The ARF2–ANT–COR15A gene cascade regulates ABA-signaling-mediated resistance of large seeds to drought in Arabidopsis

Seed size in higher plants is an important trait with respect to ecology and agriculture. Generally, larger seeds are less easily dispersed, but offer the germinating seedling a larger supply of nutrients, thus increasing its competitiveness during seedling establishment and tolerance to adverse environmental stresses (Westoby et al., 1992). By contrast, smaller seeds are more easily colonized, thus giving rise to population dispersal and spread. As for the mother plant, nutrient resources for producing seeds generally are limited, thus causing a trade-off between the number and size of the seeds produced (Venable, 1992; Ohto et al., 2005; Coomes and Grubb, 2003; Orsi and Tanksley, 2009). In agriculture, increasing seed size has been a crucial contributor to the yield increases in crop plants during domestication (Shomura et al., 2008).

The seed size depends on the development of the embryo, endosperm and seed coat tissues, which are derived from distinct cells of the ovule and have distinct complements of maternal and paternal genomes. The embryo constitutes the major volume of a mature seed in Arabidopsis, and the changes in seed mass are reflected in the size of the embryos. Thus, the mature seed size is largely affected by both the embryo cell number and cell size in Arabidopsis. Furthermore, seed size is often able to alter in an intraspecific manner in response to environmental cues (Ohto et al., 2005). Previous studies have found that the loss-of-function mutants of apetala2 (ap2), auxin response factor2 (arf2), short hypocotyl under blue1 (shb1) and da1-1 give an increased seed-mass phenotype in Arabidopsis through enlargement of the embryonic cell size or an increase in the embryonic cell number (Ohto et al., 2005; Schruff et al., 2006; Zhou et al., 2009; Li et al., 2008). By contrast, the mutants of miniseed3 and iku2 exhibit a smaller seed mass owing to a decrease in cell number, but not cell size (Luo et al., 2005). However, relatively little is known about the specific regulatory mechanisms that underlie these genes that mediate seed development in response to environmental inducers.

Phytohormone abscisic acid (ABA) is broadly involved in developmental regulation and various stress-related responses in higher plants. The germinating seedlings at different developmental stages exhibit different physiological responses to ABA signals in Arabidopsis, and the ABA-deficient mutants have significantly reduced cell vigor (Wang et al., 2011). Also, different concentrations of ABA can regulate the root and stem growth by boosting or suppressing cell proliferation and differentiation during Arabidopsis seedling development (Wang et al., 2011; Zhang et al., 2010). In addition, a number of genotypes with mutations in DNA replication machinery display a hypersensitive response to ABA during seed germination and seedling growth (Yin et al., 2009), strongly implying that ABA signaling suppresses cell proliferation by regulating DNA-replication-related proteins. In particular, ABA is a key regulator of plant responses to environmental cues, such as drought, cold and salt stresses (Guo et al., 2004). Usually, these environmental stresses induce ABA accumulation, thereby triggering many physiological actions in response to these stresses (Lim et al., 2007). For example, under water-deprived conditions, ABA can induce stomatal closure, leading to a decrease of transpirational water loss (Assmann and Wang, 2001).

The transcription factor Auxin Response Factor 2 (ARF2), originally identified as an ARF1-binding protein (therefore formerly known as ARF1-BP), binds to the AuxREs (TGTCTC) cis-element in the promoter region of auxin-regulated genes (Ulmasov et al., 1997). ARF2 has been identified as a regulator that is involved in negatively regulating ABA-mediated seed germination and primary root growth by binding to homeodomain gene HB33 (Wang et al., 2011). The lost-of-function mutant arf2 exhibits a bigger seed size phenotype, compared to the wild type, because of the extra cell proliferation (Schruff et al., 2006). At the same time, studies have revealed that the transcription factor AINTEGUMENTA (ANT), a member of the AP2-domain family, is involved in mediating cell proliferation and growth control (Klucher et al., 1996; Mizukami and Fischer, 2000; Nole-Wilson and Krizek, 2000; Krizek, 2003). The overexpressing transformants of the ANT gene also display a phenotype of larger seed size, compared to controls (Mizukami and Fischer, 2000).

Compounding evidence suggests that seedlings that have been germinated from larger seeds have an increased tolerance to abiotic stresses compared to seedlings from smaller seeds (Coomes and Grubb, 2003; Ohto et al., 2005; Moles et al., 2005; Orsi and Tanksley, 2009). However, the molecular mechanisms underlying why large-seeded seedlings give stronger tolerance to abiotic stresses are not fully understood. In this study, we found that ARF2 mediates signaling crosstalk between the drought stress response and seed development by negatively regulating ANT. Furthermore, we identified that the target gene COR15A is directly regulated by ANT. Combining the molecular and genetic evidence, we show for the first time that the ARF2–ANT–COR15A pathway forms the ABA-mediated signal cascade that regulates seed mass and drought tolerance. This study provides new evidence to enhance understanding of why large-seeded seedlings exhibit stronger tolerance to drought.

