Ethylene Response Factor6 acts as a central regulator of leaf growth under water-limiting conditions in Arabidopsis.

ETHYLENE RESPONSE FACTOR6 is a central regulator of both leaf growth inhibition and stress tolerance under osmotic stress conditions. Leaf growth is a complex developmental process that is continuously fine-tuned by the environment. Various abiotic stresses, including mild drought stress, have been shown to inhibit leaf growth in Arabidopsis (Arabidopsis thaliana), but the underlying mechanisms remain largely unknown. Here, we identify the redundant Arabidopsis transcription factors ETHYLENE RESPONSE FACTOR5 (ERF5) and ERF6 as master regulators that adapt leaf growth to environmental changes. ERF5 and ERF6 gene expression is induced very rapidly and specifically in actively growing leaves after sudden exposure to osmotic stress that mimics mild drought. Subsequently, enhanced ERF6 expression inhibits cell proliferation and leaf growth by a process involving gibberellin and DELLA signaling. Using an ERF6-inducible overexpression line, we demonstrate that the gibberellin-degrading enzyme GIBBERELLIN 2-OXIDASE6 is transcriptionally induced by ERF6 and that, consequently, DELLA proteins are stabilized. As a result, ERF6 gain-of-function lines are dwarfed and hypersensitive to osmotic stress, while the growth of erf5erf6 loss-of-function mutants is less affected by stress. Besides its role in plant growth under stress, ERF6 also activates the expression of a plethora of osmotic stress-responsive genes, including the well-known stress tolerance genes STZ, MYB51, and WRKY33. Interestingly, activation of the stress tolerance genes by ERF6 occurs independently from the ERF6-mediated growth inhibition. Together, these data fit into a leaf growth regulatory model in which ERF5 and ERF6 form a missing link between the previously observed stress-induced 1-aminocyclopropane-1-carboxylic acid accumulation and DELLA-mediated cell cycle exit and execute a dual role by regulating both stress tolerance and growth inhibition.

Drought stress is one of the most destructive environmental cues that affect plant growth and crop productivity (Boyer, 1982;Yang et al., 2010). In response to water deprivation, leaf growth is shut down by a fast and active mechanism initiated in order to save energy, as the duration and extent of the stress are unknown (Skirycz and Inzé, 2010;Skirycz et al., 2011a). This growth inhibition upon stress, however, is expected to cause yield losses that are unnecessary when the stress only lasts for short periods or when the stress is too mild to threaten the plant's survival.
The molecular processes by which mature plant organs respond to water shortage are extensively documented and are characterized by increasing abscisic acid levels activating stress-avoidance mechanisms (Xiong et al., 2002;Verslues et al., 2006;Seki et al., 2007;Schachtman and Goodger, 2008). However, the mechanisms by which stress affects actively growing plant organs are largely unknown. Although there have been many reports of transgenic Arabidopsis (Arabidopsis thaliana) lines with enhanced survival after severe water stress, an analysis of 27 of these showed no improved growth under milder, nonlethal drought conditions (Skirycz et al., 2011b). Thus, tolerance to severe drought stress and the ability of plants to continue to grow under mild stress conditions are very different traits mediated by different molecular processes. Furthermore, recent studies pointed out the importance of analyzing stress responses at the organ or tissue level, as the responses to stress both at the transcriptional level (Harb et al., 2010; and at the protein level (Baerenfaller et al., 2012) are much dependent on organ developmental stage or even on cell type identity (Dinneny et al., 2008).
In Arabidopsis, leaf development consists of three major phases during which cell proliferation, driving the growth of very young leaves, gradually switches toward cell expansion. The transition between cell proliferation and cell expansion occurs gradually, from leaf tip to leaf base, and is generally paired with a switch from the mitotic cell cycle to endoreduplication (Donnelly et al., 1999;Vlieghe et al., 2005;Anastasiou and Lenhard, 2007;Andriankaja et al., 2012;Gonzalez et al., 2012). In plants undergoing mild osmotic stress, both cell proliferation and cell expansion are affected, and as a result, leaves have fewer and smaller cells Tardieu et al., 2011). Proliferating leaves were shown to be affected in a two-step process, previously denominated the "pause-and-stop" mechanism (Skirycz et al., 2011a). In Arabidopsis, when stress occurs during early leaf development (leaves approximately 0.1 mm 2 in size), cell cycle progression is first arrested in a reversible manner by posttranslational inhibition of CYCLIN-DEPENDENT KINASE A (CDKA) activity. Only later, if the stress persists, the cell cycle pause will be converted into a definitive cell cycle exit. Cells then enter cell expansion, which is accompanied by the well-documented activation of endoreduplication and an increased DNA copy number. The exit out of the cell cycle was previously shown to be dependent on GA and DELLAs (Achard et al., 2009;Claeys et al., 2012).
In the first phase of a mild stress response in growing leaves ("pause"), we previously observed an early stress-induced increase in 1-aminocyclopropane-1-carboxylic acid (ACC) levels. ACC is the direct precursor of ethylene, a gaseous plant hormone that previously has been implicated in regulating, either positively or negatively, growth upon stress treatments. For example, ethylene was shown to either stimulate or inhibit primary root growth under low phosphate availability or under deficiency of other nutrients, respectively (Ma et al., 2003;Pierik et al., 2007). Shoot growth was shown to be positively regulated by ethylene during flooding as well as during shade avoidance, the latter as a result of increased cell expansion (Bailey-Serres and Voesenek, 2008;Jackson, 2008;Pierik et al., 2011). On the other hand, using mutants in the ethylene signaling pathway, ethylene was reported to confer growth inhibition. The ctr1 mutant, in which the ethylene signaling pathway is constitutively active, has a dwarf phenotype due to a reduction of both cell size and cell number (Roman et al., 1995;Kieber, 1997). Moreover, ethylene-insensitive mutants are generally reported to be larger than wild-type plants (Roman et al., 1995). Consistently, overexpression of the ethylene receptors increases rosette size (Cao et al., 2006(Cao et al., , 2007Wuriyanghan et al., 2009). Together, these seemingly contradictory observations can be explained by a biphasic model (Pierik et al., 2006), presenting ethylene as a growth-stimulating hormone until an optimal concentration is reached, after which ethylene inhibits growth. This optimum varies according to environmental signals, internal signals, and species-dependent factors.
Here, we identify two redundant transcription factors, ETHYLENE RESPONSE FACTOR5 (ERF5) and ERF6, as being the central regulators of leaf growth inhibition upon mild osmotic stress. ERF5 and ERF6 belong to the class of APETALA2 (AP2)/ERF transcription factors (Fujimoto et al., 2000;Nakano et al., 2006) and are situated downstream of the ethylene signaling cascade, where they regulate ethylene-responsive genes (Yoo et al., 2009). Recently, these two transcription factors were shown to be able to interact with each other at the protein level, although further investigations are necessary to confirm an in planta interaction (Son et al., 2012). Although ERF5 and ERF6 have not yet been extensively characterized, they recently have been described as being important regulators of biotic stress defense (Moffat et al., 2012;Son et al., 2012). Besides its function in response to biotic stress, ERF6 was recently shown to control the expression of reactive oxygen species-responsive genes after activation by MPK3/ MPK6 (Wang et al., 2013).
In this study, we focus on ERF6, demonstrating that it affects cell cycle exit by triggering the expression of the GA2-OXIDASE6 (OX6) gene and consequently the inactivation of GAs. Thus, ERF6 provides a link between ACC and DELLA signaling in the cell cycle pause-and-stop model, improving our understanding of growth inhibition in the proliferating leaf primordia of plants subjected to water limitation. In addition, ERF6 regulates, in a GA-and DELLA-independent manner, the expression of multiple genes associated with abiotic and biotic stress conditions, such as genes encoding the transcription factors WRKY33, MYB51, and STZ (for salt tolerance zinc finger). Thus, ERF6 plays a dual role under stress, as it activates both stress tolerance and growth inhibition, and, importantly, these two roles occur independently from each other.

