The Nigella damascena genome provides insights into transposable element-driven genomic gigantism and trait evolution Authors: Xuehao Fu, Tianyu Lei, Boka Li, Chunxi Peng, Cheng Xue, Ruoheng Jian, Yi Yuan, Xu Yao, Yuannian Jiao, Guixia Xu, Rui Zhang, Jie Cheng, Xiaofeng Yin, Hongyan Shan, Hongzhi Kong Published: 2026 ABSTRACT AbstractGiant genomes, generally dominated by transposable elements (TEs), have evolved repeatedly in angiosperms. The role of TEs in the evolution of giant genomes, however, remains largely unclear. Here, by ancestral genome size reconstruction, whole-genome sequencing, and comparative genomic and transcriptomic analyses, we reveal the processes, drivers, and consequences of genomic gigantism in the buttercup family (Ranunculaceae). We find that the giant diploid genome of Nigella damascena (10.84 Gb, ~35 times that of columbine) has evolved from an ~1.46 Gb ancestral genome by persistent TE accumulation over 60 million years. TE insertions in genic regions have generated genes with ultra-long introns or altered coding sequences, as well as Nigella-specific TE-derived genes, collectively accounting for ~20% of protein-coding genes in the N. damascena genome. TE-mediated regulatory changes and gene duplications/losses likely underpin the evolution of elaborate petals, fused carpels, and specialized secondary metabolites in N. damascena. These findings reveal how TEs drive genomic gigantism and shape specialized traits, advancing our understanding of genome size evolution.Subject terms: Phylogenetics, Plant evolution, Evolutionary biology, Comparative genomics Introduction Genome size varies tremendously across flowering plants (i.e., angiosperms), spanning over 2400-fold, from the so-far recorded smallest genome of Genlisea tuberosa (Lentibulariaceae; 61 Mb) to the largest one of Paris japonica (Melanthiaceae; 148.8 Gb)1. Giant genomes, especially those exceeding 10 Gb, have been documented in at least 43 families across 24 orders and thus evolved independently in angiosperms2,3 (Supplementary Fig. 1). Genomic gigantism is generally believed to have greatly impacted the survival of corresponding lineages and potentially contributed to the origin of the endless forms and functions of plants4,5. Theoretically, both polyploidization, also known as whole-genome duplication (WGD), and repetitive sequence amplification can drive genome expansion4. However, accumulating genomic evidence reveals that giant genomes are typically dominated by repetitive sequences, particularly transposable elements (TEs), which constitute the majority (generally > 80%) of the genome, as exemplified by Allium cepa (Amaryllidaceae), Helianthus tuberosus (Asteraceae), Paeonia ludlowii (Paeoniaceae), Lilium davidii var. unicolor (Liliaceae), and Paris polyphylla var. yunnanensis (Melanthiaceae)6–10. This recurrent pattern, therefore, highlights the important contribution of repeat-driven inflation in shaping giant genomes. Numerous case studies have demonstrated that insertion of TEs into specific genomic regions can modulate gene expression and function, induce gene loss or duplication, and even facilitate the creation of novel genes, thereby contributing to plant trait evolution11–14. Insertion of a retrotransposon (Hopscotch) in the upstream region of TEOSINTE BRANCHED1 (TB1), for instance, has caused overexpression of TB1 and a dramatic reduction in branch number during the domestication of maize15. Integration of a retrotransposon (Dasheng) into the ninth exon of the Wx allele led to a loss-of-function mutation, generating a glutinous rice cultivar16. A retrotransposon (Rider)-mediated origination of a novel gene, SUN, underpinned the shift from round to elongated tomato fruits17. At the genomic level, it has been reported that in species bearing giant genomes, such as Allium sativum, Pae. ludlowii, and Lilium regale, substantial proportions of protein-coding genes contain TEs7,18,19. In Pae. ludlowii, TEs have influenced the evolution of genes related to oil biosynthesis and floral traits (e.g., color and scent)7. Despite these advances, it remains poorly understood how TEs have shaped giant genomes over evolutionary time, and to what extent TE amplification has influenced the evolution of genes and ultimately specialized traits. The buttercup family (Ranunculaceae) appears to be an excellent system for addressing the above issues. On one hand, the Ranunculaceae represents one of the families exhibiting a wide range of genome sizes in angiosperms2,3 (Supplementary Fig. 1). Within the family, some species, such as those in the genera Nigella, Ranunculus, and Hepatica, possess large or even giant genomes and are characterized by long and bent chromosomes, known as Ranunculus-type (R-type) chromosomes2,20,21 (Fig. 1). This represents a striking contrast to relatives with smaller genomes such as Aquilegia harboring small and kidney-shaped Thalictrum-type (T-type) chromosomes and Coptis possessing small and slender Coptis-type (C-type) chromosomes2,20,21 (Fig. 1). The genomic divergence, together with a well-resolved phylogeny of this family and the availability of reference genomes spanning from ~300 Mb to ~6.0 Gb22–30, positions the Ranunculaceae as a pivotal system for investigating the evolution of genomic gigantism, particularly given the addition of giant genomes. Ancestral genome size reconstruction in the Ranunculaceae. Colored circles of different sizes indicate inferred ancestral genome sizes at internal nodes and genome sizes of extant species at tips. Chromosome diagrams of the representative species on the right are adapted from Gregory21, with hollow circles indicating satellites at the ends of chromosomes. The green stars indicate the WGD events predating the origins of Ranunculales and Delphinieae, whose occurrence times were estimated based on subsequent synteny analyses and comparison of KS distributions among syntenic paralogs and orthologs. Mya, million years ago. Source data are provided as a Source Data file. On the other hand, species within the Ranunculaceae have evolved distinctive floral and phytochemical traits. For instance, floral symmetry can be actinomorphic or zygomorphic; floral phyllotaxis can be whorled or spiral; petals and staminodes can be present or absent; petals can be flat, spurred, or peltate in shape; and carpels can be separate or fused31–33. The ranunculaceous species are also renowned for their diverse and lineage-specific phytochemical secondary metabolites, such as protoberberine alkaloids (PAs), diterpene alkaloids, ranunculin, and β-elemene34. By using Aquilegia, Thalictrum, Delphinium, and Nigella as models, the genetic programs involved in flower development and petal morphogenesis have been extensively studied. The available data suggest that the functions of floral organ identity genes are largely conserved in the family35–42, and that petal elaboration and diversification result from rearranged expression of adaxial/abaxial genes and co-option of genes that are originally not expressed in petals43–48. Despite these mechanistic insights, the genomic basis of the evolution of specialized traits in the Ranunculaceae remains poorly understood. Here, based on a robust, time-calibrated phylogenetic framework, we reconstruct ancestral genome sizes across the Ranunculaceae and trace the evolutionary trajectory of genomic gigantism. By generating a chromosome-level genome assembly of Nigella damascena, a representative species with a giant diploid genome, and conducting extensive comparative genomic and transcriptomic analyses with other ranunculaceous species with varying genome sizes (Aquilegia coerulea, 310 Mb; Coptis chinensis, 0.94 Gb; Ranunculus cassubicifolius, 2.69 Gb; and Aconitum vilmorinianum, 5.76 Gb), we reveal the critical role of TE amplification in driving genome expansion and its pervasive impacts on gene evolution. Furthermore, we explore the genomic basis underlying the evolution of specialized traits in N. damascena, including bilabiate petal elaboration, postgenital carpel fusion, β-elemene production, and PA loss. Our results not only provide insights into genome size evolution, but also underscore the critical roles of TEs in driving genomic gigantism and shaping specialized traits. Results Evolution of genome size within the family Ranunculaceae To understand the evolutionary processes of genome size in the Ranunculaceae, we performed ancestral genome size reconstruction based on a time-calibrated phylogenetic framework assembled from previous studies23,27,49–61 (Fig. 1, Supplementary Fig. 2). A total of 68 species (50 from the Ranunculaceae and 18 from other Ranunculales families as outgroups) were included, with their