichment terms with gene counts and significance indicated. (H) Bubble plot of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the 15 CuDEGs, highlighting pathways such as mineral absorption and ferroptosis. CuDEGs refer to CRGs that showed significant differential expression between BD and healthy controls, suggesting their potential involvement in BD‐related immune dysregulation, oxidative stress, and mitochondrial dysfunction. (***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05).

Functional enrichment analysis of the 15 CuDEGs provided deeper insights into their potential roles. GO analysis revealed that the most enriched biological processes included the response to copper ions (GO:0046688; adjusted p < 0.001), cellular response to copper ions (GO:0071280; adjusted p < 0.001), and cellular response to metal ions (GO:0071248; adjusted p < 0.001). Visualization of these GO terms was conducted using bubble plots (Figure 1F) and circle plots (Figure 1G). These results highlight the importance of copper ion regulation in BD, suggesting that disturbances in copper homeostasis may contribute to disease progression. KEGG pathway analysis identified significant enrichment in mineral absorption (hsa04978; adjusted p = 0.003) and ferroptosis (hsa04216; adjusted p = 0.03) pathways (Figure 1H). Ferroptosis, an iron‐ and ROS‐dependent cell death process, has been implicated in various inflammatory and autoimmune diseases.

These findings suggest that CuDEGs play crucial roles in BD development, potentially by regulating metal ion homeostasis, oxidative stress, and cell death pathways. For instance, NFE2L2 encodes NRF2, a master regulator of antioxidant defense, and its upregulation in BD suggests a compensatory response to oxidative stress [25]. Taken together, these results provide evidence for copper‐related molecular mechanisms underlying BD pathogenesis and identify potential targets for further investigation.

3.2. Identification of Hub CuDEGs via Machine Learning

We employed the LASSO regression, SVM, and RF algorithms to identify hub CuDEGs (Figure 2A–D). This analysis ultimately identified six hub CuDEGs: ANKRD9, COX11, MT1G, MT2A, MT4, and TYR (Figure 2E). Figure 2F depicts the expression levels of these six hub CuDEGs in both BD patients and healthy individuals. Most CuDEGs, except for COX11, are highly expressed in BD patients (Figure S1). To evaluate their diagnostic potential, we evaluated the ROC curves and AUC values for each hub CuDEGs. Among them, MT1G showed the highest AUC (0.739), indicating its strong ability to distinguish BD patients from healthy controls. The AUC values for the other hub CuDEGs were as follows: COX11 (0.726), ANKRD9 (0.697), MT4 (0.707), MT2A (0.690), and TYR (0.650) (Figure S2). These results demonstrated that all six hub CuDEGs have significant diagnostic value for BD. Additionally, these findings were validated using the GSE209567 data set, further supporting their robustness and reliability (Figures S3 and S4).

Figure 2.

Machine learning identifies hub CuDEGs. (A–D) Construction of cuproptosis signatures using LASSO regression, SVM, and RF algorithms. (E) Venn diagram showing the overlap of candidate genes identified by the three algorithms. (F) Line plots displaying the expression levels of hub CuDEGs.

The biological relevance of these hub CuDEGs in BD pathogenesis is noteworthy. MT1G and MT2A are metallothioneins that regulate metal ion metabolism and have been associated with oxidative stress resistance. Their upregulation in BD suggests a potential protective mechanism against copper‐induced oxidative stress [26, 27]. COX11, a cytochrome c oxidase assembly factor, is involved in mitochondrial function, and its downregulation may contribute to mitochondrial dysfunction, a known factor in autoimmune diseases [28, 29]. ANKRD9 has been linked to lipid metabolism and inflammation, suggesting that metabolic dysregulation may be a contributing factor in BD [30, 31]. Additionally, TYR encodes tyrosinase, an enzyme involved in melanin biosynthesis, and has been associated with immune modulation in various diseases [32]. Together, these findings illustrate that each hub CuDEG is mechanistically linked to known or suspected pathogenic processes in BD. The identification of these hub genes provides mechanistic insights into BD pathology, highlighting the interplay between copper metabolism, oxidative stress, mitochondrial function, and immune responses.

