Dosage-Sensitive Function of RETINOBLASTOMA RELATED and Convergent Epigenetic Control Are Required during the Arabidopsis Life Cycle

The plant life cycle alternates between two distinct multi-cellular generations, the reduced gametophytes and the dominant sporophyte. Little is known about how generation-specific cell fate, differentiation, and development are controlled by the core regulators of the cell cycle. In Arabidopsis, RETINOBLASTOMA RELATED (RBR), an evolutionarily ancient cell cycle regulator, controls cell proliferation, differentiation, and regulation of a subset of Polycomb Repressive Complex 2 (PRC2) genes and METHYLTRANSFERASE 1 (MET1) in the male and female gametophytes, as well as cell fate establishment in the male gametophyte. Here we demonstrate that RBR is also essential for cell fate determination in the female gametophyte, as revealed by loss of cell-specific marker expression in all the gametophytic cells that lack RBR. Maintenance of genome integrity also requires RBR, because diploid plants heterozygous for rbr (rbr/RBR) produce an abnormal portion of triploid offspring, likely due to gametic genome duplication. While the sporophyte of the diploid mutant plants phenocopied wild type due to the haplosufficiency of RBR, genetic analysis of tetraploid plants triplex for rbr (rbr/rbr/rbr/RBR) revealed that RBR has a dosage-dependent pleiotropic effect on sporophytic development, trichome differentiation, and regulation of PRC2 subunit genes CURLY LEAF (CLF) and VERNALIZATION 2 (VRN2), and MET1 in leaves. There were, however, no obvious cell cycle and cell proliferation defects in these plant tissues, suggesting that a single functional RBR copy in tetraploids is capable of maintaining normal cell division but is not sufficient for distinct differentiation and developmental processes. Conversely, in leaves of mutants in sporophytic PRC2 subunits, trichome differentiation was also affected and expression of RBR and MET1 was reduced, providing evidence for a RBR-PRC2-MET1 regulatory feedback loop involved in sporophyte development. Together, dosage-sensitive RBR function and its genetic interaction with PRC2 genes and MET1 must have been recruited during plant evolution to control distinct generation-specific cell fate, differentiation, and development.


Introduction
Independent evolution of multicellularity and thus the cell types has implications for adaptation of distinct developmental strategies in plants and animals [1]. Adaptive mechanisms unique to higher plants include alternation between the reduced gametophytic and dominant sporophytic generations, absence of a distinct germ line, and continuous postembryonic development. Unlike animals that develop a germline early in development, the progenitors of gametophytic cell types are derived from sporophytic cells of a mature plant, which acquire competence to undergo meiosis and subsequent mitotic divisions and to establish cell fates of gametic and accessory cell types [2,3]. Further, double fertilization of gametes leads to the development of an embryo and endosperm. Upon germination, the mature embryo develops into an adult plant by recurrent morphogenetic patterning. Therefore, plant cells must have a flexible but coordinated molecular machinery that helps to maintain their state of competence for cell fate determination and differentiation of distinct cell types during their developmental ontogeny [4][5][6]. In particular, dynamic control of cell fate and differentiation in plants is achieved by regulators of the cell cycle and chromatin complexes in distinct developmental stages, unlike stable gene repression by the same type of regulators during animal development [7,8].
The tumour suppressor Retinoblastoma (pRB) and closely related proteins are primarily known as negative regulators of the cell cycle and for their antiproliferative activity in multicellular organisms [9,10]. Specifically, pRB forms a repressive complex with E2F transcription factors to control cell cycle progression from G1 into S phase. Less is known how the pRB pathway functions beyond cell cycle, whether in coordinating cell proliferation and differentiation, or to control early cell fate establishment until late developmental processes [9,11]. In recent years, pRB homologues have been shown to be necessary in the control of cellular differentiation, stem cell maintenance, and apoptosis in diverse model systems [9,12,13] including Arabidopsis [14][15][16][17]. Evolutionary homologues of pRB, either alone or in cooperation with chromatin-associated regulators, can regulate genes involved in cell fate determination and differentiation [9,13,18], suggesting a central role of this protein in early cell fate control, as well as subsequent maintenance of the differentiated state and genome integrity [9,12].
In Arabidopsis, RETINOBLASTOMA RELATED (RBR) is the single homologue of pRB, and the pRB-E2F pathway is largely conserved [7,9]. Unlike the mouse embryo-lethal pRB knockouts, Arabidopsis knock-out alleles of RBR are defective in both female and male gametogenesis [14,15], constraining functional dissection of the pre-and post-gametophytic role of RBR in development. Studies that down-regulated RBR in distinct tissues using RBR RNA interference, virus induced gene silencing or by misexpression of a RBR-binding viral protein to compete with the native RBR, have not elucidated the genetic behaviour of a rbr null mutation during gametophyte or sporophyte development [16,17,[19][20][21]. In addition, it was unclear in these experiments if both RBR mRNA and protein levels were stably reduced throughout development, or aberrantly elevated due to the autoregulatory function of the pRB-E2F pathway [22]. Nonetheless, these studies have provided an early indication that the RBR pathway functions distinctly in different cell types to prevent cell division, endoreduplication and stem cell maintenance. Recent work demonstrated that RBR genetically interacts with the conserved epigenetic regulators of the Polycomb Repressive Complex 2 (PRC2) to control development of both male and female gametophytes [15], and that RBR control of cell fate in the male gametophyte is at least partly coupled to its genetic interaction with the cell cycle associated pollen-specific CYCLIN-DEPENDENT KINASE A1 (CDK A1) [17]. Unlike in the sporophytic leaf, RBR is repressed by a maternal and paternal PRC2 complexes during plant reproduction [15], suggesting that the RBR regulatory network can function differently depending on the developmental context. Together, the developmental role of RBR during sporophytic development remains poorly understood, primarily due to the lack of genetic tools.
