BRANCHED1 acts as a rheostat of Nuclear Factor Y-dependent epigenetic regulation in axillary bud development

Abstract

Shoot branching exhibits significant plasticity and is largely regulated by BRANCHED1 (BRC1), a key integrator of endogenous and environmental cues that acts locally to control axillary bud development. However, the gene regulatory networks governed by BRC1 and the mechanisms underlying its regulation remain poorly understood. Here, we identify a cluster of transcription factor-encoding genes that are dynamically responsive to multiple branching pathways, inversely correlated in expression with BRC1 levels, and functionally enriched in pathways related to floral organ development (FOD) and shoot system morphogenesis. Mechanistically, the heterotrimeric Nuclear Factor Y (NF-Y) complex is required for transcriptional activation of representative FOD genes, functioning through chromatin looping and H3K27me3 demethylation to establish active chromatin states at target loci. In contrast, BRC1 inhibits NF-Y complex assembly and its epigenetic regulatory effects in a dose-dependent manner, thereby enabling tunable FOD gene expression and plastic bud development. These findings position BRC1 as a molecular rheostat of NF-Y-dependent epigenetic regulation, providing mechanistic insights into the dynamic control of axillary bud development.

Shoot branching plasticity is largely regulated by BRC1. The authors show that BRC1 acts as a dose-dependent rheostat, modulating NF-Y-mediated chromatin looping and H3K27me3 demethylation at floral organ development genes to fine-tune axillary bud activity.

Introduction

Shoot branching is a critical agronomic trait that directly influences crop yield by determining the number of branches capable of producing flowers, fruits, or seeds. It is also an adaptive trait that enables plants to adjust their growth patterns or morphology in response to varying environmental conditions (e.g., light, temperature, and nutrient availability)^1–4. Optimizing branching patterns, a core aspect of ideal plant architecture, is considered an effective strategy for improving photosynthetic efficiency, biomass production, and overall plant fitness, particularly under suboptimal field conditions^1,4,5. A prominent example of crop improvement through branching optimization is the domestication and breeding of modern maize, in which selection for a high-expression allele of the teosinte branched1 (tb1) gene led to reduced axillary branches compared to its wild ancestor, teosinte⁶.

Shoot branching generally arises from two developmental processes: the initiation of axillary meristems in leaf axils and the subsequent development and growth of axillary buds into branches^7,8. In contrast to the genetically determined initiation of axillary meristems, the development of axillary buds in angiosperms is highly dynamic and plastic, influenced by an intricate interplay of diverse branching signals, including phytohormones, nutrients, and environmental factors^3,8–12. Phytohormone homeostasis and signaling, primarily involving auxin, strigolactones (SLs), cytokinins (CKs), and gibberellins (GAs), play pivotal roles in regulating axillary bud development^13,14. Auxin is considered a long-distance signal produced at the shoot apex and transported basipetally to indirectly suppress the outgrowth of lower buds, thereby establishing apical dominance^3,10,11. SLs and CKs act as local signals in axillary buds, antagonistically regulating bud outgrowth by controlling the polar distribution of auxin transporters and the expression of branching inhibitor BRANCHED1 (BRC1)^3,10,15–18. Sugars are proposed as both an energy source and an early signal, acting prior to phytohormones in initiating bud growth, partly by antagonizing SL signaling^19–25. Nutrients also exert an indispensable effect on shoot branching through concerted actions with phytohormones. This is evident from the inhibition of bud outgrowth under low nitrogen or phosphate conditions, which is at least partially attributed to enhanced SL production or signaling^12,20,26. Additionally, environmental cues, such as changes in light quality (e.g., the red/far-red light ratio or shade), photoperiod, and temperature, are implicated in regulating bud activity, probably through coordination with SL and abscisic acid (ABA) pathways^27–33.

Arabidopsis BRC1, a plant-specific TCP domain transcription factor (TF) with orthologs in many species (e.g., maize ZmTB1, pea PsBRC1, and rice FINE CULM1/OsTB1), is widely recognized as a key integrator of branching signals, acting locally within axillary buds to inhibit their outgrowth^{3,10,12,34–36}. BRC1/TB1 expression is highly responsive to a variety of phytohormones, nutrients, and environmental cues, partially explaining the roles of these signals in shoot branching^{18,25,29,34,36–40}. In addition to its transcriptional responses, BRC1/TB1 activity is modulated at the protein level through interactions with other proteins. In Arabidopsis, TIE1 and TCP15 counteract BRC1-mediated transcriptional activation of HD-ZIP encoding genes (e.g., HB21, HB40, and HB53) and the downstream ABA biosynthetic gene NCED3, leading to reduced ABA accumulation and increased branching^41–43. In rice, the paralog OsTB2 counteracts the inhibitory effect of OsTB1 on OsMADS57, which consequently reduces expression of the SL receptor gene DWARF14 (D14) and promotes tillering^44,45. Additionally, BRC1/TB1 interacts with and suppresses the florigen protein FT, thereby influencing floral transition, seasonal growth, inflorescence architecture, or aerial tuberization in axillary branches across various species^27,46–49.

Although prior studies have consistently demonstrated a strong negative correlation between BRC1 expression/activity and axillary bud outgrowth, more recent evidence complicates this relationship. Specifically, it has been shown that axillary buds lacking BRC1 expression can still remain inhibited, whereas buds with high BRC1 expression can become active^34,50. To reconcile these observations, a bud activation threshold hypothesis for BRC1 action has been proposed, whereby BRC1 does not function as an absolute on/off switch for bud activity but instead modulates the bud’s potential to respond to additional regulatory signals or pathways^1,3,34. This hypothesis suggests the presence of a sophisticated and hierarchical gene regulatory network (GRN) that integrates BRC1 with other branching regulators to determine bud activity. Several multi-omics analyses have identified a low-light-responsive GRN in which BRC1/TB1 primarily functions as a transcriptional activator, directly upregulating target genes in the ABA pathway to promote axillary bud dormancy^40,43,51,52. Despite these advances, current findings fall short of fully elucidating the molecular mechanisms underlying BRC1 action in the bud activation threshold hypothesis. Moreover, they leave open the possibility that distinct GRNs may be engaged in response to specific branching signals and that BRC1 may function as a transcriptional repressor within certain GRNs.

To investigate these possibilities, we conducted an integrative analysis of SL-responsive, decapitation-induced, and BRC1-dependent transcriptomes. This analysis identified a cluster of TF genes inversely correlated in expression with BRC1 levels and enriched for functions in floral organ development (FOD) and shoot system morphogenesis. We demonstrate that transcriptional activation of representative FOD genes requires the heterotrimeric Nuclear Factor Y (NF-Y) complex, which establishes active chromatin states at target loci by directly binding their upstream regions to mediate chromatin looping and H3K27me3 demethylation. We also show that BRC1 specifically interacts with multiple NF-YC members, disrupting NF-Y complex assembly and its epigenetic regulatory effects, thus maintaining transcriptional silencing at FOD gene loci. Altering BRC1 levels through overexpression or decapitation leads to dose-dependent changes in H3K27me3 deposition at FOD gene loci, enabling tunable expression of these targets in axillary buds. Together, our findings reveal a dynamic regulatory mechanism in which BRC1 functions as a molecular rheostat to modulate NF-Y-dependent epigenetic regulation of target genes in a dose-dependent manner, thereby integrating diverse branching signals in the control of developmental plasticity of axillary buds.

Results

Strigolactone signaling represses FOD gene expression through BRC1 upregulation

In Arabidopsis, a majority of axillary buds (>60%) typically remain dormant throughout the entire life-cycle, primarily due to the action of strigolactones (SLs), a class of carotenoid-derived phytohormones that play a crucial role in suppressing shoot branching^36,53,54. Previous transcriptomic analyses identified a few early SL-responsive genes (responsive within 4 h) involved in shoot branching, including SL biosynthetic genes MORE AXILLARY GROWTH 3/4 (MAX3/4), SL signaling repressors SUPPRESSOR OF MAX2-LIKE 6/7/8 (SMXL6/7/8), and branching-related TFs BRC1 and HB21/40^37,55–57. However, these early response genes are unlikely to fully account for the sustained dormancy of axillary buds mediated by SLs, prompting us to search for late or sustained SL-responsive genes involved in long-term bud inhibition. To this end, we performed RNA sequencing (RNA-seq) analysis on wild-type Col-0 (WT) seedlings grown for 7 days on agar medium supplemented with 2.5 μM rac-GR24 (a synthetic SL analog active in multiple SL-responsive pathways) or solvent control dimethyl sulfoxide (DMSO). Compared to the 401 differentially expressed genes (DEGs) previously identified within 4 h of SL treatment³⁷, we detected a total of 5354 DEGs (2764 upregulated and 2590 downregulated) after 7 days of rac-GR24 treatment (Supplementary Data 2), indicating that a large number of genes exhibit sustained responsiveness to SL treatment. Notably, approximately 43.75% (91 out of 208) of early upregulated genes and 35.75% (69 out of 193) of early downregulated genes were also identified in our RNA-seq data (Supplementary Fig. 1a, b), indicating that many early response genes maintain prolonged responsiveness to SL treatment.

TFs are at the center of regulatory networks that drive gene expression reprogramming to underpin a wide range of biological responses⁵⁸. To delineate the core GRN associated with sustained SL responses, the differentially expressed TFs (228 upregulated and 214 downregulated) were specifically selected for Gene Ontology (GO) and heatmap analyses (Supplementary Data 2). SL-upregulated TFs were predominantly enriched in heat and auxin response pathways (Fig. 1a, c), consistent with SLs’ roles in heat tolerance and crosstalk with auxin signaling^37,55,59. Intriguingly, SL-downregulated TFs showed significant enrichment in FOD and pattern specification processes (Fig. 1b, d). Given the limited understanding of SL-repressed GRNs in branching regulation, we then concentrated on these SL-downregulated TFs. To further determine the temporal expression patterns of representative SL-downregulated FOD genes, we employed real-time quantitative PCR (RT-qPCR) analysis to measure the transcript levels of FRUITFULL (FUL) and its closely related homolog CAULIFLOWER (CAL) at multiple time points following rac-GR24 treatment. Unlike the early response gene BRC1, FUL and CAL transcript levels showed no significant changes after 1 day of rac-GR24 treatment but were substantially downregulated after 3 days of treatment, with even stronger downregulation observed at 7 days (Fig. 1e–g), demonstrating that FUL and CAL are late and sustained SL-downregulated genes.

Fig. 1. Strigolactone signaling represses FOD gene expression through BRC1 upregulation.