When compared with other known BD biomarkers, such as inflammatory cytokines (e.g., IL‐6, TNF‐α) and acute‐phase proteins (e.g., C‐reactive protein), the AUC values of the six hub CuDEGs are comparable or slightly lower in some cases. However, the CuDEGs identified in this study provide a unique perspective by reflecting copper‐related molecular mechanisms, which are less explored in BD pathogenesis. These findings underscore the diagnostic potential of the six hub CuDEGs identified through machine learning. By elucidating their roles in BD‐related pathways, such as ferroptosis, oxidative stress regulation, and immune response modulation, they may serve as supplementary biomarkers for clinical diagnosis.

3.3. Consensus Clustering Analysis of Six Hub CuDEGs Clusters

The “Consensus Cluster Plus” package in R was utilized for consensus clustering analysis. Based on the expression levels of the six key CuDEGs, we determined that k = 2 yielded the most consistent clusters (Figure 3A). As a result, the 44 BD samples from the GEO database were classified into two groups: cluster 1 (low‐expression group, n = 33) and cluster 2 (high‐expression group, n = 11). The PCA plot demonstrated distinct gene expression patterns between these clusters (Figure 3B).

Figure 3.

Identification of hub CuDEGs clusters in BD. (A) Consensus clustering matrix when k = 2. (B) The PCA plot of subclusters. (C) The expression patterns of six hub CuDEGs were presented in the heatmap. (D) Boxplots showed the expression of six hub CuDEGs between two cuproptosis clusters. (E) The relative abundance of immune cells between two subclusters. (F) Boxplots showed the difference of immune infiltration in two subclusters. (***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05).

To further investigate the biological differences between the two subclusters, heatmaps and boxplots (Figure 3C,D) were used to visualize the expression levels of the six hub CuDEGs. Notably, ANKRD9, MT1G, MT4, and TYR exhibited significantly higher expression in cluster 2 compared to cluster 1, whereas COX11 and MT2A did not display this trend. These findings suggest that the two subclusters represent distinct molecular subtypes of BD.

Immune cell infiltration analysis provided additional insights into the biological differences between these clusters. Cluster 2, characterized by higher expression of key CuDEGs, exhibited significantly increased levels of M0 macrophages and neutrophils (p < 0.05) compared to cluster 1 (Figure 3E,F). This differential immune cell infiltration may reflect distinct immune response mechanisms in the two subclusters. Elevated neutrophil levels in cluster 2 align with the known role of neutrophils in BD pathogenesis, suggesting a more pronounced inflammatory state in this subgroup.

Clinically, these findings have potential implications. Patients in cluster 2 may represent a subgroup with more severe inflammation, requiring more aggressive anti‐inflammatory therapies. Conversely, cluster 1, with lower CuDEG expression and immune cell infiltration, may correspond to a less severe or early‐stage phenotype of BD. Future studies are warranted to validate these subclusters and explore their potential as a basis for stratified therapeutic approaches in BD management.

3.4. GSEA and GSVA of Six Hub CuDEGs

To better understand the role of CuDEGs in the pathogenesis of BD, we performed GSEA and GSVA analyses to identify the associated signaling pathways and biological processes (Figures S5–S9). The results highlighted several significant pathways and processes that may contribute to BD progression.

In the GSEA analysis, multiple CuDEGs were found to be involved in immune regulation and oxidative stress, which are critical in BD pathogenesis. For example, ANKRD9 was overexpressed in the complement and coagulation cascades (KEGG: adjusted p = 0.004), a pathway known to mediate inflammation and immune responses in BD. Similarly, TYR was associated with the cytosolic DNA‐sensing pathway (KEGG: adjusted p = 0.01), which plays a pivotal role in innate immune activation. Conversely, COX11 and MT1G were significantly underexpressed in mitochondrial gene expression and the spliceosome (KEGG: adjusted p < 0.001), suggesting a potential link to mitochondrial dysfunction and impaired cellular homeostasis in BD.

