Hormone-response Genes Are Direct in Vivo Regulatory Targets of Brahma (SWI/SNF) Complex Function*

Metazoan SWI/SNF chromatin remodeling complexes exhibit ATP-dependent activation and repression of target genes. The Drosophila Brahma (SWI/SNF) complex subunits BRM and SNR1 are highly conserved with direct counterparts in yeast (SWI2/SNF2 and SNF5) and mammals (BRG1/hBRM and INI1/hSNF5). BRM encodes the catalytic ATPase required for chromatin remodeling and SNR1 is a regulatory subunit. Importantly, SNR1 mediates ATP-independent repression functions of the complex in cooperation with histone deacetylases and direct contacts with gene-specific repressors. SNR1 and INI1, as components of their respective SWI/SNF complexes, are important for developmental growth control and patterning, with direct function as a tumor suppressor. To identify direct regulatory targets of the Brm complex, we performed oligonucleotide-based transcriptome microarray analyses using RNA isolated from mutant fly strains harboring dominant-negative alleles of snr1 and brm. Steady-state RNA isolated from early pupae was examined, as this developmental stage critically requires Brm complex function. We found the hormone-responsive Ecdysone-induced genes (Eig) were strongly misregulated and that the Brm complex is directly associated with the promoter regions of these genes in vivo. Our results reveal that the Brm complex assists in coordinating hormone-dependent transcription regulation of the Eig genes.

The metazoan SWI/SNF ATP-dependent chromatin remodeling complexes are large (1.2 MDa, 8 -11 subunits) multimeric assemblies that act to modify nucleosome structure, allowing for activation or repression of gene transcription (1,2). Prevailing models suggest that the SWI/SNF complexes are recruited to specific in vivo targets through interactions with DNA-binding transcription factors, where the ATP-dependent activities of the complex assist in gene regulation through changes in DNA-histone contacts (3,4). Subsequently, the affected nucleosomes can be covalently modified to maintain active or repressed transcription through cooperation with histone acetyltransferases and histone deacetylase complexes (5).
Whole genome analyses of yeast and mammalian SWI/SNF complex mutants revealed that the expression of ϳ5% of all genes was affected by removal of the complex (6,7). Importantly, whereas there was a detectable reduction in the expression of a small subset of genes, more transcripts were induced upon loss of the gene encoding the core ATPase (SNF2/SWI2), suggesting that the complex had direct roles in both gene activation and repression. In fact, both epigenetic functions of the yeast complex appear to rely on the ATPase activity of SNF2/ SWI2 (8). Although the yeast SWI/SNF complex is not required for vegetative growth, the metazoan SWI/SNF complex counterparts, including the Brahma (Brm) complex in Drosophila (9,10) and the related mammalian hBrm/Brg1 complexes (11), are essential (12,13). This requirement may be due to direct involvement of the Brm complexes in facilitating global gene expression by RNA polymerase II (14).
Recently, we found that SNR1 functions to mediate aspects of Brm complex transcriptional repression in specific cells through shielding ATP-dependent gene activation, a mechanism that requires the coordinated activities of a tissue-specific transcription repressor and histone deacetylase activity (39). Taking advantage of conditional dominant-negative mutant alleles of snr1 and brm, we showed that cell-type or tissuespecific regulation of Brm complex activities may play an essential role in regulating key developmental growth and patterning processes (16,23,39). This regulation involves epigenetic coregulators, such as transcription factors or enzymes that lead to modification of histones, including histone acetyltransferases or deacetylases to prevent inappropriate expression of critical target genes (23,39,40). The dominant-negative mutant alleles produce proteins that are stable and incorporated into Brm complexes, where the mutant phenotypes arise from loss of subunit function and reflect important contributions of the affected subunit to whole complex activities on in vivo target genes.
