SEARCH

SEARCH BY CITATION

Keywords:

  • brassinosteroid;
  • plant hormone;
  • ChIP-chip;
  • microarray;
  • transcription factors;
  • gene regulatory network

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Brassinosteroids (BRs) are important regulators for plant growth and development. BRs signal to control the activities of the BES1 and BZR1 family transcription factors. The transcriptional network through which BES1 and BZR regulate large number of target genes is mostly unknown. By combining chromatin immunoprecipitation coupled with Arabidopsis tiling arrays (ChIP-chip) and gene expression studies, we have identified 1609 putative BES1 target genes, 404 of which are regulated by BRs and/or in gain-of-function bes1-D mutant. BES1 targets contribute to BR responses and interactions with other hormonal or light signaling pathways. Computational modeling of gene expression data using Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNe) reveals that BES1-targeted transcriptional factors form a gene regulatory network (GRN). Mutants of many genes in the network displayed defects in BR responses. Moreover, we found that BES1 functions to inhibit chloroplast development by repressing the expression of GLK1 and GLK2 transcription factors, confirming a hypothesis generated from the GRN. Our results thus provide a global view of BR regulated gene expression and a GRN that guides future studies in understanding BR-regulated plant growth.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Brassinosteroids (BRs) are a group of plant steroid hormones that play fundamental roles in orchestrating plant responses to developmental and environmental cues. They are widely distributed throughout the plant kingdom and among various plant tissues such as roots, shoots, leaves and flowers (Clouse and Sasse, 1998). BRs regulate multiple biological processes, including cell elongation, vasculature differentiation, photomorphogenesis, senescence and stress responses. Defects in BR biosynthesis or perception result in severe dwarfism, delayed senescence, de-etiolation in dark and reduced male fertility (Chory et al., 1991; Clouse et al., 1996; Li and Chory, 1997).

Knowledge has been built up about BR signal transduction in the past decade (Kim and Wang, 2010; Li, 2010). BRs are perceived by a membrane-localized receptor kinase, BRI1 (Brassinosteroid Insensitive 1) (Li and Chory, 1997), which binds BRs with a 70 amino acid island domain and a nearby leucine-rich repeat (LRR) motif in the extracellular domain (Wang et al., 2001; Kinoshita et al., 2005). Binding of BRs leads to the release of its inhibitor BKI1, activation of BRI1’s intracellular kinase domain and association with co-receptor BAK1, through a series of phosphorylation events (Li et al., 2002; Nam and Li, 2002; Wang et al., 2005a,b, 2006, 2008; Oh et al., 2009b). Several BRI1 substrates have been identified, including BSK1 that functions through BSU1 phosphatase to regulate negative regulator BIN2 (Li and Nam, 2002; Mora-Garcia et al., 2004; Nam and Li, 2004; Ehsan et al., 2005; Tang et al., 2008; Kim et al., 2009). BIN2, a GSK3-like kinase, phosphorylates the BES1/BZR1 family transcription factors to negatively regulate BR signaling (Choe et al., 2002; Li and Nam, 2002; Pérez-Pérez et al., 2002; Peng et al., 2010). Recent studies have suggested that phosphorylation controls BES1/BZR1 in several ways, such as targeting protein to proteasome-mediated degradation, retaining protein at the cytoplasm and reducing DNA-binding affinity (Wang et al., 2002; Yin et al., 2002; Vert and Chory, 2006; Gampala et al., 2007; Ryu et al., 2007, 2010a,b).

BES1 and BZR1 are two well characterized transcription factors in the BR signaling pathway. They are unique to plants and share high homology at the amino acid level (Wang et al., 2002; Yin et al., 2002; Zhao et al., 2002). They both have an atypical basic helix–loop–helix (bHLH) DNA-binding motif in the N terminus and have been shown to bind E-box (CANNTG) and BRRE (CGTGT/CG) elements, respectively (He et al., 2005; Yin et al., 2005). The middle portion of these proteins harbors multiple BIN2 phosphorylation sites and a PEST domain. A single proline to leucine substitution in the PEST domain results in the accumulation of both phosphorylated and unphosphorylated BES1 and BZR1 in bes1-D and bzr1-D mutants, indicating that the PEST domain has an important role in controlling protein stability.

BES1 has been shown to interact with other transcription factors to synergistically activate target genes (Yin et al., 2005; Li et al., 2009) and to recruit two jumonji domain-containing histone demethylases ELF6 and REF6 (Yu et al., 2008), and a transcription elongation factor IWS1 (Li et al., 2010) to regulate gene expression. Several atypical bHLH proteins are involved in BR signaling either positively (ATBS1 and PRE1) or negatively (AIF1 and IBH1) (Wang et al., 2009; Zhang et al., 2009). Another bHLH protein, TCP1, was recently found to promote the expression of BR biosynthesis gene DWF4 (Guo et al., 2010). Genome-wide microarray experiments in Arabidopsis have demonstrated that BRs regulate the expression of many genes. For example, in light-grown seedlings, 2.5 h BR treatment induces the expression (transcript accumulation) of 342 genes and represses the expression of 296 genes (Nemhauser et al., 2004). We found that in adult plants, about 1170 genes showed altered expression after 2.5 h BR treatment (Guo et al., 2009). Treatment with BR for longer periods of time leads to successively more changes in transcript profiles (Goda et al., 2004). The transcriptional network through which BRs regulate thousands of target genes is largely unknown. In addition, the majority of BES1 direct target genes have not been identified. Toward this end, we performed chromatin immunoprecipitation coupled with high-resolution tiling arrays (ChIP-chip) and global gene expression studies to identify and characterize BES1 direct target genes in the Arabidopsis genome.

