Whole-genome analyses illuminate demographic history and lineage differentiation of polymorphic swallowtail butterfly Papilio bootes

Abstract

Background

The polymorphic swallowtail butterfly Papilio bootes displays pronounced subspecific differentiation and extensive wing-pattern variation, offering an exceptional system to investigate lineage diversification and phenotypic evolution. However, the genetic mechanisms underlying its intraspecific divergence and morphological polymorphism remain poorly understood.

Results

We assembled a chromosome-level reference genome for P. bootes and performed whole-genome resequencing of 66 individuals from six subspecies and 12 outgroup P. janaka. Population genomic analyses delineated three major genetic lineages (northern, central, southern), whose divergences aligned with Pleistocene glacial cycles. Historical hybridization was evident, with the nigricauda group originating from admixture between dealbatus and rubicundus (east) populations. The Z chromosome exhibited significantly elevated differentiation and selective sweep signatures, underscoring its role in early intraspecific divergence. Comparative analyses between black-winged and white-spotted morphs identified candidate genomic regions containing spalt and fpps, two genes with known links to wing pigmentation and melanin-associated processes, that may be associated with wing-pattern polymorphism in this species.

Conclusions

Our study demonstrates how climatic oscillations, hybridization, rapid evolution of sex chromosomes, and natural selection have jointly shaped lineage diversification and adaptive wing-pattern evolution in P. bootes, offering new insights into the genomic architecture of speciation and phenotypic diversification in Papilio butterflies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12955-w.

Background

Swallowtail butterflies (Lepidoptera: Papilionidae) have long served as important model systems in evolutionary biology because of their extraordinary diversity in wing coloration and morphology across species, sexes, populations, and seasonal forms [1–4]. Within this family, the genus Papilio, which comprises more than one-third of all papilionid species [5], is among the most intensively studied butterfly lineages. Previous work on Papilio has generated important insights into phylogenetic relationships [6, 7], biogeographic history [5], and the evolution of interspecific phenotypic variation [8]. More recently, increasing attention has shifted from broad interspecific patterns to population-level processes. In this context, population genomic studies of representative Papilio species with dense geographic and phenotypic sampling are needed to clarify how lineage divergence arises within species and how phenotypic diversity is maintained.

Papilio bootes Westwood, 1842 is a large swallowtail butterfly distributed across India, Laos, Vietnam, Myanmar, and parts of China, including Yunnan, Sichuan, and Shaanxi. It is a Batesian mimic of the toxic genus Byasa and, unlike several classic Papilio mimicry systems, shows no obvious sexual dimorphism [9–11]. Currently, eight subspecies are recognized on the basis of morphology and geography: P. b. bootes (Meghalaya, N.E. India), P. b. mixta (Nagaland, N.E. India), P. b. xamnuensis (Laos), P. b. mindoni (Dulongjiang-Irrawaddy watershed), P. b. parcesquamata (Nujiang-Salween watershed), P. b. rubicundus (Lancang-Mekong and Yuanjiang-Red River watersheds extending into northern Vietnam), P. b. nigricauda (eastern margin of the Hengduan Mountains in northwestern Yunnan and western Sichuan), and P. b. dealbatus (Daba Mountains in Chongqing and Qinling Mountains in Shaanxi). However, the extent to which these subspecies represent genetically distinct evolutionary units remains unclear.

The geographic distribution of P. bootes suggests that its evolutionary history may have been shaped by complex environmental processes. The species occupies a broad region spanning multiple mountain systems and river basins, especially across southwestern China and adjacent parts of Southeast Asia [10, 11]. In such topographically complex landscapes, climatic oscillations and geographic barriers can promote lineage differentiation by restricting dispersal, fragmenting habitats, and altering connectivity among populations [12–16]. These processes may leave detectable signals in demographic history, phylogeographic structure, and patterns of genomic differentiation. In addition, because sex chromosomes have heteromorphism, reduced recombination, reduced effective population size and often experience distinctive selective dynamics relative to autosomes [17–20], they may contribute disproportionately to early stages of population divergence. Papilio bootes therefore provides an opportunity to examine not only how climatic and geographic history have shaped lineage differentiation, but also whether genomic differentiation is distributed unevenly across the genome.

Wing-pattern evolution provides a second major dimension of interest in this species. In butterflies, wing patterns are under strong selection because they mediate multiple ecologically important functions, including aposematic signaling [21], mimicry [22, 23], crypsis [24, 25], sexual signaling, and thermoregulation [26, 27]. In Papilio, the genetic basis of mimetic wing-pattern variation is known to be diverse. Classic examples include the dsx supergene in P. polytes and P. memnon [28–30], the engrailed-invected region in P. dardanus [31], and cortex in P. clytia [32]. In P. glaucus, mimicry is influenced by two interacting sex-linked loci [33]. Notably, all of these well-studied Papilio mimicry systems are sexually dimorphic, with marked differences between females and males. By contrast, P. bootes, though a typical Batesian mimic, shows no morphological differences between sexes but exhibits remarkable intraspecific wing-pattern polymorphism. Even individuals from the same subspecies or locality differ in hindwing white spot number, degree of red pigmentation, tail shape and coloration, and the development of ventral red submarginal spots [10, 11]. This combination implies that the genetic basis of wing-pattern variation in P. bootes may differ from those described in other members of the genus, highlighting the need for genome-scale, population-level investigation.

