Identication of Both Shared and Specic Potential Molecular Mechanisms of ARVC and DCM Based on A Genome-Wide Microarray Dataset

Background (cid:0) The study aimed to detect the shared differentially expressed genes (DEGs) and specic DEGs of arrhythmogenic right ventricular cardiomyopathy (ARVC) and dilated cardiomyopathy (DCM) as well as their pathways. Methods: The GSE29819 dataset was examined for the DEGs of ARVC vs. non-failing transplant donor hearts (NF), DCM vs. NF, and ARVC vs. DCM based on 8 patients with ARVC, 7 patients with DCM, and 4 non-failing transplant donor hearts that were never actually transplanted. The shared DEGs and specic DEGs were screened out using a Venn diagram. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, Gene Ontology (GO) annotation, and protein-protein interaction (PPI) of the DEGs were determined using online analytical tools. Then, the modules and hub genes were identied using Cytoscape software. Results: A total of 684 shared DEGs of ARVC vs. NF and DCM vs. NF, 1371 specic DEGs of ARVC vs. NF, and 1075 specic DEGs of DCM vs. NF were identied. The shared DEGs were enriched in 63 biological processes (BP), 11 molecular functions (MF), 10 cellular components (CC), and 25 KEGG pathways. The DEGs of ARVC vs. DCM were enriched in 71 BPs, 19 MFs, 14 CCs, and 26 KEGG pathways. A PPI network with 187 nodes, 700 edges, and 2 modules, and another PPI network with 575 nodes, 2834 edges, and 7 modules were constructed based on the shared and specic DEGs, respectively. The top ten hub genes CCR3, CCR5, CXCL2, CXCL10, CXCR4, FPR1, APLNR, PENK, BDKRB2, GRM8, and RPS8, PRS3A, PRS12, RPS14, RPS21, RPL14, RPL18A, RPL21, RPL31 were identied for the shared and specic PPI networks,

Previously called arrhythmogenic right ventricular dysplasia, ARVC is an autosomal dominant inherited cardiomyopathy that manifests as the replacement of the right ventricular myocardium with fatty and brous tissue, and its prevalence is estimated to be 1:5000 [6]. The causative genes accounting for 50% of ARVC have been identi ed, including PKP2 (Plakophilin-2) [7], DSP (Desmoplakin) [8], DSG2 (Desmoglein-2) [9], and JUP (Junctional-plakophilin) [10]. It is a complex disease involving multiple genes; although the unique manifestations of electrocardiogram (ECG), cardiac ultrasound, and cardiac magnetic resonance imaging (MRI) have greatly improved the sensitivity and speci city of ARVC diagnosis [11], little is still known about many of the potential pathogenic genes and regulatory pathways of this disease. DCM is mainly characterized by ventricular enlargement and a decrease in the myocardial contraction force, with a prevalence of about 1:250 [12]. Some studies point out that both familial DCM and a fraction of non-familial DCM have a genetic basis [13,14]. To date, a portion of the key pathogenic genes of DCM have been identi ed, such as TTN [15], RBM20 [16], LMNA [17], and SCN5A [18]. Among them, a rare TTN variant accounts for 15-25% of DCM while RBM20 accounts for 1-3%. In addition, 6% of DCM cases is caused by rare variants of LMNA, and these mutations are associated with the highest risk of SCD. Diverse cellular proteins and pathologic mechanisms are involved in the development of DCM, such as the cytoskeleton, desmosomes, and ion channels [1]. However, a number of potentially associated genes and pathways may not have been identi ed yet.
In this study, we used a genome-wide microarray dataset in human ventricular tissues to identify DEGs between ARVC and non-failing transplant donor hearts that were not transplanted due to technical issues, DCM and these same non-failing transplant donor hearts, and ARVC and DCM. Based on these DEGs, GO and KEGG pathways were analyzed, and then we used Cytoscape software to construct PPI networks and identify the hub genes, aiming to detect the shared and unique molecular mechanisms in ARVC and DCM.

