Divergent genome fractionation and adaptation following polyploidy in the Araliaceae

Abstract

Cyclic polyploidization/diploidization is a driving force in the evolution of eukaryotes. It remains unclear how post-polyploidy reshuffling of ancestral genomes is correlated with evolutionary and ecological diversification. The Araliaceae consists of > 1500 diverse species of herbs, shrubs and trees, which have undergone several polyploidization/diploidization events and have adapted to different forest ecosystem habitats. Here, by leveraging 20 genomes/subgenomes and 561 transcriptomes, we demonstrate that the genome architecture and transcriptional landscape of extant Araliaceae have undergone remodeling associated with polyploidizations/diploidizations. Phenotypic comparisons revealed that phylogenetically distinct species of the Araliaceae have repeatedly evolved shade- and open-habitat phenotypes, accompanied by leaf morphology, efficiency of photosynthesis and carbon fixation. Biased genetic fractionation and transcriptional remodeling of the photomorphogenesis (e.g. PhyB and PIFs) and photosynthesis (e.g. Lhcb-1 and Rubisco) networks are associated with this ecological adaptation. Our study provides evidence that post-polyploidy reshuffling and fractionation correlate with phenotypic and ecological diversification.

Polyploidy is a driving force in eukaryotic evolution. Here the authors describe a macroecological-macrogenomic approach to explore how post-polyploidy reshuffling of an ancestral genome in Araliaceae correlates with phenotypic diversity and ecological adaptation.

Introduction

The evolutionary mechanisms underlying phenotypic innovation and species diversification are an enduring fundamental question in biology^1,2. Polyploidy, or whole genome duplication (WGD), has been increasingly recognized as a creative evolutionary mechanism that has impacted all eukaryotes, and especially higher plants³. It has been proposed that the creative capacity of WGD stems from the combinatorial complexity arising from polyploidy per se as well as the post-polyploidy diploidization processes, whereby divergent genetic and epigenetic variants are generated^4–6. These generalities notwithstanding, the functional and adaptive connections between polyploidization/(re)diploidization-induced variation and evolutionary innovation remain mostly elusive⁷.

At the beginning of the 20th century, phenotypic and ecological evidence observed in neo-polyploid species indicates that genome doubling (or polyploidization) per se is an important mechanism to produce advantageous phenotypes^8–10. Yet it is clear that diploidization follows polyploidization, and that a great deal of cladogenesis arises within highly diploidized groups¹¹. Accumulated ecological and phylogenetic data support the important roles of polyploidization and the attendant rediploidization process in phenotypic innovation and species diversification^3–6. In particular, modern genomic data from diverse model species demonstrate that genomic architecture and epigenetic landscape are extensively reshaped by different evolutionary mechanisms both during and after polyploidization, such as gene deletion, neofunctionalization, subfunctionalization, expression dominance, and dosage balance modulation^4,12. Over the longer term, bursts of species diversification often follow polyploidization, sometimes following a long “lag”, and accordingly, repeated polyploidizations/(re)diploidizations may have increased diversification rates of angiosperms^6,13–15. Indeed, increasing evidence from ecological and genomic data has confirmed the important contribution of polyploidization in both short- and long-term evolutionary processes^3,16–18. However, one approach to better understand the importance of WGD and diploidization in diversification is to study large and diverse clades that have adapted to various environments over long timescales.

The Araliaceae contains > 1500 species of herbs, shrubs and trees that diversified from a common ancestor ~65 million years ago (Mya)^19–21. Phenotypic innovations during polyploidization/(re)diploidization conferred descendant species high adaptability to occupy different ecological habitats of the forest ecosystem in the northern hemisphere^21–24. Here we selected 12 species from three major lineages (i.e., “Asian Palmate”, Aralia and Panax) of the Araliaceae to represent herb, shrub and tree species. In the Asian Palmate lineage, both Kalopanax septemlobus and Tetrapanax papyrifer are arboreal species that occupy the forest canopy layer and are characterized by simple, palmately lobed leaves and compound umbel inflorescences. In contrast, Eleutherococcus senticosus and Schefflera arboricola are shrub species that inhabit the mid-canopy layer and possess palmately compound leaves and umbel inflorescences. In the Aralia lineage, both A. elata and A. dasyphylla are arboreal trees with pinnate leaves and compound umbel inflorescences. However, A. continentalis is a herbaceous species that grows in shaded habitats on the floor layer of the forest ecosystem. In the Panax lineage, all species are shade-tolerant herbs that occupy the floor layer and are characterized by palmately compound leaves and solitary umbel inflorescences. In summary, phenotypic changes of the life form (i.e., tree, shrub and herb), leaf shape (i.e., simple, palmately and pinnately compound) and inflorescence structure (i.e., simple and compound umbel) have evolved repeatedly in the three lineages of Araliaceae. These attributes demonstrate that the most recent common ancestor of the Araliaceae has diversified into phenotypically distinct descendants that occupy different ecological layers of the forest ecosystem, and with varying morphologies.

The aim of this study was to address whether the reorganization and fractionation of the duplicated ancestral genomes accompanying cycles of polyploidization/(re)diploidization (referred to as post-polyploidy reshuffling) are associated with genomic variants that linked to phenotypic and ecological diversification of the different clades mentioned above. To address this, we subjected the chosen Araliaceae species and an outgroup species (Daucus carota) to simulated habitats of the open-canopy (high light and normal light) and shaded-understory (low light, shade light, and fluctuating light). Our integrated analyses of pan-genome and phenotyping revealed that: i) post-polyploidy reshuffling of the ancestral genome represents a notable evolutionary feature associated with the genome architecture and the transcriptional regulation landscape of extant species; ii) herbs, shrubs and trees exhibit different phenotypic performance and gene expression plasticity under the two contrasting habitats; and iii) biased genetic fractionation and transcriptional regulatory remodeling of the photomorphogenesis and photosynthesis networks correlate with divergent ecological adaptations. Our study demonstrates that post-polyploidy reshuffling of the ancestral genome is potentially associated with phenotypic and ecological diversification of extant Araliaceae.

Results

Polyploidization history and phenotypic diversity

The evolutionary history and genomic features of the Araliaceae were inferred based on five newly assembled genomes generated from this study in conjunction with seven previously released genomes^22,23,25,26. Basic descriptions of the five newly assembled genomes are provided in Table 1. In brief, 4′,6-diamidino-2-phenylindole (DAPI) staining revealed a diploid karyotype for A. dasyphylla and tetraploid karyotypes for the other four species (A. continentalis, K. septemlobus, S. arboricola and T. papyrifer) (Supplementary Fig. 1). Genome sizes of the five species ranged from 1.14 to 1.99 Gb, with scaffold N50 values of 44.34–85.87 Mb. All five genome assemblies showed high completeness (BUSCO completeness: 95.91–98.14%) and assembly contiguity (LAI scores: 11.97–25.91) (Table 1, Supplementary Data 1 and Supplementary Figs. 2–5).

Table 1.

Genome features of the five new genome assemblies of the five Araliaceae species

Genomic information	Aralia continentalis	Aralia dasyphylla	Kalopanax septemlobus	Schefflera arboricola	Tetrapanax papyrifer
Karyotype	2n = 4x = 48	2n = 2x = 24	2n = 4x = 48	2n = 4x = 48	2n = 4x = 48
Total contig size (bp)	1,193,193,182	1,138,933,981	1,163,260,069	1,992,865,661	1,218,956,633
Contig N50 (bp)	17,193,698	2,900,000	35,512,153	23,887,292	34,651,251
Total scaffolding length (bp)	1,201,662,065	1,139,187,981	1,163,265,269	1,992,879,361	1,218,960,033
Cumulative length of pseudo-chromosomes (bp)	1,090,685,971	1,033,011,667	1,114,547,615	1,974,405,388	1,089,887,149
Scaffold N50 (bp)	45,475,998	85,866,302	46,356,386	83,602,958	44,339,954
GC content (%)	35.22	36.16	33.53	34.15	34.94
Repeat content (%)	54.86	67.77	55.55	72.48	62.72
Predicted genes	69,390	42,130	55,543	52,918	46,033
Completeness BUSCOs (%)	98.14	97.71	97.71	95.91	96.16
Completeness CEGMA (%)	97.18	95.56	96.51	96.72	96.72
LTR assembly index (LAI)	14.94	18.41	16.64	25.91	11.97

Analyses of the whole-genome scale phylogenetics, genome collinearity, and distribution of synonymous substitutions (Ks) identified one ancient (Pg-β) and three more recent (Ac-α, Pg-α, and Ar-α) WGD events in the Araliaceae (Fig. 1a and Supplementary Figs. 6–8). The ancient WGD event (Pg-β) preceded the diversification of extant Araliaceae species and was accompanied by extensive genome reshuffling, leading to a transition from the pre-Pg-β ancestral karyotype (2n = 2x = 16) to the post-Pg-β karyotype (2n = 2x = 24)^23,24. Thereafter, three additional, more recent WGD events occurred independently in the three lineages (Asian Palmate, Aralia and Panax) at different evolutionary stages. The Ac-α event occurred within the Aralia lineage around 8.08 Mya, leading to the formation of the tetraploid species A. continentalis (post-Ac-α karyotype: 2n = 4x = 48). In the Panax lineage, the Pg-α event led to the formation of an ancestral tetraploid species (post-Pg-α karyotype: 2n = 4x = 48) around 5.04–6.79 Mya, which further split into three extant tetraploid species (P. ginseng, P. quinquefolius and P. japonicus). In contrast, the four woody species of the Asian Palmate lineage (E. senticosus, K. septemlobus, S. arboricola, and T. papyrifer) are descendants of the same ancestor that evolved through a tetraploidization event (Ar-α) (post-Ar-α karyotype: 2n = 4x = 48) around 17.31–26.43 Mya. In addition, phenotypic comparisons based on plant habit, leaf morphology and inflorescence architecture revealed that parallel phenotypic changes occurred among the three lineages (Fig. 1b). For example, parallel phenotypic changes of the life form from tree to shrub or herb are associated with the ecological transitions from open-canopy to shaded understory habitat in the lineages of Asian Palmate and Aralia. We thus conclude that in parallel with two rounds of polyploidization and (re)diploidization, extant Araliaceae species have evolved remarkable phenotypic diversity, which might be correlated with their occupation of distinct ecological niches within forest ecosystems.

