Natural variation in SL6 determines fatty acid components and seed longevity in rice Authors: Yifeng Wang, Jiezheng Ying, Jia Li, Xinyong Liu, Qianya Zhou, Shuang Jia, Guanghao Li, Wanning Liu, Zhenghao Tao, Hongzhou Li, Zhiyong Li, Jie Huang, Xiaohong Tong, Xixi Liu, Yu Cheng, Yuxuan Hou, Guangheng Zhang, Yangjun Zhang, Kun Zhou, Xiaoshan Zeng, Jian Zhang Published: 2026 ABSTRACT Seed longevity (SL) is vital for ensuring food security worldwide. However, the genetic basis of SL has been scarcely documented. Here, we report the cloning of a major SL locus, qSL6, encoding a fatty acyl-ACP thioesterase type B. SL6 is functionally conserved in regulating palmitic acid synthesis in seeds, conferring higher oxidation durability and SL in various species. Through the VP1-SL6 module, a seed desiccation–derived ABA signal is transmitted via VP1, which directly activates SL6 transcription to alter the fatty acid composition and elevate SL in seeds. The ancestral elite allele SL6HHZ harbors a virus-derived CT-rich motif cis-element in the 5′UTR, which serves as a universal, bidirectional mRNA stabilizer, contributing to the divergence between indica and japonica in terms of SL. Moreover, manipulating SL6 expression via marker-assisted selection or transgenic approaches notably improved SL in rice cultivars and F1 hybrids without affecting major agronomic traits. Our findings provided a promising genetic locus for improving SL in rice. INTRODUCTION Quality seed is the fundamental and most critical input for sustainable agriculture. Driven by rising food demand, the global commercial seed market share is projected to reach USD 140 billion by 2030. However, similar to all living organisms, seeds are subject to aging, gradually losing vigor and viability until they eventually perish (1). It is estimated that 25 to 33% of the world’s crop seeds are lost during storage each year, substantially affecting global food security (2). Seed longevity (SL), the period during which seeds maintain viability, varies widely among species, ranging from a few days to thousands of years (3). This longevity is closely linked to the composition of seed storage reserves and the conditions (4, 5). High levels of moisture, oxygen, and temperature can activate mitochondrial respiration, leading to the generation of reactive oxygen species (ROS) that oxidize storage reserves, such as lipids, which may cause molecular damage to cells (1). Conversely, chemical protectants such as vitamin E, raffinose oligosaccharides (RFOs), and flavonoids have the potential to neutralize ROS-mediated oxidation, thereby safeguarding seeds from deterioration (6–8). Furthermore, phytohormones, particularly abscisic acid (ABA), play a crucial role in enhancing SL (4, 9). SL is a complex agronomic trait governed by multiple quantitative trait loci (QTLs). To date, the genetic basis for SL has been extensively investigated in model plants. In Arabidopsis, genes such as VTE, AtFAHD1a, PIMT, ABI3, and ATM have been associated with tocopherol biosynthesis, lipid peroxidation, abscisic acid accumulation, and the DNA damage response, respectively, contributing to SL (6, 10–12). In contrast, rice research, a principal crop that provides sustenance for over half of the global population, has identified only a limited number of cloned genes despite the detection of more than 60 QTLs across various genetic populations. These include LOX family members, OsALDH7, Hsp18.2, Gly13, OsPIMT1, and OsMRBS5, most of which were identified through mutant studies and may lead to severe growth defects, potentially hindering their application in breeding practice (13–19). Through GWAS analysis, Yuan et al. (20) successfully cloned a natural allele of the indole-3-acetic acid (IAA)–amido synthetase gene GRETCHEN HAGEN3-2, which influences rice SL by regulating endogenous IAA levels, ABA signaling, and LATE EMBRYOGENESIS ABUNDANT proteins accumulation. Despite important progress in this area, the genetic basis of rice SL remains unclear, particularly regarding the natural alleles that contribute to variation in SL across species and varieties. Here, we report the map-based cloning of a major SL QTL, qSL6, which encodes a fatty acyl-ACP (acyl-carrier protein) thioesterase type B in regulating the ratio of saturated fatty acids (SFAs) to unsaturated fatty acids (UFAs) in grains, helping seeds withstand oxidation during storage and ultimately influencing SL. The ancestral elite allele, SL6HHZ, features a 62–base pair (bp) CT-rich motif cis-element in the 5′ untranslated region (5′UTR), which enhances SL6 mRNA stability, leading to a higher SFA palmitic acid (PA), lower UFA oleic acid (C18:1; OLA), and linoleic acid (C18:2; LA) contents, ultimately improving SL in indica rice varieties. Through the “VP1-SL6” module, ABA signal enhances SL, partly by influencing the UFA/SFA ratio. Manipulating SL6 expression via marker-assisted selection or transgenic approaches significantly improved seed storability in japonica backgrounds without negatively affecting major agronomic traits. RESULTS qSL6 controls SL without affecting major agronomic traits From a collection of rice germplasm, we identified two varieties, Huanghuazhan (Oryza sativa ssp. indica, HHZ) and Jizi-1560 (O. sativa ssp. japonica, JZ), that show extremely contrasting SL. The G50A, which refers to the time required for artificial aging (AA) treatment to achieve a 50% germination ratio, was significantly higher in HHZ than in JZ (fig. S1, A and B). Accordingly, HHZ exhibited markedly higher germination rate (GR), germination index (GI), and seedling vigor index (SVI) than JZ when equal AA and natural aging (NA) treatments were applied, respectively (fig. S1, B to H). To facilitate the cloning of SL QTLs, we constructed a set of chromosome segment substitution lines (CSSLs) in BC3F2 using JZ as the donor and HHZ as the recurrent parent. Through SL screening, a CSSL carrying a fragment of chromosome 6 from JZ showed substantial phenotypic variation in SL and was subsequently developed into near-isogenic lines (NILs) (fig. S2, A and B). In morphology, the NIL-JZ and NIL-HHZ were almost identical in plant height, heading date, seed setting, effective panicle number, and seed size (figs. S2C and S3, B to H). However, the NIL-HHZ exhibited 40% higher G50A, and more robust seed vigor with higher GR, GI, and SVI under either AA or NA treatments (Fig. 1, A to E, and fig. S2, E, G, and H). The results implied the existence of a major QTL, designated qSL6, specifically controlling SL on Chr 6, from which high SL originates from the HHZ allele. Map-based cloning of SL6 and genetic complementation analysis. (A) Germination phenotypes of NILs and the Com seeds after AA treatment for 0 and 25 days, respectively. On the right side of the red line were non-germinated seeds. Photographs were taken on day 7. (B) Germination rate