BBX32, an Arabidopsis B-Box protein, functions in light signaling by suppressing HY5-regulated gene expression and interacting with STH2/BBX21.

A B-box zinc finger protein, B-BOX32 (BBX32), was identified as playing a role in determining hypocotyl length during a large-scale functional genomics study in Arabidopsis (Arabidopsis thaliana). Further analysis revealed that seedlings overexpressing BBX32 display elongated hypocotyls in red, far-red, and blue light, along with reduced cotyledon expansion in red light. Through comparative analysis of mutant and overexpression line phenotypes, including global expression profiling and growth curve studies, we demonstrate that BBX32 acts antagonistically to ELONGATED HYPOCOTYL5 (HY5). We further show that BBX32 interacts with SALT TOLERANCE HOMOLOG2/BBX21, another B-box protein previously shown to interact with HY5. Based on these data, we propose that BBX32 functions downstream of multiple photoreceptors as a modulator of light responses. As such, BBX32 potentially has a native role in mediating gene repression to maintain dark adaptation.

In the absence of light below the soil surface, the elongating seedling (etiolated seedling) adopts a developmental strategy termed skotomorphogenesis, characterized by a long hypocotyl, unexpanded cotyledons protected by an apical hook, and lack of pigmentation. Exposure to light triggers deetiolation, a coordinated inhibition of hypocotyl elongation, initiation of cotyledon expansion, and stimulation of chlorophyll accumulation and chloroplast development. This developmental transition (photomorphogenesis) leads to photoautotrophic growth. Plants experience a high degree of diurnal and seasonal variance in ambient light quality (wavelength) and quantity (fluence). Information from the light signal used for directing developmental changes is perceived by multiple families of photosensory photoreceptors that detect light wavelength, intensity, duration, and direction. In Arabidopsis (Arabidopsis thaliana), red/far-red light is monitored by the phytochrome family (phyA-phyE), blue/UV-A light is perceived by the cryptochrome (cry1 and cry2) and phototropin (phot1 and phot2) families, and UV-B is perceived by a recently identified photoreceptor, UVR8 (Briggs and Huala, 1999;Briggs and Christie, 2002;Quail, 2002aQuail, , 2002bChen et al., 2004;Whitelam and Halliday, 2007;Rizzini et al., 2011).
Several studies using microarray-based analyses of seedlings grown in darkness or exposed to light have revealed that the first exposure to light triggers significant changes in gene expression. About 20% of the Arabidopsis and rice (Oryza sativa) genomes are differentially regulated between darkness and light. (Ma et al., 2001;Tepperman et al., 2001Tepperman et al., , 2004Tepperman et al., , 2006Ohgishi et al., 2004;Jiao et al., 2005). Over 40% and 25% of the genes responding early to the light signal (within 1 h; far-red and red light, respectively) are predicted to encode different classes of transcription factors and accessory proteins that assist transcription factor function, which in turn may regulate the light-mediated transcriptional network (Tepperman et al., 2001(Tepperman et al., , 2004(Tepperman et al., , 2006. This early-light-responsive set of transcriptional regulators includes proteins known to be involved in light signaling, such as ELONGATED HYPOCOTYL5 (HY5), and factors involved in circadian clock function, such as CIRCADIAN CLOCK-ASSOCIATED PROTEIN1 (CCA1) and LATE ELONGATED HYPO-COTYL (LHY; Tepperman et al., 2001). CCA1 and LHY are partially redundant MYB transcription factors and are important components of the plant circadian clock (Schaffer et al., 1998;Wang and Tobin, 1998).
HY5, a bZIP transcription factor, acts as a positive regulator of photomorphogenesis in all light conditions, downstream of phyA, phyB, cryptochromes, and UV-B (Koornneef et al., 1980;Ang and Deng, 1994;Oyama et al., 1997;Ulm et al., 2004). The hy5 mutant has elongated hypocotyls in all wavelengths of light and exhibits an increased number of lateral roots, shows defects in secondary thickening in roots, and accumulates markedly lower levels of chlorophyll and anthocyanin (Oyama et al., 1997;Holm et al., 2002). A study of in vivo binding sites of HY5 using chromatin immunoprecipitation coupled with DNA chip analysis demonstrated that HY5 preferentially binds to promoter regions throughout the genome (Lee et al., 2007). These direct HY5 binding sites include promoters of genes involved in photosynthesis and pigment synthesis, like CAB, RbcS1A, F3H, and CHS Chattopadhyay et al., 1998;Lee et al., 2007), as well as genes involved in circadian regulation, such as CCA1, LHY, TOC1, and ELF4 (Lee et al., 2007).
HY5 protein accumulates in light and is degraded in the dark (Osterlund et al., 2000;Pokhilko et al., 2011). HY5 binding sites are largely found in promoters of genes that are either induced or repressed by light. In 35S::HA:HY5 transgenic plants, HY5 was found to bind its target sites constitutively, in darkness and in different light conditions, revealing that binding of HY5 alone is not sufficient for regulating the transcriptional activities of these light-responsive genes (Lee et al., 2007). Rather, HY5 activity is dependent on the interaction of the transcription factor with transcriptional accessory proteins. Lee et al. (2007) suggested that HY5 functions high in a regulatory hierarchy, regulating a branch of the transcriptional cascade involved in lightmediated development. Recent studies have shown that HY5 plays a role at points of convergence between light and phytohormonal signaling pathways (Cluis et al., 2004;Vandenbussche et al., 2007;Chen et al., 2008). HY5 transcript was detectable in all organs of mature plants, and the effects of hy5 mutation on the activation of light-responsive promoters fused to a GUS reporter could be detected in the leaves, stems, and roots of older plants (Oyama et al., 1997;Chattopadhyay et al., 1998). Loss of function of HY5 HOMOLOG (HYH) results in weak impairment in the inhibition of hypocotyl elongation specifically in blue light, suggesting that HYH functions redundantly with HY5 under specific light conditions (Holm et al., 2002).
Here, we characterize a novel gene from the B-box family, B-BOX32 (BBX32; Khanna et al., 2009). BBX32 protein contains a single N-terminal B-box motif but lacks a CCT domain. We show that BBX32 gene expression is robustly induced by early light (within 1 h) treatment. Seedlings overexpressing BBX32 are hyposensitive to red, far-red, and blue light, similar to the hy5 mutant. We present evidence that BBX32 is a modulator of light signaling, functioning as a transcriptional accessory protein through its direct interaction with STH2, which is known to act with HY5.

