Identification of potential blood biomarkers of coronary artery disease using a cuproptosis gene set

Abstract

Purpose

To investigate whether cuproptosis-related genes contribute to coronary artery disease (CAD) pathogenesis and to develop a robust, blood-based diagnostic model.

Patients and methods

Whole-blood transcriptome profiles (GSE180081, GSE180082) were retrieved from the GEO database. After batch-effect correction (limma::removeBatchEffect) and quantile normalization, differentially expressed genes (DEGs) between CAD patients (n = 521) and controls (n = 191) were identified with FDR < 0.05 and |log2FC|≥ 1. Consensus clustering (ConsensusClusterPlus, k = 2) on 19 cuproptosis genes stratified patients into high- and low-cuproptosis activity groups. DEGs between these clusters were intersected with the CAD-DEG list to yield 818 cuproptosis-linked DEGs. A five-gene diagnostic signature (HIST1H4E, IL6ST, LST1, RN7SKP45, and SNORD50B) was selected by LASSO regression and modeled with logistic regression. Immune infiltration, ceRNA networks, and druggability were further analyzed. Local RT-qPCR in an independent cohort (12 CAD, 12 controls) confirmed expression trends.

Results

We identified 818 differentially expressed genes that were common to the CAD and cuproptosis gene sets, and these principally represented the cell–substrate junction and the positive regulation of leukemia. Furthermore, HIST1H4E, IL6ST, RN7SKP45, LST1, and SNORD50B were found to be potentially useful for the diagnosis of CAD using our diagnostic model. These genes were also found to be closely associated with immune modification. Further validation revealed that HIST1H4E, IL6ST, and LST1 are very likely to be potential biomarkers.

Conclusion

We constructed a diagnostic prediction model based on cuproptosis-related genes using whole-blood transcriptome data. Our results identify HIST1H4E, IL6ST, and LST1 as potential biomarkers for CAD risk assessment. These findings provide a novel basis for the prediction, prevention, and individualized treatment of CAD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40001-025-03276-x.

Introduction

Coronary artery disease (CAD) represents the leading cause of global mortality and the primary cause of death in developing countries [1]. The condition predominantly arises from coronary atherosclerosis and acute vascular occlusion [2], with pathogenesis involving sequential pathological processes: lipid infiltration [3, 4], oxidative stress-mediated endothelial damage [5], inflammatory activation, vascular smooth muscle cell proliferation, and progressive fibrotic remodeling culminating in atherosclerotic plaque formation [6]. Plaque progression may cause critical luminal stenosis or trigger acute thrombosis through rupture/erosion, ultimately resulting in myocardial ischemia and irreversible necrosis [7].

Currently, aspirin, statins, and β-blockers are widely used for primary and secondary prevention [8]. Severe cases require stent implantation or coronary artery bypass grafting (CABG) [9]. However, long-term medication use and repeated interventions significantly impact patients' quality of life, highlighting the urgent need for blood-based biomarkers and robust predictive models capable of early identification of high-risk individuals [10], 11.

Copper serves as a cofactor for numerous critical enzymes [12], and disruption of its homeostasis can trigger immune dysregulation [13, 14]. Cuproptosis—a recently discovered form of programmed cell death—is driven by copper ion overload within the mitochondrial tricarboxylic acid (TCA) cycle. This leads to the aggregation of acylated proteins and downregulation of iron-sulfur cluster proteins, subsequently inducing proteotoxic stress and cell death [15]. Given the central role of the immune response in CAD and the close link between cuproptosis and immunity, we hypothesized that cuproptosis-related genes may participate in CAD pathogenesis and serve as potential biomarkers.

Given copper's established roles in mitochondrial metabolism and endothelial dysfunction, we hypothesized that cuproptosis-related genes contribute to CAD pathogenesis. To test this, we analyzed whole-blood transcriptomes from GEO datasets (GSE180081/GSE180082), identifying differentially expressed genes (DEGs) between CAD patients (n = 521) and controls (n = 191). Consensus clustering using 19 cuproptosis genes stratified patients into high- and low-cuproptosis activity subgroups. The subsequent intersection of subgroup-specific DEGs with clinical phenotypes yielded 818 cuproptosis-associated CAD genes. Through LASSO regression feature selection, we constructed a three-gene diagnostic signature (HIST1H4E, IL6ST, LST1), which demonstrated robust diagnostic performance in both an external validation cohort (GSE12288) and local RT-qPCR analysis of patient-derived monocytes.