ERF5 and ERF6 Are Transcriptionally Induced in Actively Growing Leaves within 1 h of Stress Exposure
We recently investigated the effects of mild osmotic stress on the transcriptome of very young, small (approximately 0.1 mm 2 in size), and thus still actively growing leaves (Skirycz et al., , 2011a. Among the more than 1,500 genes differentially expressed following exposure to mild osmotic stress, the transcription factors ERF5 (AT5G47230) and ERF6 (AT4G17490) were induced very early after stress onset, already 1 h after stress exposure specifically in actively growing leaves, and their induction was further maintained over time (Supplemental Fig. S1A). Furthermore, analysis of publicly available transcriptome data revealed that ERF5 and ERF6 are induced by several other, often severe abiotic stresses, including drought (Supplemental Fig. S1B; Hruz et al., 2008). These data prompted us to investigate the role of ERF5 and ERF6 in integrating environmental signals into leaf growth regulation.
Within the large class of more than 120 AP2/ERF transcription factors, ERF5 and ERF6 belong to group IXb, a small group of transcriptional activators containing several other stress-responsive ERFs, such as ERF1 and ERF2 (Supplemental Fig. S2; Nakano et al., 2006;. ERF5 and ERF6 share 51% amino acid similarity, with high conservation of three functional domains: CMIX-2, CMIX-5, and the AP2/ ERF domain (Thompson et al., 1994). To investigate the importance of ERF5 and ERF6 in leaf growth under various conditions, we used single and double mutants. Single erf5 and erf6 mutants were obtained from the SALK collection and have transfer DNA insertions in the 39 untranslated region and coding sequence, respectively (Supplemental Fig. S3A; Alonso et al., 2003;Wang et al., 2013). The ERF5 and ERF6 expression levels were strongly decreased both under normal and osmotic stress conditions (Supplemental Fig. S3B). The erf5 and erf6 single mutants had no obvious growth phenotype, most likely because of functional redundancy; thus, the erf5erf6 double mutant was used for further analysis.
First, we explored how erf5erf6 plants behave under standard conditions and when exposed to various long-term abiotic stress conditions. As we were interested in measuring growth dynamics in response to stress instead of limiting our analysis to end-point measurements, we chose to grow the erf5erf6 mutant and the wild-type line on an automated phenotyping platform named the In Vitro Growth Imaging System (IGIS; see "Materials and Methods"). The IGIS platform allows for a continuous measurement of rosette size of in vitro-grown Arabidopsis plants by taking photographs every hour from germination onward until 20 d after stratification (DAS) and extracting the rosette area. The erf5erf6 mutant and the wild type were exposed to different mild abiotic stresses: osmotic stress (25 mM mannitol, mimicking mild drought stress; Skirycz et al., 2011a), oxidative stress (1.5 mM hydrogen peroxide), and salt stress (50 mM NaCl). Interestingly, growth curves representing rosette area over time (Fig. 1A) demonstrate that under standard conditions, the erf5erf6 mutant grows faster than the wild type. At the end point (20 DAS), erf5erf6 was 13% larger (Supplemental Table S1). When grown on osmotic stress from germination onward, erf5erf6 mutants tolerated the stress better than wild-type plants. The erf5erf6 mutants were less affected by the stressinduced growth inhibition and showed an increase in final rosette area of 33% compared with wild-type plants exposed to this stress (Supplemental Table S1). Importantly, the same tendency was observed for erf5erf6 plants exposed to long-term oxidative stress and was consistent in two out of the three experiments. However, the long-term hydrogen peroxide treatment introduced significant variability between the experiments, making them poorly reproducible and thus rather difficult to interpret (details per experiment are provided in Supplemental Table S1). Finally, when exposed to long-term mild salt stress (50 mM NaCl), the erf5erf6 plants were not larger than wild-type plants, suggesting that both transcription factors have no role in tolerance to salt stress. This hypothesis is further supported by the observation that exposure of plants to 50 mM NaCl does not induce ERF5 and ERF6 expression in very young, small proliferating leaves (Supplemental Fig. S4). Together, these data show that ERF5 and ERF6 are central regulators that orchestrate leaf growth under long-term mild osmotic and possibly oxidative stress conditions but not under salt stress.
Next we investigated the growth of erf5erf6 when exposed to a short-term mild osmotic stress (25 mM mannitol). Briefly, the assay consists of growing plants on a nylon mesh covering control Murashige and Skoog (MS) medium until the third leaf has completely emerged from the shoot apical meristem but is still in a fully proliferative stage, at 9 DAS. At this time point, the mesh is transferred to MS medium containing 25 mM mannitol, and the effect of the stress is analyzed daily by measuring the growth of the third leaf. As previously shown, wild-type plants exposed to 25 mM mannitol show a reduction of leaf area of about 50%, caused by a reduction of both cell number and cell size (Skirycz et al., , 2011aFig. 1B). Importantly, the growth of erf5erf6 was significantly (P = 0.0004) less affected than that of the wild type ( Fig. 1B): on average, wild-type plants showed a 42% leaf size reduction, while the leaf size of erf5erf6 was only decreased by 11% (Supplemental Table S2, C and D). As a consequence, third leaves harvested at 19 DAS from erf5erf6 plants exposed to stress are 59% larger in comparison with those of wild-type plants exposed to stress (P = 0.025). Thus, the third leaf of erf5erf6 plants exposed to short-term mild osmotic stress continued to grow almost indistinguishably from that of wild-type plants grown on standard medium. Interestingly, this reduced leaf growth inhibition is already visible at 12 DAS ( Fig.  1B), suggesting that ERF5 and ERF6 act early in leaf development.

