Animals, fungi, and plants make use of two major epigenetic repressive modifications on histones that are destined to silence genes and transposable elements (TEs). The Polycomb repressive system applies the repressive trimethylation on lysine 27 of histone H3 (H3K27me3), which mainly serves to silence genes during defined stages of development (Mozgova and Hennig, 2015; Schuettengruber et al., 2017). In contrast, the repressive di- or trimethylation on lysine 9 of histone H3 (H3K9me2/3) is mainly associated with constitutively silenced heterochromatin, present in repeat-rich regions and TEs (Saksouk et al., 2015; Wendte and Schmitz, 2018). Repression mediated by the Polycomb system is transient and dynamically regulated, while repression mediated by H3K9me2/3 is generally stable during development (Mozgova and Hennig, 2015; Saksouk et al., 2015; Schuettengruber et al., 2017; Wendte and Schmitz, 2018). The establishment of H3K27me3 is mediated by the evolutionary conserved Polycomb repressive complex 2 (PRC2). In plants, there are distinct PRC2 complexes that are active during different stages of plant development (Mozgova and Hennig, 2015). In Arabidopsis, the EMBRYONIC FLOWER (EMF) and VERNALIZATION (VRN) PRC2 regulate sporophyte development, while the FERTILIZATION-INDEPENDENT SEED (FIS) PRC2 is active in the central cell of the female gametophyte and the descendent endosperm (Mozgova and Hennig, 2015). The establishment of H3K9me2 in Arabidopsis depends on the SU(VAR)3–9 homologous proteins (SUVHs), KRYPTONITE (or SUVH4), SUVH5 and SUVH6 that act in a feedback loop with CHG methylation (CHGm) established by the CHROMOMETHYLTRANSFERASE 3 (CMT3) (Kenchanmane Raju et al., 2019; Zhang et al., 2018). In animals and plants, H3K27me3 and H3K9me2 are generally exclusive repressive marks (Wiles and Selker, 2017); however, they co-occur in certain tissue types like the plant endosperm (Klosinska et al., 2016; Moreno-Romero et al., 2016; Weinhofer et al., 2010), as well as in specific organisms as filamentous fungi and bryophytes (Carlier et al., 2020; Montgomery et al., 2020). Furthermore, redistribution of H3K27me3 to heterochromatic regions occurs when constitutive heterochromatin is disrupted, indicating that specific features of heterochromatin prevent PRC2 recruitment (Deleris et al., 2012; Jamieson et al., 2016; Peters et al., 2003; Wiles and Selker, 2017).

Erasure of H3K27me3 and H3K9me2 is mediated by different types of JUMONJI lysine demethylases (KDMs) that belong to the KDM4 and KDM3 type, respectively (Cheng et al., 2020). RELATIVE OF EARLY FLOWERING 6 (REF6) and EARLY FLOWERING 6 (ELF6) function redundantly as major erasers of H3K27me3 during Arabidopsis development (Lu et al., 2011; Yan et al., 2018). The zinc-finger (ZnF) domains of REF6 recognize the sequence motif CTCTGYTY (Y = T or C) (Cui et al., 2016; Li et al., 2016). Binding of REF6 to this motif is negatively affected by non-CG methylation (Qiu et al., 2019), revealing cross-talk between different epigenetic modifications.

Seed development in flowering plants is initiated by a double fertilization event, where one of the sperm cells fertilizes the haploid egg, giving rise to the diploid embryo and the other sperm cell fertilizes the diploid central cell, initiating development of the triploid endosperm (Zhou and Dresselhaus, 2019). The endosperm regulates the transfer of nutrients to the developing embryo, similar as the placenta in mammals. (Costa et al., 2012; Lee et al., 2010; Linkies et al., 2009). The endosperm furthermore regulates seed dormancy, a process preventing germination of seeds under favourable conditions (Nonogaki, 2019). Both processes have been linked to genomic imprinting, an epigenetic phenomenon causing parent-of-origin specific gene expression (Batista and Kohler, 2020; Gehring and Satyaki, 2017; Piskurewicz et al., 2016). Imprinted genes are epigenetically modified during gametogenesis, establishing parental-specific expression patterns that are maintained after fertilization. Genomic imprinting leading to maternally biased expression is predominantly connected to DNA demethylation of maternal alleles mediated by the DNA glycosylase DEMETER (Gehring et al., 2009; Hsieh et al., 2011; Park et al., 2016). In contrast, paternally biased expression is predominantly connected to the presence of H3K27me3, H3K9me2, and CHGm on maternal alleles (Klosinska et al., 2016; Moreno-Romero et al., 2019; Moreno-Romero et al., 2016; Moreno-Romero et al., 2017). The FIS-PRC2 establishes H3K27me3 in the central cell, which is maintained on the maternal alleles after fertilization. Loss of FIS-PRC2 function impairs CHGm establishment, indicating that H3K27me3 is the primary imprinting mark and CHGm and H3K9me2 are recruited after H3K27me3 establishment (Moreno-Romero et al., 2019). The biological meaning of the co-occurance of H3K27me3, H3K9me2, and CHGm on the maternal alleles of paternally expressed genes (PEGs) is unclear. In this study, we reveal that genes marked by H3K27me3 on their maternal alleles are targeted by REF6 and become reactivated during germination. In contrast, REF6 fails to activate genes marked by H3K27me3, H3K9me2, and CHGm, consistent with the finding that DNA methylation prevents REF6 binding (Qiu et al., 2019). Among those genes with triple repressive marks was the ABA INSENSITIVE 3 (ABI3) gene, a known positive regulator of seed dormancy (Liu et al., 2013; Shu et al., 2016; Tian et al., 2020). Genes activated by REF6 are enriched for ethylene signaling, a process activated during germination. Consistently, loss of REF6 causes increased dormancy, as recently reported (Chen et al., 2020). Together, our study unveils a mechanism that employs the combination of different repressive epigenetic modifications to time a crucial developmental process.

In this study, we uncovered the molecular function and physiological role of the endosperm-specific triple modification of the maternal genome with H3K27me3/H3K9me2 and CHGm. We found that genes marked by only H3K27me3 became reactivated during seed germination, a process that depends on the H3K27me3 demethylase REF6. In contrast, genes marked by triple repressive modifications remained silenced during germination because REF6 binding is repelled by H3K9me2/CHGm, consistent with previous work (Qiu et al., 2019). Supporting the functional requirement of REF6 to initiate germination, loss of REF6 caused higher seed dormancy (Figure 3, Figure 3—figure supplement 2; Chen et al., 2020), connected with decreased expression of many germination-inducible genes. Loss of SUVH4/5/6 also increased dormancy (Figure 6—figure supplement 1; Zheng et al., 2012), which was associated with increased expression of potential REF6 target genes. Since H3K9me2 acts in a feedback loop with CHGm (Du et al., 2012; Jackson et al., 2002; Lindroth et al., 2001; Malagnac et al., 2002), loss of SUVH4/5/6 is expected to expose unmethylated binding sites for REF6 that can then be activated by REF6. Consistent with this model, loss of CMT3 also caused increased dormancy (Figure 6b, Figure 6—figure supplement 2). Interestingly, among those upregulated SUVH4/5/6 targets was the positive regulator of seed dormancy ABI3 (Liu et al., 2013), providing a possible explanation for the increased dormancy in suvh4/5/6. Consistent with this notion we found that downregulated genes with single H3K27me3 were enriched for ABI3 binding motifs. Nevertheless, the small fraction of germination-inducible genes with triple repressive marks suggests possible additional mechanisms removing triple epigenetic marks.

Our work furthermore revealed the connection between maternal-specific repressive marks, gene imprinting, and seed dormancy. Imprinted genes were previously shown to play a role in regulating seed germination and were connected to the maternal inheritance of seed dormancy levels (Iwasaki et al., 2019; Piskurewicz et al., 2016). The non-canonical RdDM pathway was found to target and silence the paternal alleles of the dormancy-related gene ALN1 (Iwasaki et al., 2019). Our data suggests that in addition to RdDM, maternally biased expression of dormancy-related genes also requires REF6, which activates maternal H3K27me3 marked genes. We propose that both mechanisms together ensure the maternal control of seed dormancy; while REF6 regulates the time of activation of certain genes, RdDM-mediated CHHm on the paternal alleles prevents their activation. Interestingly, the timing of REF6 action probably differs for distinct genes. While some genes, like ALN1, CYP707A1, and CYP707A3 (Chen et al., 2020) are likely activated in the central cell of the female gametophyte and therefore lack detactable H3K27me3 in the endosperm, other genes are activated during endoperm development. To the latter category belongs the previously reported gene CP1, which is maternally biased in dormant seeds and biallelically expressed in non-dormant seeds (Piskurewicz et al., 2016) and marked with single H3K27me3 on the maternal alleles (Moreno-Romero et al., 2016). In contrast to genes with single H3K27me3, genes with triple repressive marks on maternal alleles are not targeted by REF6, causing the maternal alleles to be silenced and the paternal alleles to be active. Genes with triple repressive marks on REF6-binding sites were upregulated in suvh4/5/6 and were paternally biased during germination. Although more detailed parental-specific DNA methylome analyses in germinating endosperm are required to precisely determine the molecular mechanism repressing the paternal alleles of genes with maternal H3K27me3 during germination, this data supports the model that the presence of single H3K27me3 and triple repressive marks determines whether genes are maternally or paternally biased during germination, respectively (Figure 8—figure supplement 5).

The SUVH4/5/6 genes are not imprinted (Hsieh et al., 2011; Moreno-Romero et al., 2019; Pignatta et al., 2014); however, the delay of germination in suvh4/5/6 was under maternal control (Figure 6a). This indicates that maternal-specific triple repressive marks are established before fertilization in the central cell and maintained after fertilization, consistent with previous data (Moreno-Romero et al., 2019). This suggests that maternal-specific triple epigenetic marks of H3K27me3/H3K9me2 and CHGm contribute to maternally controlled seed dormancy (Iwasaki et al., 2019; Piskurewicz et al., 2016). Likewise, also REF6 is not imprinted (Hsieh et al., 2011; Pignatta et al., 2014), but ref6 had a maternal effect (Figure 3b), supporting the notion that derepression of some dormancy-controlling genes starts in the central cell of the female gametophyte. Nevertheless, the fact that some REF6 targets have detectable levels of H3K27me3 in the endosperm and since we could partially complement the ref6 mutant by expressing REF6 in the micropylar domain of the endosperm (Figure 5a), strongly suggests that REF6 also acts after fertilization. In support of this view, homozygous ref6 mutants had a stronger delay in germination than maternal ref6 mutants (Figure 3b).

