Transcriptomic and epigenetic alterations in aging-related genes in the medaka liver due to ancestral BPA exposure

Abstract

Environmental endocrine disruptors (EEDs) can accelerate the onset of age-associated liver dysfunction, yet the underlying molecular mechanisms remain poorly understood. Whether unexposed offspring experience differential aging of the liver due to ancestral environmental exposure is currently unknown. Here, we investigated the molecular underpinnings of aging in female medaka livers due to ancestral exposure to bisphenol A (BPA). Exposure occurred at the first (F0) generation, and the aging liver phenotype was examined at the third (F2) generation. Controls did not receive BPA while experimental fish were exposed to 10 μg/L BPA from 8 hours post-fertilization to 15 days post-fertilization, spanning their embryonic and perinatal development during which the liver forms and fully differentiates. The liver developed hepatic steatosis, characterized by increased expression of aging gene signatures commonly found in mice and humans. Transcriptome analysis revealed significant downregulation of 13 mitochondrial oxidative phosphorylation genes, indicating impaired mitochondrial transcriptional activity. A total of 189 aging-associated genes were identified as common between medaka and humans, which are known to regulate mitochondrial metabolism, chromatin remodeling, and inflammation. Functional enrichment linked these changes to oxidative stress, DNA damage, apoptosis, and impaired regenerative capacity. Integrated methylome–transcriptome analysis revealed hypermethylation of gene body CpG islands in core aging regulators, correlating with transcriptional repression and disruption of FoxO, PI3K-Akt, AMPK, and longevity pathways. These findings demonstrate that ancestral BPA exposure induces coordinated mitochondrial, transcriptomic, and epigenetic reprogramming of conserved aging networks, predisposing descendants to accelerated hepatic aging associated with increased severity of sex-biased NAFLD phenotype.

Bisphenol A (BPA), a ubiquitous environmental endocrine disruptor (EED), is known for its ability to bind with estrogenic receptors and to alter epigenetic profiles, leading to various adverse health effects [1, 2]. Accumulating evidence indicates that BPA exposure alters DNA methylation profiles and disrupts gene expression patterns in the liver across generations, particularly when exposure occurs during a sensitive developmental window [3, 4]. These epigenetic changes can persist even long after the exposure has ceased, leading to transgenerational reprogramming of gene regulation, including alterations in genes critical for metabolic homeostasis and proper liver function. In murine models, BPA exposure has been shown to impair genomic imprinting, suggesting that early exposure may lead to broader epigenetic changes in liver function, critical for metabolic homeostasis [5]. In rainbow trout, exposure of eggs to BPA resulted in extensive liver reprogramming across generations [4], suggesting long-lasting effects of EEDs in the developing liver. BPA induced DNA methylation in lipid metabolism genes associated with non-alcoholic fatty liver disease (NAFLD, also called MASLD). These studies underscore BPA as a potential inducer of heritable epigenetic changes that elevate the risk of metabolic disorders, including NAFLD, in unexposed future generations. Consumption of high-fat diet has been found to accelerate aging by promoting lipid deposition in the liver of aging mice, with increased insulin secretion playing a significant role in driving fatty acid synthesis in hepatocytes [6]. However, significant gaps remain in understanding the molecular mechanisms of EED-induced liver aging, particularly how epigenetic and transcriptomic changes contribute to age-related functional decline and disease susceptibility. Our study aims to shed light on these complex interactions in liver aging. As society undergoes demographic transitions, there is a growing need to understand how early-life EED exposure contributes to age-associated diseases such as NAFLD. Age is a critical factor in the prevalence and severity of NAFLD. Epidemiological studies indicate that the burden of the disease escalates with aging [7, 8]. Recent studies have illuminated the role of epigenetic factors in both aging and the pathogenesis of NAFLD. Age-related hypermethylation of specific liver genes may impair metabolic function and promote NAFLD [9, 10]. These age-dependent epigenetic modifications are thought to alter transcriptional profiles linked to the progression of liver disease.

Aging is a critical driver of liver homeostasis and disease susceptibility [11]. Oxidative stress and chronic inflammation, which are common features of both aging and NAFLD, can promote hepatocyte senescence and contribute to declining liver function [12]. High-fat diets seem to worsen NAFLD through the induction of oxidative stress, leading to the acceleration of liver aging [13]. The complexity of this NAFLD and aging association is underscored by the interplay of various physiological, metabolic, and hormonal factors. These processes are especially pronounced in postmenopausal women, who exhibit increased visceral adiposity and metabolic imbalance due to declining estrogen levels [14]. The transition from NAFLD to more severe disease states can predispose individuals to complications affecting overall health and quality of life, speeding up the onset of age-related morbidity and mortality. Epigenetic drift and transcriptomic remodeling associated with aging can interact with prior environmental insults, shaping the risk trajectory for metabolic liver diseases such as NAFLD [15, 16]. We recently demonstrated that early exposure to BPA during germ cell development caused NAFLD in the grandchildren generation, with more severity in the female offspring [17-19]. The interplay between epigenetic changes induced by ancestral BPA exposure and the altered expression of age-related genes in the liver of unexposed descendants with transgenerational NAFLD remains poorly understood. To address this knowledge gap, we investigated epigenetic and transcriptomic alterations in mammalian and human aging-related genes in the liver of female medaka (Oryzias latipes) from the BPA lineage, which developed NAFLD phenotype due to ancestral BPA exposure. Our findings aim to examine molecular dysregulation of aging-related genes in the liver of female from BPA lineage.

