Wild rice GL12 synergistically improves grain length and salt tolerance in cultivated rice

Abstract

The abounding variations in wild rice provided potential reservoirs of beneficial genes for rice breeding. Maintaining stable and high yields under environmental stresses is a long-standing goal of rice breeding but is challenging due to internal trade-off mechanisms. Here, we report wild rice GL12^W improves grain length and salt tolerance in both indica and japonica genetic backgrounds. GL12^W alters cell length by regulating grain size related genes including GS2, and positively regulates the salt tolerance related genes, such as NAC5, NCED3, under salt stresses. We find that a G/T variation in GL12 promoter determined its binding to coactivator GIF1 and transcription factor WRKY53. GIF1 promotes GL12^W expression in young panicle and WRKY53 represses GL12^W expression under salt stresses. The G/T variation also contributes to the divergence of indica and japonica subspecies. Our results provide useful resources for modern rice breeding and shed insights for understanding yield and salt tolerance trade-off mechanism.

Introduction

The domestication of cultivated rice is considered as one of the most important events in human agricultural history¹. Currently cultivated rice (Oryza sativa) is a staple food that feeds more than half of the world’s population².To meet the growing demand for rice production, it is essential to cultivate high-yielding and environmentally resilient rice varieties. However, simultaneous improvement of yield and biotic/abiotic resistance remains a huge challenge for rice breeders because yield and stress tolerance are typically negatively correlated^2,3.

Grain size and shape are major determinants for grain yield and quality, and have been selected during early rice domestication⁴. Grain size and shape are quantitative traits regulated by a series of quantitative trait loci (QTLs). In the past ten years, many QTLs/genes regulating grain size have been identified through classical mutant analysis or map-based cloning. They are involved in several important genetic pathways, mediated by phytohormones, G-proteins, proteasomal degradation, protein kinases, and transcriptional factors^4,5,6,7,8. Thus far, several genes for grain size, such as GW2, GL3.1, GS5, GW8, GS2/GL2, GS3, GL7/GW7 and GS9, have been identified in Asian rice^{9,10,11,12,13,14,15}. GS3 was the characterized QTL that regulates grain length (GL), encoding a transmembrane protein. The natural variation in GS3 contributes to grain-length differences between indica and japonica varieties^10,16. GL3.1 is a serine/threonine phosphatase that belongs to the PPKL family. It regulates grain length by mediating cell cycle progression through affecting the phosphorylation status of cell cycle proteins, such as Cyclin-T1;3, thereby controlling grain yield^11,12,17. Transcription factor OsMADS1, protein phosphatase GL3-1/GL3.1/OsPPKL1 and transcriptional regulator GS9 have been reported to control grain length by regulating cell proliferation in the grain hull^{11,12,17,18,19}. GS2 encodes transcription factor OsGRF4, which regulates grain size mainly by promoting cell expansion and slightly promoting cell proliferation. OsGRF4 is regulated by microRNA396. The GS2 allele with mutations in the targeting site of OsmiR396 in OsGRF4 disrupts the repression effects of OsmiR396 on OsGRF4, leading to increased expression of OsGRF4, and resulting in large grains and high yield. OsGRF4 physically interacts with transcription coactivators OsGIF1/2/3^5,11,20,21. However, the genetic and molecular mechanisms by which these factors control grain size still remain to be further elucidated.

As the direct progenitor of modern rice, common wild rice (Oryza rufipogon Griff.) has an important role in cultivated rice breeding and been considered as valuable resources for improving rice yield and environmental resilience. Cultivated crops, such as wheat, maize, and rice, have larger seeds than their wild relatives²². Almost all cultivated rice varieties have larger seeds than rufipogon. Although the overall phenotypic characteristics of wild rice, especially yield-related traits, are inferior to those of cultivated rice, studies have show that genes in wild rice are potentially useful for yield-related trait improvement²³. However, so far there are no positive regulation alleles for yield reported from wild rice. Cultivated rice is also a salt-sensitive plant, wild rice harbors especially genes controlling adaptation to unfavorable environments which were lost during domestication. The exploration and utilization of salt tolerance genes in wild rice resources is the main strategy of rice salt tolerance breeding. However, the elite resistant traits often linked with inferior yield traits in wild rice. Therefore, the dissection of wild rice QTLs/genes that synergistically increase yield and salt tolerance would be crucial for rice breeding.

In our previous study, we identified a QTL-qGL12 from wild rice, which regulates grain length in cultivated rice²⁴. Here, we find that the overexpression of the wild rice GL12^W allele simultaneously increases grain length and salt tolerance. A G/T variation in the GL12 promoter region, interacting with transcription factor regulator OsGIF1 (Growth-regulating factor interacting factor 1) and transcription factor OsWRKY53, is involved in the grain length and salt tolerance regulating pathways. In addition, it is found that GL12^W functions in different rice genetic backgrounds including both indica and japonica. Our results provide resources for rice breeding and insights for synergistic mechanism of rice yield and salt tolerance.