In this study, we generated a high-quality chromosome-level reference genome for P. bootes and performed whole-genome resequencing of 66 individuals representing six subspecies, together with 12 individuals of its sister species P. janaka. Using population genomic analyses, we aim to resolve the population genetic structure of P. bootes, evaluate subspecies delimitation, assess the effects of environmental and biological factors on intraspecific differentiation, and identify candidate genomic regions potentially associated with wing-spot polymorphism. Our results provide a genomic framework for understanding lineage differentiation and phenotypic diversification in P. bootes and broaden current perspectives on the evolutionary mechanisms underlying wing-pattern evolution in Papilio.

Results

Genomic data generation, genome assembly, and SNP identification

For genome assembly, we generated 10 Gb of PacBio HiFi long reads, 25 Gb of Illumina paired-end short reads, and 30 Gb of Hi-C reads from a male P. bootes individual. We also performed whole-genome resequencing of 78 Papilio individuals, including 66 P. bootes samples representing six subspecies: dealbatus (n = 7), nigricauda (n = 6), xamnuensis (n = 15), rubicundus (n = 17), mindoni (n = 13), and parcesquamata (n = 8), and 12 individuals of the outgroup species P. janaka (Fig. 1a; Additional File 1: Figs. S1–S2; Additional File 2: Table S1). In total, approximately 600 Gb of Illumina resequencing data were generated.

Fig. 1.

Sampling locations, genome assembly, and SNP distribution across chromosomes. a Geographic distribution and sequenced samples of six subspecies within P. bootes, grey regions represent P. janaka. The blue line segments represent the mountain ranges (QL: Qinling Mountains, DB: Daba Mountains, YL: Yunling Mountains, NS: NuShan Mountains, GLG: Gaoligong Mountains, AL: Ailao Mountains and WL: Wuliang Mountains). b Circos plot showing chromosome-level alignment between P. bootes (right) and P. machaon (left). The tracks represent (I) chromosome, (II) genome-wide GC percentage, (III) coding gene density, (IV) LINEs, (V) SINEs, (VI) LTRs, and (VII) DNA transposons. c Number of SNPs per 100-kb window across the P. bootes genome

De novo genome assembly produced 36 gap-free contigs with high continuity (contig N50 = 7.9 Mb; Additional File 2: Table S2). These contigs were anchored into 30 chromosomes with Hi-C data, including 29 autosomes and one Z chromosome (Fig. 1b; Additional File 1: Fig. S3). Collinearity analysis showed clear chromosome-level synteny between P. bootes and P. machaon (Fig. 1b). The assembled genome had a total length of 225 Mb and contained 12,752 predicted protein-coding genes, with high completeness as assessed by BUSCO (99.4%; Additional File 2: Table S2). The overall GC content of the genome was 35.03%. Transposable elements (TEs) totaled 54.27 Mb in length, accounting for 24.06% of the genome (Additional File 2: Table S3).

Resequencing reads from the 78 individuals mapped well to the reference genome, with an average mapping rate of 96.97% and a mean depth of 35 × (Additional File 2: Table S1). The six samples for which abdominal tissues were used had an average sequencing depth of ~ 30 ×, comparable to that of samples prepared from leg tissues (Additional File 2: Table S1). Although some abdominal samples showed relatively low mapping rates (54–84%; Additional File 2: Table S1), they still achieved sufficient genome-wide sequencing coverage (Additional File 1: Fig. S4) and were therefore retained for all subsequent population genomic analyses. After SNP calling and filtering, we retained approximately 6.52 million high-quality SNPs for downstream analyses (Fig. 1c). Most SNPs were located in intergenic regions (34.76%), followed by intronic (24.05%), upstream (17.93%), and downstream (17.79%) regions of genes (Additional File 2: Table S4).

Population genomic analyses reveal three major lineages within P. bootes

The Neighbour-Net network based on whole-genome SNPs clearly separated P. janaka from P. bootes and identified seven major genetic clusters within P. bootes (Fig. 2a). Among the six sampled subspecies, four (parc., deal., xamn., and mind.) were largely monophyletic. The subspecies nigr. showed substantial genetic exchange with deal. The subspecies rubi. formed two geographically isolated groups: rubi. (west), distributed along the upper Lancang-Mekong watershed, and rubi. (east), ranging from the Yuanjiang-Red River watershed to Vietnam.

Fig. 2.

Phylogenetic relationships and population genetic structure of P. bootes. a Neighbour-net network based on whole-genome SNPs. b Maximum-likelihood phylogenetic tree constructed from whole-genome SNPs; all nodes show 100% bootstrap support. c Principal component analysis (PCA) of all P. bootes samples based on whole-genome SNPs. d ADMIXTURE results for P. bootes populations with K = 2, 4, and 6

Maximum-likelihood (ML) analysis of the whole-genome SNP dataset fully resolved the phylogenetic relationships among the sampled P. bootes individuals, with all internal nodes receiving maximal bootstrap support (100%) (Fig. 2b). Consistent with the Neighbour-Net results, within P. bootes, five groups (deal., xamn., rubi. (w), mind., and parc.) were strongly supported as monophyletic. In contrast, nigr. and rubi. (e) were not monophyletic. The ML tree indicated that the northern deal. was the earliest-diverging group, while the southern xamn. (sampled from Laos) was nested within other lineages from Yunnan Province and Vietnam, and showed a closer relationship to rubi. (e).

