Phylogenomic analysis of Cervidae provides insights into antler origin and evolution

Abstract

Antlers represent the only known example of complete annual organ regeneration in mammals. Despite this unique regenerative capacity, the genetic mechanisms driving antler origin and morphological diversification across Cervidae remain poorly understood. This study assembled six high-quality chromosome-level genomes of cervids spanning four distinct tribes, including the first genomic reference for the tufted deer (Elaphodus cephalophus). Comparative analyses across the Cervidae lineage identified signatures of positive selection on gene networks governing stem cell differentiation and bone metabolism, with elevated expression of these genes detected in antler developmental and regenerative tissues. Tribe-specific selective pressures in Cervini and Odocoileini further revealed convergent evolution targeting core developmental pathways, notably the RAS/MAPK pathway, implicating these pathways in both the emergence and enhancement of antler traits. In contrast, relaxed selective constraints in the antlerless Chinese water deer (Hydropotes inermis) revealed disruptions in gene modules associated with tumor suppression and skeletal homeostasis, suggesting a rewiring of regulatory homeostasis. These findings highlight how antler evolution reshaped physiological trade-offs, including reduced oncogenic susceptibility and enhanced tissue regeneration and cyclic bone remodeling. This study advances current understanding of antler evolution and diversification, while providing genomic resources for mammalian regenerative biology.

INTRODUCTION

Antlers constitute the only known example of complete and cyclical organ regeneration in mammals, functioning not only as dynamic osseous structures but also as key indicators of sexual selection and social hierarchy in cervids (Cegielski et al., 2006; Clutton-Brock et al., 1982, 1988; Kruuk et al., 2002; Malo et al., 2005; Suttie, 1980; Wang et al., 2019a). The extraordinary characteristics of cervid antlers prompted the late developmental biologist Richard Goss of Brown University to remark that “The antlers of cervids are so improbable that if they had not evolved in the first place, they would never have been conceived even in the wildest fantasies of the most imaginative biologists” (Goss, 1983). Despite their striking phenotypes, antler morphology does not strictly reflect phylogenetic relationships across Cervidae. While traditionally divided into four subfamilies—Cervinae, Odocoileinae, Muntiacinae, and Hydropotinae (Wilson & Reeder, 1993)—recent mitochondrial phylogenies have reclassified the clade into four tribes: Cervini and Muntiacini within a revised Cervinae, Odocoileini and Capreolini within Capreolinae (Gilbert et al., 2006; Jiang et al., 2023; Zhang & Zhang, 2012). Cervini and Odocoileini exhibit more elaborate trifurcated antlers, whereas their sister tribes, Muntiacini and Capreolini, possess reduced or unbranched antlers. Notably, Capreolini includes the Chinese water deer (Hydropotes inermis), which lacks antlers entirely (Samejima & Matsuoka, 2020). To date, the genetic mechanisms underlying antler origin and evolution remain largely unresolved. Additionally, intriguing evolutionary questions persist in Cervidae, such as extreme karyotypic remodeling during the radiation of Muntiacini (Zhang et al., 2021b). This tribe comprises two genera, Elaphodus (tufted deer) and Muntiacus (muntjacs), which exhibit markedly reduced chromosome counts compared to the ancestral cervid karyotype of 2n=70 (Huang et al., 2006). Notably, the tufted deer (Elaphodus cephalophus), the monotypic genus in Muntiacini, contains a chromosome number of 2n=48, while that of the Reeves’ muntjac (Muntiacus reevesi) is 2n=46, raising questions regarding whether this similarity reflects shared ancestry or convergent chromosomal fusion events. Comparative analyses of genomic synteny may help elucidate whether Elaphodus retained an ancestral karyotype or underwent independent fusion events parallel to those in Muntiacus. Palindromic repeat sequences, known to induce double-strand breaks, have been identified flanking the short telomeric sequences in Muntiacus genomes, which may act as drivers of prolific chromosomal fusions (Yin et al., 2021). However, whether these palindromic structures are a synapomorphic feature of Muntiacini or unique to Muntiacus has yet to be clarified.

Cervid antler regeneration represents one of the fastest known rates of organogenesis among mammals, with growth rates up to 2.75 cm per day (Landete-Castillejos et al., 2019; Price et al., 2005). This extreme growth demands substantial metabolic resources and triggers the activation of proto-oncogenes associated with proliferative signaling (Wang et al., 2019c). Despite the theoretical increase in oncogenic risk, tumor suppressor pathways appear to be under strong evolutionary constraint, potentially explaining the exceptionally low incidence of spontaneous malignancies in cervid populations (Lombard & Witte, 1959; Wang et al., 2019c). Notably, recent research has demonstrated antineoplastic effects of proliferating antler tissue extracts against human tumor cells, further implicating active cancer surveillance mechanisms during antlerogenesis (Rossetti et al., 2024). Additionally, the significant calcium demands of rapid osteogenesis during antler regeneration induce systemic skeletal remodeling. Reports have documented seasonal mobilization of calcium from other skeletal bones, followed by post-regenerative replenishment, a process manifesting as annual cyclic physiological osteoporosis (Banks et al., 1968; Stéger et al., 2010). To date, however, the genetic basis of these coordinated physiological changes remains poorly defined. Expanding the availability of high-quality genome assemblies across phylogenetically diverse lineages will facilitate integrative phylogenomic analyses aimed at uncovering the evolutionary genetic basis of these unique traits.

Despite recent advances, high-quality genome assemblies in Cervidae remain sparse and unevenly distributed, limiting evolutionary and comparative genomic resolution. Earlier cervid genomes were assembled using diverse sequencing platforms: PacBio CLR for white-tailed deer (Odocoileus virginianus), mule deer (Odocoileus hemionus), and sika deer (Cervus nippon) (London et al., 2022; Lamb et al., 2021; Xing et al., 2023); Oxford Nanopore for Reeves’ muntjac, Gongshan muntjac (Muntiacus gongshanensis), and Chinese water deer (Yin et al., 2021); and short-read, next-generation sequencing for reindeer (Rangifer tarandus) and roe deer (Capreolus capreolus) (Chen et al., 2019).However, these earlier cervid genome datasets are frequently limited by elevated error rates or incomplete assemblies, and, critically, no genome is currently available for the tufted deer. This absence, combined with the limited genomic representation of Cervini—the most species-rich tribe within Cervidae—hinders robust evolutionary inference and increases the risk of misidentifying stochastic polymorphisms as lineage-specific signals. Generating high-quality genomes from key lineages across the cervid phylogeny is essential for reducing such errors and enabling fine-scale comparative analyses of genetic divergence. In the case of giraffes, for example, the inclusion of additional high-quality ruminant genomes has helped resolve previously reported false-positive lineage-specific mutations (Liu et al., 2021). Recent advances in long-read sequencing have yielded high-fidelity genome assemblies for red deer (Cervus elaphus) (Pemberton et al., 2021), fallow deer (Dama dama) (Barnard et al., 2023), and sika deer (Li et al., 2025) using PacBio HiFi technology, marking a major step toward resolving these gaps.

