The Arabidopsis Mediator CDK8 module genes CCT (MED12) and GCT (MED13) are global regulators of developmental phase transitions

Temporal coordination of developmental programs underlies many processes in plants and animals and plays a major role in evolution. For example, it is hypothesized that an extended phase of rapid brain growth (a characteristic of juvenile development) played a key role in the evolution of humans from their primate ancestors (Gould, 1977). In flowering plants, changes in the relative timing of juvenile and adult development can have major effects on traits such as leaf morphology and the onset of flowering (reviewed by Huijser and Schmid, 2011; Geuten and Coenen, 2013; Poethig, 2013). Resistance to herbivory, as well as changes in cell wall composition, also have an important temporal component (Abedon et al., 2006; reviewed by Boege and Marquis, 2005 and Chandler et al., 2011). Thus, the study of developmental timing is crucial to our understanding of plant and animal ontogeny, physiology and evolution.

Post-embryonic phases of plant development include germination, seedling growth, the juvenile and adult vegetative phases, and flowering. Transitions between some of these programs are abrupt, whereas other transitions are more gradual. For example, germination entails a drastic switch from dormancy and growth repression (seed) to active growth and metabolism (seedling) (Bassel et al., 2011). By contrast, gradual transitions occur during the transition from the juvenile to the adult vegetative phase: aspects such as shape, serrations and epidermal hairs change incrementally between successive leaves. Depending on the species, leaf morphological traits may become either more or less complex from the juvenile to adult vegetative phase, when plants become reproductively competent (Poethig, 2013). The reproductive transition is relatively rapid, as the shoot apical meristem changes from leaf initiation to producing branches and flowers (Amasino, 2010).

Studies of the model plants Arabidopsis and maize have been particularly important in elucidating some of the molecular mechanisms that regulate phase-specific developmental programs, and have implicated factors such as the phytohormone gibberellic acid (GA), microRNAs (miRNAs) and epigenetic marks in the regulation of genes involved in these processes. The role of GA in promoting developmental transitions in plants is well established. GA activates seed germination, when rapid cell elongation is required for the embryo to rupture the seed coat, and simultaneously represses the expression of seed-specific genes (Koornneef and van der Veen, 1980; Ogas et al., 1997). GA also promotes leaf expansion during vegetative growth, as well as the onset of adult epidermal characteristics such as trichomes, which are canonical markers of vegetative phase change in Arabidopsis and maize (Evans and Poethig, 1995; Telfer et al., 1997). In addition, GA is required for the transition from vegetative growth to flowering (Wilson et al., 1992; Yu et al., 2012). Similarly, epigenetic regulation of gene expression, in particular through Histone 3 lysine 27 (H3K27) methylation, is important for both the seed-to-seedling and vegetative-to-reproductive transitions in plants (reviewed by Crevillén and Dean, 2011). Loss of H3K27 methylation affects the embryo-to-seedling transition by increasing dormancy and causing expression of seed-specific genes in seedlings (Bouyer et al., 2011; Zhang et al., 2008), while decreased H3K27 methylation at the floral repressor FLOWERING LOCUS C (FLC) delays flowering (Bastow et al., 2004).

Another important temporal regulator in plants is miR156 (Poethig, 2013), which regulates leaf traits and flowering by repressing members of the SPL family of transcription factors. SPL proteins positively regulate the expression of miR172, which in turn represses AP2-like transcription factors that inhibit flowering (Aukerman and Sakai, 2003; Chuck et al., 2007; Wu and Poethig, 2006; Wu et al., 2009; Wang et al., 2009). It has recently been shown that miR156 exerts its effects on flowering in parallel with GA because GA induces degradation of DELLA repressor proteins that are bound to SPL transcription factors (Yu et al., 2012). miR156 is abundant in seedlings, and decreases during vegetative development, partly due to transcriptional repression by leaf-derived sugars (Wu and Poethig, 2006; Yang et al., 2011, 2013; Yu et al., 2013). However, miR156 levels decrease during vegetative development even in the absence of sugar signaling, indicating that there are additional, as yet unknown, factors that regulate miR156 abundance during vegetative development (Yang et al., 2013).