ChIP-chip has been used to generate high-resolution maps of genome-wide distributions of histone modifications (Morris et al., 2007; Opel et al., 2007; Zhang et al., 2007; Minsky et al., 2008) and transcription factor binding sites in several species. The latter can be exemplified by studies on transcription factors such as Twist, Biniou and Ladybird in Drosophila (Jakobsen et al., 2007; Sandmann et al., 2007), estrogen receptor and homeobox C6 in human cancer cells (Carroll et al., 2006; McCabe et al., 2008), as well as HY5, AGL15, PIL5 and FLP in Arabidopsis (Lee et al., 2007; Oh et al., 2009a; Zheng et al., 2009; Xie et al., 2010). These studies have led to unexpected discoveries in transcription regulation and also provided unique resources for further research.

In this study, we identified BES1 target genes based on ChIP-chip and gene expression studies and characterized BES1 in vivo binding sites. We established a BR transcriptional network based on computational modeling and confirmed the functions of many genes in the network in BR responses. Our results, therefore, provide new insights into the mechanisms by which BES1 regulates downstream gene expression and mediates BR responses in Arabidopsis.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Genome-wide identification of BES1 binding regions by ChIP-chip

In order to map in vivo BES1 binding regions throughout the genome, we utilized the Arabidopsis Tiling 1.0R Array from Affymetrix, which is composed of more than 3 million probes representing the non-repetitive genome with one probe per 35 bp DNA on average. We raised anti-BES1 antibody against the C-terminus of BES1 (aa 89–335) that does not include the most conserved N-terminal DNA binding domain. To test its specificity, the affinity-purified antibody was used to detect BES1 in wild-type, bes1-D or bzr1-D mutants (in which BES1 and BZR1 accumulate to high levels, respectively). The antibody mostly recognizes BES1 in plants as it detected dramatic increase of BES1 protein in bes1-D but not the increase of BZR1 in bzr1-D mutant (Figure 1a). We decided to use the bes1-D mutant in the ChIP-chip experiments based on the following reasons. First, the mutant BES1 protein accumulates significantly more than in the wild-type plants (Figure 1a), which assures that most of the positive intervals detected are likely derived from BES1, rather than its close homologs. Second, the mutant BES1 protein is functional as bes1-D mutant displays constitutive BR responses and can further accumulate in response to BRs (Yin et al., 2002). BES1-bound DNA was isolated with affinity-purified BES1 antibody, from bes1-D mutant seedlings with three biological repeats, using an unrelated IgG as negative control. The datasets in triplicate were analyzed by CisGenome with the Moving Average (MA) option, which has been demonstrated to have high sensitivity and accuracy (Ji and Wong, 2005; Gottardo et al., 2008). After comparing several criteria, positive intervals were defined as regions that contained at least two continuous probes above the statistic threshold (MA ≥ 3) (Figure 1b). These parameters produced most significant enrichment of BES1 targets compared with BR-regulated genes.

image

Figure 1.  Identification of BES1 target genes in Arabidopsis genome. (a) Anti-BES1 antibody mostly recognizes BES1, but not BZR1. Same amount of proteins from wild-type (WT), bes1-D or bzr1-D were used in the western blotting with affinity-purified anti-BES1 antibody (Lanes 1–3). Different dilutions of protein from bes1-D were used to estimate the accumulation of BES1 in the bes1-D mutant (Lanes 4–6). A background band was used as loading control. (b) A typical positive interval detected by CisGenome, showing the fold changes between ChIP samples and negative controls (top), MA statistics and cutoff (middle) and the detected interval (bottom). (c) Distribution of positive intervals on five chromosomes. (d) Quantitative (q) PCR validation of 22 BES1 target genes identified by ChIP-chip with chromatin prepared from bes1-D or WT plants. Totally 49 BES1 target genes are selected for confirmation by semi-quantitative PCR (Figure S1), from which 22 were selected for q-PCR analysis. Averages of enrichment by anti-BES1 antibody and standard deviations are derived from biological replicates. Green line indicates no enrichment (fold change = 1). All except for the last three genes have significant enrichment.

Download figure to PowerPoint

Under such stringent conditions (Zhang et al., 2006), we identified 1718 BES1 positive intervals across the genome, which were distributed on all five chromosomes (Figure 1c). Among all the positive intervals, none of them was from the chloroplast genome (ChrC, 150 kb) and only two were identified in the mitochondrion genome (ChrM, 370 kb), which has high homology with areas on Chr2. The result suggests that the positive regions we identified are likely true BES1 targets. All chromosomal intervals were assigned to genes if they are located in the gene promoters and/or gene bodies according to the TAIR8 annotation, which identifies 1609 putative BES1 target genes that accounts for 5.1% of the total gene population in the Arabidopsis genome (Table S1).

