Genome-wide DNA methylation analysis in pigs using long-read sequencing unveils high-altitude adaptation and allele-specific regulation

Abstract

Background

The advent of third-generation sequencing, particularly Oxford Nanopore Technologies (ONT), has revolutionized epigenetic studies by enabling direct detection of DNA methylation modifications and single-base resolution profiling of methylation patterns. While this technology has been predominantly utilized in human and bacterial research, its applications in livestock and poultry remain limited. In this study, we employed ONT sequencing to construct comprehensive 5-methylcytosine modification maps for ten representative pig breeds, explored the mechanism of high altitude adaptive methylation and allele-specific methylation events in these pigs.

Results

Through genome-wide integration of sequencing data, we identified 27,857,021 CpG sites, with 71.5% (19,836,456) shared across pigs. Comparative differential methylation analysis between high-altitude and low-altitude pigs revealed four candidate genes (CALM1, HBB, PRKCQ and RAMP1) and the Sp1 transcription factor as potential key regulators of hypoxic adaptation. Notably, Tibetan pigs exhibited promoter hypomethylation patterns at the CALM1 locus, correlating with its consistently elevated expression confirmed by public transcriptomic databases. Allele-specific methylation (ASM) analysis integrated with transcriptomic profiles demonstrated significant enrichment of ASM events in promoters or exons of allele-specific expression (ASE) genes, suggesting synergistic regulatory mechanisms between epigenetic modifications and allelic expression patterns.

Conclusions

Our results provided a high-resolution DNA methylation atlas based on long-read sequencing encompassing ten representative pigs across Eurasia, and identified hypoxic adaption-related genes (CALM1, etc.) and a transcription factor (Sp1) based on the unique physiological characteristics of Tibetan pigs. Further, combined with transcriptome data, it was demonstrated ASM and ASE events are synergistic and expressions of ASE genes may be regulated by ASM. This study offers valuable insights into the epigenetic mechanisms underlying adaptation and gene regulation in pigs.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13117-8.

Background

Native to the Qinghai-Tibet Plateau at an average elevation above 3000 m [1, 2], Tibetan pigs have successfully adapted to extreme high-altitude environments characterized by hypobaric hypoxia and enhanced ultraviolet radiation. As a crucial epigenetic regulatory mechanism, DNA methylation represents a key mechanism epigenetic underlying high-altitude adaptation [3, 4]. Nevertheless, its biological functions roles remain poorly elucidated, particularly in pigs. In the mammalian genome, methylation at the carbon number 5 of cytosine (5-methylcytosine, or 5mC) is the most predominant DNA modification, which is also called the “fifth base” [5]. Under the action of DNA methyltransferases (DNMTs), the 5'-cytosine of symmetrical CG dinucleotide was converted into 5mC. 5mC plays an important role in key biological processes and disease occurrence, such as cell differentiation, aging, gene imprinting, carcinogenesis and metabolic diseases [6–8], etc. Therefore, constructing genome-wide high-resolution DNA methylation maps is critical for comprehensively characterizing biological functional diversity, elucidating epigenetic-driven pathological processes (including aberrant gene expression and disease pathogenesis), and deciphering developmental mechanisms in species evolution as well as the underlying basis of trait formation.

Bisulfite sequencing (BS-Seq) [9] and Methylated DNA Immunoprecipitation Sequencing (MeDIP-seq) are two widely used methods for 5mC detection, but they are limited by short read lengths and difficult in analyzing repetitive sequences. In contrast, third-generation sequencing technologies, such as PacBio Single-Molecule Real-Time (SMRT) sequencing [10] and Oxford Nanopore Technologies (ONT) sequencing [11], offer long-read capabilities and single-base resolution, enabling direct detection of DNA modifications, including 5mC, even in repetitive regions. While Oxford Nanopore sequencing technology has been extensively utilized in methylation analysis across various biological studies [12, 13], its application in porcine genomic research remains relatively underexplored [14], particularly in the context of epigenetic characterization for Chinese indigenous pigs.

It is well known that pigs are an ideal biomedical model due to anatomical, genetic and physiological similarities with human [15]. In addition to providing new information for biomedical research, epigenomic studies in pigs can help reveal the molecular basis of economic traits or other traits in pigs [16, 17], which can be used to improve the efficiency of artificial selection. More than one-third of the domesticated pig breeds inhabit China (https://www.moa.gov.cn/). Distinct local breeds exhibit substantial phenotypic divergence attributable to geographical isolation and environmental adaptations, establishing a valuable foundation for investigating epigenetic regulatory divergence underlying trait variation. Current research on DNA methylation in pigs is relatively limited, particularly comparative methylome studies focusing on representative Chinese local breeds. Furthermore, the reliance on second-generation sequencing technologies presents a technical constraint, as these methods failed to detect certain methylation sites.

