CmDREBa regulates the perennial growth habit of chrysanthemum in response to prolonged chilling Authors: Tianhua Jiang, Xuening Liu, Jiayi Luo, Lei Liu, Yufei Liu, Mingzheng Han, Conglin Huang, Chang Luo, Yafei Zhao, Chao Ma Published: 2026 ABSTRACT AbstractSeasonal growth modulation is a key adaptive trait in perennials that supports overwinter survival. During winter, many perennials adopt a rosette growth habit by suppressing both vegetative extension and flowering, but the mechanism that coordinates these two distinct developmental programs remains unclear. Here, we show that prolonged chilling primarily acts at the shoot apex to confer extension growth and flowering competence in chrysanthemum (Chrysanthemum morifolium). Long-term chilling is accompanied by marked changes in gene expression in the rib zone of the shoot apex. Notably, the expression of CmDREBa is associated with the transition from extension growth to rosette growth. Chilling induces CmDREBa expression, which suppresses both extension growth and flowering; when temperatures rise, CmDREBa expression declines, allowing both processes to resume. CmDREBa directly induces the expression of MADS AFFECTING FLOWERING (MAF) genes, which in turn promote rosette growth. Together, these findings identify CmDREBa as a key regulator of chilling-responsive seasonal growth.Subject terms: Abiotic, Plant physiology, Plant development Introduction Angiosperm families initially diversified at least 140 million years ago during the Mesozoic era, when temperate seasonal environments were not present. Today, fewer than half of modern angiosperm families contain members that have acquired cold adaptation mechanisms to temperate climates through convergent evolution1–3. One of these adaptations, exhibited by many perennials, is a distinct seasonal growth transition characterized by the formation of a compact rosette during winter. In this rosette state, internodes remain extremely short, leaves are densely clustered, and shoot elongation is suppressed4–7. At the same time, flowering competence is also suppressed, reflecting a reversible blockage of the acquired ability of meristems to initiate flowering. The rosette habit preserves the meristem of the perennating organ in an indeterminate state, thereby enhancing winter survival5,8–10. Extension growth, defined as rapid stem elongation, resumes in the following spring and involves simultaneous restoration of flowering competence in the overwintering rosette8,11,12. Precise architectural organization and tightly regulated phenotypic plasticity in response to seasonal environmental cues are fundamental to the perennial life cycle. However, the molecular mechanisms that govern this seasonal growth transition in rosette-forming perennials have not been fully characterized. The perception of low temperature likely triggers two key physiological changes in overwintering plants: (1) the coordination of developmental progression to suppress vegetative and reproductive growth, and (2) the activation of chilling-sensing mechanisms that promote cold tolerance13. The suppression of premature flowering in many overwintering plants is mediated by the MADS-box transcription factor gene flowering locus C (FLC)14–16. Vernalization is a prolonged chilling exposure that represses FLC expression, thereby promoting flowering in the following spring or early summer17. In Arabidopsis (Arabidopsis thaliana), FLC belongs to the MADS affecting flowering (MAF) clade of MADS-box floral repressors, which includes FLM, MAF2, MAF3, MAF4, and MAF5. These MAF proteins are partially redundant with FLC in the mediation of vernalization-dependent flowering18–20. In chrysanthemum (Chrysanthemum morifolium), CmMAF2 was shown to regulate flowering time in response to low temperatures21, and MAF genes contribute to the regulation of meristem activity necessary for the perennial growth habit in Brassicaceae species, suggesting that they have broader roles in modulating bud growth and development over winter22–24. Cold tolerance is primarily regulated by the dehydration-responsive element binding 1/C-repeat factor (DREB1/CBF) pathway25,26. DREB1-type transcription factors activate the expression of a suite of cold-regulated genes by binding to dehydration-responsive element (DRE)/ C-repeat (CRT) cis-elements in the promoters of target genes27,28. Notably, overexpression of DREB1 can cause dwarfism, delayed flowering, and even early dormancy in woody plants, hinting at its potential role in integrating the cold response with developmental arrest29–31. The herbaceous perennial species chrysanthemum provides a classic example of a seasonal growth cycle based on overwinter rosette formation: flowering is induced under short days, after which the main stem senesces prior to winter. Beginning in late fall, the plant forms a rosetted rhizome that remains in a vegetative, minimally growing state, safeguarded against extreme conditions and thus ensuring overwinter survival. In the next spring, the rosette then initiates shoot elongation and grows rapidly in the summer (Fig. 1a). However, the molecular mechanisms that mediate the transition from rosette growth to extension growth in response to chilling in this perennial model remain unknown. Growth habit of chilling-dependent chrysanthemum in response to long-term low-temperature exposure. a The life cycle and growth habit of perennial herbaceous chrysanthemum. b Chilling-dependent (CD) chrysanthemum grown under long-day (LD) conditions at 22 °C for more than 1 year. Scale bar, 5 cm. c CD chrysanthemum grown under LD conditions at 22 °C followed by 6 weeks of low-temperature treatment under LD conditions at 4 °C. Scale bar, 5 cm. For b, c, similar results were observed in five individual plants. d Representative photographs of CD chrysanthemum in response to low temperature exposure. Plants grown under LD conditions at 22 °C were subjected to 6 weeks at low temperature (LT0–LT6) before being returned to normal temperature (22 °C) under LD conditions for 9 weeks (LT6 + 3–9). To accelerate the floral transition, plants were then transferred to short-day (SD) conditions for 5 weeks (LT6 + 9–14). Plants grown under the same daylength conditions without low-temperature treatment (C0–C20) were used as