ERF6 Represses Leaf Growth by Inhibiting Cell Division and Cell Expansion
To investigate the cellular basis of ERF6-mediated growth inhibition, inducible gain-of-function ERF6 lines were analyzed. To this end, the ERF6 sequence was C-terminally fused to that of a glucocorticoid receptor (GR) domain and expressed in transgenic plants under the control of the cauliflower mosaic virus 35S promoter (p35S). This chimeric ERF6-GR protein is expected to stay in the cytoplasm, but after addition of the steroid hormone dexamethasone (DEX), it undergoes conformational changes and migrates to the nucleus, where it becomes functional as a transcription factor (Corrado and Karali, 2009). Two glucocorticoid-inducible overexpression (IOE) lines were used for further analysis: a strong ERF6-overexpressing line referred to as ERF6 IOE -S (fold change = 7,000 as compared with the p35S-GFP-GR line, measured in young seedlings) and a weaker line denominated ERF6 IOE -W (fold change = 220). As a control, a DEX-inducible 35S::GFP-GR (GFP: IOE) line was used. To examine the effect of ERF6 overexpression specifically on actively growing leaves, plants were germinated on a nylon mesh overlaying MS medium and transferred to medium containing 5 mM DEX at 9 DAS, when the third leaf is fully proliferating. ERF6 activation from 9 DAS onward drastically reduced the growth of this leaf: at 16 DAS, leaf area reductions of 83% and 55% were measured for ERF6 IOE -S and ERF6 IOE -W, respectively, and significant reductions were already observed 48 and 72 h after DEX treatment of ERF6 IOE -S and ERF6 IOE -W, respectively ( Fig. 2A). At the cellular level, the severe growth reduction of ERF6 IOE -S at 16 DAS was caused by both smaller (57%) and fewer (59%) leaf cells (Supplemental Fig. S5). As expected, this cellular phenotype was less pronounced in the weaker ERF6 IOE -W line, in which mainly cell area was affected (Supplemental Fig. S5). Flow cytometry showed an increase in endoreduplication upon strong ERF6 overexpression, suggesting that ERF6 pushes cells from mitosis into endoreduplication and differentiation (Supplemental Fig. S6). When left on DEX for longer times (until 22 DAS), ERF6-overexpressing plants remained dwarfed and dark green, with stunted rosettes (Fig. 2B). Similar phenotypes were observed for soil-grown plants sprayed daily with a 5 mM DEX solution from 9 DAS onward (Fig. 2B). Taken together, these data show that ERF6 expression levels inversely correlate with leaf growth.
To further investigate the role of ERF6 in stressmediated growth inhibition, we exposed the ERF6 IOE -W line to short-term osmotic stress. We only used the ERF6 IOE -W line, as the ERF6 IOE -S line shows a very severe phenotype making the accurate measurement of subtle growth changes rather difficult. The ERF6 IOE -W line was grown on MS medium until 9 DAS and then transferred to medium with or without 25 mM mannitol, in combination with or without DEX. In the presence of DEX, the ERF6 IOE -W line was found to be hypersensitive to osmotic stress as compared with the wild type (Fig. 2C). When first germinated on MS medium and then transferred to either DEX or Figure 1. The erf5erf6 double mutant is more tolerant to mild osmotic stress conditions. A, Rosette area over time of wild-type plants (WT) and erf5erf6 double mutants under standard MS medium and different stress conditions. The erf5erf6 mutant shows significant tolerance to osmotic stress (MS medium supplemented with 25 mM mannitol) only. Colored shadows indicate SE. Three biological repeats were performed with at least 12 seedlings per line per treatment. B, Leaf area measurements (third leaf) of the erf5erf6 mutant and the wild type upon transfer at 9 DAS to standard or mild osmotic stress conditions. On osmotic stress, the erf5erf6 mutant is always about 50% larger than the wild type (for detailed measurements, see Supplemental Table S2). Error bars indicate SE. Three biological repeats were performed with 16 leaves per repeat. H 2 O 2 , Hydrogen peroxide. mannitol, ERF6 IOE -W plants showed a reduction in growth but were still able to develop normally. However, when transferred to mannitol + DEX, the plants failed to develop correctly and were extremely dwarfed. Together, our data confirm that ERF6 plays an important role in modulating leaf growth under stress.

