Abstract

   Dilated cardiomyopathy (DCM) causes heart failure and sudden death.
   Epigenetics is crucial in cardiomyopathy susceptibility and
   progression; however, the relationship between epigenetics,
   particularly DNA methylation, and DCM remains unknown. Therefore, this
   study identified aberrantly methylated differentially expressed genes
   (DEGs) associated with DCM using bioinformatics analysis and
   characterized their clinical utility in DCM. DNA methylation expression
   profiles and transcriptome data from public datasets of human DCM and
   healthy control cardiac tissues were obtained from the Gene Expression
   Omnibus public datasets. Then an epigenome-wide association study was
   performed. DEGs were identified in both DCM and healthy control cardiac
   tissues. In total, 3,353 cytosine-guanine dinucleotide sites annotated
   to 2,818 mRNAs were identified, and 479 DCM-related genes were
   identified. Subsequently, core genes were screened using logistic,
   least absolute shrinkage and selection operator, random forest, and
   support vector machine analyses. The overlapping of these genes
   resulted in DEGs with abnormal methylation patterns. Cross-tabulation
   analysis identified 8 DEGs with abnormal methylation. Real-time
   quantitative polymerase chain reaction confirmed the expression of
   aberrantly methylated DEGs in mice. In DCM murine cardiac tissues, the
   expressions of SLC16A9, SNCA, PDE5A, FNDC1, and HTRA1 were higher
   compared to normal murine cardiac tissues. Moreover, logistic
   regression model associated with aberrantly methylated DEGs was
   developed to evaluate the diagnostic value, and the area under the
   receiver operating characteristic curve was 0.949, indicating that the
   diagnostic model could reliably distinguish DCM from non-DCM samples.
   In summary, our study identified 5 DEGs through integrated
   bioinformatic analysis and in vivo experiments, which could serve as
   potential targets for further comprehensive investigation.

   Keywords: integrative bioinformatics analysis, dilated cardiomyopathy,
   epigenetics, DNA methylation, biomarker

Introduction

   Dilated cardiomyopathy (DCM) is a primary cardiac disorder
   characterized by the presence of ventricular systolic dysfunction
   accompanied by hypertrophy in either the left, right, or both
   ventricles. DCM is a leading contributor to the development of heart
   failure and unexpected mortality, and epidemiological studies have
   shown that DCM accounts for approximately 60% of all cardiomyopathies
   [31]^1. The worldwide prevalence of DCM is approximately 1:250 [32]^2.
   DCM is heterogeneous and lacks specific clinical manifestations during
   its early stages, thus creating severe challenges for accurate DCM
   diagnosis and management [33]^3^, [34]^4. The mortality rate of DCM
   remains high despite advancements in current understanding. The
   five-year mortality rate of patients with DCM in Europe was found to be
   15-50%, while the mortality rate of patients with DCM in China over 52
   months of follow-up reached as high as 42.24%, imposing a substantial
   cost on individuals, families, the economy, and society [35]^5^,
   [36]^6. Therefore, the enhancement of early diagnosis and treatment foe
   DCM is crucial.

   Epigenetics describes the alterations in genomic function caused by
   modifications to non-nucleotide sequences [37]^7, including RNA and DNA
   methylation and histone modification [38]^8. DNA methylation is a
   biochemical process that involves the addition of methyl groups to
   cytosine in DNA, resulting in the formation of 5-methylcytosine, which
   is facilitated by methyltransferases. While CpG sites are frequently
   associated with DNA methylation, it should be noted that DNA
   methylation can also occur at non-CpG sites [39]^9^, [40]^10. Because
   methyl groups are hydrophobic, DNA hypermethylation can modify
   chromatin structure, DNA stability, DNA structure, and interactions
   between DNA and proteins in order to regulate gene transcription
   [41]^9^, [42]^11. Epigenetics, particularly DNA methylation,
   significantly influences on cardiovascular disease susceptibility and
   changes in disease progression [43]^12^, [44]^13; however, its
   association with DCM remains undefined.