Fig. 1. Phylogeny, polyploidization history, phenotypic diversity and genome feature of 12 extant Araliaceae species.

a Phylogeny and divergence time of the Araliaceae was inferred based on single copy genes using ASTRAL, with divergence times estimated by MCMCTree. Purple circles and stars represent ancient- and recent-polyploidization events, respectively. Brown, orange and blue colors of the branches and species names represent the canopy, medium and floor layers of the forest ecosystem. “subA” and “subB” are the two subgenomes of the tetraploid species. b Morphological features of the Araliaceae and outgroup species, including leaf shape (first column: simple but lobed vs. compound), inflorescence structure (second column: umbel vs. compound umbel) and life form (third column: herb, shrub and tree). For leaf shape, only the two tree species, Kalopanax septemlobus and Tetrapanax papyrifer, have simple leaves. The four Aralia species have pinnately compound leaves. All the other Araliaceae species included here have palmately compound leaves. For inflorescence structure, the shrub species Eleutherococcus senticosus and Panax species have simple umbels. All other species have compound umbels. For life form, all Panax species and A. continentalis are herbs. The other species are trees or shrubs. c Genome constitution of the 20 genomes of the 12 Araliaceae species. The orange colors from light to dark indicate different types of repeat sequence, including retrotransposon, transposon, tandem repeat and unclear repeat. The blue colors from light to dark are dispensable, unique and core gene families. Grey color represents protein-coding genes that are not classified into gene families. Numbers on X-axis represent percentages of each category of sequence. d Proportion of the five gene duplication modes and unknown genes, including whole-genome duplication (WGD), transposed duplication (TRD), dispersed duplication (DSD), tandem duplication (TD), proximal duplication (PD), and others. The area of each bar (height × width) is proportional to the percentage of each gene type. e Circle size is the genome size of the 20 genomes of the 12 Araliaceae species.

Evolutionary dynamics of genome architecture

To address if genome architecture and gene content were reshaped during the repeated paleo- and neo-polyploidizations, we performed pan-genome analyses of 20 genomes (four genomes of four diploid and 16 subgenomes of eight tetraploid species) for both repeat sequences and protein-coding genes. Our results indicate that, during the post-polyploidy reshuffling of the ancestral genome, independent expansion of repeat sequences (from 0.29 Gb in A. continentalis to 2.04 Gb in P. notoginseng) and biased fractionation of protein-coding genes (from 17,581 in the subgenome B of E. senticosus to 47,870 in P. notoginseng) are associated with genomic divergence among extant Araliaceae (Fig. 1c–e and Supplementary Data 1–2). For the repeat sequences, rapid amplification of retrotransposons is inferred to have occurred in all extant Araliaceae lineages, with the Panax lineage (0.56–1.72 Gb in length) experiencing more recent retrotransposon expansion compared to the Aralia (0.27–0.69 Gb) and Asian Palmate lineages (0.26–0.66 Gb) (Supplementary Fig. 9 and Supplementary Data 2). For protein-coding genes, a total of 35,799 gene families were identified in the Araliaceae genomes/subgenomes and four outgroup species, only 3,875 (10.8%) of which are core gene families that are present in all of these genomes (Fig. 1c and Supplementary Data 3), demonstrating a remarkable amount of gene content evolution in the Araliaceae clade. In addition, the Asian Palmate lineage possessed relatively fewer dispensable (7,802–11,480) and unique (12–56) genes compared to the lineages of Aralia (dispensable: 12,267–14,046, unique: 56–274) and Panax (dispensable: 11,868–14,651, unique: 61–728). Different gene numbers identified among the three lineages were also observed as a common phenomenon in the other gene categories (Supplementary Data 4–5). For example, the Panax lineage contains relatively higher proportions of transposed duplication (TRD) (26.56–37.26%) and dispersed duplication (DSD) (11.40–21.29%) genes compared to the Aralia (TRD: 21.93–24.00%, DSD: 7.21–11.23%) and Asian Palmate lineages (TRD: 22.75–28.24%, DSD: 8.64–11.67%).

Functional enrichment analyses of protein-coding genes annotated a total of 2394 gene ontology (GO) terms, of which, 1532 (63.99%) are core across all extant Araliaceae (Supplementary Data 6). It is notable that GO terms related to photosynthesis (i.e., light reaction, carotenoid biosynthetic process, and electron and proton transport), photomorphogenesis, and carbon fixation (i.e., Calvin-Benson-Bassham cycle and tricarboxylic acid cycle) were enriched in the expanded gene families of the Panax lineage and other herbaceous species (A. continentalis and D. carota) (Supplementary Data 7). In contrast, GO terms that are mainly identified in woody species (the lineages of Asian Palmate and Aralia) included defense response to fungi and other organisms, specific tissue (i.e., root development, xylem development, and meristem) development, and endoplasmic reticulum (i.e., endopeptidase inhibitor activity, endoplasmic reticulum membrane, and response to endoplasmic reticulum stress) (Supplementary Data 7). These attributes suggest that concomitant with two rounds of paleo- and neo-polyploidizations, extant Araliaceae species experienced highly dynamic changes in genome architecture and gene complements.

Post-polyploidy reshuffling of the duplicated ancestral genome

The above pangenome analyses revealed high evolutionary dynamics of genome architecture among extant Araliaceae. Next, we evaluated how this post-polyploidy reshuffling is related to presumptive functional genomic diversity of extant Araliaceae. Our previous study has demonstrated that the ancestral genome of Araliaceae (pre-Pg-β karyotype: 2n = 2x = 16) was doubled by ancient Pg-β WGD (2n = 4x = 32), followed by diploidization of this tetraploid ancestor, leading to the modern Araliaceae karyotype (2n = 2x = 24)²³. Here we further describe this polyploidization/(re)diploidization process with genomic analysis of seven additional Araliaceae species (Supplementary Fig. 10 and Supplementary Data 8). A general pattern was that the eight ancestral-chromosome pairs (duplicated by Pg-β event) showed different degrees of divergence at gene numbers in the 19 modern Araliaceae genomes (two-sided Wilcoxon rank-sum test, p-value < 0.05) (Fig. 2a and Supplementary Fig. 11). In particular, different gene categories showed distinct distribution patterns along the ancestral chromosomes across the 19 modern genomes (Supplementary Fig. 12). A similar variation pattern was also observed along the 12 modern chromosomes where dynamic changes of the ancestral gene complement was identified across these species (Supplementary Fig. 13). For the repeated sequences, similar expansion trend of LTR-retrotransposons was observed among the eight ancestral chromosome pairs in all these extant species, but distinct expansion patterns were identified both between the subgenomes of tetraploid species and in inter-specific comparisons (Supplementary Fig. 9).

Fig. 2. Evolutionary dynamics and inferred functional diversity of protein-coding genes in the eight post-Pg-β chromosome pairs among extant Araliaceae species.

a Number of retained genes distributed along the eight post-Pg-β chromosome pairs in extant Araliaceae species. From left to right, the five columns represent whole genome (genome) and four gene categories (unique, rapid expansion, WGD and collinear). The eight chromosome pairs are named from top to bottom as Ar1a/Ar1b, Ar2a/Ar2b, Ar3a/Ar3b, Ar4a/Ar4b, Ar5a/Ar5b, Ar6a/Ar6b, Ar7a/Ar7b, and Ar8a/Ar8b. For each of the eight chromosome pairs, blue (above) and brown (below) colors are the numbers of retained gene identified in ancestral chromosomes “a” and “b”, respectively. The Y axis is the number of retained genes. The X axis represents the modern genomes of extant Araliaceae species. Within each of the five columns, species names from left to right at the bottom are: Aralia continentalis (subA and subB), Aralia dasyphylla, Aralia elata, Panax ginseng (subA and subB), Panax japonicus (subA and subB), Panax quinquefolius (subA and subB), Panax stipuleanatus, Eleutherococcus senticosus (subA and subB), Kalopanax septemlobus (subA and subB), Schefflera arboricola (subA and subB), and Tetrapanax papyrifer (subA and subB). Species names shown in red indicate statistically significant differences in gene numbers (two-sided Wilcoxon rank-sum test, p-value < 0.05). b Functional enrichment analyses of the retained genes in extant Araliaceae species. The five columns from left to right represent whole genome (genome) and four gene categories (unique, rapid expansion, WGD and collinear). Within each column, the order of species names is the same as in subpanel a. Each row represents a GO term. Only the top 5% GO terms that exhibit high divergence among modern genomes are shown. Colors of the heatmap from red, orange to blue represent numbers of enriched genes from high to low. Detailed information regarding GO enrichment on the eight post-Pg-β chromosome pairs are included in Supplementary Data 9. Species names in red color indicate the statistical significance of the gene number (two-sided Wilcoxon rank-sum test, p-value < 0.05).

Functional enrichment analyses of different gene categories, such as expanded/contracted gene families and gene duplicates, also identified distinct enrichment patterns among the 19 modern Araliaceae genomes. At the whole genome level, while 1532 (63.99%) and 2348 (98.08%) of the 2394 annotated GO terms are core (present in all species) and dispensable (present in at least one but not all species), the eight ancestral-chromosome pairs exhibited different degrees of functional divergence among extant Araliaceae (two-sided Wilcoxon rank-sum test, p-value < 0.05) (Fig. 2b, Supplementary Figs. 14–15 and Supplementary Data 9). In particular, GO terms that contained highly divergent gene numbers among extant Araliaceae are functionally related to ecological adaptation (i.e., photosystem II, defense response, and response to oxidative stress), molecular binding (i.e., protein binding, nucleic acid binding, RNA binding, and ATP binding), cellular structure (i.e., membrane, ribosome, nucleus, and microtubule), and basic cellular activities (i.e., translation, DNA integration, proteolysis, and signal transduction) (Supplementary Data 10). For each gene category, biased fractionation acting on duplicated genes of specific GO terms also generated gene content divergence among the extant Araliaceae (Fig. 2b, Supplementary Fig. 14, and Supplementary Data 9). Among the core, dispensable and unique gene families, for example, high divergence of the photosynthesis-related GO terms (i.e., photosystem II and photosystem II oxygen evolving complex) primarily arose from biased genetic fractionation of dispensable and unique gene families rather than of core gene families (Supplementary Data 10). These findings indicate that post-polyploidy reshuffling of the duplicated ancestral genome is associated with the divergence and molecular functional diversity of extant Araliaceae clades.

Independent evolution of the shaded- and open-habitat phenotypes

All extant Araliaceae originated from a common ancestor but they evolved distinct phenotypes and are adapted to different ecological layers of the forest ecosystem (see in Fig. 1a–b). We thus asked how this divergent adaptation to shaded-understory and open-canopy habitats is related to genomic composition and differential fractionation following polyploidization. Given that light quantity and quality are key environmental factors influencing plant growth and development, we evaluated morphological and physiological traits of three tree (A. elata, K. septemlobus, and T. papyrifer), two shrub (E. senticosus and S. arboricola), and three herbaceous (A. continentalis, P. ginseng, and P. quinquefolius) species, alongside one herbaceous outgroup (D. carota) species. The logic underlying this comparison was that these Araliaceae species have independently evolved shaded- and open-habitat phenotypes (defined as high adaptability to shaded-understory or open-canopy habitats) to adapt to the two contrasting habitats (Fig. 3a). We grew these plants under simulated open-canopy (high light and normal light) and shaded-understory (low light, shade light, and fluctuating light) habitats (see Methods). For leaf morphology, compared to normal light conditions, both seedlings and adult plants of the Araliaceae and outgroup species showed different degrees of phenotypic plasticity after long-term exposure to alternative light conditions, i.e., either open-canopy (high light) or shaded-understory (low light and shade light) habitats, as reflected by leaf shape, specific leaf area, and hypocotyl and petiole length (Supplementary Figs. 16–17 and Supplementary Data 11). For total chlorophyll content, after long-term exposure to under five light conditions of the two contrasting habitats, all eight Araliaceae and outgroup species exhibited reduced content under high light, shade light and fluctuating light, compared to that under normal light (two-sided Wilcoxon rank-sum test, p-value < 0.05) (Supplementary Fig. 17). Under low light condition, however, the three herbaceous species (A. continentalis, P. ginseng and P. quinquefolius) and two shrub species (E. senticosus and S. arboricola) accumulated higher total chlorophyll content relative to that under normal light (two-sided Wilcoxon rank-sum test, p-value < 0.05). These results confirm that light quantity and quality play important roles in regulating plant growth and development under open-canopy and shaded-understory habitats^27,28.

