Control of Paternally Expressed Imprinted UPWARD CURLY LEAF1, a Gene Encoding an F-Box Protein That Regulates CURLY LEAF Polycomb Protein, in the Arabidopsis Endosperm

Genomic imprinting, an epigenetic process in mammals and flowering plants, refers to the differential expression of alleles of the same genes in a parent-of-origin-specific manner. In Arabidopsis, imprinting occurs primarily in the endosperm, which nourishes the developing embryo. Recent high-throughput sequencing analyses revealed that more than 200 loci are imprinted in Arabidopsis; however, only a few of these imprinted genes and their imprinting mechanisms have been examined in detail. Whereas most imprinted loci characterized to date are maternally expressed imprinted genes (MEGs), PHERES1 (PHE1) and ADMETOS (ADM) are paternally expressed imprinted genes (PEGs). Here, we report that UPWARD CURLY LEAF1 (UCL1), a gene encoding an E3 ligase that degrades the CURLY LEAF (CLF) polycomb protein, is a PEG. After fertilization, paternally inherited UCL1 is expressed in the endosperm, but not in the embryo. The expression pattern of a β-glucuronidase (GUS) reporter gene driven by the UCL1 promoter suggests that the imprinting control region (ICR) of UCL1 is adjacent to a transposable element in the UCL1 5′-upstream region. Polycomb Repressive Complex 2 (PRC2) silences the maternal UCL1 allele in the central cell prior to fertilization and in the endosperm after fertilization. The UCL1 imprinting pattern was not affected in paternal PRC2 mutants. We found unexpectedly that the maternal UCL1 allele is reactivated in the endosperm of Arabidopsis lines with mutations in cytosine DNA METHYLTRANSFERASE 1 (MET1) or the DNA glycosylase DEMETER (DME), which antagonistically regulate CpG methylation of DNA. By contrast, maternal UCL1 silencing was not altered in mutants with defects in non-CpG methylation. Thus, silencing of the maternal UCL1 allele is regulated by both MET1 and DME as well as by PRC2, suggesting that divergent mechanisms for the regulation of PEGs evolved in Arabidopsis.


Introduction
Arabidopsis seeds contain three tissues that have distinct parental genome contributions, namely 1) the diploid embryo, which is the diploid fertilization product of the maternal and paternal genomes, 2) the triploid endosperm, which is the fertilization product of the diploid maternal and haploid paternal genomes, and 3) the seed coat, which is of diploid maternal origin [1,2].Although communication and interaction between these tissues are critical for proper seed development, the underlying mechanisms are largely unknown [3][4][5].The unequal parental genetic contribution affects seed development due to genome dosage and parent-oforigin effects.
The parent-of-origin-dependent differential allelic expression of a single gene is known as genomic imprinting.Thus, imprinted genes are predominantly expressed from either the maternal or paternal allele.Genomic imprinting occurs primarily in mammals and flowering plants.In Arabidopsis, imprinting takes place mainly in the endosperm, the tissue that nourishes the developing embryo [6].Several theories have been proposed to explain the evolution of imprinting, the most popular of which is that imprinting arose due to parental conflict over resource allocation to the embryo [7][8][9].Another theory for the evolution of imprinting is that it is required to limit the gene dosage of key genes during early development [10,11].The genomic imbalance between maternal and paternal dosages affects seed and embryo development in both plants and mammals.An increase in paternal dosage leads to an increase in placental or endosperm growth, whereas an increase in maternal dosage has the opposite effect [12,13].
MEDEA (MEA), the first imprinted gene to be reported in Arabidopsis, was described more than a decade ago [14].Recently, thanks to next generation sequencing of expression libraries or RNAs at the whole genome level, more than 200 loci were found to be imprinted in Arabidopsis [15][16][17][18].However, the mechanisms by which differential allelic expression is regulated have been studied for only a few imprinted genes.The expression and silencing of MEA, FER-TILIZATION INDEPENDENT SEED2 (FIS2), FLOWERING WAGENINGEN (FWA), PHERES1 (PHE1), and ADMETOS (ADM) have been characterized [14,[19][20][21].While MEA, FIS2, and FWA are maternally expressed imprinted genes (MEGs), PHE1 and ADM are paternally expressed imprinted genes (PEGs).The maternal alleles of MEA, FIS2, and FWA are activated in the central cell of the female gametophyte.Their activation requires sequential steps involving two antagonistic genes; DNA METHYLTRANSFERASE 1 (MET1), which adds a methyl group to a cytosine base, and DEMETER (DME), which functions as a demethylase.During megagametogenesis, the transcription of MET1 is down-regulated by RETINOBLAS-TOMA-RELATED 1 (RBR1) and its binding partner MULTICOPY SUPPRESSOR OF IRA1 (MSI1), resulting in partial passive hypomethylation [22].Then, DME is expressed in the central cell of the mature female gametophyte [23], where it removes residual methyl cytosine from its target genes [24,25].Thus, the maternal alleles of MEA, FIS2, and FWA are expressed in the central cell before fertilization.After fertilization, the maternal alleles are epigenetically maintained in the hypomethylated state and are continuously expressed in the endosperm.Methylated cytosines can be directly removed by DME and REPRESSOR OF SILENCING 1 (ROS1) and the resulting abasic sites are replaced with unmethylated cytosines through the base excision repair (BER) pathway [24,26].Thus, activation of the maternal alleles of MEA, FIS2, and FWA is controlled by DNA methylation.Accordingly, loss-of-function mutations in DME result in at least partial hypermethylation and silencing of the maternal alleles of MEA, FIS2, and FWA.On the other hand, silenced paternal FIS2 and FWA alleles are derepressed when met1 mutants are inherited paternally, indicating that silencing of the paternal FIS2 and FWA alleles is controlled by MET1 [20,21].By contrast, the paternal MEA allele is silenced not by DNA methylation, but by the MEA-containing FIS-Polycomb Repressive Complex 2 (FIS-PRC2), and is thus self-regulated [24,27,28].In the case of the PEG PHE1, repression of the maternal allele is controlled by FIS-PRC2 and requires the unmethylated 3 0 -repeat region of the PHE1 locus [19,29].This repeat region located distantly downstream of PHE1 is hypermethylated in the expressed paternal allele.Another PEG, ADM, which belongs to the diverse family of molecular chaperones called J-domain proteins and determines seed viability in paternal excess interploidy hybridizations, was identified as a target gene of the FIS-PRC2 in the endosperm [30].
Although imprinting evolved independently in mammals and in flowering plants [31], the imprinting mechanisms in the Arabidopsis endosperm and in the mammalian placenta or embryo are partially parallel.In mammals, several imprinted genes are particularly important for placental development [32] and most of the imprinted genes are located in clusters in the imprinting control regions (ICRs), which are enriched in CpG islands and subjected to methylation [33].Whereas imprinted genes are organized in large chromosomal clusters in mammals, imprinted plant genes appear to occur as singletons [31].Recent efforts to identify the DNA sequences responsible for imprinted expression in Arabidopsis revealed that plant ICRs are located close to the imprinted loci [19,29].
Previously, we reported that UPWARD CURLY LEAF1 (UCL1), which encodes an E3 ligase that regulates CURLY LEAF (CLF) protein in Arabidopsis, is expressed exclusively in the endosperm [34].Here, we investigated whether the expression of UCL1 is regulated by imprinting.
To answer this question, we examined the allele-specific expression of endogenous UCL1 as well as that of UCL1 reporter transgenes.We found that the ICR of UCL1 was adjacent to the transposable element in the 5 0 -upsteam region of UCL1.FIS-PRC2 is required for the repression of the maternal allele of UCL1.In addition, repression of the maternal UCL1 allele is associated with DNA methylation near the ICR.Mutations in both MET1 and DME caused derepression of the silenced maternal UCL1 allele.These results provide new insight into the epigenetic mechanisms that maintain imprinting of UCL1 in the Arabidopsis endosperm.

