DNA demethylase ROS1 negatively regulates the imprinting of DOGL4 and in Arabidopsis thaliana

Haifeng Zhua,b,c,1, Wenxiang Xiea,b,1, Dachao Xuc,1, Daisuke Mikia,b, Kai Tanga,b, Chao-Feng Huanga,b,c,2, and Jian-Kang Zhua,d,2

aShanghai Center for Plant Stress , Center for Excellence in Molecular Plant Sciences, Chinese Academy of Sciences, Shanghai 200032, China; bNational Key Laboratory of Plant Molecular Genetics, Center for Excellence in Molecular Plant Sciences, Chinese Academy of Sciences, Shanghai 200032, China; cState Key Laboratory of Crop Genetics and Germplasm Enhancement, Nanjing Agricultural University, 210095 Nanjing, China; and dDepartment of Horticulture and Landscape Architecture, Purdue University, West Lafayette, IN 47907

Contributed by Jian-Kang Zhu, August 29, 2018 (sent for review July 31, 2018; reviewed by Yong Ding and Justin Goodrich) Genomic imprinting is a form of epigenetic regulation resulting in ential expression of parental alleles in the (17, 18). De differential gene expression that reflects the parent of origin. In novo methylation through the RNA-directed DNA methylation plants, imprinted gene expression predominantly occurs in the (RdDM) pathway regulates genomic imprinting at some other loci seed endosperm. Maternal-specific DNA demethylation by the (19). In addition to DNA methylation, trimethylation of histone DNA demethylase DME frequently underlies genomic imprinting in H3 on lysine 27 (H3K27me3), catalyzed by the Polycomb Repressive endosperm. Whether other more ubiquitously expressed DNA Complex 2 (PRC2), is another important epigenetic mark involved demethylases regulate imprinting is unknown. Here, we found that in the regulation of some imprinted genes in the endosperm (20, 21). the DNA demethylase ROS1 regulates the imprinting of DOGL4. Whereas DME is preferentially expressed and functions in the DOGL4 is expressed from the maternal allele in endosperm and central cell and endosperm, other 5-methylcytosine DNA gly- displays preferential methylation and suppression of the paternal cosylases/lyases, including ROS1, DEMETER-like 2 (DML2), allele. We found that ROS1 negatively regulates imprinting by and DEMETER-like 3 (DML3), are expressed in nearly all tis- demethylating the paternal allele, preventing its hypermethyla- sues of Arabidopsis (22–25). ROS1 dysfunction causes the tran- tion and complete silencing. Furthermore, we found that DOGL4 scriptional silencing of some transgenes and endogenous genes negatively affects seed dormancy and response to the phytohor- – mone and that ROS1 controls these processes by reg- (22, 25 27). ROS1-mediated DNA demethylation and gene ulating DOGL4. Our results reveal roles for ROS1 in mitigating regulation control diverse physiological functions, including an- imprinted gene expression and regulating seed dormancy. tibacterial defense (28), and the production of stomatal stem cells (29). Indeed, genome-wide DNA methylation analyses DNA methylation | DNA demethylation | ABA response | DME revealed that ROS1 protects thousands of genomic regions from DNA hypermethylation (25), suggesting that the cellular processes enomic imprinting is an epigenetic mechanism that alters regulated by ROS1 are not fully appreciated. For instance, Ggene expression in a parent-of-origin manner whereby some genes display differential expression, depending on whether the Significance allele was inherited from the maternal or paternal parent (1, 2). This epigenetic phenomenon occurs in mammals and seed plants Genomic imprinting is a form of epigenetic regulation causing and is considered to have evolved independently in the two line- parent-of-origin differential expression of maternally or pa- ages (3). Mammals display imprinted gene expression in both the ternally inherited alleles. The DNA demethylase DME regulates placenta and embryo, whereas imprinting in angiosperms appears the imprinting of many genes in the Arabidopsis endosperm. It to occur predominantly in the seed endosperm (2, 4). The endo- is not known whether and how other DNA demethylases may sperm is a triploid tissue derived after the fertilization of the also regulate imprinting. Here, we discovered that the DNA homodiploid central cell with one sperm cell; the other sperm cell demethylase ROS1 negatively regulates DOGL4 imprinting via fertilizes the egg cell to form a diploid embryo. The seed endo- demethylation of the DOGL4 promoter on the paternal allele. sperm serves to nurture and support the growing embryo and can Additionally, we found that DOGL4 negatively regulates seed be considered analogous to the mammalian placenta (5). Although dormancy and abscisic acid (ABA) response and that ROS1 the evolutionary forces for the selection of imprinted expression in regulates seed dormancy and ABA response by controlling DOGL4 the endosperm are still debated, conflicts between maternal and expression. Our results thus suggest a different mecha- paternal genomes in resource allocation to offspring or dosage nism of regulation of gene imprinting and reveal important roles control may be important driving forces for this process (6). of ROS1 in the regulation of seed dormancy and ABA response. Genome-wide RNA sequencing of endosperm has identified – Author contributions: C.-F.H. and J.-K.Z. designed research; H.Z., W.X., D.X., D.M., K.T., hundreds of imprinted genes in Arabidopsis, rice, and maize (7 and C.-F.H. performed research; H.Z., W.X., D.X., C.-F.H., and J.-K.Z. analyzed data; and 13). However, different studies in the same species show low C.-F.H. and J.-K.Z. wrote the paper. overlap in the lists of imprinted genes; therefore, putative imprinted Reviewers: Y.D., University of Science and Technology of China; and J.G., University genes identified from genome-wide surveys must be validated. of Edinburgh. Differential DNA methylation and histone modifications be- The authors declare no conflict of interest. tween maternal and paternal alleles are major regulators of Published under the PNAS license. imprinted gene expression (2, 14). Active DNA demethylation by 1H.Z., W.X., and D.X. contributed equally to this work. the5-methylcytosineDNAglycosylase/lyaseDEMETER(DME) 2To whom correspondence may be addressed. Email: [email protected] or jkzhu@ establishes differential DNA methylation at some imprinted loci (15, purdue.edu. 16). DME removes methylated cytosines in the central cell of the This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. female gametophyte, leading to DNA hypomethylation on the ma- 1073/pnas.1812847115/-/DCSupplemental. ternal alleles in the endosperm, which ultimately results in differ- Published online September 28, 2018.

