Finding aging-related markers in fat tissue using computer analysis: key genes ELN, MXD1, and FGF21

A total of 49 DEARGs were identified by comparing DEGs in adipose tissue, ARGs, and human genes associated with aging (Figure 1a). The expression patterns of these 49 DEARGs in 12-month-old and 2-month-old mice were visualized using a heatmap (Figure 1b). Functional enrichment analysis was performed on these DEARGs to explore their biological roles (Supplementary Table S1).

For biological processes (BP), significant overlaps were found in glucose homeostasis, response to peptide hormones, thermal stimulus response, and peptide response (Figure 1c). In terms of molecular functions (MF), notable overlaps were observed in beta-catenin binding, DNA-chromatin binding, binding to unfolded protein domains, and protein chaperone folding (Figure 1d). Cellular components (CC) revealed overlap in the RNA polymerase II transcription repressor complex, euchromatin, TOR complex, and transcription regulatory complexes (Figure 1e). Disease Ontology (DO) analysis indicated overlaps with endometrial cancer, endometrioid carcinoma, hyperalimentation, and obesity (Figure 1f). KEGG pathway analysis suggested that pathways such as the longevity-regulating pathway, FoxO signaling pathway, JAK-STAT signaling pathway, Th17 cell differentiation, growth hormone synthesis/secretion/action, adiponectin signaling pathway, and AMPK signaling pathway may be involved in adipocyte changes during aging (Figure 1g). DisGeNET enrichment analysis indicated overlaps with sleep apnea, cachexia, diabetic peripheral neuropathy, diabetic neuropathies, estrogen receptor-negative breast cancer, iron overload, and sarcopenia (Figure 1h).

To identify hub DEARGs involved in adipose tissue aging, the 49 DEARGs were analyzed using the STRING database to construct a protein-protein interaction (PPI) network with 49 nodes and 295 edges (Figure 2a, Supplementary Table S2). Feature selection using LASSO regression identified five hub DEARGs—PCK1, ELN, MXD1, STAT3, and FGF21—which showed significant non-zero regression coefficients with an optimal lambda value of 0.095 (Figures 2b, c).

Further co-expression analysis was performed using the GeneMANIA database, revealing an intricate PPI network with various interaction types, including 1.04% common protein domains, 1.57% colocalization, 1.81% genetic interactions, 2.07% pathways, 8.09% additional elements, 17.96% co-expression, 22.45% anticipated interactions, and 45.00% direct interactions (Figure 2d). These genes were found to play key roles in energy metabolism, extracellular structure organization, fibroblast growth factor responses, short-chain fatty acid metabolism, protein oxidation, growth factor binding, and triglyceride metabolism, suggesting their significant involvement in the aging of adipose tissue.

To validate the expression of the identified hub DEARGs—PCK1, ELN, MXD1, STAT3, and FGF21—in human adipose tissue, data from the Integrating Platform of ADEIP were analyzed. The results showed that ELN, MXD1, and FGF21 exhibited upregulated expression in the adipose tissue of older individuals (Figure 3a).

In vitro, differentiated adipocytes were confirmed by Oil Red O staining (Figure 3d). To establish an in vitro model of cellular aging, adipocytes were treated with D-galactose (Figure 3e). H&E staining (Figure 3h) and immunofluorescence (Figure 3i) of white adipose tissue (WAT) from young and old mice further supported these findings.

The expression of the hub DEARGs was also validated by qPCR in the D-galactose-induced aging model and in WAT isolated from 2-month-old and 12-month-old mice. The mRNA levels of ELN, MXD1, and FGF21 were significantly higher in older adipocytes compared to younger controls (Figure 3b). This validation confirmed that ELN, MXD1, and FGF21 are upregulated during adipose tissue aging, supporting their potential as biomarkers for aging in adipose tissue (Figure 3c).

To explore aging-associated genes in adipose tissue, we sourced 4,314 ARGs from the GeneCards database, with each gene having a relevance score greater than 5 (Supplementary Table S3). Additionally, we obtained a set of 307 genes specifically associated with human aging from the Human Aging Genomic Resources (HAGR) database (https://genomics.senescence.info/↗; see (Supplementary Table S4).

