Gene expression data for CHD and normal samples were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE202625). The GSE202625 dataset, encompassing twenty-seven CHD blood samples and twenty-five normal blood samples, was used to validate the expression of potential marker genes.

Differential expression analyses were conducted using the R package DESeq2 (1.38.3). Normalized data were assessed to identify differentially expressed genes (DEGs) between samples. Correction for multiple hypothesis testing was executed, and p-values were adjusted utilizing the false discovery rate (FDR), and the corrected p-value was called the q-value. DEGs were deemed statistically significant at a q-value 0.05. The fold change (log₂FC) was calculated using fragments perkilobase of exon model per million mapped reads (FPKM) values, and DEGs with log₂FC $>$1 or log₂FC –1 were considered up-regulated and down-regulated, respectively.

R software (R Foundation for Statistical Computing, Vienna, Austria), along with the “biological conductor” and “ggplot” packages (3.4.2), was employed for gene ontology (GO) function and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. The DEGs were categorized into biological processes, cellular components, and molecular functions. Enrichment was determined using an q-value of 0.05.

WGCNA explores genetic combinations and associations among samples, aiding biomarkers and therapeutic gene discovery [12]. The study employed 1338 DEGs to assess network topology for scale independence and connectivity using WGCNA. Hierarchical clustering was employed to cluster genes with comparable expression patterns into modules. Module–trait correlation identified highly CHD-associated module genes.

We assessed the impact of hub genes on CHD by employing a method of least absolute shrinkage and selection operator (LASSO) regression through the R glmnet [13]. LASSO is renowned for analyzing high-dimensional data, aiming to minimize the residual sum of squares while constraining the absolute sum of the model coefficients below a threshold. Genes with p 0.05 significance were input into LASSO, yielding regression coefficients precisely equal to 0. This approach reduced data dimensionality by excluding low-weight variables and curbing overfitting due to covariate collinearity. We evaluated the diagnostic potential of these gene biomarkers by computing receiver operating characteristic (ROC) curves and measuring the corresponding area under the curve (AUC).

Human umbilical vein endothelial cells (HUVECs) (RRID: CVCL_2959, HTX2104) were obtained from Shenzhen OTWO Biotech Co., Ltd (Shenzhen, China). The cell lines were identified using short tandem repeat (STR) profiling, and the absence of mycoplasma contamination was confirmed. The sh-negative control (NC) plasmid and shZNF429 plasmid were designed and constructed by Chongqing Biomedicine Biotechnology Co., Ltd (Chongqing, China). The shZNF429 interference sequence is as follows: F-5^′-GATCCGGCTATACAAAGGAGGTTATAACTCGAGTTATAACCTCCTTTGTATAGCTTTTTTG-3^′; R-5^′-AATTCAAAAAAGCTATACAAAGGAGGTTATAACTCGAGTTATAACCTCCTTTGTATAGCCG-3^′. And the pLVX-shRNA1 vector without the insertion of the shZNF429 interference sequence serves as the NC. Specifically, two 1.5 mL EP tubes were prepared, with 25 µg NC plasmid added to the sh-NC group and 25 µg shZNF429 plasmid to the shZNF429 group. OPTI-MEM (31985070, GIBCO, Shanghai, China) was supplemented to make a final volume of 500 µL. Then, two additional 1.5 mL Eppendorf tubes (EP) were prepared by adding 75 µL transfection reagent max (24765-1, Biomedicine, Chongqing, China) and 425 µL optimized-minimal essential medium (OPTI-MEM) in each tube. After 5 minutes, the transfection reagent-containing mixture from the EP tubes was added into the plasmid-containing EP tubes, mixed thoroughly, let stand at room temperature for 15 minutes, and then added to HUVECs for passage and culture. The experimental groups were as follows: The control group: HUVEC cells were cultured in DMEM medium without serum and dual-antibody culture medium (C11995500BT, GIBCO, Shanghai, China). The model group: HUVEC cells were cultured in DMEM medium without serum and dual-antibody at a final concentration of 80 µg/mL oxidized low-density lipoprotein (ox-LDL, 20605ES05, Yeasen, Shanghai, China). The model+sh-NC group: transfected sh-NC plasmid cell suspension was adjusted to a density of 2 $\times$ 10⁵ cells/mL, and changed to culture in DMEM medium without serum and dual-antibody at a final concentration of 80 µg/mL ox-LDL. The model+shZNF429 group: transfected shZNF429 plasmid cell suspension was adjusted to a cell density of 2 $\times$ 10⁵ cells/mL for plating, and cultured in DMEM medium without serum and dual-antibody at a final concentration of 80 µg/mL ox-LDL. After 24 h of culture in each group, samples were collected for subsequent detection. The gene level of ZNF429 was knocked down by transfection with shZNF429 plasmid. All experimental results were repeated three times.