In this study, we investigated the effects of an Arabidopsis RBR knock-out allele [14,15] on the plant life cycle. Detailed analysis of rbr female gametophytes supported the role of RBR in gametophytic cell fate control. Further, we performed a tetraploid genetic analysis that provided direct evidence that at reduced levels of RBR sporophyte development is perturbed. When only one out of four functional RBR alleles was present in tetraploids triplex for rbr (rbr/rbr/rbr/RBR), specific stages of sporophytic differentiation and development were affected. The function of RBR is therefore partially haplo-insufficient during sporophytic plant development, as revealed by RBR dosage analysis in tetraploid plants. Furthermore, we provide genetic evidence that RBR functions in concert with the sporophytic PRC2 subunits to control developmental processes in the sporophyte. In short, our work not only illustrates the coordinated function of the RBR pathway in both gametophytes and the sporophyte, it also demonstrates how tetraploid genetics can be exploited to uncover a novel developmental role of an essential regulator during the entire plant life cycle.

Results/Discussion
RBR is required for cell fate determination in the female gametophyte In Arabidopsis, the fully differentiated female gametophyte (embryo sac) consists of only four cell types of clonal origin [3]: a haploid egg cell, a homo-diploid central cell derived from the fusion of two haploid polar nuclei, two synergids that facilitate entry of sperm cells into the embryo sac ( Figure 1A), and three antipodal cells that undergo early apoptosis. By characterizing one of the RBR knock-out alleles, rbr-3 [14], we could identify that loss of RBR function did not affect the mitotic divisions and cellularization in the female gametophyte [15]. In the majority of cases, however, all cell types including the central cell with unfused polar nuclei commenced proliferation in this mutant ( Figure 1B and 1C). The morphological identity of the proliferating rbr-3 cell types was previously assigned based on their positional information within the embryo sac; however, their molecular identity remained questionable. Therefore, we examined the fate of specific cell types in the absence of RBR using cell type-specific molecular markers that are characteristic for the three cell types of the mature female gametophyte. The marker lines ET1119, ET2634, and ET956 express b-glucuronidase (GUS) in the egg cell, synergid cells, and the central cell, respectively [23,24] (Figure 1D, 1H, 1L). In most proliferating rbr embryo sacs we could not detect GUS expression in the egg, synergid, and central cell ( Figure 1E, 1G, 1I, 1K, 1M, 1N). In 3-8% of the cells, cell type-specific markers showed ectopic expression that deviated from their wild-type pattern ( Figure 1F, 1G, 1J, 1K; Figure S1). These findings were further substantiated by loss of gene expression in rbr embryo sacs for central cellspecific FERTILIZATION INDEPENDENT SEED2 (FIS2) [15,25] and for two additional unpublished egg cell-specific genes (A.J. Johnston, H. Bä umlein, T. Dresselhaus, U. Grossniklaus and W. Gruissem, data not shown). Therefore, RBR is required for the identity establishment of these gametophytic cell types. In the rare cases where these markers were still present, possibly due to some RBR activity carried over from the rbr/RBR heterozygous

Author Summary
Understanding the convergent developmental mechanisms of core cell cycle genes is highly instructive in biology. When these genes are essential in development, lethality precludes mutation analysis throughout the life cycle of an organism. We subjected a homozygous lethal mutation in RETINOBLASTOMA RELATED (RBR) of Arabidopsis for tetraploid genetic analysis to study the function of RBR during the plant life cycle. In diploids, while RBRdeficient female gametophytes with features of aberrant cell fate and differentiation were analogous to what was previously reported for male gametophytes, we provide evidence that RBR controls gametic genome duplication, thus genome integrity in the gametophyte-derived progeny. Quantitative reduction of RBR in tetraploids led to identification of rbr heterozygous plants that displayed novel RBR dosage-dependent phenotypes in differentiation and development of the sporophyte albeit the absence of cell cycle defects. These phenotypes coincided with deregulation of conserved epigenetic factors such as Polycomb Repressive Complex 2 (PRC2) genes and METHYLTRANSFERASE 1 (MET1) in the sporophyte, as shown for the gametophytes as well. However, unlike the repression by the PRC2 in gametophytes, RBR is activated by the sporophytic PRC2 subunits, suggesting that distinct modules of the conserved RBR-PRC2-MET1 loop control gametophyte and sporophyte generations in plants.
megaspore mother cell, they were mis-expressed in the spatial domains of other cell types ( Figure 1F and 1J; Figure S1B, S1C, S1D, S1F). For instance, an egg cell marker and a synergid marker were expressed in the central cell and egg cell domains, respectively, in the absence of RBR. Therefore, RBR not only promotes cell differentiation but also seems to coordinate certain positional information in the female gametophyte.
