NUCLEAR FACTOR Y, Subunit A (NF-YA) Proteins Positively Regulate Flowering and Act Through FLOWERING LOCUS T

Abstract

Photoperiod dependent flowering is one of several mechanisms used by plants to initiate the developmental transition from vegetative growth to reproductive growth. The NUCLEAR FACTOR Y (NF-Y) transcription factors are heterotrimeric complexes composed of NF-YA and histone-fold domain (HFD) containing NF-YB/NF-YC, that initiate photoperiod-dependent flowering by cooperatively interacting with CONSTANS (CO) to drive the expression of FLOWERING LOCUS T (FT). This involves NF-Y and CO binding at distal CCAAT and proximal “CORE” elements, respectively, in the FT promoter. While this is well established for the HFD subunits, there remains some question over the potential role of NF-YA as either positive or negative regulators of this process. Here we provide strong support, in the form of genetic and biochemical analyses, that NF-YA, in complex with NF-YB/NF-YC proteins, can directly bind the distal CCAAT box in the FT promoter and are positive regulators of flowering in an FT-dependent manner.

Author Summary

For plants to have reproductive success, they must time their flowering with the most beneficial biotic and abiotic environmental conditions—after all, reproductive success would likely be low if flowers developed when pollinators were not present or freezing temperatures were on the horizon. Proper timing mechanisms for flowering vary significantly between different species, but can be connected to a variety of environmental cues, including water availability, temperature, and day length. Numerous labs have studied the molecular aspects of these timing mechanisms and discovered that many of these pathways converge on the gene FLOWERING LOCUS T (FT). This means that understanding precisely how this gene is regulated can teach us a lot about many plant species in both natural and agricultural settings. In the current study, we focus on day length as an essential cue for flowering in the plant species Arabidopsis thaliana. We further unravel the complexity of FT regulation by clarifying the roles of NUCLEAR FACTOR Y genes in day length perception.

Introduction

Plants undergo numerous developmental phase changes that are both species specific and intimately linked to the environments in which they evolved. One of the most important phase changes—as evidenced by the numerous pathways controlling the process—is the transition from vegetative to reproductive growth (recently reviewed in [1]). For many plant species, a potent trigger of the transition to reproductive growth is photoperiod-dependent flowering. Photoperiod-dependent species use the relative length of day and night to either activate or repress flowering such that it is timed with the appropriate environmental conditions to maximize reproductive success.

The model plant Arabidopsis thaliana (Arabidopsis) is a so-called long day plant; that is, it flowers rapidly when days are longer than ~12 hrs [2–5]. Central to measuring photoperiod is the circadian regulation of CONSTANS (CO) transcription and the light-mediated regulation of CO protein accumulation [6]. CO protein is stabilized in long days and is able to bind and transcriptionally activate FLOWERING LOCUS T (FT) [7, 8]. FT protein is the principal mobile hormone—or “florigen”—that travels from leaves, where the photoperiod signal is perceived, to the shoot apex, where the floral transition occurs [9–12]. In the shoot apex, FT activates its downstream targets, which includes APETALA 1 (AP1) and SUPPRESSOR OF CONSTANS 1 (SOC1) [13, 14]. Members of the heterotrimeric NUCLEAR FACTOR-Y (NF-Y) transcription factor family are required for activation of the FT promoter, thus initiating the downstream events leading to the floral transition [15–20].

NF-Y transcription factors are composed of three independent protein families, NF-YA, NF-YB, and NF-YC. To activate target genes, NF-YB and NF-YC dimerize in the cytoplasm and move to the nucleus where the heterodimer interacts with NF-YA to create the DNA-binding, heterotrimeric NF-Y transcription factor [21–24]. NF-Y binding is widely regarded as sequence specific to the evolutionarily conserved CCAAT motifs, with some modified sites having been reported [17, 25, 26]. All direct contacts with the pentanucleotide are made by NF-YA, while the NF-YB/NF-YC dimer primarily makes non-sequence specific contacts in adjacent regions, stabilizing the complex [27]. While ubiquitous to eukaryotes NF-Y subunits have undergone an extensive expansion in plants [28, 29]. For example, Arabidopsis has ten members of each NF-Y gene family [29].

Several NF-YB and NF-YC subunits have been demonstrated to regulate photoperiod dependent flowering [15, 18–20, 30, 31]. Briefly, nf-yb2 nf-yb3 double and nf-yc3 nf-yc4 nf-yc9 triple mutants flower very late under normally inductive photoperiods [19]. In both cases, the single mutants have either no effect or comparatively mild effects on flowering time, indicating overlapping functions for these family members. NF-YB and NF-YC proteins can physically interact with CO and loss of function mutations lead to FT expression downregulation [15, 18–20, 30]. Finally, genetic and biochemical data suggest that NF-Y complexes bind the FT promoter at a distal CCAAT box (-5.3kb from start codon), while CO binds several clustered proximal CO regulatory elements (CORE—approx. -200bp upstream from start). Chromatin loops stabilize the interactions between these two distally separated, DNA-bound complexes [8, 16, 32, 33].

