Introduction

Heart failure (HF) is a complicated chronic clinical syndrome characterized by worsening symptoms and signs caused by cardiac dysfunction as well as a major etiology of mortality, morbidity, and poor quality of life1,2. Meanwhile, cancer has a high morbidity and mortality rate globally and is the leading cause of non-cardiovascular mortality in patients with HF3,4. HF is strongly related to cancer due to the large overlap in risk factors, such as smoking, sedentary lifestyle, and obesity. Besides, HF and cancer share several pathophysiological pathways, such as inflammation response, immune disorder, genetic mutations, metabolic remodeling, and microbial composition4. However, no effective therapies exist for patients with HF complicated by tumors, leading to challenges in managing both conditions effectively. Consequently, it is imperative to gain deeper insights into the potential mechanisms and shared therapeutic targets involved in both HF and cancer. Several investigations have indicated that the immune infiltration, particularly T cells in adaptive immune responses, is highly important for the advancement of hypertrophic cardiomyopathy (HCM)-induced HF5. Among them, cytotoxic CD8+T-cells rather than CD4+T-cells are the predominant type of adaptive immune cells that play a protective role in HCM-induced HF5,6. For example, adoptive transfer of antigen-specific cytotoxic CD8+T-cells effectively alleviates cardiac fibrosis and restores cardiac dysfunction6. Cytotoxic CD8+T-cells promote the translation of cardiac macrophages into cardioprotective macrophages in heart under conditions of pressure overload7. However, CD4+T-cells promote the transition from cardiac hypertrophy to HF8. Consequently, immunotherapeutic approaches based on cytotoxic CD8+T-cells for the treatment of HF are urgently needed.

Immunotherapies based on cytotoxic CD8+T-cells exert important effects on inhibiting tumor progression9. Mechanistically, tumor cells can be efficiently eliminated by cytotoxic CD8+T-cells through the secretion of various cytotoxic molecules, including granzyme B, perforin, tumor necrosis factor-α, and interferon-γ10. The spatiotemporal exhaustion of cytotoxic CD8+T-cells within tumor microenvironment (TME) facilitates tumor escape11. The presence of inhibitory molecules, such as cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1), in cytotoxic CD8+T-cells within TME leads to poor clinical prognosis for cancer patients9. The clinical success of immune checkpoint blockade and adoptive CD8+T-cell therapy demonstrates the potential of CD8+T-cells in mediating anti-tumor responses. While the contribution of CD4+T-cells to tumor progression is widely variable, and can either promote malignancies owing to their intrinsic B cell helper activity or participate in the immune response against solid tumors12. Consequently, immunotherapeutic approaches based on cytotoxic CD8+T-cells for the treatment of diverse tumors are needed.

In recent years, bioinformatics analysis has played a crucial role in revealing the underlying mechanisms of various diseases. Weighted correlation network analysis (WGCNA) is a widely utilized tool that helps identify highly disease-correlated genes and potential therapeutic targets13. Moreover, CIBERSORT (http://cibersort.stanford.edu), an immune cell composition analysis tool for high throughput sequencing data, aids in understanding immune infiltration in complex tissues14. Despite their applications in studying a single disease15,16, their use in exploring both HF and pan-cancer has remained limited.

We obtained the GSE141910 dataset that consisted of individuals with HCM-induced HF and non-HF from the Gene Expression Omnibus (GEO) database. We employed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) for analyzing the enrichment of differentially expressed genes (DEGs) in the dataset. Our subsequent analysis involved the application of WGCNA to this dataset to determine the modules that exhibited the strongest association with the infiltration levels of CD8+T-cells and to further pinpoint the hub genes. Receiver Operating Characteristic (ROC) curves were generated to evaluate the potential diagnostic significance of hub genes for HF.

This study represents a significant milestone, as it marks the first instance of integrating WGCNA with immune infiltration analysis for the purpose of identifying potential therapeutic targets associated with CD8+T-cells in HF while simultaneously investigating their immune mechanisms in both HF and diverse types of cancers. Concomitantly, the association between hub gene expression and immune infiltration, as well as between hub gene expression and both tumor mutation burden (TMB) and microsatellite instability (MSI), have been extensively explored for the purpose of functionally validating hub genes. Cox risk analysis identified hub genes as beneficial prognostic factors in skin cutaneous melanoma (SKCM) patients. Survival curves revealed a positive correlation between the expression levels of hub genes and disease-specific survival (DSS) outcomes in SKCM. Finally, we validated the overexpression of hub genes in heart samples from mice and patients with cardiac hypertrophy-induced HF through molecular biology experiments. A flowchart of the overall data analysis is shown in Fig. 1. This study not only contributes to the understanding of potential link between HF and cancer but also proposes theoretical strategies for patients with both conditions. The identification of new therapeutic options, such as Nelarabine and Methylprednisolone, holds promise for improving the treatment outcomes of patients, particularly those with SKCM and HF.

Figure 1
figure 1

A flowchart of overall data screening strategy.

