Despite contemporary advances in medicine, cardiovascular diseases (CVDs) are still the leading cause of mortality worldwide, accounting for almost half the total number of global deaths (1). CVDs encompass a wide array of heart and vessel-related pathologies, such as coronary heart disease, hypertension, cardiomyopathy (CM), and congenital heart disease (2, 3), all eventually progressing to heart failure (HF).

Heart failure is a debilitating condition that manifests as a consequence of abnormalities in cardiac function, structure, rhythm, or conduction (4). The etiological factors of HF syndrome are often difficult to discern and vary (5, 6), with dilated cardiomyopathy (DCM) and ischemic cardiomyopathy (ICM) being among the main causes of HF in Western countries. ICM is defined by an imbalance between myocardial oxygen demand and supply resulting in myocyte loss and ventricular failure (7), while DCM is characterized by left ventricular dilation and subsequent contractile dysfunction (8).

Although HF pharmacotherapy has come a long way since diuretics and digitalis were state-of-the-art (9), a long-standing paradigm is that HF with reduced ejection fraction evolves via a “final common pathway” (10). Current therapeutic approaches, such as angiotensin-converting enzyme inhibitors (ACEIs) and β-receptor blockers, are relatively etiology agnostic and focus on symptom alleviation (11). This potentially reflects a lack of comprehension of the heterogeneous pathogenic mechanisms in the progression of DCM and ICM.

Dysregulated genes, proteins, and their corresponding biological pathways represent the molecular background of multiple diseases (12). The continuous development of omics technologies and data processing through bioinformatics shed new light on CVD molecular basis (13, 14). Although several standalone transcriptomics and proteomics studies have revealed differentially expressed molecules (DEMs), i.e., genes and proteins, in DCM and ICM, integrated multi-omics analyses of multiple datasets are still sparse.

The present study aims to elucidate the gene expression and proteomic signature of DCM and ICM and explore their molecular characteristics via bioinformatics analyses. This approach combines the assessment of mRNA and protein molecules, generating common DEMs, and provides strong evidence for their role in HF. The goal is to identify potential biomarkers and discover novel therapeutic targets by unraveling the complex architecture of HF pathogenesis. Our results accentuate the importance of Extracellular Matrix (ECM) organization in HF and stress the need for matrix-based therapies that may attenuate remodeling processes and/or promote cardiac regeneration.

As shown in the present multi-omics study, a unique molecular signature was deduced for DCM and ICM after intersection of microarray, RNASeq and mass spectrometry analyses with 10 DiSig and 15 IsSig DEMs being the most important finding in our results. In intersection of at least two common methods, DiSig is comprised of 106 DEMs (78 upregulated and 28 downregulated), while IsSig encompasses 187 DEMs (147 upregulated and 40 downregulated). Analyses of pathway enrichment were performed and PPI networks were constructed to detangle the processes and mechanisms of DCM and ICM pathology, using DiSig and IsSig molecules.

Extracellular matrix organization was the most abundantly expressed pathway in both DCM (p = 10–12) and ICM (p = 10–24), with upregulated genes encoding collagens and metalloproteases. Extracellular matrix (ECM) is a well-documented molecular pathway (38) and a major player in cardiac homeostasis. Not only does it provide structural support, but also facilitates force transmission and transduces molecular signals regulating cardiac cell function. In failing hearts, the cardiac interstitium is expanded, by augmentation of both structural and matricellular ECM proteins, resulting in alterations in extracellular matrix biochemistry. ECM plays a critical role in regulating fibrotic, inflammatory, and even regenerative responses, making it an attractive therapeutic target in HF (39).

TGFβ signaling pathway is also under investigation for the development of novel therapies for HF. TGFβ levels are elevated in HF, promoting cardiomyocyte apoptosis and cardiac hypertrophy and playing an important role in heart remodeling (40).

Transmembrane transport of ions is another important pathway found to be upregulated in both DCM and ICM. Ion channels, transporters and pumps comprise only a subset of proteins that are altered during HF with calcium playing a critical role in mediating the cardiac excitation-contraction coupling (41).