E9962–E9970 | PNAS | vol. 115 | no. 42 www.pnas.org/cgi/doi/10.1073/pnas.1812847115 Downloaded by guest on September 26, 2021 although ROS1 is expressed in the endosperm (22), whether it Appendix,Fig.S1). Consistent with the imprinting of DOGL4 in the regulates imprinted gene expression is still unknown. endosperm, endosperm showed high DOGL4 expression only Here, we found that the imprinted expression of DOGL4,a when Col WT was the female parent, whereas the embryo paralogous gene of DOG1 that controls seed dormancy (30), showed similar DOGL4 levels in both reciprocal crosses (Fig. depends on differential methylation of its promoter. We show 1C). Thus, DOGL4 is imprinted in the endosperm of Col that the paternal allele is methylated and partially repressed in background. The differential expression between the maternal endosperm. ROS1 is specifically required to protect the paternal and paternal DOGL4 alleles in endosperm was more obvious allele from excessive DNA methylation and complete silencing, within the Col ecotype (Fig. 1C)thaninCol/C24hybrids(Fig. whereas the maternal DOGL4 allele remains unmethylated and 1A), suggesting that different genetic backgrounds affect the expressed upon loss of function of ROS1.Thus,ROS1isa extent of DOGL4 imprinting. negative regulator of the imprinting of DOGL4. We discovered that ROS1 regulates seed dormancy and the abscisic acid (ABA) DOGL4 Imprinting Requires DNA Methylation of the Paternal Promoter. response via regulation of DOGL4. These findings reveal roles To investigate whether DNA methylation regulates the silencing for ROS1 in the regulation of imprinted gene expression and of the DOGL4 paternal allele in endosperm, we used RdDM- seed dormancy. defective mutants as pollen donors in crosses with dogl4-1.We Results found that the loss of the key RdDM pathway component NRPD1, RDR2, DCL3,orNRPE1 in the male parent released the DOGL4 Is an Imprinted Gene in Arabidopsis Endosperm. At4g18650 silencing of paternal DOGL4 in the F endosperm (Fig. 2A), (DOGL4) was previously identified as a putative maternally 1 expressed imprinted gene in the endosperm (9). To confirm the suggesting that RdDM-mediated DNA methylation of the pater- imprinting of DOGL4, we first analyzed reciprocal crosses of Col nal allele is required for DOGL4 imprinting in the endosperm. and C24 accessions. We utilized a G/C polymorphism to discern To determine if the DOGL4 promoter is sufficient for im- DOGL4 allele-specific transcripts by RT-PCR-RFLP (Fig. 1A) printing in endosperm, we generated transgenic DOGL4 reporter and qRT-PCR (Fig. 1B). The embryos showed slightly higher lines in Col that contained 3.7- or 1.9-kb DOGL4 promoter β levels of the C24 allele, irrespective of the cross (Fig. 1 A and B). fragments driving -glucuronidase (GUS) expression (SI Appen- dix,Fig.S2A). F derived from reciprocal crosses Although F1 endosperms derived from reciprocal crosses showed 1 higher expression of the maternal allele compared with the paternal between Col wild-type (WT) and the transgenic lines showed allele, the change was not substantially higher than the expected significantly higher GUS expression from the maternal allele than twofold difference in the absence of imprinting. Therefore, it was the paternal allele (Fig. 2 B and C and SI Appendix,Fig.S2). In unclear if there was imprinting of DOGL4 expression in the en- contrast, embryos showed similar expression of GUS from both dosperm of these crosses. Next, we crossed wild-type Col with a alleles (Fig. 2 B and C and SI Appendix,Fig.S2). These data mutant line of DOGL4 (dogl4-1), which has a T-DNA insertion indicate that the 1.9-kb DOGL4 promoter is sufficient to confer in the intron and does not produce the full gene transcript (SI imprinted expression. PLANT BIOLOGY A Endosperm Embryo

Col DOGL4 Col DOGL4 C24 C24

B Col allele Col allele 1.4 C24 allele Endosperm 2.5 C24 allele Embryo 1.2 2.0 1.0 0.8 1.5 DOGL4 DOGL4 0.6 1.0 of of 0.4 0.5 Relative expression 0.2 Relative expression 0.0 0.0 ColColh ×C24C24 C24 C24h ×ColCol ColCol/C24hC24 C24 C24/ColhCol C 2.0 Endosperm 2.0 Embryo

1.5 1.5

1.0 1.0 DOGL4 DOGL4 of

of 0.5 0.5 Relative expression Relative expression 0.0 0.0 ColhCol Colhdogl4-1 dogl4-1hCol ColhCol Colhdogl4-1 dogl4-1hCol

Fig. 1. Analysis of DOGL4 expression in endosperm and embryo. Endosperms or embryos of 7–9 DAP from various crosses were collected for RNA extraction and expression analysis. UBQ10 was used an internal control. Crosses are indicated as maternal × paternal. (A) Allele-specific RT-PCR of DOGL4 in

the endosperm (Left) or embryo (Right)ofF1 seeds from reciprocal crosses between Col and C24. (B) Allele-specific qRT-PCR analysis of DOGL4 in the en- dosperm (Left) or embryo (Right)ofF1 seeds from reciprocal crosses between Col and C24. Data are means ± SD of three technical replicates. (C) Quantitative expression analysis of DOGL4 in the endosperm (Left) or embryo (Right)ofF1 seeds from reciprocal crosses between Col and dogl4-1 mutant. Data are means ± SD of three technical replicates.

Zhu et al. PNAS | vol. 115 | no. 42 | E9963 Downloaded by guest on September 26, 2021 A 2.0 D F 1.5 ATG 1709 DOGL4 Fragment 3 - -1161 1.0 -861

DOGL4 0.5 3 2 1 4 77 of - -1709 -830 -1161 -446 -861 -468 -1201 0.0 ǁ Relative expression E C24h Col

Fragment 4 1201 -1709 - ǀ B 3.7p-GUS(L4)×WT 1.5 WT×3.7p-GUS(L4) ǁ Colh Fragment 3 C24 -1161 1.0 -861 ǀ GUS

of 0.5

Relative expression 0.0 G Embryo Endosperm

1161 Fragment 3 - -861 C 1.9p-GUS(L8)×WT

830 Fragment 2 -468 2.0 WT×1.9p-GUS(L8) - 1.5

GUS 1.0 of 77 -

-446 Fragment 1 0.5 Methylated CG Unmethylated CG Methylated CHG Unmethylated CHG Relative expressionRelative 0.0 Methylated CHH Unmethylated CHH Embryo Endosperm

