DNA methylation and hydroxymethylation combined with transcriptional profiling identify key regulators of hyperoxia-induced bronchopulmonary dysplasia

Abstract

Background

Hyperoxia-induced bronchopulmonary dysplasia (BPD) is a major cause of lung injury in premature infants. Epigenetics, particularly DNA (hydroxy)methylation, has been identified as a crucial regulator of BPD pathogenesis. This study aimed to reveal key regulators and pathogenic genes involved in hyperoxia-induced BPD via DNA (hydroxy)methylation and transcriptional analysis.

Methods and results

Multi-omics analyses including RNA-seq, reduced representation bisulfite sequencing (RRBS), and oxidative RRBS (oxRRBS) were conducted on lung tissues from hyperoxia-induced rat BPD model. Differentially methylated and hydroxymethylated regions (DMRs and DhMRs) were further detected by targeted bisulfite sequencing (TBS) and oxidative TBS (oxTBS). Differentially expressed genes (DEGs) were finally verified in hyperoxia-exposed lung tissues by qPCR, western blotting, and immunohistochemistry. Our integrated analysis identified 2058 DEGs, 62,123 DMRs, and 33,212 DhMRs in hyperoxia-induced BPD. Among them, eighteen candidate genes with altered expression patterns were revealed to be involved in BPD-related pathways. Notably, ten candidate genes, including Cxcl6, Gpr39, Hs6st2, Htatip2, Apln, Calca, Hist1h1t, Lgals3, Rarres1, and Rasl2-9, exhibited significant upregulation with both hypo-DNA methylation and hyper-DNA hydroxymethylation levels. Conversely, eight candidate genes, containing Krt76, Spon2, Abcc6, Egfl7, Gpbar1, Myh6, Tgfbi, and Tmem100, displayed pronounced downregulation associated with both hyper-DNA methylation and hypo-DNA hydroxymethylation levels. Most importantly, the upregulation of Apln and Calca was further validated in hyperoxia-induced BPD, which was characterized by reduced DNA methylation and increased DNA hydroxymethylation levels at their promoter regions.

Conclusions

This study reveals that hyperoxia triggers decreased DNA methylation together with increased DNA hydroxymethylation at promoter regions of Apln and Calca, promoting their gene expression and contributing to BPD pathogenesis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13148-025-01926-9.

Background

Bronchopulmonary dysplasia (BPD) is a chronic lung disease resulting from abnormal development in the lower respiratory tract, which is influenced by various risk factors, including prematurity, prolonged mechanical ventilation, oxygen toxicity, and maternal conditions such as chorioamnionitis [1]. BPD is a prevalent long-term complication in premature infants, affecting 30–50% of neonates born at < 28-week gestation. This condition leads to severe and lifelong illnesses, including pulmonary hypertension and neurodevelopmental impairment, which significantly affect both quality of life and long-term outcomes [2, 3]. The pathogenesis of BPD involves complex mechanisms, with genetic predisposition playing a critical role. Functional alterations in genomic pathways associated with genes encoding essential growth factors, such as vascular endothelial growth factor (VEGF), transforming growth factor-β (TGF-β), and insulin-like growth factor (IGF), are key determinants of susceptibility to BPD [4]. Additionally, specific cytokines, particularly interleukin-24 (IL-24), have been identified as key mediators of oxidative stress and pulmonary inflammatory responses that are characteristic of BPD progression [5]. Hyperoxia exposure and the role of reactive oxygen species (ROS) are the primary risk factors contributing to the development of BPD [6]. Long-term exposure to hyperoxia can cause irreversible developmental abnormalities [7, 8]. Studies have demonstrated that the ROS and DNA methylation are involved in BPD [9]. ROS triggers DNA hypermethylation, leading to the suppression of certain genes and subsequent cell apoptosis [10]. Hydroxymethylated cytosine, an intermediate step of active demethylation, has been illustrated to be initiated by oxidative stress in several pathophysiological processes [11–14]. This mechanism highlights the direct relationship between oxidative stress, epigenetic modifications, and the regulation of gene expression. However, studies on the dynamics of genome-wide DNA (hydroxy)methylation in hyperoxia-induced BPD are limited.

To evaluate whether long-term hyperoxia exposure during early lung development affects DNA (hydroxy)methylation dynamics and transcriptional patterns of key regulators, a hyperoxia-induced rat BPD model was established using rat neonates exposed to 95% oxygen from postnatal day 1 to 10. Reduced representation bisulfite sequencing (RRBS) and oxidative RRBS (oxRRBS) were employed to detect DNA methylation and hydroxymethylation levels in the lung samples from hyperoxia- or normoxia-treated rat neonates. RNA sequencing of differentially expressed genes (DEGs) was also carried out. The DEGs together with either differentially methylated regions (DMRs) or differentially hydroxymethylated regions (DhMRs) were quantitatively analyzed by targeted bisulfite sequencing (TBS) and oxidative TBS (oxTBS). Eventually, the gene expression levels in lung tissues after hyperoxia exposure were verified by qPCR, western blotting, and immunohistochemistry.

In this study, a total of 2,058 DEGs, 62,123 DMRs, and 33,212 DhMRs were identified using RNA-seq, RRBS, and oxRRBS analysis. Ten candidate genes, including Cxcl6, Gpr39, Hs6st2, Htatip2, Apln, Calca, Hist1h1t, Lgals3, Rarres1, and Rasl2-9, were upregulated with hypo-DNA methylation and hyper-DNA hydroxymethylation levels after hyperoxia exposure in BPD. In contrast, eight candidate genes containing Krt76, Spon2, Abcc6, Egfl7, Gpbar1, Myh6, Tgfbi, and Tmem100 were downregulated with hyper-DNA methylation and hypo-DNA hydroxymethylation levels after hyperoxia treatment. This investigation further confirmed that hyperoxia triggered upregulation of Apln and Calca in lung tissues from rat BPD, characterized with hypo-DNA methylation and hyper-DNA hydroxymethylation levels at their promoter regions. Taken together, this study represents the first integrated analysis of DNA methylation, DNA hydroxymethylation, and transcriptional changes in hyperoxia-induced BPD, identifying key regulators, Apln and Calca. Our findings reveal that hyperoxia induces the upregulation of Apln and Calca through the hypo-DNA methylation and hyper-DNA hydroxymethylation levels at their promoter regions, which may contribute to the progression of BPD.

