Edinburgh Research Explorer Polycomb-group proteins and FLOWERING LOCUS T Maintain commitment to flowering in Arabidopsis thaliana

The switch from vegetative to reproductive growth is extremely stable even if plants are only transiently exposed to environmental stimuli that trigger ﬂ owering. In the photoperiodic pathway, a mobile signal, ﬂ origen, encoded by FLOWERING LOCUS T ( FT ) in Arabidopsis thaliana , induces ﬂ owering. Because FT activity in leaves is not maintained after transient photoperiodic induction, the molecular basis for stable ﬂ oral commitment is unclear. Here, we show that Polycomb-group (Pc-G) proteins, which mediate epigenetic gene regulation, maintain the identity of in ﬂ orescence and ﬂ oral meristems after ﬂ oral induction. Thus, plants with reduced Pc-G activity show a remarkable increase of cauline leaves under noninductive conditions and ﬂ oral reversion when shifted from inductive to noninductive conditions. These phenotypes are almost completely suppressed by loss of FLOWERING LOCUS C ( FLC ) and SHORT VEGETATIVE PHASE , which both delay ﬂ owering and promote vegetative shoot identity. Upregulation of FLC in Pc-G mutants leads to a strong decrease of FT expression in in ﬂ orescences. We ﬁ nd that this activity of FT is needed to prevent ﬂ oral reversion. Collectively, our results reveal that ﬂ oral meristem identity is at least partially maintained by a daylength-independent role of FT whose expression is indirectly sustained by Pc-G activity.


