Genetic Analysis of brahma: The Drosophila Homolog of the Yeast Chromatin Remodeling Factor SWI2/SNF2

Corresponding author: John W. Tamkun, 350 Sinsheimer Labs, Department of Biology, University of California, Santa Cruz, CA 95064, tamkun{at}biology.ucsc.edu (E-mail).

Communicating editor: S. HENIKOFF

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The Drosophila brahma (brm) gene encodes an activator of homeotic genes related to the yeast chromatin remodeling factor SWI2/SNF2. Here, we report the phenotype of null and dominant-negative brm mutations. Using mosaic analysis, we found that the complete loss of brm function decreases cell viability and causes defects in the peripheral nervous system of the adult. A dominant-negative brm mutation was generated by replacing a conserved lysine in the ATP-binding site of the BRM protein with an arginine. This mutation eliminates brm function in vivo but does not affect assembly of the 2-MD BRM complex. Expression of the dominant-negative BRM protein caused peripheral nervous system defects, homeotic transformations, and decreased viability. Consistent with these findings, the BRM protein is expressed at relatively high levels in nuclei throughout the developing organism. Site-directed mutagenesis was used to investigate the functions of conserved regions of the BRM protein. Domain II is essential for brm function and is required for the assembly or stability of the BRM complex. In spite of its conservation in numerous eukaryotic regulatory proteins, the deletion of the bromodomain of the BRM protein has no discernible phenotype.

THE homeotic genes of the Antennapedia complex (ANTC) and the bithorax complex (BXC) encode transcription factors that specify the identities of body segments in Drosophila (DUNCAN 1987 ; KAUFMAN et al. 1990 ). Since alterations in cell fate result from either the inactivation or derepression of homeotic genes, their transcription must be regulated precisely to ensure normal development. During much of Drosophila development, the transcription of homeotic genes is controlled by two distinct groups of regulatory genes: the Polycomb group of repressors and the trithorax group of activators (KENNISON 1993 , KENNISON 1995 ; SIMON 1995 ). Although the mechanism of action of most Polycomb and trithorax group members is not well understood, several are thought to regulate transcription by altering chromatin structure. For example, the Polycomb (PC) protein contains a short segment—the chromodomain—which is conserved in a component of Drosophila heterochromatin, the HP1 protein (PARO and HOGNESS 1991 ). Based on this similarity, it has been proposed that PC represses transcription by packaging inactive homeotic genes into heterochromatin-like complexes. brahma (brm), a trithorax group gene, was identified in a screen for dominant suppressors of Pc mutations (KENNISON and TAMKUN 1988 ). A direct connection between the regulation of homeotic gene expression and chromatin was provided by the discovery that BRM is strikingly related to SWI2/SNF2, a chromatin remodeling factor in the yeast Saccharomyces cerevisiae (TAMKUN et al. 1992 ).

SWI2/SNF2 is a subunit of a 2-MD complex, the SWI/SNF complex, that assists a wide variety of sequence-specific transcription factors to activate the transcription of their target genes (WINSTON and CARLSON 1992 ; CARLSON and LAURENT 1994 ; PETERSON and TAMKUN 1995 ). Both BRM and SWI2/SNF2 contain sequence motifs closely related to those found in DNA-stimulated ATPases; these motifs are clustered in a 500–amino acid segment known as the ATPase domain (HENIKOFF 1993 ). Purified yeast SWI/SNF complex has ATPase activity and stimulates the binding of GAL4 derivatives and other transcription factors to nucleosomal DNA in vitro (COTE et al. 1994 ; OWEN-HUGHES et al. 1996 ). swi2/snf 2 mutations are suppressed by mutations in nucleosomal histones and cause alterations in chromatin structure in the vicinity of the SUC2 promoter in vivo (HIRSCHHORN et al. 1992 , HIRSCHHORN et al. 1995 ; PRELICH and WINSTON 1993 ). The SWI2/SNF2 protein thus appears to use the energy of ATP hydrolysis to counteract the repressive effects of chromatin on transcription.

A large number of other eukaryotic proteins are related to BRM and SWI2/SNF2 within the ATPase domain (EISEN et al. 1995 ). Several of these ATPases, including the yeast STH1 and Drosophila ISWI proteins, are subunits of chromatin remodeling complexes that are distinct from SWI/SNF (PETERSON 1996 ; PAZIN and KADONAGA 1997 ). For example, Drosophila ISWI is a subunit of at least three different protein complexes: the 500-kD nucleosome remodeling factor (NURF; TSUKIYAMA et al. 1995 ), the 670-kD chromatin accessibility complex (CHRAC; VARGA-WEISZ et al. 1997 ), and the 220-kD ATP-utilizing chromatin assembly and remodeling factor (ACF; ITO et al. 1997 ). The yeast STH1 protein is a subunit of a 1-MD complex known as RSC (remodels the structure of chromatin; CAIRNS et al. 1996 ). Although SWI/SNF, RSC, NURF, CHRAC, and ACF each catalyze ATP-dependent alterations in chromatin structure in vitro, differences in their biochemical properties and subunit composition suggest they are likely to play distinct roles in chromatin remodeling in vivo.

Several lines of evidence suggest that BRM and SWI2/SNF2 play similar roles in chromatin remodeling. First, the similarities between the two proteins extend well beyond the ATPase domain and define three additional conserved regions: domain I, domain II, and the bromodomain (TAMKUN et al. 1992 ; Figure 1). These similarities are likely to be significant since the regions flanking the ATPase domain are thought to contribute to the functional specificity of SWI2/SNF2 family members by mediating interactions with other proteins (LAURENT et al. 1993 ; TREICH et al. 1995 ; TREICH and CARLSON 1997 ). Biochemical studies of BRM and its putative human homologs, BRG1 (brm-SWI2/SNF2 related gene 1; KHAVARI et al. 1993 ) and HBRM (human brm; MUCHARDT and YANIV 1993 ), have provided further evidence that brm and SWI2/SNF2 are functional homologs. BRM is a subunit of a 2-MD complex that contains SNR1, the putative Drosophila homolog of the SNF5 subunit of the yeast SWI/SNF complex (DINGWALL et al. 1995 ). The human BRG1 and HBRM proteins are also subunits of 2-MD complexes that contain several proteins related to yeast SWI/SNF proteins (MUCHARDT et al. 1995 ; WANG et al. 1996A , WANG et al. 1996B ). Like SWI/SNF, the BRG1 complex facilitates the binding of transcriptional activators to nucleosomal DNA in vitro (KWON et al. 1994 ; WANG et al. 1996A ). These findings strongly suggest that brm and its vertebrate homologs encode catalytic subunits of conserved, higher eukaryotic counterparts of the yeast SWI/SNF complex.

