Conserved roles for Polycomb Repressive Complex 2 in the regulation of lateral organ development in Aquilegia x coerulea ‘Origami’

Background

Epigenetic regulation is necessary for maintaining gene expression patterns in multicellular organisms. The Polycomb Group (PcG) proteins form several complexes with important and deeply conserved epigenetic functions in both the plant and animal kingdoms. One such complex, the Polycomb Repressive Complex 2 (PRC2), is critical to many developmental processes in plants including the regulation of major developmental transitions. In addition, PRC2 restricts the expression domain of various transcription factor families in Arabidopsis, including the class I KNOX genes and several of the ABCE class MADS box genes. While the functions of these transcription factors are known to be deeply conserved, whether or not their regulation by PRC2 is similarly conserved remains an open question.

Results

Here we use virus-induced gene silencing (VIGS) to characterize the function of the PRC2 complex in lateral organ development ofAquilegia x coerulea“折纸”,降低成员eudicot跑unculales. Leaves with PRC2 down-regulation displayed a range of phenotypes including ruffled or curled laminae, additional lobing, and an increased frequency of higher order branching. Sepals and petals were also affected, being narrowed, distorted, or, in the case of the sepals, exhibiting partial homeotic transformation. Many of the petal limbs also had a particularly intense yellow coloration due to an accumulation of carotenoid pigments. We show that theA. x coerulea花卉马德斯盒基因AGAMOUS1(AqAG1),APETALA3-3(AqAP3-3) andSEPALLATA3(AqSEP3) are up-regulated in many tissues, while expression of the class I KNOX genes and several candidate genes involved in carotenoid production or degradation are largely unaffected.

Conclusions

PRC2 targeting of several floral MADS box genes may be conserved in dicots, but other known targets do not appear to be. In the case of the type I KNOX genes, this may reflect a regulatory shift associated with the evolution of compound leaves.

Maintenance of proper gene expression in differentiated cells is essential for the development of multicellular organisms and epigenetic regulation is an important player in this process Reviewed in: [1–3]．One family of proteins with deeply conserved functions in epigenetic regulation is the Polycomb Group (PcG). The PcG was first discovered inDrosophila melanogasteras repressors of the HOX genes [4]．Several PcG complexes exist in both plants and animals, each with distinct functions in epigenetic silencing Reviewed in: [5,6]．However, only the Polycomb Repressive Complex 2 (PRC2) has been well characterized in multiple plant models Reviewed in: [5,7]．The main function of the PRC2 complex is trimethylation of lysine 27 of histone H3 (H3K27), a histone modification known to suppress gene expression [8]．The PRC2 contains four core proteins; the histone methyltransferaseEnhancer of Zeste(E(z)),and three other proteins thought to enhance PRC2 binding to nucleosome [9]．These includeSuppressor of Zeste 12 (Su(z)12)andExtra Sex Combs (ESC),known respectively asEMBRYONIC FLOWER 2(EMF2) andFERTILIZATION INDEPENDENT ENDOSPERM(FIE) in plants, andMulti-Copy Suppressor of IRA 1 (MSI1) Reviewed in: [10]．TheE(z)lineage in plants has experienced an ancient duplication such that most angiosperms have at least two paralogs, known asCURLY LEAF(CLF) andSWINGER(SWN) [11]．Many plant species have additional duplications in the core PRC2 loci that allow them to form several PRC2 complexes often with distinct developmental functions [12,13]．

PRC2 is involved in a number of important developmental transitions. In the plant model systemA. thaliana, these functions include endosperm development, early repression of flowering to allow proper vegetative development, the eventual transition to flowering, and flower organogenesis [14–17]．In grasses, the PRC2 complex plays roles in floral induction (rice and barley), flower development (rice), suppressing cell divisions in the unfertilized ovule (rice), and endosperm development (rice and maize) [12,18,19]．In the moss modelPhyscomitrella patens, PRC2-dependent remodeling appear to be required for the switch from gametophyte to sporophyte development [20,21]．