Fig. 3. Morphological and physiological features of Araliaceae and outgroup species grown under simulated open-canopy and shaded-understory habitats.

a Divergent ecological adaptation of eight Araliaceae to canopy, medium and floor layers of the forest ecosystem. Colors of the tree branches and ovals represent the three ecological layers of the forest ecosystem. Within each oval, six representative phenotypes were selected to show divergent ecological adaptation of the eight Araliaceae and outgroup species to shaded-understory and open-canopy habitats. Font colors of these phenotypes represent the adaptive phenotypes that defined by the efficiency of these traits under shaded-understory and open-canopy habitat. Brown, blue and grey colors are open-habitat phenotype, shaded-habitat phenotype, and no habitat preference, respectively. Plant morphologies of the nine species are shown at the bottom. Colors of the species names represent the three ecological layers of the forest ecosystem. b Light-response curves of the four species (A. elata, P. ginseng, S. arboricola and T. papyrifer) after long-term exposure under four simulated light conditions. For each subpanel, the four colors represent light-response curves of the plants that grown under normal (orange), shade (cyan), high (blue) and low (red) light conditions, respectively. The X and Y axis represent photosynthetically active radiation and photosynthetic rate, respectively. c A-C_i response curves of the four species (A. elata, P. ginseng, S. arboricola and T. papyrifer) fitted using an approach based on the Farquhar-von Caemmerer-Berry (FvCB) model. The four colors represent A-C_i response curves of the plants that grown under normal (orange), shade (cyan), high (blue) and low (red) light conditions, respectively. The X and Y axis represent intercellular CO₂ concentration and net CO₂ assimilation rate, respectively. d The maximum quantum efficiency of PSII (Fv/Fm) of the four species (A. elata, P. ginseng, S. arboricola and T. papyrifer) that grown under normal, shade, high, low and fluctuating light (0 h, 12 h, and 5 days) conditions. Color gradients from red (0) to purple (1) represent the PSII efficiency from low to high in leaf tissue. Statistical significance among different treatments was assessed using the two-sided Wilcoxon rank-sum test (p-value < 0.05). All details are shown in Supplementary Figs. 20–21.

Leaf morphologies and chlorophyll content are thought to correlate with photosynthetic efficiency^29–31. We addressed if these Araliaceae species differ in photosynthetic rate and carbon fixation efficiency in the two contrasting habitats, only one of which corresponds to that experienced in their native settings. The results among species were not consistent with respect to the overall pattern of the shaded and open-habitat phenotypes among the Araliaceae species studied. The herbaceous and shrub species showed relatively higher tolerance to the light stress of shaded-understory compared to the tree species and the outgroup species (Fig. 3a). In the case of photosynthesis rate, both herbaceous species, P. ginseng and P. quinquefolius, which inhabit the floor layer of forest ecosystem, showed relatively higher photosynthetic rates under low and shade light compared to high light stress (Fig. 3b and Supplementary Fig. 18). Among the five woody species, the shrub species S. arboricola exhibited relatively higher photosynthetic rate under low and shade light. For carbon fixation efficiency, both herbaceous species (P. ginseng and P. quinquefolius) showed relatively higher carbon fixation rates (i.e., high Vcmax and Jmax values) under shade and low light compared to high light (Fig. 3c and Supplementary Fig. 19). In contrast, the three tree species and the outgroup species tended to maintain high efficiencies of electron transport and carbon fixation under high light (Supplementary Figs. 18–19). For the maximum quantum efficiency of photosystem II (Fv/Fm), the three herbaceous species exhibited significantly higher Fv/Fm values under low and shade light compared to high light (two-sided Wilcoxon rank-sum test, p-value < 0.01) (Fig. 3d and Supplementary Figs. 20–21). However, the five woody and outgroup species maintained relatively higher Fv/Fm values under high light. In addition, after five days of exposure to fluctuating light, both herbaceous (P. ginseng and P. quinquefolius) and the two shrub species (E. senticosus and S. arboricola) maintained relatively higher Fv/Fm values compared to the other Araliaceae and outgroup species. These findings provide physiological support to the ecological and phylogenetic inferences that the extant Araliaceae species have convergently evolved shaded- and open-habitat phenotypes to adapt to different ecological niches of the forest ecosystem.

Gene expression plasticity and regulation mechanism diversity

To explore the genomic basis underlying the divergent evolution of shaded- and open-habitat phenotypes, gene transcriptional landscapes were evaluated for the same plant samples as used for phenotypic comparisons above. At the overall level, comparative analyses based on 561 transcriptomes revealed different degrees of expression-level diversity (Hj) and specificity (Sj) under the five light conditions in both the Araliaceae species and the outgroup species (D. carota) (Fig. 4a, Supplementary Fig. 22 and Supplementary Data 12). In addition, significant associations between the expression plasticity (Hj and Sj) and life form were detected using the Kruskal-Wallis test (Hj: χ² = 81.54, p-value < 0.001, ε² = 0.1426; Sj: χ² = 81.89, p-value < 0.001, ε² > 0.1432) (Supplementary Fig. 23). In addition, strong associations were also observed between the expression plasticity and ecological habitat (Hj: χ² = 178.36, p-value < 0.001, ε² = 0.3161; Sj: χ² = 85.46, p-value < 0.001, ε² = 0.1496), further supported by Spearman’s correlation analysis (Hj: ρ = 0.46; Sj: ρ = 0.28, both p-value < 0.001) (Supplementary Fig. 23). Ploidy level of these species was also associated with the expression-level plasticity (two-sided Wilcoxon rank-sum test: Hj: r = 0.145; Sj: r = 0.244, both p-value < 0.001) (Supplementary Fig. 23). These findings suggest that variation in light quantity and quality across different ecological habitats, as well as differences in life form strategies, are important ecological factors associated with transcriptional regulation in these Araliaceae species.

Fig. 4. Gene expression plasticity and transcriptional regulatory mechanisms of photomorphogenesis among extant Araliaceae species under simulated open-canopy and shaded-understory habitats.

a Gene expression plasticity of samples collected from P. ginseng and A. elata grown under different light conditions of two contrasting habitats, including fluctuating light, and short-term and long-term exposure under normal, shade, low and high light conditions. The X- and Y-axes represent gene expression diversity (Hj) and specificity (Sj), respectively. Each symbol denotes a sample (three biological replicates) collected from different light conditions and tissues. The Hj value on X-axis reflects gene expression variations among different genes under the same light condition. The Sj value on Y-axis represents expression variations of the same gene among different light conditions and tissues. The shaded regions represent the 95% confidence intervals around the fitted linear regression lines. b Tree map of the enriched GO terms identified in P. ginseng (upper section) and A. elata (lower section) under high light stress, respectively. Each rectangle is a single cluster representative of the enriched GO terms. The representatives are joined into superclusters of loosely related GO terms and visualized with different colors. Area size of the rectangles reflects the adjusted p-value of the corresponding GO term. GO enrichment was assessed using a one-sided hypergeometric test, and p-values were adjusted for multiple comparisons using the Benjamini–Hochberg method. The treemap was generated using REVIGO to visualize and reduce redundancy among enriched GO terms. Different distribution patterns of these cluster representatives between the two species under the same light stress indicate the identified differentially expressed genes (DEGs) show distinct enrichment pattern. c Biased genetic fractionation and transcriptional regulatory remodeling of the photomorphogenesis network. At the top, rounded rectangles within the dashed boxes denote the pre-Pg-β (purple) and post-Pg-β (dark red and dark blue) ancestral genomes, respectively. Physical locations of the three photoreceptors (Phys, Crys and Phots) and two regulators (PIFs and HY5) are identified on the eight ancestral chromosomes. Grey color represents putative genes lost in the diploidization process. In the middle, the species tree indicates the phylogenetic relationships of the eight Araliaceae species. Purple stars are the ancient and recent WGD events. At the bottom, rectangular blocks represent ancestral genomes of extant Araliaceae species that were duplicated by paleo- (dark red and dark blue) and neo-polyploidization (light red and light blue) events. Each of the extant diploid and tetraploid Araliaceae species contain two and four pre-Pg-β ancestral genomes (purple), respectively. The four rows represent transcriptional patterns of the three photoreceptors (Phys, Crys and Phots) and two regulators (PIFs and HY5) in the eight Araliaceae species under the five light conditions. Expression levels of these photoreceptors and regulators are shown in the heatmap. Under normal light condition, colors from red to white indicate expression levels from high to low. For the other four light conditions (high, shade, low and fluctuating light), red and blue colors represent significant up- and down-regulation of these genes compared to normal light condition. Grey and white colors represent gene lost and non-significant up- or down-regulation of these genes. Detailed information of the transcriptional patterns of these photomorphogenic genes is provided in Supplementary Fig. 28.

To identify candidate genes associated with the responses to variable light conditions, 5402–19,887 differentially expressed genes (DEGs) were characterized from the pan-transcriptome data (Supplementary Fig. 24). Functional analyses of these DEGs identified 939 significantly enriched GO terms, 474 and 465 of which were enriched by the down- and up-regulated DEGs, respectively, in the eight Araliaceae species (Supplementary Data 13). Inter-specific comparisons of these enriched GO terms showed that 235 (49.58%) down-regulated and 203 (43.66%) up-regulated GO terms were shared among these Araliaceae species (Supplementary Fig. 25). Yet, some GO terms that are functionally related to habitat preference were identified in the eight Araliaceae species, such as photosynthesis (i.e., light harvesting, oxygen evolving complex, PS-II, PS-I, and chlorophyll and carotenoid biosynthesis) and photomorphogenesis (i.e., regulation of circadian rhythm, phototropism, and blue light signaling pathway) (Supplementary Data 13). Different enrichment patterns of these GO terms were observed in the eight Araliaceae species under the different light conditions of the two contrasting habitats (Fig. 4b and Supplementary Fig. 26). Under high light, for example, GO terms identified in the herbaceous species P. ginseng are functionally related to response to stress and protein folding. In the tree species A. elata, however, GO terms enriched under high light are functionally related to translation and photosynthesis. These attributes suggest that post-polyploidy remodeling of the transcriptional regulatory network is correlated with differential regulatory responses to altered light conditions of the two contrasting habitats.

