The NF-YC–RGL2 module integrates GA and ABA signalling to regulate seed germination in Arabidopsis

We previously revealed that NF-Y complex regulates flowering time under GA pathway36. As GA-mediated molecular response is critical for seed germination, to investigate the role of Arabidopsis NF-Y genes in germination, we compared the germination rates of the diverse NF-Y mutant seeds in either the presence or absence of the GA biosynthesis inhibitor paclobutrazol (PAC), respectively. Intriguingly, nf-yc3 nf-yc4 nf-yc9 (nf-ycT), the combinatorial null mutant of the closest homologues NF-YC3, NF-YC4 and NF-YC9 in Arabidopsis NF-YC subfamily, similar to the key GA-related germination repressor mutant rgl2 (ref. 16), showed a strong resistance to PAC, but it has no difference in germination rate with the wild-type under mock treatment (Fig. 1a,b). By contrast, there was no significant difference in germination performance between the single or double mutants of NF-YC3/4/9 and the wild type (Supplementary Fig. 1a,b). These results suggest redundant roles of NF-YC3, NF-YC4 and NF-YC9 in repression of the GA-mediated seed germination. Consistently, the seeds of 35S:NF-YC3 and 35S:NF-YC9 exhibited lower germination rates than the wild type in the presence of a low concentration (0.5 μM) of PAC (Supplementary Fig. 1c,d). The analysis of gradient PAC concentrations further confirmed that 35S:NF-YC9 was hypersensitive, and nf-ycT showed reduced sensitivity to PAC, in comparison with the wild type (Fig. 1c). Given that stratification increases Arabidopsis seeds sensitivity to GA during germination38, we also examined germination phenotype of nf-ycT mutant in the absence of stratification. Similar to those with stratification, nf-ycT seeds showed significant resistance to PAC as rgl2 under non-stratification conditions (Supplementary Fig. 2). These results indicated that NF-YC homologues negatively regulate GA-mediated seed germination.

Previous studies have shown that the GA-deficient mutant ga1 fails to germinate, which can be sufficiently recovered by exogenous GA or loss of function of RGL2 (refs 7^,16). Considering that the similar observations on the nf-ycT and rgl2 mutants, and as a P450 enzyme inhibitor, PAC might have undesirable effect on other metabolic pathways including potentially increasing ABA levels through inhibition of ABA catabolism39,40, we thus examined whether NF-YCs could also affect ga1 germination as RGL2. Consistent with previous report41, different concentrations of GA gradually increased the germination rate of ga1. nf-ycT significantly increased GA responsiveness in ga1, whereas overexpression of NF-YC9 decreased that in ga1 (Fig. 1d). Further observations showed that, besides of nf-ycT, the NF-YC double mutants also partially rescued ga1 phenotype although the nf-yc single mutants had less effect on that (Supplementary Fig. 3a,b). Combined with that NF-YC3, NF-YC4 and NF-YC9 genes expressed in germinating seeds (Supplementary Fig. 4), these results support that Arabidopsis NF-YC3, NF-YC4 and NF-YC9 redundantly function as negative regulators in GA-mediated seed germination.

The analogous genetic roles of NF-YCs and RGL2 prompted us to investigate whether NF-YCs functionally associate with RGL2 to regulate seed germination. We first found that AD-RGL2 interacted with BD-NF-YC3, BD-NF-YC4 and BD-NF-YC9 in yeast, respectively (Fig. 2a). Glutathione S-transferase (GST) pull-down assays showed that each His-NF-YC was precipitated by GST-RGL2 but not by GST alone (Fig. 2b), indicating the physical interaction between RGL2 and NF-YCs in vitro. Because NF-YC9-3FLAG fully rescued the PAC reduced-sensitivity phenotype of nf-yc9 and nf-ycT (Supplementary Fig. 5), we thus focused NF-YC9 as representative of NF-YC homologues for further investigation.

To identify whether the functional domains are required for the NF-YC and RGL2 interaction, the various truncated versions of RGL2 and NF-YC9 were used in yeast two-hybrid assays (Fig. 2c). The results showed that deletion of the GRAS domain42 (RGL2ΔG) prevented RGL2 binding to NF-YC9, whereas deletion of the RGL2 amino-terminal (RGL2ΔD) which excluded the entire DELLA domain did not affect the interaction between RGL2 and NF-YC9 (Fig. 2c). On the other hand, RGL2 interacted with full length of NF-YC9 and deletion of the amino-terminal (NF-YC9ΔN) but not with the HFD26 and the amino-terminal fragment (NF-YC9N) of NF-YC9 (Fig. 2c), indicating the carboxy-terminal fragment of NF-YC9 is necessary for interacting with RGL2 at least, although it alone revealed a self-activation in yeast. Therefore, these results suggest that the GRAS domain of RGL2 and carboxy-terminal of NF-YC9 contribute to interaction between RGL2 and NF-YC9, and may be indispensible components in potential biological function of this heterodimer.