Material and methods

Data collection

The gene expression profiles GSE180081 and GSE180082 were downloaded from the GEO database. GSE180081 comprises 96 clinical samples, 48 with CAD and 48 controls, and contains 44,265 unique genes; GSE180082 comprises 80 clinical samples, 43 with CAD and 37 controls, and contains 44,265 unique genes. GSE180081 was used as a training set, and GSE180082 was used as a validation set. In addition, we used the Drug Gene Interaction Database (DGIdb) to identify potential gene-targeting drugs [16].

Human blood collection

CAD patients were recruited preoperatively from coronary artery bypass grafting candidates at Xiangya Hospital. Venous blood samples were collected prior to the operation from CAD patients who were undergoing coronary artery bypass grafting. Controls were age-/sex-matched healthy volunteers from the same hospital’s Health Management Center. Informed consent was provided by all the participants, and the study was approved by the Ethics Committee of Xiangya Hospital, Central South University. Exclusion criteria included previous cardiac surgery, previous percutaneous coronary intervention, rheumatic heart disease, systemic lupus erythematosus, or other connective tissue or immune-mediated diseases, and the following data were collected at baseline: age; sex; and history of smoking, hypertension, hyperlipidemia, diabetes, and cerebral infarction.

Data clustering

After obtaining the cuproptosis gene dataset [15], ConsensusClusterPlus [17] was applied to the training set (GSE180081) using the 19 cuproptosis genes. Euclidean distance and k-means (k = 2–6) [18] were used, with optimal k = 2 determined by CDF stability (Supplementary Fig. S1). An item-consensus heatmap (Supplementary Fig. S1) shows clear separation between Cu_group1 and Cu_group2. The optimal result of the cluster analysis was then used to divide the cohort into two groups for subsequent analysis.

Differential expression analysis

The downloaded gene expression matrix (RPKM) was converted to TPM data, and the GSE180081 dataset was divided into CAD and nonCAD groups. All the CAD samples were grouped according to the clustering results, and differential expression analysis was conducted using the limma package in R (4.4.1) [19]. DEGs were identified using limma with P < 0.05 and |log2FC|> 1 (Benjamini–Hochberg method) [20]. We used the ggplot2 package [21] to construct volcano plots and heat maps to display the differentially expressed genes.

Functional enrichment analysis

To identify the potential functional pathways that include the differentially expressed genes, we used the clusterprofiler and fgsea packages in R 4.3.3 to search the Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) databases and conduct functional and GSEA enrichment analysis [17, 22, 23]. GSEA used MSigDB Hallmark gene sets (v7.5), with normalized enrichment scores (NES) and false discovery rate (FDR) reported (F3 C).

Identification and validation of diagnostic cuproptosis-related genes for CAD

We analyzed the intersection between the genes that were differentially expressed between the CAD and nonCAD groups and those in the cuproptosis clusters in the training set, a total of 818 genes. The least absolute shrinkage and selection operator (LASSO) in the glmnet [24] R package was applied to the overlapped differentially expressed genes in order to reduce the number of dimensions in the data and identify genes related to a CAD diagnosis [25]. Genes obtained from the LASSO regression screening were then used to construct a diagnostic model by means of logistic regression [26, 27]. In the subsequent analysis, we conducted a more in-depth study of the encoded proteins. We used the pROC [28] package in R software to conduct predictions with the training set [29] and evaluated the diagnostic ability of the model generated using receiver operating characteristic (ROC) analysis. We then evaluated the model’s diagnostic performance by applying it to the GSE180082 dataset. Models are considered potentially useful predictors when the area under the ROC curve (AUC) is > 0.60 and good predictors when the AUC is > 0.75) [30, 31].