Reshaping of the photomorphogenesis and photosynthesis regulatory networks

Our previous studies have demonstrated that regulatory networks governing photomorphogenesis and photosynthesis are involved in transcriptional responses to altered light conditions in the Araliaceae^27,28. Indeed, the above pan-genomic and phenotypic analyses also revealed important roles of the photosynthesis and photomorphogenesis networks in regulating plant growth and development under shaded-understory and open-canopy habitats. We then addressed whether the two regulatory networks were reshaped during post-polyploidy reshuffling to respond differently to light conditions of the two contrasting habitats. As shown in Fig. 4c, the ancestral genome (purple color on the top) was duplicated by Pg-β, and then diverged into a “conserved genome” (dark red, more ancestral genes retained) and a “non-conserved genome” (dark blue, more ancestral genes lost) (Supplementary Fig. 27). In the regulatory network for photomorphogenesis, genetic fractionation of the three photoreceptor (Cry, Phot and Phy) and two regulator (HY5 and PIF) genes is correlated with differential retention of members of regulatory network among species (Fig. 4c and Supplementary Figs. 27–28). At the transcriptional regulation level, after long-term exposure to normal light condition, the Phys and Phots localized in the “conserved component” showed relatively higher transcription than copies of the “non-conserved component”. However, after exposure to non-native light conditions (high or low, depending on species), differential up- and down-regulation of the three photoreceptors (Cry, Phot and Phy) and two regulators (HY5 and PIF) were observed in all eight Araliaceae species. For example, both the cryptochrome (i.e., Cry2) and phytochromes (i.e., PhyA and PhyB) were up-regulated in herbaceous species, but only the phytochromes (i.e., PhyA and PhyB) were activated in woody species.

For the photosynthesis regulatory network, we partitioned genes into five groups based on three KEGG modules (ko00195, ko00196 and ko00710), including light capture (LHC-I/II and PS-I/II), non-photochemical quenching (NPQ), electron transport (linear and cyclic electron flow (LEF and CEF)), energy conversion (ATP and NADPH), and carbon fixation (Supplementary Data 14). In general, genes of the five groups (light capture, NPQ, electron transport, energy conversion and carbon fixation) also showed differential up- or down-regulation under shaded-understory and open-canopy habitats (Fig. 5 and Supplementary Fig. 29). For example, after long-term exposure to shade, the majority of the five module genes were up-regulated or maintained a similar expression level in leaf tissue of the herbaceous species P. ginseng compared to normal light conditions. However, up-regulation of the five module genes was identified in leaf tissues of the tree species A. elata under high light. In addition, we further analyzed the regulatory patterns of five genes that were commonly identified as key regulators under the two contrasting habitats, including LHCb-1 (light capture module), psbS (NPQ module), PEPC and rubisco (rbcL and rbcS) (carbon fixation module) (Supplementary Fig. 30). It is notable that gene duplicates derived from paleo- and neo-polyploidizations exhibit different patterns of genetic fractionation and transcriptional regulation among extant Araliaceae (Supplementary Figs. 28–30). In the case of phytochrome A (PhyA), only E. senticosus and K. septemlobus retained the PhyA1 copy at subgenome B that was duplicated by paleo-polyploidization (Pg-β). In contrast, the other species retained the PhyA2 copy, which was further doubled by neo-polyploidizations in the tetraploid species. These observations suggest that WGD and post-polyploidy reshuffling of the ancestral genome are associated with differential response in the two regulatory networks that respond to variable light in the two contrasting habitats.

Fig. 5. Transcriptional landscape of the photosynthesis network in leaf tissue of Panax ginseng (a) and Aralia eleta (b) under simulated open-canopy (normal and high light) and shaded-understory (low, shade and fluctuating light) habitats.

The photosynthesis network is assigned into five modules based on KEGG, including light capture (PS-I and PS-II), NPQ, electron transport, energy conversion (ATP and NADPH) and carbon fixation (Calvin-Benson-Bassham cycle). The five rows from top to bottom represent transcriptional patterns of these photosynthesis genes in leaf tissues after long-term exposure under the five light conditions. Transcription levels of the five module genes are shown by color gradients. Colors from dark to light represent transcription levels from high to low at the five module genes. Within each module, gene names colored red and blue denote significant up- or down-regulation, respectively, under the four stress light conditions (fluctuating, high, low and shade) compared to the normal light. Detailed information of the transcriptional patterns of these photosynthetic genes is shown in Supplementary Figs. 38–40.

Discussion

The evolutionary significance of the cyclic polyploidizations/(re)diploidizations in the diversification of angiosperms has long attracted interest^11,18,32. In the past several decades, the massive increase in genomic studies from model and other species has well-documented the roles of polyploidizations/(re)diploidizations in promoting genome evolution and possibly phenotypic innovation^3,4,16–18, although the latter connections typically remain challenging to prove. Thus, it remains unclear whether the post-polyploidy reshuffling of the ancestral genome causes or is correlated with phenotypic diversity and ecological adaptation. Here we use a macroecological-macrogenomic approach to explore possible answers to these questions, by studying a large plant family that has diversified into approximately 1500 species of herbs, shrubs and trees, themselves products of several rounds of polyploidizations/diploidizations while radiating into different ecological niches of the forest ecosystem^19–21. These attributes of the Araliaceae render it a useful system to address the evolutionary mechanisms and genomic bases of how “a single [ancestral] species diversifies into an entire forest”.

In this study, we selected 12 species from three major lineages of the Araliaceae to represent three life forms (herb, shrub and tree species) and cover different habitats of the forest ecosystem. Our pan-genome analyses based on 20 genomes and 561 transcriptomes suggest that, after multiple rounds of neo- and paleo-polyploidizations (followed by rediploidization), extant Araliaceae exhibit high divergence in genome architecture, content, and transcriptional regulation landscape. In the case of the paleo-polyploidization Pg-β, the eight ancestral-chromosome pairs (duplicated by Pg-β) are reshaped in extant species for both protein-coding genes and repeat sequences. Thereafter, this genome divergence coincided with an additional round of recent polyploidization (e.g., Ac-α, Pg-α, or Ar-α) in the three Araliaceae lineages studied, respectively. In addition, we also document genome-wide biased fractionation of the duplicated genes, which is associated with high genic and presumably molecular functional diversity across extant Araliaceae. These findings support the hypothesis that post-polyploidy reshuffling of ancestral genomes is correlated with divergent adaptation among extant Araliaceae.

Our previous studies demonstrated that both herbaceous and shrub species of the Araliaceae exhibit low phenotypic plasticity and high photosynthesis efficiency under shaded habitats^27,28. To further address if this is a common phenomenon across the Araliaceae, we exposed seedlings and adult plants of eight phylogenetically disparate species under simulated open-canopy and shaded-understory habitats. A general pattern was that the herbaceous and shrub species tended to show relatively higher adaptability to shaded-understory habitats compared to the tree species, as might be expected considering their native habitats under a tree canopy. For example, the two herbaceous (P. ginseng and P. quinquefolius) and one shrub (S. arboricola) species exhibited relatively higher efficiency of photosynthesis (i.e., photosynthetic rate, Fv/Fm, and electron transport) and carbon fixation (i.e., Vcmax and Jmax) in the shaded-understory than the higher light intensity habitat. In contrast, the three woody species (A. elata, K. septemlobus and E. senticosus) exhibited relatively higher tolerance to high light, also as expected from their native habitat. Independent evolution of the shaded- and open-habitat phenotypes in phylogenetically distinct species provides correlational evidence for convergent ecological adaptation of the extant Araliaceae since they diverged from a common ancestor.

To explore the possible genomic underpinnings of this differential adaptation, we studied gene content, transcription, and functional ontologies, with a particular focus on the connections between these aspects of diversification with genomic architecture following polyploidization and rediploidization. We found that transcriptional regulation of the light signaling pathways, including photosynthesis and photomorphogenesis, is important in responsiveness to variable light intensities in natural habitats, consistent with findings in model plants^33–35. In addition, our pan-genome analyses indicated that biased genetic fractionation and expression-level remodeling of the photosynthesis and photomorphogenesis are correlated with ecological adaptation of Araliaceae species to different ecological habitats of the forest ecosystem. For example, in the regulatory network of photomorphogenesis, the retained copies of the three photoreceptors (Phots, Crys and Phys) and two regulators (HY5 and PIFs) exhibited differential regulation mechanisms in response to altered light conditions of the two contrasting forest habitats. This supports the proposed hypothesis that post-polyploidy reshuffling/diploidization has molded genome architecture and transcriptional regulation to adapt to different ecological niches. Although the links we show between genome structure and ecophysiological origins are correlative rather than causative, then lend support to the very reasonable notion that diploidization following polyploidy is not a mechanistically random or deterministic process, but that instead it is guided by fundamental ecological evolutionary principles.

Our study provides a foundation for further work aimed at functional validation of the ecophysiological-genomic connections suggested by this research. Because all the plants we studied are non-model species, which is indeed the case for the entire family, genetic transformation and genome editing systems are yet to be established, increasing the difficulty of directly validating the biological functions of the identified candidate genes. In addition, we acknowledge that genomic and gene regulatory evidence supporting phenotypic ecological change are largely correlative, and the molecular mechanisms that underlie these evolutionary transformations remain to be elucidated. Furthermore, Araliaceae is a highly species-rich family with extensive morphological and ecological diversity. This study mainly focuses on 12 representative species, and further sampling is likely to generate additional insight into the nature of the connections between post-polyploidy genomic change and ecological/phenotypic diversification. Further studies focusing on these unresolved questions, perhaps empowered by replicated ecological comparisons as demonstrated here, will provide insights with respect to the evolutionary contributions of ancestral genome reshuffling following WGD to ecological adaptation and species diversification in extant plant species.

Methods

Representative species and sample collection

Araliaceae is a highly species-rich family with extensive morphological and ecological diversity^21,36,37. To ensure the representativeness of the study samples, a total of 12 species were selected according their phylogenetic relationship, life form, ecological habitat, and ploidy level. In brief, the 12 species were selected to represent the three major phylogenetic lineages (Asian Palmate, Aralia, and Panax) that covered the three life forms (herb, shrub, and tree) and different ecological habitats of the forest vertical gradients (canopy layer, medium layer, and floor layer). In addition, these selected species also experienced the shared paleo-polyploidization (Pg-β) and followed by three independent neo-polyploidizations (Ac-α, Pg-α, and Ar-α). D. carota was selected as an outgroup in the phenotypic and transcriptomic analyses based on the following attributes: (i) it belongs to Apiaceae, the phylogenetically closest family to Araliaceae within Apiales; (ii) it is also a model species that possesses high-quality reference genome; and (iii) it is a shade-intolerant herb that inhabits contrasting habitat relative to P. ginseng (shade-tolerant herb).