Principal component analysis (PCA) revealed genetic clustering patterns broadly consistent with geographic distribution (Fig. 2c). The deal. group from Shaanxi Province (the northernmost region) and the xamn. group from Laos (the southernmost) were clearly separated from the other lineages distributed across Yunnan Province and Vietnam. The nigr. group, primarily found in Sichuan Province, occupied an intermediate position between the Shaanxi and Yunnan lineages. The rubi. (e) group, distributed from the Yuanjiang-Red River watershed to Vietnam, clustered between the Laos lineage and the other Yunnan lineages.

ADMIXTURE analysis identified K = 4 as the optimal clustering result (CV error = 0.4287; Fig. 2d; Additional File 1: Fig. S5). At K = 2 (CV error = 0.48715), P. bootes was clearly separated from the outgroup P. janaka. At K = 4, three major genetic clusters were recovered within P. bootes: a northern cluster corresponding to deal., a southern cluster corresponding to xamn., and a central cluster comprising mind., parc., and rubi. (w). Based on their geographic distributions, these clusters were designated the northern (N), southern (S), and central (C) lineages. Under this clustering, nigr. showed mixed ancestry between the northern and central lineages (N + C), whereas rubi. (e) showed mixed ancestry between the southern and central lineages (S + C). At K = 6 (CV error = 0.47171), the central lineage further separated into mind., parc., and rubi. (w), and nigr. continued to show admixture involving deal. and rubi. (e). Together, these results indicate that P. bootes is structured into three major lineages, with additional admixture involving the nigr. and rubi. (e) groups.

Demographic history, introgression and hybridization in P. bootes

We used PSMC to infer the demographic histories of the three major P. bootes lineages: northern (deal.), southern (xamn.), and central (mind., parc., rubi. (w)). The PSMC trajectories suggested that the three lineages had broadly similar effective population sizes before ~ 1.0 Ma, after which their demographic histories began to diverge (Fig. 3a). The northern lineage reached its inferred population peak between the Naynayxungla Glaciation (0.78–0.50 Ma) and the Penultimate Glaciation (0.30–0.13 Ma), whereas the southern and central lineages reached their inferred peaks between the Penultimate Glaciation and the Dali Glaciation (0.07–0.01 Ma) (Fig. 3a). MSMC analyses further suggested that the northern lineage diverged from the others around 0.3–0.5 Ma, whereas the central and southern lineages separated around 0.15 Ma (Fig. 3b). Notably, the effective population sizes (Nₑ) of all three lineages peaked during interglacial periods. During the Dali Glaciation (DG, 0.07–0.01 Ma), the Nₑ of the central and southern (xamn.) groups declined sharply, while the northern group (deal.) remained relatively stable (Fig. 3a).

Fig. 3.

Demographic history, gene flow, and hybridization patterns of P. bootes. a Demographic history of the three major lineages. Grey bars indicate the timing of the Dali Glaciation (DG, 0.07–0.01 Ma), Penultimate Glaciation (PG, 0.30–0.13 Ma), and Naynayxungla Glaciation (NG, 0.78–0.50 Ma). b Relative cross-coalescence rates between lineage pairs based on MSMC analyses. c f-branch results showing gene flow patterns among lineages. d Phylogenetic topologies for the three lineages (c/r/x, nigr. and deal.), inferred from 50-kb sliding windows across the genome. e Evolutionary scenario simulations of the hybrid origin of the nigr. group. f Schematic diagram summarizing hybridization and introgression among P. bootes groups

ADMIXTURE analyses revealed potential gene flow within P. bootes, particularly in the nigr. and rubi. (e) groups (Fig. 2d). Consistent with this result, the F-branch analyses detected significant gene flow from the central, xamn., and rubi. (e) groups (the c/r/x clade) into the nigr. group (Fig. 3c). These results were also supported by both ABBA–BABA tests and f₄-ratio statistics (Additional File 2: Table S5). By contrast, although rubi. (e) also showed an admixed ancestry profile, no corresponding signal of clear directional gene flow was recovered in the F-branch analysis.

To examine genome-wide phylogenetic variation, we reconstructed ML trees in non-overlapping sliding windows across the genome. For the genetically admixed nigr. group, the predominant topology T1 (deal., (c/r/x, nigr.)) and the secondary topology T2 (c/r/x, (deal., nigr.)) occurred at similar frequencies (55% vs 42%), both significantly exceeding T3 (nigr., (deal., c/r/x)) (3%) (Fig. 3d). The genomic regions supporting T1 and T2 were distributed broadly across the genome (Fig. 3d; with the Z chromosome being an exception, see below), suggesting that nigr. likely originated through hybridization between deal. and c/r/x rather than from partial introgression of c/r/x. In contrast, although the rubi. (e) group also displayed an admixed pattern, the frequencies of its three topologies were more balanced (T1: 46%, T2: 30%, T3: 24%; Additional File 1: Fig. S6). In particular, the relatively similar frequencies of T2 and T3 suggest that its genetic admixture likely resulted from incomplete lineage sorting (ILS) rather than introgression.