By comparing the DEMs of DiSig and IsSig, 8 upregulated molecules were found common. These genes can be categorized as cardioprotective proteins (HSPA2, SOD3), genes having a major role in remodeling processes (AEBP1, CA3, THBS4, UCHL1) and hemoglobins (HBA2, HBB). SOD3, extracellular superoxide dismutase [Cu-Zn], is an antioxidant protein (42), while HSPA2, Heat shock-related 70 kDa protein 2, is a molecular chaperone that helps maintain cardiomyocyte protein quality. It can be induced by cellular stress, promoting cell survival, as the toxic to cardiomyocytes’ misfolded proteins, directly contribute to HF (43). Exploration of strategies involving SOD3 and HSPA2 may provide therapeutic options against HF and associated systemic inflammation. The carbonic anhydrase enzyme (CA3) gene expression is induced in HF due to ventricular stretch inflicted as a consequence of increased ventricular load. Alvarez et al., have presented evidence that elevated CA3 levels can be used as biomarker for early detection of cardiac hypertrophy and HF (44). The hemoglobin types HBA2 and HBB are overexpressed in DCM and ICM patients, suggesting a potential reciprocal mechanism due to dysregulation in oxygen circulation and general hypoxemia in HF patients.

Thrombospondin-4 (TSP-4), a secreted extracellular matrix protein, is involved in myocardial remodeling by regulating the adaptive cardiac responses to pressure overload (45). Another hypertrophic factor, Ubiquitin C-terminal hydrolase L1 (UCHL1) is related to fibrosis and has proven to deubiquitinate and stabilize the epidermal growth factor receptor (EGFR), promoting cardiac hypertrophy (46). Lastly, Adipocyte enhancer-binding protein 1 (AEBP1), a positive regulator of collagen involved in the organization and remodeling of the ECM, was found upregulated in a DCM patients (19).

While IsSig and DiSig share a lot of common elements, IsSig has five additional ECM proteins associated with it (BGN, COL14A1, LUM, LTBP2, MFAP4) and two bloodstream proteins APOA1 and CFH. Collagen type XIV (COL14A1) is major fibrillar collagen produced by fibroblasts and is involved in ECM during the progression of cardiac remodeling in the failing heart (39).

Lumican (LUM) is an ECM localized proteoglycan associated with inflammatory conditions and known to bind collagens (47). Previous studies in humans and mice indicated that the LUM protein levels are increased in cardiac tissues of patients with HF compared to control hearts (48). These findings suggest that LUM may contribute to cardiac remodeling, by assisting in fibrinogenesis.

Microfibril-associated glycoprotein 4 (MFAP4) is an ECM protein that is involved in cell adhesion or intercellular interactions. It was demonstrated that MFAP4 deficiency inhibits cardiac fibrosis and ventricular arrhythmias in mice models and therefore may act as a novel therapeutic target for the prevention of cardiac remodeling in HF (49). Latent-transforming growth factor beta-binding protein 2 (LTBP2) is an ECM protein. Bai et al., demonstrated that serum LTBP-2 levels might act as a promising biomarker in HF, as LTBP-2 levels in HF patients are significantly elevated (50). Lastly, Biglycan (BGN) is a protein responsible for muscle development, regeneration, and collagen fibril assembly. In previous studies, it was proven that biglycan is required for the stability of collagen matrix formation, during ECM remodeling (51).

The two downregulated genes specifically in DCM are Myosin Heavy Chain 6 (MYH6) and Serpin Family A Member 3 (SERPINA3) with its corresponding protein GIG25 (alpha-1-antichymotrypsin). The MYH6 gene encodes instructions for the cardiac alpha (α)-myosin heavy chain, found in cardiac muscle cells, where it forms type II myosin. Type II myosin generates the mechanical force needed for cardiac muscle contraction in sarcomeres (52). Mutations in MYH6 may cause a spectrum of cardiac phenotypes associated with contractile dysregulation (53). GIG25, a protease inhibitor, is an acute phase response gene primarily upregulated during inflammation. Tanash et al. concluded that individuals with lower levels of GIG25 protein have a lower risk of developing heart incidents (54). Dysregulation of these two genes/proteins of DiSig can be used to differentiate between the two subphenotypes of HF.

In the work of Kanapeckaitė and Burokienė (55), bulk and single-cell RNA-sequencing and proteomics datasets of the human heart tissue were analyzed. Similar results of tissue remodeling and inflammatory processes were identified as pharmacological targets for DCM and ICM, respectively, despite using a different methodology. Our approaches differ significantly as in our study we accumulated a large number of human samples (in total 252 DCM, 232 ICM, and 221 control heart samples) and achieved gene/protein cross-validation through transcriptomic and proteomic analysis, while their work was based on a mixture of human and murine samples. They also utilized a two-step machine learning pipeline, while we have followed a multi-omic network approach.