Results

NAFLD phenotypes in unexposed F2 female descendants of the BPA lineage

Hepatocytes in control-lineage females exhibited normal liver morphology, characterized by basophilic cytoplasm, minimal vacuolization, and well-organized hepatocytes separated by sinusoidal spaces (Fig. 1A). In contrast, livers from females in the BPA lineage showed clear diseased phenotypes, including hepatocyte hypertrophy, architectural disorganization, cytoplasmic depletion, nuclear displacement, and extensive cytoplasmic vacuolization indicative of hepatosteatosis (Fig. 1A). Notably, 100% of BPA-lineage females developed macrovesicular steatosis.

Figure 1.

Ancestral BPA exposure induced liver histopathological alterations and mitochondrial gene downregulation in F2 livers. (A) A representative histological section of hematoxylin stained liver tissue from control (top) and BPA-lineage (bottom) F2 generation medaka (scale bar 25 µm). Control liver displayed normal hepatocyte morphology, while BPA-lineage liver showed signs of vacuolization (black arrow), and cellular disorganization, indicating hepatic damage (arrows). (B) RNA-seq sample replicates clustered by treatment group (BPA vs control), demonstrating consistent within-lineage grouping. (C) KEGG pathway enrichment analysis of dysregulated genes from the BPA lineage revealed broad functional perturbations across multiple categories, including metabolism, genetic and environmental information processing, cellular processes, organismal systems, and human disease pathways. (D) Heatmap showing the expression profiles of mitochondrial oxidative phosphorylation genes in F2 livers. Genes include components of mitochondrial respiratory complexes I (nd1, nd2, nd5), III (cytb), IV (cox1-3), and V (atp6, atp8). Ancestral BPA exposure led to a widespread downregulation of mitochondrial genes in the BPA lineage, consistent with mitochondrial dysfunction. Expression values were normalized; red indicates higher expression, and blue indicates lower expression. (E) MitoCarta3.0-based pathway analysis shows coordinated dysregulation of mitochondrial metabolic pathways in F2 livers from the BPA lineage, including the TCA cycle, NAD metabolism, gluconeogenesis, carnitine shuttle, vitamin D metabolism, and lipid biosynthesis (cholesterol, bile acid, and steroid synthesis), indicating impaired energy metabolism and features consistent with accelerated hepatic aging.

Global dysregulation in transcriptomic profile in the liver of female from BPA lineage

To assess how ancestral BPA exposure influences transcriptional changes in the liver, we first performed RNA-seq analysis of the livers of BPA lineage and control fish. The sample replicates were clustered according to BPA and control lineage group (Fig. 1B). This analysis identified a total of 5187 differentially expressed genes (DEGs) in the BPA lineage compared to the control lineage. A comprehensive KEGG pathway enrichment analysis of dysregulated genes in the liver of female from BPA lineage revealed multifaceted functional disruptions including metabolism, genetic and environmental information processing, cellular processes, organismal systems, and human diseases (Fig. 1C) (Table S1) [20]. Among metabolic pathways, oxidative phosphorylation was most prominently affected, with 55 associated genes, followed by cysteine and methionine metabolism (19 genes), cholesterol metabolism (30 genes), PPAR signaling (31 genes), and fat digestion and absorption (17 genes). Notably, the non-alcoholic fatty liver disease (NAFLD) pathway exhibited the highest gene count (62 genes) among human disease categories. Significant enrichment was also observed in the AMPK signaling (48 genes), mTOR signaling (50 genes), and HIF-1 signaling (39 genes) pathways, implicating nutrient sensing, stress response, and cellular adaptation mechanisms in the aging liver. Pathways involved in genetic information processing, such as ubiquitin-mediated proteolysis (41 genes), indicate increased protein turnover and quality control demands. Cellular process pathways including lysosome (53 genes), peroxisome (32 genes), adherens junction (33 genes), and ferroptosis (15 genes) heightened demands for autophagy, redox regulation, and membrane integrity as the liver ages. Longevity regulating pathway (32 genes) and insulin resistance (44 genes) were also substantially represented in the liver of female from BPA lineage, reflecting the systemic metabolic remodeling due to ancestral BPA exposure.

In the livers of females from the BPA lineage, a wide range of biological processes involving catabolism, metabolism, localization, organization, and transport at both the cellular and organellar levels were significantly altered (Table S2) [20]. Key catabolic pathways, including small-molecule, organic-substance, and organonitrogen-compound breakdown, were enriched, suggesting a greater demand for degrading and recycling cellular components [20]. At the same time, biosynthetic and metabolic processes such as organonitrogen-compound synthesis and organic, oxoacid, and carboxylic acid metabolism were also enriched, pointing to adaptive metabolic remodeling needed to maintain energy balance and redox homeostasis. Under the molecular function category, binding activities include RNA binding, anion, nucleotide, identical protein, carbohydrate derivative, cytoskeletal, cadherin, protein-containing complex, ribonucleotide, purine, adenyl, and vitamin binding highlight the broad decline in molecular interactions crucial for maintaining metabolic, genetic, and structural homeostasis (Table S3) [20]. There was also evidence for reduced efficiency in catalytic activities (such as hydrolase and kinase activity) and key regulatory functions (GTPase activator, enzyme regulator, transcription factor binding).

Dysregulation of mitochondrial pathways in the liver of female from the BPA lineage

During our initial transcriptomic analysis of female BPA lineage livers, we observed that all thirteen protein-coding mitochondrial genes, including NADH dehydrogenase subunits, ATP synthase subunits (atp6, atp8), cytochrome c oxidase subunits (cox1-3), and cytochrome b (cytb), were consistently downregulated (Fig. 1D). This unexpected and uniform pattern prompted a deeper investigation into mitochondrial pathways and related gene sets.

