REVERSAL OF RDO5 1, a Homolog of rice Seed dormancy 4, Interacts with bHLH57 and Controls ABA Biosynthesis and Seed Dormancy in Arabidopsis

The control of seed dormancy by abscisic acid (ABA) has been extensively studied, but the underlying mechanism is not fully understood. Here we report the characterization of two ABA-related seed dormancy regulators in Arabidopsis: ODR1 (for reversal of rdo5 ), an orthologue of the rice Seed dormancy 4 Sdr4, and the basic Helix-Loop-Helix transcription factor bHLH57. ODR1 , whose transcript levels are directly suppressed by the transcription factor ABA INSENSITIVE 3 (ABI3), negatively regulates seed dormancy by affecting ABA biosynthesis and ABA signaling. In contrast, bHLH57 positively regulates seed dormancy by inducing the expression of the genes 9-CIS-EPOXYCAROTENOID DIOXYGENASE NCED6 and NCED9 , which encode ABA biosynthetic enzymes, and thus leads to higher ABA levels. ODR1 interacts with bHLH57 and inhibits bHLH57-modulated NCED6 and NCED9 expression in the nucleus. bhlh57 loss of function alleles can partially counteract the enhanced NCED6 and NCED9 expression seen in odr1 mutants and can therefore rescue their associated hyper-dormancy phenotype. Thus, we identified a novel ABI3-ODR1-bHLH57-NCED6/9 network that provides insights into the regulation of seed dormancy by ABA biosynthesis and signaling.


INTRODUCTION
Dormancy prevents germination when seeds are exposed to short but temporary favorable periods before the return of adverse conditions, and thus delays seedling establishment until the start of the growing season. Dormancy therefore plays a vital role in plant survival and evolution (Linkies et al., 2010;Née et al., 2017b). Seed dormancy is imposed during seed maturation and released during after-ripening or by hydration at specific temperatures. Precise control of seed dormancy in crops results in fast and uniform germination after sowing and prevents pre-harvest sprouting, which would otherwise negatively impact agricultural production (Gubler et al., 2005).
Seed dormancy is a complex trait influenced by genetic and environmental factors (Graeber et al., 2012;Penfield and MacGregor 2017). Previous studies revealed that phytohormones, including abscisic acid (ABA), gibberellins (GA), ethylene, strigolactones, and brassinosteroids all play important roles in the control of seed dormancy Shu et al., 2016a). Among these hormones, ABA and GA have central and antagonistic roles: ABA enhances dormancy while GA stimulates germination. The roles of ABA and GA biosynthesis and signal transduction in the control of seed dormancy and germination have been intensively studied in the past decades (Gubler et al., 2005;Lefebvre et al., 2006;Née et al., 2017b).
During seed maturation, endogenous ABA gradually accumulates to enforce dormancy (Kanno et al., 2010). Cleavage of the ABA precursors 9-cis-violaxanthin and 9-cis-neoxanthin into the intermediate xanthoxin by 9-cis-epoxycarotenoid dioxygenase (NCED) is considered the key rate-limiting step in ABA biosynthesis (Nambara and Marion-Poll 2005). Of the nine Arabidopsis (Arabidopsis thaliana) NCED genes, seed-specific NCED6 and NCED9 significantly contribute to ABA biosynthesis during seed development. The corresponding nced6 nced9 double mutant shows a significant decrease of seed ABA content and concomitant reduced seed dormancy (Lefebvre et al., 2006), suggesting that both NCED6 and NCED9 are important for the establishment and maintenance of seed dormancy. A tight control of NCED6 and NCED9 expression is therefore vital for seed dormancy. Previous studies have reported that several transcription factors (TFs) control the expression of NCED6 and NCED9 during seed development, such as ABSCISIC ACID INSENSITIVE 4 (ABI4), DEHYDRATION-RESPONSIVE ELEMENT BINDING FACTOR 2C (DREB2C) and MYB96. These TFs activate NCED6 and NCED9 expression by binding to their promoters (Je et al., 2014;Lee et al., 2015;Shu et al., 2016b). Aside from biosynthesis, ABA degradation also plays an important role in determining endogenous ABA content and release of dormancy. Among four cytochrome P450 CYP707A family (CYP707A1-4) members in Arabidopsis, CYP707A2 is considered to be the major factor that executes ABA degradation in mature and hydrated seeds (Kushiro et al., 2004). The cyp707a2 mutant over-accumulates ABA and shows stronger seed dormancy . ABA signal transduction also influences seed dormancy, acting primarily through a cascade that comprises ABA receptors (encoded by PYRABACTIN RESISTANCE1 (PYR1)/PYR1-like (PYL) 1-13, also known as REGULATORY COMPONENTS OF ABA RECEPTORS (RCAR)), Type 2C protein phosphatases (PP2C) and SNF-related kinases (SnRK) (Ma et al., 2009;Park et al., 2009;Cutler et al., 2010;Miyakawa et al., 2013). Loss-of-function mutants in ABA signaling show decreased ABA sensitivity resulting in early release of seed dormancy (Park et al., 2009;Fuchs et al., 2014). ABI3 is a vital ABA-responsive TF that plays a central role in seed maturation and primary dormancy establishment in Arabidopsis. Mutations in ABI3 lead to reduced dormancy and premature seed germination (Mambara et al., 1995). ABI3 contains a B3 domain, which can physically associate with the RY motif found in the promoter of downstream genes, like SOMNUS (SOM) and STAY-GREEN (SGR)1/2 (Ezcurra et al., 2000;Mönke et al., 2004;Park et al., 2011;Delmas et al., 2013). In addition, ABI3 interacts with the important ABA-responsive TFs ABI4 and ABI5 to establish seed dormancy (Söderman et al., 2000;Lopez-Molina et al., 2002).
Apart from phytohormones, seed dormancy is also controlled by several genes originally identified as major quantitative trait loci (QTLs) like DELAY OF GERMINATION 1 (DOG1), DOG18/REDUCED DORMANCY 5 (RDO5), and DOG6 (Bentsink et al., 2010). DOG1 is a major seed dormancy factor in Arabidopsis (Bentsink et al., 2006). DOG1 protein levels are tightly correlated with dormancy levels in freshly harvested seeds, with higher DOG1 levels causing delayed germination . The basic leucine zipper TFs bZIP67 and Ethylene Response Factor ERF12 bind to the promoter of and control the expression of DOG1 and seed dormancy in response to cool temperatures and ethylene exposure, respectively (Bryant et al., 2019;Li et al., 2019). DOG1 transcript levels are also modulated by its antisense transcript asDOG1 (Fedak et al., 2016). The DOG1 protein interacts with ABA-HYPERSENSITIVE GERMINATION 1 (AHG1) and AHG3, which are Type 2C protein phosphatases that belong to the same clade as those interacting with the ABA receptors. Therefore, these PP2Cs are considered a converging point for DOG1-and ABA-dependent dormancy control pathways (Née et al., 2017a;Nishimura et al., 2018).
Another seed dormancy factor, RDO5/DOG18, was identified both in a mutagenesis screen and by QTL mapping. RDO5 was shown to control seed dormancy independently of ABA, and a transcriptome analysis suggested that RDO5 does so by controlling the expression of the genes encoding the PUMILIO RNA binding proteins APUM9 and homologs, revealing a post-transcriptional dormancy pathway (Xiang et al., 2014，2016). RDO5 belongs to the PP2C family of protein phosphatases but is found in a clade distinct from AHG1 and AHG3. In fact, RDO5 functions as a pseudophosphatase because it lacks phosphatase activity . Interestingly, the RDO5 and DOG1 proteins interact in Arabidopsis seeds (Née et al., 2017a), although the reason or the role of such interaction in seed dormancy is unknown.
In order to dissect the RDO5-mediated seed dormancy network, we carried out a suppressor mutagenesis screen with the low dormancy mutant rdo5-2. Here, we report the identification of one such suppressor (globally called odr, for reversal of the rdo phenotype): odr1. Mutations in ODR1 cause stronger seed dormancy. Expression of ODR1 is repressed by ABI3, and ODR1 negatively affects the expression of NCED6 and NCED9 and ABA content in freshly harvested seeds. Furthermore, ODR1 interacts with bHLH57 and prevents bHLH57-mediated induction of NCED6 and NCED9 expression. In agreement, the odr1-2 bhlh57 double mutant decreased the expression of NCED6 and NCED9 compared to bhlh57 single mutants, while rescuing the hyperdormancy phenotype of odr1-2. We therefore discovered a new seed dormancy pathway that includes ABI3, ODR1, bHLH57, NCED6, and NCED9. Because the rice orthologue of ODR1 was previously identified as a QTL in rice, our work also provides a molecular link between pre-sprouting research in rice and seed dormancy in Arabidopsis.

