The Gentiana rigescens genome and transcriptome for the investigation of flower repetitive blooming Authors: Ling Yang, Dan Chen, Yang Dong, Shanshan Chen, Weibin Wang, Chenghua Yu, Shaofeng Duan, Fan Li, Huijuan Yuan, Shengchang Duan, Timothy Charles Baldwin, Chunlian Jin, Jihua Wang, Yanli Liang, Wei Chen Published: 2025 ABSTRACT BackgroundReversible flower opening and closing is a specialized reproductive strategy to improve plant fitness in response to ambient environmental changes. As a common phenomenon in angiosperm plants, the underlying genetic and molecular mechanisms remain unknown. Previous research shows that this phenomenon is a characteristic feature of the genus Gentiana, and the external environmental cues driving petal movements differ among species. However, due to the interaction among environmental stimuli, limited progress was made in disentangling the responsible signaling pathways and putative molecular regulators.ResultsWe identified the alpine species Gentiana rigescens as a model to study repetitive flower opening and closing. Field observations and laboratory investigations confirmed that ambient temperature is the dominant factor driving repetitive flower opening and closing in G. rigescens. We constructed a chromosome-level genome assembly for G. rigescens and performed comparative transcriptomic analyses of flowers under varying natural temperature conditions. Differentially expressed genes revealed Ca2+ channel proteins as candidate temperature sensors. Both transcriptomic and experimental analyses showed that distinct calcium signaling pathways followed by dynamic cell wall modification underlie the repetitive flower blooming at different temperatures. In addition, auxin showed antagonistic effects on flower opening at high temperatures and closing at low temperatures, possibly by affecting the cell wall acidification and the activities of cell wall remodeling enzymes.ConclusionsThis study provides genetic resources for investigating the temperature-regulated repetitive flower opening and highlights the divergent signaling pathways in controlling reversible flower opening and closing.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12915-025-02483-6. Background Flowering in the life cycle of plants requires petals to open and close at the right time to facilitate pollination [1, 2]. In order to maximize reproduction success and survival in various ecological niches, many plants take on unique strategies to synchronize ambient environmental signals with petal movement in the blooming season. Compared to single-opening flowers, long-lasting blooms may take on repetitive opening and closing in a circadian fashion [3]. For example, flowers of water lily (Nymphaeaceae), poppy (Papaveraceae), and some Asteraceae species open during the day and close at night, whereas night bloomers such as Cestrum nocturnum, Cereus grandiflorus, and several Oenothera species open and close petals in the opposite time frame [2–4]. In addition, many species in alpine or arctic regions employ repeated corolla movement to maintain flowers’ internal temperature and protect pollen [5]. Current knowledge posits that corolla movement relies on unequal cell growth on two sides of the petals or a change of petal cell turgor [2, 6], but the underlying molecular mechanisms remain unclear. Despite the commonality of the phenomenon, an easy-to-operate research plant model and the associated genomic resources are still lacking in investigating flower repetitive opening and closing. Members of the Gentiana genus, usually with blue flowers, have cultural importance in many societies as flavoring ingredients, horticultural plants, and herbal medicine [7–9]. For example, the genomes of cultivated Gentiana dahurica, Gentiana macrophylla, and Gentiana straminea helped delineate the biosynthesis of medicinal secondary metabolites (e.g., gentiopicrosides and iridoids) [10–12]. By contrast, wild Gentiana flowers widely grow in the alpine terrain and bloom in a cold and dry environment with substantial day-night temperature differences [1, 13]. Field observations show that wild Gentiana flowers open to bright sunlight and close to low sunlight or physical touch [14–16], thus making Gentiana attractive candidates to investigate repetitive flower opening and closing. This repetitive process is hypothetically under the tight control of both endogenous (e.g., plant hormones) and environmental (e.g., photoperiod and temperature) signals [17]. In Gentiana scabra, biochemical investigations showed that the phosphorylation of an aquaporin protein GsPIP2;2 regulated the flower reopening upon temperature and light stimuli in a calcium-dependent manner, but the exact signaling pathway remains unresolved. So far, the knowledge gap regarding genomic and transcriptomic resources remains for systematically studying the genic network underlying temperature-induced flower repetitive opening in Gentiana. In this study, we reported a chromosome-level genome for Gentiana rigescens Franch. (hereafter G. rigescens, Fig. 1A) and transcriptome data for open and closed flowers at different temperatures. G. rigescens is a perennial herb from the high mountains of Southwestern China that flowers for 2 to 3 months in autumn. We showed that under laboratory conditions, the blooming of G. rigescens primarily responded to temperature and not to light. A combination of genomic, transcriptomic, and experimental analyses helped identify the potential temperature sensors, the involvement of downstream calcium signaling pathways, and the major effector enzymes for cell wall remodeling in regulating repetitive flower opening and closing. We revealed divergent gene expression patterns for differentially expressed genes and the antagonistic effect of auxin on flower opening and closing, which suggests that distinct signaling pathways underlie the low-temperature-induced petal closure and high-temperature-induced petal opening.Fig. 1G. rigescens de novo genome assembly and evolutionary analysis. A Photo of G. rigescens plant. B Genome features and synteny information. I, chromosome size; II, repeat content; III, gene number; IV, GC content; V, gene synteny. C Venn diagram showing orthogroups from six Gentianales species. O. pumila, C. roseus, G. sempervirens, G. rigescens, G. dahurica, and G. macrophylla. D Phylogenetic tree showing divergence times and the evolution of gene family size. E Ks distribution of intraspecific and interspecific collinear blocks G. rigescens de novo genome assembly and evolutionary analysis. A Photo of G. rigescens plant. B Genome features and synteny information. I, chromosome size; II, repeat content; III, gene number; IV, GC content; V, gene synteny. C Venn diagram showing orthogroups from six Gentianales species. O. pumila, C. roseus, G. sempervirens, G. rigescens, G. dahurica, and G. macrophylla. D Phylogenetic tree showing divergence times and the evolution of gene family size. E Ks distribution of intraspecific and interspecific collinear blocks Results De novo G. rigescens genome assembly and annotation We processed ~ 190-Gb Oxford Nanopore long sequencing reads (Additional file 1: Table S1) with NextDenovo to attain an initial G. rigescens genome assembly of ~ 3.32 Gb. We then corrected errors in the genome assembly with ~ 247-Gb Illumina short paired-end reads (Additional file 1: Table S1) and removed redundancy to obtain a contig assembly of 2.07 Gb (970 contigs, contig N50 = 3.23 Mb, Additional file 1: Table S2). The contig assembly size was close to the genome size estimated by k-mer frequency distribution analysis (~ 2.18 Gb) and slightly larger than that obtained by flow cytometry (1.86 Gb, Additional file 2: Fig. S1A and B). K-mer analysis and SmudgePlot graph reveal a diploid genome, with a heterozygosity of 2.66% and a repeat content of ~ 76%, highlighting the complexity of the G. rigescens genome (Additional file 2: Fig. S1B and C). Then, we used ~ 307-Gb chromosome conformation capture (Hi-C) sequencing data (Additional file 1: Table S1) to obtain chromosome-scale scaffolds. We managed to anchor ~ 1.92-Gb (93.10%) contig sequences into 13 pseudo-chromosomes (GC content = 38.7%; Fig. 1B, Additional file 2: Fig. S1D) with sizes ranging from 79.06 to 190.61 Mb (Additional file 1: Table S3). The complete Benchmarking Universal Single-Copy Orthologs (BUSCO) in the G. rigescens genome reached 91.00% (Additional file 1: Table S4), suggesting a relatively high-quality genome assembly. The long terminal repeat assembly index (LAI) [18] was 18.70, indicating that the G. rigescens genome reached reference-level quality. Approximately 85.66% of the assembled genome (Fig. 1B, Additional file 1: Table S5) comprised various types of repetitive elements, among which about 75.06% were long terminal repeats (LTRs; Additional file 1: Table S6). The proportion of repetitive elements was higher than G. dahurica (73.47%) and G. macrophylla (70.25%) [10, 11], which may contribute to its relatively larger genome size. We estimated insertion times in three closely related species and found that LTR retrotransposon (LTR-RT) proliferation peaked in the early Pleistocene. The burst of LTR-RT in G. rigescens (1.12 million years ago, Mya; 95% CI 0.17–2.51 Mya) occurred earlier than in G. dahurica (0.57 Mya; 95% CI 0.08–1.58 Mya) and