Genome-wide association study identifies a gene responsible for temperature-dependent rice germination

Rice is cultivated across Japan’s diverse climates (latitudes 20–45°), necessitating genetic adaptation to local environments. Prior work showed regionally enriched alleles controlling heading time and cold tolerance (e.g., COLD1), implying temperature-driven selection. Seed germination is highly temperature-sensitive, proceeding within a few days at 27–37 °C and being disrupted at lower or higher temperatures. This study asks which genetic factors underlie temperature-dependent variation in rice seed germination and whether genotype-by-environment (G × E) interactions can reveal causal genes. The authors apply a G × E GWAS in Japanese japonica cultivars to dissect genetic control of germination across contrasting temperatures (15 °C vs 30 °C), aiming to identify mechanisms and alleles contributing to geographic adaptation.

• 14-3-3 proteins are conserved eukaryotic regulators that bind phosphorylated targets, modulating activity, stability, localization, and interactions. Plants possess expanded 14-3-3 families (Arabidopsis: 15; rice: 8), grouped into ε and non-ε types, and implicated in development, stress, and hormone signalling, including ABA.
• In rice, low-temperature germination is strongly influenced by the QTL qLTG3-1, with multiple functional alleles (partial loss-of-function, loss-of-function, gain-of-function). Multi-allelic loci can reduce power in standard (bi-allelic) GWAS.
• G × E frameworks have been recommended to uncover context-specific genetic effects but gene-level dissection has been rare. Prior studies in Arabidopsis indicated G × E effects on dormancy and germination plasticity without isolating causal genes.
• Transcription factors OREB1/OsABI5 and TRAB1/OsbZIP66 mediate ABA-responsive transcription that inhibits germination and mediates stress responses. MFT family members regulate germination in cereals and Arabidopsis, with context-dependent roles in ABA signalling. Florigen complexes (FAC: 14-3-3, bZIP FD, FT-like) illustrate how 14-3-3–bZIP modules control developmental transitions.

• Plant materials and phenotyping: A panel of 164 Japanese Oryza sativa ssp. japonica varieties was grown and seeds harvested 45 days after flowering. After dormancy release (45 °C, 3 days), germination was assayed under dark conditions at 30 °C for 24 h and 15 °C for 96 h. Germination was scored when the epiblast was broken and the embryo protruded; rates were normalized to germination at 30 °C for 48 h.
• Sequencing and variant calling: Genomic DNA libraries (~500 bp inserts) were sequenced (Illumina HiSeq, 100–150 bp paired-end). Reads were aligned to Nipponbare IRGSP-1.0 (ver.7) using BWA-MEM; BAMs were processed with SAMtools. Variants were called with GATK HaplotypeCaller (gVCF mode), joint genotyped, then filtered (heterozygotes set to missing; MAF ≥5%; mapping quality ≥40; minimum count threshold 40%).
• Addressing multi-allelic variants: Tri-allelic sites were transformed into three bi-allelic lines with an in-house script to enable association testing (e.g., for qLTG3-1 functional alleles).
• Heritability: Narrow-sense heritability was estimated via rrBLUP (A.mat to compute GRM; mixed.solve to partition σg² and σe²).
• GWAS models: • Single-environment GWAS at 15 °C and 30 °C used rrBLUP’s LMM (n.PC=5) with kinship correction. Significance thresholds were Bonferroni-corrected. • G × E GWAS used a linear mixed model tailored to two environments (15 vs 30 °C), modeling fixed effects of environment and PCs, marker main effects, and random genetic and G × E effects (Hadamard product structure for G × E). Marker-specific G × E effects were tested by LRT (1 d.f.).
• LD and candidate gene prioritization: LD heatmaps defined intervals around significant peaks. Polymorphisms causing amino-acid changes or frameshifts were prioritized; genes annotated as retrotransposons or hypothetical proteins were excluded.
• Functional validation and mechanistic assays: • Expression pattern: Promoter (Hap.2)::GUS transgenics examined organs/tissues. • Complementation: Nipponbare (NPB; GF14h Hap.1, presumed LOF) transformed with GF14h Hap.2 genomic fragment or pUBQ::GF14h Hap.2 CDS; germination tested at 15 and 30 °C. • ABA responsiveness: Germination assays across ABA concentrations; qRT-PCR of ABA-responsive genes (OsRab16A, OsLea3, OsEM) in imbibed seeds at 30 °C. • Protein interactions: Yeast two-hybrid (Y2H) for GF14h (Hap.1/Hap.2) with OREB1 and TRAB1; subcellular localization of GFP fusions in rice mesophyll protoplasts; BiFC to test pairwise and ternary interactions among GF14h, OREB1 (WT, S385A, S385E), and MFT2; co-immunoprecipitation (co-IP) in protoplasts. • Transcriptional reporter: Transient protoplast assay using OsEM promoter::LUC reporter; effectors OREB1 (±ABA), GF14h (Hap.1/Hap.2), and MFT2 tested; OREB1 phosphorylation mutants assessed. • Gene editing: CRISPR/Cas9 knockout of MFT2 in NPB; ABA sensitivity of germination evaluated.
• Population genetics: Haplotype discovery for GF14h in ~3010 cultivated rice accessions (3K panel) and O. rufipogon datasets. Haplotype network (PopART) and geographic distribution analyses identified functional vs LOF haplotypes and their subpopulation/region-specific frequencies. Temporal comparisons were made between Japanese landraces and modern cultivars (pre- and post-1990) in northern vs southern regions; analogous analyses in Chinese temperate japonica were performed.