In a previous work we have shown that the lack of cell differentiation in rbr gametophytes paralleled deregulation of certain PRC2 genes and MET1 [15], whose functional orthologues were known for their role in cell specification, differentiation and also cell cycle regulation in diverse animal systems [26][27][28][29]. Both our present work and a recent report [25] have established that cell fate, cell cycle and ploidy are also impaired in certain RBRdeficient female and male gametophytic cells. There is evidence that RBR directly interacts with MULTICOPYSUPPRESSOR OF IRA1 (MSI1) and FERTILIZATION INDEPENDENT ENDOSPERM (FIE) proteins, which are members of distinct PRC2 complexes in plants [25,30,31]. This is consistent with the findings that central cells in rbr, msi1 and fie mutant female gametophytes aberrantly proliferate and they are defective either in acquiring cellular identity and/or in heterochromatin status [15,25]. The phenotypes in rbr mutant gametophytes can be partly attributed to the derepression of MET1 [15,25], which in turn might result in aberrant hypermethylation, heterochromatin maintenance and/or histone turn over. Interestingly, some of these maternal mutant phenotypes including the defective central cell fate in rbr and msi1 could be rescued by suppressing MET1 and associated global methylation, suggesting a complex epigenetic control of development [25,32]. Taken together, it is possible that the RBR-PRC2-MET1 network controls cell fate determination either independently, by co-regulating cell cycle activity, and/or by forming a repressive chromatin modifying complex both in male and female gametophytes.
RBR plays a prominent role in maintaining ploidy and genome integrity Evolutionary homologues of pRB in animal systems have been implicated in the control of ploidy and chromosomal stability [11]. For instance, pRB-deficient tumors are reported to have elevated aberrant ploidy levels, most likely due to the deregulation of mitotic cell cycle [33,34]. In Arabidopsis, impairment of the RBR-E2F pathway by ectopic expression of the viral RepA protein [19] increased the endocycles in leaf cells. Similar results were obtained when the RBR pathway was perturbed by over-expressing a D3type cyclin [35,36] or E2F/DP transcription factors [7]. In all these cases, however, it remained unclear if the ploidy changes were the primary effect of loss or reduction of RBR function. Therefore, we investigated if reduced or loss of RBR function in a genetically tractable rbr knock-out allele would change developmentally controlled ploidy. Analysis of cellular ploidy in RBR-deficient female gametophytic cell types is difficult due to the problems in isolating these miniature cells from plants that are heterozygous for rbr. During the morphological analysis of diploid rbr/RBR plants using Nomarski optics [15], we noticed that in many instances rbr gametophytic nuclei and, in particular, proliferating nuclei in the central cell region were of unusual size. Therefore, we analysed the ploidy of these nuclei by confocal microscopy and subsequent 3D reconstruction of acquired image stacks. We noticed that several rbr supernumerary nuclei derived from the unfused polar nuclei had a diploid rather than haploid chromosome number ( Figure 1B; 14 observations). This might be due to endoreduplication events in the absence of RBR activity [19], as illustrated by an egg-like cell in the inset of Figure 1C as well, where a large excess of metaphase chromosomes was observed. Given that wild-type polar nuclei are haploid [3] and that rbr polar nuclei do not fuse to form a homodiploid central cell [15], it is likely that RBR restricts not only ectopic divisions but also polyploidization of haploid polar nuclei. Thus, confocal analysis of the ovules allowed us to demonstrate that absence of RBR leads to events of elevated cellular ploidy in the female gametophyte.
Since RBR seems to control ploidy of the female gamete(s), this led us to investigate potential changes in plant genome ploidy in rbr gamete(s)-derived progeny. We had previously shown that a selfed diploid rbr/RBR plant (also referred to as rbr mutant) produced viable progeny segregating for wild-type RBR/RBR and mutant rbr/RBR genotypes, while the female gametophytically lethal rbr allele was not transmitted to the next generation [14]. Therefore, the observed rare polyploid egg cells ( Figure 1C) might not produce an offspring. If the rbr mutant produced viable male gametes with altered ploidy, we would expect that the ploidy of a subset of rbr progeny would be different from the parent plant. Indeed, we found that selfed diploid rbr/RBR plants produced 6% triploids among rbr mutant offspring (n = 56), which produced an array of aneuploid, diploid, and tetraploid plants in the next generation ( Figure S2). This phenomenon is normally not observed in diploid wild-type Arabidopsis. It is most likely that these triploid progeny resulted from the fusion of a haploid RBR egg cell with either a diploid rbr sperm cell or two haploid rbr sperm cells [14]. Unfortunately, we are unable to test these hypotheses in detail because (i) the chance occurrence of these events was estimated to be 0.160.06 = 0.006 considering the rbr transmission efficiency of 0.1 and the presence of 6% triploids among the transmitted mutant progeny, and (ii) rbr knock-out male gametophytes rarely formed sperm cells [15,17]. Together, our results suggest that RBR controls ploidy maintenance in the gametophytic cells and that it is involved in maintaining genome integrity because in its absence or down-regulation polyploid offspring are produced.
