Chromosome-scale assembly of the giant Fritillaria cirrhosa genome elucidates high-altitude adaptive trade-offs

Dear Editor,

The ecological drivers and biological consequences of genome gigantism remain a central question in evolutionary biology, echoing the classic C-value paradox. In plants, giant genomes provide a valuable opportunity to examine how repeat accumulation reshapes genome architecture, gene structure, and potentially organismal biology. Fritillaria is one of the most extreme examples of genome expansion without recent lineage-specific polyploidization, with diploid 1C values up to 85.38 Gb.¹ As a representative species of this genus, F. cirrhosa is of dual interest. Biologically, it is a rare alpine geophyte distributed in the Qinghai-Tibet Plateau and adjacent high-mountain regions characterized by harsh abiotic conditions. Medicinally, it is one of the principal botanical sources of “Chuanbeimu (川贝母),” a highly valued traditional Chinese medicine widely used for respiratory disorders. Traditionally, younger bulbs are preferred due to their distinct pharmacological efficacy, particularly in relation to steroidal alkaloid accumulation. These features raise two related questions. From an evolutionary perspective, how has a species bearing such a gigantic genome adapted to extreme alpine environments? From a developmental perspective, what molecular programs underlie the age-dependent variation in medicinal quality observed across bulb growth stages? However, the absence of a chromosome-scale reference genome has limited both evolutionary and developmental analyses. Here, we report the first chromosome-scale assembly of the F. cirrhosa mega-genome, unraveling the structural basis of genome gigantism and providing insights into its alpine adaptation and developmental remodeling of medicinal metabolites.

Chromosome-scale assembly and annotation of the F. cirrhosa mega-genome

The genome size of F. cirrhosa was estimated to be 31.03 Gb by flow cytometry and 33.34 Gb by K-mer analysis, with a heterozygosity rate of 2.38%. Assembly of 2.07 Tb Oxford Nanopore reads yielded a 39.71 Gb contig-level assembly with 12,638 contigs and 35.35% GC (Figure 1A). The larger assembly size likely reflects underestimation by K-mer modeling in highly repetitive regions together with partial retention of uncollapsed heterozygous or repetitive contigs. Hi-C data anchored 36.31 Gb (91.44%) of the assembly to 12 pseudochromosomes, consistent with the karyotype of this species. The assembly achieved a contig N50 of 10.75 Mb and a scaffold N50 of 2.79 Gb. RNA sequencing (RNA-seq) from five organs supported gene annotation, yielding 59,011 predicted protein-coding genes, 97.25% of which were functionally annotated. Assembly quality was supported by 96.2% complete BUSCOs, >99.99% short-read mapping with 98.71% genome coverage, a quality value (QV) of 40, and an long terminal repeat assembly index (LAI) of 15.97 (Figure 1A). This reference provides a practical foundation for comparative genomics, biosynthesis, conservation, and molecular breeding.

Figure 1.

Chromosome-scale assembly and genomic evolution of the F. cirrhosa genome

(A) Circos view of the 12 pseudochromosomes, showing repeat density, Gypsy and Copia density, gene density, and density of functionally annotated genes.

(B) Insertion-time (mya) distribution of intact Copia and Gypsy long terminal repeat retrotransposons (LTR-RTs), indicating continued recent retrotransposon activity that contributed to genome expansion.

(C) Phylogenetic relationships of major intact LTR-RT lineages based on reverse transcriptase domains, highlighting lineage-specific amplification within Copia and Gypsy.

(D) Distribution of synonymous substitution rates (Ks) and whole-genome duplication (WGD) analysis.

(E) Phylogenomic tree and divergence time estimation of F. cirrhosa within the Liliales and other monocot lineages, mapped against geological timescales.

(F) Whole-genome macrosynteny and localized micro-genomic architecture of cold-responsive tandem arrays.

(G) Temporal metabolic shift of key components in F. cirrhosa across five developmental years.

(H) WGCNA-based metabolite-gene correlation network showing candidate hub genes associated with medicinal and stress-related metabolites.