In this study, high-quality reference genomes were assembled for five Araliaceae species, including A. continentalis, A. dasyphylla, K. septemlobus, S. arboricola, and T. papyrifer. Specimens used for genome assembly were sourced as follows: Seeds of A. continentalis were obtained from the National Germplasm Repository; Seeds of A. dasyphylla and T. papyrifer were collected from the Shanghai Chenshan Botanical Garden; Seedlings and seeds of K. septemlobus were provided by the Hangzhou Botanical Garden; and Seedlings and seeds S. arboricola were acquired from the Hainan Tropical Rainforest National Park Administration (Supplementary Data 15).

Phenotypic and transcriptomic analyses were performed for three herbaceous species, five woody Araliaceae and one outgroup (D. carota) species. For the three herbaceous species (A. continentalis, P. ginseng, and P. quinquefolius), seeds were germinated to obtain seedlings for phenotyping and RNA extraction. For the two woody species, A. elata and T. papyrifer, both one-year-old seedlings (germinated from seeds) and adult plants were used. For the other three woody species, E. senticosus, K. septemlobus, and S. arboricola, only adult plant materials were used (because the seed germination rate is very low). Information of sample collection are listed in Supplementary Data 15. All seedlings and adult plants were grown in the greenhouse at Fudan University (Shanghai, China).

Seed stratification, germination, and plant growth conditions

All seeds used in this study were stored at 4 °C for cold stratification to promote germination. Seeds of D. carota were germinated under controlled conditions of 25/18 °C (day/night) with a 16 h light/8 h dark photoperiod. After germination, seedlings of these species were grown in greenhouse at Fudan University for further experimental use. Greenhouse growth conditions were maintained at 22/18 °C (day/night) with a 16 h light/8 h dark photoperiod under normal light intensity (55–60 μmol·m⁻²·s⁻¹; R/FR ratio = 5.6–6.0). Seeds of Araliaceae species underwent additional cold stratification treatments to improve germination efficiency. Detailed germination protocols for each species are as follows:

• i.A. continentalis: Fruits were soaked in water to facilitate manual removal of the pericarp and mesocarp. Cleaned seeds were rinsed thoroughly to remove floating and non-viable seeds, then mixed with moist fine sand at a 1:3 ratio (seed:sand, v/v). The mixture was placed in Petri dishes, covered with plastic wrap to maintain moisture, and subjected to alternating temperature conditions of 25 °C/10 °C with a 12 h light/12 h dark photoperiod for 1–2 months to induce germination.
• ii.A. elata: Fruits were processed similarly to A. continentalis. Seeds were mixed with moist sand at a 1:3 ratio, placed in Petri dishes with plastic wrap, and incubated under alternating temperatures of 20/10 °C and a 12/12 h light/dark cycle for 1–2 months to promote germination.
• iii.P. ginseng and P. quinquefolius: Fruits were processed as described above. The seeds were soaked in water for 48 h to ensure full imbibition, then mixed with moist sand at a 1:3 ratio and incubated at a constant 18 °C under a 12/12 h photoperiod for approximately three weeks to stimulate germination.
• iv.T. papyrifer: Fruits were treated as described for A. continentalis, and seeds were stratified in moist sand (1:3 ratio) under alternating temperatures of 25 °C/15 °C with a 12/12 h photoperiod for approximately two months.

Karyotype and DNA/RNA extraction

Pan-genome and phenotypic analyses were performed for 12 Araliaceae species in this study. Chromosomes numbers of the five newly assembled genomes (A. continentalis, A. dasyphylla, K. septemlobus, S. arboricola, and T. papyrifer) were examined using DAPI staining (Supplementary Fig. 1). Karyotypes of the other seven Araliaceae species were obtained from previous studies, including A. elata, P. ginseng, P. japonicus, P. notoginseng, P. quinquefolius, P. stipuleanatus, and E. senticosus^22,23,25,26. Three species of the Apiaceae (Apium graveolens³⁸, Coriandrum sativum³⁹ and D. carota⁴⁰), Vitis vinifera⁴¹, and Arabidopsis thaliana⁴² were used as outgroups. Genome sizes of the five newly assembled species were estimated using flow cytometry, with three technical replicates performed for each species (Supplementary Fig. 31). Prior to DNA and RNA extraction, leaf surfaces were cleaned with 70% ethanol followed by deionized water to remove surface contaminants. All fresh samples of these tissues were immediately flash-frozen in liquid nitrogen and stored at −80 °C until further processing. Genomic DNA for genome assembly was extracted from fresh young leaves by a commercial biotechnology service (Tiangen, Tianjin, China). Total RNA for gene annotation was extracted from fresh leaf, stem and main root of the five species by the same service provider using standard protocols (Tiangen, Tianjin, China).

Estimation of genome size

Genome size was estimated using flow cytometry following established protocols. Seeds of Solanum lycopersicum and Zea mays (inbred line A188) were used as internal reference standards. Young leaves were collected from one-month-old seedlings of both the two internal reference and Araliaceae species. Approximately 0.8 mL of ice-cold MGb isolation buffer (45 mM MgCl₂·6H₂O, 20 mM MOPS, 30 mM sodium citrate, 1% (w/v) PVP-40, 0.2% (v/v) Triton X-100, 10 mM Na₂EDTA, and 20 μL mL⁻¹ β-mercaptoethanol, pH 7.5) was used for nuclei isolation. Leaf tissues were rapidly chopped with a sharp razor blade in the isolation buffer and incubated on ice for 10 min. The homogenate was then filtered through a 40 μm nylon mesh to obtain a nuclei suspension. An appropriate volume of propidium iodide (PI; stock concentration 1 mg mL⁻¹) and RNase solution (stock concentration 1 mg mL⁻¹) was added, and samples were incubated on ice in the dark for 0.5–1 h. The final working concentrations of PI and RNase were both 50 μg mL⁻¹. For flow cytometric analysis, nuclei suspensions of the target samples were mixed with those of the internal standards at appropriate ratios. Measurements were performed using a BD FACScalibur flow cytometer equipped with a 488 nm blue laser. PI fluorescence intensity was recorded, and 10,000 nuclei were collected per measurement. The coefficients of variation (CVs) of the G1 peaks were controlled below 5%. Data acquisition and histogram analysis were conducted using ModFit v3.0 (Verity Software House, Topsham, ME, USA) software. Genome size (C value) was calculated based on the ratio of mean fluorescence intensity between the target sample and the internal reference standard, according to the formula: DNA content of sample = DNA content of internal standard × (mean fluorescence intensity of sample / mean fluorescence intensity of internal standard). This approach relies on the linear relationship between PI fluorescence intensity and nuclear DNA content.

Genome characteristics were also estimated using a k-mer–based genome survey analysis. Clean Illumina short reads were first obtained by removing adaptor sequences, low-quality reads, and reads containing excessive ambiguous bases. The filtered reads were then used for k-mer counting using Jellyfish v2.3.0⁴³. To ensure reliable estimation, k-mer 17 was tested for all samples. The resulting k-mer frequency distributions were subsequently analyzed to estimate genome size with genome size = kmer number / average kmer depth.

Whole genome sequencing

Pan-genome analyses were performed based on 12 Araliaceae species, five of which were newly assembled in this study. A preliminary genome survey was conducted for five Araliaceae species using whole-genome shotgun (WGS) sequencing. DNA libraries with an average insert size of approximately 350 bp were constructed and sequenced on the Illumina HiSeq or NovaSeq platforms (Supplementary Data 16). Raw sequencing reads were subjected to quality control using fastp v0.23.3⁴⁴, including adapter trimming and removal of low-quality reads (Q < 30). K-mer frequency distributions (k = 17) were computed from the filtered short reads using Jellyfish v2.3.0⁴³, and the resulting k-mer histograms were used for genome survey analysis (Supplementary Fig. 32).

To generate a high-quality and contiguous genome assembly, we applied third-generation long-read sequencing technology (PacBio HiFi). For HiFi sequencing, high-molecular-weight genomic DNA was extracted from fresh young leaves using the Nanobind Plant Nuclei Kit. The DNA quality and integrity were assessed using Qubit fluorometry and pulsed-field gel electrophoresis. SMRTbell libraries were generated for each species following the manufacturer’s protocol, and sequenced on the PacBio Sequel system, generating high-fidelity (HiFi) reads through circular consensus sequencing (CCS), which produces highly accurate long reads suitable for genome assembly and analysis (Supplementary Data 16).

For chromosome-level scaffolding, Hi-C libraries were constructed following a modified version of the protocol⁴⁵. Fresh young leaves were harvested and crosslinked with formaldehyde to preserve chromatin interactions. The crosslinked chromatin was then digested with the restriction enzyme DpnII, and the resulting DNA ends were labeled with biotinylated nucleotides. Proximity ligation was performed to join spatially adjacent DNA fragments, forming chimeric molecules that represent physical chromatin interactions. After ligation, the DNA was purified and sheared to an appropriate fragment size. Biotin-labeled DNA fragments were captured using streptavidin-coated magnetic beads, and sequencing libraries were prepared using Illumina-compatible adapters. The final libraries were sequenced on the Illumina HiSeq X Ten platform with 150 bp paired-end reads (Supplementary Data 16).

Genome assembly and quality control

HiFi reads were assembled using Hifiasm v0.16.1⁴⁶, integrating Hi-C data to generate high-quality assemblies. Prior to Hi-C scaffolding, the initial contig assemblies were subjected to redundancy removal to minimize the influence of heterozygous sequences. Specifically, duplicated and heterozygous contigs were identified and removed using purge_dups v1.0.1 (-l 2) (https://github.com/dfguan/purge_dups), retaining a single representative haplotype for downstream analysis. To construct chromosome-level scaffolds, Hi-C reads were aligned to contigs longer than 1000 bp using BWA-MEM⁴⁷. Scaffolding was performed using the 3D-DNA pipeline⁴⁸, followed by manual review and refinement using Juicebox v1.11.08⁴⁹ to correct misassembles and improve scaffold accuracy (Supplementary Fig. 2). During this process, Hi-C contact maps were manually inspected to further identify and remove residual heterozygous contigs. Although Hifiasm does not explicitly report sequencing-error-derived artifacts, the high base accuracy of PacBio HiFi (CCS) reads minimizes such risks. The main redundancy observed in the assemblies stems from heterozygous sequences, which were effectively removed through purge_dups filtering and Hi-C–based manual curation, resulting in a final, high-quality assembly.

To evaluate the quality of the five assembled genomes, we employed multiple assessment strategies. In brief, Illumina paired-end reads were aligned to the final genome assemblies using Minimap2 v2.26 with the “-ax sr” parameter⁵⁰. Subsequent analyses using SAMtools v1.16⁵¹ provided metrics on mapping rate. Genome completeness and continuity were evaluated using CEGMA v2.5⁵², BUSCO v5.4.3⁵³ and LTR_retriever v2.8⁵⁴. CEGMA v2.5 assessed the presence of 458 conserved core eukaryotic genes to estimate genome completeness. BUSCO v5.4.3 evaluated the assemblies against the Embryophyta lineage dataset from OrthoDB v10 in genome mode, which comprises 1,614 conserved single-copy orthologs. We also used LTR_retriever v2.8 to estimate the LTR assembly index (LAI), an important metric for assessing the completeness of long terminal repeat (LTR) retrotransposons in the genome assembly. As eight of the 12 Araliaceae are polyploid, we further assessed genome completeness and continuity for each subgenome using BUSCO v5.4.3⁵³ and LTR_retriever v2.8⁵⁴ (Supplementary Data 17). Notably, only protein-coding genes predicted from the nuclear genome were subjected for pan-genome analyses, while chloroplast genes were only applied in the transcriptomic analyses of the photosynthesis network.