We next performed bimolecular fluorescence complementation (BiFC) analysis to examine the interaction between NF-YC9 and RGL2 in plants. The results showed that the interaction fluorescence of NF-YC9-nEYFP with RGL2-cEYFP existed in the cell nuclei, but no YFP signal was detected in the negative control (Fig. 2d). To perform co-immunoprecipitation assay, we further created nf-yc9 rgl2 pNF-YC9:NF-YC9-3FLAG pRGL2:RGL2-6HA homozygous lines in which NF-YC9-3FLAG expressed at comparable levels in the germinating seeds with mock and PAC treatment, whereas RGL2-6HA was only detected in those with PAC (Supplementary Fig. 6). Both NF-YC9 and RGL2 fusion proteins retained the biological function in seeds as they are able to rescue the PAC reduced-sensitivity phenotype of mutants (Supplementary Fig. 5). The co-immunoprecipitation results of PAC-treated nf-yc9 rgl2 pNF-YC9:NF-YC9-3FLAG pRGL2:RGL2-6HA seeds further confirmed the in vivo binding of NF-YC9 to RGL2 (Fig. 2e). Taken together, these data consistently support the direct interactions between the three NF-YC homologues and RGL2 proteins.

Among DELLAs, RGL2 has been reported as the predominant repressor, and RGA, GAI and RGL1 play minor roles in GA-mediated seed germination in Arabidopsis Ler ecotype16,19. Our observations confirmed that the non-germinating phenotype of ga1 was also partially suppressed by loss of RGA function or completely rescued by the rga rgl2 double mutant in Arabidopsis Col ecotype (Supplementary Fig. 7a). Consistent with this, like RGL2, RGA was shown to interact with NF-YC3, NF-YC4 and NF-YC9 in yeast, respectively (Supplementary Fig. 7b), which implies widespread interactions between NF-YCs and DELLAs.

The interaction between NF-YCs and RGL2 in plants suggests that these proteins may function together to regulate seed germination. To verify this hypothesis, we created various combinatorial genetic backgrounds of NF-YC and RGL2 by intercrossing. Investigations of germination rate showed that rgl2 fully suppressed the hypersensitivity of 35S:NF-YC9 to PAC (Fig. 3a,b). Consistently, although 35S:NF-YC9 enhanced the germination inhibition of ga1 at low GA condition (0.01–1 μM), it did not suppress ga1 germination at higher GA level (10 μM GA; Fig. 1d). In addition, 35S:NF-YC9 also rarely affected germination of seeds grown in normal condition (Fig. 3). Since RGL2 proteins are highly accumulated in ga1 or under PAC treatment (Supplementary Fig. 6) and degraded in response to GA by the SCF^SLY1 complex23, these observations indicate that NF-YC function on germination inhibition requires RGL2. On the contrary, similar to rgl2, the nf-yc mutants promote the germination of ga1 or PAC-treated seeds (Fig. 1c,d, Supplementary Fig. 3a,b), supporting that the repressive role of RGL2 in germination is also dependent on NF-YCs. Observation that rgl2 nf-ycT had no significant difference with rgl2 further confirmed the interdependent roles of NF-YCs and RGL2 in germination inhibition (Fig. 3).

Further analysis demonstrated that transcriptional and protein levels of NF-YC3, NF-YC4 and NF-YC9 were rarely affected by rgl2 under mock or PAC treatment (Supplementary Fig. 8a,c). In turn, NF-YCs also had less effect on the transcription and protein accumulation of RGL2 (Supplementary Fig. 8b,c,d). These observations demonstrate that NF-YCs and RGL2 do not regulate each other in the mRNA or protein levels. Together, above data corroborate that NF-YCs and RGL2 interdependently repress seed germination via protein interactions.