This study generated high-quality chromosome-level genome assemblies for six cervid species using PacBio HiFi long-read sequencing combined with Hi-C chromosome conformation capture technologies. The species span four major tribes, including the tufted deer from Muntiacini; red deer, fallow deer, and hog deer (Axis porcinus) from Cervini; reindeer from Odocoileini; and Chinese water deer from Capreolini. To investigate population dynamics, additional 150 bp paired-end short-read data (insert size: 350 bp) were generated for all six species, four of which—tufted deer, hog deer, reindeer, and Chinese water deer—are currently classified as Near Threatened or higher on the IUCN Red List. For comparative and phylogenomic analyses, the dataset was expanded to include four recently published cervid genomes (sika deer, white-tailed deer, Reeves’ muntjac, and Père David’s deer (Elaphurus davidianus)), alongside outgroup taxa comprising the giraffe (Giraffa camelopardalis rothschildi), cattle (Bos taurus), goat (Capra hircus), and forest musk deer (Moschus berezovskii). These analyses revealed extensive chromosomal fusions within Muntiacini and uncovered the genetic mechanisms underlying antler evolution and broader physiological adaptations in cervids.

MATERIALS AND METHODS

Sample and data collection

Peripheral blood samples were obtained from red deer, fallow deer, Chinese water deer (Milu Park, Beijing, China), hog deer and tufted deer (Chengdu Zoo, Sichuan Province, China), and reindeer (Lushengyuan Animal and Plant Garden, Jilin Province, China). Approximately 10 mL of peripheral blood was drawn from each individual and stored in EDTA-anticoagulation tubes at −80°C. In addition, antler buds (n=3) were collected from 150-day-old sika deer provided by Jilin Agricultural University for transcriptomic sequencing. All animal procedures complied with relevant ethical regulations and were approved by the Northwestern Polytechnic University Ethics Committee Institutional Review Board (Approval No. 202201015).

Genome assemblies generated using long-read sequencing platforms (PacBio or Oxford Nanopore), along with reference annotations for human (Homo sapiens) (hg38; GCF_000001405.40), giraffe (GCA_017591445.1), goat (GCF_001704415.1), cattle (GCF_002263795.1), forest musk deer (GCA_022376915.1), white-tailed deer (GCA_014726795.1), Père David’s deer (PRJNA438286), sika deer (GCA_038088365.1), and Reeves’ muntjac (PRJNA640966), were downloaded from the NCBI database (https://www.ncbi.nlm.nih.gov/) for comparative genomic analyses.

Genomic DNA extraction

Genomic DNA was extracted using the cetyltrimethylammonium bromide (CTAB) method, followed by purification with the QIAGEN® Genomic Kit (Cat#13343, Qiagen, Germany), according to the standard operating procedures provided by the manufacturer. DNA quality was initially assessed by electrophoresis on 1% agarose gels to detect degradation and contamination. Purity was measured using a NanoDrop One UV-Vis spectrophotometer (Thermo Fisher Scientific, USA), with acceptable OD260/280 values between 1.8 and 2.0 and OD 260/230 values between 2.0 and 2.2. DNA concentration was quantified using a Qubit® 4.0 fluorometer (Invitrogen, USA).

Library preparation and sequencing

High-molecular-weight DNA libraries (n=6), representing red deer, fallow deer, tufted deer, reindeer, hog deer, and Chinese water deer, were constructed for PacBio HiFi sequencing according to the manufacturer’s protocols (Pacific Biosciences, USA). Libraries with an average insert size of 15 kb were sequenced on the PacBio Sequel II platform at approximately 30× coverage using three SMRT cells per species. Hi-C libraries were constructed and sequenced for all six species to a depth of approximately 100× (250 Gbp) to scaffold contigs into chromosome-level assemblies. In addition, short-read sequencing (50×, 125 Gbp) was performed on the DNBSEQ-T7 sequencing platform to generate data for subsequent PSMC analysis. Genome assembly was conducted using Hifiasm (v.0.19.5) with the following parameters: “hifiasm -o prefix of output files -t 30 file names of HiFi --h1 file names of Hi-C R1 --h2 file names of Hi-C R2 2> out.log”. Contigs were scaffolded into chromosome-level assemblies using Juicer and 3D-DNA pipelines (Dudchenko et al., 2017; Durand et al., 2016). Briefly, HiFi and Hi-C reads were used to obtain non-redundant haplotype-resolved contigs and haplotype-phased pseudo-autosomal region (PAR) sequences of the X and Y chromosomes. Subsequently, Hi-C data were used to generate chromosome-level assemblies, including X and Y chromosomes. Assembly completeness was evaluated using BUSCO (v.4.0.2) (Waterhouse et al., 2018) with the certa_odb10 database and the following parameters “busco -i HiFi.fasta -c 100 -m geno -o busco.out -l certa_odb10 --offline”. For genomic annotation, transcriptomic sequencing (10 Gbp per species) was performed as described previously (Chen et al., 2018). Repetitive elements were identified and masked using RepeatMasker (v.4.1.0) (Tarailo-Graovac & Chen, 2009).

Phylogenetic analysis

Phylogenetic reconstruction was performed using both concatenation-based and coalescence-based approaches across two independent datasets. The first dataset consisted of a 1.5 Gb multi-genome alignment generated for 14 species using Cactus (Armstrong et al., 2020). A guide tree for Cactus was constructed from annotated protein sequences using OrthoFinder (v.2.5.5) (Emms & Kelly, 2019) with the parameters “-M msa -T iqtree”. Aligned genomic sequences were extracted and concatenated to construct a species tree with IQ-TREE2 (v.1.6.12) (Minh et al., 2020), using the parameters “-B 1000 -T 96 -alrt 1000 -m GTR+F+R6”. To refine phylogenetic signal, alignment blocks located within at least 2 kb of single-copy, repeat-free regions were extracted using “hal2maf” in the HAL package (Hickey et al., 2013), with uninformative gaps trimmed using trimAI with the parameter “automated1” (Capella-Gutiérrez et al., 2009). Single-copy regions were identified using ‘halSingleCopyRegionsExtract’ from the Cactus package, and overlapping repetitive elements were filtered using BEDTools based on annotations from the sika deer genome. Blockwise phylogenies were constructed for each alignment segment using IQ-TREE with the parameter “-m GTR+F+R6”, and a coalescent-based species tree was then inferred using ASTRAL-III (v.5.7.1) (Zhang et al., 2018), following the framework described in Feng et al. (2022).