HyDe tests further confirmed the hybrid origin of the nigr. group, with the c/r/x clade contributing slightly more ancestry to the hybrid (P1 = deal., P2 = c/r/x, Hybrid = nigr., γ = 0.59/0.75/0.68; Additional File 2: Table S6). Coalescent-based demographic modeling in fastsimcoal2 also identified a hybrid-origin model as the best-fitting scenario (Model 4; Additional File 2: Table S7). Under this model, nigr. arose through hybridization between rubi. (e) and deal. at approximately 0.03 Ma, following the divergence of xamn. and rubi. (e) at approximately 0.12 Ma (Fig. 3e). Taken together, these results consistently support a hybrid origin for the nigr. group (Fig. 3f).

The Z chromosome shows elevated differentiation among lineages

Genome-wide scans of genetic differentiation revealed that the strongest divergence among the three major P. bootes lineages was concentrated along the Z chromosome. Fst values calculated in 10-kb sliding windows showed consistently elevated differentiation across the entire Z chromosome relative to the autosomes (Fig. 4a). Similar patterns were recovered using Dxy and PBS analyses (Additional File 1: Figs. S7–S8).

Fig. 4.

Genetic differentiation among the three major lineages of P. bootes. a Manhattan plots of Fst for pairwise comparisons among the deal., xamn., and central lineages in 10-kb sliding windows across the genome. Red lines indicate the mean Fst values across all chromosomes. b Haplotype structures on chromosome 1 and chromosome Z, where rows represent individuals and columns denote polymorphic sites. c-d Correlation between Z-chromosome heterozygosity and genomic inbreeding coefficient (F_ROH) and e–f correlation between autosomal heterozygosity and F_ROH

Genotype distributions further showed that the deal. group carried a markedly higher proportion of homozygous variant sites on the Z chromosome than the other P. bootes groups, whereas no comparable excess was observed on the autosomes (Fig. 4b). Across individuals, Z-linked homozygosity was strongly correlated with the genomic inbreeding coefficient F_ROH (r² = 0.838, P = 6.42 × 10⁻¹⁰; Fig. 4c, d), whereas autosomal homozygosity showed no significant relationship with F_ROH (r² = 0.376, P = 0.26; Fig. 4e, f).

Furthermore, Tajima’s D analyses revealed that the Z chromosome in the deal. group experienced selective sweeps (Tajima’s D < 0), whereas other groups showed signatures of balancing selection on the Z chromosome (Tajima’s D > 0) (Additional File 1: Fig. S9). In addition, most Z-linked variants in the deal. group were classified as having Moderate or Modifier effects (Additional File 1: Fig. S10), suggesting that selection predominantly acted on non-coding regions rather than causing major functional disruptions in coding sequences.

Candidate genomic regions associated with hindwing spot polymorphism in P. bootes

Papilio bootes shows extensive hindwing spot polymorphism across its range, with the strongest contrast occurring between black-winged groups (deal. and nigr.) and white-spotted groups (mind., parc., rubi., and xamn.). To investigate the genomic basis of this phenotypic difference, we compared Fst and nucleotide diversity (π) between black-winged and white-spotted populations. We defined candidate regions as windows showing both high genetic differentiation and reduced nucleotide diversity in the black-winged group relative to the white-spotted group. Based on these criteria, we identified 28 candidate windows (containing 22 genes) falling within the top 5% tails of both statistics (Fig. 5a; Additional File 2: Table S8).

Fig. 5.

Genomic signatures of divergence and selection between black-winged (B) and white-spotted groups (W) of P. bootes. a Genome-wide distribution of pairwise genetic differentiation Fst and relative nucleotide diversity ln(π_B/π_W) between the B and W groups. Each dot represents a 10-kb genomic window. The upper and right panels show the frequency distributions of ln(π_B/π_W) and Fst, respectively, and the red curves indicate the cumulative percentages. Dashed lines denote the 5% thresholds used to identify highly divergent candidate regions. b, c Genomic signatures of selection of genes spalt and fpps. The gene region is highlighted in red. Blue shaded areas indicate selective regions, and gray boxes mark coding sequence (CDS) regions

Most candidate genes have no clear documented association with wing coloration. However, two genes, spalt and fpps, were worth mentioning because of their known roles in wing pigmentation and melanin-associated processes in Lepidoptera [34, 35]. In the spalt region, the strongest signatures of differentiation and diversity reduction were located mainly in non-coding sequences, whereas the fpps signal extended across both coding and non-coding regions (Fig. 5b, c). Moreover, both genes displayed signatures of selective sweep in the black-wing group (Tajima’s D < 0; Fig. 5b and c), suggesting that these regions may have experienced recent selection associated with the black-wing phenotype.

Discussion

Genomic delineation of evolutionary lineages within P. bootes

Our whole-genome analyses provide a comprehensive view of lineage diversification in P. bootes, revealing three major genetic groups: north (deal.), central (mind., parc., rubi. (w)), and south (xamn.) that correspond closely with the species’ current geographic distribution. The strong phylogeographic structure inferred from genome-wide SNP data supports the existence of multiple evolutionary lineages, each associated with distinct river basins and mountain systems across southwestern China and Indochina.