The Seed-Specific Protein ODR1 Negatively Controls Seed Dormancy
We had previously reported the cloning and initial characterization of the positive dormancy factor RDO5, which controls seed dormancy without influencing ABA metabolism or signal transduction. Loss-of-function rdo5 mutant alleles cause strongly reduced seed dormancy (Xiang et al., 2014. Seeking to uncover the function of RDO5 in seed dormancy regulation, we carried out a -ray mutagenesis screen with rdo5-2, a T-DNA insertion mutant with substantially reduced dormancy duration (Xiang et al., 2014). We identified six mutants that suppressed the rdo5-2 dormancy phenotype, and named them odr1-6 (for reversal of the rdo phenotype). The odr1 mutant exerted the most significant effect on seed dormancy and was therefore selected for further characterization. We used bulked segregant analysis-based sequencing to identify mutations that might be responsible for the phenotype and identified an 8 bp deletion in At1g27461 that caused a reading frame shift leading to a premature stop codon (Supplemental Figure 1A and 1B). To confirm the identity of At1g27461 as ODR1, we obtained an independent mutant in the gene, the homozygous T-DNA insertion mutant SALK_022729, which we named odr1-2. This mutant carried a T-DNA insertion in the single ODR1 exon and lacked full-length ODR1 transcript (Supplemental Figure 1C and 1D). The odr1-2 allele is therefore presumed to be a null allele. Germination assays showed that odr1-2 had reduced germination compared to wild-type Col-0 ( Figure 1A and 1B). We also introduced the odr1-2 T-DNA insertion into rdo5-2 via crossing and conducted germination assays with the double mutant: it showed lower germination rates than rdo5-2, indicating that odr1-2 also effectively suppressed the dormancy phenotype seen in rdo5-2 ( Figure 1A). Additionally, to confirm the repressive role of ODR1 in seed dormancy, we generated overexpression lines in the odr1-2 background by placing a copy of ODR1 under the control of the Cauliflower Mosaic Virus 35S promoter (Supplemental Figure 1D). Germination assays showed that homozygous 35Spro:ODR1/odr1-2 seeds had similar dormancy rates as Col-0 ( Figure 1B). These results established that constitutive expression of ODR1 rescues the enhanced dormancy phenotype of odr1-2, and that ODR1 negatively impacts seed dormancy.
The ODR1 gene consists of a single exon of 1,065 bp, encoding a 355 amino acid protein. Protein sequence alignment and phylogenetic analysis demonstrated that ODR1 and its homologues are conserved within angiosperms (Supplemental Figure 2A and 2B). ODR1 was previously described as DROUGHT RESPONSIVE GENE, a seedspecific gene induced under drought conditions (Moon et al., 2016). The drg mutant was shown to be more sensitive to a number of abiotic stresses, including osmotic stress, drought and freezing. The putative rice orthologue of DRG, the zinc finger protein Sdr4, is a positive factor of seed dormancy (Sugimoto et al., 2010). In another study, germination of freshly harvested drg/Atsdr4L mutant seeds were found to be insensitive to exogenous GA (Cao et al., 2020). The SDR acronym is already assigned to designate short-chain dehydrogenase reductases in Arabidopsis, while the DRG acronym can also stand for developmentally regulated GTP-binding proteins. We therefore propose to rename DRG/AtSDR4 as ODR1 to avoid confusion and to adopt a single gene identifier representative of its underlying function.
We analyzed the subcellular localization of ODR1 by transient expression in tobacco (Nicotiana benthamiana) leaves. The ODR1-YFP fusion protein was exclusively localized to the nucleus ( Figure 1C). We next evaluated ODR1 expression levels in different Arabidopsis tissues by RT-qPCR: ODR1 transcript was undetectable in roots, stems, leaves, or flowers, but it gradually increased during seed maturation and reached its highest level in seeds 20 d after pollination, while it sharply declined in hydrated seed ( Figure 1D). This pattern of expression is consistent with the expression data for ODR1 in the Arabidopsis eFP Browser (Supplemental Figure 3). Taken together, ODR1 is specifically expressed in seeds and encodes a nuclear protein that is involved in the control of dormancy.