Our data furthermore uncovered a relationship among REF6, ethylene, and ABA. Consistent with previous work showing that loss of ABA2 does not completely suppress the delayed germination of ref6 after stratification (Chen et al., 2020), we found that the ABA biosynthesis inhibitor fluridone was not sufficient to suppress delayed germination phenotypes of ref6 (Figure 5—figure supplement 4). Similarly, the ethylene precursor ACC caused only a partial suppression of the germination delay. However, the combined treatment with fluridone and ACC completely suppressed the ref6 germination phenotype (Figure 5—figure supplement 4), supporting our and previous data revealing that REF6 is required to activate ABA catabolism and ethylene signaling through removal of H3K27me3. Since the activation of the ABA catabolism genes CYP707A1 and CYC707A3 likely occurs in the central cell while activation of ethylene signaling genes occurs during endosperm development, we propose that there is a two-step maternal control of germination by REF6. The first step occurs in the central cell and leads to the suppression of ABA signaling, while the second step occurs in the endosperm and leads to the activation of ethylene signaling. This two-step control of germination may ensure the best timing of germination. Understanding by which mechanism the timing of REF6 action is regulated will be subject of future investigations.

Previous data showed that loss of FIS-PRC2 function causes upregulation of genes with both H3K27me3 and H3K9me2 (Hsieh et al., 2011; Moreno-Romero et al., 2016), suggesting that establishment of H3K9me2 depends on H3K27me3. This idea was substantiated by data showing that deficiency of FIS-PRC2 in the central cell causes depletion of CHGm (Moreno-Romero et al., 2019), pointing that H3K9me2/CHGm depend on FIS-PRC2. Also data shown in this study are consistent with this scenario; we demonstrate that loss of H3K9me2/CHGm only caused an effect if H3K27me3 was removed by REF6, revealing that H3K27me3 is the primary repressive mark. Additionally, we found that maternal loss of SUVH4/5/6 was sufficient to cause delayed gemination (Figure 6a), while that of CMT3 was not (Figure 6c). One possible explanation is that the effects caused by loss of SUVH4/5/6 and CMT3 function differ. While H3K9me2 and CHGm are severely depleted in suvh4/5/6; loss of CMT3 mainly affects CHGm with only intermediate effects on H3K9me2 (Stroud et al., 2014; Underwood et al., 2018). Thus, genes with H3K27me3/H3K9me2 can likely be suppressed by CMT3 after fertilization.

A possible mechanism is that the FIS2-PRC2 complex or H3K27me3 recruits SUVH4/5/6 and that H3K9me2 recruits CMT3, suggesting the presence of a molecular mechanism that distinguishes target genes to be marked by single H3K27me3 or triple marks of H3K27me3/H3K9me2 and CHGm. To identify this distinguishing mechanism remains an important future endeavor.

Our data bear striking similarities to maternal H3K27me3-dependent non-canonical imprinting reported in mammals (Chen et al., 2019; Hanna et al., 2019; Inoue et al., 2017a; Inoue et al., 2017b). In non-canonical imprinting, maternal-specific H3K27me3 inherited from oocytes serves as the primary epigenetic mark for imprinted paternal-specific expression. Most of this non-canonical imprinting is transient and restricted to preimplantation embryos, with only some genes important for placental development remain imprinted in extraembryonic cells (Inoue et al., 2017a). Maintenance of non-canonical imprinting is mediated by DNA methylation, which is specifically applied on the maternal alleles of previously H3K27me3 marked regions (Chen et al., 2019; Hanna et al., 2019). Thus, similar to the findings reported in this study, genes regulated by non-canonical imprinting in mammals are transiently imprinted by inheriting maternal-specific H3K27me3 and only few genes remain imprinting by establishing DNA methylation as secondary imprint. Genomic imprinting has independently evolved in flowering plants and mammals; however, rather remarkably, similar molecular mechanisms have been coopted to regulate imprinted genes in the accessory, nutrient-transferring tissues.

The combined presence of H3K27me3 with the heterochromatic modification H3K9me3 was also reported to occur on TEs and repeat-rich regions in the protozoa Tetrahymena and Paramecium (Frapporti et al., 2019; Liu et al., 2007; Zhao et al., 2019) and the fungus Podospora anserine (Carlier et al., 2020). In P. anserine, the presence of double repressive modifications is restricted to TEs and repeat-rich regions, while H3K27me3 alone is located in genic regions, indicating a similar molecular distinction of constitutively silenced regions marked by double modifications from regions that require activation during specific developmental stages marked by H3K27me3 alone. It is tempting to speculate that the co-occurrence of H3K9me3 and H3K27me3 may prevent H3K27me3 demethylases to target and activate those regions, similar to the findings reported in this study.

In summary, in this study, we uncovered a new molecular mechanism leading to maternal-specific dormancy control that distinguishes genes marked by either H3K27me3 alone from genes marked by the generally antagonistic repressive epigenetic modifications, H3K27me3, H3K9me2, and CHGm. We uncovered how the H3K27me3 demethylase REF6 controls release of dormancy and revealed the underlying mechanism for the H3K9me2 methylases SUVH4/5/6 in controlling dormancy and propose that these regulators form a regulatory network through ABA and ethylene pathways. Our data provide new insights into how maternal-specific epigenetic modifications control dormancy and thus ensure timely germination.

All data generated in this study were deposited to the Gene Expression Omnibus and are available under the accession number GSE158907.

1. 1. Köhler C

   (2021) NCBI Gene Expression Omnibus

   ID GSE158907. Combinations of maternal-specific repressive epigenetic marks in the endosperm control seed dormancy.

   http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE158907

1. 1. Moreno-Romero J
   2. Jiang H
   3. Santos-González J
   4. Köhler C

   (2016) NCBI Gene Expression Omnibus

   ID GSE66585. Parental epigenetic asymmetry of PRC2-mediated histone modifications in the Arabidopsis endosperm.

   http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE66585
2. 1. Park K
   2. Kim MY
   3. Vickers M Park

   (2016) NCBI Gene Expression Omnibus

   ID GSE89789. DNA demethylation is initiated in the central cells of Arabidopsis and rice.

   http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE89789
3. 1. Dekkers BJ
   2. Pearce SP
   3. van Bolderen-Veldkamp M
   4. Marshall A
   5. Bassel GW
   6. King JR
   7. Wood AT
   8. Müller K
   9. Widera P
   10. Gilbert J
   11. Krasnogor N
   12. Leubner-Metzger G
   13. Holdsworth MJ
   14. Bentsink L

   (2013) NCBI Gene Expression Omnibus

   ID GSE41212. Arabidopsis thaliana seed germination timecourse.

   http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE41212
4. 1. Ibarra CA
   2. Feng X
   3. Schoft VK
   4. Hsieh TF
   5. Uzawa R
   6. Rodrigues JA
   7. Zemach A
   8. Chumak N
   9. Machlicova A
   10. Nishimura T
   11. Rojas D
   12. Fischer RL
   13. Tamaru H
   14. Zilberman D

   (2012) NCBI Gene Expression Omnibus

   ID GSE38935. Active DNA demethylation in plant companion cells reinforces transposon methylation in gametes.

   http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE38935

1. 1. Alonso JM
   2. Stepanova AN
   3. Leisse TJ
   4. Kim CJ
   5. Chen H
   6. Shinn P
   7. Stevenson DK
   8. Zimmerman J
   9. Barajas P
   10. Cheuk R
   11. Gadrinab C
   12. Heller C
   13. Jeske A
   14. Koesema E
   15. Meyers CC
   16. Parker H
   17. Prednis L
   18. Ansari Y
   19. Choy N
   20. Deen H
   21. Geralt M
   22. Hazari N
   23. Hom E
   24. Karnes M
   25. Mulholland C
   26. Ndubaku R
   27. Schmidt I
   28. Guzman P
   29. Aguilar-Henonin L
   30. Schmid M
   31. Weigel D
   32. Carter DE
   33. Marchand T
   34. Risseeuw E
   35. Brogden D
   36. Zeko A
   37. Crosby WL
   38. Berry CC
   39. Ecker JR

   (2003) Genome-wide insertional mutagenesis of Arabidopsis thaliana

   Science 301:653–657.

   https://doi.org/10.1126/science.1086391

   • Google Scholar
2. 1. Anders S
   2. Huber W

   (2010) Differential expression analysis for sequence count data

   Genome Biology 11:R106.

   https://doi.org/10.1186/gb-2010-11-10-r106

   • Google Scholar
3. 1. Bae H
   2. Sicher R
   3. Natarajan S
   4. Bailey B

   (2009) In situ expression of trehalose synthesizing genes, TPS1 and TPPB, in Arabidopsis thaliana using the GUS reporter gene

   Plant Cell, Tissue and Organ Culture (PCTOC 98:311–319.

   https://doi.org/10.1007/s11240-009-9565-3

   • Google Scholar
4. 1. Batista RA
   2. Kohler C

   (2020) Genomic imprinting in plants-revisiting existing models

   Genes & Development 34:24–36.

   https://doi.org/10.1101/gad.332924.119

   • Google Scholar
5. 1. Beaudoin N
   2. Serizet C
   3. Gosti F
   4. Giraudat J

   (2000) Interactions between abscisic acid and ethylene signaling cascades

   The Plant Cell 12:1103–1115.

   https://doi.org/10.1105/tpc.12.7.1103

   • Google Scholar
6. 1. Belmonte MF
   2. Kirkbride RC
   3. Stone SL
   4. Pelletier JM
   5. Bui AQ
   6. Yeung EC
   7. Hashimoto M
   8. Fei J
   9. Harada CM
   10. Munoz MD
   11. Le BH
   12. Drews GN
   13. Brady SM
   14. Goldberg RB
   15. Harada JJ

   (2013) Comprehensive developmental profiles of gene activity in regions and subregions of the Arabidopsis seed

   PNAS 110:E435–E444.

   https://doi.org/10.1073/pnas.1222061110

   • PubMed
   • Google Scholar
7. 1. Cao D
   2. Cheng H
   3. Wu W
   4. Soo HM
   5. Peng J

   (2006) Gibberellin mobilizes distinct della-dependent transcriptomes to regulate seed germination and floral development in arabidopsis

   Plant Physiology 142:509–525.

