The OsbZIP35-COR1-OsTCP19 module modulates cell proliferation to regulate grain length and weight in rice

Abstract

Grain size substantially influences rice quality and yield. In this study, we identified CONTROL OF SLENDER RICE 1 (COR1), a quantitative trait locus encoding an F-box protein that enhances grain length by promoting cell proliferation. The transcription factor OsbZIP35 represses COR1 expression, while COR1 interacts with OsTCP19, leading to its degradation. Knockout of either OsbZIP35 or OsTCP19 results in increased grain length, confirming their regulatory roles in grain development. The OsbZIP35-COR1-OsTCP19 module controls the expression of the cell cycle gene OsCycB1;4, thereby modulating cell proliferation and ultimately determining grain size. Five haplotypes of COR1 were identified, with COR1-Hap1 and COR1-Hap3 being elite alleles associated with longer grains. Field plot trials demonstrated that the near-isogenic line NIL-COR1^SLG increased yield by ~6.6% compared to NIL-COR1^Nip. These findings elucidate the genetic mechanisms underlying grain size regulation and offer promising strategies for improving rice yield.

Grain size is regulated by the OsbZIP35-COR1-OsTCP19 module, which offers genetic targets for improving rice yield.

INTRODUCTION

The yield of rice (Oryza sativa L.) is mainly determined by quantitative trait loci (QTLs) associated with its constituent traits, such as number of panicles, grain number per panicle, and grain weight. “Grain size” is a key trait that determines grain weight and can, therefore, affect rice yield (1). Therefore, determining important genes that regulate grain size and characterizing their molecular regulatory networks can provide a theoretical basis and genetic resources for the scientific breeding of high-yield rice (2).

In recent years, scientists have identified several QTLs or genes associated with rice grain length using map-based cloning or genome-wide association studies (GWAS). These findings provide important approaches for the development of high-yield rice varieties. Recent research progress has identified several signaling pathways that regulate grain size, including the ubiquitin proteasome pathway, G protein signaling pathway, mitogen-activated protein kinase signaling pathway, plant hormone perception and homeostasis, as well as some transcription regulatory factors (3). Among them, the ubiquitin-proteasome pathway has been shown to play a pivotal role in the regulation of rice grain size (4). GRAIN WIDTH AND WEIGHT 2 (GW2) is a negative regulator of grain width and weight, encoding a Really Interesting New Gene (RING) protein with E3 ubiquitin ligase activity, and negatively regulates cell division by targeting its substrate to the proteasome for degradation (5). GW2 can interact with WIDE GRAIN 1 (WG1) and ubiquitinate it, thereby regulating the stability of WG1 protein. WG1 can also interact directly with the transcription factor (TF) OsbZIP47 and recruit transcriptional corepressors ABERRANT SPIKELET AND PANICLE1 (ASP1) to inhibit the transcriptional activity of OsbZIP47, thereby regulating the expression of downstream genes (6). Histone acetyltransferase GRAIN WEIGHT 6a (GW6a) is an important regulator of rice grain size and yield. It interacts with the ubiquitin receptor HOMOLOG OF DA1 ON RICE CHROMOSOME 3 (HDR3), which has a ubiquitin interaction motif to delay GW6a degradation by the 26S proteasome. This interaction ultimately results to changes in grain size and weight (7). The large grain 1-D (lg1-D) is a dominant mutant that increases grain width and thousand grain weight by 30.8 and 34.5%, respectively, compared to the wild type. Map-based cloning has identified LG1 as a gene encoding ubiquitin-specific protease 15 (OsUBP15), which shows ubiquitinase activity in vitro. The down-regulation of OsUBP15 leads to the narrowing and shrinking of rice grains, emphasizing its important role in regulating the rice grain width and size (8). WIDE AND THICK GRAIN 1 (WTG1) encodes a tubular protease with deubiquitinating activity, and the phenotypic traits of transgenic materials suggest its potential to enhance both the size and yield of rice grains (9).

F-box proteins represent a substantial family within eukaryotes, distinguished by a conserved motif known as the F-box, comprising ~40 amino acids. As a substrate-targeting subunit of the S phase kinase-associated protein 1–Cullin 1–F-box protein (SCF) ubiquitin ligase complex, F-box protein plays a crucial role in the controlled degradation of cellular proteins (10, 11). They play an important role in protein ubiquitination and degradation. At present, researchers have found that plant F-box genes genetically control many key processes, such as embryogenesis, hormone response, seedling development, floral organogenesis, aging, and disease resistance (12, 13). Moreover, F-box proteins appear to be a crucial constituent in modulating the mechanisms of plant growth and development throughout the entire plant life cycle, and their expression is influenced by light and abiotic stress (14–16). There are 779 potential F-box proteins predicted in rice and can be divided into 42 families based on the domains in the C-terminal regions, which play important roles in regulating rice growth, development, and various stress responses (17). For example, OsFBX156 targets the heat shock protein 70 (OsHSP71.1) for degradation and promotes the generation of reactive oxygen species and the expression of pathogenesis-related genes, thereby positively regulating a previously unidentified mechanism of rice innate immunity (18). Similarly, OsFBX257 interacts with 14-3-3 rice proteins GF14b and GF14c and modulates drought stress adaptations by regulating root development (19). Up to now, only two F-box protein encoding genes conferring rice grain size have been identified by reverse genetics. The regulatory network involving the OsFBX206 and the OVATE family proteins governs the biosynthesis of brassinosteroids, thereby modulating the grain size and yield of rice (20). OsFBK12 also targets the substrate S-adenosyl-l-methionine synthetase (OsSAMS1) for degradation, thereby triggering alterations in ethylene levels and regulating leaf senescence and grain size (21). However, there is scarce knowledge at present regarding the function of F-box genes, especially the natural variation of F-box protein encoding genes, in influencing rice grain size.

Here, we isolate a new gene, COR1, from the large-grain germplasm, SLG, that regulates grain length and weight in rice by map-based cloning. COR1 encodes a protein that belongs to the F-box protein FBA subfamily, contains the F-box and FBA_1 superfamily domain, and positively regulates rice grain length by promoting cell proliferation. We also demonstrated the regulatory role of the OsbZIP35-COR1-OsTCP19 module in controlling rice grain length and weight, which has considerable potential for improving rice yield.

RESULTS

Map-based cloning of COR1

To identify the QTLs influencing the size of rice grains, we selected the improved temperate japonica rice variety SLG-1 (SLG) with a large grain as the donor parent and crossed it with the typical temperate japonica rice variety Nipponbare (Nip) as the recipient parent to construct the segregating population. The grain length, grain width, and 1000-grain weight, as well as plant height and panicle length of SLG, were significantly higher than those of Nip, but the grain number per panicle and seed setting were significantly lower than those of Nip, with no obvious difference in tiller number (Fig. 1, A to D, and fig. S1). We carried out QTL analysis by using the BC₄F₂ and BC₄F₃ populations. Two QTLs located ~100 kb apart from each other were discovered on the long arm of chromosome 1 and were named as qGL1a and qGL1b, respectively. The individual harboring qGL1a and qGL1b was selected to backcross with Nip for BC₅ segregating population construction (fig. S2). In this study, we focused on qGL1a localized within the region between RM11800 and RM3447. To eliminate the influence of qGL1b, we identified five recombinant plants (N2 to N6), of which the qGL1b segments were from Nip. Among them, the grains of N2 and N3 plants were shorter, similar to those of the control group (N7), suggesting that the qGL1a gene was located on the right side of Wn35234. The grains of N4, N5, and N6 plants were longer, similar to those of SLG, indicating that the qGL1a gene was located on the left side of Wn35243. Eventually, the qGL1a gene was confined to the interval between the molecular markers Wn35234 and Wn35243, with a physical distance of 8.24 kb (Fig. 1E). This candidate region encompassed one open reading frame, LOC_Os01g60920, and its promoter region. According to the Rice Genome Annotation Project (http://rice.plantbiology.msu.edu/), LOC_Os01g60920 encodes OsFBX30, which pertains to the FBA subfamily of F-box proteins and encompasses the F-box and FBA_1 superfamily domain (fig. S3). Through genomic DNA sequence analysis, it was revealed that LOC_Os01g60920 has two single-nucleotide polymorphisms (SNPs) and one indel in the promoter region between Nip and SLG, as well as two synonymous SNPs and two nonsynonymous SNPs (S35239149, G-to-A, 60Gly-to-Ser; S35239865, G-to-C, 265Glu-to-Gln) in the coding region (Fig. 1F and fig. S3). We postulate that LOC_Os01g60920 may be the candidate gene for qGL1a and designate it as CONTROL OF SLENDER RICE 1 (COR1).

Fig. 1. Map-based cloning of COR1.

