Recurrent tandem duplication of a conserved enhancer underlies parallel genetic origins of rumplessness in chickens

Abstract

Congenital caudal body truncation in chickens—the rumpless phenotype—provides a powerful model for studying vertebrate axis elongation. This trait has been linked to misexpression of IRX1 and IRX2 in the embryonic tail bud, and breed-specific candidate mutations, including a deletion upstream of IRX1 proposed for the Piao breed, have been reported, yet whether the various breeds share a common causal basis has remained unresolved. By combining whole-genome resequencing, long-read sequencing, and population genomic analyses of four geographically isolated rumpless breeds (Chinese Piao, Japanese Uzurao, Belgian Barbu de Grubbe, and Chilean Araucana), we identify distinct tandem duplications of the IRX1/IRX2 intergenic region that converge on a single ∼700 bp conserved noncoding element, which we designate EnhRp. Haplotype and breakpoint analyses reveal that these duplications arose through at least two mutational events rather than a single shared origin. Functional assays in chicken embryos demonstrate that duplicating EnhRp acts as a hypermorphic gain-of-function mutation, elevating enhancer activity in the tail bud in a dose-dependent manner, consistent with the ectopic IRX1/IRX2 expression. Our findings support a model of parallel evolution in which distinct structural variants recurrently target a specific cis-regulatory element across at least two independent lineages, highlighting the critical role of noncoding variation in shaping morphological diversity.

Graphical Abstract

Graphical Abstract.

Introduction

The elongation and termination of the vertebrate body axis is a fundamental developmental process, and the loss of the tail provides a powerful model for its study. Body truncation phenotypes in vertebrates, occurring naturally or under artificial selection, have been used to investigate the molecular mechanisms of axial elongation. Mutations in the coding region of the TBXT (T or Brachyury) gene in mice (Herrmann et al. 1990), sheep (Li et al. 2022), cats (Buckingham et al. 2013), and dogs (Haworth et al. 2001) lead to absence or short-tail phenotypes. An Alu insertion in an intron of the TBXT gene mediates alternative splicing to drive the tail-loss phenotype in humans and apes (Xia et al. 2024). The domestic chicken also exhibits a striking “rumpless” phenotype, which is characterized by the loss of caudal vertebrae and inherited as an autosomal dominant trait (Dunn and Landauer 1934; Song et al. 2015). This phenotype has appeared in geographically distinct breeds such as Chinese Piao (Song et al. 2014), Japanese Uzurao (Tsudzuki 2003), Chilean Araucana (Noorai et al. 2012), Belgian Barbu de Grubbe (Moula et al. 2012), British Rumpless Game (Allonby and Wilson 2018), and Indonesian Tukong (Figs. S1 to S3; Tribudi et al. 2023).

Previous studies have found some candidate genes and potential causative mutations linked to the rumpless phenotype. Members of the Iroquois homeobox gene family, IRX1 and IRX2, have been associated with rumplessness in the Araucana (Noorai et al. 2012) and Piao (Wang et al. 2021; Zhang, et al. 2022b). IRX genes play a critical role during vertebrate embryonic development. Whole-mount in situ hybridization of Araucana embryos shows that misexpression of IRX1 and IRX2 in the chordoneural hinge cell population of the tailbud at Hamburger–Hamilton (HH) stages (Hamburger and Hamilton 1951) 15 to 17 triggers subsequent apoptosis (Freese et al. 2014). A recent study identified a ∼4.1 kb deletion in the upstream region of IRX1, which is completely associated with rumplessness in Piao chickens (Guo et al. 2023). However, it remains unclear whether various rumpless chicken breeds share a common driver gene or causative mutation, and how the underlying gene regulatory mechanisms lead to the formation of this trait.

In the present study, we collected whole genome sequencing data of four rumpless breeds (Piao, Uzurao, Araucana, and Barbu de Grubbe) to identify the candidate genes or mutations and elucidate the genetic basis of rumplessness. We specifically investigated whether the rumpless phenotype in these geographically distinct breeds originated from a common mutation or evolved through independent genetic mechanisms.

Results

Convergent F_st outliers at the IRX1/IRX2 locus in three rumpless breeds

To address this question, we first sought to identify the genomic loci under selection by performing a genome-wide scan for differentiated genomic regions, quantified using the fixation index (F_st). We utilized our comprehensive dataset comprising Piao (n = 36), Uzurao (n = 20), Araucana (n = 9), and Barbu de Grubbe (n = 1) (Tables S1 and S2), notably including the first genome sequencing of the Japanese Uzurao breed (Fig. S3). Combined with published sequence data of worldwide domestic chickens and red junglefowl (Table S1 and Fig. S4; Wang et al. 2020), we searched for the genomic regions exhibiting comparatively high divergence between rumpless and normal breeds. A remarkable divergent region was consistently observed in both Piao and Uzurao breeds (Fig. 1a). This divergent region overlaps with loci previously reported in Piao (chr2:85,500,954 to 86,050,282) (Wang et al. 2021; Zhang et al. 2022b) and Araucana (chr2:86,458,791 to 87,189,412) (Noorai et al. 2012), and harbors two key developmental genes: IRX1 and IRX2 (Fig. 1b). A previous study in Araucana showed that misexpression of these two genes in the embryonic tail bud disrupts progenitor cell fate, initiating a cascade that ultimately leads to apoptosis (Freese et al. 2014). The convergent differentiated region observed across the three rumpless chicken breeds likely underlies the body truncation phenotype, implicating IRX1 and IRX2 as candidate genes responsible for this trait. Interestingly, in the Uzurao breed, the highest F_st outlier was actually located at a different locus containing the TSHR gene (Fig. 1a). As TSHR is a well-known domestication gene associated with growth and metabolic regulation (Rubin et al. 2010; Yang et al. 2018), this signal is likely linked to the small body size (bantam) characteristic of the Uzurao breed, rather than the rumpless trait. In Piao, the second-highest F_st outlier was located on chromosome 24 (∼6.11 to 6.23 Mb; Fig. 1a) and encompassed BCO2, a gene previously associated with carotenoid metabolism and skin color in chickens (Eriksson et al. 2008; Fallahshahroudi et al. 2019; Wang et al. 2023).

Figure 1.

Selective signatures at the IRX1/IRX2 locus in rumpless chickens. a) Genome-wide windowed F_st Manhattan plots for Piao (top) and Uzurao (bottom) compared against nonrumpless controls. b) Detailed view of the differentiated region on chromosome 2 for the Piao (top) and Uzurao (bottom) breeds. For each breed, the plot shows the DAF within the breed (dots, left y-axis) and the F_st values (solid line, right y-axis) against all other breeds. The pink shaded area highlights the critical region previously identified in Araucana chickens (Noorai et al. 2012). The gene annotation for this locus is shown in the middle.