ERF6 Inhibits Growth through a GA/DELLA-Dependent Mechanism
To elucidate how ERF6 reduces leaf growth, we performed a genome-wide analysis of genes rapidly induced by the activation of ERF6 overexpression. To this end, 9-d-old ERF6 IOE -S plants were transferred for 4 h to DEX, and subsequently, third leaf primordia (which are then smaller than 0.1 mm 2 in size) were microdissected and subjected to AGRONOMICS1 tiling arrays (Rehrauer et al., 2010). Already 4 h after DEX treatment, 344 genes were differentially expressed (false discovery rate-corrected P , 0.05), of which 332 were induced (Supplemental Table S3), suggesting that ERF6 acts as an activator of gene expression. Gene Ontology annotation analysis of the 332 up-regulated genes using BiNGO (Maere et al., 2005) revealed that the putative ERF6 targets are highly enriched in several stress-related and biological signaling process categories, such as "response to water stress," "response to chemical stimulus," "response to biotic stimulus," and "response to ethylene" (Supplemental Fig. S7), again strongly suggesting a role for ERF6 in early stress response. Importantly, when comparing the putative ERF6 target genes with the previously identified list of genes specifically induced in leaf initials within 3 h of exposure to osmotic stress (Skirycz et al., 2011a), we observed a highly significant overlap (18.5 times higher than expected by chance; Fig. 3). Out of the 332 putative ERF6 targets, 56 were found to be induced within 3 h of mannitol treatment (Supplemental Table S4). Interestingly, 14 of the 27 genes induced after 1.5 h of mannitol treatment are differentially expressed upon ERF6 induction (62.3 times more than expected by chance), again underlining the central role for ERF6 in early stress response. An additional 56 ERF6-induced genes are also found to be induced 12 and 24 h following mannitol treatment in proliferating leaves.
Genes that are rapidly induced by DEX-mediated activation of ERF6 are putative target genes. One of these genes, GA2-OX6, encoding an oxidase involved in GA inactivation, deserves particular attention because the ERF6 IOE dwarf phenotype phenocopies that of plants insensitive to GAs (Peng et al., 1997;Thomas and Sun, 2004). Moreover, from our previously established list of osmotic stress-responsive genes (Skirycz et al., 2011a), we observed that GA2-OX6 is the only GA 2-oxidase activated by mild osmotic stress (Supplemental Fig. S8). We subsequently tested whether the ERF6mediated activation of GA2-OX6 expression could explain the growth retardation caused by ERF6 activation. Time-course quantitative PCR analysis confirmed the induction of GA2-OX6 expression within 2 h after the activation of strong ERF6 overexpression (Fig. 4A). Consistently, there is a correlation between the timing and level of GA2-OX6 induction and the observed growth inhibition, as seen in the ERF6 IOE -W line, where upon DEX treatment the growth is less affected and the GA2-OX6 induction by ERF6 is slower and less pronounced (Supplemental Fig. S9). Importantly, in the erf5erf6 mutant, the growth of which is less affected by osmotic stress, the induction of GA2-OX6 following stress exposure is delayed (Supplemental Fig. S10).
Our data support a model in which ERF6 is able to activate GA2-OX6 expression, triggering GA breakdown and consequently stabilizing DELLA proteins, which are known to negatively affect growth. To further confirm this model, we investigated whether DELLAs were stabilized after ERF6 activation by crossing the ERF6 IOE lines with a GFP-tagged DELLA reporter line (pRGA:: GFP-RGA; Silverstone et al., 2001). Stabilization of RGA, the major DELLA expressed in developing leaves , following ERF6 activation can be followed by measuring GFP protein levels. Western-blot analysis using a primary antibody against GFP demonstrated the stabilization of RGA between 12 and 24 h after ERF6 activation (Fig. 4B). Finally, to confirm the involvement of GA in the ERF6-mediated growth arrest, we tested whether the growth inhibition activated by ERF6 could be abolished by crossing the ERF6 IOE -W line with a transgenic line overexpressing the rate-limiting GA biosynthetic enzyme GA 20-oxidase1 (35S-GA20-OX1; Coles et al., 1999;Gonzalez et al., 2010). When grown on DEX, the resulting plants (ERF6 IOE -W 3 35S-GA20-OX1), in contrast to ERF6 IOE -W plants, did not show any growth retardation (Fig. 4C), and the final size (at 21 DAS) of the third leaf was similar to that of control GFP:IOE plants (Fig. 4D). A similar, albeit more partial, restoration of growth was obtained for ERF6 IOE -S 3 35S-GA20-OX1 plants (Fig. 4C). Together, these data confirm that, under osmotic stress, ERF6 induces the expression of the gene encoding the GA-inactivating enzyme GA2-OX6, thereby reducing the bioactive GA levels and stabilizing the DELLA proteins.

ERF6 Activates a Plethora of Stress-Responsive Genes
We subsequently further explored the ERF6 regulon of 332 putative ERF6 target genes, which is enriched for stress-responsive genes, suggesting that ERF6 plays a role in orchestrating, besides growth, also early stressinduced gene expression. Multiple genes encoding stress-related transcription factors were identified within the overlap of putative ERF6 target genes and genes transcriptionally induced within 24 h after osmotic stress exposure (Skirycz et al., 2011a). To further uncover the link between ERF6 and this stress-related transcription factor network, we chose three representative genes, STZ, WRKY33, and MYB51, as these were previously shown to have a role in biotic and abiotic stress signaling (Sakamoto et al., 2000(Sakamoto et al., , 2004Gigolashvili et al., 2007;Jiang and Deyholos, 2009;Birkenbihl et al., 2012;Li et al., 2012;Niu et al., 2012). Additional quantitative PCR analysis confirmed the induction of these genes within 2 h after DEX-mediated ERF6 activation using the ERF6 IOE -S (Fig. 5A) and ERF6 IOE -W (Supplemental Fig. S9) transgenic lines, thereby rendering them primary candidates for being direct ERF6 targets. This is supported using a protoplast activation assay with promoter-luciferase reporter constructs (pSTZ: fLUC, pWRKY33:fLUC, and pMYB51:fLUC), in which the respective promoters were cloned upstream of the fLUC gene (encoding the firefly luciferase enzyme) and expressed together with a 35S-ERF6 or 35S-ERF5 construct in tobacco (Nicotiana tabacum) Bright Yellow-2 (BY-2) protoplasts. Binding of ERF6 or ERF5 to the promoter of interest triggers the expression of the fLUC gene and the production of the luciferase enzyme. A significant increase in luciferase activity shows the activation of pSTZ, pWRKY33, and pMYB51 by both ERF5 and ERF6 (Fig. 5B). In summary, these data strongly suggest that ERF5 and ERF6 directly activate the expression of the stress-related transcription factor genes STZ, WRKY33, and MYB51.
Finally, we investigated whether the two functions of ERF6, being on the one hand leaf growth regulation and on the other hand the activation of a stress defense transcriptional cascade, are interdependent. For this purpose, we used the ERF6 IOE -W 3 35S-GA20-OX1 and ERF6 IOE -S 3 35S-GA20-OX1 lines, in which the Figure 3. Overlap between ERF6 targets and genes rapidly induced by 25 mM mannitol. Comparison was performed of the 332 putative ERF6 targets with the previously identified list of genes specifically induced in leaf initials within hours upon exposure to osmotic stress (Skirycz et al., 2011a). Values indicated in the Venn diagram represent the number of genes induced upon 1.5-and 3-h mannitol treatment and the genes induced in leaf initials 4 h following DEX application in ERF6 IOE -S. growth inhibition is entirely and partially abolished, respectively. Importantly, when ERF6 overexpression was activated at 9 DAS by DEX treatment for 8 h, the stress defense transcriptional cascade (represented here by the expression of STZ, WRKY33, and MYB51) was still activated at least as highly as in the positive control lines (Fig. 5C). The expression of STZ and WRKY33 was even considerably higher in the ERF6 IOE -W 3 35S-GA20-OX1 plants (without growth reduction) as compared with that in ERF6 IOE -W 3 35S-GFP plants. Thus, although the growth inhibition by ERF6 was suppressed, the stress-related transcription network was still active. These results are of significant importance, as they demonstrate that growth reduction caused by mild stress can be uncoupled from the stress defense response.