RESULTS
BBX32 (AT3G21150; Khanna et al., 2009) was identified as a putative regulator of hypocotyl growth in a large-scale functional genomics project performed in Arabidopsis. Overexpression of BBX32 was observed to produce increases in hypocotyl length; as such, the gene was chosen for a detailed study in which the phenotypes of 35S::BBX32 lines were compared with the wild type and with those produced by a T-DNA insertion mutation (bbx32-1) within the gene (Fig. 1A).
Seedlings Overexpressing BBX32 Are Hyposensitive to Red, Far-Red, and Blue Light Two representative independent overexpression lines, 35S::BBX32-10 and 35S::BBX32-12, were examined; both lines showed clear hyposensitivity to continuous red (Rc), far-red (FRc), and blue (Bc) light (Fig.  1). These seedlings had elongated hypocotyls in re-sponse to increasing fluence rates of red, far-red, and blue light (Fig. 1, B-D). The 35S::BBX32-10 seedlings were relatively more hyposensitive than 35S::BBX32-12 seedlings under all light conditions. However, both lines were hyposensitive to all light conditions, with weak responsiveness to higher light intensities. By contrast, the bbx32-1 mutant exhibited marginally shorter hypocotyls under low fluence rates in comparison with its sib-Col (negative segregant) control ( Fig. 1, B-D). The bbx32-1 mutant has reduced levels of a truncated BBX32 transcript, but there was no detectable full-length BBX32 message in this mutant (Supplemental Fig. S1A). Since bbx32-1 was the only available mutant allele, we performed complementation analysis to test whether the observed phenotype was linked to the bbx32-1 mutation. Complementation of the bbx32-1 mutant with a prBBX32::BBX32 transgene rescued its short-hypocotyl phenotype (Supplemental Fig. S2). Double and triple mutants of bbx32-1 in combination with phyA, phyB, or cry1 showed that photoreceptor function under the respective light conditions was necessary for light responsiveness of bbx32-1 mutant seedlings ( Fig. 1-D). Note that the two controls showed a small difference in their responsiveness to light: EV-Col (for empty vector-Col) had shorter hypocotyls than sib-Col seedlings under all wavelengths ( Fig. 1-D). For experiments in blue light, we included cry1-Ler (for ecotype Landsberg erecta), Ler, and Col-wt (for Columbia wild type) seedlings for comparison. We observed some ecotypic differences between the control seedlings. The Ler controls were consistently shorter than the Col-wt in the dark and in blue light. The Col-wt responded similarly to Ev-Col seedlings, maintaining shorter hypocotyls than the sib-Col. The cry1/bbx32-1 (Ler/ Col background) double mutants were taller than Ler and cry1 seedlings in the dark, possibly due to the introduction of Col genes into the Ler background. The cotyledons of 35S::BBX32-10 and 35S::BBX32-12 seedlings were significantly smaller compared with the EV-Col seedlings in red light, but there was no significant difference between sib-Col and bbx32-1 mutants (Fig.  1E). These data are consistent with the conclusions from a previous study, where it was found that a large percentage of loss-of-function mutations in single genes implicated in light signaling do not result in any significant defects in deetiolation (Khanna et al., 2006). Here, we show that overexpression of BBX32 resulted in seedlings with elongated hypocotyls along with smaller cotyledons, consistent with phenotypes linked to an altered photomorphogenic response.

BBX32 Modulates Light-Regulated Gene Expression
Based on the effects on hypocotyl and cotyledon development, we surmised that overexpression of BBX32 might result in altered light-regulated gene expression. In order to test this hypothesis, we per- Figure 1. Seedlings overexpressing BBX32 exhibit hyposensitivity to continuous red (Rc), far-red (FRc), and blue (Bc) light. A, The B-box domain of BBX32 and the predicted T-DNA insertion site in the bbx32-1 mutant allele, at residue 172 of the 225 amino acids (aa) constituting the BBX32 protein, are shown. B to D, Hypocotyl growth response of the indicated genotypes exposed to increasing red (B), farred (C), or blue (D) light fluence rates. E, Relative cotyledon areas of seedlings grown under Rc light. Where not indicated, the fluence rates used were as follows: Rc (63 mmol m 22 s 21 ), FRc (3 mmol m 22 s 21 ), or Bc (16 mmol m 22 s 21 ). * P , 0.01 for hypocotyl lengths of bbx32-1 versus its negative segregant (sib-Col) in B and C and for cotyledon areas of the two BBX32overexpressing lines versus the empty vector control line (EV-Col) in E. Average values are plotted with SE. Data are representative of three biological repeats.
formed a microarray-based analysis of 35S::BBX32 and the bbx32-1 mutant seedlings and included the hy5-1 (Ler background) mutant for comparison. As expected, a large number of genes responded to the red light treatment in wild-type (control) seedlings, including known light-responsive genes such as ELIP1 and SIGE ( Fig. 2A; Supplemental Table S1). BBX32 gene expression was robustly induced by the red light treatment in wild-type seedlings (Supplemental Table S1). The gene expression profile of 35S::BBX32 seedlings showed a marked reduction in responsiveness to light ( Fig. 2A, profile 2). BBX32-overexpressing seedlings exhibited a greater change in the expression of early-light responsive genes in comparison with the hy5-1 mutant seedlings ( Fig. 2A, profiles 2 and 4). There was no significant effect of the bbx32-1 mutation in seedlings treated with 1 h of red light ( Fig. 2A, profile 6). However, a number of light-responsive genes were derepressed in the dark-grown bbx32-1 mutants ( Fig.  2A, profile 5). These data suggested that the native function of BBX32 might be in helping to maintain dark-mediated repression of gene expression and that constitutive expression of BBX32 resulted in the suppression of light-regulated gene expression. There is precedent that the overexpression of an individual gene can result in hyposensitivity to red light due to reduced phyB protein levels in seedlings (Khanna et al., 2007). However, both phyA and phyB proteins were detectable in 35S::BBX32 seedlings at levels similar to those found in EV-Col (Supplemental Fig. S3), indicating that the alterations in light responsiveness in the overexpression line were not due to changes in phyA or phyB protein levels.

