The SUMO‐conjugating enzyme OsSCE1a from wild rice regulates the functional stay‐green trait in rice

Summary

The functional stay‐green trait is a major goal of rice breeding. Here, we cloned OsSCE1a encoding SUMO‐conjugating enzyme from Yuanjiang common wild rice, which simultaneously regulates the functional stay‐green trait and growth duration. Low expression or knocking out OsSCE1a corresponded to increased chlorophyll content, photosynthetic competence, N use efficiency and a shortened growth period without affecting yield. A natural MITE‐transposon insertion/deletion in the OsSCE1a promoter is the functional variation that regulates these traits. OsSCE1a was selected during evolution and shows significant variation between indica and japonica rice. OsNAC2 interacts with the MITE to enhance OsSCE1a expression. Genetic manipulation of OsSCE1a revealed its potential for rice improvement. OsSCE1a‐mediated SUMOylation of OsGS2 suppresses GS (involved in N assimilation) enzyme activity. OsSCE1a also regulates growth duration by SUMOylating the transcription factor such as OsGBP1, which regulates the expression of the key heading gene Ghd7. Our findings shed light on the role of SUMOylation in crops and provide a strategy for increasing agricultural productivity.

Introduction

Early leaf senescence in indica (Oryza sativa ssp. indica) and hybrid rice is a common phenomenon that strongly influences yield improvement. Chlorophyll (Chl) content is an important indicator of leaf senescence. Many genes involved in Chl synthesis (Jung et al., 2003; Kong et al., 2016; Lee et al., 2005; Sakuraba et al., 2013; Sun et al., 2011; Wang et al., 2010; Wu et al., 2007; Xiong et al., 2021; Yang et al., 2016; Zhang et al., 2006; Zhou et al., 2017) and degradation (Kusaba et al., 2007; Morita et al., 2009; Piao et al., 2017; Sato et al., 2009; Shin et al., 2020) affect leaf senescence in rice. Some mutants of Chl‐degrading genes in rice show the stay‐green trait, whereas most cannot maintain photosynthetic activity. A few functional stay‐green genes that maintain both Chl content and high photosynthetic activity in leaves have been identified. Natural variation of the promoter of the nonfunctional Chl‐degrading gene OsSGR fine‐tunes OsSGR expression, leading to a stay‐green phenotype due to reduced Chl degradation. Introgressing the OsSGR allele from japonica rice into excellent indica rice varieties delayed leaf senescence and enhanced photosynthetic competence, thereby improving the grain‐filling rate and yield (Shin et al., 2020). Leaf senescence is induced by nutrient deficiency, especially under low nitrogen (N) conditions (Shin et al., 2020). OsFd‐GOGAT (Zeng et al., 2017), OsGS2 (Bao et al., 2015) and ARE1 (Wang et al., 2018a) are responsible for regulating N assimilation, thereby affecting leaf senescence. The transcription factors OsGRF4 (Li et al., 2018) and OsDREB1C (Wei et al., 2022) simultaneously activate the expression of genes determining photosynthetic capacity and N utilization, thereby increasing grain yield. In general, stay‐green rice will lead to an extension of the growth period, while yields decline as the growth period shortens. However, a few genes can advance the growth period while not damaging (Fang et al., 2019), or even increase the yield (Wang et al., 2018b; Wei et al., 2022).

SUMOylation is a dynamic, reversible post‐translational modification that finely regulates plant development and stress responses by modulating the activity, stability, location and interactions of proteins (Gill, 2004; Miura et al., 2007; Rosa and Abreu, 2019). In addition, SUMOylation functions as a critical transcriptional modulator in plant cells (Han et al., 2021a, 2021b). Similar to the ubiquitination pathway, the SUMO‐conjugation pathway consists of three enzyme cascades: E1 SUMO‐activating enzyme, E2 SUMO‐conjugating enzyme and E3 SUMO ligase. However, unlike in ubiquitination, E2 directly interacts with its substrate without E3 ligase and specifically recognizes and couples SUMO to some substrates (Gareau and Lima, 2010). To date, only a few studies have explored rice traits regulated by E2 SUMO‐conjugating enzymes (OsSCEs) (Joo et al., 2019; Nigam et al., 2008; Rosa et al., 2018; Skelly et al., 2019). Thus, the mechanism of SUMOylation in crops requires further investigation.

Common wild rice (Oryza rufipogon Griff.) contains abundant favourable alleles (Sun et al., 2001). Here, we cloned and characterized OsSCE1a, a gene encoding a SUMO‐conjugating enzyme from Yuanjiang common wild rice, and indicated OsSCE1a of low expression or OsSCE1a‐KO^93‐11 mutant enhances functional stay‐green and shortens growth duration without yield penalty in rice.

Results

Map‐based cloning and functional analysis of OsSCE1a from Yuanjiang common wild rice

We previously constructed a set of introgression lines (named 9YJ) using a Yuanjiang common wild rice (O. rufipogon) and the elite indica cultivar 93–11 (O. sativa) (Fu et al., 2010). The introgression line 9YJ30 was identified based on higher Chl content compared to 93–11 (Figure 1a,b; Figure S1). Surprisingly, the heading date of 9YJ30 was approximately 5–7 days earlier than that of 93–11 under LD conditions and approximately 3 days earlier under SD conditions (Figure 1a,c).

Figure 1.

Map‐based cloning, functional analysis and expression analysis of OsSCE1a. (a) Phenotypes of 93–11 and 9YJ30 at the heading stage. Scale bar, 20 cm. (b) SPAD values (n = 10) and (c) heading date (n = 20) under long‐day (LD) and short‐day (SD) conditions. SPAD values reflect Chl content. (d) Schematic representation of chromosome 3 to show the location (red bar) of the QTL qHSCH3, including ORF1 (LOC_Os03g03110), ORF2 (LOC_Os03g03120), ORF3 (LOC_Os03g03130, OsSCE1a) and ORF4 (LOC_Os03g03140). (e) Phenotypes of 93–11 and OsSCE1a‐KO^93‐11. Scale bar, 20 cm. (f‐i) Comparison of Chl content (n = 6), heading date in Beijing (n = 20), photosynthetic rate (Pn, n = 6) and Chl fluorescence (Fv/Fm, PSII maximum quantum yield) between 93–11 and OsSCE1a‐KO^93‐11 lines. (j) Expression levels of OsSCE1a in 93–11 and 9YJ30 in first leaves (FL), second leaves (SL), third leaves (TL) and fourth leaves (FTL) from the top at the jointing stage. (k) Expression analysis of OsSCE1a in 93–11 and 9YJ30 in FL at the heading stage (n = 3). (l‐m) Subcellular localization of OsSCE1a‐GFP fusion protein in protoplasts from albino and green seedlings, respectively. GFP, green fluorescent protein; OsMADS6, a nuclear localization marker; Chlorophyll, chloroplast autofluorescence. Scale bars, 10 μm. Data are means ± SD (standard deviation), two‐tailed Student's t‐tests, **P < 0.01: extremely significant difference, *P < 0.05: significant difference, ns: no significant difference.