• G × E GWAS identified a significant locus (Peak 1) at ~23.5 Mb on chromosome 11 associated with temperature-dependent germination. In single-environment GWAS, Peak 1 was significant at 30 °C but not at 15 °C (where its signal dropped to ~−log10 P ≈ 2.6), indicating environment-specific effect.
• Heritability: Narrow-sense heritability of germination rate was 51.5% (30 °C) and 53.2% (15 °C).
• A 4 bp coding-region InDel in LOC_Os11g39540 (GF14h), a grass-specific ε-type 14-3-3 protein, was prioritized. Nipponbare (Hap.1) carries the 4 bp deletion removing a conserved C-terminal region, consistent with a loss-of-function (LOF) allele; Hap.2/Hap.3 sequences are conserved across orthologs and functional.
• Phenotypic effect of GF14h haplotypes: Varieties with GF14h Hap.1 (LOF) exhibited significantly lower germination rates than Hap.2/3 at 30 °C, with little difference at 15 °C. Complementation of NPB (Hap.1) with GF14h Hap.2 (genome fragment or pUBQ::CDS) increased germination at 30 °C but had minimal effect at 15 °C, confirming a G × E effect.
• ABA signalling mechanism: • Functional GF14h (Hap.2) reduced ABA sensitivity of germination and lowered expression of ABA-responsive genes (OsRab16A, OsLea3, OsEM) during imbibition at 30 °C. • GF14h (Hap.2), but not Hap.1, interacted with OREB1 in Y2H; localization studies showed GF14h Hap.2 in nucleus/cytosol and GF14h Hap.1 mainly at ER; GFP-OREB1 localized to nucleus. BiFC confirmed nuclear interaction for GF14h Hap.2–OREB1. • OREB1 S385 phosphorylation status modulated interactions: S385A weakened interaction and transcriptional effects; S385E behaved similar to WT, supporting a phosphorylation-dependent complex akin to the flowering activation complex. • MFT2 localized to nucleus/cytosol, contributed to ABA-mediated germination control, and formed a ternary complex with GF14h–OREB1 in nuclei. Presence of OREB1 relocalized GF14h–MFT2 BiFC signals to the nucleus; S385A impeded this relocalization. • In OsEM::LUC reporter assays, OREB1 activated transcription (enhanced by ABA). GF14h Hap.2 partially suppressed OREB1 activation in a dose-dependent manner, whereas MFT2 alleviated GF14h-mediated suppression in a dose-dependent manner; effects required OREB1 phosphorylation (abolished with S385A). GF14h Hap.1 failed to repress OREB1.
• qLTG3-1 detection required tri-allelic-to-bi-allelic transformation; it was strongly associated at 15 °C but did not appear as a G × E locus in this design, consistent with its role in low-temperature germination rather than environment interaction across these two conditions.
• Population genetics and breeding: • Multiple GF14h LOF haplotypes arose independently and are subpopulation- and region-specific: Hap.1 (4 bp deletion) enriched in temperate japonica in East Asia and Europe (especially East Asia), Hap.6 (1 bp insertion) in indica in Indochina, Hap.7 (genome deletion) dominant in China, Hap.8 (genome deletion) in Indonesia. O. rufipogon mostly carries functional haplotypes (rare Hap.7 exception), indicating LOFs emerged during/after domestication. • In Japanese breeding, GF14h shifted toward higher LOF frequency from landraces to modern cultivars; the shift is more pronounced in northern regions. qLTG3 similarly transitioned from GOF to LOF, especially rapidly in northern Japan; Chinese temperate japonica shows the same trend. • Interpretation: LOF alleles (slower germination via enhanced ABA signalling) likely selected to reduce pre-harvest sprouting risk despite potential benefits of rapid germination under cool conditions.