G. macrophylla (0.86 Mya; 95% CI 0.17–2.02 Mya) (Additional file 2: Fig. S2). Besides the 2835 (1210 tRNAs, 1019 rRNAs, 132 miRNAs, and 474 snRNAs) nonprotein-coding genes (Additional file 1: Table S7), we predicted 55,697 protein-coding genes in the G. rigescens genome, with an average coding sequence length of 1192 bp (Additional file 1: Table S8). We also showed an inverse correlation between gene density along all pseudomolecules and the density of repetitive elements (Fig. 1B). The completeness of gene prediction showed 87.79% coverage of 1614 BUSCO genes (Additional file 1: Table S4), demonstrating the reliability of the predicted results. About 75.77% of predicted protein-coding genes have functional annotations in the public databases (Additional file 1: Table S9). Genome evolution analysis A total of 48,536 orthologous genes were identified across G. rigescens and 7 other plant species (5 Gentianales: G. dahurica, G. macrophylla, Gelsemium sempervirens, Catharanthus roseus, Ophiorrhiza pumila; 1 Lamiales: Olea europaea, 1 Asterales: Helianthus annuus), with 49,911 genes belonging to 19,057 gene families in the G. rigescens genome (Additional file 1: Table S10). A total of 4976 gene families were shared among six Gentianales species (Fig. 1C, Additional file 2: Fig. S3). It is worth noting that the number of unique gene families is much larger than those found in other Gentiana plants (Additional file 1: Table S10). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment results indicated that the G. rigescens unique gene families were mainly enriched in metabolism, secondary metabolites, environmental adaptation, and genetic information processing (Additional file 2: Fig. S4, Additional file 1: Table S11). These results are consistent with the adaptive evolution of G. rigescens in high-altitude environments. The maximum likelihood (ML) phylogenetic tree of G. rigescens and the other 7 species using 102 single-copy orthologous genes highly supports all Gentianales species forming a distinct evolutionary lineage, which is a sister group to Lamiales (Fig. 1D). Gentianaceae diverged approximately 54.35 (45.67–64.12) Mya from Gelsemiaceae. G. rigescens clusters with its congeneric G. dahurica and G. macrophylla, sharing a most recent common ancestor about 17.29 (8.88–29.42) Mya. The ancestral lineage of Gentiana exhibits the most pronounced gene family expansion (816) among all internal nodes. We also identified 526 expanded gene families and 359 contracted gene families within the G. rigescens genome (Fig. 1D). Expanded gene families show significant functional enrichment in translation, regulation of translation, and chromosome organization processes (Additional file 2: Figs. S5 and S6, Additional file 1: Tables S12 and S13). Genome synteny and WGD We performed MCScanX analysis between G. rigescens and three species (G. macrophylla, G. sempervirens, and Vitis vinifera) to investigate potential whole‑genome duplication (WGD) events during the evolutionary of G. rigescens. Dotplot of syntenic blocks between G. rigescens and V. vinifera showed a 3:1 or 4:1 orthology ratio (Additional file 2: Fig. S7) and a 1:1 relationship between G. rigescens and G. macrophylla (Additional file 2: Fig. S8). This results indicates two WGD events occurred in G. rigescens after the WGTγ event. To further validate and estimate the timing of the WGDs in G. rigescens, we analyzed the synonymous substitution rate (Ks) distributions of collinear gene pairs and intra-/interspecific syntenic relations. Gaussian mixture modeling of the Ks distributions in G. rigescens and G. macrophylla revealed three peaks at Ks values of approximately 0.04, 0.76, and 1.16 (Fig. 1E). Accordingly, the abundance of 4DTv sites distributions also exhibit three peaks (Additional file 2: Fig. S9). These results suggest that in addition to the ancient angiosperm WGT-γ event (Ks≈1.16), G. rigescens and G. macrophylla each experienced two subsequent WGD events (WGD-g1 at Ks≈0.04 and WGD-g2 at Ks≈0.76). To further test whether these duplication events are shared across the genus, we compared the Ks distribution of G. rigescens with that of G. sempervirens. G. sempervirens displayed evidence for only one WGD event, consistent with previous studies reporting no recent WGD after WGT-γ [19]. Taken together, these results indicate that both WGD-g1 and WGD-g2 likely occurred after the divergence of the Gentiana common ancestor from G. sempervirens and therefore represent Gentiana-lineage-shared WGD events. The inference accords with published results for G. macrophylla [11] and G. straminea [12]. G. rigescens as a model for temperature-regulated flower opening and closing Previous studies show that the Gentiana flower’s repetitive opening and closing responds to both photoperiod and temperature cues, making it challenging to tease apart the players in each signaling pathway [15, 20]. We observed the repetitive opening and closing of wild G. rigescens flowers at four different time points during the day and found that flowers stayed closed at 6:00 a.m. (ambient temperature 8.50 ± 0.13 °C) and started to open at 10:00 a.m. (ambient temperature 15.60 ± 0.69 °C; Fig. 2A and B, Additional file 1: Table S14). The flowers were fully open at noon (ambient temperature 22.38 ± 0.78 °C) and closed at 6:00 p.m. as temperature dropped below 10.05 ± 0.35 °C. This result suggests that ambient temperature might be an important environmental factor in the repetitive opening and closing of flowers.Fig. 2Temperature-regulated opening and closing of G. rigescens flower. A Corresponding G. rigescens blooming rate under different temperature and light regimes in the lab (n = 30, 3 replicates). B Corresponding ambient temperatures at 6:00, 10:00, 12:00, and 18:00 in the wild (n = 45). C Photos of G. rigescens flower at 6:00, 10:00, 12:00, and 18:00 in the wild. D Photos of G. rigescens flowers under different temperature and light regimes in the lab. E Time required for the opening of G. rigescens flower at different temperatures (n = 5, three replicates). F Time required for the closing of G. rigescens flower at different temperatures (n = 5, three replicates). G Scanning electron microscopy of the petal epidermal cells. Data in the graph are mean ± standard deviation. Statistical significance was determined using one-way ANOVA followed by Tukey’s multiple comparison test. Different letters indicate significant differences between treatments (P ≤ 0.05) Temperature-regulated opening and closing of G. rigescens flower. A Corresponding G. rigescens blooming rate under different temperature and light regimes in the lab (n = 30, 3 replicates). B Corresponding ambient temperatures at 6:00, 10:00, 12:00, and 18:00 in the wild (n = 45). C Photos of G. rigescens flower at 6:00, 10:00, 12:00, and 18:00 in the wild. D Photos of G. rigescens flowers under different temperature and light regimes in the lab. E Time required for the opening of G. rigescens flower at different temperatures (n = 5, three replicates). F Time required for the closing of G. rigescens flower at different temperatures (n = 5, three replicates). G Scanning electron microscopy of the petal epidermal cells. Data in the graph are mean ± standard deviation. Statistical significance was determined using one-way ANOVA followed by Tukey’s multiple comparison test. Different letters indicate significant differences between treatments (P ≤ 0.05) To investigate the primary environmental factors regulating flower repetitive opening and closing in G. rigescens, fully open and closed flowers (within 3 days after blooming) were incubated under constant light or constant dark (10 °C or 22 °C), respectively. Closed flowers at 22 °C remained fully open throughout the day, whereas open flowers at 10 °C stayed closed, regardless of light conditions (Fig. 2C, Additional file 2: Fig. S10). These results indicate that repetitive flower opening and closing of G. rigescens are primarily driven by temperature rather than photoperiod or circadian cues. Consistently, budding-stage plants (n = 30 per group) showed comparable flowering rates under constant light and dark at 22 °C, both significantly higher than the control, whereas flowering was strongly suppressed at 10 °C (Fig. 2D, Additional file 1: Table S15). These results demonstrate that temperature governs both flowering propensity and repetitive petal movements, highlighting G. rigescens as a robust model for investigating temperature-dependent floral behaviors. Further observations of G. rigescens flower opening and closing in a temperature gradient (15–28 °C for flower opening and 5–14 °C for flower closing) showed that the speed of flower opening and closing was strongly associated with the change in temperatures (Fig. 2E and F, Additional file 1: Table S16 and 17). The times for flowers to fully open or close were the shortest at 28 °C (21.70 ± 0.99 min.) and 5 °C (14.20 ± 0.40 min), respectively. Then, we manually removed various organs from individual plants to test the impact on flowering. The result revealed that even lacking other floral organs, petals retained the capacity for repetitive opening and closing (Additional file 2: Fig. S11). Petal movements of excised flowers in water incubation also confirmed this observation (Additional file 2: Fig. S11). In addition, scanning electron