We constructed an F₂ segregating population from a cross between 93–11 and 9YJ30. The QTL qHSCH3 of regulating Chl content and heading date was identified (Figure S2a,b). We selected plants that were heterozygous at QTL qHSCH3 (markers ei4 and san2) on chromosome 3 but with other introgression fragments that were homozygous and consistent with 93–11 from the F₂ population. Separate populations containing 7864 plants were constructed from the selected individuals to perform fine‐mapping (Figure S2c–g). We generated homozygous lines by screening recombinant plants to verify phenotypes to improve the accuracy of phenotype assessment (Figure S2g). Notably, the Chl content and heading date traits co‐segregated. The interval regulating Chl content and heading date was delimited to a 14.4‐kb region between markers W4 and X9 (Figure 1d; Figure S2g). This region contained four candidate genes: ORF1 with almost no expression; ORF2 encoding a transposon protein; ORF4 was excluded by functional verification (Figure S3); ORF3 (LOC_Os03g03130) encoding a SUMO‐conjugating enzyme OsSCE1a. There was no sequence variation in the coding region, but there are differences in the promoter (~2.6‐kb) of OsSCE1a between 93–11 and 9YJ30, and the 9YJ30 promoter contained a 367‐bp deletion which is a miniature inverted‐repeat transposable element (MITE) (Figure S4a,b). We transformed 93–11 with OsSCE1a knockout (OsSCE1a‐KO^93‐11) (Figure 1e) and overexpression (OsSCE1a‐OE^93‐11) vectors (Figure S4c). OsSCE1a‐KO^93‐11 transgenic plants exhibited increased Chl content and an ~6 to 8 day earlier heading date compared to 93–11 in Beijing (Figure 1f,g), whereas OsSCE1a‐OE^93‐11 transgenic plants (with high OsSCE1a expression) showed reduced Chl content and delayed heading date (Figure S4d,e). We also transformed 9YJ30 with a complementation vector from 93‐11 to 9YJ30 (OsSCE1a‐CPL^9YJ30, Figure S4i). OsSCE1a‐CPL^9YJ30‐positive plants showed increased OsSCE1a expression, reduced Chl content and delayed heading date compared to 9YJ30 (Figure S4j–l). Furthermore, we measured the net photosynthetic rate (Pn, a key photosynthetic parameter) and analysed chlorophyll fluorescence (Fv/Fm, PSII maximum quantum yield) at the heading stage. OsSCE1a‐KO^93‐11 mutant showed high Pn and Fv/Fm (Figure 1h,i), whereas OsSCE1a‐OE^93‐11 (Figure S4g,h) and OsSCE1a‐CPL^9YJ30 (Figure S4m,n) showed low Pn and Fv/Fm values compared to the wild type. In summary, OsSCE1a is the target gene that simultaneously regulates the functional stay‐green trait and early heading date.

OsSCE1a was predominately expressed in photosynthetic leaf tissue of 93–11 and exhibited the highest expression compared to other organs (Figure S5a). OsSCE1a expression decreased during leaf senescence and was higher in 93–11 than in 9YJ30 (Figure 1p,q; Figure S5b). These results indicate that differences of expression in OsSCE1a led to phenotypic differences between 93‐11 and 9YJ30. OsSCE1a protein localized to the cytoplasm and nucleus (Figure 1r), and it particularly localized to chloroplasts (Figure 1s), indicating that OsSCE1a may not only interact with some proteins related to chloroplast localization in the cytoplasm, but also with transcription factors in the nucleus, and possibly cause them to be SUMOylated.

OsSCE1a negatively regulates the functional stay‐green trait and early growth duration

To investigate the role of OsSCE1a in the stay‐green trait, we measured the Chl content in FLs from 93–11, 9YJ30 and OsSCE1a‐KO^93‐11 mutant every 7 days from the jointing stage (~80 days after planting) to the harvesting stage (~150 days) (Figure 2a–g). The Chl content was highest at the heading stage (101–108 days), followed by a decrease. Compared to 93–11, the Chl content of 9YJ30 and OsSCE1a‐KO^93‐11 mutant was higher from Days 101 to 129, and the Chl content of both lines deceased slowly compared to 93–11 (Figure 2g). However, due to the earlier maturity of 9YJ30 and OsSCE1a‐KO^93‐11 mutant, the Chl content of these plants was notably lower than that of 93–11 at the harvesting stage (Figure 2c,f,g). Except during the harvesting stage, the Pn values of 9YJ30 and OsSCE1a‐KO^93‐11 mutant were higher than that of 93–11 (Figure 2h,i), confirming that OsSCE1a regulates the functional stay‐green trait. In addition, the heading dates of 9YJ30 and OsSCE1a‐KO^93‐11 mutant were, respectively, advanced 5–7 days and 6–8 days (Figure 1a,c,e,g), which shortened their growth periods in Beijing (Figure 2j,k).

Figure 2.

OsSCE1a regulate the functional stay‐green trait and growth duration. Phenotypes of 93–11, 9YJ30 and OsSCE1a‐KO^93‐11 mutant at the heading stage (HS) (a,d), maturity stage (MS) (b,e) and harvesting stage (HT) (c,f). (g) The Chl content of 93–11, 9YJ30 and three OsSCE1a‐KO^93‐11 mutant (n = 6) on different days after planting. (h) The Pn of 93–11, 9YJ30 and three OsSCE1a‐KO^93‐11 lines at the HS, MS and HT (n = 5). (i) The diurnal changes in Pn from 8:00 am to 5:00 pm at the heading stage (n = 5). (j,k) Comparison of growth periods (days) among 93–11, 9YJ30 and three OsSCE1a‐KO^93‐11 mutant under LD and SD conditions (n = 20). Data in (b‐g) are means ± SD, two‐tailed Student's t‐tests, **P < 0.01: extremely significant difference, *P < 0.05: significant difference, ns: no significant difference.

OsSCE1a negatively regulates the N use efficiency (NUE)

To explore the physiological basis of leaf senescence, we cultured 93–11 and 9YJ30 in deionized water at the seedling stage to induce nutrient starvation. The Chl content of 9YJ30 was significantly higher than that of 93–11 under starvation (Figure 3a,b), suggesting that leaf senescence was caused by nutrient deficiency. We then carried out low N treatment in the field (Figure 3c). The heading date of 9YJ30 and OsSCE1a‐KO^93‐11 mutant was shortened by ~7 and 9 days, respectively, compared to the control (Figure S6a). The relative [LN(low N) / NN(normal N)] Chl content (Figure 3d), Pn (Figure S6b), Fv/Fm (Figure S6c), tiller number (Figure 3e) and N content in functional leaves (the upper three leaves) (Figure 3f) at the heading stage were higher in 9YJ30 and OsSCE1a‐KO^93‐11 mutant than in 93–11. In addition, compared to 93–11, the yield‐related traits of 9YJ30 and OsSCE1a‐KO^93‐11 mutant, with shorter growth periods, markedly increased under LN conditions (Figure S6d–g), whereas no marked differences were observed under normal conditions in Beijing (Figure 3g; Figure S6h–k). There was no significant difference in harvest index (the ratio of grain yield to aboveground biomass) between 93–11, 9YJ30 and OsSCE1a‐KO^93‐11 mutants (Figure S6l), which indicated that increasing NUE but the harvest index was not damaged on the basis of earlier growth duration. Compared to 93–11, the C content and N content of 9YJ30 and OsSCE1a‐KO^93‐11 mutant were higher not only under NN but also under LN conditions (Figure S6m,n), but the C/N ratios were lower (Figure S6o). The C/N ratio of 93–11 increased under LN conditions, whereas the C/N ratios of 9YJ30 and OsSCE1a‐KO^93‐11 mutant showed no marked differences under LN and NN conditions (Figure S6p–s). Furthermore, the results of seed quality determination showed that the protein content in OsSCE1a‐KO^93‐11 mutant was 11.1% and slightly increased compared to 93–11 with protein content of 10.7%, but OsSCE1a‐KO^93‐11 mutant had a lower chalkiness and higher head rice rate, and there was no significant difference in other quality‐related parameters (Figure S7). These results indicate that a lower C/N ratio is beneficial for delaying leaf senescence (Martin et al., 2002) and that OsSCE1a not only regulates photosynthesis but also regulates N use efficiency (NUE) to impact on the proper balance of C and N, thereby affecting the functional stay‐green trait.

Figure 3.

OsSCE1a regulates NUE. (a) Phenotypes of 93–11 and 9YJ30 plants grown in deionized water at the seedling stage. Scale bar, 2 cm. (b) Comparison of Chl contents of 93–11 and 9YJ30 treated with deionized water at the seedling stage (n = 6). (c) Phenotypes of 93–11, 9YJ30 and OsSCE1a‐KO^93‐11 mutant in the field at the heading stage under nitrogen deficiency conditions. (d‐g) Comparison of relative (LN/NN) SPAD value (n = 6), tiller number (n = 10), N content (n = 6) and grain yield per plant (n = 10) in 93–11, 9YJ30 and OsSCE1a‐KO^93‐11 mutant. Data in (b‐g) are means ± SD, two‐tailed Student's t‐tests, **P < 0.01: extremely significant difference, *P < 0.05: significant difference.