Identi cation Of Differentially Expressed Genes
The samples from the GSE29819 dataset were divided into three groups: an ARVC group (12 samples), a DCM group (14 samples), and a non-failing group (12 samples). Afterwards, the DEGs of the ARVC group vs. the NF group, the DCM group vs. the NF group, and the ARVC group vs. the DCM group were identi ed using an online analytical tool called GEO2R. GEO2R is used for conducting comparisons on raw data based on the GEO-query and limma R package. The results were extracted to a le, and the screening criteria were as follows: P < 0.05, and |log fold-change|>1.0. The heat maps and volcano plots of DEGs were created using an R package. Then, the shared DEGs and the speci c DEGs of the ARVC group vs. the NF group and the DCM group vs. the NF group were identi ed using the online analytical tool Draw Venn Diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/). GO and KEGG pathway analyses and integration of the PPI network GO and KEGG pathway analyses of the DEGs were conducted separately on the Database for Annotation, Visualization and Integrated Discovery (DAVID) (version 6.8). Statistical signi cance was set at P-value < 0.05. The results of the GO and KEGG pathway analyses were visualized by the R-ggplot2 package (version 3.5.3). Interactions among the shared DEGs and the DEGs of the ARVC group vs. the DCM group were evaluated using the STRING database (version 10.5) with a combined score of > 0.9, and the results were downloaded in TSV format for visualization using the Cytoscape plugin cytoHubba (version 0.1).
Then, the Cytoscape plugin Molecular Complex Detection (MCODE; version 1.31) was used to identify molecular modules of the PPI network, with a screening criterion of Degree cut-off > 10 and K-Core > 10, and cytoHubba was used to identify the top 10 hub genes according to maximal clique centrality (MCC) rank.

Identi cation of differential expression genes
A total of 2055 DEGs (826 up-regulated and 1229 down-regulated) of the ARVC group vs. NF group, 1759 DEGs (772 up-regulated and 987 down-regulated) of the DCM group vs. NF group, and 1658 DEGs (1264 up-regulated and 788 down-regulated) of ARVC group vs. DCM group were identi ed, respectively. After screening with the Venn diagram, a total of 1371 speci c DEGs of the ARVC group vs. NF group, 1075 speci c DEGs of the DCM group vs. NF group, and 684 shared DEGs of the ARVC and DCM groups were determined ( Fig. 1). Additionally, we screened out the top 100 DEGs in ascending P-value order to draw the heat map and volcano plots (Fig. 2).

Go And Kegg Pathway Enrichment Analysis
Among the shared DEGs, 63 biological processes (BP), 11 molecular functions (MF), and 10 cellular components (CC) were revealed by GO function clustering, and 25 KEGG pathway enrichments were identi ed. Of the DEGs of the ARVC group vs. DCM group, 71 BPs, 19 MFs, and 14 CCs, and 26 KEGG pathway enrichments were revealed. As for the speci c DEGs of the ARVC group vs. NF group, and the DCM group vs. NF group, 64 BPs, 19 MFs, 14 CCs, and 12 KEGG pathway enrichments, and 39 BPs, 16 MFs, 17 CCs, and 2 KEGG pathway enrichments were revealed, respectively. The entire results of GO are shown in Supplementary Table 1, Supplementary Table 2, and Supplementary Table 3. The results of KEGG pathways and the top 10 BPs, MFs, and CCs of GO based on different DEGs were selected for visualization (Fig. 3). The shared pathways of the ARVC vs. NF and the DCM vs. NF, as well as ARVC vs.

Ppi Network Construction And Hub Gene Identi cation
One PPI network with 187 nodes and 700 edges and another PPI network with 575 nodes and 2834 edges were constructed to detect the interactions among the shared DEGs and the interactions among DEGs of AVRC vs. DCM with a combined score > 0.9 ( Fig. 5A and Fig. 6A). After analysis with Degree cut-off > 10 and K-Core > 10, 2 modules ( Fig. 5B and Fig. 5C) and 5 modules (Fig. 6B-F) were identi ed, respectively. Finally, the top 10 hub genes of the two PPI networks were screened out according to the rank of MCC ( Fig. 5D and Fig. 6G).