The second dataset contained 16 827 synteny-defined protein-coding orthologous gene families across 14 species, including 10 cervids from all four subfamilies and four outgroups (giraffe, cattle, goat, and forest musk deer). Genomes were first aligned to the human reference (hg38) using SegAlign (Goenka et al., 2020), and syntenic chains were extracted using the UCSC chainNet pipeline (2024 update) (Raney et al., 2024). Orthologous and lost genes were identified with TOGA (Kirilenko et al., 2023). Single copy orthologs were extracted using “extract_codon_alignment.py” in TOGA, and codon-aware alignments were generated using MACSE2 (Ranwez et al., 2018). A total of 3 585 992 four-fold degenerate sites were extracted using a custom Perl pipeline (https://github.com/yongzhiyang2012/genome_analysis/tree/master/evolution/02.phylogenetic) and concatenated for tree construction using IQ-TREE2, with the parameters “-B 1000 -alrt 1000 -m GTR+F+R6”. Divergence times were estimated using MCMCTREE (Yang, 2007), with fossil calibrations applied to Cervidae (8.5–17.7 million years ago (Ma)) and the ancestor of Bovidae and Moschidae (17.65–26.1 Ma). Phylogenetic trees for each protein-coding ortholog gene family were inferred using IQ-TREE2 with the parameters “-seqtype CODON -m GY+F3X4+I+G”, and ASTRAL-III (v.5.7.1) was used to generate a coalescent species tree (Zhang et al., 2018). To further investigate genome-wide synteny, MCScanX (Wang et al., 2012) was used with default parameters to identify collinear blocks among cattle, red deer, fallow deer, reindeer, and Chinese water deer, with cattle selected as the reference outgroup due to the availability of a high-quality chromosome-level genome and well-curated annotations (Yin et al., 2021). Chromosome-scale syntenic block plots were constructed using SynVisio (Bandi & Gutwin, 2020). Finally, phylogenetic discordance was assessed using DiscoVista (v.1.0) and relative frequency analysis (Sayyari et al., 2018) to quantify alternative branching (discordance) among phylogenies.

Satellite and telomeric sequence content in HiFi reads

Satellite and telomeric sequence content was analyzed in tufted deer, sika deer, reindeer, and Chinese water deer, following previously described methods (Yin et al., 2021). Briefly, based on fusion sites reported in previous research on muntjac species, monomer sequences representing three satellite classes (satI, satII, and satIV) identified in the female black muntjac (Muntiacus crinifron) genome (Yin et al., 2021), along with vertebrate telomeric repeats ((TTAGGG)n), were aligned to the HiFi reads of fallow deer, hog deer, red deer, tufted deer, sika deer, reindeer, and Chinese water deer using BLAST (-evalue 0.001) (Camacho et al., 2009). Reads containing these motifs were then classified into distinct patterns and quantified, as described in a previous study (Yin et al., 2021). To identify palindromic structures, self-BLAST analyses were conducted on reads containing satellite and telomeric sequences (Camacho et al., 2009). Reads with coverage≤2 were regarded as sequencing errors and excluded from downstream analyses (Kelley et al., 2010).

Demographic analysis

Short-read data (approximately 50× coverage) generated on the DNBSEQ-T7 platform were subjected to adapter trimming and quality filtering using Fastp (Chen et al., 2018). Filtered reads were mapped to the respective reference genome using BWA-MEM. BCFtools (Danecek et al., 2021) was used to call single nucleotide polymorphisms (SNPs). Variants with depth <30× or >200×, as well as those located on sex chromosomes, were excluded using “vcfutils.pl”. Demographic history was inferred with PSMC (Li & Durbin, 2011) using the following parameters: psmc -N25 -t15 -r5 -b -p 4+25*2+4+6. Generation times for each species were obtained from the IUCN Red List (https://www.iucnredlist.org/), and mutation rates were estimated using MCMCTREE based on the four-fold degenerate site-based concatenated phylogenetic tree. The generation times applied were as follows: red deer, 7 years (Hoffmann et al., 2016); fallow deer, 1 year (Baker et al., 2017); tufted deer, 2 years (Sun et al., 2016); reindeer, 6 years (Kvalnes et al., 2024); hog deer, 5 years (Ramsey et al., 2019); and Chinese water deer, 5 years (Chen et al., 2019).

Identification of positively selected genes (PSGs)

Lineage-specific evolutionary rates were estimated using the Codeml program in the PAML package (v.4.8) (Yang, 2007). The branch-site model was applied to detect PSGs. To identify selection associated with antler origin, the Cervidae lineage was designated as the foreground branch, while all cervids and outgroup species were treated as the background branch. To examine evolutionary changes linked to more elaborate antlers, Cervini and Odocoileini were defined as the foreground branches, with the remaining cervids as background. Likelihood ratio tests (LRTs) were performed to compare the alternative model allowing codons with ω>1 against a null model constrained to ω≤1, using chi-square statistics implemented in PAML (Yang, 2007). Posterior probabilities of positive selection at specific codon sites were inferred using the Bayes Empirical Bayes (BEB) approach under the M2A model. Genes were classified as significant PSGs if LRT P-values were <0.05 and BEB posterior probabilities exceeded 0.95, while those with probabilities >0.5 were considered moderate PSGs. The free-ratio model within Codeml was used to calculate gene-wise dN/dS ratios as a measure of evolutionary rate variation. For significant PSGs, AlphaFold2 (Jumper et al., 2021; Mirdita et al., 2022) was used to predict and visualize three-dimensional protein structures. To assess relaxed selection in Chinese water deer relative to other cervids, the RELAX method (Pond & Frost, 2005) implemented in HyPhy (Pond et al., 2005) was applied. Genes showing significant relaxation (P<0.05, k<1) were identified as undergoing reduced selective constraint in the Chinese water deer.

Gene family expansion analysis

For gene family analysis, the longest isoforms of annotated protein-coding genes were extracted for 14 target species, including giraffe, goat, cattle, forest musk deer, white-tailed deer, Père David’s deer, sika deer, Reeves’ muntjac, reindeer, hog deer, Chinese water deer, tufted deer, fallow deer, and red deer using AGAT (v.0.8.0) (Dainat, 2022) and GffRead (v.0.12.7) (Pertea & Pertea, 2020). Orthologous gene families were inferred using OrthoFinder (v.2.5.5) (Emms & Kelly, 2019) with default parameters. Gene family expansions and contractions along the phylogenetic tree were estimated using CAFÉ (v.4.2.1) (De Bie et al., 2006). Families showing significant change across the entire tree and at specific nodes (P<0.05) were treated as lineage-specific extractions or expansions.