INTRODUCTION
The time of flowering in plants is critically important so that flower and fruit production is coordinated with pollinators and seasons. It is regulated by external environmental factors, notably photoperiod and temperature, which provide seasonal cues, and also by internal factors such as age; most plants are not competent to respond to inductive signals unless they have first completed a juvenile phase (reviewed in Kobayashi and Weigel, 2007;Srikanth and Schmid, 2011). Once triggered, the switch from vegetative growth to flowering is usually irreversible. This is likely important in natural environments so that plants do not make sterile flower/ shoot intermediates if inductive signals such as temperature fluctuate. Floral commitment is most apparent in annual plants, which usually do not revert to vegetative growth. In perennial plants, which undergo repeated yearly cycles of vegetative and reproductive growth, although not all meristems switch to reproductive growth in any one year, those that do usually show stable floral commitment (Tooke et al., 2005;Wang et al., 2009). However, there are a few examples of plant species and mutants that show distinct types of floral reversion: inflorescence reversion in which the inflorescence reverts to vegetative growth and flower reversion in which the flower itself has features of the inflorescence or even vegetative identity. In these species, floral reversion occurs primarily if inductive signals are not maintained (Tooke et al., 2005). Despite this, the molecular basis for floral commitment is poorly understood, even in plants like Arabidopsis thaliana where the molecular mechanisms triggering flowering are well defined.
Aerial parts of plants are formed from shoot apical meristems (SAMs), which undergo several changes in identity before producing flowers. These changes in meristem identity cause modifications in the shoot structure through changes within the basic unit of the metamer, which is composed of an internode and a node. The latter consists of a leaf or bract that harbors a secondary meristem in its axil that give rise to a side shoot (Evans and Grover, 1940;Schultz and Haughn, 1991). In Arabidopsis, the SAM sequentially produces three types of metamers: in the vegetative phase, the SAM generates rosette leaves subtending a lateral shoot; after floral induction, the stem elongates (bolting) and cauline leaves with a side shoot, the paraclade, are formed (I 1 stage), finally leading to production of flowers without bracts (I 2 stage) ( Figure 1A; Schultz and Haughn, 1991;Haughn et al., 1995). Detailed analysis of the SAM at the time of floral induction suggests that the I 1 stage is a continuation of the vegetative phase and that the SAM directly produces floral primordia on its flanks after induction Feldman, 1994, 1995). Occasionally, e.g., by shifting wild-type plants from short-day (SD) to continuous light, flowers subtended by bracts (cauline leaves) can be observed, indicating that floral bracts can occur in Arabidopsis (Hempel et al., 1998). Furthermore, analysis of SHOOTMERISTEMLESS (STM) and AINTEGUMENTA (ANT) expression, which are markers for meristems and organs respectively, indicate that primordia are initially leaf-like in character as they are marked by absence of STM expression and presence of ANT (flower development stages 0 and 1 in the classification of Smyth et al., 1990;Long and Barton, 2000). In early stage 2, the primordium is partitioned into a meristem, marked by STM expression, and a "cryptic" bract marked by absence of STM (Long et al., 1996;Long and Barton, 2000) and briefly morphologically discernable (Kwiatkowska, 2006). During later stages of flower development, cryptic bract outgrowth is suppressed; hence, it is not visible at maturity. Nonetheless, in mutants with compromised floral meristem identity, such as leafy (lfy) or apetala1 (ap1), the cryptic bract develops more fully and is visible at maturity as a bract subtending flowers or paraclades (Mandel et al., 1992;Weigel et al., 1992). The floral meristem (FM) differs from vegetative or inflorescence SAMs not only in that it makes floral organs, but also in that it is determinate and does not give rise to more meristems or branches.
Arabidopsis FLOWERING LOCUS T (FT) is the key gene mediating the vegetative to floral transition and is widely conserved in flowering plants (Kardailsky et al., 1999;Kobayashi et al., 1999;Izawa et al., 2002;Kojima et al., 2002). Thus ft mutants show delayed flowering but produce normal flowers once the transition occurs. The FT product likely corresponds to florigen, the longsought mobile signal that moves from leaves to the SAM in response to inductive photoperiods and promotes flowering (reviewed in Kobayashi and Weigel, 2007). FT is expressed in leaves in inductive photoperiods (long days in Arabidopsis) and produces a small protein that moves through the phloem to the shoot apex where it forms a complex with a bZIP transcription factor, FD, to activate targets promoting flower primordium identity including AP1 and FRUITFULL (FUL) (Abe et al., 2005;Wigge et al., 2005). A second gene promoting the floral transition is SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1 (SOC1), which encodes a MADS box transcription factor (Samach et al., 2000;Searle et al., 2006). SOC1, together with FUL, are likely targets needed for FT action; thus, soc1 ful double mutants largely prevent the early flowering triggered by 35S pro :FT transgenes. In addition, soc1 ful double mutants show reversion from inflorescence to vegetative growth, indicating that floral commitment is impaired (Melzer et al., 2008).
Once floral induction has occurred, LFY and AP1 are expressed. Expression of AP1 is initiated in floral primordia at stage 1 and is thought to mark floral commitment, whereas LFY expression initiates in floral anlagen in incipient primordia (stage 0) and is thus expressed before flowers are discernible (Mandel et al., 1992;Weigel et al., 1992). Studies using transient activation of a steroiddependent LFY transgene indicate that persistent LFY activity is needed to prevent flower-to-shoot reversion (Wagner and Meyerowitz, 2011), consistent with lfy mutants displaying floral reversion under certain conditions (Okamuro et al., 1996). LFY and AP1 proteins bind each other's promoters and upregulate one another's transcription, suggesting that once LFY and AP1 expression are activated in floral primordia they are likely selfperpetuating through positive feedback loops (Wagner et al., 1999;Adrian et al., 2009;Kaufmann et al., 2010).
In addition to flower-promoting factors like LFY and AP1, the switch to floral meristem identity also requires repression of factors that promote vegetative or inflorescence shoot identity in the early floral primordia. Thus, transcriptional profiling indicates that during the early stages of floral primordium development the majority of genes whose expression significantly changes are downregulated (Wellmer et al., 2006). The downregulated genes include SHORT VEGETATIVE PHASE (SVP) and SOC1 which are both direct targets of AP1 and promote shoot identity (Liu et al., 2009;Kaufmann et al., 2010). Paradoxically, activity of these genes is needed during the earliest stages (stages 0 to 2) of flower development and acts together with LFY to activate the floral homeotic genes that specify floral organ identity and determinacy (Liu et al., 2009). Subsequently, SVP/SOC1 need to be switched off for floral meristem identity to be maintained and flower patterning to occur. Thus, expression of SVP/SOC1 transgenes during later stages of flower development causes floral reversion (Liu et al., 2007). (A) The wild-type SAM produces three metameric types sequentially: rosette leaves are formed at the vegetative state (V), after floral induction, the stem elongates and cauline leaves subtending lateral branches, named paraclades, are generated (I 1 ), followed by production of flowers (I 2 ). After floral induction, flower production is stable even if plants are shifted from LD to SD. (B) to (E) Development of the flower metamers (according to Smyth et al. [1990], Chandler [2012], and Long and Barton [2000]): In stage 1 (B), the floral primordium emerges as cryptic bract (CB) and thus has leaf-like identity. In early stage 2 (C), the FM develops between the IM and the cryptic bract. The formation of the sepal primordia at the flanks of the FM marks stage 3 (D). The CB is morphologically not detectable. Stage 13 flower is in (E). (F) Pc-G mutants shifted from LD to SD show floral reversion and, thus, an additional I 1 phase after I 2 .
(G) to (L) Likely scenario for origin of distinct types of reversion nodes: empty cauline leaf (G), cauline leaf subtending a paraclade (H), cauline leaf subtending a flower (I), paraclade without a cauline leaf but with a rudimentary bract (RB) (J), late reversion of the FM, an inflorescence emerges inside of sepals (K), and flower with RB (L).
In Arabidopsis, as with most plants, flowering persists after transient photoperiodic induction (Corbesier et al., 1996). However, analysis of plants shifted from inductive long-day (LD) to SD conditions shows that both FT expression in leaves and SOC1 expression in the SAM is rapidly lost after the shift (Corbesier et al., 2007;Torti et al., 2012). Floral commitment has similarity with the vernalization response in that both involve a stable memory of a transient environmental stimulus (cold in the case of vernalization), raising the question of whether they share a common mechanistic basis. The memory of vernalization is mediated by Polycomb-group (Pc-G) genes that confer stable epigenetic silencing of a repressor of flowering, FLOWERING LOCUS C (FLC) (Gendall et al., 2001), which in turn represses expression of FT and SOC1 (Searle et al., 2006). Consistent with their epigenetic role, the Pc-G proteins have biochemical activity toward chromatin; in particular, a complex of four core proteins termed Polycomb Repressive Complex 2 (PRC2) catalyzes the methylation of Lys-27 on histone H3 (H3K27me3), a modification associated with transcriptional repression (reviewed in Margueron and Reinberg, 2011). In Arabidopsis, the MEDEA, CURLY LEAF (CLF), and SWINGER (SWN) genes encode homologs of Drosophila melanogaster Enhancer of zeste, the catalytic subunit of the PRC2, while FERTILIZATION INDEPENDENT SEED2, EMBRYONIC FLOWER2 (EMF2), and VERNALIZATION2 (VRN2) show similarity to a second core PRC2 component, Drosophila Suppressor of zeste 12 (Chanvivattana et al., 2004). In clf swn mutant seedlings, H3K27me3 methylation is completely lost, suggesting that Pc-G activity is eliminated (Lafos et al., 2011). Whereas double mutants for the severe alleles emf2-3 and vrn2-1 resemble clf swn mutants phenotypically (Schubert et al., 2005), combination of the weaker emf2-10 allele with vrn2-1 results in viable, fertile plants, although global H3K27me3 levels are highly reduced (Lafos et al., 2011).
Here, we show that Pc-G proteins are required for floral commitment. Thus, Pc-G mutants show floral reversion such that FMs revert to an earlier inflorescence meristem (IM) identity. Floral reversion in Pc-G mutants requires activity of the two MADS box transcription factors, FLC and SVP. We show that in plants given a transient photoperiodic induction, FT expression does not persist in leaves as previously reported (Corbesier et al., 2007), although it is activated in inflorescences. Furthermore, reduced FT expression in inflorescences, both in Pc-G mutants and as a consequence of FT mutation, causes floral reversion. Thus, additional to its known role in triggering the switch to flowering, FT has a second role in maintaining flower primordium identity in the inflorescence. Thus, Pc-G proteins maintain commitment to flowering at least partially by promoting FT activity in the inflorescence.