Annotation of protein-coding gene and repeat sequence

Protein-coding genes were predicted using the combination of ab initio, homology-based, and transcriptome-based approaches (Supplementary Fig. 3 and Supplementary Data 18). For ab initio prediction, AUGUSTUS (v2.4)⁵⁵ and SNAP (2006-07-28)⁵⁶ were employed. Homology-based prediction was performed using GeMoMa (v1.7)⁵⁷ with closely related species as references. For transcriptome-based prediction, RNA-seq data were aligned to the genome using Hisat v2.0.4 (--dta -p 10)⁵⁸, and transcripts were assembled with StringTie (v1.3.3b) (-p 2)⁵⁹. Gene structures were subsequently predicted using GeneMarkS-T (v5.1)⁶⁰. In parallel, de novo transcriptome assembly was performed using Trinity (v2.11.0)⁶¹ with default parameters. The assembled transcripts were subsequently aligned to the genome and assembled into high-quality transcript structures using PASA (v2.0.2)⁶², providing additional transcript evidence for gene annotation. Gene models generated by individual tools within each prediction category were combined into a single non-redundant gene set, resulting in three evidence-based datasets representing ab initio, transcriptome-based, and homology-based predictions, respectively. The three evidence-based gene sets were integrated using EVM (v1.1.1)⁶³. Given the higher reliability of transcriptome- and homology-based evidence, weights were assigned as follows: ab initio (0.3), RNA-seq (50), and homology-based prediction (50). The integrated gene set was further filtered to remove gene models with incorrect coding sequences (e.g., invalid start or stop codons, or frameshifts) and low-complexity sequences (e.g., those in which the top four amino acids accounted for more than 80% of the composition). Finally, the consensus gene models were refined using PASA (v2.0.2)⁶², which updated EVM predictions by adding untranslated region (UTR) annotations and alternatively spliced isoforms, yielding a high-confidence gene annotation set. In addition, we also examined the proportions of expressed genes in the five newly annotated species (Supplementary Data 19).

Repeat sequences, encompassing both tandem and interspersed repeats, were annotated to characterize the repetitive landscape of the assembled genomes (Supplementary Data 2). Given the species-specific nature of many repeats, a de novo approach was employed to construct custom repeat libraries for each Araliaceae species. De novo identification of transposable elements (TEs) was performed using RepeatModeler v2.0.1⁶⁴. The de novo identified repeats were merged with known repeat databases and redundant sequences were removed to generate a non-redundant, species-specific repeat library. This library was then used with RepeatMasker v4.1.0⁶⁵ to annotate and mask interspersed repeats in the genome assemblies. Long terminal repeat retrotransposons (LTRs) were further identified using LTR_retriever v2.8, which incorporates LTRharvest v1.5.9 and LTR_FINDER v1.1 for comprehensive LTR detection^54,66,67. Tandem repeats were identified using MIcroSAtellite identification tool (MISA v2.1) and Tandem Repeat Finder (TRF v4.09), facilitating the detection of microsatellites and other tandemly repeated sequences^68,69.

Gene functions were assigned according to the best match by aligning the protein sequences to the Swiss-Prot using BLASTp (with a threshold of E-value ≤ 1e−5)⁷⁰. The motifs and domains were annotated using InterProScan v5.62-94.0⁷¹ by searching against publicly available databases, such as ProDom, PRINTS, Pfam, SMRT, PANTHER and PROSITE. Functional annotations were also performed via online software eggNOG-mapper v2⁷² (http://eggnog-mapper.embl.de/). We integrated these results to obtain the final gene functional annotations with the corresponding Gene Ontology (GO)⁷³ and Kyoto Encyclopedia of Genes and Genomes (KEGG) items⁷⁴. GO enrichment analyses were conducted using the R package clusterProfiler⁷⁵. Functional categories with p-values < 0.05 and q-values < 0.05 were considered significantly enriched. Noncoding RNA, including tRNA, rRNA, miRNA, snRNA, and snoRNA, were identified using the Infernal v1.1.4⁷⁶ software with the Rfam database⁷⁷.

Inference of ancestral karyotype in extant Araliaceae species

To identify genomic regions derived from the same ancestral chromosome, we performed syntenic analyses for both the diploid and tetraploid species according to our previous study²³. In brief, protein sequences of these Araliaceae species were aligned to the diploid species P. stipuleanatus using DIAMOND⁷⁸, with an E-value cutoff of 1e−10 and a minimum identity threshold of 40%. Then, the program MCScanX⁷⁹ was applied to identify syntenic (collinear) blocks based on conserved gene order. A minimum of five homologous gene pairs was required to define a syntenic block (MATCH_SIZE = 5). Homologous gene pairs were identified using BLASTp with an E-value cutoff of 1e−10. The scoring scheme included a match score of 50 and a gap penalty of −1, allowing up to 25 intervening genes between collinear anchors (MAX_GAPS = 25). An overlap window of five genes was applied to merge adjacent collinear segments. To visualize the distribution of collinear blocks within each genome, we used Circos v0.65⁸⁰ and the online tool ChiPlot (https://www.chiplot.online/) (Supplementary Fig. 33). Synonymous substitution rates (Ks) were calculated for both intra-genomic collinear gene pairs (i.e., paralogous genes) and inter-genomic collinear gene pairs (i.e., homologous genes), using KaKs_Calculator 2.0 with the YN model⁸¹ (Supplementary Figs. 34–36). In addition, chromosomal breakpoint genes were identified using MCScanX⁷⁹, and further integrated with the synteny information from WGDI v0.2.1⁸² to determine the relationships between chromosomal segments (before and after the breakpoints) and their corresponding ancestral chromosomes in P. stipuleanatus (Supplementary Fig. 8). This approach enabled us to resolve the ancestral origin of each chromosome in all studied Araliaceae species (Supplementary Data 20). Connections between the ancestral and modern karyotypes are shown in Supplementary Fig. 10 and Supplementary Data 8.

However, we noted that numbers of protein-coding genes do not scale with the ploidy level of the 12 Araliaceae species. We then examined pipelines of the genome assembly and gene annotation of these species. In brief, the five newly assembled (A. continentalis, A. dasyphylla, K. septemlobus, S. arboricola, and T. papyrifer) and four previously released Panax species (P. stipuleanatus, P. ginseng, P. japonicus, and P. quinquefolius) are assembled and annotated using the same pipeline. Differences in genome architecture and gene content are supposed to have evolved during the post-polyploidy reshuffling process²³. In addition, genome assemblies and gene annotations of the other three species (A. elata, P. notoginseng, and E. senticosus) are obtained from public database that released by previous studies^22,25,26. However, we re-annotated the repeat sequences and GO terms for the three species using our pipeline. Furthermore, our comparative genomic analyses reveal that A. elata shows high genome collinearity with our Aralia species and contains reasonable gene numbers (Supplementary Fig. 8 and Supplementary Data 1). Likewise, while the tetraploid species E. senticosus contains lesser gene number (36,372), it shows high genome collinearity with the other three tetraploid species of the Asian Palmate lineage (generated in our study). To this end, the two species were included in the analyses of biased fractionation of ancestral genome. In contrast, P. notoginseng was not included in the analyses of biased genetic fractionation due to its very low level of genome collinearity with our Panax species. Given that both E. senticosus and P. notoginseng are medicinally important plants and possess similar evolutionary dynamics to the other Araliaceae species, analyses of the genome features of the two species are included in this study.

Identification of the subgenomes in tetraploid species

To obtain a more comprehensive understanding of evolutionary patterns in the Araliaceae family, we performed pan-genome analysis for the five newly assembled species and seven previously released genome assemblies, including P. ginseng, P. japonicus, P. notoginseng, P. quinquefolius, P. stipuleanatus, A. elata and E. senticosus^22,23,25,26. Furthermore, A. graveolens³⁸, C. sativum³⁹, D. carota⁴⁰ and V. vinifera⁴¹ were included as control and outgroup species for comparative purposes. P. stipuleanatus is a diploid species that exhibits clear correspondence with ancestral (post-Pg-β) chromosomes²³, thereby serving as a reliable reference for subgenome partitioning and ancestral karyotype inference (Supplementary Data 21). Subgenomes of all Panax species and A. continentalis were successfully characterized by the program SubPhaser⁸³. However, this program failed to identify subgenomes in the four Asian Palmate lineage, E. senticosus, K. septemlobus, S. arboricola, and T. papyrifer. We further applied WGDI v0.2.1⁸² to identify homologous chromosome pairs in the four Asian Palmate species through whole-genome comparisons with P. stipuleanatus. Next, we performed pairwise collinearity analyses between the four Asian Palmate species and the two subgenomes of A. continentalis. The number of collinear gene pairs shared with each A. continentalis subgenome provided an initial classification framework for subgenome assignment in the four Asian Palmate species. To refine subgenome partitioning, we extracted the 12 homologous chromosome sets from each of the four Asian Palmate species and performed OrthoFinder v2.5.5⁸⁴ analysis to identify single-copy orthologs. Maximum likelihood (ML) trees were then constructed for each homologous group using IQ-TREE v2.3.3 (parameters: -st AA -m JTT -bb 1000)⁸⁵. Subgenome identity was further verified and adjusted based on phylogenetic clustering patterns across the eight homologous chromosomes in each group. Through this integrative approach, we successfully delineated subgenomes for all tetraploid Araliaceae species (Supplementary Data 8 and 21).

Phylogeny, divergence time, gene family evolution and identification of subgenome

To reconstruct phylogenetic relationships of the 12 Araliaceae and four outgroup species, we used the BUSCO dataset (eudicots_odb10) to identify universal single-copy orthologs from the whole-genome protein sequences of each species. BUSCO searches were performed using the “--metaeuk” option to enable accurate gene prediction, and both nucleotide and corresponding protein sequences were retrieved. Genes with coverage < 80% among these species were excluded from further analysis. As a result, a total of 2321 high-confidence single-copy genes were retained. Protein sequences were aligned using MAFFT v7.526⁸⁶ with the L-INS-i algorithm, and codon-based nucleotide alignments were generated using PAL2NAL v14.1⁸⁷. Poorly aligned regions were removed using trimAl v1.5 (-clustal -nogaps)⁸⁸. For each gene, a maximum likelihood (ML) phylogenetic tree was constructed using IQ-TREE v2.3.3 (parameters: -st DNA-m GTR -bb 1000)⁸⁵. Subsequently, a species tree was inferred from all individual gene trees under the multispecies coalescent model using ASTRAL-III v5.6.3⁸⁹. In parallel, a super-matrix was generated by concatenating the cleaned codon alignments of all 2321 retained single-copy genes using AMAS v1.0⁹⁰. Phylogenetic trees based on the super-matrix were inferred using two independent ML approaches: RAxML-NG v1.0.3⁹¹ with a partitioned GTR + GAMMA model and 200 bootstrap replicates, and IQ-TREE v2.3.3⁸⁵ with a partitioned GTR model and 1000 bootstrap replicates (Supplementary Fig. 37).