Fig. 2. DOGL4 imprinting requires RdDM-mediated methylation of the paternal promoter. (A) Paternal silencing of DOGL4 in the endosperm is dependent on RdDM pathway. RdDM pathway-defective mutants including nrpd1, nrpde1, dcl3,andrdr2 were respectively used as pollen donors to cross with dogl4-1 mutant, and the expression of DOGL4 in these derived endosperms were determined and compared. Data are means ± SD of three technical replicates. (B and

C) Expression analysis of GUS reporter gene in embryo and endosperm of F1 seeds from reciprocal crosses between WT and a 3.7-kb DOGL4 promoter-driven GUS transgenic line (L4) (A) and between WT and a transgenic line (L14) with 1.9-kb DOGL4 promoter (B). Data are means ± SD of three technical replicates. (D) Diagram of bisulfite sequencing on four segments covering 1.9-kb promoter of DOGL4.(E) Bisulfite sequencing of the four segments of DOGL4 promoter in 7–9 DAP endosperm of Col wild type and the result showed that DNA methylation of paternal allele in CG sequence context at fragment 3 was increased

compared with that of maternal allele. (F) Bisulfite sequencing of fragment 3 in F1 endosperms from reciprocal crosses between Col and C24 ecotypes. (G) Bisulfite sequencing of fragment 3 in F1 endosperm of Col/nrpd1.

Given that DNA methylation regulates the imprinting of DOGL4, Loss of DME and PRC2 Reduces DOGL4 Expression Likely Indirectly we investigated whether the parental alleles may display differ- Through an Effect on Endosperm Development. Previous genome- ential DNA methylation within the 1.9-kb DOGL4 promoter. wide analyses had suggested that DME regulates DOGL4 im- We performed bisulfite sequencing of four nonoverlapping printing in the endosperm (9). To clarify the potential role of promoter fragments in endosperm (Fig. 2D). Fragment 3 (re- DME, we examined maternal and paternal DOGL4 expression ferred to as −1.0-kb region) but not the other fragment meth- in dme/C24 F1 endosperm. We found that both alleles were ylation profiles revealed two distinct CG methylation patterns, suppressed (SI Appendix, Fig. S3A), consistent with previous data consistent with differential methylation of the alleles (Fig. 2E). (9). However, reduced DOGL4 expression in dme/C24 endo- To confirm that DNA methylation occurred at the paternal al- sperm was not associated with substantially increased DNA lele, we collected endosperm from reciprocal crosses of Col and methylation within the DOGL4 promoter, which even showed C24 and utilized a G/T polymorphism within the −1.0-kb region reduced CG methylation in fragment 3 of the paternal allele (SI to distinguish the two alleles. Although there was a bias toward Appendix, Fig. S3B). sequenced clones from the Col allele, clones from the maternal To investigate whether the PRC2 complex regulates DOGL4 and paternal alleles clearly displayed low and high DNA meth- imprinting, we quantified DOGL4 expression in the endo- ylation of the −1.0-kb region, respectively (Fig. 2F). Additionally, sperm of PRC2-defective mutants, including mea/+, fie/+, the higher DNA methylation of the −1.0-kb paternal allele was msi1/+ and fis2/+. Similar to dme/+, the total expression of lost in Col/nrpd1 F1 endosperm (Fig. 2G). Together, these results DOGL4 was repressed in endosperm from all of the mutants suggest that paternal DNA methylation at the −1.0-kb region of (SI Appendix,Fig.S3C), and CG methylation of fragment the DOGL4 promoter is necessary and sufficient for imprinted 3 was lost from the paternal allele in fie/C24 endosperm (SI expression of DOGL4 in endosperm. Appendix,Fig.S3D).

E9964 | www.pnas.org/cgi/doi/10.1073/pnas.1812847115 Zhu et al. Downloaded by guest on September 26, 2021 Given that reduced methylation of DOGL4 correlated with in- expression and altered methylation of DOGL4 in the PRC2- creased expression from the maternal allele in WT plants, it is un- deficient and dme mutants are probably due to the delayed likely that the reduced levels of DOGL4 in dme and PRC2-defective and/or arrested development of endosperm, separate from their mutants arise from promoter demethylation. PRC2-deficient and defects in imprinting. dme mutant endosperms display delayed development, and em- bryos arrest before the heart stage, 4 d after pollination (DAP) ROS1 Negatively Regulates the Imprinting of DOGL4. Previous work (15, 31, 32). We postulated that the delayed development of showed that fragment 1 of the DOGL4 promoter (Fig. 2E, re- these mutants may impact DOGL4 expression, independent of ferred to as −0.5-kb region) was hypermethylated in ros1 mutant the mutant effects on imprinting. We examined DOGL4 levels plants (33). We confirmed increased CG and CHG methylation during the development of WT plants and observed expression in at the −0.5-kb region in seedlings of both ros1-1 (C24 back- the endosperm starting at 6 DAP and in the embryo by 8 DAP ground) and ros1-4 (Col background) mutants compared with (SI Appendix, Fig. S4 A and B). Similarly, GUS activity in the 3.7- WT controls (SI Appendix, Fig. S5A). Since the −1.0-kb region kb promoter pDOGL4:GUS transgenic line was detected in the displayed differential DNA methylation in WT endosperm (Fig. endosperm by 5–6 DAP and in the embryo by 8–9 DAP (SI 2 E and F), we investigated whether ROS1 might also regulate Appendix, Fig. S4C). Given that we evaluated the endosperm DNA methylation of this region. Indeed, the −1.0-kb region shortly after DOGL4 expression initiates in the WT, the low displayed increased CG and CHG methylation in seedlings of all

446 Fragment 1 (Endosperm) A C 77 - 120 Col n=17 - 100 ros1-4 n=20 n=20 80 ros1-6 ros1-4 C24 n=20 60 ros1-1 n=20 40 20 0 Fragment 3 (Embryo) -861 -1161 Cytosine methylation (%) methylation Cytosine CG CHG CHH D B Fragment 3 (Endosperm) ǁ PLANT BIOLOGY -861 -1161 ros1-1h ros1-4 ǁ ǀ ros1-1h ros1-4 ǁ ǀ ros1-4 h ros1-1 ǀ -446 Fragment 1 (Embryo) -77 ǁ E ros1-4 ros1-4 h ros1-1 ǀ Fragment 1 (Embryo) -77 F -446 Methylated CG Unmethylated CG Methylated CHG Unmethylated CHG Col Methylated CHH Unmethylated CHH