   https://doi.org/10.1104/pp.106.082289

   • PubMed
   • Google Scholar
8. Preprint

   1. Carlier F
   2. Debuchy R
   3. Maroc L
   4. Souaid C
   5. Noordermeer D
   6. Grognet P
   7. Malagnac F

   (2020) Loss of Polycomb Protein Ezh2 Causes Major Depletion of H3k27 and H3k9 Tri-Methylation and Developmental Defects in the Fungus Podospora Anserina

   bioRxiv.

   https://doi.org/10.1101/2020.08.21.261065

   • Google Scholar
9. 1. Chen W-H
   2. Li P-F
   3. Chen M-K
   4. Lee Y-I
   5. Yang C-H

   (2015) FOREVER YOUNG FLOWER negatively regulates ethylene response dna-binding factors by activating an ethylene-responsive factor to control arabidopsis floral organ senescence and abscission

   Plant Physiology 168:1666–1683.

   https://doi.org/10.1104/pp.15.00433

   • PubMed
   • Google Scholar
10. 1. Chen Z
    2. Yin Q
    3. Inoue A
    4. Zhang C
    5. Zhang Y

    (2019) Allelic h3k27me3 to allelic DNA methylation switch maintains noncanonical imprinting in extraembryonic cells

    Science Advances 5:eaay7246.

    https://doi.org/10.1126/sciadv.aay7246

    • PubMed
    • Google Scholar
11. 1. Chen HH
    2. Tong J
    3. Fu W
    4. Liang Z
    5. Ruan J
    6. Yu Y
    7. Song X
    8. Yuan L
    9. Xiao L
    10. Liu J
    11. Cui Y
    12. Huang S
    13. Li C

    (2020) The H3K27me3 Demethylase RELATIVE OF EARLY FLOWERING6 Suppresses Seed Dormancy by Inducing Abscisic Acid Catabolism

    Plant Physiology 184:1969–1978.

    https://doi.org/10.1104/pp.20.01255

    • PubMed
    • Google Scholar
12. 1. Cheng K
    2. Xu Y
    3. Yang C
    4. Ouellette L
    5. Niu L
    6. Zhou X
    7. Chu L
    8. Zhuang F
    9. Liu J
    10. Wu H
    11. Charron JB
    12. Luo M

    (2020) Histone tales: Lysine methylation, a protagonist in arabidopsis development

    Journal of Experimental Botany 71:793–807.

    https://doi.org/10.1093/jxb/erz435

    • PubMed
    • Google Scholar
13. 1. Chiu RS
    2. Nahal H
    3. Provart NJ
    4. Gazzarrini S

    (2012) The role of the arabidopsis fusca3 transcription factor during inhibition of seed germination at high temperature

    BMC Plant Biology 12:15.

    https://doi.org/10.1186/1471-2229-12-15

    • PubMed
    • Google Scholar
14. 1. Clough SJ
    2. Bent AF

    (1998) Floral dip: a simplified method for agrobacterium-mediated transformation of Arabidopsis thaliana

    The Plant Journal 16:735–743.

    https://doi.org/10.1046/j.1365-313x.1998.00343.x

    • PubMed
    • Google Scholar
15. 1. Corbineau F
    2. Xia Q
    3. Bailly C
    4. El-Maarouf-Bouteau H

    (2014) Ethylene, a key factor in the regulation of seed dormancy

    Frontiers in Plant Science 5:539.

    https://doi.org/10.3389/fpls.2014.00539

    • PubMed
    • Google Scholar
16. 1. Costa LM
    2. Yuan J
    3. Rouster J
    4. Paul W
    5. Dickinson H
    6. Gutierrez-Marcos JF

    (2012) Maternal control of nutrient allocation in plant seeds by genomic imprinting

    Current Biology 22:160–165.

    https://doi.org/10.1016/j.cub.2011.11.059

    • PubMed
    • Google Scholar
17. 1. Cui X
    2. Lu F
    3. Qiu Q
    4. Zhou B
    5. Gu L
    6. Zhang S
    7. Kang Y
    8. Cui X
    9. Ma X
    10. Yao Q
    11. Ma J
    12. Zhang X
    13. Cao X

    (2016) REF6 recognizes a specific DNA sequence to demethylate H3K27me3 and regulate organ boundary formation in Arabidopsis

    Nature Genetics 48:694–699.

    https://doi.org/10.1038/ng.3556

    • Google Scholar
18. 1. Das D
    2. St Onge KR
    3. Voesenek LA
    4. Pierik R
    5. Sasidharan R

    (2016) Ethylene- and Shade-Induced Hypocotyl Elongation Share Transcriptome Patterns and Functional Regulators

    Plant Physiology 172:718–733.

    https://doi.org/10.1104/pp.16.00725

    • Google Scholar
19. 1. Dekkers BJ
    2. Pearce S
    3. van Bolderen-Veldkamp RP
    4. Marshall A
    5. Widera P
    6. Gilbert J
    7. Drost HG
    8. Bassel GW
    9. Muller K
    10. King JR
    11. Wood AT
    12. Grosse I
    13. Quint M
    14. Krasnogor N
    15. Leubner-Metzger G
    16. Holdsworth MJ
    17. Bentsink L

    (2013) Transcriptional dynamics of two seed compartments with opposing roles in Arabidopsis seed germination

    Plant Physiology 163:205–215.

    https://doi.org/10.1104/pp.113.223511

    • Google Scholar
20. 1. Deleris A
    2. Stroud H
    3. Bernatavichute Y
    4. Johnson E
    5. Klein G
    6. Schubert D
    7. Jacobsen SE

    (2012) Loss of the DNA methyltransferase MET1 Induces H3K9 hypermethylation at PcG target genes and redistribution of H3K27 trimethylation to transposons in Arabidopsis thaliana

    PLOS Genetics 8:e1003062.

    https://doi.org/10.1371/journal.pgen.1003062

    • Google Scholar
21. 1. Du J
    2. Zhong X
    3. Bernatavichute YV
    4. Stroud H
    5. Feng S
    6. Caro E
    7. Vashisht AA
    8. Terragni J
    9. Chin HG
    10. Tu A
    11. Hetzel J
    12. Wohlschlegel JA
    13. Pradhan S
    14. Patel DJ
    15. Jacobsen SE

    (2012) Dual binding of chromomethylase domains to h3k9me2-containing nucleosomes directs DNA methylation in plants

    Cell 151:167–180.

    https://doi.org/10.1016/j.cell.2012.07.034

    • PubMed
    • Google Scholar
22. 1. Dubois M
    2. Van den Broeck L
    3. Inzé D

    (2018) The pivotal role of ethylene in plant growth

    Trends in Plant Science 23:311–323.

    https://doi.org/10.1016/j.tplants.2018.01.003

    • PubMed
    • Google Scholar
23. 1. Dubreucq B
    2. Berger N
    3. Vincent E
    4. Boisson M
    5. Pelletier G
    6. Caboche M
    7. Lepiniec L

    (2000) The Arabidopsis AtEPR1 extensin-like gene is specifically expressed in endosperm during seed germination

    The Plant Journal 23:643–652.

    https://doi.org/10.1046/j.1365-313x.2000.00829.x

    • Google Scholar
24. 1. Ebbs ML
    2. Bender J

    (2006) Locus-specific control of DNA methylation by the arabidopsis SUVH5 histone methyltransferase

    The Plant Cell 18:1166–1176.

    https://doi.org/10.1105/tpc.106.041400

    • PubMed
    • Google Scholar
25. 1. Endo A
    2. Tatematsu K
    3. Hanada K
    4. Duermeyer L
    5. Okamoto M
    6. Yonekura-Sakakibara K
    7. Saito K
    8. Toyoda T
    9. Kawakami N
    10. Kamiya Y
    11. Seki M
    12. Nambara E

    (2012) Tissue-specific transcriptome analysis reveals cell wall metabolism, flavonol biosynthesis and defense responses are activated in the endosperm of germinating Arabidopsis thaliana seeds

    Plant & Cell Physiology 53:16–27.

    https://doi.org/10.1093/pcp/pcr171

    • Google Scholar
26. 1. Feng J
    2. Meyer CA
    3. Wang Q
    4. Liu JS
    5. Shirley Liu X
    6. Zhang Y

    (2012) GFOLD: A generalized fold change for ranking differentially expressed genes from RNA-seq data

    Bioinformatics 28:2782–2788.

    https://doi.org/10.1093/bioinformatics/bts515

    • PubMed
    • Google Scholar
27. 1. Frapporti A
    2. Miro Pina C
    3. Arnaiz O
    4. Holoch D
    5. Kawaguchi T
    6. Humbert A
    7. Eleftheriou E
    8. Lombard B
    9. Loew D
    10. Sperling L
    11. Guitot K
    12. Margueron R
    13. Duharcourt S

    (2019) The Polycomb protein Ezl1 mediates H3K9 and H3K27 methylation to repress transposable elements in Paramecium

    Nature Communications 10:2710.

    https://doi.org/10.1038/s41467-019-10648-5

    • Google Scholar
28. 1. Gehring M
    2. Bubb KL
    3. Henikoff S

    (2009) Extensive demethylation of repetitive elements during seed development underlies gene imprinting

    Science 324:1447–1451.

    https://doi.org/10.1126/science.1171609

    • PubMed
    • Google Scholar
29. 1. Gehring M
    2. Satyaki PR

    (2017) Endosperm and Imprinting, Inextricably Linked

    Plant Physiology 173:143–154.

    https://doi.org/10.1104/pp.16.01353

    • Google Scholar
30. 1. Hanna CW
    2. Perez-Palacios R
    3. Gahurova L
    4. Schubert M
    5. Krueger F
    6. Biggins L
    7. Andrews S
    8. Colome-Tatche M
    9. Bourc’his D
    10. Dean W
    11. Kelsey G

    (2019) Endogenous retroviral insertions drive non-canonical imprinting in extra-embryonic tissues

    Genome Biology 20:225.

    https://doi.org/10.1186/s13059-019-1833-x

    • Google Scholar
31. 1. Hsieh TF
    2. Shin J
    3. Uzawa R
    4. Silva P
    5. Cohen S
    6. Bauer MJ
    7. Hashimoto M
    8. Kirkbride RC
    9. Harada JJ
    10. Zilberman D
    11. Fischer RL

    (2011) Regulation of imprinted gene expression in arabidopsis endosperm

    PNAS 108:1755–1762.

    https://doi.org/10.1073/pnas.1019273108

    • PubMed
    • Google Scholar
32. 1. Ibarra CA
    2. Feng X
    3. Schoft VK
    4. Hsieh TF
    5. Uzawa R
    6. Rodrigues JA
    7. Zemach A
    8. Chumak N
    9. Machlicova A
    10. Nishimura T
    11. Rojas D
    12. Fischer RL
    13. Tamaru H
    14. Zilberman D