(A) Mature grains of Nip and SLG. (B to D) Statistical results for grain length (B), grain width (C), and 1000-grain weight (D) of Nip and SLG. Data represent means ± SD (n = 15). (E) Fine mapping of COR1. Left: High-resolution mapping. Right: Phenotypes of homozygous recombinants. (F) The gene structure and allelic variation of Nip and SLG genes on COR1. The yellow and gray boxes, respectively, represent the coding region and the 5′UTR and 3′UTR. (G) Mature grains of the NILs. Scale bars, 10 mm. (H) Mature grains of NIL-COR1^Nip the CO lines. (I to K) Statistical results for grain length (I), grain width (J), and 1000-grain weight (K) of NIL-COR1^Nip and NIL-COR1^SLG. Data represent means ± SD (n = 30). (L to N) Statistical results for grain length (L), grain width (M), and 1000-grain weight (N) of NIL-COR1^Nip and the CO lines. Data represent means ± SD (n = 30). **P < 0.01, Student’s t test. Different letters indicate statistically significant differences at P = 0.05 by one-way analysis of variance (ANOVA). Scale bars, 10 mm (A, G, and K).

Subsequently, we obtained a pair of near-isogenic lines (NILs) at the qGL1a locus in the genetic background of Nip: NIL-COR1^Nip, which contains the qGL1a fragment from Nip, and NIL-COR1^SLG, which contains the qGL1a fragment from SLG (fig. S4). The grain length and 1000-grain weight of NIL-COR1^SLG were significantly higher than those of NIL-COR1^Nip; however, its grain width was significantly lower (Fig. 1, G to J). These results suggest that the qGL1a allele of SLG influences the grain size and enhances the grain weight.

To determine whether COR1 is a candidate gene for qGL1a, we constructed a complementary vector by fusing its native promoter and coding region of SLG, ProCOR1^SLG:COR1^SLG, and transferred it into NIL-COR1^Nip plants (fig. S5A). Two positive complementary plants with COR1 elevated expression, COR1-CO-1 and COR1-CO-6, were selected for further phenotypic analysis. The grain length, grain width, and 1000-grain weight were significantly higher than those of the recipient plant NIL-COR1^Nip (Fig. 1, K to N), but the grain number per panicle was significantly lower than that of NIL-COR1^Nip, without substantial alternation in plant height, tiller number, and seed setting (fig. S5, B to L). These results indicate that COR1 can compensate for the grain length and 1000-grain weight phenotype of NIL-COR1^Nip, implying that LOC_Os01g60920 is the causal gene of COR1.

COR1 positively regulates grain length and weight

To further confirm the function of COR1, we obtained two independent mutants, COR1-CR-8, and COR1-CR-24, in the Nip background through CRISPR-Cas9 genome editing, where COR1-CR-8 had a 4–base pair (bp) deletion in the first target site and a 7-bp deletion in the second target site compared to the wild-type Nip, and COR1-CR-24 had a 206-bp deletion spanning the two target sites (fig. S6A). Compared with Nip, the grain length, grain width, and 1000-grain weight of COR1-CR-8 and COR1-CR-24 significantly decreased (Fig. 2, A to D), while the grain number per panicle was significantly increased, without substantial alternation in tiller number and panicle length (fig. S6, B to J). To investigate whether the differences in the coding region of COR1 can affect grain size, we separately constructed two overexpression vectors driven by the CaMV35S promoter using the COR1 coding regions of Nip and SLG, namely, 35S: COR1^Nip and 35S: COR1^SLG, and transformed them into Nip (fig. S7A). The COR1^Nip-OE-23 and COR1^Nip-OE-31, as well as COR1^SLG-OE-10 and COR1^SLG-OE-26 with significantly high expression were selected for phenotyping analysis (fig. S7, B to E). Although the grain length and 1000-grain weight of COR1^Nip-OE-23, COR1^Nip-OE-31, COR1^SLG-OE-10, and COR1^SLG-OE-26 plants were all significantly higher than those of Nip, the grain length and 1000-grain weight of COR1^SLG-OE-10 and COR1^SLG-OE-26 plants were significantly higher than those of COR1^Nip-OE-23 and COR1^Nip-OE-31 plants, and all of them showed slender grains with significantly reduced grain width (Fig. 2, E to H). In addition, the grain number per panicle and seed setting of COR1 overexpression plants were reduced compared to those of Nip, without any obvious change in tiller number (fig. S7, G and K to M). These results suggest that COR1 positively regulates rice grain length and 1000-grain weight, and the differences in the coding region might contribute to the grain length variation between Nip and SLG.

Fig. 2. COR1 positively regulates grain length and weight.

(A) Mature grains of Nip and COR1-CR plants. (B to D) Statistical results for grain length (B), grain width (C), and 1000-grain weight (D) of Nip and the COR1-CR plants. Data represent means ± SD (n = 15). (E) Mature grains of Nip, COR1^Nip-OE, COR1^SLG-OE, and SLG plants. (F to H) Statistical results for grain length (F), grain width (G), and 1000-grain weight (H) of Nip, COR1^Nip-OE, COR1^SLG-OE, and SLG plants. Data represent means ± SD (n = 30). Different letters indicate statistically significant differences at P = 0.05 by one-way ANOVA. Scale bars, 10 mm (A and E).

COR1 promotes cell proliferation and regulates cell number

To explore the cellular basis of grain size regulation mediated by COR1, we analyzed the cell number and cell size in the central lemma of mature grains through scanning electron microscopy. We observed that in the longitudinal direction of the spikelet, the cell number of NIL-COR1^SLG was significantly greater than that of NIL-COR1^Nip, yet, in the transverse direction, it was significantly lower, which was consistent with the slender grain of NIL-COR1^SLG compared with NIL-COR1^Nip (Fig. 3, A to C). We also conducted a statistical analysis of the total cell number within the glume and found that the cell number of NIL-COR1^SLG was significantly higher than that of NIL-COR1^Nip, but there was no notable difference in cell size between them (Fig. 3, D and E). We also observed that, in comparison with Nip, the cell number was significantly decreased in COR1-CR plants in the longitudinal direction, while the cell numbers in COR1^Nip-OE and COR1^SLG-OE plants were significantly increased in the longitudinal direction but declined in the transverse direction. We statistically analyzed the total cell number within the glume and found that the cell numbers in COR1^Nip-OE and COR1^SLG-OE plants were significantly increased compared with Nip, while the cell number in COR1-CR plants was significantly reduced (fig. S8). These data imply that COR1 regulates grain size mainly through alterations in cell number rather than cell size, suggesting that it may be involved in the regulation of cell proliferation.

Fig. 3. COR1 regulates grain length by promoting cell division.

(A) Scanning electron micrographs of the outer surfaces of glumes of NIL-COR1^Nip and NIL-COR1^SLG. (B to E) Statistical results for numbers of outer epidermal cells in the longitudinal direction (B) and the transverse direction (C), number of outer epidermal cells (D), and cell size in the outer epidermis (E) of NIL-COR1^Nip and NIL-COR1^SLG. Data represent means ± SD (B to D, n = 10; and E, n = 50). (F and G) Percentage of cell numbers during the G₁, S, or G₂-M phase for NIL-COR1^Nip and NIL-COR1^SLG. Data represent means ± SD (n = 6). (H and I) Percentage of cell numbers during the G₁, S, or G₂-M phase for Nip, COR1-CR, COR1^Nip-OE, and COR1^SLG-OE plants. *P < 0.05; **P < 0.01, Student’s t test. Scale bars, 200 μm (A).

Subsequently, flow cytometry analysis was performed on young panicles of NIL-COR1^Nip and NIL-COR1^SLG to quantify the distribution of cells across different phases of the cell cycle. The results revealed that, compared to NIL-COR1^Nip, NIL-COR1^SLG exhibited a reduced proportion of cells in the G₁ phase and an increased proportion of cells in the S and G₂-M phases. These findings suggest a notable activation of cell division in NIL-COR1^SLG (Fig. 3, F and G). Additionally, we found that several key genes involved in cell cycle regulation, including CyCU4;3, CyCD3, CyCD6, and CDKA1, were significantly up-regulated in NIL-COR1^SLG (fig. S9). Flow cytometry analysis further demonstrated that, compared to Nip, the young panicles of COR1-CR plants exhibited a higher proportion of cells in the G₁ phase and a lower proportion of cells in the S and G₂-M phases. In contrast, COR1^Nip-OE and COR1^SLG-OE plants showed a reduced proportion of cells in the G₁ phase and an increased proportion of cells in the S and G₂-M phases (Fig. 3, H and I). These findings indicate that COR1 activates cell division, which subsequently leads to a increase in grain length. Collectively, these results suggest that COR1 regulates grain size by promoting cell division.

To compare the grain filling rates between NIL-COR1^Nip and NIL-COR1^SLG, we measured the dry and fresh weights of grains at various stages of the grain filling process. During the 12 to 18 days after pollination (DAF) period, the dry weight of grains in NIL-COR1^SLG increased more rapidly compared to that in NIL-COR1^Nip. From 21 to 30 DAF, both the fresh and dry weights of grains in NIL-COR1^SLG were significantly higher than those in NIL-COR1^Nip, with both parameters reaching their peak at 24 DAF (fig. S10). To further elucidate the impact of COR1 on grain filling rate, we measured the dry and fresh weights of grains at various stages of the filling process in Nip, COR1-CR, COR1^Nip-OE, and COR1^SLG-OE. The results demonstrated that, from 12 to 18 DAF, the rate of dry weight increase in COR1^Nip-OE grains was significantly higher than in Nip grains, whereas the rate in COR1-CR grains was significantly lower than in Nip grains. From 6 to 18 DAF, the dry weight increase rate in COR1^SLG-OE grains was significantly greater than in Nip grains and other materials. Additionally, from 6 to 30 DAF, both the fresh and dry weights of COR1^SLG-OE grains were significantly greater than those of Nip grains and other materials (fig. S11). These results collectively suggest that COR1 plays a crucial role in regulating grain length and grain filling rate, ultimately contributing to an enhanced 1000-grain weight.