and G50A of NILs, and the Com seeds upon AA treatment. (C) Germination rate of the NILs and the Com seeds upon NA treatment. (D and E) Germination index and Seedling vigor index of NILs, and the Com seeds upon AA treatment. (F) Map-based cloning of SL6. SL6 was mapped to the interval between RM8112 and RM3805 on the short arm of chromosome 6 and narrowed to a 65-kb region. The GR of the seeds after AA treatment for 20 days is exhibited on the right of the schematic for each representative recombinant line. Red box: S1 (stimulatory region 1), white boxes: UTRs (untranslated regions), black boxes: exons, and black lines: introns. (G) FAs compositions in the brown rice of NILs and the Com lines. (H and I) MDA and H2O2 contents of NILs and the Com seeds upon AA treatment. DW: dry weight. Error bars represent SD with biological triplicates (n = 3) in (D), (E), (G), (H), and (I), and biological triplicates (n = 3, each replicate contains 30 seeds) in (B), (C), and (F). Different letters indicate statistical differences at P < 0.01 by one-way ANOVA analysis of variance in (F). *P < 0.05, **P < 0.01, and ***P < 0.001 by two-tailed Student’s t test analysis in (B), (D), (E), (G), (H), and (I). SL6, a fatty acyl-ACP thioesterase type B, regulates fatty acid composition Through fine mapping, we finally located qSL6 within a 65-kb region between the molecular markers P1 and P2 (Fig. 1F). There is only one putative ORF (LOC_Os06g05130) encoding a fatty acyl-ACP thioesterase type B (Fig. 1F). Sanger sequencing revealed that there is only one missense variation, A2828G in the coding sequence, and a 62-bp virus-derived S1 element deletion in the 5′UTR (−1213 to −1152 in position) of the LOC_Os06g05130JZ, likely reducing the gene transcription level in NIL-JZ (Fig. 1F and fig. S3A). We therefore performed a genetic complementation experiment by transforming the LOC_Os06g05130HHZ gene, driven by its native promoter (2.5 kb), into the NIL-JZ background (fig. S2D). The genetic complemented lines (Com) showed recovery of G50A, GR, GI, and SVI in the Com seeds after AA or NA to the NIL-HHZ level (Fig. 1, A to E, and fig. S2, E, G, and H). Meanwhile, the major agronomic traits of the Com lines were statistically identical to NIL-HHZ and NIL-JZ (figs. S2C and S3, B to H). Hence, LOC_Os06g05130 is the causal gene of SL6. The fatty acids (FAs) in rice grains are mainly composed of PA, OLA, and LA, which account for ∼90% of the total FA content (21). It is known that fatty acyl–ACP thioesterase type B catalyzes the cleavage of thioester bonds in activated fatty acyl–coenzyme A (CoA) substrates and ACPs, mainly regulating PA biosynthesis (22). In this context, we quantified FA composition in parents and NILs. The results showed that HHZ and NIL-HHZ had significantly higher PA and correspondingly lower total oil and unsaturated OLA and LA content in brown rice than JZ and NIL-JZ, respectively (Fig. 1G and figs. S1, I and J, and S2F). As a consequence of the altered FA compositions, NIL-JZ produced more malondialdehyde (MDA) and hydrogen peroxide (H2O2) (Fig. 1, H and I), inhibited seed germination upon the AA treatment, and led to lower SL than NIL-HHZ. To explore the indicative FA parameters in SL, we further quantified the FA compositions of 28 rice varieties covering four diverse ecotypes. There was a strong positive correlation between rice SL and the composition of PA coefficient of determination (R2) = 0.4383, P = 0.0001), but negative correlations with the compositions of OLA (R2 = 0.3129, P = 0.0020) and LA (R2 = 0.3660, P = 0.0006) (fig. S4). Given the highest correlation between PA composition and SL, we hypothesized that PA composition could be used as a physiological indicator of rice SL. Notably, the absolute quantities of OLA and LA, but not PA, are also negatively correlated with the SL (fig. S4). SL6 conservatively modulates SL across monocot and dicot species We generated sl6 knockout mutants in HHZ and overexpression lines (OxSL6) in Nipponbare (O. sativa ssp. japonica, Nip), respectively (fig. S5, A to C and E). Compared to the control, sl6 lines had significantly lower PA and higher UFA content, resulting in higher MDA and H2O2 levels and deteriorated seed vigor, including GR, G50A, GI, and SVI (Fig. 2, A to E, and fig. S5, D and M to P). In contrast, OxSL6 lines showed the opposite trend to that of the mutants for FA composition and SL (Fig. 2, A to E, and fig. S5, D and M to P). However, OxSL6 lines led to obvious dwarfism, earlier heading, and smaller grain weight phenotypes, while sl6 lines exhibited similar agronomic traits to HHZ, except for later heading (fig. S5, A, B, and F to L). SL6 regulates grain PA, OLA and LA content and SL. (A and B) Germination phenotypes of wild type, sl6, and OxSL6 seeds after AA treatment, respectively. On the right side of the red line were nongerminated seeds. Photographs were taken on day 7. (C) FAs compositions in the brown seeds of HHZ, sl6, Nip and OxSL6. (D and E) G50A and GR of HHZ, sl6, Nip, and OxSL6 seeds upon AA treatment. (F) FAs compositions of the wild type (Columbia) and OxSL6(At) seeds in Arabidopsis thaliana. WT: wild type. (G and H) Germination phenotypes of OxSL6(At) seeds in Arabidopsis and OxSL6(Nb) seeds in tobacco after AA treatment, respectively. Photographs were taken on day 9. (I and J) Germination rate and G50A of OxSL6(At) seeds in Arabidopsis and OxSL6(Nb) in tobacco upon AA treatment. Error bars represent SD with biological triplicates (n = 3) in (C), (D), and (F); and biological triplicates (n = 3, each replicate contains 30 seeds) in (E); and biological triplicates (n = 3, each replicate contains 100 seeds) in (I) and (J). *P < 0.05 and **P < 0.01 by two-tailed Student’s t test analysis in (C), (D), (F), (I), and (J). In addition, we overexpressed the rice SL6 in Arabidopsis thaliana (At) and tobacco Nicotiana benthamiana (Nb) (figs. S6B and S7B). Similar to the OxSL6 seeds from rice, the green cotyledon of OxSL6 (At) seeds had 29.5% higher PA but lower OLA and LA compositions, when compared with that of the WT (Fig. 2F and fig. S6D). Upon AA treatment, OxSL6(At) and OxSL6(Nb) displayed more robust seed vigor with higher GR, G50A, GI, SVI, and cotyledon green rate than the controls, suggesting that SL6 is functionally conserved across monocot and dicot species (Fig. 2, G to J, and figs. S6, C to F, and S7, F to H). The OxSL6(Nb) exhibited a phenotype of dwarf, later flowering with fewer rosette leaves, whereas OxSL6(At) had no visible morphology differences in plant growth and development (figs. S6, A and G to K, and S7, A and C to E). The conserved S1 cis-element confers higher SL6 stability and SL The 62-bp insertion in the 5′UTR of SL6HHZ showed high similarity to the stimulatory region 1 (S1) of the Cauliflower Mosaic Virus 35S RNA leader sequence, which has been known as a cis-element containing a CT-rich motif to