A mediator of OsbZIP46 deactivation and degradation negatively regulates seed dormancy in rice

Preharvest sprouting (PHS) is a deleterious phenotype that occurs frequently in rice-growing regions where the temperature and precipitation are high. It negatively affects yield, quality, and downstream grain processing. Seed dormancy is a trait related to PHS. Longer seed dormancy is preferred for rice production as it can prevent PHS. Here, we map QTLs associated with rice seed dormancy and clone Seed Dormancy 3.1 (SDR3.1) underlying one major QTL. SDR3.1 encodes a mediator of OsbZIP46 deactivation and degradation (MODD). We show that SDR3.1 negatively regulates seed dormancy by inhibiting the transcriptional activity of ABIs. In addition, we reveal two critical amino acids of SDR3.1 that are critical for the differences in seed dormancy between the Xian/indica and Geng/japonica cultivars. Further, SDR3.1 has been artificially selected during rice domestication. We propose a two-line model for the process of rice seed dormancy domestication from wild rice to modern cultivars. We believe the candidate gene and germplasm studied in this study would be beneficial for the genetic improvement of rice seed dormancy.

Crop domestication is necessary and fundamental for historical development.The food surplus brought about by domestication has promoted the development of human civilization through urbanization and the evolution of the nation state 1 .In turn, the diverse cultural preferences of different historical periods have determined the direction of crop domestication that can meet the particular needs of humans [2][3][4][5] .Rice is one of the oldest domesticated crops, and is cultivated approximately 11,000 years ago 6 .Over time, rice has become the main source of energy for more than half of the world's population 7 .During the domestication of wild rice, many physiological traits and morphological features were considerably altered.These transformations include changes in awn development [8][9][10] , seed shattering [11][12][13] , grain quality 6 , plant architecture [14][15][16][17] , grain size 18 , heading date 19 , panicle size 20 , and seed dormancy 21 .
An important goal of rice domestication is to increase yield to meet demand and overcome unfavorable environmental conditions.Yield is usually affected by PHS, which generally occurs in high-temperature and high-humidity environments in the middle and lower reaches of the Yangtze River.PHS can cause enormous economic losses in crop production, including declines in seed yield and quality 22 .Longer seed dormancy can prevent PHS, thereby reducing losses and increasing yield.Unfortunately, during domestication, rice seed dormancy is largely lost 23 .Therefore, it is necessary to identify rice cultivars that have longer seed dormancy and clone seed dormancy genes to improve rice yield.
Seed dormancy is a complex trait, and many dormancy-related QTLs have been identified in rice; [24][25][26][27][28] however, few of these QTLs have been cloned 21,[29][30][31] .The cloned seed dormancy genes are associated with either abscisic acid (ABA) or gibberellin (GA), two key hormones that regulate seed dormancy 32 .Sdr4 is the first gene cloned from the mapping population constructed from the japonica cultivar Nipponbare and the indica cultivar Kasalath 21 .Sdr4 encodes an unknown function protein, and its regulation of seed dormancy depends on VP1.Only Sdr4-n occurs in japonica cultivars, whereas both Sdr4-n and Sdr4k occur in indica cultivars, which seems to indicate that Sdr4 is involved in the rice seed dormancy domestication 21 .qSD7-1 is cloned from the weedy rice cultivar SS18-2 and encodes a basic helix-loop-helix protein.
It increases the ABA content in seeds by upregulating the expression of ABA synthesis genes 29 .In addition, qSD7-1 interacts with VP1 and OsC1 to increase the seeds sensitivity to ABA 33 .qSD1-2 is isolated from SS18-2.It regulates seed dormancy via the GA pathway 30 .Recently, SD6, which is cloned from Kasalath, together with ICE2 antagonistically balanced the expression of ABA metabolic genes to control rice seed dormancy by responding to temperature signals 31 .However, if seed dormancy is too persistent, it is difficult to sprout, which increases the cost of timely farming; conversely, if dormancy is too weak, PHS will occur.Thus, it is necessary to appropriately regulate seed dormancy by identifying additional seed dormancy genes.
ABA is an important hormone that regulates seed dormancy.Its signal transduction involves many genes, such as the transcription factors ABI3 and ABI5.These genes positively regulate the transcription factors involved in rice seed dormancy [34][35][36] .Other ABA signaling response factors, such as embryonic abundant protein (EM1) 37 , late embryogenesis abundant protein (LEA3) 38 , and the mediator of OsbZIP46 deactivation and degradation (MODD), are often involved in regulating drought resistance in rice 39 .MODD, which is homologous to the Arabidopsis thaliana ABI5-binding protein AFP, negatively regulates drought tolerance in rice by regulating OsbZIP46 activity and stability 39 .However, whether it regulates rice seed dormancy remains unknown.
In this work, we clone the major QTL qSDR3.1 associated with seed dormancy.We show that the underlying gene SDR3.1,encoding a mediator of OsbZIP46 deactivation and degradation (MODD), represses rice seed dormancy by inhibiting the transcriptional activity of ABIs.Population genetics analysis reveals that SDR3.1 has been intensively selected during domestication and breeding, especially for Geng/japonica cultivars.We demonstrate the usefulness of SDR3.1 by introducing it into a rice restorer line cultivar for hybrid seed production.

