Australia, 3 School of Life Science and Technology, Xinjiang University, Urumqi , China, and SUMMARY

Transcription

1 The Plant Journal (2017) 90, doi: /tpj Mutants in the imprinted PICKLE RELATED 2 gene suppress seed abortion of fertilization independent seed class mutants and paternal excess interploidy crosses in Arabidopsis Fang Huang 1, Qian-hao Zhu 2, Anyu Zhu 2, Xiaoba Wu 2, Liqiong Xie 3, Xianjun Wu 1, Chris Helliwell 2, Abed Chaudhury 4, E. Jean Finnegan 2 and Ming Luo 2, * 1 Rice Research Institute, Sichuan Agricultural University, Wenjiang, Chengdu, Sichuan , China, 2 Commonwealth Scientific and Industrial Research Organization (CSIRO), Agriculture and Food, PO Box 1700, ACT 2601, Australia, 3 School of Life Science and Technology, Xinjiang University, Urumqi , China, and 4 HKG-Agro, Shamshernagare, Bangladesh Received 22 November 2016; revised 23 January 2017; accepted 25 January 2017; published online 3 February *For correspondence ( ming.luo@csiro.au). SUMMARY Endosperm cellularization is essential for embryo development and viable seed formation. Loss of function of the FERTILIZATION INDEPENDENT SEED (FIS) class Polycomb genes, which mediate trimethylation of histone H3 lysine27 (H3K27me3), as well as imbalanced contributions of parental genomes interrupt this process. The causes of the failure of cellularization are poorly understood. In this study we identified PICKLE RELATED 2 (PKR2) mutations which suppress seed abortion in fis1/mea by restoring endosperm cellularization. PKR2, a paternally expressed imprinted gene (PEG), encodes a CHD3 chromatin remodeler. PKR2 is specifically expressed in syncytial endosperm and its maternal copy is repressed by FIS1. Seed abortion in a paternal genome excess interploidy cross was also partly suppressed by pkr2. Simultaneous mutations in PKR2 and another PEG, ADMETOS (ADM), additively rescue the seed abortion in fis1 and in the interploidy cross, suggesting that PKR2 and ADM modulate endosperm cellularization independently and reproductive isolation between plants of different ploidy is established by imprinted genes. Genes upregulated in fis1 and downregulated in the presence of pkr2 are enriched in glycosyl-hydrolyzing activity, while genes downregulated in fis1 and upregulated in the presence of pkr2 are enriched with microtubule motor activity, consistent with the cellularization patterns in fis1 and the suppressor line. The antagonistic functions of FIS1 and PKR2 in modulating endosperm development are similar to those of PICKLE (PKL) and CURLY LEAF (CLF), which antagonistically regulate root meristem activity. Our results provide further insights into the function of imprinted genes in endosperm development and reproductive isolation. Keywords: endosperm, cellularization, reproductive isolation, interploidy cross, FIS, speciation. INTRODUCTION In flowering plants, seeds consist of three functionally different tissues: the maternally derived seed coat, the embryo and the endosperm. The embryo, which has a 1 maternal:1 paternal (1 m:1p) genome composition develops from a haploid egg fertilized by a sperm cell, while the triploid endosperm results from the homodiploid central cell fertilized with another haploid sperm cell, giving a 2 m:1p genome composition. Following fertilization, the endosperm progenitor nucleus undergoes several rounds of division without cytokinesis, resulting in a multinucleate syncytium which cellularizes and is later consumed by the developing embryo (Becker et al., 2014; Li and Berger, The Plant Journal 2017 John Wiley & Sons Ltd 2012). Cellularization is one of the key developmental processes that determines the viability and final size of seeds (Hehenberger et al., 2012; Kradolfer et al., 2013b; Rebernig et al., 2015; Scott et al., 1998; Wolff et al., 2015). Central cell initiation, syncytial endosperm proliferation and cellularization are modulated by histone H3K27me3 that is catalyzed in Arabidopsis by the FIS Polycomb Repressive Complex 2 (PRC2) (Chaudhury et al., 1997; Grossniklaus et al., 1998; K ohler et al., 2003; Luo et al., 2000; Ohad et al., 1996; Simon and Kingston, 2013). Loss of function of any of the FIS class genes causes autonomous formation of endosperm in the absence of 383

2 384 Fang Huang et al. fertilization and over-proliferation of uncellularized endosperm in fertilized fis class mutant seeds (Chaudhury et al., 1997; Grossniklaus et al., 1998; Kiyosue and Fischer, 1999; Ohad et al., 1996; Sørensen et al., 2001), suggesting that FIS genes repress proliferation and promote cellularization of endosperm. Among the FIS genes, FIS1/MEA and FIS2 are maternally expressed imprinted genes, expressed exclusively in the syncytial endosperm and central cell (Grossniklaus et al., 1998; Jullien et al., 2006; Kinoshita et al., 1999; Luo et al., 2000). The endosperm phenotype caused by the loss of the FIS function is similar to that observed in paternal excess interploidy hybridization (Dilkes et al., 2008; Erilova et al., 2009; Scott et al., 1998). Previous studies have suggested that altered endosperm cellularization in fis class seeds, or seeds with an unbalanced contribution of parental genomes from interploidy crosses, is associated with deregulation of imprinting (Haig and Westoby, 1991; Hsieh et al., 2011; Kradolfer et al., 2013a; Scott et al., 1998). Imprinting is the selective expression of genes from either the maternal or the paternal genome, which is dependent on differential epigenetic modifications established in the gametes (Gehring, 2013; K ohler et al., 2012). In flowering plants imprinting primarily occurs in the endosperm. The parental conflict theory predicts that maternally expressed imprinted genes (MEGs) should inhibit the growth of embryo-supporting tissues (such as endosperm) to distribute resources evenly across all progeny, while paternally expressed imprinted genes (PEGs) maximize the resources to individual embryos by promoting the growth of embryo-supporting tissues. This is consistent with the observation of contrasting endosperm phenotypes in reciprocal interploidy crosses (Haig and Westoby, 1989; Scott et al., 1998). Current epigenetic analyses in plants suggests that MEGs tend to be silenced by DNA