The OsDDM1b T-DNA insertion mutant (PFG_2B-60109) and wild-type cultivar Hwayoung were obtained from the Crop Biotech Institute, Department of Plant Systems Biotech, Kyung Hee University. The T-DNA information was obtained from SIGnAL database at http://signal.salk.edu/cgi-bin/RiceGE. The plants were grown in the greenhouse at 22–32°C and 80–90% humidity with a 14-h/10-h (light/dark) photoperiod. About 1, 3, 7, 10, and 25 days after fertilization (DAF) of seeds were harvested using micro-dissection needles, frozen in liquid nitrogen immediately, and stored at -80°C for total RNA extraction. A total of three biological replicates of each sample were collected for experimental analysis.

The T-DNA insertion in osddm1b was confirmed by PCR using the primers 2B-60109-Lp and 2B-60109-Rp and the T-DNA-specific primer Rb (2707). The full-length coding sequence of OsDDM1b was amplified from WT cDNA by PCR and cloned into the EcoRI/KpnI sites of the pCAMBIA2300 vector. The derived constructs were introduced into the osddm1b by Agrobacterium tumefaciens-mediated transformation. The relative expression levels of OsDDM1b were detected in 7-day-old seedling leaves of osddm1b and complementary mutants by the primers P1 and P2. The primers sequences are described in Supplementary Table 2 for genotyping identification. The gene structure of OsDDM1b was searched from Gene Structure Display Server.¹

Total RNAs of all collected samples were extracted using Plant RNeasy Mini kit (Qiagen, Hilden, Germany). About 1 μg RNA was reverse-transcribed using the PrimeScript RT-PCR kit (Takara, Kyoto, Japan) according to the manufacturer’s instruction (Cai et al., 2019). The relative expression level was detected by qRT-PCR using the Bio-Rad qRT-PCR system (Foster City, CA, United States) and SYBR Premix Ex TaqII (TaKaRa Perfect Real Time) (Zhang et al., 2020). The qRT-PCR program was 95°C for 30 s; 40 cycles of annealing at 95°C for 5 s and extension at 60°C for 35 s; and 95°C for 15 s (Zhang et al., 2020; Zhao et al., 2021). The rice OsUBQ5 gene was used as an internal control. To evaluate the relative expression levels of the examined genes, we used the comparative C_T method (Su et al., 2017). The gene-specific primers are listed in Supplementary Tables 3–6.

Several anthers were randomly selected before pollination and placed on microscopic slide for dissecting with the micro-dissection needles using a microscope. The pollen was stained with 1% I₂-KI solution and observed under the microscope (Ren et al., 2019). For the analysis of the ovule fertility, the spikelets before pollination were fixed in the FAA solution (50% ethanol: acetic acid: formaldehyde = 89:6:5) and using the whole-mount eosin B-staining confocal laser scanning microscopy (WECLSM) to observe the ovule development (Zeng et al., 2007, 2009). After fixation, eosin B staining, and clearing, the ovaries were divided from spikelets and observed using a Leica SP8 CLSM (Leica Microsystems) to screen the ovule (Zhao et al., 2020).

A 2,559-bp fragment upstream of OsDDM1b ATG codon sequence was amplified by PCR from DNA of rice leaf using the primers listed in Supplementary Table 3. Then, the products were constructed into a pENTER/D-TOPO vector (Invitrogen, CA, United States). After that, the positive clones were recombined with the pGWB533 vector by LR Clonase II enzyme (Invitrogen, CA, United States). The wild-type ZH11 (Zhonghua 11) callus was transformed using the Agrobacterium-mediated transformation using the pOsDDM1b: green fluorescent protein (GUS) recombinant construction. The transgenic plant tissues were incubated in β-glucuronidase (GUS) staining buffer overnight at 37°C (Jefferson et al., 1987) and dehydrated in an ethanol series to remove the chlorophyll (Jiang et al., 2012). The images were viewed under a Leica (M205 FA) microscope.

A 2,547-bp segment of OsDDM1b coding sequence was amplified from WT cDNA using the primers listed in Supplementary Table 3. The PCR fragments were cloned into the pENTER/D-TOPO vector (Invitrogen, CA, United States), and pENTER/D-TOPO clones were recombined into the pGWB505 vector using LR Clonase II enzyme (Invitrogen, CA, United States). The 35S:OsDDM1b-GFP recombinant construction and 35S: GFP (vector control) were transformed into Agrobacterium tumefaciens (GV3101) and infiltrated with ER-mCherry marker to tobacco leaves. The fluorescence signals were observed using a confocal microscope (SP8, Leica, Germany), and the excitation wavelength was 488 nm.