To understand how the NF-YC–RGL2 complex functions in repressing seed germination, a genome-wide transcriptomic analysis was carried out using germinating seeds of rgl2, nf-ycT and the wild type (Col) with PAC and Col with mock treatment (Supplementary Data 1). On the basis of the criteria of 1.5-fold cutoff for the genes with 5% false discovery rate, we first identified 1,326, 906 and 632 differentially expressed genes in Col_PAC versus Col_mock, nf-ycT_PAC versus Col_PAC, and rgl2_PAC versus Col_PAC subsets, which are referred to as PAC-, NF-YC- and RGL2-regulated genes (Fig. 4a, Supplementary Data 2). To define whether the PAC-regulated genes respond to GA, we executed a comparative analysis of the PAC-regulated genes profile with a previous microarray data in which the GA-regulated transcripts were identified from Ler wild-type versus ga1-3 germinating seeds43,44, and it showed a high overlap (44.6%) between these two independent data sets despite the different ecotypes and growth conditions used (Supplementary Data 2). Venn diagrams demonstrated that ∼70% (444/632) of RGL2-regulated genes and ∼41% (374/906) of NF-YC-regulated genes were also regulated by PAC mostly in an identical pattern, respectively, while ∼28% (174/632) of RGL2-regulated genes were similarly regulated by NF-YCs (Supplementary Fig. 9a), implying the high correlation of regulation between RGL2 and NF-YCs. Total 142 overlapped genes were coregulated by PAC, NF-YC and RGL2, which are considered as target genes involving in NF-YC- and RGL2-mediated seed germination (Supplementary Data 3). Strikingly, most of these coregulated genes were regulated by NF-YC, RGL2 and PAC in the same direction, and only three genes (2%) were differentially regulated (Fig. 4b, Supplementary Fig. 9b, Supplementary Data 3). Further gene ontology (GO) analysis revealed that these genes were primarily enriched in seed germination, response to hormone stimulus, cell wall modification, transferase activity, transport and other metabolic process (Supplementary Data 3). Interestingly, in the top three GO annotations, more co-upregulated genes were involved in response to ABA (12.2%), whereas lesser of them in cell wall modification (1.2%). The co-downregulated genes were remarkably enriched in cell wall-related process (28.1%) and none of them in response to ABA (0%). By contrast, the two subsets of genes were both involved in GA-mediated seed germination (32.9 and 19.3%, respectively) (Fig. 4c). These data reveal that NF-YCs and RGL2 co-target a set of common genes in response to phytohormone signals, strongly supporting the role of NF-YC-RGL2 in seed germination regulation.

Further quantitative PCR with reverse transcription (RT–PCR) analysis was performed to confirm the regulation of NF-YC–RGL2 on several selected downstream genes. Consistent with the transcriptomic analysis, PAC dramatically induced the expression of ABA responsive genes ABI5, TZF5 and MFT, and repressed that of cell wall-related genes EXP3, EXP9, XTH5 and XTH31, respectively, in the wild-type germinating seeds, whereas these PAC-triggered expression changes were compromised by nf-yc and rgl2. By contrast, these selected genes have largely comparable expression in different genetic backgrounds when grown in normal condition (Fig. 4d). Furthermore, the effect of GA on expression of NF-YC–RGL2 downstream was analysed in various genetic backgrounds. In the absence of GA (under PAC treatment or in ga1 background), 35S:NF-YC9 promoted the expression of ABI5, TZF5 and repressed that of EXP9, XTH31, respectively. However, the transcriptional regulation effect of NF-YCs on the selected genes was significantly compromised by loss of RGL2 or GA application, similarly, the effect of RGL2 on the selected genes in ga1 was also attenuated by loss of NF-YCs or GA application (Fig. 4e,f). These results confirm that NF-YC–RGL2 differentially regulates two subsets of genes that are involved in ABA response and GA-mediated cell wall modification, respectively, to repress seed germination. Meanwhile, it is also intriguing how this complex functions on activation and repression of its downstream. Interestingly, the chromatin immunoprecipitation (ChIP) assay showed that, rather than the direct transcriptional repression on the cell wall-related genes, NF-YC–RGL2 module might directly target ABA responsive gene ABI5 for transcriptional activation (Supplementary Fig. 10).