controls. Scale bar, 10 cm. e Plant height of CD chrysanthemum under control conditions or after the low-temperature treatment illustrated in (d). f Leaf number of chrysanthemum under control conditions or after low-temperature treatment. In e, f, results are shown as the mean ± standard deviation (SD), n = 5 independent biological replicates. g Grafting experiment between chilled and unchilled plants. Left panel, unchilled shoot tips (scion) were grafted onto chilled rootstocks that had been subjected to 6 weeks of low temperature (4 °C, LT6). Right panel, chilled shoot tips (LT6, scion) were grafted onto unchilled rootstocks. Inset, zoomed-in view of the apical bud, which has developed into a capitulum bud. Scale bar, 5 cm. Source data are provided as a Source data file. Here, we performed spatial transcriptomics to investigate the molecular basis of this developmental transition and reveal that CmDREBa plays a critical role in coordinating rosette growth and extension growth through the regulation of MAF genes. Results Acquisition of extension growth and flowering competence in response to prolonged chilling occurs in the shoot apex Chrysanthemum cultivars and their wild relatives can be classified as chilling-dependent (CD) or chilling-independent (CI) types, depending on whether their extension growth and acquisition of flowering competence require an obligatory chilling period. CD chrysanthemums exhibit a rosette growth habit and do not flower without prolonged chilling12,21. Consistent with these previous reports, we observed that the CD chrysanthemum cultivar forms a non-flowering rosette in the absence of chilling, but acquires the ability to undergo extension growth and flowering after exposure to low temperature (Fig. 1b, c). Indeed, plants exposed to low temperature (4 °C) for 6 weeks exhibited extension growth and were taller than control plants. Leaf number was lower than that of control plants after 6 weeks of low-temperature treatment (LT6) but resumed rising upon return to normal temperature and was eventually comparable to that of control plants (Fig. 1d–f). Low-temperature treatment of 6 weeks or 8 weeks had similar effects on growth and flowering, whereas 4 weeks of low-temperature treatment had a weaker effect (Supplementary Fig. 1). Six weeks of low temperature were therefore sufficient to fulfill the chilling requirement of CD chrysanthemum, and we used 6 weeks as the standard low-temperature treatment for the remainder of the study. We performed reciprocal grafting experiments to determine whether the chilling effect acted locally within the shoot apex or systemically through a mobile signal. To this end, we grafted shoot tips (scions) from unchilled plants onto rootstocks that had been chilled, and vice versa. The unchilled scions grafted onto chilled rootstocks remained in a rosette state and did not undergo extension growth or flowering. By contrast, the chilled scions grafted onto unchilled rootstocks exhibited extension growth and completed the floral transition, although they did not reach anthesis, while their unchilled rootstocks remained in rosette form (Fig. 1g). These results indicate that the acquisition of extension growth and flowering competence in response to prolonged chilling occurs at the shoot apex. Spatial transcriptomes of chilled and unchilled shoot apices from CD chrysanthemum plants To dissect the spatiotemporal changes in gene expression associated with chilling-induced acquisition of extension growth and flowering competence, we generated spatial transcriptome datasets of chilled shoot apices (6 weeks at 4 °C followed by 3 weeks at 22 °C, LT6 + 3) and unchilled shoot apices (9 weeks at 22 °C). The two biological replicates of each treatment exhibited good reproducibility and were used for subsequent analysis (Supplementary Fig. 2). In total, transcripts for 53,180 genes were detected from 1129 spots at a resolution of 100 μm, with an average of 5080 genes per spot (Supplementary Table. 1). We categorized the spots from the chilled and unchilled shoot apices into seven clusters, as visualized by uniform manifold approximation and projection (UMAP) (Fig. 2a and Supplementary Fig. 2). On the basis of apical bud anatomy, the clusters were designated as follows: cluster 0 (chilled rib zone), cluster 1 (leaf primordium/young leaf), cluster 2 (primordial meristem), cluster 3 (chilled vascular tissue), cluster 4 (unchilled rib zone), clusters 5 and 6 (vascular tissue). We identified feature genes for each cluster (Fig. 2b and Supplementary Data 1). For example, YABBY 1 (YAB1, also reported as filamentous flower [FIL]; Cse-sc011972.1-g050.1), the Arabidopsis ortholog of which is highly expressed in the distal axis of the leaf primordium32, is a feature gene for cluster 1. Terminal flower 1 (TFL1; Cse_sc005398.1_g060.1), a known flowering repressor in chrysanthemum33, is a feature gene for cluster 4. Spatial transcriptomes of shoot apices from CD chrysanthemum with and without long-term low-temperature exposure. a Spatial distribution showing assignment of spots to different gene-expression clusters in representative samples of CD chrysanthemum. Clusters are color-coded as shown. Scale bar, 100 μm. b Violin plots showing the expression patterns of the indicated feature genes in each cluster. The y-axis represents the expression level of each gene. c Uniform manifold approximation and projection (UMAP) plot showing the clear separation of clusters 0 and 4 along the first UMAP axis. d Gene Ontology (GO) enrichment analysis of differentially expressed genes (DEGs) between cluster 0 and cluster 4. Criteria for differential gene expression was set as P < 0.001, |log2FoldChange| ≥ 0.9 using a two-sided Wald test. ClusterProfiler was used for the enrichment analysis with a one-sided hypergeometric test. To control the false discovery rate, P-values were adjusted for multiple comparisons using the Benjamini–Hochberg (BH) procedure. UMAP analysis revealed that rib-zone tissues can be divided into two clusters based on chilling exposure of their originating tissues: cluster 0 (chilled rib zone) and cluster 4 (unchilled rib zone, Fig. 2c). The substantial distance between cluster 0 and cluster 4 in the UMAP plot suggested that gene expression patterns are dramatically