genome size data estimated using flow cytometry or obtained from published studies (Supplementary Fig. 3, Supplementary Data 1, with references listed therein). Ancestral states were reconstructed under a Brownian motion (BM) model using the anc.ml function in phytools (Supplementary Data 2), with the phylogeny and genome size data as input. The results showed that the most recent common ancestor (MRCA) of the Ranunculaceae possessed a 1.46 Gb genome and the MRCA of the core Ranunculaceae (excluding the three basal tribes) had a 2.61 Gb genome, both comparable to those of extant species with C-type chromosomes (Fig. 1). Following the diversification of the core Ranunculaceae, the genome size evolved in two contrasting directions: the genome contracted sharply to 0.61 Gb in the MRCA of Isopyreae, a tribe characterized by T-type chromosomes, whereas it expanded to 3.56 Gb in the MRCA of the remaining tribes, which harbor R-type chromosomes (Fig. 1). Subsequently, the 3.56 Gb ancestral genome showed further expansion, leading to independent evolution of large or even giant genomes in different genera, such as Hepatica, Ranunculus, Eranthis, Anemonopsis, Nigella, and Adonis (Fig. 1). For instance, along the evolution toward Nigella, the genome size increased from the ancestral 3.56 Gb to 6.83 Gb in the MRCA of the genus and continued to expand in most extant species, even exceeding 10 Gb in N. damascena and N. orientalis (Fig. 1). These results indicate that the genome size has undergone dramatic expansions or contractions during the evolution of the Ranunculaceae and that the giant genomes in extant genera originated through independent and gradual expansion processes. The giant genome of N. damascena and its key features To explore the drivers and consequences of genome expansion in the Ranunculaceae, we sequenced the genome of N. damascena, a representative species with a diploid giant genome, and analyzed its genomic features. With 274.82 Gb of HiFi long reads, we first generated an initial 10.84 Gb assembly (3192 contigs; N50 = 11.58 Mb), a size consistent with flow cytometry (10.32 Gb) and k-mer (10.82 Gb) estimates (Supplementary Figs. 3, 4a, Supplementary Data 3–6). Using 247.04 Gb of valid high-throughput chromosome conformation capture (Hi-C) contact data, we then anchored the contigs onto six pseudochromosomes, covering 95.86% (10.39 Gb) of the assembled sequences (Fig. 2a, Supplementary Fig. 4b, Supplementary Data 7, 8). The genome achieved high completeness and continuity, as evidenced by an Illumina read mapping rate exceeding 99.25% (Supplementary Data 9), a Benchmarking Universal Single-Copy Orthologs (BUSCO) completeness of 98.50% (Supplementary Fig. 4c), and an long terminal repeat (LTR) assembly index (LAI) of approximately 17 (Supplementary Fig. 4d). Genomic features of N. damascena in comparison with other Ranunculales species. a Genomic features of N. damascena. Track a represents the assembled six chromosomes, with red and black boxes on each chromosome representing assembled centromere and telomere regions, respectively. Track b represents GC content. Tracks c–g represent densities of protein-coding genes, non-coding RNAs, TEs, Gypsy, and Copia, respectively. The genomic features showing in tracks b–g are calculated in a 30 Mb sliding window, with a step size of 3 Mb. Inner lines indicate syntenic blocks. b Genome-wide synteny among seven Ranunculales species. Colored lines connect syntenic blocks, with each color corresponding to a specific chromosome in the N. damascena genome. A representative syntenic block is highlighted in green. Numbered white ellipses denote chromosomes. c Microsynteny comparisons among Ac. vilmorinianum, N. damascena, and Pap. rhoeas. Blue and green rectangles represent protein-coding genes on the forward and reverse strands, respectively. Gray lines connect syntenic gene pairs, with one set highlighted in purple and brown. d KS distributions of syntenic paralogs and orthologs after evolutionary correction. The red arrow indicates the KS peak of syntenic paralogs within N. damascena, while the gray arrow marks the KS peak of syntenic orthologs between N. damascena and Pap. rhoeas. Nida, N. damascena; Acvi, Ac. vilmorinianum; Raca, R. cassubicifolius; Aqco, Aq. coerulea; Coch, C. chinensis; Eppu, E. pubescens; Parh, Pap. rhoeas. A total of 35,747 protein-coding genes and 9.61 Gb of repetitive elements were annotated in this genome (Table 1, Supplementary Data 10–13). Further comparison with the genomes of Ac. vilmorinianum, R. cassubicifolius, Aq. coerulea, and C. chinensis revealed that several key genomic features in N. damascena were largely comparable to or smaller than those in the other four species, including the number of protein-coding genes, mean coding sequence (CDS) length, mean exon length, and mean exon number per gene (Table 1). In contrast, the mean gene length (9715 bp), mean intron length (3016 bp), and repeat content (9.61 Gb, 88.60% of the genome) were noticeably higher in N. damascena (Table 1). Specifically, of the repetitive elements in N. damascena, TEs accounted for 9.59 Gb, exceeding those in the other four species by 1.94-, 4.14-, 69.76-, and 14.49-fold, respectively (Table 1, Supplementary Data 13). Table 1Comparison of genomic features across ranunculaceous speciesParameterNigella damascenaAconitum vilmorinianumRanunculus cassubicifoliusAquilegia coeruleaCoptis chinensisAssembly size (Gb)10.845.762.690.310.94Number of chromosomes (2n)1216161418Number of protein-coding genes35,74760,35131,32230,02340,011Mean gene length (bp)97155509365028303248Mean CDS length (bp)1082121011351133964Mean exon length (bp)280304260244210Mean exon number per gene3.863.984.374.644.60Mean intron length (bp)30161441747466634Repeat content (Gb/%)9.61/88.604.95/85.912.33/86.640.14/45.500.66/70.97TE content (Gb/%)9.59/88.404.94/85.712.32/86.060.14/44.830.66/70.69Post-Ranunculaceae WGD eventsnoneone roundnonenonenoneNote: Genomic features of N. damascena that are markedly different from those of the other four ranunculaceous species are highlighted in bold. Since the untranslated regions (UTRs) of all genes were not fully annotated in the five species, the term “exon” here specifically refers to the protein-coding region. Comparison of genomic features across ranunculaceous species Note: Genomic features of N. damascena that are markedly different from those of the other four ranunculaceous species are highlighted in bold. Since the untranslated regions (UTRs) of all genes were not fully annotated in the five species, the term “exon” here specifically refers to the protein-coding region. We next conducted an intra-genomic comparison in N. damascena and identified 96 self-syntenic blocks, revealing an obvious WGD event (Fig. 2a, Supplementary Fig. 5a). Subsequent inter-genomic synteny analyses revealed one-to-two syntenic blocks between N. damascena and Ac. vilmorinianum, but one-to-one blocks with the other three ranunculaceous species (R. cassubicifolius, Aq. coerulea, C. chinensis), Epimedium pubescens (Berberidaceae), and Papaver rhoeas (Papaveraceae), suggesting a shared ancient WGD event across Ranunculales and a more recent WGD in Ac. vilmorinianum (Figs. 1, 2b, c, Supplementary Fig. 6). The number of synonymous substitutions per synonymous site (KS) analyses revealed a single peak with the KS value > 1.0 for syntenic paralogs within each species (Supplementary Fig. 5b), further supporting the ancient WGD event shared by Ranunculales species. After evolutionary rate correction, the two WGD events were dated to approximately 128.12 million years ago (Mya; adjusted KS peak value = 1.08) and 50.60 Mya (adjusted KS peak value = 0.43), respectively, predating the origins of Ranunculales and Delphinieae (Figs. 1, 2d). Collectively, these results indicate that genome expansion in N. damascena is associated with massive TE accumulation, rather than WGD, as the identified WGD event predated the divergence of Ranunculales. Patterns and dynamics of TE expansion during the evolution towards N. damascena To elucidate how TEs shape the giant genome, we first compared their accumulation patterns across N. damascena, Ac. vilmorinianum, R. cassubicifolius, Aq. coerulea, and C. chinensis. Within genic regions, N. damascena contained 162.45 Mb of TEs, largely comparable to Ac. vilmorinianum (132.59 Mb) but markedly higher than R. cassubicifolius (33.01 Mb), Aq. coerulea (11.45 Mb), and C. chinensis (36.57 Mb) (Fig. 3a, b, Supplementary Data 14). Within intergenic regions, 9.40 Gb of TEs were accumulated in N. damascena, exceeding Ac. vilmorinianum, R. cassubicifolius, Aq. coerulea, and C. chinensis by 1.97-, 4.16-, 80.44-, and 15.78-fold, respectively (Fig. 3a, b, Supplementary Data 14). Among the five major TE categories, LTRs (9.18 Gb) and DNA transposons (387.30 Mb) were substantially more abundant and longer in N. damascena than in Ac. vilmorinianum, R. cassubicifolius, Aq. coerulea, and C. chinensis, exceeding the four species by 2.07-/2.32-, 4.33-/10.61-, 112.27-/35.56-, and 17.66-/8.19-fold, respectively (Fig. 3c, Supplementary Data 13). These findings suggest that massive TE amplification in intergenic regions, accompanied by moderate accumulation in genic regions, represents a primary cause of genomic gigantism in N. damascena, with LTRs and DNA transposons serving as predominant contributors. Distribution and accumulation patterns of TEs across ranunculaceous species. a Total TE contents in the five genomes. b Representative genomic regions showing TE density and types in the five species. Two subregions are enlarged to highlight TE accumulation in genic and intergenic areas in N. damascena. Gray lines connect syntenic gene pairs. c Sizes of five TE types in the five genomes. d–i Estimated insertion time of all TEs (d) or a certain TE type (e–i) in the five species. Each column represents the proportion of all TEs (d) or a certain TE type (e–i) in a given insertion time relative to the total TE content of all five species. The proportions are calculated separately for each species. Nida, N. damascena; Acvi, Ac. vilmorinianum; Raca, R. cassubicifolius; Aqco, Aq. coerulea; Coch, C. chinensis. Mya, million years ago. We then compared TE insertion dynamics and elimination efficiency across the five ranunculaceous species. Two strategies were applied to estimate TE insertion time and accumulation dynamics. Analysis restricted to intact LTRs revealed recent activity bursts within the past 10 million years in all five species (Supplementary Fig. 7). In contrast, when all annotated TEs were considered, distinct accumulation patterns emerged across species with different chromosome types. In species with R-type chromosomes (i.e., N. damascena, Ac. vilmorinianum, and R. cassubicifolius), TEs accumulated persistently since 60 Mya, roughly coinciding with the evolutionary radiation of lineages bearing this chromosome type (Figs. 1, 3d). In C. chinensis with C-type chromosomes, TE accumulation appears to have occurred more recently, whereas no pronounced TE bursts were detected in Aq. coerulea with T-type chromosomes (Fig. 3d). In N. damascena, approximately 92% of TE insertions postdated the divergence of Nigella from its sister lineage (Delphinieae, ~59 Mya), and the TE burst around 27 Mya coincided with the early diversification of Nigella, reflecting predominantly lineage-specific accumulation (Figs. 1, 3d). Furthermore, different TE types contributed differentially to genomic gigantism in this species: LTRs and short interspersed nuclear elements (SINEs) largely mirrored the overall TE accumulation pattern, while DNA transposons, unclassified TEs, and long interspersed nuclear elements (LINEs) peaked earlier, at 37, 44, and 51 Mya, respectively (Fig. 3e–i). The TE elimination efficiency was assessed by calculating the ratio of intact LTRs to the sum of solo and truncated LTRs. The latter two LTR types represent the products of two major TE elimination mechanisms, i.e., unequal homologous recombination and double-strand break (DSB)–associated deletions62. The results showed that except for R. cassubicifolius (ratio of ~1:8), the other four species exhibited markedly high TE removal efficiency, with ratios of ~1:195 for N. damascena, ~1:199 for Ac. vilmorinianum, ~1:70 for Aq. coerulea, and ~1:117 for C. chinensis (Supplementary Data 15), regardless of their genome sizes. Taken together, these results indicate that genomic gigantism of N. damascena was driven by prolonged and gradual TE accumulation, whereas genome contraction in Aq. coerulea resulted from the combined effects of low TE activity and efficient DNA elimination. Overall impacts of TE amplification on gene evolution To assess the impacts of extensive TE amplification on gene evolution, we screened for TE-containing genes in genomes of N. damascena, Ac. vilmorinianum, R. cassubicifolius, Aq. coerulea, and C. chinensis. A total of 32,863 (91.93%) genes harboring TEs were identified in N. damascena, with 11,339, 12,795, and 31,511 genes having TEs in CDSs, introns, and flanking regions (i.e., 1 kb upstream of the translation start site and 1 kb downstream of the translation end site), respectively (Fig. 4a). The proportions of genes containing TEs in different regions and the median number/length of TEs in CDSs and flanking regions were largely comparable across species (Fig. 4a). However, genes in N. damascena contained markedly more and longer TEs within introns, with a median TE number of 12 per gene and a median length of 3658.0 bp (Fig. 4b, c). A total of 4194 genes in N. damascena had ultra-long introns with a total intronic TE length ≥ 10 kb, far exceeding the counts in R. cassubicifolius (624 genes), Aq. coerulea (26 genes) and C. chinensis (295 genes) (Supplementary Data 16). Overall impacts of TE amplification on protein-coding genes. a Number of genes containing TEs in their CDSs, introns, and flanking regions (1 kb upstream of the translation start site and 1 kb downstream of the translation end site) in the five ranunculaceous species. Numbers in parentheses represent the percentage of TE-containing genes among all protein-coding genes in each species. b, c Median number (b) and length (c) of TEs in CDSs, introns, and flanking regions of genes in the five species. In b, c, the box represents the first and third quartiles with a horizontal line indicating the median, and the whiskers extend to the minimum and maximum values, with outliers omitted. The number of genes in each set of b, c corresponds to those shown in a. d Impacts of ultra-long introns on gene expression. e Examples of two genes (E2F1, an up-regulated gene in N. damascena; UGE5, a down-regulated gene in N. damascena), showing their gene structures, TE contents, and expression profiles in the three species. Data are presented as mean values + standard deviation (SD) of three biological replicates. f Impacts of TE insertions on coding sequences. g Examples showing TE-induced exonic region insertion (KANADI 2 [KAN2]) and stop codon alteration (Nida5G03533), which lead to coding sequence divergence in N. damascena relative to the other two species. Red circles mark TE insertions that cause protein sequence changes. h Examples of two Nigella-specific TE-derived genes (Nida3G05031 and Nida4G00472) in N. damascena. In all examples, gene structures are shown from 5’ to 3’, with arrows indicating the transcriptional direction. Nida, N. damascena; Acvi, Ac. vilmorinianum; Raca, R. cassubicifolius; Aqco, Aq. coerulea; Coch, C. chinensis. Source data are provided as a Source Data file. To understand the impacts of intronic TE accumulation on gene expression, we conducted comparative transcriptomic analyses. We found that in N. damascena, the 4194 genes having ultra-long introns exhibited significantly higher expression levels than shorter genes (<10 kb in total length) across ten examined organs (i.e., roots, stems, leaves, flowers, fruits, seeds, sepals, petals, stamens, and carpels), suggesting that TE-enriched long introns were generally associated with enhanced transcriptional activity (Supplementary Fig. 8, Supplementary Data 17). When comparing the expression profiles of 2538 low-copy (1 ≤ N ≤ 2) genes with ultra-long introns in N. damascena to their orthologs with short intronic TEs (a total intronic TE length ≤ 1 kb) in Aq. coerulea and C. chinensis, 46.38% (1177 genes) were conserved, while others showed divergent expression between N. damascena and the other two species (Supplementary Data 18–22). The divergent genes included 627 up-regulated and 648 down-regulated genes, and 86 genes showing organ-dependent changes (Fig. 4d, e, Supplementary Data 21, 22). As a control, when the expression profiles of 3609 N. damascena genes having short intronic TEs were compared with their orthologs in Aq. coerulea and C. chinensis, only 24.05% (868) showed conserved expression (Supplementary Data 21, 22). A chi-square test revealed that TE-enriched genes showed significantly higher expression conservation than the control group (P < 0.01; Supplementary Data 22). These findings suggest that TE expansion has led to increased intron size, which in turn is strongly associated with enhanced transcriptional activity and gene expression conservation. To explore the consequences of TE insertions in coding regions, we identified orthologs of 11,339 N. damascena genes containing TE-derived coding sequences in the