enhance gene transcriptional level (Fig. 1F) (23). The SL6 transcription levels were significantly higher in NIL-HHZ than in the NIL-JZ (fig. S3A). In a luciferase (LUC) assay on the promoter and 5′UTR of LOC_Os06g05130, pSL6HHZ::LUC had consistently higher expression than pSL6JZ::LUC in different truncation forms containing S1 (Fig. 3, A and B). However, when the region containing the 62-bp insertion was truncated out, the pSL6-1152HHZ::LUC and pSL6-1152JZ::LUC exhibited similar expression levels, suggesting that the S1 insertion enhanced SL6 level (Fig. 3, A and B). A SL6 mRNA stability assay revealed that the half-life of SL6HHZ in NIL-HHZ was over twice that of SL6JZ in NIL-JZ (Fig. 3C). At the same time, the genetic complementation line had SL6 mRNA stability almost identical to that of NIL-HHZ, suggesting that the S1 element prevented SL6 mRNA degradation (Fig. 3C). The 62-bp insertion harbors a conserved S1 cis-element conferring higher SL6 stability and SL. (A) Schematic representation of the different truncation forms flanking the SL6 promoter and 5′UTR. The blue, red and black squares represent the promoter, 5′UTR and S1, respectively. (B) LUC assay of the vectors from (A) in rice protoplast. REN: Renilla luciferase. (C and D) Quantitation of the mRNA abundance and transcript half-life estimation of SL6 in various lines. Cor: cordycepin. (E) FAs compositions in the brown seeds of J492 and sl6⊿S1. (F) Germination rate and G50A of J492 and sl6⊿S1 seeds upon AA treatment. (G) Germination phenotypes of J492 and sl6⊿S1 seeds after AA treatment, respectively. On the right side of the red line were nongerminated seeds. Photographs were taken on day 7. (H) Relative transcription of SL6 in 15 DAP seeds of HapJZ (n = 56) and HapHHZ (n = 68) accessions. (I) Germination rate of HapJZ and HapHHZ seeds after AA treatment for 15 days. (J) Linear regressions of relative transcription of SL6 and seed vigor of the accessions. (K) Schematic representation of the different promoters from rice (NAC23 and bZIP58) or Arabidopsis (AtAP1 and AtFT) inserting with the S1 motif. (L) The effects of the different vectors from (K) on LUC/REN activity by luciferase transient transcriptional activity assay. Error bars represent SD with biological triplicates (n = 3) in (C), (D), (E), and (J); six biological replicates (n = 6) in (B) and (L); and biological triplicates (n = 3, each replicate contains 30 seeds) in (F). *P < 0.05, **P < 0.01, and ***P < 0.001 by two-tailed Student’s t test analysis in (B), (E), (F), (H), (I), and (L). To test the biological effect of S1 in rice, we knocked out the S1 element in the SL6 of a variety J492 (O. sativa ssp. japonica carrying SL6HHZ) by a CRISPR-Cas9 system targeting two single guide RNA sites flanking the S1 region. We obtained a sl6⊿S1 line harboring a 35-bp deletion in the S1 (fig. S8A). sl6⊿S1 seeds had significantly reduced SL6 transcription level and accordingly lower PA, higher LA, and OLA, which mimicked seeds from NIL-JZ (Figs. 1G and 3E, and fig. S8, B and C). In terms of SL, sl6⊿S1 seeds also showed lower GR, GI, and SVI after AA or NA treatments compared to J492 (Fig. 3, F and G, and fig. S8, D to I). The mRNA stability assay further revealed that the half-life of sl6⊿S1 was significantly lower than that of the wild type, supporting the mRNA stability results in NIL-JZ (Fig. 3D). Last, based on the indel in S1, we categorized 124 germplasm accessions into two groups: HapHHZ, which includes S1, and HapJZ, which does not include S1 (table S1). It was found that varieties in HapHHZ exhibited significantly higher transcription level of SL6 and higher seed vigor after 15 DAA (days after AA) treatments than those in HapJZ, showing a significant positive correlation (R2 = 0.6184, P < 0.0001) between SL6 transcription and SL (Fig. 3, H to J, and table S1). Hence, the evidence above suggests that the S1 indel in the 5′UTR regulates SL6 transcript stability and contributes to variations in seed PA content and SL among rice varieties. To test whether S1 is a universal transcription enhancer, we fused the S1 sequence to the 3′ end of four promoters for the LUC assay. These included NAC23 and bZIP58 from rice and AtAP1 and AtFT from Arabidopsis, which exhibit different spatio- and temporal expression patterns (Fig. 3K) (24–27). Intriguingly, the S1 element insertion, either in forward or reverse directions, significantly enhanced the expression level of LUC reporters in tobacco epidermal cells, indicating that S1 is a bidirectional, universal enhancer suitable for various plant species and genes (Fig. 3, K and L, and fig. S9). The sequence analysis of the S1 region revealed a potential palindrome structure, which is assumed to be the core element for its enhancer function (fig. S10A). To this end, we generated various mutated forms of the stem-loop structure to test their effects in driving LUC expression (fig. S10, A and B). M1, with a more stringent stem-loop structure and a higher dG, showed the strongest LUC transcription and translation in tobacco cells (fig. S10, C, D, and E). At the same time, LUC expression was somewhat compromised by other mutated S1 forms with loosened stem loops and lower dGs (fig. S10, C, D, and E). Hence, the stem-loop structure of S1 is essential for its enhancer functions. ABA regulates FA compositions in developing seeds to affect SL A GUS staining assay on pSL6JZ::GUS lines suggested that SL6 is more robustly expressed in the embryo than endosperm (fig. S11D). In addition, the GUS reporter was also detected in the coleoptile, radicle, leaf, pistil, and root (fig. S11D). Similarly, reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis revealed that SL6 is constitutively transcribed in all tested tissues, with relatively higher levels in pistils and developing seeds, particularly at 15 DAP (days after pollination), when endogenous ABA levels are at a plateau in developing seeds (fig. S11A) (28). The expression of SL6 was induced by exogenous ABA in germinative and developing seeds, suggesting that the ABA signal may mediate FA biosynthesis through SL6 (fig. S11, B and C). Transient expression of p35S::SL6JZ-GFP in epidermal cells of tobacco and rice protoplasts both showed overlapping signals of the SL6JZ–green fluorescent protein (GFP) and autofluorescence of the chloroplast, suggesting that SL6 is located in the chloroplast, which supports its potential roles in FA biosynthesis (fig. S11, E and F). Given the ABA-inducible expression pattern of SL6, we determined to investigate the potential effect of ABA in FA biosynthesis in seeds and SL. To this end, the aba1, an ABA biosynthesis mutant with a low endogenous ABA level (29, 30), was used to test FA composition and SL. Compared with the Nip control, aba1 exhibited lower SL6 transcription level and PA content, and substantially lower G50A, GR, GI, and SVI after 6 DAA, possibly