Despite the recognition of eight subspecies based on wing morphology and geographic isolation [9, 10], our genomic evidence indicates that not all of these morphologically defined units correspond to genetically distinct entities. Among the sampled subspecies, parc., mind., xamn., and deal. form well-supported monophyletic clades, while nigr. and rubi. exhibit strong genetic admixture signals. These findings highlight the limitations of morphology-based taxonomy in groups characterized by high intraspecific phenotypic plasticity and recent divergence. Morphological characters such as hindwing spot number and red pigmentation, though taxonomically informative, appear to have evolved under selection or hybridization pressures rather than as neutral lineage markers.

The existence of a hybrid group (nigr.) between deal. and rubi. (e) exemplifies how gene flow continues to shape population boundaries in P. bootes. Such a hybrid origin aligns with the geographic proximity of these taxa, deal. occupying the northern Daba and Qinling Mountains and rubi. (e) extending into the Yuanjiang–Red River region of northern Vietnam, suggesting secondary contact after historical isolation. These observations parallel findings in other butterfly complexes, such as Aglais urticae, where phylogenomic analyses have revealed cryptic hybridization and reticulate evolution [36].

Pleistocene glaciations as drivers of diversification in P. bootes

The Quaternary glacial period, characterized by extreme climatic fluctuations, spanned from the Early Pleistocene (2.58 million years ago) to about 15,000 years ago [37]. During this period, alternating glacial–interglacial cycles profoundly reshaped global species distributions, diversification patterns, and survival histories [38]. Such cyclical climatic changes acted as powerful regulators of population dynamics and genetic structure in temperate organisms by repeatedly promoting the formation of refugia and dispersal corridors [39]. Biotic refugia served as climatic buffers that ensured long-term persistence during glacial maxima [40], while postglacial expansions from these refugia typically proceeded northward as climates ameliorated [41]. Consequently, northern populations often experienced stronger selective pressures for environmental adaptation compared with southern refugial populations.

Consistent with these insights, the demographic history inferred from PSMC and MSMC analyses highlights the central role of Pleistocene climatic oscillations in shaping the diversification of P. bootes. Divergence among the major lineages of P. bootes coincided with two major glacial events: the Naynayxungla Glaciation (0.78–0.50 Ma) and the Penultimate Glaciation (0.30–0.13 Ma). These glacial periods likely reduced habitat connectivity and promoted population isolation across the complex topography inhabited by P. bootes. The temporal coincidence of lineage splits with glacial maxima suggests that climatic fluctuations imposed repeated cycles of fragmentation and reconnection, consistent with a “refugial isolation and expansion” model proposed for other East Asian montane taxa [42].

The demographic histories of the northern deal., the southern xamn., and the central lineages further support this scenario. Each lineage experienced population growth during interglacial periods, followed by contractions during glacial maxima. The relatively stable effective population size of the northern deal. lineage during the Dali Glaciation (0.07–0.01 Ma) may reflect its adaptation to colder environments. In contrast, the southern xamn. and central lineages experienced severe bottlenecks, likely because they originated in low-latitude, warmer regions and were less adapted to cold conditions, making them more vulnerable during glacial periods.

Hybridization signals in nigr. provide additional evidence of post-glacial contact among previously isolated populations. Model-based demographic inference identified a recent hybrid origin for nigr. at approximately 0.03 Ma, after the divergence of rubi. (e) and xamn. groups around 0.12 Ma. Such secondary contact events following glacial retreat likely facilitated gene flow across lineage boundaries, producing the complex genomic mosaics observed in present-day populations. This scenario assembles with similar postglacial hybridization patterns observed in Polyommatus coridon [43] and Geometridae species [44], underscoring the repeated role of Quaternary climatic oscillations in generating both diversification and admixture within Lepidoptera [45].

Accelerated differentiation and selective dynamics on the Z chromosome

In Lepidoptera, the hemizygous state of the Z chromosome in females (ZW) reduces its effective population size, thereby amplifying the effects of genetic drift and selection relative to autosomes [46]. In addition, the extensive suppression of recombination across large regions of the Z chromosome can promote the accumulation of locally adaptive alleles and strengthen linked selection [47]. Together, these factors give rise to the so-called fast-Z effect, whereby genes located on the Z chromosome evolve more rapidly than those on autosomes [19]. Our genomic analyses of P. bootes lineages provide clear support for this hypothesis. The Z chromosome consistently exhibited higher F_ST and D_XY values than autosomes, particularly between the northern deal. group and the others. This disproportionate divergence closely parallels patterns observed in many other Lepidopteran taxa [48], reflecting the distinctive inheritance and evolutionary dynamics of the Z chromosome.

