The GIGANTEA-ENHANCED EM LEVEL Complex Enhances Drought Tolerance via Regulation of Abscisic Acid Synthesis1[OPEN]

Dongwon Baek,a,2 Woe-Yeon Kim,a,b,2 Joon-Yung Cha,a,b Hee Jin Park,c,d Gilok Shin,c Junghoon Park,c Chae Jin Lim,c Hyun Jin Chun,b Ning Li,e Doh Hoon Kim,f Sang Yeol Lee,a Jose M. Pardo,g Min Chul Kim,a and Dae-Jin Yunc,d,3,4 Division of Applied Life Science (BK21 PLUS), Plant Molecular Biology and Biotechnology Research Center, Gyeongsang National University, Jinju 52828, Korea Institute of Agriculture and Life Science, Gyeongsang National University, Jinju 52828, Korea Department of Biomedical Science and Engineering, Konkuk University, Seoul 05029, Korea Institute of Glocal Disease Control, Konkuk University, Seoul 05029, Korea State Key Laboratory of Cultivation and Protection for Non-Wood Forest Trees, Ministry of Education, Central South University of Forestry and Technology, Changsha 410004, China College of Life Science and Natural Resources, Dong-A University, Busan 49315, Korea Instituto de Bioquímica Vegetal y Fotosíntesis, cicCartuja, CSIC-Universidad de Sevilla, Sevilla 41092, Spain

The productivity and distribution of plants are adversely influenced by a variety of abiotic stresses, including drought, high salinity, and extreme temperatures (Zhu, 2016). Global climate change and the resulting water shortages are expected to escalate drought episodes, which would limit plant growth and development (Dai, 2013;Zhu, 2016). Plants have evolved distinct morphological and physiological adaptations that reduce the adverse impact of water shortages (Basu et al., 2016;Gilbert and Medina, 2016;Zhu, 2016). These adaptations are predominantly mediated by endogenous plant hormones, particularly abscisic acid (ABA; Basu et al., 2016;Zhu, 2016;Li et al., 2018). When the plant senses stress signals during periods of dehydration and osmotic stress, the endogenous ABA levels increase to promote stomatal closure, reducing the transpiration rate (Hirayama and Shinozaki, 2007;Cutler et al., 2010;Wu et al., 2019). ABA is not only involved in the response to various environmental challenges, including salinity, freezing, water deficit, wounding, and pathogen attack, but also plays a role in a wide range of developmental processes, such as seed germination, early seedling development, and reproduction (Finkelstein et al., 2002;Huang et al., 2008;Cutler et al., 2010;Cao et al., 2011;Hauser et al., 2011;Qi et al., 2018;Wang et al., 2018).
ABA-mediated signaling is activated or repressed through the regulation of several enzymatic reactions involved in its biosynthesis or degradation (Nambara and Marion-Poll, 2005;Dong et al., 2015;Chen et al., 2020). The first step of ABA biosynthesis takes place in plastids, where b-carotene is converted into xanthoxin, and the final step occurs in the cytosol (Seo and Koshiba, 2002). Epoxidation of all-transzeaxanthin is catalyzed to either 9-cis-violaxanthin or all-transneoxanthin by zeaxanthin epoxidase (Finkelstein, 2013). To produce xanthoxin, the oxidative cleavage by the 9-cis-epoxycarotenoid dioxygenases (NCEDs) is a key regulatory rate-limiting step in ABA biosynthesis following exposure to abiotic stresses (Iuchi et al., 2001;Qin and Zeevaart, 2002;Lefebvre et al., 2006;Martínez-Andújar et al., 2011). In Arabidopsis (Arabidopsis thaliana), the NCED family comprises five enzymes, NCED2, NCED3, NCED5, NCED6, and NCED9, which asymmetrically cleave carotenoids (Schwartz et al., 2003). Most NCED family members play individual regulatory roles in the responses to environmental stimuli and developmental processes (Iuchi et al., 2001;Tan et al., 2003). NCED2 and NCED3 are expressed during root development, whereas NCED5, NCED6, and NCED9 are highly expressed in embryonic plants and induced during seed dormancy Frey et al., 2012). NCED3 is up-regulated upon exposure to drought and high salt stress (Barrero et al., 2006;Endo et al., 2008;Hao et al., 2009), and has been shown to cooperate with NCED5 to enhance stress-induced ABA synthesis (Frey et al., 2012). NCED6 is critical for ABA synthesis under photoreversible seed germination in Arabidopsis as the transcription of NCED6 is induced upon exposure to far-red light . In addition, several transcription factors have been shown to regulate the NCEDs under a variety of growth conditions (Jiang et al., 2012;Lee et al., 2015). WRKY57 induces the expression of NCED3 and RESPONSIVE TO DESICCA-TION 29A (RD29A) by directly binding to the W-box in their promoters (Jiang et al., 2012). The transcription factor NGATHA1 regulates expression of NCED3 by binding to the NGATHA-binding element (Sato et al., 2018), whereas HISTONE ACETYLTRANSFERASE1 (HAT1) acts as a negative regulator by binding to the HB site within the NCED3 promoter (Tan et al., 2018). Another transcription factor, MYB96, directly activates the transcription of NCED2 and NCED6 to modulate both ABA and gibberellin (GA) biosynthesis .
The levels of biologically active ABA are fine-tuned by ABA degradation and sugar-conjugation processes (Dietz et al., 2000;Xu et al., 2002;Kushiro et al., 2004;Saito et al., 2004). Sugar-conjugation represents a major pathway of ABA inactivation (Lee et al., 2006). Chemically modified and biologically inactive ABA can be recycled to rapidly increase the pool of the bioactive hormone. The b-glucosidase encoded by b-glucosidase 1 (AtBG1) hydrolyzes Glc-conjugated ABA into active ABA (Lee et al., 2006). The ABA release in deconjugation processes by AtBG1 regulates both intraand extracellular ABA levels, as well as gene expression in stress responses (Lee et al., 2006;Han et al., 2012). Mutation of AtBG1 leads to reduced levels of bioactive ABA, the increase in stomata number and impaired stomatal closure in the drought stress response (Allen et al., 2019), whereas the overexpression enhances drought tolerance (Han et al., 2012). Among the catabolic pathways, ABA 89-hydroxylation appears to be the regulatory step in a variety of physiological processes (Kushiro et al., 2004). The expression of genes encoding the ABA 89-hydroxylases, CYTOCHROME P450 FAMILY 707 SUBFAMILY A POLYPEPTIDE (CYP707A1) and CYP707A2, is transiently induced after seed imbibition, but is rapidly down-regulated during seed germination (Saito et al., 2004;Okamoto et al., 2006).
The circadian clocks of plants anticipate environmental cues and synchronize physiological responses to occur at the most optimal time of the day. The metabolic pathways of phytohormones are under circadian regulation (Covington et al., 2008;Michael et al., 2008;Grundy et al., 2015;Singh and Mas, 2018). The transcription of genes involved in the biosynthesis of auxin, salicylic acid (SA), jasmonic acid (JA), and ethylene, as well as a large proportion of genes responsive to various abiotic stresses are rhythmically regulated (Yang et al., 2004;Covington and Harmer, 2007;Cheng et al., 2013;Wasternack and Hause, 2013;Kazan, 2015). For instance, the gene encoding the rate-limiting enzyme ACC SYNTHASE 8 is rhythmically expressed for the circadian control of ethylene biosynthesis (Thain et al., 2004), whereas ACC SYNTHASE 6 expression is rhythmically regulated by TIMING OF CAB EXPRES-SION 1 (TOC1), a transcription factor playing as a core oscillator in the circadian clock (Grundy et al., 2015). TOC1 rhythmically regulates JA levels by binding to the promoter of the gene encoding the 13-lipoxyenase enzyme required for JA biosynthesis (Grundy et al., 2015). PSEUDO-RESPONSE REGULATOR 5 (PRR5) and TOC1 also contribute to oscillations in SA levels by binding to the promoters of SA biosynthesis-related genes (Huang et al., 2012;Nakamichi et al., 2012;Liu et al., 2013). ABA biosynthesis and many ABAresponsive genes are under circadian control (Mizuno and Yamashino, 2008;Singh and Mas, 2018  [OPEN] Articles can be viewed without a subscription. www.plantphysiol.org/cgi/doi/10.1104/pp.20.00779 the diurnal variations of ABA content reached a peak during daytime (Grundy et al., 2015;Adams et al., 2018). The diurnal changes of ABA abundance may be necessary for anticipating the diurnal day/night cycle in the regulation of stomata aperture, which in turn affects water consumption and the photosynthetic rate (Nováková et al., 2005;Mizuno and Yamashino, 2008;Grundy et al., 2015). Levels of bioactive ABA are principally regulated by the daily fluctuations of the ABA DEFICIENT 1 (ABA1) and NCED3 gene expression and by the polimerization-mediated activation of AtBG1 in the absence of stress (Lee et al., 2006;Fukushima et al., 2009). In tomato (Solanum lycopersicum), NCED1 showed a noncircadian diurnal accumulation during daytime (Thompson et al., 2000). The ABA-mediated stress response requires the key circadian clock regulators, CIRCADIAN CLOCK-ASSOCI-ATED1, LATE ELONGATED HYPOCOTYL (LHY), and TOC1 (Fukushima et al., 2009;Legnaioli et al., 2009;Adams et al., 2018). TOC1 negatively regulates the circadian expression of the ABA receptors and the H subunit of the magnesium-protoporphyrin IX chelatase (Legnaioli et al., 2009). In addition, TOC1 coordinates drought tolerance and seed germination through its physical interaction with the PHYTOCHROME-INTERACTING FACTORs (PIFs), as well as with several ABA-related components such as DEHYDRATION-RE-SPONSIVE ELEMENT-BINDING 1A and ABA-INSEN-SITIVE 3 (ABI3; Kurup et al., 2000;Kidokoro et al., 2009;Kudo et al., 2017). Moreover, the interaction between PIF7 and TOC1 reduces the circadian clock-associated expression of as DEHYDRATION-RESPONSIVE ELEMENT-BINDING 1C during the drought stress response (Kidokoro et al., 2009). The function of LHY in ABA physiology is complex because LHY partly represses the diurnal expression of NCED3 but at the same time promotes ABA responses, at least in part through the repression of phosphatases ABI1 and ABI2 (Adams et al., 2018). GIGANTEA (GI) was originally isolated as a regulator of the photoperiodic flowering and the circadian clock (Koornneef et al., 1991;Fowler et al., 1999;Park et al., 1999;Mizoguchi et al., 2005). GI functions upstream of CONSTANS, a floral activator that induces FLOWERING LOCUS T (FT) transcription in the circadian clock-controlled flowering pathway under long days (Koornneef et al., 1991;Suárez-López et al., 2001). GI interacts with FLAVIN-BINDING KELCH REPEAT F-BOX 1 (FKF1), a blue light receptor F-box E3 ligase, in the afternoon in a light-dependent manner, and the resulting GI-FKF1 complex targets CYCLING DOF FACTOR 1 (CDF1), a transcriptional repressor of CONSTANS (Imaizumi et al., 2005;Sawa et al., 2007). GI also plays diverse pleiotropic roles in various plant developmental processes, including light signaling, sugar metabolism, and cell wall deposition, as well as abiotic stress responses to oxidative stress, cold, drought, and salinity (Cao et al., 2007;Edwards et al., 2010;Dalchau et al., 2011;Kim et al., 2013;Riboni et al., 2013;Mishra and Panigrahi, 2015). In these responses, GI interacts with a wide range of partner proteins, such as ZEITLUPE for circadian clock regulation, FKF1 and CDF1 for flowering, SPINDLY for light signaling, and SALT-OVERLY SENSITIVE 2 (SOS2) for the salt response (Tseng et al., 2004;Kim et al., 2007;Sawa et al., 2007;Kim et al., 2013). Most recently, GI has been found to promote floral induction via the activation of FT in response to ABA signaling (Riboni et al., 2016). Other components of the circadian clock have been found to be involved in the signal transduction that maintains hormonal balance in response to environmental stresses (Legnaioli et al., 2009;Seung et al., 2012;Lee et al., 2016).
Although GI forms one or more feedback loop(s) with the core clock oscillators to maintain the rhythmicity of the plant circadian clock (Fowler et al., 1999;Park et al., 1999;Swarup et al., 1999;Salomé et al., 2008), and has been shown to participate in various stress signaling pathways (Cao et al., 2005;Penfield and Hall, 2009;Kim et al., 2013;Han et al., 2013;Riboni et al., 2013), there are no reports linking the activity of GI to hormone synthesis. In this study, a yeast two-hybrid (Y2H) analysis revealed that ENHANCED EM LEVEL (EEL), a basic Leu zipper (bZIP) transcription factor involved in ABA-regulated gene expression during seed dehydration, interacts with GI. The GI-EEL complex mediates drought tolerance by activating the diurnal expression of NCED3 to up-regulate ABA biosynthesis. GI and EEL could therefore be targeted using molecular genetics to develop crop plants better able to withstand global climate changes.

