Exploring the Role of Obesity in Dilated Cardiomyopathy Based on Bio-informatics Analysis

(1) Background: Obesity is a major risk factor for cardiovascular disease (CVD), contributing to increasing global disease burdens. Apart from heart failure, coronary artery disease, and arrhythmia, recent research has found that obesity also elevates the risk of dilated cardiomyopathy (DCM). The main purpose of this study was to investigate the underlying biological role of obesity in increasing the risk of DCM. (2) Methods: The datasets GSE120895, GSE19303, and GSE2508 were downloaded from the Gene Expression Omnibus (GEO) database. Differentially expressed genes (DEGs) were analyzed using GSE120895 for DCM and GSE2508 for obesity, and the findings were compiled to discover the common genes. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted for the common genes in RStudio. In addition, CIBERSORT was used to obtain the immune cellular composition from DEGs. The key genes were identified in the set of common genes by the least absolute shrinkage and selection operator (LASSO) algorithm, the prognostic risk models of which were verified by receiver operator characteristic (ROC) curves in GSE19303. Finally, Spearman’s correlation was used to explore the connections between key genes and immune cells. (3) Results: GO and KEGG pathway enrichment analyses showed that the main enriched terms of the common genes were transforming growth factor-beta (TGF-β), fibrillar collagen, NADPH oxidase activity, and multiple hormone-related signaling pathways. Both obesity and DCM had a disordered immune environment, especially obesity. The key genes NOX4, CCDC80, COL1A2, HTRA1, and KLHL29 may be primarily responsible for the changes. Spearman’s correlation analysis performed for key genes and immune cells indicated that KLHL29 closely correlated to T cells and M2 macrophages, and HTRA1 very tightly correlated to plasma cells. (4) Conclusions: Bio-informatics analyses performed for DCM and obesity in our study suggested that obesity disturbed the immune micro-environment, promoted oxidative stress, and increased myocardial fibrosis, resulting in ventricular remodeling and an increased risk of DCM. The key genes KLHL29 and HTRA1 may play critical roles in obesity-related DCM.


Introduction
Dilated cardiomyopathy (DCM) is a disease of the myocardium with left ventricular (LV) dilation and systolic dysfunction resulting from genetic and environmental factors, excepting coronary artery disease and abnormal loading [1]. DCM, a major cause of heart failure, is the most common indication for heart transplantation worldwide, the prevalence of which is about 40 in 100,000 [2]. Genetic factors play important roles in the development

Micro-Array Data
Gene expression profiles of DCM (GSE120895 [20], GSE19303 [21]) on platform GPL570 and gene expression profiles of obesity (GSE2508 [22]) on platform GPL92 were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/, accessed on 13 May 2022) and logarithmically transformed in RStudio. Samples of DCM and control samples were all taken from human endocardium myocardium. GSE120895 was used as the training dataset, and included 47 DCM patients and 8 controls. GSE19303, which was used to validate the key genes, contained 40 human endocardium myocardium samples from DCM at baseline, and 8 control samples, of which 33 DCM samples were obtained again six months later, after immune-adsorption with subsequent immunoglobulin substitution (IgA/IgG). The 33 patients with symptoms of HF caused by DCM did not develop cancer, infectious diseases, coronary heart disease, acute myocarditis, or other conditions that lead to HF [21]. GSE2508 contained 39 samples of adipocytes from 19 obese subjects and 20 lean individuals.

Function Enrichment Analysis of Common Genes
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted for the common genes in RStudio with package "clusterProfiler" [26], setting p-value < 0.05. Gene Ontology (GO) described the biological functions of common genes at biological process (BP), cellular component (CC), and molecular function (MF) levels.

CIBERSORT
CIBERSORT is a method for obtaining immune cellular composition from gene expression profiles using a deconvolution algorithm and leukocyte gene signature matrix LM22 [27]. For the analysis of the immune micro-environment for DCM and obesity, CIBERSORT was performed on DEGs with 1000 permutations in RStudio. The results were visualized by the "ggplot2" package.