Results

Identification and confirmation of the wild rice qGL12 candidate gene

To identify wild rice genes underlying the natural variations in grain length (GL) and 1000-grain weight (TGW), we constructed a set of chromosome segment substitution lines (CSSLs) with an indica variety ‘9311’ as the recurrent parent and the O.rufipogon as the donor parent. One major QTL for GL, qGL12, was identified across six environments, and mapped using molecular markers on chromosome (Chr.) 12²⁵. One CSSL line, CSSL41, which has a long GL phenotype and carries the QTL-qGL12, was selected for further fine mapping (Supplementary Fig. 1a). An F₂ population was constructed from a cross between CSSL41 and 9311 (Fig. 1a). Using bulked-segregant analysis (BSA), and F₃ and F₄ homozygous recombinant analysis, qGL12 was finally narrowed to a~16-kb interval between molecular markers WR12.4 and RM28586, which contains three predicted ORFs (Supplementary Fig. 1b, Fig. 1a, b). Among them, LOC_Os12g39640 encodes a MYB transcription factor and sequence analysis revealed 11 single nucleotide polymorphisms (SNPs) and six insertion-deletion (InDels) polymorphism in the 1.8-kb promoter region and four SNPs in the coding region causing amino acid change between wild rice and 9311 (Supplementary Fig. 2a). LOC_Os12g39650 encodes a tubulin and LOC_Os12g39660 encodes a calcium-transporting ATPase, and there were eight and 17 SNPs between wild rice and 9311 in these two alleles respectively (Supplementary Fig. 3). We designed a cleaved amplified polymorphic sequence (CAPS) marker associated with wild rice LOC_Os12g39640 allele (Supplementary Fig. 2b), and a near-isogenic line (NIL) of qGL12 was developed with marker-assisted selection (Fig. 1a, c). We tested the expression levels of these three genes in young panicles of the NIL and 9311, among them, expression levels of LOC_Os12g39640 and LOC_Os12g39650 were significantly higher in the NIL than in 9311 (Fig. 1d). Compare with 9311, the NIL showed significantly increased GL and TGW, with 8.1% and 12.9%, respectively (Fig. 1f, g). To identify the candidate gene, we overexpressed (OE) these three genes from wild rice in 9311 background individually, both of LOC_Os12g39640-OE and LOC_Os12g39650-OE showed significantly increased GL (5.1% and 2.2%, respectively) than 9311, and LOC_Os12g39640-OE have much more increased GL than that of LOC_Os12g39650-OE (Fig. 1e–g). Given LOC_Os12g39650 is a tubulin gene, the LOC_Os12g39640 gene is the most likely candidate gene. Moreover, we knocked out (KO) LOC_Os12g39640 in the NIL background using CRISPR/Cas9 technology. The GL and TGW of NIL-KO plants decreased significantly, and the overexpression of LOC_Os12g39640 from 9311 did not improve GL in 9311 genetic background²⁴ (Supplementary Fig. 4), which confirmed that LOC_Os12g39640 was the candidate gene and therefore termed GL12.

Fig. 1: Wild rice GL12^W increases the grain length and weight in different genetic background.

a The near isogenic line (NIL) of qGL12 and recurrent parent 9311. b Fine mapping of qGL12. c Schematic diagram showing 9311 and wild rice chromatin. d Expression analyses of three genes in the qGL12 interval. The 2–5 cm young panicles were used for RT-qPCR. The relative abundances in 9311 plants were set to be one. Values given are means ± SD (n = 3) Statistical analysis was performed by two-tailed Student’s t-test *p ≤ 0.05, **p ≤ 0.01, NS no significant difference. The p-values were 3.7 × 10^–4,0.028 and 0.086, respectively. e Grains of 9311, NIL, and overexpression lines of three genes in qGL12. Bar = 1 cm. f, g Statistical analysis of grain length (GL) and 1000-grain weight (TGW) of lines in (e). Data were given as means ± SD (n = 21/20/22/21/19, 21/20/22/21/19 biological replicates). One-tailed Student’s t-test was used to generate the P values. h Grains of GL12 overexpression lines in Zhonghua11(ZH11) genetic background. Bar = 1 cm. i, j Statistical analysis of grain length and 1000-grain weight of lines in (h). Data were given as means ± SD (n = 21/20/22/21/19, 21/20/22/21/19 biological replicates). One-tailed Student’s t-test was used to generate the P values. Source data are provided as a Source Data file.

In addition, we generated transgenic lines in japonica variety ‘Zhonghua11 (ZH11)’ background, which overexpressed GL12/LOC_Os12g39640 alleles from wild rice (GL12^W) and 9311 (GL12⁹³¹¹), respectively. The GL12^W-OE lines showed significantly increase in GL and TGW than wild type ZH11 while GL12⁹³¹¹-OE lines had no difference (Fig. 1f–j, Supplementary Fig. 5). We also overexpressed GL12^W in ‘Nipponbare (NIP)’ background and the OE lines showed significantly increase in GL and TGW than that of wild type (Supplementary Fig. 6d, e). These data demonstrated that GL12^W has stable positive effect on GL in different genetic backgrounds.

GL12 ^W improves grain length by reprograming genes expression and altering cell length

GL12 encodes a MYB transcription factor, which belongs to a family with more than 2,00 members in rice^26,27. From the NCBI database, we performed phylogenetic analysis and revealed that GL12 was clustered with MYB transcription factor gene family in rice and other species, closest with japonica homologous gene (Supplementary Fig. 7). The GL12^w::GFP fusion protein was localized in the nucleus in tobacco epidermal cells and rice protoplasts (Fig. 2a). These results indicated that the GL12 protein could function in the nucleus and was consistent with the role of GL12 as a transcription factor.