GI Interacts with EEL, a bZIP Transcription Factor
GI plays a role in the response to various abiotic stresses, such as high salinity, drought, and low temperatures (Cao et al., 2005;Penfield and Hall, 2009;Han et al., 2013;Kim et al., 2013;Riboni et al., 2013); however, the detailed molecular mechanism(s) by which GI contributes to the drought stress response remain largely unknown. We performed a mating-based Y2H screen of an Arabidopsis complementary DNA (cDNA) library to identify proteins that interact with GI. First, the auto-activation of the reporter genes was tested using a full-length GI (GI full ) and truncated proteins (GI 1-749 , GI 1-391 , GI 543-1173 , and GI 788-1173 amino acids) fused to the GAL4 DNA-binding domain (EEL-BD) of plasmid pGBK7 (Supplemental Fig. S1A). The GI 543-1173 fragment was used for the Y2H screen because it showed no auto-activation activity, whereas the GI full , GI 1-749 , GI 1-391 , and GI 788-1173 proteins exhibited autoactivation in the absence of prey partners (Supplemental Fig. S1B).
The Y2H screening with GI 543-1173 revealed seven putative GI-interacting proteins (Supplemental Table S1). Among them, we selected EEL/AtbZIP12 (At2g41070) for further study because EEL is a homolog of the bZIP transcription factor ABI5. ABI5 has critical roles in ABA signaling and ABA-dependent drought stress response (Kim et al., 2016). Moreover, EEL functions antagonistically with ABI5 to fine-tune the expression of LATE EMBRYOGENESIS-ABUNDANT genes during seed maturation (Bensmihen et al., 2002). To confirm the interaction of the native GI with EEL in the Y2H system, we used the pDEST22 prey (activation domain [AD]) and pDEST32 bait (BD) vector system ( Fig. 1A; Park et al., 2018) in which full-length GI did not show auto-activation. In this assay, the full-length coding regions of GI and EEL were translationally fused to the GAL4 transcription (GI)-AD and EEL-BD, respectively. The yeast cells that were cotransformed with the GI-AD and EEL-BD constructs were able to grow on the synthetic complete medium lacking Trp, Leu, and His and containing 25 mM 3-amino-1,2,4-triazole (3-AT), thus confirming that GI physically interacted with EEL (Fig. 1A). However, GI did not interact with ABI5, a close homolog of EEL (Supplemental Fig. S2). We further confirmed the interaction of GI and EEL by a coimmunoprecipitation assay using total proteins from Agrobacterium tumefaciensmediated Nicotiana benthamiana leaves after transient coexpression of GI and EEL (Fig. 1B). Last, the interaction between GI and EEL in vivo was tested by a bimolecular fluorescence complementation (BiFC) assay in N. benthamiana leaves. The full-length coding regions of EEL and GI cDNAs were fused with sequences encoding the N-terminal ( VN EEL or EEL VN ) and C-terminal fragments ( VC GI or GI VC ) of Venus (eYFP) fluorescent protein, respectively. Following the coexpression of VN EEL and VC GI or EEL VN and GI VC in N. benthamiana leaves, reconstituted fluorescence signals were detected in the nuclei of the leaf epidermal cells (Fig. 1C). Together, these data demonstrate that GI interacts specifically with the bZIP transcription factor EEL in the nucleus.