altered in the rib zone after chilling treatment. We therefore examined the 148 differentially expressed genes (DEGs) (p < 0.001; q ≤ 0.001; |log2[fold-change]| ≥ 0.9) between the two clusters, comprising 38 genes with higher expression in cluster 0 and 110 genes with higher expression in cluster 4 (Supplementary Data 2). Gene ontology (GO) term enrichment analysis showed that these DEGs are enriched in terms related to “reproductive process,” “multicellular organismal development,” “post-embryonic development,” and “anatomical structure development” (Fig. 2d), which may be associated with the acquisition of flowering competence and extension growth after prolonged chilling. CmDREBa induces rosette growth in chrysanthemum Among the DEGs identified from the spatial transcriptomes, we noticed the higher expression of the DREB1 transcription factor gene CmDREBa (Cse_sc015801.1_g010.1) in unchilled than in chilled shoot apices (LT6 + 3), particularly in the rib zone (Fig. 3a). We investigated the expression of CmDREBa in shoot apices of CD chrysanthemum in response to prolonged chilling by in situ hybridization and RT-qPCR. In situ hybridization confirmed the spatial transcriptomic results, revealing that CmDREBa transcripts are more abundant in unchilled shoot apices (LT0) than in chilled apices (LT6 + 3), particularly within the rib zone (Fig. 3b). RT-qPCR showed that CmDREBa transcript levels rise during chilling treatment (LT3 and LT6) but were barely detectable after a 3-week recovery period at 22 °C (LT6 + 3, Fig. 3c). Extension growth and flowering of CmDREBa-RNAi CD chrysanthemum. a Spatial profiles of CmDREBa expression in chilled and unchilled shoots by spatial transcriptome. Scale bar, 100 μm. b In situ hybridization of CmDREBa transcripts. Plants were subjected to 0, 3, or 6 weeks of low-temperature treatment (LT0–LT6) before being returned to normal temperature (22 °C) under LD conditions for 3 weeks (LT6 + 3). The red square highlights a zoomed-in view of the in situ signals in the rib zone. A sense CmDREBa probe was used as a negative control. Scale bar, 100 μm. c Relative CmDREBa expression levels in response to low temperature in CD chrysanthemum, as determined by RT-qPCR. Results are shown as means ± SD, n = 4 independent biological replicates. d Phenotypes of wild-type (WT, CD chrysanthemum) and CmDREBa-RNAi (CD-RNAi) CD chrysanthemum. Plants were cultivated at normal temperature for 6 weeks under LD conditions and then transferred to SD conditions. The plants shown were photographed 125 days after transfer to SD conditions. Scale bar, 5 cm. e Relative CmDREBa transcript levels in WT plants and CD-RNAi lines (CD-RNAi#3 and CD-RNAi#5), as determined by RT-qPCR. Results are shown as means ± SD, n = 4 independent biological replicates. f Plant height of WT and CD-RNAi lines. g Days to visible capitulum buds in WT and CD-RNAi plants. Days were recorded following the transfer to SD conditions. nb, no bloom. In f, g, results are shown as means ± SD, n = 5 independent biological replicates. In c, significant differences were determined by one-way ANOVA analysis followed by Tukey’s multiple comparison test. Data were log10-transformed to achieve a normal distribution and homogeneity of variance prior to analysis (c). In e, f, the significance of differences between WT and CD-RNAi lines was determined by two-tailed Student’s t tests. Source data are provided as a Source data file. We used the 3′ region of the CmDREBa transcript to specifically silence CmDREBa by RNA interference (RNAi) in CD chrysanthemum and selected two representative CmDREBa-RNAi lines (#3 and #5) for functional analysis (Fig. 3d). We confirmed lower CmDREBa transcript levels in these lines by RT-qPCR (Fig. 3e). We also evaluated the expression of other DREB1 genes in WT and CmDREBa-RNAi lines and determined that only CmDREBa expression is knocked down (Supplementary Fig. 3). Plants from the CmDREBa-RNAi lines (11.6 ± 0.8 cm, 12.2 ± 0.5 cm) were taller than WT plants (6.9 ± 0.8 cm) (Fig. 3f). In addition, flowering competence was accelerated in the CmDREBa-RNAi lines, which flowered after ~70 days under short-day (SD) conditions, whereas the WT failed to flower by the end of the experiment (150 days after transplant) (Fig. 3d, g). These results demonstrate that CmDREBa is a crucial regulator of chilling-induced extension growth and flowering in CD chrysanthemum. Unlike CD chrysanthemum, CI chrysanthemum can undergo extension growth and flowering in the absence of long-term chilling. We therefore examined whether CmDREBa also regulates seasonal extension growth and flowering in CI chrysanthemum using the wild chrysanthemum C. indicum. CmDREBa in CD and CI chrysanthemum are highly similar, differing by only a single amino acid in the predicted nuclear localization signal (Supplementary Fig. 4). CmDREBa expression was upregulated in both CD and CI chrysanthemum in response to long-term and short-term exposure to low temperature, suggesting that its response to chilling is conserved (Fig. 4a and Supplementary Fig. 5a). We also compared CmDREBa expression between CD and CI chrysanthemum by RT-qPCR. At 22 °C, when CD plants exhibited rosette growth, and CI plants showed extension growth, CmDREBa transcript levels were much higher (~278-fold) in CD than in CI chrysanthemum (Fig. 4b), suggesting that high CmDREBa expression is associated with the rosette growth habit of CD chrysanthemum. Extension growth and flowering of CmDREBa transgenic chilling-independent (CI) chrysanthemum. a Relative CmDREBa expression in response to low-temperature exposure in CI chrysanthemum, as determined by RT-qPCR. Results are shown as means ± SD, n = 3 independent biological replicates. b Relative CmDREBa expression in CD and CI chrysanthemum, as determined by RT-qPCR. Shoot apices were collected from CD and CI chrysanthemum grown at 22 °C under LD conditions for 1 week, when CD plants exhibited rosette growth, and CI plants showed extension growth. Results are shown as means ± SD, n = 4 independent biological replicates. c Relative CmDREBa expression in WT (CI chrysanthemum) plants and CmDREBa overexpression lines (CI-OX#1 and CI-OX#3) in CI chrysanthemum, as determined by RT-qPCR. Results are shown as means ± SD, n = 4 independent biological