In addition to its role in developmental transitions, PRC2 has been suggested to function in lateral organ development inA. thaliana. In fact, the first description of a plant PRC2 function was discovered with the characterization of theclfmutant inA. thaliana[17]．Theclfplants had severely curled leaves, smaller narrower sepals and petals, and partial homeotic transformations of sepals and petals towards carpel and stamen identity, respectively. Two MADS box genes, the C class memberAGAMOUS(AG)and the B class representativeAPETALA3(AP3) were shown to be over-expressed inclfmutants, suggesting that the PRC2 complex was required for stable repression of these genes [17]．This was particularly interesting because MADS box genes regulate homeotic floral organ identity in plants somewhat analogously to the way HOX genes regulate segment identity in animals [22–25]．Further studies have subsequently shown that the E class MADSSEPALLATA3(SEP3) is similarly up-regulated inclfmutants [26]．PRC2 has also been shown to regulate the expression of the class I KNOX genes during vegetative development. The class I KNOX genes are a family of homeobox domain-containing loci in plants that have conserved roles in promoting pluripotency in the shoot apical meristem and in compound leaf development [27,28]．Katz et al [29] found that in addition to the phenotypes reported inclfmutant plants,FIEcosuppressed plants also had loss of apical dominance and fasciated stems, rolled leaves with varying degrees of serration, loss of phyllotaxy in the inflorescence, and many problems with ovary and ovule development. They further demonstrated that several class I KNOX genes, includingBREVIPEDICELLUS(BP),KNOTTED-LIKE FROM ARABIDOPSIS THALIANA 2(KNAT2), andSHOOTMERISTEMLESS(STM), were over-expressed in rosette leaves ofFIEsilenced plants. Inclfmutants,STMandKNAT2were over-expressed butBPwas not, possibly because theCLFparalogSWNwas acting redundantly. The class I KNOX genesMOSS KNOTTED1-LIKE 2and5(MKN2andMKN5) were also shown to be over-expressed inPpFIEmutant gametophytes [21,30], suggesting that PRC2 targeting of the class I KNOX genes may be deeply conserved.

While the functions of the floral ABC class and type I KNOX genes are thought to be conserved across angiosperms, comparative studies of their regulation have largely focused on upstream transcription factors, such asLEAFYor ARP family members [31,32]．In order to begin addressing the question of whether PRC2-targeting interactions are similarly conserved, we have examined the functions of PRC2 members in lateral organ development of the emerging model systemAquilegia. The genusAquilegiais a member of an early diverging lineage of the eudicotyledonous flowering plants, the Ranunculales, that arose before the radiation of the core eudicots Reviewed in: [33]．It therefore can be used as a rough phylogenetic midpoint betweenA. thalianaand model systems in the grasses [34]．Additionally, many ecological, evolutionary and genetic studies have been conducted inAquilegiaover the past 50 years. These have taken advantage of its small genome (n = 7, approximately 300 Mbp) as well as a number of more recent genomic tools, including the fully sequencedAquilegia x coeruleagenome (http://www.phytozome.net/search.php?method=Org_Acoerulea) Reviewed in: [33,35]．The reverse genetic tool virus-induced gene silencing (VIGS) has been optimized in several species ofAquilegia[36] for both leaf and floral development [37–40]．Previously we examined the evolution and expression of the PRC2 family inAquilegia[41] and found that the genome contains a simple complement of PRC2 homologs: one copy each of the two plantE(z)homologs,AqCLFandAqSWN; anESChomolog,AqFIE; aSu(z)12homolog,AqEMF2; and a copy ofMSI1, AqMSI1. We initially assessed gene expression throughoutAquilegia vulgarisdevelopment due to its strong vernalization dependency and found no obvious tissue or stage specialization. Furthermore, the ancient paralogs,AqCLFandAqSWN,are not imprinted inAquilegiaendosperm as is seen in other plant species [19,41]．

In the current study we have used VIGS to knock down the expression ofAqFIE[Genbank: JN944599] andAqEMF2[Genbank: JN944598] in unvernalized and vernalizedAquilegia coerulea‘Origami’ plants using theANTHOCYANIN SYNTHASE(AqANS) as a marker gene. Due to limitations of the VIGS approach, it is not possible to assess many life cycle transitions, most notably flowering time, but lateral organ development can still serve as a useful model for PRC2 function. We find that PRC2 plays a role in leaf and floral organ development inA. x coerulea,particularly via down-regulation of the floral MADS box genes. This has allowed us to identify PRC2 targets that appear to be conserved between Arabidopsis andAquilegiaas well as some novel PRC2-regulated pathways.