View larger version (17K): [in this window] [in a new window]   Figure 1. —Schematic comparison of wild-type and mutant BRM proteins. The four regions that are highly conserved in BRM and its putative yeast and human homologs are represented by shaded boxes [domain I, diagonal lines; domain II, wavy lines; domain III (DNA-dependent ATPase domain), stippled; domain IV (the bromodomain), horizontal lines]. The positions of the bipartite ATP-binding site (GKT and DEGH) and other highly conserved blocks of sequence within the ATPase domain (HENIKOFF 1993 ) are marked by black boxes. The sequences altered in the BRM549-610, BRMK804R, and BRM1446-1517 proteins are compared to the sequence of the BRM protein. Regions of the BRM protein contained in the GST-BRM fusion proteins used as immunogens are underlined.

Despite these similarities, there are striking differences in the composition of the yeast SWI/SNF complex and its counterparts in higher eukaryotes. Multiple BRG1 and HBRM complexes are present in human cells, and the subunit composition of these complexes varies in different cell types (WANG et al. 1996A ). These differences may reflect specialized roles of the BRG1 and HBRM complexes in development or other processes. Indeed, studies of BRG1 have revealed potential roles in viral integration, cell cycle control, and cancer (KALPANA et al. 1994 ; DUNAIEF 1994; STROBER et al. 1996 ).

What have previous genetic studies of brm revealed about the role of chromatin remodeling factors in multicellular eukaryotes? brm is an essential gene that is expressed both maternally and zygotically (TAMKUN et al. 1992 ; BRIZUELA et al. 1994 ). Loss of maternal brm function blocks oogenesis; individuals homozygous for extreme brm alleles die as late embryos with no obvious pattern defects (BRIZUELA et al. 1994 ). Since it has not been possible to generate embryos lacking both maternal and zygotic brm function, the exact role of brm in embryonic development is not clear. Information concerning the role of brm after embryogenesis has been derived primarily from the analysis of hypomorphic brm alleles. Individuals trans-heterozygous for certain combinations of brm alleles survive to adulthood and exhibit developmental abnormalities similar to those arising from reduced expression of ANTC and BXC genes, including the transformation of first legs to second legs and the fifth abdominal segment to a more anterior identity (BRIZUELA et al. 1994 ). Because the effect of complete loss of brm function had not been examined, it was unclear whether brm is also involved in other processes.

To clarify the role of brm in Drosophila development, we examined the spatial and temporal expression of the BRM protein and used mosaic analysis to determine the null phenotype of brm mutations. As an alternative approach, we used site-directed mutagenesis to generate dominant-negative brm mutations and investigate the functions of evolutionarily conserved domains within the BRM protein.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Stocks:
Flies were raised on a cornmeal-molasses-yeast-agar medium containing Tegosept and propionic acid at 25°. Mutations and chromosome aberrations are described in LINDSLEY and ZIMM 1992 unless otherwise noted. brm¹ and brm² are described in KENNISON and TAMKUN 1988 and BRIZUELA et al. 1994 . Df(3L)th102 deletes polytene chromosome region 72A1;72D12, including brm. The FLP and FRT stocks (XU and RUBIN 1993 ), UAS-lacZ reporter 4-2-4B (BRAND and PERRIMON 1993 ), and IJ3 and 69B GAL4 insertion lines (BRAND and PERRIMON 1993 ) used in this study were obtained from the Bloomington Stock Center (Indiana University, Bloomington, IN). The e16E GAL4 insertion line is described in HARRISON et al. 1995 . y w P[ry⁺, hsFLP]12 was generously provided by T.-B. CHOU and N. PERRIMON.

Production of antibodies against the BRM protein:
Polyclonal rabbit antisera were raised against glutathione S-transferase (GST) fusion proteins containing amino acids 1504 –1638 or 505–775 of the BRM protein (Figure 1). The first fusion protein was produced by subcloning a 423-bp Sau3A fragment from brm cDNA6 (TAMKUN et al. 1992 ) into the BamHI site of pGEX3X (Pharmacia LKB, Piscataway, NJ). To produce the second fusion protein, the polymerase chain reaction (SAIKI et al. 1988 ) was used to amplify an 816-bp fragment from brm cDNA 1' (TAMKUN et al. 1992 ) subcloned in pGEM7ZF using a T7 promoter primer and a primer (5'-TGGAATTCC TTGAGCGTA-3') that converts nucleotides 2377–2382 of the brm cDNA into an EcoRI site. The amplified fragment was cloned into the EcoRI site of pGEX-KG (GUAN and DIXON 1991 ). GST-BRM fusion proteins were expressed in Escherichia coli, purified on glutathione-agarose columns, and used to immunize rabbits as described in HARLOW and LANE 1988 . Immune sera were affinity purified on columns containing either GST or GST-BRM fusion proteins coupled to Affigel 10 or Affigel 15 resins (Bio-Rad, Richmond, CA) according to the manufacturer's instructions. Antibodies were eluted with high salt, dialyzed, and assayed for specificity by Western blotting.

Rabbit antibodies against the BRM protein were used to stain whole-mount preparations of 0–22-hr Drosophila embryos as described in CARROLL and SCOTT 1985 . 12CA5 mouse monoclonal antibody (Babco, Berkeley, CA) was used to detect BRM proteins bearing an epitope tag from the influenza hemagglutinin protein (HA; WILSON et al. 1984 ; KOLODZIEJ and YOUNG 1991 ). Goat or mouse anti-rabbit secondary antibodies conjugated to horseradish peroxidase (Bio-Rad) were used to detect the bound primary antibody. To localize the BRM protein in larval tissues, third instar larvae were dissected, fixed, and stained as described by PATTATUCCI and KAUFMAN 1991 .

Clonal analysis of brm mutations:
Clones of homozygous mutant brm tissue were generated in heterozygous larvae by mitotic recombination using the FLP-FRT technique (GOLIC 1991 ; XU and RUBIN 1993 ). As an internal control, we compared the frequency, size, and phenotype of brm clones in the presence and absence of a wild-type brm transgene carried on the second chromosome (P[w⁺, brm⁺]21A). The FRT at the base of 3L at position 80B was recombined onto the brm² st cp in ri p^p chromosome. y w P[ry⁺, hsFLP]12/Y ; P[w⁺, brm⁺]21A/+; P[ry⁺, y⁺](66E) P[ry⁺, FRT](80B)/+ males were crossed to y w ; brm² P[ry⁺, FRT](80B)/TM3, Ser virgin females, and second instar larvae were exposed to a 90-min heat shock to induce expression of FLP recombinase. Body parts of experimental [y w P[ry⁺, hsFLP]12/y w ; P[ry⁺, y⁺](66E) P[ry⁺, FRT](80B)/brm² P[ry⁺, FRT](80B) and control (y w P[ry⁺, hsFLP]12/y w ; P[w⁺, brm⁺]21A/+; P[ry⁺, y⁺](66E) P[ry⁺, FRT](80B)/brm² P[ry⁺, FRT](80B)] adults were dissected in 70% ethanol, mounted in methyl salicylate and Canada balsam, and examined by bright-field microscopy. Clones of homozygous brm² tissue were identified in the adult cuticle by the loss of the y⁺ gene on the P[ry⁺, y⁺](66E) P[ry⁺, FRT](80B) chromosome.