Expanding our analysis with the MitoCarta3.0 [21] database of mitochondrial proteins and pathways, we identified widespread dysregulation not only within mitochondrial genes but also across key nuclear-encoded mitochondrial pathways (Fig. 1E). In particular, several pathways with established links to aging and metabolic regulation showed pronounced dysregulation. These include the TCA cycle, NAD biosynthesis and metabolism, gluconeogenesis, carnitine shuttle, and vitamin D metabolism. Additionally, pathways involved in lipid processing such as cholesterol, bile acid, and steroid synthesis also showed reduced gene expression. The uniform suppression of mitochondrial gene expression and dysregulation of metabolic pathways highlights a possible mechanism by which ancestral BPA exposure may accelerate hepatic aging and impair energy metabolism.

Ancestral BPA exposure alters the expression of conserved mammalian aging-associated genes

To assess how ancestral BPA exposure influences aging-related transcriptional changes in the female liver, we compared total DEGs (5187) with a curated database of aging-related genes from mammalian and medaka datasets, identifying 503 medaka aging-associated orthologs corresponding to 611 mammalian aging genes. Among 503 aging genes in medaka, 189 DEGs overlapped with RNA seq gene sets of BPA lineage fish (Fig. 2A).

Figure 2.

Identification and characterization of ancestral BPA exposure-induced aging-related genes in medaka and mammals. (A) A Venn diagram shows overlap between differentially expressed genes (DEGs) from the BPA-lineage of female medaka (Medaka-RNA Seq, red circle), identified medaka aging-related genes (blue circle), mammalian aging-related genes (green circle), and common orthologs of 503 genes in medaka (purple circle). A total of 189 genes overlapped between DEGs and aging-related genes in medaka. (B) Expression patterns of mammalian aging genes were compared with RNA seq of BPA lineage liver breakdown of the 189 overlapping genes into three categories based on expression pattern: mutually upregulated genes (Mutual Up, pink; 58 genes, 78.26%), mutually downregulated genes (Mutual Down, blue; 29 genes, 15.34%), and BPA-specific genes (gray; 101 genes, 53.43%). (C) Heatmap of common upregulated, and (D) downregulated conserved aging genes between BPA lineage livers and mammals.

We next examined the directionality of these differentially expressed aging-related genes. Among the 189 DEGs, 58 genes were mutually upregulated, while 29 were mutually downregulated in the BPA lineage livers and human aging dataset (Fig. 2A). The heatmaps further show the expression profiles of genes that are mutually upregulated (Fig. 2B) or downregulated (Fig. 2C) in BPA lineage livers. Biological replicates within each sample group exhibit similar expression patterns, demonstrating strong directional concordance. This consistency supports our earlier observation that among the 189 DEGs, subsets of genes are consistently upregulated or downregulated in both the medaka and mammalian aging datasets (Fig. 2A), highlighting shared transcriptomic signatures related to aging in response to ancestral BPA exposure. Several genes known to regulate mitochondrial metabolism (ppargc1a, got2, ndufs2), chromatin remodeling (sox11, phf14) were significantly downregulated, whereas genes linked to inflammation (psmb8, serpine1) were significantly upregulated altered in the liver of female from BPA lineage, suggesting systemic transcriptional dysregulation of aging-related hepatic functions following ancestral BPA exposure.

Cluster analysis identified a subset of genes from slc6a6, pdlim1, cebpb, rela, arpc1b, and snap23 that exhibited coordinated expression patterns in the liver of female from BPA lineage (Fig. 2B, black dashed rectangle). Notably, slc6a6, which encodes a taurine transporter, was consistently upregulated and is essential for hepatic cell volume regulation, cytoprotection, and immune modulation. The transcription factor cebpb showed increased expression and is known to drive hepatocyte proliferation [22], metabolic recovery, and immune response. rela (NF-κB p65) was also elevated, indicating activation of pathways involved in apoptosis inhibition, liver regeneration, and fibrosis regulation. Cytoskeleton-related genes, including arpc1b and snap23, exhibited coordinated expression, reflecting their roles in cell migration, actin dynamics, and hepatic stellate cell motility. Notably, pdlim1 was highly expressed, consistent with its function in hepatic stellate cell activation and fibrotic progression [17, 23]. In the heatmap of downregulated genes (Fig. 2C, black dashed rectangle), mdh, hacd1, pde4d, pgk1, and sucla2 were clustered together based on their expression pattern. The enzyme mdh (malate dehydrogenase) participates in the Krebs cycle and is essential for energy metabolism in hepatocytes [24]. hacd1and sucla2 are associated with membrane integrity and redox homeostasis, respectively [25, 26].

Conserved orthologous human aging genes are linked to DNA damage response, oxidative stress, apoptotic process, and immune system regulation

To identify evolutionarily conserved aging-associated target pathways altered by ancestral BPA exposure, we mapped 265 medaka aging genes to their 307 human orthologs (Fig. 3A). Among these, 88 aging genes were differentially expressed in the livers of BPA lineage fish, suggesting persistent transcriptomic disruption of conserved aging programs. Hierarchical clustering of these genes revealed two distinct groups: 25 genes were significantly upregulated (Log₂FC > 1, FDR < 0.05) and 24 were significantly downregulated (Log₂FC < −1, FDR < 0.05) in the BPA-lineage compared to the control lineage fish (Fig. 3B and 3C). Upregulated genes included cytokines, inflammatory, and immune response mediators (serpine1, tnf, ccn2 ptk2b, mif, c1qa), transcription factor and gene expression regulators (esr1, myc, foxm1), DNA repair (brca2, aptx, and recql4), fibrosis (serpine1, ccn2), pro-oncogenic transcription factors (myc, sncg, bub1b). Transcription factors and co-regulators included nfe2l1, sp1, sin3a, stat5a, nfe2l2, fos, nr3c1, foxo3, and ppargc1a. These genes collectively regulate antioxidant responses (nfe2l1, nfe2l2), stress-induced signaling (fos, stat5a), nuclear receptor-mediated transcription (nr3c1), metabolic gene activation (ppargc1a), and chromatin remodeling (sin3a, sp1). Several key regulators of metabolic and mitochondrial function (ppargc1a, mt1f, prkca, igf1, foxo3, nfe2l1) were significantly downregulated, suggesting impaired energy homeostasis in the liver caused by ancestral BPA exposure. Cell cycle, apoptosis, DNA damage response (mdm2 and sun1), growth factors, hormones, aging modulators (igf1, igf2, kl), and oxidative stress defense (cat, mt1f, and mt-co1) associated genes showed significantly reduced expression.

