ATAF2 integrates Arabidopsis brassinosteroid inactivation and seedling photomorphogenesis

Multiple hormonal and light signals coordinately regulate plant morphogenesis (Neff et al., 2000; Gray, 2004). The Arabidopsis thaliana hypocotyl is a robust system for studying the interplay of light and plant hormones in the regulation of plant growth and development (Collett et al., 2000; Vandenbussche et al., 2005). Genetic screens for long-hypocotyl mutants in the light led to the isolation of genes involved in photomorphogenic signaling pathways, including many of the photoreceptors (Neff and Chory, 1998). Screens for short-hypocotyl mutants in the dark have also identified additional genes involved in photomorphogenesis-regulating pathways, including those associated with hormone-mediated development (Chory et al., 1989). For example, the dark-grown short-hypocotyl mutant det2 uncovered the role of brassinosteroids (BRs) in seedling growth and development (Chory et al., 1991; Li et al., 1996).

Brassinolide (BL) is the endpoint of the BR biosynthetic pathway in many plant species and confers the strongest biological activity (Yokota, 1997). The BR signaling (Zhu et al., 2013) and metabolism (Zhao and Li, 2012) pathways have been well studied in Arabidopsis. Loss-of-function mutations in key BR biosynthesis genes leads to BR deficiency-specific dwarfism of Arabidopsis plants, which can be rescued in a dose-dependent manner by exogenous BR (Zhao and Li, 2012). By contrast, a BR-deficient dwarf phenotype is conferred by overexpression (or gain-of-function) of BR catabolic enzymes, such as the cytochrome P450s BAS1 (CYP734A1, formerly CYP72B1) and SOB7 (CYP72C1) (Neff et al., 1999; Turk et al., 2003, 2005; Thornton et al., 2010).

BRs cannot be transported between plant tissues (Savaldi-Goldstein et al., 2007; Symons et al., 2008). Since BRs are synthesized and perceived in the same tissues, homeostasis of this class of hormones is tightly regulated to ensure that their cellular levels are appropriate for a given developmental fate. BR homeostasis is also important for the establishment of plant immunity (Belkhadir et al., 2012).

In Arabidopsis, BR homeostasis is maintained in part by transcriptional feedback regulation loops that control the expression of key biosynthetic and catabolic enzymes (Tanaka et al., 2005). For example, TCP1, a putative bHLH transcription factor (TF), interacts with the promoter of the BR biosynthesis gene DWF4 (Choe et al., 1998) to modulate BR levels (Guo et al., 2010). In addition, the TF LATERAL ORGAN BOUNDARIES (LOB) binds directly to the promoter of BAS1 as an activator, and consequently reduces BR accumulation to limit growth in organ boundaries (Bell et al., 2012).

In this research, we found that an Arabidopsis NAC (NAM, ATAF1,2 and CUC2) family TF, ATAF2 (ANAC081), binds the promoters of BAS1 and SOB7 to suppress their expression. ATAF2 regulates hypocotyl elongation and root growth by suppressing BR catabolism. The expression of ATAF2 is suppressed by both BRs and light. The photomorphogenic phenotypes of ATAF2 loss- and gain-of-function seedlings can be suppressed by treatment with the BR biosynthesis inhibitor brassinazole (BRZ) (Asami et al., 2000). Moreover, the disruption of BAS1 and SOB7 abolishes the short-hypocotyl phenotype of ATAF2 loss-of-function seedlings in low fluence rate white light, which demonstrates an ATAF2-mediated connection between BR catabolism and photomorphogenesis. We propose that ATAF2 plays a central role in integrating BR homeostasis and seedling development. Considering that ATAF2 has previously been identified as a regulator of plant defense (Delessert et al., 2005; Wang et al., 2009) and the auxin biosynthesis gene NIT2 (Huh et al., 2012), ATAF2 acts as a central regulator of plant defense, hormone metabolism and light-mediated seedling development.

In a yeast one-hybrid (Y1H) screen for the potential protein interactors of the BAS1 promoter, a truncated ATAF2 missing the first 65 amino acids (ATAF2t) was found to interact with a BAS1 promoter fragment (−731 to −504) (Table S1). This fragment, pBAS1-EE, contains only an Evening Element TF binding site (EE; AAAATATCT or its complementary sequence) (Michael and McClung, 2002) when scanned with the Athena program (O'Connor et al., 2005). Thus, the EE element was proposed to be the ATAF2 binding target in the BAS1 promoter. Full-length ATAF2 interacted with pBAS1-EE in a targeted Y1H assay (Fig. 1A). The CCA1 binding site (CBS; AAAAATCT or its complementary sequence) is similar to EE in sequence. Both EE and CBS are binding targets of the circadian clock protein CCA1 (Wang and Tobin, 1998; Michael and McClung, 2002; Harmer and Kay, 2005). As there are two CBSs in the BAS1 promoter and one in the SOB7 promoter (Pan et al., 2009), three CBS-containing promoter fragments (Table S1), pBAS1-CBS1 (−844 to −786), pBAS1-CBS2 (−1832 to −1768) and pSOB7-CBS (−1623 to −1524), were used as baits in targeted Y1H assays to test the hypothesis that ATAF2 binds the EE/CBS elements in the BAS1 and SOB7 promoters. CBS was the only predicted TF binding site in all three fragments when scanned with the Athena program. ATAF2 interacted with both pBAS1-CBS1 (Fig. 1B) and pSOB7-CBS (Fig. 1C).