The transcriptional co-activator Mediator has recently emerged as a key integrator of signaling pathways during development in both animals and plants, capable of influencing RNA polymerase II transcription of mRNAs and miRNAs at all stages of the transcription process, including through epigenetic regulation (Kim et al., 2011; reviewed by Yin and Wang, 2014). Mediator is a large protein complex that is conserved in all eukaryotes and has two major components: Core Mediator, which acts as a transcriptional co-activator, and the Cyclin Dependent Kinase 8 (CDK8) module, which represses transcription by modulating the activity of Core Mediator (Bourbon, 2008; reviewed by Carlsten et al., 2013). The CDK8 module is usually thought of as a transcriptional repressor, since the active form of Core Mediator is never purified together with the CDK8 module (Carlsten et al., 2013); however, in some contexts, the CDK8 module can also act to stimulate transcription, usually through the kinase activity of the CDK8 protein on the C-terminal domain of RNA polymerase II (reviewed by Nemet et al., 2013).

Mediator is capable of integrating diverse signaling inputs to influence transcription of multiple genes. For example, the MED18 subunit of Arabidopsis Core Mediator regulates fungal resistance, responses to the phytohormone abscisic acid, and flowering time, by influencing transcription initiation, elongation and termination, as well as RNA polymerase II occupancy and deposition of Histone 3 lysine 36 (H3K36) methylation (Lai et al., 2014). Likewise, studies in yeast and humans have shown that the CDK8 module can inhibit transcription through different mechanisms, including preventing physical association of Core Mediator with RNA polymerase II, and recruiting Histone 3 lysine 9 (H3K9) and H3K27 methylation at target genes (Tsai et al., 2013; Ding et al., 2008; Chaturvedi et al., 2012). The CDK8 module is composed of CDK8, Cyclin C, MED12 and MED13 (Carlsten et al., 2013). In Arabidopsis, MED12 is encoded by CENTER CITY [CCT; also known as CRYPTIC PRECOCIOUS (CRP)], MED13 is encoded by GRAND CENTRAL [GCT; also known as MACCHI-BOU 2 (MAB2)] and CDK8 is encoded by HUA ENHANCER 3 (HEN3) (Gillmor et al., 2010; Imura et al., 2012; Ito et al., 2011; Wang and Chen, 2004). CCT and GCT have previously been implicated in temporal regulation of development, as gct and cct mutants affect the timing of radial pattern formation during early embryogenesis (Gillmor et al., 2010).

Here, we demonstrate that GCT and CCT are global regulators of developmental transitions, promoting the embryo-to-seedling, juvenile-to-adult and vegetative-to-reproductive transitions. We show that one of the most important post-embryonic roles of GCT and CCT is to repress seed-specific transcripts in seedlings, and to promote growth after germination. In addition, we demonstrate that GCT and CCT regulate the abundance of miR156 during vegetative development, and that the upregulation of miR156 in gct and cct plants plays a major role in extending the vegetative phase in these mutants. GCT and CCT also regulate levels of FLC, and overexpression of FLC is responsible for much of the delayed flowering phenotype in gct and cct mutants. Thus, CCT and GCT act as master regulators of developmental transitions during the plant life cycle by repressing the expression of key temporal control genes.

gct and cct mutants germinate well, but develop abnormally following germination (99.5% germination in progeny of gct/+, n=428; 99.1% germination in progeny of cct/+, n=441). Although loss-of-function mutations of both genes are fully penetrant, they display variable expressivity: some mutant seedlings have small leaves, short petioles and malformed cotyledons but continue to grow and develop (we refer to this as the ‘grow’ phenotype), whereas others germinate and produce a root, but do not form leaf primordia and arrest at this stage (we refer to this as the ‘arrest’ phenotype) (Fig. 1B,C). 54% of gct seedlings and 24% of cct seedlings display the arrest phenotype. This variable expressivity is not attributable to partial activity of the mutant gene products: all five alleles of gct and all three alleles of cct described by Gillmor et al. (2010) generate both grow and arrest phenotypes (data not shown). These gct and cct alleles include both RNA nulls and predicted functional nulls (supplementary material Fig. S1) (Ito et al., 2011; Imura et al., 2012). Taken together, these molecular and functional data demonstrate that null mutants for each gene produce both grow and arrest phenotypes.