Comparison of Genes Targeted by BBX32 and HY5
The two ecotypes (EV-Col and Ler) showed differences in gene responsiveness to treatment with 1 h of red light. About 61% of the induced and 44% of the repressed genes in EV-Col showed a similar response in Ler (Fig. 2B). In dark-grown 35S::BBX32 and hy5-1 mutant seedlings, the expression of a small number of genes was either elevated or reduced by 2-fold or higher (Fig. 2C). There was no overlap between the induced gene sets, and there were only four genes commonly reduced in these two genotypes in the dark (Fig. 2C). The number of genes induced over 2-fold in response to red light in the 35S::BBX32 (63 genes) and hy5-1 mutant (61 genes) was small in comparison with the number of genes that were reduced by 2-fold or higher in these lines, 504 and 172 genes, respectively (Fig. 2C). While only about 20% of induced genes were common between 35S::BBX32 and hy5-1 mutant seedlings, over 73% of the genes that failed to respond to red light in hy5-1 mutants were also reduced in 35S:: BBX32 seedlings (Fig. 2C). These data showed that a majority of genes targeted by HY5 for induction were reduced in 35S::BBX32 seedlings, despite the differences between the two ecotypes. However, a set of genes were reduced specifically in 35S::BBX32 seed-lings: 75% of the genes reduced in the 35S::BBX32 seedlings were distinct from the hy5 set. In summary, the 35S::BBX32 seedlings showed alterations in response to the early light signal, with significant overlap to the response profile of the hy5-1 mutant. However, the 35S::BBX32 seedlings showed an alteration in gene expression distinct from the hy5-1 mutant, as there was Figure 2. Comparison of global gene expression profiles in response to 1 h of red light treatment. A, Each colored bar represents a probe set that was significantly differentially expressed (P BH , 0.05, greater than 1.3-fold change) in response to the light treatment, and probe sets representing early-light-responsive genes (Tepperman et al., 2001) are shown. The various genotypes and light treatments are described on the left. For each genotype, the values were compared with their respective wild-type data. Gene responsiveness to 1 h of red light (light versus dark) for EV-Col and Ler are shown at the bottom of each panel (profiles 7 and 8). As expected, there were some notable ecotypic differences in light responsiveness between the Col and Ler backgrounds. Clusters were created using cosine correlation and Ward's minimum variance. Genes with significantly differential response between each experimental line and its corresponding control seedling are shown in Supplemental Table S1. B and C, Venn diagrams representing set comparisons of differentially expressed loci (P BH , 0.05, greater than 2.0-fold change). B, Genes induced or repressed in EV-Col and Ler in response to 1 h of red light. C, Comparison of genes differentially expressed in 35::BBX32 and hy5-1 mutant seedlings grown in the dark or in response to 1 h of red light treatment. Numbers in parentheses show the total number of genes represented in each set (bottom) and the percentage of those genes that fall in the intersection (top). a discrete set of genes affected only by the overexpression of BBX32.
We subjected the array data to functional category analysis (Gene Ontology [GO] annotations; see "Materials and Methods") to determine the functional pathways most affected in each of the genotypes. Similar functional categories were repressed in both hy5-1 and 35S::BBX32 seedlings, although the magnitude of the reduction was greater in the BBX32 overexpression lines (Fig. 3). In dark-grown hy5-1 mutants and 35S::BBX32 seedlings, the most affected functional categories included response to UV and pigment biosynthesis (Fig. 3, A and C; Supplemental Table S2), albeit there were some differences in individual gene responses within these categories. These differences could represent distinct features of the two genotypes and could also be due to ecotypic differences between the Ler and EV-Col backgrounds. The functional categories most affected after 1 h of red light treatment in both hy5-1 and 35S::BBX32 seedlings were similar, representing genes that are light responsive or are involved in light-regulated processes (Fig. 3, B and D; Supplemental Table S2). However, the order of significance of these functional categories between the two genotypes was different due to the differences in responsiveness of some of the genes. Overall, the 35S::BBX32 seedlings, like the hy5-1 mutant lines, showed a dampening of light-regulated pathways. The effect of the bbx32-1 mutation on gene expression was relatively small, with significant differences observed only in dark-grown seedlings. There was no significant effect on gene expression in the lighttreated bbx32-1 mutant seedlings. On the other hand, a distinct set of functional categories were induced in etiolated bbx32-1 seedlings ( Fig. 3E; Supplemental Table S2). The three most derepressed functional categories in the dark included genes involved in UV and light-responsive pathways followed by categories related to photosynthesis (Fig. 3E). The altered expression of light-related genes in dark-grown bbx32-1 seedlings is consistent with BBX32 playing a native role in dark-mediated gene repression.
We performed quantitative reverse transcription (qRT)-PCR experiments on independently grown seedlings to confirm the array-based results for a selected set of genes. For the qRT-PCR analysis, we included another independently transformed line, 35S::BBX32-487. BBX32 transcript levels were robustly induced by red light treatment in all wild-type (EV-Col, Ler, and sib-Col) control seedlings (Supplemental Fig. S1A). Light induction of BBX32 expression was slightly affected in the hy5-1 mutant; similarly, HY5 transcript levels were reduced in seedlings overexpressing BBX32 but were not affected in the bbx32-1 mutant (Supplemental Fig. S1B; Supplemental Table  S3). We selected about 20 genes for qRT-PCR analysis, either because of their known response to light or based upon differences in expression observed on the array. As a control set, we also selected some that showed no differences on the array. For the complete list of selected genes, primer sequences used, and the qRT-PCR data, see Supplemental Tables S3 and S4. Overall, the results from the qRT-PCR analysis confirmed the array data for the expression of each of these genes in 35S::BBX32 and hy5-1 mutant seedlings (Supplemental Fig. S1, Supplemental Materials and Methods S1, and Supplemental References S1). These data suggested that BBX32 and HY5 proteins function antagonistically in affecting the expression of these genes, including in etiolated seedlings. Arabidopsis seedlings exhibit diurnal hypocotyl growth patterns. Both light and the circadian clock coordinately regulate hypocotyl growth rates during the dark and light cycles. Seedlings grown under short-day (8 h of light/16 h of dark) conditions exhibit reduced hypocotyl growth during the day and increased growth at the end of the night (Nozue et al., 2007). It was shown that hy5 mutants displayed reduced light sensitivity during the day and exhibited higher growth rates in the dark (Nozue et al., 2007). We examined the diurnal growth patterns of 35S::BBX32 seedlings to see if they responded like the hy5 mutants under diurnal conditions and included the bbx32-1 mutants for comparison. There was no detectable difference between the diurnal growth patterns of the bbx32-1 mutant and the sib-Col control seedlings (Fig.  4A). A biphasic pattern of growth was observed in Col seedlings in the dark, marked by increased growth at the start and at the end of the night (Fig. 4B). As expected, the hy5-215 mutants exhibited an increased rate of growth in the dark and maintained a biphasic pattern of growth during the night (Fig. 4B). Like the hy5-215 mutants, both 35S::BBX32 lines showed increased growth during the night, with rates higher than the hy5 mutants (Fig. 4, B and C). These data are also consistent with our observations that the 35S:: BBX32 seedlings show differences in light responses when compared with the hy5 mutants. These results suggested that constitutive expression of BBX32 resulted in alterations in the clock-regulated control of hypocotyl growth during the night, which could be due to direct or indirect effects on clock function (see "Discussion").

