Mutation in HER1 enhances stigma exsertion and hybrid seed production in rice

Abstract

Hybrid rice has achieved a significant leap in rice yield, but the low outcrossing rate of sterile lines limits hybrid seed production due to rice’s self-pollinating habits, constraining the further development and widespread adoption of hybrid rice. Stigma exsertion is a critical trait affecting outcrossing rate. Here we characterize a rice mutant, high exsertion rate1 (her1), characterized by enlarged stigma and enhanced stigma exsertion. HER1 encodes a SET domain-containing protein that binds to histone H3 at methylated lysine 9 and regulates H3K9me2 levels. Loss of function of HER1 results in reduced H3K9me2 levels in her1, increasing the transcription of Downstream2 (DS2), which positively controls stigma size. Using the loss-of-function allele of her1, we develop a high-exsertion male-sterile line herA, which increases hybrid seed production by more than 20% in field conditions, substantiating the breeding value of HER1 in hybrid seed production.

Low outcrossing rate of sterile lines limits hybrid rice seed production. Here, the authors report that mutation of a SET domain-containing protein HER1 leads to enlarged stigma, enhanced stigma exsertion, and hybrid seed production in rice.

Introduction

Rice (Oryza sativa L.) is a highly self-pollinating crop with flowering traits that favor self-pollination over cross-pollination¹. However, hybrid rice seeds are produced when the female parent (male-sterile line) receives pollen from the male parent (restorer line) and undergoes fertilization². Therefore, the outcrossing rate of the sterile line critically limits hybrid seed production. The outcrossing rate of rice is influenced by both environmental and genetic factors. The environmental effects can be optimized through cultivation techniques³. Genetic factors associated with plant morphology, flowering habits, and the structures of floral organs are all closely related to outcrossing. Stigma size and exsertion rate are key traits determining the outcrossing rate⁴.

Research on the rice OsPINOID (OsPID) gene has provided some insight into rice stigma formation. Initially discovered in Arabidopsis, PINOID (PID) encodes a serine-threonine protein kinase involved in polar auxin transport^5,6. In rice, the reduced function of OsPID results in the absence or diminutive size of stigmas, while its overexpression leads to excessive stigma proliferation, forming up to three or even four stigmas in a single floret, accompanied by a reduced number of stamens^7,8. Defective Stigma and Panicle (DSP) encodes an AP2/ERF-domain transcription factor that directly binds to the promoter of OsPID and regulates its transcription levels, influencing rice stigma differentiation⁹.

Numerous quantitative trait loci (QTL) controlling stigma exsertion have been identified, but only a few genes have been cloned^10–20. Grain Size3 (GS3) is a major gene that controls grain length in rice^21,22, and is also the first gene cloned to regulate stigma exsertion. The loss-of-function allele of GS3 not only increases grain length but also elongates the style, making the stigma more likely to extend outside the spikelet, ultimately leading to a significant increase in exsertion rate¹¹. Through a genome-wide association study (GWAS) on 533 accessions, two genes, GS3 and Grain Width5 (GW5), associated with stigma exsertion were identified²³. GW5 encodes a calmodulin-binding protein and acts as a negative regulator of grain width²⁴. The strong functional allele of GW5 can reduce spikelet width, making it easier for stigmas to extend outside the glumes. The haplotype combination GW5gs3 can significantly enhance stigma exsertion as well as the appearance quality of rice grains²³. Based on the correlation between grain shape and stigma exsertion, three genes regulating grain size, GS3, Grain Width8 (GW8), and Grain Shape9 (GS9), were chosen, and their single and multiple mutant lines were constructed²⁵. Loss of function of these three genes leads to narrower or longer grains^21,26,27 and single mutants of all three genes showed enhanced stigma exsertion, an effect that was more pronounced in the double and triple mutants. The outcrossing rate and grain length-to-width ratio of the triple mutant lines were significantly improved²⁵. However, the effects of these genes on enhancing hybrid seed production have not been tested under field conditions.

Epigenetics plays a crucial role in regulating plant growth and development, with histone modification being one of the most critical mechanisms in this regulatory system²⁸. Histone modifications primarily include methylation, acetylation, phosphorylation, and ubiquitination, which are closely associated with gene expression, DNA damage repair and recombination²⁹. The establishment of histone methylation primarily depends on SET domain-containing proteins, which catalyze the transfer of methyl groups from S-adenosylmethionine (SAM) to target lysine residues on histones³⁰. These proteins can be classified into several subfamilies based on their structural characteristics and functions, including SET1, SET2, SU(VAR)3-9, EZH, and SMYD, among others, which specifically regulate histone methylation modifications at different lysine residues³¹. H3K9 methylation, especially H3K9me2, is a highly prevalent repressive histone mark in plants and plays crucial roles in maintaining genome stability, regulating transposon activity and controlling gene expression^32,33. In Arabidopsis thaliana, the SU(VAR)3-9 homologs (SUVHs) and SU(VAR)3-9 related proteins (SUVRs) constitute two important subfamilies of SET domain proteins, which play crucial roles in the establishment of histone H3K9 methylation³⁴. SUVHs possess a unique SRA domain, which is associated with their function to specifically recognize DNA methylation and catalyze histone methylation. For example, SUVH4, also known as KRYPTONITE (KYP), binds to methylated DNA CHG/CHH sites to establish H3K9me2³⁵. Unlike SUVHs, SUVR5 has been demonstrated to catalyze H3K9me2 independent of DNA methylation, contributing to repressing gene expression³⁶. H3K9 methylation is closely associated with various important traits in plants, including stress responses and flowering time, among others^37–40. However, there are currently no research reports on its relationship with stigma development and stigma exsertion.

Here, we report a gene, High Exsertion Rate1 (HER1), that regulates stigma exsertion by directly controlling the stigma size, which acts differently from those genes that regulate grain shape. HER1 encodes a SET domain-containing protein that is homologous to SUVR5 in Arabidopsis thaliana. HER1 specifically binds to histone H3 at methylated lysine 9 and regulates H3K9me2 levels. Loss of function of HER1 leads to a reduction in H3K9me2 levels, increasing the transcription of the downstream target, Downstream2 (DS2), and expanding stigma size. Using the loss-of-function allele of HER1, we develop a high-exsertion male-sterile line herA, which can increase hybrid seed production by more than 20% in field conditions, substantiating the breeding value of HER1 in improving hybrid seed production.

Results

Mutant her1 showed enlarged stigma and increased stigma exsertion rate

We utilized ethyl methanesulfonate (EMS) mutagenesis on the indica three-line hybrid rice maintainer II-32B, which shows purple-black stigmas, and obtained a mutant with a high stigma exsertion rate, her1. Compared to the wild-type II-32B, the stigma exsertion rate of her1 significantly increased from 36.32% to 70.63%, an increase of 94.47% (Fig. 1a, c). Microscopic measurements of pistils revealed that the overall size of the pistil in her1 was larger than that of the wild-type (Fig. 1b). Specifically, the stigma length (brush-like part) increased from 0.96 mm to 1.25 mm, the stigma width (brush-like part) increased from 0.35 mm to 0.45 mm, and the style length (non-brush-like part) increased from 0.78 mm to 0.87 mm (Fig. 1d–f). Stigma exsertion is a key factor in determining the outcrossing rate and hybrid seed production. In her1, the stigma exsertion rate was significantly increased by nearly 100% compared to the wild-type, suggesting that the gene controlling this mutant trait could have high breeding value.

Fig. 1. Mutant her1 exhibited enlarged stigma and increased exsertion rate.

a Comparison of stigma exsertion between panicles of II-32B and her1 at the flowering stage. Dashed circles indicate zooming in for a better view of stigma exsertion. Scale bars, 1 cm. b Comparison of stigma morphology between II-32B and her1. Scale bar, 500 μm. c–h Comparisons of exsertion rate (c), stigma length (d), stigma width (e), style length (f), grain length (g), and grain width (h) between II-32B and her1. i photograph of seeds from II-32B and her1. Scale bar 1 cm. j, k SEM images of the stigmas (brush-like parts) of II-32B and her1, with scale bars of 500 μm for image (j) and 100 μm for image (k). l, m SEM images of the styles of II-32B and her1, with scale bars of 100 μm for image (l) and 50 μm for image (m). The red boxes represent the boundary of the cells. n, o Comparisons of style cell length (n), and style cell width (o) between II-32B and her1. The data are presented as means ± s.d. P-values above the bars were calculated using two-tailed Student’s t tests. Source data are provided as a Source Data file.