Several disease-susceptibility loci of heart failure and cardiomyopathies have been identified in genome wide association studies (56). However, these loci were either not identified as DEGs in our study or they were only differentially expressed in proteomic analysis. It should be acknowledged that genetic variations affecting amino acid coding sequences do not affect the number of final transcripts, mutations leading to protein isoforms cannot be directly linked with transcriptomic or proteomic alterations and, additionally, the number of transcripts does not necessarily correspond to protein abundance since several expression and translation regulators exist in between (57). Further studies are therefore needed to assess the functional significance of genetic alterations, including their transcriptomic and proteomic provoked changes, which can create a predisposition to DCM and ICM.

The results of the multi-omics approach we have integrated propose a total of 17 targets that are potentially of enhanced biological significance as their dysregulation is confirmed on both transcriptomic and proteomic level. These targets can be further investigated as potential therapeutic targets, as a variety of them regulate extracellular matrix (LUM, BGN, MFAP4, COL14A1, LTBP2) and fibrosis (CA3, UCHL1), while others are associated with increased oxidative stress and inflammation in the heart (HSPA2, SOD3). Additionally, the DEMs that were found unique in each disease could serve as biomarkers, by measuring their expression levels in patient samples, leading to patient stratification between the two subtypes of heart failure. To overcome the need of tissue samples for transcriptomic and proteomic analyses, further studies can also identify the best fitted biomarkers in easily accessible biological material, such as blood or plasma.

As with most bioinformatics studies on human diseases, this study has its limitations. Although we detected gene and protein expression in cardiac tissues, as well as several related pathways and mechanisms, these findings need to be confirmed in further studies. Moreover, there are unavoidable differences in samples used such as etiologies and duration of cardiomyopathy, differences in age, gender, and medications, as well as the individual course of the disease, which contribute to the variability of gene and protein expression data. Missing metadata is a common limitation of studies based on public data and heterogeneity on the abovementioned variables is expected. In our study, however, the left ventricle samples used were collected during cardiovascular surgery, suggesting that the disease has already progressed. Subsequently, it can be speculated that the majority of patients were under medication and thus similar confounding factors are expected throughout the whole study population. In addition, true “non-failing” human ventricular tissue is not easy to obtain, as non-transplantable donor hearts are usually exposed to varying degrees of hypoxia which is known to be a potent inducer of BNP gene expression and chemokine (58). Finally, in bioinformatics studies, results can only successfully impute correlation and not causation between differentially expressed genes/proteins and disorders. To assess the causality and functional significance of dysregulated genes in DCM and ICM as for whether these targets contribute to disease pathogenesis or are changes resulting from the disease, different models both in vivo and in vitro are required; the results of our study can be used for the selection of the molecules further examined in such studies.

An obvious strength of this study is the integration of multiple independent microarray, RNASeq, and proteomic studies accumulating a large number of failing and non-failing hearts allowing for minimizing biases after normalization. To the best of our knowledge, this is the seminal study to cross-validate gene and protein expression as well as differentiate between the two subphenotypes of HF. Additionally, the rather large sample size of our study combined with the strict cutoffs used (padj < 0.05, | FC | > 2) during statistical analysis suggest that the derived results are minimum affected by random variation.

We aimed to identify the genetic and proteomic signatures of DCM and ICM, using a comprehensive multi-omics analysis. We herein demonstrate that DiSig and IsSig share common gene and protein expression elements, but also exhibit disease-specific molecular pathways. Extracellular matrix dysregulation was highlighted in both DCM and ICM, suggesting an attractive pharmacological target. In total 10 genes/proteins were highlighted in DiSig and 15 genes/proteins in IsSig. Therefore, our findings could provide insights into the pathogenesis of HF and suggest that the uncovered genes can be further investigated as possible novel diagnostic and/or therapeutic agents.

The original contributions presented in this study are included in this article/Supplementary material, further inquiries can be directed to the corresponding authors.

KP: formal analysis, methodology, and writing—original draft. ND: formal analysis, methodology, and writing—review and editing. GR, NA, and GK: writing—review and editing. VM: conceptualization, funding acquisition, writing—review and editing, and final approval. All authors made a significant intellectual contribution and read and approved the manuscript.