Expression pattern of COR1 and its encoding protein subcellular location

To further elucidate the functional role of COR1, quantitative reverse transcription polymerase chain reaction (qRT-PCR) was conducted to analyze the accumulation of COR1 mRNA transcripts during various developmental processes. The results revealed that COR1 was expressed in all examined tissues, with particularly high transcriptional levels observed in young panicles (fig. S12A). This expression pattern aligns with its proposed biological function in regulating rice grain size. Furthermore, β-glucuronidase (GUS) staining confirmed COR1 expression in roots, stems, leaves, and young panicles, corroborating the qRT-PCR findings (fig. S12B). Furthermore, we analyzed COR1 mRNA accumulation using public transcriptomic data from the plant eFP Browser database. The analysis showed that COR1 exhibited the highest expression in young leaves, followed by the shoot apical meristem and young panicles (P1 and P2). We also identified COR1 expression in grains across developmental stages S1 to S5, with peak expression observed at stage S2 (fig. S13). Additionally, transient transformation experiments demonstrated that both COR1^Nip–green fluorescent protein (GFP) and COR1^SLG-GFP fusion proteins were localized in nucleus of rice protoplasts (fig. S12C).

OsbZIP35 acts upstream of COR1 to regulate grain length

To further explore the regulatory network of COR1, we analyzed its promoter sequence and identified a variety of cis-regulatory elements (fig. S14). Among these, six potential G-box motifs were detected, which are known to serve as binding sites for basic leucine zipper (bZIP) TFs. In rice, a total of 89 bZIP TFs have been predicted, most of which have been reported to play roles in development and stress responses (22). Among these, OsbZIP35, a TF implicated in regulating grain number per panicle, was previously identified through GWAS in our laboratory and was selected for further investigation. To explore its interaction with COR1, we divided the COR1 promoter into six fragments (ProCOR1-I to ProCOR1-VI) and examined their binding affinity to OsbZIP35 using yeast one-hybrid (Y1H) assays. The results revealed that OsbZIP35 potentially binds to ProCOR1-I, ProCOR1-II, and ProCOR1-V in yeast cells (Fig. 4, A and B). To validate these findings in planta, we performed chromatin immunoprecipitation (ChIP) followed by qPCR on young panicles. The anti-flag antibody exhibited significant enrichment in the amplified fragments spanning ProCOR1-I (Fig. 4C). Furthermore, electrophoretic mobility shift assays (EMSAs) demonstrated that glutathione S-transferase (GST)–OsbZIP35 fusion proteins, but not GST alone, could bind to ProCOR1-I (Fig. 4D). Additionally, transient expression assays in rice protoplasts showed that OsbZIP35 suppressed the promoter activities of both ProCOR1^Nip and ProCOR1^SLG, with no significant differences observed between the two (Fig. 4, E and F). To further confirm the regulatory role of OsbZIP35, we generated osbzip35 mutants, bZIP35-CR-15 and bZIP35-CR-27, using the CRISPR-Cas9 system in the Nip background (fig. S15A). Notably, the expression of COR1 was significantly up-regulated in both bZIP35-CR-15 and bZIP35-CR-27 compared to Nip (Fig. 4G). Collectively, these results demonstrate that OsbZIP35 directly binds to the COR1 promoter and acts as a transcriptional repressor of COR1 expression.

Fig. 4. OsbZIP35 acts upstream of COR1.

(A) The schematic diagram for promoter and its G-box binding motifs of COR1. The red arrowheads indicate G-box binding motifs. (B) Y1H assay showing interaction between OsbZIP35 with the COR1 promoter. (C) ChIP-qPCR analysis of OsbZIP35 binding to the region of ProCOR1. Promoter region VII was used as negative control. Data represent means ± SD (n = 3). (D) EMSA assay showing the binding of OsbZIP35 to ProCOR1. (E and F) The transient dual-LUC reporter assay conducted in rice protoplasts showed the effect of OsbZIP35 on ProCOR1^Nip and ProCOR1^SLG. Data represent means ± SD (n = 6). (G) The relative expression level of COR1 in young panicles of Nip and bZIP35-CR plants. Data represent means ± SD (n = 3). (H) Mature grains of Nip and bZIP35-CR plants. (I to K) Statistical results for grain length (I), grain width (J), and 1000-grain weight (K) of Nip and bZIP35-CR plants. Data represent means ± SD (n = 15). (L) Mature grains of Nip, COR1-CR, bZIP35-CR, and COR1-CR;bZIP35-CR plants. (M to O) Statistical results for grain length (M), grain width (N), and 1000-grain weight (O) of Nip, COR1-CR, bZIP35-CR, and COR1-CR;bZIP35-CR plants. Data represent means ± SD (n = 30). *P < 0.05; **P < 0.01, Student’s t test. Different letters indicate statistically significant differences at P = 0.05 by one-way ANOVA. Scale bars, 10 mm (H and L).

To validate the functional role of OsbZIP35, we assessed the agronomic traits of bZIP35-CR-15 and bZIP35-CR-27 mutants. Both mutants exhibited significantly increased grain length and 1000-grain weight compared to Nip, suggesting that OsbZIP35 acts as a negative regulator of rice grain size and weight (Fig. 4, H to K). Concurrently, panicle length, secondary branch number, grain number per panicle, and seed setting rate were significantly reduced in the mutants, while no significant changes were observed in tiller number or primary branch number (fig. S15, B to J). To elucidate the genetic relationship between OsbZIP35 and COR1, we generated a cor1 bzip35 double mutant (bZIP35-CR; COR1-CR). Phenotypic analysis revealed that the grain size, grain weight, panicle length, secondary branch number, grain number per panicle, and seed setting rate of the double mutant were comparable to those of the cor1 single mutant (Fig. 4, L to O, and fig. S16). These findings indicate that OsbZIP35 and COR1 regulate rice grain length through a shared genetic pathway, with COR1 functioning downstream of OsbZIP35. Collectively, these results demonstrate that OsbZIP35 binds to the COR1 promoter to repress its transcription, thereby modulating rice grain development.

COR1 targets OsTCP19 and promotes its degradation

To elucidate the mechanism through which COR1 regulates rice grain size, we identified the interacting proteins by fusing the coding sequences (CDSs) of COR1 from Nip and SLG to the GAL4 DNA-binding (BD-COR1^Nip and BD-COR1^SLG) and used them as bait in yeast two-hybrid (Y2H) screening preys comprising rice yield-related genes (generated by our laboratory). We found that the yeast cells transformed with a plasmid containing the GAL4 activation domain (AD)–fused OsTCP19 (AD-TCP19) and BD-COR1^Nip or BD-COR1^SLG grew well on quadruple dropout media, whereas cells transformed with the corresponding control plasmids could not grow (Fig. 5A). To verify whether COR1 functions as an F-box protein, we constructed AD-OSK1 and performed Y2H (fig. S17). The results demonstrated that both COR1^Nip and COR1^SLG interacted with the rice OSK1 protein, suggesting the formation of the SCF^COR1 complex. Furthermore, we performed split firefly luciferase (LUC) complementation assays to verify the interaction between COR1^Nip or COR1^SLG and OsTCP19 by detecting the activated LUC activity through the luminescence signal when the fusion constructs nLUC-COR1^Nip or nLUC-COR1^SLG and cLUC-TCP19 were present together (Fig. 5B). Subsequently, we performed the coimmunoprecipitation (Co-IP) experiments to detect the in vivo interaction between COR1^Nip-MYC or COR1^SLG-MYC and TCP19-GFP in tobacco epidermal cells. Upon immunoprecipitation (IP) using an antibody to GFP, we detected bands for COR1^Nip-MYC and COR1^SLG-MYC in the TCP19-GFP samples but not in the GFP samples (Fig. 5C). These results indicate that both COR1^SLG and COR1^Nip can physically interact with OsTCP19 in vitro and in vivo.

Fig. 5. COR1 targets OsTCP19 and promotes its degradation.

(A) COR1^Nip and COR1^SLG interact with OsTCP19 in yeast cells. pGBKT7 (BD) and pGADT7 (AD) are the bait and prey vectors, respectively. (B) COR1^Nip and COR1^SLG interact with OsTCP19 in the epidermal cells of tobacco leaves (left panel). The relative LUC activities of each combination in the left panel (right panel). Data represent means ± SD (n = 3). (C) Co-IP assay showing COR1^Nip or COR1^SLG interact with OsTCP19. (D) Ubiquitination analyses of OsTCP19. (E) OsTCP19 protein level analysis. Total proteins were extracted from the protoplasts of Nip, COR1^Nip-OE, or COR1^SLG-OE transiently expressing OsTCP19. (F) Analysis of the stability of OsTCP19 proteins in the cell-free assays. Purified OsTCP19-GST proteins were incubated with total proteins from the young panicle of Nip, COR1-CR, and COR1-OE plants. MG132, a proteasome inhibitor. (G to J) Mature grains (G) and statistical results for grain length (H), grain width (I), and 1000-grain weight (J) of Nip and OsTCP19-CR plants. Data represent means ± SD (n = 15). (K to N) Mature grains (K) and statistical results for grain length (L), grain width (M), and 1000-grain weight (N) of Nip, COR1-CR, OsTCP19-CR, and COR1-CR; OsTCP19-CR plants. IB, immunoblotting; IP, immunoprecipitation. Data represent means ± SD (n = 15). Different letters indicate statistically significant differences at P = 0.05 by one-way ANOVA. Scale bars, 10 mm (G and K).