Virus-induced gene silencing

TheAquilegiaVIGS protocol was preformed as described previously [36]．TRV2-AqCLF-AqANS,TRV2-AqSWN-AqANS,TRV2-AqFIE-AqANSand TRV2-AqEMF2-AqANSconstructs were prepared by PCR amplifying approximately 300 bp regions of each gene using primers that addedEcoR1 and XbaI restriction sites to the 5′ and 3′ ends of the PRC products (see Additional file1). The PCR products were then purified and cloned into the TRV2-AqANSconstruct [36] and electroporated intoAgrobacteriumstrain GV101.A. x coeruleaseedlings were grown to approximately the 4 to 6 leaf stage and then either treated as described in Gould and Kramer [36] for unvernalized samples or as described in Sharma and Kramer [37] for plants that had been vernalized for approximately 4 weeks at 4°C [36,37]．The TRV2-AqANSand TRV2-AqFIE-AqANSconstructs were each used to treat approximately 400 plants over 4 rounds of VIGS. Approximately 250 of these plants were VIGS treated before vernalization and approximately 150 of these plants were treated after vernalization. The TRV2-AqEMF2-AqANSconstruct was used to treat approximately 100 plants; roughly 50 of these plants were treated before vernalization and 50 were treated after vernalization. Leaves, petals, and sepals showingAqANSsilencing were photographed, collected, and stored at -80°C for RNA analysis.

RT-PCR

RNA was extracted from control (AqANSsilenced) and experimental (AqFIEandAqEMF2VIGS-treated) tissue. One half of theAqANSsilenced (control) leaves were from separate unvernalized plants (C1 and C2) and half were from separate vernalized plants (C3 and C4). Five of the TRV2-AqFIE-AqANStreated leaves were from separate unvernalized plants (F2, F4, F5, F7, and F8) while three were collected from separate vernalized plants (F1, F3, and F6). All of the TRV2-AqEMF2-AqANSleaves were collected from separate vernalized plants (E1-E4). Sample numbers do not indicate order of leaf appearance but were collected at roughly the same stages of development. We selected a variety of observed phenotypes for each set of samples. For leaves, the RNeasy Mini Kit (Qiagen, Valencia, CA) was used. For petals and sepals RNA was extracted using the Pure-Link Plant RNA Reagent small scale RNA isolation protocol (Ambion, Austin, TX). RNA was treated with Turbo DNase (Ambion, Austin, TX) and cDNA was synthesized from 1 μg of total RNA using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA) and oligo (dT) primers. cDNA was diluted 1:5 prior to use.

Amplification was performed using AccuStart PCR SuperMix (Quanta Biosciences Inc, Gaithersburg, MD). The amplification program began with 1 minute activation step at 94°C, followed by a 20 second denaturing step at 94°C, a 15 second annealing step at 55°C, and a 15 second extension at 72°C, repeated for 30 cycles. This cycle number was chosen for optimal detection ofAqFIEandAqEMF2, which are expressed at relatively low levels in mature organs, especially compared to the high expression levels ofAqIPP2. All primers used are listed in Additional file1. Amplification ofISOPENTYL PYROPHOSPHATE:DIMETHYLALLYL PYROPHOSPHATE ISOMERASE2 (AqIPP2)was used as a positive control [38,42]．To test for expression ofAPETALA3-1 (AqAP3-1), APETALA3-2 (AqAP3-2), APETALA3-3 (AqAP3-3),andFUL-like- 1 (AqFL1) in VIGS-treated leaves, cDNA from several leaves were pooled together prior to amplification. The control pool consisted ofAqANS-silenced control leaves C1-4, theAqFIEVIGS-treated pool consisted ofAqFIEleaves F3-6, and theAqEMF2VIGS-treated pool consisted ofAqEMF2leaves E1-4.

qRT-PCR

cDNA was prepared from VIGS-treated tissue as described above. For the carpel sample, carpels were collected from 3 anthesis stage wild type plants and pooled together. RNA was extracted using the RNeasy Mini Kit (Qiagen, Valencia, CA) and treated as described above. cDNA from VIGS-treated tissue was then pooled together and diluted 1:10. The control sepal pool consisted ofAqANS-silenced control sepals C1-4, the control petal pool consisted ofAqANS-silenced control petals C1-4, theAqFIEsepal pool consisted ofAqFIEVIGS-treated sepals F2, 3, 5, and 6, theAqFIEpetal pool consisted ofAqFIEVIGS-treated petals F2, 3, 5, and 6, theAqEMF2sepal pool consisted ofAqEMF2VIGS-treated sepals E2, 3, and 4 s, and theAqEMF2petal pool consisted ofAqEMF2VIGS-treated petals E1 and E2. qRT-PCR was performed using PerfeCTa qPCR FastMix, Low ROX (Quant Biosciences Inc., Gaithersburg, MD) in the Stratagene Mx3005P QPCR system to study the relative expression ofAqAG1andAqAG2. AqIPP2expression was used for value normalization. All primers are listed in Additional file1.