Generation of Plant Lines with Strongly Depleted Pc-G Activity
After germination, seedlings of mutants lacking Pc-G activity fail to develop leaves or flowers but instead produce embryogenic callus (Figures 2A and 2B;Chanvivattana et al., 2004). Hence, null Pc-G mutants are uninformative for analyzing Pc-G function during vegetative and reproductive development. To reveal whether Pc-G proteins are also required to initiate and maintain reproductive identity, we circumvented this early developmental arrest by constructing a conditional mutant (iCLF; see Methods; Figure 2; Supplemental Figure 1) in which a steroid-dependent CLF transgene (CLF pro :CLF-GR described in Schubert et al., 2006) was introduced into the clf-28 swn-7 background that lacks all Pc-G activity (Lafos et al., 2011). We rescued the early defects by germinating iCLF seedlings on steroid-containing media, then withdrew steroid by transplanting to soil. When Pc-G activity was progressively depleted in this way, development continued normally except that leaves were more serrated than normal ( Figures  2B and 2C) and most flowers were sterile (Supplemental Figures  1A and 1B). Similar leaf serration is observed in the Pc-G double mutant emf2-10 vrn2-1 ( Figure 2D), which shows a strong global reduction of H3K27me3 levels (Chanvivattana et al., 2004;Lafos et al., 2011). Therefore, these two plant lines, which are deficient in different components of the PRC2, were used to investigate the role of Pc-G in the vegetative and reproductive phase.

Flower Formation Is Delayed in Plants with Reduced Pc-G Activity
To examine potential defects in flowering time regulation, we grew emf2-10 vrn2-1 and iCLF plants under LD and SD conditions and counted their rosette and cauline leaf number. The emf2-10 vrn2-1 double mutants flowered appreciably later than the wild type in SD, whereas in LD, their flowering was only slightly delayed ( Table  1). The rosette leaf number of iCLF plants was also significantly increased in both LD and SD (Supplemental Table 1). Both lines displayed additional developmental defects, including homeotic transformations in organ identity, increased floral organ number, and a lack of floral meristem determinacy evident as a meristematically active fifth whorl (Supplemental Figures 2A and 2B), suggesting misregulation of a similar set of target genes. In addition, when emf2-10 vrn2-1 mutants were grown in SD, they produced inflorescences with an abnormally high number of cauline leaves (>60) so that flowers were not produced until very late in development ( Figure 2E, Table 1). A similar, but less pronounced, phenotype and overall more variability in plant size was observed for iCLF plants (Supplemental Figure 1C and Supplemental Table 1). Most of the cauline leaves in emf2-10 vrn2-1 and iCLF in SD did not carry paraclades or any other secondary structure in their axils (73.6%, n > 750 for emf2-10 vrn2-1; 54.4%, n > 500 for iCLF). Thus, emf2-10 vrn2-1 and iCLF lines displayed an uncoupling of bolting and flower production under SD conditions, indicating an expanded I 1 phase and a requirement for Pc-G proteins to promote the transition from I 1 to I 2 ( Figure 1).
Notably, we observed single flowers in the axils of the last one to five cauline leaves of emf2-10 vrn2-1 mutants and rudimentary bracts at the base of the pedicels of the flowers in both Pc-G-depleted lines ( Figure 1L; Supplemental Figure 2G), indicating a gradual switch from the I 1 to I 2 phase. Surprisingly, many of the flowers produced in SD in the wild type also showed rudimentary bracts at the base of their pedicels (84.0% for La-0 and 42.5% for Columbia-0 (Col-0), first 20 flowers of 10 plants) and occasionally also stipules at their base (Supplemental Figures  2E and 2F). Thus, even in Arabidopsis wild-type, SD conditions induce leaf-like characteristics on flower metamers.
The leafy shoot phenotype of emf2-10 vrn2-1 in SD strongly resembled that described for ft lfy or fd lfy double mutants (Ruiz-Garcia et al., 1997;Abe et al., 2005;Wigge et al., 2005) but LFY expression was only moderately reduced (Supplemental Figures  3A, 3B, and 4). Consistent with the extensive uncoupling between bolting and flower production, expression of AP1, which is a marker for floral meristem identity, was largely eliminated during the early stages of emf2-10 vrn2-1 inflorescence development in SD but increased as flowers were produced (Supplemental Figures 3C, 3D, and 4). In addition, expression of FT followed a similar trend as AP1, whereas SOC1 and FUL, two other key regulators of flowering, showed wild-type expression levels in all stages analyzed (Supplemental Figure 4). Importantly, the floral repressor FLC but not SVP was upregulated in leafy emf2-10 vrn2-1 inflorescence apices (Supplemental Figure 4). Using a FLC pro : FLC-GUS (b-glucuronidase) reporter, we confirmed the higher promoter activity of FLC in emf2-10 vrn2-1 inflorescences compared with the wild type (Supplemental Figures 5A and 5B). Furthermore, the reporter indicated an upregulation of FLC in emf2-10 vrn2-1 rosette leaves, which may explain the delayed flowering of emf2-10 vrn2-1 in SD (Supplemental Figures 5C and 5D).
Collectively, our analyses of leafy emf2-10 vrn2-1 inflorescences revealed changes in gene expression for FLC, AP1, and FT, but not for LFY, SOC1, FUL, and SVP.  Table 2 for detailed organ counts.
Loss of FLC, SVP, and FT Modulate the emf2-10 vrn2-1 Leafy Shoot Phenotype To investigate whether the upregulation of FLC is required for the leafy shoot phenotype of emf2-10 vrn2-1 plants in continuous SD, we created emf2-10 vrn2-1 flc-5 triple mutants. Because FLC can act in a repressor complex with SVP (Li et al., 2008), we also created emf2-10 vrn2-1 svp-32 triple and emf2-10 vrn2-1 flc-5 svp-32 quadruple mutants. Surprisingly, removing FLC or SVP activity had clearly distinct effects (Table 1). In SD, lack of FLC strongly reduced the number of rosette leaves but had little effect on cauline leaf number. By contrast, removing SVP activity did not significantly affect rosette leaf number (P = 0.44 Student's t test), whereas cauline leaf number was reduced to a quarter of that in emf2-10 vrn2-1. Importantly, the combined loss of both MADS box transcription factors in emf2-10 vrn2-1 flc-5 svp-32 quadruple mutants restored cauline leaf number to near wild-type and rosette leaf number was reduced to slightly fewer than the wild type. These results suggested that the prolonged V phase of emf2-10 vrn2-1 in SD was largely caused by increased FLC, whereas the delay in the I 1 -to-I 2 transition required SVP and to lesser extent FLC activity.
The repressor complex of FLC and SVP delays flowering by binding to the promoter of FT (Li et al., 2008;Deng et al., 2011).
Thus, FLC and SVP have overlapping and distinct functions in the promotion of bolting and the transition from I 1 to I 2 in emf2-10 vrn2-1 mutant plants. In addition, based on genetic analyses of emf2-10 vrn2-1 and emf2-10 vrn2-1 ft-13 mutants, FT is active in promoting the I 1 -to-I 2 transition in emf2-10 vrn2-1 inflorescences in SD.