methylation at the paternal alleles, while PEGs tend to be silenced by H3K27me3 at the maternal alleles (Pignatta and Gehring, 2012). Loss of H3K27me3 due to fis mutations activates the maternal copies of PEGs and is considered to cause seed abortion, which is similarly seen in paternal excess interploidy crosses in certain Arabidopsis ecotypes. Suppression of endosperm abortion in fis class mutants, or in paternal excess crosses, by a paternally hypomethylated genome is associated with the downregulation of PEGs due to de novo CHG methylation. Seed defects in paternal excess crosses and in fis1/mea can be suppressed by a mutation in a PEG, ADM (Kradolfer et al., 2013a). Recent assays of more PEGs in triploid seed identified an additional three genes with similar function to ADM (Wolff et al., 2015), supporting the idea that imprinted genes modulate endosperm cellularization and are involved in the establishment of post-zygotic hybridization barriers in interploidy crosses. It is unknown how/ whether the other imprinted genes regulate cellularization and contribute to post-zygotic reproductive isolation. The antagonistic activity of the PRC2 and CHD3 proteins has been observed in non-reproductive tissues (Aichinger et al., 2011, 2009). In Arabidopsis, the vegetative phenotype of loss of function of a PRC2 gene, CLF (a FIS1/MEA homolog), can be partially suppressed by mutation in PKL, a member of the CHD3 gene family. Whether there are similar antagonistic roles between the FIS PRC2 complex and CHD3 proteins in seed is unknown. We conducted a suppressor screen using a homozygous mutant of a mea allele, fis1, and isolated four independent mutants which gave increased formation of viable seed. To avoid confusion and reflect the history of the identification of fis1 and mea mutants, we use fis1 to describe the original fis1 mutant in Ler and mea to represent a fis1/mea allele isolated from a Syngenta Arabidopsis Insertion Library (SAIL) line in this study. The suppressor mutations mapped to a CHD3 chromatin remodeling family member gene PICKLE RELATED 2 (PKR2), a PEG specifically expressed in syncytial endosperm (Gentry and Hennig, 2014; Hsieh et al., 2011; Wolff et al., 2011). The suppressor mutation also suppressed the seed defect in a paternal excess interploidy cross, supporting the notion that imprinted genes are involved in establishing post-zygotic hybridization barriers (Haig and Westoby, 1991) and adding new evidence for the opposing roles of H3K27me3 and CHD3 in modulating endosperm development. RESULTS Isolation and identification of mutations that suppress fis1 seed defects Fertilization of the fis1 mutant in the Ler background with wild-type (WT) or fis1 pollen results in aborted seeds with arrested embryos and uncellularized endosperm, which mostly fail to germinate (Figure 1a, c, fi iii) (Chaudhury et al., 1997; Grossniklaus et al., 1998). Without fertilization, ovules carrying the fis1 mutation gave seed-like structures at low frequency with autonomous endosperm Figure 1. Phenotypes of suppressor line 1. (a) Upper panel: mature seeds of self-pollinated fis1. Scale bar = 0.5 mm. Lower panel: fis1 seeds 9 days after germination. Scale bar = 1 cm. (b) Upper panel: mature seeds of self-pollinated suppressor line 1. Scale bar = 0.5 mm. Lower panel: suppressor line 1 seeds 9 days after germination. Scale bar = 1 cm. (c) Percentage of plump seeds and germinated seeds in fis1 and suppressor line 1. Numbers above bars are the total seeds been tested. (d) (i) (iii) Sections of seeds of Ler at 5, 8 and 10 days after pollination (DAP). Scale bar = 100 lm. (e) (i) (iii) Sections of seeds of pkr2-2 at 5, 8 and 10 DAP. Scale bar = 100 lm. (f) (i) (iii) Sections of seeds of fis1 at 5, 8 and 10 DAP. Scale bar = 100 lm. (g) (i) (iii) Sections of seeds of fis1 pkr2-2 (suppressor line 1) at 5, 8 and 10 DAP. Scale bar = 100 lm.

3 Antagonistic roles of PRC2 and CHD3 genes in seed 385 (a) (b) (c) (di) (ei) (fi) (gi) (dii) (eii) (fii) (gii) (diii) (eiii) (fiii) (giii)

4 386 Fang Huang et al. development (Chaudhury et al., 1997). We isolated four independent fis1 suppressor lines, which gave approximately 50% WT-like seed capable of germination (Figure 1b, c and Figure S1a in the Supporting Information). Cross pollination between the four lines gave approximately 50% WT-like seeds, suggesting that the mutations in each line are allelic (Figure S1a). The fis1 homozygote was used in reciprocal crosses with suppressor line-3; each cross gave rise to approximately 10% plump seeds, comparable to the rate of plump seed set in the self-pollinated fis1, suggesting that the suppressor mutations are recessive (Figure S1a). To clone the suppressor gene, the genomes of 10 F 2 plants showing approximately 50% plump seeds and 11 F 2 plants showing <10% plump seeds, derived from the cross between fis1 and suppressor line-1, were sequenced using the Illumina platform. Four SNPs were identified between the two populations, with one occurring in the coding region of PKR2 (At4 g31900) and the other three in nongenic regions. Sanger sequencing of the other three suppressor lines revealed that each line had a mutation in PKR2 but at a different location (Figure S1b), confirming that mutations in PKR2 suppress the seed abortion of fis1 in the suppressor lines. A suppressor line pollinated with a pkr2-1 homozygote in a Col background (Aichinger et al., 2009) also gave approximately 30% plump seeds (Figure S1b, c), confirming that the pkr2 mutations are responsible for the seed rescue of fis1. We designated our suppressor mutations pkr2-2, -3, -4 and -5 respectively. Mutation in PKR2 suppresses fis seed abortion by promoting cellularization and embryo development To understand the effect of the suppressor mutation on seed development, we isolated pkr2-2 in a WT background by backcrossing the suppressor line-1 to Ler. Seed development of Ler, pkr2-2, fis1 and fis1 pkr2-2 (suppressor line- 1) was determined by sectioning seeds harvested at 5, 8 and 10 days after pollination (DAP; Figure 1d g, Methods S1, References S1). At 5 DAP the embryo reached the heart stage and the endosperm became cellularized in Ler and pkr2-2 (Figure 1di, ei). At 8 and 10 DAP, development