UCL1 is a paternally expressed imprinted gene in the Arabidopsis endosperm
We previously reported that UCL1 is substantially expressed in flowers, young stamens, and developing seeds [34].Cytoplasmic GUS activity driven by the UCL1 promoter was detected in young stamens of floral stages 9 to 10 and then decreased significantly before fertilization.Strong GUS activity was observed in the endosperm after fertilization [34].To investigate whether UCL1 expression shows parent-of-origin specificity, we performed reciprocal crosses between UCL1_4.1k::GUS(transcriptional fusion) transgenic plants and Col-0 wild type plants (Fig. 1A-1C).GUS activity was not detected in the female gametophyte of the transgenic plants before fertilization (Fig. 1A) or in the developing seeds of UCL1_4.1k::GUStransgenic plants pollinated by the Col-0 wild-type plants (Fig. 1B).In the seeds of wild-type plants pollinated by the transgenic plants, however, cytoplasmic GUS expression was detected in the endosperm (Fig. 1C).
Next, we reciprocally crossed UCL1_4.1k::UCL1:GUS(translational fusion) transgenic plants with Col-0 wild-type plants (Fig. 1D-1F).GUS activity was detected neither in the female gametophyte (Fig. 1D) nor in the developing endosperm of the transgenic plants pollinated by wild-type plants (Fig. 1E).In contrast, when wild-type plants were pollinated by UCL1_4.1k::UCL1:GUStransgenic plants, GUS activity was detected in proliferating endosperm nuclei, but not in the embryo (Fig. 1F).Thus, these observations indicate that only the paternal allele of the UCL1 transgene is expressed in the developing endosperm.
To test whether endogenous UCL1 expression also depends on its parent-of-origin, single nucleotide polymorphisms (SNPs) among different Arabidopsis ecotypes were used to differentiate parent-specific UCL1 transcripts.While Col-0, Ler, and En-2 ecotypes have identical sequences within the UCL1 coding region, RLD and C24 ecotypes have polymorphic sites (S1 Fig) .RT-PCR analysis was performed using RNAs extracted from the developing seeds resulting from reciprocal crosses between Col-0 and RLD.The sequencing chromatogram of the amplification products showed a single peak of the SNP corresponding to the paternal allele (Fig. 1G).Taken together, these results indicate that UCL1 is expressed only from the paternally inherited allele and thus that UCL1 is a paternally expressed imprinted gene (PEG) in the endosperm.
The 5 0 -upstream region controls UCL1 imprinting According to the TAIR annotation, At1g65750, which is located upstream of the UCL1 (At1g65740) locus, encodes a non-LTR retrotransposon (LINE).Using the RepeatMasker program (http://www.repeatmasker.org/cgi-bin/WEBRepeatMasker),we identified two ATLINE1_1 transposable elements (TEs) in the Col-0 ecotype, namely 1) a long one (blue box) between 2.7 and 5.2 kb upstream and 2) a short one (red box) 1.5 and 2.0 kb upstream of the UCL1 translation start site (Fig. 2A).Interestingly, while the Col-0 and En-2 ecotypes contained the two ATLINE1_1 TEs, Ler, RLD, and C24 possess only the short ATLINE1_1 TE, which is closer to the UCL1 coding region (Fig. 2A).In addition, simple repeat sequences (Fig. 2A, green bar) were predicted at the 1.0-kb upstream region using the RepeatMasker program.
Because the transcriptional and translational fusions of the GUS transgenes driven by the UCL1 promoter recapitulated the imprinted expression of the endogenous UCL1 gene, we analyzed GUS expression using different lengths of the UCL1 promoter to identify the region necessary for UCL1 imprinting.To examine the activity of a UCL1 promoter fragment lacking the two ATLINE1_1 TEs, the 1.5-kb fragment upstream of UCL1 was transcriptionally fused to GUS (UCL1_1.5k::GUS)(S2D and S2L  We also generated transgenic plants carrying the UCL1_2.7k::UCL1:GUStranslational fusion.Consistent with the transcriptional UCL1_2.7k::GUSdata, nuclear GUS activity was detected in the endosperm only when the transgene was inherited paternally (Fig. 2D and S3G and S3O Fig).Taken together, these results suggest that the ICR of UCL1 is located in the region between 2.7 kb and 1.5 kb upstream of the UCL1 translation start site and that this region is necessary for the repression of the UCL1 maternal allele in the central cell before fertilization and in the endosperm after fertilization.
TE sequences are thought to be highly methylated due to the silencing of the invading foreign DNA [35][36][37].One short ATLINE1_1 TE is located in the 2.0-kb upstream region of the UCL1 locus (Fig. 2A, red box); thus, it is possible that this short ATLINE1_1 TE might be the target of methylation and function as the ICR of UCL1.To test this possibility, we generated constructs containing various lengths of the UCL1 promoter, corresponding to 2.0 kb, 1.9 kb, 1.7 kb, and 1.0 kb from the translation start site, fused to GUS and examined GUS staining in developing seeds of the corresponding transgenic plants after crossing with wild-type pollen.Surprisingly, all transgenic plants showed bi-allelic expression of the GUS transgenes (Figs. 3  and S4).These results clearly demonstrate that the ICR that underlies the maternal repression of UCL1 is located in the 5 0 -upstream region of this gene, between 2.7 kb and 2.0 kb from the translation start site, but that the cis-element(s) responsible for default bi-allelic expression of UCL1 is contained in the 1.0-kb upstream sequence.

