Integrative methylome and transcriptome analysis of porcine abdominal fat indicates changes in fat metabolism and immune responses during different development

Abstract

Fat is involved in synthesizing fatty acids (FAs), FA circulation, and lipid metabolism. Various genetic studies have been conducted on porcine fat but understanding the growth and specific adipose tissue is insufficient. The purpose of this study is to investigate the epigenetic difference in abdominal fat according to the growth of porcine. The samples were collected from the porcine abdominal fat of different developmental stages (10 and 26 weeks of age). Then, the samples were sequenced using MBD-seq and RNA-seq for profiling DNA methylation and RNA expression. In 26 weeks of age pigs, differentially methylated genes (DMGs) and differentially expressed genes (DEGs) were identified as 2,251 and 5,768, compared with 10 weeks of age pigs, respectively. Gene functional analysis was performed using GO and KEGG databases. In functional analysis results of DMGs and DEGs, immune responses such as chemokine signaling pathways, B cell receptor signaling pathways, and lipid metabolism terms such as PPAR signaling pathways and fatty acid degradation were identified. It is thought that there is an influence between DNA methylation and gene expression through changes in genes with similar functions. The effects of DNA methylation on gene expression were investigated using cis-regulation and trans-regulation analysis to integrate and interpret different molecular layers. In the cis-regulation analysis using 629 overlapping genes between DEGs and DMGs, immune response functions were identified, while in trans-regulation analysis through the TF-target gene network, the co-expression network of lipid metabolism-related functions was distinguished. Our research provides an understanding of the underlying mechanisms for epigenetic regulation in porcine abdominal fat with aging.

This study provides an understanding of the basic epigenetic mechanism of epigenetic regulation of pig abdominal fat with aging using multi-omics integration methods.

Introduction

Muscle growth and fat deposition are complex quantitative traits that are economically important for meat production from livestock. Phenotypic features of livestock animals, such as average daily weight gain, back thickness, and carcass composition, contribute to production efficiency. Intramuscular fat and fatty acid (FA) content, along with fattening efficiency, fertility, and immunity, are closely related to indicators of meat quality such as oxidative stability, tenderness, and juiciness of pork (Wood et al., 1999, 2008). Fat in pigs is primarily located in the abdominal, subcutaneous, intermuscular (between muscles), and intramuscular (within muscle) regions (Mourot et al., 1995; Monziols et al., 2007). These tissues are dissimilar and exhibit differences in regional production and glucose oxidative lipolysis (Budd et al., 1994; Mourot et al., 1995). Adipose tissue, which is the site of de novo FA synthesis in pigs, is also involved in FA circulation and lipid metabolism (O’Hea and Leveille, 1969). Moreover, it also produces and releases lipid hormones (lipokines such as palmitolate) and peptide hormones such as leptin, adiponectin, estrogen, and resistin (Fasshauer and Bluher, 2015). In addition, adipose tissue also secretes adipocytokines, such as tumor necrosis factor α (TNFα), that are involved in metabolic homeostasis and the development of obesity-related conditions, such as insulin resistance, inflammation, hypertension, cardiovascular disease, and metabolic disorders (Kershaw and Flier, 2004; Cao et al., 2008).

Phenotypic traits associated with fat metabolism and meat quality are determined by interactions between several genetic and environmental factors; hence, understanding these genetic mechanisms can help improve productivity. Gene markers associated with backfat, intramuscular fat content (Malek et al., 2001), and tissue-specific FA composition (Munoz et al., 2013) have been identified in expression quantitative trait loci studies. Moreover, a previous study has demonstrated the pro-inflammatory role of adipocyte differentiation and the MAPK signaling pathway by conducting transcriptome analysis in the fat tissue(Huang et al., 2017). Studies on the properties and contents of saturated FA and ω-6 polyunsaturated FA(Wang et al., 2013) as well as those focusing on reproduction (Li et al., 2012) and development have shed light on the role of the adipose tissue (Jiang et al., 2013). While several genetic markers of these traits have been identified, the previous results did not provide sufficient insight into the molecular mechanisms underlying the phenotypes observed in adipose tissue. In pigs, there are a number of factors that cause various changes in the body. The development of fat storage and the ratio of FA composition was reported to differ according to the growth stage of pigs, and other studies have demonstrated that the breed and type of feed consumed by pigs influence the lipid composition of pork (Kouba et al., 2003; Ayuso et al., 2020). In addition, differences in fat composition, carcass weight, and fatness were observed on the basis of immune castration and standard with repeated social mixing(Kress et al., 2019; Skrlep et al., 2020).

Another important factor that can influence phenotypic traits is epigenetics. Epigenetics refers to heritable modifications that change DNA or related proteins without changing the DNA sequence itself (Egger et al., 2004). These epigenetic modifications include DNA methylation, histone modifications, and RNA interference (Su et al., 2011). Among the various epigenetic mechanisms, DNA methylation is known to play an important role in gene silencing, genomic imprinting (Weber et al., 2007; Suzuki and Bird, 2008), X-chromosome inactivation (Carrel andWillard, 2005), cancer progression (Ehrlich, 2009), embryonic development, and tissue differentiation (Igarashi et al., 2008; Sasaki and Matsui, 2008; Liang et al., 2011). Many previous genomic studies have shown that epigenetic factors regulate gene expression via local methylation and affect processes like skeletal muscle development and metabolism (Guller and Russell, 2010; Neguembor et al., 2014; Svensson and Handschin, 2014; Carrio and Suelves, 2015). While DNA methylation in the promoter region of a gene typically suppresses its expression, methylation in the gene body suppresses or promotes gene expression (Dhar et al., 2021).