RNA sequencing (RNA-seq) and single-cell RNA-seq (scRNA-seq) data collection and analysis

Bulk RNA-seq data for sika deer antler bud were generated using the Illumina NovaSeq 6000 platform and deposited in the China National Center for Bioinformation (CNCB) under accession number CRA022995. RNA-seq data for sheep horn bud were retrieved from a previous study (Wang et al., 2019c) (PRJNA438286). To reduce annotation-related discrepancies between sika deer and sheep genomes, raw RNA-seq reads from both species were mapped to the sika deer reference genome using HISAT2 (v.2.2.1) (Zhang et al., 2021a) with the parameters “--very-sensitive --dta”. The average mapping rate for sheep reads was 91.33%, indicating high compatibility and reliability for downstream analysis. Gene expression quantification was conducted using StringTie (v.2.1.7) (Pertea et al., 2015), and count matrices were generated via the prepDE.py script provided with the package. Differential expression analysis between antler and horn buds was carried out using the DEseq2 package, following standard procedures (Love et al., 2014). For scRNA-seq analysis, data from regenerating antler tissue were obtained from a previous study (Qin et al., 2023). Raw sequencing reads were processed with 10× Genomics Cell Ranger (v.8.0.1). Downstream normalization, dimensionality reduction, and visualization were performed in Scanpy, following a standardized pipeline (Li et al., 2025; Wolf et al., 2018).

Lineage-specific mutation and pathway analyses

Deduced amino acid sequences encoded by PSGs, along with specific site mutations across species, were aligned and visualized using AliView (Edgar, 2004). Protein family classification, domain prediction, and identification of functionally important residues were performed using InterPro (Paysan-Lafosse et al., 2023). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed using g:Profiler (Kolberg et al., 2023) and Enrichr (https://maayanlab.cloud/Enrichr/) (Chen et al., 2013). Visualization of results was carried out using the Generic Diagramming Platform (https://BioGDP.com) (Jiang et al., 2025) and finalized in Inkscape (www.inkscape.org/).

RESULTS AND DISCUSSION

Chromosome-level genome assemblies and phylogenomic framework across Cervidae

PacBio HiFi, Hi-C, and paired-end short-read sequencing were performed for six cervid species representing all four major Cervidae tribes, including hog deer, red deer, and fallow deer from Cervini; tufted deer from Muntiacini; reindeer from Odocoileini; and Chinese water deer from Capreolini (Figure 1A) (CNCB, 2025a). After stringent filtering to remove low-quality and contaminant sequences, between 4 200 845 and 7 228 228 high-precision long reads were retained, with average read lengths ranging from 16 091.30 to 17 940.20 bp (Supplementary Table S1). The resulting assemblies spanned 3.0 to 3.5 Gb (Supplementary Table S1), exceeding the size of earlier cervid genomes (Chen et al., 2019; Yin et al., 2021) and closely aligning with the recently generated HiFi-based assembly of sika deer (Li et al., 2025). These expanded genome sizes likely reflect improved anchoring of complex repeat regions in pericentromeric and centromeric domains, enabled by the high-fidelity long-read data (Li et al., 2025). Repeat content constituted 36.55% to 44.34% of the assembled genomes, dominated by long interspersed nuclear elements (LINEs, 666.23–793.11 Mbp), short interspersed nuclear elements (SINEs, 157.16–280.11 Mbp), and satellite DNA (92.15–368.99 Mbp), with smaller contributions from long terminal repeats (LTRs, 93.37–122.11 Mbp) and DNA transposons (57.75–81.92 Mbp) (Supplementary Table S2). While LINE and SINE distributions were broadly consistent with previously reported cervid genomes (Han et al., 2023; Pemberton et al., 2021; Xing et al., 2023), the higher satellite DNA content suggests that HiFi-based genome assemblies offer higher resolution in centromeric regions, consistent with observations in Arabidopsis (Rabanal et al., 2022) and sika deer (Li et al., 2025). Merqury analysis (Rhie et al., 2020) yielded QV scores between 61.00 and 73.23 across species, corresponding to over 99.99% single-base accuracy. Assembly completeness was further supported by BUSCO scores ranging from 96.00% to 99.00%, short-read mapping rates between 93.64% and 98.46%, and contig N50 values between 34.18 and 65.04 Mbp. These chromosome-level genomes markedly increase the number of high-quality genomes in Cervini from three to five species and offer a robust genomic foundation for investigating the evolutionary diversification and lineage-specific innovations in antler development, chromosomal architecture, and ecological adaptation within Cervidae.

Figure 1.

Evolutionary relationships and population dynamics of Cervidae

A: Whole-genome phylogeny of Cervidae reconstructed from 1.5 Gb of homologous sequences. Antler morphologies (Samejima & Matsuoka, 2020) and karyotypes are shown for each species. Ma: Million years ago. B: Chromosomal synteny among cattle, red deer, reindeer, tufted deer, Reeves’ muntjac, hog deer, fallow deer, and Chinese water deer. Red ribbons indicate chromosomal fusion synteny blocks shared by tufted deer, Reeves’ muntjac and members of the Cervini tribe. C: Abundance of reads containing satI-satIV-telomeric-repeat structures with palindromic sequences in tufted deer, fallow deer, hog deer, red deer, sika deer, reindeer, and Chinese water deer, along with previously reported data for Reeves’ muntjac, Gongshan muntjac (Muntiacus gongshanensis), and black muntjac (Yin et al., 2021). Reads with sequencing coverage ≤2 were excluded as probable errors (Kelley et al., 2010). Percentages reflect the proportion of satellite-telomeric reads with palindromic repeats among all telomeric reads. D: Inferred demographic histories of six cervid species. Colored lines represent population size trajectories associated with different conservation statuses; light-colored lines represent 100 bootstrap replicates. Blue band represents geological time: Upper Pleistocene (UP); Last Glacial Maximum (LGM); Pliocene (P). IUCN categories: VU, vulnerable; EN, endangered; NT, near threatened; LC, least concern.