In the GSVA analysis, pathways related to immune function and metabolic regulation were enriched. For example, MT1G and MT2A was significantly associated with pathways such as valine, leucine, and isoleucine biosynthesis (KEGG) and the renin‐angiotensin system (KEGG), highlighting their potential roles in metabolic regulation and inflammation. Furthermore, COX11 showed significant enrichment in the calcium signaling pathway (KEGG), which is implicated in immune cell activation and cytokine production, both of which are relevant to BD.

These findings suggested that the six hub CuDEGs contributed to BD pathogenesis through diverse mechanisms, including dysregulation of immune signaling pathways, such as the complement cascade, cytosolic DNA sensing, and calcium signaling, which might have driven immune dysfunction; impaired mitochondrial function and spliceosome activity, which led to cellular stress and inflammation; and altered metabolic processes, such as amino acid biosynthesis and oxidative stress pathways, which could have further exacerbated immune dysregulation and oxidative damage in BD. Integrating the results from GSEA and GSVA analyses underscored the importance of these pathways in BD and highlighted the potential of CuDEGs as therapeutic targets or biomarkers for disease management. Future studies were recommended to validate these findings in larger cohorts and to investigate their clinical implications.

3.5. Immune Cells Analysis

We performed an analysis of immune infiltration to contrast the immune systems of healthy individuals with those of BD patients. The findings indicated that BD patients had lower levels of infiltration of resting mast cells compared to healthy controls (Figure 4A,B), suggesting a significant association between BD development and the immune system. Mast cells are known to be involved in inflammatory and autoimmune diseases by releasing histamines and cytokines that modulate immune responses. Reduced resting mast cell levels in BD patients could indicate chronic immune activation, contributing to persistent inflammation and tissue damage, which are hallmarks of BD.

Figure 4.

Immune cells analysis. (A) The relative abundance of immune cells between BD and Control. (B) Boxplots showed the difference of immune infiltration in BD and control. (C) The analysis of immune‐related functional cells in ANKRD9. (D) The analysis of immune‐related functional cells in COX11. (E) The analysis of immune‐related functional cells in MT1G. (F) The analysis of immune‐related functional cells in MT2A. (G) The analysis of immune‐related functional cells in MT4. (H) The analysis of immune‐related functional cells in TYR. (***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05).

Next, we examined differences in 29 immune‐related functional cells between high and low expression groups of the six hub CuDEGs (Figure 4C–H). Notably, within the group exhibiting low ANKRD9 gene expression, there was heightened activity of activated dendritic cells (aDCs), B cells, pro‐inflammatory cells, neutrophils, natural killer (NK) cells, T helper cells, tumor‐infiltrating lymphocytes (TIL), and cells involved in the type I interferon (IFN) response. Increased neutrophil activation is particularly relevant, as BD is characterized by neutrophilic hyperactivity, leading to excessive inflammation and tissue damage. Neutrophils play a key role in BD‐related vasculitis, and their elevated activity in patients with low ANKRD9 expression suggests that ANKRD9 may influence neutrophil‐mediated inflammatory responses.

On the other hand, cells involved in antigen‐presenting cell (APC) costimulation, mast cells, and those responding to type II IFN were more active in the group with elevated ANKRD9 gene expression. In the context of the COX11 gene, the low expression group exhibited activity in dendritic cells (DCs), mast cells, neutrophils, T follicular helper (TFH) cells, and Type II IFN response cells, while the high expression group showed activity in T cell co‐inhibition cells, T cell costimulation cells, Th1 cells, TIL, and regulatory T cells. Regulatory T cells (Tregs) play a crucial role in maintaining immune tolerance, and their increased activity in BD patients with high COX11 expression suggests a compensatory response to control chronic inflammation. This finding is particularly relevant, as dysregulated Treg function has been implicated in BD pathogenesis.