Genome-wide polyploidization has played essential role in speciation and thus evolution of plants [37], however, the factors leading to increased ploidy in plants are not completely understood [38,39]. Plant autopolyploidization can be preceded by changes in ploidy either somatically, during meiosis, or during male or female gametogenesis. In case of meiosis, asynaptic mutations and meiotic restitution might lead to formation of unreduced gametes and therefore autopolyploids [40]. Thus far, three Arabidopsis meiotic mutants, dyad [41], mitosis instead of meiosis (mime) [42] and jason [43], were reported to produce unreduced diploid instead of normal haploid gametophytes. Ploidy alterations in female gametogenesis is partly controlled by the maize INDETERMINATE GAMETOPHYTE 1 (IG1) which encodes for a gene with high similarity to ASYMMETRIC LEAVES2 (AS2) in Arabidopsis [44]. The rbr mutation we report here is the first case in Arabidopsis in which aberrations in gametogenesis could result in triploid offspring due to doubling of haploid gametic genome. A ''triploid bridge'' leading to production of diploids, aneuploids and tetraploids may act as a transition between diploids and autotetraploids and therefore could play a significant role in polyploidization [39,45]. An induction of triploid offspring as observed in rbr knock-out mutants may also occur in the wild-type, should RBR activity be altered by unknown environmental factors. Thus, RBR might have played a crucial role in plant evolution by controlling genome duplication events.
rbr mutation is recessive in the gametophytes The roles we propose for wild-type RBR in female gametophytic cell specification and differentiation as well as in maintaining genome integrity are only valid if the rbr-3 allele is gametophytically recessive and genetic reduction of RBR function had caused the observed effects. Since rbr-3 carries a T-DNA insertion in the middle of the RBR gene [14], it might generate a truncated protein with a dominant effect. In order to understand the genetic behaviour of rbr in the gametophyte, we subjected the gametophytically lethal rbr mutation to tetraploid genetic analysis. We asked if the rbr-3 mutation behaves recessive or dominant in diploid gametophytes produced by tetraploid plants by analysing seed set phenotypes and segregation of rbr genotypes in the progeny ( Figure S3; see Table 1 and the Materials and Methods section for details). An autonomously tetraploidized plant that was heterozygous for the rbr mutation ( Figure S2) was subjected to a detailed progeny test (n = 103), which identified three distinct rbr genotypes. Seed set and progeny segregation phenotypes of rbr simplex (rbr/RBR/RBR/RBR), duplex (rbr/rbr/RBR/RBR) and triplex (rbr/rbr/rbr/RBR) mutant plants significantly fit the recessive model of rbr-3 inheritance (x 2 test; p = 0.05), implicating abortion of homozygous rbr/rbr gametophytes and rbr/RBR gametophytes giving viable seeds ( Table 1).
As an additional step to confirm this genetic model, we examined cytological phenotypes of female gametophytes (FG) in these plants. The majority of FGs in both diploid and tetraploid wild-type plants were at stage FG7 upon emasculation, which is typical for wild-type Arabidopsis [15]. We noted that all rbr embryo sacs in a heterozygous rbr/RBR diploid plant showed nuclei proliferation, which significantly fit the expected ratio (x 2 = 0.20, p = 0.65, n = 194) as reported previously [14,15] (Figure S3). In case of the triplex rbr/rbr/rbr/RBR plant, three types of FG genotypes are expected: rbr/rbr, rbr/RBR and rare RBR/RBR. Given that the rbr mutation fits a recessive model of inheritance based on the tetraploid seed set phenotype and progeny test (Table 1), only those FGs that had completely lost RBR function (rbr/rbr) would be expected to show ectopic nuclear divisions, accounting for 54% proliferating FGs ( Figure S2). The observed numbers of embryo sacs with supernumerary nuclei in the triplex plants matched the expectation for proliferation of rbr/rbr embryo sacs (x 2 = 0.076, p = 0.78, n = 162). These data suggest that the viable rbr/RBR female gametophytes are likely phenotypically and functionally identical to RBR/RBR FGs of wild-type tetraploids, and RBR FGs of wild-type diploids.
Together, two independent genetic experiments of seed set and transmission analysis (Table 1) and quantitative analysis of FGs ( Figure S3) confirmed that the rbr mutation behaves recessive in the female (and male) gametophyte(s). Therefore, we can rule out a dominant (negative or positive) effect of a possible truncated version of RBR mRNA or of RBR protein. This situation is perhaps similar to previous reports that premature termination in mouse Rb exons resulted in truncated non-functional proteins [46,47]. Hence, we conclude that the rbr-3 allele is a clear loss-offunction mutation of RBR.

Sporophytic development requires dosage-sensitive function of RBR
The male and female gametophytic lethality of the rbr mutation constrains analysis of RBR function during sporophytic development. Tetraploid analysis is therefore an excellent approach to investigate a dosage dependent function of RBR in the sporophyte.
We recovered and analysed rbr tetraploid plants with different numbers of rbr-3 alleles by large-scale genotyping and segregation analysis of the tetraploid mutant progeny (see Figure 2, Figure S2A and S2B, Table 1). No homozygous tetraploid rbr quadruplex (rbr/ rbr/rbr/rbr) genotypes could be identified, confirming the genetic data (see previous section) that they do not survive postgametophytically. Specifically, we identified second generation rbr triplex plants (rbr/rbr/rbr/RBR) that showed significant quantitative reduction of RBR expression levels in leaves when compared to the wild-type tetraploids (nulliplex RBR/RBR/RBR/ RBR) ( Figure 3C) or heterozygous diploids (data not shown). A subsequent independent expression analysis reconfirmed that the third generation rbr triplex plants also maintained significantly lower RBR expression levels in a RBR dosage-dependent manner ( Figure S4).
