The yeast SWI/SNF complex and its Drosophila and mammalian homologs are thought to control gene expression by altering chromatin structure, but the mechanism and specificity of this process are not fully understood. The Drosophila osa gene, like yeast SWI1, encodes an AT‐rich interaction (ARID) domain protein. We present genetic and biochemical evidence that Osa is a component of the Brahma complex, the Drosophila homolog of SWI/SNF. The ARID domain of Osa binds DNA without sequence specificity in vitro, but it is sufficient to direct transcriptional regulatory domains to specific target genes in vivo. Endogenous Osa appears to promote the activation of some of these genes. We show evidence that some Brahma‐containing complexes do not contain Osa and that Osa is not required to localize Brahma to chromatin. These data suggest that Osa modulates the function of the Brahma complex.
The role of specific transcription factors in activating or repressing gene transcription has been well established. However, more recently it has become clear that in order for these factors to gain access to their target sites in vivo, the organized chromatin structure surrounding the DNA must be altered. A number of large protein complexes have been implicated in remodeling chromatin to promote transcription factor access (reviewed by Kadonaga, 1998; Kingston and Narlikar, 1999).
In yeast, components of the SWI/SNF complex were originally identified genetically as factors required for the expression of the HO mating type gene or the SUC2 sucrose fermentation gene, but have since been shown to affect the expression of a large number of other genes (reviewed by Peterson and Tamkun, 1995). A purified yeast or human SWI/SNF‐related complex can facilitate transcription factor binding to nucleosomal DNA in an ATP‐dependent reaction (Cote et al., 1994; Kwon et al., 1994). The SWI2/SNF2 component of the complex contains the DNA‐stimulated ATPase activity (Laurent et al., 1992, 1993). The complex can bind to the minor groove of DNA, with an affinity not affected by sequence but influenced by DNA conformation (Quinn et al., 1996). Although the SWI/SNF complex is not essential for viability (Stern et al., 1984), another yeast chromatin remodeling complex, RSC, with an STH1 subunit related to SWI2/SNF2, is essential (Cairns et al., 1996).
In Drosophila, both genetic and biochemical approaches have led to the discovery of chromatin remodeling complexes. The Polycomb group of genes is required to maintain repression of homeotic genes such as Ultrabithorax; they are thought to do so by forming a repressive chromatin structure (reviewed by Pirrotta, 1997). Some members of the trithorax group of genes were subsequently identified by their ability to suppress dominant Polycomb phenotypes, suggesting that they might be involved in homeotic gene activation at the level of chromatin (Kennison and Tamkun, 1988). Indeed, two members of this group, brahma (brm) and moira (mor), have now been shown to encode proteins related to subunits of the SWI/SNF and RSC complexes, SWI2/SNF2/STH1 and SWI3/RSC8, respectively (Tamkun et al., 1992; Crosby et al., 1999). The products of two other members of the group, absent, small and homeotic discs 1 and 2 (ash1 and ash2; Shearn, 1989), do not form part of this complex but are present in two distinct large nuclear complexes, the function of which has not been established (Papoulas et al., 1998). The products of the genes trithorax (trx) and kismet (kis) contain regions of homology to other proteins implicated in chromatin remodeling; Trx has a SET [Su(var)3–9, E(Z), Trx] domain and Kis has the domains characteristic of chromodomain‐helicase (CHD) proteins (Mazo et al., 1990; Daubresse et al., 1999). Three other complexes, nucleosome remodeling factor (NURF), chromatin accessibility complex (CHRAC) and ATP‐dependent chromatin assembly and remodeling factor (ACF), have been isolated from Drosophila embryo extracts using a biochemical assay for nucleosome array disruption (Tsukiyama and Wu, 1995; Ito et al., 1997; Varga‐Weisz et al., 1997). All three contain imitation switch (ISWI), a protein with homology to the ATPase domain of SWI2/SNF2, but differ in their other subunits (Tsukiyama et al., 1995; Ito et al., 1997; Varga‐Weisz et al., 1997).
Two human homologs of the SWI/SNF and RSC complexes can be distinguished by their SWI2/SNF2‐related subunit; one complex contains human Brahma (hBRM) and the other contains Brahma‐related gene 1 (BRG1) (Wang et al., 1996a). Five other subunits of these complexes are homologous to components of both the yeast and Drosophila complexes (Wang et al., 1996b). It has recently been shown that four components of the complex, BRG1, INI1 (the SNF5 homolog), BAF155 and BAF170 (SWI3 homologs), can reconstitute in vitro nucleosome remodeling activity almost equal to that of the full complex, suggesting that other subunits may play a modulatory role (Phelan et al., 1999).