Figure 3.

Transcriptomic alterations in aging-related genes in liver of BPA lineage medaka and humans. (A) A Venn diagram showing orthologous aging genes between humans (309) and medaka (265), with 88 aging genes identified in the liver of BPA lineage medaka using RNA-seq. (B) A heatmap of 25 significantly upregulated aging-related genes (Log₂FC > 1, FDR < 0.05). (C) A heatmap of 24 significantly downregulated aging-related genes (Log₂FC > 1, FDR < 0.05). (D-F) Gene ontology (GO) enrichment analysis of differentially expressed aging genes: (D) biological process (GO-BP), (E) molecular function (GO-MF), and (F) cellular component (GO-CC). All significant terms (FDR < 0.05) showed in purple and pink color circle. Circle sizes represent the number of genes in each term, and color intensity corresponds to significance levels (purple color: more significant, pink color: less significant).

Gene ontology (GO) enrichment analysis of these 88 differentially expressed genes indicated significant overrepresentation in key biological processes (GO-BP), such as regulation of the MAPK cascade, response to stress, DNA damage response, oxidative stress, apoptotic process, and immune system regulation (Fig. 3D). Molecular function (GO-MF) terms included transcription factor binding, double-stranded DNA binding, kinase activity, chromatin binding, and positive regulation of transcription (Fig. 3E). Cellular component (GO-CC) terms were enriched for membrane-enclosed lumen, perinuclear region of cytoplasm, and nuclear chromosome part (Fig. 3F). Together, these data suggest that ancestral BPA exposure alters the hepatic expression of aging-related genes in the liver, affecting cellular stress responses, transcriptional regulation, and chromatin-associated processes, potentially contributing to liver pathophysiology.

Ingenuity pathway analysis revealed dysregulated signaling associated with aging and oxidative stress

To investigate the molecular signatures associated with liver aging, we performed Ingenuity Pathway Analysis (IPA) on differentially expressed genes. Among total transcriptomic signatures, 88 aging genes common in medaka and humans were used to examine their biological roles on IPA. The graphical summary revealed distinct regulatory patterns associated with aging-related functions (Fig. S1 [20]). Upregulated genes (orange) were linked to enhanced apoptosis, immune cell adhesion, transmigration, accumulation of reactive oxygen species, and reduced cell viability. Conversely, downregulated genes (blue) were associated with decreased transmembrane potential of mitochondria, estrogen receptor signaling, erythropoietin signaling, cell viability of connective tissue cells, and fibroblasts. Central hub genes such as agt, fn1, and il15 appear to integrate multiple functional pathways, indicating their potential role as key mediators in age-related tissue disorders and systemic physiological decline.

Canonical pathway analysis identified several significantly enriched pathways (Fig. 4A). As per IPA prediction based on Ingenuity knowledgebase, Estrogen Receptor Signaling (Activation z-score-2.5, P-value 1.08E-14), Docosahexaenoic Acid (DHA) signaling (Activation z-score-2.65, P-value 3.23E-07), RAR activation (Activation z-score-2.65, P-value 9.83E-07), and Sirtuin signaling (Activation z-score −2.65, P-value 1.02E-06) were predicted to be inhibited, suggesting impaired metabolic and stress-response processes caused by ancestral BPA exposure (Fig. 4A). In contrast, NAFLD signaling (Activation z-score 1.13, P-value 1.50E-07), Acute Phase Response (Activation z-score 1, P-value 1.20E-06), and LPS/IL-1-mediated inhibition of RXR function (Activation z-score 1, P-value 1.98E-03) were predicted to be activated, indicating widespread perturbation in metabolic and inflammatory responses (Fig. 4A). Upstream regulator analysis predicted key transcriptional and signaling modulators driving the observed gene expression changes (Fig. 4B). The upstream regulators, mainly klf3, il15, and notch1, were revealed as significant regulators, implicating them in the regulation of liver homeostasis, oxidative stress, and inflammation during aging (Fig. 4B). Some downstream genes, including ppargc1a, foxo3, pik3r1, and fos, were downregulated, and myc, tnf, apoe, and ccn2 were predicted to be upregulated by the upstream regulators.

Figure 4.

IPA-based pathway, functional, and regulatory network analysis of differentially expressed aging-related genes in the female liver of BPA lineage medaka. (A) Canonical pathway enrichment showing top signaling pathways associated with altered aging-related genes, including estrogen receptor signaling, NAFLD signaling, RAR activation, and sirtuin signaling. Bars represent −log(P-value), and colors indicate predicted activation state based on z-scores. (B) Upstream regulator analysis predicting transcriptional regulators (KLF3, IL15, NOTCH1) potentially driving observed gene expression changes, with predicted activation (orange) or inhibition (green) relationships. (C) Disease and function enrichment highlighting top predicted phenotypes, including abnormal abdominal morphology, immune-mediated inflammatory disease, DNA fragmentation, and accumulation of reactive oxygen species (ROS). (D) Gene network illustrating regulators and targets associated with ROS accumulation. (E) Gene network depicting regulators and targets linked to DNA fragmentation. (F) Apoptosis-associated network integrating observed gene expression with predicted activation/inhibition relationships, where line colors indicate relationship type, and dashed vs solid lines represent indirect vs direct molecular interaction.

