DNA methylation modification in heterosis initiation through analyzing rice hybrid contemporary seeds

Shirong Zhou,Miqing Xing,Zhilong Zhao,Yinong Gu,Yunping Xiao,Qiaoquan Liu,Hongwi Xub,,*

a State Key Laboratory of Crop Genetics and Germplasm Enhancement,Jiangsu Plant Gene Engineering Research Center,Nanjing Agricultural University,Nanjing 210095,Jiangsu,China

b Joint Center for Single Cell Biology,School of Agriculture and Biology,Shanghai Jiao Tong University,Shanghai 200240,China

c National Key Laboratory of Plant Molecular Genetics,CAS Center for Excellence in Molecular Plant Sciences,Chinese Academy of Sciences,Shanghai 200032,China

d Shanghai OEbiotech,Shanghai 201210,China

e Co-Innovation Center for Modern Production Technology of Grain Crops of Jiangsu Province,Key Laboratory of Plant Functional Genomics of the Ministry of Education,Yangzhou University,Yangzhou 225009,Jiangsu,China

Keywords:DNA methylome Heterosis Rice Reciprocally hybrid seeds Embryo Endosperm

ABSTRACT Heterosis is an important biological phenomenon and widely applied in agriculture.Although many studies have been performed by using vegetative organs of F1 hybrid plants,how heterosis(or hybrid vigor)is initiated and formed,particularly the underlying molecular mechanism,remain elusive.Hybrid contemporary seeds of rice indica varieties 9311 and PA64 were innovatively used and analysis of DNA methylome of embryo and endosperm at early developing stages revealed the globally decreased DNA methylation.Genes,especially those relate to hormones function and transcriptional regulation present non-additive methylation.Previously identified heterosis-related superior genes are non-additively methylated in early developing hybrid contemporary seeds,suggesting that key genes/loci responsible for heterosis are epigenetically modified even in early developing hybrid seeds and hypomethylation of hybrid seeds after cross-pollination finally result in the long-term transcriptional change of F1 hybrid vegetative tissues after germination,demonstrating that altered DNA methylation in hybrid seeds is essential for initiation regulation and maintenance of heterosis exhibiting in F1 hybrid plants.Notably,a large number of genes show non-additive methylation in the endosperm of reciprocal hybrids,suggesting that endosperm might also contribute to heterosis.

1.Introduction

Heterosis,or hybrid vigor,is an important phenomenon in both animals and plants.In plants,heterosis presents that hybrids are superior to its genetically diverse parents with respect to many traits including seedling viability,photosynthetic efficiency,nutrient efficiency,growth rate,leaf area,flowering time,freezing tolerance,seed number and size,metabolite contents,biomass and yield [1].Although hybrids have been extensively exploited for increasing agronomic production,such as hybrid maize and rice,underlying molecular basis is still largely unknown.Three classic genetic hypotheses,i.e.dominance,over dominance and epistasis,have been proposed [2,3].

Through genetic mapping and genome-wide association study(GWAS),heterosis associated quantitative trait loci (QTL) or loci related to different traits have been identified and characterized from different species [4–6].In addition,transcriptome,proteome and metabolome of seedling,various vegetative or mature organs,of F1hybrid plants have been comparatively analyzed with those of two parents.Significant changes of multiple biological processes in F1hybrids,including the energy and metabolism,stress responses and aging,light and hormone signaling,flower and fruit development,indicate a global change of developmental processes after combining two distinct genomes of parents and interaction between genetic materials from parents will result in a genetic reprograming [1,7].

Cross ofArabidopsisecotype Landsbergerecta(Ler)and C24 produce hybrids with increased biomass and seed yield.Analysis of DNA methylome of Ler and C24 ecotypes suggests that heterosis may result from the differential epigenetic modification of a similar genome,which causes the changes of global transcription patterns in F1hybrid plants [8,9].Indeed,analysis using 15-dayold seedlings revealed the altered methylomes in hybrid offspring,which preferentially occur at the loci with differential methylation of parents.The context-specific biases of non-additive methylation patterns were observed,i.e.CG methylation generally increases and CHH methylation decreases comparing to those of parents.In addition,alterations in DNA methylation are correlated to the changed gene expression in hybrids [10].Another study on the methylome variations using same F1hybrid plants showed that cytosine methylation in all sequence contexts is increased across the entire genomes of both reciprocal hybrids,especially in transposable elements,and differentially methylated region of parents majorly contribute to the increased methylation in F1hybrid plants.Creating of epiHybrids by using epigenetic inbred lines(epiRILs) in test crosses to Col-0 reveals the importance of epigenetic regulation and epigenetic interactions in heterosis ofArabidopsis[11,12].In addition,growth vigor of F1hybrid plants was compromised by reduction of DNA methylation in hybrids,which highlights the role of DNA methylation in heterosis [13].

In rice,DNA methylation plays crucial roles in various processes,including seed development [14].Analysis of whole genome DNA methylome using rice shoots of 4-week-old seedling of Nipponbare (japonicavariety) and 9311 (indicavariety) and their reciprocal hybrids showed the varied DNA methylation of a multitude of genes in F1hybrid plants,especially the transposable element-related genes [15].Similar study by analyzing DNA methylome using leaves of 6-week-old seedlings showed the halfway methylation levels of F1hybrid plants between the parents[16].

