The Enzyme-Like Domain of Arabidopsis Nuclear b -Amylases Is Critical for DNA Sequence Recognition and Transcriptional Activation

Plant BZR1-BAM transcription factors contain a b -amylase (BAM) – like domain, characteristic of proteins involved in starch breakdown. The enzyme-derived domains appear to be noncatalytic, but they determine the function of the two Arabidopsis thaliana BZR1-BAM isoforms (BAM7 and BAM8) during transcriptional initiation. Removal or swapping of the BAM domains demonstrates that the BAM7 BAM domain restricts DNA binding and transcriptional activation, while the BAM8 BAM domain allows both activities. Furthermore, we demonstrate that BAM7 and BAM8 interact on the protein level and cooperate during transcriptional regulation. Site-directed mutagenesis of residues in the BAM domain of BAM8 shows that its function as a transcriptional activator is independent of catalysis but requires an intact substrate binding site, suggesting it may bind a ligand. Microarray experiments with plants overexpressing truncated versions lacking the BAM domain indicate that the pseudo-enzymatic domain increases selectivity for the preferred cis -regulatory element BBRE ( BZR1-BAM Responsive Element ). Side speci ﬁ city toward the G-box may allow crosstalk to other signaling networks. This work highlights the importance of the enzyme-derived domain of BZR1-BAMs, supporting their potential role as metabolic sensors.


INTRODUCTION
Plant growth is a complex interplay of cell division, cell expansion, and the assimilation of carbon and other nutrients, all of which have to be adjusted to the available resources. Therefore, growth rates need to be tightly connected to metabolic capacity, and the use of resources optimized and balanced between the production of energy, construction of cellular compounds, and the assimilation of reserves. This optimization is achieved by the integration of signals reporting the plant's metabolic status and developmental stage. Phytohormones are key transmitters for developmental programs. However, metabolites also influence plant growth by acting as signaling molecules as well as metabolic intermediates and substrates for biosynthesis.
Carbohydrates are the most abundant metabolites and are derived directly from photosynthesis. Sucrose is the most common carbohydrate transported throughout the plant and can be immediately used to fuel metabolism or provide building blocks for cellular compounds. Starch is the most widespread storage carbohydrate. In leaves, it is produced in chloroplasts during the day and broken down during dark periods (Zeeman et al., 2010). In this way, starch metabolism buffers photosynthesis to provide the plant with a continuous supply of sugar. The balance between carbon use and storage is responsive to both plant developmental stage and environmental conditions.
There are many reports that abundant sugars including sucrose, glucose, and fructose act as signals (Rolland et al., 2006;Hanson and Smeekens, 2009). They influence gene expression, protein translation, and posttranslational enzyme regulation. In all cases, our understanding of the signaling cascades is fragmentary at best. For glucose and fructose, isoforms of hexokinase and fructose 1,6-bisphosphatase have been shown to have second functions as hexose sensors, but a detailed understanding of their roles in a physiological context is still missing (Moore et al., 2003;Cho and Yoo, 2011). In sucrose signaling, the low-abundance molecule trehalose-6-phosphate has been proposed as a signaling intermediate or proxy, as their levels are closely correlated (Smeekens et al., 2010;Schluepmann et al., 2012). By reporting the plant's energy resources, the trehalose-6-phosphate signaling pathway affects developmental programs like the transition from vegetative growth to flowering (Wahl et al., 2013).
Recently, we reported the identification of a new class of putative sugar sensors called BZR1-BAMs, which belong to the family of b-amylases (BAMs), enzymes usually associated with starch catabolism. BAMs comprise a family 14 glucosyl hydrolase domain, the active site of which binds an a-1,4-linked glucan chain and catalyzes the hydrolysis of the disaccharide maltose from the nonreducing end. The genomes of vascular plants encode many proteins annotated as BAMs. The Arabidopsis thaliana genome encodes nine BAMs, not all of which are active enzymes (Fulton et al., 2008;Reinhold et al., 2011). Starch degradation in chloroplasts is mainly dependent on the catalytically active isoforms BAM1 and BAM3. Mutation of the respective genes causes excess starch to accumulate in the leaves (Fulton et al., 2008). Starch metabolism is also influenced by the noncatalytic isoform BAM4, as bam4 mutants also develop a starch-excess phenotype (Fulton et al., 2008). BAM4 is also targeted to the chloroplasts and was proposed to exert a regulatory function, modulating the rate of starch degradation, although the molecular mechanism by which it acts is not known. BAM7 and BAM8 are both BZR1-BAMs and differ from the remaining family members (Reinhold et al., 2011). Like BAM4, they have little or no catalytic activity, but their loss has no effect on starch levels. Instead of a chloroplastic target peptide, they possess an N-terminal transcription factor domain of the BZR-family type, which targets them to the nucleus and enables them to bind DNA. The fact that these proteins contain both a DNA binding domain and an enzyme-like domain led to the suggestion that they may sense a sugar ligand (Reinhold et al., 2011).
We demonstrated previously that BZR1-BAMs bind to the eight-letter DNA motif CACGTGTG (called BZR1-BAM Responsive Element [BBRE]) via their transcription factor domains. A combination of genetic and microarray analyses showed that BZR1-BAMs are transcription factors and activate the expression of genes having the BBRE in their promoters. Analysis of the BZR1-BAM-responsive genes suggested crosstalk with brassinosteroid signaling, consistent with the fact that BZRfamily transcription factors are themselves key players in plant steroid signaling events Yin et al., 2002;Kim and Wang, 2010). Furthermore, mutants and transgenic plants deregulated in BZR1-BAMs display abnormalities in leaf growth, traits typically regulated by phytohormones.
Based on our previous observations, we suggested a mechanism by which metabolism might influence plant development. The BAM domain of BZR1-BAMs could act as a sensor module allowing changes in the level of a metabolite ligand to be translated into a transcriptional output. Ligand binding could potentially affect the subcellular localization, DNA binding properties, and/or activity of these transcription factors. Although the identity of the putative ligand(s) for BZR1-BAMs remains unknown, we sought to provide more detailed information on the structure and conservation of the proteins and direct evidence for the importance of the BAM domain in their function. We demonstrate that orthologs of the Arabidopsis isoforms BAM7 and BAM8 are present across the plant kingdom, that the isoforms differ significantly in sequence and function, and that they can physically interact with each other. By creating truncated and site-directed mutant versions, we show that the enzyme-derived BAM domain is important for the function of BAM7 and BAM8 as transcription factors, influencing both DNA sequence recognition and transcriptional modulation.
Our findings are consistent with the idea that BZR1-BAMs have a conserved function in metabolic sensing.