Results

Differential analysis and GO and KEGG analysis

A total of 7,678 genes exhibited significant difference between the samples of patients with non-HF and those with HF, among which 3,663 genes were upregulated while 4,015 genes were downregulated. The KEGG analysis revealed that the upregulated genes were predominantly enriched in pathways related to the immune system, including Th17 cell differentiation, Th2 cell differentiation, Th1 and T-cell receptor, Toll-like receptor, Wnt, and PI3K-Akt signaling pathway (Fig. 2a). Additionally, GO analysis highlighted the crucial role of immune response-activating cell surface receptor signaling pathway in HF (Fig. 2b).

Figure 2
figure 2

Significant upregulation of immune response related pathways in patients with HF. (a) KEGG pathway analysis of DEGs. (b) GO enrichment analysis of DEGs.

Evaluation of immune cell infiltration in HF

To investigate the infiltration of immune cells in GSE14190 dataset, we employed the R package “CIBERSORT” for analyzing the RNA sequencing (RNA-seq) profile. The reference dataset “LM22” consisted of bulk RNA data from 22 purified immune cells. By using a deconvolution algorithm, we were able to determine the proportion of various immune cell subtypes in individual samples within the GSE14190 dataset by comparing them with the reference dataset. Notably, patients with HF exhibited significant difference in the infiltration levels of CD8+T-cells and T follicular helper cells (Fig. 3a). Furthermore, a correlation analysis was performed among different types of immune cells. CD8+T-cells were predominantly associated with activated NK cells that were involved in antiviral infection, antitumor, and immune regulation, with a fair correlation coefficient of 0.318 (Fig. 3b).

Figure 3
figure 3

Infiltration of diverse immune cell types in patients with HF. (a) Differential infiltration of CD8+T-cells in patients with HF and non-HF. (b) Pearson Correlation Analysis of 22 types of immune cells. *p < 0.05. **p < 0.01. ***p < 0.001.

Gene co-expression network of HF

We initially created a gene co-expression network using the R package “WGCNA” based on the expression levels of genes in the GSE14190 dataset13. Next, the average linkage and Pearson correlation coefficients were calculated, and cluster analysis was conducted for all samples within the dataset. Additionally, a scale-first network with β = 7 as the soft threshold power was successfully established (Supplemental Fig. 2a), and a hierarchical clustering tree was successfully constructed by employing a dynamic hybrid cut method, in which each leaf represented a specific gene and each branch represented a specific module that encompassed all genes exhibiting comparable expression levels (Fig. 4a). Finally, we consolidated the functionally equivalent modules into a single large module, resulting in a total of 9 modules (Fig. 4b).

Figure 4
figure 4

Establishment of a WGCNA model. (a) Gene classification into different modules by hierarchical clustering, with different colors representing different modules. (b) Heat map showing module features associated with T-cell infiltration and induced HF diseases. (c) Scatter plot of genes in brown modules. (d) Scatter plot of genes in blue modules. (e) A violin plot showing the expression levels of hub genes in the hearts of non-HF and HF patients. (f) ROC curve analyses of hub genes (GZMM, GNLY, NKG7, PTPRCAP, and ZAP70). ***p < 0.001.

Identification of hub modules and functional enrichment analysis

By choosing the proportion of 7 T-cell subtypes in each module as the trait data for WGCNA. The brown module (R2 = 0.75, p = 6.3E-139; Fig. 4c) and the blue module (R2 = 0.77, p = 1E-200; Fig. 4d) among the 9 modules exhibited the strongest correlation with CD8+T-cells. Consequently, our choice of hub modules encompassed both the blue and brown modules, from which the genes exhibiting greater linkage were plausibly genes linked to CD8+T-cells. We identified a total of 5 candidate hub genes (GZMM, GNLY, and NKG7 within the brown module, PTPRCAP and ZAP70 within the blue module) as candidate hub genes based on the truncation criteria (module affiliation > 0.8 and gene significance > 0.5). These key genes were significantly overexpressed in samples of patients with HF in comparison to those of patients with non-HF (Fig. 4e and Supplemental Fig. 2b). Subsequently, the analysis of genes in brown and blue modules was executed by utilizing the “cluster profile” R package. The results demonstrated that the genes in brown module were notably enriched in the Endocytosis, NF-kappa B, Neurotrophins, and Notch signaling pathways through KEGG analysis (Supplemental Fig. 3a). Meanwhile, Biological Process (BP) analysis revealed significant enrichment of module genes in the positive regulation of proteasomal ubiquitin. Cellular Component (CC) analysis revealed notable enrichment of module genes in both the cell-substrate junction and mitochondrial proteins. Molecular Function (MF) analysis revealed marked enrichment of module genes in both transcription corepressor activity and DNA-binding transcription factor binding (Supplemental Fig. 3b). Additionally, the genes in blue module were significantly enriched in both the PI3K-Akt and Ras signaling pathways in cancer cellular amino acid metabolic process through KEGG analysis (Supplemental Fig. 4a). Meanwhile, BP analysis revealed significant enrichment of module genes in the regulation of angiogenesis pathway and basement membrane organization. CC analysis revealed notable enrichment of module genes in both the mitochondrial matrix and myofibrils. MF analysis revealed marked enrichment of modular genes in both actin binding and growth factor binding (Supplemental Fig. 4b).