Ranunculaceae phylogenetic context. Sequence comparison revealed that 415 of these genes, while possessing orthologs in other ranunculaceous species, encoded proteins with a stretch of extra amino acids that were either specific to N. damascena (331 genes) or shared across the Nigella genus (84 genes) (Fig. 4f, g, Supplementary Data 23). These coding sequence changes were primarily caused by alterations in the initiation codon (116 genes), stop codon (179 genes), or both (6 genes), with additional contributions from exonic region insertion (57 genes), exonization (21 genes), and intron retention (36 genes) (Fig. 4f, g, Supplementary Fig. 9, Supplementary Data 23). Gene Ontology (GO) enrichment analysis showed that these genes are mainly involved in anatomical structure development and regulation of chlorophyll biosynthetic process (Supplementary Data 24). We also discovered 2473 Nigella-specific genes, whose coding sequences are largely or entirely TE-derived. Among these genes, 1548 were unique to N. damascena and 925 were shared exclusively within Nigella species (Fig. 4f, h, Supplementary Fig. 10, Supplementary Data 25). In the two gene sets, 708 (45.74%) and 393 (42.49%) were expressed in at least one of the 10 organs examined, with 245 of the 708 and 137 of the 393 genes exhibiting organ-specific or preferential expression (Supplementary Data 25). Taken together, these findings suggest that TEs in coding regions have driven gene evolution in N. damascena by reshaping existing genes or creating Nigella-specific genes. Contributions of TEs to the evolution of new morphological traits Nigella flowers are characterized by highly elaborate bilabiate petals and syncarpous gynoecia. To investigate the genomic basis of, and the potential contributions of TEs to, the evolution of these traits, we first compared floral MADS-box genes across N. damascena, Aq. coerulea, and C. chinensis (Supplementary Figs. 11, 12). Gene expression profiling revealed that most orthologs, including the petal identity gene APETALA3-3 (AP3-3) and the carpel identity gene AGAMOUS2 (AG2), exhibited conserved spatiotemporal expression patterns across species (Fig. 5a, Supplementary Data 26). In N. damascena, four genes showed expression patterns that differed from those in the other two species: NidaAP3-1 was restricted to petals/stamens and NidaAP3-2 to stamens; NidaAG1 and NidaSEPALLATA2 (NidaSEP2) showed the lowest expression levels (Fig. 5a, Supplementary Data 26). Further sequence analyses indicated that while the protein sequences and exon-intron structures of the four genes were conserved, NidaAP3-1, NidaAG1, and NidaSEP2 harbored higher TE densities in their promoters and introns compared to their orthologs (Supplementary Fig. 13). Genomic basis underlying the evolution of floral traits in N. damascena. a Expression profiles of floral MADS-box genes at four developmental stages of flowers of N. damascena, Aq. coerulea, and C. chinensis. Mature flowers of the three species and their floral diagrams are shown on the left. Sky blue, sepal (Se); purple, petal (Pe); orange, stamen (St); pink, staminode; dark green, carpel (Ca). b, c Possible mechanisms underlying the bilabiate petal elaboration. b Numbers of shared and specific target genes of NidaAP3-3, AqcoAP3-3, and CochAP3-3, with representaive targets listed for each group. c Examples showing the predicted TE-derived CArG box (red triangles) only in NidaWHIRLY 1 (NidaWHY1), NidaMED25, and NidaZIM-LIKE 2 (NidaZML2). d, e Possible mechanisms underlying the evolution of syncarpy. d Expression profiles of candidate genes involved in the cuticular wax biosynthesis of N. damascena, Aq. coerulea, and C. chinensis. Diagrams showing the syncarpous gynoecium of N. damascena and apocarpous gynoecia of Aq. coerulea and C. chinensis. Floral organs from four to eight developmental stages are sampled. Data are presented as mean values + SD of three biological replicates. e Gene structures of LCR in the three species. The N. damascena-specific DNA transposon sequence is highlighted. TIR, terminal inverted repeat of miniature inverted repeat transposable element (MITE); TSD, target site duplication of MITE. In all examples, gene structures are shown from 5’ to 3’, with arrows indicating the transcriptional direction. Source data are provided as a Source Data file. Given that the observed expression and sequence changes do not involve the petal and carpel identity genes, we next focused on other genes and programs potentially involved in the two floral innovations. We attempted to understand the evolution of bilabiate petals in N. damascena by comparing potential targets of AP3-3 orthologs across this species, Aq. coerulea (with spurred petals), and C. chinensis (with flat petals). Since the DNA-binding MADS domains are highly conserved among AP3-3 orthologs, the ChIP-seq-validated CArG box motifs from AqcoAP3-363 were used to scan the genomes of the three species for putative AP3-3 binding sites (Supplementary Fig. 14a–c). By integrating the CArG box position and expression data, we identified 7127, 7089, and 4922 putative target genes in N. damascena, Aq. coerulea, and C. chinensis, respectively (Fig. 5b, Supplementary Data 27–29). Of these targets, 3826 (53.68%) of N. damascena were shared with Aq. coerulea (1891), C. chinensis (956), or both (979) and were primarily enriched in processes including flower development, specification of floral organ number, cell division and growth, and adaxial/abaxial pattern specification (Fig. 5b, Supplementary Data 27, 30). The remaining 3301 targets were unique to N. damascena and were significantly enriched in processes such as cell cycle and trichome papilla formation (Fig. 5b, Supplementary Data 27, 30). Of the NidaAP3-3-specific targets, 48.17% (1590/3301) harbored TE-derived CArG boxes, a frequency markedly higher than those observed in AqcoAP3-3-specific (31.46%, 1085/3449) and CochAP3-3-specific (37.02%, 826/2231) targets (Fig. 5c, Supplementary Data 27–29). These results indicate that independent recruitment of NidaAP3-3-specific targets, largely through TE-mediated network rewiring, played critical roles in the evolution of elaborate bilabiate petals in N. damascena. To elucidate the evolutionary origin of syncarpy in Nigella, which involves postgenital fusion potentially associated with cuticular wax disruptions64,65, we retrieved 96 potential cuticular wax biosynthetic pathway genes in N. damascena and their orthologs from Aq. coerulea and C. chinensis for comparison (Supplementary Data 31). Gene expression profiling revealed 11 genes of N. damascena that were highly expressed during the late developmental stages of sepals, petals, and stamens, but were markedly down-regulated in carpels, hinting at an association with postgenital fusion of carpels (Fig. 5d, Supplementary Fig. 15). However, among these genes, five were similarly down-regulated in carpels of Aq. coerulea and/or C. chinensis (Supplementary Fig. 15), ruling out their specific evolutionary roles in carpel fusion in N. damascena. Of the six remaining genes, homologs of HYDROXYACYL-COA DEHYDRATASEs (HCDs) and MID-CHAIN ALKANE HYDROXYLASE1s (MAH1s) underwent recent tandem or proximal duplications (Supplementary Fig. 16) and nearly all copies in the three species contained TEs within their genic regions (Supplementary Fig. 17). In contrast, LACERATA (LCR) remained single-copied across the Ranunculaceae and NidaLCR harbored a DNA transposon in its downstream regulatory region, a feature that was absent in AqcoLCR and CochLCR (Fig. 5e). Based on these findings and the evidence that the lcr mutants in Arabidopsis thaliana exhibit postgenital organ fusion66, we propose that the lineage-specific TE insertion in the regulatory region highly likely caused carpel-specific downregulation of NidaLCR, thereby contributing to the evolution of syncarpy in N. damascena. Contributions of TEs to the evolution of phytochemical traits β-elemene, a sesquiterpene with antitumor activity, accumulates uniquely in Nigella seed oils within the Ranunculaceae34,67,68. To gain insights into its evolution, we compared sesquiterpene biosynthesis genes across N. damascena, Aq. coerulea, and C. chinensis. We found that genes encoding β-elemene precursor synthases were all present in the three species with small copy number variation and widely expressed (Fig. 6a, Supplementary Fig. 18, Supplementary Data 32). In contrast, the terpene synthase A subfamily (TPSa) genes showed dramatic copy number variation in N. damascena (22 genes), Aq. coerulea (11 genes), and C. chinensis (1 gene) (Fig. 6a). Two N. damascena TPSa genes, NidaTPSa1-1 and NidaTPSa1-2, were specifically expressed in β-elemene-enriched