To study the mechanisms of epigenetic regulation according to changes in DNA methylation, previous studies have tried to integrate the omics datasets of DNA methylation and gene expression and confirm the relationship of DNA methylation at the genome level with mRNA, microRNA, or long noncoding RNA expression. These studies, called multi-omics integration (MOI) studies, have used a variety of approaches(Kim andKim, 2021) and have suggested the direct or indirect effects of various epigenetic regulatory factors on growth (Ziller et al., 2013), breed (Xi et al., 2019), litter size (Hwang et al., 2017), and diet (Zhang et al., 2017) in pigs. Therefore, the study of transcriptome and genome-wide methylation together is expected to lead to a deeper understanding of the epigenetic regulation of the adipose tissue phenotype.

Previous studies on adipose tissue in pigs have mainly focused on the three tissues associated with muscle, excluding abdominal fat. However, abdominal fat is deposited early during development and exhibits unique biochemical properties that distinguish it from subcutaneous fat (Deveaud et al., 2004; Strissel et al., 2007; Murano et al., 2008). Since the quality of pork is dependent on lipid formation and deposition, a better understanding of the epigenetic processes underlying lipid metabolism in pigs can improve the efficiency of pork production. Therefore, we aimed to investigate and apply an MOI method for understanding the mechanism of underlying epigenetic changes and to study the epigenetic changes in methylome and transcriptome analyses of abdominal fat during growth in pigs.

Materials and Methods

Ethics statement and sample collection

A total of nine pigs that were F1 crossbreeds of Korean native and Yorkshire breeds were used in this study. The experiment was reviewed and approved by the Institutional Animal Care and Use Committee, National Institute of Animal Science, South Korea (IACUC no. NIAS2016-848). Ad libitum food and water were provided to the animals during the experiment. Piglets were randomly allocated to two groups and were raised for 10 weeks (stage 1, n = 5) and 26 weeks of age (stage 2, n = 4). The animals were sacrificed at the end of each growth stage, and the abdominal fats were collected for RNA sequencing (RNA-Seq) and methyl-CpG-binding domain sequencing (MBD-seq) analyses. The tissues were dissected into pieces (1 cm³), snap-frozen in liquid nitrogen, and stored at −80 °C until use.

DNA extraction for MBD-seq

The MBD-seq was performed to analyze the whole genomic methylated DNA. We used the MethylMiner Methylated DNA Enrichment kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. Briefly, 1 µg of genomic DNA was fragmented using adaptive focused acoustic technology (Covaris) and captured by MBD proteins. The extracted methylated DNA was eluted in a high-salt elution buffer. DNA in each eluted fraction was precipitated using glycogen, sodium acetate, and ethanol and resuspended in DNase-free water. The eluted DNA was used to generate libraries following the standard protocols of the TruSeq Nano DNA Library Prep kit (Illumina). The eluted DNA was repaired, an adaptor was ligated to the 3ʹ end, and TruSeq adapters were ligated to the fragments. Once ligation was assessed, the adapter-ligated product was polymerase chain reaction (PCR) amplified. The final purified product was quantified using quantitative PCR (qPCR) according to the qPCR Quantification Protocol, qualified using Agilent Technologies 4200 TapeStation (Agilent technologies), and sequenced using the HiSeq 4000 platform (Illumina).

Preprocessing and identification of differentially methylated genes (DMGs) for MBD-seq

The quality of raw paired-end reads from the MBD-seq data generation was checked using the FastQC (v 0.11.7). Raw reads were processed and trimmed for adapter sequences with TrimGalore software (v 0.6.0) using criteria for quality scores (less than 20) and read lengths (less than 20 bp). We only used qualified trimmed reads for further analysis. The clean reads were aligned to the porcine reference genome (Sscrofa11.1/susScr11) downloaded from the Ensembl database (https://www.ensembl.org/) using bowtie2 (v 2.3.5.1). In the bam file mapped to the reference genome, the correlation between each sample was obtained by the deepTools (v 3.3.0) plotCorrelation function using the Spearman method.

To identify the differentially methylated regions (DMRs) between the two age points of the genome, MEDIPS and the BSgenome package on R (v 3.5.0) were used. The parameters specified for this analysis were set to: window size = 500 bp, default for other options. Window size indicates the size of the adjacent windows into which the genome is divided for the length set; this window size was used for all further calculations. In this study, the insert size of paired-end reads was set to 500 bp. The “uniq” term allows more readings mapped to the same genomic location as the value is close to zero. To identify DMR, MEDIPS calculates the average fragment per kilobase of transcript per million mapped reads (FPKM) for each sample and the average relative methylation score that overlaps with the genome-wide 500-bp windows. MEDIPS also calculates the false discovery rate (FDR) value by comparing the relative methylation score signal distribution in 500bp windows. The coupling factor was set based on 10 weeks of age samples (control) for normalization. Then, normalized FPKM values were exported for each sample across the entire genome. We merged nearby differentially methylated windows into DMRs (distance parameter = 1). Significant DMRs were selected by filtering using the FDR and methylated level. Hypermethylated DMRs were defined as FDR < 0.05 where methylation level ≥ 2 (log₂ value 1), and hypomethylated DMRs were defined as FDR < 0.05 where methylation level ≤ −2 (log₂ value −1).