Microscopy

Petals from wild type,AqANSVIGS-treated, andAqEMF2VIGS-treated plants were stored at -80°C and then warmed to room temperature and mounted whole on glass slides in water. Cells were visualized in the Harvard Center for Biological Imaging on a Zeiss AxioImager Z2 microscope using trans-illumination with white light. Images were taken using a Zeiss AxioCam Mrc digital camera.

We treated both unvernalized and vernalized plants with TRV2 constructs containing eitherAqANS-AqFIEorAqANS-AqEMF2fragments. TRV2-AqANStreated plants were used as controls throughout. Phenotypes ofAqFIEandAqEMF2silenced plants were equivalent and will be discussed together. We also treated a small number of unvernalized plants withAqANS-AqCLFandAqANS-AqSWNVIGS constructs. Phenotypes from these plants were similar to those seen inAqFIEandAqEMF2,but were weaker (data not shown), most likely due to partial redundancy betweenAqCLFandAqSWN.Thus we chose to focus onAqFIEandAqEMF2VIGS-treated tissue. As is common for VIGS-treated plants, we recovered a range of phenotypes in a small percentage of VIGS-treated plants (roughly 10-15% of plants in each round) [36]．In the current experiment there is the added component that phenotypes are likely due to mis-expression of PRC2 target genes, and are therefore likely to have an added complexity due to ectopic expression of a potentially wide range of target loci.

Vegetative phenotypes

Wild typeAquilegialeaves are compound, typically bearing three leaflets that are themselves divided into two to three lobes (Figure1A). Although these leaflets are often relatively deeply lobed, they do not generally produce elongated, higher order petiolules within the leaflets. However,A. x coeruleadoes display heteroblasty over the course of its lifespan, varying leaf morphology as the individual progresses from the vegetative to the reproductive stage (Additional file2). In late reproductive adult stages, higher order petiolules may be observed in which the central lobe of each leaflet becomes itself a separate leaflet borne its own petiolule (Additional file2C). Using the terminology of Kim et al. 2003 [32], all of these leaf forms are non-peltately palmate in that the leaflets are not radially positioned around the terminus of the primary petiole.

We observedAqANSsilencing in 10-15% of treated plants across theAqFIE- andAqEMF2-VIGS experiments. In addition to theAqANS大雨如注ncing, the leaves of these plants showed a complex set of phenotypes. The most consistently observed perturbation was curled or ruffled laminae that typically curled toward the abaxial surface (~10-12% of treated plants and, thus, the majority of silenced plants) (Figure1F, H, J-L). We also observed an increased frequency of higher order branching in which fully formed petiolules developed within the leaflet, creating as many as ten or twelve distinct leaflets rather than the usual three (Figure1B-F, H, L and Additional file2E). While we have never observed such higher order branching in control leaves, either in the context of these experiments or others [40], we obtained 15 leaves from a total of 10AqFIE-andAqEMF2-treated plants that exhibited increased branching. When quantified (Additional file2E), the presence of higher order petiolules is significant at p < 0.05 for unvernalized lateral leaflets but not significant for the other stages/leaflet types. However, it is obvious that there is much more branching variation in silenced leaflets than in controls. In many cases, the margins of the laminae had additional lobing relative to control leaves (observed in 10% of treated plants) (Figure1B-E, K) and, in a small number of cases, the central lobe of the terminal leaflet was severely reduced (seen in multiple leaves from 5 plants) (Figure1D, M). Laminar area was highly variable with some leaflets appearing to have expanded area (~25 plants) (Figure1F) while others seemed reduced (~8 plants) (Figure1G, I, M). In two plants, ectopic finger-like projections were observed on the adaxial surface of laminae (Figure1F), which was never observed in control leaflets.