organs (fruits and seeds), whereas TPSa genes of Aq. coerulea and C. chinensis exhibited extremely low or undetectable expression in organs examined (Supplementary Fig. 18, Supplementary Data 32). Phylogenetic and collinearity analyses revealed that NidaTPSa1-1 and NidaTPSa1-2 arose from a recent tandem duplication possibly mediated by LTRs (Fig. 6b, c). Moreover, the LTR insertion time was estimated to be 37.03 Mya (Fig. 6d), occurring slightly earlier than the divergence of Nigella (Fig. 1). Collectively, these results suggest that LTR-mediated tandem duplication and subsequent expression specialization of TPSa subfamily genes likely underpinned the evolution of β-elemene in N. damascena. Genomic basis underlying the evolution of phytochemical traits in N. damascena. a–d Possible mechanisms underlying the evolution of β-elemene. a Copy number variation of β-elemene biosynthesis genes among N. damascena (Nida), Aq. coerulea (Aqco), and C. chinensis (Coch). b Phylogeny of TPSa subfamily genes in the Ranunculaceae. c Collinearity of the genomic regions surrounding NidaTPSa1-1/-2/-3 genes among N. damascena, Aq. coerulea, and C. chinensis. Gray lines connect syntenic orthologs. d Estimated insertion time of LTRs flanking NidaTPSa1-1/-2/-3 genes. The thick bar represents the interquartile range (IQR) between the first and third quartiles, with the median marked by a white dot and labeled with the corresponding value; the whiskers extend to the minimum and maximum values. e–h Possible mechanisms underlying the loss of protoberberine alkaloids (PAs). e Copy number variation of PA biosynthesis genes among N. damascena, Aq. coerulea, and C. chinensis. f Collinearity of genomic regions surrounding SOMT, CYP719, STOX, and BBE genes among N. damascena, Aq. coerulea and C. chinensis. Gray lines connect syntenic orthologs, with the gene pairs of interest highlighted in pink. Yellow and green lines indicate a small inversion near SOMT. g Dot plot comparisons of CochSOMT, CochCYP719f, and CochSTOX, with their putative orthologous regions in Aq. coerulea and N. damascena. Black and red dots represent sequence alignments on the same strand or reverse strand, respectively. h Estimated insertion time of LTRs located in the regions corresponding to SOMT, CYP719, and STOX in N. damascena, as shown in f. The thick bar represents the IQR between the first and third quartiles, with the median marked by a white dot and labeled with the corresponding value; the whiskers extend to the minimum and maximum values. Mya, million years ago. Source data are provided as a Source Data file. PAs, a class of benzylisoquinoline-derived metabolites, are present in the basal Ranunculaceae lineages including Coptis, but absent in genera like Aquilegia and Nigella26,34. To understand the underlying genomic basis, we compared PA biosynthesis genes in N. damascena, Aq. coerulea, and C. chinensis. As expected, all PA biosynthesis genes were present and widely expressed in roots, stems, leaves, flowers, fruits, and seeds of C. chinensis (Fig. 6e, Supplementary Fig. 19, Supplementary Data 33). In contrast, the PA pathway was incomplete in the other two species: the critical berberine bridge enzyme (BBE)-encoding gene was not expressed in any examined organs of Aq. coerulea; four key genes, including BBE, (S)-scoulerine 9-O-methyltransferase (SOMT), CYP719, and (S)-tetrahydroprotoberberine oxidase (STOX), were absent in N. damascena (Supplementary Figs. 19, 20, Supplementary Data 33). Collinearity analyses revealed no conservation at the BBE locus between Aq. coerulea and C. chinensis, suggesting lineage-specific rearrangement (Fig. 6f). In N. damascena, the syntenic blocks corresponding SOMT, CYP719, and STOX were preserved, yet no sequences similar to the three genes were detected, indicative of complete losses (Fig. 6f, g). Further TE analysis revealed that the gene losses likely resulted from LTR-mediated ectopic recombination approximately 30-36 Mya (Fig. 6f, h), also pointing to the origin of Nigella (Fig. 1). These results indicate that the independent losses of PAs in the Ranunculaceae involved different mechanisms, at least including transcriptional silencing of BBE in Aq. coerulea and LTR-mediated gene losses in N. damascena. Discussion In this study, using the Ranunculaceae as an experimental system, we investigate the evolutionary processes, drivers, and consequences of genomic gigantism. Our results reveal that giant genomes in this family have evolved independently and gradually from an ancestral genome of approximately 1.46 Gb. The genomic gigantism of N. damascena (10.84 Gb), the largest diploid genome sequenced to date in the Ranunculaceae, is primarily driven by the persistent accumulation of TEs over the past 60 million years, rather than by WGD events. Pervasive TE insertions have profoundly reshaped the gene repertoire of N. damascena, leading to the generation of 4194 genes with ultra-long introns, 415 genes with significantly altered protein-coding sequences, and 2473 Nigella-specific TE-derived genes. Furthermore, TE-mediated regulatory changes and gene duplications/losses likely underpin the evolution of key specialized traits, including the bilabiate petal elaboration, postgenital carpel fusion, β-elemene production, and PA loss. These findings provide insights into genome size evolution and underscore the critical roles of TEs in driving genomic gigantism and shaping specialized traits. The Ranunculaceae, with a > 200-fold variation in genome size (Supplementary Fig. 1), serves as an excellent system for exploring the evolution of giant genomes. Early studies distinguished three chromosome types (R-, T-, and C-types) within this family and revealed a strong correlation with genome size: R-type chromosomes associate with large genomes, T-type with small genomes, and C-type with medium-sized genomes20,21,69. It has been further inferred that both the large R-type and small T-type chromosomes originated from the medium-sized C-type chromosomes27. Our results not only corroborate this inference, but also indicate that the diverse genome sizes in the Ranunculaceae evolved from an ancestral genome of approximately 1.46 Gb. Following the divergence of the basal three tribes, which largely retained the ancestral genome size, the remaining lineages evolved in two contrasting directions: independent and gradual expansion to large or even giant genomes in some taxa, or drastic contraction to small genomes in the tribe Isopyreae. Moreover, our results showed that variation in genome size among ranunculaceous species is primarily driven by differences in TE insertion dynamics, rather than by changes in TE elimination efficiency. The pattern of genome size evolution in the Ranunculaceae mirrors trends revealed in other angiosperm families harboring giant genomes. For instance, genome sizes in the Orchidaceae vary from 341 Mb to 54.88 Gb, with an inferred ancestral size of 4.72 Gb70. Subsequent diversification involved genome downsizing in some clades (e.g., Apostasioideae, Dendrobium, and Pleurothallidinae) and gradual expansion to giant genomes in others (e.g., Cypripedioideae, Neottieae, Pogonia, Liparis, and Calypso)70. Similarly, the ancestral genomes of Melanthiaceae and Liliaceae were medium-sized (5.27 Gb and 6.58 Gb, respectively)71,72. During their evolution, some lineages maintained relatively stable genome sizes, while others underwent genomic gigantism (e.g., Parideae in the Melanthiaceae, Fritillaria and Lilium in the Liliaceae) or downsizing72,73. Collectively, despite limited evidence, general patterns of giant genome evolution have emerged: 1) giant genomes tend to originate independently from small or medium-sized ancestors in different lineages, and 2) genome size evolution often follows divergent paths, involving either gradual expansion toward gigantism or drastic contraction toward smaller genomes. Presumably, genome expansion mainly through massive TE accumulation and/or WGDs potentially can provide raw material for genetic novelty5, whereas genome contraction through various DNA removal mechanisms likely enhances replication efficiency and reduces nutrient cost62. A key question in understanding genomic gigantism is whether it arises from recent TE bursts or persistent TE accumulation over time. Accumulating evidence demonstrates that TEs constitute a dominant portion of many plant genomes, and their amplification is widely regarded as a major driver of genomic gigantism. In particular, by estimating the insertion time of intact LTRs, several studies have proposed that giant genomes arose primarily through recent LTR amplification, for example within 0-3 Mya in Lilium sargentiae74, 0-5 