Orthologous gene families were identified using OrthoFinder v2.5.5⁸⁴. Protein sequences from all selected species were combined and subjected to all-vs-all sequence similarity searches using DIAMOND (E-value cutoff: 1e−30)⁷⁸. A total of 35,799 gene families were identified, including 89 strict single-copy ortholog genes, each containing exactly one gene from each species. Protein sequences of the shared single-copy orthologs were aligned using the linsi algorithm in MAFFT v7.526⁸⁶, followed by trimming with trimAl v1.5 (-clustal -nogaps)⁸⁸ to remove poorly aligned regions. Corresponding codon alignments were generated using PAL2NAL v14.1⁸⁷ and concatenated using AMAS v1.0⁹⁰. A maximum likelihood (ML) tree was inferred from this concatenated super-matrix using IQ-TREE v2.3.3⁸⁵ with the partitioned GTR model and 1000 bootstrap replicates (Supplementary Fig. 37).

Divergence time estimation was performed based on the 89 single-copy orthologs using MCMCTREE implemented in the PAML v4.10.7 package⁹². The analysis was conducted with the following parameters: “usedata = 1”, “model = 3”, “alpha = 0” (F84 model), “clock = 2” (independent rates), “BDparas = 1 1 0.1”, “burnin = 2000”, “sampfreq = 50” and “nsample = 20,000”. Two fossil-based calibration constraints were obtained from the TimeTree database (http://www.timetree.org/): the divergence time range of 48–66.8 million years ago (Mya) between Araliaceae and Apiaceae, and the divergence time range of 112.4–125 Mya between Apiaceae and Vitis (Vitaceae). Gene family expansion and contraction events were inferred using CAFE v5⁹³, based on the estimated species divergence tree. To identify subgenomes of these tetraploid Araliaceae species, we employed three strategies: firstly, subgenomes of the tetraploid species were identified using the program SubPhaser⁸³; secondly, orthologous genes were used to reconstruct phylogenetic trees to infer the subgenome relationships; thirdly, homologous chromosome pairs among the subgenomes of the tetraploid species were identified using WGDI v0.2.1⁸². Gene duplication modes were identified using DupGen_finder⁹⁴, using Vitis vinifera⁴¹ as the outgroup species to infer ancestral gene states. Genes were classified into five duplication types: dispersed duplication (DSD), proximal duplication (PD), tandem duplication (TD), transposed duplication (TRD), and whole-genome duplication (WGD). Normality of the data matrix was assessed using the Shapiro-Wilk test (p-value < 0.05). Statistical significance of differences in gene counts on chromosomes and in gene counts associated with enriched GO terms was assessed using a two-sided Wilcoxon rank-sum test (p-value < 0.05).

Experimental design for simulated ecological niches

To investigate the light responses of Araliaceae and outgroup species under simulated ecological habitats, we collected both seedlings and adult plants from eight Araliaceae and one outgroup species to cover the lineages, life forms and habitats, including four herbaceous (A. continentalis, P. ginseng, P. quinquefolius, and D. carota) and five woody (A. elata, E. senticosus, K. septemlobus, S. arboricola, and T. papyrifer) species. Ecological habitats in natural field involve complex interactions among multiple environmental factors, such as light, temperature, and water availability. To decrease the interaction effect of different environmental factors, this study focused on how light quality and quantity affect phenotypic performance and transcriptional regulation. To this end, five light conditions were simulated based on our previous studies^27,28 and the other model species^95,96, including two simulated open-canopy (high light and normal light) and three shaded-understory (low light, shade light, and fluctuating light) conditions. Detailed light conditions for each treatment are provided in Supplementary Data 22. All seedlings and adult plants were grown under controlled conditions in a greenhouse at Fudan University (Shanghai, China). The growth conditions of normal light were 22 °C/18 °C (day/night) with a 16/8 h of light/dark photoperiod, 55–60 photosynthetic photon flux density (PPFD, μmol·m⁻²·s⁻¹) and 5.6–6.0 of the red to far-red ratio (R/FR). Once the first true leaf geminated, seedlings were subjected to four different light treatments: high light (High), low light (Low), shade light (Shade), and fluctuating light (Fluctuation). The adult plants were firstly grown under normal light (Normal) for three weeks, and then these adult plants were exposed to the same set of light treatments until the new leaves were developed.

We performed the following series of light treatment experiments on both seedlings and adult plants:

• (i)Seedling short-term light treatments (hereafter referred to as Short-term treatment) and sampling: For A. elata, P. ginseng, P. quinquefolius, and D. carota, seedlings at the first true leaf stage were subjected to four light conditions—Normal, Shade, High, and Low—for 1 h, 3 h, and 6 h. The aboveground tissues were harvested at each time point, and all sampled organs were pooled to form composite samples. For each species under each light treatment, three biological replicates were collected from phenotypically uniform individuals. After rinsing with deionized water and blotting dry, tissues were flash-frozen in liquid nitrogen and stored at −80 °C until further use.
• (ii)Seedling long-term light treatments (hereafter referred to as Long-term treatment) and sampling: Seedlings of A. continentalis, A. elata, P. ginseng, P. quinquefolius, T. papyrifer, and D. carota were sampled under long-term exposure to the four light conditions mentioned above. At the third true leaf stage, the aboveground parts were collected, including the first true leaf, its petiole, and the hypocotyl. Due to species-specific morphological characteristics, P. ginseng and P. quinquefolius lacked clear first true leaves and hypocotyls; hence, their compound leaves (comprising three leaflets) and main petioles were sampled instead. For each species and treatment, three biological replicates were collected from phenotypically similar seedlings. Tissues were rinsed, dried, flash-frozen in liquid nitrogen, and stored at −80 °C.
• (iii)Long-term treatments (hereafter referred to as Leaf development) and sampling of adult plants: For mature individuals of E. senticosus, K. septemlobus, S. arboricola, and T. papyrifer, newly emerged leaf buds were treated under the four light conditions for over three weeks. Leaves were sampled at three developmental stages—curled, emerging, and fully expanded. For each species and treatment, three biological replicates were obtained from morphologically consistent plants. After rinsing and drying, samples were frozen in liquid nitrogen and stored at −80 °C.
• (iv)Fluctuating light treatment (hereafter referred to as Fluctuation) and sampling: All nine species were subjected to fluctuating light treatments for 0 h, 12 h, and 5 days (sampled at the 12th hour of the fifth day). At each time point, leaf tissues were collected from fully expanded leaves at comparable positions and under uniform light exposure across individuals. Leaves from three uniform plants were pooled as one biological replicate, and three replicates were collected for each condition. Samples were rinsed, dried, snap-frozen in liquid nitrogen, and stored at −80 °C.

Measurement of morphological trait under different light conditions

(1) Leaf morphological diversity: To assess morphological plasticity in response to light, we conducted light treatment experiments under five light conditions across nine species. Due to limitations in sample availability, only four seed-derived species—A. elata, P. ginseng, P. quinquefolius, and D. carota—were subjected to short-term light exposure (normal, shade, high and low light) during the first true leaf stage. Leaf morphological features were documented using a Nikon COOLSCOPE II digital imaging system (Nikon Corporation, Tokyo, Japan) (Supplementary Fig. 16). In contrast, long-term light treatments were conducted on six species with successful seedling establishment (D. carota, A. continentalis, A. elata, P. ginseng, P. quinquefolius, and T. papyrifer). Seedlings were grown under the four light regimes from the first to the third true leaf stage. After sufficient morphological differentiation had developed under each treatment, leaf phenotypes were again imaged using the same Nikon system (Supplementary Fig. 16).

For woody species (E. senticosus, K. septemlobus, S. arboricola, and T. papyrifer), adult plants were grown under each light condition until newly developed leaves had fully expanded. Leaf development was monitored across three key stages: leaf curling stage (Stage I), young leaf stage (Stage II), and fully expanded stage (Stage III). At each stage, leaf morphology was imaged to facilitate comparative analysis of light-dependent morphological traits (Supplementary Fig. 16).

(2) Specific leaf area, petiole length, and hypocotyl length measurements: We measured key morphological traits under different light conditions, including specific leaf area (SLA), petiole length, and hypocotyl length (Supplementary Fig. 16). The value SLA is defined as the ratio of leaf area to leaf dry mass, is a key trait reflecting plant strategies in light capture, carbon fixation, growth rate, and environmental adaptation^97,98. Measurement procedures were as follows:

• (i)For each species and treatment condition, we selected the first true leaf from seedlings at the third-leaf stage. Leaves were imaged using the Nikon COOLSCOPE II system, and leaf area was quantified using ImageJ software (http://imagej.net/ImageJ).
• (ii)Harvested leaves were placed in labeled kraft paper envelopes and oven-dried at 40 °C for 3–5 days. Once fully dried, leaf dry weights were measured using an analytical balance (Sartorius TE212-L, Göttingen, Germany).
• (iii)SLA was calculated for each sample using the formula: SLA = Leaf area / Leaf dry weight. Five biological replicates were used per species per light treatment, and the average SLA was used for subsequent comparisons.
• (iv)In addition to SLA, we measured petiole lengths (for the first, second, and third true leaves) and hypocotyl lengths using a ruler for seedlings at the third-leaf stage. Shade-avoidance responses—such as elongation of petioles and hypocotyls—are well-documented in sun-loving species, whereas shade-tolerant species often exhibit contrasting morphological patterns^27,99,100. Statistical significance of these phenotypic differences was assessed using a two-sided Wilcoxon rank-sum test (p-value < 0.05). Thus, measuring these traits provides insight into the light-response strategies among Araliaceae species.

Measurement of physiological trait under different light conditions

(1) Chlorophyll content and chlorophyll fluorescence parameters: To evaluate whether photosynthetic efficiency was affected under different light treatments, chlorophyll content and fluorescence parameters were measured (Supplementary Fig. 17). Total chlorophyll content was assessed using a portable chlorophyll meter (SPAD-502, Konica Minolta, Japan) on the first true leaf of each species under each light condition. To minimize measurement error, leaves were randomly sampled from the upper canopy, with 25 biological replicates analyzed per treatment. Statistical significance was assessed using a two-sided Wilcoxon rank-sum test (p-value < 0.05). Additionally, chlorophyll fluorescence parameter Fv/Fm, which reflects the maximum quantum efficiency of PSII, was measured using an IMAGING-PAM chlorophyll fluorometer system (Walz, Germany) (Supplementary Figs. 20–21). Data acquisition and image processing were conducted using the ImagingWin software v2.45i provided with the instrument, and statistical significance was evaluated as above.

