Significance
The Arabidopsis DEMETER (DME) DNA demethylase is required for reproduction and endosperm gene imprinting. We investigated mechanisms that restrict DME transcription to the female and male companion cells, the central and vegetative cells, adjacent to egg and sperm cells, respectively. We delineated a region, surprisingly within the DME gene, sufficient for proper DME expression, which rescues seed abortion and aberrant endosperm DNA methylation associated with dme mutations. We discovered overlapping DNA enhancers promoting vegetative and central cell expression, which can be used to pinpoint candidate transcription factors that regulate DME in companion cells. These results show how reproductive DNA demethylation is restricted to companion cells, thereby protecting egg and sperm cells from deleterious DNA demethylation that would be transmitted to progeny.
Keywords: DNA demethylation, central cell, vegetative cell, cell-specific transcription, DNA enhancer elements
Abstract
The DEMETER (DME) DNA glycosylase initiates active DNA demethylation via the base-excision repair pathway and is vital for reproduction in Arabidopsis thaliana. DME-mediated DNA demethylation is preferentially targeted to small, AT-rich, and nucleosome-depleted euchromatic transposable elements, influencing expression of adjacent genes and leading to imprinting in the endosperm. In the female gametophyte, DME expression and subsequent genome-wide DNA demethylation are confined to the companion cell of the egg, the central cell. Here, we show that, in the male gametophyte, DME expression is limited to the companion cell of sperm, the vegetative cell, and to a narrow window of time: immediately after separation of the companion cell lineage from the germline. We define transcriptional regulatory elements of DME using reporter genes, showing that a small region, which surprisingly lies within the DME gene, controls its expression in male and female companion cells. DME expression from this minimal promoter is sufficient to rescue seed abortion and the aberrant DNA methylome associated with the null dme-2 mutation. Within this minimal promoter, we found short, conserved enhancer sequences necessary for the transcriptional activities of DME and combined predicted binding motifs with published transcription factor binding coordinates to produce a list of candidate upstream pathway members in the genetic circuitry controlling DNA demethylation in gamete companion cells. These data show how DNA demethylation is regulated to facilitate endosperm gene imprinting and potential transgenerational epigenetic regulation, without subjecting the germline to potentially deleterious transposable element demethylation.
Sexual reproduction is characterized by fertilization of an egg by a sperm cell, generating the embryo. Uniquely in angiosperms, a second sperm cell fertilizes the companion cell of the egg, the central cell, to generate the endosperm, which supports development of the embryo. During reproduction in angiosperm Arabidopsis thaliana, the DEMETER (DME) DNA glycosylase exhibits a striking expression pattern. Within the ovule, the female gametophyte is generated by mitosis of the haploid megaspore, forming a mature gametophyte of seven cells. During this process, the egg and central cell lineages are separated, and, at this point, DME expression and DNA demethylation is initiated solely in the central cell (1, 2). DME expression is switched off after fertilization (2). This precise pattern of expression initiated in the central cell, and not in the egg cell, is responsible for hypomethylation specifically in the maternal endosperm genome and not in the maternal embryo genome (3). DME expression in the central cell is essential for plant reproduction and genomic imprinting, whereby its absence results in loss of genomic imprinting, aberrant endosperm development, and early seed abortion (2, 4, 5).
In the male gametophyte, indirect evidence suggests that DME is expressed during development of the mature three-cell pollen grain, perhaps originating specifically in the vegetative cell, the companion cell of the two sperm cells (6). During reproduction, the vegetative cell generates a pollen tube that transports two sperm cells to the ovule for double fertilization. Although paternal inheritance of a DME mutation is compatible with normal seed development, it does result in decreased pollen viability and germination rates in certain ecotypes (6, 7).
DME is required to demethylate regions of DNA as part of the base-excision repair (BER) pathway. The dual activity helix-hairpin-helix glycosylase family consists of DME, REPRESSOR OF SILENCING1 (ROS1), and DEMETER-LIKE (DML) 2 and 3. Each glycosylase enzyme acts to remove 5-methylcytosine and nick the DNA backbone, followed by repair and replacement with cytosine by downstream enzymes in the BER pathway (4, 8–10). Within the glycosylase family of DNA demethylating enzymes, DME is distinguished by its highly restricted pattern of expression in gamete companion cells, as well as its profound effects on plant reproduction. The consequence of silencing the maternal DME allele is in the aberrant retention of DNA methylation on the maternal endosperm genome, including the imprinting control regions of imprinted genes (3, 4). Notably, maternal expression of MEDEA (MEA) and Fertilization Independent Seed 2 (FIS2), which form part of the floral polycomb repressive complex 2 (PRC2), involved in chromatin organization and regulation, requires DME action. Without DME-mediated DNA demethylation, the expression of these genes is lost, resulting in a loss of PRC2 and subsequent seed abortion.
DME also has a second function, which potentially impacts plant DNA methylation transgenerationally. DME-mediated DNA demethylation in companion cells is preferentially targeted to small, AT-rich, and nucleosome-depleted euchromatic transposable elements (TEs) (3). Evidence suggests that TE hypomethylation in the companion cells promotes transcription of mobile siRNA at the TEs, mediating RNA-directed DNA methylation (RdDM) in the gametes, so that the same TE sequences become hypermethylated, safeguarding the genomic integrity of the gametes (3, 11, 12). The large overlap between sites demethylated in the central cell, inferred from hypomethylated sites in the maternal endosperm genome (13) and sites demethylated in the vegetative cell, despite their different cell fates, provides evidence toward this common basal function of DME expression in gamete companion cells.
Both for the appropriate expression of imprinted genes during seed development, and for the putative role of DME in transgenerational epigenetic regulation, it is vital that DME expression is confined to the companion cells of the gametes, and not in the gametes themselves. We therefore sought to delineate the mechanisms affording this important expression profile.
Results
DME Is Expressed Specifically in the Companion Cell of the Male Gametophyte After Separation of the Sperm Cell Lineage.
During pollen development, a haploid microspore undergoes an asymmetric mitosis to produce a bicellular pollen with a generative cell engulfed in the vegetative cell. A second mitosis of the generative cell generates two sperm cells (Fig. 1 A and B). Previously, a low level of DME transcripts had been detected in mature pollen grains but not in sperm nuclei whereas DME-mediated DNA demethylation was shown to be restricted to the vegetative cell, implicating the vegetative cell as the site of DME expression (6). However, the precise pattern of DME expression during male gametophyte development is unknown. To address this issue, we measured β-glucuronidase (GUS) and green fluorescent protein (GFP) reporter expression in pollen from plants bearing the previously described 2.3pDME::GUS/GFP transgene. The 2.3pDME::GUS/GFP construct has 2.3 kb of upstream sequence and 2 kb of the DME transcriptional unit fused to GUS or GFP and is expressed in the central cell of the female gametophyte (Fig. S1A) (2, 14). GUS or GFP reporter expression was detected only in the vegetative cell nucleus of late bicellular pollen: that is, after the first asymmetric mitosis, but not in the generative or sperm cell nuclei, or at any other stages of pollen development (Fig. 1 A, Bottom and B). Real-time quantitative RT-PCR (qRT-PCR) analysis was in accord with these results, showing elevated DME RNA expression at the bicellular pollen stage, followed by rapid decreases as pollen matured (Fig. 1C). Thus, DME expression is not detected until the sperm cell lineage is separated from that of the vegetative cell, at which point DME is active specifically in the vegetative cell.
Fig. 1.