Phenotypic characterization of the two parental lines
ZH11 is a japonica cultivar with a high postharvest germination rate (83.00%), whereas Mengjialaxiaoli (MJLXL) is an Aus cultivar with a low post-harvest germination rate (4.11%;Supplementary Fig. 1a, b).The germination rates were measured immediately at 35 days after heading (DAH).Three months after harvest, the germination rates of ZH11 and MJLXL both reached to approximately 80% (Supplementary Fig. 1c, d).These results demonstrate that MJLXL is a cultivar with longer seed dormancy and ZH11 is a cultivar with shorter seed dormancy.Therefore, ZH11 and MJLXL were selected as parental lines to develop the mapping population.

qSDR3.1 is a major QTL that regulates seed dormancy
To determine the genetic mechanism underlying rice seed dormancy, we used ZH11 as the recipient parent and MJLXL as the donor parent to construct a backcross inbred lines population (Fig. 1a).First, we examined the germination rate of 153 lines derived from the BC 3 F 2 population, which were selfbred from a single BC 3 F 1 plant (the genotype of BC 3 F 1 is shown in Supplementary Data 1).The germination rate of this population exhibited a normal distribution, indicating that seed dormancy was controlled by QTLs (Supplementary Fig. 2a).Then, we identified three QTLs of seed dormancy, designated qSDR2.1,qSDR3.1, and qSDR3.2,which were located on the second and third chromosomes, respectively (Supplementary Fig. 2b).qSDR3.1 had the highest logarithm of odds value (11.75) and was mapped between markers 3-15 and 3-21 on chromosome 3 (Supplementary Fig. 2c; Supplementary Table 1).Genetic analysis of the BC 3 F 2 individuals suggested that qSDR3.1 is an incomplete dominant gene (Supplementary Fig. 2d, e).
The six homozygous recombinant lines (selected from 4146 BC 4 F 2 individuals) in the QTL region were analyzed for fine mapping (Fig. 1b).The 7-day germination rate phenotype of the recombinant lines ranged from 20.16% to 72.14%.The 7-day germination rate of R2, R4, and R6 were similar to that of MJLXL, and the 7-day germination rate of R1, R3, and R5 were similar to that of ZH11.Finally, the location of qSDR3.1 was narrowed to a 26.9-kb region between markers Indel27 and Indel29 (Fig. 1b).This region contained 4 candidate genes, namely, LOC_Os03g11520, LOC_Os03g11530, LOC_Os03g11540 and LOC_Os03g11550 (Supplementary Table 2).LOC_Os03g11520 encode a proteins with unknown function.There was no difference between the parents in LOC_Os03g11530 and LOC_Os03g11540.And there were three bases differences between the two parents in LOC_Os03g11550, which created two amino acid differences (Fig. 1c; Supplementary Fig. 3a-b).
In addition, the cDNA of LOC_Os03g11550 of R3 was consistent with that of ZH11, whereas the cDNA of LOC_Os03g11550 of R2 and R4 were consistent with that of MJLXL, indicating that the LOC_Os03g11550 genotypes cosegregated with the seed dormancy phenotypes (Supplementary Fig. 3c-e).According to the Rice Genome Annotation Database (http://rice.uga.edu/,RGAP 7), LOC_Os03g11550 comprises three exons and two introns.It encodes MODD 39 .MODD negatively regulates ABA signaling and drought resistance in rice 39 .Therefore, LOC_Os03g11550 was selected as the candidate gene of SDR3.1.
In rice, PHS frequently leads to severe decreases in grain yield and quality.Enhanced seed dormancy is useful for preventing PHS 40 .To determine whether SDR3.1 regulates PHS, we measured the PHS of ZH11, MJLXL, NIL, and the two mutant lines.On the 6th day, the PHS of NIL and MJLXL significantly decreased by 82.83% and 83.51% compare with that of ZH11, respectively.On the 5th day, the PHS of KO-1 and KO-2 significantly decreased by 48.21% and 47.79% compare with that of ZH11, respectively (Supplementary Fig. 6).Combined with the enhanced seed dormancy in NIL and the two mutant lines (Fig. 1d; Supplementary Fig. 4d, e), indicating that SDR3.1 regulates PHS.
Expression and subcellular localization of SDR3.1 qRT-PCR showed that SDR3.1 is widely expressed in 10-day leaf, glume, anther, 5-day root, 20-day root, and 5-day leaf, with the highest transcript level detected in the pistil (Supplementary Fig. 7a).Then, the dynamic transcript levels of SDR3.1 were investigated during seed development.The transcript began to accumulate 5 days after fertilization and peaked at 15 days (Supplementary Fig. 7a).
To study the biological function of SDR3.1, we fused the coding region to the N-terminus of GFP.The empty GFP vector was used as a control and mCherry (LOC_Os11g01330) was used as a marker.The results showed that the 35 S::SDR3.1-GFP protein was localized in the nucleus (Supplementary Fig. 7b).