Transcriptome profiling arrays have enabled a more detailed view of genes that are normally expressed during development (41,42), or that are misregulated as a consequence of disease (43), immune challenge (44,45), or a specific mutant gene. In the latter case, a loss-of-function mutation in ash2, a member of the Drosophila Trithorax Group (Trx-G) of homeotic (HOM) gene activators, caused widespread disruptions in gene expression that generally correlated with the ash2 mutant phenotypes (46). To help identify bona fide in vivo targets of Brm complex regulation we generated transcriptome profiles associated with loss of specific Brm complex functions at critical points in development. Whereas our results reveal that ϳ3-5% of both individual and functionally related groups of genes are significantly impacted, global gene expression is largely unaffected. We identified the hormone-regulated genes as significant in vivo targets of Brm complex regulation. Cultured cell studies further suggested that the ecdysone-regulated genes, normally expressed late in development around the larval-pupal transition, were directly regulated by the Brm complex, which appears to function as a gene-specific corepressor in the absence of hormone. Moreover, we found that the Brm chromatin remodeling complex can potentially influence gene expression through coregulation in localized genomic regions.

EXPERIMENTAL PROCEDURES
Genetic Manipulations and Strains-Fly strains and manipulations used in this study have been previously described (15,16,39). Flies (snr1 E1 and wild type OregonR) were reared at either 18 or 29°C (permissive or restrictive temperatures for snr1 E1 function), as indicated. Expression of the UAS GAL4 -Brm K804R transgene, herein referred to as UASBrm K804R (47), was controlled using the P(GawB)69B-GAL4 driver (48). Crosses were carried out at 18°C and appropriately staged pupae selected for RNA analysis. Temperature shift experiments with staged early pupae were performed as described (16) using 100 pupae of each genotype reared at 18°C, then shifted to 29°C for 24 h prior to RNA isolation.
RNA Analyses-Appropriately staged wild type, homozygous snr1 E1 and P(GawB)69B-UASBrm K804R pupae (0 -24 h after puparium formation) raised at 18 or 29°C were collected and total RNA prepared using the RNAqueous (Ambion, Inc., Woodlands, TX) extraction system according to the manufacturer's protocols. RNA samples were quantified and analyzed on an Agilent 2100 BioAnalyzer. Equal amounts of RNA were treated with DNase prior to reverse transcription using murine mammary tumor virus reverse transcriptase (RETROscript, Ambion). PCR oligo-primers (MWG Biotech, High Point, NC) were selected using the ABI Prism/Primer Express program (ABI, Foster City, CA). Sequences are available upon request.
Real time PCR were performed using iQ TM SYBR® Green and a Bio-Rad iCycler (Bio-Rad). Results were analyzed using the Genex Macros software (Bio-Rad). Semi-quantitative RT 2 -PCR were carried out using standard protocols with HotStar Taq polymerase (Qiagen Inc., Valencia, CA), as appropriate for each primer pair.
Gene Expression Array Analyses-The microarray experiments conducted in this report utilized the Affymetrix Drosophila GeneChip© for the analysis of gene expression patterns (44). The aRNA amplification and labeling reactions (4 g of RNA per sample) for these experiments were carried out according to the GeneChip© Expression Analysis Technical Manual (Affymetrix publication 701021 Rev. 3). The Gene-Chips were hybridized at 45°C for 16 h with constant rotation (60 rpm), then washed and stained on the Affymetrix Fluidics Station according to the EukGE-WS2v4 protocol. After washing and staining, fluorescent images were scanned at 2-m resolution using the Gene Array Scanner. Affymetrix software (MicroArray Suite 5.0) was used to process the raw images into Cel image files, which were loaded into GeneSpring (Agilent Technologies, Palo Alto, CA) and normalized using the robust multiarray averaging method (49). This robust multiarray averaging normalization was performed on two replicate samples for each condition, and the entire group of arrays was normalized together. The overall correlation between the expression levels for all genes on each array was extremely high (R Ͼ 0.99) for all five sets of replicate arrays (supplemental Table 1). Despite this high degree of reproducibility in replicate arrays, three different methods were used to establish empirical criteria for a gene to be called "changed" in expression in our data (supplemental Tables 2-4). A complete description of the statistical analyses is included as supplementary data. First, we calculated the thresholds for Type I errors seen in replicate arrays by analyzing the distribution of log2 ratios for all the genes in these arrays, and determining the fraction that exceeded an absolute value of 1.0. In the replicate array comparisons, a log2 difference of 1.0 was more than 2 S.D. beyond the 99.9th percentile confidence level (99% confidence level). At this confidence level, no false positives would be expected if the differences in expression were relatively minor and normally distributed. Indeed, none of these comparisons between replicate arrays showed any false positive changes at the 2-fold level. Next, we calculated the expected frequency of Type II errors based on spike-in control data (supplemental Fig. 1). These calculations revealed no Type I or Type II errors were likely included in the changed gene expression data set that we report. Third, overall concordance estimates between the replicate array comparisons were calculated, indicating very clear 99.9% confidence level cutoffs. As these cutoffs approximately equaled a 2-fold change for the comparisons between experimental and control samples, we determined the fraction of genes showing changes in expression of 2-fold or greater in any of the experimental conditions compared with baseline. All comparative data are presented as log2 ratios of the change in the experimental versus baseline condition. Changes in expres-sion of genes that belong to the same functional classes were examined based on current curated Gene Ontology databases available from the NetAffx server.