Using giraffe (Liu et al., 2021), cattle (Rosen et al., 2020), goat (Bickhart et al., 2017), and forest musk deer (Chen et al., 2022) as outgroups, multiple phylogenomic strategies were applied to reconstruct cervid evolutionary relationships. Whole-genome alignment of 14 species was performed using Cactus (Armstrong et al., 2020), yielding approximately 1.5 Gb of homologous sequences for phylogenetic inference (Figure 1A). To enhance phylogenetic resolution, repeat-overlapping regions and alignment blocks shorter than 100 bp were excluded. From the filtered dataset, 300 000 independent 1 kb repeat-free windows were selected to construct species trees using both concatenation-based and coalescent-based methods (Supplementary Figure S1A, B). A total of 16 827 orthologous protein-coding gene families were identified across the 14 species and used for separate species tree reconstructions through both concatenated alignments and gene tree reconciliation approaches (Supplementary Figure S1C, D). These trees exhibited highly consistent tribe-level topologies, with 100% bootstrap support and posterior probabilities of 1, aligning with the mitochondrial phylogeny previously reported by Jiang et al. (2023). These results provide the first genome-wide phylogenetic evidence confirming the placement of the tufted deer within Muntiacini. Within Cervini, however, topological discordance emerged. The concatenation-based tree constructed from orthologous gene alignments recovered the grouping of fallow deer and hog deer with moderate bootstrap support (84%) (Supplementary Figure S1C), in contrast to full support across other nodes. To further explore this uncertainty, the relative frequencies of different topologies were compared. Two configurations—t1 ((Père David’s deer, hog deer), fallow deer) and t2 ((Père David’s deer, fallow deer), hog deer) topologies—occurred at similar frequencies, both exceeding that of t3 ((fallow deer, hog deer), Père David’s deer) (Supplementary Figure S2). This pattern is indicative of either incomplete lineage sorting or historical hybridization among these Cervini lineages.

Karyotypic restructuring frequently contributes to speciation and can profoundly influence genomic architecture and regulatory function (Eichler & Sankoff, 2003; Liu et al., 2021; Yin et al., 2021). Comparative analysis revealed that members of Odocoileini and its sister tribe Capreolini, including the Chinese water deer, retained the ancestral cervid karyotype of 2n=70 (Figure 1A), in agreement with prior cytogenetic studies (Hartmann & Scherthan, 2004; Wang & Lan, 2000; Yang et al., 1997). Within Cervini, multiple fusion events have reduced chromosome number, resulting in karyotypes of 2n=66 for sika deer and 2n=68 for Père David’s deer, fallow deer, hog deer, and red deer (Figure 1A). In contrast, the karyotype of the tufted deer, the sole representative of Elaphodus, possessed a markedly reduced karyotype of 2n=48, consistent with earlier reports (Cao et al., 2005). Chromosome-level collinearity and centromere positioning traced 11 chromosomal fusion events underlying this karyotypic contraction in the tufted deer, including nine tandem fusions—linking the centromeric region of one ancestral chromosome to the telomeric end of another—and two Robertsonian fusions, formed by joining the proximal centromeric regions of two ancestral chromosomes. Only three tandem fusions were shared between tufted deer and Reeves’ muntjac (Figure 1B; Supplementary Figure S3 and Table S3), indicating that most chromosomal fusions arose independently in the Elaphodus and Muntiacus lineages, as reported previously (Huang et al., 2006).

High-fidelity long-read sequencing of the tufted deer genome enabled precise resolution of telomeric and subtelomeric regions, revealing structural features implicated in lineage-specific chromosomal rearrangements. Previous analyses have implicated short telomeric repeats, flanking palindromic sequences, and associated Class I and IV satellite elements in driving extensive chromosomal fusions in Muntiacini (Yin et al., 2021). The new HiFi data refined these observations by correcting terminal sequencing errors and enabling high-confidence identification of telomeric architectures across cervids (Wu et al., 2024). Comparative analysis across seven species representing all four tribes, including fallow deer, hog deer, red deer, sika deer, reindeer, tufted deer, and Chinese water deer, identified the recurrent presence of either satellite I and telomere (satI-telomere) or satellite IV and telomere (satIV-telomere) composite elements (Supplementary Figure S4). Notably, a more complex satI-satIV-telomere arrangement flanked by palindromic repeats was observed in the tufted deer, consistent with previous reports in muntjac species (Yin et al., 2021), and in fallow deer, the basal lineage within Cervini (Figure 1A). These findings suggest that palindromic satellite-telomere structures originated in the common ancestor of Cervinae (containing Cervini and Muntiacini) and subsequently underwent expansion in the ancestor of Muntiacini, leading to extensive chromosomal fusions in Muntiacini. Given recent progress in yeast and mouse models (Shao et al., 2018; Wang et al., 2022), the unique palindromic architectures found in Muntiacini may serve as valuable models for elucidating mechanisms of chromosomal instability and for advancing synthetic chromosome design.

Demographic reconstructions revealed divergent population trajectories among cervid lineages (Figure 1D). Species belonging to Muntiacini exhibited markedly greater historical effective population sizes (Ne) and more pronounced demographic fluctuations compared to other tribes prior to the Upper Pleistocene (129 000–11 700 years ago). Over the past 100 000 years, sharp Ne declines were detected in both tufted deer and hog deer, consistent with their classification as Endangered and Near Threatened by the IUCN (Smith et al., 2008; Sun et al., 2016). In contrast, reindeer exhibited a notable increase in Ne during the Last Glacial Maximum (26 500–19 000 years ago), likely reflecting population expansion in response to colder climatic conditions of the Late Pleistocene (Huybers & Wunsch, 2005; Locke, 1990), in line with previous research (Chen et al., 2019). Although elevated gene flow cannot be excluded as a contributing factor, the observed expansion supports the hypothesis that reindeer successfully adapted to high-latitude glacial environments. Following this peak, reindeer experienced a pronounced postglacial decline, paralleling the reduction in suitable habitat reported between 21 000 and 6 000 years before present (Lorenzen et al., 2011; Sommer et al., 2014), potentially attributable to anthropogenic pressures and climate change. In comparison, the Chinese water deer exhibited a recent rise in Ne over the last 10 000 years, possibly reflecting demographic recovery or range expansion. Red deer and fallow deer, however, maintained relatively stable and consistently low Ne levels across this timeframe, underscoring the need for enhanced conservation attention for these less dynamically fluctuating lineages.