Fig. 3. Effect of ROS1 mutation on DOGL4 promoter DNA methylation in endosperms and embryos. (A) Bisulfite sequencing of −1.0-kb region of DOGL4 promoter in WT and ros1 mutants. Twelve-day-old seedlings of Col, ros1-4, ros1-6, C24, and ros1-1 were used for DNA isolation and bisulfite sequencing. DNA methylation at CG and CHG contexts was increased in all ros1 mutants. (B–F) DNA methylation analysis of −1.0-kb region (Fragment 3) (B and D) and −0.5-kb

region (Fragment 1) (C–E)ofDOGL4 promoter. (B and C) DNA methylation analysis in F1 endosperms from reciprocal crosses between ros1-1 and ros1-4 (B) and in ros1-4 endosperms (C). (D–F) DNA methylation analysis in F1 embryos from reciprocal crosses between ros1-1 and ros1-4 (D), ros1-4 embryos (E), and Col embryos (F).

Zhu et al. PNAS | vol. 115 | no. 42 | E9965 Downloaded by guest on September 26, 2021 ros1 mutants compared with WT controls (Fig. 3A). Thus, Thus, our data suggest that ROS1 negatively regulates gene ROS1 protects the DOGL4 promoter from hypermethylation in imprinting by preventing the hypermethylation and complete seedlings. silencing of the paternal allele. Given that loss of ROS1 led to increased DNA methylation of the −1.0-kb region of the DOGL4 promoter in seedlings, we DOGL4 Regulates Seed Dormancy and ABA Response. DOGL4 is a investigated whether ROS1 may regulate differential DNA meth- member of the plant-specific DOG1 gene family (30). DOG1 is a ylation of the DOGL4 promoter in endosperm. F1 endosperms major regulator of seed dormancy, and knockout of DOG1 from reciprocal crosses between ros1-1 (C24) and ros1-4 (Col) renders seeds nondormant. Arabidopsis eFP browser data show showed increased CG and CHG methylation of the −1.0-kb re- that DOGL4 expression is highly enriched in seeds (34). Con- gion only at the paternal allele, but not the maternal allele, sistent with these data, our expression analysis also showed that compared with WT (compare Fig. 2F with Fig. 3B). Next, we DOGL4 expression in seeds, especially in dry seeds, was much analyzed differential DNA methylation of the −0.5-kb region in higher than that in other tissues such as roots, leaves, and flowers ros1 mutants. There are no polymorphisms in the −0.5-kb region (Fig. 5A). between Col and C24 to distinguish the paternal and maternal To investigate whether DOGL4 regulates seed dormancy, we alleles, therefore we examined ros1-4 mutant endosperm. Our evaluated the rates of freshly harvested seeds. We results suggest increased CG and CHG methylation of only one found that the germination rate of dogl4-1 seeds was lower than allele, likely the paternal allele, in the −0.5-kb region, as with that of WT (Fig. 5B). To confirm the involvement of DOGL4 in the −1.0-kb region, in ros1-4 endosperm compared with WT en- seed dormancy regulation, we used the CRISPR/Cas9 system dosperm (Figs. 2E and 3C). Thus, knockout of ROS1 increased (35) to generate a mutant allele, dogl4-2, in the highly dormant the differential DNA methylation at the DOGL4 promoter be- C24 accession. The dogl4-2 mutant has a frameshift mutation tween parental alleles in endosperm. In contrast, both alleles of (1-bp insertion at +1154 bp from ATG) in the DOGL4 gene (SI the DOGL4 promoter displayed high DNA methylation in the Appendix, Fig. S1). We found that dogl4-2 displayed increased embryos of ros1 mutant (Fig. 3 D and E) but not Col WT (Fig. seed dormancy compared with the C24 WT control (Fig. 5C). 3F). Together, these data suggest that ROS1 is required to protect These data suggest that DOGL4 promotes germination and nega- the paternal DOGL4 allele from hypermethylation in endosperm. tively regulates seed dormancy in WT plants. Next, we determined whether ROS1 regulates DOGL4 ex- The phytohormone ABA is a key factor that induces dormancy pression. Relative to WT controls, DOGL4 expression was de- (36, 37). We found that DOGL4 expression was induced by ABA creased in ros1 mutant endosperm (Fig. 4A), and even more so in treatment in both roots and shoots of Col (SI Appendix, Fig. S5 B embryos (Fig. 4B) and freshly harvested seeds (Fig. 4C), con- and C), suggesting that DOGL4 might attenuate the ABA re- sistent with DNA methylation levels. We analyzed the expression sponse. The germination rates of both dogl4-1 and dogl4-2 mu- of maternal and paternal alleles in F1 endosperms from re- tants were less than the WT controls at all ABA concentrations ciprocal crosses between ros1-1 (C24) and ros1-4 (Col) and found tested (Fig. 5 D and E). To confirm the ABA hypersensitivity that upon knockout of ROS1, DOGL4 expression was reduced phenotype of dogl4-1 mutant, we compared postgermination from the maternal allele only slightly, whereas paternal DOGL4 growth of WT and dogl4-1 mutant on agar media containing expression was completely repressed (Fig. 4 D and E). There- different concentrations of ABA. Consistently, dogl4-1 mutants fore, DOGL4 is a maternally expressed, imprinted gene in the formed green cotyledons at a slower rate than WT at all ABA ros1 mutant background in ColxC24 crosses (Fig. 4 D and E). concentrations (Fig. 5 F and G). We also depleted the expression

A BC

0.8 Endosperm 1.2 Embryo 1.5

1.0 DOGL4 DOGL4 0.6 DOGL4 a 0.8 1.0 0.4 0.6 0.4 0.5 b 0.2 0.2 c 0.0 0.0 0.0

Col ros1-4 ros1-6 C24 ros1-1 Col ros1-4 ros1-6 C24 ros1-1 of Relative expression ros1-4 ros1-6

Relative expression of Relative expression Col ros1-4 ros1-6 C24 ros1-1 Relative expression of Relative expression Col ros1-4 ros1-6 C24 ros1-1 Col E 2.0 Col allele D C24 allele DOGL4 1.6 1.2 Col DOGL4 0.8 C24 0.4 0.0 Relative expression of Relative expression