BBX32 Interacts with STH2
HY5 is known to interact with other B-box proteins like STH2/BBX21 and LZF1/STH3/BBX22, which function positively in light signaling with HY5; null mutations in STH2/BBX21 and LZF1/STH3/BBX22 genes cause phenotypes similar to the hy5 mutants (Datta et al., 2007(Datta et al., , 2008Chang et al., 2008). BBX32 might either directly interact with HY5 and/or bind one or more of the other B-box proteins that interact with HY5. To test these possibilities, we performed in vitro immunoprecipitation assays and protoplast-based protein-protein interaction assays (Fig. 5). We did not detect any evidence of direct interaction between BBX32 and HY5 in these assays (data not shown). Next, we used STH2:3XHA fusion protein as bait for immunoprecipitation with a rabbit polyclonal anti-HA (for hemagglutinin) antibody; after several . Functional category analysis of genes differentially expressed in 35S::BBX32, hy5, and bbx32-1 lines compared with control lines. Categorizing the biological processes that were significantly affected in the different genotypes and treatments was done by calculating the relative overrepresentation of annotations to different GO terms among the misexpressed genes in darkgrown (A, C, and E) or red light-treated (B and D) 35S::BBX32, hy5-1 mutant, and bbx32-1 mutant seedlings. Based on an analysis (see "Materials and Methods") of GO terms annotated to Arabidopsis genes (Berardini et al., 2004;as of January 19, 2009), the categories listed above were overrepresented (P , 0.01) among significantly misexpressed genes. For each overrepresented GO category, we list the name of the category followed by the number of genes with significantly altered expression and the total number of all genes annotated to the term. The horizontal length of the bar indicates the ratio of the number of genes annotated to the term that have significant expression changes relative to the number of genes from that category that would have been expected to show such changes by chance (based on maximizing the hypergeometric density function), and the vertical length of the bar represents the magnitude of the average absolute fold change of the genes with significantly altered gene expression that were annotated to the given GO term. The percentage of induced and repressed genes annotated to each GO term is indicated by red and green, respectively. For list of genes overrepresented in each of the GO categories, see Supplemental Table S2. washes, each pellet was divided into two equal parts and subjected to western blotting to detect the presence of bait and prey proteins, as described in "Materials and Methods." We used purified 6XHIS:NF-YB2 protein as prey to serve as a negative control, and two other control reactions lacked either the bait STH2: 3XHA or the prey MBP:BBX32. A monoclonal anti-HA antibody was able to detect the bait protein STH2: 3XHA in each pellet, except in the control reaction lacking the bait (Fig. 5A, left panel). The anti-His monoclonal antibody did not detect any 6XHIS:NF-YB2 prey protein in the negative control reaction, but it did detect the 6XHIS:HY5 fusion protein in the pellet with STH2:3XHA, confirming the previously reported interaction between STH2 and HY5 (Fig. 5A, right top panel; Datta et al., 2007). The anti-MBP monoclonal antibody detected a band only in the reaction containing MBP:BBX32 but not in the two controls lacking either the bait or the prey protein (Fig. 5A, right bottom  panel). These data suggested that the MBP:BBX32 fusion was able to bind STH2:3XHA. To exclude the possibility of MBP interaction with the bait protein, Figure 4. Hypocotyl growth curve time course. Hypocotyl growth rate (mm h 21 ) is shown for seedlings grown under 8 h of light and 16 h of dark with 95% confidence intervals (shaded). Hatched areas represent darkness, and white areas represent daytime. Time is shown on the x axis in minutes after dawn on day 4 (time 0). Note that 35S::BBX32-10 seedlings grew continuously during the night, while 35S:: BBX32-12 seedlings displayed a weak biphasic pattern of growth in the dark. The 35S::BBX32-10 seedlings contained higher levels of BBX32 transcript (Supplemental Fig. S5) and consistently had taller hypocotyls compared with the other BBX32-overexpressing lines ( Fig. 1; Supplemental Fig. S5). There is a larger beginning-of-night growth increase here as compared with Nozue et al. (2007), most likely due to differences between growth chambers.
we used an MBP:PARAMYOSIN fusion protein as a negative control. The anti-HA monoclonal antibody detected STH2:3XHA (bait) in the immunoprecipitated pellets from the three separate reactions with or without a prey protein (Fig. 5B, left panel). In these pellets, the anti-MBP monoclonal antibody did not find any significant amount of MBP:PARAMYOSIN or a crossreacting band in the reaction lacking the prey protein; however, abundant MBP:BBX32 fusion protein was present in the pellet, with STH2:3XHA in the corresponding reaction (Fig. 5B, right panel). Therefore, we concluded that BBX32 is able to bind STH2.
We performed protoplast-based protein-protein interaction assays to confirm the in vivo interaction potential between BBX32 and STH2. We used GAL4 DNA binding domain (GD) alone or GD fusions with HY5:cMyc(4X), STH2:cMyc(4X), and BBX32:cMyc(4X) to examine their interactions with HA(2X):STH2 or with chloramphenicol acetyl transferase (CAT) as the negative control (see "Materials and Methods"). Protein expression from these constructs was confirmed by western-blot analysis (Fig. 5C, right panel). There was no increase in GUS activity in negative control protoplasts cotransformed with CAT and GD alone, GD:HY5:cMyc(4X), or GD:BBX32:cMyc(4X) (Fig. 5C). Protoplasts containing CAT with GD:STH2:cMyc(4X) fusion had increased GUS activity (Fig. 5C), suggesting that GD:STH2:cMyc(4X) may have an inherent transcriptional activation potential when bound to DNA. Cotransfection of HA(2X):STH2 with GD: STH2:cMyc(4X) increased GUS activity further, possibly due to intramolecular interaction of STH2 (Fig.  5C). We found that HA(2X):STH2 did not interact with the negative control GD, but it did interact with both GD:HY5cMyc(4X) and GD:BBX32:cMyc (4X) (Fig. 5C). These data confirmed that BBX32 has the capacity to interact with STH2 and are consistent with the possibility that this interaction leads to inactivation of a protein complex containing STH2 and HY5 in vivo.