We expected that the rbr triplex plants, which displayed nearly 75% reduction in gene expression compared to the wild type ( Figure S4), could reveal quantitative effects of RBR function more readily than diploid rbr/RBR plants. Wild-type RBR/RBR and heterozygous rbr/RBR plants did not differ in sporophytic phenotypes from germination until maturity, indicating full functionality of a single wild-type RBR copy (haplo-sufficiency) at the diploid level. Although rbr triplex plants appeared to grow normally during early sporophyte development (Figure 2A versus Figure 2B), they showed several developmental phenotypes such as stunted growth habit, aberrant leaf size, altered phyllotaxy of siliques, reduced stem thickness, reduced apical dominance, and ectopic floral organs when they were about six weeks old ( Figure 2C-2N). In comparison, the growth and development of rbr nulliplex ( Figure 2C-2N), simplex and duplex rbr plants (not shown) were normal and indistinguishable at similar stages. Thus, the mutant sporophytic phenotypes only occur in rbr triplex plants, indicating that a single RBR copy cannot sustain normal growth and development and, therefore, RBR function is haplo-insufficient in the tetraploid context.
While we anticipated that reduction of RBR would alter cell division and/or cell size as was reported from other systems [9,20], organization and size of cells on the abaxial side of cauline leaves in the triplex mutant surprisingly did not deviate from the corresponding wild-type in young and mature leaves ( Figure S5).
As rbr gametophytes showed ploidy aberrations (as discussed before), we anticipated that ploidy deregulation could also be observed in the leaf sporophyte with decreased RBR dosage. However, neither diploid (rbr/RBR versus RBR/RBR) (data not shown) nor tetraploid (rbr/rbr/rbr/RBR versus RBR/RBR/RBR/ RBR) plants ( Figure S6) had significant changes in leaf ploidy when  analyzed by flow cytometry. Our genetic results are in contrast to earlier reports that deregulation of RBR had immediate consequences on cell divisions and endocycles during leaf organogenesis [19,21]. We reason that retaining one functional RBR copy in diploid and tetraploid systems is sufficient to coordinate cell cycle and specification in the leaf sporophyte and that RBR reduction using a viral RBR-binding protein or virus-induced gene silencing may affect specific functions of RBR that are required for the control of DNA endoreduplication. Next we asked if differentiation of specific cell types was altered in response to RBR dosage change. We examined trichome differentiation patterns in young rosette leaves around 15 days after germination on plates, in diploid and tetraploid plants. As expected, development of trichomes in the 3 rd and 4 th rosette leaves of diploid rbr/RBR plants did not differ from the corresponding wild-type, confirming haplo-sufficiency of RBR in diploids ( Figure 3A). In wild-type tetraploid plants, over 53% of the trichomes had three branches, 34% had four branches and 12% with 5-6 branches ( Figure 3A). These data are consistent with the increased DNA content and supernumerary branching in tetraploids as previously reported [5]. Concomitant with a reduction of RBR dosage in rbr triplex plants, however, there was a significant reduction of 4-branched (11%) and 5-6 branched trichomes (1%) along with an increase of less-differentiated 3branched trichomes (84%) ( Figure 3A). In addition, we observed a similar trend in RBR dosage-dependent reduction of 4-branched trichomes in an independent experiment ( Figure S4). Therefore, the single copy of RBR in rbr triplex is sufficient to specify the trichome cells ( Figure 2B versus Figure 2A) but not sufficient to complete full differentiation of this specialized cell type.
It has been proposed that key cell cycle genes that control ploidy restrict trichome branching [5]. Previous studies of downregulating RBR in diploid leaves provided inconclusive results for ploidy-dependent leaf and trichome differentiation. For example, suppression of RBR in Brassica napus led to elevated ploidy levels in leaves and retarded leaf and trichome development [21]. In contrast, over-expression of a RBR-binding geminivirus RepA protein in diploid Arabidopsis in order to interfere with RBR function revealed only marginal elevation of ploidy levels in mature leaves and supernumerary trichome branching patterns [19]. It is unclear, however, if the RepA protein reduced the endogenous RBR levels in these plants, or if the transcription of RBR was aberrantly elevated due to the autoregulatory function of RBR-E2F pathway [22]. We therefore asked if in single cell trichomes lower RBR levels had caused a concomitant reduction in DNA ploidy, which could explain the fewer branches. By measuring the relative DNA content of individual trichome nuclei by fluorescence microscopy, we found that DNA ploidy in rbr triplex trichome cells was comparable to corresponding tetraploid wild type, and that there was no significant difference within ploidy groups across different genotypes ( Figure S6). Thus, we conclude that cellular differentiation and morphogenesis of trichomes were affected in a RBR dosage-sensitive manner ( Figure 3A). Retaining 25% RBR in the triplex (rbr/rbr/rbr/RBR) plants does not alter the general leaf, trichome and plant ploidy, cell proliferation and trichome specification, but it appears to be insufficient to complete a full differentiation program. This could be particularly true for trichome differentiation, as is also suggested by a recent report that RBR is a target of the trichome cell specification and differentiation factors GLABRA1 and GLABRA3 [48]. Obtaining homozygous rbr trichomes by inducible methods will be required to analyze how RBR controls early specification and/or differentiation. Taken together, the observed sporophytic developmental anomalies including retarded trichome differentiation are a consequence of partial haploinsufficiency of RBR in tetraploids, but not due to RBR-mediated cell cycle deregulation.