Fig. 4. Effect of ROS1 mutation on DOGL4 expression. (A and B) Effect of ROS1 mutation on the expression of DOGL4 in endosperms and embryos. En- dosperms (A) or embryos (B)of7–9 DAP seeds from WT (Col or C24) and ros1 mutants were collected for RNA isolation. The expression of DOGL4 in WT and ros1 mutants was examined and compared. Data are means ± SD of three technical replicates. (C) Effect of ROS1 mutation on the expression of DOGL4 in freshly harvested seeds. Fresh seeds of ros1-4 and ros1-6 and Col wild-type were harvested for RNA extraction and DOGL4 expression analysis. Data are means ± SD of three biological replicates. Means with different letters are significantly different (P < 0.05, Tukey test). (D and E) Allele-specific RT-PCR (D)and

qRT-PCR (E)ofDOGL4 in the endosperm of F1 seeds from reciprocal crosses between Col and C24 or between ros1-4 and ros1-1.

E9966 | www.pnas.org/cgi/doi/10.1073/pnas.1812847115 Zhu et al. Downloaded by guest on September 26, 2021 A 900 B C 100 a 60 600 a 300 b 50 80 c 50 60 40 8 30 DOGL4 40 b b 4 20 0 20 10 Relative expression of Relative expression Germination rate (%) Germination

0 rate (%) Germination 0 Col dogl4-1 rig1-1 ros1-4 ros1-4 C24 dogl4-2 ros1-1

E 120 C24C24 D 120 ColCol a rig1-2dogl4-2 aa a a 100 b b a 100 aa rig1-1dogl4-1 a ros1-1 80 b ros1-1 80 a ros1-4ros1-4 b b a a 60 b 60 b b b 40 b b b 40 a 20 c 20 b 0 Germination rate (%) Germination

Germination rate (%) Germination 0 00.20.51 00.20.51 2 ABA concentration ( M) G 120 F ABA concentration ( M) aaa a ColCol 100 b b a rig1-1dogl4-1 80 ros1-4ros1-4 ros1-4 dogl4-1 60 40 a Col 20 b bbb 0

Control 0.5 μM ABA rate (%) cotyledon Green 00.20.51 ABA concentration ( M)

Fig. 5. Mutation of DOGL4 or ROS1 increases seed dormancy and ABA sensitivity. (A)ExpressionanalysisofDOGL4 in various tissues or organs of Col wild-type plants. (B and C) Seed dormancy of wild-type, dogl4,andros1 mutants. Freshly harvested seeds of Col wild-type, dogl4-1,andros1-4 (B)orC24wild-type,dogl4-2, and ros1-1 (C) were sown on water-soaked filter paper for 3 d, and then seed germination based on radicle emergence was scored and compared. (D and E) Germination of mature seeds in response to different concentrations of ABA. Seeds of Col, dogl4-1,andros1-4 (D) or C24 wild-type, dogl4-2,andros1-1 (E)were exposedto0,0.2,0.5,1,or2μM ABA for 2 d, and then seed germination rate was calculated and compared. (F and G) Green cotyledon formation of Col, dogl4-1, and ros1-4 in response to ABA. Seeds were grown on an agar plate containing 0, 0.2, 0.5, or 1 μM ABA for 9 d, and then the rate of green cotyledon formation was calculated and compared. Data shown are means ± SD of three biological replicates. Means with different letters are significantly different (P < 0.05, Tukey test).

of DOGL4 in the Col background by RNAi (SI Appendix, Fig. mutants to ABA. Loss-of-function of either DML2 or DML3,two

S6A). RNAi lines of DOGL4 were more sensitive to ABA than paralogues of ROS1, did not elevate the DNA methylation level of PLANT BIOLOGY WT in seed germination and cotyledon greening assays (SI Ap- DOGL4 promoter (Fig. 6D). The sensitivity to ABA in dml2-2 or pendix, Fig. S6 B and C). Together, these results show that dml3-2 mutants was also not altered (Fig. 6 E and F). Furthermore, DOGL4 negatively regulates seed dormancy and ABA response. the triple mutant rdd-2 (ros1-4 dml2-2 dml2-3) showed a similar F1 seeds from dogl4×WT should have reduced DOGL4 ex- DNA methylation level of DOGL4 promoter and ABA sensitivity pression in the endosperm compared with WT×dogl4, but both to the ros1-4 single mutant (Fig. 6 D–F). Therefore, of the three embryos should have similar DOGL4 expression. We found that ubiquitously expressed DNA demethylases, ROS1 is significantly F1 seeds of dogl4-2×C24 had stronger seed dormancy than F1 involved in the regulation of seed dormancy and ABA response. seeds of C24×dogl4-2 (SI Appendix, Fig. S6D), and that F1 seeds To confirm that the decreased expression of DOGL4 is re- of dogl4-1×Col also had higher ABA sensitivity in germination sponsible for the increased seed dormancy and ABA sensitivity than F1 seeds of Col×dogl4-1 (SI Appendix, Fig. S6E). These re- in ros1 mutants, we overexpressed DOGL4 in a ros1-4 mutant sults indicate that imprinted DOGL4 expression in the endosperm background (Fig. 7A). Phenotypic analysis in the transgenic lines may contribute to promote seed dormancy and ABA sensitivity. revealed that overexpression of DOGL4 could rescue the de- fective seed dormancy and ABA sensitivity phenotypes of ros1-4 ROS1 Modulates Seed Dormancy via Regulation of DOGL4 Expression. (Fig. 7 B and C). Thus, ROS1 negatively regulates seed dor- As DOGL4 regulates seed dormancy and ABA response (Fig. 5), mancy and ABA response via DOGL4. we hypothesized that ROS1 may also regulate these processes via controlling the expression of DOGL4. We compared seed dor- Discussion mancy phenotypes of ros1-4, dogl4-1, and Col WT. Similar to In this study, we validated DOGL4 as a maternally imprinted dogl4-1, ros1-4 showed a lower germination rate than WT (Fig. gene in endosperm. Our results revealed that DOGL4 imprinting 5B), indicating that the ros1-4 mutant has increased seed dor- depends on DNA methylation of the paternal allele in the −1.0-kb mancy. We also investigated the seed dormancy phenotype of region of DOGL4 promoter. The increased DOGL4 promoter ros1-1 from the highly dormant C24 accession. Consistently, DNA methylation and reduced DOGL4 expression in the pa- freshly harvested ros1-1 seeds displayed a lower germination rate ternal allele were suppressed in F1 endosperms when RdDM- than the C24 WT control (Fig. 5C). These results indicate that pathway defective mutants were used as male parents, suggesting ROS1 regulates seed dormancy. that differential DNA methylation might be established before Next, we investigated whether ros1 mutants show an altered fertilization. In plants, DME and PRC2 are the key regulators of ABA response. Like dogl4 mutants, the germination rates of imprinted gene expression in the endosperm (2, 14). PRC2-deficient ros1-1, ros1-4, and ros1-6 were less than their WT controls at all and dme mutant plants displayed defective endosperm devel- ABA concentrations tested (Figs. 5 D and E and 6A). Further, opment; this developmental defect likely contributes to the low ros1 mutants formed green cotyledons at a slower rate than their expression of DOGL4 in these mutants since DOGL4 is highly respective WT controls at all ABA concentrations tested (Figs. 5 expressed only later during endosperm development. Intrigu- F and G and 6 B and C), indicating an increased sensitivity of ros1 ingly, the paternally derived DNA methylation in the −1.0-kb