Fig. 2: GL12^W regulated grain size related genes and improves grain length by altering cell length.

a Subcellular localization of GL12 (35S::GL12-GFP) in tobacco leaf cells and rice protoplasts. The fusion construct 35S::GFP/RFP marker, and GL12 (35S::GL12-GFP) / RFP marker were co-transformed into rice protoplasts and tobacco leaf cells, respectively. Scale bar = 10 μm. b Heat map diagram of the expression levels for rice seed size related genes. Data were from transcriptome using 2–5 cm young panicles. The heat map was drawn according to FPKM values. Columns and rows in the heat map represent samples and genes, respectively. Color scale indicates fold changes in gene expression. c Transcription levels of the grain size related genes were confirmed through qRT-PCR analysis with three biological replicates in 9311 and NIL. OsActin was used as a control. Data show means ± SD. d–f Spikelet hulls before anthesis (d). The white dashed line indicates the sites of the cross-sections shown in (e). The images on the right panel showed the close-up views of the boxed region in (e), Scale bar = 100 μm. The red arrows indicate the lower epidermal cells. Comparisons of spikelet of cell number, and cell length between 9311, NIL and GL12^W transgenic plants were shown in (f), letters a–c in indicate the different means according to Duncan’s multiple range test (P < 0.05). g Quantitative real-time PCR analysis of cell cycle-related and expansion gene expression in 9311, NIL and GL12^W-OE lines. OsActin was used as a control. Data show means ± SD of three biological replicates. Statistical analysis was performed by two-tailed Student’s t-test (*p ≤ 0.05, **p ≤ 0.01, NS no significant difference). Scanning electron microscopic analysis of the lemma of NIP and GL12^W-OE lines (h), and statistical analysis of longitudinal cell number and cell length in the lemmas of spikelet hulls (i) n = 10 view of 10 areas in (i). Statistical analysis was performed by two-tailed Student’s t-test (***p < 0.001). Source data are provided as a Source Data file.

To identify the molecular mechanism and downstream genes regulated by GL12^W, we first performed RNA-sequencing analysis using young panicles from NIL and NIL-KO plants. As a result, 8079 differentially expressed genes (DEGs) were identified with the stringent criteria (|log 2 (FC)|> 1, and FDR < 0.05). Among these DEGs, 4615 were upregulated and 3464 were downregulated in the NIL compared with the NIL-KO lines. Gene Ontology identified that GL12^W mostly participated in biological processes, and KEGG pathway enrichment analysis revealed that GL12^W affected the expression of genes involved in metabolism, genetic information processing, environmental information processing, organismal systems, and cellular processes (Supplementary Fig. 8). In addition, we found that several grain shape related genes were downregulated in young panicles of the NIL-KO relative to the NIL (Fig. 2b), including previously reported GS2/GRF4 genes¹¹. We also confirmed the expression levels of grain shape related genes in the NIL and 9311, GS2, GRF3, and GRF6 had significant higher expression levels in the NIL than that of 9311 (Fig. 2c). These results indicated that GL12^W might regulate grain shape by affecting the expression of GRFs.

The rice spikelet hull, which is coordinately controlled by cell proliferation and cell expansion, was proposed to restrict the rice grain shape⁴. To further understand the function of GL12 in grain shape, we investigated the spikelet hulls of 9311, GL12^W-OE and NIL before fertilization. The results verified that the longer GL in the NIL and GL12^W-OE was the result of cell expansion in spikelet hull (Fig. 2d, e). Furthermore, we examined the cross-section of spikelet hulls, and showed that the outer parenchyma cell layer in GL12^W-OE and NIL plants were longer and had substantially longer spikelet perimeter than the corresponding layer in 9311 plants (Fig. 2f). Cell division and expansion in eukaryotic organisms are controlled by a highly conserved basic cell cycle machinery^28,29. To investigate the relationship of GL12 with genes regulating plant cell cycle and expansion, we analyzed the expression of eight related genes in the NIL and GL12^W-OE relative to 9311 by real-time PCR. The transcription levels of five genes, CYCD4, CYCD7, CAK1, CYCA2.1 and CYCB2.2, were greatly reduced in GL12^W-OE and NIL as compared to 9311 plants. In contrast, the expression of OsEXPB3, OsEXPB4 and OsEXPA10 were significantly elevated in the NIL and GL12^W-OE (Fig. 2g). Additionally, we checked the grain hulls of GL12^W-OE NIP lines and NIP wild type using a scanning electron microscope. There were significantly fewer longitudinal cells in spikelet hulls of GL12^W-OE than in NIP; however, the cells of spikelet hulls of GL12^W-OE were significantly longer than those of Nip (Fig. 2h, i). Thus, GL12^W functions putatively as a positive regulator upstream of cell expansion genes, and its overexpression may result in an increase in cell expansion in longitudinal direction. These findings suggested that GL12^W functions as a regulator of both cell proliferation and cell expansion.

The native promoter enhances GL12 ^W expression in young panicle and increases grain length

Compared with GL12^W, GL12⁹³¹¹ coding region has four mutant sites and has no function on GL improvement (Supplementary Fig. 2a). There were 17 SNPs/InDels in the promoters of GL12 between 9311 and wild rice (Supplementary Fig. 2a). We performed real-time PCR to detect the expression of GL12 in young panicles of 9311 and the NIL, and found that the transcription levels of GL12 in the NIL were significantly higher in all panicle development stages than those in 9311 (Fig. 3a). We then performed the transient expression assays in Nicotiana benthamiana leaves, and found that the promoter of GL12^W (Pro^W) had stronger activity than that of GL12⁹³¹¹ (Pro⁹³¹¹) (Fig. 3b, c). Furthermore, we generated transgenic plants in the NIP background that harbored the promoter and coding regions of GL12 from 9311 and NIL, respectively. We found that the GL of Pro^W:GL12^W transgenic lines were much longer than those of NIP and the Pro⁹³¹¹:GL12^W transgenic lines and NIP (4.4% and 3.5% increase than NIP and Pro⁹³¹¹:GL12^W, respectively) (Fig. 3d–f), demonstrating that GL12^W has a stronger effect on GL under its native promoter.