Photoperiod Shifts Induce Floral Reversion in Plants with Reduced Pc-G Activity
When flowering in continuous LDs, a few (<1% of plants) iCLF inflorescences reverted to an earlier developmental stage, so that after producing a series of flowers as normal they went back to producing a single paraclade, suggesting that floral commitment was impaired (Supplemental Figure 1B). To test this further, we provided a transient photoperiodic flowering induction by shifting iCLF or emf2-10 vrn2-1 seedlings from inductive LD to SD 21 to 25 d after germination (DAG). Several weeks later, iCLF and emf2-10 vrn2-1 plants made some flowers but then reverted to cauline leaves and paraclade generation before eventually resuming making flowers ( Figures 1A, 1F, 2G, and 2H; Supplemental Figure  1D). By contrast, wild-type plants began flowering in LD and did not show obvious floral reversion of IM and FM when they were shifted to SD, except the presence of rudimentary bracts (Table 2;  Supplemental Table 2 and Supplemental Figure 2). These findings indicate that Pc-G function is essential for maintenance of floral meristem identity. We observed a reduction of reversion nodes in both emf2-10 vrn2-1 and iCLF if we delayed the shift to SD, suggesting that plants became less susceptible to reversion as time in LD increased.
To resolve spatial and temporal origin of inflorescence and flower reversion in emf2-10 vrn2-1 and iCLF, we carefully analyzed the floral reversion nodes. Four main classes were revealed, which were the result of either flower reversion or inflorescence reversion or combined flower and inflorescence reversion and usually occurred progressively on the bolting stem (Table 2; Supplemental  Table 2  development, as shown in Figures 1B to 1E and 1G to 1K. Interestingly, flower reversion also occurred after sepals were initiated (<5% of strong reversion nodes; Figure 1K; Supplemental Figure  2D) or even from the fifth whorl after carpels had formed so that inflorescences emerged from within the fourth whorl siliques (<1% of strong reversion nodes; Supplemental Figures 2B and 2C). Notably, although the different reversion nodes usually appeared in the order (1) to (4) (see above), we observed fluctuations in the order of successive reversion nodes in emf2-10 vrn2-1 and iCLF (e.g., leaf/ flower nodes appeared frequently between nodes with empty cauline leaves; Supplemental Figure 2L), suggesting that meristem identity was unstable and fluctuating around a threshold between the vegetative and reproductive state.
To investigate the morphological changes in the IM and in young primordia in the early stage of floral reversion, we analyzed the distribution of ANT and STM RNA in tissue sections of shifted emf2-10 vrn2-1 apices 6 d after shift (DAS) as these two genes provide indirect markers for the cryptic bracts that subtend early stage floral primordia (Long and Barton, 2000). The cryptic bracts were more prominent in shifted emf2-10 vrn2-1 plants than in the wild type; notably, some primordia at stage 2 did not express STM and therefore likely lacked meristematic activity (Supplemental Figures 6A to 6D), suggesting that reversions occur mostly in very early primordium development.
Thus, our detailed analyses of floral reversion in Pc-G mutants indicate that Pc-G proteins are not only required for a timely transition to flower production after bolting in SD but are also needed to maintain floral formation when plants are shifted from inducing to noninducing photoperiods.