An unanswered question is whether and how these complexes are targeted to specific genes. Although SWI/SNF has no apparent DNA sequence preference (Quinn et al., 1996), mutations in SWI2 affect only a small subset of the genome, causing the transcription of some genes to be up‐regulated and others down‐regulated (Holstege et al., 1998). It is possible that transcription factors showing specific promoter interactions, such as GAL4 or Trx, can recruit the complex to their target genes (Cote et al., 1994; Rozenblatt‐Rosen et al., 1998). However, there is little understanding of what interactions might promote such recruitment.
The osa trithorax group gene of Drosophila, previously molecularly characterized under the name eyelid (Treisman et al., 1997; Vazquez et al., 1999), encodes a large protein with an AT‐rich interaction (ARID) domain like that of SWI1, and a second domain that is conserved in multicellular organisms but is not present in yeast (Treisman et al., 1997). Although Osa was not found in the previously reported Brm complex (Papoulas et al., 1998), we show here that both biochemical and genetic interactions indicate that it is a component of a subset of Brm complexes. The ARID domain of Osa binds DNA apparently without sequence specificity in vitro, but it is sufficient to target specific genes for regulation in vivo. Analysis of transgenic flies containing constitutively activating or repressing forms of the Osa protein implies that osa is required for the activation of some target genes. We suggest that Osa is required for some functions of the Brm chromatin remodeling complex, and that both its DNA‐binding capacity and other interactions contribute to promoter targeting.
osa genetically interacts with trithorax group genes
It was recently reported (Vazquez et al., 1999) that the gene we had named eyelid (eld; Treisman et al., 1997) is allelic to the previously described trithorax group gene osa (Kennison and Tamkun, 1988); we will henceforth refer to our alleles as osaeld308 and osaeld616. Vazquez et al. (1999) also showed that flies heterozygous for both osa and brm, which encodes a homolog of the yeast SWI2/SNF2 DNA‐dependent ATPase (Tamkun et al., 1992), have a held‐out wings phenotype. We have further characterized the genetic interactions of osa with brm and other members of the trithorax group.
Ectopic expression of a dominant‐negative form of Brm with a mutation in the ATP binding site (UAS‐brmK804R) disrupts many developmental processes (Elfring et al., 1998). We used an optomotor‐blind (omb)‐GAL4 driver to direct expression of UAS‐brmK804R in the central region of the wing disc; this resulted in loss of the distal wing margin, formation of ectopic campaniform sensillae and wing margin bristles, and disruptions in wing vein morphology (Elfring et al., 1998; Figure 1B). These phenotypes were strongly enhanced in animals heterozygous for osa (Figure 1C). Expression of UAS‐brmK804R at the wing margin using vestigial (vg)‐GAL4 resulted in the loss of the proximal, posterior wing margin, a phenotype that was again enhanced in osa heterozygotes (Figure 1D and E). We tested the effect of increasing osa dosage by co‐expressing a full‐length osa transcript under the control of the same vg‐GAL4 driver, and found that this completely rescued the dominant‐negative Brm phenotype (Figure 1F). Interestingly, ectopic expression of osa alone with vg‐Gal4 induced a dominant loss of proximal wing hinge structures (Figure 1G), and this phenotype was also rescued in animals co‐expressing osa and dominant‐negative brm (Figure 1F). This suggests that the functions of Osa and Brm are closely related, as a reduction in the activity of one can compensate for an excess of the other.
Ectopic expression of Osa in eye imaginal discs using eyeless (ey)‐GAL4 resulted in a variable reduction in eye size (Figure 1H). Rather than the expected suppression, we observed an enhancement of this phenotype in flies that either co‐expressed dominant‐negative Brm or were heterozygous for brm (Table I). The eye phenotype was also enhanced by mor and SNF5‐related 1 (Snr1; Figure 1 and Table I), both of which encode components of the Brm complex (Dingwall et al., 1995; Crosby et al., 1999). However, reducing the dosage of the trithorax group genes trx, ash1 or ash2 did not enhance the Osa overexpression phenotype (Table I; data not shown). As expected, a reduction in osa dosage suppressed the small eye phenotype (Table I). Clones of mor mutant cells in the eye disc exhibited a severe reduction in growth (Brizuela and Kennison, 1997), which was partially rescued if the cells were also mutant for osa (data not shown). Taken together, these data demonstrate that osa shows strong and specific genetic interactions with components of the Brm complex. However, in the wing osa appears to act in concert with brm, whereas in the eye osa opposes the functions of brm, snr1 and mor.
Osa physically interacts with the Brm complex
The Brm protein is thought to function in a large (∼2 MDa) multiprotein complex that regulates gene expression through the alteration of chromatin architecture (Papoulas et al., 1998). The genetic interactions between osa and brm suggested that the encoded proteins might function together to alter chromatin structure. To investigate this we tested whether Osa and Brm could physically interact in vivo. Nuclear extracts from Drosophila Schneider cells were immunoprecipitated with anti‐Osa antibody and blotted with antibodies against Brm or Snr1, or antibodies against ISWI or Ash2, components of different complexes that do not contain Brm (Papoulas et al., 1998). Brm and Snr1, but not ISWI or Ash2, were present in Osa immunoprecipitates (Figure 2). Similarly, Osa could be detected in nuclear extracts immunoprecipitated with either anti‐Brm or anti‐Snr1 antibody (Figure 2).