Cell Culture-Drosophila Schneider (S2) cells were cultured in standard medium. Hormone induction was achieved by addition of 20-hydroxyecdysone (Sigma) to a final concentration of 1 M for 24 h. Approximately 10 6 cells were used for RNA isolation and chromatin immunoprecipitation analysis.
RNAi, Flow Cytometry, and Chromatin Immunoprecipitation-Knockdown of snr1 and brm in cultured S2 cells was performed using RNAi. Double-stranded RNA was prepared using genespecific primers with added T7 polymerase priming sites and the RETROscript kit according to the manufacturer's protocols. dsRNA was added to cultured S2 cells and incubated for 5 days prior to assay or addition of 20-hydroxyecdysone. Knockdown efficiency was assessed by RT-PCR and Western blot using SNR1 and BRM antibodies (15). Following RNAi, cells were centrifuged and resusupended in a buffer containing propidium iodide, then analyzed by flow cytometry using a high speed BD Biosciences FACSAria cell sorter. Chromatin immunoprecipitations were performed using the chromatin immunoprecipitation assay kit according to standard manufacturer's protocols (Upstate Inc., Charlottesville, VA) for rabbit anti-SNR1 and anti-BRM antibodies (15) and primers corresponding to the divergent promoter regions of Eig71Ef-g (f5Ј-cagctttatataagttggaaacaagga-3Ј; r5Ј-tcgtagtatatcgccctgtattt-3Ј); and Eig71Eh-i (f5Ј-caacaaggaagcgaggtctt-3Ј; r5Ј-gaacgtgtcccgtggtattt-3Ј).

Transcriptome Analysis of Brm Complex
Mutants-To identify cellular targets of snr1 and brm function in an unbiased fashion, we examined steady-state gene expression levels of all known and predicted Drosophila genes using RNA isolated from wild type and both snr1 E1 and UAS-Brm K804R mutant flies. Both snr1 E1 and UASBrm K804R function as antimorphic dominant-negative alleles, as they produce defective products that elicit specific phenotypes as a consequence of stable incorporation into Brm complexes (16,47,50,51). Homozygous snr1 E1 and flies ubiquitously expressing UASBrm K804R raised at 18°C are generally viable with discernible and highly penetrant mutant phenotypes. The major temperature-sensitive developmental phase for snr1 E1 is at the larval/pupal transition (Ϯ24 h), whereas snr1 E1 homozygotes (derived from heterozygous parents) raised continuously at 29°C show late pupal lethality (16). Moreover, at the developmental time period and conditions examined, all experimental flies are viable and all snr1 E1 mutant phenotypes are fully rescued with a wild type snr1 transgene (15). Therefore, any significant gene expression changes are a consequence of specific mutant genotype and temperature of incubation and are not likely to result from differences in genetic background or general loss of viability (16,47).
Total RNA was prepared from temperature and stagematched wild type OregonR and mutant (snr1 E1 and UASBrm K804R ) pupae (0 -24 h after puparium formation), labeled and hybridized to Affymetrix Drosophila GeneChip oligo arrays containing 14,010 gene probe sets. Using the highly conservative robust multiarray averaging-normalized average expression as a guide, we found that only 0.7-1.4% (total of 299) of the genes showed mean changes in expression for any of the mutant versus wild type comparisons (Fig. 1).
Similar to swi/snf mutants in Saccharomyces cerevisiae (6,52), more genes appeared to increase in expression rather than decrease in snr1 E1 relative to wild type at the 29°C restrictive temperature (Fig. 1); although, at 18°C approximately the same number of genes showed increased or reduced expression. More genes also showed 2-fold increased expression in snr1 E1 at 29°C relative to 18°C, than showed 2-fold decreased expression. This contrasted with wild type Drosophila, in which more genes show reduced expression at elevated temperatures than are increased. Importantly, these data show that snr1 E1 and UASBrm K804R do not strongly impact global steady-state gene expression under conditions that give rise to clear mutant phenotypes.