Results

Establishment of hyperoxia-induced rat bronchopulmonary dysplasia

The establishment of hyperoxia-induced rat bronchopulmonary dysplasia (BPD) was performed as previously described [15]. Briefly, rat neonates in hyperoxia group were exposed to 95% oxygen in a plexiglass chamber containing an oxygen volume of 1.5 L/min from postnatal day 1 to day 10. To evaluate hyperoxia-induced lung injury, lung samples of rat neonates treated with either room air (RA) or 95% oxygen (O₂) throughout postnatal day 1 to 10 were assessed by H&E staining. On postnatal day 10, the lung samples exposed to hyperoxia exhibited thinner septa but larger air spaces than those treated under normoxia (Fig. 1A). Additionally, the mean linear intercept (MLI) of the lung tissues from the hyperoxic group was significantly higher than those in the normoxic group (Fig. 1B), confirming the establishment of hyperoxia-induced lung injury in rat BPD.

Fig. 1.

Assessment of hyperoxia-induced lung injury. A Representative images of hematoxylin and eosin (H&E)-stained lung sections from rat neonates exposed to room air (RA) or hyperoxia (O₂) throughout postnatal day 1 to 10. Scale bars = 100 μm. B Quantitative analysis of mean linear intercept (MLI) of lung sections (n = 5). Quantitative data were presented as mean ± SD. Statistical significance was analyzed via an unpaired Student’s t test and represented as ***P < 0.001. RA and O₂ indicate room air and hyperoxia, respectively

RNA sequencing and KEGG analysis of differentially expressed genes in hyperoxia-induced rat BPD

To explore the dynamic change of differentially expressed genes (DEGs), RNA sequencing was carried out on the lung tissue of neonatal rats from either the hyperoxic or normoxic groups. Principal component analysis (PCA) of gene expression indicated that the eight independent biological replicates for RNA-seq were clustered together, with good separation between varieties (Fig. 2A). As shown in Figure S1, the range and distribution of the gene expression values were consistent among the samples, suggesting the reliability and high quality of the RNA sequencing data. HTseq-count and DESeq2 were employed to analyze the effect of hyperoxia on gene expression. A | log2FC (fold change) | of mRNAs level ≥ 1 along with P ≤ 0.05 was considered significant. As a result, a total of 837 genes were significantly upregulated (log2FC ≥ 1), while 1221 genes were markedly downregulated (log2FC ≤ -1) (Fig. 2B). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that the DEGs in this study were notably enriched in twenty signaling pathways (Fig. 2C). Among them, the highlighted signaling pathways in Fig. 2C have been previously reported to associate with BPD including p53, TNF, Ras, PI3K-Akt, MAPK, transcriptional misregulation in cancer, rheumatoid arthritis, olfactory transduction, neuroactive ligand-receptor interaction, hematopoietic cell lineage, focal adhesion, cytokine-cytokine receptor interaction, complement and coagulation cascades, cell adhesion molecules (CAMs), axon guidance, and arachidonic acid metabolism. Besides, a complete list of all differentially expressed genes (DEGs) identified in the RNA-seq analysis is presented in Table S6.

Fig. 2.

RNA sequencing and KEGG analysis of differentially expressed genes in hyperoxia-induced rat BPD. A Principal component analysis (PCA) of RNA-seq data comparing lung tissues from hyperoxic and normoxic groups. The red and blue dots represented biological replicas from hyperoxic or normoxic groups, respectively (n = 8). B Volcano plot illustrating differentially expressed genes (DEGs) between the hyperoxic and normoxic groups. C KEGG pathway enrichment analysis of the significantly DEGs. The previously reported signaling pathways related to BPD were highlighted in red. DEGs indicates differentially expressed genes, while KEGG indicates Kyoto Encyclopedia of Genes and Genomes

Dynamic alterations of genome-wide DNA (hydroxy)methylation patterns in hyperoxia-induced rat BPD

To detect the DNA (hydroxy)methylation dynamics in hyperoxia-induced rat BPD, RRBS and oxRRBS were utilized to analyze the profiles of genomic 5-mC and 5-hmC in lung tissues from neonatal rats exposed to hyperoxia or normoxia. As demonstrated in Fig. 3A and B, the PCA of the DNA methylation and hydroxymethylation data displayed a good separation between the varieties of the eight independent biological replicates. Compared to the normoxic group, a total of 62,123 differentially methylated regions (DMRs) and 33,212 differentially hydroxymethylated regions (DhMRs) have been identified in lung tissues from hyperoxia-exposed neonatal rats (Fig. 3C and D). Besides, the distribution of DMRs and DhMRs on chromosomes was shown in Figure S2A and S2B. According to the DNA methylation distribution analysis at different chromosome loci, chromosome 1 demonstrated the highest DNA methylation level (Figure S2A). On the other hand, the DNA hydroxymethylation distribution analysis at different chromosome loci revealed that chromosome 1 exhibited the highest DNA hydroxymethylation level (Figure S2B). However, insignificant alteration of DNA methylation and hydroxymethylation patterns in sex chromosomes was found in hyperoxia-induced rat BPD.

Fig. 3.