of Ler and the pkr2-2 single mutant was similar, with the embryo at the bent cotyledon stage and the endosperm being mostly consumed (Figure 1dii iii, eii iii). Endosperm cellularization was not seen in the fis1 seeds and embryos were arrested at the torpedo stage (Figure 1fi iii), as previously observed (Chaudhury et al., 1997). No cellularization was observed in the fis1 pkr2-2 seeds, at 5 DAP (Figure 1gi). However, 40% of seeds (n = 20) showed cellularization and delayed development of the embryo at 8 DAP (Figure 1gii). At 10 DAP, 40% of seeds (n = 10) showed embryos at the early bent cotyledon stage accompanied by cellularized endosperm (Figure 1giii). The failure of endosperm cellularization is considered to be associated with embryo arrest and seed abortion in the fis1 mutant (Hehenberger et al., 2012), suggesting that the restoration of cellularization in fis1 pkr2-2 allows embryos to develop to maturity. We further tested whether the mutation in PKR2 also suppressed the seed defect in fis2, a mutant of one of the PRC2 FIS class genes (Chaudhury et al., 1997). The double homozygous mutant fis2 pkr2-2 gave about 30% viable seeds, suggesting that the pkr2 mutation also suppresses fis2 seed abortion (Figure S2a, b). PKR2 expression is restricted to the syncytial endosperm and overlaps with FIS1 expression To further understand PKR2 function in endosperm development, we made a PKR2::GFP protein fusion construct by cloning the open reading frame (ORF) of GFP to the 3 end of the PKR2 genomic sequence (Methods S1). Similarly, a FIS1::GFP protein fusion construct was made, allowing comparison of the expression pattern of these two genes. The fis1 homozygous plants transformed with the FIS1:: GFP transgene showed increased formation of viable seeds; one homozygous T 3 line showing complete seed rescue was used for further analysis. As expected, GFP expression was detected in the central cell and early syncytial endosperm (Figure S3c) (Luo et al., 2000). The suppressor line fis1 pkr2-2 was transformed with the PKR2::GFP transgene, generating 11 plants expressing GFP. All plants showed reduced formation of viable seeds (Figure S3a, b), indicating that the PKR2::GFP transgene restored PKR2 function and complemented the suppressor line phenotype. These lines showed syncytial endosperm expression of PKR2-GFP (Figure S3d). By crossing one representative GFP line to the pkr2-2 single mutant we isolated homozygous GFP lines with or without fis1 for detailed GFP characterization. Without fis1, there was no maternal PKR2::GFP expression in endosperm carrying the PKR2::GFP transgene pollinated with Ler (Figure 2ai v). In contrast, WT ovules fertilized with PKR2::GFP pollen showed nuclear-localized GFP in the syncytial endosperm starting from the two-nuclei stage and continuing until endosperm cellularization (Figure 2bi v), suggesting that PKR2 is a PEG (Hsieh et al., 2011; Wolff et al., 2011) and functions only in syncytial endosperm. There is maternal PKR2::GFP expression in the early endosperm in fis1 carrying the PKR2::GFP transgene; when pollinated by Ler, this GFP expression continued into later seed development stages probably due to the failure of endosperm cellularization in ovules carrying fis1 (Figure 2ci v). This suggests that the maternal allele of PKR2 is repressed by FIS PRC2, likely via H3K27me3. Consistent with this, PKR2 was identified as a PEG and is associated with H3K27me3 (Aichinger et al., 2009; Hsieh et al., 2011; Wolff et al., 2011).

5 Antagonistic roles of PRC2 and CHD3 genes in seed 387 Figure 2. Expression of PKR2::GFP. (a) (i) (v) PKR2::GFP pollinated with Ler showing no GFP activity in the syncytial endosperm. The maternal tissue surrounding the embryo sac shows autofluorescence (i iii) similar to fluorescence of Ler pollinated with the GFP line in the surrounding maternal tissue (see bi iii). Scale bar = 50 lm. (b) (i) (v) Ler pollinated with PKR2::GFP showing GFP activity in the syncytial endosperm (i iv), but showing no GFP at the late heart stage when endosperm is cellularized. Scale bar = 50 lm. (c) (i) (v) fis1 PKR2::GFP pollinated with Ler showing GFP in the syncytial endosperm. Note that endosperm does not cellularize due to the fis1 mutation. Scale bar = 50 lm. (ai) (bi) (ci) (aii) (bii) (cii) (aiii) (biii) (ciii) (aiv) (biv) (civ) (av) (bv) (cv)

6 388 Fang Huang et al. Figure 3. Mature seeds and germination of self-pollinated mea, pkr2-1 mea, adm mea and adm pkr2-1 mea in Col background. (a) Upper panels: mature seeds of mea, pkr2-1 mea, adm mea and adm pkr2-1 mea. Scale bar = 0.5 mm. Lower panels: 9-day germination of mea, pkr2-1 mea, adm mea and adm pkr2-1 mea. Scale bar 1 cm. (b) Percentage of plump seeds and germinated seeds of mea, pkr2-1 mea, adm mea and adm pkr2-1 mea. Numbers above the bars are the total number of seeds tested. The pkr2 and adm mutations act additively to suppress seed abortion in fis1 Mutation in another PEG, ADM, can also suppress seed abortion in fis1 (Kradolfer et al., 2013b). To investigate the genetic interaction between ADM and PKR2, we used a fis1 allele (termed mea in this section) isolated from a SAIL line (SAIL_55_B04; Col background). By crossing mea to adm and pkr2-1, all in the Col background, we produced the double mutants pkr2-1 mea and adm mea and the triple mutant adm pkr2-1 mea. The mea homozygote gave <1% plump seeds. However, the pkr2-1 mea double mutant produced about 40% plump seeds (Figure 3a, b), similar to the fis1 pkr2-2 suppressor line. The majority of plump seeds germinated and developed into viable seedlings. The adm mea mutant formed about 30% plump seeds in agreement with the previous study (Kradolfer et al., 2013b). The formation of plump seeds in the adm pkr2-1 mea triple mutant increased to approximately 77% (Figure 3a, b), suggesting that the mutations in ADM and PKR2 additively suppress seed abortion. We examined the effect of these mutations on autonomous seed development in mea, and found that single or double mutations had slightly repressed autonomous seed formation (Figure S4a, b). The adm single mutant has a slight reduction in seed weight but there was no further seed weight reduction in the adm pkr2-1 double mutant (Figure S4c, Methods S1). The pkr2 mutation promotes seed viability in paternal genome excess interploidy crosses Diploid Columbia (2n Col) pollinated with tetraploid Columbia (4n Col) results in a high frequency of aborted seed with endosperm that fails to cellularize, similar to the