In light of HFD (Histone Fold Domain dimer; NF-YB/NF-YC) interactions with CO in photoperiod-dependent flowering, immediate questions are whether NF-YA proteins are regulators of photoperiod-dependent flowering and whether this is CO-dependent and exerted through regulation of FT. Related to NF-YA roles in flowering, initial reports demonstrated that they can negatively regulate flowering as overexpression of some NF-YA genes caused late flowering [20, 34]. Because NF-YA and CO proteins share a region of sequence homology, one possibility is that they compete for occupancy on NF-YB/C dimers. In this scenario, NF-YA and CO might play opposing negative and positive roles, respectively. However, recent reports suggest a more complex scenario, given 1) Genetic evidence for the importance of the -5.3kb FT CCAAT box in flowering [16, 32, 33]; 2) DNA-bound mammalian NF-Y crystal structure showing that NF-YA makes the direct contacts with the CCAAT box and that CO contains key differences in amino acids necessary for these contacts [16, 27, 35, 36]; and 3) Both direct and indirect evidence that CO directly binds CORE sites on FT [7, 8].

Here we address the roles of NF-YA proteins in FT binding, expression regulation, and photoperiod-dependent flowering time. Using a combination of genetic and biochemical approaches, we show complete NF-Y complexes, including NF-YA, bound to the -5.3kb FT CCAAT box. We further demonstrate that NF-YA and NF-YB constructs that can drive early flowering, and perform this activity in an FT-dependent manner. Collectively our data indicate that NF-YA acts as a positive regulators of flowering and that FT is a key regulatory target of NF-Y/CO complexes in the photoperiod-dependent flowering pathway.

Results

NF-YA genes can be positive regulators of photoperiod dependent flowering

To identify NF-YAs involved in flowering, we first examined constitutive overexpression (35S promoter) in first generation (T1) transgenic plant lines for each of the 10 Arabidopsis NF-YA genes (lines described in [37]). We observed that p35S:NF-YA2 and p35S:NF-YA6 expressing plants consistently flowered earlier than Col-0. Nevertheless, confident interpretations of these data were complicated by the pleiotropic, dwarf phenotypes in most overexpressing lines. In fact, lines that constitutively overexpressed NF-YA6 were infertile and did not survive (as previously described, [37]). We were able to isolate and quantify stable, third generation transgenic p35S:NF-YA2 lines and compare them to several other stable lines for constitutively expressed NF-YA genes (Fig 1A). Two independent p35S:NF-YA2 lines flowered early (~10 leaves, compared to 13 for wild type Col-0 plants), while overexpression of other NF-YA genes either did not alter flowering or actually caused modestly later flowering. This is consistent with the original observations of Wenkel et al., 2006 [20]. We note that all of these plant lines showed similar dwarf phenotypes, suggesting that our flowering time observations were not directly correlated with this phenotype.

Fig 1. NF-YA2 is a positive regulator of photoperiod dependent flowering.

A) Flowering time quantification of two independent plant lines each (plant lines 1 and 2) for p35S:NF-YA2 (white bars), p35S:NF-YA7 (light grey bars), p35S:NF-YA8 (grey bars), and p35S:NF-YA9 (dark grey bars) (n≥12/line). B) Flowering time quantification of two independent pNF-YA2:NF-YA2 plant lines (n≥24). C) The expression pattern of pNF-YA2:GUS in leaves of 10 day old plants. D) Relative transcript abundance of CO, FT, and AP1. Asterisks in 1A and 1B represent significant differences derived from one-way ANOVA (P < 0.05) followed by Dunnett’s multiple comparison post hoc tests against Col-0. Asterisks in 1D represent significant differences derived from Student’s T-tests (P < 0.05). All experiments were repeated with identical results.

To avoid the pleiotropic effects from ectopically overexpressing NF-YA2, we additionally generated stable, native promoter transgenic plant lines (pA2:NF-YA2). Presumably due to position effects, some of these lines had high relative transcript abundance of NF-YA2 (~60 fold overexpressed) and were early flowering (Fig 1B, S1 Fig). Interestingly, these plants appeared phenotypically normal, suggesting that the dwarf phenotypes of p35S-driven lines is more related to ectopic expression than overexpression, per se. Note that our previous research on NF-Y:GUS expression patterns showed that both NF-YA2 and NF-YA6 had very strong vascular expression, consistent with the expected localization of floral promoting genes (Fig 1C and [19, 32, 33, 35, 38, 39]).

As discussed above, previous reports suggest that CO, NF-YB and NF-YC regulate flowering primarily by controlling FT expression which, in turn, rapidly upregulates AP1 [18, 19, 30, 32, 33, 40–42]. We used the stable pNF-YA2:NF-YA2-1 plant line to test if NF-YA2 regulates the same set of genes. We used the time points of seven and nine days after germination because they correlate with the initiation of flowering signals in long day grown plants [38]. NF-YB and NF-YC do not affect the relative transcript abundance of CO [18, 19, 30]; likewise, CO was not misregulated in the NF-YA2 overexpressor (Fig 1C). However, the relative transcript abundance of FT was upregulated in seven day old pNF-YA2:NF-YA2-1 plants, which was followed by significant AP1 upregulation by day nine. We did not see increased relative transcript abundance of FT in day 9, possibly due to negative feedback as downstream targets in the signaling cascade, such as AP1, are activated. We note that the moderate upregulation of FT transcript abundance due to NF-YA2 overexpression is consistent with NF-YB2 overexpression [18]. These results suggest that NF-YA2, like its NF-YB and NF-YC counterparts, regulates flowering by controlling FT expression.