The microarray data trends were verified by examining a subset of the affected and unaffected genes by semi-quantitative PCR analyses using the same RNA preparations. Consistent results were found when the gene chip array data were compared with the semi-quantitative PCR data (supplemental Fig. 2 and Ref. 23). Overall, 22 of the 24 genes tested by this method showed similar patterns of increase or decrease relative to the array data under the conditions tested.
Inspection of the genes that were most increased or decreased in any of the three main comparisons between the mutant and wild type strains revealed a number of interesting findings (supplemental Tables 5-7). We focused on the largest changes observed in the UASBrm K804R mutant background (supplemental Table 5), as well as those in the snr1 E1 mutant at both 18 and 29°C (supplemental Tables 6 and 7). SNR1 function is important in restricting cell growth, with adult snr1 E1 flies (Ϫ/Ϫ Ͼ ϩ/Ϫ) raised at 18°C exhibiting significantly increased growth (biomass and cell number) and highly penetrant wing patterning defects (16). The growth defect is manifest at both growth temperatures by late larval development (23), 24 -48 h prior to the experimental time point used for expression analyses in this study. Of the 30 genes showing the largest increase or decrease in expression in the snr1 E1 compared with wild type at 18°C, the majority also showed similar 2-fold increases or decreases in expression at 29°C; similar trends were observed among the decreased genes at the two temperatures. Consistent with snr1 involvement in growth regulation, many of the genes showing increased expression in the mutant have predicted or known functions in apoptosis, hormone binding and signaling, cellular immunity, or contribute to cell structure. Less is known regarding the genes showing decreased expression, although many have predicted or known functions in intracellular protein transport and cell cycle regulation.
Functional Pathway Targets for Brm Complex Chromatin Remodeling-Functional gene groups are those in which the encoded or deduced proteins share similar biological function, enzymatic properties, or have been implicated in a specific developmental pathway. Groupings are commonly assigned using the curated Gene Ontology databases. To determine which functional gene groups might be over-represented in the three different comparisons of interest (snr1 E1 versus wild type at 18 and 29°C, UASBrm K804R versus wild type at 18°C), we used the batch query option to cross-reference the lists of these probe sets with the Affymetrix Gene Ontology information. These tables listed the particular ontologies represented in the list, the number of probes mapped to each ontology cluster on the entire Drosophila array, and the number of probes mapped to each ontology cluster (supplemental Tables 8 -10). Only ontology clusters containing two or more probe sets, with -fold enrichments equal to 5-fold or more, and a calculated p Ͻ 0.05 were considered further. When we examined our data according to functional gene groups, we observed only a small number of groups that showed similar transcriptional effects across the mutant strains under the conditions examined. Rather, most gene groups showed robust changes in only one particular comparison. Some of the gene groups with the largest increases in one or more of the snr1 mutant comparisons included the vitelline membrane formation, mismatch repair, RNA polymerase I transcription factor activity, myosin light chain kinase activity, monophenol monooxygenase activity, vascular endothelial growth factor receptor binding, octopamine receptor activity, and transforming growth factor ␤ receptor binding groups. Gene groups with strongly decreased expression included several involved in protein and RNA metabolism, such as the nascent polypeptide-associated complex, autophagy, proteasome complex, mRNA catabolism, protein-endoplasmic reticulum retention, and leucyl aminopeptidase activity gene groups. These gene functional groups are among those known to be associated with metamorphosis and are regulated in response to rising titers of the hormone 20-hydroxyecdysone (53).
We next analyzed the impact of disrupting Brm complex function on gene expression as a consequence of chromosomal position across the Drosophila genome (Fig. 2). Using the same strategy described above for the Gene Ontology cluster detection, we also examined the total list of genes that were significantly changed in any direction in one or more of the replicate comparisons. This involved a set of 299 unique probe IDs (many transcripts were significantly changed in more than one replicate set of comparisons). Using the previous criteria to describe enrichment (-fold increases greater than 5, p Ͻ 0.05), we observed only a single physical cluster enriched on chromosome 3L. A small number of other clusters were also detected at a threshold of 2-fold enrichment and p Ͻ 0.05 and are listed along with the most robust cluster (supplemental Table 11). We plotted the position of only the stringently (99.9% confidence level) differentially expressed genes for each of the three sets of experiments shown in Fig. 1. This analysis clearly indicated a marked degree of over-representation in the chromosome 3L 16 megabase region that contains a cluster of hormone-regulated genes (Eig) and the brm gene in the SNR1/WT comparison at 29°C and BRM/WT comparison at 18°C. Thus, some genes may be sensitive to reduced Brm complex chromatin remodeling functions as a consequence of chromosomal position or proximity to a common regulated enhancer.