Fig. 3: Promoter activity assays for the GL12 alleles in 9311 and wild rice.

a Relative quantitative real-time PCR analysis of GL12 expression in young panicles of 2, 4, 6, 7, and 8 cm in 9311 and NIL. OsActin was used as a control. qRT-PCR analysis with three biological replicates. Data showed means ± SD. (The p-values were 2.9 × 10^–5, 3.5 × 10^–5, 2.7 × 10^–3,1.7 × 10^–3and 3.4 × 10^–6, respectively). b Transient expression assays in Nicotiana benthamiana leaves showing the promoter activities of GL12, and the quantification of the relative luminescence intensities were shown in (c) The quantification of the relative luminescence intensities (Data show means ± SD, n = 8). Statistical analysis was performed by one-tailed Student’s t-test. d Grains of the transgenic lines in Nipponbare background. GL12 alleles form the wild rice and 9311 were driven by both promoters. Bar = 1 cm. Statistical analysis of grain length (e) and grain weight (f) in transgenic plants (n = 30/31/31/31/29). Data were given as means ± SD. One-tailed Student’s t-test was used to generate the P values. Source data are provided as a Source Data file.

The G/T variation in GL12 promotes its binding by GIF1

The OsGIF1 (GRF-interacting factor) gene in rice, which encodes a transcription coactivator, has been reported to interact with the transcription factor GS2/OsGRF4 and positively promote grain size in rice^30,31,32. Firstly, we detected the physical interaction between GL12 and OsGIF1. The yeast two-hybrid assay (Y2H) and firefly luciferase complementation experiment showed that both GL12^W and GL12⁹³¹¹ can interact with OsGIF1 protein (Fig. 4a, b). To determine whether GL12 was regulated by GIF1, we investigated the expression levels of GL12 in the GIF1 overexpression and CRISPR/Cas9 knock out lines in ZH11 background. Compared to ZH11, the GL12 expression levels in GIF1-OE lines were significantly improved while no significantly difference was found in GIF1-KO lines (Fig. 4c). Furthermore, we generated four constructs with truncated promoters of Pro^W and Pro⁹³¹¹ of GL12 alleles from NIL and 9311, respectively (Fig. 4d). The yeast one-hybrid assay results demonstrated that GIF1 could only bind the promoter of Pro-2, and GIF1 bind the Pro^W-2 much stronger than Pro⁹³¹¹-2. (Fig. 4d; Supplementary Fig. 9). Unexpectedly, the truncation of Pro^W-2 only harbored one SNP between the wild rice and 9311, which generated a G/T substitution at the -1117th position (Fig. 4d). These results were further supported by the electrophoresis mobility shift assay (EMSA) that used biotin-labeled probes containing the G/T SNP in the Pro^W-2 (Fig. 4e). These results indicated that the G/T site contributed to the binding of GIF1 to the promoter of GL12 and that was a direct target regulated by GIF1. Moreover, we knocked out the OsGIF1 in NIL background by CRISPR/Cas9 technology, the grain length of GIF1-KO-NIL significantly decreased (Supplementary Fig. 10).

Fig. 4: The G/T variation in GL12 promotes its binding by GIF1.

a Yeast two-hybrid assay showed that GIF1 physically interacted with GL12^W and GL12⁹³¹¹. b LCI assays in N. benthamiana leaves showing the GIF1 and GL12 interaction. N. benthamiana leaves were transformed by injection of Agrobacterium GV3101 cells harboring GL12-nLUC and GIF1-cLUC plasmids. Luciferase complementation was observed for the GL12-nLUC and cLUC- GIF1 combination, the negative control was GL12-nLUC + NAC2-cLUC, NAC2 was a nuclear localization gene, whereas no obvious signal was observed for the negative controls. c The GL12 expression levels in ZH11 and GIF1 overexpression (GIF1-OE) and knockout (GIF1-KO) lines. OsActin was used as a control. qRT-PCR analysis with three biological replicates. Statistical analysis was performed by two-tailed Student’s t-test (*p ≤ 0.05, **p ≤ 0.01, NS no significant difference). Data showed means ± SD. d Schematic illustration of the four truncations of GL12 promoters. The MYB motif was highlighted in yellow and the G/T variation was marked in red. Yeast one-hybrid assays were shown. GIF1 only interacted with Pro^W-2 truncation. Other yeast one-hybrid results were shown in Supplementary Fig. 9. e EMSA assays showed that GIF1 bound to the Pro^W-2 truncation. EMSA assays were repeated three times. GST-GIF1 fusion protein was expressed in E. coli. The probes were labeled with biotin. Source data are provided as a Source Data file.