The mature seeds of arf2 mutants were observed to be obviously bigger in size than their wild-type counterparts (Fig. 1B,D; Fig. S1A,C), consistent with a previous report (Schruff et al., 2006). Cytological observations showed that the average areas of the cotyledon embryo and the cotyledon embryo cell in the arf2 mutant were ∼1.6 and ∼1.3 times larger than those of the wild type, respectively (Fig. 1A,C; Fig. S1B,D); the corresponding ratio of the area of the embryo to that of the cell was 1.23 (1.6/1.3=1.23), suggesting that the overall enlargement of arf2 mutant embryos might result from both the augmentation of cell size and the increase in cell number. Furthermore, the abnormal size of the cotyledons of the arf2-6 mutant could be restored on Murashige and Skoog (MS) medium with 3.0 µm ABA (see Fig. 1E). And, cytological observations showed that the average size of the cotyledon cells in the arf2-6 mutant was smaller than that of wild type upon treatment with 3.0 µm ABA (see Fig. 1E; Fig. S2). These observations strongly imply that the function of ARF2 in seed development might be involved in the ABA signaling pathway. Using the ELISA method for quantifying the endogenous ABA content in developing seeds of the arf2-6 mutant and wild type, we found that the endogenous ABA content in the arf2-6 mutant was significantly higher (see Fig. 4D), suggesting that the large seeds of the arf2-6 mutant are likely to be the result of the abnormal ABA signals.

Generally, ABA signaling is one of main factors that induces plant drought tolerance. To investigate the response of arf2 mutants to drought stress, the arf2 and wild-type plants were grown for two weeks in soil and subsequently subjected to water deprivation for the indicated number of days. Initial water-deprivation experiments showed that the leaves of both arf2 and wild-type plants remained green (Fig. 2A; Fig. S3A). Under water deprivation for the given number of days, most wild-type plants dried up and died, whereas the arf2 plants remained turgid and retained their green leaves (see Fig. 2A and B; Fig. S3A). Consistent with these results, the detached leaves of the arf2 plants lost water more slowly than those of the wild-type plants (Fig. 2C). Furthermore, the enhanced water-deficit survival of arf2-6 plants was closely associated with their capacity to maintain higher leaf relative water content (RWC) than the wild type at ∼16% soil water content (SWC) (Fig. S3B). The stomatal closing of the leaf surfaces in the arf2 plants was obviously increased under treatment with ABA (Fig. 2D). This finding was consistent with the expression of β-glucuronidase under the ARF2 promoter (ProARF2:GUS) on the stomata (Fig. 3F), and seed germination and seedling root growth responded in a more sensitive manner to ABA in the arf2 plants (Wang et al., 2011). Upon exposure to 10% polyethylene glycol (PEG) 6000, commonly used to mimic drought tolerance under controlled conditions, the ABA content became significantly higher in the arf2-6 mutant, as shown in Fig. 4D. Thus, the more sensitive stomatal closure in the arf2 mutant was most likely caused by abnormal ABA accumulation. However, the stomatal density and size were not significantly different between the arf2 and wild-type plants (Fig. S3). Taken together, these results indicate that ARF2 is involved in regulating the responsiveness of ABA and that the drought tolerance of arf2 mutants is caused by the ABA-mediated stomatal closure.

The mature seeds of the 35S:ANT (ANT-overexpressing) transformants were observed to be obviously bigger in size, consistent with a previous report (Mizukami and Fischer, 2000), whereas ant-knockout (ant-KO) mutants exhibited smaller seeds, compared to wild type (see Fig. 1B,D; Fig. S1A,C). Cytological observations showed that the average area of the cotyledon embryo and the cotyledon embryo cell in the 35S:ANT plants were ∼1.50 times and ∼1.28 times larger than the wild type, respectively (see Fig. 1A and C); the corresponding ratio of the area of the embryo to that of the cell was 1.17 (1.5/1.28=1.17). Similarly, this observation meant that the overall enlargement of the 35S:ANT embryos could have resulted from both an augmentation in cell size and an increase in cell number. The enlarged cotyledons of the 35S:ANT transformants could also be restored on MS medium with 3.0 µM ABA (see Figs 1E and 4A). Cytological observations showed that the average size of the cotyledon cells in the 35S:ANT transformants was also smaller than that of wild type (see Fig. S4). Consistent with this, the seed germination and seedling root growth of the 35S:ANT transformants were more sensitive to ABA signals, whereas the ant-KO mutant was less sensitive, compared to the wild type (see Fig. 4A–C). When quantifying the endogenous ABA content in developing seeds of the 35S:ANT-transformant and ant-KO mutant lines, we found that the endogenous ABA content in the 35S:ANT transformants and ant-KO mutant (SALK_022770) was significantly higher and lower than the wild type, respectively (see Fig. 4D; Fig. S1E). These observations indicate that the function of ANT is also connected to the ABA signaling pathway and that the large seeds of 35S:ANT plants are most likely to be the result of the abnormal ABA signals.