IPA predicted diseases and functions linked to aging gene signatures

Disease and function analysis highlighted processes directly relevant to liver aging (Fig. 4C). There was a predicted increase in the accumulation of reactive oxygen species (Fig. 4D), and DNA fragmentation (Fig. 4E) linked to liver aging. The genes linked to stress response tnf, apoe, myc, and esr1 were cooperatively associated with DNA damage and ROS production. Cellular viability and sprouting were predicted to decrease, reflecting impaired regenerative potential. Notably, apoptosis-related genes were associated with multiple upstream regulators, including pycr2, tnf, myc, esr1, igf1, and foxo3, suggesting that the dysregulation of these nodes may contribute to age-associated hepatocyte loss (Fig. 4F). Overall, the IPA results indicate that liver aging exhibits a disruption of metabolic, oxidative, and inflammatory pathways, with key roles in estrogen signaling, sirtuin activity, and RAR-mediated transcription. The interplay between ROS accumulation, DNA damage, and apoptosis highlights mechanistic links between altered upstream regulators and functional decline in aging hepatocytes.

The hepatotoxicity analysis indicated significant enrichment of pathways related to liver necrosis/cell death, hepatocellular carcinoma, hyperplasia/hyperproliferation, steatosis, and proliferation, with highly significant P-values and substantial molecular representation (Fig. S2 [20]). Network analysis further linked these changes to key biological functions, including cancer and organismal injury, lipid metabolism, cell death/survival, and the HIF-1 signaling pathway (Fig. S3 [20]). These findings suggest that the observed molecular alterations are strongly associated with liver injury, oncogenic transformation, oxidative stress response, and impaired regenerative processes, reflecting a complex interplay between degenerative and proliferative events during disease progression.

Ancestral BPA exposure alters DNA methylation in human ortholog of aging genes

KEGG pathway-enrichment analysis of the 88 dysregulated aging genes revealed significant overrepresentation of pathways critical to cellular longevity and metabolic control (Fig. 5A). Notably enriched pathways included longevity regulating pathways, FoxO signaling, PI3K-Akt, AMPK signaling, cellular senescence, growth hormone synthesis and cancer (Fig. 4A). Many of these are canonical regulators of aging and are also implicated in NAFLD progression and hepatocellular stress responses. These findings suggest that ancestral BPA exposure induces persistent reprogramming of evolutionarily conserved aging-associated genes and signaling pathways in the liver. UpSet analysis (Fig. 4B) shows that the FoxO pathway had the highest number of overlapping genes across multiple aging-associated pathways, followed by PI3K-Akt signaling and the longevity pathway. Notably, 22 genes were common across three or more pathways, indicating the convergence of stress response and energy sensing in transgenerational liver disease. These results suggest that ancestral BPA exposure disrupts core aging-regulatory pathways, likely contributing to the altered expression of mitochondrial and aging-related genes and the progression of NAFLD phenotypes.

Figure 5.

Integrated pathway enrichment and epigenetic regulation of aging-related signaling in BPA lineage livers. (A) KEGG pathway enrichment analysis of differentially expressed genes showing top pathways linked to aging, including longevity regulation, endocrine resistance, AMPK signaling, FoxO signaling, estrogen signaling, and PI3K–Akt signaling. Color scale represents −log₁₀(FDR), and bar length indicates fold enrichment. (B) Upset plot showing pathway overlaps and shared genes among significantly enriched pathways. (C and D) Integration of methylation (gene body CpG islands) and RNA-seq data reveals hypermethylation-associated downregulation of key genes (mdm2, pik3r1, prkca, irs2, foxm1) across multiple pathways.

To investigate the epigenetic basis of aging gene dysregulation, we performed integrated analysis of 5-methylcytosine (5mC) profiles and transcriptomic data obtained from the whole-genome bisulfite sequencing (WGBS) and RNAseq, respectively, of the same liver tissue. We focused on five genes, mdm2, pik3r1, prkca, irs2, and foxm1, which were involved in multiple aging-associated pathways including PI3K-Akt, AMPK, FoxO signaling, longevity regulation, and cellular senescence (Fig. 5C). These genes also showed consistent differential methylation and gene expression changes between BPA and control lineages. Furthermore, we observed that the pik3r1 gene was found common across all pathways with significant changes in gene body CpG islands methylation in the BPA lineage livers. All five genes displayed increased gene body CpG islands methylation in the BPA lineage compared to controls, with values ranging from ∼41% to 49% in the BPA lineage vs ∼17% to 29% in controls. Specifically, mdm2 exhibited both promoter and gene body CpG islands hypermethylation (48.17% BPA vs 24.36% in control lineage), correlating with its significant downregulation (Log₂FC = −1.357) in the BPA lineage. Similarly, pik3r1, prkca, and irs2 showed strong downregulation (Log₂FC from −1.26 to −1.59) along with increased methylation levels (Fig. 5D).

The dot plot showed differentially methylated CpG sites across selected genes. Each panel represents a CpG-rich (CpG Island within gene body) genomic region, with red and blue dots indicating the number of methylated and unmethylated reads, respectively, at each CpG site. The red-to-blue ratio is generally higher in BPA lineage samples compared to controls, indicating that these regions are hypermethylated in the BPA lineage relative to the control lineage. Consistent hypermethylation was observed across multiple genes, including foxm1 and irs2, in the gene body CpG islands of the BPA lineage livers (Fig. S4 [20]). Collectively, these findings suggest that ancestral BPA exposure leads to epigenetic alteration on core aging regulators, some of which act as nodes in multiple longevity and stress-response pathways. The concordance between altered 5mC levels and transcriptional repression supported a functional role of methylation in mediating environmental reprogramming of hepatic aging gene networks (Fig. 4D).