7 Antagonistic roles of PRC2 and CHD3 genes in seed 389 defects seen in fis class seed (Scott et al., 1998; Tiwari et al., 2010). Mutations in ADM and several other PEGs suppress seed abortion in triploid seeds derived from osd1. In osd1 unreduced (diploid) pollen is formed, and hence seeds with triploid embryos and tetraploid endosperm are produced after pollination (Kradolfer et al., 2013b; Wolff et al., 2015). The pkr2 mutation appeared not to suppress such triploid seed abortion in osd1. Instead of using the osd1 mutant, we converted pkr2-1, adm and adm pkr2-1 in the Col background to tetraploid to test the roles of PEGs in interploidy crosses (Methods S1, Table S7). The 2n Col pollinated with the 4n Col gave approximately 3% plump seed (Figures 4a, b and S5), and a similar frequency of plump seed formation was observed if diploid pkr2-1 was pollinated with 4n Col. The 2n Col pollinated with pkr2-1 4n gave more plump seeds, consistent with PKR2 being a PEG. The cross between pkr2-1 2n pollinated with pkr2-1 4n achieved a slightly higher seed rescue, indicating that the maternal allele of PKR2 may not be completely silenced (Figures 4a, b and S5). Similar to previous findings (Kradolfer et al., 2013b), the maternally derived adm mutation showed only limited suppression of seed abortion in 2n 9 4n crosses in contrast to the paternally derived adm (Figures 4a, b and S5). Compared with adm, pkr2 has a reduced ability to suppress seed abortion, which may explain why pkr2 appeared not to suppress triploid seed abortion in osd1 as osd1 gives mixed diploid and triploid seeds (Kradolfer et al., 2013b; Wolff et al., 2015). By Figure 4. Paternal genome excess interploidy crosses involving in pkr2-1 4n, adm 4n, adm pkr2-1 4n and Col 4n as pollen donors, and pkr2-1 2n, adm 2n, adm pkr2-1 2n and Col 2n as pollen recipients. (a) Mature seeds of interploidy crosses. Genotypes involved are indicated. Scale bar = 0.5 mm; corresponding germination of these seeds is shown in Figure S5. (b) Percentage of plump seeds and germinated seeds of interploidy crosses. Numbers above bars are the total number of seeds tested.

8 390 Fang Huang et al. converting the adm pkr2 double mutant into 4n we found similar suppression to that when adm 4n was used to pollinate the 2n Col. However, additive suppression of seed abortion was observed when adm pkr2 2n was pollinated by the 4n double mutant (Figures 4a, b and S5). Transcriptome analysis shows that some genes deregulated in fis1 are normalized by pkr2 To test whether the rescue of fis1 seed abortion was reflected by a genome-wide normalization of gene expression in the suppressor line, we generated transcriptome data from each of WT Ler, pkr2-2, fis1 and fis1 pkr2-2 seed with three biological repeats at 3 and 6 DAP (Methods S1). A hierarchical clustering analysis showed that three biological repeats of each genotype harvested at the same time point are correctly grouped together, suggesting that the expression data are reliable (Figure S6). Differentially expressed genes were defined as those in the mutants that had either increased or decreased expression levels compared with the WT (Table S1). We identified 695 genes which were upregulated in fis1 at 3 DAP, 364 of which overlapped with genes upregulated in fis1 pkr2-2, suggesting that the remaining 331 genes (47%) upregulated in fis1 are downregulated in fis1 pkr2-2 (termed up-normalized; Figure 5a, Table S2). Similarly, 398 downregulated genes (58%) were also normalized (termed down-normalized) by introducing the pkr2-2 mutation into a fis1 background at 3 DAP (Figure 5b, Table S2). At 6 DAP, 819 upregulated genes (38%) in fis1 as well as 2511 downregulated genes (75%) in fis1 were also normalized in fis1 pkr2-2 (Figure 5a, b, Table S2). The transcriptome profile of the pkr2-2 single mutant is similar to that of the WT, with only a small number of genes being differentially expressed; these genes were enriched for a few Gene Ontology (GO) terms (Figure 5a, b, Table S3). The differentially expressed genes in pkr2-2 hardly overlapped with those in fis1 and fis1 pkr2-2 (Figure 5a, b). Considering that there are no obvious endosperm defects in pkr2-2, we reasoned that the genes differentially expressed in pkr2-2 might not be biologically relevant to the suppression role of the pkr2 mutation. There was little overlap in normalized genes between the two stages, suggesting that different genes are involved in different developmental stages (Figure S7a, Table S2). There were 673 (up) and 350 (down) genes deregulated in seed at 3 DAP, and 578 (up) and 122 (down) genes deregulated in the 6-DAP seeds that are unique to fis1 pkr2-2, (Figure 5a, b), suggesting that the fis1 and pkr2 double mutant deregulates genes which are not affected by either mutation alone. Alternatively, this may only reflect the fact that the rescue of embryos or endosperm in fis1 pkr2-2 is a Figure 5. Analysis of differentially expressed genes in fis1, pkr2-2, and fis1 pkr2-2 mutant seeds compared with Ler seeds. (a) Left panel: Venn diagram showing overlap of genes upregulated in fis1 (fold change 1.5, P < 0.05), upregulated in fis1 pkr2-2 (fold change 1.5, P < 0.05) and downregulated in pkr2-2 (fold change 1.5, P < 0.05) at 3 days after pollination (DAP). Right panel: Venn diagram showing overlap of genes upregulated in fis1 (fold change 1.5, P < 0.05), upregulated in fis1 pkr2-2 (fold change 1.5, P < 0.05) and downregulated in pkr2-2 (fold change 1.5, P < 0.05) at 6 DAP. (b) Left panel: Venn diagram showing overlap of genes downregulated in fis1 (fold change 1.5, P < 0.05), downregulated in fis1 pkr2-2 (fold change 1.5, P < 0.05) and upregulated in pkr2-2 (fold change 1.5, P < 0.05) at 3 DAP. Right panel: Venn diagram showing overlap of genes downregulated in fis1 (fold change 1.5, P < 0.05), downregulated in fis1 pkr2-2 (fold change 1.5, P < 0.05) and upregulated in pkr2-2 (fold change 1.5, P < 0.05) at 6 DAP.