Genome-wide analysis of DNA (hydroxy)methylation patterns in hyperoxia-induced rat BPD. A, B Principal component analysis (PCA) of DNA methylome (A) and hydroxymethylome (B) of lung tissues from hyperoxic and normoxic groups. The red and blue dots represented samples from hyperoxic or normoxic groups, respectively (n = 8). C, D Heatmap displaying differentially methylated and hydroxymethylated regions (DMRs and DhMRs) between hyperoxic and normoxic groups. E, F Dynamic changes of genome-wide DNA methylation (E) and hydroxymethylation (F) patterns in different gene structural elements. G Correlation analysis of genome-wide differential DNA methylation and hydroxymethylation patterns. PCA and PC indicate principal component analysis and principal component, while DMR and DhMR indicate differentially methylated region and differentially hydroxymethylated region. RA and O₂ indicate room air and hyperoxia, respectively

Furthermore, the DNA methylation and hydroxymethylation levels in different genome regions are analyzed in Fig. 3E and F. Interestingly, a slightly increased DNA methylation pattern along with a reduced DNA hydroxymethylation pattern in the regions of promoter, intron, 3’untranslated region, and genebody was detected in hyperoxia group (Fig. 3E and F). Moreover, it was demonstrated that the changes of genome-wide DNA methylation patterns in both the hyperoxic and normoxic groups represented the opposite trend to that of DNA hydroxymethylation patterns in each gene element (Fig. 3E and F). As shown in Fig. 3G, a negative correlation between the differential DNA methylation regions and differential DNA hydroxymethylation regions was also found in hyperoxia-induced rat BPD. Taken together, our findings suggest that the dynamic changes in genome-wide DNA (hydroxy)methylation patterns could be considered as essential characteristics in hyperoxia-induced rat BPD.

KEGG and GO analysis of transcriptomic and DNA (hydroxy)methylomic alterations at promoter regions in hyperoxia-induced rat BPD

To identify the epigenetic regulator of hyperoxia-induced BPD, we focused on the comprehensive analyses of DNA methylation and hydroxymethylation patterns at promoter regions of DEGs identified through RNA-seq analysis. The DEGs exhibiting significant differential methylation or hydroxymethylation at promoter regions underwent functional characterization through Gene Ontology (GO) and KEGG pathway enrichment analyses. Previous studies have demonstrated that transcriptional repression is associated with hypermethylation or hypohydroxymethylation at promoter regions, whereas transcriptional activation correlates with hypomethylation or hyperhydroxymethylation at these regulatory elements [16]. To elucidate the DNA modification of gene expression, the DEGs were integrated with DMRs or DhMRs in hyperoxia-induced rat BPD. Subsequently, the DEGs exhibiting inverse correlation with DNA methylation or positive correlation with DNA hydroxymethylation at promoter regions underwent KEGG pathway enrichment analysis (Figure S3). KEGG analysis showed that the DEGs with negative changes of DNA methylation pattern at promoter regions were significantly enriched in BPD-associated pathways, including cGMP-PKG signaling pathways, regulation of actin cytoskeleton, platelet activation, olfactory transduction, neuroactive ligand–receptor interaction, hypertrophic cardiomyopathy (HCM), focal adhesion, dilated cardiomyopathy (DCM), complement and coagulation cascades, and cardiac muscle contraction (Figure S3A). Additionally, novel pathways were also identified to enrich in hyperoxia-induced rat BPD, such as cAMP signaling pathway, thyroid hormone signaling pathway, proximal tubule bicarbonate reclamation, protein digestion and absorption, mineral absorption, legionellosis, insulin resistance, aldosterone-regulated sodium reabsorption, adrenergic signaling in cardiomyocyte, and ABC transporters (Figure S3A). GO analysis demonstrated significant enrichment in biological processes related to extracellular space, negative regulation of blood pressure, regulation of the force of heart contraction, angiogenesis, and cytoplasm, which were later verified in association with Apln or Calca expression (Figures S4A and S5A).

Conversely, the DEGs with positive changes in DNA hydroxymethylation pattern at promoter regions showed significant enrichment in BPD-related pathways, containing PI3K-Akt, MAPK, viral myocarditis, phagosome, HCM, focal adhesion, ECM-receptor interaction, DCM, complement and coagulation cascades, CAMs, cardiac muscle contraction, and apelin signaling pathway (Figure S3B). Notably, additional pathways were identified to be enriched in hyperoxia-induced rat BPD, such as protein digestion and absorption, pancreatic secretion, malaria, human papillomavirus infection, glycosphingolipid biosynthesis, glycosaminoglycan biosynthesis, cellular senescence, and adrenergic signaling in cardiomyocytes (Figure S3B). GO analysis further revealed the enrichment in hormone activity, extracellular space, feeding behavior, perinuclear region of cytoplasm, angiogenesis, cytoplasm, monocyte chemotaxis, and regulation of heart rate, with subsequent validation confirming associations with Apln and Calca expression (Figures S4B and S5B).

Integrative analysis of transcriptomic and DNA (hydroxy)methylomic changes identified eighteen key regulatory genes in hyperoxia-induced rat BPD