While testing the drought tolerance of the 35S:ANT and wild-type plants, the physiological changes and survival status of leaves upon water deprivation were observed. Similar to the phenotype of the arf2 mutants, the 35S:ANT plants displayed a stronger tolerance to drought than the wild type (Fig. 2A and B; Fig. S3A). Consistent with this observation, the detached leaves from the 35S:ANT plants lost water more slowly (Fig. 2C); furthermore, the enhanced water-deficit survival of the 35S:ANT plants was closely associated with their capacity to maintain higher leaf RWC than the wild type at ∼16% SWC (Fig. S3B). Further, the performance of stomata was analyzed, and the stomatal aperture was obviously reduced in the 35S:ANT plants and enhanced in the ant-KO mutant plants upon treatment with ABA (Fig. 2E). The analysis of the expression of β-glucuronidase under the ANT promoter (ProANT:GUS) indicated the ANT was preferentially expressed in the stomata (Fig. 3F). Accordingly, seed germination and seedling root growth were more responsive to ABA treatment in the 35S:ANT plants, but less sensitive in the ant-KO plants, compared to the wild type (Fig. 4A–C). Upon exposure to 10% PEG 6000, the ABA content became significantly higher in the 35S:ANT seedlings, as shown in Fig. 4D. Thus, drought tolerance of the 35S:ANT plants was most likely caused by abnormal ABA signaling. Overall, these results indicate that ANT is involved in regulating ABA-mediated drought tolerance. Clearly, the 35S:ANT transformants exhibited a phenotype that is consistent with that of the arf2 mutant plants (Figs 1 and 2).

As noted previously, the expression level of ANT was significantly higher in the arf2 mutants compared to that in the wild type (Schruff et al., 2006). To further dissect the potential link between ARF2 and ANT, we first investigated the relevance of the expression of ARF2 and ANT in wild-type leaves. Using 10-, 15- and 20-day-old leaves and real-time PCR analyses, we found that the expression of ARF2 gradually enhanced with increasing leaf age and, correspondingly, that the expression of ANT decreased (see Fig. 3A). Using young and old leaves, and developing siliques as source material, we further compared the expression differences of ANT between the arf2 mutants and the wild type; the level of expression of ANT in arf2 mutants was significantly higher than that in the wild type (see Fig. 3B and C). In addition, we introduced the ProARF2:GUS and Pro ANT:GUS constructs into the wild type. As shown in Fig. 3D and E, the ProARF2:GUS construct was strongly expressed in the root differentiation zone, but not in the meristem and elongation zones; whereas ProANT:GUS was strongly expressed in the root meristem, but not in the elongation and differentiation zones. Also, ProARF2:GUS was strongly expressed in the mature leaf abaxial epidermis, including the epidermis cells and stomata, whereas ProANT:GUS was expressed at a very low level only in stomata (Fig. 3F and G). However, when ProANT:GUS was introduced into the arf2 mutants, we found that ANT was strongly expressed in both leaf epidermis cells (including stomata) and the root tissues in transformants (see Fig. 3D–G). These results strongly imply that ARF2 negatively regulates ANT expression.

Inspecting the structural characteristics of the ANT promoter region, two reverse AuxREs cis-elements (GAGACA) were identified – i.e. in positions −376 to −370 and −470 to −463, respectively (see Fig. S4A). To test whether ANT is directly regulated by ARF2, we performed both an electrophoresis mobility shift assay (EMSA) and chromatin immunoprecipitation (ChIP) assays. For the EMSA assay, as shown in Fig. S4, when the target protein (GST–ARF2N1-470) and the labeled P3-P4 (the region from -601 to -276) DNA probes were added, a shifted DNA-binding band was detected; when the unlabeled P3-P4 DNA probes were added to the reaction mixture, the DNA-binding band was eliminated (Fig. S4B). Furthermore, the target protein did not bind to mutated DNA probes (mP3-P4) (Fig. S4C), and the amount of bound protein was significantly reduced in the presence of excess unlabeled P3-P4 fragments (Fig. S4D). These results clearly reveal that the ARF2 DNA-binding domain directly and specifically binds to the cis-element AuxREs in the promoter of ANT in vitro. For the ChIP analysis, as shown in Fig. 3H, amplification of the DNA with P3-P4 (-601 to -276), covering a region with two reverse conserved N-terminal DNA-binding sites (GAGACA), resulted in a greater amount of PCR product than amplification with primers 1 and 2 (P1-P2; -1405 to -1081), covering a region that does not contain reverse conserved N-terminal DNA-binding sites (GAGACA), which also applied to the coding sequence (CDS). These results indicate that ARF2 directly binds to the ANT promoter in vivo.