ODR1 Strongly Impacts ABA-Mediated Dormancy
ABA plays a crucial role in the control of seed dormancy and germination. To test whether the enhanced dormancy of odr1-2 was associated with ABA, we first measured the ABA content in freshly harvested seeds from different genotypes, including Col-0, odr1-2, 35Spro:ODR1/odr1-2, rdo5-2 and rdo5-2 odr1-2. ABA levels in odr1-2 freshly harvested seeds were approximately 40% higher compared to those in Col-0 and the overexpression lines (Figure 2A), which is consistent with the stronger dormancy seen in odr1-2 seeds. The rdo5-2 odr1-2 double mutant also showed higher ABA levels compared to the rdo5-2 single mutant, which itself was similar to Col-0 in freshly harvested seeds ( Figure 2B), as previously reported (Xiang et al., 2014). These results suggest that ODR1 might influence ABA metabolism. Next, we evaluated the germination behavior of freshly harvested seeds of Col-0 and odr1-2 in the presence of the ABA biosynthesis inhibitor fluridone. The odr1-2 mutant responded more strongly to fluridone compared to Col-0 by exhibiting lower germination rates, suggestive of heightened ABA sensitivity ( Figure 2C). Overall, these results demonstrate that the stronger seed dormancy in odr1-2 is largely caused by a high level of endogenous ABA in dry seeds and de novo ABA synthesis during seed hydration. Finally, we evaluated the ABA sensitivity of Col-0 and odr1-2 after-ripened seeds. Seeds of both genotypes germinated fully and synchronously after stratification on half-strength Murashige and Skoog growth medium in the absence of ABA. However, odr1-2 seeds were more sensitive to ABA as shown by reduced germination rates when the medium was supplemented with 0.5 μM ABA. This suggests that loss-of-function of ODR1 affects ABA sensitivity of seeds ( Figure 2D).
Since ABA content was higher in odr1-2 seeds compared to Col-0, we evaluated the expression levels of pivotal ABA metabolism genes (Supplemental Figure 4), including NCEDs (NCED2, NCED3, NCED5, NCED6, NCED9, involved in ABA biosynthesis) and CYP707As (CYP707A1-4, linked to ABA degradation) in Col-0 and odr1-2 freshly harvested and in seeds that had been hydrated for 6 hours. Nearly all of the NCEDs (except NCED5) and CYP707A2 were significantly upregulated in dry and/or hydrated seeds in odr1-2 compared with Col-0 ( Figure 2E). Differential expression of NCEDs could well explain the increased ABA content of odr1-2 freshly harvested seeds, while the elevated expression levels of CYP707A2 might indicate a feedback reaction to the increased ABA content in odr1-2 seeds. However, ODR1 does not activate gene expression by direct binding to promoters, as we failed to detect ODR1 binding ability to any of the NCEDs and CYP707As promoters in yeast onehybrid assays (Supplemental Figure 5). In addition, ODR1 lacks self-transcriptional activity in yeast two-hybrid assays, which might have been expected from a transcription factor ( Figure 4A). We also found that ODR1 indirectly regulated DREB2C and ABI4 gene expression in dry or/and hydrated seeds but did not physically interact with the encoded proteins (Supplemental Figure 6). Taken together, these results suggest that ODR1 controls seed dormancy by affecting the expression of genes involved in ABA metabolism but is unlikely to behave as a TF.
We had shown previously that RDO5 controls seed dormancy in an ABAindependent manner. However, loss-of-function mutants of odr1-1 and odr1-2 both suppressed the weak dormancy phenotype of rdo5-2 in double mutant combinations, which indicated that the ODR1-mediated ABA metabolism pathway was epistatic to RDO5 for seed dormancy. To further confirm this relationship, we generated the rdo5-2 cyp707a2 double mutant by crossing. Germination assays showed that the rdo5-2 cyp707a2 double mutant had a stronger dormancy compared to rdo5-2 ( Figure 2F). This result demonstrated that the cyp707a2 mutant, which causes an over-accumulation of ABA in seeds, is epistatic to rdo5-mediated reduced seed dormancy. Surprisingly, the dormancy level of rdo5-2 cyp707a2 was even higher than that of the cyp707a2 single mutant. This might be due to a feedback response to the rdo5 mutation in order to keep seeds in a dormant status, which becomes noticeable in the cyp707a2 mutant background. This is supported by increased transcript levels following 6 hours of seed hydration in the rdo5-2 background for two PYR/PYL/RCAR genes (At4g17870, At2g38310) and a SnRK2 gene (At1g78290) involved in ABA signaling (Xiang et al., 2014). We also found that DOG1, the core factor controlling seed dormancy, was upregulated in odr1-2, but again without direct interaction between the encoded protein and ODR1. Higher DOG1 expression might also contribute to the hyper-dormancy phenotype of the odr1-2 mutant (Supplemental Figure 7).
Overall, ODR1 controls seed dormancy at least partially through modulation of ABA biosynthesis, which acts downstream of the RDO5-mediated dormancy pathway.