Discussion
The shared pathways of ARVC and DCM compared to HF The shared DEGs were enriched on the 25 pathways, as shown in Figure 2. Cytokines play important role in regulation of immune function and are associated with the occurrence and development of a large number of human diseases, and many cardiovascular diseases have been shown to be associated with uncontrolled cytokines [19]. The activity of intracellular Janus kinases (JAKs) is associated with the assembly of the cytokine-receptor complex in the classic cytokine signaling pathway [20]. JAKs phosphorylate and activate the signal transducer and activator of transcription (STAT), which subsequently modulates gene expression [21,22]. IL-2, IL-4, and IL-6 are the most common types of cytokine; among them, IL-6 rst forms a dimer with IL-6R prior to binding its cognate receptor gp130, which is constitutively associated with JAK family tyrosine (Tyr) kinases and can be phosphorylated by JAKs. In addition, IL-6 also activates phosphatidylinositol 3-kinase (PI3K) pathways and extracellular signal-regulated kinase (ERK)1/2, following recruitment of SH2-containing protein Tyr phosphatase 2 to JAK-phosphorylated gp130 [23]. PI3Kα/Akt signaling leads to phosphorylation of Na v 1.5 on a site that regulates its gating properties, thus suppressing persistent I Na . At some point, the decrease of Na v 1.5 on the cell surface may result in a decrease in peak I Na after the inhibition of PI3Kα, which could slow action potential conduction and further induce arrhythmia if large enough [24]. These factors interact with pathways, some of which, if unregulated or unbalanced, may induce arrhythmias.
Many previous reports have proposed theories related to ARVC and DCM, including immunity, apoptosis, and gene mutation-related theories [25,26]. FoxO activity is involved in immune response, apoptosis, oxidative stress, aging, and other biological processes, mainly regulated by the PI3K/Art signaling pathway [27]. Similarly, the p53 signal pathway and TNF signal pathway are important components of apoptosis pathways. Consistent with previous reports, our study found the FoxO signal pathway, the p53 signal pathway, and TNF signaling pathways among the shared genes. The results suggest that immunity and apoptosis may be the common pathogenesis of ARVC and DCM, and intervention based on cell immunity and apoptosis may have signi cance for the prevention and treatment of ARVC and DCM.
Electrical remodeling is an important mechanism of arrhythmia and heart failure. After the integrity of myocardial bers is destroyed, the anisotropy of electrical activity increases, which promotes conduction disorders, inducing arrhythmias and heart failure. Shimizu H. et al [28] revealed that overload of Ca 2+ can destroy the integrity of myocardial bers by activating a calcineurin-FoxO-MuRF1-proteasome signaling pathway. Mota R. et al [29] reported a novel link between atrogin-1-mediated regulation of FoxO1/3 activity, reduced collagen deposition, and brosis in the aged heart. Bagchi AK. et al [30] showed that under stress, IL-10-mediated toll-like receptor 4 (TLR4) signaling suppresses apoptosis as well as brosis, while TLR2 has the opposite effect. Similarly, the hypoxia-inducible factor-1 (HIF-1) signal pathway also takes part in the development of brosis [31]. In the present study, the FoxO signaling pathway, HIF-1 signaling pathway, and toll-like receptor signaling pathway were all discovered to be shared DEGs. Thus, pathway-based interventions may help preserve the integrity of primary myocardial bers, improve electrical remodeling, and reduce the occurrence of heart failure in patients with ARVC or DCM.
The speci c pathways of ARVC compared to HF and DCM compared to HF Speci c pathways of ARVC compared to HF were identi ed in our study. Among them, the ECM-receptor interaction had already been proven to be involved in the brosis of myocardium [32]. The rest of the pathways tended toward neuro-regulation (neuroactive ligand-receptor interaction, serotonergic synapse, dopaminergic synapse, Fc gamma R-mediated phagocytosis) and infection (Leishmaniasis and Staphylococcus aureus infection). By contrast, the speci c pathways of DCM compared to HF were related to ATP-binding cassette (ABC) transporters and glycerolipid metabolism. ABC transporters like ABCA1, ABCA4, and ABCA5 are all expressed in human platelets, and they regulate platelet function [33]. The ATP-binding cassette transporter P-glycoprotein (ABCB1) may affect the bioavailability and elimination of digoxin, while ABCA8 and ABCA9 are indispensable components of the ATP-sensitive potassium (K ATP ) channel [34]. Glucose and lipid metabolism is important for myocardial cells, and some research shows that metabolic disorders are a cause of chronic heart failure, and that several parameters are even biological indicators of prognosis [35]. According to our results, development of ARVC and DCM occur via their unique pathways, and these pathways may provide some evidence to support targeted intervention for the two diseases, respectively.