Accumulating epigenetic,genomic,proteomic and metabolic data greatly broad the understanding on molecular basis of heterosis,although there are some inconsistence among different studies.However,it is noticed that most of the omics-based studies are performed by using vegetative organs of F1hybrid plants,while the alterations of various biological and physiological processes are indeed the consequences/performance of heterosis.The heterosis initiates since double fertilization that later develops into hybrid embryo and endosperm synchronously.Investigation of DNA methylome changes during early embryo and endosperm development of hybrid contemporary seeds will reflect the changes after hybridization as a result of cross-pollination and provides informative clues to help to elucidate the initiation,formation and regulatory mechanisms of heterosis.

Riceindicavarieties 9311 and PA64S (a sterility line derived from PA64) are the paternal and maternal parents of Chinese super-hybrid rice LYP9 (Liang You Pei Jiu),which is a successful rice hybrid cultivar showing strong heterosis [17].By using 9311 and PA64 as parental lines,early embryo and endosperm of hybrid seeds(6 days after cross-pollination)were used to analyze the DNA methylome.Globally decreased DNA methylations were detected,especially in endosperm,of reciprocal hybrids.Genes involving in various processes,particularly those related to transcriptional regulation and hormone functions,show non-additive methylation pattern in hybrids,suggesting the crucial role of DNA methylation of hybrid seeds in heterosis initiation and maintenance.

2.Material and methods

2.1.Plant material and growth conditions

Rice(Oryza sativaL.)varieties 9311 and PA64 were used to generate reciprocal hybrids.Hybrids are designated 93 × PA when 9311 was used as the maternal line,and PA × 93 when PA64 is used as the maternal line respectively.All rice plants are grown in fields under natural growing conditions in Shanghai,China.The cross-pollinated seeds are harvested 6 days after pollination,and embryos and endosperms were isolated for subsequent bisulfite-treated DNA sequencing.All panicles were harvested from plants around dusk each day and immediately frozen by liquid nitrogen to avoid the impact from circadian rhythm.DNAs were extracted from embryos and endosperms by using a cetyltrimethyl-ammonium bromide method and treated with RNase.

2.2.Bisulfite sequencing

The sequencing library was constructed with EpiGnome Methyl-Seq Kit (EGMK81312),and resultant libraries were paired-end sequenced by an illumina HiSeq 2000.Image analysis,base calling,and quality calibration were performed through the standard Illumina pipeline.Only high-quality reads were used for further analysis after filtering low-quality reads and reads containing primer/adaptor sequences by NGS QC Toolkit version 2.3 [18]with default parameters.Qualified filtered reads were aligned to the rice genome (http://www.mbkbase.org/R498/) using Bismark version v0.16.3 with default options [19].

2.3.Methylation analysis

The bulk methylation levels of each sample were calculated and represented by the ratio of methylated cytosines to all cytosines of the CG,CHG and CHH contexts.TEs and non-TE genes were obtained according toindicarice R498 genome annotation(http://www.mbkbase.org/R498/) and used to compare the distribution patterns of DNA methylation between TE and non-TE genes.For TEs,we extracted the coordinates of 3-kb upstream of TE start site(TES)region,3-kb downstream of TE terminal site(TET)region and TE region,then each of them was divided into 30 bins.Average CpG/CHG/CHH methylation values for each bin were calculated by averaging the methylation level of every CpG/CHG/CHH cytosine within the range.A similar profiling method was applied to non-TE genes,the methylation level of 3-kb upstream of transcription start site (TSS) region,3-kb downstream of transcription terminal site (TTS) region and gene body region were calculated and profiled.

2.4.Identification of non-additively methylated genes

The DMRs(differentially methylated regions)between two parents,between hybrids and average methylation levels of two parents(MPV),or between hybrids and parents,were analyzed with a dynamic fragmentation strategy using CGmap Tools [20].Only cytosines covered by bisulfite sequencing at least 5 times in both samples were selected for DMR analysis.

DMRs are dynamically defined using the default criteria,i.e.the length of DMR should be at least 100 bp,which contains at least 3 cytosines.An unpairedt-test is carried out to compare the methylation levels of shared cytosines within the fragment by thresholds ofP-values <0.05 and fold change of methylation levels at least 10%.

DMRs were grouped to different patterns according to combined results by comparing methylation levels between two parents,between hybrids and MPV,and between hybrids and parent.Taking in account that endosperm is triploid,the MPV was calculated by weighted mean of one paternal and double maternal methylation level of two parents.

DMRs were classified as additive when methylation level of a region in hybrid is equal to MPV,or as non-additive when methylation level in hybrid is deviated from MPV.Non-additively methylated DMRs are further subdivided into four patterns when methylation level of parents is different,‘‘High” or ‘‘Low” (methylation level of a hybrid is similar to the highest or lowest methylation level of parent),and‘‘Above”or‘‘Below”(methylation level of a hybrid is above or below that of both parent).In condition that parents are equal methylated,non-additively methylated DMRs are further subdivided into patterns‘‘Above”or‘‘Below”(methylation level of a hybrid is above or below that of both parents).