Two Distinct BZR1-BAM Isoforms Are Conserved in Vascular Plants
The genome of Arabidopsis encodes two BZR1-BAM transcription factor isoforms, BAM7 and BAM8, which specifically bind the BBRE cis-regulatory element in vitro. However, only BAM8 was shown to substantially activate gene transcription via the BBRE motif in protoplasts and transgenic plants (Reinhold et al., 2011). We mined the published genome data of 33 vascular plants, two mosses, and six algae in Phytozome v9.1 (http://www.phytozome.net/) for the presence of BZR1-BAM isoforms. The two-domain structure of BZR1-BAMs was absent from the moss and algal genomes but was annotated in 32 of the 33 vascular plant genomes (the exception being sorghum [Sorghum bicolor]). In 26 genomes, specific BAM7 or BAM8 isoforms could be identified, which separated into distinct clades when aligned ( Figure 1A; Supplemental Table 1). This separation was still evident when only the transcription factorderived BZR1 domain or the pseudo-enzymatic BAM domain was used for the alignments (Supplemental Figures 1A and 1B Differences in the conserved amino acid sequences of BAM7 and BAM8 isoforms were found in the N-terminal part upstream of the DNA binding domains ( Figure 1B; Supplemental Data Sets 1 to 4). While BAM7 isoforms contain a highly conserved acidic region and a stretch rich in Gln and Gly residues (Q/G-rich region; Figure 1C), BAM8 isoforms carry a conserved region rich in Arg and Ala residues (R/A-rich region; Figure 1D). Furthermore, there were differences between BAM7 and BAM8 dispersed over the pseudo-enzymatic BAM domain. Interestingly, both isoforms lacked conservation of the inner loop domain, which moves upon substrate binding in active BAMs (Kang et al., 2005). BAM8 isoforms also had a low conservation of the outer flexible loop structure, which encloses the substrate in the active site during catalysis in active BAMs (Mikami et al., 1993(Mikami et al., , 1994. In the Arabidopsis proteins, both glutamates implicated in catalysis in active BAMs (Glu-186 and Glu-380;Kang et al., 2004) are conserved. However, our alignments revealed that while Glu-186 is conserved in all sequences, Glu-380 is substituted in several BAM7 and BAM8 isoforms. The presence of two BZR1-BAM isoforms and the sequence differences between them are suggestive of distinct functions for BAM7 and BAM8.

The Enzyme-Like Domains of BAM7 and BAM8 Have Distinct Functions during Transcriptional Activation
We previously demonstrated that both BAM7 and BAM8 fulllength proteins can bind to the BBRE motif, although transcription factor activity could only be demonstrated for BAM8 (Reinhold et al., 2011). We reasoned that this difference might lie in the enzyme-like domains. Therefore, we compared the transcriptional activation activity of HA-tagged, full-length BZR1-BAMs and truncated versions lacking their BAM domain in a protoplast reporter gene assay. Consistent with previous results, the full-length version of BAM8 activated the transcription of the BBREcontrolled luciferase reporter gene (Figure 2; Reinhold et al., 2011). Intriguingly, the N-terminal part of BAM8 alone (amino acids 1 to 245) activated transcription via the BBRE even more strongly than the full-length protein. In contrast, the BZR1 domain alone of BAM8 (amino acids 81 to 245) was less effective than the full-length protein. The BAM7 full-length protein could not induce reporter gene expression when compared with a transfection of control DNA. However, the N-terminal part of BAM7 (amino acids 1 to 229) activated reporter gene expression slightly, whereas its BZR1 domain alone (amino acids 65 to 229) was ineffective.
Next, we swapped the BAM domains of BAM7 and BAM8 in a highly conserved region at the start of the BAM domain. We tested in vitro DNA binding of the chimeric proteins in a DNAprotein interaction (DPI)-ELISA assay (Brand et al., 2010(Brand et al., , 2013 and in vivo transactivation activity in leaf protoplasts. Intriguingly, fusion of the BAM domain of BAM8 to the N-terminal part of BAM7 (BAM78) resulted in a protein that bound to the BBRE motif in vitro and activated reporter gene expression in protoplasts. In contrast, the chimeric construct combining the BAM domain of BAM7 with the N terminus of BAM8 (BAM87) bound only weakly to the BBRE motif and was ineffective in the reporter gene assay ( Figure 3). This suggests that the BAM domain of BAM7 reduces DNA binding of both BZR1 domains, presumably explaining the inability of BAM7 and BAM87 to activate gene expression. In contrast, the BAM domain of BAM8 allows DNA binding and transcriptional activation by the BAM7 and BAM8 N-terminal domains.
Next, we tested the subcellular localization of both artificial proteins by tagging them with yellow fluorescent protein (YFP) and expressing them transiently in tobacco (Nicotiana benthamiana) leaves (Supplemental Figure 2). Both chimeric proteins localized to the cell nucleus. However, for BAM87, the fluorescence signal was not evenly dispersed in the nucleus but rather localized to discrete foci.

BZR1-BAM Isoforms Interact during the Regulation of Gene Transcription
We were interested if the two BZR1-BAM isoforms influence each other during transcriptional activation. Therefore, the transcription factor activity of BAM7 and BAM8 alone, or the combination of both, was investigated in protoplasts using the BBRE reporter. While BAM7 did not influence the expression of the BBRE reporter, the presence of BAM8 caused transcriptional activation. However, when BAM8 was coexpressed with BAM7, the transcription of the luciferase reporter was reduced to an intermediate level between those of the single transformations ( Figure 4A). The reduction of reporter gene expression was not caused by lower expression levels of BAM8 since its protein levels remained unaffected in the presence of BAM7 ( Figure 4B).
To find out if this effect of BAM7 on BAM8 depends on physical interaction, we tested for protein-protein interaction between BAM7 and BAM8 using bimolecular fluorescence complementation (BiFC; Schütze et al., 2009). When BAM7 was tested for self-interaction by coexpression of BAM7-YFPn and BAM7-YFPc, no clear nuclear fluorescence was detected ( Figure 5). Instead, fluorescent dots in the cytosol were observed. Whether these dots represent unspecific aggregation of the overexpressed proteins or specific sequestration of BAM7 homo-oligomers remains unclear. In contrast, coexpression of BAM8-YFPn with BAM8-YFPc reconstituted the YFP fluorescence mainly in the nucleus, suggesting nuclear homo-oligomerization of BAM8 proteins. When interaction between BAM7 and BAM8 was tested, the YFP fluorescence was also mainly detected in the cell nucleus but enriched in confined subnuclear foci (see insets in Figure 5).
The Conserved Ligand Binding Site of BAM8 Is Essential for Its Ability to Activate Gene Transcription by Binding the BBRE Motif Our data suggest that the BAM domains regulate the transcription factor activity of BZR1-BAM proteins, affecting both   Table 1) were aligned and used to construct a maximum-likelihood tree with 1000 bootstrap replicates. Arabidopsis BAM1 and BAM3 were used as outliers to root the tree. (B) Protein sequences of BAM7 (31 sequences) and BAM8 (27 sequences) orthologs were aligned using ClustalW, and the degree of conservation is plotted. BAM7 sequences of M. domestica (2), Setaria italica, and Linum usitatissimum, and BAM8 sequences of Solanum tuberosum and Oryza sativa were excluded to avoid large gaps in the alignment. Conserved regions and residues of BZR-family transcription factors and active BAMs are represented in a protein model. The x axis indicates amino acid position in alignment (Supplemental Data Set 2) and should not be confused with the specific numbering of the Arabidopsis proteins.