Analysis of GCT and CCT expression in a public dataset revealed that both genes are broadly expressed at low levels throughout post-embryonic development (supplementary material Fig. S2) (Schmid et al., 2005; Winter et al., 2007). The functional relationship between GCT and CCT was investigated by constructing plants mutant for both genes. The genotypes of the phenotypically mutant progeny of a self-pollinated gct/+ cct/+ plant were determined using allele-specific molecular markers. No double mutants were identified among the 27 plants with the grow phenotype, but 3 out of 18 plants with the arrest phenotype were homozygous for both mutations; the phenotypes of these gct cct double mutants are shown in Fig. 1D. GCT and CCT are not linked, so the expected frequency of double mutants among progeny displaying a mutant phenotype is 1/7; the ratio of double mutants (0/27) among grow phenotype seedlings is significantly different from this ratio (χ² goodness of fit test, P=0.03), whereas the frequency of double mutants among arrest phenotype seedlings (3/18) is not significantly different from this ratio. We conclude that loss of both GCT and CCT consistently leads to growth arrest. These results, as well as the phenotypic similarity of gct and cct mutations, suggest that these genes operate together or in parallel to regulate a core set of processes required for post-embryonic development.

To investigate the molecular basis for the post-embryonic phenotype of gct and cct, and the vegetative and reproductive changes described below, we used Affymetrix microarrays to compare steady-state mRNA levels in 7-day-old (7 d) wild-type (wt) seedlings and 9 d gct and cct seedlings, when the first two leaf primordia of these three genotypes were 1 mm in length (Fig. 2A). Compared with wt, transcripts for 1267 probe sets were more abundant in gct and 1373 were less abundant in gct (supplementary material Table S2), whereas transcripts for 925 probe sets were more abundant in cct and 1051 were less abundant in cct (supplementary material Table S3) (Limma, with a Benjamini-Hochberg adjustment, P<0.05). In total, 490 probe sets increased and 683 decreased in both gct and cct (supplementary material Table S4), an overlap that is greater than would be expected by chance (Fisher's exact test, P=0). Thus, GCT and CCT share a large, but not totally overlapping, set of direct and indirect gene targets.

An analysis of genes with transcripts that increase or decrease in both gct and cct is presented in supplementary material Table S5. These enriched terms (including GO terms) suggest a role for GCT and CCT in a number of growth-related processes, including the secretory pathway (such as GPI-anchored proteins and N-linked glycosylation), and cell wall remodeling. Among the most highly overrepresented GO terms were those related to seed-specific processes and flowering (supplementary material Table S5). An inspection of the genes that are most highly overexpressed in both gct and cct revealed many seed-specific transcripts, as well as the flowering repressor FLC; flowering genes that showed significant decreases in expression included FLOWERING LOCUS T (FT) and SUPPRESSOR OF CONSTANS 1 (SOC1) (both of which are repressed by FLC), as well as CONSTANS (CO) (supplementary material Table S4). These results suggest that, in addition to controlling the timing of embryonic development (Gillmor et al., 2010), GCT and CCT regulate post-embryonic developmental transitions.

As mentioned above, a number of seed-specific transcripts were among the most highly overexpressed genes in both gct and cct seedlings (Fig. 2B; supplementary material Table S4). In support of a role for GCT and CCT in regulating the seed-to-seedling transition, we found a significant overlap between probe sets upregulated in both gct and cct and those increasing at the green cotyledon stage of embryogenesis (Fig. 2C; Fisher's exact test right-tail, P≤1.61×10⁻⁴) (Willmann et al., 2011). Thus, gct and cct seedlings show molecular signatures of late embryogenesis.

To confirm and extend these results, we used qRT-PCR to measure the abundance of four seed-specific transcripts (At2S2, CRU1, OLEO1, ATTI7) in 6 d, 14 d and 21 d wt and mutant plants; only mutant plants with the grow phenotype were used for this experiment. All four genes were expressed at high levels in 6 d gct seedlings, and their expression declined gradually in this mutant over the next 2 weeks (Fig. 2D). At2S2, OLEO1 and ATTI7 were also misexpressed in 6 d cct seedlings, but were much less elevated than in gct (Fig. 2E) and also disappeared faster in this genotype. Consistent with the results of the microarray analysis, CRU1 was unaffected by cct. We also compared the expression of these genes in 7 d gct and cct seedlings with the grow and arrest phenotypes (Fig. 2F). In every case, these genes were expressed at significantly higher levels in seedlings with the arrest phenotype. For example, At2S2 was elevated 3-fold in grow gct seedlings and 88-fold in arrest gct seedlings, while OLEO1 was elevated 4-fold in grow gct seedlings and 162-fold in arrest gct seedlings. These results demonstrate that GCT and CCT repress the expression of embryo-specific genes during post-embryonic development.