Direct Regulation of Brm-responsive Gene Targets-Individual reductions of Brahma (BRM) ATPase or SNR1 functions acquired through the incorporation of defective subunits (BRM K804R and SNR1 E1 ) lead to opposite growth and patterning phenotypes that are suppressed when both mutant subunits are present in the same complex, indicating that SNR1 serves to regulate the ATP-dependent activities of the Brm complex (16,23,39). We therefore reasoned that genes showing a differential response to reduced SNR1 and BRM function at this developmental stage might be relevant direct targets for Brm complex regulation.
The transcriptome data were analyzed to identify individual genes that displayed dissimilar transcription responses to snr1 and brm mutants (supplemental Tables 5-7). A group of late expressed ecdysone inducible genes (Eig) displayed differential expression associated with reduced SNR1 and BRM function (Fig. 3). These responses are consistent with our observed conditional snr1 E1 mutant phenotypes and developmental requirements for snr1 function at the larval/pupal transition (16,39). The ecdysone-regulated Eig71Eh and Eig71Ei genes showed significantly increased expression (ϳ7-fold) associated with reduced SNR1 function at both the permissive and restrictive temperatures (18/29°C) and decreased expression in response to reduced BRM function (ϳ4 -7-fold) at 18°C (Fig. 3B). These response trends were confirmed by real-time quantitative RT-PCR (Fig. 3C) as well as semi-quantitative RT-PCR (supplemental Fig. 2). The quantitative estimates of expression change using the two independent assays revealed a strong correlation (r ϭ 0.74 -0.91; p Ͻ 0.05). However, expression of genes encoding components of the ecdysone receptor complex (EcR, USP), the EcR corepressor (SMRTR), or ecdysone biosynthetic enzymes were not markedly affected in either snr1 or brm mutants arguing against a general defect in hormone regulation or synthesis (supplementary data). These results are strongly consistent with recent results examining the genomic response to ecdysone signaling at the same developmental stage using a similar transcriptome profile analysis (53).
To test whether Brm complex regulation of the Eig genes was direct or indirect, we employed a cell culture (S2 cell) system to allow for controlled hormone-dependent gene regulation and RNA-mediated gene silencing of snr1 and brm. Use of dsRNA against either snr1 or brm resulted in reducing their individual steady-state mRNA and protein levels (Fig. 4). We further observed that dsRNA directed against snr1 resulted in reduced BRM protein accumulation, and conversely, knockdown of brm resulted in reduced SNR1. Therefore, consistent with genetic studies, reductions of either the SNR1 or BRM core subunits in cultured S2 cells results in decreased stability of the other subunit, and presumably the entire Brm complex. In addition, knockdown of snr1 or brm individually as well as together (Fig.  4C) resulted in increased cell volume and nuclear size. Side scatter analysis revealed that brm RNAi resulted in approximately a 50% increase in cellular complexity relative to the control RNAi cells, which serves as an indirect measure of the change in nuclear size (Fig. 4D). Although the basis of this morphological change is unknown, it was not observed in untreated  S2 cells or in control knockdowns of unrelated genes (Fig. 4, C  and D), and it is consistent with predicted roles of INI1 and SNR1 in controlling cell cycle progression and actin cytoskeleton organization through regulation of the retinoblastoma-E2F and Rho pathways (54).