of 18
The Plant Cell DNA binding and subcellular localization. However, Reinhold et al. (2011) showed that neither protein had significant b-amylase catalytic function and proposed that the BAM domain may serve to bind a ligand. To explore this hypothesis, we introduced mutations in the BAM domain, aiming first to eliminate any residual catalytic activity without affecting binding properties and second to disrupt the binding pocket of the BAM domain and thereby prevent ligand binding ( Figure 6A). We modeled the active site of BAM8 using the published crystal structure of the soybean b-amylase 1 enzyme (Gm-BMY1) from which the two catalytic Glu residues (Glu-186 and Glu-380) in the active site were identified (Kang et al., 2004(Kang et al., , 2005. Although the two essential Glu residues are present, the overall surface charge of the BAM8 binding cleft is less electro-negative (Supplemental Figure 3), which could partly explain the loss of catalytic activity shown previously (Reinhold et al., 2011).
To eliminate residual catalytic activity, we mutated the Glu corresponding to Glu-186 in Gm-BMY1 to a Gln (Glu422Gln for BAM7 and Glu429Gln for BAM8). The equivalent substitution in Gm-BMY1 (Glu186Gln) caused a 16,000-fold decrease in activity (Kang et al., 2005). According to our model, this amino acid exchange does not influence the structure of the binding pocket (Supplemental Figure 3). To prevent binding of putative ligands, we created another mutant variant in which the Glu corresponding to Glu-380 in Gm-BMY1 was exchanged with an Arg (Glu618Arg for BAM7 and Glu623Arg for BAM8). In addition to rendering the proteins noncatalytic (Kang et al., 2004), this significantly larger, basic amino acid should alter the size and surface charge of the binding pocket, generating nonligand binding mutant proteins.
We also created a third mutant version of both BZR1-BAMs to rule out the possibility that the BAM domain without a functional DNA binding domain affects gene expression. BZR1-BAM transcription factors contain a basic domain similar to the DNA binding residues of basic helix-loop-helix (bHLH) transcription factors (Reinhold et al., 2011). We targeted Glu-13 (position number according to Toledo-Ortiz et al., 2003), which is highly conserved throughout the bHLH transcription factor family, and even the subtle exchange of this amino acid to an Asp prevents DNA binding (Fisher and Goding, 1992). Therefore, DNA binding mutants of both BZR1-BAMs were created by exchanging the equivalent Glu (Glu83 for BAM7 and Glu100 for BAM8) to Asp. We confirmed that the introduced mutations had no influence on nuclear localization by transient expression of YFP fusion proteins in tobacco leaves (Supplemental Figure 4). (A) The chimeric versions BAM78 (BAM7 BZR1 domain fused to the BAM8 BAM domain) and BAM87 (BAM8 BZR1 domain fused to the BAM7 BAM domain) were cloned by swapping the BZR1 and BAM domains at a Tyr (Tyr-251 for BAM7 and Tyr-258 for BAM8) in a highly conserved region. (B) DNA binding activity of full-length and chimeric BZR1-BAMs was measured using DPI-ELISA assays. Therefore, proteins were recombinantly expressed with a C-terminal HIS-tag in Escherichia coli, and the bacterial cell lysates were used in the assay. Transactivation activity was measured using transactivation assays in Arabidopsis protoplasts as described in  Salmon sperm DNA was used as control DNA (CTRL). The luciferase gene (LUC) under control of the minimal 35S promoter in combination with three repetitions of either the BBRE (CACGTGTG) or a mutated version thereof (mBBRE, CACTTGTG) was used as a reporter (Reinhold et al., 2011). The b-glucuronidase (GUS) gene was used as transfection control. Error bars represent SD of three technical replicates. The experiment was repeated three times independently with similar results. One representative example is shown. Asterisks indicate a significant difference from control transfections (t test, *P # 0.05).
[See online article for color version of this figure.] b-Amylase-Like Transcription Factors 5 of 18 As previously reported (Reinhold et al., 2011), we could not detect catalytic activity for wild-type BZR1-BAMs. Unsurprisingly, the active site mutant variants were not different (Supplemental Figure 5). Next, we assessed the impact of the mutations on DNA binding and transcriptional activation ( Figure 6). Again, BAM7 full-length protein bound weakly to DNA and could not activate reporter gene expression via the BBRE motif. The same result was observed for the catalytic mutant BAM7-Glu422Gln (BAM7-cat) and the ligand binding mutant BAM7-Glu618Arg (BAM7-lig). The DNA binding mutant BAM7-Glu83Asp (BAM7dna) completely lost its affinity toward the BBRE and did not activate reporter gene expression ( Figure 6B). In contrast to BAM7, BAM8 full-length protein acted as a transcriptional activator of the reporter gene. This ability was lost by mutating the DNA binding residue Glu-100 (BAM8-dna). We could still detect marginal DNA binding, but with efficiencies well below that for the wild-type protein ( Figure 6C). The catalytic mutant BAM8-Glu429Gln (BAM8-cat) bound DNA and activated reporter gene expression at a level comparable to the wild-type protein. Intriguingly, the ability to bind DNA and activate reporter gene transcription was lost in BAM8-Glu623Arg (BAM8-lig).
These results suggest that for BAM8, mutational change of the BAM domain to disrupt the binding pocket decreases binding to the BBRE and impairs its function as a transcriptional activator. This is consistent with the idea that it may need to bind a substrate/ligand to activate gene expression. However, loss only of catalysis does not impair its function. Owing to the lack of detectable transcription factor activity, it is unclear how the mutations affect the BAM7 protein.

In Vitro Transcription Factor Functionality of BZR1-BAMs Correlates with Abnormalities in Shoot Growth
To study the functionality of the mutant BZR1-BAM proteins in vivo, the mutant and truncated versions of the BAM7 and BAM8 proteins were constitutively overexpressed as C-terminal YFP fusion proteins in the respective mutant background. For each mutant or truncated variant, several independent transgenic lines were created and analyzed for phenotypic alterations.
When grown on soil, transgenic plants overexpressing the BAM7 full-length proteins BAM7, BAM7-cat, and BAM7-lig showed no apparent phenotypic differences from the wild type ( Figure 7A). In contrast, overexpression of BAM7-dna mutant protein caused a reduction in shoot fresh weight and development of epinastic leaves when compared with the wild type (Supplemental Figure 6). Plants overexpressing only the N-terminal part of BAM7 protein (BAM7-N-OX) developed smaller, compact rosettes. This phenotype was comparable to that of plants overexpressing only the N-terminal part of BAM8 (BAM8-N-OX) ( Figure 7A). Interestingly, plants overexpressing BAM8-cat protein resembled plants overexpressing the wild-type BAM8 protein (BAM8-OX). Both BAM8-OX and BAM8-cat-OX plants developed smaller rosettes with rounded, hyponastic leaves ( Figure 7B). However, plants overexpressing the ligand binding mutant protein BAM8-lig were indistinguishable from the wild type.
The phenotypes of the transgenic plants demonstrate that mutant BAM8 proteins without the ability to bind DNA and/or activate target gene transcription also do not cause developmental phenotypes. In contrast, of the mutant BAM7 proteins, the DNA binding mutant did impart a subtle growth phenotype similar to the bam7 bam8 double mutant.

Developmental Effects of BZR1-BAM Overexpression Is Reflected at the Transcriptome Level
To connect developmental abnormalities of plants overexpressing functional BZR1-BAM versions to specific changes in gene expression, we analyzed global transcript levels by microarray experiments for all 14 genotypes (the wild type, bam7 and bam8 single mutants, the bam7 bam8 double mutant, and lines overexpressing the wild-type, truncated, and mutated forms of the BAM7 and BAM8 proteins). The expression level of BAM7 and BAM8 wild-type and mutated versions were comparable between different overexpression lines ( Figure 7C), and the proteins were all readily detectable ( Figure 7D).
Across all genotypes, we found 2097 genes (see Figure 8) that changed more than 1.5-fold compared with the wild type (with a false discovery rate of <5%). First, we compared our data to (A) Transcription factor activity of full-length BZR1-BAMs was measured using Arabidopsis protoplast transactivation assays as described in Figure 2. For cotransformation, plasmids containing the coding sequence of BAM7 or BAM8 with an N-terminal HA tag were combined at equal molar ratios. The experiment was repeated three times independently with similar results. One representative example is shown. Error bars represent SD of three technical replicates. Asterisks indicate a significant difference from control transfections (t test, *P # 0.05). (B) Transient expression of BZR1-BAM proteins by protoplasts in (A) was confirmed by protein gel blotting using polyclonal antibodies raised against BAM7 (top panel) and BAM8 (bottom panel). Protein size is indicated in kilodaltons.