EEL and GI Are Involved in ABA Biosynthesis
Circadian clock components are essential for seed dormancy through their maintenance of hormonal balance, especially ABA and GA, and are known to affect ABA synthesis and signaling (Penfield and Hall, 2009;Grundy et al., 2015;Adams et al., 2018). Moreover, ABA levels show diurnal rhythms and peak 3 to 4 h after dawn and before dusk in a long-day photoperiod (Grundy et al., 2015;Adams et al., 2018). To determine whether EEL and GI affect the daily ABA Figure 1. Interaction between the GI and EEL proteins. A, Proteinprotein interaction assay using a Y2H system. Prey is the pDEST22 plasmid with the AD domain of GAL4, and Bait is the pDEST32 plasmid with BD domain. Yeast cells cotransformed with GI-AD and EEL-BD were plated on the control synthetic complete (Sc) medium lacking Trp and Leu (Sc-TL) and selective medium Sc lacking Trp, Leu, and His (Sc-TLH) with 25 mM 3-AT. The combinations with empty vector plasmids were used as negative controls. B, Coimmunoprecipitation assay with EEL and GI proteins. Total proteins extracted from N. benthamiana leaves coinfiltrated with GI-GFP and myc-EEL constructs. Input levels of epitope tagged proteins in total protein extracts were analyzed by immunoblotting with antimyc and anti-GFP antibodies. Immunoprecipitated myc-tagged proteins were probed with anti-GFP antibody to detect coimmunoprecipitation of GI-GFP with myc-EEL. C, GI and EEL interaction using BiFC assays in N. benthamiana cells. The VN and VC represent the N-and C-terminal domain of Venus (eYFP), respectively. The GI-EEL complex was localized to the nucleus of the N. benthamiana leaf epidermal cells. Plasmid combinations of VN EEL and VC GI (top) or EEL VN and GI VC (bottom) are indicated above the images. The combinations with empty vector plasmids were used as negative controls. Scale bars 5 100 mm. metabolism, we examined the expression of genes encoding ABA biosynthesis enzymes in the single eel and gi-1, and double eel gi-1 mutants. ABA1, ABA2, ABA3, and NCED3 are key regulators and rate-limiting enzymes of ABA biosynthesis. The reverse transcription quantitative PCR (RT-qPCR) analysis revealed that the expression of NCED3 at Zeitgeber time (ZT)4 (i.e. 4 h after dawn) was significantly down-regulated (;3-fold lower) in the eel, gi-1, and eel gi-1 mutants in comparison with the wild type under normal growth conditions ( Fig. 2A). In contrast, the expression levels of ABA1 and ABA2 were similar in both the wild type and the mutants ( Fig. 2A). Unexpectedly, ABA3 showed reduced expression relative to the wild type only in the eel mutant, but not the gi-1 or the double mutant ( Fig. 2A). Importantly, we found that NCED3 transcripts accumulated gradually during daytime and declined sharply at night, and that this photoperiodic transcription required both EEL and GI ( Fig. 2A). This result led us to check the transcriptional changes of the other NCED family genes in the eel, gi-1, and eel gi-1 mutants. The expression of NCED5 was halved in the eel gi-1 double mutant compared with the wild type, but no statistically significant change in NCED5 expression was observed in the eel and gi-1 single mutants (Supplemental Fig. S3). The expression patterns of the other NCED genes did not differ in any of the genotypes tested (Supplemental Fig. S3). These results suggest that EEL and GI positively coregulate the diurnal expression of NCED3 and NCED5, although the later gene was observed only in the eel gi-1 double mutant.
Because NCED3 expression showed GI-dependent diurnal oscillations and NCED3 function is linked to ABA synthesis (Iuchi et al., 2001), we determined the ABA content in the wild type, gi-1, and GIoverexpressing (GI-OX) seedlings at ZT4, which coincides with the reported diurnal maxima of nonstress ABA in long days (Grundy et al., 2015), and ZT12 when the NCED3 expression was maximal (Fig. 2B). Results showed that the ABA content in seedlings of the gi-1 mutant was significantly reduced relative to the wild type in equal conditions (Fig. 2C). By contrast, GI overexpression had no effect on ABA accumulation at ZT4 and produced a modest increase at ZT12 relative to the wild type. These results indicate that the diurnal accumulation of endogenous ABA is positively regulated by the GI protein, most likely through the regulation of NCED3 transcription. Although the gi-1 mutant showed some degree of stress-induced ABA Figure 2. The diurnal expression of the ABA biosynthesis-related gene NCED3 requires EEL and GI. A, Transcript levels of NCED3, ABA1, ABA2, and ABA3 in wild-type (WT) plants, eel, gi-1, and eel gi-1 mutants. The 10-d-old seedlings grown on one-half strength MS medium under long-day cycles were sampled 4 h after dawn (ZT4) and submitted to total RNA extraction. The transcript levels of NCED3, ABA1, ABA2, and ABA3 were measured using RT-qPCR. The TUBULIN2 was used as an internal control for normalization. Error bars represent the SD from three biological replicates, each with three technical replicates. Asterisks represent significant differences from the wild type by Student's t test (**P # 0.01). B, Transcript levels of NCED3 were analyzed in wildtype plants and gi-1 or eel mutants grown on one-half stength MS medium for 10 d under a long-day photoperiod. Transcript levels were measured using RT-qPCR from total RNA extracted from seedlings at different ZT times. The white and black bars below the plot indicate the light and darkness periods, respectively. TUBULIN2 was used as an internal control for normalization. Error bars represent the SD from three biological replicates, each with three technical replicates. Asterisks represent significant differences from wild type by Student's t test (*0.01 , P # 0.05 and **P # 0.01). C, ABA content in 10-d-old seedlings of wild-type, gi-1, and GI-OX plants grown on one-half strength MS medium under long-day cycles and sampled 4 and 12 h after dawn (ZT4 and ZT12). ABA contents (nanogram per gram of fresh weight [F.W.]) were measured from 20 whole seedlings of each genotype. Error bars represent the SD from three biological replicates, each with three technical replicates. Lowercase letters indicate significantly different values (P # 0.05) determined by one-way ANOVA. synthesis, the total ABA produced under dehydration stress was reduced in the gi-1 seedlings compared with the wild type. The commensurate reduction of ABA content in the gi-1 mutant before and after dehydration suggests that GI is less relevant for the enhanced ABA synthesis elicited by dehydration, which could still be observed in the gi-1 mutant, than for the diurnal production of ABA.
The GI-EEL Complex Activates NCED3 Expression through Binding to the Promoter of NCED3 Several transcription factors are involved in regulating the expression of ABA biosynthesis-related genes to maintain ABA homeostasis (Jiang et al., 2012;Lee et al., 2015). Among them, Arabidopsis activating factor 1 (ATAF1), a NAC transcription factor, transcriptionally regulates NCED3 by binding to the non-ABA response element (ABRE) consensus binding site TTGCGTA (Jensen et al., 2013). To determine whether GI and EEL regulate the transcription of NCED3 directly or indirectly, we examined the physical interaction of the EEL and GI proteins with the promoter of NCED3 in planta. We performed a chromatin immunoprecipitation (ChIP) assay with HAtagged GI-overexpressing (GI-OX) and myc-tagged EEL-overexpressing (EEL-OX) transgenic plants. For this, the NCED3 promoter was divided into six different regions to design amplicons used for ChIP (Fig. 3A). Significantly more amplicon 5 (P5) was precipitated in the GI-OX and EEL-OX plants than in the wild type (Fig. 3, B and C), suggesting that GI and EEL regulate NCED3 expression by binding to the P5 region in the NCED3 promoter. An in silico analysis demonstrated that amplicon P5 contains a cis-acting ABRE (CACGTGGC) regulatory element with a consensus G-box (CACGTG; Fig. 4A). EEL is known to function by directly binding to the ABRE motif in the promoter of LATE EMBRYOGENESIS ABUNDANT1 (EM1; Bensmihen et al., 2002). To determine whether EEL can directly bind to the putative ABRE motif in the NCED3 promoter, we performed an electrophoretic mobility shift assay (EMSA) using EEL fused to GST produced in Escherichia coli. The EEL recombinant protein bound the ABRE motif in the NCED3 promoter (Fig. 4B). To analyze the specificity of this cis-motifbinding activity, we added unlabeled core probes as inhibitors in the EMSA. The nonlabeled oligonucleotides containing the ABRE motif competed with the labeled ABRE probe and reduced their binding to EEL-GST proportionately with the concentration of unlabeled probes added (Fig. 4B). These results indicate that EEL directly binds to the ABRE motif on the promoter of NCED3.
To examine how GI and EEL regulate the transcription of NCED3, transient expression assays were performed using Arabidopsis protoplasts. To make the reporter construct, the promoter region of NCED3 was transcriptionally fused to the upstream region of the GUS gene. In addition, constructs encoding GFPtagged GI and myc-tagged EEL were generated as effector constructs, both under the control of the Cauliflower mosaic virus (CaMV) 35S promoter (Fig. 5A). The luciferase (LUC) gene under the control of the CaMV 35S promoter was used for signal readout normalization. . EEL and GI associate with the NCED3 promoter in vivo. A, Schematic drawing of the NCED3 locus and locations of the ChIP assay amplicons (P1 to P6). The 1,000 bp upstream of the transcription start site on the NCED3 genes was used. B and C, The ChIP assay of the NCED3 chromatin regions associated with GI and EEL. The ChIP assays were performed on nuclear proteins extracted from 10-d-old seedling of wild type (WT) and those of GI-OX (B) or EEL-OX (C) seedlings. Plants were grown on one-half strength MS under long-day conditions. Samples were prepared for the ChIP analysis using an anti-HA (B) or antimyc antibody (C). The immunoprecipitated DNA was amplified using RT-qPCR with specific primers for the amplicons. The TUBULIN2 was used as an internal control for normalization. The fold enrichment is the ratio of GI-OX or EEL-OX to wild-type signal. N.D., Not detected. Error bars represent the SD from three biological replicates, each with three technical replicates. Asterisks represent significant differences from the wild type by Student's t test (**P # 0.01).
The reporter and effector plasmids were cotransformed into the protoplasts, and the GUS and LUC activities were measured. Both of the GI-GFP and EEL-myc proteins activated the NCED3 promoter-driven GUS activity. Cotransformation with the GI-GFP and EELmyc effectors had an additive effect on the transactivation of the reporter compared to EEL-myc alone (Fig. 5B). These results were further supported by transient expression of nontagged GI and EEL proteins in N. benthamiana leaves using the NCED3:GUS construct as the reporter (Fig. 5C). Together, these results indicate that both EEL and GI are able to activate NCED3 transcription.