The Eig genes are part of a cluster of 10 coordinately regulated ecdysone-responsive genes located at cytological position 71E on the third chromosome (55). These genes encode for small cysteine-rich peptides that are related to vertebrate defensins, possibly involved in antimicrobial protection (55). The expression of the Eig genes during the prepupal period is tightly coupled to rising titers of ecdysone hormone at the larval-pupal transition during development through direct activation by the Broad-complex (BR-C) and E74A transcription factors (56). The Eig genes are grouped into five divergently transcribed pairs with short intergenic regions (ϳ260 bp) that comprise the transcription regula-tory elements (Fig. 5A). These genes are not normally expressed in the embryonic-derived S2 cells, nor are they significantly induced within 24 h following addition of 20-OH ecdysone (10 Ϫ6 M) to the culture medium (Fig. 5B), despite the presence of the EcR/USP hormone receptor, as well as the E74A and BR-C activators (Fig. 5B). However, hormone-dependent expression of the Eig genes was observed 48 h following hormone addition (Fig. 5B), 3 suggesting that delayed transcription of the Eig genes may reflect an important feature of a coordinated ecdysone-dependent transcriptional cascade.
To address whether the Brm complex might be directly involved in the regulation of the Eig genes through repression as suggested from the microarray analysis, we employed RNAi-mediated knockdown of both snr1 and brm in S2 cells. Following RNAi-mediated knockdown of snr1 or brm or both simultaneously we observed a rapid hormone-dependent induction of the Eig genes within 24 h (Fig. 5C). The inductive effect was maintained for at least 6 days, suggesting that the ongoing high level expression of the Eig genes was not dependent on subsequent Brm complex function. As controls we measured the expression of other ecdysone-regulated genes, including Eip55E, Eip74EF, Eip71CD, and Eip75B, and found that these were unaffected by knockdown of either snr1 or brm in cultured S2 cells. 3 Thus, consistent with our expression profiling microarray data obtained from the snr1 E1 mutant, the Brm complex may contribute to transcriptional repression of the Eig genes in the absence of hormone and assist in coordinating the ecdysone-stimulated transcription cascade.
We next examined whether the Brm complex directly regulates expression of the Eig genes by testing for chromatin binding. Formaldehyde cross-linked bulk chromatin was prepared from cultured S2 cells in the absence of ecdysone, sheared by sonication, and immunoprecipitated with antibodies directed against SNR1 or BRM (15). Promoter-specific primers corresponding to the divergently transcribed Eig71Ef-g and Eig71Eh-i genes were used for amplification (Fig. 5A). Both the SNR1 and BRM proteins were found associated with the Eig 3 C. B. Zraly and A. K. Dingwall, unpublished results. A, RT-PCR analysis of snr1 and brm transcripts following RNAi in cultured S2 cells. dsRNA (30 g each) was added to 10 6 cells for 6 days prior to analysis of steady-state transcript levels (C, control dsRNA). B, SNR1 and BRM protein levels following RNAi knock-down were measured by Western blot. Note that SNR1 and BRM show some decrease when the other subunit is reduced, but not when an unrelated (control) gene was reduced by RNAi. C, cells exhibit altered cellular and nuclear morphology upon dsRNA knockdown of snr1, brm, or both genes together. D, side scatter analysis of propidium iodide (PI)-negative S2 cells following RNAi to either a control gene or brm. gene promoter sequences (Fig. 5D). Furthermore, RNAi-mediated knockdown of either snr1 or brm resulted in loss of the Brm complex from the Eig gene promoters (Fig. 5D). Because we observed delayed expression of the Eig genes in S2 cells following the addition of hormone, we next asked whether this was inversely correlated with Brm complex binding. Consistent with our hypothesis that the Brm complex was associated with Eig repression, we observed reduced BRM binding to the Eig71Eh-i promoter 3 days after addition of ecdysone. 3 Differential Brm Complex Regulation of Adjacent Genes-One possible explanation for coregulation of the Eig genes is their physical proximity on the chromosome, such that factors influencing the expression of one pair might "spread" to neighboring genes. Our microarray data are consistent with this view, as significant coregulation of expressed discrete loci is observed along the lengths of the individual chromosome arms (Fig. 2). These data plots suggest that chromatin remodeling factors, such as the SWI/SNF or Brm complexes, might indirectly influence the expression of some genes through localized or restricted affects on chromatin structure associated with nearby regulated target genes.