Gene Ontology (GO) analysis of transcription factors (TFs) identified as upregulated (a) or downregulated (b) in wild-type (WT) seedlings treated with 2.5 μM rac-GR24 (compared to solvent control DMSO) for 7 days, assessed by RNA-seq (n = 3 biological replicates; p value < 0.05, fold-change [FC] > 1.2). The top 10 enriched biological processes are ranked by the count of associated TFs. Color scale represents −log₁₀ adjusted p values determined using a one-sided hypergeometric test with Benjamini–Hochberg correction; darker red indicates higher statistical significance. Heatmaps showing expression patterns of upregulated (c) or downregulated (d) TFs in response to rac-GR24 treatment. Each column represents a replicate and each row corresponds to a gene. Color scale represents normalized Z–scores, with darker red indicating higher expression. Genes with similar expression patterns are grouped by hierarchical clustering. Representatives of upregulated TFs or downregulated FOD genes are listed. e–g FOD genes exhibit a late and sustained transcriptional response to rac-GR24 treatment. Relative expression levels of BRC1 (e), FUL (f), and CAL (g) were measured in WT seedlings treated with 2.5 μM rac-GR24 or DMSO (mock) for 1, 3, or 7 days. h–j SL signaling pathway is required for repressing FOD gene expression under rac-GR24 treatment. Relative expression levels of BRC1 (h), FUL (i), and CAL (j) were measured in WT, d14-1, max2-3, max3-11, and brc1-2 seedlings treated with rac-GR24 or DMSO (mock) for 7 days. k, l Upregulation of BRC1 strongly represses FOD gene expression. Relative expression levels of FUL (k) and CAL (l) were measured in 7-day-old seedlings of WT and two independent BRC1 overexpression lines (BRC1-OE2 and BRC1-OE12, expressing BRC1 under the constitutive CaMV 35S promoter). For RT-qPCR assays, data are means ± s.d., with each circle representing a biological replicate (n = 4). All P values were determined by unpaired two-tailed Student’s t-tests, with P < 0.05 considered statistically significant.

To determine whether rac-GR24-mediated downregulation of FOD genes depends on the SL signaling pathway, we measured FUL and CAL expression in SL biosynthetic mutant max3-11 and signaling mutants d14-1, max2-3, and brc1-2. RT-qPCR analysis showed that rac-GR24-mediated repression of FUL and CAL was abolished in d14-1, max2-3, and brc1-2 mutants, but restored to wild-type levels in max3-11 (Fig. 1h–j and Supplementary Fig. 1c–e), suggesting that an intact SL signaling pathway is required for repressing these FOD genes. The insensitivity of the brc1 mutant to rac-GR24-mediated repression of FOD genes led us to hypothesize that SL-induced upregulation of BRC1 is responsible for the downregulation of FOD genes. To test this, we measured FUL and CAL expression in two independent BRC1 overexpression lines (BRC1-OE2 and BRC1-OE12, expressing full-length BRC1 under the constitutive CaMV 35S promoter). As expected, both FUL and CAL were strongly repressed in the BRC1 overexpression lines (Fig. 1k, l). Taken together, these results suggest that SL signaling represses FOD gene expression by upregulating BRC1.

Decapitation induces FOD gene expression through BRC1 downregulation

The developmental plasticity entails that dormant buds can be reactivated to produce branches under permissive conditions (e.g., decapitation, low planting density, or high sugar availability)^25,33,36,60, usually coinciding with BRC1 downregulation^{20,25,36,38,40}. After the shoot apex removal, the uppermost primary rosette (R1) branch escapes basipetal inhibition from the apical bud and attains a dominant position in the branching hierarchy, making it an ideal model for assessing the extent to which decapitation-induced bud release correlates with BRC1 downregulation. We monitored the elongation of the uppermost R1 branch in WT and brc1-2 plants daily after decapitation treatment (Fig. 2a). Our measurements showed that decapitation rapidly induces elongation of the uppermost R1 branch in WT (at a rate comparable to that in intact brc1-2 mutant), but has no significant impact on brc1-2 mutant (Fig. 2b, c), suggesting that decapitation-induced bud release is largely related to BRC1 downregulation, particularly in the uppermost R1 bud adjacent to the main shoot. To identify the core GRN underlying decapitation-induced bud release and its connection with BRC1 downregulation, we conducted RNA-seq analysis on axillary tissues (including axillary buds and subtending rosette leaf axil tissues) of decapitated WT and intact brc1-2 plants, with intact WT plants as control (Supplementary Data 3). Venn diagram revealed that 60.47% of differentially expressed TFs in decapitated WT (387 out of 640) overlapped with those in intact brc1-2 mutant (387 out of 612; Fig. 2d), indicating that decapitation-induced transcriptional changes are closely related to BRC1 downregulation.

Fig. 2. Decapitation induces FOD gene expression through BRC1 downregulation.

a Schematic of Arabidopsis shoot apex decapitation showing the uppermost primary rosette (R1) branch used for length measurement and axillary tissues (shaded area) harvested for gene expression analysis. b Representative images of intact or decapitated WT and brc1-2 plants at 7 days post-decapitation. Arrowheads indicate elongated R1 branches. Scale bar, 2 cm. c Lengths of the uppermost R1 branch in WT and brc1-2 plants measured daily for 14 days post-decapitation. Data are means ± s.e.m. (n = 10 plants). P values (for lengths at 14 days) were determined by unpaired two-tailed Student’s t-tests. d Venn diagram showing overlap of differentially expressed TFs in axillary tissues of decapitated WT and intact brc1-2 (compared to intact WT) and in seedlings of rac-GR24-treated WT (compared to DMSO-treated WT), assessed by RNA-seq (n = 3 biological replicates; p value < 0.05, FC > 1.2). Scatterplots of expression patterns for 145 overlapping TFs showing positive correlation between intact brc1-2 and decapitated WT (e), and negative correlations between rac-GR24-treated WT and decapitated WT (f) or intact brc1-2 (g). Each dot represents an individual TF plotted by Log₂FC values. Percentages in quadrants denote proportions relative to total overlapping TFs. h Heatmap with K-means clustering of 145 overlapping TFs. Box plots (left) show Z–scored expression values for TFs in each cluster (n = 9, 76, 7, 46, and 7 TFs for C1–C5, respectively); center lines indicate medians, box bounds represent the 25th and 75th percentiles, and whiskers extend to 1.5× the interquartile range. Representative C2 and C4 TFs are listed (right). i GO analysis of C4 TFs. The top 8 enriched biological processes are ranked by the count of associated TFs. Color scale represents adjusted p values determined using a one-sided hypergeometric test with Benjamini–Hochberg correction. j–l Decapitation-induced FOD gene expression correlates with BRC1 downregulation. Relative expression levels of BRC1 (j), FUL (k), and CAL (l) were measured in axillary tissues of WT and brc1-2 plants at 18 h or 36 h post-decapitation. Data are means ± s.d. (n = 5 biological replicates). P values were determined by unpaired two-tailed Student’s t-tests.

Having separately identified the SL-responsive and decapitation-induced GRNs, we sought to refine a co-regulated GRN that dynamically responds to multiple branching pathways. To this end, we conducted an integrative analysis of transcriptomes from axillary tissues of decapitated WT and intact brc1-2, as well as from seedlings of rac-GR24-treated WT. Venn diagram analysis revealed 145 overlapping TFs co-regulated across all three conditions (Fig. 2d and Supplementary Data 3). Pairwise scatterplot analysis showed that 90.34% (131 out of 145) of the overlapping TFs exhibited similar expression patterns (either co-upregulated or co-downregulated) in decapitated WT and intact brc1-2 (Fig. 2e), reinforcing the notion that decapitation-induced transcriptional changes are closely associated with BRC1 downregulation. In contrast, 81.38% (118 out of 145) and 85.52% (124 out of 145) of the overlapping TFs exhibited opposite expression patterns when comparing decapitated WT with rac-GR24-treated WT and intact brc1-2 with rac-GR24-treated WT, respectively (Fig. 2f, g). Therefore, these overlapping TFs constitute a core GRN co-regulated by decapitation and SL treatment in a BRC1-dependent manner.

To explore the possible biological functions of the co-regulated GRN, we performed K-means clustering analysis on the 145 overlapping TFs, dividing them into five distinct clusters (Fig. 2h and Supplementary Data 3). The 76 Cluster-2 (C2) genes exhibited the most significant enrichment in ABA-related pathways (Supplementary Fig. 2a). This finding aligns with prior evidence that ABA signaling acts downstream of BRC1 to regulate bud dormancy⁴³, and further suggests that both decapitation and SL signaling can influence bud activity through ABA signaling pathways. Notably, the 46 Cluster-4 (C4) genes were significantly enriched in pathways associated with FOD and shoot system morphogenesis (Fig. 2i), highlighting FOD genes as common targets dynamically regulated by both decapitation and SL treatment.

To further validate the relationship between decapitation-induced FOD gene expression and BRC1 downregulation, we measured FUL and CAL expression in axillary tissues of WT and brc1-2 plants after decapitation. RT-qPCR analysis showed that FUL and CAL expression in WT plants increased at 18 h post-decapitation, coinciding with significant BRC1 downregulation, but returned to basal levels at 36 h as BRC1 expression recovered (Fig. 2j–l). In contrast, FUL and CAL expression in brc1-2 remained elevated relative to WT and exhibited insensitivity to decapitation treatment (Fig. 2j–l). These results suggest that decapitation-induced FOD gene expression depends on BRC1 downregulation in axillary buds. To investigate the role of the SL signaling pathway in decapitation-induced FOD gene expression, we measured FUL and CAL expression in axillary tissues of d14-1, max3-11, and brc1-2 after decapitation. Similar to brc1-2, all mutants exhibited higher basal expression of FUL and CAL compared to WT; however, unlike the insensitivity observed in brc1-2, decapitation further elevated FUL and CAL expression in d14-1 and max3-11 mutants (Supplementary Fig. 2b–d). These findings suggest that decapitation-induced FOD gene expression is partially dependent on SL signaling, consistent with previous evidence that decapitation-induced BRC1 downregulation involves diverse signaling pathways^14,18,25.

BRC1 interacts with NF-YCs in axillary buds

Although previous studies have characterized BRC1 as a transcriptional activator of downstream target genes^40,43,51,52, we discovered its additional role as a transcriptional repressor of FOD gene expression (Fig. 1k, l). To elucidate the molecular basis of its repression function, we performed a yeast two-hybrid (Y2H) screening to identify BRC1-interacting proteins. Due to the strong self-activation of full-length BRC1 when fused to the GAL4 DNA-binding domain (Fig. 3a), we utilized an N-terminal truncated version of BRC1 (BRC1-DIV) as bait to screen an Arabidopsis cDNA library. This screening identified 11 candidate BRC1 interactors, including NF-YC3 and NF-YC4, which belong to the C subunit of the NF-Y complex (Supplementary Data 1). Given that the Arabidopsis genome encodes 13 NF-YC members⁶¹, we systematically tested BRC1 interactions with all of them. Y2H assay showed that, in addition to NF-YC3 and NF-YC4 identified in the initial screening, BRC1 interacts with NF-YC1, NF-YC2, NF-YC6, and NF-YC9 (Fig. 3b). The pull-down assay further confirmed that BRC1 directly interacts with multiple NF-YCs in vitro (Fig. 3c). To examine whether BRC1 interacts with NF-YCs in vivo, we conducted a bimolecular fluorescence complementation (BiFC) assay in Arabidopsis mesophyll protoplasts. Strong Venus fluorescence signals were observed in the nucleus when BRC1 was co-expressed with multiple NF-YCs (Fig. 3d), validating the in vivo interactions between BRC1 and NF-YCs. In addition, BRC2, a paralog of BRC1 with a partially redundant but less prominent role in axillary bud inhibition³⁶, also interacted with NF-YCs in the Y2H assay (Supplementary Fig. 3a).