Identification and Verification of Key Genes
The least absolute shrinkage and selection operator (LASSO) algorithm was used to identify the key genes among the common genes with R package "glmnet" [28]. Subsequently, lambda.min and lambda.1se (standard error, SE) were selected to construct the prognostic risk models in GSE120895 and GSE2508, respectively, which were then verified by receiver operator characteristic (ROC) curves in GSE19303. Gene expression profiles of 15 common genes from GSE19303 were used to verify the values of the models to distinguish between DCM and control, and to assess the ability to discriminate obese and lean in the DCM group. Genes from the intersection of the models (with lambda.min) of GSE120895 and GSE2508 were identified as key genes. In order to further evaluate the diagnosis value of key genes, we drew ROC curves and calculated the area under the curve (AUC) for each gene in SPSS.

Spearman's Correlation Analysis between Key Genes and Infiltrating Immune Cells
Spearman's correlation analysis was performed on the obese DCM set using the R packages "psych" and "ggcorrplot", to explore the connections between the key genes and immune cells, with 0.05 as a p-value cutoff.
In addition, we separately intersected common genes with immune genes and ferroptosis genes, and found that NOX4 was the only common gene in both intersections ( Figure 1D,E).  In addition, we separately intersected common genes with immune genes and ferroptosis genes, and found that NOX4 was the only common gene in both intersections ( Figure 1D,E).

Analysis of Common Genes at Functional Level
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed for 15 common genes to decipher biological functions in DCM. Gene Ontology (GO) enrichment analysis showed that the main enriched terms in the biological process (BP) category included the response to transforming growth factorbeta (TGF-β), several kinds of extracellular organization, and the FasL biosynthetic process ( Figure 2A). The collagen-containing extracellular matrix, fibrillar collagen and NADPH oxidase complex were enriched in the cellular component (CC) category ( Figure 2B). Extracellular matrix structural constituents, growth factor binding, heparin binding, and the activity of several enzymes (endopeptidase, NADPH oxidase, phosphodiesterase) were the major enriched terms in the molecular function (MF) category ( Figure 2C). According to Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, the terms mainly enriched were the AGE-RAGE signaling pathway in diabetic complications, cGMP-PKG, cAMP and PI3K-Akt signaling pathways, human papillomavirus infection, regulation of lipolysis in adipocytes, and multiple hormone (relaxin, aldosterone, vasopressin, adrenergic, thyroid hormone, insulin, glucagon)-related signaling pathways ( Figure 2D). This showed that the endocrine system plays an important role in DCM.

Analysis of Common Genes at Functional Level
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed for 15 common genes to decipher biological functions in DCM. Gene Ontology (GO) enrichment analysis showed that the main enriched terms in the biological process (BP) category included the response to transforming growth factor-beta (TGF-β), several kinds of extracellular organization, and the FasL biosynthetic process (Figure 2A). The collagen-containing extracellular matrix, fibrillar collagen and NADPH oxidase complex were enriched in the cellular component (CC) category ( Figure 2B). Extracellular matrix structural constituents, growth factor binding, heparin binding, and the activity of several enzymes (endopeptidase, NADPH oxidase, phosphodiesterase) were the major enriched terms in the molecular function (MF) category ( Figure 2C). According to Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, the terms mainly enriched were the AGE-RAGE signaling pathway in diabetic complications, cGMP-PKG, cAMP and PI3K-Akt signaling pathways, human papillomavirus infection, regulation of lipolysis in adipocytes, and multiple hormone (relaxin, aldosterone, vasopressin, adrenergic, thyroid hormone, insulin, glucagon)-related signaling pathways ( Figure 2D). This showed that the endocrine system plays an important role in DCM.