DMRs will be regarded as reciprocal non-additively DMRs when methylation change in both reciprocal hybrids show identical(e.g.both ‘‘Low” in 93 × PA and PA × 93) or similar (e.g.‘‘Low” in 93×PA and‘‘Below”in PA×93)methylation change trends.Those DMRs show non-additive methylation only in one hybrid (e.g.‘‘non-additive”in 93×PA but‘‘additive”in PA×93)or show conflict methylation change trends (e.g.‘‘Low” in 93 × PA and ‘‘High”in PA × 93) in reciprocal hybrids will not be considered.

To assign the non-additive DMRs to individual genes,only DMRs located within 2 kb of gene upstream or downstream regions and gene bodies are assigned to the nearby genes.Genes with multi assigned DMRs are treated as one non-additively methylated gene no matter the non-additive pattern of each DMR is accordant or conflict.

2.5.Gene ontology analysis

To annotate the functions of identified non-additively methylated gene sets,gene ontology information was obtained from gene ontology website (http://www.geneontology.org/),Rice Genome Annotation (http://www.mbkbase.org/R498/),and analyzed.Hypergeometric test was used to estimate the significance of overrepresentation and GO terms and performed multiple testing corrections by Benjamini &Hochberg False Discovery Rate (FDR)with a corrected FDR <0.05 and at least 5 annotated genes were kept.

3.Results

3.1.Decreased DNA methylation levels in developing seeds of reciprocal hybrids

To characterize the role of DNA methylation in rice heterosis initiation,indicavarieties 9311 and PA64 were used as parents.Both 9311 and PA64 are inbred cultivars (Oryza sativasubsp.indica) and have no common ancestor in traceable pedigrees.As one of the top 3 restore lines (ranking by hybrid rice growing area in China),9311 is widely used in hybrid rice production.PA64S(the derived male sterility line of PA64)is one of the top 10 Chinese male sterility lines (according to the ranking by either hybrid rice growing area or hybrid cultivar number,http://www.ricedata.cn/variety/index.htm),thus both 9311 and PA64 are representative parents that show heterosis in rice.

The reciprocal hybrids(designated as 93×PA or PA×93,PA64 or 9311 is the maternal line) indeed show obvious heterosis(Fig.1A).Both embryos and endosperms at early developing stages(6 day after pollination) from the self-pollinated seeds of parental lines and reciprocal cross-pollinated seeds were collected for bisulfite-treated DNA sequencing (Fig.1B).Due to trace amounts of samples,we paid much attention on the data quality.High quality methylome data were obtained by generating～150 million clean reads (～13 billion clean bases) for each genome (Table S1),which accounts for over 35 fold coverage relative to rice genome.All clean reads were mapped to the high qualityindicarice reference genome R498 [21] and all samples have a mapping rate over 89% (Table S2).False-positive methylation rate determined by methylated chloroplast genome sequences revealed that the bisulfite conversion rate was over 99% for all samples (Table S3),confirming the high quality of methylation data,which are used for further analysis.

Comparisons of bulk methylation levels of two parents by calculating methylated and unmethylated cytosines showed the similar methylation patterns,i.e.the highest and lowest level at CG or CHH respectively.Endosperm is significantly less methylated than embryo in all sequence contexts,and the most significant reduction of methylation was detected at CHH(Fig.1C).Further analysis of methylation patterns of parents and reciprocal hybrids by estimating the methylation percentage of each methylcytosine showed that the distribution of CG methylation is roughly similar in all examined samples,while CHG and CHH methylation are different in embryos or endosperms (Fig.2).There are more lower methylated and less higher methylated CHG in endosperm.Regarding CHH,there are more lower methylated sites in endosperm and more medium methylated sites in embryo,while more extremely higher methylated CHH sites in endosperm than in embryo.These results indicate that endosperm is hypomethylated both in parents and hybrids,which is consistent with the previous studies in rice andArabidopsis[14,22–24].

Although relatively fewer striking differences were detected between parents and hybrids,there is a significant decrease of methylation in the cross-pollinated hybrids,especially in embryos.The self-pollinated parental embryos have a bulk methylation ratio of 49.6%(9311)and 50.0%(PA64)in CG context,while in reciprocal cross-pollinated embryos,CG methylation ratio reduced to 45.2%(93 × PA) or 45.3% (PA × 93),which is equivalent to 91% of the mid-parent value (MPV,average of two parents).Similarly,bulk methylation ratio of CHG reduced to 89.5% and 91.1% of MPV,CHH reduced to 70.5% and 83.6% of MPV (Fig.1C).Decreased bulk methylation level is also observed in hybrid endosperms(PA×93,Fig.1C).In addition,analysis of methylation distribution revealed the decrease of relatively higher methylated sites and increase of relatively lower methylated sites in hybrid embryos at CHG and CHH contexts,while the CG context variation is less obvious(Fig.2).