Floral Reversion in Pc-G Mutants Depends on FLC and SVP
We next sought to determine whether the Pc-G target genes FLC and SVP are not only required for a delayed bolting and transition from I 1 to I 2 in Pc-G mutants, but also responsible for floral reversion. Analysis of FLC expression by RT-PCR and an FLC pro : FLC-GUS reporter in inflorescence apices revealed an increase in emf2-10 vrn2-1 and iCLF plants relative to the wild type, most notably when plants were shifted to SD (Figures 3A and 4A;Supplemental Figures 5E to 5H and 7A). For SVP, RT-PCR revealed a significant upregulation after the shift for emf2-10 vrn2-1 but not for iCLF inflorescence apices ( Figure 4A; Supplemental Figure 7B). Analysis of the spatial expression pattern by in situ hybridization showed stronger accumulation of SVP mRNA in inflorescence meristems and in the vasculature beneath the meristem of emf2-10 vrn2-1 plants shifted to SD compared with emf2-10 vrn2-1 plants continuously grown in LD ( Figure 3B). However, the overall expression pattern of SVP was not changed in shifted stage 2 emf2-10 vrn2-1 floral primordia and was excluded from the cryptic bract as in the wild type (Supplemental Figures 6E and 6F). Chromatin immunoprecipitation (ChIP) assays using wild-type inflorescence apices revealed H3K27me3 enrichment at FLC and SVP ( Figure 4B; Supplemental Figure 7D), confirming that Pc-G proteins directly regulate FLC and SVP in the inflorescence as well as in seedlings. While H3K27me3 at FLC was strongly reduced in emf2-10 vrn2-1 and iCLF in LD and SD, no decrease was detected at SVP ( Figure  4B; Supplemental Figure 7D).
We concluded that floral reversion in Pc-G mutants depends on additive and synergistic function of FLC and SVP and that Pc-G proteins are needed to keep FLC repressed and H3K27me3 decorated in wild-type inflorescences when plants are shifted from LD to SD. As the analyzed Pc-G mutants represent only partial loss-of-Pc-G-function, SVP H3K27me3 may be maintained by the residual Pc-G activity present in the mutants.
SVP, as part of a repressor complex, delay flowering by binding to the promoters of SOC1 and FT in leaves (Hepworth et al., 2002;Helliwell et al., 2006;Searle et al., 2006;Fujiwara et al., 2008;Li et al., 2008;Jang et al., 2009;Deng et al., 2011). Both genes were expressed in both LD, SD, and after LD to SD shift in wild-type inflorescence apices (Figures 3C and 5A; Supplemental Figures 4 and  8A). Therefore, regulation of FT expression is apparently different in leaves and inflorescences as its expression in leaves was strongly reduced when plants were shifted to SD as previously reported (Corbesier et al., 2007;Hiraoka et al., 2013) (Figures 3D and 5B). We confirmed that FT is active in wild-type inflorescences after the shift to SD using FT pro :LUCIFERASE and FT pro :GUS reporters, which revealed strongest expression in the vasculature of the stem and of pedicels (Figures 3 and 5), consistent with recent studies (Adrian et al., 2010;Hiraoka et al., 2013). Plants grown in continuous SDs showed weaker but detectable FT expression in the inflorescence ( Figure 5E), suggesting that high FT expression in the inflorescence under noninductive conditions may require a transient photoperiodic induction. We then analyzed flowering-promoting genes in Pc-G mutant backgrounds. Neither the expression of the FLC/SVP target gene SOC1 nor of FUL, which is needed with SOC1 to prevent floral reversion (Melzer et al., 2008), was significantly decreased in shifted emf2-10 vrn2-1 mutants compared with shifted wild-type plants (Supplemental Figure 8A). Similarly, the expression of AP1 was also not altered, whereas LFY expression was increased in emf2-10 vrn2-1 inflorescences compared with wild-type plants both in LD and SD (Supplemental Figure 8B). By contrast, FT expression in emf2-10 vrn2-1 and iCLF inflorescences was strongly reduced in LD relative to the wild type and further reduced when plants were shifted to SD (Figures 3C and 5A; Supplemental Figure 7C). It was restored to near wild-type levels in emf2-10 vrn2-1 flc-5 and emf2-10 vrn2-1 flc-5 svp-32 inflorescences but not in emf2-10 vrn2-1 svp-32, indicating that FLC but not SVP activity caused the FT downregulation ( Figure 5A). Thus, only the activity of the floweringpromoting gene FT was reduced in emf2-10 vrn2-1 after transient photoperiodic induction, whereas the key meristem identity genes SOC1, FUL, and LFY were unchanged or even elevated. The correlation between FT level and floral reversion together with our genetic analysis showing that the leafy inflorescence phenotype of emf2-10 vrn2-1 plants grown in continuous SD is largely due to reduced FT activity (Figures 2E and 2F; Supplemental Figure 4) suggested that reduced FT expression might be the basis for floral reversion in shifted emf2-10 vrn2-1 plants. To further test this, we used an estradiol-inducible FT construct in transgenic emf2-10 vrn2-1 plants (iFT emf2-10 vrn2-1) to activate FT in specific tissues and at distinct time points. When FT was induced by (A) and (B) In situ RNA hybridization: a shift from LD to SD leads to FLC (A) and SVP (B) upregulation in emf2-10 vrn2-1 (ev), but not in wild-type inflorescence apices (27 DAG). Expression is detected in the vasculature of the stem (arrowheads) and in the inflorescence meristem (arrow). A control experiment using STM RNA probes verified that these differences did not reflect variation in sample preservation (Supplemental Figure 6). Bars = 100 mm. treating whole plants or apices with estradiol after shifting to SD, floral reversion was partially suppressed (Figures 5F to 5I, Table 3;  Supplemental Table 3). Together, these results suggest that FT activity in emf2-10 vrn2-1 inflorescences is required to prevent floral reversion in short days, when FT activity is lacking in leaves.

FT Maintains Inflorescence and Floral Meristem Identity
Our results raised the question of whether FT activity also has a role in maintaining floral commitment in wild-type inflorescences. FT acts redundantly with LFY to activate AP1 in floral meristems (Ruiz-García et al., 1997;Abe et al., 2005;Wigge et al., 2005), so that ft lfy double mutants bolt and generate cauline leaves but never produce flowers, suggesting a role for FT in specifying floral meristem identity. To reveal a role for ft single mutants in maintaining floral commitment, we performed a careful analysis of metamers in the ft-10 (Col) null allele. In LD, these ft mutant plants resemble SD-grown wild-type plants in lacking FT activity in leaves, but differ in that they also lack FT activity in inflorescences. As previously reported, ft-10 mutants flowered late in LD and SD and showed an increase in cauline leaf number, particularly in SD where ft-10 resembled iCLF (Supplemental Table 1). In addition, we observed inflorescence and flower reversion in ft-10 in both LD and SD, but never in the corresponding wild type ( Figure 2L, Table  4). About 70% of the pedicels of ft-10 mutant flowers carried rudimentary bracts under LD (Supplemental Figure 2H; first 10 flowers of 18 plants) phenocopying the rudimentary bracts in the wild type in SD (Supplemental Figures 2E and 2F). Importantly, the ft-10 mutants initiated flower production, but then showed an oscillation of leaf and flower emergence in the following nodes suggesting that FT activity in inflorescences is required for the maintenance of IM and FM identity.