APC costimulation cells, chemokine (C‐C motif) receptor (CCR) cells, and DCs were more active in the high expression group for the MT1G gene. Likewise, aDCs, parainflammation, type I IFN response cells, and type II IFN response cells exhibited increased activity in the group with elevated MT2A gene expression. Parainflammation, a low‐grade chronic inflammatory state, has been increasingly recognized in BD, particularly in the progression of vascular inflammation and mucocutaneous lesions. The association between MT2A and parainflammation suggests that MT2A might influence chronic inflammatory processes in BD.

In the group with low MT4 gene expression, immature dendritic cells (iDCs) exhibited higher activity, whereas in the group with elevated MT4 gene expression, Tfh cells showed increased activity. Tfh cells play a key role in B cell activation and antibody production. Their increased activity in BD patients with high MT4 expression could contribute to autoantibody production, a feature observed in BD patients with systemic involvement.

Subsequently, we performed an immune infiltration analysis specifically on the six hub CuDEGs (Figure 5A). The findings revealed strong inverse relationships between B cell memory and the ANKRD9 gene (Figure 5B, Figure S10A), along with notable positive associations between macrophages M0, macrophages M1, neutrophils, and the ANKRD9 gene (Figure 5C–E, Supporting Figure S10A). Macrophage polarization is critical in autoimmune diseases, with M1 macrophages promoting pro‐inflammatory responses and M2 macrophages contributing to tissue repair. The positive correlation between ANKRD9 and M1 macrophages suggests that ANKRD9 might be involved in the persistent inflammatory response observed in BD.

Figure 5.

Immune infiltration cells analysis. (A) The correlation analysis between 15 CuDEGs and immune infiltration cells. (B–E) ANKRD9. (F–J) COX11. (K) MT1G. (L) MT4. (M) TYR. (***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05).

There were significant positive correlations between activated DCs, T cells CD4 naïve, and the COX11 gene (Figure 5F,J, Figure S10B), as well as significant negative correlations between macrophages M0, macrophages M1, neutrophils, and the COX11 gene (Figure 5G–I, Supporting Figure 10B). A strong positive relationship was detected between M1 macrophages and the MT1G gene (Figure 5K, Figure S10C), whereas no notable associations were identified between immune cells and the MT2A gene (Figure S10D). A notable positive relationship was observed between M0 macrophages and the MT4 gene (Figure 5L, Figure S10E), while a marked negative association was found between naïve CD4 T cells and the TYR gene (Figure 5M, Figure S10F).

The observed immune cell associations provide potential therapeutic implications. Given the significant role of neutrophils in BD‐related vasculitis and tissue injury, targeting neutrophil activation—potentially through modulating ANKRD9 expression—could be explored as a therapeutic strategy. Similarly, the involvement of Tregs and Th1 cells suggests that immunomodulatory therapies aimed at restoring T cell balance may benefit BD patients. These findings highlight the need for future studies to investigate whether targeting CuDEGs can modulate immune responses and alleviate BD symptoms.

3.6. CeRNA Network Construction of Six Hub CuDEGs

A CeRNA network was developed by integrating data from the miRanda, TargetScan, miRDB, and SpongeScan databases, including six key CuDEGs. The resulting network comprises 33 nodes, including 3 hub CuDEGs, 8 miRNAs, and 22 lncRNAs (Figure 6). Specifically, 8 lncRNAs were identified as competitively binding with TYR, regulated by hsa‐miR‐1208, hsa‐miR‐27a‐5p, and hsa‐miR‐450b‐5p. Furthermore, 11 long noncoding RNAs were identified to interact with COX11, influenced by hsa‐miR‐141‐3p, hsa‐miR‐7‐5p, and hsa‐miR‐770‐5p. Furthermore, 6 lncRNAs were discovered to target MT1G, regulated by hsa‐miR‐27a‐3p and hsa‐miR‐944.