Next,we compared the methylation profiles of transposable elements (TEs) and non-TE genes.Methylation patterns in each cytosine context between the TSS (transcriptional start site;TES for TEs,TE start site) and TTS (transcriptional terminal site;TET for TEs,TE terminal site)as well as 3-kb promoter and 3-kb downstream region of parents and hybrids were plotted.Although the methylation patterns (absolute methylation levels) of non-TE genes and TEs at CG,CHG and CHH are different in all examined samples,the general trends among parents and hybrid seeds are similar.Methylations at CG of non-TE genes significantly decrease near the TSS and TTS,with relatively high methylation level in gene body as well as promoter and TTS downstream regions is observed in both embryo and endosperm.Methylation level of gene body at CHG and CHH is significantly lower than those in promoter and TTS downstream region,both in embryo and endosperm (Fig.3A).The gene body of TEs showed relatively high methylation at CG and CHG,while with similar methylation at CHH,compared with the methylation levels at upstream and TET downstream region(Fig.3B).

Comparison of methylation profile of non-TE genes between parents and hybrids showed that CG methylations in reciprocal hybrid embryos are clearly lower than both parents at promoter and downstream regions of genes (Fig.3A).Similar trends were observed in TEs.Methylation level at CHG and CHH of hybrid embryos is also lower than parents in both non-TEs and TEs.A more obvious decrease of methylation in hybrids was observed in promoter and TTS downstream region comparing with gene body,especially in the non-TE genes.Interestingly,clearly decreased methylation at CG,CHG and CHH are detected both in non-TE and TEs in endosperm of PA × 93,but not in that of 93 × PA (Fig.3),which suggests that decreased methylation in endosperm of PA × 93 is not the effect from crosspollination but the dosage effect of reciprocal triploid endosperms or the parent of origin effects.These results confirm a decrease of DNA methylation of hybrid seeds,especially in hybrid embryos.

Fig.2.Distribution of methylation status.Percentage of methylation at CG,CHG,CHH cytosine context in embryos and endosperms of rice parents and reciprocal hybrids were shown.

Fig.3.Methylation patterns of 9311 and PA64 and cross-pollinated hybrid contemporary seeds.DNA methylation profiles of non-TE genes(A)and transposable element(TE,B)at three cytosine contexts in embryo and endosperm of the parents and hybrids were shown.The R498 annotated genes or transposons were aligned at 5′ or 3′ end,and distributions of average methylation level in gene body and 3-kb flanking sequences on both sides were plotted.TSS and TTS indicate the transcriptional start site ortranscriptional terminal site of non-TE gene.For TE,the start site and end site are indicated as TE start site (TES) or TE terminal site (TET),respectively.

3.2.Hypomethylation of hybrid embryo and endosperm are differentially associated with parental methylation status

To investigate the association of DNA methylation of hybrids with parents,all methylated cytosines of parental embryos and endosperms are firstly divided into four categories according to methylation levels,including those of 9311 is higher methylated than PA64 (9311 >PA64),those of 9311 is lower methylated than PA64 (9311 0),no methylation in those of 9311 and PA64 (9311=PA64=0).

In hybrid embryos,in all four categories and each context,the methylation level was different from the MPV as determined by paired-samplettest.However,the MPV and hybrids in each category was not consistent regarding the difference extent.Methylation levels of CG and CHG in hybrids are nearly same to the MPV when parents are differentially methylated,but equally methylated CG and CHG sites of parents are slightly hypomethylated in hybrids.The unmethylated CG and CHG sites keep the parental status or show an imperceptible increase of methylation(Figs.4A,S1,S2).Nevertheless,CHH site of hybrids show reduced methylation level compared to MPV no matter it is differentially or equally methylated in parents,with a slight increase of methylation of unmethylated sites of parents (Figs.4A,S3).

In hybrid endosperms,in all four categories and each context,the methylation level was different from the MPV similar as that in embryo.The difference extent was varied among different contexts in each category.Methylation at CG and CHG sites of hybrids is close to MPV or they are slightly hypomethylated,however,both differentially methylated or equal methylated CHH sites of hybrids are hypomethylated (Figs.4B,S1,S2,S3).

These results indicate that in hybrid embryos and endosperms,hypomethylation mainly occurs at three contexts,especially those were parentally equally methylated,while for parental differentially methylated cytosines,hypomethylation mainly occurs at CHH sites of hybrid embryos and endosperms.

3.3.Identification of genes with non-additive methylation

DNA methylation generally suppresses gene expressions and genes showing non-additive methylation in hybrids were particularly analyzed.After extracting the methylation data for all samples,differentially methylated regions (DMRs) in any of two samples were identified as described in the method section.Identified DMRs were then grouped into different patterns after comparing the methylation levels between two parents,between hybrids and MPV,and between hybrids and parents,i.e.‘‘additive”when methylation level of a region in hybrid is equal to MPV,or‘‘non-additive” when methylation level of a region in hybrid is deviated from MPV.