Initially, ARF2 was identified as factor that is sensitive to auxin signals. To further examine the potential differences in the responses of ARF2 and ANT to ABA and auxin signals, the vector DR5:GUS, which allows identification of auxin accumulation (Ulmasov et al., 1997), was transformed into the arf2-6, 35S:ANT and wild-type plants, and the signal of auxin accumulation in different tissues was inspected. Generally, the auxin accumulation in young leaves (10–20% expanded) usually exhibited a gradient – i.e. a relatively higher level at the base and a relatively lower level at the tip, which regulates the cell proliferation and elongation patterns in leaf development (Chen et al., 2001; Benkova et al., 2003). Our experiments showed that on the tips of the 10-day-old arf2-6, 35S:ANT and wild-type leaves, the auxin accumulation and/or auxin signals could not be detected in any of the samples tested (see Fig. 7). By contrast, when treated with 1.5 µm ABA, the leaf tips of the arf2-6 and 35S:ANT transformants exhibited more visible auxin signals than the wild type (see sub-panels a,c,e in Fig. 5A,B). Further, when treated with 3.0 µm ABA, the leaf tips of the arf2-6 and 35S:ANT transformants exhibited stronger auxin signals than the wild type (see sub-panels b,d,f in Fig. 5A and B). Correspondingly, the leaf sizes of the arf2-6 and 35S:ANT transformants were dramatically reduced after the treatment with ABA, compared to those of the wild type (see sub-panels a–f in Fig. 5A). These results clearly indicate that treatment with ABA significantly enhances auxin accumulation in the leaf blades of the arf2-6 and 35S:ANT transformants, compared to that in the wild type. All of the above findings suggest that abnormal ABA signaling significantly induces auxin accumulation in cotyledons of the arf2-6 mutants and 35S:ANT transformants, and leads to the suppression of elongation of their cotyledon.

It has previously been shown that the two AP2 domains of ANT can bind selectively to gCAC(A/G)N(A/T)TcCC(a/g)ANG(c/t) cis-elements that are involved in regulating the COR15A promoter (Nole-Wilson and Krizek, 2000; Krizek, 2003). The COR15A gene, considered to be a marker gene of the response to drought stress, is involved in response to environmental stresses through an ABA-dependent signaling pathway (Gilmour et al., 1998,, 2004). We found that COR15A was significantly downregulated in the ant-KO plants, compared to its expression in wild type (Fig. 6B and C), suggesting that the positive regulation of ANT on COR15A occurs at the transcriptional level. To determine whether COR15A is directly regulated by ANT through binding to the promoter in vivo, a ChIP assay was performed. The results showed that the regions C1 and C2 (containing consensus sequences) led to greater amounts of PCR product than the amount obtained by the region C3 (not containing consensus sequences) (Fig. 6D). Furthermore, expression of β-glucuronidase under the COR15A promoter (ProCOR15A:GUS) was weak on the leaves, hypocotyls and roots in the wild-type background, whereas expression of ProCOR15A:GUS was strong on the corresponding organs in the 35S:ANT background (Fig. 6E). These findings indicate that ANT plays a major role in activating the COR15A promoter under non-stress conditions.

To dissect whether ANT and COR15A form a signaling transduction pathway to regulate seed mass, firstly, the cor15a-knockout mutant (cor15a-KO) seed mass was further checked. As shown in Fig. 1B,D and Fig. S1A,C, the mutant cor15a-KO had smaller seeds than the wild type. Cytological observation indicated that the average areas of the cotyledon embryo and the cotyledon embryo cells in the cor15a seeds were ∼33% and ∼24% smaller than the wild type, respectively (Fig. S1B,D), meaning that the decreased size of the cor15a embryo results from decreased embryo cell size and cell number. Further, the seed mass, embryo size and embryo cell area of the 35S:ANTcor15a double mutants exhibited a similar phenotype to the cor15a plants (see Fig. 1A–D; Fig. S1A–D). These findings indicate that mutation of COR15A suppresses the seed mass of 35S:ANT plants, probably because ANT and COR15A form an ANT–COR15A gene cascade to regulate seed mass.

To better demonstrate that the ARF2–ANT–COR15A gene cascade regulates seed mass, we crossed the arf2-6 mutant with a cor15a. Our findings indicated that the resulting arf2-6cor15a plant had a smaller seed mass than the wild type, suggesting that cor15a is epistatic to arf2-6 (Fig. S1F,G). In conclusion, ANT acts upstream of COR15A to regulate seed mass.