EEL and GI Enhance Plant Tolerance under Drought Stress Conditions
GI has multiple functions in various plant environmental responses, especially drought and saline stresses (Han et al., 2013;Kim et al., 2013;Riboni et al., 2013;Riboni et al., 2016). By contrast, EEL participates in ABA-regulated gene expression during seed dehydration but has no known role on water-stressed plants (Bensmihen et al., 2002). To characterize the functions of EEL and GI in the drought stress response, the loss-offunction eel and gi-1 mutants and overexpressing transgenic plants were exposed to drought conditions for 11 d, followed by 1 d of rewatering. After rewatering, eel and gi-1 mutants showed a 9.52% and 15.48% survival rate compared with more than 60% in the wild type (Fig. 6, A and C). A different mutant allele, gi-2, also showed the hypersensitive phenotype to drought stress (Supplemental Fig. S4). However, the The probe containing the ABRE binding site motif was biotin labeled for use in the reaction. Unlabeled probes were also included in the reaction as competitors in the specified ratios to the biotin-labeled probe. The arrow indicates the EEL protein and ABRE probe complex. Figure 5. Transcriptional activity assay of GI and EEL. A, A schematic representation of the effector and reporter constructs used in the transient expression assay. B, Protoplasts were isolated from the leaves of 3-week-old Arabidopsis plants, and were cotransfected with the reporter plasmids NCED3:GUS and 35S:LUC, and with one of the effector plasmids (empy vector-GFP, GI-GFP, empty vector-myc, and EEL-myc). The 35S:LUC plasmid was used for signal normalization. The GUS reporter activity in each sample combination is presented as the GUS/ LUC ratio. C, The NCED3 promoter was fused to GUS and coexpressed in N. benthamiana leaves together with different combinations of EEL and GI. The images of GUS staining at top show leaves expressing the indicated constructs. In the middle, the quantification of GUS activity is presented. The bottom shows transcript levels of GI, EEL, or GUS in infiltrated N. benthamiana leaves quantified using RT-PCR. 18S rRNA expression was detected as a loading control. Error bars represent the SD from three independent experiments. Lowercase letters indicate significantly different values (P # 0.05) determined by one-way ANOVA.
overexpression of EEL (EEL-OX; 71.43%) and GI (GI-OX; 69.05%) only weakly enhanced the tolerance of these plants to drought stress in comparison with the wild type (Fig. 6, A and C). Drought stress leads to dehydration because the water lost by transpiration is not replaced. The transpiration rate is therefore used as a physiological parameter associated to the drought tolerance or sensitivity of plants (Basu et al., 2016). To measure the rate of water loss under dehydration stress, rosette leaves of the eel and gi-1 mutants, and EEL-OX, GI-OX, and wild-type plants were detached, and their fresh weights were measured over a 2-h period (Fig. 6, B and D). The detached leaves of the eel and gi-1 mutants lost water more rapidly than wildtype leaves, and EEL-OX and GI-OX genotypes had only marginally decreased water loss (Fig. 6, B and D). The hypersensitivity of eel and gi mutants to drought stress suggested that both EEL and GI positively regulate the drought response.
To specifically characterize the function of the EEL-GI complex in the drought stress response, the eel gi-1 double mutants were also exposed to drought conditions for 9 d, followed by 1 d of rewatering, together with the single eel and gi-1 mutants used as parents. Survival of the double mutants was only marginally worse than that of the single mutants, indicating that the simultaneous loss of EEL and GI proteins had no additive effects and they likely work in the same process (Fig. 7A). To measure the rate of water loss under drought stress, rosette leaves of the two-independent eel gi-1 mutants and wild-type plants were detached, and their fresh weights were measured over a 2-h period (Fig. 7B). The detached leaves of the eel gi-1 double mutant plants lost much more water than the wild type, but again they did not depart from the phenotype of the single mutants (Fig. 7B).