SDR3.1 physically interacts with ABI5 and represses its transcriptional activation
To determine whether SDR3.1 is a transcription factor, the recombinant vector pGBKT7-SDR3.1 was transferred into the yeast (Saccharomyces cerevisiae) strain Y2HGold.Cells expressing SDR3.1 protein could not grow on the systematic tri-deficient medium (Fig. 2a), indicating that SDR3.1 was not a transcription factor.
Similarly, we divided ABI5 into two segments; an N-terminus domain (AD-ABI5-N) and a C-terminus bZIP domain (AD-ABI5-C; Supplementary Fig. 8c).Deletion of the C-terminus residue of ABI5 did not affect its interaction with SDR3.1, whereas deletion of the N-terminus fragment abolished the interaction of ABI5 with SDR3.1 (Supplementary Fig. 8d).Therefore, the N-terminus of ABI5 was essential for the SDR3.1-ABI5interaction.
To clarify the function of ABI5, we disrupted it (LOC_Os01g64000) in ZH11 via CRISPR/Cas9, and two homozygous mutants, abi5-1 and abi5-2, were obtained through sequencing.These mutants contained a 1-bp frame-shift insertion or deletion in the first exon (Supplementary Fig. 9a), and the mutants had lower ABI5 expression levels than that of ZH11 (Supplementary Fig. 9b).Moreover, we overexpressed ABI5 in ZH11 (Supplementary Fig. 9c).The results indicated that the mutant lines (97.92% and 98.33%) had higher germination rate than that of ZH11 (88.75%).In contrast, the germination rate of the overexpressed lines (36.67% and 46.67%) were significantly lower than that of ZH11 (88.75%) (Supplementary Fig. 9d-e).These results indicated that ABI5 promoted seed dormancy.
Because SDR3.1 inhibits seed dormancy and ABI5 promotes seed dormancy, we investigated whether the SDR3.1 protein inhibits the transcriptional activation of ABI5.Therefore, identifying ABI5 target genes was necessary.In Arabidopsis, the EMBRYONIC ABUNDANT PROTEIN1 (EM1) is a target of ABI5 41 , but EM1 has not been reported as a target gene of ABI5 in rice.Therefore, we used a yeast one-hybrid (Y1H) assay to confirm that ABI5 could bind to the promoter of EM1 in rice (Supplementary Fig. 10a), and we narrowed the binding region within 200 bp (Supplementary Fig. 10b).In addition, we used an electrophoretic mobility shift assay (EMSA) to confirm the interaction between ABI5 and the promoter of EM1 (Supplementary Fig. 10c).To identify additional downstream target genes of ABI5, we used Cleavage Under Targets and Tagmentation (cut&Tag) technology combined with Y1H and EMSA to verify that ABI5 could bind to the promoter of LEA3 (Supplementary Fig. 10d-f).
Furthermore, we disrupted EM1 (LOC_Os05g28210) and LEA3 (LOC_Os05g46480) in ZH11 via CRISPR/Cas9, and four homozygous mutants, namely, em1-1, em1-2, lea3-1 and lea3-2 were obtained through sequencing.These mutants contained a 32-bp or 28-bp frameshift deletion in the first exon of EM1 and an 89-bp frame-shift deletion or 1-bp frame-shift insertion in the second exon of LEA3 (Supplementary Fig. 11a-b).The em1/lea3 plants had lower gene expression levels than that of ZH11 (Supplementary Fig. 11c).Moreover, we overexpressed EM1 and LEA3 in ZH11, respectively (Supplementary Fig. 11c).The mutant lines had higher germination rate than that of ZH11.In contrast, the germination rate of the overexpressed lines were significantly lower than that of ZH11 (Supplementary Fig. 11d-e).These results indicated that EM1 and LEA3 are important downstream genes involved in the seed dormancy regulation pathway.
To verify the relationship between SDR3.1 and ABI5, we constructed a double mutant in ZH11 (Supplementary Fig. 12a-b).The germination rate of the double mutants had no significantly difference compare with that of ZH11 (Supplementary Fig. 12c-d).These results further verified that SDR3.1 repressed the transcriptional activity of ABI5.