GL12 ^W improves salt tolerance in rice

Besides GL12^W could increase the grain length and grain weight, we found that GL12^W could also improve salt tolerance. After extensive salt treatment (10 days of 150 mM NaCl treatment, recovered for seven days), both the NIL and GL12^W-OE showed significant higher salt tolerance than that of 9311, and the NIL showed much stronger salt tolerance than GL12^W-OE lines. The survival rate of the NIL, GL12^W-OE lines, and 9311 were 80%, 60% and no more than 30%, respectively. And the fresh weight and dry weight of the NIL were significantly higher than that of 9311 and GL12^W-OE lines. Meanwhile the NIL-KO line showed hypersensitivity to salt stress (Fig. 5a, b). Moreover, the overexpression of GL12^W in the NIP genetic background also significantly improved salt tolerance, compared with NIP wild type plants (Supplementary Fig. 11). There was no significant difference in salt tolerance between GL12⁹³¹¹-OE lines and 9311 (Supplementary Fig. 12). We then investigated the GL12 expression pattern under salt treatment. Before salt treatments, GL12 had higher basic expression levels in leaves of 9311 than that of NIL; after salt treatment, the GL12 expression in 9311 was induced rapidly compared with that in the NIL (Fig. 5c). These results indicated that GL12^W and GL12⁹³¹¹ have different expression patterns under salt treatment. Moreover, we analyzed the expressions of salt tolerance-related genes in the NIL and 9311. Compared with 9311, the NCED3 and NAC5 in the NIL were significantly induced after salt treatments (Fig. 5d, e). We further determined the effects on salt tolerance of OsGIF1 and GS2, and found that the GIF1-OE lines showed higher salt tolerance while GIF1-KO lines were salt sensitive compared with wild type ZH11 (Supplementary Fig. 13). The GS2-NIL line, which had increased GS2 expression levels¹¹, also showed higher salt tolerance than its genetic background ZH11 (Supplementary Fig. 13). These results indicated that GL12^W also played an important role in regulating the salt stress response in rice.

Fig. 5: GL12^W improves salt tolerance in rice.

a Photos of 9311, NIL and GL12^W overexpression lines before and after salt treatments. 20-day-old seedlings were treated with 150 mM NaCl for10 days and recovered for seven days. b Statistical analysis of survival rates and fresh and dry weight of seedlings in (a). Five biological replicates were used to determine the survival rate. Data were means ± SD (n = 5). Statistical analysis was performed by two-tailed Student’s t-test (*p ≤ 0.05, **p ≤ 0.01, NS no significant difference). c GL12 expression analysis in 9311 and NIL. 20-day-old seedlings were treated with 150 mM NaCl for indicated time. qRT-PCR analysis with three biological replicates. Data were means ± SD. Expression analysis of rice salt tolerance related genes NAC5 (d) and NCED3 (e) in 9311 and NIL, 20-day-old seedlings were treated with 150 mM NaCl. OsActin was used as a control. qRT-PCR analysis with three biological replicates. Data showed means ± SD. Source data are provided as a Source Data file.

The G/T variation in GL12 promoter determines its repression by WRKY53

Compared with GL12⁹³¹¹, the GL12^W expression was higher in young panicles but lower in leaves under salt stress (Figs. 3a, 5c). We postulated that GL12^W might be regulated by another transcription factor. Because OsWRKY53 was reported to be a key regulator of salt tolerance in rice³³, we investigated whether GL12 was involved in the OsWRKY53 regulated salt tolerance pathway. The OsWRKY53-OE and CRISPR/Cas9 knockout lines in japonica variety ‘LongGeng 11’ (LG11) background were used to determine the GL12 expression levels. The GL12 expression was significantly increased in the WRKY53-KO lines and significantly decreased in the WRKY53-OE lines, compared with wild type LG11 (Fig. 6a), which was consistent with the WRKY53-KO lines showed stronger salt tolerance³³. To study the interaction between WRKY53 and GL12, we analyzed the promoter region of Pro-1 to Pro-4 of GL12^W and GL12⁹³¹. Consistent with GIF1, the Y1H experiments demonstrated that WRKY53 could only bind the truncated promoter Pro-2 (Fig. 6b; Supplementary Fig. 9), which harbored only one G/T variation between the wild rice and 9311 (Fig. 4d). Interestingly, Y1H and EMSA proved that the binding ability of Pro^W-2 with WRKY53 was stronger than that of Pro⁹³¹¹-2 (Fig. 6c). We further tested the capacity of WRKY53 protein on driving the expression of GL12 using dual-luciferase assays. The promoter of GL12 was fused to the luciferase reporter gene and co-introduced into rice protoplast with constructs driving the expression of WRKY53 (Fig. 6d). In the presence of WRKY53, the luciferase (LUC) activity driven by the promoter of GL12^W was largely reduced than that of GL12⁹³¹¹ (Fig. 6e, f). These results strongly suggested that WRKY53 regulated salt tolerance by repressing the GL12^W expression.

Fig. 6: The G/T variation in GL12 determines its repression by WRKY53.

a The GL12 expression levels in Longgeng11 (LG11) wild type, and WRKY53 overexpression and knockout lines. Young leaves from 20-day-old seedlings were used for RNA extraction and Real-time PCR analysis. OsActin was used as a control. qRT-PCR analysis with three biological replicates. Data showed means ± SD. Statistical analysis was performed by two-tailed Student’s t-test (*p ≤ 0.05, **p ≤ 0.01, NS no significant difference). b Yeast one-hybrid assays proved that the transcription factor WRKY53 only bound to the Pro-2. Other yeast one-hybrid assay results were shown in Supplementary Fig. 9. c EMAS assays showed that WRKY53 bound to Pro^W-2. EMSA assays were repeated three times. MBP-WRKY53 fusion protein was expressed in E. coli. The probes were labeled with biotin. d Schematic diagrams of the effectors and reporters used in the dual-LUC experiments. Transient expression assays in N. benthamiana protoplasts showing that WRKY53 could reduce the transcriptional activity of GL12^W (f), but not the GL12⁹³¹¹ (e). The activities of firefly luciferase (LUC) and renilla luciferase (REN) were determined 16 h post transformation. The relative luciferase activities in different samples were calculated by normalizing the LUC values against REN. LUC/REN analysis with five biological replicates. Statistical analysis was performed by two-tailed Student’s t-test (*p ≤ 0.05, **p ≤ 0.01; NS, not significant). Data show means ± SD. Source data are provided as a Source Data file.