DME is specifically expressed in the vegetative nucleus of late bicellular stage pollen. (A) Sequential development of flowers (Top) and corresponding pollen development in 2.3 kb DME::GUS transgenic plants with DAPI (Middle) and GUS staining (Bottom). (B) The 2.3 kb DME::GFP expression (Left) in microspore (Top), bicellular (Middle), and tricellular (Bottom) stage pollen grains stained with DAPI (Right). G, generative nucleus; N, microspore nucleus; S, sperm cell nucleus; V, vegetative nucleus. (Scale bars: 5 μm.) (C) qRT-PCR analysis of DME expression in WT pollen development after normalization with ACT1, ACT3, and ACT12 expression. The four different stages analyzed using qRT-PCR are indicated in A. Values are plotted relative to the expression of DME in stage 4 mature pollen, which was set at 1.0, and represent the average of triplicate measurements ± SD.
Fig. S1.
Diagram and expression of a 2.3pDME::GUS reference line and two complementing constructs. (A) Diagram (Top) and expression pattern (Middle and Bottom) of a 2.3pDME::GUS reference line. The number in black before the slash is based on a TSS that we defined using 5′ RACE (see Fig. S2). The number in red after the slash is based on the translation start site of At5g04560.2, the major splice variant. (Middle) Images of seedlings at 7 days after germination (DAG 7), showing shoot meristems on the Left and root tips on the Right. (Bottom) Pistils on the Left and an ovule at higher magnification on the Right. (B) Diagram of the two cDME complementation constructs with 2.3kb upstream sequences (2.3kb cDME) or deleted 5′ UTR (+46 cDME). Both constructs contain the first and second introns.
The DME Promoter Lies Within the DME Transcriptional Unit and Contains both Positive and Negative Regulatory Elements.
To identify the elements that promote the striking pattern of DME expression in male and female companion cells, we systematically deleted portions of our 2.3pDME::GUS reference construct (Fig. 2A). Deletion of the entire 5′ region, from −2.3 kb to +46 bp downstream of the transcriptional start site (TSS), as defined by 5′ RACE (Fig. 2 A and B and Fig. S2), had no effect on DME::GUS expression in the central and vegetative cells. For each of these deletion constructs, both temporal and spatial DME::GUS expression profiles in transgenic plants reflected those of the reference construct (Fig. 2 A and B and Fig. S1A). We then deleted a larger block of sequence, up to 395 bp downstream of the TSS, at which point DME expression was decreased, and, finally, deletion of DME transcriptional unit sequence to 473 bp downstream of the TSS led to the complete loss of DME::GUS expression in both central and vegetative cells (Fig. 2 A and B). These data indicate that the regulatory sequences that are required for the proper expression of DME in the central and vegetative cells lie between 46 and 473 bp downstream of the TSS.
Fig. 2.
Diagram of the DME::GUS reporter constructs and expression of the T-DNA insertion lines in the DME region. (A) The name, staining intensity, and the coordinates for each construct are shown. CC, central cells; VC, vegetative cell of pollen; −, none; +, moderate; ++, strong. (B) GUS staining is shown in ovules and pollen. DAPI-stained pollen grains are shown in the Bottom row. Plants expressing transgenes 2.3kb to +395 displayed GUS expression in the central cell nucleus (arrow) and vegetative cell nucleus. No GUS expression was detected in +473 transgenic plants, and 2.3kb Pro. plants exhibited GUS expression only in the synergid cells (arrowhead). (Scale bars: ovule, 50 μm; pollen, 20 μm.) (C) dme T-DNA insertion alleles at 72 nt upstream (CS857766) and at 25 nt upstream (SALK-036171) of the TSS. Black box, translated exon; gray box, untranslated exon; first line, 5′ flanking sequences; other lines, intron. (D) qRT-PCR analysis of DME expression in homozygous dme mutant seedlings after normalization with ACT1, ACT3, and ACT12 expression. Values are plotted relative to the expression of DME in Ler WT, which was set at 1.0, and represent the average of triplicate measurements ± SD.
Fig. S2.
The 5′ RACE analysis of DME using inflorescence RNA. (A) Diagram of the two alternative forms of the DME 5′ region. Light gray box, 5′ UTR; dark gray box, translated exon; front line, 5′ flanking region; second line, first intron; black arrow, 5′ RACE primer; GSP, gene-specific primer; TSS, transcription start site. (B) Agarose gel electrophoresis of DME 5′ RACE products from total inflorescence RNA using the GSP2 primer. (C) qRT-PCR analysis to compare the relative expression levels of the two DME splice variants, At5g04560.1 and At5g04560.2, in aerial tissue and root tissue of 2-week-old seedling and in immature inflorescences. Expression levels are normalized to the ACT2 housekeeping gene and represent the average of triplicate measurements ± SD. (D) Relative distribution of DME TSS determined by 5′ RACE clones. Nucleotide positions relative to the start codon are indicated on the x axis. The graph summarizes the results of 24 RACE clones.
To verify genetically that DNA sequences upstream of the TSS do not regulate DME expression, we obtained two transfer DNA (T-DNA) insertion mutants from the Arabidopsis Biological Resource Center (ABRC): CS857766, which has a T-DNA insertion 72 bp upstream (−72) of the TSS, and SALK-036171, which has a T-DNA insertion 25 bp upstream (−25) of the TSS (Fig. 2C). Homozygous mutants of either line were developmentally and morphologically indistinguishable from WT and did not exhibit any defects in fertility or seed viability (Table S1), suggesting that DME is appropriately expressed and functions normally in these mutants. DME is also expressed in sporophyte tissues (14), and we found the level of DME expression in homozygous CS857766 and SALK-036171 seedlings to be the same as in WT seedlings (Fig. 2D).
Table S1.
Seed abortion in dme alleles
In transgenic plants where the sequence downstream from +83 was deleted and the upstream portion fused to GUS directly, “2.3kb Pro DME::GUS,” GUS expression was absent from the central and vegetative cells (Fig. 2 A and B; central cell nucleus within ovule indicated with arrow). However, strong ectopic GUS activity was observed in the synergid cells of mature female gametophytes in plants expressing this transgene (Fig. 2 A and B, arrowhead, and Table S4). Thus, a putative suppressor element that usually represses DME expression in synergid cells is present downstream of +83 bp. The lack of a nuclear localization signal (NLS) in this construct resulted in staining of the synergid cells’ cytoplasm.
Table S4.
Expression of T1 pDME:GUS transgenic plants
| Construct | Sporophytic (/total) | Central cell (/total) | Vegetative cell (/total) | Total |
| 2.3kb | 14 (0.74) | 18 (0.95) | 18 (0.95) | 19 |
| 0.5kb | 8 (0.80) | 10 (1) | 10 (1) | 10 |
| +7 | 9 (1) | 9 (1) | 9 (1) | 9 |
| +20 | 0 (0) | 18 (0.95) | 18 (0.95) | 19 |
| +33 | 5 (0.29) | 17 (1) | 17 (1) | 17 |
| +46 | 0 (0) | 22 (1) | 22 (1) | 22 |
| +395 | 0 (0) | 17 (1) | 14 (0.82) | 17 |
| +473 | 0 (0) | 0 (0) | 0 (0) | 13 |
| 2.3kb Δ5′ | 18 (0.90) | 20 (1) | 20 (1) | 20 |
| 2.3kb Pro. | 13 (0.72) | 0 (0) | 0 (0) | 18 |
| TU0 | 12 (1) | 11 (0.92) | 11 (0.92) | 12 |
| TU12 | 5 (0.50) | 0 (0) | 0 (0) | 10 |
| TU13 | 14 (0.82) | 0 (0) | 0 (0) | 17 |
| TU14 | 6 (0.55) | 9 (0.82) | 9 (0.82) | 11 |
| TU23 | 0 (0) | 0 (0) | 0 (0) | 15 |
| TU24 | 0 (0) | 14 (0.88) | 14 (0.88) | 16 |
| TU25 | 2 (0.17) | 12 (1) | 12 (1) | 12 |
| TU34 | 1 (0.08) | 10 (0.83) | 10 (0.83) | 12 |
| TU35 | 1 (0.11) | 9 (1) | 9 (1) | 9 |
| TU45 | 0 (0) | 10 (0.67) | 9 (0.60) | 15 |
| TU0_ΔSP | 3 (0.20) | 14 (0.93) | 14 (0.93) | 15 |
| TU0_ΔPOL | 10 (0.67) | 4 (0.27) | 3 (0.20) | 15 |
| TU0_ΔCC1 | 15 (1) | 11 (0.73) | 11 (0.73) | 15 |
| TU0_ΔCC2 | 20 (1) | 20 (1) | 20 (1) | 20 |
| TU0_ΔCC3 | 14 (0.93) | 4 (0.27) | 4 (0.27) | 15 |
| TU0_ΔHB | 6 (0.55) | 2 (0.18) | 2 (0.18) | 11 |
Expressing DME Polypeptide in the Central Cell with a Minimal Reproductive Promoter Rescues Seed Abortion and Aberrant DNA Methylation Associated with the dme-2 Mutation.