In embryo,a total of 4440 DMRs in CG context,3527 DMRs in CHG context and 11,660 DMRs in CHH context show nonadditive methylation in both reciprocal hybrids.In endosperm,6492 CG context,6329 CHG context and 1547 CHH context DMRs show non-additive methylation in both reciprocal hybrids(Fig.4C;Table S4).Most of the non-additive DMRs are parentally equally methylated,except for CHG methylation of embryo (53% of nonadditive CHG DMRs are parentally differentially methylated)(Fig.4D),which suggests that non-allelic epigenetic interaction in seed significantly contribute the heterosis.To further investigate how methylation in reciprocal non-additive DMRs is changed in hybrids,we further detailedly subdivided these DMRs into six patterns,‘‘High”or‘‘Low”(methylation level of reciprocal hybrids are similar to the highest or lowest methylation level of parents),‘‘Above” or ‘‘Below” (methylation level of reciprocal hybrids are above or below that of both parents),‘‘High/Above” or ‘‘Low/Below” (methylation level in one hybrid is similar to the highest or lowest methylation level of the parent,but the reversed hybrid is above or below that of both parents).

Analysis showed that ‘‘Below” is the major pattern of DMRs in both embryo and endosperm,and in all contexts(Fig.4E),suggesting that methylation level of both reciprocal hybrids are lower than the lowest parent in major of the non-additive region,and the over-dominant demethylation in hybrids happens at CG,CHG and CHH sites.

To assign the non-additive DMRs to individual genes,only DMRs located within 2 kb of gene upstream regions,gene bodies and 2 kb of gene downstream regions were assigned to the nearby genes.A total of 2792,2199,and 8692 genes showing non-additive methylation in both reciprocal hybrids in CG,CHG,and CHH context were identified in embryo,and 5163,3067,and 1227 genes in CG,CHG,and CHH context respectively were identified in endosperm (Fig.5A;Table S5;methylation levels of three circadian rhythm-related genes were shown as representative,Fig.S4).Few genes have multi DMRs assigned and show different nonadditive patterns among DMRs,and only a small portion of these genes are commonly found in all three CG,CHG,and CHH contexts.In sum,10,958 and 7632 non-redundant genes in embryo or endosperm show non-additive methylation in both reciprocal hybrids(Table S5).

3.4.Previously known heterosis-related superior genes and QTL show non-additive methylation in early developing hybrid seed

Although vegetative organs of F1hybrid plants display heterosis,initiation of which might be determined since double fertilization.Differential methylation in the early developing embryos and endosperms of hybrid contemporary seeds could result in the altered gene expression during vegetative growth of F1hybrid plants.Previous genome-wide transcriptional profiling studies using 9311,PA64s and LYP9 have identified a lot of differentially expressed genes that show non-additive expression [7,25],which were compared to the genes with non-additive methylation detected in our study.Of all 2581 genes showing different expressions between LYP9 and its parents(heterosis-related genes),637 genes in embryo showed non-additive methylation (Table S6),and these genes are enriched in gene ontology classes of metabolic process,oxidation–reduction process,carbohydrate metabolic process,and photosynthesis.Meanwhile,466 genes in endosperm showed nonadditive methylation(Table S7),and they are enriched in the categories of metabolic process,carbohydrate metabolic process,photosynthesis,and metal ion transport(Fig.S5).

Fig.4.Association of DNA methylation in hybrids with parental methylation status.(A,B)Average genome methylation levels at three cytosine contexts in embryos(A)and endosperms(B)of parents and hybrids.Four groups are classified based on the cytosine methylation levels,including cytosine contexts of 9311 are higher methylated than those of PA64(9311 >PA64);cytosine contexts of 9311 is lower methylated than PA64(93110);cytosine contexts showing no methylation in 9311 and PA64(9311=PA64=0).MPV,average methylation level of the two parents.(C)Statistic of DMRs (differentially methylated regions) and genes showing non-additive methylation in both reciprocal hybrids (93 × PA and PA × 93).(D) Comparison of methylation status of DMRs showing non-additive methylation in both reciprocal hybrids(PA×93 and 93×PA)to the parents.(E)Non-additive patterns of DMRs showing non-additive methylation in both reciprocal hybrids PA × 93 and 93 × PA.Reciprocally non-additive DMRs were classified when methylation change in both reciprocal hybrids is identical (e.g.both ‘‘Low” in 93 × PA and PA × 93) or shows similar change trends (e.g.‘‘Low” in 93 × PA and ‘‘Below” in PA × 93).‘‘High” or ‘‘Low” means methylation level of reciprocal hybrids is similar to the highest or lowest methylation level of parents,and ‘‘Above” or ‘‘Below” means the methylation level of reciprocal hybrids is above or below that of both parents.‘‘High/Above”means similar methylation change trends of reciprocal hybrids that show‘‘High”in one hybrid and‘‘Above”in reversed hybrid.‘‘Low/Below” have the similar meaning.In condition that parents are equal methylated,non-additively methylated DMRs are further subdivided into patterns ‘‘Above” or ‘‘Below” (methylation level of one hybrid is above or below that of both parents).