Natural variations in GL12 contribute to the divergence of indica and japonica subspecies

To identify whether GL12 has been artificially selected during rice domestication, we further analyzed the genomic sequences of this gene in rice germplasm resources from the 3K Rice Genomes Project^34,35, the Rice Super Pangenome Information Resource Database (RiceSuperPIRdb)³⁶ and our own re-sequencing data. Firstly, we compared the nucleotide diversity (PI-value) of the 100-kb region flanking GL12 with our re-sequencing data (56 indica, 51 japonica, and 52 wild rice accessions). The results showed that the PI values of the GL12 locus in wild rice were significantly higher than those in indica cultivars, but no significant difference with those japonica cultivars. In addition, the PI ratio in indica to that in wild rice was significantly lower than in japonica to that in wild rice in the GL12 locus (Fig. 7a, b).

Fig. 7: Natural variations in GL12 contribute to the divergence of indica and japonica subspecies.

a, b The relative ratio of nucleotide diversity analysis in the complete chromosome 12 of cultivated and common wild rice. The GL12 locus and its flanking genomic regions were marked. Dotted line area indicates a selective sweep of GL12 locus in indica. c Distribution of grain length for the two haplotypes in indica rice accessions, n = 15/894 means the accessions number of HapG and HapT. The haplotypes and number of accessions from the 3K Rice Genomes Project dataset were shown. Data show means ± SD. Statistical analysis was performed by two-tailed Student’s t-test (*p ≤ 0.05, p-value was 0.0104). d Haplotype networks generated from the promoter sequence of GL12 in both cultivated rice and various groups of wild rice accessions. The circle size is proportional to the sample size. Grey spots on the lines indicate the mutational steps between two haplotypes. e A proposed working model for the role of GL12 in the regulation of rice grain length and salt tolerance. Wild rice GL12^W is a functional and GL12⁹³¹¹ is a nonfunctional allele. The G/T variation in GL12 promoter determined its binding with transcription coactivator GIF1 and transcription factor WRKY53. GIF1 promotes GL12^W expression levels in young panicles, and then regulates the downstream grain size related genes, such as GS2. WRKY53 represses GL12^W expression under salt stresses, and GL12^W also regulates rice salt response genes, such as NCED3 and NAC5. Therefore, the wild rice allele GL12^W synergistically improves both the rice grain length and salt tolerance through the regulation by GIF1 and WRKY53. The image was drawn by Figdraw. Source data are provided as a Source Data file.

Furthermore, with the data from the 3K Rice Genome Project and rice super pangenome data, we found that the G/T substitution (1117-bp upstream of the start codon) in GL12 was unbalanced between the two cultivated rice subspecies, the majority of indica varieties were haplotype-T at the G/T site whereas most of japonica varieties were haplotype-G (Supplementary Fig. 14). We further investigated the role of G/T variation in the promoter of GL12 on GL using the data from 3K Rice Genomes Project of the indica varieties. Among them, 15 and 894 indica varieties carried the G and T types, respectively. The indica varieties with the haplotype-G exhibited longer grains than those with the haplotype-T (Fig. 7c). The haplotype classification results showed that the 228 rice varieties could be divided into a total of 20 haplotypes (Fig. 7d). Interestingly, the rice varieties within the blue dashed line are all Hap-T at promoter -1117 position, while those within the red dashed line are Hap-G. Among them, most japonica are Hap-G, indica are Hap-T, and wild rice are distributed in both haplotypes. And the Aus rice tends towards the indica type. All these findings suggested the genetic differentiation of GL12 between japonica and indica subspecies, which exhibited significantly reduced nucleotide diversity in indica compared with wild rice, suggesting a strong artificial selection in the GL12 locus of indica.

We proposed a possible working model for the network and showed that GL12 played an essential role in grain length and salt tolerance, which was synergistically regulated by GIF1 and WRKY53 at the G/T variation site (Fig. 7e).

Discussion

Early attempts to identify positive yield-related genes proved unsuccessful because most of the agronomic traits, especially yield-related characters of wild rice were inferior to those of cultivated rice. In this study, we identified wild rice gene GL12^W that improved rice GL and salt tolerance, which is synergistically regulated by GIF1 and WRKY53. The wild rice CSSL population we development in our previous study has greatly facilitated the discovery of unknown useful genes in wild rice³⁷. We identified the GL12^W gene using CSSL population and found it improves both the GL and salt tolerance synergistically in cultivated rice.