of 18
The Plant Cell transcript profiles of BAM8-OX and bam7 bam8, which were generated previously in our laboratory (Reinhold et al., 2011). The developmental stage of the plant material used and the growth protocol differed compared with the previous study (3week-old rosettes versus 14-d-old seedlings). Approximately 20% of the genes deregulated compared with the wild type in our experiments were also changed in the previous experiment. The majority changed into the same direction (i.e., up-or downregulated). Importantly, the trend for genes carrying the BBRE motif in their promoters ( Figures 8A and 8B) was the same in both experiments (i.e., these genes were upregulated in the BAM8-OX and downregulated in the bam7 bam8 double mutant). Next, we compared the transcript profiles of BZR1-BAM transgenic lines and knockout mutants in a principal component analysis. The first three components covered more than 75% of the whole variance in the data set ( Figure 8C). The transgenic lines BAM7-N-OX and BAM8-N-OX showed the largest transcriptional changes (731 and 732 differentially expressed genes, respectively) and clustered together, separately from the other genotypes. Genes that changed both in BAM7-N-OX and BAM7 and BAM8 proteins fused to the complementary halves of YFP were transiently coexpressed in leaves of N. benthamiana by infiltration of Agrobacterium tumefaciens. The 1st and the 2nd proteins were fused to C-terminal (YFPc) and N-terminal part (YFPn) of YFP, respectively. Agrobacterium cells transformed with a binary p35S:p19 construct were coinfiltrated to suppress gene silencing (Voinnet et al., 2003). Arrowheads label cytosolic speckles formed by BAM7. Insets show magnified cytosolic speckles of BAM7 or nuclei containing BAM8 or BAM7/BAM8. Green, BiFC fluorescence; red, chlorophyll autofluorescence. Bar = 10 mm. b-Amylase-Like Transcription Factors 7 of 18 BAM8-N-OX were mainly regulated into the same direction, and the BBRE motif was found in the promoters of both upand downregulated genes ( Figure 8D). The other genotypes (overexpressing full-length versions of BAM7 and BAM8 or deficient in them) arranged in the principal component analysis on an axis with the bam7 bam8 double mutant and the BAM8-OX at the extremities.
The lines overexpressing the active site mutants of BAM7 (BAM7-cat and BAM7-lig), which behaved similarly to the wildtype BAM7 protein in DNA binding assays ( Figure 6B) clustered close to BAM7-OX. The overexpression of the DNA binding mutant BAM7-dna caused a transcript profile most similar to that of the bam7 bam8 double mutant. In this case, the underlying transcriptional changes were partly dependent on the BBRE motif, which was mainly present in the group of repressed genes ( Figure 8E). This is interesting, as it suggests that the BAM7-dna mutant protein somehow inhibits the function of the endogenous BAM8 protein.
Of the overexpressed BAM8 active-site mutants (BAM8-cat and BAM8-lig), only BAM8-cat caused transcriptional changes similar to those generated by overexpression of the wild-type BAM8 protein. Genes that changed in both BAM8-OX and BAM8-cat-OX were deregulated almost exclusively the same direction, and the BBRE motif was only present in the promoters of induced genes, suggesting that BAM8-cat acts similarly to wild-type BAM8 as an activator of gene transcription via the BBRE motif ( Figure 8F). In contrast, the BAM8 mutant incapable of binding a putative ligand (BAM8-lig) did not cause comparable transcriptional changes ( Figure 8G). In fact, BAM8-lig clustered together with the BAM8 DNA binding mutant (BAM8dna) and the bam8 single mutant, suggesting that these two mutant versions are incapable of substituting for the endogenous BAM8 protein.

The Presence of the BAM Domain Causes Distinct Transcriptional Changes
We clustered the genes that were differentially regulated in the mutants to obtain groups that were influenced by the presence and function of the BAM domain ( Figure 9A). This resulted in 17 groups containing between one and 44 genes. Many of the genes that changed in BAM7-N-OX and/or BAM8-N-OX remained largely unaffected in the BAM7-OX and/or BAM8-OX (Groups 1, 2, 4, 5, 12, 16, and 17) and can therefore be considered as "BZR1 domain dependent" genes, although there were differences between BAM7-N-OX and BAM8-N-OX in the expression pattern of some of these groups (Groups 2, 4, 12, and 17). This is interesting, as it suggests these truncated forms are not functionally equivalent. Furthermore, a set of genes changed simultaneously in BAM7-N-OX/BAM8-N-OX and BAM8-OX (Groups 3 and 7) and was assigned to the category of "BZR1/BAM domain dependent" genes. In addition, we identified a distinct gene set, which was predominantly induced in BAM8-OX but not in one of the N termini overexpressing plants (Group 6). This category of genes was considered as "BAM domain dependent" (Supplemental Table 2).
We analyzed the occurrence of the BBRE motif in the 21500bp promoter region of the different gene groups and used the mBBRE sequence as control ( Figure 9B). Interestingly, the BBRE motif was only overrepresented in the categories "BZR1/BAM domain dependent" and 'BAM domain dependent', but not in the category "BZR1 domain dependent." The mBBRE was not enriched in any group. Within the categories "BZR1/BAM domain dependent" and "BAM domain dependent," there were many known proteins involved in transcriptional regulation (e.g., SWINGER [SWN], HB-4, Figure 6. Putative Ligand Binding Activity of BAM8 Is Required for Its Function as a Transcriptional Activator. (A) Scheme of BZR1-BAM proteins showing position and predicted impact of mutagenized residues. Site-directed mutagenesis was used to target residues involved in DNA binding (Glu-83 in BAM7/Glu-100 in BAM8), catalytic function (Glu-422 in BAM7/Glu-429 in BAM8), and ligand binding (Glu-618 in BAM7/Glu-623 in BAM8). (B) DNA binding and transactivation activity of BAM7 mutant versions (B7-FL, BAM7 full-length; B7-dna, BAM7 DNA binding mutant; B7-cat, BAM7 catalytic mutant; B7-lig, BAM7 ligand binding mutant) were measured using DPI-ELISA and Arabidopsis protoplast transactivation assays, respectively, as described in Figures 2 and 3. Error bars represent SD of three technical replicates. Asterisks indicate a significant difference from control transfections (t test, *P # 0.05).

of 18
The Plant Cell ANAC102, ESE3, MYB56, and MBD11), but also cell cycle modulators (e.g., CYCD1;1, CYCP3;1, and LCD1). Of the genes of these two categories, 30% carry at least one BBRE motif in their promoter region (Supplemental Table 2). The microarray analyses are broadly consistent with the findings from DNA binding and reporter gene assays and from the growth phenotypic analysis. Furthermore, the data suggest that the presence of the BAM domain specifies the group of target genes and the strength/direction of BBRE gene deregulation.

