Identification of a novel immune infiltration-related gene signature, MCEMP1, for coronary artery disease

Screening and enrichment analysis of differentially expressed genes (DEGs)

The GEO DataSets database stores curated gene expression DataSets, as well as original Series and Platform records in the GEO repository (https://www.ncbi.nlm.nih.gov/gds/?term=). CAD-related gene expression data (GSE24519, GSE61145, and GSE113079) was obtained from the GEO DataSets database. The GSE24519 and GSE61145 datasets, used as the metadata datasets, included 67 CAD patient samples and 21 normal control samples, and the GSE113079 dataset, used as a validation dataset, included 93 CAD patient samples and 48 normal control samples. All samples from these databases were derived from human peripheral blood.

The R “SVA” package (https://rdrr.io/bioc/sva/src/R/sva-package.R) was used for background correction and standardization of the metadata. DEGs between CAD samples and normal samples were identified using the R “limma” package (https://bioconductor.org/packages/release/bioc/html/limma.html) with a significance threshold of p < 0.05 and |log2 FC| > 1. Volcano and heat maps were generated using the online SangerBox software (http://sangerbox.com/). Gene ontology (GO) and pathway analyses were conducted to identify the biological functions of DEGs, performed using the R “clusterProfiler” package (https://guangchuangyu.github.io/software/clusterProfiler/) and Metascape database (http://metascape.org).

Selection and verification of hub genes

Two machine learning algorithms—support vector machine-recursive feature elimination (SVM-RFE) and least absolute shrinkage and selection operator (LASSO)—were applied to search for CAD-related gene signatures in the metadata cohort. The SVM-RFE algorithm (https://github.com/johncolby/SVM-RFE) was implemented using the R “mlbench” and “caret” packages, while the LASSO model (https://glmnet.stanford.edu/) was executed by the R “glmnet,” “survival,” and “survminer” packages. A Venn diagram was plotted using the R “VennDiagram” package to display the intersection of hub genes filtered by both the SVM-RFE and LASSO algorithms.

The GSE113079 dataset was employed to validate the expression of overlapping gene signatures between CAD patients and the control group.

Association between CAD-related candidate gene signatures and immune-related features

The CIBERSORT program (https://cibersortx.stanford.edu/) was utilized to quantify the proportion of immune cell infiltration (22 different cell types) in CAD gene expression profiles using the R “ggplot2,” “corrplot,” and “forcats” packages. A gene set enrichment analysis (GSEA, https://www.broadinstitute.org/gsea/) was conducted to explore the underlying functional mechanisms of the hub genes using the R “ggplot2,” “clusterProfiler,” “enrichplot,” and “ggridges” packages.

HUVEC culture and treatment

Human umbilical vein endothelial cells (HUVECs) were purchased from iCell Bioscience (iCell-h110; iCell Bioscience, Shanghai, China). These HUAECs were cultured within an endothelial cell growth medium (PriMed-iCell-002, Shanghai, China) supplemented with 5% (v/v) FBS, 1% endothelial cell growth supplement (ECGS), and 1% penicillin streptomycin at 37 °C in 5% CO₂. HUVECs within passages 3–7 were applied for the experiments. Then, human ox-LDL (Yiyuan Biotechnologies, Guangzhou, China) at concentrations of 50 and 100 μg/mL were added into the cell cultures according to the experimental requirements (Kang et al., 2021; Di et al., 2021).

Knockdown interference sequences for MCEMP1 were provided by GenePharma (Shanghai, China). The optimal working concentration of these interference sequences was selected experimentally prior to the formal experiments. Theinterference sequence information for MCEMP1 is as follows: siMCEMP1#1, sense 5′-GACCCAGACUAUGAGAAUATT-3′, anti-sense 5′- UAUUCUCAUAGUCUGGGUCTT-3′, siMCEMP1#2, sense 5′-CAGCCUUCAUCAUGGUGAATT-3′, anti-sense 5′-UUCACCAUGAUGAAGGCUGTT-3′, siMCEMP1#3, sense 5′-GAGCAAGAUUGAUAGAUUATT-3′, anti-sense 5′-UAAUCUAUCAAUCUUGCUCTT′.