Variation in the OsSCE1a promoter influences the stay‐green trait and heading date

We measured the Chl content of 115 rice cultivars from 39 different countries and regions, including 49 indica and 66 japonica cultivars, and sequenced ~6.6‐kb genomic fragments within the 2601‐bp 5′‐flanking region, 2859‐bp coding region and 1120‐bp 3′‐flanking region of OsSCE1a. Haplotype analysis showed that Hap^93‐11 (majority of indica rice, MITE insertion) and Hap^9YJ30 (majority of japonica rice, MITE deletion) accounted for the highest proportion of haplotypes (Figure S8). An association test showed that all variation sites of OsSCE1a were strongly correlated with Chl content (Figure 4a) and that Chl content was significantly higher in plants harbouring Hap^9YJ30 vs. Hap^93‐11 (Figure 4b). We performed association analysis of polymorphisms in OsSCE1a with heading date in 2055 O. sativa accessions from the IRGCIS database (Mansueto et al., 2017). All polymorphic loci of OsSCE1a were significantly correlated with heading date (Figure 4c). We randomly selected 20 Hap^93‐11 and 20 Hap^9YJ30 cultivars from among the 115 rice cultivars and performed RT‐PCR of plants after 30 days of culture in the greenhouse. OsSCE1a was expressed at significantly lower levels in Hap^9YJ30 cultivars than in Hap^93‐11 cultivars (Figure 4d). These results indicate that structural variation of the OsSCE1a promoter regulates both the stay‐green trait and heading date by controlling the expression of OsSCE1a.

Figure 4.

Variation in the OsSCE1a promoter influences the stay‐green trait and heading date. (a) Association test between all variation sites of OsSCE1a and Chl contents in 115 rice cultivars (49 indica and 66 japonica). The 12 haplotypes (−2571–3961) of OsSCE1a were classified into nine variant types (S1–S9). Hap^93‐11 and Hap^9YJ30 represent the two haplotypes of OsSCE1a in 93–11 and 9YJ30, respectively. (b) Comparison of Chl content in cultivars with Hap^93‐11 and Hap^9YJ30 (n = 5 biological repeats). (c) Association test of Hap^93‐11 and 20 Hap^9YJ30 with heading date in the 2055 O. sativa accessions from the IRGCIS database. (d) Relative expression of OsSCE1a in 20 Hap^93‐11 and 20 Hap^9YJ30 cultivars that were randomly selected from among the 115 rice cultivars (n = 3 biological repeats). (e) Transient expression assays of the OsSCE1a promoter. M1, MITE deletion; M2, MITE insertion (n = 3). (f) Yeast one‐hybrid and (g) EMSA of OsNAC2 and the promoter region of OsSCE1a. (h) Transient expression assay using rice protoplasts harbouring OsNAC2 and the promoter of OsSCE1a (n = 3). (i) Phenotypes of OsNAC2‐KO^93‐11 mutant and 93–11 at the heading stage. (j‐m) Comparison of Chl content, heading date, Pn and Fv/Fm in OsNAC2‐KO^93‐11 mutant and 93–11 (n = 6). Values are means ± SD; all P‐values were calculated by two‐tailed Student's t‐tests, **P < 0.01: extremely significant difference, *P < 0.05: significant difference, ns: no significant difference. (n) Nucleotide polymorphism and neutrality tests of OsSCE1a. N, Number of sequences. L, selected region. S, number of variable sites. H, number of haplotypes. π, Nucleotide diversity. (o) F _ST between Ind and Jap in115 germplasms. (p) Distribution of MITE‐I and MITE‐D in O. rufipogon and 115 germplasms.

The transient expression assays in rice protoplasts indicated that the MITE insertion promotes OsSCE1a expression (Figure 4e). By predicting transcription factor binding sites for MITE, we determined that OsSCE1a might be regulated by OsNAC2 (matched sequence: TTTGATTGAACCACAGGAAAA). OsNAC2 is a negative regulator of plant height and flowering time (Chen et al., 2015) that positively regulates salt‐induced cell death (Mao et al., 2018). Yeast one‐hybrid (Figure 4f), EMSA (Figure 4g) and transient expression assays (Figure 4h) confirmed that OsNAC2 directly binds to the MITE region in the OsSCE1a promoter and activates its expression. We generated OsNAC2‐KO mutant in the 93–11 background (OsNAC2‐KO^93‐11, Figure 4i). OsNAC2‐KO^93‐11 mutant exhibited significantly higher Chl content, earlier heading date and higher Pn and Fv/Fm values than the control (Figure 4j–m). Therefore, OsNAC2 regulates the expression of OsSCE1a, thereby affecting the stay‐green trait and growth duration in rice.

We analysed the nucleotide diversity (π) of OsSCE1a based on publicly available genome resequencing data (Huang et al., 2012; Mansueto et al., 2017) and the result showed that OsSCE1a positive selection during domestication (Figure 4n; Figure S9a). The fixation index (F _ST) analysis pointed to a significant difference between indica and japonica subspecies (Figure 4o; Figure S9b). The rate of MITE‐D was higher in both indica and japonica rice compared to O. rufipogon, and the rate of MITE‐D was far higher in japonica than in indica (Figure 4p; Figure S9c). These results suggest that the MITE was partially eliminated during the evolution of O. sativa rice and was more often selected in japonica rice. Perhaps, the MITE deletion in the OsSCE1a promoter is a favourable variation in O. sativa.

Genetic manipulation of OsSCE1a reveals the potential of this gene for rice improvement

To further verify the role of OsSCE1a in conferring an earlier heading date and the stay‐green trait, we transferred OsSCE1a‐KO into elite indica rice cultivar Huanghuazhan (HHZ) and the superior japonica rice cultivars Wuyungeng 27 (WYG) and Daohuaxiang (DHX) to generate OsSCE1a‐KO^HHZ, OsSCE1a‐KO^WYG and OsSCE1a‐KO^DHX, respectively (Figure 5a–c). The Chl content, Pn and Fv/Fm were significantly elevated in OsSCE1a‐KO mutants (Figure 5d–f; Figure S10a–f), and the heading date were advanced (Figure 5g–i) compared to the wild types.

Figure 5.

Genetic manipulation of OsSCE1a reveals the potential of this gene for rice improvement. (a‐c) Phenotypes of Huanghuazhan (HHZ) and OsSCE1a‐KO^HHZ, Wuyungeng27 (WYG) and OsSCE1a‐KO^WYG, and Daohuaxiang (DHX) and OsSCE1a‐KO^DHX mutants, respectively. Scale bars, 20 cm. (d‐f) Graphs of Chl content (n = 6) vs. (g‐i) heading date (n = 20). Phenotypes of 93–11, 9YJ30 and F₁ plants from a cross between 93–11 and 9YJ30 (j), 93–11, OsSCE1a‐KO^93‐11 mutant, and F₁ plants from a cross between 93–11 and OsSCE1a‐KO^93‐11 mutant (m), 93–11, Y58S and 93–11, Y58S and OsSCE1a‐KO^93‐11 mutant (p) at the heading stage. Scale bar, 20 cm. (k, l, n, o, q, r) Comparison of Chl content (n = 6) and heading date (n = 20) among the three hybrid combinations. Values are means ± SD; all P‐values were calculated by two‐tailed Student's t‐tests, **P < 0.01: extremely significant difference, ns: no significant difference.

There was no significant difference in grain yield per plant in 9YJ30, OsSCE1a‐KO^93‐11 mutant compared to indica rice 93–11, and in OsSCE1a‐KO^HHZ mutant compared to indica rice HHZ in Beijing (Figure S11i–l) and Hunan (Figure S11a–h). However, the OsSCE1a‐KO mutants showed a shorter growth duration than the controls (Figure 2j–k; Figure S11m,n), which is valuable for rice breeding, potentially expanding the planting range to the north and enhancing multi‐season planting in the south. In Beijing, WYG contained almost no filled grains, whereas the seed‐setting rate of OsSCE1a‐KO^WYG mutant was ~60% (data not shown). However, the yield traits of OsSCE1a‐KO^DHX mutant slightly declined in Beijing (data not shown). Therefore, for rice accessions whose growth periods are too long (such as WYG) or too short (such as DHX), knocking out OsSCE1a might not lead to high enough yields when planted in Beijing, but such plants might be successful when grown at a suitable latitude, pointing to the application potential of OsSCE1a.

F₁ plants from a cross between 93–11 and 9YJ30 (Figure 5j), and 93–11 and OsSCE1a‐KO^93‐11 mutant (Figure 5k) showed increased Chl levels (Figure 5m,n) and an earlier heading date (Figure 5p,q) than 93–11. The Pn (Figure S12a,b) and Fv/Fm (Figure S12d,e) were also higher in these plants than in 93–11. When we hybridized 93–11 and OsSCE1a‐KO^93‐11 mutant with the sterile line Y58S, the Chl content of Y58S × OsSCE1a‐KO^93‐11 (Figure 5l,o), Pn (Figure S12c) and Fv/Fm (Figure S12f) increased, and its growth period was shorter than that of Y58S × 93–11 (Figure 5r). In summary, high photosynthetic efficiency and early maturity showed dominant traits, which has great potential for utilizing rice heterosis.