DISCUSSION
BBX32 was identified as playing a role in lightregulated hypocotyl growth from a large-scale functional genomics screen. Seedlings with constitutive expression of BBX32 were hyposensitive to red, blue, and far-red light. These seedlings exhibited alterations in light-regulated development of hypocotyls and cotyledons, which are characteristic of changes in photomorphogenesis. In contrast, the bbx32-1 mutant seedlings displayed only weak, but partially opposite, phenotypes. There was only one mutant allele (bbx32-1) available for this study; therefore, we performed complementation analysis. The weak phenotype displayed by the bbx32-1 mutant seedlings was rescued by a prBBX32::BBX32 transgene (Supplemental Fig. S2), suggesting that the observed hypersensitivity to light was linked to this locus. Our microarray and qRT-PCR-based gene expression analyses of bbx32-1 mutant seedlings revealed altered expression of light-responsive genes in the dark. In particular, the expression of light-stimulated genes, ELIP, CHS, and SIGE, appeared to be derepressed in dark-grown bbx32-1 seedlings (Supplemental Fig. S1). This lack of gene repression in the dark-grown bbx32-1 mutant seedlings is consistent with the hypothesis that BBX32 acts with other proteins in a complex that functions as a modulator of light-regulated gene expression. Furthermore, a number of genes belonging to GO categories related to photosynthesis or carbon assimilation were induced in the bbx32-1 mutants grown in the dark (Fig. 3). Analysis of additional alleles and multiple mutant combinations with potential functional cofactors will help in confirming the role of the native BBX32 as an accessory protein in modifying gene expression in the dark. Our results with multiple independent transgenic 35S:: BBX32 lines provide evidence that the constitutive expression of BBX32 reduces light responsiveness and promotes gene expression patterns that are characteristic of seedlings grown in darkness.

BBX32 Functions in HY5-Regulated Development
The hy5 mutant was selected for comparison with 35S::BBX32 seedlings because, like hy5 mutants, these seedlings displayed hyposensitivity to multiple light wavelengths, suggesting that BBX32 functions antagonistically to HY5 in light-regulated development, downstream of multiple photoreceptors. We undertook three different approaches to study the possible relationship between BBX32 and HY5: first, we analyzed global gene expression in seedlings in response to early red light treatment; second, we compared the diurnal hypocotyl growth patterns of these seedlings grown under short days (8 h of light/16 h of dark); and third, we examined genetic relationships between BBX32 and HY5 using combinations of genetic mutants and transgenic lines. Our studies with hy5-1/ bbx32-1 double mutants and the 35S::BBX32/35S::HY5 lines showed that the hy5 mutation was epistatic in the  Figs. S4-S6). These results suggested that BBX32 acts in the same light signaling pathway as HY5. Seedlings of the hy5-1/bbx32-1 double mutant appeared to be slightly shorter than the hy5-1 single mutant under blue and white light (Supplemental Figs. S4-S6). This could be due to a possible repression of HYH activity by the native BBX32 protein present in the hy5-1 single mutant, whereas this repression will be absent in the hy5-1/bbx32-1 double mutant. In our microarray analysis, there were ecotypic differences between Ler and EV-Col. Seventy-five percent of all the genes induced by light in Ler were commonly induced in EV-Col (Fig. 2B). The 35S:: BBX32 seedlings displayed a greater magnitude change in gene expression in response to red light than the hy5-1 mutants. Consistent with this observation, the 35S::BBX32 seedlings had longer hypocotyls than hy5-1 mutants under all wavelengths of light (Supplemental Fig. S4). These data support the hypothesis that BBX32 is an accessory protein to other transcription factors, in addition to HY5.