As mentioned above, large-seeded seedlings are generally more robust and exhibit a stronger tolerance to environmental stresses, compared to small-seeded seedlings (Muller-Landau, 2010). Although ecologists have examined the survival advantage of large-seeded and small-seeded species at the macroscopic level, and have explained the trade-off strategy between colonization and competition in which smaller-seeded species are superior at colonizing and larger-seeded species are superior competitors (Coomes and Grubb, 2003; Westoby et al., 1992), little is known about the potential molecular mechanisms underlying why large-seeded seedlings exhibit a stronger tolerance to environmental stresses than the small-seeded seedlings. Our current study has demonstrated that the ARF2 transcription factor serves as a molecular link that integrates the regulation of seed development into plant responses to drought stresses and has provided novel insights into understanding why large-seeded seedlings are more efficient at withstanding abiotic stresses than smaller-seed seedlings.

The transcription factor ARF2 was originally identified as a repressor that is involved in negatively regulating ABA-mediated seed germination and primary root growth (Wang et al., 2011). Usually, with ABA treatment, the expression of ARF2 is significantly upregulated (Wang et al., 2011). Our current study demonstrated that the ARF2 loss-of-function mutants accumulate ABA, consequently resulting in extra cell proliferation or larger seed mass, and an increase of stomatal closing or drought tolerance. It seems clear that ARF2 acts not only in response to ABA signals but also as a regulator of ABA accumulation. As Wang et al. (2011) have proposed previously, the function of ARF2 in the network of ABA signaling might be twofold because it responds to different ABA inducers – i.e. ABI3, ABI4 and ABI5 act upstream of ARF2, resulting in the regulation of ARF2 expression, whereas ABI1 and ABI2 act further downstream, to regulate ABA accumulation (see Fig. 8). Not surprisingly, the change in ABA accumulation triggers consequent physiological actions in the regulation of growth and development and responses to environmental stresses (Lim et al., 2007; Assmann and Wang, 2001). Interestingly, evidence has been presented that ARF2 modulates the hypocotyl bending of the hookless1 mutant through auxin signaling in the apical hook (Li et al., 2004), meaning that ARF2 might be involved in auxin signaling. Our current study confirms that ARF2 is able to mediate auxin signaling by changing the levels of ABA, which causes crosstalk between ABA and auxin.

The transcription factor ANT plays a pivotal role in controlling cell proliferation, which determines the overall size of the organ (Klucher et al., 1996; Mizukami and Fischer, 2000). This study demonstrated that the expression of ANT was negatively regulated by ARF2 through binding to the ANT promoter region and that the regulation of ANT modulates seed mass by promoting cell proliferation. In particular, the expression of ANT increased significantly under drought conditions or upon ABA treatment (see Fig. 4E–G). Thus, the regulation of ANT by ARF2 might be, at least partly, dependent on ABA signaling.

It was well known that COR15A and other COR genes are involved in responding to environmental stresses, particularly drought, cold and dehydration, and responses to these conditions are associated with ABA signaling (Stockinger et al., 1997; Gilmour et al., 2004; Yamaguchi-Shinozaki and Shinozaki, 2006). Moreover, COR15A is involved in regulating plant growth and development, for example, leaf senescence (Yang et al., 2011). In this study, the cor15a mutants exhibited smaller seed size, indicating that the gene is involved in controlling seed mass. The phenotype of the 35S:ANTcor15a double mutant showed that mutation of COR15A can inhibit the large seed mass of the 35S:ANT plants (see Fig. 1). Particularly, the arf2, 35S:ANT and cor15a mutant plants all exhibited an altered seed mass because of altered embryo sizes. These findings indicate that the ARF2–ANT–COR15A signal cascade forms an ABA-mediated signal transduction pathway that regulates seed mass.

Studies have identified that the ABA2, ABI5 and SHB1 genes are involved in controlling seed size (Cheng et al., 2014). Specifically, ABA2 is positively regulated upon ABA accumulation, activating ABI5 in ABA signaling, consequently negatively modulating SHB1 expression through the direct binding between ABI5 and the ABRE cis-elements in the SHB1 promoter region (see Fig. 7). Further, SHB1 has been associated with both MINI3 and IKU2 in the regulation of seed development (Zhou et al., 2009). In this study, we identified that the ARF2–ANT–COR15A signal cascade participates in controlling seed size within ABA-mediated signaling through ABA–auxin crosstalk, which causes an increase in cell proliferation. These studies indicate that the mechanisms to control the size of seeds are complex and involve diverse networks. Clearly, our current study adds new evidence to help the understanding of the molecular mechanisms that control seed size.

Based on our current study, a potential role of the ARF2–ANT–COR15A signal cascade in integrating ABA signals into the regulation of seed mass and drought tolerance is proposed. As shown in Fig. 7, ARF2 responds to ABA signals and mediates ABA accumulation by triggering ABI1 and ABI2. ARF2 negatively regulates the expression of ANT, consequently ANT directly regulates the function of COR15A to regulate seed mass and stress resistance. The ARF2–ANT–COR15A–ABA-mediated signal cascade that we have identified in this study might be different from that which regulates seed development through the SHB1–ABA-mediated pathway, which has been proposed by Cheng et al. (2014).