Immune infiltration analysis

To explore the differences between the subgroups in the diagnostic model, we divided the training set into a high-risk group and a low-risk group on the basis of the median risk score from the model. Next, we used GSVA single-sample gene set enrichment analysis (ssGSEA) [32] to identify the differences in immune cells and immune function between groups and used the Cibersort [33] package to analyze the representation of 22 types of immune cells in the high-risk group [34, 35]. Finally, we employed the ESTIMATE algorithm [36] to predict the immune status of and analyze the differences in the levels of immune cell infiltration between the high-risk and low-risk groups.

Prediction of the effects of immunomodulation

To predict the effects of immunomodulation, we first combined the data regarding immunosuppressants, immune activators, and MHC-related genes in the TISIDB database [37] with the expression matrix of the CAD patients, then conducted Pearson correlation analysis of the relationships between these biomarkers using the corrplot [38] package in R. Specifically, correlation analysis was performed for the relationships between these two sets of genes, and one-way linear regression analysis was performed for immune loci with close correlations.

Single gene set enrichment analysis (ssGSEA)

GSEA was used to identify the set of genes with concordant differences from the expression matrix of all the identified genes, and we chose this method because it permits genes with even small differences in expression to be taken into account [39]. To identify the function of the genes used to construct the diagnostic model, single-gene GSEA analysis was performed on the GSE180081 dataset using the clusterProfiler package [40].

ceRNA network construction

The MicroCosm database was searched using the multiMiR package [41] to identify micro (mi)RNAs that are predicted to target the identified genes of interest. In addition, a search for long non-coding (lnc)RNAs associated with the identified miRNAs was conducted using the starBase database [42], with all the node information being collated in Cytoscape (v3.9.0) [43] for visualization.

mRNA expression of the genes of interest

Peripheral blood was collected from 12 participants in the CAD and nonCAD groups, and the peripheral blood mononuclear cells were isolated and RNA extracted using an RNA Extraction Kit (Accurate Biology, AG21023, Changsha, China). We then reverse-transcribed the RNA to cDNA using a Reverse Transcription Kit (Accurate Biology, AG11728, Changsha, China) and analyzed target gene expression using real-time quantitative fluorescence polymerase chain reaction (RT-PCR, Genecopoeia, QP001, Changsha, China). All the procedures were conducted according to the kit manufacturers’ instructions.

Statistical analysis

All statistical analysis was performed using R software version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria) [44]. The limma [19] package was used to identify differentially expressed genes, and ROC curve analysis was performed using the pROC [28] package. Pearson correlation analysis was used to evaluate the relationships between variables, and differences were regarded as statistically significant when P < 0.05.

Results

We identified differential genes between CAD and controls in the GSE180081 cohort, as well as between-group differential genes in all CAD patients grouped according to their cuproptosis gene dataset clustering group. After taking the intersection of these two sets of differential genes, we screened for potential biomarkers of CAD in combination with clinical features. These and other study procedures are outlined in a flowchart in Fig. 1.

Fig. 1.

Flowchart for the study

Identification of DEGs in the GSE180081 cohort

The GSE180081 dataset obtained from the GEO database comprised 96 samples (48 from CAD patients and 48 from those without). Of the 44,265 genes represented in the GSE180081 dataset, 943 were differentially expressed (CAD-DEG); 645 were upregulated and 298 were downregulated (Log2(FC) > 1; P < 0.05) (Fig. 2A, B). For the CAD group, we clustered the training set based on its relationship to the cuproptosis gene dataset (from k = 2 to k = 6) (Supplementary Fig. S1). Based on the results of the cluster analysis, we divided the samples into two groups (Cu_group1, Cu_group2), which comprised 21 and 27 samples, respectively (Fig. 2C). DEGs between Cu_group1 and Cu_group2 were identified using limma, with covariates (age, sex) included in the linear model. Here, 1,108 genes were differentially expressed (Cu-DEG), of which 793 were upregulated and 369 were downregulated (Log2(FC) > 1; P < 0.05) (Fig. 2D, E). The clusters did not correlate with known CAD subtypes but reflected differential cuproptosis-related gene expression. The gene set at the intersection of CAD-DEG and Cu-DEG, referred to as Co-DEG, comprised 818 differentially expressed genes, of which 231 were upregulated and 587 were downregulated (Fig. 2F).

Fig. 2.