This investigation examined hyperoxia-mediated alterations in the transcriptome and DNA (hydroxy)methylome in a rat model of BPD. Through integrative analysis, we identified eighteen candidate genes exhibiting significant differential expression concurrent with altered DNA methylation and hydroxymethylation patterns at their promoter regions. Specifically, we observed 76 upregulated genes associated with hypo-DNA methylation and 51 upregulated genes displaying hyper-DNA hydroxymethylation at their promoter regions (Fig. 4A). Additionally, 122 downregulated genes exhibited hyper-DNA methylation, while 52 downregulated genes showed hypo-DNA hydroxymethylation at their promoter regions (Fig. 4B). Among these differentially regulated genes, ten upregulated candidates (Cxcl6, Gpr39, Hs6st2, Htatip2, Apln, Calca, Hist1h1t, Lgals3, Rarres1, and Rasl2-9) and eight downregulated candidates (Krt76, Spon2, Abcc6, Egfl7, Gpbar1, Myh6, Tgfbi, and Tmem100) were identified as potentially crucial mediators of hyperoxia-induced lung injury in rat BPD (Fig. 4A and B). These genes were selected based on their functional relevance to BPD pathogenesis, as supported by prior literature and pathway analysis [17–34]. The transcriptional profiles and DNA (hydroxy)methylation patterns of these 18 candidate genes under hyperoxic and normoxic conditions were presented in Fig. 4C to E. Notably, quantitative analysis revealed that Apln and Calca exhibited significant upregulation in the hyperoxic group, with 2.57-fold and 2.79-fold increases in mRNA expression when compared to normoxic controls (Fig. 4C and Table S3). Correspondingly, the DNA methylation level of Apln and Calca decreased by 0.52-fold and 0.58-fold after hyperoxic exposure (Fig. 4D and Table S4). Conversely, the DNA hydroxymethylation level of Apln and Calca in hyperoxic group was increased by 3.53-fold and 4.70-fold than normoxic group, indicating hyperoxia-induced upregulation of Apln and Calca in rat BPD (Fig. 4E and Table S5).

Fig. 4.

Multi-omics analyses of transcriptomic and DNA (hydroxy)methylomic changes identified eighteen key regulatory genes in hyperoxia-induced rat BPD. A Venn diagram illustrating the intersection of upregulated differentially expressed genes (up-DEGs) with hypomethylated differentially methylated regions (hypo-DMRs) (purple) and up-DEGs with hyperhydroxymethylated differentially hydroxymethylated regions (hyper-DhMRs) (yellow) at promoter regions (n = 8). B Venn diagram showing the overlap between downregulated differentially expressed genes (down-DEGs) with hypermethylated DMRs (hyper-DMRs) (purple) and down-DEGs with hypohydroxymethylated DhMRs (hypo-DhMRs) (yellow) at promoter regions (n = 8). C Heatmap representation of the eighteen identified DEGs in hyperoxic and normoxic groups. RNA-seq data were normalized by log2 normcounts, with values ranging from − 0.38 to 14.30 (n = 8). D Heatmap illustration of DNA methylation level at promoter regions for the eighteen candidate genes. The analysis of DMRs was normalized by mean methylation level, with values ranging from 0.09 to 0.90 (n = 8). E Heatmap illustration of DNA hydroxymethylation level at promoter regions for the eighteen candidate genes. The analysis of DhMRs was normalized by mean hydroxymethylation level, with values ranging from 0.01 to 0.29 (n = 8). DMR and DhMR indicate differentially methylated region and differentially hydroxymethylated region, while DEG indicates differentially expressed gene. RA and O₂ indicate room air and hyperoxia, respectively

Hyperoxia-induced upregulation of Apln and Calca with hypo-DNA methylation and hyper-DNA hydroxymethylation at promoter regions in rat BPD

To validate the DNA methylation and hydroxymethylation patterns of eighteen candidate genes, we employed TBS and oxTBS for quantitative assessment of DNA methylation and hydroxymethylation levels at protomer regions. Consistently, the analysis revealed that DNA methylation levels of Apln and Calca decreased by 0.74-fold and 0.92-fold after hyperoxic exposure (Fig. 5A and B). Concurrently, DNA hydroxymethylation levels exhibited upregulation by 1.47-fold and 1.19-fold, respectively (Fig. 5C and D). These findings demonstrated that hyperoxia-induced upregulation of Apln and Calca, accompanied by hypo-DNA methylation and hyper-DNA hydroxymethylation levels at promoter regions. To further verify Apln and Calca expression in hyperoxia-exposed lung tissues, we assessed mRNA and protein levels via qPCR and western blot analysis (Fig. 6A to C). Compared to the normoxic group, mRNA expression of Apln and Calca in hyperoxic group increased by 1.62-fold and 2.55-fold, respectively (Fig. 6A), while protein level showed enhancement by 1.57-fold and 2.99-fold, respectively (Fig. 6B and C). Immunohistochemical staining of lung sections further confirmed the hyperoxia-induced upregulation of Apln and Calca expression (Fig. 6D), corroborating the results from RNA-seq, qRCR, and western blot analysis.

Fig. 5.

Increased DNA methylation with decreased DNA hydroxymethylation levels at promoter regions of Apln and Calca in hyperoxia-induced rat BPD. A, B Quantitative assessment of DNA methylation levels at promoter regions of Apln (A) and Calca (B) in lung tissues from hyperoxic and normoxic groups (n = 8). C, D Quantitative assessment of DNA hydroxymethylation levels at promoter regions of Apln (C) and Calca (D) in lung tissues from hyperoxic and normoxic groups. Quantitative data were presented as mean ± SD. Statistical significance was analyzed via an unpaired Student’s t test and represented as *P < 0.05 and **P < 0.01. RA and O₂ indicate room air and hyperoxia, while CpG indicates cytosine–phosphate–guanine

Fig. 6.

Upregulation of Apln and Calca in hyperoxia-induced rat BPD. A Quantitative real-time PCR analysis of relative mRNA expression level of Apln and Calca in lung tissues exposed to hyperoxia or normoxia (n = 9). B, C Western blotting (B) and quantitative analysis (C) of the relative protein levels of Apln and Calca in lung tissues exposed to hyperoxia or normoxia (n = 3). D Immunohistochemical staining of Apln or Calca in lung sections from hyperoxic and normoxic groups. Scale bars = 100 μm. Quantitative data were presented as mean ± SD. Statistical significance was analyzed via an unpaired Student’s t test and represented as ***P < 0.001. RA and O₂ indicate room air and hyperoxia, respectively