The arf2-6, arf2-7 (Okushima et al., 2005) and 35S:ARF2:FLAG (Wang et al., 2011), ant-8 (CS3944) (described by ABRC) and 35S:ANT transgenic plants (Mizukami and Fischer, 2000) on the Col-0 background have been described previously. The arf2-6 and arf2-7 homozygous seeds have similar phenotypes (such as large seed mass, delayed flowering and leaf senescence, etc.), as have been described by Okushima et al. (2005).

cor15a (SALK_054513), cor15a (SALK_008461) and ant-KO (CS3944) seeds were obtained from the Arabidopsis Biological Resource Center (ABRC; Ohio State University). The arf2-6, arf2-7 and 35S:ARF2:FLAG homozygous seeds were kindly provided by Professor Z. Z. Gong (China Agricultural University, China). The ant-KO (SALK_022770) and 35S:ANT (hygromycin B; Ph2GW7) homozygous seeds were kindly provided by Professor H. G. Nam (Daegu Gyeongbuk Institute of Science and Technology, Korea).

Homozygous lines of cor15a (SALK_054513) and cor15a (SALK_008461) were obtained through herbicide selection for three or more generations and analysis of segregation ratios. Absence of gene expression in these mutants was verified by reverse transcriptase (RT)-PCR analysis before use. The arf2 ant-KO/+ lines were obtained from F2 seedlings of arf2-6×ant-KO (SALK_022770)/+ lines that did not show any defect in apical hook formation when grown in the dark (Li et al., 2004). Then, the arf2 ant-KO homozygous lines were obtained from arf2 ant-KO/+ lines through hygromycin selection for three or more generations and analysis of segregation ratios. Absence of ARF2 and ANT expression in the arf2 ant-KO mutant was verified by RT-PCR analysis before use. The 35S:ANTcor15a homozygous lines were obtained from 35S:ANTcor15a/+ through hygromycin and herbicide selection for three or more generations and analysis of segregation ratios. Absence of COR15A gene expression in 35S:ANTcor15a lines was verified by RT-PCR analysis before use; overexpression of the ANT gene in 35S:ANTcor15a lines was also verified by RT-PCR analysis before use.

We firstly generated (+/−) arf2-6cor15a double mutant (F1) seeds. (+/−) arf2-6cor15a double mutants (F1) seeds were sown, and (+/+, +/−, −/−) arf2-6cor15a double mutants (F2) seeds were obtained. Then, we selected (+/−, −/−) arf2-6cor15a seeds (F2), which are smaller size than those of wild type, sowed them and gained (+/+, +/−, −/−) arf2-6cor15a seeds (F3). We randomly selected (+/−, −/−) a few arf2-6cor15a lines (F3) with smaller seed size compared with those of wild type and their developed small seeds were identified by RT-PCR (data not shown). Our findings indicated that ARF2 and COR15A expression in these developed small seeds could be not detected, suggesting that these developed small seeds are (−/−) arf2-6cor15a double mutant (F3).

Plants exhibiting the arf2-6 mutant phenotype [large rosette leaves and large seeds grown in white light (Schruff et al., 2006)] in the F2 populations were screened for ProANT:GUS expression in roots. F3 seeds were collected from those exhibiting expression, and lines expressing GUS in all F3 plants were used for subsequent analysis, as described previously (Zgurski et al., 2005). Plants exhibiting the arf2-6 mutant and 35S:ANT phenotype [large rosette leaves and large seeds grown in white light (Schruff et al., 2006; Mizukami and Fischer, 2000)] in the F2 populations were screened for Dr5:GUS expression in roots. F3 seeds were collected from those exhibiting expression, and lines expressing GUS in all F3 plants were used for subsequent analysis. ProANT:GFP was introduced into the 35S:ARF2 background by Agrobacterium-mediated transformation of 35S:ARF2 homozygous plants. Transformants were selected on hygromycin B (Wang et al., 2011) for three or more generations and analyzed to determine segregation ratios. Plants exhibiting the 35S:ANT phenotype [large rosette leaves and large seeds grown in white light (Mizukami and Fischer, 2000)] in the F2 populations were screened for ProCOR15A:GUS expression in roots. F3 seeds were collected from those exhibiting expression, and lines expressing GUS in all F3 plants were used for subsequent analysis. Transgenic plants were generated using the Agrobacterium-tumefaciens-mediated floral dip method (Meng, 2015; Meng and Yao, 2015).

The seeds were subjected to 4°C for 3 days, and then sown onto solid MS medium supplemented with 1% sucrose at pH 5.8 and 0.8% agar. The plants grown on agar were maintained in a growth room under 16–8 h light–dark cycles with cool white fluorescent light at 21±2°C. Plants grown in soil-less medium were maintained in a controlled environment growth room under 16–8 h light–dark cycles with cool white fluorescent light at 21±2°C.