Identification of breed-specific tandem duplications

To pinpoint the specific genetic mutation associated with the rumpless phenotype within the IRX1/IRX2 candidate region, we sought to identify potential structural variants (SVs) that might be missed by short-read sequencing. To this end, we performed long-read sequencing (PacBio Continuous Long Read [CLR] or Oxford Nanopore R9) on two female Piao chickens (Table S2). Within the most differentiated genomic region in the Piao breed, we identified two large SVs: an 8.6 kb duplication (chr2:86,775,691 to 86,784,315, referred to as pDUP) (Fig. 2) and a 4.1 kb deletion (chr2:86,914,916 to 86,919,098, referred to as pDEL) (Fig. S5). Genotyping revealed that one sample was heterozygous for both SVs (Fig. S6). The pDEL was reported in a previous study using PacBio long-read sequencing (Zhang et al. 2022a,  et al. 2022b), whereas the pDUP is a novel SV discovered in this study.

Figure 2.

Identification of distinct tandem duplications in diverse rumpless breeds. a) Visualization (SAMplot) of duplications in the candidate region for four different rumpless breeds using short-read sequencing data. The plots show read coverage (filled area, right y-axis) and insert size of paired-end reads (paired-dots with linked line, left y-axis). Evidence for duplications, including split reads (dots) and discordant paired-end reads, supports the presence of distinct tandem duplications (uDUP, pDUP, aDUP) in Uzurao, Piao, Araucana, and Barbu de Grubbe chickens. The identified breakpoints and coordinates for each duplication are indicated. b) Minimiro plot of the long-read assembled Piao genome versus the chicken reference genome (galGal6) in the pDUP region. The alignment shows the reference genome sequence (bottom track) and the assembled region from the rumpless Piao genome (top track). The shaded area illustrates how a single genomic segment is tandemly duplicated, resulting in a head-to-tail arrangement. The genomic positions of the pDUP, aDUP, and uDUP in galGal6 coordinates are indicated by colored bars. RepeatMasker annotations for both genomes are also displayed.

We next applied SAMplot (Belyeu et al. 2021) to screen and visualize the distorted mapping signal around the two SVs across all rumpless chickens with short-read sequencing data. Fortunately, we identified two overlapping duplications in Araucana (chr2:86,776,886 to 86,784,186, 7.3 kb in length, referred to as aDUP) and Uzurao (chr2:86,766,157 to 86,785,101, 18.9 kb in length, referred to as uDUP), each with distinct breakpoints (Figs. 2a and S6 to S10). Interestingly, one individual from the Barbu de Grubbe chicken shared an identical aDUP with Araucana, suggesting a shared ancestral origin of the mutation for these two breeds. In contrast, the pDEL was unique to the Piao breed and absent from all other rumpless chickens examined (Figs. S6 to S8).

To determine the precise genomic architecture and linkage relationship of the SVs in the Piao breed, we utilized our long-read sequencing data. By applying the long-read-based phasing method, WhatsHap (Martin et al. 2023), to phase those two SVs on a heterozygous individual, we found that they are located on the same haplotype block (chr2:86,515,069 to 86,988,746) (Fig. S6). We assessed whether the pDUP and pDEL can be accurately de novo assembled in the previously published Piao long-read Oxford Nanopore genome assembly (NCBI GenBank assembly accession GCA_030914265.2). Comparing the Piao genome with the chicken galGal6 assembly revealed that these two regions can be fully reconstructed (Figs. 2b and S5). Crucially, this analysis confirmed that the pDUP is tandemly duplicated (Fig. 2b). Similarly, short-read evidence, specifically split reads and discordant insert sizes approximating the duplication length, supported the tandem nature of both the uDUP and aDUP (Figs. S9 and S10).

Population validation and evolutionary history of SVs

Having defined the precise genomic structure of these SVs, we next sought to validate their association with the rumpless phenotype at a population level. To do this, we reanalyzed publicly available sequencing datasets for rumpless chickens (Table S2; Freese et al. 2014; Beauclair et al. 2019; Bortoluzzi et al. 2020; Guo et al. 2023), integrating them with our newly generated resequencing data for the Piao breed. After filtering out individuals with low sequencing coverage (<10×), we employed two SV genotyping tools: SVTyper (Chiang et al. 2015) and Duphold (Pedersen and Quinlan 2019), to determine the genotypes for all rumpless breeds. The results revealed a strong correlation for both pDUP and pDEL (Table S2). Among the 21 rumpless Piao individuals, 10 were heterozygous for both the pDUP and the pDEL, while the remaining 11 were homozygous for both variants, indicating strong linkage. Conversely, in the three available individuals recorded as tailed, we found no corresponding SV signals (Table S2), although we note their low sequencing depth (<7×). We extended this validation to the Araucana breed and observed strong concordance with phenotypic records (Fig. S10), notably finding that the phenotypically tailed individual lacked the aDUP. Furthermore, the single sequenced rumpless Barbu de Grubbe individual was found to be homozygous for the aDUP. For the Uzurao breed, the considerable length of the uDUP presented challenges for genotyping. Nevertheless, we confirmed its consistent presence across all rumpless Uzurao individuals, supporting the robust genotype-phenotype association (Table S2). Together, these results demonstrate a consistent genotype–phenotype association between the recurrent tandem duplications and rumplessness across multiple genetically distinct breeds.

To further investigate the evolutionary history of these SVs and to test whether they share a common ancestry, we first analyzed the linkage disequilibrium (LD) between the duplications and flanking single nucleotide polymorphism (SNPs) (Fig. S11). In Araucana, the aDUP is in complete linkage with specific surrounding SNPs, while the pDUP in Piao exhibits a strong LD signature across the flanking region. In contrast, the uDUP in Uzurao did not show strong linkage with flanking SNPs, a lack of LD signal that likely stems from the difficulty in accurately genotyping such a large duplication using short-read data. To further characterize the ancestral relationships among these SVs, we performed haplotype clustering and network analysis after excluding SNPs within the uDUP and pDEL regions to minimize genotyping artifacts. Within a 730 kb interval, rumpless individuals from Uzurao, Piao, and Araucana clustered with entirely different groups of tailed chicken breeds (Fig. S12). Intriguingly, within this 730 kb region, we identified a specific 242 kb sub-interval where the haplotype patterns of Piao and Araucana were closely related (Figs. 3a and S13). This shared haplotype pattern suggests a potential common ancestral background around this locus between these two breeds, even though their broader genomic backgrounds have diverged. Notably, while this specific 242 kb haplotype background is shared between Piao and Araucana, it was also detected in nonrumpless breeds such as Red Junglefowl and Luxi Game that lack the duplications (Fig. 3a). Haplotype network analysis further confirmed the close genetic relationship of the rumpless-associated haplotypes in Piao and Araucana, contrasting with the independent origin of uDUP (Fig. 3b). These results suggest that while uDUP arose independently, pDUP and aDUP potentially share an evolutionary history involving either a common ancestral background or introgression followed by secondary deletion.