Mya in Pae. ludlowii and L. davidii var. unicolor7,9, and 0-6 Mya in Al. sativum18, Allium fistulosum75, Paeonia ostii76, and L. regale19 (Supplementary Data 34). Applying a similar strategy, we revealed that intact LTRs accumulated in N. damascena within the past 10 million years. However, ancestral genome size reconstructions showed that the current giant genomes of N. damascena and the aforementioned species resulted from persistent genome expansion over the past tens of million years (Supplementary Data 34). We argue that this apparent discrepancy likely stems from the methodological constraints in dating based solely on intact LTRs, for two main reasons. First, intact LTRs represent only a small fraction of all TEs, thus capturing only part of TE evolutionary history. Second, due to their susceptibility to decay, intact LTRs are predominantly recent in origin; consequently, relying on them alone inevitably underestimates ancient TE amplification events. Instead of focusing on intact LTRs, many animal studies have investigated the genomic gigantism by estimating the insertion time of all TEs, as exemplified by the African lungfish (Protopterus annectens; genome size: 40.05 Gb) and the Antarctic krill (Euphausia superba; genome size: 48.01 Gb)77,78. This comprehensive dating strategy has also been applied to plants. For instance, a recent study of Par. polyphylla var. yunnanensis revealed that over three-fourths of TE insertions occurred prior to 29 Mya, strongly suggesting the critical contribution of ancient TE accumulation to genome expansion10. Applying this method to N. damascena, we observed TE accumulation beginning around 60 Mya, and approximately 92% of TE insertions postdated the divergence of Nigella from its sister lineage Delphinieae, with a major amplification burst around 27 Mya. The inferred dynamics of TE activity in both Par. polyphylla var. yunnanensis and N. damascena correspond closely to the key phases of genome expansion identified through ancestral genome size reconstruction. Furthermore, our study, together with the previous work10, indicates that different TE categories exhibited distinct evolutionary patterns, involving either rapid accumulation in sharp bursts at specific time points or steady expansion over extended periods. Collectively, these studies suggest that, in different lineages of angiosperms, genomic gigantism was not primarily driven by a single, recent burst of LTR activity, but rather by the gradual and persistent accumulation of various TEs over tens of millions of years. Future studies exploring potential triggers of this sustained TE proliferation, such as paleoclimatic events and historical population dynamics79, or other unknown factors, will provide more insights into the evolution of giant genomes. The assembly of the N. damascena genome has advanced our understanding of critical roles of TEs in shaping the evolution of genes and traits. While only a small fraction of TEs is located within genic regions, they have profoundly influenced the evolution of protein-coding genes. First, TEs have substantially increased intron size, yielding a mean intron length of 3016 bp and generating 4194 genes with TE-enriched ultra-long introns. Previous studies in Lilium and Pinus tabuliformis have shown that ultra-long introns do not necessarily impair transcriptional efficiency9,19,80. In N. damascena, we not only observed similar patterns, but also found that 46.38% of genes with TE-enriched ultra-long introns showed conserved expression, a proportion even higher than that of genes with short intronic TEs. Presumably, this high transcriptional activity and conservation may be attributable to elevated DNA methylation in TE-rich introns, as observed in Pi. tabuliformis, Pae. ludlowii, and L. regale7,19,80. Second, TEs have altered protein coding sequences and reshaped gene composition. Specifically, TE insertions, which occurred in Nigella, have introduced extra amino acids in 415 existing genes, through various mechanisms, including alterations of the initiation codon or stop codon, exonic region insertion, exonization, and intron retention. TE insertions also generated 2473 Nigella-specific genes largely derived from TEs. Although the functions of these lineage-specific genes await further characterization, over half were expressed, with some exhibiting tissue specificity, suggesting their potential roles in N. damascena. TE-mediated cis-regulatory divergence contributed to petal elaboration and carpel fusion. The petals of N. damascena are bilabiate, with highly specialized appendages such as long hairs, short trichomes, and pseudonectaries48,81, which distinguish them from the spurred petals of Aq. coerulea and the flat petals of C. chinensis, although all are considered to be derived from stamens and known as nectary leaves31,82. Previous studies have shown that the differences in petal morphology within the Ranunculaceae are primarily attributed to shifts in adaxial/abaxial gene expression, differentiated regional growth, and gene co-option43–48. Here, we found that while approximately half of the AP3-3 targets were shared across species, species-specific targets appear to correlate with distinct specialized characters of each petal type. For instance, in N. damascena, NidaAP3-3-specific targets included genes involved in bilabiate structure formation (e.g., NidaASYMMETRIC LEAVES 2 [NidaAS2]46), trichome development (e.g., NidaCAPRICE [NidaCPC]), growth regulation (e.g., NidaAUXIN RESPONSE FACTOR 8 [NidaARF8] and NidaMEDIATOR SUBUNIT 25 [NidaMED25]), and pigmentation (e.g., NidaLUTEIN DEFICIENT 2 [NidaLUT2]). In Aq. coerulea, AqcoAP3-3-specific targets included genes regulating spur development (e.g., AqcoPOPOVICH [AqcoPOP]43) and chlorophyll degradation (e.g., AqcoPHEOPHYTINASE [AqcoPPH], AqcoPHEOPHORBIDE A OXYGENASE [AqcoPAO], and AqcoACCELERATED CELL DEATH 2 [AqcoACD2]). More importantly, 48.17% of NidaAP3-3-specific targets contained TE-derived CArG boxes, a markedly higher proportion than in the other two species. In addition, our results revealed that the carpel-specific downregulation of NidaLCR, likely associated with a DNA transposon-mediated regulatory change, may have led to the origin of syncarpy. These findings demonstrate how TEs rewire gene regulatory networks and drive evolution of morphological traits in N. damascena, although further functional validation is needed. TEs act as double-edged swords in the evolution of phytochemical traits, promoting both innovation of β-elemene through gene duplication and loss of PAs through gene deletion. It has been suggested that the serially arranged duplicated genes associated with secondary metabolite biosynthesis usually arose from unequal crossover, which occurs frequently in regions rich in TEs or in genomes with TE contents exceeding 70%62,83. Our findings reinforce this idea, as the candidate genes related to the β-elemene biosynthesis in N. damascena, NidaTPSa1-1 and NidaTPSa1-2, were tandemly duplicated and located in regions highly enriched with LTRs. Besides facilitating the acquisition of certain phytochemical traits, we also found that TE-mediated unequal crossover also contributed to their loss. Specifically, genes involved in the biosynthesis of PAs, such as SOMT, CYP719, and STOX, were lost in N. damascena, likely due to LTR-mediated ectopic recombination. Taken together, these findings highlight the significant potential of TEs in driving the evolution of genes and traits. More studies are warranted to decipher a broader functional landscape of TEs in giant genomes. Methods Plant materials and growth conditions Seeds of N. damascena and Aq. coerulea were sown in nutrient soil and cultivated under conditions of 25 °C, 60% relative humidity, and a 12-h-light/12-h-dark photoperiod. Seedlings of Aq. coerulea with four to six true leaves were vernalized for one month at 4 °C under short-day conditions (8-h-light/16-h-dark), then returned to the initial growth conditions. Plants of C. chinensis were purchased from Lichuan, Hubei Province. Ancestral genome size reconstruction To understand the evolution of giant genomes in the Ranunculaceae, we first estimated genome sizes of 20 ranunculaceous species using flow cytometry (Supplementary Fig. 3, Supplementary Data 1). For each species and its appropriate internal reference of known genome size, approximately 30 mg of fresh leaves was chopped in 1 mL of ice-cold Galbraith’s buffer (45 mM MgCl₂, 20 mM MOPS, 30 mM sodium citrate, and 0.1% [v/v] Triton X-100, pH 7.0). The resulting cell suspension was filtered by a 400-mesh nylon filter and stained with propidium iodide (100 μg/mL) for 10 min. The stained nuclei were analyzed on