Sex chromosomes are considered to play a key role in establishing reproductive barriers and speciation processes [49, 50]. Therefore, elevated Z-chromosome differentiation may reflect early stages of speciation, where sex-linked incompatibilities accumulate faster than autosomal ones [51, 52]. Correspondingly, we observed extensive homozygous variations across the entire Z chromosome in the northern deal. group, along with clear signatures of selective sweeps (Tajima’s D < 0), suggesting adaptive evolution potentially associated with environmental traits. In contrast, other P. bootes lineages exhibited patterns of balancing selection (Tajima’s D > 0) on the Z chromosome, suggesting ongoing maintenance of genetic diversity, possibly associated with polymorphic traits such as mimicry and colour variation. The deal. group may therefore represent a population undergoing incipient speciation, with genetic differentiation first emerging on the sex chromosome. These findings are consistent with the “large-Z effect,” which posits that sex chromosomes play a disproportionately important role in species divergence and the establishment of reproductive barriers [53, 54]. Consequently, in P. bootes, the Z chromosome likely serves both as a repository of adaptive variation and as a genomic foundation for the early stages of speciation.

Genetic basis and evolutionary significance of hindwing spot polymorphism in Papilio bootes

The hindwing spot polymorphism of P. bootes is one of the most visually interesting aspects of its phenotypic diversity. The distinct contrast between black-winged (deal. and nigr.) and white-spotted groups (mind., parc., rubi. and xamn.) provides an opportunity to investigate the genetic underpinnings of adaptive colouration within a sexually monomorphic species.

Our genome-wide analyses identified two candidate genes, spalt and fpps, that may be associated with hindwing spot polymorphism in P. bootes. Both genes have been reported to be involved in melanin synthesis and wing-pattern formation of Lepidoptera [34, 35]. The spalt transcription factor family regulates spatial patterning during wing development, controlling the distribution of pigment cell precursors [55], while fpps is essential in the mevalonate pathway for synthesizing isoprenoid compounds used in cuticular pigmentation [56]. Although these two genes emerge as plausible candidates, our inferences are based primarily on comparative genomic analyses under a predefined analytical framework and do not yet establish a direct causal relationship between either gene and the hindwing spot polymorphism of P. bootes. Therefore, the results presented here should be interpreted cautiously.

Notably, the selection signals of spalt and fpps were largely confined to their non-coding regions, suggesting that changes in gene regulation, rather than alterations in protein function, may contribute to the coloration polymorphism of P. bootes. This pattern is consistent with growing evidence that adaptive wing-pattern diversification in butterflies is often associated with evolutionary changes in non-coding regions [57]. In addition, the pronounced sweep-like signals detected around these loci in the black-winged groups imply that the black phenotype may have experienced strong selection. One possible explanation is Batesian mimicry, as the black morphs of P. bootes may imitate sympatrically co-occurring toxic black-winged Byasa species [9], thereby gaining protection from visually hunting predators. Alternatively, the dark pigmentation in the deal. and nigr. groups may provide a thermoregulatory advantage in colder environments, given that both occur in the northernmost part of the species’ range. Similar associations between darker pigmentation and colder conditions have been reported in Pontia occidentalis, in which darker individuals are more common in cool spring environments and may benefit from enhanced heat absorption [27]. Sexual selection may also contribute indirectly to the maintenance of this trait if mate preference is associated with wing coloration, as has been documented in other butterflies [58]. Taken together, these observations suggest that the black-wing phenotype of P. bootes is likely shaped by natural selection, although the exact selective mechanism remains unresolved.

In contrast to the classic Papilio mimicry systems, which are typically female-limited and often controlled by major-effect switch loci such as dsx [28–30], engrailed–invected [31], or cortex [32], the hindwing spot polymorphism of P. bootes appears to use a different genetic mechanism. Because P. bootes is sexually monomorphic and its phenotypic variation exists among lineages rather than between sexes, its maintenance is likely shaped more directly by ecological selection than by sex-specific selective conflict [59–61]. Although the evidence remains preliminary, the candidate genes identified here, spalt and fpps, suggest that the genetic mechanisms underlying wing-pattern diversification in Papilio may be more diverse than previously thought and remain far from fully understood. Our findings therefore broaden the current view of wing-pattern evolution in Papilio by indicating that similar phenotypic contrasts may arise and be maintained through distinct genetic architectures and selective contexts.

Conclusions

Whole-genome SNP analyses of P. bootes revealed pronounced intraspecific genetic structure, identifying three major genetic lineages: northern, central, and southern. This genetic subdivision does not fully correspond to the traditional subspecies classification based on wing morphology and coloration, suggesting that morphological variation alone does not adequately reflect the evolutionary relationships within this species. Demographic reconstructions indicated that divergence among these lineages occurred during two major Pleistocene glaciations, the Naynayxungla and Penultimate glaciations, highlighting the important role of glacial cycles in shaping the population history of P. bootes. Historical gene flow analyses further showed that the nigr. group is a hybrid lineage formed through admixture between the deal. and rubi. (e) groups, indicating that inter-lineage gene exchange has been an important component of the species’ evolutionary history. Notably, the Z chromosome in the northern deal. group exhibited extensive genetic differentiation and elevated homozygosity. These patterns suggest that the Z chromosome may have played an important role in promoting early-stage intraspecific divergence. In addition, genome-wide selection scans comparing black-winged and white-spotted individuals detected strong selection signals near spalt and fpps, suggesting that these genes may be associated with wing-spot polymorphism. Collectively, this study integrates phylogenetic, demographic, and selection-based genomic evidence to show how multiple forces, including Pleistocene climatic oscillations, historical hybridization, and sex-linked selection, have jointly shaped lineage diversification, genomic differentiation, and wing-pattern evolution in P. bootes.