    (2012) Active DNA demethylation in plant companion cells reinforces transposon methylation in gametes

    Science 337:1360–1364.

    https://doi.org/10.1126/science.1224839

    • PubMed
    • Google Scholar
33. 1. Inagaki S
    2. Miura-Kamio A
    3. Nakamura Y
    4. Lu F
    5. Cui X
    6. Cao X
    7. Kimura H
    8. Saze H
    9. Kakutani T

    (2010) Autocatalytic differentiation of epigenetic modifications within the arabidopsis genome

    The EMBO Journal 29:3496–3506.

    https://doi.org/10.1038/emboj.2010.227

    • PubMed
    • Google Scholar
34. 1. Inoue A
    2. Jiang L
    3. Lu F
    4. Suzuki T
    5. Zhang Y

    (2017a) Maternal H3K27ME3 controls DNA methylation-independent imprinting

    Nature 547:419–424.

    https://doi.org/10.1038/nature23262

    • PubMed
    • Google Scholar
35. 1. Inoue A
    2. Jiang L
    3. Lu F
    4. Zhang Y

    (2017b) Genomic imprinting of xist by maternal h3k27me3

    Genes & Development 31:1927–1932.

    https://doi.org/10.1101/gad.304113.117

    • PubMed
    • Google Scholar
36. 1. Iwasaki M
    2. Hyvärinen L
    3. Piskurewicz U
    4. Lopez-Molina L

    (2019) Non-canonical RNA-directed DNA methylation participates in maternal and environmental control of seed dormancy

    eLife 8:e37434.

    https://doi.org/10.7554/eLife.37434

    • PubMed
    • Google Scholar
37. 1. Jackson JP
    2. Lindroth AM
    3. Cao X
    4. Jacobsen SE

    (2002) Control of CPNPG dna methylation by the kryptonite Histone h3 methyltransferase

    Nature 416:556–560.

    https://doi.org/10.1038/nature731

    • PubMed
    • Google Scholar
38. 1. Jamieson K
    2. Wiles ET
    3. McNaught KJ
    4. Sidoli S
    5. Leggett N
    6. Shao Y
    7. Garcia BA
    8. Selker EU

    (2016) Loss of HP1 causes depletion of h3k27me3 from facultative heterochromatin and gain of h3k27me2 at constitutive heterochromatin

    Genome Research 26:97–107.

    https://doi.org/10.1101/gr.194555.115

    • PubMed
    • Google Scholar
39. 1. Kenchanmane Raju SK
    2. Ritter EJ
    3. Niederhuth CE

    (2019) Establishment, maintenance, and biological roles of non-CG methylation in plants

    Essays in Biochemistry 63:743–755.

    https://doi.org/10.1042/EBC20190032

    • PubMed
    • Google Scholar
40. 1. Klosinska M
    2. Picard CL
    3. Gehring M

    (2016) Conserved imprinting associated with unique epigenetic signatures in the Arabidopsis genus

    Nature Plants 2:16145.

    https://doi.org/10.1038/nplants.2016.145

    • PubMed
    • Google Scholar
41. 1. Kushiro T
    2. Okamoto M
    3. Nakabayashi K
    4. Yamagishi K
    5. Kitamura S
    6. Asami T
    7. Hirai N
    8. Koshiba T
    9. Kamiya Y
    10. Nambara E

    (2004) The Arabidopsis cytochrome P450 CYP707A encodes ABA 8’-hydroxylases: key enzymes in ABA catabolism

    The EMBO Journal 23:1647–1656.

    https://doi.org/10.1038/sj.emboj.7600121

    • PubMed
    • Google Scholar
42. 1. Lee KP
    2. Piskurewicz U
    3. Turecková V
    4. Strnad M
    5. Lopez-Molina L

    (2010) A seed coat bedding assay shows that rgl2-dependent release of abscisic acid by the endosperm controls embryo growth in arabidopsis dormant seeds

    PNAS 107:19108–19113.

    https://doi.org/10.1073/pnas.1012896107

    • PubMed
    • Google Scholar
43. 1. Li C
    2. Gu L
    3. Gao L
    4. Chen C
    5. Wei C-Q
    6. Qiu Q
    7. Chien C-W
    8. Wang S
    9. Jiang L
    10. Ai L-F
    11. Chen C-Y
    12. Yang S
    13. Nguyen V
    14. Qi Y
    15. Snyder MP
    16. Burlingame AL
    17. Kohalmi SE
    18. Huang S
    19. Cao X
    20. Wang Z-Y
    21. Wu K
    22. Chen X
    23. Cui Y

    (2016) Concerted genomic targeting of H3K27 demethylase REF6 and chromatin-remodeling atpase BRM in arabidopsis

    Nature Genetics 48:687–693.

    https://doi.org/10.1038/ng.3555

    • PubMed
    • Google Scholar
44. 1. Lindroth AM
    2. Cao X
    3. Jackson JP
    4. Zilberman D
    5. McCallum CM
    6. Henikoff S
    7. Jacobsen SE

    (2001) Requirement of CHROMOMETHYLASE3 for maintenance of CpXpG methylation

    Science 292:2077–2080.

    https://doi.org/10.1126/science.1059745

    • Google Scholar
45. 1. Linkies A
    2. Muller K
    3. Morris K
    4. Tureckova V
    5. Wenk M
    6. Cadman CS
    7. Corbineau F
    8. Strnad M
    9. Lynn JR
    10. Finch-Savage WE
    11. Leubner-Metzger G

    (2009) Ethylene interacts with abscisic acid to regulate endosperm rupture during germination: a comparative approach using Lepidium sativum and Arabidopsis thaliana

    The Plant Cell 21:3803–3822.

    https://doi.org/10.1105/tpc.109.070201

    • Google Scholar
46. 1. Linkies A
    2. Leubner-Metzger G

    (2012) Beyond gibberellins and abscisic acid: how ethylene and jasmonates control seed germination

    Plant Cell Reports 31:253–270.

    https://doi.org/10.1007/s00299-011-1180-1

    • Google Scholar
47. 1. Liu Y
    2. Taverna SD
    3. Muratore TL
    4. Shabanowitz J
    5. Hunt DF
    6. Allis CD

    (2007) RNAi-dependent H3K27 methylation is required for heterochromatin formation and DNA elimination in Tetrahymena

    Genes & Development 21:1530–1545.

    https://doi.org/10.1101/gad.1544207

    • PubMed
    • Google Scholar
48. 1. Liu X
    2. Zhang H
    3. Zhao Y
    4. Feng Z
    5. Li Q
    6. Yang H-Q
    7. Luan S
    8. Li J
    9. He Z-H

    (2013) Auxin controls seed dormancy through stimulation of abscisic acid signaling by inducing arf-mediated abi3 activation in arabidopsis

    PNAS 110:15485–15490.

    https://doi.org/10.1073/pnas.1304651110

    • PubMed
    • Google Scholar
49. 1. Lu F
    2. Cui X
    3. Zhang S
    4. Jenuwein T
    5. Cao X

    (2011) Arabidopsis ref6 is a histone h3 lysine 27 demethylase

    Nature Genetics 43:715–719.

    https://doi.org/10.1038/ng.854

    • PubMed
    • Google Scholar
50. 1. MacGregor DR
    2. Zhang N
    3. Iwasaki M
    4. Chen M
    5. Dave A
    6. Lopez-Molina L
    7. Penfield S

    (2019) ICE1 and ZOU determine the depth of primary seed dormancy in Arabidopsis independently of their role in endosperm development

    The Plant Journal 98:277–290.

    https://doi.org/10.1111/tpj.14211

    • Google Scholar
51. 1. Malagnac F
    2. Bartee L
    3. Bender J

    (2002) An Arabidopsis SET domain protein required for maintenance but not establishment of DNA methylation

    The EMBO Journal 21:6842–6852.

    https://doi.org/10.1093/emboj/cdf687

    • Google Scholar
52. 1. Martinez-Andujar C
    2. Ordiz MI
    3. Huang Z
    4. Nonogaki M
    5. Beachy RN
    6. Nonogaki H

    (2011) Induction of 9-cis-epoxycarotenoid dioxygenase in Arabidopsis thaliana seeds enhances seed dormancy

    PNAS 108:17225–17229.

    https://doi.org/10.1073/pnas.1112151108

    • Google Scholar
53. 1. Maruyama K
    2. Todaka D
    3. Mizoi J
    4. Yoshida T
    5. Kidokoro S
    6. Matsukura S
    7. Takasaki H
    8. Sakurai T
    9. Yamamoto YY
    10. Yoshiwara K
    11. Kojima M
    12. Sakakibara H
    13. Shinozaki K
    14. Yamaguchi-Shinozaki K

    (2012) Identification of cis-acting promoter elements in cold- and dehydration-induced transcriptional pathways in Arabidopsis, rice, and soybean

    DNA Research 19:37–49.

    https://doi.org/10.1093/dnares/dsr040

    • Google Scholar
54. 1. Montgomery SA
    2. Tanizawa Y
    3. Galik B
    4. Wang N
    5. Ito T
    6. Mochizuki T
    7. Akimcheva S
    8. Bowman JL
    9. Cognat V
    10. Marechal-Drouard L
    11. Ekker H
    12. Hong SF
    13. Kohchi T
    14. Lin SS
    15. Liu LD
    16. Nakamura Y
    17. Valeeva LR
    18. Shakirov EV
    19. Shippen DE
    20. Wei WL
    21. Yagura M
    22. Yamaoka S
    23. Yamato KT
    24. Liu C
    25. Berger F

    (2020) Chromatin Organization in Early Land Plants Reveals an Ancestral Association between H3K27me3, Transposons, and Constitutive Heterochromatin

    Current Biology 30:573–588.

    https://doi.org/10.1016/j.cub.2019.12.015

    • Google Scholar
55. 1. Moreno-Romero J
    2. Jiang H
    3. Santos-Gonzalez J
    4. Kohler C

    (2016) Parental epigenetic asymmetry of PRC2-mediated histone modifications in the Arabidopsis endosperm

    The EMBO Journal 35:1298–1311.

    https://doi.org/10.15252/embj.201593534

    • Google Scholar
56. 1. Moreno-Romero J
    2. Santos-González J
    3. Hennig L
    4. Köhler C

    (2017) Applying the INTACT method to purify endosperm nuclei and to generate parental-specific epigenome profiles