The spikelets were fixed in FAA (50% ethanol, 5% acetic acid, and 3.7% formaldehyde) at 4°C overnight and dehydrated in a series of ethanol for observing the spikelet cell size and number. After fixing with chloroform, the samples were embedded in Paraplast Plus (Sigma). The samples were sliced into the 8-μm-thickness samples using an RM2245 rotary microtome (Leica). Sections were dewaxed in xylene and gradually rehydrated and dehydrated before staining with toluidine blue for light microscopy. For scanning electron microscope (SEM), fresh materials were applied directly to the scanning electron microscope (Ren et al., 2018; Yu et al., 2020). The cell number and cell size in the outer parenchyma layer of the spikelet hulls were measured by ImageJ² (Yu et al., 2020).

For the lamina joint test, the seeds were germinated in water at 37°C and grown hydroponically in a nutrient solution containing 0, 0.01, 0.1, and 1 μm epi-BL for 1 week. Then, the second leaf lamina joint was measured. For the coleoptile elongation assay, the seeds were sterilized with 10% hypochlorous acid and germinated under dark conditions in half-strength MS medium supplemented with 0, 0.01, 0.1, and 1 μm epi-BL. Coleoptile length were measured after a week (Liu S. Y. et al., 2015).

Plants were grown hydroponically in nutrient solution for 10 days. The shoots and roots (equivalent to 1 g of fresh weight) of wild type and osddm1b were harvested and ground with phosphate-buffered saline solution immediately. BR endogenous contents were analyzed using the BR ELISA kit according to the manufacturer’s protocol (SINOBESTBIO, Shanghai, China) by UV spectrophotometer in 450 nm.

The development of seeds is important to get high yield, and many genes highly regulate the developmental process. To explore the expression level of the Snf2 family genes during seed development, we collected the seeds from 1, 3, 7, 10, and 25 DAF. We checked the gene expression level shown in Figure 1 and Supplementary Figure 1. The results showed that OsCHR712 and OsCHR730 were predominantly expressed in early developmental stages of seeds (S1), and OsCHR703, OsCHR735, and OsCHR741 were preferentially expressed in both S1 and S5 stages (Figure 1A). In addition, OsCHR701, OsCHR707, OsCHR711, OsCHR739, and OsCHR745 were preferentially expressed in both S1 and S4 stages (Figure 1B). There were 23 Snf2 genes that showed similar expression pattern which were predominantly expressed in the S4 stage, such as OsCHR702, OsCHR704, OsCHR705, OsCHR706, OsCHR708, OsCHR709, OsCHR710, OsCHR713, OsCHR715, OsCHR719, OsCHR721, OsCHR722, OsCHR725, OsCHR727, OsCHR729, OsCHR731, OsCHR732, OsCHR736, OsCHR737, OsCHR742 (Figure 1C), OsCHR720, OsCHR726, and OsCHR740 (Supplementary Figure 1A). These results indicated that Snf2, a sizeable family that exhibits specific expression patterns, may have unique functions in different developmental stages of seed in rice.

The previous studies showed that two homologous genes in rice (Oryza sativa var. Nipponbare) offered 60% identity to DDM1 (AT5G66750) (Higo et al., 2012). The two genes, OsDDM1a (LOC_Os09g27060) and OsDDM1b (LOC_Os03g51230), share 93% amino acid identity to each other (Higo et al., 2012; Tan et al., 2016). Here, we identified a T-DNA insertion approximately 0.55 kb downstream of the translation start site ATG in osddm1b (2B-60109.L) from Hwayoung background (Figure 2A), which were identified in the SIGnAL database.³ The homozygous plants for the T-DNA insertion were identified and obtained by PCR (Figure 2H). In addition, the PCR products of homozygous plants were sequenced, and the sequencing results were aligned with the genome sequence (OsDDM1b) and vector sequence (2707), and then, we found the T-DNA insertion site (Supplementary Figure 3).