To evaluate the impact of gene mutation on the growth and development of her1 plants, we conducted agronomic trait assessments on both wild-type II-32B and her1. Compared to the wild-type, the grain size of her1 was slightly smaller, with grain length reduced by 0.16 mm (1.88%) and grain width reduced by 0.07 mm (2.02%) (Fig. 1g–i). But the ratio of grain length to width remained unchanged (Supplementary Fig. 1a). The seed setting rate of her1 decreased from 78.48% to 71.29% (Supplementary Fig. 1b). However, there were no significant differences in panicle number per plant (tiller number), panicle length, and grain number per panicle between her1 and the wild-type (Supplementary Fig. 1c–e). The reduction in grain size resulted in a 1.35 g (5.33%) decrease in the thousand-grain weight of her1 (Supplementary Fig. 1f). Potassium iodide staining was used to investigate whether the change in the seed setting rate of her1 was related to its pollen fertility, but there was no significant difference between II-32B and her1 (Supplementary Fig. 2).

To further explore the stigma trait variations in her1, we observed the stigmas of both wild-type and her1 using scanning electron microscopy (SEM). The density and thickness of stigma brushes in her1 were increased compared to the wild-type, and the stigma of her1 appeared overall more robust and substantial (Fig. 1j, k). In the style, the epidermal cells of her1 showed notable protrusions, appearing plumper compared to the wild-type (Fig. 1l, m). The length of the epidermal cells in her1 styles decreased while their width increased, making them shorter and wider compared to the wild-type (Fig. 1n, o). These traits indicate that the increased number and thickened size of the stigma brushes in her1 are the primary factors contributing to the increase in stigma volume. Moreover, it can be inferred that the increased total length of the her1 stigma is mainly caused by an increase in the number of longitudinal cells in the style rather than changes in cell morphology.

HER1 encodes a SET domain-containing protein

To determine the inheritance pattern of the mutant traits in her1, we conducted reciprocal crosses between her1 and the wild-type II-32B. The BC₁F₁ plants exhibited the wild-type phenotype regardless of the direction of the cross, and the segregation patterns of two BC₁F₂ populations both followed a 3:1 ratio (Supplementary Table 1), suggesting that the mutant traits of her1 is governed by a single recessive nuclear gene.

We screened 50 extreme mutant-phenotype individuals from the BC₁F₂ population for whole-genome resequencing and MutMap analysis. The SNP-index map showed a significant linkage peak on chromosome 2, containing four SNPs with an SNP-index of 1.0 (Fig. 2a). SNP1 and SNP2 were located in intergenic and promoter regions, respectively. SNP3 and SNP4 fell within coding regions, while SNP4 results in a synonymous mutation, which does not have a functional effect (Supplementary Table 2). SNP3 was located in the coding region of the gene LOC_Os02g47900, named HER1, which was considered the candidate gene controlling the mutant traits. HER1 is comprised of seven exons, and in her1, the 97th base in exon 4 is mutated from A to T, causing a change in the encoded amino acid from arginine (AGA) to a stop codon (TGA) (Fig. 2b). HER1 encodes a SET domain-containing protein, also known as SET Domain Group706 (SDG706). The mutation in her1 leads to the truncation of the SET domain and complete loss of the post-SET domain (Fig. 2c).

Fig. 2. HER1 encodes a SET-domain protein.

a MutMap analysis of the mutant bulk from the BC₁F₂ population with II-32B as control. b The structure of the HER1 gene. Gray boxes represent the 5´ and 3´ UTRs, while black boxes denote the exons. A mutation from A to T appeared in her1 compared to II-32B. c The protein structure of HER1. A complete SET domain is present in the C-terminus of HER1, consisting of Pre-SET, SET, and Post-SET. The mutation in her1 caused a change in the encoded amino acid from arginine to a stop codon, leading to a truncation in the SET domain. d, e Phenotypes of II-32B, her1, the knockout lines her1-ko1/2, and the complementation lines HER1-CP1/2, with scale bars of 500 μm for image (d), and 1 cm for image (e). f, g Comparisons of stigma length and exsertion rate among II-32B, her1, her1-ko1/2, and HER1-CP1/2. The data are presented as means ± s.d. P-values were calculated using one-way ANOVA followed by Dunnett’s T3 multiple comparisons test. P-values for each line compared to II-32B are shown above the bars. Source data are provided as a Source Data file.

HER1 was constitutively expressed in rice plants, with expression observed in the roots, leaves, panicles, glumes, pistils, and stamens. Its expression level in the pistil was higher than that in other tissues (Supplementary Fig. 3). To validate the biological function of HER1 in regulating stigma size and exsertion, we constructed a CRISPR/Cas9 knockout vector and a functional complementation vector for HER1, which were introduced into the II-32B and her1 backgrounds, respectively (Supplementary Fig. 4a, b). The phenotypes of the two knockout lines were similar to those of her1, exhibiting significantly increased stigma size and exsertion rate (Fig. 2d–g). Conversely, the phenotypes of the two complementation lines in the her1 background were restored to the wild-type level, with no significant differences in either stigma length or exsertion rate (Fig. 2d–g). These results demonstrate that the traits of her1 are caused by the mutation of HER1, which has a biological function in regulating stigma size and exsertion.

We investigated the biological effects of HER1 in the japonica rice variety Zhonghua11 (ZH11). The two knockout lines her1-zh-ko1/2 in the ZH11 background showed larger stigmas compared to the wild-type (Supplementary Fig. 5a–c), and the stigma exsertion rate significantly increased from 2.48% to 8.66% and 9.75%, respectively, which is more than a threefold increase compared to the wild-type (Supplementary Fig. 5d). In addition, a ubiquitin promoter-driven overexpression vector of HER1 was constructed and introduced into II-32B (Supplementary Fig. 4c, d). Overexpression of HER1 did not alter the stigma size or exsertion rate in transgenic lines (Supplementary Fig. 5e–i). Phenotypic variation only occurs when there is a deficiency of HER1.

HER1 specifically binds to methylated histone H3K9 and regulates H3K9me2 levels

SET domain-containing proteins are universally localized in the nucleus and exhibit histone lysine methyltransferase (HKMT) activity^37–40. In the covalent modification of histones, HKMTs typically act as writers, establishing methylation marks on various lysine residues of histones, particularly H3 and H4, thereby regulating chromatin conformation and related processes to influence gene expression⁴¹. Using BLASTP, proteins with sequence similarity to HER1 were screened in rice and Arabidopsis. Phylogenetic analysis revealed that HER1 shares sequence similarity with the H3K9 methyltransferase protein families SUVHs and SUVRs in Arabidopsis, and is highly homologous to the SUVR5, suggesting its potential function in the regulation of H3K9 methylation (Supplementary Fig. 6).

To study the molecular function of HER1 protein, we first transiently expressed a 35S promoter-driven HER1-eGFP fusion protein in onion epidermal cells, and localized HER1 to the nucleus, consistent with a previous report⁴² (Supplementary Fig. 7a). Immunoblotting analyses were performed using specific antibodies targeting different methylation states of various lysine sites on histone H3 in the stigmas of wild-type and her1. The H3K9me2 level in the stigmas of her1 was reduced by more than 1-fold compared to the wild-type, while the levels of H3K9me1, H3K9me3, and other lysine sites showed minor changes (Fig. 3a). Both knockout and overexpression of HER1 affected the H3K9me2 levels in vivo (Fig. 3b). These results indicate that HER1 protein regulates H3K9me2 levels in plants and may influence the expression of downstream genes through this epigenetic mechanism, thereby modulating stigma exsertion rate in rice.