    Nature Protocols 12:238–254.

    https://doi.org/10.1038/nprot.2016.167

    • Google Scholar
57. 1. Moreno-Romero J
    2. Del Toro-De Leon G
    3. Yadav VK
    4. Santos-Gonzalez J
    5. Kohler C

    (2019) Epigenetic signatures associated with imprinted paternally expressed genes in the Arabidopsis endosperm

    Genome Biology 20:41.

    https://doi.org/10.1186/s13059-019-1652-0

    • Google Scholar
58. 1. Mozgova I
    2. Hennig L

    (2015) The Polycomb Group Protein Regulatory Network

    Annual Review of Plant Biology 66:269–296.

    https://doi.org/10.1146/annurev-arplant-043014-115627

    • Google Scholar
59. 1. Nonogaki H

    (2019) Seed germination and dormancy: The classic story, new puzzles, and evolution

    Journal of Integrative Plant Biology 61:541–563.

    https://doi.org/10.1111/jipb.12762

    • Google Scholar
60. 1. Park K
    2. Kim MY
    3. Vickers M
    4. Park JS
    5. Hyun Y
    6. Okamoto T
    7. Zilberman D
    8. Fischer RL
    9. Feng X
    10. Choi Y
    11. Scholten S

    (2016) DNA demethylation is initiated in the central cells of Arabidopsis and rice

    PNAS 113:15138–15143.

    https://doi.org/10.1073/pnas.1619047114

    • Google Scholar
61. 1. Peters AH
    2. Kubicek S
    3. Mechtler K
    4. O’Sullivan RJ
    5. Derijck AA
    6. Perez-Burgos L
    7. Kohlmaier A
    8. Opravil S
    9. Tachibana M
    10. Shinkai Y
    11. Martens JH
    12. Jenuwein T

    (2003) Partitioning and plasticity of repressive histone methylation states in mammalian chromatin

    Molecular Cell 12:1577–1589.

    https://doi.org/10.1016/s1097-2765(03)00477-5

    • Google Scholar
62. 1. Pignatta D
    2. Erdmann RM
    3. Scheer E
    4. Picard CL
    5. Bell GW
    6. Gehring M

    (2014) Natural epigenetic polymorphisms lead to intraspecific variation in Arabidopsis gene imprinting

    eLife 3:e03198.

    https://doi.org/10.7554/eLife.03198

    • Google Scholar
63. 1. Piskurewicz U
    2. Iwasaki M
    3. Susaki D
    4. Megies C
    5. Kinoshita T
    6. Lopez-Molina L

    (2016) Dormancy-specific imprinting underlies maternal inheritance of seed dormancy in Arabidopsis thaliana

    eLife 5:e19573.

    https://doi.org/10.7554/eLife.19573

    • Google Scholar
64. 1. Qiu Q
    2. Mei H
    3. Deng X
    4. He K
    5. Wu B
    6. Yao Q
    7. Zhang J
    8. Lu F
    9. Ma J
    10. Cao X

    (2019) DNA methylation repels targeting of Arabidopsis REF6

    Nature Communications 10:2063.

    https://doi.org/10.1038/s41467-019-10026-1

    • Google Scholar
65. Software

    1. R Development Core Team

    (2016) R: A language and environment for statistical computing, version 3.3. 2

    R Foundation for Statistical Computing, Vienna, Austria.

    http://www.r-project.org
66. 1. Saksouk N
    2. Simboeck E
    3. Dejardin J

    (2015) Constitutive heterochromatin formation and transcription in mammals

    Epigenetics & Chromatin 8:3.

    https://doi.org/10.1186/1756-8935-8-3

    • Google Scholar
67. 1. Schuettengruber B
    2. Bourbon HM
    3. Di Croce L
    4. Cavalli G

    (2017) Genome Regulation by Polycomb and Trithorax: 70 Years and Counting

    Cell 171:34–57.

    https://doi.org/10.1016/j.cell.2017.08.002

    • Google Scholar
68. 1. Shu K
    2. Liu X-D
    3. Xie Q
    4. He Z-H

    (2016) Two faces of one seed: Hormonal regulation of dormancy and germination

    Molecular Plant 9:34–45.

    https://doi.org/10.1016/j.molp.2015.08.010

    • PubMed
    • Google Scholar
69. 1. Stroud H
    2. Do T
    3. Du J
    4. Zhong X
    5. Feng S
    6. Johnson L
    7. Patel DJ
    8. Jacobsen SE

    (2014) Non-CG methylation patterns shape the epigenetic landscape in Arabidopsis

    Nature Structural & Molecular Biology 21:64–72.

    https://doi.org/10.1038/nsmb.2735

    • Google Scholar
70. 1. Supek F
    2. Bošnjak M
    3. Škunca N
    4. Šmuc T

    (2011) REVIGO summarizes and visualizes long lists of gene ontology terms

    PLOS ONE 6:e21800.

    https://doi.org/10.1371/journal.pone.0021800

    • PubMed
    • Google Scholar
71. 1. Tao Z
    2. Shen L
    3. Gu X
    4. Wang Y
    5. Yu H
    6. He Y

    (2017) Embryonic epigenetic reprogramming by a pioneer transcription factor in plants

    Nature 551:124–128.

    https://doi.org/10.1038/nature24300

    • PubMed
    • Google Scholar
72. 1. Tian T
    2. Liu Y
    3. Yan H
    4. You Q
    5. Yi X
    6. Du Z
    7. Xu W
    8. Su Z

    (2017) Agrigo v2.0: A GO analysis toolkit for the agricultural community, 2017 update

    Nucleic Acids Research 45:W122–W129.

    https://doi.org/10.1093/nar/gkx382

    • PubMed
    • Google Scholar
73. 1. Tian R
    2. Wang F
    3. Zheng Q
    4. Niza V
    5. Downie AB
    6. Perry SE

    (2020) Direct and indirect targets of the arabidopsis seed transcription factor ABSCISIC ACID insensitive3

    The Plant Journal 103:1679–1694.

    https://doi.org/10.1111/tpj.14854

    • PubMed
    • Google Scholar
74. 1. Trapnell C
    2. Pachter L
    3. Salzberg SL

    (2009) TopHat: discovering splice junctions with RNA-Seq

    Bioinformatics 25:1105–1111.

    https://doi.org/10.1093/bioinformatics/btp120

    • PubMed
    • Google Scholar
75. 1. Underwood CJ
    2. Choi K
    3. Lambing C
    4. Zhao X
    5. Serra H
    6. Borges F
    7. Simorowski J
    8. Ernst E
    9. Jacob Y
    10. Henderson IR
    11. Martienssen RA

    (2018) Epigenetic activation of meiotic recombination near Arabidopsis thaliana centromeres via loss of h3k9me2 and non-cg DNA methylation

    Genome Research 28:519–531.

    https://doi.org/10.1101/gr.227116.117

    • PubMed
    • Google Scholar
76. 1. Weinhofer I
    2. Hehenberger E
    3. Roszak P
    4. Hennig L
    5. Köhler C

    (2010) H3k27me3 profiling of the endosperm implies exclusion of polycomb group protein targeting by DNA methylation

    PLOS Genetics 6:10.

    https://doi.org/10.1371/journal.pgen.1001152

    • PubMed
    • Google Scholar
77. 1. Wendte JM
    2. Schmitz RJ

    (2018) Specifications of targeting heterochromatin modifications in plants

    Molecular Plant 11:381–387.

    https://doi.org/10.1016/j.molp.2017.10.002

    • PubMed
    • Google Scholar
78. 1. Wiles ET
    2. Selker EU

    (2017) H3K27 methylation: A promiscuous repressive chromatin mark

    Current Opinion in Genetics & Development 43:31–37.

    https://doi.org/10.1016/j.gde.2016.11.001

    • PubMed
    • Google Scholar
79. 1. Wuest SE
    2. Vijverberg K
    3. Schmidt A
    4. Weiss M
    5. Gheyselinck J
    6. Lohr M
    7. Wellmer F
    8. Rahnenfuhrer J
    9. von Mering C
    10. Grossniklaus U

    (2010) Arabidopsis female gametophyte gene expression map reveals similarities between plant and animal gametes

    Current Biology 20:506–512.

    https://doi.org/10.1016/j.cub.2010.01.051

    • PubMed
    • Google Scholar
80. 1. Yan W
    2. Chen D
    3. Smaczniak C
    4. Engelhorn J
    5. Liu H
    6. Yang W
    7. Graf A
    8. Carles CC
    9. Zhou DX
    10. Kaufmann K

    (2018) Dynamic and spatial restriction of polycomb activity by plant histone demethylases

    Nature Plants 4:681–689.

    https://doi.org/10.1038/s41477-018-0219-5

    • PubMed
    • Google Scholar
81. 1. Zhang H
    2. Lang Z
    3. Zhu JK

    (2018) Dynamics and function of DNA methylation in plants

    Nature Reviews. Molecular Cell Biology 19:489–506.

    https://doi.org/10.1038/s41580-018-0016-z

    • PubMed
    • Google Scholar
82. 1. Zhao X
    2. Xiong J
    3. Mao F
    4. Sheng Y
    5. Chen X
    6. Feng L
    7. Dui W
    8. Yang W
    9. Kapusta A
    10. Feschotte C
    11. Coyne RS
    12. Miao W
    13. Gao S
    14. Liu Y

    (2019) Rnai-dependent polycomb repression controls transposable elements in tetrahymena

    Genes & Development 33:348–364.

    https://doi.org/10.1101/gad.320796.118

    • PubMed
    • Google Scholar
83. 1. Zheng J
    2. Chen F
    3. Wang Z
    4. Cao H
    5. Li X
    6. Deng X
    7. Soppe WJJ
    8. Li Y
    9. Liu Y

    (2012) A novel role for histone methyltransferase kyp/suvh4 in the control of arabidopsis primary seed dormancy

    The New Phytologist 193:605–616.

    https://doi.org/10.1111/j.1469-8137.2011.03969.x

    • PubMed
    • Google Scholar
84. 1. Zhou LZ
    2. Dresselhaus T

    (2019) Friend or foe: Signaling mechanisms during double fertilization in flowering seed plants

    Current Topics in Developmental Biology 131:453–496.

    https://doi.org/10.1016/bs.ctdb.2018.11.013

    • PubMed
    • Google Scholar

Acceptance summary:

Sato et al. describe the role of histone modifications and DNA methylation in the control of seed dormancy in Arabidopsis. The authors show that DNA methylation associated with a specific histone modification (H3K9me2) prevents the REF6 enzyme from removing another histone modification (H3K27me3). Genes marked with both histone modifications are therefore resistant to activation, and this is important for gene regulation in the endosperm during seed

Decision letter after peer review:

Thank you for submitting your article "Combinations of maternal-specific repressive epigenetic marks in the endosperm control seed dormancy" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by Daniel Zilberman as the Reviewing Editor and Jürgen Kleine-Vehn as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Steven Penfield (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. We are also appending Steven Penfield's original review because we feel it may be helpful in drafting a revised manuscript, but only the revisions indicated in this letter need to be directly addressed.