Zhu et al. PNAS | vol. 115 | no. 42 | E9967 Downloaded by guest on September 26, 2021 ACB 120 ros1-4 Col 120 Col a a a Col aa a 100 Ros1-1dml2-2 100 a a a ros1-4 ros1-4 ros1-6 a ros1-6 a b 80 ros1-6 80 b a dml3-2 60 60 b a b b 40 c 40 a 20 bb 20

Germination rate (%) Germination b b

0 rate (%) cotyledon Green 0 00.20.512 00.20.5 ABA concentration (μM) ABA concentration (μM) Control 0.5μM ABA

DECol F n=20 Col ros1-4 120 Col 100 n=20 120 ros1-4 aaaaa a aa dml2-2 aaaaa ros1-4 dml2-2n=19 a a a dml3-2 100 80 100 a dml2-2 dml3-2n=20 a a rdd-2 a ab 80 a dml3-2 rdd-2 n=20 80 a rdd-2 60 b a a 60 60 b b b b 40 c 40 40 a ab

20 rate (%) Germination 20 20 c c

Cytosine methylation (%) methylation Cytosine 0 0 0 rate (%) Green cotylendon CG CHG CHH 00.20.51 00.20.5 ABA concentration (μM) ABA concentration (μM)

Fig. 6. Mutation of ROS1 but not DML2 and DML3 affects ABA sensitivity. (A) Seed germination of Col WT, ros1-4,andros1-6 under different concentrations of ABA. Seeds were exposed to 0, 0.2, 0.5, 1, or 2 μM ABA for 3 d, and then seed germination rate was calculated and compared. (B) Green cotyledon formation of Col, ros1-4, and ros1-6 in response to ABA. Seeds were exposed to 0, 0.2, or 0.5 μM ABA for 7 d, and then the rate of green cotyledon formation was calculated and compared. (C) Picture of green cotyledon formation of Col, ros1-4, ros1-6, dml2-2,anddml3-2 in response to ABA. Seeds were grown on an agar plate containing 0 or 0.5 μMABA for 9 d. (D) DNA methylation analysis of −1.0 kb region of DOGL4 promoter. Twelve-day-old seedlings of Col wild-type, ros1-4, dml2-2, dml3-2,andrdd-2 (ros1-4 dml2-2 dml3-2) was sampled for bisulfite sequencing. (E) Seed germination of Col, ros1-4, dml2-2,anddml3-2 under different concentrations of ABA. Seeds were exposed to 0, 0.2, 0.5, and 1 μM ABA for 3 d, and then seed germination rate was calculated and compared. (F) Green cotyledon for- mation of Col, ros1-4, dml2-2, dml3-2,andrdd-2 under different concentrations of ABA. Seeds were exposed to 0, 0.2, or 0.5 μM ABA for 9 d, and then the rate of green cotyledon formation was calculated and compared. Data shown are means ± SD of three biological replicates. Means with different letters are significantly different (P < 0.05, Tukey test).

DOGL4 promoter in WT endosperms was lost in dme/C24 and DOGL4 is a member of the plant-specific DOG1 gene family, fie/C24 endosperms (SI Appendix, Fig. S3). The underlying which regulates seed dormancy. Mutation of DOG1 renders mechanism of this effect on paternal DOGL4 methylation re- seeds nondormant (30). We found that DOGL4 also controls mains to be elucidated. seed dormancy but, in contrast to DOG1 mutations, mutations of Among the family of four 5-methylcytosine DNA glycosylases/ DOGL4 enhanced seed dormancy (Fig. 5). Like DOG1, DOGL4 lyases, we found that only ROS1 negatively regulates DNA is highly and preferentially expressed in seeds (Fig. 5A). dogl4 methylation of the DOGL4 promoter. We found that multiple mutants did not show pleiotropic phenotypes, which suggests regions within the DOGL4 promoter were hypermethylated on that DOGL4 is specifically involved in the regulation of seed the paternal allele in ros1 endosperms, which might underlie the dormancy. We found that DOGL4 expression in the endosperm full suppression of paternal DOGL4 expression. By contrast, may contribute to regulation of seed dormancy and ABA re- mutation of ROS1 did not affect DOGL4 promoter DNA meth- sponse (SI Appendix,Fig.S6D and E). These observations support ylation and gene expression from the maternal allele in the en- arecentviewthatexpressionofmaternally imprinted genes in the dosperm. Consequently, the differential expression between endosperm plays important roles in modulating seed dormancy parental alleles was increased in ros1 mutants. Our results thus (40). DOG1 regulates seed germination through inhibiting the ex- revealed a role of ROS1 in mitigating imprinting effects by pression of gibberellin-regulated genes encoding cell-wall remod- demethylating the otherwise highly methylated and suppressed eling proteins (41). Whether DOGL4 has an opposite role in paternal allele. Further, our findings suggest that ROS1 and the regulating the expression of these genes, and whether DOGL4 RdDM pathway antagonize each other to control the DNA antagonizes DOG1 to regulate seed dormancy, remain to be methylation of DOGL4 promoter of the paternal allele in the investigated in the future. endosperm and, in turn, DOGL4 imprinting. ROS1 positively regulates DOGL4 expression through demethylation Whether ROS1 regulates imprinted gene expression more of the DOGL4 promoter region. We found that seed dormancy and broadly remains to be determined. Since mutation of ROS1 did ABA sensitivity were increased in ros1 mutants compared with not influence DNA methylation of the DOGL4 maternal allele in their WT controls, phenocopying dogl4 mutants. Furthermore, the endosperm, we speculate that ROS1 regulates DOGL4 pro- overexpression of DOGL4 in a ros1 mutant background suppressed moter methylation of the paternal allele in pollen before fertil- the enhanced seed dormancy and ABA hypersensitivity phenotypes ization. The maternal allele in the endosperm is derived from of ros1 mutant plants. These results indicate that the higher seed central cells where DME is preferentially expressed and the CG dormancy and ABA sensitivity germinating seeds in ros1 mutants maintenance methyltransferase MET1 is repressed (15, 38, 39). are at least partly caused by reduced expression of DOGL4.In We found that DME does not regulate maternal DOGL4 pro- summary, we have discovered a role for ROS1 in mitigating the moter DNA methylation; however, MET1 repression might paternal-specific silencing of DOGL4 in Arabidopsis endosperm and contribute to the relatively low DNA methylation of the mater- found that ROS1 negatively regulates seed dormancy and ABA nal DOGL4 allele in both WT and ros1 endosperms. Further sensitivity by promoting the expression of DOGL4 (SI Appendix, work is required to test this hypothesis. Fig. S7).