This study represents comprehensive exploration of the feasibility of ONT sequencing for DNA methylation profiling in pigs. We systematically characterized the DNA methylation landscapes across various pigs, elucidating both shared and divergent methylation patterns among different populations. Notably, our investigation of Tibetan pigs, with their distinctive physiological adaptations, led to the identification of differentially methylated genes potentially associated with hypoxia adaptation. Furthermore, through integrative analysis of allele-specific methylation (ASM) and allele-specific expression (ASE), we uncovered novel insights into the genetic regulation of DNA methylation and its impact on transcriptional control. These findings establish a robust theoretical framework for advancing pig germplasm resource utilization and facilitating the exploitation of superior genetic traits in pig breeding programs.

Methods

Biological samples

A total of ten male individuals from ten pig breeds (Anqingliubai, Laiwu, Meishan, Min, Rongchang, Tibetan, Wuzhishan, CW (Chinese wild), EW (European wild) and Yorkshire) were sampled across Eurasia. These Chinese native pigs cover six geographic types, including the North China, South China, Central China, East China, Southwest China, and Plateau of China. Peripheral blood were collected from live pigs using 5 mL K2-EDTA vacutainer blood collection tube (Cat# YA1300, Solarbio, Beijing, China) and mixed by inversion. The tubes were placed on dry ice and then stored at − 80 °C freezer until DNA extraction.

DNA extraction and sequencing

High-molecular-weight DNA was extracted based on the CTAB method [18] and purified with Blood & Cell Culture DNA Kit (Cat# 13,343, Qiagen, Beijing, China) following the manufacturer’s protocols. The concentration and purity of DNA were detected using NanoDrop™ (Termo Fisher Scientifc, Waltham, Massachusetts, USA) and Qubit Fluorometer (Invitrogen, Carlsbad, California, USA) with optical density 260/280 ranging from 1.8 to 2.0 and 260/230 ranging from 2.0 to 2.2. The integrity of DNA was detected using pulsed field electrophoresis. And the large DNA segments (> 20 kb) were retrieved using the BluePippin™ System (Sage Science, Beverly, MA) by gel cutting.

Nanopore Sequencing: Sequencing library preparation (20-kb) was carried out following the general guidelines from Oxford Nanopore Technologies using ONT Template prep kit (SQK-LSK109) and NEB Next FFPE DNA Repair Mix kit. The high-quality libraries were sequenced on the ONT PromethION platform with Corresponding R9.4.1 pore flow cell and ONT sequencing reagents kit (EXP-FLP001.PRO.6). The real-time sequencing data were collected and base calling was performed using Guppy (version 1.5) software.

Short-read sequencing and single nucleotide polymorphism (SNP) calling: The sequencing libraries (150 bp paired-end with insert sizes of 350 bp) were constructed following standard Illumina library preparation protocols by Beijing Biomarker Biotechnology Co., Ltd., using the Illumina HiSeq X platform. Then, the clean reads were aligned to pig reference genome Sscrofa11.1 using BWA-MEM [19] (version 0.7.17-r1188). SAMtools [20] (version 1.9), Picard SortSam and Picard MarkDuplicates ((http://broadinstitute.github.io/picard/) were used for file format converting, sorting and marking potential PCR duplicates, respectively. The Genome Analysis Toolkit [21] (GATK, v3.7) best practice was used for SNP-calling in HaplotypeCaller mode. The hard filters with “QD < 2.0 || FS > 60.0 || MQRankSum < −12.5 || ReadPosRankSum < −8.0 || MQ < 40.0” to obtain a high-confidence SNP set.

5mC modification detection and filtering

Minimap2 (version 2.28) [22] with the setting minimap2 –ax map-ont was used to align basecalled reads to Sscrofa11.1 reference genome. Subsequently, Nanopolish (version 0.13.2) [23] with the default parameters was used to build an index mapping from basecalled reads to the signals measured by the sequencer and classify nucleotides as methylated or not. The bam and index files were generated using SAMtools (version 1.9) [20]. Nanopolish uses the raw nanopore signal and the alignment of reads to the reference genome to calculate, for each read covering a CpG site, a log-likelihood ratio, comparing the methylated and unmethylated models. For each CpG site, the methylation status of CpG sites was determined according to the log-likelihood ratio as follows: (1) CpG site with log-likelihood ratio > 2.5 was classified as methylated; (2) CpG site with log-likelihood ratio < −2.5 was classified as unmethylated.

The methylation level at each CpG site was calculated as:

where is the number of reads classified as methylated and is the number of reads classified as unmethylated at that CpG site. To ensure reliable estimation, only CpG sites covered by at least 10 valid reads were retained for downstream analyses. For comparisons among samples, shared CpG sites were defined as CpG sites that were covered by at least 10 valid reads in each sample included in the comparison.