microscopy showed that the corolla epidermal cells on the abaxial side became turgid at 22 °C (Fig. 2G), suggesting that cellular osmotic change may be the causative mechanism. Differentially expressed genes in G. rigescens flowers at different temperatures We selected three times (6:00 a.m., AM6; 12:00 p.m., PM12; 6:00 p.m., PM6) for transcriptomic profiling of G. rigescens flowers to find candidate genes and related whole regulated networks underlying temperature-regulated flower repetitive opening and closing. We generated ~ 64.56-Gb clean RNA sequencing data from 9 samples (3 samples per time) to calculate the fragments per kilobase of transcript per million mapped fragments (FPKM) of 55,679 protein-coding genes. We argue that genuine gene candidates should meet two criteria: (1) Differential gene expression patterns between open and closed blooms (AM6-PM12 and PM12-PM6), and (2) similar gene expression levels between closed blooms (AM6 and PM6). We identified 2203 and 1702 differentially expressed genes (DEGs) in the AM6-PM12 and PM12-PM6 groups, respectively (Fig. 3A and B), which contained 1070 shared DEGs. They intersected with non-differentially expressed genes in the AM6-PM6 group for the final 994 DEGs (442 up-regulated, 551 down-regulated; Fig. 3C; Additional file 2: Fig. S12; Additional file 1: Table S18). About 78.10% of these DEGs have functional annotations in the TAIR database. Gene Ontology (GO) enrichment analyses found 183 significantly enriched terms (P ≤ 0.05), including processes related to pollen tube development and growth, cell wall and cell membrane functions, and the regulation of pectinesterase activity (Fig. 3D, Additional file 1: Table S19). Similarly, the KEGG enrichment analyses revealed significant terms in the categories of carbohydrate metabolism, various secondary metabolite biosynthesis, and plant hormone signal transduction (Additional file 2: Fig. S13, Additional file 1: Table S20). These results correspond to the abundance of DEGs with annotations highlighting putative pectinesterases, proton-dependent oligopeptide transporters, glycoside hydrolases, various transferases, and oxidases.Fig. 3Differentially expressed genes in G. rigescens flowers at different times. A Volcano plot of DEGs between 6:00 am (AM6) and 12:00 am (PM12). B Volcano plot of DEGs between PM12 and 6:00 pm (PM6). C Venn diagram of DEGs between open and closed blooms (AM6 vs. PM12 and PM12 vs. PM6) while exhibiting no differential expression between the two closed‑bloom conditions (AM6 vs. PM6). D GO function enrichment analysis of 994 DEGs. E Time-series expression analysis of 994 DEGs. F GO enrichment of Pattern 1-Pattern 3 DEGs (Top 10). G GO enrichment of Pattern 4 DEGs (Top 10) Differentially expressed genes in G. rigescens flowers at different times. A Volcano plot of DEGs between 6:00 am (AM6) and 12:00 am (PM12). B Volcano plot of DEGs between PM12 and 6:00 pm (PM6). C Venn diagram of DEGs between open and closed blooms (AM6 vs. PM12 and PM12 vs. PM6) while exhibiting no differential expression between the two closed‑bloom conditions (AM6 vs. PM6). D GO function enrichment analysis of 994 DEGs. E Time-series expression analysis of 994 DEGs. F GO enrichment of Pattern 1-Pattern 3 DEGs (Top 10). G GO enrichment of Pattern 4 DEGs (Top 10) Time-series expression based on a loose clustering algorithm using Mfuzz [21] identified 4 expression patterns among 994 DEGs (Fig. 3E, Additional file 1: Table S21). Notably, patterns 1–3 showed higher gene expression levels during flower closure (6:00 a.m. and 6:00 p.m.) than those during flower opening (12:00 a.m.). In contrast, Pattern 4 included the largest anthesis-associated DEGs (443 genes) and showed gene expression upregulation during flower opening (12:00 a.m.). Functional enrichment analysis of DEGs in Patterns 1–3 revealed pathways related to the regulation of hormone levels, cell periphery, pollen tube development and growth, plasma membrane and region, pollination, cell wall, and cellular response (Fig. 3F, Additional file 1: Table S22, 23, 24). In comparison, anthesis-associated DEGs in Pattern 4 are enriched in the categories of the regulation of catalytic activity, pectinesterase and inhibitor activity, oxidoreductase activity, regulation of molecular function, response to hormone, cell wall, external encapsulating structure, and enzyme inhibitor activity (Fig. 3G, Additional file 1: Table S25). These data collectively identified key nodes in the molecular mechanisms underlying temperature-dependent flower opening and closing, including putative upstream hormonal signaling pathways and downstream cell wall metabolism. In addition, the DEGs related to pollen tube growth and development supported the idea that repetitive flower opening and closing is in concert with plant reproduction. Putative temperature sensors in G. rigescens flowers The perception of ambient temperature change presumably relies on master temperature sensors in the cell membrane [22]. Early work proposed calcium transporter annexin 1 (ANN1) and the cyclic nucleotide-gated calcium channels (CNGC), the TOLERANCE DIVERGENCE 1 (COLD1) protein in rice, and several receptor-like protein kinases (RLKs) as candidates [23–27]. The current working mechanism asserts that the activity of membrane-localized channels corresponds to membrane fluidity and permeability in eliciting rapid and transient fluctuations in cytosolic Ca2+ concentration [22, 28]. These fluctuations convert external environmental signals into various intracellular biochemical responses. We speculated that plasma membrane-localized Ca2+-permeable channels might function as temperature sensors in the petal movement of G. rigescens flowers. Among the DEGs in G. rigescens flowers are several Ca2+ transport-related genes belonging to putative ANN, CNGC, and autoinhibited Ca2+-ATPase (ACA) gene family (Additional file 1: Table S18). We used Arabidopsis thaliana as a reference for the whole-genome identification and phylogenetic analysis of the ANN gene family in four Gentianaceae species (G. rigescens, G. macrophylla, G. dahurica, E. grandiflorum). A total of 61 ANN genes were identified and classified into four groups, with one cluster in group IV showing significant expansion (Fig. 4A). Further investigation showed that the duplication patterns of the ANN gene family vary Across Gentianaceae species. In G. rigescens, G. dahurica, and G. macrophylla, the expansion of ANN genes is predominantly driven by dispersed duplication (DSD), proximal duplication (PD), and transposed duplication (TRD), respectively (Fig. 4A and C). The prediction and analysis of cis-acting elements in the ANN gene promoter regions revealed that all members contain many light and hormone response-related cis-acting elements (Additional file 2: Fig. S14). Specifically, low-temperature response elements were observed in some gene promoter regions (Fig. 4D, Additional file 2: Fig. S14). The two ANN family differentially expressed genes (Grig09G002125.t1 and Grig03G002764.t1) are homologs of Arabidopsis ANN1 (AT1G35720) and ANN8 (AT5G12380), respectively (Fig. 4A). They exhibit high expression levels during G. rigescens flowering at noon (Fig. 4B). We used RT-qPCR to verify the transcriptomic data of the two genes (Fig. 4E). Previous research shows that Arabidopsis ANN1 not only regulates cytosolic Ca2+ elevation in seedlings in response to heat stress but also mediates plant cold tolerance [23, 24]. ANN8 primarily regulates H2O2 and Ca2+ signal transduction, which is critical in immune response and stress signal transduction pathways in plants [29].Fig. 4Identification of putative temperature sensors in G. rigescens flowers. A The phylogenetic tree of ANN genes from A. thaliana, Eustoma grandiflorum, G. rigescens, G. dahurica, and G. macrophylla. B The expression levels of putative temperature sensors in G. rigescens flowers. C Estimated the number of ANN genes from different gene duplication types in the genomes of A. thaliana, G. rigescens, and G. dahurica. Colored circles are sized to represent the number of all genes in each gene duplication type. WGD, whole genome duplication; TD, tandem duplicate; TRD, transposed duplication; PD, proximal duplication; DSD, dispersed duplication. D Distribution and number of cis-acting elements in the promoter regions of ANN and CNGC DEGs. Red color scheme represents pytohormone response related, yellow color scheme represents light response related, and blue color scheme represents low temperature response related. E RT-qPCR analysis of ANN family genes in DEGs. F The expression levels of putative calcium signaling sensors and effectors in G. rigescens flowers. H The expression levels of G. rigescens flowers aquaporin DEGs. Data in the graph are mean ± standard deviation (n = 3). ∗ Indicates P ≤ 0.05 Identification of putative temperature sensors in G. rigescens flowers. A The phylogenetic tree of ANN genes from A. thaliana, Eustoma grandiflorum, G. rigescens, G. dahurica, and G. macrophylla. B The expression levels of putative temperature sensors in G. rigescens flowers. C Estimated the number of ANN genes from different gene duplication types in the genomes of A. thaliana, G. rigescens, and G. dahurica. Colored circles are sized to represent the number of all genes in each gene duplication type. WGD, whole