due to elevated MDA and H2O2 inhibition (fig. S12). Conversely, exogenous ABA treatment significantly increased SL in both HHZ and sl6 backgrounds (fig. S13). Notably, sl6 exhibited lower sensitivity to exogenous ABA than HHZ, suggesting that SL6 is a key node in the ABA-mediated regulatory pathway of FA biosynthesis (fig. S13C). VP1 directly activates the transcription of SL6 According to the plant regulomics database (31), SL6 transcription may be regulated by B3-type transcription factors. Among the rice B3 family members, VP1/OsABI3 attracted our particular attention as it is induced by ABA and involved in the rice seed dormancy and preharvest sprouting (32). The expression of SL6 was significantly reduced in vp1 mutant and elevated in VP1 overexpression (OxVP1) lines (fig. S14, A and B). Meanwhile, ABA substantially activated SL6 transcription in the Nip. In contrast, this activation was weaker in vp1, suggesting that VP1 may act as an upstream regulator mediating the ABA signal to SL6 (Fig. 4A). Four layers of evidence supported the VP1-SL6 regulatory relationship. First, VP1 could bind the promoters of SL6HHZ and SL6JZ and activate reporter gene expression in the yeast one-hybrid (Y1H) system (Fig. 4, B, C, and D). VP1 showed more vigorous activation activity on pSL6HHZ than on pSL6JZ. However, the VP1-mediated activation was gradually compromised on pSL61-681 and pSL61-409, respectively (Fig. 4, B, C, and D). It suggests that the RY/Sph and S1 cis-elements were indispensable for the function of the SL6 promoter. Second, the B3 domain of VP1 also specifically binds to the P1 region containing an RY/Sph cis-element of SL6 in an electrophoretic mobility shift assay (EMSA) (Fig. 4, E to G). At the same time, mutation of the core sequence “CATGCA” of the RY/Sph motif on the F1 probe effectively impaired the VP1 binding (Fig. 4, E to G). Third, the in vivo binding of VP1 on SL6 promoter was significantly enriched in the P1 region as revealed by a chromatin immunoprecipitation–qPCR (ChIP-qPCR) assay on pVP1::VP1-GFP young seedlings (Fig. 4, G and H). Exogenous ABA treatment further strengthened the enrichment of VP1 on this region (Fig. 4H). Last, a transient Luciferase (LUC) transcriptional activity assay confirmed that VP1 could transactivate the SL6HHZ::LUC and SL6JZ::LUC reporters in rice protoplasts, and the expression of the reporters was increased when exogenous ABA treatment was applied (Fig. 4, I and J). Together, this suggests that VP1 directly binds to the RY/Sph core element of the SL6 promoter to transactivate transcription. VP1 promotes SL by activating SL6. (A) Relative transcription of SL6 in 15 DAP seeds of Nip and vp1 mutants. Seeds were grown on 1/2 MS medium containing 0 or 10 μM ABA for 12 hours before sampling. (B) Schematic representation of the SL6 promoter vectors. LacZ: β-galactosidase. (C and D) Y1H assay and β-galactosidase activity of VP1 binding to the different promoter forms of SL6 as described in (B). (E and F) EMSA on the binding of VP1 to SL6 promoter and RY/Sph motif (CATGCA). The mutated nucleotides were marked in lowercase. (G) Probe positions on SL6. Gray box: genome region, black line: promoter region. Numbers indicate the distances (bps) of probes to the transcription start site. Red star: RY/Sph motif (CATGCA, −1424 to −1418) in P1. (H) ChIP-qPCR assay showing enrichment of VP1 at the SL6 promoter. The enrichment values were normalized to the input. (I and J) LUC assay of VP1 on SL6 transcription. mproSL6JZ: LUC, RY/Sph motif mutated. AtUbi3:rLUC was used as an internal control. (K) FAs compositions in the brown seeds of the Nip, vp1 and OxVP1 lines. (L and M) Germination rate and G50A of various seeds upon AA treatment. (N) Germination phenotypes of various seeds after AA treatment, respectively. On the right side of the red line were nongerminated seeds. Photographs were taken on day 7. Error bars represent SD with biological triplicates (n = 3) in (A), (D), (H), (K), and (M); six biological replicates (n = 6) in (J); and biological triplicates (n = 3, each replicate contains 30 seeds) in (L). Different letters indicate statistical differences at P < 0.05 by one-way ANOVA analysis of variance. *P < 0.05, **P < 0.01, and ***P < 0.001 by two-tailed Student’s t test analysis. Similar to sl6, vp1 had lower PA content and higher OLA, LA, MDA, and H2O2 contents, resulting in lower seed vigor after AA treatment (Fig. 4, K to N, and fig. S14, C to F). In contrast, OxVP1 seeds showed opposite trends in FA compositions and higher SL, mimicking OxSL6 seeds (Fig. 4, K to N, and fig. S14, C to F). To verify the genetic relationship between VP1 and SL6, we further developed OxSL6/vp1 lines to test their SL (fig. S14B). As a result, the overexpression of SL6 significantly enhanced GR, G50A, GI, and SVI of vp1 (Fig. 4, L to N, and fig. S14, E and F), suggesting that VP1 regulates SL through its downstream target SL6. SL6 underwent selection with a latitude-based distribution Using published genome sequences (33, 34) and molecular marker targeting the S1 indel (table S2), the SL6 genotype was analyzed across various wild rice types and modern cultivars. In Oryza brachyantha (FF genome), Oryza punctata (BB genome), and AA genome species such as Oryza meridionalis, Oryza glumipatula, Oryza glaberrima, and Oryza barthii, no homologs of SL6 were identified (Fig. 5A). In contrast, SL6 was classified as HapHHZ in all tested accessions of Oryza nivara and Oryza longistaminata, and in the majority of Oryza rufipogon (Fig. 5A). This classification suggests that the SL6HHZ haplotype may represent the ancestral form in rice evolution. Approximately 0.55 million years ago, SL6 began to diverge between indica and japonica rice varieties, marked by the deletion of the S1 element in japonica varieties (Fig. 5A). In terms of the distribution in subspecies of 4607 rice accessions around the world, HapHHZ was dominated by aus, indica, and VI/aromatic, accounting for 89.84, 73.42, and 65.96%, respectively, whereas HapJZ was mainly japonica and intermediate, accounting for 99.21 and 72.09%, respectively (Fig. 5B and table S3). Furthermore, we examined the geographic distribution of SL6 haplotypes within 934 modern rice cultivars from China (table S4). HapJZ accessions were predominantly distributed in higher-latitude regions, and the proportion of HapHHZ decreased with latitude (Fig. 5C and table S4). This pattern indicates a notable regional differentiation between the two haplotypes. In line with this, the two haplotypes also showed substantial genetic divergence for SL6, with a fixation index (Fst) of 0.591 (Fig. 5F). As depicted in Fig. 5D, the nucleotide diversity (π) of SL6 in japonica cultivars was significantly lower (π = 0.000158) compared to that observed in indica (π = 0.002244) and aus subpopulations (π = 0.006253) (Fig. 5D). The neutral test (Tajima’s