SDR3.1 negatively regulates ABA signaling
Because ABI5 is the central factor corresponding to ABA signaling, we examined the expression levels of ABA response genes in ZH11, MJLXL and transgenic plants.In addition to OsEM1 and OsLEA3, we also detected the expression levels of three genes that are related to ABA regulation (OsbZIP72, OsRab16A and OsTRAB1).The results showed that the expressions were suppressed significantly by SDR3.1 (Fig. 3a-e).However, the expression levels of OsbZIP72, OsRab16A and OsTRAB1 were not activated by ABI5 (Fig. 3c-e).Therefore, SDR3.1 might regulate the response to ABA signals through more than just ABI5.Based on these findings, we tested whether SDR3.1 inhibits another seed dormancy-related transcription factor, OsABI3.The results showed that SDR3.1 inhibited the transcriptional activity of OsABI3 (Supplementary Fig. 13a-b).
In addition, we examined the ABA and GA contents in seeds, and the results showed that SDR3.1 regulated seed dormancy not by regulating ABA and GA contents (Fig. 3f; Supplementary Fig. 13c-f).Then, we examined the sensitivity of ZH11 and the two mutant lines to ABA after breaking seed dormancy.As expected, under normal conditions, there was no significant differences in germination rate between ZH11 and the two mutant lines (Fig. 3g-i).However, after 5 μM or 10 μM ABA treatment, the germination rate of the two mutant lines were significantly lower than that of ZH11 (Fig. 3g-i).Therefore, the two mutant lines were more sensitive to ABA.The above results indicated that SDR3.1 negatively regulated ABA signaling by repressing ABIs.
Due to the similar germination rate between Hap1 and Hap2, as well as between Hap3 and Hap4, we speculated that different SNPs may have caused synonymous mutations.We subsequently analyzed the differences in amino acids at the five different SNP sites.We found only two of five SNPs (SNP-3399 and -305 in Fig. 4b) changed the amino acids; the other three SNPs (SNP-3591, -348 and -75 in Fig. 4b) did not change the amino acids.Therefore, the amino acid sequences of SDR3.1 of Hap1 and Hap2 were the same (Ser 102 +Asp 156 ), the amino acid sequences of SDR3.1 of Hap3 and Hap4 were also the same (Asn 102 +Glu 156 ), and Hap5 was Ser 102 +Glu 156 .Among the combinations tested, Asn 102 +Glu 156 was associated with the highest germination rate (64.13%), followed by Ser 102 +Asp 156 (27.72%), and Ser 102 +Glu 156 was associated with the lowest germination rate (4.95%) (Fig. 4e).The combination of Asn 102 +Glu 156 was mostly appear in the Geng/japonica cultivar (85.19%), the combination of Ser 102 +Asp 156 was mostly appear in the Xian/indica cultivar (53.95%), and the combination of Ser 102 +Glu 156 was mostly appear in the Aus cultivar (98.28%) (Fig. 4f).These results implied that Aus had longest seed dormancy, followed by Xian/indica, while Geng/japonica had the shortest seed dormancy.The genetic effects of different alleles are strongly affected by the population structure as Asian cultivated rice exhibit dramatic differences in Xian/indica and Geng/japonica rice cultivars.Therefore, we compared seed dormancy between the Xian/indica and Geng/japonica cultivars in both Hap1 and Hap2 with Ser 102 +Asp 156 , and the results showed that the germination rate of Geng/japonica was significantly higher than that of Xian/indica (Fig. 4g), indicating that the genetic effect of dormancy was influenced by population structure.
To investigate whether the effect of SDR3.1 on seed dormancy was related to its expression levels, we measured its expression levels in some rice cultivars from each haplotype.There was no significant correlation between the germination rate and the expression levels of each haplotype (Supplementary Fig. 21a-e).There were also no significant differences in expression levels among the five haplotypes, except between haplotype 4 and haplotype 5 or between haplotype 1 and haplotype 4 (Supplementary Fig. 21f).These results indicated that the regulation of seed dormancy by SDR3.1 depended on its protein activity more than its expression levels.
Conversion of longer dormancy in wild rice to shorter dormancy in Xian/indica and Geng/japonica requires SDR3.1 aa 102 and aa 156 To clarify the evolution of SDR3.1 in rice, we analyzed the amino acid sequences of the 4 wild rice accessions (Supplementary Fig. 14).The amino acid sequences at positions 102 and 156 were Ser and Glu, respectively, which was consistent with the sequence of Hap5 (Fig. 5a).We subsequently sequenced 28 wild rice accessions and found that most of the amino acid sequences at positions 102 and 156 were Ser and Glu (Fig. 5b).The absolute Tajima values for Xian/indica and Geng/ japonica were both greater than 2, their π values decreased sharply around SDR3.1, and Geng/japonica's π values were lower than that of Xian/indica (Fig. 5c-d; Supplementary Fig. 22).These findings suggested that SDR3.1 was selected during rice domestication and was more strongly selected in Geng/japonica.
Based on the varying degrees of selective effects between the Xian/indica and Geng/japonica cultivars, which may have an impact on the germination rate.We investigated the germination rate of some cultivars among different rice subspecies.The results showed that the germination rate was as follows: Geng/japonica (such as TD-156, TD-187, TD-198, and TD-229) > Xian/indica (such as TD-45, TD-13, TD-92, and TD-104) > Aus (such as TD-238 and TD-256) (Supplementary Fig. 23).To determine the molecular functions of aa 102 and aa 156 , we constructed effectors of five haplotypes (Supplementary Fig. 24a), cotransformed rice protoplasts with ABI5, and detected the LUC activity of EM1 and LEA3.The results showed that Hap3 and Hap4 had the strongest inhibitory effects on ABI5 transcriptional activity, Hap1 and Hap2 had the second strongest inhibitory effects, and Hap5 had the weakest inhibitory effect (Supplementary Fig. 24b).Furthermore, we constructed a site-directed mutation mutant containing the aa 102 or aa 156 substitutions; namely, we changed the asparagine acid (N) to the serine acid (S) at site 102 in ZH11 and changed the aspartic acid (D) to glutamic acid (E) at site 156 in YK17.The results showed that the germination rate of the mutants were significantly lower than that of ZH11 or YK17 (Fig. 5e-f).Taken together, these results indicated that the conversion of the longer seed dormancy of wild rice to the shorter seed dormancy of Xian/indica and Geng/japonica required aa 102 or aa 156 : a change from Ser 102 to Asn 102 to produce Geng/japonica and a change from Glu 156 to Asp 156 to produce Xian/indica (Fig. 5g).