To determine whether Osa was associated with the high molecular weight Brm complex, we fractionated Schneider cell nuclear extracts through a glycerol gradient and immunoblotted with antibodies against the various proteins. Figure 3A shows that Osa, Brm and Snr1 co‐sediment in the bottom third of the gradient, suggesting that they are part of a large protein complex. Although Osa and Brm are present in similar fractions, Snr1 sediments in the bottom half of the gradient and could also be part of another complex that does not contain Osa or Brm. Alternatively, the anti‐Snr1 antibody might be much more sensitive, detecting very low levels of the Snr1 protein. When glycerol gradient fractions were immunoprecipitated with anti‐Osa antibody, Osa, Brm and Snr1 were co‐precipitated in the same region of the gradient in which they co‐sediment (Figure 3B). ISWI and Ash2 both showed broad sedimentation patterns, appearing in the bottom half of the gradient, but neither protein was immunoprecipitated from the gradient fractions with anti‐Osa antibody (data not shown). Thus, in vivo, Osa is found in a large complex with Brm and Snr1, but does not bind to proteins in other chromatin remodeling complexes.
To identify other proteins that co‐sediment and co‐immunoprecipitate with Osa, pooled gradient fractions 18–20 were immunoprecipitated with either anti‐Osa or anti‐Brm antibodies. Silver staining of the resulting purified protein complexes showed that all the bands in the Osa complex (Figure 3C, lane 3) were also present in the Brm complex (lane 4). We detected all the Brm complex components previously described (Papoulas et al., 1998) except BAP74. Two bands with molecular weights of 360 and 300 kDa were shown by mass spectrometry analysis to correspond to Osa. Although only a single osa RNA species has been detected on Northern blots (Vazquez et al., 1999), it is possible that two alternatively spliced products were not well resolved; alternatively, one of the bands could represent a proteolytically cleaved or post‐translationally modified form of the protein. Two bands present at a molecular weight of 155–160 kDa were likewise shown by mass spectrometry analysis to correspond to Mor (BAP155; Papoulas et al., 1998; Crosby et al., 1999).
Interestingly, several protein bands present in the Brm complex were absent from the Osa complex. Proteins of ∼190 and 380 kDa (P190* and P380*), as well as at least five additional very large proteins, depicted in Figure 3C, were detected in the complex immunoprecipitated with anti‐Brm antibody but not in the complex immunoprecipitated with anti‐Osa antibody. This suggests that Brm is a component of at least two distinct complexes, and that Osa is only present in a subset of these.
Osa is a non‐specific DNA‐binding protein
Osa contains a region that shares homology with the ARID family of DNA‐binding proteins (Treisman et al., 1997). The ARID domain mediates the binding of DNA at AT‐rich sequences, with contacts in the minor groove (Herrscher et al., 1995; Gregory et al., 1996; Yuan et al., 1998). Some ARID domain proteins, such as mouse Bright and Drosophila Dead ringer (Dri), recognize specific DNA sequences (Herrscher et al., 1995; Gregory et al., 1996), whereas others, such as human MRF, appear to interact with DNA without sequence specificity (Huang et al., 1996). The yeast SWI1 protein also contains an ARID domain and may mediate non‐sequence‐specific binding of the SWI/SNF complex to DNA (Quinn et al., 1996). We were interested in determining whether Osa could bind DNA and with what specificity.
Dri was isolated by its ability to bind to a multimer of the consensus binding sequence (NP) for the homeodomain of Engrailed (En) (Gregory et al., 1996). Because of the homology between Osa and Dri, we tested whether the predicted DNA‐binding domain of Osa could also bind to NP. A fusion protein containing a region of Osa including its ARID domain linked to glutathione S‐transferase (GST–osaDB; see Figure 6A), but not GST alone (data not shown), was able to retard the migration of a labeled oligonucleotide containing three copies of the NP sequence (NP3) in a gel mobility shift assay (Figure 4A). This DNA–protein complex could be supershifted by the addition of an antibody specific to Osa, but not by an unrelated antibody. This demonstrated that the ARID domain of Osa encodes a functional DNA‐binding domain, and that the anti‐Osa antibody, which was generated against a peptide that partially overlaps the DNA‐binding domain (Treisman et al., 1997; Figure 6A) is still able to recognize its epitope when Osa is bound to DNA. Osa also bound in vitro to the enhancer regions of two genes with altered expression in osa mutants, even‐skipped (eve) and Ultrabithorax (Ubx; data not shown).