replicates. d Representative photographs of WT and CI-OX chrysanthemum plants. Insets show rosettes at the tops of CI-OX plants. Plants were cultivated at normal temperature for 4 weeks under LD conditions and then transferred to SD conditions. Scale bar, 5 cm. e Plant height of WT and CI-OX plants. f Leaf number of WT and CI-OX plants. g Days to visible capitulum buds in WT and CI-OX plants. nb, no bloom. h Representative photographs of WT and CI-dreba mutant plants. Plants were cultivated at normal temperature for 4 weeks under LD conditions and then transferred to SD conditions. Scale bar, 5 cm. i Plant height of WT and Cmdreba biallelic mutants (CI-dreba) in CI chrysanthemum. j Days to visible capitulum buds in WT and CI-dreba mutant plants. In e–g, i, j, results are shown as means ± SD, n = 5 independent biological replicates. In a, significant differences were determined by one-way ANOVA analysis followed by Tukey’s multiple comparison test. Data were log10-transformed to achieve a normal distribution and homogeneity of variance prior to analysis (a). In b, c, e, f, i, j, significant differences between WT, CI-OX, or CI-dreba lines were determined by two-tailed Student’s t tests. Source data are provided as a Source data file. The CmDREBa promoters from CD and CI chrysanthemum exhibit high sequence identity (98.4%; Supplementary Fig. 6). To investigate potential epigenetic regulation of CmDREBa expression, we divided the promoter region into six contiguous fragments and examined the enrichment of three key histone modifications (H3K4me3, H3K27me3, and H3K36me3) in CD and CI plants (Supplementary Fig. 7). Much higher levels of H3K36me3, a mark generally linked to transcriptional activation, were detected on the P3 and P4 fragments of CD chrysanthemum. Differences in H3K36me3 deposition may therefore contribute to the differences in CmDREBa expression observed between CD and CI chrysanthemum. We also examined the expression of all DREB1 family members (Supplementary Fig. 5b) between buds in the extension-growth or rosette-growth state from CD and CI chrysanthemum by RT-qPCR (Supplementary Fig. 5c). In CI chrysanthemum, rosette-growth buds had higher expression of all DREB1 genes than extension-growth buds, possibly reflecting an induction by chilling. Likewise, in CD chrysanthemum, the expression of CmDREBa-1, CmDREBa-2, and CmDREBc-1, as well as CmDREBa, was higher in rosette-growth buds than in extension-growth buds (Supplementary Fig. 5c), suggesting that these genes may function redundantly in the regulation of rosette growth and extension growth. To further investigate the regulation of extension growth and flowering by CmDREBa, we overexpressed CmDREBa in CI chrysanthemum (CI-OX#1 and #3). RT-qPCR confirmed significantly higher CmDREBa expression in the CI-OX lines than in WT plants (Fig. 4c). Plants of the CI-OX lines were significantly shorter than WT plants and formed a rosette at the top of the shoot (Fig. 4d, e, and Supplementary Fig. 8). However, leaf numbers were comparable among WT and CI-OX plants (Fig. 4f), suggesting that CmDREBa overexpression mainly represses extension growth. Under SD conditions, there were no visible flower buds in CI-OX plants by the end of the experiment (150 days after transplant) (Fig. 4d, g). Because rosette formation is a key survival strategy under low temperatures, we exposed the WT and CmDREBa overexpression lines to chilling at 0–2 °C for 6 weeks. After returning to normal growth conditions, the CI-OX plants had more surviving buds than the WT plants (Supplementary Fig. 9), indicating that CmDREBa enhances the cold tolerance of vegetative meristems. We also obtained two Cmdreba biallelic knockout mutants in CI chrysanthemum using clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated nuclease 9 (Cas9)-mediated gene editing (Supplementary Fig. 10 and Supplementary Data 3). There were no significant differences in plant height among the Cmdreba mutants and the WT (Fig. 4h, i), but the Cmdreba mutants did flower significantly earlier (after 27.6 ± 0.9 and 27.2 ± 1.3 days) than the WT (after 42.8 ± 1.1 days) (Fig. 4h, j). CmDREBa is a direct regulator of MAFs expression To characterize the downstream network through which CmDREBa regulates extension growth and flowering, we performed DNA affinity purification sequencing (DAP-seq) to identify CmDREBa-binding sites globally. We performed two independent biological replicates for DAP-seq, together with DNA-input negative-control libraries. We identified a total of 126,400 peaks in the two biological replicates (Fig. 5a). CmDREBa-binding sites were highly enriched between the transcription start site (TSS) and the transcription end site (TES) of gene loci (Fig. 5b), although most of the binding sites (66.0%) were located within intergenic regions (Fig. 5c). Among the enriched motifs in the CmDREBa-bound regions was the DRE motif (CCGAC) (Fig. 5d and Supplementary Data 4), a typical binding site for members of the DREB1 family34. GO enrichment analysis revealed that genes located near CmDREBa-binding sites are enriched in terms related to “regulation of meristem development,” “plant organ morphogenesis,” “floral organ morphogenesis,” and “regulation of cellular response to stress” (Fig. 5e). These results suggest that CmDREBa regulates both vegetative and reproductive growth, in addition to stress tolerance. Genome-wide identification of CmDREBa-binding sites by DAP-seq. a Shared and unique high-confidence CmDREBa-binding peaks identified in two biological replicates. b Metaplot of CmDREBa-binding peaks. TSS transcription start site, TES transcription end site. c Distribution of CmDREBa-binding peaks across genomic features. d DNA logos of enriched CmDREBa-binding sites. e GO enrichment analysis of the CmDREBa-bound genes identified by DAP-seq. ClusterProfiler was used for the enrichment analysis with a one-sided hypergeometric test. To control the false discovery rate, P-values were adjusted for multiple comparisons using the Benjamini–Hochberg (BH) procedure. To identify direct target genes of CmDREBa, we performed a large-scale screen for DEGs between WT and CmDREBa-OX CI chrysanthemum by transcriptome deep sequencing (RNA-seq). We identified 601 upregulated DEGs and 351 downregulated DEGs in