   Over the past decade, the application of advanced sequencing techniques
   and the integration of bioinformatics analyses have proven valuable in
   revealing previously unidentified genes and pathways involved in
   disease mechanisms. In this research, we conducted a thorough
   bioinformatic analysis of gene expression and DNA methylation data
   sourced from the Gene Expression Omnibus (GEO) database of the National
   Center for Biotechnology Information and conducted in vivo experiments
   to identify abnormally methylated genes in DCM. Simultaneously, we
   developed a logistic regression prediction model to assess the
   potential clinical utility of these hub genes in diagnosing DCM.

Materials and Methods

Epigenome-Wide Association Study

   An epigenome-wide association study (EWAS) is an association analysis
   tool utilizing DNA methylation data across the wide genome that aims
   used to analyze the relationships between complex phenotypes and
   epigenetic modification. Methylation microarray data from the
   [45]GSE81337[46]^14 dataset was analyzed using the "CpGassoc" tool
   ([47]https://CRAN.R-project.org/package=CpGasso) in R (version 4.2.0)
   ([48]https://www.r-project.org) to identify CpG sites associated with
   DCM. By applying this tool to the [49]GSE81337 dataset, we were able to
   narrow down their focus to CpG sites that are located in functionally
   relevant regions of the genome, specifically in promoter regions and
   exons. To ensure the reliability and significance of the findings, Only
   CpG sites located in the promoter region, including 200 base internal
   regions (TSS200), 1,500 base internal regions (TSS1500), and the first
   exon regions from the transcription start sites (TSS) were considered.
   The screening requirement of a false discovery rate (FDR) of < 0.001
   was applied to ensure that the identified CpG sites are statistically
   significant and unlikely to be false positives.

Differential Gene Analysis

   Comparative analysis of gene was performed to identify genes related to
   DCM. First, R software was used to transform the platform and matrix
   information files. After the initial data transformation, the data
   underwent standardization using the "normalizeBetweenArrays" purpose of
   the "limma"
   ([50]http://www.bioconductor.org/packages/release/bioc/html/limma.html)
   package. Finally, the differentially expressed genes (DEGs) of DCM in
   the [51]GSE42955[52]^15, [53]GSE79962[54]^16, [55]GSE57338[56]^17,
   [57]GSE84796[58]^18, and [59]GSE111544[60]^18 microarray datasets were
   identified using the limma package [61]^19. The limma package offers
   robust statistical methods to detect significant differences in gene
   expression between groups of samples. In addition to the statistical
   analysis, a literature review was also used to acquire DCM proteomic
   differentially expressed proteins directly [62]^20. Differentially
   expressed proteins and DEGs with |log Fold change (FC)| ≥ 0.5 and P <
   0.05 were deemed statistically significant. However, due to excessive
   DEGs with |logFC| ≥ 0.5 in the combined [63]GSE84796[64]^18 and
   [65]GSE111544[66]^18 datasets, the screening criteria were narrowed to
   |logFC| ≥ 1.

Weighted Gene Co-Expression Network Analysis

   Weighted gene co-expression network analysis (WGCNA) is a biological
   approach for analyzing gene expression patterns across numerous samples
   to uncover meaningful gene modules and their relationships with
   clinical traits [67]^21. WGCNA was performed using the "WGCNA" package
   in R software ([68]https://CRAN.R-project.org/package=WGCNA). First,
   this transcriptome data of the [69]GSE141910[70]^22 dataset with the
   largest number of samples were preprocessed to build a gene
   relationship network. Subsequently, different gene modules were then
   identified by constructing a cluster tree to identify distinct gene
   modules based on a gene co-expression network. Finally, a correlation
   analysis was conducted between gene modules and the DCM phenotype to
   identify the essential genes in the DCM modules. The threshold was
   fixed at 5 so that the scale-free network map structure R^2 > 0.8. The
   key genes associated with DCM were identified by performing an
   intersection analysis between DEGs, differentially expressed proteins,
   and the key genes of essential WGCNA modules in each dataset.

Construction of Protein-Protein Interaction Network

   The STRING database serves as a valuable resource for the analysis and
   prediction of protein-protein interactions (PPI) [71]^23. In our study,
   we utilized STRING (version 11.0) ([72]https://string-db.org/) to
   assess the protein interactions of interest. Specifically, we screened
   for interaction pairs with interaction scores greater than 0.4. To
   facilitate visualization and interpretation of the PPI network, we
   employed Cytoscape (version 3.8.2) [73]^24.