Fig. 3. BRC1 interacts with NF-YCs in axillary buds.

a Domain mapping through Y2H assay showing strong self-activation of full-length BRC1 (FL) but not the N-terminal truncated variant (DIV). BRC1 or its truncated variants (DI to DV) were fused to the GAL4 DNA-binding domain (BD) and co-transformed with an empty GAL4 activation domain (AD). Transformed yeast cells were grown on non-selective SD−Trp/Leu (−TL) medium or selective SD−Trp/Leu/His medium supplemented with 10 mM 3-amino-1,2,4-triazole (−TLH + 3AT). Conserved TCP and R domains are indicated, with amino acid positions numbered. b Y2H assay demonstrating interactions between the N-terminal truncated BRC1 (DIV) and multiple NF-YCs. Yeast cells co-transformed with BD- and AD-fusion constructs were grown on selective SD−Trp/Leu/His medium containing 10 mM 3-AT. c Pull-down assay confirming direct interactions between BRC1 and NF-YCs in vitro. GST−NF-YCs were incubated with immobilized MBP or MBP−BRC1, followed by immunoblot analysis using anti-GST antibody (top panel). 5% of purified NF-YCs is shown as input control, and Ponceau S staining (Ponc.) indicates equal amounts of bait proteins (bottom panel). One representative experiment from three biological replicates is shown. d BiFC assay verifying interactions between BRC1 and NF-YCs in the nuclei of Arabidopsis mesophyll protoplasts. Venus panel shows protein interactions, and the chlorophyll panel indicates chloroplast autofluorescence. A representative result from three independent experiments is shown. Scale bars, 10 µm. e Y2H assay demonstrating the requirement of the R domain for BRC1 interactions with NF-YCs. Different AD-fused BRC1 variants were co-transformed with BD-fused NF-YCs, and yeast transformants were grown on selective SD−Trp/Leu/His medium containing 10 mM 3-AT. All Y2H assays were performed more than three times with consistent results. GUS histochemical staining of pNF-YC3::GUS plants (expressing GUS reporter driven by native NF-YC3 promoter) showing NF-YC3 expression activity in axillary buds subtended by rosette (f) and cauline (g) leaves. Black arrowheads indicate axillary buds. Representative images are shown from more than five independent plants. Scale bars, 1 mm.

BRC1, a class II TCP family TF, is characterized by a conserved TCP domain responsible for DNA binding and protein interactions, as well as an arginine-rich R domain of unknown function^36,62. To determine which domain of BRC1 mediates interaction with NF-YCs, we generated a series of truncated BRC1 variants and tested their interactions with NF-YCs. Y2H assays revealed that the R domain of BRC1 is essential for interaction with NF-YCs (Fig. 3e). Similarly, we generated truncated NF-YC3 fragments and demonstrated the conserved histone-fold domain (HFD) as the region responsible for interacting with BRC1 (Supplementary Fig. 3b).

BRC1 is specifically expressed in young axillary buds³⁶, prompting us to analyze the spatiotemporal expression patterns of NF-YCs in axillary shoots using pNF-YC3::GUS and pNF-YC4::GUS transgenic lines (expressing GUS reporter driven by native promoters of NF-YC3 and NF-YC4, respectively). Histochemical staining of GUS activity revealed that NF-YC3 and NF-YC4 are actively expressed in axillary buds at nearly all developmental stages (Fig. 3f, g and Supplementary Fig. 3c), showing substantial overlap with the axillary bud-specific expression pattern of BRC1. Together, these results demonstrate that BRC1 and its paralog BRC2 physically interact with multiple NF-YCs in axillary buds.

NF-YCs function downstream of BRC1 in regulating axillary bud development

NF-YCs are widely implicated in numerous aspects of plant development and stress responses, including photomorphogenesis, seed germination, flowering time, salt tolerance, and disease resistance^63–71. The observed interaction between BRC1 and NF-YCs raises the possibility that NF-YCs may play a role in axillary bud development. To test this hypothesis, we quantitatively analyzed axillary bud phenotypes in WT, brc1-2, triple nf-yc3 nf-yc4 nf-yc9 (yc349), and quadruple brc1-2 nf-yc3 nf-yc4 nf-yc9 (brc1-2 yc349) mutants at various time points. Under long-day (LD) conditions, yc349 and brc1-2 yc349 mutants exhibited delayed flowering relative to WT and brc1-2 (Supplementary Fig. 4a, b), consistent with the known role of NF-YCs in photoperiod-dependent flowering⁷⁰. After the flowering transition, axillary buds in yc349 and brc1-2 yc349 showed apparent growth retardation compared to WT and brc1-2, respectively (Fig. 4a and Supplementary Fig. 4c). After 60 days of growth under LD conditions, the number of elongated R1 branches in yc349 and brc1-2 yc349 was slightly reduced compared to WT and brc1-2, respectively, though brc1-2 yc349 still produced more R1 branches than WT (Supplementary Fig. 4c, d). At later stages, R1 buds in yc349 eventually grew out similarly to WT, as did those in brc1-2 yc349 relative to brc1-2 (Supplementary Fig. 4c). These observations indicate that loss of NF-YC function retards axillary bud growth but does not affect bud formation.

Fig. 4. NF-YCs function downstream of BRC1 in regulating axillary bud development.

a Representative images of WT, brc1-2, yc349 (nf-yc3 nf-yc4 nf-yc9), and brc1-2 yc349 plants grown for 50 days under long-day (LD) conditions. b Quantitative analysis of the number of rosette branches per rosette leaves (R1/RL) in different genotypes grown for 60 days under LD conditions. Data are means ± s.e.m. (n = 20 plants). Different letters above bars indicate significant differences (P < 0.05) determined using one-way ANOVA with Tukey’s HSD tests. c Developmental stages of buds in the axils of cotyledons (positions 1 and 2) and rosette leaves (positions 3 to 12) in different genotypes. Upper panel, stacked bar charts of bud developmental stages when the shoot apex reached 1 cm (n = 20 plants per genotype); lower panel, close-ups of the rosette base for each genotype. Bud developmental stages are defined as empty (E), meristem (M), leaf primordium (LP), vegetative (V1 to V3), or reproductive (R). d–f NF-YCs function downstream of BRC1 to regulate FOD gene expression in axillary buds. Relative expression levels of BRC1 (d), FUL (e), and CAL (f) were measured in axillary tissues of the indicated plants. Data are means ± s.e.m. (n = 4 biological replicates). Different letters above bars indicate significant differences (P < 0.05) determined using one-way ANOVA with Tukey’s HSD tests. g Representative images of intact or decapitated WT and yc349 plants at 7 days post-decapitation, with insets showing close-ups of the rosette base in intact (upper) or decapitated (lower) yc349 plants. Arrowheads indicate elongated R1 branches. Scale bar, 2 cm. h Lengths of the uppermost R1 branch in WT and yc349 plants measured on indicated days post-decapitation. Data are means ± s.e.m. (n = 8 plants). P values (for lengths at 14 days) were determined by unpaired two-tailed Student’s t-tests. i–k NF-YCs are required for decapitation-induced FOD gene expression in axillary buds. Relative expression levels of BRC1 (i), FUL (j), and CAL (k) were measured in axillary tissues of WT and yc349 plants at 18 h post-decapitation. Data are means ± s.e.m. (n = 4 biological replicates). P values were determined by unpaired two-tailed Student’s t-tests.

Since flowering time directly influences rosette leaf (RL) number and is therefore strongly correlated with R1 branch number³², we further quantified the R1/RL ratio across genotypes to minimize the confounding effects of delayed flowering in yc349 and brc1-2 yc349 mutants. Only 10% of R1 buds grew out in the yc349 mutant, significantly lower than in WT; similarly, approximately 20% of R1 buds grew out in the brc1-2 yc349 mutant, notably lower than in both WT and brc1-2 (Fig. 4b). These findings suggest that NF-YCs function downstream of BRC1 to regulate axillary bud growth. In addition, a lower proportion of R1 buds grew out in the d14-1 yc349 quadruple mutant compared to both WT and d14-1 (Supplementary Fig. 4e–g), suggesting that NF-YCs also function downstream of SL signaling to regulate axillary bud growth.

Following floral transition of the shoot apical meristem, axillary buds initiate and develop in a basipetal sequence, progressing from the uppermost leaf axil toward lower leaf axils^72,73. This orderly developmental progression is accelerated in brc1 loss-of-function mutants, in which most axillary buds prematurely develop to the reproductive phase shortly after bolting of the main shoot³⁶. In contrast, axillary bud development in yc349 was markedly delayed compared to WT, with nearly all buds remaining in the vegetative phase (Fig. 4c). Likewise, brc1-2 yc349 exhibited a developmental progression akin to that of yc349, albeit slightly advanced (Fig. 4c), indicating that the accelerated bud development characteristic of brc1-2 requires functional NF-YCs. Consistent with their early axillary bud phenotypes, the expression of FUL and CAL genes in axillary tissues was elevated in brc1-2 but reduced in both yc349 and brc1-2 yc349 (Fig. 4d–f). Together with the observed expression activity of FUL in axillary buds and its positive role in shoot branching (Supplementary Fig. 4h–j), these findings support the conclusion that NF-YCs function downstream of BRC1 to regulate FOD gene expression during axillary bud development.

Having shown the genetic relationship between NF-YCs and BRC1 in axillary bud development, we further investigated the role of NF-YCs in decapitation-induced bud release and FOD gene expression. Our measurements showed that elongation of the uppermost R1 branch was markedly delayed in decapitated yc349 compared with decapitated WT (Fig. 4g, h), suggesting that NF-YCs are required for decapitation-induced bud release, possibly by influencing the length of the lag phase during bud activation⁵⁰. Notably, the elongation rate in decapitated yc349 exceeded that in intact yc349 (Fig. 4g, h), potentially due to functional redundancy among other NF-YC members or the involvement of NF-YC-independent pathways. Consistent with these observations, RT-qPCR analysis showed that decapitation-induced expression of FUL and CAL in axillary tissues of yc349 was significantly lower than in WT, despite a slight induction of these genes in yc349 after decapitation (Fig. 4i–k), supporting the necessity of NF-YCs for decapitation-induced FOD gene expression.