ROC curve analysis of key genes

ROC curves were generated using the expression data from GSE141910 to evaluate the potential diagnostic significance of 5 hub genes. The result showed that the area under the curve (AUC) for 5 hub genes in the datasets exceeded 0.7, except for PTPRCAP which had an AUC of 0.645, demonstrating that the remaining 4 hub genes have high sensitivity and specificity for HF17, suggesting they can be served as biomarkers for the diagnosis of HF (Fig. 4f).

The expression levels of HF hub genes in pan-cancer analysis

Pan-cancer analysis indicated that NKG7 was highly expressed, GNLY and GZMM were moderately expressed, and PTPRCAP was rarely expressed in tumors. Therefore, PTPRCAP was excluded for the subsequent analysis (Fig. 5a). To conduct an assessment of hub gene expression across various cancers with the Log2(TPM + 1) scale, we utilized RNA-seq data obtained from the GEPIA database. Our findings indicated that GZMM exhibited high expression levels in KIRC, PAAD, and SKCM. GNLY demonstrated heightened expression in CESC, ESCA, HNSC, KICH, KIRC, LAML, OV, PAAD, SKCM, STAD, and TGCT. NKG7 was highly expressed in KIRC, LAML, PAAD, and THYM. ZAP70 was highly expressed in KIRC, LAML, PAAD, and THYM (Fig. 5b).

Figure 5
figure 5

The expression levels of HF hub genes in different cancer types based on TCGA database. (a) Increased expression levels of HF hub genes in various cancers and undetected expression of PTPRCAP in various cancers. (b) The expression levels of HF hub genes in pan-cancer. Red represents significant upregulation in the tumor. Green represents significant downregulation in the tumor.

Correlation analysis of MSI and TMB for the hub genes in pan-cancer

The TMB is defined as the number of non-synonymous mutations in somatic cells within a specific genomic region and indirectly reflects the ability of tumors to produce new antigens and predict the efficacy of immunotherapy for tumors18. MSI is the phenomenon where replication errors resulting from DNA mismatch repair function are not corrected and continue to accumulate, resulting in changes in the length or base composition of microsatellite sequences19. A high MSI is negatively related to the effectiveness of tumor immunotherapy18. Radar images showed a significant correlation between GZMM and TMB in various cancers, such as UCEC, THYM, THCA, TGCT, STAD, PAAD, LGG, LAML, HNSC, DLBC, and COAD. GNLY had a strongly significant correlation with TMB in THYM, STAD, LUSC, LGG, KIGC, COAD, CHOL, and BRCA. NKG7 was associated with TMB in ACC, UCEC, THYM, THCA, TGCT, PRAD, PAAD, LUAD, LGG, KIRP, KIRC, COAD, CHOL, and BRCA. ZAP70 had a significantly correlation with TMB in THYM, THCA, STAD, PAAD, MESO, LUAD, LIHC, LGG, DLBC, CHOL, and ACC (Fig. 6a). Meanwhile, we investigated the potential correlation between the expression levels of hub genes and the MSI of diverse cancers. Our findings revealed that the expression of the GZMM gene was linked to MSI in multiple tumors including SKCM, OV, LGG, and COAD. Similarly, the expression of the GNLY gene was associated with MSI in SKCM, TGCT, COAD, and PRAD. Additionally, NKG7 gene was shown to be associated with MSI in TGCT, OV, LUSC, KIRP, and COAD. Furthermore, ZAP70 gene was linked to MSI in SKCM, PRAD, LUAD, ESCA, and OV (Fig. 6b).

Figure 6
figure 6

Correlation analysis of the expression of HF hub genes and both the TMB and MSI in pan-cancer. (a) Correlation analysis of the expression levels of HF hub genes and TMB in pan-cancer. (b) Correlation analysis of the expression levels of HF hub genes and MSI in pan-cancer. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001.

Gene enrichment analysis based on Gene Set Enrichment Analysis (GSEA)

After utilizing “c5.all.v2023.2.Hs.symbols.gmt” as the reference genome, our analysis showed that the list of gene sets exhibiting a strong correlation with GNLY was dominantly enriched in the antimicrobial humoral immune response and negative regulation of blood vessel endothelial cell migration of tumors. The list of gene sets highly correlated with ZAP70 was mainly enriched in the immunoglobulin production, IgG immunoglobin complex, and T-cell receptor pathways of tumors. The list of gene sets highly correlated with NKG7 was primarily enriched in the cytolytic granule and chemokine activity of tumors. The list of gene sets exhibiting a strong correlation with GZMM was dominantly enriched in the cytolytic granule and antibacterial humoral response of tumors (Supplemental Fig. 5). Taken together, the hub genes play a vital role not only in regulating immune responses in CD8+T-cells in HF, but also in controlling immune responses during tumor development.