We next attempted to determine the sequence specificity of DNA binding by Osa using a PCR‐based random oligonucleotide selection assay (Thiesen and Bach, 1990; Wilson et al., 1993). However, this produced no clear consensus sequence other than an apparent preference for AT‐rich sequences (data not shown). As a more direct test of specificity, we allowed the Osa DNA‐binding domain to bind to the 50 kb genome of phage lambda cut into 123 fragments. Osa bound to all these fragments with similar affinity at all salt and competitor DNA concentrations tested (Figure 4B; data not shown). Under the same conditions, the sequence‐specific DNA‐binding protein E47 (Sun and Baltimore, 1991) selected only a few fragments (Figure 4B, lane 6). Furthermore, in gel mobility shift assays a mutated form of NP3, in which the invariant central Ts were replaced by Cs, was able to compete with labeled NP3 for binding to GST–osaDB as efficiently as the wild‐type oligonucleotide (data not shown). Thus, the specific sequence of NP does not affect its ability to bind Osa.
We were not able to produce full‐length Osa protein in vitro, but to examine the DNA‐binding activity of the full‐length protein in vivo we stained polytene chromosomes with anti‐Osa antibody. Unlike many sequence‐specific DNA‐binding proteins that have been shown to recognize specific polytene bands (Kuzin et al., 1994; Serrano et al., 1995), Osa was localized along the entire length of all the chromosomes (Figure 5B). Osa antibody staining was present not only in the bands that stained strongly with DAPI, but also in the interband regions where DNA staining is less prominent (Figure 5A and D). As a control, we showed that antibodies directed against Male‐specific‐lethal‐1 (Msl‐1; Hilfiker et al., 1994) specifically stain the male X chromosome under the same conditions (data not shown). These data suggest that osa encodes a functional DNA‐binding protein that binds DNA without sequence specificity in vitro and is associated extensively with chromosomal DNA in vivo.
Osa is not required for Brm localization to chromatin
Brm and its homologs in other species are components of large multiprotein complexes that associate with and remodel chromatin; however, it is not clear which components of these complexes are responsible for chromatin binding. It has been suggested that SWI1 functions to target the yeast SWI/SNF complex to DNA (Quinn et al., 1996). Osa has a DNA‐binding domain related to that of SWI1 and, like the SWI/SNF complex, it binds DNA without sequence specificity. We therefore tested whether Osa might function to target Brm complexes to chromosomal DNA.
An antibody to Brm stains polytene chromosomes along their entire length in a pattern similar to that of DAPI staining for DNA (Figure 5C and D). We made clones of osa mutant cells in the salivary gland to determine whether Osa was required for this localization. Chromosomes from cells within these clones showed no anti‐Osa staining, demonstrating the specificity of the antibody, but retained a wild‐type pattern of staining with anti‐Brm (Figure 5F, G and H). This demonstrates that Osa is not required for the association of Brm complexes with chromosomal DNA.
An activation domain is sufficient for some Osa functions
The above results show that osa interacts genetically with brm, and that Osa is a component of a large multiprotein complex containing Brm. Brm‐related complexes are thought to promote transcription by altering the architecture of nucleosomal DNA, thus generating a conformation that is more favorable to binding by transcription factors and the basal transcriptional machinery. Some genes, such as even‐skipped, show reduced levels of expression in osa mutant embryos, supporting the role of Osa as an activator of gene expression. However, other genes, such as engrailed, show expanded domains of expression in osa mutants (Treisman et al., 1997). These genes could be directly activated or repressed by Osa, or their changes in expression level could be secondarily due to the regulation of other transcription factors by Osa. The lack of specificity of DNA binding by Osa in vitro prevented us from demonstrating direct action by altering Osa binding sites in the promoters of potential target genes. As an alternative approach, we sought to preserve Osa's target specificity in vivo and to determine the effect of making it an obligate activator or repressor of transcription. We therefore fused either the exogenous activator domain of VP16 or the repressor domain of Engrailed to the DNA‐binding domain of Osa (Figure 6A). The effects of misexpressing these activator (UAS‐osaAD) and repressor (UAS‐osaRD) forms of Osa under the control of the GAL4‐responsive UAS sequences were compared with those caused by misexpressing the full‐length wild‐type Osa protein. The Osa DNA‐binding domain appeared to be sufficient for chromosomal localization of these fusion proteins, as an antibody to VP16 detected the OsaAD protein along the length of polytene chromosomes (data not shown).