Figure 6.

lncRNA‐miRNA‐mRNA regulatory network.

4. Discussion

This study provided new insights into the role of CuDEGs in BD pathogenesis by integrating transcriptomic data, immune infiltration analysis, and bioinformatics approaches. The identification of 15 CuDEGs, with distinct expression profiles between BD patients and healthy controls, highlighted the central role of copper‐related molecular pathways in immune regulation, oxidative stress, and mitochondrial function. Among these, six hub CuDEGs—ANKRD9, COX11, MT1G, MT2A, MT4, and TYR—were identified using machine learning techniques, which demonstrated significant diagnostic value. These findings suggested that CuDEGs were not only biomarkers of disease but also key regulators in BD development.

Differential immune infiltration patterns observed in this study emphasized the importance of immune regulation in BD. Elevated neutrophils and M0 macrophages in BD patients aligned with the known inflammatory nature of the disease, suggesting that CuDEGs might have modulated the activation and recruitment of innate immune cells. Neutrophilic hyperactivation has been widely reported in BD, contributing to endothelial damage and excessive inflammatory responses. Similarly, M0 macrophages, which can differentiate into either pro‐inflammatory (M1) or anti‐inflammatory (M2) subtypes, play a critical role in chronic inflammation in BD [33]. For instance, the positive correlations between ANKRD9 and M0/M1 macrophages, and between MT4 and M0 macrophages, further supported their involvement in promoting inflammatory responses. Conversely, COX11 and TYR showed inverse correlations with key immune cell types, indicating their potential roles in maintaining immune homeostasis.

Copper plays a pivotal role in maintaining mitochondrial function, influencing metabolic processes, mitophagy, and cellular signaling, with significant implications for various diseases, including cancer [34]. As a regulator of cell signaling pathways, copper modulates critical interactions in processes such as phosphorylation [35]. Additionally, copper serves as an essential cofactor for enzymes involved in mitochondrial respiration, antioxidant defense, and maintaining connective tissue integrity. Proper copper homeostasis is crucial to prevent cellular damage caused by oxidative stress, underscoring its importance in both normal physiology and disease pathogenesis [36]. Notably, previous studies have demonstrated that oxidative stress is a key driver of BD pathology, leading to endothelial dysfunction and immune system dysregulation [37]. Our study further supports this concept by revealing that altered CuDEG expression is closely linked to oxidative stress‐related pathways, such as ferroptosis and mitochondrial dysfunction.

Elevated copper levels have been observed in patients with BD, especially in severe cases [13, 14, 15]. This increase is linked to higher oxidative stress markers, contributing to inflammation and tissue damage [14, 15]. Copper dysregulation exacerbates oxidative stress by promoting ROS and weakening antioxidant defenses [14, 15, 16]. These findings highlight copper's role in BD pathogenesis and its potential as a therapeutic target. Consistent with our findings, earlier research has shown that patients with BD exhibit upregulation of oxidative stress‐related genes and downregulation of antioxidant mechanisms, leading to an imbalance that favors inflammatory responses [38]. Antioxidant therapies to normalize copper levels or counteract its effects may benefit BD patients [16, 18].