PRC2 controls the silencing of maternal UCL1
FIS-PRC2, containing the four core polycomb group proteins MEA, FIS2, FIE, and MSI1, regulates not only seed development, but also genomic imprinting in Arabidopsis [19,20,24,28].Because UCL1 is a paternally expressed and maternally silenced imprinted gene in the endosperm after fertilization, we tested whether FIS-PRC2 is involved in maternal UCL1 repression.Firstly, we analyzed UCL1 expression in mea-3 homozygous mutant seeds.Real-time quantitative reverse transcription-polymerase chain reaction (qRT-RCR) revealed a strong increase in expression in four regions of UCL1 in the mea-3 homozygous mutant seeds (Fig. 4A and Fig. 4B).This suggests that the silenced maternal UCL1 allele might be de-repressed in mea-3 seeds.
Secondly, using RT-PCR and cleaved amplified polymorphic sequence (CAPS) markers of different ecotypes, we analyzed the expression of the maternally and paternally derived UCL1 alleles.Consistent with the transgene data, maternal UCL1 was silenced and paternal UCL1 was expressed in the developing seeds.However, when we used mea-3 homozygous plants or fie-1 heterozygous plants pollinated by RLD wild-type plants, the maternal UCL1 allele was activated, resulting in bi-allelic expression in the developing seeds (Fig. 4C).This finding shows that MEA and FIE are indeed required for the repression of the UCL1 maternal allele.To confirm the allele-specific expression of endogenous UCL1, we verified the sequence chromatogram of distinguished SNPs between Ler and RLD.The sequencing chromatogram showed that pollination of the maternally inherited mea or fie mutant with RLD pollen caused activation of the maternal UCL1 allele, which was silenced in Ler wild-type plants (Fig. 4D).Interestingly, not only the Col-0 and En-2 ecotypes, which carry the 5.2-kb upstream sequences containing the two TEs, but also the Ler, RLD, and C24 ecotypes, which have only the 2.7-kb upstream region containing the short TE, show imprinting (Fig. 1G, Fig. 2A, Fig. 4C  and 4D).
Thirdly, we examined the expression of the maternal UCL1_4.1k::GUS and UCL1_4.1k::UCL1:GUS transgenes in the wild-type and mea-3 background.To determine whether the paternal UCL1 expression pattern is affected by mutation of MEA, we pollinated wild-type stigmas with pollen derived from plants hemizygous for the GUS transgenes and heterozygous for the mea-3 mutation.No differences were observed in the seeds, indicating that the absence of MEA in the paternally derived genome does not affect UCL1 imprinting in the endosperm (Fig. 5B compared to 5A and 5F compared to 5E).Furthermore, inheritance of the fie-1 mutant through the male parent did not alter the endogenous UCL1 imprinting pattern (S5 Fig) .Conversely, when wild-type pollen was used to pollinate plants hemizygous for the GUS transgenes and heterozygous for mea-3, the maternally inherited UCL1_4.1k::GUS and UCL1_4.1k::UCL1:GUStransgenes were de-repressed, suggesting that MEA is required for silencing of the maternally derived UCL1 allele in the endosperm (Fig. 5D compared to 5C and 5H compared to 5G).
The initiation of endosperm development before fertilization is repressed by the FIS-PRC2 complex [38].Among the polycomb group proteins constituting this complex, the fie mutant showed a stronger phenotype of diploid central cell proliferation, resulting in a higher percentage of autonomous endosperm development in the silique when fertilization was blocked [39,40].To elucidate whether maternal UCL1 expression in the central cell of the female gametophyte is repressed by FIE, plants that were heterozygous for the fie-1 mutation and hemizygous for the UCL1_4.1k::GUSor UCL1_4.1k::UCL1:GUStransgene were emasculated and GUS expression was examined in the ovules.Whereas no GUS expression was detected in the female gametophyte either one-or two-days after emasculation when the transgenes were in the wildtype background, some female gametophytes of the fie-1 heterozygous mutants showed GUS signals after emasculation.Cytoplasmic GUS was detected in the central cell and in the autonomous endosperm after emasculation in the plants heterozygous for the fie-1 mutation and hemizygous for the UCL1_4.1k::GUStransgene (Fig. 5I and 5J).Likewise, the GUS signal was detected not only in the nucleus of the diploid central cell, but also in the nuclei of the dividing central cells of the autonomous endosperm of plants heterozygous for fie-1 and hemizygous for the UCL1_4.1k::UCL1:GUStransgene when emasculated (Fig. 5K to 5L).
These results demonstrate that functional MEA and FIE which are components of a FIS-PRC2 complex is responsible for the repression of maternal UCL1 expression in the central cell of the female gametophyte prior to fertilization and in the endosperm after fertilization.