We also observed that COR1^SLG showed stronger interaction with OsTCP19 than COR1^Nip both in yeast cells and tobacco epidermal cells (Fig. 5, A to C). Then, we used AlphaFold 3 for protein complex modeling to investigate the structural interactions between COR1^Nip or COR1^SLG and OsTCP19. The modeling results revealed distinct interaction patterns between COR1^Nip or COR1^SLG and OsTCP19. The COR1^SLG variant demonstrated more extensive interactions with OsTCP19, forming 13 predicted contact pairs, while COR1^Nip showed four interaction pairs (fig. S18 and table S1). Two nonsynonymous mutations exist between the alleles COR1^Nip and COR1^SLG, leading to amino acid substitutions at positions 60 (G60S) and 265 (E265Q) (fig. S19). To investigate whether these substitutions, individually or in combination, affect protein structure and protein interaction strength, we used FoldX 5.1 to calculate the binding free energies (ΔG_bind) for different COR1-OsTCP19 complexes (fig. S20A and table S2). The COR1^G60S-OsTCP19 complex showed a significant increase in binding affinity compared with COR1^Nip-OsTCP19, with ΔG_bind = −22.856 kcal mol⁻¹ (ΔΔG versus COR1^Nip = −11.153 kcal mol⁻¹). In contrast, the COR1^E265Q-OsTCP19 complex exhibited destabilization, as evidenced by ΔG_bind = −6.053 kcal mol⁻¹ (ΔΔG versus COR1^Nip = +5.651 kcal mol⁻¹). Notably, the COR1^SLG-OsTCP19 complex achieved ΔG_bind = −31.49 kcal mol⁻¹ (ΔΔG versus COR1^Nip = −19.790 kcal mol⁻¹), demonstrating the strongest binding affinity. Furthermore, the coupling energy of COR1^SLG (ΔΔG_int = −14.288 kcal mol⁻¹) was strongly negative, indicating negative epistasis: The two mutations synergistically stabilized the protein interface to a much greater extent than expected from their individual effects.

Next, α-galactosidase (α-Gal) activity assays showed that the interaction between COR1^SLG and OsTCP19 was the strongest, followed by COR1^G60S-OsTCP19, which exhibited a stronger interaction than COR1^Nip-OsTCP19. In contrast, COR1^E265Q showed the weakest interaction with OsTCP19 (fig. S20B). These findings were consistent with the binding free energy changes predicted by FoldX 5.1 analysis. Collectively, these results demonstrate that both amino acid substitutions significantly affect protein interaction strength and structural stability, with the nonsynonymous G60S mutation potentially playing a critical role in functional alteration.

Given that COR1 encodes an F-box protein, it is possible that it functions as an E3 ubiquitin ligase involved in the 26S proteasome–dependent degradation pathway of OsTCP19. Therefore, we first detect the accumulation of TCP19-GFP protein in COR1-CR and wild type. The results showed that, compared with Nip, the accumulation of TCP19-GFP proteins was notably increased in COR1-CR (Fig. 5D). In contrast, the ubiquitination level of OsTCP19 weakened in COR1-CR compared to that in Nip (Fig. 5D). Moreover, we found that the accumulation of TCP19-GFP protein was the highest in Nip, the second in COR1^Nip-MYC plants, and the least in COR1^SLG-MYC (Fig. 5E). These results suggest that COR1-mediated ubiquitination might enhance the degradation of OsTCP19. To further explore the degradation of OsTCP19 by COR1, total proteins extracted from 5-cm young panicles of Nip, COR1-CR, COR1^Nip-OE, and COR1^SLG-OE plants were incubated with TCP19-GST proteins, and the protein content of TCP19-GST was detected every 2 hours. It was found that the degradation rate of TCP19-GST protein was slower when incubated with COR1-CR proteins than when incubated with Nip, while it was faster when incubated with COR1^Nip-OE and was fastest with COR1^SLG-OE proteins (Fig. 5F). Moreover, the degradation of TCP19-GST was notably inhibited when treated with MG132 (Fig. 5F), suggesting its degradation via the 26S proteasome pathway.

Next, we obtained two independent mutants, namely, TCP19-CR-10 and TCP19-CR-11, in the Nip background through the CRISPR-Cas9 system (fig. S21A). The grain length, grain width, and 1000-grain weight, as well as tiller number, of TCP19-CR-10 and TCP19-CR-11 were significantly higher than those of Nip, while the secondary branch number, grain number, and seed setting were significantly reduced, with no obvious change in primary branch number (Fig. 5, G to J; and fig. S21, B to J), suggesting that OsTCP19 negatively regulates grain length and grain weight. To dissect the genetic relationship between COR1 and OsTCP19, we used CRISPR-Cas9 technology to generate a cor1 tcp19 double mutant (COR1-CR; TCP19-CR) in the Nip background (fig. S22A). The grain length, grain width, and 1000-grain weight, as well as tiller number, panicle branch number, and grain number per panicle, of the cor1 tcp19 mutant were similar to those of tcp19 single mutant (Fig. 5, K to N; and fig. S22, B to J), implying that COR1 and OsTCP19 collaboratively regulate grain length and weight and that COR1 acts upstream of OsTCP19.

Isolation of OsbZIP35-COR1-OsTCP19 downstream genes

To identify the target downstream genes of the COR1-OsTCP19 module, we compared the RNA sequencing profiles of COR1-CR and TCP19-CR young panicles with those of Nip. The results showed that, compared with Nip, there were 2602 up-regulated differentially expressed genes (DEGs) and 2853 down-regulated DEGs in COR1-CR, while there were 397 up-regulated DEGs and 757 down-regulated DEGs in TCP19-CR (Fig. 6A and data S1). Considering that COR1 promotes the degradation of TCP19, we screened for overlapping genes between the up-regulated DEGs of COR1-CR versus Nip and the down-regulated DEGs of TCP19-CR versus Nip, as well as the down-regulated DEGs of COR1-CR versus Nip and the up-regulated DEGs of TCP19-CR versus Nip, and identified 90 DEGs in total (Fig. 6B). Subsequently, we compared the sequences of the overlapping genes with the Eukaryotic Orthologous Groups of Proteins (KOG) functional database. Based on the annotation information of the matched genes, we annotated the genes in the dataset and performed statistical analysis and visualization of the KOG classification. Notably, three genes were enriched in functional categories related to cell cycle control, cell division, and chromosome partitioning (Fig. 6C). Among them, CycB1;4 encodes the B-type cyclins regulating the cell cycle in the G₂-M phase and was selected for further analysis. We first conducted qRT-PCR experiments and found that the expression level of CycB1;4 was significantly lower in COR1-CR plants but was significantly higher in COR1-OE, bZIP35-CR, and TCP19-CR plants than that of Nip, which was consistent with the results of transcriptome sequencing (Fig. 6D). Next, we performed the transient expression experiment in rice protoplasts and found that OsTCP19 suppressed the promoter activities of CycB1;4. However, when we coexpressed the COR1^Nip or COR1^SLG effector with OsTCP19, the promoter activities of CycB1;4 were significantly up-regulated, and the up-regulation amplitude of the COR1^SLG effector was significantly higher than that of the COR1^Nip effector (Fig. 6E). The results suggest that the COR1-OsTCP19 module may regulate the transcription of CycB1;4 in rice.

Fig. 6. Isolation of OsZIP35-COR1-OsTCP19 downstream genes.

(A) Volcano plots of total gene expression profiles in COR1-CR or TCP19-CR compared to Nip. FC, fold change. (B) The Venn diagram indicates the gene overlap of the up-regulation of COR1-CR and the down-regulation of TCP19-CR or the down-regulation of COR1-CR and the up-regulation of TCP19-CR. (C) KOG enrichment analysis clarifies the principal variations in the expression of overlapping genes. (D) The relative expression level of CycB1;4 in young panicles of Nip, COR1-CR, COR1^Nip-OE, COR1^SLG-OE, bzip35-CR, and TCP19-CR plants. Data represent means ± SD (n = 3). (E) The dual-LUC assays showing the effect of TCP and COR1 on the promoter activity of ProCycB1;4. Data represent means ± SD (n = 3). (F and G) Percentage of cell numbers during the G₁, S, or G₂-M phase. Data represent means ± SD (n = 3). *P < 0.05; **P < 0.01, Student’s t test. Different letters indicate statistically significant differences at P = 0.05 by one-way ANOVA.