Immune Cell Infiltration
Both DCM and obesity showed differences from control in immune cell infiltration ( Figure 3). The immune environment of obesity was more disordered, where most kinds of immune cells had a dysregulation. Compared to control, both DCM and obesity groups had a higher proportion of resting memory CD4 and naive T cells, as well as resting NK cells. The number of naive and memory B cells in DCM and obesity was less than that in control.

Immune Cell Infiltration
Both DCM and obesity showed differences from control in immune cell infiltration ( Figure 3). The immune environment of obesity was more disordered, where most kinds of immune cells had a dysregulation. Compared to control, both DCM and obesity groups had a higher proportion of resting memory CD4 and naive T cells, as well as resting NK cells. The number of naive and memory B cells in DCM and obesity was less than that in control.

Identification and Verification of Key Genes
Key genes were further identified from common genes by LASSO in the training datasets ( Figure 4). Lambda.min and lambda.1se (standard error, SE) were 0.014 and 0.058 ( Figure 4A), which were used to construct LASSO regression models in GSE120895. Lambda.min and lambda.1se of LASSO regression models in the GSE2508 datasets were 0.018 and 0.062 ( Figure 4C). NOX4, MBD6, PDE3B, ADAMTS15, TPPP3, CCDC80, PHLDA1, COL1A2, HTRA, and KLHL29 in the LASSO model with lambda.min were selected in GSE120895, while the model with lambda.min in GSE2508 missed MBD6,
ROC curves further verified the value of the two models to make a distinction between DCM and control in the validation datasets, the AUCs of which were 0.99 and 0.98 ( Figure 4E), respectively. However, the differences between lean and obese DCM were not obvious, with both AUCs being <0.7 ( Figure 4F).  ROC curves further verified the value of the two models to make a distinction between DCM and control in the validation datasets, the AUCs of which were 0.99 and 0.98 ( Figure 4E), respectively. However, the differences between lean and obese DCM were not obvious, with both AUCs being <0.7 ( Figure 4F).
The five key genes were all obviously up-regulated in DCM ( Figure 5A,B), the AUCs of which were calculated in GSE19303 (Table 1). We chose 40 DCM samples at different BMI baseline levels and four lean control samples to perform ROC analysis and calculate the AUCs ( Table 2), indicating that the key genes had significant diagnostic value for DCM. Compared to lean or overweight DCM, the AUCs of CCDC80, NOX4, and COL1A2 dropped a lot in obese DCM and the control group, while HTRA1 (AUC = 0.9, p = 0.024) and KLHL29 (AUC = 0.8, p = 0.066) showed stable values within normal limits. A total of 33 patients were followed up after IgA/IgG, including 20 responders and 13 non-responders. Only KLHL29 had a statistically significant decrease in responders ( Figure 5C,D).

Discussion
Given the increasing global disease burden of CVD, there is no doubt that obesity as one of its main risk factors contributes to the burden [29]. Heart failure accounts for a majority of the CVD burden, for which DCM is one of the main causes [2]. Although the relationship between DCM and obesity has been explored in literature, and the scientific community acknowledges obesity as a risk factor for DCM, the pathological link is not clear [13,14]. Bio-informatics was used to reveal potential connections between DCM and obesity in this study. Initially, we derived 15 common genes between DCM and obesity via an analysis of DEGs. Enrichment analysis showed that the common genes mainly played a role in TGF-β, fibrillar collagen, NADPH oxidase activity, and multiple hormonerelated signaling pathways, which suggested that metabolic disorders, oxidative stress, and myocardial fibrosis might play a role in the obesity-induced development of DCM. In addition, both DCM and obesity had a similar immune micro-environment, possibly involving NOX4, related to immune response and ferroptosis. NOX4, CCDC80, COL1A2, HTRA1, and KLHL29 were identified as key genes by LASSO regression analysis, among which HTRA1 and KLHL29 showed a close association with immune system infiltration in DCM and obesity.