The NF-YB2^E65R mutation prevents NF-YA subunits from entering into NF-Y complexes

Because of the apparent difficulties in working directly with NF-YAs, the likely overlapping functionality between family members in flowering (e.g., Hou reports that nf-ya2 mutants have no flowering delay, [17]), and lethality [43, 44], we decided to indirectly manipulate NF-YA function by altering its ability to interact with the HFD dimer. In mammals, the NF-YB^E92R mutant protein specifically loses interaction with NF-YA, but not NF-YC [22]. Crystal structure analysis of the NF-Y complex demonstrated that this glutamic acid makes multiple contacts with NF-YA Arg249 and Arg253 [27]. Alignments between human and Arabidopsis NF-YB proteins show that this glutamic acid (E65 in Arabidopsis NF-YB2) is completely conserved (Fig 2A) and examination of other published alignments also confirm this conservation in the monocot lineage [35, 45–47]. Thus, we reasoned that NF-YB2^E65R mutations would eliminate the ability of NF-YA to enter floral promoting NF-Y complexes and allow us to further test the hypothesis that NF-YA proteins are positive regulators of photoperiod-dependent flowering.

Fig 2. NF-YB2^E65R loses interaction with NF-YA subunits.

A) Alignment of the core domain of human and Arabidopsis NF-YB subunits. * marks the position of the conserved glutamic acid required for interaction with NF-YA in humans [27]. B) NF-YB2 and NF-YB2^E65R interact with NF-YC3, NF-YC4, and NF-YC9 in Y2H assays. C) NF-YB2, but not NF-YB2^E65R, interacts with NF-YA2 when NF-YC9 is expressed using a bridge vector in yeast three-hybrid assays. DBD: DNA binding domain, AD: activation domain, EV: empty vector control.

We first used yeast two hybrid assays to test if NF-YB2^E65R could interact with NF-YC3, NF-YC4, and NF-YC9 and found that both NF-YB2 and NF-YB2^E65R were able to physically interact with the NF-YCs (Fig 2B). Since NF-YA heterotrimerizes with HFD dimers and not individually with NF-YB or NF-YC [48], we used yeast three hybrid assays to test the ability of NF-YA2 to enter into a complex with NF-YB2^E65R and NF-YC9 (Fig 2C). As predicted, NF-YA2/NF-YB2/NF-YC9 complexes formed, but the NF-YB2^E65R variant prevented formation of the trimeric NF-Y complex. Thus, the NF-YB2^E65R provides a powerful genetic tool to test the requirement for NF-YA in photoperiod-dependent flowering.

The NF-YB2^E65R mutation prevents rescue of a late flowering nf-yb2 nf-yb3 mutant

As previously described [16], here and throughout this study, we examined T1 plants as it gave a better representation of flowering time responses by eliminating bias associated with the selection of individual transgenes. For each transgene we examined 15–20 individual plants and for a few key lines we generated two independent T3 transgenic lines from individuals that were representative of the observed T1 average plant.

We predicted that p35S:NF-YB2^E65R would be unable to drive early flowering in wild type Col-0 or rescue the nf-yb2 nf-yb3 late flowering phenotype. We tested this by overexpressing both p35S:NF-YB2 and p35S:NF-YB2^E65R in each background and measuring flowering responses in 15–20 randomly selected T1 plants for each construct. We found that p35S:NF-YB2 showed a trend towards earlier flowering in Col-0, but only caused significantly earlier flowering in a subset of independent experiments (Fig 3A, non-significant example shown). However, p35S:NF-YB2 nf-yb2 nf-yb3 plants flowered ~20 leaves earlier than the parental mutant (Fig 3B–3D). With the p35S:NF-YB2^E65R version, Col-0 actually flowered significantly later than normal (likely indicating dominant interference with the endogenous complexes) and there was no rescue of the nf-yb2 nf-yb3 late flowering phenotype (Fig 3A–3D).

Fig 3. p35S:NF-YB2^E65R cannot rescue the nf-yb2 nf-yb3 late flowering phenotype.

A) Flowering time quantification of 15–20 randomly selected T1 p35S:NF-YB2 and p35S:NF-YB2^E65R plants in the Col-0 background. B) Flowering time quantification of 15–20 randomly selected T1 p35S:NF-YB2 and p35S:NF-YB2^E65R plants in the nf-yb2 nf-yb3 background. C) Flowering time quantification of two independent, stable T3 p35S:NF-YB2 and p35S:NF-YB2^E65R plant lines in the nf-yb2 nf-yb3 background (n≥12). D) Representative plants of p35S:NF-YB2 and p35S:NF-YB2^E65R in the nf-yb2 nf-yb3 background. E) Relative transcript abundance of NF-YB2, FT and AP1 in stable T3 p35S:NF-YB2 and p35S:NF-YB2^E65R plants in the nf-yb2 nf-yb3 background. Asterisks in 3A, 3B and 3C represent significant differences derived from one-way ANOVA (P < 0.05) followed by Dunnett’s multiple comparison post hoc tests against nf-yb2 nf-yb3.

To confirm that NF-YB^E65R was localizing properly, we compared plants expressing NF-YB2-YFP and NF-YB2^E65R-YFP and found that both had identical nuclear localization patterns (S2A Fig). Additionally, we measured NF-YB protein accumulation in late flowering p35S:NF-YB2^E65R T1 plants (all >31 leaves at flowering) compared to a well-characterized, stable, early flowering p35S:NF-YB2 line (all proteins were translationally fused to the HA epitope). The p35S:NF-YB2^E65R T1 lines showed the expected variation in NF-YB protein accumulation; note that even lines that strongly accumulated NF-YB2^E65R could not rescue late flowering (S2B and S2C Fig; e.g., in S2B Fig compare protein accumulation in p35S:NF-YB2^E65R lines 6, 10, 11, and 12 to the stable p35S:NF-YB2 line). Stable, single insertion T3 lines, selected based on high NF-YB2^E65R protein accumulation, showed the same pattern of late flowering (Fig 3C and S2D Fig). Finally, we compared stable p35S:NF-YB2 nf-yb2 nf-yb3 and p35S:NF-YB2^E65R nf-yb2 nf-yb3 for relative transcript abundance of NF-YB2, FT, and AP1 (Fig 3E). Although both lines had very high, ~equivalent NF-YB2 transcript abundance, p35S:NF-YB2 resulted in increased FT and AP1 transcript abundance while p35S:NF-YB2^E65R significantly suppressed both. Collectively, this data shows that the p35S:NF-YB2^E65R plant lines are not able to drive the typical flowering responses associated with NF-YB2 overexpression and a possible explanation for this would be the loss of interaction with NF-YA.