The Presence of the BAM Domain Increases Specificity of DNA Sequence Recognition and Regulates the Expression of Nuclear and Plastidial Proteins
The differences in the transcriptomes of plants overexpressing the full-length BAM8 protein, compared with the truncated N termini, led us to investigate whether the BAM domain influences DNA sequence specificity in vivo. Therefore, we compared the presence of the BBRE (CACGTGTG) and the G-box (CACGTG; a subsequence of the BBRE) in the promoter regions of the deregulated genes ( Figure 10; Supplemental Table 3). As controls, we tested the occurrence of the mBBRE (CACTTGTG) and a mutated version of the G-box (CACTTG), neither of which was overrepresented in any of the gene sets ( Figures 10B and  10D). We could not detect enrichment for the BBRE or the G-box in the set of genes deregulated in transgenics overexpressing BAM7 (BAM7-OX) or in the single mutants bam7 and bam8 ( Figures 10A and 10C). However, overexpression of BAM8 caused deregulation of many genes containing these motifs in their promoters. Both motifs were only enriched in promoters of genes induced in BAM8-OX. The opposite was true for genes deregulated in the bam7 bam8 double mutant, where repressed genes containing the motifs in their promoters were enriched. These data confirm previous observations that BAM8 acts as an activator of gene expression via the BBRE motif (Reinhold et al., 2011). In contrast, overexpression of the N termini alone appeared to cause more changes independent of the BBRE. The degree of enrichment of the BBRE in the promoters of genes was decreased compared with BAM8 overexpression, although an enrichment of the G-box was still observed.
These differences in gene activation by the full-length and truncated forms of the BZR1-BAM proteins led us to investigate the sequence specificity of the core BZR1 domain. Therefore, we compared the binding specificity toward the BBRE, the G-box, and mutated versions thereof in vitro. The BBRE motif was bound by the BZR1 domains of BAM7 and BAM8 with b-Amylase-Like Transcription Factors 9 of 18 comparable affinity ( Figure 11A). In addition, the BZR1 domain of BAM7 also interacted with the G-box, with binding signal half as strong as that for the BBRE. It also interacted weakly with the mutated motifs mBBRE and mG-box. In contrast, BAM8 bound only weakly to the G-box, and no interaction was detected with the mutated motifs. We also tested the motif specificity by competing the binding to the biotinylated BBRE with nonbiotinylated probes of the BBRE, the G-box, and their mutated variants as controls. The interaction of both BZR1 domains with the BBRE probe could be prevented with increasing concentrations of nonbiotinylated BBRE DNA ( Figures 11B and 11C), but not with any of the other competitor probes. Thus, these data confirm that the BZR1 domains of BZR1-BAM proteins are flexible in sequence recognition and can recognize the G-box as well as the original BBRE motif, although not with comparable affinity. An unbiased search for motifs in the promoters of the deregulated genes reaffirmed this picture. The most highly represented eight-letter motif in the promoters of genes downregulated in bam7 bam8 plants and in genes upregulated in BAM8-OX plants was the BBRE (CACGTGTG, with E-values of 1.2 3 10 211 and 4.5 3 10 216 , respectively). However, the most highly represented motifs in the transgenics BAM7-N-OX (ACACGTGT) and BAM8-N-OX (CACGTGTA) differed, and their enrichment was far less significant (E-values of 1.5 3 10 21 and 2.4 3 10 23 , respectively). Interestingly, all three motifs contain the G-box at their core.
This difference in motif specificity could explain the phenotypic differences between plants overexpressing full-length BZR1-BAMs and the N-terminal truncations. Although both BZR1 domains are able to bind the BBRE in vitro, they appear to target a broader range of G-box-based DNA motifs, resulting in large transcriptional changes independent of the BBRE motif. Therefore, the BAM domain not only affects DNA binding per se, but also DNA sequence specificity of the BZR1 domain in the protein's N terminus.
Most functional cis-regulatory elements are present in a 100to 1000-bp region upstream of the transcriptional start site of target genes (Berendzen et al., 2006;Biggin, 2011). When we analyzed the genome-wide distribution of the BBRE motif, we (A) Scatterplot of log 2 (fold change) values (log 2 FC) of 82 genes that changed more than 1.5-fold in BAM8-overexpressing (BAM8-OX) plants compared with the wild type both at the seedling and rosette stages. Transcript data from seedlings are derived from Reinhold et al. (2011). Crossed dots report the presence of the BBRE cis-element (CACGTGTG) in the 21500-bp promoter region of the respective gene. Linear trend lines with R 2 values are indicated. (B) Scatterplot of log 2 (fold change) of 34 genes that changed more than 1.5-fold compared with the wild type in bam7 bam8 plants both at the seedling and rosette stages. (C) Principal component analysis of 2097 genes whose expression changed more than 1.5-fold compared with the wild type in at least one of 14 genotypes analyzed (the wild type, the bam7 and bam8 single mutants, the bam7 bam8 double mutant, and the lines overexpressing [OX] the mutant protein versions described in Figure 6). The coverage of the variance in the data set by the principal components 1-3 (PC1-PC3) is given as percentages.
(D) to (G) Scatterplots of log 2 (fold change) of selected genotypes compared with the wild type. Labeling is as in (A).

of 18
The Plant Cell identified 653 genes carrying at least one motif in their 1500-bp upstream region (referred to as "BBRE genes"). The highest frequency of the BBRE in this selection of promoters was in the region 200 bp upstream of the transcriptional start site ( Figure  12A). Interestingly, functional classification of BBRE genes showed enrichment of genes encoding nuclear and chloroplastic proteins ( Figure 12B; Supplemental Table 4). While the group of BBRE genes encoding nuclear proteins unsurprisingly had assigned functions in nucleic acid metabolism ("DNA or RNA binding" and "transcription factor activity"), the group of BBRE genes encoding chloroplastic proteins did not cover a common pathway or process (Supplemental Figure 7). However, we checked the expression pattern of these genes in our microarray data and found 18.8 and 19.4% (of the BBRE genes encoding nuclear and chloroplastic proteins, respectively) deregulated compared with the wild type in at least one of the 13 genotypes (Supplemental Table 5). Clustering of the expression Figure 9. The Presence of the BAM Domain Causes Distinct Transcriptional Changes.
(A) Log 2 (fold change) values (log 2 FC) of genes (n = 278) whose expression changed more than 2-fold compared with the wild type in at least one of the mutant or overexpressing (OX) genotypes were clustered into a heat map. Genes clustered into 17 groups. The number of genes in each group is given in parenthesis.
(B) The occurrence of the cis-element BBRE (CACGTGTG) and the control sequence mBBRE (CACTTGTG) in the 21500-bp region of genes belonging to different gene groups was analyzed. The significance of overrepresentation is given as the negative logarithm of the P value, determined by hypergeometric distribution.  Table 5). These regulatory proteins could contribute to the developmental abnormalities that are triggered by BZR1-BAM mutation or overexpression/ectopic expression.

DISCUSSION
There is good evidence that metabolic signals are integrated to influence developmental programs in plants. Here, we provide direct evidence for the importance of the BAM domain of BZR1-BAM transcription factors during DNA binding and transcriptional activation, consistent with the proposed role of these proteins as metabolite sensors. However, despite their sequence similarity, the two BZR1-BAM proteins in Arabidopsis seem to differ in their exact functions. Our data show that BAM7 binds DNA with low affinity without activating gene transcription, while BAM8 binds with a higher affinity and mediates substantial activation of gene expression, an ability lost if the DNA binding sequence of the BZR1 domain is mutated.