The binding of NF-YC–RGL2 to chromatin provoked us to speculate whether this complex serves as a transcriptional activator to directly regulate the ABI5 gene. Because NF-Y was reported to specifically bind to the CCAAT-box in promoter of target genes26,45, we analysed the ABI5 genomic DNA and chose 12 fragments (P1 to P12), which covered all six CCAAT-boxes of the ABI5 region, for next examination (Fig. 5a). ChIP analyses of PAC-treated nf-yc9 pNF-YC9:NF-YC9-3FLAG and rgl2 pRGL2:RGL2-6HA seeds revealed that both NF-YC9 and RGL2 were associated with the genomic region near the adjacent fragments P7 and P8 with the highest enrichments (Fig. 5a). ChIP–reChIP analysis further verified that NF-YC9 and RGL2 co-localized to the same region of ABI5 (Fig. 5a). Since P7 and P8 fragments contained two CCAAT elements (designated as CCAAT-2 and CCAAT-3, respectively), to examine whether these elements are involved in the NF-YC–RGL2 regulation on ABI5 expression, we performed transient expression assays using ∼1.8 kb fragment of native or various mutated ABI5 promoters fused to the β-glucuronidase (GUS) reporter gene. The effector constructs of 35S:NF-YC9 and 35S:RGL2 were individually or together transfected with reporters into Arabidopsis mesophyll protoplasts (Fig. 5b). Addition of RGL2 or NF-YC9 activated the expression of ABI5. Notably, in comparison with that expressing RGL2 alone, the higher GUS activity was detected when co-expressing NF-YC9 and RGL2 (Fig. 5b). However, when site-specific mutations (CCAAT to ACATA) were introduced into the CCAAT elements in ABI5 promoter, the expression of ABI5 was strikingly impaired by disruption of the CCAAT-2 or CCAAT-3 (Mut2 or Mut3) but not by Mut1 or Mut4, either each or both NF-YC9 and RGL2 existed (Fig. 5b). These results indicate that the CCAAT elements located at P7 and P8 are essential for NF-YC–RGL2-mediated activation of ABI5. In addition, other DELLA proteins also contribute to ABI5 expression activation together with NF-YC9 in a variable extent (Supplementary Fig. 11). Since the different circumstances between cells of protoplasts and germinating seeds, the biological roles of these DELLAs on ABI5 in plants still need to be further determinated.

Because no DNA binding domain is identified in DELLA proteins, to make clear how RGL2 and NF-YC9 recognize ABI5 promoter, we performed protein–DNA affinity pull-down assay in which a heterotrimeric NF-Y, composed of NF-YC9 and two NF-Y subunits core domains from yeast (NF-YA core and NF-YB core), was used for CCAAT element binding as previously reported46. Interestingly, RGL2 indirectly associated with the ABI5 promoter fragment DNA containing CCAAT-2/3 (ABI5-2,3), but not with the mutated ABI5-2,3 or ABI5 promoter fragment containing CCAAT-4 (ABI5-4), via interacting with NF-YA core/NF-YB core/NF-YC9 complex (Supplementary Fig. 12). However, the DNA affinity of RGL2-NF-YC9 was not detected in vitro (Supplementary Fig. 12), which is consistent with the previous reports that NF-YC subunit has no DNA binding ability26. These results indicate that RGL2, not itself, but via NF-Y complex, recognizes ABI5 promoter region.

To further examine whether NF-YC–RGL2 regulates ABI5 expression through the CCAAT elements during seed germination, we created ABI5:GUS transgenic plants and two mutated lines containing the Mut2 or Mut3 version of CCAAT elements. Among 11 transformants harbouring Mut2 (mABI5:GUS) with PAC treatment, 9 displayed significantly reduced GUS staining in comparison with ABI5:GUS seeds (data not shown). The similar results were observed in the transgenic lines harbouring Mut3 (m3-ABI5:GUS) (Supplementary Fig. 13a,b). These observations verified that CCAAT-2 and CCAAT-3 are critical for GA-mediated ABI5 expression during seed germination. Furthermore, with PAC treatment, the staining of ABI5:GUS seeds was remarkably weaker in both nf-ycT and rgl2 than that in the wild-type background, whereas there was no significant difference of GUS staining among these germinating seeds with mock treatment (Fig. 5c,d). As expected, ABI5:GUS but not mABI5:GUS seeds displayed an increased GUS staining in 35S:NF-YC9 compared with the wild-type background (Fig. 5c,d). Thus, these findings strongly support the idea that NF-YC–RGL2 module activates ABI5 expression via binding to the specific CCAAT elements. It was noted that, although PAC resulted in increased GUS staining in the ABI5:GUS seeds, but it had an opposite effect on the mABI5:GUS seeds probably owing to other unknown regulations caused by the disruption of CCAAT element.