ODR1 is a Direct Target of ABI3
A previous chromatin immunoprecipitation followed by hybridization to tiling arrays (ChIP-chip) and transcriptome analysis predicted that ODR1 was one of the 98 direct targets of ABI3 (Mönke et al., 2012). Based on this finding, we analyzed the ODR1 promoter sequence and found two potential RY motifs (CATGCA) known to be binding sites for ABI3 ( Figure 3A). To validate the control of ODR1 expression by ABI3, we carried out a yeast one-hybrid assay, whose results indicated that ABI3 specifically binds to the ODR1 promoter at the proximal RY motif (CATGCA-363), but not at the more distal motif (CATGCA-700) ( Figure 3B). Furthermore, we performed ChIP followed by quantification of immunoprecipitated chromatin by qPCR with chromatin extracted from seedlings overexpressing ABI3 fused to a FLAG tag (35Spro:ABI3-FLAG; Supplemental Figure 8) (Park et al., 2011). We found that the DNA fragment containing the proximal RY motif (CATGCA-363) of the ODR1 promoter was highly enriched compared to other DNA fragments ( Figure 3C). To investigate how ABI3 controlled ODR1 expression, we then carried out a transient expression assay using N. benthamiana leaves, which showed that ABI3 repressed the expression of ODR1 ( Figure 3D). In addition, expression levels of ODR1 in abi3-1 and 35Spro:ABI3-FLAG seeds demonstrated that overexpression of ABI3 reduced ODR1 expression, while the weak allele abi3-1 showed enhanced ODR1 expression during seed hydration compared to the wild-type control ( Figure 3E). Finally, we used the genome editing technique CRISPR/Cas9 (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-Associated nuclease) to inactivate ODR1 in an abi3-1 background. After genomic PCR verification and Sanger sequencing, three lines with different editing scars in ODR1 (containing a 1 bp insertion, a 32 bp and a 55 bp deletion, respectively) were identified and named odr1-3, odr1-4, and odr1-5 (Supplemental Figure 9).
Germination assays of wild-type Ler, abi3-1, abi3-1 odr1-4, and abi3-1 odr1-5 showed that removal of ODR1 activity could largely suppress the weak dormancy rates of abi3-1 ( Figure 3F), suggesting that ODR1 genetically acts downstream of ABI3 and that ABI3 requires functional ODR1 to control seed dormancy. Taken together, we found that ABI3, which is an essential factor of ABA signaling transduction and seed dormancy, represses ODR1 expression by binding to its promoter.

ODR1 Interacts with the bHLH-type Protein bHLH57 in the Nucleus
We performed a yeast two-hybrid screen using ODR1 as bait to identify potential interacting partners and identified bHLH57 (At4g01460), a putative basic helix-loophelix-type TF. The direct physical interaction between ODR1 and bHLH57 was confirmed using a yeast two-hybrid assay using their full-length coding sequences. A yeast strain containing both ODR1 and bHLH57 was able to grow on synthetic medium lacking His, Leu and Trp and supplemented with 2 mM 3-AT, in contrast to the negative controls ( Figure 4A). We also performed bimolecular fluorescence complementation (BiFC) assays by co-transfecting different combinations of proteins fused to fragments of YFP into N. benthamiana leaves. Yellow fluorescence signal was detected in the nuclei of N. benthamiana cells co-expressing ODR1-YFPn and bHLH57-YFPc, while the expression of either fusion protein alone did not result in measurable YFP signal ( Figure 4B). Finally, we confirmed the interaction between ODR1 and bHLH57 with pull-down experiments using total protein extracts from E.coli producing bHLH57-GST and ODR1-6×His. The bHLH57-GST fusion protein was able to pull down ODR1-6×His, indicating that bHLH57 interacts with ODR1 in vitro ( Figure 4C). These results therefore confirm that ODR1 physically interacts with bHLH57, suggesting that they may modulate seed dormancy as a complex.
The bhlh57-1 allele carries the insertion ~50 bp upstream of the ATG and reduced bHLH57 transcript levels to 10% of Col-0. The T-DNA is inserted within the first exon in bhlh57-2 and abrogates bHLH57 transcript accumulation ( Figure 5A; Supplemental Figure 10A). Germination assays showed that both mutants had weaker dormancy compared to Col-0, indicating that bHLH57 positively controls seed dormancy ( Figure   5B). bHLH57 localizes to the nucleus and is expressed in various plant organs, including roots, stems, and leaves, but is very low in flowers ( Figure 5C and D).
bHLH57 transcript levels gradually decreased during seed maturation and declined after seed hydration ( Figure 5D).
Considering the reduced seed dormancy phenotype of bhlh57 mutants and the interaction between ODR1 and bHLH57, we speculated that the ABA content and ABA sensitivity of bhlh57-2 seeds might also be affected. Indeed, ABA content in bhlh57-2 freshly harvested seeds was significantly reduced compared to Col-0 seeds ( Figure 5E).
ABA sensitivity of bhlh57-2 seeds was however similar to that of Col-0 (Supplemental Figure 10B), suggesting that the reduced dormancy of bhlh57-2 seeds was related to ABA metabolism. Next, we analyzed the expression of NCEDs and CYP707As in bhlh57-2 freshly harvested seeds. NCED6 and NCED9 transcripts were both significantly down-regulated in bhlh57-2 compared with Col-0 ( Figure 5F), indicating that bHLH57 normally induces the expression of NCED6 and NCED9. In addition, CYP707A2 expression was elevated in bhlh57-2, collectively contributing to the decreased ABA content in mutant seeds. In contrast, both CYP707A3 and CYP707A4 were down-regulated, which could be due to an ABA metabolic feedback in bhlh57-2 ( Figure 5F). These results, in combination with the interaction between ODR1 and bHLH57 and the altered transcript levels of NCED6 and NCED9 in the odr1-2 mutant, suggest that bHLH57 may directly control NCED6 and NCED9 expression.
bHLH TFs generally modulate the expression of their target genes through binding to E-box (CANNTG) motifs in their promoters (Atchley and Fitch 1997). The 1,400 bp promoter region upstream of NCED6 contains five E-box motifs, and an E-box and a G-box motif (CACGTG) were also found to be present in the 600 bp promoter region upstream of NCED9 ( Figure 6A). To determine whether bHLH57 could bind to the promoters of NCED6 and NCED9, we carried out yeast one-hybrid and ChIP-qPCR assays. The yeast one-hybrid assays showed that bHLH57 did bind to the NCED6 and NCED9 promoters, but could not bind to the promoters of NCED2, NCED3, CYP707A2-4 or DOG1 ( Figure 6B; Supplemental Figure 11A). Furthermore, we determined that bHLH57 bound to the first, fourth, and fifth E-box motifs in the NCED6 promoter, and to the G-box of the NCED9 promoter (Supplemental Figure 11B). ChIP-qPCR assays with 35Spro:bHLH57-GFP transgenic seeds (Supplemental Figure 12) further confirmed the binding between bHLH57 and the promoters of NCED6 and NCED9: the bHLH57 protein preferentially bound to the first, third, fourth, and fifth Ebox motifs in the NCED6 promoter, and to both the E-box and G-box motifs in the NCED9 promoter ( Figure 6C). We also performed a transient expression assay in N.
benthamiana leaves to confirm the positive control of bHLH57 in NCED6 and NCED9 expression in plants. bHLH57 activated both NCED6 and NCED9 expression relative to negative controls ( Figure 6D). Collectively, our data demonstrate that bHLH57 induces NCED6 and NCED9 expression through binding to specific E/G-boxes within their promoters, thereby controlling seed dormancy by influencing ABA metabolism.