An electrophoresis mobility shift assay (EMSA) provided additional evidence that ATAF2 interacts with a CBS-containing DNA fragment (Fig. 1D). ATAF2 bound with biotin-labeled pBAS1-CBS1, suggesting the formation of an ATAF2–pBAS1-CBS1 complex. With increasing concentrations of the unlabeled probes, the specific binding signal of pBAS1-CBS1 with ATAF2 was gradually abolished (Fig. 1D), which further suggests the physical interaction between ATAF2 and pBAS1-CBS1.

To further test whether the interaction between ATAF2 and the BAS1/SOB7 promoters is mediated by the EE/CBS elements, site-directed mutagenesis was performed for both pBAS1-EE (the EE element was mutated from AAAATATCT to AACATATCT, named pBAS1-EEm) and pSOB7-CBS (the CBS element was mutated from AGATTTTT to AGATTCTT, named pSOB7-CBSm). Neither pBAS1-EEm (Fig. 1E) nor pSOB7-CBSm (Fig. 1F) was able to interact with ATAF2 in targeted Y1H assays, which indicates that the EE/CBS elements are responsible for the binding of ATAF2.

Additionally, neither ATAF2 nor ATAF2t protein showed interaction with pBAS1-CBS2 in the targeted Y1H assay (Fig. S1), which indicates that the existence of EE/CBS sequence alone is insufficient to mediate interaction with ATAF2.

ATAF2 can either activate (Wang et al., 2009; Huh et al., 2012) or suppress (Delessert et al., 2005) the expression of downstream target genes. To test the effects of ATAF2 on BAS1 and SOB7 expression, the transcript accumulation of BAS1 and SOB7 in ATAF2 gain- and loss-of-function seedlings was examined by qRT-PCR (unless otherwise stated, 4-day-old seedlings grown at 25°C were used for all total RNA extraction and hypocotyl measurement experiments). BAS1 transcript accumulation was significantly decreased in ATAF2 overexpression (ATAF2ox) lines and increased in ataf2 null mutants (ataf2-1, ataf2-2) in both darkness (Fig. 2A) and 80 μmol m⁻² s⁻¹ continuous white light (Fig. 2B). A similar relationship was observed in the case of SOB7 transcript accumulation, except that SOB7 expression in ATAF2ox seedlings was not significantly reduced in white light (Fig. 2C,D).

The effects of ATAF2 on the expression of BR biosynthesis genes were also tested. Unlike the BR-inactivation genes BAS1 and SOB7, two key BR biosynthesis genes, DWF4 (Choe et al., 1998) and CPD (Szekeres et al., 1996), did not show any dramatic changes of expression in the gain- or loss-of-function ATAF2 mutants (Fig. S2A-D), which indicates that ATAF2 does not affect BR biosynthesis directly under the conditions tested. In darkness, the slight changes in DWF4 and CPD transcript accumulation in loss- and gain-of-function ATAF2 mutants (Fig. S2A,C) might be caused by transcriptional feedback regulation due to altered BR levels (Tanaka et al., 2005).

GUS analysis of ATAF2-GUS fusion transgenic Arabidopsis plants revealed that ATAF2 is expressed at varying levels in all plant tissues (Delessert et al., 2005). Similarly, GUS analysis showed that BAS1 and SOB7 have distinct, as well as overlapping, expression patterns and that these are limited to certain tissues, such as the shoot apex, root tip or root elongation zone (Sandhu et al., 2012). The inconsistency of expression patterns between ATAF2 and BAS1/SOB7 proteins indicates that ATAF2 might function as an in vivo repressor to restrict the expression of BAS1 and SOB7 in specific tissues. To test this hypothesis, pBAS1:BAS1-GUS and pSOB7:SOB7-GUS (genomic DNA translational fusions with 1.6 and 2.1 kb of their native promoters, respectively) transgenes were expressed in the ataf2-2 background. The BAS1 promoter used includes the ATAF2 interaction-positive pBAS1-EE and pBAS1-CBS1 fragments, but not the ATAF2 interaction-negative pBAS1-CBS2 fragment. The SOB7 promoter used includes the ATAF2 interaction-positive pSOB7-CBS element. Approximately 10% of the T1 primary transformants showed BR deficiency-associated dwarf (BR-dwarf) phenotypes for both transgenes (Fig. S3). Homozygous T3 plants were identified from the BR-dwarf single-locus T-DNA insertion lines for each transgene.