Functional Specialization of BZR1-BAMs
Extensive database searches revealed that distinct BAM7 and BAM8 orthologs are conserved in a range of other vascular plants, which is suggestive of functional specialization of the two isoforms. The sequence differences are not restricted to one of the conserved domains. Therefore, differences in multiple characteristics of the proteins might account for their contrasting functionality in our experiments. We demonstrate that the complete N termini of BAM7 and BAM8 activate gene expression more strongly than the BZR1 domains alone, suggesting that there are intrinsic activation domains N-terminal of the BZR1 domain. However, the differences in transactivation activity between the complete N terminus and BZR1 domain were less obvious for  (A) The motif distribution of the BBRE (CACGTGTG) on the sense and antisense strand was determined on automatically assembled promoter sequence data sets of the Arabidopsis genome. The number of motifs per site is displayed as total number (blue) and moving average trend line (black). The distance from the transcriptional start site is given on the x axis.
(B) Functional categorization of genes containing the BBRE motif in the 21500-bp region of their transcriptional start site. GO cellular component prediction for the selected set of genes is presented. The significance of overrepresented gene categories was determined by hypergeometric distribution (*P # 5$10 25 ).

of 18
The Plant Cell BAM7 than for BAM8. Intriguingly, only the BAM domain of BAM8 allows substantial transcriptional activation by a full-length BZR1-BAM protein (i.e., the native BAM8 and the BAM78 chimeric protein), while the BAM domain of BAM7 significantly reduces transactivation. We present evidence that BAM7 and BAM8 proteins interact to form homomultimeric and heteromultimeric complexes. Reporter gene assays in protoplasts showed that transactivation activity was reduced when BAM7 was coexpressed with BAM8. This could mean that heteromeric complexes form between the two proteins and that these have reduced or no transactivation activity when compared with BAM8 homomers. According to this model, the two BZR1-BAM isoforms might antagonistically fine-tune each other to cooperatively regulate the level of gene transcription. It is interesting that when BAM7 and BAM8 interact, it occurs in, or causes the formation of, discrete nuclear foci and that different patterns are seen for the homomeric forms. The nuclear foci are conspicuously similar to the localization of the inactive chimeric BAM87 protein. This subnuclear sequestration correlates with the presence of the BAM domain of BAM7 and might reflect a mechanism by which it modulates the activity of BAM8.
The formation of different complexes of transcription factors, as proposed here, may enable transcriptional flexibility and is not without precedent. Some bHLH transcription factors interact with non-DNA binding homologs via their HLH domains and are thereby prevented from regulating gene expression (Hornitschek et al., 2009;Zhang et al., 2009). Interestingly, the N termini of BZR1-BAMs (and of BZR-family transcription factors in general) have sequence similarity to bHLH proteins (Yin et al., 2005;Reinhold et al., 2011). One possible explanation for the observed subnuclear localization and transcription factor activity of the BAM7-BAM8 complexes might be that BAM7 sequesters BAM8 away from promoters and ultimately causes reduced expression levels of BAM8 target genes. However, if BAM7 simply acts as a repressor of BAM8, why does the loss of BAM7 in bam7 bam8 double mutants cause a much stronger transcriptional phenotype than the bam8 mutation? Only the simultaneous loss of both BAM7 and BAM8 results in the deactivation (or repression) of genes dependent on the BBRE motif, which is more indicative of a redundant function of the two isoforms. However, such redundancy conflicts with our findings that BAM7 cannot activate gene transcription alone and even inhibits BAM8 in protoplast reporter assays. Our data do not allow us to resolve this issue, but it is plausible that BZR1-BAMs compete with additional transcription factors for the same promoter motifs. In both of the single mutants, the other BZR1-BAM can bind the BBRE motif and either activate gene expression or prevent binding of the BBRE motif by other factors, particularly transcriptional repressors, or both. Further work will be required to discriminate between the possible roles of BZR1-BAMs as transcriptional activators themselves or as "derepressors" that function by outcompeting repressors on the promoters of target genes, allowing activation by other as yet unidentified factors. According to this hypothesis, repressors could bind the BBRE motif only in the bam7 bam8 double mutant. The BAM7-dna-OX transgenics provide indirect support for this idea, as these plants have a bam7 bam8-like growth and transcriptional phenotype. This would not be expected if the only role of BAM7 was to sequester BAM8 away from the promoters of the target genes.
It is also possible that BZR1-BAMs may be dependent on other interacting proteins. This might also account for some of our apparently contradictory observations. For example, if BAM7 were dependent on a protein that was absent or limiting in our in vitro and in vivo studies (where BAM7 was overexpressed), it could explain the lack of obvious transcription factor activity. Even if this is the case, our data remain valuable as they establish that BAM8 is not dependent in the same way.

The Influence of the BAM Domain in DNA Binding
Our results not only demonstrate a role of the BAM domain during DNA binding, but also suggest an influence on sequence specificity of the binding. While the full-length versions of BZR1-BAMs preferentially bind the BBRE, the truncated versions lacking the BAM domains show a broader range of target sequences, including the G-box. This can also be seen in the transcriptional changes caused by overexpression. While the genes deregulated by the full-length proteins were highly enriched for the BBRE motif, the N-terminal domains deregulated more genes, many of which had the G-box in their promoters. This might explain the fact that, although the BAM7 N terminus only slightly induces BBRE-dependent reporter gene expression in protoplasts, it causes major transcriptional changes in planta. The resulting phenotypic alterations are similar to those caused by the N terminus of BAM8. Expression of either of the N termini alone causes mostly deregulation of genes that do not contain the BBRE in their promoters. We propose that the removal of the BAM domain causes deregulation of the DNA binding and/or activation domains, resulting in a less specific function.
The genome-wide distribution of the BBRE peaks at 2200 bp upstream of the transcriptional start site of putative target genes. This places the BBRE as a genuine functional cis-regulatory element that regulates a distinct group of genes. Interestingly, the BBRE motif is enriched in promoter regions of genes encoding for nuclear proteins, many of which are themselves regulatory in nature. We therefore propose that BZR1-BAMs regulate the expression of a small number of proteins that in turn trigger a broader secondary response. Among the deregulated BBRE genes were the transcription factors SWN and ANAC102, among others. SWN is a Polycomb-group protein involved in chromatin remodeling and the determination of cell fate (Chanvivattana et al., 2004;Lafos et al., 2011), but was also proposed to regulate growth (Spillane et al., 2007). ANAC102 was shown to be a transcription factor involved in stress responses during seed germination (Christianson et al., 2009). In addition, the BZR1-BAM-responsive gene set contains factors involved in cell division including LCD1 (allelic to RETICULATA), which influences cell density in the palisade parenchyma and is important for normal leaf development (Barth and Conklin, 2003;González-Bayón et al., 2006), and the cyclin p3;1 (CYCP3;1). Interestingly, CYCP3;1 belongs to the PHO80-like proteins (Torres Acosta et al., 2004). The PHO80 protein from yeast is involved in phosphate signaling, but also glycogen metabolism and carbon source utilization. A role in nutritional signaling was also proposed for PHO80-like proteins of plants (Torres Acosta et al., 2004).

b-Amylase-Like Transcription Factors
Site-directed mutagenesis of the BAM domain further illustrates its importance for regulation of gene transcription. Although enzymatic assays, together with our sequence analysis, suggest that BZR1-BAMs are noncatalytic enzyme-derived proteins, we excluded an effect of possible residual activity by creating catalytic mutants of the Arabidopsis proteins. In vitro DNA binding assays and reporter gene activation assays in protoplasts showed that the catalytic mutants behave like the wild-type proteins. Furthermore, the wild-type and catalytic mutant BAM8 proteins caused similar alterations of rosette morphology and transcriptome upon overexpression in vivo. Despite the apparent insensitivity to the loss of catalytic activity, the substrate/ligand binding ability of BZR1-BAMs seems to be of fundamental importance. For BAM8, the putative ligand binding mutant completely lost DNA binding affinity toward the BBRE motif. Thus, BAM8 could not act as a transcriptional activator anymore, neither in protoplasts nor transgenic plants. These data suggest that interaction of a ligand with the substrate binding domain may be necessary for BAM8 to function as a transcriptional activator. When unable to bind this putative ligand, or when the ligand is absent, BAM8 would be inactive and unable to induce expression of the target BBRE genes.