DNA methylation affects UCL1 imprinting
DNA methylation and histone methylation are important epigenetic mechanisms regulating genomic imprinting in animals and plants [20,23,41].While the maternally expressed imprinted FWA and FIS2 genes have a differentially methylated region (DMR) in their promoter regions, the paternally expressed imprinted PHE1 gene has a DMR in the 3 0 -downstream region.To test whether UCL1 imprinting is regulated by DNA methylation, we emasculated plants that were heterozygous for the dme-2 mutation and hemizygous for the UCL1_4.1k::GUS or UCL1_4.1k::UCL1:GUStransgene.Plants hemizygous for the UCL1_4.1k::GUSor UCL1_4.1k::UCL1:GUStransgene in the wild-type background were used as a negative control (Fig. 6A, 6E, and 6I) and plants heterozygous for the fie-1 mutation and hemizygous for the UCL1_4.1k::GUSor UCL1_4.1k::UCL1:GUStransgene were used as a positive control (Fig. 6B, 6F, and 6J).After emasculation, maternal GUS activity was detected in the central cell of the dme-2 female gametophyte (Fig. 6C).After pollination with wild-type pollen, maternal GUS activity was detected in the endosperm of the dme-2 mutant (Fig. 6G and 6K).The maternal UCL1 allele was silenced by FIS2-PRC2 (Figs. 4 and 5).Given that DME is required for the activation of the MEA and FIS2 maternal alleles [20,23,24], the maternal UCL1 expression in the dme-2 mutant might be due to the lack of FIS-PRC2 in the central cell and the endosperm rather than to the hypermethylation of the endosperm DNA.
We also investigated UCL1 imprinting in the met1-6 mutant [42], which exhibits global hypomethylation of the genome [43].We emasculated plants heterozygous for the met1-6 mutation and hemizygous for the UCL1_4.1k::GUSor UCL1_4.1k::UCL1:GUStransgene, and examined the resulting GUS activity in ovules.Maternal GUS expression was detected in the central cell of the met1-6 mutant female gametophyte (Fig. 6D).We also performed crosses of plants hemizygous for the GUS transgene and heterozygous for met1-6 as females with wildtype pollen.The maternally inherited UCL1_4.1k::GUS and UCL1_4.1k::UCL1:GUStransgenes in the maternal met1-6 mutants were de-repressed in the endosperm (Fig. 6H and 6L).By contrast, we could not detect any GUS signal in crosses using plants that were heterozygous for argonaute 4 (ago4-1), rna-dependent rna polymerase 2 (rdr2-1), or dicer-like 3 (dcl3-1), which are involved in asymmetric methylation through RNA-dependent DNA methylation (RdDM) (S6 Fig).Taken together, these results imply that the repression of the UCL1 maternal allele is not related to the asymmetric RdDM pathway, but is related to symmetric CpG DNA methylation.A possible mechanism whereby DNA methylation mediates UCL1 imprinting will be discussed below.
To further our understanding of UCL1 methylation patterns, we analyzed publicly available CpG methylation data in wild-type and dme-2 endosperms [36].CpG methylation is significantly lower in the wild-type endosperm than in the wild-type embryo or dme-2 endosperm, indicating that the UCL1 promoter is demethylated by DME in the central cell (Fig. 7 and S1 Table ; p<0.05 in both wild-type endosperm-dme endosperm, and wild-type endospermwild-type embryo comparisons).Hypomethylation of the UCL1 promoter is more significant around the short ATLINE1_1 TE, and extends into the ICR region, which is located between 2.0 kb and 2.7 kb upstream of the translation start site of UCL1.The CpG methylation profile of the UCL1 promoter is consistent with the observation that DME activity is required for silencing of the UCL1 maternal allele, presumably by allowing FIS-PRC2 to bind to and establish silencing in the ICR.By contrast, no significant differences in non-CpG methylation were found between the embryo and endosperm, or between the wild-type and dme endosperm in the UCL1 promoter, supporting the observation that mutations in the RdDM pathway do not affect UCL1 imprinting.