Discussion

Emerging evidence suggests that early-life environmental exposures can program long-term metabolic and epigenetic landscapes, with implications for aging and chronic disease. Our study demonstrates that ancestral BPA exposure results in transgenerational alterations in liver transcriptome and epigenome in female medaka (Oryzias latipes), culminating in premature hepatic aging. We observed significant downregulation of mitochondrial oxidative phosphorylation genes, dysregulation of conserved human and mammalian aging-related genes, and concordant DNA hypermethylation in key metabolic and regulatory genes. Our results provide direct evidence that ancestral EED exposures can impose a long-lasting molecular “aging signature” in the liver of unexposed descendants, supporting the developmental origins of health and disease (DOHaD) hypothesis.

The expression of mitochondrial genes and their corresponding functions exhibit significant changes with aging, affecting various physiological systems and contributing to the decline in cellular function. Changes in mitochondrial gene expression are known to occur with aging and can directly influence oxidative phosphorylation efficiency and ATP synthesis, leading to increased oxidative stress and cellular dysfunction [27]. One of the most striking observations is the downregulation of mitochondrial gene expression. The downregulation of nearly all 13 mitochondrial protein-coding genes in the BPA-lineage livers suggests a systemic impairment in oxidative phosphorylation (OXPHOS), which is essential for cellular energy homeostasis. In mice, mitochondrial gene expression declines markedly in later stages of life, especially in the senescent stage of the life cycle [28]. In line with our observations, literature suggests that older individuals typically exhibit lower expression of critical genes associated with mitochondrial respiration, including those coding for the respiratory complexes such as nd1, nd2, and cox1-3 [29]. Mitochondrial dysfunction associated with this downregulation has been linked to several age-associated conditions, including heightened oxidative stress and diminished antioxidant defense mechanisms within cells [30, 31]. This decline contributes significantly to metabolic dysregulation and increases cellular oxidative damage, emphasizing the need for interventions that may support mitochondrial health and counteract the effects of aging in future generations.

In addition to mitochondrial gene disruption, the transcriptome of BPA-lineage livers revealed widespread dysregulation of conserved aging-associated genes. A subset of these genes overlapped with known mammalian aging genes, indicating that BPA-induced transcriptional reprogramming targets deeply conserved pathways of aging regulation. Notably, genes involved in mitochondrial function (ppargc1a, got2, ndufs2), chromatin remodeling (sox11, phf14), and inflammation (psmb8, serpinf1) were significantly altered, pointing to a broader impact on hepatic function that spans across metabolism, genomic stability, and immune regulation. Consistent with our results, downregulation of PPARGC1A, GOT2, and NDUFS2 expression leads to reduced mitochondrial function and increased susceptibility to metabolic disorders [32, 33]. SOX11 and PHF14 (PHD finger protein 14) play a role in chromatin remodeling during aging, particularly by influencing chromatin accessibility in response to cellular stress and inflammation [34]. In aged tissues, PSMB8 contributes to the inflammatory response and is associated with the senescence-associated secretory phenotype (SASP), exacerbating tissue inflammation and promoting age-related degeneration [35]. Together, these findings suggest that ancestral BPA exposure reprograms conserved aging pathways in the liver, linking mitochondrial dysfunction, chromatin remodeling, and inflammation to premature hepatic aging and increased disease susceptibility.

Importantly, several of these transcriptional changes mirrored expression patterns observed in aged human livers, further supporting the notion of ancestral BPA-induced premature aging. The mapping of 88 dysregulated medaka aging genes to their human orthologs allowed us to uncover significant perturbations in stress and aging-related pathways such as DNA damage response, oxidative stress, apoptosis, and immune system regulation. These alterations were accompanied by dysregulation of transcriptional regulators (myc, foxm1, foxo3, nr3c1, stat5a), anti-aging (klotho, kl), metabolic controllers (foxo3, igf1, nfe2l1), and oxidative stress response (cat, mt1f, mt-co1), providing functional evidence of a disrupted hepatic aging program.

KEGG and Gene Ontology analyses further emphasized the convergence of ancestral BPA exposure-induced changes on longevity and stress-responsive signaling pathways, including the PI3K-Akt, FoxO, AMPK, and senescence pathways. The PI3K-Akt signaling pathway is known for its critical role in cellular growth, proliferation, and survival, while the FoxO family of transcription factors regulates various cellular responses, including stress resistance, metabolism, and apoptosis, thereby contributing to longevity [36, 37]. The role of AMPK, often regarded as a metabolic sensor, complements this regulatory axis. Studies have shown that AMPK activation can lead to increased expression of FoxO target genes involved in oxidative stress response and nutrient sensing, essential for promoting cellular homeostasis within aging liver tissues [38]. The interplay between these signaling pathways highlights important therapeutic avenues for mitigating liver aging and cellular senescence.

The idiopathic progressive loss of metabolic homeostasis is partly mediated by reduced Sirtuin activity. Inhibition of Sirtuin signaling in BPA lineage livers, as predicted by our IPA analysis, likely exacerbates mitochondrial dysfunction and promotes the accumulation of reactive oxygen species (ROS), amplifying oxidative damage and impairing regenerative capacity [39]. Another study supported that overexpression of SIRT1 can restore mitochondrial morphology and function [40]. The predicted downregulation of RAR activation indicates compromised retinoid signaling, which is essential for hepatocyte differentiation, lipid metabolism, and liver regeneration [41]. Disruption of vitamin A metabolism and RAR signaling has been implicated in the progression of hepatic fibrosis and impaired repair responses, which may worsen with age. Downregulation of PPARGC1A (PGC-1α) and FOXO3 observed in our network analysis may further impair mitochondrial-nuclear crosstalk, leading to transcriptional silencing of genes essential for energy metabolism and stress resistance. The downregulation of PGC-1α impairs this regulatory network, leading to reduced expression of nuclear-encoded mitochondrial genes and, consequently, decreased mitochondrial biogenesis and function [42]. These findings suggest that ancestral BPA exposure disrupts conserved Sirtuin- and retinoid-mediated regulatory networks, thereby accelerating mitochondrial dysfunction, impairing metabolic resilience, and predisposing the liver to progressive, age-associated disease.