NF-YA2 and NF-YA6 heterotrimerize with NF-YB2 and NF-YC3 in vitro to bind the -5.3kb CCAAT box

We previously showed that NF-YB2 and NF-YC3, together with mouse NF-YA, are able to bind a 31bp, CCAAT-containing oligonucleotide from the FT -5.3kb site [16]. At that time we were unsure of the likely Arabidopsis NF-YA(s) involved in flowering: with the data presented here and a recent publication [17] showing that NF-YA2 and NF-YA6 can act as positive regulators of flowering, we used EMSA to test if NF-YA2 and NF-YA6 are able to bind a probe encompassing the -5.3kb CCAAT box on FT. In the presence of NF-YB2/NF-YC3 dimers, NF-YA2 and NF-YA6 bound the CCAAT probe in a concentration-specific manner (Fig 4). However, neither NF-YA2 nor NF-YA6 could individually bind the CCAAT probe. Further, CO did not bind the CCAAT probe, individually or in the presence of the NF-YB2/NF-YC3 dimer (Fig 4 and S3 Fig). We additionally tested equivalent concentrations of NF-YA2 with NF-YB2^E65R/NF-YC3 and found that this combination completely lost the ability to bind the CCAAT probe. This data demonstrates that plant NF-Y complexes interact with and bind the FT -5.3kb CCAAT box in a manner that is similar, if not identical, to their mammalian counterparts.

Fig 4. NF-YA2 and NF-YA6 bind the FT -5.3kb CCAAT box as a trimer with NF-YB2 and NF-YC3.

NF-Y trimerization and FT CCAAT binding was assessed by EMSA analysis. An FT CCAAT probe was incubated with wild type (WT, lanes 2–8; 20) or E65R mutant (B2^E65R, lanes 15–18; 21) NF-YB2/NF-YC3 dimers (60 nM) in the presence of NF-YA2 (lanes 3–5; 16–18), or NF-YA6 (lanes 6–8) at increasing molar ratios (3, 4.5 or 6 fold), or CO (lanes 20, 21; 6 fold molar ratio). As controls, NF-YA2 (lane 9), NF-YA6 (lane 10), or CO (lane 22) were incubated alone with the probe, at the highest concentration of the dose curve (360 nM), in the absence of NF-YB2/NF-YC3. Lanes 1, 11, 14, 19: probe alone, without protein additions; lanes 12, 13: empty lanes. The NF-Y/DNA complex is indicated by a labelled arrowhead. fp: free probe.

p35S:NF-YB2^E65R fused to a strong activation domain is not able to induce flowering in a CONSTANS-deficient mutant

A potential criticism of using NF-YB2^E65R as a tool to demonstrate an NF-YA requirement in flowering is that we do not know if mutant NF-YB had altered interactions with other components involved in photoperiod-dependent flowering. In particular, we did not know if it might impact CO recruitment or binding to its CORE site. One approach to test this question might be a Y2H assay between NF-YB2 or NF-YB2^E65R and CO. However, as we have previously published, we are unable to detect a direct interaction between NF-YB and CO or CO-Like proteins in Y2H assays [19]. We additionally attempted to replicate the Y3H experiments above, replacing NF-YA for CO, but did not observe yeast growth with either NF-YB2 or NF-YB2^E65R. However, we were concerned that this lack of a Y3H interaction could simply be a technical problem with the assay. Therefore, to further examine the potential impact of the NF-YB2^E65R on CO function, we extended previous experiments by Tiwari et al. [49]. Here they showed that NF-YB2 overexpression could not rescue the late flowering phenotype of a CO loss of function mutant (co-9 allele); alternatively, when a strong transcriptional activation domain (called EDLL) was fused to NF-YB2, it was able to drive very early flowering in co-9 plants. Therefore, we reasoned that if the NF-YB2^E65R flowering responses were due to loss of interaction with CO, NF-YB2^E65R-EDLL would be able to drive early flowering in a co mutant. On the other hand, if NF-YA interactions were relevant in flowering, we should expect that an NF-YB2^E65R-EDLL construct would not be able to drive early flowering or rescue a co mutant.