Furthermore, we performed flow cytometry analysis on young panicles of Nip, bZIP35-CR, and TCP19-CR. The results indicated that, compared with Nip, the cell proportion in the G₁ phase of bZIP35-CR and TCP19-CR plants was lower, while the cell proportions in the S phase and G₂-M phase were higher, suggesting that cell division was activated in bZIP35-CR and TCP19-CR plants (Fig. 6, F and G, and figs. S23 and S24). Meanwhile, we also conducted flow cytometry analysis on cor1 bzip35 (COR1-CR; bZIP35-CR) and cor1 tcp19 (COR1-CR; TCP19-CR) double mutants and found that the cell proportion in the G₁ phase of cor1 bzip35 plants was higher, while the cell proportions in the S phase and the G₂-M phase were lower than Nip and bZIP35-CR plants, showing a similar pattern as COR1-CR plants (Figs. 3, H and I, and 6F and fig. S23). The cell proportion in the G₁ phase of cor1 tcp19 plants was lower, while the cell proportions in the S phase and the G₂-M phase were higher than Nip, showing a similar pattern as TCP19-CR plants (Fig. 6G and fig. S24). To investigate the role of OsbZIP35 and OsTCP19 in cell division regulation, we analyzed cell number and size in mature grain central glumes via scanning electron microscopy (figs. S25 and S26). Results showed that, in both longitudinal and transverse sections, cell numbers in OsbZIP35-CR, OsTCP19-CR, and cor1 tcp19 double mutants were significantly higher than in Nip, whereas the cor1 bzip35 double mutant had fewer cells than Nip. Statistical analysis of total cell numbers and areas within glumes confirmed these trends: OsbZIP35-CR, OsTCP19-CR, and cor1 tcp19 mutants had increased cell numbers, while cor1 bzip35 mutants showed reduced cell numbers. Notably, no genotype-dependent differences in cell area were observed, consistent with flow cytometry results. Together, these results imply that the OsbZIP35-COR1-OsTCP19 module may affect the expression of CycB1;4, thus regulating cell division and grain length.

Natural variation, origin, and evolution of COR1

To investigate the natural variation in germplasm, we analyzed the sequences of COR1 in 244 wild and 676 cultivated rice accessions and identified five haplotypes (Hap1 to Hap5) based on the nonsynonymous SNPs in the coding region, including S35239149 and S35239865 variations that were different between Nip and SLG (Fig. 7A and data S2). Among them, Hap1 and Hap2 are mainly present in the indica subpopulation; Hap3 and Hap4 are mainly present in the japonica subpopulation, while Hap5 is present only in the wild rice. Nip and SLG belong to Hap4 and Hap3, respectively (Fig. 7A and data S2). The accessions harboring Hap1 or Hap3 exhibited significantly longer grain than the accessions harboring Hap2 in the indica subpopulation, and the accessions harboring Hap3 showed slightly longer grain than the accessions harboring Hap4, implying that Hap1 and Hap3 were the superior haplotypes in indica and japonica, respectively (Fig. 7, B to D). The phylogenetic analysis revealed that Hap1 to Hap4 were present in both wild and cultivated rice, implying that they might directly derive from wild rice, while Hap3 only present in cultivated rice (fig. S27).

Fig. 7. Variation, origin, and evolution of COR1 in germplasm.

(A) Haplotypes of COR1 identified in 244 wild and 676 cultivated rice accessions. Wild rice: ruf1, ruf2, niv1, and niv2; cultivated rice: aromatic (aro), aus, indica, temperate japonica (tej), and tropical japonica (trj). (B) Statistical results for grain length of indica accessions containing Hap1, Hap2, or Hap3. Data represent means ± SD (n = 365, 20, and 21). (C and D) Statistical results for grain length of temperate japonica (C) and tropical japonica (D) accessions containing Hap3 or Hap4. Data represent means ± SD (n = 65 and 80; n = 10 and 13). (E) Haplotype network analysis of COR1. (F) The schematic diagram of predicted domestication paths of COR1. (G and H) Geographic distribution of COR1 haplotypes across Asian rice populations (65°E to 142°E, −10°N to 55°N). Distribution of Hap3, Hap4, and Hap5 (G) and Distribution of Hap1, Hap2, and Hap5 (H). Individual dots represent single accessions; pie charts indicate multiple accessions at similar coordinates, with size proportional to accession numbers and segments showing subspecies composition. (I) The frequencies of COR1-Hap3 and COR1-Hap4 in landraces (LAN) and improved varieties (IMP). (J to M) Mature grains (J) and statistical results for grain length (K), grain width (L), and 1000-grain weight (M) of Yuefu and IL200 plants. Data represent means ± SD (n = 9). *P < 0.05; **P < 0.01, Student’s t test. Different letters indicate statistically significant differences at P = 0.05 by one-way ANOVA. Scale bar, 10 mm (J).

To explore the origin and evolution of COR1, we identified 16 haplotypes (H1 to H16) based on 44 SNPs in the promoter and gene body region (data S2 and table S3). Furthermore, we conducted haplotype network analysis and found that there were two separate domestication paths of COR1, namely, Hap5-Hap2-Hap1 in indica and Hap5-Hap4-Hap3 in japonica, of which two-step standing variation occurred following the mutation from S35240004 to S35239087 and from S35239042 to S35239149 and S35239865, respectively (Fig. 7, E and F). Geographic distribution of COR1 haplotypes demonstrated that from this Hap5, two major evolutionary pathways emerged: Hap2 and Hap1 showing higher frequencies in the southern regions, particularly associated with indica rice populations (Fig. 7G); the Hap5-Hap4-Hap3 path shows a northward expansion pattern. Hap4 appears as a transitional form predominantly distributed in central and eastern China, while Hap3 shows higher frequency in the northern regions, particularly in japonica rice populations (Fig. 7H). Next, we analyzed the nucleotide diversity and Tajima’s D value of COR1 and its 100-kb flanking region and found that the nucleotide diversity of COR1 was much lower in indica and tropical japonica than in wild rice; the Tajima’s D value of COR1 was significantly negative in indica and tropical japonica (table S4), indicating that COR1 has undergone directional selection in both indica and tropical japonica.

To explore the utilization of COR1 in japonica breeding, we analyzed the frequency of COR1-Hap3 and COR1-Hap4 in landraces (LAN) and improved varieties (IMP) of temperate japonica and tropical japonica. The results showed that the frequency of COR1-Hap3 increased in IMP than that in LAN in temperate japonica, while it was not obviously selected (Fig. 7I), implying considerable potential for breeding. Next, we isolated an introgression line, IL200, that was constructed by backcrossing Yuefu (the recipient parent belongs to COR1-Hap3) with IRAT109 (the donor parent belongs to COR1-Hap4) (fig. S28). Phenotypic analysis showed that, compared to Yuefu, IL200 exhibited significantly reduced grain length, grain width, and 1000-grain weight. In contrast, no significant differences were observed in the number of primary branches, filled grains number per panicle, spikelet number per panicle, or seed setting (Fig. 7, J to L, and fig. S29). These results imply that COR1-Hap3 is an elite allele for improving rice grain length and weight.

Potential for utilization in breeding

To evaluate the potential of COR1 in breeding, we investigated the other agronomic traits of NIL-COR1^SLG and NIL-COR1^Nip. There were no significant differences between them in terms of plant height, tiller number, panicle length, and seed setting. However, the secondary branch number and grain number per panicle of NIL-COR1^SLG were significantly lower than those of NIL-COR1^Nip, while the primary branch number and 1000-grain weight of NIL-COR1^SLG were significantly higher than those of NIL-COR1^Nip, resulting in grain yield per plant of NIL-COR1^SLG increased by 7.6% compared to that of NIL-COR1^Nip (Figs. 8, A to J, and 1J). Furthermore, we performed the field trials and found that the plot yield of NIL-COR1^SLG increased by 6.6% compared to that of NIL-COR1^Nip (Fig. 8, K and L). Moreover, the chalky grain rate and chalkiness degree of NIL-COR1^SLG were similar to those of NIL-COR1 ^Nip (Fig. 8M and fig. S30). These results indicate that COR1^SLG could be used to improve rice yield without reducing grain quality.

Fig. 8. Potential for utilization in breeding.

(A and B) The plant (A) and panicle (B) architecture of NIL-COR1^Nip and NIL-COR1^SLG. (C to J) Statistical results for plant height (C), tiller number (D), panicle length (E), primary branch number (F), secondary branch number (G), grain number per panicle (H), seed setting (I), and grain yield per plant (J) of NIL-COR1^Nip and NIL-COR1^SLG. Data in (C) to (F), (H), and (I) (n = 30) and (G) and (J) (n = 15) represent means ± SD. (K and L) The field plot trials of NIL-COR1^Nip and NIL-COR1^SLG. The field plot covers an area of 2 m². Data represent means ± SD (n = 3). (M) The milled rice of NIL-COR1^Nip and NIL-COR1^SLG. (N and O) Grain and panicle morphology of LJ31 and COR1^SLG-OE plants. n.s., not significant; *P < 0.05; **P < 0.01, Student’s t test. Scale bars, 10 mm (N), 20 mm (B, K, and O), and 20 cm (A).