The +46 pDME::GUS/GFP transgene has the shortest sequence that correctly regulates reporter expression in the central cell and vegetative cells, without deleting internal DME coding sequences (Fig. 2 A and B and Fig. S3A). We therefore considered this transgene to contain the minimal reproductive promoter that could be used to drive the correct reproductive expression of a full-length DME polypeptide in a functional assay. We then constructed a +46 pDME::cDME transgene (Fig. S1B) to determine the functional significance of DME expression driven by this minimal reproductive promoter. We transformed dme-2 heterozygotes with the +46 pDME::cDME transgene (Fig. S3B). The dme-2 mutation is a loss-of-function null allele, and, in self-pollinated dme-2 heterozygous mutant plants, 50% of the F1 progeny seed inherit the maternal dme-2 mutant allele and abort their development whereas inheritance of the paternal mutant dme-2 allele has no effect on seed viability (2). To test for +46 pDME::cDME transgene function, we analyzed whether it could rescue seed abortion in transgenic lines. In self-pollinated plants that were hemizygous for a single transgene locus, and heterozygous for dme-2, 25% of the F1 seed inherited the mutant maternal dme-2 allele and aborted their development, and 25% inherited both the mutant maternal dme-2 allele and the transgene. Therefore, full complementation of the mutant maternal dme-2 allele by the +46 pDME::cDME transgene results in 25% seed abortion (2), which we observed (Fig. S3B and Table S2). Moreover, self pollination of plants heterozygous for dme-2 and hemizygous for +46 pDME::cDME generated plants homozygous for both the dme-2 mutation and the +46 pDME::cDME transgene, which displayed the same low seed abortion rate (<1%) as both WT plants and homozygous dme-2 plants expressing the homozygous 2.3kb pDME::cDME control transgene (Fig. S3B and Table S2), demonstrating the functional activity of the minimal reproductive promoter.
Fig. S3.
DME expression driven by the +46 minimal reproductive promoter transgene rescues dme-2-mediated seed abortion. (A) The +46 pDME:GFP transgene is expressed in the central cell before fertilization, but not in the endosperm after fertilization. (Scale bars: 50 μm.) (B) Siliques of dme-2 mutants containing 2.3kb cDME or +46 cDME complementing constructs. See also Fig. S1 and Table S2. (Scale bar: 1 mm.)
Table S2.
Seed abortion ratio of dme-2 mutants containing two different complementing constructs
| Genotype | No. of siliques | Total viable seed | Total aborted seed | Average seed abortion ratio ± SD, % | χ2 for 1:1 | χ2 for 3:1 | P for 1:1 | P for 3:1 |
| Col(gl) | 8 | 434 | 1 | 0.2 ± 0.6 | ||||
| 2.3kb cDME/- ;DME/dme-2 F1 | 10 | 344 | 147 | 29.9 ± 5.9 | 0.0114 | >0.90 | ||
| +46 cDME/- ;DME/dme-2 F1 | 10 | 324 | 132 | 28.9 ± 5.2 | 0.0516 | >0.80 | ||
| 2.3kb cDME/- ;dme-2/dme-2 F2 | 13 | 286 | 271 | 48.7 ± 6.7 | 0.5251 | >0.45 | ||
| 2.3kb cDME/2.3kb cDME ; \dme-2/dme-2 F2 | 6 | 264 | 6 | 2.2 ± 3.5 | ||||
| +46 cDME/- ;dme-2/dme-2 F2 | 9 | 189 | 183 | 49.2 ± 5.2 | 0.7557 | >0.35 | ||
| +46 cDME/+46 cDME ;dme-2/dme-2 F2 | 10 | 552 | 1 | 0.2 ± 0.5 | ||||
| DME/dme-2 | 7 | 142 | 145 | 50.5 ± 9.5 | 0.8594 | >0.35 |
Seed abortion ratio of dme-2 mutants containing two different complementing constructs. We selected heterozygous dme-2 mutants with single copy transgenes using the antibiotic resistant ratio: For the 2.3kb cDME control promoter transgene, we observed an 81:17 resistant:sensitive ratio. For the +46 cDME minimal reproductive promoter transgene, we observed a 95:22 resistant:sensitive ratio. We detected a significant reduction in seed abortion, 28.9% (n = 456), with the +46 cDME minimal reproductive promoter transgene, compared with 50.5% seed abortion (n = 287) in dme-2/DME control plants. Similar observations of 29.9% seed abortion (n = 491) were obtained using the control 2.3-kb DME promoter, which indicates that expressing the DME polypeptide using the minimal reproductive promoter is sufficient to complement seed abortion caused by the dme-2 loss-of-function mutation. P, probability that the deviation from the indicated segregation ratio of viable:aborted seeds is due to chance. See also Figs. S2 and S3B.
Seed abortion resulting from the dme-2 mutation is caused, at least in part, by the resultant aberrant expression pattern of imprinted components of the PRC2 in endosperm (2, 4, 15–17). In the absence of DME, PRC2 is defective, and endosperm development is severely compromised, resulting in embryo abortion (18). Because seed abortion is rescued by the +46 pDME::cDME transgene, we hypothesized that DME expression driven by the minimal reproductive promoter is able to demethylate the central cell genome-wide, including specific PRC2 genes, resulting in a functional endosperm with a distinctive pattern of maternal endosperm genome hypomethylation compared with the paternal endosperm genome. To test this hypothesis, we pollinated dme-2/dme-2 homozygous Col(gl) (Columbia ecotype, homozygous for the glabrous mutation) plants that were also homozygous for the +46 pDME::cDME transgene, with WT Ler (Landsberg ecotype homozygous for the erecta mutation) pollen. F1 seeds were harvested at 9 days after pollination, endosperm was obtained by manual seed dissection, and genomic DNA was isolated. Maternal and paternal genomes were distinguished by Col versus Ler single nucleotide polymorphisms, and DNA methylation profiles were obtained by next generation bisulphite sequencing of DNA (3). We analyzed the methylome of F1 endosperm from dme-2/dme-2 homozygotes that were homozygous for the +46 pDME::cDME transgene (dme-2; +46 cDME) and compared it with a WT control (Col-0 crossed to Ler), and with the methylome of seeds inheriting the dme-2 mutation maternally (3). We found that the maternal allele of F1 dme-2; +46 cDME endosperm is normally methylated at maternally (e.g., FIS2, FWA) and paternally (e.g., YUK10, PHE1) expressed imprinted gene loci and resembles the WT maternal allele whereas these loci are hypermethylated in dme-2 (Fig. 3A). Genome-wide, the hypermethylation phenotype seen in dme-2 maternal endosperm, demonstrated by the increased density of genomic sites with a fractional methylation level between 0.5 and 1 (Fig. 3B, dme-2 minus WT kernel density trace), is fully complemented in dme-2; +46 cDME endosperm and resembles the WT endosperm methylome (Fig. 3B, dme-2; +46 cDME minus WT trace, and Fig. S4) whereas the paternal allele is unaffected (Fig. 3C). Thus, the minimal reproductive promoter promotes functional DME expression required for DNA demethylation.