Because NF-YCs and RGL2 interact to directly regulate ABI5 expression, we wondered whether NF-YCs and RGL2 are cooperative and interdependent on such transcriptional regulation. ChIP analyses of rgl2 nf-yc9 pNF-YC9:NF-YC9-3FLAG and nf-ycT rgl2 pRGL2:RGL2-6HA showed that the absence of RGL2 significantly impaired the affinity of NF-YC9 to P7/8 fragments in ABI5 promoter. In turn, binding of RGL2 to the same locus was attenuated by nf-ycT (Fig. 6a). Furthermore, GA application also weakened the DNA binding of NF-YC9 in PAC-treated seeds (Supplementary Fig. 14). The results suggest that NF-YCs and RGL2 cooperatively bind to the CCAAT elements to regulate ABI5 transcription.

ABI5 functions as the central ABA signalling component to repress seed germination, which is epistatic to RGL2 (ref. 23). To investigate the genetic role of NF-YCs in ABI5-mediated inhibition of seed germination, we created abi5 35S:NF-YC9 combinatorial line and found that loss of ABI5 strikingly suppressed the hypersensitivity of 35S:NF-YC9 to PAC in germinating seeds (Fig. 6b,c). Consistent with this, the expression of two ABI5 target genes EM1 and EM6 was decreased in abi5 35S:NF-YC9 compared with 35S:NF-YC9 seeds (Fig. 6d). In turn, overexpression of ABI5 remarkably rescued the PAC-reduced sensitivity of nf-ycT (Fig. 6b,c), and EM1 and EM6 genes expressed at comparable levels in 35S:ABI5 and nf-ycT 35S:ABI5 seeds (Fig. 6d). These results, together with previous report23, suggest that ABI5 act epistatic to NF-YCs and RGL2 during seed germination. We next examined the ABI5 expression in response to immediate RGL2 activity using a steroid-inducible RGL2 (RGL2-GR) in ga1-3 rgl2-1 background. In contrast to no change in ga1-3 rgl2-1, ABI5 expression was rapidly induced by dexamethasone combined with cycloheximide in ga1-3 rgl2-1 35S:RGL2-GR without de novo protein synthesis, whereas it was compromised in ga1-3 rgl2-1 nf-ycT 35S:RGL2-GR (Fig. 6e), providing a further molecular evidence to support genetic relationship between NF-YC–RGL2 and ABI5.

Previous studies showed that GA promotes seed germination by triggering RGL2 degradation, and RGL2 indirectly regulates ABI5 expression through stimulating ABA biosynthesis, thus inhibiting germination23,47,48. Here, our observations support a direct regulation of RGL2 on ABI5. To determine whether such regulation is ABA dependent or not, we examined the expression of ABI5 in ABA synthesis defective mutants aba1-5 and aba2-1. Interestingly, the ABI5 expression was still induced, even though with lesser elevated extent than the wild-type, in PAC-treated aba1 and aba2 seeds compared with mock-treated (Fig. 6f), implying ABI5 might be regulated by GA or RGL2 regardless of ABA biosynthesis.

Germination process contains the rupture of the testa before the endosperm rupture. ABA suppresses seed germination mainly through inhibition of endosperm rupture rather than that of testa rupture, whereas RGL2 accumulation prevents both testa and endosperm rupture under GA deficiency10,23. We here investigated the detailed role of NF-YCs in testa and endosperm rupture. Under normal condition, the germination rates of all examined seeds were 100% at 120 HAS (Fig. 7a,b). In the presence of ABA, compared with the wild-type, nf-ycT, rgl2 and abi5 showed higher testa and endosperm rupture rates, but the 35S:NF-YC9 seeds had the lower rate on those (Fig. 7a,b). Furthermore, 35S:ABI5 dramatically attenuated or abolished the ABA insensitivity of nf-ycT and rgl2, while abi5 35S:NF-YC9 still remained high testa and endosperm rupture rates as abi5. In addition, loss of ABI5 function also significantly rescued non-germinating phenotype of ga1 under either ABA or mock treatment (Fig. 7a,b). The results indicate that, similar to RGL2, NF-YCs repress both testa and endosperm rupture via ABI5 in ABA-mediated seed germination.

To further determine the regulatory effect of NF-YC–RGL2 module on ABA-mediated seed germination, we examined the transcriptional level of several representative NF-YC–RGL2 coregulated genes in different seeds under ABA treatment. As expected, all selected genes in the ABA-treated seeds of nf-ycT, rgl2 and 35S:NF-YC9 were expressed in similar regulatory patterns with those under PAC treatment (Fig. 7c, Fig. 4d,e, Supplementary Fig. 15). These results collectively support the conclusion that NF-YC–RGL2 module integrates ABA and GA signalling to regulate seed germination.