Genes associated with antler origin and evolution

To explore the genetic architecture underlying antler emergence, positive selection analyses were conducted across antler-bearing cervids, excluding the antlerless Chinese water deer to minimize signal dilution from relaxed selective constraint. This approach identified 152 PSGs within Cervidae (Figure 2A). Functional enrichment analyses revealed strong overrepresentation of pathways involved in stem cell differentiation, bone metabolism and remodeling, immune response, and signal transduction (Figure 2B). In parallel, 34 significantly expanded gene families (SEGFs) were detected in cervids (Figure 2A), with dominant enrichment in processes related to immune response and protein translation (Supplementary Table S4). Transcriptomic profiling of embryonic antler buds from sika deer (CNCB, 2025b), combined with comparative analysis against sheep (Ovis aries) horn bud transcriptomes (Wang et al., 2019c), identified 3 907 genes exhibiting significantly elevated expression in antler buds (Supplementary Figure S5). These genes were predominantly associated with ATP synthesis, RNA splicing, cell migration, and proliferative capacity (Supplementary Table S5). Eighteen of these antler-enriched transcripts corresponded to cervid PSGs, with annotated roles in angiogenesis, cell differentiation, lipid transport, and immune response, implicating these loci in antler origin (Supplementary Table S6). Among this subset, DCN, FST, IGFALS, and LUZP2 showed pronounced expression in the PRRX1+ mesenchymal lineage, comprising PRRX1+ mesenchymal cells (PMCs), proliferative PMCs, and antler blastema progenitor cells (ABPCs) (Li et al., 2025; Qin et al., 2023) identified within the antler regeneration periosteum (Figure 2C; Supplementary Table S7). DCN encodes decorin, a key regulator of collagen fibrillogenesis through high-affinity interactions with extracellular matrix components (Lieveld et al., 2014). A cervid-specific positively selected substitution was detected at amino acid position 67 (K67A; posterior probability=0.953), proximal to the leucine-rich repeat N-terminal domain, accompanied by a lineage-specific structural variant spanning residues 39–50 (Figure 2D). These alterations may modify decorin-mediated matrix organization, a process integral to tissue remodeling and regeneration (Goetsch & Niesler, 2016). FST encodes follistatin, a known promoter of bone regeneration through inhibition of myostatin (MSTN)-driven osteoclastogenesis (Tang et al., 2020; Wallner et al., 2017). Both IGFALS and LUZP2 contribute to organ growth and developmental regulation through growth factor stabilization and cytoskeletal signaling pathways (Domené et al., 2004; Rao et al., 2023). The convergence of cervid-specific sequence evolution with elevated expression in antlerogenic tissues highlights these four genes as central components of the molecular network governing antler formation and regeneration. Their coordinated involvement in extracellular matrix dynamics, skeletal remodeling, and proliferative control underscores their relevance as priority targets for future mechanistic studies of mammalian organ regeneration.

Figure 2.

Genes associated with the origin and evolution of deer antlers

A: Distribution of genes under different selective regimes mapped onto the cervid phylogeny, including positively selected genes (PSGs; green), significantly expanded gene families (SEGFs; red), genes under relaxed selection (GRSs; purple; Chinese water deer), and genes lost (orange; Chinese water deer). B: Gene Ontology (GO) enrichment analysis of cervid-shared PSGs after exclusion of Chinese water deer. Red stars represent antler regeneration-related processes. C: UpSet plot of highly differentially expressed genes in antlers, highly expressed genes in PMC lineage, and Cervidae PSGs. Numbers above columns represent number of genes. (a) Highly expressed genes in antler buds; (b) Highly expressed genes in PMCs lineage; (c) Cervidae PSGs. D: DCN was identified as a PSG in cervids within the extracellular matrix remodeling pathway. Positively selected sites inferred for the Cervidae ancestral lineage are indicated by red bars and the structural variant is highlighted by a red dashed box. Cervid species are highlighted in orange. E: Top four significantly enriched GO processes associated with PSGs in Cervini and Odocoileini, as well as GRSs and lost genes in Chinese water deer. F: GO biological processes shared between Cervidae and Cervini/Odocoileini. G: GO biological processes shared between Chinese water deer and Cervini/Odocoileini. Shared processes are represented by green and blue stars. H: Venn diagram depicting overlap among Chinese water deer GRSs and lost genes, Cervini PSGs, and Odocoileini PSGs. Numbers within circles represent number of genes.

The high-quality genome assembly of the Chinese water deer provides a powerful framework for exploring the genetic basis of antler formation from the perspective of secondary trait loss. By focusing on antler absence, 40 genes exhibiting complete loss and 188 genes showing signatures of relaxed selection (GRSs) were identified in the antlerless Chinese water deer. Among the lost genes, six loci (TMPRSS5, ZNF550, TIGD7, EFCC1, RXFP2, and RADX) were highly expressed in sika deer antler buds, with RXFP2 and RADX also showing high expression in the periosteum during antler regeneration (Supplementary Table S8). Previous studies have reported convergent pseudogenization of RXFP2 associated with antler loss in the Chinese water deer and forest musk deer (Moschus chrysogaster and Moschus leucogaster) (Wang et al., 2019c), a pattern further corroborated by the present analysis. Moreover, 31 GRSs in the Chinese water deer were highly expressed in sika deer antler buds and were significantly enriched in biological processes involving protein autophosphorylation, calcium ion transport, cell motility, and immune response (Supplementary Table S9). Functional enrichment analyses revealed that genes lost in the Chinese water deer were predominantly associated with cytidine-to-uridine RNA editing, while GRSs were primarily enriched in the calcium ion transmembrane transport into the cytosol (Figure 2E; Supplementary Table S10).

Evolutionary modulation of male secondary traits offers a valuable opportunity to dissect the genetic mechanisms shaping sexually selected traits under the influence of both sexual and natural selection pressures (Hine et al., 2011). Given the hypothesis that antler elaboration has been subject to intensified selection (Mittell et al., 2024), genetic changes were further examined in Cervini and Odocoileini, two tribes characterized by pronounced antler morphology. This analysis identified 47 PSGs in Cervini and 118 PSGs in Odocoileini, along with tribe-specific significantly expanded gene families. GO enrichment analysis revealed that PSGs in Cervini were primarily associated with regulation of hormone secretion and the MAPK signaling pathway (Figure 2E; Supplementary Table S11), whereas SEGFs in this tribe were enriched in protein translation and immune-related processes (Supplementary Table S12). The Odocoileini PSGs were enriched in RNA catabolic processes, positive regulation of B cell activation, and chondroitin sulfate proteoglycan biosynthesis (Figure 2E; Supplementary Table S13), while SEGFs showed marked enrichment in processes related to calcium ion transport and enzyme activity regulation (Supplementary Table S14).

Comparative analysis of PSCs across Cervidae, Cervini, and Odocoileini revealed a striking lack of overlap among these gene sets (Supplementary Figure S6), indicating that distinct genetic targets were under selection during different phases of cervid evolutionary history. No evidence was detected for individual genes undergoing continuous positive selection across lineages, in contrast to patterns reported for shared targets of selection in Homininae and Ponginae (Bakewell et al., 2007). This disparity likely reflects the greater diversity of phenotypic trajectories among cervids and supports the view that antler elaboration in Cervini and Odocoileini evolved through independent evolutionary routes.