Seed dormancy improvement of the high-quality rice restorer line
Zhonghui261 (ZH261) is a high-quality restorer line, but its seed dormancy persistence needs to be further improved.Therefore, we introduced SDR3.1 from MJLXL into ZH261.We obtained BC 1 F 1 populations with 3565 plants by hybridizing MJLXL with ZH261 (Fig. 6a).Then, the plants that were heterozygous for SDR3.1 were selected by molecular markers detection, and the single plant named BC 1 F 1 -80 with the highest degree of genetic background reversion was selected by sequencing with a rice 40 K gene chip.The progeny of the BC 1 F 1 -80 selfing were marker detected to select lines with homozygous alleles of MJLXL at SDR3.1, and the single plant named BC 1 F 2 -39 with the highest degree of genetic background reversion was subsequently selected by sequencing with a rice 40 K gene chip (Fig. 6b).BC 1 F 2 -39 exhibited 88.78% genetic similarity to ZH261 (Supplementary Fig. 25a-c).As expected, BC 1 F 2 -39 had significantly lower germination rate than that of ZH261 (Fig. 6d-e), whereas the 1000-grain weight, panicle length, grain length, and grain width (Fig. 6c, f) had no significantly differences than those of ZH261.The seed dormancy of the other two sister lines also increased significantly, indicating that SDR3.1 indeed increased the seed dormancy of the improved lines (Supplementary Fig. 25d-i).