Fig. 3. HER1 protein specifically binds to methylated histone H3K9 and regulates H3K9me2 levels.

a Immunoblotting analyses on the levels of H3K4me2/3, H3K9me1/2/3, and H3K36me1/2/3 in II-32B and her1 stigmas, with H3 as the internal control. Fold changes are shown above the bands. Log₂FCs, which represent the log₂-transformed fold changes, are displayed on the right. b Immunoblotting analyses on H3K9me2 levels in stigmas of II-32B, her1-ko1, and HER1-OE1, with H3 as the internal control. Numbers in the upper panel indicate the fold changes compared to II-32B. c In vitro methyltransferase assay of GST-HER1 protein using recombinant mononucleosomes as the substrate, with G9a as the positive control, and GST as the negative control. d GST pull-down assay of GST-HER1 protein using calf thymus histones or recombinant mononucleosomes as substrates, with GST as the negative control. The image of Coomassie Brilliant Blue (CBB) staining is shown in the upper panel, while the immunoblotting analysis is displayed in the lower panel, using the α-H3 antibody. Red arrows mark the pulled-down calf thymus histones. e, f Peptide pull-down assays of various biotinylated H3 peptides with different methylation patterns, using both His-HER1 (e) and His-HER1^1-624 (f) proteins as substrates. Control groups were performed with empty streptavidin beads (no peptides). H3 denotes histone H3 peptides without methylation. g, h Surface plasmon resonance (SPR) analysis of HER1 binding to unmodified H3 (g) and H3K9me2 (h) peptides. HER1 was immobilized on the sensor chip, and increasing concentrations (0.3125–10 μM) of peptides were injected. Response units (RU) are shown as a function of time. Source data are provided as a Source Data file.

Using recombinant HER1 in methyltransferase assays, neither western blotting nor mass spectrometry (MS) revealed increased H3K9 methylation compared with controls, and no enzymatic activity was detected when HER1 was further tested using a commercial H3K9 methyltransferase activity kit (ab113453, Abcam) (Fig. 3c and Supplementary Fig. 7b–f). All these results are consistent with a previous report⁴². Therefore, we suggest that HER1 alone may not possess typical HKMT activity but instead interacts with histones through other mechanisms, thereby affecting H3K9me2 levels.

To verify this hypothesis, we performed a GST pull-down assay with GST-HER1 fusion protein, using calf thymus histones or recombinant mononucleosomes as substrates. Calf thymus histones are naturally purified and rich in various covalent modifications, whereas the histones in recombinant mononucleosomes are unmodified. HER1 specifically bound to calf thymus histones but not to recombinant mononucleosomes, suggesting that HER1 may have a preference for specific covalent modifications on histones (Fig. 3d).

We synthesized various biotinylated peptides of histone H3 with different methylation patterns (including H3K4, H3K9, H3K27, H3K36, and unmethylated H3) and performed a peptide pull-down assay using the His-tagged HER1 protein as the substrate. HER1 specifically bound to H3K9me1, H3K9me2, and H3K9me3 peptides but did not bind to H3K4me3, H3K27me3, or H3K36me3 peptides, nor did it bind to unmodified H3 peptides (Fig. 3e). HER1^1-624 protein, expressed using the coding sequence in her1 (Fig. 2c), could not bind to any form of histone H3 peptide (Fig. 3f). To further validate these results, we performed surface plasmon resonance (SPR) analysis using immobilized GST-HER1 protein. Consistent with the peptide pull-down assay, SPR analysis showed that HER1 selectively bound to H3K9me1, H3K9me2, and H3K9me3 peptides, but not to unmodified H3 peptides or peptides methylated at other lysine residues (Fig. 3g, h and Supplementary Fig. 8a–e). The SPR measurements revealed differences in binding affinities among the H3K9-methylated peptides, with H3K9me2 exhibiting the lowest K_D value, suggesting a relatively higher binding affinity compared with the other H3K9-methylated peptides (Supplementary Table 3). In addition, we divided HER1 into three parts, HER1-P1, HER1-P2, and HER1-P3 (P3 contains pre-SET, SET, and post-SET domains), and performed peptide pull-down assays with H3, H3K4me3, H3K9me1/2/3, H3K27me3, and H3K36me3 peptides to test the binding region of the HER1 protein (Supplementary Fig. 8f). Among the three fragments, only HER1-P3, which encompasses the entire SET domain architecture, demonstrated binding affinity to H3K9me peptides, but not to peptides with other modification types (Supplementary Fig. 8g). This finding was further validated using His-tagged constructs (Supplementary Fig. 8h). Based on the observed changes in H3K9me2 levels in her1, as well as in knockout and overexpression lines, it can be inferred that although HER1 may lack HKMT activity, it can specifically bind to H3K9me through P3 (containing SET domain) and maintain its methylation status.

HER1 regulates stigma morphology by modulating the transcription of downstream DS2

Based on the effect of HER1 on histone methylation in vivo (Fig. 3a, b), a Cut&Tag assay was performed using the α-H3K9me2 antibody on stigmas of the wild-type and her1. Global H3K9me2 levels exhibited a notable decline in her1 compared to the wild-type (Fig. 4a), aligning with immunoblotting results (Fig. 3a). Differential analysis on the H3K9me2 peaks identified 550 down-regulated peaks and 255 up-regulated peaks in her1 compared to the wild-type, co-locating with 487 and 236 genes, respectively (Fig. 4b, c). A specific polyclonal antibody, α-HER1, was prepared in rabbits for the identification of downstream targets of the HER1 protein. The Cut&Tag analysis with α-HER1 revealed predominant enrichment of HER1 in the promoter, intergenic, and gene body regions (encompassing CDS and introns), with up to 4559 peaks enriched on the promoters, implying an impact of HER1 on gene transcription (Fig. 4d). Moreover, HER1-enriched peaks exhibited a significant overlap with H3K9me2-enriched peaks in the wild-type, indicating a close relationship between HER1 and di-methylation of H3K9 (Fig. 4e).

Fig. 4. Cut&Tag and RNA sequencing reveled the downstream target of HER1.

a Comparison of reads density across H3K9me2 peaks between II-32B and her1. b Volcano plot of down-regulated and up-regulated H3K9me2 peaks in her1 compared to II-32B, with q values ≤ 0.05. c Counts of genes co-located with down-regulated and up-regulated H3K9me2 peaks. d Distribution of HER1 peaks across various genomic structures. The bars indicate the number of peaks uniquely situated or overlapping on 3´ UTR, 5´ UTR, intron, intergenic, exon, and promoter regions. The numbers in the bottom-left corner represent the total sum of peaks distributed on the structures. e Venn diagram showing the overlapping peaks between H3K9me2 and HER1 in II-32B. f Volcano plot of the DEGs in her1 compared to II-32B from RNA-seq, with q values ≤ 0.05 and fold changes ≥ 2 or ≤ 0.5. g Venn diagram illustrating the overlap among HER1-enriched genes, up-regulated DEGs, and genes expressed in stigma cell cluster reported previously⁴⁶. Five genes in the intersection of the three datasets were identified as candidate downstream targets of HER1. h Genome browser tracks showing H3K9me2 enrichment on DS2 in both II-32B and her1, along with HER1 enrichment in II-32B. Scale bar 500 bp. i Comparisons of DS2 expression in leaf, glume, and stigma between II-32B and her1 determined by RT-qPCR. OsActin1 was used as the internal control. The data are presented as means ± s.d. P-values above the bars were calculated using two-way ANOVA. Source data are provided as a Source Data file.

RNA-seq analyses on stigmas of II-32B and her1 detected 1542 differentially expressed genes (DEGs), with 320 genes down-regulated and 1222 genes up-regulated in her1 compared to the II-32B (Fig. 4f). Considering the repressive effect of H3K9me2 on gene expression^34,43–45, we screened genes that were both enriched by HER1 and up-regulated in expression level (in her1), and performed an overlap analysis with genes previously identified in the stigma cell cluster⁴⁶, identifying five genes as possible downstream targets of HER1 (Fig. 4g and Supplementary Table 4). Principal component analysis (PCA) confirmed the consistency between biological replicates for both Cut&Tag-seq and RNA-seq (Supplementary Fig. 9).