Correlation analysis between the expression of HF hub genes and immune cell subtypes and between the expression of HF hub genes and the TME in pan-cancer

We conducted a correlation analysis to investigate the potential correlation between the expression of HF hub genes and various immune cell subtypes, including C1 (wound healing), C2 (IFN-γ dominant), C3 (inflammatory), C4 (lymphocyte depleted), C5 (immunologically quiet), and C6 (TGF-β dominant) 20,21 in pan-cancer. Our findings demonstrated that the HF hub genes exhibited high expression levels in C2 and C6 but low expression levels in C5 (Supplemental Fig. 6a). The TME plays a crucial role in influencing cellular heterogeneity in tumor cells, leading to the development of multidrug resistance and the progression of cancer21. Therefore, it was reasonable to investigate the correlation between the expression of HF hub genes and the cellular heterogeneity in TME. We utilized the ESTIMATE algorithm to quantify the immune and stromal scores in pan-cancer. Our results revealed a stronger positive correlation between the expression of HF hub genes and ImmuneScores (Supplemental Fig. 6b) compared to StromalScores (Supplemental Fig. 6c), suggesting that the expression of hub gene is related to the infiltration levels of immune cells in TME.

Identifying the correlation between the expression of HF hub genes and CD8+T-cells in pan-cancer

We identified the composition of diverse immune cell types in pan-cancer using the “CIBERSORT” algorithm and performed Pearson correlation analysis between the expression of hub genes and the infiltration of CD8+T-cells in various cancers. Our findings revealed a positive correlation between the expression of hub genes and the infiltration of CD8+T-cells in pan-cancer (Fig. 7).

Figure 7
figure 7

Correlation analysis of the expression levels of HF hub genes and the infiltrating of CD8+-T cell in pan-cancer. (a) GZMM. (b) GNLY. (c) NKG7. (d) ZAP70.

Prognostic value of the HF hub genes in SKCM

We investigated the prognostic risk of HF hub genes by Cox risk analysis in pan-cancer. All of the hub genes had detrimental prognostic effects on GBM, COAD, KIRP, TGCT, THYM, and UVM (HR > 1, p < 0.05). In contrast, the 4 hub genes were found to function as beneficial prognostic factors in SKCM, BLCA, SARC, CESC, THCA, BRCA, and UCEC (HR < 1, p < 0.05) (Supplemental Fig. 76). Meanwhile, we examine the prognostic significance of the HF hub genes for pan-cancer. Notably, the survival curves revealed a most positive correlation between the expression levels of these hub genes and Disease-Specific Survival (DSS) prognosis in SKCM patients rather than other cancer types, as determined from the UCSC Xena database, indicating that SKCM patients with elevated HF hub gene expression exhibited the highest likelihood of survival compared to other tumors (Fig. 8a). Additionally, the expression of hub genes was positively correlated with the infiltration of CD8+T-cells in SKCM (Fig. 8b). Overall, patients with HF and SKCM showed the greatest potential benefits in terms of hub gene therapy compared to other pan-cancer.

Figure 8
figure 8

SKCM was the most beneficial cancer for targeting hub genes. (a) DSS survival curves comparison of high and low expression of HF hub genes in SKCM. (b) HF hub genes were correlated with CD8+T-cells in SKCM.

The localization and transcription factors (TFs) prediction of hub genes

The regulation of key biological processes and functions by mRNAs is closely related to their localization within cells. The mRNA locater database revealed that GZMM, NKG7, and ZAP70 were mainly localized in the cytoplasm, while GNLY was mainly localized in the extracellular region (Supplemental Fig. 7b). We identified 24 TFs of GZMM, GNLY, NKG7, and ZAP70 from the NetworkAnalyst database. These TFs included Yin Yang 1 (YY1), FOXC1, RELA, TP53, based on which the hub gene-associated TF-gene networks were successfully constructed (Supplemental Fig. 7c).

Drug sensitivity analysis of the hub genes

To examine the possible correlation between the expression of hub genes and drug sensitivity in various cancer cell lines obtained from the CellMiner™ database, we conducted a correlation analysis by arranging the correlation coefficients in descending order. The results demonstrated that ZAP70, GZMM, and NKG7 were the main drug targets with significantly positive correlation with Nelarabine and Methylprednisolone which might have potential therapeutic effects on patients with HF and cancer (Fig. 9, Nelarabine, Cor = 0.980, p < 0.001; Methylprednisolone, Cor = 0.942, p < 0.001).

Figure 9
figure 9

Drug sensitivity analysis of HF hub genes.

Validation of the correlation between the expression of hub genes and HF by molecular biology techniques

To further validate the correlation between the expression of the hub genes and HF, subsequent molecular biology techniques were used to analyze heart samples from mice and humans. Our results indicated that GMZZ, ZAP70, and NKG7 exhibited high expression levels in the hearts of mice undergoing transverse aortic constriction (TAC) for a duration of 8 weeks (Fig. 10a–c) and in the hearts of patients with HCM-induced HF (Fig. 10d–f) compared to the control groups.