Identification of CAD-DEGs and Cu-DEGs in the GSE180081 cohort. A, B Volcano and heat maps showing the genes that were differentially expressed between the CAD and nonCAD groups (CAD-DEGs). C Results of the cluster analysis of the CAD samples and cuproptosis gene sets using the GSE180081 data set. D, E Volcano and heat maps showing the genes that were differentially expressed in the CAD group in the cluster analysis of cuproptosis genes (Cu-DEGs). F The intersection of CAD-DEG with Cu-DEG, showing that 231 genes were upregulated and 587 genes were downregulated. p < 0.05 was considered to indicate significant enrichment

Functional analysis of the Co-DEGs

To elucidate the biological functions and potential pathways of the Co-DEGs, GO enrichment and KEGG pathway analysis were performed. As shown in the scatter graph of the GO analysis (Fig. 3A), the differential genes were principally derived from the positive regulation of cell activation, positive regulation of leukocyte activation, actin filament organization, leukocyte cell–cell adhesion, and positive regulation of lymphocyte activation pathways, which implies that leukocyte activation and interactions are important in this interaction. The KEGG analysis showed that the Co-DEGs were principally associated with infectious diseases (Fig. 3B). Thus, when the results of the two analyses are combined, the differentially expressed genes appear to be principally associated with immune function.

Fig. 3.

Results of the functional analysis of the Co-DEGs. A Results of the GO analysis showing that numerous Co-DEGs are present in pathways associated with cell activation and inflammation. B Results of the KEGG analysis showing that numerous Co-DEGs are present in pathways associated with multiple infections and inflammation. C Results of GSEA enrichment for the Co-DEGs. D Chordal diagram showing the key genes involved in five pathways with close correlations in the GO analysis. p < 0.05 was considered to indicate significant enrichment

The inflammatory response of vascular endothelial cells plays a crucial role in the development of CAD [45], and many of the results obtained from the enrichment analysis are related to inflammation regulation, which suggests that these differential genes are likely to mediate the inflammatory state of CAD and participate in the course of the disease. This also provides guidance for further research on the inflammatory state involved. Cuproptosis-related genes were more strongly linked to immune pathways than apoptosis or ferroptosis. Their unique association with mitochondrial metabolism (e.g., FDX1, LIAS) may explain this specificity. GSEA analysis showed that the genes of interest were principally derived from the Apical junction, Complement, and Hypoxia pathways (Fig. 3C). To better visualize the key differentially expressed genes in the Co-DEG set, we used a chordal diagram to show the genes involved in the pathways identified in the GO analysis (Fig. 3D).

Diagnostic Co-DEGs genes for CAD

We used the aforementioned 818 Co-DEGs in what follows. LASSO regression screening identified five genes that were associated with the diagnostic features of CAD: HIST1H4E, IL6ST, RN7SKP45, LST1, and SNORD50B (Fig. 4A, B, Supplementary Table S1). In addition, these five genes were able to distinguish the CAD and control groups in the LASSO regression model (Fig. 4C). Next, we used these five genes to construct a logistic regression model that we analyzed using ROC analysis. This analysis showed that the model was capable of satisfactorily identifying CAD in the training set (AUC 0.83, 95% CI: 0.75–0.92, Fig. 4D), so we also tested it on the GSE180082 dataset, where it resulted in an AUC of 0.62 (Fig. 4D).

Fig. 4.

Co-DEGs include genes that are diagnostic for CAD. A LASSO regression cross-validation error curve. B LASSO regression coefficient solution path used to determine the optimal coefficient. C Diagnosis-related genes, identified using LASSO regression that could distinguish the CAD and control groups. D ROC analysis results. Analysis of the GSE180081 training set yielded an AUC of 0.83, and analysis of the GSE180082 test set yielded an AUC of 0.62

Decision curve analysis (Supplementary Fig. S2) demonstrates optimal clinical utility with a net benefit surpassing "treat-all" strategy when threshold probabilities exceed 10%.

Although the model we constructed achieved satisfactory results in the training set, the AUC in the test set was quite a bit lower. This may be due to the batch effects between different sequencing data, or it may have been caused by some baseline differences in the sequencing population. Since CAD is both a complex and latent disease, it poses a great challenge in terms of diagnosis. Thus, even though the AUC of our diagnostic declined when predicting the test dataset, it can still provide relatively effective predictive power and offer some reference for early clinical risk assessment. Additionally, these results also provide evidence that copper-death-related genes play a certain role in the development of coronary heart disease, which lays a theoretical foundation for further research.