Among these five genes, DS2 (LOC_Os03g02460), which is specifically expressed in stigmas according to the ePlant database⁴⁷ (https://bar.utoronto.ca/eplant_rice/; Supplementary Fig. 10a), showed significant enrichment of HER1 in both its promoter region and gene body, along with a notable decrease in H3K9me2 levels (Fig. 4h). Chromatin immunoprecipitation quantitative PCR (ChIP-qPCR) analysis using the α-HER1 antibody further validated the enrichment of HER1 on DS2 gene, and in the her1 mutant, the enrichment of HER1 on DS2 was significantly reduced due to the loss of key binding site of the HER1 protein (Supplementary Fig. 10b, c). Whole-genome bisulfite sequencing (WGBS) was also performed in both II-32B and her1. Unlike the HER1 binding and H3K9me2 enrichment observed at its gene region (Fig. 4h), DS2 exhibited very low DNA methylation in both II-32B and her1 (Supplementary Fig. 10d). Importantly, no appreciable differences in DNA methylation patterns were observed at the DS2 locus between II-32B and her1, despite clear differences in H3K9me2 levels and DS2 expression. These results suggest that HER1 regulates H3K9me2 and DS2 expression in a DNA methylation-independent manner, akin to what has been reported for SUVR5 in Arabidopsis³⁶.

DS2 encodes a short-chain dehydrogenase/reductase (SDR) with an unknown function. The RT-qPCR assay demonstrated high DS2 expression in the stigma, with a significantly increased expression level in the stigma of her1 compared to that of the wild-type (Fig. 4i). Similar to the her1 mutant, two independent DS2 overexpression lines showed significantly enlarged stigmas, with stigma length increased from 1.41 mm to 1.65 mm and 1.68 mm, respectively (Fig. 5a–c). Stigma exsertion rate of both DS2-OE1 and DS-OE2 also significantly increased compared to the wild-type (Fig. 5d). We further knocked out DS2 in both II-32B and her1 to create ds2 single mutant and her1-ds2 double mutant. The ds2 single mutant showed a significant decrease in both stigma length and stigma width compared to II-32B, while loss of function of DS2 in her1 significantly limited the stigma size (Fig. 5e–g). These results further indicated that HER1 regulates the expression of DS2 by modulating H3K9me2 levels, ultimately controlling the stigma size and exsertion (Fig. 5h).

Fig. 5. Genetic relationship between HER1 and DS2.

a Photograph of pistils from II-32B, DS2-OE1, and DS2-OE2, scale bar 500 μm. b Relative expression levels of DS2 in II-32B, DS2-OE1, and DS2-OE2. OsActin1 was used as the internal control. c, d Comparisons of stigma length (c) and stigma exsertion rate (d) among II-32B, DS2-OE1, and DS2-OE2. The data are presented as means ± s.d. P-values were calculated using one-way ANOVA. e Photograph of pistils from II-32B, her1, her1-ds2, and ds2. f, g Comparisons of stigma length (f) and stigma width (g) among II-32B, her1, her1-ds2, and ds2. The data are presented as means ± s.d. P-values were calculated using one-way ANOVA followed by Dunnett’s T3 multiple comparisons test. h A model depicting a HER1-H3K9me2-DS2 module regulating stigma morphology and exsertion in rice. In the wild-type, the normally functioning HER1 protein binds to histone H3K9me, preventing its methylation level from being reduced by other factors, such as possible demethylases. As a result, the histone H3K9me2 remains at a normal level, and the transcription of downstream DS2 is repressed, leading to a normal stigma size and exsertion rate. In the her1 mutant, the functional disruption prevents HER1 from binding to histone H3K9me, resulting in a reduction in H3K9me2 levels and an increase in DS2 expression. This ultimately leads to the phenotype of increased stigma size and exsertion rate. Source data are provided as a Source Data file.

The her1 allele enhances hybrid seed production

The increased exsertion rate observed in the her1 mutant suggests that HER1 may have significant breeding value. We searched for natural mutations of HER1 in 533 rice accessions²³ and found 21 variations in HER1 and its promoter region, including 18 SNPs and 3 InDels, with only six mutations located in the coding region (Supplementary Fig. 11a, b). All variations were categorized into six major haplotypes, named Hap1-Hap6, with 137, 108, 75, 72, 69, and 35 accessions included, respectively (Supplementary Fig. 11a). Only the accessions with Hap2, Hap3, and Hap4 showed an increase in total stigma length compared to those carrying Hap1, while no significant differences were observed for other haplotypes (Supplementary Fig. 11c). Hap2, Hap3, and Hap4 were almost exclusively present in the indica subpopulations (IndI and IndII), whereas Hap1 and Hap6 were predominantly found in the temperate japonica (TeJ) and tropical japonica (TrJ), respectively; Hap5 was more prevalent in TrJ and aus subpopulations, showing minimal presence in indica rice (Supplementary Fig. 11d). To avoid the impacts of genetic background, we compared the phenotypes of accessions carrying Hap2, Hap3, and Hap4 in indica groups, while those with Hap1, Hap5, and Hap6 were analyzed in japonica and aus groups. The results showed no significant differences in total stigma length among Hap2, Hap3, and Hap4, nor were there any differences observed among Hap1, Hap5, and Hap6 (Supplementary Fig. 11e, f).

To further analyze the effects of different HER1 haplotypes, we selected three accessions from each of Hap1-Hap4, which had shown statistically significant phenotypic differences, and performed qPCR analysis of the DS2 gene in their stigmas. Among the analyzed accessions, only one accession, W108 (Hap2), showed decreased expression compared to the three specific accessions, C069 (Hap2), W259 (Hap3), and W272 (Hap3), while no significant differences were observed among other accessions (Supplementary Fig. 11g). This differential expression did not show a clear association with the phenotype. Furthermore, we found that the variation explained by haplotypes was less than the variation explained by subpopulations (Supplementary Table 5). Given this lack of correlation between HER1 haplotypes, DS2 expression, and phenotypes, the increased total stigma length observed in accessions with Hap2, Hap3, and Hap4 may be caused by differences in subpopulation genetic architecture, rather than natural variations in HER1. In addition, nucleotide diversity (π) of HER1 and its flanking regions (50 kb upstream and 50 kb downstream) was analyzed in both cultivated rice (Oryza sativa) and wild rice (Oryza rufipogon). The nucleotide diversity value for HER1 is much lower than its flanking regions in both cultivated and wild populations (Oryza rufipogon), suggesting that the minimal natural variation in the locus of HER1 might be the result of natural selection rather than artificial selection (Supplementary Fig. 11h).

Therefore, we introduced the loss-of-function allele from her1 into the sterile line II-32A (matched with II-32B) through backcross breeding, naming the new line herA. Compared to II-32A, herA exhibited enlarged stigma and a significant increase in stigma exsertion rate, rising from 47.48% to 75.42% (Fig. 6a–c). In 2022, we conducted a preliminary outcrossing experiment on herA and II-32A, using II-32B as the pollen donor. Compared to II-32A, the outcrossing rate of herA increased by 23.03%, and the seed production number per plant increased by 19.16% (Supplementary Fig. 12).

Fig. 6. The loss-of-function allele of HER1 enhanced the outcrossing habit and hybrid seed production of the sterile line II-32A.

a Comparison of stigma size and exsertion stage between II-32A and herA. Scale bar 1 mm. b, c Comparison of stigma exsertion rate between II-32A and herA. d Aerial photograph of the seed production experimental field, which measures 65.87 m in length and 6.50 m in width, covering an area of 428.15 m². e, f Schematic diagram of the planting pattern. B denotes the maintainer rows, while A denotes the sterile rows. g A photo depicting the seeds produced by a single plant of II-32A and herA, with II-32B as the male parent. h–j Comparisons of outcrossing rate (h), seed production yield per field plot (10 plants) (i), and seed production per 666.7 m² (j) between II-32A and herA, with II-32B as the male parent. The data are presented as means ± s.d. P-values above the bars were calculated using two-tailed Student’s t tests. Source data are provided as a Source Data file.