D) value for SL6 was −2.68471 in japonica cultivars, whereas it was recorded as 2.83611 in aus and 0.017443 in indica subpopulations (Fig. 5E). These results suggest that SL6 experienced positive selection in japonica rice. SL6 undergoes selection with a latitude-based distribution and shows great promise for improving SL. (A) Genotype analysis and distribution pattern of the SL6 haplotypes in wild rice and cultivars based on the S1 Indel. The pie charts on the right show the frequencies of the SL6 haplotypes in each species. Ma, million years ago. (B) Haplotype frequency of SL6 among 4607 rice accessions. (C) Geographic distribution of SL6 haplotypes in 934 Chinese modern cultivars. (D to F) π, Tajima’s D, and Fst analyses in the part of chromosome 6. The red line represents a selective sweep region surrounding the SL6 locus. (G) The seedling growth of SL6HHZ(L), SL6JZ(L), SL6HHZ(N) and SL6JZ(N) transgenic seeds after NA treatment. Bars, 5 cm. (H to K) Germination rate and G50N of SL6HHZ(L), SL6JZ(L), SL6HHZ(N) and SL6JZ(N) transgenic seeds upon NA treatment. G50N: the time span of NA treatment for the seeds to reach a 50% GR. (L) Germination phenotypes of NIL-SH/H (J), NIL-SH/J (J), and NIL-SJ/J (J) seeds after AA treatment. Photographs were taken on day 7. (M) Germination rate and G50A of NIL-SH/H (J), NIL-SH/J (J), and NIL-SJ/J (J) seeds upon AA treatment. (N) Germination phenotypes of F1-SH/J and F1-SJ/J seeds after AA treatment for 0 and 15 days, respectively. On the right side of the red line were nongerminated seeds. Photographs were taken on day 7. (O) Germination rate and G50A of F1-SH/J and F1-SJ/J seeds upon AA treatment. Error bars represent SD with biological triplicates (n = 3) in (I) and (K) and biological triplicates (n = 3, each replicate contains 30 seeds) in (H), (J), (M), and (O). **P < 0.01 and ***P < 0.001 by two-tailed Student’s t test analysis in (I), (K), (M), and (O). SL6 effectively improves the SL of rice Although higher SL6 expression is associated with greater SL in rice, excessive SL6 expression may lead to dwarfism, delayed heading, and lower seed weight, which hinder its application in breeding (fig. S5). To increase SL6 transcription to an appropriate level, we generated pSL6HHZ::SL6HHZ and pSL6JZ::SL6JZ constructs and genetically transformed them into two elite japonica varieties harboring SL6JZ, Longken 229 (LK229), and Nangeng 46 (NG46), respectively. In contrast to the overexpression lines that exhibited more than a 10-fold increase in SL6 levels, the SL6HHZ(L)(pSL6HHZ::SL6HHZ transgenic line in LK229), SL6JZ(L), SL6HHZ(N), and SL6JZ(N) lines showed an elevation in SL6 ranging from four to seven times (figs. S5E and S15G). Notably, this increase did not result in any significant changes to the major agronomic traits of these lines, including flowering date, effective panicle number, seed setting, and grain size (fig. S15, A, B, and H to N). Nevertheless, following a NA treatment for 24 months, the SL6HHZ(L), SL6JZ(L), SL6HHZ(N), and SL6JZ(N) seeds exhibited substantial improvement in germination performance, including higher G50N, GR, GI, and SVI, when compared to the controls (Fig. 5, G to K, and fig. S15, C to F). The results demonstrated the feasibility of improving SL through transgenic modulation of SL6 expression at an appropriate high level. Alternatively, through backcross breeding and marker-assisted selection, we introgressed the SL6HHZ locus into the cultivar JZ, generating NIL-S(J)s (fig. S16A). As expected, the transcription levels of SL6, G50A, GR, GI, and SVI in NIL-SH/H(J) (NIL carrying homozygous SL6HHZ in JZ background) were significantly higher than in NIL-SJ/J(J) under AA treatments, without affecting other major agronomic traits (Fig. 5, L and M, and fig. S16, B to M). Intriguingly, the SL6 heterozygous NIL-SH/J(J) seeds also showed higher G50A than the NIL-SJ/J(J). However, the extent was not as high as in homozygous NIL-SH/H(J) seeds (Fig. 5, L and M, and fig. S16, D and E), indicating that the dominant effect of SL6HHZ could be used to improve the SL of hybrid rice. To this end, we further generated F1 seeds by crossing NIL-HHZ and NIL-JZ (HHZ background) with the elite japonica-type male-sterile line CJ88A (Chunjiang 88A carrying SL6JZ), respectively (fig. S17A). As expected, the F1-SH/J (F1 derived from CJ88A♀ × NIL-HHZ♂) seeds exhibited higher SL6 transcription level and seed vigor than those of F1-SJ/J (F1 derived from CJ88A♀ × NIL-JZ♂) seeds (Fig. 5, N and O, and fig. S17, B to D). Together, these results strongly suggest that introgressions of the SL6HHZ locus hold great promise for enhancing SL in japonica inbreds and indica-japonica hybrids. DISCUSSION SL6 determines SL by regulating FA compositions The current study identified a major QTL, qSL6, which encodes a fatty acyl-ACP thioesterase type B. Biochemically, the FAT family, comprising both FATA and FATB groups, facilitates the hydrolysis of ACP to generate free FAs (35, 36). FATB specifically hydrolyzes saturated acyl-ACP chains ranging from C8 to C18, with a notable preference for C16 (37). This activity can enhance both the overall quantity and the relative proportion of SFAs (37, 38). Aligning with findings in Arabidopsis, maize, and soybean, in which disruption of FATB reduced SFAs (39–41), this research underscores the highly conserved role of SL6 in positively regulating PA and SFA composition in both rice and Arabidopsis, as illustrated in Figs. 1 and 2. Lipids, especially those rich in UFAs with two or more double bonds, are susceptible to oxidation when exposed to high moisture, oxygen, and elevated temperatures. This oxidation generates free radicals that can damage cellular membranes and contribute to seed deterioration (42, 43). Various genes associated with FA biosynthesis and metabolic pathways have been linked to seed vigor and longevity in species such as Arabidopsis, cotton, and Brassica napus (44–48). In rice, the LOX3 gene, which encodes lipoxygenase, exacerbates FA peroxidation, leading to reduced levels of UFAs and soluble sugars derived from triacylglycerols, ultimately impairing seedling growth due to insufficient energy supply (13, 14). This study underscores the critical influence of FA composition on SL. ABA promotes SFA and reduces UFA compositions through the VP1-SL6 pathway Homeoviscous adaptation, which refers to the adaptation of the cell membrane lipid composition to maintain adequate membrane fluidity, has been a well-adapted, responsive mechanism for biological individuals to cope with unfavorable environments (49). Nevertheless, the mechanism by which homeoviscous adaptation is triggered in plants remains unclear. Recently, ABA has been identified as a key signal determining FA compositions in plants, although its effect varies among species (50). Exogenous ABA can elevate the ratio of UFA, such as C18:1 and C18:2, in