Despite the absence of shared genes, functional convergence emerged at the pathway level. PSGs identified at the Cervidae level and within Cervini were both significantly enriched in GO terms related to cell differentiation, proliferation, migration, and regulation of organ size, with strong involvement of the MAPK signaling pathway (Figure 2F). Notably, enrichment profiles of Cervini PSGs and GRSs in the Chinese water deer overlapped in processes related to cell migration, regulation of organ size, and calcium ion homeostasis (Figure 2G). These biological functions closely mirror cellular strategies employed by antlerogenic periosteum cells during antler formation and growth (Goss & Powel, 1985; Li & Suttie, 2000), suggesting mechanistic continuity between antler origin and subsequent enhancement within Cervini. In Odocoileini, enrichment patterns differed in emphasis. PSGs at both the Cervidae and Odocoileini levels, as well as GRSs in the Chinese water deer, were significantly enriched in processes related to transport and metabolism of nitrogen-containing compounds (Figure 2F, G). This pattern implicates nutrient uptake and metabolic regulation as important contributors to antler enhancement in Odocoileini, consistent with experimental evidence showing that nutritional restriction modulates metabolic activity through effects on organ growth in sheep (Burrin et al., 1990).

Pathway-level inspection further highlighted recurrent involvement of the RAS/MAPK signaling network. Five Cervidae PSGs associated with regulation of organ size were mapped to this pathway (Supplementary Figure S7). Additional components included FLT4 and MAP4K4 among Cervini PSGs, RALGAPA2 and PLA2G4B among Odocoileini PSGs, and six lost genes or GRSs in the Chinese water deer (Supplementary Figure S7). These findings indicate sustained selective pressure on the RAS/MAPK pathway across cervid lineages, despite lineage-specific turnover in the underlying genes. Integration of gene expression data reinforced this conclusion. Quantification of PSG expression across antler cell populations revealed enrichment of proliferation-related pathways in Cervidae, Cervini, and Odocoileini, whereas differentiation-related pathways were preferentially enriched in Cervidae and Odocoileini (Supplementary Figure S8A, B). Targeting of different genetic components within shared regulatory pathways likely facilitated lineage-specific trajectories of antler elaboration. Collectively, these results suggest that antler evolution relied on recurrent selection acting on conserved developmental and metabolic pathways, particularly those governing cell proliferation, differentiation, and RAS/MAPK signaling, rather than on persistent selection of identical genes. Genes evolving under altered selection pressure across tribe-level lineages were examined by comparing the PSGs in Cervini and Odocoileini with genes lost or under relaxed selection in the Chinese water deer. This comparison identified TEX51 as a gene under relaxed selection in the Chinese water deer but subject to positive selection in Cervini (Figure 2H). Previous studies have demonstrated a significant association between genetic variation at this locus and population-level differences in bone density (He et al., 2023). Within Odocoileini, three PSGs (SMCHD1, MS4A8, and STRC) exhibited relaxed selection in the Chinese water deer, while one Odocoileini PSG (SLC18A1) was completely absent from the Chinese water deer genome (Figure 2H). Functional annotation indicates involvement of SMCHD1 in bone development (Benetti et al., 2022), MS4A8 in the regulation of smooth muscle function within the gastrointestinal tract of mice (Ishibashi et al., 2001), and SLC18A1 in monoamine transport into synaptic vesicles (Wimalasena, 2011).

Cancer resistance and bone metabolic remodeling in cervids associated with rapid antler growth

Antler regeneration in cervids proceeds at exceptional rates, accompanied by transient activation of proto-oncogenes and coordinated induction of tumor suppressor pathways. Prior analyses have suggested that widespread anticancer capacity in these species may result from strong selection on tumor suppressor genes (Wang et al., 2019c), reinforced by recent functional data (Rossetti et al., 2024). To further delineate the genetic mechanisms underlying the balance between rapid antler growth and cancer resistance, the present study incorporated comparative data from species bearing either hypertrophied or vestigial antlers, including taxa with antlers less than 2 cm in length (Hooijer & Granger, 1951), to evaluate potential correlations between antler morphology and tumor-suppressive adaptations.

Four PSGs implicated in apoptosis and DNA repair were newly identified in Cervidae: TP53, KLK10, HINFP, and CD226 (Figure 3A). Evolutionary rate analysis across tribes revealed no significant associations between antler size and divergence patterns for these loci (Supplementary Table S15). Nevertheless, the inclusion of additional cervid genomes enabled robust classification of TP53 as a PSG, which has not been recognized previously (Wang et al., 2019c). Sequence-level analysis showed that sites under positive selection (posterior probability>0.8) were concentrated in the DNA-binding domain of TP53 (Balagurumoorthy et al., 1995), with fixed substitutions across cervid lineages (Figure 3B). Structural modeling further demonstrated that these substitutions occurred in or near α-helical motifs. A threonine-to-isoleucine replacement in the helix core led to diminished hydrogen bonding capacity, while a second mutation involving a serine residue outside the α-helix—normally forming strong hydrogen bonds—also converted to isoleucine. These alterations are predicted to modify protein-DNA interactions and downstream regulatory effects (Supplementary Figure S9A–D). Functional characterization of the remaining three PSGs reinforces their role in tumor suppression. KLK10 inhibits tumorigenesis in nude mice (Goyal et al., 1998) and suppresses proliferation of human esophageal cancer cells (Li et al., 2015). HINFP plays a critical role in maintaining genomic stability, with its deletion linked to cancer development (Nirala et al., 2021). CD226 encodes a cell surface receptor essential for mediating antitumor immune responses (Bi, 2022).

Figure 3.

Co-evolution of cancer resistance and cyclic physiological osteoporosis during antler growth in cervids

A: Regulatory pathways linking rapid antler regeneration with cancer suppression, highlighting PSGs in Cervidae and GRSs in Chinese water deer. B: TP53 was identified as under significant positive selection in the Cervidae lineage, with three positively selected sites at amino acid positions 131, 263, and 269 (indicated by red bars). C: FZD6 was identified as a positively selected component of the Wnt signaling pathway in cervids, with positively selected sites highlighted by red bars. Cervid species in B and C are highlighted in orange. D: Evolutionary landscape of genes related to bone metabolism. Central pathways are surrounded by lineage-specific evolutionary patterns. Genes from Cervidae, Cervini, Odocoileini, and Chinese water deer are denoted by frames colored in orange, red, blue, and green, respectively. PSGs: Positively selected genes; SEGFs: Significantly expanded gene families; GRSs: Genes under relaxed selection in Chinese water deer (Capreolini).