We first examined overexpressed (35S) NF-YB2-EDLL in the same co-9 loss of function allele. Consistent with Tiwari et al., we found that co-9 plants transgenic for p35S:NF-YB2-EDLL flowered as early as 8–15 leaves (the co-9 average was 40 leaves with no plants earlier than 30 leaves). However, co-9 plants expressing either p35S:NF-YB2 or p35S:NF-YB2^E65R-EDLL did not flower early and were indistinguishable from co-9. Here the impacts of the E65R mutation cannot be attributed to loss of interaction with CO. To further examine the E65R mutation, we additionally examined these constructs in Col-0 and the co-2 mutant (Ler ecotype, Fig 5A and 5B): once again, while NF-YB2 alone did not drive early flowering, NF-YB2-EDLL overexpressing plants were consistently earlier. However, in each case, NF-YB2^E65R-EDLL either caused later flowering (presumably the dominant negative effect suggested above, Fig 3A) or had no effect. We then used the nf-yb2 nf-yb3 background where NF-YB2 plants flowered at a mean of ~21 leaves and NF-YB2-EDLL flowered at ~12 leaves (Fig 5C); NF-YB2^E65R-EDLL was once again unable to alter flowering time. Short day grown plants, which mimic a co mutant because CO is unable to accumulate [2], told the same story—NF-YB2-EDLL, but not NF-YB2^E65R-EDLL, caused earlier flowering (Fig 5D). Finally, we repeated the entire transgenic panel in the loss of function ft-10 mutant (Fig 5E). Importantly, all constructs, including NF-YB2-EDLL, failed to cause significantly earlier flowering in the ft-10 genetic background. Collectively, this data adds additional evidence for NF-YA proteins as positive, FT-dependent regulators of photoperiod-dependent flowering.

Fig 5. NF-YB2-EDLL, but not NF-YB2^E65R-EDLL, rescues late flowering in an FT-dependent manner.

Flowering time quantification for 15–20 randomly selected T1 plants of p35S:NF-YB2, p35S:NF-YB2-EDLL, and p35S:NF-YB2^E65R-EDLL in A) Col-0 B) co-2 C) b2b3 D) short days E) ft-10. Asterisks represent significant differences derived from one-way ANOVA (P < 0.05) followed by Bonferroni’s multiple comparison tests (_*** P < 0.001; * P < 0.05).

NF-YA2-EDLL induces flowering in a CONSTANS-deficient mutant

We additionally hypothesized that if NF-YA2 is able to interact with NF-YB/NF-YC dimers on the FT promoter, attaching the EDLL domain to the pNF-YA2:NF-YA2 construct would also induce flowering in co mutants. If true, this would significantly extend the NF-YB2^E65R and EMSA results above, ameliorating possible concerns about relying on NF-YB2^E65R as a proxy measure of NF-YA function. Again, we first tested flowering responses of 15–20 randomly selected T1 plants in the Col-0 background. As expected, both pNF-YA2:NF-YA2 and pNF-YA2:NF-YA2-EDLL drove earlier flowering in Col-0 (Fig 6A). Alternatively, in the co-2 background, pNF-YA2:NF-YA2-EDLL induced much earlier flowering (~20 leaves earlier than co-2), whereas the control pNF-YA2:NF-YA2 did not (Fig 6B). As with NF-YB2-EDLL (Fig 5E), NF-YA2-EDLL was completely unable to induce flowering in the ft-10 background (Fig 6C), indicating once again an FT-dependent, positive role for NF-YA proteins in flowering.

Fig 6. pNF-YA2:NF-YA2-EDLL can induce flowering in the absence of CO.

Flowering time quantification for 15–20 randomly selected T1 plants in A) Col-0, B) co-2, and C) ft-10. Asterisks represent significant differences derived from one-way ANOVA (P < 0.05) followed by Bonferroni’s multiple comparison tests.

Discussion

Our initial understanding of NF-Y roles in flowering was primarily driven by evidence of physical interactions between individual NF-Y subunits and CO, as well as in planta overexpression analyses [15, 20]. Thereafter, loss of function mutations in HFD subunits identified specific NF-YB and NF-YC genes involved in flowering [18, 30, 50]. Demonstrating roles for NF-YAs has proven more difficult, since they appear to have redundant functions, mutants of a few members are lethal (including NF-YA2) and overexpressing them leads to substantially deleterious pleiotropic effects [37, 43, 44, 51, 52]. Here we have attempted to work around these difficulties with a variety of biochemical and genetic approaches. We provide a compelling body of evidence that NF-YA2 and NF-YA6, and perhaps other NF-YAs, can activate FT expression, and are FT-dependent, positive regulators of flowering.

Previously, NF-YAs were believed to act as negative regulators of flowering, because overexpression of two NF-YA genes, NF-YA1 and NF-YA4, led to later flowering [20]. We noticed the same response with NF-YA7 and NF-YA9 overexpressors. In another study by Leyva-Gonzalez [51], this was also the outcome of generalized overexpression of NF-YAs. NF-YA2 has also been demonstrated to be a negative regulator of a novel flowering response pathway, which is independent of the photoperiod dependent pathway, termed stress-mediated flowering responses [34]. Nevertheless, our observation of early flowering in NF-YA2 overexpression lines is consistent with those recently reported [17]. Although, Hou et al. [17] suggested that NF-YA2 was a positive regulator of flowering time, it was suggested that this was mediated by interaction with a novel, non-CCAAT cis regulatory element called NFYBE in the SOC1 promoter, and not the binding and regulation of FT expression. This seems at odds with existing evidence, as well as experiments presented here.

As reported for co mutants [40], multiple groups have demonstrated that nf-yb and nf-yc mutants also had strongly reduced FT expression and that these reductions were directly correlated with alterations in flowering time [18–20, 30, 53]. Likewise, overexpression of NF-YB and NF-YC genes was associated with FT upregulation [18, 30, 49, 54, 55]. Mutations in cis-regulatory elements bound by either CO or NF-Y complexes in the FT promoter (CCAAT and/or CORE, respectively) also reduced FT expression in a manner that was directly correlated with the severity of flowering delays [16, 18, 19, 32]. Further, constitutive overexpression of CO drove increased FT expression and early flowering, but these phenotypes were strongly reduced or eliminated in nf-yb and nf-yc mutants or when the -5.3kb CCAAT site was eliminated [19, 32, 49]. Finally, multiple labs have shown in vivo and in vitro binding of NF-Y and CO proteins to the FT promoter and mutations in the associated CCAAT and CORE regulatory elements additively reduced FT expression and delayed flowering [7, 8, 16, 32]. Thus, it remains very well-supported that photoperiod-dependent flowering is mediated through direct regulation of FT by CO and NF-Y complexes.