genome duplication; TD, tandem duplicate; TRD, transposed duplication; PD, proximal duplication; DSD, dispersed duplication. D Distribution and number of cis-acting elements in the promoter regions of ANN and CNGC DEGs. Red color scheme represents pytohormone response related, yellow color scheme represents light response related, and blue color scheme represents low temperature response related. E RT-qPCR analysis of ANN family genes in DEGs. F The expression levels of putative calcium signaling sensors and effectors in G. rigescens flowers. H The expression levels of G. rigescens flowers aquaporin DEGs. Data in the graph are mean ± standard deviation (n = 3). ∗ Indicates P ≤ 0.05 The two CNGC genes (Grig09G002231.t1 and Grig11G001911.t1) share homology protein with Arabidopsis CGNC6 (AT2G23980) and CGNC18 (AT5G14870), respectively (Additional file 2: Fig. S15). The promoter regions of the CNGC gene family contain cis-acting elements similar in type to those of ANN (Fig. 4D, Additional file 2: Fig. S16). Arabidopsis CNGC6 is a putative thermal sensor for the heat stress response, which increases Ca2+ influx and heat shock protein expression at higher temperatures [30]. Activated CNGC18 in response to cytosolic cyclic nucleotides (cAMP) is present at the tip of pollen tubes to create a tip-focused Ca2+ gradient and regulate pollen tube growth and orientation [31, 32]. In contrast, the expression levels of these two G. rigescens CNGC genes are three times higher in closed than open flowers (Fig. 4B), which suggests potential involvement in cold-related responses. In addition, we found two ACA genes with Grig05G001543.t1 and Grig05G001544.t1 homologous to Arabidopsis ACA9 (AT3G21180) and Grig10G003065.t1 to ACA8 (AT5G57110). As Ca2+-ATPases balance Ca2+ homeostasis by controlling the efflux of Ca2+ from the cytoplasm into other cell compartments, it is unsurprising that ACA8 facilitates cold adaptation by regulating cytoplasmic Ca2+ concentrations [33], and ACA9 contributes to pollen tube burst [34]. In G. rigescens, all three ACA genes exhibit expression upregulation during flower closure (Fig. 4B, Additional file 2: Fig. S14). Taken together, these results identify ANN, CNGC, and ACA genes as strong candidates for temperature-responsive regulators in G. rigescens flower repetitive opening and closing. While functional validation is needed, their expression patterns and evolutionary features suggest potential contributions to the temperature-dependent petal movement. Downstream calcium signaling sensors and effectors Cellular calcium-sensing receptors and effectors link cytosol Ca2+ levels to various downstream target gene expressions, which subsequently impact ion transport, metabolic homeostasis, immunity, and other physiological processes [35, 36]. Major plant Ca2+ receptors include calmodulin (CaM), CaM-like proteins (CMLs), calcineurin B-like proteins (CBLs), CBL-interacting protein kinases (CIPKs), and calcium-dependent protein kinases (CDPKs) [37]. In G. rigescens flowers, we found one CML, one CIPK, and two CDPKs in the DEG list that shared high sequence homology with their Arabidopsis counterparts (Fig. 4F, Additional file 1: Table S18). The G. rigescens CML3 (Grig06G003805.t1), CIPK7 (Grig08G002687.t1), and CDPK6/CPK3 (Grig03G001683.t1) show high expression levels during flower closure under low ambient temperatures (Fig. 4F). Functional annotation indicates that CML3 acts as a potential Ca2+ sensor. Arabidopsis CIPK7 interacts with CBL1 in responding to cold stress under unknown mechanisms [38]. CPK3 promotes abscisic acid-induced stomatal closure by modulating ion channels in guard cells [39]. These patterns suggest that CML3, CIPK7, and CPK3 may participate in flower closure at low temperatures in G. rigescens. By contrast, Grig06G000821.t1, a homolog of Arabidopsis CDPK16/CPK16 (AT2G17890), exhibited approximately 4–6 times higher expression levels during flower opening at noon (Fig. 4F), implying potential functional role at higher ambient temperatures. Previous investigations in G. scabra show that CPK16 phosphorylates the aquaporin protein PIP2;2 in a calcium-dependent manner under temperature and light stimulation, which activates water transport to regulate flower reopening [20]. Given the genetic similarity, a similar physiological pathway might exist in G. rigescens, which underlies the change in cell turgidity and shape (Fig. 2G). Notably, the expression of G. rigescens aquaporin genes (PIP2;1, TIP4;1, and TIP5;1) was lower during flower opening than during flower closure at cold ambient temperature (Fig. 4H). This result may reflect additional mechanisms underlying cellular osmoregulation, water and ion homeostasis, nutrient transport, and cell membrane stability at low temperatures [40]. Together, these expression patterns indicate that distinct sets of Ca2⁺ sensors and effectors are differentially engaged during flower repetitive opening and closure in G. rigescens, with CML3, CIPK7, and CPK3 associated with closure at low temperatures and CPK16 potentially contributing to reopening under warmer conditions. G. rigescens flower repetitive opening and closing requires dynamic cell wall remodeling Plant cells’ morphogenesis and growth rely on the cell wall, a rigid matrix composed of cellulose, hemicellulose, and pectin [41, 42]. We hypothesize that the repetitive opening and closing of G. rigescens flowers requires concerted changes in cell volume, turgor pressure, and cell wall rigidity [41]. Supporting this, DEGs were significantly enriched in pectinesterase activity, cell wall organization, and carboxylic ester hydrolase activity, featuring abundant genes associated with hemicellulose metabolism, pectin metabolism, expansins, and cellulose synthesis (Fig. 3D, Additional file 1: Table S18). These findings suggest that cell wall remodeling acts downstream of calcium signaling during floral movements. Cellulose microfibrils confer cell wall stiffness, and its synthesis at the cell membrane relies on the CELLULOSE SYNTHASE A (CESA) enzymes [43]. A DEG identified in this study is G. rigescens CESA6 (Grig13G003238.t1), which shows a lower gene expression level at noon during flower opening (Additional file 2: Fig. S17). Similarly, a DEG encoding COMPANION OF CELLULOSE SYNTHASE 1 (CC1; Grig07G001855.t1), which stabilizes the microtubule–CESA machinery under stress [44], also showed lower expression during opening. In the cell wall architecture, xyloglucans tether adjacent cellulose together to form a structural network [41]. Xyloglucan-modifying enzymes, such as xyloglucan endotransglycosylase/hydrolases (XTH) and glycoside hydrolases (GH), reduce cell wall stiffness by breaking down these hemicellulose polymers [41]. We identified four XTHs and 32 GHs among the DEGs, and the gene expression patterns are divergent (Fig. 5A). While most genes show higher expressions at cold temperatures during flower closure, one XTH and six GHs manifest the opposite expression pattern. The expansins (EXP) proteins break the hydrogen bond between cellulose and hemicellulose to loosen cell walls and promote wall synthesis [45, 46]; they were also differentially expressed, with nine DEGs exhibiting patterns similar to CESA and CC1.Fig. 5The effects of dynamic cell wall modification in G. rigescens petal movement. A The heat maps and RT-qPCR analysis of PME, PMEI, PL, EXP, XTH, and GH family DEGs. B Schematic of external spray treatment. The whole inflorescence was sprayed with 10-μm mixed pectinase inhibitors and 10-μm DMSO as the control group. After 12 h of treatment, the flowers were cultured at 22 °C and 10 °C, respectively, and the opening and closing of flowers were observed. C The time required for flower opening and closure upon treatment with pectinase inhibitors (n = 5, three replicates). Statistical significance between two groups was determined using a two-tailed Student’s t-test. D RT-qPCR analysis of PME family genes upon treatment with pectinase inhibitors (n = 3). Data in the graph are mean ± standard deviation, ∗ Indicates P ≤ 0.05 The effects of dynamic cell wall modification in G. rigescens petal movement. A The heat maps and RT-qPCR analysis of PME, PMEI, PL, EXP, XTH, and GH family DEGs. B Schematic of external spray treatment. The whole inflorescence was sprayed with 10-μm mixed pectinase inhibitors and 10-μm DMSO as the control group. After 12 h of treatment, the flowers were cultured at 22 °C and 10 °C, respectively, and the opening and closing of flowers were observed. C The time required for flower opening and closure upon treatment with pectinase inhibitors (n = 5, three replicates). Statistical significance between two groups was determined using a two-tailed Student’s t-test. D RT-qPCR analysis of PME family genes upon treatment with pectinase inhibitors (n = 3). Data in the graph are mean ± standard deviation, ∗ Indicates P ≤ 0.05 Pectin network accounts for almost half of the cell wall dry mass [47]. Its synthesis and integrity are under the regulation of many wall proteins, including pectin methylesterases (PME), pectin methylesterase inhibitors (PMEI), and pectin lyases (PL) [47]. PME removes methyl groups from pectin to create negatively charged “egg-box junctions,” where calcium ionically cross-links adjacent pectin chains for increased stability [48]. Intuitively, PMEI