The altered expression of hub CuDEGs identified in this study likely reflected dysregulated copper homeostasis, which could have exacerbated oxidative stress and inflammation in BD. For example, ANKRD9 influenced metabolic pathways, lipid processing, and tumor suppression by modulating protein interactions via the ubiquitin ligase complex [30, 31] ANKRD9's significant overexpression in secretory granule membranes and transcription factor binding, as well as its positive correlation with M0/M1 macrophages and neutrophils, suggested a role in immune activation and inflammation. The role of COX11 in mitochondrial copper transport linked it directly to oxidative stress and energy metabolism, both of which were dysregulated in BD [28, 29] Similarly, MT1G and MT2A, known for their roles in metal ion detoxification and oxidative stress regulation, were associated with pathways such as ferroptosis and the RIG‐I‐like receptor pathway, suggesting their contribution to both cell death and immune response regulation [26, 27] MT4, known for its role in tissue homeostasis and cancer progression [39], showed increased oxygen‐binding activity and positive associations with M0 macrophages, implicating it in inflammatory regulation. TYR, essential for melanin production through pathways like STAT3 and CREB‐MITF [32], demonstrated upregulation in innate immune responses and DNA‐sensing pathways, coupled with an inverse correlation with naïve CD4 T cells, highlighting its immunomodulatory potential. These findings reinforced the hypothesis that copper dysregulation was a driving factor in BD pathogenesis, offering novel avenues for therapeutic intervention.

Using the six hub CuDEGs, unsupervised clustering analysis identified two distinct clusters linked to cuproptosis, revealing molecular heterogeneity in BD. Cluster 2 showed increased M0 macrophages and neutrophils compared to cluster 1, indicating different immunological landscapes and potential pathogenic mechanisms. Elevated levels of M0 macrophages and neutrophils in cluster 2 align with previous findings highlighting the role of innate immune cells in BD development [40]. Macrophages exhibit plasticity, influencing inflammation, while neutrophils, as first responders, contribute to tissue damage in BD [41]. Additionally, BD patients had lower resting mast cell infiltration compared to healthy controls, suggesting an immune imbalance linked to BD pathology [42].

To further explore regulatory mechanisms, a CeRNA network was developed using three hub diagnostic indicators, providing insights into interactions among noncoding RNAs, mRNAs, and key markers. For instance, hsa‐miR‐27a‐5p and hsa‐miR‐141‐3p were identified as key regulators of TYR and COX11, respectively, suggesting potential targets for RNA‐based therapies. RNA‐based therapeutic approaches have shown promise in modulating immune responses in autoimmune diseases [43], and our findings suggest that similar strategies could be explored for BD. Targeting these regulatory networks could modulate CuDEG expression and downstream effects, paving the way for innovative treatment strategies. These findings emphasize BD's complexity and highlight the value of integrating machine learning and bioinformatics to understand disease mechanisms and identify therapeutic targets [44]. By linking CuDEGs to immune dysregulation and oxidative stress, our study provides a new perspective on BD pathogenesis and potential intervention strategies. Identifying distinct clusters and analyzing immune cell infiltration offer deeper insights into BD's immunopathogenesis, paving the way for more targeted and effective treatments.

This study had several constraints that should be addressed. First, our findings were primarily based on bioinformatics analyses and relied on data from public databases, which restricted access to the original sequencing data. Publicly available datasets often lack detailed metadata, including patient‐specific clinical records, treatment history, and disease severity scores, which may introduce confounding variables that could not be controlled in our study. Consequently, we were unable to provide detailed demographic and clinical characteristics of the included samples, which may have introduced selection bias. Additionally, the relatively small sample size, particularly for specific racial groups, may have reduced the robustness of our conclusions and limited the generalizability of our findings to the broader population. The underrepresentation of certain racial or ethnic groups in transcriptomic databases could lead to biased gene expression signatures, potentially limiting the applicability of our findings across diverse populations. Future studies should include larger, multi‐center cohorts with greater diversity to address these limitations. Furthermore, the datasets used may not fully represent the genetic, environmental, and lifestyle diversity across populations. Variations in these factors, along with genetic ancestry, could influence the expression levels of CuDEGs and related pathways, potentially impacting the observed associations. While racial differences and other potential confounders could significantly affect gene expression and disease mechanisms, this study lacked the necessary data and sample diversity to perform statistical adjustments for these factors. Future research should aim to collect datasets with more detailed demographic information and utilize advanced statistical methods, such as stratified analysis or multivariate regression, to better account for these variables.