In osddm1b, the dwarfism was observed in the seedlings and mature plants (Figures 2B,I and Supplementary Figure 2A and Table 1). The length of first internode was similar to the wild type. However, the second, third, and fourth internodes of osddm1b were shortened, which suggests that abnormal internode elongation leads to dwarf phenotype (Figure 2J and Supplementary Figure 2B). The reproductive organs of the osddm1b, such as spikelets, anthers, and pistils, were much smaller than those of wild type (Figures 2C–F). The panicle morphology in osddm1b exhibited a significantly reduced growth of 11.03% than the wild type at mature stage (Figures 2G,K). Then, we compared male and female gametophyte morphology in the osddm1b and wild-type plants. The staining of pollen with iodine potassium iodide (I₂-KI) solution and the observation of ovule with eosin B by confocal showed that the male (Supplementary Figures 2C–D) and female gametophytes (Supplementary Figures 2E,F) have no noticeable difference compared to the wild type. However, the seed setting in osddm1b was decreased to 84.39% (Supplementary Figure 2G).

To further confirm the effect of OsDDM1b on grain size, we examined the grain length, width, thickness, and 1,000-grain weight. The results showed that osddm1b had significantly reduced grain length of 8.15% than the wild type (Figures 3A,D, Supplementary Figure 2I,J), but it increased 17.59% in grain width (Figures 3B,E and Supplementary Figures 2K,M) and 10.67% in grain thickness, respectively (Figures 3C,F and Supplementary Figures 2L,N). Even though there was a change in the grain size, there was no noticeable change in the 1,000-grain weight (Supplementary Figure 2H). To verify the role of OsDDM1b in determining grain size in rice, we performed a genetic complementation test. An overexpression plasmid containing full-length coding sequence was inserted by the transformation into the osddm1b. Then, we got two positive complementary lines. The overexpression of OsDDM1b in osddm1b could partially complement the grain size defection of mutant (Figures 3A–C). The grain size (length, width, and thickness) was significantly different between the complementary lines and osddm1b (Figures 3D–F). Particularly, the grain width in the complementary lines was similar to wild type (Figure 3E). The OsDDM1b expression level was dramatically reduced in 7-day-old young leaves in osddm1b but increased in the complementary lines compared with those of the wild type (Figure 3G). The other defection phenotypes could be partially rescued in the complementation lines. This result indicated that the mutation in OsDDM1b was responsible for the phenotype of grain size exhibited in osddm1b.

The spikelet hullin osddm1b became shorter and more comprehensive compared with wild type before fertilization (Figure 4A). Since the cell division and/or cell expansion are responsible for the alteration in the final spikelet hull size and grain size, we carefully examined the cross-section of the central part of lemma and palea in mature spikelets and compared the epidermal cells of osddm1b and HY by scanning electron microscope. The hull cross-section of the spikelet hullin osddm1b revealed a significant increase of 27.84% in the length of the total cells and 8.15% in cell number of outer parenchyma in the osddm1b. In comparison, the cell width was decreased by 18.07% in the outer parenchyma cells (Figures 4B–F). It also changed the number of rows of specialized cells with a rigid wall in the upper epidermis (Supplementary Figures 4A–B). The data suggested that the cell division was significantly increased in a transverse direction in the osddm1b spikelet hulls. We also investigated the expression of genes involved in the cell cycle, such as CYCA2.2, CYClaZm, E2F2, CYCB2, MCM3, MCM4, and MCM5. We found that these genes were upregulated in young panicles of osddm1b, which suggests that the increased cell number might have resulted from the elevated expression of genes that promote cell proliferation (Supplementary Figure 4C). In addition, we observed the average length and width of longitudinal epidermal cells in outer and inner glumes of wild type and osddm1b by scanning electron microscopy. An apparent 58.43% decrease in cell width was observed in the outer glum, but 10.41% in cell width was increased in the inner glum. Then, there is no significant difference in outer and inner glumes’ cell length in osddm1b (Figures 4G–I). These observations suggest that OsDDM1b might promote latitudinal growth by increasing cell proliferation.

To investigate the subcellular localization of OsDDM1b, we performed a transient transformation assay. When OsDDM1b-green fluorescent protein (GFP) was transformed into tobacco leaves, the signals mainly targeted the nucleus and cell membrane (Figure 5A).