We retrieved the CRA015560 mRNA expression profiling dataset from the Genome Sequence Archive (GSA) (https://ngdc.cncb.ac.cn/gsa/↗) to analyze differentially expressed genes (DEGs) between young and aged mice. Comparative analysis was performed using the limma R package to identify significant gene expression differences between the youthful and aged cohort. Genes were considered differentially expressed if they met the criteria of |fold-change (FC)| > 1.5 and p-value < 0.05.

To investigate the biological significance of DEGs and differentially expressed age-related genes (DEARGs) in adipose tissue aging, functional enrichment analysis was conducted. Gene annotations were sourced from the Human Phenotype Ontology (HPO), Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) within the Molecular Signatures Database (MSigDB) and DAVID databases. Enrichment analyses for DEGs were performed using the R package ClusterProfiler alongside DAVID. Functional categories and pathways with a p-value < 0.05 were considered statistically significant.

To identify key biomarkers within DEARGs, we performed a PPI analysis using the STRING database, applying a confidence score threshold of >0.4. The interaction network was visualized in Cytoscape 3.8.1. We used the MCODE plugin within Cytoscape to extract major functional modules, with selection parameters set to K-Core = 2, Cut Grade = 2, Maximum Depth = 100, and Cut Node Score = 0.3. Each gene node was evaluated using the Maximum Clique Centrality (MCC) score within the cytoHubba plugin of Cytoscape, and the top 10 genes with the highest MCC scores were identified as key nodes. Further refinement of important biomarkers was achieved through Cox regression analysis with the least absolute shrinkage and selection operator (LASSO) for gene selection. The optimal penalty parameter was chosen using triple cross-validation to minimize cross-validation error, as implemented in the 'glmnet' R package. Additionally, we utilized GeneMANIA (http://genemania.org/↗; last accessed on May 1, 2023) to validate PPI networks of key biomarkers.

We employed the Age-Dependent Expression and Immune Profiles (ADEIP) database to validate hub biomarkers. ADEIP provides gene expression and cell proportion data across different age groups and genders (http://gb.whu.edu.cn/ADEIP/↗). Using this resource, we examined the expression profiles of identified biomarkers across age groups to confirm their relevance to aging.

Animal experiments were conducted in accordance with the ethical guidelines set by the Chongqing University Animal Experiment Ethics Committee. C57BL/6 mice were obtained from Gempharmatech Co., Ltd (Guangzhou, China). Mice were housed in a temperature-controlled facility with a 12-hour light-dark cycle and provided ad libitum access to food and water. Male mice of specific ages were used for all experiments, with at least three animals per group in each experiment. Mice were anesthetized by continuous inhalation of isoflurane (3%). Euthanasia was performed by exsanguination under anesthesia, followed by cervical dislocation as a secondary euthanasia procedure.

The 3T3-L1 preadipocyte cell line was sourced from the Stem Cell Bank of the Chinese Academy of Sciences. Cells were cultured in Dulbecco's Modified Eagle Medium (DMEM, Vivacell) supplemented with 10% newborn calf serum (CS, Adamas Life, C8231). Cells were grown until confluence, at which point adipogenesis was induced. The differentiation process began with pre-culturing cells in basic medium for 2 days, followed by treatment with a differentiation medium containing 10 µg/mL insulin, 0.25 µM dexamethasone, and 500 µM IBMX for 2 days. Subsequently, cells were cultured in a growth medium containing only insulin for an additional 2 days. Fully differentiated adipocytes were cultured in DMEM supplemented with 10% fetal bovine serum (FBS, Vivacell), with the medium changed every 2 days until adipogenesis was complete. Differentiated adipocytes were identified by their characteristic morphology of multivesicular lipid droplets, which stained red with Oil Red O solution. To induce cellular senescence, adipocytes were exposed to 40 g/L D-(+)-galactose (Beyotime, ST1218) for 48 hours.