All Arabidopsis plants used in this study are in Col background except for ga1-3 rgl2-1 and ga1-3 rgl2-1 35S:RGL2-GR in Ler background, and abi5-1 in Ws background. The nf-yc3-2 (GK-051E10), nf-yc4-1 (SALK_032163), nf-yc9-1 (SALK_058903), ga1 (SALK_109115), rgl2 (SALK_124231) and rga-28 (SALK_089146) seeds were obtained from the ‘The Arabidopsis Information Resource (TAIR, http://www.arabidopsis.org/↗). Transgenic lines of 35S:NF-YC3-6HA, 35S:NF-YC9-6HA and nf-yc9-1 pNF-Y9:NF-YC9-3FLAG and the mutants of aba1-5 and aba2-1 (refs 13^,36). Each transgenic line used to compare under different genetic backgrounds in this study is the same line which was introduced into various mutants by crossing. The seeds used for germination comparison were harvested in the same batch of plants grown at 22 °C under long days (16 h light/8 h dark). Dry seeds were obtained and stored in a dry condition (25% humidity, 25 °C) for at least 4 weeks of after-ripening before performing the germination test.

Genes referenced in this article can be found in the Arabidopsis Genome Initiative database under the following accession numbers: NF-YC3 (AT1G54830), NF-YC4 (AT5G63470), NF-YC9 (AT1G08970), RGL2 (AT3G03450), RGA (AT2G01570), RGL1 (AT1G66350), GAI (AT1G14920), RGL3 (AT5G17490), ABI5 (AT2G36270), TZF5 (AT5G44260), MFT (AT1G18100), EM1 (AT3G51810), EM6 (AT2G40170), XTH5 (AT5G13870), XTH31 (AT3G44990), EXP3 (AT2G37640), EXP9 (AT5G02260), PP2A (AT1G13320), TIP41-like (AT4G34270) and TUB8 (AT5G23860).

The after-ripened seeds were sterilized and washed with 75% (v/v) ethanol with 0.5% (v/v) Triton X-100 (Sigma-Aldrich) for 1 min, and washed twice with absolute ethanol, then were plated on sterile filter paper for air drying. Subsequently, the sterilized seeds were sown on half-strength MS medium (0.025% MES, pH 5.7) containing 0.8% (w/v) Bacto Agar (Difco/BD) supplemented with 0.01% (w/v) ethanol (Mock), PAC (Sigma-Aldrich), ABA (Sigma-Aldrich), GA₃ (Sigma-Aldrich) or DEX (Sigma-Aldrich) upon the experiment requirement. All the plates were kept at 4 °C in darkness for 3 days for stratification and then transferred to an illumination incubator at 22 °C with 16 h light/8 h dark condition for further analysis. At least 100 seeds for each genotype were used in three biological replicates. The germination event was defined as the first sign of radicle emergence and recorded at different time points until 120 h of incubation.

For the pRGL2:RGL2-6HA construct, an ∼3.7 kb genomic fragment of RGL2 without stop codon was amplified and cloned into the pHY105-6HA binary vector57. To construct 35S:RGL2-6HA, the CDS encoding RGL2 was amplified and cloned into pGreen-35S-6HA. For the ABI5:GUS construct, ∼1.8 kb promoter of ABI5 was cloned into the pHY107 vector harbouring the GUS reporter57. The primers used for plasmid construction are listed in Supplementary Table 1. Transgenic plants harbouring pNF-YC9:NF-YC9-3FLAG were selected on 1/2 MS medium supplemented with gentamicin, while other transgenic plants were selected by basta.

The coding regions of NF-YC3, NF-YC4, NF-YC9, RGL2 and RGA or truncated versions of NF-YC9 and RGL2 were amplified and cloned into pGBKT7 and pGADT7 (Clontech), respectively. The primers used are listed in Supplementary Table 1. Yeast two-hybrid assays were performed using the Yeastmaker Yeast Transformation System 2 (Clontech). Yeast AH109 cells were co-transformed with the specific bait and prey constructs. All yeast transformants were grown on SD/-Trp/-Leu or SD/-Trp/-Leu/-His/-Ade medium for selection or interaction test.

