A chicken DNA methylation clock for the prediction of broiler health

Chickens are a globally dominant livestock species and a major source of meat and eggs, making factors influencing growth, pathogen resistance, and meat quality of high economic and scientific interest. While genetic analyses have driven substantial livestock improvement, additional strategies are needed to meet increasing demand for meat by 2050. Epigenetic information, particularly DNA methylation, offers context-dependent regulatory insights complementary to genetics. DNA methylation in animals predominantly occurs at CpG dinucleotides and is catalyzed by DNMT enzymes, including DNMT1 and DNMT3A/3B, with DNMT3L acting as a cofactor for germline methylation in mammals. However, genome-wide DNA methylation patterns in non-mammalian vertebrates, including chickens, remain underexplored. Prior chicken methylome studies have yielded conflicting observations (e.g., non-CpG methylation and repeat hypomethylation), and comprehensive datasets to develop methylation-based biomarkers are lacking. DNA methylation clocks, multivariate biomarkers predicting chronological age from CpG methylation, have proven accurate in humans and other species, and differences between methylation age and chronological age can indicate biological aging and health status. Developing a chicken clock is challenging due to the broiler's short commercial lifespan (~42 days) and absence of avian clocks to date. This study analyzes genome-wide methylation across multiple chicken tissues and ages to establish a multi-tissue DNA methylation clock and demonstrates its application for assessing broiler health.

Previous chicken methylation studies transitioned from indirect methods to whole-genome bisulfite sequencing (WGBS), but some early results (e.g., non-CpG methylation, repeat hypomethylation) conflicted with patterns known in vertebrates. The dynamics of chicken DNA methylation have been only partially explored and comprehensive, multi-tissue, multi-timepoint datasets suitable for biomarker development have been scarce. DNA methylation clocks have been developed in humans, mice, and other animals (chimpanzees, dogs, wolves, whales), typically from single tissues and limited CpG panels, with underexplored functionality and application potential in animals. Clocks can reveal divergences between methylation and chronological age predictive of mortality and health outcomes; inflammation and other pathologies are associated with age acceleration while anti-aging interventions can decelerate clock age. Given the importance of poultry health monitoring and the need for robust biomarkers, a chicken methylation clock could address gaps in health assessment tools, particularly for intestinal health issues common in broilers.

Study design and samples: Broiler chickens (Gallus gallus domesticus, Ross308) were used. Multiple WGBS datasets were generated across tissues and ages: training samples included ileum, jejunum, spleen, and breast muscle at 3, 15/16, and 34/35 days; validation samples included independent breast muscle sets at 14 and 28 days; additional high-coverage jejunum datasets at days 14 and 28 supported LMR mapping and age-related analyses. In total, 56 single-base resolution methylomes were generated (Table 1). Animal use was IACUC-approved.

Phylogenetics: DNMT homologs were identified by BLAST; DNMT1, DNMT3A, DNMT3B, and DNMT2 homologs were detected in chicken. DNMT3L was probed using 24 Swiss-Prot DNMT3L sequences against the chicken genome and a broader survey of 421 NCBI/refseq vertebrate genomes to place DNMT3L presence/absence phylogenetically.

Whole-genome bisulfite sequencing: DNA was isolated (Invitrogen PureLink) and WGBS performed at the DKFZ Genomics and Proteomics Core Facility using standard protocols. Conversion rates were >99.5–>99.9% (one set 98.2%). Coverage ranged ~34×–122×.

Read mapping and methylation calling: Reads were trimmed and mapped with BSMAP v2.5 to chicken genome galGal5. Duplicates were removed (Picard). Methylation ratios were computed with methratio.py (BSMAP). Only CpGs with ≥3× coverage were analyzed.

Genome feature analyses: Violin plots of methylation ratios used 2 kb sliding windows; annotations from Ensembl (galGal5). Promoters defined as 1 kb upstream of TSS; repeats from UCSC RepeatMasker tracks.

LMR identification and TF motif analysis: LMRs were called using MethylSeekR per tissue (lung, breast muscle, sperm) and pooled, yielding ~47,012 LMRs (~17 Mb). Additional jejunum datasets (d14, d28) identified 33,422 LMRs; differentially methylated LMRs between ages were defined by mean methylation difference >0.1. Transcription factor binding motif enrichment within LMRs was assessed with HOMER using known vertebrate motif matrices.

Clock feature preprocessing and normalization: CpGs on sex chromosomes and CpGs present in dbSNP were removed. For the CpG clock, only CpGs with strand-specific coverage >10 in every sample were retained (257,913 CpGs). For the LMR clock, average methylation per LMR was computed from CpGs with strand-specific coverage >5 present in every sample (67,651 LMRs). To address tissue-specific offsets (reflecting different maturation trajectories seen in PCA), per-feature tissue-wise mean methylation was subtracted from each sample’s value (performed for both CpG and LMR features).

Clock training and validation: Penalized regression (glmnet) regressed chronological age on normalized features. For the LMR clock, alpha was set to 0.9 (elastic net), selecting 32 LMRs (weights in Table S4); 6-fold cross-validated RMSE was 1.6 days. For the CpG clock, alpha 0.7 selected 45 CpGs (Table S5); 6-fold CV RMSE was 3.4 days. Feature genomic enrichments (promoter, gene body, intergenic) were computed relative to genome fractions. Independent validation used 6 breast muscle samples (14 and 28 days) from a separate trial; RMSEs were 2.6 and 3.4 days (LMR) and 2.3 and 3.7 days (CpG) for the two age groups, respectively.