(2) Light response curve measurements: To further assess the photosynthetic performance of plants under varying light regimes, light response curves were measured as follows:

• (i)Measurement conditions: All measurements were conducted in the greenhouse under the corresponding constant light treatments and a temperature of 22 °C. Mature leaves from plants subjected to long-term treatments were selected for analysis. Light response curves were measured using a LI-COR LI-6800 portable photosynthesis system.
• (ii)Instrument settings: After calibration and confirmation of normal instrument performance, the following environmental parameters were configured: airflow rate (500 μmol·s⁻¹), pressure valve setting (0.1 kPa), relative humidity (RH, 60%), CO₂ concentration (400 μmol·mol⁻¹), leaf chamber fan speed (10,000 rpm), and chamber temperature (25 °C). Light intensity was applied in a stepwise gradient of 0, 20, 60, 80, 100, 200, 300, 500, 800, 1000, 1500, and 2000 μmol·m⁻²·s⁻¹. At each light level, a waiting period of 10–30 minutes was applied to allow sufficient light acclimation.
• (iii)Data acquisition and analysis: Measurements were recorded and logged using the LI-6800 system. For each species under each light condition, three biological replicates were included. Data were then analyzed in R by fitting the response curve to a rectangular hyperbola correction model, from which key photosynthetic parameters were extracted, including: maximum photosynthetic rate (Amax), light saturation point (LSP), light compensation point (LCP) and dark respiration rate (Rₑ) (Supplementary Figs. 18–19).

Gene expression profiling and expression plasticity

All samples of the nine species used for phenotypic assessments were also used for gene expression analyses (Supplementary Data 12). Total RNA was extracted following the manufacturer’s protocol. After quality control of the RNA samples, cDNA libraries were constructed and sequenced on the Illumina NovaSeq platform with a paired-end 150-bp (PE150) reads configuration. Clean reads generated from cDNA libraries were aligned to the reference genome using Hisat v2.0.4⁵⁸. The genome index was built with hisat2-build, and reads were mapped using the parameters --dta -p 20. The resulting SAM files were converted into BAM format using SAMtools v1.16 (samtools view -bS)⁵¹, followed by sorting with SAMtools sort to generate sorted BAM files. Transcript assembly and quantification were performed using StringTie v1.3.3b⁵⁹ to calculate Fragments Per Kilobase of transcript per Million mapped reads (FPKM) and Transcripts Per Million (TPM) values for each gene. Differentially expressed genes (DEGs) were identified using DESeq2 v1.36.0¹⁰¹ based on raw count matrices from three biological replicates. Genes were considered significantly differentially expressed with an adjusted p-value (FDR) < 0.05 and |log₂(FoldChange)| ≥ 1, corresponding to a ≥ 2-fold difference in expression between groups or tissues.

To further characterize transcriptional responses across light conditions, gene expression plasticity was quantified using two complementary indices: expression diversity (Hj), which captures the overall breadth and variability of gene expression responses across conditions, and expression specificity (Sj), which measures the degree to which gene expression is preferentially associated with particular conditions. All calculations of Hj and Sj were performed using FPKM- and TPM-normalized gene expression values. Higher Hj values indicate greater transcriptional variability, whereas higher Sj values reflect more condition-specific and canalized expression patterns¹⁰².

Associations between the gene expression plasticity and biological/ecological factors

To examine if gene-expression plasticity (Hj and Sj) is associated with biological and ecological factors, we employed distinct statistical approaches to explain the nature of ploidy, habitat, and life form. For each species, the Hj and Sj values were calculated using FPKM and TPM expression matrices. All statistical analyses were conducted in R using the packages rstatix (v0.7.2)¹⁰³, ggpubr (v0.6.1)¹⁰⁴, and corrplot (v0.95)¹⁰⁵.

For ploidy level, Wilcoxon rank-sum test was used to compare the Hj and Sj values between the diploid and tetraploid, with the effect size reported as r (calculated as Z/N^1/2). Spearman’s rank correlation was further performed as a complementary approach to assess monotonic trend, with the diploid and tetraploid coded as 0 and 1, respectively. This ordinal coding is justified by the evolutionary trajectories of polyploidization event because the tetraploid species are derived from diploid ancestors.

For the ecological habitats, as the floor layer, medium layer, and canopy layer/open habitat follow a natural ordinal structure based on light availability and vertical stratification, Kruskal-Wallis test with Dunn’s post-hoc comparison (Benjamini-Hochberg correction) was performed to detect pairwise differences, with effect size reported as epsilon-squared (ε²). Spearman’s rank correlation was further conducted as a complementary approach to assess monotonic trend, with the habitat categories coded as 0 (floor layer), 1 (medium layer), and 2 (canopy layer/open habitat).

For the life form, as the three plant types (herb, shrub, and tree) lack a justified monotonic order, we employed Kruskal-Wallis test with the effect size reported as ε². Then, Dunn’s post-hoc test with Benjamini-Hochberg correction was applied for pairwise comparisons. Spearman’s rank correlation was not applied to life form due to the absence of a biologically meaningful ordering assumption.

Identification of photomorphogenic- and photosynthetic-related genes and network inference

Orthologs involved in the photomorphogenesis and photosynthesis networks were identified according to our previous studies^27,28. To investigate the transcriptional regulatory landscape underlying photomorphogenesis in Araliaceae species, we first identified putative photoreceptor genes—including cryptochromes (Crys), phototropins (Phots), and phytochromes (Phys)—as well as regulatory genes such as HY5 and phytochrome-interacting factors (PIFs). The protein sequences of these genes from Arabidopsis thaliana were retrieved from the TAIR database⁴² and used as queries to perform BLASTp searches against the annotated proteomes of each Araliaceae species. Homologs with sequence identity ≥ 40% and E-value < 1e−5 were retained as candidate genes for downstream analysis.

Given that the enzymatic steps involved in photosynthesis and carbon fixation pathways have been well characterized, we identified the relevant genes using reference information from the Kyoto Encyclopedia of Genes and Genomes (KEGG) (https://www.kegg.jp/). Specifically, we retrieved all enzyme-coding genes involved in the photosynthesis (KEGG pathway map00195, https://www.kegg.jp/pathway/map00195) and carbon fixation in photosynthetic organisms (map00710, https://www.kegg.jp/pathway/map00710) pathways, and obtained their corresponding KEGG Orthology (KO) numbers and associated A. thaliana protein sequences. To identify orthologous genes involved in the photosynthesis pathway in Araliaceae species and D. carota, we performed BLASTp searches using the Arabidopsis protein sequences as queries against the predicted proteomes of the target species. Blast results were filtered using the following criteria: sequence identity > 30% and E-value < 1e−5. For the carbon fixation pathway, we directly searched for enzyme-encoding genes in the genome annotation files of the Araliaceae species and D. carota using the KO numbers of each enzyme as queries. Genes annotated with matching KO identifiers were retrieved, and their gene copy numbers were counted for each species.

To address the underlying gene regulation mechanisms of photosynthesis efficiency, gene regulatory network of photosynthesis network was divided into five modules based on KEGG pathway,, including light capture (LHC-I/II and PS-I/II), non-photochemical quenching (NPQ), electron transport (linear and cyclic electron flow (LEF and CEF)), energy conversion (ATP and NADPH), and carbon fixation (https://www.kegg.jp/pathway/ko00195, https://www.kegg.jp/brite/ko00194 and https://www.kegg.jp/pathway/ko00710) (Supplementary Figs. 38–40 and Supplementary Data 14).

Functional enrichment analysis and tree map visualization

To investigate the functional classification of genes responding to light stress in the shaded-understory and open-canopy habitats, functional enrichment analysis was performed on significantly differentially expressed genes (DEGs) of Araliaceae species. GO enrichment analysis was conducted separately for up-regulated and down-regulated DEGs using the R package clusterProfiler⁷⁵. GO terms with a p-value < 0.05 and a false discovery rate (q-value) < 0.05 were considered significantly enriched. Finally, GO terms from DEGs under both short- and long-term light conditions were pooled and subjected to REVIGO¹⁰⁶ for redundant terms simplification and tree map generation. S. lycopersicum was selected as the reference species at the suggestion of the developer, and all other parameters were set to default values.

Gene expression and co-expression network analysis

Gene transcriptional levels were assessed using both Fragments Per Kilobase of transcript per Million mapped reads (FPKM) and Transcripts Per Million (TPM). Comparison of the two expression matrices revealed broadly consistent expression patterns across all species (Supplementary Figs. 41–42). Weighted gene co-expression network analysis (WGCNA) was performed using the R package WGCNA v1.74¹⁰⁷ to identify gene modules associated with chlorophyll content and Fv/Fm values across all nine species. Log-transformed FPKM expression matrices were used as input after filtering out lowly expressed genes, defined as those with a raw counts > 10 in fewer than three samples. The following parameters were applied: TOMType = “signed”, corType = “pearson”, mergeCutHeight = 0.25, and minModuleSize = 30. Modules significantly associated with the target traits were identified based on correlation coefficients and corresponding p-values. Module membership (MM) and gene significance (GS) were calculated, and genes with |MM | > 0.8 and |GS | > 0.2 were defined as hub genes. Gene Ontology (GO) enrichment analysis was performed separately for hub genes from modules significantly correlated with chlorophyll content or Fv/Fm values. All results are provided in Supplementary Fig. 43.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

L.F.L. conceived this project. L.F.L., X.F.W, B.L., and J.F.W. designed and supervised the project. J.Q.L, X.F.W., Y.Q.N., Y.X.F., J.H.H., and H.X.Y. conducted the experiments and analyzed the data. X.F.W., Y.Q.N., J.Q.L, J.H.H., H.X.Y., B.L., L.F.L., and J.F.W. wrote the manuscript. All authors discussed the results and approved the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was supported by the National Key Research and Development Program of China (2024YFA0918000), the National Natural Science Foundation of China (31970235; 32500199) and the Guangdong Basic and Applied Basic Research Foundation (2026A1515010301).

Data availability

All data supporting the findings of this study are available in this paper. The raw sequencing data, including Illumina short reads, HiFi reads, Hi-C reads, and RNA short reads, have been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center (NGDC), China National Center for Bioinformation (CNCB) under BioProject PRJCA045967 with the accession numbers CRA029809 and CRA029989. The final genome assemblies and annotation files are available in the Genome Warehouse (GWH) at NGDC under the same BioProject (PRJCA045967 [https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA045967]) with accession numbers of SAMC5875150–SAMC5875154. The genome assemblies have been deposited in the National Center for Biotechnology Information (NCBI) under BioProject PRJNA1321913 with accession numbers JBWDIF000000000–JBWDIJ000000000. Source data are provided with this paper.

Code availability

Custom scripts for reproducing the analyses and figures are deposited at https://github.com/JiaqingLei/araliaceae-polyploidy-ecoadaptation, with a detailed step-by-step analytical protocol integrated into the repository’s README file.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77785-6.