EEL and GI Contribute to ABA Homeostasis and Stomatal Closure in the Drought Stress Response
NCED3 expression is induced by water deficit, and has been associated with plant tolerance during drought stress (Iuchi et al., 2001). To investigate whether EEL and GI affect the transcription of NCED3 during drought stress, 10-d-old seedlings of the wild type, and eel, gi-1, and eel gi-1 mutants were dehydrated for different time periods (0-60 min) on petri-dishes. The results of RT-qPCR analysis indicated that NCED3 expression was rapidly induced in the wild type, but it was much less up-regulated during dehydration in the mutant plants compared with the wild type (Fig. 8A). Time-course leaf ABA contents were also measured to examine whether EEL and GI affected dehydrationinduced ABA levels. ABA accumulation in the eel, gi-1, and eel gi-1 mutants was significantly lower than in the wild-type plants under both normal and dehydration conditions and similar to those resulting from the loss of NCED3 activity (Fig. 8B).
Drought stress induces stomatal closure (Sirichandra et al., 2009). To investigate whether EEL and GI influence drought stress-mediated changes in stomatal physiology, the stomatal patterning and closure responses were determined in the eel, gi-1, and eel gi-1 mutants. The Figure 6. Characterization of the drought responses of the eel and gi-1 mutants. Drought stress response of wild-type (WT), eel, and EEL-OX (A and B) or gi-1 and GI-OX (C and D) plants. The plants were grown in soil with sufficient water for 2 weeks (top in A and C), then water was withheld for 9 d (middle in A and C). The drought stress was then alleviated by rewatering the plants for 1 d (bottom in A and C). The survival rates of the plants were determined from three replicates, each of which involved at least 12 plants. B and D, Water loss by transpiration was measured from detached leaves of 4-week-old wild type, eel, and EEL-OX (B) or gi-1, and GI-OX (D) plants. The water loss at each time point was calculated as a percentage of the initial fresh weight (n 5 10). Error bars represent the SD from three independent experiments. Asterisks represent significant differences from the wild type by Student's t test (*0.01 , P # 0.05 and **P # 0.01).
stomatal density and guard cell sizes were similar in leaves of all genotypes at the same developmental stage (Fig. 8C). The stomatal apertures were also similar in leaves floated on stomata-opening buffer. However, when leaves were exposed to dehydrating conditions, the stomata of the eel, gi-1, and eel gi-1 mutants closed much less than those of the wild type (Fig. 8, C and D). Together, these results indicated that the impaired stomatal closure of the eel, gi-1, and eel gi-1 mutants was mainly caused by their low levels of stress-induced ABA. Therefore, EEL and GI enhance the plant tolerance by regulating ABA homeostasis and stomatal closure in the drought stress response.

DISCUSSION
Various abiotic stresses, such as heat, cold, salinity, and dehydration, affect the circadian expression of stress-responsive genes (Covington et al., 2008;Singh and Mas, 2018). The expression of genes involved in biosynthesis of the phytohormone ABA and regulating drought stress response is under control of the circadian clock (Nambara and Marion-Poll, 2005;Agarwal and Jha, 2010;Basu et al., 2016;Adams et al., 2018). However, how ABA is rhythmically accumulated through diurnal biosynthesis remains poorly understood. In Arabidopsis, the endogenous ABA level peaks during the day (Grundy et al., 2015;Adams et al., 2018), in agreement with our observation of the diurnal expression pattern of NCED3 (Fig. 2). Circadian clock components such as PRR5, PRR7, and TOC1, are indirectly involved in the increase of the ABA levels, whereas LHY functions to repress NCED3 and ABA synthesis (Nakamichi et al., 2010;Huang et al., 2012;Liu et al., 2013;Adams et al., 2018). Here, we have shown that GI, a clock component involved in the regulation of circadian rhythms and photoperiodic flowering, makes a complex with the bZIP transcription factor EEL to regulate the expression of NCED3, the gene encoding a key rate limiting enzyme in ABA synthesis. NCED3 showed a diurnal oscillation in which the transcript accumulated during daytime and declined at night (Fig. 2). The expression during daytime was strictly dependent on GI and EEL (Fig. 2). Indeed, NCED3 expression recapitulates the diurnal pattern of GI protein abundance (Yu et al., 2008), suggesting that GI activity contributes to the circadian amplitude of NCED3 expression and ABA oscillations. This is consistent with the known role of GI in gating the light input into the photoperiodic pathway of flowering (Imaizumi et al., 2005;Sawa et al., 2007). GI is not a DNA-binding protein per se but influences gene expression through the interaction with DNA-finding transcription factors that recruit GI to specific gene promoters (Imaizumi et al., 2005;Sawa et al., 2007;Fornara et al., 2009;Kubota et al., 2017). Here, we show that GI and EEL interact to target the NCED3 gene promoter to gate the light information that dictates the diurnal oscillations of NCED3 and endogenous ABA synthesis. We suggest that EEL provides the target specificity for the NCED3 promoter and that GI cooperates with EEL in gating the light input in the transcriptional regulation of NCED3.
NCED3 expression is also highly responsive to dehydration and contributes to the stress-induced ABA synthesis (Iuchi et al., 2001;Chang et al., 2020), and the abundance of NCED3 transcripts and ABA contents were reduced in the gi and eel mutants under dehydrating conditions (Fig. 8). The overall reduction in ABA content in the gi-1 and eel mutants correlated with the dehydration-sensitive phenotype (Fig. 8). However, the gi-1 and eel mutants retained some ability to induce NCED3 expression and ABA synthesis upon dehydration treatment, in agreement with the known regulation Figure 7. Characterization of the drought stress responses of the eel gi-1 double mutants. A, Wild-type (WT) plants, and eel, gi-1, and eel gi-1 mutants were grown in soil with sufficient water for 3 weeks (top), then water was withheld for 9 d (middle). The drought-stressed plants were then rewatered for 1 d (bottom), after which their survival rates were assessed. Each experiment comprised at least 12 plants, and three replicates were performed. B, Water loss by transpiration was measured from the leaves of wild-type plants, eel, gi-1, and eel gi-1 mutants. The shoots of 3-week-old plants were detached, and their water loss at each time point was calculated as a percentage of their initial fresh weight (n 5 10). Error bars represent SD from three independent experiments. Asterisks represent significant differences from the wild type by Student's t test (*0.01 , P # 0.05 and **P # 0.01).
of NCED3 by additional factors (Jiang et al., 2012;Adams et al., 2018;Sato et al., 2018;Tan et al., 2018). Together, these results imply that the EEL-GI complex is principally required for the regulation of the diurnal fluctuations of ABA contents by gating the light input while contributing to the amplification of the dehydration signal.