Floral phenotypes

Wild typeA. x coerulea花朵拥有五种器官类型:萼片,花瓣,stamens, staminodia and carpels [39]．We have focused on the sepals and petals because they showed strong phenotypes in the silenced flowers. Wild type sepals are flat and ovate with an entire margin (Figure2A-B). The petals are notable for the presence of a long hollow nectar spur, which forms near the attachment point (Figure2A). This feature divides the organ into two regions, the proximal spur and the distal limb. Spurs inA. x coeruleaare typically 5-6 cm in length and slightly curved. The limb region is relatively flat with a rounded, weakly lobed margin (Figure2C). In 25 flowers from vernalizedAqFIE-andAqEMF2-treated plants, we observed sepals that were narrower than wildtype organs and dramatically folded towards the adaxial surface (Figure2D, F, L, P). In severely affected flowers, petals were narrowed and stunted (10 flowers, Figure2D, G, Q) or exhibited sharply bent spurs (12 flowers, Figure2H-I, K, M, Q). In twoAqEMF2大雨如注nced flowers, the sepals exhibited chimeric petal identity including ectopic spur formation (Figure2M-N). Perhaps most surprising, many of the perianth organs had a definite yellow hue, with the petal limbs showing particularly intense yellow coloration (observed in at least one organ from 15 flowers) (Figure2E, I-M, O). Such coloration was not observed inAqANS大雨如注nced control flowers (Figure2A-C). Examination of theAqFIE- andAqEMF2大雨如注nced organs under high magnification reveals that yellow pigment is deposited in plastids (Additional file3A), consistent with carotenoids rather than the vacuole-based aurones that are produced in someAquilegiaspecies [43,44]．

Assessment of AqFIE and AqEMF2 down-regulation

Due to limited RNA availability, we used standard RT-PCR to assess target gene down-regulation in leaves, sepals and petals compared to their expression inAqANSsilenced control tissue. Even inAqANSsilenced (control) tissue,AqFIEandAqEMF2are expressed at low levels relative to the loading controlAqIPP2.The experimental samples (F1-8 and E1-4) were selected to represent all of the observed phenotypes and were derived from separate plants. This analysis demonstrated that in the TRV2-AqFIE-AqANStreated plants,AqFIEwas strongly down-regulated, being undetectable in a number of samples (Figure3A and Figure4). Likewise,AqEMF2expression is reduced to undetectable levels in most testedAqEMF2大雨如注nced samples (Figure3A and Figure4). We also tested forAqEMF2inAqFIE-treated plants and vice versa, and found thatAqEMF2levels are often reduced inAqFIE-treated leaves, although the reciprocal is generally not true (Figure3A). Furthermore, we tested the other PRC2-complex members,AqCLFandAqSWN, and found no consistent evidence of their down-regulation in either type of silenced tissue (Additional file4A).

Assessment of candidate gene expression

We tested for ectopic expression of a wide panel of potential target genes, with a focus on the floral organ identity loci and type I KNOX homologs (Figure3, Figure5, and Additional file4B). We again compared the expression of these genes to expression inAqANS-silenced control tissue. One of the twoA. x coeruleathe C class MADS box genes,AGAMOUS 1(AqAG1) (see Additional file1for all gene identification numbers), is consistently up-regulated in silenced leaves and floral organs. The secondAGAMOUShomologs,AGAMOUS2(AqAG2) may also be slightly up-regulated in some of the leaves, althoughAqAG2shows basal expression in control floral organs (Figures3and5). The threeA. x coerulea SEPALLATAparalogs (AqSEP1, AqSEP2,andAqSEP3) are somewhat difficult to assess because they are variably expressed in control leaves butAqSEP3in particular seems to be up-regulated inAqEMF2大雨如注nced leaves (Figure3B). These genes were not assessed in floral organs because they are already broadly expressed in these tissues.A. x coeruleaalso has three paralogs of the B class MADS box gene,APETALA3(AqAP3-1, AqAP3-2,andAqAP3-3). The petal-specificAqAP3-3locus is highly up-regulated inAqEMF2大雨如注nced sepals, which also showed chimeric sepal/petal identity in several cases (Figure5A). Additionally two of the threeAP3paralogs are moderately up-regulated in PRC2 VIGS-treated leaves (Additional file4B), but the expression ofAqAP3-1andAqAP3-2is unaffected in mature sepals and petals (Figure5A and B). We also looked at the expression ofFUL-like 1(AqFL1), which is normally expressed in early leaves, but no ectopic expression was detected (Figure5and Additional file4B).

Next, we tested for up-regulation of three of the fiveA. x coeruleaclass I KNOX genes. No significant ectopic KNOX gene expression could be detected in the leaves. Weak expression of theSHOOTMERISTEMLESS2 (AqSTM2) andKNOTTED(AqKN) homologs is detected inAqEMF2大雨如注nced leaves (Figure3C), however, we also occasionally detected comparable expression of these genes in control (AqANS-silenced) leaves, so it is difficult to ascribe significance to this expression. Given the lack of clear up-regulation in leaves and due to a limited amount of floral RNA, class I KNOX gene expression was not tested in the floral organs.

