Mendelian randomization and transcriptome analysis reveal depression-driven regulatory patterns of the immune microenvironment in myocardial infarction and heart failure

Figures

• 

  Figure 1. MR analysis of MHD and CVD. (A) Influence of MHD traits on MI by MR models. (B) Influence of MHD traits on HF by MR models. Effect estimates represent the change in outcome risk per unit increase in the log-odds of genetic liability to the exposure. (C) Reverse MR of MDD and MI. (D) Reverse MR of MDD and HF.

• 

  Figure 2. NHANES cohort analysis of depression and CVD. (A) RCS model of depression and HF. (B) RCS model of depression and CAD. (C) RCS model of depression and AP. (D) RCS model of depression and MI. (E) Logistic regression associations of depression with CVD in adults.

• 

  Figure 3. Identification of DEGs. (A) PCA of genes associated with healthy and MDD samples before batch effect removal. (B) PCA of genes associated with healthy and MDD samples after batch effect removal. (C) The volcano plot displays the expression changes of DEGs between healthy and MDD. (D) Heatmap shows 20 significantly dysregulated differentially expressed genes between healthy and MDD samples. (E) The volcano plot displays the expression changes of DEGs between MI patients at admission and at 6 months post-discharge. (F) The volcano plot displays the expression changes of DEGs between MI patients at admission and stable CAD patients without a history of MI. (G) Venn diagram shows the number of genes commonly upregulated and commonly downregulated between the two MI groups. (H) The volcano plot displays the expression changes of DEGs between HF and non-HF groups of STEMI patients. (I) Heatmap shows 20 significantly dysregulated differentially expressed genes between HF and non-HF groups of STEMI patients.

• 

  Figure 4. Identification of MDD-related genes in MI and HF through machine learning approaches. (A) Venn diagram shows 202 MDD-related genes commonly dysregulated in both MI and HF. (B, C) Parameter selection for the LASSO model in MI (B) and HF (C). The left panels display the coefficient profiles, illustrating how coefficients of each gene shrink to zero as the L1 norm penalty increases. The right panels show the 10-fold cross-validation curves (binomial deviance vs. log(λ)). The vertical dotted lines mark the optimal λ values based on the minimum criteria (λmin) and the 1-standard-error criteria (λ1se). (D) Venn diagram shows the 6 MDD marker genes shared by MI and HF. (E, F) Boxplot comparing the risk scores of the LASSO model between controls, MI and HF. (G, I) ROC curve analysis of 6 MDD marker genes in MI and HF. (H, J) ROC curve analysis of the 6 MDD marker genes in the external validation cohorts of MI and HF.

• 

  Figure 5. Independent transcriptomic validation and diagnostic evaluation of the RS model in MI. (A) Volcano and the heatmap plot of the DEGs from control and MI groups. (B) KEGG enrichment analyses based on the DEGs between control and MI groups. (C) GSEA analysis of KEGG pathways. (D) Box plots showing the expression of MDD marker genes in control and MI patients. (E) Box plot showing the distribution of risk scores between control and MI patients. (F) ROC curve analysis evaluating the diagnostic efficacy of the Risk Score (RS) model (trained on the GSE59867 discovery dataset) in the independent internal MI cohort.

• 

  Figure 6. Analysis of the immune microenvironment in CVD and its correlation with MDD marker genes. (A, B) Differential abundance of infiltrating immune cells between control, MI, and HF groups was quantified using the CIBERSORT algorithm. (C, D) Differential enrichment of immune response gene sets between control, MI, and HF groups was analyzed using ssGSEA. (E, F) Comparison of HLA gene set expression levels between control, MI, and HF groups. (G–L) The dot-plot demonstrate the correlations between each dysregulated infiltrating immune cells, immune response gene sets and HLA gene, and each dysregulated autophagy gene. (G, J) Spearman correlation analysis of the six MDD marker genes with infiltrating immune cell abundance in MI (G) and HF (J). (H, K) Spearman correlation analysis of the six MDD marker genes with immune response gene set scores in MI (H) and HF (K). (I, L) Spearman correlation analysis of the six MDD marker genes with HLA family gene expression in MI (I) and HF (L).

• 

  Figure 7. Identification of MDD-mediated patterns in CVD. (A, F) Cumulative distribution function curves of MI and HF for k = 2–6 clusters. (B, G) Relative change in the area under the Cumulative distribution function curve of MI and HF for k = 2–6 clusters. (C, H) Heatmap of the consensus matrices for MI and HF samples. (D, I) Principal component analysis of the transcriptomes under two MDD regulatory patterns in MI and HF, showing a clear distinction in transcriptomic features. (E, J) Box plots showing the expression differences of 6 MDD marker genes under different metabolic regulatory patterns in MI and HF.

• 

  Figure 8. Differences in the immune microenvironment among CVD subtypes driven by MDD. (A, B) Differences in infiltrating immune cell abundance under two MDD regulatory patterns in MI and HF. (C, D) Differences in immune response gene set activity between the two MDD regulatory patterns in MI and HF. (E, F) Differences in HLA gene expression between the two MDD regulatory patterns in MI and HF.

• 

  Figure 9. Biological functions underlying CVD expression patterns. (A) GO-BP enrichment analysis of genes shared across CVD subtypes. (B) KEGG enrichment analysis of genes shared across CVD subtypes. (C, D) The 16 most significantly different HALLMARKS pathways between two MDD-related expression patterns identified by GSVA analysis. (E, F) The 20 most significantly different KEGG pathways between two MDD-related expression patterns identified by GSVA analysis.