Fig.5.Non-additively methylated genes in cross-pollinated hybrid seeds.(A) Statistics of non-additively methylated genes of embryo and endosperm in both reciprocal hybrids PA × 93 and 93 × PA.(B,C) Enriched GO processes of non-additively methylated genes in embryo (B) and endosperm (C) in both reciprocal hybrids PA × 93 and 93×PA.Hypergeometric test was used to estimate the significance of overrepresentation and GO terms,and performed multiple testing corrections by Benjamini&Hochberg false discovery rate (FDR) with a corrected FDR <0.05 and at least 5 annotated genes were kept.

Further comparison between methylation status of hybrid contemporary seeds and expression level in vegetative organs of F1hybrid plants at different stages resulted in the identification of 250 and 154 genes with negative correlation between expression and methylation in embryo and endosperm,respectively (Tables S8,S9).Interestingly,these genes are mostly enriched in categories of metabolism and transport,suggesting that the altered gene expressions in F1hybrid plants maybe directly regulated by DNA methylation in early stage of hybrid seeds.

In addition,non-additively methylated genes involve in circadian rhythm,flowering time regulation and panicle development show non-additive expression in LYP9 F1plants [7],which is consistent with study usingArabidopsisvegetative tissues [26].Interestingly,among the total 61 identified non-additively expressed genes in F1hybrid plants(LYP9,Table 1)related to above processes,21 genes were found being non-additively methylated in early developing embryos,including 7 of 16 known flowering-related genes (OsCO3,OsDof12,OsMADS55,OsMADS14,OsMADS1,OsMADS47,OsMADS50/OsSOC1),12 of 34 circadian-associated factors(OsPRR37,OsPRR73,OsCCA1,OsLHY,OsZTL2,OsFTL12,OsCNR11,OsCNR10,qKGW1,SRS3,GIF1,andDEP1),and 2 of 11 paniclebranching regulators (TAW1,DEP1).

Circadian clock genes play important roles in heterosis by regulating the carbon metabolism and stress responses in F1hybrid plants,and non-additive expression of these genes was found both inArabidopsisand rice [1,7,26–29].OsCCA1,OsLHY,OsPRR37,andOsZTL2are the putative circadian clock core components and show non-additive methylation both in embryo and endosperm,suggesting that methylation status has been changed in early developing hybrid seeds.OsPRR37/DTH7/Hd2,a major quantitative locus(QTL) determining photoperiod sensitivity and grain yield in rice[30],DEP1,a key gene determining rice panicle architecture,high nitrogen using efficiency and high yield [31,32],andGIF1,a cellwall invertase required for carbon partitioning during early grain-filling [33],also show non-additive methylation in hybrids.In addition,genes related to flowering includingOsPRR73,OsCO3,OsMADS50/OsSOC1,TAW1,andOsEMF1[7] are non-additively methylated as well in hybrid embryos or endosperms.These results indicate that altered methylation status in early developing hybrid seeds might cause the transcriptional changes of related gene during vegetative growth of F1hybrid plants.

GWAS of genetic basis of rice heterosis revealed multiple loci for yield related phenotypes[4].Analysis of the methylation status of these loci showed that major QTLDEP1,Ghd7(for heading date),NAL1(related to high yield),andTMS5also present non-additive methylation in hybrids (Table 2).

A previous proteomics study using rice leaves of F1hybrid plants identified some non-additively expressed proteins that involve in photosynthesis,glycolysis and stress response [34],and some of them indeed show non-additive methylation in hybrid embryos and endosperms (Table 3),further indicating a possible regulation of them by DNA methylation and suggesting that key genes/loci responsible for heterosis are epigenetically modified in hybrid seeds even at early developing stages (before seed maturation).

3.5.Non-additive methylation-regulated hormone signaling and transcription-related genes in heterosis

To investigate the primary biological progresses regulated by DNA methylation in heterosis,we further analyze the gene ontology (GO) enrichment of non-additively methylated genes.Results showed that GO terms including metabolic process,response to hormones,oxygen transport,negative regulation of transcription,steroid biosynthetic process and phospholipid transport are enriched in embryo non-additively methylated genes (Fig.5B).

Phytohormones play crucial roles in plant growth and development.Analysis revealed that a total of 191 genes involving in phytohormone synthesis or signaling are non-additively methylated in hybrids’embryo(Table S10),including those related to biosynthesis,polar transport,perception and signaling of auxin;biosynthesis,inactive and signaling of cytokinin and abscisic acid;as well as biosynthesis and signaling of brassinolide,gibberellin,ethylene,jasmonic acid,salicylic acid and strigolactone.Interestingly,majority of these genes are CHH context non-additive,and most of them show‘‘Below”or‘‘Low”pattern,suggesting the hypomethylation of these genes in hybrid embryos.

Table 1 Identified flowering,circadian rhythm and panicle branching-related genes that are non-additively transcribed in LYP9 show non-additive methylation in hybrid embryo or endosperm.

Table 2 Identified heterosis genes show non-additive methylation in hybrid embryo.