CmDREBa-OX plants relative to the WT (Supplementary Fig. 11 and Supplementary Data 5). We then compared the DEGs from RNA-seq with the genes with CmDREBa-binding sites identified by DAP-seq, revealing a set of 262 overlapping genes (Fig. 6a and Supplementary Data 6). GO enrichment analysis of these putative direct CmDREBa targets suggested their involvement in “defense response,” “reproductive process,” and “cell division,” among others (Fig. 6b). Among the DEGs related to reproduction was the MAF gene CmMAF-like 1 (Cse_sc001511_g040, Supplementary Fig. 12), the expression of which was upregulated in CmDREBa-OX chrysanthemum (Supplementary Data 6). Previous studies have highlighted MAF genes as strong candidates for the regulation of seasonal growth and flowering. Identification of target genes directly regulated by CmDREBa. a Venn diagram illustrating the overlap between CmDREBa-bound genes identified via DAP-seq and CmDREBa-regulated genes identified by RNA-seq. b GO enrichment analysis of target genes directly regulated by CmDREBa. ClusterProfiler was used for the enrichment analysis with a one-sided hypergeometric test. To control the false discovery rate, P-values were adjusted for multiple comparisons using the Benjamini–Hochberg (BH) procedure. c, f Chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) validation of CmDREBa binding to the promoters of CmMAF-like1 and CmMAF-like2 in CD and CI chrysanthemum. DRE/CRT elements are present within peak 1 and peak 2. Results are shown as means ± SD, n = 3 independent biological replicates. CmDREBa promotes the transcription of CmMAF-like1 (d) and CmMAF-like2 (g) in dual-luciferase reporter assays performed in Nicotiana benthamiana leaves. Results are shown as means ± SD, n = 6 independent biological replicates. Electrophoretic mobility shift assays (EMSAs) confirming that CmDREBa-GST binds to the promoters of CmMAF-like1 (e) and CmMAF-like2 (h), which contain DRE/CRT elements. In c, d, f, g, significant differences were determined by two-tailed Student’s t tests. Source data are provided as a Source data file. To better understand the potential regulation of CmMAF-like1 by CmDREBa, we examined the region surrounding the CmMAF-like1 locus in the C. indicum genome35. We identified an 875-kb region on chromosome 7 that contains eight putative genes, five of which are MAF genes, suggesting the presence of a MAF gene cluster (Supplementary Fig. 13a). DAP-seq identified two CmDREBa-binding peaks in the promoter regions of CmMAF-like1 and CmMAF-like2 (fold enrichment >5; Supplementary Fig. 13b and Supplementary Data 7). The CmMAF-like1 and CmMAF-like2 promoter regions in CD and CI chrysanthemum showed high sequence similarity, and both contained conserved putative CmDREBa-binding sites (Supplementary Figs. 14 and 15). We confirmed the binding of CmDREBa to two distinct promoter peaks in vivo by chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) in both CD and CI chrysanthemum (Fig. 6c, f). In addition, dual luciferase-based transient transactivation assays demonstrated that CmDREBa activates CmMAF-like1 and CmMAF-like2 transcription (Fig. 6d, g). Electrophoretic mobility shift assays (EMSAs) confirmed that recombinant purified CmDREBa fused to glutathione S-transferase (GST) binds to the DRE/CRT elements in the CmMAF-like1 and CmMAF-like2 promoters (Fig. 6e, h). Furthermore, RT-qPCR analysis showed that CmMAF-like1 and CmMAF-like2 expression is upregulated in CmDREBa-OX CI chrysanthemum plants and downregulated in Cmdreba mutants relative to the WT (Fig. 7a–d). Together, these results demonstrate that CmDREBa can activate the expression of CmMAF-like1 and CmMAF-like2. Expression of CmMAF-like genes in response to low temperature. Relative CmMAF-like1 (a) and CmMAF-like2 (b) transcript levels in WT (CI chrysanthemum) plants and CmDREBa overexpression lines (CI-OX#1 and CI-OX#3) in CI chrysanthemum, as determined by RT-qPCR. Results are shown as means ± SD, n = 4 independent biological replicates. Relative CmMAF-like1 (c) and CmMAF-like2 (d) transcript levels in WT plants and Cmdreba biallelic mutants (CI-dreba#12 and CI-dreba#19) in CI chrysanthemum, as determined by RT-qPCR. Results are shown as means ± SD, n = 3 independent biological replicates. e In situ hybridization of CmMAF-like transcripts. Plants were subjected to 0, 3, or 6 weeks at low temperature (LT0–LT6) before being returned to normal temperature (22 °C) under LD conditions for 3 weeks (LT6 + 3). The red square highlights a zoomed-in view of the in situ signals in the rib zone. A sense CmMAF-like probe was used as a negative control. Scale bar, 100 μm. Relative CmMAF-like1 transcript levels in CD (f) or CI (g) chrysanthemum plants exposed to different treatments. Relative CmMAF-like2 transcript levels in CD (h) or CI (i) chrysanthemum exposed to different treatments. In f–i, results are shown as means ± SD, n = 4 independent biological replicates in CD chrysanthemum or 3 independent biological replicates in CI chrysanthemum. Relative expression of CmMAF-like1 (j) and CmMAF-like2 (k) in CD and CI chrysanthemum, as determined by RT-qPCR. Shoot apices were collected from CD and CI chrysanthemum grown at 22 °C under LD conditions for 1 week, when CD plants exhibited rosette growth, and CI plants showed extension growth. Results are shown as means ± SD, n = 3 independent biological replicates. In a–d, j, k, significant differences between CI-WT and CI-OX lines, CI-WT and CI-dreba lines, or CD and CI chrysanthemum were determined by two-tailed Student’s t tests. In f–i, significant differences were determined by one-way ANOVA analysis followed by Tukey’s multiple comparison test. Data were log10-transformed to achieve a normal distribution and homogeneity of variance prior to analysis. Source data are provided as a Source data file. CmMAF-like genes inhibit extension growth and flowering Because CmMAF-like1 and CmMAF-like2 are direct targets of CmDREBa, we investigated their functions in extension growth and flowering in chrysanthemum. CmMAF-like1 and CmMAF-like2 are highly similar (93.10% identity), and we therefore used a conserved fragment of coding sequence as a probe to detect the expression of both genes in shoot apices by in situ hybridization (Supplementary Fig. 16). CmMAF-like1 and