The notum of the adult fly contains a regular pattern of small (microchaetae) and large (macrochaetae) bristles (Figure 6B). Expression of the osa transgenes in the developing notum using a GAL4 insertion in the pannier (pnr) gene resulted in a dominant alteration of bristle formation. Ectopic expression of osa caused the loss of both micro‐ and macrochaetae, and defects in the midline of the notum, scutellum and abdomen (Figure 6C). Expression of UAS‐osaAD with the same GAL4 driver led to a very similar phenotype (Figure 6D), and co‐expression of UAS‐osa and UAS‐osaAD induced a stronger, apparently additive phenotype (Figure 6E). Expression of UAS‐osaRD with pnr‐GAL4 had the opposite effect, inducing the formation of ectopic macrochaetae on the notum (Figure 6F). Co‐expression of UAS‐osa with UAS‐osaRD rescued the bristle loss phenotype caused by the expression of UAS‐osa alone (Figure 6G). Thus, targeting an activation domain to Osa‐regulated genes has an effect similar to overexpression of the full‐length protein, while a repressor domain has the opposite effect.
In the wing, expression of UAS‐osaRD with omb‐GAL4 produced ectopic campaniform sensillae and wing margin bristles (Figure 6H). This phenotype was enhanced in flies heterozygous for osa (Figure 6I), suggesting that it results from interference with wild‐type osa function. It is also very similar to the effect of expression of dominant‐negative brm (Elfring et al., 1998). Expression of UAS‐osaAD caused the opposite phenotype, loss of campaniform sensillae (Figure 6J). Expression of full‐length osa with this driver resulted in dominant pupal lethality; although a small number of flies expressing osa did eclose, their wings were deformed, making a phenotypic comparison difficult.
The observation that UAS‐osaAD and UAS‐osaRD cause specific phenotypes in the developing wing disc, related to those caused by full‐length Osa, implies that the DNA‐binding domain of Osa has functional specificity in spite of its lack of DNA sequence specificity in vitro. Binding to other proteins could contribute to its ability to act on specific promoters. Expressing the DNA‐binding domain alone had no effect (data not shown), suggesting that its promoter interactions are not strong enough to compete significantly with endogenous Osa. The similar effects of UAS‐osa and UAS‐osaAD and opposite effects of UAS‐osaRD also indicate that, in the wing imaginal disc, Osa functions as an activator of gene expression.
Osa associates with the Brm complex
We have found that the Osa protein co‐immunoprecipitates and co‐sediments with the Brm and Snr1 proteins, indicating that Osa associates with the Brm complex. Although Osa was not previously found as a component of the purified complex, some large unidentified proteins were co‐purified and one of these could correspond to Osa (Papoulas et al., 1998). It has been reported that a human protein, p270, which has not been completely sequenced but also has an ARID domain, is associated with the BRG1 complex (Dallas et al., 1998). However, it is unlikely that Osa is an essential subunit of every Brm complex. A number of proteins are precipitated with anti‐Brm but not anti‐Osa. While it is possible that the binding of anti‐Osa antibody to its epitope prevents the association of these proteins with Osa, or that these proteins bind the anti‐Brm antibody rather than Brm itself, the simplest explanation is that Brm is a component of at least two distinct complexes and that these proteins are present in Brm complexes from which Osa is absent. This would be consistent with the different phenotypic effects caused by osa and brm mutations. Mutations in brm or mor cause greatly reduced cell growth and viability, as well as oogenesis defects that prevent the production of maternally mutant embryos (Brizuela et al., 1994; Brizuela and Kennison, 1997; Elfring et al., 1998). In contrast, cells mutant for a null allele of osa are able to proliferate, and maternal osa is required for normal embryogenesis but not for oogenesis (Treisman et al., 1997; Vazquez et al., 1999). Genetic interactions between osa and components of the Brm complex show some tissue specificity; in the wing, loss of osa and loss of brm enhance each other, but in the eye, loss of brm enhances the effect of osa overexpression.
Osa is thus unlikely to be a strict functional homolog of SWI1, which is essential for the stability and function of the yeast SWI/SNF complex (Peterson and Herskowitz, 1992). Alternatively, other subunits of the Drosophila complex not present in yeast may be partially redundant with Osa; one possibility is the HMG box protein BAP111 (Papoulas et al., 1998), which has a vertebrate counterpart, BAF57 (Wang et al., 1998). Brm is still stable and localized to chromatin in the absence of Osa; it has recently been reported that an AT‐hook DNA‐binding domain is present in human and yeast homologs of Brm and contributes to their association with chromatin (Bourachot et al., 1999). Other possible explanations are that this localization could be mediated by the above DNA‐binding proteins, or could be due to interactions of other subunits of the complex with protein components of chromatin; a third alternative is that the remaining Brm is present in complexes that do not contain Osa. As only SWI2/SNF2, SWI3 and SNF5‐related subunits are essential for chromatin remodeling in vitro (Phelan et al., 1999), other DNA‐binding proteins may play a specificity‐determining or regulatory role.