Summary:

Sato et al. describe the role of histone marks and DNA methylation in the control of seed dormancy in Arabidopsis. This paper will be of interest to scientists working on epigenetics, seed development and germination. Most of the claims are supported by data and the authors have generated many genomic datasets and knowledge that could be useful for the study of seed dormancy in other plant species. The authors show that the REF6/H3K27me3 suvh456/H3K9me2 interaction is important for gene regulation in the endosperm during germination. The authors also investigate the role of REF6 in germination responses in seeds. This aspect of the study is intriguing, but less strongly supported and is in apparent disagreement with recently published conclusions.

Essential revisions:

1. The reviewers and editors felt that the most important revisions concern the narrative that REF6 and SUVH4-6 control dormancy via ethylene responses. The mechanism by which REF6 regulates dormancy is an essential component of the manuscript. However, the presented mechanism is not convincingly supported by the supplied data and differs from the conclusions of a recently published paper (Chen et al., Plant Physiology 2020), which implicates ABA catabolism during seed development. The authors should establish the relationship between REF6, ethylene and ABA signalling, and ideally reconcile their findings with those of Chen et al., 2020.

2. Related to the above point, seed dormancy should be assessed in the absence of dormancy-breaking stimuli, especially nitrate in the MS medium. Sowing freshly harvested seed on minimal media would result in higher quality data and be the normal approach in the field, showing that seed lots indeed show dormancy (germination <50% is better). Then the authors would clearly be able to show that dormancy rather than germination is affected.

3. The authors are not always clear on the mutant allele used in different experiments. A major caveat is that the ref6-1 and ref6-3 mutants used in germination experiments (Figure 3, Figure 3¬—figure supplement 2 and Figure 3-—figure supplement 3) and most of the molecular analyses are not a full knockout (null) mutants (Yan et al., 2019 and Antunez et al., 2020). The use of partial-loss-of-function REF6 alleles could be problematic, in particular for germination assays with seeds from reciprocal crosses. To address this issue, authors should use a loss-of-function mutant line ref6C, which they have used to generate the triple elf6-3 ref6C jmj13G mutant.

Similarly, the interpretations arising from the endosperm transcriptome analysis with ref6-1 mutant plants should be carefully considered in their conclusions.

4. The interpretation that suvh4/5/6 cause a maternal effect in germination due to the activity of these histone methylases in the central cell is not fully supported. Genetic analyses would be required to distinguish sporophytic and gametophytic effects of these mutants on seed germination. Similarly, the parent-of-origin effects on germination arising in the cmt3 mutant are far from clear. The role of suvh4/5/6 and cmt3 in parent-of-origin effects on germination should be critically assessed and described in the manuscript accordingly.

5. The expression of REF6 in the micropylar endosperm and embryo using EPR1::REF6 and TPS1::REF6 should be confirmed experimentally (eg. RT-PCR with dissected tissues).

6. The strong claims about the effects of CHGm on REF6 should be moderated. The model communicated by the authors involves a binary distinction between genes that only have H3K27me3 (accessible to REF6) and genes that also have H3K9me2 and CHGm (inaccessible to REF6). However, the data do not support such a strong conclusion. For example, Figure 2c shows significant but modest differences between control genes (about 90% are not upregulated) and genes with REF6 target domains that lack CHGm (about 80% not upregulated). Furthermore, because CHGm is not absolute and rarely reaches 50%, the effects of CHGm on REF6 recruitment in endosperm are likely to be quantitative rather than qualitative. These limitations should be clearly explained.

7. Statistical approaches are not always deployed appropriately. Throughout the manuscript, statistical comparisons are performed between multiple test groups and a single control group. This is appropriate when differences with the control group are key, but not appropriate when differences between the test groups are of primary interest. In such cases, the authors should use ANOVA or a similar approach that can simultaneously evaluate multiple groups.

An example of this issue is in Figure 2c. The lower four bars indicate that genes without CHGm are significantly upregulated, but not genes with CHGm. However, there are many fewer CHGm genes. The key issue is whether genes with and without CHGm behave differently from each other. A similar example is in Figure 5b. Genes with and without CHGm are likely not different from each other, but genes with CHGm do not reach significance in comparison to the control group because they are fewer (116) than genes without CHGm (488). Figure 8a also suffers from this problem. The top three blue boxes are marked as significant, and the following three grey boxes are not, but the blue and grey boxes are probably not significantly different from each other. The authors should carefully reassess the statistical comparisons and adjust their conclusions accordingly.

8. The regulation of genomic imprinting during germination by REF6 and SUVH4/5/6 is supported by a small number of genes (Figure 8). Similarly, the silencing of paternal alleles of maternally-expressed genes is linked to a very small level of CHH methylation on genes. The authors should acknowledge that regulation of genomic imprinting during germination by REF6 and SUVH4/5/6 is not fully demonstrated by the data shown, and the authors should moderate their claims accordingly.

Reviewer #2:

The study by Sato et al. examines the role of chromatin remodelling in the endosperm during dormancy and germination control. The challenge with any study if this kind is that individual epigenetic processes are very general, required at multiple stages for the regulation of seed development, germination and associated metabolism. When a phenotype is observed in mutants lacking a chromatin remodelling process this may be a product of multiple effects, or specific effects on a limited number of loci at one developmental stage. In this study the authors show that the REF6/H3K27me3 suvh456/H3K9me2 interaction is important for gene regulation in the endosperm during germination. I find this interesting and the authors understandably would like to expand this to show that functionally REF6 is required for germination responses in seeds. In my view this aspect of the study is less successful, perhaps reflecting the fact that the authors are world leaders in the field of plant epigenetics but less experienced in the study of seed physiology.

1. Lets start with reviewing what I believe is shown to the highest degree of confidence. In my view the data strongly support the conclusion that the ref6 mutants have germination phenotype which is conferred through endosperm-specific REF6 activity. Although the phenotypes are very weak, the authors take care to generate consistent data across experiments. Cmt3 and suvh456 mutants seem to have a maternal effect on dormancy consistent with previous studies of the maternal roles of these genes in the endosperm. The interaction between H3K27me3, H3K9me2, REF6 and CHG methylation is clearly described on the context of chromatin remodelling during germination, and I agree looks similar in seeds to that shown in a recent study (Qi et al., 2019). It seems clear that REF6-mediated epigenetic remodelling plays a role in the induction of ethylene responsive genes but also for genes in cell wall remodelling, lipid metabolism and other processes known to occur in the endosperm during germination, perhaps in response to ethylene. I would contrast the function of REF6 with that of PICKLE (Ogas et al., 1998) which is required for silencing of the seed expressed genes upon the transition to vegetative growth.

2. There is one experimental issue that make some of the data hard to interpret. The authors would ideally take more care to distinguish clearly between dormancy and germination of non-dormant seeds. Firstly, the dormancy assays are confounded by the fact that a nitrate source was added to the germination medium, a strong dormancy-breaking signal. As such all seed lots behave as if non-dormant in all the experiments, although a difference in germination speed between mutant and wt is recorded. It is not clear from the data provided whether without nitrate the seeds would actually show dormancy, but my feeling is that they would, but only then can it be studied. Certainly, where seeds lots germinate 100% after a few days of imbibition the field would not ascribe any dormancy to them, and would not usually use such experiments to make judgements about dormancy or use the word dormancy to describe the phenotypes. However, the authors rightly intuit that differences in germination speed can result from residual activity of dormancy-inducing pathways even if they have insufficient activity to completely prevent germination (Bassel and Finch-Savage, 2016). Evidence that this is happening is that germination speed is further increased by cold stratification, another dormancy-breaking signal. However, sowing freshly harvested seed on minimal media would result in higher quality data and be the normal approach in the field, showing that seed lots indeed show dormancy (germination <50% is better). Then the authors would clearly be able to show that dormancy rather than germination is affected.

3. The main implied narrative in the abstract that these marks cooperate to regulate dormancy loss through reactivation of maternal copies of ethylene response genes is only partly supported by the data in my view. There is no evidence presented to show that only maternal copies are activated, or describing how imprinting affects the activation rate. So the wider significance was less clear, in contrast to Piskurewicz 2016 who showed that dormancy status affected the degree of imprinting, and that imprinting of these genes affected the rate of protein catabolism or dormancy loss.

4. In my view more supporting evidence is required to show that REF6 directly affects endosperm responses to ethylene. I would inhibit germination with paclobutrazol to remove secondary effects and apply ethylene or a precursor, comparing the transcriptional response of wt and ref6 seeds.

5. If REF6 does affect the ethylene response, more evidence is required to show that this is linked to dormancy loss rather than germination promotion. The latter will occur in both dormant and non dormant seeds, and the seeds examined here are either non-dormant or have unknown dormancy because of the problems mentioned previously. The data in figure 1 only show an increase in the expression of H3K27me3-marked ethylene-inducible genes after testa rupture, ie after the germination process has already begun. This also suggests that the process is linked to the mechanics of germination rather than dormancy. This ought to be manifest in germination speed phenotypes even in non-dormant seeds, eg in a ref6 aba2 double mutant. Conversely, if REF6 is involved in dormancy loss maybe there is no dormancy phenotype until a dormancy-breaking signal is applied.

6. Timing of REF6 activity. Primary dormancy is established during seed development and seed maturation. Thus if seeds have different levels of primary dormancy, with some exceptions the cause is usually found in the processes that occur prior to desiccation. It is unclear when the EPR1:REF6 construct is expressed, whether it is also active prior to desiccation when it may affect the establishment of dormancy, rather than dormancy loss during germination. The unfortunate recent publication (Chen et al., 2020) which shows that REF6 affects ABA metabolism during seed maturation also seems to me to be a plausible mechanism for the ref6 dormancy phenotype. However, disagreement over the cause of the phenotype does not necessarily detract from the fact that REF6 is clearly affecting gene expression in the endosperm during germination, it just means that the significance of the process is less clear for the observed phenotype.