Elegant genetic experiments previously demonstrated that SOC1 activation is downstream of FT [42]. Therefore, if NF-Ys are directly binding and activating SOC1 to activate photoperiod-dependent flowering, FT loss of function alleles (such as ft-10 used here) should not have impaired this function. However, we find that p35S:NF-YA-EDLL and p35S:NF-YB-EDLL cannot drive early flowering in the absence of FT, strongly suggesting that SOC1 is not their only target in photoperiod-dependent flowering. We do not rule out the possibility that the NF-Y are also involved in the direct regulation of SOC1 and this may be an important element of the gibberellic acid flowering pathway [56]. However, regulation of SOC1 alone cannot readily explain the flowering phenotypes discussed here.

Regulation of the FT promoter is influenced by a plethora of pathways and numerous cis-regulatory elements continue to emerge [16, 32, 33, 57]. One of these is the -5.3kb CCAAT enhancer site, where both deletions and mutations significantly delay flowering time [16, 32, 58]. We provide here formal in vitro evidence that complexes formed by NF-YA2 and NF-YA6, associated with NF-YB2/NF-YC3, robustly and specifically bind to this site. Interestingly, the phenotype of the -5.3kb CCAAT mutant was not as strong as those from nf-y HFD loss of function alleles [16, 18], implying that there must be additional CCAAT sites bound by the NF-Y trimer in the FT promoter or that NF-Y subunits also regulate non-CCAAT sites. Another set of important sites responsible for CO activation, CORE, are in the proximal promoter [8, 32]. Indeed, the near complete loss of photoperiod-dependent flowering responses in nf-yb2 nf-yb3 and co mutants strongly argues that NF-Y complexes and CO must be necessary for function at both cis-regulatory regions. In keeping with this, we recently showed that NF-Y, bound to the -5.3kb CCAAT, and CO, bound to CORE sites, physically interact via a chromatin loop. Further, simultaneous mutations in the -5.3kb CCAAT, CORE1 and CORE2 sites in the FT promoter nearly eliminated rescue of an ft-10 mutant [16]. The importance of the -5.3 kb CCAAT element implies a role of the sequence-specific subunit NF-YA; however, the interactions of the HFD subunits with CO, and the resulting enhancer-promoter connections through CORE, made the direct demonstration of NF-YA function in FT expression and flowering all the more important.

NF-YB2^E65R overexpressors were not able to rescue the late flowering phenotype of the nf-yb2 nf-yb3 mutant. We formally excluded that this was due to expression levels and we could also exclude that the mutant folded incorrectly for two reasons: 1) Recombinant production in E. coli recovered WT and E65R as soluble proteins when co-expressed with NF-YC3, and indeed both were easily purified, and 2) The mutant had a dominant negative effect on flowering time when overexpressed in Col-0 plants. A similar conclusion on the dominant negative nature of the glutamic acid mutation was made for rat NF-YB (CBF-A) in vitro [22], but this is the first demonstration that it could also act in vivo. The likeliest explanation for the dominant negative behavior of NF-YB2^E65R is related to formation of HFD heterodimers impaired in trimer formation, and hence normal NF-Y function—i.e., it is possible that they subtract functional NF-YCs, which would otherwise enter the normal trimerization/CCAAT-binding processes. Obviously, we cannot formally rule out the possibility that the NF-YB2^E65R mutant lost interaction with proteins other than NF-YA and that this resulted in the lack of rescue of late flowering.

To rule out the possibility that the NF-YB2^E65R flowering phenotypes were possibly due to loss of interaction with CO, we used the EDLL transactivation domain. CO was previously demonstrated to provide an activation domain for the NF-Y complex and NF-YB2 was able to drive flowering in the absence of CO when fused to the EDLL activation domain [49]. However, in the current study, p35S:NF-YB2^E65R-EDLL was not able to induce flowering, indicating that while CO provides an activation domain for the NF-Y complex, the HFD dimer is non-functional in the absence of NF-YA. Our EMSA data further connects an NF-YA requirement to the capacity to bind at CCAAT elements. Finally, the flowering phenotypes for pNF-YA2:NF-YA2-EDLL were essentially the same as p35S:NF-YB2-EDLL. Both constructs were able to induce flowering in co mutants, were not able to induce flowering in ft-10 mutants, and drove earlier flowering in Col-0.

Collectively, these data strongly suggest that NF-YA proteins regulate photoperiod-dependent flowering by directly binding and modifying the expression of FT.

Methods

Multiple sequence alignments

Protein sequences were obtained from TAIR (http://www.arabidopsis.org [59] or National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/) and manipulated in TextWrangler (http://www.barebones.com) Multiple sequence alignments were made using ClustalX [60] and shaded within Geneious (http://www.geneious.com/).