To determine how many of the BES1 targets identified in bes1-D mutant are also targets in wild-type, we selected 49 genes ranked from 4 to 1644 from the positive intervals and conducted ChIP-PCR with bes1-D mutant and wild-type plants by semi-quantitative and quantitative PCR (Figures 1d and S1). All 49 genes are confirmed to be BES1 targets in bes1-D and about 45 of them are also confirmed in wild-type plants. The enrichment in wild-type is in general less robust than in bes1-D, reinforcing the importance of using bes1-D in Chip-chip analysis. The result may also explain, at least in part, the fact that BRs usually induce gene expression by about two to four-fold (Yin et al., 2002). Taken together, the results indicate that about 90% of BES1 targets identified in bes1-D are physiological targets of BES1 in wild-type plants.

BES1 directly regulates many BR-responsive genes

We first compared the 1609 BES1 targets identified by ChIP-chip with 2163 BR-regulated genes (Tables S2 and S3) at different developmental stages, including the ones from this study (see next) and those published previously (Goda et al., 2004; Nemhauser et al., 2004; Guo et al., 2009). Eighty-five out of 1188 (7.2%) BR-induced genes and 165 out of 975 (17.0%) BR-repressed genes overlapped with the targets of BES1 in vivo (Figure 2a). Therefore, a total of 250 (11.6%) BR-responsive genes were direct BES1 targets (Table S4).

image

Figure 2.  Many of BES1 targets are regulated by BRs and/or in bes1-D mutant. (a) BR-regulated genes are compiled from 10-day-old (Nemhauser et al., 2004), 14-day-old (this study), 24-day-old (Guo et al., 2009) wild-type or 7-day-old det2 plants (Goda et al., 2004). Genes differentially expressed in bes1-D are derived from 14-day-old plants. Of the 1609 BES1 target genes, only 1090 are represented in Affymetrix microarrays. (b) Venn diagram shows overlaps among BES1 target genes, BR-regulated genes and genes that are differentially expressed in bes1-D mutants. (c) Clustering analysis of BES1 targets that are either regulated in bes1-D (indicated at right, purple, up; blue: down) or by BRs (indicated at left, red: up; green: down). (d) Clustering analysis of BES1 targets that did not appear to be regulated in bes1-D or BRs by statistic analysis (see Experimental procedures) under tested conditions.

Download figure to PowerPoint

To determine how many BES1 target genes are differentially expressed in the bes1-D mutants (in which BES1 accumulates to high level, Figure 1a), we performed microarray experiments using bes1-D seedlings with or without BL treatment. A total of 4194 genes are either down- or up-regulated in bes1-D mutant (Figure 2a,b, Tables S5 and S6). About 272 of them are BES1 direct targets, 118 of which are regulated by BRs. Interestingly, 154 BES1 target genes that are regulated in the bes1-D did not appear to be responsive to BRs (Table S7). BR regulation of these genes may not be detectable under tested conditions in wild-type but their regulations by BES1 are magnified in bes1-D. These genes thus are also considered verified BES1 targets. All together, 404 (250 + 154) out of 1609 putative BES1 targets, are differentially expressed in response to BR and/or in bes1-D (Figure 2a,c). These 404 genes can be either up- or down-regulated by BRs and BES1 (Figure 2c), suggesting that BES1 can function either as an activator or as a repressor. Interestingly, 686 BES1 targets, which do not appear to be regulated by BRs or BES1, have low levels of expression in general, implying that they may be regulated by BR at different growth or developmental stages (Figure 2d).

We performed a Z-test on whether the percentage of BES1 target genes that are regulated by BRs (11.6%) or in bes1-D (6.5%) is significantly higher than the percentage of BES1 targets at the whole genome level (5.1%), i.e., whether we have true enrichment of BES1 targets among the genes that are regulated by BRs or in bes1-D. The p-values are very small (0 for BR-regulated BES1 targets and 0.0000188 for BES1 targets regulated in bes1-D). Taking consideration of multiple testing, we still found that there are significant enrichments of BES1 targets among the genes regulated by BRs or in bes1-D. The 404 verified BES1 target genes that are either regulated by BRs or by bes1-D were chosen for further analysis (Tables S4 and S7).

Characteristics of BES1 in vivo binding sites

To gain a global view of the distribution of BES1 binding regions on the BES1 target genes, we determined the distances from the centers of the positive intervals to the transcription start sites (TSS). TSS for each gene is determined according to the TAIR8 annotation. Although it is well established that eukaryotic transcription factors can function over long distances, BES1 tended to bind to DNA sequences near the TSS (Figure 3a,b). In all BES1 targets (Figure 3a) and in the BR-regulated BES1 targets (Figure 3b), the highest frequencies of interval occurrence appeared to be around the TSS (–500–500) with a slight preference to the promoter side. So we conclude that BES1 tends to bind to DNA elements in close proximity to the TSS to regulate gene expression.

image

Figure 3.  Characterization of BES1 target sites. (a,b) Distances from the centers of the positive intervals to TSS for all BES1 target genes (a) or BR-responsive BES1 target genes (b). (c,d) Conserved motifs found by de novo motif discovery from BES1 positive intervals in the top 85 BR down-regulated (c) or up-regulated (d) genes.