Additionally, our findings of Apln in regulation of hyperoxia-induced BPD were in accordance with previous studies demonstrating its association with BPD pathogenesis [17]. Although the function of Calca in hyperoxia-induced BPD has not been reported previously, our functional GO enrichment analysis revealed both Apln and Calca were involved in BPD related signaling pathways (Figure S4 and S5). GO analysis of hyperoxia-induced DEGs with DMRs or DhMRs indicated potential roles for Apln and Calca in regulating extracellular space, negative regulation of blood pressure, hormone activity, and feeding behavior (Figure S5). Furthermore, it was revealed that Apln was prone to in regulation of the force of heart contraction, angiogenesis, and perinuclear region of cytoplasm, while Calca was associated with regulation of heart rate, cytoplasm, and monocyte chemotaxis (Figure S5). Interestingly, our investigation revealed significantly reduced DNA methylation of Lgals3 in the hyperoxic group compared to normoxic controls, while Tgfbi, Egfl7, and Tmem100 exhibited significantly elevated methylation levels (Figure S6). In addition, the DNA hydroxymethylation level of Rarres1 increased in the hyperoxic group (Figure S7). Collectively, these findings provide evidence of epigenetic regulation in hyperoxia-induced BPD through the upregulation of Apln and Calca, characterized by decreased DNA methylation and increased DNA hydroxymethylation levels at promoter regions (Fig. 7). This suggests a mechanistic basis underlying their transcriptional activation.

Fig. 7.

Schematics of the proposed mechanism regarding hyperoxia-induced upregulation of key regulators (Apln and Calca) in bronchopulmonary dysplasia (BPD), attributed to reduced DNA methylation and increased DNA hydroxymethylation levels at promoter regions

Discussion

This study provides novel insights into the epigenetic mechanisms underlying hyperoxia-induced BPD through comprehensive analysis of genome-wide DNA methylation and hydroxymethylation patterns. Our findings highlight the crucial role of DNA (hydroxy)methylation in regulating gene expression at promoter regions during BPD pathogenesis. Through multi-omics analysis of lung samples from hyperoxia-induced rat BPD, including RNA-seq, RRBS, and oxRRBS, we identified 2,058 DEGs, 62,123 DMRs, and 33,212 DhMRs. The integrated analysis of DEGs along with DMRs or DhMRs revealed several enriched signaling pathways related to BPD, emphasizing the regulation of DNA methylation and hydroxymethylation in BPD pathogenesis.

DNA methylation at promoter regions demonstrates an inverse correlation with transcriptional activity, whereas DNA hydroxymethylation at promoter regions exhibits positive association with gene expression. Dynamic alterations of DNA methylation and demethylation at promoter regions exert a regulatory role on gene expression and subsequent biological functions [29]. Besides, DNA hydroxymethylation, serving as an important demethylation intermediate, functions as a primary epigenetic modulator of target gene expression [35]. Previous studies have demonstrated that even minor epigenetic alterations can induce substantial changes in transcriptional activity [36, 37]. On this basis, this study focused on analyzing the DEGs with dynamic changes of DNA methylation and hydroxymethylation patterns at promoter regions after hyperoxia treatment in BPD. Inspiringly, a slightly increase of DNA methylation together with reduced DNA hydroxymethylation levels in the regions of promoter, intron, 3’untranslated region, and genebody was detected in the hyperoxia group. Multi-omics analyses of transcriptome and DNA (hydroxy)methylome identified ten candidate genes (Cxcl6, Gpr39, Hs6st2, Htatip2, Apln, Calca, Hist1h1t, Lgals3, Rarres1, and Rasl2-9) upregulated with hypo-DNA methylation and hyper-DNA hydroxymethylation levels at promoter regions in hyperoxia-induced rat BPD. On the contrary, eight candidate genes (Krt76, Spon2, Abcc6, Egfl7, Gpbar1, Myh6, Tgfbi, and Tmem100) were downregulated with hyper-methylation and hypo-hydroxymethylation levels at promoter regions. Notably, these findings were in line with previous studies demonstrating hyperoxia induced significant decrease of Tgfbi, Egfl7, and Tmem100 but increase of Rarres1 [18–21].

Specifically, this study also provides evidence of the involvement of Apln and Calca in hyperoxia-induced lung injury. The downregulation of DNA methylation but upregulation of hydroxymethylation levels at protomer regions of Apln and Calca was quantitatively verified by TBS and oxTBS analyses. Moreover, the upregulation of Apln and Calca at both the mRNA and protein levels in lung tissues from rat BPD was verified, implying their contribution to BPD pathogenesis. Apln (Apelin) encodes a peptide ligand for the G protein-coupled APJ receptor, which plays a critical role in vascular development and angiogenesis. Our findings regarding Apln are consistent with previous studies demonstrating its crucial role in the pathophysiology of BPD by impairing both alveolar and vascular development [17, 38]. The increased expression of Apln in hyperoxia-induced BPD suggests a potential involvement of apelin/APJ signaling in the pathophysiology of arrested alveolarization and pulmonary hypertension, which are key hallmarks of experimental BPD. Calca (Calcitonin gene-related peptide) encodes a neuropeptide involved in vasodilation and inflammation. Although previous studies have linked Calca ablation to accelerated pulmonary fibrosis in rat models [24], its role in hyperoxia-induced BPD has not been previously reported. Our GO analysis revealed the involvement of Calca in BPD-related signaling pathways, and we further demonstrated that hyperoxia induces Calca upregulation through hypo-DNA methylation and hyper-hydroxymethylation at its promoter regions. This study is the first to identify the regulatory role of Calca in BPD pathogenesis. While the precise mechanisms remain to be fully elucidated, our findings provide compelling evidence supporting its fundamental contribution to alveolar simplification, likely through excessive vasodilation and disruption of the alveolar–capillary barrier [24, 39, 40].