ODR1 Inhibits the Regulation of NCED6 and NCED9 by bHLH57
Our results above demonstrated an interaction between ODR1 and bHLH57. We also documented the opposite phenotypes on seed dormancy, ABA content, and NCED6 and NCED9 expression exhibited by the odr1-2 and bhlh57-2 mutants. These observations prompted us to investigate whether ODR1 may prevent the induction of NCED6 and NCED9 expression by bHLH57. To test this hypothesis, we performed a transient dual-luciferase assay by co-expressing 35Spro:ODR1 and/or 35Spro:bHLH57 with the luciferase (LUC) reporters NCED6pro:LUC or NCED9pro:LUC in N.
benthamiana leaves. Co-expression of ODR1 and bHLH57 indeed resulted in a significant decrease in LUC activity compared to expression of bHLH57 alone ( Figure   7A), suggesting that the upregulation of NCED6 and NCED9 by bHLH57 is blocked by ODR1. Furthermore, we generated the odr1-2 bhlh57-2 double mutant (Supplemental Figure 13) to evaluate its germination behavior and pattern of NCED6 and NCED9 expression. Transcript levels for NCED6 and NCED9 were both upregulated in odr1-2 seeds, but this was largely counteracted when bHLH57 was also inactivated in the odr1-2 bhlh57-2 double mutant ( Figure 7B). Germination rates indicated that the dormancy level of the odr1-2 bhlh57-2 double mutant was significantly lower than that of odr1-2 ( Figure 7C), demonstrating that ODR1 and bHLH57 had opposite roles in the control of dormancy and that ODR1 requires bHLH57 for its full function in seed dormancy. Overall, these results suggest that ODR1 inhibits the upregulation of NCED6 and NCED9 by bHLH57 in seeds.
Interestingly, the odr1-2 bhlh57-2 double mutant was still more dormant than bhlh57-2 ( Figure 7C), and also showed increased expression of NCED6 and NCED9 compared to wild-type ( Figure 7B). This indicates that ODR1 also controls NCED6 and NCED9 expression through other factors. Collectively, these results demonstrate that ODR1 negatively controls NCED6 and NCED9 expression partly through direct interaction and inhibition of bHLH57, which induces NCED6 and NCED9 expression.

DISCUSSION
The identification and characterization of seed dormancy genes and understanding their roles in the dormancy mechanism is important for crop genetic improvement. In this study, we isolated two new dormancy factors, ODR1 and bHLH57, and provide genetic, biochemical and molecular evidence to support the signaling network in which they work alongside ABI3 and NCED6/9. We propose that this network functions in a positive feedback loop with ABA: higher levels of ABA lead to the upregulation of ABI3 during seed maturation. ABI3 then binds physically to the ODR1 promoter and represses its expression. In the absence of the ODR1-imposed inhibition, bHLH57 can bind to the promoters of NCED6 and NCED9 and elevates their expression levels, resulting in enhanced ABA biosynthesis and stronger seed dormancy. This control may be gradually amplified via a positive feedback loop involving the further induction of ABI3 by ABA (Figure 8). Cao et al. (2020) recently reported a role for ODR1/SEED DORMANCY 4-LIKE (AtSdr4L) in seed dormancy. Based on a physiological and genetic analysis, the authors concluded that this role was mainly mediated by the GA biosynthesis and signaling pathways. Indeed, odr1/atsdr4l loss of function was associated with higher expression of the GA biosynthesis genes GA20-OXIDASE1 (GA20OX1) and GA20OX2 (Cao et al., 2020). Taken together, we propose that ODR1/AtSdr4L, a homolog of OsSdr4, plays a negative role in seed dormancy by adjusting the ABA and GA balance during seed maturation and germination (Figure 8).