Download figure to PowerPoint

BES1 was previously found to bind to the E-box (CANNTG) (Yin et al., 2005) and its close homolog BZR1 was found to bind to the BRRE (BR Response Element, CGTGT/CG) to regulate gene expression (He et al., 2005). To further define BES1 in vivo binding sites, we performed de novo motif discovery of the BR-regulated BES1 targets using the cosmo algorithm (Bembom et al., 2007) and found that while E-boxes are present in both BR-repressed and BR-induced genes, BRRE is more dominant in BR-repressed genes (Figure 3c,d). We performed gel mobility shift assay with labeled DNA probes containing BRRE as well as CACGTG and CACTTG E-boxes (Figure S2). BES1 can bind to both BRRE and E-boxes (CACGTG and CACTTG). The stronger binding of BES1 to BRRE element than to E-boxes is likely due to the fact that BES1 needs a heterodimer partner to bind E-box more efficiently (Yin et al., 2005). Taken together, the in vitro and in vivo results suggest that BES1 can target both BRRE and E-boxes to regulate gene expression.

BES1 target genes are involved in multiple aspects of BR pathway

BES1 directly targets genes of different functional groups, including signaling molecules, enzymes, and many with unknown functions (Figure 4a). Some BR-responsive BES1 target genes contributed to cell elongation and cellular growth, such as cell wall modifying enzymes (Tables S2, S3, S4 and S7). BES1 positively regulated itself and this self-amplification mechanism could increase the hormone response efficiency. On the other hand, BES1 repressed the expression of several BR biosynthesis enzymes such as CPD and DWF4, as well as the expression of BR signaling components BRI1, and BES1 homologs BEH1 and BEH2, thereby attenuating BR responses in a feedback inhibition loop (Figure S3).

image

Figure 4.  Functional classification of BR-responsive BES1 target genes. (a) Molecular functions of BR-responsive BES1-target genes. (b) Gene ontology analysis. Categories with gene enriched were presented and p-values were indicated on top of bars. ABA: abscisic acid; GA: gibberellin; JA: jasmonic acid.

Download figure to PowerPoint

BES1 also tightly connects the BR pathway to other hormone responses in Arabidopsis. Genes involved in auxin (= 0.000), gibberellin (= 0.001), light (= 0.002), and abscisic acid (= 0.003) signaling were over-represented in BR-responsive BES1 target genes according to gene ontology analysis (Figure 4b), which was consistent with the observed interactions among these hormones (Moller and Chua, 1999). For example, it is well documented that BRs and auxin function cooperatively to regulate many developmental processes such as cell elongation (Hardtke, 2007). Our results showed that BES1 targeted not only auxin responsive genes such as SAUR-like proteins and the IAA1 transcription factor, but also auxin efflux facilitators (PIN4, PIN7) which control the spatial distribution of auxin and have profound effects on auxin action (Paponov et al., 2005). The results suggest that one mechanism for BR–auxin interaction is through BES1 regulated genes involved in auxin responses and transport.

BES1 initiates a hierarchical transcription network downstream of BR signaling

We found that 34 BES1 targeted transcription factors (BTFs) are responsive to BR treatment (Figure 2a) and hypothesize that these transcription factors form a transcriptional network to mediate BR responses. If the BTFs indeed function to mediate BR responses, we should be able to observe some expression correlation among them, which can be demonstrated by reconstruction of a gene regulatory network (GRN). For this purpose, we employed ARACNe (Algorithm for the Reconstruction of Accurate Cellular Networks), which was developed and successfully used to infer a GRN in human B cells (Basso et al., 2005). ARACNe calculates expression correlations between genes based on mutual information and picks out statistically significant correlations, presented as edges in GRNs (Basso et al., 2005). We used 199 BR-responsive transcription factors (including 34 BTFs) as well as two BR-signaling components, BRI1 and BIN2, as probes to reconstruct a GRN based on the publicly available microarray datasets. A sub-network including BTFs and several other BR-regulated transcription factors is presented in Figure 5. The BR-regulated transcription factors including BTFs are extensively interconnected, suggesting strong expression correlation and possible interactions among them. Interestingly, BES1 expression does not closely correlate with most BTFs, presumably because BES1 activity is mainly regulated at the protein level (only moderate transcriptional change in response to BR) so BES1 direct targets may not show direct connections in this network based on transcript levels (Yin et al., 2002). Nevertheless, the network confirms some previously published gene interactions. For example, it is known that BR induces the bHLH genes BEE1, BEE2 and BEE3, which function redundantly to mediate some BR responses (Friedrichsen et al., 2002). Consistent with the genetic results, interactions are detected between BES1, BEE2 and BEE3. Furthermore, we recently found that MYB30, connected to several BES1 targets as well as BR signaling gene BIN2 in the network, is a BES1 target gene that functions to amplify BR signaling (Li et al., 2009).

image

Figure 5.  A transcriptional network for BES1-regulated gene expression. GRN inferred by ARACNe. Green and red circles represent BR up- or down-regulated BES1 targets, respectively. Pink circles represent BR-regulated transcription factors or BIN2 that are not BES1 direct targets. *Genes with T-DNA mutants analyzed.

Download figure to PowerPoint

The network also predicts previously unknown interactions such as those between BES1, PIL6, GLK1 and GLK2. PIL6, a bHLH factor involved in light signal transduction (Fujimori et al., 2004), is a BES1 direct target and interacts with many other BR-regulated transcription factors and GLK1/GLK2. GLK1 and GLK2 are related transcriptional factors that function redundantly to promote chloroplast development (Fitter et al., 2002; Waters et al., 2008, 2009). BR mutants have premature chloroplast development and ectopic expression of light-regulated genes when grown in the dark (Chory et al., 1991). The interactions between BES1 and GLK1 and GLK2 suggest that BES1 can prevent chloroplast development in the dark by repressing expression of GLK1 and GLK2.