The role of hyperoxia-induced oxidative stress in disrupting alveolarization and vascular development has been extensively documented [2], with recent studies providing further insights into the epigenetic contributions to BPD pathogenesis [41]. Despite the growing interest in the epigenetic regulation of BPD, the availability of samples from preterm infants to investigate these mechanisms remains limited. Previous epigenome-wide association studies (EWAS) have identified DNA methylation loci associated with BPD in formalin-fixed lung samples from neonates with various morbidities [42]. Additionally, recent studies have highlighted changes in the cord blood methylome, including alterations in lung development, immune inflammation, and platelet activation, as potential early indicators of the risk factor of BPD [43]. Our multi-omics analysis highlighted pathways previously implicated in BPD, including p53, TNF, Ras, PI3K-Akt, MAPK, transcriptional misregulation in cancer, rheumatoid arthritis, olfactory transduction, neuroactive ligand–receptor interaction, hematopoietic cell lineage, focal adhesion, cytokine–cytokine receptor interaction, complement and coagulation cascades, cell adhesion molecules (CAMs), axon guidance, and arachidonic acid metabolism [44–55]. Notably, we observed minimal epigenetic changes on the sex chromosomes in hyperoxia-induced BPD, suggesting stable regulation of sex chromosomes under hyperoxic conditions [56]. Additionally, novel pathways were enriched in our rat BPD model, such as the cAMP signaling pathway, thyroid hormone signaling pathway, proximal tubule bicarbonate reclamation, protein digestion and absorption, mineral absorption, legionellosis, insulin resistance, aldosterone-regulated sodium reabsorption, adrenergic signaling in cardiomyocyte, and ABC transporters. These findings expand the current molecular landscape of BPD and underscore the complexity of hyperoxia-induced epigenetic remodeling.

It is worth noting that there are certain limitations of this study. First, while we identified Calca as a potential epigenetic regulator in BPD, we did not perform additional functional studies to establish a causal relationship between DNA (hydroxy)methylation and Calca transcription. Future studies should focus on the functional validation of Calca through gene knockdown or overexpression experiments to better clarify its role in hyperoxia-induced lung injury. Furthermore, the transcriptional and epigenetic changes observed in whole lung tissues likely reflect contributions from multiple cell types involved in BPD pathogenesis. For instance, the upregulation of Apln and Calca may primarily occur in endothelial cells, given their roles in vascular regulation [24, 38–40]. Similarly, the downregulation of Egfl7 and Tgfbi could predominantly affect alveolar epithelial cells, as these genes are essential for alveolar repair and surfactant production [18, 20]. Future studies employing single-cell RNA-seq and cell type-specific methylation profiling will be crucial to fully elucidate the cellular heterogeneity of these changes and identify precise therapeutic targets. Furthermore, the application of DNA methylation editing tools and cell type-specific models will be essential to establish causal relationships between 5mC/5hmC modifications at the Apln and Calca promoters and their functional roles in BPD pathogenesis.

Conclusions

In summary, this study suggests that hyperoxia promotes gene expression of key regulators, particularly Apln and Calca, via downregulating DNA methylation and upregulating DNA hydroxymethylation at promoter regions. Our results illustrate that DNA (hydroxy)methylation plays a vital role in the expression level of key regulators in hyperoxia-induced BPD, providing novel epigenetic regulation in BPD. Although the precise functional roles of Apln and Calca in BPD pathogenesis require further elucidation, these findings establish their potential significance as both diagnostic biomarkers and therapeutic targets for BPD.

Methods

Animal study

The animal experiment was carried out with the approval from The Institutional Animal Care and Use Committee of Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China (SIAT-IACUC-200824-CXY-A1391). Pregnant Sprague–Dawley rats obtained from BesTest (Zhuhai, China) were housed individually in cages and allowed to deliver their offspring full-term (22 days) with ad libitum access to laboratory food and water. Within 12 h after birth, rat neonates were randomly assigned to hyperoxia (O₂) or room air (RA) exposed groups (n = 8). The experimental model of BPD was induced as previously described [15]. Briefly, rat neonates in the hyperoxia group were exposed to 95% oxygen in a plexiglass chamber with an oxygen volume of 1.5 L/min from postnatal day 1 to day 10. Additionally, rat neonates in the RA-exposed group were maintained in ambient room air. On postnatal day 10, the rat neonates were sacrificed, and the lung tissues were collected and fixed in situ with 4% paraformaldehyde. Subsequently, the lung tissues were embedded in paraffin and sectioned for hematoxylin and eosin (H&E) staining and immunohistochemistry.

H&E staining and lung morphometry

The right lung tissues were collected and fixed in 4% paraformaldehyde. After a series of dehydration processes, the tissues were then embedded in paraffin and cut into 5 μm sections. The paraffin sections were subsequently stained with H&E, and images were captured under a light microscope (Leica, Germany). To assess the morphological and developmental status of the lung tissue, lung morphometry was evaluated using the mean linear intercept (MLI). Briefly, ten non-overlapping images of lung tissues were captured under a light microscope and later analyzed for the average alveolar diameter. To calculate the MLI, at least 1000 alveoli per animal were measured, excluding other structures such as vessels and airways. Alveoli with an area more than 100 μm² were analyzed to calculate the absolute alveolar diameter.

RNA sequencing and data analysis

Total RNA was extracted from lung tissues of eight hyperoxia-exposed and eight normoxia-control rat neonates using Trizol. The TruSeq RNA Sample Preparation Kit V2 (Illumina, America) was utilized to prepare cDNA libraries following the manufacturer’s instructions. Initially, oligo-dT magnetic beads were employed to enrich eukaryotic mRNA, and fragmentation reagent was added to obtain fragmented mRNA. The first-strand cDNA was synthesized using fragmented mRNA as a template with random hexamer primers. After the synthesis of the second-strand cDNA, the double-stranded cDNA was repaired with sticky ends and modified with adenylation at the 3′ end. The cDNA was then connected with the Illumina adenylation sequencing adaptor. The DNA fragments were size-selected (200–500 bp), and PCR amplification was conducted for fragment enrichment. The quality of the constructed cDNA libraries was evaluated using the Agilent 2100 Bioanalyzer together with ABI StepOnePlus Real-Time PCR System. Genome-wide transcript analysis was performed using the Illumina sequencing platform.