Fig. 3.
DME expression driven by the +46 transgene can correct the methylation phenotype of homozygous dme-2 mutant endosperm. (A) Snapshots of CG DNA methylation at selected imprinted loci. Each track represents a different genotype: crimson trace, WT subtracted from dme-2 homozygous endosperm expressing the +46 transgene; orange trace, WT subtracted from dme-2 heterozygous endosperm; green tracks are raw CG methylation data in the three genotypes compared. Differential methylation at both maternally expressed (FIS2, FWA) and paternally expressed (YUK10, PHE1) imprinted loci (i.e., maternal hypomethylation of imprinting control regions) is regained in dme-2 homozygous endosperm when the +46 transgene is expressed. Gray boxes show the imprinting control regions at each locus, and arrows show the direction of gene transcription. (B) Kernel density plots of CG methylation differences between the maternal alleles of dme-2 homozygous endosperm expressing the +46 transgene and WT (i, crimson trace) and dme-2 heterozygous endosperm and WT (ii, orange trace,). Hypermethylation of the dme-2 mutant endosperm is evident in the increased density at a fractional methylation difference of between 0.5 and 1 in ii and is corrected by the +46 transgene as seen by the loss of this density increase in i. (C) Kernel density plots of CG methylation differences between the paternal alleles of dme-2 homozygous endosperm expressing the +46 transgene and WT (i, blue trace) and dme-2 heterozygous endosperm and WT (ii, aquamarine trace). Methylation of the paternal (WT Ler) alleles is the same in each genotype, showing that the +46 transgene does not affect methylation postfertilization.
Fig. S4.
DME expression driven by the +46 minimal reproductive promoter transgene can correct the methylation phenotype of homozygous dme-2 mutant endosperm. (A) Kernel density plots of CG methylation differences between the maternal alleles of dme-2 mutant endosperm (3) and dme-2 mutant endosperm expressing the +46 transgene (i, red trace) and dme-2 heterozygous endosperm and WT (ii, orange trace). Similar hypermethylation of the dme-2 mutant endosperm relative to both the WT and the dme-2 mutant expressing the +46 transgene is evident in the increased density at a fractional methylation difference of between 0.5 and 1 in both plots. Compared with the dme-2 mutant endosperm, the same genomic sites are relatively hypomethylated in WT and the +46 transgene DNA. (B) Kernel density plots of CG methylation differences between the maternal alleles of dme-2 mutant endosperm and dme-2 mutant endosperm expressing the +46 transgene for (i) all genomic sites (ii) DME target sites only.
A 357-bp Region of the DME Transcriptional Unit Is both Necessary and Sufficient to Generate the Appropriate DME Expression Profile During Female Gametophyte Development.
To identify where the precise regulatory elements that control DME expression in the central cell are located, we carried out further deletions within the 2-kb region that we had so far identified to be necessary and sufficient for fully functional DME activity. This gain-of-function (GOF) construct series is denoted “Truncated 5′-UTR” (TU) (Fig. 4A and Fig. S5), for which we used increasingly smaller portions of the 748-bp-long −90 to +658 region around the DME TSS to drive GUS expression. The TU0 reporter construct, containing the full −90 to +658 region, showed the same expression pattern and intensity as the reference 2.3pDME::GUS construct, except for GUS expression in the cytoplasm of cells expressing GUS because the endogenous nuclear localization sequence of DME is downstream of 658 bp, and therefore absent from all TU constructs (Fig. 4 and Figs. S5 and S6).
Fig. 4.
Diagram of the DME::GUS reporter constructs for fine mapping of cis-elements and their expression patterns. The TU (truncated 5′-UTR) series of constructs. (A) The name, staining intensity, and coordinates for each construct are shown. CC, central cells; VC, vegetative cell of pollen; −, none; +, moderate; ++, strong. (B) GUS staining is shown in ovules and pollen. DAPI-stained pollen grains are shown in the Bottom row. TU0, TU34, and TU45 transgenic plants exhibited GUS expression in the central cell and pollen. No GUS expression was detected in TU12 and TU23 plants. (Scale bars: 50 μm.)
Fig. S5.
Diagram of the DME::GUS reporter constructs for fine mapping of cis-elements. The TU (truncated 5′ UTR) series of constructs. The name, staining intensity, and the coordinates for each construct are shown. CC, central cells; SDL, seedling; −, none; +, moderate; ++, strong.
Fig. S6.
Catalog of the expression patterns of the TU DME:GUS construct series. TU12, TU13, and TU23 plants showed weak GUS signal only in sporophytic tissues. By contrast, TU34, TU35, and TU45 plants showed GUS expression only in central cells. (Scale bars: seedlings, 1,000 μm; shoot meristem and pistil, 200 μm; ovule, 50 μm.)
From our GOF TU series, the minimal sequence that we found to be necessary and sufficient to drive DME expression in the central cell was 357 bp in length, from +202/+559 (transgene TU34) (Fig. 4). TU23 (+46/+415) plants did not show any GUS expression, but TU34 plants displayed GUS activity in the central cell (Fig. 4). Because our previous deletion to 473 bp downstream of the TSS led to the complete loss of DME::GUS expression (Fig. 2A), we deduced that the central cell regulatory region lies in a 57-bp fragment between the +416 and +472 positions. We also observed reduced GUS expression in the central cell in TU45 (+363/+658), indicating that quantitative regulation of central cell expression also involves a region between +202 and +362, denoted the quantitative regulatory element (QE) (Fig. 5A).
Fig. 5.
Internal deletion/substitution of cis-elements. (A) Summary of DME cis-regulatory elements. Dark gray box, translated exon; light gray box, 5′-UTR; line, first intron; red line, sporophytic element (SPE); blue line, central cell element (CCE); green line, pollen vegetative cell element (VCE) ; dotted line, quantitative regulatory element (QE). (B) Diagram of DME::GUS internal deletion and substitution constructs of the cis-elements. CC, central cells; VC, vegetative cell of pollen; −, none; (+), weak; ++, strong; Δ, deletions or substitutions. (C) GUS staining is shown in ovules and pollen. DAPI-stained pollen grains are shown in the bottom of each pollen. TU0_ΔSP, same GUS expression pattern as TU0; TU0_ΔPOL, central cell and pollen GUS disappeared; TU0_ΔCC1 and TU0_ΔCC2, only the pollen expression disappeared; TU0_ΔCC3, central cell and pollen GUS disappeared. TU0_ΔHB, central cell GUS was significantly reduced and pollen GUS disappeared. (Scale bars: ovule, 50 μm; pollen, 20 μm.)
DME Expression in Sporophytic Tissues Is Regulated by Distinct DNA Sequences.
DME is expressed in the sporophyte shoot apical meristem (SAM), leaf primordia, and the root apical meristem (Fig. S1A and Figs. S5–S8) and is required for floral and vegetative developmental patterning (2, 14). To determine the relationship between the regulation of DME in reproductive and sporophytic tissues, we further investigated the regulatory regions of DME to elucidate those required for sporophytic DME expression. We identified a 349-bp region, from −90 to +259 that is necessary and sufficient for DME expression in sporophytic tissues (TU12, Figs. S5 and S6). Next, we generated constructs to narrow this region, identifying 13 bp close to the TSS, between +7 and +19, required for the sporophytic expression of DME, which we designated as a necessary sporophytic enhancer (SPE) (Fig. 5A and Figs. S7 and S8). Deletion specifically of the SPE (TU0_ΔSP) results in loss of sporophytic, but not reproductive, DME expression (Fig. 5 B and C).
Fig. S8.