Materials and methods

Sample collection, DNA extraction, library construction and sequencing

An adult male P. b. mindoni individual was collected from the wild in Yingjiang, Dehong, Yunnan Prov., China, for PacBio HiFi and Hi-C sequencing. The head, thorax, and abdomen were dissected, rinsed with sterile water, and immediately frozen in liquid nitrogen. High-molecular-weight genomic DNA was extracted from head and thorax using the TIANGEN Blood & Tissue Kit (TIANGEN, Beijing, China), and its quality and concentration were assessed by 1.0% TBE agarose gel electrophoresis and a Qubit 3.0 Fluorometer. For PacBio HiFi sequencing, 20-kb SMRTbell libraries were constructed using the SMRTbell Express Template Prep Kit 2.0 (Pacific Biosciences) and sequenced on the PacBio Sequel II System in HiFi mode to generate consensus reads. Hi-C libraries were prepared from abdomen tissue following the standard Hi-C protocol [62], and 150-bp paired-end (PE150) reads were generated on the Illumina NovaSeq platform.

In addition, whole-genome resequencing was performed for 66 P. bootes individuals and 12 P. janaka individuals (Additional File 2: Table S1; Additional File 1: Fig. S1-S2). Genomic DNA was extracted from legs, abdomens, or thoraxes using the TIANGEN Blood & Tissue Kit. For abdominal samples, the digestive tract was carefully removed prior to DNA extraction to minimize exogenous DNA contamination. Illumina sequencing libraries with insert sizes of 300–500 bp were constructed using the NEBNext Ultra DNA Library Prep Kit (New England Biolabs, USA) and sequenced in paired-end 150 bp (PE150) mode on the Illumina NovaSeq platform.

Genome assembly and annotation

Hifiasm (version 0.16) [63] was used to assemble the HiFi reads with default parameters and PURGE_DUPS (version 1.2.5) [64] was used to identify and remove redundant contigs. The initially assembled contigs were then anchored to the chromosomes by Hi-C reads using Juicer (version 1.5.7) [65] and 3D-DNA (version 180922) [66] and errors were fixed with the Juicebox Assembly Tools [67]. To validate the accuracy of our genome assembly, we aligned the P. bootes genome to the genomic sequences of P. machaon using Mummer (version 4.0) [68] with the parameters “–mum –mincluster 500”. The sex chromosome was initially identified based on its average mapping coverage, which was approximately half that of the autosomes, and was further confirmed by genomic collinearity analyses.

Tandem repeats were identified using Repeats Finder (version 4.04) [69], while RepeatMasker (version 4.0.7) [70] and RepeatModeler (version 1.0.11) [71] were used to detect known Papilio repeats and to perform de novo repeat prediction. Long terminal repeat (LTR) retrotransposons were identified using LTR_retriever (version 2.9.0) [72]. Protein-coding genes were predicted using the GeMoMa pipeline (version 1.6.1) [73] in combination with the search engine MMseqs2 (version 13.45111) [74] based on related species listed in Additional File 2: Table S9, with filtering parameters set to iAA ≥ 0.8 and ce/rce = 1. TBLASTN (E-value ≤ 1e − 5) (version 2.11.0) [75] and GeneWise (version 2–4-1) [76] were used to identify the best-hit regions in the genome. De novo gene prediction was performed using Augustus (version 3.3.1) [77], GlimmerHMM (version 3.0.4) [78] and SNAP (version 2017–03-01) [79]. The final gene set was generated by integrating homolog-based and de novo predictions using EVidenceModeler (version 1.1.1) [80]. Annotation completeness was assessed with BUSCO (version 4.1.2) [81] using the lepidoptera_odb10 dataset.

Reads filtering and SNP calling

Illumina short reads were mapped to the P. bootes reference genome using BWA (version 0.7.17-r1188) [82] with default parameters. The resulting BAM files were sorted, and duplicate reads were removed using Samtools (version 1.5) [83] and Picard Tools (version 3.4.0, https://github.com/broadinstitute/picard). Reads with mapping quality < 20 were discarded prior to SNP calling. Single nucleotide polymorphisms (SNPs) were called using the HaplotypeCaller and GenotypeGVCFs modules in the Genome Analysis Toolkit (version 3.8) [84]. SNPs were removed if they met any of the following criteria: (1) QD < 2.0, QUAL < 30.0, FS > 60.0, MQ < 40.0, MQRankSum < − 12.5, ReadPosRankSum < − 8.0, or SOR > 3.0; (2) mean sequencing depth per site was < 1/3 × or > 3 × of the average depth across individuals; (3) missing rate > 0.1; or (4) minor allele frequency (MAF) < 0.05. Filtered SNPs were annotated and predicted their functional effects using SnpEff (version 5.0) [85]. All filtered SNPs were exported in Variant Call Format (VCF) files for downstream analyses.