BBX32 Functions in a Regulatory Protein Complex to Modulate Changes in Gene Expression
Previous studies have shown that several B-box proteins, including STH2 and LZF1 (STH3), which both directly interact with HY5, play roles in light signaling (Datta et al., 2007(Datta et al., , 2008Chang et al., 2008;Kumagai et al., 2008). Despite extensive efforts, we did not detect any direct interaction between BBX32 and HY5, nor did we find an interaction of BBX32 with DNA. Our data suggest that it is likely that BBX32 acts Figure 6. Coexpression of 35S::BBX32 reduces the light responsiveness of 35S::STH2 seedlings. Hypocotyl lengths of seedlings from various genotypes grown in darkness (A) or red light (45 mmol m 22 s 21 ; B) are plotted along the left y axis and overlaid with the transcript levels of BBX32 and STH2 relative to a negative segregant (sib-Col) control (log 2 ratio), as measured by qRT-PCR using total RNA extracted from 4-d-old dark-grown seedlings treated to 1 h of red light. Error bars indicate SD in transcript levels. P , 0.01 for hypocotyl lengths in red light of various genotypes versus sib-Col.
as a member of a larger group of proteins modulating the response of the HY5 complex to light signal transduction. Consistent with the above prediction, STH2 was shown to function independently and with HY5 to regulate photomorphogenesis (Datta et al., 2007). It was suggested that STH2 may bind HY5 and possibly other cofactors to regulate transcription to promote light signaling (Datta et al., 2007). We selected STH2 to test whether BBX32 could bind other B-box proteins to form potentially inactive heterodimers. BBX32 interacted with STH2 in both in vitro immunoprecipitation assays and protoplast-based interaction assays (Fig. 5). These results show that BBX32 is capable of direct interactions with other B-box proteins to potentially modulate their activities. Our emerging understanding from this study as well as other work (Datta et al., 2007(Datta et al., , 2008Chang et al., 2008;Kumagai et al., 2008) suggests that B-box proteins as a whole may act to modulate the activity of other B-box proteins as well as the activity of specific transcription factors, allowing the plant to fine-tune a transcriptional response to specific endogenous or environmental inputs. STH2 and HY5 are thought to be targeted for protein degradation through COP1. Future experiments will reveal if BBX32 protein levels are also regulated by COP1.
Under diurnal short-day (8 h of light/16 h of dark) conditions, the 35S::BBX32 seedlings continued to grow during the night; only one of the two independent lines, 35S::BBX32-12, showed a weak biphasic pattern of growth, whereas the hy5-215 mutants showed two peaks of growth during the night, like the wild-type seedlings (Fig. 4). These data suggested that the growth pattern of 35S::BBX32 seedlings at night was less responsive to regulation by the circadian clock than that of the hy5-215 mutant. These results can be explained either by an effect of BBX32 on clock-regulated growth or by indirect effects on clock function due to a greater repression of light signaling in 35S::BBX32 seedlings. The diurnal growth rhythms displayed by the 35S::BBX32 seedlings were similar to the rhythms observed for seedlings overexpressing PIF4 and PIF5 ( Fig. 4; Nozue et al., 2007). An alternative hypothesis is that BBX32 acts in conjunction with one or more of the PIF proteins to regulate hypocotyl growth. However, seedlings overexpressing PIF4 or PIF5 phenocopy phyB mutants (Huq and Quail, 2002;Khanna et al., 2007), and PIF proteins have been shown to promote phyB polyubiquitination and degradation (Khanna et al., 2007;Leivar et al., 2008;Jang et al., 2010). On the other hand, seedlings overexpressing BBX32 phenocopy hy5 mutants and contain phyB protein levels similar to those of wild-type seedlings (Supplemental Fig. S3). BBX32 and PIFs may have overlapping and distinct roles in light signal transduction. Furthermore, the expression of some of the B-box genes (DBB1a, DBB1b, DBB3, STO, and STH1) was shown to be controlled by the circadian clock and to peak at distinct phases, and these genes were implicated in regulating hypocotyl growth (Kumagai et al., 2008). BBX32 may interact with one or more of these B-box proteins to regulate their activities in regulatory protein complexes with or without STH2 and HY5. BBX32 is likely to function as an accessory protein to a protein complex that modulates transcriptional responses downstream of multiple photoreceptors (Fig. 7) and to integrate external and internal cues to promote dark adaptation.
In conclusion, this study provides evidence that BBX32 protein modulates light signaling by acting antagonistically to HY5. Constitutive expression of BBX32 caused hyposensitivity to red, far-red, and blue light. However, the 35S::BBX32 seedlings exhibited a higher magnitude of differences in light-regulated gene expression and hypocotyl growth than the hy5 mutant. By contrast, the single bbx32-1 gene mutation caused weak phenotypic effects and displayed derepression of light-responsive genes in the dark. These results suggest a native role for BBX32 in maintaining dark-mediated patterns of gene expression and hypocotyl elongation. We found that BBX32 is capable of interaction with STH2 and potentially with other B-box proteins. Therefore, it is probable that BBX32 belongs to a regulatory protein complex that acts to modulate light-regulated gene expression early in the hierarchy of light signal transduction.  PCR amplified from an Arabidopsis Col cDNA library containing the entire HY5 open reading frame plus 58 bp of 5#-untranslated sequence and 73 bp of 3#-untranslated sequence. For Arabidopsis overexpression studies, both HY5 and the BBX32 clone were introduced into a standard binary vector harboring a kanamycin selection marker driven by a NOS promoter and the selected coding region downstream from the cauliflower mosaic virus (CaMV) 35S promoter. STH2 was introduced into a similar binary vector except that it contained a sulfonamide selection marker driven by the CaMV 35S promoter. For the creation of the BBX32/STH2 double overexpression lines, the STH2 construct described above was supertransformed (see below) into an established BBX32-overexpressing line (BBX32-10-6). Creation of BBX32/HY5 double-overexpressing lines and isolation of the hy5-1/bbx32-1 double mutant are described in Supplemental Materials and Methods S1. All lines were verified to contain the transgenes by PCR.

Generation of Overexpression Lines
Arabidopsis plants were transformed by the floral dip method (Clough and Bent, 1998;Bechtold et al., 2003). For each of the genes, between 20 and 40 independent primary transformants were isolated on selection medium. Transformants were PCR genotyped to confirm that they harbored the correct transgene, and gene expression was verified by RT-PCR on RNA extracted from 7-d-old seedlings. For each of the genes, one or more representative lines that showed substantial levels of overexpression versus the wild type was selected and used in subsequent experiments. For detailed studies on the effects on BBX32 overexpression, each experiment was performed on one or more of three independently transformed lines (35S::BBX32-10-6, 35S::BBX32-12-2, and 35S::BBX32-487). These three lines exhibited and maintained stable phenotypes over several generations. Based upon genomic Southern-blot analysis, 35S::BBX32-10-6 is predicted to harbor two separate insertions of the T-DNA, while 35S::BBX32-12-2 and 35S::BBX32-487 have multiple copies of the T-DNA (Supplemental Fig. S7).