Furthermore, OsSCE1a is closely related to Zm00001eb403740, which encodes an E2 ubiquitin‐conjugating enzyme 15 (uce15) in maize (Figure S13a). The sequences of these proteins are highly conserved across rice and maize (96.88%) (Figure S13b). It is worth verifying whether this homologous gene in maize has similar regulatory functions.

OsSCE1a SUMOylates OsGS2 and suppresses GS enzyme activity

OsSCE1a encodes a SUMO‐conjugating E2 enzyme. IP‐MS showed that OsSUMO1 is a putative interactor of OsSCE1a (Figure S14a). Furthermore, yeast two‐hybrid (Figure S14b), bimolecular fluorescence complementation (BiFC) assays (Figure 6a; Figure S14c) and co‐immunoprecipitation (Co‐IP) (Figure 6b) demonstrated that OsSCE1a interacts with OsSUMO1. We also identified OsGS2 in the IP‐MS data (Figure S14d). OsGS2, a glutamine synthetase, is the key chloroplastic enzyme involved in N assimilation. BiFC (Figure 6a) and Co‐IP (Figure 6c–d) assays indicated that OsSCE1a or OsSUMO1 interacts with OsGS2 in chloroplasts. We used the SUMOylation system (Okada et al., 2009), and the AtSCE1a was replaced by OsSCE1a for SUMOylation analysis. By using anti‐His‐tag and anti‐SUMO antibodies, the molecular weight shift both from 89 KD (42 KD MBP + 47 KD OsGS2) to 107 KD (42 KD MBP + 47 KD OsGS2 + 18 KD His‐AtSUMO1) due to conjugation of AtSUMO1 was observed in samples containing AtSUMO(GG), but not in samples containing AtSUMO(AA) (Figure 6e). The His‐AtSAE2 (Figure 6e) and AtSAE1a‐S (Figure S14e) as E1 enzymes were both detected using anti‐His and anti‐S‐Tag antibodies. OsSCE1a functions as E2 enzyme detected using anti‐S‐Tag antibody (Figure S14e). Compared to 93–11, the total GS enzyme activity of 9YJ30 and OsSCE1a‐KO^93‐11 mutant at heading stage markedly increased (Figure 6f). In conclusion, OsSCE1a directly interacted with OsGS2 and mediated SUMOylation of OsGS2 for suppressing GS enzyme activity.

Figure 6.

OsSCE1a interacts with OsSUMO1 and OsGS2 to regulate GS (Glutamine synthetase) enzyme activity. (a) BiFC assay of the interaction of OsSCE1a with OsGS2 and the interaction of OsSUMO1 with OsGS2 in chloroplasts. (b‐d) Interactions between OsSCE1a and OsSUMO1, OsSCE1a and OsGS2, and OsSUMO1 and OsGS2 confirmed by co‐immunoprecipitation (Co‐IP). (e) In vitro reconstructed SUMOylation assay in E. coli cells showing that OsSCE1a SUMOylated OsGS2. GG, AtSUMO1 with the C‐terminal Gly‐Gly sequence can be covalently attached to the target protein; AA, AtSUMO1 with its C‐terminal Ala‐Ala sequence cannot be covalently attached to the target protein. Coomassie blue staining shows the equal loading of total substrate proteins; the location of the target gene is shown. IB: immunoblotting; His‐AtSAE2 was detected using anti‐His antibody. His‐AtSUMO1 was detected using anti‐SUMO and anti‐His antibody. (f) Comparison of GS enzyme activity in 93–11, 9YJ30 and three OsSCE1a‐KO mutant (n = 3). Data in (f) are means ± SD, two‐tailed Student's t‐tests, **P < 0.01: extremely significant difference.

SUMOylation and ubiquitination are competitive and for preventing the target proteins from being ubiquitination, SUMOylation can increase the stability of the target proteins (Park et al., 2011). However, in our study, the 9YJ30 and OsSCE1a‐KO^93‐11 mutant with low SUMOylation showed high GS enzyme activity, which indicated that SUMO conjugation to target proteins changes the spatial conformation of the substrate and then modulates the activity of enzymes that catalyse the conversion of substrates to products.

OsSCE1a targets and SUMOylates transcription factor such as OsGBP1 to regulate growth period

We performed RNA‐seq using FLs from 93–11 and OsSCE1a‐KO^93‐11 mutant at the heading stage. The differentially expressed genes (DEGs) were enriched in multiple biological processes, and many DEGs related to senescence were identified (Figure 7a,b; Supplemental Data 1 and 2). DEGs associated with heading date were also identified, including OsMADS1, OsMADS14 and OsMADS18 and so on which positive regulates heading date (upregulated, Supplemental Data 1) (Fornara et al., 2004; Jeon et al., 2000; Kim et al., 2007), and Ghd7 (downregulated, Supplemental Data 2) (Xue et al., 2008). In a previous study, upregulating Ghd7 expression delayed heading, and overexpressing Ghd7 inhibited the Ehd1‐Hd3a/RFT1 flowering pathway (Xue et al., 2008; Zong et al., 2021). RT‐PCR analysis confirmed that Ghd7 expression was higher in 93–11, but the expression levels of the florigen genes OsMADS1, OsMADS14, OsMADS18, Hd3a and RFT1 (Komiya et al., 2008) were lower in 93–11 than in 9YJ30 and OsSCE1a‐KO^93‐11 mutant (Figure 7c–h). Together, OsSCE1a promotes OsMADS1, OsMADS14, OsMADS18, Hd3a and RFT1 expression and inhibits Ghd7 expression, suggesting OsSCE1a might influence the expression of genes associated with heading date and then regulate growth duration.

Figure 7.

OsSCE1a SUMOylates OsGBP1 to regulate Ghd7 expression. (a) Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of differentially expressed genes (DEGs) in the FL at the heading stage. (b) Hierarchical clustering of DEGs associated to senescence. The numbers in the boxes represent the number of genes per cluster. (c‐h) Expression verification of genes by RT‐PCR involved in flowering time including Ghd7, OsMADS1, OsMADS14, OsMADS18, Hd3a and RFT1. n = 3, Data are means ± SD, two‐tailed Student's t‐tests were performed to determine significant differences. (i) Yeast one‐hybrid assay and (j) electrophoretic mobility shift assay (EMSA) of OsGBP1 and the promoter region of Ghd7. (k) Schematic representation of the predicted high‐probability SUMOylation sites and mutations in OsGBP1. OsGBP1^R indicates the lysine of OsGBP1 (OsGBP1^K) mutated to arginine. (l) Transient expression assay using rice protoplasts harbouring OsGBP1 and the promoter of Ghd7. Luciferase (LUC) activity was normalized to that of Renilla luciferase (REN). Data are shown as means ± SD (n = 3). Different letters indicate significant differences (one‐way ANOVA followed by post hoc Tukey test; P < 0.05). (m) In vitro reconstructed SUMOylation assay indicating that OsGBP1 undergoes OsSCE1a‐mediated SUMOylation. Coomassie blue staining shows the equal loading of total substrate proteins; the location of the target gene is shown. IB: immunoblotting; His‐AtSAE2 was detected using anti‐His antibody. His‐AtSUMO1 was detected using anti‐SUMO and anti‐His antibodies.

We predicted the transcription factor binding sites in the 2 K promoter region of Ghd7 and identified the transcription factor OsGBP1, which delays flowering (Gong et al., 2018). A yeast one‐hybrid assay (Figure 7i) and EMSA (Figure 7j) confirmed that OsGBP1 could bind to Ghd7. We also predicted high‐probability SUMOylation sites of OsGBP1 (Figure 7k,l). We mutated the three lysines into arginines (OsGBP1^R) and performed a transient expression experiment. The SUMOylation of OsGBP1 promoted the activation of the Ghd7 promoter (Figure 7m; Figure S14f). Furthermore, SUMOylation assays showed that OsGBP1 was SUMOylated by OsSCE1a. Therefore, the OsSCE1a‐mediated SUMOylation of OsGBP1 regulates heading date, thereby affecting the growth duration.