Total RNA was extracted from the tissues indicated in the figures using the TRIZOL reagent (Invitrogen), as has been described by Yu et al. (2008). SYBR green was used to monitor the kinetics of PCR product in real-time RT-PCR, as has been described by Yu et al. (2008). Gene-specific PCR primers are described below. For analyzing ARF2 expression in young or old seedlings and developing siliques of wild type (Col-0), primers F 5′-GAGTTTTGACTACCTCTGGTTAA-3′ and R 5′-GATAAAACCACCAATTTCACCTC-3′ were used. For analyzing ANT expression in young or old seedlings and developing siliques of wild type (Col-0), primers F 5′-AGGTGGCAAGCACGGATTGGT-3′ and R 5′-AGGCAACGCGAAAATCGCCG-3′ were used. For analyzing COR15A expression in ant-KO seedlings, primers F 5′-AGCGGAGCCAAGCAGAGCAG-3′ and R 5′-TGCCGCCTTGTTTGCGGCTT-3′ were used. These experiments were repeated at least two times with similar results.

Average seed weight was tested by weighing mature dry seeds in batches of 100 using an electronic analytical balance (Mettler Toledo). The chlorophyll of young cotyledons was removed by using an ethanol gradient (30%, 50%, 70%, 90% and 100%), for 30 min in each solution. These cotyledons were then photographed using a HIROX three-dimensional video microscope, and the cotyledon cell size was measured by using ImageJ software. Measurements of cotyledon area were made through scanning these organs to form a digital image and then calculating the area using ImageJ software. Mature seeds were photographed at a relevant magnification using a HIROX three-dimensional video microscope.

Mature dried seeds were imbibed for 60 to 100 min and dissected under a microscope in order to isolate mature embryos. The embryos were incubated overnight in buffer (30 mM sodium phosphate, pH 7.0, 10 mM EDTA, 1% Triton X-100 and 1% DMSO) at 37°C, fixed for 1 h in buffer (FAA with 10% formalin, 5% acetic acid and 45% ethanol) and 0.01% Triton X-100, and dehydrated with an ethanol series, as described by Ohto et al. (2005). Then, the embryos were treated for 1 to 2 h in Hoyer's buffer (3:0.8:0.4 of chloral hydrate:water:glycerol). Using a HIROX three-dimensional video microscope, under relevant magnification, we observed the treated embryos. Using ImageJ software, cotyledon and hypocotyl area, and average epidermal cell size in the central region of cotyledons and hypocotyls were measured, as described by Ohto et al. (2005).

For water loss measurements, 6–8 leaves per individual mutant and wild-type plant that had been grown under normal conditions for 3 weeks were excised, and fresh weight was determined at the designated time intervals. Four replicates were performed for each line. Water loss was represented as the percentage of initial fresh weight at each time point.

Stomata were opened by exposing plants for 12 h to light and high humidity, and leaves were incubated for 2 h in stomatal-opening solution containing 50.0 mM KCl, 10.0 mM CaCl₂ and 10.0 mM Mes, pH 6.0. Stomatal apertures were measured 1 h after adding 3.0 mM ABA. Data represent means±s.d. (n>80 stomata; *P<0.05, **P<0.01). The stomatal aperture was measured as previously described by Song et al. (2005). Subsequently, the epidermis was placed onto a slide and photographed under a Hirox three-dimensional video microscope. These experiments were repeated at least three times with similar results.

For ARF2, ANT and COR15A promoter analysis, promoter–GUS constructs of At5g62000, At4g37750 and At2g42540 were created by inserting ∼2.0 kb, ∼1.0 kb and ∼1.1 kb promoter fragments, respectively, which were amplified using a primer into pCB308R, as previously described by Lei et al. (2007). For amplifying ARF2 promoter fragments, primers F 5′-ggggacaagtttgtacaaaaaagcaggctTTTCTCGTCCTTTTCCTCTCAA-3′ and R 5′-ggggaccactttgtacaagaaagctgggtAGCTTCAATCATTTCAACCGC-3′ were used. For amplifying ANT promoter fragments, primers F 5′-ggggacaagtttgtacaaaaaagcaggctAGCTTATAATGTGACAAAAGTTA-3′ and R 5′-ggggaccactttgtacaagaaagctgggtCTAATAATTAGGTTTCTTGTCACTT-3′ were used. For amplifying COR15A promoter fragments, primers F 5′-ggggacaagtttgtacaaaaaagcaggctCTTCGGAACAACAACAAGAGTT-3′ and R 5′-ggggaccactttgtacaagaaagctgggtTGTAATCATATTTGTGGTTTTCAG-3′ were used. To generate the ANT:HA plasmid, the primers F 5′-GTTTGTACAAAAAAGCAGGCTATGCAACAGCACCTGATGCAGAT-3′ and R 5′-CTTTGTACAAGAAAGCTGGGTTCAATTCCCATCATCTGATGATTTC-3′ were used. For amplifying ANT promoter fragments, primers F 5′-ggggacaagtttgtacaaaaaagcaggctAGCTTATAATGTGACAAAAGTTA-3′ and R 5′-ggggaccactttgtacaagaaagctgggtCTAATAATTAGGTTTCTTGTCACTT-3′ were used. In this section, lowercase letters denote sequences from the connector primers, whereas capital letters denote the sequences of targeted genes.