To further investigate the potential application of COR1 in molecular breeding, we selected Longjing 31 (LJ31), a commercially important rice variety widely cultivated in the northern cold-region rice area, as the recipient. We introduced an overexpression vector containing the coding region of COR1 from SLG into LJ31. Phenotypic analysis revealed that the overexpression plants exhibited a significant increase in grain length, 1000-grain weight, and grain yield per plant compared to the wild-type LJ31 (Fig. 8, N and O; and fig. S21, E to G and N). These findings demonstrate that the COR1 allele holds potential for application in rice breeding programs aimed at enhancing grain length and yield across diverse cultivars.

DISCUSSION

Grain size and grain weight are critical agronomic traits that determine rice yield. In the present study, we identified a novel gene, COR1, which encodes an F-box protein and positively regulates rice grain length by promoting cell proliferation. Further investigations revealed that OsbZIP35 binds to the promoter of COR1 and modulates its expression. The COR1^SLG allele, which is associated with larger grains, results in enhanced degradation of OsTCP19 compared to the COR1^Nip allele. This degradation alleviates the transcriptional repression of CycB1:4, thereby stimulating cell proliferation (Fig. 9). Our findings elucidate the molecular mechanism underlying the regulation of grain size and grain weight by the OsbZIP35-COR1-TCP19 module.

Fig. 9. The working model.

A predicted model demonstrates that the OsbZIP35 TF binds to the COR1 promoter and acts as a transcriptional repressor. Notably, sequence variations in the coding regions of COR1^Nip and COR1^SLG alter their interaction and ubiquitination efficiency with OsTCP19, leading to differential transcriptional repression of CycB1;4. This regulatory divergence ultimately affects cell proliferation and contributes to grain length variation.

In numerous plant species, seed number and seed size are often negatively correlated in evolutionary processes, and a certain relationship exists between these two traits (23). In rice, a trade-off is observed between grain number per panicle and grain size. Although many genes related to grain number per panicle and grain weight have been cloned, and several clear regulatory pathways have been documented (3, 24), studies on the synergistic balance between these two traits remain limited. The GRAIN SIZE AND NUMBER1 (GSN1)–mitogen-activated protein kinase molecular module has been reported to fine-tune the balance between grain number and grain size in rice. Reduced expression of GSN1 leads to increased grain size but decreased grain number, while elevated expression results in increased grain number but smaller grains (25). LARGE2 encodes a HECT domain E3 ubiquitin ligase, OsUPL2, which governs panicle size and grain number per panicle in rice. The large2 mutant produces larger panicles with an increased grain number and broader grains (26). In the present study, we observed that COR1-CR plants, which have relatively smaller grains, exhibited a significantly higher grain number per panicle compared to the wild type (Fig. 2, A to C; and fig. S6, C, I, and J). Conversely, transgenic plants overexpressing COR1^Nip and COR1^SLG, which exhibited longer grains, had a significantly lower grain number per panicle than the wild type (Fig. 2, E to H; and fig. S7, D, K, and L). These results suggest that there may be a trade-off between the effects of COR1 on grain size and grain number. This finding provides insights into the molecular regulation mechanisms underlying inflorescence morphogenesis and its plasticity and may serve as a target for the coordinated improvement of yield and quality in rice.

Analysis of grain size in NILs demonstrated that NIL-COR1^SLG exhibited significantly greater grain length and 1000-grain weight compared to NIL-COR1^Nip, but a significant reduction in grain width was observed (Fig. 1, H to J). However, in the complementation experiment, introducing the COR1 coding region and promoter from SLG (ProCOR1^SLG:COR1^SLG) into NIL-COR1^Nip resulted in significant increases in grain length, 1000-grain weight, and grain width (Fig. 1, L to N). This outcome appears contradictory to the reduced grain width observed in NIL-COR1^SLG, as no reduction in grain width was evident in the complementation experiment. This discrepancy may be attributable to subtle genetic background differences. Despite the construction of NILs through multiple generations of backcrossing, residual SLG fragments may still exist, potentially leading to the incomplete exclusion of QTLs associated with grain width. In the COR1-CR mutant, the complete loss of COR1 function caused reductions in both longitudinal and transverse cell numbers (fig. S8), resulting in shorter grains with proportionally narrower widths. Conversely, in COR1-CO, moderate expression of COR1^SLG driven by its native promoter may have optimized the balance of cell proliferation in both longitudinal and transverse dimensions.

Upon identifying the candidate gene COR1 through fine mapping, DNA sequence analysis uncovered four SNPs within the coding regions of COR1 between the two parental lines, Nip and SLG. Notably, SNPs S35239149 and S35239865 led to amino acid alterations (Fig. 1F and fig. S3). Our findings demonstrated that both COR1^Nip-OE and COR1^SLG-OE transgenic plants exhibited significantly increased grain length compared to Nip. Despite similar COR1 expression levels, the grain length of COR1^Nip-OE-31 was markedly shorter than that of COR1^SLG-OE-10 (Fig. 2, E and F, and fig. S7E), suggesting that the S35239149 and S35239865 variations are critical for the functional divergence between COR1^Nip and COR1^SLG. To further investigate, we used AlphaFold 3 to model the protein complexes of COR1^Nip and COR1^SLG with OsTCP19, revealing 4 and 13 predicted contact pairs, respectively (fig. S18 and table S1). Notably, several key interaction sites were localized near the amino acid variations (60Gly-to-Ser and 265Glu-to-Gln) in the COR1^SLG-OsTCP19 complex, but not in the COR1^Nip-OsTCP19 complex (table S1). This indicates that the S35239149 and S35239865 variations play a pivotal role in grain length variation. The binding free energies (ΔG_bind) calculation revealed that the two mutations synergistically stabilize the protein interface to a much greater extent than would be expected on the basis of their individual effects alone. Additionally, the α-GAL activity assays demonstrated that the interaction strength between COR1^SLG and OsTCP19 was the highest, followed by COR1^G60S and OsTCP19, which exhibited a stronger interaction than COR1^Nip. In contrast, COR1^E265Q showed the weakest interaction with OsTCP19 (fig. S20). In conclusion, the nonsynonymous G60S mutation strengthens the interaction between COR1 and TCP19, serving as a critical determinant of their interaction affinity. Further studies are required to determine which specific site exerts a more notable influence on this phenotypic trait.

We identified 16 haplotypes (H1 to H16) based on 44 SNPs in the promoter and coding regions of COR1, which were further classified into five haplotypes (Hap1 to Hap5) according to nonsynonymous SNPs (Fig. 7A and table S3). Notably, accessions carrying Hap5 were exclusively found in Oryza nivara wild rice (niv2) from South and Southeast Asia. Accessions containing Hap1 and Hap2, primarily present in indica varieties, likely originated directly from Hap5, while those with Hap4 were predominantly found in japonica varieties as well as in Oryza rufipogon and O. nivara wild rice from South China and South/Southeast Asia (Fig. 7, A, E, G, H). The distribution of these haplotypes exhibits a strong correlation with rice subspecies distribution: Hap3 is predominantly associated with japonica varieties in temperate regions, whereas Hap1 and Hap2 show a stronger association with indica varieties in tropical and subtropical zones. This distribution pattern suggests that the evolution of COR1 haplotypes was likely shaped by both natural selection for environmental adaptation and artificial selection during domestication. While Hap3 facilitates grain elongation, its influence may be counterbalanced by other genes selected during domestication that constrain grain width. Consequently, this interaction leads to a modest increase in grain length while preserving the overall round grain size characteristic of japonica varieties. Hap3 is likely to enhance yield through improved grain filling rates (fig. S10) and enhanced cell proliferation, potentially outweighing its effect on grain elongation. This is consistent with the observed 6.6% yield increase in NIL-COR1^SLG (Fig. 8, L and M), indicating that Hap3’s agronomic value extends beyond grain morphology. Notably, OsTCP19 plays a important role in regulating cold tolerance, lodging resistance, and nitrogen response (27–29). These research findings suggest that COR1 may also have a certain function in the response to abiotic stress, which requires further in-depth research. Furthermore, the central position of Hap5 in the haplotype network implies its role as the ancestral haplotype, serving as the evolutionary foundation from which other haplotypes diversified during rice domestication and adaptation to diverse ecological niches. Additionally, it was unexpected that accessions carrying S35239087, a mutation causing premature translation termination and resulting in a truncated COR1 protein, exhibited larger grains in the indica subgroup (Fig. 7A). The functional impact of this truncated COR1 requires further validation to confirm whether it represents a gain-of-function mutation.

The SCF complex, a well-characterized multisubunit ring E3 ubiquitin ligase, plays a pivotal role in tagging target proteins with ubiquitin for subsequent degradation via the 26S proteasome system. F-box proteins, as core components of the SCF complex, are essential for ubiquitin-mediated proteolysis and serve as critical regulators in diverse physiological processes (30). In this study, we identified the OsbZIP35-COR1 module as a key regulator targeting OsTCP19 to modulate grain size through the regulation of cell proliferation (Fig. 9). KOG analysis of 90 DEGs further revealed enrichment in functional categories, such as signal transduction mechanisms, secondary metabolite biosynthesis, and transport and catabolism (Fig. 6C). Moreover, overexpression of COR1 in both Nip and LJ31 genetic backgrounds significantly enhanced grain size and grain weight. Field experiments demonstrated that NIL-COR1^SLG, carrying the COR1^SLG allele, exhibited a 6.6% higher yield compared to NIL-COR1^Nip (Fig. 8, L and M), primarily due to a significant increase in 1000-grain weight (Figs. 1J and 8J), with no adverse effects on grain quality (Fig. 8K and fig. S30). Furthermore, we evaluated the function of COR1 under different backgrounds (Figs. 7, I to M, and 8, N and O; and fig. S31), showing its trade-offs between grain number and size. In the future, the editing of its untranslated region [including the promoter, 5′ untranslated region (5′UTR) and 3′UTR] and combination editing of other grain yield related genes can be used in breeding practices. These results underscore the potential of the elite COR1 allele as a valuable genetic resource for breeding high-yielding and stable rice varieties.