Compared with GL12⁹³¹¹, GL12^W improved the GL, TGW and salt tolerance in both indica and japonica genetic backgrounds (Figs. 1, 5; Supplementary Figs. 6, 9). We demonstrated that GL12^W regulated global gene expression, including GS2, and improves the GL by altering cell length (Fig. 2). We found a G/T variation at 1,117-bp upstream of the start codon of GL12 playing an essential role in binding with OsGIF1 and OsWRKY53, which positively and negatively regulated the GL12 expression, respectively (Figs. 3, 6). Previous reports demonstrated that OsGIF1 interacted with GS2 to control grain-size predominantly by increasing cell expansion in the grain hull^11,20,30. But whether these two genes involving in rice salt tolerance remains unknown, and no reports on whether GIF1 interacted with DNA and regulates gene expression directly. Our results showed the GIF1 physically interacted with GL12 but no differences were found between GL12^W and GL12⁹³¹¹ (Fig. 4a, b). GIF1 interacted with GL12^W promoter and the G/T site determined its binding to GIF1 and increased the GL12^W expression levels in young panicle (Fig. 4c–e). We also found that GIF1 and GS2 positively regulated salt tolerance in rice (Supplementary Fig. 13). Our results provided insights into the GIF1-GL12^W-GS2 module how to regulate the grain length and salt tolerance. Nevertheless, further studies are needed to explore the detailed mechanism and regulatory network of this module.

The NIL of GL12^W showed longer GL and stronger salt tolerance than GL12^W-OE lines. In addition, our transgenic experiments demonstrated that GL12^W had a stronger effect on the GL under its native promoter (Fig. 3d–f). Interestingly, compared with GL12⁹³¹¹, GL12^W has higher expression levels in young panicle but lower expression levels under salt stresses, and the latter might be caused by the repression of WRKY53 (Fig. 6). The expression levels of GL12^W synergistically regulated by GIF1 and WRKY53, reach an optimal balance in NIL, that explains NIL had much longer grain length and higher salt tolerance than GL12^W-OE lines. We further demonstrated that the G/T variation contributed to the divergence of indica and japonica subspecies. GL12 might have been directionally selected in indica but not in japonica during rice domestication (Fig. 7d). We investigated the agronomic and yield-related traits of 9311, NIL and GL12^W-OE lines, GL12^W has improved plant height, panicle length, but significantly decreased the grain number per panicle, same results were found in GL12^W-OE lines in NIP background (Supplementary Fig. 15). Unfortunately, the GL12^W did not improve yield significantly under both normal and salt stress conditions. This indicated that GL12^W might have a negatively effect on the grain number per spike, which can explain why the wild rice GL12^W has not been selected from wild rice to indica varieties during the rice artificial selection breeding.

Improving the salt tolerance without reducing yield is one of the major goals for modern rice breeding. The wild rice GL12^W gene, regulated by GIF1 and WRKY53, provides the endogenous cues for wild rice gene utilization and further paves the way for future investigation of the detailed biochemical mechanisms that synergistically improve both the grain length and salt tolerance. In summary, the discovery of this gene not only helps understand the mechanism for improving grain length and environmental resilience simultaneously for rice improvement but also provides the useful genetic resources for rice breeding programs.

Methods

Plant materials and growth conditions

Rice (O. sativa L.) materials include two wild-type japonica rice varieties, ZH11 and NIP, and one indica rice 9311. The overexpression lines were derived from ZH11 and NIP, and the knockout mutants were from NIL. All rice plants were cultivated in a natural paddy condition at Beijing and Sanya in Hainan province, China. Standard cultivation practices were used, and pests and diseases were managed in rice cultivation. Mature and dry seeds were used for measuring grain length, width, and TGW by image analysis method using SC-E software (Hangzhou Wanshen Detection Technology, Hangzhou, China). The plant height, panicle length, tiller number, and grain number were obtained from the measurement of the main culm. All agronomic traits and grain size measurements were based on at least 20 individuals, which were randomly selected.

Vector construction and rice transformation

To construct the overexpression vector, the full-length coding sequence of GL12 was amplified from the wild rice and 9311 and cloned into a pBWA(V)HS vector. Genes were under the control of double 35S promoter in overexpression lines. The Pro⁹³¹¹::GL12⁹³¹¹, Pro⁹³¹¹::GL12^W, Pro^W::GL12⁹³¹¹ and Pro^W::GL12^W transgenic lines also used the pBWA(V)HS vector, but 35S promoter fragment of the pBWA(V)HS vector were replaced by the GL12 promoter from the wild rice or 9311 respectively, the construct vectors were introduced into the calli of NIP and by Agrobacterium mediated transformation. To generate knockout mutants, two GL12 site-specific target sequences were selected in the fifth exon and introduced into the CRISPR/Cas9 system, and the plasmid were transformed into GL12^W-NIL to generate the GL12 knockout mutants. Knockout lines of OsGIF1 in the GL12^W-NIL background were also generated by CRISPR/Cas9 technology.

RNA extraction and RT-PCR

Total RNA was extracted from young leaves and panicles using the RNA Easy Fast Plant Tissue Kit (TIANGEN, Beijing, China) and reverse-transcribed into the first-strand cDNA with a PrimeScript RT Reagent Kit (Invitrogen, Carlsbad, CA). Real-time qPCR experiments were performed using a SuperReal PreMix Plus (SYBR Green) (TIANGEN) on an Applied Biosystems 7500 Real-Time PCR system following the manufacturer’s instructions. All the qRT-PCR analyses were performed on at least three independent biological replicates. Rice ACTIN was used as an internal control for all the qRT-PCR analyses. Primers for qRT-PCR are listed in Supplementary Data 1.