Immune landscape analysis

We then carried out an in-depth investigation of these five genes, including analyses of their immunological roles and the regulation and expression of the protein-coding genes. Though cuproptosis genes were present in Co-DEGs, their pathway enrichment was indirect, suggesting synergistic roles with immune pathways. Based on the aforementioned diagnostic model, we conducted a CAD risk assessment for all samples and divided them into high-risk and low-risk groups according to the median of the risk scores. The risk groups identified using the proposed model were combined with the expression matrix to compare the immune cell infiltration of the groups using three methods: ssGSEA, estimate, and Cibersort. The ssGSEA analysis showed differences in the numbers of 14 of the 28 types of immune cells, and all of these except activated CD4 T cells were present in larger numbers in the high-risk group (Fig. 5A, B). Each of the five identified genes is closely associated with the function of multiple immune cells, and the LST1 gene in particular is associated with 20 types of immune cells and therefore may be a key candidate for future research (Fig. 5C).

Fig. 5.

Results of the immune landscape analysis. A Results of the ssGSEA immune infiltration analysis showing that 13 out of 28 types of immune cells are expressed more in the high-risk group and that activated CD4 T cells are expressed less. B Heat map of ssGSEA immune infiltration. C Results of the ssGSEA immune infiltration analysis of the five identified genes showing that all were closely associated with the abundance of multiple types of immune cells (“*” refers to P < 0.05; “**” refers to P < 0.01; “***” refers to P < 0.001). D Estimate immune scores, showing higher scores in the high-risk group. E Results of the Cibersort analysis demonstrating immune infiltration. p < 0.05 was considered to indicate significant enrichment

The immunization scores obtained using the estimate package showed differences between the high-risk and low-risk groups, with a higher overall immunization score in the high-risk group, consistent with the ssGSEA findings (Fig. 5D). Similarly, Cibersort analysis also showed differences in immunity between the two groups but with the high-risk group showing a lower number of memory CD4⁺ T cells (Fig. 5E). Thus, although the exact immune statuses of the two groups remain to be determined by laboratory experiments, our results demonstrate clear differences between the subgroups of the diagnostic model constructed using the cuproptosis gene set. Despite methodological differences, all tools converged on heightened immune activity in high-risk patients. ssGSEA and CIBERSORT both implicated macrophages, while ESTIMATE underscored stromal-immune crosstalk (Supplementary Table S4). Tool-specific findings (e.g., CIBERSORT’s M2 macrophages) may reflect granularity in cell-type resolution.

The identified genes are closely associated with CAD-related pathways

To ensure that multicollinearity during the logistic regression analysis did not interfere with the findings, we performed correlation analysis on the five genes used to construct the model (Fig. 6A) and found that there were no close correlations between any of them. The results of the differential expression analysis indicated that all five genes were expressed less in the CAD group compared to the control group. Furthermore, to make sure that the five genes endowed the model with high reliability, we corrected them with multiple tests and found that their adj P-values were also statistically significant (Table 1). In addition, we performed GSEA analysis for each of the five genes (Fig. 6B–F) and found that they are associated with pathways involved in cardiovascular diseases and the regulation of inflammation, including Leukocyte transendothelial migration, Type I diabetes mellitus, Viral myocarditis, arrhythmogenic right ventricular cardiomyopathy, and glycine, serine, and threonine metabolism. Therefore, the roles of these five genes in CAD should be further explored.

Fig. 6.

The five identified genes are closely associated with CAD-related pathways. A Results of the correlation analysis of the five identified genes. B–F Results of the single-gene GSEA analysis of HIST1H4E, LST1, RN7SKP45, IL6ST, and SNORD50B, respectively

Table 1.