Enrichment Analysis

   The “clusterProfiler” package [74]^25
   ([75]https://CRAN.R-project.org/package=grandR) in R software was
   employed to gain a deeper understanding of the functional roles of the
   key genes associated with DCM. Gene Ontology (GO) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of
   relevent genes linked to DCM were performed, followed by selecting key
   GO functional enrichment entries using the “GOSemSim” package [76]^26.
   The functional enrichment entry screening criteria were P values of <
   0.05.

Analysis of Differentially Expressed Genes with Abnormal Methylation

   To further identify key DCM-related genes with strong diagnostic value
   and biological significance, we implemented a multi-step analysis using
   various statistical and machine learning techniques. Initially,
   logistic regression models were performed first using the "glm"
   function in the R software "stats" package, and an ROC curve was
   plotted utilizing the "pROC" package. Next, the R software "glmnet"
   package ([77]https://CRAN.R-project.org/package=glmnet) was used on the
   positive results of the logistic model. Then relevant analysis involved
   the implementation of LASSO regression, which aims to minimize absolute
   shrinkage and selection. A λ value of 13 was chosen as it minimized the
   deviation of the LASSO regression model. In parallel, the
   "randomForestSRC" package
   ([78]https://CRAN.R-project.org/package=randomForestSRC) was used to
   perform random forest analysis, and the "e1071" package
   ([79]https://github.com/cran/e1071) was used to train support vector
   machine models to further feature-screen key genes. Both random forest
   and SVM models were used to further screen for key genes based on their
   feature importance. The genes ranking among the top 10 were selected
   from the random forest analysis and support vector machine model and
   intersected. Finally, overlapping genes between above results and the
   findings from the EWAS analysis to obtain a set of abnormally
   methylated DEGs that are potentially linked to DCM.

Animals

   Male C57BL/6 mice, aged eight weeks, were acquired from Beijing Vital
   River Laboratory Animal Technology Co., Ltd. (located in Beijing,
   China) and reared in a specific pathogen-free environment at the
   Laboratory Animal Center of Tongji Medical College, Huazhong University
   of Science and Technology. These facilities maintained optimal
   conditions, including a room temperature range of 20-25 °C, humidity
   between 60-70%, and a regulated 12-hour light/dark cycle. Prior to
   experimentation, all mice were acclimatized for one week without any
   experimentation. The Animal Care and Utilization Committee of Huazhong
   University of Science and Technology, China, approved all animal
   experiments adhering strictly to the Guide for the Care and Use of
   Laboratory Animals (National Institutes of Health Publication 8th
   Edition, 2011).

Model Construction for Dilated Cardiomyopathy

   DCM was induced by injecting doxorubicin hydrochloride (DOX) (No.
   HY-15142; MedChemExpress, Monmouth Junction, NJ, USA) into male
   C57BL/6J mice to verify the expression of abnormally methylated DEGs in
   the DCM cardiac tissue [80]^27^, [81]^28. The male C57BL/6J mice, aged
   8 weeks and free from specific pathogens were randomly divided into
   saline and DOX groups. Mice were injected intraperitoneally with saline
   and DOX (5 mg/kg) every three days, ten times in total.
   Echocardiography was performed to assess the successful establishment
   of the DCM mode. The mice were fully anesthetized when the
   tail-pinching reflex disappeared. Mice were subject to euthanasia using
   a carbon dioxide chamber, followed by cervical dislocation for
   investigations involving the isolation of mouse tissues.

Echocardiographic Analysis

   Transthoracic echocardiography was performed 30 days after saline and
   DOX injections using a Vevo 2100 high-resolution microimaging system
   (VisualSonics, Toronto, Canada). Mice were placed on a warmed cushion
   after anesthetization by isoflurane inhalation (4% for induction, 1.7%
   for experiment), and their chest hair were shaved off. The body
   temperature was controlled at 36-37℃. The M-mode tracings were obtained
   from the short-axis 2-D view of the left ventricle. Measurements were
   taken for the left ventricular dimensions during end-diastole and
   end-systole (LVIDd and LVIDs), while left ventricular ejection fraction
   (LVEF) and left ventricular fractional shortening (LVFS) were
   calculated. The analysis of all loops and images from five cardiac
   cycles was conducted in a blinded manner, with the average value being
   utilized.