The Eig gene cluster also includes an additional embedded gene (Eig71Ee) that is not coordinately regulated with the adjacent divergently transcribed loci (55). Eig71Ee (also known as I71-7 (57)) is normally expressed during late third instar larvae (intermolt) prior to the adjacent Eig71E genes. The deduced Eig71Ee protein is strongly related to Drosophila glue proteins (e.g. SGS-3 and SGS-4) and has also been implicated in antibacterial and immune responses (58). Surprisingly, our expression profiling and RT-PCR data indicate that Eig71Ee expression is increased in the UASBrm K804R mutant and decreased in the snr1 E1 mutant; thus, regulation by the Brm complex is opposite to the surrounding Eig genes (supplemental Tables 5-7; Fig. 3). Temperature shift experiments using our snr1 E1 allele were employed to verify the role of SNR1 in regulating Eig71Ee gene expression. Flies, both wild type and snr1 E1 mutant, were raised at the permissive temperature (18°C) then shifted to the restrictive (29°C) for 24 h during the temperature-sensitive period for SNR1 function at the larval-pupal transition (16). Effects of the temperature shift on expression of the Eig genes was nearly identical to the results from growth at constant temperature, except for Eig71Ee that showed complete loss of expression in the snr1 E1 mutant upon temperature shift (supplemental Fig. 2). 3 One possibility is that Eig71Ee is positively regulated by the Brm complex during late third instar larval development, with subsequent repression during early pupariation (the developmental time period corresponding to our expression array samples). In this scenario, the Eig71Ee gene would likely be repressed by ATPase-dependent chromatin remodeling coincident with up-regulation of the nearby Eig genes. The Eig71Ee gene is not expressed in cultured S2 cells under any of our experimental conditions (data not shown), suggesting that positive regulatory factors controlling Eig71Ee transcription are absent in the embryonic-derived S2 cells.

DISCUSSION
The metazoan SWI/SNF complexes are essential for growth and development. In flies, loss of the Brm complex is lethal, although reduced complex functions are associated with a variety of growth and patterning defects. In mammals, early wide- spread loss of the Brg1 complex is embryonic lethal, whereas loss of some core components in specific tissues is often associated with a variety of cancers and may be linked to several other human diseases. As a first step in understanding the critical roles chromatin remodeling plays in development, it is essential to identify the genes that are targets for complex functions. Genetic screens in Drosophila using conditional or dominant-negative mutant alleles of genes encoding the Brm complex components have provided important insights; whereas analysis of tumors or tumor-derived cell lines in humans and mammalian knock-out models of a few genes encoding subunits of the complex have added significant, although limited understanding to how certain genes are targeted and regulated.
Brm Complex Function in Repressing Eig Gene Transcription-Studies of our conditional snr1 E1 mutant revealed critical requirements for SNR1 and Brm complex functions in Drosophila wing patterning and growth control during late larval and early pupal development, coincident with the last major proliferation and cellular determination events prior to the emergence of adults (16,23,39). The snr1 E1 allele is an antimorph that exhibits both temperature-and dosage-dependent mutant phenotypes that can be dominantly enhanced and suppressed by heterozygous mutations in other Brm complex components. Importantly, the snr1 E1 phenotypes are suppressed by reducing Brm complex function with mutant brm alleles (39), suggesting that the phenotypes arise from misregulated Brm complex activities. Conversely, mutant phenotypes arising from expression of the dominant-negative UASBrm K804R are suppressed by the snr1 E1 allele. Based on the reciprocal suppression of mutant phenotypes associated with specific dominant-negative snr1 and brm alleles, we reasoned that important in vivo gene targets would likely display opposite transcription profiles in response to reduced function of a specific Brm complex subunit.
Based on molecular and genetic studies, we previously proposed a model of Brm complex-dependent gene repression that depended on constraint of ATP-dependent chromatin remodeling mediated through interaction of SNR1 with gene-specific repressors and histone deacetylase activities (23,39). Transcriptome profiling using the same dominant-negative antimorphic Brm complex mutations has now allowed us to identify a subset of hormone-responsive genes as potential direct targets of Brm complex-dependent repression during the developmental period defined as the most sensitive to loss of SNR1 function (16). The Eig genes are among the late responding gene targets in the ecdysone pathway (55,56) and they exhibited increased expression in a snr1 E1 mutant, whereas their expression was decreased in a UASBrm K804R mutant background. To verify the mutant transcriptome data and explore potential mechanisms, we employed a Drosophila cell culture system where the Eig genes are not normally expressed without addition of exogenous hormone. Despite the presence of the EcR/USP ecdysone receptor, as well as the E74A and BR-C activators that are rapidly induced coincident with rising hormone titers (59), the late ecdysone responsive Eig genes were not transcribed until 2 days following addition of hormone in S2 cells, suggesting that they were actively repressed. Consistent with our profiling data, knockdown of the Brm complex using RNAi resulted in the rapid induction of Eig gene transcription, indicating that the normally delayed expression was at least partly dependent on constraining Brm complex activity.