We massively propagated the seeds of herA and II-32A and conducted a field seed production trial in 2023. The experimental field was approximately 65.87 m in length and 6.50 m in width, covering an area of about 428.15 m² (Fig. 6d). To ensure uniform pollen distribution, the trial utilized an interplanting method between the sterile (A) and maintainer (B) lines. Each B row was alternated with two A rows, with each A row containing half a row of herA and half a row of II-32A (Fig. 6e, f). The sterile line herA showed a notable increase in seed production per plant compared to II-32A (Fig. 6g). Statistical analyses revealed that the outcrossing rate of herA increased from 40.80% to 50.37% compared to II-32A (an increase of 23.46%), the seed production yield per field plot (10 plants) increased from 124.57 g to 150.46 g (an increase of 20.79%), and the seed production yield per 666.7 m² increased significantly from 175.76 kg to 215.08 kg, showing a substantial increase of 22.37% (Fig. 6h–j).

As a recessive allele, the negative effects of her1 may be masked by the dominant allele from the restorer line in the hybrid rice (F₁) generation⁴⁸. We generated six hybrid rice combinations by crossing the male sterile lines II-32A and herA with three restorer lines, Shuhui12, Shuhui527, and Fuhui838 (Supplementary Fig. 13a). Agronomic trait analysis revealed that the hybrid rice combinations developed using herA and those using II-32A showed no significant differences (P > 0.05) in plant height, panicle number per plant, panicle length, grain number per panicle, seed setting rate, thousand-grain weight, and yield per plant (Supplementary Fig. 13b, h). Final yield of every field plot was tested, and none of them revealed significant differences between the hybrid rice combinations developed using herA and those using II-32A (Supplementary Fig. 13i). These results confirmed the significant practical value of HER1 and provided precious material resources for high-outcrossing sterile line breeding, contributing to hybrid seed production.

Discussion

We have isolated a gene, HER1, that controls stigma morphology and exsertion in rice. The loss-of-function allele, her1, nearly doubled the stigma exsertion rate from 36.32% to 70.63%, increasing the outcrossing rate and hybrid seed production by more than 20%.

Current research on the regulatory mechanisms of rice stigma exsertion is limited and predominantly related to grain shape. A number of genes affecting stigma exsertion have been described. The loss-of-function allele of GS3 not only increases grain length but also affects the morphology of floral organs, particularly increasing the length of the pistil, enhancing stigma exsertion^11,23. The strong functional allele of GW5 can significantly reduce spikelet width, making it easier for stigmas to extend outside the glumes²³. The regulatory mechanisms of GW8 and GS9 are similar to those of GS3 and GW5, affecting stigma exsertion by modulating the length, width or length-width ratio of spikelets²⁵. However, the effect of these genes on enhancing hybrid seed production has not been confirmed by field trials.

Unlike these genes, HER1 controls stigma exsertion by directly regulating the length of the stigma and style, independent of changes in the size or length-width ratio of spikelets. HER1 encodes a SET domain-containing protein that is homologous to SUVR5 in Arabidopsis thaliana. SUVR5 was reported to function as an H3K9 methyltransferase, modifying H3K9me2 levels and controlling a subset of genes involved in environmental stress responses and pathogen defense³⁶. In this study, we found that HER1 specifically recognized and bound to histone H3K9me but did not catalyze the methylation of H3K9. However, the levels of H3K9me2 were decreased in the her1 mutant, suggesting that HER1 may lack the typical HKMT activity, but instead interacts with H3K9me to maintain its methylation levels. The maintenance of methylation levels relies on both histone methylation and demethylation factors. Methylation is executed by HKMTs, while demethylation is carried out by histone demethylases such as IBM1 and JMJ27 in Arabidopsis^49–51, and JMJ706 in rice⁵². Subsequent research should link HER1 to the histone demethylases, analyze their synergistic effects on H3K9me, and further investigate the molecular mechanisms by which HER1 regulates H3K9me2 levels.

HER1 targets the downstream gene, DS2, and regulates its transcription level by modulating di-methylation of H3K9, ultimately influencing stigma exsertion. A single-cell transcriptome analysis on rice pistils found that DS2 is highly expressed in the stigma cell cluster (cluster 5), suggesting that DS2 may play a specific role in affecting the development of rice stigma⁴⁶. DS2 encodes an SDR family protein with unknown function. Many plant SDRs participate in developmental regulation by catalyzing redox steps in hormone or secondary-metabolite pathways^53–57. We hypothesize that it functions in one of these metabolic routes to modulate stigma development. Specifically, DS2 may catalyze redox reactions that influence the balance between cell division and cell expansion in stigma tissues. Its pronounced accumulation in stigma further suggests DS2 could serve as a key regulatory node in stigma-specific metabolic networks that determine the final morphology of the stigma. Future research should explore the molecular function of the DS2 protein to enhance our understanding regarding the regulation of stigma morphology and exsertion.

Based on our existing results, we have outlined the epigenetic regulatory mechanisms by which HER1 controls stigma exsertion in rice. In the wild-type, the normally functioning HER1 protein can closely bind to histone H3K9me, preventing its methylation level from being decreased by other factors, such as potential demethylases. As a result, the histone H3K9me2 levels remain normal, and the transcription of downstream DS2 is repressed, leading to a normal stigma size and exsertion rate. In the her1 mutant, the functional disruption prevents HER1 from binding to histone H3K9me, resulting in a reduction in H3K9me2 levels and up-regulation of DS2 expression. This ultimately leads to the phenotype of enlarged stigma and increased exsertion rate (Fig. 5h).

Certain histone modification reader proteins can bring SET domain-containing proteins together to create complexes, which in turn enhance methyltransferase activity or accelerate the spread of histone modifications across chromatin. The Arabidopsis Polycomb Repressive Complex 1 (PRC1) can specifically recognize H3K27me3 and contribute to the establishment and propagation of H3K27me3 at specific loci through the PRC1-PRC2 interaction mode⁵⁸. In Drosophila, the histone H3K36 methyltransferase Absent, small, or homeotic 1 (Ash1) can form a protein complex with the histone modification reader protein MORF4-related gene on chromosome 15 (Mrg15), which has been shown to significantly enhance the methyltransferase activity of Ash1⁵⁹. Although we did not detect HKMT activity of HER1, its presence cannot be ruled out. HER1-interacting proteins and potential complexes formed by HER1 should be further identified in investigating the molecular function of HER1.

As the staple food for more than half of the world’s population, rice production is critically important⁶⁰. Consequently, whether it is the study of genes related to rice yield, stress resistance, or outcrossing habit, the fundamental purpose is to serve practical rice production. During the exploration of gene functions, researchers often focus more on the beneficial phenotypes produced by genes, while sometimes neglecting the accompanying adverse effects. In breeding practice, these negative impacts can be difficult to eliminate, ultimately leading to only a few genes being effectively utilized⁶¹. Hybrid rice performs better than conventional rice, but its adoption by farmers, especially in less developed countries, is hindered by the high cost of hybrid seeds. Based on the results of the field trial, the loss-of-function allele of HER1 significantly increased hybrid seed production by 22.37%, with only minor negative effects on other agronomic traits. As a recessive allele, all the negative effects of her1 were masked by the dominant allele from the restorer line in the hybrid rice generation (F₁). Building on the successful experience in II-32A, we knocked out HER1 in the indica three-line hybrid rice maintainers Shu6B and 148B, as well as the two-line sterile Ju182S (Supplementary Fig. 14). These improved lines will be directly used in hybrid rice production. We hope that HER1 can become one of the few effectively utilized genes and be widely applied in China and other major hybrid-rice-producing countries around the world.