Results of limma analysis demonstrating low expression of five genes *

Gene	Log2(FC)	AveExpr	t		P-value	adj. P-value	B	Change
HIST1H4E	− 1.28	6.82	− 8.44		3.65E − 13	1.66E − 11	19.16	DOWN
IL6ST	− 1.26	6.70	− 7.78		8.85E − 12	1.74E − 10	16.23	DOWN
RN7SKP45	− 1.39	6.89	− 7.7		1.28E − 11	2.38E − 10	15.75	DOWN
LST1	− 1.42	7.37	− 7.66		1.62E − 11	2.90E − 10	15.41	DOWN
SNORD50B	− 1.21	6.70	− 6.93		5.17E − 10	6.40E − 09	12.48	DOWN

^* log₂ (FC): estimate of the log₂-fold-change corresponding to the effect or contrast; AveExpr: average log₂-expression for the probe over all arrays and channels, same as Amean in the MarrayLM object; t: moderated t-statistic; P-value: raw p-value; adj. P-value: adjusted p-value or q-value; B: log-odds that the gene is differentially expressed; Change: differential trends in individual gene

Associations of the identified genes with immunosuppressant, immune activator, and MHC-related genes

We next analyzed the relationships of well-established immunosuppressant genes (Fig. 7A), immune activator genes (Fig. 7B), and MHC genes (Fig. 7C) with the above five identified genes and found that the identified genes closely correlated with BTLA and VTCN1 among the immunosuppressant genes, CD276 and TMEM173 among the immune activator genes, and HLA-DQA1 and TABPBP among the MHC-related genes. Subsequently, each group of correlations was examined. HIST1H4E was found to be closely associated with BTLA and TMEM173, and LST1 was found to be closely associated with TMEM173 and TAPBP (Fig. 7D). Therefore, the expression of HIST1H4E and LST1 may be regulated by these genes, which warrants further research.

Fig. 7.

A, B, C Correlations of the expression of the identified genes with those of immunosuppressant, immune agonist, and MHC-associated genes. D More detailed presentations of A–C

Creation of a ceRNA network based on the identified genes and the prediction of drugs that can target these genes

Despite the successes achieved from the above research, we unfortunately found that SNORD50B is a noncoding RNA [46] and that RN7SKP45 is a pseudogene [47]. Thus, in order to identify biomarkers of practical value, we focused all subsequent research on the other three genes (HIST1H4E, IL6ST, LST1). We constructed a ceRNA network based on the three genes using information from the MicroCosm and starBase databases and the multiMiR package. The network contained 59 nodes (3 genes, 47 miRNAs, 9 lncRNAs) and 59 edges (Fig. 8A). We identified many miRNA regulatory networks targeting LST1, with 27 of the miRNAs involved in regulation, and the principal associated lncRNAs included DANCR, NEAT1, and SNHG22. The regulatory network of IL6ST was also found to be diverse, with 15 miRNAs found to be involved in regulation and five associated lncRNAs, including let-7b-5p and miR-628-5p. Six miRNAs were found to be involved in the regulation of HIST1H4E, among which miR-148a-3p was found to have two related lncRNAs: NUTM2A-AS1 and AL049840.4.

Fig. 8.

Construction of a ceRNA network and the prediction of gene-targeting drugs. A ceRNA network of the identified genes. B Gene-targeting drugs identified using the DGIdb database

We also attempted to identify drugs that may target the identified genes using the DGIdb database, and these results, obtained using the Cytoscape software, are shown in Fig. 8B. Three drugs that target IL6ST and one that targets LST1 were identified. According to our research, the abnormal expression of HIST1H4E, IL6ST, and LST1 in the blood of CAD patients may predict the occurrence of CAD. However, whether the expression changes in blood biomarkers indicate further acceleration of the course of CAD, and whether drug intervention in blood biomarkers can prevent the development of CAD, are questions we believe are worth discussing. Therefore, we also explored potential gene-targeted drugs.

RT-qPCR analysis of the expression of the identified genes in mononuclear cells obtained from the peripheral blood of CAD patients and healthy controls

We collected fresh blood from 12 CAD patients and 12 healthy people who were undergoing a physical examination at the same hospital. Validation cohorts were age-matched with no significant differences in hypertension or smoking (P > 0.05) (Supplementary Table S2). RNA was extracted from peripheral monocytes, and RT-qPCR was performed to measure the expression of LST1, IL6ST, and HIST1H4E. The primer sequences for LST1, IL6ST, and HIST1H4E are listed in Table 2.