Discussion

How to increase the productivity of the limited cultivated land is an important issue. Leaf senescence and the duration of the growth period are important traits of rice that are directly related to yield and adaptability to different environments. The functional stay‐green trait, which maintains Chl content and photosynthetic activity, ultimately promoting yield, is a goal pursued by breeders. However, there are few reports on the natural variation related to functional stay‐green genes that can directly be applied to rice variety improvement. In general, stay‐green rice will lead to an extension of the growth period, while yields decline as the growth period shortens. Up to now, there is a lack of genes that can simultaneously regulate these two traits to coordinate this contradiction. Here, we identified OsSCE1a, encoding a SUMO‐conjugating enzyme, from Yuanjiang common wild rice. A natural MITE‐transposon insertion/deletion in the OsSCE1a promoter influences expression of OsSCE1a and then regulates the functional stay‐green trait and growth duration.

When the plant enters the reproductive stage, the demand for assimilates increases obviously, which in turn causes nutrient stress on the leaves, leading to functional decline. The low nitrogen content of leaves will cause premature senescence at grain‐filling stage (Borrell et al., 2001). Glutamine synthetase (GS) is a key catalytic enzyme for the assimilation and utilization of inorganic nitrogen to organic nitrogen in plants (Fuentes et al., 2001). Overexpression of GS may affect the regulation and maintenance of photosynthesis, plant growth and development (Fuentes et al., 2001). The dysfunction of OsGS2 results in chlorosis, shorter plant height, reduced tiller number and dry weight (Bao et al., 2015; Cai et al., 2009). The rice orange leaf phytoplasma protein 1(SRP1) recruited OsGS2 to assist its localization to chloroplasts, and the direct interaction between SRP1 and OsGS2 attenuated its enzymatic activity, which suppressed the synthesis of chlorophyll precursors glutamate (Glu) and glutamine (Gln), and the induction of leaf chlorosis (Zhang et al., 2024). Similarly, the chlorosis phenotype is also observed in barley (Brestic et al., 2014) and Arabidopsis (Ferreira et al., 2019), suggesting that low GS activity caused leaf yellowing. This study proves that OsSCE1a can regulate N utilization efficiency during reproductive growth and further affect functional stay‐green trait.

Unlike ubiquitination, SUMOylation involves only a few enzymes, such as two E1, three E2 (OsSCE1a, OsSCE1b and OsSCE1c) (Rosa et al., 2018) and three E3 ligase in rice. The overexpression of E2 (OsSCEs) genes in rice stimulated SUMOylation (Joo et al., 2019), and the OsSCEs affected traits such as abiotic stress responses, plant height, grain weight, grain yield, heading date and immunity (Joo et al., 2019; Nigam et al., 2008; Rosa et al., 2018; Skelly et al., 2019). In a previous study, a mutant with an insertion in the OsSCE1a promoter that reduced its expression showed a slightly reduced global SUMOylation level and early flowering (Rosa et al., 2018). Our findings indicate that OsSCE1a mediates the SUMOylation of OsGS2 (a protein related to N assimilation) inhibiting the activities of the proteins, leading to 9YJ30 with low expression of OsSCE1a and OsSCE1a‐KO mutant showed enhanced N utilization, thus exhibiting the functional stay‐green trait.

Furthermore, an important role of SUMOylation is to modify transcription factors, which can inhibit or promote their transcriptional activation activity (Han et al., 2021a, 2021b). OsSCE1a may SUMOylate some transcription factors, such as OsGBP1 related to flowering time, thereby promoting the expression of Ghd7 (a key gene involved in the Ghd7‐Ehd1‐Hd3a/RFT1 flowering pathway), thus delaying the heading date. Therefore, 9YJ30 (with low OsSCE1a expression) and knockout mutants showed early heading. However, the earlier heading did not affect the yield in 93–11. Perhaps the higher N assimilation efficiency in these plants promoted plant growth and development, thereby delaying senescence and ultimately maintaining yield. Meanwhile, due to the regulation of growth period by OsSCE1a, the functional stay‐green rice does not extend its growth period.

Indica rice shows early leaf senescence, while japonica rice shows late leaf senescence (Abdelkhalik et al., 2005). Japonica rice has a shorter growth period than indica rice to adapt to northern environments (Zhao et al., 2019). OsSCE1a showed obvious differentiation between indica and japonica rice. OsNAC2 bound to the MITE in the OsSCE1a promoter to improve its expression in indica rice, whereas the MITE deletion in the OsSCE1a promoter in japonica rice led to its reduced expression, higher Chl contents and a shorter growth duration. Perhaps the japonica haplotype of OsSCE1a could be transferred into indica rice via conventional breeding, or perhaps OsSCE1a could be knocked out in various rice varieties to improve these traits. In addition, a dominant functional stay‐green trait and earlier heading date could help overcome premature senescence, a common trait in hybrid rice, as well as the inheritance of super parent‐late maturing in hybrid rice. In summary, OsSCE1a from common wild rice simultaneously regulates functional stay‐green and growth duration. This provides a strategy for resolving the contradiction between stay‐green and growth period, as well as between growth period and yield, in order to increasing agricultural productivity. Our functional analysis of OsSCE1a lays the foundation for analysing the roles of SUMOylation in regulating various traits in crops.

Methods

Plant materials and growth conditions

The introgression line 9YJ30, with high Chl content, was derived from a cross between an accession of common wild rice (O. rufipogon Griff.) collected from Yuanjiang County, Yunnan Province, China, as the donor and elite indica cultivar 93–11 as the recurrent parent (Sun et al., 2001).

Gene mapping, gene expression analysis and analysis of agronomic traits were performed using the Yuanjiang‐derived introgression line 9YJ30 and the recurrent parent 93–11. The plants were grown in the experimental field at China Agriculture University in Beijing (40.1° N, 116.1° E) and Hunan Hybrid Rice Research Center in Changsha (Hunan Province) (28° N, 112° E) under natural long‐day (LD) conditions and in the National Biological Breeding Zone in Sanya (Hainan Province) (18.3° N, 109.1° E) under natural short‐day (SD) conditions. Heading date was recorded as the number of days from germination to the day of panicle emergence from the leaf sheath.

To further investigate the selection mechanism of OsSCE1a, 115 germplasm resources from 39 different countries and regions and ~ 2055 cultivated rice accessions (Mansueto et al., 2017) were used for linkage analysis (listed in Supplemental Data 3,4). A total of 446 wild rice accessions (Huang et al., 2012) (Supplemental Data S5) were used to analyse the geographical distribution of the MITE in the promoter of OsSCE1a.

Measuring chlorophyll (Chl) content, net photosynthetic rate (Pn) and chlorophyll fluorescence (Fv/Fm)

The Chl content of the first leaves (FLs) from the top of each plant at the heading stage was evaluated using a SPAD‐502 meter (Soil Plant Analyzer Development, Minolta Camera Co., Osaka, Japan), a portable device for nondestructively estimating leaf Chl content (Li et al., 2014), following the manufacturer's instructions. SPAD values were measured in the upper, middle and base of the leaf; the mean values represent the values from individual plants (Zhang et al., 2014).

The Pn values of FLs in field‐grown plants at the heading stage were measured between 9:30 am and 12:00 noon on a sunny day with a portable photosynthesis measuring system (LI‐6400XT, http://www.ecotek.com.cn). Diurnal changes in photosynthetic rate were measured from 8:00 am to 5:00 pm in the field in Beijing. The parameters of this system were follows: red and blue light with light intensity set to 1000 μmol CO₂ m⁻²·s⁻¹; constant air temperature in the leaf chamber; CO₂ levels controlled using small CO₂ cylinders; and airflow maintained at 400 μmol CO₂ mol⁻¹. Five or six plants were randomly selected to measure the Pn in the middle of the leaves. Data were recorded after the parameters of each group had stabilized.

FLs from plants at the heading stage were selected to measure chlorophyll fluorescence. A dark acclimation clip was used to hold the upper middle of the FL for 20 min. Chlorophyll fluorescence, a nondestructive measure of photosynthetic capacity, was measured using a MINI‐PAM II Series chlorophyll fluorescence system (Heinz‐Walz Instruments, Effeltrich, Germany). Parameters were automatically calculated by the independently operated MINI‐PAM‐II or by WinControl‐3 software. The maximum quantum yield of PSII electron transport was calculated using the following equation: F_V/Fm = (F_m − F₀)/F_m, where F_V is variable chlorophyll fluorescence (Zaltsman et al., 2005).

All values of the above indexes in the control 93–11 were measured at different heading dates.