Perhaps most critically, the integration of DNA methylation profiling with transcriptomic data revealed that a subset of aging-related genes showed strong concordance between increased 5-methylcytosine (5mC) levels and gene repression. Genes such as mdm2, irs2, and prkca displayed promoter and gene body CpG islands hypermethylation, correlating with significant transcriptional downregulation. The mice deficient in IRS2 exhibit heightened hepatic insulin resistance with inflammation, an exacerbation of metabolic dysregulation that reinforces the importance of IRS2 in the aging liver [43]. These findings support the hypothesis that BPA induces epigenetic changes in ancestral generations, potentially affecting the liver functions in unexposed descendants. Intriguingly, the upregulation of foxm1 despite gene body CpG islands hypermethylation suggests complex, possibly non-canonical mechanisms of methylation-mediated regulation, which warrants further mechanistic investigation. Collectively, this study demonstrates that ancestral BPA exposure reprograms the hepatic aging trajectory through transcriptional and epigenetic disruption of conserved genes and pathways. These findings provide insights into a mechanistic basis for how early-life environmental exposures can promote premature liver aging and metabolic dysfunction across generations, with implications for public health. Further research into the plasticity of the epigenome during critical developmental windows and its long-term impact on organ-specific aging may reveal novel targets for intervention in environmentally induced chronic diseases.

Finally, to ensure appropriate interpretation of our pathway analysis results, we acknowledge important methodological caveats. Ingenuity Pathway Analysis (IPA) relies on an internal curated knowledge base that may occasionally associate pathways with tissues or functions outside the primary study context. For instance, predictions related to ovarian-specific processes such as “Quantity of Graafian follicle” in liver tissue may reflect shared molecular mechanisms across tissues or database limitations rather than direct biological relevance to the liver. We have, therefore, focused our discussion on the most tissue-relevant and biologically plausible findings while remaining aware of these inherent limitations in computational pathway analysis tools.

Materials and methods

Animal care and maintenance

Animal care, preparation of BPA exposure solution, maintenance of animal colony, and sample collection were performed as previously described [17, 19, 44]. The Hd-rR strain of medaka was used in this study [45]. All exposure procedures, animal handling, and euthanasia protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of North Carolina Greensboro (#20-002 and #22-1169) and the University of Missouri (#43230 Amendment 2.1). Adult medaka (4 months old) were maintained in 20 L glass aquaria under a 14:10 hours light-dark cycle in a recirculating water system with 25% water exchange every 4 hours at 26 ± 1 °C and were fed three times daily with Otohime granular food and newly hatched brine shrimp (Artemia nauplii). After mating of six breeding pairs, eggs were collected, detangled with forceps under a brightfield stereoscope, and viable fertilized eggs were selected for exposure. Six breeding pairs produced more than 100 fertilized eggs. A total of 30 eggs were randomly selected for each exposure.

Chemical exposure and production of BPA-exposed transgenerational offspring

Bisphenol A (BPA) has been shown to disrupt metabolic health across a range of concentrations [46-48]. A concentration of 10 μg/L BPA was selected based on a dose-response analysis, reflecting environmentally relevant levels reported in many regions worldwide [44, 49]. Three individual tanks were assigned per exposure group, and a total of two exposure groups were set up. Throughout the exposure period, measured BPA levels remained within <10% of the nominal concentration, with uptake estimated at 20 pg/mg egg/day, as previously described [44, 49, 50]. Exposure was initiated at 8 hours post-fertilization (hpf) and continued until 15 days post-fertilization (dpf), with daily renewal of BPA solution. This exposure window was chosen to encompass key epigenetic reprogramming events in germ cells while avoiding early embryonic reprogramming occurring between fertilization and the blastula stage (≤8 hpf). The 10 μg/L BPA concentration used for F0 exposure falls within the environmentally relevant range documented in both human exposure studies and aquatic ecosystems. Global environmental monitoring demonstrates BPA concentrations in surface waters ranging from 39 ng/L to 56 μg/L [51, 52]. These ranges support the environmental relevance of the exposure dose used in this study. Following exposure, both the F0 generation and subsequent offspring were reared in clean water without further BPA treatment. Notably, this window coincides with critical periods of sex determination and liver differentiation [53]. Generation of the BPA-exposed transgenerational lines of medaka was previously described [44]. Briefly, following BPA exposure for 15 days, larvae were raised in clean water until they became adults, designated as F0 adults (intragenerational due to direct exposure to BPA). At 120 days postfertilization, all experimental medaka reached sexual maturity and spawned [54]. A total of six pairs of fish from the F0 generation were bred at 120 days postfertilization to produce F1 (intergenerational). The same methodology was used to produce the F2 generation (transgenerational) that was used in this study. To collect tissue samples, three fish from each tank were euthanized with buffered MS-222 (250 mg/L). There were three tanks per lineage. In total, liver tissues from 9 individual fish were collected from each lineage.