Table 2.

Primer sequences used in the study

Gene name	Forward	Reverse
LST1	GCCCCTGATCATTTCGCCTA	CTGGGACCAGGACAGAAGGT
IL6ST	TGTTTTAACTATCCCTGCCTGT	CATTTGCTCTCTGCTAAGTTCC
HIST1H4E	TAAGGTCCTGCGAGATAACATC	GTAAGTCACAGCATCACGAATC
ACTB	CATGTACGTTGCTATCCAGGC	CTCCTTAATGTCACGCACGAT

We found that the expression of LST1 and HIST1H4E was significantly lower in the CAD group (Fig. 9), which is consistent with the results of our bioinformatic analysis. In contrast, IL6ST was expressed at a significantly higher level in the CAD group (Fig. 9), suggesting that further research regarding this gene is merited. Thus, LST1, HIST1H4E, and IL6ST may represent biomarkers of CAD, and our results may serve to provide a basis for further study of these genes in the future.

Fig. 9.

Expression of the identified genes in blood samples from individuals with and without CAD. RT-qPCR analysis of the peripheral blood mononuclear cells showed low LST1 and HIST1H4E expression and high IL6ST expression in the CAD group

Discussion

Selection and establishment of methods for predicting CAD

Coronary artery disease (CAD) represents a leading global public health burden. Early identification of at-risk individuals is therefore essential for reducing CAD-related morbidity and mortality. Current clinical diagnosis relies principally on imaging modalities—notably coronary computed tomography angiography (CCTA) and invasive coronary angiography. However, developing accurate non-invasive methods for CAD prediction remains critical to enabling effective prevention and personalized management strategies. [48]. Therefore, it would be of great significance to identify peripheral blood predictors of CAD.

A growing body of research suggests that cell death plays important roles throughout the pathogenesis of CAD [49]. Both apoptosis and autophagy are upregulated in CAD patients, for example [50]. Some previous studies have already sought to identify biomarkers of CAD and to construct diagnostic models based on iron-related cell death [29, 51, 52], but the relationship between a newly identified form of cell death, cuproptosis, and CAD has yet to be elucidated [15].

This study conducted an in-depth analysis of blood transcriptomic profiles from CAD patients and controls, integrated with cuproptosis-related genes, to identify biomarkers for CAD prevention, prediction, and personalized treatment. We functionally characterized five candidate genes (HIST1H4E, RN7SKP45, IL6ST, LST1, and SNORD50B). Single-gene GSEA revealed associations between all five genes and inflammation/cardiovascular pathways. However, RN7SKP45 and SNORD50B showed no significant functional enrichment in disease-relevant mechanisms. Consequently, we prioritized HIST1H4E, IL6ST, and LST1 for further investigation.

Biological characteristics and functions of HIST1H4E, IL6ST, and LST1 in CAD

The leukocyte-specific transcript 1 (LST1) gene, which encodes a member of MHC class III, contains five noncoding exons and four coding exons [53, 54]. High LST1 expression has been reported in inflammatory bowel disease [55] and rheumatoid arthritis [56], suggesting that it may regulate inflammation. LST1 is polymorphic, being not only a soluble protein but also a transmembrane protein [57]. Therefore, it may be a transmembrane adapter protein (TRAP) and may be associated with cell surface receptors that initiate intracellular signaling cascades [58]. There is evidence that LST1 has an important role in inflammation and immunity [55, 59], but the underlying mechanisms have yet to be described. Therefore, the association between LST1 and CAD deserves further exploration.

HIST1H4E, also known as H4C5, is a member of the histone H4 family. Histone modification is a key regulator of the tightness of chromatin structure. Moreover, the overexpression of histone deacetylase 4 (HDAC4) in H9c2 cardiomyocytes leads to mitochondrial dysfunction and increases the probability of cell death [60]. Histone deacetylase (HDAC) inhibitors significantly reduce the area of myocardial infarction [61], and in addition, numerous studies have shown a protective effect of HDAC inhibitors in cardiomyocytes [62–64]. HIST1H4E has been found to be associated with melanoma metastasis [65], it can also be used as an independent predictor of the prognosis of melanoma [66]. However, the question of whether HIST1H4E is useful for the diagnosis and/or treatment of CAD has yet to be answered.