QTL mapping

QTL analysis of the F₂ segregation population associated with Chl content was analysed by Haley‐Knott regression of R/QTL (1000 permutations, P < 0.05). QTL analysis of the F₂ segregation population related to heading date was performed using MapManager QTXb 20 (Manly et al., 2001).

Plasmid construction and genetic transformation

To generate ORF3 (LOC_Os03g03130, OsSCE1a) and ORF4 (LOC_Os03g03140) knockout plants, CRISPR/Cas9 vectors with guide RNAs targeting OsSCE1a and ORF4 were constructed with 20‐bp gene‐specific spacer sequences. T₂ homozygous transgene‐positive lines were used for subsequent experiments. To generate the OsSCE1a complementation construct, a 6580‐bp genomic fragment from 93–11 containing the entire OsSCE1a sequence with the 2601‐bp 5′‐flanking region and the 1120‐bp 3′‐flanking region was amplified. Then, it was inserted into binary vector pCAMBIA1300 (http://www.cambia.org) between the KpnI and BamHI sites to form the complementation construct (OsSCE1a‐CPL^9YJ30). The ORF4 complementation vector was constructed using a 3469‐bp genomic fragment from 9YJ30 harbouring the 1335‐bp 5′‐flanking region and the 708‐bp 3′‐flanking region (ORF4‐CPL^93‐11). To generate the ORF3 overexpression construct, the 483‐bp OsSCE1a‐OE^9YJ30 sequence was amplified and inserted into the pCAMBIA1301 vector (http://www.cambia.org) between the BamHI and KpnI sites under the control of the maize Ubiquitin promoter. All primers are listed in Supplemental Data S6. Rice cultivar 93–11 was separately transformed with the OsSCE1a‐OE and ORF4‐CPL constructs, and 9YJ30 was separately transformed with OsSCE1a‐CPL and the knockout vector of ORF4. In addition, the elite indica variety Huanghuazhan (HHZ, grown in southern China) and japonica accessions Wuyungeng 27 (WYG, grown in Jiangsu, China) and Daohuaxiang (DHX, grown in Wuchang, Heilongjiang province, China) were transformed with the OsSCE1a knockout vector.

Reverse‐transcription quantitative PCR

Total RNA was extracted from the samples using an RNAprep Pure Plant Kit (TIANGEN, Beijing, China). Total RNA (~2 μg) was used for cDNA synthesis using FastKing gDNA Dispelling RT Supermix according to the manufacturer's instructions. RT‐PCR was performed using a CFX96 Real‐Time System (BIO‐RAD, US). The rice Ubiquitin (LOC_Os03g13170) gene was used as an internal control to normalize the gene expression data using the relative quantification method (2^−ΔΔCT ) (Michaelidou et al., 2013). Each reaction contained 10 ng of first‐strand cDNA, 10 μM gene‐specific primers and 10 μL of real‐time PCR SYBR Mix (TB Green® Premix Ex Taq™, Takara). The sequences of all primers used in RT‐PCR are listed in Supplemental Data S6.

Confocal laser‐scanning microscopy

For subcellular localization of OsSCE1a, two vectors were constructed (Supplemental Data S6): p35S:OsMADS6‐RFP (Li et al., 2010) (expressing a nuclear localization marker) and CaMV35S:OsSCE1a‐GFP. The resulting vector and control plasmids were co‐transformed into rice protoplasts. To examine the chloroplast localization of OsSCE1a, protoplasts extracted from 93–11 seedlings cultured for 14 days on MS medium under a 14‐h‐light/10‐h‐dark cycle were used for analysis.

For the BiFC assays, OsSCE1a, OsSUMO1 and OsGS2 were amplified from 93–11 cDNA (Supplemental Data S6) and recombined into pUC‐SPYNE and pUC‐SPYCE to construct the OsSCE1a^C, OsSUMO1^C, OsSUMO1^N and OsGS2^N vectors, respectively. The relevant sets of vectors were co‐transformed into rice protoplasts.

Fluorescence was examined under a confocal laser‐scanning microscope (LSM900) at 14–16 h after transformation.

RNA‐seq and differential gene expression analysis

RNA samples were extracted from FL of field‐grown 93–11 and OsSCE1a‐KO^93‐11 mutant at the heading stage in Beijing. Paired‐end libraries were constructed and sequenced on the Illumina HiSeq 2500 platform at the Novogene Company (China). The raw reads were mapped to the Nipponbare reference genome. The FPKM (fragments per kilobase per million mapped reads) values of each gene were calculated and the DEGs (|log₂(fold change)| ≥ 0.585, FDR < 0.05) between 93–11 and OsSCE1a‐KO^93‐11 mutant identified (Trapnell et al., 2010). GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) analyses were conducted using agriGO (Tian et al., 2017) (http://systemsbiology.cau.edu.cn/agriGOv2/) and the PlantGSEA toolkit (Yi et al., 2013) (http://structuralbiology.cau.edu.cn/PlantGSEA), respectively.

Measuring total nitrogen (N) and carbon (C) content

To measure total N contents, photosynthetic leaf tissues (FLs, SLs [second leaves], and TLs [third leaves]) were heated in an oven at 105°C for 30 min and dried at 70°C for 3 days. Carbon and nitrogen concentrations were measured in finely homogenized samples using an IsoPrime 100 analyser (Elementar, Germany). Total N levels were determined by the Kjeldahl method as described previously (Yi et al., 2013).

Rice quality measurements

Briefly, about 600 g of full grains harvested from 93–11 and OsSCE1a‐KO^93‐11 mutant were dried for quality analysis. The main measures included brown rice rate, milled rice rate, head rice rate, chalky rice rate, chalkiness, grain length, grain width, length‐width rate, amylose content, gel consistency, alkali spreading value and protein content. The determination is completed by a professional third‐party testing institution (The Rice and Product Inspection and Testing Center of Hunan Rice Research Institute) and according to the Rice Quality Measurement Standards (Ministry of Agriculture, People's Republic of China, 1988).

Transient expression assay in rice protoplasts

The promoter sequences upstream of the OsSCE1a translation start site in 93–11 and 9YJ30 were amplified using specific primers designed with site‐directed mutations at the InDels and SNPs. All upstream fragments were separately inserted into pGreenII 0800 upstream of the firefly luciferase (LUC) gene. The CaMV 35S promoter‐driven Renilla luciferase (REN) gene was used as an internal control.

To analyse transcriptional activation or inhibition, the 93‐11‐OsSCE1a ^pro‐LUC, 9YJ30‐OsSCE1a ^pro‐LUC and Ghd7 ^pro ‐LUC vectors were used as reporters. The coding sequences (CDSs) of OsNAC2 and OsGBP1 were amplified from 93–11 and fused into the pGreenII 62‐SK vector as effectors. The reporters and effectors were co‐transformed into rice protoplasts.

For the transient expression assays, 6–8 μg of plasmids was introduced into 200 μL rice protoplasts using the polyethylene glycol–mediated method and incubated for 16 h at 28 °C in the dark (Shen et al., 2014). Luciferase activity was assayed using the Dual‐Luciferase Reporter Assay System (Promega).

Yeast one‐hybrid assay

The CDSs of OsNAC2 and OsGBP1 were cloned into pB42AD to produce the pB42AD‐OsNAC2 and pB42AD‐OsGBP1 constructs, respectively. The MITE insertion in the promoter of OsSCE1a and the Ghd7 promoter containing OsGBP1 binding sites were inserted into pLacZi2μ to produce the reporters OsSCE1a ^MITE‐pLacZi. Using the PEG/LiAc method, the pB42AD‐OsNAC2 and pB42AD‐OsGBP1 constructs or empty pB42AD vector (control) were co‐transformed with each reporter or empty pLacZi2μ vector (control) into yeast strain EGY48. The clones were cultured on SD/‐Ura‐Trp plates containing X‐Gal for blue colour development for 3 days to detect possible interactions between OsNAC2 or OsGBP1 and the reporter.

Electrophoretic mobility shift assay (EMSA)

The cDNA of OsNAC2 and OsGBP1 containing the DNA‐binding domain was amplified and cloned into the vector pET32a to construct His‐OsNAC2 and His‐OsGBP1, respectively, and transformed into Escherichia coli BL21(DE3) cells. The 5′‐biotin‐labelled DNA fragments listed in Supplemental Data 6 were synthesized; unlabeled DNA fragments of the same sequences were used as the competitors in this assay. Competition for binding was performed with cold (unlabeled) probe containing the binding site. The probes were incubated with nuclear extract at room temperature for 30 min. The entire reaction mixture was run on a non‐denaturing 0.5× TBE 6% polyacrylamide gel for 1 h at 60 V at 4 °C and transferred onto Biodyne® B nylon membranes (Pall Corporation). EMSA was performed using a LightShift Chemiluminescent EMSA kit (Thermo Fisher Scientific).