Histology

   The entire heart was removed and cleaned of blood, and the cardiac
   tissue was fixed overnight at room temperature (20-25 °C) with 4%
   paraformaldehyde. The tissues were paraffin-fixed to prepare 3-μm
   paraffin slices. The paraffin slices from the papillary muscle layer
   were stained with hematoxylin and eosin (HE) and Masson's trichrome.
   The visualization of all stained sections was conducted using a NIKON
   ECLIPSE E100 microscope (Nikon, Tokyo, Japan).

Real-Time Quantitative Polymerase Chain Reaction

   The extraction of total RNA was performed using TRIzol isolation
   reagent (Vazyme Biotec, Nanjing, China), followed by reverse
   transcription of the RNA into cDNA using HiScript RT SuperMix (Vazyme
   Biotec, Nanjing, China). To conduct real-time quantitative polymerase
   chain reaction (RT-qPCR), we utilized sequence-specific primers, ChamQ
   SYBR qPCR Master Mix (Vazyme Biotec), and a CFX96 Real-Time PCR
   Detection System (Bio-Rad Laboratories, Hercules, CA, USA). Each
   reaction was repeated thrice. The data were normalized to GAPDH via the
   2^-ΔΔCT method. The primer sequences utilized for amplifying the target
   genes can be found in [82]Table S1.

Construction of Logistic Regression Prediction Model

   Logistic regression is a widely used classification technique for
   predicting a classification based on a set of variables [83]^29. In
   this study, the transcriptome expression values of each abnormally
   methylated DEG were used to predict the sample type (DCM or non-DCM).
   Based on the two sample groups (DCM or non-DCM), the continuous
   independent variable for each DEG with abnormal methylation was
   represented by the transcriptome expression value, while the sample
   type was considered a binary variable. The R software was used to build
   a logistic regression model using the generalized linear model
   function. In addition, we also plotted the ROC curve of the logistic
   regression prediction model to assess the clinical diagnostic potential
   of abnormally methylated DEGs in DCM.

Statistical Analyses

   The statistical analyses were conducted utilizing GraphPad Prism8
   (GraphPad Software, San Diego, CA, USA). All experimental data are
   expressed as the mean ± standard deviation. A comparison between
   different groups was conducted using an independent sample t-test
   analysis. Statistical significance was considered at P < 0.05.

   Details regarding the extended methods are provided in the
   [84]Supplementary Material.

Results

Identification of Dilated Cardiomyopathy-Associated Cytosine-Guanine
Dinucleotide Sites

   To identify DEGs in the hearts of patients with DCM, we searched the
   GEO database ([85]https://www.ncbi.nlm.nih.gov/geo/) using the keywords
   "Dilated cardiomyopathy". After screening, we downloaded seven datasets
   with a sample size of at least 10 for analysis. These comprised the DNA
   methylation dataset [86]GSE81337[87]^14 and the original gene
   expression datasets [88]GSE42955[89]^15, [90]GSE79962[91]^16,
   [92]GSE57338[93]^17, [94]GSE84796[95]^18, [96]GSE111544[97]^18, and
   [98]GSE141910[99]^22.

   The screening process for aberrantly methylated DEGs in DCM is shown in
   Fig. [100]1.

Figure 1.

   [101]Figure 1
   [102]Open in a new tab

   Flowchart identifying abnormally methylated differentially expressed
   genes (DEGs). DCM, dilated cardiomyopathy; EWAS, epigenome-wide
   association study; LASSO, least absolute shrinkage and selection
   operator

   The analysis of EWAS was conducted on DNA methylation data from the
   [103]GSE81337[104]^14 dataset to obtain genome-wide critical CpG sites
   in DCM. A Manhattan plot illustrating the distribution of DCM-critical
   CpG sites on different chromosomes is presented (Fig. [105]2a). Each
   point in the plot represents a DCM-critical CpG site; the dashed line
   horizontally represents log10 (FDR), and CpG sites exceeding the dashed
   line have FDR < 0.001. A total of 3,353 CpG sites located in the
   promoter regions were identified as being involved in DCM development
   and were mostly located on chromosomes 1 and 2, which mapped to 2,818
   mRNAs (Fig. [106]2a). The majority of CpG sites within the genome are
   situated in the island region (Fig. [107]2b). In addition, analysis of
   the precise spreading of CpG sites within the promoter region revealed
   that a predominant localization of these sites in TSS1500 (Fig.
   [108]2c).