Figure 3.

Haplotype analysis of the rumpless-associated locus. a) Hierarchical clustering and heatmap of phased haplotypes across a 242 kb region encompassing the duplication locus. Individual breeds are indicated by colors. The location of the IRX2 gene and the site of the tandem duplications (vertical line) are indicated at the top. While the SNP-based haplotype background associated with the duplications is shared between Piao and Araucana, this specific haplotype is not unique to rumpless birds; it also occurs in nonrumpless birds, including Red Junglefowl and Luxi Game, which lack the duplications. Notably, within the Araucana breed, the haplotypes carrying the aDUP exhibit significant genetic divergence from the haplotypes without duplication (marked bysolid dots). b) Haplotype network of the 242 kb region. The network illustrates the relationships among haplotypes derived from the above 242 kb region. Each circle (node) represents a unique haplotype, with the area proportional to its frequency in the population after removing singleton haplotypes. Colors correspond to specific breeds as shown in the legend. The lines connecting the nodes represent mutational steps, with numerals indicating the number of mutations between major haplotypes.

We therefore hypothesized that similar mutational mechanisms might underlie these distinct duplication events. To test this hypothesis, we performed high-resolution junction mapping using breed-specific synthetic references representing the tandem structures of pDUP, aDUP, and uDUP. Remapping the sequencing reads to these tandem-duplicated sequences revealed the breakpoint junctions at single-base resolution. Specifically, we identified small deletions of 3 bp in pDUP, 2 bp in aDUP, and 1 bp in uDUP (Fig. S14). Crucially, these junctional indels were flanked by distinct microhomology sequences at the breakpoints, a hallmark of pathways such as microhomology-mediated end joining or microhomology-mediated break-induced replication (McVey and Lee 2008; Hastings et al. 2009; Carvalho and Lupski 2016). These breakpoint features, together with the haplotype results, suggest that the rumpless-associated duplications arose through at least two independent recurrent mutational events.

Identification of a candidate enhancer within the overlapping duplicated region

We next examined whether the overlapping region of the three duplications contains functional regulatory elements that could explain the molecular mechanism of the trait. The intergenic region between IRX1 and IRX2 is large with a length of ∼497 kb in chickens, ∼665 kb in mice, and ∼844 kb in humans. This large interval was recognized as a gene desert in vertebrates (Nobrega et al. 2003; de la Calle-Mustienes et al. 2005), which is devoid of protein-coding genes and is rich in distant enhancers. The shared duplicated region in chicken is located upstream of IRX1 and IRX2 with a distance of ∼325 and ∼154 kb, respectively, and is distinct from two reported enhancers in mouse (chr2:86,871,080 to 86,871,967 and chr2:87,059,251 to 87,060,277, liftover from mouse mm10 genome) (Fig. 4a; Tena et al. 2011). Meanwhile, IRX1 and IRX2 are known to share a conserved 3D regulatory landscape and multiple enhancers in mice, Xenopus, and zebrafish (Tena et al. 2011). Given that these genes display specific spatio-temporal expression patterns in chicken embryonic tissues, including limbs (Díaz-Hernández et al. 2013), somites, and neural tube (Freese et al. 2014), we hypothesize that the overlapping duplicated region functions as a distant cis-regulatory element. Comprehensive annotation of this overlapping genomic interval reveals that, beyond the core ∼700 bp conserved element, this region contains some of interspersed repetitive sequences (SINE/CR1 and LINE/L2 families) and lacks any other protein-coding genes or annotated cis-regulatory elements (Figs. 2b and S15).

Figure 4.

Epigenomic and evolutionary analysis pinpoints a conserved regulatory element within the overlapping region of tandem duplications. a) Overview of the ∼2 Mb genomic region on chicken chromosome 2 containing the IRX gene cluster. Tracks from top to bottom show: Hi-C contact map from embryonic day 15 skeletal muscle, highlighting the local chromatin architecture; ATAC-seq profiles from dissected tissues of HH14 chicken embryos (DS, differentiated somite; MS, maturing somite; ES, epithelial somite; PSM, presomitic mesoderm) and HH18 tissues (limbs and flank tissues); ChIP-seq profiles for key histone modifications (H3K27ac, H3K4me1, H3K27me3) and the insulator protein CTCF from various embryonic stages; PhyloP conservation scores across avian and mammalian species; and genome alignment blocks for mouse (mm10) and human (hg38). Key gene and SV locations are indicated at the bottom. Dashed vertical lines denote the regions magnified in (b) and (c). b) Zoomed-in view of the region encompassing the three identified rumplessness-associated tandem duplications (uDUP, pDUP, aDUP), which all converge upon a putative enhancer designated EnhRp (highlighted by a gray shaded area). The multi-omics tracks reveal that the EnhRp locus corresponds to a prominent open chromatin peak (ATAC-seq) and is strongly marked by H3K27ac, a hallmark of active enhancers, particularly in posterior trunk cells. Deep evolutionary conservation of the EnhRp element is demonstrated by both PhyloP scores and cross-species alignment blocks. c) Zoomed-in view of the genomic and epigenomic context surrounding the pDEL variant.

To further annotate the chicken genome with regulatory elements, we reanalyzed five public ChIP-seq, ATAC-seq, and Hi-C datasets from different developmental stages and tissues of the chicken embryo (Table S3; Seki et al. 2017; Sackton et al. 2019; Jhanwar et al. 2021; Mok et al. 2021; Hintermann et al. 2022; Li et al. 2024). By combining these data with evolutionary conservation scores (PhyloP) from birds and mammals (Feng et al. 2020; Genereux et al. 2020), we identified a ∼700 bp conserved noncoding element (CNE) (chr2:86,777,011 to 86,777,711, galGal6) (Figs. 4b and S15). Furthermore, we can find the homology sequences by lifting over this enhancer to human (chr5:3,023,534 to 3,024,214, hg38) and mouse (chr13:72,425,767 to 72,426,458, mm10). However, no functional element annotation was found by searching in the mouse and human ENCODE database (Luo et al. 2020). In the developing limbs and body axis (Mok et al. 2021), we can identify the gene expression of IRX1 and IRX2 and open chromatin of this CNE (Fig. 4a and b). Although these two genes are not expressed in the calf and thigh muscle at embryonic day 15, the Hi-C data provide the 3D chromatin context of this locus. Importantly, because this dataset is not derived from tail tissue, it does not provide evidence for a direct or tail-bud-specific functional interaction between EnhRp and the IRX1/IRX2 genes. Nevertheless, the presence of a highly conserved element across birds and mammals, strong open chromatin, and a significant H3K27ac signal indicates a candidate active enhancer, which we have named EnhRp.