CmMAF-like2 transcripts were more abundant in unchilled (LT0) than in chilled apices (LT6 + 3), particularly within the rib zone (Fig. 7e). RT-qPCR showed that CmMAF-like1 and CmMAF-like2 exhibit similar expression patterns (Fig. 7f–i), with CD chrysanthemum showing significantly higher CmMAF-like1 and CmMAF-like2 expression levels than CI chrysanthemum under normal growth conditions (Fig. 7j, k). We next made use of their highly similar coding sequences (Supplementary Fig. 16) to simultaneously silence both CmMAF-like1 and CmMAF-like2 via RNAi in CD chrysanthemum (MAF-like-RNAi). We confirmed the lower expression of CmMAF-like1 and CmMAF-like2 in the MAF-like-RNAi lines by RT-qPCR (Fig. 8a, b), whereas the expression of the other three MAF genes in the gene cluster was unchanged (Supplementary Fig. 17). The MAF-like-RNAi CD chrysanthemum plants were significantly taller than WT plants (Fig. 8c, d) and were able to bloom without chilling (Fig. 8c, e). We also investigated whether the expression of selected DEGs identified in CmDREBa-OX plants was also altered in MAF-like-RNAi plants. We selected three DEGs based on their predicted roles in cell division, flowering, or stress tolerance: CmMYB3R-like36,37, ETHYLENE RESPONSE FACTOR 105 (CmERF105)38,39, and CmMYC240. RT-qPCR revealed that their expression differed significantly between MAF-like-RNAi plants and WT plants (Supplementary Fig. 18). When we overexpressed CmMAF-like1 in CI chrysanthemum (MAF-CI-OX, Fig. 8f), the resulting MAF-CI-OX plants were significantly shorter than WT plants (Fig. 8g, h). Under SD conditions, there were no visible flower buds on the MAF-CI-OX plants by the end of the experiment (100 days after transplant) (Fig. 8g, i). Extension growth and flowering of CmMAF-like transgenic chrysanthemum. Relative expression of CmMAF-like1 (a) or CmMAF-like2 (b) in WT plants and MAF-like-RNAi lines in CD chrysanthemum, as determined by RT-qPCR. Results are shown as means ± SD, n = 3 independent biological replicates. c Representative photograph of WT (CD chrysanthemum) and MAF-like-RNAi CD chrysanthemum. Plants were cultivated at normal temperature for 6 weeks under LD conditions before being transferred to SD conditions. The plants shown were photographed 78 days after transfer to SD conditions. Scale bar, 5 cm. d Plant height of WT and MAF-like-RNAi CD chrysanthemum. e Days to visible capitulum buds in WT and MAF-like-RNAi CD chrysanthemum. Days were recorded following the transfer to SD conditions. nb no bloom. Results are shown as means ± SD, n = 5 independent biological replicates (d, e). f Relative expression of CmMAF-like1 in WT, CmMAF-like1-overexpressing CI chrysanthemum lines (MAF-CI-OX#7 and MAF-CI-OX#8), a CmDREBa-overexpressing CI chrysanthemum line (DREBa-CI-OX#3), and CmDREBa-overexpressing CI chrysanthemum lines in which MAF-like expression is suppressed by RNAi (MAF-RNAi DREBa-OX#1, MAF-RNAi DREBa-OX#2), as determined by RT-qPCR. Results are shown as means ± SD, n = 3 independent biological replicates. g Representative photograph of WT and different transgenic CI chrysanthemum lines. Plants were cultivated at normal temperature for 4 weeks under LD conditions and then transferred to SD conditions. Scale bar, 5 cm. h Plant height of WT and different transgenic CI chrysanthemum lines. i Days to visible capitulum buds in WT and different transgenic CI chrysanthemum lines. nb no bloom. In h, i, results are shown as means ± SD, n = 5 independent biological replicates. In a, b, d, e, significant differences between WT and MAF-CD-RNAi lines were determined by two-tailed Student’s t tests. In f, h, different lowercase letters indicate significant differences determined by one-way ANOVA analysis followed by Tukey’s multiple comparison test; exact P values and test statistics are provided in Source data. Data were log10-transformed to achieve a normal distribution and homogeneity of variance prior to analysis (f). Source data are provided as a Source data file. To obtain genetic evidence that CmDREBa influences extension growth and flowering by regulating the expression of CmMAF-like1 and CmMAF-like2, we knocked down CmMAF-like1/2 in CmDREBa-overexpressing CI chrysanthemum (MAF-RNAi DREBa-OX, Fig. 8f). The shorter plant stature and flowering repression associated with CmDREBa overexpression were partially suppressed by knockdown of CmMAF-like1/2, returning to levels close to or similar to those of WT plants (Fig. 8g–i). Discussion Vernalization is defined as the acquisition or acceleration of flowering competence through prolonged exposure to cold temperatures11,41. As a critical phenological adaptation, vernalization exhibits significant variation between annual and perennial Brassicaceae species42. In particular, the shift between perennial and annual flowering strategies in Brassicaceae is largely governed by the dosage-dependent effects of FLC and related MAF genes24. However, many herbaceous perennials, such as chrysanthemum and strawberry (Fragaria × ananassa), have evolved a different type of perennial growth habit5,8. In this case, prolonged chilling does not trigger a typical vernalization response; instead, it also influences vegetative growth patterns. The CD chrysanthemum cultivar used in this study serves as an excellent model for studying perennial growth habits. It naturally grows as a rosette and only breaks from this state after prolonged cold exposure. Our findings demonstrate that CD chrysanthemum loses its capacity for elongation and flowering in the rosette state, although leaf production persists (Fig. 1a–f). This finding suggests that the rosette state of chrysanthemums is quantitative and aligns more closely with a state of semi-dormancy, as described previously8. Dormancy in perennials is an adaptive strategy for withstanding harsh winter conditions that enhances seasonal stress tolerance and resilience. Abiotic stress response pathways are therefore likely to contribute to dormancy. In the woody plant hybrid aspen (Populus tremula × Populus alba), SD conditions induce the expression of genes encoding abscisic acid (ABA) receptors and elevate levels of ABA43, which is known to inhibit symplastic intercellular communication through plasmodesmata to establish dormancy44. In this study, we characterized