Discussion
Given the increasing global disease burden of CVD, there is no doubt that obesity as one of its main risk factors contributes to the burden [29]. Heart failure accounts for a majority of the CVD burden, for which DCM is one of the main causes [2]. Although the relationship between DCM and obesity has been explored in literature, and the scientific community acknowledges obesity as a risk factor for DCM, the pathological link is not clear [13,14]. Bio-informatics was used to reveal potential connections between DCM and obesity in this study. Initially, we derived 15 common genes between DCM and obesity via an analysis of DEGs. Enrichment analysis showed that the common genes mainly played a role in TGF-β, fibrillar collagen, NADPH oxidase activity, and multiple hormonerelated signaling pathways, which suggested that metabolic disorders, oxidative stress, and myocardial fibrosis might play a role in the obesity-induced development of DCM.
In addition, both DCM and obesity had a similar immune micro-environment, possibly involving NOX4, related to immune response and ferroptosis. NOX4, CCDC80, COL1A2, HTRA1, and KLHL29 were identified as key genes by LASSO regression analysis, among which HTRA1 and KLHL29 showed a close association with immune system infiltration in DCM and obesity.
Obesity exposes the body to chronic inflammation. It impacts immune system function by disrupting the structure and function of lymphoid tissue and altering the distribution of white blood cells [30]. Immune cells are recruited to respond to the myocardial inflammation in DCM, whether from an infection or due to an auto-immune response [31]. Our study revealed that obesity showed a more disturbed immune environment compared to DCM, in which almost all immune cell populations are altered. The increased adipose tissue in obesity may be responsible for the disordered immune environment [32]. We found that M2 macrophages increased in both DCM and obesity, but especially in obesity. Adipose tissue macrophages (ATM) play a central role in obesity-associated inflammation. M2 macrophages are alternatively activated macrophages playing a role in the anti-inflammatory response, and undergo conversion to M1 as obesity progresses [33]. Increased infiltration of M2 macrophages into the myocardium in DCM is independently associated with cardiac fibrosis, leading to a poor prognosis [34]. T-cell infiltration is another typical inflammatory infiltration, which plays a part in the pathogenesis of inflammation in DCM [34]. HTRA1 and KLHL29 had a high correlation with CD8 T cells, CD4 naive T cells, gamma delta (γδ) T cells, M2 macrophages, and plasma cells. HTRA1, a highly conserved serine protease, is ubiquitous in various organisms, involved in certain signaling pathways, and participating in various disease pathogeneses [35]. HTRA1 and oxidative stress act synergistically to promote macrophage infiltration and inflammation in age-related macular degeneration (AMD) [36]. Although the role of HTRA1 in DCM has not been reported, D Colak [37] found HTRA1 significantly up-regulated (6.9-fold) in DCM and suggested that HTRA1 may contribute to cardiomyopathy pathways. Another interesting discovery suggested that HTRA1 was mainly expressed in plasma cells in inflamed gingival tissue [38], echoing our findings that HTRA1 had a high correlation coefficient to plasma cells (cor = 0.8, p < 0.05). It is reasonable to consider that HTRA1 plays a significant role in the cardiomyopathy inflammation of DCM. HTRA1 expression is increased in obese patients, especially in insulin-resistant (IR) adipose tissue, possibly related to the developmental and functional deficits of the adipocytes [39]. The levels of HTRA1, a negative regulator of mesenchymal stem cell (MSC) adipogenesis, and matrix metalloproteinase-13 (MMP-13) proteins, which played an important role in the pathophysiology of adipose tissue, were evidently high in the visceral adipose tissue of IR obese patients and also in the cardiac tissue of DCM patients [38][39][40], lending support to HTRA1 mediating the potential pathogenesis of DCM and obesity. KLHL29 is a protein-coding gene belonging to the conserved Kelch-like (KLHL) gene family whose expression is associated with micro-fragmented adipose tissue (MF), exerting an anti-inflammatory effect in osteoarthritis [41]. There is little research on KLHL29, but in our study KLHL29 was the only gene obviously decreased with statistical significance after immunotherapy in responders. Furthermore, KLHL29 had good AUCs and a close relation to CD8 T cells, CD4 naive T cells, gamma delta (γδ) T cells, and M2 macrophages in our study, which indicated that KLHL29 had a potential association with obesity and DCM, especially immunologically.