To further characterize the potential regulatory properties of EnhRp, we performed a transcription factor (TF) binding motif analysis using the JASPAR 2026 database (Table S4; Ovek Baydar et al. 2026). Within this 700 bp sequence, we identified motifs for multiple TFs associated with vertebrate body axis development. Notably, EnhRp harbors motifs for HOX genes (HOXA5 and HOXB6) and their essential cofactors (MEIS1, MEIS2, and PBX2), alongside effectors of the WNT pathway (TCF7L1, TCF7L2, and LEF1) and components of the retinoic acid (RA) signaling pathway, including RXRA and RARG. Furthermore, we detected motifs for SOX2, a key marker of neuromesodermal progenitors (NMPs) involved in posterior body elongation (Bénazéraf and Pourquié 2013). Although motif presence alone does not establish in vivo TF occupancy, the diversity of developmental TF motifs identified within EnhRp suggests that it may function as a regulatory element responsive to multiple signaling inputs. Nevertheless, the high conservation of these motifs across chicken, human, and mouse (Figs. S16 and S17) highlights EnhRp as a potentially conserved regulatory element involved in vertebrate axis development.

Duplicated enhancer drives gain-of-function activity and downstream molecular defects

Having identified EnhRp as a highly conserved regulatory element recurrently targeted by the rumplessness-associated duplications, we next sought to functionally validate its enhancer activity and test the hypothesis that its tandem duplication represents a gain-of-function (GoF) mutation. To this end, we performed an in vivo reporter assay in wild-type chicken embryos using constructs containing either a single copy (1 $\times$  EnhRp) or two tandem copies (2 $\times$  EnhRp) of the enhancer (Fig. S18). Our results demonstrate that the tandemly duplicated 2 $\times$  EnhRp construct drives markedly stronger and more widespread EGFP expression in the tail bud compared with the weak but detectable activity of the 1 $\times$  EnhRp construct (Fig. 5a). Consistent with this qualitative pattern, quantification of relative enhancer activity showed that both the 1 $\times$  EnhRp and 2 $\times$  EnhRp constructs had significantly higher activity than the control construct (two-sided Welch's t-tests with Holm adjustment, adjusted P = 0.016, 0.0039; Fig. 5b), and that the 2 $\times$  EnhRp construct had significantly higher activity than the 1 $\times$  EnhRp construct (adjusted P = 0.016). Together, these reporter-assay results show that the tandem duplication is a hypermorphic GoF regulatory mutation, capable of driving gene expression in the tail bud.

Figure 5.

Dose-dependent enhancer activity of EnhRp in the chicken tail bud. a) Representative images of chicken embryo tail buds 24 h after co-electroporation with reporter constructs and a transfection control. Bright-field images show tail bud morphology. The mCherry channel confirms successful electroporation in all experiments. Top row: Embryos electroporated with a control plasmid show widespread mCherry expression (transfection control) but no EGFP signal (7/7 embryos). Middle row: Embryos electroporated with the 1 $\times$  EnhRp construct (representing the wild-type allele) display weak but detectable EGFP expression (7/7 embryos) restricted to the tail bud tip (white arrowheads). Bottom row: Embryos electroporated with the 2$\times$EnhRp construct (mimicking the rumpless duplication) exhibit markedly stronger and more widespread EGFP expression (10/12 embryos) in the tail bud progenitor region compared with the 1× construct (white arrowheads). The dose-dependent increase in reporter activity supports the hypothesis that the tandem duplication functions as a hypermorphic gain-of-function mutation. Scale bars, 500 μm. b) Quantification of relative enhancer activity. The relative enhancer activity was calculated as EGFP fluorescence intensity divided by mCherry fluorescence intensity in the electroporated region. Each circle represents one embryo (control, n = 3; 1 × EnhRp, n = 5; 2 × EnhRp, n = 5); bars and error bars indicate mean ± SEM. Pairwise comparisons were performed using two-sided Welch's t-tests with Holm adjustment. One asterisk indicates adjusted P < 0.05; two asterisks indicate adjusted P < 0.01.

Given that this GoF enhancer is predicted to drive ectopic expression of its target genes, we next employed the same in vivo electroporation strategy to test whether misexpression of IRX1 and IRX2 is sufficient to initiate the molecular cascade leading to the rumpless phenotype. Misexpression of IRX1 and IRX2 in the presomitic mesoderm (PSM) decreased WNT3A expression (86%, 6/7 embryos) compared with the control (Fig. S19a), in which all embryos showed normal expression (100%, 7/7 embryos). Similarly, analysis of FGF8 expression revealed that, compared with the control (100%, 8/8 embryos), FGF8 expression (63%, 5/8 embryos) was reduced and spatially restricted to the dorsal side of the PSM (Fig. S19b). Together, these results indicate that ectopic expression of IRX1 and IRX2 in the PSM downregulates WNT3A and FGF8 expression. This molecular signature mirrors the defects previously reported in Araucana rumpless embryos (Freese et al. 2014). Although we attempted to rear these electroporated embryos to later stages to observe the final skeletal phenotype, we were unable to recover viable hatchlings due to embryonic lethality following the procedure. Nevertheless, the successful recapitulation of the downstream molecular defects strongly suggests that the ectopic induction of IRX1 and IRX2, driven by the duplicated 2 $\times$  EnhRp, likely underlies the rumpless phenotype.

Discussion

Our study provides compelling evidence for parallel genetic origins underlying rumplessness in domestic chickens. These results provide another notable example of the profound impact of SVs in the noncoding genome on phenotypic evolution, a mechanism that has been increasingly recognized as a key driver of diversity in domesticated animals (Georges et al. 2019; Chen et al. 2024). We identify recurrent tandem duplications encompassing a ∼700 bp CNE that are associated with rumplessness across distinct breeds. This CNE exhibits enhancer activity in tail-bud reporter assays. Together with previous evidence of IRX1/IRX2 misexpression, these findings support a plausible regulatory model for rumplessness, although a direct regulatory link to these genes has yet to be established. This work supports parallel genetic origins of rumplessness associated with recurrent structural variation encompassing a conserved regulatory element, highlighting the role of noncoding DNA in morphological diversity. Moreover, it demonstrates how modifying the regulation of conserved developmental genes through noncoding elements can drive significant phenotypic change.