E9968 | www.pnas.org/cgi/doi/10.1073/pnas.1812847115 Zhu et al. Downloaded by guest on September 26, 2021 A B 8 100 a a c a 6 d 80 60 4 b 40 DOGL4 2 a 20 b Relative expression of 0 rate (%) Germination 0 ColCol ros1-4 R17 R22 Col ros1-4 R17 R22 ros1-4;pUBQ10-DOGL4 ros1-4;pUBQ10-DOGL4

Col C 100 a ros1-4 a a a R17 ros1-4;pUBQ10-DOGL4 b b R22 80 b bc 60 aa c b 40 c b c c 20 aa a

Germination rate (%) Germination d 0 0 0.2 0.5 0.75 1 ABA concentration (μM)

Fig. 7. Overexpression of DOGL4 rescues the seed dormancy and ABA sensitivity phenotypes of ros1-4 mutant plants. (A) Expression analysis of DOGL4 in freshly harvested seeds of Col, ros1-4, and two independent DOGL4 overexpression lines with ros1-4 mutant background. (B) Seed dormancy of Col, ros1-4, and two DOGL4 overexpression lines with ros1-4 mutant background. Freshly harvested seeds were sown on water-soaked filter paper for 3 d, and then seed germination based on radicle emergence was scored and compared. (C) Seed germination of Col, ros1-4, and two DOGL4 overexpression lines under different concentrations of ABA. Seeds were exposed to 0, 0.2, 0.5, 0.75, or 1 μM ABA for 2 d, and then seed germination rate was calculated and compared. Data shown are means ± SD of three biological replicates. Means with different letters are significantly different (P < 0.05, Tukey test).

Materials and Methods products were cloned and sequenced, and DNA methylation was analyzed Plant Materials and Growth Conditions. Arabidopsis thaliana ecotypes using Kismeth tool. See SI Appendix, SI Materials and Methods. Columbia-0 (Col-0) and C24 were obtained from the Arabidopsis Stock Centre PLANT BIOLOGY (https://www.arabidopsis.org). The mutants were either ordered from ABRC or Seed Dormancy and ABA Response. Radicle emergence of freshly harvested generated in our laboratory (SI Appendix,TableS1). Plants were grown in a seeds was scored for seed dormancy determination. For ABA response tests, growth room with 16 h of light and 8 h of darkness at 20–24 °C. after-ripened seeds were grown on plates containing different concentrations of ABA, and radicle emergence and green cotyledon formation were recorded RNA and DNA Isolation from Dissected Seed Embryo and Endosperm. Seeds of for seed germination rate and postgermination growth, respectively. See SI 7–9 d after pollination were dissected with a fine forceps on a slide under a Appendix, SI Materials and Methods. dissecting microscope, and RNA or DNA was then isolated from the dissected seed embryo and endosperm. See SI Appendix, SI Materials and Methods. GUS Analysis. Vector construct and GUS analysis for pDOGL4:GUS transgenic lines were described in detail in SI Appendix, SI Materials and Methods. RT-PCR and Quantitative PCR. Various tissues of control or treated plants were collected for RNA isolation and RT-PCR analysis. See SI Appendix, SI Materials Vector Constructs for DOGL4 and Phenotypic Analysis of Transgenic Lines. and Methods. Vector constructs and phenotypic analysis for DOGL4 RNAi, knockout, and overexpression transgenic lines were described in detail in SI Appendix, SI Allele-Specific Expression Analysis of DOGL4. A dCAPS marker and allele- Materials and Methods. specific primers were designed based on a G/C polymorphism between Col and C24 at DOGL4 and were used for allele-specific expression analysis of ACKNOWLEDGMENTS. We thank Life Science Editors for editorial assistance. DOGL4. See SI Appendix, SI Materials and Methods. This work was supported by the Shanghai Center for Plant Stress Biology, Jiangsu Science Fund for Distinguished Young Scholars Grant BK20150027, DNA Methylation Analysis. Extracted DNA was sodium-bisulfite converted and Chinese Academy of Sciences, and Strategic Priority Research Program Grant purified and then amplified using two rounds of nested primers. The PCR XDB27040000 of the Chinese Academy of Sciences.