Figure 10
figure 10

Validation of hub genes in vivo. (a) mRNA levels of hub gene in the heart samples of human with HF and non-HF. (b, c) Western blotting analysis and quantification of the expression levels of hub genes in heart samples of individuals with HF and non-HF. (d) mRNA levels of hub gene in the heart samples of mice with HF and non-HF. (e, f) Western blotting analysis and quantification of the expression levels of hub genes in heart samples of mice with HF and non-HF. **p < 0.01.

Discussion

HF and cancer are leading causes of mortality, morbidity, and reduced quality of life. Despite advancements in medication, their co-occurrence is increasingly common. Numerous studies emphasize the crucial significance of adaptive immune response in the pathogenesis of both HF and cancer9,12,36. Additionally, immunotherapies based on cytotoxic CD8+T-cells have yielded promising outcomes in cancer treatment, such as chimeric antigen receptor (CAR)-CD8+T-cells22, T-cell receptors (TCR)-CD8+T-cells5, and immune checkpoint blockade22. However, inappropriate immunotherapies for anti-cancer based on CD8+T-cells led to the disruption of cardiac immune homeostasis and the development of cardiac immune-related adverse events. For example, cytotoxic CD8+T-cells targeting alpha-myosin promote the progression of myocarditis and HF23. Pembrolizumab targets cytotoxic CD8+T-cells for melanoma treatment develops HF24. While properly administered CD8+T-cell-based immunotherapies show potential in heart disease treatment, CAR-CD8+T-cells have been shown to target cardiac fibrosis6, and TCR-specific CD8+T-cells significantly repress the development of myocarditis25. Notably, HF and cancer share common risk factors and pathophysiological pathways4,26. Therefore, investigating therapeutic approaches for intervention to substantially improve the outcomes of both diseases is imperative.

In this study, we applied WGCNA to GSE141910 dataset to identify 2 modules that strongly correlated with CD8+T-cells. The genes within brown module were predominantly enriched in Notch signaling pathway that exerts cardioprotective effects on the progression of HF by repressing inflammation, inhibiting myocardial hypertrophy, and promoting tissue revascularization27. The genes within blue module were primarily enriched in PI3K-Akt signaling pathway and mitochondrial quality control. The PI3K-Akt signaling pathway plays a detrimental role in in HF by regulating cell morphology, angiogenesis, and inflammation28,29. Notably, aberrant activation of both the PI3K-Akt and Notch pathways promote the loss of control of tumor cell growth, survival, and metabolism30,31, suggesting that these pathways are common pathophysiological pathways of HF and cancer and primary candidates for therapeutic intervention. Additionally, mitochondrial quality control, which involves dynamic processes such as mitochondrial biogenesis, fission, fusion, and mitophagy, has recently been implicated in the development of HF32,33. Meanwhile, mitochondrial quality control imparts substantial flexibility to the growth, survival, and metabolism of tumor cells in challenging environments like nutrient scarcity, hypoxia, and cancer therapies, making it a crucial contributor to tumorigenesis as well34.

4 hub genes, NKG7, GNLY, GZMM, and ZAP70, were screened by using WGCNA on GSE141910. Their expression levels were positively correlated with CD8+T-cell infiltration in HCM-induced HF and pan-cancer. 4 hub genes have high sensitivity and specificity for HF and have the prognostic significance for cancer. Among the 4 hub genes, NKG7 gene is located on chromosome 19. NKG7 protein is a transmembrane protein with a cytoplasmic domain and an extracellular domain and exhibits significant expression levels and localization within the cytotoxic granules of CD8+T-cells35. NKG7 functions as an intrinsic factor for the cytotoxic function of CD8+T-cells by promoting efficient cytolysis via the perforin/granzyme pathway36. NKG7 have not been reported to be associated with HF, but function as a T-cell-intrinsic therapeutic target for improving antitumor cytotoxicity37. GNLY gene is encoded on chromosome 236. GNLY protein is present in cytolytic granules released by cytotoxic CD8+T-cells via exocytosis and participate in granule-mediated cytolysis of tumor cells and microbes38. GNLY has not been documented as linked to HF. However, elevated GNLY levels indicate a decreased risk of local recurrence in breast cancer39. GZMM gene is located on chromosome 5q11-q12, 14q11.2, and 19p13.240. GZMM protein is a chymotrypsin-like serine protease that belongs to the granzyme family in CD8+T-cells41. GZMM can induce cytolysis either in a caspase-dependent manner or by cleaving cytoskeletal components42,43. GZMM have not been linked to HF. However, lower levels of GZMM expression are indicative of a poor prognosis in individuals with follicular lymphoma44. ZAP70 gene is encoded on chromosome 2q11.2. ZAP70 protein is a member of the Syk family kinase and is predominantly expressed in thymocytes and peripheral T-cells45. ZAP70 comprises 2 N-terminal SH2 domains that are indispensable for the transfer of ZAP70 from cytoplasm to cytomembrane to directly interact with the TCR complex required for downstream signaling46. Humans and mice lacking ZAP70 develop severe combined immunodeficiency due to the inability to produce functional T-cells47. ZAP70 is correlated with immune cell infiltration in myocardial, and serve as a positive prognostic biomarker for laryngeal cancer48,49.