Site-directed mutagenesis and generation of transgenic strains:
Two brm genomic DNA fragments were used as the templates for site-directed mutagenesis. The first was a 14.4 -kb BamHI-EcoRI fragment (BR14.4) that spans the entire brm gene (TAMKUN et al. 1992 ). The second was a modified version of this DNA fragment (BR14.4T) that encodes a BRM protein with a C-terminal tag from the influenza HA protein followed by six histidines (DINGWALL et al. 1995 ). P-element transgenes encoding either untagged (P[w⁺, brm⁺]21A) or tagged (P[w⁺, brm⁺]92C) BRM protein rescue the recessive lethality of extreme brm alleles (BRIZUELA et al. 1994 ; DINGWALL et al. 1995 ).

To generate the brm^1446-1517 mutation lacking the bromodomain, we PCR-amplified fragments from a subcloned 3.6-kb BamHI-EcoRI fragment of BR14.4 using primers that hybridize to either the T7 or SP6 promoters flanking the cloning site and the mutagenic primers 5'-CGGACAGAGCTCGTCCAT-3' or 5'-ACTCGAGAGCTCTGCAAAAG-3'. These primers introduce a SacI site at the site previously occupied by the bromodomain. After cleavage with SacI, the amplified fragments were ligated together, substituted for the 3.6-kb BamHI-EcoRI genomic DNA fragment in BR14.4, and subcloned in the P-element transformation vector pCaSpeR (PIRROTTA 1988) to generate P[w⁺, brm^1446-1517].

The ATP-binding site mutation brm^K804R and domain II deletion brm^549-610 were generated by site-directed mutagenesis (KUNKEL et al. 1985) of a 1.3-kb BamHI-EcoRI fragment from BR14.4 using the mutagenic oligonucleotides 5'-TGAATG GTTCTACCCAAAC-3' and 5'-GGTCGTCCTTGTGCTGGT GATTCATCACGGC-3', respectively. The mutant fragments were used to replace the corresponding wild-type fragments within BR14.4T. The resulting fragments were subcloned in pCaSpeR to generate P[w⁺, brm^K804R] and P[w⁺, brm^549-610]. Transgenic strains bearing the P[w⁺, brm^1446-1517], P[w⁺, brm^K804R] or P[w⁺, brm^549-610] transgenes were generated by P-element–mediated transformation as described in TAMKUN et al. 1991 . To control against errors introduced during PCR amplification or site-directed mutagenesis, all relevant regions were sequenced on one strand (SANGER et al. 1977 ) before transformation.

Functional analysis of brm transgenes:
An insertion of P[w+, brm^K804R] on the second chromosome (P[w⁺, brm^K804R]17D) was tested for the ability to rescue the hemizygous lethality of brm². w/Y ; P[w⁺, brm^K804R]17D/+; Df(3L)th102, h ri Sb ca²/+ males were mated to virgin w/w ; brm²/ln(3LR)DcxF, ru h D females and the progeny were scored for w ; P[w⁺, brm^K804R]17D/+; brm²/Df(3L)th102, h ri Sb ca² adults. Two independent insertions of the P[w⁺, brm^1446-1517] transgene on the second chromosome (22-1 and 13-1) were tested for the ability to rescue brm² mutations. For both insertions, w; P[w⁺, brm^1446-1517]/+; brm² st cp in ri p^p/TM3, Sb virgin females were crossed to w ; P[w⁺, brm^1446-1517]/+; brm² st cp in ri p^p/TM3, Sb males, and the progeny were scored for P[w⁺, brm^1446-1517]/+; brm² st cp in ri p^p/brm² st cp in ri p^p adults.

Two independent insertions of the P[w⁺, brm^549-610] transgene on the X chromosome (24 -1 and 4 -3) and a single insertion on the second chromosome (3-3) were tested for the ability to rescue the hemizygous lethality of brm². For both insertions on the X chromosome, w P[w⁺, brm^549-610]/Y; Df(3L)th102, h ri Sb ca²/+ males were crossed to w ; brm² st cp in ri p^p/TM3, Ser virgins, and the progeny were scored for w P[w⁺, brm^549-610]/w ; Df(3L)th102, h ri Sb ca²/brm² st cp in ri p ^p adults. To test the ability of four copies of the P[w⁺, brm^549-610] transgene to rescue the hypomorphic brm¹ allele, w P[w⁺, brm^549-610]4-3 P[w⁺, brm^549-610]24-1/Y ; mwh brm¹ st cu sr e ^s ca/ln(3LR)CxD males were crossed to w P[w⁺, brm^549-610]4-3 P[w⁺, brm^549-610]24-1/+; mwh brm¹ st cu sr e^s ca/ln(3LR)CxD females and the progeny were scored for w P[w⁺, brm^549-610]4-3 P[w⁺, brm^549-610]24-1; mwh brm¹ st cu sr e^s ca/mwh brm¹ st cu sr e ^s ca progeny.

As a positive control for all of the above experiments, we showed that a wild-type brm transgene (P[w⁺, brm⁺]21A) was able to rescue the homozygous and hemizygous lethality of brm¹ and brm². No homozygous or hemizygous brm adults were observed in the progeny of the above crosses in the absence of a rescuing brm transgene.

Generation and analysis of a GAL4 -responsive brm^K804R transgene:
Using brm cDNA clones and the BR14.4T brm genomic DNA fragment, we created a 6.7-kb DNA fragment that contains the coding sequence for the HA-tagged BRM protein flanked by 66 nucleotides of 5' genomic DNA and ~1.3 kb of 3' genomic DNA, including the brm polyadenylation site. This fragment extends from the PstI site 66 nucleotides upstream of the brm initiation codon to the BamHI site at coordinate -22 of a chromosome walk through the brm-arl region (TAMKUN et al. 1992 ), but lacks the introns within this region of genomic DNA. A 0.76-kb Nsl I/HindIII genomic DNA fragment within this fragment (which encodes residues 682–937 of BRM) was replaced with the corresponding fragment of the brm^K804R transgene described above. The resulting fragment was placed under the control of the GAL4 UAS and minimal hsp70 promoter in the P-element transformation vector pUAST (BRAND and PERRIMON 1993 ). A homozygous viable insertion of this transgene (P[w⁺, UAS_GALhsp70:brm^K804R]2-2) on the third chromosome was generated by P-element transformation, as described in TAMKUN et al. 1991 .