The phytohormone GA has a similar function to GCT and CCT in that it promotes seed germination and represses embryonic identity in young seedlings (Koornneef and van der Veen, 1980; Ogas et al., 1997). To determine if GA operates via GCT and CCT, we examined the effect of this hormone on the expression of At2S2, CRU1, OLEO1, ATTI7 in gct and cct mutants. These genes were strongly repressed by GA in both mutants (Fig. 2D,E), indicating that GA represses embryonic gene expression independently of GCT and CCT.

In Arabidopsis, the transition from the juvenile to the adult phase of vegetative development is marked by changes in leaf shape (round to elongated), leaf serration (smooth to serrated) and abaxial trichome production (absent to present) (Röbbelen, 1957; Chien and Sussex, 1996; Telfer et al., 1997). gct and cct delayed the appearance of the adult forms of all of these traits (Fig. 3). In long-day conditions, gct and cct plants had over twice as many leaves without abaxial trichomes as wt plants (Table 1). gct and cct also had rounder leaves than wt (Fig. 3E,F) and began to produce serrated leaves later and with a slower rate of increase in serration number than wt plants (Fig. 3E,G). Importantly, the rate of leaf initiation was almost identical in wt, gct and cct, indicating that the increased number of juvenile leaves in gct and cct is not due to an increase in the rate of leaf production (Fig. 3H).

The phytohormone GA promotes several of the processes regulated by GCT and CCT, including the onset of abaxial trichome production and the transition to flowering (Chien and Sussex, 1996; Telfer et al., 1997; Wilson et al., 1992). To determine whether GCT and CCT act through the GA pathway, we examined how variation in the amount of GA affects the phenotype of gct and cct. Exogenously applied GA₃ accelerated abaxial trichome production and flowering in gct and cct, whereas ga1-3 – a mutation that blocks GA biosynthesis – further delayed abaxial trichome production and flowering time in these mutants (Table 1; Fig. 3I-K). The observation that gct and cct do not interfere with the sensitivity of plants to GA suggests that GCT and CCT regulate vegetative phase change and flowering in parallel to this hormone.

The vegetative phenotype of gct and cct is similar to that of plants overexpressing miR156, the miRNA that acts as a master regulator of the juvenile-to-adult transition in Arabidopsis (Wu and Poethig, 2006). To determine if GCT and CCT regulate vegetative phase change via miR156, we examined the effect of gct and cct on the expression of miR156, as well as the expression of direct (SPL3 and SPL9) and indirect (miR172) targets of this miRNA (Wu et al., 2009) (Fig. 4). Northern blot analysis revealed that miR156 levels were approximately twice as high in gct and cct mutants as in wt plants at 7 d, 14 d and 21 d after germination. miR156 levels decreased steadily in all genotypes, so that at 21 d the amount of miR156 was about half that seen at 7 d. As reported previously, miR156 and miR172 show opposite patterns of accumulation (Wu et al., 2009). In agreement with this, we found that the increased expression of miR156 in gct and cct was associated with a decrease in the expression of miR172 at 7 d, 14 d and 21 d (Fig. 4A). Similarly, the mRNA levels of the transcription factors SPL3 and SPL9, which are direct targets of miR156, were reduced in both gct and cct (Fig. 4C), although in the case of SPL9 this effect was only observed at 21 d. gct and cct had no significant effect on the expression of miR159, miR161 and miR168 (Fig. 4C). These results suggest that GCT and CCT specifically regulate miR156, rather than being generally involved in miRNA transcription.

To determine if the increased expression of miR156 in gct and cct mutants is responsible for their effect on vegetative phase change, we reduced miR156 levels in these mutants using a 35S::MIM156 transgene. This transgene expresses an imperfect miR156 target site, which reduces free miR156 in the cell by sequestering this miRNA in inactive RISC complexes (Franco-Zorrilla et al., 2007). Consistent with previous reports (Wu et al., 2009; Wang et al., 2009), wt plants carrying a single 35S::MIM156 transgene failed to express juvenile leaf traits: they produced abaxial trichomes starting with leaf 1 (instead of leaf 5.5) (Fig. 5A), and all of their rosette leaves had the pattern of leaf serrations (Fig. 5B) and the elongated shape (Fig. 5C) characteristic of adult leaves. 35S::MIM156 had this same effect in the gct and cct mutant backgrounds: transgenic gct and cct plants produced abaxial trichomes on leaf 1 (instead of 16 and 15.5, respectively) (Fig. 5A), and had more serrated (Fig. 5B) and more highly elongated (Fig. 5C) leaves than wt plants. Although a single copy of 35S::MIM156 was totally epistatic to gct and cct with regard to abaxial trichome production, it was unable to completely suppress their effect on leaf serrations and the leaf length-to-width ratio. This could mean that the pathways affecting these morphological traits are sensitive to very small amounts of miR156, or that gct and cct have both miR156-dependent and miR156-independent roles in leaf development.