Oxidative stress is significantly associated with obesity and DCM [42,43]. NOX4 is a kind of isoform of NOX whose main biological function is to generate reactive oxygen species (ROS), expressed in various cardiovascular tissues and playing a complex role in the development of CVD [44]. NOX4 expression increasing in catalase-knockout mice adipocytes resulted in both adipogenesis and lipogenesis [45]. The increasing fatty acids in adipocytes induces the activation of NADPH oxidase (a main enrichment term in our study), increasing oxidative stress and production of ROS, ultimately causing metabolic syndrome [46]. Our results showed that NOX4 had an elevated expression in both DCM and obesity, connecting with immune response and ferroptosis. Ferroptosis promoted by oxidative stress-induced lipid peroxidation increases myocardial fibrosis, resulting in cardiomyopathy [47]. The activation of nucleotide-binding domain and leucine-rich repeat pyrin domain containing 3 (NLRP3) inflammasomes was closely associated with ferroptosis [48]. NOX4 may be involved in DCM progression by activating NLRP3 inflammasomes [49]. It is reasonable to assume that obesity promotes the development of DCM via NOX4 participating in the immune response and ferroptosis by activating NADPH oxidase and NLRP3 inflammasomes. The pathogenesis association needs to be explored via further study.
Intramyocardial fibrillar collagen increased significantly in patients with DCM, especially type I collagen [50]. COL1A2 encodes the pro-alpha2 chain of type I collagen, which is a main fibrillar collagen produced by heart fibroblasts [51]. TGF-β up-regulates the expression of COL1A2 in cardiac tissues and vascular smooth muscle, antagonizing the inhibitory effect of interferon gamma (IFN-γ) on it [51,52]. Obesity participates in cardiac fibrosis by up-regulating the expression of COL1A1 and COL1A2 in cardiac fibroblasts and further significantly increasing myocardial collagen content [53]. CCDC80, a protein secreted by adipocytes that regulates lipogenesis, is significantly elevated in visceral adipose tissue (VAT) in obesity [54]. CCDC80 has a close association with the fibrillin-1 affected by TGF-β signaling, and may be involved in the regulation of vascular tone [55]. Given that COL1A2, closely related to CCDC80, mediates the production of fibrillar collagen, we may assume that the combined action of CCDC80 and COL1A2 promotes myocardial fibrosis and accelerates the development of cardiomyopathy in the presence of obesity risk factors.
This study revealed the underlying relation between DCM and obesity by bio-informatic analysis, with a focus on oxidative stress, collagen synthesis, immune response, and ferroptosis. Antonini-Canterin et al. developed another index which showed a better performance than BMI in evaluating body fat-related cardiovascular risk, named the waist-corrected BMI (wBMI), calculated as the waist circumference (WC) × BMI [56]. Indeed, numerous studies have shown that BMI is not a good indicator of the risk of obesity comorbidities [57]. Performing a comprehensive clinical evaluation of obesity can provide a deeper understanding of obesity-related DCM. Further experiments and clinical trials are needed to validate our results and explore potential mechanisms of pathophysiology.

Conclusions
This study based on bio-informatics analyses found that obesity might exacerbate the development of DCM by having a stimulatory effect on collagen synthesis, influenced by immune response and ferroptosis related to oxidative stress. The key genes, NOX4, CCDC80, COL1A2, HTRA1, and KLHL29 were significantly expressed in DCM, with KLHL29 and HTRA1 especially closely associated with the immune response, which might be markers for obesity-induced DCM and be potential avenues for exploration in immunotherapy.