In the present study, we utilized the CellMiner database to ascertain prospective therapeutic agents capable of promoting the upregulation of the 4 hub genes. Among the 11 drugs, Nelarabine, emerged as a robust candidate, has shown remarkable efficacy in inducing remission among individuals afflicted with relapsed and refractory T-cell acute lymphoblastic leukemia (T-ALL) 50. Concomitantly, Nelarabine effectively induces remission of extensive cardiac infiltration in T-ALL patients51. Methylprednisolone has demonstrated significant clinical activity in patients with multiple myeloma52. In addition, methylprednisolone combined with rituximab demonstrates efficacy in the treatment of patients with chronic lymphocytic leukemia and small lymphocytic lymphoma53. Methylprednisolone alleviates myocardial infarction size and protects hypoxic heart muscle by preventing reduction in coronary flow in marginally ischemic tissue, protecting the fine ultrastructure of the heart, and delaying the increase in resting tension54,55,56. Few studies of cardiovascular disease were found for 11 predicted drugs in our study, except for the 2 agents mentioned above. Therefore, this research lays an important foundation for future investigations into effective therapeutic agents for both cancer and HF.

We constructed a TF–gene network containing 24 common TFs associated with NKG7, GNLY, GZMM, and ZAP70 to identify the linkage between hub genes. Numerous studies have generally reported that YY1 expression is increased in failing hearts and plays a protective role in diverse cardiovascular diseases by repressing the transcriptional activity of the alpha myosin heavy chain promoter57. Concurrently, YY1 overexpression is correlated with poorer prognosis in several tumor types, including prostate cancer, gastrointestinal cancer, and breast cancer58,59,60. YY1 transcription factor may serve as a therapeutic target for HCM-induced HF-cancer interactions.

We validated the expression of hub genes in heart samples of both human and mice with HF by using molecular biology techniques. However, GNLY was only expressed in human hearts and exhibited no expression in mice hearts61. Consequently, GNLY was excluded from 4 hub genes for further validation.

Conclusion

Our study revealed the significant roles of NKG7, GZMM, and ZAP70 in the immune response to HF through the modulation of the PI3K/Akt and Notch signaling pathways, as well as mitochondrial dynamics. These findings emphasized the potential diagnostic biomarkers and candidate therapeutic drugs, offering a promising avenue for addressing both tumors and HF simultaneously. Importantly, our results suggested that patients with SKCM and HF could particularly benefit from the implementation of these dual-target therapies, bringing us closer to more effective treatment options for this specific group of individuals.

This study has several limitations. First, we primarily focus on cytotoxic CD8+T-cells and hub genes in HF and cancer, neglecting other immune cell types and pathways that may be involved in these diseases. Future studies should adopt a more comprehensive approach to encompass a wider range of immune cell types and pathways. Second, although the hub genes have been implicated in HF and cancer, the exact mechanisms by which they contribute to these diseases remain unclear. Further investigations are necessary to elucidate the specific roles of these genes in disease progression. Thirdly, the sample size of our dataset is relatively small due to the challenge in obtaining human heart tissues. Finally, it is important to validate the functional significance of the transcriptional regulatory relationships and associated TFs in HF and cancer. Consequently, studies in the future need to be conducted to evaluate their potential as therapeutic targets.

Materials and Methods

Acquisition and processing of gene expression data

The GSE141910 dataset was downloaded from the GEO (https://www.ncbi.nlm.nih.gov/geo/), containing 195 heart samples from patients with HCM-induced HF and 161 heart samples from patients with non-HF62. First, the “voom” method in “limma” package was utilized to normalize the dataset (Supplemental Fig. 1) and conduct differential analysis63. The raw count data were converted into the Transcripts Per Million (TPM) format for differential analysis. The GPL16791 platform information was subsequently utilized to convert the gene IDs to gene symbols.

GO and KEGG pathway analysis

Log2Fold Change (logFC) > 1 and an adjusted p value < 0.05 were considered to indicate statistical significance for differential expression analysis. To analyze the upregulated pathways in HF, we selected the upregulated genes that met the filtering criteria for enrichment in GO and KEGG pathways using the “cluster profile” R package64,65.

Evaluation of immune cell infiltration in HF

To analyze the infiltration of immune cells in the GSE14190 dataset, we utilized the CIBERSORT algorithm, which is commonly employed to determine the proportions of 22 distinct types of immune cells by computing TPM data14,66.

Construction of the co-expression network

WGCNA is employed to depict patterns of gene association among diverse samples13. Its advantages lie in the identification of gene sets exhibiting synergistic changes and the identification of potential biomarkers based on their interconnectivity and association with phenotype. In the research utilizing the “WGCNA” R package, we constructed a gene co-expression network based on GSE14190 RNA-seq data, specifically targeting genes with differential expression. The process involved several key steps: the expression levels of each transcript were initially converted into a similarity matrix by using Pearson correlation coefficients. This similarity matrix was then transformed into an adjacency matrix. The transformation used the formula amn =|cMn|^b, where cMn signifies the Pearson correlation for gene pairs, and amn indicates the adjacency links between these gene pairs. To effectively convert the adjacency matrix into a topological overlap matrix, we identified the most suitable value for β. Finally, we set a module size minimum threshold at 20 and employed a bottom-up algorithm to categorize genes with akin expression patterns into distinct modules.