To induce expression of the transgene, homozygous y Df(1)w67c2; P[w⁺, UAS_GALhsp70:brm^K804R]2-2 virgins were crossed to males bearing the GAL4 insertions 69B, IJ3, or e16E. As a negative control in these experiments, virgin females homozygous for a P-element insertion bearing a UAS-lacZ reporter gene (4 -2-4B; BRAND and PERRIMON 1993 ) were crossed to homozygous 69B males in parallel. To determine the lethal phase and phenotype of individuals expressing the BRM^K804R protein under the control of the 69B GAL4 insertion, eggs were collected on grape juice plates for 3 hr at either 20° or 25°. 36 hr later, the number of unhatched eggs were counted to score embryonic lethality, and first instar larvae were transferred to vials at the same temperature. The number of pupae and adults per vial were counted to assess larval and pupal lethality. At least 150 embryos were collected at each temperature for both the control and experimental crosses.

Electrophoresis and Western blotting:
SDS-polyacrylamide gel electrophoresis and Western blotting were performed as described previously (TSUKIYAMA et al. 1995 ). Drosophila embryos, larvae, pupae, or adults were homogenized in boiling electrophoresis sample buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 0.72 M ß-mercaptoethanol, 10% glycerol), boiled for an additional 10 min, and centrifuged at 10,000 rpm for 10 min at room temperature. Embryos were dechorionated in 50% sodium hypochlorite before homogenization. Protein was quantitated using the Bio-Rad protein assay, and 30 µg total protein extract was loaded per lane.

Gel filtration chromatography:
To prepare nondenatured protein extracts, 0–12-hr embryos were dechorionated and homogenized in an equal volume of ice-cold extraction buffer (50 mM Hepes, pH7.6, 385 mM NaCl, 0.1% Tween 20, 0.1 mM EGTA, 1.1 mM MgCl₂, 100 µg/ml PMSF, and 1 µg/ml each aprotinin, leupeptin, chymostatin and Pepstatin A) using a Dounce homogenizer. The homogenate was centrifuged for 1 hr at 55,000 g. After addition of glycerol (to 10% v/v), the supernatant was aliquoted and frozen in liquid nitrogen. Before gel filtration chromatography, extracts were passed over a G-25 column and eluted in column buffer (50 mM sodium phosphate, pH 7.8, 400 mM NaCl, 0.05% Tween 20, 0.1 mM EGTA, 1.0 mM MgCl₂, 0.1 mM DTT, 10% glycerol, and 1 µg/ml each aprotinin, leupeptin, chymostatin, and Pepstatin A). 2 mg of extract was diluted to 200 µl with column buffer and fractionated on a Superose 6 10/30 FPLC column (Pharmacia) equilibrated in column buffer. 0.5-ml fractions were collected and analyzed by Western blotting.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Developmental expression of the BRM protein:
Previous studies revealed that the expression of brm RNA is both temporally and spatially regulated (TAMKUN et al. 1992 ; ELFRING et al. 1994 ). brm RNA is expressed at high levels throughout the early embryo but becomes restricted to the ventral nerve cord and brain by stage 15 of embryogenesis. No brm transcripts are detected between stage 16 of embryogenesis and hatching. Low levels of brm RNA are observed in larvae, pupae, and adult females, and no brm RNA has been detected in adult male flies. Based on the restricted pattern of brm RNA expression, it has been proposed that brm is unlikely to play a general role in transcription (TAMKUN et al. 1992 ; ELFRING et al. 1994 ). To further examine this issue, we characterized the expression of BRM protein during Drosophila development.

Rabbit polyclonal antisera were raised against different GST-BRM fusion proteins (Figure 1). One of the fusion proteins contains a region extending from within the bromodomain to the C terminus of BRM (residues 1504–1638). The other fusion protein contains a region of BRM (residues 505–776) that spans domain II and extends into the ATPase domain. Whole-sera and affinity-purified antibodies against both fusion proteins, but not preimmune sera, detect a 200-kD protein on Western blots of Drosophila extracts (Figure 2). This molecular weight is very close to that predicted for BRM based on its sequence (~185 kD). These antibodies (but not preimmune sera) also detect the BRM protein in whole-mount preparations of both embryos and larvae (Figure 3).

View larger version (32K): [in this window] [in a new window]   Figure 2. —Developmental expression of the BRM protein. Thirty micrograms each of protein extracted from Drosophila melanogaster embryos (0–3, 3–6, 6–12, 12–18, and 18–22 hr); first (L1), second (L2), and third (L3) instar larvae, pupae (P), adult males (M), and females (F) were electrophoresed on an 8% SDS-polycacrylamide gel and analyzed by Western blotting. BRM protein levels are highest in embryos and pupae; much lower levels of BRM are present in larvae and adults.

View larger version (140K): [in this window] [in a new window]   Figure 3. —Expression of the BRM protein in embryos and imaginal discs. Whole-mount preparations of Drosophila embryos or larvae were stained with affinity-purified antibodies directed against the BRM protein. (A) Late syncytial embryo. (B) Lateral view of stage 15 embryo. (C) Ventral view of stage 17 embryo. (D) Prothoracic leg disc. (E) Mesothoracic leg disc. (F) Metathoracic leg disc. The BRM protein is expressed in nuclei throughout the developing embryo and larva.

Unlike brm RNA, BRM protein is present at all stages of development, as revealed by Western blotting (Figure 2). BRM protein is expressed at relatively high levels throughout embryogenesis and in pupae; lower amounts of BRM are present in larvae and adult flies. We investigated the level of BRM protein in developing embryos by quantitative Western blotting using a GST-BRM fusion protein as a standard. Approximately equivalent immunoreactivity was observed with 2 ng of purified fusion protein and protein extracted from 40 3–6-hr embryos (data not shown). Since ~6,000 nuclei are present at this stage of development, we estimate that there are at least 100,000 molecules of BRM protein per nucleus at its peak stage of expression. This level of expression corresponds to approximately one molecule of BRM protein per 20 nucleosomes, contrasting sharply with the relatively low abundance of SWI2/SNF2 in yeast cells (approximately several hundred molecules per nucleus; COTE et al. 1994 ).

We also examined the spatial expression of BRM protein by immunostaining whole-mount preparations of embryos and larvae. BRM protein is present at similar levels in nuclei throughout the early embryo (Figure 3A). The BRM protein continues to be expressed ubiquitously during the remainder of embryogenesis, although its levels are somewhat enriched in the ventral nerve cord and brain in late embryos (Figure 3, Figure B and Figure C). In late third instar larvae, BRM protein is expressed at relatively uniform levels in nuclei of the imaginal discs (Figure 3, D–F) and other diploid and polytene tissues, including the polytene nuclei of the salivary gland (data not shown). Thus, in contrast to the previously reported patterns of brm RNA expression, the BRM protein is ubiquitously expressed throughout the developing organism.