To further examine the expression profile of OsDDM1b, a construct was produced in which the OsDDM1b promoter (approximately 2.5 kb upstream of the ATG site) was fused to the β-glucuronidase gene (GUS) to transform wild-type (ZH11) plants. GUS signals were detected in the vegetative organs, such as root tips, young and mature leaf blades, stem, and internodes (Supplementary Figures 5A–D). We also detected at different developmental stages of the reproductive organs, such as spikelets, stamens, pistils, and seeds. GUS expression was visualized during the developmental stage of spikelet hulls except for the mature stage (Figure 5B). The developmental period of the spikelet can provide a reference for the pistil and stamen. Then, we estimated the developmental stage of pistil and stamen according to the length of the spikelet. Filaments of early stamens and the stamens of 4–5 mm and mature spikelets had the GUS signals (Supplementary Figure 5E). It was detected in the basal of pistils during the early stage (≤5 mm spikelet) and stigma of the mature stage (Supplementary Figure 5F). The GUS staining in seeds showed that it was highly expressed in the early stage [such as 5 h after fertilization (HAF), 1 DAF, 3 DAF, and 10 DAF] and mature seeds (Figure 5C), consistent with the results of the qRT-PCR. This expression profile indicated that OsDDM1b functioned in different young tissues, in line with the results presented above that OsDDM1b regulated cell division and proliferation rate.

The phenotype of osddm1b shown, such as erect and dark green leaves, dwarfism, and small grains, is typical BR response phenotypes in rice (Yamamuro et al., 2000; Hong et al., 2003). So, we hypothesize that OsDDM1b is involved in the BR response. We evaluated the sensitivity of osddm1b to 24-epibrassinolide (24-epiBL) using a lamina joint assay. As shown in the results, the second lamina of the wild type exhibited a 24-epiBL dosage-dependent inclination. By contrast, osddm1b did not show a blatant response (Figures 6A,B and Supplementary Figure 6A). The previous studies suggested that BR promoting coleoptile growth was an another indicator of a BR-responsive phenotype (Yamamuro et al., 2000; Li et al., 2002; Je et al., 2010). To analyze the response of coleoptiles to BR, we also performed a coleoptile elongation assay for BR sensitivity (Yamamuro et al., 2000). The wild-type and osddm1b seeds were germinated and grown under dark conditions in half MS medium containing different concentrations of 24-epiBL. The results showed that the relative elongation of coleoptile treatments was much weaker in osddm1b than in the wild type under BR treatment (Figure 6C), which suggests that osddm1b was less sensitive to BRs than the wild type.

According to the above results, loss-of-function of OsDDM1b decreased the response to BR treatments, which suggests that OsDDM1b may involve in BR signaling. Then, we analyzed the expression levels of BR-related genes, which include OsBAK1, OsBRI1, OsBU1, OsDLT, OsILI1, OsLIC, OsMDP1, OsBZR1, OsD2, and OsGSK2 in the wild type and osddm1b (Yamamuro et al., 2000; Hong et al., 2003; Ito et al., 2005; Duan et al., 2006; Bai et al., 2007; Koh et al., 2007; Wang et al., 2008; Tong et al., 2009, 2012; Zhang et al., 2009). In roots, the OsBAK1, OsBRI1, OsILI1, and OsBZR1 were slightly upregulated in osddm1b (Figure 6D). However, the expression level of OsBU1 and OsDLT, positive regulator of BR signaling, was significantly lower in osddm1b than in the wild type. The expression level of OsLIC, OsMDP1, and OsGSK2, as the negative regulator of BR, was also upregulated in osddm1b. OsD2, as a representative BR biosynthetic gene, was slightly upregulated in osddm1b. These results suggested that OsDDM1b influenced BR homeostasis and signaling.

It had been reported that the regulation of feedback-inhibited genes in BR biosynthesis usually defects mainly in BR signaling in mutants (Bai et al., 2007; Chen et al., 2013). To confirm this, we analyzed the expression of OsBRI1, OsCPD, OsD11, and OsDWARF (Yamamuro et al., 2000; Hong et al., 2002; Tanabe et al., 2005) genes in the wild type and osddm1b in the absence or presence of 24-epiBL. In osddm1b, BR’s inhibition of these feedback-inhibited genes was released, and they were all strongly suppressed in the mutant than wild type (Figure 6E). To find the essential answers of OsDDM1b in regulating BR biosynthesis, the amounts of endogenous BRs were measured in the mutants and wild type (Supplementary Figure 6B). The contents of endogenous BRs were increased by 36.01% in osddm1b compared with wild type, which indicated that OsDDM1b was required for the feedback inhibition of BR biosynthetic and signaling genes.