Methods

Plant materials and growth conditions

The rice varieties used in this study mainly include the indica three-line maintainer II-32B and the japonica variety ZH11. II-32B shows purple-black stigmas, facilitating the investigation and screening of stigma-related traits. The her1 mutant was isolated from an EMS-induced mutant library of II-32B. The breeding application utilized several varieties: the sterile line II-32A (corresponding to II-32B), the indica three-line maintainer Shu6B (bred by our team), the indica three-line maintainer 148B (bred by Hunan Taohuayuan Agricultural Technology Co., Ltd.), and the indica two-line sterile Ju182S (bred by Jiangxi Xiannong Seed Industry Co., Ltd.). Shu6B, 148B, and Ju182S all exhibit high combining ability, good quality, and strong resistance. However, their poor outcrossing habits result in low seed production, indicating significant potential for improvement. All of the materials were grown in the experimental bases of Sichuan Agricultural University in Chengdu, Sichuan (summer), and Lingshui, Hainan (winter).

Evaluation of stigma exsertion rate and other traits

The stigma exsertion can be classified into single exsertion and dual exsertion, based on the number of exserted stigmas. The single stigma exsertion rate (SSER) refers to the percentage of spikelets with single exserted stigma relative to the total number of spikelets. Similarly, the dual stigma exsertion rate (DSER) denotes the percentage of spikelets with dual exserted stigmas. The stigma exsertion rate was universally calculated as the total stigma exsertion rate (TSER) in this study, i.e., stigma exsertion rate (%) = TSER (%) = SSER (%) + DSER (%).

The stigmas used for size measurement were obtained from spikelets on the day of flowering. For each line, at least 10 samples were collected and immediately placed under an optical microscope for observation and measurement. Stigma length and width were measured as the length and width of the brush-like part. Style length was measured as the length of the non-brush-like part.

$$\mathrm{Total}\mathrm{stigma}\mathrm{length}=\mathrm{stigma}\mathrm{length}+\mathrm{style}\mathrm{length}$$ 1

For SEM analysis⁶², the stigmas were also obtained from spikelets on the day of flowering, but were immediately placed into a 2.5% glutaraldehyde fixative. After 2 h of fixation, the samples were dehydrated in a gradient series of increasing ethanol concentrations (30, 50, 70, 80, 90, 95, and 100%, each for 15 minutes), and then dried with a critical point dryer (K850, Quorum Technologies). Finally, the samples were coated with gold–palladium using a polaron sputter coater (MC1000, Hitachi) and examined using a scanning electron microscope (Gemini SEM 300, Zeiss).

The samples for measurements of grain length, grain width, seed setting rate, panicle number per plant, panicle length, grain number per panicle, and thousand-grain weight were collected from the experimental base in Chengdu. At least 10 biological replicates were randomly selected and measured for each trait.

Genetic analysis and gene mapping

For genetic analysis, her1 and the wild-type II-32B were crossed reciprocally, and the stigma-related phenotypes of BC₁F₁ plants were observed to determine if they were consistent with either the wild-type or the mutant. By self-pollination, BC₁F₂ populations from both reciprocal crosses were obtained. The stigma-related phenotype of each individual in these populations was identified, and the number of plants with wild-type and mutant phenotypes was recorded. Chi-squared tests were used to verify the segregation ratio and determine the inheritance pattern of the mutant phenotype.

For MutMap analysis, leaves of 50 mutant-phenotype plants from the BC₁F₂ population were pooled in equal proportions, forming a mutant bulk for whole-genome resequencing. The wild-type II-32B was also sequenced as a control. High-quality DNA extraction, library construction, and high-throughput sequencing were performed in collaboration with Wuhan IGENEBOOK Biotechnology Co., Ltd. (on the Illumina NovaSeq 6000 sequencing platform). Fastp (v0.23.4)⁶³ was used to remove sequencing adapters and low-quality data. Clean reads were aligned to the reference genome (MSU7.0, rice.uga.edu) using BWA (v0.7.17)⁶⁴, and potential PCR duplicates were removed using GATK MarkDuplicates (v4.3.0.0). SNP detection, SNP-index calculation, and linkage analysis were performed using the MutMap Pipeline (v2.3.4)⁶⁵, filtering for loci with a SNP-index of 1 within highly confident intervals. Functional annotation was performed using SnpEff (v5.2)⁶⁶ to identify candidate genes and their mutation modes in her1. Protein structures and functional domains were analyzed using the InterPro database (www.ebi.ac.uk/interpro).

Vector construction and genetic transformation

To generate the HER1 knockout construct, a multiplex CRISPR/Cas9 expression vector was generated using the toolkit developed by Yaoguang Liu’s group⁶⁷. Two targets for HER1 and their primers, HER1-KO-Target1-U6a and HER1-KO-Target2-U6b (Supplementary Table 6), were designed using the TargetDesign⁶⁸ web tool. Two sgRNA boxes driven by OsU6a and OsU6b promoters were amplified using overlapping PCR, with pYLsgRNA-OsU6a and pYLsgRNA-OsU6b as templates, and inserted into pYLCRISPR/Cas9P_ubi-H to obtain the final HER1-KO vector.

To knockout DS2, one target and its primers, DS2-KO-U6a (Supplementary Table 6), were designed. The sgRNA box was amplified using overlapping PCR, with pYLsgRNA-OsU6a as template, and inserted into pYLCRISPR/Cas9P_ubi-H to obtain the complete DS2-KO vector.

For the functional complementation assay, the full-length HER1 gene sequence, including 2 kb of its promoter region, was amplified from the genomic DNA of II-32B using the HER1-CP primers with homologous recombination arms (Supplementary Table 6). It was then integrated into pCAMBIA1300 at the EcoRI and SalI sites to construct the HER1-CP vector, using the In-Fusion Snap Assembly Cloning Kit (638946, Takara).

To overexpress HER1, the coding sequence of HER1 was amplified from the cDNA of II-32B stigma using the HER1-OE primers (Supplementary Table 6), and then integrated into pBWA(V)HU-GFP (modified from pCAMBIA1300 with the ubiquitin promoter and a GFP tag, Wuhan BioRun Biosciences Co., Ltd.) to construct the HER1-OE vector.

To overexpress DS2, the coding sequence of DS2 was amplified from the cDNA of II-32B stigma using the DS2-OE primers (Supplementary Table 6), and then integrated into pCAMBIA2300-35s-eGFP (modified from pCAMBIA2300 with the 35 s promoter and an eGFP tag) at the BamHI and SalI sites to construct the DS2-OE vector.

The vectors were introduced into rice plants via Agrobacterium tumefaciens-mediated transformation⁶⁹.

RNA extraction and RT-qPCR

Total RNA was isolated using TRIzol reagent through sample homogenization, chloroform-induced phase separation, and isopropanol precipitation of the aqueous phase, followed by washing the pellet with 75% ethanol⁷⁰. The first-strand cDNA was synthesized from 1 µg total RNA using the HiScript IV RT SuperMix for qPCR (R423, Vazyme), according to the manufacturer’s instructions. Transcript-specific primers were designed and listed in Supplementary Table 7. RT-qPCR was performed on the QX400 platform (Sichuan Jielaimei Technology Co., Ltd.) using the ChamQ SYBR qPCR Master Mix (Q311, Vazyme), following the manufacturer’s instructions. OsActin1 was used as the internal control. Each sample underwent at least three technical replicates.

Subcellular localization analysis

The coding sequence of HER1 was amplified from the cDNA of II-32B stigma using the HER1-SL primers (Supplementary Table 6), and then integrated into pCAMBIA2300-35s-eGFP (modified from pCAMBIA2300 with the 35 s promoter and an eGFP tag) at the BamHI and SalI sites to construct the HER1-SL vector.

The empty vector, pCAMBIA2300-35S-eGFP, and the HER1-SL vector were used for subcellular localization analysis of the HER1 protein. The plasmids were coated onto gold microcarriers and delivered into onion epidermal cells through biolistic acceleration. Specifically, the microcarriers were accelerated by compressed helium gas into the target tissue. The bombarded samples were subsequently maintained in a dark, humid chamber for 24 h before being examined under a fluorescence microscope (BX63, Olympus)⁷¹.