9 Antagonistic roles of PRC2 and CHD3 genes in seed 391 continuum from seeds showing complete rescue to seeds being aborted, because fis1 and these seeds showing variable phenotypes displayed a complex transcriptome profile. Five GO terms were significantly enriched among the down-normalized genes at 3 DAP, while 13 were significantly enriched among the down-normalized genes at 6 DAP in the suppressor line (Table S4). The enrichment of genes in microtubule motor activity at 6 DAP correlates with the restoration of cellularization of the suppressor line, as the upregulation of microtubule motor activity could be beneficial for phragmoplast and cell wall formation in the endosperm (Table S4) (Lee and Liu, 2004; Lee et al., 2001). Among the up-normalized genes, only one GO term, hydrolase activity, was significantly enriched in fis1 pkr2-2 at 3 DAP. But 13 GO terms were significantly enriched at 6 DAP, with three terms related to glycosylhydrolyzing activity (Table S4). Glycosyl-hydrolyzing enzymes can degrade pectin and interrupt cell wall formation (Xiao et al., 2014). Thus, decreased activity of these genes is also coupled with cellularization in fis1 pkr2-2. Suppression of endosperm abortion in the fis class mutants, as well as in the paternal excess interploidy cross by the paternally hypomethylated genome, is associated with the downregulation of PEGs due to de novo CHG methylation (Schatlowski et al., 2014; Wolff et al., 2015). We examined 103 PEGs (Pignatta et al., 2014) in our normalized gene set and found that only 14 were up-normalized at 3 DAP, 6 DAP or both time points, including SUVH7 at 3 DAP, mutation of which suppresses abortion of triploid seed (Table S5) (Wolff et al., 2015). Sixteen PEGs were down-normalized at 3 DAP, 6 DAP or both time points (Table S5). This suggests that pkr2 rescue of fis1 may partly be associated with those PEGs, such as SUVH7, with decreased expression or with other unknown genes. Increased expression of AGAMOUS-LIKE MADS-box genes (AGLs) has been functionally associated with seed abortion in fis1 and interploidy crosses (Hehenberger et al., 2012; Kradolfer et al., 2013b; Rebernig et al., 2015; Wolff et al., 2015). We scrutinized 105 AGLs (Par enicova et al., 2003) in the normalized gene set and found that only six genes were up-normalized at 3 DAP, 6 DAP or both time points and eight were down-normalized at 3 DAP or 6 DAP (Table S5). Most of the AGLs which were normalized in adm, including AGL62 and PHE1 which have been functionally connected to seed abortion in the interploidy cross and the fis mutants, were not normalized in fis1 pkr2-2, suggesting that the pkr2 mutation rescues fis1 seed without normalizing the key AGLs. In Arabidopsis, the vegetative phenotypes due to the loss of function of a PRC2 gene, CLF (a FIS1 homolog), can be partially suppressed by mutation in a CHD3 family gene PKL (At2G25170), a homolog of PKR2 (Aichinger et al., 2011, 2009). The same study also showed that PKR2 acts redundantly with PKL in suppressing cell dedifferentiation in the root, and deregulated genes in the pkl pkr2 double mutants are enriched with H3K27me3. While we did not observe any role for PKL in seeds in a pkl mea double mutant or the pkl pkr2-1 mea triple mutants (Figure S8) we did find that the genes normalized by the pkr2 mutation were significantly enriched for genes marked by H3K27me3 in the endosperm (Figure S7b, c) (Moreno- Romero et al., 2016). DISCUSSION We report the identification of an imprinted gene PKR2, a CHD3 homolog, containing a chromo domain and a SNF2- related helicase/atpase domain (Gentry and Hennig, 2014; Ogas et al., 1999), which acts antagonistically to the PRC2 FIS genes in endosperm to mediate endosperm cellularization and seed viability. Loss of PKR2 function suppresses seed abortion in fis mutants and in paternal excess interploidy crosses by promoting endosperm cellularization. The pkr2 mutation also slightly suppressed autonomous endosperm formation in fis1. PKR2 was active in the early syncytial endosperm where the FIS genes are expressed. The pkr2 mutation also suppressed abortion of fis2 seed, suggesting that suppression is not specific to the fis1 mutation but likely to the loss of H3K27me3 due to the mutation in PRC2 members, which causes seed abortion. Previous results suggested that PKR2 is a PEG, which is associated with differential DNA methylation and H3K27me3 (Aichinger et al., 2009; Hsieh et al., 2011; Wolff et al., 2011). Using PKR2::GFP, we confirmed the imprinted status at this locus and further showed that the fis1 mutation caused the activation of maternal PKR2::GFP. The genetic analysis also suggests that loss of FIS1 function activates the maternal copy of PKR2, as suppression of fis1 seed abortion required homozygous pkr2 mutations (Figure S1a). The FIS PRC2 complex functions to repress proliferation of endosperm and promote its cellularization (Chaudhury et al., 1997). Thus, PKR2 acts in an antagonistic manner to promote endosperm proliferation and delay cellularization. Another PEG, ADM, is similarly regulated to PKR2 and the adm mutation can also suppress fis seed abortion (Kradolfer et al., 2013b), suggesting that the FIS PRC2 complex promotes endosperm cellularization and viable seed formation by repressing the maternal copies of these two PEGs. Consistent with this, PEGs are preferentially targeted by H3K27me (Pignatta and Gehring, 2012). Therefore, loss of FIS function causes over-expression of many PEGs by activating the maternal alleles of these genes, and this contributes to over-proliferation of endosperm and seed abortion. Similarly, in the 2n 9 4n paternal genome excess crosses, the abortive phenotype can also be suppressed by mutations in PKR2 and/or ADM (Figure 4; Kradolfer et al., 2013b). It is plausible that the additional copy of the paternal genome in 2n 9 4n endosperm

