The Gene Expression Omnibus (GEO) is an international public repository where researchers globally share a wide range of genomic data, including microarray chips, second-generation sequencing, and other forms of high-throughput genomic data, for free download. Therefore, we chose to download MI-related bulk datasets, namely GSE104187, GSE153485, GSE206281, and a single-cell dataset GSE214611, from GEO. There are four samples in the GSE104187 dataset, including two MI samples and two normal samples; ten samples in the GSE153485 dataset, with five MI samples and five normal samples; six samples in the GSE206281 dataset, consisting of three MI samples and three normal samples; seven samples in the GSE214611 dataset, comprising four MI samples and three normal samples. Additionally, we sorted out 49 genes associated with cuproptosis from the published literature ( Supplementary Table 1 ) (1, 8, 9).

We integrated the three bulk datasets and subsequently carried out batch effect elimination using the SVA package (Version 3.48), following which we conducted data standardization using the limma package (Version 3.56).

In our effort to identify genes influencing MI progression, we performed a differential expression analysis on the integrated dataset using the amplicon package (Version 1.19), and set a criterion of |logFC|>0.2 & adjust. P-value<0.05 for screening differentially expressed genes (DEGs). The results were visually represented through volcano maps using the ggplot2 package (Version 1.16.0).

The clusterProfiler package (Version 4.6.2), org.Hs.eg.db package (Version 3.16.0) and ggplot2 package (Version 3.4.2) were used for performing GO and KEGG functional enrichment analysis of the DEGs, with parameters set as follows: p-value Cutoff<0.05, q-value Cutoff<0.05, and pAdjustMethod to Benjamini-Hochberg (BH). Venn diagrams were generated using the ggVennDiagram (Version 1.2). gseKEGG in clusterProfiler (Version 4.8) were applied for performing single-sample gene set enrichment analysis (ssGSEA).

In our analysis aimed at screening out key genes influencing MI progression, we employed the randomFores package (Version 4.7) for conducting random forest analysis. Setting the seed to 123 and the number of trees to 500, we finally selected the top seven genes based on their importance SD. At the same time, we employed the glmnet package (Version 4.1-7) for lasso analysis, specifying the family as “binomial”. Through binomial machine learning, we obtained six characteristic genes.

To investigate whether there is a correlation between key genes and immunity and whether immunity plays an important role in the disease, we conducted immunoinfiltration analysis using the CIBERSORT package.

We performed single-cell analysis on the GSE214611 dataset using the Seurat package (Version 4.3.0). Quality control of the single-cell data was conducted, and cells meeting the following criteria were retained: (1) cells with a gene count ranging from 200 to 5000; (2) cells with a mitochondrial ratio below 5%; (3) cells with an RNA count of less than 20,000.

For performing animal experiments, we purchased 12-week-old male C57BL/6 mice from Zhejiang Vital River Laboratory Animal Technology Co., Ltd. and raised them under constant temperature and humidity (23°C, 65%) within a 12/12-hour light/dark regime. All experimental procedures were conducted in strict adherence to relevant provisions outlined in the National Laboratory Animal Welfare Ethics, and this animal study was approved by the Laboratory Animal Welfare Ethics Committee, with approval number of ZJCLA-IACUC-20040153.

We randomly divided 20 mice into two groups: MI and Sham. Nine mice survived in the MI group, while ten survived in the Sham group. Afterwards, three mice from each group were randomly selected for TTC staining, and six for related molecular experiments. In the MI group, mice underwent endotracheal intubation, followed by ligation of the LAD coronary artery to induce MI. After inducing anesthesia until the muscle strength decreased to level 0, the mice were taken out and fixed on a sticky mouse board. Anesthesia was maintained by pressing the nasal cavity against the mask. A small incision, approximately 1.5cm above the heart, was made in the left chest 3-4 rib space using small scissors. Using a 6-0 silk thread, the anterior descending branch was ligated at 2 mm below the left atrial appendage, tightening the infarct blood vessel and muscle. The needle insertion depth was about 1.5mm. The left auricle displayed purplish-red congestion, and the myocardium gradually changed from bright red to pale after vasculature, with gradual weakening of contraction. By connecting the limb leads of the small animal electrocardiogram detection system, if the electrocardiogram revealed an upward arching of the ST segment, accompanied by the appearance of J waves after the ST segment, the successful establishment of the MI model was indicated. After turning off the anesthetic mask, the skin on both sides was aligned using 3-0 silk for pouch sutures to close the chest. The incision was disinfected with alcohol. Mice were positioned prone on a specialized small animal special electric blanket to facilitate awakening and rewarming. In the Sham group, mice underwent an identical procedure, excluding the step of coronary artery ligation.

Human cardiomyocytes AC16 (Pricella, CL-0790) were cultured in DMEM/F12 medium containing 10% fetal bovine serum at 37°C in an incubator with 5% CO₂. To create the cell model, we exposed the cells to oxygen-glucose deprivation (OGD), during which they were cultured in sugar-free DMEM/F12 medium in a hypoxic incubator at 37°C with 94% N₂, 1% O₂, and 5% CO₂ for 24 h. Cells in the control group were cultured under standard conditions.