Discussion
The imprinting control region of UCL1 is located in the 5 0 upstream region GUS transgene expression and endogenous UCL1 expression as analyzed using allele-specific SNPs indicate that UCL1 is an imprinted gene in the endosperm after fertilization (Fig. 1).Whereas the paternal UCL1 allele is expressed in the endosperm, the maternal allele is silenced.To identify the imprinting control region (ICR) of UCL1, we first analyzed the flanking sequences near the UCL1 coding region.Although TAIR annotation suggests that At1g65750 has a single non-LTR retrotransposon (LINE), which is located upstream of UCL1 (At1g65740), At1g65750 is likely composed of two ATLINE1_1 transposable elements (TEs), according to the RepeatMasker program.Interestingly, the number of ATLINE1_1 TEs is not identical in different ecotypes.While both ATLINE1_1 TEs are present in the Col-0 and En-2 ecotypes, only the short ATLINE1_1 TE, which is closer to the UCL1 coding region, is in the Ler, RLD, and C24 ecotypes (Fig. 2).However, UCL1 still showed imprinting in the Ler and RLD ecotypes (Fig. 4), suggesting that the absence of the long ATLINE1_1 TE in the further upstream region does not affect UCL1 imprinting and that the ICR is located within the region 2.7 kb upstream of the translation start site of UCL1.Deletion analysis of the UCL1 promoter confirmed that the ICR of UCL1 is located between 2.0 kb and 2.7 kb upstream of UCL1 (Fig. 3).
Imprinted genes are often associated with natural parasitic elements such as TEs and tandem repeats [29,44].Imprinting has been proposed to arise as a byproduct of silencing of invading foreign DNA [36,45].TEs sometimes have negative effects on genome integrity; however, they also provide a source for genetic and epigenetic diversity during evolution.TEs have been implicated as targets of methylation by various mechanisms [46][47][48].DNA methylation silences these potentially damaging DNA elements and also the neighboring genes.Imprinted genes in Arabidopsis are frequently in close proximity to differentially methylated regions (DMRs), most of which correspond to short TEs [6,16,35,36].Regulation of gene expression by DNA methylation could be selected for if imprinting of these genes mediates the parental conflict or gene dosage balance in the triploid endosperm [49].Because the short ATLINE1_1 TE is present in the UCL1 upstream region and CpG residues of the ATLINE1-1 TE are indeed methylated in the embryo (Fig. 7) [36], the short ATLINE1-1 TE is a likely UCL1 ICR candidate.However, the GUS transgene driven by the region 2.0 kb upstream of UCL1 showed bi-allelic expression in the presence of the short ATLINE1-1 TE.Therefore, the ICR of UCL1 is considered to be located between 2.0 kb and 2.7 kb upstream of the translation start site of UCL1.Consistently, the repeat sequences in the 1.0-kb upstream region did not affect UCL1 imprinting, indicating that the repeat sequences are not the ICR of UCL1 (Fig. 3).
ICRs and DMRs have been identified in plant imprinted genes [31].The DMR of PHE1 is located in the 3 0 downstream region that contains triple repeats [29].The MEA ICR was reported to be located in the 5 0 upstream region [50].The SINE-related sequence located in the 5 0 upstream region of FWA was identified as being sufficient for imprinting [44,51].Although the MEA ICR was reported to be independent of DME and MET1, other ICR-like sequences are controlled by DNA methylation.Nonetheless, the effects of DNA methylation on ICR-like sequences are different for each imprinted gene.Lack of sequence homology in the plant ICRlike sequences suggests that divergent imprinting mechanisms exist in plants, and that these mechanisms require sequence-specific imprinting factors.
PRC2 and DNA methylation control UCL1 imprinting MEA, FIS2, FIE, and MSI1 form the seed-specific FIS-PRC2 and are required for proper seed development.MEA and FIS2 are maternally expressed imprinted genes in the endosperm [14,20].Whereas FIS2 imprinting is solely regulated by DNA methylation, MEA imprinting is more complex; maternal MEA is activated by DME DNA glycosyalse and paternal MEA is silenced by maternally-expressed MEA-containing PRC2 [20,24].Thus, maternal MEA activation is established by DME-mediated DNA demethylation in the central cell and, in turn, the maternally expressed MEA-containing PRC2 silences its own paternal MEA allele.In contrast, UCL1 is a paternally expressed imprinted gene in the endosperm.The expression of the maternal UCL1 allele was de-repressed in the mea and fie mutants (Figs. 4 and 5), demonstrating that maternal UCL1 is silenced by FIS-PRC2.
DNA methylation is involved in silencing of many imprinted genes, including FWA [52].Maternal FWA is demethylated and activated by DME, and paternal FWA is hypermethylated and repressed in the endosperm [21].A short interspersed nuclear element (SINE)-related sequence in the FWA upstream region is sufficient for FWA imprinting and shows a differential methylation pattern depending on the parental origin [21,44,51].When the wild-type stigma is pollinated with hypomethylated mutant pollen, such as met1 or ddm1, the silenced paternal FWA allele is derepressed in the endosperm.By contrast, the repressed PHE1 maternal allele is hypomethylated and the expressed PHE1 paternal allele is hypermethylated at its 3 0 repeat region in the endosperm [29].DNA methylation is necessary for the paternal expression of PHE1.Therefore, DNA methylation of the imprinted genes can produce opposite outcomes, i.e., activation or repression.It has been suggested that FIS-PRC2 preferentially binds to hypomethylated regions of DNA and then silences nearby genes [29,37].While FIS-PRC2 binds to and represses the hypomethylated maternal PHE1 allele, the active paternal PHE1 allele is hypermethylated in the differentially methylated 3 0 repeat region, and this blocks FIS-PRC2 binding.