The coding regions of NF-YC9 and RGL2 were cloned into the serial pGreen binary vectors containing C- or N-terminal fusions of EYFP to generate 35S:NF-YC9-nEYFP and 35S:RGL2-cEYFP. Plasmids were co-transformed into Arabidopsis mesophyll protoplasts by the PEG-mediated transient transformation64, and then cultured for 12 h and observed for BiFC analysis using a confocal laser scanning microscope (LSM 510 META, Zeiss).

The coding regions of NF-YC3, NF-YC4, NF-YC9 and RGL2 were cloned into the pQE30 (QIAGEN) and pGEX-4T-1 (Pharmacia) vectors to produce His-NF-YC3, His-NF-YC4, His-NF-YC9 and GST-RGL2 proteins, respectively. Primers used for constructions are listed in Supplementary Table 1. GST and His fusion recombinant proteins were induced by IPTG and expressed in E. coli Rosetta (DE3, Novagen). The soluble His and GST fusion proteins were purified using Ni-NTA agarose beads (30210, QIAGEN) or Glutathione Sepharose Beads (17-0756-01, Amersham Biosciences) according to the manufacturers' instruction. For pull-down assays, 2 μg of His-NF-YCs were incubated in the binding buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl and 1 mM EDTA) with immobilized GST or GST fusion protein at 4 °C for 2 h. After washing with binding buffer, proteins retained on the beads were subsequently resolved by SDS loading buffer and then run SDS–PAGE to detect with anti-His (AbM59012-18-PU, BGI) at a dilution of 1:5,000 or anti-GST antibody (AB101-02, Tiangen) at a dilution of 1:2,000. Uncropped scans of western blot results are shown in Supplementary Fig. 16.

The 5 μM PAC-treated nf-yc9 rgl2 pNF-YC9:NF-YC9-3FLAG pRGL2:RGL2-6HA seeds were kept under light for 12 h. Total proteins were extracted with extraction buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 5 mM DTT, 1% Triton X-100), and were incubated with Protein G PLUS/Protein A-Agarose Suspension (IP10, CALBIOCHEN) plus either anti-FLAG antibody (F3165, Sigma) or preimmune serum (IgG) in the co-immunoprecipitation buffer (50 mM Hepes, pH 7.5, 150 mM KCl, 10 μM ZnSO₄, 5 mM MgCl₂, 1% Triton X-100) at 4 °C for 2 h. After being washed by co-immunoprecipitation buffer three times, the proteins bound to beads were resolved by SDS–PAGE and detected by anti-FLAG (F3165, Sigma) at a dilution of 1:10,000 or anti-HA antibody (sc-7392, Santa Cruz) at a dilution of 1:2,000. Uncropped scans of western blot results are shown in Supplementary Fig. 16.

The after-ripened seeds (4 weeks) harvested in a same batch were grown on half-strength MS medium (0.025% MES, pH 5.7) containing 5 μM PAC under light for 12 h. Total RNA was extracted from harvested seeds by Plant RNA Kit (R6827, Omega) and sent to BGI for RNA-seq analysis. The used RNA samples have been strictly detected upon the RNA sequencing standard and the libraries constructed using Ultra RNA sample preparation kit (Illumina) reached high quality before RNA-sequencing. Sequencing was performed using an Illumina HiSeq2000 according to the standard protocol. Total RNA-Seq reads were mapped to the Arabidopsis TAIR10 genome. The differentially expressed genes were identified by the program Cuffdiff with the criteria set as fold change >1.5 and FDR-adjusted P values <0.05. Three valid biological replicates were used for the transcriptomic analysis. The gene expression patterns were graphically represented in a heat map by cluster analysis tool in Heml software65. GO analysis was performed by the GO Annotation of TAIR66.

The treatment of seeds was performed upon various experiments. Total RNA was extracted using the Plant RNA Kit (Omega) and reverse transcribed using the M-MLV reverse transcriptase (Promega). Quantitative RT–PCR was performed in triplicates on Roche LightCycler480 real-time system with the SYBR Premix ExTaq Mix (DRR041A, TaKaRa) following the manufacturer's instruction. The relative expression level was normalized to that of PP2A internal control. The primers used for gene expression analysis are listed in Supplementary Table 1.