RBR expression is dynamically regulated consistent with its function in gametophyte and sporophyte development
RBR is an essential cell cycle regulatory gene that is expressed in the sporophyte (embryo, leaves, root and shoot meristems) and the ovule including the embryo sac [14][15][16]49]. To gain better insight into the dynamic expression pattern of RBR throughout the plant life cycle, we analysed RBR RNA and protein accumulation by in situ hybridization and a transgenic RBR protein reporter line RBR::RFP, respectively [49]. We observed RBR::RFP expression throughout sporophyte development in leaves and seedlings (not shown), also during trichome development ( Figure 4P). In reproductive tissues, RBR mRNA was detected in developing ovules and anthers ( Figure 4A and 4B). In particular, RBR expression was detected in the functional megaspore, the progenitor cell type of the female gametophyte ( Figure 4D). In a fully differentiated embryo sac, we observed RBR mRNA expressed in all the embryo sac cell types such as egg cell, central cell and synergid cells, in addition to the sporophytic cells of the ovule ( Figure 4F). In contrast, RBR::RFP fusion protein was localized primarily in the central cells [49] suggesting posttranslational regulation of RBR in the egg apparatus. A recent study detected RBR::RFP throughout the male gametophyte development [17]. In summary, expression of RBR in all cell types of the gametophytes and the sporophyte, including trichome cells, is consistent with the requirement of RBR for cellular proliferation, cell fate and differentiation of the gametophytic cells, sporophytic development and trichome differentiation.

RBR participates in an epigenetic network to control sporophyte development and trichome differentiation
The dynamic expression of RBR throughout plant development, and its cell-cycle inter-and independent functions reported thus far suggests that RBR is also involved in other regulatory networks. Evolutionary homologues of pRB and epigenetic factors such as PRC2 proteins and DNA maintenance methyltransferase (Dnmt1) have essential roles in controlling cell differentiation and development both in plants and animals [8,9,29]. In animal systems, it has been established that Enhancer of zeste homolog 2 (Ezh2), a core member of PRC2, recruits DNA methyltransferase 1 (Dnmt1), and the resulting maintenance of DNA methylation facilitates formation of more repressive complexes to control distinct developmental processes [50,51]. There is evidence that both PRC2 genes and Dnmt1 exert their function in a cell cycledependent manner. For instance, several PRC2 members and Dnmt1 homologues seem to be directly repressed by the pRB-E2F complexes in plants and animals [29,51]. Furthermore, PRC2 dynamically regulates pRB or RBR via its inherent H3K27me3 activity and possibly through its continuous association throughout the cell cycle [15,52,53]. In Arabidopsis, there are three distinct orthologues of Ezh2, namely CURLY LEAF (CLF), which positively regulates cell size and elongation in the leaf sporophyte; MEDEA (MEA), which negatively regulates cell proliferation and cell size during seed development; and SWINGER (SWN), which enhances the function of both CLF and MEA [8,54]. Similarly, three orthologues of Supressor of zeste 12 (Suz12), which are known to be associated with cell cycle and cell differentiation in animal systems, exist in the Arabidopsis genome. FERTILIZATION INDEPENDENT SEED 2 (FIS2) functions similar to MEA during seed development; VERNALIZATION 2 (VRN2) and EMBRYONIC FLOWER 2 (EMF2) are associated with distinct sporophytic pathways [8]. MET1 is the Arabidopsis orthologue of Dnmt1, which is a key target of RBR and a modifier of several PRC2 genes, and it is critical for coordinated cell division, specification and differentiation of the embryos, and also throughout the sporophytic development [15,[55][56][57]. The mechanisms by which pRB, PRC2 and Dnmt1 homologues control cellular differentiation and development are not completely understood in plant and animal systems. We recently reported that RBR, several PRC2 genes and MET1 are co-regulated by a negative feedback mechanism during gametophyte differentiation and development [15]. Here we asked if a similar mechanism exists in the leaf sporophyte as well.
First, we compared the expression levels of sporophytic PRC2 genes and MET1 in rbr triplex and heterozygous diploid rbr mutant leaf tissues in relation to their corresponding tetraploid or diploid wild-type tissues. Our initial expression analysis in tetraploids suggested that plant to plant variation in expression was quite high. Therefore, we analyzed leaves from individual plants as independent replicates. Prior to gene expression in tetraploids, we examined the expression of MET1 and PRC2 genes in the diploid wild type and rbr/RBR leaf tissues, but we did not detect significant differences in expression levels (not shown). In contrast, we observed that CLF and MET1 were upregulated and VRN2 was downregulated in rbr triplex leaves when compared to the tetraploid wild type (Figure 3C), suggesting the importance of RBR dosage for gene regulation in the tetraploid context. This experiment, however, did not reveal if acute genetic downregulation of RBR below 25% would be required for the deregulation of the PRC2 genes SWN and EMF2. Together, RBR regulates MET1 and the PRC2 genes CLF and VRN2 during leaf development in a dosage-dependent manner. Given that RBR-PRC2-MET1 regulatory network functions during gametophyte development [15], we conclude that RBR control of PRC2 and MET1 is important throughout the plant life cycle.