Transposable element expansions drive genome giantization

Repetitive sequences account for 85.23% of the F. cirrhosa genome, with long terminal repeat (LTR) retrotransposons accounting for 68.05%, including 32.42% Gypsy and 8.58% Copia. TE landscape analyses identified major historical amplification waves at ∼10.38 million years ago (mya) for Copia and ∼9.02 mya for Gypsy, while intact LTRs indicate continued recent activity of 0.42 mya for Gypsy and 0.17 mya for Copia elements (Figure 1B). Genome expansion was dominated by a limited number of lineages: Retand, Athila, and Tekay within Gypsy and Ale, Bianca, and Tork within Copia (Figure 1C). Retand elements, which are associated with heterochromatin formation and ancient nested repeat blocks,² were broadly distributed across all 12 chromosomes, exhibiting dispersed local hotspots rather than a dominant chromosome-central concentration. This Retand-biased expansion contrasts with the Athila/Tekay-dominated amplification reported in lily genomes,^3,4,5 indicating distinct TE evolutionary trajectories within Liliaceae.

Despite massive repeat accumulation, rate-corrected synteny and Ks analyses detected no evidence of recent lineage-specific whole-genome duplication (WGD). Instead, the F. cirrhosa genome retains signatures of the ancient monocot τ event and a subsequent WGD shared by Fritillaria, Lilium, and Nomocharis (Figure 1D). Phylogenomic analysis based on 165 single-copy orthologs confirmed the sister relationship between F. cirrhosa and Lilium, with divergence estimated at ∼26.1 mya during the Oligocene-Miocene transition (Figure 1E). Together, these results indicate that repeat expansion after lineage divergence, rather than recent polyploidization, drove genome gigantization in F. cirrhosa.

Genomic trade-offs in high-altitude adaptation

The reference genome provides a framework for dissecting the genomic basis of alpine adaptation. Comparative analyses revealed that expanded gene families are enriched in pathways associated with energy metabolism, oxidative phosphorylation, carbohydrate metabolism, hormone signaling, and glutathione-mediated redox homeostasis, whereas contracted families are concentrated in plant-pathogen interaction and MAPK signaling pathways. This pattern is consistent with emerging evidence that adaptive evolution in alpine plants involves both expansion of abiotic stress-related functions and contraction of biotic defense repertoires.⁶ Positive selection further highlighted key regulators of environmental resilience. Strong selection signals were detected in EREBP, a central regulator of abiotic stress responses, and ATP2A, which mediates calcium-dependent signal resetting. Selection on ICL additionally suggests optimization of the glyoxylate cycle, potentially enhancing lipid mobilization during prolonged dormancy under alpine conditions.

Selection-pressure analyses using HyPhy supported this shift in evolutionary priorities. Several immunity-related genes exhibited significant signatures of relaxed selection (ω < 1, K < 1, p < 0.05), including thiC, PIGN, and genes encoding galacturan 1,4-alpha-galacturonidase. In contrast, genes associated with UV and cold responses, including TOPB1, ATP2A, and ICL, showed intensified selection (K > 1, p < 0.05). Together, these results indicate a genome-wide reallocation of evolutionary constraint toward abiotic stress tolerance.

Comparative genome analyses further identified lineage-specific tandem expansion of several cold-responsive gene families, including Histone H2A, DREB, and dehydrin genes (Figure 1F). These tandem arrays are substantially expanded in F. cirrhosa relative to both Lilium species and are embedded within TE-rich genomic regions dominated by Gypsy and Copia retrotransposons. The close association between tandem duplications and dense TE accumulation suggests that TE-mediated non-allelic homologous recombination may have facilitated local gene amplification, providing a structural substrate for adaptation to alpine environments.⁷

Temporal metabolic reprogramming and defense strategy trade-offs

The chromosome-scale assembly also enabled a genome-guided analysis of age-dependent medicinal quality. Broadly targeted metabolomics across bulbs from year 1 to year 5 revealed a developmental shift in metabolite composition. Bioactive alkaloids and nucleosides declined with age, whereas lipids, lignin-related compounds, and stress-associated saccharides accumulated progressively (Figure 1G). Within the broader framework of coordinated resource allocation, plants can dynamically balance resource investment among growth, defense, and physiological resilience.⁸ This pattern is consistent with the traditional preference for younger bulbs and suggests a life-history transition from early-stage medicinal metabolite accumulation toward later-stage structural and physiological maintenance under alpine stress. These contrasting trajectories are consistent with the reprogramming of resource allocation, potentially reflecting a shift from early investment in alkaloid-based chemical defense toward later investment in structural maintenance and tolerance to alpine abiotic stresses, including ultraviolet radiation. We interpret this trend cautiously because relative abundance alone cannot distinguish active metabolic redirection from dilution caused by biomass expansion in older bulbs.