To perform ChIP assays, the nf-yc9 pNF-YC9:NF-YC9-3FLAG, rgl2 pRGL2:RGL2-6HA and the Col wild-type seeds were incubated with mock, 5 μM PAC or 5 μM PAC plus 1 μM GA for 12 h and harvested for fixation. Chromatins were isolated and sonicated to generate DNA fragment with an average size around 250–500 bp. The solubilized chromatins were immunoprecipitated by Protein G PLUS agarose (16–201, Millipore) with anti-FLAG (F3165, Sigma) and anti-HA (sc-7392x, Santa Cruz), and the co-immunoprecipitated DNA was recovered and analysed by quantitative PCR (qPCR) with SYBR Premix ExTaq Mix (DRR041A, TaKaRa Bio). For ChIP–reChIP assays, nf-yc9 rgl2 pNF-YC9:NF-YC9-3FLAG pRGL2:RGL2-6HA and pRGL2:RGL2-6HA seeds were incubated under 5 μM PAC for 12 h and harvested for fixation. The sonicated chromatins were immunoprecipitated by anti-HA agarose conjugate (the first ChIP), and then washed by the ChIP buffer and eluted with 10 mM dithiothreitol (DTT). The eluted chromatins were diluted 20-fold with dilution buffer (1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl at pH 8.1, 150 mM NaCl) and immunoprecipitated again by anti-FLAG (the second ChIP). The recovered DNA was purified and quantified by qPCR. Relative enrichment fold was calculated by normalizing the amount of a target DNA fragment against that of a genomic fragment of a reference gene TUB8, and then against the respective input DNA samples. The enrichment of a PP2A genomic fragment was used as the negative control. The primers used are listed in Supplementary Table 1.

To generate the ABI5:GUS reporter construct, ∼1.8 kb ABI5 promoter was cloned into HY107 containing GUS gene36. The mABI5:GUS constructs carrying mutated CCAAT elements were generated by overlapping PCR reaction from the ABI5:GUS construct. The 35S:NF-YC9-6HA and various DELLA constructs were used as effectors, and a construct containing the firefly luciferase driven by 35S promoter in pGreen-35S was used as an internal control to evaluate the protoplast transfection efficiency. All the primers used for transient expression assay are listed in Supplementary Table 1. Arabidopsis mesophyll protoplasts were prepared, transfected and cultured for 24 h before protein extraction and detection. Activities of GUS and luciferase were measured using microfluorometer (Cary Eclipse, USA). Relative GUS activity was calculated by normalizing against the luciferase activity, and the data presented were the averages of three biological replicates.

The coding regions of yeast NF-YA (158-214aa) and yeast NF-YB (35-127aa) core domains for NF-Y combination67 were cloned into the pQE30 vector to produce fusion proteins of His-Yeast NF-YA core and His-Yeast NF-YB, respectively. Primers used for constructions are listed in Supplementary Table 1. His-NF-YC9 and GST-RGL2 fusion proteins were induced and purified as described above. Protein–DNA affinity assay was conducted with a modified method68. 1 μg of His-NF-YC9, 1 μg of His-Yeast NF-YA and 1 μg of His-Yeast NF-YB proteins were pre-incubated with various DNA fragments (0.5 μg each) produced by PCR using specific primes (Supplementary Table 1) in the binding buffer (100 mM phosphate, 150 mM NaCl and 1 mM EDTA, pH 7.5) at 4 °C for 4 h. Then, the incubated protein–DNA samples were mixed with GST or GST-RGL2 immobilized by Glutathione Sepharose Beads (17-0756-01, Amersham Biosciences) at 4 °C for additional 4 h. After being washed with binding buffer three times, proteins and DNA retained on the beads were subsequently eluted by elution buffer (2% SDS, 50 mM Tris-HCl, 150 mM NaCl and 1 mM EDTA, pH 8.0) at 65 °C for 15 min twice, and the DNA pulled down was extracted and quantified by qPCR using specific primers (Supplementary Table 1). The protein–DNA binding efficiency was calculated by normalizing the amount of DNA precipitated by proteins against that of the respective input DNA, and PP2A genomic DNA fragment was amplified and used as an internal control in this assay. The precipitated proteins bound to beads were detected by anti-His or anti-GST antibody as described above. Uncropped scans of western blot results are shown in Supplementary Fig. 16.

The RNA-seq data used in this study have been deposited in the Gene Expression Omnibus database under accession number GSE84753↗. All other data supporting the findings of this study are available within the article and its Supplementary Information files or on request from the corresponding author.

The RNA-seq data used in this study have been deposited in the Gene Expression Omnibus database under accession number GSE84753↗. All other data supporting the findings of this study are available within the article and its Supplementary Information files or on request from the corresponding author.