NCED9 Expression
We identified ODR1 as a suppressor of the low dormant rdo5-2 mutant during a ray mutagenesis screen. Considering the germination behavior and ABA content of rdo5-2, odr1-2 single mutants and rdo5-2 odr1-2 double mutant, we hypothesized that ODR1 controls seed dormancy by modulating ABA biosynthesis, which is independent of RDO5-mediated seed dormancy. ODR1 interacts with bHLH57 and represses the induction of NCED6 and NCED9 expression and ABA biosynthesis directly (Figure 4, 5, 6 and 7). Therefore, ODR1 and bHLH57 control ABA metabolism and dormancy in opposite directions. The mechanisms underlying these control pathways are consistent with the reported role of ABA in seed dormancy (Lefebvre et al., 2006;Seo et al., 2006). In Arabidopsis, NCED6 and NCED9 are the main contributors to ABA biosynthesis during seed maturation; in agreement, the nced6 nced9 double mutant has significantly reduced ABA levels and weaker seed dormancy (Lefebvre et al., 2006). In contrast, induction of NCED6 during seed development increased seed dormancy (Martínez-Andújar et al., 2011).
Studies have reported that the expression of NCED6 and NCED9 is directly promoted by TFs, including ABI4, DREB2C, and MYB96 (Je et al., 2014;Lee et al., 2015;Shu et al., 2016b). Therefore, it is interesting to note that our data also support an indirect function for ODR1 on ABA biosynthesis by decreasing the transcription of ABI4 and DREB2, as well as that of the core dormancy gene DOG1 (Supplemental Figures 6 and 7). This indicates that there are still other pathways in ODR1-mediated dormancy that need further investigation.
It was previously reported that two bHLH TFs, SPATULA (SPT) and PHYTOCHROME INTERACTING FACTOR1 (PIF1), control seed germination and GA3OX expression in response to light and temperature signaling in Arabidopsis (Penfield et al., 2005;Oh et al., 2006). bHLH TFs also function in seed dormancy in other plant species (Gao et al., 2018;Zhao et al., 2019). These findings collectively demonstrate the conservation and diversity of the bHLH gene family in the control of dormancy. Our study extends a role for the TF bHLH57 in seed dormancy by promoting NCED6 and NCED9 expression.
It is noteworthy that our genetic analysis demonstrated that the loss of bHLH57 function in the odr1-2 background was only partially able to counteract the higher NCED6 and NCED9 expression levels and associated hyper-dormancy phenotype brought on by the loss of ODR1 ( Figure 7B and 7C). These results indicate that ODR1 probably also controls NCED6/9 by factors other than bHLH57, like the abovementioned ABI4 and DREB2C. Furthermore, other dormancy factors such as GA and DOG1 have recently also been reported to be involved in ODR1/AtSdr4L-mediated seed dormancy (Figure 8, Cao et al., 2020).

ODR1 is Negatively Controlled by ABI3 during Seed Dormancy Establishment
ABI3 is a vital ABA-responsive TF that directly binds to the RY motif in the promoter of its target genes via a B3 domain, and it plays a central role in seed maturation and the establishment of primary seed dormancy (Ezcurra et al., 2000;Mönke et al., 2004). Mutations in ABI3 lead to reduced dormancy and premature seeds germination (Nambara et al., 1995). ChIP-chip and transcriptome analysis in a previous study identified a set of 98 genes (including ODR1) involved in seed development, seed protein and lipid accumulation as direct targets of ABI3 (Mönke et al., 2012). In our study, we confirmed the binding activity of ABI3 to the proximal RY motif in the promoter of ODR1 by yeast one-hybrid and ChIP-qPCR assays ( Figure 3B and 3C).
Furthermore, we demonstrated that expression of ODR1 is repressed by ABI3 using gene expression analysis and transient expression assays ( Figure 3D and 3E). Finally, inactivation of ODR1 in the abi3-1 background by CRISPR/Cas9 genome editing supported the position of ABI3 upstream of ODR1 in the control of seed dormancy ( Figure 3F).

AtODR1 and its rice Homolog OsSdr4 have Opposite Roles in Seed Dormancy
Seed dormancy is a key decision point in plant development. Hence, maintaining suitable levels of dormancy is of great importance and has been under strong natural selection during plant evolution. In our study, we found that ODR1 and its homologs are conserved in angiosperms (Supplemental Figure 2), indicating that they might have a conserved function in seed dormancy. Sdr4 is a putative ODR1 orthologue in rice with about 35% amino acid identity (and 49% similarity). However, Sdr4 positively affects seed dormancy in several rice cultivars (Sugimoto et al., 2010), which is the opposite of ODR1 in Arabidopsis. We speculate that such divergence in function might be related to their associated ABA-responsive factors: ABI3 and OsVP1. In rice, OsVP1 induces the expression of Sdr4 during seed development and maturation (Sugimoto et al., 2010), whereas ABI3 represses ODR1 expression in Arabidopsis seeds. These opposite effects confer a deeper dormancy in odr1-2, and pre-harvest sprouting and non-dormant phenotypes in sdr4. Furthermore, the expression levels of the two closest rice homologues of Arabidopsis DOG1 were lower in the rice sdr4 mutant compared to wild type, while our results showed that DOG1 was upregulated in odr1-2 freshly harvested and hydrated Arabidopsis seeds (Supplemental Figure 7). The different expression behavior of DOG1 and its rice homologues may contribute to the distinct mutant phenotypes. We hypothesize that the relationship between ODR1 and DOG1 might be conserved between dicots and monocots. Given the opposite modes of action displayed by ABI3 and OsVP1 on ODR1 and Sdr4 in seed dormancy, it will be of interest to assess the effects of Sdr4 on ABA metabolism and dormancy in rice. Moreover, considering the relatively low identity between ODR1 and Sdr4, we speculate that the opposite dormancy phenotypes of ODR1 and sdr4 in dormancy may also be attributed to sequence variation in their coding sequences, which could lead to a divergence in protein function.