To test the hypothesis, we examined the chloroplast development in bes1-D grown under light, in which chloroplasts are developed so the inhibitory effect of BES1 can be assayed. Consistent with the idea that BES1 functions through GLK1 and GLK2 to inhibit chloroplast development, bes1-D seedlings are pale green and have reduced amounts of chlorophyll (Figure 6a,b). These reductions are not accompanied by drastic alterations in chloroplast ultrastructure or in the ability of the mutant to form thylakoids and grana, but rather, bes1-D chloroplasts have dramatically enlarged plastoglobules (PG) (Figure 6b,c). Plastoglobules are lipoprotein bodies that participate in a variety of metabolic pathways, primarily involving lipids, carotenoids and tocopherols (vitamin E) (Grennan, 2008). Some of these compounds protect thylakoid membranes from oxidative stress, and in agreement with this role, plastoglobule size and/or number increase under various abiotic and biotic stress conditions (Brehelin and Kessler, 2008). The finding of abnormally large plastoglobules in bes1-D chloroplasts therefore points toward an impact of BES1 accumulation on chloroplast biogenesis and/or function. Since BES1 acts to repress the expression of GLK1 and GLK2 (Figure 6e), we examined the expression of the top 20 genes that are regulated by GLK1 and GLK2 (Waters et al., 2009) and found that at least seven of them have significantly reduced expression in bes1-D (Figure 6f). Many of these genes encode chloroplast proteins, e.g. light-harvesting chlorophyll binding proteins (LHCB), consistent with the defect of bes1-D in chloroplast function. In summary, our data clearly demonstrate that BES1 adversely affects chloroplast function, perhaps via inhibition of GLK1 and GLK2 expression. Our results, therefore, confirm one of the hypotheses generated from the network.

image

Figure 6.  BES1 can inhibit chloroplast function, likely through the inhibition of GLK1 and GLK2 expression. (a) 10-day-old seedlings of wild-type (WT) and bes1-D seedlings. (b) Total chlorophyll is reduced in bes1-D. (c,d) EM images show chloroplasts from wild-type (WT) and bes1-D. Bars represent 0.5 mm. PG: plastoglobule. (e) GLK1 and GLK2 expression is reduced by BL and/or in bes1-D. (f) GLK1 and GLK2-upregulated genes are reduced in bes1-D. The gene expression levels in (e) and (f) are from the microarray experiments described in this study. The differences between WT and bes1-D are significant as tested by two-way anova F test (P < 0.05).

Download figure to PowerPoint

To further validate the GRN, we identified T-DNA knockout mutants of 15 BTF genes for which T-DNA knockout mutants are available. We determined the BR responses by the hypocotyls elongation assays under two conditions: (i) with BR biosynthesis inhibitor BRZ, which reduces endogenous BR level and thus hypocotyl lengths in the dark; and (ii) with BL treatment, which promotes hypocotyl elongation in the light. Remarkably, mutants for 13 individual genes displayed BR response phenotypes either in the dark or light conditions (Figure 7), among which, BES1 (At1g19350), ZFP8 (At2g41940), MYB30 (At3g28910), At1g64380 and At1g79700 are BR-induced and IAA7 (At3g23050), LHY1 (At1g01060), At1g01520, ATHB5 (At5g65310), TCP(At2g45680), At2g43060, RD26 (At4g27410) and HAT1 (At4g17460) are BR repressed (Figure 5). Interestingly, with the exception of IAA7 and LHY1, the mutants for BR-induced genes are more sensitive to BRZ or less responsive to BL, while the mutants of BR-repressed genes are more resistant to BRZ or more responsive to BL (Figure 7b,d). The results suggest that BES1 functions to up- or down-regulate many transcriptional factors and each of them contributes to a portion of BR responses.

image

Figure 7.  The BRZ and BR responses of BES1 target gene mutants. (a) Hypocotyl lengths were measured with 6-day-old seedlings in the absence or presence of 1000 nm BRZ in the dark. (b) The BRZ sensitivity is calculated as ratio of hypocotyl lengths in the presence BRZ over that of control for each genotype. The green line indicated the wild-type BRZ sensitivity. (c) Hypocotyl lengths were measured with 2-week-old seedlings grown in the light in the absence or presence of 100 nm BL. (d) BL-induction fold is calculated, with green line indicating the induction fold in wild-type. For hypocotyl lengths, averages and standard deviation are calculated from 10–20 seedlings. All the mutants shown have significantly different BRZ/BR responses compared to wild-type, as determined by Student’s t-test (P < 0.05).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Transcriptional regulation is pivotal for many hormones to exert their biological effects in both plants and animals. To systematically investigate BES1-regulated transcription downstream of BR signaling, we performed ChIP-chip experiments along with the gene expression analysis and identified direct target genes of BES1 in the Arabidopsis genome. We demonstrated that BES1 binds to divergent DNA elements to activate or repress transcription. Interestingly, BES1-targeted transcription factors form a GRN that appears to mediate many BR responses and interactions with other signaling pathways. Thus we have established a substantial framework for BES1-regulated genes in Arabidopsis that should aid us in deciphering the mechanisms of many BR-regulated processes.