Figure 2.

   [109]Figure 2
   [110]Open in a new tab

   Dilated cardiomyopathy (DCM)-associated cytosine-guanine dinucleotide
   (CpG) sites (a) Manhattan plot of the DCM-associated CpG sites across
   all chromosomes. Horizontal axis: chromosome; longitudinal axis: log10
   (P value). (b) Distribution pattern of CpG sites within the CpG
   islands. (c) Promoter region CpG sites distribution.

Identification of Differentially Expressed Genes with Abnormal Methylation in
Dilated Cardiomyopathy

   The expression matrix data of the standardized [111]GSE42955[112]^15,
   [113]GSE79962[114]^16, [115]GSE57338[116]^17, and [117]GSE84796 +
   [118]GSE111544 [119]^18 datasets were used ([120]Fig. S1). A total of
   374, 714, 637, and 610 DEGs, respectively, were identified between the
   DCM group and the non-DCM group in the above datasets (Fig. [121]3a).
   There was a notable disparity in the expression levels of DEGs observed
   between the DCM and non-DCM cohorts (Fig. [122]3a). The obtained DEGs
   were utilized for subsequent analyses. The heatmap visualized all of
   the identified DEGs (Fig. [123]3b).

Figure 3.

   [124]Figure 3
   [125]Open in a new tab

   Differentially expressed genes (DEGs) of the [126]GSE42955,
   [127]GSE79962, [128]GSE57338, and [129]GSE84796 + [130]GSE111544
   datasets. (a) Volcano map of the [131]GSE42955, [132]GSE79962,
   [133]GSE57338, and [134]GSE84796 + [135]GSE111544 datasets. (b) Heatmap
   of the [136]GSE42955, [137]GSE79962, [138]GSE57338, and [139]GSE84796 +
   [140]GSE111544 datasets

   Next, to comprehensively identify key modules and hub genes associated
   with DCM, we employed WGCNA to construct a co-expression network using
   the [141]GSE141910[142]^22 dataset. The samples were subjected to
   clustering analysis by the removal of any outliers (Fig. [143]4a), and
   a soft threshold of 5 was chosen based on scale-free topological
   criteria to establish a weighted adjacency matrix (Fig. [144]4b). The
   construction of a co-expression network was performed based on the
   optimal soft threshold, and a gene clustering tree was generated.
   Subsequently, 12 modules resulting from module clustering were analyzed
   for conservation (Fig. [145]4c and [146]4d). Brown modules with the
   smallest P values and the highest correlation indices were considered
   the most relevant to the DCM features (Fig. [147]4e).

Figure 4.

   [148]Figure 4
   [149]Open in a new tab

   Application of the weighted gene co-expression network analysis (WGCNA)
   on the [150]GSE141910 dataset. (a) Sample clustering plot. (b)
   Measuring the topological fit index and average connectivity for
   scale-free networks. (c) Gene clustering tree plot. (d) Modular
   conservation analysis. (e) Association analysis plot between each gene
   module and dilated cardiomyopathy (DCM) phenotype

   The abovementioned DEGs (Fig. [151]3a) and key module genes were
   intersected with proteomic differentially expressed proteins in DCM
   reported in the literature [152]^20 to identify repetitive genes,
   revealing 479 essential DCM-related genes (Fig. [153]5a and [154]5b).
   The STRING database for protein interactions was used to investigate
   these 479 DCM-related essential genes, and a total of 479 nodes and
   1,702 edges of the PPI network were obtained, indicating interactions
   between genes and proteins (Fig. [155]5c).

Figure 5.

   [156]Figure 5
   [157]Open in a new tab

   Identification and protein-protein interaction (PPI) network of key
   genes linked to dilated cardiomyopathy (DCM). (a) Venn diagram. (b)
   Upset plot. (c) PPI network diagram of key genes associated with DCM

   DCM-related essential genes were significantly enriched, and the top 10
   GO biological process (BP) (Fig. [158]6a), GO cellular component (CC)
   (Fig. [159]6b), GO molecular function (MF) (Fig. [160]6c), and KEGG
   (Fig. [161]6d) pathways were identified. GO analysis indicated that the
   DEGs showed enrichment in pathways related to remodeling of the
   extracellular matrix pathways, indicating potential alterations in the
   extracellular matrix associated with DCM (Fig. [162]6a-c).