The Mechanism and Biological Context of BZR1-BAM Function
We do not yet fully understand the molecular mechanism by which BZR1-BAMs function, specifically how the BAM domain mediates its effect on the BZR1 domain. A direct intramolecular interaction between the BAM and BZR1 domains to regulate DNA binding could be envisaged, as could intermolecular interactions between adjacent domains of interacting proteins. Presumably, when DNA-bound, the protein also functions as a docking station to help assemble the other factors needed during transcriptional activation. Such domain interactions or interactions with other proteins might themselves depend on other factors, for example, the binding of the putative ligand by the BAM domain or posttranslation modifications, such as reversible protein phosphorylation.
Previous findings showed that BZR1-BAMs regulate a significant number of genes that are also regulated by brassinosteroids (Reinhold et al., 2011). Although BZR1-BAMs prefer the BBRE motif to all other tested motifs, we demonstrated binding of, and gene deregulation via, the shorter G-box motif. This side specificity toward the G-box could partly explain the crosstalk with brassinosteroid signaling pathways. Recently, it has been reported that BZR-family transcription factors, interacting with phytochrome interacting factors (PIFs), cooperatively bind the G-box and synergistically regulate expression (Oh et al., 2012). Since the BBRE contains the G-box, these regulators of hormone and light signaling pathways could also affect the expression of BZR1-BAM target genes. Additional work will reveal if BZR1-BAMs interact directly with transcription factors involved in brassinosteroid signaling or if they compete for the promoter regions of common target genes. Reinhold et al. (2011) presented evidence that BZR1-BAMs counteract brassinosteroid signaling, so competition at the level of DNA binding may well occur. It seems plausible that BZR-family proteins, described as potent transcriptional repressors, inhibit BBRE gene expression when both BZR1-BAMs are missing (i.e., in the bam7 bam8 double mutant and the BAM7-dna-OX lines, as hypothesized above).
Assuming the existence of a BZR1-BAM ligand, it would promote BBRE gene expression through the activation of BAM8. Since BAM8 overexpression induces BR-repressed genes (Reinhold et al., 2011), the putative ligand might initially be viewed as a growth-retarding signal. At first, this seems counterintuitive since BZR1-BAMs are proposed to act as sensors of cellular carbohydrates, high levels of which might be expected to promote growth. However, it should be noted that BZR1-BAMs appear to affect the fine-tuning of plant architecture rather than growth rates per se. The primary impact of their deregulation is on plant form through differential growth of specific parts of the rosette (i.e., the petiole length or the leaf lamina). BAM8-OX plants have compact rosettes and the upwardly curled leaves, while bam7 bam8 mutants have elongated petioles and downwardly curled leaves. Interestingly, both BAM8-OX and bam7 bam8 plants are smaller than the wild type. This could mean that neither has a leaf architecture optimal for the growth conditions used.
Our observations inevitably leave open questions about BZR1-BAM function, including the possible biological meaning of the putative ligand signal. It is possible that under conditions when carbon assimilation and sugar levels are sufficiently high, ligand levels would also be high, and signaling through BZR1-BAMs might promote an economical plant stature by preventing further extension growth toward the light. In contrast, when sugar (and ligand) levels are low, growth would be altered to increase petiole length to enhance light capture. In this way, BZR1-BAM signaling could be viewed as causing the same types of growth responses usually associated with light signaling pathways. Low light levels, or a low red to far-red light ratio, promote petiole elongation, the so called shade avoidance and neighbor detection responses. These responses are believed to help the plant compete for light to enable photosynthesis (Casal, 2013). Consequently, it is plausible that such a response is moderated, or indeed accentuated, by metabolic signals (carbon sufficiency and starvation, respectively). This would also explain the similarities of the BBRE target motif with that recognized by transcription factors associated with light responses such as the PIF proteins (Leivar and Quail, 2011). However, it has to be noted that, at present, the existence of such a ligand for metabolic signaling through BZR1-BAMs is not certain, despite our results. Therefore, the ideas presented here remain speculative. Future attempts should aim to understand the regulation of BZR1-BAMs, including the isolation and identification of any ligand(s) they bind.

Plant Material
Arabidopsis thaliana (accession Columbia) was grown in a nutrient-rich, medium-grade, peat-based compost in Percival AR95 growth chambers (CLF Plant Climatics) at a constant temperature of 20°C, 70% humidity, and a 12-h light period with constant, uniform light intensity of 150 µmol m 22 s 21 . Genotyping of the Tilling line of BAM7 (bam7-1) and the T-DNA insertion line of BAM8 (bam8-1) was described previously (Reinhold et al., 2011).

of 18
The Plant Cell

Molecular Cloning
Standard molecular techniques were performed as described (Sambrook and Russel, 2001). Site-directed mutagenesis and domain swapping were conducted by a Gateway technology (Invitrogen) based method described before (Atanassov et al., 2009). Primers used are listed in Supplemental Table 6. PCR products with Gateway recombination sites were obtained using pDONR221 clones as template and m13_fwd primer (59-GTAAAACGACGGCCAGT-39) with gene-specific reverse primers and m13_rev (59-GGAAACAGCTATGACCATG-39) with gene-specific forward primer. Resulting PCR products were fused in an overlap extension reaction and recombined into pEarleyGate201 (Earley et al., 2006) for Nterminal HA fusion or in pB7YWG2.0 (Karimi et al., 2002) for C-terminal YFP fusion and transformed into chemo-competent DH5l cells. For HIStagged protein versions, the coding sequences were amplified using primers with NheI and NotI restriction sites to clone the resulting PCR product into the pET21a(+) vector (Novagen, Merck).

Stable Transformation of Arabidopsis
Arabidopsis plants were stably transformed with Agrobacterium tumefaciens (GV3101) as described previously (Clough and Bent, 1998). Wildtype and mutated versions of BZR1-BAMs were expressed under control of the cauliflower mosaic virus 35S promoter as C-terminal YFP fusion proteins in the corresponding single knockout background (i.e., BAM8-OX in the bam8-1 mutant, BAM7-OX in the bam7-1 mutant, etc.). Selection of transformed lines was enabled by BASTA resistance encoded on the pB7YWG2.0 vector. Per mutated version, between 14 and 28 primary transformants (T1) were selected and analyzed for the levels of overexpressed proteins by immunoblotting using monoclonal anti-YFP antibodies (1:5000 dilution, JL-8; Clontech Laboratories). At least two independent lines were progressed to the homozygous state. Different T3 seeds were screened on BASTA containing plates to ensure homozygosity of T2 ancestry.

Transient Transformation of Nicotiana benthamiana
For subcellular localization, proteins were transiently expressed as Cterminal YFP fusions in epidermal cells of tobacco leaves. Agrobacterium (GV3101) cells were transformed with pB7YWG2.0 plasmids harboring the gene of interest. Single colonies were grown in liquid cultures under selective conditions (rifampicin, gentamycin, and kanamycin) at 28°C for 48 h. Cells were spun down at 2500g for 15 min at 20°C and resuspended in 50 mM MES, pH 5.7, and 10 mM MgCl 2 to an OD 600 of 0.8 to 1. The silencing suppressor p19 (Voinnet et al., 2003) was constitutively coexpressed for BiFC assays (Schütze, Harter and Chaban, 2009). In brief, Agrobacterium cultures transformed with plasmids harboring the coding sequence for the YFPn fusion, the YFPc fusion, and the p19 silencing suppressor were mixed at equal ratio and infiltrated into the lower epidermis of N. benthamiana leaves. Fluorescence was imaged 3 to 4 d after infiltration using a Zeiss LSM780 confocal microscope.