The collected mouse heart tissue sections were immersed into 2% TTC solution in a light-protected environment, and incubated in a 37°C incubator for 30 min. To ensure complete staining, the container was gently agitated every five minutes. Subsequently, images were captured to observe the heart tissue infarction area.

The heart tissue sections were subjected to HE staining using relevant staining kit (Beyotime, C0105M) to visualize and analyze the morphological characteristics of the heart tissue.

We used DLAT Rabbit pAb (ABclonal, A14530) to detect Dlat oligomerization in the mouse heart tissue or cardiomyocytes.

We extracted RNA from mouse heart tissue and cardiomyocytes using the Trizol reagent (Beyotime, R0016). Subsequently, cDNA synthesis was carried out using the SuperScript™ VILO™ cDNA Synthesis Kit (Invitrogen, 11754050). The quantitative PCR procedure was executed according to the instructions provided by the BeyoFast™ SYBR Green qPCR Mix (Beyotime, D7501). The primers used in this experiment are mmu-Ube2d3-F: CACGAGAACAGTTGCCCAAGATTC; mmu-Ube2d3-R: CAGCATGTTAGAGCAGTGAGGATTG; hsa-Ube2d3-F: GGGCATTACCATTCTTCACCTTTCC; hsa-Ube2d3-R: GTGGAAGTGGAGGAGTCACAGATG; mmu-GAPDH-F: GGCAAATTCAACGGCACAGTCAAG; mmu-GAPDH-R: TCGCTCCTGGAAGATGGTGATGG; hsa-GAPDH-F: GATCATCAGCAATGCCTCCT; hsa-GAPDH-R: TGTGGTCATGAGTCCTTCCA.

We extracted the total protein from mouse heart tissue and cardiomyocytes with a lysis buffer containing 1% protease inhibitor. Protein quantification was carried out using the BCA method (Beyotime, #P0012S), followed by subsequent procedures including electrophoresis, membrane transfer, antibody incubation, and color development. The primary antibodies included Ube2d3 (Invitrogen, PA5-119881), FDX1 (Abcam, ab108257), FDX1 (Invitrogen, PA5-59653), SLC31A1 (Invitrogen, PA1-16586), Dlat (ABclonal Technology, A8814), Dlat (Invitrogen, PA5-59298) and Lipoic acid (abcam, ab58724). GAPDH (Invitrogen, PA1-988) was selected as the internal reference protein.

To assess neutrophil infiltration in the mouse heart tissue, we used the Anti-Neutrophil antibody (abcam, ab53457) for immunofluorescence analysis.

The expression of respective inflammatory factors in the mouse heart tissue was assessed using ELISA assay kits for IL-6 (Beyotime, PI326), IL-1β (Beyotime, PI301), TNF-α (Beyotime, PT512), and MCP-1 (Beyotime, PC125).

Cell viability was assessed following the instructions provided by the MTT kit (Beyotime, C0009S).

Cardiomyocyte damage was assessed following the instructions provided by the LDH Cytotoxicity Assay Kit (Beyotime, C0017). Cell death was evaluated according to the instructions provided by the One Step TUNEL Apoptosis Assay Kit (Beyotime, C1086).

Statistical analysis in the bioinformatics section was carried out using the R language. In the experimental part, IBM SPSS Statistics software was used for statistical data analysis, and GraphPad Prism 9 was subsequently used to draw bar charts. Results were expressed as mean ( x ¯ ) ± standard deviation (SD), with all measurement data conforming to a normal distribution. Data comparison between groups was performed using t-tests. P<0.05 indicates a statistically significant difference.

Initially, we performed differential analysis on CRGs in the three downloaded bulk datasets. This analysis revealed significant differences in 26 CRGs in the GSE153485 dataset ( Figure 1A ), five CRGs in the GSE206281 dataset ( Figure 1B ), and four CRGs in the GSE104187 dataset ( Figure 1C ). To investigate DEGs affecting MI development, we merged the three bulk datasets, performed batch effect removal ( Supplementary Figures 1A, B ) and conducted data normalization ( Supplementary Figures 1C, D ).

To identify genes that influence MI progression, we analyzed the differences in the combined dataset, comparing samples from the disease and control groups. A total of 5543 DEGs were identified, comprising 3097 down-regulated and 2446 up-regulated ones ( Figure 2A ). Subsequent functional enrichment analysis categorized these DEGs based on their functions, revealing their primary enrichment in ribonucleoprotein complex, actin cytoskeleton, actin binding and PI3K-AKT signaling pathway ( Figures 2B, C ). Given the focus on cuproptosis in this study, we aimed to screen out genes associated with cuproptosis that might influence MI progression. To achieve this goal, we plotted a Venn diagram illustrating the overlap of DEGs with CRGs, resulting in the identification of 16 cuproptosis-related DEGs ( Figure 2D ), and a heatmap was subsequently generated to visualize their gene expression patterns in each group ( Figure 2E ).