Somatic clonal analysis of brm mutations:
Since brm mutations are recessive lethal mutations, we used somatic clonal analysis to analyze the role of brm after embryogenesis. Homozygous clones of brm mutant tissue marked with yellow were generated in second instar larvae using the FLP-FRT technique (GOLIC 1991 ; XU and RUBIN 1993 ). A strong brm allele, brm², was used for the majority of these experiments; evidence that brm² is a protein null allele is presented below. As an internal control for effects on cell viability or division, we compared the frequency, size, and phenotype of clones of brm mutant tissue in the presence or absence of a wild-type brm transgene (P[w⁺, brm⁺]21A) on the second chromosome.

A total of 716 mutant clones were scored and compared with 1396 clones generated in control siblings (Table 1). The size and frequency of the clones in control and experimental individuals revealed that loss of brm activity has dramatic effects on cell viability in the imaginal discs; the size and frequency of experimental clones in the head and thoracic segments were significantly reduced relative to the controls (Table 1). In contrast, the frequency and size of control and experimental clones in the abdomen were similar (Table 1). These data indicate that brm is essential for the development of imaginal tissues but not abdominal histoblasts. It is also possible that sufficient brm RNA or protein persisted after clone induction to allow the development of the abdominal segments.

Examination of the phenotype of brm² clones revealed unanticipated defects in the adult peripheral nervous system. The mechanosensory bristles of brm² clones in the head, thoracic, and abdominal segments were either duplicated, stunted, or fused (Table 1; Figure 4A). In many cases, the sockets were also malformed, absent, or duplicated (Figure 4A). These defects were not observed in clones generated in individuals bearing a brm⁺ transgene, indicating that they are caused by the brm² mutation and not another mutation on the brm²-FRT chromosome. No homeotic transformations were observed in experimental clones in any of the abdominal segments. The small number and size of experimental clones in the head and thoracic segments did not allow us to score alterations in cell fate arising from loss of brm activity in these segments.

View larger version (86K): [in this window] [in a new window]   Figure 4. —Bristle defects associated with brm mutations. (A) Somatic clone of homozygous brm2 tissue in an abdominal segment marked with yellow. Mechanosensory bristle defects associated with loss of brm function include malformation and loss or twinning of shafts and/or sockets. (B) Expression of the BRMK805R protein under the control of the 69B GAL4 driver causes twinning of mechanosensory bristles. Both the socket and shaft are duplicated in a bristle marked by an arrow in an abdominal segment of a P[w+, UASGALhsp70:brmK804R]2-2/P[w+, hsp70:GAL4]69B adult.

Generation and analysis of a dominant-negative brm mutation:
As an alternative method for studying the role of brm in Drosophila development, we generated and analyzed dominant-negative brm mutations. This approach was based on previous studies of the yeast SWI2/SNF2 and human BRG1 genes, which showed that mutations in the ATP-binding site of both proteins eliminate their activity without affecting their ability to interact with other proteins (KHAVARI et al. 1993 ; PETERSON et al. 1994 ; COTE et al. 1994 ). The mutant proteins thus have dominant-negative effects on transcription when expressed in yeast or human cells. We reasoned that similar brm mutations could be valuable tools for analyzing brm function in vivo.

We previously showed that a transgene encoding a BRM protein bearing an epitope tag from the influenza virus HA protein at its C terminus could rescue the recessive lethality of brm mutations (DINGWALL et al. 1995 ). We used site-directed mutagenesis to create a mutation in this transgene that replaces the conserved lysine (amino acid 804) in the ATP-binding site of the BRM protein with an arginine (Figure 1). Transgenic strains bearing homozygous viable insertions of the mutant transgene on either the X or second chromosome (P[w⁺, brm^K804R]22D and P[w⁺, brm^K804R]17D, respectively) were generated by P-element–mediated transformation. The brm^K804R transgene is expressed at levels comparable to a brm⁺ transgene, as assayed by probing Western blots of proteins extracted from P[w⁺, brm^K804R]22D or P[w⁺, brm⁺]92C embryos with a monoclonal antibody against an HA epitope tag (data not shown). However, the P[w⁺, brm^K804R]17D transgene was unable to rescue the hemizygous lethality of brm², indicating that the ATP-binding site of brm is critical for its function in vivo.

To determine whether the brm^K804R mutation interferes with the assembly of the BRM complex, we examined the native molecular weight of the BRM^K804R protein in embryos using gel filtration chromatography. Proteins extracted from P[w⁺, brm^K804R]22D or P[w⁺, brm⁺]92C embryos were fractionated on a Superose 6 FPLC column, and the eluted fractions were assayed for BRM protein by Western blotting (Figure 5). The apparent native molecular weights of the BRM^K804R and BRM proteins are identical (~2 MD), indicating that the BRM^K804R protein is efficiently incorporated into the BRM complex.

View larger version (35K): [in this window] [in a new window]   Figure 5. —Gel filtration analysis of wild-type and mutant BRM proteins. Whole embryo extract (2 mg) was applied to a Superose 6 gel filtration column; 0.5-ml fractions were collected and assayed for BRM protein by Western blotting using a monoclonal antibody against an HA epitope at the C-terminus of the BRM proteins. The elution volumes of protein standards are marked by arrows. The BRM and BRMK804R proteins elute from the column with apparent native molecular weights of ~2 MD. Note that the average molecular weight of the BRM549-610 protein is slightly smaller than that of BRM protein.

Since BRM^K804R is incorporated into the BRM complex, it should have dominant-negative effects on brm function in vivo. To test this possibility, we examined the effect of varying the relative dosage of the brm^K804R transgene and the wild-type brm gene from 0:2 to 2:1 (Table 2). As anticipated, the BRM^K804R protein interferes with the function of the endogenous BRM protein. Individuals bearing one or two copies of the brm^K804R transgene are viable, but frequently exhibit partial transformations of haltere to wing, as evidenced by an increase in haltere size and the appearance of ectopic bristles on the capitellum (Figure 6). Approximately one third of P[w⁺, brm^K804R]22D adults exhibit this transformation, which is presumably caused by the decreased expression of the Ultrabithorax (Ubx) gene. Increasing the ratio of BRM^K804R to BRM to 2:1 was lethal. This effect was revealed when we tried to generate adult flies heterozygous for brm² or a brm deficiency [Df(3L)th102] that had two copies of a brm^K804R transgene (P[w⁺, brm^K804R]17D) on the second chromosome (Table 2). Thus, the brm^K804R mutation behaves as an antimorphic allele of brm.

View larger version (110K): [in this window] [in a new window]   Figure 6. —Expression of the BRMK804R protein in imaginal discs causes homeotic transformations. (A) Wild-type haltere. (B) Haltere from a P[w+, brmK804R]22D male. A partial transformation of haltere to wing is evident from the increased size of the haltere and appearance of ectopic bristles on the capitellum.