Pc-G Proteins Are Needed for Floral Commitment
Pc-G proteins have been shown to regulate numerous traits in plants, and a common underlying theme is that they maintain changes in developmental phase by repressing genes that promote an earlier phase. For example, after seed germination, the Pc-G permit transition from embryonic to seedling identity by repressing regulatory genes, such as FUSCA3, that are normally expressed earlier during embryonic development and specify embryonic identity (Chanvivattana et al., 2004;Makarevich et al., 2006). Here, by genetically manipulating Pc-G activity, we reveal roles for Pc-G in maintaining phase transitions as they promote commitment to flowering after transient photoperiodic induction and induce the switch from cauline leaf to flower production in noninductive conditions. These roles are consistent with a function of Pc-G in sustaining developmental transitions as they involve repression of genes such as FLC that are normally expressed during the earlier vegetative phase and repress the floral transition (Michaels and Amasino, 1999;De Lucia et al., 2008;Pien et al., 2008). Our genetic analyses indicate that activities of both FLC and SVP are critical for reversion and the switch to flower production in emf2-10 vrn2-1 mutants ( Figure 6). Both genes are direct targets of Pc-G, as they are enriched for H3K27me3 in inflorescence tissue. This is necessary for FLC regulation: FLC loses H3K27me3 and is upregulated in iCLF and emf2-10 vrn2-1 mutants. For SVP, the role of Pc-G is less clear; we did not see consistent decreases in H3K27me3 enrichment at SVP in Pc-G mutants and observed only minor differences in expression in inflorescences. It is therefore possible that the role of SVP in floral reversion is indirect (at least for the Pc-G mutants (A) Quantitative RT-PCR analyses of FLC and SVP mRNA expression in emf2-10 vrn2-1 (ev) inflorescence apices (harvested 8 h after lights on) normalized to elF4, relative to expression in La-0 (LD). Each bar represents the mean of five biological replicates 6 SE. Asterisks indicate significant change of expression (Student's t test: P # 0.05) compared with the equally treated wild-type control (La-0). (B) ChIP assay: FLC and SVP chromatin are enriched for H3K27me3 in inflorescence apices of the wild type (La-0). H3K27me3 is strongly reduced at the FLC locus in emf2-10 vrn2-1 but not at SVP. (+) H3K27me3 antibody; (2) no antibody control. Data represent mean of two biological replicates 6 SE. (C) Loss of both FLC and SVP suppress floral reversion in emf2-10 vrn2-1 plants; reversion nodes per plant, reversions of IM to tSAM (indicated by cauline leaves), and of FM to IM (indicated by paraclades) are distinguished 6 SE (n $ 40). See Table 2 for detailed counts of the different reversion node types. evfs, emf2-10 vrn2-1 flc-5 svp-32.
[See online article for color version of this figure. ] analyzed in this study), i.e., it is required as a cofactor for another, unknown factor ("X") that is misexpressed in Pc-G mutants or that chromatin of SVP target genes promoting reversion are more accessible in Pc-G mutants and activated ( Figure 6). As the Pc-G mutants analyzed here have residual EMF2 (in emf2-10 vrn2-1 mutants) or CLF activity (in iCLF), SVP H3K27me3 may be conferred by the EMF-PRC2, whereas FLC H3K27me3 and repression may be mediated by both EMF-PRC2 and VRN-PRC2 (Chanvivattana et al., 2004; Figure 6). Nonetheless, it seems likely that the role of Pc-G, FLC, and SVP in floral commitment is conserved in flowering plants. The Pc-G proteins are present in all plants, and recent research suggests that FLC is much more widely conserved in angiosperms than previously thought, having been identified in monocots (Reeves et al., 2007;Ruelens et al., 2013). FLC plays SE. ev, emf2-10 vrn2-1; evs, emf2-10 vrn2-1 svp-32; evf, emf2-10 vrn2-1 flc-5; evfs, emf2-10 vrn2-1 flc-5 svp-32; plants were 27 DAG and inflorescence apices were harvested 8 h after lights on. (C) to (E) FT pro :LUCIFERASE activity in wild-type shoots 50 DAG in LD (C), 11 DAS from LD to SD (D), and in continuous SD (E). Note that after LD to SD shift, no expression in leaves is seen, whereas expression is readily detected in inflorescence apices (arrowhead). Color coding represents relative counts per second (cps). a role in floral reversion in natural environments as the ecotype Skye-0 shows FLC-dependent floral reversion (Poduska et al., 2003). In addition, SVP is widely conserved, and in barley (Hordeum vulgare), a monocot that is distantly related to Arabidopsis, SVP orthologs are also expressed during vegetative development and repressed during FM development (Trevaskis et al., 2007). Furthermore, experiments using transgenics show that misexpression of SVP orthologs in flowers causes floral reversion, particularly when plants were grown in noninductive short days (Trevaskis et al., 2007). It is therefore plausible that the Pc-G/FLC/SVP module provides a general mechanism for floral commitment.