In summary, these results demonstrate that miR156 is required for the juvenilized phenotype of gct and cct, and support the conclusion that this phenotype is largely attributable to the increased amount of miR156 in these mutants.

gct and cct also have a significant effect on the transition to flowering (Fig. 3A). During this transition, the shoot apex produces several flower-bearing branches (co-florescences) subtended by leaves (bracts), and then produces only flowers. In long days, wt plants made an average of 3.1 bracts and flowered 24.7 days after planting (DAP) (Table 1). gct plants made an average of 7.2 bracts and flowered 50.8 DAP, while cct plants produced an average of 7.8 bracts and flowered 41.3 DAP (Table 1). gct and cct produced smaller flowers than normal, with occasional fused stamen filaments and a variable number of petals and stamens (Fig. 3B-D; data not shown). gct flowers are completely male and female sterile, whereas the flowers of cct are initially completely sterile, but will produce a small amount of seed after several months of growth. Thus, as previously reported for cct (crp) by Imura et al. (2012), GCT and CCT promote floral induction and also play a role in floral morphogenesis.

Our microarray analysis revealed that the late flowering phenotype of gct and cct is associated with reduced expression of the floral inducers FT, SOC1, SEPALLATA3 (SEP3), FRUITFULL (FUL) and CO and with elevated expression of the floral repressor FLC (Fig. 6A; supplementary material Tables S2-S4), which is consistent with the known regulatory interactions between these genes (Fig. 6B) (Amasino, 2010). To validate and extend these results, we quantified FLC, FT and SOC1 levels in wt, gct and cct plants at 6 d and 21 d. In wt plants, FLC transcripts remained steady at 6 d and 21 d, whereas SOC1 and FT increased dramatically (consistent with the vegetative-to-floral transition having occurred by 21 d) (Fig. 6C,D). In gct and cct plants, we found that levels of FLC transcripts were much higher than in wt at 6 d, and had decreased several fold by 21 d (Fig. 6C), whereas the levels of FT and SOC1 increased slightly from 6 d to 21 d (Fig. 6D,E). These changes in FT and SOC1 gene expression are not attributable to differences in leaf identity between wt, gct and cct, as FT and SOC1 show essentially the same expression in juvenile and adult leaves (supplementary material Fig. S3). Although FLC is expressed more highly in juvenile than in adult leaves (supplementary material Fig. S3), a secondary effect of leaf identity is unlikely to explain the large differences in FLC levels between wt, gct and cct because the largest differences in FLC were observed at 6 d, when all three genotypes have only two juvenile leaf primordia (Fig. 3E-H). Thus, gct and cct mutants show large increases in expression of the floral repressor FLC, and, in turn, show decreased expression of the FLC targets FT and SOC1.

As shown in Fig. 4, gct and cct also have higher levels of miR156, a miRNA that has previously been shown to cause a small delay in flowering in long-day conditions when overexpressed (Wang et al., 2009). This raises the possibility that the increased level of FLC in gct and cct might be attributable to the increased level of miR156 in these mutants (Fig. 4). To test this, we examined FLC levels in 35S::miR156A plants and found a ∼30% decrease in the level of FLC compared with wt (supplementary material Fig. S4). Because gct and cct show elevated levels of both miR156 and FLC, we conclude that the increase in FLC in gct and cct is not caused by the increase in miR156.