To obtain high-quality reads for transcript analysis, the Trimmomatic software was employed to truncate the sequence adaptors and screen the clean reads. These clean reads were aligned with the whole-rat reference genome using Hierarchical Indexing for Spliced Alignment of Transcripts [57]. The alignment results were further processed by SAMtools. The quantitation of transcript expression levels was performed using HTseq-count, and the differential gene expression analysis between groups was carried out using DESeq2 software [58]. Fragment Per Kilobase Per Million Fragment values were calculated for each gene for downstream analysis, and the differential genes were filtered and displayed using R language [59]. Furthermore, GO and KEGG enrichment analysis was conducted to obtain the biological function and enriched pathway.

Reduced representation bisulfite sequencing (RRBS) and oxidative reduced representation bisulfite sequencing (oxRRBS)

Genomic DNA was extracted from lung tissues (n = 8) using the DNeasy Blood & Tissue Kit (Qiagen), followed by library preparation using the Acegen Rapid RRBS Library Prep Kit (Acegen, Cat. No. AG0422) according to the manufacturer’s instructions. Briefly, 100–300 ng genomic DNA was digested with MspI and subjected to end-repair and 3′-dA-tailing before being ligated to a methylated Illumina linker. Half of the product was then subjected to bisulfite treatment and conversion for RRBS library construction, while the other half was selected for bisulfite conversion (TrueMethyl® oxBS Module) after oxidation of products in the 100–350 bp range for oxRRBS library construction. PCR amplification was carried out with Illumina 8 bp index primers for 12 cycles, and the amplified products were purified using AMPure XP. The fragment range of the libraries was detected by qPCR using Agilent 2100. The constructed RRBS and oxRRBS libraries were subjected to DNA sequencing on Illumina Nova 6000 using paired-end 150 bp sequencing strategy.

Analysis of RRBS and oxRRBS data

Initial quality assessment of raw sequence reads was conducted utilizing FastQC, followed by adapter trimming and quality filtering through Trimmomatic software [60]. The bisulfite-converted sequences were then aligned to the whole rat reference genome using BSMAP software with defined parameter. Only unique map readings with a sequence depth coverage of methylated cytosine higher than 5 were retained for subsequent data analysis. The methylation level (ML) of individual cytosines was quantified using the ratio of the methylated CpG cytosine sequencing depth (mC) to total sequencing depth of individual cytosines (mC + umC), where ML = mC / (mC + umC), with mC and umC representing the number of reads supporting methylation and unmethylated cytosine, respectively. The methylation rate of RRBS and oxRRBS libraries was calculated, and the difference between the methylation rates of the two libraries was considered the hydroxymethylation rate.

Differentially methylated region (DMR) and differentially hydroxymethylated region (DhMR) were identified using Metilene software [61]. A DMR/DhMR was required to contain at least five cytosine regions, and the distance between adjacent cytidine regions was less than 200 bp. CG-DMR/hCG-DhMR candidate regions were defined only if the average methylation level difference between the corresponding populations was more than 0.1. The regions with a P value < 0.05 in the 2D KS-test and Benjamini & Hochberg (BH) corrected P value < 0.05 were determined as final DMRs/DhMRs. GO analysis was performed to reveal biological processes of potential genes from DEG, DMRs, and DhMRs. Pathway enrichment was considered significant at Q value ≤ 0.05. In addition, based on KEGG database and DEGs, DMRs, and DhMRs annotation results, functional enrichment analysis was performed on genes with promoter regions overlapped with DMRs and DhMRs [62].

Targeted bisulfite sequencing (TBS) and oxidative targeted bisulfite sequencing (oxTBS)

Bisulfite and oxidative bisulfite conversion of genomic DNA (n = 8) was performed using the TECAN TrueMethyl oxBS Module kit according to the manufacturer’s protocol. Briefly, 0.5 μg genomic DNA was purified and denatured, and then divided into two equivalent reaction aliquots. One aliquot underwent chemical oxidation followed by bisulfite conversion, while the other underwent mock oxidation with water followed by bisulfite conversion. The bisulfite and oxidative bisulfite-converted DNAs were subjected to purification and eluted in 20 μL buffer. Gene-specific DNA methylation was evaluated by bisulfite PCR coupled with next-generation sequencing according to previously established method [63]. In brief, bisulfite or oxidative bisulfite-converted DNA (10 μL) underwent 35-cycle amplification utilizing bisulfite sequencing primers (Table S1) and KAPA ReadyMix PCR Kit (Kapa Biosystems). For each sample, PCR products of multiple genes were equally pooled, modified with 5′-phosphorylation and 3′-dA-tail, and ligated to barcoded adapter using T4 DNA ligase (NEB). Barcoded libraries from all samples were finally sequenced on the Illumina platform. TBS/oxTBS and RRBS/oxRRBS analyses were performed on anatomically distinct lung regions from the same animals to minimize inter-individual confounding.

Analysis of TBS and oxTBS data

Raw sequence data underwent quality control processing utilizing Trimmomatic software, implementing a sliding window approach for adapter removal and quality filtration. Sequences were truncated when the mean base quality within the window fell below a threshold of 15. Processed sequences were aligned to target amplicon sequences utilizing BSMAP software with defined parameters. The methylation levels of the CG site for the TBS or oxTBS library were calculated using a python algorithm in BSMAP software. The methyl value was calculated using the following formula: Methyl value = C-reads / (C-reads + T-reads) × 100%, where C-reads and T-reads represented sequence reads supporting and not supporting methylation, respectively. Furthermore, the hydroxymethylation level of cytosine bases was also calculated using the maximum likelihood estimation method (MLML) [64]. Finally, the statistical and visual display of the methylation and hydroxymethylation levels was performed using the R language. Differential analysis between two groups was conducted using a two-tailed t tests, with significance criteria established as absolute methylation level difference > 0.1 and P < 0.05.