MATERIALS AND METHODS

Plant materials and growth condition

SLG-1 (ssp. japonica, SLG) and Nipponbare (ssp. japonica, Nip) were crossed to produce the F₁ generation, which was subsequently backcrossed with Nip to generate BC₄F₂ and BC₄F₃ populations for primary QTL mapping (31). The plants were cultivated under natural rice-growing conditions in either Beijing or Sanya, Hainan Province. The planting density was maintained at 23.3 cm between rows and 13.3 cm between plants. The plot yield experiment was carried out at the Shangzhuang Experimental Station of China Agricultural University in May 2024. NIL-COR1^Nip and NIL-COR1^SLG were sown separately to ensure as uniform a distribution as possible. One month after seedling cultivation, seedlings with consistent growth were carefully selected for transplantation. Each plot consisted of 21 rows × 5 plants per row, with a row spacing of 25 cm and a plant spacing of 15 cm, and three replicates were established. Two rows of protective plants were evenly planted around each plot to minimize experimental errors caused by edge effects.

Traits measurement

The grain length, width, and 1000-grain weight were measured by using the Wanshen SC-G automatic seed analysis system. The 1000-grain weight was calculated using fully filled grains. Measurements of panicle length, primary branch number, secondary branch number, grain number per panicle, yield per plant, seed setting rate, and plant height were carried out on the main panicle. The number of tillers was recorded at the maturity stage. The chalkiness and chalkiness rate were evaluated with the Wanshen SC-E Rice Appearance Quality Detector.

RNA extraction and qRT-PCR

The Aidlab RNA Extraction Kit was used for RNA isolation. Subsequently, the Aidlab Reverse Transcription Kit was used for reverse transcription. Real-time qPCR was performed using the ChamQ Universal SYBR qPCR Master Mix (Vazyme Company) on the Applied Biosystems QuantStudio 1 Plus device.

Plasmid construction and plant transformation

To generate a CRISPR-Cas9 vector, two 20-bp protospacer adjacent motif (PAM) sequences were selected from the CDS of COR1, OsbZIP35, and OsTCP19 for specific targeting. These sequences were then cloned into the sgRNA-Cas9 vector, following the protocol as described previously (32). For the complementation construct ProCOR1^SLG:COR1^SLG, a 1919-bp promoter fragment from SLG and the COR1 CDS were amplified from SLG and cloned into the Pme I and Sac I sites of the plant binary vector pMDC163. To construct the overexpression vectors, the COR1 coding region of Nip and SLG (excluding the stop codon) was amplified and separately cloned into the Hind III and Kpn I sites of the binary plant expression vector pSuper1300-MYC, generating the Pro35S:COR1^Nip-MYC and Pro35S:COR1^SLG-MYC vectors. For subcellular localization, the COR1 coding region (without stop codon) from Nip and SLG was amplified and cloned into the Kpn I and Hind III sites of the binary plant expression vector pSuper1300-GFP, resulting in the Pro35S:COR1^Nip-GFP and Pro35S:COR1^SLG-GFP vectors. Similarly, the CDS of OsbZIP35 was amplified and used to construct the Pro35S:OsbZIP35-GFP and the Pro35S:3×Flag-OsbZIP35 vectors. For the ProCOR1:GUS vector, the 1919-bp promoter region of COR1 was amplified and cloned into the Pme I and Sac I sites of the vector pMDC162. All DNA constructs in this study were amplified using Tks Gflex DNA polymerase (TAKARA). The correct vectors were introduced into the EHA105 strain of Agrobacterium tumefaciens, and rice callus tissue was transformed via Agrobacterium-mediated transformation (33).

GUS staining analysis

GUS staining was performed according to the previously established protocol (34). Briefly, young panicles at different developmental stages, along with roots, stems, and leaves harvested from ProCOR1:GUS transgenic plants, were incubated at 37°C in a staining solution containing 1 mM X-Gluc, 0.4 mM K₃Fe(CN)₆, 0.4 mM K₄Fe(CN)₆, 50 mM NaPO₄ (pH 7.0), and 0.1% (v/v) Triton X-100 for 8 hours. Following the staining procedure, chlorophyll was removed by washing the tissues with 70% ethanol.

Subcellular localization analysis

To investigate subcellular localization, COR1^Nip-GFP and COR1^SLG-GFP vectors were generated and transient expressed in rice protoplasts isolated from 10-day-old seedlings grown under dark conditions using the polyethylene glycol (PEG)–mediated transformation method. Following transformation, the protoplasts were incubated in darkness for 16 hours. Fluorescence signals were then visualized using an LSM 880 confocal laser scanning microscope (Zeiss, Oberkochen, Germany).

Flow cytometry analysis

For flow cytometry, ~10 young panicles (~2 cm in length) were finely chopped in 4 ml of cold nuclear isolation buffer containing 3 mM dithiothreitol (pH 8.0), 5 mM Hepes, 50 mM KCl, 10 mM MgSO₄, and 0.25% (v/v) Triton X-100. The resulting crude extracts, containing isolated nuclei, were filtered through a 48-μm nylon mesh to remove debris. The nuclei were then stained with propidium iodide (5 μg ml⁻¹; Sigma-Aldrich) and immediately analyzed using a CyFlow Ploidy Analyzer (Sysmex Partec, Germany).

Measurement of grain filling rate

During the heading stage of rice, spikelets with abundant florets were selected at 10:00 a.m. The glumes of blooming florets were marked using an oil-based pen, with ~30 glumes marked per spikelet. Each marked panicle was labeled with a small tag indicating the date of marking. A total of 40 panicles were marked for the experiment. Sampling was conducted at 3-day intervals postflowering and pollination (i.e., on the 3rd, 6th, 9th, 12th, 15th, 18th, 21st, 24th, 27th, and 30th days). At each sampling point, previously marked glumes were collected, with three replicates per time point and 30 seeds per replicate. The fresh weight of the seeds was measured. To ensure consistency, seeds from similar positions on the panicle were selected, and those exhibiting uneven growth were excluded. After fresh weight measurement, the seeds were transferred to a constant temperature oven at 42°C for thorough drying. Once dried, the dry weight of the seeds was recorded.

Scanning electron microscopy analysis

The spikelet hulls of mature grains were examined using a scanning electron microscope (S-3000N & Quorum PP3000T, Hitachi). Following imaging, the cell area was quantified using ImageJ software. Additionally, the number of cells in both the longitudinal and transverse directions was statistically analyzed.

Transient transcriptional activity assays

The 1919-bp promoter region of COR1 was cloned from the genomic DNA of Nip and SLG and fused into the pGreenII0800-LUC vector as a reporter construct. The OsbZIP35-GFP fusion protein was used as the effector, with the empty vector serving as the negative control. Similarly, the 1887-bp promoter region of CycB1;4 was cloned from the genomic DNA of Nip and inserted into the pGreenII0800-LUC vector as a reporter construct. The COR1^Nip-GFP, COR1^SLG-GFP, and OsTCP19-GFP fusion proteins were used as effectors, while the empty vector was used as the negative control. For each cotransfection experiment, 2 μg of the reporter plasmid, 2 μg of the effector plasmid, and 2 μg of the empty vector were cotransfected into 100 μl of rice protoplasts using the PEG method. After incubation in the dark at 28°C for 16 hours, the relative LUC activity was measured using the Dual-LUC Reporter Assay System.

Yeast one-hybrid assays

Six distinct promoter regions containing the G-box motif were amplified and cloned into the Eco RI and Xho I sites of the pLacZi2μ vector to construct reporter plasmids. The CDS of OsbZIP35 was cloned into the Eco RI and Xho I sites of the pB42AD vector to generate the effector plasmid. The pB42AD-OsbZIP35 plasmid and the reporter constructs were cotransformed into the yeast EGY48 strain. Transformants were cultured on SD/-Trp-Ura dropout plates supplemented with X-β-gal to assess protein-DNA interactions.

ChIP-qPCR assays

In accordance with the method described by Weng et al. (35), ChIP experiments were carried out using 10-cm young panicles of Pro35S:3×Flag-OsbZIP35 transgenic plants. Flag antibodies were used for detection.

Electrophoretic mobility shift assays

The Chemiluminescent EMSA Kit (Beyotime Biotechnology) was used for EMSA. The oligonucleotide probes were synthesized by DeOiping and labeled with biotin. The CDS of OsbZIP35 was integrated into the pGEX-4 T-1 vector, Escherichia coli BL21 was transformed, and the GST-OsbZIP35 protein was purified. EMSA was conducted in accordance with the manufacturer’s instructions (Beyotime, GS009).