oil palm, siberian apricot, and rapeseed (51, 52). In contrast, we observed a positive effect of ABA on the SFA/UFA ratio (fig. S12), consistent with reports in Arabidopsis and microalgae (53, 54). Actually, very similar phenomena have been reported, as ABA can promote or, conversely, inhibit flowering in rice (55, 56). Given the effect of phytohormones such as ABA is highly sensitive and dosage dependent, we assume the opposite responses to ABA in different species might be contributed to many potential aspects, such as the dosage used, timing and tissue types of treatment, and the fatty acid component backgrounds. Generally, SFA confers higher rigidity but less fluidity to cell membranes. It is reasonable to propose that desiccation stress during seed maturation elevates ABA levels, subsequently up-regulating SL6. This upregulation enhances SFA synthesis, thereby conferring increased rigidity and integrity to cell membranes throughout seed development. Additionally, ABA has been reported to repress rice flowering (57). In this study, we also found that sl6 showed early flowering, whereas OxSL6 showed the opposite (fig. S5G), further supporting that SL6 is involved in ABA-mediated signal transduction. More importantly, we identified a VP1-SL6 signaling pathway that mediates the ABA signal to PA synthesis, contributing to higher SL (Figs. 4 and 6). VP1 binds to the RY/Sph cis-element of the SL6 promoter (Fig. 4). This activation of SL6 transcription results in higher PA accumulation, which enhances the resistance to lipid peroxidation and ultimately increases SL. Notably, VP1 might not be the sole upstream regulator of SL6 transcription. According to the plant regulomics database (31), numerous transcription factors, such as the ABA-inducible seed vigor regulator bZIP23, could also bind to the SL6 promoter (58), suggesting more complex ABA signaling pathways in regulating FA components. A working model illustrating SL6 regulates SL. Arrowheads represent positive regulation. S1 region enhances SL6 transcription and determines indica-japonica differentiation Although a strong association between SL6 level and SL has been established, the overdosage of SL6 may cause severe alterations in major agronomic traits, hindering its utilization in breeding (fig. S5). Fortunately, the current study identified an elite allele, SL6HHZ, that maintains a sufficiently high level of SL6 without causing a growth penalty (figs. S3 and S15). Compared to SL6JZ, SL6HHZ harbors a 62-bp insertion of a virus-derived S1 element, which increases the SL6HHZ expression by approximately four times (Fig. 3, A and B, and fig. S3A). S1 consistently conferred higher SL6 levels in the rice protoplast system with native transactivation effectors (Fig. 3, A and B) and in the yeast system with the specific effector VP1 (Fig. 4, B to D), suggesting that S1 executes its functions regardless of the transactivators. The S1 leader sequence promoted the expression of various genes in rice, Arabidopsis, and tobacco, probably by stabilizing the resulting mRNA (Figs. 2, G to J and 3, C, D, K, and L, and fig. S9). Therefore, S1 could be used as a universal, bidirectional enhancer element in biotechnology, like the well-known Ω element from the leader of Tobacco Mosaic Virus RNA (59). Previous studies suggested that the repeated CT-rich motif in S1 is involved in interactions with plant nuclear proteins (23), and that RNA binding proteins (RBPs) can recognize and bind mRNA to enhance its stability in plants (60). Similar cases have been reported in APUM9 and UBA2A in Arabidopsis (61, 62), which strongly supports this hypothesis. Therefore, further in-depth studies identifying the specific RBPs that bind the S1 motif would be an effective strategy to elucidate the molecular mechanism underlying the S1-mediated mRNA stability. Among a selection of 124 rice germplasms, accessions containing the S1 allele exhibited significantly higher SL6 transcription levels and enhanced seed SL compared to those lacking S1 (Fig. 3, H, I, and J). Furthermore, the in situ knockout of S1 in J492 notably decreased SL6 expression, SFA compositions, and SL (Fig. 3, D to G, and fig. S8). These genetic findings indicate that S1 is a functional variant contributing to SL across various rice species. HapHHZ, which carries the S1 allele, is predominantly found in indica rice varieties adapted to low-latitude regions, typically under high-temperature, high-humidity conditions (Fig. 5, A to F). In contrast, HapJZ, which lacks the S1 allele, is primarily associated with japonica varieties situated in high-latitude areas (Fig. 5, A to F). Given that indica rice typically shows greater SL than japonica (58, 63), SL6 is likely a crucial factor contributing to the divergence between indica and japonica rice, as well as to the latitude-driven geographic distribution of rice in relation to aging tolerance. We emphasized the potential to enhance SL in rice, particularly in japonica varieties, by using SL6. In the context of LK229 and NG46, two prominent japonica cultivars from northern and central China, respectively, we demonstrated that the SL6 transgene, driven by its native promoter, can substantially elevate SL6 transcription to a properly high level, thereby enhancing seed vigor after storage without affecting major agronomic traits (Fig. 5, G to K, and fig. S15). Conventional cross-breeding with indica donor varieties that carry SL6HHZ, such as HHZ, has proven highly effective in improving SL, as demonstrated by NILs in JZ backgrounds and F1 hybrid seeds established in this study (Fig. 5, L to O, and figs. S16 and S17). Particularly, hybrid rice is superior to its parents in agronomic traits with ∼20% higher yield and has developed a market of more than 3 billion US dollars in over 40 countries, especially in Asian countries. Nevertheless, hybrid rice seeds are more susceptible to deterioration due to the extensive application of gibberellic acid during seed production, which has made the SL a bottleneck for the hybrid rice industry. Our work sheds light on how to solve this issue by using the SL6HHZ allele. Alternatively, the advancement of gene editing technology opens up possibilities for in situ knock-in of the 62-bp S1 element into the 5′UTR of SL6JZ, offering great potential to precisely and effectively improve the aging tolerance of rice varieties. MATERIALS AND METHODS Plant materials and growth conditions HHZ is an indica rice cultivar, and JZ, Nip, J492, LK229, and NG46 are japonica cultivars used in this study. CJ88A is a japonica-type male sterile line adopted in the rice three-line hybrid system. A set of CSSLs in BC3F2, using HHZ as the recurrent parent and JZ as the donor, was generated to screen for QTLs regulating SL. The genome screening of JZ, HHZ, and NILs was performed using a Green Super Rice 40K gene chip with 36,584 SNP markers by the