Genome-wide DNA methylation profiles in the endosperm revealed that TEs and repetitive sequences are hypomethylated in the endosperm as compared to the embryo [35].Methylome data revealed that the UCL1 promoter region containing the short ATLINE1_1 TE has much lower levels of CpG methylation in the endosperm than in the embryo, and that hypomethylation in the endosperm is DME-dependent, indicating that the UCL1 promoter is demethylated by DME in the central cell prior to fertilization (Fig. 7).The UCL1 maternal allele is derepressed in the dme mutant female gametophyte and endosperm, suggesting that proper demethylation is required for establishing maternal UCL1 silencing.Silencing and maintenance of repressed maternal UCL1 also depend on FIS-PRC2, which is consistent with the notion that demethylation of tandem repeats downstream of maternal PHE1 allows binding of FIS-PRC2 and subsequent silencing [15,29,37].
Although DNA demethylation is a global phenomenon, only selected sequences are targeted by FIS-PRC2, suggesting that DNA demethylation is necessary, but not sufficient for targeting of FIS-PRC2 [53].The ICR of UCL1 is located further upstream of the short ATLINE1_1 TE, whereas CpG hypomethylation occurs at the AT-rich region and at the short ATLINE1_1 TE (Fig. 7).Therefore, the short ATLINE1_1 TE is probably demethylated by DME, but the sequence 700 bp upstream of ATLINE1_1 TE may be required for targeting of FIS-PRC2, functioning as a polycomb response element, and spreading H3K27me3 for stable repression.This hypothesis is supported by the observation that the 2.0-kb promoter fragment containing the short TE drives biallelic expression of the transgene (Fig. 3).Future work is needed to further define the ICR and identify its associated epigenetic marks, such as H3K27me3.
Unexpectedly, the maternal UCL1 allele was also activated when a met1 mutant was crossed with wild-type pollen.This can possibly be explained by titration of PRC2 binding; met1 mutants cause global hypomethylation in the genome and if a certain or fixed amount of FIS-PRC2 is available, FIS-PRC2 can move to the newly exposed FIS-PRC2-binding sites in met1 mutants.Thus, the previously silenced UCL1 maternal allele is activated.This hypothesis can be tested by comparing the results of chromosome immunoprecipitation (ChIP) assays using PRC2 antibody or H3K27me3 antibody in met1 or the wild type.Another possibility is that a cryptic promoter causes de-repression in the met1 mutant.Interestingly, the maternal PHE1 allele is not reactivated in the endosperm when the met1 mutant was crossed with wild-type pollen.Thus, PHE1 and UCL1 imprinting is individually fine-tuned, although both genes are PEGs that are regulated by DMRs and FIS-PRC2 binding.
Why the paternal UCL1 allele is not expressed in the mature male gametophyte [34] remains unclear.One possible explanation is that the endosperm-specific activator (or transcription factor) that is absent in pollen induces paternal UCL1 expression in the endosperm.Alternatively, the unknown repressor in the mature pollen grain that is absent in the endosperm may inhibit the expression of the paternal UCL1 allele in the male gametophyte.
Control of paternal UCL1 imprinting is distinct from that of PHE1 Although many PEGs have been identified by genome-wide approaches [15,16], PHE1 is the only one that has been thoroughly examined in Arabidopsis [29,54].PHE1 imprinting depends on a distantly located region downstream of the PHE1 locus.PHE1 expression depends on the presence of methylation of this downstream region, whereas PHE1 repression is associated with the absence of methylation at this region [54].PHE1 imprinting is controlled by direct tandem repeats in the downstream region [54].On the other hand, FIS-PRC2 binding to the PHE1 promoter region and DNA demethylation of the 3 0 region of PHE1 are both necessary and sufficient for stable maternal PHE1 repression [54].However, although maternal UCL1 repression requires DME-mediated demethylation and binding of FIS-PRC2, expression of the paternal UCL1 allele does not seem to be regulated by DNA methylation, as is PHE1.Furthermore, the ICR located in the 700-bp region of the UCL1 promoter appears not to be involved in paternal UCL1 expression, because the fragment 1 kb upstream of the translation start site of UCL1 confers biallelic expression of UCL1 in the endosperm (Fig. 3 and S4 Fig) .While the maternal PHE1 allele was not reactivated in mutants that are defective in DNA methylation [29], the maternal UCL1 allele was de-repressed in the met1-6 mutant (Fig. 6H, 6L, and 6M).Therefore, our study provides insights into the divergent imprinting mechanisms that arose during the evolution of flowering plants.