Discussion
Rice germplasm resources are a diverse collection of rice varieties that have been conserved for their genetic diversity and potential utility in breeding programs.The seed hull gene Bh4 43 , the awn development gene An-1 8 , and the plant architecture gene PROG1 14 were cloned from wild rice, and the hypoxia tolerance gene GF14h was recently cloned from natural weedy rice 44,45 .However, rice seed dormancy gene cloning is rare 21,46 .In this study, we cloned a major QTL, SDR3.1, from the longer seed dormancy cultivar MJLXL.This finding suggested that cloning elite genes by using rice germplasm resources is feasible.During the process of cloning SDR3.1, we found that the germination rate of COM-1 was significantly higher than that of COM-2 and COM-3 (Fig. 1d).The copy number results showed that the copy number of COM-1 was almost twice that of COM-2 and COM-3 (Fig. 1e), and copy number variation affects gene function 47 .Therefore, we speculated that the significant differences in the germination rate among COM-1, COM-2 and COM-3 is due to copy number variation.
Seed dormancy and germination are regulated by internal and external signals 48 .Among these signals, the hormone ABA is an important regulator that inhibits seed germination 32 .Consistent with this, most seed dormancy genes are involved in ABA synthesis and ABA signal transduction 21,33,46,49 .qSDR3.1 participates in the ABA signal transduction pathway to regulate seed dormancy through interaction with ABI5 (Fig. 2b-d).ABI5 is a key regulator of ABA signal transduction 50 .The abi5 mutant has been found to be insensitive to ABA during germination, whereas plants that overexpress ABI5 are hypersensitive to ABA and exhibit enhanced seed dormancy 34,35 .GA 51 , sugar 35 , cytokinin 52 , the brassinolide (BR) signaling factor BES1 53 and the jasmonic acid (JA) signaling inhibitor JAZ 54,55 can antagonize the effect of ABA by inhibiting the activity of ABI5 or degrading the ABI5 protein to promote seed germination or postgerminative growth.Conversely, the circadian clock gene PRR5 41 and the BR signaling inhibitor BIN2 56 inhibited seed germination by promoting the expression of ABI5 and maintaining its protein stability.Moreover, ABI5 regulates reactive oxygen species homeostasis by activating CATALASE1 transcription during seed germination 57 .In the present study, knocking out ABI5 reduced rice seed dormancy, while overexpressing ABI5 enhanced rice seed dormancy (Supplementary Fig. 9d-e), indicating that ABI5 positively regulates rice seed dormancy.SDR3.1 inhibited the transcriptional activity of ABI5 (Fig. 2f-g), thus blocking the signal transduction of ABA, which was not conducive to seed dormancy, and knocking out SDR3.1 enhanced dormancy.
Interestingly, gene expression experiments showed that SDR3.1 regulated more ABA signaling genes than did ABI5 (Fig. 3c-e), suggesting that SDR3.1 may not only affect seed dormancy through ABI5.ABI3 is another important transcription factor that regulates ABA signaling to affect seed dormancy 36 , and the inhibition of ABI3 transcriptional activity by SDR3.1 was confirmed in this study (Supplementary Fig. 13a-b).Therefore, SDR3.seed dormancy by inhibiting the transcriptional activity of ABIs (Supplementary Fig. 26).In addition, the ABA content in MJLXL was significantly higher than that in ZH11 (Fig. 3f), indicating that the seed dormancy of MJLXL is influenced not only by ABA signaling but also by ABA content.This partly explains why MJLXL seeds hardly germinated on the 7th day (Supplementary Fig. 1a-b), but the germination rate of KO-1 and KO-1 plants reached 10-20% (Supplementary Fig. 4d-e).The seed dormancy mechanism of MJLXL and the molecular mechanism through which SDR3.1 regulates seed dormancy involve a more complex network, which requires further in-depth research.Domestication is a process of natural or artificial selection.After 10,000 years of selection, crops such as rice, wheat, corn, soybean, and sunflower have been successfully brought to peoples' tables 58 .Identification of domestication-related genes helps us understand the process of domestication.Nucleotide polymorphism analysis revealed that the π values of An-1 and OsNCED3 were reduced in cultivated rice, suggesting that these two genes were subject to artificial domestication 8,49 .In the present study, the π value for SDR3.1 was reduced, suggesting that SDR3.1 was subject to artificial domestication.Furthermore, the π value of Geng/japonica was the lowest, followed by that of Xian/indica, and the π value of Aus was the highest (Fig. 5d), suggesting that SDR3.1 was subjected to stronger selection in Geng/ japonica, which lost most of its seed dormancy.
The search for functional nucleotide polymorphisms (FNPs) became increasingly relevant as single-base substitution technologies matured 59 .PAO5 -578G , an FNP in the promoter region, reduced the expression level and promoted mesocotyl elongation, which was beneficial for direct seeding of rice 60 .In the present study, haplotype analysis and natural population germination experiments showed that two FNPs caused two amino acid changes in SDR3.1 protein that had a large effect on seed dormancy (Fig. 4e).Asn 102 and Glu 156 in Geng/ japonica had the shorter seed dormancy, Ser 102 and Asp 156 in Xian/ indica had the second longest seed dormancy, and Ser 102 and Glu 156 in Aus had the longest seed dormancy (Fig. 4e).Replacing 102 nd amino acid in the ZH11 or 156 th amino acid in the YK17 significantly reduced the germination rate (Fig. 5e-f), further indicating the importance of aa 102 and aa 156 for seed dormancy.
Homology analysis revealed that SDR3.1 had the highest homology with wild rice, and most wild rice was a combination of Ser 102 and Glu 156 (Fig. 5a-b).Although molecular phylogenetic analyses suggest that Xian/indica and Geng/japonica originated independently 61,62 , studies of domestication genes suggest a single domestication origin 11,12,14,15,43 .Therefore, the domestication of wild rice into Geng/ japonica or Xian/indica mainly depended on changes in two amino acids: aa 102 was determined for Geng/japonica and aa 156 was determined for Xian/indica (Fig. 5g).In addition, Geng/japonica was first domesticated from wild rice and subsequently hybridized with wild rice from different regions to produce Xian/indica 61,63 , which explains why a portion of Asn 102 and Glu 156 are combined in wild rice (Fig. 5b).
Domestication is also a process that has led to a dramatic reduction in biodiversity.In the context of the homogenization of the global food supply 64 , biodiversity reduction is incompatible with the need for increased crop diversity in the second wave of the Green Revolution; low crop diversity is a threat to food security 65 .The creation of germplasm, including via de novo domestication 66,67 , to increase crop diversity is important for food security 68,69 .Therefore, it is essential to retrieve genes lost during domestication and assess their regulatory mechanisms for creating germplasms.In this study, we introduced the SDR3.1 allele of MJLXL into ZH261, which was consistent with expectations and significantly reduced the PHS of ZH261 without altering other agronomic traits (Fig. 6).This finding indicates that we successfully created rice germplasm resistant to PHS, which will provide important insights into the problem of PHS in rice planting area in southern China.In addition, pyramiding different seed dormancy genes is a strategy for cultivating rice varieties with resistance to PHS 44 .Our research provides a basis for this.