Transcriptional regulation and histone methylation related biological progress are enriched in non-additively methylated genes of reciprocal cross-pollinated embryos,and further analysis identified 665 transcription factors or regulators (including histone methylation related factors) which belong to different families and involve in regulation of various aspects of plant growth,development and stress resistance (Table S11).Most of these factors show ‘‘Low” methylation pattern in hybrids,suggesting a broad range hypomethylation of transcription factors that might cause the expression change of them in F1plants,hence regulating the growth and development,as well as stress responses of F1plants,which is somehow consistent with the study,by using various tissues of F1plants,showing the global change of multiple genes;however,whether expression of these genes are indeed regulatedby altered methylation and hence to regulate the downstream genes need further investigations.

Table 3 Identified proteins that are non-additively expressed in LYP9 show non-additive methylation in hybrid embryo or endosperm.

3.6.Genes involved in various biological processes show non-additive methylation in hybrid endosperms

Endosperm provides nutrients and is essential for embryo development and seed maturation.As much as 7632 genes show non-additive methylation in reciprocal hybrid endosperms,which involve in various biological processes and terms including DNA methylation,response to water,SRP-dependent co-translation,Larabinose metabolic process,transcription,DNA-templated,metabolic process,nitrogen compound metabolic process are enriched(Fig.5C;Table S5).Interestingly,many non-additively methylated genes in endosperms are related to endosperm development and synthesis of storage substances,includingOsFIE1andENL1(endosperm development),GluA,GluB-1,GluB6,andOsVPE1(protein synthesis),ISA3,FLO2andSSG4(starch synthesis);GL3.1,GW2,GW7,GS2,andOsDA1(grain size).However,21 previously identified non-additively expressed genes in LYP9 F1plants [7],which are reported to participate in circadian rhythm,flowering time regulation and panicle development,also show a non-additive methylation pattern in hybrid endosperm (Table 1),suggesting that methylation change in endosperm probably plays a role other than regulating the nutrient storage and endosperm development.

4.Discussion

Hybrid plants starts from double fertilization and study of maize early seed development reveals that heterosis emerges in early embryos shortly after hybridization [35].Presented study by analyzing DNA methylome of embryos and endosperms at early developing stage after cross-pollination (technically realistic and earliest samples can be obtained currently) demonstrates the decreased DNA methylation in reciprocally hybrid seeds,and suggests the global reprograming of DNA methylation in hybrids.The facts that many previously identified heterosis-related superior genes and differentially expressed genes of F1hybrid plants show the non-additive methylation in contemporary hybrid seeds indicate that altered DNA methylation in contemporary hybrid seeds likely affect and maintain the heterosis exhibiting in F1hybrid plants,which demonstrates the crucial roles of DNA methylation in rice heterosis determination and significantly expands the understanding on the molecular mechanism of this important phenomenon.

4.1.Hypomethylation of hybrid seeds after cross-pollination

Epigenetic reprogramming includes demethylation of DNA and remodeling of histones,which has been reported in flowering plants and mammals [36].Large-scale epigenetic reprogramming was observed inArabidopsisendosperm [23],and a recent study revealed the mechanism for the reprogramming of embryonic chromatin states inArabidopsis.The activeFLCchromatin state that is established in pro-embryo can be transmitted through cell division to post-seed development stages,leading to the epigenetic memory of embryonic active gene expression in post-embryonic phases [37].Indeed,a significant hypomethylation was detected in hybrid contemporary embryos and endosperms (embryos show much decrease),which may result in the long-term activation of lots of genes.Significantly changed DNA methylation in vegetative tissue of F1hybrid plants[10,13,15,16]may due to the methylation changes in hybrid embryos at early stages after cross-pollination and the ‘memory’ of embryonic methylation status may result in the change of transcriptome of F1hybrid vegetative tissues after germination.

Studies using 15-day-oldArabidopsisseedlings showed the increased DNA methylation in reciprocal F1plants [13],while that using rice leaves of 6-week-old plants revealed that global DNA methylation level of reciprocal hybrids were halfway between those of parents [16].Our analysis revealed an apparently decreased DNA methylation level in both reciprocal hybrid contemporary seeds,indicating a significant DNA demethylation after hybridization.InArabidopsisF1hybrid plants,the sites with increased methylation are mainly in those with differential methylation of parents,while in rice hybrid contemporary seeds,the sites with decreased methylation are mainly in those with similar methylation of parents,suggesting that altered methylation of hybrid contemporary seeds may not be the interaction of allelic genes,but the interaction of non-allelic genes.

4.2.Complex regulation of heterosis through various processes

Comparative transcriptome analysis of different vegetative tissues across developmental stages among 9311,PA64s and LYP9 has identified some candidate heterosis-related genes[7,25].Comparative proteomics studies between LYP9 and its parents identified 222 unique differentially expressed proteins,of which 119 non-additive proteins are up-regulated [34].Genomic analysis of hybrid rice varieties to investigate 38 agronomic traits through genome-wide association study(GWAS) resulted in the identification of 130 heterosis associated loci [4,5].In addition,studies revealed a crucial role of epigenetic changes in hybrid nucleus[8].Our analysis showed that amounts of genes,particularly transcriptional regulation related genes,present non-additive methylation in early developing hybrid seeds,and most of them show‘‘Low-parent” pattern.Reduced methylation of key transcription factors may result in the altered transcription of their target genes and further contribute to heterosis of F1hybrid plants.