NF-Y mediates H3K27me3 demethylation and chromatin looping at FOD gene loci

NF-YC, together with NF-YA and NF-YB subunits, forms the conserved NF-Y heterotrimeric TF, which is found in nearly all eukaryotes. NF-YB and NF-YC dimerize through their HFDs to create a nucleosome-like structure, into which NF-YA subsequently assembles to complete the heterotrimer, enabling sequence-specific recognition of CCAAT boxes within promoters and enhancers^64,65,67,68. In animals, NF-Y is widely recognized as a pioneer factor that establishes permissive chromatin conformations, facilitating the recruitment of master TFs or basal transcription machinery to target regions^74–77. In plants, although its pioneer activity remains to be conclusively demonstrated, NF-Y is well characterized as an epigenetic mediator that recruits various histone-modifying enzymes (e.g., REF6 demethylase, HDA15 deacetylase, and PRMT6 arginine methyltransferase) to reprogram epigenetic marks (e.g., H3K27me3, H4ac, and H3R2me2a) surrounding its binding sites^69,78–81.

Due to the lack of genome-wide analyses correlating NF-Y binding regions with specific chromatin marks in plants, we compared NF-YC2 and NF-YB2 ChIP-seq (chromatin immunoprecipitation followed by sequencing) signals from pNF-YC2::NF-YC2-YFP and pNF-YB2::NF-YB2-YFP seedlings with H3K27me3 ChIP-seq and ATAC-seq (Assay for Transposase-Accessible Chromatin with sequencing) signals from WT seedlings using public datasets^82–85. Heatmaps and metagene plots revealed that NF-YC2 and NF-YB2 ChIP-seq (hereafter referred to as NF-Y ChIP-seq) peaks highly correlate with ATAC-seq peaks across genome-wide loci, enriched predominantly in proximal promoter regions near transcription start sites (TSS). In contrast, chromatin regions marked by NF-Y ChIP-seq and ATAC-seq peaks exhibited low H3K27me3 deposition (Supplementary Fig. 5a, b). These findings suggest that NF-Y binding sites are located in highly active promoter regions characterized by increased chromatin accessibility and reduced H3K27me3 deposition.

The genome-wide correlation between NF-Y binding and active chromatin landscapes prompted us to examine their distribution patterns specifically at C4 gene loci. NF-Y ChIP-seq peaks were enriched at 26 of the 46 C4 gene loci, with 80.42% located in proximal promoter regions (<1 kb from the TSS) and 4.9% in distal regulatory regions (>2 kb from the TSS; Fig. 5a–c and Supplementary Data 4), supporting the essential role of NF-Y in regulating FOD gene expression. Heatmaps and metagene plots confirmed that NF-Y binding regions at C4 gene loci, consistent with genome-wide patterns, exhibited high chromatin accessibility and low H3K27me3 levels (Supplementary Fig. 5c, d). This correlation was further illustrated at representative FOD gene loci, including FUL, SEP3, and HDG5 (Fig. 5d and Supplementary Fig. 5e, f).

Fig. 5. NF-Y mediates H3K27me3 demethylation and chromatin looping at FOD gene loci.

a Venn diagram showing overlap between NF-Y (NF-YC2/NF-YB2) target genes and C4 TFs. b Pie chart showing percentage distribution of NF-Y binding peaks across distinct genomic regions of C4 gene loci. c Heatmap displaying NF-Y ChIP-seq peaks within 5 kb upstream of the transcription start site (TSS) for 46 C4 gene loci. Color scale represents normalized ChIP-seq signal intensity. Representative genes are listed. d Integrative Genomics Viewer (IGV) tracks showing NF-Y binding, H3K27me3 deposition, and chromatin accessibility in the FUL upstream region. H3K27me3_1 and H3K27me3_2, two independent experiments; ATAC_Rep1 and ATAC_Rep2, two biological replicates. Black boxes represent untranslated regions (UTRs), and arrows denote the transcriptional orientation of genes. The dashed frame highlights a chromatin region with strong NF-Y binding, low H3K27me3 deposition, and high chromatin accessibility. e Schematic of the FUL upstream region showing CCAAT boxes, DpnII restriction sites, and genomic fragments analyzed by ChIP-qPCR (F1 to F5, horizontal lines) or 3C-qPCR (B1 to B5, gray blocks) assays. White and black arrowheads indicate positions of anchor and 3C primers, respectively. f ChIP-qPCR assay demonstrating NF-YC3 binding at CCAAT box sites upstream of the FUL locus in 35S::NF-YC3−6HA seedlings. The % input was calculated by normalizing fragments enriched by anti-HA antibody to input chromatin. Non-immune IgG was used as a negative control, and the ACT2 gene was amplified as a nonspecific target. g EMSA confirming direct binding of the NF-Y complex to distal (F1) and proximal (F5) CCAAT box sites in the FUL upstream region. “−” and “+” indicate the absence and presence of proteins or probes, respectively. An unlabeled cold probe was used as a competitor. Representative blots are shown from three biological replicates with similar results. h ChIP-qPCR assay showing H3K27me3 deposition in the FUL upstream region in WT, yc349, and ya2yb23yc9 (nf-ya2 nf-yb2 nf-yb3 nf-yc9) seedlings. The % input was calculated by normalizing fragments enriched by anti-H3K27me3 antibody to input chromatin, with ACT2 gene enrichment as a negative control. i 3C-qPCR assay showing chromatin interactions between NF-Y binding regions at the FUL locus in WT, yc349, and ya2yb23yc9 seedlings. Relative interaction frequency is shown between the anchor region (B5) and adjacent regions (B1 to B4). For ChIP-qPCR and 3C-qPCR assays, data are means ± s.e.m. (n = 3 biological replicates). All P values were determined by unpaired two-tailed Student’s t-tests.

To further characterize the distribution of NF-Y binding sites at FOD gene loci, we performed ChIP-qPCR (chromatin immunoprecipitation followed by quantitative PCR) to assess NF-YC3 binding capacity at different CCAAT box sites upstream of the FUL locus. Consistent with NF-Y ChIP-seq signals, NF-YC3 showed high binding capacity at the proximal CCAAT box site (F5) near the TSS (Fig. 5e, f). In addition, NF-YC3 demonstrated binding capacity at distal CCAAT box sites, including strong binding to a CCAAT box (F1) located ~4.8 kb upstream of the TSS (Fig. 5e, f). This binding of NF-Y to both proximal and distal CCAAT boxes was further validated by EMSA (Electrophoretic Mobility Shift Assay), showing that the complete NF-Y complex, rather than individual subunits, binds directly to synthetic probes corresponding to the FUL upstream sequences (Fig. 5g).

NF-Y has been shown to mediate H3K27me3 demethylation at target genes by recruiting the REF6 demethylase or antagonizing the CLF methyltransferase^69,86,87. To address whether NF-Y regulates H3K27me3 demethylation at FOD gene loci, we measured H3K27me3 levels around NF-Y binding sites upstream of the FUL locus in WT, triple yc349, and quadruple nf-ya2 nf-yb2 nf-yb3 nf-yc9 (ya2yb23yc9) mutants. ChIP-qPCR analysis showed significantly elevated H3K27me3 levels around NF-Y binding sites in both mutants compared to WT (Fig. 5h), suggesting that NF-Y is required to mediate H3K27me3 demethylation in the FUL upstream region.

NF-Y binding to distal CCAAT boxes in the upstream regions of C4 gene loci, including FUL, is reminiscent of its action at the FT locus, where NF-Y mediates chromatin looping by binding to a distal CCAAT box site⁶⁶. To determine whether NF-Y contributes to the formation of local chromatin structures at FOD gene loci, we performed 3C-qPCR (Quantitative Chromosome Conformation Capture) analysis to measure chromatin interactions between different NF-Y binding regions at the FUL locus. A strong interaction was detected between the distal (B1) and proximal (B5) NF-Y binding regions in WT, whereas such interaction was significantly decreased in both mutants (Fig. 5i), suggesting that NF-Y is required to mediate chromatin loop formation between these regions.

To evaluate the regulatory roles of distinct NF-Y binding sites in FUL expression, we generated pFUL::GUS, pFULΔF1::GUS, and pFULΔF5::GUS transgenic lines (expressing GUS reporter driven by the wild-type or NF-Y binding site-deleted FUL upstream sequences). Histochemical and quantitative analysis of GUS activity showed that deletion of the distal NF-Y binding site (F1) significantly reduced GUS expression in rosette leaves, suggesting that this distal NF-Y binding site functions as an enhancer element for gene activation (Supplementary Fig. 5g, h). In contrast, deletion of a single proximal NF-Y binding site (F5) did not visibly affect GUS expression (Fig. 5e and Supplementary Fig. 5g, h), possibly due to functional redundancy among multiple NF-Y binding sites near the TSS.

Together, these findings suggest that NF-Y binds to the upstream regions of FOD gene loci to mediate H3K27me3 demethylation and chromatin looping, thereby establishing active chromatin states for gene activation.

BRC1 dose-dependently disrupts NF-Y-mediated chromatin looping at the FUL locus

Having shown the inverse correlation between FOD gene expression and BRC1 levels (Figs. 1 and 2), the dependency of FOD gene activation on NF-Y-mediated epigenetic regulation (Fig. 5), and the molecular and genetic interactions between BRC1 and NF-YCs (Figs. 3 and 4), we reasoned that BRC1 serves as a transcriptional repressor of FOD genes through modulating NF-Y function. Significantly, BRC1 interacts with the HFD of NF-YCs (Supplementary Fig. 3b), a domain previously shown to mediate interactions with other NF-Y subunits and nonspecific DNA binding^67,74,88. This raised the question of whether BRC1 influences NF-Y complex assembly or DNA-binding capability. To address this, we first examined interactions between BRC1 and all three NF-Y subunits. Y2H assay showed that BRC1 specifically interacts with NF-YCs, but not with NF-YAs or NF-YBs (Supplementary Fig. 6a). We next employed a competitive yeast three-hybrid (Y3H) assay to evaluate the influence of BRC1 on NF-Y complex assembly. Strikingly, BRC1 specifically disrupted interactions between NF-YCs and NF-YAs, but did not affect interactions between NF-YCs and NF-YBs (Fig. 6a, b). To validate these findings in vivo, we utilized a multi-color BiFC system to visualize interactions among NF-Y subunits in the presence or absence of BRC1 in Arabidopsis mesophyll protoplasts. In agreement with the Y3H results, co-expression of BRC1 disrupted interactions between NF-YC3 and NF-YA2, but not between NF-YC3 and NF-YB3 (Fig. 6c and Supplementary Fig. 6b). To further quantitatively assess the extent to which BRC1 influences NF-Y assembly, we conducted an in vitro competitive pull-down assay using recombinant proteins produced in bacteria. Increasing concentrations of BRC1 progressively abolished the NF-YC3–NF-YA2 interaction, without affecting the NF-YC3–NF-YB3 interaction (Fig. 6d, e). These results collectively demonstrate that BRC1 specifically interacts with the NF-YC subunit to prevent the assembly of the NF-YA subunit into the heterotrimeric NF-Y complex in a dose-dependent manner.