Methods
Plant material and growth conditions Zhonghua 11 (ZH11), a Geng/japonica cultivar, was used as an acceptor, and MJLXL, an Aus cultivar from Bangladesh, was used as a donor.The single plant whose genotypes of all markers (108 pairs of SSR polymorphic markers evenly distributed on 12 chromosomes were used for background screening), except markers P1 and P2, were consistent with those of ZH11 was selected as the NIL (from the BC 4 F 2 population).The transgenic plants were all based on ZH11 or YK17.A population composed of 291 rice accessions was used for domestication analysis.The plants were grown in Hangzhou and Hainan under normal field management conditions.

Evaluation of seed germination
Seed germination was conducted as previously described by He et al. 70 .Briefly, approximately 100 seeds were placed in a 90-mm petri dish covered with two filter papers.Ten milliliters of distilled water was added, and the dishes were placed in a dark incubator at 25 °C for 7 days to allow for germination.Germination was considered to have occurred when the shoot length was half of the seed length or the radicle length was the full length of the seed.Three biological replicates were performed.

Map-based cloning
Primary gene mapping was conducted in a B 3 C 2 population with 153 plants derived from the cross between ZH11 and MJLXL.Approximately 4146 plants of the BC 4 F 2 population were subjected to recombinant screening with markers P3 and P4.qSDR3.1 was ultimately narrowed to a 26.9 kb interval between markers Indel27 and Indel29.For the complementation experiments, the entire DNA sequence of SDR3.1 from ZH11 was inserted into the pCAMBIA1300 vector and subsequently transformed into NIL.The CRISPR/Cas9 vector BGK03 Biogle (Hangzhou, China) was used for genome editing.The primers used in this assay are listed in Supplementary Data 2.

Expression analysis
Total RNA was extracted with TRIzol (TransGen).RNA reverse transcription was performed using ReverTra Ace (TOYOBO Biotech).qRT-PCR was performed using Applied Biosystems QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, USA).The primers used in this assay are listed in Supplementary Data 2.

Copy number analysis
The copy number of the COMs were detected via qRT-PCR.The pUC plasmid containing the SDR3.1 gene fragment was used as a standard to generate a standard curve.The primers used in this assay are listed in Supplementary Data 2.

Y2H assay
The coding sequences of SDR3.1 were inserted into pGBKT7 (Clontech, Dalian, China) as a 'bait' vector, and the coding sequences of ABI5 were inserted into pGADT7 (Clontech) as a 'prey' vector.The two plasmid vectors were cotransformed into the yeast strain Y2HGold.Transformants were spotted on SD-Leu-Trp media.After 28 °C cultivation for 3 days, five colonies were randomly selected, spotted on SD-Leu-Trp-His-Ade media and incubated for 3 days.The primers used for cloning are listed in Supplementary Data 2.

Pull-down assay
The coding sequences of SDR3.1 were inserted into the pGEX-4T-1 vector to create a fusion protein with GST.The coding sequences of ABI5 were inserted into pET28a to generate a His fusion protein.Both the empty pGEX-4T-1 vector and the corresponding vector were introduced into the E. coli strain Rosetta.The GST fusion protein was induced with 0.1 mmol/L isopropyl β-D-1-thiogalactopyranoside (IPTG) at 24 °C for 12 h.The His fusion protein was induced with 1.0 mmol/L IPTG at 16 °C for 12 h.For pull-down assays, 2 μg of His::ABI5 prey protein was incubated with 2 μg of immobilized GST or GST::SDR3.1 bait proteins.The mixtures were incubated on a shaker for 5 h (all operations were performed at 4 °C), after which the proteins were separated via 8% SDS-PAGE.Then the pulled down proteins were immunoblotted with anti-GST or anti-His antibodies.The mouse monoclonal anti-GST, mouse monoclonal anti-His and HRP-goat antimouse antibodies used for this study were obtained from Beyotime at a dilution of 1:10,000 (Beyotime: AF0174, AF2870 and A0192).The primers used in this assay are listed in Supplementary Data 2.