ChIP-chip is an effective approach to map global transcription factor binding sites, because of its high resolution, high throughput and less bias compared with conventional PCR-based methods. By ChIP-chip with a 35 bp resolution whole genome tiling array, we identified 1609 potential BES1 target genes and 404 of them are differentially expressed in response to BR treatment or in bes1-D mutant (Figure 2a). The large number of putative BES1 target genes is comparable with those of other transcription factors published. For example, HY5 and PIL5 were found to bind 3894 and 748 genes in Arabidopsis (Lee et al., 2007), while SOX2 and NANOG targeted 1271 and 1687 genes in human (Boyer et al., 2005).

Several lines of evidence suggest that the targets we identified are true BES1 targets. First, BES1 in vivo binding sites derived from BES1 direct target genes are largely consistent with previously in vitro DNA binding results (He et al., 2005; Yin et al., 2005). Second, many previously well known BR targets, such as DWF4, CPD, BRI1 and many auxin-responsive genes, are identified by the ChIP-chip experiments. Third, gene expression studies demonstrate that a significant portion (37%) of BES1 target genes are regulated by BRs and/or BES1, while most of the BES1 targets that did not appear to be regulated by BRs/BES1 have in general very low expression.

Why not all the BES1 targets are regulated by BRs or BES1? First of all, the microarray experiments we used as comparison here only examined the gene expression at three developmental stages after short (0.5–3 h) BR exposure; using plants at different developmental stages with longer BR treatment may reveal more BR-regulated genes. Moreover, BES1 binding to promoters is often not sufficient to regulate gene expression and other cooperating factors are required for BES1 to activate or repress gene expression. This finding is in consistent with our recent finding that BES1 cooperates with one of its target genes, MYB30, in the activation of a subset of BR target genes (Li et al., 2009). Another example comes from MHCII gene expression in human. Three transcription factors (RFX, NF-Y and X2BP) always bind to the promoter but MHCII expression only takes place in B cells where protein CIITA is available and physically interacts with these transcription factors (Masternak et al., 2000; Merika and Thanos, 2001). So it seems that transcription factor binding potentiates gene expression but transcription is merely allowed when transcription factors are ‘activated’ under certain developmental or environmental conditions (Merika and Thanos, 2001; Carroll et al., 2006). Therefore, it is not surprising that, for all the transcription factors of which the target genes are globally determined, only a portion of their target genes are known to be differentially expressed. For example, HY5 has 3894 target genes, about 19% of them are differentially expressed in hy5 mutant (Lee et al., 2007). AGL15 has 1706 target genes, 48 are activated and 97 are repressed as revealed by gene expression studies in a AGL15 overexpression and agl15 agl18 mutants (Zheng et al., 2009). PIL5 has 748 binding sites from ChIP-chip experiments and 166 of them (or 22%) showed differential expression in microarray analysis (Oh et al., 2009a).

Characterization of BES1 positive intervals in these target genes reveals interesting features of BES1-regulated gene expression. In contrast to the conventional thinking that regulatory elements are in promoters, about half of BES1 positive intervals are actually located in gene bodies, even several kb downstream of the transcription start sites (TSS). They seem to be as functional as those in promoters, because half of BR-responsive BES1 target genes have BES1 positive intervals in their gene bodies. In addition, BES1 tends to bind DNA elements in the proximity to the TSS since more than 50% of positive intervals fall into the –500–1500 regions. Similar observations have also been made in the study of Arabidopsis transcription factor HY5 (Lee et al., 2007). On the other hand, only 4% of the estrogen receptor binding sites are located in promoter-proximal regions in human cells (Carroll et al., 2006). This difference may stem from the lack of redundant sequences in the Arabidopsis genome compared to the human genome.

Most significantly, the identification of BES1 direct target genes leads to the establishment and functional confirmation of a BR transcriptional network that include BES1-targeted as well as BR-regulated transcription factors (Figure 5). The fact that BTFs show extensive expression correlation with each other supports the hypothesis that BES1 initiates a transcriptional network to control gene expression and BR responses. Indeed, 13 out 15 BES1 target genes selected from the network showed BR response phenotypes when knocked-out (Figure 7). The results demonstrated that BES1 controls the expression of many transcription factors and each of them contributes to BR responses. Our conclusion is corroborated by a similar study in the control of human leukemia cell growth arrest and differentiation (Suzuki et al., 2009). More than 52 transcriptional factors have been identified to contribute to the process, positively or negatively, in a transcriptional network, but none of them is both necessary and sufficient to drive the differentiation process. This genetic redundancy probably ensures that a process is properly regulated under various developmental or environmental conditions.

The BR transcriptional network has successfully predicted some important interactions that explain well known BR responses. For example, several lines of evidence support the hypothesis that BES1 acts to repress GLK1 and GLK2, thereby inhibiting chloroplast biogenesis and/or function. Consistent with our results, it was recently found that BPG2, a protein involved in chloroplast protein accumulation, is involved in BR signaling (Komatsu et al., 2009). In the future, it will be interesting to determine if BES1 represses GLK1 and GLK2 expression through PIL6, as the network suggests, and to study how BR signaling regulates chloroplast development under different conditions (dark/light).