Fig. 6. BRC1 dose-dependently disrupts NF-Y-mediated chromatin looping at the FUL locus.

a Y2H assay showing interactions among NF-Y subunits. Transformed yeast cells were grown on selective SD−Trp/Leu/His medium containing 50 mM 3-AT. b Y3H assay revealing that BRC1 disrupts interactions of NF-YCs with NF-YAs but not with NF-YBs. Transformed yeast cells were grown on selective SD−Met/Trp/Leu/His medium containing 10 mM 3-AT. BD−DIV, BD-fused N-terminal truncated BRC1. c Multi-color BiFC assay confirming that BRC1 disrupts interactions of NF-YC3 with NF-YA2 but not with NF-YB3 in Arabidopsis mesophyll protoplasts. Cerulean panel shows interaction between NF-YC3 and NF-YA2 or NF-YB3, and the Venus panel indicates interaction between NF-YC3 and BRC1. A representative result from three independent experiments is shown. Scale bars, 10 µm. Competitive pull-down assays showing that BRC1 dose-dependently disrupts interactions of NF-YC3 with NF-YA2 (d) but not with NF-YB3 (e). MBP−NF-YC3 was incubated with immobilized GST−NF-YA2 or GST−NF-YB3 in the presence of increasing amounts of MBP−BRC1. Ponceau S staining (Ponc.) indicates equal amounts of bait proteins, and 5% of purified NF-YC3 or BRC1 is shown as input control. Competition with MBP proteins served as a negative control. One representative experiment of three biological replicates is shown. Competitive EMSA showing that BRC1 dose-dependently disrupts NF-Y assembly at the distal (f) but not the proximal (g) CCAAT box site upstream of the FUL locus. Representative blots are shown from three biological replicates with similar results. h 3C-qPCR assay showing relative interaction frequency between NF-Y binding regions at the FUL locus in axillary tissues of WT, brc1-2, BRC1-OE2, and BRC1-OE12 plants. Data are means ± s.e.m. (n = 3 biological replicates). P values were determined by unpaired two-tailed Student’s t-tests. i GUS histochemical staining of growing buds detached from pFUL::GUS or pFULΔF1::GUS transgenic plants (expressing GUS reporter driven by wild-type or distal NF-Y binding site-deleted FUL upstream sequences). Representative images are shown from more than three independent transgenic plants. j Quantitative GUS activity in 0.5-cm-long buds of the indicated plants, expressed as pmol 4-methylumbelliferone (MU) produced per min per mg of total protein. Data are means ± s.e.m. (n = 3 biological replicates). P value was determined by unpaired two-tailed Student’s t-tests.

Structural studies have proven that NF-YA is responsible for directly recognizing the CCAAT box in target sequences^67,74,88. To investigate whether BRC1 influences the binding capacity of NF-Y at CCAAT box sites, we performed competitive EMSA to assess NF-Y binding to target sites in the FUL upstream region under varying BRC1 concentrations. Although BRC1 alone did not bind directly to these target sites, it dose-dependently disrupted binding of the canonical NF-Y complex to a synthetic probe corresponding to the distal CCAAT box site (F1; Fig. 6f). By contrast, no such effect was observed at the proximal CCAAT box site (F5; Fig. 6g). Notably, BRC1 effectively competed with NF-YA2, forming a non-canonical BRC1/NF-YC3/NF-YB3 complex that occupied the distal F1 site (Fig. 6f). Consistent with these in vitro findings, dual-crosslinking ChIP-qPCR assays confirmed that NF-YC3 and BRC1 occupy the same distal F1 site upstream of the FUL locus in axillary buds (Supplementary Fig. 6c). In addition, comparison of NF-YC2 and BRC1 ChIP-seq signals across 46 C4 gene loci revealed substantial overlap in their binding peaks (Supplementary Fig. 6d and Supplementary Data 4), suggesting that BRC1 may influence NF-Y assembly at numerous target sites. These results collectively demonstrate that BRC1 preferentially interferes with NF-Y assembly at the distal CCAAT box site upstream of the FUL locus in a dose-dependent manner.

To evaluate the consequences of BRC1 occupancy at CCAAT box sites in planta, we performed 3C-qPCR to analyze chromatin interactions between NF-Y binding regions at the FUL locus in axillary tissues of genotypes with varying BRC1 expression. The interaction frequency between the distal (B1) and proximal (B5) NF-Y binding regions was significantly reduced in BRC1 overexpression lines (BRC1-OE2 and BRC1-OE12) compared to WT, but was slightly increased in brc1-2 mutant (Fig. 6h and Supplementary Fig. 6e). Likewise, decapitation-induced BRC1 downregulation resulted in enhanced interaction between these regions in WT axillary tissues (Supplementary Fig. 6f). These findings suggest that BRC1 negatively regulates chromatin loop formation at the FUL locus in a dose-dependent manner.

Given the enhancer role of the distal NF-Y binding site (Supplementary Fig. 5g) and the inhibitory effect of BRC1 on chromatin looping at the FUL locus (Fig. 6h), we reasoned that BRC1 competes for occupancy at the enhancer element, preventing its interaction with the proximal promoter and consequently influencing gene activation. In line with our expectation, disruption of enhancer–promoter interaction by removing the distal NF-Y binding site significantly reduced GUS expression activity in axillary buds of pFULΔF1::GUS compared to pFUL::GUS (Fig. 6i, j).

Together, these findings suggest that BRC1 disrupts NF-Y assembly at specific CCAAT box sites in a dose-dependent manner, thereby interfering with NF-Y-mediated chromatin looping required for FOD gene activation.

BRC1 dose-dependently inhibits NF-Y-dependent H3K27me3 demethylation at FOD gene loci

As mentioned earlier, most axillary buds in Arabidopsis remain dormant throughout the plant’s lifespan, raising the possibility that plants might employ epigenetic regulatory mechanisms to exert prolonged control over growth- or development-related genes in these buds. Given the negative correlation between NF-Y binding and H3K27me3 deposition at genome-wide loci, including the C4 gene loci (Supplementary Fig. 5a–d), we hypothesized that BRC1 occupancy at CCAAT box sites could influence NF-Y-mediated H3K27me3 demethylation at these loci. To test this, we performed CUT&Tag (Cleavage Under Targets and Tagmentation) to analyze H3K27me3 deposition at C4 gene loci in axillary tissues of WT and BRC1 overexpression plants (BRC1-OE12). Metagene plots showed significantly increased H3K27me3 deposition within 2 kb upstream of the TSS at C4 gene loci in BRC1-OE12 plants compared to WT (Fig. 7a, b, Supplementary Fig. 7a, b, and Supplementary Data 4), a trend also evident at representative FOD gene loci such as FUL, CAL, and HDG5 (Fig. 7c and Supplementary Fig. 7c–e). These findings indicate that elevated BRC1 levels lead to increased H3K27me3 deposition in upstream regulatory regions of FOD gene loci.

Fig. 7. BRC1 dose-dependently inhibits NF-Y-dependent H3K27me3 demethylation at FOD gene loci.

a Metagene plots showing the distribution of H3K27me3 signal within 5 kb upstream of the TSS for 46 C4 gene loci in axillary tissues of WT and BRC1-OE12 plants. H3K27me3 signal was calculated from two independent CUT&Tag experiments, with IgG used as a negative control. b Box plots showing H3K27me3 signal intensity at the −2 kb, −1 kb, and TSS positions indicated in (a); center lines indicate medians, box bounds represent the 25th and 75th percentiles, and whiskers extend to 1.5× the interquartile range (n = 46 gene loci). P values were determined by paired two-tailed Student’s t-tests. c IGV tracks illustrating H3K27me3 deposition in the FUL upstream region. Dashed frames highlight chromatin regions with significant differences in H3K27me3 levels. d, e ChIP-qPCR assay showing H3K27me3 deposition in the FUL upstream region in axillary tissues of the indicated plants. Data are means ± s.e.m., with each circle representing a biological replicate (n = 3). P values were determined by unpaired two-tailed Student’s t-tests. f Representative images of WT, brc1-2, BRC1-OE2, and BRC1-OE12 plants grown for 40 days under LD conditions. White arrowheads indicate elongated R1 branches. g A proposed model illustrating BRC1 as a rheostat of NF-Y-dependent epigenetic regulation in axillary bud development.

To better understand the relationship between BRC1 levels and H3K27me3 deposition at FOD gene loci, we performed ChIP-qPCR to compare H3K27me3 levels in the FUL upstream region in axillary tissues of genotypes with varying BRC1 expression. H3K27me3 deposition was significantly elevated in BRC1 overexpression lines (BRC1-OE2 and BRC1-OE12) compared to WT, but was slightly reduced in the brc1-2 mutant (Fig. 7d). Similarly, decapitation-induced BRC1 downregulation led to reduced H3K27me3 deposition in the FUL upstream region in WT axillary tissues (Supplementary Fig. 6g). These findings suggest that BRC1 positively regulates H3K27me3 deposition in a dose-dependent manner. To assess whether BRC1’s influence on H3K27me3 deposition is NF-Y-dependent, we analyzed H3K27me3 levels in the FUL upstream region in axillary tissues of WT, brc1-2, yc349, and brc1-2 yc349 plants. ChIP-qPCR assays showed that H3K27me3 deposition was significantly restored in the brc1-2 yc349 mutant compared to the brc1-2 mutant (Fig. 7e), suggesting that BRC1-mediated changes in H3K27me3 deposition require a functional NF-Y complex. In agreement with the levels of H3K27me3 deposition and FOD gene expression across genotypes, axillary bud growth was apparently retarded in BRC1 overexpression lines but accelerated in the brc1-2 mutant compared to WT (Fig. 7f).

Together, these findings suggest that BRC1 inhibits NF-Y-dependent H3K27me3 demethylation at FOD gene loci in a dose-dependent manner, thus enabling sustained epigenetic silencing of these developmental genes in axillary buds.

Discussion

In animals, the “pioneer factor co-binding” and “TF competition” models are well-established paradigms for elucidating the interplay between NF-Y and other transcriptional regulators in the control of gene expression^{75–77,89,90}. In plants, although NF-Y-associated regulatory models remain less clearly defined, emerging evidence suggests mechanistic similarities to animal systems^{63,69,81,91,92}. Building on these paradigms, this study proposes an “epigenetic rheostat” model, positioning BRC1 as a molecular rheostat that dose-dependently modulates NF-Y-mediated epigenetic regulation of target genes in axillary buds (Fig. 7g). In decapitated, SL-deficient, or BRC1-downregulated plants, low BRC1 levels permit the assembly of NF-YA, NF-YB, and NF-YC subunits into a heterotrimeric NF-Y complex. This complex specifically binds to CCAAT box sites, mediating enhancer–promoter looping and H3K27me3 demethylation to activate FOD genes (e.g., FUL), thereby promoting axillary bud development (other co-activators may coordinate with NF-Y in this process). In SL-treated or BRC1-upregulated plants, high BRC1 levels enable competition with NF-YA, resulting in the formation of a non-canonical BRC1/NF-YC/NF-YB complex. This complex preferentially occupies distal enhancers, disrupting NF-Y-mediated epigenetic regulatory effects to silence FOD genes, thus arresting axillary bud development. In WT plants under apical dominance, intermediate BRC1 levels enable dose-dependent modulation of NF-Y assembly, fine-tuning FOD gene expression in response to changing branching signals. Such a model not only elaborates the mechanistic interplay between BRC1 and NF-Y in regulating development-related FOD genes in plants but also introduces a key paradigm for the dynamic/reversible modulation of NF-Y activity across eukaryotic systems.