Plant materials and growth conditions
All plants used in this study were Arabidopsis thaliana in the Columbia (Col-0) ecotype, except for the mea-3 [55] and fie-1 [40] mutants, which were isolated in the Ler ecotype.The met1-6 mutant [42] used in this study was in the Col-gl ecotype and only the first met1-6 homozygous mutants generated from a self-pollinated met1-6 heterozygote that had never been homozygous were used.Plants were grown in Sunshine Mix 5 under long-day (16 h/8 h) conditions at 23°C.Col-0 ecotype plants were used for Agrobacterium-mediated transformation by the floral dip method [56].

Histochemical GUS assay
The histochemical GUS assay was performed as previously described [34].Briefly, the samples were fixed in 90% acetone and then soaked in staining solution containing 1 mg/ml X-GlcA (5-bromo-4-chloro-3-indolyl glucuronide) in 50 mM Na 2 HPO 4 (pH 7.2), 5 mM potassium ferricyanide/ferrocyanide, and 0.1% Triton X-100 at 37°C overnight after washing in the same staining solution without X-GlcA.The next day, the staining buffer was removed, and the samples were mounted in clearing solution (2.5 g chloral hydrate, 0.7 ml H 2 O, 0.3 ml glycerol) or 1×PBS for microscopy.

Microscopy
The mounted tissue samples were observed on a Zeiss Axio Imager A1 microscope with a DIC filter and photographed using an AxioCamHRc camera (Carl Zeiss).

RT-PCR and quantitative real time qRT-PCR
Total RNA was extracted from tissue ground in liquid nitrogen using the RNeasy Mini Kit (Qiagen) and the RNase-free DNase Kit (Qiagen), according to the manufacturers' instructions, and messenger RNA was extracted from ground tissue with liquid nitrogen using the Dynabeads mRNA DIRECT TM Kit (Invitrogen DYNAL), according to the manufacturer's instructions.Following DNase-I treatment, 2 μg of total RNA or the entire amount of mRNA isolated from each sample was reverse transcribed into cDNA using oligo(dT) primer (18 mer) and the RevertAid TM First Strand cDNA Synthesis Kit (Fermentas).Real time qRT-PCR was performed as previously described [34].One tenth of the final volume of the reverse transcription (RT) product was used for each PCR reaction.Expression of gene transcripts was quantitated by iQ5 (Bio-Rad) and data were analyzed using the iCycler iQ system software (Bio-Rad).All PCR mixtures contained 10 μl of iQ SYBR Green Supermix (Bio-Rad), 0.5 μl of forward and reverse primers (10 μM), and 5 μl of 50 times diluted RT product per well.Samples were normalized against β2 tubulin or actin levels.