Catalog of the expression patterns of the DME::GUS deletion series. The 2.3 kb, 0.5 kb, +7 DME::GUS, and 2.3kb Δ5′ constructs showed GUS expression in both sporophytic tissues and central cells. The +20, +33, +46, and +396 were expressed only in central cells. No expression was detected in +473 DME::GUS plants. The 2.3 kb Pro. exhibited DME:GUS expression in sporophytic tissues, but not in the central cell nucleus. Ectopic expression was detected in the micropylar end of the embryo sac. (Scale bars: seedling, 1,000 μm; shoot meristem and pistil, 200 μm; ovule, 50 μm.)
Fig. S7.
Diagram of the DME::GUS reporter construct series and their expression patterns. CC, central cells; SDL, seedlings; −, none; +, moderate; ++, strong.
Overlapping 15- and 47-Base Pair Regions Are Necessary for DME Expression in the Central and Vegetative Cells, Respectively.
As stated previously, a 57-bp element necessary for central cell DME expression lies between the +416 and +472 positions. To establish whether this sequence was also sufficient to drive DME expression, we generated constructs containing one to four copies of this 57-bp fragment with and without the minimal CaMV 35S promoter downstream, but none of these constructs exhibited any GUS expression in any tissue; therefore, we were unable to conclude that this sequence is sufficient for expression (Fig. S9). Nevertheless, to investigate this region further, we generated fine-deletion constructs TU0_ΔPOL (Δ+416/+462), TU0_ΔCC1 (Δ+416/+431), TU0_ΔCC2 (Δ+432/+447), and TU0_ΔCC3 (Δ+448/+462) (Fig. 5B) to establish the sequence necessary for regulating central cell expression. GUS activity was detected in the central cell in TU0_ ΔCC1 and TU0_ΔCC2, but not in TU0_ΔCC3 or TU0_ΔPOL plants (Fig. 5 B and C); therefore, the sequence necessary for central cell expression, denoted the “CCE,” is ∼15 bp in length and is located between +448 and +462 nt (Fig. 5A). Vegetative cell DME expression is present in TU0_ΔSP but disappears in TU0_ΔPOL and in each of TU0_ΔCC1, TU0_ΔCC2, and TU0_ΔCC3 (VC in Fig. 5 B and C), demonstrating that vegetative cell expression of DME specifically requires the 47-bp +416/+462 sequence, denoted the “VCE,” which encompasses, but is broader than, the +448 /+462 CCE (Fig. 5A).
Fig. S9.
Diagram of the gain-of-function VCE tandem repeat constructs with and without the 35S minimal promoter. A scheme is shown at left. The name of the construct is shown, and the presence of staining in seedlings (SDL) and central cells (CC) is indicated.
The 15-bp CCE Sequence, Shared by the VCE, Is Required for DME Expression and Is Predicted to Bind Several Key Transcription Factors.
DME expression in the vegetative and central cells is thought to have a common function, in regulation of transposon silencing in the germline. As such, the 15-bp common region of the VCE and CCE elements is of particular intrigue. This sequence contains the 9-bp “CATTTATTG” motif, which is strikingly similar to the pseudopalindromic targets of the Arabidopsis Homeobox HD-ZIP family of plant-specific transcription factors: for example, the recognition sequence “CAAT(T/A)ATTG” of subfamily 1 (19, 20). To examine the role of this AT-rich sequence in the expression of DME, 7 bp of an AT-rich sequence in TU0 was changed to a GC-rich sequence (Fig. 5B). This change resulted in a significant reduction of GUS activity in the central cell of TU0_ΔHB plants and the complete absence of GUS expression in the vegetative cell of pollen (Fig. 5C). Thus, the pseudopalindromic sequence is required for normal central and vegetative cell DME expression.
Our identification of precise coordinates for key regulatory elements of DME expression enabled us to carry out preliminary investigations to reveal potential interacting transcription factors. A recent genome-wide analysis to characterize regulatory elements and transcription factor binding sites used a novel high throughput DNA affinity purification sequencing assay (DAP-seq), generating a “cistrome” map for 30% of transcription factors in Arabidopsis (21). By correlating our VCE and CCE coordinates with this cistrome dataset, we were able to identify 40 potential candidates that bind these regions in vitro and may therefore be involved in DME regulation in reproductive tissues (Table S3). Among these candidates are 10 HD-ZIP transcription factors, spanning the four subfamilies, which is consistent with our finding functional targets of the HD-ZIP family in the common region of the VCE and CCE elements.
Table S3.
Transelement candidates of CCE/VCE binding from published DAP-seq data
| Name | AGI | DAP-seq† | VCE reads‡ |
| HD-ZIP I | |||
| ATHB23 | AT1G26960 | ** | 11F6R |
| ATHB6 | AT2G22430 | * | 2F4R |
| ATHB40 | AT4G36740 | * | 2F2R |
| LMI1 | AT5G03790 | * | 9F18R |
| HD-ZIP II | |||
| HAT22 | AT4G37790 | * | 4F1R |
| HD-ZIP III | |||
| ATHB15 | AT1G52150 | * | 5F |
| PHB | AT2G34710 | * | 5F1R |
| ATHB8 | AT4G32880 | * | 5F 8R |
| HD-ZIP IV | |||
| ANL2 | AT4G00730 | ** | 10F10R |
| HDG4 | AT4G17710 | * | 3F5R |
| KNOX | |||
| KNAT1 | AT4G08150 | * | 5F8R |
| WOX | |||
| WOX4 | AT1G46480 | * | 1F1R |
| bHLH | |||
| HEC3 | AT5G09750 | * | 6R |
| bZIP | |||
| TGA3 | AT1G22070 | * | 4R |
| CAMTA | |||
| CAMTA2 | AT5G64220 | * | 3F6R |
| DBP | |||
| DBP | AT3G51470 | * | 1F3R |
| FAR1 | |||
| FAR1_FRS12 | AT5G18960 | * | 5F |
| FHA | |||
| FHA | AT2G21530 | * | 5R |
| FHA | AT5G47790 | * | 4R |
| GEPB | |||
| GEPB | AT4G25210 | * | 5F6R |
| GRF | |||
| GRF6 | AT2G06200 | * | 5F9R |
| HSF | |||
| HSFA6A | AT5G43840 | * | 1F4R |
| HSFA4C | AT5G45710 | * | 1F2R |
| MADS | |||
| AGL63 | AT1G31140 | *** | 19F20R |
| AGL11 | AT4G09960 | * | 2F6R |
| AGL98 | AT5G39810 | * | 5R |
| AGL52 | AT4G11250 | * | 23F15R |
| AGL23 | AT1G65360 | * | 5R |
| MYB-LIKE | |||
| HD-like | AT4G01280 | *** | 15F12R |
| EPR1 | AT1G18330 | *** | 26F27R |
| LCL1 | AT5G02840 | *** | 26F24R |
| LHY1 | AT1G01060 | *** | 14F15R |
| RVE1 | AT5G17300 | *** | 14F13R |
| LCL5 | AT3G09600 | *** | 27F28R |
| HD-like | AT3G10113 | *** | 19F27R |
| NAC | |||
| ANAC018 | AT1G52880 | 8F9R | |
| ORPHAN | |||
| ARR22 | AT3G04280 | *** | 9F12R |
| REM | |||
| REM19 | AT1G49480 | * | 16F13R |
| REMB3 | AT2G31460 | * | 5R |
| SBP | |||
| SLP1 | AT1G07010 | * | 7F |
Factors were included if the center of the VCE element overlapped with the peak of the reads. Peak size is indicated by number of *, with * indicating a minor peak and *** indicating distinctive peak.
Reads are listed here as the number of reads that mapped to either the Watson (F) or Crick (R) DNA strands. Reads were normalized according to ref. 21.