Mapping of ODR1
To clone the ODR1 gene, we followed a next-generation sequencing (NGS)-based bulked-segregant analysis (BSA) approach. First, odr-1 was backcrossed with its wild type rdo5-2 four times to minimize background effects. Then, a segregating population was produced by outcrossing ODR1-1 with Ler. F2 progeny with a homozygous rdo5-2 background were selected for phenotyping. Bulk DNA was in two groups based on their dormant or non-dormant phenotype. After sequencing of bulk DNA, we performed linkage analysis based on whole genome distributed SNP markers and detailed genome sequence analyses in the candidate region.

Constructs and Plant Transformation
To generate 35Spro:ODR1/odr1-2 plants, the full-length coding sequence (CDS) of ODR1 was amplified and cloned into the pFAST-R01 vector (Shimada et al., 2010) and The ODR1 genome-edited lines in the abi3-1 background were generated using the pYAO-based CRISPR/Cas9 system according to the instructions (Yan et al., 2015).

Seed Germination Assays
Seed germination assays were performed as previously described (Xiang et al., 2014;Née et al., 2017). For ABA response assays, freshly harvested seeds were

Subcellular Localization and BiFC Assay
To generate constructs for subcellular localization analysis, the CDS of ODR1 and bHLH57 were amplified and clone in frame with enhanced yellow fluorescent protein (eYFP) driven by the 35S promoter in pEarleyGate101 (Earley, et al., 2006). Vectors were transiently transfected into N. benthamiana leaves as previously described (Liu et al., 2009). Yellow fluorescence signal was detected using a confocal laser scanning microscope (Zeiss) at the excitation wavelength of 513 nm. For Bimolecular Fluorescence Complementation (BiFC) assays, the CDS of ODR1 and bHLH57 and truncated bHLH57 (bHLH57ΔN) were amplified and inserted into pBatTL-B sYFPn (YFPn) and pBatTL-B sYFPc (YFPc) vectors, respectively. The BiFC assay was performed as described (Née et al., 2017a).

RT-qPCR Assay
RNA Isolation Aid (Invitrogen) as described previously (Kushiro et al., 2004). Firststrand cDNA was synthesized using All-In-One RT MasterMix (Applied Biological Materials) according to the product instructions. qPCR was then performed with PowerUpTM SYBR® Green Master Mix (Life technologies) with gene-specific primers (Supplemental Table 1). PP2A (At1g69960) was used as the internal control. Three independent biological replicates with three technological repeats each were performed.

Yeast Two-hybrid and One-hybrid Assay
The yeast two-hybrid library screen was performed with the Arabidopsis Mate & Plate TM Library (Clontech) according to the user manual. For confirmation of interaction between ODR1 and bHLH57, bHLH57 CDS was amplified and inserted into the pGADT7 vector and co-transformed into yeast strain AH109 harboring pGBKT7-ODR1 and confirmed with the method described previously (Liu et al., 2016).
Interaction between TTG1 and MYB5 was used as positive control (Gonzalez et al., 2009). Yeast one-hybrid assays were performed as described previously (Li et al., 2011).
The CDS of ABI3, ODR1, bHLH57 and ABI4 were amplified and individually fused in frame with the GAL4 activation domain (AD) in the pGADT7-Rec2 vector digested with SmaI. The various pairs of recombined pHIS2 and pGADT7-Rec2 plasmids were co-transformed into the yeast strain AH109 individually and grown on SD/-Leu/-Trp/-His/-Ade medium with 50 mM 3-AT. Binding of ABI4 to the promoters of NCED6 and CYP707A1 was used as positive controls (Shu et al., 2016b).

GST Pull-Down Assay
To generate constructs for GST pull-down assays, the CDS of ODR1 and bHLH57 were amplified and fused in frame with a His tag or GST in the vectors pET-28a and modified pGEX, to generate vectors expressing ODR1-6×His and GST-bHLH57 proteins in E.coli (BL21). GST pull-down assays were carried out with the PierceTM GST Protein Interaction Pull-Down Kit (Thermo) according to the user manual. Briefly, isolated bait lysate GST or GST-bHLH57 was incubated with glutathione agarose resin in TBS solution (25 mM Tris·HCl, 150 Mm NaCl, pH 7.2) for 1 hour at 4°C with gentle rocking. After immobilization of the bait protein, the prey lysate was incubated with the glutathione agarose resin-bait protein complex in TBS solution for 1 hour at 4°C.

Transient Expression Assay
To generate plasmids for the transient expression assays, the respective promoters of ODR1, NCED6, and NCED9 (same promoter fragments used in yeast one-hybrid assays) were amplified and inserted into pGreenII-0800-LUC vector to produce luminometer (Ye et al., 2017). Three biological replicates were measured for each sample.

ChIP-qPCR Assay
ChIP assays were carried out following the procedure described previously (Wang et al., 2016). Briefly, approximately 1 g hydrated seeds were cross-linked with 1% formaldehyde solution for 15 min under vacuum, and the fixation reaction was then terminated by adding 2 M glycine to a final concentration of 0.125 M. After grinding all the samples into a powder, chromatin was isolated and sheared by sonication to 300-1,000 bp. The genomic DNA fragments were immunoprecipitated by the addition of anti-FLAG (Sigma, Catalog # F1804) or anti-GFP antibodies (abcam, Catalog # ab290).
The precipitated DNA was recovered with the EpiQuikTM Plant ChIP Kit (Epigentek, P-2014-24) and analyzed by qPCR with specific primers (Supplemental Table 1). Three biological replicates were measured for each sample.