counters PME’s action in the cell wall. The degradation of pectin polymers depends on PL. We identified among the DEGs 25 PMEs, 10 PMEIs, and 5 PLs in the plant (Fig. 5A). Similar to XTHs and GHs, the expression heat map reveals the divergent patterns of PMEs and PMEIs. By contrast, all PL genes show upregulation during flower closure in cold temperatures (Fig. 5A). We verified the transcriptomic data of select genes using the RT-qPCR method (Fig. 5A). To further validate the requirement of cell wall remodeling in G. rigescens’ flower repetitive opening and closing, we leveraged commercial pectinase inhibitors to observe the effect on G. rigescens’ flower opening and closing (Fig. 5B). We found that pectinase inhibitor treatment doubled the times for the petals to open and close fully by significantly reducing some PME gene expressions (Fig. 5C and D, Additional file 1: Table S26). Auxin promotes G. rigescens flower opening and inhibits flower closing Auxin-induced cell wall acidification is essential in plant growth and development, particularly pivotal in cell elongation and morphogenesis [49, 50]. To assess the involvement of auxin in the repetitive petal movements of G. rigescens, we examined auxin signaling genes. We found many DEGs in the Aux/IAAs, AUXIN RESPONSE FACTORS (ARFs), and SMALL AUXIN UP-REGULATED RNAs (SAURs) family with high expression levels in closed flowers at low temperatures (Fig. 6A). RT-qPCR validation of select genes showed the same expression pattern (Fig. 6A).Fig. 6The effects of auxin on petal movement of G. rigescens. A The heatmaps and RT-qPCR analysis of AUX, ARF, PIN, and SAUR family DEGs. B Schematic of external spray treatment. The whole inflorescence was sprayed with 10-μm inhibitors of polar transport of growth hormones (TIBA) and 10-μm DMSO as the control group. After 12 h of treatment, the flowers were cultured at 22 °C and 10 °C, respectively, and the opening and closing of flowers were observed. C The time required for flower opening and closure upon treatment with TIBA (n = 5, three replicates). Statistical significance between two groups was determined using a two-tailed Student’s t-test. D RT-qPCR analysis of AUX, PME, EXP, and XTH family genes upon treatment with TIBA. Data in the graph are mean ± standard deviation. ∗ Indicates P ≤ 0.05 The effects of auxin on petal movement of G. rigescens. A The heatmaps and RT-qPCR analysis of AUX, ARF, PIN, and SAUR family DEGs. B Schematic of external spray treatment. The whole inflorescence was sprayed with 10-μm inhibitors of polar transport of growth hormones (TIBA) and 10-μm DMSO as the control group. After 12 h of treatment, the flowers were cultured at 22 °C and 10 °C, respectively, and the opening and closing of flowers were observed. C The time required for flower opening and closure upon treatment with TIBA (n = 5, three replicates). Statistical significance between two groups was determined using a two-tailed Student’s t-test. D RT-qPCR analysis of AUX, PME, EXP, and XTH family genes upon treatment with TIBA. Data in the graph are mean ± standard deviation. ∗ Indicates P ≤ 0.05 To further validate the effect of auxin on flower repetitive opening and closing in G. rigescens, we sprayed the flowers with auxin transport inhibitor 2,3,5-triiodobenzoic acid (TIBA). TIBA treatment delayed the time to full bloom by 100% and quickened the time to full closure by 50% (Fig. 6B and C, Additional file 1: Table S27). Previous work showed that auxin may directly or indirectly regulate the activity of AUTOINHIBITED H + ATPASEs (AHAs) [51], which transport protons into the outer cell wall [50]. The acidification of the cell wall activates enzymes, such as EXP, XTH, and PME, to promote cell wall relaxation. Two G. rigescens AHA8 genes (Grig05G004175.t1 and Grig13G002358.t1) exhibited high expression during flower closure (Additional file 1: Table S18). In concordance with this observation, TIBA treatment significantly decreased the expression of AUX, EXP, XTH, and PME genes (Fig. 6D). These results indicate that auxin promotes flower opening and inhibits flower closing by modulating the gene expressions in the cell wall remodeling process. Discussion Repetitive opening and closing of flowers is a critical evolutionary adaptation to environmental fluctuations, conferring reproductive fitness in alpine plants [15]. In Gentiana, petal movement has been documented across diverse species, triggered by cues such as temperature, light, thunderstorms, and mechanical stimuli [1, 5, 16, 20]. In G. rigescens, our findings demonstrate that temperature acts as the dominant environmental regulator of floral movements, in contrast to prior examples where both light and temperature jointly affect the movement. We assembled a chromosome-scale genome of G. rigescens, which provided a high-quality foundation for investigating Gentiana’s evolutionary history. Our analyses indicate that G. rigescens experienced two WGD events at approximately 2.9 Mya and 54.36 Mya, broadly consistent with findings in other Gentiana species [10–12]. Notably, the inferred time of WGD-g2 in our study is slightly later than in some previous reports, which may reflect the fact that this duplication occurred around the same time as lineage divergence. Consequently, the signals of genome duplication and speciation are partially overlapping, making it difficult to completely disentangle the two [52]. Although the precise timing warrants further investigation, convergent evidence supports the occurrence of two lineage-specific WGDs in Gentiana. Importantly, these duplication events likely facilitated the expansion of gene families associated with secondary metabolism and stress responses, thereby providing genetic innovations for the adaptive evolution of Gentiana species in alpine environments. By assembling a chromosome-level genome for G. rigescens and integrating it with transcriptome profiling and experimental observations, we inferred a regulatory framework underlying temperature-induced flower repetitive opening and closure. To focus on genes primarily responsive to temperature, we first selected those showing no differential expression between two closed-flower samples (AM6 vs. PM6), thereby minimizing the influence of light or circadian regulation. This set was then intersected with DEGs between closed-flower (AM6 or PM6) and open-flower (PM12) samples, yielding a subset that predominantly reflects temperature-dependent regulation. This framework involves three interconnected modules: (1) The sensing of ambient temperature with membrane calcium transporters and the activation of calcium signaling transduction, (2) dynamic cell‑wall remodeling, and (3) auxin‑mediated cell‑wall acidification and enzyme activation (Fig. 7). Given the identification of DEGs with divergent expression patterns and the different roles of auxin in regulating petal movement, our results suggest that distinct signaling pathways underlie the low-temperature-induced petal closure and high-temperature induced petal opening. Future studies should employ controlled-environment experiments with precisely manipulated temperature gradients to further validate the proposed regulatory network while excluding potential confounding factors such as humidity fluctuations or mechanical stimuli.Fig. 7Putative regulation network of G. rigescens-dependent-temperature reversible flower opening and closing. A Low temperatures induce the influx of Ca2+, activating CNGC calcium channels, which leads to changes in cytoplasmic Ca2+ concentration. Cytoplasmic Ca2+ binds to CDPKs, CIPKs, and CBLs, activating relevant transcription factors. This results in the upregulation of genes associated with cell wall modification or hormone responses. Auxin-responsive genes may also be activated by Ca2+ or calcium-binding proteins, such as the upregulation of SAURs, which inhibits the dephosphorylation of AHAs by PP2Cs, maintaining AHAs in an active phosphorylated state and leading to an acidic cell wall environment. Cell wall acidification activates cell wall modification-related enzymes, including EXP, XTH, PME, PL, and PG, modulating the cell wall flexibility and loosening. PME increased significantly upon the reduction of temperature, which caused a considerable amount of HG to undergo demethylation and subsequently form calcium bridges with free Ca2+ in the cell wall. Concurrently, changes in cytoplasmic Ca2+ levels regulate various ion transporters such as ACAs, AHAs, and CHX to maintain intracellular ion homeostasis. B High temperature activates ANNs and triggers Ca.2+ influx, which binds to CPK16, causing a conformational change. This leads to the phosphorylation of PIPs and the activation of water transport. Water transport provides the turgor pressure required for petal cell expansion, leading to flower opening. Activated CPK16 may also activate transcription factors, leading to up-regulation of genes related to cell wall modification or hormone response. This figure was created using BioRender (https://biorender.com/) Putative regulation network of G. rigescens-dependent-temperature reversible flower opening and closing. A Low temperatures induce the influx of Ca2+, activating CNGC calcium channels, which leads to changes in cytoplasmic Ca2+ concentration. Cytoplasmic Ca2+ binds to CDPKs, CIPKs, and CBLs, activating