ERFs Are Rapidly Induced by Osmotic Stress
Osmotic stress was previously demonstrated to induce the expression of a large number of genes in actively growing leaves (more than 1,500 genes). Within 1.5 and 3 h of stress exposure, only a small number (27 and 193, respectively) were rapidly induced (Skirycz et al., 2011a). This suggests that stress signaling occurs through cascades in which, in a simplified view, a few rapidly activated transcription factors orchestrate other transcription factors, which in turn switch on their own targets. In this context, ERF5 and ERF6, two transcription factors that are transcriptionally induced within 1 h of osmotic stress, belong to a very confined group of early stress regulators. Intriguingly, the expression of the ERF6 targets (STZ, WRKY33, and MYB51) is induced within 1 h of osmotic stress as well, suggesting that there is yet another mechanism than ERF6 transcription that triggers STZ, WRKY33, and MYB51 expression. A possible explanation for the fast induction of STZ, WRKY33, and MYB51 could be that ERF6 is expressed at basal levels under standard conditions (Andriankaja et al., 2012;Wang et al., 2013) and that ERF6 is posttranscriptionally activated by osmotic stress. A similar mechanism was recently shown to occur during oxidative stress, where ERF6 is phosphorylated by two mitogen-activated protein kinases (MPK3 and MPK6; Wang et al., 2013). Both kinases act downstream of ACC and independently of the EIN2 signaling pathway (Yoo and Sheen, 2008), and we have previously demonstrated their presence during very early osmotic stress (Skirycz et al., 2011a). Importantly, MPK3 and MPK6 were recently shown to physically interact with ERF5 and ERF6, and phosphorylation of ERF5 and ERF6 by MPK3 and MPK6 Figure 4. ERF6 regulates GA levels through transcriptional control of the GA2-OX6 gene. A, Induction of GA2-OX6 following the activation of ERF6 overexpression. Two hours after transfer to DEX of ERF6 IOE -S, GA2-OX6 is significantly induced. Expression was measured in proliferating third leaves, and values are normalized to their expression in the GFP:IOE control line exposed to the same treatment. Error bars indicate SE of three repeats with 64 young third leaves per repeat. B, Stabilization of the DELLA protein RGA upon the activation of ERF6 overexpression shown by western blot, targeting the GFP domain of the RGA-GFP fusion protein in pRGA:GFP-RGA 3 ERF6 IOE -S seedlings. DELLA stabilization first clearly appears 24 h after ERF6 activation. Three biological replicates were performed. C, Growth complementation assay. By crossing the two independent ERF6:IOE lines with a 35S-GA20-OX line (ectopic GA overproduction), the dwarfed phenotype could be partially and fully complemented in ERF6 IOE -S and ERF6 IOE -W lines, respectively. Treatment with DEX was applied at 9 DAS, and photographs were taken at 21 DAS. D, Measurements of third leaves at 21 DAS of GFP:IOE, ERF6 IOE -W, and ERF6 IOE -W 3 35S-GA20-OX upon treatment with DEX at 9 DAS. Error bars indicate SE of three repeats with 12 leaves per repeat. [See online article for color version of this figure.] could be demonstrated (Popescu et al., 2009;Son et al., 2012;Wang et al., 2013). The very rapid transcriptional induction of ERF6 following stress exposure results from an autoactivation loop in which phosphorylated ERF6 activates its own expression, independent from induction by the upstream EIN3 and EIL1 transcription factors (Supplemental Fig. S11). Therefore, we propose the hypothesis that ERF5 and ERF6, present at the basal level prior to stress exposure, are upon stress treatment rapidly phosphorylated through MPK3 and MPK6 and thereby converted into active transcription factors, able to rapidly regulate their own expression and the expression of STZ, MYB51, and WRKY33.

ERF5 and ERF6 Form the Connection between ACC Accumulation and the GA/DELLA Response in Leaves Subjected to Mild Drought Stress
Recently, we have shown that growth inhibition by mild osmotic stress response occurs in two steps: first, a pause step, in which the cell cycle is temporally arrested in an ACC-dependent manner by inhibition of CDKA; and later, if stress is maintained, a stop mechanism, which pushes cells irreversibly out of the cell cycle and into cell differentiation (Skirycz et al., 2011a). The last step was shown to be mediated by DELLAs, driving cells into early endoreduplication and thus mitotic exit (Claeys et al., 2012). However, it was not yet clear how the early ACC accumulation causes DELLA stabilization. Here, we propose that ERF5 and ERF6 form the connection between stress sensing and GA/DELLA signaling (Fig. 6A). The presence of GA2-OX6 among the putative ERF6 targets strongly supported this hypothesis. The GA2-OX6 gene encodes a GA-inactivating enzyme and its induction thus decreases the levels of bioactive GA, thereby stabilizing the DELLAs (Fig. 6A). An analysis of various transgenic lines with altered ERF6 levels and exposure of these lines to standard and mild osmotic stress conditions revealed a remarkable correlation between the levels of ERF6 expression and the severity of growth inhibition (Fig. 6B). ERF6 IOE -S seedlings, which are characterized by very strong ERF6 overexpression, are completely dwarfed. As osmotic stress transcriptionally induces ERF6 expression and most likely triggers ERF6 activation, it is expected that osmotic stress would aggravate the growth phenotype. This is exactly what was observed for ERF6 IOE -W plants, which are smaller but still develop normally. Exposing these weak Figure 5. ERF6 regulates the stress-related transcription factors STZ, MYB51, and WRKY33. A, Induction of STZ, MYB51, and WRKY33 following the activation of ERF6 overexpression. Within 2 h of the transfer of ERF6 IOE -S to DEX, STZ, WRKY33, and MYB51 are significantly induced. Expression was measured in proliferating third leaves, and values are normalized to their expression in the GFP:IOE control line exposed to the same treatment. Error bars indicate SE of three repeats with 64 young third leaves per repeat. B, ERF5/ERF6-dependent activation of the promoters of MYB51, WRKY33, and STZ by the protoplast activation assay. Indicated values are luciferine detection levels normalized to the negative control. Asterisks indicate significantly different values from the control at the 1% (**) and 5% (*) significance levels. Error bars indicate SE, and eight biological repeats were performed. C, Induction of STZ, MYB51, and WRKY33 expression 8 h after the activation of ERF6 overexpression in ERF6 IOE 3 35S-GA20-OX plants. Although the dwarfed growth phenotype is partially and completely rescued in ERF6 IOE -S 3 35S-GA20-OX and ERF6 IOE -W 3 35S-GA20-OX plants, respectively (Fig. 4B), the stress-related transcription factors are still induced to the same extent as in the positive control lines (ERF6 IOE -S 3 GFP:IOE and ERF6 IOE -W 3 GFP:IOE, respectively). Expression values are normalized to their expression in the control line (GFP:IOE). Error bars indicate SE of three repeats with 64 young third leaves per repeat.
Interestingly, the phenotype of ERF6-overexpressing plants strongly resembles that of the 35S:gai-GR line, in which a mutant GA-insensitive version of the DELLA protein GAI is overexpressed (Claeys et al., 2012). Both lines show the same cellular phenotype in the presence of DEX, with less and smaller epidermal cells. Importantly, the cellular phenotype observed in ERF6-overexpressing plants matches that of the epidermal cells of plants exposed to mild osmotic stress . The reduced cell number, however, could only be obtained by strong ERF6 overexpression and was not clear after weak ERF6 overexpression, indicating that ERF6 mainly works on cell expansion and to a lesser extent on cell division, in accordance with how DEL-LAs inhibit root and probably shoot growth (Achard et al., 2009). Finally, in the ERF6-overexpressing plants, proliferating cells of young leaves are pushed faster into endoreduplication, a process mediated through GA/ DELLA signaling that was also observed under osmotic stress (Skirycz et al., 2011a). In conclusion, several lines of evidence show that ERF5 and ERF6 provide a link between ACC accumulation and DELLA signaling in the pause-and-stop model.
The involvement of DELLA and/or ethylene signaling in growth and stress responses is not restricted to osmotic stress. A well-known example is the involvement of C-REPEAT/DROUGHT-RESPONSIVE ELEMENT BINDING FACTOR1 (CBF1) in freezing acclimation and growth inhibition under cold stress, upstream of several GA 2-oxidases (GA2-OX3, GA2-OX6, and GA2-OX1), thereby causing DELLA accumulation (Achard et al., 2008). Another AP2 transcription factor, DDF1, was shown to play a similar function under salt stress by regulating the expression of the GA2-OX7 gene (Magome et al., 2008). The data presented here demonstrated an analogous role for ERF6 in regulating growth under mild osmotic and oxidative stress. Moreover, whereas CBF1 acts upstream of several GA 2-oxidases (GA2-OX3, GA2-OX6, and GA2-OX1), we provide evidence that osmotic stress specifically involves GA2-OX6 (Supplemental Fig. S8). Thus, although hormonal interactions between ethylene and GA/ DELLA might regulate general growth mechanisms that are shared between different abiotic cues, the identity of the molecular players involved is highly condition and tissue specific. We speculate that, although these transcription factors all activate, on the one hand, the DELLA-mediated growth inhibition, they each activate, on the other hand, a specific cluster of stress tolerance genes according to the type of stress.