The Ensembl database (https://www.ensembl.org/) was used to annotate DMGs from significant DMRs. Depending on the DMRs, we annotated the associated genomic elements as follows: promoter, exon, and intron. The promoter region was defined as 2,000 bp upstream of the transcription start site (TSS).

Significant DMGs were defined by specifying a cutoff value. The cutoff values were calculated for each gene region, by considering the ratio of the window size used to discover the DMRs for each gene feature. The median value was used to calculate the cover for each gene feature. When different features were annotated for the same gene, DMRs with more significant methylation levels were defined as DMGs to avoid overlapping genes.

Transcriptome sequencing and analysis of differentially expressed genes (DEGs)

Total RNA was extracted from the dissected frozen tissues using TRIzol reagent (Invitrogen), and the quality of the extract was checked using NanoDrop (ThermoFisher Scientific, Waltham, MA, USA). After performing quality control, the samples were used for library preparation. The sequencing library was prepared by random fragmentation of the cDNA sample, followed by 5ʹ and 3ʹ adapter ligation. Adapter-ligated fragments were then PCR-amplified and gel purified. For cluster generation, the sequencing library was loaded into a flow cell where fragments were captured on a lawn of surface-bound oligos complementary to the library adapters. Each fragment was then amplified into distinct, clonal clusters through bridge amplification. After cluster generation, the RNA-Seq was performed using the Illumina HiSeq 4000 sequencer.

The quality of RNA-Seq data was checked by examining the raw paired-end sequencing data in FastQC. Then, the reads were trimmed with Trimmomatic (v 0.38) to remove low-quality bases and adapters. Next, the reads were revalued in FastQC to confirm quality improvement. Subsequently, high-quality reads were aligned with the pig (Sus scrofa) reference genome obtained from the Ensembl database (https://www.ensembl.org/) using default options in HISAT2 (v 2.1.0). We converted BAM files into SAM files using SAMtools (v 1.9). Next, we estimated the count of mapped reads for each annotated gene in the Sus scrofa 11.1.98 gene transfer file format using featureCounts (v 2.0.0). We obtained the expression of 31,907 genes after serial preprocessing.

Genes with a quantification value of 10 or less were removed to improve the statistical ability to identify DEGs. Since this value was individually constructed and sequenced for each time point, normalized read counts were obtained using the trimmed mean of M-value (TMM) method (control = 10 weeks of age) by edgeR package (v 3.24.3) in R to ensure accurate quantification. Multidimensional scaling (MDS) analysis was performed to check the similarity between normalized values of samples and plotted using ggplot2 (v 3.3.1). Normalized RNA-Seq values were applied to compare gene expression levels at 10 weeks and 26 weeks of age using a generalized linear model. The resulting P-value was adjusted to control FDR using Benjamini and Hochberg’s method. Significant DEGs were identified by FDR < 0.05 and classified as upregulated or downregulated genes according to whether fold change (FC) in expression at 26 weeks of age relative to that at 10 weeks of age was greater than 2 (log₂ = 1) or less than −2 (log₂ = −1).

Functional enrichment and pathway analysis of DMGs and DEGs

DEG sets were compared between 10 weeks and 26 weeks of age group and subjected to functional enrichment and pathway analyses using the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) databases through the Database for Annotation, Visualization, and Integrated Discovery (v 6.8) (Dennis et al., 2003) bioinformatics tools. GO enrichment analysis of biological processes, cellular components, and molecular function was performed simultaneously, and significant terms were represented by a fold enrichment level and a −log₁₀  P value. The REVIGO visualization tool was used to make treemaps for the enriched GO terms. KEGG pathway analysis results were visualized using ggplot.

We applied TMM values from RNA-Seq data to gene set enrichment analysis (GSEA) (Subramanian et al., 2005). GSEA identifies genes (grouped as gene sets) enriched in expression and methylation compared to the control by using the hallmark database. To estimate the statistical significance (nominal P-value) of the enrichment score (ES), we conducted an empirical 1,000 gene set permutation test procedure that preserves the complex correlation structure of the gene expression data. The ES is the maximum deviation from zero encountered in the random walk and uses a weighted Kolmogorov–Smirnov-like statistic. GSEA results were visualized via bubble plot using R, and the significant gene set is shown through an enrichment plot and heatmap.

MOI of transcriptome and methylome

Two approaches were used to validate the interactions between DNA methylation and gene expression. First, we screened for overlapping DEGs and DMGs and divided them into four groups: hypomethylation-downregulation, hypomethylation-upregulation, hypermethylation-downregulation, and hypermethylation-upregulation. All overlapping differentially expressed and methylated genes were enriched using the KEGG database. Methylation and expression changes were visualized using Pathview (https://pathview.uncc.edu).