relevant transcription factors. This results in the upregulation of genes associated with cell wall modification or hormone responses. Auxin-responsive genes may also be activated by Ca2+ or calcium-binding proteins, such as the upregulation of SAURs, which inhibits the dephosphorylation of AHAs by PP2Cs, maintaining AHAs in an active phosphorylated state and leading to an acidic cell wall environment. Cell wall acidification activates cell wall modification-related enzymes, including EXP, XTH, PME, PL, and PG, modulating the cell wall flexibility and loosening. PME increased significantly upon the reduction of temperature, which caused a considerable amount of HG to undergo demethylation and subsequently form calcium bridges with free Ca2+ in the cell wall. Concurrently, changes in cytoplasmic Ca2+ levels regulate various ion transporters such as ACAs, AHAs, and CHX to maintain intracellular ion homeostasis. B High temperature activates ANNs and triggers Ca.2+ influx, which binds to CPK16, causing a conformational change. This leads to the phosphorylation of PIPs and the activation of water transport. Water transport provides the turgor pressure required for petal cell expansion, leading to flower opening. Activated CPK16 may also activate transcription factors, leading to up-regulation of genes related to cell wall modification or hormone response. This figure was created using BioRender (https://biorender.com/) As a messenger molecule, Ca2+ converts environmental stimuli such as cold, heat, salt stress, and pathogen infection into physiological signals within plant cells [24, 53]. In line with this model, we found calcium transporters and downstream calcium signaling effectors in the list of DEGs (Fig. 4B, F). However, how G. rigescens differentiates cold and heat-induced intracellular Ca2+ signaling is unknown. A recent study using optogenetically modified ion channels suggests that discrete ion fluxes lead to distinct physiological responses [54]. In addition, calcium-binding CDPK and CIPK can activate the calcium pump ACAs, forming a feedback loop to maintain intracellular calcium homeostasis [55, 56]. The activated calcium-binding proteins may directly or indirectly regulate the activity of other ion transport proteins to maintain the stability of the intracellular environment [35]. Therefore, it is interesting to investigate whether additional ion fluxes are involved in the temperature-dependent flower repetitive opening and closing in G. rigescens in the future. Even though cellular osmotic changes and differential cell growth are vital mechanisms in controlling plant organ movement, we show that dynamic cell wall remodeling is crucial to the temperature-regulated flower opening and closure in G. rigescens (Fig. 5). Cell wall relaxation and expansion correspond to water dynamics, which rely on a myriad of cell wall-modifying enzymes. It is possible that the calcium signaling effectors, such as CDPKs and CIPKs, activate the expression and function of cell wall remodeling enzymes via transcription factors such as MYB, NAC, and WRKY, but the details remain unknown. Investigation into the expression regulation of each enzyme will help understand the mechanisms by which these enzymes manifest divergent expression patterns at different temperatures. Furthermore, the dynamics of the pectin network based on Ca2+ cross-linking and decrosslinking indicates a second role for calcium in controlling petal movement in G. rigescens. This result suggests that a homeostasis between bound and unbound Ca2+ in pectin exists besides the homeostasis among the cell wall, the cytosol, and the other organellar compartments. Finally, we show that auxin has antagonistic effects on flower opening at high temperatures and closing at low temperatures (Fig. 6), and blocking auxin transport delays flower blooming and quickens flower closure. The antagonistic effects suggest the presence of a complex hormonal signaling crosstalk in controlling flower repetitive opening and closing. Even though the exact mechanism is unknown, we hypothesize that AHAs may play a role in the process via the acidification of the extracellular environment. A low pH environment activates expansins, which relax the cell wall by breaking the hydrogen bonds between cellulose microfibrils, thereby promoting cell wall expansion [50]. The low pH environment also activates XTH and pectin-metabolizing enzymes, thereby modulating the cell wall flexibility and resulting in petal movement. Distinct temperature-induced Ca2+ influxes may result in disparate polar transport and distribution of auxin. It is also possible that distinct downstream effectors regulate the expression of Ca2+ signaling components, such as Ca2+ channels, calmodulin, and Ca2+-dependent protein kinases, forming a feedback loop that creates a complex regulatory network. Conclusions By providing a chromosome-level genome and temperature-specific flower transcriptomes for G. rigescens, this study establishes a global framework for understanding temperature-dependent flowering and repetitive petal movements. Integrating genomic and transcriptomic insights with phenotypic observations, we reveal that Ca2⁺ signaling, dynamic cell wall remodeling, and auxin-mediated processes collectively orchestrate repetitive flower opening and closure. G. rigescens, thus representing a robust model for dissecting temperature-regulated flower repetitive opening and closing, and future studies under finely controlled environmental conditions will be essential to resolve the contributions of additional cues and hormonal crosstalk in alpine floral adaptation. Methods Materials In this research project, G. rigescens was grown in Jiufang village, Manwan Town, Yunxian County, Lincang City, Yunnan Province (23 °C 56′ ~ 24° C 46′N, 99 °C 43′ ~ 100 °C 33′E), at an altitude of 2440 m. Observations and sampling were conducted from October 2018 to December 2018, October 2019 to December 2019, and November 2020. The flowers of G. rigescens were collected at 6:00 (AM6), 12:00 (PM12), and 18:00 (PM6). Samples were snap-frozen in liquid nitrogen and then stored at –80 °C. Flowers at different temperatures and illumination treatments In the field, 30 randomly selected G. rigescens flowers in full opened (within 3 days after blooming) were recorded at different times. Observations and measurements of the time and ambient temperature required for flower opening and closing in natural conditions were taken for three consecutive days. Statistics is as follows: one-way ANOVA with Tukey’s multiple comparison test. Besides, the plants of G. rigescens in full opened (within 3 days after blooming) were grown for 3 days under (10,000 Lx) illumination and darkness (22 °C or 10 °C) to observe whether the flowers opened or closed. Thirty G. rigescens plants of each group in the budding stage were transplanted into an illuminated plant growth chamber (RDN type top-mounted artificial climate box, model number RND-1000D-3, Ningbo Southeast Instrument Co., Ltd., China). The culture humidity was 60%, and the culture illumination and temperature were as follows: 22 °C, illumination 10,000 lx; 22 °C, darkness; 10 °C, illumination 10,000 lx; and 10 °C, darkness. Plants in the control group were maintained under natural conditions for 2 weeks.The culture lasted 2 weeks before plants were observed in each group. Statistics is as follows: one-way ANOVA with Tukey’s multiple comparison test. The experiments were repeated three times in separate growth chambers under the same conditions. Fully opened flowers of G. rigescens (within 3 days after blooming) were incubated in growth chambers at 14 °C, 13 °C, 12 °C, 11 °C, 10 °C, 8 °C, and 5 °C under illumination (10,000 lx). At each temperature gradient, whether the flowers closed and the time required for closure was recorded, five plants per treatment. Similarly, completely closed flowers (within 3 days after blooming) were incubated at 15 °C, 17 °C, 20 °C, 22 °C, 25 °C, and 28 °C under illumination (10,000 lx). At each temperature gradient, flower opening and closure were recorded, and the time required for flower opening was measured, five plants per treatment. All treatments were independently repeated three times. Statistics is as follows: one-way ANOVA with Tukey’s multiple comparison test. Determination of the floral organ in sensing ambient temperature After different parts of G. rigescens were removed, the plants were incubated at 22 °C and 10 °C (10,000 Lx or darkness), respectively, to observe flower opening and closing. The comparison groups are as follows: (a) Without leaves, (b) without leaves and sepals, and (c) without leaves, sepals, pistils, and stamens. Single flowers were collected and incubated at 22 °C and 10 °C (10,000 Lx or darkness), respectively, to observe flower opening and closing. A second experiment using single flowers without sepals was also performed. Treatment with pectinase and auxin inhibitor Mixed pectinase inhibitor (Pythonbio, Guangzhou, China) and auxin transport inhibitor (TIBA, Macklin, Shanghai, China) were dissolved in dimethyl sulfoxide (DMSO) to a final concentration of 10 μM [3, 57]. After flower closure, the inflorescences of G. rigescens were sprayed with either pectinase inhibitor or TIBA every 2 h (three applications in total). A control group was