ERFs Regulate Many Stress Resistance Genes in a GA/DELLA-Independent Way
Genome-wide identification of putative ERF6 target genes provided a list of 332 genes highly enriched for Figure 6. How ERF5 and ERF6 regulate leaf growth and stress defense under osmotic stress. A, Immediately upon exposure to osmotic stress, ACC accumulates in the actively growing leaves, where it is converted to ethylene. Ethylene further activates the signaling pathway involving MPK3 and MPK6. These kinases phosphorylate the basal amount of ERF5 and ERF6 proteins present in the cell prior to stress exposure. The activated ERF5 and ERF6 then execute their dual functions: on the one hand, the activation of the stress defense transcriptional cascade with direct transcriptional activation of WRKY33, STZ, and MYB51, and on the other hand, the activation of leaf growth inhibition. This occurs through the transcriptional activation of the gene encoding the GA-inactivating enzyme GA2-OX6, thereby decreasing the bioactive GA concentration and stabilizing the DELLA proteins. B, In accordance with the model presented in A, ERF6 transcript levels and stressmediated activity inversely correlate with leaf growth (for discussion, see text). Bar = 1 cm. [See online article for color version of this figure.] genes involved in stress response and signaling. Among them were the well-known stress-related transcription factors STZ, WRKY33, and MYB51. Cotransfection of protoplasts with a promoter-luciferase reporter and 35S-ERF5 or 35S-ERF6 shows that STZ, WRKY33, and MYB51 are most likely direct ERF5 and ERF6 target genes. STZ is a transcriptional repressor activated under severe salt stress to control survival mechanisms (Sakamoto et al., 2000(Sakamoto et al., , 2004Mittler et al., 2006). MYB51 is a homeodomain-like transcription factor known to regulate the biosynthesis of indole-glucosinolates, a class of secondary metabolites involved in defense against herbivores (Gigolashvili et al., 2007). Finally, WRKY33 is a transcriptional activator involved in plant survival under high-salt, cold, and severe osmotic stress (Jiang and Deyholos, 2009). Furthermore, WRKY33 has a role in biotic stress defense, where it regulates the balance between necrotrophic and biotrophic pathogen responses (Lippok et al., 2007;Pandey and Somssich, 2009;Birkenbihl et al., 2012). In another recent study dealing with biotic stress, WRKY33 also was found downstream of ERF5, but surprisingly, and in contrast to the data presented here, ERF5 overexpression appeared to downregulate WRKY33 (Son et al., 2012). A possible reason for this discrepancy is that in the latter study, plants constitutively overexpressing ERF5 were used, possibly activating negative feedback loops suppressing ERF5 activity. In contrast, using the inducible ERF6 overexpression line, we could show that WRKY33 expression increases gradually over time, demonstrating that under osmotic stress conditions, ERF6 works as a transcriptional activator to regulate WRKY33 expression. Supporting our observations, ERF6 was recently shown to be necessary for WRKY33 induction under oxidative stress (Wang et al., 2013). Consistent with the activation of stress tolerance genes by ERF5 and ERF6, overexpression of SlERF5 in tomato (Solanum lycopersicum) plants was recently shown to confer tolerance to drought stress (Pan et al., 2012).
ERF6 regulates two diverse processes: on the one hand, the activation of the stress defense transcriptional cascade, and on the other hand, the regulation of growth inhibition. Our data show that both processes can be uncoupled. Overexpression of the gene encoding the GA biosynthetic enzyme GA20-OX1 (Coles et al., 1999) in ERF6-overexpressing plants suppressed the growth reduction phenotype but left the ERF6mediated induction of stress response genes intact. Thus, in plants in which the stress signaling pathway is activated through ERF6 overexpression, the stress tolerance factors remain activated even when growth inhibition is completely suppressed. This is similar to CBF1, which regulates a cluster of cold-responsive genes in a DELLA-independent way (Achard et al., 2008). From an agricultural point of view, this means that it should be possible to generate crops that are less affected by mild drought in terms of growth but are still able to activate their stress defense mechanisms.