Second, the trans-effect was confirmed using the transcription factor (TF) database. TFs were identified in pigs using PANTHER (http://www.pantherdb.org/), and the target genes of TFs (among the DEGs or DMGs) were identified with Enrichr (https://maayanlab.cloud/Enrichr/) using a human database because targets of TFs are better established in human studies. Further, two databases of ChEA (http://amp.pharm.mssm.edu/lib/chea.jsp) and ENCODE (https://www.encodeproject.org/) were used to classify the target genes of the TFs, and common TF target genes were selected in both databases. Functional enrichment analysis was performed using the BioPlanet database that supports interactive browsing, retrieval, and analysis of pathways, exploration of pathway connections, and pathway search by gene targets, category, and availability of corresponding bioactivity assay. Finally, a network of TF gene targets and functions was constructed using Cytoscape.

Results

Overview of DNA methylome and transcriptome profiling

We performed MBD-seq on the abdominal fat of 10 and 26 weeks of age pigs to understand DNA methylation during pig growth. A total of 110 million reads were generated in the whole sample. On average, 12.8 million reads were generated for 10 weeks of age pigs, and 11.5 million reads were generated for 26 weeks of age pigs (Supplementary Table S1).

We observed correlations of 0.8 or more between samples within the same growth phase, which were clustered with respect to the growth stage (Figure 1A). MDS analysis used to normalize 19,309 DMRs confirmed clustering according to different growth phases (Supplementary Table S2), allowing further analysis (Figure 1B) and confirming methylation levels near the gene body (Figure 1C). Methylation levels were higher in 10 weeks of age samples than those in 26 weeks of age samples. Upstream DNA methylation decreased toward the TSS, but high levels of methylation were observed in the gene body. Methylation again decreased downstream of the transcription end site but partially recovered thereafter.

Figure 1.

Genome-wide DNA methylation patterns and genomic annotation distribution in abdominal fat according to the swine growth period. (A) Spearman correlation matrix among the individual samples based on the raw methylation counts. (B) MDS plot displaying the variation between normalized counts of individual samples. Colors represent the age of samples in the experiment. (C) DNA methylation patterns around gene bodies in the growth stage are measured by MBD-seq. (D) Upset plot for the number of DMRs annotated by genomic regions and their intersections. (E) Bar plot for DMRs and DMGs in different intragenic regions and intergenic. Bar thickness distinguishes DMRs and DMGs: the thick bar represents DMRs, and the thin bar represents DMGs.

Next, we examined DMRs annotated with each genetic feature. DMRs uniquely located in the intron accounted for 9,216 (47.7%) of the total DMRs, followed by 5,786 (30%) located in the intergenic region. The exon and intron domains, which are classified as the gene body, shared a number of domains, and 12,831 DMRs (66.5%) uniquely intersected these domains. Promoters spanned all regions, and no independently annotated DMRs were identified (Figure 1D). DMGs were finally identified through the cutoff for each gene feature, as suggested in Supplementary Table S3.

Moreover, we identified 12,682 DMRs (hypomethylation 10,983 and hypermethylation 1,699) in intron regions, 6,255 intergenic DMRs (hypomethylation 5,375 and hypermethylation 880), 4,039 DMRs in exons (hypomethylation 3,658 and hypermethylation 381), and 472 DMRs in promoters (hypomethylation 422 and hypermethylation 50). These annotated DMRs were selected as DMGs when the coverage cutoff values were greater than 25%, 100%, and 34.4% in the promoter, exon, and intron, respectively. Hence, 63 DMGs (13.3%; 62 hypo DMGs n = 62; 1 hyper DMGs) in the promoters, 1,589 (39.3%; 1,449 hypo DMGs; 140 hyper DMGs) in the exons, and 1,673 (13.2%; 1,529 hypo DMGs; 144 hyper DMGs) in the introns were selected for further analysis (Figure 1E).

After screening these DMGs to coverage, we obtained a total of 2,251 DMGs (Supplementary Table S4). We found 12 overlapping genes among promoters, exons, and introns, among which 11 genes were hypo DMGs (GRAMD4, RASGEF1B, ENSSSCG00000031903, ENSSSCG00000032403, ENSSSCG00000033726, ENSSSCG00000047071, ENSSSCG00000036186, EFNA1, ENSSSCG0000005138, EPS8L2, and RAB11FIP3) and CASQ2 was the only hyper DMG (Supplementary Figure S1).

To understand the epigenetic regulation of gene expression, RNA-Seq was performed on the same abdominal fat samples. Illumina paired-end (PE) sequencing yielded 310 million raw reads at an average of 34.5 million reads per sample, which were trimmed by about 2.3% after quality control. On average, over 93.8% of the clean reads were aligned to the pig reference genome, and over 61.3% were unique (Supplementary Table S1). The transcriptomes at different growth stages (10 and 26 weeks of age; Supplementary Figure S2A) were compared to determine the DEGs. The total number of DEGs was 5,768, of which 3,047 were upregulated, and 2,721 were downregulated; 54.8% of genes showed a fourfold difference or greater (Supplementary Figure S2B).