In order to compare the expression strength of exactly the same pBAS1:BAS1-GUS or pSOB7:SOB7-GUS genomic insertion sites in wild-type (Col-0) and ataf2-2 backgrounds, these selected homozygous single-locus T-DNA insertion BR-dwarf lines were crossed with Col-0, and homozygous pBAS1:BAS1-GUS/Col-0, pBAS1:BAS1-GUS/ataf2-2, pSOB7:SOB7-GUS/Col-0 and pSOB7:SOB7-GUS/ataf2-2 lines were identified from the segregating F2 populations. Whereas many pBAS1:BAS1-GUS/ataf2-2 and pSOB7:SOB7-GUS/ataf2-2 plants retained the BR-dwarf phenotype, no BR-dwarfs were identified in their pBAS1:BAS1-GUS/Col-0 and pSOB7:SOB7-GUS/Col-0 siblings (Fig. 3A,B). In addition, GUS analysis demonstrated that the disruption of ATAF2 allowed the expansion of BAS1 and SOB7 expression to almost all plant tissues (Fig. 3C,D). In conclusion, the disruption of ATAF2 dramatically increased the expression of BAS1 and SOB7 proteins, demonstrating its role as a negative regulator of these BR-inactivating enzymes.

The expression levels of BAS1 and SOB7 influence seedling growth in response to exogenous BRs (Neff et al., 1999; Turk et al., 2003, 2005; Thornton et al., 2010). As a transcriptional repressor of BAS1 and SOB7, the expression level of ATAF2 is also likely to influence seedling growth in response to exogenous BRs. To test this hypothesis, Col-0, the ataf2 null and the ATAF2ox mutants were grown on media with gradient concentrations of BL (0, 10, 100 and 1000 nM) in darkness or 80 μmol m⁻² s⁻¹ white light. At all three BL concentrations, Col-0 hypocotyl growth was inhibited in darkness and enhanced in white light, which is consistent with the findings of Turk et al. (2003). When compared with Col-0, the ataf2 null mutants were less responsive to all exogenous BL treatments, with longer hypocotyls in the dark and shorter hypocotyls in the light. By contrast, the ATAF2ox mutants generally displayed the opposite response of shorter hypocotyls in the dark and longer hypocotyls in the light (Fig. 4A-D), the only exception being that, in white light, ATAF2ox mutants only showed a significant hypocotyl growth phenotype with 100 nM BL treatment. These results indicate that 100 nM BL might be an ideal dose for the identification of hypocotyl elongation phenotypes for both dark and white light conditions.

Alteration of endogenous BR levels may also cause hypocotyl growth phenotypes in ATAF2 mutants. When grown in the dark, ataf2-1 and ATAF2ox seedlings showed opposite hypocotyl growth phenotypes in response to treatments with the BR biosynthesis inhibitor BRZ (Fig. 4E,F). In all three BRZ treatment conditions (20, 100 and 500 nM), BRZ slightly promoted hypocotyl growth of Col-0 and the ATAF2ox mutants, but inhibited that of ataf2 null mutants. Upon BRZ treatment, ataf2 null mutants always had a shorter hypocotyl than that of Col-0, whereas ATAF2ox mutants were always taller than Col-0. BRZ treatment assays further demonstrated that the hypocotyl growth phenotypes of ATAF2 mutants are due to the changes in BR levels.

BRZ has been reported to inhibit the hypocotyl growth of dark-grown seedlings at lower temperatures (Asami et al., 2000; Nagata et al., 2000). Thus, the BRZ treatment assays were also conducted on dark-grown seedlings at 20°C (Fig. 4G). At this temperature, BRZ inhibited the hypocotyl growth of all three genotypes. ATAF2ox mutants were always less responsive to BRZ treatments than Col-0, whereas ataf2 null mutants exhibited the opposite phenotype. When grown at 20°C, ATAF2ox mutants were shorter than both Col-0 and ataf2 null mutants in the absence of BRZ.