Isolation of Mutant Alleles and Genetic Crosses
Database searches revealed a single mutant allele (Salk Institute Genomic Analysis Laboratory; Alonso et al., 2003) with a T-DNA insertional disruption of the BBX32 gene, bbx32-1 (SALK_059534), available from the Arabidopsis Biological Resource Center. The insertion site of the disruptive T-DNA is predicted to be downstream of the B-box domain, approximately 670 nucleotides from the start codon at residue 172 of the 225 amino acids constituting the BBX32 protein (Fig. 1A). We used gene-specific (5#-AACTCCACCGC-TCTTTCTCTC-3# and 5#-TTGGATTACCATTATTCCGTTTC-3#) and T-DNA (5#-ATTTTGCCGATTTCGGAAC-3#) primers flanking the predicted insertion site to isolate bbx32-1 homozygous lines. A wild-type sibling (sib-Col) was maintained as the control. In addition, we used two pairs of BBX32-specific primers: one pair targeted the region spanning the predicted T-DNA insertion site (BBX32-A [spanning]), and the other primer pair targeted sequences upstream of the T-DNA (BBX32-B [upstream]; see qRT-PCR section below; for primer sequences, see Supplemental Table S3). We found that the bbx32-1 mutant lacks any detectable BBX32 transcript spanning the predicted T-DNA insertion site; however, with primers targeted upstream of the insertion site, a possibly truncated BBX32 mRNA was detected at levels lower than those in control (sib-Col) seedlings (Supplemental Fig. S1A). Based on these results, it is unlikely that the bbx32-1 mutant expressed any significant quantity of functional full-length BBX32 protein.
Double and triple mutants between bbx32-1, phyA-211, and phyB-9 were selected from F2 populations, created by crossing bbx32-1 to a phyA-211/phyB-9 double mutant in the Col background (a gift from Dr. Peter H. Quail), by a combination of PCR and phenotypic characterization. Using a similar method, double mutants between bbx32-1 and cry1 (CS70; available from the Arabidopsis Biological Resource Center; accession Ler) were identified from F2 segregating populations, PCR genotyped for bbx32-1, and screened in blue light to identify plants containing both mutations. Double mutants between bbx32-1 and sth2-1 (a gift from Dr. Magnus Holm) were made by crossing and PCR genotyping the F2 population for bbx32-1 and sth2-1.

Plant Growth Conditions
For plate-based assays, seeds were sterilized and grown on plates containing growth medium without Suc as described (Khanna et al., 2006). An E-30LED plant growth chamber (Percival) was used for red, far-red, and blue light treatments, and an AR75L chamber (Percival) was used for white light. Fluence rates were measured using a spectroradiometer (EPP2000-VIS-50; StellarNet). Hypocotyl and cotyledon area measurements were performed using digital images (Canon G9) analyzed with ImageJ (version 1.47) software (National Institutes of Health). Approximately 30 seedlings per genotype were used for hypocotyl measurements, and 10 seedlings (20 cotyledons) were used for cotyledon area measurements. Average values were plotted with SE values. All experiments were performed with two or more biological replicates. For soil-grown plants, Arabidopsis seeds were chlorine gas sterilized, cold stratified for 4 d at 4°C, plated on 80% Murashige and Skoog agar containing 0.3% Suc, and transferred to growth chambers. Seedlings were transplanted to soil after 7 d and grown under cool-white fluorescent light at 22°C in Sunshine mix soil.

Microarray Analysis
Seedlings of 35S::BBX32-10-6, bbx32-1, and hy5-1 mutants and their respective controls, EV-Col, sib-Col, and Ler, were grown on plates as described above. Seedlings were grown in the dark for 4 d, and some seedlings were treated with 1 h of monochromatic red light (10 mmol m 22 s 21 ). Three biological replicates, each consisting of approximately 300 seedlings, were harvested for two time points: dark (time 0) and after 1 h of red light treatment. Total RNA was isolated from liquid-nitrogen-pulverized tissue, and biotin-labeled target RNA was produced from 4 mg of total RNA using the MessageAmp II kit from Ambion (catalog no. 1751) according to the manufacturer's instructions. Hybridizations were done according to the standard Affymetrix protocol, with 6.5 mg of labeled copy RNA for each array. To ensure that all samples were high quality, RNA quantity and quality were determined with the Nanodrop 1000 spectrophotometer and the QIAxcel capillary gel electrophoresis system. Arrays were processed and scanned with the Affymetrix GeneChip Workstation (Hybridization Oven 640, Fluidics Station 450, and Scanner 3000) using the GCOS version 1.4.036 software.
All transcript profiling experiments were performed using a custom fullgenome Arabidopsis Affymetrix GeneChip (mbiAth1a520184) microarray designed by Mendel Biotechnology. Based on the TAIR6 (www.arabidopsis. org) November 15, 2005, release of the Arabidopsis genome, this array unambiguously probes 29,544 unique loci. More details about this custom array can be found under Gene Expression Omnibus accession number GPL10037 at the National Center for Biotechnology Information (www.ncbi. nlm.nih.gov/geo).
Microarray data were preprocessed in R (R Development Core Team, 2008), where the data were first background corrected using Robust Multichip Average background correction (Bolstad et al., 2003;Irizarry et al., 2003) and normalized with a quantile normalization. Probe sets were then summarized using a linear probe-level model (Bolstad, 2004), which provided both an estimate of absolute expression level and an estimated error. Both of these data sets were imported into the Rosetta Resolver system (Rosetta Biosoftware) for gene expression data analysis. All derived data sets, including scan profiles, average intensity experiments, and intensity ratios, were analyzed within the Experiment Definition context of the Resolver system. Significance estimates (P values) for intensity ratios defined in this way were subsequently adjusted by the method of Benjamini and Hochberg (1995), resulting in multitest-corrected P BH values. The Affymetrix data files (.CEL), preprocessed profiles, and supplemental files described herein have been deposited at the National Center for Biotechnology Information (www. ncbi.nlm.nih.gov/geo) under Gene Expression Omnibus accession number GSE21174.
The annotation of the Arabidopsis genome (Berardini et al., 2004) to categories in GO (Ashburner et al., 2000) has been used as the basis for overrepresentation analysis of various logical sets of genes. Overrepresentation analysis examines the intersection between a given gene list and genes annotated to all included GO terms and returns the likelihood of observing the resulting degree of overlap. In this study, we defined gene lists based on expression intensity ratios and used a combination of significance threshold (P , 0.01) and fold change, depending on genotype (greater than 1.5-fold for bbx32-1 and greater than 2.0-fold for 35S::BBX32 and hy5-1), as the selection criterion. The calculations for the overrepresentation analysis were performed with freely available software called ermineJ (Lee et al., 2005), and the results were then postprocessed with custom Perl (www.perl.org) and R scripts to calculate and collate additional information about the overrepresented categories. The number of genes expected both to be annotated with a given GO term and to show a significant change in gene expression was calculated based on maximizing the hypergeometric density function given the size of the genome, the number of genes passing our significance parameters in a given ratio, and the total number of genes annotated to the GO category.