Functional annotation through GO and KEGG pathway enrichment

GO and KEGG pathway enrichment analysis was performed by dividing the combinations into four categories: entire DMG, exon unique DMG, exon and intron common DMG, and intron unique DMG. In the GO analysis, lipid storage and triglyceride biosynthesis were the most representative terms for entire DMGs, followed by skeletal muscle tissue growth (Figure 2A). Sodium ion homeostasis and vitamin transport pathways, including hormone secretion of various substances, were associated with unique DMGs (Figure 2B) while TORC1 signaling and protein citrullination, including NIK per nuclear factor kappa B (NFκB) signaling and lipid storage localization, were important for exon and intron common DMGs (Figure 2C). Positive regulation of protein complex assembly was one of the representative terms identified among intron unique DMGs, but cell cycle and triglyceride and steroid metabolic processes were also important (Figure 2D).

Figure 2.

Functional enrichment analyses of DMGs and methylation genes. (A to D) Treemap of the biological process of GO in DMG by gene region. In order: total DMGs, exon unique DMGs, common (exon and intron overlapped DMGs), and intron unique DMGs. Representative enriched GO terms are indicated by bold letters, and the square sizes are different according to the P-value score of −log based 10. (E) Top 30 of KEGG pathway enrichment analyses for DMGs. The node size and color indicate the gene count associated with the pathway term and the significance level. (F) Circle diameter corresponds to the false discovery rate (FDR) q value. Color intensity represents the normalized enrichment score. Yellow, hypermethylated gene sets; blue, hypomethylated gene sets. Line thickness connecting the gene set nodes represents the degree of gene overlap between the two sets.

Details of the KEGG pathway analysis shown in Figure 2E can be found in Supplementary Table S5. Among the identified results, 30 pathways with a significant value (P < 0.01) were represented (Figure 2E). We identified KEGG pathways directly or indirectly involved in lipid metabolisms, such as ABC transport, FA degradation, insulin resistance, and FA metabolism, and those involved in immune system processes, such as FC gamma R-mediated phagocytosis, B cell receptor signaling pathway, and chemokine signaling pathway, through pathway analysis. Moreover, we found that genes belonging to the exon and intron common DMG were similar to those of total DMGs.

None of the genes in DMRs on the promoter region had statistically significant enrichment results. After enrichment analysis for GO and KEGG, 16 terms were found in GO (BP n = 8, CC n = 7, MF = 1) and one term was found in KEGG when no threshold was specified. Angiogenesis and positive regulation of apoptotic processes and epithelial proliferation were identified as important growth-related pathways in GO while cytokine and other immune response-related pathways were also identified in GO and KEGG enrichments.

GSEA based on the KEGG database was performed using normalized FPKM counts to verify the DMRs and DMGs. Through the analysis, we only found drug metabolism of other enzymes for hypermethylated regions and genes (Figure 2F). In addition, lipolysis and insulin regulatory pathways were identified, similar to the results of the KEGG analysis. However, the most significant pathways with the highest normalized enrichment score (NES) values were those regulating longevity, while the pathway with the most edges was related to the regulation of lipolysis in adipocytes. Thirteen genes (LOC100511937, ADCY2, ADCY1, AKT2, PIK3R1, LOC100515824, INSR, ADCY9, IRS2, ADCY3, ADCY5, PIK3CD, and ADCY8) overlapped between the pathway with the largest node size and with the most edges, while an additional seven genes (SOD2, IGF1, MTOR, PRKAA2, HRAS, RPTOR, and PRKAG2) overlapped between the genes of two different longevity-regulating pathways. These results indicate a relationship between growth regulation and fat metabolism.

Data validation was performed using RNA-Seq to confirm the change in gene expression by DNA methylation, which was epigenetically modified according to growth. The transcriptomes at different growth stages (10 weeks of age and 26 weeks of age) were compared to determine the DEGs. MDS analysis was used to normalize 26,839 expression values and confirmed clustering according to different growth phases (Supplementary Figure S2A). The total number of DEGs was 5,768, of which 3,047 were upregulated, and 2,721 were downregulated (Supplementary Figure S2B and Table S6).

GO analysis revealed downregulated DEGs related to immune responses, such as inflammatory response, response to chemistry, and regulation of cytokine production, as well as those related to cell cycle and DNA recombination. Skeletal and muscle development, as well as lipid metabolism pathways, were most enriched with upregulated DEGs. In addition, we found biological insight related to development and morphogenesis, such as the regulation of cell proliferation, among the overall DEGs (Supplementary Figure S3A–C). Similarly, in the KEGG pathway analysis, we identified pathways involved in the immune response and lipid metabolism associated with DEGs. In addition, peroxisome- and primary immunodeficiency-related pathways were only associated with upregulated and downregulated DEGs, respectively (Supplementary Figure S3D and Table S7).