Construction of genome-wide 5mC map

To systematically investigate DNA methylation dynamics in different genomic regions, we performed methylation quantification across distinct genomic regions. Gene body regions (transcription start site (TSS) to transcription termination site (TTS)) and adjacent 10-kb flanking regions were partitioned into 20 and 10 equidistant bins, respectively, using CGmaptools (version 0.1.2) [24] for genome-wide methylation trend analysis. In addition, the methylation levels of CpG islands (CGI) and their flanking regions (CGI shores and CGI shelves) were also calculated. CGI shores were defined as the genomic regions located within 0–2 kb upstream or downstream of CGI, whereas CGI shelves were defined as the regions located 2–4 kb upstream or downstream of CGI. Off-CGI refers to genomic regions outside CGI, CGI shores, and CGI shelves. The CGI regions were downloaded from UCSC Table Browser.

The methylation level of genomic regions (exons, introns, 5’ UTRs, 3’ UTRs, promoters and intergenic regions) was determined using the formula: Methylation level = (Number of methylated CpG sites)/(Total number of covered CpG sites) × 100%.

Detection of differentially-methylated regions between Tibetan pig and other Chinese indigenous pigs

The DSS R package [25] was used to identify differentially-methylated loci (DML) and differentially methylated regions (DMRs). Considering the spatial correlation of methylation sites, smooth method [26] was applied to estimate average methylation levels. The significance of DML between the two groups was determined by Wald's test (FDR < 0.05), and the minimum absolute value of methylation difference between the two groups was 0.3 [27]. We established stringent DMR calling criteria through DML integration: significance threshold (FDR < 0.05), minimum genomic span (≥ 100 bp), CpG density requirement (≥ 10 sites), inter-region spacing (≥ 150 bp), and CG content threshold (≥ 50% per DMR). Differentially methylated genes (DMGs) within DMRs were annotated using the Biomart database. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) enrichment analyses were performed for further gene function analysis using DAVID Bioinformatics Resources 6.8 [28].

To elucidate the mechanisms of high-altitude hypoxia adaptation, we conducted a promoter motif enrichment analysis on the DMRs identified between high-altitude (Tibetan) and low-altitude pig breeds (Anqingliubai, Laiwu, Meishan, Min, Rongchang, and Wuzhishan). Wild boars and European domestic pigs were excluded from this analysis to avoid extra noise caused by evolutionary divergence. We screened the DMRs located in promoter regions and divided DMRs into two groups, one with high methylation levels in highland group (called H-DMRs) and the other with high methylation levels in lowland group (called L-DMRs). Motif analysis using Homer [29] was performed on the two types of DMRs respectively to find transcription factors that could reveal the differences in methylation levels between the two groups.

Detection of ASM

The core methodology for ASM detection using ONT sequencing leverages haplotype-phased reads partitioning (reference (HP1) vs. alternative (HP2) alleles), enabling concurrent assessment of allele-specific methylation divergence through comparative analysis of haplotype-resolved methylation patterns [30]. The Whatshap [31] was used to phase heterozygous SNPs from short-read sequencing. Then, the phased VCF files, the methylation call files and alignment files obtained from the ONT data were supplied to NanoMethPhase (version 1.2.0) [30] to phase long reads (divided into HP1 and HP2) and CpG methylations. Specifically, in the phased VCF generated by WhatsHap, the two alleles at each heterozygous SNPs were assigned to HP1 or HP2 according to the phased genotype. For each long read, we counted the number of phased SNPs alleles supporting HP1 and HP2 and further considered their base qualities. A read was assigned to HP1 if it satisfied the following criteria: (1) the read overlapped at least two phased heterozygous SNPs within the same phase block; (2) the number of HP1-supporting phased SNP alleles was greater than the number of HP2-supporting alleles; (3) the haplotype-support ratio, defined as the number of HP2-supporting alleles divided by the number of HP1-supporting alleles, was less than 0.7; (4) the average base quality of HP1-supporting alleles was higher than that of HP2-supporting alleles, or the number of HP1-supporting alleles with base quality was higher than that of HP2-supporting alleles. Conversely, a read was assigned to HP2 using the symmetrical criteria, namely when the number, ratio, and quality evidence of HP2-supporting alleles were stronger than those of HP1.

DMRs between the two haplotype pairs were identified by NanoMethPhase dma module based on DSS R package using dispersion shrinkage method. To avoid the interference of the X-inactivation, we identified ASM only on autosomes [32].

Enrichment of ASM

To further explore the distribution and functional characteristics of ASM. First, we investigated the distribution of ASM in different genomic regions of the whole genome, including standard positional regions (promoter, exon, intron, 5’UTRs, 3’UTRs and intergenic regions), imprinted gene regions and CpG islands. The information of genomic regions was downloaded from Ensembl (release 104). Promoters were defined as the regions from upstream 1.5 kb to downstream 0.5 kb of TSS. Imprinted genes were obtained from the Geneimprint website (https://www.geneimprint.com/). Then the fold enrichment of ASM in each individual for each genomic element (e.g., promoter and exon) was calculated by (C/A)/(B/D), where A, B, C, D are the number of bases in ASM, the number of bases in a genomic element, the number of overlapping bases between ASM and a genomic element, and the number of bases in the genome, respectively. In addition, we mapped the ASM to the 15 chromatin states annotations of pigs identified by Pan et al. [33] to revealed the enrichment characteristic of ASM.