Discussion
Here, we show that the regulation of DME expression is mirrored in both male and female gametophytes, developing simultaneously upon germline differentiation in distinct reproductive organs. DME expression is restricted to the vegetative cell nucleus after the first asymmetric mitosis, at the late bicellular stage of pollen development (Fig. 1), which is concurrent with separation of the generative and vegetative cell lineages, so that the demethylation activity of DME is restricted to the vegetative cell whereas the sperm genome remains highly methylated at DME targets. This expression profile is likewise reflected in the female gametophyte. During female gametogenesis, the third mitotic division is followed immediately by cellularization and differentiation, generating antipodal cells at the chalazal pole, and the egg cell, synergids, and two polar nuclei at the micropylar pole (22). It is immediately after this differentiation step that DME expression is activated so that expression is confined primarily to the polar nuclei, which fuse to form the central cell, and is absent from the egg (2).
Expression of DME in companion cells, and the evasion of DME expression in gametes, is key for understanding the function of DNA demethylation during plant reproduction. This pattern explains how the maternal endosperm genome is hypomethylated compared with the paternal endosperm genome. Maternally hypomethylated loci are either directly or indirectly (via PRC2 activity) responsible for parent-of-origin gene expression: i.e., gene imprinting, in the endosperm (5, 13). The fact that DME is not expressed in the egg or sperm cells is responsible, at least in part, for the similarity of the maternal and paternal embryo methylomes (3) and, therefore, the fact that genes displaying parent-of-origin expression in endosperm do not do so in the embryo (5, 23). Maternal genome hypomethylation is required for seed development, but the demethylation of the vegetative cell does not directly affect seed viability. Instead, demethylation of both the central and vegetative cells at DME targets, such as small, AT-rich, and nucleosome-depleted euchromatic TEs, likely promotes expression of TEs in these cells. Demethylated companion cells do not pass on their genome to the next generation; therefore, the genomic instability resulting from transposon transcription is not deleterious to the species as a whole. Rather, there is evidence to suggest that the RdDM pathway then promotes corresponding TE methylation in the egg and sperm cells, respectively (3, 11, 12). In this way, the companion cell acts sacrificially, reinforcing and protecting the genomic integrity of egg and sperm, which will be inherited by the next generation. The function of DME expression in companion cells provides support for the unique importance of double fertilization involving companion cells during evolution.
We explored the regulatory sequences that contribute to this remarkable expression profile by producing a comprehensive array of iteratively deleted reporter transgenes for the regions upstream of the DME translational start site. With the exception of a negative regulatory region that suppresses DME expression in female gametophyte synergid cells, all other regulatory elements reduced DME expression when lost or mutated (Figs. 2 and 4), suggesting that the majority of transcriptional regulation of DME is positive. The lack of DME expression in the fertilized endosperm, needed to preserve regions of DNA demethylation that are specific to the maternal endosperm genome (3), is therefore likely caused by a decrease in activity of a positive regulator or regulators, rather than the appearance of a negative regulator.
Using our deletion transgenes, we found that sequences regulating DME expression were contained within its transcriptional unit. We designated the +46 transgene, which consists of 592 bp of sequence before the translational start site, as the minimal reproductive promoter and used this +46 minimal promoter in a functional construct to drive expression of DME cDNA. The expression of this transgene rescued both the seed abortion and genome-wide DNA methylation phenotypes of dme-2 heterozygous and homozygous mutants, showing that the expression timing, level, and tissue specificity of DME expression in reproductive tissues is recapitulated with a promoter sequence of 592 bp contained within the transcriptional unit. Within this sequence, we identified a 47-bp VCE, overlapping with a 15-bp CCE, necessary for regulation of the vegetative and central cell DME expression patterns, respectively. The CCE and VCE are distinct from the 13-bp SPE close to the TSS that promotes DME expression in the sporophyte. Each of the three enhancer sequences is conserved in closely related Brassicaea family members, such as Arabidopsis lyrata, Capsella rubella, and Brassica rapa (Fig. S10), but they are missing from the DME homologs ROS1, DML2, and DML3, which are expressed much more broadly than DME (24) and do not contribute to demethylation in the central cell.
Fig. S10.
DME and ROS1 homolog comparisons in publically available Brassica family DNA sequences. Lines above the DNA sequence indicate the cis-regulatory elements found by these experiments. SPE in the DME CT-repeats (red box) and the 9 bp of sequence that are similar to the pseudopalindromic target sequence that is similar to Arabidopsis thaliana Homeobox 1 (Athb-1) (blue box) are well conserved in Brassica family. A.Ly, Arabidopsis lyrata; A.Th, Arabidopsis thaliana; B.Ra, Brassica rapa; C.Ru, Capsella rubella.
Because the CCE is contained entirely within the VCE, it is possible that the control of DME expression in each of the companion cells of the gametes shares a common regulatory pathway. The overlapping VCE/CCE sequence of 15 bp (+448/ +462) is AT-rich, including 9 bp with striking similarity to the pseudopalindromic “CAAT(T/A)ATTG” sequence, which is a target of the HD-ZIP plant-specific homeobox transcription factor family (25–27). Substitution of this motif led to a large reduction in central cell DME expression and ablation of vegetative cell DME expression, showing that this pseudopalindromic sequence is required for correct DME regulation.
Using the coordinates that we derived for the VCE and CCE and our analyses of the recently published DAP-seq “cistrome” collection (18), we were able to catalog a list of 40 potential transcription factors, including 10 HD-ZIPs, that bind to these elements in vitro (21). MADS-box transcription factor AGL80 is required for DME expression in the central cell (28) so it is likely that MADS-box binding domains are present in this regulatory region, and several MADS-box transcription factors were found to bind to the VCE by DAP-seq (Table S3) (21); however, AGL80 was not specifically tested in the DAP-seq screen.
If there is a transcription factor that coregulates DME in the central and vegetative cells, it would be expressed in both these tissues, at a similar time to DME itself. However, as we show here, DME expression in the male gametophyte is confined to a short period of the bicellular pollen stage and is often not detected in pollen expression datasets (29–31). Thus, to identify potential transcription factors that may bind the shared sequence of the VCE and CCE and regulate DME, precise establishment of their endogenous expression profile using reporter genes, and their effect on DME expression when mutated or ectopically expressed, will be required in the future.
In summary, we show here that DME expression during reproduction is confined to a narrow window of time, and to single companion cells, in female and male gametophytes, which is necessary for its role in seed viability, gene imprinting, and transgenerational transposon silencing. We delineate specific, conserved enhancer sequences required for the precise expression pattern of DME and identify candidate transcription factors by their in vitro binding patterns at the VCE and CCE, information that will be valuable in the future to delineate the regulatory pathways that control DME expression.
Materials and Methods
Please SI Materials and Methods full details of methods.
Plant Materials and Growing Conditions.
All of the promoter constructs used in this study were transformed into Arabidopsis Columbia gl. The dme-1 homozygous mutant allele is in Landsberg er (Ler) background (2). Heterozygous dme-2 in Col(gl) was used for the complementation test.
Next Generation Bisulphite Sequencing.
F1 endosperm was hand-microdissected at 8 to 10 days after pollination of homozygous dme-2; +46 cDME plants, or WT Col-0, with Ler pollen, allowing distinction of parental alleles. Bisulphite sequencing libraries were prepared as described previously (3), and in SI Materials and Methods.
Recombinant Plasmid Construction.
Methods for generating deletion constructs and TU_GUS (Table S4) are described in SI Materials and Methods. Primers are listed in Table S5.
Table S5.