Inflammation trial and kinome analysis: To evaluate clock sensitivity to health status, systemic inflammation was induced via repeated intraperitoneal injections of a CpG oligonucleotide (25 µg in 0.2 ml PBS) from day 1 post-hatch; GpC served as control. At multiple timepoints (d14–16 and d35), jejunum tissues were collected. Kinome peptide arrays profiled phosphorylation signaling; differential phosphorylation events were analyzed via PIIKA2 and interpreted by GO and STRING. Immune activation peaked at day 15, decreased at day 16, and was substantially reduced by day 35. DNA methylation age acceleration was computed from clock residuals relative to chronological age.

• DNMT repertoire: Chicken harbors DNMT1, DNMT3A, DNMT3B, and DNMT2 homologs, but lacks DNMT3L. A survey of 421 vertebrate genomes indicates DNMT3L is present in most mammals and many reptiles but absent in birds, monotremes, amphibians, and fish.
• Methylome architecture: Chicken somatic methylation landscapes resemble those of mouse and elephant shark, with dense CpG methylation across the genome; in contrast to mammals, chicken sperm DNA is globally hypomethylated. Repeats are highly methylated; promoters and 5′-UTRs are hypomethylated; exons, introns, and 3′-UTRs are highly methylated. Significant tissue-specific differences in methylation were observed (P = 2.2×10^-16, Wilcoxon rank-sum test). Sperm is less methylated than somatic tissues, opposite to mammalian sperm hypermethylation.
• Low-methylated regions (LMRs): Approximately 47,000 LMRs (~17 Mb) were identified and are enriched for conserved transcription factor binding sites. LMR methylation patterns robustly separate tissues in PCA, evidencing tissue specificity. In jejunum (d14 vs. d28), 33,422 LMRs were identified; 1,728 showed differential methylation (|Δ| > 0.1), with 964 hypermethylated and 764 hypomethylated at day 28, demonstrating age-dependent dynamics.
• Chicken DNA methylation clocks: An LMR-based multi-tissue clock using 32 LMRs achieved 6-fold cross-validation RMSE of 1.6 days; a CpG-based clock using 45 CpGs achieved RMSE of 3.4 days. Both clocks showed enrichment of markers in promoters. Independent validation in breast muscle (n=6; 14 and 28 days) yielded RMSEs of 2.6 and 3.4 days (LMR) and 2.3 and 3.7 days (CpG), demonstrating accuracy and generalization.
• Health application: In a systemic inflammation model (CpG oligonucleotide injections), kinome profiling indicated peak immune activation at day 15 with reduction by day 35. Correspondingly, significant epigenetic age acceleration at days 14–16 was detected by both LMR and CpG clocks (P < 0.05 vs. day 35), linking intestinal inflammation to accelerated methylation aging in broilers.

This work establishes that the chicken methylome shares vertebrate features yet differs from mammals in key respects, notably the absence of DNMT3L and hypomethylation of sperm DNA. The conservation of low-methylated regions and their enrichment for transcription factor binding implicates LMRs as central, tissue-specific regulatory elements in chicken, mirroring mammalian systems. Leveraging coordinated methylation changes within LMRs enabled construction of a multi-tissue DNA methylation clock that accurately estimates age across the short broiler lifespan. A critical step was normalization to correct tissue-specific maturation offsets evident in PCA; this aligns with prior observations that epigenetic age trajectories vary across tissues, especially in developing organisms. Compared to a CpG-only clock, the LMR clock achieved superior cross-validated accuracy, likely due to reduced noise and greater biological interpretability of multi-CpG regions tied to regulatory activity. Applying the clocks to an inflammation model demonstrated that epigenetic age acceleration coincides with heightened immune signaling, showcasing the utility of methylation clocks as health biomarkers in poultry. Such biomarkers may enable objective monitoring of intestinal and systemic health in broiler production and could be extended to other tissues and livestock species.

The study defines the chicken as a representative DNMT3L-deficient vertebrate methylome, documents conserved and tissue-specific regulation at transcription factor-associated low-methylated regions, and develops a multi-tissue DNA methylation clock that accurately predicts broiler age over a 0–42 day window. The clock detects epigenetic age acceleration under experimentally induced inflammation, supporting its use as a health marker in poultry. Future work should expand training datasets across tissues and ages, develop tissue-specific and combined models to further improve accuracy, and evaluate clock responses to diverse diseases and in other livestock species for broader agricultural health monitoring.

• Training and validation were conducted over a limited age range corresponding to the commercial broiler lifespan and specific tissues, which may constrain generalizability to other ages, lines, or tissues without further calibration.
• Tissue-specific maturation trajectories required normalization; larger datasets would enable more refined, tissue-specific clocks and combined models.
• Health application was demonstrated using a single inflammation model and primarily intestinal tissue; broader disease contexts and multi-tissue validations are needed to establish robustness.
• The study used the galGal5 assembly and filtered SNPs from dbSNP; unannotated variants or assembly differences could affect CpG inclusion and clock transferability across chicken populations.