To determine whether the late flowering phenotype of gct and cct is indeed attributable to the increased expression of FLC, we crossed the flc-3 null mutation into gct and cct mutant backgrounds. The flowering time of these double mutants was intermediate between the two parents: gct flc-3 double mutants flowered ∼7 days earlier than gct, while cct flc-3 double mutants flowered 6 days earlier than cct (Table 1; Fig. 6F), but neither genotype flowered as early as flc-3. The levels of FT and SOC1 in these double mutants were consistent with their intermediate phenotype. Whereas flc-3 completely blocked the effect of cct on SOC1 expression in 21 d plants (Fig. 6E), it had only a small effect on FT expression in cct mutants (Fig. 6D). flc-3 had little or no effect on the expression of FT and SOC1 in a gct mutant background (Fig. 6D,E). This result indicates that the elevated expression of FLC contributes to the late flowering phenotype of cct and gct, but is not entirely responsible for this phenotype.

Multicellular organisms pass through several developmental phases during their life cycle. In Arabidopsis, these stages include embryogenesis and seed formation, germination, juvenile and adult phases of vegetative growth, and a reproductive phase that culminates in gamete production and fertilization. We have previously shown that the transcriptional repressors GCT (MED13) and CCT (MED12) regulate the timing of pattern formation during Arabidopsis embryogenesis (Gillmor et al., 2010). In the current study we show that GCT and CCT play a global role in the regulation of developmental timing, promoting the seed-to-seedling transition, vegetative phase change, and flowering.

Microarray analysis demonstrated that GCT and CCT share close to half of their direct or indirect gene targets, with 490 transcripts increasing in both gct and cct (out of a total 1267 genes increasing in gct and 925 in cct), and 683 transcripts decreasing in both mutants (out of 1373 genes decreasing in gct and 1051 in cct). The number of transcripts that change in level is comparable with the findings of RNA profiling in single cell types in Drosophila [S2 cells, where a total of 361 genes were found to change in med12 (kohtalo) or med13 (skuld) mutants] and yeast (where a total of 900 genes were found to increase or decrease), as is the highly significant overlap that we observed between the transcripts that change in common between med12 and med13 mutants (Kuuluvainen et al., 2014; Van de Peppel et al., 2005). It is commonly accepted that MED12 and MED13 primarily act as transcriptional repressors (Carlsten et al., 2013), and thus many of the transcripts whose steady-state levels decrease in gct and cct mutants are unlikely to be direct targets of GCT and CCT regulation. However, there are also reports of the CDK8 module acting as a transcriptional activator (Carrera et al., 2008; Donner et al., 2007). No genome-wide chromatin immunoprecipitation (ChIP) profiling experiment has been reported for either MED12 or MED13 in plants or animals. Despite the challenges inherent in ChIP analysis using such large proteins, determination of the direct targets of MED12 and MED13 is an important goal for future research in both Arabidopsis and animals.

In addition to the molecular data discussed above, genetic analysis provides functional evidence that GCT and CCT regulate many of the same processes. For example, gct and cct single mutants express either a weak ‘grow’ phenotype characterized by moderate defects in vegetative and reproductive timing, or a strong ‘arrest’ phenotype of arrested growth. By contrast, gct cct double mutants express only the arrest phenotype. This result suggests that, in the absence of either MED12 or MED13, the CDK8 module is partially functional and can still promote post-embryonic growth, albeit in a less effective manner, leading to the variable expressivity of the gct and cct grow and arrest phenotypes. This result, and the observation that gct cct double mutants only have the severe arrest phenotype, suggest that GCT and CCT have closely related, but not completely redundant functions. This is consistent with studies of MED12 and MED13 in yeast and animals, where they have been shown to affect transcription of many of the same genes and to have nearly identical mutant phenotypes (Samuelsen et al., 2003; Janody et al., 2003; van de Peppel et al., 2005; Yoda et al., 2005). Additionally, in vitro studies have demonstrated that, although MED12 and MED13 are capable of individually blocking transcription, they act most efficiently together (Knuesel et al., 2009).