GI Regulation of ABA Metabolism and Stress Responses
ABA is generally considered to be a floral repressor, in contrast with GAs that are flowering accelerators (Blazquez et al., 1998;Conti et al., 2014). Exogenous ABA treatment inhibits flowering by reducing the expression of FT, a floral integrator (Blazquez et al., 1998;Domagalska et al., 2010), whereas endogenous ABA promotes flowering via the up-regulation of FT, as part of the drought response (Riboni et al., 2013). Short-term drought or water shortage promotes the floral transition as a drought-escape (DE) response via the up-regulation of FT to avoid prolonged exposure to drought (Riboni et al., 2013). The DE response does not occur under short-day conditions or in the gi mutant, indicating that DE requires GI and the expression of its downstream targets FT and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (Riboni et al., 2013). These observations all suggest that there is a molecular cross talk between ABA signaling and the photoperiodic pathway to flowering.
Several regulatory components of the circadian rhythm are involved in the regulation of the signaling pathways of diverse stresses (Franks et al., 2007;Legnaioli et al., 2009;Penfield and Hall, 2009). GI, a key regulator of the photoperiodic flowering and the circadian clock, also plays important roles in the responses to various stresses, including cold, drought, salt, and oxidative stress (Kurepa et al., 1998;Cao et al., 2005;Kim et al., 2013;Riboni et al., 2013). GI is a negative regulator in salt stress signaling via the inhibitory interaction with the SOS2 protein kinase that is essential Figure 8. The expression of NCED3, ABA levels, and stomatal closure in wild-type (WT) plants, the eel, gi-1, and eel gi-1 mutants under drought stress condition. A, Transcript levels of NCED3 in wild-type plants, and of eel, gi-1, and eel gi-1 mutants over 1-h dehydration stress. The 10-d-old seedlings grown on one-half strenght MS medium under long-day cycles were sampled at ZT4 (control nontreated sample) and again 30 and 60 min after a dehydration treatment. NCED3 transcript levels were measured using RT-qPCR. The expression of TUBULIN2 was used as an internal control for normalization. Error bars represent the SD of three biological replicates, each with three technical replicates. B, ABA content in seedlings treated as in A. ABA contents were measured from 20 whole seedlings of each genotype. Error bars represent the SD from four independent experiments. C, The rosette leaf epidermis of wild-type plants, eel, gi-1, and eel gi-1 mutants were floated in stomatal opening solution for 2-h, and then removed and placed onto filter paper for 1 h for the dehydration treatment. Stomata on the abaxial surface were observed using scanning electron microscopy. Scale bars 5 10 mm. D, Measurement of stomatal apertures (width/length) in wild-type plants, eel, gi-1, and eel gi-1 mutants before and after dehydration for 1 h. Error bars represent the SD from three independent experiments, with at least 30 stomata measured per genotype and per treatment. Asterisks represent significant differences from the wild type by Student's t test (*0.01 , P # 0.05 and **P # 0.01).
for the activation of the Na 1 /H 1 antiporter SOS1 (Quintero et al., 2011;Kim et al., 2013). Salinity promoted the proteasomal degradation of GI, with the subsequent release of SOS2 and the activation of SOS1 . The SOS pathway is primarily involved in counteracting sodicity stress, independently of the water stress imposed by high salinity, and is considered an ABA-independent response (Xiong et al., 2002;Ji et al., 2013). Therefore, although salt and drought stresses both enhance the levels of endogenous ABA and NCED3 expression, GI could have different functions in the salt and drought stress responses. First, gi mutants show enhanced salt-tolerance but are hypersensitive to drought stress and lack the early flowering likened to the drought escape response (Han et al., 2013;Riboni et al., 2013). This could indicate that GI protein integrity must be preserved under drought stress because GI must accumulate to promote flowering. Prior research had shown that GI is involved in the drought stress response, but the underlying mechanism was not fully understood except that GI interacted with micro RNA172 to regulate the expression of the gene encoding the WRKY44 transcription factor (Han et al., 2013). Here we show that GI, together with EEL, promoted the diurnal expression of NCED3 and mediated stomatal closure in the drought stress response (Fig. 8, C and D). Although the clock components PRR5, PRR7, and TOC1 are also involved in the control of stomatal aperture and expression of ABA responsive genes, how these clock elements regulate the rhythmicity of ABA biosynthesis remains largely unknown (Grundy et al., 2015). Together, these findings suggest that core clock components are intimately associated with plant responses to abiotic stresses.
Our results reveal that GI interacts with a bZIP transcription factor, EEL (Fig. 1), and that the GI-EEL complex binds to the NCED3 promoter region containing an ABRE to induce its expression for de novo ABA biosynthesis (Figs. 2, 3, 4, and 8). NCED5, but none of the other NCEDs (NCED2, NCED6, and NCED9), appear to be influenced by the GI-EEL complex (Supplemental Fig. S3). This is consistent with the known role of NCED5 to enhance stress-induced ABA synthesis in addition to NCED3 (Frey et al., 2012). The results of the transcriptional activation of the NCED3 promoter by GI and EEL proteins using Arabidopsis protoplast and N. benthamiana agro-infiltration systems (Fig. 5) suggest that GI and EEL act as positive effectors in a transcriptional activator complex. The GI protein interacts with other transcription factors in the photoperiodic flowering pathway, such as CDF1, FLOW-ERING BHLH, FKF1, and TEOSINTE BRANCHED 1/CYCLOIDEA/ PROLIFERATING CELL NUCLEAR ANTIGEN FACTOR 4 (Imaizumi et al., 2005;Sawa et al., 2007;Fornara et al., 2009;Kubota et al., 2017). Although GI formed complexes with these transcription factors to modulate their activity, GI did not bind directly to the DNA of target genes (Imaizumi et al., 2005;Sawa et al., 2007;Fornara et al., 2009;Kubota et al., 2017). The expression pattern of NCED3 under the long-day condition is similar to the steady accumulation of the GI protein during daytime (Yu et al., 2008). Thus, NCED3 expression was induced when the GI expression was started at ZT4, and remained constant thereafter. The rhythmical fluctuations of ABA are also known to be regulated by the PRR5, PRR7, and PRR9 clock components (Fukushima et al., 2009). The loss-of-function mutant of LHY showed an altered rhythmical accumulation of ABA, with a reduction of ABA content at dusk (Adams et al., 2018). It has been suggested that LHY may repress light-dependent NCED3 expression (Adams et al., 2018). Together, our data show that GI and EEL stimulate diurnal ABA biosynthesis and plant drought tolerance by upregulating the transcriptional expression of NCED3, but whether the EEL-GI complex operates to relieve inhibition by LHY is presently unknown.