Building module feature relationships

Pearson Correlation Analysis was performed for the premium module based on the correlation between module features and T-cell subtypes and between module features and HF. To determine the modules significantly correlated with CD8+T-cells, we established the criteria of p < 0.05 and a correlation coefficient ≥ 0.4. Based on these findings, the genes with the highest 2 correlation coefficient with CD8+T-cells were selected as the hub modules.

Functional and pathway enrichment analysis

To clarify the biological roles and signaling pathways of the identified modules, we conducted GO and KEGG analysis using the “cluster profile” R package64,65.

Identification and validation of hub genes

Candidate hub genes were selected based on 2 criteria: module connectivity and their association with clinical traits. Module connectivity is defined by the absolute value of the Pearson correlation between genes (module membership). The clinical trait relationship is defined by the total value of the Pearson correlation between each gene and trait (gene significance). We set thresholds for selecting candidate hub genes: a module membership exceeding 0.8 and a gene significance above 0.5.

ROC Curve Analysis of hub Genes

ROC curve analysis was conducted to categorize the sensitivity and specificity of the hub genes in diagnosing HF. We explored the AUC using the statistical tool SPSS (https://www.ibm.com/analytics/spss-statistics-software).

Validation of the hub genes in pan-cancer

The gene expression RNA sequencing (HTSeq-FPKM), clinicopathological data, immune subtype, and survival data of 33 cancer types were downloaded from the database UCSC Xena (https://xena.ucsc.edu/, originating from The Cancer Genome Atlas Program (TCGA) database) 20,67. Box plots and heatmaps were further constructed using the R package “ggpubr”. Detailed information about the pan-cancer datasets presently evaluated was shown in Supplementary Table 168. For the TCGA pan-cancer analysis, the expression levels of the hub genes were extracted and integrated with Perl software. The method “Wilcox. Test” was applied to analyze the expression of the hub genes in different cancer types. GEPIA2 (http://gepia2.cancer-pku.cn/#general) is a reliable website based on TCGA database and Genotype-Tissue Expression (GTEx) normal sample database and is used to compare the differential expression levels of hub genes in pan-cancer69. The TPM data were used to determine the differential expression levels between the normal and tumor groups. A differential box plot was downloaded from the GEPIA2 database (http://gepia2.cancer-pku.cn).

Correlation analysis between the expression of hub genes and their immunotherapeutic efficacy in pan-cancer

Correlation analysis between key gene expression and both the TMB and MSI in pan-cancer was performed through radar imaging.

Immune correlation analysis of the hub genes in pan-cancer

The correlation analysis of immune subtypes for hub genes was primarily performed using R packages such as “limma” and “reshape2”. To calculate the scores for stromal and immune cells, we utilized the ESTIMATE algorithm from the R package “estimate” and “limma” to assess the infiltration of stromal/immune cells in pan-cancer tissues70. Meanwhile, the visualization of the hub genes and CD8+T-cells in 33 pan-cancer was performed via R packages including “impute”, “limma”, “corrplot”, “ggplot2”, and “ggpubr”.

Survival analysis based on the expression levels of the hub genes in human cancers

The DSS refers to the period from initial diagnosis to death caused by a specific disease, and at this point, attention is beginning to be paid to whether the cause of death was caused by a specific disease71. Survival analysis was performed by the Kaplan–Meier method and the log-rank test (p < 0.05). The median expression level of the hub genes was the cutoff value for dividing patients into 2 equal groups, namely, the high-expression and low-expression groups. A survival curve was delineated by the R packages “survminer” and “survival”. A Cox risk analysis was conducted to determine the association between the expression of the hub genes and the prognosis of pan-cancer. Multiple-forest plots were drawn using the R packages “survival” and “forestplot”. A p value < 0.05 was set as the threshold value.

The localization of hub genes and the construction of hub gene-associated TF-gene regulatory networks

The mRNA locater database (http://bio-bigdata.cn/mRNALocater) provides 3 different machine learning models with relatively stable performance based on the optimal feature subset using algorithms including XGBoost, LightGBM, and CatBoost, to calculate accurate eukaryotic mRNA subcellular localization72. We used this tool to predict the location of the hub genes within cells. TFs play important roles in mRNA transcriptional regulation by recognizing specific DNA sequences. The NetworkAnalyst database (https://www.networkanalyst.ca/) was utilized to predict TFs73. Cytoscape software (https://cytoscape.org/) was used to visualize these regulatory relationships.

Gene enrichment analysis based on GSEA

GSEA was utilized to determine the significant difference in a set of genes between 2 group states74. We performed a single-gene GSEA of hub genes, the pathways associated with the hub genes with high expression were identified using the “clusterProfilerR” package. The reference gene set c5.all.v2023.2.Hs.symbols.gmt was chosen from the Molecular Signature Database (MSigDB, https://www.gsea-msigdb.org/gsea/index.jsp), and a p value < 0.05 was used as the screening criterion.