GSEA results using normalized TMM counts revealed changes in DEGs that were consistent with the results of the KEGG pathway enrichment analysis (Supplementary Figure S3E). The largest subnetwork is connected through the chemokine signaling pathway with negative NES values and the relaxin signaling pathway with positive NES values, indicating effects on cancer, diseases, and immune responses. In addition, longevity and insulin-related pathways were connected through coefficient values and included in the network. However, the lipolysis pathway was separately included in a subnetwork centered on the cGMP-PKG signaling pathway. An additional subnetwork of carbon metabolism centered on glucose, citrate, and pyruvate metabolism also had upregulated genes. At the same time, a subnetwork containing FA metabolism was also differentially expressed.

Cis- and trans-regulation effects in DNA methylation and gene expression

The patterns of DMGs and DEGs in the entire chromosomes were examined to assess changes between DNA methylation and gene expression (Figure 3). DMGs and DEGs were evenly distributed across all chromosomes in the genome, except for the Y chromosome, where DMGs and DEGs were unevenly distributed. The methylation levels ranged from −6 to 6, and the expression levels ranged from −9 to 11. It was confirmed that the gene expression showed a more extensive change pattern compared to methylation. Overlapping DMGs and DEGs were estimated to confirm cis-regulation effects. Of the total 31,907 reference genes, DEGs are 5,761 (18%) and DMGs are 2,251 (7%), indicating that the effect of growth is greater on gene expression. However, the total number of overlapping genes between DMGs and DEGs was 629. The overlapping 629 genes account for 11% in DEGs and 28% in DMGs, which reveals that genes with changes in DNA methylation account for more among genes indicating the cis-regulation effect. Then, these genes were further divided into four categories: (1) hypomethylated DMGs and upregulated DEGs, (2) hypomethylated DMGs and downregulated DEGs, (3) hypermethylated DMGs and upregulated DEGs, and (4) hypermethylated DMGs and downregulated DEGs. Genes in the first two categories accounted for more than 90% of the overlapping genes, wherein methylation is distributed over features of the gene body.

Figure 3.

Distribution of DNA methylation and RNA expression pattern across the genome. (A) The outside most ring indicates the ideogram of the pig genome in a clockwise rotation. (B), (D) Each dot represents an annotated gene. (C), (E) The track contains orange (hyper), blue (hypo), red (up), and navy (down) colors, indicating DMR methylation levels and RNA expression levels, respectively. The white dot on the line indicates the peak of each area. (F) The bar designates the overlapping of DMGs and DEGs. Individually color symbolizes the methylation and expression level relationship between DMGs and DEGs. (G) This text identifies gene features that are methylated in overlapped genes. (H) Venn diagram showing the degree of overlap according to the methylation and expression level of DMGs and DEGs.

We expected that the overlapping genes from the results may play an initial mediating role in the pathways. Pathway analysis of the overlapping genes revealed significant pathways related to immune response, cellular activity, and material transport, consistent with the top 30 DMGs and pathways like FA degradation and glycerolipid metabolism, consistent with the top 30 DEGs (Figure 4A). KEGG pathway analysis on overlapping genes revealed involvement in the FA degradation pathway and the Rap1 signaling pathway that also regulates other pathways (Figure 4B and C) between the DMGs and DEGs.

Figure 4.

Gene changes related to the KEGG pathway in the overlapping genes of DEG and DMG. (A) The combination dot plot was constructed based on significant pathways (FDR < 0.05), and a gene regulation change between DNA methylation and expression category. (B) Genes overlapping between DEGs and DMGs are indicated using an asterisk. Methylation and expression level is presented as log2 difference methylation level and log2 FC values, respectively.

As the ratio of DEGs to DMGs per chromosome was not consistent, we performed a functional enrichment analysis based on the TF database to confirm interactions between methylation and gene expression by trans-effect. A total of 415 TFs were identified among the DEGs and DMGs; 32 source TFs and 273 target TFs were identified by filtering only common TF values through the ChEA and ENCODE databases. Among source TFs, the number of DEGs, DMGs, and overlapping genes of both DEGs and DMGs was 23, 6, and 3, respectively, while that of target TFs was 187, 63, and 23, respectively.

Total TF and target gene network and functions are provided in Supplementary Figure S4 and Table S8. Functional enrichment analysis of TF genes revealed pathways involved in disease and fat metabolism, of which adipogenesis was shown to be the most significant (Figure 5). Pathways related to fat metabolism were ranked at the top and constructed into a subnetwork. PPARG, PPARA, RXRA, SREBF1, NCOR1, and NCOR2 were involved in all pathways on the subnetwork, whereas CEBPB, EBF1, and CEBPA were commonly involved in adipocyte differentiation and adipogenesis, and MED16, MED24, MED25, MED26M, CDK19, and NCOA3 were involved in the transcriptional regulation of adipocytes and the regulation of lipid metabolism. Thus, source TFs were verified to regulate genes related to lipid metabolism by regulating target genes that may not function in each pathway.

Figure. 5.

Subnetworks and interactions of genes related to lipid metabolism. The shape of the node distinguishes the TF from the target gene and term, and the outer line of the node indicates the origin of each gene. The node size is displayed only for TF, indicating the number of genes each TF targets on the network. In the bar graph inside the node, the left and right sides represent methylation and expression, respectively, and the color of the bar represents the level. In addition, the genes bound to the same term are presented as a heatmap. The number in the term indicates the significance level.