The plasticity of shoot branching entails dynamic regulatory mechanisms that sustain axillary bud inhibition under adverse conditions and relieve this inhibition when favorable conditions resume^3,10,12. This study presents a coherent mechanistic model in which two distinct branching pathways, SL signaling and decapitation, are integrated through BRC1 expression dynamics to modulate the NF-Y-mediated epigenetic regulation of FOD genes. SL signaling upregulates BRC1 expression to repress NF-Y-mediated epigenetic regulatory effects, leading to the accumulation of repressive chromatin marks at FOD gene loci that ensure sustained transcriptional silencing of these developmental genes in axillary buds under strong apical dominance. In contrast, decapitation downregulates BRC1 expression to derepress NF-Y-mediated epigenetic regulatory effects, resulting in the establishment of active chromatin states at FOD gene loci that enable their transcriptional activation in axillary buds upon removal of apical dominance. In this way, BRC1 functions not merely as a transcriptional repressor but as a molecular rheostat that integrates distinct hormonal and environmental inputs into tunable epigenetic outputs at target loci. Whether this BRC1-regulated GRN also functions in the integration of additional branching pathways, such as nutrient or sugar signaling, remains a compelling question for future investigation.

BRC1 is hypothesized to regulate the bud activation threshold without being strictly necessary or sufficient to inhibit bud growth³⁴, a role that aligns well with its rheostatic function in regulating FOD genes. In this scenario, BRC1 does not act as a conventional molecular switch that directly binds FOD gene loci to control their transcriptional on/off states; instead, it functions as a molecular rheostat that sets the threshold for NF-Y-dependent transcriptional activation of FOD genes in a dose-dependent manner, thereby integrating branching signals to modulate the activation potential of axillary buds. Mechanistically, elevated BRC1 levels do not require direct DNA binding to repress FOD genes, but instead function by inhibiting NF-Y-mediated active chromatin states at target loci. Conversely, reduced BRC1 levels are not sufficient to directly activate FOD genes, but serve to relieve the inhibition on NF-Y-mediated epigenetic regulatory effects. The actual activation of FOD genes therefore requires the NF-Y complex to coordinate with as-yet undefined co-activators, initiating transcription in response to intrinsic developmental cues (e.g., meristem identity and floral transition) and external environmental signals (e.g., photoperiod and temperature). This “epigenetic rheostat” model helps to understand observations such as reduced rosette branching in brc1-2 yc349 and ft-10 brc1-2 brc2-1 plants under LD conditions³⁴ (where BRC1 loss is insufficient to activate rosette bud growth due to a deficiency of FOD-related activators) and enhanced rosette branching in FUL and FT overexpression plants (where BRC1 expression is ineffective to suppress rosette bud growth due to an abundance of FOD-related activators)^32,93,94.

Following initiation, axillary meristems sequentially produce leaf primordia during the vegetative phase and floral meristems during the reproductive phase, forming buds capable of elongating into branches under favorable conditions^72,73. This developmental progression is markedly accelerated in brc1 mutants, in which most axillary buds transition to the reproductive phase significantly earlier than in WT³⁶. Previous studies demonstrated that BRC1 interacts with and suppresses FT, a multifunctional protein implicated in early meristem development, floral transition, and outgrowth of axillary buds^30,48,94. BRC1 was also shown to directly activate HB21/40/53, a group of genes that contribute to ABA-regulated axillary bud dormancy and floral bud arrest^43,95. Consistent with these roles, this study shows that BRC1, acting as a repressor of axillary bud development, negatively regulates many meristem identity genes associated with the FOD process. Moreover, we identified an interaction between BRC1 and AGAMOUS-LIKE 16 (AGL16; Supplementary Data 1), a known transcriptional repressor of numerous FOD genes⁹⁶. These findings collectively underscore the central role of BRC1 in integrating multiple GRNs that govern meristem development and phase transitions in axillary buds. Notably, BRC1/TB1 has been shown to regulate aerial bud tuberization in potato⁴⁶, as well as spikelet development in wheat and barley^47,97, highlighting its pleiotropic roles across diverse developmental processes in axillary branches. Future research should further explore additional functions of BRC1 in axillary buds, particularly in specialized lateral organs of crop species.

In addition to the C4 genes, we identified 76 C2 genes whose expression patterns positively correlate with BRC1 levels. Many C2 genes, including HB53, TCP1, and MYB114, were previously characterized as early SL-responsive genes involved in shoot branching, leaf development, and anthocyanin biosynthesis, respectively³⁷. Our study further demonstrates that their expression remains persistently responsive to SL signaling in a BRC1-dependent manner. Previous studies have well established that BRC1 functions as a transcriptional activator by directly binding to and activating bud dormancy-related genes (e.g., HB21/40/53)^43,51,52. Our study extends this understanding by demonstrating that BRC1 can also function as a transcriptional repressor by disrupting NF-Y-dependent active chromatin landscapes at FOD genes. Moreover, tomato BRC1 has been reported to directly repress the CK biosynthetic gene LOG4 while activating CK and GA catabolic genes (CKX7, GA2ox4, and GA2ox5)⁹⁸. These findings highlight the mechanistic complexity of BRC1-mediated regulation and provide a foundation for further investigation into its target-specific regulatory modes.

Another notable feature reflecting the mechanistic complexity of BRC1-mediated regulation is its site selectivity. Although BRC1 can compete with NF-YA subunits to prevent NF-Y complex assembly (in the absence of DNA; Fig. 6d, e), this competition occurs only at specific CCAAT box sites (e.g., the distal F1 site but not the proximal F5 site) and does not rely on direct BRC1 binding to either site (Fig. 6f, g). These observations suggest that such site-selective competition is determined by the specific CCAAT box site or, more explicitly, by the sequence context flanking the core CCAAT motif. Prior studies have shown that nucleotide variations flanking the core CCAAT motif can substantially alter NF-Y/DNA binding affinity and complex stability^88,99,100. Accordingly, we speculate that this site selectivity arises as a consequence of intrinsic differences in NF-Y/DNA complex stability at the distal versus proximal CCAAT box sites. Specifically, the NF-Y/DNA complex assembled at the proximal F5 site appears sufficiently stable to resist BRC1 competition, whereas the complex assembled at the distal F1 site is less stable and thus more susceptible to BRC1 competition. Future efforts should systematically evaluate BRC1 competition across additional CCAAT box sites upstream of FOD gene loci to further elucidate this flanking-sequence-dependent selectivity.

Finally, the mechanistic complexity of BRC1-mediated regulation is further shaped by its distinctive spatiotemporal expression pattern, which is strictly confined to developing axillary buds. It is well established that axillary buds at different leaf axil positions develop in a basipetal sequence and exhibit differential sensitivity to environmental cues. Loss of BRC1 leads to the outgrowth of nearly all axillary buds, rendering them insensitive to treatments such as decapitation and SL application^36,60,101. Conversely, excessive BRC1 overexpression severely arrests seedling growth, and its ectopic expression at the shoot apex delays flowering^40,48. These observations emphasize the critical importance of precise spatial and temporal control of BRC1 expression in axillary buds. Given BRC1’s rheostatic role in setting the bud activation potential, we propose that variations in BRC1 expression among individual buds underlie differences in their developmental sequence and environmental responsiveness. To test this hypothesis, emerging single-cell and spatial-omics technologies offer promising tools for resolving the spatiotemporal dynamics of BRC1 expression and its downstream GRN activity across specific leaf axil positions. Such efforts may advance our understanding of BRC1-mediated position-specific bud activity and inform strategies for manipulating BRC1-regulated GRNs to optimize crop architecture.

Methods

Plant materials and growth conditions

The Arabidopsis thaliana ecotype Columbia (Col-0) was used as the wild type (WT). The max3-11 (SALK_023975), max2-3 (SALK_092836), d14-1 (CS913109), brc1-2 (SALK_091920), and nf-yc3 nf-yc4 nf-yc9 (yc349; SALK_034838 SALK_032163 SALK_058903) mutants were previously described^36,53,70,102. The pNF-YC3::GUS (CS67034) and pNF-YC4::GUS (CS67035) lines were obtained from the Arabidopsis Biological Resource Center. The nf-ya2 nf-yb2 nf-yb3 nf-yc9 (ya2yb23yc9) and 35S::NF-YC3−6HA lines were kindly provided by Dr. Xingliang Hou (University of the Chinese Academy of Sciences, China) as previously described⁶⁹. The brc1-2 yc349 and d14-1 yc349 mutants were generated by genetic crosses using the relevant parental lines. All homozygous T-DNA insertional mutants were genotyped by using primers listed in Supplementary Data 5.

Arabidopsis seeds were surface-sterilized, sown on 1/2 Murashige and Skoog (MS) agar plates (containing 1/2 MS powder, 0.7% agar, and 1% sucrose, pH 5.8), chilled at 4 °C for 3 days, and then transferred to a growth chamber (at 22 °C with a 16 h light/8 h dark photoperiod). For regular growth, 7-day-old seedlings were transferred to soil and grown under the same conditions. For mesophyll protoplast preparation, 7-day-old seedlings were transplanted into autoclaved soil and grown for 4 weeks in a growth chamber (at 22 °C with a 10 h light/14 h dark photoperiod and a light intensity of 120 µmol m⁻² s⁻¹).

For rac-GR24 treatment, seeds were surface-sterilized, immersed in water, and stratified at 4 °C for 3 days. For the 1-day treatment, seeds were sown on 1/2 MS agar plates for 6 days before being transferred to plates containing 2.5 µmol rac-GR24 (PhytoTech Labs, G3324) or DMSO (Sigma-Aldrich, BP231) for 1 day. For the 3-day treatment, seeds were sown on 1/2 MS agar plates for 4 days, then moved to plates with the chemicals for an additional 3 days. For the 7-day treatment, seeds were sown directly on 1/2 MS agar plates containing the chemicals and grown for 7 days until harvest. All treatments were conducted in a growth chamber (at 22 °C with a 16 h light/8 h dark photoperiod).

Generation of transgenic plants

To generate a BRC1 overexpression line, the full-length coding sequence of BRC1 was PCR-amplified from Arabidopsis cDNA and cloned into the plant binary vector pEarleyGate202. For the pFUL::GUS construct, the native FUL promoter containing 5465 bp upstream of the start codon was amplified from Arabidopsis genomic DNA and assembled into the plant binary vector pCAMBIA1300Z. To create the pFULΔF1::GUS and pFULΔF5::GUS constructs, the relevant F1 or F5 fragment was deleted through site-directed mutagenesis using the pFUL::GUS construct as the template.