Re-evaluating the causal mutation of rumplessness

For over a decade, the genetic origin of the rumpless phenotype remained contentious. Previous studies proposed breed-specific candidates, such as two SNPs in Chilean Araucana (Freese et al. 2014) and a 4.2 kb deletion in Chinese Piao chickens (Zhang et al. 2022b; Guo et al. 2023). However, our multidimensional evidence suggests that these earlier candidates were likely markers in LD rather than the functional drivers. Specifically, we found that the two Araucana SNPs are absent in all other rumpless breeds, including Piao, Uzurao, and Barbu de Grubbe (Table S5). Furthermore, neither those SNPs nor the pDEL overlap with any active epigenetic peaks in the embryonic tail bud (Figs. 4c and S20). While the tandem duplication appears to be the universal contributor across lineages, we cannot entirely rule out a potential secondary or modulatory role for pDEL in the specific development of the Piao rumpless phenotype.

The identification of tandem duplications of the EnhRp enhancer provides a more robust and universal explanation. We observed that aDUP is in complete linkage with breed-specific SNPs in Araucana (Fig. S11), which likely explains why traditional fine-mapping in earlier studies was restricted to SNP-based markers. Although the historical “RB1” recombinant bird (Freese et al. 2014) was thought to exclude this intergenic locus, the absence of this sample prevents the re-genotyping of SVs. Nevertheless, our functional assays and the presence of shared duplications across four geographically isolated breeds suggest that the structural variation contributes to the causal mechanism. SVs typically exert a more profound impact on the regulatory landscape than SNPs (Spielmann et al. 2018; Scott et al. 2021), and the duplication of a highly conserved enhancer like EnhRp provides a clear path from genomic variation to the hypermorphic gain-of-function IRX1/2 phenotype.

Parallel genetic origins and complex evolutionary scenarios

Our results reveal a complex and fascinating evolutionary history of the rumpless trait. Haplotype analysis at the 730 and 242 kb scale strongly supports the independent origin of the Uzurao breed, which clusters with entirely different lineages compared with Piao and Araucana (Figs. 3, S12 and S13). However, the observation that duplication-containing haplotypes from Piao and Araucana cluster together within a narrower 242 kb window suggests that these duplications may have arisen on a shared ancestral background.

These observations suggest two possible evolutionary hypotheses: ancestral haplotype sharing, where a specific ancestral haplotype was shared between these lineages, followed by independent duplication events; and single introgression, wherein the duplication was introduced from one breed to another and subsequently underwent a secondary genomic rearrangement (Fig. S21). Notably, the identification of nonrumpless individuals carrying the 242 kb haplotype but lacking the duplication, such as the Red Junglefowl and Luxi Game samples observed in our study (Fig. 3a), lends support to the ancestral haplotype sharing model. If future studies identify additional nonrumpless breeds carrying this exact haplotype, it would help clarify whether these duplications were independent mutational events on a shared ancestral background or a single event spread via introgression. Regardless, the existence of at least two highly divergent mutational lineages (Uzurao vs. the Piao/Araucana clade) provides a classic example of parallel evolution.

Enhancer modularity and the dosage-based regulatory mechanism

IRX1 and IRX2 are key regulators in limb morphogenesis and cartilage patterning (Díaz-Hernández et al. 2013; Zhang et al. 2024), yet the duplication of their distal enhancer (EnhRp) in rumpless chickens specifically truncates the tail without affecting limb development. This specificity is likely due to the modularity of vertebrate enhancers. While EnhRp appears to be a specialized regulatory module for the posterior body axis, other independent enhancers likely regulate IRX1/2 expression in the limb buds. Because these limb-specific elements are not duplicated in rumpless breeds, limb development remains unaffected. This modularity allows for significant morphological innovation in one body part without deleterious pleiotropic effects in another.

Beyond modularity, our comparative evidence across multiple breeds points toward a dosage-based regulatory mechanism rather than long-range chromatin architecture changes. Although the duplication in the Uzurao breed (uDUP) encompasses a nearby CTCF-binding site, this site is not duplicated in both the Piao and Araucana breeds (Fig. 4b). Given that all three geographically isolated lineages exhibit an identical rumpless phenotype, the disruption of TAD boundaries is likely not the primary driver. Instead, the tandem duplication of the shared ∼700 bp EnhRp element represents the most parsimonious molecular trigger for this parallel evolution, suggesting that increasing the dosage of a specific conserved enhancer is sufficient to drive major morphological shifts.

A developmental model for rumplessness in chickens

Integrating our findings with previous results on Araucana chickens (Fig. 6a; Freese et al. 2014) allows us to propose a unified model (Fig. 6b) for the gene regulatory network underlying the rumpless phenotype. This model suggests the defect as a disruption of the NMPs’ normal temporal maturation (Fig. 6b). The process is initiated by a specific GoF mutation: the duplication of EnhRp, a conserved cis-regulatory element located upstream of the IRX1 and IRX2 genes. This duplication confers an enhancer activity that is expected to drive the ectopic expression of IRX1 and IRX2 in the tail bud NMPs—a territory where they are normally silent. The presence of these proneural factors disrupts NMP maturation, potentially accelerating the loss of NMP bipotentiality and self-renewal capacity shown in our model (Fig. 6b). Such imbalance forces a pronounced shift in cell fate that prematurely resolves their intrinsic bipotentiality toward a predominantly neural identity. Molecular evidence for this lineage shift includes the rapid downregulation of essential mesoderm genes (TBX6, TBXT) and the reciprocal upregulation of the neural marker SOX2. This disruption of cell identity leads to the subsequent collapse of the signaling environment (WNT3A, FGF8, and CYP26A1) required to maintain the progenitor population. Consequently, a sharp decrease in cell proliferation and an increase in apoptosis lead to the terminal depletion of the progenitor pool. This arrest of somitogenesis, marked by the loss of MESO1 and DACT2 expression, directly accounts for the failure to form the most posterior vertebrae, resulting in the definitive rumpless phenotype. Together, our model provides a plausible mechanistic path from a noncoding structural variation to a major morphological transformation.

Figure 6.