IP‐MS (immunoprecipitation‐mass spectrometry)

Total proteins were extracted from CaMV35S:OsSCE1a‐GFP transgenic and ZH11 (control) plants with IP buffer (50 mM Tris–HCl pH = 7.5, 150 mM NaCl, 10 mM MgCl₂, 0.1% NP‐40, 3 mM DTT and 1 mL 100× Cocktail). After mixing 150 μL total proteins with 50 μL 4× SDS loading buffer, the sample was denatured in boiling water for 10 min and kept as input. The remaining protein solution was coated with anti‐GFP antibody (ab290, Abcam, https://www.abcam.com/) and incubated for 3 h at 4 °C. The GFP beads were washed three times with wash buffer (1× PBS) and combined with 100 μL 1× PBS. Possible interacting proteins were analysed in the Biological Mass Spectrometry Laboratory at the College of Biological Sciences, China Agricultural University.

Yeast two‐hybrid assay

The cDNAs of OsSCE1a and OsSUMO1 were amplified and ligated into the pGBKT7 and pGADT7 vectors to generate BD‐OsSCE1a and AD‐OsSUMO1, respectively. These constructs were used to transform yeast (Saccharomyces cerevisiae) strain Y2H Gold. The yeast cells were cultured on SD medium lacking Leu and Trp (DDO: SD/−Leu/−Trp) at 30 °C for 2 days, transferred to QDO: SD/−Trp/−Leu/‐His/−Ade and incubated for 3 days. The combination of AD‐NAL1 and BD‐FZP was used as a positive control (Huang et al., 2018), and pGADT7 and pGBKT7 were used as negative controls.

Co‐immunoprecipitation

For the Co‐IP assay, the CDSs of OsSCE1a, OsSUMO1 and OsGS2 were amplified from 93–11 and cloned into pAN580 (pAN580‐GFP, pAN580‐Flag and pAN580‐HA) vectors to produce the pAN580‐GFP‐OsSCE1a, pAN580‐GFP‐SUMO1, pAN580‐HA‐SUMO1 and pAN580‐Flag‐OsGS2 fusion plasmids, respectively. pAN580‐GFP‐OsSCE1a + pAN580‐HA‐SUMO1, pAN580‐GFP‐OsSCE1a + pAN580‐Flag‐OsGS2 and pAN580‐GFP‐SUMO1 + pAN580‐Flag‐OsGS2 were transformed into rice protoplasts. Total proteins from rice protoplasts were incubated with agarose beads (36403ES08, Yeasen) and anti‐GFP antibody (AE012, ABclonal). The immunoprecipitated proteins were detected by immunoblotting with anti‐GFP (ABclonal), anti‐Flag (M185‐3L, MBL) and anti‐HA (AE008, ABclonal) antibodies.

SUMOylation assay

The SUMOylation assay was performed as described previously, with minor modifications (Sun et al., 2001). The BglII/XhoI fragment of OsSCE1a was cloned into pCDFDuet‐SUMO1‐SCE1a to generate pCDFDuet‐SUMO1‐OsSCE1a. The CDSs of OsGS2 and OsGBP1 were cloned into pMAL‐c5X to generate the MBP‐OsGS2 and MBP‐OsGBP1 plasmids. E. coli Rosetta (DE3) cells were transformed with pACYCDuet‐AtSAE1a‐AtSAE2 and used to prepare competent cells. The MBP‐OsGS2/MBP‐OsGBP1 and pCDFDuet‐AtSUMO1(AA or GG)‐OsSCE1a constructs were transformed into competent E. coli cells. The transformed E. coli cells were induced at 16°C until OD₆₀₀ was 1.0, followed by analysis of SUMOylation by immunoblotting (with anti‐His antibody, BE7001, Easybio; anti‐SUMO antibody, PHY0160S, PhytoAB; and anti‐S‐tag antibody, ab199310, Abcam). Coomassie brilliant blue staining was performed to confirm the equal loading of total proteins.

Sequencing analysis

Amino acid sequences homologous to OsSCE1a were downloaded from the National Center for Biotechnology Information website. Then, phylogenetic trees were constructed using the neighbour‐joining algorithm in MEGA version 7 (Kumar et al., 2016). An ~6.6‐kb genomic fragment harbouring the OsSCE1a sequence with the 2601‐bp 5′‐flanking region, 2859‐bp coding region and 1120‐bp 3′‐flanking region was amplified using two pairs of primers. SeqMan 5.0 software from DNASTAR was used to sequence the contigs, and MEGA7 was used for multiple sequence alignment. For nucleotide diversity analysis, the genotype information was processed and imported into DnaSP 5.10 software (Librado and Rozas, 2009) to calculate π and Tajima's D. Published data from 1083 rice accessions (Huang et al., 2012) were used to analyse the fixation index (F _ST) on chromosome 3 between indica and japonica using VCFtools, with a 10‐kb window size (Danecek et al., 2011).

Primers

All primers in this study are listed in Supplemental Data S6.

Statistical analysis

Two‐tailed Student's t‐test was used for data analysis with SPSS version 16 (SPSS Inc., Chicago, IL). Statistical significance was set at P < 0.05.

Funding

This research was supported by the Key Program of the National Natural Science Foundation of China (Grant 32 130 079), Chinese Universities Scientific Fund (Grant 2024TC162), the 2115 Talent Development Program of China Agricultural University and the Guided Project of Sanya Yazhou Bay Science and Technology City (Grant Number SYND‐2021‐4).

Author contributions

X.Y. performed most of the experiments and analysed some of the data. Y.L. helped perform the yeast one‐hybrid, transcriptional activation, EMSA, BiFC and SUMOylation assays. D.L. initially located the QTLs related to stay‐green trait by using introgression lines (9YJs) and screened an introgression line 9YJ30 with high Chl content. J.W. verified preliminary locating result with secondary isolated populations constructed with 93–11 and 9YJ30. J.P. performed fine‐mapping analysis of OsSCE1a. J.Z. and L.L. participated in fine‐mapping of OsSCE1a and phenotypic verification. J.S., Y.X. and Y.L. helped perform the experiments. X.M. and X.Z. analysed the distribution of Hap^93‐11 and Hap^9YJ30 in the 446 Oryza rufipogon rice panel. L.T., F.L., H.S., P.G., R.X., P.Z. and Z.Z. guided the experiments. C.S. provided a good experimental platform and guided the experiments. K.Z. performed and guided some of the experiments and data analysis, wrote part of the manuscript and modified the manuscript. Y.F. designed and guided the experiments, wrote part of the manuscript and modified the manuscript. All authors read and approved the final manuscript.

Competing interests

The authors declare no competing interests.

Supporting information

Figure S1. Graphic depiction of genotype between 93–11 and 9YJ30. The black boxes of represent the chromosome segments of YJCWR, and the white boxes represent the 93–11 genetic background.

Figure S2. Mapping of OsSCE1a. Initial mapping of qHSCH3 (OsSCE1a), a QTL involved in Chl content (a) and heading date (b). ¹: The phenotypic variance explained by the QTL; ²: Additive effects, positive value means YJCWR carried alleles decreasing Chl content. ³: Dominant effect. (c) Phenotypes of 93–11, 9YJ30, and R8 plants. R8 is a recombinant plant. (d, e) Comparison of Chl content (n = 6) and heading date (n = 20) under LD and SD conditions. (f) The interval was narrowed down to a 79.8‐kb region between markers D4 and pg4. Scale bar, 50‐kb. (g) Fine‐mapping of OsSCE1a to a 14.4‐kb interval between markers W4 and X9, containing four ORFs (ORF1–ORF4). R1 through R8 are homozygous lines of recombinant plants. The white and grey regions indicate regions homozygous for the 93–11 genome and the 9YJ30 genome, respectively. Scale bar, 5‐kb. n = 6 for Chl content, n = 20 for heading date, with error bars showing standard deviation.