In summary, our study has identified many BES1 target genes that are involved in plant growth, BR biosynthesis and signaling. Our experiments have also implicated BES1 targets in interaction between BR and other hormonal or light signaling pathways. Using ChIP-chip, global gene expression studies and computational modeling, we have generated a GRN that confirms many known gene interactions and, significantly, provides testable hypotheses about BR function. The network we have generated has been validated by genetic studies showing that mutants of many genes in the network display BR response phenotypes and by establishing a molecular link between BES1 and chloroplast development. Future efforts will be focused on functional characterization of BES1 targeted transcription factors and their interactions in the regulation of plant growth, development and response to environmental changes.

Experimental procedures

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Plant materials and growth conditions

Arabidopsis thaliana ecotype Columbia (Col-0) and Enkheim-2 (En-2) were the wild types. bes1-D is the gain-of-function mutant described before (Yin et al., 2002). Plants were grown on MS plates or in soil under long day (16 h light/8 h dark) conditions at 22°C.

BES1 antibody production

The truncated BES1 (aa 89–335) fused with MBP was expressed, purified from E. coli and used to inject rabbits. The BES1 antibody was purified with GST–BES1 (aa 89–335) coupled to CNBr-activated sepharose (Phamacia).

Sample preparation for ChIP-chip

Chromatin immunoprecipitation was performed with 14-day-old bes1-D seedlings using protocol modified by Pikaard group (http://sites.bio.indiana.edu/~pikaardlab/) based on published methods (Gendrel et al., 2002; Nelson et al., 2006). Briefly, plant tissues were cross-linked with formaldehyde and nuclei were isolated using sucrose gradients. Chromatin was sonicated to generate fragments with the average size of 500 bp and precipitated using anti-BES1 antibody or an unrelated IgG as negative control. Immunocomplexes were harvested by protein A beads, washed and reverse cross-linked by boiling in the presence of Chelex resin (Bio-Rad, http://www.bio-rad.com/). Three biological replicates were carried out through the whole process.

DNA samples from ChIP were then processed following the Affymetrix instructions. ChIP-PCR with known BES1 target sites was first performed to evaluate sample qualities. Validated samples were subsequently subjected to linear and PCR amplification where dUTP was incorporated into the products, which were fragmented later by the uracil DNA glycosylase and APE1 enzymes. DNA fragments were then labeled with biotin using the GeneChip® WT double-stranded DNA terminal labeling kit (Affymetrix, http://www.affymetrix.com/). Hybridization to the GeneChip Arabidopsis Tiling 1.0R Array from Affymetrix and image scanning were performed at the GeneChip Facility of Iowa State University (http://www.biotech.iastate.edu/facilities/genechip/Genechip.htm).

ChIP-chip data analysis

Signal normalization and detection of BES1 positive intervals were assisted by the CisGenome software (http://www.biostat.jhsph.edu/~hji/cisgenome/). Signal intensities were computed from perfect match (PM) probes and log2-transformed before quantile normalization. The Moving Average (MA) method was applied to calculate test-statistic for each probe by combining information from probes within a 15-probe (about 500 bp) window. Neighboring probes yielding MA-statistics equal to or larger than 3.0 were combined into positive intervals, in which at least two continuous probes should be above the threshold and at most five continuous probes were allowed to be below the threshold.

Gene expression analysis

Microarray experiments were performed with 2-week-old WT (En2) and bes1-D seedlings grown under light. The plants were treated with either 1 μm BL or mock control in liquid 1/2MS medium for 3 h. Triplicate samples were collected and processed for RNA extraction with RNeasy Plant Mini Kit (Qiagen, http://www.qiagen.com/). The probe labeling, hybridization to Affymetrix ATH1 Genome Arrays, and scanning were performed in GeneChip facility at Iowa State University according to manufacture’s instructions. The microarray data were normalized by statistical software R using MAS 5.0 method in Affymetrix package, and tested using limma package. The false discovery rate (FDR) is controlled at 5% (Benjamini and Hochberg, 1995). Clustering analysis was performed with the GENESPRING program (Silicon Genetics) using Pearson correlation.

ChIP-PCR validation

Primers were designed for selected intervals such that the PCR fragments covered interval centers. PCR reactions were carried out with two biological replicates. Primer sequences are described in Table S8.

Bioinformatics analysis

Positive intervals were assigned to annotated genes if they resided in promoters and/or gene bodies according to TAIR8 annotation. Here the promoter was defined as the entire 5′ intergenic region of one gene or half of it if this gene shared this region with its 5′ neighboring gene. This process was accomplished with R.

De novo motif discovery was performed with an R package cosmo with the motif distribution model of ‘ZOOPS’ or ‘TCM’ (Bembom et al., 2007). Sequence logos were generated with seqLogo (http://bioconductor.org/packages/2.1/bioc/html/seqLogo.html).

Gene ontology analysis was assisted by the GOstats package (Falcon and Gentleman, 2007). The ‘gene universe’ was consisted of all Arabidopsis protein coding genes and annotation was based on the data in the R package ath1121501. The hyperGTest function was employed to detect over-represented gene categories.