Molecular and developmental dynamics of axis termination in rumpless embryos. a) Summary of developmental and molecular features in Araucana (rumpless) embryos, with data redrawn and synthesized from Freese et al. (2014). The top graph plots the number of somites versus HH stages, revealing a significant divergence and arrest of somitogenesis in Araucana embryos starting at approximately HH20. The bottom heatmap summarizes the temporal sequence of gene expression changes observed in the Araucana tail bud relative to controls. The cascade is initiated by the ectopic expression of IRX1 and IRX2 at HH15, followed by the sequential downregulation of key genes associated with the mesodermal (TBX6, TBXT), progenitor maintenance (CYP26A1, WNT3A, FGF8), and somitogenesis (MESO1, DACT2) genes, along with the upregulation of the neural fate marker SOX2 at HH18. b) A conceptual model for the normal temporal maturation of the NMP pool during vertebrate axis elongation. The Waddington landscape (top) illustrates how the NMP population (purple sphere) transitions over developmental time. Early NMPs are highly bipotential and self-renewing, contributing to both mesodermal and neural lineages. As axis elongation proceeds, the NMP pool matures, intrinsically reducing its self-renewal capacity and bipotentiality, and becoming progressively biased toward a neural fate before its eventual depletion. The tapering bars below schematically depict this process, showing the gradual loss of self-renewal and mesodermal potential, and the concomitant rise in neural fate commitment. This model provides a framework for understanding how the aberrant molecular events in Araucana embryos (shown in a) hijack and prematurely trigger the final stages of this normal developmental program.

While our study shows an association among the CNE duplication, EnhRp enhancer activity, and IRX1/IRX2 misexpression, further investigations could solidify this model. For instance, direct functional testing using CRISPR/Cas9 to precisely insert an additional copy of the enhancer into the wild-type chickens, or using CRISPR interference (CRISPRi; eg dCas9-KRAB) to epigenetically silence the ectopic EnhRp activity in the tail buds of rumpless embryos, would provide definitive proof of causality. Moreover, the CNE's high degree of conservation across birds and mammals suggests an important, evolutionarily constrained function in normal development, although its precise role in different animals remains to be elucidated. To this end, identifying the specific TFs that bind to the 700 bp CNE, as well as the downstream target genes of IRX1 and IRX2 in both normal and ectopic tail bud, would further clarify the regulatory mechanism. Intriguingly, the parallel genetic origin of the rumpless phenotype suggests the IRX1/IRX2 locus is a structural variation hotspot, likely because specific genomic features in the region facilitate recurrent duplication events, a hypothesis requiring further sequence analysis. Importantly, while our study focused on four rumpless breeds, expanding our analysis to additional rumpless breeds would more comprehensively reveal the phenotype's genetic architecture. Nevertheless, our findings highlight the necessity of investigating noncoding SVs to fully understand the genetic basis of complex traits and evolutionary innovations.

Materials and methods

Sample collection, library construction, and genome re-sequencing

We collected a total of 23 Piao chickens from the Experimental Breeding Chicken Farm of the Yunnan Agricultural University (Yunnan, China), and 20 Uzurao chickens from Japan. Genomic DNA was extracted from blood sample using a DNeasy Tissue Kit (Qiagen) according to the manufacturer's recommended instructions. The qualified libraries were sequenced on Illumina NovaSeq 6000 platform with PE150 strategy in Novogene Bioinformatics Technology Co., Ltd (Beijing, China).

Short-read filter, alignment, and SNP calling

Raw sequencing reads were trimmed adaptor and low-quality bases using Trimmomatic (v0.32) (Bolger et al. 2014) with “ILLUMINACLIP:TruSeq3-PE.fa:2:30:10:2:True LEADING:3 TRAILING:3 SLIDINGWINDOW:4:20 MINLEN:50.” Filtered reads were aligned to chicken genome galGal6 (download from Ensembl) using bwa MEM (v0.7.17) algorithm (Li 2013). After removing PCR-duplicated reads by “samtools rmdup” (Danecek et al. 2021), we generated GVCF (genomic VCFs) files for each sample using Genome Analysis Toolkit (GATK, v4.3.0.0) (McKenna et al. 2010) HaplotypeCaller tool. Finally, we genotyped the SNPs from combined GVCF files using GATK GenotypeGVCFs tool. The filtered SNPs (hard filtering QUAL < 30, QD < 2.0, MQ < 40.0, FS > 60.0, MQRankSum < 12.5, ReadsPosRankSum < −8.0) were for downstream analysis.

Population genetics and selective sweep analysis

To reconstruct the phylogenetic relationships among chicken populations, we generated a maximum-likelihood tree using TreeMix (v1.13) (Pickrell and Pritchard 2012). We set a block size of 500 SNPs to account for LD and used Gallus gallus spadiceus (GGS) as the outgroup. To identify differentiated genomic regions, the rumpless breeds were compared against a pooled control group consisting of multiple nonrumpless domestic chicken breeds (excluding the Red Junglefowl/GGS outgroup) based on the fixation index (F_st) using VCFtools (v0.1.16) (Danecek et al. 2011). Pairwise comparisons were performed between rumpless breeds (Chinese Piao and Japanese Uzurao) and Chinese nonrumpless chickens using a 50 kb sliding window with a 25 kb step size. Additionally, the derived allele frequency (DAF) was computed for each population using VCFtools, with ancestral states defined by alleles fixed (frequency ≥ 0.9) in the red junglefowl (GGS) population.

Long-read genome sequencing

To resolve complex SVs, we utilized long-read sequencing data from two female Piao chickens. Oxford Nanopore (ONT) sequencing data for one individual (ID: Piao_P17) was obtained from our previous study (Wu et al. 2024). The sequencing library was prepared using the SQK-LSK109 ligation sequencing kit and sequenced using R9.4.1 chemistry on a PromethION P48 platform. For the second individual (ID: Piao_P9), we generated new PacBio CLR data in this study. Briefly, 2 mL of whole blood was collected from the wing vein into EDTA-2K anticoagulant tubes and stored at −80 °C. High-molecular-weight genomic DNA was extracted using the Nanobind DNA Extraction Kit (PacBio, CA, USA) following the manufacturer's protocol. To enrich for long reads, DNA fragments were size-selected with a cutoff of ∼25 kb using the BluePippin system (Sage Science, MA, USA). Sequencing libraries were constructed using the SMRTbell Express Template Prep Kit (PacBio) according to the vendor's instructions. The libraries were subsequently sequenced on the PacBio Sequel II platform (PacBio).