Grain size includes length, width, and thickness, as a quantitative trait locus (QTL), which is vital for grain yield and quality. Thus far, multiple QTLs have been cloned and studied, such as BG1, GL3.1, GL6, GL7, GLW7, GS2, GS5, GS9, GSN1, MKKK10, qLGY3, qTGW3, GW2, GW5, GW8, DEP1, RGB1, and XIAO, which contributed to rice yield (Song et al., 2007; Huang et al., 2009; Li et al., 2011; Jiang et al., 2012; Qi et al., 2012; Wang et al., 2012, 2015, 2019; Hu et al., 2015, 2018; Liu L. C. et al., 2015; Si et al., 2016; Liu et al., 2017, 2018; Guo et al., 2018; Xu et al., 2018; Ying et al., 2018; Zhao et al., 2018). Snf2 family genes, as the chromatin remodeling factors, regulate the transcription of the genes, and our data showed that loss-of-function of OsDDM1b resulted in abnormal grain size. To analyze the expression level of grain size-related genes in osddm1b, we detected several genes in inflorescence (length 8 cm) and seeds (1 DAF) in wild type and osddm1b using qRT-PCR. The results showed that GL3.1, GL7, GS2, GS5, GSN1, MKKK10, GW8, DEP1, RGB1, and XIAO were upregulated in osddm1b compared with wild type in inflorescence or seeds (Figure 7), and GL3.1, GL7, GS2, and DEP1 were significantly upregulated. It indicates that OsDDM1b negatively regulates their expression level, which affects the grain length (Figure 7A). The loss-of-function of GSN1, MKKK10, and XIAO caused small grain size, which was highly expressed in inflorescence and seeds in osddm1b, which suggests that OsDDM1b might negatively regulate the expression of these genes (Figures 7A,B). The expression of GS5 and GW2, which affects the milk-filling rate in rice, increased significantly in 1 DAF seeds of osddm1b. However, in-depth study is required for identifying that the loss-of-function of OsDDM1b influences milk-filling rate in rice.

Snf2, as a sizeable gene family, has conserved domains (SNF2_N, DEAD and Helicase_C, HELICs) as in animal counterparts, which suggests that they may have a similar function to regulate the gene expression. There are 42 proteins in chromDB database that belong to Snf2 family in rice. The proteins of OsCHR723 and OsCHR744 (homologs of MOM1/CHR15 in Arabidopsis) do not have the Helicase_C domain, which is not the member of Snf2 family (Hu et al., 2013). Many genes exhibit specific expression patterns at the different developmental stages of various reproductive tissues (inflorescence, stamens, pistils, and seeds). In addition, some genes have been reported to regulate the abiotic or biotic stress responses, such as OsALT1 (alkaline tolerance 1, OsCHR706) (Guo et al., 2014) and OsBRHIS1 (Li et al., 2015). Here, we report the comparative expression profile of 40 Snf2 genes during the reproductive development to enrich the expression patterns and insights into the functions of this family. However, the responses of the Snf2 gene family under abiotic or biotic stresses have not been systematically analyzed.

DDM1 is a nucleosome remodeling ATPase and regulates the DNA damage that is critical influence for the chromatin structure in Arabidopsis. The RNAi mutation of DDM1 has a typical phenotype of demethylation (Shaked et al., 2006). The repeated sequence, even a single-copy sequence, is induced toward hypomethylation after repeated self-pollination, which activates the transposons (Kakutani et al., 1996). The mutation of DDM1 showed reduced fertility only after repeated self-pollination and produced several kinds of developmental abnormalities. DDM1 is conserved in regulating DNA methylation. For example, Lsh (lymphocyte-specific helicase), a homolog gene of DDM1, causes genomic DNA hypomethylation in the mouse (Dennis et al., 2001). The antisense OsDDM1 lines exhibit dwarf phenotypes, and some sublines were observed to be progressive loss of fertility even complete infertility. In addition, methylation is severely reduced in antisense OsDDM1 lines, and reduction is evident in later generations but not in the genetic transformation process. It is still unknown whether the demethylation reduction was mediated by the repression of the OsDDM1a/OsCHR746, OsDDM1b/OsCHR741, or both of them. Our data showed that these two homologous genes, OsDDM1a/OsCHR746 and OsDDM1b/OsCHR741, had similar expression patterns and slight differences during seed development, which indicates that the two genes might have diverse functions.