The observation that overexpression of osa has effects opposite to those caused by its loss of function suggests that despite its widespread expression (Treisman et al., 1997) Osa is limiting in vivo; this is probably also the case for SWI1 (C.Peterson, personal communication). Interestingly, an excess of Osa can compensate for the presence of dominant‐negative Brm in the wing, suggesting that other components are not limiting, so that increasing Osa is sufficient to restore the normal number of functional complexes. Overexpressing osa in a wild‐type background may increase the number of Brm complexes carrying out Osa‐regulated functions. Reducing brm function in the wing ameliorates the effects of excess osa; since ectopic osa requires brm for its activity, it is likely to be acting through the Brm complex rather than by an alternative mechanism. However, in the eye disc Osa antagonizes Brm function, perhaps redirecting Brm into the Osa‐containing type of complex and thus reducing the number of complexes that do not contain Osa. The number of these complexes would be further reduced by lowering the level of Brm.
The Osa DNA‐binding domain is not sequence‐specific but has some target specificity in vivo
We have shown that Osa contains a functional DNA‐binding domain that has no apparent sequence specificity beyond a preference for AT‐rich regions. This DNA‐binding domain thus resembles that of the related MRF (Huang et al., 1996), as well as reproducing some properties of the intact SWI/SNF complex that could reflect binding by SWI1; the major band that can be cross‐linked to DNA has a molecular weight similar to that of SWI1 (Quinn et al., 1996). The similar domains present in Bright and Dri do have a consensus recognition sequence (Herrscher et al., 1995; Gregory et al., 1996); this specificity may be provided by an extended region of homology between the two proteins that is not conserved in Osa. Although we have not been able to produce full‐length Osa protein in vitro, it is unlikely that another region of the protein confers sequence‐specific recognition, as we observe binding of endogenous Osa protein along the entire length of the polytene chromosomes rather than to specific bands. However, it is possible that association of Osa with protein components of chromatin contributes to this staining pattern.
Several observations suggest that Osa has more specific effects on gene expression than its DNA‐binding capability would predict. First, many genes are expressed normally in osa maternally and zygotically mutant embryos (Treisman et al., 1997), although we would expect Osa to be able to bind to every promoter region. Secondly, osa dosage affects the phenotype caused by Antennapedia (Antp) misexpression in the antennal disc only when this misexpression is driven by the Antp P2 promoter and not when it is driven by the hsp70 or sas promoters, showing that Osa has promoter‐specific effects (Vazquez et al., 1999). Finally, the Osa ARID domain, which is sufficient to bind any DNA in vitro, appears to target an activation or repression domain to only a specific subset of genes. This suggests that Osa's access to DNA in vivo may depend on interaction with other factors. The in vivo specificity observed for the isolated DNA‐binding domain implies that this domain is sufficient for some such interactions. Interestingly, the ARID domain is the only region of homology between Osa and SWI1 and may therefore contain the determinants necessary for incorporation into the SWI/SNF or BRM complex. Alternatively, its interactions with other DNA‐bound proteins could allow it to associate with specific promoters. Further experiments will be necessary to address this issue.
Osa can promote transcriptional activation
Lack of osa in the embryo leads to the loss of expression of certain genes but to the misexpression of others (Treisman et al., 1997). To investigate whether Osa can directly activate or repress transcription, we compared the effects of gain or loss of osa function with those caused by expressing the Osa DNA‐binding domain fused to exogenous transcriptional activator or repressor domains. Although these Osa fusion proteins may not use the same mechanism as wild‐type Osa to affect gene expression, the similarity of observed phenotypes allows us to infer the direction of the changes in gene expression normally caused by Osa. The results show that expression of the activator form in the wing or notum resembles overexpression of the full‐length protein, while expression of the repressor form resembles and is enhanced by loss of osa function. This suggests that during wing disc development Osa functions to activate gene expression. However, the effects of expressing these transgenes in other tissues suggest that Osa may also act to repress gene expression in some contexts (R.T.Collins, unpublished data).
The yeast SWI/SNF complex may likewise function both to activate and to repress transcription; genome‐wide analysis shows that more genes show elevated rather than reduced expression in a swi2 mutant strain (Holstege et al., 1998), although it is not known which are direct targets. The related yeast complex RSC appears to repress at least one identified target (Moreira and Holmberg, 1999). Human hBRM has also been shown to assist Rb to repress E2F1‐regulated transcription (Trouche et al., 1997). It is also possible that Osa represses transcription by antagonizing the activity of the Brm complex rather than by facilitating its normal function. It is not clear how the multiple chromatin remodeling complexes present in the cell may interact. Although we did not detect any physical association of Osa with Ash2, which is in a complex smaller than the Brm complex (Papoulas et al., 1998), the absence of osa can rescue the oogenesis defect of ash2 mutants (our unpublished data). Thus, the two complexes may have common target genes in addition to the homeotic genes.