Generation of overexpression constructs

The p35S:NF-YB2 and the ten p35S:NF-YA constructs were previously described [37, 47], as was the 35S promoter [61]. NF-YB2^E65R was amplified from cDNA using mutagenic PCR. pNF-YA2:NF-YA2 was amplified using genomic DNA with the promoter region starting approximately 1 KB upstream of the start codon. The proof reading enzyme Pfu Ultra II (cat#600670; Agilent Technologies) was used for PCR reactions and the resulting fragments were ligated into GATEWAY^™ entry vector pENTR/D-TOPO (cat#45–0218; Invitrogen). The EDLL domain [49] was amplified from cDNA and contained Acs1 sites, which were used to clone the EDLL domain into the pENTR/D-TOPO backbone of NF-YB2, NF-YA2 and NF-YB2^E65R entry clones. All entry clones generated were sequenced and other than the point mutation were identical to sequences at TAIR (http://www.arabidopsis.org [59]. Entry clones were sub-cloned into the following destination vectors using the GATEWAY^™ LR Clonease II reaction kit (cat#56485; Invitrogen): NF-YB2^E65R, NF-YB2-EDLL and NF-YB2^E65R-EDLL into pEarlyGate101 (YFP-HA tag) for selection using BASTA resistance [62] and pK7FWG2 (GFP tag) for selection using kanamycin resistance [63]; pNF-YA2:NF-YA2 and pNF-YA2:NF-YA2-EDLL into pEarlyGate301 (HA tag) [62]. S1 Table lists primer sequences used for cloning and mutagenesis.

Plant transformation, cultivation and flowering time experiments

Arabidopsis thaliana ecotype Columbia (Col-0) was the wild type for all experiments, except the co-2 allele in the Ler ecotype. nf-yb2 nf-yb3, ft-10, co-2, and co-9 [42, 47, 64, 65] were previously described. Plants were transformed using Agrobacterium mediated floral dipping [66]. Plants were cultivated in a custom-built walk-in chamber under standard long day conditions (16h light/8h dark) or short day conditions (8h light/ 16h dark) using plant growth conditions previously described [37]. All transgenic plants were selected by spraying with the herbicide glufosinate-ammonium (BASTA), except in the co-9 background where selection was done on 50μg/ml kanamycin plates. Leaf number at flowering was measured as the total number of rosette and cauline leaves on the primary axis at flowering.

Protein expression and purification

The cDNAs encoding for NF-YA2 (aa 134–207) and NF-YA6 (aa 170–237) were obtained by gene synthesis (Eurofins Genomics) and cloned into pnEA/tH [67] by restriction ligation with NdeI and BamHI to obtain C-terminal 6His-tag fusions. The CCT domain of CO (aa 290–352), with the addition of a 5’ ATG, was cloned into pnEA/tH via PCR amplification followed by restriction ligation with XhoI and MunI to obtain C-terminal 6His-tag fusions. Clones were verified by sequence analysis. NF-YB2 mutant cDNA, encoding for aa 24–116 with residue E65 mutated to R (NF-YB2^E65R) was obtained by gene synthesis and subcloned in pET15b to obtain N-terminal 6His-tag fusion. 6His-NF-YB2 or 6His-NF-YB2^E65R/NF-YC3 soluble HFD dimers were produced by co-expression in E. coli BL21(DE3) and purified by ion metal affinity chromatography (IMAC) as described in [68]. NF-YA2-6His, NF-YA6-6His or CO-6His were expressed in BL21(DE3) by IPTG induction (0.4mM IPTG for 4h at 25C) and purified by IMAC (HisSelect, SIGMA-Aldrich) in buffer A (10mM Tris pH 8.0, 400mM NaCl, 2mM MgCl₂, 5mM imidazole). Purified proteins were eluted in Buffer A containing 100mM imidazole, and dialysed against Buffer B (10mM Tris-Cl pH 8.0, 400mM NaCl, 2mM DTT, 10% glycerol).

Electrophoretic Mobility Shift Assays

EMSA analyses were performed essentially as previously described [16, 67, 68]. Heterotrimer formation and CCAAT-box DNA-binding of WT or mutant NF-YB2/NF-YC3 dimers was assessed by addition of purified NF-YAs (or CO) using the Cy5-labeled FT CCAAT probe [16]. DNA binding reactions (16μl) (20nM FT CCAAT probe, 12mM Tris-HCl pH 8.0, 50mM KCl, 62.5mM NaCl, 0.5mM EDTA, 5mM MgCl₂, 2.5mM DTT, 0.2 mg/ml BSA, 5% glycerol, 6.25ng/μl poly dA-dT) were incubated with WT or mutant NF-YB2/NF-YC3 dimers (60nM), with or without NF-YA2 or -YA6 (or CO), as indicated in Fig 4. Proteins were pre-mixed in Buffer B containing 0.1 mg/ml BSA, then added to DNA binding mixes. After 30min incubation at 30C, binding reactions were loaded on 6% polyacrylamide gels and separated by electrophoresis in 0.25X TBE. Fluorescence gel images were obtained and analyzed with a Chemidoc^™ MP system and ImageLab^™ software (Bio-Rad).

Western blot

Total protein was extracted by grinding in lysis buffer (20mM Tris, pH 8.0, 150mM NaCl, 1mM EDTA, pH 8.0, 1% Triton X-100, 1% SDS with fresh 5mM DTT, 10mM protease inhibitor). NF-YB2-YFP/HA and NF-YB2^E65R-YFP/HA were detected using high affinity anti-HA primary antibody (cat#11 867 423 001; Roche) and goat anti-rat secondary antibody (cat#SC-2032; Santa Cruz Biotechnology). Horseradish peroxidase-based ECL plus reagent was used for visualization in a Bio-Rad ChemiDoc XRS imaging system. The membrane was stained with Ponceau S (cat#P3504; Sigma-Aldrich) to determine equivalent loading and transfer efficiency.