To further identify key cuproptosis-related DEGs that affect MI, two different machine learning algorithms were applied: LASSO algorithm and random forest algorithm. Through lasso analysis, we screened out six characteristic genes ( Figures 3A, B ), while through random forest analysis, we identified 16. To narrow down the selection of key genes, we selected the top seven genes with the largest importance SD value for subsequent analysis ( Figure 3C ). Then, we obtained four hub genes, Dlat, Ube2d1, Ube2d3, and Dbt, by taking intersection of the results of the two machine learning analyses ( Figure 3D ), and plotted their expression boxplots ( Figure 3E ). To understand the pathways regulated by these four key genes, ssGSEA analysis was conducted. The results revealed that Dlat ( Supplementary Figure 2A ), Ube2d1 ( Supplementary Figure 2B ), Ube2d3 ( Supplementary Figure 2C ), and Dbt ( Supplementary Figure 2D ) may all be involved in Citrate cycle (TCA cycle). This outcome suggests their potential involvement in cuproptosis.

We performed an immunoinfiltration analysis on the datasets and presented the differences in 22 types of immune cells between the disease and the control groups ( Figure 4A ), revealing significant differences in plasma cells, NK cells activated, macrophages M0, macrophages M1, macrophages M2, dendritic cells activated, mast cells resting, mast cells activated and neutrophils between groups. Meanwhile, we analyzed the correlation between the hub genes and immune cells, as well as between different immune cells ( Figure 4B ). The results showed that both Ube2d3 and Dlat were significantly positively correlated with neutrophils.

We performed single-cell analysis on the GSE214611 dataset. Quality control procedures were applied to the cells within the dataset ( Supplementary Figures 3A, B ). All the cells were subsequently classified into 21 cell clusters ( Figure 5A ), and subjected to cell annotation ( Figure 5B ), resulting in the identification of five major cell types: B cells, Endothelial cells, Fibroblasts, Macrophages, and Stromal cells. Notably, Ube2d3 showed significant differences in Endothelial cells, Fibroblasts, Macrophages, and Stromal cells ( Figure 5C ), and it displayed the most widespread distribution among these cell types ( Figure 5D ).

Based on the results of single-cell analysis as well as relevant literature (10), we selected the CRG Ube2d3 for follow-up studies. To elucidate the mechanism of Ube2d3 in MI, we constructed a mouse model of MI. Through TTC staining, we observed relatively obvious white infarct area in the heart tissue of mice in the MI group ( Figure 6A ). Histological examination via HE staining revealed evident myocardial cell necrosis, edema, neutrophil infiltration, and arrangement disorder in the myocardial tissue of mice in the MI group, whereas mice the Sham group displayed neatly arranged fibers with only a minimal degree of neutrophil infiltration within the myocardial tissue ( Figure 6B ). Furthermore, according to immunofluorescence results, substantial neutrophil infiltration was observed in the myocardial tissue of mice in the MI group ( Figure 6C ). Besides, we also measured the expression of inflammatory factors, including IL-6, IL-1β, TNF-α and MCP-1, in the myocardial tissue. The results indicated a significant elevation of these inflammatory factors in the MI group ( Figure 6D ), signifying a state of immune activation in the myocardial tissue. Through qRT-PCR and WB detection, we found a notable upregulation in the expression of Ube2d3 in the myocardial tissue in the MI group ( Figures 6E, F ), which was consistent with the results of our bioinformatics analysis. Additionally, the expression of cuproptosis-related proteins, FDX1 and SLC31A1, as well as the lipoylation level of Dlat in the myocardial tissue of mice was found to be significantly increased in the MI group ( Figures 6G, H ). Meanwhile, we also observed a significant aggregation of Dlat ( Figure 6I ). The above results collectively suggest that cuproptosis may indeed play an important role in the progression of MI.

We treated AC16 cells with OGD, and measured Ube2d3 expression levels via qRT-PCR and WB detection. As a results, Ube2d3 expression was revealed to be significantly elevated in the OGD group ( Figures 7A, B ), which aligned with our bioinformatics analysis results. Concomitantly, we observed a significant reduction in cell viability in the OGD group ( Figure 7C ) and a remarkable increase in cell death rates ( Figures 7D, E ) based on relevant test results. Additionally, the expressions of FDX1 and SLC31A1, as well as the lipoylation level of Dlat were significantly increased in the OGD group ( Figures 7F, G ). Meanwhile, we also observed a significant aggregation of Dlat ( Figure 7H ). These results suggest that during MI development, Ube2d3 is closely related to cuproptosis and can aggravate hypoxic injury of cardiomyocytes.

To further investigate the mechanism of action of Ube2d3, we knocked down the expression of Ube2d3 in OGD-treated AC16 cells ( Figures 8A, B ), and found increased cell activity ( Figure 8C ), significantly reduced number of cell death ( Figures 8D, E ), markedly lowered expressions of FDX1 and SLC31A1, as well as obviously decreased lipoylation level of Dlat ( Figures 8F, G ) in the OGD+si-Ube2d3 group when compared with the OGD+si-NC group. Moreover, Dlat did not display significant aggregation ( Figure 8H ) in the OGD+si-Ube2d3 group. Therefore, we infer that Ube2d3 can promote hypoxic damage in cardiomyocytes AC16 by regulating cuproptosis.