Phylogeny construction and population genetic structure

The VCF file was converted into a PHYLIP matrix using vcf2phylip v2.6 (https://github.com/edgardomortiz/vcf2phylip), retaining only variable sites and generating an alignment with 6.52 Mb sites. This alignment was used to construct a Neighbour-Net phylogenetic network with SplitsTree (version 6.5.1) [86] and a maximum-likelihood (ML) phylogeny with RAxML (version 8.0) [87] under the GTRGAMMA model and 1,000 bootstrap replicates. Principal component analysis (PCA) was performed using SmartPCA implemented in the EIGENSOFT (version 6.1.4) [88]. ADMIXTURE (version 1.3.0) [89] was used to infer population structure across all P. bootes populations, testing ancestral cluster numbers (K) from 2 to 8.

Demographic and divergence histories and gene flow analyses

Changes in effective population size (Nₑ) over time were inferred using PSMC (version 0.6.5-r67) [90]. Individuals with high sequencing depth were selected to ensure high-quality consensus sequences. The parameters were set to -N25, t15, -r5, -p ‘4 + 25*2 + 4 + 6’. The mutation rate of P. bootes was set to 3 × 10⁻⁹ per site per generation [91], with a generation time of one year. MSMC2 (version 2.1.1) [92] was used to infer population separation history. SNP sites were phased using Beagle (version 5.2) [93], and the same samples were used to estimate the relative cross-coalescence rate (RCCR), with a threshold of 0.5 applied to estimate divergence times between populations.

To evaluate patterns of gene flow, we performed D-statistic analyses using Dsuite (version 0.4-r43) [94]. The D and f₄ ratios statistics among all populations were calculated with the Ditros module. D and f statistics were further computed for each branch of the population tree using the Fbranch module. To identify genomic regions potentially affected by introgression, the genome was divided into non-overlapping 50 kb windows, and a maximum-likelihood tree was constructed for each window using RAxML with GTRGAMMA model.

Hybridization events were tested using HyDe (version 1.0.2) [95], which estimates the admixture proportion (γ; where the hybrid population is sister to P1 [1–γ] or P2 [γ]) and corresponding p-values. A coalescent-based demographic reconstruction was conducted using fastsimcoal2 (version 2.8) [96] under the composite likelihood framework. A folded site frequency spectrum (SFS) was generated using easySFS (version 0.0.1) [97]. We designed nine alternative demographic models incorporating divergence and hybridization histories with varying levels of gene flow (Additional File 1: Fig. S11). Parameter estimation was based on 100,000 simulations for each likelihood estimation and 40 cycles of the likelihood maximisation algorithm. For each candidate model, fastsimcoal2 was run with 100 independent replicates, and the best-fitting model was selected according to maximum likelihood values and Akaike Information Criterion (AIC) scores.

Detection of selection signatures in P. bootes populations

We identified genomic regions under selection in P. bootes based on F_ST [98], D_xy [99] and population branch statistic (PBS) [100]. The F_ST and D_XY were calculated in 10-kb sliding windows to quantify genetic differentiation among the three regional P. bootes groups (central, xamn., and deal.) using the script ‘popgenWindows.py’ (https://github.com/simonhmartin/). PBS was used to detect highly differentiated genomic regions. For each 10-kb window, the PBS statistic was computed as: PBS = (T^dx + T^dc − T^xc)/2, where T = − log (1 − F_ST). Here, T^ij represents the estimated branch length between populations i and j based on pairwise F_ST values calculated by popgenWindows.py. The superscripts d, x, and c correspond to deal. (target population), xamn. (first reference population), and central (second reference population), respectively. The PBS value for a population reflects the degree of allele frequency change at a specific locus since its divergence from the other two populations. Additionally, to identify loci associated with wing colour variation, we calculated F_ST and nucleotide diversity (π) between black-winged populations (deal. and nigr.) and white-winged populations (central and xamn.).

Inbreeding and genetic load analysis

Runs of homozygosity (ROHs) were identified using PLINK (version 1.9) [101] with the following parameters: ‘–homozyg -homozyg-window-snp 20 –homozygwindow- het 1 -homozyg-kb 100’. The genomic inbreeding coefficient (F_ROH) was calculated as the ratio of the total ROH length to the total genome length. Genome-wide heterozygosity was estimated as the proportion of heterozygous sites across the genome using PLINK with the ‘–het’ parameter. Tajima’s D was computed with Pixy (version 2.0.0) [102] in 10-kb sliding windows.

Abbreviations

N.E.

Northeast

dsx

doublesex

HiFi

High Fidelity

Hi-C

High-throughput chromosome conformation capture

Authors’ contributions

Peng Zhang and Shao-Ji Hu conceived and supervised the study. Shao-Ji Hu and Adam Miles Cotton provided the samples. Jingyi Zeng and Hao Li performed the experiments and processed the raw sequencing data. Xiao Xu conducted the data analyses. Peng Zhang and Xiao Xu drafted the manuscript, and Dan Liang contributed to manuscript revision. All authors discussed the results, provided critical feedback, and approved the final version of the manuscript.

Funding

This work is supported by the Outstanding Youth Team Project of the Natural Science Foundation of Guangdong Province (No. 2023B1515040002) and the National Natural Science Foundation of China (Grants No. 32370541, 32370477, 42561013).

Data availability

Genome assembly has been deposited in the Genome Warehouse under accession number GWHGSFO00000000. All samples have been deposited in the National Genomics Data Center under BioProject PRJCA048768. Raw sequencing data has been deposited in the Genome Sequence Archive with accession number CRA031752.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.