Discussion

Molecular and functional changes during mammalian aging are continuous, complex, and difficult to study. Epigenetic changes in the genome, such as DNA methylation, play central roles in regulating gene expression (Jones andTakai, 2001) and mediate cellular and organismal aging (Booth and Brunet, 2016). While mammalian development and cancer have been extensively studied, epigenetic regulation of growth, metabolism, body size, and obesity may also play a role in aging. In this study, we applied an integrated analysis of DMGs and DEGs to reveal differences in DNA methylation and gene expression in the abdominal fat of pigs that were 10 and 26 weeks of age. We showed that total DNA methylation levels were higher at 10 weeks of age than at 26 weeks of age (Figure 1C and E). These results are consistent with the results of previous studies showing genome-wide DNA hypomethylation characteristics, excluding hypermethylation of specific regions such as promoters (Fraga et al., 2007). DNA methylation levels within genes were much higher than those of promoters and those of transcription start and end sites in general. Hypermethylation and hypomethylation of intragenic regions are known to regulate gene expression in the porcine genome (Hellman and Chess, 2007; Laurent et al., 2010). Intragenome hypermethylation typically reduces gene expression but can also increase transcription (Lorincz et al., 2004; Cokus et al., 2008; Ball et al., 2009; Laurent et al., 2010), whereas DNA hypomethylation affects genes involved in cell growth and development, histone remodeling, apoptosis, and cell proliferation. Differential methylation genes in peritoneal fat across ages included genes involved in signaling and immune-related pathways and fat metabolic pathways. This is consistent with the GSEA results in this study that revealed connections between growth, longevity, and fat metabolism.

Although abdominal fat develops in a relatively early stage of growth in pigs compared to other fat, RNA-seq yielded several DEGs across the two ages studied. Although the number of upregulated genes was similar to that of the downregulated genes, the upregulated genes showed higher significance levels and expression changes. Gene regulation in abdominal fat at 26 weeks of age was more dynamic than that at 10 weeks of age. DEG functional analysis enriched muscle growth and fat accumulation pathways, and whole mRNA profiling detected comprehensive molecular networks (Supplementary Figure S3). Moreover, tube development, lipid metabolism, and PPAR signaling pathways were upregulated at 26 weeks of age while immune response, cell activation, and cytokine-related pathways were downregulated.

Adipose tissue consists of connective tissue matrix, blood vessels, nerve tissue, and a large number of adipocytes and non-adipocytes. Non-adipocytes include inflammatory cells (macrophages), immune cells, preadipocytes, and fibroblasts. Abdominal fat has more cells, blood vessels, and innervation than subcutaneous fat, which contains a large number of inflammatory and immune cells, has less ability to differentiate adipocytes, and has a higher proportion of large fat cells (Ibrahim, 2010). Adipose tissue has immune functions such as the synthesis and secretion of TNFα (Hotamisligil et al., 1993), expression of toll-like receptors, and lipopolysaccharide reactivity (Lin et al., 2000). In addition, adipose tissue produces systemic low-grade pro-inflammatory sequelae during obesity (Bluher, 2009; Trayhurn, 2013). Expansion of adipose tissue was followed by macrophage infiltration, the formation of crown-like structures, and increased inflammatory markers in plasma (Wentworth et al., 2010; Klimcakova et al., 2011) including interleukin (IL), leptin, plasminogen activator inhibitors-1, TNFα, and C-reactive protein. These inflammatory markers contain adipokines, which are involved in the differentiation of immune cells, the development of the immune system, and the development of hemopoietic and lymphoid organs. In this study, this gene expression was found to be regulated by DNA methylation in the gene body. For instance, VAV2 is a hypomethylated DMG and a downregulated DEG involved in the development and activation of T cells and B cells, essential for developing an adaptive immune system, whereas the hypomethylated upregulated AKT promotes NFκB activation by linking with the CARMA-1/MALT1/BCL10 complex of T cells (Narayan et al., 2006; Qiao et al., 2008). Selective inhibitors of AKT1 and AKT2 (AKTi-1/2) inhibit the NFκB-dependent TNFα while PI3K can promote NFκB activation. In addition, genes that induce inflammatory markers, such as IL6R, ICAM1, PPP1R3B, CCL19, ITK, and LCK, were expressed more at 10 weeks of age, whereas APOE and PNPLA3 genes related to FAs and lipoproteins were expressed more at 26 weeks of age. Macrophage markers like CD19, CD68, CD80, CD86, CD163, and CCL22 were also upregulated at 10 weeks of age, which can alleviate obesity-related chronic inflammation by increasing macrophage infiltration (Wellen and Hotamisligil, 2003; Burrell et al., 2004).