Because ATAF2 has been reported to regulate the expression of the auxin biosynthesis gene NIT2 (Huh et al., 2012), the involvement of auxin in ATAF2-mediated seedling phenotypes was tested, along with that of BRs. In contrast to the distinct hypocotyl growth phenotypes in ATAF2 mutants caused by BL and BRZ, treatment with either auxin indole-3-acetic acid (IAA) or the auxin transport inhibitor naphthylphthalamic acid (NPA) (Ruegger et al., 1997) did not significantly affect the hypocotyl growth phenotypes of ATAF2 loss- and gain-of-function seedlings in darkness (Fig. S4A,B). These results suggest that BRs are more important than auxins in ATAF2-mediated hypocotyl growth.

In addition to hypocotyl elongation, root growth is another important parameter for measuring seedling development. Thus, the effects of ATAF2 on seedling primary and lateral root growth were examined by BRZ and NPA treatment. With no BRZ applied, Col-0, ataf2 null and ATAF2ox mutants showed similar primary root growth (Fig. 5A). Compared with the wild type and ATAF2ox mutant, ataf2-1 was more sensitive to exogenous BRZ treatments with respect to the elongation of the primary root (Fig. 5A). By contrast, ataf2-1, Col-0 and ATAF2ox seedlings showed similar primary root growth in response to exogenous NPA treatments (Fig. S5A), indicating that BRs are more important than auxins in ATAF2-mediated primary root growth regulation.

Even in the absence of BRZ, ATAF2 loss- and gain-of-function seedlings showed opposite lateral root growth phenotypes (Fig. 5B). ataf2-1 had more lateral root growth than Col-0, whereas overexpression of ATAF2 inhibited lateral root growth (Fig. 5B). Exogenous BRZ treatments significantly attenuated the lateral root growth phenotype of ataf2-1, but not that of ATAF2ox (Fig. 5B). With BRZ concentrations greater than 100 nM, the lateral root growth phenotype of ataf2-1 was similar to that of the wild type. An NPA treatment assay was carried out in the presence of 100 nM BRZ to investigate whether disruption of auxin transport would cause a difference in lateral root growth between Col-0 and ATAF2 mutants. The results showed that the lateral root growth phenotypes of ataf2-1 and ATAF2ox under 100 nM BRZ were not significantly affected, compared with Col-0, by additional NPA treatments (Fig. S5B).

In a fluence rate response assay (Fig. 6A,B), the hypocotyl growth of ATAF2 loss- and gain-of-function and Col-0 seedlings showed no significant difference both in darkness and under high fluence rates of white light (80 and 100 μmol m⁻² s⁻¹), which was consistent with observations above (Fig. 4A-D). However, ATAF2 gain- and loss-of-function mutants displayed opposite hypocotyl length phenotypes in subsaturating fluence rates of white light (10, 20 and 40 μmol m⁻² s⁻¹). Fluence rate response analysis demonstrated that ATAF2 stimulates hypocotyl growth in a light intensity-dependent manner (Fig. 6A,B). ATAF2 acts as a suppressor of seedling photomorphogenesis under subsaturating white light conditions.

The short-hypocotyl phenotype of ataf2-1 under low fluence rate (10 μmol m⁻² s⁻¹) white light was retained in monochromatic red (Fig. 6C), far-red (Fig. 6D) or blue (Fig. 6E) light. In order to investigate the contributions of photoreceptors to ATAF2-mediated photomorphogenesis, ataf2-1 was crossed with null mutants of PHYB, PHYA and CRY1 to create corresponding double mutants. Under the same fluence rate of monochromatic light (10 μmol m⁻² s⁻¹), the short-hypocotyl phenotype of ataf2-1 was retained but attenuated (Table S2) when knocking out PHYB (Fig. 6F), PHYA (Fig. 6G) or CRY1 (Fig. 6H). These results indicate that ATAF2-mediated photomorphogenesis is regulated by multiple photoreceptors. Among the three photoreceptors examined, PHYA played the most significant role in regulating ATAF2-mediated seedling photomorphogenesis (Fig. 6D,G, Table S2).

In addition to far-red light, PHYA is known to respond to light in a wide range of wavelengths, including red (Tepperman et al., 2006; Franklin et al., 2007), blue and UV-A (Chun et al., 2001). Removal of PHYA was able to abolish the short-hypocotyl phenotype of ataf2-1 under 10 μmol m⁻² s⁻¹ red (Fig. 6I) or blue (Fig. 6J) light. These results indicate that the regulatory function of PHYA in ATAF2-mediated photomorphogenesis operates in light conditions of various wavelengths.