AlthoughAqAG1is consistently over-expressed inAqFIEandAqEMF2silenced sepals and petals, we never saw any evidence of carpel identity in these organs. We therefore pooled cDNA from severalAqANS-(control), AqFIE-,andAqEMF2-treated petals and sepals and used qRT-PCR to further examined the expression ofAqAG1andAqAG2in these organs as well as in wild type carpels (Figure5C and D). We found that whileAqAG1was clearly up-regulated inAqFIEandAqEMF2silenced organs compared to the controls,AqAG1expression was still much lower than in wild type carpels (about 0.05 to 0.2 fold). In contrast,AqAG2expression was similar in control and PRC2 silenced tissue, but much lower than in wild type carpels.

Lastly, in an effort to investigate the carotenoid production, we identified the likelyA. x coeruleahomologs of a range of components of the carotenoid pathway inA. thaliana, including enzymes involved in production (PHYTOENE SYNTHASE(PSY) andCAROTENOID ISOMERASE(CRTISO)) and breakdown (CAROTENOID CLEAVAGE DIOXYGENASE 4(CCD4) and9-CIS-EPOXYCAROTENOID DIOXYGENASE 3(NCED3)) of carotenoids [45]．A. x coeruleahas two copiesCCD4(AqCCD4andAqCCD4L)and two genes that are closely related toA. thaliana PSY(AqPSYL1andAqPSYL2). Previous studies inA. thalianahave indicated that bothCRTISOandNCED3are positively epigenetically regulated by other SET domain containing proteins so we were particularly interested in the expression of these genes inAqFIEandAqEMF2down-regulated tissue [46,47]．We used RT-PCR to examine the expression of these six genes inAqANS-(control),AqFIE-,andAqEMF2-treated petals (Additional file3B). Given the observed phenotypes, we might expect the expression ofAqPSYL1,AqPSYL2, orAqCRTISOto be up-regulated orAqCCD4, AqCCD4L, orAqNCED3to be down-regulated. Unfortunately, no clear patterns are apparent from these reactions.

AqFIEandAqEMF2VIGS-treated plants displayed a range of lateral organ phenotypes. Silenced leaves often had ruffled or curled lamina, additional lobing, and an increased frequency of higher order branching. The perianth organs were generally narrower than wild type organs. Sepals were also curled and petals were stunted or had bent spurs, while petal limbs also had a particularly intense yellow coloration seemingly due to an accumulation of carotenoid pigments in these cells. Many of the phenotypes we observed are similar to those seen inclfmutants andFIEcosuppressedA. thaliana, including curled leaves and narrow perianth organs [17,29]．Unlikeclfmutants andAGover-expressers inA. thaliana,dramatic transformation towards carpel identity was not observed in theAqFIE-andAqEMF2-treated sepals or petals. However, the level ofAqAG1expression in these organs was much less than what is seen in wild typeAquilegiacarpels. Interestingly, the distinct folded morphology of the sepals may suggest slight transformation towards carpel identity as silenced leaves were folded towards the abaxial surface while the sepals were dramatically folded towards the adaxial surface, which is similar to the folding pattern of theAquilegia心皮(48]．

It is interesting to note that inAqFIEsilenced leaves,AqEMF2is also down-regulated. The reverse is not true inAqEMF2silenced leaves, andAqEMF2expression is not affected inAqFIEsilenced floral organs. This result suggests that PRC2 may be directly or indirectly regulatingAqEMF2expression inA. x coerulealeaves, which could account for the generally more severe phenotypes observed inAqFIEsilenced leaves compared toAqEMF2silenced leaves.AqEMF2is the only member of the complex that appears to be PRC2-regulated as the expression ofAqCLFandAqSWNis not affected in PRC2 down-regulated leaves. In general, the potential for this type of cross-regulation is relatively unexplored inA. thalianaand, therefore, bears further study.