a MoFlo-XDP flow cytometer (Beckman Coulter, USA) and genome sizes were calculated using Summit v5.2 software. We also obtained genome size data of 30 additional ranunculaceous species and 18 outgroup species from previously published studies (Supplementary Data 1, with references listed therein). A time-calibrated phylogeny including the above species (Supplementary Fig. 2) was assembled following an approach similar to that used in the Timetree Project84. In brief, the backbone topology, which reflected relationships and divergence times among relevant families and tribes, was adopted from the comprehensive analysis of He et al27. Subsequently, we manually grafted previously published well-established phylogenies within tribe or genus23,49–61 onto the backbone27 using the tree editing tool in Mesquite v3.7085. The phylogeny was then time-scaled by adjusting branch lengths in Mesquite to correspond with divergence times reported in the original studies. Outgroup taxa were chosen from the six non-Ranunculaceae families of Ranunculales, all with reliable genome size estimates and well-resolved phylogenetic relationships. To select an appropriate evolutionary model for ancestral genome size reconstruction, three commonly used models, i.e., BM, Ornstein–Uhlenbeck (OU), Early Burst (EB), and white-noise (WN; a model lacking phylogenetic signal as a control), were assessed using the fitContinuous function in the R package geiger v2.0.1186. BM was identified as the best-fit model based on weighted corrected Akaike Information Criterion (AICc) values (Supplementary Data 2). Under this model, ancestral genome sizes at different evolutionary nodes were inferred using the anc.ml function and visualized with contMap, both implemented in the R package phytools v1.9.1687. Genome sequencing, assembly, and quality assessment For genome sequencing of N. damascena, genomic DNA was extracted from the young leaves using the DNAsecure Plant Kit (TIANGEN). Using the Illumina NovaSeq system and PacBio Sequel II platform, 1018.20 Gb of paired-end 150 bp short reads and 274.82 Gb of single-molecule long reads were yielded, respectively (Supplementary Data 3, 5). For Hi-C sequencing, seedlings were fixed in formaldehyde, and the cross-linked DNA was digested with MboI restriction endonuclease. By sequencing on the Illumina NovaSeq platform in the 150 bp paired-end mode, 247.04 Gb of valid Hi-C data were generated (Supplementary Data 7). The genome size of N. damascena was estimated using flow cytometry (as described above) and k-mer analysis. For the latter, Jellyfish v2.2.10 was used to generate a 21-mer frequency distribution of Illumina paired-end short reads88, and GenomeScope v2.0 was used to estimate genome size89. For genome assembly, PacBio HiFi reads were de novo assembled using hifiasm v0.2.0 with default parameters90 and redundant sequences were removed using Purge_dups v1.2.591. The resulting contigs were further anchored and oriented to six pseudochromosomes according to the Hi-C contact map, using LACHESIS with the parameter of “CLUSTER_N = 6”92. Misassembled contigs were manually corrected based on Hi-C contact frequency using Juicebox v1.9.093. The genome assembly quality was assessed by following methods: 1) the Illumina paired-end reads were mapped back to the assembly for consistent assessment using BWA v0.7.17, with the parameters of “mem -M -k 30” (https://arxiv.org/abs/1303.3997); 2) BUSCO v5.4.7 with the embryophyta_odb10 database was used to evaluate genome completeness94; and 3) LAI was employed to infer the assembly continuity95. Genome annotation Three approaches were used to predict protein-coding genes in the N. damascena genome. For the transcriptome-based prediction, total RNAs were first extracted separately from six different organs (i.e., roots, stems, leaves, flowers, fruits, and seeds) at various developmental stages using the RNAprep Pure Plant Plus Kit (TIANGEN). Each sample was sequenced on the Illumina NovaSeq system (yielding ~12 Gb of 150 bp paired-end short reads) and the PacBio Sequel II platform (producing ~2 Gb of single-molecule long reads), respectively (Supplementary Data 10). The resulting data were then used to predict gene models using PASA v2.1 with the parameter of “-m = 50”96. For the homology-based prediction, gene models were annotated using Genewise v2.2.0 based on homologous gene sets from the Ar. thaliana, Aq. coerulea, C. chinensis, Papaver somniferum and Macleaya cordata 22,26,97–99. For the ab initio prediction, Augustus v3.3.1, with the parameter “--genemodel = partial”, and GeneMark v4.33, were applied100,101. Gene models derived from these three approaches were integrated using EVidenceModeler (EVM)102. The integrated gene set was functionally annotated by BLAST searches103 (E-value ≤ 1 × 10-5) against the following databases, including SwissProt (https://www.uniprot.org), NT and NR (https://www.ncbi.nlm.nih.gov/), PFAM, eggNOG (http://eggnogdb.embl.de/), GO, and KEGG. Transcription factors were identified using PlanTFDB v5.0 (http://planttfdb.gao-lab.org/prediction.php)104. Two approaches were used to identify repetitive sequences in the N. damascena genome. For the ab initio prediction, a de novo repeat library was constructed using RepeatModeler v2.0.2a105 and an LTR-specific library was generated using LTR_retriever v3.0.4, LTRharvest v1.6.1, and LTR_finder v1.07, all with default parameters106–108. For the homology-based approach, the genome was searched against the Repbase database v28.05109. The two libraries were then used to annotate TEs with RepeatMasker v4.1.2-p1110. The solo and intact LTRs were identified by LTR_retriever v3.0.4, with the solo LTRs further filtered by TEsorter v1.5.1111. The truncated LTRs were retrieved by removing solo and intact LTRs from the total LTRs using BEDTools v2.31.0112. tRNAs were predicted using tRNAscan-SE v1.3.1 with default parameters113. rRNAs, snRNAs, and miRNAs were identified by mapping the genome to the Rfam database (http://rfam.xfam.org/) with Infernal v1.1.3 under default settings114. Identification of WGD events To identify WGD events in N. damascena, we performed synteny analyses both within its genome and with genomes of six other Ranunculales species22,26,28,30,99,115. Putative syntenic genes were identified through an all-against-all alignment of protein sequences using BLASTP v2.5.0+ (E-value ≤ 1 × 10-5)103. The BLAST results were submitted to MCScanX for inferring syntenic blocks under default parameters116. For each block, we calculated KS of syntenic anchor genes using the Nei-Gojobori method in KaKs_Calculator v2.0117. The median KS values of these anchor genes were used to estimate the divergence level of each syntenic block. The occurrence times of WGD events were estimated based on the divergence time between N. damascena and Pap. rhoeas (109.6 Mya) and the corresponding corrected KS peak value (0.92) for this divergence. Following the approach of Wang et al.118, KS peak values were corrected as follows: (1) using the paralogous KS peak value of Aq. coerulea to calibrate all species to the ancient Ranunculales WGD event, (2) using the orthologous KS peak value between Pap. rhoeas and R. cassubicifolius to calibrate the divergence of the Papaveraceae from other Ranunculales lineages, and (3) using the orthologous KS peak value between Aq. coerulea and N. damascena to calibrate the divergence between Aq. coerulea and R. cassubicifolius, as well as that between N. damascena and R. cassubicifolius. Estimation of TE insertion time The TE insertion time was estimated using two strategies. The first was applied specifically to intact LTRs, for which the nucleotide distance (K) was calculated based on the pairwise alignment of its 5’ and 3’ terminal repeats using the Kimura two-parameter (K2P) model119. The second strategy involved all annotated TEs, for which the K value of each TE was calculated based on the pairwise alignment of the TE sequence and its corresponding consensus sequence. In both strategies, the TE insertion time (T) was estimated based on the following formula where r (2.85 × 10-9) is the average substitution rate per site per year. The r value was calculated based on nucleotide sequences of 1007 single-copy genes from the five ranunculaceous and two outgroup species (Pap. rhoeas and E. pubescens) using r8s v1.81 with default parameters120, with the divergence times between the Papaveraceae and the Ranunculaceae (102.9-117.2 Mya) and between the Berberidaceae and the Ranunculaceae (84.0-97.1 Mya) as calibration points27. Ortholog identification and gene gain/loss analysis Protein-coding sequences of 33 species (30 from the Ranunculaceae, two from the Berberidaceae, and