Bimolecular fluorescence complementation assay
The coding sequences of SDR3.1 were connected to the vector pSPYCE, and the coding sequences of ABI5 was connected to the vector pSPYNE.The plasmids were subsequently introduced into rice protoplasts 72 .Fluorescence was detected with a Zeiss LSM 710 NLO confocal microscope.The primers used in this assay are listed in Supplementary Data 2.

Fig. 1 |Fig. 2 |
Fig. 1 | Fine mapping of SDR3.1.a Roadmap of constructing near-isogenic lines and fine-mapping populations.b SDR3.1 was located between markers Indel27 and Indel29 on chromosome 3. c Identification of candidate genes of SDR3.1.d Germination rate of ZH11, the NIL, and three complementary lines 35 DAH (n = 3 independent experiments).e Copy numbers of SDR3.1 in ZH11 and three complementary lines.f PHS phenotypes of ZH11, the NIL, and three complementary lines 35 DAH.Data are presented as the mean ± SD.Source data are provided as a Source Data file.

Fig. 3 |
Fig. 3 | SDR3.1 negatively regulates ABA signaling.a-e The relative expression levels of ABA-responsive genes between ZH11 and MJLXL, among the NIL and three COM lines, among ZH11 and two SDR3.1 knockout lines, among ZH11, two ABI5 knockout lines and two OE lines.f ABA contents between ZH11 and MJLXL, among the NIL and three COM lines; among ZH11 and two SDR3.1 knockout lines; and among ZH11, two ABI5 knockout lines and two OE lines.g Germination performance of two SDR3.1 knockout lines and ZH11 seeds treated with 0, 5, or 10 μM ABA at

Fig. 4 |
Fig. 4 | Natural variation in SDR3.1 is associated with differences in seed dormancy.a Geographic distribution of the five haplotypes among 2356 accessions of 3 K RGs.N: North latitude; S: South latitude; W: West longitude; E: East longitude.b Haplotypes of SDR3.1 in 2584 accessions of 3 K RG according to five SNPs in the coding sequences region.c The 7-day germination rate of the five haplotypes (291 rice accessions) 35 DAH.The short horizontal lines represent the number of cultivars.d Rice subspecies distribution of each haplotypes.e The 7-day germination rate of the three combinations (291 rice accessions) 35 DAH.The short horizontal lines represent the number of cultivars.f Rice subspecies distribution of each amino acid sequences.g The differences of germination rate between the Xian/ indica and Geng/japonica in the combination of Ser 102 + Asp 156 .The short horizontal lines represent the number of cultivars.In c, e Data are presented as the mean ± SD, and P values are indicated by two-tailed Student's t test.Source data are provided as a Source Data file.

Fig. 5 |
Fig. 5 | Domestication analysis of SDR3.1.a Analysis of two key amino acids (aa) of SDR3.1 in wild rice.The red arrows mark the positions of the two key amino acids (aa 102 and aa 156 ).b Two combinations of two key aa of SDR3.1 in 28 wild rice accessions.cd Tajima's D and nucleotide diversity (π) for the ~2-Mb genomic region flanking SDR3.1 in the 3 K RGs; SDR3.1 is indicated by a red dashed line.e Germination rate of ZH11 and N102S 35 DAH (n = 3 independent experiments); PHS phenotypes of ZH11 and N102S 35 DAH.f Germination rate of YK17 and D156E 35 DAH (n = 3 independent experiments); PHS phenotypes of YK17 and D156E 35 DAH.g Proposed two-line model for the process of rice seed dormancy domestication.During the domestication of wild rice to Geng/ japonica, Ser 102 of SDR3.1 was converted to Asn 102 , which caused most of the dormancy lost.During the process of domestication to Xian/indica, Glu 156 was converted to Asp 156 , which caused some of the dormancy lost.The values in parentheses indicate the average germination rate.In e, f Data are presented as the mean ± SD, and P values are indicated by two-tailed Student's t test.Source data are provided as a Source Data file.

Fig. 6 |
Fig. 6 | Dormancy improvement of the high-quality rice restorer line.a The breeding scheme used for the development of BC 1 F 2 -39 by the introgression of MJLXL (donor parent, DP) into ZH261 (recurrent parent, RP).b Plant architectures of ZH261 and BC 1 F 2 -39 at the grain-filling stage.Bar = 10 cm.c Panicle architectures of ZH261 and BC 1 F 2 -39 at the grain-filling stage.Bar = 5 cm.d Germination rate of ZH261 and BC 1 F 2 -39.e PHS phenotypes of ZH261 and BC 1 F 2 -39.f Comparison of agronomic traits between ZH261 and BC 1 F 2 -39, included the 1000-grain weight, panicle length, grain length, and grain width.Data are presented as the mean ± SD, P values are indicated by two-tailed Student's t test.n = 3 independent experiments (the figure on the left in f) or n = 10 independent experiments (the three figures on the right in f).Source data are provided as a Source Data file.