Figure S3. Functional analysis of ORF4 (LOC_Os03g03140). (a) Genetic structure the sequence differences of ORF4. Black boxes represent exons. The solid black lines between the black boxes represent introns. White boxes represent the 5′‐UTR and 3′‐UTR. Scale bar, 100‐bp. Red lines represents SNPs and one 18‐bp Indel. Phenotypes of 9YJ30 and ORF4‐KO^9YJ30 mutant (b), Comparison of the Chl content (c) and heading date (d). Phenotypes of 93–11 and ORF4‐CPL^93‐11 (e). Scale bars, 20 cm. Comparison of the relative expression (f), Chl content (g) and heading date (h). n = 6 repeats for Chl contents, n = 20 repeats for heading date. Data are means ± SD, two‐tailed Student's t‐tests, ns: no significant difference.

Figure S4. Gene structure of ORF3 (OsSCE1a) and phenotypic analysis of complemented transgenic material. (a) The gene structure of OsSCE1a. Black boxes represent exons. I (−2544–2178) represents the 367‐bp insertion, and D represents the 367‐bp deletion. The yellow box represents the 367‐bp MITE. The solid black lines between the black boxes represent introns. White boxes represent the 5′‐UTR and 3′‐UTR. Red lines represent SNPs and one 8‐bp Indel. Scale bar, 100‐bp. (b) MITE insertion of a 367‐bp sequence in the promoter of OsSCE1a. (c) Phenotypes of 93–11 and overexpression plants (OsSCE1a‐OE^93‐11). Scale bar, 20 cm. (d) Expression analysis of OsSCE1a in 93–11 and OsSCE1a‐OE^93‐11. (e‐h) Comparison of Chl content (n = 6), heading date (n = 20), Pn (n = 6) and Fv/Fm (n = 6) between 93–11 and OsSCE1a‐OE^93‐11. (i) Phenotypes of 9YJ30 and OsSCE1a‐CPL^9YJ30. Scale bar, 20 cm. (j) Expression analysis of OsSCE1a in 93–11 and OsSCE1a‐CPL^9YJ30. (k–n) Comparison of Chl content (n = 6), heading date (n = 20), Pn (n = 6), and Fv/Fm (n = 6) between 9YJ30 and OsSCE1a‐CPL^9YJ30. Data are means ± SD, two‐tailed Student's t‐tests. **P < 0.01: extremely significant difference, *P < 0.05: significant difference.

Figure S5. Expression analysis of OsSCE1a. (a) The expression levels of OsSCE1a in different organs of 93–11 at the heading stage. (b) Comparison of OsSCE1a expression in 93–11 and 9YJ30 at the seedling stage grown in a greenhouse for 30 days. n = 3; two‐tailed Student's t‐tests were performed to determine significant differences, **P < 0.01: significant difference.

Figure S6. Analysis of the functional stay‐green trait. (a) Comparison of heading date of 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant under LN conditions. (b, c) Comparison of relative (LN/NN) Pn and Fv/Fm between 93–11, 9YJ30, and OsSCE1a‐KO^93‐11. (d–g) Comparison of tiller number, grain number per plant, 1000‐grain weight, and grain yields per plant between 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant under LN conditions (n = 10). (h–k) Comparison of tiller number, grain number per plant and 1000‐grain weight, and grain yields per plant between 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant under NN conditions (n = 10). (l) Comparison of harvest index (the ratio of grain yield to aboveground biomass) between 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant under NN conditions in 2024, respectively (n = 10). (m, n) Comparison of C content and N content under LN and NN conditions (n = 6). (o) Comparison of C/N ratio between 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant under LN or NN conditions (n = 6). (p‐r) Comparison of C/N ratio of 93–11 or 9YJ30 or OsSCE1a‐KO^93‐11 mutant under LN and NN conditions (n = 6). (s) Comparison of relative (LN/NN) C/N ratio (n = 6) in 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant. Two‐tailed Student's t‐tests were performed to determine significant differences, *P < 0.05: significant difference, **P < 0.01: significant difference, ns: no significant difference.

Figure S7. Comparison of seed quality‐related parameters of 93–11 and OsSCE1a‐KO^93‐11 mutant (n = 3). Two‐tailed Student's t‐tests were performed to determine significant differences, *P < 0.05: significant difference, ns: no significant difference.

Figure S8. Haplotype analysis of OsSCE1a from 115 rice cultivars. (a, c) Twelve haplotypes of OsSCE1a. I (S1, −2170) indicates the 367‐bp insertion, and D indicates the 367‐bp deletion. Ind, indica population; Jap, japonica population. (b) The SNPs and Indels in 12 haplotypes of OsSCE1a. Black boxes represent exons. The solid black lines between the black boxes represent introns. White boxes represent the 5′‐UTR and 3′‐UTR. Yellow boxes represent the 367‐bp insertion, and red lines represent SNP variation sites. Scale bar, 100‐bp.

Figure S9. Domestication analysis of OsSCE1a. (a) Sliding‐window analysis of nucleotide polymorphism (π) in OsSCE1a. The values were calculated for each sliding window of 10‐kb with an increment of 1‐kb. (b) Fixation index (F _ST) on chromosome 3 between indica and japonica subspecies. (c) Geographical distribution of 446 O. rufipogon accessions with MITE insertions and deletions. i: MITE insertion; d: MITE deletion.

Figure S10. The photosynthetic characteristics of OsSCE1a‐KO mutant. Comparison of Pn (a, b, c) and Fv/Fm (d, e, f) between HHZ and OsSCE1a‐KO^HHZ mutant, WYG and OsSCE1a‐KO^WYG mutant, and DHX and OsSCE1a‐KO^DHX mutant. n = 6, Data are means ± SD, two‐tailed Student's t‐tests, **P < 0.01.

Figure S11. Yield‐related traits and heading dates of OsSCE1a‐KO mutant in the 93–11 and HHZ backgrounds in Hunan and Beijing. (a‐d) Comparison of tiller number, grain number per plant, 1000‐grain weight, and grain yield per plant in 93–11, 9YJ30, and OsSCE1a‐KO^93‐11 mutant in Hunan (n = 10). (e‐l) Comparison of tiller number, grain number per plant, 1000‐grain weight, and grain yields per plant in HHZ and OsSCE1a‐KO^HHZ mutant in Hunan (e‐h) (n = 10) and Beijing (i‐l) (n = 10). (m, n) Comparison of heading date between HHZ and OsSCE1a‐KO^HHZ mutant in Hunan and Beijing (n = 20). Data are means ± SD, two‐tailed Student's t‐tests, *P < 0.05: significant difference, **P < 0.01: significant difference, ns: no significant difference.

Figure S12. Comparison of Pn and Fv/Fm values in different rice lines. Comparison of Pn (a‐c) and Fv/Fm (d‐f) values between 93–11, 9YJ30, and F₁ (93–11 × 9YJ30), 93–11, OsSCE1a‐KO^93‐11 mutant and F₁ (93–11× OsSCE1a‐KO^93‐11), 93–11, F₁ (Y58S × 93–11), and F₁ (Y58S × OsSCE1a‐KO^93‐11), respectively. n = 6. Data are means ± SD, two‐tailed Student's t‐tests, *P < 0.05: significant difference, **P < 0.01: significant difference, ns: no significant difference.

Figure S13. Phylogenetic analysis of OsSCE1a. (a) Phylogenetic tree of OsSCE1a and homologues in other plant species. The neighbour‐joining tree was constructed using MEGA 7 software. (b) Comparison of the amino acid sequences of OsSCE1a in rice and Zm00001eb403740 (uce15) in maize (Zea mays).

Figure S14. OsSCE1a interacts with OsSUMO1 and OsGS2, and OsSUMO1 interacts with OsGS2. (a, d) Mass spectrometry of OsSCE1a and OsSUMO1 (score = 159), and OsSCE1a and OsGS2 (score = 74). (b, c) Yeast two‐hybrid and BiFC assays of the interaction of OsSCE1a with OsSUMO1. QDO indicates SD/−Trp/−Leu/‐His/−Ade, and DDO indicates the control medium SD/−Trp/−Leu. Scale bars, 10 μm. (e, f) The E1 activating enzyme AtSAE1a and the E2 conjugating enzyme OsSCE1a were detected by immunoblotting with anti‐S antibody. IB: immunoblotting.

Data S1.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and the Supplemental Information and are available from the corresponding author upon request. The NCBI accession numbers of the genomic and cDNA sequences of the OsSCE1a allele from 93–11 and 9YJ30 are OQ559940 and OQ559941, respectively. The RNA‐seq data have been deposited in the Sequence Read Archive (SRA) under accession code PRJNA939027.