SDS-PAGE and immunoblotting

SDS-PAGE was performed using high-quality gradient precast gels (M00928, GenScript), according the manufacturer’s instructions. The samples were denatured in SDS loading buffer (MB01015, GenScript) at 95 °C for 10 min, then electrophoresed at a constant voltage of 100 V for 1 h, and transferred to PVDF membrane (Immobilon^®-P, Millipore) at a constant current of 280 mA for 2 h. Then the membrane was sequentially blocked at 4 °C in 5% milk for 1 h, incubated with the primary antibody at 4 °C overnight, and incubated with the secondary antibody at room temperature for 1 h, all on a rocking shaker. Imaging was carried out using ECL Chemiluminescence Kits (E422 and E423, Vazyme) on the e-BLOT Touch Imager (Shanghai e-BLOT Life Science Co., Ltd.).

Analysis of histone methylation levels in vivo

Histone methylation analysis was performed following the previously reported protocol⁷² with some modifications. In brief, fresh stigmas from different lines were flash-frozen with liquid nitrogen, then quickly ground into powder, and mixed in nuclei extraction buffer (250 mM sucrose; 15 mM piperazine-N, N-bis (pH 6.8); 5 mM MgCl₂; 60 mM KCl; 15 mM NaCl; 1 mM CaCl₂; 0.5% Triton X-100; 1 mM PMSF; protease inhibitor Cocktail (Roche) at 4 °C for 1 h with rotation. Following centrifugation at 4 °C and 3000 × g, the pellets were resuspended with 0.2 M HCl, vortexed thoroughly, and incubated on ice for 30 min. After centrifugation at 4 °C and 14,000 × g, the supernatant was transferred to a new EP tube, and the pH was adjusted to 8.0 using 1 M Tris-HCl. The samples were denatured and analyzed through SDS-PAGE and immunoblotting. Several antibodies against various lysine sites on histone H3 were used, including H3K4me2 (07-030, Millipore), H3K4me3 (ab8580, abcam), H3K9me1 (07-450, Millipore), H3K9me2 (ab1220, abcam), H3K9me3 (07-442, Millipore), H3K36me1(ab9048, abcam), H3K36me2 (ab9049, abcam), and H3K36me3 (ab9050, abcam), with α-H3 (ab1791, abcam) as a control, all at a dilution ratio of 1:3000. Fiji (v1.54 f)⁷³ was used for grayscale calculations.

Protein expression and purification

To prokaryotically express His-HER1 and GST-HER1 fusion proteins, the coding sequence of HER1 was amplified from the cDNA of rice stigma using His-HER1 and GST-HER1 primers with different homologous recombination arms (Supplementary Table 6), and then integrated into pET-28a and pGEX-4T-1, respectively, using the In-Fusion Snap Assembly Cloning Kit (638946, Takara) to construct the His-HER1 and GST-HER1 vectors. Protein induction was performed using Escherichia coli BL21 (DE3), cultured in 1000 ml of 2 × YT medium with 0.2 mM IPTG and the corresponding antibiotics (Kan⁺ for His-HER1, and Amp⁺ for GST-HER1). The GST-HER1 protein for methyltransferase activity and GST pull-down assays was purified using Glutathione Beads (SA008005, Changzhou Smart-Lifesciences), and the His-HER1 protein for peptide pull-down assay was purified using the Ni-NTA Pre-Packed Gravity Column (C600791, Sangon Biotech), following the manufacturer’s instructions. His-HER1 was further concentrated and buffer-exchanged with a displacement buffer (20 mM Tris-HCl (pH 8.5); 100 mM NaCl) using the Amicon® Ultra Centrifugal Filter (UFC8030, Millipore). His-HER1^1-624, and truncated HER1 Protein, HER1-P1, HER1-P2, and HER1-P3 (GST and His tagged), were also induced and purified for use in the peptide pull-down assay. HER1^1-624 was coded by the sequence from the her1 mutant and amplified using the His-HER1^1-624_. Primers are listed in Supplementary Table 6.

Histone methyltransferase activity assay

In vitro histone methyltransferase activity assays were performed following a method⁷⁴ with modifications. 2 μg of protein, 2 μg of substrate, and 0.1 mM of S-adenosyl methionine (SAM) were incubated in a 50 μl mixture with 50 mM Tris-HCl (pH 8.5), 20 mM KCl, 10 mM MgCl₂, 100 μM ZnCl₂, 10 mM β-mercaptoethanol, and 250 mM sucrose at 37 °C for 6 h. Recombinant mononucleosomes, kindly offered by Prof. Aiwu Dong, were used as the substrate, and recombinant G9a protein (31425, Active Motif) served as a positive control. Reaction was stop by incubating at 95 °C for 10 min in the SDS loading buffer. SDS-PAGE and immunoblotting were performed as described above. Methylation was detected by mass spectrometry and western blotting using the α-H3K9me2 antibody (ab1220, abcam). For further validation, HKMT activity was also tested using a commercial H3K9 methyltransferase activity kit (ab113453, Abcam), according to the manufacturer’s instructions.

Pull-down assays

In the GST pull-down assay, 20 μl of glutathione beads coupled with GST-HER1 and 2 μg of substrate were incubated in 500 μl of the GST pull-down buffer comprising 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 50 μM ZnCl₂ at 4 °C with rotation for 2 h. Calf thymus histones (37244-51-2, Sigma-Aldrich) or recombinant mononucleosomes were employed as the substrates. After washing three times with the GST pull-down buffer, the samples were analyzed by SDS-PAGE, Coomassie Brilliant Blue staining, and immunoblotting using α-H3 antibody (ab1791, abcam).

The peptide pull-down assay was performed following a procedure⁷⁵ with some slight modifications. 20 μl of Pierce Streptavidin Magnetic beads (88816, Thermo Scientific) and 2 μg of biotinylated H3 peptides were first incubated in 500 μl of the peptide pull-down buffer (50 mM Tris-HCl (pH 7.5); 200 mM NaCl; 50 μM ZnCl₂; 0.1% Triton X-100) at 4 °C with rotation for 2 h. The beads coupled with peptides were subsequently washed three times with 1 ml of peptide pull-down buffer, and incubated with His-HER1 or His-HER1^1-624 in the peptide pull-down buffer at 4 °C with rotation for 2 h. After washing three times with peptide pull-down buffer, the samples were analyzed by SDS-PAGE and immunoblotting using α-His antibody (bsm-33004M, Bioss). Animo acid sequences of all the peptides are listed in Supplementary Table 8.

Surface plasmon resonance assays

Surface plasmon resonance (SPR) analysis was performed using a Biacore 1 K instrument (Cytiva). GST-HER1 protein was immobilized onto flow cells in a CM5 sensor chip (BR-1005-30, Cytiva) via an amine-coupling kit (BR100050, Cytiva), according to the manufacturer’s instructions. Solvent correction was performed using a running buffer consisting of 1 × PBS-P + supplemented with 5% (v/v) DMSO. Histone H3 peptides, including H3K4me3, H3K9me1, H3K9me2, H3K9me3, H3K36me1, H3K36me2, and unmethylated H3 peptides, were serially diluted in running buffer and injected over the immobilized protein at a flow rate of 30 μL/min with an association time of 60 s. After each injection, the sensor surface was regenerated using 10 mM glycine-HCl (pH 2.0). SPR sensorgrams were globally fitted to a 1:1 Langmuir binding model using Biacore Insight Evaluation software (Cytiva). Animo acid sequences of all the peptides are listed in Supplementary Table 8.