treated with DMSO solution. After 12 h of treatment, plants were transferred to conditions of 22 °C and illumination of 10,000 lx to record the flower-opening process and the time required for full blooming. After 12 h, these plants were transferred to conditions of 10 °C and illumination of 10,000 lx to observe flower closure and the time required for complete closure. Each experiment included five flowers and was repeated three times. Statistical significance was determined using one-way ANOVA followed by Tukey’s multiple comparison test. Scanning electron microscopy of the perianth abaxial surface A 1 cm in length and width tissue sample excised from the perianth of both open and closed flowers was observed in the morphology of the abaxial epidermal cells with HITACHI Regulus8100 scanning electron microscope. At least 10 flowers or 85 cells were measured in triplicate during each experiment. Sequencing Using a modified CTAB method, genomic DNA was extracted from the fresh leaves of a single plant. The quality and concentration of genomic DNA were measured using 1% agarose gel electrophoresis and a Qubit 3.0 fluorometer (Invitrogen, Carlsbad, CA, USA). Total RNA from flowers, leaves, stems, and roots was extracted using the Plant RNA Kit (Tiangen, Beijing, China). The Oxford Nanopore library was constructed using 10 μg of genomic DNA following the Ligation Sequencing Kit SQK-LSK109 protocol. Four short paired-end libraries were constructed with insertion sizing from 210 bp, 330 bp, 430 bp, and 510 bp (NEBNext Ultra II DNA Kit, NEB, USA). The Illumina NovaSeq 6000 platform was used to sequence paired-end reads for all these libraries. In addition, the Hi-C library was prepared following an improved modification protocol by our laboratory [58]. Furthermore, roots, stems, leaves, and different states of flowers (AM6, PM12, and PM6) were collected from the same single plant. RNA-seq libraries were produced with NEBNext Ultra RNA Library Prep Kit (NEB, USA). Genome size estimation and ploidy analysis Based on flow cytometry, the genome size of G. rigescens was estimated. As described previously, 20 mg of fresh leaves was prepared for flow cytometric analysis. Analyses of nuclear samples were performed using the BD FACScalibur flow cytometer (BD Biosciences, USA), and the estimated genome size of G. rigescens was 1.86 Gb with the internal standard of Zea mays ssp. mays var. B73 (2300 Mb) [59]. The genome size was also estimated with GenomeScope2 (Version: 2.0) [60] using the distribution of 21-mer frequencies generated from Illumina reads using Jellyfish (Version: 2.3.1) [61]. The genome ploidy was predicted with 21-mer frequencies using Smudgeplot (Version: 0.4.0dev) [60]. Genome assembly To assemble long reads, NextDenovo2.2 (https://github.com/Nextomics/NextDenovo) used the “correct-then-assemble” strategy to perform de novo genome assembly of G. rigescens. The Oxford Nanopore sequencing data were self-corrected using NextCorrect (NextDenovo2.2) with the parameters as follows: read cutoff = 1 k, seed cutoff = 25.17 k, and block size = 10 g. NextGraph (NextDenovo2.2) was used to assemble the corrected reads: random_round = 100, minimap2_options_cns = -x ava-ont -t 32 -k17 -w17, and nextgraph_options = -a 1. NextPolish was used to correct errors based on the assembled genome with Illumina short paired-end reads [62]. After that, the assembled genome was processed through the Purge Haplotigs pipeline to eliminate redundant sequences. Scaffolding was also performed using Hi-C based 3d-DNA [63]. The Hi-C library was produced using the improved standard procedure described below. By filtering raw reads from sequencing, we ensured that clean reads would be used in the information analysis. After the filter, 1.03-Gb clean reads were used as input for the Juicer [64] (version 1.5.6), 3d-DNA version 180,922 Hi-C analysis and scaffolding pipeline. BWA mem version 0.7.17-r1188 was used to align the Illumina Hi-C reads to the contigs. The contigs were then ordered and oriented by the 3d-DNA pipeline with the parameter “–editor-repeat-coverage 3”. By using Juicebox, we visualized the Hi-C contact matrix and manually corrected misassemblies and misjoins according to neighboring contacts. The manually validated assembly was used to generate pseudomolecules sorted by size using the finalized section from 3d-DNA. Annotation of repetitive elements We combined ab initio and evidence-based approaches to predict repetitive elements. RepeatMasker was used to search for repeat content in the assembly using the Viridiplantae database from Repeatbase, which was classified from the repeat protein database (Release16.10; http://www.girinst.org/repbase/index.html) using default settings to identify and mask the well-characterized TEs [65]. RepeatMasker was used to identify known repeats for homolog-based prediction (Version: open-4.0.9). A de novo prediction was performed using LTRharvest (version 4.9.1) and LTR_FINDER_parallel (Version v1.07) [66], and the results were integrated using LTR_retriever (Version: v2.8) [67]. The LTR insert time was based on the LTR_retriever results and calculated using the equation T = K/2μ, where K is the LTR divergence rate and μ is the neutral mutation rate (1.3 × 10−8 mutations per site per year) [68, 69]. Use Repeatmodeler (version 2.0) to generate another de novo repeat library. RepeatMasker is rerun against the masked genome assembly, and the merged de novo repeat library is used as the query library [70]. Using TRF (version 4.07b), tandem repeats were identified within the genome assembly. Prediction and annotation of protein-coding genes Glimmer HMM [71], SNAP [72], GenScan [71], and Augustus [73] were used for the ab initio prediction of the protein-coding genes under default settings. In order to make homology-based predictions, the protein sequences of G. rigescens, G. macrophylla, G. dahurica, O. pumila [74], O. europaea [75], Catharanthus roseus [76], G. sempervirens [77], and H. annuus [78] were downloaded. The alignment to the G. rigescens genome was performed using TBLASTN v2.2.29 + with a cutoff E-value of 1e-5. Further gene models were predicted using GeneWise v2.4.1 on each genomic region of each TBLASTN hit. RNA-Seq reads were aligned with the genome using TopHat, and transcripts were assembled using Cufflinks [79]. A transcriptome-based gene model was generated by aligning assembled transcripts with the genome assembly using the Program to Assemble Spliced Alignments (PASA) v2.4.1. AUGUSTUS v3.2.2 was trained with the script “autoAugTrain.pl” using transcriptome-based gene models from PASA. The protein-coding genes’ structure was then predicted using AUGUSTUS, a transcriptome-trained algorithm. Repetitive elements were masked using EvidenceModeler v1.1.1, which combined the gene models into a weighted consensus gene structure (the weights are as follows: homology-set > transcripts-set > ab initio-set and redundant genes are deleted) [58]. An E-value of 1e-5 was set for t-RNAscan-SE [80] to predict tRNA in genomic sequences. The RNA sequences in the Rfam database were used to predict rRNA by BLASTN, with an E-value set to 1e-5. BLASTN was also used to predict miRNA and snRNA genes compared to the Rfam database with an E-1 E-value. Gene functional annotation was performed by searching against the public database using BLASTP. To screen the proteins for functional domains, we used SwissProt, TrEMBL, KEGG, and InterPro. Gene Ontology (GO) terms are taken out of the corresponding InterPro entries. Evolution and WGDs The gene family clustering was performed with protein-coding gene sequences of G. rigescens, G. macrophylla, G. dahurica, O. pumila, O. europaea, C. roseus, G. sempervirens, and H. annuus using OrthoFinder (Version: 2.5.4) [81]. The aligned multiple single-copy orthologous genes were used to construct a rooted species tree using the ML method. The ML tree was used to infer the species divergence times with the MCMCTREE in PAML (Version: 4.9j) [82]. Tree node calibrations were based on known divergence times from the TimeTree web service (http://www.timetree.org/). CAFE5 [83] was applied to identify gene families that had undergone expansion and contraction. The gamma modeling with four gamma rate categories and the root equilibrium frequency with Poisson distribution were set. All-to-all Blastp analysis of protein sequences was performed within/between G. rigescens, G. macrophylla, and G. dahurica, respectively, using an E-value cutoff of 1e-10. Syntenic regions within each species or between two species were then identified using MCscanX [84] based on the all-to-all BLASTP results. Protein sequences of paralogs or homologous gene pairs in the identified syntenic regions were first aligned using MUSCLE [85], and the protein alignments were then converted to the CDS alignments. Finally, synonymous divergence values were calculated on these CDS alignments using KaKs_Calculator (Version 1.2) [86] with the NG method. Analysis of DEGs of G. rigescens flower FPKM values were filtered to keep values > 0.1 in at least one sample/stage and then transformed and normalized into TPM. After scaling TPM to log2 (TPM + 1), limma_3.46.0 was used to fit linear models for the arrays and compute moderated t-statistics and moderated F-statistics. Based on empirical Bayes moderation of standard