Each well of a 96-well plate was supplied with 10 µL of CCK-8 solution (C0038, Beyotime, Shanghai, China). Blank wells were filled with culture medium and CCK-8 solution, while those for control with cells, culture medium, and CCK-8 solution. Following one-hour incubation, absorbance at 450 nm was measured using a microplate reader (CMax Plus, Molecular Devices, San Jose, CA, USA).

For flow cytometry detection of cell proliferation, BeyoClick™ EdU-555 Cell Proliferation Kit (C0075S, Beyotime, Shanghai, China) was used following the operation steps. Cells were labeled with EdU, digested, resuspended, and added with 1 mL of 4% paraformaldehyde fixative solution. Following PBS treatment, 1 mL of permeabilization solution was supplied. Then, 0.5 mL of Click solution was added to each tube to evenly cover the samples, followed by incubation at room temperature for 30 minutes. The cells were detected using a flow cytometer (1026M, Betalco, Changzhou, China) and data were analyzed with FlowJo_V10 software (FlowJo LLC, Ashland, OR, USA).

Flow cytometry apoptosis was detected using an Annexin V-FITC/PI apoptosis detection kit (40302ES60, Yeasen, Shanghai, China), following the manufacturer’s protocol. After the collection of cell suspension from each group, the cells were centrifuged and the supernatant was discarded. The cells were gently resuspended with 195 µL of AnnexinV-FITC-A binding solution. Then, 5 µL of AnnexinV-FITC was added and mixed, followed by centrifugation and removal of the supernatant. Subsequently, 190 µL of AnnexinV-FITC-A binding solution was gently resuspended with cells. After adding the staining solution, the samples were placed in an ice bath in the dark before detection, and data were analyzed with FlowJo_V10 software.

ELISA experiments were conducted following the instructions provided with the respective assay kits. The ELISA kits for human IL-6, IL-1$\beta$, and TNF-$\alpha$ were purchased from Quanzhou Ruixinbio Co., Ltd (Quanzhou, China). The kits for intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1) were procured from Shanghai Qifa Laboratory Reagent Co., Ltd (Shanghai, China).

Total RNA was extracted from each cell group using the TRIzol method, and the concentration and purity of RNA were determined using a UV–visible spectrophotometer (OD260/OD280). Complementary DNA (cDNA) synthesis was performed using the cDNA synthesis kit instructions (TSK302M, Tsingke, Beijing, China), and the qPCR reaction system was prepared (TSE002, Tsingke, Beijing, China). Following this, reactions were conducted in a real-time fluorescent quantitative PCR instrument (Bio-Rad Laboratories, Hercules, CA, USA), with the housekeeping gene GAPDH employed for normalization. Gene expression levels were determined using the 2^{-Δ⁢Δ⁢Ct} method. Primer sequences can be found in Table 1.

The cell total protein was lysed on ice. In total, 500 µg of total protein from each sample was mixed with buffer (8015011, Dakewei, Shenzhen, China) (resulting in a protein concentration of approximately 3.3 µg/µL after mixing) and subjected to metal bath denaturation. Sixty micrograms of denatured protein were loaded for electrophoresis at 80 V until the samples entered the concentrating gel, followed by an increase in voltage to 120 V. Primary antibodies against ZNF429 (ab171422, abcam, Shanghai, China), VCAM-1 (A19131, abclonal, Wuhan, China), ICAM-1 (A20472, abclonal, Wuhan, China), and GAPDH (A19056, abclonal, Wuhan, China) were diluted at 1:1000 and incubated overnight at 4 °C. Secondary antibodies (AS014, abclonal, Wuhan, China) were then diluted at 1:2000 and incubated at room temperature for 1 hour. The membrane was uniformly covered with ECL exposure solution and detected using an exposure instrument (Universal Hood II, Bio-Rad, Hercules, CA, USA).