Recent data suggest that RBR can function downstream of chromatin regulators like PRC2 or transcription factors such as SCARECROW, GLABRA1 and GLABRA3 during distinct stages of plant development [15,16,48]. We therefore asked if PRC2 would reciprocally regulate RBR in the leaf sporophyte. We used two different double mutants that disrupted CLF and SWN, and VRN2 and EMF2, respectively. Sporophytic PRC2 activity is considerably reduced in these double mutants and, consequently, development of the leaf sporophyte is impaired [58]. Although trichomes were correctly specified in the mutant leaves, we observed that their branching was incomplete ( Figure 3B). The majority of the trichomes (92-95%) in diploid wild type leaves differentiated to the mature 3-branched stage. The clf;swn double mutant showed the most severe phenotype in which the majority of trichomes were 2-branched (78%) while only 15% differentiated to the 3-branch stage. The vrn2;emf2 double mutant showed similar phenotypes ( Figure 3B), although the percentage of trichomes that fully differentiated was higher than that of clf;swn, likely because the emf2 allele used here was a weak loss-of-function allele of EMF2 [58]. These data collectively suggest that a novel PRC2-dependent epigenetic mechanism operates to control trichome differentiation in addition to leaf development. Intriguingly, expression levels of both RBR and MET1 were significantly reduced in the mutant leaves ( Figure 3B, 3D, 3E), suggesting that the sporophytic PRC2 complexes activate both RBR and MET1 in leaves. Previous work demonstrated that MEA is derepressed in leaves of PRC2 mutants and that MEA is a direct target of the sporophytic PRC2 [59]. Given that MEA represses paternal RBR in fertilized female gametophytes [15], it is probable that indirect repression of RBR by the sporophytic MEA might have led to reduction of RBR levels in PRC2 mutant leaves. Alternatively, reduction of MET1 in PRC2 mutants is consistent with a previous observation in an animal system that depletion of Ezh2 led to downregulation of Dnmt1 concomitant with local reduction of H3K27me3 activity [60]. Therefore, we propose that both RBR and its target gene MET1 are likely independently activated during leaf development and trichome differentiation either by a cell cycle dependent CLF-PRC2 activity, or indirectly via repression by MEA, which in turn is controlled by the CLF-PRC2 ( Figure 5). However, due to the complexity of tetraploid wild type and mutant plants used in this study, diploid plants deregulating RBR in a temporal and spatial manner will be necessary to test our hypothesis.
A convergent RBR-mediated epigenetic mechanism controls development of the gametophytes and the sporophyte in plants We have provided here direct evidence that RBR has an instructive and dosage-dependent role in cell fate determination, differentiation and development in Arabidopsis. This function is partly mediated by a regulatory loop between RBR and epigenetic regulators such as PRC2 genes and MET1, which operates distinctly in the gametophyte [15] and the sporophyte generations. When RBR function is abolished, such as in female or male rbr gametophytes [15,17], proper cell fate assignment does not occur. However, quantitative reduction of RBR expression in rbr triplex mutant sporophyte does not prevent cell specification but impairs full differentiation, consistent with an earlier study in which stem cell differentiation was delayed when RBR was reduced [16]. Since we did not observe significant changes in ploidy levels in response to reduced RBR dosage, it is likely that RBR-mediated developmental functions can also be cell cycle-independent, similar to pRB control of cell cycle-unrelated processes in animals [18,61,62]. In support of the RBR-PRC2-MET1 epigenetic network that we have identified, a recent study revealed that DNA methylation of RBR, CLF, SWN, VRN2 and EMF2 loci is regulated by MET1 in Arabidopsis sporophyte [63]. We propose that the dynamic modulation of this RBR-PRC2-MET1 circuit was adopted to accommodate the regulation of distinct developmental processes in both gametophyte and sporophyte generations.

Histological analyses
Confocal analysis of ovules and spatial analysis of GUS activity in ovules and seed tissues were performed as described earlier [6,15]. Scanning electron micrographs of leaves were prepared as published [65].

Ploidy measurements in trichomes
Cauline leaves of mutant and wild-type tissues were fixed in a formaldehyde/glutaraldehyde fixative. Intact trichome cells were isolated from leaf epidermis by an established protocol based on removal of Ca 2+ ions [66]. Nuclear images of guard cells from leaves (n = 42 and 54 in wild type and mutant, respectively), and of the trichome cells of the wild-type (n = 80) and the mutant (n = 42) were recorded for fluorescence measurement upon DAPI staining by confocal microscopy (LSM 510META, Carl Zeiss, Jena, Germany). DAPI was visualized with a 364 nm laser line in combination with a 380-475 nm bandpass filter. Recordings were made with a 20x objective at zoom 4, with maximum pinhole. Fluorescence intensity was analysed with the LSM software (release 3.2). Data normalization with average fluorescence values of the tetraploid guard cell nuclei (4C) and arbitrary clustering of data points were performed as described elsewhere [67].

RNA in situ hybridization
Semi-thin paraffin sections of inflorescences, emasculated pistils, and siliques [68] were used for hybridization with the hydrolyzed digoxygenin-UTP-labeled riboprobes (Roche Diagnostics, Basel, Switzerland) that were prepared using a RBR cDNA expression clone as a template. In situ hybridization was performed as before [6].

Tetraploid genetic analysis of rbr-3
Tests for dominance/recessiveness were performed as previously described [69]. The expected phenotypic ratios for recessive and dominant genetic models were calculated considering (a) reduced transmission efficiency of the rbr-3 allele and (b) maximal double reduction. Transmission efficiency (TE) is an estimate of inheritance of a mutant allele versus the wild-type allele by female or male gametes [14]. It is calculated as a ratio of number of mutant plants to wild-type plants in progenies from reciprocal crosses of a heterozygous mutant. The rbr allele is not transmitted through female gametes (TE R(rbr) = 0) [14]; the transmission of rbr through pollen was estimated as 0.1 based on both TE R(rbr) in diploid condition, and recovery of triplex plants in triplex progeny (as a ratio of triplex to duplex plants). Double reduction describes the situation in polyploids, in which a heterozygous individual produces homozygous gametes [69]. This can occur if quadrivalents are formed and recombination occurs between the centromere and the locus of interest. Through chromatid segregation both alleles of the sister chromatids can co-exist in the same gamete. Thus, the frequency of double reduction depends on the distance between the locus in question and the centromere. Because the RBR locus is ,45 cM away from the centromere, these loci can be considered unlinked. Therefore, we used the maximal double reduction frequency of 1/6 for our calculations (see Text S1).