the critical role of CmDREBa in regulating extension growth and flowering in response to prolonged chilling. Previous studies have shown that CmDREBa expression is induced by short-term chilling in chrysanthemum45, and our observations reveal that CmDREBa expression is induced during prolonged chilling (Fig. 3b, c), suggesting that it plays a role in regulating seasonal growth in chrysanthemum. Expression and functional analyses demonstrated that CmDREBa expression levels are associated with competence for extension growth and flowering. Because CmDREBa expression is significantly lower in CI chrysanthemum than in CD chrysanthemum (Fig. 4b), we reasoned that the rosette phenotype might be linked to elevated CmDREBa expression. This hypothesis was supported by the observation that overexpression of CmDREBa in CI chrysanthemum indeed produced a rosette phenotype (Fig. 4d and Supplementary Fig. 8). By contrast, Cmdreba mutants in the CI background, which already exhibit very low endogenous CmDREBa expression under normal growth conditions, showed no significant change in plant height but did display an early-flowering phenotype (Fig. 4h, i). Nevertheless, these mutants still exhibited an early-flowering phenotype, which may be attributable to late-stage alterations in CmDREBa expression in CI chrysanthemum. Our data also suggest that differences in H3K36me3 levels may help explain the differences in CmDREBa expression between CI and CD chrysanthemum (Supplementary Fig. 7), although this possibility requires experimental assessment. A potential connection between freezing tolerance and vernalization has been suggested in previous works. In einkorn wheat (Triticum monococcum), mutants of the major vernalization gene VRN-1 exhibited diminished freezing tolerance and lower expression of several DREB genes46. In Arabidopsis, DREB1a (also reported as CBF3) induces the expression of COOLAIR, a cold-induced antisense RNA transcribed from the FLC locus47. Here, analysis of the gene network downstream of CmDREBa using DAP-seq and RNA-seq suggested that CmDREBa is extensively involved in regulating both vegetative and reproductive growth (Fig. 6b). We demonstrated that CmDREBa regulates the expression of MAF-like genes (Fig. 6c–h). In Arabidopsis, the MAF gene family, the members of which are closely related to FLC, also plays a role in vernalization48. In chrysanthemum, previous research has shown that CmMAF2 regulates flowering time in response to low temperatures21, suggesting a conserved function for MAFs in chilling-induced flowering. Knockdown of CmMAF2 resulted in taller plants and accelerated flowering time21, a phenotype similar to that of MAF-like-RNAi chrysanthemum plants generated in our study (Fig. 8). An FLC-like gene was recently identified in chrysanthemum and shown to participate in the ambient temperature–mediated flowering pathway49. Nevertheless, our transcriptome analysis indicated that the expression of this FLC-like gene was not significantly altered in response to CmDREBa manipulation. Together, our results demonstrate that CmDREBa regulates extension growth and flowering in response to low temperature, at least in part, by modulating the expression of MAFs. Methods Plant materials and growth conditions C. morifolium cv. “Fenkou” was used as a representative CD chrysanthemum, and the wild chrysanthemum C. indicum as a representative CI chrysanthemum. Plants were propagated by tissue culture on half-strength Murashige and Skoog (MS) medium for 5 weeks and then transplanted into pots (9-cm diameter, 1:1 [v/v] peat and vermiculite) and grown in growth chambers (22 ± 1 °C, 30–40% relative humidity) under a long-day (LD) photoperiod (16-h light/8-h dark, 100–120 μmol m−2 s−1 illumination provided by fluorescent lamps [SINOL, SN-T5, 16 W]). Low-temperature treatments were carried out at 4 °C under LD conditions. To accelerate flowering, plants were transferred to a growth chamber (22 ± 1 °C, 30–40% relative humidity) under a short-day (SD) photoperiod (8-h light/16-h dark). Spatial transcriptome sequencing and data analysis Spatial transcriptome sequencing was performed on the BMKMANU S1000 platform. Samples were embedded and sectioned at a 10-µm thickness using a pre-cooled cryostat50. Toluidine blue staining and tissue permeabilization were performed according to the manufacturer’s instructions. Libraries were prepared with a TruSeq Library Prep Kit (Illumina) and sequenced on the NovaSeq platform (Illumina) to a minimum sequencing depth of 150,000 read pairs per spatial spot by IntegraGen (Evry, France). Raw data were analyzed with BSTMatrix v2.3 (http://www.bmkmanu.com/portfolio/tools) and aligned to the Chrysanthemum seticuspe reference genome51 using STAR genome aligner v.2.5.1b. Image adjustments were made using BSTViewer (v1.42), and the corresponding level 13 matrix (100 μm) was used for downstream analyses. Data normalization, clustering, and marker gene identification were performed using the Seurat R package (v4.0.1), followed by marker gene annotation and functional enrichment analysis using clusterProfiler (v3.18.1). RNA extraction and RT-qPCR analysis Total RNA was isolated from terminal buds (two buds per sample) using TRIzol reagent (TaKaRa, Japan). First-strand cDNA was reverse transcribed from 1 µg of total RNA per sample using HiScript II Q Select RT SuperMix (Vazyme). qPCR was performed with 1 μL of cDNA template using a KAPATM SYBR quantitative PCR Kit (Kapa Biosystems) on a Step One Plus real-time PCR system (Applied Biosystems). The chrysanthemum Ubiquitin gene (GenBank accession EU862325) was used as the reference gene for normalization. The primers used are listed in Supplementary Data 8. In situ hybridization Terminal buds were vacuum infiltrated for 2 h in formaldehyde-alcohol-acetic acid (FAA) solution consisting of 3.7% (w/v) formaldehyde, 5% (v/v) acetic acid, and 50% (v/v) ethanol. Samples were then transferred to fresh FAA solution and fixed at 4 °C overnight. Antisense and sense probes were generated using T7 and SP6 RNA polymerases, respectively, with specific primers using the DIG RNA Labeling Kit (Cat. No. 11175025910, Sigma-Aldrich) according to the manufacturer’s protocol. After post-hybridization washes and RNase A treatment to eliminate unhybridized