Fixed cells were rinsed on the coverslip three times with PBS, each time for 5 minutes. After blocking with goat serum (C0265, Beyotime, Shanghai, China) for 30 minutes, primary antibodies against P-selectin (CD62P, 1:200, bs-0561R, Bioss, Beijing, China) and E-selectin (CD62E, 1:200, bs-1273R, Bioss, Beijing, China) were incubated at 4 °C overnight and washed three times. Secondary antibodies (1:200, GB21303, Servicebio, Wuhan, China) were added and incubated at room temperature for 1.5 hours and rinsed with PBS three times. DAPI staining (C1005, Beyotime, Shanghai, China) was performed, and the sections were sealed with an anti-fluorescence quenching agent (P0126, Beyotime, Shanghai, China) before observing changes under a microscope.

All experimental results were repeated three times. One-way ANOVA analysis and plotting were performed using Prism GraphPad 8.0 software (Dotmatics, Boston, MA, USA). Data display indicates mean $\pm$ standard deviation (x $\pm$ s), and p 0.05 was considered statistically significant.

This study utilized the R DESeq2 package to identify 1338 genes with distinct expression patterns between normal and CHD samples ($|$log₂FC$|$ $>$1, q-value 0.05). Within the set of 1338 DEGs, 700 genes were upregulated and 638 were downregulated (Fig. 1A). To screen for functional and pathway enrichments linked to CHD-related DEGs, we conducted GO and KEGG enrichment analyses. GO analysis was employed to analyze the biological process (BP), cellular component (CC), and molecular function (MF) of the DEGs. GO analyses showed notable enrichment within specific biological processes, including leukocyte-mediated immunity and lymphocyte-mediated immunity. The chart displaying CC analysis revealed significant enrichment of DEGs within structures, such as the T-cell receptor complex, secretory granule membrane, and plasma membrane signaling receptor complex. In the MF category, DEGs demonstrated prominent enrichment in functions such as antigen binding, pattern recognition receptor activity, and immunoglobulin receptor binding (Fig. 1B). Regarding KEGG analysis, the outcomes demonstrated significant enrichment of DEGs in pathways including the ribosome, mitogen-activated protein kinase (MAPK) signaling pathway, coronavirus disease (COVID-19), ribosome biogenesis in eukaryotes, and the oxytocin signaling pathway (Fig. 1C).

Fig. 1.

Screening of coronary heart disease biomarkers based on bioinformatics analysis. (A) Volcano map depicting differentially expressed genes (DEGs) comparison between the coronary heart disease (CHD) group and control group, featuring distinctive color codes for gene expression alterations. (B) Top 10 gene ontology (GO) terms in the GSE202625 dataset. (C) Top 30 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways in the GSE202625 dataset. (D) The scale-free fitting index of soft threshold power. (E) Cluster dendrogram of DEGs using various metrics. Each branch signifies a gene, while colors below denote co-expression modules. (F) Interactions among co-expression modules. (G) Matrix illustrating the correlations and corresponding p-values between traits and modules. (H) Top 10 GO enrichment terms of DEGs within the turquoise and brown modules. (I) Top 30 KEGG enrichment pathways of the DEGs within the turquoise and brown modules. (J) Determination of the number of factors. (K) The important prognostic variables were selected using the least absolute shrinkage and selection operator (LASSO) regression method. (L) Logistic regression model to identify the area under the curve (AUC) of disease samples. The abscissa is the false positive rate, and the ordinate is the true positive rate.