Using a mix buffer [1 mM X-gluc, 60 mM NaPO₄ buffer, 0.4 mM of K₃Fe(CN)₆/K4Fe(CN)₆ and 0.1% (v/v) Triton X-100], samples (transgenic plants harboring and expressing ProARF2:GUS, Pro ANT:GUS, Pro COR15A:GUS DR5:GUS) were stained, and then incubated at 37°C for 8 h. After staining of GUS, chlorophyll was removed using a 30, 50, 70, 90 and 100% ethanol series, with an incubation of 30 min at each concentration. GUS staining was performed as previously described by Meng et al. (2015).

Leaf tissues of 2-week-old transgenic lines overexpressing 35S:ARF2:FALG and 35S:ANT:HA were used in this assay. ChIP was performed, as described previously (Meng, 2015; Meng and Yao, 2015; Meng and Liu, 2015). A FLAG-M2-tag-specific monoclonal antibody was used for ChIP analysis in overexpressing 35S:ARF2:FALG lines. A hemagglutinin (HA)-tag-specific monoclonal antibody was used for ChIP analysis in overexpressing 35S:ANT:HA lines. Green fluorescent protein (GFP)-tag-specific monoclonal antibody was used as a control in the above experiments. The ChIP DNA products were analyzed by using quantitative PCR with primers that had been synthesized to amplify ∼300-bp DNA fragments in the promoter region of ANT and COR15A in the ChIP analysis. The primer sequences used for ARF2 ChIP analysis were: ANT-1 (P1 5′-AGCTTATAATGTGACAAAAGTTATT-3′; P2 5′-TGTCTTGGGTTATTTTGTGGTG-3′); ANT-2 (P3 5′-TAGATACAGTATAAACTAACTTTAA-3′; P4 5′-CTAATAATTAGGTTTCTTGTCACTT-3′). The primer sequences used for ANT ChIP analysis were: COR15A-1 (P1 5′-CTTCGGAACAACAACAAGAGTT-3′; P2 5′-TTAAATTTTTACAAAATTAAATT-3′); COR15A-2 (P3 5′-AGGAGATGTTACTGTCCGTCAG-3′; P4 5′-ATGAGTTGAAACCACAAACCATT-3′) and COR15A-3 (P3 5′-GGCTTTTGGTAGATTTGGGCTTG-3′; P4 5′-ACGTGTAATCATATTTGTGGTTT-3′).

The plasmid pGEX-5X-1 for ARF2 was used in this experiment. The coding sequence of ARF2 was amplified by using the primer pair (5′-GGATCCATGGCGAGTTCGGA-3′ and 5′-GAATTCAGATTGCCAGACAAC-3′), and then cloned into the BamHI and EcoRI restriction sites of the pET28a plasmid to generate the final plasmid. Recombinant glutathione S-transferase binding protein (GST)-tagged ARF2 was extracted from transformed Escherichia coli (Rosetta2) after 10 h of incubation at 16°C following induction with 10 μM isopropylβ-d-1-thiogalactopyranoside. These recombinant proteins were purified using GST-agarose affinity.

The EMSA protocol has been described previously (Meng et al., 2015). The biotin-labeled ANT DNA fragments (5′-GAAAAAGAGACAAAAGGAGGGAATTTAGAAATGAGGTGGTGAAGGTATGTTGGATTGTTGTGGAACGATATGGTCAATAAAGCATATCGCATTATTGGAGAGACATTACAT-3′) and mutated ANT DNA fragments (5′-GAAAAAGTGTCAAAAGGAGGGAATTTAGAAATGGTGAAGGTATGTTGGATTGTTGTGGAACGATATGGTCAATAAAGCATATCGCATTATTGGAGTGTCATTACAT-3′) were synthesized, annealed and used as probes, and the biotin unlabeled same DNA fragments were used as competitors in this assay. The probes were incubated with the ARF2 fusion protein at room temperature for 20 min in a binding buffer [50 mM HEPES-KOH (pH 7.5), 375 mM KCl, 6.25 mM MgCl, 1 mM DTT, 0.5 mg/ml BSA, glycerol 25%].

We thank Professors Hong-Gil Nam (DGIST, Korea) and Cheng-Bin Xiang (University of Science and Technology of China, China) for their support during the initial stages of this work. We extend our thanks to Zi-Qing Miao (University of Science and Technology of China, China) for completing the ANT:HA plasmid and gaining ANT:HA seeds.