The study identifies GF14h as a causal gene underlying temperature-dependent variation in rice seed germination via a G × E GWAS framework. GF14h acts predominantly at 30 °C, consistent with a model where at low temperature (15 °C) elevated ABA levels and OREB1 abundance mask GF14h’s modulatory effects, while at optimal temperature lower ABA/OREB1 permits GF14h to suppress ABA signalling and promote germination. Mechanistically, GF14h (ε-type 14-3-3) forms a nuclear complex with OREB1, reducing its transactivation of ABA-responsive genes; MFT2 joins to form a ternary complex that relieves GF14h-mediated suppression, fine-tuning ABA responses. This module parallels the flowering activation complex in composition (14-3-3–bZIP–FT-like) yet differs in interaction properties and regulatory consequences. The findings reconcile conflicting literature by highlighting isoform- and context-specific roles of 14-3-3s in ABA signalling. Population analyses show independent emergence and regional enrichment of GF14h LOF haplotypes, mirroring breeding priorities: in northern regions and modern varieties, selection favored slower germination (stronger ABA signalling) to mitigate pre-harvest sprouting risk. Together with the reshaping of qLTG3-1 alleles, the results illuminate how breeders leveraged standing variation in germination pathways to expand rice cultivation into cooler climates while managing agronomic trade-offs.

This work demonstrates that a 4 bp InDel in GF14h (ε-type 14-3-3) underlies genotype-by-environment variation in rice germination, acting mainly at optimal temperature by modulating ABA signalling through a GF14h–OREB1–MFT2 transcriptional complex. Functional validation and molecular assays establish GF14h as a negative regulator of ABA signalling that promotes germination, while MFT2 counterbalances this suppression in a phosphorylation-dependent manner. Population genetics reveals multiple independent GF14h LOF alleles and their geographic/temporal enrichment, indicating selection for slower germination to reduce pre-harvest sprouting risk in northern and modern cultivars. Future directions include: dissecting additional components and co-regulators of the GF14h–OREB1–MFT2 module; clarifying roles of non-ε 14-3-3 isoforms in ABA responses; expanding G × E analyses across more environmental gradients (moisture, light, diverse temperatures); integrating pyramiding strategies that balance rapid seedling establishment with sprouting resistance.

• Environmental scope was limited to two temperature regimes (15 °C and 30 °C) under controlled dark conditions; broader environmental factors (e.g., fluctuating temperatures, light, soil conditions) were not tested.
• The GWAS panel comprised 164 Japanese japonica varieties, which may limit generalizability to other subpopulations (indica, aus) without further validation.
• Some protein–protein interactions showed discrepancies between yeast and plant cell assays, suggesting additional plant-specific factors influence complex formation.
• qLTG3-1 required special handling of multi-allelic variants for detection; other multi-allelic loci might remain underpowered or undetected.
• The mechanistic model emphasizes ABA and OREB1; direct in vivo evidence for all proposed co-regulators and genome-wide target effects under field conditions remains to be established.