Cell proliferation and apoptosis assay

After a 24-h treatment with ox-LDL, 2 × 10³ cells (100 μL) were inoculated within each well of the 96-well culture plate. About 3–4 h later, when cells had adhered to the wall, 100 μL of RPMI 1,640 complete medium and 10 μL of CCK-8 were added. Then, the cells were cultured in 5% CO₂ at 37 °C for 2 h. The OD450 values were determined with a microplate reader (Infinite M200, Tecan, Salzburg, Austria) at the wavelength of 450 nm.

The Muse^® Cell Analyzer (Milipore, Burlington, MA, USA) was used to assess the effects of MCEMP1 on apoptosis in ox-LDL treated HUVECs. The cells were suspended in 100 μL of 1% BSA, and an equal volume of apoptosis working fluid was added. Apoptosis was detected after 20 min of gentle resistance. HUVECs were harvested after the indicated treatments, re-suspended at a concentration of 10⁶/mL, and stained with a Muse Annexin V & Dead Cell Kit in darkness. The Muse^® Cell analyzer (CogtoFLEX, BECKMAN) was used to quantify the percentage of apoptotic cells. The apoptotic rate was defined as the percentage of “UR+LR” cells.

Lactate dehydrogenase (LDH) and nitric oxide (NO) assay

HUVECs were cultured in 96-well plates at a density of 1 × 10⁴ cells/mL, and pretreated with ox-LDL (100 μg/mL) for 48 h. The supernatants from ox-LDL treated HUVEC were collected by centrifugation at 5,000 g for 10 min. Then, the LDH content and NO content in the media was assessed using an LDH activity kit (BC0685; Solarbio, Beijing, China) and a NO assay kit (AKNM005C; Boxbio, Beijing, China), respectively.

Quantitative real-time polymerase chain reaction (qRT-PCR)

Total RNA was extracted from the HUVECs using TRIzol reagent (Accurate Biology, Changsha, China). The RNA was reverse-transcribed into cDNA using a PrimeScript RT Reagent Kit (TakaRa Biotechnology, Dalian, China). Real-time PCR was performed using SYBR green fluorescence (TakaRa) for the relative quantification of mRNA expression. Quantification was achieved using the 2^−ΔΔCt method, with GAPDH used to normalize mRNA levels. The following primers were used in this study: DIRC2 forward primer: 5′-GTCCATAGTTGGAGGCTGTGTTG-3′ and reverse primer: 5′-ACGTGGATGACAGTGTAGCTCC-3′; MCEMP1 forward primer: 5′-CTGAGATGTCCAAGGAGCTGCT-3′ and reverse primer: 5′-TGGTGATGCTCTGCTGAACGGA-3′; GAPDH forward primer: 5′-GGAGCGAGATCCCTCCAAAAT-3′ and reverse primer: 5′-GGCTGTTGTCATACTTCTCATGG-3′.

Western blotting

Proteins were extracted from HUVECs using RIPA lysis buffer (Solarbio, Beijing, China) for 20 min, followed by centrifugation at 12,000 rpm for 15 min at 4 °C to collect the proteins. Protein concentrations were measured by a BCA protein assay kit (Beyotime, Shanghai, China). Equal amounts of proteins (60 μg) and a pre-stained protein ladder (YESEN, Shanghai, China) were separated by 10% SDS-PAGE gel, electro-transferred to nitrocellulose membranes, and incubated with primary antibodies overnight at 4 °C. The anti-DIRC2 antibody (1:1,000, 26750-1-AP; Proteintech), anti-MCEMP1 antibody (1:3,000, 25700-1-AP, proteintech), anti-ICAM1 antibody (1:500, 200350-F12; ZEN BIO), and anti-GAPDH antibody (1:2,000, 60004-1-Ig; Proteintech) were used in this study. The membranes were incubated with secondary antibodies for 1.5 h at room temperature. Bands with antigen-antibody complexes were visualized using Immobilon ECL substrate (YESEN, Shanghai, China), and the blots were imaged with an E-BLOT luminescent image analyzer (e-BLOT, Shanghai, China).