Figure 6.

   [163]Figure 6
   [164]Open in a new tab

   Histogram of enrichment analysis of key genes associated with dilated
   cardiomyopathy (DCM). (a) Scores of the enrichment in the top 10 of
   Gene Ontology (GO) biological process (BP) analysis of enriched
   pathways. (b) Top 10 enrichment scores of GO cellular component (CC)
   pathways enrichment analysis. (c) Scores of the enrichment in the top
   10 of GO molecular function (MF) pathways enrichment analysis. (d) Top
   10 enrichment scores of Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analysis

   Logistic regression analysis was performed, and the [165]Fig. S2
   displays the ROC curves. The LASSO regression, random forest, and
   support vector machine models were used for feature screening of the
   positive logistic results. The intersection of 13, 10, and 10 key
   genes, identified using LASSO regression analysis (Fig. [166]7a and
   [167]7b), random forest analysis (Fig. [168]7c and [169]7d), and
   support vector machine analysis, respectively, was performed to obtain
   the hub genes. Eight DEGs exhibiting abnormal methylation patterns were
   identified: choline dehydrogenase (CHDH), regulator of G protein
   signaling 9 binding protein (RGS9BP), solute carrier family 16 member 9
   (SLC16A9), Fibronectin type III domain-containing protein 1 (FNDC1),
   phosphodiesterase 5A (PDE5A), HTRA1, synuclein alpha (SNCA), and
   neuronal pentraxin 2 (NPTX2) (Fig. [170]7e).

Figure 7.

   [171]Figure 7
   [172]Open in a new tab

   Identification for aberrantly methylated differentially expressed genes
   (DEGs) using various computational algorithms. (a-b) Regression model
   utilizing the LASSO technique for minimizing absolute shrinkage and
   selecting variables. (c-d) The random forest model. (e) Venn plots of
   candidate genes in the LASSO regression, random forest, and support
   vector machine models

In Vivo Validation of Differentially Expressed Genes with Abnormal
Methylation

   We used doxorubicin to induce a DCM model in mice[173]^27^, [174]^28.
   Subsequently, the success of the model was assessed through
   echocardiography and quantification of cardiac function in mice. Mice
   with DCM exhibited enlarged heart chambers, thin cardiac walls,
   impaired myocardial mobility and diminished systolic function (Fig.
   [175]8a). Compared to the control group, mice with DCM exhibited a
   significant decrease in LVEF and LVFS, as well as a notable increase in
   LVIDs. However, no statistically significant difference was observed in
   LVIDd (Fig. [176]8b), suggesting successfully construction of the DCM
   mouse model.

Figure 8.

   [177]Figure 8
   [178]Open in a new tab

   Echocardiographic images and statistical data of mice. (a)
   Representative M-mode echocardiographic images of the left ventricle in
   the control and doxorubicin hydrochloride (DOX) groups. (b) Analysis of
   left ventricular ejection fraction (LVEF), left ventricular fractional
   shortening (LVFS), left ventricular internal diameter in systole
   (LVIDs), and left ventricular internal diameter in diastole (LVIDd). n
   = 6 per group. Data in (b) were analyzed using independent sample
   t-test analyses. Data are expressed as the means ± standard deviation.
   ***P < 0.001

   HE staining revealed myocardial fiber disintegration and disarray in
   the DOX group. Additionally, cardiac muscle cells displayed hypertrophy
   accompanied by vacuolar degeneration, disrupted arrangement, enlarged
   and deformed nuclei with intense staining (Fig. [179]9a). Masson's
   staining indicated that the DOX group had more blue-dyed collagen
   fibers than the control group had, indicating more interstitial
   fibrosis. The myocardial small artery wall undergoes simultaneous
   thickening and lumen narrowing (Fig. [180]9b).

Figure 9.