The pathways of ARVC compared to DCM Compared to the pathways based on the shared DEGs, some of the pathways that are enriched in ARVC vs. DCM are unique, including ABC transporters, signaling pathways that regulate the pluripotency of stem cells, type II diabetes mellitus, fat digestion and absorption, bile secretion, complement and coagulation cascades, and in ammatory bowel disease (IBD). Type II diabetes mellitus, fat digestion and absorption, and bile secretion are correlated with glucose and lipid metabolism. As we mentioned above, metabolic syndrome is a risk factor for heart failure. Metabolic disorders such as those related to glucose, fat, or protein metabolism may contribute to heart failure. The complement system is a major element of immune response, and also plays an important role in the development of IBD [36]. Whether the complement system activated by in ammatory bowel disease has any effect on the development of ARVC or DCM needs further study.
Hub genes of the PPI network As the Figure 5 and Figure 6 show, we identi ed the top 10 hub genes of the ARVC vs. NF and DCM vs. NF, and ARVC vs. DCM, respectively. These genes are involved with metabolism, in ammation, immune, cell apoptosis, or other critical biological processes. Proenkephalin (PENK), which is related to renal function, is a stable endogenous opioid biomarker and has been reported to be a prognostic indicator of heart failure [37]. What's more, it is also a modulator of IL-10 [38], a cytokine involved in in ammation and immune response, which also have important roles in cardiovascular disease. BDKRB2 is related to hypertension as a target of angiotensin II type 1 receptor signaling, and its polymorphism is related to the glucose metabolism [39,40]. Endothelial APLNR is critical for apelin signaling and its glucose-lowering effects [41]. CCR3 and CCR5 are chemokine receptors, and the former plays a pivotal role in leukocyte chemotaxis [42]. CXC chemokines regulate the recruitment of neutrophils via CXCR1 and CXCR2 in humans [43]. In addition, CXCL10 has demonstrated a novel function in mediating monocyte production of pro-in ammatory cytokines [44]. Another gene, CXCR4, was postulated to mediate atherosclerosis and in ammation in a recent study [45]. These genes played a role in immunity and/or in ammation, which has been demonstrated to be related to the development of cardiovascular disease.
In the context of the hub genes of the PPI network based on ARVC vs. DCM, RPS3A was a key factor in modulating the brown fat-speci c gene UCP-1 and carbon metabolic enzymes in EAT for preventing CAD [46]. CDK11p46 and RPS8 are associated with each other, and both are involved in cell apoptosis; similarly, over-expression of RPS14 can inhibit Rb phosphorylation and result in cell cycle arrest and senescence [47,48]. It is well known that diabetes and renal dysfunction can both affect cardiac function, but interestingly, RPS12 has been identi ed as a pathogenic gene of diabetic kidney disease by a genome-wide association study (GWAS) [49]. Finally, the mutations of RPL18A and RPL31 have been proven to be associated with Diamond-Blackfan anemia [50], which may decrease the blood supply to myocardial cells to some degree.
Although the function and the pathway of the above hub genes have been reported previously, they have not been veri ed before, as a large enough sample of myocardium tissues with ARVC and DCM is di cult to collect. Meanwhile, it can also be inferred that since the mechanisms of ARVC and DCM are regulated by multiple genes and multiple pathways, they require more comprehensive and targeted intervention.

Limitations
There are some limitations to our study. Firstly, the samples from non-failing donors may differ from the normal population, which may limit the applicability of our results between patients with ARVC or DCM and normal population. Secondly, since genes interact with each other, we arti cially screened the intersection of different groups of DEGs for further analysis and may have inadvertently excluded some genes with potential links, which may make the analysis of diseases one-sided to some degree. Thirdly, our study was only based on the GSE29819 dataset, and the results need to be validated with a further, more rigorous investigation and a large sample size .

Conclusions
In this study, genome-wide differentially expressed genes were used to identify the functions and mechanism of shared and speci c genes of patients with ARVC and/or DCM compared to non-failing donor heart patients. Our ndings may help to provide a better understanding of the functions and roles of these DEGs in ARVC and DCM and provide a reference for future treatment strategies. However, further studies are required to validate the role of these DEGs and pathways involved in these two diseases.      The protein-protein interaction (PPI) network of ARVC vs. NF and DCM vs. NF. A. PPI network. The sequence of the edges' colors is red-orange-blue from high combined score to low combined score. B-C.
The modules of the PPI network. Yellow, module 1; green, module 2. D. The top hub genes of the PPI network. The sequence of colors is red-orange-yellow from high ranking to low ranking.