Histological tissue processing

As mentioned in the animal care and maintenance section, three independent 20L glass tanks were used to raise control fish and three tanks for BPA-lineage fish. From each tank, three fish were sampled. A total of three tanks were assigned per exposure group. Therefore, livers from 9 fish/exposure group were used for histological processing. Fish were euthanized with MS-222 (250 mg/L), a half of the liver was collected for molecular analysis, and the other half was fixed in Bouin's solution for histological analysis. After tissue processing, liver samples were embedded in low-temperature paraffin, sectioned at 5 μm thickness in a microtome, and stained with hematoxylin following the optimized histological protocol [17]. In total, liver tissues from 9 fish per group were histologically examined. More than 100 sections were prepared per liver. All sections were microscopically examined. Five representative slides were selected per liver to further examine phenotypic alterations. Fatty liver pathology was confirmed based on the presence of vacuolated hepatocytes, which is an indicator of steatosis in the liver.

Nucleic acids extraction, library preparation, and sequencing

For RNA-seq and WGBS analyses, RNA and DNA from the the livers of the three fish within each tank were pooled to create a single pooled sample per tank which served as a biological replicate, allowing for assessment of biological variability between tanks while reducing technical variations.

Liver from control and BPA-lineage female fish were collected for total RNA extraction using the Quick RNA/DNA Miniprep Plus Kit (#D-7003, Zymo Research, CA, USA) following the manufacturer's protocol [50, 55, 56]. RNA integrity was assessed by bleach gel electrophoresis [57], and concentrations were measured with a Nanodrop 2000 and Qubit fluorometer (Thermo Fisher, Waltham, MA). Total RNA integrity and purity were assessed using an Agilent 2100 Bioanalyzer before proceeding with library preparation. Transcriptome libraries were constructed with the NEBNext Ultra II RNA Kit (New England Biolabs) according to the manufacturer's instructions and sequenced on an Illumina HiSeq X platform (Novogene, CA, USA) using a 150 bp paired-end strategy, 20 to 40 million reads per replicate. Genomic DNA quality was verified by agarose gel electrophoresis, requiring intact high-molecular-weight bands. DNA concentration and purity were quantified by Qubit and NanoDrop, ensuring suitable input for bisulfite conversion. Whole-genome bisulfite sequencing (WGBS) libraries were prepared as previously described [56]. Briefly, 100 ng of genomic DNA from each sample was spiked with 0.5% unmethylated E. coli DNA and processed using the NEBNext^® Ultra™ II FS DNA Library Prep Kit (NEB, E6177) following the manufacturer's instructions. WGBS libraries were sequenced on an Illumina NextSeq 500 platform (University of Missouri–Columbia DNA Sequencing Core) using a 150 bp single-end strategy.

Bioinformatic analyses

RNA-seq analysis

Raw sequencing reads were preprocessed with Fastp v0.23.2, an efficient FASTQ preprocessor that performs adapter trimming, quality filtering, per-read quality pruning, and comprehensive quality control in a single pass [58]. The resulting high-quality reads were aligned to the Oryzias latipes reference genome (ASM223467v1) using STAR v2.7.7a [59], with transcript annotation provided via the Ensembl GTF file (sjdbGTFfile = Oryzias_latipes.ASM223467v1.108.gtf) to improve splice junction mapping accuracy).

DESeq2 v1.34.0 [60] was used for differential gene expression analyses, employing standard median-of-ratios normalization. Sample clustering was performed on regularized log-transformed count matrices (rlog) to assess batch effects and sample concordance. Differentially expressed genes (DEGs) were defined based on an FDR-adjusted P-value of <0.05, using the Benjamini–Hochberg procedure to control multiple testing.

Cluster heatmaps were generated using ClustVis [61] with row-wise z-score normalization. Both rows and columns were hierarchically clustered using Euclidean distance and Ward's minimum variance linkage method. In addition, Ingenuity Pathway Analysis was performed to further interpret the functional implications of the data [62]. Aging-related genes were compared to publicly available mammalian (127 publicly available microarray and RNA-Seq datasets from mice, rats and humans) [63] and human aging [64] gene sets. Gene ontology data were visualized by using Expression analyst [65]. Shiny GO [66].

WGBS analysis

Methylation data from bisulfite sequencing were processed using the nf-core/methylseq [67] automated bioinformatics pipeline (v1.6.1), which is a community-curated, scalable, and reproducible workflow designed specifically for methylation analyses. Within this pipeline, the Bismark [68] workflow was employed to align bisulfite-treated sequencing reads to the Oryzias latipes reference genome (assembly ASM223467v1).

Following alignment, Bismark coverage files were generated for each individual sample, containing methylation percentages at each CpG site along with coverage metrics detailing the number of methylated and unmethylated cytosines. These coverage files served as the input for downstream analyses using the methylKit [69] R package for differential methylation analysis. To ensure data reliability, CpG sites were filtered to retain only those with a minimum coverage of five reads per site. We used an adjusted P-value threshold of <.05 combined with a minimum methylation difference of 10% between control and BPA-lineage groups to identify differentially methylated regions. For promoter-specific analyses, the core promoter region was defined as the 200 bp window flanking the transcription start site (−100 bp upstream to +100 bp downstream), which was examined for differential methylation patterns. We also analyzed differential methylation within gene-body CpG islands. First, CpG islands were identified de novo using the makeCGI R package using default parameters (available at https://www.haowulab.org/software/makeCGI). We calculated differential methylation for those CpG islands overlapping gene bodies to investigate gene-body specific methylation changes. The transcriptome and DNA methylome datasets analyzed in this study are part of a larger project on transgenerational epigenetic inheritance and are deposited to GEO under accession numbers GSE285665 (methylome) and GSE252744 (transcriptome) [70].

Funding

The research was supported by funds from the US National Institutes of Health (National Institute of Environmental Health Sciences [R01ES032452, R21ES027123] and Eunice Kennedy Shriver National Institute of Child Health and Human Development [R21HD098621]) and Mizzou Forward Faculty Startup Package of the University of Missouri to R.K.B.

Disclosures

The authors have nothing to disclose.

Data availability

Some of the data sets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.