A lot of overlap of heterosis related non-additively expressed and non-additively methylated genes in hybrid seeds at early developmental stages (Tables 1,S6) indicates the accuracy of our study.However,it is noticed that there are still some nonadditively expressed genes don’t show non-additive methylation in hybrid contemporary seeds,such as theRH8/DTH8/Ghd8/LHD1(Fig.S6),a key gene contributing to heterosis in LYP9.Expression of these genes may not be regulated directly by DNA methylation,instead by other DNA methylation regulated genes or with a methylation independent manner.By using the early developing hybrid contemporary seeds to study the heterosis will help to shield the secondary effect,reflecting an advantage of this approach.

Genes involved in biosynthesis and signaling of various hormones were detected being non-additively methylated in hybrid embryos,suggesting an important role of hormones in heterosis.Auxin contributes to hybrid vigor inArabidopsis[38,39],and quite a lot of auxin biosynthesis related genes show ‘‘Low/below” nonadditive methylation pattern suggests these genes are hypomethylated in hybrid seeds and might be activated in young embryos.‘‘Low/below” methylation pattern of many cytokinin signaling related genes in hybrids’embryo will activate the cytokinin signaling to regulate cell proliferation and differentiation of hybrids.Studies in maize reveal the enhanced gibberellin (GA) sensitivity and endogenous GA content in hybrids [40],as well as the rice heterosis of F1hybrid seedlings [41].Our studies indicated that GA synthesis is much enhanced in developing hybrid seeds as several genes involved in GA biosynthesis (CPS:3,KS:3,KS:6,KS:8,KO:4 and KAO) show ‘‘Low-parent” pattern in both reciprocal hybrid embryos,which might stimulate the long-term expression of these genes in F1hybrid plants.It is suggested that auxin and GA may play key roles in heterosis through epigenetics modification.

4.3.Potential roles of endosperm in heterosis

Comparing to the obviously decreased methylation in embryos of both reciprocal hybrids,it is interesting to observe a bulk significantly decreased methylation in the endosperms of PA × 93 hybrid but less so in those of 93 × PA.This might because endosperm is a triploid tissue,i.e.in contrast to the 1:1 ratio of maternal to paternal genomes in embryo,the ratio of maternal to paternal genomes is 2:1 in endosperm.The reciprocal hybrids have different gene dosage in endosperm,resulting in a parent-origin effect of global DNA methylation in endosperm.Considering that both 93 × PA and PA × 93 hybrids show obvious heterosis,and a bulk decreased methylation is only detected in PA × 93 endosperm,it is suggested that DNA methylation change in PA × 93 endosperm does not globally contribute to hybrid vigor and the non-additively DMRs identified in both reciprocal hybrids attribute to related heterosis.

Triploid endosperm is usually regarded as a sink tissue and of great importance for acquiring various resources from maternal tissue for young hybrid embryo development.In addition,endosperm is crucial for hybrid seed germination and provides nutrition at early stage of seedling growth [42,43].Our studies showed that quite a lot of genes show non-additive methylation in hybrid endosperms in addition to embryos (Table S5),which suggests that endosperm also contributes to heterosis,potentially by regulating the embryo development.

Genes involving in ‘‘endosperm unrelated” process show nonadditive methylation in hybrid endosperms,such as flower development regulating genes.In addition to the possible multifunction of these genes in different tissues,there is possible comprehensive communication between endosperm and embryo,although still less is known yet.InArabidopsis,it is proposed that small RNAs generated in endosperm might direct non-CG DNA methylation in embryo [44,45],which could be a hint for this interaction.Processes related to various aspects are enriched in endosperm,indicating the complex regulation of heterosis and raising the possibility that DNA methylation change in hybrid endosperm may affect the gene expression in embryo and contribute to heterosis indirectly.

In summary,global hypomethylation of hybrid contemporary seeds results in activation of a large numbers of genes in F1hybrid plants,indicating that DNA methylation in hybrid seeds is essential for initiation and maintenance of heterosis.Non-additive methylation of various genes in hybrid embryo will have long-term effects in regulating vegetative growth and development of F1hybrid plants.In addition,endosperm might indirectly contribute to heterosis possibly by communicating with and regulating the development of embryo.

5.Data availability

The raw data generated of this study are available in Sequence Read Archive(SRA)database of the National Center for Biotechnology Information (https://trace.ncbi.nlm.nih.gov/Traces/sra/) under accession numbers SRP133858.

CRediT authorship contribution statement

Shirong Zhou,Meiqing Xing,and Hongwei Xue:designed the projects.Shirong Zhou and Zhilong Zhao:collected the materials.Shirong Zhou,Yincong Gu,and Meiqing Xing:analyzed the data.Yunping Xiao:helped the sequencing.Shirong Zhou and Meiqing Xing:drafted the manuscript.Qiaoquan Liu:discussed the project and helped to draft the manuscript.Hongwei Xue:wrote the paper.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the Ministry of Science and Technology of China (2012CB944804),the National Transformation Science and Technology Program (2016ZX08001006-009),and the National Key Research and Development Program of China(2016YFD0100501,2016YFD0100902).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.12.003.