Differing Roles of FLC and SVP in Reversion
Unexpectedly, our genetic analysis revealed that FLC and SVP have distinct roles in floral reversion. First, the emf2-10 vrn2-1 flc-5 svp-32 quadruple mutant shows less floral reversion than either the emf2-10 vrn2-1 flc-5 or the emf2-10 vrn2-1 svp-32 triple mutant, suggesting that FLC and SVP act in parallel. Second, removing FLC activity in the emf2-10 vrn2-1 background increases FT expression, whereas removing SVP activity does not, suggesting that the role of SVP in promoting reversion is independent of FT ( Figure 6). Third, the phenotypes of the two triple mutants differ, with SVP activity being more specifically required for the production of cauline leaves. Arabidopsis is unusual in that its flowers are not subtended by a leaf-like organ termed a bract. However, detailed analysis has suggested that after the floral transition, the youngest primordia to arise on the flanks of the inflorescence meristem are leaf-like organs, which usually do not grow out during later flower stages and hence are termed cryptic bracts (Figure 1) (Long and Barton, 2000;Kwiatkowska, 2006;Chandler, 2012). The different types of reversion nodes observed in the Pc-G mutants likely reflect the time at which floral primordium identity is compromised and reversion occurs. For example, cauline leaves with "empty" axils may arise if floral primordium identity is inhibited extremely early, likely before or around the time the FM is specified on the adaxial side of cryptic bract primordia. Inhibition at later stages may result in cauline leaves subtending paraclades, paraclades only, or in weaker cases flowers subtended by cauline leaves or rudimentary bracts (Figure 1). SVP therefore seems to inhibit early stages in floral primordium development, whereas FLC has more continuous effects, perhaps consistent with the more general misexpression of FLC in emf2-10 vrn2-1 inflorescences.

FT Is Needed for Floral Commitment
The FT gene was originally identified based on its requirement to accelerate time to flowering in LD in Arabidopsis. Subsequently, FT or its orthologs have been found to act more generally and to control other traits, including bud dormancy, potato tuberization, leaf dissection, and lateral bud outgrowth (Böhlenius et al., 2006;Shalit et al., 2009;Navarro et al., 2011;Hiraoka et al., 2013). Our analyses reveal a daylength-independent role for FT in the maintenance of IM and FM identity.
We and others have detected FT pro :GUS reporter expression in the vasculature of the pedicels of older flowers in inflorescences in LD and SD ( Figures 3C and 5C) (Adrian et al., 2010;Hiraoka et al., 2013). We believe this reflects native FT expression as RT-PCR detects FT mRNA in inflorescence tips of shifted wild-type plants but much less so in shifted emf2-10 vrn2-1 plants. Additionally, the reporters that contain 8.1 kb of sequences upstream of FT reliably reflect the expression of FT in vasculature in leaves in LD but not SD and the same promoter fragment driving FT cDNA complements null ft-10 mutations (Adrian et al., 2010). Importantly, our genetic analysis shows that the FT expression we detect in the inflorescence is functionally relevant. In particular, we show that ft mutants grown in LD or SD conditions show floral reversion after bolting (Table 4). Thus, although FT plays a negligible role in triggering flowering in SD, as it is not expressed in leaves (Kobayashi and Weigel, 2007), once bolting and floral induction has occurred it is transcriptionally activated in the inflorescence and here has a role in maintaining floral commitment. Our results further suggest that 1.6 6 0.5 0.6 6 0.2 8.8 6 1.4 1.9 6 0.3 12.8 6 1.6 ev iFT + estradiol 0.1 6 0.1* 0 0** 0* 0.1 6 0.1** Data are 6 SE. n = 12 plants. CL, cauline leaves; PC, paraclades; F, flowers; ev, emf2-10 vrn2-1; ev iFT, emf2-10 vrn2-1 iFT. Significantly fewer reversion nodes than ev 2 estradiol (Student's t test): *P #0.01 and **P < 0.001.  the decreased FT expression in emf2-10 vrn2-1 inflorescences is partly responsible for their loss of floral commitment. First, in emf2-10 vrn2-1 flc svp quadruple mutants, which show drastically reduced floral reversion relative to emf2-10 vrn2-1 double mutants, FT expression in inflorescences is restored to near wild-type levels. Second, when a FT transgene is induced in LD to SD shifted emf2-10 vrn2-1 plants, floral reversion is reduced. Since removing SVP activity in the emf2-10 vrn2-1 background reduces floral reversion without restoring FT activity, it is likely that additional genes besides FT are involved. It is also notable that emf2-10 vrn2-1 ft-13 triple mutants show a more extreme phenotype than do emf2-10 vrn2-1 double mutants in SD. This reflects the fact that emf2-10 vrn2-1 reduces but does not eliminate FT activity and implies that there are ev-independent pathways that act in parallel to regulate FT.