Quantification of ABA
The measurement of ABA was performed as described previously with small modifications (Müller and Munné-Bosch 2011), Briefly, 20 mg of seeds were homogenized in a pre-cooled methanol/isopropanol (20:80 v/v) solution containing 0.2% formic acid using a TissueLyser (JX-24) with zirconia beads for 3 min at 30 Hz and ABA was extracted at -20°C overnight. The supernatant was collected after 4°C, 14,000 g centrifugation for 15 min and dried with a flow of nitrogen. The residue was dissolved with 100 μL cold methanol solution containing internal standard d6-ABA (CDN Isotopes). Quantification of ABA was performed with a UPLC-MS/MS system consisting of a UPLC system (Waters) and a triple quadruple tandem mass spectrometer (AB Sciex). Three independent biological replicates were performed for each sample.

Accession Numbers
Sequence data from this investigation can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: ODR1 and PP2A (At1g13320).   (B) ABA content in Col-0, rdo5-2 and rdo5-2 odr1-2 freshly harvested seeds. Values are means ± SD of three biological repeats. **, P < 0.01 by Student's t-test.

Supplemental Data
(C) Germination rates of Col-0 and odr1-2 seeds after 1 week of storage, tested in the presence of 0.05% ethanol (control) or 0.05% ethanol plus the indicated concentrations of fluridone. Values are means ± SD of three independent batches of seeds per genotype.
(D) Germination rates of Col-0 and odr1-2 freshly harvested seeds on 1/2 MS medium supplemented with 0 or 0.5 µM ABA. Values are means ± SD of three independent batches of seeds per genotype.
(E) RT-qPCR analysis of NCEDs and CYP707As transcript levels in Col-0 and odr1-2 freshly harvested dry and seeds hydrated for 6 h. The expression values were normalized to PP2A. ND, Not detected.
Values are means ± SD of three biological repeats. * or **, P < 0.05 or 0.01 by Student's t-test.
Values are means ± SD of three independent batches of seeds per genotype. (A) A schematic diagram of the ODR1 promoter. Three fragments were amplified for yeast one-hybrid: the -1 to -457 bp fragment (containing RY2 element (RY2pro)), the -458 to -1,100 bp fragment (containing RY1 element (RY1pro)) and the -1 to -1,100 bp fragment (ODR1pro). Three fragments for ChIP-qPCR detection (named RY1, RY2 and P1) are depicted with parallel black lines below the promoter. Scale bar, 100 bp.
(C) ChIP-qPCR assay: ABI3 mainly binds to the RY2 motif-containing fragment of the ODR1 promoter.
Immunoprecipitation was performed using seeds hydrated for 6 h from Col-0 and 35Spro:ABI3-FLAG with anti-FLAG antibody or anti-lgG. qPCR was performed with specific primers listed in Supplementary  (F) Germination rates after different periods of dry storage of wild-type Ler, abi3-1, abi3-1 odr1-4 and abi3-1 odr1-5 seeds. Values are means ± SD of three independent batches of seeds per genotype.   and -1,170 bp to -1,400 bp (E-box1pro)) of NCED6 promoter and fragments (from -1 bp to -224 bp (Gboxpro) and -225 bp to -600 bp (E-boxpro)) of the NCED9 promoter were amplified for a yeast one-hybrid assay. The relative positions of the ChIP-qPCR-amplified fragments in the NCED6 promoter (named E-box1, E-box2, E-box3, F1, E-box4, E-box5) and in the NCED9 promoter (named E-box, F2, G-box) are depicted with parallel black lines below the promoter. Scale bars, 100 bp.
(B) Yeast one-hybrid assay: direct binding of bHLH57 to the NCED6 and NCED9 promoters. 50 mM 3-AT was added in the -Leu/-Trp/-His medium. ABI4 binding to NCED6 promoter was used as a positive control.
(C) ChIP-qPCR assay: bHLH57 binds to the E/G-box motifs of the NCED6 (left) and NCED9 (right) promoters. Immunoprecipitation was performed using seeds hydrated for 6 h from Col-0 and 35Spro:bHLH57-GFP with anti-GFP or anti-lgG antibody. qPCR was performed with specific primers listed in Supplementary Table 1. The comparison is between the nonimmune control (anti-IgG) and the immunoprecipitation (anti-GFP). Values are means ± SD of three biological replicates. * or **, P < 0.05 or 0.01 by Student's t-test.  (B) RT-qPCR analysis of NCED6 and NCED9 transcript levels in Col-0, odr1-2, bhlh57-2 and odr1-2 bhlh57-2 freshly harvested seeds. The expression values were normalized to PP2A. Values are means ± SD of three biological repeats. **, P < 0.01 by Student's t-test.
(C) Germination rates of Col-0, odr1-2, bhlh57-2 and odr1-2 bhlh57-2 after-ripened seeds after 1 week of storage. Values are means ± SD of three independent batches of seeds per genotype. * or **, P < 0.05 or 0.01 by Student's t-test. Over the course of seed maturation, increasing levels of ABA lead to the upregulation of ABI3. ABI3 binds to the promoter of ODR1 and represses its expression, resulting in the alleviation of ODR1-imposed inhibition of bHLH57. bHLH57 will then bind to the promoter of NCED6 and NCED9 and elevate their expression levels, leading to higher ABA biosynthesis and seed dormancy. This regulation may be amplified by the stimulation of ABI3 by ABA. Recently, Cao et al. (2020), demonstrated that ODR1/AtSdr4L also controls seed germination through GA biosynthesis in hydrated seeds. Taken together, we propose that ODR1/AtSdr4L plays a negative role in seed dormancy by adjusting the ABA and GA balance during seed maturation and germination. Our data also supports an indirect function for ODR1 on ABA biosynthesis by decreasing the transcription of ABI4 and DREB2C, as well as that of the core dormancy gene DOG1. This indicates that there are still other pathways in ODR1-mediated dormancy that need further investigation.