Interleukin-6 signal transducer (IL6ST) is a member of the cytokine receptor family that mediates the initiation of intracellular signaling by ligands such as IL-6 and plays important roles in growth, immune homeostasis, infection, trauma, autoimmune diseases, and cancer [67]. It is also a prognostic marker for breast cancer [68] and promotes the proliferation and migration of liver cancer cells [69]. Additionally, recent studies have shown that it alleviates endothelial injury [70, 71] and participates in myocardial protection [72].

Strengths and limitations of the present study, and suggestions for future research

In the present study, we used a cuproptosis gene set to construct a diagnostic model, an approach seldom seen in the recent literature. Based on blood transcriptomics, we identified novel biomarkers and constructed a predictive model that demonstrates superior performance compared to existing models [29, 73, 74]. To elucidate the underlying mechanisms, we compared immune cell profiles between high-risk and low-risk groups. Notably, the high-risk group exhibited significantly elevated levels of multiple immune cell types. These findings align with the observed enrichment of Co-DEGs in immune-related pathways, collectively suggesting that immune dysregulation contributes critically to CAD pathogenesis—a role that is well-established in the literature. [75, 76]. As a result, studies have been conducted to search for immunology-related blood biomarkers of CAD [77].

To provide a basis for further studies of the genes identified in this paper, we analyzed HIST1H4E, IL6ST, and LST1 for potentially relevant interactions with immunosuppressant, immune activator, and MHC-related genes, and attempted to identify drugs that target these genes. In addition, we constructed ceRNA networks that can provide a bioinformatic basis for future in-depth basic research. To validate our bioinformatic findings, we collected peripheral blood from CAD patients and healthy individuals from whom RNA was extracted from monocytes to measure the expression of HIST1H4E, IL6ST, and LST1 and thereby look for evidence that these three genes may represent biomarkers of CAD with the hope of inspiring further, in-depth mechanistic studies.

There are some limitations to the present study, however. First, the patient data in the GEO database are sparse, which may have affected the reliability of the model constructed and may have led to the omission of some potential biomarkers. Second, we performed a retrospective analysis, which may have resulted in selection bias. Therefore, prospective studies should be performed in the future to corroborate the reliability of our identified biomarkers. The small validation cohort may limit generalizability. Future studies with larger, independent cohorts are needed to confirm HIST1H4E, IL6ST, and LST1 as biomarkers. Nevertheless, we have conducted a preliminary screen for markers of CAD using publicly available data and made an initial validation with patient data, the results of which we hope will serve as the basis for prospective studies of HIST1H4E, IL6ST, and LST1k and further exploration of their roles in the pathogenesis of CAD.

Conclusion

In summary, we identified HIST1H4E, IL6ST, and LST1 as genes that are associated with cuproptosis and are biomarkers of CAD and used them to construct a new CAD prediction model. Furthermore, we explored the relevance of these genes to immunity and identified potential targeting drugs and regulatory networks. These results may help guide the development of targeted prevention and personalized treatment strategies for CAD in the future.

Author contributions

Jia Li analyzed the data, completed the figures, and wrote the manuscript. Ping Hu processed the raw data. Can-E Tang, Zhanwei Zhu, and Fanyan Luo designed the research. Can-E Tang, Kaibo Lei, Lin Wang and Fanyan Luo reviewed and edited the manuscript. Lin Wang guided data analysis and supplementary revisions. All authors contributed to the article and approved the submitted version.

Funding

This work was supported by the National Natural Science Foundation of China (82471620), the Hunan Province Natural Science Foundation regional joint fund (2024JJ7631) and Hunan Natural Science Foundation (2022JJ70075), and the National Natural Science Foundation of China (81974112).

Data availability

Publicly available datasets were analyzed in this study. These data can be found here: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE180081 (GPL14761), https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE180082 (GPL14761).

Declarations

Ethics approval and consent to participate

The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Medical Ethics Committee, Xiangya Hospital, Central South University (201803209).

Informed consent statement

Informed consent was obtained from all subjects involved in the study.

Conflicts of interest

The authors declare no competing interests.