Primer list
| Allele/Target | Name | Primer sequence |
| Primers for genotyping | ||
| WT sibling of dme-1 | B13F | 5’-CTGATCAGATGCCCTTCTCC-3’ |
| B13R | 5’-CCCAATCCATTGGTCTTGTC-3’ | |
| dme-1 | B13F | 5’-CTGATCAGATGCCCTTCTCC-3’ |
| SKI015-LB | 5’-TTGACCATCATACTCATTGCTG-3’ | |
| WT sibling of dme-2 | B33F | 5’-CACTTGTTCCCTATGAGAGC-3’ |
| B33R | 5’-CACTGATTGTGATGTTCCAC-3’ | |
| dme-2 | B33R | 5’-CACTGATTGTGATGTTCCAC-3’ |
| SKI015-LB | 5’-TTGACCATCATACTCATTGCTG-3’ | |
| WT sibling of CS857766 | −90 DME F | 5’-GCAACAACGTCCTCGTGAA–3’ |
| +659 DME R | 5’-CCGGATCAGCCCTCGAATTC-3’ | |
| CS857766 | −90 DME F | 5’-GCAACAACGTCCTCGTGAA–3’ |
| p745 | 5'-AACGTCCGCAATGTGTTATTAAGTTGTC-3' | |
| WT sibling of SALK-036171 | −90 DME F | 5’-GCAACAACGTCCTCGTGAA–3’ |
| +659 DME R | 5’-CCGGATCAGCCCTCGAATTC-3’ | |
| SALK-036171 | −90 DME F | 5’-GCAACAACGTCCTCGTGAA–3’ |
| LB3 | 5’-TAGCATCTGAATTTCATAACCAATCTCGATACAC-3’ | |
| Primer for RT or qRT-PCR | ||
| DME | cDNA5 | 5’-CAGAAGTGTGGAGGGAAAGCGTCTGGC-3’ |
| SKEN5 | 5’-GCAATGCGTTTGCTTTCTTCCAGTCATCT-3’ | |
| ACT1 | ACT1F | 5’-TCTTGATCTTGCTGGTCGTG-3’ |
| ACT3 | ||
| ACT12 | ACT1R | 5’-AATGGTGATCACTTGCCCATC-3’ |
| ACT2 | ACT2_TAQ_F | 5’-CCATCCAAGCTGTTCTCTCC-3’ |
| ACT2_TAQ_R | 5’-GACGGAGGATGGCATGAGGAAG-3’ | |
| At5G04560.2 | DME.2_TAQrpt_F | 5’-CACAAGCTTGCTGAGTGGTGG-3’ |
| DME.2_TAQrpt_R | 5’-CAGACTGACCCAACTGCTTCTC-3’ | |
| At5G04560.1 | DME.1_F_TAQ | 5’-GTCAATGTGAGTGATCAAATC-3’ |
| DME.1_R_TAQ | 5’-GGTCCATCTGTTCAAAACCATG-3’ |
SI Materials and Methods
Recombinant Plasmid Construction.
The generation of the reference lines 2.3pDME::GUS and 2.3pDME::GFP was previously described (2). All of the 5′ deletion constructs were generated by PCR-based cloning using a 2.3kb DME::GUS template in the pBI101.1 vector (Clontech Ltd.). All of the TU_GUS lines used for fine mapping of the cis-elements were generated in the pDW137 vector (Clontech Ltd.) using the 2.3kb DME::GUS template. Internal deletions in the TU lines were generated by chimeric PCR. Substitution TU constructs were generated by PCR-mediated site-directed mutagenesis. GUS expressions of all transgenic lines used in this study from initial independent T1 lines are summarized in Table S4.
Transgenic Seed Lines.
CS857766 and SALK-036171 seeds from the ABRC stock center are in Columbia-0. Plants were grown under conditions described previously (32).
Histochemical GUS Staining, GFP Fluorescence, and Microscopy.
Methods for GUS staining, fixing tissues, and Axio Imager A1 (Carl Zeiss) microscopy are as described previously (33). GFP fluorescence in the gamete cells and the DAPI staining were observed with LM700 (Carl Zeiss) confocal microscopy.
Gene Expression Analysis.
Total RNAs were isolated using RNAqueous (Ambion). After RNase-free DNase (TaKaRa Bio) treatment, 3 μg of total RNAs were used to synthesize cDNA using oligo-dT primers and M-MLV reverse transcriptase (Ambion). RNA levels were quantified by qPCR (CFX96; Bio-Rad) using iQ SYBR Green supermix (Bio-Rad), and data were analyzed with the CFX manager software (Bio-Rad). For control normalizations, we used Actin gene expression. Our ACT primer set (Table S5) could amplify ACT1, ACT3, and ACT12 in a single reaction. Thus, transcript levels were normalized to the total sum of ACT1, ACT 3, and ACT 12 expression in Figs. 1 and 2. For the relative comparison of two DME splice variants (Fig. S2), we used ACT2 gene expression for the normalization. Relative expression levels were quantified by the ΔΔCt method {[cycle threshold (Ct) of gene of interest] − [cycle threshold (Ct) of the reference genes]}. Real-time SYBR-green dissociation curves showed one species of amplicon for each primer combination. All of the primers for qRT-PCR are listed in Table S5.
Identification of DME Regulatory Regions—TU0 Vector Series.
Serial truncations of the 5′-UTR were named “TU.” The DME start codon is at +639 so we amplified the region from −90 bp to +658 bp. This fragment, TU0, contains all of the regulatory elements of endogenous DME, except the NLS. The TU0 fragment was ligated into two different GUS-containing vectors, pBI101.1 and pDW137, to control for differences in vector efficiency. The TU0 reporters from both vectors showed the same expression pattern and intensity as the 2.3kb DME::GUS construct, except for GUS expression in the cytoplasm of cells expressing DME, due to the lack of an NLS in TU0 (Fig. 4B and Fig. S5).
Identification of DME Regulatory Regions—Element Deletion.
Transgenes TU0_ ΔSP (Δ+7/+45), TU0_ ΔPOL (Δ+416/+462), TU0_ ΔCC1 (Δ+416/+431), TU0_ ΔCC2 (Δ+432/+447), and TU0_ ΔCC3 (Δ+448/+462) were generated by deleting the sequences necessary for sporophytic tissue, and central cell DME expression from TU0 (Fig. 5B). Because +472 is the end point of the first intron, +462 was chosen instead of +472 for the deletion to prevent potential abnormal splicing. TU0_ ΔSP plants showed GUS expression only in the central and vegetative cells, and TU0_ ΔCC and TU0_ ΔCC3 plants showed GUS expression only in sporophytic tissues, confirming that we had successfully identified the locations of the sporophytic element (SPE) and central cell element (CCE) of the DME gene (Fig. 5A).
Identification of DME Regulatory Regions—Element Substitution.
Transgene TU0_ ΔHB (Δ+450/+456) was generated by altering the known regulatory cis-element sequence in TU0 (Fig. 5). GUS expression of TU0_ ΔHB plants was not detected in vegetative cells of male gametes and was significantly reduced in central cells of female gametes.
Acknowledgments
We thank Jihyun Yu (Seoul National University) for thoughts about the promoter assays and for providing vectors. This study used the Vincent J. Coates Genomics Laboratory at the University of California, Berkeley, supported by NIH S10 Instrumentation Grants S10RR029668 and S10RR027303. This work was supported by Next-Generation BioGreen 21 Program Grant PJ01101802 (to Y.C.), NRF of Korea Grant 2014R1A2A2A01004887 (to Y.C.), and NIH Grant GM69415 (to R.L.F.). J.-S.P. was supported by the Stadelmann-Lee Scholarship Fund, Seoul National University.
Footnotes
The authors declare no conflict of interest.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1620592114/-/DCSupplemental.