We next examined the effect of expressing high levels of the BRM^K804R protein in restricted temporal and spatial patterns using the GAL4 system of BRAND and PERRIMON 1993 . A gene encoding HA-tagged BRM^K804R protein was placed under the control of a GAL4-regulated promoter in the vector pUAST (BRAND and PERRIMON 1993 ), and a homozygous viable insertion of the dominant-negative transgene on the third chromosome (P[w⁺, UAS_GALhsp70:brm^K804R]2-2) was generated by P-element–mediated transformation. Expression of this trangene was induced by crossing this transformant to strains that express GAL4 in a variety of different patterns. Expression of the dominant-negative BRM protein in patterns identical to the segmentation genes hairy (using the GAL4 insertion IJ3; BRAND and PERRIMON 1993 ) or engrailed (using the GAL4 insertion e16E; HARRISON et al. 1995 ) had no effect on embryonic viability or segmentation (data not shown). The lack of an embryonic phenotype resulting from embryonic expression of the BRM^K804R protein may be caused by the high maternal expression of wild-type BRM protein, which is sufficient to allow embryogenesis to proceed to near completion in the absence of zygotic brm function. Expression of the BRM^K804R protein in imaginal tissues after embryogenesis led to greatly reduced viability. For example, expression of BRM^K804R under the control of the GAL4 insertion 69B (BRAND and PERRIMON 1993 ) caused high levels of lethality (>99%). At 25° or higher, the majority of individuals expressing BRM^K804R pupated but failed to differentiate.

The GAL4 system is inherently temperature sensitive; higher levels of induction are observed at elevated temperatures (BRAND et al. 1994). When the temperature at which crosses were maintained was reduced, we could increase the number of individuals expressing the BRM^K804R that survived to adulthood. This allowed us to score adult phenotypes resulting from expression of BRM^K804R under the control of the 69B GAL4 insertion. Individuals reared at 20° displayed partial transformation of first leg to second leg, as evidenced by a reduction in the number of sex comb teeth on the first leg (data not shown). This phenotype is also seen in adults trans-heterozygous for hypomorphic brm alleles (BRIZUELA et al. 1994 ), and it is presumably caused by decreased expression of the Sex combs reduced (Scr) gene. Adults reared at 20° also displayed twinning of mechanosensory bristles (Figure 4B), a phenotype similar to that observed in clones of brm² tissue.

Expression of the BRM^K804R protein induced by the 69B driver at 20° also had dramatic effects on the size and morphology of the wing; mutant wings were reduced in size, and the L5 and the posterior cross-vein (PCV) were usually absent (Figure 7, A and B). Defects in the campaniform sensilla, a class of sensory organs important for flight, were also observed with high frequency because of expression of the BRM^K804R protein in imaginal discs (Figure 7). These defects fell into four classes: missing sensilla, duplication or triplication of sensilla, transformation of sensilla into bristles, and the appearance of ectopic sensilla (Table 3; Figure 7). Ectopic sensilla and bristles were observed most frequently on the L3 vein. Three sensilla (L3-1, L3-2, and L3-3) and no bristles are normally found on this vein. By contrast, approximately one-half of mutant wings displayed one or two additional sensilla on L3. Ectopic bristles were observed on this vein in approximately one-fifth of mutant wings.

View larger version (103K): [in this window] [in a new window]   Figure 7. —Wing defects associated with expression of the BRMK804R protein. Wings were dissected from control (P[w+, UASGALhsp70:brmK804R]2-2/P[UAS-lacZ 4-1-2]; A, C, and E) or mutant (P[w+, UASGALhsp70:brmK804R]2-2/P[w+, hsp70:GAL4]69B; B, D, F, and G) individuals. (A) Normal wing. The positions of the campaniform sensilla are marked by arrows. (B) Mutant wing. Note the reduced size of the wing, the disruption of the L5 vein, and the absence of the posterior cross-vein (PCV). The positions of the campaniform sensilla along the L3 vein (including one ectopic sensillum) are marked by arrows. (C and D) Magnified views of the proximal and distal twin sensilla of the margin (p-TSM and d-TSM) of wings from normal (C) or mutant (D) individuals. Note the transformation of the d-TSM to a bristle in D. (E and F) Magnified views of the anterior cross-vein sensillum (ACV) in wings from normal (E) or mutant (F) individuals. Note the duplication of the ACV sensillum in F. (G) Magnification of the L3 vein of a wing from a mutant individual. An ectopic bristle distal to the L3-3 sensillum is marked by an arrow.

Functional analysis of evolutionarily conserved domains of the BRM protein:
Comparison of the sequences of the BRM protein and its putative homologs in yeast and humans have suggested that they contain at least four functional domains. The largest of these domains, the ATPase domain, is highly conserved in all members of the SWI2/SNF2 family of ATPases (EISEN et al. 1995 ). The regions flanking this domain are thought to contribute to the functional specificity of individual SWI2/SNF2 family members by mediating interactions with other proteins. Three short segments outside the ATPase domain—domain I, domain II, and the bromodomain—are conserved in the yeast SWI2/SNF2 protein, the Drosophila BRM protein, and the human BRG1 and HBRM proteins (TAMKUN et al. 1992 ; KHAVARI et al. 1993 ; MUCHARDT and YANIV 1993 ). The conservation of these domains from yeast to humans suggests that they are critical for the function of the BRM protein. To test this possibility, we examined the consequences of deleting two of these conserved regions of the BRM protein: domain II and the bromodomain.

The bromodomain is dispensable for brm function:
To investigate the function of the bromodomain, we created a transgene encoding a BRM protein lacking amino acids 1446–1517, including the 46–amino acid segment (residues 1461–1505), which is most highly conserved in other bromodomains (Figure 1). Two independent, homozygous, viable insertions of this transgene on the second chromosome (P[w⁺, brm^1446-1517]22-1 and P[w⁺, brm^1446-1517]13-1) were tested for the ability to rescue brm mutations. Surprisingly, the bromodomain appeared to be dispensable for zygotic brm function; both brm^1446-1517 transgenes were found to rescue the hemizygous lethality of an extreme brm allele, brm². P[w⁺, brm^1446-1517]; brm²/Df(3L)th102 individuals are viable, phenotypically normal, and fertile.

If the BRM protein consists of independent functional domains, interallelic complementation could account for the ability of P[w⁺, brm^1446-1517] to rescue the brm² mutation. This concern was justified, since brm² had not been shown to be a protein null allele, and previous genetic studies have suggested that BRM may act as a multimer (BRIZUELA et al. 1994 ). To investigate this possibility, we examined whether brm² is a protein null allele by Western blotting. The difference in size between BRM^1446-1517 and BRM is sufficient for the two proteins to be resolved by electrophoresis on low-percentage SDS-polyacrylamide gels (Figure 8, lane 3). Polyclonal antibodies against two different regions of the BRM protein (residues 505–776 and 1504–1638) detected both the wild-type and BRM^1446-1517 proteins in adults bearing the P[w⁺, brm^1446-1517]22-1 transgene (Figure 8, lane 3). By contrast, these antibodies detect only the smaller BRM^1446-1517 protein in P[w⁺, brm^1446-1517]22-1; brm²/Df(3L)th102 adults (Figure 8, lane 4). Similar results were obtained using the P[w⁺, brm^1446-1517]13-1 transgene. These results demonstrate that the brm² allele encodes either no protein, or a severely truncated protein lacking residues 505–1638 of BRM, including the entire ATPase domain, the nuclear localization signal and the bromodomain. We therefore conclude that brm² is a null allele, and that the bromodomain is dispensable for both maternal and zygotic brm function.