A New Role for EEL in ABA Biosynthesis during Drought Stress
The bZIP transcription factors in Arabidopsis are reported to regulate the expression of genes involved in various abiotic stress responses (Yang et al., 2009;Alves et al., 2013;Kim et al., 2015). The bZIP family includes 75 distinct members classified into 13 groups (A to L, and S) according to their sequence similarity and functions (Kim, 2006). Group A genes are involved in ABA signaling, and are divided into two categories, the ABI5/AtDPBF family members (ABI5, EEL, DPBF2/ AtbZIP67, DPBF4, and AREB3) and AREB/ABF family members (AREB1/ABF2, AREB2/ABF4, ABF1, and ABF3; Choi et al., 2000;Bensmihen et al., 2005;Fujita et al., 2005). The ABI5/AtDPBF family members, including EEL, transcriptionally regulate systems mediating ABA-dependent stress signaling during seed maturation and developmental processes (Finkelstein and Lynch, 2000;Bensmihen et al., 2005). Accordingly, EEL is strongly expressed in seeds, where EEL functions as either a homodimer or in a heterodimer complex with ABI5 to interact with the cis-acting regulatory element ABRE of genes such as EM1 and EM6, during embryo maturation (Bensmihen et al., 2002;Carles et al., 2002). However, EEL is also expressed in other plant tissues at lower levels (https://www. arabidopsis.org/), and our RT-qPCR analysis showed the presence of EEL transcripts in vegetative tissues, including root, rosette leaves, cauline leaves, stem, and flowers, although the levels were low (Supplemental Fig. S5). Moreover, EEL regulated the expression of STAYGREEN1 (SGR1) in the chlorophyll degradation pathway during leaf senescence (Sakuraba et al., 2016). ABI5, which shows a preferential expression in seeds like EEL, was also involved in several abiotic stress responses of whole plants, such as drought, salt, and high temperature (Lim et al., 2013;Song et al., 2013;Skubacz et al., 2016;Chang et al., 2019). Here we show that the loss-of-function mutant eel exhibited drought hypersensitivity (Figs. 6A and 7A), and was found to contain lower levels of endogenous and stress-induced ABA (Fig. 8B), and faster water loss upon dehydration than the wild type (Figs. 6B and 7B). In addition, the significant decrease of NCED3 expression in the eel mutant indicates that EEL positively controls ABA biosynthesis by acting as a transcriptional activator of NCED3 (Figs. 2 and 8A). Although the NCED3 promoter contains two putative ABRE cis-acting regulatory elements (Supplemental Fig. S6; Baek et al., 2017), EEL was associated only with the ABRE site in the P5 region of the NCED3 promoter in our EMSA and ChIP experiments (Figs. 3 and 4). Although EEL and ABI5 can associate as either homodimers or heterodimers (Bensmihen et al., 2002), we observed that ABI5 was not able to bind to the ABRE on the NCED3 promoter, suggesting that only EEL induces NCED3 expression specifically (Supplemental Fig. S7). Apart from EEL, other transcription factors contribute to regulate NCED3 expression according to various consensus binding sites and conditions. For example, ATAF1, a NAC transcription factor, regulates NCED3 transcription by binding to the non-ABRE consensus binding site TTGCGTA (Jensen et al., 2013), that is, AtAF1 and EEL transcription factors use different binding sites in the NCED3 promoter. In addition, AtAF1 is related to plant growth and flowering time, whereas EEL is involved in seed germination and, as we show here, the dehydration stress response of seedlings and mature plants. Although most NCED family members have a few putative ABRE and/or ABRE-like cis-acting regulatory elements on their promoters, the expression levels of these other genes do not seem to be largely affected by EEL, indicating that EEL regulates NCED3 specifically.

CONCLUSIONS
In summary, we have shown that the GI-EEL complex regulates the diurnal oscillation of ABA biosynthesis by means of the transcriptional activation of NCED3. Overall ABA contents after dehydration stress were also reduced in eel, gi-1, and eel gi-1 mutants, which were all hypersensitive to drought stress. In addition, GI and EEL act together to regulate stomatal closure. Plants regularly experience basal levels of water deficiency by evapotranspiration on a daily basis, and circadian clock-controlled ABA biosynthesis and the resulting stomatal closure after dawn are essential preemptive measures for maintaining water homeostasis. This study shows that GI, a circadian clock component and flowering time regulator, is also essential for plant acclimation to daily water demands by elevating the amount of endogenous ABA in cooperation with the transcription factor EEL. Collectively, the interdependence of ABA signaling and the circadian clock highlights an adaptive strategy to deal with recurrent daily strains and adverse environments.

Y2H Screen and Interaction Assay
To identify GI-interacting proteins, a Y2H screen was performed using the Matchmaker Gold Yeast Two-Hybrid System (Takara Bio), which is based on the mating of two haploid yeast strains that independently express the bait and prey fusion proteins. The full-length and truncated GI sequences were amplified using PCR and cloned into the pGBK7 bait vector (Supplemental Fig. S1A). These constructs were transformed into Saccharomyces cerevisiae Y187 strain used in the yeast mating protocol. Only the truncated protein GI 543-1173 fragment could be used for the Y2H screen because it showed no auto-activation activity. To confirm the protein-protein interactions found in the library screen, the full-length GI or EEL sequences were cloned into the pDEST22 prey vector (GI-AD) or pDEST32 bait vector (EEL-BD) and cotransformed into the yeast cells ( Fig. 1A; Park et al., 2018). Of note is that full-length GI did not show autoactivation in this alternative Y2H system. Protein-protein interactions were determined by the growth of yeast colonies on the synthetic complete medium lacking Trp and Leu or Trp, Leu and His (Takara Bio) agar media containing 5bromo-4-chloro-3-indolyl-b-d-galactopyranoside acid (40 mg mL 21 ) or 3-AT (25 mM).

Coimmunoprecipitation Assays
The leaves of 3-week-old Nicotiana benthamiana plants were coinfiltrated with Agrobacterium tumefaciens carrying 35S:GI-GFP and 35S:myc-EEL together with the p19 plasmid (Park et al., 2018). Total proteins were extracted from coinfiltrated leaves and reacted for immunoprecipitation using antimyc antibody (Roche) and protein A agarose (Invitrogen). For immunoblotting, membranes were incubated with the appropriate anti-GFP (Abcam), and detected using ECL-detection reagent (GE Healthcare). The coimmunoprecipitation assays were performed in three independent replicates.