To investigate the regulatory architecture associated with this transition, we integrated transcriptomic and metabolomic data using Weighted Gene Co-expression Network Analysis (WGCNA). Modules positively associated with alkaloids and nucleosides were distinct from those associated with saccharides and lipids. Within these networks, several expanded genes emerged as candidate hubs, including transcription factors (BZIP53 and AOP1) and a biosynthetic enzyme (GGPP1S) co-expressed with cyclic nucleotides (cGMP and cAMP), alongside VAR3, which positively correlated with steroidal alkaloids (Figure 1H). Expression profiling of metabolite-associated candidates showed that several hubs were developmentally regulated, supporting active transcriptional contributions to metabolic remodeling, although upstream terpenoid pathway genes showed heterogeneous patterns rather than uniform repression. The resulting network therefore provides a genome-anchored candidate framework for future absolute quantification and functional validation of medicinal metabolism.

Linking genomic architecture to developmental metabolic shifts

The repeat-rich genome of F. cirrhosa is also characterized by extensive gene elongation. Genes average 17.8 kb, whereas coding sequences remain compact at 2.52 kb, indicating that genome-scale structural inflation is driven primarily by intron expansion. Notably, 12.8% of genes exceed 50 kb. Although F. cirrhosa has a giant genome, its average gene length is substantially shorter than those reported for L. davidii (57.6 kb) and L. sargentiae (42.8 kb), suggesting lineage-specific evolution of gene architecture within Liliaceae rather than a single uniform long-gene model across giant-genome lilies.

Intron-length distributions revealed a bimodal pattern, indicating that elongation is not uniform across gene classes. Conserved core expressed genes were more strongly expanded than evolutionarily dynamic gene families. Consistent with the “economy selection” hypothesis,⁹ core gene length showed a global negative correlation with mean expression. TE density around exons also increased with gene length, implicating TE insertion as a major driver of structural inflation. These results provide a genome-scale explanation for how repeat accumulation can reshape gene architecture while preserving relatively compact coding regions.

This structural constraint was not identical across developmental modules. Hub genes associated with early-stage alkaloid and nucleoside accumulation showed a largely negative relationship between gene length and expression, whereas age-upregulated genes associated with lipids and stress responses showed a non-linear pattern. These late-stage genes also carried higher densities of Copia insertions, predominantly within introns rather than promoter-proximal regions. Given that TE amplification has been shown to modulate higher-order chromatin architecture and transcriptional regulation,¹⁰ these observations raise the possibility that the F. cirrhosa genome is not a uniform transcriptional environment and that local TE-associated architecture may shape developmental metabolic programs. However, these structure-expression associations remain correlative, and their evolutionary origin may reflect insertion bias, drift, or relaxed purifying selection rather than direct adaptation. For that reason, the main value of this analysis is to define a mechanistic hypothesis space, not to claim that TE accumulation has already been shown to control medicinal quality.

In summary, we present the first chromosome-scale genome assembly of F. cirrhosa and show that its giant genome was shaped primarily by lineage-specific transposable element (TE) expansion rather than recent polyploidization. The genome also reveals signatures of alpine adaptation, including stress-related gene family remodeling, tandem expansion of cold-responsive loci, and shifts in selective constraint. Finally, integration of the reference genome with developmental metabolomics and transcriptomics provides a framework for understanding age-dependent medicinal quality formation. This resource lays the groundwork for future studies of giant-genome evolution, alpine plant biology, and specialized metabolism in F. cirrhosa.

Data and code availability

The chromosome-scale genome assembly and associated multi-omics datasets generated in this study have been deposited in the China National GeneBank DataBase (CNGBdb) under BioProject accession number CNP0009600. More detailed results for this work are provided at https://doi.org/10.6084/m9.figshare.32562357.

Funding and acknowledgments

The research was funded by the National Key Research and Development Program of China (2025YFC3509201); the Scientific and Technological Innovation Project of the China Academy of Chinese Medical Sciences (CI2026A03905, CI2024C003YN, and CI2024E003); the Research and Large-scale Demonstration Project on Breeding of Rare and Endangered Medicinal Plants of NATCM (GHC-2024-ZFGM-405); the Beijing Natural Science Foundation (no. 7254508); the Innovation Team and Talents Cultivation Program of the National Administration of Traditional Chinese Medicine (ZYYCXTD-D-202408); and the Fundamental Research Funds for the Central Public Welfare Research Institutes (grant/award nos. Z2023014, ZZ18-YQ-036, ZZ16-XRZ-067, ZXKT22061, and ZXKT25005).

Declaration of interests

The authors declare no competing interests.

Published Online: September 7, 2026