10 392 Fang Huang et al. and associated overexpression of multiple PEGs are responsible for the over-proliferation of endosperm and seed abortion. Suppression of endosperm abortion in fis class mutants by a paternally hypomethylated genome (Adams et al., 2000; Luo et al., 2000; Xiao et al., 2006) is associated with the downregulation of PEGs, including ADM and PKR2, due to de novo CHG methylation (Schatlowski et al., 2014). We did not observe a noticeable phenotype in the pkr2 single mutant, though in our study the seed size in adm is slightly reduced, suggesting that each imprinted gene may have only a minor effect on seed development in a WT background and stacking mutations or copies of multiple PEGs may result in a change in seed size. Alternatively, the function of these genes may only be readily observed in endosperm containing unbalanced parental genomes (caused by interploidy cross), or in endosperm of the fis class mutants in which imprinted genes are deregulated. The functional analysis of ADM in a previous study (Kradolfer et al., 2013b) and PKR2 in this study provides strong support for the parental conflict theory, predicting that PEGs function to promote growth of supporting tissues (such as endosperm in plant and placenta in mammals) that sustain embryo growth and MEGs (such as FIS1/MEA and FIS2) function to inhibit growth of supporting tissues. The antagonistic activity of other PRC2 and CHD3 proteins operates in vegetative tissues. In Arabidopsis, the vegetative phenotype of loss of function of a PRC2 gene CLF, a MEA/FIS1 homolog, can be partially suppressed by a mutation in PKL, a homolog of PKR2. Loss of function of PKL results in reduced expression of many PRC2 target genes, suggesting that PKL is required for transcriptional activity of PRC2 targets and that PKL functions in transcriptional activation by an as yet unknown mechanism (Aichinger et al., 2011, 2009). CHD3/CHD4 proteins in animals can function as transcriptional activators by associating with histone acetyltransferases or by other unknown mechanisms (Murawska et al., 2008; Scimone et al., 2010; Shimono et al., 2003; Williams et al., 2004). In addition, they have a repressive function in transcription when they are part of a nucleosome remodeling and deacetylase multiprotein complex which confers ATP-dependent chromatin remodeling and deacetylation (Bouazoune and Brehm, 2006). Upregulation of genes in the early endosperm of fis1 was presumably caused by loss of the repressive mark H3K27me3. The expression of some of these fis1-dependent upregulated genes is normalized in fis1 pkr2-2, suggesting that PKR2 is required for upregulation of these potential FIS1 target genes. Downregulation of other genes in fis1 is probably secondary in response to the failed PRC2 complex and is also normalized in fis1 pkr2-2. In this regard, PKR2 probably acts as a transcriptional activator of PRC2-targeted genes in endosperm, as does PKL in other tissues (Aichinger et al., 2011, 2009). Our results showed that the genes normalized in fis1 pkr2-2 were enriched with H3K27me3, suggesting that the opposing functions of FIS and PKR2 allow correct transcription programs for the balance of cell proliferation and differentiation in endosperm. Our results shed light on how PRC2 and CHD3 chromatin remodelers regulate development in a different tissue. Our transcriptome analysis shows that pathways related to degradation and formation of cell walls are enriched among the up- and down-normalized genes, respectively, in the suppressor line, which is consistent with the cellularization patterns observed in mutants. Genes related to motor and microtubule motor activity were enriched among the down-normalized genes in the suppressor line (Table S4). Some of these genes encode kinesins or kinesin-related proteins that are microtubulebased motor proteins. Microtubule motor activity is essential for the construction of a new cell wall in the phragmoplast (Murata et al., 2013), where formation of the cell plate during cytokinesis requires coordinated microtubule reorganization and vesicle transport (Lee and Liu, 2004; Lee et al., 2001). HINKEL is one kinesin-related gene that is normalized in fis1 pkr2-2. HINKEL localizes to the cell plates and loss of function of HINKEL causes defective cytokinesis in both endosperm and embryo (Strompen et al., 2002). The up-normalized genes were enriched with hydrolase activity and specifically in glycosyl-hydrolyzing activity. Glycosyl-hydrolyzing enzymes can degrade pectin, which is assumed to be a key step in the deconstruction of plant cell walls (Xiao et al., 2014). Increased expression of these genes in fis1 and normalized expression in the suppressor line correlates with the cellularization patterns in fis1 and fis1 pkr2-2. Therefore, reduced expression of genes that are potentially required to build the phragmoplast, together with increased expression of genes that degrade cell walls, correlates with the observed failure of cellularization in fis1 seeds. Similarly, the expression of genes with microtubule motor activity and glycosyl-hydrolyzing activity is similarly affected in triploid seeds (Rebernig et al., 2015). This suggests that deregulated expression of imprinted genes in fis1 and 3n seeds perturbs the same pathways in cell division and cell wall formation during endosperm development. The key AGLs involved in regulating endosperm cellularization are deregulated in adm but not affected in pkr2, suggesting that ADM and PKR2 may regulate endosperm cellularization via different mechanisms. This is consistent with the additive effects of the adm and pkr2 mutations in suppressing seed abortion in fis1 and the interploidy cross. The mechanism that drives the formation of new species is a fundamental question in evolutionary biology. Speciation refers to mechanisms that cause reproductive isolation between two groups of organisms. Polyploidization and