ERF5 and ERF6 Regulate Growth under Multiple, But Not All, Abiotic Stresses
The erf5erf6 double mutant is more tolerant to both short-term and long-term osmotic stress, most likely because GA2-OX6 expression is no longer activated (Supplemental Fig. S10). Although less clear due to experimental variability, a similar tendency is observed for plants exposed to long-term oxidative stress. Surprisingly, although salt stress is generally known to be closely related to osmotic stress, erf5erf6 plants do not tolerate mild salt stress better than wild-type plants. This observation was supported by expression analysis demonstrating that ERF5 and ERF6 have no role in salt stress signaling in actively growing leaves. Consistently, in proliferating leaves, mild salt stress does not induce the expression of the GA2-OX6 gene, in contrast to osmotic and oxidative stress. It is thus likely that growth inhibition induced by mild salt stress occurs independently of the ERF5/ERF6-centered growth regulatory pathway. Both drought and salt stress are characterized by the reduced ability to take up water, causing cellular dehydration and wilting. In addition, since salt ions are taken up by plant cells, plants have to cope with toxic levels of Na + . In the majority of species, NaCl concentrations above 40 mM cause toxicity (Munns and Tester, 2008). In this study, plants were exposed to 50 mM NaCl and the observed growth reduction was probably mainly caused by NaCl toxicity and thus less related to osmotic stress defense. Although salt and osmotic stress show similarities, genome-wide expression studies revealed that large sets of genes are specifically induced by only one of these stresses (Denby and Gehring, 2005). Depending on the duration and exact conditions by which the stresses were applied, the overlap of genes responding to both abiotic stresses varied on average between 10% and 40%. When comparing only expression analyses performed on shoot tissue, the overlap was reduced to 3%. Thus, early stress-sensing and signaling responses in Arabidopsis shoots are mainly specific to either salt or osmotic stress, and our data clearly support this notion in actively growing leaves. Both osmotic stress and salt stress implicate the ethylene precursor ACC as an early signal (Zhang et al., 2011), and the molecular mechanisms by which this difference between salt and osmotic stress response is established in actively growing leaves are far from resolved. We speculate that the intermediate regulator acting between ACC and ERF5/ERF6 in the cascade is active after either osmotic or salt stress and activates or inhibits ERF5/ERF6 specifically in this condition. A putative candidate for such a regulator is NEK6, a kinase transcriptionally induced by ACC and by salt stress (Skirycz et al., 2011a;Zhang et al., 2011) but not by osmotic stress (Skirycz et al., 2011a). We further speculate that upon salt stress, the NEK6 kinase either rapidly phosphorylates and thereby inactivates ERF5/ ERF6 or inhibits the ethylene biosynthesis pathway (Zhang et al., 2011), establishing a slower but stable ERF5/ERF6 inhibition. It is likely that CBF1 is the functional equivalent of ERF6 under salt stress conditions (Achard et al., 2008). This probably allows the activation of genes conferring tolerance to sodium toxicity, which are not activated by ERF6.

CONCLUSION
In this study, we provide a missing link between ACC accumulation and DELLA stabilization in the pause-and-stop mechanism by which Arabidopsis leaf growth is shut down under osmotic stress. We uncovered a dual regulatory role for the transcription factors ERF5 and ERF6 and propose them to be central elements in a signaling network summarized in Figure  6A. Mild osmotic stress triggers the accumulation of ACC and, most likely, initiates an ACC-dependent signaling cascade involving MPK3 and MPK6. These mitogen-activated protein kinases activate ERF5 and ERF6, which in turn initiate transcription of the GA2-OX6 gene encoding a GA-degrading enzyme. GA breakdown stabilizes DELLA proteins and represses growth. In parallel, ERF5 and ERF6 activation triggers the expression of stress tolerance factor genes such as STZ, WRKY33, and MYB51. Interestingly, the ERF5-and ERF6-mediated growth inhibition and the activation of the stress-responsive network can be uncoupled. This uncoupling holds great potential for engineering crops that are less inhibited by mild stress while maintaining stress tolerance.

Plant Lines
The single erf5 and erf6 mutants of Arabidopsis (Arabidopsis thaliana) were obtained from the SALK collection, references SALK_076967 (erf5) and SALK_030723 (erf6). The pRGA:RGA-GFP line was a kind gift of Prof. Dr. Tai-ping Sun (Duke University). The 35S-GA20-OX1 line used was previously described by Gonzalez et al. (2010) and originally was a gift from P. Hedden (Coles et al., 1999). All lines used are in the Columbia background.

In Vitro Plant Growth Conditions
Seedlings were grown in vitro on one-half-strength MS medium (Murashige and Skoog, 1962) containing 1% Suc at 21°C under a 16-h-day (110 mmol m 22 s 21 ) and 8-h-night regime. For long-term experiments where no transfer was needed, 9 g L 21 agar was added to the medium. For short-term experiments involving transfer, 6.5 g L 21 agar was used, and the growth medium was overlaid with nylon mesh (Prosep) of 20-mm pore size to facilitate transfer. For expression analysis and growth experiments, 64 and 16 seeds, respectively, were equally distributed on a 14-cm-diameter petri dish. The different ERF6 gain-and loss-offunction lines were always grown together with the appropriate control on one plate to enable correct comparisons.

Exposure to Short-Term Osmotic Stress and/or Glucocorticoid-Induced Activation of ERF6
Plants were grown on a nylon mesh covering control MS medium until the third leaf had completely emerged from the shoot apical meristem but was still in a fully proliferative stage, at 9 DAS. At this time point, the mesh was transferred to plates with one-half-strength MS medium containing 25 mM D-mannitol (plant culture tested; Sigma), 5 mM DEX (Sigma), or a combination of both. For expression analysis and growth experiments, all seedlings were transferred to DEX, including the GFP:IOE control lines, to account for the possible effects of DEX on growth or gene expression.

Growth Analysis
Growth analysis was performed on the third true leaf harvested at different time points after transfer to DEX. After clearing with 70% ethanol, leaves were mounted in lactic acid on microscope slides. For each experiment, about 15 to 20 leaves were photographed with a binocular microscope, and abaxial epidermal cells (100-200) were drawn for three representative leaves with a DMLB microscope (Leica) fitted with a drawing tubus and a differential interference contrast objective. Photographs of leaves and drawings were used to measure leaf area and cell size, respectively, using ImageJ version 1.37o (http://rsb.info.nih.gov/ij/), and average cell numbers were calculated by dividing leaf area by cell area.

Sampling RNA for Expression Analysis
Samples were obtained from three independent experiments and from multiple plates within the experiment. Whole seedlings were harvested rapidly in an excess of RNAlater solution (Ambion) and, after overnight storage at 4°C, dissected using a binocular microscope on a cooling plate with precision microscissors. Dissected leaves were transferred to a new tube, frozen in liquid nitrogen, and ground with a Retsch machine and 3-mm metal balls. RNA was extracted with Trizol (Invitrogen) and further purified with the RNeasy Mini Kit (Qiagen). DNA digestion was done on columns with RNase-free DNase I (Roche).