Enzyme-linked immunosorbent assay (ELISA)

HUVECs were cultured in 96-well plates at a density of 1 × 10⁴ cells/mL and pretreated with ox-LDL (100 μg/mL) for 48 h. The cultured HUVEC supernatants were collected by centrifugation at 5,000 g for 10 min. The levels of IL-1β, IL-6, and TNF-α were detected using respective ELISA kits (Jianglai Biotechnology, Shanghai, China).

Screening and enrichment analyses of DEGs in CAD

In our study, a comprehensive analysis was conducted to identify DEGs in CAD. A total of 73 DEGs (four down-regulated genes and 69 up-regulated genes, Table S1) were discovered between 67 CAD patient samples and 21 normal control samples from the metadata cohorts (GSE24519 and GSE61145) cohort. Volcano (Fig. 1A) and heatmap (Fig. 1B) plots depicted these DEGs.

Further functional characterization of these DEGs was performed using GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. The GO analysis (Fig. 1C) results indicated potential associations of DEGs with a range of biological processes, including the regulation of platelet aggregation, defense response or response to bacterium, regulation of aspartic-type endopeptidase activity involved in the amyloid precursor protein catabolic process, regulation of homotypic cell-cell adhesion, and positive regulation of cytokine production. Additionally, the KEGG pathway analysis (Fig. 1D) revealed significant involvement of DEGs in pathways such as the NF-kappa B signaling pathway, lipid and atherosclerosis, the B cell receptor signaling pathway, the IL-17 signaling pathway, among others.

Selection and verification of hub genes

To identify key genes that might serve as potential biomarkers or therapeutic targets, SVM-RFE and LASSO logistic regression algorithms were employed. The SVM-RFE algorithm identified 18 gene signatures in the determined DEGs (Fig. 2A), while the LASSO logistic regression algorithm recognized 16 gene signatures (Fig. 2B). Through intersection analysis, seven overlapping hub genes (HSDL2, PGLYRP1, CAMP, MCEMP1, IRAK3, DIRC2, and PFKFB3) were finally selected in the metadata cohort (Fig. 2C).

The GSE113079 dataset (93 CAD patient samples and 48 normal samples) was used to validate the accuracy of the approach and the expression levels of the seven hub genes (Fig. 2D). Notably, the expression levels of PGLYRP1 and IRAK3 were not detected in the validation set. Furthermore, CAMP and HSDL2 expression levels did not show significant differences between CAD and normal samples. Interestingly, PFKFB3 expression was found to be down-regulated in CAD samples, contrasting with the metadata results. However, DIRC2 (p = 0.011) and MCEMP1 (p = 0.00047) were significantly up-regulated in CAD samples compared to control samples, aligning with the metadata. Therefore, DIRC2 and MCEMP1 were selected for further investigation in the follow-up studies.

Correlation of DIRC2 and MCEMP1 with immune cell infiltration and signal pathways

In our study, we investigated the correlation between the expression levels of two key genes, DIRC2 and MCEMP1, and the infiltration of immune cells in CAD tissues. The proportions of 22 different types of immune cell infiltrates were calculated and compared between CAD and normal tissue (Fig. S1A). Our findings revealed that the degree of infiltration of regulatory T cells (Tregs, p = 4.3e-4), resting dendritic cells (p = 8.4e-4), and resting mast cells (p = 0.03) in CAD tissues was lower than in normal tissues. Conversely, CAD samples displayed a higher infiltration of activated mast cells (p = 6.3e-3) and neutrophils (p = 1.2e-3). Additionally, the correlation among these infiltrating immune cells was calculated (Fig. S1B), showing that activated mast cells had a negative correlation with resting mast cells but a positively associated with neutrophils.