View larger version (39K): [in this window] [in a new window]   Figure 8. —Analysis of BRM and BRM1446-1517 proteins by Western blotting. Extracts from wild-type (lanes 1 and 5); P[w+, brm+]21A (lane 2); P[w+, brm1446-1517]22-1 (lane 3) and P[w+, brm1446-1517]22-1; brm2/Df(3L)th102 females were fractionated on a 5% SDS-polyacrylamide gel and analyzed by Western blotting using a polyclonal antibody against the BRM protein. Antibodies against BRM recognize both the BRM protein and the smaller BRM1446-1517 proteins in transgenic flies bearing the P[w+, brm1446-1517] transgene (lane 3). In P[w+, brm1446-1517] adults hemizygous for the brm2 allele, only the BRM1446-1517 protein is detected (lane 4).

Domain II is required for the assembly of the BRM complex:
We next examined the function of domain II of the BRM protein. This 62–amino acid domain of BRM (residues 549–610) is located N-terminal to the ATPase domain, and is 48% identical to the corresponding region of SWI2/SNF2 (TAMKUN et al. 1992 ). This domain is also conserved in the putative human homologs of BRM, BRG1, and hbrm (83% identity to BRM in both proteins; KHAVARI et al. 1993 ; MUCHARDT and YANIV 1993 ). Using site-directed mutagenesis, we generated a transgene encoding an epitope-tagged BRM protein (BRM^549-610) that lacks the 61 amino acids between residues 549 and 610, including all of domain II (Figure 1). A homozygous viable insertion of the brm^549-610 transgene on the X chromosome (P[w⁺, brm^549-610]4 -3) was unable to rescue the hemizygous lethality of brm². Similar results were obtained with two independent transformants(P[w⁺, brm^549-610]24 -1 and P[w⁺, brm^549-610]3-3), suggesting that domain II is essential for the activity of the BRM protein.

Why does brm^549-610 fail to rescue brm mutations? Deletion of domain II has a reproducible effect on the level of BRM present in the Drosophila embryo. Using Western blotting, we found that the levels of epitope-tagged BRM^549-610 protein in P[w⁺, brm^549-610]4 -3 embryos were fourfold lower than the level of epitope-tagged BRM protein in P[w⁺, brm⁺]92C embryos (data not shown). Similar results were obtained using two independent transformants of brm^549-610 (P[w⁺, brm^549-610]24 -1 and P[w⁺, brm^549-610]3-3). These results suggested that the failure of domain II to rescue brm mutations might be caused by decreased levels of BRM protein, as opposed to the elimination of a function specific to domain II. However, four copies of a P[w⁺, brm^549-610] transgene were unable to rescue the recessive lethality of brm¹, a hypomorphic brm allele, indicating that domain II is essential for brm function.

To determine if domain II is required for assembly into the BRM complex, we examined the native molecular weight of the epitope-tagged BRM^549-610 protein in extracts prepared from P[w⁺, brm^549-610]3-3 embryos. Although no monomeric BRM^549-610 protein was present in these embryos, a significant proportion of the mutant BRM protein has a native molecular weight (~700 kD) smaller than that of the wild-type BRM protein (Figure 5). This size difference was also observed in extracts prepared from an independent transformant, P[w⁺, brm^549-610]24-1, and was confirmed by re-probing the Western blots with polyclonal antisera that recognize both the BRM^549-610 and endogenous BRM proteins. Although we did not examine the subunit composition of the wild-type and mutant complexes, these results indicate that domain II contributes to the formation or stability of the BRM complex.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Our characterization of null and dominant-negative brm mutations confirms and extends previous studies which showed that brm plays an important role in the control of cell fate (KENNISON and TAMKUN 1988 ; TAMKUN et al. 1992 ; BRIZUELA et al. 1994 ). Partial loss of brm function leads to a variety of homeotic transformations, including the transformations of haltere to wing and first leg to second leg. These transformations are identical to those associated with mutations in the Ubx and Scr genes, respectively. Complete loss of brm function was lethal, however, indicating that the function of brm is not limited to homeotic gene regulation. Although we are not certain how many genes require brm for their expression, the ubiquitous expression of high levels of the BRM protein (approximately one molecule of BRM per 20 nucleosomes in embryos) suggests that it may play a fairly general role in transcription or other processes.

The effect of brm mutations on the number, position, and identities of adult sensory organs indicates that brm plays an unanticipated role in the development of the adult peripheral nervous system. The mechanosensory structures of the peripheral nervous system consist of four unique cells derived from a common sensory organ precursor (SOP) cell (JAN and JAN 1993 ). Division of the SOP cell gives rise to two daughter cells, IIa and IIb. The IIa cell gives rise to the trichogen (shaft) cell and the tormogen (socket) cell; the IIb cell gives rise to the neuron and thecogen (sheath) cell. Some of the mechanosensory bristle defects observed in clones of mutant brm tissue, including abnormal shaft size and morphology and malformed or missing sockets, may be secondary consequences of decreased cell viability. Other defects associated with the brm² and brm^K804R mutations, including the appearance of ectopic or duplicated sensory organs, may result from alterations in cell fate. For example, the transformation of the IIb cell into a IIa cell would lead to the twinning of both bristle and socket at the expense of the neuron and sheath cells. Alternatively, the duplication of sensory organs may be caused by extra divisions of the SOP cell or its descendants resulting from loss of brm function.

How might brm control the fate or proliferation of the cells that form mechanosensory organs? The simplest explanation is that brm regulates the activity of one or more neurogenic genes. Mutations in several neurogenic genes cause sensory bristle abnormalities similar to those associated with brm mutations. For example, loss of Notch function early in pupal development causes a proliferation of SOP cells, resulting in the formation of supernumerary bristles and a low frequency of stunted, fused, and composite bristles (HARTENSTEIN and POSAKONY 1990 ). Mutations in Suppressor of Hairless (Su(H)) cause transformations of socket to shaft, presumably caused by increased expression of Hairless (H; SCHWEISGUTH and POSAKONY 1994 ). Mutations in twins cause duplication of both sockets and bristles (SHIOMI et al. 1994 ). Interactions between brm and one or more of these genes could account for the defects observed in brm clones.