Identification of genes with ASE in relevant tissues

We used the ASE and non-ASE sites identified in the published study by Yao et al. [34], which analyzed 151 whole-blood RNA-seq samples primarily from Tibetan and Yorkshire pigs, etc. In that study, ASE was identified by using phased VCF files derived from imputed RNA genotype data to distinguish the two haplotypes and quantify allele-specific expression. Genes were classified as mono-allelically expressed genes if they harbored at least one site with a significant allelic imbalance (FDR < 0.05). Only genes with consistent allelic imbalance directions were retained as mono-allelically expressed genes. Conversely, genes were classified as bi-allelically expressed genes if they contained a site with no significant allelic imbalance (P value > 0.2) [35]. To identify tissue-associated mono-allelically expressed genes and bi-allelically expressed genes, the frequency of occurrence: mean + 1SD was selected as the threshold, and a gene was identified as the two types of genes in blood tissue when the occurrence frequency of the gene in all individuals was greater than the threshold value. For mono-allelically expressed genes, the number of samples in which mono-allelic expression was detected was first counted, and the mean and standard deviation of the gene-level occurrence counts were then calculated. The same procedure was applied to bi-allelically expressed genes. Blood-associated mono-allelically and bi-allelically expressed genes were defined using an occurrence-frequency threshold of mean + 1SD. If genes with occurrence counts exceeding this threshold across all individuals were classified as blood-associated mono-allelically or bi-allelically expressed genes, respectively.

Detection of AMRs within imprinted regions

To evaluate the accuracy of ASM identification, we identified genes containing ASMRs (allele specific methylation regions), defined as ASMGs (allele specific methylation genes). Meanwhile, we used Chi-squared test (P < 0.01) to determine whether ASMGs was significantly overlap with imprinting control regions.

Results

Summary of ONT data and annotation of CpG sites

Oxford Nanopore Technologies sequencing analysis of ten pigs yielded 8.29 million reads (146.7 Gb mean throughput), achieving 58.6 × average coverage (SD = 8.9 ×) (see Additional file 1: Table S1) with 27,857,021 autosomal CpG sites (coverage ≥ 10), including 19,836,456 shared CpG loci (71.5%) (Fig. 1a, b). By annotating genome-wide CpG sites, we found that CpG sites in introns, exons, and intergenic regions accounted for 50%, 8% and 32%, respectively (Fig. 1c). Combining CGI annotation, it was found that most of CpG sites were located off the CGIs, but the density of CpG sites in the CGI regions was higher (Fig. 1c). Genome-wide methylation profiling identified hypermethylation across all chromosomes, contrasting with mitochondrial DNA hypomethylation (0.0227–0.0341) (Fig. 1d). In addition, we analyzed the proportion of reference genome CpG sites covered, as well as the number of CpG sites gained or lost due to SNPs. As shown in the Table 1, the autosomal CpG calling rate varied among breeds, ranging from 85.00% in Tibetan pigs to 98.69% in Meishan pigs. Regarding SNP-induced changes, we observed substantial breed-specific variation in both CpG site loss and de novo CpG formation.

Fig. 1.

Basic information of CpG sites across pigs. a The proportion of shared CpGs and variable CpGs. b The circos plots of the density of CpG sites on autosomes. The outermost layer of both circos plots (left and right) shows the CpG distribution density (100-kb windows) of the reference genome. c Distribution (left and middle) and density (right) of shared and reference-genome CpG sites across genomic features, CpG islands, and their flanking regions. CGI, CpG islands; CGI shores, genomic regions located within 0–2 kb upstream or downstream of CGIs; CGI shelves, regions located 2–4 kb upstream or downstream of CGIs; off-CGI, genomic regions outside CGIs, CGI shores, and CGI shelves. d Characteristics of methylation levels at chromosome levels

Table 1.