11 Antagonistic roles of PRC2 and CHD3 genes in seed 393 the resultant reproductive isolation between the polyploid and the ancestral diploid are a major path for sympatric speciation in plants (Otto and Whitton, 2000; Ramsey and Schemske, 1998). It has long been speculated that imprinting is involved in establishing reproductive isolation between plants of different ploidy levels (Haig and Westoby, 1989, 1991). Endosperm abortion in F 1 hybrids between related species can be overcome by changing the ploidy level of one parent, suggesting the involvement of imprinted genes in reproductive isolation between species (Johnston et al., 1980; Josefsson et al., 2006). The identification of PKR2 in this study, ADM and several other PEGs (Kradolfer et al., 2013b; Wolff et al., 2015), mutations of which suppress triploid seed abortion in a paternal excess interploidy cross, provides direct evidence about the function of imprinted genes in establishing reproductive isolation between diploid and tetraploid Arabidopsis plants. ADM is not conserved outside the Brassicaceae (Kradolfer et al., 2013b), but PKR2 is conserved in different species. The rice counterpart of PKR2 is an imprinted gene (Luo et al., 2011), raising the possibility that imprinted CHD3 genes may function as one of the speciation genes in a broader context. Post-hybridization barriers between related species may be effectively established via conserved and non-conserved imprinted genes. In this regard, these PEGs can be considered to be speciation genes, while the FIS class genes play an indirect role by regulating imprinting. Identification of the suppressor activity of PKR2 provides further insight into the function of imprinting and the molecular mechanism controlling reproductive isolation which leads to speciation (Burkart-Waco et al., 2015; Wolff et al., 2015). EXPERIMENTAL PROCEDURES Plant material and growth conditions Ler, fis1 (a mea mutant), fis2 and pi2 were used in this study (Luo et al., 1999). T-DNA insertion lines for mea (SAIL_55_B04), pkr2-1 (SALK_109423) and pkl (SAIL_276_F01) in the Col accession were obtained from the Arabidopsis Stock Centre ( arabidopsis.info/). The adm-2 mutant (Kradolfer et al., 2013b) was kindly provided by Claudia K ohler. Primers used to validate the T-DNA insertion lines are shown in Table S6. All plants were grown in growth chambers under 16-h light/8-h dark at 20 C. The artificial light was a fluorescent light source producing 120 lmol photons m 2 sec 1. Flowers were emasculated 2 days before pollination. Germination analysis Seeds were surface sterilized in a sealed container for 1 h using chlorine gas liberated by adding 3 ml of hydrochloric acid to 100 ml of commercial bleach. Seeds were plated on MS medium containing 1% sucrose. After 2 days at 4 C, the plates were transferred to a growth room with conditions of 22 C/18 C (day/night) and a 16-h light/8-h dark cycle. Germination frequency was determined after 9 days. Pictures were captured using Nikon COOLPIX ( and processed using Photoshop CC (Adobe, Ethyl methanesulfonate mutagenesis and gene mapping fis1 homozygous seeds were mutagenized with ethyl methanesulfonate (EMS) as described by Chaudhury et al. (1997). From an M 2 population consisting of about plants grown in 200 pots, four plants showing approximately 50% plump seeds were identified from four pots independently and were considered to be lines (designated suppressor lines 1 to 4) containing suppressors of fis1. To identify the suppressor gene of fis1, suppressor line 1 was crossed to the fis1 homozygote to establish a F 2 population, from which 10 plants showing >50% plump seeds and 11 plants showing <10% plump seeds were individually harvested. High-throughput sequencing libraries using DNA from the fis1 homozygote and the 21 individually harvested plants were generated using the Illumina NEBNext Ultra DNA Library Prep Kit ( following the manufacturer s protocol and sequenced on the Illumina HiSeq2000 platform ( After trimming of adaptor sequences and removing low-quality reads and using the FASTX-Toolkit ( we first aligned clean reads from the fis1 homozygote to the TAIR10 Arabidopsis reference genome using the CLC Genomics Workbench (version 6.0.4) to generate a consensus sequence of the fis1 homozygote. The parameter settings used were: mismatch cost, 2; insertion and deletion cost, 3; length fraction, 0.9; similarity fraction, We then aligned reads from the 10 plants showing the mutant phenotype to the consensus sequence of the fis1 homozygote using the aforementioned parameter settings and ignoring non-specifically matched reads; the mapping result was used to identify single nucleotide polymorphism (SNP) between the mutant and the fis1 homozygote using the CLC Genomics Workbench with the default settings. Raw and processed reads were deposited in the Gene Expression Omnibus database (accession no. SUB ). ACKNOWLEDGEMENTS We thank Cathy Miller from the Australian National Herbarium for providing flow cytometry analysis, Rosemary White and Mark Talbot for help with microscopy, Kerrie Ramm and Neil Smith for help with experiments and Elizabeth S. Dennis and W. James Peacock for critical reading of the manuscript. Chinese Academy of Sciences-Commonwealth Scientific and Industrial Research Organisation Fund Grant GJHZ1122 (to M.L.). CONFLICTS OF INTEREST The authors declare no conflicts of interest. SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article. Figure S1. Genetic analysis of suppressor lines and cloning of the suppressor genes. Figure S2. The effect of pkr2 on the fis2 seed abortion. Figure S3. Complementation of fis1 pkr2-2 with PKR2::GFP and expression of PKR2::GFP and FIS1::GFP. Figure S4. pkr2 and adm slightly suppress mea autonomous endosperm formation. Figure S5. Germination of various interploidy crosses. Figure S6. Dendrogram showing the clusters of expression profiles of the normalized genes generated by DESeq for all the samples.

12 394 Fang Huang et al. Figure S7. Overlapping normalized genes in fis1 pkr2-2 between 3 DAP and 6 DAP stages and with H3K27 m3 targeted genes in endosperm. Figure S8. Mature seeds and germination of pkl mea and pkl pkr2-1 mea. Table S1. All gene reads and differential gene list. Table S2. List of genes deregulated in fis1 but normalized in fis1 pkr2-2 and their annotated functions. Table S3. Enriched Gene Ontology terms of deregulated genes in pkr2-2. Table S4. Enriched Gene Ontology terms of the normalized genes in fis1 pkr2-2. Table S5. Normalized AGAMOUS-LIKE MADS-box genes and paternally expressed imprinted genes in fis1 pkr2-2. Table S6. Primers used in this study. Table S7. Chloroplast numbers in guard cells of candidate triploid lines. Methods S1. Generation of plasmids and transgenic lines, microscopy, RNA sequencing and high-throughput RNA sequence analysis, generation of tetraploid lines, and seed weight analysis. References S1. Literature citations used in the Supporting Information. REFERENCES Adams, S., Vinkenoog, R., Spielman, M., Dickinson, H.G. and Scott, R.J. (2000) Parent-of-origin effects on seed development in Arabidopsis thaliana require DNA methylation. Development, 127, Aichinger, E., Villar, C.B., Farrona, S., Reyes, J.C., Hennig, L. and K ohler, C. (2009) CHD3 proteins and polycomb group proteins antagonistically determine cell identity in Arabidopsis. PLoS Genet. 