This study adhered to ethical guidelines; however, ensuring equitable representation of diverse racial and ethnic populations requires further consideration, including appropriately tailored informed consent procedures. Potential biases introduced by racial group differences in datasets could also impact the results. To minimize these biases, future research should implement strategies such as blinding data analyses and using objective outcome measures.

Future research directions include investigating the molecular mechanisms by which CuDEGs regulate immune signaling, oxidative stress, and mitochondrial function in BD. In particular, experimental validation should be conducted to confirm the functional roles of the identified hub CuDEGs, such as using qRT‐PCR, Western blot, and immunohistochemistry to confirm differential expression and localization in patient‐derived samples. Functional studies using siRNA knockdown and CRISPR‐Cas9 in immune cell models can help elucidate their impact on inflammatory cytokine production and oxidative stress. Additionally, investigating copper homeostasis, oxidative stress levels, and mitochondrial function through specialized assays will provide mechanistic insights into CuDEG‐mediated pathways. From a therapeutic perspective, targeting CuDEGs via small‐molecule inhibitors, RNA‐based therapies, or immunomodulatory approaches could offer novel strategies for BD treatment. Moreover, exploring RNA‐based therapies or small molecules targeting CuDEG‐related pathways could provide new insights into disease mechanisms and potential therapeutic approaches. Finally, datasets with detailed environmental and lifestyle information should be incorporated to better assess their impact on disease progression. A systematic integration of clinical, genetic, and environmental data will be crucial in developing a precision medicine approach for BD.

5. Conclusion

Using comprehensive bioinformatics analyses, this study revealed significant associations between CRGs and immune cell infiltration, demonstrating the diversity of immune responses among BD patients with distinct cuproptosis subtypes. Six hub CuDEGs—ANKRD9, COX11, MT1G, MT2A, MT4, and TYR—were identified as potential diagnostic markers, showing strong ability to distinguish BD subtypes. This study is the first to explore the role of cuproptosis in BD, providing new insights into its involvement in immune regulation, oxidative stress, and mitochondrial function. These findings contribute to a deeper understanding of BD pathogenesis and offer directions for future research.

Author Contributions

Si Chen: conceptualization, data curation, writing – review and editing, investigation, supervision, writing – original draft, visualization, validation, methodology, software. Rui Nie: writing – review and editing, visualization. Yan Wang: visualization. Haixia Luan: writing – review and editing. Chao Wang: writing – review and editing. Yuan Gui: writing – review and editing. Xiaoli Zeng: conceptualization, funding acquisition, writing – review and editing. Hui Yuan: conceptualization, writing – review and editing, funding acquisition.

Ethics Statement

This study was conducted using publicly available datasets obtained from the Gene Expression Omnibus (GEO) repository. The Institutional Review Board (IRB) of Beijing Anzhen Hospital, Capital Medical University reviewed the study and granted a waiver of ethical approval and participant consent due to the use of secondary, deidentified data. The study complies with all relevant ethical regulations regarding the use of human participant data.

Conflicts of Interest

The authors declare no conflicts of interest.

Transparency Statement

The lead author Xiaoli Zeng, Hui Yuan affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Supporting information

Supporting Figure 1.

Supporting Figure 2.

Supporting Figure 3.

Supporting Figure 4.

Supporting Figure 5.

Supporting Figure 6.

Supporting Figure 7.

Supporting Figure 8.

Supporting Figure 9.

Supporting Figure 10.

Supporting Table 1.

supmat.

Data Availability Statement

This study utilized publicly available datasets (GSE17114 and GSE209567), which were obtained from the gene expression omnibus (GEO) repository (https://www.ncbi.nlm.nih.gov/geo/). The datasets analyzed during the current study are available from the corresponding authors upon reasonable request.