By evaluating the scale-free topology fitting index for a specific power and network connectivity strength, we ensured alignment of the corresponding soft threshold power and a scale-free network. Network topology analysis was used to determine a threshold force of 4, and a hierarchical clustering tree was generated based on this value (Fig. 1D,E). The resultant hierarchical clustering tree delineated 17 co-expressed gene modules that were identified for subsequent exploration. Further assessment addressed the independence among these 17 co-expression modules, revealing a minimal overlap between their constituent genes (Fig. 1F). Notably, the module–trait correlation heat map highlighted that the turquoise and brown modules had the highest correlations with CHD characteristics, with correlation coefficients of 0.80 (p = 0.0002) and 0.76 (p = 0.0006, Fig. 1G), respectively. Functional and pathway enrichment analyses were performed for DEGs within the turquoise and brown modules. The results of GO analysis highlighted predominant enrichment in key processes, such as regulation of protein ubiquitination, positive regulation of protein modification via small protein conjugation or removal, regulation of protein modification via small protein conjugation or removal, positive regulation of protein ubiquitination, ribosome, structural constituent of ribosome, and phosphatidylinositol bisphosphate phosphatase activity (Fig. 1H). KEGG analysis revealed substantial enrichment in pathways including the MAPK signaling pathway, signaling pathways regulating pluripotency of stem cells, the Hippo signaling pathway, Kaposi sarcoma-associated herpesvirus infection, and the ribosome (Fig. 1I).

LASSO analysis of DEGs from turquoise and brown modules was used to identify hub genes for CHD. Following LASSO analysis, the selection of these vital markers was accomplished through the precision of the penalty parameter adjustment, facilitated by a 10-fold cross-validation approach (Fig. 1J). LASSO regression analysis yielded two features, indicating that we reduced the number of latent variables to 2 (Fig. 1K). Consequently, two hub genes, SLC16A6 and ZNF429, demonstrated a discernible association with CHD. Subsequent ROC curve analysis further illuminated the discriminative potential of the logistic regression model based on these two marker genes. AUC values of SLC16A6 and ZNF429 were 0.952 and 0.984, respectively, underscoring the relatively high predictive efficacy of these hub genes in distinguishing between normal and CHD samples. This outcome underscores the considerable utility of the identified hub genes as potential indicators of CHD diagnosis and prediction (Fig. 1L). Based on the above analysis, we found that the AUC value of ZNF429 is higher. Therefore, ZNF429 was chosen for cell experimental validation.

IL-6, IL-1$\beta$, and TNF-$\alpha$ are common inflammatory cytokines, while VCAM-1 and ICAM-1 are adhesion molecules associated with CHD [14, 15]. Previous studies have shown that IL-6, IL-1$\beta$, TNF-$\alpha$, VCAM-1, and ICAM-1 levels are significantly increased in patients with CHD [16, 17]. Based on bioinformatics analysis, we identified one hub gene, ZNF429. To further validate the roles of ZNF429 in CHD, we established a HUVEC cell model and interfered with ZNF429. In this study, qPCR detection results indicated that compared to control, mRNA expression levels of ZNF429, VCAM-1, and ICAM-1 were markedly increased in the model group and model+sh-NC group, confirming successful modeling (p 0.01, p 0.001). Furthermore, compared to the model, there was no significant difference in the mRNA expression of ZNF429, VCAM-1, and ICAM-1 in model+sh-NC (p $>$ 0.05); in fact, they were greatly decreased in model+shZNF429 (p 0.001). mRNA expression of ZNF429, VCAM-1, and ICAM-1 in model+shZNF429 decreased greatly vs. model and model+sh-NC groups (p 0.001, Fig. 2A). The Western blot analysis of the protein expression of ZNF429, VCAM-1, and ICAM-1 also exhibited a similar trend (Fig. 2B). ELISA results indicated significantly higher levels of IL-1$\beta$, IL-6, TNF-$\alpha$, VCAM-1, and ICAM-1 in the model group and model+sh-NC group compared to the control group (p 0.001). In contrast, compared to the model group, there were no significant differences in the model+sh-NC group (p $>$ 0.05), while the levels of IL-1$\beta$, IL-6, TNF-$\alpha$, VCAM-1, and ICAM-1 were significantly reduced in the model+shZNF429 group (p 0.01, p 0.001). The described factors in the model+shZNF429 group were greatly lower than those in the model+sh-NC group (p 0.01, p 0.001, Fig. 2C). These results indicate that ZNF429 plays critical roles in regulating cell inflammation and adhesion.