Drug sensitivity and immune subtype correlation analysis of the hub genes

The CellMiner™ database is based on 60 types of cancer cells listed by the Cancer Research Institute of the National Cancer Institute66,75. Drug sensitivity data were downloaded from the CellMiner™ database (https://discover.nci.nih.gov/cellminer/home.do). For the data visualization, the R packages “ggpubr”, “limma”, “ggplot2”, and “impute” were used.

Human Heart Samples

The human myocardium specimens were collected from 8 individuals who undergone surgeries for cardiac valve replacement. These samples were divided into 2 categories based on echocardiographic findings: non-HCM (a septum thickness ≤ 11 mm) and HCM (a septum thickness > 11 mm). Before gathering heart tissues, informed consent was obtained from the potential organ donors upon receiving permission. Human experiments complied with the Ethical guidelines of The Code of Ethics of the World Medical Association (Declaration of Helsinki). The research protocol was reviewed and approved by the Medical Ethics Committee of Union Hospital Affiliated to Huazhong University of Science and Technology before the commencement of the study.

Mice and TAC surgery

The male C57BL/6 mice aged 10 weeks were obtained from Gempharmatech Co., Ltd (Nanjing, China). The TAC operation was performed to construct a model of pressure overload-induced HF. Mice were immobilized on a heating pad in the supine position. Anesthesia was induced via the intraperitoneal injection of pentobarbital sodium. The anesthetized mice were connected to a volume-cycled rodent ventilator. The proximal region of the sternum was accessed via median thoracotomy and blunt dissection. This approach allowed visualization of the transverse aorta situated between the innominate and left common carotid arteries. By tightly securing a 7–0 silk suture around the transverse aorta using a 28-gauge needle, a constant and permanent constriction of the aorta artery was induced. After ligation, the needle was gently withdrawn, and the chest cavity was closed layer by layer. Sham mice underwent the same operation except for the construction of the aorta. All animal experiments were performed in accordance with the United Kingdom Animals (Scientific Procedures) Act 1986 and the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals (2020). Prior to the study, the research protocol was reviewed and approved by the Medical Ethics Committee of the General Hospital of Northern Theater Command. We affirmed that this study strictly followed the ARRIVE guidelines (https://arriveguidelines.org).

Reagents

ZAP70 antibody (CST, USA, #84,835), GZMM antibody (Origene, USA, #TA367583), NKG7 antibody (CST, USA, #84,835), GAPDH antibody (CST, USA, #92310SF).

Western blotting

RIPA lysis buffer containing 1 × protease inhibitor cocktail was employed to lyse the heart tissues at 0 °C for 30 min. The protein concentration was assessed using the BCA Protein Assay Kit. The extracted protein, ranging from 20 to 40 µg, was electrophoresed on SDS-PAGE gels and subsequently transferred onto PVDF membranes. The membrane was subsequently blocked for 30 min with 5% skim milk. Next, the membrane was incubated overnight at 4 °C with specific primary antibodies. Afterward, peroxidase-conjugated secondary antibodies were added and incubated for 1 h at room temperature. Due to the low specificity of antibodies, the PDVF membranes were cropped before being exposed to an Amersham Imager 680 (GE, USA), and protein expression levels were quantified utilizing ImageJ software V1.8.0 (https://imagej.en.softonic.com/).

RNA extraction and real-time polymerase chain reaction (RT-PCR)

The total RNA from left ventricular tissues was extracted by using TRIzol reagent. Reverse transcription into cDNA was achieved by utilizing the PrimeScript RT Kit with gDNA Eraser. Afterwards, the mRNA levels of the targeted genes were determined via quantitative PCR using SYBR1 Premix Ex Taq II and normalized to the housekeeping gene gapdh. The sequences of the primers were displayed in Supplemental Table 2.

Statistical analysis

Bioinformatics statistical analysis was performed using R software v4.3.1 (https://www.r-project.org/). The data from molecular biology experiments are presented as the means ± standard errors of the means (SEMs) and were analyzed using GraphPad Prism 8 (GraphPad Software, CA, USA). Differences between 2 groups were assessed using the unpaired Student’s two-tailed t test. Normal distribution of the data was analyzed using a Shapiro Wilk test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Ethics approval

Human experiments were reviewed and approved by the Medical Ethics Committee of Union Hospital Affiliated to Huazhong University of Science and Technology. Before gathering heart tissues, informed consent was obtained from the potential organ donors upon receiving permission. Human experiments complied with the Ethical guidelines of The Code of Ethics of the World Medical Association (Declaration of Helsinki). All animal experiments were reviewed and approved by the Medical Ethics Committee of the General Hospital of Northern Theater Command. All animal experiments were performed in accordance with the United Kingdom Animals (Scientific Procedures) Act 1986 and the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals (2020). We affirmed that this study strictly followed the ARRIVE guidelines (https://arriveguidelines.org).