The dwarfism phenotype was also observed in the T-DNA insertion mutation of OsDDM1b (Figures 2B,H,I and Supplementary Figures 2B,C). The male and female gametophyte formation was found to be normal as expected. However, the seed setting rate was partially decreased to 84.39%, which suggests that OsDDM1b might play a role in pollen tube growth or embryo development after fertilization (Supplementary Figure 2D–H). The dwarfism and partial fertility are similar to antisense OsDDM1 lines. The single mutation of OsDDM1a or OsDDM1b showed a normal phenotype during vegetative or reproductive stages, even in several generations (Tan et al., 2016). However, our data showed that osddm1b exhibited dwarfism and grain size variations. The T-DNA insertion in Tan’s mutation of OsDDM1b was in the third intron. However, in our experiment, the mutation was in the second exon (Figure 2A and Supplementary Figure 2A, see footnote 3), which suggests that this difference may be due to the position of T-DNA insertion. OsDDM1a and OsDDM1b, as the homologous genes, have a significantly similar genome and amino acid sequences. The primers of qRT-PCR were designed by aligning the CDS of two genes and finding the different bases among T-DNA insertion positions to improve primers’ specificity and obtain more accurate expression data (Figure 2A).

As an important agronomic trait, grain size is essential for crop genetics and breeding, which includes length, width, and thickness. Several genes have been reported to play the critical roles in regulating grain size. Cell proliferation and expansion are coordinately significant in the spikelet hull, which controls the final grain size. The expression levels of both OsDDM1 are more enriched in proliferating young tissues (Hu et al., 2013), which is consistent with the expression pattern of OsDDM1b by GUS staining (Figures 5B,C and Supplementary Figure 5). The results of the paraffin section and SEM suggested that OsDDM1b might promote latitudinal growth by increasing cell proliferation to contribute to the grain size (Figure 4 and Supplementary Figure 4). Many regulators are involved in controlling grain size, such as G protein signaling (DEP1, RGB1) (Huang et al., 2009; Zhang et al., 2021), the ubiquitin–proteasome pathway (GW2) (Song et al., 2007), the MAPK signaling pathway (GSN1, MKKK10, and MAPK6) (Liu S. Y. et al., 2015; Guo et al., 2018; Xu et al., 2018), phytohormone signaling (BG1, GL3.1, and GS5) (Li et al., 2011; Qi et al., 2012; Liu L. C. et al., 2015), transcriptional regulation (GL7, GLW7/OsSPL13, GS2/OsGRF4, GS9, GW8/OsSPL16, and qLGY3/OsMADS1) (Wang et al., 2012, 2015; Hu et al., 2015; Si et al., 2016; Liu et al., 2018; Zhao et al., 2018), and other regulators such as protein kinase (qTGW3/OsGSK5) (Hu et al., 2018). The creation of insertion mutations is a valuable tool for active transposons to analyze the function of endogenous genes (Hirochika, 2001). The abnormal grain size was caused by the defection of OsDDM1b, which affected the expression of grain size regulators.

There may be a balance mechanism in plants when the mutation of OsDDM1b displayed decreased grain length. In contrast, grain width and thickness were increased due to the compensation mechanism or regulated grain filling rate to balance the 1,000-grain weight with wild type. The higher expression levels of GS5 produced more comprehensive grains and increased grain filling rate. Then, GS5 was upregulated in the inflorescence (length 8 cm) and seeds (1 DAF) of osddm1b, which suggests that OsDDM1b might negatively regulate GS5 to affect the grain width. However, whether OsDDM1b can influence the grain filling rate is still unknown.

Above all, OsDDM1b (decrease in DNA methylation) contributes to grain size by regulating latitudinal cell proliferation and cell width to enhance the grain width and thickness, which decreases the grain length reversely. In addition, OsDDM1b, as a chromatin remodeling regulator, influences BR homeostasis and signaling pathway, the expression levels of BR, and grain size-related genes (Supplementary Figure 7). DNA methylation is involved in embryo and endosperm development and is required to determine the methylation pattern in osddm1b. The relationship between grain size and dynamic DNA methylation is necessary to be revealed with the future studies. This helps to understand the mechanism of epigenetic regulation in controlling grain size in rice.