Since ATAF2-regulated hypocotyl growth was shown to involve both BRs and light, the effects of these pathways on ATAF2 transcript accumulation were analyzed by qRT-PCR. Exogenous BL suppressed ATAF2 transcript accumulation in Col-0 seedlings grown in both darkness and continuous white light (Fig. 7A,B). In Col-0 seedlings, ATAF2 transcript accumulation was also suppressed by continuous white light in a fluence rate-dependent manner (Fig. 7C), with high fluence rates being more effective in suppressing ATAF2 transcript accumulation. Additionally, a 1-h white light treatment for dark-grown seedlings was sufficient to significantly reduce ATAF2 transcript accumulation (Fig. 7D). These results revealed that ATAF2 is transcriptionally regulated by both BR and light signals. The white light-induced suppression of ATAF2 accumulation was retained in the same fluence rate (10 μmol m⁻² s⁻¹) of monochromatic red, far-red or blue light (Fig. 7E), which is consistent with ATAF2-mediated photomorphogenesis being regulated by multiple photoreceptors and with the major regulator, PHYA, functioning in a wide range of wavelengths.

The removal of PHYA did not affect ATAF2 transcript accumulation in dark-grown seedlings (Fig. 7F). The light-induced suppression of ATAF2 accumulation was attenuated by the removal of PHYA in 10 μmol m⁻² s⁻¹ white, red, far-red or blue light (Fig. 7G). This is also consistent with our observation that ATAF2-mediated photomorphogenesis is regulated by multiple photoreceptors, with PHYA playing a major role in light conditions of various wavelengths (Fig. 6A-J).

To test whether the ATAF2-mediated photomorphogenic phenotype is caused by changes in endogenous BR levels, Col-0 and ATAF2 loss- and gain-of-function seedlings grown under low fluence rate (10 μmol m⁻² s⁻¹) white light were treated with increasing concentrations (0, 20, 100 and 500 nM) of BRZ. Exogenous BRZ treatments attenuated the hypocotyl growth phenotypes of ATAF2 loss- and gain-of-function seedlings (Fig. 8A,B). The addition of 500 nM BRZ was sufficient to eliminate the hypocotyl length differences of all three genotypes (Fig. 8A,B). By contrast, NPA treatments did not affect the hypocotyl growth phenotypes of ATAF2 loss- and gain-of-function seedlings under identical light conditions (Fig. S6). BRZ and NPA response assays demonstrated that BRs play a more important role than auxins in ATAF2-mediated seedling photomorphogenesis.

To test the function of BAS1 and SOB7 in the ATAF2-mediated photomorphogenic phenotype, a bas1-2 sob7-1 ataf2-1 triple mutant was created and compared with the bas1-2 sob7-1 double mutant for hypocotyl growth in response to different fluence rates (0, 10 and 80 μmol m⁻² s⁻¹) of white light (Fig. 8C). Removal of BAS1 and SOB7 completely abolished the short-hypocotyl phenotype caused by disruption of ATAF2 (Fig. 8C). This indicates that ATAF2 suppresses seedling photomorphogenesis by repressing expression of the BR-inactivation genes BAS1 and SOB7 (Fig. 8D).

ATAF2 is an NAC family TF that is localized in the nucleus (Wang et al., 2009). NAC TFs have a highly conserved N-terminal NAC domain and diverse C-terminal domains. The NAC domain can bind both DNA and protein. The C-terminal domains are responsible for the activation or suppression of downstream target genes (Ernst et al., 2004). ATAF2 has been reported to bind a 25 bp A/T-rich consensus sequence (Wang and Culver, 2012) and a 36 bp A/T-rich sequence (Huh et al., 2012). Here, we have shown that ATAF2 also binds the A/T-rich EE/CBS elements (Fig. 1A-F). Interestingly, ATAF2 interacts with two members of the AT-hook motif-containing nuclear-localized (AHL) protein family, which specifically bind A/T-rich DNA (Zhao et al., 2013). The similarity in binding targets between ATAF2 and AHLs might facilitate their interaction in vivo. AHLs are present in all plant species and are involved in a wide range of biological processes, including the regulation of Arabidopsis hypocotyl growth (Zhao et al., 2013, 2014). Together, these observations suggest that ATAF2 and certain AHL proteins might work as functional partners in this fundamental plant development process.

The observation that ATAF2 and CCA1 share common promoter binding sites indicates that they might interact in vivo to regulate downstream gene expression coordinately (binding together to function) or antagonistically (competing for the same binding site or restricting each other from binding targets by direct protein-protein interaction). There is also a putative CBS (−577 to −570) in the ATAF2 promoter, which suggests that CCA1 might also regulate the expression of ATAF2 by directly binding its promoter.

Recently, more than 1000 CCA1 binding regions were identified via the ChIP-seq approach (Nagel et al., 2015), although the BAS1, SOB7 and ATAF2 promoters were not identified as CCA1 targets. Nagel et al. (2015) used 10-day-old Arabidopsis plants, whereas we used 3- to 4-day-old seedlings. The age difference and/or growth conditions might explain the differences between the two studies. In addition, the ChIP-seq approach might not identify all CCA1 binding sites, especially those where binding is transient or weak.