Summary of CpG site statistics across pig breeds

Breed	Total CpG sites	CpG sites on autosomes	CpG sites lossa	De novo CpG sitesb	CpG calling rate on autosomesc
Anqingliubai	27,464,835	26,442,954	1,725,935	1,710,341	97.01%
Laiwu	25,240,997	24,458,529	1,703,673	1,700,280	89.73%
Meishan	28,050,553	26,903,070	1,728,186	1,730,791	98.69%
Min	25,685,583	24,948,723	1,653,089	1,654,567	91.52%
Rongchang	26,147,801	25,293,542	1,930,965	1,898,948	92.79%
Tibetan	23,764,857	23,169,193	1,990,137	1,939,601	85.00%
Wuzhishan	27,057,671	26,178,343	2,005,569	1,953,043	96.04%
Yorkshire	26,106,120	25,294,273	1,142,744	1,160,766	92.79%
EW	26,896,925	25,985,443	1,060,494	1,072,666	95.33%
CW	27,933,907	26,879,943	2,500,546	2,422,544	98.61%

^aCpG sites loss: the number of CpG sites lost due to SNPs

^bDe novo CpG sites: the number of CpG sites arising from SNPs

^cThe CpG calling rate on autosomes is defined as the proportion of known CpG sites on the autosomes that were successfully detected by third-generation sequencing compared to the reference genome

Shared and heterogeneous patterns of DNA methylation across pigs

After quality control, we obtained 23.76 to 28.05 million CpG sites (Table 1). These sites exhibited three key characteristics: (1) Uneven chromosomal distribution with telomeric and CGI enrichment (Fig. 1b); (2) Predominant hypermethylation (58.5–75.9% methylation states) with limited hypomethylated (≤ 20%) or intermediate (20–40%) sites (Fig. 2a); (3) Dynamic methylation landscapes showing elevated gene body methylation versus flanking 5-kb regions, with striking TSS-associated hypomethylation (Fig. 2b). CGI cores demonstrated depressed methylation relative to progressively increasing levels in adjacent shores and shelves (Fig. 2b).

Fig. 2.

Genomic methylation information in pigs. a The proportion of CpG sites at different methylation levels. b The dynamics of DNA methylation across genes and CpG islands and their flanking regions. c Genome-wide methylation levels across individuals. d The correlation heat map of methylation in individual level

Genome-wide CpG methylation levels ranged from 66.36% to 77.14% (Fig. 2c). Methylation hierarchy of genomic elements remained consistent: exons > 3'UTRs > introns > promoters > 5'UTRs (see Additional file 2: Figure S1), with promoters showing significantly reduced methylation compared to enhancers (see Additional file 2: Figure S2). Shared CpG loci exhibited strong correlation (Pearson's r = 0.70–0.88), with higher methylation concordance was observed within domestic pigs (Fig. 2d).

Differentially-methylated regions between Tibetan pig and other Chinese indigenous pigs

To investigate epigenetic adaptations to high-altitude environments in Tibetan pigs, we performed comparative methylome analyses against lowland counterparts. A total of 62,767 significant different methylation loci (DML) (FDR < 0.05) in the autosome were identified using dispersion shrinkage method. The absolute mean difference of DML between the two groups was 45.61%. By merging DML, 328 different methylation regions (DMRs) were obtained (see Additional file 3: Table S4). The average length and the average CpG number of DMRs was 369 bp and 20, respectively. These DMRs covered 206 genes (see Additional file 3: Table S5), including 83 hypermethylated genes and 123 hypomethylated genes of Tibetan pigs. Functional enrichment analysis identified GO terms and KEGG pathways related to energy metabolism, such as ATP binding, GTPase activator activity and protein kinase activity, et al (Fig. 3a). Moreover, some known hypoxic-response-related pathways [36, 37] were significantly enriched, including the vascular smooth muscle contraction and protein digestion and absorption.

Fig. 3.

Differential methylation analysis between high-altitude and low-altitude pigs. a KEGG enrichment results of DMGs between high-altitude and low-altitude pigs. b Differential methylation levels for the four DMGs in two different phenotypes (high-altitude and low-altitude pigs) in public datasets. c Expression of CALM1 in four tissues of Tibetan pig and other pigs in public datasets (Expression data sourced from the PIGOME database: Pigome: http://pigome.com/). d Expression of Sp1 factor in heart and lung tissues of Tibetan pig and other pigs (expression data sourced from the PIGOME database: Pigome: http://pigome.com/)

Our analysis identified hypoxia-associated differentially methylated genes, including Calmodulin 1 (CALM1), Protein kinase C theta (PRKCQ), and Receptor activity-modifying protein 1 (RAMP1) in vascular smooth muscle contraction pathway and Hemoglobin Subunit Beta (HBB) gene (Fig. 3b and see Additional file 2: Figure S3). These genes exhibited Tibetan-specific methylation patterns, suggesting their epigenetic regulation may critically mediate high-altitude adaptation, thus meriting focused investigation into their hypoxia-responsive molecular mechanisms. Notably, a DMR was identified in the CALM1 promoter (see Additional file 2: Figure S4). Integration with multi-tissue transcriptomic datasets revealed CALM1 upregulation in Tibetan pigs compared to low-altitude pigs, with pronounced expression in skeletal muscle, cardiac tissue and hypothalamus tissue, contrasting with pulmonary downregulation (Fig. 3c).