   [181]Figure 9
   [182]Open in a new tab

   Photomicrographs of pathological staining in the control and
   doxorubicin hydrochloride (DOX) groups. (a) Photomicrographs of
   hematoxylin and eosin staining. (b) Representative photomicrographs of
   Masson's staining. Scale bar: 100 μm

   The successful establishment of the DCM model was confirmed by
   echocardiography and histological staining. Subsequently, we assessed
   the expression levels of the aforementioned DEGs in cardiac tissue. The
   SLC16A9, SNCA, PDE5A, FNDC1, and HTRA1 transcript levels were found to
   be significantly elevated in the cardiac tissues of the DCM group than
   those of the control group (Fig. [183]10). Conversely, the other three
   genes (CHDH, RGS9BP, and NPTX2) were not sufficiently expressed to be
   detected in the mouse cardiac tissue.

Figure 10.

   [184]Figure 10
   [185]Open in a new tab

   Expressions of SLC16A9, SNCA, PDE5A, FNDC1, and HTRA1 mRNA of cardiac
   tissue in dilated cardiomyopathy (DCM) mice, determined using
   quantitative real-time polymerase chain reaction. Each group contained
   four samples. (a) SLC16A9 transcripts in cardiac tissue. (b) SNCA
   transcripts in cardiac tissue. (c) PDE5A transcripts in cardiac tissue.
   (d) FNDC1 transcripts in cardiac tissue. (e) HTRA1 transcripts in
   cardiac tissue. Data in (a-e) were analyzed using independent sample
   t-test analyses. Data are expressed as the means ± standard deviation.
   * P < 0.05, ** P < 0.01, **** P < 0.0001

Construction of Logistic Regression Prediction Model

   Five DCM-related DEGs exhibiting abnormal methylation (SLC16A9, SNCA,
   PDE5A, FNDC1, and HTRA1) were used to create logistic regression
   models, and the dependent variables were either of the DCM or non-DCM
   sample type. The risk score was calculated as follows: Risk score =
   3.94 × FNDC1 + 7.26 × HTRA1 + 1.91 × PDE5A + 3.14 × SLC16A9 - 2.22 ×
   SNCA. The [186]GSE141910[187]^22 and [188]GSE57338[189]^17 datasets
   were used as the training and validation sets, respectively, to
   evaluate the accuracy of the established logistic regression prediction
   model. The training dataset revealed significant variations in risk
   ratings between DCM and non-DCM samples (P < 0.01) ([190]Fig. S3).
   Additionally, the risk scores between the DCM and non-DCM samples in
   the validation set differed significantly (P < 0.01) (Fig. [191]11a).
   This finding demonstrates that the logistic regression prediction model
   effectively distinguished DCM from non-DCM samples. Furthermore, the
   expression of the 5 DCM-related DEGs with abnormal methylation varied
   significantly between the DCM and non-DCM samples (Fig. [192]11b-f),
   thereby demonstrating the critical role of these genes in DCM. The area
   under the curve of the logistic regression prediction model was 0.949,
   indicating a high level of the accuracy in the predictions of this
   model (Fig. [193]11g). Thus, the potential diagnostic efficacy of the 5
   DCM-related DEGs with abnormal methylation was demonstrated.

Figure 11.

   [194]Figure 11
   [195]Open in a new tab

   The results of the logistic regression prediction model of the
   [196]GSE57338 dataset. (a) A significant difference in risk scores
   between dilated cardiomyopathy (DCM) and non-DCM samples was observed.
   (b-f) The expression values of 5 aberrantly methylated DEGs (SLC16A9,
   SNCA, PDE5A, FNDC1, and HTRA1) differed significantly between DCM and
   non-DCM samples. (g) Receiver operating characteristic (ROC) curve of
   the logistic regression prediction model. ****P < 0.0001

Discussion

   Due to the difficulties in diagnosing DCM, many patients are typically
   presented with grade III-IV cardiac function, resulting in a dismal
   prognosis [197]^30. As the pathophysiology of DCM remains elusive,
   research efforts in this field have increasingly pivoted towards
   advancing early detection and treatment strategies, aiming to mitigate
   myocardial damage and improve patient outcomes [198]^31. Gene
   expression microarrays, next-generation transcriptome sequencing, and
   crucial bioinformatics components have been extensively used to study
   cardiovascular diseases and to offer various opportunities for
   molecular treatment, molecular prediction, and drug targeting
   [199]^32^, [200]^33.