Production of Recombinant Protein
For production of recombinant expressed proteins, HIS-tagged versions were overexpressed in Escherichia coli BL21 (DE) (Stratagene) using the pET21a(+) vector (Novagene). Protein production was induced by 1 mM isopropyl b-D-1-thiogalactopyranoside and performed at 16°C for 20 h. Cells were harvested by centrifugation at 5000g, at 4°C for 15 min, and resuspended in lysis buffer (50 mM HEPES, pH 7.5, 0.5 M NaCl, 10% [v/v] glycerol, and 30 mM imidazole) and broken using a M-110P microfluidizer processor (Microfluidics). Protein expression was confirmed and quantified by SDS-PAGE and immunoblotting (Supplemental Figure 7) using anti-HIS antibodies (Qiagen). For purification, the protein was bound to a nickel-loaded HiTrap chelating column (GE Healthcare). The imidazole concentration of the buffer was raised to 80 mM for washing and to 250 mM for elution of the protein. Buffer exchange to 50 mM HEPES, pH 7.5, 150 mM NaCl, 0.5 mM EDTA, and 10% (v/v) glycerol was performed using NAP 25 columns (GE Healthcare). Protein was stored in aliquots at 280°C.

Modeling of At-BAM8 Protein Structure
MODELER (Eswar et al., 2007) was used to model the structure of the BAM8 b-amylase-like domain using the structure of Gm-BMY1 as a template (1Q6C.pdb). The sequences of Gm-BMY1 and At-BAM8 b-amylase domains share 41% sequence identity. Five models were built, and the structure with the lowest DOPE score was chosen for further analysis. No restraint violations were reported near the active site of the models. Least squared superpositions of the template and model structures, and figures, were made using PyMol (DeLano, 2002). For the Glu429Gln mutation, the structure of Gm-BMY1 with equivalent mutation (Glu186Gln) was used as a template (1V3H.pdb). 1V3H.pdb was solved with maltopentaose bound in the active site. For the Glu623Arg mutation, Gm-BMY1 in the apo conformation (1BYA.pdb) was used as a template since this is the structure that an incoming substrate/ligand would encounter.

Detection of Amylolytic Activity
b-Amylase activity of BZR1-BAM wild-type and mutant proteins were measured using the Betamyl Assay kit using the manufacturer's instructions (Megazyme). Per reaction, 5 µg of purified BZR1-BAM protein or 5 µg total protein of BZR1-BAM expressing E. coli BL21 was assayed. As control, 0.1 µg recombinant BAM3 or 5 µg total protein from BAM3 expressing cells was used. Reactions were done in triplicate and stopped after 15, 30, and 45 min. The colorimetric reaction was quantified using an Infinite M1000 microtiter plate reader (Tecan Trading).

DPI-ELISA
DNA binding properties of BZR1-BAM wild-type and mutated proteins were determined using the DPI-ELISA method as described before (Brand et al., 2010(Brand et al., , 2013. The amount of total protein input was normalized to the level of heterologous expressed protein (Supplemental Figure 9). For direct binding assays, 2 pmol biotinylated oligonucleotides and 30 µg total protein extract was used. For competition experiments, 30 µg total protein extract were incubated with 0, 2, 10, and 50 pmol nonbiotinylated oligonucleotides before 2 pmol biotinylated oligos were added.

Transactivation Assays
Transactivation assays in Arabidopsis mesophyll protoplasts were performed as described earlier (Yoo et al., 2007;Reinhold et al., 2011) with slight modifications. A pUC18 vector carrying the LUC gene under control of three repetitions of the BBRE (CACGTGTG) or the mBBRE (CACTTGTG) upstream of the minimal 35S promoter was used as the reporter construct (Reinhold et al., 2011). As effector plasmids, pEar-leyGate201 constructs containing the genes of interest were used. A Ubq10:GUS:nosT construct (Yoo et al., 2007) served as transfection control and salmon sperm DNA (Sigma-Aldrich) was used as control DNA. After transformation, protoplasts were incubated 16 h in WI solution (4 mM MES, pH 5.7, 500 mM mannitol, and 20 mM KCl) supplemented with 15 mM sucrose. Enzymatic assays were performed as described by Yoo et al. (2007). Three technical replicates were performed per plasmid combination, and the whole experiment was repeated three times. The expression of effector proteins was checked by immunoblotting using polyclonal BAM7 (1:2500 dilution) and BAM8 antibodies (1:3000 dilution) previously produced in our lab (Reinhold et al., 2011).

Transcriptional Profiling
Three-week-old soil-grown Arabidopsis plants were harvested 4 h into the 12-h light period and immediately frozen in liquid nitrogen. Fourteen individual plants were pooled per replicate. Two biological replicates per genotype were grown for the microarray analyses. Sample preparation for ATH1 GeneChips was performed exactly as described previously (Wenke et al., 2012).
Overlap with upstream open reading frames was prevented. The occurrence of different motifs was determined using the Motif discovery tool and the Pattern matching tool (both from RSAT). Statistical significance of overrepresentation was calculated using the hypergeometric distribution function in Microsoft Excel using the motif occurrence in promoters genome-wide as background. The gene ontology tool from TAIR9 (http://Arabidopsis.org) was used to group genes into functional categories. Heat maps based on hierarchical clustering were generated with the statistical computing program R (http://www. r-project.org/) using the gplot packages. Comparisons of gene sets were conducted with Microsoft Excel and Microsoft Access. Principal component analyses were generated using GeneSpring Software (Agilent Technologies). Positional overrepresentation of the BBRE motif was calculated using Motif Mapper program (http://www2.mpiz-koeln.mpg.de/coupland/ coupland/mm3/html/) and visualized using promoter motif distribution curves (Berendzen et al., 2006).

Phylogenetic Analyses
BZR1-BAM homologs from different species were identified using the homolog search tool of the Phytozome v9.1 database (www.phytozome. net). Homologous proteins of Arabidopsis BAM7 and BAM8 were collected and manually checked for the presence of both the BZR1 and the BAM domain. After selection of one splice variant per homolog, all sequences were aligned together with Arabidopsis BAM1, BAM3, BZR1, and BES1 and used for the construction of a phylogenetic tree using the nearest neighbor method in MEGA5 (Tamura et al., 2011). The clades containing Arabidopsis BAM7 and BAM8 were assigned as BAM7 homologs and BAM8 homologs, respectively. After removal of all duplicate sequences, the final phylogenetic tree was constructed by the maximum likelihood method with 1000 bootstrap replicates in MEGA5. The degree of residue conservation was calculated and visualized using CLC Genomics Workbench 6 (CLC bio).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure 1. The Transcription Factor-Derived BZR1 Domain and the Pseudo-Enzymatic BAM Domain Both Contribute to the Differences between BAM7 and BAM8 Isoforms. Supplemental Figure 9. Detection of Recombinant HIS-Tagged BZR1-BAM Proteins by Immunoblotting.
Supplemental Table 3. The BBRE Motif Is Overrepresented in the Promoters of BZR1-BAM Deregulated Genes.
Supplemental Table 4. "BBRE Genes" Encoding Chloroplastic and Nuclear Proteins Which Are Deregulated in BZR1-BAM Mutants and Transgenic Overexpressing Lines.
Supplemental Table 5. Expression Clusters of "BBRE Genes" Encoding Chloroplastic and Nuclear Proteins Which Are Deregulated in BZR1-BAM Mutants and Transgenic Overexpressing Lines.
Supplemental Table 6. Primers Used for Cloning.