The results of our combined genetic, transcriptome array, and cell culture data supports the view that the repression of the Eig genes is not a consequence of Brm complex ATPase activity; rather, the coordinated expression of the late hormone response genes may be in part due to constraining chromatin remodeling at the late gene promoters. In this scenario, late gene targets of the ecdysone regulatory hierarchy are repressed following the initiation of the regulatory cascade, perhaps to delay expression of certain late genes until completion of the pathway requires their expression. An interesting possibility is that an early ecdysone-response protein functions as a genespecific repressor until the pathway reaches the appropriate point in the hierarchy to activate the late genes, perhaps through a switch from repressor to activator function. Thus, at the appropriate point in the cascade, the repression is released and the late genes are transcribed. Although the identities of these factors are uncertain, we have not observed obvious changes in promoter acetylation status following Brm complex knockdown, 3 suggesting that histone deacetylase activity is not likely to be involved. Unlike the situation we observed in wing patterning where genes must be kept repressed for proper vein development, the Eig genes are expressed at a later step in the cascade and would not be expected to be epigenetically silenced. Rather, the delayed transcription may be a consequence of targeted repression, perhaps as one critical component of the typical hormone response pathway. The identities of these factors and their mechanisms of action are the subject of our ongoing investigations.
Implications for Controlling Growth and Cancer Development-The metazoan SNF5-related genes are essential for normal development and survival. INI1/hSNF5 is the mammalian counterpart of Drosophila SNR1. INI1/hSNF5 is a potent tumor suppressor as loss of heterozygosity is strongly correlated with aggressive rhabdoid tumors of the kidneys, spleen, and liver, as well as aggressive brain (AT/RT) tumors in infants and young children (30,60). Significantly, these aggressive cancers have not been associated with loss of any other SWI/SNF complex subunit (61). Mammalian studies using rhabdoid tumor cell lines or mouse INI1/hSNF5 knock-out models have implicated widespread inappropriate expression of cell cycle regulators as a critical misregulated process eventually leading to aggressive cancers (25-27, 29, 62, 63). A recent report using rhabdoid tumor-derived cell lines that lack INI1/hSNF5 suggested that it may not be required for all BRG1-dependent transcription activation functions (64). Inactivation of INI1/hSNF5 in the developing murine liver using Cre-mediated recombination results in increased proliferation and defects in glycogen storage (7). Following INI1/hSNF5 inactivation, transcriptome analysis revealed defective activation of ϳ70% of the genes normally up-regulated during liver development, including genes involved in glycogen metabolism and cell-cell adhesion. Many of the genes showing reduced expression are functionally similar to genes that show reduced expression in the Drosophila UASBrm K804R mutants (this study), suggesting that the misregulated genes are dependent on ATPase-chromatin remod-eling for their transcription activation. Thus, the mammalian INI1/hSNF5 knock-out studies may reflect loss of the entire complex rather than revealing unique and important functions of the INI1/hSNF5 subunit.
The metazoan hormone-responsive genes are activated or repressed at specific developmental times and in discrete tissues, requiring highly coordinated regulation. Numerous reports have documented direct associations between the SWI/ SNF-related ATP-dependent chromatin remodeling complexes and mammalian nuclear receptors, including the androgen, estrogen, thyroid, retinoic acid, peroxisome proliferatoractivated receptor ␥, and glucocorticoid receptors, as well as the ecdysone receptor in Drosophila. Importantly, all major subclasses of nuclear receptors in mammals have counterparts in Drosophila (65), suggesting that the relationships between chromatin remodeling and nuclear receptor function are well conserved features of metazoan development. Recent work has also shown that the ISWI-based NURF complex in Drosophila is important to mediate some aspects of ecdysone-dependent gene induction (66), suggesting multiple roles for chromatin remodeling in hormone-regulated development. Our present findings reveal that the hormone-regulated genes may be among the most critical SWI/SNF chromatin remodeling complex targets and raise an interesting possibility that these targets may be misexpressed in aggressive rhabdoid tumors.