As we did not know how many rbr-3 alleles were present in the tetraploid plants, we compared the observed data to six different models with one, two, or three rbr-3 alleles, them being dominant and recessive, respectively. First we recorded the seed set/sterility phenotypes of a total of 103 progeny plants originating from a selfed autonomously tetraploidized plant, which was heterozygous for the rbr mutation ( Figure S2A, S2B). Out of the 103 progeny plants, we took one tetraploid plant group (consisting of 52 plants) with similar seed set and subjected the sterility phenotype and progeny segregation data for 6 different models [simplex, recessive (rbr/RBR/RBR/RBR); simplex, dominant (rbr D /RBR/RBR/RBR); duplex, recessive (rbr/rbr/RBR/RBR); duplex, dominant (rbr D / rbr D /RBR/RBR); triplex, recessive (rbr/rbr/rbr/RBR); triplex, dominant (rbr D /rbr D /rbr D /RBR)]. These plants were identified as duplex-recessive for rbr (Table 1). Subsequently, two other distinct tetraploid phenotype groups were fit to simplex-recessive and triplex-recessive models (41 and 2 plants, respectively) (see Figure  S2 and Table 1 for details). Progeny analysis of one of these two triplex plants identified in this experiment confirmed stability of the seed set phenotype over subsequent generation ( Figure S4).

Quantitative real-time RT-PCR
RNA extraction and reverse transcription were performed as described [15]. Quantitative real-time measurements were performed using SYBR Green Fast Master Mix reagent in an ABI Prism 7500 Sequence Detection System (Applied Biosystems) (Applied Biosystems), according to the manufacturer's instructions. For each condition, 2 technical replicates and 3 biological replicates were used. Relative gene expression levels were normalized to the expression levels of a control gene, PP2A (At1g13320) [70]. Primers used in this work are listed in Table S1. Figure S1 Deregulation of cell-specific markers in rbr female gametophyte. (A-D) the enhancer detector ET1119 (egg cell marker) GUS expression in wild-type and rbr female gametophytes. (A) A wild-type embryo sac at maturity showing a typical eggspecific GUS expression pattern of ET1119 at 2 days after emasculation (red arrow). Green arrow marks synergids. (B-D) Mis-expression of the egg-specific GUS marker in rbr mutant embryo sacs. In some small number of cases, the ectopic GUS staining was restricted to the egg (red arrow) and central cell region (black arrow) (B) (2 observations) or the whole FG (C, black arrow) (1 observation). (D) In this particular rbr embryo sac, two big cells in the place of an egg were stained (red arrows) (1 observation). Note that the synergids appeared morphologically normal, but they also showed GUS expression (green arrows). (E) Synergidspecific expression of ET2634 in the wild-type (green arrow). (F) Rare mis-expression phenotype of ET2634 in rbr embryo sac. Black arrow points to egg apparatus (mainly synergid-derived proliferation) with a weaker GUS. Scale bars: 30 mm.  Table 1 for details). (C) A second generation progeny test of an rbr triplex plant (n = 93) (see Figure S4 for additional data). Found at: doi:10.1371/journal.pgen.1000988.s002 (2.17 MB DOC) Figure S3 Comparison of female gametophyte phenotypes in rbr triplex versus the corresponding tetraploid wild-type confirms gametophytic recessiveness of the rbr-3 allele. (A) Histogram of female gametophyte (FG) phenotypes in diploid rbr plants (rbr/ RBR) in comparison to the corresponding wild-type (RBR/RBR).

Supporting Information
In the expected genetic model we considered that FGs homozygous for rbr hyper-proliferate and they are lethal, typical of the rbr gametophytic lethal mutation [15]. (B) Histogram of female gametophyte (FG) phenotypes in tetraploid triplex plants (rbr/rbr/rbr/RBR) in comparison to the tetraploid wild type (RBR/ RBR/RBR/RBR). The expected ratio of FG phenotypes in the triplex plant was calculated based on a genetic model for recessiveness (in our case full loss of function) considering double reduction (see Table 1). obs: observed FG phenotypes; exp: expected FG phenotypes; class FG7: mature 4-celled wild-type female gametophyte; class Proliferation: FGs with ectopic cell proliferation. Total counts for RBR/RBR and rbr/RBR ovules were 101 and 194, for RBR/RBR/RBR/RBR and rbr/rbr/rbr/RBR 108 and 162, respectively. Found at: doi:10.1371/journal.pgen.1000988.s003 (0.45 MB TIF) Figure S4 Quantitative reduction of RBR expression and concomitant reduction of the characteristic 4-branched trichomes in tetraploid leaves confirms RBR dosage-dependent trichome differentiation. Note that reduction in RBR levels ultimately correlated with reduction in seed set. Analyzed are four distinct genotypes, RBR/RBR/RBR/RBR (n = 178 and 415, for seed set and trichome counts, respectively); rbr/RBR/RBR/RBR (n = 259 and 387); rbr/rbr/RBR/RBR (n = 214 and 395); and rbr/rbr/rbr/ RBR (n = 217 and 432). *significantly different in the mutant when compared to the corresponding nulliplex (p = 0.05).