Greenfafa genetic diagnosis platform (Wuhan, China). Rice varieties (124) used for AA analysis was sourced from the core rice germplasm of Huazhong Agricultural University (64), and 934 modern cultivar samples used for natural variation analysis for SL6 were collected from China National Rice Research Institute, Zhejiang Academy of Agricultural Sciences, Northeast Institute of Geography and Agroecology, Xinjiang Rice Research Institute, Crop Research Institute of CAAS, Hunan Rice Research Institute, Guangdong Rice Research Institute, Guangxi Rice Research Institute, and Yangzhou University, respectively. For the overexpression of SL6 and VP1, the CDSs of these genes were cloned into the binary vector PU1301 driven by the maize ubiquitin promoter, using Kpn I and Bam HI sites for SL6 and Bam HI for VP1, respectively. sl6 and sl6⊿S1 mutants were generated using pYLCRISPR/Cas9-MH vector by CRISPR-Cas9 system. For the complementation test for SL6, the CDS of SL6 was cloned into the binary vector pCAMBIA1300 by using Kpn I and Bam HI sites, and 2.5 kb of the SL6HHZ promoter was inserted before the CDS of SL6 by using Eco RI and Kpn I sites to generate the pSL6HHZ::gSL6HHZ vector. All the binary plasmids were transformed into the mentioned rice cultivars using Agrobacterium-mediated transformation. The primers used were listed in table S2. SL6-overexpressing lines in Arabidopsis and tobacco were generated by BioRun Biotech Co. Ltd. (Wuhan, China). All the plant materials mentioned above were cultivated at the CNRRI experimental stations in Hangzhou, China (120.2°E, 30.3°N) from May to October, and the rice plants were grown under natural conditions with an inter-plant spacing of 16 cm × 26 cm, and field management practices were conducted according to routine agricultural methods (65). SL6-overexpressing lines in Arabidopsis and tobacco were grown in a growth chamber with a 16-hour day (24°C)/12-hour night (21°C) photoperiod. Measurements of SL SL was determined as described previously (13). All seeds were harvested at maturity and dried at 42°C for 1 week to ∼13% moisture content to break seed dormancy (66). The GR of all lines reached more than 90% after dormancy release. For the AA treatment, the seeds were treated at 42°C and 88% relative humidity in a closed aging chamber (Model: LH-150S, TopCloud-agri Tech, Hangzhou, China). Triple biological repeats (each replicate contains 200 seeds) of each sample were used for the AA treatment, collected at the indicated time points, and then desiccated at room temperature for 2 days before germination assays. For the NA treatment, the seeds were stored under natural conditions (27.5° to 28.5°C, 65 to 70% relative humidity) and harvested at regular intervals. After the aging treatments described above, the seeds were evaluated for germination ability as previously described (67). Brief, 30 seeds per replicate were sown on two layers of filter paper and imbibed in petri dishes (diameter of 9 cm) with 10 ml of distilled water refreshed every day and placed in a growth chamber with a 12-hour day (30°C)/12-hour night (26°C) photoperiod with 60% relative humidity for 7 days to reach the maximum GR of the wild-type seeds under mock condition. Germination ability was recorded daily, and coleoptile growth to 5 mm in length was defined as complete germination (67). The AA treatments of Arabidopsis and tobacco seeds were performed as described previously (68, 69). Briefly, the seeds of Arabidopsis were treated at 42°C and 85% relative humidity, and the seeds of tobacco were placed in petri dishes (diameter of 9 cm) with 10 ml of distilled water and treated at 45°C with 100% relative humidity and then collected over a time course, with the Arabidopsis seeds desiccated at room temperature for 24 hours. The tobacco seeds were dried at 25°C to ∼4.5% moisture content for AA treatment. One hundred surface-sterilized seeds per replicate were sown on half-strength Murashige and Skoog medium (pH 5.8) containing 0.3% plant agar and placed in a growth chamber with a 16-hour day (24°C)/12-hour night (21°C) photoperiod. The seed GRs and cotyledon green rate were recorded daily. GI was defined as follows: GI = ∑(Gt/t), where Gt is the number of germinated seeds on day t, and SVI was defined as follows: SVI = GI × S, where S represents the dry weight of 10 seedlings (66). The time to reach 50% germination under artificial or NA treatment was defined as G50A or G50N and calculated from the linear regression curve of seed GR versus aging treatment time. The MDA and H2O2 content of the brown rice was detected using an MDA Content Assay Kit or an H2O2 Content Assay Kit (Solarbio, Beijing, China) according to the manufacturer’s instructions. Quantification of FAs in brown rice Identification and quantification of fatty acids in brown rice were performed on a Gas Chromatography-Mass Spectrometry (GC-MS) platform at Sanshu Biotech Co. Ltd. (Nantong, China) as described previously with a few modifications (70). Briefly, 1 g of brown rice was added to 4 ml of chloroform and 0.9% NaCl solution, and the mixture was mixed. After centrifuging at 3500 rpm for 15 min at room temperature, the lower layer of liquid was mixed with 2 ml of dichloromethane. After centrifuging at 3500 rpm for 15 min, the bottom layer was dried with nitrogen, then treated with 2 ml of methanol (containing 5% sulfuric acid), and heated to 80°C for 2 hours in a water bath. Then, the cooling mixture was added with 2 ml of hexane and 1 ml of distilled water. After centrifuging at 2000 rpm for 5 min, the supernatant was dried under nitrogen and dissolved in isooctane. The fatty acid content was determined using an Agilent GC/MS system 7820 (Agilent, California, USA) according to the manufacturer’s instructions. The fatty acid content was acquired using MassHunter GC/MS Acquisition and quantified using Quant-My-Way software (Agilent, California, USA). The oil contained all identified fatty acids. RNA isolation, RT-qPCR and GUS staining Total RNA of all tissues except developing seeds was extracted using TRIzol (Invitrogen, Carlsbad, USA) according to the manufacturer’s instructions. The RNA from the developing seed was isolated using a modified SDS-TRIzol method as described previously (71). A 2 μg of total RNA was reverse-transcribed to cDNA using MMLV reverse transcriptase (Clontech, CA, USA). RT-qPCR was performed using SYBR qPCR Mix (Toyobo, Shanghai, China) and a CFX96 real-time PCR detection system (Bio-Rad, Hercules, USA). The relative transcription of each gene was normalized to Ubiquitin (LOC_Os03g13170) in rice, AtUBC9 (AT4G27960) in Arabidopsis, and NbGAPDH in tobacco (72), and the LUC mRNA expression was normalized to REN (73) using the 2-ΔΔCT method (74). The primers used were listed in table S2. 