After sequencing verification, the constructs were transformed into WT plants using the standard Agrobacterium-mediated flower dip method. Homozygous T₃- or T₄-generation plants of BRC1-OE2, BRC1-OE12, pFUL::GUS, pFULΔF1::GUS, or pFULΔF5::GUS were employed for subsequent experiments. The PCR primers used for plasmid constructions and Sanger sequencing are listed in Supplementary Data 5.

Sample collection, decapitation treatment, and phenotypic analysis

Plants were grown in a growth chamber under long-day (LD) conditions (22 °C, 16 h light/8 h dark photoperiod). Axillary tissues (including axillary buds and subtending rosette leaf axil tissues) of each genotype were collected by removing the main inflorescence, roots, and rosette leaves (with their petioles) when the main inflorescence reached approximately 1 cm in height after bolting.

For decapitation treatment, the entire main shoot below the lowest cauline node was excised using sterilized scissors. For gene expression analysis, decapitation was performed when the main inflorescence reached 1 cm in height, and axillary tissues were collected at 18 h or 36 h post-decapitation from at least 8 plants (intact or decapitated) per genotype. For axillary branch elongation analysis, decapitation was conducted when the main inflorescence reached 5 cm in height, and lengths of the uppermost primary rosette (R1) branch were measured daily from at least 8 plants (intact and decapitated) per genotype.

For early axillary bud phenotype analysis, 20 individuals per genotype were dissected using a stereoscopic microscope when the main inflorescence reached 1 cm in height, thereby minimizing the influence of flowering-time variation among genotypes. The developmental stage of each axillary bud was classified according to a previously described method³⁶. For adult plant phenotype analysis, plants were grown in soil and photographed at various time points. The numbers of R1 branches (>1 cm in length) and rosette leaves (RL) were recorded for at least 16 plants per genotype after 60 days of growth under LD conditions, by which time all genotypes had bolted.

RNA-seq and data analysis

Total RNA was extracted from seedlings or axillary tissues using TRIzol reagent (Invitrogen, 15596026) following the manufacturer’s protocol. RNA-seq libraries were prepared with the RNA Library Prep Kit for Illumina (NEB, E7775) and sequenced on the NovaSeq 6000 platform (Illumina) to produce paired-end 150 bp reads. Raw reads were trimmed to remove adapters, and low-quality reads were filtered out using Cutadapt (v.3.5). High-quality reads were aligned to the Arabidopsis reference genome (TAIR10) using HISAT2 (v.2.0.5) with default parameters. Aligned reads were sorted and converted into BAM files using SAMtools (v.1.13), and gene expression was quantified by counting paired mapped reads with FeatureCounts (v.2.0.1). DEGs were identified using DESeq2 (v.1.32.0) with thresholds of p value < 0.05 and fold-change >1.2. Differentially expressed TFs were clustered using the K-means algorithm and visualized with the R package ClusterGVis. GO analysis was conducted using agriGO (v.2.0) and visualized with the R package ggplot2.

CUT&Tag and data analysis

For Arabidopsis nuclei extraction, 0.2 g of freshly harvested axillary tissue was ground in 10 mL of extraction buffer (20 mM HEPES, 250 mM sucrose, 1 mM MgCl₂, 5 mM KCl, 40% glycerol, 0.25% Triton X-100, 0.1% β-mercaptoethanol, and 2× Complete protease inhibitor cocktail, pH 7.4) using a pre-cooled mortar and pestle at 4 °C. The extract was sequentially filtered through 70 µm and 40 µm cell strainers and centrifuged at 1900 × g for 20 min at 4 °C. After discarding the supernatant, the nuclei pellet was resuspended in 1 mL of extraction buffer, transferred to a 1.5 mL centrifuge tube, and centrifuged at 1900 × g for 5 min at 4 °C. The isolated nuclei were immediately processed using the CUT&Tag Assay Kit (CST, 77552) following the manufacturer’s manual. Briefly, the nuclei were washed twice with 1 mL of wash buffer, immobilized onto Concanavalin A-coated beads, and resuspended in 100 µL of antibody binding buffer. Then, 2.5 µL of anti-H3K27me3 antibody (CST, 9733) or normal rabbit IgG (as a negative control) was added, and the mixture was incubated overnight at 4 °C. After incubation, the supernatant was removed, and 50 µL of digitonin buffer containing 1.25 µL of anti-rabbit IgG antibody was added, followed by a 2 h incubation at 4 °C. Next, 2 µL of pAG-Tn5 was added, and the mixture was incubated for 1 h at room temperature for pAG-Tn5 binding. After the addition of Mg²⁺, the mixture was incubated at 37 °C for 1 h to activate tagmentation. The reaction was terminated by adding 6.75 µL of 0.5 M EDTA, 8.25 µL of 10% SDS, and 3 µL of 20 mg/mL Proteinase K, followed by overnight incubation at 58 °C to reverse crosslinking. After centrifugation, the supernatant was transferred to a new centrifuge tube, and the tagmented DNA was purified using a spin column and eluted in 30 µL of nuclease-free water. Finally, the DNA products were amplified for 16 cycles using DNA/RNA UD Indexes (Illumina, 20026121), purified with AMPure XP beads (Beckman, A63881), and sequenced on the Illumina NovaSeq 6000 platform.

For data analysis, raw reads were processed using fastp (v.0.23.1) to trim adapters, remove low-quality bases, and filter out poor-quality reads. Clean reads were aligned to the Arabidopsis reference genome (TAIR10) using Bowtie2 (v.2.3.5.1) with default settings. Aligned reads were sorted and filtered using SAMtools (v.1.13) with the parameters “-F 1804 -f 2 -q 30” to retain high-quality, properly paired reads. Duplicate reads were removed using Picard tools (v.2.23.3). Deduplicated BAM files from two biological replicates were merged and converted to bigWig format using bamCoverage from deepTools (v.3.5.0), normalized by reads per kilobase per million (RPKM), for visualization in IGV (v.2.8.0). Peak calling for H3K27me3 modifications was performed using MACS2 (v.2.2.7.1) with parameters for broad peak detection (“--broad --broad-cutoff 0.05”). Peak annotation was conducted using the R package ChIPseeker (v.1.24.0). Quantification of read counts per peak was performed using DiffBind (v.2.16.2), and differential analysis of H3K27me3 peaks was performed with DESeq2 (v.1.32.0). Data visualization, including heatmaps and profile plots centered on peaks, was carried out using deepTools (v.3.5.0).

ChIP-seq and ATAC-seq data analysis

NF-YC2 and NF-YB2 ChIP-seq datasets generated from pNF-YC2::NF-YC2-YFP and pNF-YB2::NF-YB2-YFP seedlings (GSE80564)⁸³, along with H3K27me3 ChIP-seq datasets (GSE47202, GSE89357)^82,84 and ATAC-seq datasets (GSE146948)⁸⁵ generated from WT seedlings, were obtained from the ChIP-Hub database¹⁰³. BRC1 ChIP-seq datasets generated from GFP::BRC1^ind seedlings (GSE155028)⁵² were retrieved from the NCBI Gene Expression Omnibus (GEO) database. Data analysis was performed following procedures similar to those described for CUT&Tag. Briefly, cleaned reads were aligned to the Arabidopsis reference genome (TAIR10) using Bowtie2 (v.2.3.5.1) for ATAC-seq and BWA (v.0.7.17) for ChIP-seq. To improve peak detection accuracy, deduplicated BAM files from biological replicates were merged prior to peak calling. For ATAC-seq, peaks were identified with “-q 0.05 -f BAMPE --nomodel --extsize 200 --shift -100” to define accessible chromatin regions. For TF ChIP-seq, narrow peaks were called using “-p 1e-3 -f BAMPE.” For H3K27me3 ChIP-seq, broad peaks were called using “--broad --broad-cutoff 0.05.” Heatmaps and profile plots centered on peaks were generated using computeMatrix and plotHeatmap from deepTools (v.3.5.0) to visualize enrichment patterns across samples.

ChIP-qPCR

Chromatin was extracted using the ChromaFlash Plant Chromatin Extraction Kit (Epigentek, P-2022), and ChIP assays were performed with the EpiQuik Plant ChIP Kit (Epigentek, P-2014) according to the manufacturer’s instructions. Briefly, chromatin from seedlings or axillary tissues was extracted, sheared, and immunoprecipitated using anti-HA (Sigma-Aldrich, H9658), anti-Flag (Sigma-Aldrich, F3165), or anti-H3K27me3 (CST, 9733) antibodies. Following overnight incubation at 4 °C, enriched DNA was released from the antibody-protein-DNA complex, de-crosslinked, and purified using a spin column. For dual-crosslinking ChIP assays, tissues were first crosslinked with 1.5 mM ethylene glycol bis-succinimidyl succinate (EGS; ThermoFisher, 21565) in PBS under vacuum for 30 min, followed by fixation with 1% formaldehyde for an additional 10 min. Subsequent steps were performed according to the standard ChIP protocol described above. The eluted DNA was quantified and analyzed by qPCR using iQ SYBR Green Supermix (Bio-Rad, 170-8882). The enrichment of target DNA fragments was calculated as a percentage of input DNA. The ACT2 locus was used as a nonspecific target control. All primers used for qPCR analyses are listed in Supplementary Data 5.

3C-qPCR

Chromosome Conformation Capture (3C) assays were conducted as previously described with minor modifications¹⁰⁴. Briefly, 3 g of seedlings or axillary tissues were crosslinked with 2% formaldehyde under vacuum for 20 min. Nuclei were isolated using 30 mL of extraction buffer (as used in the CUT&Tag assay), filtered sequentially through 70 µm and 40 µm cell strainers, and pelleted by centrifugation at 1900 × g for 20 min at 4 °C. The nuclei pellets were washed twice with 1 mL of 1× rCutSmart buffer (NEB, B6004), resuspended in 500 µL of 1.2× rCutSmart buffer containing 0.3% SDS, and incubated for 40 min at 37 °C, followed by an additional 20 min at 65 °C, Next, Triton X-100 was added to a final concentration of 2%, and the mixture was incubated for 60 min at 37 °C to sequester the SDS. Chromatin DNA was digested with 600 U of DpnII (NEB, R0543) for 20 h at 37 °C with rotation. DpnII was inactivated by adding 1% SDS and incubating for 25 min at 65 °C.