Confocal imaging

p35S:NF-YB2-YFP and p35S:NF-YB2^E65R:YFP in nf-yb2 nf-yb3 background, and nf-yb2 nf-yb3 seeds were cold stratified in the dark for 48-h then germinated and grown on B5 media under 24hr light. Six to seven-day-old seedlings were counterstained with propidium iodide (PI) (50μg/mL) for five minutes, washed in DI water for five minutes and whole mounted in fresh DI water on standard slides. Hypocotyls were imaged with an Olympus FluoView 500 using a 60X WLSM objective. XYZ scans were taken with line sequential scanning mode where fluorescent signals were sampled using a filter-based detection system optimized for YFP and PI with chloroplast autofluorescence also detected in the latter. YFP was excited using a 488nm Argon laser whereas PI was excited using a 543nm Helium Neon laser. Approximately 50 serial sections were imaged with a cubic voxel size of 414nm x 414nm x 414nm. Image processing took place in ImageJ (http://rsb.info.nih.gov/ij/) where average intensity projections were taken from YFP and PI channels and merged.

Yeast two-hybrid (Y2H) and three-hybrid (Y3H) analysis

Entry clones of NF-YA2 and NF-YC9, which were previously described [19, 37], were subcloned into pDEST^™22 (Invitrogen) and pTFT1 [69] respectively to obtain an activation domain (AD) and bridge construct. The DNA binding domain (DBD) and AD constructs for NF-YB2, CO and NF-YC9 were previously described [19]. The plasmids were transferred to the yeast strains MaV203 (Invitrogen) for Y2H and PJ69-4α [70] for Y3H analysis. Protein interactions were tested according to the ProQuest^™ manual (Invitrogen). For the X-Gal assay nitrocellulose membranes were frozen in liquid nitrogen and placed on a filter paper saturated with Z-buffer containing X-Gal (5-bromo-4-chloro-3-indoxyl-beta-D-galactopyranoside, Gold Biotechnology, cat#Z4281L). For the synthetic dropout medium lacking the amino acid Histidine, 5mM 3-amino-1,2,4-triazole (3-AT) was added to eliminate nonspecific activation. All experiments were repeated with identical results.

qPCR analysis

Total RNA was collected from seven-day-old or nine-day-old seedlings according to instructions in the E.Z.N.A Plant RNA Kit (cat#R6827-01; Omega Biotek). For each genotype, we analyzed three or four independent biological replicates. First-strand cDNA synthesis was performed as previously described [37]. For qPCR a CFX Connect^™ Real-Time PCR Detection System (Bio-Rad) with the SYBR Green qPCR Master Mix (cat#K0222; Fermentas) was used. Results were analyzed using CFX Manager^™ (Bio-Rad) where samples were normalized to a constitutively expressed reference gene At2G32170 [71]. S1 Table lists primer sequences used for qPCR analysis.

Supporting Information

S1 Table. List of Primers.

(DOCX)

S1 Fig. NF-YA2 is expressed in pNF-YA2:NF-YA2 plants.

Quantification of NF-YA2 expression in pNF-YA2:NF-YA2-1 plants used for qPCR analysis. Asterisks represent significant differences derived from student’s T-test (P < 0.05).

(PDF)

S2 Fig. NF-YB2^E65R is expressed in the nf-yb2 nf-yb3 background.

(A) Confocal images of NF-YB2 and NF-YB2^E65R protein localization in stable plant lines. (B) Protein expression in 12 individual T1 p35S:NF-YB2^E65R plants compared to a stable strongly expressed p35S:NF-YB2 in the nf-yb2 nf-yb3 background. Regardless of the level of expression, p35S:NF-YB2^E65R (flowering at > 31 leaves) was not able to rescue the nf-yb2 nf-yb3 late flowering phenotype, whereas p35S:NF-YB2 was readily able to rescue the late flowering phenotype (flowering at 12 leaves). (C) Protein expression in 12 individual T1 p35S:NF-YB2^E65R plants with individual flowering times shown. (D) Protein expression in two stable plant lines each for p35S:NF-YB2 and p35S:NF-YB2^E65R in the nf-yb2 nf-yb3 background. Note here that all four stable lines had the same approximate level of protein expression, however the p35S:NF-YB2 lines were able to rescue the nf-yb2 nf-yb3 late flowering phenotype, whereas p35S:NF-YB2^E65R were not able to rescue nf-yb2 nf-yb3.

(PDF)

S3 Fig. FT CCAAT is bound by the NF-Y trimer and not by the CO/NF-YB2/NF-YC3 trimer.

EMSA assays were performed using the fluorescently labeled FT CCAAT DNA probe (20 nM). DNA binding was assayed by addition of CO (lanes 2–5) or NF-YAs, as indicated (lanes 6–9; 10–13), at increasing protein concentration (60, 120, 180, or 240 nM) in the presence of the WT NF-YB2/NF-YC3 HFD dimer (WT, 40 nM). As a control, the probe was incubated with the HFD dimer alone (40 nM, lane 1), or with the NF-YA2 protein (YA2, 240 nM, lane 14). The NF-Y/DNA complex is indicated by a labeled arrowhead. fp: free probe.

(PDF)

Data Availability

All relevant data are within the paper and its Supporting Information files.

Funding Statement

This work was funded by the National Science Foundation (US, http://www.nsf.gov/) award 1149822 to BFH. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.