   Reduced DNA methylation levels of the trypsin receptor 3 gene impact
   mRNA production, thereby elevating the likelihood of myocardial
   infarction [201]^34. Fifty-two genome-wide CpG methylation sites linked
   to myocardial infarction were found in a follow-up study involving
   11,461 people. The biological activity of these CpG sites showed that
   they are primarily associated with calcium metabolism and renal
   function [202]^35. According to Chinese cohort studies, racial
   disparities exist in the correlation between DNA methylation and common
   coronary heart disease risk factors [203]^36^, [204]^37. Conclusively,
   DNA methylation is highly associated with cardiovascular illness;
   however, there is currently insufficient evidence linking DNA
   methylation to DCM

   In this study, we used bioinformatics techniques to thoroughly analyze
   six transcriptome datasets and one DNA methylation dataset of DCM to
   identify DEGs for DCM-related aberrant methylation. Overall, 3,353 CpG
   sites, which may map to 2,818 mRNAs, were associated with DCM
   development. These sites were primarily concentrated on chromosomes 1
   and 2. A total of 479 key DCM-related genes were identified by means of
   differential analysis, WGCNA, and differential expression protein
   screening. Eight DCM-related DEGs with abnormal methylation were
   identified using logistic regression, LASSO regression, random forest,
   and support vector machine analyses.

   Of the 8 genes exhibiting aberrant DCM-related methylation, 5 genes -
   SLC16A9, SNCA, PDE5A, FNDC1, and HTRA1 - displayed significantly higher
   expression levels in the cardiac tissue of DCM mice compared to healthy
   counterparts. However, the expression of the remaining three genes was
   too low to be reliably evaluated. SLC16A9, a transporter gene, oversees
   the intestinal and renal excretion of uric acid. Abnormal expression of
   SLC16A9 can hinder uric acid excretion, resulting in hyperuricemia,
   [205]^38, which is a recognized risk factor for cardiovascular diseases
   [206]^39^, [207]^40. SNCA affects neurotransmitter release by encoding
   α synuclein, which affects normal neuronal function and was the first
   causative gene found to be associated with hereditary Parkinson's
   disease; however, its association with cardiovascular disease has not
   yet been elucidated [208]^41. PDE5A is highly expressed in vascular
   smooth muscle cells and mediates vasodilation [209]^42^, [210]^43.
   Sildenafil increases abdominal aortic aneurysms by inhibiting PDE5A
   [211]^44. FNDC1 is a biomarker of aortic calcified valves and interacts
   with lipid components in plasma to promote the development of
   inflammatory responses during aortic valve calcification [212]^45.
   Human serine protease HTRA1 involves several physiological processes,
   including mitochondrial homeostasis regulation, apoptosis, and cell
   signal transduction [213]^46. The structure and function abnormalities
   of HTRA1 lead to transforming growth factor-β expression variations,
   impacting cardiovascular disease progression [214]^47^, [215]^48. HTRA1
   methylation could be used as a possible diagnostic tool to diagnose
   strokes [216]^49. Our logistic regression prediction model, based on
   these 5 DCM-related DEGs with aberrant methylation, successfully
   distinguished DCM samples from non-DCM samples, further illustrating
   that epigenetic regulation of these 5 DCM-related DEGs may play an
   essential role in the pathogenesis of DCM.

   This study demonstrated a correlation between DNA methylation and DCM;
   however, it had several limitations. Most data were obtained from
   public databases, and while some clinical data were needed for more
   comprehensive research, there was no particular direct mechanism of
   action study to support the mechanism of action.

Conclusion

   We comprehensively analyzed the transcriptome and DNA methylation data
   of DCM and identified 8 DCM-related DEGs with aberrant methylation.
   Five (SLCA6A9, SNCA, PDE5A, FNDC1, and HTRA1) were highly expressed in
   the cardiac tissue of DCM mouse models. In addition, the logistic model
   established for these 5 genes showed that DCM samples could be
   accurately distinguished from non-DCM samples, suggesting that these
   genes are associated with the occurrence and prognosis of DCM. These
   discoveries have increased our understanding of DCM and offer novel
   guidelines for future therapeutic strategies. These 5 genes could be
   used in future studies to determine their role in DCM pathogenesis and
   to elucidate their mechanism of action.

Supplementary Material

   Supplementary materials and methods, figures and table.
   [217]ijmsv21p1769s1.pdf^ (4.2MB, pdf)

Acknowledgments