Genome-Wide Expression Changes
For the identification of genome-wide expression changes, samples of the strong ERF6-overexpressing line (ERF6 IOE -S) and the control line (GFP:IOE) were harvested 4 h after transfer to DEX. Two micrograms of pure RNA samples was hybridized to AGRONOMICS1 Arabidopsis Tiling Arrays (Rehrauer et al., 2010) at the VIB Microarray Facility. Obtained expression data were processed with Robust Multichip Average (background correction, normalization, and summarization) as implemented in BioConductor (Irizarry et al., 2003a(Irizarry et al., , 2003bGentleman et al., 2004). The Brainarray "agronomics1tair9genecdf" Chip Definition File was used to assign probes to genes (Brainarray). The Bio-Conductor package Limma as well as the Rank Products method were used to identify differentially expressed genes (Breitling et al., 2004;Smyth, 2004). To compare gene expression with and without ERF6 induction (ERF6 IOE -S 3 GFP: IOE), moderated Student's t test statistics were calculated using the eBayes function, and P values were corrected for multiple testing using topTable (Hochberg and Benjamini, 1990). False discovery rate-corrected P , 0.05 was used as a cutoff.

Flow Cytometry
For flow cytometry analysis, 16 leaves were chopped with a razor blade in CyStain UV Precise P buffers (Partec) according to the manufacturer's instructions. The nuclei were analyzed with a CyFlow flow cytometer with FloMax Software (Partec).

RGA:GFP Quantification
Amounts of RGA:GFP protein in either DEX-treated or nontreated ERF6 IOE -S plants were quantified by western blotting. Complete seedlings were harvested in liquid nitrogen 48 h after transfer to DEX or control medium and ground with a Retsch machine. Protein extraction was done by adding extraction buffer (Van Leene et al., 2007) to ground samples, followed by two freeze-thaw steps and two centrifugation steps (20,817g, 10 min, 4°C), whereby the supernatant was collected each time. Western-blot analysis was performed with primary rabbit anti-GFP antibodies (Santa Cruz; diluted 1:200) and secondary horseradish peroxidase-conjugated donkey anti-rabbit antibodies (GE Healthcare; diluted 1:10,000). A chemiluminescence procedure (NEN Life Science Products) was used for detection.

Long-Term Stress Exposure with IGIS
For long-term exposure to abiotic stress in combination with automated phenotypic analysis, plants were grown on the IGIS platform in the same conditions as described in "In Vitro Plant Growth Conditions." The one-halfstrength MS medium contained 9 g L 21 , and stresses were applied by adding 25 mM mannitol (osmotic stress), 50 mM NaCl (salt stress), or 1.5 mM hydrogen peroxide (oxidative stress). The platform allows for a detailed rosette growth analysis of in vitro-grown Arabidopsis plants and can hold up to 10 petri dishes. On each plate, the erf5erf6 mutant was grown next to the appropriate control (azygous for both transfer DNA constructs). Images were captured on an hourly basis, using near-infrared technology to visualize plants in the dark. Individual rosettes were extracted automatically by image analysis processing. A data analysis pipeline compiles the measurements and constructs rosette growth curves. Details about the IGIS platform will be published later.

Protoplast Activation Assay
The protoplast activation assay was performed as described previously (De Sutter et al., 2005;Pauwels et al., 2010). All transformation constructs were obtained using the Gateway cloning system, and all liquid handlings were done on the Tecan Genesis automated platform (De Sutter et al., 2005). The protoplast activation assay was performed in a 3-d-old tobacco BY-2 cell culture, subcultured from a 6-to 10-d-old culture. BY-2 cells were protoplasted using a 1% cellulase (Kyowa Chemical Products) and 0.1% pectolyase (Kyowa Chemical Products) enzyme solution in a 0.4% mannitol (Sigma) buffer. Protoplasts were then washed, counted, and diluted to 500,000 mL 21 . For every transcription factor-promoter combination, 100 mL (50,000 protoplasts) was used. To confirm direct binding of ERF5/ERF6 on the promoters of STZ, MYB51, and WRKY33, protoplasts were cotransfected with 35S-ERF5 or 35S-ERF6 (in p2GW7) and pSTZ-fLUC, pMYB51-fLUC, or pWRKY33-fLUC (in pM42GW7). Promoters were defined as the 2,000 bp upstream of the start codon. fLUC encodes the firefly luciferase enzyme. Every protoplast sample was transfected with 2 mg per construct as well as with 2 mg of normalization construct expressing the Renilla luciferase (rLUC) enzyme. Transformed protoplasts were further grown by gentle shaking overnight in BY-2 medium to allow expression of the constructs. The next day, the BY-2 medium was removed and protoplasts were lysed in Cell Culture Lysis Reagent (Promega). Protoplast content was transferred to Nunc plates (Thermo Scientific), and fLUC and rLUC activities were measured using the Dual Luciferase Assay (Promega) and the LumiStar Galaxy (De Sutter et al., 2005). Measured fLUC activities were then normalized to rLUC activities.
Microarray data from this article were deposited in the Gene Expression Omnibus database (GSE45830).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. ERF5 and ERF6 are induced by abiotic stresses.
Supplemental Figure S2. ERF5 and ERF6 are homologous genes and share highly conserved functional domains.
Supplemental Figure S3. Gene structure and expression analysis of erf5 and erf6 mutants.
Supplemental Figure S4. Gene expression of ERF5, ERF6, and their targets under osmotic, salt, and oxidative stress.
Supplemental Figure S6. Endoreduplication index in third leaves of plants overexpressing ERF6.
Supplemental Figure S7. BiNGO analysis of the putative ERF6 target genes.
Supplemental Figure S8. Expression pattern of the multiple GA2-OX genes upon ERF6 overexpression and under osmotic stress.
Supplemental Figure S9. Induction of the ERF6 targets upon ERF6 overexpression in ERF6 IOE -S and ERF6 IOE -W.
Supplemental Figure S10. Induction of GA2-OX6 in the erf5erf6 double mutant following osmotic stress exposure.
Supplemental Figure S11. ERF6 is transcriptionally induced by ERF6 itself and not by EIN3 and EIL1.
Supplemental Table S1. Measurements of erf5erf6 rosettes under different long-term abiotic stress conditions.
Supplemental Table S2. Measurements of erf5erf6 third leaves under control and osmotic stress conditions.
Supplemental Table S3. List of putative ERF6 target genes.
Supplemental Table S4. Genes induced within 3 h of osmotic stress exposure.