The relationships between the two hub genes (DIRC2 and MCEMP1) and the 22 types of immune cell infiltrates in CAD were further explored (Figs. 3A and 3B). DIRC2 was negatively associated with regulatory T cells (Tregs, p < 0.001), CD8 T cells (p = 0.001), resting dendritic cells (p = 0.015), and CD4 memory activated T cells (p = 0.033), but positively associated with activated NK cells (p = 0.028) and neutrophils (p = 0.001, Fig. 3A). Similarly, MCEMP1 was negatively associated with regulatory T cells (Tregs, p < 0.001), resting dendritic cells (p = 0.009), and CD4 memory activated T cells (p = 0.020), but positively associated with macrophages M0 (p = 0.048), CD4 naive T cells (p = 0.028), monocytes (p = 0.005), activated mast cells (p = 0.003), and neutrophils (p = 0.002, Fig. 3B).

To further understand the potential roles of DIRC2 and MCEMP1 in the pathogenesis of CAD, GSEA was used to identify the related signal pathways (Figs. 3C and 3D). DIRC2 was positively linked to Kaposi sarcoma-associated herpesvirus, salmonella infection, shigellosis, and tuberculosis, and negatively linked to ribosomes (Fig. 3C). On the other hand, MCEMP1 was positively linked to lipid and atherosclerosis, lysosomes, non-alcoholic fatty liver disease, phagosomes, and Yersinia infection (Fig. 3D). These findings indicate that DIRC2 and MCEMP1 may play an important role in the pathogenesis of CAD, potentially through their involvement in these immune-related signaling pathways.

The expression of DIRC2 and MCEMP1 in ox-LDL-treated HUVECs

In ox-LDL-treated HUVECs, the HUVECs were exposed to ox-LDL (50 and 100  μg/mL) for 24 h. The impact on cell viability was assessed using a CCK-8 assay, which indicated that the cultured HUVECs displayed reduced cell viability after both 50 and 100 μg/mL ox-LDL treatment (Fig. 4A). Additionally, ox-LDL induced cellular injury in HUVECs, as evidenced by remarkably increased LDH and decreased NO in all tested concentrations (Figs. 4B and 4C). Moreover, the expression of ICAM1 was upregulated in response to both concentrations of ox-LDL (Fig. 4D).

Western blot assay and qRT-PCR revealed that both the mRNA and protein levels of DIRC2 and MCEMP1 were upregulated in ox-LDL-treated HUVECs (Fig. 5), which was consistent with the results obtained from bioinformatics analysis. Based on these results, the MCEMP1 gene with a high expression level was selected for subsequent experiments.

Low expression of MCEMP1 promoted the dysfunction of ox-LDL-treated HUVECs

To explore the role of MCEMP1 in the dysfunction of ox-LDL-treated HUVECs, interference sequences targeting MCEMP1 were synthesized. The interference efficiency was evaluated in HUVECs, as shown in Fig. S2. Both siMCEMP1-1 and siMCEMP1-2 were selected for subsequent experiments. Interference sequences for MCEMP1 significantly reduced the ox-LDL-induced expression of MCEMP1, especially siMCEMP1-1 and siMCEMP1-2 (both p < 0.001, Figs. 6A and 6B).

The effect of MCEMP1 knockdown on the function of ox-LDL-treated HUVECs was investigated. The results, shown in Figs. 6C–6E, indicated that MCEMP1 knockdown promoted cell proliferation (p < 0.01, Fig. 6C) but inhibited apoptosis (p < 0.001, Fig. 6D). Additionally, MCEMP1 knockdown appeared to decrease the expression of inflammatory cytokines IL-1β, IL-6, and TNF-α (Fig. 6E). Collectively, these findings suggested that MCEMP1 knockdown could inhibit the inflammation response in ox-LDL-treated HUVECs.