2.5 kb of the SL6JZ promoter was cloned into the binary vector pCAMBIA1300-GUS by using Hind III and Kpn I sites, and the recombinant vector was introduced into Nipponbare via Agrobacterium-mediated transformation. Different tissues of the transgenic plants were harvested and incubated in a solution containing 50 mM sodium phosphate buffer (pH 7.0), 5 mM K3Fe(CN)6, 5 mM K4Fe(CN)6, 0.1% Triton X-100, and 1 mM X-Gluc at 37°C overnight. Images were observed under a stereo microscope SZX16 (Olympus, Tokyo, Japan) after destaining. Measurements of mRNA half-life The experiments were performed as described previously (75). Two-week-old seedlings were treated with distilled water or 1 mM cordycepin (3′ deoxyadenosine, a transcriptional inhibitor) (Aladdin, Shanghai, China), and the leaves were harvested over a time course after the treatments. Total RNAs extracted from the samples were applied for qRT-PCR analysis. The mRNA half-life was calculated from the linear regression curve of SL6 transcript levels plotted against the indicated time points, with or without cordycepin treatment. Electrophoretic mobility shift assay Recombinant GST-VP1-B3 (530 to 640 amino acids containing the B3 domain of VP1) was purified from Transetta (DE3) chemically competent cells (Transgen, Beijing, China). The EMSA probes, 60 bp in length and spanning the SL6 promoter and genomic regions, were commercially synthesized by SUNYA biotech (Hangzhou, China) and labeled using an EMSA Probe Biotin Labeling Kit (Beyotime, Shanghai, China). The probes used were listed in table S2. Equal amounts of recombinant proteins and 20-fmol labeled probes with or without nonlabeled competitive probes were incubated with EMSA/Gel shift binding buffer at 25°C for 20 min and resolved on 6% acrylamide gels. The labeled DNA probes were detected using the LightShift Chemiluminescent EMSA Kit (Thermo Fisher Scientific, Waltham, USA). ChIP-qPCR assay A ChIP-qPCR assay was carried out as described previously (57). A 2 kb of the VP1 promoter was cloned into the binary vector 2300-GFP by using Eco RI and Xba I sites, and the CDS of VP1 was inserted after the VP1 promoter by using Xba I and Sal I sites to generate the pVP1::VP1-GFP vector. The primers used were listed in table S2. The recombinant vector was transformed into Nipponbare based on Agrobacterium-mediated transformation. Approximately 2 g of leaves from transgenic plants were cross-linked with formaldehyde to isolate chromatin. Then, the isolated chromatin was sonicated to produce DNA fragments ranging from 200 to 700 bp. The DNA/protein complex was immunoprecipitated with anti-GFP (Abmart, Shanghai, China), and the immunoprecipitated DNA was purified with phenol/chloroform (1:1, v/v) after reverse cross-linking and proteinase K treatment. qPCR was performed using immunoprecipitated and input DNA with gene-specific primers, and results were calculated according to the manual for the Magna ChIP HiSens Kit (Millipore, Darmstadt, Germany). Luciferase transient transcriptional activity assay The experiments were performed as described previously (76). The CDS of VP1 was cloned into the “None” vector as an effector by using Xba I and Eco RI sites, and different forms of SL6 promoter were amplified from HHZ and JZ1560 and cloned into the 190fLUC (firefly LUC) vectors as reporters by using Hind III and Bgl II sites, respectively. The primers used were listed in table S2. Then, the effector, reporter, and AtUbi3:rLUC (Renilla LUC, an internal control) were cotransformed into rice protoplasts and incubated for 12 hours at room temperature in the dark. The relative luciferase activities (fLUC/rLUC) were detected using the Dual Luciferase Reporter Gene Assay Kit (Beyotime, Shanghai, China) and the Tecan infinite M200 system (Tecan, Mannedorf, Switzerland). Yeast one-hybrid assays The Y1H assays were performed using the Matchmaker One-Hybrid System (Clontech, CA, USA). The CDS of VP1 was cloned into the pB42AD vector by using the Eco RI site, and different truncation forms of the SL6 promoter were cloned into the pLacZi vector by using Eco RI and Kpn I sites. The recombinant constructs were cotransformed into the yeast strain EGY48 and detected on selective media lacking tryptophan and uracil, with 1× BU salts, X-Gal (80 mg/liter; 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside), 2% galactose, and 1% raffinose added. β-Galactosidase activity in the Y1H assay was measured as reported previously (77). Briefly, transformed yeast was cultured with SD/-Trp/-Ura solution to OD600 ≈ 0.6. Total proteins were extracted to detect the β-galactosidase activity with the reaction buffer (40 mM Na2HPO4, 60 mM NaH2PO4, 1 mM MgSO4, 10 mM KCl, and 50 mM β-mercaptoethanol), ortho-nitrophenyl-β-d-galactopyranoside (4 mg/ml) substrate, and stopping buffer (1.5 M CaCO3). Protein transient expression assay For the subcellular localization of SL6, the CDS of SL6 was amplified (excluding the stop codons) and cloned into the pCAMBIA1300-35S-GFP vector, by using Kpn I and Bam HI sites. The primers used were listed in table S2. The recombinant vector was transformed into rice protoplasts for 12 hours or into Nicotiana benthamiana leaves for 2 days using Agrobacterium-mediated transformation, respectively, and fluorescence was captured with a Zeiss LSM710 confocal laser scanning microscope (Carl Zeiss AG, Jena, Germany). For the luciferase activity assay on the SL6 promoters, different truncation or mutated forms of the SL6 promoter were amplified from HHZ and JZ1560 and cloned into the pGreenII 0800-LUC vector, respectively. The recombinant vectors were transformed into the Agrobacterium strain GV3101 and coinfiltrated into N. benthamiana leaves. The fluorescence was detected using luciferin (1 mM luciferin and 0.01% Triton X-100) substrate and the Chemiluminescence Imaging system (Tanon, Shanghai, China) at 2 days after infiltration. The fluorescence intensities were quantified using TanonImage software according to the manufacturer’s instructions. Natural variation analysis for SL6 The primers were designed to identify the 62-bp indel in the 5′UTR of SL6 across 34 wild rice germplasms, 124 core rice accessions, and 934 modern cultivars. Information on the primers and the rice varieties used is listed in tables S1, S2, and S4. The frequency of genotypes at the S1 variation sites for SL6 was determined across the 3 K Rice Genome Project database (78) (https://brs-snpseek.duckdns.org/) and RiceVarMap v2.0 (79) (https://ricevarmap.ncpgr.cn/). The π and Tajima’s D in the 2-Mb region flanking SL6 of different subpopulations were calculated using Cygwin and vcftools (version 3.0). Quantification and statistical analysis Data were analyzed with GraphPad Prism 10.6.1 software (www.graphpad.com/scientific-software/prism/) to compare differences. Values in the figures are means ± SD with biological triplicates unless otherwise noted. Multiple t test was used to detect the least significant difference at P < 0.05.