In our analysis of candidate target genes, we found thatAqAG1is often ectopically expressed in PRC2 down-regulated tissue.AqAP3-3andAqSEP3are also up-regulated in some organs, but expression of the class I KNOX genes and several candidate genes involved in carotenoid production or degradation seem largely unaffected. Mutations inAGandSEP3are known to suppress the curled leaf phenotype inclf突变体植物虽然表达这些尼box genes, which themselves function together in a complex [49], is thought to be the cause of the curled leaf phenotype [26]．It is, therefore, possible that over-expression ofAqAG1andAqSEP3is similarly responsible for many of the observed phenotypes inAqFIEandAqEMF2silenced leaves. These findings lead us to conclude that PRC2-based regulation ofAGandSEP3homologs is deeply conserved in eudicots. It has recently been shown that several chromatin remodeling factors associate with MADS complexes and one model is that an important function of MADS domain complexes may be to recruit chromatin remodeling complexes to target loci in order to alter transcription of these genes and direct organ development [50,51]．For example, RELATIVE OF EARLY FLOWERING 6 (REF6) was enriched in protein complexes that were isolated via immunoprecipitation using tagged ABCE class MADS box proteins [50]．REF6 has been shown to specifically demethylate H3K27me3, the histone modification deposited by PRC2 [52]．Activation ofSEP3byAPETALA1(AP1) inA. thalianaresults in the reduction of H3K27me3 at theSEP3promoter, suggesting thatAP1may recruitREF6to theSEP3promoter in order to help induceSEP3gene function [50]．Our data suggests that this key dependency on epigenetic regulation for the switch from vegetative to floral development may be important outside ofA. thaliana. There are some complications, however. Of the twoA. x coerulea AGhomologs, only one,AqAG1, is strongly regulated by PRC2. Perhaps consistent with this observation, sequencing of theAquilegiagenome (http://www.phytozome.net/search.php?method=Org_Acoerulea) reveals thatAqAG1does contain the large regulatory second intron that is common toAGhomologs [53,54] whileAqAG2’s second intron is much smaller. These results suggest that PRC2 regulation can be directed in a paralog-specific fashion and may even play some role in the distinct expression patterns observed among these gene copies [39]．

The class I KNOX genes are directly or indirectly regulated by PRC2 in bothA. thalianaandPhyscomitrella, however, we detected little or no increase in KNOX gene expression in ourAqFIEandAqEMF2silenced leaves. This is somewhat surprising because of the higher order branching that we observed in silenced leaves, including several of the tested RNA samples. The class I KNOX genes are thought to play a role in compound leaf development in a number of species. In many, but not all, compound leafed taxa where KNOX gene expression has been studied, includingAquilegia, it has been shown that the genes are expressed in the shoot apical meristem and down-regulated in incipient leaf primordia (P0), but subsequently turned back on in early leaf primordia [28]．Down-regulation of class I KNOX genes in the leaves of models such as tomato orCardaminecauses reduced branching while over-expression leads to increased branching [55,56], suggesting that KNOX genes act to maintain indeterminacy in compound leaves and promote leaflet initiation.

There are several possible explanations for why we did not observe significant ectopic KNOX gene expression in our VIGS-treated leaves. First, it is possible the KNOX genes were ectopically expressed early in leaf development when the higher order branching actually developed, but were later down-regulated by redundant mechanisms, such asASYMMETRIC LEAVES 1(AS1)-mediated repression [57,58]．InA. thaliana AS1mediated silencing of some of the KNOX genes has been shown to require the PRC2 complex and it is thought that AS1 and AS2 directly recruit the PRC2 complex to KNOX loci [59]．However, it is important to remember that in other taxa with compound leaves, the KNOX andAS1homologs have lost their mutually exclusive regulatory interactions and are expressed together at later stages [32]．This may suggest that theAS1-dependent epigenetic silencing of KNOX genes that has been described in several simple-leafed models [57,58] does not hold for plants with compound leaves. Along these lines, it is also possible that the increased branching phenotypes are due to other factors, such as accelerated phase change or novel genetic mechanisms regulating leaflet branching inAquilegia. For instance, a recent functional study of the geneAqFL1inA. x coerulearevealed that it promotes proper leaf margin development, a unique finding for homologs of this gene lineage [40]．这就提出了一个实际的可能性rs other than the KNOX genes contribute to compound leaf branching inAquilegia.