RNA-Seq and MBD-seq represent the transcriptome and methylome of abdominal adipose tissue, respectively. However, there is a limit to understanding phenotypes with results only at the molecular level. By integrating the results at each molecular level, greater insight can be gained into understanding the epigenetic regulation of gene expression. First, we tried to integrate the two omics data and identify cis-effects between DNA methylation and gene expression through the overlap of DMGs and DEGs (Figure 3). Intriguingly, KEGG enrichment analysis revealed overlapping genes in the FA degradation pathway. Fat deposition is determined by lipogenesis, lipolytic enzymes, and a complex balance between FA transport and FA utilization (Zhao et al., 2010). Among the seven genes involved in the FA degradation pathway, only ALDH7A1 was hypermethylated, whereas ACADVL, ACOX1, CPT1C, ECHS1, ECIT2, and HADHB were hypomethylated. CPT1 regulates the entry of FA into mitochondria (rate-limiting step of FA degradation pathway), where they undergo mitochondrial β-oxidation (Teng et al., 2016). The rate of adipogenesis in adipose tissue decreases during aging (Gellhorn and Benjamin, 1965), whereas fat deposition increases with age (McMeekan, 1940). Fat deposition increases as the rate of fat generation decreases, which may be due to the balance between synthesis and degradation. In humans, the ABC transport pathway is known to be involved in the transport of lipids and lipid-related compounds (Tarling et al., 2013). The present study suggests that the hypomethylated genes could contribute to ABC transport, indicating that fat degradation and transport functions may vary across age.

PPAR signaling is closely related to FA and sterol metabolism as well as adipogenesis and differentiation in adipose tissue (Farmer, 2005). We identified two upregulated genes associated with the PPAR signaling pathway. A lipid droplet coat protein, perilipin 1, has conserved N-terminal sequences (PAT domain) and an affinity for intracellular neutral lipid storage droplets (Kimmel et al., 2010). The findings regarding these genes in this study are consistent with those of studies on adipocytes in humans (Grahn et al., 2013) and mice (Sun et al., 2013). A previous study has shown that triglycerides can be inhibited from fat degradation to increase their content in adipocytes (Li et al., 2018). Similarly, FA binding protein 3 (which was higher at 26 weeks of age), which is involved in the absorption, transport, and metabolism of FA and is mainly expressed in the heart and skeletal muscle, has an important effect on intramuscular fat content (Li et al., 2007; Zhao et al., 2009; Ovilo et al., 2002).

Next, we estimated the trans-effect between DMGs and DEGs using the TF database. Pathways involved in fat metabolism on the TF subnetwork were all directly or indirectly linked to five target genes (RXRA, SREBF1, PPARA, NCOR1, andNCOR2) and one TF (PPARG). PPARG, a member of the nuclear receptor superfamily necessary and sufficient for adipogenesis, is a key regulator of adipocyte differentiation (Rosen et al., 2002). CEBP is an important transcriptional regulator of lipogenic lipid production. In addition, CEBPA is a trans-activator of PPARG, which works together to promote adipogenesis (Wu et al., 1999). PPARG forms a protein complex that reduces the transcription of the target gene (Yu et al., 2005). NCOR1 and NCOR2 encode proteins that inhibit the transcription of thyroid hormone and retinoic acid receptors through histone deacetylation (Davie and Chadee, 1998). In this study, both NCOR1 and NCOR2 were slightly downregulated, which could enhance adipocyte differentiation, in part through higher PPARG transcriptional activity (Yu et al., 2005). In addition to their role in cholesterol synthesis, SREBF1 and SREBF2 can induce mRNA expression of genes involved in FA synthesis (Horton et al., 2003), and SREBF1 can upregulate downstream fat-inducing genes such as ACACA, FASN, and AGPAT6(Kadegowda et al., 2009). Taken together, we observed epigenetic changes in abdominal fat and confirmed that these changes are involved in fat metabolism and immune responses. Overall, our study identified the genes that showed epigenetic changes with growth that appear in lipolysis and production, suggesting that the changes in fat deposits with growth constitute the difference between breakdown and production. In addition, the genes showing epigenetic changes can be additionally applied as molecular biomarkers.

Conclusions

Fat deposition is an important factor that determines the economic traits of pigs. Although MOI analysis provided an insufficient understanding of the phenotype of localization at the single molecule level, it provided a better understanding of the control of epigenetic changes. Inflammatory and immune responses in abdominal fat are reactions that can be partially explained by aging-related differences in tissue characteristics and responses to factors related to obesity such as insulin resistance. The degradation, synthesis, and circulation of fat are key factors that can explain the fat accumulation process. Our research helps to understand the development of abdominal fat during aging and provides a basis for the future exploration of molecular epigenetic mechanisms underlying fat metabolism and growth.

Glossary

Abbreviations

DEGs

differentially expressed genes

DMRs

differentially methylated regions

ES

enrichment score

FA

fatty acid

FPKM

fragment per kilobase of transcript per million mapped reads

GO

Gene Ontology

GSEA

gene set enrichment analysis

KEGG

Kyoto Encyclopedia of Genes and Genomes

MBD-seq

methyl-CpG-binding domain sequencing

MDS

Multidimensional scaling

MOI

multi-omics integration

NFκB

Nuclear factor kappa B

RNA-Seq

RNA sequencing

TES

transcription end site

TF

transcription factor

TSS

transcription start site

TMM

trimmed mean of M-value

TNFα

tumor necrosis factor α

Conflict of Interest Statement

The authors declare that they have no conflict of interest.