one from the Papaveraceae) were retrieved from previously published studies (Supplementary Data 18, with references listed therein). These sequences, along with those from N. damascena, were clustered into putative gene families/subfamilies (orthogroups) by OrthoFinder v2.5.4121, with the parameters of “-S blast -M msa -T raxml -t 128”. Gene family gain/loss events were inferred based on the orthogroup clustering results using the Wagner parsimony method in Count122. Gene expression profiling RNA-seq data from mature organs (i.e., roots, stems, leaves, flowers, fruits, and seeds) and floral organs at different developmental stages of N. damascena, Aq. coerulea, and C. chinensis were used to generate gene expression profiles, with three replicates per organ at each stage. The developmental stages of flowers in N. damascena and C. chinensis were defined according to morphological events corresponding to those in Aq. coerulea123. RNA-seq data for floral organs of N. damascena and roots, stems, and leaves of C. chinensis were retrieved from published studies26,48. All remaining data from 174 samples were obtained in this study (Supplementary Data 17). For each sample, the total RNA was extracted and sequenced as described above, yielding an average of 6 Gb of paired-end (150 bp) clean data. The clean reads were mapped to their corresponding reference genomes using HISAT2 v2.21 with default parameters124. Transcripts per million (TPM) values of genes in a given sample were calculated using Stringtie v2.21125 and the averaged TPM value from three replicates was used to represent the gene expression level. To ensure comparability of gene expression levels across species and organs, we normalized the averaged TPM values following the approach described by Brawand et al.126. Specifically, the highly conserved 1:1:1 single-copy genes were first identified from N. damascena, Aq. coerulea, and C. chinensis. These genes were then ranked based on their TPM values in each sample, and 147 genes with the most conserved ranks were selected, whose rankings consistently fell within the interquartile range across samples (Supplementary Data 19). The median TPM values of these genes in each sample were further calculated, and adjusted to a common value using derived scaling factors (Supplementary Data 20). Finally, these scaling factors were applied to normalize the TPM values of all genes within each sample. All TPM values shown in this study are normalized unless otherwise specified. Evaluation of the impacts of TE insertions on genes TE insertion sites were defined according to their overlap with predicted genomic regions (CDSs, introns, and flanking regions). The numbers and lengths of TE insertions in each genomic region were determined as follows: if a TE was entirely located within a single genomic region, it was counted as one insertion and the TE length was recorded as the insertion length; if a TE spanned multiple regions (e.g., CDS and flanking regions), it was counted as one insertion in each corresponding region, and the length within each region was recorded as the TE insertion length in that region. Gene expression profiles from ten matched organs of N. damascena, Aq. coerulea, and C. chinensis, including six mature organs (i.e., roots, stems, leaves, flowers, fruits, and seeds) and four types of floral organs (i.e., sepals, petals, stamens, and carpels), were used to evaluate the impact of intronic TE accumulation on gene expression. Note that for sepals, petals, stamens, and carpels, the gene expression level was determined in terms of the highest TPM value among the developmental stages. A gene was classified as having conserved expression if its expression levels across all organs in N. damascena were similar to those of its orthologs in the other two species. A gene was considered up- or down-regulated in N. damascena if it showed at least twofold expression difference in any given organ relative to its orthologs, while showing conserved expression across all other organs. Organ-dependent expression changes were defined as cases where a gene exhibited markedly higher expression in some organs but lower in others compared to its orthologs in the other two species. A Chi-square test assessing the association between intronic TE content and gene expression divergence was conducted using R v4.3.2. A gene was defined as a Nigella-specific TE-derived gene through two steps. First, genes in N. damascena whose protein-coding sequences partially or completely overlapped with TE sequences were identified. Second, their orthologs were identified using OrthoFinder as described above, and only those lacking orthologs outside the Nigella genus were considered as Nigella-specific TE-derived genes. Phylogenetic analyses of genes associated with floral and phytochemical traits Floral MADS-box genes from N. damascena41, cuticular wax and sesquiterpene biosynthesis genes from Ar. thaliana127,128, and PA biosynthesis genes from C. chinensis as queries26 were used as queries to search for corresponding homologs in the genomes or transcriptomes of Ranunculales species, using BLASTP v2.5.0 + 103, with an E-value threshold of 1.0 × 10-5. Amino acid sequences of each gene lineage were aligned with MAFFT v7.505 and the corresponding codon-based nucleotide alignment was generated using PAL2NAL v.14129,130. The maximum likelihood tree was constructed using IQ-TREE 2.1.4-beta with the parameters of “-bb 1000 -nt AUTO”131. Prediction of potential target genes of AP3-3 orthologs The protein structures of NidaAP3-3, AqcoAP3-3, and CochAP3-3 were predicted using AlphaFold Server132. To identify the potential target genes of NidaAP3-3 and CochAP3-3, the four ChIP-seq-validated CArG box motifs from AqcoAP3-3 were used to scan against the genomes of N. damascena, Aq. coerulea, and C. chinensis63. Genes containing at least one CArG box within their regulatory regions (from 3 kb upstream of the translation start site to 1 kb downstream of the translation end site, excluding coding sequences) and expressed in petals were considered candidate targets. A CArG box was defined as TE-derived if it completely or partially overlapped with a TE sequence (Supplementary Fig. 14d). Gene Ontology enrichment analysis GO enrichment analyses of focal gene sets were performed using their homologous genes in Ar. thaliana as input. Enriched biological process terms were identified using g:Profiler133, with a Benjamini–Hochberg false discovery rate (FDR) threshold of 0.05. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Informaiton Peer Review file Description of Additional Supplementary Files Supplementary Data 1-34 Reporting Summary Source data Source data Acknowledgements We thank Prof. Li Guo and Jie Sun for annotating TEs and members of the Kong laboratory for helpful discussions. We also thank the American Philosophical Society for granting permission to adapt the original chromosome drawings from Gregory (1941) for this publication. Author contributions X.F., H.S. and H.K. designed the project; X.F., C.P., C.X., R.J. and Y.Y. collected samples and prepared sequencing libraries; X.F., T.L. and B.L. conducted data analyses with the help of Y.J. and G.X.; X.F., H.S. and H.K. wrote the manuscript with contributions from T.L., B.L., C.P., X.Yao, Y.J., G.X., R.Z., J.C. and X.Yin. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding This work was supported by the National Natural Science Foundation of China (32221001, 32370231, 32200185). Data availability The genome sequencing data for N. damascena and the transcriptome sequencing data for N. damascena, Aq. coerulea, and C. chinensis have been deposited to National Genomics Data Center under accession PRJCA041492. The genome assembly and annotations of N. damascena are available at Figshare [https://figshare.com/s/7e80a04b8c87b97b24ea]. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Xuehao Fu, Tianyu Lei, Boka Li. Contributor Information Hongyan Shan, Email: shanhongyan@ibcas.ac.cn. Hongzhi Kong, Email: hzkong@ibcas.ac.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-75891-z. References Associated Data Supplementary Materials Supplementary Informaiton Peer Review file Description of Additional Supplementary Files Supplementary Data 1-34 Reporting Summary Source data Data Availability Statement The genome sequencing data for N. damascena and the transcriptome sequencing data for N. damascena, Aq. coerulea, and C. chinensis have been deposited to National Genomics Data Center under accession PRJCA041492. The genome assembly and annotations of N. damascena are available at Figshare [https://figshare.com/s/7e80a04b8c87b97b24ea]. Source data are provided with this paper.