ATAF2 acts as a repressor of BAS1 and SOB7 expression (Fig. 2A-D). ATAF2 expression is suppressed by BRs (Fig. 7A,B). Therefore, ATAF2-BAS1/SOB7-BR-ATAF2 forms a feedback regulation loop. The BR biosynthesis genes DWF4 and CPD are not involved in this ATAF2 regulatory loop in our growth conditions (Fig. S2A-D). This is consistent with the fact that no EE/CBS element is detected when a 2.5 kb region of the DWF4 or CPD promoter is scanned with the Athena program. In addition to transcriptionally repressing BAS1 and SOB7, the emergence of BR-deficient dwarf plants in pBAS1:BAS1-GUS/ataf2-2 and pSOB7:SOB7-GUS/ataf2-2 transgenic populations (Fig. 3A,B, Fig. S3) indicates that the disruption of ATAF2 dramatically increases the expression of BAS1 and SOB7 proteins. In addition, comparison of identical pBAS1:BAS1-GUS and pSOB7:SOB7-GUS transgene insertions in Col-0 and ataf2-2 backgrounds supports the hypothesis that ATAF2 modulates the expression patterns of BAS1 and SOB7, restricting their expression to certain tissues, and controlling their expression within certain levels (Fig. 3C,D).

Although some pBAS1:BAS1-GUS/ataf2-2 and pSOB7:SOB7-GUS/ataf2-2 transgenic plants have BR-dwarf phenotypes, the original ataf2-2 mutant is not dwarf. There are at least four possible explanations for these observations. First, the transgenic plants harbor additional copies of BAS1 or SOB7. Second, the genomic positions of the transgenes differ from those of endogenous BAS1 or SOB7. Third, remote cis-regulatory elements might not be included in the promoter regions cloned in the constructs. Fourth, the fusion of ATAF2 to GUS protein might help to stabilize ATAF2. In Arabidopsis, there are examples in which GUS and GFP tags increase the stability of several SMALL AUXIN UP RNA (SAUR) proteins (Chae et al., 2012; Spartz et al., 2012). In this case, BAS1-GUS or SOB7-GUS translational fusion proteins might have greater stability than native BAS1 or SOB7, and the increased accumulation of BAS1-GUS or SOB7-GUS might lead to the BR-deficient dwarf phenotypes in pBAS1:BAS1-GUS/ataf2-2 or pSOB7:SOB7-GUS/ataf2-2, respectively. In conclusion, any of the possible reasons mentioned above could result in sufficient expression of BAS1 or SOB7 in transgenic ataf2-2 to cause the dwarf phenotype due to increased BR inactivation.

The GUS expression patterns in pBAS1:BAS1-GUS/Col-0 and pSOB7:SOB7-GUS/Col-0 are similar, but not exactly identical to, those reported previously (Sandhu et al., 2012), which might also be a consequence of differences in T-DNA insertion position among ectopic expression lines. Additionally, differences in the genetic background might affect the GUS expression pattern observed. Wild-type (Col-0) plants were used to generate GUS expression lines in this work, whereas Sandhu et al. (2012) adopted a bas1-2 sob7-1 double-mutant background for pBAS1:BAS1-GUS and pSOB7:SOB7-GUS expression.

The BL and BRZ response assay results (Fig. 4A-G) further support the hypothesis that ATAF2 acts as a repressor of BR catabolism. The reduced ATAF2 expression in ataf2 null mutants leads to higher transcript accumulation of BAS1 and SOB7, which enhances the BR inactivation capacity in seedlings and confers more resistance to BL treatment. Conversely, higher expression of ATAF2 in ATAF2ox mutants leads to lower levels of BAS1 and SOB7 transcript accumulation, which reduces the BR inactivation capacity in these seedlings, conferring more sensitivity to BL treatment. In darkness, ATAF2ox seedlings show a BL-oversensitivity phenotype under all three BL treatment conditions (Fig. 4A,B). In white light, a high concentration of exogenous BL (1 μM) masks the BL-oversensitivity phenotype of ATAF2ox seedlings. A low concentration of exogenous BL (10 nM) is not sufficient to induce the phenotype (Fig. 4C,D). However, the hypocotyl phenotype was evident upon treatment with 100 nM BL (Fig. 4C,D). Both ataf2 null mutants had BL-insensitivity phenotypes under all six light- and dark-grown BL treatment conditions (Fig. 4A-D).