In addition to the conserved role in regulatingAG, AP3,andSEP3, A. x coeruleaPRC2 may target novel pathways, including those regulating carotenoid production or degradation. InA. thalianapatches of yellow anther-like tissue are observed onclfmutant petals [17]．However, the yellow pigmentation we observed is due to the accumulation of carotenoids in the plastids rather than to a partial homeotic transformation. While genes in the carotenoid pathway are not known to be suppressed by PRC2, some loci are positively epigenetically regulated inA. thaliana. Previous studies have shown that a major enzyme in the carotenoid biosynthesis pathway, CRTISO, requires the chromatin modifying enzyme SET DOMAIN GROUP 8 (SDG8) to maintain its expression [46]．NCED3, an enzyme that cleaves some types of carotenoids as a part of abscisic acid (ABA) synthesis, is similarly epigenetically regulated by theA. thalianatrithorax homolog ATX1 [47]．While none of the genes we tested were consistently up- or down-regulated inAqFIEandAqEMF2silenced petals, carotenoid production is very genetically complex and we were unable to test all of the candidate loci [60]．Thus, it seems likely that PRC2 regulates an as yet unidentified enzyme in this pathway inA. x coerulea.

• A critical role for PRC2 in maintaining the repression ofAG,SEP3, and possiblyAP3appears to be conserved across eudicots. This conservation underscores the importance of chromatin remodeling factors in regulating the floral transition and the proper localization of floral organ identity.

• Class I KNOX genes are not ectopically expressed in PRC2 down-regulated tissue inA. x coerulea, possibly due to a regulatory shift associated with the evolution of compound leaves.

• A. x coeruleaPRC2 plays a significant role in regulating the carotenoid pathway in floral organs, which has not been observed in other taxa.

• This study, the first to examine PRC2 function in angiosperms outsideA. thalianaor the grasses, highlights how little we still know about the general conservation or targeting mechanisms underlying PRC2 function in major developmental transitions.

AG:

AGAMOUS

ANS:

Anthocyanin synthase

AP1:

APETALA1

AP3:

APETALA3

Aq:

Aquilegia

BP:

Brevipedicellus

bp:

Base pair(s)

CCD:

CAROTENOID CLEAVAGE DIOXYGENASE

cDNA:

DNA complementary to RNA

CLF:

CURLY LEAF

cm:

Centimeter

CRTISO:

CAROTENOID ISOMERASE

DNA:

Deoxyribonucleic acid

DNase:

Deoxyribonuclease

E(z):

Enhancer of zeste

EMF2:

EMBRYONIC FLOWER 2

ESC:

Extra sex combs

Eudicots:

Eudicotyledonous

FIE:

FERTILIZATION INDEPENDENT ENDOSPERM

FIS:

FERTILIZATION INDEPENDENT SEED

FLC:

FLOWERING LOCUS c

FL1:

FRUITFUL-like 1

H3K27:

Histone H3 Lysine 27

HOX:

Homeobox

IPP2:

Isopentyl pyrophosphatedimethylallyl pyrophosphate isomerase:

KN:

KNOTTED

KNAT2:

KNOTTED-like from Arabidopsis thaliana 2

KNOX:

knotted1homeobox gene

MADS:

MCM1, agamous, deficiens, SRF

MEA:

Medea

MKN:

Moss knotted1-like

MSI1:

Multi copy suppressor of IRA 1

n:

Chromosome number

NCED:

9-CIS-epoxycarotenoid dioxygenase

oligo:

Oligodeoxyribonucleotide

PcG:

Polycomb group

Pp:

Physcomitrella patens

PCR:

Polymerase chain reaction

PRC1:

Polycomb repressive complex1

PRC2:

Polycomb repressive complex2

PSY:

Phytoene Synthase

qRT-PCR:

Quantitative real time polymerase chain reaction

REF6:

Relative of early flowering 6

RNA:

Ribonucleic acid

RT-PCR:

Reverse transcriptase PCR

SEP:

SEPALLATA

STM:

SHOOTMERISTEMLESS

Su(z)12:

Suppressor of zeste12

SWN:

SWINGER

TRV:

Tobacco rattle virus

VIGS:

Virus-induced gene silencing.

The authors would like to thank members of the Kramer lab and 2 anonymous reviewers for critical comments on the manuscript.

Affiliations

1. Department of Organismic and Evolutionary Biology, Harvard University, 16 Divinity Ave., Cambridge, MA, 02138, USA

   Emily J Gleason & Elena M Kramer

2. Department of Molecular and Cellular Biology, Harvard University, 16 Divinity Ave., Cambridge, MA, 02138, USA

   Emily J Gleason

Corresponding author

Correspondence toElena M Kramer.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

EJG helped to conceive of the study, carried out the experiments, and drafted the manuscript. EMK helped to conceive of the study, supervised the experiments, and helped draft the manuscript. Both authors read and approved the final manuscript.

This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and Permissions