7. Also influencing my conclusions is the observation that the suvh456 effect is not limited to the endosperm but also observable in the embryo, where ABI3 is also higher expressed than wt (Figure 7d). This suggests that suvh456 mutant seeds have higher ABA levels or some global effect on dormancy that affects the whole seed and is capable of up-regulating ABI3 expression. I found the authors' discussion of ABI3 hard to follow. It is surprising that ABI3 reaches maximal expression in the mature seed, but is marked with the repressive H3K27me3. If I understand correctly the authors suggest that in suvh456 ABI3 is further activated by REF6. Again this is surprising because their seeds are germinating and during germination ABI3 expression falls rapidly (if the seeds were dormant then this would be more plausible). So it is possible that ABI3 regulation is several steps removed from the activity of the epigenetic process, or that REF6 acts on ABI3 earlier in seed development. Recent paper Chen et al., 2020 seems to shed some light on this, showing that ref6 seeds have higher ABA content. This would lead to an increase in dormancy and higher ABI3 expression, even in the embryo, because endospermic ABA is transported to the embryo to maintain dormancy (Kang et al., 2015).

So in conclusion I think the authors have discovered an interesting process which could well be very important for gene regulation during germination in the endosperm, but have not convinced me that the observed phenotype is due to disruption of this process. However, there may be other consequences that are observable if endosperm responses to ethylene are investigated in more detail, rather than focussing on the effect on primary dormancy. Alternatively, if the authors wish to explain the germination phenotype it may be worth expanding the study to events during the establishment of primary dormancy, taking into account the recent published observations of Chen et al., 2020.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Combinations of maternal-specific repressive epigenetic marks in the endosperm control seed dormancy" for further consideration by eLife. Your revised article has been evaluated by Jürgen Kleine-Vehn (Senior Editor), Daniel Zilberman (Reviewing Editor) and the original reviewers.

The manuscript has been greatly improved and the reviewers and editors are generally satisfied that the revisions address the majors points raised after initial review. A single important remaining issue remains, as explained in the three related major points below. Please consider these points and address them in the manuscript as appropriate.

1. Although they germinate slower than wild type, ref6 mutant response to ACC is similar to wt in terms of the type of germination increase promoted. Certainly ref6 does not appear to have low sensitivity to ACC. Therefore the authors' data suggest that ethylene responses are normal in ref6 mutants. The authors conclude that because ref6 germination deficiency is reversed by ACC (with fluridone), ref6 must be acting through ethylene signalling. How the authors reached this conclusion is not clear.

2. The authors may not be correctly interpretating the aba2 rescue of ref6 in Chen 2020. Chen 2020 show that aba2 more-or-less fully rescues the ref6 phenotype, and indeed any other finding would be very surprising. There are small differences but these are well within normal biological variation. The issue with the fluridone experiment is that fluridone inhibits ABA synthesis in imbibed seeds, and cyp707a2 is important for ABA catabolism in imbibed seeds. But according to Chen et al. REF6 mainly affects CYP707A1 and A3, which are active primarily during seed maturation, and therefore affect germination through the amount of ABA left over in mature seeds (Okamoto et al., 2006). In that context the failure of fluridone to completely suppress the ref6 phenotype, if caused by up-regulation of CYP707A1/3, is unsurprising.

3. New genetic data is presented that shows that the activity of REF6 in the central cell is sufficient to lead to most of the phenotypic effects. The H3K27me3 data used for figure 1 is from developing endosperm, after REF6 could have acted in the gametophyte. But the authors argue that REF6 is affecting de-methylation during dormancy loss or germination, or at least after the stage tissue was collected for the 2016 paper. So is the activity of REF6 described in the rest of the paper the weak phenotype attributed to the paternal effect? Could the differences (which are small) be simply due to gene dosage effects in the endosperm, if REF6 is a rate limiting factor, even if REF6 is not imprinted? If dormancy is determined by central cell REF6 activity, then it can't relate to the effects on imprinted gene expression during germination.

If the above is true, the statement in the manuscript title is mis-leading, because dormancy is determined by central cell ref6 function – the events during germination are clearly affected by REF6, but lack an obvious phenotypic outcome (in Col-0 at least). This is also pertinent to the normal ACC response in ref6 mutants. A clear statement on this issue is hence necessary.

Essential revisions:

1. The reviewers and editors felt that the most important revisions concern the narrative that REF6 and SUVH4-6 control dormancy via ethylene responses. The mechanism by which REF6 regulates dormancy is an essential component of the manuscript. However, the presented mechanism is not convincingly supported by the supplied data and differs from the conclusions of a recently published paper (Chen et al., Plant Physiology 2020), which implicates ABA catabolism during seed development. The authors should establish the relationship between REF6, ethylene and ABA signalling, and ideally reconcile their findings with those of Chen et al., 2020.

To address this concern, we performed several experiments and in silico analyses. First, to test the contribution of the ethylene and ABA pathways to the delayed germination phenotypes in ref6, we analyzed germination of ref6 upon treatment with the ethylene precursor 1-aminocyclopropane-1carboxylic acid (ACC), and the ABA biosynthesis inhibitor fluridone, which were reported to induce seed germination (Linkies et al., 2009; Martinez-Andujar et al., 2011). Treatment with only ACC or fluridone failed to completely suppress the delayed germination of ref6 (Figure 5—figure supplement 4), consistent with previous data showing that loss of the ABA biosynthetic gene ABA2 does not completely rescue the delayed germination phenotype of ref6 after stratification (Chen et al., 2020). In contrast, fluridone could completely suppress the cyp707a2 germination phenotype (Figure 5—figure supplement 4), revealing that fluridone can effectively suppress ABA biosynthesis as previously shown (Martinez-Andujar et al., 2011). Supporting our transcriptome data showing that ethylene pathways were suppressed in ref6 (Figure 5c), we found that treatment with both, ACC and fluridone, could completely rescue the delayed germination phenotype of ref6 after stratification(Figure 5— figure supplement 4). We thus conclude that suppression of the ethylene pathway together with the previously revealed induction of the ABA pathway (Chen et al., 2020) account for the delay of germination in ref6. We included and discussed this new data on P11, L243 to L261.

Second, we have further investigated the contribution of SUVH4/5/6 in regulating germination. As shown in our previous version of the manuscript, the ABI3 gene has triple repressive marks on the maternal alleles and was upregulated in suvh456 endosperm during germination (Figure 7). We furthermore found that genes with single H3K27me3 tend to be suppressed in suvh456 endosperm probably through indirect effects (Figure 7—figure supplement 2). In the revised manuscript we include data showing that B3 type transcription factors-binding motifs (RY motifs) are enriched among the downregulated genes with single H3K27me3 in suvh456 (Figure 7—figure supplement 5). This data suggests that upregulated ABI3 in suvh456 suppresses genes with single H3K27me3. Supporting this idea, we found that genes with single H3K27me3 are significantly enriched among direct ABI3 target genes (Tian et al., 2020), while genes with single H3K9me2 and double H3K27me3/H3K9me2 are not enriched (Figure 7—figure supplement 5). Together, this data reveals a complex network between ABA and ethylene pathways in the endosperm that are both epigenetically and transcriptionally regulated by multiple factors. We have included this new data on P14, L332 to L338 and discussed on P17, L417 to L418, P18, L431 to L437, P19, L461 to L476 and P21, L525 to L526.

2. Related to the above point, seed dormancy should be assessed in the absence of dormancy-breaking stimuli, especially nitrate in the MS medium. Sowing freshly harvested seed on minimal media would result in higher quality data and be the normal approach in the field, showing that seed lots indeed show dormancy (germination <50% is better). Then the authors would clearly be able to show that dormancy rather than germination is affected.

To address this comment, we performed analyses of seed dormancy on minimal media (Figure 3figure supplement 3a, Figure 6—figure supplement 1a and Figure 6—figure supplement 2).

In addition to that, we further confirmed seed dormancy phenotypes in ref6 single and suvh456 triple mutant seeds that were developed under low temperature (Figure 3—figure supplement 3b, Figure 6figure supplement 1c). We included this new data on P8, L177 and P12, L276.

3. The authors are not always clear on the mutant allele used in different experiments. A major caveat is that the ref6-1 and ref6-3 mutants used in germination experiments (Figure 3, Figure 3¬—figure supplement 2 and Figure 3—figure supplement 3) and most of the molecular analyses are not a full knockout (null) mutants (Yan et al., 2019 and Antunez et al., 2020). The use of partial-loss-of-function REF6 alleles could be problematic, in particular for germination assays with seeds from reciprocal crosses. To address this issue, authors should use a loss-of-function mutant line ref6C, which they have used to generate the triple elf6-3 ref6C jmj13G mutant.

Similarly, the interpretations arising from the endosperm transcriptome analysis with ref6-1 mutant plants should be carefully considered in their conclusions.

To address this concern, we performed several new experiments and in silico analyses that support our conclusions reached with ref6-1 and ref6-3 alleles (see figures below).

In ref6-1/ref6-3 alleles, the T-DNAs are inserted into the coding region between the JumonjiC domain and Zinc-finger domain, and weak expression of REF6 mRNA was detected (Yan et al., 2019). This raises the possibility that the Zinc-finger domain-independent functions of REF6 still remain active in ref6-1 or ref6-3. It was previously shown that REF6 can be recruited by MADS-box transcription factors independently of the REF6 binding domain (Yan et al., 2019).

1. We obtained the ref6c null allele and compared the phenotype with ref6-1 and ref6-3 alleles. (Additional figure 1a), and also analyzed the parental-specific effects of ref6c (Additional figure 1b). The results show that the phenotypes of the mutant alleles are highly comparable, justifying the use of the ref6-1 and ref6-3 alleles for this study.

2. We investigated the possibility that REF6 is recruited to MADS-box motifs by analyzing whether predicted REF6 targets have REF6 binding sites. As shown in the manuscript, the presence of REF6binding sites is associated with the activation of genes with single H3K27me3 and double H3K27me3/H3K9me2 (Figure 2c). Consistently, REF6-binding sites are significantly enriched within gene bodies of upregulated genes with single H3K27me3 and double H3K27me3/H3K9me2 (additional figure 1c, left panel), while MADS-box binding sites on promoter regions are not enriched among the upregulated genes (additional figure 1c, right panel). Furthermore, we found that there were substantially fewer MADS-box binding sites (around 10%) than REF6-binding sites (40-60%) among upregulated genes. This data suggests that the possible recruitment of REF6 by MADS-box transcription factors in ref6-1 and ref6-3 is unlikely to have a large contribution to total gene activation. This data are in support with our observations that the ref6-1/ref6-3 alleles behave very similar to the ref6c allele and justify their use in this study.