Network construction

The BES1-related transcription factor network was constructed using the ARACNe (Algorithm for the Reconstruction of Accurate Cellular Networks) software (Margolin et al., 2006). Expression data from 933 Affymetrix ATH1 Genome Arrays was retrieved from publicly available NASC and AtGenExpress databases (http://www.arabidopsis.info/;http://www.arabidopsis.org/info/expression/ATGenExpress.jsp). Signal intensities were normalized using the standard MAS 5.0 algorithm, provided by the R and Bioconductor packages (Gentleman et al., 2004). The microarray data was scaled to set the mean to 100, excluding probes with an average intensity below 100 and coefficient of variance ≤30%. Quantile normalization, as implemented in R, was then applied for further filtering of the normalized data. Approximately 14 497 genes passed the filtering criteria and were used as input for ARACNe. A total of 201 BR-related genes (199 BR-regulated transcription factors, BRI1 and BIN2) were included as probes for network construction, in which the p-value for the mutual information (MI) criteria in ARACNe was set to 1e−06 and DPI (Data Processing Inequality) tolerance was set to 0.1.

Gel mobility shift assay

Gel mobility shift assay (GMSA) was carried out as described previously (Yin et al., 2005).

Chlorophyll measurement and transmission electron microscope (TEM)

Chlorophyll contents were determined as previously described using rosettes from mature WT and mutant seedlings at the same stage of growth using 2-week-old light grown seedlings (Lichtenthaler, 1987). For TEM analysis, mutant and wild-type cotyledon tissue was fixed in a solution consisting of 2% glutaraldehyde/2% paraformaldehyde in a 0.1 m cacodylate buffer (pH 7.2). Samples were washed in 0.1 m cacodylate buffer and then post fixed in 1% buffered osmium tetraoxide. After another 0.1 m cacodylate buffer wash, the tissue was en bloc stained with 3% uranyl acetate. Samples were dehydrated in an ethanol series (25, 50, 70, 95, 100%) followed by a 100% acetone dehydration step. Directly after dehydration, samples were infiltrated with a ratio of acetone to Spurr’s resin mixture (3:1, 1:1, 1:3, pure resin). Samples were placed in molds with fresh 100% Spurr’s resin at 60°C to polymerize. Resin was sectioned on a Reichert Ultracut S ultramicrotome with a glass knife to a thickness of 70 nm and placed on formavar coated copper grids. Images were captured on a JEOL 2100 STEM.

T-DNA knockout mutants and BRZ/BR responses

The T-DNA insertion lines are obtained from ABRC (http://www.arabidopsis.org/) (Alonso et al., 2003). The accession numbers for the mutants and primers used to identify homozygous mutant plants are listed in Table S8. BRZ and BR responses are carried out as described (Guo et al., 2009; Li et al., 2009).

Accession numbers:

The Chip-chip data was deposited in the NCBI GEO database (access no. GSE24684).

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank ABRC for T-DNA knockout seeds, Hongkai Ji at University of Johns Hopkins for suggestions about CisGenome, Huaming Chen and Joe Ecker at the Salk Institute for help with the annotation of positive intervals, Tadao Asami at Tokyo University for providing BRZ, Jiqing Peng at the ISU microarray facility for performing microarrays and scanning, and Dr. Harry Horner for help with EM studies. The research was supported by NSF CCF-0811804 to S.A., DOE (Energy Biosciences grant no. DE–FG02–94ER20147) to S.R.R., a faculty start-up fund from Iowa State University and NSF grant (IOS0546503) to Y.Y.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Figure S1. Semi-quantitative PCR validation of 49 BES1 target genes. ChIP was performed with bes1-D and wild-type (WT) plants as described in methods. While ‘aBES1’ and ‘con’ indicate ChIP reactions with ant-BES1 and an unrelated antibodies, respectively. The 5srRNA was used as internal control. All the genes are confirmed targets in bes1-D; 44 out of 49 (with the exception of last five genes) are also target genes in WT. From 49 genes, 22 including five that did not show clear enrichment in semi-quantitative PCR were selected for q-PCR analysis, which showed that all except for three are BES1 targets in WT (Figure 1(d)). Gene numbers and their ranks in ChIP positive list are indicated. The genes that were further analyzed by qPCR are highlighted in red.

Figure S2. BES1 binds to both E-box and BRRE DNA elements. Gel mobility shift assay were performed with DNA probes containing indicated elements. Each lane contains about 100 ng BES1 and increasing concentrations of labeled probes (0.5, 1, 2, 4 and 8 ng in 20 ml binding reaction). The free and BES1-bound probes are indicated on the right.

Figure S3. A model for the functions of BES1-target genes. BES1 regulates itself in a positive feedback loop and several other BR-biosynthesis and signaling components in a negative feedback loop. In addition, BES1 directly targets many transcriptional factors to regulate genes required for BR responses.

Table S1. Putative BES1 target genes in Arabidopsis thaliana.

Table S2. BR up-regulated genes in Arabidopsis thaliana.

Table S3. BR down-regulated genes in Arabidopsis thaliana.

Table S4. BES1 target genes regulated by BRs.

Table S5. Genes down-regulated in bes1-D mutant.

Table S6. Genes up-regulated in bes1-D mutant.

Table S7. BES1 target genes differentially expressed in bes1-D but not regulated by BRs.

Table S8. Primers used in ChIP verification and T-DNA knockout genotyping as well as accession numbers for T-DNA knockout mutants.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
TPJ_4449_sm_FiguresS1-S3.pdf1183KSupporting info item
TPJ_4449_sm_Legends.doc26KSupporting info item
TPJ_4449_sm_TablesS1-S8.doc1682KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.