The CDK8 module of Mediator represses transcription by at least two mechanisms. Structural studies in yeast and humans demonstrate that this module binds Core Mediator, blocking the Core Mediator-RNA polymerase II interaction (Elmlund et al., 2006; Tsai et al., 2013). The MED13, MED12 and CDK8 subunits all contribute to binding, with MED13 playing the most important role, followed by MED12 and CDK8 (Tsai et al., 2013). These results are consistent with the mutant phenotypes of these proteins in Arabidopsis, as the expressivity of the growth arrest phenotype is higher for gct (med13) than for cct (med12). Human MED12 also represses gene expression by recruiting H3K9 and H3K27 methylation to target loci through its interaction with the histone methyltransferase G9a (EHMT2) (Ding et al., 2008; Chaturvedi et al., 2009, 2012). Steric hindrance of RNA polymerase II-Core Mediator interactions represents a relatively flexible form of transcriptional repression, whereas H3K9 and H3K27 methylation is likely to have more stable effects on gene expression. The mutant phenotype of the gct cct double mutant suggests that both activities are necessary for full repression by the CDK8 module. H3K9 and H3K27 methylation have well-established roles in the seed-to-seedling and vegetative-to-reproductive transitions. The PRC2 complex represses the expression of embryonic genes after germination by recruiting deposition of H3K27 methylation (Bouyer et al., 2011). Likewise, the flowering repressor FLC is repressed by both H3K9 and H3K27 marks (Bastow et al., 2004). It will be important to determine whether the effect of GCT and CCT on these genes is mediated by these histone modifications.

Our results indicate that the delayed vegetative phase change phenotype of gct and cct is attributable to an increase in the abundance of miR156 and a corresponding decrease in levels of the downstream miRNA miR172. miR156 was elevated in gct and cct throughout shoot development, and declined at the same time in mutant and wt plants, suggesting that GCT and CCT might regulate the amplitude, rather than the temporal pattern, of miR156 expression. It will be important to determine if GCT and CCT directly regulate the transcription of specific MIR156 genes or whether they modulate the transcription of an upstream regulator of one or more MIR156 genes, in order to investigate the molecular mechanism of this effect. Whatever this mechanism might be, it is interesting that the relatively small change in the amount of miR156 in these mutants (∼2-fold) has such a drastic effect on the expression of phase-specific traits. This suggests that fine-tuning of MIR156 expression by GCT and CCT is crucial for normal phase change, and raises the possibility that spatial or temporal variation in the activity of these proteins might contribute to natural variation in the timing of the vegetative phase change. Our results also suggest that GCT and CCT play an important role in controlling the timing of the vegetative-to-floral transition by direct or indirect regulation of FLC, as well as through additional, yet to be determined, pathways. Using physiological and genetic experiments, we demonstrated that GCT and CCT regulate the seed-to-seedling transition, vegetative phase change and flowering in parallel with the phytohormone GA.

Developmental transitions are characterized by a decrease in the expression of genes that specify the pre-existing developmental phase and by an increase in the expression of genes that promote the subsequent phase. In principle, the decline in gene expression during these transitions could be mediated by the loss of factors that promote gene expression, by the appearance of factors that repress gene expression, or by a combination of these events. The pleiotropic phenotype of gct and cct implies that active transcriptional repression by GCT (MED13) and CCT (MED12) plays a major role during many temporal transitions in Arabidopsis. This paradigm has been established for a few developmental transitions in Arabidopsis – the best example being the repression of FLC expression during vernalization – but the generality of transcriptional repression as a mechanism for developmental transition had not been fully appreciated until now. Our results demonstrate that GCT and CCT are not required for the downregulation of gene expression during phase transitions per se, but rather for the magnitude and perhaps the stability of the repressed state. It is reasonable to assume that they perform this function in association with temporally regulated transcription factors, and a major challenge for the future is to identify these transcription factors and the ways in which they interact with GCT and CCT to repress gene expression.

All seed stocks were in the Columbia ecotype. Seeds were sown on soil (Farfard #2) and placed at 4°C for 3 days, before moving flats to growth chambers or laboratory growth racks under 16 h fluorescent illumination (140 µmol/m²/s) at 22°C. Days to first open flower were measured beginning on the day flats were placed in the growth chamber. For GA treatments, plants were sprayed daily with 100 µM GA₃ (Sigma-Aldrich). For the microarray experiments, plants were grown under long-day conditions (16 h light:8 h dark; 100-125 μmol/m²/s) in a Conviron growth chamber at a constant 23°C under a 1:1 ratio of T8 Sylvania Octron 4100K Ecologic and GroLite WS fluorescent lamps (Interlectric).