4. The interpretation that suvh4/5/6 cause a maternal effect in germination due to the activity of these histone methylases in the central cell is not fully supported. Genetic analyses would be required to distinguish sporophytic and gametophytic effects of these mutants on seed germination. Similarly, the parent-of-origin effects on germination arising in the cmt3 mutant are far from clear. The role of suvh4/5/6 and cmt3 in parent-of-origin effects on germination should be critically assessed and described in the manuscript accordingly.

To address this concern, we analyzed the gametophytic effect of suvh456 on germination using F1 seeds generated by pollinating heterozygous suvh456 mutants with wild-type pollen. The results support the conclusion that suvh456 caused delayed germination through gametophytic effects (Figure 6b). We described this new data on P12, L287 to L290.

Additionally, we analyzed the cause of the maternal effect of ref6 by analyzing the phenotypes of F1 seeds generated by pollinating heterozygous ref6-1 with wild-type pollen. The results support the conclusion that the maternal effect of ref6 is caused by a gametophytic effect (Figure 3c). We described this new data on P8, L183 to P9, L191.

The cmt3-11 mutant did not have a maternal effect on germination (Figure 6d), indicating that CMT3 acts after fertilization on H3K9me2 marks established in the female gametophyte. One possible explanation for the difference between suvh456 and cmt3 mutants is different effect on H3K9me2; while loss of SUVH4/5/6 causes severe depletion of both, H3K9me2 and CHGm (Stroud et al., 2014; Underwood et al., 2018), loss of CMT3 causes depletion of CHGm but has only intermediate effects on on H3K9me2 (Stroud et al., 2014; Inagaki et al., 2010). This suggests that genes with H3K27me3/H3K9me2 are suppressed by the paternal CMT3 after fertilization. We have discussed this data on P13, L298 to L301 and P20, L485 to L489.

5. The expression of REF6 in the micropylar endosperm and embryo using EPR1::REF6 and TPS1::REF6 should be confirmed experimentally (eg. RT-PCR with dissected tissues).

We confirmed the tissue-specific expression of REF6 in EPR1::REF6;ref6-1 and TPS1::REF6;ref6-1 (Figure 4—figure supplement 1).

6. The strong claims about the effects of CHGm on REF6 should be moderated. The model communicated by the authors involves a binary distinction between genes that only have H3K27me3 (accessible to REF6) and genes that also have H3K9me2 and CHGm (inaccessible to REF6). However, the data do not support such a strong conclusion. For example, Figure 2c shows significant but modest differences between control genes (about 90% are not upregulated) and genes with REF6 target domains that lack CHGm (about 80% not upregulated). Furthermore, because CHGm is not absolute and rarely reaches 50%, the effects of CHGm on REF6 recruitment in endosperm are likely to be quantitative rather than qualitative. These limitations should be clearly explained.

To address this concern, we moderated our claim about the relationship between CHGm directly on REF6 binding sites and REF6 accessibility. As the editor suggests about 10% of genes with H3K27me3/H3K9me2 and CHGm on REF6 binding sites are induced during germination, suggesting unknown mechanisms can remove the triple repressive marks or paternal alleles of some of those genes are activated during germination. It will be interesting to study whether and how the strong suppressing mechanism can be released. More detailed analyses of individual genes with triple repressive marks will be required to reveal the complete molecular mechanisms establishing and removing triple repressive marks. We have discussed these points on P17, L418 to L420.

7. Statistical approaches are not always deployed appropriately. Throughout the manuscript, statistical comparisons are performed between multiple test groups and a single control group. This is appropriate when differences with the control group are key, but not appropriate when differences between the test groups are of primary interest. In such cases, the authors should use ANOVA or a similar approach that can simultaneously evaluate multiple groups.

An example of this issue is in Figure 2c. The lower four bars indicate that genes without CHGm are significantly upregulated, but not genes with CHGm. However, there are many fewer CHGm genes. The key issue is whether genes with and without CHGm behave differently from each other. A similar example is in Figure 5b. Genes with and without CHGm are likely not different from each other, but genes with CHGm do not reach significance in comparison to the control group because they are fewer (116) than genes without CHGm (488). Figure 8a also suffers from this problem. The top three blue boxes are marked as significant, and the following three grey boxes are not, but the blue and grey boxes are probably not significantly different from each other. The authors should carefully reassess the statistical comparisons and adjust their conclusions accordingly.

To address this concern, we reanalyzed our data using pairwise statistical tests with multiple testing correction. The statistical methods are described in the legend of each figure.

8. The regulation of genomic imprinting during germination by REF6 and SUVH4/5/6 is supported by a small number of genes (Figure 8). Similarly, the silencing of paternal alleles of maternally-expressed genes is linked to a very small level of CHH methylation on genes. The authors should acknowledge that regulation of genomic imprinting during germination by REF6 and SUVH4/5/6 is not fully demonstrated by the data shown, and the authors should moderate their claims accordingly.

To address this concern, we moderated our claims and specifically noted that further studies are required to validate the conclusions. Modifications are on P16, L376 to L377, L395 to L397 and P18, L441 to L443.

We also reanalyzed the DNA methylation data using pairwise Wilcoxon test and included up-and downstream regions in the analyses (Figure 8—figure supplement 2, 3 and 4). Also for this analyses we toned down the conclusions and included the modifications on P16, L383 to L397.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

1. Although they germinate slower than wild type, ref6 mutant response to ACC is similar to wt in terms of the type of germination increase promoted. Certainly ref6 does not appear to have low sensitivity to ACC. Therefore the authors' data suggest that ethylene responses are normal in ref6 mutants. The authors conclude that because ref6 germination deficiency is reversed by ACC (with fluridone), ref6 must be acting through ethylene signalling. How the authors reached this conclusion is not clear.

We agree with the reviewer that the ref6 mutant can respond to ethylene; however, also ethylene biosynthesis is negatively affected in ref6 mutants (Figure 5—figure supplement 3); thus by adding ACC, ethylene production likely increases and thus ethylene signalling. Furthermore, since ethylene biosynthesis and signalling are in a feedback loop and ethylene was shown to directly regulate its biosynthesis (doi.org/10.1111/nph.16873), it is well possible that by adding ACC and thus highly increasing ethylene levels, ethylene signalling is enhanced. Since the combined treatment of fluridone and ACC completely suppressed the ref6 mutant phenotype, we consider our interpretation that REF6 acts (as well) through ethylene signalling justified. We discussed the requirement of REF6 to induce ethylene biosynthesis on P11, L239 to P11, L244.

2. The authors may not be correctly interpretating the aba2 rescue of ref6 in Chen 2020. Chen 2020 show that aba2 more-or-less fully rescues the ref6 phenotype, and indeed any other finding would be very surprising. There are small differences but these are well within normal biological variation. The issue with the fluridone experiment is that fluridone inhibits ABA synthesis in imbibed seeds, and cyp707a2 is important for ABA catabolism in imbibed seeds. But according to Chen et al. REF6 mainly affects CYP707A1 and A3, which are active primarily during seed maturation, and therefore affect germination through the amount of ABA left over in mature seeds (Okamoto et al., 2006). In that context the failure of fluridone to completely suppress the ref6 phenotype, if caused by up-regulation of CYP707A1/3, is unsurprising.

The data from Chen et al., 2000 show that the ref6 aba2 double mutant has lower germination rates than the aba2 single mutant after stratification (Figure 2C). While the difference has not been statistically evaluated, a germination difference of 50% versus 90% at 1.5 DAP seems to be highly significant. Since the data are generated in biological triplicates; we consider it unlikely that the difference is solely due to biological variation.

3. New genetic data is presented that shows that the activity of REF6 in the central cell is sufficient to lead to most of the phenotypic effects. The H3K27me3 data used for figure 1 is from developing endosperm, after REF6 could have acted in the gametophyte. But the authors argue that REF6 is affecting de-methylation during dormancy loss or germination, or at least after the stage tissue was collected for the 2016 paper. So is the activity of REF6 described in the rest of the paper the weak phenotype attributed to the paternal effect? Could the differences (which are small) be simply due to gene dosage effects in the endosperm, if REF6 is a rate limiting factor, even if REF6 is not imprinted? If dormancy is determined by central cell REF6 activity, then it can't relate to the effects on imprinted gene expression during germination.

If the above is true, the statement in the manuscript title is mis-leading, because dormancy is determined by central cell ref6 function – the events during germination are clearly affected by REF6, but lack an obvious phenotypic outcome (in Col-0 at least). This is also pertinent to the normal ACC response in ref6 mutants. A clear statement on this issue is hence necessary.

While ref6 has a maternal effect, the fact that many REF6 targets have detectable levels of H3K27me3 in the endosperm and since we could partially complement the ref6 mutant by expressing REF6 in the micropylar domain of the endosperm (Figure 5a), strongly suggests that REF6 also acts after fertilization. In support of this, homozygous ref6 mutants have a stronger delay in germination than mutants inheriting a maternal ref6 and a wild-type paternal allele (Figure 3b). Nevertheless, independent of when REF6 acts; the consequence of REF6 activity occurs during germination. We had discussed the idea that there are two types of genes that are either activated in the central cell or later on in the endosperm in the manuscript lines 472 to 495. In summary, we do not think that our title is misleading, the combination of epigenetic modifications present in the endosperm (independent on when they are established) determines the activity status of the alleles during germination.

Author details

1. Hikaru Sato

   Department of Plant Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences and Linnean Centre for Plant Biology, Uppsala, Sweden

   Contribution

   Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing - original draft, Writing - review and editing

   Competing interests

   none

   "This ORCID iD identifies the author of this article:" 0000-0001-7628-0414

2. Juan Santos-González

   Department of Plant Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences and Linnean Centre for Plant Biology, Uppsala, Sweden

   Contribution

   Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization

   Competing interests

   none

   "This ORCID iD identifies the author of this article:" 0000-0002-8712-9776

3. Claudia Köhler

   1. Department of Plant Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences and Linnean Centre for Plant Biology, Uppsala, Sweden
   2. Max Planck Institute of Molecular Plant Physiology, Potsdam-Golm, Germany

   Contribution

   Conceptualization, Data curation, Funding acquisition, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing - review and editing

   For correspondence

   claudia.kohler@slu.se

   Competing interests

   none

   "This ORCID iD identifies the author of this article:" 0000-0002-2619-4857

• 1,281

  Page views

• 272

  Downloads

• 6

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.