Motif analysis provided a key clue to elucidate the transcriptional regulation mechanism of differential methylation regions. By annotating the DMRs, a total of 25 DMRs that overlap the promoter regions were obtained, including 14 H-DMRs and 11 L-DMRs (see Additional file 1: Table S2). H-DMRs revealed enrichment of the Sp1 transcription factor, which binds GC-rich motifs to regulate genes governing cell proliferation, apoptosis, differentiation, and immune responses [38, 39]. Notably, public available database (http://pigome.com/) showed that Sp1 expression was lower in Tibetan pigs than in low-altitude breeds (Fig. 3d). These findings suggest a potential association between Sp1 and the observed methylation differences.

Detection of allele-specific methylation

We implemented allele-specific methylation detection by integrating long-read methylation calls, phased alignment files, and heterozygous SNP-containing VCFs, the imprinted gene was used to verify the accuracy of the ASM identification (see Additional file 2: Figure S5). In all samples, more than 6.45 million reads were labeled by at least one phased SNPs, and on average, 70.3% (M ± SD, 4.52 M ± 1.40 M) of the pig methylome could be assigned to one of two haplotypes (reference haplotype and alternative haplotype).

In our study, an average of 136.4 k ASM sites were identified in each individual, accounting for 0.40% ~ 0.63% of CpG sites (see Additional file 1: Table S3). We detected 99,567 ASM events that were present in at least one-third of the individuals (≥ 33%; FDR < 0.05). To explore the distribution characteristics of ASM across the genome, we calculated the enrichment of ASM in genomic elements and 15 chromatin states. The results showed that ASM events were significantly enriched in promoters (1.5-fold), exonic regions (1.2-fold), imprinted genes (6.9-fold), and CpG islands (5.1-fold) (chi-square test, P < 0.001) (Fig. 4a). A total of 1,619 ASM loci were shared across Chinese indigenous porcine populations (Fig. 4b; see Additional file 2: Figure S6). Meanwhile, we mapped the ASM to the chromatin states defined by the Pan et al. annotation [33], revealed a significant enrichment in poised TSS, weak transcribed at gene, flank active TSS without ATAC and strongly active promoters (Fig. 4c).

Fig. 4.

The results of ASM. a Enrichment of ASM in various genomic elements. **P < 0.01, ***P < 0.001, chi-square test. b Manhattan plot of the genome-wide landscape of ASM for blood samples. The red highlighted mark represents the common ASM shared by all individuals. c ASM enrichment in 15 chromatin states. (E1: Active TSS; E2: Flanking active TSS without ATAC; E3: Transcription at gene 5’ and 3’; E4: Weak transcribed at gene; E5: Transcribed region with ATAC; E6: Active enhancer; E7: Medium enhancer with ATAC; E8: Weak active enhancer; E9: Active enhancer without ATAC; E10: Poised enhancer; E11: ATAC islands; E12: Poised TSS; E13: Repressed Polycomb; E14: Weak repressed Polycomb; and E15: Quiescent/Low). d The ASM enrichment levels of promoters in mono-allelically and bi-allelically expressed genes in blood tissue. *P < 0.05, **P < 0.01, ***P < 0.001, chi-square test. e The ASM enrichment levels of exons in mono-allelically and bi-allelically expressed genes in blood tissue. ns: not significant, ***P < 0.001, chi-square test. f The heatmap represents methylation levels of phased CpG sites shown separately by haplotype for imprinted genes. Methylation levels gradient from high (red) to low (green), while grey-filled cells indicate imprinted genes that were not identified in the corresponding breed

To investigate whether ASM plays a role in allelic-specific expression, we integrated ASM and ASE for analysis, using RNA-seq data from blood samples to identify ASE genes. Based on our selection criteria in Methods, there were 946 mono-allelically expressed genes (see Additional file 3: Table S6) and 2090 bi-allelically expressed genes (see Additional file 3: Table S7) in blood tissue. In both promoter and exon regions, ASM showed significant depletion (P < 0.001) in bi-allelically expressed genes in almost all individuals, with ASM enrichment factors in ASE genes being much higher than those in bi-allelically expressed genes, except for Yorkshire (Fig. 4d, e). Together, these results demonstrate a significant correlation between ASM and ASE, providing direct evidence for their association in pig blood tissue.

Based on ASM, we identified several imprinted genes. Some imprinted genes (e.g., IGF2R, KCNQ1) were identified based on allelic DMR (Fig. 4f and see Additional file 2: Figure S5). Additionally, we analyzed the overlap between ASM genes (ASMGs) and known imprinted genes in each pig. As expected, imprinted genes significantly overlapped with ASMGs in all pigs (Table 2).

Table 2.