Y2H assays

The CDS of COR1 was amplified from the genomic DNA of Nip and SLG, respectively, and cloned into the pGBKT7 vector to generate the fusion proteins BD-COR1^Nip and BD-COR1^SLG. These fusion proteins were used as baits to screen for interacting proteins from a laboratory-constructed cDNA library of yield-related genes using the Y2H system. The Y2H assay was performed according to the manufacturer’s instructions (Clontech) in the yeast strain AH109. To validate the interactions, the CDS of OsTCP19 was amplified from Nip DNA and cloned into the pGADT7 vector to create the AD-OsTCP19 fusion construct. The BD-COR1^Nip and BD-COR1^SLG constructs were then cotransformed with AD-OsTCP19 into yeast AH109. The transformed yeast cells were plated on SD/-Trp- Leu, as well as SD/-Trp- Leu-His-Ade, dropout plates to assess the interactions.

Quantitative determination of α-Gal

Resuspend the monoclonal yeast in 100 μl of sterile water and use a pipette to spot 10 μl on SD/-Trp-Leu and SD-Trp-Leu-His-Ade + X-α-gal. Incubate at 28°C for 2 to 3 days. Select positive clones and inoculate them into the corresponding selective media. Cultivate overnight at 28°C with shaking at 200 rpm until the optical density at 600 nm (OD₆₀₀) reaches 0.5 to 1.0, and record the OD₆₀₀ value of the culture at this time. Transfer 1 ml of the well-mixed culture to a 1.5-ml centrifuge tube and centrifuge at 12,000 rpm for 2 min at room temperature. Carefully collect the supernatant and transfer it to a new centrifuge tube. Take 16 μl of the supernatant, add 48 μl of reaction solution, and incubate at 30°C for 60 min. Terminate the reaction by adding 136 μl of stop solution. Measure the OD₄₁₀ value using a spectrophotometer. Calculate the α-Gal enzyme activity (milliunits/milliliter*cell) using the following formula: α-Gal activity (milliunits/milliliter*cell) = OD₄₁₀ × V_f × 1000/[(ε × b) × t × V_i × OD₆₀₀], V_f is the final reaction volume; ε × b is the molar absorptivity of PNPG at 410 nm multiplied by the path length (for a 200-μl reaction system, this value is 10.5 ml/μmol); t is the reaction time in minutes; and V_i is the volume of the supernatant added to the reaction mixture.

Co-IP assays

The plasmids Pro35S:COR1^Nip-MYC and Pro35S:COR1^SLG-MYC were individually cotransformed with Pro35S:OsTCP19-GFP plasmid into A. tumefaciens (strain EHA105). The transformed Agrobacterium suspensions were then coinfiltrated with the p19 silencing suppressor into Nicotiana benthamiana leaves. For immunoblot analysis, antibodies against GFP (Sigma-Aldrich, catalog no. M4439, diluted at 1:3000), MYC (Sigma-Aldrich, catalog no. SAB4200681, diluted at 1:3000), and actin (Sigma-Aldrich, catalog no. A3853, diluted at 1:3000) were used.

LUC complementation image assays

The CDS of COR1 from Nip and SLG was amplified from their respective genomic DNA and subsequently fused into the pCAMBIA1300-nLUC vector. Similarly, the CDS of OsTCP19 was cloned into the pCAMBIA1300-cLUC vector. The resulting constructs, COR1^Nip-nLUC and COR1^SLG-nLUC, were cotransformed with OsTCP19-cLUC into A. tumefaciens (strain EHA105) and then infiltrated into tobacco leaves. After 48 hours of incubation, the leaves were sprayed with 1 mM beetle luciferin solution (Promega, E1650), and the LUC signal was visualized using a cooled charge-coupled device imaging system (Vilber, Fusion FX7 EDGE). Relative LUC activity was quantified using the Dual-LUC Reporter Assay System.

Protein degradation assays

To assess the protein stability of OsTCP19 in the COR1-CR mutant, COR1^Nip-OE, and COR1^SLG-OE plants, protoplasts were isolated from seedlings grown in darkness for 10 days. The Pro35S:OsTCP19-GFP plasmid was introduced into 100 μl of protoplasts from each genotype using the PEG-mediated transformation method. Following 16 hours of incubation in darkness, total proteins were extracted using a lysis buffer containing 50 mM tris-HCl (pH 7.5), 150 mM NaCl, 10 mM NP-40, 3 mM dithiothreitol (DTT), and one tablet of protease inhibitor cocktail per 50 ml of buffer. For immunoblot analysis, antibodies against GFP (Sigma-Aldrich, catalog no. M4439, diluted at 1:3000), MYC (Sigma-Aldrich, catalog no. SAB4200681, diluted at 1:3000), and actin (Sigma-Aldrich, catalog no. A3853, diluted at 1:3000) were used.

To further investigate whether COR1 enhances ubiquitination of OsTCP19, total proteins were incubated with anti-GFP–conjugated beads (TransGen Biotech, HT801-02) at 4°C for 4 hours. Ubiquitination (Ub) levels were then detected by immunoblotting using an anti-Ub antibody (Abmart, catalog no. PL026378S; diluted at 1:3000).

Cell-free degradation assays

To investigate the differences in protein degradation rates of COR1 among the COR1-CR mutant, COR1^Nip-OE, and COR1^SLG-OE plants, young panicles (5 cm in length) from each genotype were rapidly frozen in liquid nitrogen and ground into a fine powder under low-temperature conditions. Total proteins were extracted using a lysis buffer containing 50 mM tris-HCl (pH 7.5), 150 mM NaCl, 10 mM NP-40, 3 mM DTT, and one tablet of protease inhibitor cocktail per 50 ml of buffer.

The CDS of OsTCP19 was cloned into the pGEX-4 T-1 vector and expressed in E. coli BL21. The GST-OsTCP19 fusion protein was subsequently purified. Equal amounts of total protein extracts from the young panicles were incubated with the purified GST-OsTCP19 protein at room temperature for specified durations. The resulting samples were then subjected to immunoblot analysis using anti-GST (Sigma-Aldrich, catalog no. G1160, dilution at 1:3000) and anti-actin (Sigma-Aldrich, catalog no. A3853; dilution at 1:3000) antibodies.

Transcriptome analysis

High-quality total RNA was extracted from 10 young panicles (each ~1 cm in length) of COR1-CR, TCP19-CR, and wild-type plants, respectively. RNA sequencing was performed using the HiSeq2000 platform provided by Biomarker Technologies (Beijing). The DEGs were identified and used for KOG analysis using the BMKCloud platform (www.biocloud.net).

Variation, origin, and evolution analysis

The SNPs used for haplotype identification, phylogenetic analysis, and haplotype network analysis were obtained from previously published studies (36, 37). Phylogenetic analysis, nucleotide diversity assessment, and selection analysis were conducted as previously (38). The haplotype network was constructed using PopART 1.7 software (39). Geographic distribution analysis was performed using Python 3.13.1, with the following packages: pandas (1.4.0), geopandas (1.0.1), matplotlib (3.10.0), shapely (2.0.6), and numpy (2.2.1). The base map for geographic visualization was obtained from the Resource and Environment Science and Data Center of the Chinese Academy of Sciences (www.resdc.cn/).

Structural modeling and processing

To investigate and elucidate the structural interactions between COR1 and OsTCP19, comprehensive protein complex modeling was performed using the AlphaFold 3 webserver (40). Key confidence metrics extracted from the JSON output files included the Predicted Template Modeling Score (pTM), Interface Predicted Template Modeling Score (ipTM), Chain-Pair ipTM, and Minimum Chain pTM. Interaction interfaces were analyzed to identify residue-residue contacts using PyMOL, and contact pairs for each variant were documented.

Computational mutagenesis and binding-energy calculations were conducted on four structural templates using FoldX version 5.1. Each PDB file was preprocessed with the RepairPDB command (ionStrength of 0.05 M, pH 7.0, and vdwDesign of 2) to correct side-chain conformations and optimize hydrogen-bond networks. The repaired models were then analyzed with the AnalyseComplex command (chains A and B) to estimate binding free energies (ΔG_bind). Relative changes in binding affinity (ΔΔG) were quantified by subtracting the ΔG_bind of COR1^Nip from that of each mutant. The interaction coupling energy (ΔΔG_int) for the double mutant was calculated according to the formula outlined in the notes of table S2. To structurally interpret these energetic effects, COR1^Nip and COR1^SLG were superimposed in PyMOL 2.x by aligning all Cα atoms; iterative atom rejection yielded 577 Cα pairs with a global RMSD of 23.87 Å. Local Cα-Cα distances between residues 60 and 265 in chain COR1 were measured using a custom Python 3.8 script that parses PDB files via Biopython 1.78 and computes Euclidean norms through NumPy 1.20. All command-line parameters and source code are provided in the Supplementary Materials.

Primers

The information of primers used in this study is listed in data S3.

Declaration of generative AI and AI-Assisted technologies in the writing process

During the preparation of this work, the authors used Deepseek V3 in order to improve the language, using the following prompt: “Please correct the grammar and spelling errors in this manuscript.” After using Deepseek, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Supplementary Materials

The PDF file includes:

Figs. S1 to S31

Tables S1 to S4

Legends for data S1 to S4

Other Supplementary Material for this manuscript includes the following:

Data S1 to S4