Another possibility has been suggested by recent studies demonstrating that the human BRG1 and HBRM proteins cooperate with retinoblastoma tumor suppressor proteins to regulate cell cycle progression. Physical association between Rb family members and the BRG1 and HBRM proteins have been demonstrated using both the yeast two-hybrid system and coimmunoprecipitation assays (DUNAIEF et al. 1994 ; SINGH et al. 1995 ). Rb, but not BRG1, is present in a human tumor cell line, SW13; transfection of BRG1 or HBRM into these cells causes a dramatic decrease in cell division and the restoration of normal cellular morphology (DUNAIEF et al. 1994 ; STROBER 1996). These findings suggest that disruption of the interaction between BRG1 and Rb leads to increased rates of cell division and malignant transformation. Loss of brm function could similarly alter the activity of a Drosophila tumor suppressor protein, leading to extra divisions of SOP cells or their descendants. The recent identification of RBF, a Drosophila Rb family member (DU et al. 1996 ), should allow this possibility to be tested directly.

The peripheral nervous system defects associated with brm mutations are also similar to those associated with mutations in ash2, another trithorax group gene. ash2 mutant hemizygotes display duplications of bristles and sockets, transformation of campaniform sensilla to bristles, and the appearance of ectopic bristles on wing veins (ADAMSON and SHEARN 1996 ). Overexpression of the Polycomb group genes Posterior sex combs (Psc) and Suppressor of zeste 2 (Su(z)2) in third instar larvae also results in a range of bristle abnormalities very similar to those observed in brm-deficient clones, including twinned, composite, and stunted bristles and partially duplicated or missing sockets (BRUNK et al. 1991 ; SHARP et al. 1994 ). These findings suggest that ash2 and brm function antagonistically to Psc and Su(z)2 during the development of the peripheral nervous system.

Comparison of the sequences of BRM and its putative homologs in yeast and humans suggests that these proteins contain at least four functional domains: domain I, domain II, the ATPase domain, and the bromodomain. The ATPase domain is required for the catalytic activity of SWI2/SNF2 family members, while the regions flanking this domain are thought to contribute to their functional specificity by mediating interactions with other proteins. For example, domain I of the yeast SWI2/SNF2 protein interacts with the SNF11 subunit of the SWI/SNF complex (TREICH et al. 1995 ). The significance of the interaction between SNF11 and SWI2/SNF2 is unknown, however, since domain I is not essential for SWI2/SNF2 function in vivo, and yeast cells lacking SNF11 are viable and phenotypically normal (TREICH et al. 1995 ).

As previously shown for other SWI2/SNF2 family members (LAURENT et al. 1993 ; KHAVARI et al. 1993 ; PETERSON et al. 1994 ), we found that a mutation in the ATP-binding site of the BRM protein eliminates its function in vivo without disrupting its interactions with other proteins. By contrast, deletion of domain II causes a small but reproducible decrease in the size of the BRM complex, suggesting that it lacks one or more subunits. Although the effect of deleting domain II of the yeast SWI2/SNF2 protein has not been examined, the two-hybrid system has revealed an interaction between this domain and the SWI3 subunit of the SWI/SNF complex (TREICH et al. 1995 , 1997). Although a Drosophila homolog of SWI3 has not yet been identified, proteins related to SWI3 are present in the human BRG1 and hbrm complexes (WANG et al. 1996B ). These observations strongly suggest that domain II of the BRM protein interacts with an as yet unidentified Drosophila relative of SWI3.

What is the function of the bromodomain? More than a dozen bromodomain proteins have been identified, including a subset of chromatin remodeling factors (SWI2/SNF2, STH1, BRM, BRG1, and HBRM), type A histone acetyltransferases (GCN5, PCAF, TAF_II250, and p300/CBP; BROWNELL et al. 1996 ; YANG et al. 1996 ; MIZZEN et al. 1996 ; OGRYZKO et al. 1996 ), and other proteins involved in transcriptional activation, including SPT7 (GANSHEROFF et al. 1995 ) and FSH/RING3 (HAYNES et al. 1989 ; BECK et al. 1992 ). Most bromodomain proteins are components of multiprotein complexes, and they activate transcription by interacting with other proteins. For example, TAF_II250 interacts with the TATA-binding protein (HISATAKE et al. 1993 ; RUPPERT et al. 1993 ), and p300/CBP interacts with CREB, E1A, PCAF, c-jun, c-fos, c-Myb, MyoD, and TFIIB (OGRYZKO et al. 1996 ; YANG et al. 1996 ). Type A histone acetyltransferases, which contain bromodomains, acetylate nucleosomal histones and are associated with transcriptional activation (BROWNELL and ALLIS 1996 ). The bromodomain is thus an excellent candidate for a domain that mediates interactions with gene-specific regulatory proteins or general transcription factors. Such interactions might be important for targeting chromatin remodeling factors and histone acetyltransferases to specific regions of chromatin.

Although the bromodomain is required for full GCN5 function in vivo (CANDAU et al. 1997 ), it is dispensable for the function of two other yeast proteins, SWI2/SNF2 and SPT7 (LAURENT et al. 1993 ; GANSHEROFF et al. 1995 ). The bromodomain is not required for the enzymatic activity of histone acetyltransferases in vitro (MIZZEN et al. 1996 ; OGRYZKO et al. 1996 ). We reasoned that Drosophila would provide a more stringent system for investigating the function of the bromodomain. To our surprise, we found that the bromodomain is dispensible for both maternal and zygotic brm function. The bromodomain is, therefore, either not essential for the function of the BRM protein or it is functionally redundant. Given the conservation of the bromodomain from yeast to humans, we favor the latter of these possibilities. The BRM complex may contain more than one bromodomain protein, or other proteins may have functions that overlap those of the BRM complex. Either of these possibilities would account for the ability of brm^1446-1517 to rescue brm mutations.

It has recently become apparent that multiple complexes containing members of the SWI2/SNF2 family of ATPases are involved in chromatin remodeling in eukaryotic cells (TSUKIYAMA and WU 1997 ; PAZIN and KADONAGA 1997 ). For example, the Drosophila ISWI protein (ELFRING et al. 1994 ), which is closely related to BRM within the ATPase domain, is a subunit of at least three chromatin remodeling complexes: NURF, CHRAC, and ACF (TSUKIYAMA et al. 1995 ; VARGA-WEISZ et al. 1997 ; ITO et al. 1997 ). Each of these complexes was identified using biochemical assays for chromatin remodeling, but their functions in vivo are unknown. The ability of these complexes to catalyze ATP-dependent alterations in chromatin structure in vitro suggests that their in vivo activities may overlap those of the BRM complex. Numerous other SWI2/SNF2 family members are also present in Drosophila, including CHD1 (STOKES and PERRY 1995 ), 89B helicase (GOLDMAN-LEVI et al. 1994), RAD54 (KOOISTRA et al. 1997 ), and lodestar (GIRDHAM and GLOVER 1991 ). Additional work will be necessary to clarify the distinct and common roles of these proteins in transcription, development, and other processes involving chromatin.

	ACKNOWLEDGMENTS