probes, the sections were blocked and incubated with an anti-DIG-AP antibody (1:1250; sheep Fab fragments, Cat. No. 11093274910, Roche) at room temperature for 2 h. Signal detection was performed by incubating the sections in NBT (Cat. No. 11383213001, Roche) / BCIP (Cat. No. 11383221001, Roche) substrate solution in the dark until visual signals appeared52. The primers used are listed in Supplementary Data 8. Chrysanthemum transformation To generate the CmDREBa and MAF-like RNAi constructs, specific 329-bp (CmDREBa) or 332-bp (MAF-like1/2) sense and antisense fragments were amplified using two primer sets with either AscI/SwaI or PacI/BamHI sites present at primer ends and then inserted into the pFGC1008 vector downstream of the constitutive cauliflower mosaic virus (CaMV) 35S promoter. To construct the CmDREBa and CmMAF-like1 overexpression constructs, the full-length CmDREBa or CmMAF-like1 coding sequence in frame with that of GFP was digested by XbaI/KpnI and cloned into the pBI121 vector driven by the CaMV 35S promoter. To construct the CRISPR/Cas9 construct, three single guide RNAs targeting CmDREBa were designed using CRISPR-P 2.0 (http://crispr.hzau.edu.cn/cgi-bin/CRISPR2/SCORE) and inserted into the pDIRECT-Ci-opti vector53. The primers used are listed in Supplementary Data 8. Plasmids were introduced into Agrobacterium (Agrobacterium tumefaciens) strain EHA105, and transgenic chrysanthemum plants were generated by Agrobacterium-mediated transformation29. DAP sequencing and data analysis Genomic DNA from C. indicum was sonicated to obtain fragments of 100–400 bp using the Bioruptor Plus system (Diagenode). The fragmented DNA was end-repaired, followed by the addition of a 3′ A-tail and adaptor ligation. Recombinant CmDREBa was produced as a fusion protein containing the HaloTag sequence and then purified using HaloTag ligand-conjugated magnetic beads (Promega, USA). The genomic DNA library was incubated with recombinant purified Halo-CmDREBa at room temperature for 1 h. DNA that specifically bound to Halo-CmDREBa was then eluted, recovered, and amplified by PCR. The enriched DNA was sequenced on the Illumina NovaSeq 6000 platform. After sequencing, clean DAP-seq reads were aligned to the C. indicum genome35. Peak-calling analysis was performed to identify DNA fragments bound by CmDREBa across the genome. Reads spanning an additional 2 kb upstream and downstream of the transcription start site (TSS) and transcription end site (TES) were quantified using deepTools (version 3.5.12.0)54. Peaks with a q < 0.05 were identified using MACS2 (version 2.2.7.1)55. Enriched peaks were visualized using the Integrative Genomics Viewer (IGV, version 2.8.2)56, and motif discovery was performed using Homer (version 4.11)57. The experiment was performed with two biological replicates. RNA sequencing and data analysis Total RNA was extracted from three biological replicates of wild-type and transgenic plants and used for library preparation as described above. The RNA-seq libraries were sequenced on the Illumina HiSeq 2000 platform to obtain paired-end reads by Novogene Co., Ltd (Beijing, China). Raw sequencing reads were aligned to the reference genome using HISAT2 software. Differential expression analysis was performed using DESeq2, with significance thresholds set at fold change (FC) ≥ 2 and P ≤ 0.0558. Chromatin immunoprecipitation quantitative PCR (ChIP-qPCR) assay Terminal buds (0.2 g fresh weight) were harvested from 35S:CmDREBa-GFP transgenic and 35S:GFP transgenic chrysanthemum plants. ChIP was performed using EpiQuik Plant ChIP Kit (Cat No. 2409047, Epigentek) according to the manufacturer’s instructions. Chromatin DNA was sonicated to shear the DNA into fragments ranging from 300 to 500 bp. A rabbit polyclonal anti-GFP antibody (1:100; Cat No. 50430-2-AP, Proteintech) was used to immunoprecipitated the protein–DNA complexes. The qPCR results for the ChIP product were normalized to those for the input DNA to calculate DNA enrichment. Primers used are listed in Supplementary Data 8. Electrophoretic mobility shift assay (EMSA) EMSAs were performed using a Light Shift Chemiluminescent EMSA kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The full-length CmDREBa coding sequence was inserted into the pGEX-4T-2 vector; the resulting construct was introduced into the E. coli Rosetta strain. Production of the fusion protein was induced with 0.4 mM isopropyl β-D-1-thiogalactopyranoside in 200 mL LB medium at 16 °C for 12 h. The fusion protein was extracted and purified using glutathione agarose resin (Cat No. 78602, Thermo Scientific™). The biotin-labeled probes were designed as described in Supplementary Data 8. Dual-luciferase reporter assay in Nicotiana benthamiana Effector constructs for CmDREBa were created by cloning its full-length coding sequence into the pGreenII 0029 62-SK vector. The CmMAF-like1 and CmMAF-like2 promoters were cloned and separately inserted into the pGreenII 0800-LUC vector to generate reporter constructs. All constructs were introduced into the Agrobacterium strain GV3101 harboring the pMP90 and pSoup plasmids. Cells were resuspended in infiltration buffer (10 mM MgCl2, 200 mM acetosyringone, 10 mM MES, pH 5.6) to an OD600 of 1.0, and then infiltrated into young leaves of N. benthamiana plants. After 3 days, the activities of firefly luciferase (LUC) and Renilla luciferase (REN) were measured using dual-luciferase reporter assay reagents (Cat No. E1910, Promega) on a GloMax 20/20 luminometer (Promega). The primers used are listed in Supplementary Data 8. Statistical analyses Quantitative data are presented as mean ± SD. Statistical analyses were performed using GraphPad Prism 10. Statistical significance between two groups was determined using a two-tailed Student’s t test, whereas comparisons among multiple groups were performed using one-way ANOVA followed by Tukey’s multiple comparison test. In the figures, different letters indicate statistically significant differences. Exact P values and test statistics are provided in the Source data, and detailed statistical information is shown in the figure legends. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Supplementary Data 7 Supplementary Data 8 Reporting Summary Transparent Peer Review File Source data Source data