The relationship between BRs and the hypocotyl growth phenotypes of ATAF2 mutants was further demonstrated by BRZ response analysis (Fig. 4E-G). In our assays, BRZ promotes the hypocotyl elongation of dark-grown seedlings (except ATAF2 loss-of-function mutants) at 25°C (Fig. 4E,F) but inhibits the elongation at 20°C (Fig. 4G). This observation suggests that the sensitivity of dark-grown hypocotyls to endogenous BR levels is temperature dependent. Gray et al. (1998) reported that high temperature promotes hypocotyl elongation that is mainly mediated by auxin, with a lesser involvement of BRs.

In all the BR, BRZ and NPA treatment assays conducted, only two distinct light conditions (darkness and 80 μmol m⁻² s⁻¹ white light) were used. These two conditions were selected because all hypocotyl lengths were similar for the non-treatment controls, which facilitated the evaluation of BR treatment effects. Since it is well known that seedling photomorphogenesis is modulated by the coordination of light and hormonal signals, the involvement of light intensity and quality in ATAF2-regulated seedling growth was investigated. The fluence rate response assay demonstrates that ATAF2 stimulates hypocotyl growth in a light intensity-dependent manner (Fig. 6A,B). In darkness, the hypocotyl growth regulatory activity of ATAF2 was not significantly apparent. Under high fluence rates of white light, the light intensities were strong enough to suppress the photomorphogenic activities of ATAF2.

ATAF2 transcript accumulation is suppressed by light in a fluence rate-dependent manner (Fig. 7C,D), which at least partially explains why ATAF2 loss- and gain-of-function and Col-0 seedlings have no significant difference in hypocotyl length under high fluence rates of light. ATAF2 has been reported to activate the expression of the auxin biosynthesis gene NIT2 (Huh et al., 2012). Thus, both BRs and auxins could be involved in the photomorphogenic phenotypes of ATAF2 mutants. BRZ (Fig. 8A,B) and NPA (Fig. S6) response assays demonstrate that BRs have a more important role in this response than auxins. The short-hypocotyl phenotype of ATAF2 loss-of-function mutants relies on the expression of BAS1 and SOB7 (Fig. 8C), which further demonstrates that ATAF2-mediated photomorphogenic phenotypes depend on BR inactivation catalyzed by BAS1 and SOB7. Together, these results demonstrate the presence of a light- ATAF2-BAS1/SOB7-BR-photomorphogenesis signaling network. The BL-response phenotypes of ATAF2ox seedlings were more dramatic in darkness than in white light (Fig. 4A-D). These results provide indirect evidence that light-mediated reduction of ATAF2 expression may attenuate ATAF2-mediated BR-response phenotypes.

ATAF2 loss- and gain-of-function seedlings retain their hypocotyl length phenotypes in monochromatic red, far-red and blue light (Fig. 6C-E). The phenotypes are attenuated (Fig. 6F-H, Table S2) by the removal of PHYB, PHYA or CRY1, with disruption of PHYA being the most dramatic (Fig. 6D,G, Table S2). These results suggest that multiple photoreceptors are involved in ATAF2-mediated seedling photomorphogenesis, with PHYA playing an important role. Turk et al. (2003) reported that the interaction between light and the BAS1-mediated BR-inactivation pathway is mainly dependent on far-red light. The regulation of ATAF2-mediated photomorphogenesis is consistent with this previous observation and might constitute part of the light-BR intersection. The observation that the ATAF2-mediated photomorphogenic phenotype under monochromatic red or blue light also requires PHYA (Fig. 6I,J) further demonstrates the close relationship between PHYA and ATAF2 function.

BRZ treatments reduce both primary and lateral root growth (Fig. 5A,B), which is consistent with previous observations that BRs can promote the growth of both tissues (Müssig et al., 2003; Bao et al., 2004). There have been numerous reports revealing the complexity of BR activity in controlling root growth and development (Hacham et al., 2011; Fridman et al., 2014; Vragović et al., 2015). BRZ (Fig. 5A,B) and NPA (Fig. S5A,B) response assays demonstrate that BRs are responsible for the primary and lateral root growth phenotypes modulated by ATAF2. The root phenotypes of ATAF2 loss-of-function mutants indicate that this TF acts as a repressor of lateral root growth while acting as an activator of primary root growth in the presence of BRZ. Thus, it is possible that ATAF2 has a role in determining resource allocation and differential growth between the primary and lateral roots.

Overall, we propose that ATAF2 acts as an integrator of BR catabolism and seedling photomorphogenesis (Fig. 8D). ATAF2 suppresses expression of the BR-inactivating genes BAS1 and SOB7. As a result, ATAF2 promotes the elevation of BR levels. In turn, BRs suppress ATAF2 expression to form an ATAF2-BAS1/SOB7-BR-ATAF2 feedback regulation loop. On the other hand, ATAF2 promotes hypocotyl growth at least partially via the pathway of BAS1/SOB7-mediated BR degradation. Light suppresses hypocotyl growth through multiple pathways, one of which is by suppressing ATAF2 in a fluence rate-dependent manner. PHYA plays a major role in the ATAF2-mediated seedling photomorphogenic pathway. In this model, ATAF2 plays a key role in connecting seedling development with BR homeostasis and photomorphogenesis. Given that ATAF2 was previously reported to regulate plant defense (Delessert et al., 2005; Wang et al., 2009) and auxin biosynthesis (Huh et al., 2012), ATAF2 acts as a central regulator of plant defense, hormone metabolism and light-mediated seedling development.