Eleven PSGs in Cervidae were associated with pathways governing tissue regeneration and tumor suppression, including growth factor signaling, Wnt pathway activation, extracellular matrix remodeling, and regulation of stem cell properties. Among these, NT5E, a mesenchymal stem cell marker in both humans and mice, enhances stemness properties and regenerative capacity (Kimura et al., 2021). FZD6, a typical G protein-coupled receptor mediating Wnt signaling, regulates cell division, proliferation, adhesion, and migration (Beghini & Postorino, 2019; Zou et al., 2017) and displayed significant positive selection across cervid lineages (Figure 3C). Compared to the outgroup, three sites under positive selection were identified, including A465V, located within the ligand-binding interface of the FZD6 protein (Kozielewicz et al., 2020). A mutation at this site may alter pathway activation, suggesting that changes in FZD6 may have contributed to the evolution of antler regenerative capacity. Intriguingly, the T5I substitution in FZD6 was located within the transmembrane α-helix, where reduced hydrogen bonding potential in cervids suggests altered molecular interaction dynamics (Supplementary Figure S9E–H). None of the genes lost in the Chinese water deer exhibited evidence of positive selection in other cervids; however, 15 genes under relaxed selection in water deer showed selective signatures in other cervids. These genes were functionally related to immune regulation, transcriptional regulation, and metabolic homeostasis (Supplementary Table S16), including TP53 and FGF1, both associated with regeneration and tumor suppression (Figure 3A). FGF1, a member of the fibroblast growth factor family, plays a crucial role in differentiation and tissue regeneration (Conte et al., 2009). Additionally, several genes involved in growth factor signaling (F2R and FLT1), extracellular matrix remodeling (MMP28 and FN1), and tumor suppressor pathways (HIPK2, TP53BP1, SMCHD1, and RADX) displayed relaxed selection in the Chinese water deer. These patterns indicate that degeneration of antlers in this lineage coincided with gradual uncoupling of regenerative and tumor-suppressive constraints. Collectively, these findings deepen understanding of how the trade-off between regeneration and cancer resistance has been differentially shaped across cervid evolution.

The annual regeneration of antlers imposes an extreme mineral demand exceeding dietary intake, which typically contains only 7%–8% calcium (Estevez et al., 2009; Muir et al., 1987). To compensate, deer undergo a profound, reversible form of skeletal demineralization termed cyclic physiological osteoporosis (CPO), characterized by seasonal bone loss and restoration (Banks et al., 1968; Borsy et al., 2009; Estevez et al., 2009; Stéger et al., 2010). In this study, analysis of Cervidae PSGs revealed several genes involved in bone remodeling and mineral homeostasis, including GPR137, COL4A2, EIF2AK2, IRRK1, and P2RX7, all associated with bone resorption, turnover, and mineralization (Figure 3D). Twelve additional PSGs were distributed across functional categories related to hormone-mediated bone regulation (NPPC, IGFALS, OSM, and SLC7A8), calcium mobilization (FGF1 and VDAC2), osteoblast activation (EIF2AK2, IL6ST, and NPPC), osteoclast activation (FCER1A, EIF2AK2, and IRRK1), energy utilization (PPP1CA, FGF1, IGFALS, and GBE1), and nutrient absorption (SLC7A8) (Figure 3D). Interestingly, loss of ZNF675, a known suppressor of osteoclast-mediated bone resorption (Shin et al., 2002), in the Chinese water deer further suggested that these genes may be linked to the unique systemic bone metabolism capabilities of antlered cervids (Figure 3D). Several water deer GRSs were associated with bone metabolism, hormone regulation, calcium mobilization, energy metabolism, and nutrient absorption pathways, including 12 GRSs linked to bone metabolism and associated hormonal regulation and 10 GRSs linked to calcium mobilization (Figure 3D). Genes involved in vitamin absorption pathways (Benoit et al., 2021; Ganapathy et al., 2004; Guo et al., 2013) also showed lineage-specific selection, including SLC19A1 in Cervini and FOLH1 and TCN2 in Odocoileini, suggesting that enhanced vitamin absorption constitutes an adaptive requirement for rapid antlerogenesis. Comparative analysis of bone metabolism regulators with a curated list of 1 499 human osteoporosis-related genes (Wang et al., 2019b) revealed four overlapping loci under positive selection in Cervidae: TP53 and P2RX7, related to bone metabolism, IL6ST, involved in bone formation, and GBE1, associated with energy utilization (Figure 3D). In contrast, three osteoporosis-related genes exhibited relaxed selection in the Chinese water deer, including ADCY10 and HNF1A, related to hormone regulation, and F2R, implicated in calcium regulation. These evolutionary patterns point to functionally relevant divergence in skeletal regulation, with potential translational value for understanding antler-associated bone physiology. The identified genes constitute a targeted resource for future investigations into the molecular basis of cyclic physiological osteoporosis, a reversible state of skeletal demineralization that remains poorly understood.

Collectively, the emergence of antlers in cervids appears to have driven extensive physiological remodeling across multiple systems, including somatic regeneration, cyclic physiological osteoporosis, mineral and vitamin metabolism, and oncogenic control. These findings underscore the relevance of deer as a powerful biomedical model for studying large-scale tissue regeneration and systemic homeostatic adaptations.

CONCLUSION

Comparative genomic analyses across cervid lineages revealed lineage-specific genomic variations associated with antler evolution and related physiological traits. Notably, convergence was observed at the pathway level, particularly in bone metabolism, rather than at the gene level, suggesting shared selective pressures shaping antler enhancement in Cervini and Odocoileini. This study presents a robust analytical framework for elucidating the evolutionary origins of lineage-specific traits in cervids, paving the way for future studies on antler regeneration, reduced oncogenic risk, and cyclic physiological osteoporosis reversal, with translational potential for biomedical applications.

SUPPLEMENTARY DATA

Supplementary data to this article can be found online.

Transparent Peer Review file

Transparent Peer Review file

Funding Statement

This work was supported by the National Natural Science Foundation of China (32388102, 32030016, 32220103005) and New Cornerstone Investigator Program to W.W., Financial Program of BJAST (11000025T000003328806) to Z.Y.Z., National Natural Science Foundation of China (32122083) to Z.P.L., and China Postdoctoral Science Foundation (2024M764241) to L.S.L.

DATA AVAILABILITY

All raw sequencing data generated in this study were submitted to the Genome Sequence Archive of China National Center for Bioinformation (GSA; https://ngdc.cncb.ac.cn/gsa) under accession numbers CRA022343 (genomic data) and CRA022995 (single-cell RNA-seq data). The genome assembles generated in this study were deposited in the Zenodo repository (https://doi.org/10.5281/zenodo.15594434), NCBI database (BioProjectID PRJNA1330408), and Science Data Bank (https://www.scidb.cn/s/FRRfiu).