Subcellular localization

To generate the 35S::GL12-GFP fusion, the full length cDNA of GL12 without the stop codon was cloned into vector pCAMBIA 1300. The empty 35S: GFP, 35S::GL12-GFP and NLS-RFP plasmids were transformed into Agrobacterium tumefaciens strain GV3101, and infiltrated into 3‐week‐old N. benthamiana leaves. The Agrobacteria harboring the corresponding construct and NLS-RFP were co-infiltrated into N. benthamiana leaves. At 2–4 day after infiltration, fluorescent signals in leaves were detected by laser confocal microscopy (Carl Zeiss, LSM780, Germany).

Histological analysis

The fresh young spikelet hulls were fixed using FAA (40% formaldehyde: 50% alcohol: 5% acetic acid) and embedded in Paraplast Plus (Sigma, Germany) then cut into 10-μm-thick sections. Cross sections were observed by light microscopy (Olympus, Japan), and the cell number and cell area in the outer parenchyma cell layer of hulls were measured using ImageJ software³⁸. The outer surfaces of the spikelet hulls were observed using a scanning electron microscope (GeminiSEM 300, Carl Zeiss, Germany). Cell size and cell number were measured with ImageJ software.

Dual-Luciferase transcriptional activity assay

To construct the Pro⁹³¹¹::LUC and Pro^W::LUC plasmids, we amplified a 2 kb promoter sequence from 9311 and wild rice and cloned them into the pGreen0800II-LUC vector, respectively; the resulting vectors were used as reporters. The full-length cDNA of WRKY53 and GIF1 were cloned into the pGreen0800II-62-SK vector to construct 35S: WRKY53 and 35S: GIF1, which were used as the effector. Transient transactivation assays were performed using the tobacco (N. benthamiana) leaf system. The LUC and REN activities were measured by the Dual-Luciferase Reporter Assay System (Promega, USA), and the LUC activity was normalized to REN activity. Primers used for these constructs are listed in Supplementary Data 1.

Yeast one-hybrid assay

The full-length cDNA of WRKY53 and GIF1 were cloned into the pB42AD vector to generate prey constructs. Approximately 2 kb of the GL12 promoter from wild rice and 9311 were cloned into the lacZ (β-galactosidase) reporter plasmid pLACZi as bait constructs. Yeast strain EGY48 was used for transformation. All the procedures were performed according to the manufacturer’s protocol (Clontech, USA).

Luciferase complementation imaging (LCI) Assay

The LCI assays for the interaction between GL12 and GIF1 were performed in N. benthamiana leaves. The full-length GL12 and GIF1 coding regions were fused with the N-terminal and C-terminal parts of the luciferase reporter gene, respectively. The nLUC-GL12 and cLUC-GIF1 constructs were transformed into Agrobacterium GV3101 cells and co-infiltrated into N. benthamiana leaves. One same subcellular localization gene, NAC2, was used as the negative control^39,40. The LUC activity was analyzed using Tanon 5200 plant imaging system (Tanon Science & Technology, China).

Electrophoretic mobility shift assay

Electrophoretic mobility shift assay (EMSA) was used to examine the binding of the WRKY53 and GIF1 proteins to the motif in the promoter regions of Pro⁹³¹¹ and Pro^W, respectively. We synthesized 5′- biotin labeled DNA fragments containing binding sites as probes and used unlabeled DNA fragments of the same sequence for competitive reactions. The MBP-OsWRKY53 and GST-GIF1 fusion protein were expressed in E. coli pMal-c2x at 16 °C for 20–24 h in the presence of 0.1 mM isopropyl β-D-1-thiogalactopyranoside. The LightShift™ Chemiluminescent EMSA Kit was used to perform EMSA according to the manufacturer’s instructions (Thermo Fisher Scientific, USA). The binding reactions were incubated for 30 min at room temperature and then resolved by electrophoresis on 6% (w/v) native polyacrylamide gels in 0.5×TBE buffer. Probe sequences are shown in Supplementary Data 1.

Yeast two-hybrid assay

Matchmaker Gold Yeast Two-Hybrid System (Clontech, USA) was used for Y2H assays. The CDS of GL12 was cloned in prey vector pGBKT7, and the GIF1 were cloned into the bait vector pGADT7. The prey and bait vectors were transformed into the Y2H Gold yeast. All the procedures were performed according to the manufacturer’s protocol (Clontech, USA). Relevant primers are listed in Supplementary Data 1.

Haplotype and evolutionary analyses

For the Hap-G and Hap-T distribution analysis, 1772 (596 japonica, 1176 indica) and 909 (79 japonica, 191 indica, 16 rifipogon) accessions from the 3K Rice Genomes Project and rice pan-genome database³⁶, respectively, were used (Supplementary Data 2 and 3). The nucleotide diversity (π) of 52 wild rice, 56 indica, and 51 japonica accessions were calculated in 50-kb sliding windows using VCFtools⁴¹ and the ratio of genetic diversity in all accessions was calculated. The genomic sequences and grain length of 909 cultivated indica rice were obtained from the 3K Rice Genomes Project database. The grain length of two types of promoters of GL12 with T or G haplotype was compared using t-test. The geographical distribution of 909 cultivated rice were obtained from the 3K Rice Genomes Project database and marked on map using R-version (Team, 2014). The genotype of 219 accessions including wild rice were from rice pan-genome database³⁶. Fifteen SNPs and five InDels of GL12 in the promoter were aligned with ClustalW. Haplotype frequency data were processed with DnaSP5⁴² and visualized Median-joining networks were generated by Popart⁴³.

Reporting summary

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