Overlap between ASMGs and known imprinted genes

Samples	Allelic DMRs	ASMG	Overlapa
Anqingliubai	790	472	13**
Laiwu	782	495	8**
Meishan	612	376	8**
Min	735	431	8**
Rongchang	731	446	14**
Tibetan	798	500	15**
Wuzhishan	701	446	15**

^aThere are 36 known imprinted genes in pig genome. For each individual, the forth column shows the number of overlapped genes between the ASMGs and known imprinted genes

^**P < 0.01 Chi-square test

Discussion

As 5mC is one of the most extensive epigenetic modifications, exploring its role is fundamental to unravel the susceptibility mechanism of complex traits. Since traditional genetic variation cannot explain all the heritability of diseases or traits, epigenetic modifications have the potential to make up for deficiencies [40]. With the development of sequencing technology, the application of third-generation sequencing technology in epigenetic modification has become increasingly popular in recent years, including the construction of DNA or RNA modification maps [41, 42]. Though ONT sequencing is now capable of detecting 5mC and 5hmC simultaneously, the modified base detection at a single base was very high for 5mC (precision of 0.99 and recall of 0.97) [43], however, the false positive rate increased in 5hmC detection [43] with 5hmC-trained base-calling models were largely lacking. Therefore, this study mainly focus on 5mC. In this study, we profiled genome-wide DNA 5mC methylation using Nanopore sequencing and constructed a comprehensive epigenome atlas of 5mC modifications in ten Chinese indigenous and European pigs. These datasets are anticipated to enhance understanding of DNA methylation's potential roles and functional mechanisms in mammalian organisms. Our analysis reveals an average methylation level of approximately 74% across all individual samples (Fig. 2c), showing a high methylation level, which is consistent with the reported methylation levels in mammalian tissues or cells, such as ~ 72.5% in developing porcine skeletal muscle [44] and ~ 72% in human cells [45]. Genome-wide methylation dynamics and transcriptional element analyses confirmed hypomethylation in promoter regions (Fig. 2b), which is usually associated with active transcription of genes [46–48]. Conversely, gene bodies exhibited hypermethylation (Fig. 2b), particularly within exonic and intronic regions (see Additional file 2: Figure S1). While functional significance of methylation of gene body remains ambiguous, some studies suggest its potential role in enhancing gene expression [5], though the exact mechanism requires further elucidation. Genome-wide methylation dynamic characterization analysis showed complete consistency with WGBS benchmarks [44]. These findings collectively validate robustness and reliability of third-generation sequencing for methylation profiling.

Differential methylation analysis between high-altitude and low-altitude pigs revealed hypoxia adaptation-associated epigenetic signatures. Within the vascular smooth muscle contraction pathway, four candidate genes (CALM1, PRKCQ, RAMP1 and HBB) emerged as potential genes of high-altitude adaptation. CALM1, a member of the EF-hand calcium-binding protein family, regulates cardiac ion channel function through calcium-dependent activation [49]. Experimental evidence indicates elevated calmodulin concentrations enhance endothelial nitric oxide synthase (eNOS) activity, stimulating nitric oxide (NO) production. The NO derived from eNOS activates soluble guanylate cyclase in vascular smooth muscle to produce cyclic guanosine monophosphate (cGMP), which promotes vasodilation and plays a buffering role against hypoxia-induced vasoconstriction [50]. Notably, CALM1 forms functional complexes with KCNQ1 (another differentially methylated gene identified in our study), which regulates electrophysiological activity of potassium channel through calcium ion binding [51]. Notably, one identified differentially methylated region (chr7: 112,157,849–112,158,933, see Additional file 2: Figure S4) overlaps with the promoter of CALM1, exhibiting significantly reduced methylation in Tibetan pigs (0.196) compared to low-altitude pigs(0.846, Fig. 3b). Given the established inverse correlation between promoter methylation and gene expression [48], this epigenetic divergence likely drives transcriptional differences. Supporting this hypothesis, venous blood transcriptome analysis revealed 4.05-fold upregulation of CALM1 (a key eNOS pathway component) in high-altitude residents (3,500 m) versus sea-level counterparts [52]. Furthermore, comparative skeletal muscle analyses between indigenous Tibetan pigs (4,000 m) and lowland-adapted descendants (500 m after 10 generations) demonstrated differential methylation (0.60 vs. 0.67) within this shared DMR, suggesting evolutionary maintenance of this regulatory feature (https://www.dna-asmdb.com/). The HBB gene [53], encoding the beta-globin subunit of hemoglobin, plays a fundamental role in oxygen transport and is critically involved in adaptation hematologic disease. Genomic analyses of wild and domestic sheep reveal that specific genetic variants in this gene [54], which have been retained or introgressed from wild relatives, play a key role in the high-altitude adaptation observed in domestic populations. In addition to genetic variants, the epigenetic regulation, particularly DNA methylation of this gene, may represent another layer of regulation in hypoxia adaptation. This suggests a complex mechanism involving both sequence variation and epigenetic modification underpinning high-altitude adaptation. Nevertheless, given the lack of sufficient biological replicates in the methylation analysis, these results should be regarded as preliminary and requires further confirmation in larger populations.