Fig. 2.

Differences in cell characteristics detected via quantitative polymerase chain reaction (qPCR), Western blot, and enzyme-linked immunosorbent assay (ELISA). (A) qPCR detection of the mRNA expression of ZNF429, VCAM-1, and ICAM-1 in cells. (B) Western blot detection of the mRNA expression of ZNF429, VCAM-1, and ICAM-1 in cells. (C) ELISA detection of IL-1$\beta$, IL-6, TNF-$\alpha$, VCAM-1, and ICAM-1 in the cells. Compared to control, * p 0.05, ** p 0.01, *** p 0.001. Compared to model, # p 0.05, ## p 0.01, ### p 0.001. Compared to model+sh-NC group, && p 0.01, &&& p 0.001. n = 3.

CCK-8 assay results demonstrated a significantly lower proliferation rate in the model group and model+sh-NC group compared to the control (p 0.001). However, there were no significant differences in the model+sh-NC group and model group (p $>$ 0.05). Notably, the model+shZNF429 group showed significant differences compared to both the model group and the model+sh-NC group ((p 0.01, p 0.001, Fig. 3A). The results of flow cytometry implied that compared with control, cell proliferation ability of model and model+sh-NC groups was greatly decreased (p 0.001). Compared with the model group, there was no significant difference in the proliferation ability of the model+sh-NC group (p $>$ 0.05), but the proliferation ability of the model+shZNF429 group was notably increased (p 0.001). In addition, the proliferation ability of model+shZNF429 was greatly higher than that of model+sh-NC (p 0.001, Fig. 3B). Subsequently, we measured cell apoptosis. As shown in Fig. 4, compared with control, the apoptosis rate of model and model+sh-NC groups was significantly increased (p 0.001). However, the model+shZNF429 group was significantly lower than the model and model+sh-NC groups (p 0.01). These results indicated that interference with ZNF429 attenuated a decrease in cell proliferation rate but an increase in apoptosis rate caused by CHD.

Fig. 3.

Comparative analysis of cell proliferation in each group. (A) Cell proliferation detected via the cell counting kit-8 (CCK-8) assay. (B) Cell proliferation was evaluated via flow cytometry. PE-H subset: positive peak, indicating how many (%) proliferating cells were contained. Compared to control, *** p 0.001. Compared to model, ## p 0.01, ### p 0.001. Compared to the model+sh-NC group, &&& p 0.001. n = 3.

Fig. 4.

Cell apoptosis detected through flow cytometry. Compared to control, *** p 0.001. Compared to model, ### p 0.001. Compared to the model+sh-NC group, &&& p 0.001. n = 3.

CD62P and CD62E are selectins. Studies have found that CD62P is significantly elevated in patients with acute coronary syndrome [18, 19, 20], while CD62E plays an essential role in inflammation [21]. Therefore, we measured the levels of CD62Pand CD62E after knocking down ZNF429. The immunofluorescence detection results are shown in Fig. 5. CD62P and CD62E levels in model and model+sh-NC groups were significantly increased vs control (p 0.001). Compared with the model group, there was no significant difference in the model+sh-NC group (p $>$ 0.05), but the levels of CD62P and CD62E in the model+shZNF429 group were greatly reduced (p 0.05). Additionally, CD62P and CD62E levels in the model+shZNF429 group were substantially lower vs model+sh-NC (p 0.05, Fig. 5A,B).

Fig. 5.

Immunofluorescence detection of CD62P and CD62E levels in cells. (A) The effect of interference with ZNF429 on the levels of CD62P in each group (400$\times$). (B) The effect of interference with ZNF429 on the levels of CD62E in each group (400$\times$). Scale bar = 50 µm. Compared to control, ** p 0.01, *** p 0.001. Compared to model, # p 0.05. Compared to model+sh-NC, & p 0.05. n = 3.