Structural variation analysis

Long reads of the two Piao samples were aligned to the chicken reference genome galGal6 using Minimap2 (v2.24) (Li 2018). SV calling from long reads alignment was performed using Sniffles2 (Smolka et al. 2024), and the regions of SVs were verified by SAMplot (v1.3.0) (Belyeu et al. 2021) across all rumpless breeds with short-reads sequencing. The pDUP and pDEL in Piao_P9 were phased by the WhatsHap (v2.4) (Martin et al. 2023) based on the long-read alignment and the SNPs from the short-read alignment. For accurately genotyping the duplications (pDUP, aDUP, and uDUP) and the deletion (pDEL) across all rumpless breeds, we employed a dual-strategy approach using two complementary genotypers: SVTyper (v0.7.1) (Chiang et al. 2015), which leverages discordant mate-pairs and split-read evidence, and Duphold (v0.2.3) (Pedersen and Quinlan 2019), which evaluates depth-based fold changes. In instances where the two methods yielded discordant genotypes, a scenario most frequent among heterozygous calls, the final genotypes were determined through manual curation, integrating evidence from read-depth profiles and alignment patterns. Genotyping the 18.9-kb uDUP was considerably more challenging than genotyping the smaller variants. We therefore integrated SVTyper calls with Duphold read-depth estimates and manually inspected alignment and coverage patterns using SAMplot (Table S2).

LD and haplotype analysis

To investigate the linkage between these SVs and surrounding genomic variants, we first excluded SNPs located within the genomic coordinates of the SVs (pDUP and pDEL) regions to avoid potential genotyping artifacts. Joint haplotype inference and phasing of the SVs and flanking SNPs were then performed using Beagle (v5.5) (Browning et al. 2021). The LD between the SVs and SNPs was quantified using the haplotype-frequency-based r² metric, calculated via PLINK (v2.0) (Chang et al. 2015) based on the phased dataset.

To test the alternative hypotheses of a single ancient mutation (followed by boundary instability) versus true parallel evolution, we reconstructed the evolutionary relationships of the haplotypes flanking the SVs. Phased haplotypes were extracted for two regions: a broad 730 kb critical region (chr2:86,456,875 to 87,187,678) and a narrow 242 kb region (chr2:86,604,325 to 86,846,645) containing the SV locus. For each scale, haplotype hierarchical clustering and visualization was performed using the average clustering method as implemented in the R package pheatmap (v1.0.12) (Kolde 2019). A haplotype network (Templeton et al. 1992) was also constructed using R package pegas (v1.4) (Paradis 2010) using phased SNPs from the 242 kb flanking region of the SV breakpoints (excluding the SV region itself) to resolve the fine-scale genomic backgrounds of the mutations. To limit the number of haplotypes to display, we only included the Piao, Araucana, Uzurao, Wuding, Daweishan, Luxi Game, Cobb, White Leghorn, and Red Junglefowl.

Analysis of duplication junctions and mutational mechanisms

To investigate the molecular mechanisms underlying the recurrent duplications, we constructed breed-specific synthetic reference sequences for uDUP, pDUP, and aDUP. Each reference consisted of two tandemly repeated segments based on the precisely identified breakpoints, flanked by 500 bp of reference sequence. Raw sequencing reads from each breed were remapped to their respective synthetic references using BWA-MEM (v0.7.17). The resulting alignments were visually inspected at the duplication junctions using Integrative Genomics Viewer (Robinson et al. 2017) to identify potential junctional indels and microhomology motifs.

Regulatory element annotation

For regulatory element annotation of SVs, we collected open published Hi-C (Li et al. 2024), ChIP-seq (Seki et al. 2017; Hintermann et al. 2022), and ATAC-seq (Sackton et al. 2019; Mok et al. 2021) data of chicken embryos across different tissues and developmental stages from the National Center for Biotechnology Information (NCBI) Sequence Read Archive database or the China National Center for Bioinformation Genome Sequence Archive (GSA) database. Data processing was performed by nextflow pipelines (Di Tommaso et al. 2017): atacseq (https://nf-co.re/atacseq/2.1.1) (Patel et al. 2023), chipseq (https://nf-co.re/chipseq/2.0.0) (Ewels et al. 2022), and hic (https://nf-co.re/hic/2.1.0/) (Servant et al. 2023). Genome conserved region annotations for 363 avian species (Feng et al. 2020) and 241 mammal species (Genereux et al. 2020) were downloaded from UCSC Genome Browser (Perez et al. 2025). Genome coordinates conversion to the chicken genome galGal6 was performed using the UCSC LiftOver tool and chain files. Epigenetic sequencing data were visualized using pyGenomeTracks (Lopez-Delisle et al. 2021).

Motif analysis

To identify potential TF binding sites within the CNE EnhRp, the 701 bp nucleotide sequence of EnhRp (chr2:86,777,011 to 86,777,711 in the galGal6 reference genome) was extracted. Motif scanning was performed using the FIMO tool (v5.5.7, MEME suite) (Grant et al. 2011) based on the JASPAR 2026 CORE vertebrate nonredundant database (Ovek Baydar et al. 2026). To account for genome-wide sequence composition, a background Markov model was generated using the fasta-get-markov tool based on the chicken reference genome. A P-value threshold of 10⁻⁴ was applied for motif discovery. Only motifs that were located within evolutionarily conserved elements as defined by the UCSC 100-way vertebrate genome alignments were retained as high-confidence candidate binding sites. The orthologous EnhRp sequences in the human (chr5:3,023,534 to 3,024,214, hg38) and mouse (chr13:72,425,767 to 72,426,458, mm10) genomes were retrieved using the UCSC LiftOver tool.

In vivo electroporation and reporter assay

Electroporation was performed according to previously published protocols (Sato et al. 2002; Nakaya et al. 2004), with some modifications. After incubating the eggs until HH 8, 5 mL albumin was removed through a hole created at the broader end of the egg. Subsequently, the eggshells were removed using curved scissors. The vitelline membrane was removed using a tungsten needle to expose the surface of the embryo. Then, 100 to 200 μL black ink/phosphate buffered saline (PBS) solution (1:40 ink in PBS; rOtring, Germany; S0194660) was injected beneath the embryo using a 1 mL syringe. The tk-EGFP vector containing candidate enhancer regions and/or the pCAGGS-mCherry vector were diluted with distilled water to a final concentration of 1 μg/μL and mixed with Brilliant Blue G (diluted in water to a final concentration of 0.1%; Tokyo Chemical Industry, Japan; B1146). For ectopic expression experiments, pCAGGS-chick IRX1 and IRX2 vectors were diluted with distilled water to a final concentration of 1 μg/μL.