Hypocotyl Growth Curve
Hypocotyl growth curve experiments were performed during a diurnal photoperiod, essentially as described (Nozue et al., 2007). Images were collected using a PixeLINK PL-781 camera controlled by National Instruments LABView software. Data analysis was performed in R (R Development Core Team, 2008), as described (Nozue et al., 2007), except that Loess and spline smoothing were used instead of a running average.
Total protein extracts were prepared 16 h after autoinduction at 28°C using a quick-freeze extraction method in the presence of 13 Bugbuster, 7.5 kilounits mL 21 lysozyme, 25 units mL 21 Benzonase (Novagen), and protease inhibitors (Complete Mini Protease Inhibitor tablet; Roche). The MBP:BBX32 fusion protein was purified according to the manufacturer's recommendation (New England Biolabs). The 6XHIS-tagged fusion proteins were purified under native conditions on His-bind Quick 300 cartridges (Novagen). Protein samples were dialyzed for 6 to 12 h (D tubes; Novagen) against a solution of 20 mM HEPES, pH 7.9, 100 mM KCl, 0.5 mM EDTA, and glycerol added to a 30% final concentration prior to storage at 220°C. The enrichment of the fusion proteins was evaluated by SDS-PAGE and western blot using either a mouse anti-6XHIS antibody (Qiagen) or an anti-MBP monoclonal antibody (New England Biolabs), followed by detection using a goat-derived antimouse IgG horseradish peroxidase-conjugated antibody (Thermo Fisher Scientific, Pierce). Total protein concentration was evaluated by bicinchoninic acid (BCA Protein Assay Kit; Novagen).

Protein-Protein Interaction Assays (Immunoprecipitation)
Bait (2 pmol) and prey (4 pmol) proteins, purified from a heterologous bacterial expression system, were incubated at 4°C for 40 min in a Trisbuffered saline binding buffer (500 mL final) containing 5 mM MgCl 2 , 1 mM dithiothreitol, 0.1% Nonidet P-40, and protease inhibitors (including Complete Mini Protease Inhibitor [Roche], 2 mM phenylmethylsulfonyl fluoride, and 1 mM each of aprotinin, leupeptin, and pepstatin). The protein complex was then immunoprecipitated using a rabbit polyclonal anti-HA antibody (sc-805; Santa Cruz Biotechnology) and 30 mL of protein A/G PLUS-agarose beads (sc-2003; Santa Cruz Biotechnology) according to the manufacturer's recommendation. The pellet was washed five times in Tris-buffered saline wash buffer (0.1% Nonidet P-40 with protease inhibitors), and proteins were eluted with 65 mL of 13 sample buffer. A total of 30 mL of eluate from each reaction was used to detect STH2 protein (bait), and 30 mL of the same eluate was used to detect each prey. Western blot analysis was performed using monoclonal anti-HA (Sigma-Aldrich), anti-6XHIS (Qiagen), or anti-MBP (New England Biolabs) as primary antibodies and either a mouse or rabbit horseradish peroxidase-conjugated secondary antibody (Thermo Fisher Scientific, Pierce). Chemiluminescence detection was done using Supersignal western substrates (Thermo Fisher Scientific, Pierce) and captured on x-ray film (GE Healthcare Bio-Sciences, Amersham).

Reporter Gene Construct
The Gal4UAS(4X):GUS reporter gene constructs contained four copies of a Gal4 binding site (CTCCGCTCGGAGGACAGTACTCC) cloned upstream of a CaMV minimal promoter and had the NOS terminator downstream of the GUS reporter gene.

Protoplast Transfection Assays
Isolation of protoplasts from Arabidopsis leaves and transfection have been described previously (Tiwari et al., 2006). Briefly, protoplasts were isolated from 3-to 4-week-old plants grown under long-day conditions (16 h of light and 8 h of dark). Reporter plasmid (10 mg) and effector plasmid (5 mg) were used in transfection assays. Transfected protoplasts were kept in the dark for at least 16 h at room temperature before the GUS activity was measured. All assays were performed in triplicate, and at least two independent transfections were carried out for each experiment. The transfected protoplasts were lysed in 23 sample loading buffer (Sigma S3401), and the expression of effector proteins was confirmed by western blotting using anti-cMYC (Upstate Biotechnology) or anti-HA (Sigma Aldrich) antibodies.
BBX32 sequence data can be found in the Arabidopsis Genome Initiative database (At3g21150).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Array results are highly correlated with results from an independent qRT-PCR experiment.
Supplemental Figure S3. Western-blot analysis of phyA and phyB protein levels in seedlings overexpressing BBX32.
Supplemental Figure S4. Hypocotyl length responsiveness of seedlings correlates to BBX32 expression in various genetic backgrounds.
Supplemental Figure S5. Hypocotyl length responsiveness to red light compared with BBX32 and HY5 gene expression in seedlings.
Supplemental Figure S6. Hypocotyl length responsiveness of wild-type and mutant siblings in the Ler/Col background.
Supplemental Figure S7. Genomic Southern hybridization using pr35S sequence as a probe.
Supplemental Table S1. Genes responding significantly differentially in each of the experimental lines compared with respective control seedlings.
Supplemental Materials and Methods S1. Materials and Methods for Supplemental Figures S1 to S7.
Supplemental References S1. References for Supplemental Figures S1 to S7 and Supplemental Materials and Methods S1.