Plant Materials and Growth Conditions
The Arabidopsis (Arabidopsis thaliana) eel mutant (SALK_021965), gi-1 mutant, and CaMV 35S promoter driven GI-OX transgenic plants (ecotype Col-0; Kim et al., 2007) were used in this study. The eel gi-1 double mutants were generated by crossing gi-1 with eel, and then isolated in the F2 progeny by diagnostic PCR. Plants were grown on one-half strength Murashige and Skoog (MS) media (1.5% [w/v] Suc, 0.6% [w/v] agar, pH 5.7) at 23°C. For the germination assay, the seeds were sown on a one-half strength MS agar medium supplemented with different concentrations of ABA, and 5-d-old seedlings with green cotyledons were scored as resistant to ABA inhibition. For the drought treatments, water was withheld from 3-week-old plants for 9 d, and their survival ratio was measured on day 10 after 1 d of rewatering. The drought experiments were performed for five independent replicates, each using at least 12 plants.

RT-qPCR Analysis
Total RNA was isolated from 10-d-old seedlings using a RNeasy Kit (Qiagen) following the manufacturer's instructions. The RNA was treated with DNase I (Qiagen) to remove contamination from genomic DNA. For the RT-qPCR analysis, the first-strand cDNA was synthesized from 1 mg of total RNA using a cDNA synthesis kit (Thermo Fisher Scientific). The QuantiSpeed SYBR No-Rox Mix (PhileKorea) was used for the RT-qPCR reactions as follows: 50°C for 10 min, 95°C for 2 min, and 50 cycles of 95°C for 5 s, and 60°C for 30 s. TUBULIN2 expression was used for normalization. The relative expression levels of all samples were automatically calculated from three biological replicates using the CFX Manager software program (Bio-Rad Laboratories). The RT-qPCR analyses were performed in three biological replicates, each with three technical replicates. The primers used for the RT-qPCR analyses are listed in Supplemental Table S2.

Generation of Transgenic Plants
To generate EEL-overexpressing transgenic plants, the full-length cDNA of the EEL gene was inserted into the pGWB17 vector (with myc tag) under the control of the constitutive CaMV 35S promoter using the gateway system (Nakagawa et al., 2007;Ali et al., 2018). The primers used in the PCR are listed in Supplemental Table S2. The construct was introduced into A. tumefaciens GV3101, then transformed into the wild-type plants by floral dipping. Transgenic plants were selected for hygromycin resistance, and their genotypes were confirmed using RT-PCR. The homologous T 3 generation plants were used for further experiments.

ChIP Assay
The ChIP assays were performed as described by Saleh et al. (2008) using nuclear proteins extracted from the leaves (100 mg) of 3-week-old wild type, EEL (fused myc tag)-overexpressing, and GI (fused GFP tag)-overexpressing plants. Monoclonal antimyc (Cell Signaling Technology) or monoclonal anti-GFP (Thermo Fisher Scientific) antibodies were used for the immunoprecipitation. The amount of immunoprecipitated DNA was quantified using RT-qPCR. The ChIP assays were performed in three independent replicates. The primers used in the ChIP assays are listed in Supplemental Table S2.

EMSA
The EMSA was performed using the Lightshift Chemiluminescent EMSA kit (Thermo Fisher Scientific) according to the manufacturer's instructions (Yang et al., 2018). The probes were labeled with the 39 end biotin (Cosmo Genetech), oligonucleotides spanning the ABRE binding site motif on the NCED3 promoter. The DNA binding took place in a 20-min reaction at 25°C in binding buffer (10 mM Tris pH 7.5, 50 mM KCl, 1 mM dithiothreitol) containing 50 mM KCl, 0.05% (w/v) Nonidet P-40, 5 mM MgCl 2 , 10 mM EDTA, 2.5% (w/v) glycerol, 50 ng mL 21 of poly (dI-dC), and various concentrations of purified bacterially expressed GST-EEL protein. For the competition assay, 2-, 5-, and 10-fold amounts of unlabeled probe were incubated with the GST-EEL protein before the labeled probe was added to the reaction. The reaction mixture was subjected to electrophoresis on a 6% (w/v) polyacrylamide gel in 0.5 3 TBE buffer at 100 V for 2 h, transferred onto a nylon membrane, and then crosslinked. The biotin-labeled DNA was detected using chemiluminescence (Thermo Fisher Scientific). The EMSA experiments were performed in three independent replicates.

Analysis of Transcriptional Activity
The plasmids indicated in the figure legends were introduced into protoplasts obtained from 3-week-old Arabidopsis wild-type plants using polyethylene glycol-mediated transformation (Baek et al., 2013). The expression of the fusion constructs was monitored and imaged using a Zeiss Axioplan fluorescence microscope (Carl Zeiss), and the transcriptional activity of the EEL or GI proteins was analyzed in the protoplasts as described previously (Baek et al., 2013). The fluorescence was measured using a SpectraMax GEMINI XPS spectrofluorometer (Molecular Devices) and SoftMax Pro-5 software (Molecular Devices). The GUS activity was normalized to the LUC activity to eliminate experimental variation between samples. Each experiment was replicated three-independent times.

Gravimetric Water Loss Assay
The shoots of 4-week-old plants were detached from the root and weighed immediately. The shoots were placed on a plate at room temperature and weighed at various time intervals. The loss of fresh weight was calculated as a percentage of the initial weight of the plant. At least five biological replicates were performed for each sample.

Stomatal Aperture Assays
Three or four leaves of 10-d-old seedlings were detached and floated on stomatal opening buffer (5 mM MES, 5 mM KCl, 50 mM CaCl 2 , pH 5.6) under light conditions for 3 h. And then, to treat drought stress, leaves samples treated with dehydration for 1 h using filter paper for air dry. After drought stress treatment, the leaves were sequentially fixed by 2.5% (v/v) glutaraldehyde and 1% (w/v) OsO 4 in the dark condition. Images of stomata were captured by scanning electron microscopy (JSM-6380LV; JEOL, Akishima, Japan). The stomatal aperture was determined from measurements of 40 to 60 stomata per treatment. Each experiment was replicated three times.

Measurement of ABA Content
Endogenous ABA was extracted from 10-d-old seedlings (100 mg) and analyzed using a Phytodetek ABA test kit (Agdia), following the manufacturer's protocols. At least three biological repeats and two technical repeats were performed for each sample.

Statistical Analyses
The statistical analyses including Student's t test were performed by using the Excel 2010 program (Microsoft). The RT-qPCR analyses were performed three-independent experiments the average values of 2 DCT were used to determine the differences, and the data indicated as means 6 SD. A significant difference was considered at 0.01 , P-value # 0.05 and P-value # 0.01. Where indicated, ANOVA by one-way ANOVA (MS Excel software) with Tukey's honestly significant difference (HSD) mean-separation test of significance for each experiments (P-value # 0.05) was applied.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Auto-activation between the GI protein and the GAL4 AD in the Matchmaker Y2H screen system.
Supplemental Figure S2. Y2H assay of GI and ABI5 proteins.
Supplemental Figure S3. The expression of NCED family genes in wild type plants, eel, gi-1, and eel gi-1 mutants.
Supplemental Figure S4. Characterization of the drought stress responses of the gi-2 mutant.
Supplemental Figure S5. The expression of NCED3, GI and EEL in various tissues of Arabidopsis.
Supplemental Figure S6. Putative ABRE cis-acting regulatory elements in the promoters of the NCED family.
Supplemental Figure S7. EMSA using ABI5 and the ABRE binding site motif in the NCED3 promoter.
Supplemental Table S1. Summary of GI-interacting proteins revealed in a Y2H screen.
Supplemental Table S2. List of primers used in this study.