5, e Aichinger, E., Villar, C.B., Di Mambro, R., Sabatini, S. and K ohler, C. (2011) The CHD3 chromatin remodeler PICKLE and polycomb group proteins antagonistically regulate meristem activity in the Arabidopsis root. Plant Cell, 23, Becker, M.G., Hsu, S.-W., Harada, J.J. and Belmonte, M.F. (2014) Genomic dissection of the seed. Front Plant Sci. 5, 464. Bouazoune, K. and Brehm, A. (2006) ATP-dependent chromatin remodeling complexes in Drosophila. Chromosome Res. 14, Burkart-Waco, D., Ngo, K., Lieberman, M. and Comai, L. (2015) Perturbation of parentally biased gene expression during interspecific hybridization. PLoS ONE, 10, e Chaudhury, A.M., Ming, L., Miller, C., Craig, S., Dennis, E.S. and Peacock, W.J. (1997) Fertilization-independent seed development in Arabidopsis thaliana. Proc. Natl Acad. Sci. 94, Dilkes, B.P., Spielman, M., Weizbauer, R., Watson, B., Burkart-Waco, D., Scott, R.J. and Comai, L. (2008) The maternally expressed WRKY transcription factor TTG2 controls lethality in interploidy crosses of Arabidopsis. PLoS Biol. 6, e308. Erilova, A., Brownfield, L., Exner, V., Rosa, M., Twell, D., Mittelsten Scheid, O., Hennig, L. and Kohler, C. (2009) Imprinting of the polycomb group gene MEDEA serves as a ploidy sensor in Arabidopsis. PLoS Genet. 5, e Gehring, M. (2013) Genomic imprinting: insights from plants. Genetics, 47, 187. Gentry, M. and Hennig, L. (2014) Remodelling chromatin to shape development of plants. Exp. Cell Res. 321, Grossniklaus, U., Vielle-Calzada, J.-P., Hoeppner, M.A. and Gagliano, W.B. (1998) Maternal control of embryogenesis by MEDEA, a polycomb group gene in Arabidopsis. Science, 280, Haig, D. and Westoby, M. (1989) Parent-specific gene expression and the triploid endosperm. Am. Nat. 134, Haig, D. and Westoby, M. (1991) Genomic imprinting in endosperm: its effect on seed development in crosses between species, and between different ploidies of the same species, and its implications for the evolution of apomixis. Philos. Trans. R. Soc. Lond. B Biol. Sci. 333, Hehenberger, E., Kradolfer, D. and K ohler, C. (2012) Endosperm cellularization defines an important developmental transition for embryo development. Development, 139, Hsieh, T.F., Shin, J., Uzawa, R. et al. (2011) Regulation of imprinted gene expression in Arabidopsis endosperm. Proc. Natl Acad. Sci. USA, 108, Johnston, S.A., Nijs, T.P.M.D., Peloquin, S.J. and Hanneman, R.E. (1980) The significance of genic balance to endosperm development in interspecific crosses. Theor. Appl. Genet. 57, 5 9. Josefsson, C., Dilkes, B. and Comai, L. (2006) Parent-dependent loss of gene silencing during interspecies hybridization. Curr. Biol. 16, Jullien, P.E., Kinoshita, T., Ohad, N. and Berger, F. (2006) Maintenance of DNA methylation during the Arabidopsis life cycle is essential for parental imprinting. Plant Cell, 18, Kinoshita, T., Yadegari, R., Harada, J.J., Goldberg, R.B. and Fischer, R.L. (1999) Imprinting of the MEDEA polycomb gene in the Arabidopsis endosperm. Plant Cell, 11, Kiyosue, T. and Fischer, R.L. (1999) Control of fertilization-independent endosperm development by the MEDEA polycomb gene in Arabidopsis. Proc. Natl Acad. Sci. USA, 96, K ohler, C., Hennig, L., Bouveret, R., Gheyselinck, J., Grossniklaus, U. and Gruissem, W. (2003) Arabidopsis MSI1 is a component of the MEA/FIE Polycomb group complex and required for seed development. EMBO J. 22, K ohler, C., Wolff, P. and Spillane, C. (2012) Epigenetic mechanisms underlying genomic imprinting in plants. Annu. Rev. Plant Biol. 63, Kradolfer, D., Hennig, L. and K ohler, C. (2013a) Increased maternal genome dosage bypasses the requirement of the FIS polycomb repressive complex 2 in Arabidopsis seed development. PLoS Genet. 9, e Kradolfer, D., Wolff, P., Jiang, H., Siretskiy, A. and K ohler, C. (2013b) An imprinted gene underlies postzygotic reproductive isolation in Arabidopsis thaliana. Dev. Cell, 26, Lee, Y.-R.J. and Liu, B. (2004) Cytoskeletal motors in Arabidopsis. Sixty-one kinesins and seventeen myosins. Plant Physiol. 136, Lee, Y.R.J., Giang, H.M. and Liu, B. (2001) A novel plant kinesin-related protein specifically associates with the phragmoplast organelles. Plant Cell, 13, Li, J. and Berger, F. (2012) Endosperm: food for humankind and fodder for scientific discoveries. New Phytol. 195, Luo, M., Bilodeau, P., Koltunow, A., Dennis, E.S., Peacock, W.J. and Chaudhury, A.M. (1999) Genes controlling fertilization-independent seed development in Arabidopsis thaliana. Proc. Natl Acad. Sci. 96, Luo, M., Bilodeau, P., Dennis, E.S., Peacock, W.J. and Chaudhury, A. (2000) Expression and parent-of-origin effects for FIS2, MEA, and FIE in the endosperm and embryo of developing Arabidopsis seeds. Proc. Natl Acad. Sci. 97, Luo, M., Taylor, J.M., Spriggs, A., Zhang, H., Wu, X., Russell, S., Singh, M. and Koltunow, A. (2011) A genome-wide survey of imprinted genes in rice seeds reveals imprinting primarily occurs in the endosperm. PLoS Genet. 7, e Moreno-Romero, J., Jiang, H., Santos-Gonzalez, J. and Kohler, C. (2016) Parental epigenetic asymmetry of PRC2-mediated histone modifications in the Arabidopsis endosperm. EMBO J. 35, Murata, T., Sano, T., Sasabe, M., Nonaka, S., Higashiyama, T., Hasezawa, S., Machida, Y. and Hasebe, M. (2013) Mechanism of microtubule array expansion in the cytokinetic phragmoplast. Nat. Commun. 4, doi: /ncomms2967. Murawska, M., Kunert, N., van Vugt, J., L angst, G., Kremmer, E., Logie, C. and Brehm, A. (2008) dchd3, a novel ATP-dependent chromatin remodeler associated with sites of active transcription. Mol. Cell. Biol. 28, Ogas, J., Kaufmann, S., Henderson, J. and Somerville, C. (1999) PICKLE is a CHD3 chromatin-remodeling factor that regulates the transition from embryonic to vegetative development in Arabidopsis. Proc. Natl Acad. Sci. 96, Ohad, N., Margossian, L., Hsu, Y.-C., Williams, C., Repetti, P. and Fischer, R.L. (1996) A mutation that allows endosperm development without fertilization. Proc. Natl Acad. Sci. 93, Otto, S.P. and Whitton, J. (2000) Polyploid incidence and evolution. Annu. Rev. Genet. 34, Par enicova, L., de Folter, S., Kieffer, M. et al. (2003) Molecular and phylogenetic analyses of the complete MADS-Box transcription factor family