Cut&Tag, RNA-seq and ChIP-qPCR assays

A polyclonal antibody against HER1 was developed by Wuhan Gene Create Biotechnology Co., Ltd. in rabbits, utilizing a synthesized peptide with the sequence NLGKDDGSQK. Cut&Tag assays were performed following a protocol⁷⁶ with modifications. Briefly, fresh stigmas were collected from pre-flowering spikelets, and each sample was replicated two times biologically. Nuclei were isolated and incubated with primary antibodies against α-HER1 and α-H3K9me2 (ab1220, abcam). After incubation with secondary antibodies, pA-Tn5 transposase was bound to the antibody-chromatin complex. Tagmentation was activated by the addition of Mg²⁺, and the resulting DNA fragments were purified and amplified via PCR to generate libraries for high-throughput sequencing. The library construction and high-throughput sequencing were performed in collaboration with Wuhan IGENEBOOK Biotechnology Co., Ltd. (on the Illumina NovaSeq 6000 sequencing platform). Fastp (v0.23.4)⁶³ was utilized to remove the adapters low-quality reads, and the clean reads were mapped to the reference genome (MSU7.0, rice.uga.edu) using BWA (v0.7.17)⁶⁴. Potential PCR duplicates were removed by GATK MarkDuplicates (v4.3.0.0), and MACS (v2.1.1)⁷⁷ software was utilized for peak calling. PCA was performed by R (4.2.2) to assess the consistency between biological replicates.

For the RNA-seq analyses, fresh stigmas of both wild-type and her1 were collected with three biological replicates for each sample. High-quality RNA extraction, RNA-seq library construction, and high-throughput sequencing were conducted in collaboration with Wuhan IGENEBOOK Biotechnology Co., Ltd. (on the Illumina NovaSeq 6000 sequencing platform). Fastp (v0.23.4)⁶³ was used to remove sequencing adapters and low-quality data, and the clean reads were aligned to the reference genome (MSU7.0, rice.uga.edu) using Hisat2 (v2.2.1)⁷⁸. FeatureCounts (v2.0.6)⁷⁹ was used to calculate the number of reads mapped to each gene, analyzing the gene expression levels across the entire genome. Then the differentially expressed genes between the wild-type and her1 were detected using edgeR (v3.40.2)⁸⁰ in R (4.2.2). PCA was performed R (4.2.2) to assess the consistency between biological replicates.

ChIP-qPCR assays were performed following a protocol⁷² with modifications. Briefly, fresh stigmas of II-32B and her1 were cross-linked with formaldehyde, and chromatin was fragmented into 200–500 bp segments via sonication. The sheared chromatin was then immunoprecipitated using the same antibody used for Cut&Tag. After reversal of cross-links and DNA purification, the enriched DNA fragments were analyzed by qPCR to determine the relative occupancy of HER1 at target genomic regions. Primers are listed in Supplementary Table 7.

WGBS assays

Fresh stigmas of both wild-type and her1 were collected with two biological replicates for each sample. High-quality DNA extraction, bisulfite treatment, WGBS library construction, and high-throughput sequencing were conducted in collaboration with Shanghai BIOZERON Biotechnology Co., Ltd. (on the Illumina NovaSeq 6000 sequencing platform). Fastp (v0.23.4)⁶³ was utilized to remove the adapters low-quality reads, and the clean reads were mapped to the reference genome (MSU7.0, rice.uga.edu) using BSMAP (v2.74). Alignment results were visualized using IGV (v2.16.2).

Haplotype analysis and hybrid seed production trial

The 533 accessions for haplotype analysis were kindly provided by Prof. Qifa Zhang. Genotypes were analyzed and classified into haplotypes using the geneHapR (v1.2.4)⁸¹ package in R (4.2.2). Phenotype and subpopulation data were obtained from a previous study²³. The genomic data of 28 Oryza rufipogon accessions were obtained from the Rice Super Pan-genome Information Resource Database⁸² (RiceSuperPIRdb, www.ricesuperpir.com). Nucleotide diversity (π) of 533 cultivated rice and 28 Oryza rufipogon accessions was analyzed by PopART (1.7) and visualized using R (4.2.2).

Hybridization was performed using the sterile line II-32A as the female parent and her1 as the male parent. Individuals exhibiting the mutant phenotype in each generation were chosen as male parents and underwent repeated backcrossing to II-32A until the resultant line displayed stable inheritance, complete sterility, and a high stigma exsertion rate. The developed high-exertion sterile line was named herA. A large number of seeds of herA and II-32A were propagated, and a seed production trial was conducted in the experimental base in Chengdu, using II-32B as the male parent.

For consistent pollen distribution, the seed production trial utilized a sterile-maintainer interplanting system with a planting density of 15,000 plants per 666.7 m². In the rows of sterile lines, half of each row is planted with herA and the other half with II-32A. Before reaching the flowering phase, each plant underwent leaf pruning and treatment with 920 gibberellins. At the flowering stage, manual pollination was carried out thrice daily with 30-minute intervals until all lines reached the grain filling stage. Upon full maturity of the sterile lines, an extensive collection of individual herA and II-32A plants was conducted using a multi-point random sampling approach to assess their outcrossing rate and seed production yield.

Three restorer lines, Shuhui12, Shuhui527, and Fuhui838, were used for the development of hybrid rice combinations and yield evaluation. The male sterile lines II-32A and herA were respectively crossed with three restorer lines. The F1 hybrids were planted in field plots with 10 plants × 20 rows (200 plants per field plot), covering an area of approximately 10.66 m². Plant height, panicle number per plant, panicle length, grain number per panicle, seed setting rate, thousand-grain weight, and yield per plant were measured at least 10 times (biological replicates). In addition, total grain yield was determined for each experimental plot to assess field-level productivity.

Statistics and reproducibility

All statistical analyses were performed using GraphPad Prism (v10.2.3). Numbers (n) of samples or replicates are indicated in the figures or figure legends. For 2-sample comparisons, significance was analyzed using the two-tailed unpaired Student’s t test, with the P-values presented in the figures. For multiple comparisons, significance was analyzed using one/two-way ANOVA. All western blot results are representative of at least three independent experiments with similar results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

D.G. and K.D. designed and performed the experiments, analyzed the data, and wrote the manuscript with input and biological insight from all co-authors. P.X. and Y.L. (Yongxiang Liao) performed EMS mutagenesis and screened and obtained the mutant. D.G. and Y.L. (Yongxiang Liao) conducted the hybrid seed production trial in the field. K.D. and S.L. performed the haplotype analysis and CRISPR/Cas9 assays. M.L. (Mengyuan Li), Y.L. (Yong Li), X.Y., J.D., J.Z., M.Z., F.J., and G.Z. assisted in the experiments. Z.Z., S.Y., D.X., H.Z. (Hongyu Zhang), X.C., M.L. (Ming Luo), and A.A. modified the manuscript. Y.T. planted the rice materials. H.Z. (Hao Zhou), X.W., and Y.L. (Yongxiang Liao) conceived and supervised this study, and reviewed and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks Yuqing He, Frederic Pontvianne, Rongxin Shen and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

MutMap and WGBS data generated in this study have been deposited in the SRA database under accession PRJNA1141160. RNA-seq and Cut&Tag data have been deposited in the GEO database under accession GSE273302 and GSE273303, respectively. Sequence data can be found in the MSU database for japonica rice Nipponbare (https://rice.uga.edu/) and the MBK database for indica rice R498 (https://mbkbase.org/) under the following accession HER1 (LOC_Os02g47900 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os02g47900]; OsR498G0204626900.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G0204626900.01]), DS1 (LOC_Os01g46970 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os01g46970]; OsR498G0101691600.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G0101691600.01]), DS2 (LOC_Os03g02460 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os03g02460]; OsR498G0305119600.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G0305119600.01]), DS3 (LOC_Os03g20680 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os03g20680]; OsR498G0305848400.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G0305848400.01]), DS4 (LOC_Os10g35070 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os10g35070]; OsR498G1018963700.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G1018963700.01]), DS5 (LOC_Os11g27329 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os11g27329]; OsR498G1120094910.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G1120094910.01]), OsActin1 (LOC_Os03g50885 [https://rice.uga.edu/cgi-bin/ORF_infopage.cgi?orf=LOC_Os03g50885]; OsR498G0306938500.01 [https://www.mbkbase.org/rice/customGT?locus=OsR498G0306938500.01]). Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69149-x.