SA-β-gal activity was assessed using a senescence β-galactosidase staining kit (C0602, Beyotime, Shanghai, China), following the manufacturer's instructions. The percentage of SA-β-gal-positive cells was quantified to assess the extent of cellular senescence.

Oil Red O staining was performed using a modified Oil Red O staining kit (Beyotime, C0158), according to the manufacturer's instructions. After staining, cells were subjected to microscopic analysis to assess lipid accumulation.

Total RNA was extracted from mouse white adipose tissue (WAT) and cultured cells. For tissue RNA extraction, approximately 50 mg of WAT was placed in tissue grinding tubes with zirconium beads and 500 μL TRIzol reagent (Invitrogen, 15596026CN). Tissues were homogenized using a Servicebio™ homogenizer with two 30-second pulses at 6 m/s, separated by a 15-second pause. For cell RNA isolation, the NcmSpin Cell/Tissue Total RNA Kit (NCM Biotech, M5105) was used. RNA concentration and quality were assessed with a Thermo Scientific™ NanoDrop™ One spectrophotometer. One microgram of RNA was reverse transcribed into complementary DNA (cDNA) using the ABScript Neo RT Master Mix (ABclonal, RK20433), along with the gDNA Remover Kit. Quantitative PCR (qPCR) was performed using a Bio-Rad™ CFX96 Touch Real-Time PCR Detection System with 2X Universal SYBR Green Fast qPCR Mix (ABclonal, RK21203) and gene-specific primers (). The ΔΔ-CT method was applied to analyze gene expression, with Actin as the internal control. 1

Slides were prepared according to a previous report (44).

Hematoxylin and Eosin (H&E) Staining: Tissue sections were rehydrated through a graded alcohol series (xylene for 3 minutes, 100% alcohol for 1 minute twice, 95% alcohol for 1 minute twice, and water for 1 minute), followed by staining with H&E. The slides were dehydrated in reverse order, mounted with Cytoseal 60, and analyzed using an Olympus VS200 inverted microscope.

Immunofluorescence Staining: Slides were rehydrated according to the following protocol: xylene (20 min x 3 times); 100% reagent alcohol (5 min x 2 times); 95% reagent alcohol (5 min x 1 time); 85% reagent alcohol (5 min x 1 time); 75% reagent alcohol (5 min x 1 time); 50% reagent alcohol (5 min x 1 time); PBS (5 min x 1 time). Sodium citrate solution (BOSTER, AR0024) was utilized for antigen retrieval. The slides were rinsed three times with 1x PBS. Slides were incubated with the primary antibody in 1X TBS containing 5% goat serum at 4°C overnight after being blocked for 30 minutes with 5% goat serum in 1X TBS. Primary antibodies included Rabbit anti-Perilipin A (1/200; HUABIO ET1703-38) and Mouse anti-Ucp1 (1/200; Santa Cruz sc-518024). Invitrogen secondary antibodies were diluted to 1/500 and then incubated for 1 hour at room temperature. The following secondary antibodies were used: 555 Goat anti-Rabbit or 488 Goat anti-Mouse. The slides were cleaned and dyed using DAPI (Servicebio G1012) for 10 minutes, followed by application of mounting solution (Solarbio S2100). Fluorescent images were recorded using the Olympus VS200 inverted microscopy system.

Statistical significance was assessed using a two-tailed Student's t-test to compare two groups, and one-way ANOVA was used for multiple group comparisons. Data are expressed as means ± SEM. P-values less than 0.05 were considered statistically significant. For microscopy images, three randomly selected fields from at least three mice per cohort were analyzed using NIH Fiji ImageJ software. H&E, Oil Red O, SA-β-gal, and immunofluorescence images were captured from a minimum of three to four replicates per group. RNA-seq data were analyzed for statistical significance, heat map generation, and KEGG pathway analysis using RStudio. Graphical representation and statistical analysis were performed using GraphPad Prism 7-9, and initial data collection was performed using Microsoft Excel.