An L-shaped platinum anode (BEX, Japan; LF613P3) was inserted into the hole used to inject the black ink/PBS solution. The tip of the platinum anode was fixed slightly posterior to Hensen's node and at a depth of 2 mm from the embryo. The DNA solution was applied over the caudal lateral epiblast (CLE) posterior to Hensen's node at ss4 (HH 8) of the chick embryo (Sato et al. 2002) using a glass pipette. A sharpened tungsten needle (The Nilaco Corporation, Japan; W-461267) serving as the cathode was placed over the CLE within the prospective PSM region. Electric pulses (25 ms pulse-on and 50 ms pulse-off at 7 V, three times) were applied using a GEB14 electroporator (BEX). After electroporation, the cathode was gently withdrawn from the amnion, and 30 μL of a penicillin-streptomycin solution (1:100 penicillin–streptomycin in PBS; Invitrogen, Thermo Fisher Scientific, MA, USA; 15,140-122) was added near the embryo. The eggshell window was firmly sealed with plastic tape. The following regions were amplified by PCR using specific primer pairs and inserted upstream of tk-EGFP: EnhRp (F-5′-CTCGAGTAACTTCCAGTCAGGTGTTTCCAATAG-3′, R-5′-GTCGACGTCATGCCTTGAATGAACTATCTTCCTATAC-3′). Full length chick IRX1 and IRX2 were amplified by PCR using specific primer pairs (IRX1: F-5′-GATATCACCATGTCCTTCCCCCAGCTGGGC-3′, R-5′-GCGGCCGCTCAAGCCGAAGGGAGAGCTGTT-3′; IRX2: F-5′-AGATCTACCATGTCCTATCCTCAGGGTTACCTC-3′, R-5′-GATATCCTATAGGTATGGTTGTACTCCTACTGTG-3′).

Assessment of enhancer activity

The tk-EGFP vector containing 1 $\times$EnhRp or 2 $\times$EnhRp enhancer regions and/or the pCAGGS-mCherry vector was diluted with distilled water to a final concentration of 1 μg/μL and mixed with Brilliant Blue G (diluted in water to a final concentration of 0.1%; Tokyo Chemical Industry; B1146). Enhancer activity was assessed as previously described (Saito et al. 2019). Embryos were harvested and placed in PBS. Fluorescence was observed using a Leica M205FA fluorescence stereomicroscope (Leica, Germany) with the following bandpass filters: GFP3 (excitation, 450 to 490 nm; emission, 500 to 550 nm) and TXR (excitation, 540 to 580 nm; emission, 610 nm long-pass). Images were captured using a Leica DFC450C camera in RGB color mode (1,280 × 960 pixels, 2 × 2 color binning). For quantification of fluorescence intensity, images were processed using ImageJ software (NIH, https://imagej.net/ij/index.html). The electroporated area was manually selected and cropped from images captured using the GFP3 filter. The corresponding areas were cropped from images captured using the TXR filter. The images were then split into red, green, and blue channels, and the green and red channel images were used to quantify fluorescence intensity from the GFP3 and TXR images, respectively. For images captured using the GFP3 filter, background intensity was subtracted, and the area used for fluorescence intensity measurements was selected. The mean fluorescence intensity within the selected area of each image was then quantified. Relative enhancer activity was calculated for each embryo as the EGFP/mCherry fluorescence-intensity ratio. Individual embryos were treated as independent experimental units. Pairwise differences among control, 1 × EnhRp, and 2 × EnhRp constructs were assessed using two-sided Welch's t-tests, with P-values adjusted for the three comparisons by the Holm method.

Whole-mount in situ hybridization

Whole-mount in situ hybridization was conducted as previously described (Dietrich et al. 1997; Matsubara et al. 2017). Unless otherwise stated, gentle rocking was performed in all steps. Collected embryos were fixed in 4% paraformaldehyde (PFA) in PBS overnight at 4 °C, without shaking. The embryos were then placed in 25% methanol in PBST (0.1% Tween 20 in PBS) and dehydrated for 10 min. Dehydration was performed by a stepwise increase in the concentration of methanol (50%, 75%, and 100% methanol, 10 min each step). Embryos were washed with 50% methanol in PBST for 10 min and twice with PBST for 5 min. The embryos were bleached in 6% H₂O₂ in PBST for 15 min. After washing twice with PBST for 5 min, the embryos were treated twice with detergent (1% NP-40, 1% SDS, 0.5% deoxycholate, 50 mM Tris pH 8.0, 1 mM EDTA, and 150 mM NaCl) for 20 min. Embryos were post-fixed in 0.2% glutaraldehyde in 4% PFA/PBS for 20 min. After washing twice with PBST for 5 min, the embryos were prehybridized in hybridization mix (50% formamide, 5× SSC pH 5.0, 250 μg/mL tRNA, 100 μg/mL heparin, 2% SDS) at 70 °C for 1 h. Then, a 1 μg/mL DIG- labeled probe was added to the hybridization mix, and embryos were incubated overnight at 70 °C. The next day, the embryos were washed four times for 30 min at 70 °C with wash solution (50% formamide, 2 × SSC pH 5.0, 1% SDS) and then with 50% wash solution in TBST (0.1% Tween 20) at 70 °C for 10 min, followed by cooling to room temperature. After washing with TBST three times for 30 min, the embryos were blocked in 20% bovine serum in TBST for 1 h at 4 °C. Embryos were placed in 20% bovine serum in TBST containing 1:2,000 Anti-Digoxigenin-AP Fab fragments (Roche, Switzerland; 11093274910), and the antibody reaction was performed overnight at 4 °C. The next day, after washing with TBST three times for 5 min and then seven times for 1 h, the embryos were washed overnight with TBST at 4 °C. The following day, the embryos were treated four times for 10 min with NTMT (100 mM NaCl, 100 mM Tris-HCl pH 9.5, 50 mM MgCl₂, and 1% Tween 20), with levamisole added to a final concentration of 2 mM. Embryos were placed in NTMT containing 10% PVA, 3.5 μL/mL BCIP (Roche, 11383221001), and 3.5 μL/mL NBT (Roche, 11383213001), and color development was performed without shaking. After confirming sufficient coloring, the embryos were washed twice with TBST for 5 min to stop color development. Embryos were fixed in 4% PFA/PBS and stored at 4 °C.

Supplementary material

Supplementary material is available at Molecular Biology and Evolution online.

Funding

This study was supported by the National Key Research and Development Program of China (2023YFF1002600), the Young Scientists Fund of the National Natural Science Foundation of China (no. 32100342), the Yunnan Fundamental Research Projects (202401AV070007), the Yunnan Provincial Science and Technology Department “Academician (Expert) Workstation” Project (202605AF350077), and the Yunnan Revitalization Talent Support Program Young Talent Project.

Data availability

The raw sequencing data generated in this study have been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center (NGDC) under BioProject accession number PRJCA034621.