1. Reik W, Walter J (2001) Genomic imprinting: Parental influence on the genome. Nat 8. Luo M, et al. (2011) A genome-wide survey of imprinted genes in rice seeds reveals Rev Genet 2:21–32. imprinting primarily occurs in the endosperm. PLoS Genet 7:e1002125. 2. Köhler C, Wolff P, Spillane C (2012) Epigenetic mechanisms underlying genomic im- 9. Hsieh TF, et al. (2011) Regulation of imprinted gene expression in Arabidopsis en- printing in plants. Annu Rev Plant Biol 63:331–352. dosperm. Proc Natl Acad Sci USA 108:1755–1762. 3. Feil R, Berger F (2007) Convergent evolution of genomic imprinting in plants and 10. McKeown PC, et al. (2011) Identification of imprinted genes subject to parent- mammals. Trends Genet 23:192–199. of-origin specific expression in Arabidopsis thaliana seeds. BMC Plant Biol 11:113. 4. Bauer MJ, Fischer RL (2011) Genome demethylation and imprinting in the endosperm. 11. Waters AJ, et al. (2011) Parent-of-origin effects on gene expression and DNA meth- Curr Opin Plant Biol 14:162–167. ylation in the maize endosperm. Plant Cell 23:4221–4233. 5. Berger F (2003) Endosperm: The crossroad of seed development. Curr Opin Plant Biol 12. Wolff P, et al. (2011) High-resolution analysis of parent-of-origin allelic expression in 6:42–50. the Arabidopsis endosperm. PLoS Genet 7:e1002126. 6. Haig D (2013) Kin conflict in seed development: An interdependent but fractious 13. Zhang M, et al. (2011) Extensive, clustered parental imprinting of protein-coding and collective. Annu Rev Cell Dev Biol 29:189–211. noncoding RNAs in developing maize endosperm. Proc Natl Acad Sci USA 108: 7. Gehring M, Bubb KL, Henikoff S (2009) Extensive demethylation of repetitive ele- 20042–20047. ments during seed development underlies gene imprinting. Science 324:1447–1451. 14. Gehring M (2013) Genomic imprinting: Insights from plants. Annu Rev Genet 47:187–208.

Zhu et al. PNAS | vol. 115 | no. 42 | E9969 Downloaded by guest on September 26, 2021 15. Choi Y, et al. (2002) DEMETER, a DNA glycosylase domain protein, is required for 29. Yamamuro C, et al. (2014) Overproduction of stomatal lineage cells in Arabidopsis endosperm gene imprinting and seed viability in arabidopsis. Cell 110:33–42. mutants defective in active DNA demethylation. Nat Commun 5:4062. 16. Hsieh TF, et al. (2009) Genome-wide demethylation of Arabidopsis endosperm. 30. Bentsink L, Jowett J, Hanhart CJ, Koornneef M (2006) Cloning of DOG1, a quantitative Science 324:1451–1454. trait locus controlling seed dormancy in Arabidopsis. Proc Natl Acad Sci USA 103: 17. Gehring M, et al. (2006) DEMETER DNA glycosylase establishes MEDEA polycomb 17042–17047. gene self-imprinting by allele-specific demethylation. Cell 124:495–506. 31. Kiyosue T, et al. (1999) Control of fertilization-independent endosperm development 18. Kinoshita T, et al. (2004) One-way control of FWA imprinting in Arabidopsis endo- by the MEDEA polycomb gene in Arabidopsis. Proc Natl Acad Sci USA 96:4186–4191. – sperm by DNA methylation. Science 303:521 523. 32. Köhler C, et al. (2003) Arabidopsis MSI1 is a component of the MEA/FIE Polycomb 19. Vu TM, et al. (2013) RNA-directed DNA methylation regulates parental genomic im- group complex and required for seed development. EMBO J 22:4804–4814. – printing at several loci in Arabidopsis. Development 140:2953 2960. 33. Huang CF, et al. (2013) A Pre-mRNA-splicing factor is required for RNA-directed DNA 20. Schuettengruber B, Cavalli G (2009) Recruitment of polycomb group complexes and methylation in Arabidopsis. PLoS Genet 9:e1003779. their role in the dynamic regulation of cell fate choice. Development 136:3531–3542. 34. Winter D, et al. (2007) An “electronic fluorescent pictograph” browser for exploring 21. Jiang H, Köhler C (2012) Evolution, function, and regulation of genomic imprinting in and analyzing large-scale biological data sets. PLoS One 2:e718. plant seed development. J Exp Bot 63:4713–4722. 35. Feng Z, et al. (2013) Efficient genome editing in plants using a CRISPR/Cas system. Cell 22. Gong Z, et al. (2002) ROS1, a repressor of transcriptional gene silencing in Arabidopsis, Res 23:1229–1232. encodes a DNA glycosylase/lyase. Cell 111:803–814. 36. Finkelstein R, Reeves W, Ariizumi T, Steber C (2008) Molecular aspects of seed dor- 23. Agius F, Kapoor A, Zhu JK (2006) Role of the Arabidopsis DNA glycosylase/lyase mancy. Annu Rev Plant Biol 59:387–415. ROS1 in active DNA demethylation. Proc Natl Acad Sci USA 103:11796–11801. 37. Holdsworth MJ, Bentsink L, Soppe WJ (2008) Molecular networks regulating Arabidopsis 24. Lister R, et al. (2008) Highly integrated single-base resolution maps of the epigenome – in Arabidopsis. Cell 133:523–536. seed maturation, after-ripening, dormancy and germination. New Phytol 179:33 54. 25. Qian W, et al. (2012) A histone acetyltransferase regulates active DNA demethylation 38. Jullien PE, et al. (2008) Retinoblastoma and its binding partner MSI1 control im- in Arabidopsis. Science 336:1445–1448. printing in Arabidopsis. PLoS Biol 6:e194. 26. Zhu J, Kapoor A, Sridhar VV, Agius F, Zhu JK (2007) The DNA glycosylase/lyase 39. Jullien PE, Susaki D, Yelagandula R, Higashiyama T, Berger F (2012) DNA methylation ROS1 functions in pruning DNA methylation patterns in Arabidopsis. Curr Biol 17: dynamics during sexual reproduction in Arabidopsis thaliana. Curr Biol 22:1825–1830. 54–59. 40. Piskurewicz U, et al. (2016) Dormancy-specific imprinting underlies maternal in- 27. Lei M, et al. (2015) Regulatory link between DNA methylation and active demethy- heritance of seed dormancy in Arabidopsis thaliana. eLife 5:e19573. lation in Arabidopsis. Proc Natl Acad Sci USA 112:3553–3557. 41. Graeber K, et al. (2014) DELAY OF GERMINATION 1 mediates a conserved coat- 28. Yu A, et al. (2013) Dynamics and biological relevance of DNA demethylation in dormancy mechanism for the temperature- and gibberellin-dependent control of Arabidopsis antibacterial defense. Proc Natl Acad Sci USA 110:2389–2394. seed germination. Proc Natl Acad Sci USA 111:E3571–E3580.

E9970 | www.pnas.org/cgi/doi/10.1073/pnas.1812847115 Zhu et al. Downloaded by guest on September 26, 2021