References
- 1.Park K, et al. DNA demethylation is initiated in the central cells of Arabidopsis and rice. Proc Natl Acad Sci USA. 2016;113(52):15138–15143. doi: 10.1073/pnas.1619047114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Choi Y, et al. DEMETER, a DNA glycosylase domain protein, is required for endosperm gene imprinting and seed viability in arabidopsis. Cell. 2002;110(1):33–42. doi: 10.1016/s0092-8674(02)00807-3. [DOI] [PubMed] [Google Scholar]
- 3.Ibarra CA, et al. Active DNA demethylation in plant companion cells reinforces transposon methylation in gametes. Science. 2012;337(6100):1360–1364. doi: 10.1126/science.1224839. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Gehring M, et al. DEMETER DNA glycosylase establishes MEDEA polycomb gene self-imprinting by allele-specific demethylation. Cell. 2006;124(3):495–506. doi: 10.1016/j.cell.2005.12.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hsieh TF, et al. Regulation of imprinted gene expression in Arabidopsis endosperm. Proc Natl Acad Sci USA. 2011;108(5):1755–1762. doi: 10.1073/pnas.1019273108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Schoft VK, et al. Function of the DEMETER DNA glycosylase in the Arabidopsis thaliana male gametophyte. Proc Natl Acad Sci USA. 2011;108(19):8042–8047. doi: 10.1073/pnas.1105117108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Xiao W, et al. Imprinting of the MEA Polycomb gene is controlled by antagonism between MET1 methyltransferase and DME glycosylase. Dev Cell. 2003;5(6):891–901. doi: 10.1016/s1534-5807(03)00361-7. [DOI] [PubMed] [Google Scholar]
- 8.Morales-Ruiz T, et al. DEMETER and REPRESSOR OF SILENCING 1 encode 5-methylcytosine DNA glycosylases. Proc Natl Acad Sci USA. 2006;103(18):6853–6858. doi: 10.1073/pnas.0601109103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ortega-Galisteo AP, Morales-Ruiz T, Ariza RR, Roldán-Arjona T. Arabidopsis DEMETER-LIKE proteins DML2 and DML3 are required for appropriate distribution of DNA methylation marks. Plant Mol Biol. 2008;67(6):671–681. doi: 10.1007/s11103-008-9346-0. [DOI] [PubMed] [Google Scholar]
- 10.Gong Z, et al. ROS1, a repressor of transcriptional gene silencing in Arabidopsis, encodes a DNA glycosylase/lyase. Cell. 2002;111(6):803–814. doi: 10.1016/s0092-8674(02)01133-9. [DOI] [PubMed] [Google Scholar]
- 11.Slotkin RK, et al. Epigenetic reprogramming and small RNA silencing of transposable elements in pollen. Cell. 2009;136(3):461–472. doi: 10.1016/j.cell.2008.12.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Martínez G, Panda K, Köhler C, Slotkin RK. Silencing in sperm cells is directed by RNA movement from the surrounding nurse cell. Nat Plants. 2016;2:16030. doi: 10.1038/nplants.2016.30. [DOI] [PubMed] [Google Scholar]
- 13.Hsieh TF, et al. Genome-wide demethylation of Arabidopsis endosperm. Science. 2009;324(5933):1451–1454. doi: 10.1126/science.1172417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Kim M, et al. Temporal and spatial downregulation of Arabidopsis MET1 activity results in global DNA hypomethylation and developmental defects. Mol Cells. 2008;26(6):611–615. [PMC free article] [PubMed] [Google Scholar]
- 15.Grossniklaus U, Vielle-Calzada JP, Hoeppner MA, Gagliano WB. Maternal control of embryogenesis by MEDEA, a polycomb group gene in Arabidopsis. Science. 1998;280(5362):446–450. doi: 10.1126/science.280.5362.446. [DOI] [PubMed] [Google Scholar]
- 16.Köhler C, et al. Arabidopsis MSI1 is a component of the MEA/FIE Polycomb group complex and required for seed development. EMBO J. 2003;22(18):4804–4814. doi: 10.1093/emboj/cdg444. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Luo M, Bilodeau P, Dennis ES, Peacock WJ, Chaudhury A. Expression and parent-of-origin effects for FIS2, MEA, and FIE in the endosperm and embryo of developing Arabidopsis seeds. Proc Natl Acad Sci USA. 2000;97(19):10637–10642. doi: 10.1073/pnas.170292997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hehenberger E, Kradolfer D, Köhler C. Endosperm cellularization defines an important developmental transition for embryo development. Development. 2012;139(11):2031–2039. doi: 10.1242/dev.077057. [DOI] [PubMed] [Google Scholar]
- 19.Charité J, et al. Transducing positional information to the Hox genes: Critical interaction of cdx gene products with position-sensitive regulatory elements. Development. 1998;125(22):4349–4358. doi: 10.1242/dev.125.22.4349. [DOI] [PubMed] [Google Scholar]
- 20.Sessa G, Morelli G, Ruberti I. The Athb-1 and -2 HD-Zip domains homodimerize forming complexes of different DNA binding specificities. EMBO J. 1993;12(9):3507–3517. doi: 10.1002/j.1460-2075.1993.tb06025.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.O’Malley RC, et al. Cistrome and epicistrome features shape the regulatory DNA landscape. Cell. 2016;165(5):1280–1292. doi: 10.1016/j.cell.2016.04.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Drews GN, Koltunow AM. The female gametophyte. Arabidopsis Book. 2011;9:e0155. doi: 10.1199/tab.0155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gehring M, Missirian V, Henikoff S. Genomic analysis of parent-of-origin allelic expression in Arabidopsis thaliana seeds. PLoS One. 2011;6(8):e23687. doi: 10.1371/journal.pone.0023687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Penterman J, et al. DNA demethylation in the Arabidopsis genome. Proc Natl Acad Sci USA. 2007;104(16):6752–6757. doi: 10.1073/pnas.0701861104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Chan RL, Gago GM, Palena CM, Gonzalez DH. Homeoboxes in plant development. Biochim Biophys Acta. 1998;1442(1):1–19. doi: 10.1016/s0167-4781(98)00119-5. [DOI] [PubMed] [Google Scholar]
- 26.Palena CM, Gonzalez DH, Chan RL. A monomer-dimer equilibrium modulates the interaction of the sunflower homeodomain leucine-zipper protein Hahb-4 with DNA. Biochem J. 1999;341(Pt 1):81–87. [PMC free article] [PubMed] [Google Scholar]
- 27.Palena CM, Tron AE, Bertoncini CW, Gonzalez DH, Chan RL. Positively charged residues at the N-terminal arm of the homeodomain are required for efficient DNA binding by homeodomain-leucine zipper proteins. J Mol Biol. 2001;308(1):39–47. doi: 10.1006/jmbi.2001.4563. [DOI] [PubMed] [Google Scholar]
- 28.Portereiko MF, et al. AGL80 is required for central cell and endosperm development in Arabidopsis. Plant Cell. 2006;18(8):1862–1872. doi: 10.1105/tpc.106.040824. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Borges F, et al. Comparative transcriptomics of Arabidopsis sperm cells. Plant Physiol. 2008;148(2):1168–1181. doi: 10.1104/pp.108.125229. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Qin Y, et al. Penetration of the stigma and style elicits a novel transcriptome in pollen tubes, pointing to genes critical for growth in a pistil. PLoS Genet. 2009;5(8):e1000621. doi: 10.1371/journal.pgen.1000621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Boavida LC, Borges F, Becker JD, Feijó JA. Whole genome analysis of gene expression reveals coordinated activation of signaling and metabolic pathways during pollen-pistil interactions in Arabidopsis. Plant Physiol. 2011;155(4):2066–2080. doi: 10.1104/pp.110.169813. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Hyun Y, et al. The catalytic subunit of Arabidopsis DNA polymerase α ensures stable maintenance of histone modification. Development. 2013;140(1):156–166. doi: 10.1242/dev.084624. [DOI] [PubMed] [Google Scholar]
- 33.Yadegari R, et al. Mutations in the FIE and MEA genes that encode interacting polycomb proteins cause parent-of-origin effects on seed development by distinct mechanisms. Plant Cell. 2000;12(12):2367–2382. doi: 10.1105/tpc.12.12.2367. [DOI] [PMC free article] [PubMed] [Google Scholar]