ASM is a phenomenon of imbalances in alleles’ DNA methylation. Leveraging phasing capability of long-read sequencing combined with heterozygous SNPs identified through short-read sequencing, we resolved two haplotypes (HP1, HP2) and performed inter-haplotype methylation comparisons to detect ASM. This approach demonstrates enhanced resolution for allele-specific methylation profiling compared to conventional methods [55]. Significant ASM enrichment at imprinted genes, CpG islands, promoter regions, and exonic sequences aligns with observations in human cell lines [56], illustrating the possibility of identifying ASM by third-generation sequencing (Fig. 4a). ASM serves as a hallmark of genomic imprinting, enabling identification of potential imprinted regions. Our results demonstrated significant enrichment of ASM at known imprinted loci (Table 2). Notably, not all documented imprinted genes were detectable through ASM in this study, likely attributable to: 1) lack of imprinted tissues, 2) methodological limitations including false positives and resolution constraints, 3) developmental stage limitations (adult samples capturing post-natal rather than embryonic imprinting patterns) [57], and 4) inter-individual epigenetic variation [32]. ASM, initially characterized in imprinted genomic regions, demonstrates wide distribution across non-imprinted loci, suggesting fundamental regulatory functions beyond imprinting. Our analysis reveals significant ASM enrichment in promoter regions (Fig. 4a), indicative of its potential role in transcriptional regulation [58]. When we focus on the chromatin states of ASM, we found ASM is mostly in states of weak transcribed at gene, flank active TSS without ATAC and strongly active promoters, etc.(Fig. 4c). These findings demonstrated that ASM is functionally linked to transcriptional regulation, providing a theoretical basis for the colocalization of ASM with gene regulatory elements. We further observed ASM enriched in regions having bivalent chromatin signatures (co-occurring active and repressive histone methylation modifications) at TSSs, suggesting that the establishment of these poised regions requires the stable presence of methylation [59]. To further understand the relationship between ASM and ASE, we integrated RNA-seq and ONT datasets. Analysis revealed uneven ASM distribution between ASE genes and biallelically expressed genes at promoter regions. This distribution bias was similarly observed in exonic regions (Fig. 4d, e). Collectively, these findings demonstrate coordinated ASM-ASE occurrence [60], indicating ASE gene expression is likely regulated through ASM-mediated epigenetic mechanisms.

Conclusions

In this study, we constructed the 5mC map of representative pig breeds using nanopore sequencing, elucidated the methylation variation in across pigs at a single-base resolution level, and identified hypoxic adaption-related genes (CALM1, HBB, PRKCQ and RAMP1) and a transcription factor (Sp1) based on the unique physiological characteristics of Tibetan pigs. Furthermore, combined with transcriptome data analysis, it was demonstrated ASM and ASE events is synergistic and expressions of ASE genes may be regulated by ASM.

Abbreviations

5mC

The 5th carbon of cytosine

BS-seq

Bisulfite sequencing (BS-Seq)

MeDIP-seq

Methylated DNA Immunoprecipitation Sequencing

SMRT

Single-Molecule Real-Time Technologies

ONT

Oxford Nanopore Technologies

ASM

Allele-specific methylation

ASE

Allele-specific expression

CW

Chinese wild

EW

European wild

TSS

Transcription start site

TTS

Transcription termination site

CGI

CpG islands

DMR

Differentially-methylated regions

DML

Differentially-methylated loci

DMG

Differentially-methylated genes

H-DMRs

High differentially-methylated regions

L-DMRs

Low differentially-methylated regions

CALM1

Calmodulin 1

PRKCQ

Protein kinase C theta

RAMP1

Receptor activity modifying protein 1

Authors’ contributions

X.D.D., Q.Z., Y.F.J., Y.G. designed and supervised the project. Y.G. performed the majority of the analysis with the help of Y.F.J.; W.Y.Y. performed the ASE analysis. A.Q. performed part of the data analysis and visualization. C.L.W. and R.H.X. collected the samples needed for the experiment. X.D.D., Y.F.J. and L.Z.F. revised the manuscript. All authors approved the final version of the manuscript.

Funding

This work was supported by grants from the National Key Research and Development Project (2023YFF1001104 and 2023YFD1300200), China Agriculture Research System of MOF and MARA, the 2115 Talent Development Program of China Agricultural University.

Data availability

The datasets generated and/or analysed during the current study are available in the Genome Sequence Archive (https://ngdc.cncb.ac.cn/bioproject/; the accession number PRJCA005901) and the processed genome-wide DNA methylation signal tracks are available in the zenodo database (https://doi.org/10.5281/zenodo.17668868).

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of China Agricultural University (Approval No.AW12204202-1–1). Blood samples were collected for routine procedures by licensed veterinarians and all experiments strictly followed the guidelines of this committee. The experimental animals were not anesthetized or euthanized in this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.