Sequencing of At1g65750 and At1g65760 in different Arabidopsis ecotypes
To amplify At1g65750 and At1g65760 in different Arabidopsis ecotypes, the primer sets JCW637/638 and JCW639/640 (S2 Table ) were used.Agarose gel electrophoresis revealed that a band of 4.6 kb was amplified from the genomic DNA of Col-0 and En-2.By contrast, the

Fig 1 .
Fig 1. UCL1 is paternally expressed in the endosperm.(A-C) Ovule and seeds derived from reciprocal crosses between the UCL1_4.1k::GUStransgenic plant and Col-0 wild type.(A) Expression of the maternally derived UCL1_4.1k::GUStransgene in a wild-type ovule after emasculation.B) Expression of the maternally derived UCL1_4.1k::GUStransgene in a wild-type seed at 1 day after pollination (DAP).(C) Expression of the paternally derived UCL1_4.1k::GUStransgene in the wild-type seed at 1 DAP.(D-F) Ovules and seeds resulting from reciprocal crosses between a UCL1_4.1k::UCL1:GUStransgenic plant and the wild type.(D) Expression of the maternally derived UCL1_4.1k::UCL1:GUStransgene in a wild-type ovule after emasculation.(E) Expression of the maternally derived UCL1_4.1k::UCL1:GUStransgene in a wild-type seed at 1 DAP.(F) Expression of the paternally derived UCL1_4.1k::GUStransgene in a wild-type seed at 1 DAP.Scale bars: 20 μm.(G) Sequencing chromatograms of RT-PCR products of UCL1 showing allelespecific expression at a polymorphic site indicated in S1 Fig. Endosperm RNA was prepared in samples derived from reciprocal crosses between Col-0 and RLD ecotypes.doi:10.1371/journal.pone.0117431.g001 Fig) and the transgenic plants were reciprocally crossed with Col-0 wild-type plants.Cytoplasmic GUS activity was detected in the central cell of the female gametophyte before fertilization in the UCL1_1.5k::GUStransgenic plants (S2D Fig), but not in the female gametophyte of the UCL1_4.1k::GUSplants (Fig. 1A and S2B Fig).After fertilization, the cytoplasmic GUS signal was detected not only in the self-fertilized seeds of UCL1_1.5k::GUStransgenic plants (S2L Fig) but also in the seeds of wild-type plants pollinated

Fig 2 .
Fig 2. Structure of the UCL1 locus in different Arabidopsis ecotypes.(A) Overview of the UCL1 locus in different Arabidopsis ecotypes.The blue and red boxes indicate distinct TEs in the ATLINE1_1 family of the LINE/L1 superfamily in At1065750.Ler, RLD, and C24 do not include the long ATLINE1_1 TE, whereas Col-0 and En-2 do.The numbers are in base pairs (bp) from the translation start site of UCL1.(B) Expression of the maternally derived UCL1_2.7k::GUStransgene in a wild-type seed at 1 DAP.(C) Expression of the maternally derived UCL1_1.5k::GUStransgene in a wild-type seed at 1 DAP.(D) Expression of the maternally derived UCL1_2.7k::UCL1:GUStransgene in a wild-type seed at 1 DAP.(E) Expression of the maternally derived UCL1_1.5k::UCL1:GUStransgene in a wild-type seed at 1 DAP.Scale bars: 50 μm.doi:10.1371/journal.pone.0117431.g002

Fig 3 .
Fig 3.The ICR of UCL1 is located between two LINE/L1 TEs.In the diagram of the promoter region of UCL1, the blue and red boxes indicate the distinct TEs in the ATLINE1_1 family of the LINE/L1 superfamily in At1065750.The numbers to the left of the lines indicate the size of the promoters (in bp) fused to the GUS transgene.Transgenic plants carrying the GUS transgene fused to various lengths of the UCL1 promoter were generated and the expression of the maternally derived transgenes was analyzed in seeds at 1 DAP, similarly as in Fig. 2. doi:10.1371/journal.pone.0117431.g003

Fig 4 .
Fig 4. MEA proteins are required for the repression of the maternal allele of UCL1.(A) Location of qRT-PCR primer sets used to detect the expression of UCL1.(B) Comparison of the expression levels of UCL1 in wild-type Ler and mea-3 endosperm at 3 DAP.The expression of UCL1 in the mea-3/mea-3 mutant was set to 1 and the error bar represents the standard deviation of three independent samples.(C) Analysis of the allele-specific expression of UCL1 using a CAPS marker.RT-PCR analysis was performed on RNA isolated from the endosperms of RLD females crossed with Ler males, Ler females crossed with RLD males, and mea-3 or fie-1 females (Ler background) crossed with RLD males.These products were digested with EcoRI.The Ler allele shows a 276 bp band, whereas the RLD allele was cut into a 222-bp band after EcoRI digestion.(D) Sequencing chromatograms of the RT-PCR products of UCL1 at the distinguished SNP regions showing allele-specific expression.RNAs were isolated from endosperms resulting from reciprocal crosses between RLD and Ler ecotypes and in crosses between the female mea-3 or fie-1 mutant and the male RLD.doi:10.1371/journal.pone.0117431.g004

Fig 7 .
Fig 7. CpG methylation patterns of the 5 0 upstream region of UCL1 in the endosperm and embryo.Only CpG sites with fractional CpG methylation that is significantly different between the embryo and endosperm, and between the wild-type endosperm and dme-2 endosperm are shown.Numbers on the xaxis represent CpG site positions (in bp) relative to the UCL1 translational start site.doi:10.1371/journal.pone.0117431.g007