Persistent FT Activity Is Needed for Floral Commitment
How does FT activity in the inflorescence maintain floral commitment in plants shifted to noninductive conditions? In transiently induced plants, it is likely that the LFY/AP1 autoregulatory loop stabilizes FM identity once LFY/AP1 expression has been initiated in young floral primordia, but this does not explain how flower primordium identity is repeatedly established on the flanks of the SAM. Here, movement of mobile FT protein into the SAM from flower primordia may be important for continued production of flowers. Although the generality of this mechanism is unknown, characterization of FT orthologs from legumes has revealed a correlation between floral commitment and FT expression in inflorescences (Hecht et al., 2011;Sun et al., 2011), and genetic studies in tomato (Solanum lycopersicum) show that FT is needed for floral commitment as well as for floral induction (Molinero-Rosales et al., 2004). In addition, floral commitment in Impatiens balsamina requires a leafderived, mobile signal, which is consistent with florigen being the signal (Tooke and Battey, 2000). Thus, there are other examples in diverse plant species that stable floral commitment does not only rely on signaling events in the meristem but requires continuous signaling from lateral organs producing a mobile signal.
In conclusion, FT activity in inflorescences is required for floral commitment, and the Pc-G genes are needed to permit this. However, an important question remains as to how FT transcription is activated or maintained in inflorescences in LD-to-SD-shifted plants. FT itself is unlikely to provide the "memory" of transient inductive signals in shifted plants, but rather is needed for maintaining commitment. Thus, shifting experiments show that plants that are only at a very early stage in inflorescence development lacking stage 3 or older flower primordia at the time of shifting (Torti et al., 2012) are fully committed, yet we do not detect FT transcription in the IM or very early flower primordia. Genes such as FUL, whose expression persists in the IM and which are needed for floral commitment, are strong candidates to provide the initial memory (Melzer et al., 2008;Torti et al., 2012). We detected FT transcription in the vasculature of young flowers (Figure 3) and speculate that FT may be activated downstream of floral organ identity or other genes active during flower development. Regardless of how this occurs, the activation of FT in young flowers would provide a mobile signal (FT protein) that reinforces flower primordium identity on the flanks of the IM. In this way, a promotive signal from older flowers back to the IM would make flowering selfperpetuating once initiated. At least two differently composed PRC2 complexes, the EMF2-PRC2 and the VRN2-PRC2, prevent daylength-dependent floral reversion. (1) SVP is a target of H3K27me3 in inflorescences that is likely mediated by the EMF2-PRC2 as loss of VRN2 (in emf2-10 vrn2-1) or SWN (in iCLF) do not reduce H3K27me3 levels at SVP. (2) In both lines with reduced Pc-G activity, H3K27me3 is strongly reduced at the FLC locus and FLC mRNA accumulates. (3) At least one other H3K27me3 target (X) promotes daylength-dependent floral reversion (5), since lack of SVP (4) and FLC (6) only incompletely suppress daylength-dependent floral reversion in emf2-10 vrn2-1 mutants. Factor X is likely misexpressed and loses H3K27me3 in partially depleted Pc-G mutants and may be a cofactor for SVP. The upregulation of FLC (6) but not SVP (4) represses FT expression in emf2-10 vrn2-1 mutants. (7) ft mutants show photoperiod-independent floral reversion, whereas induced FT suppresses daylength-dependent floral reversion in emf2-10 vrn2-1 mutants. In addition, FT is also regulated by pathways acting in parallel to ev (not indicated; see Discussion).

Construction of iCLF
We generated the iCLF (clf-28 swn-7 CLF pro :CLF-GR) line by introducing the previously described CLF pro :CLF-GR ) steroiddependent transgene into the clf-28 swn-7 background by floral dip transformation of swn-7 clf-28/+ plants. In the presence of dexamethasone steroid, iCLF plants are near wild-type (Supplemental Figure 1) as CLF activity complements both swn-7 and clf-28 mutations due to the functional redundancy between the SWN and CLF genes (Chanvivattana et al., 2004). In the absence of steroid, iCLF plants have a weaker phenotype than clf-28 swn-7 mutants (Supplemental Figure 1), suggesting that the CLF pro :CLF-GR transgene is slightly leaky. In depletion experiments, iCLF plants were grown on tissue culture plates containing 10 mM dexamethasone for 10 or 14 d and then transferred to soil, resulting in a progressive decrease in Pc-G activity from day 10 to 14 onwards. iCLF FLC pro :GUS plants were generated by crossing FLC pro :GUS into the clf-50 swn-3 CLF pro :CLF-GR background.

Shift Experiments
Seeds were sown on soil, grown for 21 d (emf2-10 vrn2-1 and La-0) or 29 d (iCLF and Col-0) in LD, then shifted to SD and retained there for at least 4 weeks before examination of the phenotype. Rosette leaves and inflorescence apices, excluding flowers older than stage 13 (Smyth et al., 1990), were harvested 27 DAG (emf2-10 vrn2-1 and La-0) or 34 DAG (iCLF and Col-0) for gene expression analyses. Additionally, the length of the initial period in LD was varied from 7 d to 5 weeks to determine whether sensitivity to reversion varied with developmental age. Because emf2-10 vrn2-1 flc-5 and emf2-10 vrn2-1 flc-5 svp-32 flowered slightly earlier than emf2-10 vrn2-1 in LD (Table 1), differences in reversion might reflect that they were at a more advanced developmental stage at the time of shift. The time in LD was therefore varied, between 7 and 28 d, before shifting to SD so that each genotype was shifted at a range of different stages.

RNA Extraction and Real-Time RT-PCR Analysis
Total RNA of two to six biological replicates of leaves and inflorescence tips, excluding open flowers (see Shift Experiments above), was extracted with TRIZOL (Invitrogen) and cDNA was synthesized using M-MLV reverse transcriptase (Promega). The primers used are listed in Supplemental Table 4.

In Situ RNA Hybridization
In situ RNA hybridization was performed using digoxigenin-labeled mRNA probes as described previously (Chanvivattana et al., 2004). STM, ANT, and LFY antisense probes were generated using plasmids generously provided by R. Simon. FLC (Pien et al. 2008) and SVP plasmids for antisense probes (Hartmann et al., 2000) were kindly provided by C. Dean and P. Huijser, respectively. AP1 antisense probe was generously provided by Coral Vincent and George Coupland (Torti et al., 2012).

ChIP Assay
ChIP assays were performed as described previously . The primers used are listed in Supplemental Table 5.

Luciferase Imaging
Initially, plants were sprayed with a solution of 20 mM luciferine (Synchem) to degrade accumulated LUCIFERASE and kept for 1 h. Plants were then sprayed again and imaged in a NightOWL II LB 983 (Berthold Technologies). Exposition time was 10 min and resolution 1 3 1 pixels.

GUS Staining
Detection of GUS activity in tissue preparations were performed as described with minor modifications (Jefferson et al., 1987).

Imaging
Photoshop adjustment involved only image exposure using adjustment levels and sharpening using unsharp mask.

Supplemental Data
The following materials are available in the online version of this article. Supplemental Figure 8. qRT-PCR Analyses of LFY, SOC1, FUL, and AP1 in emf2-10 vrn2-1 (ev) and Wild-Type Inflorescence Apices. Supplemental Table 3. Suppression of Floral Reversion in ev by High FT.