errors towards a global value, calculate the log-odds of differential expression. DESeq2 [87] software was performed to identify differentially expressed genes (DEGs) with the following parameters: |log2 (fold change)|≥ 1 and FDR adjusted P ≤ 0.05. Gene functional annotation and enrichment analysis were performed using EggNOG-mapper v2 and OrmicShare Tools, respectively. These genes not expressed at the flowering stage (PM12) and those not expressed at the closing stage (AM6 and PM6) were selected for the same annotation and enrichment methods as above. Time-series expression based on a loose clustering algorithm used Mfuzz [21]. ANN and CNGC gene family analysis We used HMMER v3.3.2 [88] with the “–cut_t” parameter to identify genes, employing Pfam HMM models (ANN: PF00191, CNGC: PF00027 and PF00520) as queries for two gene families. A. thaliana genes were downloaded from the TAIR database (https://www.arabidopsis.org/) and used as outgroups. Protein sequences for ANNs and CNGCs were aligned with MAFFT [89], and poor alignments were trimmed using trimAl [90]. ML phylogenetic trees were constructed with IQ-TREE [91] and visualized with iTOL software (https://itol.embl.de/). Different types of gene duplication were predicted using the “DupGen_finder-unique.pl” module of DupGen_finder [92], employing default settings and A. thaliana serving as the reference. MEME [93] (http://meme-suite.org/tools/meme) was used to analyze the conserved motifs, with the prediction threshold set to 15. The conserved domains and motifs in ANNs and CNGCs were conducted in the GSDS [94]. The 2-kb sequences upstream of the identified genes were analyzed in PlantCARE to predict promoter cis-acting elements. Visualization of conserved domains, conserved motifs, heatmaps, and cis-acting elements was performed using the Evolview website [95]. RT-qPCR assay The EvaGreen 2 × qPCR MasterMix Kit (ABM, Vancouver, Canada) was used in a QuanstudioTM 5 Real-Time System (Thermo Fisher Scientific, Inc.) to validate the expression of related DEGs observed in RNA-Seq data. For first-strand cDNA synthesis, total RNA extracted from whole flower samples was reverse transcribed using the RevertAid™ First-Strand cDNA Synthesis Kit (TransGen Biotech, Beijing, China). The primers were designed using Primer-Blast (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). The primers used for RT-qPCR analyses are listed in Additional file 1: Table S28. Using the Quanstudio™ 5 Real-Time PCR software, the relative expression values were calculated using the 2−ΔΔCt method. Statistical analyses GraphPad Prism 8 and SPSS 25.0 were used for statistical analysis and mapping. Statistical significance was assessed using two-tailed Student’s t-tests (for pairwise comparisons) or one-way ANOVA followed by Tukey’s multiple comparison test (for multiple groups). In the case of P ≤ 0.05, the difference was considered significant. To calculate the relative multiple changes in transcripts, RT-qPCR CT values were normalized, and REST was used to calculate relative multiple changes in transcripts [96]. Supplementary Information Additional file 1: Tables S1–S28. Table S1-Statistics of sequencing data for G. rigescens. Table-S2 Summary of G. rigescens genome assembly. Table S3-Summary of Hi-C assisted assembly pseudomolecule lengths. Table-S4 BUSCO assessment of the completeness of the G. rigescens genome assembly. Table-S5 Prediction of repetitive sequences in the G. rigescens genome. Table-S6 Categories of TEs predicted in the G. rigescens genome. Table-S7 Statistics of non-coding RNAs. Table-S8 Gene numbers and features of G. rigescens. Table-S9 Gene function annotation in the G. rigescens genome. Table-S10 Orthologous protein-coding genes clustering. Table-S11 Results of KEGG enrichment analysis of species-specific orthogroups in G. rigescens. Table-S12 Results of GO enrichment analysis of expanded gene families in G. rigescens. Table-S13 Results of KEGG enrichment analysis of expanded gene families in G. rigescens. Table-S14 Ambient temperatures at 6:00, 10:00, 12:00 and 18:00 in the wild. Table-S15 The G. rigescens flowering rate under different temperature and light regimes in the lab. Table-S16 Time required for the opening of G. rigescens flower at different temperatures. Table-S17 Time required for the closing of G. rigescens flower at different temperatures. Table-S18 Differentially expressed genes in G. rigescens flowers at different temperatures. Table-S19 GO enrichment results of DEGs. Table-S20 KEGG enrichment results of DEGs. Table-S21 Time-series expression analysis of DEGs. Table-S22 GO enrichment results of pattern 1. Table-S23 GO enrichment results of pattern 2. Table-S24 GO enrichment results of pattern 3. Table-S25 GO enrichment results of pattern 4. Table-S26 The time required of flower opening and closure upon treatment with pectinase inhibitors. Table-S27 The time required of flower opening and closure upon treatment with TIBA. Table-S28 The list of primers used in qRT-PCR analyses.Additional file 2: Figures S1–S17. FigS1-Determination of genome size of G. rigescens by flow cytometry and Hi-C interactome. Fig-S2 LTR retrotransposon (LTR-RT) insertion times of 3 Gentiana species. Fig-S3 Upset plot showing orthogroups from six Gentianales species. Fig-S4 KEGG enrichment analysis of species-specific orthogroups in G. rigescens (Top 20). Fig-S5 GO enrichment analysis of expanded gene families in G. rigescens (Top 20). FigS6 KEGG enrichment analysis of expanded gene families in G. rigescens (Top 20). Fig-S7 Dotplot comparing the G. rigescens and V. vinifera genomes. Fig-S8 Dotplot comparing the G. rigescens and G. macrophylla genomes. Fig-S9 The distribution of 4DTv value of intraspecific and interspecific collinear blocks. Fig-S10 Flower opening and closing observation under different temperature, time and light regimes. Fig-S11 Flower organ removed to determines the sensing part. Fig-S12 Volcano plots of AM6 and PM6. Fig-S13 KEGG enrichment analysis of (Top 20). Fig-S14 ANN gene family analysis. Fig-S15 CNGC gene family phylogenetic tree of 5 species. Fig-S16 CNGC gene family analysis. Fig-S17 The heatmap of putative regulation network related genes in G. rigescens petal movement. Additional file 1: Tables S1–S28. Table S1-Statistics of sequencing data for G. rigescens. Table-S2 Summary of G. rigescens genome assembly. Table S3-Summary of Hi-C assisted assembly pseudomolecule lengths. Table-S4 BUSCO assessment of the completeness of the G. rigescens genome assembly. Table-S5 Prediction of repetitive sequences in the G. rigescens genome. Table-S6 Categories of TEs predicted in the G. rigescens genome. Table-S7 Statistics of non-coding RNAs. Table-S8 Gene numbers and features of G. rigescens. Table-S9 Gene function annotation in the G. rigescens genome. Table-S10 Orthologous protein-coding genes clustering. Table-S11 Results of KEGG enrichment analysis of species-specific orthogroups in G. rigescens. Table-S12 Results of GO enrichment analysis of expanded gene families in G. rigescens. Table-S13 Results of KEGG enrichment analysis of expanded gene families in G. rigescens. Table-S14 Ambient temperatures at 6:00, 10:00, 12:00 and 18:00 in the wild. Table-S15 The G. rigescens flowering rate under different temperature and light regimes in the lab. Table-S16 Time required for the opening of G. rigescens flower at different temperatures. Table-S17 Time required for the closing of G. rigescens flower at different temperatures. Table-S18 Differentially expressed genes in G. rigescens flowers at different temperatures. Table-S19 GO enrichment results of DEGs. Table-S20 KEGG enrichment results of DEGs. Table-S21 Time-series expression analysis of DEGs. Table-S22 GO enrichment results of pattern 1. Table-S23 GO enrichment results of pattern 2. Table-S24 GO enrichment results of pattern 3. Table-S25 GO enrichment results of pattern 4. Table-S26 The time required of flower opening and closure upon treatment with pectinase inhibitors. Table-S27 The time required of flower opening and closure upon treatment with TIBA. Table-S28 The list of primers used in qRT-PCR analyses. Additional file 2: Figures S1–S17. FigS1-Determination of genome size of G. rigescens by flow cytometry and Hi-C interactome. Fig-S2 LTR retrotransposon (LTR-RT) insertion times of 3 Gentiana species. Fig-S3 Upset plot showing orthogroups from six Gentianales species. Fig-S4 KEGG enrichment analysis of species-specific orthogroups in G. rigescens (Top 20). Fig-S5 GO enrichment analysis of expanded gene families in G. rigescens (Top 20). FigS6 KEGG enrichment analysis of expanded gene families in G. rigescens (Top 20). Fig-S7 Dotplot comparing the G. rigescens and V. vinifera genomes. Fig-S8 Dotplot comparing the G. rigescens and G. macrophylla genomes. Fig-S9 The distribution of 4DTv value of intraspecific and interspecific collinear blocks. Fig-S10 Flower opening and closing observation under different temperature, time and light regimes. Fig-S11 Flower organ removed to determines the sensing part. Fig-S12 Volcano plots of AM6 and PM6. Fig-S13 KEGG enrichment analysis of (Top 20). Fig-S14 ANN gene family analysis. Fig-S15 CNGC gene family phylogenetic tree of 5 species. Fig-S16 CNGC gene family analysis. Fig-S17 The heatmap of putative regulation network related genes in G. rigescens petal movement.