We have found that many of the phenotypes of osa mutations resemble those caused by ectopic wingless (wg) expression, although wg itself is not ectopically expressed in osa mutant cells (Treisman et al., 1997). We now have additional evidence that wg target genes are activated in the absence of osa and repressed in the presence of excess osa (R.T.Collins and J.E.Treisman, manuscript in preparation). Because Wg signaling is transmitted by an HMG box transcription factor, dTCF (Brunner et al., 1997; Riese et al., 1997; van de Wetering et al., 1997), vertebrate homologs of which have been shown to act by bending DNA to bring other transcription factors into proximity (Giese et al., 1992), it is likely to be sensitive to chromatin structure (Sheridan et al., 1995). We suggest that Osa maintains a chromatin structure incompatible with transcription on wg‐regulated genes, and that in the absence of Osa these promoters can be stimulated even without the activation domain contributed by Armadillo (van de Wetering et al., 1997). Interactions such as these between developmental signals and chromatin architecture are of great interest in understanding the different responses of different cells to such signals.
Materials and methods
A fragment of Osa encoding amino acids 965–1198 was amplified using PCR primers that introduced BamHI and EcoRI sites, and was subcloned into pGEX‐4T‐1 (Pharmacia) to generate GST–osaDB. Recombinant protein was expressed in Escherichia coli and purified on glutathione–agarose beads. Purified protein was either retained on the beads or eluted with 5 mM glutathione in 50 mM Tris pH 8.0. For the gel mobility shift assay, 200 ng of 32P end‐labeled oligonucleotide consisting of three copies of the engrailed homeodomain consensus binding sequence (Desplan et al., 1988), NP3 [(GATCTCAATTAAT)3], incubated with 250 ng of purified GST–osaDB and increasing amounts of either anti‐Osa (Treisman et al., 1997) or anti‐En (DiNardo et al., 1985) antibody in gel shift buffer [20 mM HEPES pH 7.6, 1 mM EDTA, 10 mM (NH2)SO4, 1 mM dithiothreitol (DTT), 0.2% Tween‐20 and 200 mM KCl] supplemented with 50 μg/ml sonicated salmon sperm DNA and 25 μg/ml bovine serum albumin (BSA) for 30 min at room temperature. Protein–DNA complexes were separated on an 8% non‐denaturing polyacrylamide gel and exposed for autoradiography. For the GST pulldown assay, lambda DNA was digested with HindIII and Sau3AI and the fragments were end labeled with [32P]dATP. Labeled DNA (0.8 μg) was incubated with 50 ng of GST–DB or GST–E47 (provided by Xiao‐Hong Sun) bound to glutathione–agarose beads for 1 h at 4°C in binding buffer [gel shift buffer supplemented with 25 μg/ml each BSA, poly(dI–dC) and poly(dA–dT)]. The beads were washed three times with binding buffer at 4°C, and bound DNA was eluted by boiling in formamide loading buffer (90% formamide, 1 × TBE, 0.04% bromophenol blue, 0.04% xylene cyanol) and separated on a 6% sequencing gel followed by autoradiography. The random oligonucleotide selection assay (SELEX) was performed as described (Jun and Desplan, 1996).
Immunoprecipitation and glycerol gradient sedimentation
Drosophila Schneider cell nuclear extracts were prepared as described (Pugh, 1995). Immunoprecipitations and washes were carried out in HEM buffer (25 mM HEPES–KOH pH 7.9, 0.1 mM EDTA, 12.5 mM MgCl2) containing 0.3 M KCl and 20% glycerol. Glycerol gradient sedimentation was carried out according to Tanese (1997). Nuclear extract in buffer C (20 mM HEPES–KOH pH 7.9, 0.42 M NaCl, 0.2 mM EDTA, 1.5 mM MgCl2, 0.5 mM DTT, 25% glycerol) was diluted to <10% glycerol. Approximately 1.25 mg (in 250 μl) of nuclear extract was applied to a 5 ml, 10–30% gradient of glycerol in HEM buffer plus 0.1 M KCl. Samples were centrifuged in a Beckman SW50.1 rotor at 43 000 r.p.m. for 13 h at 4°C. Twenty 250 μl fractions were collected from top (F1) to bottom (F20) of each gradient. Proteins in odd‐numbered fractions (75 μl) were precipitated with trichloroacetic acid and separated by SDS–PAGE. Immunoblotting with α‐Osa, α‐Brm, α‐Snr1, α‐ISWI and α‐Ash2 was performed on the protein samples prepared from the same gradient run. Molecular mass standards were run in a parallel gradient. For immunoprecipitations, 250 μl of even‐numbered gradient fractions were used to immunoprecipitate proteins with α‐Osa antibody followed by sequential immunoblotting with α‐Osa, α‐Brm and α‐Snr1 antibodies. For silver staining of Osa and Brm complexes, glycerol gradient fractions F18–F20 from two parallel gradients were pooled (1200 μl), divided into three and each pooled fraction was incubated with either α‐Osa antibody‐conjugated protein A–Sepharose, α‐Brm antibody and protein A–Sepharose, or protein A–Sepharose alone.