Real-time qPCR

Total RNA was isolated from the lung tissues (n = 9) utilizing TRIzol reagent. RNA concentrations were subsequently assessed using a NanoDrop. Total RNA was later reverse transcribed into complementary DNA (cDNA) by PrimeScript™ RT Master mix (Takara Bio, Inc.). Quantitative PCR (qPCR) analysis was conducted using the SYBR Green Premix Pro Taq HS qPCR kit (Accurate Biotechnology Co., Ltd.) on a LightCycler 480 Real-Time PCR System. The thermal cycling parameters consisted of initial denaturation at 95 °C for 30 s, followed by 40 amplification cycles of denaturation at 95 °C for 5 s and combined annealing/extension at 60 °C for 30 s. The relative expression levels of Apln and Calca were normalized to the expression of the housekeeping gene (β-actin) using the 2^–ΔΔCt methodology. The primer sequences used for qPCR were shown in Table S2.

Western blotting

Total proteins were extracted from lung tissues (n = 3) using RIPA lysis buffer supplemented with phosphatase and protease inhibitor cocktail (Thermo Fisher Scientific Inc.). Protein concentrations were quantified via the bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific Inc.). Protein lysates (30 μg) were subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and subsequently electrotransferred onto polyvinylidene fluoride (PVDF) membranes (Millipore). The membranes were blocked with 5% non-fat milk in Tris-buffered saline containing 0.1% Tween-20 (TBST) for one hour at room temperature, followed by overnight incubation at 4 °C with the following primary antibodies: anti-Apln antibody (1:1000 dilution, DF13350, Affinity), anti-Calca antibody (1:500 dilution, DF7386, Affinity), and anti-β-actin antibody (1:1000 dilution, 8H10D10, Cell Signaling Technology). After washing with TBST, the membranes were incubated with horseradish peroxidase (HRP)-conjugated goat anti rabbit antibody (1:5000 dilution, ZB-2301, GSJB-BIO) or HRP-conjugated goat anti-mouse antibody (1:5000 dilution, ZB-2305, GSJB-BIO) for 60 mins. Eventually, the bands were visualized using Western Lightning® Plus-ELC kit (PerkinElmer lnc.) and documented using the Amersham Imager 600 (Amersham Biosciences). Densitometric analysis of each protein band was performed by Image J software, and target protein expression levels were normalized to β-actin.

Immunohistochemistry

The paraffin sections (5 μm) of lung tissue were deparaffinized with xylene and then subjected to sequential rehydration process by gradient ethanol. To block endogenous peroxidase activity, the sections were treated with 3% hydrogen peroxide in methanol for 10 min at room temperature. For antigen retrieval process, treated sections were placed into 10 mM citric acid buffer (pH 6.0) and then microwaved for 20 min. After blocking with serum, the sections were subjected to overnight incubation at 4 °C with primary antibodies: anti-Apln antibody (1:1000 dilution, DF13350, Affinity), anti-Calca antibody (1:100 dilution, DF7386, Affinity). Subsequently, the sections were incubated with biotinylated secondary antibodies and streptavidin peroxidase in a humidified chamber for 30 mins at room temperature. The sections were later developed with freshly made 3,3′-diaminobenzidine (DAB) substrate solution, followed by counterstaining with hematoxylin, dehydration and mounting. Eventually, image acquisition was performed using a light microscope (Leica).

Statistical analysis

Quantitative data were presented as mean ± standard deviation (SD). Statistical comparisons between two groups were performed using an unpaired two-tailed Student′s t test by GraphPad software. Statistical significance was established when P value was less than 0.05, with significance levels denoted as follows: *P < 0.05, **P < 0.01 and ***P < 0.001.

Abbreviations

BPD

Bronchopulmonary dysplasia

ROS

Reactive oxygen species

5mC

5-Methylcytosine

5hmC

5-Hydroxymethylcytosine

O₂

Hyperoxia

RA

Room air

RRBS

Reduced representation bisulfite sequencing

oxRRBS

Oxidative reduced representation bisulfite sequencing

TBS

Targeted bisulfite sequencing

oxTBS

Oxidative targeted bisulfite sequencing

MLI

Mean linear intercept

PCA

Principal component analysis

DEGs

Differentially expressed genes

DMRs

Differentially methylated regions

DhMRs

Differentially hydroxymethylated regions

HCM

Hypertrophic cardiomyopathy

DCM

Dilated cardiomyopathy

CAMs

Cell adhesion molecules

Author contributions

H.-T.L., T.Q., H.-M.Z., and Q.-H.L. contributed to study designation and conduction. Y.-Y.M., Y.-F.L., Z.-L.H., Y.-Y.D., and Y.-Y.X. performed data collection and analysis. H.-T.L., X.-Y.J., and D.-S.H. performed manuscript writing. L.S., X.-Y.C., and X.-Y.J. contributed to supervision, study designation, manuscript editing, and funding acquisition. All authors read and approved the final manuscript.

Funding

This work was supported by the Project of Educational Commission of Guangdong Province of China (2024KTSCX119), the Tertiary Education Scientific research project of Guangzhou Municipal Education Bureau (2024312127), the Open research fund from Guangdong Provincial Key Laboratory of Molecular Target and Clinical Pharmacology (J24413004 and J24413003), the National Natural Science Foundation of China (82300471), the Medical Scientific Research Foundation of Guangdong Province (B2025502), Sanming Project of Medicine in Shenzhen (SZSM202211001) and Shenzhen Fund for Guangdong Provincial High Level Clinical Key Specialties (SZGSP009), and Shenzhen Key Laboratory of Maternal and Child Health and Diseases (ZDSYS20230626091559006).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Consent for publication

All authors have agreed to publish this manuscript on Clinical Epigenetics.

Competing interests

The authors declare no competing interests.