Arabidopsis plants used in this study are in the Columbia (Col-0) background. T-DNA insertion mutants were obtained from the Arabidopsis Biological Resource Center (ABRC). For ataf2-1 (SALK_136355), the T-DNA insertion site is in the second exon; for ataf2-2 (SALK_015750), the T-DNA is located in the second intron. Quantitative RT-PCR (qRT-PCR) analysis (see below) demonstrated that both ataf2-1 and ataf2-2 are null alleles. For the generation of ATAF2ox lines, the ATAF2 coding sequence was overexpressed in Col-0 under the control of the CaMV 35S promoter. The pBAS1:BAS1-GUS and pSOB7:SOB7-GUS constructs were described previously (Turk et al., 2003; Sandhu et al., 2012). Homozygous single-locus T-DNA insertion lines were selected for all three transgenic events. The histochemical GUS analysis was described previously (Sandhu et al., 2012). Photoreceptor mutants (all in the Col-0 background) phyB-9 (Reed et al., 1993), phyA-211 (Reed et al., 1994) and cry1-103 (Liscum and Hangarter, 1991) were crossed with ataf2-1 to isolate double mutants with ataf2-1 from F2 populations. The bas1-2 sob7-1 double mutant (Turk et al., 2005) was crossed with ataf2-1 to isolate the bas1-2 sob7-1 ataf2-1 triple mutant. Identification by PCR of bas1-2, sob7-1, phyB-9, phyA-211 and cry1-103 mutations was described previously (Turk et al., 2005; Ward et al., 2005; Sandhu et al., 2012). Unless otherwise stated, 4-day-old seedlings were used for total RNA extraction and hypocotyl measurements, and 7-day-old seedlings were used for root growth measurements. Unless otherwise stated, seedlings were grown in growth chambers at 25°C for all physiological and molecular assays.

The Y1H assays were carried out using the Gateway-compatible system (Deplancke et al., 2006). Briefly, the promoter DNA fragments (baits) were amplified and fused to the reporter gene HIS3, and then integrated into the yeast genome. The yeast bait clones with the lowest self-activation levels were selected independently for each assay and used to either screen an Arabidopsis cDNA library for interacting protein prey or to test interactions with the targeted prey. The activation of HIS3 was tested by yeast tolerance to 3-aminotriazole (3-AT, a competitive inhibitor of the His3p enzyme). Sequences of DNA fragments used for the Y1H screen and targeted Y1H are listed in Table S1. All constructs used in this research that were generated by PCR were sequence verified. Site-directed mutagenesis of baits was performed using the QuikChange Lightning Kit (Agilent Technologies). His-tagged ATAF2 protein was purified for EMSA by procedures described previously (Zhao et al., 2013).

Total RNA was extracted using the RNeasy Plant Kit (Qiagen). On-column DNase digestion was performed to eliminate genomic DNA contamination. Total cDNA was synthesized using the SuperScript III First-Strand Synthesis System (Invitrogen). qRT-PCR reactions were performed using SsoFast EvaGreen Supermix with the Low Rox Kit (Bio-Rad) and the 7500 Fast Real-Time PCR System (Applied Biosystems). Relative expression levels were determined by normalizing to the transcript levels of UBQ10. For each qRT-PCR assay, 4-day-old seedlings of three biological replicates were grown under the same treatment conditions at the same time. qRT-PCR primers are listed in Table S3.

Seed sterilization, media composition, plating, growth conditions and hypocotyl measurement were as previously described (Sandhu et al., 2012; Zhao et al., 2013). For BL (Turk et al., 2003), BRZ (Asami et al., 2000), IAA (Collett et al., 2000) and NPA (Ruegger et al., 1997) treatment assays, seeds were put on the treatments from the beginning of the experiments. The 30 tallest seedlings from the group of three independent replicates were used for each sample in all hypocotyl measurements. For root growth measurement, seedlings were grown on vertical plates in white light. Ten seedlings with the longest primary or lateral root growth from the group of three independent replicates were used for each sample in all the root measurements. All seedlings were scanned/photographed and measured using NIH ImageJ (Schneider et al., 2012).

We thank Dr Hanjo Hellmann (Washington State University) for providing the Arabidopsis cDNA prey library.