For immunostaining of polytene chromosomes, salivary glands from climbing third instar larvae were dissected into PBT [phosphate‐buffered saline (PBS), 0.2% Triton X‐100] and fixed for 30 s in a drop of 3.7% formaldehyde in PBS, 1% Triton X‐100. Glands were then transferred to a drop of 3.7% formaldehyde, 50% acetic acid for 5 min and squashed under glass cover slips. Slides were frozen in liquid N2 and cover slips flipped off. Chromosomes were blocked for 1 h at 4°C in PBT, 10% normal donkey serum (PBSTS) and incubated overnight at 4°C with anti‐Osa antibody (diluted 1:1 in PBT), anti‐Brm antibody (Elfring et al., 1998; diluted 1:5 in PBSTS), anti‐VP16 antibody (Santa Cruz Biotechnology; diluted 1:50 in PBSTS) or anti‐Msl‐1 antibody (provided by Bruce Baker; diluted 1:100 in PBSTS). Slides were washed three times for 15 min at room temperature with PBT and incubated with FITC‐conjugated donkey anti‐rabbit (for Brm), or Texas Red‐conjugated donkey anti‐mouse (for Osa and VP16), or anti‐rat (for Msl‐1) secondary antibody (diluted 1:200 in PBSTS) for 1 h at room temperature. After three 15 min washes at room temperature in PBT slides were rinsed in 0.1 M phosphate buffer, 0.2% Triton X‐100. Chromosomes were counter‐stained with DAPI and observed with a Leica TCS NT confocal microscope. Adult wings were mounted in Canada balsam:methyl salicylate (2:1).
Generation of transgenic strains
To generate UAS‐osa, the sequence encoding the full‐length Osa open reading frame with 453 bp of 5′UTR and 754 bp of 3′UTR was subcloned in three steps using EcoRI and XbaI sites into the P‐element transformation vector pUAST. The other constructs, also in pUAST, contained 453 bp of 5′UTR and the first seven amino acids of Osa (MNEKIKS) followed by amino acids 965–1198. For UAS‐osaAD, these Osa sequences were followed by amino acids 481–559 of VP16 (Sadowski et al., 1988). For UAS‐osaRD, the same sequences were followed by amino acids 168–282 of Drosophila Engrailed (Tolkunova et al., 1998). All fragments were generated by PCR (details available on request). Transgenic strains carrying the transgenes UAS‐osa, UAS‐osaDBD, UAS‐osaRD and UAS‐osaAD were generated by P‐element‐mediated transformation as described (Rubin and Spradling, 1982). Several independent insertions of each construct were examined.
Alleles and other transgenic lines used were osaeld308 (Treisman et al., 1997), brm2 (Tamkun et al., 1992), mor1, trxE2 (Kennison and Tamkun, 1988), snr1P1 (Dingwall et al., 1995), ash122 (Tripoulas et al., 1994), ash21 (Adamson and Shearn, 1996), UAS‐brmK804R (Elfring et al., 1998), omb‐GAL4 (Lecuit et al., 1996), vg‐GAL4 (Simmonds et al., 1995), ey‐GAL4 (Hazelett et al., 1998), pnr‐GAL4 (Calleja et al., 1996) and UAS‐FLP (Duffy et al., 1998). To make osa mutant clones in salivary glands, males of genotype FRT82,osaeld308/TM6B were crossed to females of genotype w,hsFLP122; FRT82, Sb, hs‐[pi]myc and the larvae were heat shocked at 38.5°C for 1 h during first instar to induce expression of hsFLP. Recombinant chromosomes carrying FRT82, osaeld308 and either mor1 or ash21 were used for analysis of double‐mutant phenotypes.
We thank Spyros Artavanis‐Tsakonas, Bruce Baker, Marc Bourouis, Claude Desplan, Steve DiNardo, Andrew Dingwall, Ulrike Heberlein, Armen Manoukian, Alex Mazo, Allen Shearn, Xiao‐Hong Sun, John Tamkun and the Bloomington Drosophila stock center for fly stocks and reagents. Drosophila Schneider cells used to prepare the nuclear extracts were grown and harvested at the National Cell Culture Center (Minneapolis, MN) whose service is hereby acknowledged. We are grateful to Mary Ann Gawinowicz at the Protein Chemistry Core Facility, Columbia University, for performing mass spectrometry analysis. We thank Angus Wilson for providing Drosophila Schneider cell nuclear extracts. We thank Ruth Lehmann and Michelle Starz‐Gaiano for help with confocal microscopy. We are grateful to Ian Oliver for expert technical assistance, and we thank the members of the Treisman and Lehmann laboratories for stimulating discussions. The manuscript was improved by the thoughtful comments of Claude Desplan, Jeff Lee, Ariel Ruiz i Altaba, Chris Yohn and Corinne Zaffran. This work was supported by NIH grants GM56131 to J.E.T. and GM51314 to N.T.
- Copyright © 1999 European Molecular Biology Organization