For the ligation reaction, 750 µL of digested nuclei were mixed with 4 mL of 1× T4 DNA ligation buffer (NEB, B0202) containing 1% Triton X-100. After incubation for 1 h at 37 °C, 2000 U of T4 DNA ligase (NEB, M0202) was added, and the mixture was incubated at 16 °C overnight, followed by 45 min at room temperature. De-crosslinking was carried out by overnight incubation at 65 °C in the presence of 200 µg/mL proteinase K. The 3C ligation products were purified by phenol:chloroform:alcohol (25:24:1) extraction, and the precipitated DNA was dissolved in 150 µL of ddH₂O. In parallel, the pFUL::GUS construct was digested and re-ligated to generate a control template for normalizing variations in PCR amplification efficiency across different primer pairs. Quantification of 3C DNA ligation products was performed by real-time PCR and relative interaction frequency was calculated using the 2^−ΔΔCt method. Briefly, Ct values were first normalized to an internal loading control amplified from a DpnII-free genomic region (B5) to correct for differences in template input; then calibrated against the corresponding pFUL::GUS ligation controls to account for variations in primer amplification efficiency; and finally scaled relative to the primer pair exhibiting the highest interaction frequency (B1 to B5) in the WT sample to enable cross-sample comparisons. Primers used for 3C-qPCR are listed in Supplementary Data 5.

RT-qPCR

Total RNA from seedlings or axillary tissues was isolated using TRIzol reagent (Invitrogen, 15596026) and reverse transcribed into cDNA using the Maxima First Strand cDNA Synthesis Kit (ThermoFisher, K1641). Transcript levels were quantified by real-time qPCR with iQ SYBR Green Supermix (Bio-Rad, 170-8882) on a Bio-Rad CFX Opus 96 thermal cycler. Relative gene expression was normalized to the internal control ACTIN8 using the 2^−ΔΔCt method. All experiments were performed with at least four biological replicates. Primers used are listed in Supplementary Data 5.

BiFC and multi-color BiFC

BiFC and multi-color BiFC assays were performed using transient expression vectors and procedures as previously described¹⁰⁵. The full-length coding sequences of NF-YC3 and NF-YC4 were cloned into pSAT6-cCFP to generate the YC3-cCFP and YC4-cCFP constructs. The BRC1 coding sequence was cloned into pSAT4-nVenus to create the BRC1-nVenus construct. The full-length sequences of NF-YA2 and NF-YB3 were cloned into pSAT1-nCerulean to generate the YA2-nCerulean and NF-YB3-nCerulean constructs. After sequencing verification, equal amounts of plasmids were mixed and transformed into Arabidopsis mesophyll protoplasts, with an empty vector serving as a negative control. After overnight incubation at room temperature, protoplasts were harvested, and fluorescence signals were detected using a confocal microscope (LSM 710, Carl Zeiss). Experiments were repeated at least three times, with a minimum of 10 protoplasts imaged per replicate.

Recombinant protein preparation

To construct MBP-fusion, GST-fusion, or 6His-fusion plasmids, coding sequences were amplified from Arabidopsis cDNA and inserted into pMal-C5X (for MBP−BRC1 and MBP−NF-YC3), pGEX-6P-1 (for GST−NF-YC3, GST−NF-YC4, GST−NF-YC9, GST−NF-YB3, and GST−NF-YA2), or pETDuet-1 (for 6His−NF-YB3/NF-YC3, co-expressing two genes in a single vector). Primers used for plasmid construction are listed in Supplementary Data 5.

Recombinant proteins were expressed in Escherichia coli strain BL21 (DE3) and induced with 0.3 mM isopropyl β-D-thiogalactoside (IPTG) for 18 h at 16 °C. GST-fusion proteins were purified using Glutathione Sepharose 4B (Cytiva, 17075605) and eluted with 10 mM reduced L-glutathione (Sigma-Aldrich, G4251). MBP-fusion proteins were purified using Amylose Resin (NEB, E8021) and eluted with 20 mM maltose (Sigma-Aldrich, M9171). 6His-fusion proteins were purified using Ni-NTA Agarose (Qiagen, 30210) and eluted with 250 mM imidazole (Sigma-Aldrich, I2399). All recombinant proteins were further purified by HiTrap Desalting chromatography (Cytiva, 17-5087-01) using a buffer containing 10 mM HEPES (pH 7.4), 150 mM NaCl, and 1 mM DTT.

Pull-down and competitive pull-down

For pull-down assays, 150 µL of Amylose resin was washed five times with reaction buffer (50 mM Tris-HCl, pH 7.8, 100 mM NaCl, 0.1% Tween 20, 10% glycerol, and 20 mM β-mercaptoethanol) to remove residual ethanol. Approximately 30 µg of purified MBP−BRC1 or MBP proteins was added to the pre-washed resin and incubated for 2 h at 4 °C with gentle shaking. The resin was then washed five times with reaction buffer to remove unbound proteins. Next, 10 µg of GST−NF-YC proteins was added to the resin. After 1 h incubation at 4 °C with gentle shaking, the resin was washed five times with reaction buffer to remove unbound proteins. Finally, the bound proteins were released by boiling in SDS sample buffer at 100 °C for 5 min.

For competitive pull-down assays, approximately 30 µg of purified GST−NF-YA2 or GST−NF-YB3 was added to pre-washed Glutathione Sepharose 4B resin and incubated for 2 h at 4 °C with gentle shaking. After washing the resin five times with reaction buffer, 10 µg of MBP−NF-YC3 protein was added and incubated for 1 h at 4 °C. Then, gradient amounts of MBP−BRC1 protein were added for competitive binding. After another 1 h incubation at 4 °C, the resin was washed five times with reaction buffer to remove unbound proteins. Finally, the bound proteins were released by boiling in SDS sample buffer at 100 °C for 5 min.

Immunoblotting was performed using anti-MBP (1:5000 dilution; NEB, E8032), anti-GST (1:1000 dilution; Santa Cruz, sc-138), and anti-mouse IgG-HRP (1:2000 dilution; Santa Cruz, sc-516102) antibodies. The relative input levels of MBP-fusion and GST-fusion proteins were assessed by Ponceau S staining.

Y2H screening, Y2H and Y3H assays

For Y2H screening, the Y2HGold strain was transformed with the bait vector pGBKT7-BRC1-DIV (an N-terminal truncated variant of BRC1) and then mated with the Y187 strain harboring the Normalized Mate & Plate Library - Universal Arabidopsis (Takara, 630487). After mating, diploid yeast cells were plated on selective media SD−Trp/Leu/His supplemented with 5 mM 3-amino-1,2,4-triazole (3AT). Positive colonies were restreaked onto a fresh selection plate, and colony PCR was performed to amplify and sequence the inserted fragments.

For Y2H and Y3H assays, coding sequences of NF-YAs, NF-YBs, NF-YCs, BRC1, and BRC1 derivatives were cloned into pGBKT7 (BD), pGADT7 (AD), or pBridge (BD) vectors. Different combinations of prey and bait constructs were co-transformed into the AH109 yeast strain. Transformants were initially cultured on non-selective SD−Trp/Leu medium to confirm the presence of both constructs. Subsequently, well-growing colonies were spotted on selective media (SD−Trp/Leu/His or SD−Met/Trp/Leu/His) containing 10 mM or 50 mM 3AT to test protein interactions. Plates were incubated at 30 °C, and images were taken after 3 days to assess colony growth. In parallel, transformations with empty vectors were conducted to serve as negative controls. All yeast experiments were performed more than three times with consistent results. Primers used for vector construction are listed in Supplementary Data 5.

EMSA

DNA probes corresponding to the CCAAT box sites in the FUL upstream sequences were synthesized and biotin-labeled at the 3′ end by Integrated DNA Technologies (IDT). EMSA was conducted using the LightShift Chemiluminescent EMSA Kit (ThermoFisher, 20148) following the manufacturer’s instructions. Briefly, 20 fmol of the labeled probe was incubated with the specified recombinant proteins in 1× binding buffer (10 mM Tris, 50 mM KCl, 1 mM DTT, 50 ng/μL Poly dI•dC, pH 7.5) at room temperature for 30 min. For competition assays, 400 fmol (20× excess) or 4 pmol (200× excess) of unlabeled probe was added to the reaction mixture. The experiment was repeated three times with consistent results.

Histochemical and quantitative analysis of GUS activity

For GUS histochemical staining, plant tissues were prefixed in ice-cold 90% acetone for 20 min and subsequently incubated overnight at 37 °C in staining buffer containing 50 mM phosphate buffer (pH 7.0), 0.5 mM K₃Fe(CN)₆, 0.5 mM K₄Fe(CN)₆, 0.05% Triton X-100, and 2 mM 5-bromo-4-chloro-3-indolyl-β-D-glucuronic acid. After staining, tissues were destained in 70% ethanol to remove chlorophyll. Images were acquired using a Zeiss Lumar epifluorescence stereo microscope.

Quantitative analysis of GUS activity was performed as previously described with minor modifications¹⁰⁶. Briefly, approximately 0.1 g of plant tissue was ground to a fine powder in liquid nitrogen and extracted with 0.5 mL of extraction buffer (50 mM phosphate buffer, pH 7.2, 10 mM EDTA, 0.1% SDS, 0.1% Triton X-100, and 1× Complete protease inhibitor cocktail). Total protein concentration was determined using a BCA protein assay kit (ThermoFisher, 23227). Subsequently, 75 µL of total protein extract was mixed with 300 µL of extraction buffer and 375 µL of 2 mM 4-methylumbelliferyl-β-D-glucuronide (Sigma-Aldrich, M9130) and incubated at 37 °C in the dark. Aliquots (100 µL) were collected at 0 and 40 min and immediately mixed with 900 µL of 0.2 M Na₂CO₃ to terminate the reaction. Fluorescence was measured with excitation at 360 nm and emission at 410 nm using a PerkinElmer EnVision plate reader. A standard curve was generated by serial dilution of 4-methylumbelliferone (MU; Sigma-Aldrich, M1381) to quantify MU production in each sample. GUS activity was calculated as the amount of MU produced per minute and normalized to total protein content.

Statistical analysis

Statistical analyses were performed using GraphPad Prism (v.8.3.0). Sample size (n) refers to the number of biological replicates or individual samples. Statistical significance was determined by one-way ANOVA with Tukey’s HSD tests or two-tailed Student’s t-test. P < 0.05 or lower was considered statistically significant, while P > 0.05 was considered non-significant. All sample sizes, statistical tests and P values are indicated in the figures or figure legends. Adobe Illustrator CS6 was used in organizing the images and figures.

Reporting summary

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

Author contributions

C.Y. and Y.Z. conceived the work and designed the experiments; C.Y., M.Y., and J.Z. performed the experiments; C.Y., M.Y., and H.Z. analyzed the data; X.S. and W.L. assisted in data interpretation and manuscript revision; C.Y. and S.L. wrote the manuscript with contributions from all authors.

Peer review

Peer review information

Nature Communications thanks Chunyu Zhang and the other anonymous reviewer for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported in part by the National Natural Science Foundation of China (grant nos. 32000204 to C.Y., 82171844 and 81970898 to Y.Z.) and the China Postdoctoral Science Foundation (grant no. 2019M660012 to C.Y.).

Data availability

The RNA-seq and CUT&Tag data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession PRJNA1466707. All other data supporting the findings of this study are available within the article and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77029-7.