The reference alleles gct-2 (ABRC stock CS65889; referred to as gct throughout the paper) and cct-1 (ABRC stock CS65890; referred to as cct throughout the paper) described by Gillmor et al. (2010) were used for all analyses. The gct-2 mutation was genotyped using the gct-2 dCAPS primer pair, and the cct-1 mutation was genotyped using the cct-1 dCAPS primer pair. dCAPS primers were designed using dCAPS Finder 2.0 (Neff et al., 2002; http://helix.wustl.edu/dcaps/). The ga1-3 allele, originally in the Ler ecotype (Wilson et al., 1992), was introgressed six times into the Col ecotype, and was genotyped based on its morphological phenotype. The flc-3 allele was used for analysis of the FLC gene (Michaels and Amasino, 1999). flc-3 mutants were genotyped using the flc-3 primer pair. The 35S::MIM156 target mimicry line for miR156 was genotyped based on its morphological phenotype (Franco-Zorrilla et al., 2007). Sequences of genotyping primers are listed in supplementary material Table S1.

The microarray experiment was performed on three biological replicates of 7 d wt and 9 d gct and cct mutants with the ‘grow’ phenotype, each replicate being a pool of at least 20 seedlings. Different time points were used because they were when the first two leaf primordia of these three genotypes were 1 mm in length. Total RNA was extracted with TRI Reagent (Sigma-Aldrich), and further purified using RNeasy columns (Qiagen) with on-column DNase treatment (Qiagen). Biotin-labeled cDNA targets for hybridization to Affymetrix Arabidopsis ATH1 microarrays were prepared essentially as described by Willmann et al. (2011). The University of Pennsylvania Microarray Core Facility hybridized the arrays. Raw data have been deposited at Gene Expression Omnibus under accession number GSE56155.

The microarrays were gcRMA normalized in R (http://www.r-project.org/) and filtered using MAS5.0 presence/absence calls to remove any probe sets not expressed in at least three samples. The remaining 16,826 non-control probe sets were tested for differential expression in R using Limma (Bioconductor) with contrast and a Benjamini-Hochberg multiple test correction (MTC), P≤0.05. Enrichment of GO terms, SP_PIR keywords, PIR Superfamilies, and SMART, INTERPRO, COG_ONTOLOGY and UP_SEQ features within the lists of probe sets increasing and decreasing in both gct and cct was performed using DAVID Bioinformatics Resources 6.7 (http://david.abcc.ncifcrf.gov/home.jsp; Dennis et al., 2003; Huang et al., 2009) with the 16,826 expressed probe sets as a background. The probe sets differentially expressed in both gct and cct were also compared with the 862 probe sets shown to specifically increase in green cotyledon stage embryos by Willmann et al. (2011) (Cluster 9).

For quantitative RT-PCR (qRT-PCR), total RNA was isolated as described above for ten pooled seedlings per time point, per biological replicate. Reverse transcription was performed with the oligo(dT)₂₀ primer using the SuperScript II First-Strand Synthesis System for RT-PCR (Invitrogen). Quantitative PCR was performed with Fast SYBR Green Master Mix (Applied Biosystems) on a StepOnePlus RT-PCR System (Applied Biosystems). Transcript levels were normalized against EUKARYOTIC TRANSLATION INITIATION FACTOR 4A (EIF4A). Primers for qRT-PCR were designed using AtRTPrimer (http://pbil.kaist.ac.kr/AtRTPrimer; Han and Kim, 2006) and are listed in supplementary material Table S1.

For small RNA northern blots, 30 μg total RNA was separated on 8 M urea/15% denaturing polyacrylamide gels and electrically transferred to a Hybond-N+ nylon membrane (GE Healthcare). Blots were hybridized with [γ-³²P]ATP-labeled complementary oligonucleotide probes for 3 h at 40°C in Rapid-hyb hybridization buffer (GE Healthcare). A U6 RNA-complementary oligonucleotide probe was used as a loading control. Blots were washed twice at 40°C in 5×SSC, 0.1% SDS for 20 min, and were scanned with a Storm 820 phosphorimager (Amersham Biosciences). Oligonucleotide probes used are listed in supplementary material Table S1. Signal intensities were quantified using ImageJ (NIH).

Analysis of leaf traits, such as size, shape and the presence of abaxial trichomes, was performed at ∼4 weeks in order to allow leaves to reach their final size. The presence of abaxial trichomes was scored using a dissecting microscope. Leaf length and width ratios, as well as the number of serrations, were quantified by taping leaves to paper, scanning and analyzing in Photoshop (Adobe).

We thank Keith Earley and Enrique Ibarra-Laclette for advice on qRT-PCR, and Doris Wagner and Kimberly Gallagher for helpful discussions during the course of this study.